{ "best_metric": 1.661559820175171, "best_model_checkpoint": "storage/pretrain-pythia-eva-1b/checkpoint-36513", "epoch": 0.9, "eval_steps": 4057, "global_step": 36513, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0, "learning_rate": 1.6420361247947455e-08, "loss": 6.5851, "step": 1 }, { "epoch": 0.0, "learning_rate": 3.284072249589491e-08, "loss": 6.4887, "step": 2 }, { "epoch": 0.0, "learning_rate": 4.926108374384237e-08, "loss": 6.3495, "step": 3 }, { "epoch": 0.0, "learning_rate": 6.568144499178982e-08, "loss": 6.5821, "step": 4 }, { "epoch": 0.0, "learning_rate": 8.210180623973728e-08, "loss": 6.4662, "step": 5 }, { "epoch": 0.0, "learning_rate": 9.852216748768474e-08, "loss": 6.5383, "step": 6 }, { "epoch": 0.0, "learning_rate": 1.1494252873563219e-07, "loss": 6.5175, "step": 7 }, { "epoch": 0.0, "learning_rate": 1.3136288998357964e-07, "loss": 6.4691, "step": 8 }, { "epoch": 0.0, "learning_rate": 1.477832512315271e-07, "loss": 6.5622, "step": 9 }, { "epoch": 0.0, "learning_rate": 1.6420361247947457e-07, "loss": 6.4294, "step": 10 }, { "epoch": 0.0, "learning_rate": 1.80623973727422e-07, "loss": 6.3797, "step": 11 }, { "epoch": 0.0, "learning_rate": 1.9704433497536947e-07, "loss": 6.6216, "step": 12 }, { "epoch": 0.0, "learning_rate": 2.1346469622331695e-07, "loss": 6.4242, "step": 13 }, { "epoch": 0.0, "learning_rate": 2.2988505747126437e-07, "loss": 6.6012, "step": 14 }, { "epoch": 0.0, "learning_rate": 2.4630541871921185e-07, "loss": 6.3381, "step": 15 }, { "epoch": 0.0, "learning_rate": 2.627257799671593e-07, "loss": 6.5961, "step": 16 }, { "epoch": 0.0, "learning_rate": 2.7914614121510676e-07, "loss": 6.2781, "step": 17 }, { "epoch": 0.0, "learning_rate": 2.955665024630542e-07, "loss": 6.2479, "step": 18 }, { "epoch": 0.0, "learning_rate": 3.1198686371100166e-07, "loss": 6.2217, "step": 19 }, { "epoch": 0.0, "learning_rate": 3.2840722495894914e-07, "loss": 6.1143, "step": 20 }, { "epoch": 0.0, "learning_rate": 3.4482758620689656e-07, "loss": 6.1351, "step": 21 }, { "epoch": 0.0, "learning_rate": 3.61247947454844e-07, "loss": 6.1898, "step": 22 }, { "epoch": 0.0, "learning_rate": 3.776683087027915e-07, "loss": 6.1329, "step": 23 }, { "epoch": 0.0, "learning_rate": 3.9408866995073894e-07, "loss": 6.1016, "step": 24 }, { "epoch": 0.0, "learning_rate": 4.1050903119868637e-07, "loss": 6.1137, "step": 25 }, { "epoch": 0.0, "learning_rate": 4.269293924466339e-07, "loss": 6.1154, "step": 26 }, { "epoch": 0.0, "learning_rate": 4.433497536945813e-07, "loss": 5.9414, "step": 27 }, { "epoch": 0.0, "learning_rate": 4.5977011494252875e-07, "loss": 6.2825, "step": 28 }, { "epoch": 0.0, "learning_rate": 4.7619047619047623e-07, "loss": 6.1232, "step": 29 }, { "epoch": 0.0, "learning_rate": 4.926108374384237e-07, "loss": 6.0958, "step": 30 }, { "epoch": 0.0, "learning_rate": 5.090311986863711e-07, "loss": 6.1994, "step": 31 }, { "epoch": 0.0, "learning_rate": 5.254515599343186e-07, "loss": 5.8622, "step": 32 }, { "epoch": 0.0, "learning_rate": 5.418719211822661e-07, "loss": 5.6854, "step": 33 }, { "epoch": 0.0, "learning_rate": 5.582922824302135e-07, "loss": 5.5414, "step": 34 }, { "epoch": 0.0, "learning_rate": 5.747126436781609e-07, "loss": 5.5135, "step": 35 }, { "epoch": 0.0, "learning_rate": 5.911330049261084e-07, "loss": 5.5106, "step": 36 }, { "epoch": 0.0, "learning_rate": 6.075533661740559e-07, "loss": 5.4272, "step": 37 }, { "epoch": 0.0, "learning_rate": 6.239737274220033e-07, "loss": 5.2755, "step": 38 }, { "epoch": 0.0, "learning_rate": 6.403940886699508e-07, "loss": 5.2681, "step": 39 }, { "epoch": 0.0, "learning_rate": 6.568144499178983e-07, "loss": 5.3265, "step": 40 }, { "epoch": 0.0, "learning_rate": 6.732348111658457e-07, "loss": 5.3502, "step": 41 }, { "epoch": 0.0, "learning_rate": 6.896551724137931e-07, "loss": 5.31, "step": 42 }, { "epoch": 0.0, "learning_rate": 7.060755336617405e-07, "loss": 5.3397, "step": 43 }, { "epoch": 0.0, "learning_rate": 7.22495894909688e-07, "loss": 5.0401, "step": 44 }, { "epoch": 0.0, "learning_rate": 7.389162561576356e-07, "loss": 5.0474, "step": 45 }, { "epoch": 0.0, "learning_rate": 7.55336617405583e-07, "loss": 4.7106, "step": 46 }, { "epoch": 0.0, "learning_rate": 7.717569786535305e-07, "loss": 4.4159, "step": 47 }, { "epoch": 0.0, "learning_rate": 7.881773399014779e-07, "loss": 4.7304, "step": 48 }, { "epoch": 0.0, "learning_rate": 8.045977011494253e-07, "loss": 4.5225, "step": 49 }, { "epoch": 0.0, "learning_rate": 8.210180623973727e-07, "loss": 4.5025, "step": 50 }, { "epoch": 0.0, "learning_rate": 8.374384236453203e-07, "loss": 4.4896, "step": 51 }, { "epoch": 0.0, "learning_rate": 8.538587848932678e-07, "loss": 4.4441, "step": 52 }, { "epoch": 0.0, "learning_rate": 8.702791461412152e-07, "loss": 4.4616, "step": 53 }, { "epoch": 0.0, "learning_rate": 8.866995073891626e-07, "loss": 4.4557, "step": 54 }, { "epoch": 0.0, "learning_rate": 9.031198686371101e-07, "loss": 4.3686, "step": 55 }, { "epoch": 0.0, "learning_rate": 9.195402298850575e-07, "loss": 4.4061, "step": 56 }, { "epoch": 0.0, "learning_rate": 9.359605911330049e-07, "loss": 4.2169, "step": 57 }, { "epoch": 0.0, "learning_rate": 9.523809523809525e-07, "loss": 4.3458, "step": 58 }, { "epoch": 0.0, "learning_rate": 9.688013136288999e-07, "loss": 4.241, "step": 59 }, { "epoch": 0.0, "learning_rate": 9.852216748768474e-07, "loss": 4.1275, "step": 60 }, { "epoch": 0.0, "learning_rate": 1.001642036124795e-06, "loss": 4.1716, "step": 61 }, { "epoch": 0.0, "learning_rate": 1.0180623973727423e-06, "loss": 4.304, "step": 62 }, { "epoch": 0.0, "learning_rate": 1.0344827586206898e-06, "loss": 4.1138, "step": 63 }, { "epoch": 0.0, "learning_rate": 1.0509031198686371e-06, "loss": 4.1431, "step": 64 }, { "epoch": 0.0, "learning_rate": 1.0673234811165846e-06, "loss": 4.0476, "step": 65 }, { "epoch": 0.0, "learning_rate": 1.0837438423645322e-06, "loss": 4.0917, "step": 66 }, { "epoch": 0.0, "learning_rate": 1.1001642036124797e-06, "loss": 3.971, "step": 67 }, { "epoch": 0.0, "learning_rate": 1.116584564860427e-06, "loss": 4.0108, "step": 68 }, { "epoch": 0.0, "learning_rate": 1.1330049261083746e-06, "loss": 4.0155, "step": 69 }, { "epoch": 0.0, "learning_rate": 1.1494252873563219e-06, "loss": 3.7625, "step": 70 }, { "epoch": 0.0, "learning_rate": 1.1658456486042694e-06, "loss": 3.8784, "step": 71 }, { "epoch": 0.0, "learning_rate": 1.1822660098522167e-06, "loss": 3.9348, "step": 72 }, { "epoch": 0.0, "learning_rate": 1.1986863711001645e-06, "loss": 3.9208, "step": 73 }, { "epoch": 0.0, "learning_rate": 1.2151067323481118e-06, "loss": 3.9604, "step": 74 }, { "epoch": 0.0, "learning_rate": 1.2315270935960593e-06, "loss": 3.7948, "step": 75 }, { "epoch": 0.0, "learning_rate": 1.2479474548440066e-06, "loss": 3.9058, "step": 76 }, { "epoch": 0.0, "learning_rate": 1.2643678160919542e-06, "loss": 3.8846, "step": 77 }, { "epoch": 0.0, "learning_rate": 1.2807881773399017e-06, "loss": 3.9386, "step": 78 }, { "epoch": 0.0, "learning_rate": 1.297208538587849e-06, "loss": 3.77, "step": 79 }, { "epoch": 0.0, "learning_rate": 1.3136288998357965e-06, "loss": 3.8729, "step": 80 }, { "epoch": 0.0, "learning_rate": 1.330049261083744e-06, "loss": 3.926, "step": 81 }, { "epoch": 0.0, "learning_rate": 1.3464696223316914e-06, "loss": 3.8623, "step": 82 }, { "epoch": 0.0, "learning_rate": 1.362889983579639e-06, "loss": 3.7682, "step": 83 }, { "epoch": 0.0, "learning_rate": 1.3793103448275862e-06, "loss": 3.7256, "step": 84 }, { "epoch": 0.0, "learning_rate": 1.3957307060755338e-06, "loss": 3.8846, "step": 85 }, { "epoch": 0.0, "learning_rate": 1.412151067323481e-06, "loss": 3.6289, "step": 86 }, { "epoch": 0.0, "learning_rate": 1.4285714285714286e-06, "loss": 3.7105, "step": 87 }, { "epoch": 0.0, "learning_rate": 1.444991789819376e-06, "loss": 3.816, "step": 88 }, { "epoch": 0.0, "learning_rate": 1.4614121510673235e-06, "loss": 3.7356, "step": 89 }, { "epoch": 0.0, "learning_rate": 1.4778325123152712e-06, "loss": 3.8339, "step": 90 }, { "epoch": 0.0, "learning_rate": 1.4942528735632185e-06, "loss": 3.8153, "step": 91 }, { "epoch": 0.0, "learning_rate": 1.510673234811166e-06, "loss": 3.9224, "step": 92 }, { "epoch": 0.0, "learning_rate": 1.5270935960591136e-06, "loss": 3.7095, "step": 93 }, { "epoch": 0.0, "learning_rate": 1.543513957307061e-06, "loss": 3.6971, "step": 94 }, { "epoch": 0.0, "learning_rate": 1.5599343185550085e-06, "loss": 3.7456, "step": 95 }, { "epoch": 0.0, "learning_rate": 1.5763546798029558e-06, "loss": 3.7414, "step": 96 }, { "epoch": 0.0, "learning_rate": 1.5927750410509033e-06, "loss": 3.7478, "step": 97 }, { "epoch": 0.0, "learning_rate": 1.6091954022988506e-06, "loss": 3.8594, "step": 98 }, { "epoch": 0.0, "learning_rate": 1.6256157635467982e-06, "loss": 3.5268, "step": 99 }, { "epoch": 0.0, "learning_rate": 1.6420361247947455e-06, "loss": 3.7405, "step": 100 }, { "epoch": 0.0, "learning_rate": 1.658456486042693e-06, "loss": 3.8386, "step": 101 }, { "epoch": 0.0, "learning_rate": 1.6748768472906405e-06, "loss": 3.6257, "step": 102 }, { "epoch": 0.0, "learning_rate": 1.6912972085385879e-06, "loss": 3.6435, "step": 103 }, { "epoch": 0.0, "learning_rate": 1.7077175697865356e-06, "loss": 3.637, "step": 104 }, { "epoch": 0.0, "learning_rate": 1.724137931034483e-06, "loss": 3.6229, "step": 105 }, { "epoch": 0.0, "learning_rate": 1.7405582922824304e-06, "loss": 3.4795, "step": 106 }, { "epoch": 0.0, "learning_rate": 1.756978653530378e-06, "loss": 3.6883, "step": 107 }, { "epoch": 0.0, "learning_rate": 1.7733990147783253e-06, "loss": 3.5677, "step": 108 }, { "epoch": 0.0, "learning_rate": 1.7898193760262728e-06, "loss": 3.6282, "step": 109 }, { "epoch": 0.0, "learning_rate": 1.8062397372742201e-06, "loss": 3.7552, "step": 110 }, { "epoch": 0.0, "learning_rate": 1.8226600985221677e-06, "loss": 3.8331, "step": 111 }, { "epoch": 0.0, "learning_rate": 1.839080459770115e-06, "loss": 3.6486, "step": 112 }, { "epoch": 0.0, "learning_rate": 1.8555008210180625e-06, "loss": 3.6244, "step": 113 }, { "epoch": 0.0, "learning_rate": 1.8719211822660098e-06, "loss": 3.5468, "step": 114 }, { "epoch": 0.0, "learning_rate": 1.8883415435139574e-06, "loss": 3.7307, "step": 115 }, { "epoch": 0.0, "learning_rate": 1.904761904761905e-06, "loss": 3.5827, "step": 116 }, { "epoch": 0.0, "learning_rate": 1.9211822660098524e-06, "loss": 3.6416, "step": 117 }, { "epoch": 0.0, "learning_rate": 1.9376026272577998e-06, "loss": 3.7442, "step": 118 }, { "epoch": 0.0, "learning_rate": 1.9540229885057475e-06, "loss": 3.6705, "step": 119 }, { "epoch": 0.0, "learning_rate": 1.970443349753695e-06, "loss": 3.6212, "step": 120 }, { "epoch": 0.0, "learning_rate": 1.986863711001642e-06, "loss": 3.8365, "step": 121 }, { "epoch": 0.0, "learning_rate": 2.00328407224959e-06, "loss": 3.7478, "step": 122 }, { "epoch": 0.0, "learning_rate": 2.019704433497537e-06, "loss": 3.7035, "step": 123 }, { "epoch": 0.0, "learning_rate": 2.0361247947454845e-06, "loss": 3.6755, "step": 124 }, { "epoch": 0.0, "learning_rate": 2.052545155993432e-06, "loss": 3.5676, "step": 125 }, { "epoch": 0.0, "learning_rate": 2.0689655172413796e-06, "loss": 3.6006, "step": 126 }, { "epoch": 0.0, "learning_rate": 2.085385878489327e-06, "loss": 3.6042, "step": 127 }, { "epoch": 0.0, "learning_rate": 2.1018062397372742e-06, "loss": 3.6705, "step": 128 }, { "epoch": 0.0, "learning_rate": 2.118226600985222e-06, "loss": 3.557, "step": 129 }, { "epoch": 0.0, "learning_rate": 2.1346469622331693e-06, "loss": 3.5229, "step": 130 }, { "epoch": 0.0, "learning_rate": 2.1510673234811166e-06, "loss": 3.726, "step": 131 }, { "epoch": 0.0, "learning_rate": 2.1674876847290643e-06, "loss": 3.6432, "step": 132 }, { "epoch": 0.0, "learning_rate": 2.1839080459770117e-06, "loss": 3.464, "step": 133 }, { "epoch": 0.0, "learning_rate": 2.2003284072249594e-06, "loss": 3.5185, "step": 134 }, { "epoch": 0.0, "learning_rate": 2.2167487684729067e-06, "loss": 3.634, "step": 135 }, { "epoch": 0.0, "learning_rate": 2.233169129720854e-06, "loss": 3.5591, "step": 136 }, { "epoch": 0.0, "learning_rate": 2.2495894909688014e-06, "loss": 3.4171, "step": 137 }, { "epoch": 0.0, "learning_rate": 2.266009852216749e-06, "loss": 3.6392, "step": 138 }, { "epoch": 0.0, "learning_rate": 2.2824302134646964e-06, "loss": 3.507, "step": 139 }, { "epoch": 0.0, "learning_rate": 2.2988505747126437e-06, "loss": 3.5361, "step": 140 }, { "epoch": 0.0, "learning_rate": 2.315270935960591e-06, "loss": 3.5413, "step": 141 }, { "epoch": 0.0, "learning_rate": 2.331691297208539e-06, "loss": 3.5215, "step": 142 }, { "epoch": 0.0, "learning_rate": 2.348111658456486e-06, "loss": 3.5962, "step": 143 }, { "epoch": 0.0, "learning_rate": 2.3645320197044334e-06, "loss": 3.3897, "step": 144 }, { "epoch": 0.0, "learning_rate": 2.380952380952381e-06, "loss": 3.4889, "step": 145 }, { "epoch": 0.0, "learning_rate": 2.397372742200329e-06, "loss": 3.5991, "step": 146 }, { "epoch": 0.0, "learning_rate": 2.4137931034482762e-06, "loss": 3.5501, "step": 147 }, { "epoch": 0.0, "learning_rate": 2.4302134646962236e-06, "loss": 3.7437, "step": 148 }, { "epoch": 0.0, "learning_rate": 2.446633825944171e-06, "loss": 3.4372, "step": 149 }, { "epoch": 0.0, "learning_rate": 2.4630541871921186e-06, "loss": 3.4, "step": 150 }, { "epoch": 0.0, "learning_rate": 2.479474548440066e-06, "loss": 3.5379, "step": 151 }, { "epoch": 0.0, "learning_rate": 2.4958949096880133e-06, "loss": 3.3221, "step": 152 }, { "epoch": 0.0, "learning_rate": 2.512315270935961e-06, "loss": 3.5102, "step": 153 }, { "epoch": 0.0, "learning_rate": 2.5287356321839083e-06, "loss": 3.3635, "step": 154 }, { "epoch": 0.0, "learning_rate": 2.545155993431856e-06, "loss": 3.4808, "step": 155 }, { "epoch": 0.0, "learning_rate": 2.5615763546798034e-06, "loss": 3.4331, "step": 156 }, { "epoch": 0.0, "learning_rate": 2.5779967159277507e-06, "loss": 3.4487, "step": 157 }, { "epoch": 0.0, "learning_rate": 2.594417077175698e-06, "loss": 3.4468, "step": 158 }, { "epoch": 0.0, "learning_rate": 2.6108374384236458e-06, "loss": 3.3622, "step": 159 }, { "epoch": 0.0, "learning_rate": 2.627257799671593e-06, "loss": 3.3991, "step": 160 }, { "epoch": 0.0, "learning_rate": 2.6436781609195404e-06, "loss": 3.3134, "step": 161 }, { "epoch": 0.0, "learning_rate": 2.660098522167488e-06, "loss": 3.5598, "step": 162 }, { "epoch": 0.0, "learning_rate": 2.6765188834154355e-06, "loss": 3.349, "step": 163 }, { "epoch": 0.0, "learning_rate": 2.692939244663383e-06, "loss": 3.6029, "step": 164 }, { "epoch": 0.0, "learning_rate": 2.70935960591133e-06, "loss": 3.4134, "step": 165 }, { "epoch": 0.0, "learning_rate": 2.725779967159278e-06, "loss": 3.4509, "step": 166 }, { "epoch": 0.0, "learning_rate": 2.742200328407225e-06, "loss": 3.2554, "step": 167 }, { "epoch": 0.0, "learning_rate": 2.7586206896551725e-06, "loss": 3.4265, "step": 168 }, { "epoch": 0.0, "learning_rate": 2.7750410509031202e-06, "loss": 3.5939, "step": 169 }, { "epoch": 0.0, "learning_rate": 2.7914614121510676e-06, "loss": 3.4803, "step": 170 }, { "epoch": 0.0, "learning_rate": 2.807881773399015e-06, "loss": 3.3886, "step": 171 }, { "epoch": 0.0, "learning_rate": 2.824302134646962e-06, "loss": 3.343, "step": 172 }, { "epoch": 0.0, "learning_rate": 2.84072249589491e-06, "loss": 3.459, "step": 173 }, { "epoch": 0.0, "learning_rate": 2.8571428571428573e-06, "loss": 3.4302, "step": 174 }, { "epoch": 0.0, "learning_rate": 2.8735632183908046e-06, "loss": 3.359, "step": 175 }, { "epoch": 0.0, "learning_rate": 2.889983579638752e-06, "loss": 3.4022, "step": 176 }, { "epoch": 0.0, "learning_rate": 2.9064039408866996e-06, "loss": 3.4336, "step": 177 }, { "epoch": 0.0, "learning_rate": 2.922824302134647e-06, "loss": 3.3491, "step": 178 }, { "epoch": 0.0, "learning_rate": 2.9392446633825943e-06, "loss": 3.3608, "step": 179 }, { "epoch": 0.0, "learning_rate": 2.9556650246305424e-06, "loss": 3.2793, "step": 180 }, { "epoch": 0.0, "learning_rate": 2.9720853858784898e-06, "loss": 3.3848, "step": 181 }, { "epoch": 0.0, "learning_rate": 2.988505747126437e-06, "loss": 3.3194, "step": 182 }, { "epoch": 0.0, "learning_rate": 3.004926108374385e-06, "loss": 3.3892, "step": 183 }, { "epoch": 0.0, "learning_rate": 3.021346469622332e-06, "loss": 3.3461, "step": 184 }, { "epoch": 0.0, "learning_rate": 3.0377668308702795e-06, "loss": 3.4829, "step": 185 }, { "epoch": 0.0, "learning_rate": 3.054187192118227e-06, "loss": 3.6172, "step": 186 }, { "epoch": 0.0, "learning_rate": 3.0706075533661745e-06, "loss": 3.2807, "step": 187 }, { "epoch": 0.0, "learning_rate": 3.087027914614122e-06, "loss": 3.3598, "step": 188 }, { "epoch": 0.0, "learning_rate": 3.103448275862069e-06, "loss": 3.1897, "step": 189 }, { "epoch": 0.0, "learning_rate": 3.119868637110017e-06, "loss": 3.1812, "step": 190 }, { "epoch": 0.0, "learning_rate": 3.1362889983579642e-06, "loss": 3.57, "step": 191 }, { "epoch": 0.0, "learning_rate": 3.1527093596059115e-06, "loss": 3.2921, "step": 192 }, { "epoch": 0.0, "learning_rate": 3.169129720853859e-06, "loss": 3.4092, "step": 193 }, { "epoch": 0.0, "learning_rate": 3.1855500821018066e-06, "loss": 3.3335, "step": 194 }, { "epoch": 0.0, "learning_rate": 3.201970443349754e-06, "loss": 3.2035, "step": 195 }, { "epoch": 0.0, "learning_rate": 3.2183908045977012e-06, "loss": 3.3308, "step": 196 }, { "epoch": 0.0, "learning_rate": 3.234811165845649e-06, "loss": 3.3702, "step": 197 }, { "epoch": 0.0, "learning_rate": 3.2512315270935963e-06, "loss": 3.2708, "step": 198 }, { "epoch": 0.0, "learning_rate": 3.2676518883415436e-06, "loss": 3.2133, "step": 199 }, { "epoch": 0.0, "learning_rate": 3.284072249589491e-06, "loss": 3.2844, "step": 200 }, { "epoch": 0.0, "learning_rate": 3.3004926108374387e-06, "loss": 3.4254, "step": 201 }, { "epoch": 0.0, "learning_rate": 3.316912972085386e-06, "loss": 3.3302, "step": 202 }, { "epoch": 0.01, "learning_rate": 3.3333333333333333e-06, "loss": 3.2297, "step": 203 }, { "epoch": 0.01, "learning_rate": 3.349753694581281e-06, "loss": 3.2615, "step": 204 }, { "epoch": 0.01, "learning_rate": 3.3661740558292284e-06, "loss": 3.2075, "step": 205 }, { "epoch": 0.01, "learning_rate": 3.3825944170771757e-06, "loss": 3.139, "step": 206 }, { "epoch": 0.01, "learning_rate": 3.399014778325123e-06, "loss": 3.4068, "step": 207 }, { "epoch": 0.01, "learning_rate": 3.415435139573071e-06, "loss": 3.3227, "step": 208 }, { "epoch": 0.01, "learning_rate": 3.4318555008210185e-06, "loss": 3.3399, "step": 209 }, { "epoch": 0.01, "learning_rate": 3.448275862068966e-06, "loss": 3.3, "step": 210 }, { "epoch": 0.01, "learning_rate": 3.4646962233169136e-06, "loss": 3.4119, "step": 211 }, { "epoch": 0.01, "learning_rate": 3.481116584564861e-06, "loss": 3.1982, "step": 212 }, { "epoch": 0.01, "learning_rate": 3.497536945812808e-06, "loss": 3.4147, "step": 213 }, { "epoch": 0.01, "learning_rate": 3.513957307060756e-06, "loss": 3.2973, "step": 214 }, { "epoch": 0.01, "learning_rate": 3.5303776683087033e-06, "loss": 3.1672, "step": 215 }, { "epoch": 0.01, "learning_rate": 3.5467980295566506e-06, "loss": 3.4486, "step": 216 }, { "epoch": 0.01, "learning_rate": 3.563218390804598e-06, "loss": 3.3802, "step": 217 }, { "epoch": 0.01, "learning_rate": 3.5796387520525457e-06, "loss": 3.2541, "step": 218 }, { "epoch": 0.01, "learning_rate": 3.596059113300493e-06, "loss": 3.2933, "step": 219 }, { "epoch": 0.01, "learning_rate": 3.6124794745484403e-06, "loss": 3.1309, "step": 220 }, { "epoch": 0.01, "learning_rate": 3.628899835796388e-06, "loss": 3.1423, "step": 221 }, { "epoch": 0.01, "learning_rate": 3.6453201970443354e-06, "loss": 3.1727, "step": 222 }, { "epoch": 0.01, "learning_rate": 3.6617405582922827e-06, "loss": 3.1662, "step": 223 }, { "epoch": 0.01, "learning_rate": 3.67816091954023e-06, "loss": 3.1687, "step": 224 }, { "epoch": 0.01, "learning_rate": 3.6945812807881777e-06, "loss": 3.0247, "step": 225 }, { "epoch": 0.01, "learning_rate": 3.711001642036125e-06, "loss": 3.1621, "step": 226 }, { "epoch": 0.01, "learning_rate": 3.7274220032840724e-06, "loss": 3.2499, "step": 227 }, { "epoch": 0.01, "learning_rate": 3.7438423645320197e-06, "loss": 3.1547, "step": 228 }, { "epoch": 0.01, "learning_rate": 3.7602627257799674e-06, "loss": 3.3796, "step": 229 }, { "epoch": 0.01, "learning_rate": 3.7766830870279148e-06, "loss": 3.1841, "step": 230 }, { "epoch": 0.01, "learning_rate": 3.793103448275862e-06, "loss": 3.2647, "step": 231 }, { "epoch": 0.01, "learning_rate": 3.80952380952381e-06, "loss": 3.1345, "step": 232 }, { "epoch": 0.01, "learning_rate": 3.825944170771757e-06, "loss": 3.2813, "step": 233 }, { "epoch": 0.01, "learning_rate": 3.842364532019705e-06, "loss": 3.3354, "step": 234 }, { "epoch": 0.01, "learning_rate": 3.858784893267652e-06, "loss": 3.103, "step": 235 }, { "epoch": 0.01, "learning_rate": 3.8752052545155995e-06, "loss": 3.2316, "step": 236 }, { "epoch": 0.01, "learning_rate": 3.891625615763547e-06, "loss": 3.3631, "step": 237 }, { "epoch": 0.01, "learning_rate": 3.908045977011495e-06, "loss": 3.2681, "step": 238 }, { "epoch": 0.01, "learning_rate": 3.924466338259442e-06, "loss": 3.3214, "step": 239 }, { "epoch": 0.01, "learning_rate": 3.94088669950739e-06, "loss": 3.3264, "step": 240 }, { "epoch": 0.01, "learning_rate": 3.957307060755337e-06, "loss": 3.0598, "step": 241 }, { "epoch": 0.01, "learning_rate": 3.973727422003284e-06, "loss": 3.3492, "step": 242 }, { "epoch": 0.01, "learning_rate": 3.990147783251232e-06, "loss": 3.2358, "step": 243 }, { "epoch": 0.01, "learning_rate": 4.00656814449918e-06, "loss": 3.1994, "step": 244 }, { "epoch": 0.01, "learning_rate": 4.022988505747127e-06, "loss": 3.2168, "step": 245 }, { "epoch": 0.01, "learning_rate": 4.039408866995074e-06, "loss": 3.0989, "step": 246 }, { "epoch": 0.01, "learning_rate": 4.055829228243022e-06, "loss": 3.2448, "step": 247 }, { "epoch": 0.01, "learning_rate": 4.072249589490969e-06, "loss": 3.3528, "step": 248 }, { "epoch": 0.01, "learning_rate": 4.088669950738917e-06, "loss": 3.2588, "step": 249 }, { "epoch": 0.01, "learning_rate": 4.105090311986864e-06, "loss": 3.1567, "step": 250 }, { "epoch": 0.01, "learning_rate": 4.121510673234811e-06, "loss": 3.1919, "step": 251 }, { "epoch": 0.01, "learning_rate": 4.137931034482759e-06, "loss": 3.0843, "step": 252 }, { "epoch": 0.01, "learning_rate": 4.154351395730706e-06, "loss": 3.0697, "step": 253 }, { "epoch": 0.01, "learning_rate": 4.170771756978654e-06, "loss": 3.1874, "step": 254 }, { "epoch": 0.01, "learning_rate": 4.1871921182266015e-06, "loss": 3.2276, "step": 255 }, { "epoch": 0.01, "learning_rate": 4.2036124794745484e-06, "loss": 3.0092, "step": 256 }, { "epoch": 0.01, "learning_rate": 4.220032840722496e-06, "loss": 3.1903, "step": 257 }, { "epoch": 0.01, "learning_rate": 4.236453201970444e-06, "loss": 3.1196, "step": 258 }, { "epoch": 0.01, "learning_rate": 4.252873563218391e-06, "loss": 3.1427, "step": 259 }, { "epoch": 0.01, "learning_rate": 4.2692939244663386e-06, "loss": 3.1781, "step": 260 }, { "epoch": 0.01, "learning_rate": 4.2857142857142855e-06, "loss": 3.1266, "step": 261 }, { "epoch": 0.01, "learning_rate": 4.302134646962233e-06, "loss": 3.3254, "step": 262 }, { "epoch": 0.01, "learning_rate": 4.318555008210181e-06, "loss": 3.0318, "step": 263 }, { "epoch": 0.01, "learning_rate": 4.334975369458129e-06, "loss": 3.1273, "step": 264 }, { "epoch": 0.01, "learning_rate": 4.3513957307060764e-06, "loss": 3.1153, "step": 265 }, { "epoch": 0.01, "learning_rate": 4.367816091954023e-06, "loss": 3.2513, "step": 266 }, { "epoch": 0.01, "learning_rate": 4.384236453201971e-06, "loss": 3.1967, "step": 267 }, { "epoch": 0.01, "learning_rate": 4.400656814449919e-06, "loss": 3.0774, "step": 268 }, { "epoch": 0.01, "learning_rate": 4.417077175697866e-06, "loss": 3.0525, "step": 269 }, { "epoch": 0.01, "learning_rate": 4.4334975369458135e-06, "loss": 3.1925, "step": 270 }, { "epoch": 0.01, "learning_rate": 4.44991789819376e-06, "loss": 3.121, "step": 271 }, { "epoch": 0.01, "learning_rate": 4.466338259441708e-06, "loss": 3.2348, "step": 272 }, { "epoch": 0.01, "learning_rate": 4.482758620689656e-06, "loss": 3.165, "step": 273 }, { "epoch": 0.01, "learning_rate": 4.499178981937603e-06, "loss": 3.0947, "step": 274 }, { "epoch": 0.01, "learning_rate": 4.5155993431855505e-06, "loss": 3.1683, "step": 275 }, { "epoch": 0.01, "learning_rate": 4.532019704433498e-06, "loss": 3.1191, "step": 276 }, { "epoch": 0.01, "learning_rate": 4.548440065681445e-06, "loss": 3.1713, "step": 277 }, { "epoch": 0.01, "learning_rate": 4.564860426929393e-06, "loss": 3.2359, "step": 278 }, { "epoch": 0.01, "learning_rate": 4.581280788177341e-06, "loss": 3.0722, "step": 279 }, { "epoch": 0.01, "learning_rate": 4.5977011494252875e-06, "loss": 3.1559, "step": 280 }, { "epoch": 0.01, "learning_rate": 4.614121510673235e-06, "loss": 3.117, "step": 281 }, { "epoch": 0.01, "learning_rate": 4.630541871921182e-06, "loss": 3.0061, "step": 282 }, { "epoch": 0.01, "learning_rate": 4.64696223316913e-06, "loss": 3.2252, "step": 283 }, { "epoch": 0.01, "learning_rate": 4.663382594417078e-06, "loss": 3.0077, "step": 284 }, { "epoch": 0.01, "learning_rate": 4.6798029556650245e-06, "loss": 3.4167, "step": 285 }, { "epoch": 0.01, "learning_rate": 4.696223316912972e-06, "loss": 3.0553, "step": 286 }, { "epoch": 0.01, "learning_rate": 4.71264367816092e-06, "loss": 3.0878, "step": 287 }, { "epoch": 0.01, "learning_rate": 4.729064039408867e-06, "loss": 3.1273, "step": 288 }, { "epoch": 0.01, "learning_rate": 4.745484400656815e-06, "loss": 3.2525, "step": 289 }, { "epoch": 0.01, "learning_rate": 4.761904761904762e-06, "loss": 3.2411, "step": 290 }, { "epoch": 0.01, "learning_rate": 4.77832512315271e-06, "loss": 3.2551, "step": 291 }, { "epoch": 0.01, "learning_rate": 4.794745484400658e-06, "loss": 3.0349, "step": 292 }, { "epoch": 0.01, "learning_rate": 4.811165845648605e-06, "loss": 3.0671, "step": 293 }, { "epoch": 0.01, "learning_rate": 4.8275862068965525e-06, "loss": 3.1478, "step": 294 }, { "epoch": 0.01, "learning_rate": 4.844006568144499e-06, "loss": 3.3131, "step": 295 }, { "epoch": 0.01, "learning_rate": 4.860426929392447e-06, "loss": 3.0436, "step": 296 }, { "epoch": 0.01, "learning_rate": 4.876847290640395e-06, "loss": 3.1537, "step": 297 }, { "epoch": 0.01, "learning_rate": 4.893267651888342e-06, "loss": 3.2208, "step": 298 }, { "epoch": 0.01, "learning_rate": 4.9096880131362895e-06, "loss": 3.1871, "step": 299 }, { "epoch": 0.01, "learning_rate": 4.926108374384237e-06, "loss": 3.0949, "step": 300 }, { "epoch": 0.01, "learning_rate": 4.942528735632184e-06, "loss": 3.1578, "step": 301 }, { "epoch": 0.01, "learning_rate": 4.958949096880132e-06, "loss": 3.2802, "step": 302 }, { "epoch": 0.01, "learning_rate": 4.97536945812808e-06, "loss": 3.1986, "step": 303 }, { "epoch": 0.01, "learning_rate": 4.9917898193760265e-06, "loss": 3.0599, "step": 304 }, { "epoch": 0.01, "learning_rate": 5.008210180623974e-06, "loss": 3.3097, "step": 305 }, { "epoch": 0.01, "learning_rate": 5.024630541871922e-06, "loss": 3.1083, "step": 306 }, { "epoch": 0.01, "learning_rate": 5.04105090311987e-06, "loss": 3.0029, "step": 307 }, { "epoch": 0.01, "learning_rate": 5.057471264367817e-06, "loss": 3.1314, "step": 308 }, { "epoch": 0.01, "learning_rate": 5.073891625615764e-06, "loss": 2.9775, "step": 309 }, { "epoch": 0.01, "learning_rate": 5.090311986863712e-06, "loss": 2.9198, "step": 310 }, { "epoch": 0.01, "learning_rate": 5.106732348111659e-06, "loss": 3.1329, "step": 311 }, { "epoch": 0.01, "learning_rate": 5.123152709359607e-06, "loss": 3.1062, "step": 312 }, { "epoch": 0.01, "learning_rate": 5.1395730706075545e-06, "loss": 2.9927, "step": 313 }, { "epoch": 0.01, "learning_rate": 5.1559934318555014e-06, "loss": 3.1253, "step": 314 }, { "epoch": 0.01, "learning_rate": 5.172413793103449e-06, "loss": 3.1752, "step": 315 }, { "epoch": 0.01, "learning_rate": 5.188834154351396e-06, "loss": 3.0748, "step": 316 }, { "epoch": 0.01, "learning_rate": 5.205254515599344e-06, "loss": 3.0121, "step": 317 }, { "epoch": 0.01, "learning_rate": 5.2216748768472915e-06, "loss": 3.2436, "step": 318 }, { "epoch": 0.01, "learning_rate": 5.2380952380952384e-06, "loss": 3.1088, "step": 319 }, { "epoch": 0.01, "learning_rate": 5.254515599343186e-06, "loss": 2.9768, "step": 320 }, { "epoch": 0.01, "learning_rate": 5.270935960591134e-06, "loss": 3.1633, "step": 321 }, { "epoch": 0.01, "learning_rate": 5.287356321839081e-06, "loss": 3.0039, "step": 322 }, { "epoch": 0.01, "learning_rate": 5.3037766830870286e-06, "loss": 3.2203, "step": 323 }, { "epoch": 0.01, "learning_rate": 5.320197044334976e-06, "loss": 2.949, "step": 324 }, { "epoch": 0.01, "learning_rate": 5.336617405582923e-06, "loss": 3.1402, "step": 325 }, { "epoch": 0.01, "learning_rate": 5.353037766830871e-06, "loss": 2.9637, "step": 326 }, { "epoch": 0.01, "learning_rate": 5.369458128078819e-06, "loss": 2.9818, "step": 327 }, { "epoch": 0.01, "learning_rate": 5.385878489326766e-06, "loss": 3.145, "step": 328 }, { "epoch": 0.01, "learning_rate": 5.402298850574713e-06, "loss": 3.1523, "step": 329 }, { "epoch": 0.01, "learning_rate": 5.41871921182266e-06, "loss": 2.9487, "step": 330 }, { "epoch": 0.01, "learning_rate": 5.435139573070608e-06, "loss": 3.1634, "step": 331 }, { "epoch": 0.01, "learning_rate": 5.451559934318556e-06, "loss": 2.9358, "step": 332 }, { "epoch": 0.01, "learning_rate": 5.467980295566503e-06, "loss": 3.1873, "step": 333 }, { "epoch": 0.01, "learning_rate": 5.48440065681445e-06, "loss": 3.101, "step": 334 }, { "epoch": 0.01, "learning_rate": 5.500821018062398e-06, "loss": 3.0391, "step": 335 }, { "epoch": 0.01, "learning_rate": 5.517241379310345e-06, "loss": 3.1939, "step": 336 }, { "epoch": 0.01, "learning_rate": 5.533661740558293e-06, "loss": 2.9653, "step": 337 }, { "epoch": 0.01, "learning_rate": 5.5500821018062405e-06, "loss": 3.0502, "step": 338 }, { "epoch": 0.01, "learning_rate": 5.566502463054187e-06, "loss": 3.0543, "step": 339 }, { "epoch": 0.01, "learning_rate": 5.582922824302135e-06, "loss": 3.0244, "step": 340 }, { "epoch": 0.01, "learning_rate": 5.599343185550082e-06, "loss": 3.1227, "step": 341 }, { "epoch": 0.01, "learning_rate": 5.61576354679803e-06, "loss": 2.9331, "step": 342 }, { "epoch": 0.01, "learning_rate": 5.6321839080459775e-06, "loss": 3.0732, "step": 343 }, { "epoch": 0.01, "learning_rate": 5.648604269293924e-06, "loss": 3.0928, "step": 344 }, { "epoch": 0.01, "learning_rate": 5.665024630541872e-06, "loss": 3.1667, "step": 345 }, { "epoch": 0.01, "learning_rate": 5.68144499178982e-06, "loss": 3.179, "step": 346 }, { "epoch": 0.01, "learning_rate": 5.697865353037767e-06, "loss": 2.976, "step": 347 }, { "epoch": 0.01, "learning_rate": 5.7142857142857145e-06, "loss": 3.1855, "step": 348 }, { "epoch": 0.01, "learning_rate": 5.730706075533662e-06, "loss": 2.9312, "step": 349 }, { "epoch": 0.01, "learning_rate": 5.747126436781609e-06, "loss": 3.0412, "step": 350 }, { "epoch": 0.01, "learning_rate": 5.763546798029557e-06, "loss": 3.0242, "step": 351 }, { "epoch": 0.01, "learning_rate": 5.779967159277504e-06, "loss": 2.8561, "step": 352 }, { "epoch": 0.01, "learning_rate": 5.7963875205254515e-06, "loss": 2.8717, "step": 353 }, { "epoch": 0.01, "learning_rate": 5.812807881773399e-06, "loss": 2.9486, "step": 354 }, { "epoch": 0.01, "learning_rate": 5.829228243021346e-06, "loss": 2.9893, "step": 355 }, { "epoch": 0.01, "learning_rate": 5.845648604269294e-06, "loss": 2.9564, "step": 356 }, { "epoch": 0.01, "learning_rate": 5.862068965517242e-06, "loss": 3.1564, "step": 357 }, { "epoch": 0.01, "learning_rate": 5.8784893267651886e-06, "loss": 2.9595, "step": 358 }, { "epoch": 0.01, "learning_rate": 5.894909688013136e-06, "loss": 3.109, "step": 359 }, { "epoch": 0.01, "learning_rate": 5.911330049261085e-06, "loss": 2.9301, "step": 360 }, { "epoch": 0.01, "learning_rate": 5.927750410509033e-06, "loss": 2.8475, "step": 361 }, { "epoch": 0.01, "learning_rate": 5.9441707717569795e-06, "loss": 3.0868, "step": 362 }, { "epoch": 0.01, "learning_rate": 5.960591133004927e-06, "loss": 3.069, "step": 363 }, { "epoch": 0.01, "learning_rate": 5.977011494252874e-06, "loss": 3.1375, "step": 364 }, { "epoch": 0.01, "learning_rate": 5.993431855500822e-06, "loss": 3.0486, "step": 365 }, { "epoch": 0.01, "learning_rate": 6.00985221674877e-06, "loss": 3.0029, "step": 366 }, { "epoch": 0.01, "learning_rate": 6.0262725779967165e-06, "loss": 2.9402, "step": 367 }, { "epoch": 0.01, "learning_rate": 6.042692939244664e-06, "loss": 2.9839, "step": 368 }, { "epoch": 0.01, "learning_rate": 6.059113300492612e-06, "loss": 3.1261, "step": 369 }, { "epoch": 0.01, "learning_rate": 6.075533661740559e-06, "loss": 3.0259, "step": 370 }, { "epoch": 0.01, "learning_rate": 6.091954022988507e-06, "loss": 2.9274, "step": 371 }, { "epoch": 0.01, "learning_rate": 6.108374384236454e-06, "loss": 2.9325, "step": 372 }, { "epoch": 0.01, "learning_rate": 6.124794745484401e-06, "loss": 3.0881, "step": 373 }, { "epoch": 0.01, "learning_rate": 6.141215106732349e-06, "loss": 3.0715, "step": 374 }, { "epoch": 0.01, "learning_rate": 6.157635467980296e-06, "loss": 2.9397, "step": 375 }, { "epoch": 0.01, "learning_rate": 6.174055829228244e-06, "loss": 3.1977, "step": 376 }, { "epoch": 0.01, "learning_rate": 6.1904761904761914e-06, "loss": 3.2123, "step": 377 }, { "epoch": 0.01, "learning_rate": 6.206896551724138e-06, "loss": 3.1492, "step": 378 }, { "epoch": 0.01, "learning_rate": 6.223316912972086e-06, "loss": 3.0198, "step": 379 }, { "epoch": 0.01, "learning_rate": 6.239737274220034e-06, "loss": 2.9013, "step": 380 }, { "epoch": 0.01, "learning_rate": 6.256157635467981e-06, "loss": 2.9261, "step": 381 }, { "epoch": 0.01, "learning_rate": 6.2725779967159284e-06, "loss": 2.98, "step": 382 }, { "epoch": 0.01, "learning_rate": 6.288998357963876e-06, "loss": 3.1952, "step": 383 }, { "epoch": 0.01, "learning_rate": 6.305418719211823e-06, "loss": 2.9479, "step": 384 }, { "epoch": 0.01, "learning_rate": 6.321839080459771e-06, "loss": 3.067, "step": 385 }, { "epoch": 0.01, "learning_rate": 6.338259441707718e-06, "loss": 2.9753, "step": 386 }, { "epoch": 0.01, "learning_rate": 6.3546798029556655e-06, "loss": 2.957, "step": 387 }, { "epoch": 0.01, "learning_rate": 6.371100164203613e-06, "loss": 2.9392, "step": 388 }, { "epoch": 0.01, "learning_rate": 6.38752052545156e-06, "loss": 3.0369, "step": 389 }, { "epoch": 0.01, "learning_rate": 6.403940886699508e-06, "loss": 2.9061, "step": 390 }, { "epoch": 0.01, "learning_rate": 6.420361247947456e-06, "loss": 3.1193, "step": 391 }, { "epoch": 0.01, "learning_rate": 6.4367816091954025e-06, "loss": 3.024, "step": 392 }, { "epoch": 0.01, "learning_rate": 6.45320197044335e-06, "loss": 3.0894, "step": 393 }, { "epoch": 0.01, "learning_rate": 6.469622331691298e-06, "loss": 3.0344, "step": 394 }, { "epoch": 0.01, "learning_rate": 6.486042692939245e-06, "loss": 3.054, "step": 395 }, { "epoch": 0.01, "learning_rate": 6.502463054187193e-06, "loss": 3.0629, "step": 396 }, { "epoch": 0.01, "learning_rate": 6.5188834154351395e-06, "loss": 3.0556, "step": 397 }, { "epoch": 0.01, "learning_rate": 6.535303776683087e-06, "loss": 3.0769, "step": 398 }, { "epoch": 0.01, "learning_rate": 6.551724137931035e-06, "loss": 2.9297, "step": 399 }, { "epoch": 0.01, "learning_rate": 6.568144499178982e-06, "loss": 3.0945, "step": 400 }, { "epoch": 0.01, "learning_rate": 6.58456486042693e-06, "loss": 2.9557, "step": 401 }, { "epoch": 0.01, "learning_rate": 6.600985221674877e-06, "loss": 2.9409, "step": 402 }, { "epoch": 0.01, "learning_rate": 6.617405582922824e-06, "loss": 3.046, "step": 403 }, { "epoch": 0.01, "learning_rate": 6.633825944170772e-06, "loss": 2.9091, "step": 404 }, { "epoch": 0.01, "learning_rate": 6.65024630541872e-06, "loss": 2.8753, "step": 405 }, { "epoch": 0.01, "learning_rate": 6.666666666666667e-06, "loss": 2.9637, "step": 406 }, { "epoch": 0.01, "learning_rate": 6.683087027914614e-06, "loss": 2.9936, "step": 407 }, { "epoch": 0.01, "learning_rate": 6.699507389162562e-06, "loss": 3.0563, "step": 408 }, { "epoch": 0.01, "learning_rate": 6.715927750410509e-06, "loss": 2.9969, "step": 409 }, { "epoch": 0.01, "learning_rate": 6.732348111658457e-06, "loss": 2.9349, "step": 410 }, { "epoch": 0.01, "learning_rate": 6.748768472906404e-06, "loss": 3.0395, "step": 411 }, { "epoch": 0.01, "learning_rate": 6.765188834154351e-06, "loss": 2.9778, "step": 412 }, { "epoch": 0.01, "learning_rate": 6.781609195402299e-06, "loss": 3.0885, "step": 413 }, { "epoch": 0.01, "learning_rate": 6.798029556650246e-06, "loss": 2.7541, "step": 414 }, { "epoch": 0.01, "learning_rate": 6.814449917898194e-06, "loss": 2.9472, "step": 415 }, { "epoch": 0.01, "learning_rate": 6.830870279146142e-06, "loss": 3.0147, "step": 416 }, { "epoch": 0.01, "learning_rate": 6.84729064039409e-06, "loss": 2.9154, "step": 417 }, { "epoch": 0.01, "learning_rate": 6.863711001642037e-06, "loss": 3.0613, "step": 418 }, { "epoch": 0.01, "learning_rate": 6.880131362889985e-06, "loss": 3.1807, "step": 419 }, { "epoch": 0.01, "learning_rate": 6.896551724137932e-06, "loss": 3.1891, "step": 420 }, { "epoch": 0.01, "learning_rate": 6.912972085385879e-06, "loss": 3.224, "step": 421 }, { "epoch": 0.01, "learning_rate": 6.929392446633827e-06, "loss": 2.9942, "step": 422 }, { "epoch": 0.01, "learning_rate": 6.945812807881774e-06, "loss": 3.014, "step": 423 }, { "epoch": 0.01, "learning_rate": 6.962233169129722e-06, "loss": 2.9916, "step": 424 }, { "epoch": 0.01, "learning_rate": 6.9786535303776695e-06, "loss": 3.0952, "step": 425 }, { "epoch": 0.01, "learning_rate": 6.995073891625616e-06, "loss": 2.8681, "step": 426 }, { "epoch": 0.01, "learning_rate": 7.011494252873564e-06, "loss": 2.8858, "step": 427 }, { "epoch": 0.01, "learning_rate": 7.027914614121512e-06, "loss": 2.9331, "step": 428 }, { "epoch": 0.01, "learning_rate": 7.044334975369459e-06, "loss": 2.8987, "step": 429 }, { "epoch": 0.01, "learning_rate": 7.0607553366174065e-06, "loss": 3.0192, "step": 430 }, { "epoch": 0.01, "learning_rate": 7.0771756978653534e-06, "loss": 2.9244, "step": 431 }, { "epoch": 0.01, "learning_rate": 7.093596059113301e-06, "loss": 2.9007, "step": 432 }, { "epoch": 0.01, "learning_rate": 7.110016420361249e-06, "loss": 2.8617, "step": 433 }, { "epoch": 0.01, "learning_rate": 7.126436781609196e-06, "loss": 2.9427, "step": 434 }, { "epoch": 0.01, "learning_rate": 7.1428571428571436e-06, "loss": 2.9586, "step": 435 }, { "epoch": 0.01, "learning_rate": 7.159277504105091e-06, "loss": 2.9235, "step": 436 }, { "epoch": 0.01, "learning_rate": 7.175697865353038e-06, "loss": 2.7811, "step": 437 }, { "epoch": 0.01, "learning_rate": 7.192118226600986e-06, "loss": 3.0233, "step": 438 }, { "epoch": 0.01, "learning_rate": 7.208538587848934e-06, "loss": 3.1764, "step": 439 }, { "epoch": 0.01, "learning_rate": 7.224958949096881e-06, "loss": 2.9551, "step": 440 }, { "epoch": 0.01, "learning_rate": 7.241379310344828e-06, "loss": 2.8024, "step": 441 }, { "epoch": 0.01, "learning_rate": 7.257799671592776e-06, "loss": 3.0501, "step": 442 }, { "epoch": 0.01, "learning_rate": 7.274220032840723e-06, "loss": 2.9685, "step": 443 }, { "epoch": 0.01, "learning_rate": 7.290640394088671e-06, "loss": 2.9345, "step": 444 }, { "epoch": 0.01, "learning_rate": 7.307060755336618e-06, "loss": 3.0088, "step": 445 }, { "epoch": 0.01, "learning_rate": 7.323481116584565e-06, "loss": 3.1488, "step": 446 }, { "epoch": 0.01, "learning_rate": 7.339901477832513e-06, "loss": 2.7991, "step": 447 }, { "epoch": 0.01, "learning_rate": 7.35632183908046e-06, "loss": 3.0313, "step": 448 }, { "epoch": 0.01, "learning_rate": 7.372742200328408e-06, "loss": 3.027, "step": 449 }, { "epoch": 0.01, "learning_rate": 7.3891625615763555e-06, "loss": 2.8705, "step": 450 }, { "epoch": 0.01, "learning_rate": 7.405582922824302e-06, "loss": 2.9695, "step": 451 }, { "epoch": 0.01, "learning_rate": 7.42200328407225e-06, "loss": 3.0606, "step": 452 }, { "epoch": 0.01, "learning_rate": 7.438423645320198e-06, "loss": 3.0774, "step": 453 }, { "epoch": 0.01, "learning_rate": 7.454844006568145e-06, "loss": 2.8841, "step": 454 }, { "epoch": 0.01, "learning_rate": 7.4712643678160925e-06, "loss": 2.8234, "step": 455 }, { "epoch": 0.01, "learning_rate": 7.487684729064039e-06, "loss": 2.7665, "step": 456 }, { "epoch": 0.01, "learning_rate": 7.504105090311987e-06, "loss": 3.0935, "step": 457 }, { "epoch": 0.01, "learning_rate": 7.520525451559935e-06, "loss": 2.8777, "step": 458 }, { "epoch": 0.01, "learning_rate": 7.536945812807882e-06, "loss": 2.841, "step": 459 }, { "epoch": 0.01, "learning_rate": 7.5533661740558295e-06, "loss": 2.9809, "step": 460 }, { "epoch": 0.01, "learning_rate": 7.569786535303777e-06, "loss": 2.9127, "step": 461 }, { "epoch": 0.01, "learning_rate": 7.586206896551724e-06, "loss": 2.9799, "step": 462 }, { "epoch": 0.01, "learning_rate": 7.602627257799672e-06, "loss": 2.9642, "step": 463 }, { "epoch": 0.01, "learning_rate": 7.61904761904762e-06, "loss": 2.9737, "step": 464 }, { "epoch": 0.01, "learning_rate": 7.635467980295567e-06, "loss": 2.8454, "step": 465 }, { "epoch": 0.01, "learning_rate": 7.651888341543514e-06, "loss": 3.1071, "step": 466 }, { "epoch": 0.01, "learning_rate": 7.668308702791462e-06, "loss": 3.0705, "step": 467 }, { "epoch": 0.01, "learning_rate": 7.68472906403941e-06, "loss": 2.9455, "step": 468 }, { "epoch": 0.01, "learning_rate": 7.701149425287356e-06, "loss": 3.0895, "step": 469 }, { "epoch": 0.01, "learning_rate": 7.717569786535304e-06, "loss": 2.928, "step": 470 }, { "epoch": 0.01, "learning_rate": 7.733990147783253e-06, "loss": 3.1855, "step": 471 }, { "epoch": 0.01, "learning_rate": 7.750410509031199e-06, "loss": 3.018, "step": 472 }, { "epoch": 0.01, "learning_rate": 7.766830870279147e-06, "loss": 2.917, "step": 473 }, { "epoch": 0.01, "learning_rate": 7.783251231527095e-06, "loss": 2.9024, "step": 474 }, { "epoch": 0.01, "learning_rate": 7.799671592775042e-06, "loss": 2.9158, "step": 475 }, { "epoch": 0.01, "learning_rate": 7.81609195402299e-06, "loss": 3.0149, "step": 476 }, { "epoch": 0.01, "learning_rate": 7.832512315270938e-06, "loss": 2.8738, "step": 477 }, { "epoch": 0.01, "learning_rate": 7.848932676518884e-06, "loss": 2.8854, "step": 478 }, { "epoch": 0.01, "learning_rate": 7.865353037766832e-06, "loss": 2.9755, "step": 479 }, { "epoch": 0.01, "learning_rate": 7.88177339901478e-06, "loss": 2.8857, "step": 480 }, { "epoch": 0.01, "learning_rate": 7.898193760262727e-06, "loss": 2.9783, "step": 481 }, { "epoch": 0.01, "learning_rate": 7.914614121510675e-06, "loss": 2.8654, "step": 482 }, { "epoch": 0.01, "learning_rate": 7.93103448275862e-06, "loss": 2.8885, "step": 483 }, { "epoch": 0.01, "learning_rate": 7.947454844006569e-06, "loss": 2.9086, "step": 484 }, { "epoch": 0.01, "learning_rate": 7.963875205254516e-06, "loss": 2.9317, "step": 485 }, { "epoch": 0.01, "learning_rate": 7.980295566502464e-06, "loss": 2.8626, "step": 486 }, { "epoch": 0.01, "learning_rate": 7.996715927750412e-06, "loss": 2.9236, "step": 487 }, { "epoch": 0.01, "learning_rate": 8.01313628899836e-06, "loss": 2.8887, "step": 488 }, { "epoch": 0.01, "learning_rate": 8.029556650246306e-06, "loss": 2.7896, "step": 489 }, { "epoch": 0.01, "learning_rate": 8.045977011494253e-06, "loss": 2.8531, "step": 490 }, { "epoch": 0.01, "learning_rate": 8.062397372742201e-06, "loss": 2.7791, "step": 491 }, { "epoch": 0.01, "learning_rate": 8.078817733990149e-06, "loss": 2.8783, "step": 492 }, { "epoch": 0.01, "learning_rate": 8.095238095238097e-06, "loss": 2.9844, "step": 493 }, { "epoch": 0.01, "learning_rate": 8.111658456486044e-06, "loss": 2.9629, "step": 494 }, { "epoch": 0.01, "learning_rate": 8.12807881773399e-06, "loss": 2.9869, "step": 495 }, { "epoch": 0.01, "learning_rate": 8.144499178981938e-06, "loss": 2.8378, "step": 496 }, { "epoch": 0.01, "learning_rate": 8.160919540229886e-06, "loss": 3.0601, "step": 497 }, { "epoch": 0.01, "learning_rate": 8.177339901477834e-06, "loss": 2.7791, "step": 498 }, { "epoch": 0.01, "learning_rate": 8.193760262725781e-06, "loss": 2.8387, "step": 499 }, { "epoch": 0.01, "learning_rate": 8.210180623973727e-06, "loss": 2.7115, "step": 500 }, { "epoch": 0.01, "learning_rate": 8.226600985221675e-06, "loss": 2.9688, "step": 501 }, { "epoch": 0.01, "learning_rate": 8.243021346469623e-06, "loss": 3.0555, "step": 502 }, { "epoch": 0.01, "learning_rate": 8.25944170771757e-06, "loss": 2.8896, "step": 503 }, { "epoch": 0.01, "learning_rate": 8.275862068965518e-06, "loss": 3.0547, "step": 504 }, { "epoch": 0.01, "learning_rate": 8.292282430213466e-06, "loss": 3.0317, "step": 505 }, { "epoch": 0.01, "learning_rate": 8.308702791461412e-06, "loss": 2.8379, "step": 506 }, { "epoch": 0.01, "learning_rate": 8.32512315270936e-06, "loss": 3.0226, "step": 507 }, { "epoch": 0.01, "learning_rate": 8.341543513957308e-06, "loss": 2.9728, "step": 508 }, { "epoch": 0.01, "learning_rate": 8.357963875205255e-06, "loss": 2.9839, "step": 509 }, { "epoch": 0.01, "learning_rate": 8.374384236453203e-06, "loss": 2.9483, "step": 510 }, { "epoch": 0.01, "learning_rate": 8.390804597701149e-06, "loss": 3.0991, "step": 511 }, { "epoch": 0.01, "learning_rate": 8.407224958949097e-06, "loss": 3.0402, "step": 512 }, { "epoch": 0.01, "learning_rate": 8.423645320197045e-06, "loss": 2.6352, "step": 513 }, { "epoch": 0.01, "learning_rate": 8.440065681444992e-06, "loss": 2.7741, "step": 514 }, { "epoch": 0.01, "learning_rate": 8.45648604269294e-06, "loss": 2.9179, "step": 515 }, { "epoch": 0.01, "learning_rate": 8.472906403940888e-06, "loss": 3.0372, "step": 516 }, { "epoch": 0.01, "learning_rate": 8.489326765188834e-06, "loss": 2.9171, "step": 517 }, { "epoch": 0.01, "learning_rate": 8.505747126436782e-06, "loss": 2.8597, "step": 518 }, { "epoch": 0.01, "learning_rate": 8.52216748768473e-06, "loss": 2.7778, "step": 519 }, { "epoch": 0.01, "learning_rate": 8.538587848932677e-06, "loss": 2.9832, "step": 520 }, { "epoch": 0.01, "learning_rate": 8.555008210180625e-06, "loss": 2.8357, "step": 521 }, { "epoch": 0.01, "learning_rate": 8.571428571428571e-06, "loss": 2.9343, "step": 522 }, { "epoch": 0.01, "learning_rate": 8.587848932676519e-06, "loss": 2.7481, "step": 523 }, { "epoch": 0.01, "learning_rate": 8.604269293924466e-06, "loss": 2.7859, "step": 524 }, { "epoch": 0.01, "learning_rate": 8.620689655172414e-06, "loss": 3.0025, "step": 525 }, { "epoch": 0.01, "learning_rate": 8.637110016420362e-06, "loss": 2.679, "step": 526 }, { "epoch": 0.01, "learning_rate": 8.65353037766831e-06, "loss": 2.7496, "step": 527 }, { "epoch": 0.01, "learning_rate": 8.669950738916257e-06, "loss": 3.0186, "step": 528 }, { "epoch": 0.01, "learning_rate": 8.686371100164205e-06, "loss": 2.9842, "step": 529 }, { "epoch": 0.01, "learning_rate": 8.702791461412153e-06, "loss": 2.9619, "step": 530 }, { "epoch": 0.01, "learning_rate": 8.719211822660099e-06, "loss": 2.9746, "step": 531 }, { "epoch": 0.01, "learning_rate": 8.735632183908047e-06, "loss": 2.8566, "step": 532 }, { "epoch": 0.01, "learning_rate": 8.752052545155994e-06, "loss": 2.9157, "step": 533 }, { "epoch": 0.01, "learning_rate": 8.768472906403942e-06, "loss": 2.9799, "step": 534 }, { "epoch": 0.01, "learning_rate": 8.78489326765189e-06, "loss": 2.8283, "step": 535 }, { "epoch": 0.01, "learning_rate": 8.801313628899838e-06, "loss": 2.9076, "step": 536 }, { "epoch": 0.01, "learning_rate": 8.817733990147784e-06, "loss": 2.9008, "step": 537 }, { "epoch": 0.01, "learning_rate": 8.834154351395731e-06, "loss": 2.7306, "step": 538 }, { "epoch": 0.01, "learning_rate": 8.85057471264368e-06, "loss": 2.8967, "step": 539 }, { "epoch": 0.01, "learning_rate": 8.866995073891627e-06, "loss": 2.9749, "step": 540 }, { "epoch": 0.01, "learning_rate": 8.883415435139575e-06, "loss": 2.8738, "step": 541 }, { "epoch": 0.01, "learning_rate": 8.89983579638752e-06, "loss": 2.9993, "step": 542 }, { "epoch": 0.01, "learning_rate": 8.916256157635468e-06, "loss": 3.1402, "step": 543 }, { "epoch": 0.01, "learning_rate": 8.932676518883416e-06, "loss": 2.8676, "step": 544 }, { "epoch": 0.01, "learning_rate": 8.949096880131364e-06, "loss": 2.8097, "step": 545 }, { "epoch": 0.01, "learning_rate": 8.965517241379312e-06, "loss": 2.9117, "step": 546 }, { "epoch": 0.01, "learning_rate": 8.98193760262726e-06, "loss": 2.9904, "step": 547 }, { "epoch": 0.01, "learning_rate": 8.998357963875205e-06, "loss": 3.0154, "step": 548 }, { "epoch": 0.01, "learning_rate": 9.014778325123153e-06, "loss": 2.7072, "step": 549 }, { "epoch": 0.01, "learning_rate": 9.031198686371101e-06, "loss": 2.9116, "step": 550 }, { "epoch": 0.01, "learning_rate": 9.047619047619049e-06, "loss": 2.9349, "step": 551 }, { "epoch": 0.01, "learning_rate": 9.064039408866996e-06, "loss": 2.9663, "step": 552 }, { "epoch": 0.01, "learning_rate": 9.080459770114942e-06, "loss": 2.8554, "step": 553 }, { "epoch": 0.01, "learning_rate": 9.09688013136289e-06, "loss": 2.7274, "step": 554 }, { "epoch": 0.01, "learning_rate": 9.113300492610838e-06, "loss": 2.7846, "step": 555 }, { "epoch": 0.01, "learning_rate": 9.129720853858786e-06, "loss": 2.9203, "step": 556 }, { "epoch": 0.01, "learning_rate": 9.146141215106733e-06, "loss": 2.9291, "step": 557 }, { "epoch": 0.01, "learning_rate": 9.162561576354681e-06, "loss": 2.8553, "step": 558 }, { "epoch": 0.01, "learning_rate": 9.178981937602627e-06, "loss": 2.9293, "step": 559 }, { "epoch": 0.01, "learning_rate": 9.195402298850575e-06, "loss": 2.7911, "step": 560 }, { "epoch": 0.01, "learning_rate": 9.211822660098523e-06, "loss": 3.1461, "step": 561 }, { "epoch": 0.01, "learning_rate": 9.22824302134647e-06, "loss": 2.9946, "step": 562 }, { "epoch": 0.01, "learning_rate": 9.244663382594418e-06, "loss": 2.8756, "step": 563 }, { "epoch": 0.01, "learning_rate": 9.261083743842364e-06, "loss": 2.7486, "step": 564 }, { "epoch": 0.01, "learning_rate": 9.277504105090312e-06, "loss": 2.8989, "step": 565 }, { "epoch": 0.01, "learning_rate": 9.29392446633826e-06, "loss": 2.9941, "step": 566 }, { "epoch": 0.01, "learning_rate": 9.310344827586207e-06, "loss": 2.8425, "step": 567 }, { "epoch": 0.01, "learning_rate": 9.326765188834155e-06, "loss": 2.814, "step": 568 }, { "epoch": 0.01, "learning_rate": 9.343185550082103e-06, "loss": 2.8232, "step": 569 }, { "epoch": 0.01, "learning_rate": 9.359605911330049e-06, "loss": 2.9691, "step": 570 }, { "epoch": 0.01, "learning_rate": 9.376026272577997e-06, "loss": 2.8744, "step": 571 }, { "epoch": 0.01, "learning_rate": 9.392446633825945e-06, "loss": 2.8432, "step": 572 }, { "epoch": 0.01, "learning_rate": 9.408866995073892e-06, "loss": 2.8598, "step": 573 }, { "epoch": 0.01, "learning_rate": 9.42528735632184e-06, "loss": 2.8274, "step": 574 }, { "epoch": 0.01, "learning_rate": 9.441707717569786e-06, "loss": 2.9046, "step": 575 }, { "epoch": 0.01, "learning_rate": 9.458128078817734e-06, "loss": 2.8006, "step": 576 }, { "epoch": 0.01, "learning_rate": 9.474548440065682e-06, "loss": 2.9253, "step": 577 }, { "epoch": 0.01, "learning_rate": 9.49096880131363e-06, "loss": 2.8963, "step": 578 }, { "epoch": 0.01, "learning_rate": 9.507389162561577e-06, "loss": 2.8064, "step": 579 }, { "epoch": 0.01, "learning_rate": 9.523809523809525e-06, "loss": 2.9247, "step": 580 }, { "epoch": 0.01, "learning_rate": 9.54022988505747e-06, "loss": 2.8559, "step": 581 }, { "epoch": 0.01, "learning_rate": 9.55665024630542e-06, "loss": 2.9894, "step": 582 }, { "epoch": 0.01, "learning_rate": 9.573070607553368e-06, "loss": 3.0434, "step": 583 }, { "epoch": 0.01, "learning_rate": 9.589490968801316e-06, "loss": 2.876, "step": 584 }, { "epoch": 0.01, "learning_rate": 9.605911330049262e-06, "loss": 2.9218, "step": 585 }, { "epoch": 0.01, "learning_rate": 9.62233169129721e-06, "loss": 2.7567, "step": 586 }, { "epoch": 0.01, "learning_rate": 9.638752052545157e-06, "loss": 2.9258, "step": 587 }, { "epoch": 0.01, "learning_rate": 9.655172413793105e-06, "loss": 2.8308, "step": 588 }, { "epoch": 0.01, "learning_rate": 9.671592775041053e-06, "loss": 2.7621, "step": 589 }, { "epoch": 0.01, "learning_rate": 9.688013136288999e-06, "loss": 2.7671, "step": 590 }, { "epoch": 0.01, "learning_rate": 9.704433497536947e-06, "loss": 2.9633, "step": 591 }, { "epoch": 0.01, "learning_rate": 9.720853858784894e-06, "loss": 2.8357, "step": 592 }, { "epoch": 0.01, "learning_rate": 9.737274220032842e-06, "loss": 2.7225, "step": 593 }, { "epoch": 0.01, "learning_rate": 9.75369458128079e-06, "loss": 2.8289, "step": 594 }, { "epoch": 0.01, "learning_rate": 9.770114942528738e-06, "loss": 3.071, "step": 595 }, { "epoch": 0.01, "learning_rate": 9.786535303776684e-06, "loss": 2.8532, "step": 596 }, { "epoch": 0.01, "learning_rate": 9.802955665024631e-06, "loss": 2.794, "step": 597 }, { "epoch": 0.01, "learning_rate": 9.819376026272579e-06, "loss": 2.8561, "step": 598 }, { "epoch": 0.01, "learning_rate": 9.835796387520527e-06, "loss": 2.8629, "step": 599 }, { "epoch": 0.01, "learning_rate": 9.852216748768475e-06, "loss": 2.8307, "step": 600 }, { "epoch": 0.01, "learning_rate": 9.86863711001642e-06, "loss": 2.862, "step": 601 }, { "epoch": 0.01, "learning_rate": 9.885057471264368e-06, "loss": 2.9639, "step": 602 }, { "epoch": 0.01, "learning_rate": 9.901477832512316e-06, "loss": 2.9969, "step": 603 }, { "epoch": 0.01, "learning_rate": 9.917898193760264e-06, "loss": 2.7601, "step": 604 }, { "epoch": 0.01, "learning_rate": 9.934318555008212e-06, "loss": 2.9226, "step": 605 }, { "epoch": 0.01, "learning_rate": 9.95073891625616e-06, "loss": 2.8822, "step": 606 }, { "epoch": 0.01, "learning_rate": 9.967159277504105e-06, "loss": 2.8314, "step": 607 }, { "epoch": 0.01, "learning_rate": 9.983579638752053e-06, "loss": 2.7884, "step": 608 }, { "epoch": 0.02, "learning_rate": 1e-05, "loss": 2.8921, "step": 609 }, { "epoch": 0.02, "learning_rate": 1.0016420361247949e-05, "loss": 3.0234, "step": 610 }, { "epoch": 0.02, "learning_rate": 1.0032840722495896e-05, "loss": 2.8869, "step": 611 }, { "epoch": 0.02, "learning_rate": 1.0049261083743844e-05, "loss": 2.9116, "step": 612 }, { "epoch": 0.02, "learning_rate": 1.006568144499179e-05, "loss": 2.947, "step": 613 }, { "epoch": 0.02, "learning_rate": 1.008210180623974e-05, "loss": 2.8598, "step": 614 }, { "epoch": 0.02, "learning_rate": 1.0098522167487686e-05, "loss": 2.9305, "step": 615 }, { "epoch": 0.02, "learning_rate": 1.0114942528735633e-05, "loss": 2.9539, "step": 616 }, { "epoch": 0.02, "learning_rate": 1.0131362889983581e-05, "loss": 2.904, "step": 617 }, { "epoch": 0.02, "learning_rate": 1.0147783251231529e-05, "loss": 2.9523, "step": 618 }, { "epoch": 0.02, "learning_rate": 1.0164203612479475e-05, "loss": 2.7849, "step": 619 }, { "epoch": 0.02, "learning_rate": 1.0180623973727424e-05, "loss": 2.8304, "step": 620 }, { "epoch": 0.02, "learning_rate": 1.019704433497537e-05, "loss": 2.7671, "step": 621 }, { "epoch": 0.02, "learning_rate": 1.0213464696223318e-05, "loss": 2.7975, "step": 622 }, { "epoch": 0.02, "learning_rate": 1.0229885057471264e-05, "loss": 2.9554, "step": 623 }, { "epoch": 0.02, "learning_rate": 1.0246305418719214e-05, "loss": 2.8097, "step": 624 }, { "epoch": 0.02, "learning_rate": 1.026272577996716e-05, "loss": 3.1845, "step": 625 }, { "epoch": 0.02, "learning_rate": 1.0279146141215109e-05, "loss": 2.8219, "step": 626 }, { "epoch": 0.02, "learning_rate": 1.0295566502463055e-05, "loss": 2.9517, "step": 627 }, { "epoch": 0.02, "learning_rate": 1.0311986863711003e-05, "loss": 2.9416, "step": 628 }, { "epoch": 0.02, "learning_rate": 1.0328407224958949e-05, "loss": 3.1006, "step": 629 }, { "epoch": 0.02, "learning_rate": 1.0344827586206898e-05, "loss": 2.9226, "step": 630 }, { "epoch": 0.02, "learning_rate": 1.0361247947454844e-05, "loss": 2.9891, "step": 631 }, { "epoch": 0.02, "learning_rate": 1.0377668308702792e-05, "loss": 2.7515, "step": 632 }, { "epoch": 0.02, "learning_rate": 1.039408866995074e-05, "loss": 2.9259, "step": 633 }, { "epoch": 0.02, "learning_rate": 1.0410509031198688e-05, "loss": 2.7347, "step": 634 }, { "epoch": 0.02, "learning_rate": 1.0426929392446634e-05, "loss": 2.8992, "step": 635 }, { "epoch": 0.02, "learning_rate": 1.0443349753694583e-05, "loss": 2.7215, "step": 636 }, { "epoch": 0.02, "learning_rate": 1.0459770114942529e-05, "loss": 2.7915, "step": 637 }, { "epoch": 0.02, "learning_rate": 1.0476190476190477e-05, "loss": 2.7345, "step": 638 }, { "epoch": 0.02, "learning_rate": 1.0492610837438425e-05, "loss": 2.9734, "step": 639 }, { "epoch": 0.02, "learning_rate": 1.0509031198686372e-05, "loss": 2.8804, "step": 640 }, { "epoch": 0.02, "learning_rate": 1.0525451559934318e-05, "loss": 2.6979, "step": 641 }, { "epoch": 0.02, "learning_rate": 1.0541871921182268e-05, "loss": 2.9037, "step": 642 }, { "epoch": 0.02, "learning_rate": 1.0558292282430214e-05, "loss": 2.8751, "step": 643 }, { "epoch": 0.02, "learning_rate": 1.0574712643678162e-05, "loss": 3.106, "step": 644 }, { "epoch": 0.02, "learning_rate": 1.0591133004926108e-05, "loss": 2.947, "step": 645 }, { "epoch": 0.02, "learning_rate": 1.0607553366174057e-05, "loss": 2.7023, "step": 646 }, { "epoch": 0.02, "learning_rate": 1.0623973727422003e-05, "loss": 2.8747, "step": 647 }, { "epoch": 0.02, "learning_rate": 1.0640394088669953e-05, "loss": 2.7913, "step": 648 }, { "epoch": 0.02, "learning_rate": 1.0656814449917899e-05, "loss": 2.7605, "step": 649 }, { "epoch": 0.02, "learning_rate": 1.0673234811165846e-05, "loss": 2.9538, "step": 650 }, { "epoch": 0.02, "learning_rate": 1.0689655172413792e-05, "loss": 2.9748, "step": 651 }, { "epoch": 0.02, "learning_rate": 1.0706075533661742e-05, "loss": 2.9252, "step": 652 }, { "epoch": 0.02, "learning_rate": 1.0722495894909688e-05, "loss": 2.8891, "step": 653 }, { "epoch": 0.02, "learning_rate": 1.0738916256157637e-05, "loss": 2.8432, "step": 654 }, { "epoch": 0.02, "learning_rate": 1.0755336617405583e-05, "loss": 2.8643, "step": 655 }, { "epoch": 0.02, "learning_rate": 1.0771756978653531e-05, "loss": 2.831, "step": 656 }, { "epoch": 0.02, "learning_rate": 1.0788177339901477e-05, "loss": 2.7062, "step": 657 }, { "epoch": 0.02, "learning_rate": 1.0804597701149427e-05, "loss": 2.7621, "step": 658 }, { "epoch": 0.02, "learning_rate": 1.0821018062397373e-05, "loss": 2.7257, "step": 659 }, { "epoch": 0.02, "learning_rate": 1.083743842364532e-05, "loss": 2.7112, "step": 660 }, { "epoch": 0.02, "learning_rate": 1.0853858784893268e-05, "loss": 2.9596, "step": 661 }, { "epoch": 0.02, "learning_rate": 1.0870279146141216e-05, "loss": 2.7735, "step": 662 }, { "epoch": 0.02, "learning_rate": 1.0886699507389162e-05, "loss": 2.7134, "step": 663 }, { "epoch": 0.02, "learning_rate": 1.0903119868637111e-05, "loss": 2.9306, "step": 664 }, { "epoch": 0.02, "learning_rate": 1.091954022988506e-05, "loss": 2.7009, "step": 665 }, { "epoch": 0.02, "learning_rate": 1.0935960591133005e-05, "loss": 2.6943, "step": 666 }, { "epoch": 0.02, "learning_rate": 1.0952380952380955e-05, "loss": 2.9482, "step": 667 }, { "epoch": 0.02, "learning_rate": 1.09688013136289e-05, "loss": 2.7401, "step": 668 }, { "epoch": 0.02, "learning_rate": 1.0985221674876848e-05, "loss": 2.8896, "step": 669 }, { "epoch": 0.02, "learning_rate": 1.1001642036124796e-05, "loss": 2.7903, "step": 670 }, { "epoch": 0.02, "learning_rate": 1.1018062397372744e-05, "loss": 2.8666, "step": 671 }, { "epoch": 0.02, "learning_rate": 1.103448275862069e-05, "loss": 2.867, "step": 672 }, { "epoch": 0.02, "learning_rate": 1.105090311986864e-05, "loss": 2.7914, "step": 673 }, { "epoch": 0.02, "learning_rate": 1.1067323481116585e-05, "loss": 2.811, "step": 674 }, { "epoch": 0.02, "learning_rate": 1.1083743842364533e-05, "loss": 2.6179, "step": 675 }, { "epoch": 0.02, "learning_rate": 1.1100164203612481e-05, "loss": 2.9184, "step": 676 }, { "epoch": 0.02, "learning_rate": 1.1116584564860429e-05, "loss": 2.7353, "step": 677 }, { "epoch": 0.02, "learning_rate": 1.1133004926108375e-05, "loss": 2.8947, "step": 678 }, { "epoch": 0.02, "learning_rate": 1.1149425287356324e-05, "loss": 2.8175, "step": 679 }, { "epoch": 0.02, "learning_rate": 1.116584564860427e-05, "loss": 2.797, "step": 680 }, { "epoch": 0.02, "learning_rate": 1.1182266009852218e-05, "loss": 2.9274, "step": 681 }, { "epoch": 0.02, "learning_rate": 1.1198686371100164e-05, "loss": 2.7782, "step": 682 }, { "epoch": 0.02, "learning_rate": 1.1215106732348113e-05, "loss": 3.0161, "step": 683 }, { "epoch": 0.02, "learning_rate": 1.123152709359606e-05, "loss": 2.6967, "step": 684 }, { "epoch": 0.02, "learning_rate": 1.1247947454844009e-05, "loss": 2.9369, "step": 685 }, { "epoch": 0.02, "learning_rate": 1.1264367816091955e-05, "loss": 2.8865, "step": 686 }, { "epoch": 0.02, "learning_rate": 1.1280788177339903e-05, "loss": 2.8555, "step": 687 }, { "epoch": 0.02, "learning_rate": 1.1297208538587849e-05, "loss": 3.0064, "step": 688 }, { "epoch": 0.02, "learning_rate": 1.1313628899835798e-05, "loss": 2.7844, "step": 689 }, { "epoch": 0.02, "learning_rate": 1.1330049261083744e-05, "loss": 2.7102, "step": 690 }, { "epoch": 0.02, "learning_rate": 1.1346469622331692e-05, "loss": 2.9729, "step": 691 }, { "epoch": 0.02, "learning_rate": 1.136288998357964e-05, "loss": 2.8341, "step": 692 }, { "epoch": 0.02, "learning_rate": 1.1379310344827587e-05, "loss": 2.7548, "step": 693 }, { "epoch": 0.02, "learning_rate": 1.1395730706075534e-05, "loss": 2.9072, "step": 694 }, { "epoch": 0.02, "learning_rate": 1.1412151067323483e-05, "loss": 2.879, "step": 695 }, { "epoch": 0.02, "learning_rate": 1.1428571428571429e-05, "loss": 2.7804, "step": 696 }, { "epoch": 0.02, "learning_rate": 1.1444991789819377e-05, "loss": 2.9697, "step": 697 }, { "epoch": 0.02, "learning_rate": 1.1461412151067325e-05, "loss": 2.885, "step": 698 }, { "epoch": 0.02, "learning_rate": 1.1477832512315272e-05, "loss": 2.8528, "step": 699 }, { "epoch": 0.02, "learning_rate": 1.1494252873563218e-05, "loss": 2.7255, "step": 700 }, { "epoch": 0.02, "learning_rate": 1.1510673234811168e-05, "loss": 2.8015, "step": 701 }, { "epoch": 0.02, "learning_rate": 1.1527093596059114e-05, "loss": 3.0244, "step": 702 }, { "epoch": 0.02, "learning_rate": 1.1543513957307062e-05, "loss": 2.7702, "step": 703 }, { "epoch": 0.02, "learning_rate": 1.1559934318555008e-05, "loss": 2.7424, "step": 704 }, { "epoch": 0.02, "learning_rate": 1.1576354679802957e-05, "loss": 2.9578, "step": 705 }, { "epoch": 0.02, "learning_rate": 1.1592775041050903e-05, "loss": 2.6184, "step": 706 }, { "epoch": 0.02, "learning_rate": 1.1609195402298852e-05, "loss": 2.8182, "step": 707 }, { "epoch": 0.02, "learning_rate": 1.1625615763546799e-05, "loss": 2.7153, "step": 708 }, { "epoch": 0.02, "learning_rate": 1.1642036124794746e-05, "loss": 2.6587, "step": 709 }, { "epoch": 0.02, "learning_rate": 1.1658456486042692e-05, "loss": 2.84, "step": 710 }, { "epoch": 0.02, "learning_rate": 1.1674876847290642e-05, "loss": 2.847, "step": 711 }, { "epoch": 0.02, "learning_rate": 1.1691297208538588e-05, "loss": 2.7114, "step": 712 }, { "epoch": 0.02, "learning_rate": 1.1707717569786536e-05, "loss": 2.7807, "step": 713 }, { "epoch": 0.02, "learning_rate": 1.1724137931034483e-05, "loss": 2.7065, "step": 714 }, { "epoch": 0.02, "learning_rate": 1.1740558292282431e-05, "loss": 2.9328, "step": 715 }, { "epoch": 0.02, "learning_rate": 1.1756978653530377e-05, "loss": 2.7084, "step": 716 }, { "epoch": 0.02, "learning_rate": 1.1773399014778327e-05, "loss": 2.7227, "step": 717 }, { "epoch": 0.02, "learning_rate": 1.1789819376026273e-05, "loss": 2.6793, "step": 718 }, { "epoch": 0.02, "learning_rate": 1.180623973727422e-05, "loss": 2.772, "step": 719 }, { "epoch": 0.02, "learning_rate": 1.182266009852217e-05, "loss": 2.7017, "step": 720 }, { "epoch": 0.02, "learning_rate": 1.1839080459770116e-05, "loss": 2.6086, "step": 721 }, { "epoch": 0.02, "learning_rate": 1.1855500821018065e-05, "loss": 2.6564, "step": 722 }, { "epoch": 0.02, "learning_rate": 1.1871921182266011e-05, "loss": 2.8501, "step": 723 }, { "epoch": 0.02, "learning_rate": 1.1888341543513959e-05, "loss": 2.8569, "step": 724 }, { "epoch": 0.02, "learning_rate": 1.1904761904761905e-05, "loss": 2.7639, "step": 725 }, { "epoch": 0.02, "learning_rate": 1.1921182266009855e-05, "loss": 2.8491, "step": 726 }, { "epoch": 0.02, "learning_rate": 1.19376026272578e-05, "loss": 2.8427, "step": 727 }, { "epoch": 0.02, "learning_rate": 1.1954022988505748e-05, "loss": 2.6894, "step": 728 }, { "epoch": 0.02, "learning_rate": 1.1970443349753696e-05, "loss": 2.551, "step": 729 }, { "epoch": 0.02, "learning_rate": 1.1986863711001644e-05, "loss": 2.8514, "step": 730 }, { "epoch": 0.02, "learning_rate": 1.200328407224959e-05, "loss": 2.7949, "step": 731 }, { "epoch": 0.02, "learning_rate": 1.201970443349754e-05, "loss": 2.9083, "step": 732 }, { "epoch": 0.02, "learning_rate": 1.2036124794745485e-05, "loss": 2.6063, "step": 733 }, { "epoch": 0.02, "learning_rate": 1.2052545155993433e-05, "loss": 2.5265, "step": 734 }, { "epoch": 0.02, "learning_rate": 1.206896551724138e-05, "loss": 2.7834, "step": 735 }, { "epoch": 0.02, "learning_rate": 1.2085385878489329e-05, "loss": 2.7218, "step": 736 }, { "epoch": 0.02, "learning_rate": 1.2101806239737275e-05, "loss": 2.6014, "step": 737 }, { "epoch": 0.02, "learning_rate": 1.2118226600985224e-05, "loss": 2.9296, "step": 738 }, { "epoch": 0.02, "learning_rate": 1.213464696223317e-05, "loss": 2.9574, "step": 739 }, { "epoch": 0.02, "learning_rate": 1.2151067323481118e-05, "loss": 2.6756, "step": 740 }, { "epoch": 0.02, "learning_rate": 1.2167487684729064e-05, "loss": 2.6837, "step": 741 }, { "epoch": 0.02, "learning_rate": 1.2183908045977013e-05, "loss": 2.8348, "step": 742 }, { "epoch": 0.02, "learning_rate": 1.220032840722496e-05, "loss": 2.9813, "step": 743 }, { "epoch": 0.02, "learning_rate": 1.2216748768472909e-05, "loss": 2.7209, "step": 744 }, { "epoch": 0.02, "learning_rate": 1.2233169129720855e-05, "loss": 2.5951, "step": 745 }, { "epoch": 0.02, "learning_rate": 1.2249589490968803e-05, "loss": 2.599, "step": 746 }, { "epoch": 0.02, "learning_rate": 1.2266009852216749e-05, "loss": 2.8543, "step": 747 }, { "epoch": 0.02, "learning_rate": 1.2282430213464698e-05, "loss": 2.6909, "step": 748 }, { "epoch": 0.02, "learning_rate": 1.2298850574712644e-05, "loss": 2.793, "step": 749 }, { "epoch": 0.02, "learning_rate": 1.2315270935960592e-05, "loss": 2.8621, "step": 750 }, { "epoch": 0.02, "learning_rate": 1.233169129720854e-05, "loss": 2.7897, "step": 751 }, { "epoch": 0.02, "learning_rate": 1.2348111658456487e-05, "loss": 2.6769, "step": 752 }, { "epoch": 0.02, "learning_rate": 1.2364532019704433e-05, "loss": 2.8284, "step": 753 }, { "epoch": 0.02, "learning_rate": 1.2380952380952383e-05, "loss": 2.8075, "step": 754 }, { "epoch": 0.02, "learning_rate": 1.2397372742200329e-05, "loss": 2.8362, "step": 755 }, { "epoch": 0.02, "learning_rate": 1.2413793103448277e-05, "loss": 2.8461, "step": 756 }, { "epoch": 0.02, "learning_rate": 1.2430213464696224e-05, "loss": 2.8585, "step": 757 }, { "epoch": 0.02, "learning_rate": 1.2446633825944172e-05, "loss": 2.9363, "step": 758 }, { "epoch": 0.02, "learning_rate": 1.2463054187192118e-05, "loss": 2.8026, "step": 759 }, { "epoch": 0.02, "learning_rate": 1.2479474548440068e-05, "loss": 2.7491, "step": 760 }, { "epoch": 0.02, "learning_rate": 1.2495894909688014e-05, "loss": 2.7793, "step": 761 }, { "epoch": 0.02, "learning_rate": 1.2512315270935961e-05, "loss": 2.9823, "step": 762 }, { "epoch": 0.02, "learning_rate": 1.2528735632183907e-05, "loss": 2.5452, "step": 763 }, { "epoch": 0.02, "learning_rate": 1.2545155993431857e-05, "loss": 2.7227, "step": 764 }, { "epoch": 0.02, "learning_rate": 1.2561576354679803e-05, "loss": 2.7004, "step": 765 }, { "epoch": 0.02, "learning_rate": 1.2577996715927752e-05, "loss": 2.8626, "step": 766 }, { "epoch": 0.02, "learning_rate": 1.2594417077175698e-05, "loss": 2.6387, "step": 767 }, { "epoch": 0.02, "learning_rate": 1.2610837438423646e-05, "loss": 2.7516, "step": 768 }, { "epoch": 0.02, "learning_rate": 1.2627257799671592e-05, "loss": 2.8094, "step": 769 }, { "epoch": 0.02, "learning_rate": 1.2643678160919542e-05, "loss": 2.9126, "step": 770 }, { "epoch": 0.02, "learning_rate": 1.2660098522167488e-05, "loss": 2.7349, "step": 771 }, { "epoch": 0.02, "learning_rate": 1.2676518883415435e-05, "loss": 2.6979, "step": 772 }, { "epoch": 0.02, "learning_rate": 1.2692939244663383e-05, "loss": 2.6731, "step": 773 }, { "epoch": 0.02, "learning_rate": 1.2709359605911331e-05, "loss": 2.7304, "step": 774 }, { "epoch": 0.02, "learning_rate": 1.2725779967159277e-05, "loss": 2.8046, "step": 775 }, { "epoch": 0.02, "learning_rate": 1.2742200328407226e-05, "loss": 2.9729, "step": 776 }, { "epoch": 0.02, "learning_rate": 1.2758620689655174e-05, "loss": 2.6846, "step": 777 }, { "epoch": 0.02, "learning_rate": 1.277504105090312e-05, "loss": 2.8199, "step": 778 }, { "epoch": 0.02, "learning_rate": 1.279146141215107e-05, "loss": 2.8411, "step": 779 }, { "epoch": 0.02, "learning_rate": 1.2807881773399016e-05, "loss": 2.8773, "step": 780 }, { "epoch": 0.02, "learning_rate": 1.2824302134646963e-05, "loss": 2.5222, "step": 781 }, { "epoch": 0.02, "learning_rate": 1.2840722495894911e-05, "loss": 2.7862, "step": 782 }, { "epoch": 0.02, "learning_rate": 1.2857142857142859e-05, "loss": 2.7487, "step": 783 }, { "epoch": 0.02, "learning_rate": 1.2873563218390805e-05, "loss": 2.6665, "step": 784 }, { "epoch": 0.02, "learning_rate": 1.2889983579638754e-05, "loss": 2.6066, "step": 785 }, { "epoch": 0.02, "learning_rate": 1.29064039408867e-05, "loss": 2.8017, "step": 786 }, { "epoch": 0.02, "learning_rate": 1.2922824302134648e-05, "loss": 2.753, "step": 787 }, { "epoch": 0.02, "learning_rate": 1.2939244663382596e-05, "loss": 2.8276, "step": 788 }, { "epoch": 0.02, "learning_rate": 1.2955665024630544e-05, "loss": 2.8662, "step": 789 }, { "epoch": 0.02, "learning_rate": 1.297208538587849e-05, "loss": 2.8011, "step": 790 }, { "epoch": 0.02, "learning_rate": 1.298850574712644e-05, "loss": 2.6928, "step": 791 }, { "epoch": 0.02, "learning_rate": 1.3004926108374385e-05, "loss": 2.7844, "step": 792 }, { "epoch": 0.02, "learning_rate": 1.3021346469622333e-05, "loss": 2.7352, "step": 793 }, { "epoch": 0.02, "learning_rate": 1.3037766830870279e-05, "loss": 2.8469, "step": 794 }, { "epoch": 0.02, "learning_rate": 1.3054187192118228e-05, "loss": 2.8832, "step": 795 }, { "epoch": 0.02, "learning_rate": 1.3070607553366174e-05, "loss": 2.6997, "step": 796 }, { "epoch": 0.02, "learning_rate": 1.3087027914614124e-05, "loss": 2.6352, "step": 797 }, { "epoch": 0.02, "learning_rate": 1.310344827586207e-05, "loss": 2.902, "step": 798 }, { "epoch": 0.02, "learning_rate": 1.3119868637110018e-05, "loss": 2.7077, "step": 799 }, { "epoch": 0.02, "learning_rate": 1.3136288998357964e-05, "loss": 2.702, "step": 800 }, { "epoch": 0.02, "learning_rate": 1.3152709359605913e-05, "loss": 2.9853, "step": 801 }, { "epoch": 0.02, "learning_rate": 1.316912972085386e-05, "loss": 2.6848, "step": 802 }, { "epoch": 0.02, "learning_rate": 1.3185550082101809e-05, "loss": 2.8113, "step": 803 }, { "epoch": 0.02, "learning_rate": 1.3201970443349755e-05, "loss": 2.7626, "step": 804 }, { "epoch": 0.02, "learning_rate": 1.3218390804597702e-05, "loss": 2.7606, "step": 805 }, { "epoch": 0.02, "learning_rate": 1.3234811165845649e-05, "loss": 2.8451, "step": 806 }, { "epoch": 0.02, "learning_rate": 1.3251231527093598e-05, "loss": 2.6814, "step": 807 }, { "epoch": 0.02, "learning_rate": 1.3267651888341544e-05, "loss": 2.7606, "step": 808 }, { "epoch": 0.02, "learning_rate": 1.3284072249589492e-05, "loss": 2.7214, "step": 809 }, { "epoch": 0.02, "learning_rate": 1.330049261083744e-05, "loss": 2.8047, "step": 810 }, { "epoch": 0.02, "learning_rate": 1.3316912972085387e-05, "loss": 2.5982, "step": 811 }, { "epoch": 0.02, "learning_rate": 1.3333333333333333e-05, "loss": 2.6631, "step": 812 }, { "epoch": 0.02, "learning_rate": 1.3349753694581283e-05, "loss": 2.6288, "step": 813 }, { "epoch": 0.02, "learning_rate": 1.3366174055829229e-05, "loss": 2.7623, "step": 814 }, { "epoch": 0.02, "learning_rate": 1.3382594417077177e-05, "loss": 2.6713, "step": 815 }, { "epoch": 0.02, "learning_rate": 1.3399014778325124e-05, "loss": 2.8514, "step": 816 }, { "epoch": 0.02, "learning_rate": 1.3415435139573072e-05, "loss": 2.7942, "step": 817 }, { "epoch": 0.02, "learning_rate": 1.3431855500821018e-05, "loss": 2.682, "step": 818 }, { "epoch": 0.02, "learning_rate": 1.3448275862068967e-05, "loss": 2.7128, "step": 819 }, { "epoch": 0.02, "learning_rate": 1.3464696223316914e-05, "loss": 2.7141, "step": 820 }, { "epoch": 0.02, "learning_rate": 1.3481116584564861e-05, "loss": 2.8461, "step": 821 }, { "epoch": 0.02, "learning_rate": 1.3497536945812807e-05, "loss": 2.8318, "step": 822 }, { "epoch": 0.02, "learning_rate": 1.3513957307060757e-05, "loss": 2.7876, "step": 823 }, { "epoch": 0.02, "learning_rate": 1.3530377668308703e-05, "loss": 2.73, "step": 824 }, { "epoch": 0.02, "learning_rate": 1.3546798029556652e-05, "loss": 2.744, "step": 825 }, { "epoch": 0.02, "learning_rate": 1.3563218390804598e-05, "loss": 2.8226, "step": 826 }, { "epoch": 0.02, "learning_rate": 1.3579638752052546e-05, "loss": 2.772, "step": 827 }, { "epoch": 0.02, "learning_rate": 1.3596059113300492e-05, "loss": 2.5341, "step": 828 }, { "epoch": 0.02, "learning_rate": 1.3612479474548442e-05, "loss": 2.6286, "step": 829 }, { "epoch": 0.02, "learning_rate": 1.3628899835796388e-05, "loss": 2.6861, "step": 830 }, { "epoch": 0.02, "learning_rate": 1.3645320197044335e-05, "loss": 2.6292, "step": 831 }, { "epoch": 0.02, "learning_rate": 1.3661740558292285e-05, "loss": 2.7622, "step": 832 }, { "epoch": 0.02, "learning_rate": 1.367816091954023e-05, "loss": 2.8582, "step": 833 }, { "epoch": 0.02, "learning_rate": 1.369458128078818e-05, "loss": 2.6358, "step": 834 }, { "epoch": 0.02, "learning_rate": 1.3711001642036126e-05, "loss": 2.8596, "step": 835 }, { "epoch": 0.02, "learning_rate": 1.3727422003284074e-05, "loss": 2.9069, "step": 836 }, { "epoch": 0.02, "learning_rate": 1.374384236453202e-05, "loss": 2.8612, "step": 837 }, { "epoch": 0.02, "learning_rate": 1.376026272577997e-05, "loss": 2.6703, "step": 838 }, { "epoch": 0.02, "learning_rate": 1.3776683087027916e-05, "loss": 2.6982, "step": 839 }, { "epoch": 0.02, "learning_rate": 1.3793103448275863e-05, "loss": 2.8489, "step": 840 }, { "epoch": 0.02, "learning_rate": 1.3809523809523811e-05, "loss": 2.7558, "step": 841 }, { "epoch": 0.02, "learning_rate": 1.3825944170771759e-05, "loss": 2.7095, "step": 842 }, { "epoch": 0.02, "learning_rate": 1.3842364532019705e-05, "loss": 2.8971, "step": 843 }, { "epoch": 0.02, "learning_rate": 1.3858784893267654e-05, "loss": 2.5864, "step": 844 }, { "epoch": 0.02, "learning_rate": 1.38752052545156e-05, "loss": 2.8406, "step": 845 }, { "epoch": 0.02, "learning_rate": 1.3891625615763548e-05, "loss": 2.8482, "step": 846 }, { "epoch": 0.02, "learning_rate": 1.3908045977011496e-05, "loss": 2.6917, "step": 847 }, { "epoch": 0.02, "learning_rate": 1.3924466338259444e-05, "loss": 2.5909, "step": 848 }, { "epoch": 0.02, "learning_rate": 1.394088669950739e-05, "loss": 2.7957, "step": 849 }, { "epoch": 0.02, "learning_rate": 1.3957307060755339e-05, "loss": 2.7344, "step": 850 }, { "epoch": 0.02, "learning_rate": 1.3973727422003285e-05, "loss": 2.7757, "step": 851 }, { "epoch": 0.02, "learning_rate": 1.3990147783251233e-05, "loss": 2.7714, "step": 852 }, { "epoch": 0.02, "learning_rate": 1.4006568144499179e-05, "loss": 2.6501, "step": 853 }, { "epoch": 0.02, "learning_rate": 1.4022988505747128e-05, "loss": 2.782, "step": 854 }, { "epoch": 0.02, "learning_rate": 1.4039408866995074e-05, "loss": 2.634, "step": 855 }, { "epoch": 0.02, "learning_rate": 1.4055829228243024e-05, "loss": 2.7888, "step": 856 }, { "epoch": 0.02, "learning_rate": 1.407224958949097e-05, "loss": 2.6579, "step": 857 }, { "epoch": 0.02, "learning_rate": 1.4088669950738918e-05, "loss": 2.8083, "step": 858 }, { "epoch": 0.02, "learning_rate": 1.4105090311986864e-05, "loss": 2.7141, "step": 859 }, { "epoch": 0.02, "learning_rate": 1.4121510673234813e-05, "loss": 2.6732, "step": 860 }, { "epoch": 0.02, "learning_rate": 1.4137931034482759e-05, "loss": 2.7307, "step": 861 }, { "epoch": 0.02, "learning_rate": 1.4154351395730707e-05, "loss": 2.6981, "step": 862 }, { "epoch": 0.02, "learning_rate": 1.4170771756978655e-05, "loss": 2.6998, "step": 863 }, { "epoch": 0.02, "learning_rate": 1.4187192118226602e-05, "loss": 2.8481, "step": 864 }, { "epoch": 0.02, "learning_rate": 1.4203612479474548e-05, "loss": 2.8585, "step": 865 }, { "epoch": 0.02, "learning_rate": 1.4220032840722498e-05, "loss": 2.852, "step": 866 }, { "epoch": 0.02, "learning_rate": 1.4236453201970444e-05, "loss": 2.7153, "step": 867 }, { "epoch": 0.02, "learning_rate": 1.4252873563218392e-05, "loss": 2.7001, "step": 868 }, { "epoch": 0.02, "learning_rate": 1.426929392446634e-05, "loss": 2.8441, "step": 869 }, { "epoch": 0.02, "learning_rate": 1.4285714285714287e-05, "loss": 2.8834, "step": 870 }, { "epoch": 0.02, "learning_rate": 1.4302134646962233e-05, "loss": 2.8205, "step": 871 }, { "epoch": 0.02, "learning_rate": 1.4318555008210183e-05, "loss": 2.7187, "step": 872 }, { "epoch": 0.02, "learning_rate": 1.4334975369458129e-05, "loss": 2.7195, "step": 873 }, { "epoch": 0.02, "learning_rate": 1.4351395730706076e-05, "loss": 2.7008, "step": 874 }, { "epoch": 0.02, "learning_rate": 1.4367816091954022e-05, "loss": 2.6969, "step": 875 }, { "epoch": 0.02, "learning_rate": 1.4384236453201972e-05, "loss": 2.6902, "step": 876 }, { "epoch": 0.02, "learning_rate": 1.4400656814449918e-05, "loss": 2.7135, "step": 877 }, { "epoch": 0.02, "learning_rate": 1.4417077175697867e-05, "loss": 2.8863, "step": 878 }, { "epoch": 0.02, "learning_rate": 1.4433497536945813e-05, "loss": 2.5947, "step": 879 }, { "epoch": 0.02, "learning_rate": 1.4449917898193761e-05, "loss": 2.7898, "step": 880 }, { "epoch": 0.02, "learning_rate": 1.4466338259441707e-05, "loss": 2.7492, "step": 881 }, { "epoch": 0.02, "learning_rate": 1.4482758620689657e-05, "loss": 2.7531, "step": 882 }, { "epoch": 0.02, "learning_rate": 1.4499178981937603e-05, "loss": 2.7766, "step": 883 }, { "epoch": 0.02, "learning_rate": 1.4515599343185552e-05, "loss": 2.6416, "step": 884 }, { "epoch": 0.02, "learning_rate": 1.4532019704433498e-05, "loss": 2.6613, "step": 885 }, { "epoch": 0.02, "learning_rate": 1.4548440065681446e-05, "loss": 2.7033, "step": 886 }, { "epoch": 0.02, "learning_rate": 1.4564860426929395e-05, "loss": 2.8505, "step": 887 }, { "epoch": 0.02, "learning_rate": 1.4581280788177341e-05, "loss": 2.7446, "step": 888 }, { "epoch": 0.02, "learning_rate": 1.459770114942529e-05, "loss": 2.6237, "step": 889 }, { "epoch": 0.02, "learning_rate": 1.4614121510673235e-05, "loss": 2.5605, "step": 890 }, { "epoch": 0.02, "learning_rate": 1.4630541871921185e-05, "loss": 2.6905, "step": 891 }, { "epoch": 0.02, "learning_rate": 1.464696223316913e-05, "loss": 2.6751, "step": 892 }, { "epoch": 0.02, "learning_rate": 1.466338259441708e-05, "loss": 2.8001, "step": 893 }, { "epoch": 0.02, "learning_rate": 1.4679802955665026e-05, "loss": 2.9338, "step": 894 }, { "epoch": 0.02, "learning_rate": 1.4696223316912974e-05, "loss": 2.8525, "step": 895 }, { "epoch": 0.02, "learning_rate": 1.471264367816092e-05, "loss": 2.7823, "step": 896 }, { "epoch": 0.02, "learning_rate": 1.472906403940887e-05, "loss": 2.7653, "step": 897 }, { "epoch": 0.02, "learning_rate": 1.4745484400656815e-05, "loss": 2.7855, "step": 898 }, { "epoch": 0.02, "learning_rate": 1.4761904761904763e-05, "loss": 2.8734, "step": 899 }, { "epoch": 0.02, "learning_rate": 1.4778325123152711e-05, "loss": 2.8628, "step": 900 }, { "epoch": 0.02, "learning_rate": 1.4794745484400659e-05, "loss": 2.5742, "step": 901 }, { "epoch": 0.02, "learning_rate": 1.4811165845648605e-05, "loss": 2.6983, "step": 902 }, { "epoch": 0.02, "learning_rate": 1.4827586206896554e-05, "loss": 2.6605, "step": 903 }, { "epoch": 0.02, "learning_rate": 1.48440065681445e-05, "loss": 2.628, "step": 904 }, { "epoch": 0.02, "learning_rate": 1.4860426929392448e-05, "loss": 2.6483, "step": 905 }, { "epoch": 0.02, "learning_rate": 1.4876847290640396e-05, "loss": 2.5714, "step": 906 }, { "epoch": 0.02, "learning_rate": 1.4893267651888343e-05, "loss": 2.8266, "step": 907 }, { "epoch": 0.02, "learning_rate": 1.490968801313629e-05, "loss": 2.613, "step": 908 }, { "epoch": 0.02, "learning_rate": 1.4926108374384239e-05, "loss": 2.7683, "step": 909 }, { "epoch": 0.02, "learning_rate": 1.4942528735632185e-05, "loss": 2.8371, "step": 910 }, { "epoch": 0.02, "learning_rate": 1.4958949096880133e-05, "loss": 2.6685, "step": 911 }, { "epoch": 0.02, "learning_rate": 1.4975369458128079e-05, "loss": 2.5473, "step": 912 }, { "epoch": 0.02, "learning_rate": 1.4991789819376028e-05, "loss": 2.5588, "step": 913 }, { "epoch": 0.02, "learning_rate": 1.5008210180623974e-05, "loss": 2.6917, "step": 914 }, { "epoch": 0.02, "learning_rate": 1.5024630541871924e-05, "loss": 2.7494, "step": 915 }, { "epoch": 0.02, "learning_rate": 1.504105090311987e-05, "loss": 2.6785, "step": 916 }, { "epoch": 0.02, "learning_rate": 1.5057471264367817e-05, "loss": 2.647, "step": 917 }, { "epoch": 0.02, "learning_rate": 1.5073891625615764e-05, "loss": 2.8027, "step": 918 }, { "epoch": 0.02, "learning_rate": 1.5090311986863713e-05, "loss": 2.7284, "step": 919 }, { "epoch": 0.02, "learning_rate": 1.5106732348111659e-05, "loss": 2.6311, "step": 920 }, { "epoch": 0.02, "learning_rate": 1.5123152709359607e-05, "loss": 2.6963, "step": 921 }, { "epoch": 0.02, "learning_rate": 1.5139573070607554e-05, "loss": 2.8173, "step": 922 }, { "epoch": 0.02, "learning_rate": 1.5155993431855502e-05, "loss": 2.7345, "step": 923 }, { "epoch": 0.02, "learning_rate": 1.5172413793103448e-05, "loss": 2.4926, "step": 924 }, { "epoch": 0.02, "learning_rate": 1.5188834154351398e-05, "loss": 2.6502, "step": 925 }, { "epoch": 0.02, "learning_rate": 1.5205254515599344e-05, "loss": 2.7585, "step": 926 }, { "epoch": 0.02, "learning_rate": 1.5221674876847292e-05, "loss": 2.4895, "step": 927 }, { "epoch": 0.02, "learning_rate": 1.523809523809524e-05, "loss": 2.7277, "step": 928 }, { "epoch": 0.02, "learning_rate": 1.5254515599343187e-05, "loss": 2.588, "step": 929 }, { "epoch": 0.02, "learning_rate": 1.5270935960591133e-05, "loss": 2.8173, "step": 930 }, { "epoch": 0.02, "learning_rate": 1.528735632183908e-05, "loss": 2.8703, "step": 931 }, { "epoch": 0.02, "learning_rate": 1.530377668308703e-05, "loss": 2.93, "step": 932 }, { "epoch": 0.02, "learning_rate": 1.5320197044334976e-05, "loss": 2.6415, "step": 933 }, { "epoch": 0.02, "learning_rate": 1.5336617405582924e-05, "loss": 2.5235, "step": 934 }, { "epoch": 0.02, "learning_rate": 1.5353037766830872e-05, "loss": 2.682, "step": 935 }, { "epoch": 0.02, "learning_rate": 1.536945812807882e-05, "loss": 2.8062, "step": 936 }, { "epoch": 0.02, "learning_rate": 1.5385878489326767e-05, "loss": 2.6328, "step": 937 }, { "epoch": 0.02, "learning_rate": 1.540229885057471e-05, "loss": 2.7555, "step": 938 }, { "epoch": 0.02, "learning_rate": 1.5418719211822663e-05, "loss": 2.7135, "step": 939 }, { "epoch": 0.02, "learning_rate": 1.5435139573070607e-05, "loss": 2.584, "step": 940 }, { "epoch": 0.02, "learning_rate": 1.5451559934318555e-05, "loss": 2.792, "step": 941 }, { "epoch": 0.02, "learning_rate": 1.5467980295566506e-05, "loss": 2.7309, "step": 942 }, { "epoch": 0.02, "learning_rate": 1.548440065681445e-05, "loss": 2.7828, "step": 943 }, { "epoch": 0.02, "learning_rate": 1.5500821018062398e-05, "loss": 2.7057, "step": 944 }, { "epoch": 0.02, "learning_rate": 1.5517241379310346e-05, "loss": 2.6336, "step": 945 }, { "epoch": 0.02, "learning_rate": 1.5533661740558294e-05, "loss": 2.7758, "step": 946 }, { "epoch": 0.02, "learning_rate": 1.555008210180624e-05, "loss": 2.7161, "step": 947 }, { "epoch": 0.02, "learning_rate": 1.556650246305419e-05, "loss": 2.7929, "step": 948 }, { "epoch": 0.02, "learning_rate": 1.5582922824302137e-05, "loss": 2.7692, "step": 949 }, { "epoch": 0.02, "learning_rate": 1.5599343185550085e-05, "loss": 2.7945, "step": 950 }, { "epoch": 0.02, "learning_rate": 1.5615763546798032e-05, "loss": 2.764, "step": 951 }, { "epoch": 0.02, "learning_rate": 1.563218390804598e-05, "loss": 2.8246, "step": 952 }, { "epoch": 0.02, "learning_rate": 1.5648604269293924e-05, "loss": 2.7343, "step": 953 }, { "epoch": 0.02, "learning_rate": 1.5665024630541875e-05, "loss": 2.8568, "step": 954 }, { "epoch": 0.02, "learning_rate": 1.568144499178982e-05, "loss": 2.9388, "step": 955 }, { "epoch": 0.02, "learning_rate": 1.5697865353037768e-05, "loss": 2.6727, "step": 956 }, { "epoch": 0.02, "learning_rate": 1.5714285714285715e-05, "loss": 2.8475, "step": 957 }, { "epoch": 0.02, "learning_rate": 1.5730706075533663e-05, "loss": 2.604, "step": 958 }, { "epoch": 0.02, "learning_rate": 1.574712643678161e-05, "loss": 2.6926, "step": 959 }, { "epoch": 0.02, "learning_rate": 1.576354679802956e-05, "loss": 2.6596, "step": 960 }, { "epoch": 0.02, "learning_rate": 1.5779967159277506e-05, "loss": 2.7054, "step": 961 }, { "epoch": 0.02, "learning_rate": 1.5796387520525454e-05, "loss": 2.5024, "step": 962 }, { "epoch": 0.02, "learning_rate": 1.58128078817734e-05, "loss": 2.588, "step": 963 }, { "epoch": 0.02, "learning_rate": 1.582922824302135e-05, "loss": 2.5355, "step": 964 }, { "epoch": 0.02, "learning_rate": 1.5845648604269294e-05, "loss": 2.7151, "step": 965 }, { "epoch": 0.02, "learning_rate": 1.586206896551724e-05, "loss": 2.6551, "step": 966 }, { "epoch": 0.02, "learning_rate": 1.587848932676519e-05, "loss": 2.5449, "step": 967 }, { "epoch": 0.02, "learning_rate": 1.5894909688013137e-05, "loss": 2.6764, "step": 968 }, { "epoch": 0.02, "learning_rate": 1.5911330049261085e-05, "loss": 2.6892, "step": 969 }, { "epoch": 0.02, "learning_rate": 1.5927750410509033e-05, "loss": 2.6786, "step": 970 }, { "epoch": 0.02, "learning_rate": 1.594417077175698e-05, "loss": 2.7565, "step": 971 }, { "epoch": 0.02, "learning_rate": 1.5960591133004928e-05, "loss": 2.6325, "step": 972 }, { "epoch": 0.02, "learning_rate": 1.5977011494252876e-05, "loss": 2.8269, "step": 973 }, { "epoch": 0.02, "learning_rate": 1.5993431855500824e-05, "loss": 2.7231, "step": 974 }, { "epoch": 0.02, "learning_rate": 1.6009852216748768e-05, "loss": 2.7893, "step": 975 }, { "epoch": 0.02, "learning_rate": 1.602627257799672e-05, "loss": 2.6422, "step": 976 }, { "epoch": 0.02, "learning_rate": 1.6042692939244663e-05, "loss": 2.7245, "step": 977 }, { "epoch": 0.02, "learning_rate": 1.605911330049261e-05, "loss": 2.6509, "step": 978 }, { "epoch": 0.02, "learning_rate": 1.607553366174056e-05, "loss": 2.7717, "step": 979 }, { "epoch": 0.02, "learning_rate": 1.6091954022988507e-05, "loss": 2.8229, "step": 980 }, { "epoch": 0.02, "learning_rate": 1.6108374384236454e-05, "loss": 2.8082, "step": 981 }, { "epoch": 0.02, "learning_rate": 1.6124794745484402e-05, "loss": 2.4371, "step": 982 }, { "epoch": 0.02, "learning_rate": 1.614121510673235e-05, "loss": 2.7005, "step": 983 }, { "epoch": 0.02, "learning_rate": 1.6157635467980298e-05, "loss": 2.7363, "step": 984 }, { "epoch": 0.02, "learning_rate": 1.6174055829228242e-05, "loss": 2.6276, "step": 985 }, { "epoch": 0.02, "learning_rate": 1.6190476190476193e-05, "loss": 2.7863, "step": 986 }, { "epoch": 0.02, "learning_rate": 1.6206896551724137e-05, "loss": 2.9224, "step": 987 }, { "epoch": 0.02, "learning_rate": 1.622331691297209e-05, "loss": 2.679, "step": 988 }, { "epoch": 0.02, "learning_rate": 1.6239737274220033e-05, "loss": 2.8262, "step": 989 }, { "epoch": 0.02, "learning_rate": 1.625615763546798e-05, "loss": 2.5974, "step": 990 }, { "epoch": 0.02, "learning_rate": 1.627257799671593e-05, "loss": 2.8959, "step": 991 }, { "epoch": 0.02, "learning_rate": 1.6288998357963876e-05, "loss": 2.7051, "step": 992 }, { "epoch": 0.02, "learning_rate": 1.6305418719211824e-05, "loss": 2.7412, "step": 993 }, { "epoch": 0.02, "learning_rate": 1.632183908045977e-05, "loss": 2.6583, "step": 994 }, { "epoch": 0.02, "learning_rate": 1.633825944170772e-05, "loss": 2.759, "step": 995 }, { "epoch": 0.02, "learning_rate": 1.6354679802955667e-05, "loss": 2.7965, "step": 996 }, { "epoch": 0.02, "learning_rate": 1.6371100164203615e-05, "loss": 2.8181, "step": 997 }, { "epoch": 0.02, "learning_rate": 1.6387520525451563e-05, "loss": 2.6072, "step": 998 }, { "epoch": 0.02, "learning_rate": 1.640394088669951e-05, "loss": 2.6677, "step": 999 }, { "epoch": 0.02, "learning_rate": 1.6420361247947455e-05, "loss": 2.8347, "step": 1000 }, { "epoch": 0.02, "learning_rate": 1.6436781609195406e-05, "loss": 2.7284, "step": 1001 }, { "epoch": 0.02, "learning_rate": 1.645320197044335e-05, "loss": 2.7962, "step": 1002 }, { "epoch": 0.02, "learning_rate": 1.6469622331691298e-05, "loss": 2.8786, "step": 1003 }, { "epoch": 0.02, "learning_rate": 1.6486042692939246e-05, "loss": 2.6965, "step": 1004 }, { "epoch": 0.02, "learning_rate": 1.6502463054187193e-05, "loss": 2.6881, "step": 1005 }, { "epoch": 0.02, "learning_rate": 1.651888341543514e-05, "loss": 2.7066, "step": 1006 }, { "epoch": 0.02, "learning_rate": 1.653530377668309e-05, "loss": 2.8541, "step": 1007 }, { "epoch": 0.02, "learning_rate": 1.6551724137931037e-05, "loss": 2.6913, "step": 1008 }, { "epoch": 0.02, "learning_rate": 1.6568144499178984e-05, "loss": 2.7483, "step": 1009 }, { "epoch": 0.02, "learning_rate": 1.6584564860426932e-05, "loss": 2.7375, "step": 1010 }, { "epoch": 0.02, "learning_rate": 1.660098522167488e-05, "loss": 2.8925, "step": 1011 }, { "epoch": 0.02, "learning_rate": 1.6617405582922824e-05, "loss": 2.5918, "step": 1012 }, { "epoch": 0.02, "learning_rate": 1.6633825944170775e-05, "loss": 2.5681, "step": 1013 }, { "epoch": 0.02, "learning_rate": 1.665024630541872e-05, "loss": 2.6411, "step": 1014 }, { "epoch": 0.03, "learning_rate": 1.6666666666666667e-05, "loss": 2.8153, "step": 1015 }, { "epoch": 0.03, "learning_rate": 1.6683087027914615e-05, "loss": 2.7107, "step": 1016 }, { "epoch": 0.03, "learning_rate": 1.6699507389162563e-05, "loss": 2.5308, "step": 1017 }, { "epoch": 0.03, "learning_rate": 1.671592775041051e-05, "loss": 2.8284, "step": 1018 }, { "epoch": 0.03, "learning_rate": 1.673234811165846e-05, "loss": 2.7887, "step": 1019 }, { "epoch": 0.03, "learning_rate": 1.6748768472906406e-05, "loss": 2.6572, "step": 1020 }, { "epoch": 0.03, "learning_rate": 1.6765188834154354e-05, "loss": 2.5718, "step": 1021 }, { "epoch": 0.03, "learning_rate": 1.6781609195402298e-05, "loss": 2.643, "step": 1022 }, { "epoch": 0.03, "learning_rate": 1.679802955665025e-05, "loss": 2.6444, "step": 1023 }, { "epoch": 0.03, "learning_rate": 1.6814449917898194e-05, "loss": 2.7091, "step": 1024 }, { "epoch": 0.03, "learning_rate": 1.683087027914614e-05, "loss": 2.6489, "step": 1025 }, { "epoch": 0.03, "learning_rate": 1.684729064039409e-05, "loss": 2.8123, "step": 1026 }, { "epoch": 0.03, "learning_rate": 1.6863711001642037e-05, "loss": 2.8192, "step": 1027 }, { "epoch": 0.03, "learning_rate": 1.6880131362889985e-05, "loss": 2.9243, "step": 1028 }, { "epoch": 0.03, "learning_rate": 1.6896551724137932e-05, "loss": 2.7025, "step": 1029 }, { "epoch": 0.03, "learning_rate": 1.691297208538588e-05, "loss": 2.7617, "step": 1030 }, { "epoch": 0.03, "learning_rate": 1.6929392446633828e-05, "loss": 2.6991, "step": 1031 }, { "epoch": 0.03, "learning_rate": 1.6945812807881776e-05, "loss": 2.6991, "step": 1032 }, { "epoch": 0.03, "learning_rate": 1.6962233169129723e-05, "loss": 2.7258, "step": 1033 }, { "epoch": 0.03, "learning_rate": 1.6978653530377668e-05, "loss": 2.7349, "step": 1034 }, { "epoch": 0.03, "learning_rate": 1.699507389162562e-05, "loss": 2.6072, "step": 1035 }, { "epoch": 0.03, "learning_rate": 1.7011494252873563e-05, "loss": 2.5839, "step": 1036 }, { "epoch": 0.03, "learning_rate": 1.702791461412151e-05, "loss": 2.754, "step": 1037 }, { "epoch": 0.03, "learning_rate": 1.704433497536946e-05, "loss": 2.9766, "step": 1038 }, { "epoch": 0.03, "learning_rate": 1.7060755336617407e-05, "loss": 2.7632, "step": 1039 }, { "epoch": 0.03, "learning_rate": 1.7077175697865354e-05, "loss": 2.7245, "step": 1040 }, { "epoch": 0.03, "learning_rate": 1.7093596059113302e-05, "loss": 2.5412, "step": 1041 }, { "epoch": 0.03, "learning_rate": 1.711001642036125e-05, "loss": 2.5641, "step": 1042 }, { "epoch": 0.03, "learning_rate": 1.7126436781609197e-05, "loss": 2.6077, "step": 1043 }, { "epoch": 0.03, "learning_rate": 1.7142857142857142e-05, "loss": 2.7972, "step": 1044 }, { "epoch": 0.03, "learning_rate": 1.7159277504105093e-05, "loss": 2.5811, "step": 1045 }, { "epoch": 0.03, "learning_rate": 1.7175697865353037e-05, "loss": 2.6951, "step": 1046 }, { "epoch": 0.03, "learning_rate": 1.7192118226600985e-05, "loss": 2.7122, "step": 1047 }, { "epoch": 0.03, "learning_rate": 1.7208538587848933e-05, "loss": 2.6318, "step": 1048 }, { "epoch": 0.03, "learning_rate": 1.722495894909688e-05, "loss": 2.8888, "step": 1049 }, { "epoch": 0.03, "learning_rate": 1.7241379310344828e-05, "loss": 2.627, "step": 1050 }, { "epoch": 0.03, "learning_rate": 1.7257799671592776e-05, "loss": 2.7724, "step": 1051 }, { "epoch": 0.03, "learning_rate": 1.7274220032840724e-05, "loss": 2.5949, "step": 1052 }, { "epoch": 0.03, "learning_rate": 1.729064039408867e-05, "loss": 2.7705, "step": 1053 }, { "epoch": 0.03, "learning_rate": 1.730706075533662e-05, "loss": 2.7793, "step": 1054 }, { "epoch": 0.03, "learning_rate": 1.7323481116584567e-05, "loss": 2.8435, "step": 1055 }, { "epoch": 0.03, "learning_rate": 1.7339901477832515e-05, "loss": 2.6684, "step": 1056 }, { "epoch": 0.03, "learning_rate": 1.7356321839080462e-05, "loss": 2.5647, "step": 1057 }, { "epoch": 0.03, "learning_rate": 1.737274220032841e-05, "loss": 2.7834, "step": 1058 }, { "epoch": 0.03, "learning_rate": 1.7389162561576355e-05, "loss": 2.7411, "step": 1059 }, { "epoch": 0.03, "learning_rate": 1.7405582922824306e-05, "loss": 2.6073, "step": 1060 }, { "epoch": 0.03, "learning_rate": 1.742200328407225e-05, "loss": 2.5437, "step": 1061 }, { "epoch": 0.03, "learning_rate": 1.7438423645320198e-05, "loss": 2.7698, "step": 1062 }, { "epoch": 0.03, "learning_rate": 1.7454844006568146e-05, "loss": 2.6315, "step": 1063 }, { "epoch": 0.03, "learning_rate": 1.7471264367816093e-05, "loss": 2.7567, "step": 1064 }, { "epoch": 0.03, "learning_rate": 1.748768472906404e-05, "loss": 2.4731, "step": 1065 }, { "epoch": 0.03, "learning_rate": 1.750410509031199e-05, "loss": 2.7263, "step": 1066 }, { "epoch": 0.03, "learning_rate": 1.7520525451559937e-05, "loss": 2.5472, "step": 1067 }, { "epoch": 0.03, "learning_rate": 1.7536945812807884e-05, "loss": 2.698, "step": 1068 }, { "epoch": 0.03, "learning_rate": 1.755336617405583e-05, "loss": 2.6141, "step": 1069 }, { "epoch": 0.03, "learning_rate": 1.756978653530378e-05, "loss": 2.8143, "step": 1070 }, { "epoch": 0.03, "learning_rate": 1.7586206896551724e-05, "loss": 2.7617, "step": 1071 }, { "epoch": 0.03, "learning_rate": 1.7602627257799675e-05, "loss": 2.624, "step": 1072 }, { "epoch": 0.03, "learning_rate": 1.761904761904762e-05, "loss": 2.7713, "step": 1073 }, { "epoch": 0.03, "learning_rate": 1.7635467980295567e-05, "loss": 2.7413, "step": 1074 }, { "epoch": 0.03, "learning_rate": 1.7651888341543515e-05, "loss": 2.7871, "step": 1075 }, { "epoch": 0.03, "learning_rate": 1.7668308702791463e-05, "loss": 2.7332, "step": 1076 }, { "epoch": 0.03, "learning_rate": 1.768472906403941e-05, "loss": 2.6768, "step": 1077 }, { "epoch": 0.03, "learning_rate": 1.770114942528736e-05, "loss": 2.7145, "step": 1078 }, { "epoch": 0.03, "learning_rate": 1.7717569786535306e-05, "loss": 2.5336, "step": 1079 }, { "epoch": 0.03, "learning_rate": 1.7733990147783254e-05, "loss": 2.5934, "step": 1080 }, { "epoch": 0.03, "learning_rate": 1.7750410509031198e-05, "loss": 2.6108, "step": 1081 }, { "epoch": 0.03, "learning_rate": 1.776683087027915e-05, "loss": 2.5281, "step": 1082 }, { "epoch": 0.03, "learning_rate": 1.7783251231527094e-05, "loss": 2.6578, "step": 1083 }, { "epoch": 0.03, "learning_rate": 1.779967159277504e-05, "loss": 2.5852, "step": 1084 }, { "epoch": 0.03, "learning_rate": 1.781609195402299e-05, "loss": 2.7089, "step": 1085 }, { "epoch": 0.03, "learning_rate": 1.7832512315270937e-05, "loss": 2.7792, "step": 1086 }, { "epoch": 0.03, "learning_rate": 1.7848932676518885e-05, "loss": 2.5348, "step": 1087 }, { "epoch": 0.03, "learning_rate": 1.7865353037766832e-05, "loss": 2.7083, "step": 1088 }, { "epoch": 0.03, "learning_rate": 1.788177339901478e-05, "loss": 2.8066, "step": 1089 }, { "epoch": 0.03, "learning_rate": 1.7898193760262728e-05, "loss": 2.6174, "step": 1090 }, { "epoch": 0.03, "learning_rate": 1.7914614121510676e-05, "loss": 2.8148, "step": 1091 }, { "epoch": 0.03, "learning_rate": 1.7931034482758623e-05, "loss": 2.6716, "step": 1092 }, { "epoch": 0.03, "learning_rate": 1.7947454844006568e-05, "loss": 2.7495, "step": 1093 }, { "epoch": 0.03, "learning_rate": 1.796387520525452e-05, "loss": 2.6386, "step": 1094 }, { "epoch": 0.03, "learning_rate": 1.7980295566502463e-05, "loss": 2.715, "step": 1095 }, { "epoch": 0.03, "learning_rate": 1.799671592775041e-05, "loss": 2.694, "step": 1096 }, { "epoch": 0.03, "learning_rate": 1.801313628899836e-05, "loss": 2.6821, "step": 1097 }, { "epoch": 0.03, "learning_rate": 1.8029556650246306e-05, "loss": 2.7169, "step": 1098 }, { "epoch": 0.03, "learning_rate": 1.8045977011494254e-05, "loss": 2.6506, "step": 1099 }, { "epoch": 0.03, "learning_rate": 1.8062397372742202e-05, "loss": 2.5975, "step": 1100 }, { "epoch": 0.03, "learning_rate": 1.807881773399015e-05, "loss": 2.7116, "step": 1101 }, { "epoch": 0.03, "learning_rate": 1.8095238095238097e-05, "loss": 2.7693, "step": 1102 }, { "epoch": 0.03, "learning_rate": 1.8111658456486042e-05, "loss": 2.564, "step": 1103 }, { "epoch": 0.03, "learning_rate": 1.8128078817733993e-05, "loss": 2.6451, "step": 1104 }, { "epoch": 0.03, "learning_rate": 1.8144499178981937e-05, "loss": 2.6483, "step": 1105 }, { "epoch": 0.03, "learning_rate": 1.8160919540229885e-05, "loss": 2.6198, "step": 1106 }, { "epoch": 0.03, "learning_rate": 1.8177339901477833e-05, "loss": 2.7205, "step": 1107 }, { "epoch": 0.03, "learning_rate": 1.819376026272578e-05, "loss": 2.5422, "step": 1108 }, { "epoch": 0.03, "learning_rate": 1.821018062397373e-05, "loss": 2.5803, "step": 1109 }, { "epoch": 0.03, "learning_rate": 1.8226600985221676e-05, "loss": 2.6443, "step": 1110 }, { "epoch": 0.03, "learning_rate": 1.8243021346469624e-05, "loss": 2.7236, "step": 1111 }, { "epoch": 0.03, "learning_rate": 1.825944170771757e-05, "loss": 2.682, "step": 1112 }, { "epoch": 0.03, "learning_rate": 1.827586206896552e-05, "loss": 2.5128, "step": 1113 }, { "epoch": 0.03, "learning_rate": 1.8292282430213467e-05, "loss": 2.726, "step": 1114 }, { "epoch": 0.03, "learning_rate": 1.8308702791461415e-05, "loss": 2.5596, "step": 1115 }, { "epoch": 0.03, "learning_rate": 1.8325123152709362e-05, "loss": 2.6734, "step": 1116 }, { "epoch": 0.03, "learning_rate": 1.834154351395731e-05, "loss": 2.6099, "step": 1117 }, { "epoch": 0.03, "learning_rate": 1.8357963875205254e-05, "loss": 2.6, "step": 1118 }, { "epoch": 0.03, "learning_rate": 1.8374384236453206e-05, "loss": 2.6615, "step": 1119 }, { "epoch": 0.03, "learning_rate": 1.839080459770115e-05, "loss": 2.7987, "step": 1120 }, { "epoch": 0.03, "learning_rate": 1.8407224958949098e-05, "loss": 2.6542, "step": 1121 }, { "epoch": 0.03, "learning_rate": 1.8423645320197045e-05, "loss": 2.681, "step": 1122 }, { "epoch": 0.03, "learning_rate": 1.8440065681444993e-05, "loss": 2.7291, "step": 1123 }, { "epoch": 0.03, "learning_rate": 1.845648604269294e-05, "loss": 2.623, "step": 1124 }, { "epoch": 0.03, "learning_rate": 1.847290640394089e-05, "loss": 2.5561, "step": 1125 }, { "epoch": 0.03, "learning_rate": 1.8489326765188836e-05, "loss": 2.7116, "step": 1126 }, { "epoch": 0.03, "learning_rate": 1.8505747126436784e-05, "loss": 2.6191, "step": 1127 }, { "epoch": 0.03, "learning_rate": 1.852216748768473e-05, "loss": 2.8955, "step": 1128 }, { "epoch": 0.03, "learning_rate": 1.853858784893268e-05, "loss": 2.5762, "step": 1129 }, { "epoch": 0.03, "learning_rate": 1.8555008210180624e-05, "loss": 2.4601, "step": 1130 }, { "epoch": 0.03, "learning_rate": 1.8571428571428575e-05, "loss": 2.6352, "step": 1131 }, { "epoch": 0.03, "learning_rate": 1.858784893267652e-05, "loss": 2.5294, "step": 1132 }, { "epoch": 0.03, "learning_rate": 1.8604269293924467e-05, "loss": 2.7213, "step": 1133 }, { "epoch": 0.03, "learning_rate": 1.8620689655172415e-05, "loss": 2.5679, "step": 1134 }, { "epoch": 0.03, "learning_rate": 1.8637110016420363e-05, "loss": 2.6229, "step": 1135 }, { "epoch": 0.03, "learning_rate": 1.865353037766831e-05, "loss": 2.5993, "step": 1136 }, { "epoch": 0.03, "learning_rate": 1.8669950738916258e-05, "loss": 2.7109, "step": 1137 }, { "epoch": 0.03, "learning_rate": 1.8686371100164206e-05, "loss": 2.6329, "step": 1138 }, { "epoch": 0.03, "learning_rate": 1.8702791461412154e-05, "loss": 2.5743, "step": 1139 }, { "epoch": 0.03, "learning_rate": 1.8719211822660098e-05, "loss": 2.6962, "step": 1140 }, { "epoch": 0.03, "learning_rate": 1.873563218390805e-05, "loss": 2.5913, "step": 1141 }, { "epoch": 0.03, "learning_rate": 1.8752052545155994e-05, "loss": 2.6572, "step": 1142 }, { "epoch": 0.03, "learning_rate": 1.876847290640394e-05, "loss": 2.6446, "step": 1143 }, { "epoch": 0.03, "learning_rate": 1.878489326765189e-05, "loss": 2.6762, "step": 1144 }, { "epoch": 0.03, "learning_rate": 1.8801313628899837e-05, "loss": 2.8179, "step": 1145 }, { "epoch": 0.03, "learning_rate": 1.8817733990147784e-05, "loss": 2.6381, "step": 1146 }, { "epoch": 0.03, "learning_rate": 1.8834154351395732e-05, "loss": 2.6698, "step": 1147 }, { "epoch": 0.03, "learning_rate": 1.885057471264368e-05, "loss": 2.5152, "step": 1148 }, { "epoch": 0.03, "learning_rate": 1.8866995073891628e-05, "loss": 2.7097, "step": 1149 }, { "epoch": 0.03, "learning_rate": 1.8883415435139572e-05, "loss": 2.7274, "step": 1150 }, { "epoch": 0.03, "learning_rate": 1.8899835796387523e-05, "loss": 2.7593, "step": 1151 }, { "epoch": 0.03, "learning_rate": 1.8916256157635468e-05, "loss": 2.7432, "step": 1152 }, { "epoch": 0.03, "learning_rate": 1.893267651888342e-05, "loss": 2.5602, "step": 1153 }, { "epoch": 0.03, "learning_rate": 1.8949096880131363e-05, "loss": 2.6403, "step": 1154 }, { "epoch": 0.03, "learning_rate": 1.896551724137931e-05, "loss": 2.6862, "step": 1155 }, { "epoch": 0.03, "learning_rate": 1.898193760262726e-05, "loss": 2.8183, "step": 1156 }, { "epoch": 0.03, "learning_rate": 1.8998357963875206e-05, "loss": 2.5518, "step": 1157 }, { "epoch": 0.03, "learning_rate": 1.9014778325123154e-05, "loss": 2.4328, "step": 1158 }, { "epoch": 0.03, "learning_rate": 1.9031198686371102e-05, "loss": 2.6995, "step": 1159 }, { "epoch": 0.03, "learning_rate": 1.904761904761905e-05, "loss": 2.5512, "step": 1160 }, { "epoch": 0.03, "learning_rate": 1.9064039408866997e-05, "loss": 2.614, "step": 1161 }, { "epoch": 0.03, "learning_rate": 1.908045977011494e-05, "loss": 2.5777, "step": 1162 }, { "epoch": 0.03, "learning_rate": 1.9096880131362893e-05, "loss": 2.486, "step": 1163 }, { "epoch": 0.03, "learning_rate": 1.911330049261084e-05, "loss": 2.4466, "step": 1164 }, { "epoch": 0.03, "learning_rate": 1.9129720853858785e-05, "loss": 2.7253, "step": 1165 }, { "epoch": 0.03, "learning_rate": 1.9146141215106736e-05, "loss": 2.7097, "step": 1166 }, { "epoch": 0.03, "learning_rate": 1.916256157635468e-05, "loss": 2.6653, "step": 1167 }, { "epoch": 0.03, "learning_rate": 1.917898193760263e-05, "loss": 2.5102, "step": 1168 }, { "epoch": 0.03, "learning_rate": 1.9195402298850576e-05, "loss": 2.5802, "step": 1169 }, { "epoch": 0.03, "learning_rate": 1.9211822660098524e-05, "loss": 2.6619, "step": 1170 }, { "epoch": 0.03, "learning_rate": 1.922824302134647e-05, "loss": 2.8876, "step": 1171 }, { "epoch": 0.03, "learning_rate": 1.924466338259442e-05, "loss": 2.7899, "step": 1172 }, { "epoch": 0.03, "learning_rate": 1.9261083743842367e-05, "loss": 2.6234, "step": 1173 }, { "epoch": 0.03, "learning_rate": 1.9277504105090315e-05, "loss": 2.5682, "step": 1174 }, { "epoch": 0.03, "learning_rate": 1.9293924466338262e-05, "loss": 2.5462, "step": 1175 }, { "epoch": 0.03, "learning_rate": 1.931034482758621e-05, "loss": 2.5436, "step": 1176 }, { "epoch": 0.03, "learning_rate": 1.9326765188834154e-05, "loss": 2.618, "step": 1177 }, { "epoch": 0.03, "learning_rate": 1.9343185550082105e-05, "loss": 2.54, "step": 1178 }, { "epoch": 0.03, "learning_rate": 1.935960591133005e-05, "loss": 2.5659, "step": 1179 }, { "epoch": 0.03, "learning_rate": 1.9376026272577998e-05, "loss": 2.7513, "step": 1180 }, { "epoch": 0.03, "learning_rate": 1.9392446633825945e-05, "loss": 2.8212, "step": 1181 }, { "epoch": 0.03, "learning_rate": 1.9408866995073893e-05, "loss": 2.5841, "step": 1182 }, { "epoch": 0.03, "learning_rate": 1.942528735632184e-05, "loss": 2.6999, "step": 1183 }, { "epoch": 0.03, "learning_rate": 1.944170771756979e-05, "loss": 2.6129, "step": 1184 }, { "epoch": 0.03, "learning_rate": 1.9458128078817736e-05, "loss": 2.6215, "step": 1185 }, { "epoch": 0.03, "learning_rate": 1.9474548440065684e-05, "loss": 2.7737, "step": 1186 }, { "epoch": 0.03, "learning_rate": 1.949096880131363e-05, "loss": 2.6422, "step": 1187 }, { "epoch": 0.03, "learning_rate": 1.950738916256158e-05, "loss": 2.7369, "step": 1188 }, { "epoch": 0.03, "learning_rate": 1.9523809523809524e-05, "loss": 2.6786, "step": 1189 }, { "epoch": 0.03, "learning_rate": 1.9540229885057475e-05, "loss": 2.6797, "step": 1190 }, { "epoch": 0.03, "learning_rate": 1.955665024630542e-05, "loss": 2.5962, "step": 1191 }, { "epoch": 0.03, "learning_rate": 1.9573070607553367e-05, "loss": 2.5807, "step": 1192 }, { "epoch": 0.03, "learning_rate": 1.9589490968801315e-05, "loss": 2.6877, "step": 1193 }, { "epoch": 0.03, "learning_rate": 1.9605911330049263e-05, "loss": 2.6268, "step": 1194 }, { "epoch": 0.03, "learning_rate": 1.962233169129721e-05, "loss": 2.5114, "step": 1195 }, { "epoch": 0.03, "learning_rate": 1.9638752052545158e-05, "loss": 2.6946, "step": 1196 }, { "epoch": 0.03, "learning_rate": 1.9655172413793106e-05, "loss": 2.6006, "step": 1197 }, { "epoch": 0.03, "learning_rate": 1.9671592775041054e-05, "loss": 2.6422, "step": 1198 }, { "epoch": 0.03, "learning_rate": 1.9688013136288998e-05, "loss": 2.6742, "step": 1199 }, { "epoch": 0.03, "learning_rate": 1.970443349753695e-05, "loss": 2.6367, "step": 1200 }, { "epoch": 0.03, "learning_rate": 1.9720853858784893e-05, "loss": 2.4731, "step": 1201 }, { "epoch": 0.03, "learning_rate": 1.973727422003284e-05, "loss": 2.6862, "step": 1202 }, { "epoch": 0.03, "learning_rate": 1.975369458128079e-05, "loss": 2.6708, "step": 1203 }, { "epoch": 0.03, "learning_rate": 1.9770114942528737e-05, "loss": 2.6885, "step": 1204 }, { "epoch": 0.03, "learning_rate": 1.9786535303776684e-05, "loss": 2.667, "step": 1205 }, { "epoch": 0.03, "learning_rate": 1.9802955665024632e-05, "loss": 2.6272, "step": 1206 }, { "epoch": 0.03, "learning_rate": 1.981937602627258e-05, "loss": 2.662, "step": 1207 }, { "epoch": 0.03, "learning_rate": 1.9835796387520528e-05, "loss": 2.7252, "step": 1208 }, { "epoch": 0.03, "learning_rate": 1.9852216748768472e-05, "loss": 2.4926, "step": 1209 }, { "epoch": 0.03, "learning_rate": 1.9868637110016423e-05, "loss": 2.7068, "step": 1210 }, { "epoch": 0.03, "learning_rate": 1.9885057471264367e-05, "loss": 2.8058, "step": 1211 }, { "epoch": 0.03, "learning_rate": 1.990147783251232e-05, "loss": 2.8214, "step": 1212 }, { "epoch": 0.03, "learning_rate": 1.9917898193760263e-05, "loss": 2.5559, "step": 1213 }, { "epoch": 0.03, "learning_rate": 1.993431855500821e-05, "loss": 2.6368, "step": 1214 }, { "epoch": 0.03, "learning_rate": 1.995073891625616e-05, "loss": 2.562, "step": 1215 }, { "epoch": 0.03, "learning_rate": 1.9967159277504106e-05, "loss": 2.6681, "step": 1216 }, { "epoch": 0.03, "learning_rate": 1.9983579638752054e-05, "loss": 2.6342, "step": 1217 }, { "epoch": 0.03, "learning_rate": 2e-05, "loss": 2.6407, "step": 1218 }, { "epoch": 0.03, "learning_rate": 1.9999999968133372e-05, "loss": 2.6336, "step": 1219 }, { "epoch": 0.03, "learning_rate": 1.9999999872533483e-05, "loss": 2.6736, "step": 1220 }, { "epoch": 0.03, "learning_rate": 1.9999999713200338e-05, "loss": 2.5926, "step": 1221 }, { "epoch": 0.03, "learning_rate": 1.9999999490133933e-05, "loss": 2.6607, "step": 1222 }, { "epoch": 0.03, "learning_rate": 1.9999999203334276e-05, "loss": 2.5891, "step": 1223 }, { "epoch": 0.03, "learning_rate": 1.9999998852801363e-05, "loss": 2.6958, "step": 1224 }, { "epoch": 0.03, "learning_rate": 1.99999984385352e-05, "loss": 2.4376, "step": 1225 }, { "epoch": 0.03, "learning_rate": 1.9999997960535786e-05, "loss": 2.7754, "step": 1226 }, { "epoch": 0.03, "learning_rate": 1.9999997418803125e-05, "loss": 2.5514, "step": 1227 }, { "epoch": 0.03, "learning_rate": 1.9999996813337225e-05, "loss": 2.5891, "step": 1228 }, { "epoch": 0.03, "learning_rate": 1.9999996144138083e-05, "loss": 2.4656, "step": 1229 }, { "epoch": 0.03, "learning_rate": 1.999999541120571e-05, "loss": 2.6091, "step": 1230 }, { "epoch": 0.03, "learning_rate": 1.9999994614540105e-05, "loss": 2.5857, "step": 1231 }, { "epoch": 0.03, "learning_rate": 1.9999993754141276e-05, "loss": 2.7733, "step": 1232 }, { "epoch": 0.03, "learning_rate": 1.9999992830009228e-05, "loss": 2.3786, "step": 1233 }, { "epoch": 0.03, "learning_rate": 1.9999991842143965e-05, "loss": 2.4217, "step": 1234 }, { "epoch": 0.03, "learning_rate": 1.9999990790545497e-05, "loss": 2.8478, "step": 1235 }, { "epoch": 0.03, "learning_rate": 1.999998967521383e-05, "loss": 2.7481, "step": 1236 }, { "epoch": 0.03, "learning_rate": 1.9999988496148968e-05, "loss": 2.7659, "step": 1237 }, { "epoch": 0.03, "learning_rate": 1.9999987253350922e-05, "loss": 2.6376, "step": 1238 }, { "epoch": 0.03, "learning_rate": 1.9999985946819697e-05, "loss": 2.6984, "step": 1239 }, { "epoch": 0.03, "learning_rate": 1.99999845765553e-05, "loss": 2.715, "step": 1240 }, { "epoch": 0.03, "learning_rate": 1.9999983142557747e-05, "loss": 2.7379, "step": 1241 }, { "epoch": 0.03, "learning_rate": 1.9999981644827043e-05, "loss": 2.6724, "step": 1242 }, { "epoch": 0.03, "learning_rate": 1.9999980083363194e-05, "loss": 2.6123, "step": 1243 }, { "epoch": 0.03, "learning_rate": 1.9999978458166214e-05, "loss": 2.5053, "step": 1244 }, { "epoch": 0.03, "learning_rate": 1.999997676923611e-05, "loss": 2.6209, "step": 1245 }, { "epoch": 0.03, "learning_rate": 1.99999750165729e-05, "loss": 2.7699, "step": 1246 }, { "epoch": 0.03, "learning_rate": 1.9999973200176587e-05, "loss": 2.7, "step": 1247 }, { "epoch": 0.03, "learning_rate": 1.999997132004719e-05, "loss": 2.6426, "step": 1248 }, { "epoch": 0.03, "learning_rate": 1.999996937618471e-05, "loss": 2.5717, "step": 1249 }, { "epoch": 0.03, "learning_rate": 1.999996736858917e-05, "loss": 2.4832, "step": 1250 }, { "epoch": 0.03, "learning_rate": 1.999996529726058e-05, "loss": 2.8067, "step": 1251 }, { "epoch": 0.03, "learning_rate": 1.9999963162198953e-05, "loss": 2.5969, "step": 1252 }, { "epoch": 0.03, "learning_rate": 1.99999609634043e-05, "loss": 2.5438, "step": 1253 }, { "epoch": 0.03, "learning_rate": 1.9999958700876636e-05, "loss": 2.5949, "step": 1254 }, { "epoch": 0.03, "learning_rate": 1.9999956374615975e-05, "loss": 2.7466, "step": 1255 }, { "epoch": 0.03, "learning_rate": 1.9999953984622336e-05, "loss": 2.5276, "step": 1256 }, { "epoch": 0.03, "learning_rate": 1.9999951530895732e-05, "loss": 2.3676, "step": 1257 }, { "epoch": 0.03, "learning_rate": 1.9999949013436177e-05, "loss": 2.8463, "step": 1258 }, { "epoch": 0.03, "learning_rate": 1.9999946432243685e-05, "loss": 2.5586, "step": 1259 }, { "epoch": 0.03, "learning_rate": 1.9999943787318282e-05, "loss": 2.5866, "step": 1260 }, { "epoch": 0.03, "learning_rate": 1.9999941078659972e-05, "loss": 2.577, "step": 1261 }, { "epoch": 0.03, "learning_rate": 1.9999938306268786e-05, "loss": 2.67, "step": 1262 }, { "epoch": 0.03, "learning_rate": 1.9999935470144726e-05, "loss": 2.6071, "step": 1263 }, { "epoch": 0.03, "learning_rate": 1.9999932570287824e-05, "loss": 2.5913, "step": 1264 }, { "epoch": 0.03, "learning_rate": 1.999992960669809e-05, "loss": 2.5827, "step": 1265 }, { "epoch": 0.03, "learning_rate": 1.9999926579375546e-05, "loss": 2.4684, "step": 1266 }, { "epoch": 0.03, "learning_rate": 1.999992348832021e-05, "loss": 2.539, "step": 1267 }, { "epoch": 0.03, "learning_rate": 1.9999920333532104e-05, "loss": 2.4934, "step": 1268 }, { "epoch": 0.03, "learning_rate": 1.999991711501125e-05, "loss": 2.6837, "step": 1269 }, { "epoch": 0.03, "learning_rate": 1.999991383275766e-05, "loss": 2.5815, "step": 1270 }, { "epoch": 0.03, "learning_rate": 1.9999910486771365e-05, "loss": 2.5488, "step": 1271 }, { "epoch": 0.03, "learning_rate": 1.999990707705238e-05, "loss": 2.673, "step": 1272 }, { "epoch": 0.03, "learning_rate": 1.9999903603600726e-05, "loss": 2.7209, "step": 1273 }, { "epoch": 0.03, "learning_rate": 1.9999900066416427e-05, "loss": 2.7309, "step": 1274 }, { "epoch": 0.03, "learning_rate": 1.999989646549951e-05, "loss": 2.552, "step": 1275 }, { "epoch": 0.03, "learning_rate": 1.9999892800849993e-05, "loss": 2.5301, "step": 1276 }, { "epoch": 0.03, "learning_rate": 1.99998890724679e-05, "loss": 2.663, "step": 1277 }, { "epoch": 0.03, "learning_rate": 1.9999885280353256e-05, "loss": 2.547, "step": 1278 }, { "epoch": 0.03, "learning_rate": 1.999988142450608e-05, "loss": 2.5974, "step": 1279 }, { "epoch": 0.03, "learning_rate": 1.9999877504926404e-05, "loss": 2.6336, "step": 1280 }, { "epoch": 0.03, "learning_rate": 1.999987352161425e-05, "loss": 2.673, "step": 1281 }, { "epoch": 0.03, "learning_rate": 1.9999869474569643e-05, "loss": 2.545, "step": 1282 }, { "epoch": 0.03, "learning_rate": 1.9999865363792606e-05, "loss": 2.6546, "step": 1283 }, { "epoch": 0.03, "learning_rate": 1.999986118928317e-05, "loss": 2.4439, "step": 1284 }, { "epoch": 0.03, "learning_rate": 1.9999856951041364e-05, "loss": 2.6897, "step": 1285 }, { "epoch": 0.03, "learning_rate": 1.9999852649067206e-05, "loss": 2.492, "step": 1286 }, { "epoch": 0.03, "learning_rate": 1.9999848283360726e-05, "loss": 2.7056, "step": 1287 }, { "epoch": 0.03, "learning_rate": 1.999984385392196e-05, "loss": 2.5565, "step": 1288 }, { "epoch": 0.03, "learning_rate": 1.9999839360750927e-05, "loss": 2.7482, "step": 1289 }, { "epoch": 0.03, "learning_rate": 1.999983480384766e-05, "loss": 2.6686, "step": 1290 }, { "epoch": 0.03, "learning_rate": 1.9999830183212186e-05, "loss": 2.6624, "step": 1291 }, { "epoch": 0.03, "learning_rate": 1.999982549884454e-05, "loss": 2.6561, "step": 1292 }, { "epoch": 0.03, "learning_rate": 1.999982075074474e-05, "loss": 2.4695, "step": 1293 }, { "epoch": 0.03, "learning_rate": 1.9999815938912826e-05, "loss": 2.573, "step": 1294 }, { "epoch": 0.03, "learning_rate": 1.999981106334883e-05, "loss": 2.5891, "step": 1295 }, { "epoch": 0.03, "learning_rate": 1.9999806124052774e-05, "loss": 2.6114, "step": 1296 }, { "epoch": 0.03, "learning_rate": 1.9999801121024695e-05, "loss": 2.5695, "step": 1297 }, { "epoch": 0.03, "learning_rate": 1.9999796054264627e-05, "loss": 2.5041, "step": 1298 }, { "epoch": 0.03, "learning_rate": 1.9999790923772598e-05, "loss": 2.6042, "step": 1299 }, { "epoch": 0.03, "learning_rate": 1.9999785729548648e-05, "loss": 2.7473, "step": 1300 }, { "epoch": 0.03, "learning_rate": 1.99997804715928e-05, "loss": 2.4659, "step": 1301 }, { "epoch": 0.03, "learning_rate": 1.9999775149905095e-05, "loss": 2.483, "step": 1302 }, { "epoch": 0.03, "learning_rate": 1.9999769764485558e-05, "loss": 2.4492, "step": 1303 }, { "epoch": 0.03, "learning_rate": 1.9999764315334235e-05, "loss": 2.7784, "step": 1304 }, { "epoch": 0.03, "learning_rate": 1.9999758802451153e-05, "loss": 2.5377, "step": 1305 }, { "epoch": 0.03, "learning_rate": 1.999975322583635e-05, "loss": 2.638, "step": 1306 }, { "epoch": 0.03, "learning_rate": 1.999974758548986e-05, "loss": 2.7081, "step": 1307 }, { "epoch": 0.03, "learning_rate": 1.9999741881411724e-05, "loss": 2.6399, "step": 1308 }, { "epoch": 0.03, "learning_rate": 1.999973611360197e-05, "loss": 2.4492, "step": 1309 }, { "epoch": 0.03, "learning_rate": 1.999973028206064e-05, "loss": 2.5639, "step": 1310 }, { "epoch": 0.03, "learning_rate": 1.999972438678777e-05, "loss": 2.6076, "step": 1311 }, { "epoch": 0.03, "learning_rate": 1.9999718427783396e-05, "loss": 2.7145, "step": 1312 }, { "epoch": 0.03, "learning_rate": 1.999971240504756e-05, "loss": 2.6954, "step": 1313 }, { "epoch": 0.03, "learning_rate": 1.99997063185803e-05, "loss": 2.6519, "step": 1314 }, { "epoch": 0.03, "learning_rate": 1.9999700168381647e-05, "loss": 2.6487, "step": 1315 }, { "epoch": 0.03, "learning_rate": 1.9999693954451648e-05, "loss": 2.6984, "step": 1316 }, { "epoch": 0.03, "learning_rate": 1.9999687676790342e-05, "loss": 2.7562, "step": 1317 }, { "epoch": 0.03, "learning_rate": 1.9999681335397768e-05, "loss": 2.5591, "step": 1318 }, { "epoch": 0.03, "learning_rate": 1.9999674930273968e-05, "loss": 2.4755, "step": 1319 }, { "epoch": 0.03, "learning_rate": 1.999966846141898e-05, "loss": 2.5512, "step": 1320 }, { "epoch": 0.03, "learning_rate": 1.9999661928832838e-05, "loss": 2.7235, "step": 1321 }, { "epoch": 0.03, "learning_rate": 1.99996553325156e-05, "loss": 2.6878, "step": 1322 }, { "epoch": 0.03, "learning_rate": 1.99996486724673e-05, "loss": 2.5435, "step": 1323 }, { "epoch": 0.03, "learning_rate": 1.9999641948687976e-05, "loss": 2.5428, "step": 1324 }, { "epoch": 0.03, "learning_rate": 1.9999635161177678e-05, "loss": 2.6111, "step": 1325 }, { "epoch": 0.03, "learning_rate": 1.9999628309936444e-05, "loss": 2.5641, "step": 1326 }, { "epoch": 0.03, "learning_rate": 1.9999621394964323e-05, "loss": 2.5308, "step": 1327 }, { "epoch": 0.03, "learning_rate": 1.9999614416261354e-05, "loss": 2.5661, "step": 1328 }, { "epoch": 0.03, "learning_rate": 1.9999607373827583e-05, "loss": 2.5132, "step": 1329 }, { "epoch": 0.03, "learning_rate": 1.9999600267663056e-05, "loss": 2.6467, "step": 1330 }, { "epoch": 0.03, "learning_rate": 1.999959309776782e-05, "loss": 2.5418, "step": 1331 }, { "epoch": 0.03, "learning_rate": 1.9999585864141915e-05, "loss": 2.718, "step": 1332 }, { "epoch": 0.03, "learning_rate": 1.9999578566785392e-05, "loss": 2.6736, "step": 1333 }, { "epoch": 0.03, "learning_rate": 1.9999571205698297e-05, "loss": 2.6766, "step": 1334 }, { "epoch": 0.03, "learning_rate": 1.9999563780880676e-05, "loss": 2.597, "step": 1335 }, { "epoch": 0.03, "learning_rate": 1.9999556292332575e-05, "loss": 2.6169, "step": 1336 }, { "epoch": 0.03, "learning_rate": 1.9999548740054043e-05, "loss": 2.7857, "step": 1337 }, { "epoch": 0.03, "learning_rate": 1.999954112404513e-05, "loss": 2.5228, "step": 1338 }, { "epoch": 0.03, "learning_rate": 1.9999533444305878e-05, "loss": 2.6633, "step": 1339 }, { "epoch": 0.03, "learning_rate": 1.9999525700836347e-05, "loss": 2.6289, "step": 1340 }, { "epoch": 0.03, "learning_rate": 1.9999517893636575e-05, "loss": 2.6091, "step": 1341 }, { "epoch": 0.03, "learning_rate": 1.999951002270662e-05, "loss": 2.5873, "step": 1342 }, { "epoch": 0.03, "learning_rate": 1.9999502088046526e-05, "loss": 2.6163, "step": 1343 }, { "epoch": 0.03, "learning_rate": 1.9999494089656348e-05, "loss": 2.5699, "step": 1344 }, { "epoch": 0.03, "learning_rate": 1.9999486027536135e-05, "loss": 2.6644, "step": 1345 }, { "epoch": 0.03, "learning_rate": 1.999947790168594e-05, "loss": 2.5241, "step": 1346 }, { "epoch": 0.03, "learning_rate": 1.9999469712105812e-05, "loss": 2.5737, "step": 1347 }, { "epoch": 0.03, "learning_rate": 1.9999461458795806e-05, "loss": 2.5259, "step": 1348 }, { "epoch": 0.03, "learning_rate": 1.999945314175597e-05, "loss": 2.5358, "step": 1349 }, { "epoch": 0.03, "learning_rate": 1.9999444760986364e-05, "loss": 2.7222, "step": 1350 }, { "epoch": 0.03, "learning_rate": 1.999943631648704e-05, "loss": 2.5576, "step": 1351 }, { "epoch": 0.03, "learning_rate": 1.9999427808258046e-05, "loss": 2.6894, "step": 1352 }, { "epoch": 0.03, "learning_rate": 1.999941923629944e-05, "loss": 2.7106, "step": 1353 }, { "epoch": 0.03, "learning_rate": 1.9999410600611274e-05, "loss": 2.5892, "step": 1354 }, { "epoch": 0.03, "learning_rate": 1.999940190119361e-05, "loss": 2.6252, "step": 1355 }, { "epoch": 0.03, "learning_rate": 1.99993931380465e-05, "loss": 2.4836, "step": 1356 }, { "epoch": 0.03, "learning_rate": 1.9999384311169994e-05, "loss": 2.6914, "step": 1357 }, { "epoch": 0.03, "learning_rate": 1.9999375420564156e-05, "loss": 2.6077, "step": 1358 }, { "epoch": 0.03, "learning_rate": 1.999936646622904e-05, "loss": 2.4968, "step": 1359 }, { "epoch": 0.03, "learning_rate": 1.99993574481647e-05, "loss": 2.5123, "step": 1360 }, { "epoch": 0.03, "learning_rate": 1.9999348366371197e-05, "loss": 2.6105, "step": 1361 }, { "epoch": 0.03, "learning_rate": 1.999933922084859e-05, "loss": 2.7461, "step": 1362 }, { "epoch": 0.03, "learning_rate": 1.9999330011596937e-05, "loss": 2.5902, "step": 1363 }, { "epoch": 0.03, "learning_rate": 1.9999320738616292e-05, "loss": 2.6899, "step": 1364 }, { "epoch": 0.03, "learning_rate": 1.999931140190672e-05, "loss": 2.634, "step": 1365 }, { "epoch": 0.03, "learning_rate": 1.9999302001468276e-05, "loss": 2.6642, "step": 1366 }, { "epoch": 0.03, "learning_rate": 1.999929253730102e-05, "loss": 2.5123, "step": 1367 }, { "epoch": 0.03, "learning_rate": 1.9999283009405017e-05, "loss": 2.6011, "step": 1368 }, { "epoch": 0.03, "learning_rate": 1.9999273417780322e-05, "loss": 2.4857, "step": 1369 }, { "epoch": 0.03, "learning_rate": 1.9999263762427003e-05, "loss": 2.439, "step": 1370 }, { "epoch": 0.03, "learning_rate": 1.999925404334511e-05, "loss": 2.6492, "step": 1371 }, { "epoch": 0.03, "learning_rate": 1.9999244260534717e-05, "loss": 2.5318, "step": 1372 }, { "epoch": 0.03, "learning_rate": 1.9999234413995884e-05, "loss": 2.6753, "step": 1373 }, { "epoch": 0.03, "learning_rate": 1.999922450372867e-05, "loss": 2.5122, "step": 1374 }, { "epoch": 0.03, "learning_rate": 1.9999214529733136e-05, "loss": 2.4639, "step": 1375 }, { "epoch": 0.03, "learning_rate": 1.999920449200935e-05, "loss": 2.4794, "step": 1376 }, { "epoch": 0.03, "learning_rate": 1.9999194390557378e-05, "loss": 2.4608, "step": 1377 }, { "epoch": 0.03, "learning_rate": 1.9999184225377283e-05, "loss": 2.5519, "step": 1378 }, { "epoch": 0.03, "learning_rate": 1.9999173996469125e-05, "loss": 2.6777, "step": 1379 }, { "epoch": 0.03, "learning_rate": 1.9999163703832973e-05, "loss": 2.5559, "step": 1380 }, { "epoch": 0.03, "learning_rate": 1.999915334746889e-05, "loss": 2.5929, "step": 1381 }, { "epoch": 0.03, "learning_rate": 1.9999142927376948e-05, "loss": 2.6444, "step": 1382 }, { "epoch": 0.03, "learning_rate": 1.9999132443557206e-05, "loss": 2.6889, "step": 1383 }, { "epoch": 0.03, "learning_rate": 1.9999121896009738e-05, "loss": 2.5744, "step": 1384 }, { "epoch": 0.03, "learning_rate": 1.9999111284734604e-05, "loss": 2.7131, "step": 1385 }, { "epoch": 0.03, "learning_rate": 1.999910060973188e-05, "loss": 2.4057, "step": 1386 }, { "epoch": 0.03, "learning_rate": 1.9999089871001624e-05, "loss": 2.6226, "step": 1387 }, { "epoch": 0.03, "learning_rate": 1.999907906854391e-05, "loss": 2.5373, "step": 1388 }, { "epoch": 0.03, "learning_rate": 1.999906820235881e-05, "loss": 2.5381, "step": 1389 }, { "epoch": 0.03, "learning_rate": 1.9999057272446388e-05, "loss": 2.7238, "step": 1390 }, { "epoch": 0.03, "learning_rate": 1.9999046278806714e-05, "loss": 2.7034, "step": 1391 }, { "epoch": 0.03, "learning_rate": 1.9999035221439863e-05, "loss": 2.6794, "step": 1392 }, { "epoch": 0.03, "learning_rate": 1.9999024100345898e-05, "loss": 2.6941, "step": 1393 }, { "epoch": 0.03, "learning_rate": 1.9999012915524894e-05, "loss": 2.6864, "step": 1394 }, { "epoch": 0.03, "learning_rate": 1.9999001666976926e-05, "loss": 2.5105, "step": 1395 }, { "epoch": 0.03, "learning_rate": 1.999899035470206e-05, "loss": 2.5903, "step": 1396 }, { "epoch": 0.03, "learning_rate": 1.999897897870037e-05, "loss": 2.6216, "step": 1397 }, { "epoch": 0.03, "learning_rate": 1.9998967538971925e-05, "loss": 2.5377, "step": 1398 }, { "epoch": 0.03, "learning_rate": 1.9998956035516802e-05, "loss": 2.8332, "step": 1399 }, { "epoch": 0.03, "learning_rate": 1.999894446833508e-05, "loss": 2.5122, "step": 1400 }, { "epoch": 0.03, "learning_rate": 1.999893283742682e-05, "loss": 2.5807, "step": 1401 }, { "epoch": 0.03, "learning_rate": 1.9998921142792106e-05, "loss": 2.704, "step": 1402 }, { "epoch": 0.03, "learning_rate": 1.9998909384431006e-05, "loss": 2.7469, "step": 1403 }, { "epoch": 0.03, "learning_rate": 1.99988975623436e-05, "loss": 2.5018, "step": 1404 }, { "epoch": 0.03, "learning_rate": 1.999888567652996e-05, "loss": 2.7478, "step": 1405 }, { "epoch": 0.03, "learning_rate": 1.9998873726990166e-05, "loss": 2.5675, "step": 1406 }, { "epoch": 0.03, "learning_rate": 1.9998861713724286e-05, "loss": 2.5906, "step": 1407 }, { "epoch": 0.03, "learning_rate": 1.9998849636732406e-05, "loss": 2.6535, "step": 1408 }, { "epoch": 0.03, "learning_rate": 1.99988374960146e-05, "loss": 2.7122, "step": 1409 }, { "epoch": 0.03, "learning_rate": 1.9998825291570937e-05, "loss": 2.5847, "step": 1410 }, { "epoch": 0.03, "learning_rate": 1.9998813023401506e-05, "loss": 2.6265, "step": 1411 }, { "epoch": 0.03, "learning_rate": 1.9998800691506384e-05, "loss": 2.6304, "step": 1412 }, { "epoch": 0.03, "learning_rate": 1.9998788295885644e-05, "loss": 2.7191, "step": 1413 }, { "epoch": 0.03, "learning_rate": 1.9998775836539365e-05, "loss": 2.6052, "step": 1414 }, { "epoch": 0.03, "learning_rate": 1.9998763313467637e-05, "loss": 2.6539, "step": 1415 }, { "epoch": 0.03, "learning_rate": 1.9998750726670523e-05, "loss": 2.5327, "step": 1416 }, { "epoch": 0.03, "learning_rate": 1.9998738076148116e-05, "loss": 2.625, "step": 1417 }, { "epoch": 0.03, "learning_rate": 1.9998725361900492e-05, "loss": 2.5893, "step": 1418 }, { "epoch": 0.03, "learning_rate": 1.9998712583927734e-05, "loss": 2.6583, "step": 1419 }, { "epoch": 0.04, "learning_rate": 1.9998699742229923e-05, "loss": 2.6736, "step": 1420 }, { "epoch": 0.04, "learning_rate": 1.9998686836807136e-05, "loss": 2.6705, "step": 1421 }, { "epoch": 0.04, "learning_rate": 1.9998673867659465e-05, "loss": 2.56, "step": 1422 }, { "epoch": 0.04, "learning_rate": 1.999866083478698e-05, "loss": 2.4497, "step": 1423 }, { "epoch": 0.04, "learning_rate": 1.999864773818978e-05, "loss": 2.7264, "step": 1424 }, { "epoch": 0.04, "learning_rate": 1.9998634577867935e-05, "loss": 2.4774, "step": 1425 }, { "epoch": 0.04, "learning_rate": 1.999862135382153e-05, "loss": 2.5991, "step": 1426 }, { "epoch": 0.04, "learning_rate": 1.9998608066050654e-05, "loss": 2.2574, "step": 1427 }, { "epoch": 0.04, "learning_rate": 1.9998594714555394e-05, "loss": 2.6638, "step": 1428 }, { "epoch": 0.04, "learning_rate": 1.999858129933583e-05, "loss": 2.5892, "step": 1429 }, { "epoch": 0.04, "learning_rate": 1.999856782039205e-05, "loss": 2.6466, "step": 1430 }, { "epoch": 0.04, "learning_rate": 1.9998554277724136e-05, "loss": 2.55, "step": 1431 }, { "epoch": 0.04, "learning_rate": 1.999854067133218e-05, "loss": 2.5168, "step": 1432 }, { "epoch": 0.04, "learning_rate": 1.9998527001216264e-05, "loss": 2.4874, "step": 1433 }, { "epoch": 0.04, "learning_rate": 1.9998513267376474e-05, "loss": 2.5988, "step": 1434 }, { "epoch": 0.04, "learning_rate": 1.9998499469812905e-05, "loss": 2.5203, "step": 1435 }, { "epoch": 0.04, "learning_rate": 1.9998485608525638e-05, "loss": 2.584, "step": 1436 }, { "epoch": 0.04, "learning_rate": 1.9998471683514767e-05, "loss": 2.5705, "step": 1437 }, { "epoch": 0.04, "learning_rate": 1.9998457694780375e-05, "loss": 2.6429, "step": 1438 }, { "epoch": 0.04, "learning_rate": 1.9998443642322552e-05, "loss": 2.586, "step": 1439 }, { "epoch": 0.04, "learning_rate": 1.9998429526141394e-05, "loss": 2.5616, "step": 1440 }, { "epoch": 0.04, "learning_rate": 1.9998415346236985e-05, "loss": 2.4083, "step": 1441 }, { "epoch": 0.04, "learning_rate": 1.9998401102609413e-05, "loss": 2.8248, "step": 1442 }, { "epoch": 0.04, "learning_rate": 1.9998386795258776e-05, "loss": 2.5086, "step": 1443 }, { "epoch": 0.04, "learning_rate": 1.999837242418516e-05, "loss": 2.6115, "step": 1444 }, { "epoch": 0.04, "learning_rate": 1.9998357989388656e-05, "loss": 2.5901, "step": 1445 }, { "epoch": 0.04, "learning_rate": 1.9998343490869365e-05, "loss": 2.5362, "step": 1446 }, { "epoch": 0.04, "learning_rate": 1.9998328928627365e-05, "loss": 2.587, "step": 1447 }, { "epoch": 0.04, "learning_rate": 1.999831430266276e-05, "loss": 2.3438, "step": 1448 }, { "epoch": 0.04, "learning_rate": 1.999829961297564e-05, "loss": 2.5714, "step": 1449 }, { "epoch": 0.04, "learning_rate": 1.9998284859566098e-05, "loss": 2.6058, "step": 1450 }, { "epoch": 0.04, "learning_rate": 1.9998270042434225e-05, "loss": 2.4851, "step": 1451 }, { "epoch": 0.04, "learning_rate": 1.9998255161580122e-05, "loss": 2.6155, "step": 1452 }, { "epoch": 0.04, "learning_rate": 1.9998240217003882e-05, "loss": 2.5739, "step": 1453 }, { "epoch": 0.04, "learning_rate": 1.9998225208705594e-05, "loss": 2.5171, "step": 1454 }, { "epoch": 0.04, "learning_rate": 1.999821013668536e-05, "loss": 2.5719, "step": 1455 }, { "epoch": 0.04, "learning_rate": 1.9998195000943277e-05, "loss": 2.3991, "step": 1456 }, { "epoch": 0.04, "learning_rate": 1.9998179801479434e-05, "loss": 2.7034, "step": 1457 }, { "epoch": 0.04, "learning_rate": 1.9998164538293937e-05, "loss": 2.6868, "step": 1458 }, { "epoch": 0.04, "learning_rate": 1.9998149211386882e-05, "loss": 2.6426, "step": 1459 }, { "epoch": 0.04, "learning_rate": 1.999813382075836e-05, "loss": 2.5057, "step": 1460 }, { "epoch": 0.04, "learning_rate": 1.9998118366408472e-05, "loss": 2.5198, "step": 1461 }, { "epoch": 0.04, "learning_rate": 1.999810284833732e-05, "loss": 2.7108, "step": 1462 }, { "epoch": 0.04, "learning_rate": 1.9998087266544997e-05, "loss": 2.5393, "step": 1463 }, { "epoch": 0.04, "learning_rate": 1.9998071621031606e-05, "loss": 2.5288, "step": 1464 }, { "epoch": 0.04, "learning_rate": 1.9998055911797247e-05, "loss": 2.6246, "step": 1465 }, { "epoch": 0.04, "learning_rate": 1.9998040138842022e-05, "loss": 2.6355, "step": 1466 }, { "epoch": 0.04, "learning_rate": 1.9998024302166027e-05, "loss": 2.6718, "step": 1467 }, { "epoch": 0.04, "learning_rate": 1.9998008401769363e-05, "loss": 2.7769, "step": 1468 }, { "epoch": 0.04, "learning_rate": 1.9997992437652134e-05, "loss": 2.413, "step": 1469 }, { "epoch": 0.04, "learning_rate": 1.999797640981444e-05, "loss": 2.6846, "step": 1470 }, { "epoch": 0.04, "learning_rate": 1.999796031825639e-05, "loss": 2.3686, "step": 1471 }, { "epoch": 0.04, "learning_rate": 1.9997944162978074e-05, "loss": 2.4744, "step": 1472 }, { "epoch": 0.04, "learning_rate": 1.9997927943979606e-05, "loss": 2.6763, "step": 1473 }, { "epoch": 0.04, "learning_rate": 1.9997911661261082e-05, "loss": 2.6466, "step": 1474 }, { "epoch": 0.04, "learning_rate": 1.999789531482261e-05, "loss": 2.5431, "step": 1475 }, { "epoch": 0.04, "learning_rate": 1.9997878904664293e-05, "loss": 2.5508, "step": 1476 }, { "epoch": 0.04, "learning_rate": 1.9997862430786235e-05, "loss": 2.5525, "step": 1477 }, { "epoch": 0.04, "learning_rate": 1.999784589318854e-05, "loss": 2.5174, "step": 1478 }, { "epoch": 0.04, "learning_rate": 1.999782929187132e-05, "loss": 2.697, "step": 1479 }, { "epoch": 0.04, "learning_rate": 1.999781262683467e-05, "loss": 2.4665, "step": 1480 }, { "epoch": 0.04, "learning_rate": 1.9997795898078703e-05, "loss": 2.6632, "step": 1481 }, { "epoch": 0.04, "learning_rate": 1.999777910560353e-05, "loss": 2.5687, "step": 1482 }, { "epoch": 0.04, "learning_rate": 1.9997762249409247e-05, "loss": 2.5914, "step": 1483 }, { "epoch": 0.04, "learning_rate": 1.999774532949597e-05, "loss": 2.484, "step": 1484 }, { "epoch": 0.04, "learning_rate": 1.99977283458638e-05, "loss": 2.4558, "step": 1485 }, { "epoch": 0.04, "learning_rate": 1.9997711298512852e-05, "loss": 2.5679, "step": 1486 }, { "epoch": 0.04, "learning_rate": 1.999769418744323e-05, "loss": 2.5915, "step": 1487 }, { "epoch": 0.04, "learning_rate": 1.999767701265505e-05, "loss": 2.5443, "step": 1488 }, { "epoch": 0.04, "learning_rate": 1.999765977414841e-05, "loss": 2.7757, "step": 1489 }, { "epoch": 0.04, "learning_rate": 1.999764247192343e-05, "loss": 2.4381, "step": 1490 }, { "epoch": 0.04, "learning_rate": 1.9997625105980215e-05, "loss": 2.6169, "step": 1491 }, { "epoch": 0.04, "learning_rate": 1.9997607676318873e-05, "loss": 2.3936, "step": 1492 }, { "epoch": 0.04, "learning_rate": 1.9997590182939522e-05, "loss": 2.672, "step": 1493 }, { "epoch": 0.04, "learning_rate": 1.9997572625842273e-05, "loss": 2.4639, "step": 1494 }, { "epoch": 0.04, "learning_rate": 1.9997555005027235e-05, "loss": 2.4896, "step": 1495 }, { "epoch": 0.04, "learning_rate": 1.999753732049452e-05, "loss": 2.559, "step": 1496 }, { "epoch": 0.04, "learning_rate": 1.9997519572244243e-05, "loss": 2.5389, "step": 1497 }, { "epoch": 0.04, "learning_rate": 1.9997501760276515e-05, "loss": 2.307, "step": 1498 }, { "epoch": 0.04, "learning_rate": 1.999748388459145e-05, "loss": 2.5494, "step": 1499 }, { "epoch": 0.04, "learning_rate": 1.999746594518916e-05, "loss": 2.6158, "step": 1500 }, { "epoch": 0.04, "learning_rate": 1.999744794206976e-05, "loss": 2.408, "step": 1501 }, { "epoch": 0.04, "learning_rate": 1.999742987523337e-05, "loss": 2.6096, "step": 1502 }, { "epoch": 0.04, "learning_rate": 1.99974117446801e-05, "loss": 2.5784, "step": 1503 }, { "epoch": 0.04, "learning_rate": 1.999739355041007e-05, "loss": 2.6382, "step": 1504 }, { "epoch": 0.04, "learning_rate": 1.9997375292423388e-05, "loss": 2.7709, "step": 1505 }, { "epoch": 0.04, "learning_rate": 1.999735697072018e-05, "loss": 2.402, "step": 1506 }, { "epoch": 0.04, "learning_rate": 1.9997338585300556e-05, "loss": 2.5563, "step": 1507 }, { "epoch": 0.04, "learning_rate": 1.9997320136164633e-05, "loss": 2.6153, "step": 1508 }, { "epoch": 0.04, "learning_rate": 1.999730162331253e-05, "loss": 2.5703, "step": 1509 }, { "epoch": 0.04, "learning_rate": 1.999728304674437e-05, "loss": 2.723, "step": 1510 }, { "epoch": 0.04, "learning_rate": 1.9997264406460263e-05, "loss": 2.5055, "step": 1511 }, { "epoch": 0.04, "learning_rate": 1.9997245702460336e-05, "loss": 2.5203, "step": 1512 }, { "epoch": 0.04, "learning_rate": 1.9997226934744703e-05, "loss": 2.7098, "step": 1513 }, { "epoch": 0.04, "learning_rate": 1.9997208103313484e-05, "loss": 2.6788, "step": 1514 }, { "epoch": 0.04, "learning_rate": 1.99971892081668e-05, "loss": 2.5223, "step": 1515 }, { "epoch": 0.04, "learning_rate": 1.9997170249304768e-05, "loss": 2.4223, "step": 1516 }, { "epoch": 0.04, "learning_rate": 1.999715122672752e-05, "loss": 2.601, "step": 1517 }, { "epoch": 0.04, "learning_rate": 1.9997132140435162e-05, "loss": 2.5674, "step": 1518 }, { "epoch": 0.04, "learning_rate": 1.9997112990427824e-05, "loss": 2.5906, "step": 1519 }, { "epoch": 0.04, "learning_rate": 1.9997093776705627e-05, "loss": 2.6461, "step": 1520 }, { "epoch": 0.04, "learning_rate": 1.9997074499268694e-05, "loss": 2.6265, "step": 1521 }, { "epoch": 0.04, "learning_rate": 1.999705515811715e-05, "loss": 2.5464, "step": 1522 }, { "epoch": 0.04, "learning_rate": 1.9997035753251114e-05, "loss": 2.6448, "step": 1523 }, { "epoch": 0.04, "learning_rate": 1.999701628467071e-05, "loss": 2.5571, "step": 1524 }, { "epoch": 0.04, "learning_rate": 1.999699675237606e-05, "loss": 2.6594, "step": 1525 }, { "epoch": 0.04, "learning_rate": 1.99969771563673e-05, "loss": 2.5474, "step": 1526 }, { "epoch": 0.04, "learning_rate": 1.999695749664454e-05, "loss": 2.4697, "step": 1527 }, { "epoch": 0.04, "learning_rate": 1.9996937773207913e-05, "loss": 2.5233, "step": 1528 }, { "epoch": 0.04, "learning_rate": 1.9996917986057544e-05, "loss": 2.5979, "step": 1529 }, { "epoch": 0.04, "learning_rate": 1.999689813519356e-05, "loss": 2.7578, "step": 1530 }, { "epoch": 0.04, "learning_rate": 1.999687822061609e-05, "loss": 2.4089, "step": 1531 }, { "epoch": 0.04, "learning_rate": 1.999685824232525e-05, "loss": 2.624, "step": 1532 }, { "epoch": 0.04, "learning_rate": 1.9996838200321176e-05, "loss": 2.4727, "step": 1533 }, { "epoch": 0.04, "learning_rate": 1.9996818094604e-05, "loss": 2.7195, "step": 1534 }, { "epoch": 0.04, "learning_rate": 1.9996797925173837e-05, "loss": 2.3906, "step": 1535 }, { "epoch": 0.04, "learning_rate": 1.999677769203083e-05, "loss": 2.6431, "step": 1536 }, { "epoch": 0.04, "learning_rate": 1.9996757395175094e-05, "loss": 2.5119, "step": 1537 }, { "epoch": 0.04, "learning_rate": 1.999673703460677e-05, "loss": 2.7319, "step": 1538 }, { "epoch": 0.04, "learning_rate": 1.999671661032598e-05, "loss": 2.4952, "step": 1539 }, { "epoch": 0.04, "learning_rate": 1.999669612233286e-05, "loss": 2.4898, "step": 1540 }, { "epoch": 0.04, "learning_rate": 1.9996675570627536e-05, "loss": 2.636, "step": 1541 }, { "epoch": 0.04, "learning_rate": 1.999665495521014e-05, "loss": 2.4657, "step": 1542 }, { "epoch": 0.04, "learning_rate": 1.9996634276080805e-05, "loss": 2.5545, "step": 1543 }, { "epoch": 0.04, "learning_rate": 1.9996613533239665e-05, "loss": 2.551, "step": 1544 }, { "epoch": 0.04, "learning_rate": 1.9996592726686844e-05, "loss": 2.4912, "step": 1545 }, { "epoch": 0.04, "learning_rate": 1.9996571856422485e-05, "loss": 2.6282, "step": 1546 }, { "epoch": 0.04, "learning_rate": 1.999655092244671e-05, "loss": 2.4909, "step": 1547 }, { "epoch": 0.04, "learning_rate": 1.9996529924759662e-05, "loss": 2.6857, "step": 1548 }, { "epoch": 0.04, "learning_rate": 1.999650886336147e-05, "loss": 2.4567, "step": 1549 }, { "epoch": 0.04, "learning_rate": 1.999648773825227e-05, "loss": 2.6434, "step": 1550 }, { "epoch": 0.04, "learning_rate": 1.9996466549432195e-05, "loss": 2.5056, "step": 1551 }, { "epoch": 0.04, "learning_rate": 1.9996445296901382e-05, "loss": 2.3922, "step": 1552 }, { "epoch": 0.04, "learning_rate": 1.9996423980659964e-05, "loss": 2.4312, "step": 1553 }, { "epoch": 0.04, "learning_rate": 1.9996402600708078e-05, "loss": 2.549, "step": 1554 }, { "epoch": 0.04, "learning_rate": 1.999638115704586e-05, "loss": 2.4758, "step": 1555 }, { "epoch": 0.04, "learning_rate": 1.999635964967345e-05, "loss": 2.7157, "step": 1556 }, { "epoch": 0.04, "learning_rate": 1.999633807859098e-05, "loss": 2.6848, "step": 1557 }, { "epoch": 0.04, "learning_rate": 1.999631644379859e-05, "loss": 2.3347, "step": 1558 }, { "epoch": 0.04, "learning_rate": 1.999629474529642e-05, "loss": 2.6066, "step": 1559 }, { "epoch": 0.04, "learning_rate": 1.9996272983084603e-05, "loss": 2.4772, "step": 1560 }, { "epoch": 0.04, "learning_rate": 1.9996251157163283e-05, "loss": 2.4714, "step": 1561 }, { "epoch": 0.04, "learning_rate": 1.9996229267532594e-05, "loss": 2.6145, "step": 1562 }, { "epoch": 0.04, "learning_rate": 1.9996207314192678e-05, "loss": 2.4552, "step": 1563 }, { "epoch": 0.04, "learning_rate": 1.9996185297143677e-05, "loss": 2.4633, "step": 1564 }, { "epoch": 0.04, "learning_rate": 1.999616321638573e-05, "loss": 2.5588, "step": 1565 }, { "epoch": 0.04, "learning_rate": 1.9996141071918974e-05, "loss": 2.813, "step": 1566 }, { "epoch": 0.04, "learning_rate": 1.9996118863743558e-05, "loss": 2.5658, "step": 1567 }, { "epoch": 0.04, "learning_rate": 1.9996096591859616e-05, "loss": 2.5188, "step": 1568 }, { "epoch": 0.04, "learning_rate": 1.9996074256267293e-05, "loss": 2.5469, "step": 1569 }, { "epoch": 0.04, "learning_rate": 1.9996051856966732e-05, "loss": 2.8548, "step": 1570 }, { "epoch": 0.04, "learning_rate": 1.9996029393958075e-05, "loss": 2.6556, "step": 1571 }, { "epoch": 0.04, "learning_rate": 1.9996006867241462e-05, "loss": 2.5251, "step": 1572 }, { "epoch": 0.04, "learning_rate": 1.999598427681704e-05, "loss": 2.4369, "step": 1573 }, { "epoch": 0.04, "learning_rate": 1.9995961622684958e-05, "loss": 2.4885, "step": 1574 }, { "epoch": 0.04, "learning_rate": 1.999593890484535e-05, "loss": 2.3643, "step": 1575 }, { "epoch": 0.04, "learning_rate": 1.9995916123298368e-05, "loss": 2.4696, "step": 1576 }, { "epoch": 0.04, "learning_rate": 1.9995893278044152e-05, "loss": 2.4733, "step": 1577 }, { "epoch": 0.04, "learning_rate": 1.9995870369082854e-05, "loss": 2.4982, "step": 1578 }, { "epoch": 0.04, "learning_rate": 1.9995847396414615e-05, "loss": 2.3785, "step": 1579 }, { "epoch": 0.04, "learning_rate": 1.9995824360039582e-05, "loss": 2.4495, "step": 1580 }, { "epoch": 0.04, "learning_rate": 1.9995801259957907e-05, "loss": 2.4931, "step": 1581 }, { "epoch": 0.04, "learning_rate": 1.9995778096169728e-05, "loss": 2.5126, "step": 1582 }, { "epoch": 0.04, "learning_rate": 1.9995754868675198e-05, "loss": 2.5264, "step": 1583 }, { "epoch": 0.04, "learning_rate": 1.999573157747447e-05, "loss": 2.5207, "step": 1584 }, { "epoch": 0.04, "learning_rate": 1.9995708222567683e-05, "loss": 2.4986, "step": 1585 }, { "epoch": 0.04, "learning_rate": 1.999568480395499e-05, "loss": 2.5508, "step": 1586 }, { "epoch": 0.04, "learning_rate": 1.9995661321636542e-05, "loss": 2.6339, "step": 1587 }, { "epoch": 0.04, "learning_rate": 1.9995637775612484e-05, "loss": 2.484, "step": 1588 }, { "epoch": 0.04, "learning_rate": 1.999561416588297e-05, "loss": 2.5856, "step": 1589 }, { "epoch": 0.04, "learning_rate": 1.9995590492448152e-05, "loss": 2.5373, "step": 1590 }, { "epoch": 0.04, "learning_rate": 1.9995566755308175e-05, "loss": 2.5642, "step": 1591 }, { "epoch": 0.04, "learning_rate": 1.9995542954463193e-05, "loss": 2.61, "step": 1592 }, { "epoch": 0.04, "learning_rate": 1.9995519089913364e-05, "loss": 2.5707, "step": 1593 }, { "epoch": 0.04, "learning_rate": 1.999549516165883e-05, "loss": 2.342, "step": 1594 }, { "epoch": 0.04, "learning_rate": 1.999547116969975e-05, "loss": 2.7033, "step": 1595 }, { "epoch": 0.04, "learning_rate": 1.999544711403627e-05, "loss": 2.4757, "step": 1596 }, { "epoch": 0.04, "learning_rate": 1.9995422994668556e-05, "loss": 2.6541, "step": 1597 }, { "epoch": 0.04, "learning_rate": 1.9995398811596747e-05, "loss": 2.3045, "step": 1598 }, { "epoch": 0.04, "learning_rate": 1.9995374564821008e-05, "loss": 2.5102, "step": 1599 }, { "epoch": 0.04, "learning_rate": 1.9995350254341485e-05, "loss": 2.5041, "step": 1600 }, { "epoch": 0.04, "learning_rate": 1.9995325880158342e-05, "loss": 2.559, "step": 1601 }, { "epoch": 0.04, "learning_rate": 1.9995301442271725e-05, "loss": 2.6674, "step": 1602 }, { "epoch": 0.04, "learning_rate": 1.99952769406818e-05, "loss": 2.5851, "step": 1603 }, { "epoch": 0.04, "learning_rate": 1.9995252375388716e-05, "loss": 2.4487, "step": 1604 }, { "epoch": 0.04, "learning_rate": 1.9995227746392628e-05, "loss": 2.702, "step": 1605 }, { "epoch": 0.04, "learning_rate": 1.9995203053693703e-05, "loss": 2.5937, "step": 1606 }, { "epoch": 0.04, "learning_rate": 1.9995178297292086e-05, "loss": 2.7134, "step": 1607 }, { "epoch": 0.04, "learning_rate": 1.999515347718794e-05, "loss": 2.5263, "step": 1608 }, { "epoch": 0.04, "learning_rate": 1.9995128593381427e-05, "loss": 2.3755, "step": 1609 }, { "epoch": 0.04, "learning_rate": 1.99951036458727e-05, "loss": 2.3201, "step": 1610 }, { "epoch": 0.04, "learning_rate": 1.999507863466192e-05, "loss": 2.6593, "step": 1611 }, { "epoch": 0.04, "learning_rate": 1.999505355974925e-05, "loss": 2.4874, "step": 1612 }, { "epoch": 0.04, "learning_rate": 1.999502842113484e-05, "loss": 2.5343, "step": 1613 }, { "epoch": 0.04, "learning_rate": 1.9995003218818862e-05, "loss": 2.6802, "step": 1614 }, { "epoch": 0.04, "learning_rate": 1.9994977952801468e-05, "loss": 2.5752, "step": 1615 }, { "epoch": 0.04, "learning_rate": 1.9994952623082823e-05, "loss": 2.5349, "step": 1616 }, { "epoch": 0.04, "learning_rate": 1.999492722966309e-05, "loss": 2.6643, "step": 1617 }, { "epoch": 0.04, "learning_rate": 1.9994901772542427e-05, "loss": 2.4183, "step": 1618 }, { "epoch": 0.04, "learning_rate": 1.9994876251721e-05, "loss": 2.5563, "step": 1619 }, { "epoch": 0.04, "learning_rate": 1.9994850667198963e-05, "loss": 2.5032, "step": 1620 }, { "epoch": 0.04, "learning_rate": 1.999482501897649e-05, "loss": 2.6237, "step": 1621 }, { "epoch": 0.04, "learning_rate": 1.999479930705374e-05, "loss": 2.5567, "step": 1622 }, { "epoch": 0.04, "learning_rate": 1.9994773531430877e-05, "loss": 2.561, "step": 1623 }, { "epoch": 0.04, "learning_rate": 1.9994747692108065e-05, "loss": 2.4816, "step": 1624 }, { "epoch": 0.04, "learning_rate": 1.9994721789085466e-05, "loss": 2.7591, "step": 1625 }, { "epoch": 0.04, "learning_rate": 1.999469582236325e-05, "loss": 2.4524, "step": 1626 }, { "epoch": 0.04, "learning_rate": 1.999466979194158e-05, "loss": 2.615, "step": 1627 }, { "epoch": 0.04, "learning_rate": 1.9994643697820624e-05, "loss": 2.6107, "step": 1628 }, { "epoch": 0.04, "learning_rate": 1.9994617540000546e-05, "loss": 2.704, "step": 1629 }, { "epoch": 0.04, "learning_rate": 1.999459131848151e-05, "loss": 2.5974, "step": 1630 }, { "epoch": 0.04, "learning_rate": 1.999456503326369e-05, "loss": 2.5484, "step": 1631 }, { "epoch": 0.04, "learning_rate": 1.9994538684347246e-05, "loss": 2.4891, "step": 1632 }, { "epoch": 0.04, "learning_rate": 1.9994512271732356e-05, "loss": 2.4971, "step": 1633 }, { "epoch": 0.04, "learning_rate": 1.999448579541918e-05, "loss": 2.5457, "step": 1634 }, { "epoch": 0.04, "learning_rate": 1.9994459255407882e-05, "loss": 2.5982, "step": 1635 }, { "epoch": 0.04, "learning_rate": 1.999443265169865e-05, "loss": 2.6029, "step": 1636 }, { "epoch": 0.04, "learning_rate": 1.999440598429163e-05, "loss": 2.5431, "step": 1637 }, { "epoch": 0.04, "learning_rate": 1.999437925318701e-05, "loss": 2.6321, "step": 1638 }, { "epoch": 0.04, "learning_rate": 1.9994352458384952e-05, "loss": 2.4363, "step": 1639 }, { "epoch": 0.04, "learning_rate": 1.9994325599885628e-05, "loss": 2.4413, "step": 1640 }, { "epoch": 0.04, "learning_rate": 1.9994298677689215e-05, "loss": 2.4708, "step": 1641 }, { "epoch": 0.04, "learning_rate": 1.9994271691795874e-05, "loss": 2.5263, "step": 1642 }, { "epoch": 0.04, "learning_rate": 1.9994244642205784e-05, "loss": 2.6184, "step": 1643 }, { "epoch": 0.04, "learning_rate": 1.9994217528919115e-05, "loss": 2.5361, "step": 1644 }, { "epoch": 0.04, "learning_rate": 1.9994190351936042e-05, "loss": 2.6887, "step": 1645 }, { "epoch": 0.04, "learning_rate": 1.999416311125674e-05, "loss": 2.3315, "step": 1646 }, { "epoch": 0.04, "learning_rate": 1.9994135806881374e-05, "loss": 2.5833, "step": 1647 }, { "epoch": 0.04, "learning_rate": 1.9994108438810125e-05, "loss": 2.6217, "step": 1648 }, { "epoch": 0.04, "learning_rate": 1.999408100704317e-05, "loss": 2.5946, "step": 1649 }, { "epoch": 0.04, "learning_rate": 1.9994053511580677e-05, "loss": 2.5371, "step": 1650 }, { "epoch": 0.04, "learning_rate": 1.9994025952422823e-05, "loss": 2.5511, "step": 1651 }, { "epoch": 0.04, "learning_rate": 1.9993998329569783e-05, "loss": 2.5812, "step": 1652 }, { "epoch": 0.04, "learning_rate": 1.9993970643021735e-05, "loss": 2.4966, "step": 1653 }, { "epoch": 0.04, "learning_rate": 1.999394289277886e-05, "loss": 2.5373, "step": 1654 }, { "epoch": 0.04, "learning_rate": 1.9993915078841325e-05, "loss": 2.4164, "step": 1655 }, { "epoch": 0.04, "learning_rate": 1.9993887201209315e-05, "loss": 2.4447, "step": 1656 }, { "epoch": 0.04, "learning_rate": 1.9993859259883003e-05, "loss": 2.5859, "step": 1657 }, { "epoch": 0.04, "learning_rate": 1.999383125486257e-05, "loss": 2.4694, "step": 1658 }, { "epoch": 0.04, "learning_rate": 1.9993803186148195e-05, "loss": 2.2847, "step": 1659 }, { "epoch": 0.04, "learning_rate": 1.9993775053740056e-05, "loss": 2.5493, "step": 1660 }, { "epoch": 0.04, "learning_rate": 1.999374685763833e-05, "loss": 2.6761, "step": 1661 }, { "epoch": 0.04, "learning_rate": 1.9993718597843193e-05, "loss": 2.5994, "step": 1662 }, { "epoch": 0.04, "learning_rate": 1.9993690274354836e-05, "loss": 2.3318, "step": 1663 }, { "epoch": 0.04, "learning_rate": 1.9993661887173433e-05, "loss": 2.474, "step": 1664 }, { "epoch": 0.04, "learning_rate": 1.9993633436299168e-05, "loss": 2.7015, "step": 1665 }, { "epoch": 0.04, "learning_rate": 1.9993604921732223e-05, "loss": 2.7741, "step": 1666 }, { "epoch": 0.04, "learning_rate": 1.999357634347277e-05, "loss": 2.7702, "step": 1667 }, { "epoch": 0.04, "learning_rate": 1.9993547701521004e-05, "loss": 2.3123, "step": 1668 }, { "epoch": 0.04, "learning_rate": 1.9993518995877097e-05, "loss": 2.477, "step": 1669 }, { "epoch": 0.04, "learning_rate": 1.9993490226541235e-05, "loss": 2.5788, "step": 1670 }, { "epoch": 0.04, "learning_rate": 1.999346139351361e-05, "loss": 2.5091, "step": 1671 }, { "epoch": 0.04, "learning_rate": 1.9993432496794397e-05, "loss": 2.4487, "step": 1672 }, { "epoch": 0.04, "learning_rate": 1.999340353638378e-05, "loss": 2.5651, "step": 1673 }, { "epoch": 0.04, "learning_rate": 1.999337451228194e-05, "loss": 2.4224, "step": 1674 }, { "epoch": 0.04, "learning_rate": 1.9993345424489078e-05, "loss": 2.6801, "step": 1675 }, { "epoch": 0.04, "learning_rate": 1.9993316273005364e-05, "loss": 2.3631, "step": 1676 }, { "epoch": 0.04, "learning_rate": 1.9993287057830986e-05, "loss": 2.5931, "step": 1677 }, { "epoch": 0.04, "learning_rate": 1.9993257778966138e-05, "loss": 2.5938, "step": 1678 }, { "epoch": 0.04, "learning_rate": 1.9993228436411e-05, "loss": 2.559, "step": 1679 }, { "epoch": 0.04, "learning_rate": 1.9993199030165758e-05, "loss": 2.7639, "step": 1680 }, { "epoch": 0.04, "learning_rate": 1.9993169560230602e-05, "loss": 2.5811, "step": 1681 }, { "epoch": 0.04, "learning_rate": 1.999314002660572e-05, "loss": 2.5911, "step": 1682 }, { "epoch": 0.04, "learning_rate": 1.99931104292913e-05, "loss": 2.4243, "step": 1683 }, { "epoch": 0.04, "learning_rate": 1.9993080768287535e-05, "loss": 2.7168, "step": 1684 }, { "epoch": 0.04, "learning_rate": 1.9993051043594603e-05, "loss": 2.713, "step": 1685 }, { "epoch": 0.04, "learning_rate": 1.9993021255212705e-05, "loss": 2.5942, "step": 1686 }, { "epoch": 0.04, "learning_rate": 1.9992991403142026e-05, "loss": 2.5435, "step": 1687 }, { "epoch": 0.04, "learning_rate": 1.9992961487382754e-05, "loss": 2.3986, "step": 1688 }, { "epoch": 0.04, "learning_rate": 1.999293150793508e-05, "loss": 2.4886, "step": 1689 }, { "epoch": 0.04, "learning_rate": 1.99929014647992e-05, "loss": 2.5088, "step": 1690 }, { "epoch": 0.04, "learning_rate": 1.9992871357975304e-05, "loss": 2.6392, "step": 1691 }, { "epoch": 0.04, "learning_rate": 1.999284118746358e-05, "loss": 2.4358, "step": 1692 }, { "epoch": 0.04, "learning_rate": 1.999281095326422e-05, "loss": 2.4153, "step": 1693 }, { "epoch": 0.04, "learning_rate": 1.9992780655377423e-05, "loss": 2.4641, "step": 1694 }, { "epoch": 0.04, "learning_rate": 1.9992750293803375e-05, "loss": 2.5405, "step": 1695 }, { "epoch": 0.04, "learning_rate": 1.9992719868542276e-05, "loss": 2.5714, "step": 1696 }, { "epoch": 0.04, "learning_rate": 1.9992689379594314e-05, "loss": 2.5605, "step": 1697 }, { "epoch": 0.04, "learning_rate": 1.9992658826959687e-05, "loss": 2.4281, "step": 1698 }, { "epoch": 0.04, "learning_rate": 1.999262821063859e-05, "loss": 2.5933, "step": 1699 }, { "epoch": 0.04, "learning_rate": 1.9992597530631217e-05, "loss": 2.4403, "step": 1700 }, { "epoch": 0.04, "learning_rate": 1.9992566786937765e-05, "loss": 2.508, "step": 1701 }, { "epoch": 0.04, "learning_rate": 1.9992535979558422e-05, "loss": 2.4838, "step": 1702 }, { "epoch": 0.04, "learning_rate": 1.9992505108493397e-05, "loss": 2.6505, "step": 1703 }, { "epoch": 0.04, "learning_rate": 1.9992474173742877e-05, "loss": 2.4682, "step": 1704 }, { "epoch": 0.04, "learning_rate": 1.9992443175307065e-05, "loss": 2.5945, "step": 1705 }, { "epoch": 0.04, "learning_rate": 1.999241211318615e-05, "loss": 2.4347, "step": 1706 }, { "epoch": 0.04, "learning_rate": 1.9992380987380344e-05, "loss": 2.7048, "step": 1707 }, { "epoch": 0.04, "learning_rate": 1.9992349797889836e-05, "loss": 2.4863, "step": 1708 }, { "epoch": 0.04, "learning_rate": 1.9992318544714828e-05, "loss": 2.5476, "step": 1709 }, { "epoch": 0.04, "learning_rate": 1.999228722785551e-05, "loss": 2.518, "step": 1710 }, { "epoch": 0.04, "learning_rate": 1.9992255847312094e-05, "loss": 2.5149, "step": 1711 }, { "epoch": 0.04, "learning_rate": 1.9992224403084776e-05, "loss": 2.7247, "step": 1712 }, { "epoch": 0.04, "learning_rate": 1.9992192895173753e-05, "loss": 2.4246, "step": 1713 }, { "epoch": 0.04, "learning_rate": 1.9992161323579232e-05, "loss": 2.5312, "step": 1714 }, { "epoch": 0.04, "learning_rate": 1.9992129688301406e-05, "loss": 2.584, "step": 1715 }, { "epoch": 0.04, "learning_rate": 1.9992097989340486e-05, "loss": 2.5943, "step": 1716 }, { "epoch": 0.04, "learning_rate": 1.9992066226696664e-05, "loss": 2.5016, "step": 1717 }, { "epoch": 0.04, "learning_rate": 1.999203440037015e-05, "loss": 2.4874, "step": 1718 }, { "epoch": 0.04, "learning_rate": 1.9992002510361145e-05, "loss": 2.6803, "step": 1719 }, { "epoch": 0.04, "learning_rate": 1.9991970556669852e-05, "loss": 2.5179, "step": 1720 }, { "epoch": 0.04, "learning_rate": 1.9991938539296474e-05, "loss": 2.7006, "step": 1721 }, { "epoch": 0.04, "learning_rate": 1.9991906458241213e-05, "loss": 2.4545, "step": 1722 }, { "epoch": 0.04, "learning_rate": 1.999187431350428e-05, "loss": 2.5037, "step": 1723 }, { "epoch": 0.04, "learning_rate": 1.999184210508587e-05, "loss": 2.4522, "step": 1724 }, { "epoch": 0.04, "learning_rate": 1.99918098329862e-05, "loss": 2.4351, "step": 1725 }, { "epoch": 0.04, "learning_rate": 1.9991777497205468e-05, "loss": 2.492, "step": 1726 }, { "epoch": 0.04, "learning_rate": 1.9991745097743882e-05, "loss": 2.6947, "step": 1727 }, { "epoch": 0.04, "learning_rate": 1.9991712634601644e-05, "loss": 2.4487, "step": 1728 }, { "epoch": 0.04, "learning_rate": 1.999168010777897e-05, "loss": 2.5868, "step": 1729 }, { "epoch": 0.04, "learning_rate": 1.9991647517276064e-05, "loss": 2.5008, "step": 1730 }, { "epoch": 0.04, "learning_rate": 1.9991614863093127e-05, "loss": 2.4069, "step": 1731 }, { "epoch": 0.04, "learning_rate": 1.9991582145230376e-05, "loss": 2.3404, "step": 1732 }, { "epoch": 0.04, "learning_rate": 1.9991549363688013e-05, "loss": 2.552, "step": 1733 }, { "epoch": 0.04, "learning_rate": 1.9991516518466252e-05, "loss": 2.4362, "step": 1734 }, { "epoch": 0.04, "learning_rate": 1.99914836095653e-05, "loss": 2.6175, "step": 1735 }, { "epoch": 0.04, "learning_rate": 1.9991450636985364e-05, "loss": 2.4559, "step": 1736 }, { "epoch": 0.04, "learning_rate": 1.999141760072666e-05, "loss": 2.516, "step": 1737 }, { "epoch": 0.04, "learning_rate": 1.9991384500789395e-05, "loss": 2.3234, "step": 1738 }, { "epoch": 0.04, "learning_rate": 1.999135133717378e-05, "loss": 2.5976, "step": 1739 }, { "epoch": 0.04, "learning_rate": 1.9991318109880023e-05, "loss": 2.6336, "step": 1740 }, { "epoch": 0.04, "learning_rate": 1.999128481890835e-05, "loss": 2.6268, "step": 1741 }, { "epoch": 0.04, "learning_rate": 1.9991251464258954e-05, "loss": 2.4362, "step": 1742 }, { "epoch": 0.04, "learning_rate": 1.9991218045932058e-05, "loss": 2.5472, "step": 1743 }, { "epoch": 0.04, "learning_rate": 1.9991184563927874e-05, "loss": 2.3669, "step": 1744 }, { "epoch": 0.04, "learning_rate": 1.9991151018246616e-05, "loss": 2.57, "step": 1745 }, { "epoch": 0.04, "learning_rate": 1.999111740888849e-05, "loss": 2.4393, "step": 1746 }, { "epoch": 0.04, "learning_rate": 1.9991083735853726e-05, "loss": 2.5852, "step": 1747 }, { "epoch": 0.04, "learning_rate": 1.9991049999142524e-05, "loss": 2.2621, "step": 1748 }, { "epoch": 0.04, "learning_rate": 1.9991016198755108e-05, "loss": 2.3632, "step": 1749 }, { "epoch": 0.04, "learning_rate": 1.9990982334691686e-05, "loss": 2.5957, "step": 1750 }, { "epoch": 0.04, "learning_rate": 1.9990948406952477e-05, "loss": 2.4682, "step": 1751 }, { "epoch": 0.04, "learning_rate": 1.99909144155377e-05, "loss": 2.7641, "step": 1752 }, { "epoch": 0.04, "learning_rate": 1.999088036044757e-05, "loss": 2.5025, "step": 1753 }, { "epoch": 0.04, "learning_rate": 1.9990846241682306e-05, "loss": 2.3171, "step": 1754 }, { "epoch": 0.04, "learning_rate": 1.999081205924212e-05, "loss": 2.4488, "step": 1755 }, { "epoch": 0.04, "learning_rate": 1.999077781312723e-05, "loss": 2.4599, "step": 1756 }, { "epoch": 0.04, "learning_rate": 1.999074350333786e-05, "loss": 2.4091, "step": 1757 }, { "epoch": 0.04, "learning_rate": 1.9990709129874226e-05, "loss": 2.5171, "step": 1758 }, { "epoch": 0.04, "learning_rate": 1.999067469273655e-05, "loss": 2.4, "step": 1759 }, { "epoch": 0.04, "learning_rate": 1.9990640191925042e-05, "loss": 2.64, "step": 1760 }, { "epoch": 0.04, "learning_rate": 1.999060562743993e-05, "loss": 2.4833, "step": 1761 }, { "epoch": 0.04, "learning_rate": 1.9990570999281433e-05, "loss": 2.4514, "step": 1762 }, { "epoch": 0.04, "learning_rate": 1.999053630744977e-05, "loss": 2.375, "step": 1763 }, { "epoch": 0.04, "learning_rate": 1.9990501551945168e-05, "loss": 2.6245, "step": 1764 }, { "epoch": 0.04, "learning_rate": 1.999046673276784e-05, "loss": 2.5647, "step": 1765 }, { "epoch": 0.04, "learning_rate": 1.999043184991801e-05, "loss": 2.5944, "step": 1766 }, { "epoch": 0.04, "learning_rate": 1.999039690339591e-05, "loss": 2.3769, "step": 1767 }, { "epoch": 0.04, "learning_rate": 1.9990361893201746e-05, "loss": 2.5491, "step": 1768 }, { "epoch": 0.04, "learning_rate": 1.999032681933575e-05, "loss": 2.7689, "step": 1769 }, { "epoch": 0.04, "learning_rate": 1.9990291681798154e-05, "loss": 2.3985, "step": 1770 }, { "epoch": 0.04, "learning_rate": 1.9990256480589164e-05, "loss": 2.6371, "step": 1771 }, { "epoch": 0.04, "learning_rate": 1.999022121570902e-05, "loss": 2.456, "step": 1772 }, { "epoch": 0.04, "learning_rate": 1.9990185887157938e-05, "loss": 2.5992, "step": 1773 }, { "epoch": 0.04, "learning_rate": 1.9990150494936147e-05, "loss": 2.5699, "step": 1774 }, { "epoch": 0.04, "learning_rate": 1.999011503904387e-05, "loss": 2.6124, "step": 1775 }, { "epoch": 0.04, "learning_rate": 1.9990079519481332e-05, "loss": 2.5582, "step": 1776 }, { "epoch": 0.04, "learning_rate": 1.9990043936248764e-05, "loss": 2.451, "step": 1777 }, { "epoch": 0.04, "learning_rate": 1.999000828934639e-05, "loss": 2.6411, "step": 1778 }, { "epoch": 0.04, "learning_rate": 1.9989972578774438e-05, "loss": 2.5562, "step": 1779 }, { "epoch": 0.04, "learning_rate": 1.9989936804533135e-05, "loss": 2.482, "step": 1780 }, { "epoch": 0.04, "learning_rate": 1.998990096662271e-05, "loss": 2.6269, "step": 1781 }, { "epoch": 0.04, "learning_rate": 1.9989865065043386e-05, "loss": 2.5765, "step": 1782 }, { "epoch": 0.04, "learning_rate": 1.99898290997954e-05, "loss": 2.6396, "step": 1783 }, { "epoch": 0.04, "learning_rate": 1.998979307087898e-05, "loss": 2.4686, "step": 1784 }, { "epoch": 0.04, "learning_rate": 1.998975697829435e-05, "loss": 2.4325, "step": 1785 }, { "epoch": 0.04, "learning_rate": 1.9989720822041742e-05, "loss": 2.5932, "step": 1786 }, { "epoch": 0.04, "learning_rate": 1.998968460212139e-05, "loss": 2.4124, "step": 1787 }, { "epoch": 0.04, "learning_rate": 1.998964831853352e-05, "loss": 2.5347, "step": 1788 }, { "epoch": 0.04, "learning_rate": 1.998961197127837e-05, "loss": 2.3878, "step": 1789 }, { "epoch": 0.04, "learning_rate": 1.998957556035616e-05, "loss": 2.5515, "step": 1790 }, { "epoch": 0.04, "learning_rate": 1.9989539085767137e-05, "loss": 2.4528, "step": 1791 }, { "epoch": 0.04, "learning_rate": 1.9989502547511527e-05, "loss": 2.5266, "step": 1792 }, { "epoch": 0.04, "learning_rate": 1.9989465945589557e-05, "loss": 2.5603, "step": 1793 }, { "epoch": 0.04, "learning_rate": 1.9989429280001464e-05, "loss": 2.594, "step": 1794 }, { "epoch": 0.04, "learning_rate": 1.9989392550747487e-05, "loss": 2.4503, "step": 1795 }, { "epoch": 0.04, "learning_rate": 1.998935575782785e-05, "loss": 2.5379, "step": 1796 }, { "epoch": 0.04, "learning_rate": 1.99893189012428e-05, "loss": 2.466, "step": 1797 }, { "epoch": 0.04, "learning_rate": 1.998928198099256e-05, "loss": 2.5171, "step": 1798 }, { "epoch": 0.04, "learning_rate": 1.9989244997077372e-05, "loss": 2.6003, "step": 1799 }, { "epoch": 0.04, "learning_rate": 1.998920794949747e-05, "loss": 2.5715, "step": 1800 }, { "epoch": 0.04, "learning_rate": 1.9989170838253095e-05, "loss": 2.5945, "step": 1801 }, { "epoch": 0.04, "learning_rate": 1.9989133663344473e-05, "loss": 2.5535, "step": 1802 }, { "epoch": 0.04, "learning_rate": 1.9989096424771848e-05, "loss": 2.6203, "step": 1803 }, { "epoch": 0.04, "learning_rate": 1.9989059122535458e-05, "loss": 2.4611, "step": 1804 }, { "epoch": 0.04, "learning_rate": 1.9989021756635537e-05, "loss": 2.5231, "step": 1805 }, { "epoch": 0.04, "learning_rate": 1.9988984327072327e-05, "loss": 2.507, "step": 1806 }, { "epoch": 0.04, "learning_rate": 1.9988946833846064e-05, "loss": 2.5469, "step": 1807 }, { "epoch": 0.04, "learning_rate": 1.9988909276956987e-05, "loss": 2.5038, "step": 1808 }, { "epoch": 0.04, "learning_rate": 1.9988871656405335e-05, "loss": 2.4887, "step": 1809 }, { "epoch": 0.04, "learning_rate": 1.998883397219135e-05, "loss": 2.3138, "step": 1810 }, { "epoch": 0.04, "learning_rate": 1.998879622431527e-05, "loss": 2.5544, "step": 1811 }, { "epoch": 0.04, "learning_rate": 1.998875841277734e-05, "loss": 2.5796, "step": 1812 }, { "epoch": 0.04, "learning_rate": 1.9988720537577793e-05, "loss": 2.4969, "step": 1813 }, { "epoch": 0.04, "learning_rate": 1.998868259871687e-05, "loss": 2.4141, "step": 1814 }, { "epoch": 0.04, "learning_rate": 1.9988644596194827e-05, "loss": 2.5846, "step": 1815 }, { "epoch": 0.04, "learning_rate": 1.9988606530011893e-05, "loss": 2.5131, "step": 1816 }, { "epoch": 0.04, "learning_rate": 1.9988568400168315e-05, "loss": 2.5108, "step": 1817 }, { "epoch": 0.04, "learning_rate": 1.9988530206664335e-05, "loss": 2.6831, "step": 1818 }, { "epoch": 0.04, "learning_rate": 1.99884919495002e-05, "loss": 2.6164, "step": 1819 }, { "epoch": 0.04, "learning_rate": 1.9988453628676146e-05, "loss": 2.5158, "step": 1820 }, { "epoch": 0.04, "learning_rate": 1.9988415244192424e-05, "loss": 2.6603, "step": 1821 }, { "epoch": 0.04, "learning_rate": 1.9988376796049276e-05, "loss": 2.5123, "step": 1822 }, { "epoch": 0.04, "learning_rate": 1.9988338284246947e-05, "loss": 2.4671, "step": 1823 }, { "epoch": 0.04, "learning_rate": 1.9988299708785685e-05, "loss": 2.483, "step": 1824 }, { "epoch": 0.04, "learning_rate": 1.9988261069665732e-05, "loss": 2.3186, "step": 1825 }, { "epoch": 0.05, "learning_rate": 1.9988222366887337e-05, "loss": 2.655, "step": 1826 }, { "epoch": 0.05, "learning_rate": 1.9988183600450743e-05, "loss": 2.3804, "step": 1827 }, { "epoch": 0.05, "learning_rate": 1.9988144770356204e-05, "loss": 2.4302, "step": 1828 }, { "epoch": 0.05, "learning_rate": 1.998810587660396e-05, "loss": 2.4248, "step": 1829 }, { "epoch": 0.05, "learning_rate": 1.9988066919194265e-05, "loss": 2.5566, "step": 1830 }, { "epoch": 0.05, "learning_rate": 1.9988027898127363e-05, "loss": 2.4637, "step": 1831 }, { "epoch": 0.05, "learning_rate": 1.9987988813403506e-05, "loss": 2.495, "step": 1832 }, { "epoch": 0.05, "learning_rate": 1.9987949665022937e-05, "loss": 2.597, "step": 1833 }, { "epoch": 0.05, "learning_rate": 1.998791045298591e-05, "loss": 2.4874, "step": 1834 }, { "epoch": 0.05, "learning_rate": 1.998787117729268e-05, "loss": 2.4566, "step": 1835 }, { "epoch": 0.05, "learning_rate": 1.9987831837943488e-05, "loss": 2.4699, "step": 1836 }, { "epoch": 0.05, "learning_rate": 1.998779243493859e-05, "loss": 2.5135, "step": 1837 }, { "epoch": 0.05, "learning_rate": 1.9987752968278237e-05, "loss": 2.603, "step": 1838 }, { "epoch": 0.05, "learning_rate": 1.998771343796268e-05, "loss": 2.4771, "step": 1839 }, { "epoch": 0.05, "learning_rate": 1.9987673843992165e-05, "loss": 2.6437, "step": 1840 }, { "epoch": 0.05, "learning_rate": 1.9987634186366955e-05, "loss": 2.5841, "step": 1841 }, { "epoch": 0.05, "learning_rate": 1.9987594465087297e-05, "loss": 2.4142, "step": 1842 }, { "epoch": 0.05, "learning_rate": 1.9987554680153442e-05, "loss": 2.5191, "step": 1843 }, { "epoch": 0.05, "learning_rate": 1.9987514831565648e-05, "loss": 2.3644, "step": 1844 }, { "epoch": 0.05, "learning_rate": 1.9987474919324167e-05, "loss": 2.6145, "step": 1845 }, { "epoch": 0.05, "learning_rate": 1.9987434943429255e-05, "loss": 2.4448, "step": 1846 }, { "epoch": 0.05, "learning_rate": 1.998739490388116e-05, "loss": 2.3758, "step": 1847 }, { "epoch": 0.05, "learning_rate": 1.998735480068015e-05, "loss": 2.4093, "step": 1848 }, { "epoch": 0.05, "learning_rate": 1.998731463382647e-05, "loss": 2.4187, "step": 1849 }, { "epoch": 0.05, "learning_rate": 1.9987274403320375e-05, "loss": 2.5134, "step": 1850 }, { "epoch": 0.05, "learning_rate": 1.9987234109162127e-05, "loss": 2.7425, "step": 1851 }, { "epoch": 0.05, "learning_rate": 1.9987193751351985e-05, "loss": 2.4725, "step": 1852 }, { "epoch": 0.05, "learning_rate": 1.99871533298902e-05, "loss": 2.5678, "step": 1853 }, { "epoch": 0.05, "learning_rate": 1.9987112844777034e-05, "loss": 2.7348, "step": 1854 }, { "epoch": 0.05, "learning_rate": 1.998707229601274e-05, "loss": 2.5031, "step": 1855 }, { "epoch": 0.05, "learning_rate": 1.9987031683597583e-05, "loss": 2.5506, "step": 1856 }, { "epoch": 0.05, "learning_rate": 1.9986991007531816e-05, "loss": 2.542, "step": 1857 }, { "epoch": 0.05, "learning_rate": 1.99869502678157e-05, "loss": 2.4335, "step": 1858 }, { "epoch": 0.05, "learning_rate": 1.99869094644495e-05, "loss": 2.3524, "step": 1859 }, { "epoch": 0.05, "learning_rate": 1.9986868597433465e-05, "loss": 2.6045, "step": 1860 }, { "epoch": 0.05, "learning_rate": 1.9986827666767867e-05, "loss": 2.3648, "step": 1861 }, { "epoch": 0.05, "learning_rate": 1.9986786672452957e-05, "loss": 2.4227, "step": 1862 }, { "epoch": 0.05, "learning_rate": 1.9986745614489006e-05, "loss": 2.3647, "step": 1863 }, { "epoch": 0.05, "learning_rate": 1.9986704492876265e-05, "loss": 2.4571, "step": 1864 }, { "epoch": 0.05, "learning_rate": 1.9986663307615006e-05, "loss": 2.3138, "step": 1865 }, { "epoch": 0.05, "learning_rate": 1.9986622058705488e-05, "loss": 2.2214, "step": 1866 }, { "epoch": 0.05, "learning_rate": 1.998658074614797e-05, "loss": 2.4711, "step": 1867 }, { "epoch": 0.05, "learning_rate": 1.9986539369942716e-05, "loss": 2.73, "step": 1868 }, { "epoch": 0.05, "learning_rate": 1.9986497930089997e-05, "loss": 2.4973, "step": 1869 }, { "epoch": 0.05, "learning_rate": 1.9986456426590074e-05, "loss": 2.5346, "step": 1870 }, { "epoch": 0.05, "learning_rate": 1.9986414859443204e-05, "loss": 2.4132, "step": 1871 }, { "epoch": 0.05, "learning_rate": 1.998637322864966e-05, "loss": 2.3403, "step": 1872 }, { "epoch": 0.05, "learning_rate": 1.9986331534209704e-05, "loss": 2.5196, "step": 1873 }, { "epoch": 0.05, "learning_rate": 1.9986289776123603e-05, "loss": 2.3953, "step": 1874 }, { "epoch": 0.05, "learning_rate": 1.9986247954391626e-05, "loss": 2.5451, "step": 1875 }, { "epoch": 0.05, "learning_rate": 1.998620606901403e-05, "loss": 2.5464, "step": 1876 }, { "epoch": 0.05, "learning_rate": 1.9986164119991094e-05, "loss": 2.6568, "step": 1877 }, { "epoch": 0.05, "learning_rate": 1.9986122107323078e-05, "loss": 2.5453, "step": 1878 }, { "epoch": 0.05, "learning_rate": 1.9986080031010253e-05, "loss": 2.3784, "step": 1879 }, { "epoch": 0.05, "learning_rate": 1.998603789105288e-05, "loss": 2.419, "step": 1880 }, { "epoch": 0.05, "learning_rate": 1.998599568745124e-05, "loss": 2.425, "step": 1881 }, { "epoch": 0.05, "learning_rate": 1.998595342020559e-05, "loss": 2.4839, "step": 1882 }, { "epoch": 0.05, "learning_rate": 1.9985911089316207e-05, "loss": 2.6127, "step": 1883 }, { "epoch": 0.05, "learning_rate": 1.9985868694783356e-05, "loss": 2.4266, "step": 1884 }, { "epoch": 0.05, "learning_rate": 1.9985826236607312e-05, "loss": 2.5028, "step": 1885 }, { "epoch": 0.05, "learning_rate": 1.9985783714788344e-05, "loss": 2.4331, "step": 1886 }, { "epoch": 0.05, "learning_rate": 1.9985741129326718e-05, "loss": 2.4033, "step": 1887 }, { "epoch": 0.05, "learning_rate": 1.9985698480222716e-05, "loss": 2.345, "step": 1888 }, { "epoch": 0.05, "learning_rate": 1.99856557674766e-05, "loss": 2.4046, "step": 1889 }, { "epoch": 0.05, "learning_rate": 1.998561299108864e-05, "loss": 2.5788, "step": 1890 }, { "epoch": 0.05, "learning_rate": 1.9985570151059122e-05, "loss": 2.6023, "step": 1891 }, { "epoch": 0.05, "learning_rate": 1.998552724738831e-05, "loss": 2.6118, "step": 1892 }, { "epoch": 0.05, "learning_rate": 1.9985484280076475e-05, "loss": 2.564, "step": 1893 }, { "epoch": 0.05, "learning_rate": 1.9985441249123895e-05, "loss": 2.603, "step": 1894 }, { "epoch": 0.05, "learning_rate": 1.9985398154530846e-05, "loss": 2.5298, "step": 1895 }, { "epoch": 0.05, "learning_rate": 1.99853549962976e-05, "loss": 2.6563, "step": 1896 }, { "epoch": 0.05, "learning_rate": 1.9985311774424433e-05, "loss": 2.5018, "step": 1897 }, { "epoch": 0.05, "learning_rate": 1.998526848891162e-05, "loss": 2.4526, "step": 1898 }, { "epoch": 0.05, "learning_rate": 1.9985225139759435e-05, "loss": 2.364, "step": 1899 }, { "epoch": 0.05, "learning_rate": 1.9985181726968156e-05, "loss": 2.3259, "step": 1900 }, { "epoch": 0.05, "learning_rate": 1.998513825053806e-05, "loss": 2.4892, "step": 1901 }, { "epoch": 0.05, "learning_rate": 1.9985094710469426e-05, "loss": 2.3694, "step": 1902 }, { "epoch": 0.05, "learning_rate": 1.998505110676253e-05, "loss": 2.415, "step": 1903 }, { "epoch": 0.05, "learning_rate": 1.9985007439417645e-05, "loss": 2.4911, "step": 1904 }, { "epoch": 0.05, "learning_rate": 1.9984963708435058e-05, "loss": 2.3653, "step": 1905 }, { "epoch": 0.05, "learning_rate": 1.998491991381504e-05, "loss": 2.3297, "step": 1906 }, { "epoch": 0.05, "learning_rate": 1.9984876055557874e-05, "loss": 2.4852, "step": 1907 }, { "epoch": 0.05, "learning_rate": 1.9984832133663842e-05, "loss": 2.6383, "step": 1908 }, { "epoch": 0.05, "learning_rate": 1.9984788148133217e-05, "loss": 2.4328, "step": 1909 }, { "epoch": 0.05, "learning_rate": 1.998474409896628e-05, "loss": 2.7599, "step": 1910 }, { "epoch": 0.05, "learning_rate": 1.9984699986163322e-05, "loss": 2.6467, "step": 1911 }, { "epoch": 0.05, "learning_rate": 1.9984655809724614e-05, "loss": 2.3627, "step": 1912 }, { "epoch": 0.05, "learning_rate": 1.998461156965044e-05, "loss": 2.4606, "step": 1913 }, { "epoch": 0.05, "learning_rate": 1.9984567265941086e-05, "loss": 2.5038, "step": 1914 }, { "epoch": 0.05, "learning_rate": 1.998452289859683e-05, "loss": 2.5974, "step": 1915 }, { "epoch": 0.05, "learning_rate": 1.998447846761795e-05, "loss": 2.7116, "step": 1916 }, { "epoch": 0.05, "learning_rate": 1.9984433973004742e-05, "loss": 2.6136, "step": 1917 }, { "epoch": 0.05, "learning_rate": 1.998438941475748e-05, "loss": 2.5895, "step": 1918 }, { "epoch": 0.05, "learning_rate": 1.998434479287645e-05, "loss": 2.4849, "step": 1919 }, { "epoch": 0.05, "learning_rate": 1.998430010736194e-05, "loss": 2.4153, "step": 1920 }, { "epoch": 0.05, "learning_rate": 1.9984255358214226e-05, "loss": 2.3967, "step": 1921 }, { "epoch": 0.05, "learning_rate": 1.9984210545433603e-05, "loss": 2.528, "step": 1922 }, { "epoch": 0.05, "learning_rate": 1.998416566902035e-05, "loss": 2.4224, "step": 1923 }, { "epoch": 0.05, "learning_rate": 1.9984120728974754e-05, "loss": 2.4341, "step": 1924 }, { "epoch": 0.05, "learning_rate": 1.9984075725297107e-05, "loss": 2.5864, "step": 1925 }, { "epoch": 0.05, "learning_rate": 1.998403065798769e-05, "loss": 2.4336, "step": 1926 }, { "epoch": 0.05, "learning_rate": 1.998398552704679e-05, "loss": 2.5388, "step": 1927 }, { "epoch": 0.05, "learning_rate": 1.9983940332474702e-05, "loss": 2.6106, "step": 1928 }, { "epoch": 0.05, "learning_rate": 1.99838950742717e-05, "loss": 2.4167, "step": 1929 }, { "epoch": 0.05, "learning_rate": 1.9983849752438087e-05, "loss": 2.4866, "step": 1930 }, { "epoch": 0.05, "learning_rate": 1.9983804366974145e-05, "loss": 2.5346, "step": 1931 }, { "epoch": 0.05, "learning_rate": 1.9983758917880164e-05, "loss": 2.6999, "step": 1932 }, { "epoch": 0.05, "learning_rate": 1.9983713405156437e-05, "loss": 2.6814, "step": 1933 }, { "epoch": 0.05, "learning_rate": 1.9983667828803246e-05, "loss": 2.395, "step": 1934 }, { "epoch": 0.05, "learning_rate": 1.998362218882089e-05, "loss": 2.4308, "step": 1935 }, { "epoch": 0.05, "learning_rate": 1.998357648520965e-05, "loss": 2.3528, "step": 1936 }, { "epoch": 0.05, "learning_rate": 1.998353071796983e-05, "loss": 2.303, "step": 1937 }, { "epoch": 0.05, "learning_rate": 1.998348488710171e-05, "loss": 2.441, "step": 1938 }, { "epoch": 0.05, "learning_rate": 1.998343899260559e-05, "loss": 2.3107, "step": 1939 }, { "epoch": 0.05, "learning_rate": 1.9983393034481763e-05, "loss": 2.5038, "step": 1940 }, { "epoch": 0.05, "learning_rate": 1.9983347012730516e-05, "loss": 2.4467, "step": 1941 }, { "epoch": 0.05, "learning_rate": 1.9983300927352144e-05, "loss": 2.4566, "step": 1942 }, { "epoch": 0.05, "learning_rate": 1.998325477834694e-05, "loss": 2.5395, "step": 1943 }, { "epoch": 0.05, "learning_rate": 1.9983208565715203e-05, "loss": 2.7184, "step": 1944 }, { "epoch": 0.05, "learning_rate": 1.9983162289457224e-05, "loss": 2.5646, "step": 1945 }, { "epoch": 0.05, "learning_rate": 1.99831159495733e-05, "loss": 2.3869, "step": 1946 }, { "epoch": 0.05, "learning_rate": 1.9983069546063724e-05, "loss": 2.4319, "step": 1947 }, { "epoch": 0.05, "learning_rate": 1.9983023078928788e-05, "loss": 2.4083, "step": 1948 }, { "epoch": 0.05, "learning_rate": 1.9982976548168796e-05, "loss": 2.5014, "step": 1949 }, { "epoch": 0.05, "learning_rate": 1.998292995378404e-05, "loss": 2.677, "step": 1950 }, { "epoch": 0.05, "learning_rate": 1.998288329577482e-05, "loss": 2.5321, "step": 1951 }, { "epoch": 0.05, "learning_rate": 1.998283657414143e-05, "loss": 2.5515, "step": 1952 }, { "epoch": 0.05, "learning_rate": 1.9982789788884167e-05, "loss": 2.6443, "step": 1953 }, { "epoch": 0.05, "learning_rate": 1.9982742940003334e-05, "loss": 2.6492, "step": 1954 }, { "epoch": 0.05, "learning_rate": 1.9982696027499226e-05, "loss": 2.6443, "step": 1955 }, { "epoch": 0.05, "learning_rate": 1.9982649051372147e-05, "loss": 2.3679, "step": 1956 }, { "epoch": 0.05, "learning_rate": 1.9982602011622386e-05, "loss": 2.5035, "step": 1957 }, { "epoch": 0.05, "learning_rate": 1.9982554908250253e-05, "loss": 2.5509, "step": 1958 }, { "epoch": 0.05, "learning_rate": 1.998250774125604e-05, "loss": 2.5259, "step": 1959 }, { "epoch": 0.05, "learning_rate": 1.998246051064006e-05, "loss": 2.225, "step": 1960 }, { "epoch": 0.05, "learning_rate": 1.9982413216402602e-05, "loss": 2.5075, "step": 1961 }, { "epoch": 0.05, "learning_rate": 1.998236585854397e-05, "loss": 2.241, "step": 1962 }, { "epoch": 0.05, "learning_rate": 1.9982318437064467e-05, "loss": 2.2824, "step": 1963 }, { "epoch": 0.05, "learning_rate": 1.9982270951964396e-05, "loss": 2.5254, "step": 1964 }, { "epoch": 0.05, "learning_rate": 1.9982223403244057e-05, "loss": 2.4151, "step": 1965 }, { "epoch": 0.05, "learning_rate": 1.998217579090376e-05, "loss": 2.4546, "step": 1966 }, { "epoch": 0.05, "learning_rate": 1.99821281149438e-05, "loss": 2.5679, "step": 1967 }, { "epoch": 0.05, "learning_rate": 1.998208037536449e-05, "loss": 2.4828, "step": 1968 }, { "epoch": 0.05, "learning_rate": 1.9982032572166125e-05, "loss": 2.6446, "step": 1969 }, { "epoch": 0.05, "learning_rate": 1.998198470534901e-05, "loss": 2.4818, "step": 1970 }, { "epoch": 0.05, "learning_rate": 1.9981936774913456e-05, "loss": 2.6108, "step": 1971 }, { "epoch": 0.05, "learning_rate": 1.9981888780859765e-05, "loss": 2.5069, "step": 1972 }, { "epoch": 0.05, "learning_rate": 1.9981840723188245e-05, "loss": 2.4232, "step": 1973 }, { "epoch": 0.05, "learning_rate": 1.9981792601899204e-05, "loss": 2.4598, "step": 1974 }, { "epoch": 0.05, "learning_rate": 1.998174441699294e-05, "loss": 2.6656, "step": 1975 }, { "epoch": 0.05, "learning_rate": 1.9981696168469772e-05, "loss": 2.5541, "step": 1976 }, { "epoch": 0.05, "learning_rate": 1.998164785633e-05, "loss": 2.5363, "step": 1977 }, { "epoch": 0.05, "learning_rate": 1.9981599480573927e-05, "loss": 2.5432, "step": 1978 }, { "epoch": 0.05, "learning_rate": 1.9981551041201873e-05, "loss": 2.5173, "step": 1979 }, { "epoch": 0.05, "learning_rate": 1.998150253821414e-05, "loss": 2.4168, "step": 1980 }, { "epoch": 0.05, "learning_rate": 1.998145397161104e-05, "loss": 2.5408, "step": 1981 }, { "epoch": 0.05, "learning_rate": 1.9981405341392877e-05, "loss": 2.5538, "step": 1982 }, { "epoch": 0.05, "learning_rate": 1.998135664755997e-05, "loss": 2.5897, "step": 1983 }, { "epoch": 0.05, "learning_rate": 1.998130789011262e-05, "loss": 2.3956, "step": 1984 }, { "epoch": 0.05, "learning_rate": 1.9981259069051145e-05, "loss": 2.4487, "step": 1985 }, { "epoch": 0.05, "learning_rate": 1.9981210184375852e-05, "loss": 2.634, "step": 1986 }, { "epoch": 0.05, "learning_rate": 1.9981161236087054e-05, "loss": 2.5908, "step": 1987 }, { "epoch": 0.05, "learning_rate": 1.998111222418506e-05, "loss": 2.5679, "step": 1988 }, { "epoch": 0.05, "learning_rate": 1.998106314867019e-05, "loss": 2.5354, "step": 1989 }, { "epoch": 0.05, "learning_rate": 1.9981014009542747e-05, "loss": 2.5047, "step": 1990 }, { "epoch": 0.05, "learning_rate": 1.9980964806803052e-05, "loss": 2.578, "step": 1991 }, { "epoch": 0.05, "learning_rate": 1.9980915540451417e-05, "loss": 2.4157, "step": 1992 }, { "epoch": 0.05, "learning_rate": 1.998086621048815e-05, "loss": 2.6041, "step": 1993 }, { "epoch": 0.05, "learning_rate": 1.9980816816913574e-05, "loss": 2.4587, "step": 1994 }, { "epoch": 0.05, "learning_rate": 1.9980767359727998e-05, "loss": 2.4244, "step": 1995 }, { "epoch": 0.05, "learning_rate": 1.9980717838931738e-05, "loss": 2.3664, "step": 1996 }, { "epoch": 0.05, "learning_rate": 1.9980668254525112e-05, "loss": 2.3468, "step": 1997 }, { "epoch": 0.05, "learning_rate": 1.9980618606508434e-05, "loss": 2.6629, "step": 1998 }, { "epoch": 0.05, "learning_rate": 1.998056889488202e-05, "loss": 2.4736, "step": 1999 }, { "epoch": 0.05, "learning_rate": 1.998051911964619e-05, "loss": 2.5344, "step": 2000 }, { "epoch": 0.05, "learning_rate": 1.998046928080126e-05, "loss": 2.6767, "step": 2001 }, { "epoch": 0.05, "learning_rate": 1.9980419378347545e-05, "loss": 2.4198, "step": 2002 }, { "epoch": 0.05, "learning_rate": 1.9980369412285363e-05, "loss": 2.3768, "step": 2003 }, { "epoch": 0.05, "learning_rate": 1.9980319382615036e-05, "loss": 2.4587, "step": 2004 }, { "epoch": 0.05, "learning_rate": 1.998026928933688e-05, "loss": 2.7332, "step": 2005 }, { "epoch": 0.05, "learning_rate": 1.998021913245122e-05, "loss": 2.5807, "step": 2006 }, { "epoch": 0.05, "learning_rate": 1.9980168911958366e-05, "loss": 2.6973, "step": 2007 }, { "epoch": 0.05, "learning_rate": 1.9980118627858643e-05, "loss": 2.5528, "step": 2008 }, { "epoch": 0.05, "learning_rate": 1.9980068280152377e-05, "loss": 2.4058, "step": 2009 }, { "epoch": 0.05, "learning_rate": 1.998001786883988e-05, "loss": 2.5124, "step": 2010 }, { "epoch": 0.05, "learning_rate": 1.9979967393921473e-05, "loss": 2.53, "step": 2011 }, { "epoch": 0.05, "learning_rate": 1.9979916855397484e-05, "loss": 2.6262, "step": 2012 }, { "epoch": 0.05, "learning_rate": 1.9979866253268234e-05, "loss": 2.5551, "step": 2013 }, { "epoch": 0.05, "learning_rate": 1.9979815587534045e-05, "loss": 2.5829, "step": 2014 }, { "epoch": 0.05, "learning_rate": 1.9979764858195235e-05, "loss": 2.4925, "step": 2015 }, { "epoch": 0.05, "learning_rate": 1.9979714065252135e-05, "loss": 2.4412, "step": 2016 }, { "epoch": 0.05, "learning_rate": 1.9979663208705062e-05, "loss": 2.5116, "step": 2017 }, { "epoch": 0.05, "learning_rate": 1.9979612288554344e-05, "loss": 2.715, "step": 2018 }, { "epoch": 0.05, "learning_rate": 1.9979561304800304e-05, "loss": 2.4537, "step": 2019 }, { "epoch": 0.05, "learning_rate": 1.9979510257443272e-05, "loss": 2.6018, "step": 2020 }, { "epoch": 0.05, "learning_rate": 1.9979459146483566e-05, "loss": 2.45, "step": 2021 }, { "epoch": 0.05, "learning_rate": 1.9979407971921514e-05, "loss": 2.4572, "step": 2022 }, { "epoch": 0.05, "learning_rate": 1.9979356733757442e-05, "loss": 2.5152, "step": 2023 }, { "epoch": 0.05, "learning_rate": 1.997930543199168e-05, "loss": 2.4008, "step": 2024 }, { "epoch": 0.05, "learning_rate": 1.997925406662455e-05, "loss": 2.6711, "step": 2025 }, { "epoch": 0.05, "learning_rate": 1.997920263765638e-05, "loss": 2.6692, "step": 2026 }, { "epoch": 0.05, "learning_rate": 1.9979151145087504e-05, "loss": 2.3972, "step": 2027 }, { "epoch": 0.05, "learning_rate": 1.9979099588918244e-05, "loss": 2.4119, "step": 2028 }, { "epoch": 0.05, "learning_rate": 1.9979047969148926e-05, "loss": 2.56, "step": 2029 }, { "epoch": 0.05, "learning_rate": 1.997899628577989e-05, "loss": 2.4644, "step": 2030 }, { "epoch": 0.05, "learning_rate": 1.9978944538811454e-05, "loss": 2.5632, "step": 2031 }, { "epoch": 0.05, "learning_rate": 1.9978892728243953e-05, "loss": 2.5235, "step": 2032 }, { "epoch": 0.05, "learning_rate": 1.997884085407772e-05, "loss": 2.3571, "step": 2033 }, { "epoch": 0.05, "learning_rate": 1.997878891631308e-05, "loss": 2.5838, "step": 2034 }, { "epoch": 0.05, "learning_rate": 1.9978736914950364e-05, "loss": 2.273, "step": 2035 }, { "epoch": 0.05, "learning_rate": 1.9978684849989907e-05, "loss": 2.3891, "step": 2036 }, { "epoch": 0.05, "learning_rate": 1.9978632721432042e-05, "loss": 2.5089, "step": 2037 }, { "epoch": 0.05, "learning_rate": 1.9978580529277096e-05, "loss": 2.3975, "step": 2038 }, { "epoch": 0.05, "learning_rate": 1.9978528273525402e-05, "loss": 2.3398, "step": 2039 }, { "epoch": 0.05, "learning_rate": 1.9978475954177303e-05, "loss": 2.527, "step": 2040 }, { "epoch": 0.05, "learning_rate": 1.997842357123312e-05, "loss": 2.6457, "step": 2041 }, { "epoch": 0.05, "learning_rate": 1.9978371124693193e-05, "loss": 2.4876, "step": 2042 }, { "epoch": 0.05, "learning_rate": 1.997831861455785e-05, "loss": 2.4089, "step": 2043 }, { "epoch": 0.05, "learning_rate": 1.9978266040827437e-05, "loss": 2.5895, "step": 2044 }, { "epoch": 0.05, "learning_rate": 1.997821340350228e-05, "loss": 2.3113, "step": 2045 }, { "epoch": 0.05, "learning_rate": 1.9978160702582722e-05, "loss": 2.1949, "step": 2046 }, { "epoch": 0.05, "learning_rate": 1.9978107938069087e-05, "loss": 2.436, "step": 2047 }, { "epoch": 0.05, "learning_rate": 1.997805510996172e-05, "loss": 2.532, "step": 2048 }, { "epoch": 0.05, "learning_rate": 1.9978002218260956e-05, "loss": 2.3739, "step": 2049 }, { "epoch": 0.05, "learning_rate": 1.9977949262967136e-05, "loss": 2.368, "step": 2050 }, { "epoch": 0.05, "learning_rate": 1.997789624408059e-05, "loss": 2.4017, "step": 2051 }, { "epoch": 0.05, "learning_rate": 1.997784316160166e-05, "loss": 2.4731, "step": 2052 }, { "epoch": 0.05, "learning_rate": 1.9977790015530684e-05, "loss": 2.6119, "step": 2053 }, { "epoch": 0.05, "learning_rate": 1.9977736805868e-05, "loss": 2.5459, "step": 2054 }, { "epoch": 0.05, "learning_rate": 1.9977683532613948e-05, "loss": 2.2893, "step": 2055 }, { "epoch": 0.05, "learning_rate": 1.997763019576887e-05, "loss": 2.4133, "step": 2056 }, { "epoch": 0.05, "learning_rate": 1.99775767953331e-05, "loss": 2.3549, "step": 2057 }, { "epoch": 0.05, "learning_rate": 1.9977523331306982e-05, "loss": 2.4206, "step": 2058 }, { "epoch": 0.05, "learning_rate": 1.997746980369086e-05, "loss": 2.4579, "step": 2059 }, { "epoch": 0.05, "learning_rate": 1.9977416212485066e-05, "loss": 2.521, "step": 2060 }, { "epoch": 0.05, "learning_rate": 1.9977362557689947e-05, "loss": 2.5446, "step": 2061 }, { "epoch": 0.05, "learning_rate": 1.9977308839305846e-05, "loss": 2.479, "step": 2062 }, { "epoch": 0.05, "learning_rate": 1.997725505733311e-05, "loss": 2.4366, "step": 2063 }, { "epoch": 0.05, "learning_rate": 1.997720121177207e-05, "loss": 2.57, "step": 2064 }, { "epoch": 0.05, "learning_rate": 1.9977147302623074e-05, "loss": 2.4659, "step": 2065 }, { "epoch": 0.05, "learning_rate": 1.997709332988647e-05, "loss": 2.5366, "step": 2066 }, { "epoch": 0.05, "learning_rate": 1.9977039293562597e-05, "loss": 2.3933, "step": 2067 }, { "epoch": 0.05, "learning_rate": 1.99769851936518e-05, "loss": 2.5238, "step": 2068 }, { "epoch": 0.05, "learning_rate": 1.997693103015443e-05, "loss": 2.6411, "step": 2069 }, { "epoch": 0.05, "learning_rate": 1.9976876803070825e-05, "loss": 2.6619, "step": 2070 }, { "epoch": 0.05, "learning_rate": 1.997682251240133e-05, "loss": 2.4575, "step": 2071 }, { "epoch": 0.05, "learning_rate": 1.9976768158146297e-05, "loss": 2.5614, "step": 2072 }, { "epoch": 0.05, "learning_rate": 1.9976713740306068e-05, "loss": 2.3725, "step": 2073 }, { "epoch": 0.05, "learning_rate": 1.997665925888099e-05, "loss": 2.5126, "step": 2074 }, { "epoch": 0.05, "learning_rate": 1.997660471387141e-05, "loss": 2.5548, "step": 2075 }, { "epoch": 0.05, "learning_rate": 1.9976550105277682e-05, "loss": 2.5048, "step": 2076 }, { "epoch": 0.05, "learning_rate": 1.9976495433100144e-05, "loss": 2.3046, "step": 2077 }, { "epoch": 0.05, "learning_rate": 1.997644069733915e-05, "loss": 2.5298, "step": 2078 }, { "epoch": 0.05, "learning_rate": 1.9976385897995053e-05, "loss": 2.5188, "step": 2079 }, { "epoch": 0.05, "learning_rate": 1.9976331035068193e-05, "loss": 2.4736, "step": 2080 }, { "epoch": 0.05, "learning_rate": 1.9976276108558922e-05, "loss": 2.5755, "step": 2081 }, { "epoch": 0.05, "learning_rate": 1.9976221118467598e-05, "loss": 2.4852, "step": 2082 }, { "epoch": 0.05, "learning_rate": 1.997616606479456e-05, "loss": 2.4683, "step": 2083 }, { "epoch": 0.05, "learning_rate": 1.997611094754017e-05, "loss": 2.2947, "step": 2084 }, { "epoch": 0.05, "learning_rate": 1.997605576670477e-05, "loss": 2.3787, "step": 2085 }, { "epoch": 0.05, "learning_rate": 1.9976000522288718e-05, "loss": 2.5914, "step": 2086 }, { "epoch": 0.05, "learning_rate": 1.997594521429236e-05, "loss": 2.4676, "step": 2087 }, { "epoch": 0.05, "learning_rate": 1.9975889842716057e-05, "loss": 2.4757, "step": 2088 }, { "epoch": 0.05, "learning_rate": 1.9975834407560154e-05, "loss": 2.4674, "step": 2089 }, { "epoch": 0.05, "learning_rate": 1.9975778908825006e-05, "loss": 2.4768, "step": 2090 }, { "epoch": 0.05, "learning_rate": 1.9975723346510972e-05, "loss": 2.4004, "step": 2091 }, { "epoch": 0.05, "learning_rate": 1.9975667720618398e-05, "loss": 2.5371, "step": 2092 }, { "epoch": 0.05, "learning_rate": 1.9975612031147643e-05, "loss": 2.387, "step": 2093 }, { "epoch": 0.05, "learning_rate": 1.9975556278099065e-05, "loss": 2.355, "step": 2094 }, { "epoch": 0.05, "learning_rate": 1.997550046147301e-05, "loss": 2.4901, "step": 2095 }, { "epoch": 0.05, "learning_rate": 1.9975444581269844e-05, "loss": 2.5151, "step": 2096 }, { "epoch": 0.05, "learning_rate": 1.997538863748992e-05, "loss": 2.5117, "step": 2097 }, { "epoch": 0.05, "learning_rate": 1.9975332630133592e-05, "loss": 2.4263, "step": 2098 }, { "epoch": 0.05, "learning_rate": 1.9975276559201217e-05, "loss": 2.3646, "step": 2099 }, { "epoch": 0.05, "learning_rate": 1.9975220424693154e-05, "loss": 2.4587, "step": 2100 }, { "epoch": 0.05, "learning_rate": 1.9975164226609763e-05, "loss": 2.4163, "step": 2101 }, { "epoch": 0.05, "learning_rate": 1.9975107964951396e-05, "loss": 2.3495, "step": 2102 }, { "epoch": 0.05, "learning_rate": 1.997505163971842e-05, "loss": 2.4709, "step": 2103 }, { "epoch": 0.05, "learning_rate": 1.9974995250911184e-05, "loss": 2.3493, "step": 2104 }, { "epoch": 0.05, "learning_rate": 1.9974938798530053e-05, "loss": 2.5485, "step": 2105 }, { "epoch": 0.05, "learning_rate": 1.997488228257539e-05, "loss": 2.2638, "step": 2106 }, { "epoch": 0.05, "learning_rate": 1.9974825703047547e-05, "loss": 2.5106, "step": 2107 }, { "epoch": 0.05, "learning_rate": 1.997476905994689e-05, "loss": 2.6089, "step": 2108 }, { "epoch": 0.05, "learning_rate": 1.9974712353273783e-05, "loss": 2.3998, "step": 2109 }, { "epoch": 0.05, "learning_rate": 1.997465558302858e-05, "loss": 2.3565, "step": 2110 }, { "epoch": 0.05, "learning_rate": 1.9974598749211648e-05, "loss": 2.3897, "step": 2111 }, { "epoch": 0.05, "learning_rate": 1.9974541851823344e-05, "loss": 2.3518, "step": 2112 }, { "epoch": 0.05, "learning_rate": 1.997448489086404e-05, "loss": 2.5026, "step": 2113 }, { "epoch": 0.05, "learning_rate": 1.9974427866334087e-05, "loss": 2.3636, "step": 2114 }, { "epoch": 0.05, "learning_rate": 1.997437077823386e-05, "loss": 2.5834, "step": 2115 }, { "epoch": 0.05, "learning_rate": 1.9974313626563714e-05, "loss": 2.4909, "step": 2116 }, { "epoch": 0.05, "learning_rate": 1.9974256411324016e-05, "loss": 2.4676, "step": 2117 }, { "epoch": 0.05, "learning_rate": 1.9974199132515137e-05, "loss": 2.2061, "step": 2118 }, { "epoch": 0.05, "learning_rate": 1.997414179013743e-05, "loss": 2.393, "step": 2119 }, { "epoch": 0.05, "learning_rate": 1.997408438419127e-05, "loss": 2.3264, "step": 2120 }, { "epoch": 0.05, "learning_rate": 1.997402691467702e-05, "loss": 2.3092, "step": 2121 }, { "epoch": 0.05, "learning_rate": 1.9973969381595045e-05, "loss": 2.4195, "step": 2122 }, { "epoch": 0.05, "learning_rate": 1.9973911784945712e-05, "loss": 2.5877, "step": 2123 }, { "epoch": 0.05, "learning_rate": 1.997385412472939e-05, "loss": 2.5384, "step": 2124 }, { "epoch": 0.05, "learning_rate": 1.9973796400946445e-05, "loss": 2.4009, "step": 2125 }, { "epoch": 0.05, "learning_rate": 1.9973738613597246e-05, "loss": 2.5247, "step": 2126 }, { "epoch": 0.05, "learning_rate": 1.997368076268216e-05, "loss": 2.3991, "step": 2127 }, { "epoch": 0.05, "learning_rate": 1.9973622848201556e-05, "loss": 2.3966, "step": 2128 }, { "epoch": 0.05, "learning_rate": 1.99735648701558e-05, "loss": 2.5451, "step": 2129 }, { "epoch": 0.05, "learning_rate": 1.997350682854527e-05, "loss": 2.5274, "step": 2130 }, { "epoch": 0.05, "learning_rate": 1.997344872337033e-05, "loss": 2.5421, "step": 2131 }, { "epoch": 0.05, "learning_rate": 1.997339055463135e-05, "loss": 2.466, "step": 2132 }, { "epoch": 0.05, "learning_rate": 1.99733323223287e-05, "loss": 2.5714, "step": 2133 }, { "epoch": 0.05, "learning_rate": 1.9973274026462756e-05, "loss": 2.437, "step": 2134 }, { "epoch": 0.05, "learning_rate": 1.9973215667033884e-05, "loss": 2.4416, "step": 2135 }, { "epoch": 0.05, "learning_rate": 1.9973157244042457e-05, "loss": 2.3741, "step": 2136 }, { "epoch": 0.05, "learning_rate": 1.997309875748885e-05, "loss": 2.435, "step": 2137 }, { "epoch": 0.05, "learning_rate": 1.9973040207373438e-05, "loss": 2.3697, "step": 2138 }, { "epoch": 0.05, "learning_rate": 1.997298159369659e-05, "loss": 2.425, "step": 2139 }, { "epoch": 0.05, "learning_rate": 1.9972922916458675e-05, "loss": 2.4899, "step": 2140 }, { "epoch": 0.05, "learning_rate": 1.9972864175660076e-05, "loss": 2.5062, "step": 2141 }, { "epoch": 0.05, "learning_rate": 1.997280537130116e-05, "loss": 2.4637, "step": 2142 }, { "epoch": 0.05, "learning_rate": 1.997274650338231e-05, "loss": 2.4601, "step": 2143 }, { "epoch": 0.05, "learning_rate": 1.9972687571903897e-05, "loss": 2.3245, "step": 2144 }, { "epoch": 0.05, "learning_rate": 1.9972628576866292e-05, "loss": 2.5509, "step": 2145 }, { "epoch": 0.05, "learning_rate": 1.9972569518269876e-05, "loss": 2.4078, "step": 2146 }, { "epoch": 0.05, "learning_rate": 1.9972510396115025e-05, "loss": 2.6183, "step": 2147 }, { "epoch": 0.05, "learning_rate": 1.997245121040212e-05, "loss": 2.5397, "step": 2148 }, { "epoch": 0.05, "learning_rate": 1.997239196113153e-05, "loss": 2.6345, "step": 2149 }, { "epoch": 0.05, "learning_rate": 1.9972332648303632e-05, "loss": 2.3894, "step": 2150 }, { "epoch": 0.05, "learning_rate": 1.997227327191881e-05, "loss": 2.4292, "step": 2151 }, { "epoch": 0.05, "learning_rate": 1.9972213831977445e-05, "loss": 2.4471, "step": 2152 }, { "epoch": 0.05, "learning_rate": 1.9972154328479908e-05, "loss": 2.5191, "step": 2153 }, { "epoch": 0.05, "learning_rate": 1.997209476142658e-05, "loss": 2.2442, "step": 2154 }, { "epoch": 0.05, "learning_rate": 1.9972035130817846e-05, "loss": 2.4914, "step": 2155 }, { "epoch": 0.05, "learning_rate": 1.997197543665408e-05, "loss": 2.5001, "step": 2156 }, { "epoch": 0.05, "learning_rate": 1.9971915678935666e-05, "loss": 2.5265, "step": 2157 }, { "epoch": 0.05, "learning_rate": 1.997185585766298e-05, "loss": 2.4467, "step": 2158 }, { "epoch": 0.05, "learning_rate": 1.9971795972836416e-05, "loss": 2.5371, "step": 2159 }, { "epoch": 0.05, "learning_rate": 1.9971736024456337e-05, "loss": 2.4309, "step": 2160 }, { "epoch": 0.05, "learning_rate": 1.997167601252314e-05, "loss": 2.1932, "step": 2161 }, { "epoch": 0.05, "learning_rate": 1.99716159370372e-05, "loss": 2.4251, "step": 2162 }, { "epoch": 0.05, "learning_rate": 1.9971555797998902e-05, "loss": 2.3604, "step": 2163 }, { "epoch": 0.05, "learning_rate": 1.997149559540863e-05, "loss": 2.3573, "step": 2164 }, { "epoch": 0.05, "learning_rate": 1.9971435329266768e-05, "loss": 2.3152, "step": 2165 }, { "epoch": 0.05, "learning_rate": 1.9971374999573693e-05, "loss": 2.2996, "step": 2166 }, { "epoch": 0.05, "learning_rate": 1.9971314606329804e-05, "loss": 2.3473, "step": 2167 }, { "epoch": 0.05, "learning_rate": 1.997125414953547e-05, "loss": 2.4362, "step": 2168 }, { "epoch": 0.05, "learning_rate": 1.9971193629191086e-05, "loss": 2.5142, "step": 2169 }, { "epoch": 0.05, "learning_rate": 1.997113304529704e-05, "loss": 2.469, "step": 2170 }, { "epoch": 0.05, "learning_rate": 1.997107239785371e-05, "loss": 2.5402, "step": 2171 }, { "epoch": 0.05, "learning_rate": 1.9971011686861486e-05, "loss": 2.543, "step": 2172 }, { "epoch": 0.05, "learning_rate": 1.997095091232076e-05, "loss": 2.6279, "step": 2173 }, { "epoch": 0.05, "learning_rate": 1.9970890074231906e-05, "loss": 2.402, "step": 2174 }, { "epoch": 0.05, "learning_rate": 1.9970829172595324e-05, "loss": 2.3656, "step": 2175 }, { "epoch": 0.05, "learning_rate": 1.99707682074114e-05, "loss": 2.455, "step": 2176 }, { "epoch": 0.05, "learning_rate": 1.9970707178680517e-05, "loss": 2.5015, "step": 2177 }, { "epoch": 0.05, "learning_rate": 1.997064608640307e-05, "loss": 2.549, "step": 2178 }, { "epoch": 0.05, "learning_rate": 1.9970584930579447e-05, "loss": 2.4805, "step": 2179 }, { "epoch": 0.05, "learning_rate": 1.9970523711210036e-05, "loss": 2.5349, "step": 2180 }, { "epoch": 0.05, "learning_rate": 1.9970462428295228e-05, "loss": 2.4358, "step": 2181 }, { "epoch": 0.05, "learning_rate": 1.9970401081835414e-05, "loss": 2.5686, "step": 2182 }, { "epoch": 0.05, "learning_rate": 1.9970339671830983e-05, "loss": 2.459, "step": 2183 }, { "epoch": 0.05, "learning_rate": 1.9970278198282333e-05, "loss": 2.414, "step": 2184 }, { "epoch": 0.05, "learning_rate": 1.9970216661189844e-05, "loss": 2.3363, "step": 2185 }, { "epoch": 0.05, "learning_rate": 1.997015506055392e-05, "loss": 2.6125, "step": 2186 }, { "epoch": 0.05, "learning_rate": 1.9970093396374944e-05, "loss": 2.7667, "step": 2187 }, { "epoch": 0.05, "learning_rate": 1.997003166865332e-05, "loss": 2.4303, "step": 2188 }, { "epoch": 0.05, "learning_rate": 1.996996987738943e-05, "loss": 2.462, "step": 2189 }, { "epoch": 0.05, "learning_rate": 1.9969908022583676e-05, "loss": 2.4909, "step": 2190 }, { "epoch": 0.05, "learning_rate": 1.9969846104236446e-05, "loss": 2.5135, "step": 2191 }, { "epoch": 0.05, "learning_rate": 1.996978412234814e-05, "loss": 2.4271, "step": 2192 }, { "epoch": 0.05, "learning_rate": 1.996972207691915e-05, "loss": 2.5641, "step": 2193 }, { "epoch": 0.05, "learning_rate": 1.996965996794987e-05, "loss": 2.367, "step": 2194 }, { "epoch": 0.05, "learning_rate": 1.9969597795440704e-05, "loss": 2.4388, "step": 2195 }, { "epoch": 0.05, "learning_rate": 1.9969535559392036e-05, "loss": 2.5159, "step": 2196 }, { "epoch": 0.05, "learning_rate": 1.9969473259804272e-05, "loss": 2.3555, "step": 2197 }, { "epoch": 0.05, "learning_rate": 1.9969410896677805e-05, "loss": 2.4606, "step": 2198 }, { "epoch": 0.05, "learning_rate": 1.9969348470013036e-05, "loss": 2.5282, "step": 2199 }, { "epoch": 0.05, "learning_rate": 1.9969285979810357e-05, "loss": 2.5636, "step": 2200 }, { "epoch": 0.05, "learning_rate": 1.9969223426070168e-05, "loss": 2.3119, "step": 2201 }, { "epoch": 0.05, "learning_rate": 1.996916080879287e-05, "loss": 2.3538, "step": 2202 }, { "epoch": 0.05, "learning_rate": 1.9969098127978865e-05, "loss": 2.2887, "step": 2203 }, { "epoch": 0.05, "learning_rate": 1.9969035383628547e-05, "loss": 2.6115, "step": 2204 }, { "epoch": 0.05, "learning_rate": 1.996897257574232e-05, "loss": 2.2777, "step": 2205 }, { "epoch": 0.05, "learning_rate": 1.9968909704320577e-05, "loss": 2.5996, "step": 2206 }, { "epoch": 0.05, "learning_rate": 1.9968846769363726e-05, "loss": 2.5101, "step": 2207 }, { "epoch": 0.05, "learning_rate": 1.9968783770872167e-05, "loss": 2.311, "step": 2208 }, { "epoch": 0.05, "learning_rate": 1.99687207088463e-05, "loss": 2.5061, "step": 2209 }, { "epoch": 0.05, "learning_rate": 1.9968657583286525e-05, "loss": 2.4921, "step": 2210 }, { "epoch": 0.05, "learning_rate": 1.9968594394193248e-05, "loss": 2.4455, "step": 2211 }, { "epoch": 0.05, "learning_rate": 1.996853114156687e-05, "loss": 2.483, "step": 2212 }, { "epoch": 0.05, "learning_rate": 1.9968467825407796e-05, "loss": 2.3688, "step": 2213 }, { "epoch": 0.05, "learning_rate": 1.9968404445716426e-05, "loss": 2.6913, "step": 2214 }, { "epoch": 0.05, "learning_rate": 1.9968341002493166e-05, "loss": 2.5572, "step": 2215 }, { "epoch": 0.05, "learning_rate": 1.996827749573842e-05, "loss": 2.4626, "step": 2216 }, { "epoch": 0.05, "learning_rate": 1.9968213925452594e-05, "loss": 2.4811, "step": 2217 }, { "epoch": 0.05, "learning_rate": 1.9968150291636093e-05, "loss": 2.5309, "step": 2218 }, { "epoch": 0.05, "learning_rate": 1.9968086594289318e-05, "loss": 2.4192, "step": 2219 }, { "epoch": 0.05, "learning_rate": 1.996802283341268e-05, "loss": 2.4465, "step": 2220 }, { "epoch": 0.05, "learning_rate": 1.9967959009006586e-05, "loss": 2.5487, "step": 2221 }, { "epoch": 0.05, "learning_rate": 1.9967895121071442e-05, "loss": 2.5086, "step": 2222 }, { "epoch": 0.05, "learning_rate": 1.9967831169607648e-05, "loss": 2.3709, "step": 2223 }, { "epoch": 0.05, "learning_rate": 1.9967767154615624e-05, "loss": 2.5862, "step": 2224 }, { "epoch": 0.05, "learning_rate": 1.9967703076095766e-05, "loss": 2.579, "step": 2225 }, { "epoch": 0.05, "learning_rate": 1.9967638934048495e-05, "loss": 2.2975, "step": 2226 }, { "epoch": 0.05, "learning_rate": 1.9967574728474207e-05, "loss": 2.4497, "step": 2227 }, { "epoch": 0.05, "learning_rate": 1.996751045937332e-05, "loss": 2.514, "step": 2228 }, { "epoch": 0.05, "learning_rate": 1.9967446126746236e-05, "loss": 2.3987, "step": 2229 }, { "epoch": 0.05, "learning_rate": 1.9967381730593374e-05, "loss": 2.4973, "step": 2230 }, { "epoch": 0.05, "learning_rate": 1.9967317270915137e-05, "loss": 2.4029, "step": 2231 }, { "epoch": 0.06, "learning_rate": 1.996725274771194e-05, "loss": 2.3276, "step": 2232 }, { "epoch": 0.06, "learning_rate": 1.9967188160984193e-05, "loss": 2.3617, "step": 2233 }, { "epoch": 0.06, "learning_rate": 1.9967123510732305e-05, "loss": 2.5004, "step": 2234 }, { "epoch": 0.06, "learning_rate": 1.9967058796956697e-05, "loss": 2.4065, "step": 2235 }, { "epoch": 0.06, "learning_rate": 1.9966994019657772e-05, "loss": 2.3997, "step": 2236 }, { "epoch": 0.06, "learning_rate": 1.9966929178835946e-05, "loss": 2.3478, "step": 2237 }, { "epoch": 0.06, "learning_rate": 1.996686427449163e-05, "loss": 2.3587, "step": 2238 }, { "epoch": 0.06, "learning_rate": 1.9966799306625243e-05, "loss": 2.4152, "step": 2239 }, { "epoch": 0.06, "learning_rate": 1.9966734275237193e-05, "loss": 2.4456, "step": 2240 }, { "epoch": 0.06, "learning_rate": 1.99666691803279e-05, "loss": 2.3969, "step": 2241 }, { "epoch": 0.06, "learning_rate": 1.9966604021897774e-05, "loss": 2.5358, "step": 2242 }, { "epoch": 0.06, "learning_rate": 1.9966538799947233e-05, "loss": 2.48, "step": 2243 }, { "epoch": 0.06, "learning_rate": 1.9966473514476694e-05, "loss": 2.5373, "step": 2244 }, { "epoch": 0.06, "learning_rate": 1.996640816548657e-05, "loss": 2.5059, "step": 2245 }, { "epoch": 0.06, "learning_rate": 1.9966342752977278e-05, "loss": 2.4301, "step": 2246 }, { "epoch": 0.06, "learning_rate": 1.9966277276949236e-05, "loss": 2.5089, "step": 2247 }, { "epoch": 0.06, "learning_rate": 1.9966211737402864e-05, "loss": 2.5216, "step": 2248 }, { "epoch": 0.06, "learning_rate": 1.9966146134338573e-05, "loss": 2.4444, "step": 2249 }, { "epoch": 0.06, "learning_rate": 1.9966080467756788e-05, "loss": 2.5815, "step": 2250 }, { "epoch": 0.06, "learning_rate": 1.996601473765792e-05, "loss": 2.4168, "step": 2251 }, { "epoch": 0.06, "learning_rate": 1.9965948944042397e-05, "loss": 2.5155, "step": 2252 }, { "epoch": 0.06, "learning_rate": 1.996588308691063e-05, "loss": 2.3382, "step": 2253 }, { "epoch": 0.06, "learning_rate": 1.9965817166263048e-05, "loss": 2.3928, "step": 2254 }, { "epoch": 0.06, "learning_rate": 1.996575118210006e-05, "loss": 2.5048, "step": 2255 }, { "epoch": 0.06, "learning_rate": 1.9965685134422092e-05, "loss": 2.5401, "step": 2256 }, { "epoch": 0.06, "learning_rate": 1.9965619023229566e-05, "loss": 2.4024, "step": 2257 }, { "epoch": 0.06, "learning_rate": 1.99655528485229e-05, "loss": 2.4217, "step": 2258 }, { "epoch": 0.06, "learning_rate": 1.996548661030252e-05, "loss": 2.4584, "step": 2259 }, { "epoch": 0.06, "learning_rate": 1.996542030856885e-05, "loss": 2.2867, "step": 2260 }, { "epoch": 0.06, "learning_rate": 1.9965353943322303e-05, "loss": 2.5463, "step": 2261 }, { "epoch": 0.06, "learning_rate": 1.996528751456331e-05, "loss": 2.2346, "step": 2262 }, { "epoch": 0.06, "learning_rate": 1.9965221022292295e-05, "loss": 2.4548, "step": 2263 }, { "epoch": 0.06, "learning_rate": 1.9965154466509675e-05, "loss": 2.3482, "step": 2264 }, { "epoch": 0.06, "learning_rate": 1.9965087847215877e-05, "loss": 2.5405, "step": 2265 }, { "epoch": 0.06, "learning_rate": 1.9965021164411328e-05, "loss": 2.3402, "step": 2266 }, { "epoch": 0.06, "learning_rate": 1.9964954418096452e-05, "loss": 2.3657, "step": 2267 }, { "epoch": 0.06, "learning_rate": 1.9964887608271673e-05, "loss": 2.5135, "step": 2268 }, { "epoch": 0.06, "learning_rate": 1.9964820734937424e-05, "loss": 2.5308, "step": 2269 }, { "epoch": 0.06, "learning_rate": 1.9964753798094115e-05, "loss": 2.3998, "step": 2270 }, { "epoch": 0.06, "learning_rate": 1.996468679774219e-05, "loss": 2.4227, "step": 2271 }, { "epoch": 0.06, "learning_rate": 1.996461973388207e-05, "loss": 2.3646, "step": 2272 }, { "epoch": 0.06, "learning_rate": 1.996455260651417e-05, "loss": 2.4328, "step": 2273 }, { "epoch": 0.06, "learning_rate": 1.996448541563894e-05, "loss": 2.4727, "step": 2274 }, { "epoch": 0.06, "learning_rate": 1.996441816125679e-05, "loss": 2.3823, "step": 2275 }, { "epoch": 0.06, "learning_rate": 1.9964350843368162e-05, "loss": 2.4351, "step": 2276 }, { "epoch": 0.06, "learning_rate": 1.9964283461973475e-05, "loss": 2.572, "step": 2277 }, { "epoch": 0.06, "learning_rate": 1.9964216017073164e-05, "loss": 2.3126, "step": 2278 }, { "epoch": 0.06, "learning_rate": 1.9964148508667654e-05, "loss": 2.3995, "step": 2279 }, { "epoch": 0.06, "learning_rate": 1.996408093675738e-05, "loss": 2.4557, "step": 2280 }, { "epoch": 0.06, "learning_rate": 1.9964013301342772e-05, "loss": 2.3954, "step": 2281 }, { "epoch": 0.06, "learning_rate": 1.996394560242426e-05, "loss": 2.3844, "step": 2282 }, { "epoch": 0.06, "learning_rate": 1.9963877840002276e-05, "loss": 2.384, "step": 2283 }, { "epoch": 0.06, "learning_rate": 1.9963810014077248e-05, "loss": 2.5267, "step": 2284 }, { "epoch": 0.06, "learning_rate": 1.9963742124649617e-05, "loss": 2.5784, "step": 2285 }, { "epoch": 0.06, "learning_rate": 1.9963674171719808e-05, "loss": 2.2859, "step": 2286 }, { "epoch": 0.06, "learning_rate": 1.9963606155288254e-05, "loss": 2.4472, "step": 2287 }, { "epoch": 0.06, "learning_rate": 1.9963538075355392e-05, "loss": 2.4946, "step": 2288 }, { "epoch": 0.06, "learning_rate": 1.9963469931921656e-05, "loss": 2.4975, "step": 2289 }, { "epoch": 0.06, "learning_rate": 1.996340172498748e-05, "loss": 2.4636, "step": 2290 }, { "epoch": 0.06, "learning_rate": 1.99633334545533e-05, "loss": 2.4668, "step": 2291 }, { "epoch": 0.06, "learning_rate": 1.996326512061954e-05, "loss": 2.4457, "step": 2292 }, { "epoch": 0.06, "learning_rate": 1.9963196723186652e-05, "loss": 2.4088, "step": 2293 }, { "epoch": 0.06, "learning_rate": 1.996312826225506e-05, "loss": 2.2263, "step": 2294 }, { "epoch": 0.06, "learning_rate": 1.996305973782521e-05, "loss": 2.4827, "step": 2295 }, { "epoch": 0.06, "learning_rate": 1.996299114989753e-05, "loss": 2.6027, "step": 2296 }, { "epoch": 0.06, "learning_rate": 1.996292249847246e-05, "loss": 2.5027, "step": 2297 }, { "epoch": 0.06, "learning_rate": 1.9962853783550436e-05, "loss": 2.6513, "step": 2298 }, { "epoch": 0.06, "learning_rate": 1.9962785005131905e-05, "loss": 2.4869, "step": 2299 }, { "epoch": 0.06, "learning_rate": 1.9962716163217294e-05, "loss": 2.4049, "step": 2300 }, { "epoch": 0.06, "learning_rate": 1.9962647257807047e-05, "loss": 2.5132, "step": 2301 }, { "epoch": 0.06, "learning_rate": 1.99625782889016e-05, "loss": 2.2972, "step": 2302 }, { "epoch": 0.06, "learning_rate": 1.9962509256501396e-05, "loss": 2.3903, "step": 2303 }, { "epoch": 0.06, "learning_rate": 1.9962440160606876e-05, "loss": 2.3178, "step": 2304 }, { "epoch": 0.06, "learning_rate": 1.996237100121848e-05, "loss": 2.4901, "step": 2305 }, { "epoch": 0.06, "learning_rate": 1.996230177833664e-05, "loss": 2.6105, "step": 2306 }, { "epoch": 0.06, "learning_rate": 1.996223249196181e-05, "loss": 2.2782, "step": 2307 }, { "epoch": 0.06, "learning_rate": 1.996216314209442e-05, "loss": 2.3101, "step": 2308 }, { "epoch": 0.06, "learning_rate": 1.9962093728734924e-05, "loss": 2.4025, "step": 2309 }, { "epoch": 0.06, "learning_rate": 1.9962024251883755e-05, "loss": 2.6065, "step": 2310 }, { "epoch": 0.06, "learning_rate": 1.996195471154136e-05, "loss": 2.3726, "step": 2311 }, { "epoch": 0.06, "learning_rate": 1.996188510770818e-05, "loss": 2.5544, "step": 2312 }, { "epoch": 0.06, "learning_rate": 1.996181544038466e-05, "loss": 2.5545, "step": 2313 }, { "epoch": 0.06, "learning_rate": 1.9961745709571242e-05, "loss": 2.5447, "step": 2314 }, { "epoch": 0.06, "learning_rate": 1.9961675915268377e-05, "loss": 2.5291, "step": 2315 }, { "epoch": 0.06, "learning_rate": 1.99616060574765e-05, "loss": 2.5561, "step": 2316 }, { "epoch": 0.06, "learning_rate": 1.9961536136196064e-05, "loss": 2.4137, "step": 2317 }, { "epoch": 0.06, "learning_rate": 1.996146615142751e-05, "loss": 2.4177, "step": 2318 }, { "epoch": 0.06, "learning_rate": 1.996139610317129e-05, "loss": 2.6189, "step": 2319 }, { "epoch": 0.06, "learning_rate": 1.9961325991427843e-05, "loss": 2.443, "step": 2320 }, { "epoch": 0.06, "learning_rate": 1.996125581619762e-05, "loss": 2.4075, "step": 2321 }, { "epoch": 0.06, "learning_rate": 1.9961185577481067e-05, "loss": 2.3313, "step": 2322 }, { "epoch": 0.06, "learning_rate": 1.9961115275278632e-05, "loss": 2.5175, "step": 2323 }, { "epoch": 0.06, "learning_rate": 1.996104490959076e-05, "loss": 2.4572, "step": 2324 }, { "epoch": 0.06, "learning_rate": 1.9960974480417907e-05, "loss": 2.2784, "step": 2325 }, { "epoch": 0.06, "learning_rate": 1.9960903987760514e-05, "loss": 2.4018, "step": 2326 }, { "epoch": 0.06, "learning_rate": 1.996083343161904e-05, "loss": 2.4754, "step": 2327 }, { "epoch": 0.06, "learning_rate": 1.9960762811993925e-05, "loss": 2.3874, "step": 2328 }, { "epoch": 0.06, "learning_rate": 1.996069212888562e-05, "loss": 2.4821, "step": 2329 }, { "epoch": 0.06, "learning_rate": 1.996062138229458e-05, "loss": 2.5449, "step": 2330 }, { "epoch": 0.06, "learning_rate": 1.9960550572221252e-05, "loss": 2.21, "step": 2331 }, { "epoch": 0.06, "learning_rate": 1.996047969866609e-05, "loss": 2.5342, "step": 2332 }, { "epoch": 0.06, "learning_rate": 1.9960408761629547e-05, "loss": 2.3823, "step": 2333 }, { "epoch": 0.06, "learning_rate": 1.996033776111207e-05, "loss": 2.388, "step": 2334 }, { "epoch": 0.06, "learning_rate": 1.9960266697114116e-05, "loss": 2.4997, "step": 2335 }, { "epoch": 0.06, "learning_rate": 1.9960195569636137e-05, "loss": 2.2856, "step": 2336 }, { "epoch": 0.06, "learning_rate": 1.9960124378678586e-05, "loss": 2.5602, "step": 2337 }, { "epoch": 0.06, "learning_rate": 1.9960053124241912e-05, "loss": 2.612, "step": 2338 }, { "epoch": 0.06, "learning_rate": 1.995998180632658e-05, "loss": 2.6453, "step": 2339 }, { "epoch": 0.06, "learning_rate": 1.9959910424933034e-05, "loss": 2.4463, "step": 2340 }, { "epoch": 0.06, "learning_rate": 1.9959838980061734e-05, "loss": 2.5098, "step": 2341 }, { "epoch": 0.06, "learning_rate": 1.9959767471713138e-05, "loss": 2.4461, "step": 2342 }, { "epoch": 0.06, "learning_rate": 1.9959695899887692e-05, "loss": 2.5266, "step": 2343 }, { "epoch": 0.06, "learning_rate": 1.995962426458586e-05, "loss": 2.5838, "step": 2344 }, { "epoch": 0.06, "learning_rate": 1.9959552565808098e-05, "loss": 2.3408, "step": 2345 }, { "epoch": 0.06, "learning_rate": 1.9959480803554864e-05, "loss": 2.4559, "step": 2346 }, { "epoch": 0.06, "learning_rate": 1.9959408977826607e-05, "loss": 2.3169, "step": 2347 }, { "epoch": 0.06, "learning_rate": 1.9959337088623795e-05, "loss": 2.1711, "step": 2348 }, { "epoch": 0.06, "learning_rate": 1.9959265135946882e-05, "loss": 2.3928, "step": 2349 }, { "epoch": 0.06, "learning_rate": 1.9959193119796328e-05, "loss": 2.5503, "step": 2350 }, { "epoch": 0.06, "learning_rate": 1.9959121040172588e-05, "loss": 2.3125, "step": 2351 }, { "epoch": 0.06, "learning_rate": 1.9959048897076122e-05, "loss": 2.4866, "step": 2352 }, { "epoch": 0.06, "learning_rate": 1.9958976690507395e-05, "loss": 2.3847, "step": 2353 }, { "epoch": 0.06, "learning_rate": 1.9958904420466864e-05, "loss": 2.5374, "step": 2354 }, { "epoch": 0.06, "learning_rate": 1.995883208695499e-05, "loss": 2.4161, "step": 2355 }, { "epoch": 0.06, "learning_rate": 1.9958759689972233e-05, "loss": 2.5094, "step": 2356 }, { "epoch": 0.06, "learning_rate": 1.9958687229519054e-05, "loss": 2.5967, "step": 2357 }, { "epoch": 0.06, "learning_rate": 1.9958614705595918e-05, "loss": 2.3151, "step": 2358 }, { "epoch": 0.06, "learning_rate": 1.9958542118203282e-05, "loss": 2.6164, "step": 2359 }, { "epoch": 0.06, "learning_rate": 1.9958469467341613e-05, "loss": 2.612, "step": 2360 }, { "epoch": 0.06, "learning_rate": 1.9958396753011376e-05, "loss": 2.3888, "step": 2361 }, { "epoch": 0.06, "learning_rate": 1.9958323975213028e-05, "loss": 2.3635, "step": 2362 }, { "epoch": 0.06, "learning_rate": 1.9958251133947037e-05, "loss": 2.6267, "step": 2363 }, { "epoch": 0.06, "learning_rate": 1.9958178229213866e-05, "loss": 2.4901, "step": 2364 }, { "epoch": 0.06, "learning_rate": 1.9958105261013977e-05, "loss": 2.5222, "step": 2365 }, { "epoch": 0.06, "learning_rate": 1.995803222934784e-05, "loss": 2.3596, "step": 2366 }, { "epoch": 0.06, "learning_rate": 1.995795913421592e-05, "loss": 2.507, "step": 2367 }, { "epoch": 0.06, "learning_rate": 1.995788597561868e-05, "loss": 2.3691, "step": 2368 }, { "epoch": 0.06, "learning_rate": 1.995781275355659e-05, "loss": 2.3254, "step": 2369 }, { "epoch": 0.06, "learning_rate": 1.9957739468030112e-05, "loss": 2.4794, "step": 2370 }, { "epoch": 0.06, "learning_rate": 1.9957666119039714e-05, "loss": 2.4218, "step": 2371 }, { "epoch": 0.06, "learning_rate": 1.9957592706585864e-05, "loss": 2.5488, "step": 2372 }, { "epoch": 0.06, "learning_rate": 1.9957519230669032e-05, "loss": 2.4827, "step": 2373 }, { "epoch": 0.06, "learning_rate": 1.9957445691289686e-05, "loss": 2.4531, "step": 2374 }, { "epoch": 0.06, "learning_rate": 1.995737208844829e-05, "loss": 2.4646, "step": 2375 }, { "epoch": 0.06, "learning_rate": 1.995729842214532e-05, "loss": 2.4997, "step": 2376 }, { "epoch": 0.06, "learning_rate": 1.9957224692381243e-05, "loss": 2.3956, "step": 2377 }, { "epoch": 0.06, "learning_rate": 1.9957150899156525e-05, "loss": 2.6118, "step": 2378 }, { "epoch": 0.06, "learning_rate": 1.995707704247164e-05, "loss": 2.4463, "step": 2379 }, { "epoch": 0.06, "learning_rate": 1.9957003122327062e-05, "loss": 2.2598, "step": 2380 }, { "epoch": 0.06, "learning_rate": 1.9956929138723256e-05, "loss": 2.4027, "step": 2381 }, { "epoch": 0.06, "learning_rate": 1.9956855091660693e-05, "loss": 2.5323, "step": 2382 }, { "epoch": 0.06, "learning_rate": 1.995678098113985e-05, "loss": 2.4023, "step": 2383 }, { "epoch": 0.06, "learning_rate": 1.9956706807161197e-05, "loss": 2.3653, "step": 2384 }, { "epoch": 0.06, "learning_rate": 1.9956632569725205e-05, "loss": 2.4489, "step": 2385 }, { "epoch": 0.06, "learning_rate": 1.995655826883235e-05, "loss": 2.4176, "step": 2386 }, { "epoch": 0.06, "learning_rate": 1.9956483904483104e-05, "loss": 2.3683, "step": 2387 }, { "epoch": 0.06, "learning_rate": 1.9956409476677944e-05, "loss": 2.4022, "step": 2388 }, { "epoch": 0.06, "learning_rate": 1.9956334985417338e-05, "loss": 2.1839, "step": 2389 }, { "epoch": 0.06, "learning_rate": 1.9956260430701767e-05, "loss": 2.2988, "step": 2390 }, { "epoch": 0.06, "learning_rate": 1.9956185812531702e-05, "loss": 2.373, "step": 2391 }, { "epoch": 0.06, "learning_rate": 1.995611113090762e-05, "loss": 2.486, "step": 2392 }, { "epoch": 0.06, "learning_rate": 1.995603638583e-05, "loss": 2.5486, "step": 2393 }, { "epoch": 0.06, "learning_rate": 1.9955961577299313e-05, "loss": 2.6137, "step": 2394 }, { "epoch": 0.06, "learning_rate": 1.995588670531604e-05, "loss": 2.3567, "step": 2395 }, { "epoch": 0.06, "learning_rate": 1.9955811769880653e-05, "loss": 2.2618, "step": 2396 }, { "epoch": 0.06, "learning_rate": 1.9955736770993635e-05, "loss": 2.5468, "step": 2397 }, { "epoch": 0.06, "learning_rate": 1.9955661708655464e-05, "loss": 2.3423, "step": 2398 }, { "epoch": 0.06, "learning_rate": 1.9955586582866614e-05, "loss": 2.2088, "step": 2399 }, { "epoch": 0.06, "learning_rate": 1.995551139362757e-05, "loss": 2.5635, "step": 2400 }, { "epoch": 0.06, "learning_rate": 1.9955436140938805e-05, "loss": 2.436, "step": 2401 }, { "epoch": 0.06, "learning_rate": 1.9955360824800798e-05, "loss": 2.326, "step": 2402 }, { "epoch": 0.06, "learning_rate": 1.9955285445214037e-05, "loss": 2.3426, "step": 2403 }, { "epoch": 0.06, "learning_rate": 1.9955210002178995e-05, "loss": 2.5018, "step": 2404 }, { "epoch": 0.06, "learning_rate": 1.9955134495696157e-05, "loss": 2.6177, "step": 2405 }, { "epoch": 0.06, "learning_rate": 1.9955058925766e-05, "loss": 2.5518, "step": 2406 }, { "epoch": 0.06, "learning_rate": 1.995498329238901e-05, "loss": 2.3989, "step": 2407 }, { "epoch": 0.06, "learning_rate": 1.9954907595565667e-05, "loss": 2.5082, "step": 2408 }, { "epoch": 0.06, "learning_rate": 1.9954831835296454e-05, "loss": 2.4482, "step": 2409 }, { "epoch": 0.06, "learning_rate": 1.995475601158185e-05, "loss": 2.461, "step": 2410 }, { "epoch": 0.06, "learning_rate": 1.9954680124422346e-05, "loss": 2.1398, "step": 2411 }, { "epoch": 0.06, "learning_rate": 1.9954604173818423e-05, "loss": 2.3854, "step": 2412 }, { "epoch": 0.06, "learning_rate": 1.995452815977056e-05, "loss": 2.4172, "step": 2413 }, { "epoch": 0.06, "learning_rate": 1.9954452082279244e-05, "loss": 2.6076, "step": 2414 }, { "epoch": 0.06, "learning_rate": 1.9954375941344965e-05, "loss": 2.5812, "step": 2415 }, { "epoch": 0.06, "learning_rate": 1.99542997369682e-05, "loss": 2.4369, "step": 2416 }, { "epoch": 0.06, "learning_rate": 1.995422346914944e-05, "loss": 2.3038, "step": 2417 }, { "epoch": 0.06, "learning_rate": 1.9954147137889172e-05, "loss": 2.5195, "step": 2418 }, { "epoch": 0.06, "learning_rate": 1.9954070743187877e-05, "loss": 2.3101, "step": 2419 }, { "epoch": 0.06, "learning_rate": 1.9953994285046047e-05, "loss": 2.5157, "step": 2420 }, { "epoch": 0.06, "learning_rate": 1.9953917763464167e-05, "loss": 2.5613, "step": 2421 }, { "epoch": 0.06, "learning_rate": 1.9953841178442728e-05, "loss": 2.5258, "step": 2422 }, { "epoch": 0.06, "learning_rate": 1.995376452998221e-05, "loss": 2.5689, "step": 2423 }, { "epoch": 0.06, "learning_rate": 1.9953687818083112e-05, "loss": 2.3872, "step": 2424 }, { "epoch": 0.06, "learning_rate": 1.9953611042745912e-05, "loss": 2.5104, "step": 2425 }, { "epoch": 0.06, "learning_rate": 1.995353420397111e-05, "loss": 2.3977, "step": 2426 }, { "epoch": 0.06, "learning_rate": 1.9953457301759188e-05, "loss": 2.6027, "step": 2427 }, { "epoch": 0.06, "learning_rate": 1.9953380336110642e-05, "loss": 2.3131, "step": 2428 }, { "epoch": 0.06, "learning_rate": 1.9953303307025956e-05, "loss": 2.5119, "step": 2429 }, { "epoch": 0.06, "learning_rate": 1.9953226214505624e-05, "loss": 2.2171, "step": 2430 }, { "epoch": 0.06, "learning_rate": 1.9953149058550144e-05, "loss": 2.3293, "step": 2431 }, { "epoch": 0.06, "learning_rate": 1.9953071839159994e-05, "loss": 2.3959, "step": 2432 }, { "epoch": 0.06, "learning_rate": 1.9952994556335675e-05, "loss": 2.4729, "step": 2433 }, { "epoch": 0.06, "learning_rate": 1.9952917210077683e-05, "loss": 2.4399, "step": 2434 }, { "epoch": 0.06, "learning_rate": 1.99528398003865e-05, "loss": 2.3292, "step": 2435 }, { "epoch": 0.06, "learning_rate": 1.995276232726263e-05, "loss": 2.4588, "step": 2436 }, { "epoch": 0.06, "learning_rate": 1.995268479070656e-05, "loss": 2.3684, "step": 2437 }, { "epoch": 0.06, "learning_rate": 1.9952607190718787e-05, "loss": 2.2467, "step": 2438 }, { "epoch": 0.06, "learning_rate": 1.99525295272998e-05, "loss": 2.5225, "step": 2439 }, { "epoch": 0.06, "learning_rate": 1.9952451800450108e-05, "loss": 2.4597, "step": 2440 }, { "epoch": 0.06, "learning_rate": 1.9952374010170194e-05, "loss": 2.4517, "step": 2441 }, { "epoch": 0.06, "learning_rate": 1.9952296156460553e-05, "loss": 2.6624, "step": 2442 }, { "epoch": 0.06, "learning_rate": 1.995221823932169e-05, "loss": 2.3518, "step": 2443 }, { "epoch": 0.06, "learning_rate": 1.9952140258754092e-05, "loss": 2.4899, "step": 2444 }, { "epoch": 0.06, "learning_rate": 1.9952062214758263e-05, "loss": 2.505, "step": 2445 }, { "epoch": 0.06, "learning_rate": 1.9951984107334698e-05, "loss": 2.4752, "step": 2446 }, { "epoch": 0.06, "learning_rate": 1.9951905936483897e-05, "loss": 2.3082, "step": 2447 }, { "epoch": 0.06, "learning_rate": 1.9951827702206358e-05, "loss": 2.7838, "step": 2448 }, { "epoch": 0.06, "learning_rate": 1.995174940450257e-05, "loss": 2.4975, "step": 2449 }, { "epoch": 0.06, "learning_rate": 1.9951671043373045e-05, "loss": 2.3205, "step": 2450 }, { "epoch": 0.06, "learning_rate": 1.995159261881828e-05, "loss": 2.4338, "step": 2451 }, { "epoch": 0.06, "learning_rate": 1.995151413083877e-05, "loss": 2.2859, "step": 2452 }, { "epoch": 0.06, "learning_rate": 1.9951435579435015e-05, "loss": 2.6502, "step": 2453 }, { "epoch": 0.06, "learning_rate": 1.9951356964607517e-05, "loss": 2.5425, "step": 2454 }, { "epoch": 0.06, "learning_rate": 1.995127828635678e-05, "loss": 2.3422, "step": 2455 }, { "epoch": 0.06, "learning_rate": 1.9951199544683307e-05, "loss": 2.7338, "step": 2456 }, { "epoch": 0.06, "learning_rate": 1.995112073958759e-05, "loss": 2.2796, "step": 2457 }, { "epoch": 0.06, "learning_rate": 1.995104187107014e-05, "loss": 2.3475, "step": 2458 }, { "epoch": 0.06, "learning_rate": 1.995096293913146e-05, "loss": 2.2934, "step": 2459 }, { "epoch": 0.06, "learning_rate": 1.9950883943772047e-05, "loss": 2.2437, "step": 2460 }, { "epoch": 0.06, "learning_rate": 1.995080488499241e-05, "loss": 2.3769, "step": 2461 }, { "epoch": 0.06, "learning_rate": 1.995072576279305e-05, "loss": 2.4804, "step": 2462 }, { "epoch": 0.06, "learning_rate": 1.9950646577174472e-05, "loss": 2.2799, "step": 2463 }, { "epoch": 0.06, "learning_rate": 1.9950567328137177e-05, "loss": 2.3927, "step": 2464 }, { "epoch": 0.06, "learning_rate": 1.995048801568168e-05, "loss": 2.4239, "step": 2465 }, { "epoch": 0.06, "learning_rate": 1.9950408639808478e-05, "loss": 2.2429, "step": 2466 }, { "epoch": 0.06, "learning_rate": 1.9950329200518076e-05, "loss": 2.4122, "step": 2467 }, { "epoch": 0.06, "learning_rate": 1.9950249697810985e-05, "loss": 2.5327, "step": 2468 }, { "epoch": 0.06, "learning_rate": 1.995017013168771e-05, "loss": 2.4052, "step": 2469 }, { "epoch": 0.06, "learning_rate": 1.9950090502148755e-05, "loss": 2.3769, "step": 2470 }, { "epoch": 0.06, "learning_rate": 1.9950010809194635e-05, "loss": 2.5836, "step": 2471 }, { "epoch": 0.06, "learning_rate": 1.9949931052825852e-05, "loss": 2.3638, "step": 2472 }, { "epoch": 0.06, "learning_rate": 1.9949851233042917e-05, "loss": 2.4491, "step": 2473 }, { "epoch": 0.06, "learning_rate": 1.9949771349846336e-05, "loss": 2.4127, "step": 2474 }, { "epoch": 0.06, "learning_rate": 1.994969140323662e-05, "loss": 2.5614, "step": 2475 }, { "epoch": 0.06, "learning_rate": 1.994961139321428e-05, "loss": 2.4748, "step": 2476 }, { "epoch": 0.06, "learning_rate": 1.994953131977982e-05, "loss": 2.4388, "step": 2477 }, { "epoch": 0.06, "learning_rate": 1.9949451182933755e-05, "loss": 2.4108, "step": 2478 }, { "epoch": 0.06, "learning_rate": 1.99493709826766e-05, "loss": 2.5487, "step": 2479 }, { "epoch": 0.06, "learning_rate": 1.994929071900886e-05, "loss": 2.3537, "step": 2480 }, { "epoch": 0.06, "learning_rate": 1.994921039193104e-05, "loss": 2.5024, "step": 2481 }, { "epoch": 0.06, "learning_rate": 1.994913000144367e-05, "loss": 2.6025, "step": 2482 }, { "epoch": 0.06, "learning_rate": 1.994904954754725e-05, "loss": 2.4231, "step": 2483 }, { "epoch": 0.06, "learning_rate": 1.994896903024229e-05, "loss": 2.627, "step": 2484 }, { "epoch": 0.06, "learning_rate": 1.9948888449529315e-05, "loss": 2.2861, "step": 2485 }, { "epoch": 0.06, "learning_rate": 1.9948807805408824e-05, "loss": 2.4629, "step": 2486 }, { "epoch": 0.06, "learning_rate": 1.9948727097881343e-05, "loss": 2.3918, "step": 2487 }, { "epoch": 0.06, "learning_rate": 1.9948646326947382e-05, "loss": 2.496, "step": 2488 }, { "epoch": 0.06, "learning_rate": 1.9948565492607455e-05, "loss": 2.352, "step": 2489 }, { "epoch": 0.06, "learning_rate": 1.9948484594862077e-05, "loss": 2.3113, "step": 2490 }, { "epoch": 0.06, "learning_rate": 1.9948403633711767e-05, "loss": 2.3308, "step": 2491 }, { "epoch": 0.06, "learning_rate": 1.9948322609157036e-05, "loss": 2.1463, "step": 2492 }, { "epoch": 0.06, "learning_rate": 1.9948241521198406e-05, "loss": 2.4318, "step": 2493 }, { "epoch": 0.06, "learning_rate": 1.9948160369836388e-05, "loss": 2.4802, "step": 2494 }, { "epoch": 0.06, "learning_rate": 1.9948079155071504e-05, "loss": 2.5069, "step": 2495 }, { "epoch": 0.06, "learning_rate": 1.9947997876904267e-05, "loss": 2.3826, "step": 2496 }, { "epoch": 0.06, "learning_rate": 1.99479165353352e-05, "loss": 2.3331, "step": 2497 }, { "epoch": 0.06, "learning_rate": 1.994783513036482e-05, "loss": 2.4696, "step": 2498 }, { "epoch": 0.06, "learning_rate": 1.994775366199364e-05, "loss": 2.4065, "step": 2499 }, { "epoch": 0.06, "learning_rate": 1.9947672130222187e-05, "loss": 2.3197, "step": 2500 }, { "epoch": 0.06, "learning_rate": 1.9947590535050978e-05, "loss": 2.4443, "step": 2501 }, { "epoch": 0.06, "learning_rate": 1.9947508876480532e-05, "loss": 2.4138, "step": 2502 }, { "epoch": 0.06, "learning_rate": 1.994742715451137e-05, "loss": 2.6175, "step": 2503 }, { "epoch": 0.06, "learning_rate": 1.9947345369144014e-05, "loss": 2.5236, "step": 2504 }, { "epoch": 0.06, "learning_rate": 1.9947263520378984e-05, "loss": 2.3287, "step": 2505 }, { "epoch": 0.06, "learning_rate": 1.99471816082168e-05, "loss": 2.4453, "step": 2506 }, { "epoch": 0.06, "learning_rate": 1.994709963265799e-05, "loss": 2.2658, "step": 2507 }, { "epoch": 0.06, "learning_rate": 1.994701759370307e-05, "loss": 2.3746, "step": 2508 }, { "epoch": 0.06, "learning_rate": 1.9946935491352564e-05, "loss": 2.4138, "step": 2509 }, { "epoch": 0.06, "learning_rate": 1.9946853325606995e-05, "loss": 2.2662, "step": 2510 }, { "epoch": 0.06, "learning_rate": 1.9946771096466893e-05, "loss": 2.5632, "step": 2511 }, { "epoch": 0.06, "learning_rate": 1.9946688803932774e-05, "loss": 2.3655, "step": 2512 }, { "epoch": 0.06, "learning_rate": 1.9946606448005168e-05, "loss": 2.4134, "step": 2513 }, { "epoch": 0.06, "learning_rate": 1.9946524028684596e-05, "loss": 2.4412, "step": 2514 }, { "epoch": 0.06, "learning_rate": 1.9946441545971584e-05, "loss": 2.3495, "step": 2515 }, { "epoch": 0.06, "learning_rate": 1.994635899986666e-05, "loss": 2.3635, "step": 2516 }, { "epoch": 0.06, "learning_rate": 1.994627639037035e-05, "loss": 2.5635, "step": 2517 }, { "epoch": 0.06, "learning_rate": 1.9946193717483178e-05, "loss": 2.4287, "step": 2518 }, { "epoch": 0.06, "learning_rate": 1.994611098120567e-05, "loss": 2.3618, "step": 2519 }, { "epoch": 0.06, "learning_rate": 1.9946028181538356e-05, "loss": 2.4565, "step": 2520 }, { "epoch": 0.06, "learning_rate": 1.9945945318481764e-05, "loss": 2.5377, "step": 2521 }, { "epoch": 0.06, "learning_rate": 1.9945862392036424e-05, "loss": 2.4765, "step": 2522 }, { "epoch": 0.06, "learning_rate": 1.994577940220286e-05, "loss": 2.4572, "step": 2523 }, { "epoch": 0.06, "learning_rate": 1.99456963489816e-05, "loss": 2.3886, "step": 2524 }, { "epoch": 0.06, "learning_rate": 1.9945613232373175e-05, "loss": 2.3818, "step": 2525 }, { "epoch": 0.06, "learning_rate": 1.9945530052378118e-05, "loss": 2.3166, "step": 2526 }, { "epoch": 0.06, "learning_rate": 1.9945446808996956e-05, "loss": 2.4972, "step": 2527 }, { "epoch": 0.06, "learning_rate": 1.9945363502230225e-05, "loss": 2.4801, "step": 2528 }, { "epoch": 0.06, "learning_rate": 1.9945280132078446e-05, "loss": 2.4266, "step": 2529 }, { "epoch": 0.06, "learning_rate": 1.994519669854216e-05, "loss": 2.2692, "step": 2530 }, { "epoch": 0.06, "learning_rate": 1.994511320162189e-05, "loss": 2.4416, "step": 2531 }, { "epoch": 0.06, "learning_rate": 1.9945029641318174e-05, "loss": 2.3447, "step": 2532 }, { "epoch": 0.06, "learning_rate": 1.9944946017631545e-05, "loss": 2.6419, "step": 2533 }, { "epoch": 0.06, "learning_rate": 1.994486233056253e-05, "loss": 2.3012, "step": 2534 }, { "epoch": 0.06, "learning_rate": 1.994477858011167e-05, "loss": 2.558, "step": 2535 }, { "epoch": 0.06, "learning_rate": 1.9944694766279492e-05, "loss": 2.2951, "step": 2536 }, { "epoch": 0.06, "learning_rate": 1.9944610889066537e-05, "loss": 2.4876, "step": 2537 }, { "epoch": 0.06, "learning_rate": 1.9944526948473335e-05, "loss": 2.4157, "step": 2538 }, { "epoch": 0.06, "learning_rate": 1.9944442944500422e-05, "loss": 2.4981, "step": 2539 }, { "epoch": 0.06, "learning_rate": 1.994435887714833e-05, "loss": 2.3824, "step": 2540 }, { "epoch": 0.06, "learning_rate": 1.9944274746417604e-05, "loss": 2.5073, "step": 2541 }, { "epoch": 0.06, "learning_rate": 1.9944190552308772e-05, "loss": 2.2943, "step": 2542 }, { "epoch": 0.06, "learning_rate": 1.994410629482237e-05, "loss": 2.3884, "step": 2543 }, { "epoch": 0.06, "learning_rate": 1.994402197395894e-05, "loss": 2.6293, "step": 2544 }, { "epoch": 0.06, "learning_rate": 1.9943937589719017e-05, "loss": 2.4278, "step": 2545 }, { "epoch": 0.06, "learning_rate": 1.994385314210314e-05, "loss": 2.3638, "step": 2546 }, { "epoch": 0.06, "learning_rate": 1.9943768631111845e-05, "loss": 2.4462, "step": 2547 }, { "epoch": 0.06, "learning_rate": 1.994368405674567e-05, "loss": 2.4645, "step": 2548 }, { "epoch": 0.06, "learning_rate": 1.9943599419005163e-05, "loss": 2.2308, "step": 2549 }, { "epoch": 0.06, "learning_rate": 1.994351471789085e-05, "loss": 2.5393, "step": 2550 }, { "epoch": 0.06, "learning_rate": 1.9943429953403276e-05, "loss": 2.5718, "step": 2551 }, { "epoch": 0.06, "learning_rate": 1.9943345125542987e-05, "loss": 2.4171, "step": 2552 }, { "epoch": 0.06, "learning_rate": 1.994326023431052e-05, "loss": 2.4057, "step": 2553 }, { "epoch": 0.06, "learning_rate": 1.994317527970641e-05, "loss": 2.5057, "step": 2554 }, { "epoch": 0.06, "learning_rate": 1.9943090261731207e-05, "loss": 2.3756, "step": 2555 }, { "epoch": 0.06, "learning_rate": 1.994300518038545e-05, "loss": 2.4166, "step": 2556 }, { "epoch": 0.06, "learning_rate": 1.994292003566968e-05, "loss": 2.2767, "step": 2557 }, { "epoch": 0.06, "learning_rate": 1.994283482758444e-05, "loss": 2.4425, "step": 2558 }, { "epoch": 0.06, "learning_rate": 1.994274955613027e-05, "loss": 2.3505, "step": 2559 }, { "epoch": 0.06, "learning_rate": 1.994266422130772e-05, "loss": 2.5026, "step": 2560 }, { "epoch": 0.06, "learning_rate": 1.9942578823117332e-05, "loss": 2.3233, "step": 2561 }, { "epoch": 0.06, "learning_rate": 1.9942493361559644e-05, "loss": 2.3543, "step": 2562 }, { "epoch": 0.06, "learning_rate": 1.994240783663521e-05, "loss": 2.4629, "step": 2563 }, { "epoch": 0.06, "learning_rate": 1.9942322248344572e-05, "loss": 2.2999, "step": 2564 }, { "epoch": 0.06, "learning_rate": 1.9942236596688272e-05, "loss": 2.2035, "step": 2565 }, { "epoch": 0.06, "learning_rate": 1.9942150881666857e-05, "loss": 2.4196, "step": 2566 }, { "epoch": 0.06, "learning_rate": 1.9942065103280876e-05, "loss": 2.3296, "step": 2567 }, { "epoch": 0.06, "learning_rate": 1.9941979261530874e-05, "loss": 2.3761, "step": 2568 }, { "epoch": 0.06, "learning_rate": 1.9941893356417397e-05, "loss": 2.4485, "step": 2569 }, { "epoch": 0.06, "learning_rate": 1.9941807387940993e-05, "loss": 2.3754, "step": 2570 }, { "epoch": 0.06, "learning_rate": 1.9941721356102212e-05, "loss": 2.5591, "step": 2571 }, { "epoch": 0.06, "learning_rate": 1.9941635260901602e-05, "loss": 2.3695, "step": 2572 }, { "epoch": 0.06, "learning_rate": 1.994154910233971e-05, "loss": 2.2847, "step": 2573 }, { "epoch": 0.06, "learning_rate": 1.9941462880417085e-05, "loss": 2.4516, "step": 2574 }, { "epoch": 0.06, "learning_rate": 1.9941376595134276e-05, "loss": 2.2812, "step": 2575 }, { "epoch": 0.06, "learning_rate": 1.9941290246491837e-05, "loss": 2.5293, "step": 2576 }, { "epoch": 0.06, "learning_rate": 1.9941203834490313e-05, "loss": 2.4251, "step": 2577 }, { "epoch": 0.06, "learning_rate": 1.994111735913026e-05, "loss": 2.3972, "step": 2578 }, { "epoch": 0.06, "learning_rate": 1.9941030820412223e-05, "loss": 2.3183, "step": 2579 }, { "epoch": 0.06, "learning_rate": 1.9940944218336757e-05, "loss": 2.2324, "step": 2580 }, { "epoch": 0.06, "learning_rate": 1.9940857552904415e-05, "loss": 2.2697, "step": 2581 }, { "epoch": 0.06, "learning_rate": 1.994077082411575e-05, "loss": 2.3864, "step": 2582 }, { "epoch": 0.06, "learning_rate": 1.9940684031971313e-05, "loss": 2.4107, "step": 2583 }, { "epoch": 0.06, "learning_rate": 1.9940597176471656e-05, "loss": 2.4665, "step": 2584 }, { "epoch": 0.06, "learning_rate": 1.9940510257617333e-05, "loss": 2.4321, "step": 2585 }, { "epoch": 0.06, "learning_rate": 1.9940423275408896e-05, "loss": 2.3344, "step": 2586 }, { "epoch": 0.06, "learning_rate": 1.9940336229846905e-05, "loss": 2.5224, "step": 2587 }, { "epoch": 0.06, "learning_rate": 1.9940249120931914e-05, "loss": 2.4638, "step": 2588 }, { "epoch": 0.06, "learning_rate": 1.9940161948664473e-05, "loss": 2.332, "step": 2589 }, { "epoch": 0.06, "learning_rate": 1.994007471304514e-05, "loss": 2.2671, "step": 2590 }, { "epoch": 0.06, "learning_rate": 1.9939987414074472e-05, "loss": 2.5518, "step": 2591 }, { "epoch": 0.06, "learning_rate": 1.9939900051753024e-05, "loss": 2.4251, "step": 2592 }, { "epoch": 0.06, "learning_rate": 1.9939812626081355e-05, "loss": 2.4494, "step": 2593 }, { "epoch": 0.06, "learning_rate": 1.993972513706002e-05, "loss": 2.4114, "step": 2594 }, { "epoch": 0.06, "learning_rate": 1.993963758468958e-05, "loss": 2.4603, "step": 2595 }, { "epoch": 0.06, "learning_rate": 1.993954996897059e-05, "loss": 2.4568, "step": 2596 }, { "epoch": 0.06, "learning_rate": 1.9939462289903606e-05, "loss": 2.3874, "step": 2597 }, { "epoch": 0.06, "learning_rate": 1.993937454748919e-05, "loss": 2.4483, "step": 2598 }, { "epoch": 0.06, "learning_rate": 1.9939286741727903e-05, "loss": 2.4933, "step": 2599 }, { "epoch": 0.06, "learning_rate": 1.9939198872620303e-05, "loss": 2.39, "step": 2600 }, { "epoch": 0.06, "learning_rate": 1.9939110940166948e-05, "loss": 2.3215, "step": 2601 }, { "epoch": 0.06, "learning_rate": 1.9939022944368397e-05, "loss": 2.5798, "step": 2602 }, { "epoch": 0.06, "learning_rate": 1.993893488522522e-05, "loss": 2.5896, "step": 2603 }, { "epoch": 0.06, "learning_rate": 1.9938846762737965e-05, "loss": 2.4113, "step": 2604 }, { "epoch": 0.06, "learning_rate": 1.9938758576907203e-05, "loss": 2.6446, "step": 2605 }, { "epoch": 0.06, "learning_rate": 1.9938670327733497e-05, "loss": 2.4061, "step": 2606 }, { "epoch": 0.06, "learning_rate": 1.9938582015217404e-05, "loss": 2.5954, "step": 2607 }, { "epoch": 0.06, "learning_rate": 1.993849363935949e-05, "loss": 2.2868, "step": 2608 }, { "epoch": 0.06, "learning_rate": 1.9938405200160315e-05, "loss": 2.4743, "step": 2609 }, { "epoch": 0.06, "learning_rate": 1.9938316697620447e-05, "loss": 2.3566, "step": 2610 }, { "epoch": 0.06, "learning_rate": 1.9938228131740445e-05, "loss": 2.4934, "step": 2611 }, { "epoch": 0.06, "learning_rate": 1.9938139502520878e-05, "loss": 2.2672, "step": 2612 }, { "epoch": 0.06, "learning_rate": 1.993805080996231e-05, "loss": 2.4075, "step": 2613 }, { "epoch": 0.06, "learning_rate": 1.9937962054065304e-05, "loss": 2.4212, "step": 2614 }, { "epoch": 0.06, "learning_rate": 1.993787323483043e-05, "loss": 2.5203, "step": 2615 }, { "epoch": 0.06, "learning_rate": 1.993778435225825e-05, "loss": 2.3933, "step": 2616 }, { "epoch": 0.06, "learning_rate": 1.993769540634933e-05, "loss": 2.5104, "step": 2617 }, { "epoch": 0.06, "learning_rate": 1.9937606397104238e-05, "loss": 2.31, "step": 2618 }, { "epoch": 0.06, "learning_rate": 1.9937517324523543e-05, "loss": 2.5103, "step": 2619 }, { "epoch": 0.06, "learning_rate": 1.9937428188607813e-05, "loss": 2.4887, "step": 2620 }, { "epoch": 0.06, "learning_rate": 1.9937338989357613e-05, "loss": 2.4901, "step": 2621 }, { "epoch": 0.06, "learning_rate": 1.9937249726773514e-05, "loss": 2.414, "step": 2622 }, { "epoch": 0.06, "learning_rate": 1.9937160400856082e-05, "loss": 2.5943, "step": 2623 }, { "epoch": 0.06, "learning_rate": 1.993707101160589e-05, "loss": 2.5036, "step": 2624 }, { "epoch": 0.06, "learning_rate": 1.9936981559023504e-05, "loss": 2.4427, "step": 2625 }, { "epoch": 0.06, "learning_rate": 1.9936892043109498e-05, "loss": 2.3784, "step": 2626 }, { "epoch": 0.06, "learning_rate": 1.993680246386444e-05, "loss": 2.3951, "step": 2627 }, { "epoch": 0.06, "learning_rate": 1.9936712821288907e-05, "loss": 2.4764, "step": 2628 }, { "epoch": 0.06, "learning_rate": 1.993662311538346e-05, "loss": 2.2588, "step": 2629 }, { "epoch": 0.06, "learning_rate": 1.9936533346148676e-05, "loss": 2.6225, "step": 2630 }, { "epoch": 0.06, "learning_rate": 1.9936443513585124e-05, "loss": 2.3817, "step": 2631 }, { "epoch": 0.06, "learning_rate": 1.9936353617693385e-05, "loss": 2.4443, "step": 2632 }, { "epoch": 0.06, "learning_rate": 1.993626365847402e-05, "loss": 2.543, "step": 2633 }, { "epoch": 0.06, "learning_rate": 1.9936173635927614e-05, "loss": 2.2658, "step": 2634 }, { "epoch": 0.06, "learning_rate": 1.9936083550054733e-05, "loss": 2.4641, "step": 2635 }, { "epoch": 0.06, "learning_rate": 1.993599340085595e-05, "loss": 2.5628, "step": 2636 }, { "epoch": 0.06, "learning_rate": 1.9935903188331848e-05, "loss": 2.4072, "step": 2637 }, { "epoch": 0.07, "learning_rate": 1.9935812912482996e-05, "loss": 2.5188, "step": 2638 }, { "epoch": 0.07, "learning_rate": 1.993572257330997e-05, "loss": 2.5707, "step": 2639 }, { "epoch": 0.07, "learning_rate": 1.9935632170813345e-05, "loss": 2.3057, "step": 2640 }, { "epoch": 0.07, "learning_rate": 1.99355417049937e-05, "loss": 2.4962, "step": 2641 }, { "epoch": 0.07, "learning_rate": 1.993545117585161e-05, "loss": 2.4572, "step": 2642 }, { "epoch": 0.07, "learning_rate": 1.993536058338765e-05, "loss": 2.5581, "step": 2643 }, { "epoch": 0.07, "learning_rate": 1.99352699276024e-05, "loss": 2.2054, "step": 2644 }, { "epoch": 0.07, "learning_rate": 1.9935179208496436e-05, "loss": 2.3448, "step": 2645 }, { "epoch": 0.07, "learning_rate": 1.9935088426070338e-05, "loss": 2.339, "step": 2646 }, { "epoch": 0.07, "learning_rate": 1.9934997580324685e-05, "loss": 2.4339, "step": 2647 }, { "epoch": 0.07, "learning_rate": 1.9934906671260052e-05, "loss": 2.5245, "step": 2648 }, { "epoch": 0.07, "learning_rate": 1.9934815698877024e-05, "loss": 2.5226, "step": 2649 }, { "epoch": 0.07, "learning_rate": 1.993472466317618e-05, "loss": 2.3678, "step": 2650 }, { "epoch": 0.07, "learning_rate": 1.993463356415809e-05, "loss": 2.675, "step": 2651 }, { "epoch": 0.07, "learning_rate": 1.993454240182335e-05, "loss": 2.4102, "step": 2652 }, { "epoch": 0.07, "learning_rate": 1.993445117617253e-05, "loss": 2.3365, "step": 2653 }, { "epoch": 0.07, "learning_rate": 1.9934359887206214e-05, "loss": 2.3731, "step": 2654 }, { "epoch": 0.07, "learning_rate": 1.9934268534924985e-05, "loss": 2.3906, "step": 2655 }, { "epoch": 0.07, "learning_rate": 1.993417711932943e-05, "loss": 2.3808, "step": 2656 }, { "epoch": 0.07, "learning_rate": 1.9934085640420122e-05, "loss": 2.278, "step": 2657 }, { "epoch": 0.07, "learning_rate": 1.993399409819765e-05, "loss": 2.4403, "step": 2658 }, { "epoch": 0.07, "learning_rate": 1.9933902492662597e-05, "loss": 2.4224, "step": 2659 }, { "epoch": 0.07, "learning_rate": 1.9933810823815547e-05, "loss": 2.2392, "step": 2660 }, { "epoch": 0.07, "learning_rate": 1.9933719091657083e-05, "loss": 2.3774, "step": 2661 }, { "epoch": 0.07, "learning_rate": 1.9933627296187786e-05, "loss": 2.3116, "step": 2662 }, { "epoch": 0.07, "learning_rate": 1.9933535437408248e-05, "loss": 2.4531, "step": 2663 }, { "epoch": 0.07, "learning_rate": 1.9933443515319052e-05, "loss": 2.3092, "step": 2664 }, { "epoch": 0.07, "learning_rate": 1.9933351529920784e-05, "loss": 2.5974, "step": 2665 }, { "epoch": 0.07, "learning_rate": 1.993325948121403e-05, "loss": 2.4474, "step": 2666 }, { "epoch": 0.07, "learning_rate": 1.9933167369199372e-05, "loss": 2.5318, "step": 2667 }, { "epoch": 0.07, "learning_rate": 1.9933075193877404e-05, "loss": 2.4544, "step": 2668 }, { "epoch": 0.07, "learning_rate": 1.9932982955248712e-05, "loss": 2.4338, "step": 2669 }, { "epoch": 0.07, "learning_rate": 1.993289065331388e-05, "loss": 2.3812, "step": 2670 }, { "epoch": 0.07, "learning_rate": 1.9932798288073498e-05, "loss": 2.4401, "step": 2671 }, { "epoch": 0.07, "learning_rate": 1.9932705859528157e-05, "loss": 2.3382, "step": 2672 }, { "epoch": 0.07, "learning_rate": 1.9932613367678447e-05, "loss": 2.4128, "step": 2673 }, { "epoch": 0.07, "learning_rate": 1.993252081252495e-05, "loss": 2.4551, "step": 2674 }, { "epoch": 0.07, "learning_rate": 1.9932428194068264e-05, "loss": 2.4184, "step": 2675 }, { "epoch": 0.07, "learning_rate": 1.9932335512308977e-05, "loss": 2.47, "step": 2676 }, { "epoch": 0.07, "learning_rate": 1.993224276724768e-05, "loss": 2.4154, "step": 2677 }, { "epoch": 0.07, "learning_rate": 1.9932149958884958e-05, "loss": 2.3183, "step": 2678 }, { "epoch": 0.07, "learning_rate": 1.9932057087221413e-05, "loss": 2.4433, "step": 2679 }, { "epoch": 0.07, "learning_rate": 1.9931964152257628e-05, "loss": 2.4974, "step": 2680 }, { "epoch": 0.07, "learning_rate": 1.9931871153994197e-05, "loss": 2.3255, "step": 2681 }, { "epoch": 0.07, "learning_rate": 1.993177809243172e-05, "loss": 2.31, "step": 2682 }, { "epoch": 0.07, "learning_rate": 1.993168496757078e-05, "loss": 2.4365, "step": 2683 }, { "epoch": 0.07, "learning_rate": 1.993159177941198e-05, "loss": 2.3714, "step": 2684 }, { "epoch": 0.07, "learning_rate": 1.9931498527955902e-05, "loss": 2.5484, "step": 2685 }, { "epoch": 0.07, "learning_rate": 1.9931405213203154e-05, "loss": 2.5737, "step": 2686 }, { "epoch": 0.07, "learning_rate": 1.993131183515432e-05, "loss": 2.5332, "step": 2687 }, { "epoch": 0.07, "learning_rate": 1.993121839381e-05, "loss": 2.4716, "step": 2688 }, { "epoch": 0.07, "learning_rate": 1.993112488917079e-05, "loss": 2.5238, "step": 2689 }, { "epoch": 0.07, "learning_rate": 1.9931031321237283e-05, "loss": 2.4324, "step": 2690 }, { "epoch": 0.07, "learning_rate": 1.9930937690010075e-05, "loss": 2.4291, "step": 2691 }, { "epoch": 0.07, "learning_rate": 1.993084399548977e-05, "loss": 2.512, "step": 2692 }, { "epoch": 0.07, "learning_rate": 1.9930750237676956e-05, "loss": 2.5673, "step": 2693 }, { "epoch": 0.07, "learning_rate": 1.9930656416572235e-05, "loss": 2.3555, "step": 2694 }, { "epoch": 0.07, "learning_rate": 1.9930562532176203e-05, "loss": 2.4824, "step": 2695 }, { "epoch": 0.07, "learning_rate": 1.9930468584489462e-05, "loss": 2.4135, "step": 2696 }, { "epoch": 0.07, "learning_rate": 1.993037457351261e-05, "loss": 2.4137, "step": 2697 }, { "epoch": 0.07, "learning_rate": 1.993028049924624e-05, "loss": 2.417, "step": 2698 }, { "epoch": 0.07, "learning_rate": 1.993018636169096e-05, "loss": 2.4649, "step": 2699 }, { "epoch": 0.07, "learning_rate": 1.9930092160847364e-05, "loss": 2.5106, "step": 2700 }, { "epoch": 0.07, "learning_rate": 1.9929997896716053e-05, "loss": 2.3803, "step": 2701 }, { "epoch": 0.07, "learning_rate": 1.9929903569297632e-05, "loss": 2.4066, "step": 2702 }, { "epoch": 0.07, "learning_rate": 1.99298091785927e-05, "loss": 2.4633, "step": 2703 }, { "epoch": 0.07, "learning_rate": 1.9929714724601858e-05, "loss": 2.5346, "step": 2704 }, { "epoch": 0.07, "learning_rate": 1.9929620207325708e-05, "loss": 2.4145, "step": 2705 }, { "epoch": 0.07, "learning_rate": 1.992952562676485e-05, "loss": 2.3513, "step": 2706 }, { "epoch": 0.07, "learning_rate": 1.9929430982919893e-05, "loss": 2.2556, "step": 2707 }, { "epoch": 0.07, "learning_rate": 1.9929336275791435e-05, "loss": 2.3067, "step": 2708 }, { "epoch": 0.07, "learning_rate": 1.9929241505380078e-05, "loss": 2.5019, "step": 2709 }, { "epoch": 0.07, "learning_rate": 1.9929146671686432e-05, "loss": 2.3568, "step": 2710 }, { "epoch": 0.07, "learning_rate": 1.99290517747111e-05, "loss": 2.3247, "step": 2711 }, { "epoch": 0.07, "learning_rate": 1.992895681445468e-05, "loss": 2.6249, "step": 2712 }, { "epoch": 0.07, "learning_rate": 1.9928861790917785e-05, "loss": 2.4797, "step": 2713 }, { "epoch": 0.07, "learning_rate": 1.992876670410102e-05, "loss": 2.3749, "step": 2714 }, { "epoch": 0.07, "learning_rate": 1.9928671554004988e-05, "loss": 2.4155, "step": 2715 }, { "epoch": 0.07, "learning_rate": 1.9928576340630297e-05, "loss": 2.309, "step": 2716 }, { "epoch": 0.07, "learning_rate": 1.9928481063977553e-05, "loss": 2.5222, "step": 2717 }, { "epoch": 0.07, "learning_rate": 1.992838572404736e-05, "loss": 2.571, "step": 2718 }, { "epoch": 0.07, "learning_rate": 1.9928290320840335e-05, "loss": 2.4218, "step": 2719 }, { "epoch": 0.07, "learning_rate": 1.9928194854357075e-05, "loss": 2.3477, "step": 2720 }, { "epoch": 0.07, "learning_rate": 1.99280993245982e-05, "loss": 2.3946, "step": 2721 }, { "epoch": 0.07, "learning_rate": 1.9928003731564306e-05, "loss": 2.5039, "step": 2722 }, { "epoch": 0.07, "learning_rate": 1.9927908075256016e-05, "loss": 2.4893, "step": 2723 }, { "epoch": 0.07, "learning_rate": 1.9927812355673924e-05, "loss": 2.4952, "step": 2724 }, { "epoch": 0.07, "learning_rate": 1.9927716572818654e-05, "loss": 2.463, "step": 2725 }, { "epoch": 0.07, "learning_rate": 1.992762072669081e-05, "loss": 2.376, "step": 2726 }, { "epoch": 0.07, "learning_rate": 1.9927524817291006e-05, "loss": 2.3476, "step": 2727 }, { "epoch": 0.07, "learning_rate": 1.9927428844619843e-05, "loss": 2.2553, "step": 2728 }, { "epoch": 0.07, "learning_rate": 1.992733280867795e-05, "loss": 2.2981, "step": 2729 }, { "epoch": 0.07, "learning_rate": 1.9927236709465926e-05, "loss": 2.3667, "step": 2730 }, { "epoch": 0.07, "learning_rate": 1.9927140546984387e-05, "loss": 2.3594, "step": 2731 }, { "epoch": 0.07, "learning_rate": 1.9927044321233946e-05, "loss": 2.4173, "step": 2732 }, { "epoch": 0.07, "learning_rate": 1.9926948032215216e-05, "loss": 2.3865, "step": 2733 }, { "epoch": 0.07, "learning_rate": 1.9926851679928813e-05, "loss": 2.2365, "step": 2734 }, { "epoch": 0.07, "learning_rate": 1.9926755264375348e-05, "loss": 2.3817, "step": 2735 }, { "epoch": 0.07, "learning_rate": 1.9926658785555437e-05, "loss": 2.4097, "step": 2736 }, { "epoch": 0.07, "learning_rate": 1.9926562243469693e-05, "loss": 2.5837, "step": 2737 }, { "epoch": 0.07, "learning_rate": 1.9926465638118736e-05, "loss": 2.418, "step": 2738 }, { "epoch": 0.07, "learning_rate": 1.9926368969503173e-05, "loss": 2.4473, "step": 2739 }, { "epoch": 0.07, "learning_rate": 1.992627223762363e-05, "loss": 2.5182, "step": 2740 }, { "epoch": 0.07, "learning_rate": 1.992617544248072e-05, "loss": 2.5747, "step": 2741 }, { "epoch": 0.07, "learning_rate": 1.9926078584075057e-05, "loss": 2.3687, "step": 2742 }, { "epoch": 0.07, "learning_rate": 1.992598166240726e-05, "loss": 2.3873, "step": 2743 }, { "epoch": 0.07, "learning_rate": 1.992588467747795e-05, "loss": 2.3306, "step": 2744 }, { "epoch": 0.07, "learning_rate": 1.992578762928774e-05, "loss": 2.3577, "step": 2745 }, { "epoch": 0.07, "learning_rate": 1.9925690517837252e-05, "loss": 2.3959, "step": 2746 }, { "epoch": 0.07, "learning_rate": 1.99255933431271e-05, "loss": 2.4823, "step": 2747 }, { "epoch": 0.07, "learning_rate": 1.992549610515791e-05, "loss": 2.2306, "step": 2748 }, { "epoch": 0.07, "learning_rate": 1.9925398803930295e-05, "loss": 2.4197, "step": 2749 }, { "epoch": 0.07, "learning_rate": 1.9925301439444884e-05, "loss": 2.4396, "step": 2750 }, { "epoch": 0.07, "learning_rate": 1.992520401170229e-05, "loss": 2.4496, "step": 2751 }, { "epoch": 0.07, "learning_rate": 1.9925106520703136e-05, "loss": 2.4209, "step": 2752 }, { "epoch": 0.07, "learning_rate": 1.9925008966448044e-05, "loss": 2.413, "step": 2753 }, { "epoch": 0.07, "learning_rate": 1.9924911348937638e-05, "loss": 2.366, "step": 2754 }, { "epoch": 0.07, "learning_rate": 1.9924813668172533e-05, "loss": 2.5068, "step": 2755 }, { "epoch": 0.07, "learning_rate": 1.992471592415336e-05, "loss": 2.4377, "step": 2756 }, { "epoch": 0.07, "learning_rate": 1.9924618116880736e-05, "loss": 2.4831, "step": 2757 }, { "epoch": 0.07, "learning_rate": 1.9924520246355284e-05, "loss": 2.4464, "step": 2758 }, { "epoch": 0.07, "learning_rate": 1.9924422312577636e-05, "loss": 2.3507, "step": 2759 }, { "epoch": 0.07, "learning_rate": 1.9924324315548407e-05, "loss": 2.3709, "step": 2760 }, { "epoch": 0.07, "learning_rate": 1.9924226255268225e-05, "loss": 2.4223, "step": 2761 }, { "epoch": 0.07, "learning_rate": 1.9924128131737714e-05, "loss": 2.329, "step": 2762 }, { "epoch": 0.07, "learning_rate": 1.99240299449575e-05, "loss": 2.5062, "step": 2763 }, { "epoch": 0.07, "learning_rate": 1.992393169492821e-05, "loss": 2.2993, "step": 2764 }, { "epoch": 0.07, "learning_rate": 1.9923833381650472e-05, "loss": 2.3615, "step": 2765 }, { "epoch": 0.07, "learning_rate": 1.9923735005124908e-05, "loss": 2.4334, "step": 2766 }, { "epoch": 0.07, "learning_rate": 1.9923636565352146e-05, "loss": 2.3034, "step": 2767 }, { "epoch": 0.07, "learning_rate": 1.9923538062332813e-05, "loss": 2.3469, "step": 2768 }, { "epoch": 0.07, "learning_rate": 1.992343949606754e-05, "loss": 2.3838, "step": 2769 }, { "epoch": 0.07, "learning_rate": 1.9923340866556954e-05, "loss": 2.3995, "step": 2770 }, { "epoch": 0.07, "learning_rate": 1.9923242173801678e-05, "loss": 2.4055, "step": 2771 }, { "epoch": 0.07, "learning_rate": 1.992314341780235e-05, "loss": 2.4876, "step": 2772 }, { "epoch": 0.07, "learning_rate": 1.9923044598559592e-05, "loss": 2.2967, "step": 2773 }, { "epoch": 0.07, "learning_rate": 1.992294571607404e-05, "loss": 2.4732, "step": 2774 }, { "epoch": 0.07, "learning_rate": 1.992284677034632e-05, "loss": 2.52, "step": 2775 }, { "epoch": 0.07, "learning_rate": 1.9922747761377062e-05, "loss": 2.4732, "step": 2776 }, { "epoch": 0.07, "learning_rate": 1.99226486891669e-05, "loss": 2.3911, "step": 2777 }, { "epoch": 0.07, "learning_rate": 1.9922549553716464e-05, "loss": 2.4186, "step": 2778 }, { "epoch": 0.07, "learning_rate": 1.9922450355026387e-05, "loss": 2.5154, "step": 2779 }, { "epoch": 0.07, "learning_rate": 1.9922351093097303e-05, "loss": 2.3905, "step": 2780 }, { "epoch": 0.07, "learning_rate": 1.9922251767929838e-05, "loss": 2.4769, "step": 2781 }, { "epoch": 0.07, "learning_rate": 1.9922152379524627e-05, "loss": 2.4154, "step": 2782 }, { "epoch": 0.07, "learning_rate": 1.992205292788231e-05, "loss": 2.4188, "step": 2783 }, { "epoch": 0.07, "learning_rate": 1.9921953413003516e-05, "loss": 2.5032, "step": 2784 }, { "epoch": 0.07, "learning_rate": 1.9921853834888875e-05, "loss": 2.3742, "step": 2785 }, { "epoch": 0.07, "learning_rate": 1.992175419353903e-05, "loss": 2.4369, "step": 2786 }, { "epoch": 0.07, "learning_rate": 1.992165448895461e-05, "loss": 2.4843, "step": 2787 }, { "epoch": 0.07, "learning_rate": 1.9921554721136255e-05, "loss": 2.3747, "step": 2788 }, { "epoch": 0.07, "learning_rate": 1.9921454890084596e-05, "loss": 2.583, "step": 2789 }, { "epoch": 0.07, "learning_rate": 1.9921354995800274e-05, "loss": 2.387, "step": 2790 }, { "epoch": 0.07, "learning_rate": 1.992125503828392e-05, "loss": 2.5128, "step": 2791 }, { "epoch": 0.07, "learning_rate": 1.9921155017536176e-05, "loss": 2.4998, "step": 2792 }, { "epoch": 0.07, "learning_rate": 1.992105493355768e-05, "loss": 2.3756, "step": 2793 }, { "epoch": 0.07, "learning_rate": 1.9920954786349066e-05, "loss": 2.5111, "step": 2794 }, { "epoch": 0.07, "learning_rate": 1.9920854575910974e-05, "loss": 2.4895, "step": 2795 }, { "epoch": 0.07, "learning_rate": 1.992075430224404e-05, "loss": 2.4507, "step": 2796 }, { "epoch": 0.07, "learning_rate": 1.992065396534891e-05, "loss": 2.318, "step": 2797 }, { "epoch": 0.07, "learning_rate": 1.9920553565226216e-05, "loss": 2.3362, "step": 2798 }, { "epoch": 0.07, "learning_rate": 1.99204531018766e-05, "loss": 2.4844, "step": 2799 }, { "epoch": 0.07, "learning_rate": 1.9920352575300707e-05, "loss": 2.4167, "step": 2800 }, { "epoch": 0.07, "learning_rate": 1.9920251985499174e-05, "loss": 2.3811, "step": 2801 }, { "epoch": 0.07, "learning_rate": 1.9920151332472645e-05, "loss": 2.2845, "step": 2802 }, { "epoch": 0.07, "learning_rate": 1.9920050616221754e-05, "loss": 2.4723, "step": 2803 }, { "epoch": 0.07, "learning_rate": 1.9919949836747148e-05, "loss": 2.4497, "step": 2804 }, { "epoch": 0.07, "learning_rate": 1.9919848994049467e-05, "loss": 2.3416, "step": 2805 }, { "epoch": 0.07, "learning_rate": 1.991974808812936e-05, "loss": 2.3655, "step": 2806 }, { "epoch": 0.07, "learning_rate": 1.991964711898746e-05, "loss": 2.2119, "step": 2807 }, { "epoch": 0.07, "learning_rate": 1.991954608662442e-05, "loss": 2.3503, "step": 2808 }, { "epoch": 0.07, "learning_rate": 1.9919444991040883e-05, "loss": 2.6043, "step": 2809 }, { "epoch": 0.07, "learning_rate": 1.991934383223749e-05, "loss": 2.4436, "step": 2810 }, { "epoch": 0.07, "learning_rate": 1.991924261021488e-05, "loss": 2.3826, "step": 2811 }, { "epoch": 0.07, "learning_rate": 1.991914132497371e-05, "loss": 2.3856, "step": 2812 }, { "epoch": 0.07, "learning_rate": 1.9919039976514615e-05, "loss": 2.2775, "step": 2813 }, { "epoch": 0.07, "learning_rate": 1.991893856483825e-05, "loss": 2.4635, "step": 2814 }, { "epoch": 0.07, "learning_rate": 1.9918837089945254e-05, "loss": 2.4687, "step": 2815 }, { "epoch": 0.07, "learning_rate": 1.991873555183628e-05, "loss": 2.3693, "step": 2816 }, { "epoch": 0.07, "learning_rate": 1.9918633950511967e-05, "loss": 2.5988, "step": 2817 }, { "epoch": 0.07, "learning_rate": 1.991853228597297e-05, "loss": 2.5404, "step": 2818 }, { "epoch": 0.07, "learning_rate": 1.9918430558219938e-05, "loss": 2.4046, "step": 2819 }, { "epoch": 0.07, "learning_rate": 1.991832876725351e-05, "loss": 2.2941, "step": 2820 }, { "epoch": 0.07, "learning_rate": 1.9918226913074344e-05, "loss": 2.4495, "step": 2821 }, { "epoch": 0.07, "learning_rate": 1.9918124995683086e-05, "loss": 2.5241, "step": 2822 }, { "epoch": 0.07, "learning_rate": 1.9918023015080382e-05, "loss": 2.3863, "step": 2823 }, { "epoch": 0.07, "learning_rate": 1.991792097126689e-05, "loss": 2.4179, "step": 2824 }, { "epoch": 0.07, "learning_rate": 1.991781886424325e-05, "loss": 2.3826, "step": 2825 }, { "epoch": 0.07, "learning_rate": 1.991771669401012e-05, "loss": 2.4572, "step": 2826 }, { "epoch": 0.07, "learning_rate": 1.991761446056815e-05, "loss": 2.2764, "step": 2827 }, { "epoch": 0.07, "learning_rate": 1.9917512163917992e-05, "loss": 2.3697, "step": 2828 }, { "epoch": 0.07, "learning_rate": 1.9917409804060297e-05, "loss": 2.5279, "step": 2829 }, { "epoch": 0.07, "learning_rate": 1.9917307380995717e-05, "loss": 2.5697, "step": 2830 }, { "epoch": 0.07, "learning_rate": 1.99172048947249e-05, "loss": 2.3564, "step": 2831 }, { "epoch": 0.07, "learning_rate": 1.991710234524851e-05, "loss": 2.265, "step": 2832 }, { "epoch": 0.07, "learning_rate": 1.9916999732567195e-05, "loss": 2.451, "step": 2833 }, { "epoch": 0.07, "learning_rate": 1.991689705668161e-05, "loss": 2.5927, "step": 2834 }, { "epoch": 0.07, "learning_rate": 1.9916794317592404e-05, "loss": 2.484, "step": 2835 }, { "epoch": 0.07, "learning_rate": 1.9916691515300236e-05, "loss": 2.3679, "step": 2836 }, { "epoch": 0.07, "learning_rate": 1.9916588649805763e-05, "loss": 2.4851, "step": 2837 }, { "epoch": 0.07, "learning_rate": 1.991648572110964e-05, "loss": 2.1877, "step": 2838 }, { "epoch": 0.07, "learning_rate": 1.991638272921252e-05, "loss": 2.2428, "step": 2839 }, { "epoch": 0.07, "learning_rate": 1.9916279674115064e-05, "loss": 2.4667, "step": 2840 }, { "epoch": 0.07, "learning_rate": 1.9916176555817923e-05, "loss": 2.2843, "step": 2841 }, { "epoch": 0.07, "learning_rate": 1.991607337432176e-05, "loss": 2.4266, "step": 2842 }, { "epoch": 0.07, "learning_rate": 1.9915970129627227e-05, "loss": 2.5209, "step": 2843 }, { "epoch": 0.07, "learning_rate": 1.9915866821734984e-05, "loss": 2.4259, "step": 2844 }, { "epoch": 0.07, "learning_rate": 1.9915763450645694e-05, "loss": 2.4758, "step": 2845 }, { "epoch": 0.07, "learning_rate": 1.9915660016360008e-05, "loss": 2.485, "step": 2846 }, { "epoch": 0.07, "learning_rate": 1.9915556518878593e-05, "loss": 2.4329, "step": 2847 }, { "epoch": 0.07, "learning_rate": 1.9915452958202103e-05, "loss": 2.4336, "step": 2848 }, { "epoch": 0.07, "learning_rate": 1.99153493343312e-05, "loss": 2.479, "step": 2849 }, { "epoch": 0.07, "learning_rate": 1.9915245647266544e-05, "loss": 2.5013, "step": 2850 }, { "epoch": 0.07, "learning_rate": 1.9915141897008796e-05, "loss": 2.315, "step": 2851 }, { "epoch": 0.07, "learning_rate": 1.9915038083558617e-05, "loss": 2.5643, "step": 2852 }, { "epoch": 0.07, "learning_rate": 1.9914934206916672e-05, "loss": 2.3617, "step": 2853 }, { "epoch": 0.07, "learning_rate": 1.9914830267083617e-05, "loss": 2.22, "step": 2854 }, { "epoch": 0.07, "learning_rate": 1.991472626406012e-05, "loss": 2.3005, "step": 2855 }, { "epoch": 0.07, "learning_rate": 1.991462219784684e-05, "loss": 2.3389, "step": 2856 }, { "epoch": 0.07, "learning_rate": 1.9914518068444443e-05, "loss": 2.3225, "step": 2857 }, { "epoch": 0.07, "learning_rate": 1.991441387585359e-05, "loss": 2.3815, "step": 2858 }, { "epoch": 0.07, "learning_rate": 1.9914309620074946e-05, "loss": 2.4021, "step": 2859 }, { "epoch": 0.07, "learning_rate": 1.9914205301109178e-05, "loss": 2.4893, "step": 2860 }, { "epoch": 0.07, "learning_rate": 1.9914100918956944e-05, "loss": 2.375, "step": 2861 }, { "epoch": 0.07, "learning_rate": 1.9913996473618918e-05, "loss": 2.2503, "step": 2862 }, { "epoch": 0.07, "learning_rate": 1.9913891965095762e-05, "loss": 2.3459, "step": 2863 }, { "epoch": 0.07, "learning_rate": 1.991378739338814e-05, "loss": 2.3889, "step": 2864 }, { "epoch": 0.07, "learning_rate": 1.9913682758496723e-05, "loss": 2.3303, "step": 2865 }, { "epoch": 0.07, "learning_rate": 1.991357806042217e-05, "loss": 2.5017, "step": 2866 }, { "epoch": 0.07, "learning_rate": 1.9913473299165157e-05, "loss": 2.2523, "step": 2867 }, { "epoch": 0.07, "learning_rate": 1.9913368474726345e-05, "loss": 2.4417, "step": 2868 }, { "epoch": 0.07, "learning_rate": 1.991326358710641e-05, "loss": 2.6194, "step": 2869 }, { "epoch": 0.07, "learning_rate": 1.991315863630601e-05, "loss": 2.262, "step": 2870 }, { "epoch": 0.07, "learning_rate": 1.991305362232582e-05, "loss": 2.2347, "step": 2871 }, { "epoch": 0.07, "learning_rate": 1.9912948545166514e-05, "loss": 2.5409, "step": 2872 }, { "epoch": 0.07, "learning_rate": 1.991284340482875e-05, "loss": 2.4382, "step": 2873 }, { "epoch": 0.07, "learning_rate": 1.991273820131321e-05, "loss": 2.4119, "step": 2874 }, { "epoch": 0.07, "learning_rate": 1.9912632934620552e-05, "loss": 2.2602, "step": 2875 }, { "epoch": 0.07, "learning_rate": 1.9912527604751458e-05, "loss": 2.5064, "step": 2876 }, { "epoch": 0.07, "learning_rate": 1.9912422211706598e-05, "loss": 2.4926, "step": 2877 }, { "epoch": 0.07, "learning_rate": 1.991231675548664e-05, "loss": 2.3979, "step": 2878 }, { "epoch": 0.07, "learning_rate": 1.9912211236092252e-05, "loss": 2.2794, "step": 2879 }, { "epoch": 0.07, "learning_rate": 1.9912105653524116e-05, "loss": 2.5153, "step": 2880 }, { "epoch": 0.07, "learning_rate": 1.9912000007782896e-05, "loss": 2.4093, "step": 2881 }, { "epoch": 0.07, "learning_rate": 1.9911894298869275e-05, "loss": 2.4923, "step": 2882 }, { "epoch": 0.07, "learning_rate": 1.9911788526783917e-05, "loss": 2.1336, "step": 2883 }, { "epoch": 0.07, "learning_rate": 1.9911682691527505e-05, "loss": 2.3094, "step": 2884 }, { "epoch": 0.07, "learning_rate": 1.9911576793100708e-05, "loss": 2.4379, "step": 2885 }, { "epoch": 0.07, "learning_rate": 1.99114708315042e-05, "loss": 2.3819, "step": 2886 }, { "epoch": 0.07, "learning_rate": 1.9911364806738657e-05, "loss": 2.5368, "step": 2887 }, { "epoch": 0.07, "learning_rate": 1.9911258718804758e-05, "loss": 2.4075, "step": 2888 }, { "epoch": 0.07, "learning_rate": 1.9911152567703178e-05, "loss": 2.4109, "step": 2889 }, { "epoch": 0.07, "learning_rate": 1.9911046353434593e-05, "loss": 2.4981, "step": 2890 }, { "epoch": 0.07, "learning_rate": 1.9910940075999678e-05, "loss": 2.475, "step": 2891 }, { "epoch": 0.07, "learning_rate": 1.9910833735399114e-05, "loss": 2.3567, "step": 2892 }, { "epoch": 0.07, "learning_rate": 1.9910727331633575e-05, "loss": 2.3136, "step": 2893 }, { "epoch": 0.07, "learning_rate": 1.991062086470374e-05, "loss": 2.3937, "step": 2894 }, { "epoch": 0.07, "learning_rate": 1.9910514334610293e-05, "loss": 2.3717, "step": 2895 }, { "epoch": 0.07, "learning_rate": 1.9910407741353906e-05, "loss": 2.4696, "step": 2896 }, { "epoch": 0.07, "learning_rate": 1.9910301084935257e-05, "loss": 2.5576, "step": 2897 }, { "epoch": 0.07, "learning_rate": 1.9910194365355032e-05, "loss": 2.4434, "step": 2898 }, { "epoch": 0.07, "learning_rate": 1.9910087582613912e-05, "loss": 2.3808, "step": 2899 }, { "epoch": 0.07, "learning_rate": 1.990998073671257e-05, "loss": 2.5165, "step": 2900 }, { "epoch": 0.07, "learning_rate": 1.9909873827651695e-05, "loss": 2.203, "step": 2901 }, { "epoch": 0.07, "learning_rate": 1.990976685543196e-05, "loss": 2.3885, "step": 2902 }, { "epoch": 0.07, "learning_rate": 1.9909659820054055e-05, "loss": 2.4726, "step": 2903 }, { "epoch": 0.07, "learning_rate": 1.990955272151866e-05, "loss": 2.5721, "step": 2904 }, { "epoch": 0.07, "learning_rate": 1.990944555982645e-05, "loss": 2.4966, "step": 2905 }, { "epoch": 0.07, "learning_rate": 1.9909338334978115e-05, "loss": 2.3427, "step": 2906 }, { "epoch": 0.07, "learning_rate": 1.990923104697434e-05, "loss": 2.3411, "step": 2907 }, { "epoch": 0.07, "learning_rate": 1.990912369581581e-05, "loss": 2.3246, "step": 2908 }, { "epoch": 0.07, "learning_rate": 1.99090162815032e-05, "loss": 2.514, "step": 2909 }, { "epoch": 0.07, "learning_rate": 1.9908908804037203e-05, "loss": 2.3374, "step": 2910 }, { "epoch": 0.07, "learning_rate": 1.9908801263418496e-05, "loss": 2.4053, "step": 2911 }, { "epoch": 0.07, "learning_rate": 1.9908693659647774e-05, "loss": 2.2614, "step": 2912 }, { "epoch": 0.07, "learning_rate": 1.9908585992725714e-05, "loss": 2.4169, "step": 2913 }, { "epoch": 0.07, "learning_rate": 1.9908478262653012e-05, "loss": 2.3698, "step": 2914 }, { "epoch": 0.07, "learning_rate": 1.9908370469430344e-05, "loss": 2.4274, "step": 2915 }, { "epoch": 0.07, "learning_rate": 1.9908262613058405e-05, "loss": 2.282, "step": 2916 }, { "epoch": 0.07, "learning_rate": 1.9908154693537876e-05, "loss": 2.4524, "step": 2917 }, { "epoch": 0.07, "learning_rate": 1.9908046710869452e-05, "loss": 2.3402, "step": 2918 }, { "epoch": 0.07, "learning_rate": 1.9907938665053817e-05, "loss": 2.2566, "step": 2919 }, { "epoch": 0.07, "learning_rate": 1.990783055609166e-05, "loss": 2.4467, "step": 2920 }, { "epoch": 0.07, "learning_rate": 1.9907722383983667e-05, "loss": 2.5085, "step": 2921 }, { "epoch": 0.07, "learning_rate": 1.9907614148730533e-05, "loss": 2.479, "step": 2922 }, { "epoch": 0.07, "learning_rate": 1.9907505850332946e-05, "loss": 2.4865, "step": 2923 }, { "epoch": 0.07, "learning_rate": 1.9907397488791594e-05, "loss": 2.3966, "step": 2924 }, { "epoch": 0.07, "learning_rate": 1.990728906410717e-05, "loss": 2.4634, "step": 2925 }, { "epoch": 0.07, "learning_rate": 1.990718057628036e-05, "loss": 2.2698, "step": 2926 }, { "epoch": 0.07, "learning_rate": 1.9907072025311863e-05, "loss": 2.3042, "step": 2927 }, { "epoch": 0.07, "learning_rate": 1.9906963411202366e-05, "loss": 2.5282, "step": 2928 }, { "epoch": 0.07, "learning_rate": 1.9906854733952566e-05, "loss": 2.5439, "step": 2929 }, { "epoch": 0.07, "learning_rate": 1.990674599356315e-05, "loss": 2.3895, "step": 2930 }, { "epoch": 0.07, "learning_rate": 1.990663719003481e-05, "loss": 2.4203, "step": 2931 }, { "epoch": 0.07, "learning_rate": 1.9906528323368248e-05, "loss": 2.5616, "step": 2932 }, { "epoch": 0.07, "learning_rate": 1.990641939356415e-05, "loss": 2.3899, "step": 2933 }, { "epoch": 0.07, "learning_rate": 1.9906310400623213e-05, "loss": 2.4422, "step": 2934 }, { "epoch": 0.07, "learning_rate": 1.9906201344546134e-05, "loss": 2.3314, "step": 2935 }, { "epoch": 0.07, "learning_rate": 1.99060922253336e-05, "loss": 2.2435, "step": 2936 }, { "epoch": 0.07, "learning_rate": 1.9905983042986318e-05, "loss": 2.3735, "step": 2937 }, { "epoch": 0.07, "learning_rate": 1.9905873797504974e-05, "loss": 2.3953, "step": 2938 }, { "epoch": 0.07, "learning_rate": 1.990576448889027e-05, "loss": 2.5211, "step": 2939 }, { "epoch": 0.07, "learning_rate": 1.9905655117142897e-05, "loss": 2.3128, "step": 2940 }, { "epoch": 0.07, "learning_rate": 1.990554568226356e-05, "loss": 2.2502, "step": 2941 }, { "epoch": 0.07, "learning_rate": 1.990543618425295e-05, "loss": 2.3834, "step": 2942 }, { "epoch": 0.07, "learning_rate": 1.9905326623111766e-05, "loss": 2.3696, "step": 2943 }, { "epoch": 0.07, "learning_rate": 1.990521699884071e-05, "loss": 2.1958, "step": 2944 }, { "epoch": 0.07, "learning_rate": 1.990510731144048e-05, "loss": 2.4645, "step": 2945 }, { "epoch": 0.07, "learning_rate": 1.990499756091177e-05, "loss": 2.2428, "step": 2946 }, { "epoch": 0.07, "learning_rate": 1.990488774725528e-05, "loss": 2.2938, "step": 2947 }, { "epoch": 0.07, "learning_rate": 1.9904777870471718e-05, "loss": 2.3536, "step": 2948 }, { "epoch": 0.07, "learning_rate": 1.9904667930561776e-05, "loss": 2.378, "step": 2949 }, { "epoch": 0.07, "learning_rate": 1.9904557927526154e-05, "loss": 2.3442, "step": 2950 }, { "epoch": 0.07, "learning_rate": 1.9904447861365562e-05, "loss": 2.4045, "step": 2951 }, { "epoch": 0.07, "learning_rate": 1.9904337732080693e-05, "loss": 2.3341, "step": 2952 }, { "epoch": 0.07, "learning_rate": 1.9904227539672255e-05, "loss": 2.4935, "step": 2953 }, { "epoch": 0.07, "learning_rate": 1.9904117284140944e-05, "loss": 2.3109, "step": 2954 }, { "epoch": 0.07, "learning_rate": 1.9904006965487467e-05, "loss": 2.4961, "step": 2955 }, { "epoch": 0.07, "learning_rate": 1.990389658371253e-05, "loss": 2.307, "step": 2956 }, { "epoch": 0.07, "learning_rate": 1.9903786138816824e-05, "loss": 2.4772, "step": 2957 }, { "epoch": 0.07, "learning_rate": 1.9903675630801067e-05, "loss": 2.4312, "step": 2958 }, { "epoch": 0.07, "learning_rate": 1.9903565059665958e-05, "loss": 2.2703, "step": 2959 }, { "epoch": 0.07, "learning_rate": 1.99034544254122e-05, "loss": 2.3124, "step": 2960 }, { "epoch": 0.07, "learning_rate": 1.99033437280405e-05, "loss": 2.339, "step": 2961 }, { "epoch": 0.07, "learning_rate": 1.9903232967551563e-05, "loss": 2.5344, "step": 2962 }, { "epoch": 0.07, "learning_rate": 1.9903122143946093e-05, "loss": 2.4072, "step": 2963 }, { "epoch": 0.07, "learning_rate": 1.99030112572248e-05, "loss": 2.3419, "step": 2964 }, { "epoch": 0.07, "learning_rate": 1.9902900307388387e-05, "loss": 2.3884, "step": 2965 }, { "epoch": 0.07, "learning_rate": 1.990278929443757e-05, "loss": 2.3092, "step": 2966 }, { "epoch": 0.07, "learning_rate": 1.990267821837304e-05, "loss": 2.3608, "step": 2967 }, { "epoch": 0.07, "learning_rate": 1.9902567079195522e-05, "loss": 2.1454, "step": 2968 }, { "epoch": 0.07, "learning_rate": 1.9902455876905714e-05, "loss": 2.5874, "step": 2969 }, { "epoch": 0.07, "learning_rate": 1.9902344611504324e-05, "loss": 2.2891, "step": 2970 }, { "epoch": 0.07, "learning_rate": 1.990223328299207e-05, "loss": 2.4511, "step": 2971 }, { "epoch": 0.07, "learning_rate": 1.9902121891369655e-05, "loss": 2.5337, "step": 2972 }, { "epoch": 0.07, "learning_rate": 1.9902010436637787e-05, "loss": 2.2721, "step": 2973 }, { "epoch": 0.07, "learning_rate": 1.990189891879718e-05, "loss": 2.4957, "step": 2974 }, { "epoch": 0.07, "learning_rate": 1.990178733784855e-05, "loss": 2.5044, "step": 2975 }, { "epoch": 0.07, "learning_rate": 1.9901675693792596e-05, "loss": 2.3635, "step": 2976 }, { "epoch": 0.07, "learning_rate": 1.990156398663004e-05, "loss": 2.4766, "step": 2977 }, { "epoch": 0.07, "learning_rate": 1.990145221636159e-05, "loss": 2.3282, "step": 2978 }, { "epoch": 0.07, "learning_rate": 1.9901340382987954e-05, "loss": 2.553, "step": 2979 }, { "epoch": 0.07, "learning_rate": 1.990122848650985e-05, "loss": 2.433, "step": 2980 }, { "epoch": 0.07, "learning_rate": 1.9901116526927994e-05, "loss": 2.3545, "step": 2981 }, { "epoch": 0.07, "learning_rate": 1.9901004504243097e-05, "loss": 2.4639, "step": 2982 }, { "epoch": 0.07, "learning_rate": 1.9900892418455864e-05, "loss": 2.5455, "step": 2983 }, { "epoch": 0.07, "learning_rate": 1.9900780269567025e-05, "loss": 2.4083, "step": 2984 }, { "epoch": 0.07, "learning_rate": 1.9900668057577278e-05, "loss": 2.2829, "step": 2985 }, { "epoch": 0.07, "learning_rate": 1.9900555782487354e-05, "loss": 2.332, "step": 2986 }, { "epoch": 0.07, "learning_rate": 1.990044344429796e-05, "loss": 2.3668, "step": 2987 }, { "epoch": 0.07, "learning_rate": 1.9900331043009814e-05, "loss": 2.4685, "step": 2988 }, { "epoch": 0.07, "learning_rate": 1.9900218578623628e-05, "loss": 2.3142, "step": 2989 }, { "epoch": 0.07, "learning_rate": 1.990010605114013e-05, "loss": 2.3454, "step": 2990 }, { "epoch": 0.07, "learning_rate": 1.9899993460560023e-05, "loss": 2.4389, "step": 2991 }, { "epoch": 0.07, "learning_rate": 1.989988080688403e-05, "loss": 2.3002, "step": 2992 }, { "epoch": 0.07, "learning_rate": 1.9899768090112875e-05, "loss": 2.4589, "step": 2993 }, { "epoch": 0.07, "learning_rate": 1.9899655310247273e-05, "loss": 2.2138, "step": 2994 }, { "epoch": 0.07, "learning_rate": 1.9899542467287938e-05, "loss": 2.3913, "step": 2995 }, { "epoch": 0.07, "learning_rate": 1.989942956123559e-05, "loss": 2.4343, "step": 2996 }, { "epoch": 0.07, "learning_rate": 1.989931659209096e-05, "loss": 2.298, "step": 2997 }, { "epoch": 0.07, "learning_rate": 1.989920355985475e-05, "loss": 2.4654, "step": 2998 }, { "epoch": 0.07, "learning_rate": 1.9899090464527698e-05, "loss": 2.3214, "step": 2999 }, { "epoch": 0.07, "learning_rate": 1.9898977306110508e-05, "loss": 2.5262, "step": 3000 }, { "epoch": 0.07, "learning_rate": 1.9898864084603916e-05, "loss": 2.2876, "step": 3001 }, { "epoch": 0.07, "learning_rate": 1.9898750800008636e-05, "loss": 2.3539, "step": 3002 }, { "epoch": 0.07, "learning_rate": 1.989863745232539e-05, "loss": 2.4006, "step": 3003 }, { "epoch": 0.07, "learning_rate": 1.9898524041554903e-05, "loss": 2.476, "step": 3004 }, { "epoch": 0.07, "learning_rate": 1.9898410567697897e-05, "loss": 2.2224, "step": 3005 }, { "epoch": 0.07, "learning_rate": 1.9898297030755093e-05, "loss": 2.3866, "step": 3006 }, { "epoch": 0.07, "learning_rate": 1.9898183430727216e-05, "loss": 2.2502, "step": 3007 }, { "epoch": 0.07, "learning_rate": 1.989806976761499e-05, "loss": 2.2123, "step": 3008 }, { "epoch": 0.07, "learning_rate": 1.989795604141914e-05, "loss": 2.4025, "step": 3009 }, { "epoch": 0.07, "learning_rate": 1.9897842252140393e-05, "loss": 2.4503, "step": 3010 }, { "epoch": 0.07, "learning_rate": 1.9897728399779467e-05, "loss": 2.5254, "step": 3011 }, { "epoch": 0.07, "learning_rate": 1.9897614484337094e-05, "loss": 2.4521, "step": 3012 }, { "epoch": 0.07, "learning_rate": 1.9897500505814e-05, "loss": 2.3393, "step": 3013 }, { "epoch": 0.07, "learning_rate": 1.989738646421091e-05, "loss": 2.5089, "step": 3014 }, { "epoch": 0.07, "learning_rate": 1.989727235952855e-05, "loss": 2.2799, "step": 3015 }, { "epoch": 0.07, "learning_rate": 1.9897158191767648e-05, "loss": 2.3477, "step": 3016 }, { "epoch": 0.07, "learning_rate": 1.989704396092893e-05, "loss": 2.4233, "step": 3017 }, { "epoch": 0.07, "learning_rate": 1.989692966701313e-05, "loss": 2.3439, "step": 3018 }, { "epoch": 0.07, "learning_rate": 1.9896815310020964e-05, "loss": 2.3165, "step": 3019 }, { "epoch": 0.07, "learning_rate": 1.9896700889953173e-05, "loss": 2.3044, "step": 3020 }, { "epoch": 0.07, "learning_rate": 1.9896586406810484e-05, "loss": 2.4348, "step": 3021 }, { "epoch": 0.07, "learning_rate": 1.9896471860593622e-05, "loss": 2.4651, "step": 3022 }, { "epoch": 0.07, "learning_rate": 1.9896357251303322e-05, "loss": 2.4016, "step": 3023 }, { "epoch": 0.07, "learning_rate": 1.9896242578940307e-05, "loss": 2.4266, "step": 3024 }, { "epoch": 0.07, "learning_rate": 1.989612784350532e-05, "loss": 2.4487, "step": 3025 }, { "epoch": 0.07, "learning_rate": 1.9896013044999082e-05, "loss": 2.4464, "step": 3026 }, { "epoch": 0.07, "learning_rate": 1.9895898183422323e-05, "loss": 2.3412, "step": 3027 }, { "epoch": 0.07, "learning_rate": 1.9895783258775786e-05, "loss": 2.4915, "step": 3028 }, { "epoch": 0.07, "learning_rate": 1.98956682710602e-05, "loss": 2.3563, "step": 3029 }, { "epoch": 0.07, "learning_rate": 1.989555322027629e-05, "loss": 2.4686, "step": 3030 }, { "epoch": 0.07, "learning_rate": 1.9895438106424794e-05, "loss": 2.435, "step": 3031 }, { "epoch": 0.07, "learning_rate": 1.9895322929506447e-05, "loss": 2.4701, "step": 3032 }, { "epoch": 0.07, "learning_rate": 1.989520768952198e-05, "loss": 2.403, "step": 3033 }, { "epoch": 0.07, "learning_rate": 1.9895092386472135e-05, "loss": 2.4404, "step": 3034 }, { "epoch": 0.07, "learning_rate": 1.989497702035764e-05, "loss": 2.566, "step": 3035 }, { "epoch": 0.07, "learning_rate": 1.9894861591179227e-05, "loss": 2.3509, "step": 3036 }, { "epoch": 0.07, "learning_rate": 1.989474609893764e-05, "loss": 2.418, "step": 3037 }, { "epoch": 0.07, "learning_rate": 1.9894630543633612e-05, "loss": 2.5747, "step": 3038 }, { "epoch": 0.07, "learning_rate": 1.989451492526788e-05, "loss": 2.4614, "step": 3039 }, { "epoch": 0.07, "learning_rate": 1.989439924384118e-05, "loss": 2.3752, "step": 3040 }, { "epoch": 0.07, "learning_rate": 1.9894283499354244e-05, "loss": 2.4073, "step": 3041 }, { "epoch": 0.07, "learning_rate": 1.989416769180782e-05, "loss": 2.4612, "step": 3042 }, { "epoch": 0.08, "learning_rate": 1.989405182120264e-05, "loss": 2.3016, "step": 3043 }, { "epoch": 0.08, "learning_rate": 1.989393588753944e-05, "loss": 2.5779, "step": 3044 }, { "epoch": 0.08, "learning_rate": 1.9893819890818968e-05, "loss": 2.4505, "step": 3045 }, { "epoch": 0.08, "learning_rate": 1.989370383104195e-05, "loss": 2.292, "step": 3046 }, { "epoch": 0.08, "learning_rate": 1.989358770820914e-05, "loss": 2.3162, "step": 3047 }, { "epoch": 0.08, "learning_rate": 1.989347152232127e-05, "loss": 2.2214, "step": 3048 }, { "epoch": 0.08, "learning_rate": 1.989335527337908e-05, "loss": 2.4427, "step": 3049 }, { "epoch": 0.08, "learning_rate": 1.9893238961383315e-05, "loss": 2.4669, "step": 3050 }, { "epoch": 0.08, "learning_rate": 1.9893122586334713e-05, "loss": 2.4004, "step": 3051 }, { "epoch": 0.08, "learning_rate": 1.9893006148234012e-05, "loss": 2.3849, "step": 3052 }, { "epoch": 0.08, "learning_rate": 1.9892889647081964e-05, "loss": 2.3804, "step": 3053 }, { "epoch": 0.08, "learning_rate": 1.9892773082879307e-05, "loss": 2.4852, "step": 3054 }, { "epoch": 0.08, "learning_rate": 1.989265645562678e-05, "loss": 2.4385, "step": 3055 }, { "epoch": 0.08, "learning_rate": 1.9892539765325133e-05, "loss": 2.39, "step": 3056 }, { "epoch": 0.08, "learning_rate": 1.98924230119751e-05, "loss": 2.5016, "step": 3057 }, { "epoch": 0.08, "learning_rate": 1.9892306195577437e-05, "loss": 2.4811, "step": 3058 }, { "epoch": 0.08, "learning_rate": 1.989218931613288e-05, "loss": 2.4319, "step": 3059 }, { "epoch": 0.08, "learning_rate": 1.989207237364218e-05, "loss": 2.2413, "step": 3060 }, { "epoch": 0.08, "learning_rate": 1.9891955368106078e-05, "loss": 2.2655, "step": 3061 }, { "epoch": 0.08, "learning_rate": 1.989183829952532e-05, "loss": 2.4919, "step": 3062 }, { "epoch": 0.08, "learning_rate": 1.9891721167900652e-05, "loss": 2.4528, "step": 3063 }, { "epoch": 0.08, "learning_rate": 1.989160397323282e-05, "loss": 2.2146, "step": 3064 }, { "epoch": 0.08, "learning_rate": 1.9891486715522574e-05, "loss": 2.4959, "step": 3065 }, { "epoch": 0.08, "learning_rate": 1.9891369394770658e-05, "loss": 2.1597, "step": 3066 }, { "epoch": 0.08, "learning_rate": 1.9891252010977827e-05, "loss": 2.4412, "step": 3067 }, { "epoch": 0.08, "learning_rate": 1.9891134564144817e-05, "loss": 2.2745, "step": 3068 }, { "epoch": 0.08, "learning_rate": 1.9891017054272386e-05, "loss": 2.4468, "step": 3069 }, { "epoch": 0.08, "learning_rate": 1.989089948136128e-05, "loss": 2.3456, "step": 3070 }, { "epoch": 0.08, "learning_rate": 1.989078184541225e-05, "loss": 2.4472, "step": 3071 }, { "epoch": 0.08, "learning_rate": 1.9890664146426037e-05, "loss": 2.4964, "step": 3072 }, { "epoch": 0.08, "learning_rate": 1.9890546384403404e-05, "loss": 2.4229, "step": 3073 }, { "epoch": 0.08, "learning_rate": 1.9890428559345097e-05, "loss": 2.4327, "step": 3074 }, { "epoch": 0.08, "learning_rate": 1.989031067125186e-05, "loss": 2.5735, "step": 3075 }, { "epoch": 0.08, "learning_rate": 1.9890192720124453e-05, "loss": 2.3864, "step": 3076 }, { "epoch": 0.08, "learning_rate": 1.9890074705963625e-05, "loss": 2.4554, "step": 3077 }, { "epoch": 0.08, "learning_rate": 1.9889956628770128e-05, "loss": 2.4564, "step": 3078 }, { "epoch": 0.08, "learning_rate": 1.9889838488544715e-05, "loss": 2.3, "step": 3079 }, { "epoch": 0.08, "learning_rate": 1.9889720285288135e-05, "loss": 2.558, "step": 3080 }, { "epoch": 0.08, "learning_rate": 1.9889602019001147e-05, "loss": 2.3852, "step": 3081 }, { "epoch": 0.08, "learning_rate": 1.9889483689684502e-05, "loss": 2.5065, "step": 3082 }, { "epoch": 0.08, "learning_rate": 1.9889365297338957e-05, "loss": 2.3302, "step": 3083 }, { "epoch": 0.08, "learning_rate": 1.9889246841965263e-05, "loss": 2.4905, "step": 3084 }, { "epoch": 0.08, "learning_rate": 1.9889128323564172e-05, "loss": 2.3789, "step": 3085 }, { "epoch": 0.08, "learning_rate": 1.9889009742136448e-05, "loss": 2.4538, "step": 3086 }, { "epoch": 0.08, "learning_rate": 1.988889109768284e-05, "loss": 2.4656, "step": 3087 }, { "epoch": 0.08, "learning_rate": 1.9888772390204106e-05, "loss": 2.3785, "step": 3088 }, { "epoch": 0.08, "learning_rate": 1.9888653619701003e-05, "loss": 2.2709, "step": 3089 }, { "epoch": 0.08, "learning_rate": 1.9888534786174288e-05, "loss": 2.5274, "step": 3090 }, { "epoch": 0.08, "learning_rate": 1.988841588962472e-05, "loss": 2.4497, "step": 3091 }, { "epoch": 0.08, "learning_rate": 1.988829693005305e-05, "loss": 2.3594, "step": 3092 }, { "epoch": 0.08, "learning_rate": 1.9888177907460048e-05, "loss": 2.3976, "step": 3093 }, { "epoch": 0.08, "learning_rate": 1.9888058821846465e-05, "loss": 2.505, "step": 3094 }, { "epoch": 0.08, "learning_rate": 1.988793967321306e-05, "loss": 2.2303, "step": 3095 }, { "epoch": 0.08, "learning_rate": 1.9887820461560587e-05, "loss": 2.4365, "step": 3096 }, { "epoch": 0.08, "learning_rate": 1.988770118688982e-05, "loss": 2.4162, "step": 3097 }, { "epoch": 0.08, "learning_rate": 1.9887581849201508e-05, "loss": 2.4172, "step": 3098 }, { "epoch": 0.08, "learning_rate": 1.9887462448496412e-05, "loss": 2.4203, "step": 3099 }, { "epoch": 0.08, "learning_rate": 1.9887342984775297e-05, "loss": 2.5061, "step": 3100 }, { "epoch": 0.08, "learning_rate": 1.9887223458038926e-05, "loss": 2.3411, "step": 3101 }, { "epoch": 0.08, "learning_rate": 1.9887103868288053e-05, "loss": 2.4254, "step": 3102 }, { "epoch": 0.08, "learning_rate": 1.9886984215523445e-05, "loss": 2.3844, "step": 3103 }, { "epoch": 0.08, "learning_rate": 1.988686449974587e-05, "loss": 2.4366, "step": 3104 }, { "epoch": 0.08, "learning_rate": 1.9886744720956082e-05, "loss": 2.4833, "step": 3105 }, { "epoch": 0.08, "learning_rate": 1.9886624879154848e-05, "loss": 2.3552, "step": 3106 }, { "epoch": 0.08, "learning_rate": 1.9886504974342932e-05, "loss": 2.2953, "step": 3107 }, { "epoch": 0.08, "learning_rate": 1.9886385006521095e-05, "loss": 2.4372, "step": 3108 }, { "epoch": 0.08, "learning_rate": 1.988626497569011e-05, "loss": 2.4828, "step": 3109 }, { "epoch": 0.08, "learning_rate": 1.9886144881850734e-05, "loss": 2.4597, "step": 3110 }, { "epoch": 0.08, "learning_rate": 1.9886024725003734e-05, "loss": 2.301, "step": 3111 }, { "epoch": 0.08, "learning_rate": 1.9885904505149877e-05, "loss": 2.4577, "step": 3112 }, { "epoch": 0.08, "learning_rate": 1.9885784222289928e-05, "loss": 2.1374, "step": 3113 }, { "epoch": 0.08, "learning_rate": 1.9885663876424656e-05, "loss": 2.3069, "step": 3114 }, { "epoch": 0.08, "learning_rate": 1.9885543467554827e-05, "loss": 2.4822, "step": 3115 }, { "epoch": 0.08, "learning_rate": 1.9885422995681207e-05, "loss": 2.4869, "step": 3116 }, { "epoch": 0.08, "learning_rate": 1.9885302460804564e-05, "loss": 2.5206, "step": 3117 }, { "epoch": 0.08, "learning_rate": 1.9885181862925666e-05, "loss": 2.5121, "step": 3118 }, { "epoch": 0.08, "learning_rate": 1.9885061202045283e-05, "loss": 2.5011, "step": 3119 }, { "epoch": 0.08, "learning_rate": 1.9884940478164186e-05, "loss": 2.3315, "step": 3120 }, { "epoch": 0.08, "learning_rate": 1.988481969128314e-05, "loss": 2.3567, "step": 3121 }, { "epoch": 0.08, "learning_rate": 1.9884698841402917e-05, "loss": 2.3447, "step": 3122 }, { "epoch": 0.08, "learning_rate": 1.9884577928524286e-05, "loss": 2.2951, "step": 3123 }, { "epoch": 0.08, "learning_rate": 1.988445695264802e-05, "loss": 2.2942, "step": 3124 }, { "epoch": 0.08, "learning_rate": 1.988433591377489e-05, "loss": 2.5095, "step": 3125 }, { "epoch": 0.08, "learning_rate": 1.9884214811905664e-05, "loss": 2.3551, "step": 3126 }, { "epoch": 0.08, "learning_rate": 1.9884093647041115e-05, "loss": 2.2924, "step": 3127 }, { "epoch": 0.08, "learning_rate": 1.9883972419182018e-05, "loss": 2.3448, "step": 3128 }, { "epoch": 0.08, "learning_rate": 1.9883851128329144e-05, "loss": 2.5287, "step": 3129 }, { "epoch": 0.08, "learning_rate": 1.9883729774483264e-05, "loss": 2.4617, "step": 3130 }, { "epoch": 0.08, "learning_rate": 1.9883608357645153e-05, "loss": 2.5641, "step": 3131 }, { "epoch": 0.08, "learning_rate": 1.9883486877815587e-05, "loss": 2.6486, "step": 3132 }, { "epoch": 0.08, "learning_rate": 1.9883365334995335e-05, "loss": 2.4461, "step": 3133 }, { "epoch": 0.08, "learning_rate": 1.9883243729185176e-05, "loss": 2.4858, "step": 3134 }, { "epoch": 0.08, "learning_rate": 1.9883122060385886e-05, "loss": 2.2637, "step": 3135 }, { "epoch": 0.08, "learning_rate": 1.9883000328598235e-05, "loss": 2.5487, "step": 3136 }, { "epoch": 0.08, "learning_rate": 1.9882878533823005e-05, "loss": 2.287, "step": 3137 }, { "epoch": 0.08, "learning_rate": 1.9882756676060968e-05, "loss": 2.3813, "step": 3138 }, { "epoch": 0.08, "learning_rate": 1.9882634755312904e-05, "loss": 2.3503, "step": 3139 }, { "epoch": 0.08, "learning_rate": 1.9882512771579585e-05, "loss": 2.2915, "step": 3140 }, { "epoch": 0.08, "learning_rate": 1.9882390724861794e-05, "loss": 2.5377, "step": 3141 }, { "epoch": 0.08, "learning_rate": 1.9882268615160304e-05, "loss": 2.3277, "step": 3142 }, { "epoch": 0.08, "learning_rate": 1.9882146442475896e-05, "loss": 2.2974, "step": 3143 }, { "epoch": 0.08, "learning_rate": 1.9882024206809348e-05, "loss": 2.3441, "step": 3144 }, { "epoch": 0.08, "learning_rate": 1.9881901908161444e-05, "loss": 2.375, "step": 3145 }, { "epoch": 0.08, "learning_rate": 1.988177954653295e-05, "loss": 2.4121, "step": 3146 }, { "epoch": 0.08, "learning_rate": 1.9881657121924662e-05, "loss": 2.3869, "step": 3147 }, { "epoch": 0.08, "learning_rate": 1.9881534634337352e-05, "loss": 2.3748, "step": 3148 }, { "epoch": 0.08, "learning_rate": 1.9881412083771798e-05, "loss": 2.4722, "step": 3149 }, { "epoch": 0.08, "learning_rate": 1.9881289470228788e-05, "loss": 2.3544, "step": 3150 }, { "epoch": 0.08, "learning_rate": 1.98811667937091e-05, "loss": 2.211, "step": 3151 }, { "epoch": 0.08, "learning_rate": 1.9881044054213512e-05, "loss": 2.5487, "step": 3152 }, { "epoch": 0.08, "learning_rate": 1.9880921251742812e-05, "loss": 2.2218, "step": 3153 }, { "epoch": 0.08, "learning_rate": 1.988079838629778e-05, "loss": 2.2429, "step": 3154 }, { "epoch": 0.08, "learning_rate": 1.98806754578792e-05, "loss": 2.441, "step": 3155 }, { "epoch": 0.08, "learning_rate": 1.9880552466487857e-05, "loss": 2.3074, "step": 3156 }, { "epoch": 0.08, "learning_rate": 1.988042941212453e-05, "loss": 2.3732, "step": 3157 }, { "epoch": 0.08, "learning_rate": 1.988030629479001e-05, "loss": 2.5157, "step": 3158 }, { "epoch": 0.08, "learning_rate": 1.9880183114485078e-05, "loss": 2.4435, "step": 3159 }, { "epoch": 0.08, "learning_rate": 1.9880059871210518e-05, "loss": 2.4399, "step": 3160 }, { "epoch": 0.08, "learning_rate": 1.9879936564967118e-05, "loss": 2.577, "step": 3161 }, { "epoch": 0.08, "learning_rate": 1.9879813195755658e-05, "loss": 2.3193, "step": 3162 }, { "epoch": 0.08, "learning_rate": 1.987968976357693e-05, "loss": 2.3715, "step": 3163 }, { "epoch": 0.08, "learning_rate": 1.9879566268431726e-05, "loss": 2.3558, "step": 3164 }, { "epoch": 0.08, "learning_rate": 1.987944271032082e-05, "loss": 2.3998, "step": 3165 }, { "epoch": 0.08, "learning_rate": 1.9879319089245007e-05, "loss": 2.487, "step": 3166 }, { "epoch": 0.08, "learning_rate": 1.9879195405205075e-05, "loss": 2.3952, "step": 3167 }, { "epoch": 0.08, "learning_rate": 1.987907165820181e-05, "loss": 2.3892, "step": 3168 }, { "epoch": 0.08, "learning_rate": 1.9878947848236004e-05, "loss": 2.292, "step": 3169 }, { "epoch": 0.08, "learning_rate": 1.9878823975308443e-05, "loss": 2.3399, "step": 3170 }, { "epoch": 0.08, "learning_rate": 1.9878700039419917e-05, "loss": 2.5163, "step": 3171 }, { "epoch": 0.08, "learning_rate": 1.9878576040571215e-05, "loss": 2.294, "step": 3172 }, { "epoch": 0.08, "learning_rate": 1.987845197876313e-05, "loss": 2.3866, "step": 3173 }, { "epoch": 0.08, "learning_rate": 1.9878327853996452e-05, "loss": 2.4849, "step": 3174 }, { "epoch": 0.08, "learning_rate": 1.987820366627197e-05, "loss": 2.3138, "step": 3175 }, { "epoch": 0.08, "learning_rate": 1.9878079415590476e-05, "loss": 2.4411, "step": 3176 }, { "epoch": 0.08, "learning_rate": 1.9877955101952764e-05, "loss": 2.3644, "step": 3177 }, { "epoch": 0.08, "learning_rate": 1.9877830725359626e-05, "loss": 2.4194, "step": 3178 }, { "epoch": 0.08, "learning_rate": 1.987770628581185e-05, "loss": 2.4277, "step": 3179 }, { "epoch": 0.08, "learning_rate": 1.9877581783310237e-05, "loss": 2.2013, "step": 3180 }, { "epoch": 0.08, "learning_rate": 1.9877457217855576e-05, "loss": 2.4586, "step": 3181 }, { "epoch": 0.08, "learning_rate": 1.9877332589448657e-05, "loss": 2.3765, "step": 3182 }, { "epoch": 0.08, "learning_rate": 1.9877207898090283e-05, "loss": 2.3548, "step": 3183 }, { "epoch": 0.08, "learning_rate": 1.987708314378124e-05, "loss": 2.3057, "step": 3184 }, { "epoch": 0.08, "learning_rate": 1.9876958326522332e-05, "loss": 2.5128, "step": 3185 }, { "epoch": 0.08, "learning_rate": 1.9876833446314348e-05, "loss": 2.4172, "step": 3186 }, { "epoch": 0.08, "learning_rate": 1.9876708503158083e-05, "loss": 2.4206, "step": 3187 }, { "epoch": 0.08, "learning_rate": 1.9876583497054338e-05, "loss": 2.3909, "step": 3188 }, { "epoch": 0.08, "learning_rate": 1.9876458428003912e-05, "loss": 2.4482, "step": 3189 }, { "epoch": 0.08, "learning_rate": 1.987633329600759e-05, "loss": 2.3533, "step": 3190 }, { "epoch": 0.08, "learning_rate": 1.9876208101066182e-05, "loss": 2.2361, "step": 3191 }, { "epoch": 0.08, "learning_rate": 1.987608284318048e-05, "loss": 2.2812, "step": 3192 }, { "epoch": 0.08, "learning_rate": 1.9875957522351283e-05, "loss": 2.5414, "step": 3193 }, { "epoch": 0.08, "learning_rate": 1.9875832138579392e-05, "loss": 2.3256, "step": 3194 }, { "epoch": 0.08, "learning_rate": 1.9875706691865605e-05, "loss": 2.3111, "step": 3195 }, { "epoch": 0.08, "learning_rate": 1.9875581182210717e-05, "loss": 2.3834, "step": 3196 }, { "epoch": 0.08, "learning_rate": 1.987545560961553e-05, "loss": 2.3341, "step": 3197 }, { "epoch": 0.08, "learning_rate": 1.9875329974080854e-05, "loss": 2.4552, "step": 3198 }, { "epoch": 0.08, "learning_rate": 1.987520427560748e-05, "loss": 2.3574, "step": 3199 }, { "epoch": 0.08, "learning_rate": 1.9875078514196208e-05, "loss": 2.1796, "step": 3200 }, { "epoch": 0.08, "learning_rate": 1.987495268984784e-05, "loss": 2.408, "step": 3201 }, { "epoch": 0.08, "learning_rate": 1.9874826802563185e-05, "loss": 2.3098, "step": 3202 }, { "epoch": 0.08, "learning_rate": 1.987470085234304e-05, "loss": 2.4268, "step": 3203 }, { "epoch": 0.08, "learning_rate": 1.9874574839188206e-05, "loss": 2.3116, "step": 3204 }, { "epoch": 0.08, "learning_rate": 1.987444876309949e-05, "loss": 2.5717, "step": 3205 }, { "epoch": 0.08, "learning_rate": 1.9874322624077694e-05, "loss": 2.5107, "step": 3206 }, { "epoch": 0.08, "learning_rate": 1.9874196422123622e-05, "loss": 2.4202, "step": 3207 }, { "epoch": 0.08, "learning_rate": 1.9874070157238078e-05, "loss": 2.4081, "step": 3208 }, { "epoch": 0.08, "learning_rate": 1.9873943829421866e-05, "loss": 2.3647, "step": 3209 }, { "epoch": 0.08, "learning_rate": 1.9873817438675794e-05, "loss": 2.2484, "step": 3210 }, { "epoch": 0.08, "learning_rate": 1.9873690985000667e-05, "loss": 2.3844, "step": 3211 }, { "epoch": 0.08, "learning_rate": 1.9873564468397287e-05, "loss": 2.3816, "step": 3212 }, { "epoch": 0.08, "learning_rate": 1.9873437888866464e-05, "loss": 2.2713, "step": 3213 }, { "epoch": 0.08, "learning_rate": 1.9873311246409e-05, "loss": 2.3739, "step": 3214 }, { "epoch": 0.08, "learning_rate": 1.987318454102571e-05, "loss": 2.3897, "step": 3215 }, { "epoch": 0.08, "learning_rate": 1.98730577727174e-05, "loss": 2.3941, "step": 3216 }, { "epoch": 0.08, "learning_rate": 1.9872930941484872e-05, "loss": 2.5299, "step": 3217 }, { "epoch": 0.08, "learning_rate": 1.9872804047328936e-05, "loss": 2.2526, "step": 3218 }, { "epoch": 0.08, "learning_rate": 1.9872677090250402e-05, "loss": 2.2563, "step": 3219 }, { "epoch": 0.08, "learning_rate": 1.9872550070250084e-05, "loss": 2.3718, "step": 3220 }, { "epoch": 0.08, "learning_rate": 1.9872422987328783e-05, "loss": 2.4984, "step": 3221 }, { "epoch": 0.08, "learning_rate": 1.9872295841487313e-05, "loss": 2.3926, "step": 3222 }, { "epoch": 0.08, "learning_rate": 1.9872168632726488e-05, "loss": 2.1587, "step": 3223 }, { "epoch": 0.08, "learning_rate": 1.987204136104711e-05, "loss": 2.4944, "step": 3224 }, { "epoch": 0.08, "learning_rate": 1.987191402645e-05, "loss": 2.3657, "step": 3225 }, { "epoch": 0.08, "learning_rate": 1.9871786628935963e-05, "loss": 2.3723, "step": 3226 }, { "epoch": 0.08, "learning_rate": 1.987165916850581e-05, "loss": 2.394, "step": 3227 }, { "epoch": 0.08, "learning_rate": 1.987153164516036e-05, "loss": 2.3695, "step": 3228 }, { "epoch": 0.08, "learning_rate": 1.987140405890042e-05, "loss": 2.3273, "step": 3229 }, { "epoch": 0.08, "learning_rate": 1.9871276409726806e-05, "loss": 2.4385, "step": 3230 }, { "epoch": 0.08, "learning_rate": 1.987114869764033e-05, "loss": 2.2934, "step": 3231 }, { "epoch": 0.08, "learning_rate": 1.9871020922641806e-05, "loss": 2.3418, "step": 3232 }, { "epoch": 0.08, "learning_rate": 1.9870893084732047e-05, "loss": 2.4615, "step": 3233 }, { "epoch": 0.08, "learning_rate": 1.987076518391187e-05, "loss": 2.3267, "step": 3234 }, { "epoch": 0.08, "learning_rate": 1.9870637220182095e-05, "loss": 2.4449, "step": 3235 }, { "epoch": 0.08, "learning_rate": 1.9870509193543525e-05, "loss": 2.39, "step": 3236 }, { "epoch": 0.08, "learning_rate": 1.9870381103996988e-05, "loss": 2.3444, "step": 3237 }, { "epoch": 0.08, "learning_rate": 1.987025295154329e-05, "loss": 2.4191, "step": 3238 }, { "epoch": 0.08, "learning_rate": 1.987012473618326e-05, "loss": 2.464, "step": 3239 }, { "epoch": 0.08, "learning_rate": 1.9869996457917707e-05, "loss": 2.54, "step": 3240 }, { "epoch": 0.08, "learning_rate": 1.986986811674745e-05, "loss": 2.1341, "step": 3241 }, { "epoch": 0.08, "learning_rate": 1.9869739712673305e-05, "loss": 2.4036, "step": 3242 }, { "epoch": 0.08, "learning_rate": 1.9869611245696093e-05, "loss": 2.3794, "step": 3243 }, { "epoch": 0.08, "learning_rate": 1.986948271581663e-05, "loss": 2.3455, "step": 3244 }, { "epoch": 0.08, "learning_rate": 1.986935412303574e-05, "loss": 2.2534, "step": 3245 }, { "epoch": 0.08, "learning_rate": 1.986922546735424e-05, "loss": 2.4156, "step": 3246 }, { "epoch": 0.08, "learning_rate": 1.986909674877295e-05, "loss": 2.3953, "step": 3247 }, { "epoch": 0.08, "learning_rate": 1.9868967967292688e-05, "loss": 2.2923, "step": 3248 }, { "epoch": 0.08, "learning_rate": 1.986883912291428e-05, "loss": 2.4746, "step": 3249 }, { "epoch": 0.08, "learning_rate": 1.986871021563854e-05, "loss": 2.2315, "step": 3250 }, { "epoch": 0.08, "learning_rate": 1.98685812454663e-05, "loss": 2.3785, "step": 3251 }, { "epoch": 0.08, "learning_rate": 1.986845221239837e-05, "loss": 2.294, "step": 3252 }, { "epoch": 0.08, "learning_rate": 1.986832311643558e-05, "loss": 2.262, "step": 3253 }, { "epoch": 0.08, "learning_rate": 1.986819395757875e-05, "loss": 2.3057, "step": 3254 }, { "epoch": 0.08, "learning_rate": 1.9868064735828704e-05, "loss": 2.4695, "step": 3255 }, { "epoch": 0.08, "learning_rate": 1.9867935451186268e-05, "loss": 2.4098, "step": 3256 }, { "epoch": 0.08, "learning_rate": 1.986780610365226e-05, "loss": 2.2772, "step": 3257 }, { "epoch": 0.08, "learning_rate": 1.986767669322751e-05, "loss": 2.3564, "step": 3258 }, { "epoch": 0.08, "learning_rate": 1.986754721991284e-05, "loss": 2.4578, "step": 3259 }, { "epoch": 0.08, "learning_rate": 1.9867417683709076e-05, "loss": 2.3022, "step": 3260 }, { "epoch": 0.08, "learning_rate": 1.9867288084617042e-05, "loss": 2.3612, "step": 3261 }, { "epoch": 0.08, "learning_rate": 1.9867158422637565e-05, "loss": 2.3035, "step": 3262 }, { "epoch": 0.08, "learning_rate": 1.9867028697771476e-05, "loss": 2.319, "step": 3263 }, { "epoch": 0.08, "learning_rate": 1.9866898910019597e-05, "loss": 2.3192, "step": 3264 }, { "epoch": 0.08, "learning_rate": 1.9866769059382752e-05, "loss": 2.2083, "step": 3265 }, { "epoch": 0.08, "learning_rate": 1.986663914586177e-05, "loss": 2.3888, "step": 3266 }, { "epoch": 0.08, "learning_rate": 1.9866509169457488e-05, "loss": 2.4685, "step": 3267 }, { "epoch": 0.08, "learning_rate": 1.9866379130170725e-05, "loss": 2.2665, "step": 3268 }, { "epoch": 0.08, "learning_rate": 1.9866249028002313e-05, "loss": 2.2103, "step": 3269 }, { "epoch": 0.08, "learning_rate": 1.986611886295308e-05, "loss": 2.4633, "step": 3270 }, { "epoch": 0.08, "learning_rate": 1.9865988635023856e-05, "loss": 2.3224, "step": 3271 }, { "epoch": 0.08, "learning_rate": 1.986585834421547e-05, "loss": 2.5194, "step": 3272 }, { "epoch": 0.08, "learning_rate": 1.9865727990528757e-05, "loss": 2.4245, "step": 3273 }, { "epoch": 0.08, "learning_rate": 1.9865597573964545e-05, "loss": 2.3357, "step": 3274 }, { "epoch": 0.08, "learning_rate": 1.986546709452366e-05, "loss": 2.4657, "step": 3275 }, { "epoch": 0.08, "learning_rate": 1.986533655220694e-05, "loss": 2.3528, "step": 3276 }, { "epoch": 0.08, "learning_rate": 1.9865205947015213e-05, "loss": 2.5046, "step": 3277 }, { "epoch": 0.08, "learning_rate": 1.9865075278949317e-05, "loss": 2.3266, "step": 3278 }, { "epoch": 0.08, "learning_rate": 1.986494454801008e-05, "loss": 2.5406, "step": 3279 }, { "epoch": 0.08, "learning_rate": 1.9864813754198335e-05, "loss": 2.2304, "step": 3280 }, { "epoch": 0.08, "learning_rate": 1.986468289751492e-05, "loss": 2.1979, "step": 3281 }, { "epoch": 0.08, "learning_rate": 1.9864551977960665e-05, "loss": 2.4022, "step": 3282 }, { "epoch": 0.08, "learning_rate": 1.9864420995536404e-05, "loss": 2.4459, "step": 3283 }, { "epoch": 0.08, "learning_rate": 1.9864289950242976e-05, "loss": 2.391, "step": 3284 }, { "epoch": 0.08, "learning_rate": 1.986415884208121e-05, "loss": 2.3554, "step": 3285 }, { "epoch": 0.08, "learning_rate": 1.9864027671051948e-05, "loss": 2.5209, "step": 3286 }, { "epoch": 0.08, "learning_rate": 1.9863896437156022e-05, "loss": 2.4128, "step": 3287 }, { "epoch": 0.08, "learning_rate": 1.986376514039427e-05, "loss": 2.4831, "step": 3288 }, { "epoch": 0.08, "learning_rate": 1.9863633780767525e-05, "loss": 2.254, "step": 3289 }, { "epoch": 0.08, "learning_rate": 1.9863502358276633e-05, "loss": 2.3051, "step": 3290 }, { "epoch": 0.08, "learning_rate": 1.9863370872922423e-05, "loss": 2.3185, "step": 3291 }, { "epoch": 0.08, "learning_rate": 1.9863239324705737e-05, "loss": 2.3357, "step": 3292 }, { "epoch": 0.08, "learning_rate": 1.9863107713627408e-05, "loss": 2.4746, "step": 3293 }, { "epoch": 0.08, "learning_rate": 1.9862976039688285e-05, "loss": 2.3348, "step": 3294 }, { "epoch": 0.08, "learning_rate": 1.9862844302889203e-05, "loss": 2.3819, "step": 3295 }, { "epoch": 0.08, "learning_rate": 1.9862712503230993e-05, "loss": 2.5215, "step": 3296 }, { "epoch": 0.08, "learning_rate": 1.9862580640714506e-05, "loss": 2.4079, "step": 3297 }, { "epoch": 0.08, "learning_rate": 1.9862448715340577e-05, "loss": 2.2987, "step": 3298 }, { "epoch": 0.08, "learning_rate": 1.9862316727110048e-05, "loss": 2.4752, "step": 3299 }, { "epoch": 0.08, "learning_rate": 1.9862184676023765e-05, "loss": 2.4914, "step": 3300 }, { "epoch": 0.08, "learning_rate": 1.9862052562082562e-05, "loss": 2.4629, "step": 3301 }, { "epoch": 0.08, "learning_rate": 1.9861920385287283e-05, "loss": 2.4111, "step": 3302 }, { "epoch": 0.08, "learning_rate": 1.986178814563877e-05, "loss": 2.3204, "step": 3303 }, { "epoch": 0.08, "learning_rate": 1.9861655843137873e-05, "loss": 2.3964, "step": 3304 }, { "epoch": 0.08, "learning_rate": 1.9861523477785426e-05, "loss": 2.2655, "step": 3305 }, { "epoch": 0.08, "learning_rate": 1.9861391049582274e-05, "loss": 2.3828, "step": 3306 }, { "epoch": 0.08, "learning_rate": 1.9861258558529267e-05, "loss": 2.4076, "step": 3307 }, { "epoch": 0.08, "learning_rate": 1.9861126004627246e-05, "loss": 2.389, "step": 3308 }, { "epoch": 0.08, "learning_rate": 1.986099338787705e-05, "loss": 2.2713, "step": 3309 }, { "epoch": 0.08, "learning_rate": 1.9860860708279534e-05, "loss": 2.4727, "step": 3310 }, { "epoch": 0.08, "learning_rate": 1.9860727965835537e-05, "loss": 2.4079, "step": 3311 }, { "epoch": 0.08, "learning_rate": 1.9860595160545907e-05, "loss": 2.3441, "step": 3312 }, { "epoch": 0.08, "learning_rate": 1.9860462292411494e-05, "loss": 2.4345, "step": 3313 }, { "epoch": 0.08, "learning_rate": 1.9860329361433136e-05, "loss": 2.3566, "step": 3314 }, { "epoch": 0.08, "learning_rate": 1.9860196367611692e-05, "loss": 2.2959, "step": 3315 }, { "epoch": 0.08, "learning_rate": 1.9860063310948e-05, "loss": 2.3565, "step": 3316 }, { "epoch": 0.08, "learning_rate": 1.985993019144291e-05, "loss": 2.344, "step": 3317 }, { "epoch": 0.08, "learning_rate": 1.9859797009097274e-05, "loss": 2.5332, "step": 3318 }, { "epoch": 0.08, "learning_rate": 1.9859663763911935e-05, "loss": 2.3487, "step": 3319 }, { "epoch": 0.08, "learning_rate": 1.9859530455887746e-05, "loss": 2.3035, "step": 3320 }, { "epoch": 0.08, "learning_rate": 1.985939708502556e-05, "loss": 2.3523, "step": 3321 }, { "epoch": 0.08, "learning_rate": 1.9859263651326222e-05, "loss": 2.5598, "step": 3322 }, { "epoch": 0.08, "learning_rate": 1.985913015479058e-05, "loss": 2.4421, "step": 3323 }, { "epoch": 0.08, "learning_rate": 1.9858996595419493e-05, "loss": 2.5289, "step": 3324 }, { "epoch": 0.08, "learning_rate": 1.9858862973213805e-05, "loss": 2.4253, "step": 3325 }, { "epoch": 0.08, "learning_rate": 1.985872928817437e-05, "loss": 2.3234, "step": 3326 }, { "epoch": 0.08, "learning_rate": 1.9858595540302045e-05, "loss": 2.3213, "step": 3327 }, { "epoch": 0.08, "learning_rate": 1.9858461729597677e-05, "loss": 2.3139, "step": 3328 }, { "epoch": 0.08, "learning_rate": 1.9858327856062115e-05, "loss": 2.4682, "step": 3329 }, { "epoch": 0.08, "learning_rate": 1.9858193919696217e-05, "loss": 2.269, "step": 3330 }, { "epoch": 0.08, "learning_rate": 1.985805992050084e-05, "loss": 2.2652, "step": 3331 }, { "epoch": 0.08, "learning_rate": 1.9857925858476833e-05, "loss": 2.3968, "step": 3332 }, { "epoch": 0.08, "learning_rate": 1.9857791733625055e-05, "loss": 2.2762, "step": 3333 }, { "epoch": 0.08, "learning_rate": 1.9857657545946352e-05, "loss": 2.4861, "step": 3334 }, { "epoch": 0.08, "learning_rate": 1.985752329544159e-05, "loss": 2.4223, "step": 3335 }, { "epoch": 0.08, "learning_rate": 1.9857388982111617e-05, "loss": 2.407, "step": 3336 }, { "epoch": 0.08, "learning_rate": 1.985725460595729e-05, "loss": 2.2515, "step": 3337 }, { "epoch": 0.08, "learning_rate": 1.985712016697947e-05, "loss": 2.7166, "step": 3338 }, { "epoch": 0.08, "learning_rate": 1.985698566517901e-05, "loss": 2.4128, "step": 3339 }, { "epoch": 0.08, "learning_rate": 1.9856851100556765e-05, "loss": 2.2659, "step": 3340 }, { "epoch": 0.08, "learning_rate": 1.98567164731136e-05, "loss": 2.3783, "step": 3341 }, { "epoch": 0.08, "learning_rate": 1.9856581782850364e-05, "loss": 2.3501, "step": 3342 }, { "epoch": 0.08, "learning_rate": 1.9856447029767926e-05, "loss": 2.31, "step": 3343 }, { "epoch": 0.08, "learning_rate": 1.9856312213867137e-05, "loss": 2.2582, "step": 3344 }, { "epoch": 0.08, "learning_rate": 1.985617733514886e-05, "loss": 2.4487, "step": 3345 }, { "epoch": 0.08, "learning_rate": 1.9856042393613946e-05, "loss": 2.329, "step": 3346 }, { "epoch": 0.08, "learning_rate": 1.9855907389263268e-05, "loss": 2.4168, "step": 3347 }, { "epoch": 0.08, "learning_rate": 1.985577232209768e-05, "loss": 2.3602, "step": 3348 }, { "epoch": 0.08, "learning_rate": 1.985563719211804e-05, "loss": 2.4949, "step": 3349 }, { "epoch": 0.08, "learning_rate": 1.985550199932521e-05, "loss": 2.2475, "step": 3350 }, { "epoch": 0.08, "learning_rate": 1.985536674372006e-05, "loss": 2.3791, "step": 3351 }, { "epoch": 0.08, "learning_rate": 1.9855231425303443e-05, "loss": 2.3435, "step": 3352 }, { "epoch": 0.08, "learning_rate": 1.9855096044076225e-05, "loss": 2.3802, "step": 3353 }, { "epoch": 0.08, "learning_rate": 1.9854960600039267e-05, "loss": 2.3459, "step": 3354 }, { "epoch": 0.08, "learning_rate": 1.9854825093193438e-05, "loss": 2.4401, "step": 3355 }, { "epoch": 0.08, "learning_rate": 1.9854689523539594e-05, "loss": 2.5769, "step": 3356 }, { "epoch": 0.08, "learning_rate": 1.9854553891078602e-05, "loss": 2.2549, "step": 3357 }, { "epoch": 0.08, "learning_rate": 1.9854418195811324e-05, "loss": 2.3642, "step": 3358 }, { "epoch": 0.08, "learning_rate": 1.9854282437738633e-05, "loss": 2.2295, "step": 3359 }, { "epoch": 0.08, "learning_rate": 1.9854146616861382e-05, "loss": 2.4455, "step": 3360 }, { "epoch": 0.08, "learning_rate": 1.9854010733180448e-05, "loss": 2.5623, "step": 3361 }, { "epoch": 0.08, "learning_rate": 1.9853874786696692e-05, "loss": 2.2485, "step": 3362 }, { "epoch": 0.08, "learning_rate": 1.985373877741098e-05, "loss": 2.3941, "step": 3363 }, { "epoch": 0.08, "learning_rate": 1.9853602705324177e-05, "loss": 2.4606, "step": 3364 }, { "epoch": 0.08, "learning_rate": 1.9853466570437153e-05, "loss": 2.3427, "step": 3365 }, { "epoch": 0.08, "learning_rate": 1.985333037275078e-05, "loss": 2.1578, "step": 3366 }, { "epoch": 0.08, "learning_rate": 1.9853194112265918e-05, "loss": 2.21, "step": 3367 }, { "epoch": 0.08, "learning_rate": 1.9853057788983442e-05, "loss": 2.3626, "step": 3368 }, { "epoch": 0.08, "learning_rate": 1.9852921402904217e-05, "loss": 2.5008, "step": 3369 }, { "epoch": 0.08, "learning_rate": 1.9852784954029106e-05, "loss": 2.4251, "step": 3370 }, { "epoch": 0.08, "learning_rate": 1.9852648442358994e-05, "loss": 2.3734, "step": 3371 }, { "epoch": 0.08, "learning_rate": 1.985251186789474e-05, "loss": 2.4831, "step": 3372 }, { "epoch": 0.08, "learning_rate": 1.9852375230637217e-05, "loss": 2.1376, "step": 3373 }, { "epoch": 0.08, "learning_rate": 1.985223853058729e-05, "loss": 2.3786, "step": 3374 }, { "epoch": 0.08, "learning_rate": 1.9852101767745844e-05, "loss": 2.312, "step": 3375 }, { "epoch": 0.08, "learning_rate": 1.9851964942113738e-05, "loss": 2.2391, "step": 3376 }, { "epoch": 0.08, "learning_rate": 1.985182805369185e-05, "loss": 2.4175, "step": 3377 }, { "epoch": 0.08, "learning_rate": 1.9851691102481053e-05, "loss": 2.3142, "step": 3378 }, { "epoch": 0.08, "learning_rate": 1.985155408848222e-05, "loss": 2.2254, "step": 3379 }, { "epoch": 0.08, "learning_rate": 1.9851417011696218e-05, "loss": 2.3612, "step": 3380 }, { "epoch": 0.08, "learning_rate": 1.9851279872123927e-05, "loss": 2.1776, "step": 3381 }, { "epoch": 0.08, "learning_rate": 1.9851142669766216e-05, "loss": 2.4084, "step": 3382 }, { "epoch": 0.08, "learning_rate": 1.9851005404623967e-05, "loss": 2.4467, "step": 3383 }, { "epoch": 0.08, "learning_rate": 1.9850868076698046e-05, "loss": 2.3644, "step": 3384 }, { "epoch": 0.08, "learning_rate": 1.9850730685989336e-05, "loss": 2.6293, "step": 3385 }, { "epoch": 0.08, "learning_rate": 1.9850593232498706e-05, "loss": 2.441, "step": 3386 }, { "epoch": 0.08, "learning_rate": 1.9850455716227038e-05, "loss": 2.3721, "step": 3387 }, { "epoch": 0.08, "learning_rate": 1.9850318137175205e-05, "loss": 2.2747, "step": 3388 }, { "epoch": 0.08, "learning_rate": 1.985018049534408e-05, "loss": 2.2529, "step": 3389 }, { "epoch": 0.08, "learning_rate": 1.985004279073455e-05, "loss": 2.315, "step": 3390 }, { "epoch": 0.08, "learning_rate": 1.9849905023347486e-05, "loss": 2.4644, "step": 3391 }, { "epoch": 0.08, "learning_rate": 1.9849767193183764e-05, "loss": 2.4079, "step": 3392 }, { "epoch": 0.08, "learning_rate": 1.984962930024427e-05, "loss": 2.3311, "step": 3393 }, { "epoch": 0.08, "learning_rate": 1.9849491344529874e-05, "loss": 2.1977, "step": 3394 }, { "epoch": 0.08, "learning_rate": 1.9849353326041462e-05, "loss": 2.4379, "step": 3395 }, { "epoch": 0.08, "learning_rate": 1.9849215244779915e-05, "loss": 2.4501, "step": 3396 }, { "epoch": 0.08, "learning_rate": 1.9849077100746105e-05, "loss": 2.3972, "step": 3397 }, { "epoch": 0.08, "learning_rate": 1.984893889394092e-05, "loss": 2.2227, "step": 3398 }, { "epoch": 0.08, "learning_rate": 1.9848800624365237e-05, "loss": 2.2962, "step": 3399 }, { "epoch": 0.08, "learning_rate": 1.9848662292019936e-05, "loss": 2.3204, "step": 3400 }, { "epoch": 0.08, "learning_rate": 1.9848523896905903e-05, "loss": 2.4422, "step": 3401 }, { "epoch": 0.08, "learning_rate": 1.9848385439024014e-05, "loss": 2.4194, "step": 3402 }, { "epoch": 0.08, "learning_rate": 1.9848246918375158e-05, "loss": 2.4794, "step": 3403 }, { "epoch": 0.08, "learning_rate": 1.9848108334960215e-05, "loss": 2.3725, "step": 3404 }, { "epoch": 0.08, "learning_rate": 1.9847969688780066e-05, "loss": 2.4583, "step": 3405 }, { "epoch": 0.08, "learning_rate": 1.98478309798356e-05, "loss": 2.4019, "step": 3406 }, { "epoch": 0.08, "learning_rate": 1.9847692208127696e-05, "loss": 2.4382, "step": 3407 }, { "epoch": 0.08, "learning_rate": 1.9847553373657242e-05, "loss": 2.4039, "step": 3408 }, { "epoch": 0.08, "learning_rate": 1.9847414476425117e-05, "loss": 2.2083, "step": 3409 }, { "epoch": 0.08, "learning_rate": 1.9847275516432214e-05, "loss": 2.3488, "step": 3410 }, { "epoch": 0.08, "learning_rate": 1.9847136493679415e-05, "loss": 2.398, "step": 3411 }, { "epoch": 0.08, "learning_rate": 1.9846997408167606e-05, "loss": 2.3497, "step": 3412 }, { "epoch": 0.08, "learning_rate": 1.9846858259897674e-05, "loss": 2.3822, "step": 3413 }, { "epoch": 0.08, "learning_rate": 1.98467190488705e-05, "loss": 2.5078, "step": 3414 }, { "epoch": 0.08, "learning_rate": 1.984657977508698e-05, "loss": 2.4615, "step": 3415 }, { "epoch": 0.08, "learning_rate": 1.9846440438548e-05, "loss": 2.5655, "step": 3416 }, { "epoch": 0.08, "learning_rate": 1.9846301039254444e-05, "loss": 2.5254, "step": 3417 }, { "epoch": 0.08, "learning_rate": 1.9846161577207207e-05, "loss": 2.2573, "step": 3418 }, { "epoch": 0.08, "learning_rate": 1.9846022052407168e-05, "loss": 2.1916, "step": 3419 }, { "epoch": 0.08, "learning_rate": 1.9845882464855225e-05, "loss": 2.5203, "step": 3420 }, { "epoch": 0.08, "learning_rate": 1.9845742814552263e-05, "loss": 2.4792, "step": 3421 }, { "epoch": 0.08, "learning_rate": 1.9845603101499172e-05, "loss": 2.2828, "step": 3422 }, { "epoch": 0.08, "learning_rate": 1.9845463325696847e-05, "loss": 2.375, "step": 3423 }, { "epoch": 0.08, "learning_rate": 1.9845323487146172e-05, "loss": 2.2065, "step": 3424 }, { "epoch": 0.08, "learning_rate": 1.9845183585848045e-05, "loss": 2.472, "step": 3425 }, { "epoch": 0.08, "learning_rate": 1.9845043621803354e-05, "loss": 2.3109, "step": 3426 }, { "epoch": 0.08, "learning_rate": 1.9844903595012993e-05, "loss": 2.2734, "step": 3427 }, { "epoch": 0.08, "learning_rate": 1.9844763505477847e-05, "loss": 2.3688, "step": 3428 }, { "epoch": 0.08, "learning_rate": 1.984462335319882e-05, "loss": 2.392, "step": 3429 }, { "epoch": 0.08, "learning_rate": 1.98444831381768e-05, "loss": 2.5398, "step": 3430 }, { "epoch": 0.08, "learning_rate": 1.984434286041268e-05, "loss": 2.3505, "step": 3431 }, { "epoch": 0.08, "learning_rate": 1.9844202519907357e-05, "loss": 2.4896, "step": 3432 }, { "epoch": 0.08, "learning_rate": 1.9844062116661717e-05, "loss": 2.316, "step": 3433 }, { "epoch": 0.08, "learning_rate": 1.9843921650676665e-05, "loss": 2.3882, "step": 3434 }, { "epoch": 0.08, "learning_rate": 1.9843781121953093e-05, "loss": 2.411, "step": 3435 }, { "epoch": 0.08, "learning_rate": 1.9843640530491894e-05, "loss": 2.2738, "step": 3436 }, { "epoch": 0.08, "learning_rate": 1.984349987629397e-05, "loss": 2.4886, "step": 3437 }, { "epoch": 0.08, "learning_rate": 1.9843359159360208e-05, "loss": 2.382, "step": 3438 }, { "epoch": 0.08, "learning_rate": 1.9843218379691513e-05, "loss": 2.3777, "step": 3439 }, { "epoch": 0.08, "learning_rate": 1.984307753728878e-05, "loss": 2.3991, "step": 3440 }, { "epoch": 0.08, "learning_rate": 1.9842936632152904e-05, "loss": 2.4832, "step": 3441 }, { "epoch": 0.08, "learning_rate": 1.9842795664284788e-05, "loss": 2.1975, "step": 3442 }, { "epoch": 0.08, "learning_rate": 1.9842654633685324e-05, "loss": 2.4322, "step": 3443 }, { "epoch": 0.08, "learning_rate": 1.9842513540355416e-05, "loss": 2.537, "step": 3444 }, { "epoch": 0.08, "learning_rate": 1.9842372384295963e-05, "loss": 2.3583, "step": 3445 }, { "epoch": 0.08, "learning_rate": 1.9842231165507862e-05, "loss": 2.2569, "step": 3446 }, { "epoch": 0.08, "learning_rate": 1.9842089883992014e-05, "loss": 2.3405, "step": 3447 }, { "epoch": 0.08, "learning_rate": 1.984194853974932e-05, "loss": 2.545, "step": 3448 }, { "epoch": 0.09, "learning_rate": 1.984180713278068e-05, "loss": 2.5311, "step": 3449 }, { "epoch": 0.09, "learning_rate": 1.9841665663086998e-05, "loss": 2.4614, "step": 3450 }, { "epoch": 0.09, "learning_rate": 1.9841524130669173e-05, "loss": 2.1939, "step": 3451 }, { "epoch": 0.09, "learning_rate": 1.9841382535528107e-05, "loss": 2.1968, "step": 3452 }, { "epoch": 0.09, "learning_rate": 1.98412408776647e-05, "loss": 2.3926, "step": 3453 }, { "epoch": 0.09, "learning_rate": 1.984109915707986e-05, "loss": 2.2391, "step": 3454 }, { "epoch": 0.09, "learning_rate": 1.984095737377449e-05, "loss": 2.3194, "step": 3455 }, { "epoch": 0.09, "learning_rate": 1.9840815527749488e-05, "loss": 2.4995, "step": 3456 }, { "epoch": 0.09, "learning_rate": 1.9840673619005763e-05, "loss": 2.445, "step": 3457 }, { "epoch": 0.09, "learning_rate": 1.984053164754422e-05, "loss": 2.2992, "step": 3458 }, { "epoch": 0.09, "learning_rate": 1.9840389613365757e-05, "loss": 2.4236, "step": 3459 }, { "epoch": 0.09, "learning_rate": 1.9840247516471288e-05, "loss": 2.3589, "step": 3460 }, { "epoch": 0.09, "learning_rate": 1.9840105356861713e-05, "loss": 2.4374, "step": 3461 }, { "epoch": 0.09, "learning_rate": 1.983996313453794e-05, "loss": 2.2795, "step": 3462 }, { "epoch": 0.09, "learning_rate": 1.9839820849500877e-05, "loss": 2.4179, "step": 3463 }, { "epoch": 0.09, "learning_rate": 1.9839678501751427e-05, "loss": 2.3438, "step": 3464 }, { "epoch": 0.09, "learning_rate": 1.98395360912905e-05, "loss": 2.5754, "step": 3465 }, { "epoch": 0.09, "learning_rate": 1.9839393618119006e-05, "loss": 2.4549, "step": 3466 }, { "epoch": 0.09, "learning_rate": 1.9839251082237844e-05, "loss": 2.3169, "step": 3467 }, { "epoch": 0.09, "learning_rate": 1.9839108483647932e-05, "loss": 2.3171, "step": 3468 }, { "epoch": 0.09, "learning_rate": 1.9838965822350175e-05, "loss": 2.4019, "step": 3469 }, { "epoch": 0.09, "learning_rate": 1.9838823098345482e-05, "loss": 2.3206, "step": 3470 }, { "epoch": 0.09, "learning_rate": 1.983868031163476e-05, "loss": 2.3978, "step": 3471 }, { "epoch": 0.09, "learning_rate": 1.9838537462218928e-05, "loss": 2.4066, "step": 3472 }, { "epoch": 0.09, "learning_rate": 1.9838394550098886e-05, "loss": 2.4353, "step": 3473 }, { "epoch": 0.09, "learning_rate": 1.983825157527555e-05, "loss": 2.4524, "step": 3474 }, { "epoch": 0.09, "learning_rate": 1.9838108537749835e-05, "loss": 2.3851, "step": 3475 }, { "epoch": 0.09, "learning_rate": 1.983796543752264e-05, "loss": 2.4193, "step": 3476 }, { "epoch": 0.09, "learning_rate": 1.983782227459489e-05, "loss": 2.4074, "step": 3477 }, { "epoch": 0.09, "learning_rate": 1.9837679048967494e-05, "loss": 2.2152, "step": 3478 }, { "epoch": 0.09, "learning_rate": 1.9837535760641364e-05, "loss": 2.3894, "step": 3479 }, { "epoch": 0.09, "learning_rate": 1.9837392409617407e-05, "loss": 2.3802, "step": 3480 }, { "epoch": 0.09, "learning_rate": 1.9837248995896547e-05, "loss": 2.355, "step": 3481 }, { "epoch": 0.09, "learning_rate": 1.983710551947969e-05, "loss": 2.267, "step": 3482 }, { "epoch": 0.09, "learning_rate": 1.9836961980367754e-05, "loss": 2.3779, "step": 3483 }, { "epoch": 0.09, "learning_rate": 1.9836818378561656e-05, "loss": 2.2789, "step": 3484 }, { "epoch": 0.09, "learning_rate": 1.9836674714062308e-05, "loss": 2.2339, "step": 3485 }, { "epoch": 0.09, "learning_rate": 1.9836530986870627e-05, "loss": 2.4483, "step": 3486 }, { "epoch": 0.09, "learning_rate": 1.9836387196987527e-05, "loss": 2.2642, "step": 3487 }, { "epoch": 0.09, "learning_rate": 1.9836243344413923e-05, "loss": 2.2431, "step": 3488 }, { "epoch": 0.09, "learning_rate": 1.983609942915074e-05, "loss": 2.3579, "step": 3489 }, { "epoch": 0.09, "learning_rate": 1.9835955451198887e-05, "loss": 2.3472, "step": 3490 }, { "epoch": 0.09, "learning_rate": 1.9835811410559286e-05, "loss": 2.2141, "step": 3491 }, { "epoch": 0.09, "learning_rate": 1.9835667307232853e-05, "loss": 2.3696, "step": 3492 }, { "epoch": 0.09, "learning_rate": 1.9835523141220504e-05, "loss": 2.3423, "step": 3493 }, { "epoch": 0.09, "learning_rate": 1.9835378912523163e-05, "loss": 2.3346, "step": 3494 }, { "epoch": 0.09, "learning_rate": 1.9835234621141748e-05, "loss": 2.4033, "step": 3495 }, { "epoch": 0.09, "learning_rate": 1.9835090267077178e-05, "loss": 2.234, "step": 3496 }, { "epoch": 0.09, "learning_rate": 1.983494585033037e-05, "loss": 2.4076, "step": 3497 }, { "epoch": 0.09, "learning_rate": 1.9834801370902248e-05, "loss": 2.4446, "step": 3498 }, { "epoch": 0.09, "learning_rate": 1.9834656828793734e-05, "loss": 2.3625, "step": 3499 }, { "epoch": 0.09, "learning_rate": 1.983451222400574e-05, "loss": 2.3936, "step": 3500 }, { "epoch": 0.09, "learning_rate": 1.9834367556539204e-05, "loss": 2.3947, "step": 3501 }, { "epoch": 0.09, "learning_rate": 1.9834222826395034e-05, "loss": 2.2057, "step": 3502 }, { "epoch": 0.09, "learning_rate": 1.9834078033574157e-05, "loss": 2.4409, "step": 3503 }, { "epoch": 0.09, "learning_rate": 1.9833933178077497e-05, "loss": 2.3096, "step": 3504 }, { "epoch": 0.09, "learning_rate": 1.9833788259905975e-05, "loss": 2.4529, "step": 3505 }, { "epoch": 0.09, "learning_rate": 1.9833643279060514e-05, "loss": 2.1875, "step": 3506 }, { "epoch": 0.09, "learning_rate": 1.983349823554204e-05, "loss": 2.2264, "step": 3507 }, { "epoch": 0.09, "learning_rate": 1.983335312935148e-05, "loss": 2.4152, "step": 3508 }, { "epoch": 0.09, "learning_rate": 1.9833207960489755e-05, "loss": 2.1502, "step": 3509 }, { "epoch": 0.09, "learning_rate": 1.9833062728957792e-05, "loss": 2.4866, "step": 3510 }, { "epoch": 0.09, "learning_rate": 1.9832917434756514e-05, "loss": 2.2755, "step": 3511 }, { "epoch": 0.09, "learning_rate": 1.983277207788685e-05, "loss": 2.3542, "step": 3512 }, { "epoch": 0.09, "learning_rate": 1.9832626658349723e-05, "loss": 2.333, "step": 3513 }, { "epoch": 0.09, "learning_rate": 1.9832481176146063e-05, "loss": 2.3492, "step": 3514 }, { "epoch": 0.09, "learning_rate": 1.9832335631276795e-05, "loss": 2.3346, "step": 3515 }, { "epoch": 0.09, "learning_rate": 1.983219002374285e-05, "loss": 2.4245, "step": 3516 }, { "epoch": 0.09, "learning_rate": 1.9832044353545155e-05, "loss": 2.4313, "step": 3517 }, { "epoch": 0.09, "learning_rate": 1.9831898620684632e-05, "loss": 2.3049, "step": 3518 }, { "epoch": 0.09, "learning_rate": 1.983175282516222e-05, "loss": 2.4065, "step": 3519 }, { "epoch": 0.09, "learning_rate": 1.983160696697884e-05, "loss": 2.2137, "step": 3520 }, { "epoch": 0.09, "learning_rate": 1.9831461046135426e-05, "loss": 2.3786, "step": 3521 }, { "epoch": 0.09, "learning_rate": 1.9831315062632908e-05, "loss": 2.4601, "step": 3522 }, { "epoch": 0.09, "learning_rate": 1.9831169016472215e-05, "loss": 2.3547, "step": 3523 }, { "epoch": 0.09, "learning_rate": 1.9831022907654276e-05, "loss": 2.2857, "step": 3524 }, { "epoch": 0.09, "learning_rate": 1.9830876736180025e-05, "loss": 2.3791, "step": 3525 }, { "epoch": 0.09, "learning_rate": 1.9830730502050392e-05, "loss": 2.2702, "step": 3526 }, { "epoch": 0.09, "learning_rate": 1.983058420526631e-05, "loss": 2.2729, "step": 3527 }, { "epoch": 0.09, "learning_rate": 1.983043784582871e-05, "loss": 2.4582, "step": 3528 }, { "epoch": 0.09, "learning_rate": 1.983029142373853e-05, "loss": 2.2448, "step": 3529 }, { "epoch": 0.09, "learning_rate": 1.9830144938996695e-05, "loss": 2.356, "step": 3530 }, { "epoch": 0.09, "learning_rate": 1.9829998391604148e-05, "loss": 2.375, "step": 3531 }, { "epoch": 0.09, "learning_rate": 1.9829851781561814e-05, "loss": 2.3604, "step": 3532 }, { "epoch": 0.09, "learning_rate": 1.982970510887063e-05, "loss": 2.3861, "step": 3533 }, { "epoch": 0.09, "learning_rate": 1.9829558373531534e-05, "loss": 2.2528, "step": 3534 }, { "epoch": 0.09, "learning_rate": 1.9829411575545457e-05, "loss": 2.5854, "step": 3535 }, { "epoch": 0.09, "learning_rate": 1.9829264714913343e-05, "loss": 2.4469, "step": 3536 }, { "epoch": 0.09, "learning_rate": 1.9829117791636114e-05, "loss": 2.3596, "step": 3537 }, { "epoch": 0.09, "learning_rate": 1.9828970805714718e-05, "loss": 2.1592, "step": 3538 }, { "epoch": 0.09, "learning_rate": 1.9828823757150085e-05, "loss": 2.5128, "step": 3539 }, { "epoch": 0.09, "learning_rate": 1.9828676645943158e-05, "loss": 2.3671, "step": 3540 }, { "epoch": 0.09, "learning_rate": 1.982852947209487e-05, "loss": 2.5577, "step": 3541 }, { "epoch": 0.09, "learning_rate": 1.9828382235606164e-05, "loss": 2.3929, "step": 3542 }, { "epoch": 0.09, "learning_rate": 1.9828234936477974e-05, "loss": 2.2195, "step": 3543 }, { "epoch": 0.09, "learning_rate": 1.9828087574711237e-05, "loss": 2.3507, "step": 3544 }, { "epoch": 0.09, "learning_rate": 1.9827940150306894e-05, "loss": 2.3566, "step": 3545 }, { "epoch": 0.09, "learning_rate": 1.9827792663265888e-05, "loss": 2.3889, "step": 3546 }, { "epoch": 0.09, "learning_rate": 1.9827645113589157e-05, "loss": 2.3883, "step": 3547 }, { "epoch": 0.09, "learning_rate": 1.982749750127764e-05, "loss": 2.4737, "step": 3548 }, { "epoch": 0.09, "learning_rate": 1.9827349826332282e-05, "loss": 2.3096, "step": 3549 }, { "epoch": 0.09, "learning_rate": 1.9827202088754017e-05, "loss": 2.1267, "step": 3550 }, { "epoch": 0.09, "learning_rate": 1.982705428854379e-05, "loss": 2.3997, "step": 3551 }, { "epoch": 0.09, "learning_rate": 1.9826906425702545e-05, "loss": 2.2991, "step": 3552 }, { "epoch": 0.09, "learning_rate": 1.982675850023122e-05, "loss": 2.5593, "step": 3553 }, { "epoch": 0.09, "learning_rate": 1.9826610512130765e-05, "loss": 2.1494, "step": 3554 }, { "epoch": 0.09, "learning_rate": 1.982646246140212e-05, "loss": 2.3609, "step": 3555 }, { "epoch": 0.09, "learning_rate": 1.9826314348046223e-05, "loss": 2.3843, "step": 3556 }, { "epoch": 0.09, "learning_rate": 1.9826166172064024e-05, "loss": 2.3405, "step": 3557 }, { "epoch": 0.09, "learning_rate": 1.9826017933456464e-05, "loss": 2.3028, "step": 3558 }, { "epoch": 0.09, "learning_rate": 1.9825869632224495e-05, "loss": 2.3547, "step": 3559 }, { "epoch": 0.09, "learning_rate": 1.982572126836905e-05, "loss": 2.2898, "step": 3560 }, { "epoch": 0.09, "learning_rate": 1.9825572841891087e-05, "loss": 2.423, "step": 3561 }, { "epoch": 0.09, "learning_rate": 1.9825424352791544e-05, "loss": 2.2236, "step": 3562 }, { "epoch": 0.09, "learning_rate": 1.9825275801071367e-05, "loss": 2.3576, "step": 3563 }, { "epoch": 0.09, "learning_rate": 1.9825127186731508e-05, "loss": 2.4482, "step": 3564 }, { "epoch": 0.09, "learning_rate": 1.982497850977291e-05, "loss": 2.3118, "step": 3565 }, { "epoch": 0.09, "learning_rate": 1.982482977019652e-05, "loss": 2.4392, "step": 3566 }, { "epoch": 0.09, "learning_rate": 1.9824680968003295e-05, "loss": 2.2962, "step": 3567 }, { "epoch": 0.09, "learning_rate": 1.982453210319417e-05, "loss": 2.3352, "step": 3568 }, { "epoch": 0.09, "learning_rate": 1.9824383175770104e-05, "loss": 2.4024, "step": 3569 }, { "epoch": 0.09, "learning_rate": 1.982423418573204e-05, "loss": 2.3382, "step": 3570 }, { "epoch": 0.09, "learning_rate": 1.982408513308093e-05, "loss": 2.3196, "step": 3571 }, { "epoch": 0.09, "learning_rate": 1.9823936017817728e-05, "loss": 2.6844, "step": 3572 }, { "epoch": 0.09, "learning_rate": 1.9823786839943376e-05, "loss": 2.3359, "step": 3573 }, { "epoch": 0.09, "learning_rate": 1.982363759945883e-05, "loss": 2.3312, "step": 3574 }, { "epoch": 0.09, "learning_rate": 1.982348829636504e-05, "loss": 2.3552, "step": 3575 }, { "epoch": 0.09, "learning_rate": 1.9823338930662957e-05, "loss": 2.3665, "step": 3576 }, { "epoch": 0.09, "learning_rate": 1.9823189502353533e-05, "loss": 2.2818, "step": 3577 }, { "epoch": 0.09, "learning_rate": 1.9823040011437726e-05, "loss": 2.3703, "step": 3578 }, { "epoch": 0.09, "learning_rate": 1.9822890457916476e-05, "loss": 2.5881, "step": 3579 }, { "epoch": 0.09, "learning_rate": 1.9822740841790747e-05, "loss": 2.3593, "step": 3580 }, { "epoch": 0.09, "learning_rate": 1.9822591163061494e-05, "loss": 2.3568, "step": 3581 }, { "epoch": 0.09, "learning_rate": 1.982244142172966e-05, "loss": 2.3078, "step": 3582 }, { "epoch": 0.09, "learning_rate": 1.9822291617796213e-05, "loss": 2.2674, "step": 3583 }, { "epoch": 0.09, "learning_rate": 1.9822141751262096e-05, "loss": 2.351, "step": 3584 }, { "epoch": 0.09, "learning_rate": 1.9821991822128267e-05, "loss": 2.4303, "step": 3585 }, { "epoch": 0.09, "learning_rate": 1.9821841830395688e-05, "loss": 2.5922, "step": 3586 }, { "epoch": 0.09, "learning_rate": 1.9821691776065307e-05, "loss": 2.3957, "step": 3587 }, { "epoch": 0.09, "learning_rate": 1.982154165913808e-05, "loss": 2.4007, "step": 3588 }, { "epoch": 0.09, "learning_rate": 1.9821391479614976e-05, "loss": 2.5419, "step": 3589 }, { "epoch": 0.09, "learning_rate": 1.9821241237496937e-05, "loss": 2.3758, "step": 3590 }, { "epoch": 0.09, "learning_rate": 1.9821090932784928e-05, "loss": 2.4442, "step": 3591 }, { "epoch": 0.09, "learning_rate": 1.9820940565479905e-05, "loss": 2.4281, "step": 3592 }, { "epoch": 0.09, "learning_rate": 1.982079013558283e-05, "loss": 2.2753, "step": 3593 }, { "epoch": 0.09, "learning_rate": 1.982063964309466e-05, "loss": 2.3088, "step": 3594 }, { "epoch": 0.09, "learning_rate": 1.9820489088016346e-05, "loss": 2.249, "step": 3595 }, { "epoch": 0.09, "learning_rate": 1.9820338470348862e-05, "loss": 2.2424, "step": 3596 }, { "epoch": 0.09, "learning_rate": 1.9820187790093158e-05, "loss": 2.4092, "step": 3597 }, { "epoch": 0.09, "learning_rate": 1.9820037047250196e-05, "loss": 2.3812, "step": 3598 }, { "epoch": 0.09, "learning_rate": 1.9819886241820942e-05, "loss": 2.6247, "step": 3599 }, { "epoch": 0.09, "learning_rate": 1.9819735373806344e-05, "loss": 2.1985, "step": 3600 }, { "epoch": 0.09, "learning_rate": 1.981958444320738e-05, "loss": 2.4736, "step": 3601 }, { "epoch": 0.09, "learning_rate": 1.9819433450025002e-05, "loss": 2.2836, "step": 3602 }, { "epoch": 0.09, "learning_rate": 1.981928239426017e-05, "loss": 2.298, "step": 3603 }, { "epoch": 0.09, "learning_rate": 1.981913127591386e-05, "loss": 2.2715, "step": 3604 }, { "epoch": 0.09, "learning_rate": 1.981898009498702e-05, "loss": 2.3036, "step": 3605 }, { "epoch": 0.09, "learning_rate": 1.9818828851480623e-05, "loss": 2.3565, "step": 3606 }, { "epoch": 0.09, "learning_rate": 1.981867754539563e-05, "loss": 2.3787, "step": 3607 }, { "epoch": 0.09, "learning_rate": 1.9818526176733006e-05, "loss": 2.2805, "step": 3608 }, { "epoch": 0.09, "learning_rate": 1.9818374745493713e-05, "loss": 2.2768, "step": 3609 }, { "epoch": 0.09, "learning_rate": 1.9818223251678716e-05, "loss": 2.3053, "step": 3610 }, { "epoch": 0.09, "learning_rate": 1.9818071695288983e-05, "loss": 2.3123, "step": 3611 }, { "epoch": 0.09, "learning_rate": 1.9817920076325482e-05, "loss": 2.3238, "step": 3612 }, { "epoch": 0.09, "learning_rate": 1.9817768394789176e-05, "loss": 2.4014, "step": 3613 }, { "epoch": 0.09, "learning_rate": 1.9817616650681032e-05, "loss": 2.3729, "step": 3614 }, { "epoch": 0.09, "learning_rate": 1.9817464844002015e-05, "loss": 2.4947, "step": 3615 }, { "epoch": 0.09, "learning_rate": 1.9817312974753102e-05, "loss": 2.4619, "step": 3616 }, { "epoch": 0.09, "learning_rate": 1.981716104293525e-05, "loss": 2.2651, "step": 3617 }, { "epoch": 0.09, "learning_rate": 1.9817009048549433e-05, "loss": 2.3008, "step": 3618 }, { "epoch": 0.09, "learning_rate": 1.9816856991596612e-05, "loss": 2.3532, "step": 3619 }, { "epoch": 0.09, "learning_rate": 1.9816704872077767e-05, "loss": 2.3877, "step": 3620 }, { "epoch": 0.09, "learning_rate": 1.9816552689993864e-05, "loss": 2.3293, "step": 3621 }, { "epoch": 0.09, "learning_rate": 1.981640044534587e-05, "loss": 2.5135, "step": 3622 }, { "epoch": 0.09, "learning_rate": 1.9816248138134756e-05, "loss": 2.4982, "step": 3623 }, { "epoch": 0.09, "learning_rate": 1.9816095768361494e-05, "loss": 2.4015, "step": 3624 }, { "epoch": 0.09, "learning_rate": 1.9815943336027056e-05, "loss": 2.3791, "step": 3625 }, { "epoch": 0.09, "learning_rate": 1.981579084113241e-05, "loss": 2.4774, "step": 3626 }, { "epoch": 0.09, "learning_rate": 1.9815638283678535e-05, "loss": 2.3246, "step": 3627 }, { "epoch": 0.09, "learning_rate": 1.981548566366639e-05, "loss": 2.3272, "step": 3628 }, { "epoch": 0.09, "learning_rate": 1.9815332981096966e-05, "loss": 2.2423, "step": 3629 }, { "epoch": 0.09, "learning_rate": 1.9815180235971217e-05, "loss": 2.3587, "step": 3630 }, { "epoch": 0.09, "learning_rate": 1.981502742829013e-05, "loss": 2.3365, "step": 3631 }, { "epoch": 0.09, "learning_rate": 1.9814874558054675e-05, "loss": 2.3979, "step": 3632 }, { "epoch": 0.09, "learning_rate": 1.9814721625265826e-05, "loss": 2.4454, "step": 3633 }, { "epoch": 0.09, "learning_rate": 1.9814568629924556e-05, "loss": 2.3088, "step": 3634 }, { "epoch": 0.09, "learning_rate": 1.9814415572031845e-05, "loss": 2.3104, "step": 3635 }, { "epoch": 0.09, "learning_rate": 1.981426245158866e-05, "loss": 2.4455, "step": 3636 }, { "epoch": 0.09, "learning_rate": 1.9814109268595986e-05, "loss": 2.3149, "step": 3637 }, { "epoch": 0.09, "learning_rate": 1.9813956023054793e-05, "loss": 2.2502, "step": 3638 }, { "epoch": 0.09, "learning_rate": 1.981380271496606e-05, "loss": 2.4315, "step": 3639 }, { "epoch": 0.09, "learning_rate": 1.9813649344330765e-05, "loss": 2.1942, "step": 3640 }, { "epoch": 0.09, "learning_rate": 1.9813495911149884e-05, "loss": 2.4203, "step": 3641 }, { "epoch": 0.09, "learning_rate": 1.9813342415424396e-05, "loss": 2.4891, "step": 3642 }, { "epoch": 0.09, "learning_rate": 1.981318885715528e-05, "loss": 2.4039, "step": 3643 }, { "epoch": 0.09, "learning_rate": 1.981303523634351e-05, "loss": 2.2923, "step": 3644 }, { "epoch": 0.09, "learning_rate": 1.9812881552990074e-05, "loss": 2.445, "step": 3645 }, { "epoch": 0.09, "learning_rate": 1.981272780709594e-05, "loss": 2.199, "step": 3646 }, { "epoch": 0.09, "learning_rate": 1.9812573998662097e-05, "loss": 2.4848, "step": 3647 }, { "epoch": 0.09, "learning_rate": 1.9812420127689524e-05, "loss": 2.2097, "step": 3648 }, { "epoch": 0.09, "learning_rate": 1.9812266194179193e-05, "loss": 2.3589, "step": 3649 }, { "epoch": 0.09, "learning_rate": 1.9812112198132095e-05, "loss": 2.3165, "step": 3650 }, { "epoch": 0.09, "learning_rate": 1.981195813954921e-05, "loss": 2.2228, "step": 3651 }, { "epoch": 0.09, "learning_rate": 1.9811804018431518e-05, "loss": 2.3581, "step": 3652 }, { "epoch": 0.09, "learning_rate": 1.981164983478e-05, "loss": 2.2384, "step": 3653 }, { "epoch": 0.09, "learning_rate": 1.981149558859564e-05, "loss": 2.3136, "step": 3654 }, { "epoch": 0.09, "learning_rate": 1.9811341279879425e-05, "loss": 2.2992, "step": 3655 }, { "epoch": 0.09, "learning_rate": 1.981118690863233e-05, "loss": 2.4358, "step": 3656 }, { "epoch": 0.09, "learning_rate": 1.9811032474855345e-05, "loss": 2.3391, "step": 3657 }, { "epoch": 0.09, "learning_rate": 1.9810877978549453e-05, "loss": 2.2801, "step": 3658 }, { "epoch": 0.09, "learning_rate": 1.981072341971564e-05, "loss": 2.293, "step": 3659 }, { "epoch": 0.09, "learning_rate": 1.9810568798354884e-05, "loss": 2.4296, "step": 3660 }, { "epoch": 0.09, "learning_rate": 1.981041411446818e-05, "loss": 2.5309, "step": 3661 }, { "epoch": 0.09, "learning_rate": 1.9810259368056507e-05, "loss": 2.385, "step": 3662 }, { "epoch": 0.09, "learning_rate": 1.981010455912086e-05, "loss": 2.4731, "step": 3663 }, { "epoch": 0.09, "learning_rate": 1.9809949687662216e-05, "loss": 2.5227, "step": 3664 }, { "epoch": 0.09, "learning_rate": 1.9809794753681563e-05, "loss": 2.5656, "step": 3665 }, { "epoch": 0.09, "learning_rate": 1.9809639757179893e-05, "loss": 2.1777, "step": 3666 }, { "epoch": 0.09, "learning_rate": 1.9809484698158196e-05, "loss": 2.2878, "step": 3667 }, { "epoch": 0.09, "learning_rate": 1.9809329576617448e-05, "loss": 2.4062, "step": 3668 }, { "epoch": 0.09, "learning_rate": 1.9809174392558655e-05, "loss": 2.3582, "step": 3669 }, { "epoch": 0.09, "learning_rate": 1.980901914598279e-05, "loss": 2.3847, "step": 3670 }, { "epoch": 0.09, "learning_rate": 1.980886383689085e-05, "loss": 2.352, "step": 3671 }, { "epoch": 0.09, "learning_rate": 1.980870846528383e-05, "loss": 2.3644, "step": 3672 }, { "epoch": 0.09, "learning_rate": 1.9808553031162708e-05, "loss": 2.3684, "step": 3673 }, { "epoch": 0.09, "learning_rate": 1.9808397534528483e-05, "loss": 2.18, "step": 3674 }, { "epoch": 0.09, "learning_rate": 1.9808241975382146e-05, "loss": 2.312, "step": 3675 }, { "epoch": 0.09, "learning_rate": 1.980808635372468e-05, "loss": 2.2099, "step": 3676 }, { "epoch": 0.09, "learning_rate": 1.980793066955709e-05, "loss": 2.4659, "step": 3677 }, { "epoch": 0.09, "learning_rate": 1.980777492288036e-05, "loss": 2.4426, "step": 3678 }, { "epoch": 0.09, "learning_rate": 1.9807619113695483e-05, "loss": 2.3565, "step": 3679 }, { "epoch": 0.09, "learning_rate": 1.9807463242003454e-05, "loss": 2.2889, "step": 3680 }, { "epoch": 0.09, "learning_rate": 1.980730730780527e-05, "loss": 2.3345, "step": 3681 }, { "epoch": 0.09, "learning_rate": 1.980715131110191e-05, "loss": 2.3663, "step": 3682 }, { "epoch": 0.09, "learning_rate": 1.9806995251894386e-05, "loss": 2.2503, "step": 3683 }, { "epoch": 0.09, "learning_rate": 1.9806839130183687e-05, "loss": 2.2435, "step": 3684 }, { "epoch": 0.09, "learning_rate": 1.9806682945970802e-05, "loss": 2.2966, "step": 3685 }, { "epoch": 0.09, "learning_rate": 1.980652669925673e-05, "loss": 2.0788, "step": 3686 }, { "epoch": 0.09, "learning_rate": 1.980637039004247e-05, "loss": 2.3021, "step": 3687 }, { "epoch": 0.09, "learning_rate": 1.9806214018329016e-05, "loss": 2.4267, "step": 3688 }, { "epoch": 0.09, "learning_rate": 1.9806057584117362e-05, "loss": 2.2919, "step": 3689 }, { "epoch": 0.09, "learning_rate": 1.980590108740851e-05, "loss": 2.2366, "step": 3690 }, { "epoch": 0.09, "learning_rate": 1.9805744528203454e-05, "loss": 2.2297, "step": 3691 }, { "epoch": 0.09, "learning_rate": 1.9805587906503188e-05, "loss": 2.4115, "step": 3692 }, { "epoch": 0.09, "learning_rate": 1.9805431222308723e-05, "loss": 2.4131, "step": 3693 }, { "epoch": 0.09, "learning_rate": 1.9805274475621045e-05, "loss": 2.2541, "step": 3694 }, { "epoch": 0.09, "learning_rate": 1.9805117666441153e-05, "loss": 2.264, "step": 3695 }, { "epoch": 0.09, "learning_rate": 1.9804960794770057e-05, "loss": 2.4258, "step": 3696 }, { "epoch": 0.09, "learning_rate": 1.9804803860608746e-05, "loss": 2.3179, "step": 3697 }, { "epoch": 0.09, "learning_rate": 1.980464686395823e-05, "loss": 2.3687, "step": 3698 }, { "epoch": 0.09, "learning_rate": 1.98044898048195e-05, "loss": 2.2792, "step": 3699 }, { "epoch": 0.09, "learning_rate": 1.9804332683193563e-05, "loss": 2.0929, "step": 3700 }, { "epoch": 0.09, "learning_rate": 1.980417549908142e-05, "loss": 2.2663, "step": 3701 }, { "epoch": 0.09, "learning_rate": 1.9804018252484067e-05, "loss": 2.3319, "step": 3702 }, { "epoch": 0.09, "learning_rate": 1.9803860943402514e-05, "loss": 2.5138, "step": 3703 }, { "epoch": 0.09, "learning_rate": 1.980370357183776e-05, "loss": 2.4099, "step": 3704 }, { "epoch": 0.09, "learning_rate": 1.9803546137790807e-05, "loss": 2.4249, "step": 3705 }, { "epoch": 0.09, "learning_rate": 1.9803388641262662e-05, "loss": 2.292, "step": 3706 }, { "epoch": 0.09, "learning_rate": 1.9803231082254324e-05, "loss": 2.4023, "step": 3707 }, { "epoch": 0.09, "learning_rate": 1.98030734607668e-05, "loss": 2.467, "step": 3708 }, { "epoch": 0.09, "learning_rate": 1.9802915776801096e-05, "loss": 2.2874, "step": 3709 }, { "epoch": 0.09, "learning_rate": 1.9802758030358214e-05, "loss": 2.3325, "step": 3710 }, { "epoch": 0.09, "learning_rate": 1.980260022143916e-05, "loss": 2.5252, "step": 3711 }, { "epoch": 0.09, "learning_rate": 1.9802442350044938e-05, "loss": 2.4039, "step": 3712 }, { "epoch": 0.09, "learning_rate": 1.980228441617656e-05, "loss": 2.3328, "step": 3713 }, { "epoch": 0.09, "learning_rate": 1.9802126419835027e-05, "loss": 2.4312, "step": 3714 }, { "epoch": 0.09, "learning_rate": 1.980196836102135e-05, "loss": 2.3162, "step": 3715 }, { "epoch": 0.09, "learning_rate": 1.980181023973653e-05, "loss": 2.4233, "step": 3716 }, { "epoch": 0.09, "learning_rate": 1.9801652055981582e-05, "loss": 2.4672, "step": 3717 }, { "epoch": 0.09, "learning_rate": 1.980149380975751e-05, "loss": 2.426, "step": 3718 }, { "epoch": 0.09, "learning_rate": 1.9801335501065327e-05, "loss": 2.5275, "step": 3719 }, { "epoch": 0.09, "learning_rate": 1.9801177129906034e-05, "loss": 2.3399, "step": 3720 }, { "epoch": 0.09, "learning_rate": 1.9801018696280647e-05, "loss": 2.2315, "step": 3721 }, { "epoch": 0.09, "learning_rate": 1.980086020019018e-05, "loss": 2.3784, "step": 3722 }, { "epoch": 0.09, "learning_rate": 1.9800701641635627e-05, "loss": 2.5206, "step": 3723 }, { "epoch": 0.09, "learning_rate": 1.9800543020618014e-05, "loss": 2.3528, "step": 3724 }, { "epoch": 0.09, "learning_rate": 1.9800384337138345e-05, "loss": 2.223, "step": 3725 }, { "epoch": 0.09, "learning_rate": 1.980022559119763e-05, "loss": 2.1571, "step": 3726 }, { "epoch": 0.09, "learning_rate": 1.9800066782796886e-05, "loss": 2.2782, "step": 3727 }, { "epoch": 0.09, "learning_rate": 1.9799907911937124e-05, "loss": 2.222, "step": 3728 }, { "epoch": 0.09, "learning_rate": 1.9799748978619354e-05, "loss": 2.3122, "step": 3729 }, { "epoch": 0.09, "learning_rate": 1.9799589982844593e-05, "loss": 2.4214, "step": 3730 }, { "epoch": 0.09, "learning_rate": 1.9799430924613844e-05, "loss": 2.1615, "step": 3731 }, { "epoch": 0.09, "learning_rate": 1.9799271803928134e-05, "loss": 2.266, "step": 3732 }, { "epoch": 0.09, "learning_rate": 1.9799112620788473e-05, "loss": 2.2038, "step": 3733 }, { "epoch": 0.09, "learning_rate": 1.979895337519587e-05, "loss": 2.3612, "step": 3734 }, { "epoch": 0.09, "learning_rate": 1.9798794067151343e-05, "loss": 2.3806, "step": 3735 }, { "epoch": 0.09, "learning_rate": 1.979863469665591e-05, "loss": 2.4015, "step": 3736 }, { "epoch": 0.09, "learning_rate": 1.9798475263710584e-05, "loss": 2.3317, "step": 3737 }, { "epoch": 0.09, "learning_rate": 1.9798315768316384e-05, "loss": 2.3803, "step": 3738 }, { "epoch": 0.09, "learning_rate": 1.979815621047432e-05, "loss": 2.466, "step": 3739 }, { "epoch": 0.09, "learning_rate": 1.979799659018542e-05, "loss": 2.3623, "step": 3740 }, { "epoch": 0.09, "learning_rate": 1.979783690745069e-05, "loss": 2.347, "step": 3741 }, { "epoch": 0.09, "learning_rate": 1.9797677162271152e-05, "loss": 2.3601, "step": 3742 }, { "epoch": 0.09, "learning_rate": 1.9797517354647828e-05, "loss": 2.2918, "step": 3743 }, { "epoch": 0.09, "learning_rate": 1.979735748458173e-05, "loss": 2.3987, "step": 3744 }, { "epoch": 0.09, "learning_rate": 1.9797197552073882e-05, "loss": 2.3632, "step": 3745 }, { "epoch": 0.09, "learning_rate": 1.9797037557125297e-05, "loss": 2.3133, "step": 3746 }, { "epoch": 0.09, "learning_rate": 1.9796877499737005e-05, "loss": 2.4062, "step": 3747 }, { "epoch": 0.09, "learning_rate": 1.9796717379910014e-05, "loss": 2.2855, "step": 3748 }, { "epoch": 0.09, "learning_rate": 1.9796557197645355e-05, "loss": 2.133, "step": 3749 }, { "epoch": 0.09, "learning_rate": 1.979639695294404e-05, "loss": 2.4014, "step": 3750 }, { "epoch": 0.09, "learning_rate": 1.9796236645807098e-05, "loss": 2.3638, "step": 3751 }, { "epoch": 0.09, "learning_rate": 1.9796076276235546e-05, "loss": 2.4116, "step": 3752 }, { "epoch": 0.09, "learning_rate": 1.979591584423041e-05, "loss": 2.2463, "step": 3753 }, { "epoch": 0.09, "learning_rate": 1.9795755349792706e-05, "loss": 2.3534, "step": 3754 }, { "epoch": 0.09, "learning_rate": 1.9795594792923464e-05, "loss": 2.5528, "step": 3755 }, { "epoch": 0.09, "learning_rate": 1.9795434173623702e-05, "loss": 2.3213, "step": 3756 }, { "epoch": 0.09, "learning_rate": 1.9795273491894448e-05, "loss": 2.2525, "step": 3757 }, { "epoch": 0.09, "learning_rate": 1.979511274773672e-05, "loss": 2.4261, "step": 3758 }, { "epoch": 0.09, "learning_rate": 1.9794951941151547e-05, "loss": 2.0353, "step": 3759 }, { "epoch": 0.09, "learning_rate": 1.9794791072139957e-05, "loss": 2.2764, "step": 3760 }, { "epoch": 0.09, "learning_rate": 1.979463014070297e-05, "loss": 2.3681, "step": 3761 }, { "epoch": 0.09, "learning_rate": 1.979446914684161e-05, "loss": 2.1107, "step": 3762 }, { "epoch": 0.09, "learning_rate": 1.9794308090556912e-05, "loss": 2.3524, "step": 3763 }, { "epoch": 0.09, "learning_rate": 1.9794146971849893e-05, "loss": 2.3687, "step": 3764 }, { "epoch": 0.09, "learning_rate": 1.9793985790721583e-05, "loss": 2.4954, "step": 3765 }, { "epoch": 0.09, "learning_rate": 1.979382454717301e-05, "loss": 2.2405, "step": 3766 }, { "epoch": 0.09, "learning_rate": 1.9793663241205204e-05, "loss": 2.2695, "step": 3767 }, { "epoch": 0.09, "learning_rate": 1.9793501872819187e-05, "loss": 2.2886, "step": 3768 }, { "epoch": 0.09, "learning_rate": 1.9793340442015994e-05, "loss": 2.4341, "step": 3769 }, { "epoch": 0.09, "learning_rate": 1.9793178948796646e-05, "loss": 2.1665, "step": 3770 }, { "epoch": 0.09, "learning_rate": 1.9793017393162182e-05, "loss": 2.3671, "step": 3771 }, { "epoch": 0.09, "learning_rate": 1.9792855775113628e-05, "loss": 2.4411, "step": 3772 }, { "epoch": 0.09, "learning_rate": 1.979269409465201e-05, "loss": 2.2891, "step": 3773 }, { "epoch": 0.09, "learning_rate": 1.9792532351778358e-05, "loss": 2.314, "step": 3774 }, { "epoch": 0.09, "learning_rate": 1.979237054649371e-05, "loss": 2.4719, "step": 3775 }, { "epoch": 0.09, "learning_rate": 1.979220867879909e-05, "loss": 2.2851, "step": 3776 }, { "epoch": 0.09, "learning_rate": 1.9792046748695538e-05, "loss": 2.1744, "step": 3777 }, { "epoch": 0.09, "learning_rate": 1.979188475618408e-05, "loss": 2.3264, "step": 3778 }, { "epoch": 0.09, "learning_rate": 1.979172270126575e-05, "loss": 2.2105, "step": 3779 }, { "epoch": 0.09, "learning_rate": 1.9791560583941572e-05, "loss": 2.217, "step": 3780 }, { "epoch": 0.09, "learning_rate": 1.9791398404212596e-05, "loss": 2.358, "step": 3781 }, { "epoch": 0.09, "learning_rate": 1.9791236162079842e-05, "loss": 2.4141, "step": 3782 }, { "epoch": 0.09, "learning_rate": 1.9791073857544352e-05, "loss": 2.2677, "step": 3783 }, { "epoch": 0.09, "learning_rate": 1.979091149060716e-05, "loss": 2.4261, "step": 3784 }, { "epoch": 0.09, "learning_rate": 1.9790749061269296e-05, "loss": 2.2753, "step": 3785 }, { "epoch": 0.09, "learning_rate": 1.9790586569531792e-05, "loss": 2.2483, "step": 3786 }, { "epoch": 0.09, "learning_rate": 1.9790424015395696e-05, "loss": 2.4694, "step": 3787 }, { "epoch": 0.09, "learning_rate": 1.9790261398862034e-05, "loss": 2.5676, "step": 3788 }, { "epoch": 0.09, "learning_rate": 1.9790098719931845e-05, "loss": 2.364, "step": 3789 }, { "epoch": 0.09, "learning_rate": 1.978993597860617e-05, "loss": 2.3597, "step": 3790 }, { "epoch": 0.09, "learning_rate": 1.978977317488604e-05, "loss": 2.3024, "step": 3791 }, { "epoch": 0.09, "learning_rate": 1.9789610308772497e-05, "loss": 2.3135, "step": 3792 }, { "epoch": 0.09, "learning_rate": 1.9789447380266574e-05, "loss": 2.4041, "step": 3793 }, { "epoch": 0.09, "learning_rate": 1.9789284389369313e-05, "loss": 2.3934, "step": 3794 }, { "epoch": 0.09, "learning_rate": 1.9789121336081756e-05, "loss": 2.1607, "step": 3795 }, { "epoch": 0.09, "learning_rate": 1.9788958220404937e-05, "loss": 2.1614, "step": 3796 }, { "epoch": 0.09, "learning_rate": 1.9788795042339893e-05, "loss": 2.2991, "step": 3797 }, { "epoch": 0.09, "learning_rate": 1.9788631801887674e-05, "loss": 2.3473, "step": 3798 }, { "epoch": 0.09, "learning_rate": 1.978846849904931e-05, "loss": 2.3603, "step": 3799 }, { "epoch": 0.09, "learning_rate": 1.9788305133825848e-05, "loss": 2.2927, "step": 3800 }, { "epoch": 0.09, "learning_rate": 1.9788141706218328e-05, "loss": 2.444, "step": 3801 }, { "epoch": 0.09, "learning_rate": 1.978797821622779e-05, "loss": 2.2082, "step": 3802 }, { "epoch": 0.09, "learning_rate": 1.9787814663855278e-05, "loss": 2.2795, "step": 3803 }, { "epoch": 0.09, "learning_rate": 1.9787651049101835e-05, "loss": 2.6093, "step": 3804 }, { "epoch": 0.09, "learning_rate": 1.97874873719685e-05, "loss": 2.3825, "step": 3805 }, { "epoch": 0.09, "learning_rate": 1.978732363245632e-05, "loss": 2.3582, "step": 3806 }, { "epoch": 0.09, "learning_rate": 1.9787159830566332e-05, "loss": 2.4357, "step": 3807 }, { "epoch": 0.09, "learning_rate": 1.9786995966299593e-05, "loss": 2.4048, "step": 3808 }, { "epoch": 0.09, "learning_rate": 1.978683203965713e-05, "loss": 2.2076, "step": 3809 }, { "epoch": 0.09, "learning_rate": 1.9786668050640004e-05, "loss": 2.39, "step": 3810 }, { "epoch": 0.09, "learning_rate": 1.978650399924925e-05, "loss": 2.4496, "step": 3811 }, { "epoch": 0.09, "learning_rate": 1.9786339885485916e-05, "loss": 2.2945, "step": 3812 }, { "epoch": 0.09, "learning_rate": 1.978617570935105e-05, "loss": 2.4506, "step": 3813 }, { "epoch": 0.09, "learning_rate": 1.97860114708457e-05, "loss": 2.367, "step": 3814 }, { "epoch": 0.09, "learning_rate": 1.9785847169970905e-05, "loss": 2.4645, "step": 3815 }, { "epoch": 0.09, "learning_rate": 1.9785682806727718e-05, "loss": 2.3617, "step": 3816 }, { "epoch": 0.09, "learning_rate": 1.9785518381117187e-05, "loss": 2.3455, "step": 3817 }, { "epoch": 0.09, "learning_rate": 1.978535389314036e-05, "loss": 2.4947, "step": 3818 }, { "epoch": 0.09, "learning_rate": 1.9785189342798277e-05, "loss": 2.2615, "step": 3819 }, { "epoch": 0.09, "learning_rate": 1.9785024730092e-05, "loss": 2.3039, "step": 3820 }, { "epoch": 0.09, "learning_rate": 1.9784860055022567e-05, "loss": 2.5551, "step": 3821 }, { "epoch": 0.09, "learning_rate": 1.9784695317591033e-05, "loss": 2.4098, "step": 3822 }, { "epoch": 0.09, "learning_rate": 1.978453051779845e-05, "loss": 2.4377, "step": 3823 }, { "epoch": 0.09, "learning_rate": 1.9784365655645864e-05, "loss": 2.4581, "step": 3824 }, { "epoch": 0.09, "learning_rate": 1.9784200731134326e-05, "loss": 2.494, "step": 3825 }, { "epoch": 0.09, "learning_rate": 1.9784035744264887e-05, "loss": 2.3045, "step": 3826 }, { "epoch": 0.09, "learning_rate": 1.9783870695038604e-05, "loss": 2.2894, "step": 3827 }, { "epoch": 0.09, "learning_rate": 1.978370558345652e-05, "loss": 2.4257, "step": 3828 }, { "epoch": 0.09, "learning_rate": 1.9783540409519695e-05, "loss": 2.3684, "step": 3829 }, { "epoch": 0.09, "learning_rate": 1.9783375173229177e-05, "loss": 2.3071, "step": 3830 }, { "epoch": 0.09, "learning_rate": 1.978320987458602e-05, "loss": 2.5651, "step": 3831 }, { "epoch": 0.09, "learning_rate": 1.978304451359128e-05, "loss": 2.2746, "step": 3832 }, { "epoch": 0.09, "learning_rate": 1.9782879090246006e-05, "loss": 2.4136, "step": 3833 }, { "epoch": 0.09, "learning_rate": 1.978271360455126e-05, "loss": 2.2087, "step": 3834 }, { "epoch": 0.09, "learning_rate": 1.978254805650809e-05, "loss": 2.4419, "step": 3835 }, { "epoch": 0.09, "learning_rate": 1.9782382446117552e-05, "loss": 2.3957, "step": 3836 }, { "epoch": 0.09, "learning_rate": 1.9782216773380704e-05, "loss": 2.381, "step": 3837 }, { "epoch": 0.09, "learning_rate": 1.97820510382986e-05, "loss": 2.4893, "step": 3838 }, { "epoch": 0.09, "learning_rate": 1.9781885240872298e-05, "loss": 2.1361, "step": 3839 }, { "epoch": 0.09, "learning_rate": 1.9781719381102853e-05, "loss": 2.5265, "step": 3840 }, { "epoch": 0.09, "learning_rate": 1.9781553458991323e-05, "loss": 2.3737, "step": 3841 }, { "epoch": 0.09, "learning_rate": 1.978138747453876e-05, "loss": 2.3319, "step": 3842 }, { "epoch": 0.09, "learning_rate": 1.9781221427746235e-05, "loss": 2.3559, "step": 3843 }, { "epoch": 0.09, "learning_rate": 1.9781055318614794e-05, "loss": 2.2195, "step": 3844 }, { "epoch": 0.09, "learning_rate": 1.9780889147145502e-05, "loss": 2.4274, "step": 3845 }, { "epoch": 0.09, "learning_rate": 1.9780722913339413e-05, "loss": 2.4772, "step": 3846 }, { "epoch": 0.09, "learning_rate": 1.9780556617197592e-05, "loss": 2.3168, "step": 3847 }, { "epoch": 0.09, "learning_rate": 1.9780390258721095e-05, "loss": 2.3965, "step": 3848 }, { "epoch": 0.09, "learning_rate": 1.9780223837910982e-05, "loss": 2.3223, "step": 3849 }, { "epoch": 0.09, "learning_rate": 1.9780057354768318e-05, "loss": 2.5737, "step": 3850 }, { "epoch": 0.09, "learning_rate": 1.977989080929416e-05, "loss": 2.1356, "step": 3851 }, { "epoch": 0.09, "learning_rate": 1.977972420148957e-05, "loss": 2.3705, "step": 3852 }, { "epoch": 0.09, "learning_rate": 1.9779557531355615e-05, "loss": 2.3634, "step": 3853 }, { "epoch": 0.09, "learning_rate": 1.9779390798893346e-05, "loss": 2.3158, "step": 3854 }, { "epoch": 0.1, "learning_rate": 1.9779224004103835e-05, "loss": 2.3399, "step": 3855 }, { "epoch": 0.1, "learning_rate": 1.9779057146988142e-05, "loss": 2.4921, "step": 3856 }, { "epoch": 0.1, "learning_rate": 1.9778890227547335e-05, "loss": 2.1601, "step": 3857 }, { "epoch": 0.1, "learning_rate": 1.9778723245782473e-05, "loss": 2.2755, "step": 3858 }, { "epoch": 0.1, "learning_rate": 1.9778556201694615e-05, "loss": 2.3322, "step": 3859 }, { "epoch": 0.1, "learning_rate": 1.9778389095284837e-05, "loss": 2.477, "step": 3860 }, { "epoch": 0.1, "learning_rate": 1.9778221926554197e-05, "loss": 2.4708, "step": 3861 }, { "epoch": 0.1, "learning_rate": 1.977805469550376e-05, "loss": 2.3746, "step": 3862 }, { "epoch": 0.1, "learning_rate": 1.9777887402134598e-05, "loss": 2.43, "step": 3863 }, { "epoch": 0.1, "learning_rate": 1.977772004644777e-05, "loss": 2.2995, "step": 3864 }, { "epoch": 0.1, "learning_rate": 1.9777552628444346e-05, "loss": 2.4275, "step": 3865 }, { "epoch": 0.1, "learning_rate": 1.9777385148125395e-05, "loss": 2.3262, "step": 3866 }, { "epoch": 0.1, "learning_rate": 1.977721760549198e-05, "loss": 2.2819, "step": 3867 }, { "epoch": 0.1, "learning_rate": 1.977705000054517e-05, "loss": 2.3613, "step": 3868 }, { "epoch": 0.1, "learning_rate": 1.9776882333286036e-05, "loss": 2.5178, "step": 3869 }, { "epoch": 0.1, "learning_rate": 1.9776714603715647e-05, "loss": 2.2343, "step": 3870 }, { "epoch": 0.1, "learning_rate": 1.9776546811835064e-05, "loss": 2.3433, "step": 3871 }, { "epoch": 0.1, "learning_rate": 1.9776378957645366e-05, "loss": 2.3515, "step": 3872 }, { "epoch": 0.1, "learning_rate": 1.9776211041147616e-05, "loss": 2.4062, "step": 3873 }, { "epoch": 0.1, "learning_rate": 1.9776043062342888e-05, "loss": 2.3622, "step": 3874 }, { "epoch": 0.1, "learning_rate": 1.9775875021232253e-05, "loss": 2.2728, "step": 3875 }, { "epoch": 0.1, "learning_rate": 1.977570691781678e-05, "loss": 2.1769, "step": 3876 }, { "epoch": 0.1, "learning_rate": 1.977553875209754e-05, "loss": 2.3359, "step": 3877 }, { "epoch": 0.1, "learning_rate": 1.9775370524075607e-05, "loss": 2.2896, "step": 3878 }, { "epoch": 0.1, "learning_rate": 1.9775202233752053e-05, "loss": 2.356, "step": 3879 }, { "epoch": 0.1, "learning_rate": 1.9775033881127948e-05, "loss": 2.456, "step": 3880 }, { "epoch": 0.1, "learning_rate": 1.9774865466204366e-05, "loss": 2.2987, "step": 3881 }, { "epoch": 0.1, "learning_rate": 1.9774696988982382e-05, "loss": 2.231, "step": 3882 }, { "epoch": 0.1, "learning_rate": 1.977452844946307e-05, "loss": 2.4192, "step": 3883 }, { "epoch": 0.1, "learning_rate": 1.9774359847647496e-05, "loss": 2.3326, "step": 3884 }, { "epoch": 0.1, "learning_rate": 1.977419118353675e-05, "loss": 2.2436, "step": 3885 }, { "epoch": 0.1, "learning_rate": 1.9774022457131895e-05, "loss": 2.4179, "step": 3886 }, { "epoch": 0.1, "learning_rate": 1.9773853668434006e-05, "loss": 2.491, "step": 3887 }, { "epoch": 0.1, "learning_rate": 1.977368481744417e-05, "loss": 2.2966, "step": 3888 }, { "epoch": 0.1, "learning_rate": 1.977351590416345e-05, "loss": 2.3088, "step": 3889 }, { "epoch": 0.1, "learning_rate": 1.9773346928592928e-05, "loss": 2.4241, "step": 3890 }, { "epoch": 0.1, "learning_rate": 1.977317789073368e-05, "loss": 2.4212, "step": 3891 }, { "epoch": 0.1, "learning_rate": 1.977300879058679e-05, "loss": 2.2416, "step": 3892 }, { "epoch": 0.1, "learning_rate": 1.9772839628153323e-05, "loss": 2.3779, "step": 3893 }, { "epoch": 0.1, "learning_rate": 1.977267040343437e-05, "loss": 2.4363, "step": 3894 }, { "epoch": 0.1, "learning_rate": 1.9772501116431e-05, "loss": 2.309, "step": 3895 }, { "epoch": 0.1, "learning_rate": 1.9772331767144298e-05, "loss": 2.417, "step": 3896 }, { "epoch": 0.1, "learning_rate": 1.977216235557534e-05, "loss": 2.3673, "step": 3897 }, { "epoch": 0.1, "learning_rate": 1.9771992881725202e-05, "loss": 2.2616, "step": 3898 }, { "epoch": 0.1, "learning_rate": 1.9771823345594978e-05, "loss": 2.3008, "step": 3899 }, { "epoch": 0.1, "learning_rate": 1.9771653747185734e-05, "loss": 2.4534, "step": 3900 }, { "epoch": 0.1, "learning_rate": 1.9771484086498556e-05, "loss": 2.5867, "step": 3901 }, { "epoch": 0.1, "learning_rate": 1.9771314363534527e-05, "loss": 2.277, "step": 3902 }, { "epoch": 0.1, "learning_rate": 1.9771144578294724e-05, "loss": 2.3038, "step": 3903 }, { "epoch": 0.1, "learning_rate": 1.9770974730780236e-05, "loss": 2.2398, "step": 3904 }, { "epoch": 0.1, "learning_rate": 1.977080482099214e-05, "loss": 2.3301, "step": 3905 }, { "epoch": 0.1, "learning_rate": 1.9770634848931524e-05, "loss": 2.3225, "step": 3906 }, { "epoch": 0.1, "learning_rate": 1.9770464814599464e-05, "loss": 2.4688, "step": 3907 }, { "epoch": 0.1, "learning_rate": 1.977029471799705e-05, "loss": 2.2716, "step": 3908 }, { "epoch": 0.1, "learning_rate": 1.9770124559125362e-05, "loss": 2.3341, "step": 3909 }, { "epoch": 0.1, "learning_rate": 1.9769954337985485e-05, "loss": 2.3218, "step": 3910 }, { "epoch": 0.1, "learning_rate": 1.9769784054578505e-05, "loss": 2.2431, "step": 3911 }, { "epoch": 0.1, "learning_rate": 1.976961370890551e-05, "loss": 2.5026, "step": 3912 }, { "epoch": 0.1, "learning_rate": 1.9769443300967583e-05, "loss": 2.2759, "step": 3913 }, { "epoch": 0.1, "learning_rate": 1.976927283076581e-05, "loss": 2.3253, "step": 3914 }, { "epoch": 0.1, "learning_rate": 1.9769102298301273e-05, "loss": 2.5119, "step": 3915 }, { "epoch": 0.1, "learning_rate": 1.976893170357507e-05, "loss": 2.4422, "step": 3916 }, { "epoch": 0.1, "learning_rate": 1.976876104658828e-05, "loss": 2.3682, "step": 3917 }, { "epoch": 0.1, "learning_rate": 1.9768590327341986e-05, "loss": 2.2721, "step": 3918 }, { "epoch": 0.1, "learning_rate": 1.9768419545837288e-05, "loss": 2.4441, "step": 3919 }, { "epoch": 0.1, "learning_rate": 1.976824870207527e-05, "loss": 2.1179, "step": 3920 }, { "epoch": 0.1, "learning_rate": 1.9768077796057016e-05, "loss": 2.6184, "step": 3921 }, { "epoch": 0.1, "learning_rate": 1.9767906827783623e-05, "loss": 2.5883, "step": 3922 }, { "epoch": 0.1, "learning_rate": 1.976773579725617e-05, "loss": 2.3401, "step": 3923 }, { "epoch": 0.1, "learning_rate": 1.9767564704475762e-05, "loss": 2.4276, "step": 3924 }, { "epoch": 0.1, "learning_rate": 1.9767393549443473e-05, "loss": 2.3544, "step": 3925 }, { "epoch": 0.1, "learning_rate": 1.976722233216041e-05, "loss": 2.4035, "step": 3926 }, { "epoch": 0.1, "learning_rate": 1.976705105262765e-05, "loss": 2.4357, "step": 3927 }, { "epoch": 0.1, "learning_rate": 1.976687971084629e-05, "loss": 2.2543, "step": 3928 }, { "epoch": 0.1, "learning_rate": 1.9766708306817424e-05, "loss": 2.2258, "step": 3929 }, { "epoch": 0.1, "learning_rate": 1.9766536840542147e-05, "loss": 2.3299, "step": 3930 }, { "epoch": 0.1, "learning_rate": 1.9766365312021546e-05, "loss": 2.0786, "step": 3931 }, { "epoch": 0.1, "learning_rate": 1.976619372125671e-05, "loss": 2.2811, "step": 3932 }, { "epoch": 0.1, "learning_rate": 1.9766022068248745e-05, "loss": 2.3767, "step": 3933 }, { "epoch": 0.1, "learning_rate": 1.9765850352998738e-05, "loss": 2.4254, "step": 3934 }, { "epoch": 0.1, "learning_rate": 1.9765678575507784e-05, "loss": 2.3571, "step": 3935 }, { "epoch": 0.1, "learning_rate": 1.9765506735776978e-05, "loss": 2.3734, "step": 3936 }, { "epoch": 0.1, "learning_rate": 1.9765334833807414e-05, "loss": 2.3397, "step": 3937 }, { "epoch": 0.1, "learning_rate": 1.9765162869600186e-05, "loss": 2.4182, "step": 3938 }, { "epoch": 0.1, "learning_rate": 1.9764990843156397e-05, "loss": 2.433, "step": 3939 }, { "epoch": 0.1, "learning_rate": 1.976481875447714e-05, "loss": 2.228, "step": 3940 }, { "epoch": 0.1, "learning_rate": 1.9764646603563508e-05, "loss": 2.2248, "step": 3941 }, { "epoch": 0.1, "learning_rate": 1.97644743904166e-05, "loss": 2.4127, "step": 3942 }, { "epoch": 0.1, "learning_rate": 1.976430211503752e-05, "loss": 2.1877, "step": 3943 }, { "epoch": 0.1, "learning_rate": 1.9764129777427353e-05, "loss": 2.41, "step": 3944 }, { "epoch": 0.1, "learning_rate": 1.9763957377587208e-05, "loss": 2.3969, "step": 3945 }, { "epoch": 0.1, "learning_rate": 1.9763784915518182e-05, "loss": 2.4182, "step": 3946 }, { "epoch": 0.1, "learning_rate": 1.9763612391221372e-05, "loss": 2.2749, "step": 3947 }, { "epoch": 0.1, "learning_rate": 1.9763439804697876e-05, "loss": 2.3913, "step": 3948 }, { "epoch": 0.1, "learning_rate": 1.9763267155948802e-05, "loss": 2.5717, "step": 3949 }, { "epoch": 0.1, "learning_rate": 1.9763094444975238e-05, "loss": 2.346, "step": 3950 }, { "epoch": 0.1, "learning_rate": 1.9762921671778298e-05, "loss": 2.4077, "step": 3951 }, { "epoch": 0.1, "learning_rate": 1.9762748836359073e-05, "loss": 2.337, "step": 3952 }, { "epoch": 0.1, "learning_rate": 1.9762575938718668e-05, "loss": 2.0625, "step": 3953 }, { "epoch": 0.1, "learning_rate": 1.9762402978858187e-05, "loss": 2.3885, "step": 3954 }, { "epoch": 0.1, "learning_rate": 1.976222995677873e-05, "loss": 2.4032, "step": 3955 }, { "epoch": 0.1, "learning_rate": 1.9762056872481396e-05, "loss": 2.3451, "step": 3956 }, { "epoch": 0.1, "learning_rate": 1.9761883725967295e-05, "loss": 2.3756, "step": 3957 }, { "epoch": 0.1, "learning_rate": 1.9761710517237527e-05, "loss": 2.3153, "step": 3958 }, { "epoch": 0.1, "learning_rate": 1.9761537246293198e-05, "loss": 2.524, "step": 3959 }, { "epoch": 0.1, "learning_rate": 1.976136391313541e-05, "loss": 2.4099, "step": 3960 }, { "epoch": 0.1, "learning_rate": 1.976119051776527e-05, "loss": 2.3261, "step": 3961 }, { "epoch": 0.1, "learning_rate": 1.976101706018388e-05, "loss": 2.2987, "step": 3962 }, { "epoch": 0.1, "learning_rate": 1.9760843540392346e-05, "loss": 2.3399, "step": 3963 }, { "epoch": 0.1, "learning_rate": 1.976066995839178e-05, "loss": 2.303, "step": 3964 }, { "epoch": 0.1, "learning_rate": 1.9760496314183278e-05, "loss": 2.3845, "step": 3965 }, { "epoch": 0.1, "learning_rate": 1.9760322607767954e-05, "loss": 2.366, "step": 3966 }, { "epoch": 0.1, "learning_rate": 1.9760148839146917e-05, "loss": 2.2899, "step": 3967 }, { "epoch": 0.1, "learning_rate": 1.9759975008321267e-05, "loss": 2.2967, "step": 3968 }, { "epoch": 0.1, "learning_rate": 1.9759801115292114e-05, "loss": 2.2003, "step": 3969 }, { "epoch": 0.1, "learning_rate": 1.975962716006057e-05, "loss": 2.2549, "step": 3970 }, { "epoch": 0.1, "learning_rate": 1.975945314262774e-05, "loss": 2.18, "step": 3971 }, { "epoch": 0.1, "learning_rate": 1.975927906299474e-05, "loss": 2.2664, "step": 3972 }, { "epoch": 0.1, "learning_rate": 1.9759104921162668e-05, "loss": 2.2854, "step": 3973 }, { "epoch": 0.1, "learning_rate": 1.9758930717132642e-05, "loss": 2.2811, "step": 3974 }, { "epoch": 0.1, "learning_rate": 1.975875645090577e-05, "loss": 2.3965, "step": 3975 }, { "epoch": 0.1, "learning_rate": 1.9758582122483165e-05, "loss": 2.3448, "step": 3976 }, { "epoch": 0.1, "learning_rate": 1.9758407731865933e-05, "loss": 2.2852, "step": 3977 }, { "epoch": 0.1, "learning_rate": 1.975823327905519e-05, "loss": 2.4351, "step": 3978 }, { "epoch": 0.1, "learning_rate": 1.9758058764052045e-05, "loss": 2.2324, "step": 3979 }, { "epoch": 0.1, "learning_rate": 1.9757884186857612e-05, "loss": 2.365, "step": 3980 }, { "epoch": 0.1, "learning_rate": 1.9757709547473003e-05, "loss": 2.3341, "step": 3981 }, { "epoch": 0.1, "learning_rate": 1.975753484589933e-05, "loss": 2.2003, "step": 3982 }, { "epoch": 0.1, "learning_rate": 1.9757360082137712e-05, "loss": 2.1781, "step": 3983 }, { "epoch": 0.1, "learning_rate": 1.9757185256189252e-05, "loss": 2.3083, "step": 3984 }, { "epoch": 0.1, "learning_rate": 1.9757010368055075e-05, "loss": 2.3703, "step": 3985 }, { "epoch": 0.1, "learning_rate": 1.975683541773629e-05, "loss": 2.271, "step": 3986 }, { "epoch": 0.1, "learning_rate": 1.975666040523401e-05, "loss": 2.1451, "step": 3987 }, { "epoch": 0.1, "learning_rate": 1.975648533054936e-05, "loss": 2.2896, "step": 3988 }, { "epoch": 0.1, "learning_rate": 1.9756310193683445e-05, "loss": 2.1208, "step": 3989 }, { "epoch": 0.1, "learning_rate": 1.9756134994637388e-05, "loss": 2.3912, "step": 3990 }, { "epoch": 0.1, "learning_rate": 1.9755959733412303e-05, "loss": 2.3772, "step": 3991 }, { "epoch": 0.1, "learning_rate": 1.9755784410009308e-05, "loss": 2.4168, "step": 3992 }, { "epoch": 0.1, "learning_rate": 1.9755609024429513e-05, "loss": 2.2628, "step": 3993 }, { "epoch": 0.1, "learning_rate": 1.9755433576674048e-05, "loss": 2.1576, "step": 3994 }, { "epoch": 0.1, "learning_rate": 1.9755258066744027e-05, "loss": 2.5562, "step": 3995 }, { "epoch": 0.1, "learning_rate": 1.9755082494640564e-05, "loss": 2.3189, "step": 3996 }, { "epoch": 0.1, "learning_rate": 1.975490686036478e-05, "loss": 2.2616, "step": 3997 }, { "epoch": 0.1, "learning_rate": 1.97547311639178e-05, "loss": 2.2299, "step": 3998 }, { "epoch": 0.1, "learning_rate": 1.9754555405300738e-05, "loss": 2.1533, "step": 3999 }, { "epoch": 0.1, "learning_rate": 1.9754379584514717e-05, "loss": 2.2718, "step": 4000 }, { "epoch": 0.1, "learning_rate": 1.9754203701560852e-05, "loss": 2.2428, "step": 4001 }, { "epoch": 0.1, "learning_rate": 1.9754027756440268e-05, "loss": 2.3378, "step": 4002 }, { "epoch": 0.1, "learning_rate": 1.9753851749154092e-05, "loss": 2.3082, "step": 4003 }, { "epoch": 0.1, "learning_rate": 1.9753675679703434e-05, "loss": 2.2318, "step": 4004 }, { "epoch": 0.1, "learning_rate": 1.9753499548089423e-05, "loss": 2.3227, "step": 4005 }, { "epoch": 0.1, "learning_rate": 1.9753323354313183e-05, "loss": 2.3871, "step": 4006 }, { "epoch": 0.1, "learning_rate": 1.9753147098375837e-05, "loss": 2.4189, "step": 4007 }, { "epoch": 0.1, "learning_rate": 1.97529707802785e-05, "loss": 2.3847, "step": 4008 }, { "epoch": 0.1, "learning_rate": 1.9752794400022306e-05, "loss": 2.4624, "step": 4009 }, { "epoch": 0.1, "learning_rate": 1.9752617957608377e-05, "loss": 2.2374, "step": 4010 }, { "epoch": 0.1, "learning_rate": 1.9752441453037828e-05, "loss": 2.3792, "step": 4011 }, { "epoch": 0.1, "learning_rate": 1.97522648863118e-05, "loss": 2.2794, "step": 4012 }, { "epoch": 0.1, "learning_rate": 1.9752088257431403e-05, "loss": 2.2344, "step": 4013 }, { "epoch": 0.1, "learning_rate": 1.9751911566397772e-05, "loss": 2.1985, "step": 4014 }, { "epoch": 0.1, "learning_rate": 1.975173481321203e-05, "loss": 2.3272, "step": 4015 }, { "epoch": 0.1, "learning_rate": 1.9751557997875307e-05, "loss": 2.3005, "step": 4016 }, { "epoch": 0.1, "learning_rate": 1.975138112038872e-05, "loss": 2.26, "step": 4017 }, { "epoch": 0.1, "learning_rate": 1.975120418075341e-05, "loss": 2.2961, "step": 4018 }, { "epoch": 0.1, "learning_rate": 1.9751027178970498e-05, "loss": 2.3473, "step": 4019 }, { "epoch": 0.1, "learning_rate": 1.9750850115041105e-05, "loss": 2.3753, "step": 4020 }, { "epoch": 0.1, "learning_rate": 1.9750672988966373e-05, "loss": 2.2594, "step": 4021 }, { "epoch": 0.1, "learning_rate": 1.975049580074742e-05, "loss": 2.4254, "step": 4022 }, { "epoch": 0.1, "learning_rate": 1.9750318550385382e-05, "loss": 2.4195, "step": 4023 }, { "epoch": 0.1, "learning_rate": 1.9750141237881388e-05, "loss": 2.4205, "step": 4024 }, { "epoch": 0.1, "learning_rate": 1.9749963863236567e-05, "loss": 2.4133, "step": 4025 }, { "epoch": 0.1, "learning_rate": 1.9749786426452046e-05, "loss": 2.3169, "step": 4026 }, { "epoch": 0.1, "learning_rate": 1.974960892752896e-05, "loss": 2.2928, "step": 4027 }, { "epoch": 0.1, "learning_rate": 1.9749431366468438e-05, "loss": 2.2538, "step": 4028 }, { "epoch": 0.1, "learning_rate": 1.9749253743271613e-05, "loss": 2.1939, "step": 4029 }, { "epoch": 0.1, "learning_rate": 1.9749076057939616e-05, "loss": 2.2468, "step": 4030 }, { "epoch": 0.1, "learning_rate": 1.9748898310473582e-05, "loss": 2.354, "step": 4031 }, { "epoch": 0.1, "learning_rate": 1.9748720500874642e-05, "loss": 2.307, "step": 4032 }, { "epoch": 0.1, "learning_rate": 1.9748542629143927e-05, "loss": 2.4264, "step": 4033 }, { "epoch": 0.1, "learning_rate": 1.9748364695282577e-05, "loss": 2.5541, "step": 4034 }, { "epoch": 0.1, "learning_rate": 1.974818669929172e-05, "loss": 2.4948, "step": 4035 }, { "epoch": 0.1, "learning_rate": 1.9748008641172486e-05, "loss": 2.4124, "step": 4036 }, { "epoch": 0.1, "learning_rate": 1.9747830520926025e-05, "loss": 2.3893, "step": 4037 }, { "epoch": 0.1, "learning_rate": 1.974765233855346e-05, "loss": 2.2957, "step": 4038 }, { "epoch": 0.1, "learning_rate": 1.974747409405593e-05, "loss": 2.2196, "step": 4039 }, { "epoch": 0.1, "learning_rate": 1.9747295787434567e-05, "loss": 2.3029, "step": 4040 }, { "epoch": 0.1, "learning_rate": 1.974711741869052e-05, "loss": 2.4269, "step": 4041 }, { "epoch": 0.1, "learning_rate": 1.974693898782491e-05, "loss": 2.4258, "step": 4042 }, { "epoch": 0.1, "learning_rate": 1.9746760494838883e-05, "loss": 2.4119, "step": 4043 }, { "epoch": 0.1, "learning_rate": 1.9746581939733578e-05, "loss": 2.1029, "step": 4044 }, { "epoch": 0.1, "learning_rate": 1.9746403322510125e-05, "loss": 2.4329, "step": 4045 }, { "epoch": 0.1, "learning_rate": 1.974622464316967e-05, "loss": 2.3097, "step": 4046 }, { "epoch": 0.1, "learning_rate": 1.9746045901713352e-05, "loss": 2.3889, "step": 4047 }, { "epoch": 0.1, "learning_rate": 1.9745867098142302e-05, "loss": 2.3882, "step": 4048 }, { "epoch": 0.1, "learning_rate": 1.974568823245767e-05, "loss": 2.3995, "step": 4049 }, { "epoch": 0.1, "learning_rate": 1.9745509304660587e-05, "loss": 2.3462, "step": 4050 }, { "epoch": 0.1, "learning_rate": 1.97453303147522e-05, "loss": 2.3466, "step": 4051 }, { "epoch": 0.1, "learning_rate": 1.9745151262733646e-05, "loss": 2.5073, "step": 4052 }, { "epoch": 0.1, "learning_rate": 1.9744972148606064e-05, "loss": 2.275, "step": 4053 }, { "epoch": 0.1, "learning_rate": 1.9744792972370604e-05, "loss": 2.3194, "step": 4054 }, { "epoch": 0.1, "learning_rate": 1.97446137340284e-05, "loss": 2.3186, "step": 4055 }, { "epoch": 0.1, "learning_rate": 1.9744434433580597e-05, "loss": 2.338, "step": 4056 }, { "epoch": 0.1, "learning_rate": 1.9744255071028338e-05, "loss": 2.2871, "step": 4057 }, { "epoch": 0.1, "eval_loss": 1.9626988172531128, "eval_runtime": 95.4873, "eval_samples_per_second": 616.281, "eval_steps_per_second": 4.817, "step": 4057 }, { "epoch": 0.1, "learning_rate": 1.9744075646372764e-05, "loss": 2.2196, "step": 4058 }, { "epoch": 0.1, "learning_rate": 1.9743896159615022e-05, "loss": 2.4482, "step": 4059 }, { "epoch": 0.1, "learning_rate": 1.9743716610756257e-05, "loss": 2.4918, "step": 4060 }, { "epoch": 0.1, "learning_rate": 1.974353699979761e-05, "loss": 2.323, "step": 4061 }, { "epoch": 0.1, "learning_rate": 1.974335732674022e-05, "loss": 2.3002, "step": 4062 }, { "epoch": 0.1, "learning_rate": 1.9743177591585244e-05, "loss": 2.4473, "step": 4063 }, { "epoch": 0.1, "learning_rate": 1.9742997794333822e-05, "loss": 2.4235, "step": 4064 }, { "epoch": 0.1, "learning_rate": 1.9742817934987097e-05, "loss": 2.2918, "step": 4065 }, { "epoch": 0.1, "learning_rate": 1.974263801354622e-05, "loss": 2.4672, "step": 4066 }, { "epoch": 0.1, "learning_rate": 1.9742458030012337e-05, "loss": 2.3218, "step": 4067 }, { "epoch": 0.1, "learning_rate": 1.9742277984386593e-05, "loss": 2.3377, "step": 4068 }, { "epoch": 0.1, "learning_rate": 1.9742097876670136e-05, "loss": 2.3693, "step": 4069 }, { "epoch": 0.1, "learning_rate": 1.9741917706864117e-05, "loss": 2.3184, "step": 4070 }, { "epoch": 0.1, "learning_rate": 1.974173747496968e-05, "loss": 2.3304, "step": 4071 }, { "epoch": 0.1, "learning_rate": 1.9741557180987973e-05, "loss": 2.4275, "step": 4072 }, { "epoch": 0.1, "learning_rate": 1.9741376824920147e-05, "loss": 2.296, "step": 4073 }, { "epoch": 0.1, "learning_rate": 1.9741196406767355e-05, "loss": 2.4109, "step": 4074 }, { "epoch": 0.1, "learning_rate": 1.9741015926530744e-05, "loss": 2.2235, "step": 4075 }, { "epoch": 0.1, "learning_rate": 1.9740835384211463e-05, "loss": 2.166, "step": 4076 }, { "epoch": 0.1, "learning_rate": 1.9740654779810664e-05, "loss": 2.3533, "step": 4077 }, { "epoch": 0.1, "learning_rate": 1.97404741133295e-05, "loss": 2.2753, "step": 4078 }, { "epoch": 0.1, "learning_rate": 1.9740293384769113e-05, "loss": 2.5222, "step": 4079 }, { "epoch": 0.1, "learning_rate": 1.9740112594130668e-05, "loss": 2.2449, "step": 4080 }, { "epoch": 0.1, "learning_rate": 1.9739931741415307e-05, "loss": 2.2534, "step": 4081 }, { "epoch": 0.1, "learning_rate": 1.973975082662419e-05, "loss": 2.3666, "step": 4082 }, { "epoch": 0.1, "learning_rate": 1.9739569849758467e-05, "loss": 2.2414, "step": 4083 }, { "epoch": 0.1, "learning_rate": 1.973938881081929e-05, "loss": 2.2832, "step": 4084 }, { "epoch": 0.1, "learning_rate": 1.9739207709807814e-05, "loss": 2.2243, "step": 4085 }, { "epoch": 0.1, "learning_rate": 1.9739026546725195e-05, "loss": 2.5561, "step": 4086 }, { "epoch": 0.1, "learning_rate": 1.9738845321572583e-05, "loss": 2.3561, "step": 4087 }, { "epoch": 0.1, "learning_rate": 1.973866403435114e-05, "loss": 2.5825, "step": 4088 }, { "epoch": 0.1, "learning_rate": 1.9738482685062018e-05, "loss": 2.2882, "step": 4089 }, { "epoch": 0.1, "learning_rate": 1.9738301273706367e-05, "loss": 2.157, "step": 4090 }, { "epoch": 0.1, "learning_rate": 1.973811980028535e-05, "loss": 2.1976, "step": 4091 }, { "epoch": 0.1, "learning_rate": 1.973793826480012e-05, "loss": 2.0459, "step": 4092 }, { "epoch": 0.1, "learning_rate": 1.9737756667251843e-05, "loss": 2.3793, "step": 4093 }, { "epoch": 0.1, "learning_rate": 1.9737575007641664e-05, "loss": 2.2104, "step": 4094 }, { "epoch": 0.1, "learning_rate": 1.9737393285970747e-05, "loss": 2.3839, "step": 4095 }, { "epoch": 0.1, "learning_rate": 1.9737211502240253e-05, "loss": 2.302, "step": 4096 }, { "epoch": 0.1, "learning_rate": 1.973702965645133e-05, "loss": 2.5385, "step": 4097 }, { "epoch": 0.1, "learning_rate": 1.973684774860515e-05, "loss": 2.4263, "step": 4098 }, { "epoch": 0.1, "learning_rate": 1.973666577870286e-05, "loss": 2.4592, "step": 4099 }, { "epoch": 0.1, "learning_rate": 1.973648374674563e-05, "loss": 2.2211, "step": 4100 }, { "epoch": 0.1, "learning_rate": 1.9736301652734616e-05, "loss": 2.3469, "step": 4101 }, { "epoch": 0.1, "learning_rate": 1.9736119496670977e-05, "loss": 2.417, "step": 4102 }, { "epoch": 0.1, "learning_rate": 1.9735937278555878e-05, "loss": 2.4543, "step": 4103 }, { "epoch": 0.1, "learning_rate": 1.9735754998390477e-05, "loss": 2.3485, "step": 4104 }, { "epoch": 0.1, "learning_rate": 1.9735572656175934e-05, "loss": 2.1116, "step": 4105 }, { "epoch": 0.1, "learning_rate": 1.9735390251913416e-05, "loss": 2.152, "step": 4106 }, { "epoch": 0.1, "learning_rate": 1.973520778560408e-05, "loss": 2.3431, "step": 4107 }, { "epoch": 0.1, "learning_rate": 1.97350252572491e-05, "loss": 2.3108, "step": 4108 }, { "epoch": 0.1, "learning_rate": 1.9734842666849623e-05, "loss": 2.3611, "step": 4109 }, { "epoch": 0.1, "learning_rate": 1.9734660014406824e-05, "loss": 2.2884, "step": 4110 }, { "epoch": 0.1, "learning_rate": 1.9734477299921867e-05, "loss": 2.3937, "step": 4111 }, { "epoch": 0.1, "learning_rate": 1.973429452339591e-05, "loss": 2.2743, "step": 4112 }, { "epoch": 0.1, "learning_rate": 1.9734111684830127e-05, "loss": 2.3016, "step": 4113 }, { "epoch": 0.1, "learning_rate": 1.9733928784225672e-05, "loss": 2.3127, "step": 4114 }, { "epoch": 0.1, "learning_rate": 1.9733745821583717e-05, "loss": 2.4676, "step": 4115 }, { "epoch": 0.1, "learning_rate": 1.9733562796905432e-05, "loss": 2.2489, "step": 4116 }, { "epoch": 0.1, "learning_rate": 1.9733379710191975e-05, "loss": 2.24, "step": 4117 }, { "epoch": 0.1, "learning_rate": 1.9733196561444522e-05, "loss": 2.3854, "step": 4118 }, { "epoch": 0.1, "learning_rate": 1.973301335066423e-05, "loss": 2.269, "step": 4119 }, { "epoch": 0.1, "learning_rate": 1.9732830077852276e-05, "loss": 2.1354, "step": 4120 }, { "epoch": 0.1, "learning_rate": 1.973264674300982e-05, "loss": 2.5266, "step": 4121 }, { "epoch": 0.1, "learning_rate": 1.973246334613804e-05, "loss": 2.277, "step": 4122 }, { "epoch": 0.1, "learning_rate": 1.9732279887238098e-05, "loss": 2.183, "step": 4123 }, { "epoch": 0.1, "learning_rate": 1.973209636631116e-05, "loss": 2.4046, "step": 4124 }, { "epoch": 0.1, "learning_rate": 1.9731912783358404e-05, "loss": 2.3785, "step": 4125 }, { "epoch": 0.1, "learning_rate": 1.9731729138380995e-05, "loss": 2.4183, "step": 4126 }, { "epoch": 0.1, "learning_rate": 1.973154543138011e-05, "loss": 2.1881, "step": 4127 }, { "epoch": 0.1, "learning_rate": 1.9731361662356904e-05, "loss": 2.4219, "step": 4128 }, { "epoch": 0.1, "learning_rate": 1.9731177831312567e-05, "loss": 2.3602, "step": 4129 }, { "epoch": 0.1, "learning_rate": 1.973099393824826e-05, "loss": 2.3673, "step": 4130 }, { "epoch": 0.1, "learning_rate": 1.9730809983165157e-05, "loss": 2.5534, "step": 4131 }, { "epoch": 0.1, "learning_rate": 1.9730625966064433e-05, "loss": 2.3147, "step": 4132 }, { "epoch": 0.1, "learning_rate": 1.9730441886947256e-05, "loss": 2.4365, "step": 4133 }, { "epoch": 0.1, "learning_rate": 1.9730257745814803e-05, "loss": 2.3539, "step": 4134 }, { "epoch": 0.1, "learning_rate": 1.9730073542668245e-05, "loss": 2.2955, "step": 4135 }, { "epoch": 0.1, "learning_rate": 1.9729889277508758e-05, "loss": 2.2503, "step": 4136 }, { "epoch": 0.1, "learning_rate": 1.9729704950337515e-05, "loss": 2.2167, "step": 4137 }, { "epoch": 0.1, "learning_rate": 1.9729520561155695e-05, "loss": 2.4192, "step": 4138 }, { "epoch": 0.1, "learning_rate": 1.9729336109964466e-05, "loss": 2.4115, "step": 4139 }, { "epoch": 0.1, "learning_rate": 1.972915159676501e-05, "loss": 2.2879, "step": 4140 }, { "epoch": 0.1, "learning_rate": 1.97289670215585e-05, "loss": 2.2869, "step": 4141 }, { "epoch": 0.1, "learning_rate": 1.972878238434611e-05, "loss": 2.4274, "step": 4142 }, { "epoch": 0.1, "learning_rate": 1.972859768512902e-05, "loss": 2.5728, "step": 4143 }, { "epoch": 0.1, "learning_rate": 1.972841292390841e-05, "loss": 2.4264, "step": 4144 }, { "epoch": 0.1, "learning_rate": 1.972822810068545e-05, "loss": 2.2551, "step": 4145 }, { "epoch": 0.1, "learning_rate": 1.972804321546132e-05, "loss": 2.5189, "step": 4146 }, { "epoch": 0.1, "learning_rate": 1.9727858268237208e-05, "loss": 2.4271, "step": 4147 }, { "epoch": 0.1, "learning_rate": 1.972767325901428e-05, "loss": 2.3267, "step": 4148 }, { "epoch": 0.1, "learning_rate": 1.972748818779372e-05, "loss": 2.3563, "step": 4149 }, { "epoch": 0.1, "learning_rate": 1.9727303054576706e-05, "loss": 2.259, "step": 4150 }, { "epoch": 0.1, "learning_rate": 1.9727117859364425e-05, "loss": 2.3716, "step": 4151 }, { "epoch": 0.1, "learning_rate": 1.972693260215805e-05, "loss": 2.2168, "step": 4152 }, { "epoch": 0.1, "learning_rate": 1.9726747282958763e-05, "loss": 2.2489, "step": 4153 }, { "epoch": 0.1, "learning_rate": 1.9726561901767744e-05, "loss": 2.3229, "step": 4154 }, { "epoch": 0.1, "learning_rate": 1.9726376458586178e-05, "loss": 2.2536, "step": 4155 }, { "epoch": 0.1, "learning_rate": 1.9726190953415244e-05, "loss": 2.3355, "step": 4156 }, { "epoch": 0.1, "learning_rate": 1.9726005386256125e-05, "loss": 2.3138, "step": 4157 }, { "epoch": 0.1, "learning_rate": 1.9725819757110004e-05, "loss": 2.3247, "step": 4158 }, { "epoch": 0.1, "learning_rate": 1.972563406597807e-05, "loss": 2.2865, "step": 4159 }, { "epoch": 0.1, "learning_rate": 1.9725448312861493e-05, "loss": 2.3857, "step": 4160 }, { "epoch": 0.1, "learning_rate": 1.9725262497761468e-05, "loss": 2.3199, "step": 4161 }, { "epoch": 0.1, "learning_rate": 1.9725076620679174e-05, "loss": 2.4186, "step": 4162 }, { "epoch": 0.1, "learning_rate": 1.97248906816158e-05, "loss": 2.4719, "step": 4163 }, { "epoch": 0.1, "learning_rate": 1.9724704680572527e-05, "loss": 2.2921, "step": 4164 }, { "epoch": 0.1, "learning_rate": 1.9724518617550542e-05, "loss": 2.3783, "step": 4165 }, { "epoch": 0.1, "learning_rate": 1.9724332492551032e-05, "loss": 2.3508, "step": 4166 }, { "epoch": 0.1, "learning_rate": 1.9724146305575175e-05, "loss": 2.2966, "step": 4167 }, { "epoch": 0.1, "learning_rate": 1.9723960056624173e-05, "loss": 2.3676, "step": 4168 }, { "epoch": 0.1, "learning_rate": 1.97237737456992e-05, "loss": 2.3263, "step": 4169 }, { "epoch": 0.1, "learning_rate": 1.972358737280145e-05, "loss": 2.2411, "step": 4170 }, { "epoch": 0.1, "learning_rate": 1.972340093793211e-05, "loss": 2.3517, "step": 4171 }, { "epoch": 0.1, "learning_rate": 1.9723214441092363e-05, "loss": 2.3222, "step": 4172 }, { "epoch": 0.1, "learning_rate": 1.9723027882283404e-05, "loss": 2.4455, "step": 4173 }, { "epoch": 0.1, "learning_rate": 1.9722841261506422e-05, "loss": 2.344, "step": 4174 }, { "epoch": 0.1, "learning_rate": 1.97226545787626e-05, "loss": 2.2949, "step": 4175 }, { "epoch": 0.1, "learning_rate": 1.9722467834053132e-05, "loss": 2.348, "step": 4176 }, { "epoch": 0.1, "learning_rate": 1.9722281027379212e-05, "loss": 2.2239, "step": 4177 }, { "epoch": 0.1, "learning_rate": 1.9722094158742027e-05, "loss": 2.3493, "step": 4178 }, { "epoch": 0.1, "learning_rate": 1.9721907228142765e-05, "loss": 2.468, "step": 4179 }, { "epoch": 0.1, "learning_rate": 1.9721720235582617e-05, "loss": 2.2974, "step": 4180 }, { "epoch": 0.1, "learning_rate": 1.972153318106278e-05, "loss": 2.2856, "step": 4181 }, { "epoch": 0.1, "learning_rate": 1.9721346064584446e-05, "loss": 2.3765, "step": 4182 }, { "epoch": 0.1, "learning_rate": 1.9721158886148805e-05, "loss": 2.4754, "step": 4183 }, { "epoch": 0.1, "learning_rate": 1.972097164575705e-05, "loss": 2.4903, "step": 4184 }, { "epoch": 0.1, "learning_rate": 1.9720784343410375e-05, "loss": 2.2562, "step": 4185 }, { "epoch": 0.1, "learning_rate": 1.972059697910997e-05, "loss": 2.3185, "step": 4186 }, { "epoch": 0.1, "learning_rate": 1.9720409552857036e-05, "loss": 2.4747, "step": 4187 }, { "epoch": 0.1, "learning_rate": 1.9720222064652763e-05, "loss": 2.416, "step": 4188 }, { "epoch": 0.1, "learning_rate": 1.9720034514498346e-05, "loss": 2.2472, "step": 4189 }, { "epoch": 0.1, "learning_rate": 1.9719846902394986e-05, "loss": 2.3529, "step": 4190 }, { "epoch": 0.1, "learning_rate": 1.971965922834387e-05, "loss": 2.4241, "step": 4191 }, { "epoch": 0.1, "learning_rate": 1.97194714923462e-05, "loss": 2.2449, "step": 4192 }, { "epoch": 0.1, "learning_rate": 1.971928369440317e-05, "loss": 2.3489, "step": 4193 }, { "epoch": 0.1, "learning_rate": 1.9719095834515974e-05, "loss": 2.3875, "step": 4194 }, { "epoch": 0.1, "learning_rate": 1.9718907912685815e-05, "loss": 2.358, "step": 4195 }, { "epoch": 0.1, "learning_rate": 1.971871992891389e-05, "loss": 2.2771, "step": 4196 }, { "epoch": 0.1, "learning_rate": 1.9718531883201396e-05, "loss": 2.23, "step": 4197 }, { "epoch": 0.1, "learning_rate": 1.971834377554953e-05, "loss": 2.3513, "step": 4198 }, { "epoch": 0.1, "learning_rate": 1.9718155605959494e-05, "loss": 2.1413, "step": 4199 }, { "epoch": 0.1, "learning_rate": 1.9717967374432483e-05, "loss": 2.3427, "step": 4200 }, { "epoch": 0.1, "learning_rate": 1.97177790809697e-05, "loss": 2.4826, "step": 4201 }, { "epoch": 0.1, "learning_rate": 1.9717590725572348e-05, "loss": 2.4963, "step": 4202 }, { "epoch": 0.1, "learning_rate": 1.971740230824162e-05, "loss": 2.3317, "step": 4203 }, { "epoch": 0.1, "learning_rate": 1.9717213828978715e-05, "loss": 2.3513, "step": 4204 }, { "epoch": 0.1, "learning_rate": 1.9717025287784844e-05, "loss": 2.3113, "step": 4205 }, { "epoch": 0.1, "learning_rate": 1.9716836684661206e-05, "loss": 2.2599, "step": 4206 }, { "epoch": 0.1, "learning_rate": 1.9716648019608997e-05, "loss": 2.5042, "step": 4207 }, { "epoch": 0.1, "learning_rate": 1.9716459292629427e-05, "loss": 2.4805, "step": 4208 }, { "epoch": 0.1, "learning_rate": 1.9716270503723695e-05, "loss": 2.2645, "step": 4209 }, { "epoch": 0.1, "learning_rate": 1.9716081652893003e-05, "loss": 2.3839, "step": 4210 }, { "epoch": 0.1, "learning_rate": 1.9715892740138555e-05, "loss": 2.3507, "step": 4211 }, { "epoch": 0.1, "learning_rate": 1.971570376546156e-05, "loss": 2.3849, "step": 4212 }, { "epoch": 0.1, "learning_rate": 1.9715514728863218e-05, "loss": 2.3724, "step": 4213 }, { "epoch": 0.1, "learning_rate": 1.9715325630344734e-05, "loss": 2.2151, "step": 4214 }, { "epoch": 0.1, "learning_rate": 1.971513646990731e-05, "loss": 2.3138, "step": 4215 }, { "epoch": 0.1, "learning_rate": 1.971494724755216e-05, "loss": 2.3406, "step": 4216 }, { "epoch": 0.1, "learning_rate": 1.9714757963280483e-05, "loss": 2.24, "step": 4217 }, { "epoch": 0.1, "learning_rate": 1.971456861709349e-05, "loss": 2.4953, "step": 4218 }, { "epoch": 0.1, "learning_rate": 1.971437920899238e-05, "loss": 2.2195, "step": 4219 }, { "epoch": 0.1, "learning_rate": 1.9714189738978372e-05, "loss": 2.3014, "step": 4220 }, { "epoch": 0.1, "learning_rate": 1.9714000207052665e-05, "loss": 2.2118, "step": 4221 }, { "epoch": 0.1, "learning_rate": 1.9713810613216466e-05, "loss": 2.2368, "step": 4222 }, { "epoch": 0.1, "learning_rate": 1.971362095747099e-05, "loss": 2.3367, "step": 4223 }, { "epoch": 0.1, "learning_rate": 1.9713431239817438e-05, "loss": 2.3419, "step": 4224 }, { "epoch": 0.1, "learning_rate": 1.9713241460257028e-05, "loss": 2.3186, "step": 4225 }, { "epoch": 0.1, "learning_rate": 1.971305161879096e-05, "loss": 2.3935, "step": 4226 }, { "epoch": 0.1, "learning_rate": 1.9712861715420452e-05, "loss": 2.3676, "step": 4227 }, { "epoch": 0.1, "learning_rate": 1.971267175014671e-05, "loss": 2.4199, "step": 4228 }, { "epoch": 0.1, "learning_rate": 1.9712481722970948e-05, "loss": 2.3144, "step": 4229 }, { "epoch": 0.1, "learning_rate": 1.9712291633894368e-05, "loss": 2.3925, "step": 4230 }, { "epoch": 0.1, "learning_rate": 1.9712101482918193e-05, "loss": 2.1889, "step": 4231 }, { "epoch": 0.1, "learning_rate": 1.9711911270043628e-05, "loss": 2.2024, "step": 4232 }, { "epoch": 0.1, "learning_rate": 1.971172099527189e-05, "loss": 2.3662, "step": 4233 }, { "epoch": 0.1, "learning_rate": 1.971153065860419e-05, "loss": 2.3652, "step": 4234 }, { "epoch": 0.1, "learning_rate": 1.971134026004174e-05, "loss": 2.2799, "step": 4235 }, { "epoch": 0.1, "learning_rate": 1.971114979958575e-05, "loss": 2.3633, "step": 4236 }, { "epoch": 0.1, "learning_rate": 1.971095927723744e-05, "loss": 2.3597, "step": 4237 }, { "epoch": 0.1, "learning_rate": 1.971076869299802e-05, "loss": 2.2723, "step": 4238 }, { "epoch": 0.1, "learning_rate": 1.9710578046868712e-05, "loss": 2.3836, "step": 4239 }, { "epoch": 0.1, "learning_rate": 1.971038733885072e-05, "loss": 2.2047, "step": 4240 }, { "epoch": 0.1, "learning_rate": 1.9710196568945267e-05, "loss": 2.2172, "step": 4241 }, { "epoch": 0.1, "learning_rate": 1.9710005737153563e-05, "loss": 2.2635, "step": 4242 }, { "epoch": 0.1, "learning_rate": 1.9709814843476832e-05, "loss": 2.345, "step": 4243 }, { "epoch": 0.1, "learning_rate": 1.9709623887916288e-05, "loss": 2.3471, "step": 4244 }, { "epoch": 0.1, "learning_rate": 1.9709432870473143e-05, "loss": 2.4079, "step": 4245 }, { "epoch": 0.1, "learning_rate": 1.970924179114862e-05, "loss": 2.3185, "step": 4246 }, { "epoch": 0.1, "learning_rate": 1.9709050649943936e-05, "loss": 2.4118, "step": 4247 }, { "epoch": 0.1, "learning_rate": 1.9708859446860307e-05, "loss": 2.3012, "step": 4248 }, { "epoch": 0.1, "learning_rate": 1.9708668181898952e-05, "loss": 2.3715, "step": 4249 }, { "epoch": 0.1, "learning_rate": 1.9708476855061092e-05, "loss": 2.3765, "step": 4250 }, { "epoch": 0.1, "learning_rate": 1.9708285466347946e-05, "loss": 2.2284, "step": 4251 }, { "epoch": 0.1, "learning_rate": 1.970809401576073e-05, "loss": 2.5399, "step": 4252 }, { "epoch": 0.1, "learning_rate": 1.9707902503300665e-05, "loss": 2.2695, "step": 4253 }, { "epoch": 0.1, "learning_rate": 1.970771092896898e-05, "loss": 2.192, "step": 4254 }, { "epoch": 0.1, "learning_rate": 1.9707519292766885e-05, "loss": 2.3122, "step": 4255 }, { "epoch": 0.1, "learning_rate": 1.9707327594695606e-05, "loss": 2.4073, "step": 4256 }, { "epoch": 0.1, "learning_rate": 1.970713583475637e-05, "loss": 2.4154, "step": 4257 }, { "epoch": 0.1, "learning_rate": 1.9706944012950388e-05, "loss": 2.385, "step": 4258 }, { "epoch": 0.1, "learning_rate": 1.9706752129278893e-05, "loss": 2.3273, "step": 4259 }, { "epoch": 0.11, "learning_rate": 1.97065601837431e-05, "loss": 2.4827, "step": 4260 }, { "epoch": 0.11, "learning_rate": 1.9706368176344234e-05, "loss": 2.2108, "step": 4261 }, { "epoch": 0.11, "learning_rate": 1.9706176107083524e-05, "loss": 2.4293, "step": 4262 }, { "epoch": 0.11, "learning_rate": 1.970598397596219e-05, "loss": 2.2994, "step": 4263 }, { "epoch": 0.11, "learning_rate": 1.9705791782981457e-05, "loss": 2.5205, "step": 4264 }, { "epoch": 0.11, "learning_rate": 1.970559952814255e-05, "loss": 2.2037, "step": 4265 }, { "epoch": 0.11, "learning_rate": 1.9705407211446693e-05, "loss": 2.3275, "step": 4266 }, { "epoch": 0.11, "learning_rate": 1.970521483289511e-05, "loss": 2.3443, "step": 4267 }, { "epoch": 0.11, "learning_rate": 1.9705022392489035e-05, "loss": 2.3157, "step": 4268 }, { "epoch": 0.11, "learning_rate": 1.9704829890229687e-05, "loss": 2.318, "step": 4269 }, { "epoch": 0.11, "learning_rate": 1.97046373261183e-05, "loss": 2.1894, "step": 4270 }, { "epoch": 0.11, "learning_rate": 1.9704444700156093e-05, "loss": 2.1922, "step": 4271 }, { "epoch": 0.11, "learning_rate": 1.9704252012344295e-05, "loss": 2.174, "step": 4272 }, { "epoch": 0.11, "learning_rate": 1.9704059262684142e-05, "loss": 2.3005, "step": 4273 }, { "epoch": 0.11, "learning_rate": 1.9703866451176853e-05, "loss": 2.4878, "step": 4274 }, { "epoch": 0.11, "learning_rate": 1.9703673577823663e-05, "loss": 2.1242, "step": 4275 }, { "epoch": 0.11, "learning_rate": 1.9703480642625795e-05, "loss": 2.2621, "step": 4276 }, { "epoch": 0.11, "learning_rate": 1.9703287645584486e-05, "loss": 2.2372, "step": 4277 }, { "epoch": 0.11, "learning_rate": 1.970309458670096e-05, "loss": 2.2271, "step": 4278 }, { "epoch": 0.11, "learning_rate": 1.9702901465976453e-05, "loss": 2.3402, "step": 4279 }, { "epoch": 0.11, "learning_rate": 1.970270828341219e-05, "loss": 2.4307, "step": 4280 }, { "epoch": 0.11, "learning_rate": 1.970251503900941e-05, "loss": 2.3825, "step": 4281 }, { "epoch": 0.11, "learning_rate": 1.9702321732769334e-05, "loss": 2.3147, "step": 4282 }, { "epoch": 0.11, "learning_rate": 1.9702128364693202e-05, "loss": 2.3036, "step": 4283 }, { "epoch": 0.11, "learning_rate": 1.9701934934782242e-05, "loss": 2.4024, "step": 4284 }, { "epoch": 0.11, "learning_rate": 1.9701741443037693e-05, "loss": 2.2557, "step": 4285 }, { "epoch": 0.11, "learning_rate": 1.9701547889460783e-05, "loss": 2.4463, "step": 4286 }, { "epoch": 0.11, "learning_rate": 1.9701354274052745e-05, "loss": 2.3782, "step": 4287 }, { "epoch": 0.11, "learning_rate": 1.9701160596814818e-05, "loss": 2.3487, "step": 4288 }, { "epoch": 0.11, "learning_rate": 1.9700966857748226e-05, "loss": 2.1644, "step": 4289 }, { "epoch": 0.11, "learning_rate": 1.9700773056854216e-05, "loss": 2.1155, "step": 4290 }, { "epoch": 0.11, "learning_rate": 1.9700579194134017e-05, "loss": 2.325, "step": 4291 }, { "epoch": 0.11, "learning_rate": 1.9700385269588867e-05, "loss": 2.4189, "step": 4292 }, { "epoch": 0.11, "learning_rate": 1.9700191283219998e-05, "loss": 2.3194, "step": 4293 }, { "epoch": 0.11, "learning_rate": 1.969999723502865e-05, "loss": 2.5128, "step": 4294 }, { "epoch": 0.11, "learning_rate": 1.9699803125016058e-05, "loss": 2.3413, "step": 4295 }, { "epoch": 0.11, "learning_rate": 1.969960895318346e-05, "loss": 2.3962, "step": 4296 }, { "epoch": 0.11, "learning_rate": 1.9699414719532095e-05, "loss": 2.2162, "step": 4297 }, { "epoch": 0.11, "learning_rate": 1.9699220424063194e-05, "loss": 2.2174, "step": 4298 }, { "epoch": 0.11, "learning_rate": 1.9699026066778004e-05, "loss": 2.2424, "step": 4299 }, { "epoch": 0.11, "learning_rate": 1.969883164767776e-05, "loss": 2.473, "step": 4300 }, { "epoch": 0.11, "learning_rate": 1.96986371667637e-05, "loss": 2.2441, "step": 4301 }, { "epoch": 0.11, "learning_rate": 1.9698442624037066e-05, "loss": 2.4074, "step": 4302 }, { "epoch": 0.11, "learning_rate": 1.9698248019499097e-05, "loss": 2.4267, "step": 4303 }, { "epoch": 0.11, "learning_rate": 1.969805335315103e-05, "loss": 2.1776, "step": 4304 }, { "epoch": 0.11, "learning_rate": 1.969785862499411e-05, "loss": 2.4866, "step": 4305 }, { "epoch": 0.11, "learning_rate": 1.9697663835029577e-05, "loss": 2.2608, "step": 4306 }, { "epoch": 0.11, "learning_rate": 1.969746898325867e-05, "loss": 2.1739, "step": 4307 }, { "epoch": 0.11, "learning_rate": 1.9697274069682633e-05, "loss": 2.4381, "step": 4308 }, { "epoch": 0.11, "learning_rate": 1.9697079094302713e-05, "loss": 2.3209, "step": 4309 }, { "epoch": 0.11, "learning_rate": 1.9696884057120145e-05, "loss": 2.2833, "step": 4310 }, { "epoch": 0.11, "learning_rate": 1.9696688958136173e-05, "loss": 2.2793, "step": 4311 }, { "epoch": 0.11, "learning_rate": 1.9696493797352043e-05, "loss": 2.0415, "step": 4312 }, { "epoch": 0.11, "learning_rate": 1.9696298574768997e-05, "loss": 2.3136, "step": 4313 }, { "epoch": 0.11, "learning_rate": 1.9696103290388283e-05, "loss": 2.2324, "step": 4314 }, { "epoch": 0.11, "learning_rate": 1.969590794421114e-05, "loss": 2.3188, "step": 4315 }, { "epoch": 0.11, "learning_rate": 1.9695712536238817e-05, "loss": 2.1131, "step": 4316 }, { "epoch": 0.11, "learning_rate": 1.9695517066472562e-05, "loss": 2.3567, "step": 4317 }, { "epoch": 0.11, "learning_rate": 1.969532153491361e-05, "loss": 2.3861, "step": 4318 }, { "epoch": 0.11, "learning_rate": 1.969512594156322e-05, "loss": 2.3261, "step": 4319 }, { "epoch": 0.11, "learning_rate": 1.9694930286422628e-05, "loss": 2.2221, "step": 4320 }, { "epoch": 0.11, "learning_rate": 1.969473456949309e-05, "loss": 2.3312, "step": 4321 }, { "epoch": 0.11, "learning_rate": 1.9694538790775848e-05, "loss": 2.5012, "step": 4322 }, { "epoch": 0.11, "learning_rate": 1.969434295027215e-05, "loss": 2.3481, "step": 4323 }, { "epoch": 0.11, "learning_rate": 1.9694147047983244e-05, "loss": 2.3396, "step": 4324 }, { "epoch": 0.11, "learning_rate": 1.9693951083910385e-05, "loss": 2.2556, "step": 4325 }, { "epoch": 0.11, "learning_rate": 1.969375505805481e-05, "loss": 2.3756, "step": 4326 }, { "epoch": 0.11, "learning_rate": 1.969355897041778e-05, "loss": 2.2812, "step": 4327 }, { "epoch": 0.11, "learning_rate": 1.9693362821000534e-05, "loss": 2.3675, "step": 4328 }, { "epoch": 0.11, "learning_rate": 1.969316660980433e-05, "loss": 2.3761, "step": 4329 }, { "epoch": 0.11, "learning_rate": 1.9692970336830418e-05, "loss": 2.2919, "step": 4330 }, { "epoch": 0.11, "learning_rate": 1.969277400208005e-05, "loss": 2.2097, "step": 4331 }, { "epoch": 0.11, "learning_rate": 1.969257760555447e-05, "loss": 2.5636, "step": 4332 }, { "epoch": 0.11, "learning_rate": 1.9692381147254936e-05, "loss": 2.4043, "step": 4333 }, { "epoch": 0.11, "learning_rate": 1.9692184627182697e-05, "loss": 2.4607, "step": 4334 }, { "epoch": 0.11, "learning_rate": 1.9691988045339008e-05, "loss": 2.3105, "step": 4335 }, { "epoch": 0.11, "learning_rate": 1.969179140172512e-05, "loss": 2.3878, "step": 4336 }, { "epoch": 0.11, "learning_rate": 1.9691594696342286e-05, "loss": 2.246, "step": 4337 }, { "epoch": 0.11, "learning_rate": 1.9691397929191763e-05, "loss": 2.3843, "step": 4338 }, { "epoch": 0.11, "learning_rate": 1.96912011002748e-05, "loss": 2.2665, "step": 4339 }, { "epoch": 0.11, "learning_rate": 1.9691004209592654e-05, "loss": 2.2942, "step": 4340 }, { "epoch": 0.11, "learning_rate": 1.9690807257146582e-05, "loss": 2.3968, "step": 4341 }, { "epoch": 0.11, "learning_rate": 1.9690610242937836e-05, "loss": 2.2639, "step": 4342 }, { "epoch": 0.11, "learning_rate": 1.9690413166967676e-05, "loss": 2.3537, "step": 4343 }, { "epoch": 0.11, "learning_rate": 1.9690216029237352e-05, "loss": 2.4935, "step": 4344 }, { "epoch": 0.11, "learning_rate": 1.969001882974812e-05, "loss": 2.2849, "step": 4345 }, { "epoch": 0.11, "learning_rate": 1.9689821568501246e-05, "loss": 2.3911, "step": 4346 }, { "epoch": 0.11, "learning_rate": 1.968962424549798e-05, "loss": 2.3068, "step": 4347 }, { "epoch": 0.11, "learning_rate": 1.968942686073958e-05, "loss": 2.4017, "step": 4348 }, { "epoch": 0.11, "learning_rate": 1.96892294142273e-05, "loss": 2.4791, "step": 4349 }, { "epoch": 0.11, "learning_rate": 1.9689031905962414e-05, "loss": 2.2808, "step": 4350 }, { "epoch": 0.11, "learning_rate": 1.9688834335946162e-05, "loss": 2.2783, "step": 4351 }, { "epoch": 0.11, "learning_rate": 1.9688636704179812e-05, "loss": 2.1319, "step": 4352 }, { "epoch": 0.11, "learning_rate": 1.9688439010664625e-05, "loss": 2.3832, "step": 4353 }, { "epoch": 0.11, "learning_rate": 1.9688241255401858e-05, "loss": 2.2485, "step": 4354 }, { "epoch": 0.11, "learning_rate": 1.968804343839277e-05, "loss": 2.2794, "step": 4355 }, { "epoch": 0.11, "learning_rate": 1.968784555963863e-05, "loss": 2.2271, "step": 4356 }, { "epoch": 0.11, "learning_rate": 1.968764761914069e-05, "loss": 2.2095, "step": 4357 }, { "epoch": 0.11, "learning_rate": 1.9687449616900212e-05, "loss": 2.3012, "step": 4358 }, { "epoch": 0.11, "learning_rate": 1.9687251552918462e-05, "loss": 2.3266, "step": 4359 }, { "epoch": 0.11, "learning_rate": 1.9687053427196704e-05, "loss": 2.3339, "step": 4360 }, { "epoch": 0.11, "learning_rate": 1.9686855239736192e-05, "loss": 2.3868, "step": 4361 }, { "epoch": 0.11, "learning_rate": 1.96866569905382e-05, "loss": 2.4014, "step": 4362 }, { "epoch": 0.11, "learning_rate": 1.9686458679603984e-05, "loss": 2.243, "step": 4363 }, { "epoch": 0.11, "learning_rate": 1.9686260306934808e-05, "loss": 2.1047, "step": 4364 }, { "epoch": 0.11, "learning_rate": 1.9686061872531944e-05, "loss": 2.3073, "step": 4365 }, { "epoch": 0.11, "learning_rate": 1.9685863376396647e-05, "loss": 2.4189, "step": 4366 }, { "epoch": 0.11, "learning_rate": 1.9685664818530185e-05, "loss": 2.2503, "step": 4367 }, { "epoch": 0.11, "learning_rate": 1.9685466198933826e-05, "loss": 2.3301, "step": 4368 }, { "epoch": 0.11, "learning_rate": 1.9685267517608836e-05, "loss": 2.2567, "step": 4369 }, { "epoch": 0.11, "learning_rate": 1.9685068774556476e-05, "loss": 2.2456, "step": 4370 }, { "epoch": 0.11, "learning_rate": 1.968486996977802e-05, "loss": 2.2781, "step": 4371 }, { "epoch": 0.11, "learning_rate": 1.968467110327473e-05, "loss": 2.2987, "step": 4372 }, { "epoch": 0.11, "learning_rate": 1.9684472175047874e-05, "loss": 2.5273, "step": 4373 }, { "epoch": 0.11, "learning_rate": 1.9684273185098725e-05, "loss": 2.4508, "step": 4374 }, { "epoch": 0.11, "learning_rate": 1.9684074133428542e-05, "loss": 2.2382, "step": 4375 }, { "epoch": 0.11, "learning_rate": 1.9683875020038602e-05, "loss": 2.4011, "step": 4376 }, { "epoch": 0.11, "learning_rate": 1.968367584493017e-05, "loss": 2.3675, "step": 4377 }, { "epoch": 0.11, "learning_rate": 1.9683476608104516e-05, "loss": 2.209, "step": 4378 }, { "epoch": 0.11, "learning_rate": 1.9683277309562907e-05, "loss": 2.2457, "step": 4379 }, { "epoch": 0.11, "learning_rate": 1.968307794930662e-05, "loss": 2.3177, "step": 4380 }, { "epoch": 0.11, "learning_rate": 1.968287852733692e-05, "loss": 2.2692, "step": 4381 }, { "epoch": 0.11, "learning_rate": 1.9682679043655078e-05, "loss": 2.2945, "step": 4382 }, { "epoch": 0.11, "learning_rate": 1.9682479498262367e-05, "loss": 2.2399, "step": 4383 }, { "epoch": 0.11, "learning_rate": 1.9682279891160063e-05, "loss": 2.226, "step": 4384 }, { "epoch": 0.11, "learning_rate": 1.968208022234943e-05, "loss": 2.2838, "step": 4385 }, { "epoch": 0.11, "learning_rate": 1.9681880491831745e-05, "loss": 2.3229, "step": 4386 }, { "epoch": 0.11, "learning_rate": 1.9681680699608283e-05, "loss": 2.1665, "step": 4387 }, { "epoch": 0.11, "learning_rate": 1.9681480845680313e-05, "loss": 2.298, "step": 4388 }, { "epoch": 0.11, "learning_rate": 1.9681280930049108e-05, "loss": 2.285, "step": 4389 }, { "epoch": 0.11, "learning_rate": 1.9681080952715946e-05, "loss": 2.148, "step": 4390 }, { "epoch": 0.11, "learning_rate": 1.9680880913682103e-05, "loss": 2.414, "step": 4391 }, { "epoch": 0.11, "learning_rate": 1.9680680812948844e-05, "loss": 2.2919, "step": 4392 }, { "epoch": 0.11, "learning_rate": 1.968048065051746e-05, "loss": 2.3014, "step": 4393 }, { "epoch": 0.11, "learning_rate": 1.9680280426389214e-05, "loss": 2.6097, "step": 4394 }, { "epoch": 0.11, "learning_rate": 1.9680080140565385e-05, "loss": 2.1054, "step": 4395 }, { "epoch": 0.11, "learning_rate": 1.9679879793047252e-05, "loss": 2.2988, "step": 4396 }, { "epoch": 0.11, "learning_rate": 1.967967938383609e-05, "loss": 2.4448, "step": 4397 }, { "epoch": 0.11, "learning_rate": 1.9679478912933178e-05, "loss": 2.4346, "step": 4398 }, { "epoch": 0.11, "learning_rate": 1.967927838033979e-05, "loss": 2.4523, "step": 4399 }, { "epoch": 0.11, "learning_rate": 1.9679077786057207e-05, "loss": 2.1536, "step": 4400 }, { "epoch": 0.11, "learning_rate": 1.967887713008671e-05, "loss": 2.0928, "step": 4401 }, { "epoch": 0.11, "learning_rate": 1.967867641242957e-05, "loss": 2.5001, "step": 4402 }, { "epoch": 0.11, "learning_rate": 1.9678475633087078e-05, "loss": 2.3298, "step": 4403 }, { "epoch": 0.11, "learning_rate": 1.96782747920605e-05, "loss": 2.3801, "step": 4404 }, { "epoch": 0.11, "learning_rate": 1.9678073889351127e-05, "loss": 2.3149, "step": 4405 }, { "epoch": 0.11, "learning_rate": 1.967787292496023e-05, "loss": 2.0626, "step": 4406 }, { "epoch": 0.11, "learning_rate": 1.9677671898889097e-05, "loss": 2.2775, "step": 4407 }, { "epoch": 0.11, "learning_rate": 1.9677470811139012e-05, "loss": 2.6611, "step": 4408 }, { "epoch": 0.11, "learning_rate": 1.967726966171125e-05, "loss": 2.3497, "step": 4409 }, { "epoch": 0.11, "learning_rate": 1.967706845060709e-05, "loss": 2.2973, "step": 4410 }, { "epoch": 0.11, "learning_rate": 1.9676867177827823e-05, "loss": 2.2788, "step": 4411 }, { "epoch": 0.11, "learning_rate": 1.9676665843374726e-05, "loss": 2.4265, "step": 4412 }, { "epoch": 0.11, "learning_rate": 1.9676464447249085e-05, "loss": 2.4506, "step": 4413 }, { "epoch": 0.11, "learning_rate": 1.9676262989452182e-05, "loss": 2.4549, "step": 4414 }, { "epoch": 0.11, "learning_rate": 1.9676061469985304e-05, "loss": 2.3231, "step": 4415 }, { "epoch": 0.11, "learning_rate": 1.9675859888849727e-05, "loss": 2.4178, "step": 4416 }, { "epoch": 0.11, "learning_rate": 1.9675658246046748e-05, "loss": 2.2465, "step": 4417 }, { "epoch": 0.11, "learning_rate": 1.9675456541577645e-05, "loss": 2.1829, "step": 4418 }, { "epoch": 0.11, "learning_rate": 1.9675254775443703e-05, "loss": 2.384, "step": 4419 }, { "epoch": 0.11, "learning_rate": 1.9675052947646204e-05, "loss": 2.2367, "step": 4420 }, { "epoch": 0.11, "learning_rate": 1.9674851058186447e-05, "loss": 2.2804, "step": 4421 }, { "epoch": 0.11, "learning_rate": 1.967464910706571e-05, "loss": 2.2567, "step": 4422 }, { "epoch": 0.11, "learning_rate": 1.9674447094285278e-05, "loss": 2.3121, "step": 4423 }, { "epoch": 0.11, "learning_rate": 1.9674245019846445e-05, "loss": 2.2134, "step": 4424 }, { "epoch": 0.11, "learning_rate": 1.9674042883750494e-05, "loss": 2.2547, "step": 4425 }, { "epoch": 0.11, "learning_rate": 1.9673840685998717e-05, "loss": 2.4599, "step": 4426 }, { "epoch": 0.11, "learning_rate": 1.9673638426592397e-05, "loss": 2.4573, "step": 4427 }, { "epoch": 0.11, "learning_rate": 1.967343610553283e-05, "loss": 2.1332, "step": 4428 }, { "epoch": 0.11, "learning_rate": 1.9673233722821302e-05, "loss": 2.1529, "step": 4429 }, { "epoch": 0.11, "learning_rate": 1.9673031278459104e-05, "loss": 2.2378, "step": 4430 }, { "epoch": 0.11, "learning_rate": 1.967282877244752e-05, "loss": 2.2421, "step": 4431 }, { "epoch": 0.11, "learning_rate": 1.9672626204787853e-05, "loss": 2.3814, "step": 4432 }, { "epoch": 0.11, "learning_rate": 1.9672423575481382e-05, "loss": 2.4874, "step": 4433 }, { "epoch": 0.11, "learning_rate": 1.9672220884529404e-05, "loss": 2.3531, "step": 4434 }, { "epoch": 0.11, "learning_rate": 1.967201813193321e-05, "loss": 2.3403, "step": 4435 }, { "epoch": 0.11, "learning_rate": 1.9671815317694098e-05, "loss": 2.2774, "step": 4436 }, { "epoch": 0.11, "learning_rate": 1.9671612441813347e-05, "loss": 2.2194, "step": 4437 }, { "epoch": 0.11, "learning_rate": 1.9671409504292264e-05, "loss": 2.2468, "step": 4438 }, { "epoch": 0.11, "learning_rate": 1.9671206505132135e-05, "loss": 2.227, "step": 4439 }, { "epoch": 0.11, "learning_rate": 1.9671003444334253e-05, "loss": 2.2912, "step": 4440 }, { "epoch": 0.11, "learning_rate": 1.9670800321899914e-05, "loss": 2.1846, "step": 4441 }, { "epoch": 0.11, "learning_rate": 1.9670597137830417e-05, "loss": 2.3113, "step": 4442 }, { "epoch": 0.11, "learning_rate": 1.967039389212705e-05, "loss": 2.4899, "step": 4443 }, { "epoch": 0.11, "learning_rate": 1.9670190584791113e-05, "loss": 2.23, "step": 4444 }, { "epoch": 0.11, "learning_rate": 1.9669987215823898e-05, "loss": 2.4136, "step": 4445 }, { "epoch": 0.11, "learning_rate": 1.9669783785226704e-05, "loss": 2.1982, "step": 4446 }, { "epoch": 0.11, "learning_rate": 1.9669580293000823e-05, "loss": 2.4345, "step": 4447 }, { "epoch": 0.11, "learning_rate": 1.9669376739147563e-05, "loss": 2.2076, "step": 4448 }, { "epoch": 0.11, "learning_rate": 1.966917312366821e-05, "loss": 2.525, "step": 4449 }, { "epoch": 0.11, "learning_rate": 1.9668969446564063e-05, "loss": 2.2783, "step": 4450 }, { "epoch": 0.11, "learning_rate": 1.966876570783643e-05, "loss": 2.2618, "step": 4451 }, { "epoch": 0.11, "learning_rate": 1.9668561907486593e-05, "loss": 2.1176, "step": 4452 }, { "epoch": 0.11, "learning_rate": 1.9668358045515866e-05, "loss": 2.3623, "step": 4453 }, { "epoch": 0.11, "learning_rate": 1.966815412192554e-05, "loss": 2.2454, "step": 4454 }, { "epoch": 0.11, "learning_rate": 1.9667950136716917e-05, "loss": 2.048, "step": 4455 }, { "epoch": 0.11, "learning_rate": 1.96677460898913e-05, "loss": 2.4153, "step": 4456 }, { "epoch": 0.11, "learning_rate": 1.966754198144998e-05, "loss": 2.3529, "step": 4457 }, { "epoch": 0.11, "learning_rate": 1.966733781139427e-05, "loss": 2.2675, "step": 4458 }, { "epoch": 0.11, "learning_rate": 1.9667133579725467e-05, "loss": 2.3708, "step": 4459 }, { "epoch": 0.11, "learning_rate": 1.9666929286444868e-05, "loss": 2.2874, "step": 4460 }, { "epoch": 0.11, "learning_rate": 1.966672493155378e-05, "loss": 2.5997, "step": 4461 }, { "epoch": 0.11, "learning_rate": 1.96665205150535e-05, "loss": 2.3388, "step": 4462 }, { "epoch": 0.11, "learning_rate": 1.966631603694534e-05, "loss": 2.493, "step": 4463 }, { "epoch": 0.11, "learning_rate": 1.9666111497230595e-05, "loss": 2.2876, "step": 4464 }, { "epoch": 0.11, "learning_rate": 1.966590689591057e-05, "loss": 2.2515, "step": 4465 }, { "epoch": 0.11, "learning_rate": 1.9665702232986576e-05, "loss": 2.4415, "step": 4466 }, { "epoch": 0.11, "learning_rate": 1.9665497508459904e-05, "loss": 2.2436, "step": 4467 }, { "epoch": 0.11, "learning_rate": 1.966529272233187e-05, "loss": 2.2561, "step": 4468 }, { "epoch": 0.11, "learning_rate": 1.9665087874603775e-05, "loss": 2.2379, "step": 4469 }, { "epoch": 0.11, "learning_rate": 1.9664882965276927e-05, "loss": 2.2757, "step": 4470 }, { "epoch": 0.11, "learning_rate": 1.9664677994352626e-05, "loss": 2.3175, "step": 4471 }, { "epoch": 0.11, "learning_rate": 1.9664472961832188e-05, "loss": 2.2045, "step": 4472 }, { "epoch": 0.11, "learning_rate": 1.9664267867716913e-05, "loss": 2.546, "step": 4473 }, { "epoch": 0.11, "learning_rate": 1.9664062712008106e-05, "loss": 2.3486, "step": 4474 }, { "epoch": 0.11, "learning_rate": 1.966385749470708e-05, "loss": 2.2194, "step": 4475 }, { "epoch": 0.11, "learning_rate": 1.9663652215815143e-05, "loss": 2.3245, "step": 4476 }, { "epoch": 0.11, "learning_rate": 1.9663446875333597e-05, "loss": 2.2783, "step": 4477 }, { "epoch": 0.11, "learning_rate": 1.966324147326376e-05, "loss": 2.3585, "step": 4478 }, { "epoch": 0.11, "learning_rate": 1.9663036009606932e-05, "loss": 2.2741, "step": 4479 }, { "epoch": 0.11, "learning_rate": 1.9662830484364427e-05, "loss": 2.4124, "step": 4480 }, { "epoch": 0.11, "learning_rate": 1.9662624897537554e-05, "loss": 2.3695, "step": 4481 }, { "epoch": 0.11, "learning_rate": 1.966241924912762e-05, "loss": 2.2584, "step": 4482 }, { "epoch": 0.11, "learning_rate": 1.9662213539135944e-05, "loss": 2.397, "step": 4483 }, { "epoch": 0.11, "learning_rate": 1.9662007767563833e-05, "loss": 2.1612, "step": 4484 }, { "epoch": 0.11, "learning_rate": 1.9661801934412596e-05, "loss": 2.1619, "step": 4485 }, { "epoch": 0.11, "learning_rate": 1.966159603968355e-05, "loss": 2.187, "step": 4486 }, { "epoch": 0.11, "learning_rate": 1.9661390083377997e-05, "loss": 2.4009, "step": 4487 }, { "epoch": 0.11, "learning_rate": 1.9661184065497263e-05, "loss": 2.304, "step": 4488 }, { "epoch": 0.11, "learning_rate": 1.9660977986042652e-05, "loss": 2.1935, "step": 4489 }, { "epoch": 0.11, "learning_rate": 1.9660771845015478e-05, "loss": 2.2694, "step": 4490 }, { "epoch": 0.11, "learning_rate": 1.966056564241706e-05, "loss": 2.3262, "step": 4491 }, { "epoch": 0.11, "learning_rate": 1.9660359378248705e-05, "loss": 2.287, "step": 4492 }, { "epoch": 0.11, "learning_rate": 1.9660153052511733e-05, "loss": 2.4763, "step": 4493 }, { "epoch": 0.11, "learning_rate": 1.9659946665207456e-05, "loss": 2.387, "step": 4494 }, { "epoch": 0.11, "learning_rate": 1.9659740216337196e-05, "loss": 2.4039, "step": 4495 }, { "epoch": 0.11, "learning_rate": 1.965953370590226e-05, "loss": 2.5378, "step": 4496 }, { "epoch": 0.11, "learning_rate": 1.9659327133903966e-05, "loss": 2.2576, "step": 4497 }, { "epoch": 0.11, "learning_rate": 1.9659120500343633e-05, "loss": 2.2383, "step": 4498 }, { "epoch": 0.11, "learning_rate": 1.965891380522258e-05, "loss": 2.1753, "step": 4499 }, { "epoch": 0.11, "learning_rate": 1.965870704854212e-05, "loss": 2.269, "step": 4500 }, { "epoch": 0.11, "learning_rate": 1.9658500230303574e-05, "loss": 2.2911, "step": 4501 }, { "epoch": 0.11, "learning_rate": 1.9658293350508256e-05, "loss": 2.293, "step": 4502 }, { "epoch": 0.11, "learning_rate": 1.9658086409157485e-05, "loss": 2.2359, "step": 4503 }, { "epoch": 0.11, "learning_rate": 1.965787940625259e-05, "loss": 2.2539, "step": 4504 }, { "epoch": 0.11, "learning_rate": 1.9657672341794877e-05, "loss": 2.3785, "step": 4505 }, { "epoch": 0.11, "learning_rate": 1.965746521578567e-05, "loss": 2.2486, "step": 4506 }, { "epoch": 0.11, "learning_rate": 1.965725802822629e-05, "loss": 2.1941, "step": 4507 }, { "epoch": 0.11, "learning_rate": 1.9657050779118057e-05, "loss": 2.1498, "step": 4508 }, { "epoch": 0.11, "learning_rate": 1.9656843468462294e-05, "loss": 2.3275, "step": 4509 }, { "epoch": 0.11, "learning_rate": 1.9656636096260317e-05, "loss": 2.0894, "step": 4510 }, { "epoch": 0.11, "learning_rate": 1.9656428662513454e-05, "loss": 2.4999, "step": 4511 }, { "epoch": 0.11, "learning_rate": 1.9656221167223024e-05, "loss": 2.3339, "step": 4512 }, { "epoch": 0.11, "learning_rate": 1.965601361039035e-05, "loss": 2.3723, "step": 4513 }, { "epoch": 0.11, "learning_rate": 1.9655805992016754e-05, "loss": 2.3543, "step": 4514 }, { "epoch": 0.11, "learning_rate": 1.965559831210356e-05, "loss": 2.4466, "step": 4515 }, { "epoch": 0.11, "learning_rate": 1.965539057065209e-05, "loss": 2.4486, "step": 4516 }, { "epoch": 0.11, "learning_rate": 1.965518276766367e-05, "loss": 2.5231, "step": 4517 }, { "epoch": 0.11, "learning_rate": 1.9654974903139627e-05, "loss": 2.3546, "step": 4518 }, { "epoch": 0.11, "learning_rate": 1.9654766977081277e-05, "loss": 2.427, "step": 4519 }, { "epoch": 0.11, "learning_rate": 1.9654558989489955e-05, "loss": 2.0924, "step": 4520 }, { "epoch": 0.11, "learning_rate": 1.965435094036698e-05, "loss": 2.3159, "step": 4521 }, { "epoch": 0.11, "learning_rate": 1.965414282971368e-05, "loss": 2.3927, "step": 4522 }, { "epoch": 0.11, "learning_rate": 1.965393465753138e-05, "loss": 2.2813, "step": 4523 }, { "epoch": 0.11, "learning_rate": 1.9653726423821415e-05, "loss": 2.2584, "step": 4524 }, { "epoch": 0.11, "learning_rate": 1.9653518128585097e-05, "loss": 2.414, "step": 4525 }, { "epoch": 0.11, "learning_rate": 1.965330977182377e-05, "loss": 2.2561, "step": 4526 }, { "epoch": 0.11, "learning_rate": 1.9653101353538744e-05, "loss": 2.1613, "step": 4527 }, { "epoch": 0.11, "learning_rate": 1.9652892873731362e-05, "loss": 2.307, "step": 4528 }, { "epoch": 0.11, "learning_rate": 1.965268433240295e-05, "loss": 2.5126, "step": 4529 }, { "epoch": 0.11, "learning_rate": 1.9652475729554834e-05, "loss": 2.2194, "step": 4530 }, { "epoch": 0.11, "learning_rate": 1.9652267065188345e-05, "loss": 2.2636, "step": 4531 }, { "epoch": 0.11, "learning_rate": 1.9652058339304813e-05, "loss": 2.4053, "step": 4532 }, { "epoch": 0.11, "learning_rate": 1.9651849551905565e-05, "loss": 2.1478, "step": 4533 }, { "epoch": 0.11, "learning_rate": 1.9651640702991938e-05, "loss": 2.2641, "step": 4534 }, { "epoch": 0.11, "learning_rate": 1.9651431792565257e-05, "loss": 2.404, "step": 4535 }, { "epoch": 0.11, "learning_rate": 1.9651222820626854e-05, "loss": 2.2871, "step": 4536 }, { "epoch": 0.11, "learning_rate": 1.9651013787178066e-05, "loss": 2.4881, "step": 4537 }, { "epoch": 0.11, "learning_rate": 1.9650804692220223e-05, "loss": 2.2904, "step": 4538 }, { "epoch": 0.11, "learning_rate": 1.9650595535754652e-05, "loss": 2.3769, "step": 4539 }, { "epoch": 0.11, "learning_rate": 1.965038631778269e-05, "loss": 2.4591, "step": 4540 }, { "epoch": 0.11, "learning_rate": 1.9650177038305677e-05, "loss": 2.217, "step": 4541 }, { "epoch": 0.11, "learning_rate": 1.964996769732494e-05, "loss": 2.2497, "step": 4542 }, { "epoch": 0.11, "learning_rate": 1.964975829484181e-05, "loss": 2.1443, "step": 4543 }, { "epoch": 0.11, "learning_rate": 1.9649548830857627e-05, "loss": 2.4204, "step": 4544 }, { "epoch": 0.11, "learning_rate": 1.9649339305373726e-05, "loss": 2.449, "step": 4545 }, { "epoch": 0.11, "learning_rate": 1.964912971839144e-05, "loss": 2.3358, "step": 4546 }, { "epoch": 0.11, "learning_rate": 1.9648920069912106e-05, "loss": 2.3277, "step": 4547 }, { "epoch": 0.11, "learning_rate": 1.964871035993706e-05, "loss": 2.344, "step": 4548 }, { "epoch": 0.11, "learning_rate": 1.964850058846764e-05, "loss": 2.1159, "step": 4549 }, { "epoch": 0.11, "learning_rate": 1.9648290755505178e-05, "loss": 2.3406, "step": 4550 }, { "epoch": 0.11, "learning_rate": 1.9648080861051015e-05, "loss": 2.2766, "step": 4551 }, { "epoch": 0.11, "learning_rate": 1.9647870905106492e-05, "loss": 2.2467, "step": 4552 }, { "epoch": 0.11, "learning_rate": 1.964766088767294e-05, "loss": 2.4258, "step": 4553 }, { "epoch": 0.11, "learning_rate": 1.96474508087517e-05, "loss": 2.3586, "step": 4554 }, { "epoch": 0.11, "learning_rate": 1.964724066834412e-05, "loss": 2.1842, "step": 4555 }, { "epoch": 0.11, "learning_rate": 1.9647030466451526e-05, "loss": 2.2389, "step": 4556 }, { "epoch": 0.11, "learning_rate": 1.964682020307526e-05, "loss": 2.2208, "step": 4557 }, { "epoch": 0.11, "learning_rate": 1.9646609878216666e-05, "loss": 2.3258, "step": 4558 }, { "epoch": 0.11, "learning_rate": 1.9646399491877087e-05, "loss": 2.4446, "step": 4559 }, { "epoch": 0.11, "learning_rate": 1.9646189044057858e-05, "loss": 2.4533, "step": 4560 }, { "epoch": 0.11, "learning_rate": 1.9645978534760324e-05, "loss": 2.2571, "step": 4561 }, { "epoch": 0.11, "learning_rate": 1.9645767963985826e-05, "loss": 2.2489, "step": 4562 }, { "epoch": 0.11, "learning_rate": 1.9645557331735703e-05, "loss": 2.309, "step": 4563 }, { "epoch": 0.11, "learning_rate": 1.9645346638011297e-05, "loss": 2.2882, "step": 4564 }, { "epoch": 0.11, "learning_rate": 1.964513588281396e-05, "loss": 2.3803, "step": 4565 }, { "epoch": 0.11, "learning_rate": 1.9644925066145024e-05, "loss": 2.5923, "step": 4566 }, { "epoch": 0.11, "learning_rate": 1.964471418800584e-05, "loss": 2.2175, "step": 4567 }, { "epoch": 0.11, "learning_rate": 1.9644503248397748e-05, "loss": 2.3443, "step": 4568 }, { "epoch": 0.11, "learning_rate": 1.9644292247322093e-05, "loss": 2.4271, "step": 4569 }, { "epoch": 0.11, "learning_rate": 1.9644081184780222e-05, "loss": 2.5832, "step": 4570 }, { "epoch": 0.11, "learning_rate": 1.964387006077348e-05, "loss": 2.2303, "step": 4571 }, { "epoch": 0.11, "learning_rate": 1.964365887530321e-05, "loss": 2.1581, "step": 4572 }, { "epoch": 0.11, "learning_rate": 1.964344762837076e-05, "loss": 2.2329, "step": 4573 }, { "epoch": 0.11, "learning_rate": 1.9643236319977474e-05, "loss": 2.3365, "step": 4574 }, { "epoch": 0.11, "learning_rate": 1.9643024950124698e-05, "loss": 2.2973, "step": 4575 }, { "epoch": 0.11, "learning_rate": 1.9642813518813786e-05, "loss": 2.2548, "step": 4576 }, { "epoch": 0.11, "learning_rate": 1.964260202604608e-05, "loss": 2.3965, "step": 4577 }, { "epoch": 0.11, "learning_rate": 1.9642390471822928e-05, "loss": 2.1653, "step": 4578 }, { "epoch": 0.11, "learning_rate": 1.9642178856145676e-05, "loss": 2.3008, "step": 4579 }, { "epoch": 0.11, "learning_rate": 1.9641967179015683e-05, "loss": 2.5354, "step": 4580 }, { "epoch": 0.11, "learning_rate": 1.9641755440434286e-05, "loss": 2.2839, "step": 4581 }, { "epoch": 0.11, "learning_rate": 1.964154364040284e-05, "loss": 2.2431, "step": 4582 }, { "epoch": 0.11, "learning_rate": 1.9641331778922695e-05, "loss": 2.2276, "step": 4583 }, { "epoch": 0.11, "learning_rate": 1.96411198559952e-05, "loss": 2.3009, "step": 4584 }, { "epoch": 0.11, "learning_rate": 1.9640907871621707e-05, "loss": 2.4072, "step": 4585 }, { "epoch": 0.11, "learning_rate": 1.9640695825803563e-05, "loss": 2.3427, "step": 4586 }, { "epoch": 0.11, "learning_rate": 1.9640483718542128e-05, "loss": 2.197, "step": 4587 }, { "epoch": 0.11, "learning_rate": 1.9640271549838745e-05, "loss": 2.3078, "step": 4588 }, { "epoch": 0.11, "learning_rate": 1.964005931969477e-05, "loss": 2.4032, "step": 4589 }, { "epoch": 0.11, "learning_rate": 1.9639847028111556e-05, "loss": 2.3366, "step": 4590 }, { "epoch": 0.11, "learning_rate": 1.9639634675090457e-05, "loss": 2.4225, "step": 4591 }, { "epoch": 0.11, "learning_rate": 1.9639422260632824e-05, "loss": 2.4015, "step": 4592 }, { "epoch": 0.11, "learning_rate": 1.963920978474001e-05, "loss": 2.2463, "step": 4593 }, { "epoch": 0.11, "learning_rate": 1.963899724741337e-05, "loss": 2.3338, "step": 4594 }, { "epoch": 0.11, "learning_rate": 1.9638784648654262e-05, "loss": 2.2873, "step": 4595 }, { "epoch": 0.11, "learning_rate": 1.9638571988464033e-05, "loss": 2.498, "step": 4596 }, { "epoch": 0.11, "learning_rate": 1.9638359266844048e-05, "loss": 2.2107, "step": 4597 }, { "epoch": 0.11, "learning_rate": 1.9638146483795658e-05, "loss": 2.3854, "step": 4598 }, { "epoch": 0.11, "learning_rate": 1.9637933639320222e-05, "loss": 2.4991, "step": 4599 }, { "epoch": 0.11, "learning_rate": 1.963772073341909e-05, "loss": 2.3913, "step": 4600 }, { "epoch": 0.11, "learning_rate": 1.9637507766093625e-05, "loss": 2.0905, "step": 4601 }, { "epoch": 0.11, "learning_rate": 1.9637294737345182e-05, "loss": 2.5453, "step": 4602 }, { "epoch": 0.11, "learning_rate": 1.9637081647175117e-05, "loss": 2.457, "step": 4603 }, { "epoch": 0.11, "learning_rate": 1.9636868495584792e-05, "loss": 2.2676, "step": 4604 }, { "epoch": 0.11, "learning_rate": 1.9636655282575563e-05, "loss": 2.2861, "step": 4605 }, { "epoch": 0.11, "learning_rate": 1.963644200814879e-05, "loss": 2.2545, "step": 4606 }, { "epoch": 0.11, "learning_rate": 1.963622867230583e-05, "loss": 2.2657, "step": 4607 }, { "epoch": 0.11, "learning_rate": 1.9636015275048045e-05, "loss": 2.4061, "step": 4608 }, { "epoch": 0.11, "learning_rate": 1.9635801816376797e-05, "loss": 2.2695, "step": 4609 }, { "epoch": 0.11, "learning_rate": 1.963558829629344e-05, "loss": 2.2303, "step": 4610 }, { "epoch": 0.11, "learning_rate": 1.963537471479934e-05, "loss": 2.2374, "step": 4611 }, { "epoch": 0.11, "learning_rate": 1.9635161071895856e-05, "loss": 2.2862, "step": 4612 }, { "epoch": 0.11, "learning_rate": 1.9634947367584353e-05, "loss": 2.2861, "step": 4613 }, { "epoch": 0.11, "learning_rate": 1.963473360186619e-05, "loss": 2.3094, "step": 4614 }, { "epoch": 0.11, "learning_rate": 1.963451977474273e-05, "loss": 2.2277, "step": 4615 }, { "epoch": 0.11, "learning_rate": 1.9634305886215332e-05, "loss": 2.1501, "step": 4616 }, { "epoch": 0.11, "learning_rate": 1.9634091936285366e-05, "loss": 2.5033, "step": 4617 }, { "epoch": 0.11, "learning_rate": 1.963387792495419e-05, "loss": 2.4172, "step": 4618 }, { "epoch": 0.11, "learning_rate": 1.9633663852223172e-05, "loss": 2.3147, "step": 4619 }, { "epoch": 0.11, "learning_rate": 1.9633449718093677e-05, "loss": 2.2565, "step": 4620 }, { "epoch": 0.11, "learning_rate": 1.9633235522567063e-05, "loss": 2.1798, "step": 4621 }, { "epoch": 0.11, "learning_rate": 1.96330212656447e-05, "loss": 2.2671, "step": 4622 }, { "epoch": 0.11, "learning_rate": 1.963280694732796e-05, "loss": 2.1424, "step": 4623 }, { "epoch": 0.11, "learning_rate": 1.9632592567618195e-05, "loss": 2.3907, "step": 4624 }, { "epoch": 0.11, "learning_rate": 1.963237812651678e-05, "loss": 2.3226, "step": 4625 }, { "epoch": 0.11, "learning_rate": 1.9632163624025083e-05, "loss": 2.4005, "step": 4626 }, { "epoch": 0.11, "learning_rate": 1.9631949060144464e-05, "loss": 2.0877, "step": 4627 }, { "epoch": 0.11, "learning_rate": 1.9631734434876297e-05, "loss": 2.1488, "step": 4628 }, { "epoch": 0.11, "learning_rate": 1.9631519748221948e-05, "loss": 2.3537, "step": 4629 }, { "epoch": 0.11, "learning_rate": 1.9631305000182786e-05, "loss": 2.445, "step": 4630 }, { "epoch": 0.11, "learning_rate": 1.9631090190760177e-05, "loss": 2.1421, "step": 4631 }, { "epoch": 0.11, "learning_rate": 1.9630875319955488e-05, "loss": 2.184, "step": 4632 }, { "epoch": 0.11, "learning_rate": 1.9630660387770096e-05, "loss": 2.337, "step": 4633 }, { "epoch": 0.11, "learning_rate": 1.9630445394205363e-05, "loss": 2.35, "step": 4634 }, { "epoch": 0.11, "learning_rate": 1.9630230339262668e-05, "loss": 2.2961, "step": 4635 }, { "epoch": 0.11, "learning_rate": 1.9630015222943376e-05, "loss": 2.2608, "step": 4636 }, { "epoch": 0.11, "learning_rate": 1.9629800045248855e-05, "loss": 2.3976, "step": 4637 }, { "epoch": 0.11, "learning_rate": 1.9629584806180483e-05, "loss": 2.3116, "step": 4638 }, { "epoch": 0.11, "learning_rate": 1.962936950573963e-05, "loss": 2.0516, "step": 4639 }, { "epoch": 0.11, "learning_rate": 1.962915414392766e-05, "loss": 2.1396, "step": 4640 }, { "epoch": 0.11, "learning_rate": 1.962893872074596e-05, "loss": 2.4235, "step": 4641 }, { "epoch": 0.11, "learning_rate": 1.9628723236195893e-05, "loss": 2.385, "step": 4642 }, { "epoch": 0.11, "learning_rate": 1.9628507690278837e-05, "loss": 2.2051, "step": 4643 }, { "epoch": 0.11, "learning_rate": 1.9628292082996162e-05, "loss": 2.2789, "step": 4644 }, { "epoch": 0.11, "learning_rate": 1.9628076414349243e-05, "loss": 2.2625, "step": 4645 }, { "epoch": 0.11, "learning_rate": 1.9627860684339456e-05, "loss": 2.4795, "step": 4646 }, { "epoch": 0.11, "learning_rate": 1.9627644892968173e-05, "loss": 2.272, "step": 4647 }, { "epoch": 0.11, "learning_rate": 1.9627429040236773e-05, "loss": 2.2932, "step": 4648 }, { "epoch": 0.11, "learning_rate": 1.9627213126146634e-05, "loss": 2.3775, "step": 4649 }, { "epoch": 0.11, "learning_rate": 1.9626997150699127e-05, "loss": 2.0753, "step": 4650 }, { "epoch": 0.11, "learning_rate": 1.9626781113895626e-05, "loss": 2.4055, "step": 4651 }, { "epoch": 0.11, "learning_rate": 1.9626565015737513e-05, "loss": 2.2374, "step": 4652 }, { "epoch": 0.11, "learning_rate": 1.9626348856226162e-05, "loss": 2.3805, "step": 4653 }, { "epoch": 0.11, "learning_rate": 1.9626132635362957e-05, "loss": 2.2878, "step": 4654 }, { "epoch": 0.11, "learning_rate": 1.962591635314927e-05, "loss": 2.3156, "step": 4655 }, { "epoch": 0.11, "learning_rate": 1.9625700009586483e-05, "loss": 2.2412, "step": 4656 }, { "epoch": 0.11, "learning_rate": 1.9625483604675968e-05, "loss": 2.3484, "step": 4657 }, { "epoch": 0.11, "learning_rate": 1.9625267138419115e-05, "loss": 2.328, "step": 4658 }, { "epoch": 0.11, "learning_rate": 1.9625050610817294e-05, "loss": 2.196, "step": 4659 }, { "epoch": 0.11, "learning_rate": 1.9624834021871887e-05, "loss": 2.1957, "step": 4660 }, { "epoch": 0.11, "learning_rate": 1.962461737158428e-05, "loss": 2.4662, "step": 4661 }, { "epoch": 0.11, "learning_rate": 1.962440065995585e-05, "loss": 2.2809, "step": 4662 }, { "epoch": 0.11, "learning_rate": 1.9624183886987974e-05, "loss": 2.1811, "step": 4663 }, { "epoch": 0.11, "learning_rate": 1.962396705268204e-05, "loss": 2.3711, "step": 4664 }, { "epoch": 0.11, "learning_rate": 1.962375015703943e-05, "loss": 2.4624, "step": 4665 }, { "epoch": 0.12, "learning_rate": 1.962353320006152e-05, "loss": 2.4198, "step": 4666 }, { "epoch": 0.12, "learning_rate": 1.96233161817497e-05, "loss": 2.1748, "step": 4667 }, { "epoch": 0.12, "learning_rate": 1.962309910210535e-05, "loss": 2.2249, "step": 4668 }, { "epoch": 0.12, "learning_rate": 1.962288196112985e-05, "loss": 2.2394, "step": 4669 }, { "epoch": 0.12, "learning_rate": 1.962266475882459e-05, "loss": 2.3712, "step": 4670 }, { "epoch": 0.12, "learning_rate": 1.962244749519095e-05, "loss": 2.1849, "step": 4671 }, { "epoch": 0.12, "learning_rate": 1.9622230170230317e-05, "loss": 2.2929, "step": 4672 }, { "epoch": 0.12, "learning_rate": 1.9622012783944076e-05, "loss": 2.2347, "step": 4673 }, { "epoch": 0.12, "learning_rate": 1.962179533633361e-05, "loss": 2.3293, "step": 4674 }, { "epoch": 0.12, "learning_rate": 1.9621577827400308e-05, "loss": 2.4978, "step": 4675 }, { "epoch": 0.12, "learning_rate": 1.9621360257145554e-05, "loss": 2.399, "step": 4676 }, { "epoch": 0.12, "learning_rate": 1.9621142625570738e-05, "loss": 2.3173, "step": 4677 }, { "epoch": 0.12, "learning_rate": 1.962092493267724e-05, "loss": 2.3511, "step": 4678 }, { "epoch": 0.12, "learning_rate": 1.9620707178466456e-05, "loss": 2.3097, "step": 4679 }, { "epoch": 0.12, "learning_rate": 1.962048936293977e-05, "loss": 2.551, "step": 4680 }, { "epoch": 0.12, "learning_rate": 1.9620271486098564e-05, "loss": 2.3038, "step": 4681 }, { "epoch": 0.12, "learning_rate": 1.962005354794424e-05, "loss": 2.2988, "step": 4682 }, { "epoch": 0.12, "learning_rate": 1.9619835548478178e-05, "loss": 2.3244, "step": 4683 }, { "epoch": 0.12, "learning_rate": 1.9619617487701768e-05, "loss": 2.228, "step": 4684 }, { "epoch": 0.12, "learning_rate": 1.96193993656164e-05, "loss": 2.2119, "step": 4685 }, { "epoch": 0.12, "learning_rate": 1.9619181182223467e-05, "loss": 2.3126, "step": 4686 }, { "epoch": 0.12, "learning_rate": 1.9618962937524356e-05, "loss": 2.3406, "step": 4687 }, { "epoch": 0.12, "learning_rate": 1.961874463152046e-05, "loss": 2.3287, "step": 4688 }, { "epoch": 0.12, "learning_rate": 1.9618526264213168e-05, "loss": 2.2641, "step": 4689 }, { "epoch": 0.12, "learning_rate": 1.9618307835603875e-05, "loss": 2.3541, "step": 4690 }, { "epoch": 0.12, "learning_rate": 1.961808934569397e-05, "loss": 2.0965, "step": 4691 }, { "epoch": 0.12, "learning_rate": 1.961787079448485e-05, "loss": 2.3475, "step": 4692 }, { "epoch": 0.12, "learning_rate": 1.9617652181977906e-05, "loss": 2.2282, "step": 4693 }, { "epoch": 0.12, "learning_rate": 1.961743350817453e-05, "loss": 2.4124, "step": 4694 }, { "epoch": 0.12, "learning_rate": 1.9617214773076113e-05, "loss": 2.4457, "step": 4695 }, { "epoch": 0.12, "learning_rate": 1.9616995976684054e-05, "loss": 2.3855, "step": 4696 }, { "epoch": 0.12, "learning_rate": 1.9616777118999742e-05, "loss": 2.399, "step": 4697 }, { "epoch": 0.12, "learning_rate": 1.961655820002458e-05, "loss": 2.4277, "step": 4698 }, { "epoch": 0.12, "learning_rate": 1.9616339219759958e-05, "loss": 2.4234, "step": 4699 }, { "epoch": 0.12, "learning_rate": 1.961612017820727e-05, "loss": 2.4402, "step": 4700 }, { "epoch": 0.12, "learning_rate": 1.9615901075367917e-05, "loss": 2.2327, "step": 4701 }, { "epoch": 0.12, "learning_rate": 1.961568191124329e-05, "loss": 2.4733, "step": 4702 }, { "epoch": 0.12, "learning_rate": 1.961546268583479e-05, "loss": 2.5101, "step": 4703 }, { "epoch": 0.12, "learning_rate": 1.9615243399143815e-05, "loss": 2.3007, "step": 4704 }, { "epoch": 0.12, "learning_rate": 1.9615024051171754e-05, "loss": 2.4953, "step": 4705 }, { "epoch": 0.12, "learning_rate": 1.9614804641920018e-05, "loss": 2.3438, "step": 4706 }, { "epoch": 0.12, "learning_rate": 1.9614585171389994e-05, "loss": 2.3432, "step": 4707 }, { "epoch": 0.12, "learning_rate": 1.9614365639583087e-05, "loss": 2.2224, "step": 4708 }, { "epoch": 0.12, "learning_rate": 1.9614146046500694e-05, "loss": 2.3292, "step": 4709 }, { "epoch": 0.12, "learning_rate": 1.9613926392144216e-05, "loss": 2.2959, "step": 4710 }, { "epoch": 0.12, "learning_rate": 1.9613706676515055e-05, "loss": 2.1372, "step": 4711 }, { "epoch": 0.12, "learning_rate": 1.96134868996146e-05, "loss": 2.4486, "step": 4712 }, { "epoch": 0.12, "learning_rate": 1.961326706144427e-05, "loss": 2.2193, "step": 4713 }, { "epoch": 0.12, "learning_rate": 1.961304716200545e-05, "loss": 2.2363, "step": 4714 }, { "epoch": 0.12, "learning_rate": 1.961282720129955e-05, "loss": 2.3593, "step": 4715 }, { "epoch": 0.12, "learning_rate": 1.9612607179327966e-05, "loss": 2.1981, "step": 4716 }, { "epoch": 0.12, "learning_rate": 1.9612387096092107e-05, "loss": 2.2928, "step": 4717 }, { "epoch": 0.12, "learning_rate": 1.9612166951593373e-05, "loss": 2.3035, "step": 4718 }, { "epoch": 0.12, "learning_rate": 1.9611946745833164e-05, "loss": 2.1438, "step": 4719 }, { "epoch": 0.12, "learning_rate": 1.9611726478812887e-05, "loss": 2.4845, "step": 4720 }, { "epoch": 0.12, "learning_rate": 1.9611506150533947e-05, "loss": 2.3214, "step": 4721 }, { "epoch": 0.12, "learning_rate": 1.9611285760997743e-05, "loss": 2.2784, "step": 4722 }, { "epoch": 0.12, "learning_rate": 1.961106531020568e-05, "loss": 2.4485, "step": 4723 }, { "epoch": 0.12, "learning_rate": 1.9610844798159172e-05, "loss": 2.3723, "step": 4724 }, { "epoch": 0.12, "learning_rate": 1.9610624224859614e-05, "loss": 2.362, "step": 4725 }, { "epoch": 0.12, "learning_rate": 1.9610403590308417e-05, "loss": 2.3713, "step": 4726 }, { "epoch": 0.12, "learning_rate": 1.9610182894506987e-05, "loss": 2.3593, "step": 4727 }, { "epoch": 0.12, "learning_rate": 1.960996213745673e-05, "loss": 2.337, "step": 4728 }, { "epoch": 0.12, "learning_rate": 1.9609741319159052e-05, "loss": 2.2843, "step": 4729 }, { "epoch": 0.12, "learning_rate": 1.960952043961536e-05, "loss": 2.2576, "step": 4730 }, { "epoch": 0.12, "learning_rate": 1.960929949882706e-05, "loss": 2.3493, "step": 4731 }, { "epoch": 0.12, "learning_rate": 1.9609078496795565e-05, "loss": 2.2754, "step": 4732 }, { "epoch": 0.12, "learning_rate": 1.960885743352228e-05, "loss": 2.2551, "step": 4733 }, { "epoch": 0.12, "learning_rate": 1.9608636309008616e-05, "loss": 2.3304, "step": 4734 }, { "epoch": 0.12, "learning_rate": 1.9608415123255984e-05, "loss": 2.2573, "step": 4735 }, { "epoch": 0.12, "learning_rate": 1.960819387626579e-05, "loss": 2.4187, "step": 4736 }, { "epoch": 0.12, "learning_rate": 1.9607972568039444e-05, "loss": 2.2555, "step": 4737 }, { "epoch": 0.12, "learning_rate": 1.9607751198578358e-05, "loss": 2.2456, "step": 4738 }, { "epoch": 0.12, "learning_rate": 1.9607529767883943e-05, "loss": 2.4185, "step": 4739 }, { "epoch": 0.12, "learning_rate": 1.9607308275957608e-05, "loss": 2.3555, "step": 4740 }, { "epoch": 0.12, "learning_rate": 1.960708672280077e-05, "loss": 2.4185, "step": 4741 }, { "epoch": 0.12, "learning_rate": 1.9606865108414835e-05, "loss": 2.1555, "step": 4742 }, { "epoch": 0.12, "learning_rate": 1.960664343280122e-05, "loss": 2.3681, "step": 4743 }, { "epoch": 0.12, "learning_rate": 1.960642169596133e-05, "loss": 2.3088, "step": 4744 }, { "epoch": 0.12, "learning_rate": 1.9606199897896587e-05, "loss": 2.1937, "step": 4745 }, { "epoch": 0.12, "learning_rate": 1.9605978038608404e-05, "loss": 2.3519, "step": 4746 }, { "epoch": 0.12, "learning_rate": 1.960575611809819e-05, "loss": 2.1563, "step": 4747 }, { "epoch": 0.12, "learning_rate": 1.960553413636736e-05, "loss": 2.3108, "step": 4748 }, { "epoch": 0.12, "learning_rate": 1.9605312093417336e-05, "loss": 2.3106, "step": 4749 }, { "epoch": 0.12, "learning_rate": 1.9605089989249525e-05, "loss": 2.2677, "step": 4750 }, { "epoch": 0.12, "learning_rate": 1.9604867823865345e-05, "loss": 2.3707, "step": 4751 }, { "epoch": 0.12, "learning_rate": 1.9604645597266213e-05, "loss": 2.198, "step": 4752 }, { "epoch": 0.12, "learning_rate": 1.9604423309453544e-05, "loss": 2.1698, "step": 4753 }, { "epoch": 0.12, "learning_rate": 1.9604200960428756e-05, "loss": 2.1959, "step": 4754 }, { "epoch": 0.12, "learning_rate": 1.960397855019326e-05, "loss": 2.2806, "step": 4755 }, { "epoch": 0.12, "learning_rate": 1.9603756078748487e-05, "loss": 2.2326, "step": 4756 }, { "epoch": 0.12, "learning_rate": 1.9603533546095843e-05, "loss": 2.4764, "step": 4757 }, { "epoch": 0.12, "learning_rate": 1.960331095223675e-05, "loss": 2.2237, "step": 4758 }, { "epoch": 0.12, "learning_rate": 1.9603088297172625e-05, "loss": 2.3088, "step": 4759 }, { "epoch": 0.12, "learning_rate": 1.960286558090489e-05, "loss": 2.3558, "step": 4760 }, { "epoch": 0.12, "learning_rate": 1.9602642803434965e-05, "loss": 2.3181, "step": 4761 }, { "epoch": 0.12, "learning_rate": 1.9602419964764268e-05, "loss": 2.3631, "step": 4762 }, { "epoch": 0.12, "learning_rate": 1.9602197064894218e-05, "loss": 2.3799, "step": 4763 }, { "epoch": 0.12, "learning_rate": 1.9601974103826234e-05, "loss": 2.4477, "step": 4764 }, { "epoch": 0.12, "learning_rate": 1.9601751081561743e-05, "loss": 2.3026, "step": 4765 }, { "epoch": 0.12, "learning_rate": 1.9601527998102164e-05, "loss": 2.4227, "step": 4766 }, { "epoch": 0.12, "learning_rate": 1.9601304853448916e-05, "loss": 2.2761, "step": 4767 }, { "epoch": 0.12, "learning_rate": 1.9601081647603423e-05, "loss": 2.4327, "step": 4768 }, { "epoch": 0.12, "learning_rate": 1.9600858380567108e-05, "loss": 2.152, "step": 4769 }, { "epoch": 0.12, "learning_rate": 1.9600635052341394e-05, "loss": 2.2658, "step": 4770 }, { "epoch": 0.12, "learning_rate": 1.9600411662927707e-05, "loss": 2.0642, "step": 4771 }, { "epoch": 0.12, "learning_rate": 1.9600188212327463e-05, "loss": 2.225, "step": 4772 }, { "epoch": 0.12, "learning_rate": 1.9599964700542092e-05, "loss": 2.3707, "step": 4773 }, { "epoch": 0.12, "learning_rate": 1.959974112757302e-05, "loss": 2.1919, "step": 4774 }, { "epoch": 0.12, "learning_rate": 1.959951749342167e-05, "loss": 2.1888, "step": 4775 }, { "epoch": 0.12, "learning_rate": 1.9599293798089464e-05, "loss": 2.1039, "step": 4776 }, { "epoch": 0.12, "learning_rate": 1.959907004157783e-05, "loss": 2.4229, "step": 4777 }, { "epoch": 0.12, "learning_rate": 1.9598846223888194e-05, "loss": 2.3506, "step": 4778 }, { "epoch": 0.12, "learning_rate": 1.9598622345021987e-05, "loss": 2.3283, "step": 4779 }, { "epoch": 0.12, "learning_rate": 1.959839840498063e-05, "loss": 2.2723, "step": 4780 }, { "epoch": 0.12, "learning_rate": 1.959817440376555e-05, "loss": 2.381, "step": 4781 }, { "epoch": 0.12, "learning_rate": 1.959795034137818e-05, "loss": 2.3755, "step": 4782 }, { "epoch": 0.12, "learning_rate": 1.9597726217819938e-05, "loss": 2.2389, "step": 4783 }, { "epoch": 0.12, "learning_rate": 1.959750203309226e-05, "loss": 2.0896, "step": 4784 }, { "epoch": 0.12, "learning_rate": 1.959727778719658e-05, "loss": 2.4398, "step": 4785 }, { "epoch": 0.12, "learning_rate": 1.9597053480134315e-05, "loss": 2.2916, "step": 4786 }, { "epoch": 0.12, "learning_rate": 1.9596829111906904e-05, "loss": 2.2383, "step": 4787 }, { "epoch": 0.12, "learning_rate": 1.9596604682515777e-05, "loss": 2.3253, "step": 4788 }, { "epoch": 0.12, "learning_rate": 1.9596380191962355e-05, "loss": 2.4379, "step": 4789 }, { "epoch": 0.12, "learning_rate": 1.9596155640248075e-05, "loss": 2.3103, "step": 4790 }, { "epoch": 0.12, "learning_rate": 1.959593102737437e-05, "loss": 2.2691, "step": 4791 }, { "epoch": 0.12, "learning_rate": 1.959570635334267e-05, "loss": 2.2834, "step": 4792 }, { "epoch": 0.12, "learning_rate": 1.9595481618154403e-05, "loss": 2.2018, "step": 4793 }, { "epoch": 0.12, "learning_rate": 1.9595256821811008e-05, "loss": 2.2688, "step": 4794 }, { "epoch": 0.12, "learning_rate": 1.959503196431391e-05, "loss": 2.2394, "step": 4795 }, { "epoch": 0.12, "learning_rate": 1.9594807045664552e-05, "loss": 2.3882, "step": 4796 }, { "epoch": 0.12, "learning_rate": 1.9594582065864358e-05, "loss": 2.1168, "step": 4797 }, { "epoch": 0.12, "learning_rate": 1.959435702491477e-05, "loss": 2.3658, "step": 4798 }, { "epoch": 0.12, "learning_rate": 1.959413192281721e-05, "loss": 2.4657, "step": 4799 }, { "epoch": 0.12, "learning_rate": 1.9593906759573125e-05, "loss": 2.1289, "step": 4800 }, { "epoch": 0.12, "learning_rate": 1.9593681535183947e-05, "loss": 2.2095, "step": 4801 }, { "epoch": 0.12, "learning_rate": 1.9593456249651107e-05, "loss": 2.4021, "step": 4802 }, { "epoch": 0.12, "learning_rate": 1.9593230902976047e-05, "loss": 2.2309, "step": 4803 }, { "epoch": 0.12, "learning_rate": 1.95930054951602e-05, "loss": 2.1893, "step": 4804 }, { "epoch": 0.12, "learning_rate": 1.9592780026205002e-05, "loss": 2.3532, "step": 4805 }, { "epoch": 0.12, "learning_rate": 1.959255449611189e-05, "loss": 2.3534, "step": 4806 }, { "epoch": 0.12, "learning_rate": 1.95923289048823e-05, "loss": 2.3116, "step": 4807 }, { "epoch": 0.12, "learning_rate": 1.9592103252517676e-05, "loss": 2.3011, "step": 4808 }, { "epoch": 0.12, "learning_rate": 1.959187753901945e-05, "loss": 2.3565, "step": 4809 }, { "epoch": 0.12, "learning_rate": 1.9591651764389064e-05, "loss": 2.1929, "step": 4810 }, { "epoch": 0.12, "learning_rate": 1.9591425928627955e-05, "loss": 2.5104, "step": 4811 }, { "epoch": 0.12, "learning_rate": 1.9591200031737562e-05, "loss": 2.3136, "step": 4812 }, { "epoch": 0.12, "learning_rate": 1.9590974073719325e-05, "loss": 2.4626, "step": 4813 }, { "epoch": 0.12, "learning_rate": 1.9590748054574685e-05, "loss": 2.2499, "step": 4814 }, { "epoch": 0.12, "learning_rate": 1.9590521974305085e-05, "loss": 2.2506, "step": 4815 }, { "epoch": 0.12, "learning_rate": 1.9590295832911957e-05, "loss": 2.2874, "step": 4816 }, { "epoch": 0.12, "learning_rate": 1.9590069630396752e-05, "loss": 2.1106, "step": 4817 }, { "epoch": 0.12, "learning_rate": 1.9589843366760907e-05, "loss": 2.4525, "step": 4818 }, { "epoch": 0.12, "learning_rate": 1.9589617042005868e-05, "loss": 2.4187, "step": 4819 }, { "epoch": 0.12, "learning_rate": 1.9589390656133068e-05, "loss": 2.3422, "step": 4820 }, { "epoch": 0.12, "learning_rate": 1.958916420914396e-05, "loss": 2.3443, "step": 4821 }, { "epoch": 0.12, "learning_rate": 1.9588937701039988e-05, "loss": 2.1894, "step": 4822 }, { "epoch": 0.12, "learning_rate": 1.9588711131822583e-05, "loss": 2.4161, "step": 4823 }, { "epoch": 0.12, "learning_rate": 1.9588484501493202e-05, "loss": 2.2468, "step": 4824 }, { "epoch": 0.12, "learning_rate": 1.958825781005328e-05, "loss": 2.4308, "step": 4825 }, { "epoch": 0.12, "learning_rate": 1.958803105750427e-05, "loss": 2.3662, "step": 4826 }, { "epoch": 0.12, "learning_rate": 1.9587804243847616e-05, "loss": 2.4105, "step": 4827 }, { "epoch": 0.12, "learning_rate": 1.9587577369084756e-05, "loss": 2.4701, "step": 4828 }, { "epoch": 0.12, "learning_rate": 1.9587350433217144e-05, "loss": 2.3821, "step": 4829 }, { "epoch": 0.12, "learning_rate": 1.9587123436246224e-05, "loss": 2.2706, "step": 4830 }, { "epoch": 0.12, "learning_rate": 1.9586896378173437e-05, "loss": 2.2089, "step": 4831 }, { "epoch": 0.12, "learning_rate": 1.958666925900024e-05, "loss": 2.4315, "step": 4832 }, { "epoch": 0.12, "learning_rate": 1.9586442078728075e-05, "loss": 2.328, "step": 4833 }, { "epoch": 0.12, "learning_rate": 1.9586214837358388e-05, "loss": 2.1617, "step": 4834 }, { "epoch": 0.12, "learning_rate": 1.9585987534892632e-05, "loss": 2.101, "step": 4835 }, { "epoch": 0.12, "learning_rate": 1.958576017133225e-05, "loss": 2.369, "step": 4836 }, { "epoch": 0.12, "learning_rate": 1.9585532746678698e-05, "loss": 2.4891, "step": 4837 }, { "epoch": 0.12, "learning_rate": 1.9585305260933423e-05, "loss": 2.3631, "step": 4838 }, { "epoch": 0.12, "learning_rate": 1.9585077714097872e-05, "loss": 2.2916, "step": 4839 }, { "epoch": 0.12, "learning_rate": 1.9584850106173496e-05, "loss": 2.4044, "step": 4840 }, { "epoch": 0.12, "learning_rate": 1.958462243716175e-05, "loss": 2.3606, "step": 4841 }, { "epoch": 0.12, "learning_rate": 1.9584394707064075e-05, "loss": 2.3319, "step": 4842 }, { "epoch": 0.12, "learning_rate": 1.9584166915881935e-05, "loss": 2.3028, "step": 4843 }, { "epoch": 0.12, "learning_rate": 1.9583939063616776e-05, "loss": 2.348, "step": 4844 }, { "epoch": 0.12, "learning_rate": 1.9583711150270045e-05, "loss": 2.3082, "step": 4845 }, { "epoch": 0.12, "learning_rate": 1.9583483175843203e-05, "loss": 2.4511, "step": 4846 }, { "epoch": 0.12, "learning_rate": 1.9583255140337702e-05, "loss": 2.422, "step": 4847 }, { "epoch": 0.12, "learning_rate": 1.958302704375499e-05, "loss": 2.3169, "step": 4848 }, { "epoch": 0.12, "learning_rate": 1.9582798886096524e-05, "loss": 2.4401, "step": 4849 }, { "epoch": 0.12, "learning_rate": 1.9582570667363757e-05, "loss": 2.2649, "step": 4850 }, { "epoch": 0.12, "learning_rate": 1.9582342387558145e-05, "loss": 2.4238, "step": 4851 }, { "epoch": 0.12, "learning_rate": 1.9582114046681143e-05, "loss": 2.2095, "step": 4852 }, { "epoch": 0.12, "learning_rate": 1.9581885644734204e-05, "loss": 2.3653, "step": 4853 }, { "epoch": 0.12, "learning_rate": 1.9581657181718784e-05, "loss": 2.3722, "step": 4854 }, { "epoch": 0.12, "learning_rate": 1.9581428657636345e-05, "loss": 2.2609, "step": 4855 }, { "epoch": 0.12, "learning_rate": 1.9581200072488332e-05, "loss": 2.2651, "step": 4856 }, { "epoch": 0.12, "learning_rate": 1.9580971426276214e-05, "loss": 2.2433, "step": 4857 }, { "epoch": 0.12, "learning_rate": 1.958074271900144e-05, "loss": 2.1686, "step": 4858 }, { "epoch": 0.12, "learning_rate": 1.9580513950665472e-05, "loss": 2.2349, "step": 4859 }, { "epoch": 0.12, "learning_rate": 1.9580285121269767e-05, "loss": 2.1514, "step": 4860 }, { "epoch": 0.12, "learning_rate": 1.958005623081578e-05, "loss": 2.2345, "step": 4861 }, { "epoch": 0.12, "learning_rate": 1.9579827279304973e-05, "loss": 2.2196, "step": 4862 }, { "epoch": 0.12, "learning_rate": 1.9579598266738806e-05, "loss": 2.2704, "step": 4863 }, { "epoch": 0.12, "learning_rate": 1.9579369193118734e-05, "loss": 2.1983, "step": 4864 }, { "epoch": 0.12, "learning_rate": 1.9579140058446225e-05, "loss": 2.3351, "step": 4865 }, { "epoch": 0.12, "learning_rate": 1.9578910862722736e-05, "loss": 2.2744, "step": 4866 }, { "epoch": 0.12, "learning_rate": 1.9578681605949722e-05, "loss": 2.2128, "step": 4867 }, { "epoch": 0.12, "learning_rate": 1.957845228812865e-05, "loss": 2.2611, "step": 4868 }, { "epoch": 0.12, "learning_rate": 1.957822290926098e-05, "loss": 2.273, "step": 4869 }, { "epoch": 0.12, "learning_rate": 1.9577993469348175e-05, "loss": 2.3502, "step": 4870 }, { "epoch": 0.12, "learning_rate": 1.9577763968391694e-05, "loss": 2.3429, "step": 4871 }, { "epoch": 0.12, "learning_rate": 1.9577534406393004e-05, "loss": 2.3497, "step": 4872 }, { "epoch": 0.12, "learning_rate": 1.9577304783353567e-05, "loss": 2.3078, "step": 4873 }, { "epoch": 0.12, "learning_rate": 1.957707509927484e-05, "loss": 2.2064, "step": 4874 }, { "epoch": 0.12, "learning_rate": 1.9576845354158297e-05, "loss": 2.3455, "step": 4875 }, { "epoch": 0.12, "learning_rate": 1.9576615548005396e-05, "loss": 2.2992, "step": 4876 }, { "epoch": 0.12, "learning_rate": 1.9576385680817605e-05, "loss": 2.2139, "step": 4877 }, { "epoch": 0.12, "learning_rate": 1.9576155752596384e-05, "loss": 2.2744, "step": 4878 }, { "epoch": 0.12, "learning_rate": 1.9575925763343204e-05, "loss": 2.423, "step": 4879 }, { "epoch": 0.12, "learning_rate": 1.9575695713059527e-05, "loss": 2.3011, "step": 4880 }, { "epoch": 0.12, "learning_rate": 1.9575465601746822e-05, "loss": 2.3964, "step": 4881 }, { "epoch": 0.12, "learning_rate": 1.9575235429406555e-05, "loss": 2.3767, "step": 4882 }, { "epoch": 0.12, "learning_rate": 1.957500519604019e-05, "loss": 2.4112, "step": 4883 }, { "epoch": 0.12, "learning_rate": 1.9574774901649198e-05, "loss": 2.5083, "step": 4884 }, { "epoch": 0.12, "learning_rate": 1.9574544546235045e-05, "loss": 2.2188, "step": 4885 }, { "epoch": 0.12, "learning_rate": 1.9574314129799198e-05, "loss": 2.1742, "step": 4886 }, { "epoch": 0.12, "learning_rate": 1.957408365234313e-05, "loss": 2.2687, "step": 4887 }, { "epoch": 0.12, "learning_rate": 1.9573853113868303e-05, "loss": 2.3444, "step": 4888 }, { "epoch": 0.12, "learning_rate": 1.9573622514376192e-05, "loss": 2.3691, "step": 4889 }, { "epoch": 0.12, "learning_rate": 1.9573391853868263e-05, "loss": 2.4128, "step": 4890 }, { "epoch": 0.12, "learning_rate": 1.957316113234599e-05, "loss": 2.3002, "step": 4891 }, { "epoch": 0.12, "learning_rate": 1.957293034981084e-05, "loss": 2.2705, "step": 4892 }, { "epoch": 0.12, "learning_rate": 1.9572699506264285e-05, "loss": 2.2601, "step": 4893 }, { "epoch": 0.12, "learning_rate": 1.9572468601707796e-05, "loss": 2.4364, "step": 4894 }, { "epoch": 0.12, "learning_rate": 1.9572237636142848e-05, "loss": 2.3974, "step": 4895 }, { "epoch": 0.12, "learning_rate": 1.9572006609570908e-05, "loss": 2.3938, "step": 4896 }, { "epoch": 0.12, "learning_rate": 1.957177552199345e-05, "loss": 2.2185, "step": 4897 }, { "epoch": 0.12, "learning_rate": 1.9571544373411945e-05, "loss": 2.3629, "step": 4898 }, { "epoch": 0.12, "learning_rate": 1.9571313163827874e-05, "loss": 2.2843, "step": 4899 }, { "epoch": 0.12, "learning_rate": 1.9571081893242702e-05, "loss": 2.2455, "step": 4900 }, { "epoch": 0.12, "learning_rate": 1.9570850561657907e-05, "loss": 2.3453, "step": 4901 }, { "epoch": 0.12, "learning_rate": 1.957061916907496e-05, "loss": 2.2198, "step": 4902 }, { "epoch": 0.12, "learning_rate": 1.957038771549534e-05, "loss": 2.0798, "step": 4903 }, { "epoch": 0.12, "learning_rate": 1.957015620092052e-05, "loss": 2.1739, "step": 4904 }, { "epoch": 0.12, "learning_rate": 1.9569924625351976e-05, "loss": 2.2744, "step": 4905 }, { "epoch": 0.12, "learning_rate": 1.956969298879118e-05, "loss": 2.2139, "step": 4906 }, { "epoch": 0.12, "learning_rate": 1.9569461291239617e-05, "loss": 2.4413, "step": 4907 }, { "epoch": 0.12, "learning_rate": 1.9569229532698758e-05, "loss": 2.0851, "step": 4908 }, { "epoch": 0.12, "learning_rate": 1.9568997713170076e-05, "loss": 2.4568, "step": 4909 }, { "epoch": 0.12, "learning_rate": 1.9568765832655057e-05, "loss": 2.2802, "step": 4910 }, { "epoch": 0.12, "learning_rate": 1.9568533891155176e-05, "loss": 2.3131, "step": 4911 }, { "epoch": 0.12, "learning_rate": 1.9568301888671908e-05, "loss": 2.1981, "step": 4912 }, { "epoch": 0.12, "learning_rate": 1.9568069825206733e-05, "loss": 2.4355, "step": 4913 }, { "epoch": 0.12, "learning_rate": 1.9567837700761135e-05, "loss": 2.2588, "step": 4914 }, { "epoch": 0.12, "learning_rate": 1.9567605515336586e-05, "loss": 2.2103, "step": 4915 }, { "epoch": 0.12, "learning_rate": 1.9567373268934568e-05, "loss": 2.3046, "step": 4916 }, { "epoch": 0.12, "learning_rate": 1.9567140961556563e-05, "loss": 2.3218, "step": 4917 }, { "epoch": 0.12, "learning_rate": 1.9566908593204054e-05, "loss": 2.358, "step": 4918 }, { "epoch": 0.12, "learning_rate": 1.9566676163878515e-05, "loss": 2.11, "step": 4919 }, { "epoch": 0.12, "learning_rate": 1.9566443673581434e-05, "loss": 2.206, "step": 4920 }, { "epoch": 0.12, "learning_rate": 1.9566211122314288e-05, "loss": 2.1853, "step": 4921 }, { "epoch": 0.12, "learning_rate": 1.956597851007856e-05, "loss": 2.3164, "step": 4922 }, { "epoch": 0.12, "learning_rate": 1.9565745836875733e-05, "loss": 2.3727, "step": 4923 }, { "epoch": 0.12, "learning_rate": 1.9565513102707295e-05, "loss": 2.2542, "step": 4924 }, { "epoch": 0.12, "learning_rate": 1.956528030757472e-05, "loss": 2.2679, "step": 4925 }, { "epoch": 0.12, "learning_rate": 1.95650474514795e-05, "loss": 2.5163, "step": 4926 }, { "epoch": 0.12, "learning_rate": 1.9564814534423113e-05, "loss": 2.3751, "step": 4927 }, { "epoch": 0.12, "learning_rate": 1.9564581556407046e-05, "loss": 2.2997, "step": 4928 }, { "epoch": 0.12, "learning_rate": 1.9564348517432786e-05, "loss": 2.2539, "step": 4929 }, { "epoch": 0.12, "learning_rate": 1.9564115417501814e-05, "loss": 2.2077, "step": 4930 }, { "epoch": 0.12, "learning_rate": 1.9563882256615616e-05, "loss": 2.2787, "step": 4931 }, { "epoch": 0.12, "learning_rate": 1.9563649034775684e-05, "loss": 2.3525, "step": 4932 }, { "epoch": 0.12, "learning_rate": 1.9563415751983498e-05, "loss": 2.2896, "step": 4933 }, { "epoch": 0.12, "learning_rate": 1.9563182408240543e-05, "loss": 2.1364, "step": 4934 }, { "epoch": 0.12, "learning_rate": 1.9562949003548315e-05, "loss": 2.3439, "step": 4935 }, { "epoch": 0.12, "learning_rate": 1.9562715537908296e-05, "loss": 2.2607, "step": 4936 }, { "epoch": 0.12, "learning_rate": 1.9562482011321973e-05, "loss": 2.2213, "step": 4937 }, { "epoch": 0.12, "learning_rate": 1.9562248423790836e-05, "loss": 2.2541, "step": 4938 }, { "epoch": 0.12, "learning_rate": 1.956201477531637e-05, "loss": 2.2732, "step": 4939 }, { "epoch": 0.12, "learning_rate": 1.9561781065900072e-05, "loss": 2.2524, "step": 4940 }, { "epoch": 0.12, "learning_rate": 1.9561547295543427e-05, "loss": 2.4376, "step": 4941 }, { "epoch": 0.12, "learning_rate": 1.9561313464247927e-05, "loss": 2.3774, "step": 4942 }, { "epoch": 0.12, "learning_rate": 1.9561079572015055e-05, "loss": 2.2923, "step": 4943 }, { "epoch": 0.12, "learning_rate": 1.956084561884631e-05, "loss": 2.365, "step": 4944 }, { "epoch": 0.12, "learning_rate": 1.956061160474318e-05, "loss": 2.3625, "step": 4945 }, { "epoch": 0.12, "learning_rate": 1.9560377529707158e-05, "loss": 2.2647, "step": 4946 }, { "epoch": 0.12, "learning_rate": 1.9560143393739732e-05, "loss": 2.3159, "step": 4947 }, { "epoch": 0.12, "learning_rate": 1.95599091968424e-05, "loss": 2.298, "step": 4948 }, { "epoch": 0.12, "learning_rate": 1.9559674939016647e-05, "loss": 2.3117, "step": 4949 }, { "epoch": 0.12, "learning_rate": 1.9559440620263973e-05, "loss": 2.2151, "step": 4950 }, { "epoch": 0.12, "learning_rate": 1.9559206240585865e-05, "loss": 2.4904, "step": 4951 }, { "epoch": 0.12, "learning_rate": 1.9558971799983822e-05, "loss": 2.2476, "step": 4952 }, { "epoch": 0.12, "learning_rate": 1.9558737298459338e-05, "loss": 2.3467, "step": 4953 }, { "epoch": 0.12, "learning_rate": 1.9558502736013906e-05, "loss": 2.3288, "step": 4954 }, { "epoch": 0.12, "learning_rate": 1.955826811264902e-05, "loss": 2.3221, "step": 4955 }, { "epoch": 0.12, "learning_rate": 1.955803342836618e-05, "loss": 2.1132, "step": 4956 }, { "epoch": 0.12, "learning_rate": 1.9557798683166874e-05, "loss": 2.2624, "step": 4957 }, { "epoch": 0.12, "learning_rate": 1.95575638770526e-05, "loss": 2.4077, "step": 4958 }, { "epoch": 0.12, "learning_rate": 1.955732901002486e-05, "loss": 2.1593, "step": 4959 }, { "epoch": 0.12, "learning_rate": 1.955709408208515e-05, "loss": 2.1746, "step": 4960 }, { "epoch": 0.12, "learning_rate": 1.955685909323496e-05, "loss": 2.1321, "step": 4961 }, { "epoch": 0.12, "learning_rate": 1.9556624043475796e-05, "loss": 2.181, "step": 4962 }, { "epoch": 0.12, "learning_rate": 1.955638893280915e-05, "loss": 2.3401, "step": 4963 }, { "epoch": 0.12, "learning_rate": 1.9556153761236525e-05, "loss": 2.2888, "step": 4964 }, { "epoch": 0.12, "learning_rate": 1.9555918528759416e-05, "loss": 2.161, "step": 4965 }, { "epoch": 0.12, "learning_rate": 1.9555683235379326e-05, "loss": 2.2054, "step": 4966 }, { "epoch": 0.12, "learning_rate": 1.955544788109775e-05, "loss": 2.2565, "step": 4967 }, { "epoch": 0.12, "learning_rate": 1.9555212465916193e-05, "loss": 2.2789, "step": 4968 }, { "epoch": 0.12, "learning_rate": 1.9554976989836156e-05, "loss": 2.231, "step": 4969 }, { "epoch": 0.12, "learning_rate": 1.955474145285913e-05, "loss": 2.1059, "step": 4970 }, { "epoch": 0.12, "learning_rate": 1.9554505854986628e-05, "loss": 2.3527, "step": 4971 }, { "epoch": 0.12, "learning_rate": 1.9554270196220144e-05, "loss": 2.279, "step": 4972 }, { "epoch": 0.12, "learning_rate": 1.9554034476561183e-05, "loss": 2.2638, "step": 4973 }, { "epoch": 0.12, "learning_rate": 1.955379869601125e-05, "loss": 2.2267, "step": 4974 }, { "epoch": 0.12, "learning_rate": 1.955356285457184e-05, "loss": 2.343, "step": 4975 }, { "epoch": 0.12, "learning_rate": 1.955332695224446e-05, "loss": 2.3222, "step": 4976 }, { "epoch": 0.12, "learning_rate": 1.9553090989030618e-05, "loss": 2.3, "step": 4977 }, { "epoch": 0.12, "learning_rate": 1.955285496493181e-05, "loss": 2.2349, "step": 4978 }, { "epoch": 0.12, "learning_rate": 1.9552618879949544e-05, "loss": 2.4036, "step": 4979 }, { "epoch": 0.12, "learning_rate": 1.9552382734085325e-05, "loss": 2.3692, "step": 4980 }, { "epoch": 0.12, "learning_rate": 1.955214652734066e-05, "loss": 2.265, "step": 4981 }, { "epoch": 0.12, "learning_rate": 1.955191025971705e-05, "loss": 2.4269, "step": 4982 }, { "epoch": 0.12, "learning_rate": 1.9551673931216005e-05, "loss": 2.3268, "step": 4983 }, { "epoch": 0.12, "learning_rate": 1.955143754183903e-05, "loss": 2.2528, "step": 4984 }, { "epoch": 0.12, "learning_rate": 1.955120109158763e-05, "loss": 2.1984, "step": 4985 }, { "epoch": 0.12, "learning_rate": 1.955096458046331e-05, "loss": 2.182, "step": 4986 }, { "epoch": 0.12, "learning_rate": 1.9550728008467584e-05, "loss": 2.4356, "step": 4987 }, { "epoch": 0.12, "learning_rate": 1.9550491375601955e-05, "loss": 2.3364, "step": 4988 }, { "epoch": 0.12, "learning_rate": 1.9550254681867932e-05, "loss": 2.3158, "step": 4989 }, { "epoch": 0.12, "learning_rate": 1.9550017927267022e-05, "loss": 2.2226, "step": 4990 }, { "epoch": 0.12, "learning_rate": 1.9549781111800738e-05, "loss": 2.2554, "step": 4991 }, { "epoch": 0.12, "learning_rate": 1.9549544235470582e-05, "loss": 2.1696, "step": 4992 }, { "epoch": 0.12, "learning_rate": 1.9549307298278073e-05, "loss": 2.3547, "step": 4993 }, { "epoch": 0.12, "learning_rate": 1.9549070300224716e-05, "loss": 2.1985, "step": 4994 }, { "epoch": 0.12, "learning_rate": 1.954883324131202e-05, "loss": 2.2495, "step": 4995 }, { "epoch": 0.12, "learning_rate": 1.95485961215415e-05, "loss": 2.333, "step": 4996 }, { "epoch": 0.12, "learning_rate": 1.9548358940914663e-05, "loss": 2.4954, "step": 4997 }, { "epoch": 0.12, "learning_rate": 1.9548121699433025e-05, "loss": 2.2768, "step": 4998 }, { "epoch": 0.12, "learning_rate": 1.9547884397098095e-05, "loss": 2.3931, "step": 4999 }, { "epoch": 0.12, "learning_rate": 1.9547647033911385e-05, "loss": 2.2998, "step": 5000 }, { "epoch": 0.12, "learning_rate": 1.954740960987441e-05, "loss": 2.0562, "step": 5001 }, { "epoch": 0.12, "learning_rate": 1.954717212498868e-05, "loss": 2.3452, "step": 5002 }, { "epoch": 0.12, "learning_rate": 1.9546934579255716e-05, "loss": 2.2641, "step": 5003 }, { "epoch": 0.12, "learning_rate": 1.9546696972677023e-05, "loss": 2.2276, "step": 5004 }, { "epoch": 0.12, "learning_rate": 1.954645930525412e-05, "loss": 2.3745, "step": 5005 }, { "epoch": 0.12, "learning_rate": 1.954622157698852e-05, "loss": 2.312, "step": 5006 }, { "epoch": 0.12, "learning_rate": 1.954598378788174e-05, "loss": 2.2452, "step": 5007 }, { "epoch": 0.12, "learning_rate": 1.954574593793529e-05, "loss": 2.1953, "step": 5008 }, { "epoch": 0.12, "learning_rate": 1.9545508027150696e-05, "loss": 2.2154, "step": 5009 }, { "epoch": 0.12, "learning_rate": 1.9545270055529466e-05, "loss": 2.4241, "step": 5010 }, { "epoch": 0.12, "learning_rate": 1.954503202307312e-05, "loss": 2.3012, "step": 5011 }, { "epoch": 0.12, "learning_rate": 1.954479392978317e-05, "loss": 2.3237, "step": 5012 }, { "epoch": 0.12, "learning_rate": 1.9544555775661144e-05, "loss": 2.3224, "step": 5013 }, { "epoch": 0.12, "learning_rate": 1.9544317560708552e-05, "loss": 2.2093, "step": 5014 }, { "epoch": 0.12, "learning_rate": 1.9544079284926912e-05, "loss": 2.2485, "step": 5015 }, { "epoch": 0.12, "learning_rate": 1.9543840948317746e-05, "loss": 2.2348, "step": 5016 }, { "epoch": 0.12, "learning_rate": 1.954360255088257e-05, "loss": 2.1822, "step": 5017 }, { "epoch": 0.12, "learning_rate": 1.9543364092622905e-05, "loss": 2.1691, "step": 5018 }, { "epoch": 0.12, "learning_rate": 1.9543125573540274e-05, "loss": 2.3368, "step": 5019 }, { "epoch": 0.12, "learning_rate": 1.954288699363619e-05, "loss": 2.3369, "step": 5020 }, { "epoch": 0.12, "learning_rate": 1.9542648352912177e-05, "loss": 2.135, "step": 5021 }, { "epoch": 0.12, "learning_rate": 1.954240965136976e-05, "loss": 2.1553, "step": 5022 }, { "epoch": 0.12, "learning_rate": 1.9542170889010452e-05, "loss": 2.242, "step": 5023 }, { "epoch": 0.12, "learning_rate": 1.954193206583578e-05, "loss": 2.2453, "step": 5024 }, { "epoch": 0.12, "learning_rate": 1.9541693181847264e-05, "loss": 2.1626, "step": 5025 }, { "epoch": 0.12, "learning_rate": 1.9541454237046433e-05, "loss": 2.3006, "step": 5026 }, { "epoch": 0.12, "learning_rate": 1.95412152314348e-05, "loss": 2.3938, "step": 5027 }, { "epoch": 0.12, "learning_rate": 1.9540976165013896e-05, "loss": 2.1884, "step": 5028 }, { "epoch": 0.12, "learning_rate": 1.9540737037785237e-05, "loss": 2.2993, "step": 5029 }, { "epoch": 0.12, "learning_rate": 1.9540497849750356e-05, "loss": 2.2152, "step": 5030 }, { "epoch": 0.12, "learning_rate": 1.9540258600910772e-05, "loss": 2.4111, "step": 5031 }, { "epoch": 0.12, "learning_rate": 1.9540019291268015e-05, "loss": 2.4445, "step": 5032 }, { "epoch": 0.12, "learning_rate": 1.95397799208236e-05, "loss": 2.3409, "step": 5033 }, { "epoch": 0.12, "learning_rate": 1.9539540489579058e-05, "loss": 2.132, "step": 5034 }, { "epoch": 0.12, "learning_rate": 1.953930099753592e-05, "loss": 2.2666, "step": 5035 }, { "epoch": 0.12, "learning_rate": 1.9539061444695706e-05, "loss": 2.3665, "step": 5036 }, { "epoch": 0.12, "learning_rate": 1.9538821831059943e-05, "loss": 2.299, "step": 5037 }, { "epoch": 0.12, "learning_rate": 1.953858215663016e-05, "loss": 1.9798, "step": 5038 }, { "epoch": 0.12, "learning_rate": 1.9538342421407887e-05, "loss": 2.1742, "step": 5039 }, { "epoch": 0.12, "learning_rate": 1.9538102625394648e-05, "loss": 2.2672, "step": 5040 }, { "epoch": 0.12, "learning_rate": 1.9537862768591975e-05, "loss": 2.3087, "step": 5041 }, { "epoch": 0.12, "learning_rate": 1.953762285100139e-05, "loss": 2.5779, "step": 5042 }, { "epoch": 0.12, "learning_rate": 1.953738287262443e-05, "loss": 2.2583, "step": 5043 }, { "epoch": 0.12, "learning_rate": 1.953714283346262e-05, "loss": 2.3617, "step": 5044 }, { "epoch": 0.12, "learning_rate": 1.953690273351749e-05, "loss": 2.4893, "step": 5045 }, { "epoch": 0.12, "learning_rate": 1.953666257279057e-05, "loss": 2.1021, "step": 5046 }, { "epoch": 0.12, "learning_rate": 1.953642235128339e-05, "loss": 2.3744, "step": 5047 }, { "epoch": 0.12, "learning_rate": 1.9536182068997486e-05, "loss": 2.2632, "step": 5048 }, { "epoch": 0.12, "learning_rate": 1.9535941725934388e-05, "loss": 2.1682, "step": 5049 }, { "epoch": 0.12, "learning_rate": 1.9535701322095618e-05, "loss": 2.3253, "step": 5050 }, { "epoch": 0.12, "learning_rate": 1.9535460857482722e-05, "loss": 2.1676, "step": 5051 }, { "epoch": 0.12, "learning_rate": 1.9535220332097224e-05, "loss": 2.2251, "step": 5052 }, { "epoch": 0.12, "learning_rate": 1.953497974594066e-05, "loss": 2.3451, "step": 5053 }, { "epoch": 0.12, "learning_rate": 1.9534739099014565e-05, "loss": 2.3564, "step": 5054 }, { "epoch": 0.12, "learning_rate": 1.9534498391320465e-05, "loss": 2.3531, "step": 5055 }, { "epoch": 0.12, "learning_rate": 1.9534257622859904e-05, "loss": 2.1524, "step": 5056 }, { "epoch": 0.12, "learning_rate": 1.953401679363441e-05, "loss": 2.3815, "step": 5057 }, { "epoch": 0.12, "learning_rate": 1.9533775903645523e-05, "loss": 2.2425, "step": 5058 }, { "epoch": 0.12, "learning_rate": 1.9533534952894774e-05, "loss": 2.4012, "step": 5059 }, { "epoch": 0.12, "learning_rate": 1.95332939413837e-05, "loss": 2.269, "step": 5060 }, { "epoch": 0.12, "learning_rate": 1.9533052869113832e-05, "loss": 2.317, "step": 5061 }, { "epoch": 0.12, "learning_rate": 1.9532811736086716e-05, "loss": 2.3322, "step": 5062 }, { "epoch": 0.12, "learning_rate": 1.9532570542303884e-05, "loss": 2.344, "step": 5063 }, { "epoch": 0.12, "learning_rate": 1.9532329287766872e-05, "loss": 2.4986, "step": 5064 }, { "epoch": 0.12, "learning_rate": 1.953208797247722e-05, "loss": 2.1331, "step": 5065 }, { "epoch": 0.12, "learning_rate": 1.9531846596436465e-05, "loss": 2.4265, "step": 5066 }, { "epoch": 0.12, "learning_rate": 1.9531605159646142e-05, "loss": 2.2319, "step": 5067 }, { "epoch": 0.12, "learning_rate": 1.95313636621078e-05, "loss": 2.4713, "step": 5068 }, { "epoch": 0.12, "learning_rate": 1.9531122103822967e-05, "loss": 2.1696, "step": 5069 }, { "epoch": 0.12, "learning_rate": 1.9530880484793187e-05, "loss": 2.2538, "step": 5070 }, { "epoch": 0.12, "learning_rate": 1.9530638805019997e-05, "loss": 2.3072, "step": 5071 }, { "epoch": 0.13, "learning_rate": 1.953039706450494e-05, "loss": 2.236, "step": 5072 }, { "epoch": 0.13, "learning_rate": 1.953015526324956e-05, "loss": 2.172, "step": 5073 }, { "epoch": 0.13, "learning_rate": 1.9529913401255397e-05, "loss": 2.4264, "step": 5074 }, { "epoch": 0.13, "learning_rate": 1.9529671478523987e-05, "loss": 2.206, "step": 5075 }, { "epoch": 0.13, "learning_rate": 1.9529429495056872e-05, "loss": 2.4535, "step": 5076 }, { "epoch": 0.13, "learning_rate": 1.95291874508556e-05, "loss": 2.1917, "step": 5077 }, { "epoch": 0.13, "learning_rate": 1.952894534592171e-05, "loss": 2.3784, "step": 5078 }, { "epoch": 0.13, "learning_rate": 1.9528703180256753e-05, "loss": 2.2242, "step": 5079 }, { "epoch": 0.13, "learning_rate": 1.9528460953862255e-05, "loss": 2.2664, "step": 5080 }, { "epoch": 0.13, "learning_rate": 1.9528218666739775e-05, "loss": 2.3632, "step": 5081 }, { "epoch": 0.13, "learning_rate": 1.9527976318890853e-05, "loss": 2.3738, "step": 5082 }, { "epoch": 0.13, "learning_rate": 1.952773391031703e-05, "loss": 2.1564, "step": 5083 }, { "epoch": 0.13, "learning_rate": 1.9527491441019858e-05, "loss": 2.1536, "step": 5084 }, { "epoch": 0.13, "learning_rate": 1.9527248911000875e-05, "loss": 2.3092, "step": 5085 }, { "epoch": 0.13, "learning_rate": 1.952700632026163e-05, "loss": 2.2953, "step": 5086 }, { "epoch": 0.13, "learning_rate": 1.952676366880367e-05, "loss": 2.12, "step": 5087 }, { "epoch": 0.13, "learning_rate": 1.952652095662854e-05, "loss": 2.2282, "step": 5088 }, { "epoch": 0.13, "learning_rate": 1.9526278183737787e-05, "loss": 2.387, "step": 5089 }, { "epoch": 0.13, "learning_rate": 1.952603535013296e-05, "loss": 2.3478, "step": 5090 }, { "epoch": 0.13, "learning_rate": 1.9525792455815602e-05, "loss": 2.2596, "step": 5091 }, { "epoch": 0.13, "learning_rate": 1.9525549500787266e-05, "loss": 2.1748, "step": 5092 }, { "epoch": 0.13, "learning_rate": 1.9525306485049498e-05, "loss": 2.4118, "step": 5093 }, { "epoch": 0.13, "learning_rate": 1.952506340860385e-05, "loss": 2.4047, "step": 5094 }, { "epoch": 0.13, "learning_rate": 1.952482027145187e-05, "loss": 2.2685, "step": 5095 }, { "epoch": 0.13, "learning_rate": 1.95245770735951e-05, "loss": 2.3464, "step": 5096 }, { "epoch": 0.13, "learning_rate": 1.95243338150351e-05, "loss": 2.3538, "step": 5097 }, { "epoch": 0.13, "learning_rate": 1.9524090495773416e-05, "loss": 2.3059, "step": 5098 }, { "epoch": 0.13, "learning_rate": 1.9523847115811602e-05, "loss": 2.1221, "step": 5099 }, { "epoch": 0.13, "learning_rate": 1.9523603675151204e-05, "loss": 2.448, "step": 5100 }, { "epoch": 0.13, "learning_rate": 1.952336017379378e-05, "loss": 2.3252, "step": 5101 }, { "epoch": 0.13, "learning_rate": 1.9523116611740872e-05, "loss": 2.2656, "step": 5102 }, { "epoch": 0.13, "learning_rate": 1.9522872988994042e-05, "loss": 2.2451, "step": 5103 }, { "epoch": 0.13, "learning_rate": 1.9522629305554838e-05, "loss": 2.2383, "step": 5104 }, { "epoch": 0.13, "learning_rate": 1.9522385561424815e-05, "loss": 2.3246, "step": 5105 }, { "epoch": 0.13, "learning_rate": 1.9522141756605525e-05, "loss": 2.3408, "step": 5106 }, { "epoch": 0.13, "learning_rate": 1.9521897891098524e-05, "loss": 2.4182, "step": 5107 }, { "epoch": 0.13, "learning_rate": 1.9521653964905362e-05, "loss": 2.406, "step": 5108 }, { "epoch": 0.13, "learning_rate": 1.9521409978027596e-05, "loss": 2.4305, "step": 5109 }, { "epoch": 0.13, "learning_rate": 1.9521165930466783e-05, "loss": 2.3711, "step": 5110 }, { "epoch": 0.13, "learning_rate": 1.952092182222448e-05, "loss": 2.2576, "step": 5111 }, { "epoch": 0.13, "learning_rate": 1.9520677653302233e-05, "loss": 2.225, "step": 5112 }, { "epoch": 0.13, "learning_rate": 1.952043342370161e-05, "loss": 2.3367, "step": 5113 }, { "epoch": 0.13, "learning_rate": 1.952018913342416e-05, "loss": 2.1994, "step": 5114 }, { "epoch": 0.13, "learning_rate": 1.951994478247144e-05, "loss": 2.3468, "step": 5115 }, { "epoch": 0.13, "learning_rate": 1.951970037084501e-05, "loss": 2.2761, "step": 5116 }, { "epoch": 0.13, "learning_rate": 1.951945589854643e-05, "loss": 2.3742, "step": 5117 }, { "epoch": 0.13, "learning_rate": 1.9519211365577253e-05, "loss": 2.3834, "step": 5118 }, { "epoch": 0.13, "learning_rate": 1.951896677193904e-05, "loss": 2.2946, "step": 5119 }, { "epoch": 0.13, "learning_rate": 1.9518722117633347e-05, "loss": 2.3476, "step": 5120 }, { "epoch": 0.13, "learning_rate": 1.9518477402661743e-05, "loss": 2.4518, "step": 5121 }, { "epoch": 0.13, "learning_rate": 1.951823262702577e-05, "loss": 2.3419, "step": 5122 }, { "epoch": 0.13, "learning_rate": 1.9517987790727008e-05, "loss": 2.2607, "step": 5123 }, { "epoch": 0.13, "learning_rate": 1.9517742893767004e-05, "loss": 2.4756, "step": 5124 }, { "epoch": 0.13, "learning_rate": 1.9517497936147324e-05, "loss": 2.3025, "step": 5125 }, { "epoch": 0.13, "learning_rate": 1.951725291786953e-05, "loss": 2.3715, "step": 5126 }, { "epoch": 0.13, "learning_rate": 1.951700783893518e-05, "loss": 2.225, "step": 5127 }, { "epoch": 0.13, "learning_rate": 1.9516762699345836e-05, "loss": 2.2383, "step": 5128 }, { "epoch": 0.13, "learning_rate": 1.951651749910306e-05, "loss": 2.2734, "step": 5129 }, { "epoch": 0.13, "learning_rate": 1.951627223820842e-05, "loss": 2.3979, "step": 5130 }, { "epoch": 0.13, "learning_rate": 1.951602691666348e-05, "loss": 2.1244, "step": 5131 }, { "epoch": 0.13, "learning_rate": 1.951578153446979e-05, "loss": 2.2866, "step": 5132 }, { "epoch": 0.13, "learning_rate": 1.951553609162893e-05, "loss": 2.3155, "step": 5133 }, { "epoch": 0.13, "learning_rate": 1.9515290588142452e-05, "loss": 2.287, "step": 5134 }, { "epoch": 0.13, "learning_rate": 1.951504502401193e-05, "loss": 2.1916, "step": 5135 }, { "epoch": 0.13, "learning_rate": 1.9514799399238923e-05, "loss": 2.1767, "step": 5136 }, { "epoch": 0.13, "learning_rate": 1.9514553713825e-05, "loss": 2.3105, "step": 5137 }, { "epoch": 0.13, "learning_rate": 1.9514307967771727e-05, "loss": 2.1924, "step": 5138 }, { "epoch": 0.13, "learning_rate": 1.951406216108067e-05, "loss": 2.4712, "step": 5139 }, { "epoch": 0.13, "learning_rate": 1.9513816293753387e-05, "loss": 2.1009, "step": 5140 }, { "epoch": 0.13, "learning_rate": 1.951357036579146e-05, "loss": 2.3621, "step": 5141 }, { "epoch": 0.13, "learning_rate": 1.9513324377196444e-05, "loss": 2.2716, "step": 5142 }, { "epoch": 0.13, "learning_rate": 1.951307832796991e-05, "loss": 2.3141, "step": 5143 }, { "epoch": 0.13, "learning_rate": 1.951283221811343e-05, "loss": 2.2209, "step": 5144 }, { "epoch": 0.13, "learning_rate": 1.9512586047628573e-05, "loss": 2.3408, "step": 5145 }, { "epoch": 0.13, "learning_rate": 1.95123398165169e-05, "loss": 2.3607, "step": 5146 }, { "epoch": 0.13, "learning_rate": 1.951209352477999e-05, "loss": 2.2813, "step": 5147 }, { "epoch": 0.13, "learning_rate": 1.9511847172419408e-05, "loss": 2.2748, "step": 5148 }, { "epoch": 0.13, "learning_rate": 1.951160075943672e-05, "loss": 2.1749, "step": 5149 }, { "epoch": 0.13, "learning_rate": 1.9511354285833503e-05, "loss": 2.2985, "step": 5150 }, { "epoch": 0.13, "learning_rate": 1.9511107751611328e-05, "loss": 2.4595, "step": 5151 }, { "epoch": 0.13, "learning_rate": 1.9510861156771758e-05, "loss": 2.3254, "step": 5152 }, { "epoch": 0.13, "learning_rate": 1.9510614501316373e-05, "loss": 2.3379, "step": 5153 }, { "epoch": 0.13, "learning_rate": 1.9510367785246744e-05, "loss": 2.28, "step": 5154 }, { "epoch": 0.13, "learning_rate": 1.951012100856444e-05, "loss": 2.1605, "step": 5155 }, { "epoch": 0.13, "learning_rate": 1.9509874171271038e-05, "loss": 2.5227, "step": 5156 }, { "epoch": 0.13, "learning_rate": 1.9509627273368107e-05, "loss": 2.2159, "step": 5157 }, { "epoch": 0.13, "learning_rate": 1.950938031485722e-05, "loss": 2.2432, "step": 5158 }, { "epoch": 0.13, "learning_rate": 1.9509133295739954e-05, "loss": 2.376, "step": 5159 }, { "epoch": 0.13, "learning_rate": 1.9508886216017885e-05, "loss": 2.2832, "step": 5160 }, { "epoch": 0.13, "learning_rate": 1.9508639075692587e-05, "loss": 2.3352, "step": 5161 }, { "epoch": 0.13, "learning_rate": 1.9508391874765626e-05, "loss": 2.3451, "step": 5162 }, { "epoch": 0.13, "learning_rate": 1.950814461323859e-05, "loss": 2.3294, "step": 5163 }, { "epoch": 0.13, "learning_rate": 1.9507897291113052e-05, "loss": 2.3788, "step": 5164 }, { "epoch": 0.13, "learning_rate": 1.950764990839058e-05, "loss": 2.2254, "step": 5165 }, { "epoch": 0.13, "learning_rate": 1.950740246507276e-05, "loss": 2.4082, "step": 5166 }, { "epoch": 0.13, "learning_rate": 1.9507154961161166e-05, "loss": 2.3685, "step": 5167 }, { "epoch": 0.13, "learning_rate": 1.9506907396657374e-05, "loss": 1.9401, "step": 5168 }, { "epoch": 0.13, "learning_rate": 1.9506659771562964e-05, "loss": 2.1803, "step": 5169 }, { "epoch": 0.13, "learning_rate": 1.9506412085879515e-05, "loss": 2.3643, "step": 5170 }, { "epoch": 0.13, "learning_rate": 1.9506164339608598e-05, "loss": 2.2869, "step": 5171 }, { "epoch": 0.13, "learning_rate": 1.9505916532751804e-05, "loss": 2.2375, "step": 5172 }, { "epoch": 0.13, "learning_rate": 1.9505668665310703e-05, "loss": 2.2497, "step": 5173 }, { "epoch": 0.13, "learning_rate": 1.950542073728688e-05, "loss": 2.254, "step": 5174 }, { "epoch": 0.13, "learning_rate": 1.950517274868191e-05, "loss": 2.3319, "step": 5175 }, { "epoch": 0.13, "learning_rate": 1.9504924699497378e-05, "loss": 2.6058, "step": 5176 }, { "epoch": 0.13, "learning_rate": 1.9504676589734862e-05, "loss": 2.3214, "step": 5177 }, { "epoch": 0.13, "learning_rate": 1.9504428419395945e-05, "loss": 2.2255, "step": 5178 }, { "epoch": 0.13, "learning_rate": 1.950418018848221e-05, "loss": 2.4067, "step": 5179 }, { "epoch": 0.13, "learning_rate": 1.9503931896995233e-05, "loss": 2.2163, "step": 5180 }, { "epoch": 0.13, "learning_rate": 1.9503683544936605e-05, "loss": 2.2977, "step": 5181 }, { "epoch": 0.13, "learning_rate": 1.9503435132307904e-05, "loss": 2.2674, "step": 5182 }, { "epoch": 0.13, "learning_rate": 1.9503186659110714e-05, "loss": 2.2965, "step": 5183 }, { "epoch": 0.13, "learning_rate": 1.950293812534662e-05, "loss": 2.3179, "step": 5184 }, { "epoch": 0.13, "learning_rate": 1.9502689531017198e-05, "loss": 2.1861, "step": 5185 }, { "epoch": 0.13, "learning_rate": 1.9502440876124044e-05, "loss": 2.3992, "step": 5186 }, { "epoch": 0.13, "learning_rate": 1.9502192160668736e-05, "loss": 2.2293, "step": 5187 }, { "epoch": 0.13, "learning_rate": 1.9501943384652862e-05, "loss": 2.329, "step": 5188 }, { "epoch": 0.13, "learning_rate": 1.9501694548078e-05, "loss": 2.2813, "step": 5189 }, { "epoch": 0.13, "learning_rate": 1.950144565094575e-05, "loss": 2.2093, "step": 5190 }, { "epoch": 0.13, "learning_rate": 1.9501196693257688e-05, "loss": 2.3615, "step": 5191 }, { "epoch": 0.13, "learning_rate": 1.9500947675015406e-05, "loss": 2.2779, "step": 5192 }, { "epoch": 0.13, "learning_rate": 1.9500698596220483e-05, "loss": 2.3746, "step": 5193 }, { "epoch": 0.13, "learning_rate": 1.9500449456874513e-05, "loss": 2.1278, "step": 5194 }, { "epoch": 0.13, "learning_rate": 1.9500200256979086e-05, "loss": 2.4046, "step": 5195 }, { "epoch": 0.13, "learning_rate": 1.9499950996535783e-05, "loss": 2.3684, "step": 5196 }, { "epoch": 0.13, "learning_rate": 1.9499701675546196e-05, "loss": 2.2527, "step": 5197 }, { "epoch": 0.13, "learning_rate": 1.9499452294011916e-05, "loss": 2.4451, "step": 5198 }, { "epoch": 0.13, "learning_rate": 1.949920285193453e-05, "loss": 2.4339, "step": 5199 }, { "epoch": 0.13, "learning_rate": 1.949895334931563e-05, "loss": 2.1717, "step": 5200 }, { "epoch": 0.13, "learning_rate": 1.9498703786156806e-05, "loss": 2.4115, "step": 5201 }, { "epoch": 0.13, "learning_rate": 1.9498454162459643e-05, "loss": 2.1319, "step": 5202 }, { "epoch": 0.13, "learning_rate": 1.949820447822574e-05, "loss": 2.3541, "step": 5203 }, { "epoch": 0.13, "learning_rate": 1.949795473345668e-05, "loss": 2.3407, "step": 5204 }, { "epoch": 0.13, "learning_rate": 1.949770492815406e-05, "loss": 2.4941, "step": 5205 }, { "epoch": 0.13, "learning_rate": 1.9497455062319478e-05, "loss": 2.3067, "step": 5206 }, { "epoch": 0.13, "learning_rate": 1.9497205135954515e-05, "loss": 2.2536, "step": 5207 }, { "epoch": 0.13, "learning_rate": 1.9496955149060767e-05, "loss": 2.3679, "step": 5208 }, { "epoch": 0.13, "learning_rate": 1.949670510163983e-05, "loss": 2.3387, "step": 5209 }, { "epoch": 0.13, "learning_rate": 1.9496454993693293e-05, "loss": 2.0991, "step": 5210 }, { "epoch": 0.13, "learning_rate": 1.949620482522276e-05, "loss": 2.099, "step": 5211 }, { "epoch": 0.13, "learning_rate": 1.9495954596229817e-05, "loss": 2.3511, "step": 5212 }, { "epoch": 0.13, "learning_rate": 1.949570430671606e-05, "loss": 2.3742, "step": 5213 }, { "epoch": 0.13, "learning_rate": 1.9495453956683085e-05, "loss": 2.2047, "step": 5214 }, { "epoch": 0.13, "learning_rate": 1.9495203546132486e-05, "loss": 2.063, "step": 5215 }, { "epoch": 0.13, "learning_rate": 1.9494953075065863e-05, "loss": 2.2823, "step": 5216 }, { "epoch": 0.13, "learning_rate": 1.949470254348481e-05, "loss": 2.1949, "step": 5217 }, { "epoch": 0.13, "learning_rate": 1.949445195139092e-05, "loss": 2.2917, "step": 5218 }, { "epoch": 0.13, "learning_rate": 1.9494201298785797e-05, "loss": 2.3158, "step": 5219 }, { "epoch": 0.13, "learning_rate": 1.949395058567103e-05, "loss": 2.153, "step": 5220 }, { "epoch": 0.13, "learning_rate": 1.9493699812048227e-05, "loss": 2.2677, "step": 5221 }, { "epoch": 0.13, "learning_rate": 1.949344897791898e-05, "loss": 2.2216, "step": 5222 }, { "epoch": 0.13, "learning_rate": 1.9493198083284887e-05, "loss": 2.2611, "step": 5223 }, { "epoch": 0.13, "learning_rate": 1.949294712814755e-05, "loss": 2.3744, "step": 5224 }, { "epoch": 0.13, "learning_rate": 1.949269611250857e-05, "loss": 2.3273, "step": 5225 }, { "epoch": 0.13, "learning_rate": 1.949244503636954e-05, "loss": 2.3025, "step": 5226 }, { "epoch": 0.13, "learning_rate": 1.9492193899732065e-05, "loss": 2.3878, "step": 5227 }, { "epoch": 0.13, "learning_rate": 1.9491942702597747e-05, "loss": 2.3333, "step": 5228 }, { "epoch": 0.13, "learning_rate": 1.9491691444968183e-05, "loss": 2.3872, "step": 5229 }, { "epoch": 0.13, "learning_rate": 1.9491440126844977e-05, "loss": 2.2328, "step": 5230 }, { "epoch": 0.13, "learning_rate": 1.9491188748229733e-05, "loss": 2.1617, "step": 5231 }, { "epoch": 0.13, "learning_rate": 1.9490937309124045e-05, "loss": 2.281, "step": 5232 }, { "epoch": 0.13, "learning_rate": 1.9490685809529522e-05, "loss": 2.2806, "step": 5233 }, { "epoch": 0.13, "learning_rate": 1.9490434249447766e-05, "loss": 2.2962, "step": 5234 }, { "epoch": 0.13, "learning_rate": 1.949018262888038e-05, "loss": 2.3732, "step": 5235 }, { "epoch": 0.13, "learning_rate": 1.9489930947828967e-05, "loss": 2.0822, "step": 5236 }, { "epoch": 0.13, "learning_rate": 1.948967920629513e-05, "loss": 2.2951, "step": 5237 }, { "epoch": 0.13, "learning_rate": 1.9489427404280478e-05, "loss": 2.3474, "step": 5238 }, { "epoch": 0.13, "learning_rate": 1.9489175541786608e-05, "loss": 2.1827, "step": 5239 }, { "epoch": 0.13, "learning_rate": 1.948892361881513e-05, "loss": 2.303, "step": 5240 }, { "epoch": 0.13, "learning_rate": 1.9488671635367656e-05, "loss": 2.4746, "step": 5241 }, { "epoch": 0.13, "learning_rate": 1.948841959144578e-05, "loss": 2.3113, "step": 5242 }, { "epoch": 0.13, "learning_rate": 1.9488167487051115e-05, "loss": 2.2346, "step": 5243 }, { "epoch": 0.13, "learning_rate": 1.9487915322185266e-05, "loss": 2.1102, "step": 5244 }, { "epoch": 0.13, "learning_rate": 1.948766309684984e-05, "loss": 2.1833, "step": 5245 }, { "epoch": 0.13, "learning_rate": 1.9487410811046447e-05, "loss": 2.2549, "step": 5246 }, { "epoch": 0.13, "learning_rate": 1.9487158464776696e-05, "loss": 2.2552, "step": 5247 }, { "epoch": 0.13, "learning_rate": 1.9486906058042185e-05, "loss": 2.2323, "step": 5248 }, { "epoch": 0.13, "learning_rate": 1.9486653590844535e-05, "loss": 2.2983, "step": 5249 }, { "epoch": 0.13, "learning_rate": 1.9486401063185347e-05, "loss": 2.4709, "step": 5250 }, { "epoch": 0.13, "learning_rate": 1.948614847506624e-05, "loss": 2.445, "step": 5251 }, { "epoch": 0.13, "learning_rate": 1.948589582648881e-05, "loss": 2.3054, "step": 5252 }, { "epoch": 0.13, "learning_rate": 1.9485643117454677e-05, "loss": 2.1823, "step": 5253 }, { "epoch": 0.13, "learning_rate": 1.948539034796545e-05, "loss": 2.3965, "step": 5254 }, { "epoch": 0.13, "learning_rate": 1.9485137518022737e-05, "loss": 2.2155, "step": 5255 }, { "epoch": 0.13, "learning_rate": 1.948488462762815e-05, "loss": 2.1706, "step": 5256 }, { "epoch": 0.13, "learning_rate": 1.9484631676783307e-05, "loss": 2.3577, "step": 5257 }, { "epoch": 0.13, "learning_rate": 1.9484378665489815e-05, "loss": 2.2899, "step": 5258 }, { "epoch": 0.13, "learning_rate": 1.9484125593749283e-05, "loss": 2.2677, "step": 5259 }, { "epoch": 0.13, "learning_rate": 1.9483872461563332e-05, "loss": 2.3645, "step": 5260 }, { "epoch": 0.13, "learning_rate": 1.9483619268933567e-05, "loss": 2.2933, "step": 5261 }, { "epoch": 0.13, "learning_rate": 1.9483366015861608e-05, "loss": 2.2616, "step": 5262 }, { "epoch": 0.13, "learning_rate": 1.9483112702349067e-05, "loss": 2.2505, "step": 5263 }, { "epoch": 0.13, "learning_rate": 1.9482859328397558e-05, "loss": 2.2531, "step": 5264 }, { "epoch": 0.13, "learning_rate": 1.9482605894008696e-05, "loss": 2.2894, "step": 5265 }, { "epoch": 0.13, "learning_rate": 1.9482352399184097e-05, "loss": 2.23, "step": 5266 }, { "epoch": 0.13, "learning_rate": 1.9482098843925375e-05, "loss": 2.1422, "step": 5267 }, { "epoch": 0.13, "learning_rate": 1.948184522823415e-05, "loss": 2.3014, "step": 5268 }, { "epoch": 0.13, "learning_rate": 1.9481591552112032e-05, "loss": 2.3929, "step": 5269 }, { "epoch": 0.13, "learning_rate": 1.9481337815560642e-05, "loss": 2.2017, "step": 5270 }, { "epoch": 0.13, "learning_rate": 1.9481084018581594e-05, "loss": 2.0215, "step": 5271 }, { "epoch": 0.13, "learning_rate": 1.948083016117651e-05, "loss": 2.3532, "step": 5272 }, { "epoch": 0.13, "learning_rate": 1.9480576243347008e-05, "loss": 2.2424, "step": 5273 }, { "epoch": 0.13, "learning_rate": 1.94803222650947e-05, "loss": 2.3466, "step": 5274 }, { "epoch": 0.13, "learning_rate": 1.9480068226421213e-05, "loss": 2.1483, "step": 5275 }, { "epoch": 0.13, "learning_rate": 1.9479814127328155e-05, "loss": 2.2321, "step": 5276 }, { "epoch": 0.13, "learning_rate": 1.9479559967817156e-05, "loss": 2.0312, "step": 5277 }, { "epoch": 0.13, "learning_rate": 1.9479305747889833e-05, "loss": 2.3687, "step": 5278 }, { "epoch": 0.13, "learning_rate": 1.9479051467547805e-05, "loss": 2.2186, "step": 5279 }, { "epoch": 0.13, "learning_rate": 1.947879712679269e-05, "loss": 2.2446, "step": 5280 }, { "epoch": 0.13, "learning_rate": 1.9478542725626114e-05, "loss": 2.2749, "step": 5281 }, { "epoch": 0.13, "learning_rate": 1.9478288264049697e-05, "loss": 2.4013, "step": 5282 }, { "epoch": 0.13, "learning_rate": 1.947803374206506e-05, "loss": 2.3622, "step": 5283 }, { "epoch": 0.13, "learning_rate": 1.9477779159673823e-05, "loss": 2.3453, "step": 5284 }, { "epoch": 0.13, "learning_rate": 1.9477524516877614e-05, "loss": 2.1363, "step": 5285 }, { "epoch": 0.13, "learning_rate": 1.947726981367805e-05, "loss": 2.3061, "step": 5286 }, { "epoch": 0.13, "learning_rate": 1.947701505007676e-05, "loss": 2.4778, "step": 5287 }, { "epoch": 0.13, "learning_rate": 1.947676022607536e-05, "loss": 2.2066, "step": 5288 }, { "epoch": 0.13, "learning_rate": 1.947650534167548e-05, "loss": 2.2729, "step": 5289 }, { "epoch": 0.13, "learning_rate": 1.9476250396878743e-05, "loss": 2.3193, "step": 5290 }, { "epoch": 0.13, "learning_rate": 1.947599539168678e-05, "loss": 2.2615, "step": 5291 }, { "epoch": 0.13, "learning_rate": 1.9475740326101203e-05, "loss": 2.2237, "step": 5292 }, { "epoch": 0.13, "learning_rate": 1.947548520012365e-05, "loss": 2.3087, "step": 5293 }, { "epoch": 0.13, "learning_rate": 1.9475230013755735e-05, "loss": 2.423, "step": 5294 }, { "epoch": 0.13, "learning_rate": 1.9474974766999096e-05, "loss": 2.2244, "step": 5295 }, { "epoch": 0.13, "learning_rate": 1.9474719459855356e-05, "loss": 2.2741, "step": 5296 }, { "epoch": 0.13, "learning_rate": 1.947446409232614e-05, "loss": 2.3259, "step": 5297 }, { "epoch": 0.13, "learning_rate": 1.9474208664413073e-05, "loss": 2.3393, "step": 5298 }, { "epoch": 0.13, "learning_rate": 1.947395317611779e-05, "loss": 2.3542, "step": 5299 }, { "epoch": 0.13, "learning_rate": 1.9473697627441917e-05, "loss": 2.2631, "step": 5300 }, { "epoch": 0.13, "learning_rate": 1.947344201838708e-05, "loss": 2.1983, "step": 5301 }, { "epoch": 0.13, "learning_rate": 1.947318634895491e-05, "loss": 2.3964, "step": 5302 }, { "epoch": 0.13, "learning_rate": 1.9472930619147037e-05, "loss": 2.4197, "step": 5303 }, { "epoch": 0.13, "learning_rate": 1.9472674828965087e-05, "loss": 2.2909, "step": 5304 }, { "epoch": 0.13, "learning_rate": 1.9472418978410698e-05, "loss": 2.2041, "step": 5305 }, { "epoch": 0.13, "learning_rate": 1.9472163067485492e-05, "loss": 2.0531, "step": 5306 }, { "epoch": 0.13, "learning_rate": 1.9471907096191106e-05, "loss": 2.2425, "step": 5307 }, { "epoch": 0.13, "learning_rate": 1.947165106452917e-05, "loss": 2.2875, "step": 5308 }, { "epoch": 0.13, "learning_rate": 1.9471394972501313e-05, "loss": 2.0862, "step": 5309 }, { "epoch": 0.13, "learning_rate": 1.947113882010917e-05, "loss": 2.1737, "step": 5310 }, { "epoch": 0.13, "learning_rate": 1.947088260735437e-05, "loss": 2.3127, "step": 5311 }, { "epoch": 0.13, "learning_rate": 1.947062633423855e-05, "loss": 2.1499, "step": 5312 }, { "epoch": 0.13, "learning_rate": 1.9470370000763346e-05, "loss": 2.2171, "step": 5313 }, { "epoch": 0.13, "learning_rate": 1.9470113606930387e-05, "loss": 2.2765, "step": 5314 }, { "epoch": 0.13, "learning_rate": 1.94698571527413e-05, "loss": 2.3455, "step": 5315 }, { "epoch": 0.13, "learning_rate": 1.9469600638197737e-05, "loss": 2.0454, "step": 5316 }, { "epoch": 0.13, "learning_rate": 1.9469344063301317e-05, "loss": 2.2469, "step": 5317 }, { "epoch": 0.13, "learning_rate": 1.9469087428053684e-05, "loss": 2.2147, "step": 5318 }, { "epoch": 0.13, "learning_rate": 1.9468830732456472e-05, "loss": 2.2899, "step": 5319 }, { "epoch": 0.13, "learning_rate": 1.9468573976511313e-05, "loss": 2.3122, "step": 5320 }, { "epoch": 0.13, "learning_rate": 1.9468317160219847e-05, "loss": 2.1601, "step": 5321 }, { "epoch": 0.13, "learning_rate": 1.9468060283583708e-05, "loss": 2.3416, "step": 5322 }, { "epoch": 0.13, "learning_rate": 1.9467803346604542e-05, "loss": 2.3638, "step": 5323 }, { "epoch": 0.13, "learning_rate": 1.9467546349283975e-05, "loss": 2.3485, "step": 5324 }, { "epoch": 0.13, "learning_rate": 1.946728929162365e-05, "loss": 2.415, "step": 5325 }, { "epoch": 0.13, "learning_rate": 1.9467032173625206e-05, "loss": 2.2094, "step": 5326 }, { "epoch": 0.13, "learning_rate": 1.946677499529028e-05, "loss": 2.379, "step": 5327 }, { "epoch": 0.13, "learning_rate": 1.9466517756620512e-05, "loss": 2.2392, "step": 5328 }, { "epoch": 0.13, "learning_rate": 1.946626045761754e-05, "loss": 2.3502, "step": 5329 }, { "epoch": 0.13, "learning_rate": 1.9466003098283006e-05, "loss": 2.4427, "step": 5330 }, { "epoch": 0.13, "learning_rate": 1.946574567861855e-05, "loss": 2.4569, "step": 5331 }, { "epoch": 0.13, "learning_rate": 1.9465488198625813e-05, "loss": 2.2294, "step": 5332 }, { "epoch": 0.13, "learning_rate": 1.9465230658306434e-05, "loss": 2.1772, "step": 5333 }, { "epoch": 0.13, "learning_rate": 1.9464973057662053e-05, "loss": 2.3333, "step": 5334 }, { "epoch": 0.13, "learning_rate": 1.946471539669432e-05, "loss": 2.3247, "step": 5335 }, { "epoch": 0.13, "learning_rate": 1.9464457675404867e-05, "loss": 2.1561, "step": 5336 }, { "epoch": 0.13, "learning_rate": 1.946419989379534e-05, "loss": 2.2323, "step": 5337 }, { "epoch": 0.13, "learning_rate": 1.9463942051867385e-05, "loss": 2.2179, "step": 5338 }, { "epoch": 0.13, "learning_rate": 1.9463684149622642e-05, "loss": 2.4729, "step": 5339 }, { "epoch": 0.13, "learning_rate": 1.9463426187062756e-05, "loss": 2.2239, "step": 5340 }, { "epoch": 0.13, "learning_rate": 1.946316816418937e-05, "loss": 2.2857, "step": 5341 }, { "epoch": 0.13, "learning_rate": 1.946291008100413e-05, "loss": 2.2889, "step": 5342 }, { "epoch": 0.13, "learning_rate": 1.9462651937508682e-05, "loss": 2.1916, "step": 5343 }, { "epoch": 0.13, "learning_rate": 1.9462393733704665e-05, "loss": 2.1906, "step": 5344 }, { "epoch": 0.13, "learning_rate": 1.946213546959373e-05, "loss": 2.4085, "step": 5345 }, { "epoch": 0.13, "learning_rate": 1.946187714517753e-05, "loss": 2.3211, "step": 5346 }, { "epoch": 0.13, "learning_rate": 1.9461618760457696e-05, "loss": 2.3866, "step": 5347 }, { "epoch": 0.13, "learning_rate": 1.9461360315435883e-05, "loss": 2.238, "step": 5348 }, { "epoch": 0.13, "learning_rate": 1.9461101810113737e-05, "loss": 2.3158, "step": 5349 }, { "epoch": 0.13, "learning_rate": 1.9460843244492905e-05, "loss": 2.2805, "step": 5350 }, { "epoch": 0.13, "learning_rate": 1.946058461857504e-05, "loss": 2.2796, "step": 5351 }, { "epoch": 0.13, "learning_rate": 1.946032593236178e-05, "loss": 2.4489, "step": 5352 }, { "epoch": 0.13, "learning_rate": 1.9460067185854787e-05, "loss": 2.3054, "step": 5353 }, { "epoch": 0.13, "learning_rate": 1.94598083790557e-05, "loss": 2.3552, "step": 5354 }, { "epoch": 0.13, "learning_rate": 1.945954951196617e-05, "loss": 2.2873, "step": 5355 }, { "epoch": 0.13, "learning_rate": 1.945929058458785e-05, "loss": 2.382, "step": 5356 }, { "epoch": 0.13, "learning_rate": 1.9459031596922386e-05, "loss": 2.1313, "step": 5357 }, { "epoch": 0.13, "learning_rate": 1.9458772548971436e-05, "loss": 2.2214, "step": 5358 }, { "epoch": 0.13, "learning_rate": 1.945851344073664e-05, "loss": 2.4111, "step": 5359 }, { "epoch": 0.13, "learning_rate": 1.9458254272219656e-05, "loss": 2.2696, "step": 5360 }, { "epoch": 0.13, "learning_rate": 1.945799504342214e-05, "loss": 2.1622, "step": 5361 }, { "epoch": 0.13, "learning_rate": 1.9457735754345734e-05, "loss": 2.3123, "step": 5362 }, { "epoch": 0.13, "learning_rate": 1.9457476404992096e-05, "loss": 2.2058, "step": 5363 }, { "epoch": 0.13, "learning_rate": 1.9457216995362885e-05, "loss": 2.3247, "step": 5364 }, { "epoch": 0.13, "learning_rate": 1.9456957525459744e-05, "loss": 2.3022, "step": 5365 }, { "epoch": 0.13, "learning_rate": 1.945669799528433e-05, "loss": 2.4055, "step": 5366 }, { "epoch": 0.13, "learning_rate": 1.94564384048383e-05, "loss": 2.3826, "step": 5367 }, { "epoch": 0.13, "learning_rate": 1.9456178754123304e-05, "loss": 2.3415, "step": 5368 }, { "epoch": 0.13, "learning_rate": 1.9455919043140997e-05, "loss": 2.2152, "step": 5369 }, { "epoch": 0.13, "learning_rate": 1.9455659271893038e-05, "loss": 2.1201, "step": 5370 }, { "epoch": 0.13, "learning_rate": 1.9455399440381083e-05, "loss": 2.3092, "step": 5371 }, { "epoch": 0.13, "learning_rate": 1.9455139548606787e-05, "loss": 2.4002, "step": 5372 }, { "epoch": 0.13, "learning_rate": 1.9454879596571805e-05, "loss": 2.2201, "step": 5373 }, { "epoch": 0.13, "learning_rate": 1.945461958427779e-05, "loss": 2.2558, "step": 5374 }, { "epoch": 0.13, "learning_rate": 1.9454359511726406e-05, "loss": 2.2753, "step": 5375 }, { "epoch": 0.13, "learning_rate": 1.945409937891931e-05, "loss": 2.2929, "step": 5376 }, { "epoch": 0.13, "learning_rate": 1.945383918585815e-05, "loss": 2.1012, "step": 5377 }, { "epoch": 0.13, "learning_rate": 1.9453578932544602e-05, "loss": 2.1944, "step": 5378 }, { "epoch": 0.13, "learning_rate": 1.945331861898031e-05, "loss": 2.3998, "step": 5379 }, { "epoch": 0.13, "learning_rate": 1.9453058245166934e-05, "loss": 2.2586, "step": 5380 }, { "epoch": 0.13, "learning_rate": 1.9452797811106143e-05, "loss": 2.1728, "step": 5381 }, { "epoch": 0.13, "learning_rate": 1.9452537316799587e-05, "loss": 2.261, "step": 5382 }, { "epoch": 0.13, "learning_rate": 1.945227676224893e-05, "loss": 2.1579, "step": 5383 }, { "epoch": 0.13, "learning_rate": 1.945201614745584e-05, "loss": 2.2728, "step": 5384 }, { "epoch": 0.13, "learning_rate": 1.9451755472421963e-05, "loss": 2.3087, "step": 5385 }, { "epoch": 0.13, "learning_rate": 1.945149473714897e-05, "loss": 2.3194, "step": 5386 }, { "epoch": 0.13, "learning_rate": 1.9451233941638523e-05, "loss": 2.2607, "step": 5387 }, { "epoch": 0.13, "learning_rate": 1.945097308589228e-05, "loss": 2.277, "step": 5388 }, { "epoch": 0.13, "learning_rate": 1.9450712169911905e-05, "loss": 2.2825, "step": 5389 }, { "epoch": 0.13, "learning_rate": 1.945045119369906e-05, "loss": 2.2178, "step": 5390 }, { "epoch": 0.13, "learning_rate": 1.9450190157255412e-05, "loss": 2.2478, "step": 5391 }, { "epoch": 0.13, "learning_rate": 1.9449929060582623e-05, "loss": 2.3923, "step": 5392 }, { "epoch": 0.13, "learning_rate": 1.944966790368236e-05, "loss": 2.1336, "step": 5393 }, { "epoch": 0.13, "learning_rate": 1.9449406686556277e-05, "loss": 2.2595, "step": 5394 }, { "epoch": 0.13, "learning_rate": 1.9449145409206048e-05, "loss": 2.2911, "step": 5395 }, { "epoch": 0.13, "learning_rate": 1.9448884071633335e-05, "loss": 2.1934, "step": 5396 }, { "epoch": 0.13, "learning_rate": 1.9448622673839808e-05, "loss": 2.2265, "step": 5397 }, { "epoch": 0.13, "learning_rate": 1.9448361215827127e-05, "loss": 2.359, "step": 5398 }, { "epoch": 0.13, "learning_rate": 1.944809969759696e-05, "loss": 2.2089, "step": 5399 }, { "epoch": 0.13, "learning_rate": 1.9447838119150977e-05, "loss": 2.2652, "step": 5400 }, { "epoch": 0.13, "learning_rate": 1.944757648049084e-05, "loss": 2.1, "step": 5401 }, { "epoch": 0.13, "learning_rate": 1.9447314781618223e-05, "loss": 2.1526, "step": 5402 }, { "epoch": 0.13, "learning_rate": 1.9447053022534787e-05, "loss": 2.3554, "step": 5403 }, { "epoch": 0.13, "learning_rate": 1.9446791203242205e-05, "loss": 2.2876, "step": 5404 }, { "epoch": 0.13, "learning_rate": 1.9446529323742146e-05, "loss": 2.2184, "step": 5405 }, { "epoch": 0.13, "learning_rate": 1.944626738403627e-05, "loss": 2.2201, "step": 5406 }, { "epoch": 0.13, "learning_rate": 1.944600538412626e-05, "loss": 2.2983, "step": 5407 }, { "epoch": 0.13, "learning_rate": 1.944574332401378e-05, "loss": 2.1636, "step": 5408 }, { "epoch": 0.13, "learning_rate": 1.9445481203700494e-05, "loss": 2.3261, "step": 5409 }, { "epoch": 0.13, "learning_rate": 1.944521902318808e-05, "loss": 2.4225, "step": 5410 }, { "epoch": 0.13, "learning_rate": 1.944495678247821e-05, "loss": 2.5502, "step": 5411 }, { "epoch": 0.13, "learning_rate": 1.944469448157255e-05, "loss": 2.2236, "step": 5412 }, { "epoch": 0.13, "learning_rate": 1.9444432120472775e-05, "loss": 2.4445, "step": 5413 }, { "epoch": 0.13, "learning_rate": 1.9444169699180552e-05, "loss": 2.3528, "step": 5414 }, { "epoch": 0.13, "learning_rate": 1.9443907217697565e-05, "loss": 2.2806, "step": 5415 }, { "epoch": 0.13, "learning_rate": 1.9443644676025474e-05, "loss": 2.5363, "step": 5416 }, { "epoch": 0.13, "learning_rate": 1.944338207416596e-05, "loss": 2.1367, "step": 5417 }, { "epoch": 0.13, "learning_rate": 1.9443119412120695e-05, "loss": 2.3245, "step": 5418 }, { "epoch": 0.13, "learning_rate": 1.944285668989135e-05, "loss": 2.1147, "step": 5419 }, { "epoch": 0.13, "learning_rate": 1.9442593907479607e-05, "loss": 2.2228, "step": 5420 }, { "epoch": 0.13, "learning_rate": 1.9442331064887132e-05, "loss": 2.1798, "step": 5421 }, { "epoch": 0.13, "learning_rate": 1.9442068162115605e-05, "loss": 2.3911, "step": 5422 }, { "epoch": 0.13, "learning_rate": 1.94418051991667e-05, "loss": 2.3482, "step": 5423 }, { "epoch": 0.13, "learning_rate": 1.944154217604209e-05, "loss": 2.2597, "step": 5424 }, { "epoch": 0.13, "learning_rate": 1.9441279092743462e-05, "loss": 2.3117, "step": 5425 }, { "epoch": 0.13, "learning_rate": 1.944101594927248e-05, "loss": 2.2062, "step": 5426 }, { "epoch": 0.13, "learning_rate": 1.944075274563083e-05, "loss": 2.3285, "step": 5427 }, { "epoch": 0.13, "learning_rate": 1.9440489481820185e-05, "loss": 2.466, "step": 5428 }, { "epoch": 0.13, "learning_rate": 1.9440226157842223e-05, "loss": 2.2955, "step": 5429 }, { "epoch": 0.13, "learning_rate": 1.9439962773698623e-05, "loss": 2.2279, "step": 5430 }, { "epoch": 0.13, "learning_rate": 1.9439699329391066e-05, "loss": 2.1398, "step": 5431 }, { "epoch": 0.13, "learning_rate": 1.9439435824921228e-05, "loss": 2.326, "step": 5432 }, { "epoch": 0.13, "learning_rate": 1.9439172260290787e-05, "loss": 2.2961, "step": 5433 }, { "epoch": 0.13, "learning_rate": 1.9438908635501427e-05, "loss": 2.3533, "step": 5434 }, { "epoch": 0.13, "learning_rate": 1.9438644950554828e-05, "loss": 2.2901, "step": 5435 }, { "epoch": 0.13, "learning_rate": 1.9438381205452668e-05, "loss": 2.2829, "step": 5436 }, { "epoch": 0.13, "learning_rate": 1.943811740019663e-05, "loss": 2.3371, "step": 5437 }, { "epoch": 0.13, "learning_rate": 1.943785353478839e-05, "loss": 2.3524, "step": 5438 }, { "epoch": 0.13, "learning_rate": 1.9437589609229635e-05, "loss": 2.556, "step": 5439 }, { "epoch": 0.13, "learning_rate": 1.9437325623522043e-05, "loss": 2.1724, "step": 5440 }, { "epoch": 0.13, "learning_rate": 1.9437061577667302e-05, "loss": 2.3476, "step": 5441 }, { "epoch": 0.13, "learning_rate": 1.9436797471667093e-05, "loss": 2.2421, "step": 5442 }, { "epoch": 0.13, "learning_rate": 1.94365333055231e-05, "loss": 2.3833, "step": 5443 }, { "epoch": 0.13, "learning_rate": 1.9436269079237002e-05, "loss": 2.2633, "step": 5444 }, { "epoch": 0.13, "learning_rate": 1.9436004792810488e-05, "loss": 2.2826, "step": 5445 }, { "epoch": 0.13, "learning_rate": 1.9435740446245238e-05, "loss": 2.2927, "step": 5446 }, { "epoch": 0.13, "learning_rate": 1.943547603954294e-05, "loss": 2.2942, "step": 5447 }, { "epoch": 0.13, "learning_rate": 1.9435211572705278e-05, "loss": 2.3418, "step": 5448 }, { "epoch": 0.13, "learning_rate": 1.943494704573394e-05, "loss": 2.2569, "step": 5449 }, { "epoch": 0.13, "learning_rate": 1.9434682458630607e-05, "loss": 2.2596, "step": 5450 }, { "epoch": 0.13, "learning_rate": 1.943441781139697e-05, "loss": 2.2925, "step": 5451 }, { "epoch": 0.13, "learning_rate": 1.9434153104034713e-05, "loss": 2.1939, "step": 5452 }, { "epoch": 0.13, "learning_rate": 1.9433888336545524e-05, "loss": 2.3497, "step": 5453 }, { "epoch": 0.13, "learning_rate": 1.943362350893109e-05, "loss": 2.1916, "step": 5454 }, { "epoch": 0.13, "learning_rate": 1.9433358621193097e-05, "loss": 2.1402, "step": 5455 }, { "epoch": 0.13, "learning_rate": 1.9433093673333235e-05, "loss": 2.438, "step": 5456 }, { "epoch": 0.13, "learning_rate": 1.9432828665353197e-05, "loss": 2.2477, "step": 5457 }, { "epoch": 0.13, "learning_rate": 1.9432563597254664e-05, "loss": 2.361, "step": 5458 }, { "epoch": 0.13, "learning_rate": 1.9432298469039336e-05, "loss": 2.3056, "step": 5459 }, { "epoch": 0.13, "learning_rate": 1.9432033280708887e-05, "loss": 2.2852, "step": 5460 }, { "epoch": 0.13, "learning_rate": 1.943176803226502e-05, "loss": 2.2038, "step": 5461 }, { "epoch": 0.13, "learning_rate": 1.943150272370942e-05, "loss": 2.2495, "step": 5462 }, { "epoch": 0.13, "learning_rate": 1.9431237355043783e-05, "loss": 2.261, "step": 5463 }, { "epoch": 0.13, "learning_rate": 1.9430971926269798e-05, "loss": 2.2362, "step": 5464 }, { "epoch": 0.13, "learning_rate": 1.943070643738915e-05, "loss": 2.4153, "step": 5465 }, { "epoch": 0.13, "learning_rate": 1.943044088840354e-05, "loss": 2.2229, "step": 5466 }, { "epoch": 0.13, "learning_rate": 1.9430175279314655e-05, "loss": 1.9972, "step": 5467 }, { "epoch": 0.13, "learning_rate": 1.942990961012419e-05, "loss": 2.2974, "step": 5468 }, { "epoch": 0.13, "learning_rate": 1.9429643880833836e-05, "loss": 2.3627, "step": 5469 }, { "epoch": 0.13, "learning_rate": 1.9429378091445295e-05, "loss": 2.3116, "step": 5470 }, { "epoch": 0.13, "learning_rate": 1.942911224196025e-05, "loss": 2.2848, "step": 5471 }, { "epoch": 0.13, "learning_rate": 1.94288463323804e-05, "loss": 2.2506, "step": 5472 }, { "epoch": 0.13, "learning_rate": 1.9428580362707437e-05, "loss": 2.2608, "step": 5473 }, { "epoch": 0.13, "learning_rate": 1.9428314332943064e-05, "loss": 2.3333, "step": 5474 }, { "epoch": 0.13, "learning_rate": 1.9428048243088968e-05, "loss": 2.1886, "step": 5475 }, { "epoch": 0.13, "learning_rate": 1.942778209314685e-05, "loss": 2.2326, "step": 5476 }, { "epoch": 0.14, "learning_rate": 1.9427515883118402e-05, "loss": 2.114, "step": 5477 }, { "epoch": 0.14, "learning_rate": 1.9427249613005323e-05, "loss": 2.5535, "step": 5478 }, { "epoch": 0.14, "learning_rate": 1.942698328280931e-05, "loss": 2.3098, "step": 5479 }, { "epoch": 0.14, "learning_rate": 1.9426716892532063e-05, "loss": 2.4625, "step": 5480 }, { "epoch": 0.14, "learning_rate": 1.9426450442175276e-05, "loss": 2.3405, "step": 5481 }, { "epoch": 0.14, "learning_rate": 1.9426183931740647e-05, "loss": 2.3533, "step": 5482 }, { "epoch": 0.14, "learning_rate": 1.942591736122988e-05, "loss": 2.2952, "step": 5483 }, { "epoch": 0.14, "learning_rate": 1.9425650730644663e-05, "loss": 2.2482, "step": 5484 }, { "epoch": 0.14, "learning_rate": 1.9425384039986706e-05, "loss": 2.3277, "step": 5485 }, { "epoch": 0.14, "learning_rate": 1.9425117289257704e-05, "loss": 2.2226, "step": 5486 }, { "epoch": 0.14, "learning_rate": 1.9424850478459358e-05, "loss": 2.3698, "step": 5487 }, { "epoch": 0.14, "learning_rate": 1.942458360759337e-05, "loss": 2.3099, "step": 5488 }, { "epoch": 0.14, "learning_rate": 1.942431667666144e-05, "loss": 2.1461, "step": 5489 }, { "epoch": 0.14, "learning_rate": 1.9424049685665266e-05, "loss": 2.3707, "step": 5490 }, { "epoch": 0.14, "learning_rate": 1.9423782634606553e-05, "loss": 2.1898, "step": 5491 }, { "epoch": 0.14, "learning_rate": 1.9423515523487e-05, "loss": 2.3049, "step": 5492 }, { "epoch": 0.14, "learning_rate": 1.9423248352308316e-05, "loss": 2.3617, "step": 5493 }, { "epoch": 0.14, "learning_rate": 1.9422981121072194e-05, "loss": 2.2801, "step": 5494 }, { "epoch": 0.14, "learning_rate": 1.9422713829780345e-05, "loss": 2.3088, "step": 5495 }, { "epoch": 0.14, "learning_rate": 1.9422446478434473e-05, "loss": 2.2587, "step": 5496 }, { "epoch": 0.14, "learning_rate": 1.9422179067036276e-05, "loss": 2.2542, "step": 5497 }, { "epoch": 0.14, "learning_rate": 1.942191159558746e-05, "loss": 2.3315, "step": 5498 }, { "epoch": 0.14, "learning_rate": 1.9421644064089732e-05, "loss": 2.2851, "step": 5499 }, { "epoch": 0.14, "learning_rate": 1.94213764725448e-05, "loss": 2.4719, "step": 5500 }, { "epoch": 0.14, "learning_rate": 1.942110882095436e-05, "loss": 2.141, "step": 5501 }, { "epoch": 0.14, "learning_rate": 1.9420841109320126e-05, "loss": 2.2735, "step": 5502 }, { "epoch": 0.14, "learning_rate": 1.94205733376438e-05, "loss": 2.4288, "step": 5503 }, { "epoch": 0.14, "learning_rate": 1.942030550592709e-05, "loss": 2.2399, "step": 5504 }, { "epoch": 0.14, "learning_rate": 1.94200376141717e-05, "loss": 2.3266, "step": 5505 }, { "epoch": 0.14, "learning_rate": 1.9419769662379346e-05, "loss": 1.989, "step": 5506 }, { "epoch": 0.14, "learning_rate": 1.9419501650551728e-05, "loss": 2.2895, "step": 5507 }, { "epoch": 0.14, "learning_rate": 1.9419233578690554e-05, "loss": 2.3672, "step": 5508 }, { "epoch": 0.14, "learning_rate": 1.9418965446797536e-05, "loss": 2.4199, "step": 5509 }, { "epoch": 0.14, "learning_rate": 1.9418697254874385e-05, "loss": 2.2989, "step": 5510 }, { "epoch": 0.14, "learning_rate": 1.9418429002922805e-05, "loss": 2.3259, "step": 5511 }, { "epoch": 0.14, "learning_rate": 1.9418160690944508e-05, "loss": 2.327, "step": 5512 }, { "epoch": 0.14, "learning_rate": 1.94178923189412e-05, "loss": 2.2314, "step": 5513 }, { "epoch": 0.14, "learning_rate": 1.9417623886914597e-05, "loss": 2.2239, "step": 5514 }, { "epoch": 0.14, "learning_rate": 1.9417355394866413e-05, "loss": 2.3965, "step": 5515 }, { "epoch": 0.14, "learning_rate": 1.941708684279835e-05, "loss": 2.3181, "step": 5516 }, { "epoch": 0.14, "learning_rate": 1.9416818230712123e-05, "loss": 2.4217, "step": 5517 }, { "epoch": 0.14, "learning_rate": 1.9416549558609444e-05, "loss": 2.3986, "step": 5518 }, { "epoch": 0.14, "learning_rate": 1.9416280826492023e-05, "loss": 2.3487, "step": 5519 }, { "epoch": 0.14, "learning_rate": 1.9416012034361583e-05, "loss": 2.4629, "step": 5520 }, { "epoch": 0.14, "learning_rate": 1.9415743182219825e-05, "loss": 2.2944, "step": 5521 }, { "epoch": 0.14, "learning_rate": 1.941547427006847e-05, "loss": 2.1152, "step": 5522 }, { "epoch": 0.14, "learning_rate": 1.9415205297909226e-05, "loss": 2.2672, "step": 5523 }, { "epoch": 0.14, "learning_rate": 1.9414936265743812e-05, "loss": 2.1711, "step": 5524 }, { "epoch": 0.14, "learning_rate": 1.941466717357394e-05, "loss": 2.1346, "step": 5525 }, { "epoch": 0.14, "learning_rate": 1.9414398021401327e-05, "loss": 2.1607, "step": 5526 }, { "epoch": 0.14, "learning_rate": 1.941412880922769e-05, "loss": 2.3718, "step": 5527 }, { "epoch": 0.14, "learning_rate": 1.9413859537054737e-05, "loss": 2.2626, "step": 5528 }, { "epoch": 0.14, "learning_rate": 1.941359020488419e-05, "loss": 2.3637, "step": 5529 }, { "epoch": 0.14, "learning_rate": 1.9413320812717766e-05, "loss": 2.1937, "step": 5530 }, { "epoch": 0.14, "learning_rate": 1.9413051360557184e-05, "loss": 2.3229, "step": 5531 }, { "epoch": 0.14, "learning_rate": 1.941278184840415e-05, "loss": 2.2577, "step": 5532 }, { "epoch": 0.14, "learning_rate": 1.94125122762604e-05, "loss": 2.2692, "step": 5533 }, { "epoch": 0.14, "learning_rate": 1.9412242644127633e-05, "loss": 2.4457, "step": 5534 }, { "epoch": 0.14, "learning_rate": 1.941197295200758e-05, "loss": 2.2791, "step": 5535 }, { "epoch": 0.14, "learning_rate": 1.9411703199901955e-05, "loss": 2.2217, "step": 5536 }, { "epoch": 0.14, "learning_rate": 1.941143338781248e-05, "loss": 2.1099, "step": 5537 }, { "epoch": 0.14, "learning_rate": 1.9411163515740872e-05, "loss": 2.2585, "step": 5538 }, { "epoch": 0.14, "learning_rate": 1.9410893583688854e-05, "loss": 2.3523, "step": 5539 }, { "epoch": 0.14, "learning_rate": 1.941062359165814e-05, "loss": 2.2774, "step": 5540 }, { "epoch": 0.14, "learning_rate": 1.941035353965046e-05, "loss": 2.1781, "step": 5541 }, { "epoch": 0.14, "learning_rate": 1.9410083427667524e-05, "loss": 2.3418, "step": 5542 }, { "epoch": 0.14, "learning_rate": 1.9409813255711065e-05, "loss": 2.3136, "step": 5543 }, { "epoch": 0.14, "learning_rate": 1.9409543023782797e-05, "loss": 2.1514, "step": 5544 }, { "epoch": 0.14, "learning_rate": 1.9409272731884447e-05, "loss": 2.3152, "step": 5545 }, { "epoch": 0.14, "learning_rate": 1.9409002380017732e-05, "loss": 2.2893, "step": 5546 }, { "epoch": 0.14, "learning_rate": 1.940873196818438e-05, "loss": 2.3795, "step": 5547 }, { "epoch": 0.14, "learning_rate": 1.9408461496386114e-05, "loss": 2.4098, "step": 5548 }, { "epoch": 0.14, "learning_rate": 1.9408190964624655e-05, "loss": 2.296, "step": 5549 }, { "epoch": 0.14, "learning_rate": 1.940792037290173e-05, "loss": 2.3524, "step": 5550 }, { "epoch": 0.14, "learning_rate": 1.9407649721219065e-05, "loss": 2.3714, "step": 5551 }, { "epoch": 0.14, "learning_rate": 1.9407379009578376e-05, "loss": 2.1711, "step": 5552 }, { "epoch": 0.14, "learning_rate": 1.9407108237981402e-05, "loss": 2.1831, "step": 5553 }, { "epoch": 0.14, "learning_rate": 1.9406837406429857e-05, "loss": 2.2595, "step": 5554 }, { "epoch": 0.14, "learning_rate": 1.9406566514925473e-05, "loss": 2.2409, "step": 5555 }, { "epoch": 0.14, "learning_rate": 1.9406295563469973e-05, "loss": 2.3002, "step": 5556 }, { "epoch": 0.14, "learning_rate": 1.9406024552065093e-05, "loss": 2.1433, "step": 5557 }, { "epoch": 0.14, "learning_rate": 1.9405753480712547e-05, "loss": 2.303, "step": 5558 }, { "epoch": 0.14, "learning_rate": 1.940548234941407e-05, "loss": 2.3296, "step": 5559 }, { "epoch": 0.14, "learning_rate": 1.9405211158171387e-05, "loss": 2.2528, "step": 5560 }, { "epoch": 0.14, "learning_rate": 1.9404939906986232e-05, "loss": 2.3048, "step": 5561 }, { "epoch": 0.14, "learning_rate": 1.940466859586033e-05, "loss": 2.2638, "step": 5562 }, { "epoch": 0.14, "learning_rate": 1.9404397224795407e-05, "loss": 2.4968, "step": 5563 }, { "epoch": 0.14, "learning_rate": 1.9404125793793197e-05, "loss": 2.3229, "step": 5564 }, { "epoch": 0.14, "learning_rate": 1.9403854302855427e-05, "loss": 2.2428, "step": 5565 }, { "epoch": 0.14, "learning_rate": 1.9403582751983832e-05, "loss": 2.3138, "step": 5566 }, { "epoch": 0.14, "learning_rate": 1.9403311141180136e-05, "loss": 2.217, "step": 5567 }, { "epoch": 0.14, "learning_rate": 1.9403039470446077e-05, "loss": 2.2815, "step": 5568 }, { "epoch": 0.14, "learning_rate": 1.940276773978338e-05, "loss": 2.469, "step": 5569 }, { "epoch": 0.14, "learning_rate": 1.9402495949193783e-05, "loss": 2.353, "step": 5570 }, { "epoch": 0.14, "learning_rate": 1.9402224098679014e-05, "loss": 2.2407, "step": 5571 }, { "epoch": 0.14, "learning_rate": 1.9401952188240804e-05, "loss": 2.2404, "step": 5572 }, { "epoch": 0.14, "learning_rate": 1.9401680217880895e-05, "loss": 2.2471, "step": 5573 }, { "epoch": 0.14, "learning_rate": 1.940140818760101e-05, "loss": 2.2507, "step": 5574 }, { "epoch": 0.14, "learning_rate": 1.9401136097402885e-05, "loss": 2.4302, "step": 5575 }, { "epoch": 0.14, "learning_rate": 1.9400863947288256e-05, "loss": 2.2086, "step": 5576 }, { "epoch": 0.14, "learning_rate": 1.940059173725886e-05, "loss": 2.3203, "step": 5577 }, { "epoch": 0.14, "learning_rate": 1.9400319467316432e-05, "loss": 2.5296, "step": 5578 }, { "epoch": 0.14, "learning_rate": 1.94000471374627e-05, "loss": 2.3762, "step": 5579 }, { "epoch": 0.14, "learning_rate": 1.9399774747699407e-05, "loss": 2.2741, "step": 5580 }, { "epoch": 0.14, "learning_rate": 1.9399502298028283e-05, "loss": 2.158, "step": 5581 }, { "epoch": 0.14, "learning_rate": 1.939922978845107e-05, "loss": 2.3044, "step": 5582 }, { "epoch": 0.14, "learning_rate": 1.93989572189695e-05, "loss": 2.1502, "step": 5583 }, { "epoch": 0.14, "learning_rate": 1.9398684589585314e-05, "loss": 2.3227, "step": 5584 }, { "epoch": 0.14, "learning_rate": 1.939841190030025e-05, "loss": 2.3417, "step": 5585 }, { "epoch": 0.14, "learning_rate": 1.939813915111604e-05, "loss": 2.3885, "step": 5586 }, { "epoch": 0.14, "learning_rate": 1.939786634203443e-05, "loss": 2.3433, "step": 5587 }, { "epoch": 0.14, "learning_rate": 1.9397593473057153e-05, "loss": 2.375, "step": 5588 }, { "epoch": 0.14, "learning_rate": 1.9397320544185952e-05, "loss": 2.37, "step": 5589 }, { "epoch": 0.14, "learning_rate": 1.939704755542257e-05, "loss": 2.3449, "step": 5590 }, { "epoch": 0.14, "learning_rate": 1.9396774506768732e-05, "loss": 2.2842, "step": 5591 }, { "epoch": 0.14, "learning_rate": 1.9396501398226193e-05, "loss": 2.2711, "step": 5592 }, { "epoch": 0.14, "learning_rate": 1.9396228229796687e-05, "loss": 2.2956, "step": 5593 }, { "epoch": 0.14, "learning_rate": 1.9395955001481954e-05, "loss": 2.1401, "step": 5594 }, { "epoch": 0.14, "learning_rate": 1.9395681713283737e-05, "loss": 2.3657, "step": 5595 }, { "epoch": 0.14, "learning_rate": 1.9395408365203784e-05, "loss": 2.2839, "step": 5596 }, { "epoch": 0.14, "learning_rate": 1.939513495724383e-05, "loss": 2.2779, "step": 5597 }, { "epoch": 0.14, "learning_rate": 1.9394861489405618e-05, "loss": 2.2454, "step": 5598 }, { "epoch": 0.14, "learning_rate": 1.939458796169089e-05, "loss": 2.3368, "step": 5599 }, { "epoch": 0.14, "learning_rate": 1.939431437410139e-05, "loss": 2.1458, "step": 5600 }, { "epoch": 0.14, "learning_rate": 1.9394040726638873e-05, "loss": 2.1517, "step": 5601 }, { "epoch": 0.14, "learning_rate": 1.9393767019305067e-05, "loss": 2.3067, "step": 5602 }, { "epoch": 0.14, "learning_rate": 1.939349325210172e-05, "loss": 2.4014, "step": 5603 }, { "epoch": 0.14, "learning_rate": 1.939321942503058e-05, "loss": 2.1907, "step": 5604 }, { "epoch": 0.14, "learning_rate": 1.9392945538093392e-05, "loss": 2.4238, "step": 5605 }, { "epoch": 0.14, "learning_rate": 1.93926715912919e-05, "loss": 2.4492, "step": 5606 }, { "epoch": 0.14, "learning_rate": 1.9392397584627852e-05, "loss": 2.3366, "step": 5607 }, { "epoch": 0.14, "learning_rate": 1.9392123518102996e-05, "loss": 2.1108, "step": 5608 }, { "epoch": 0.14, "learning_rate": 1.9391849391719074e-05, "loss": 2.4509, "step": 5609 }, { "epoch": 0.14, "learning_rate": 1.9391575205477834e-05, "loss": 2.2516, "step": 5610 }, { "epoch": 0.14, "learning_rate": 1.9391300959381024e-05, "loss": 2.174, "step": 5611 }, { "epoch": 0.14, "learning_rate": 1.9391026653430396e-05, "loss": 2.2575, "step": 5612 }, { "epoch": 0.14, "learning_rate": 1.9390752287627692e-05, "loss": 2.1687, "step": 5613 }, { "epoch": 0.14, "learning_rate": 1.9390477861974663e-05, "loss": 2.2364, "step": 5614 }, { "epoch": 0.14, "learning_rate": 1.939020337647306e-05, "loss": 2.32, "step": 5615 }, { "epoch": 0.14, "learning_rate": 1.938992883112463e-05, "loss": 2.3475, "step": 5616 }, { "epoch": 0.14, "learning_rate": 1.9389654225931123e-05, "loss": 2.2968, "step": 5617 }, { "epoch": 0.14, "learning_rate": 1.938937956089429e-05, "loss": 2.4147, "step": 5618 }, { "epoch": 0.14, "learning_rate": 1.9389104836015882e-05, "loss": 2.3594, "step": 5619 }, { "epoch": 0.14, "learning_rate": 1.938883005129765e-05, "loss": 2.2879, "step": 5620 }, { "epoch": 0.14, "learning_rate": 1.9388555206741338e-05, "loss": 2.2441, "step": 5621 }, { "epoch": 0.14, "learning_rate": 1.9388280302348713e-05, "loss": 2.2394, "step": 5622 }, { "epoch": 0.14, "learning_rate": 1.938800533812151e-05, "loss": 2.1284, "step": 5623 }, { "epoch": 0.14, "learning_rate": 1.9387730314061497e-05, "loss": 2.1752, "step": 5624 }, { "epoch": 0.14, "learning_rate": 1.9387455230170413e-05, "loss": 2.3325, "step": 5625 }, { "epoch": 0.14, "learning_rate": 1.938718008645002e-05, "loss": 2.3042, "step": 5626 }, { "epoch": 0.14, "learning_rate": 1.938690488290207e-05, "loss": 2.3955, "step": 5627 }, { "epoch": 0.14, "learning_rate": 1.9386629619528318e-05, "loss": 2.2642, "step": 5628 }, { "epoch": 0.14, "learning_rate": 1.938635429633051e-05, "loss": 2.206, "step": 5629 }, { "epoch": 0.14, "learning_rate": 1.9386078913310413e-05, "loss": 2.2018, "step": 5630 }, { "epoch": 0.14, "learning_rate": 1.938580347046977e-05, "loss": 2.1842, "step": 5631 }, { "epoch": 0.14, "learning_rate": 1.9385527967810352e-05, "loss": 2.3237, "step": 5632 }, { "epoch": 0.14, "learning_rate": 1.93852524053339e-05, "loss": 2.01, "step": 5633 }, { "epoch": 0.14, "learning_rate": 1.9384976783042172e-05, "loss": 2.3591, "step": 5634 }, { "epoch": 0.14, "learning_rate": 1.9384701100936936e-05, "loss": 2.104, "step": 5635 }, { "epoch": 0.14, "learning_rate": 1.9384425359019938e-05, "loss": 2.1932, "step": 5636 }, { "epoch": 0.14, "learning_rate": 1.938414955729294e-05, "loss": 2.3862, "step": 5637 }, { "epoch": 0.14, "learning_rate": 1.9383873695757696e-05, "loss": 2.3, "step": 5638 }, { "epoch": 0.14, "learning_rate": 1.9383597774415966e-05, "loss": 2.3229, "step": 5639 }, { "epoch": 0.14, "learning_rate": 1.9383321793269515e-05, "loss": 2.2997, "step": 5640 }, { "epoch": 0.14, "learning_rate": 1.9383045752320092e-05, "loss": 2.225, "step": 5641 }, { "epoch": 0.14, "learning_rate": 1.938276965156946e-05, "loss": 2.1555, "step": 5642 }, { "epoch": 0.14, "learning_rate": 1.9382493491019386e-05, "loss": 2.3535, "step": 5643 }, { "epoch": 0.14, "learning_rate": 1.9382217270671617e-05, "loss": 2.3933, "step": 5644 }, { "epoch": 0.14, "learning_rate": 1.9381940990527922e-05, "loss": 2.2877, "step": 5645 }, { "epoch": 0.14, "learning_rate": 1.9381664650590058e-05, "loss": 2.2962, "step": 5646 }, { "epoch": 0.14, "learning_rate": 1.9381388250859786e-05, "loss": 2.2428, "step": 5647 }, { "epoch": 0.14, "learning_rate": 1.9381111791338874e-05, "loss": 2.1997, "step": 5648 }, { "epoch": 0.14, "learning_rate": 1.9380835272029078e-05, "loss": 2.2542, "step": 5649 }, { "epoch": 0.14, "learning_rate": 1.9380558692932162e-05, "loss": 2.3168, "step": 5650 }, { "epoch": 0.14, "learning_rate": 1.9380282054049886e-05, "loss": 2.3344, "step": 5651 }, { "epoch": 0.14, "learning_rate": 1.9380005355384014e-05, "loss": 2.1498, "step": 5652 }, { "epoch": 0.14, "learning_rate": 1.9379728596936315e-05, "loss": 2.2514, "step": 5653 }, { "epoch": 0.14, "learning_rate": 1.9379451778708548e-05, "loss": 2.2754, "step": 5654 }, { "epoch": 0.14, "learning_rate": 1.9379174900702477e-05, "loss": 2.2578, "step": 5655 }, { "epoch": 0.14, "learning_rate": 1.937889796291987e-05, "loss": 2.1414, "step": 5656 }, { "epoch": 0.14, "learning_rate": 1.937862096536249e-05, "loss": 2.2614, "step": 5657 }, { "epoch": 0.14, "learning_rate": 1.93783439080321e-05, "loss": 2.2496, "step": 5658 }, { "epoch": 0.14, "learning_rate": 1.937806679093047e-05, "loss": 2.2476, "step": 5659 }, { "epoch": 0.14, "learning_rate": 1.937778961405936e-05, "loss": 2.1407, "step": 5660 }, { "epoch": 0.14, "learning_rate": 1.9377512377420546e-05, "loss": 2.3705, "step": 5661 }, { "epoch": 0.14, "learning_rate": 1.9377235081015785e-05, "loss": 2.3064, "step": 5662 }, { "epoch": 0.14, "learning_rate": 1.937695772484685e-05, "loss": 2.3307, "step": 5663 }, { "epoch": 0.14, "learning_rate": 1.937668030891551e-05, "loss": 2.2925, "step": 5664 }, { "epoch": 0.14, "learning_rate": 1.937640283322353e-05, "loss": 2.3254, "step": 5665 }, { "epoch": 0.14, "learning_rate": 1.9376125297772675e-05, "loss": 2.2355, "step": 5666 }, { "epoch": 0.14, "learning_rate": 1.9375847702564722e-05, "loss": 2.1505, "step": 5667 }, { "epoch": 0.14, "learning_rate": 1.9375570047601435e-05, "loss": 2.402, "step": 5668 }, { "epoch": 0.14, "learning_rate": 1.9375292332884584e-05, "loss": 2.2841, "step": 5669 }, { "epoch": 0.14, "learning_rate": 1.9375014558415942e-05, "loss": 2.3203, "step": 5670 }, { "epoch": 0.14, "learning_rate": 1.9374736724197274e-05, "loss": 2.2328, "step": 5671 }, { "epoch": 0.14, "learning_rate": 1.9374458830230355e-05, "loss": 2.2269, "step": 5672 }, { "epoch": 0.14, "learning_rate": 1.9374180876516955e-05, "loss": 2.2891, "step": 5673 }, { "epoch": 0.14, "learning_rate": 1.9373902863058848e-05, "loss": 2.4369, "step": 5674 }, { "epoch": 0.14, "learning_rate": 1.93736247898578e-05, "loss": 2.1057, "step": 5675 }, { "epoch": 0.14, "learning_rate": 1.9373346656915585e-05, "loss": 2.2995, "step": 5676 }, { "epoch": 0.14, "learning_rate": 1.937306846423398e-05, "loss": 2.3223, "step": 5677 }, { "epoch": 0.14, "learning_rate": 1.9372790211814752e-05, "loss": 2.2429, "step": 5678 }, { "epoch": 0.14, "learning_rate": 1.937251189965968e-05, "loss": 2.2039, "step": 5679 }, { "epoch": 0.14, "learning_rate": 1.9372233527770533e-05, "loss": 2.3965, "step": 5680 }, { "epoch": 0.14, "learning_rate": 1.9371955096149092e-05, "loss": 2.3667, "step": 5681 }, { "epoch": 0.14, "learning_rate": 1.9371676604797124e-05, "loss": 2.2007, "step": 5682 }, { "epoch": 0.14, "learning_rate": 1.9371398053716407e-05, "loss": 2.2048, "step": 5683 }, { "epoch": 0.14, "learning_rate": 1.9371119442908717e-05, "loss": 2.4038, "step": 5684 }, { "epoch": 0.14, "learning_rate": 1.9370840772375826e-05, "loss": 2.3935, "step": 5685 }, { "epoch": 0.14, "learning_rate": 1.9370562042119516e-05, "loss": 2.2394, "step": 5686 }, { "epoch": 0.14, "learning_rate": 1.937028325214156e-05, "loss": 2.4572, "step": 5687 }, { "epoch": 0.14, "learning_rate": 1.9370004402443734e-05, "loss": 2.257, "step": 5688 }, { "epoch": 0.14, "learning_rate": 1.9369725493027816e-05, "loss": 2.202, "step": 5689 }, { "epoch": 0.14, "learning_rate": 1.9369446523895584e-05, "loss": 2.1028, "step": 5690 }, { "epoch": 0.14, "learning_rate": 1.9369167495048817e-05, "loss": 2.1671, "step": 5691 }, { "epoch": 0.14, "learning_rate": 1.936888840648929e-05, "loss": 2.2282, "step": 5692 }, { "epoch": 0.14, "learning_rate": 1.9368609258218788e-05, "loss": 2.2634, "step": 5693 }, { "epoch": 0.14, "learning_rate": 1.936833005023908e-05, "loss": 2.2592, "step": 5694 }, { "epoch": 0.14, "learning_rate": 1.936805078255196e-05, "loss": 2.1264, "step": 5695 }, { "epoch": 0.14, "learning_rate": 1.9367771455159194e-05, "loss": 2.3614, "step": 5696 }, { "epoch": 0.14, "learning_rate": 1.9367492068062565e-05, "loss": 2.2955, "step": 5697 }, { "epoch": 0.14, "learning_rate": 1.936721262126386e-05, "loss": 2.2591, "step": 5698 }, { "epoch": 0.14, "learning_rate": 1.9366933114764856e-05, "loss": 2.0279, "step": 5699 }, { "epoch": 0.14, "learning_rate": 1.9366653548567336e-05, "loss": 2.3266, "step": 5700 }, { "epoch": 0.14, "learning_rate": 1.936637392267308e-05, "loss": 2.1839, "step": 5701 }, { "epoch": 0.14, "learning_rate": 1.936609423708387e-05, "loss": 2.3559, "step": 5702 }, { "epoch": 0.14, "learning_rate": 1.9365814491801487e-05, "loss": 2.2411, "step": 5703 }, { "epoch": 0.14, "learning_rate": 1.936553468682772e-05, "loss": 2.0617, "step": 5704 }, { "epoch": 0.14, "learning_rate": 1.9365254822164344e-05, "loss": 2.2665, "step": 5705 }, { "epoch": 0.14, "learning_rate": 1.936497489781315e-05, "loss": 2.3481, "step": 5706 }, { "epoch": 0.14, "learning_rate": 1.9364694913775916e-05, "loss": 2.2758, "step": 5707 }, { "epoch": 0.14, "learning_rate": 1.936441487005443e-05, "loss": 2.2333, "step": 5708 }, { "epoch": 0.14, "learning_rate": 1.936413476665048e-05, "loss": 2.2442, "step": 5709 }, { "epoch": 0.14, "learning_rate": 1.9363854603565843e-05, "loss": 2.335, "step": 5710 }, { "epoch": 0.14, "learning_rate": 1.936357438080231e-05, "loss": 2.3238, "step": 5711 }, { "epoch": 0.14, "learning_rate": 1.936329409836167e-05, "loss": 2.4171, "step": 5712 }, { "epoch": 0.14, "learning_rate": 1.9363013756245703e-05, "loss": 2.3024, "step": 5713 }, { "epoch": 0.14, "learning_rate": 1.9362733354456197e-05, "loss": 2.4386, "step": 5714 }, { "epoch": 0.14, "learning_rate": 1.936245289299494e-05, "loss": 2.3152, "step": 5715 }, { "epoch": 0.14, "learning_rate": 1.936217237186372e-05, "loss": 2.4369, "step": 5716 }, { "epoch": 0.14, "learning_rate": 1.936189179106432e-05, "loss": 2.2576, "step": 5717 }, { "epoch": 0.14, "learning_rate": 1.936161115059854e-05, "loss": 2.4431, "step": 5718 }, { "epoch": 0.14, "learning_rate": 1.936133045046816e-05, "loss": 2.1299, "step": 5719 }, { "epoch": 0.14, "learning_rate": 1.936104969067497e-05, "loss": 2.0127, "step": 5720 }, { "epoch": 0.14, "learning_rate": 1.9360768871220754e-05, "loss": 2.2805, "step": 5721 }, { "epoch": 0.14, "learning_rate": 1.9360487992107312e-05, "loss": 2.3042, "step": 5722 }, { "epoch": 0.14, "learning_rate": 1.9360207053336432e-05, "loss": 2.0754, "step": 5723 }, { "epoch": 0.14, "learning_rate": 1.9359926054909895e-05, "loss": 2.2243, "step": 5724 }, { "epoch": 0.14, "learning_rate": 1.9359644996829505e-05, "loss": 2.343, "step": 5725 }, { "epoch": 0.14, "learning_rate": 1.9359363879097046e-05, "loss": 2.3847, "step": 5726 }, { "epoch": 0.14, "learning_rate": 1.935908270171431e-05, "loss": 2.1741, "step": 5727 }, { "epoch": 0.14, "learning_rate": 1.9358801464683093e-05, "loss": 2.296, "step": 5728 }, { "epoch": 0.14, "learning_rate": 1.935852016800518e-05, "loss": 2.2559, "step": 5729 }, { "epoch": 0.14, "learning_rate": 1.9358238811682367e-05, "loss": 2.2499, "step": 5730 }, { "epoch": 0.14, "learning_rate": 1.9357957395716455e-05, "loss": 2.3378, "step": 5731 }, { "epoch": 0.14, "learning_rate": 1.9357675920109225e-05, "loss": 2.4544, "step": 5732 }, { "epoch": 0.14, "learning_rate": 1.9357394384862476e-05, "loss": 2.276, "step": 5733 }, { "epoch": 0.14, "learning_rate": 1.9357112789978008e-05, "loss": 2.2819, "step": 5734 }, { "epoch": 0.14, "learning_rate": 1.935683113545761e-05, "loss": 2.2998, "step": 5735 }, { "epoch": 0.14, "learning_rate": 1.9356549421303076e-05, "loss": 2.2254, "step": 5736 }, { "epoch": 0.14, "learning_rate": 1.9356267647516203e-05, "loss": 2.3902, "step": 5737 }, { "epoch": 0.14, "learning_rate": 1.9355985814098786e-05, "loss": 2.295, "step": 5738 }, { "epoch": 0.14, "learning_rate": 1.9355703921052625e-05, "loss": 2.3732, "step": 5739 }, { "epoch": 0.14, "learning_rate": 1.9355421968379513e-05, "loss": 2.2105, "step": 5740 }, { "epoch": 0.14, "learning_rate": 1.9355139956081247e-05, "loss": 2.2601, "step": 5741 }, { "epoch": 0.14, "learning_rate": 1.9354857884159627e-05, "loss": 2.3258, "step": 5742 }, { "epoch": 0.14, "learning_rate": 1.9354575752616452e-05, "loss": 2.285, "step": 5743 }, { "epoch": 0.14, "learning_rate": 1.9354293561453512e-05, "loss": 2.4097, "step": 5744 }, { "epoch": 0.14, "learning_rate": 1.9354011310672615e-05, "loss": 2.3492, "step": 5745 }, { "epoch": 0.14, "learning_rate": 1.935372900027555e-05, "loss": 2.1761, "step": 5746 }, { "epoch": 0.14, "learning_rate": 1.9353446630264124e-05, "loss": 2.3672, "step": 5747 }, { "epoch": 0.14, "learning_rate": 1.9353164200640134e-05, "loss": 2.2435, "step": 5748 }, { "epoch": 0.14, "learning_rate": 1.935288171140538e-05, "loss": 2.3649, "step": 5749 }, { "epoch": 0.14, "learning_rate": 1.9352599162561668e-05, "loss": 2.3265, "step": 5750 }, { "epoch": 0.14, "learning_rate": 1.935231655411079e-05, "loss": 2.3456, "step": 5751 }, { "epoch": 0.14, "learning_rate": 1.935203388605455e-05, "loss": 2.0878, "step": 5752 }, { "epoch": 0.14, "learning_rate": 1.935175115839475e-05, "loss": 2.1869, "step": 5753 }, { "epoch": 0.14, "learning_rate": 1.9351468371133193e-05, "loss": 2.0986, "step": 5754 }, { "epoch": 0.14, "learning_rate": 1.9351185524271684e-05, "loss": 2.4979, "step": 5755 }, { "epoch": 0.14, "learning_rate": 1.9350902617812015e-05, "loss": 2.2183, "step": 5756 }, { "epoch": 0.14, "learning_rate": 1.9350619651756002e-05, "loss": 2.3826, "step": 5757 }, { "epoch": 0.14, "learning_rate": 1.9350336626105442e-05, "loss": 2.4297, "step": 5758 }, { "epoch": 0.14, "learning_rate": 1.9350053540862136e-05, "loss": 2.4191, "step": 5759 }, { "epoch": 0.14, "learning_rate": 1.9349770396027897e-05, "loss": 2.3006, "step": 5760 }, { "epoch": 0.14, "learning_rate": 1.934948719160452e-05, "loss": 2.1833, "step": 5761 }, { "epoch": 0.14, "learning_rate": 1.9349203927593816e-05, "loss": 2.181, "step": 5762 }, { "epoch": 0.14, "learning_rate": 1.9348920603997587e-05, "loss": 2.2884, "step": 5763 }, { "epoch": 0.14, "learning_rate": 1.934863722081764e-05, "loss": 1.985, "step": 5764 }, { "epoch": 0.14, "learning_rate": 1.9348353778055786e-05, "loss": 2.2653, "step": 5765 }, { "epoch": 0.14, "learning_rate": 1.9348070275713822e-05, "loss": 2.3174, "step": 5766 }, { "epoch": 0.14, "learning_rate": 1.934778671379356e-05, "loss": 2.36, "step": 5767 }, { "epoch": 0.14, "learning_rate": 1.934750309229681e-05, "loss": 2.306, "step": 5768 }, { "epoch": 0.14, "learning_rate": 1.9347219411225374e-05, "loss": 2.1752, "step": 5769 }, { "epoch": 0.14, "learning_rate": 1.934693567058106e-05, "loss": 2.4276, "step": 5770 }, { "epoch": 0.14, "learning_rate": 1.9346651870365684e-05, "loss": 2.205, "step": 5771 }, { "epoch": 0.14, "learning_rate": 1.9346368010581048e-05, "loss": 2.2899, "step": 5772 }, { "epoch": 0.14, "learning_rate": 1.9346084091228962e-05, "loss": 2.1809, "step": 5773 }, { "epoch": 0.14, "learning_rate": 1.9345800112311236e-05, "loss": 2.2594, "step": 5774 }, { "epoch": 0.14, "learning_rate": 1.9345516073829686e-05, "loss": 2.2969, "step": 5775 }, { "epoch": 0.14, "learning_rate": 1.934523197578611e-05, "loss": 2.4419, "step": 5776 }, { "epoch": 0.14, "learning_rate": 1.9344947818182324e-05, "loss": 2.2766, "step": 5777 }, { "epoch": 0.14, "learning_rate": 1.9344663601020142e-05, "loss": 2.236, "step": 5778 }, { "epoch": 0.14, "learning_rate": 1.9344379324301377e-05, "loss": 2.3916, "step": 5779 }, { "epoch": 0.14, "learning_rate": 1.9344094988027833e-05, "loss": 2.1288, "step": 5780 }, { "epoch": 0.14, "learning_rate": 1.9343810592201327e-05, "loss": 2.3559, "step": 5781 }, { "epoch": 0.14, "learning_rate": 1.9343526136823675e-05, "loss": 2.237, "step": 5782 }, { "epoch": 0.14, "learning_rate": 1.9343241621896682e-05, "loss": 2.1117, "step": 5783 }, { "epoch": 0.14, "learning_rate": 1.9342957047422163e-05, "loss": 2.2198, "step": 5784 }, { "epoch": 0.14, "learning_rate": 1.934267241340194e-05, "loss": 2.3967, "step": 5785 }, { "epoch": 0.14, "learning_rate": 1.9342387719837815e-05, "loss": 2.2581, "step": 5786 }, { "epoch": 0.14, "learning_rate": 1.9342102966731616e-05, "loss": 2.4847, "step": 5787 }, { "epoch": 0.14, "learning_rate": 1.9341818154085143e-05, "loss": 2.2561, "step": 5788 }, { "epoch": 0.14, "learning_rate": 1.9341533281900223e-05, "loss": 2.2971, "step": 5789 }, { "epoch": 0.14, "learning_rate": 1.9341248350178665e-05, "loss": 2.2083, "step": 5790 }, { "epoch": 0.14, "learning_rate": 1.9340963358922285e-05, "loss": 2.1401, "step": 5791 }, { "epoch": 0.14, "learning_rate": 1.9340678308132906e-05, "loss": 2.2754, "step": 5792 }, { "epoch": 0.14, "learning_rate": 1.934039319781234e-05, "loss": 2.1966, "step": 5793 }, { "epoch": 0.14, "learning_rate": 1.93401080279624e-05, "loss": 2.0485, "step": 5794 }, { "epoch": 0.14, "learning_rate": 1.933982279858491e-05, "loss": 1.9761, "step": 5795 }, { "epoch": 0.14, "learning_rate": 1.9339537509681684e-05, "loss": 2.3004, "step": 5796 }, { "epoch": 0.14, "learning_rate": 1.9339252161254546e-05, "loss": 2.3517, "step": 5797 }, { "epoch": 0.14, "learning_rate": 1.9338966753305305e-05, "loss": 2.3567, "step": 5798 }, { "epoch": 0.14, "learning_rate": 1.933868128583579e-05, "loss": 2.3176, "step": 5799 }, { "epoch": 0.14, "learning_rate": 1.9338395758847816e-05, "loss": 2.2003, "step": 5800 }, { "epoch": 0.14, "learning_rate": 1.93381101723432e-05, "loss": 2.3046, "step": 5801 }, { "epoch": 0.14, "learning_rate": 1.933782452632376e-05, "loss": 2.3975, "step": 5802 }, { "epoch": 0.14, "learning_rate": 1.9337538820791328e-05, "loss": 2.2162, "step": 5803 }, { "epoch": 0.14, "learning_rate": 1.9337253055747717e-05, "loss": 2.2313, "step": 5804 }, { "epoch": 0.14, "learning_rate": 1.933696723119475e-05, "loss": 2.2143, "step": 5805 }, { "epoch": 0.14, "learning_rate": 1.9336681347134246e-05, "loss": 2.4974, "step": 5806 }, { "epoch": 0.14, "learning_rate": 1.9336395403568028e-05, "loss": 2.3421, "step": 5807 }, { "epoch": 0.14, "learning_rate": 1.9336109400497924e-05, "loss": 2.2121, "step": 5808 }, { "epoch": 0.14, "learning_rate": 1.933582333792575e-05, "loss": 2.211, "step": 5809 }, { "epoch": 0.14, "learning_rate": 1.933553721585333e-05, "loss": 2.099, "step": 5810 }, { "epoch": 0.14, "learning_rate": 1.933525103428249e-05, "loss": 2.4334, "step": 5811 }, { "epoch": 0.14, "learning_rate": 1.933496479321505e-05, "loss": 2.2109, "step": 5812 }, { "epoch": 0.14, "learning_rate": 1.9334678492652844e-05, "loss": 2.37, "step": 5813 }, { "epoch": 0.14, "learning_rate": 1.9334392132597682e-05, "loss": 2.347, "step": 5814 }, { "epoch": 0.14, "learning_rate": 1.9334105713051402e-05, "loss": 2.2616, "step": 5815 }, { "epoch": 0.14, "learning_rate": 1.9333819234015823e-05, "loss": 2.2495, "step": 5816 }, { "epoch": 0.14, "learning_rate": 1.933353269549277e-05, "loss": 2.2387, "step": 5817 }, { "epoch": 0.14, "learning_rate": 1.933324609748408e-05, "loss": 2.2532, "step": 5818 }, { "epoch": 0.14, "learning_rate": 1.933295943999156e-05, "loss": 2.3615, "step": 5819 }, { "epoch": 0.14, "learning_rate": 1.9332672723017053e-05, "loss": 2.2451, "step": 5820 }, { "epoch": 0.14, "learning_rate": 1.933238594656238e-05, "loss": 2.2526, "step": 5821 }, { "epoch": 0.14, "learning_rate": 1.9332099110629372e-05, "loss": 2.2309, "step": 5822 }, { "epoch": 0.14, "learning_rate": 1.9331812215219857e-05, "loss": 2.3291, "step": 5823 }, { "epoch": 0.14, "learning_rate": 1.9331525260335655e-05, "loss": 2.3728, "step": 5824 }, { "epoch": 0.14, "learning_rate": 1.9331238245978605e-05, "loss": 2.173, "step": 5825 }, { "epoch": 0.14, "learning_rate": 1.9330951172150533e-05, "loss": 2.2882, "step": 5826 }, { "epoch": 0.14, "learning_rate": 1.933066403885327e-05, "loss": 2.2875, "step": 5827 }, { "epoch": 0.14, "learning_rate": 1.9330376846088643e-05, "loss": 2.3102, "step": 5828 }, { "epoch": 0.14, "learning_rate": 1.9330089593858482e-05, "loss": 2.2375, "step": 5829 }, { "epoch": 0.14, "learning_rate": 1.932980228216462e-05, "loss": 2.2382, "step": 5830 }, { "epoch": 0.14, "learning_rate": 1.932951491100889e-05, "loss": 2.0411, "step": 5831 }, { "epoch": 0.14, "learning_rate": 1.9329227480393118e-05, "loss": 2.0599, "step": 5832 }, { "epoch": 0.14, "learning_rate": 1.9328939990319142e-05, "loss": 2.2425, "step": 5833 }, { "epoch": 0.14, "learning_rate": 1.9328652440788785e-05, "loss": 2.4508, "step": 5834 }, { "epoch": 0.14, "learning_rate": 1.9328364831803892e-05, "loss": 2.3485, "step": 5835 }, { "epoch": 0.14, "learning_rate": 1.932807716336629e-05, "loss": 2.1791, "step": 5836 }, { "epoch": 0.14, "learning_rate": 1.9327789435477807e-05, "loss": 2.1337, "step": 5837 }, { "epoch": 0.14, "learning_rate": 1.9327501648140287e-05, "loss": 2.3878, "step": 5838 }, { "epoch": 0.14, "learning_rate": 1.932721380135556e-05, "loss": 2.2962, "step": 5839 }, { "epoch": 0.14, "learning_rate": 1.9326925895125457e-05, "loss": 2.2215, "step": 5840 }, { "epoch": 0.14, "learning_rate": 1.9326637929451818e-05, "loss": 2.3276, "step": 5841 }, { "epoch": 0.14, "learning_rate": 1.9326349904336473e-05, "loss": 2.3406, "step": 5842 }, { "epoch": 0.14, "learning_rate": 1.9326061819781263e-05, "loss": 2.2342, "step": 5843 }, { "epoch": 0.14, "learning_rate": 1.932577367578802e-05, "loss": 2.3142, "step": 5844 }, { "epoch": 0.14, "learning_rate": 1.9325485472358584e-05, "loss": 2.0983, "step": 5845 }, { "epoch": 0.14, "learning_rate": 1.9325197209494786e-05, "loss": 2.3483, "step": 5846 }, { "epoch": 0.14, "learning_rate": 1.932490888719847e-05, "loss": 2.3576, "step": 5847 }, { "epoch": 0.14, "learning_rate": 1.932462050547147e-05, "loss": 2.1394, "step": 5848 }, { "epoch": 0.14, "learning_rate": 1.9324332064315627e-05, "loss": 2.3775, "step": 5849 }, { "epoch": 0.14, "learning_rate": 1.9324043563732777e-05, "loss": 2.0959, "step": 5850 }, { "epoch": 0.14, "learning_rate": 1.932375500372476e-05, "loss": 2.2431, "step": 5851 }, { "epoch": 0.14, "learning_rate": 1.932346638429341e-05, "loss": 2.2748, "step": 5852 }, { "epoch": 0.14, "learning_rate": 1.932317770544057e-05, "loss": 2.3101, "step": 5853 }, { "epoch": 0.14, "learning_rate": 1.9322888967168082e-05, "loss": 2.1958, "step": 5854 }, { "epoch": 0.14, "learning_rate": 1.9322600169477785e-05, "loss": 2.2369, "step": 5855 }, { "epoch": 0.14, "learning_rate": 1.9322311312371517e-05, "loss": 2.2045, "step": 5856 }, { "epoch": 0.14, "learning_rate": 1.932202239585112e-05, "loss": 2.2404, "step": 5857 }, { "epoch": 0.14, "learning_rate": 1.932173341991844e-05, "loss": 2.2294, "step": 5858 }, { "epoch": 0.14, "learning_rate": 1.9321444384575314e-05, "loss": 2.2746, "step": 5859 }, { "epoch": 0.14, "learning_rate": 1.9321155289823586e-05, "loss": 2.2943, "step": 5860 }, { "epoch": 0.14, "learning_rate": 1.9320866135665097e-05, "loss": 2.4545, "step": 5861 }, { "epoch": 0.14, "learning_rate": 1.932057692210169e-05, "loss": 2.4246, "step": 5862 }, { "epoch": 0.14, "learning_rate": 1.9320287649135205e-05, "loss": 2.2877, "step": 5863 }, { "epoch": 0.14, "learning_rate": 1.9319998316767496e-05, "loss": 2.3942, "step": 5864 }, { "epoch": 0.14, "learning_rate": 1.9319708925000395e-05, "loss": 2.166, "step": 5865 }, { "epoch": 0.14, "learning_rate": 1.9319419473835755e-05, "loss": 2.2577, "step": 5866 }, { "epoch": 0.14, "learning_rate": 1.9319129963275414e-05, "loss": 2.3831, "step": 5867 }, { "epoch": 0.14, "learning_rate": 1.9318840393321224e-05, "loss": 2.3677, "step": 5868 }, { "epoch": 0.14, "learning_rate": 1.931855076397503e-05, "loss": 2.275, "step": 5869 }, { "epoch": 0.14, "learning_rate": 1.9318261075238668e-05, "loss": 2.4245, "step": 5870 }, { "epoch": 0.14, "learning_rate": 1.9317971327113998e-05, "loss": 2.0993, "step": 5871 }, { "epoch": 0.14, "learning_rate": 1.931768151960286e-05, "loss": 2.3568, "step": 5872 }, { "epoch": 0.14, "learning_rate": 1.9317391652707097e-05, "loss": 2.3559, "step": 5873 }, { "epoch": 0.14, "learning_rate": 1.931710172642856e-05, "loss": 2.2445, "step": 5874 }, { "epoch": 0.14, "learning_rate": 1.93168117407691e-05, "loss": 2.1759, "step": 5875 }, { "epoch": 0.14, "learning_rate": 1.9316521695730563e-05, "loss": 2.1458, "step": 5876 }, { "epoch": 0.14, "learning_rate": 1.9316231591314793e-05, "loss": 2.216, "step": 5877 }, { "epoch": 0.14, "learning_rate": 1.931594142752365e-05, "loss": 2.2787, "step": 5878 }, { "epoch": 0.14, "learning_rate": 1.931565120435897e-05, "loss": 2.3876, "step": 5879 }, { "epoch": 0.14, "learning_rate": 1.931536092182261e-05, "loss": 2.1357, "step": 5880 }, { "epoch": 0.14, "learning_rate": 1.931507057991642e-05, "loss": 2.2021, "step": 5881 }, { "epoch": 0.14, "learning_rate": 1.9314780178642253e-05, "loss": 2.2522, "step": 5882 }, { "epoch": 0.15, "learning_rate": 1.931448971800195e-05, "loss": 2.148, "step": 5883 }, { "epoch": 0.15, "learning_rate": 1.9314199197997373e-05, "loss": 2.2468, "step": 5884 }, { "epoch": 0.15, "learning_rate": 1.931390861863037e-05, "loss": 2.2414, "step": 5885 }, { "epoch": 0.15, "learning_rate": 1.9313617979902788e-05, "loss": 2.3024, "step": 5886 }, { "epoch": 0.15, "learning_rate": 1.931332728181649e-05, "loss": 2.3179, "step": 5887 }, { "epoch": 0.15, "learning_rate": 1.931303652437331e-05, "loss": 2.2029, "step": 5888 }, { "epoch": 0.15, "learning_rate": 1.9312745707575124e-05, "loss": 2.233, "step": 5889 }, { "epoch": 0.15, "learning_rate": 1.931245483142377e-05, "loss": 2.1724, "step": 5890 }, { "epoch": 0.15, "learning_rate": 1.9312163895921107e-05, "loss": 2.332, "step": 5891 }, { "epoch": 0.15, "learning_rate": 1.931187290106899e-05, "loss": 2.3571, "step": 5892 }, { "epoch": 0.15, "learning_rate": 1.931158184686927e-05, "loss": 2.2984, "step": 5893 }, { "epoch": 0.15, "learning_rate": 1.931129073332381e-05, "loss": 2.4471, "step": 5894 }, { "epoch": 0.15, "learning_rate": 1.9310999560434456e-05, "loss": 2.2858, "step": 5895 }, { "epoch": 0.15, "learning_rate": 1.9310708328203068e-05, "loss": 2.2106, "step": 5896 }, { "epoch": 0.15, "learning_rate": 1.93104170366315e-05, "loss": 2.2374, "step": 5897 }, { "epoch": 0.15, "learning_rate": 1.931012568572161e-05, "loss": 2.2637, "step": 5898 }, { "epoch": 0.15, "learning_rate": 1.9309834275475256e-05, "loss": 2.3692, "step": 5899 }, { "epoch": 0.15, "learning_rate": 1.9309542805894295e-05, "loss": 2.2823, "step": 5900 }, { "epoch": 0.15, "learning_rate": 1.9309251276980584e-05, "loss": 2.4537, "step": 5901 }, { "epoch": 0.15, "learning_rate": 1.9308959688735976e-05, "loss": 2.2706, "step": 5902 }, { "epoch": 0.15, "learning_rate": 1.9308668041162338e-05, "loss": 2.2558, "step": 5903 }, { "epoch": 0.15, "learning_rate": 1.9308376334261526e-05, "loss": 2.4116, "step": 5904 }, { "epoch": 0.15, "learning_rate": 1.93080845680354e-05, "loss": 2.2395, "step": 5905 }, { "epoch": 0.15, "learning_rate": 1.930779274248581e-05, "loss": 2.3511, "step": 5906 }, { "epoch": 0.15, "learning_rate": 1.9307500857614633e-05, "loss": 2.1711, "step": 5907 }, { "epoch": 0.15, "learning_rate": 1.930720891342371e-05, "loss": 2.1125, "step": 5908 }, { "epoch": 0.15, "learning_rate": 1.930691690991492e-05, "loss": 2.23, "step": 5909 }, { "epoch": 0.15, "learning_rate": 1.930662484709011e-05, "loss": 2.1687, "step": 5910 }, { "epoch": 0.15, "learning_rate": 1.930633272495115e-05, "loss": 2.3962, "step": 5911 }, { "epoch": 0.15, "learning_rate": 1.9306040543499897e-05, "loss": 2.2075, "step": 5912 }, { "epoch": 0.15, "learning_rate": 1.9305748302738214e-05, "loss": 2.3782, "step": 5913 }, { "epoch": 0.15, "learning_rate": 1.9305456002667967e-05, "loss": 2.4043, "step": 5914 }, { "epoch": 0.15, "learning_rate": 1.9305163643291015e-05, "loss": 2.2522, "step": 5915 }, { "epoch": 0.15, "learning_rate": 1.9304871224609224e-05, "loss": 2.183, "step": 5916 }, { "epoch": 0.15, "learning_rate": 1.9304578746624453e-05, "loss": 2.1317, "step": 5917 }, { "epoch": 0.15, "learning_rate": 1.930428620933857e-05, "loss": 2.3405, "step": 5918 }, { "epoch": 0.15, "learning_rate": 1.930399361275344e-05, "loss": 2.1723, "step": 5919 }, { "epoch": 0.15, "learning_rate": 1.930370095687093e-05, "loss": 2.346, "step": 5920 }, { "epoch": 0.15, "learning_rate": 1.9303408241692897e-05, "loss": 2.3669, "step": 5921 }, { "epoch": 0.15, "learning_rate": 1.930311546722121e-05, "loss": 2.2417, "step": 5922 }, { "epoch": 0.15, "learning_rate": 1.9302822633457745e-05, "loss": 2.2211, "step": 5923 }, { "epoch": 0.15, "learning_rate": 1.930252974040435e-05, "loss": 2.2137, "step": 5924 }, { "epoch": 0.15, "learning_rate": 1.9302236788062908e-05, "loss": 2.3222, "step": 5925 }, { "epoch": 0.15, "learning_rate": 1.9301943776435275e-05, "loss": 2.1163, "step": 5926 }, { "epoch": 0.15, "learning_rate": 1.9301650705523325e-05, "loss": 2.2236, "step": 5927 }, { "epoch": 0.15, "learning_rate": 1.9301357575328925e-05, "loss": 2.2831, "step": 5928 }, { "epoch": 0.15, "learning_rate": 1.930106438585394e-05, "loss": 2.2644, "step": 5929 }, { "epoch": 0.15, "learning_rate": 1.9300771137100242e-05, "loss": 2.2206, "step": 5930 }, { "epoch": 0.15, "learning_rate": 1.9300477829069696e-05, "loss": 2.2522, "step": 5931 }, { "epoch": 0.15, "learning_rate": 1.9300184461764173e-05, "loss": 2.1514, "step": 5932 }, { "epoch": 0.15, "learning_rate": 1.929989103518555e-05, "loss": 2.3577, "step": 5933 }, { "epoch": 0.15, "learning_rate": 1.9299597549335685e-05, "loss": 2.3904, "step": 5934 }, { "epoch": 0.15, "learning_rate": 1.9299304004216456e-05, "loss": 2.2315, "step": 5935 }, { "epoch": 0.15, "learning_rate": 1.9299010399829733e-05, "loss": 2.2587, "step": 5936 }, { "epoch": 0.15, "learning_rate": 1.929871673617738e-05, "loss": 2.2983, "step": 5937 }, { "epoch": 0.15, "learning_rate": 1.929842301326128e-05, "loss": 2.3419, "step": 5938 }, { "epoch": 0.15, "learning_rate": 1.9298129231083304e-05, "loss": 2.2277, "step": 5939 }, { "epoch": 0.15, "learning_rate": 1.9297835389645313e-05, "loss": 2.2802, "step": 5940 }, { "epoch": 0.15, "learning_rate": 1.929754148894919e-05, "loss": 2.2955, "step": 5941 }, { "epoch": 0.15, "learning_rate": 1.9297247528996806e-05, "loss": 2.2655, "step": 5942 }, { "epoch": 0.15, "learning_rate": 1.9296953509790032e-05, "loss": 2.1195, "step": 5943 }, { "epoch": 0.15, "learning_rate": 1.9296659431330745e-05, "loss": 2.2347, "step": 5944 }, { "epoch": 0.15, "learning_rate": 1.9296365293620818e-05, "loss": 2.2775, "step": 5945 }, { "epoch": 0.15, "learning_rate": 1.9296071096662123e-05, "loss": 2.2509, "step": 5946 }, { "epoch": 0.15, "learning_rate": 1.9295776840456536e-05, "loss": 2.4569, "step": 5947 }, { "epoch": 0.15, "learning_rate": 1.929548252500594e-05, "loss": 2.264, "step": 5948 }, { "epoch": 0.15, "learning_rate": 1.9295188150312202e-05, "loss": 2.3677, "step": 5949 }, { "epoch": 0.15, "learning_rate": 1.9294893716377195e-05, "loss": 2.4003, "step": 5950 }, { "epoch": 0.15, "learning_rate": 1.9294599223202808e-05, "loss": 2.2908, "step": 5951 }, { "epoch": 0.15, "learning_rate": 1.929430467079091e-05, "loss": 2.2044, "step": 5952 }, { "epoch": 0.15, "learning_rate": 1.929401005914338e-05, "loss": 2.343, "step": 5953 }, { "epoch": 0.15, "learning_rate": 1.9293715388262094e-05, "loss": 2.2423, "step": 5954 }, { "epoch": 0.15, "learning_rate": 1.929342065814893e-05, "loss": 2.3797, "step": 5955 }, { "epoch": 0.15, "learning_rate": 1.9293125868805765e-05, "loss": 2.1814, "step": 5956 }, { "epoch": 0.15, "learning_rate": 1.9292831020234486e-05, "loss": 2.2083, "step": 5957 }, { "epoch": 0.15, "learning_rate": 1.9292536112436963e-05, "loss": 2.386, "step": 5958 }, { "epoch": 0.15, "learning_rate": 1.929224114541508e-05, "loss": 2.4367, "step": 5959 }, { "epoch": 0.15, "learning_rate": 1.9291946119170715e-05, "loss": 2.333, "step": 5960 }, { "epoch": 0.15, "learning_rate": 1.929165103370575e-05, "loss": 2.0872, "step": 5961 }, { "epoch": 0.15, "learning_rate": 1.9291355889022068e-05, "loss": 2.2783, "step": 5962 }, { "epoch": 0.15, "learning_rate": 1.9291060685121542e-05, "loss": 2.2771, "step": 5963 }, { "epoch": 0.15, "learning_rate": 1.929076542200606e-05, "loss": 2.2759, "step": 5964 }, { "epoch": 0.15, "learning_rate": 1.92904700996775e-05, "loss": 2.3065, "step": 5965 }, { "epoch": 0.15, "learning_rate": 1.9290174718137754e-05, "loss": 2.4363, "step": 5966 }, { "epoch": 0.15, "learning_rate": 1.928987927738869e-05, "loss": 2.1725, "step": 5967 }, { "epoch": 0.15, "learning_rate": 1.92895837774322e-05, "loss": 2.2465, "step": 5968 }, { "epoch": 0.15, "learning_rate": 1.9289288218270164e-05, "loss": 2.1114, "step": 5969 }, { "epoch": 0.15, "learning_rate": 1.9288992599904467e-05, "loss": 2.137, "step": 5970 }, { "epoch": 0.15, "learning_rate": 1.9288696922336995e-05, "loss": 2.3601, "step": 5971 }, { "epoch": 0.15, "learning_rate": 1.928840118556963e-05, "loss": 2.4795, "step": 5972 }, { "epoch": 0.15, "learning_rate": 1.9288105389604256e-05, "loss": 2.319, "step": 5973 }, { "epoch": 0.15, "learning_rate": 1.928780953444276e-05, "loss": 2.3562, "step": 5974 }, { "epoch": 0.15, "learning_rate": 1.9287513620087027e-05, "loss": 2.341, "step": 5975 }, { "epoch": 0.15, "learning_rate": 1.928721764653894e-05, "loss": 2.2487, "step": 5976 }, { "epoch": 0.15, "learning_rate": 1.9286921613800393e-05, "loss": 2.2058, "step": 5977 }, { "epoch": 0.15, "learning_rate": 1.9286625521873266e-05, "loss": 2.4065, "step": 5978 }, { "epoch": 0.15, "learning_rate": 1.9286329370759448e-05, "loss": 2.3375, "step": 5979 }, { "epoch": 0.15, "learning_rate": 1.928603316046083e-05, "loss": 2.276, "step": 5980 }, { "epoch": 0.15, "learning_rate": 1.9285736890979293e-05, "loss": 2.2113, "step": 5981 }, { "epoch": 0.15, "learning_rate": 1.928544056231673e-05, "loss": 2.2347, "step": 5982 }, { "epoch": 0.15, "learning_rate": 1.9285144174475024e-05, "loss": 2.1374, "step": 5983 }, { "epoch": 0.15, "learning_rate": 1.9284847727456073e-05, "loss": 2.1876, "step": 5984 }, { "epoch": 0.15, "learning_rate": 1.928455122126176e-05, "loss": 2.3454, "step": 5985 }, { "epoch": 0.15, "learning_rate": 1.9284254655893978e-05, "loss": 2.2328, "step": 5986 }, { "epoch": 0.15, "learning_rate": 1.928395803135461e-05, "loss": 2.2939, "step": 5987 }, { "epoch": 0.15, "learning_rate": 1.9283661347645557e-05, "loss": 2.3175, "step": 5988 }, { "epoch": 0.15, "learning_rate": 1.9283364604768704e-05, "loss": 2.2368, "step": 5989 }, { "epoch": 0.15, "learning_rate": 1.9283067802725943e-05, "loss": 2.1788, "step": 5990 }, { "epoch": 0.15, "learning_rate": 1.9282770941519167e-05, "loss": 2.2982, "step": 5991 }, { "epoch": 0.15, "learning_rate": 1.9282474021150263e-05, "loss": 2.3609, "step": 5992 }, { "epoch": 0.15, "learning_rate": 1.928217704162113e-05, "loss": 2.3974, "step": 5993 }, { "epoch": 0.15, "learning_rate": 1.9281880002933656e-05, "loss": 2.2665, "step": 5994 }, { "epoch": 0.15, "learning_rate": 1.9281582905089737e-05, "loss": 2.0869, "step": 5995 }, { "epoch": 0.15, "learning_rate": 1.9281285748091263e-05, "loss": 2.2852, "step": 5996 }, { "epoch": 0.15, "learning_rate": 1.928098853194013e-05, "loss": 2.4194, "step": 5997 }, { "epoch": 0.15, "learning_rate": 1.928069125663824e-05, "loss": 2.1281, "step": 5998 }, { "epoch": 0.15, "learning_rate": 1.928039392218747e-05, "loss": 2.366, "step": 5999 }, { "epoch": 0.15, "learning_rate": 1.9280096528589733e-05, "loss": 2.1876, "step": 6000 }, { "epoch": 0.15, "learning_rate": 1.9279799075846913e-05, "loss": 2.1527, "step": 6001 }, { "epoch": 0.15, "learning_rate": 1.927950156396091e-05, "loss": 2.2284, "step": 6002 }, { "epoch": 0.15, "learning_rate": 1.927920399293362e-05, "loss": 2.2059, "step": 6003 }, { "epoch": 0.15, "learning_rate": 1.9278906362766934e-05, "loss": 2.3539, "step": 6004 }, { "epoch": 0.15, "learning_rate": 1.927860867346276e-05, "loss": 2.2774, "step": 6005 }, { "epoch": 0.15, "learning_rate": 1.9278310925022986e-05, "loss": 2.1975, "step": 6006 }, { "epoch": 0.15, "learning_rate": 1.9278013117449514e-05, "loss": 2.2412, "step": 6007 }, { "epoch": 0.15, "learning_rate": 1.927771525074424e-05, "loss": 2.1722, "step": 6008 }, { "epoch": 0.15, "learning_rate": 1.9277417324909065e-05, "loss": 2.3389, "step": 6009 }, { "epoch": 0.15, "learning_rate": 1.9277119339945883e-05, "loss": 2.3039, "step": 6010 }, { "epoch": 0.15, "learning_rate": 1.92768212958566e-05, "loss": 2.4345, "step": 6011 }, { "epoch": 0.15, "learning_rate": 1.927652319264311e-05, "loss": 2.2713, "step": 6012 }, { "epoch": 0.15, "learning_rate": 1.9276225030307315e-05, "loss": 2.2096, "step": 6013 }, { "epoch": 0.15, "learning_rate": 1.927592680885111e-05, "loss": 2.1983, "step": 6014 }, { "epoch": 0.15, "learning_rate": 1.9275628528276406e-05, "loss": 2.1427, "step": 6015 }, { "epoch": 0.15, "learning_rate": 1.9275330188585096e-05, "loss": 2.1261, "step": 6016 }, { "epoch": 0.15, "learning_rate": 1.9275031789779082e-05, "loss": 2.2843, "step": 6017 }, { "epoch": 0.15, "learning_rate": 1.9274733331860272e-05, "loss": 2.1673, "step": 6018 }, { "epoch": 0.15, "learning_rate": 1.9274434814830567e-05, "loss": 2.2937, "step": 6019 }, { "epoch": 0.15, "learning_rate": 1.927413623869186e-05, "loss": 2.2182, "step": 6020 }, { "epoch": 0.15, "learning_rate": 1.927383760344606e-05, "loss": 2.2509, "step": 6021 }, { "epoch": 0.15, "learning_rate": 1.9273538909095077e-05, "loss": 2.3083, "step": 6022 }, { "epoch": 0.15, "learning_rate": 1.9273240155640803e-05, "loss": 2.2009, "step": 6023 }, { "epoch": 0.15, "learning_rate": 1.927294134308515e-05, "loss": 2.3075, "step": 6024 }, { "epoch": 0.15, "learning_rate": 1.9272642471430018e-05, "loss": 2.4008, "step": 6025 }, { "epoch": 0.15, "learning_rate": 1.9272343540677315e-05, "loss": 2.3621, "step": 6026 }, { "epoch": 0.15, "learning_rate": 1.9272044550828942e-05, "loss": 2.126, "step": 6027 }, { "epoch": 0.15, "learning_rate": 1.9271745501886808e-05, "loss": 2.1364, "step": 6028 }, { "epoch": 0.15, "learning_rate": 1.9271446393852823e-05, "loss": 2.2536, "step": 6029 }, { "epoch": 0.15, "learning_rate": 1.9271147226728885e-05, "loss": 2.2139, "step": 6030 }, { "epoch": 0.15, "learning_rate": 1.9270848000516907e-05, "loss": 2.3563, "step": 6031 }, { "epoch": 0.15, "learning_rate": 1.9270548715218787e-05, "loss": 2.1044, "step": 6032 }, { "epoch": 0.15, "learning_rate": 1.9270249370836443e-05, "loss": 2.4313, "step": 6033 }, { "epoch": 0.15, "learning_rate": 1.9269949967371783e-05, "loss": 2.2455, "step": 6034 }, { "epoch": 0.15, "learning_rate": 1.9269650504826706e-05, "loss": 2.1152, "step": 6035 }, { "epoch": 0.15, "learning_rate": 1.9269350983203128e-05, "loss": 2.1864, "step": 6036 }, { "epoch": 0.15, "learning_rate": 1.9269051402502952e-05, "loss": 2.4047, "step": 6037 }, { "epoch": 0.15, "learning_rate": 1.9268751762728094e-05, "loss": 2.3515, "step": 6038 }, { "epoch": 0.15, "learning_rate": 1.9268452063880456e-05, "loss": 2.3229, "step": 6039 }, { "epoch": 0.15, "learning_rate": 1.9268152305961956e-05, "loss": 2.3399, "step": 6040 }, { "epoch": 0.15, "learning_rate": 1.92678524889745e-05, "loss": 2.2974, "step": 6041 }, { "epoch": 0.15, "learning_rate": 1.9267552612920004e-05, "loss": 2.1545, "step": 6042 }, { "epoch": 0.15, "learning_rate": 1.926725267780037e-05, "loss": 2.2026, "step": 6043 }, { "epoch": 0.15, "learning_rate": 1.9266952683617512e-05, "loss": 2.3303, "step": 6044 }, { "epoch": 0.15, "learning_rate": 1.926665263037335e-05, "loss": 2.2994, "step": 6045 }, { "epoch": 0.15, "learning_rate": 1.9266352518069787e-05, "loss": 2.2804, "step": 6046 }, { "epoch": 0.15, "learning_rate": 1.9266052346708743e-05, "loss": 2.2428, "step": 6047 }, { "epoch": 0.15, "learning_rate": 1.9265752116292128e-05, "loss": 2.2377, "step": 6048 }, { "epoch": 0.15, "learning_rate": 1.926545182682185e-05, "loss": 2.3494, "step": 6049 }, { "epoch": 0.15, "learning_rate": 1.9265151478299833e-05, "loss": 2.2161, "step": 6050 }, { "epoch": 0.15, "learning_rate": 1.9264851070727985e-05, "loss": 2.1917, "step": 6051 }, { "epoch": 0.15, "learning_rate": 1.926455060410822e-05, "loss": 2.2492, "step": 6052 }, { "epoch": 0.15, "learning_rate": 1.9264250078442455e-05, "loss": 2.1134, "step": 6053 }, { "epoch": 0.15, "learning_rate": 1.926394949373261e-05, "loss": 2.1116, "step": 6054 }, { "epoch": 0.15, "learning_rate": 1.926364884998059e-05, "loss": 2.2002, "step": 6055 }, { "epoch": 0.15, "learning_rate": 1.926334814718832e-05, "loss": 2.2406, "step": 6056 }, { "epoch": 0.15, "learning_rate": 1.926304738535771e-05, "loss": 2.1771, "step": 6057 }, { "epoch": 0.15, "learning_rate": 1.9262746564490685e-05, "loss": 2.2782, "step": 6058 }, { "epoch": 0.15, "learning_rate": 1.9262445684589155e-05, "loss": 2.2965, "step": 6059 }, { "epoch": 0.15, "learning_rate": 1.926214474565504e-05, "loss": 2.4149, "step": 6060 }, { "epoch": 0.15, "learning_rate": 1.926184374769026e-05, "loss": 2.2668, "step": 6061 }, { "epoch": 0.15, "learning_rate": 1.926154269069673e-05, "loss": 2.1033, "step": 6062 }, { "epoch": 0.15, "learning_rate": 1.9261241574676367e-05, "loss": 2.3264, "step": 6063 }, { "epoch": 0.15, "learning_rate": 1.92609403996311e-05, "loss": 2.3074, "step": 6064 }, { "epoch": 0.15, "learning_rate": 1.9260639165562835e-05, "loss": 2.1501, "step": 6065 }, { "epoch": 0.15, "learning_rate": 1.92603378724735e-05, "loss": 2.2393, "step": 6066 }, { "epoch": 0.15, "learning_rate": 1.926003652036502e-05, "loss": 2.1562, "step": 6067 }, { "epoch": 0.15, "learning_rate": 1.9259735109239305e-05, "loss": 2.1218, "step": 6068 }, { "epoch": 0.15, "learning_rate": 1.925943363909828e-05, "loss": 2.3671, "step": 6069 }, { "epoch": 0.15, "learning_rate": 1.9259132109943864e-05, "loss": 2.1755, "step": 6070 }, { "epoch": 0.15, "learning_rate": 1.925883052177799e-05, "loss": 2.2555, "step": 6071 }, { "epoch": 0.15, "learning_rate": 1.9258528874602564e-05, "loss": 2.1643, "step": 6072 }, { "epoch": 0.15, "learning_rate": 1.925822716841952e-05, "loss": 2.0946, "step": 6073 }, { "epoch": 0.15, "learning_rate": 1.9257925403230776e-05, "loss": 2.2723, "step": 6074 }, { "epoch": 0.15, "learning_rate": 1.9257623579038255e-05, "loss": 2.2691, "step": 6075 }, { "epoch": 0.15, "learning_rate": 1.9257321695843885e-05, "loss": 2.2742, "step": 6076 }, { "epoch": 0.15, "learning_rate": 1.9257019753649585e-05, "loss": 2.2488, "step": 6077 }, { "epoch": 0.15, "learning_rate": 1.925671775245728e-05, "loss": 2.2142, "step": 6078 }, { "epoch": 0.15, "learning_rate": 1.92564156922689e-05, "loss": 2.1135, "step": 6079 }, { "epoch": 0.15, "learning_rate": 1.9256113573086364e-05, "loss": 2.3434, "step": 6080 }, { "epoch": 0.15, "learning_rate": 1.9255811394911598e-05, "loss": 2.2609, "step": 6081 }, { "epoch": 0.15, "learning_rate": 1.9255509157746534e-05, "loss": 2.3368, "step": 6082 }, { "epoch": 0.15, "learning_rate": 1.925520686159309e-05, "loss": 2.2342, "step": 6083 }, { "epoch": 0.15, "learning_rate": 1.92549045064532e-05, "loss": 2.414, "step": 6084 }, { "epoch": 0.15, "learning_rate": 1.9254602092328782e-05, "loss": 2.3787, "step": 6085 }, { "epoch": 0.15, "learning_rate": 1.9254299619221773e-05, "loss": 2.3036, "step": 6086 }, { "epoch": 0.15, "learning_rate": 1.9253997087134094e-05, "loss": 2.3346, "step": 6087 }, { "epoch": 0.15, "learning_rate": 1.925369449606768e-05, "loss": 2.4551, "step": 6088 }, { "epoch": 0.15, "learning_rate": 1.9253391846024453e-05, "loss": 2.2716, "step": 6089 }, { "epoch": 0.15, "learning_rate": 1.9253089137006343e-05, "loss": 2.1975, "step": 6090 }, { "epoch": 0.15, "learning_rate": 1.9252786369015282e-05, "loss": 2.2256, "step": 6091 }, { "epoch": 0.15, "learning_rate": 1.9252483542053198e-05, "loss": 2.1991, "step": 6092 }, { "epoch": 0.15, "learning_rate": 1.9252180656122018e-05, "loss": 2.2651, "step": 6093 }, { "epoch": 0.15, "learning_rate": 1.9251877711223677e-05, "loss": 2.4055, "step": 6094 }, { "epoch": 0.15, "learning_rate": 1.9251574707360107e-05, "loss": 2.351, "step": 6095 }, { "epoch": 0.15, "learning_rate": 1.9251271644533235e-05, "loss": 2.1999, "step": 6096 }, { "epoch": 0.15, "learning_rate": 1.9250968522744993e-05, "loss": 2.0978, "step": 6097 }, { "epoch": 0.15, "learning_rate": 1.9250665341997314e-05, "loss": 2.3714, "step": 6098 }, { "epoch": 0.15, "learning_rate": 1.925036210229213e-05, "loss": 2.2174, "step": 6099 }, { "epoch": 0.15, "learning_rate": 1.9250058803631374e-05, "loss": 2.2087, "step": 6100 }, { "epoch": 0.15, "learning_rate": 1.924975544601698e-05, "loss": 2.2678, "step": 6101 }, { "epoch": 0.15, "learning_rate": 1.924945202945088e-05, "loss": 2.1875, "step": 6102 }, { "epoch": 0.15, "learning_rate": 1.9249148553935004e-05, "loss": 2.363, "step": 6103 }, { "epoch": 0.15, "learning_rate": 1.9248845019471293e-05, "loss": 2.3767, "step": 6104 }, { "epoch": 0.15, "learning_rate": 1.924854142606168e-05, "loss": 2.1842, "step": 6105 }, { "epoch": 0.15, "learning_rate": 1.9248237773708094e-05, "loss": 2.4574, "step": 6106 }, { "epoch": 0.15, "learning_rate": 1.924793406241248e-05, "loss": 2.2783, "step": 6107 }, { "epoch": 0.15, "learning_rate": 1.9247630292176765e-05, "loss": 2.1682, "step": 6108 }, { "epoch": 0.15, "learning_rate": 1.9247326463002892e-05, "loss": 2.2111, "step": 6109 }, { "epoch": 0.15, "learning_rate": 1.9247022574892792e-05, "loss": 2.3114, "step": 6110 }, { "epoch": 0.15, "learning_rate": 1.9246718627848402e-05, "loss": 2.3221, "step": 6111 }, { "epoch": 0.15, "learning_rate": 1.9246414621871662e-05, "loss": 2.2688, "step": 6112 }, { "epoch": 0.15, "learning_rate": 1.9246110556964508e-05, "loss": 2.1193, "step": 6113 }, { "epoch": 0.15, "learning_rate": 1.924580643312888e-05, "loss": 2.4242, "step": 6114 }, { "epoch": 0.15, "learning_rate": 1.9245502250366713e-05, "loss": 2.2358, "step": 6115 }, { "epoch": 0.15, "learning_rate": 1.9245198008679944e-05, "loss": 2.2684, "step": 6116 }, { "epoch": 0.15, "learning_rate": 1.9244893708070518e-05, "loss": 2.1636, "step": 6117 }, { "epoch": 0.15, "learning_rate": 1.9244589348540372e-05, "loss": 2.3047, "step": 6118 }, { "epoch": 0.15, "learning_rate": 1.9244284930091447e-05, "loss": 2.4006, "step": 6119 }, { "epoch": 0.15, "learning_rate": 1.9243980452725676e-05, "loss": 2.0873, "step": 6120 }, { "epoch": 0.15, "learning_rate": 1.9243675916445012e-05, "loss": 2.1908, "step": 6121 }, { "epoch": 0.15, "learning_rate": 1.9243371321251384e-05, "loss": 1.9275, "step": 6122 }, { "epoch": 0.15, "learning_rate": 1.924306666714674e-05, "loss": 2.2139, "step": 6123 }, { "epoch": 0.15, "learning_rate": 1.924276195413302e-05, "loss": 2.248, "step": 6124 }, { "epoch": 0.15, "learning_rate": 1.9242457182212166e-05, "loss": 2.3006, "step": 6125 }, { "epoch": 0.15, "learning_rate": 1.924215235138612e-05, "loss": 2.2507, "step": 6126 }, { "epoch": 0.15, "learning_rate": 1.9241847461656823e-05, "loss": 2.1769, "step": 6127 }, { "epoch": 0.15, "learning_rate": 1.9241542513026223e-05, "loss": 2.2759, "step": 6128 }, { "epoch": 0.15, "learning_rate": 1.924123750549626e-05, "loss": 2.0987, "step": 6129 }, { "epoch": 0.15, "learning_rate": 1.924093243906888e-05, "loss": 2.3188, "step": 6130 }, { "epoch": 0.15, "learning_rate": 1.9240627313746025e-05, "loss": 2.3344, "step": 6131 }, { "epoch": 0.15, "learning_rate": 1.924032212952964e-05, "loss": 2.2402, "step": 6132 }, { "epoch": 0.15, "learning_rate": 1.924001688642167e-05, "loss": 2.1535, "step": 6133 }, { "epoch": 0.15, "learning_rate": 1.9239711584424065e-05, "loss": 2.3397, "step": 6134 }, { "epoch": 0.15, "learning_rate": 1.9239406223538765e-05, "loss": 2.3127, "step": 6135 }, { "epoch": 0.15, "learning_rate": 1.9239100803767717e-05, "loss": 2.2744, "step": 6136 }, { "epoch": 0.15, "learning_rate": 1.9238795325112867e-05, "loss": 2.3729, "step": 6137 }, { "epoch": 0.15, "learning_rate": 1.923848978757617e-05, "loss": 2.0536, "step": 6138 }, { "epoch": 0.15, "learning_rate": 1.9238184191159562e-05, "loss": 2.3636, "step": 6139 }, { "epoch": 0.15, "learning_rate": 1.9237878535864997e-05, "loss": 2.1073, "step": 6140 }, { "epoch": 0.15, "learning_rate": 1.9237572821694418e-05, "loss": 2.314, "step": 6141 }, { "epoch": 0.15, "learning_rate": 1.923726704864978e-05, "loss": 2.2184, "step": 6142 }, { "epoch": 0.15, "learning_rate": 1.9236961216733026e-05, "loss": 2.4679, "step": 6143 }, { "epoch": 0.15, "learning_rate": 1.9236655325946113e-05, "loss": 2.2714, "step": 6144 }, { "epoch": 0.15, "learning_rate": 1.9236349376290983e-05, "loss": 2.3218, "step": 6145 }, { "epoch": 0.15, "learning_rate": 1.9236043367769587e-05, "loss": 2.3425, "step": 6146 }, { "epoch": 0.15, "learning_rate": 1.9235737300383878e-05, "loss": 2.3268, "step": 6147 }, { "epoch": 0.15, "learning_rate": 1.9235431174135805e-05, "loss": 2.3378, "step": 6148 }, { "epoch": 0.15, "learning_rate": 1.923512498902732e-05, "loss": 2.4821, "step": 6149 }, { "epoch": 0.15, "learning_rate": 1.9234818745060375e-05, "loss": 2.3205, "step": 6150 }, { "epoch": 0.15, "learning_rate": 1.9234512442236918e-05, "loss": 2.2346, "step": 6151 }, { "epoch": 0.15, "learning_rate": 1.9234206080558908e-05, "loss": 2.2042, "step": 6152 }, { "epoch": 0.15, "learning_rate": 1.923389966002829e-05, "loss": 2.2657, "step": 6153 }, { "epoch": 0.15, "learning_rate": 1.9233593180647017e-05, "loss": 2.211, "step": 6154 }, { "epoch": 0.15, "learning_rate": 1.923328664241705e-05, "loss": 2.3535, "step": 6155 }, { "epoch": 0.15, "learning_rate": 1.9232980045340338e-05, "loss": 2.177, "step": 6156 }, { "epoch": 0.15, "learning_rate": 1.9232673389418833e-05, "loss": 2.3434, "step": 6157 }, { "epoch": 0.15, "learning_rate": 1.9232366674654495e-05, "loss": 2.4088, "step": 6158 }, { "epoch": 0.15, "learning_rate": 1.923205990104927e-05, "loss": 2.3101, "step": 6159 }, { "epoch": 0.15, "learning_rate": 1.9231753068605124e-05, "loss": 2.0428, "step": 6160 }, { "epoch": 0.15, "learning_rate": 1.9231446177324006e-05, "loss": 2.4766, "step": 6161 }, { "epoch": 0.15, "learning_rate": 1.923113922720787e-05, "loss": 2.2396, "step": 6162 }, { "epoch": 0.15, "learning_rate": 1.9230832218258676e-05, "loss": 2.0881, "step": 6163 }, { "epoch": 0.15, "learning_rate": 1.9230525150478385e-05, "loss": 2.3324, "step": 6164 }, { "epoch": 0.15, "learning_rate": 1.923021802386894e-05, "loss": 2.2296, "step": 6165 }, { "epoch": 0.15, "learning_rate": 1.9229910838432318e-05, "loss": 2.1811, "step": 6166 }, { "epoch": 0.15, "learning_rate": 1.9229603594170463e-05, "loss": 2.3303, "step": 6167 }, { "epoch": 0.15, "learning_rate": 1.9229296291085334e-05, "loss": 2.2661, "step": 6168 }, { "epoch": 0.15, "learning_rate": 1.9228988929178893e-05, "loss": 2.2573, "step": 6169 }, { "epoch": 0.15, "learning_rate": 1.9228681508453095e-05, "loss": 2.3144, "step": 6170 }, { "epoch": 0.15, "learning_rate": 1.9228374028909908e-05, "loss": 2.2996, "step": 6171 }, { "epoch": 0.15, "learning_rate": 1.922806649055128e-05, "loss": 2.333, "step": 6172 }, { "epoch": 0.15, "learning_rate": 1.9227758893379183e-05, "loss": 2.19, "step": 6173 }, { "epoch": 0.15, "learning_rate": 1.9227451237395567e-05, "loss": 2.1189, "step": 6174 }, { "epoch": 0.15, "learning_rate": 1.92271435226024e-05, "loss": 2.3266, "step": 6175 }, { "epoch": 0.15, "learning_rate": 1.922683574900164e-05, "loss": 2.2319, "step": 6176 }, { "epoch": 0.15, "learning_rate": 1.9226527916595244e-05, "loss": 2.3023, "step": 6177 }, { "epoch": 0.15, "learning_rate": 1.922622002538518e-05, "loss": 2.2597, "step": 6178 }, { "epoch": 0.15, "learning_rate": 1.9225912075373414e-05, "loss": 2.2467, "step": 6179 }, { "epoch": 0.15, "learning_rate": 1.92256040665619e-05, "loss": 2.2335, "step": 6180 }, { "epoch": 0.15, "learning_rate": 1.9225295998952606e-05, "loss": 2.4137, "step": 6181 }, { "epoch": 0.15, "learning_rate": 1.9224987872547495e-05, "loss": 2.2367, "step": 6182 }, { "epoch": 0.15, "learning_rate": 1.9224679687348526e-05, "loss": 2.2672, "step": 6183 }, { "epoch": 0.15, "learning_rate": 1.922437144335767e-05, "loss": 2.3341, "step": 6184 }, { "epoch": 0.15, "learning_rate": 1.922406314057689e-05, "loss": 2.2871, "step": 6185 }, { "epoch": 0.15, "learning_rate": 1.922375477900815e-05, "loss": 2.3591, "step": 6186 }, { "epoch": 0.15, "learning_rate": 1.922344635865341e-05, "loss": 2.0931, "step": 6187 }, { "epoch": 0.15, "learning_rate": 1.9223137879514645e-05, "loss": 2.2129, "step": 6188 }, { "epoch": 0.15, "learning_rate": 1.9222829341593813e-05, "loss": 2.359, "step": 6189 }, { "epoch": 0.15, "learning_rate": 1.922252074489289e-05, "loss": 2.3884, "step": 6190 }, { "epoch": 0.15, "learning_rate": 1.922221208941383e-05, "loss": 2.3214, "step": 6191 }, { "epoch": 0.15, "learning_rate": 1.922190337515861e-05, "loss": 2.327, "step": 6192 }, { "epoch": 0.15, "learning_rate": 1.9221594602129192e-05, "loss": 2.2555, "step": 6193 }, { "epoch": 0.15, "learning_rate": 1.922128577032755e-05, "loss": 2.2041, "step": 6194 }, { "epoch": 0.15, "learning_rate": 1.9220976879755645e-05, "loss": 2.3122, "step": 6195 }, { "epoch": 0.15, "learning_rate": 1.9220667930415457e-05, "loss": 2.3167, "step": 6196 }, { "epoch": 0.15, "learning_rate": 1.922035892230894e-05, "loss": 2.2841, "step": 6197 }, { "epoch": 0.15, "learning_rate": 1.922004985543807e-05, "loss": 2.2664, "step": 6198 }, { "epoch": 0.15, "learning_rate": 1.9219740729804822e-05, "loss": 2.2569, "step": 6199 }, { "epoch": 0.15, "learning_rate": 1.9219431545411158e-05, "loss": 2.1586, "step": 6200 }, { "epoch": 0.15, "learning_rate": 1.9219122302259056e-05, "loss": 2.23, "step": 6201 }, { "epoch": 0.15, "learning_rate": 1.921881300035048e-05, "loss": 2.2855, "step": 6202 }, { "epoch": 0.15, "learning_rate": 1.9218503639687407e-05, "loss": 2.2695, "step": 6203 }, { "epoch": 0.15, "learning_rate": 1.9218194220271804e-05, "loss": 2.396, "step": 6204 }, { "epoch": 0.15, "learning_rate": 1.9217884742105647e-05, "loss": 2.0273, "step": 6205 }, { "epoch": 0.15, "learning_rate": 1.9217575205190904e-05, "loss": 2.2044, "step": 6206 }, { "epoch": 0.15, "learning_rate": 1.921726560952955e-05, "loss": 2.106, "step": 6207 }, { "epoch": 0.15, "learning_rate": 1.9216955955123563e-05, "loss": 2.4033, "step": 6208 }, { "epoch": 0.15, "learning_rate": 1.9216646241974908e-05, "loss": 2.1572, "step": 6209 }, { "epoch": 0.15, "learning_rate": 1.9216336470085565e-05, "loss": 2.3333, "step": 6210 }, { "epoch": 0.15, "learning_rate": 1.9216026639457503e-05, "loss": 2.2518, "step": 6211 }, { "epoch": 0.15, "learning_rate": 1.92157167500927e-05, "loss": 2.2567, "step": 6212 }, { "epoch": 0.15, "learning_rate": 1.9215406801993135e-05, "loss": 2.3687, "step": 6213 }, { "epoch": 0.15, "learning_rate": 1.9215096795160776e-05, "loss": 2.2359, "step": 6214 }, { "epoch": 0.15, "learning_rate": 1.9214786729597605e-05, "loss": 2.3495, "step": 6215 }, { "epoch": 0.15, "learning_rate": 1.9214476605305592e-05, "loss": 2.2559, "step": 6216 }, { "epoch": 0.15, "learning_rate": 1.921416642228672e-05, "loss": 2.0241, "step": 6217 }, { "epoch": 0.15, "learning_rate": 1.9213856180542957e-05, "loss": 2.291, "step": 6218 }, { "epoch": 0.15, "learning_rate": 1.921354588007629e-05, "loss": 2.2399, "step": 6219 }, { "epoch": 0.15, "learning_rate": 1.9213235520888694e-05, "loss": 2.254, "step": 6220 }, { "epoch": 0.15, "learning_rate": 1.9212925102982142e-05, "loss": 2.3337, "step": 6221 }, { "epoch": 0.15, "learning_rate": 1.9212614626358616e-05, "loss": 2.2068, "step": 6222 }, { "epoch": 0.15, "learning_rate": 1.9212304091020097e-05, "loss": 2.3071, "step": 6223 }, { "epoch": 0.15, "learning_rate": 1.921199349696856e-05, "loss": 2.0125, "step": 6224 }, { "epoch": 0.15, "learning_rate": 1.9211682844205987e-05, "loss": 2.1357, "step": 6225 }, { "epoch": 0.15, "learning_rate": 1.921137213273436e-05, "loss": 2.3095, "step": 6226 }, { "epoch": 0.15, "learning_rate": 1.921106136255565e-05, "loss": 2.1853, "step": 6227 }, { "epoch": 0.15, "learning_rate": 1.921075053367185e-05, "loss": 2.2279, "step": 6228 }, { "epoch": 0.15, "learning_rate": 1.9210439646084933e-05, "loss": 2.2927, "step": 6229 }, { "epoch": 0.15, "learning_rate": 1.921012869979688e-05, "loss": 2.2266, "step": 6230 }, { "epoch": 0.15, "learning_rate": 1.920981769480968e-05, "loss": 2.1569, "step": 6231 }, { "epoch": 0.15, "learning_rate": 1.920950663112531e-05, "loss": 2.1507, "step": 6232 }, { "epoch": 0.15, "learning_rate": 1.920919550874575e-05, "loss": 2.378, "step": 6233 }, { "epoch": 0.15, "learning_rate": 1.9208884327672986e-05, "loss": 2.1573, "step": 6234 }, { "epoch": 0.15, "learning_rate": 1.9208573087909003e-05, "loss": 2.1812, "step": 6235 }, { "epoch": 0.15, "learning_rate": 1.920826178945578e-05, "loss": 2.1935, "step": 6236 }, { "epoch": 0.15, "learning_rate": 1.920795043231531e-05, "loss": 2.1533, "step": 6237 }, { "epoch": 0.15, "learning_rate": 1.9207639016489565e-05, "loss": 2.3761, "step": 6238 }, { "epoch": 0.15, "learning_rate": 1.9207327541980535e-05, "loss": 2.144, "step": 6239 }, { "epoch": 0.15, "learning_rate": 1.920701600879021e-05, "loss": 2.1383, "step": 6240 }, { "epoch": 0.15, "learning_rate": 1.9206704416920572e-05, "loss": 2.3255, "step": 6241 }, { "epoch": 0.15, "learning_rate": 1.9206392766373602e-05, "loss": 2.203, "step": 6242 }, { "epoch": 0.15, "learning_rate": 1.9206081057151292e-05, "loss": 2.2182, "step": 6243 }, { "epoch": 0.15, "learning_rate": 1.9205769289255633e-05, "loss": 2.1559, "step": 6244 }, { "epoch": 0.15, "learning_rate": 1.92054574626886e-05, "loss": 2.2712, "step": 6245 }, { "epoch": 0.15, "learning_rate": 1.920514557745219e-05, "loss": 2.3973, "step": 6246 }, { "epoch": 0.15, "learning_rate": 1.9204833633548383e-05, "loss": 2.2325, "step": 6247 }, { "epoch": 0.15, "learning_rate": 1.9204521630979176e-05, "loss": 2.1822, "step": 6248 }, { "epoch": 0.15, "learning_rate": 1.920420956974655e-05, "loss": 2.2265, "step": 6249 }, { "epoch": 0.15, "learning_rate": 1.92038974498525e-05, "loss": 2.3679, "step": 6250 }, { "epoch": 0.15, "learning_rate": 1.9203585271299008e-05, "loss": 2.385, "step": 6251 }, { "epoch": 0.15, "learning_rate": 1.920327303408807e-05, "loss": 2.2274, "step": 6252 }, { "epoch": 0.15, "learning_rate": 1.920296073822167e-05, "loss": 2.0642, "step": 6253 }, { "epoch": 0.15, "learning_rate": 1.920264838370181e-05, "loss": 2.2294, "step": 6254 }, { "epoch": 0.15, "learning_rate": 1.920233597053047e-05, "loss": 2.1362, "step": 6255 }, { "epoch": 0.15, "learning_rate": 1.9202023498709638e-05, "loss": 2.2555, "step": 6256 }, { "epoch": 0.15, "learning_rate": 1.9201710968241316e-05, "loss": 2.1331, "step": 6257 }, { "epoch": 0.15, "learning_rate": 1.9201398379127492e-05, "loss": 2.2572, "step": 6258 }, { "epoch": 0.15, "learning_rate": 1.920108573137016e-05, "loss": 2.2976, "step": 6259 }, { "epoch": 0.15, "learning_rate": 1.92007730249713e-05, "loss": 2.3611, "step": 6260 }, { "epoch": 0.15, "learning_rate": 1.920046025993292e-05, "loss": 2.358, "step": 6261 }, { "epoch": 0.15, "learning_rate": 1.9200147436257013e-05, "loss": 2.1659, "step": 6262 }, { "epoch": 0.15, "learning_rate": 1.9199834553945566e-05, "loss": 2.4942, "step": 6263 }, { "epoch": 0.15, "learning_rate": 1.9199521613000572e-05, "loss": 2.3774, "step": 6264 }, { "epoch": 0.15, "learning_rate": 1.919920861342403e-05, "loss": 2.3064, "step": 6265 }, { "epoch": 0.15, "learning_rate": 1.9198895555217937e-05, "loss": 2.0724, "step": 6266 }, { "epoch": 0.15, "learning_rate": 1.9198582438384284e-05, "loss": 2.1155, "step": 6267 }, { "epoch": 0.15, "learning_rate": 1.9198269262925065e-05, "loss": 2.2969, "step": 6268 }, { "epoch": 0.15, "learning_rate": 1.9197956028842276e-05, "loss": 2.3717, "step": 6269 }, { "epoch": 0.15, "learning_rate": 1.919764273613792e-05, "loss": 2.3617, "step": 6270 }, { "epoch": 0.15, "learning_rate": 1.919732938481399e-05, "loss": 2.3145, "step": 6271 }, { "epoch": 0.15, "learning_rate": 1.919701597487248e-05, "loss": 2.3675, "step": 6272 }, { "epoch": 0.15, "learning_rate": 1.919670250631539e-05, "loss": 2.2934, "step": 6273 }, { "epoch": 0.15, "learning_rate": 1.919638897914472e-05, "loss": 2.1347, "step": 6274 }, { "epoch": 0.15, "learning_rate": 1.9196075393362462e-05, "loss": 2.284, "step": 6275 }, { "epoch": 0.15, "learning_rate": 1.919576174897062e-05, "loss": 2.2194, "step": 6276 }, { "epoch": 0.15, "learning_rate": 1.9195448045971196e-05, "loss": 2.1565, "step": 6277 }, { "epoch": 0.15, "learning_rate": 1.919513428436618e-05, "loss": 2.4631, "step": 6278 }, { "epoch": 0.15, "learning_rate": 1.919482046415758e-05, "loss": 2.4418, "step": 6279 }, { "epoch": 0.15, "learning_rate": 1.919450658534739e-05, "loss": 2.2382, "step": 6280 }, { "epoch": 0.15, "learning_rate": 1.9194192647937614e-05, "loss": 2.2452, "step": 6281 }, { "epoch": 0.15, "learning_rate": 1.919387865193025e-05, "loss": 2.3528, "step": 6282 }, { "epoch": 0.15, "learning_rate": 1.9193564597327303e-05, "loss": 2.1896, "step": 6283 }, { "epoch": 0.15, "learning_rate": 1.9193250484130773e-05, "loss": 2.0517, "step": 6284 }, { "epoch": 0.15, "learning_rate": 1.9192936312342663e-05, "loss": 2.1216, "step": 6285 }, { "epoch": 0.15, "learning_rate": 1.9192622081964973e-05, "loss": 2.1273, "step": 6286 }, { "epoch": 0.15, "learning_rate": 1.9192307792999708e-05, "loss": 2.2115, "step": 6287 }, { "epoch": 0.15, "learning_rate": 1.9191993445448863e-05, "loss": 2.163, "step": 6288 }, { "epoch": 0.16, "learning_rate": 1.9191679039314455e-05, "loss": 2.3864, "step": 6289 }, { "epoch": 0.16, "learning_rate": 1.919136457459848e-05, "loss": 2.2503, "step": 6290 }, { "epoch": 0.16, "learning_rate": 1.9191050051302942e-05, "loss": 2.3956, "step": 6291 }, { "epoch": 0.16, "learning_rate": 1.9190735469429845e-05, "loss": 2.2929, "step": 6292 }, { "epoch": 0.16, "learning_rate": 1.9190420828981198e-05, "loss": 2.1782, "step": 6293 }, { "epoch": 0.16, "learning_rate": 1.9190106129959006e-05, "loss": 2.2615, "step": 6294 }, { "epoch": 0.16, "learning_rate": 1.918979137236527e-05, "loss": 2.1862, "step": 6295 }, { "epoch": 0.16, "learning_rate": 1.9189476556201997e-05, "loss": 2.2895, "step": 6296 }, { "epoch": 0.16, "learning_rate": 1.91891616814712e-05, "loss": 2.3794, "step": 6297 }, { "epoch": 0.16, "learning_rate": 1.9188846748174876e-05, "loss": 2.2008, "step": 6298 }, { "epoch": 0.16, "learning_rate": 1.918853175631504e-05, "loss": 2.2664, "step": 6299 }, { "epoch": 0.16, "learning_rate": 1.9188216705893698e-05, "loss": 2.3429, "step": 6300 }, { "epoch": 0.16, "learning_rate": 1.918790159691285e-05, "loss": 2.2495, "step": 6301 }, { "epoch": 0.16, "learning_rate": 1.9187586429374517e-05, "loss": 2.4021, "step": 6302 }, { "epoch": 0.16, "learning_rate": 1.91872712032807e-05, "loss": 2.3171, "step": 6303 }, { "epoch": 0.16, "learning_rate": 1.918695591863341e-05, "loss": 2.1965, "step": 6304 }, { "epoch": 0.16, "learning_rate": 1.9186640575434655e-05, "loss": 2.3792, "step": 6305 }, { "epoch": 0.16, "learning_rate": 1.9186325173686445e-05, "loss": 2.2875, "step": 6306 }, { "epoch": 0.16, "learning_rate": 1.9186009713390793e-05, "loss": 2.224, "step": 6307 }, { "epoch": 0.16, "learning_rate": 1.9185694194549707e-05, "loss": 2.3099, "step": 6308 }, { "epoch": 0.16, "learning_rate": 1.9185378617165198e-05, "loss": 2.5091, "step": 6309 }, { "epoch": 0.16, "learning_rate": 1.9185062981239276e-05, "loss": 2.2486, "step": 6310 }, { "epoch": 0.16, "learning_rate": 1.9184747286773956e-05, "loss": 2.395, "step": 6311 }, { "epoch": 0.16, "learning_rate": 1.918443153377125e-05, "loss": 2.169, "step": 6312 }, { "epoch": 0.16, "learning_rate": 1.918411572223317e-05, "loss": 2.311, "step": 6313 }, { "epoch": 0.16, "learning_rate": 1.9183799852161723e-05, "loss": 2.1971, "step": 6314 }, { "epoch": 0.16, "learning_rate": 1.918348392355893e-05, "loss": 2.2233, "step": 6315 }, { "epoch": 0.16, "learning_rate": 1.9183167936426798e-05, "loss": 2.2451, "step": 6316 }, { "epoch": 0.16, "learning_rate": 1.9182851890767344e-05, "loss": 2.1859, "step": 6317 }, { "epoch": 0.16, "learning_rate": 1.9182535786582585e-05, "loss": 2.4185, "step": 6318 }, { "epoch": 0.16, "learning_rate": 1.918221962387453e-05, "loss": 2.3032, "step": 6319 }, { "epoch": 0.16, "learning_rate": 1.9181903402645197e-05, "loss": 2.2554, "step": 6320 }, { "epoch": 0.16, "learning_rate": 1.9181587122896604e-05, "loss": 2.3201, "step": 6321 }, { "epoch": 0.16, "learning_rate": 1.9181270784630763e-05, "loss": 2.2297, "step": 6322 }, { "epoch": 0.16, "learning_rate": 1.918095438784969e-05, "loss": 2.3967, "step": 6323 }, { "epoch": 0.16, "learning_rate": 1.9180637932555406e-05, "loss": 2.3409, "step": 6324 }, { "epoch": 0.16, "learning_rate": 1.9180321418749922e-05, "loss": 2.3528, "step": 6325 }, { "epoch": 0.16, "learning_rate": 1.9180004846435257e-05, "loss": 2.2481, "step": 6326 }, { "epoch": 0.16, "learning_rate": 1.917968821561343e-05, "loss": 2.4347, "step": 6327 }, { "epoch": 0.16, "learning_rate": 1.9179371526286458e-05, "loss": 2.1059, "step": 6328 }, { "epoch": 0.16, "learning_rate": 1.9179054778456362e-05, "loss": 2.0502, "step": 6329 }, { "epoch": 0.16, "learning_rate": 1.9178737972125156e-05, "loss": 2.4418, "step": 6330 }, { "epoch": 0.16, "learning_rate": 1.9178421107294863e-05, "loss": 2.2939, "step": 6331 }, { "epoch": 0.16, "learning_rate": 1.9178104183967502e-05, "loss": 2.1687, "step": 6332 }, { "epoch": 0.16, "learning_rate": 1.9177787202145093e-05, "loss": 2.3007, "step": 6333 }, { "epoch": 0.16, "learning_rate": 1.917747016182965e-05, "loss": 2.183, "step": 6334 }, { "epoch": 0.16, "learning_rate": 1.9177153063023203e-05, "loss": 2.1231, "step": 6335 }, { "epoch": 0.16, "learning_rate": 1.9176835905727768e-05, "loss": 2.4206, "step": 6336 }, { "epoch": 0.16, "learning_rate": 1.9176518689945365e-05, "loss": 2.3527, "step": 6337 }, { "epoch": 0.16, "learning_rate": 1.917620141567802e-05, "loss": 2.3036, "step": 6338 }, { "epoch": 0.16, "learning_rate": 1.9175884082927754e-05, "loss": 2.3203, "step": 6339 }, { "epoch": 0.16, "learning_rate": 1.9175566691696588e-05, "loss": 2.2215, "step": 6340 }, { "epoch": 0.16, "learning_rate": 1.9175249241986545e-05, "loss": 2.1782, "step": 6341 }, { "epoch": 0.16, "learning_rate": 1.9174931733799648e-05, "loss": 2.3006, "step": 6342 }, { "epoch": 0.16, "learning_rate": 1.917461416713792e-05, "loss": 2.2287, "step": 6343 }, { "epoch": 0.16, "learning_rate": 1.917429654200339e-05, "loss": 2.1672, "step": 6344 }, { "epoch": 0.16, "learning_rate": 1.9173978858398072e-05, "loss": 2.3461, "step": 6345 }, { "epoch": 0.16, "learning_rate": 1.9173661116324e-05, "loss": 2.32, "step": 6346 }, { "epoch": 0.16, "learning_rate": 1.9173343315783196e-05, "loss": 2.3784, "step": 6347 }, { "epoch": 0.16, "learning_rate": 1.9173025456777688e-05, "loss": 2.3096, "step": 6348 }, { "epoch": 0.16, "learning_rate": 1.9172707539309495e-05, "loss": 2.2838, "step": 6349 }, { "epoch": 0.16, "learning_rate": 1.917238956338065e-05, "loss": 2.2189, "step": 6350 }, { "epoch": 0.16, "learning_rate": 1.9172071528993176e-05, "loss": 2.1185, "step": 6351 }, { "epoch": 0.16, "learning_rate": 1.9171753436149102e-05, "loss": 2.257, "step": 6352 }, { "epoch": 0.16, "learning_rate": 1.9171435284850455e-05, "loss": 2.2841, "step": 6353 }, { "epoch": 0.16, "learning_rate": 1.9171117075099262e-05, "loss": 2.2433, "step": 6354 }, { "epoch": 0.16, "learning_rate": 1.917079880689755e-05, "loss": 2.2278, "step": 6355 }, { "epoch": 0.16, "learning_rate": 1.917048048024735e-05, "loss": 2.2748, "step": 6356 }, { "epoch": 0.16, "learning_rate": 1.9170162095150688e-05, "loss": 2.3404, "step": 6357 }, { "epoch": 0.16, "learning_rate": 1.9169843651609593e-05, "loss": 2.3604, "step": 6358 }, { "epoch": 0.16, "learning_rate": 1.91695251496261e-05, "loss": 2.2777, "step": 6359 }, { "epoch": 0.16, "learning_rate": 1.9169206589202233e-05, "loss": 2.3273, "step": 6360 }, { "epoch": 0.16, "learning_rate": 1.9168887970340022e-05, "loss": 2.2874, "step": 6361 }, { "epoch": 0.16, "learning_rate": 1.9168569293041504e-05, "loss": 2.2482, "step": 6362 }, { "epoch": 0.16, "learning_rate": 1.9168250557308703e-05, "loss": 2.2827, "step": 6363 }, { "epoch": 0.16, "learning_rate": 1.9167931763143656e-05, "loss": 2.3589, "step": 6364 }, { "epoch": 0.16, "learning_rate": 1.916761291054839e-05, "loss": 2.1889, "step": 6365 }, { "epoch": 0.16, "learning_rate": 1.9167293999524935e-05, "loss": 2.2193, "step": 6366 }, { "epoch": 0.16, "learning_rate": 1.9166975030075338e-05, "loss": 2.2865, "step": 6367 }, { "epoch": 0.16, "learning_rate": 1.9166656002201615e-05, "loss": 2.2451, "step": 6368 }, { "epoch": 0.16, "learning_rate": 1.9166336915905804e-05, "loss": 2.0973, "step": 6369 }, { "epoch": 0.16, "learning_rate": 1.9166017771189945e-05, "loss": 2.2492, "step": 6370 }, { "epoch": 0.16, "learning_rate": 1.9165698568056066e-05, "loss": 2.327, "step": 6371 }, { "epoch": 0.16, "learning_rate": 1.9165379306506204e-05, "loss": 2.1831, "step": 6372 }, { "epoch": 0.16, "learning_rate": 1.9165059986542392e-05, "loss": 2.36, "step": 6373 }, { "epoch": 0.16, "learning_rate": 1.9164740608166667e-05, "loss": 2.2141, "step": 6374 }, { "epoch": 0.16, "learning_rate": 1.9164421171381064e-05, "loss": 2.2513, "step": 6375 }, { "epoch": 0.16, "learning_rate": 1.9164101676187616e-05, "loss": 2.146, "step": 6376 }, { "epoch": 0.16, "learning_rate": 1.9163782122588363e-05, "loss": 2.2492, "step": 6377 }, { "epoch": 0.16, "learning_rate": 1.9163462510585335e-05, "loss": 2.2234, "step": 6378 }, { "epoch": 0.16, "learning_rate": 1.916314284018058e-05, "loss": 2.3229, "step": 6379 }, { "epoch": 0.16, "learning_rate": 1.916282311137613e-05, "loss": 2.0792, "step": 6380 }, { "epoch": 0.16, "learning_rate": 1.916250332417402e-05, "loss": 2.3495, "step": 6381 }, { "epoch": 0.16, "learning_rate": 1.916218347857629e-05, "loss": 2.1685, "step": 6382 }, { "epoch": 0.16, "learning_rate": 1.916186357458498e-05, "loss": 2.4205, "step": 6383 }, { "epoch": 0.16, "learning_rate": 1.9161543612202126e-05, "loss": 2.2136, "step": 6384 }, { "epoch": 0.16, "learning_rate": 1.916122359142977e-05, "loss": 2.3483, "step": 6385 }, { "epoch": 0.16, "learning_rate": 1.916090351226995e-05, "loss": 2.2188, "step": 6386 }, { "epoch": 0.16, "learning_rate": 1.9160583374724707e-05, "loss": 2.244, "step": 6387 }, { "epoch": 0.16, "learning_rate": 1.916026317879608e-05, "loss": 2.1447, "step": 6388 }, { "epoch": 0.16, "learning_rate": 1.9159942924486112e-05, "loss": 2.1688, "step": 6389 }, { "epoch": 0.16, "learning_rate": 1.9159622611796843e-05, "loss": 2.2329, "step": 6390 }, { "epoch": 0.16, "learning_rate": 1.9159302240730315e-05, "loss": 2.1965, "step": 6391 }, { "epoch": 0.16, "learning_rate": 1.9158981811288566e-05, "loss": 2.1298, "step": 6392 }, { "epoch": 0.16, "learning_rate": 1.915866132347364e-05, "loss": 2.3946, "step": 6393 }, { "epoch": 0.16, "learning_rate": 1.9158340777287582e-05, "loss": 2.2333, "step": 6394 }, { "epoch": 0.16, "learning_rate": 1.9158020172732434e-05, "loss": 2.2342, "step": 6395 }, { "epoch": 0.16, "learning_rate": 1.915769950981024e-05, "loss": 2.3215, "step": 6396 }, { "epoch": 0.16, "learning_rate": 1.9157378788523042e-05, "loss": 2.2403, "step": 6397 }, { "epoch": 0.16, "learning_rate": 1.9157058008872885e-05, "loss": 2.2363, "step": 6398 }, { "epoch": 0.16, "learning_rate": 1.9156737170861813e-05, "loss": 2.1753, "step": 6399 }, { "epoch": 0.16, "learning_rate": 1.915641627449187e-05, "loss": 2.318, "step": 6400 }, { "epoch": 0.16, "learning_rate": 1.9156095319765103e-05, "loss": 2.2349, "step": 6401 }, { "epoch": 0.16, "learning_rate": 1.9155774306683555e-05, "loss": 2.4392, "step": 6402 }, { "epoch": 0.16, "learning_rate": 1.9155453235249274e-05, "loss": 2.2702, "step": 6403 }, { "epoch": 0.16, "learning_rate": 1.9155132105464306e-05, "loss": 2.3235, "step": 6404 }, { "epoch": 0.16, "learning_rate": 1.91548109173307e-05, "loss": 2.1708, "step": 6405 }, { "epoch": 0.16, "learning_rate": 1.91544896708505e-05, "loss": 2.2206, "step": 6406 }, { "epoch": 0.16, "learning_rate": 1.9154168366025748e-05, "loss": 2.3576, "step": 6407 }, { "epoch": 0.16, "learning_rate": 1.9153847002858504e-05, "loss": 2.111, "step": 6408 }, { "epoch": 0.16, "learning_rate": 1.9153525581350808e-05, "loss": 2.036, "step": 6409 }, { "epoch": 0.16, "learning_rate": 1.915320410150471e-05, "loss": 2.2723, "step": 6410 }, { "epoch": 0.16, "learning_rate": 1.9152882563322262e-05, "loss": 2.3795, "step": 6411 }, { "epoch": 0.16, "learning_rate": 1.9152560966805507e-05, "loss": 2.0997, "step": 6412 }, { "epoch": 0.16, "learning_rate": 1.91522393119565e-05, "loss": 2.544, "step": 6413 }, { "epoch": 0.16, "learning_rate": 1.915191759877729e-05, "loss": 2.2869, "step": 6414 }, { "epoch": 0.16, "learning_rate": 1.9151595827269928e-05, "loss": 2.3588, "step": 6415 }, { "epoch": 0.16, "learning_rate": 1.9151273997436458e-05, "loss": 2.3336, "step": 6416 }, { "epoch": 0.16, "learning_rate": 1.9150952109278938e-05, "loss": 2.4282, "step": 6417 }, { "epoch": 0.16, "learning_rate": 1.9150630162799425e-05, "loss": 2.4027, "step": 6418 }, { "epoch": 0.16, "learning_rate": 1.9150308157999956e-05, "loss": 2.1874, "step": 6419 }, { "epoch": 0.16, "learning_rate": 1.9149986094882594e-05, "loss": 2.1608, "step": 6420 }, { "epoch": 0.16, "learning_rate": 1.9149663973449385e-05, "loss": 2.2041, "step": 6421 }, { "epoch": 0.16, "learning_rate": 1.9149341793702393e-05, "loss": 2.1933, "step": 6422 }, { "epoch": 0.16, "learning_rate": 1.914901955564366e-05, "loss": 2.1444, "step": 6423 }, { "epoch": 0.16, "learning_rate": 1.914869725927524e-05, "loss": 2.3356, "step": 6424 }, { "epoch": 0.16, "learning_rate": 1.9148374904599197e-05, "loss": 2.151, "step": 6425 }, { "epoch": 0.16, "learning_rate": 1.9148052491617576e-05, "loss": 2.3982, "step": 6426 }, { "epoch": 0.16, "learning_rate": 1.9147730020332438e-05, "loss": 2.2702, "step": 6427 }, { "epoch": 0.16, "learning_rate": 1.9147407490745833e-05, "loss": 2.3389, "step": 6428 }, { "epoch": 0.16, "learning_rate": 1.914708490285982e-05, "loss": 2.192, "step": 6429 }, { "epoch": 0.16, "learning_rate": 1.9146762256676452e-05, "loss": 2.2619, "step": 6430 }, { "epoch": 0.16, "learning_rate": 1.9146439552197787e-05, "loss": 2.0572, "step": 6431 }, { "epoch": 0.16, "learning_rate": 1.9146116789425888e-05, "loss": 2.2669, "step": 6432 }, { "epoch": 0.16, "learning_rate": 1.9145793968362804e-05, "loss": 2.2981, "step": 6433 }, { "epoch": 0.16, "learning_rate": 1.9145471089010588e-05, "loss": 2.2798, "step": 6434 }, { "epoch": 0.16, "learning_rate": 1.9145148151371312e-05, "loss": 2.2367, "step": 6435 }, { "epoch": 0.16, "learning_rate": 1.914482515544702e-05, "loss": 2.0508, "step": 6436 }, { "epoch": 0.16, "learning_rate": 1.9144502101239785e-05, "loss": 2.2774, "step": 6437 }, { "epoch": 0.16, "learning_rate": 1.914417898875165e-05, "loss": 2.1886, "step": 6438 }, { "epoch": 0.16, "learning_rate": 1.9143855817984686e-05, "loss": 2.2167, "step": 6439 }, { "epoch": 0.16, "learning_rate": 1.914353258894095e-05, "loss": 2.2288, "step": 6440 }, { "epoch": 0.16, "learning_rate": 1.91432093016225e-05, "loss": 2.2269, "step": 6441 }, { "epoch": 0.16, "learning_rate": 1.9142885956031395e-05, "loss": 2.3066, "step": 6442 }, { "epoch": 0.16, "learning_rate": 1.91425625521697e-05, "loss": 2.3838, "step": 6443 }, { "epoch": 0.16, "learning_rate": 1.9142239090039475e-05, "loss": 2.2175, "step": 6444 }, { "epoch": 0.16, "learning_rate": 1.914191556964278e-05, "loss": 2.2053, "step": 6445 }, { "epoch": 0.16, "learning_rate": 1.9141591990981682e-05, "loss": 2.3918, "step": 6446 }, { "epoch": 0.16, "learning_rate": 1.9141268354058232e-05, "loss": 2.1971, "step": 6447 }, { "epoch": 0.16, "learning_rate": 1.9140944658874504e-05, "loss": 2.2767, "step": 6448 }, { "epoch": 0.16, "learning_rate": 1.9140620905432558e-05, "loss": 2.3737, "step": 6449 }, { "epoch": 0.16, "learning_rate": 1.9140297093734454e-05, "loss": 2.3006, "step": 6450 }, { "epoch": 0.16, "learning_rate": 1.913997322378226e-05, "loss": 2.1782, "step": 6451 }, { "epoch": 0.16, "learning_rate": 1.913964929557803e-05, "loss": 2.328, "step": 6452 }, { "epoch": 0.16, "learning_rate": 1.9139325309123847e-05, "loss": 2.3231, "step": 6453 }, { "epoch": 0.16, "learning_rate": 1.913900126442176e-05, "loss": 2.1765, "step": 6454 }, { "epoch": 0.16, "learning_rate": 1.913867716147384e-05, "loss": 2.2473, "step": 6455 }, { "epoch": 0.16, "learning_rate": 1.9138353000282154e-05, "loss": 2.1802, "step": 6456 }, { "epoch": 0.16, "learning_rate": 1.9138028780848766e-05, "loss": 2.2579, "step": 6457 }, { "epoch": 0.16, "learning_rate": 1.913770450317574e-05, "loss": 2.0586, "step": 6458 }, { "epoch": 0.16, "learning_rate": 1.9137380167265147e-05, "loss": 2.1313, "step": 6459 }, { "epoch": 0.16, "learning_rate": 1.9137055773119054e-05, "loss": 2.3926, "step": 6460 }, { "epoch": 0.16, "learning_rate": 1.9136731320739526e-05, "loss": 2.2061, "step": 6461 }, { "epoch": 0.16, "learning_rate": 1.9136406810128628e-05, "loss": 2.4195, "step": 6462 }, { "epoch": 0.16, "learning_rate": 1.9136082241288437e-05, "loss": 2.2189, "step": 6463 }, { "epoch": 0.16, "learning_rate": 1.913575761422101e-05, "loss": 2.1776, "step": 6464 }, { "epoch": 0.16, "learning_rate": 1.913543292892843e-05, "loss": 2.2358, "step": 6465 }, { "epoch": 0.16, "learning_rate": 1.9135108185412755e-05, "loss": 2.1843, "step": 6466 }, { "epoch": 0.16, "learning_rate": 1.913478338367606e-05, "loss": 2.2083, "step": 6467 }, { "epoch": 0.16, "learning_rate": 1.9134458523720413e-05, "loss": 2.1508, "step": 6468 }, { "epoch": 0.16, "learning_rate": 1.9134133605547883e-05, "loss": 2.4099, "step": 6469 }, { "epoch": 0.16, "learning_rate": 1.9133808629160547e-05, "loss": 2.1808, "step": 6470 }, { "epoch": 0.16, "learning_rate": 1.913348359456047e-05, "loss": 2.2471, "step": 6471 }, { "epoch": 0.16, "learning_rate": 1.913315850174972e-05, "loss": 2.2206, "step": 6472 }, { "epoch": 0.16, "learning_rate": 1.913283335073038e-05, "loss": 2.2052, "step": 6473 }, { "epoch": 0.16, "learning_rate": 1.913250814150452e-05, "loss": 2.2128, "step": 6474 }, { "epoch": 0.16, "learning_rate": 1.9132182874074203e-05, "loss": 2.1295, "step": 6475 }, { "epoch": 0.16, "learning_rate": 1.9131857548441512e-05, "loss": 2.1788, "step": 6476 }, { "epoch": 0.16, "learning_rate": 1.9131532164608516e-05, "loss": 2.2141, "step": 6477 }, { "epoch": 0.16, "learning_rate": 1.9131206722577292e-05, "loss": 2.1111, "step": 6478 }, { "epoch": 0.16, "learning_rate": 1.9130881222349908e-05, "loss": 2.4105, "step": 6479 }, { "epoch": 0.16, "learning_rate": 1.9130555663928442e-05, "loss": 2.4067, "step": 6480 }, { "epoch": 0.16, "learning_rate": 1.9130230047314973e-05, "loss": 2.2495, "step": 6481 }, { "epoch": 0.16, "learning_rate": 1.912990437251157e-05, "loss": 2.4652, "step": 6482 }, { "epoch": 0.16, "learning_rate": 1.912957863952031e-05, "loss": 2.1256, "step": 6483 }, { "epoch": 0.16, "learning_rate": 1.9129252848343275e-05, "loss": 2.2327, "step": 6484 }, { "epoch": 0.16, "learning_rate": 1.912892699898253e-05, "loss": 2.2733, "step": 6485 }, { "epoch": 0.16, "learning_rate": 1.9128601091440162e-05, "loss": 2.2836, "step": 6486 }, { "epoch": 0.16, "learning_rate": 1.9128275125718242e-05, "loss": 2.1804, "step": 6487 }, { "epoch": 0.16, "learning_rate": 1.9127949101818853e-05, "loss": 2.366, "step": 6488 }, { "epoch": 0.16, "learning_rate": 1.912762301974407e-05, "loss": 2.3261, "step": 6489 }, { "epoch": 0.16, "learning_rate": 1.9127296879495967e-05, "loss": 2.206, "step": 6490 }, { "epoch": 0.16, "learning_rate": 1.9126970681076628e-05, "loss": 2.1143, "step": 6491 }, { "epoch": 0.16, "learning_rate": 1.912664442448813e-05, "loss": 2.2327, "step": 6492 }, { "epoch": 0.16, "learning_rate": 1.9126318109732555e-05, "loss": 2.3099, "step": 6493 }, { "epoch": 0.16, "learning_rate": 1.9125991736811977e-05, "loss": 2.1392, "step": 6494 }, { "epoch": 0.16, "learning_rate": 1.9125665305728482e-05, "loss": 2.3656, "step": 6495 }, { "epoch": 0.16, "learning_rate": 1.9125338816484147e-05, "loss": 2.1447, "step": 6496 }, { "epoch": 0.16, "learning_rate": 1.9125012269081056e-05, "loss": 2.1708, "step": 6497 }, { "epoch": 0.16, "learning_rate": 1.9124685663521286e-05, "loss": 2.3703, "step": 6498 }, { "epoch": 0.16, "learning_rate": 1.912435899980692e-05, "loss": 2.0944, "step": 6499 }, { "epoch": 0.16, "learning_rate": 1.9124032277940044e-05, "loss": 2.2649, "step": 6500 }, { "epoch": 0.16, "learning_rate": 1.9123705497922733e-05, "loss": 2.2987, "step": 6501 }, { "epoch": 0.16, "learning_rate": 1.9123378659757077e-05, "loss": 2.2638, "step": 6502 }, { "epoch": 0.16, "learning_rate": 1.9123051763445153e-05, "loss": 2.1719, "step": 6503 }, { "epoch": 0.16, "learning_rate": 1.9122724808989047e-05, "loss": 2.1332, "step": 6504 }, { "epoch": 0.16, "learning_rate": 1.9122397796390847e-05, "loss": 2.1809, "step": 6505 }, { "epoch": 0.16, "learning_rate": 1.9122070725652627e-05, "loss": 2.2395, "step": 6506 }, { "epoch": 0.16, "learning_rate": 1.912174359677648e-05, "loss": 2.3618, "step": 6507 }, { "epoch": 0.16, "learning_rate": 1.9121416409764487e-05, "loss": 2.1235, "step": 6508 }, { "epoch": 0.16, "learning_rate": 1.9121089164618735e-05, "loss": 2.1266, "step": 6509 }, { "epoch": 0.16, "learning_rate": 1.9120761861341314e-05, "loss": 2.169, "step": 6510 }, { "epoch": 0.16, "learning_rate": 1.91204344999343e-05, "loss": 2.2137, "step": 6511 }, { "epoch": 0.16, "learning_rate": 1.9120107080399787e-05, "loss": 1.9948, "step": 6512 }, { "epoch": 0.16, "learning_rate": 1.9119779602739855e-05, "loss": 2.3428, "step": 6513 }, { "epoch": 0.16, "learning_rate": 1.91194520669566e-05, "loss": 2.1532, "step": 6514 }, { "epoch": 0.16, "learning_rate": 1.9119124473052106e-05, "loss": 2.1566, "step": 6515 }, { "epoch": 0.16, "learning_rate": 1.911879682102846e-05, "loss": 2.2408, "step": 6516 }, { "epoch": 0.16, "learning_rate": 1.9118469110887745e-05, "loss": 2.1854, "step": 6517 }, { "epoch": 0.16, "learning_rate": 1.911814134263206e-05, "loss": 2.3575, "step": 6518 }, { "epoch": 0.16, "learning_rate": 1.9117813516263485e-05, "loss": 2.3095, "step": 6519 }, { "epoch": 0.16, "learning_rate": 1.9117485631784115e-05, "loss": 2.2834, "step": 6520 }, { "epoch": 0.16, "learning_rate": 1.9117157689196038e-05, "loss": 2.2698, "step": 6521 }, { "epoch": 0.16, "learning_rate": 1.9116829688501345e-05, "loss": 2.1421, "step": 6522 }, { "epoch": 0.16, "learning_rate": 1.911650162970212e-05, "loss": 2.2834, "step": 6523 }, { "epoch": 0.16, "learning_rate": 1.9116173512800463e-05, "loss": 2.3994, "step": 6524 }, { "epoch": 0.16, "learning_rate": 1.911584533779846e-05, "loss": 2.1525, "step": 6525 }, { "epoch": 0.16, "learning_rate": 1.9115517104698205e-05, "loss": 2.2024, "step": 6526 }, { "epoch": 0.16, "learning_rate": 1.911518881350179e-05, "loss": 2.3037, "step": 6527 }, { "epoch": 0.16, "learning_rate": 1.9114860464211307e-05, "loss": 2.2491, "step": 6528 }, { "epoch": 0.16, "learning_rate": 1.9114532056828845e-05, "loss": 2.369, "step": 6529 }, { "epoch": 0.16, "learning_rate": 1.9114203591356498e-05, "loss": 2.2496, "step": 6530 }, { "epoch": 0.16, "learning_rate": 1.9113875067796366e-05, "loss": 2.1556, "step": 6531 }, { "epoch": 0.16, "learning_rate": 1.9113546486150537e-05, "loss": 2.2952, "step": 6532 }, { "epoch": 0.16, "learning_rate": 1.9113217846421103e-05, "loss": 2.2927, "step": 6533 }, { "epoch": 0.16, "learning_rate": 1.9112889148610167e-05, "loss": 2.3183, "step": 6534 }, { "epoch": 0.16, "learning_rate": 1.9112560392719814e-05, "loss": 2.1535, "step": 6535 }, { "epoch": 0.16, "learning_rate": 1.9112231578752145e-05, "loss": 2.4459, "step": 6536 }, { "epoch": 0.16, "learning_rate": 1.9111902706709256e-05, "loss": 2.2035, "step": 6537 }, { "epoch": 0.16, "learning_rate": 1.9111573776593238e-05, "loss": 2.1455, "step": 6538 }, { "epoch": 0.16, "learning_rate": 1.9111244788406197e-05, "loss": 2.2738, "step": 6539 }, { "epoch": 0.16, "learning_rate": 1.911091574215022e-05, "loss": 2.2943, "step": 6540 }, { "epoch": 0.16, "learning_rate": 1.9110586637827405e-05, "loss": 2.4071, "step": 6541 }, { "epoch": 0.16, "learning_rate": 1.911025747543986e-05, "loss": 2.1445, "step": 6542 }, { "epoch": 0.16, "learning_rate": 1.9109928254989668e-05, "loss": 2.1093, "step": 6543 }, { "epoch": 0.16, "learning_rate": 1.9109598976478937e-05, "loss": 2.0867, "step": 6544 }, { "epoch": 0.16, "learning_rate": 1.910926963990976e-05, "loss": 2.014, "step": 6545 }, { "epoch": 0.16, "learning_rate": 1.9108940245284243e-05, "loss": 2.2877, "step": 6546 }, { "epoch": 0.16, "learning_rate": 1.910861079260448e-05, "loss": 2.2782, "step": 6547 }, { "epoch": 0.16, "learning_rate": 1.9108281281872574e-05, "loss": 2.3334, "step": 6548 }, { "epoch": 0.16, "learning_rate": 1.910795171309062e-05, "loss": 2.2631, "step": 6549 }, { "epoch": 0.16, "learning_rate": 1.910762208626072e-05, "loss": 2.2717, "step": 6550 }, { "epoch": 0.16, "learning_rate": 1.9107292401384984e-05, "loss": 2.3339, "step": 6551 }, { "epoch": 0.16, "learning_rate": 1.91069626584655e-05, "loss": 2.3263, "step": 6552 }, { "epoch": 0.16, "learning_rate": 1.9106632857504374e-05, "loss": 2.2493, "step": 6553 }, { "epoch": 0.16, "learning_rate": 1.910630299850371e-05, "loss": 2.2469, "step": 6554 }, { "epoch": 0.16, "learning_rate": 1.9105973081465612e-05, "loss": 2.2232, "step": 6555 }, { "epoch": 0.16, "learning_rate": 1.910564310639218e-05, "loss": 2.3281, "step": 6556 }, { "epoch": 0.16, "learning_rate": 1.9105313073285515e-05, "loss": 2.1629, "step": 6557 }, { "epoch": 0.16, "learning_rate": 1.910498298214772e-05, "loss": 2.362, "step": 6558 }, { "epoch": 0.16, "learning_rate": 1.9104652832980907e-05, "loss": 2.309, "step": 6559 }, { "epoch": 0.16, "learning_rate": 1.910432262578717e-05, "loss": 2.1804, "step": 6560 }, { "epoch": 0.16, "learning_rate": 1.9103992360568617e-05, "loss": 2.3234, "step": 6561 }, { "epoch": 0.16, "learning_rate": 1.9103662037327357e-05, "loss": 2.1506, "step": 6562 }, { "epoch": 0.16, "learning_rate": 1.9103331656065492e-05, "loss": 2.1173, "step": 6563 }, { "epoch": 0.16, "learning_rate": 1.9103001216785126e-05, "loss": 2.2498, "step": 6564 }, { "epoch": 0.16, "learning_rate": 1.9102670719488366e-05, "loss": 2.2515, "step": 6565 }, { "epoch": 0.16, "learning_rate": 1.910234016417732e-05, "loss": 2.3459, "step": 6566 }, { "epoch": 0.16, "learning_rate": 1.9102009550854097e-05, "loss": 2.0744, "step": 6567 }, { "epoch": 0.16, "learning_rate": 1.9101678879520795e-05, "loss": 2.3641, "step": 6568 }, { "epoch": 0.16, "learning_rate": 1.9101348150179532e-05, "loss": 2.3416, "step": 6569 }, { "epoch": 0.16, "learning_rate": 1.910101736283241e-05, "loss": 2.2852, "step": 6570 }, { "epoch": 0.16, "learning_rate": 1.9100686517481538e-05, "loss": 2.2128, "step": 6571 }, { "epoch": 0.16, "learning_rate": 1.9100355614129022e-05, "loss": 2.2816, "step": 6572 }, { "epoch": 0.16, "learning_rate": 1.910002465277698e-05, "loss": 2.3648, "step": 6573 }, { "epoch": 0.16, "learning_rate": 1.9099693633427508e-05, "loss": 2.239, "step": 6574 }, { "epoch": 0.16, "learning_rate": 1.909936255608273e-05, "loss": 2.3141, "step": 6575 }, { "epoch": 0.16, "learning_rate": 1.9099031420744744e-05, "loss": 2.3937, "step": 6576 }, { "epoch": 0.16, "learning_rate": 1.909870022741567e-05, "loss": 2.1917, "step": 6577 }, { "epoch": 0.16, "learning_rate": 1.9098368976097613e-05, "loss": 2.0732, "step": 6578 }, { "epoch": 0.16, "learning_rate": 1.9098037666792685e-05, "loss": 2.1521, "step": 6579 }, { "epoch": 0.16, "learning_rate": 1.9097706299502995e-05, "loss": 2.2852, "step": 6580 }, { "epoch": 0.16, "learning_rate": 1.909737487423066e-05, "loss": 2.2858, "step": 6581 }, { "epoch": 0.16, "learning_rate": 1.909704339097779e-05, "loss": 2.2491, "step": 6582 }, { "epoch": 0.16, "learning_rate": 1.90967118497465e-05, "loss": 2.105, "step": 6583 }, { "epoch": 0.16, "learning_rate": 1.90963802505389e-05, "loss": 2.2651, "step": 6584 }, { "epoch": 0.16, "learning_rate": 1.9096048593357103e-05, "loss": 2.2528, "step": 6585 }, { "epoch": 0.16, "learning_rate": 1.9095716878203222e-05, "loss": 2.1403, "step": 6586 }, { "epoch": 0.16, "learning_rate": 1.9095385105079376e-05, "loss": 2.1163, "step": 6587 }, { "epoch": 0.16, "learning_rate": 1.9095053273987675e-05, "loss": 2.1591, "step": 6588 }, { "epoch": 0.16, "learning_rate": 1.909472138493024e-05, "loss": 2.1365, "step": 6589 }, { "epoch": 0.16, "learning_rate": 1.9094389437909177e-05, "loss": 2.2474, "step": 6590 }, { "epoch": 0.16, "learning_rate": 1.9094057432926606e-05, "loss": 2.1498, "step": 6591 }, { "epoch": 0.16, "learning_rate": 1.9093725369984647e-05, "loss": 2.2902, "step": 6592 }, { "epoch": 0.16, "learning_rate": 1.9093393249085407e-05, "loss": 2.2295, "step": 6593 }, { "epoch": 0.16, "learning_rate": 1.9093061070231014e-05, "loss": 2.3867, "step": 6594 }, { "epoch": 0.16, "learning_rate": 1.909272883342358e-05, "loss": 2.1889, "step": 6595 }, { "epoch": 0.16, "learning_rate": 1.9092396538665215e-05, "loss": 2.3435, "step": 6596 }, { "epoch": 0.16, "learning_rate": 1.909206418595805e-05, "loss": 2.3488, "step": 6597 }, { "epoch": 0.16, "learning_rate": 1.9091731775304195e-05, "loss": 2.2511, "step": 6598 }, { "epoch": 0.16, "learning_rate": 1.909139930670577e-05, "loss": 2.1521, "step": 6599 }, { "epoch": 0.16, "learning_rate": 1.9091066780164895e-05, "loss": 2.2412, "step": 6600 }, { "epoch": 0.16, "learning_rate": 1.909073419568369e-05, "loss": 2.2091, "step": 6601 }, { "epoch": 0.16, "learning_rate": 1.909040155326427e-05, "loss": 2.2405, "step": 6602 }, { "epoch": 0.16, "learning_rate": 1.9090068852908762e-05, "loss": 2.2114, "step": 6603 }, { "epoch": 0.16, "learning_rate": 1.908973609461928e-05, "loss": 2.2439, "step": 6604 }, { "epoch": 0.16, "learning_rate": 1.908940327839795e-05, "loss": 2.1399, "step": 6605 }, { "epoch": 0.16, "learning_rate": 1.908907040424689e-05, "loss": 2.2426, "step": 6606 }, { "epoch": 0.16, "learning_rate": 1.908873747216822e-05, "loss": 2.2102, "step": 6607 }, { "epoch": 0.16, "learning_rate": 1.9088404482164065e-05, "loss": 2.2113, "step": 6608 }, { "epoch": 0.16, "learning_rate": 1.9088071434236548e-05, "loss": 2.1331, "step": 6609 }, { "epoch": 0.16, "learning_rate": 1.908773832838779e-05, "loss": 2.4256, "step": 6610 }, { "epoch": 0.16, "learning_rate": 1.9087405164619915e-05, "loss": 2.1437, "step": 6611 }, { "epoch": 0.16, "learning_rate": 1.9087071942935044e-05, "loss": 2.3161, "step": 6612 }, { "epoch": 0.16, "learning_rate": 1.90867386633353e-05, "loss": 2.196, "step": 6613 }, { "epoch": 0.16, "learning_rate": 1.908640532582281e-05, "loss": 2.2753, "step": 6614 }, { "epoch": 0.16, "learning_rate": 1.9086071930399703e-05, "loss": 2.2746, "step": 6615 }, { "epoch": 0.16, "learning_rate": 1.9085738477068092e-05, "loss": 2.2323, "step": 6616 }, { "epoch": 0.16, "learning_rate": 1.9085404965830114e-05, "loss": 2.3894, "step": 6617 }, { "epoch": 0.16, "learning_rate": 1.9085071396687883e-05, "loss": 2.1489, "step": 6618 }, { "epoch": 0.16, "learning_rate": 1.9084737769643536e-05, "loss": 2.1998, "step": 6619 }, { "epoch": 0.16, "learning_rate": 1.9084404084699192e-05, "loss": 2.1165, "step": 6620 }, { "epoch": 0.16, "learning_rate": 1.9084070341856985e-05, "loss": 2.2441, "step": 6621 }, { "epoch": 0.16, "learning_rate": 1.908373654111903e-05, "loss": 2.1063, "step": 6622 }, { "epoch": 0.16, "learning_rate": 1.9083402682487466e-05, "loss": 2.1489, "step": 6623 }, { "epoch": 0.16, "learning_rate": 1.9083068765964416e-05, "loss": 2.4032, "step": 6624 }, { "epoch": 0.16, "learning_rate": 1.908273479155201e-05, "loss": 2.3113, "step": 6625 }, { "epoch": 0.16, "learning_rate": 1.9082400759252374e-05, "loss": 2.1739, "step": 6626 }, { "epoch": 0.16, "learning_rate": 1.9082066669067637e-05, "loss": 2.3654, "step": 6627 }, { "epoch": 0.16, "learning_rate": 1.9081732520999932e-05, "loss": 2.1083, "step": 6628 }, { "epoch": 0.16, "learning_rate": 1.9081398315051382e-05, "loss": 2.0624, "step": 6629 }, { "epoch": 0.16, "learning_rate": 1.9081064051224122e-05, "loss": 2.2094, "step": 6630 }, { "epoch": 0.16, "learning_rate": 1.9080729729520286e-05, "loss": 2.185, "step": 6631 }, { "epoch": 0.16, "learning_rate": 1.9080395349941997e-05, "loss": 2.177, "step": 6632 }, { "epoch": 0.16, "learning_rate": 1.9080060912491387e-05, "loss": 2.3701, "step": 6633 }, { "epoch": 0.16, "learning_rate": 1.907972641717059e-05, "loss": 2.3238, "step": 6634 }, { "epoch": 0.16, "learning_rate": 1.9079391863981742e-05, "loss": 2.2791, "step": 6635 }, { "epoch": 0.16, "learning_rate": 1.9079057252926966e-05, "loss": 2.2288, "step": 6636 }, { "epoch": 0.16, "learning_rate": 1.90787225840084e-05, "loss": 2.26, "step": 6637 }, { "epoch": 0.16, "learning_rate": 1.907838785722818e-05, "loss": 2.0594, "step": 6638 }, { "epoch": 0.16, "learning_rate": 1.9078053072588434e-05, "loss": 2.1417, "step": 6639 }, { "epoch": 0.16, "learning_rate": 1.9077718230091297e-05, "loss": 2.1449, "step": 6640 }, { "epoch": 0.16, "learning_rate": 1.90773833297389e-05, "loss": 2.1093, "step": 6641 }, { "epoch": 0.16, "learning_rate": 1.9077048371533386e-05, "loss": 2.195, "step": 6642 }, { "epoch": 0.16, "learning_rate": 1.907671335547688e-05, "loss": 2.2361, "step": 6643 }, { "epoch": 0.16, "learning_rate": 1.9076378281571525e-05, "loss": 2.2964, "step": 6644 }, { "epoch": 0.16, "learning_rate": 1.9076043149819453e-05, "loss": 2.0709, "step": 6645 }, { "epoch": 0.16, "learning_rate": 1.90757079602228e-05, "loss": 2.2924, "step": 6646 }, { "epoch": 0.16, "learning_rate": 1.90753727127837e-05, "loss": 2.2121, "step": 6647 }, { "epoch": 0.16, "learning_rate": 1.90750374075043e-05, "loss": 2.125, "step": 6648 }, { "epoch": 0.16, "learning_rate": 1.907470204438672e-05, "loss": 2.0764, "step": 6649 }, { "epoch": 0.16, "learning_rate": 1.907436662343311e-05, "loss": 2.2758, "step": 6650 }, { "epoch": 0.16, "learning_rate": 1.9074031144645606e-05, "loss": 2.2262, "step": 6651 }, { "epoch": 0.16, "learning_rate": 1.907369560802634e-05, "loss": 2.405, "step": 6652 }, { "epoch": 0.16, "learning_rate": 1.907336001357746e-05, "loss": 2.352, "step": 6653 }, { "epoch": 0.16, "learning_rate": 1.9073024361301095e-05, "loss": 2.2111, "step": 6654 }, { "epoch": 0.16, "learning_rate": 1.9072688651199392e-05, "loss": 2.3442, "step": 6655 }, { "epoch": 0.16, "learning_rate": 1.9072352883274488e-05, "loss": 2.2493, "step": 6656 }, { "epoch": 0.16, "learning_rate": 1.907201705752852e-05, "loss": 2.2895, "step": 6657 }, { "epoch": 0.16, "learning_rate": 1.907168117396363e-05, "loss": 2.2281, "step": 6658 }, { "epoch": 0.16, "learning_rate": 1.907134523258196e-05, "loss": 2.1233, "step": 6659 }, { "epoch": 0.16, "learning_rate": 1.9071009233385657e-05, "loss": 2.3065, "step": 6660 }, { "epoch": 0.16, "learning_rate": 1.9070673176376847e-05, "loss": 2.227, "step": 6661 }, { "epoch": 0.16, "learning_rate": 1.9070337061557686e-05, "loss": 2.2642, "step": 6662 }, { "epoch": 0.16, "learning_rate": 1.907000088893031e-05, "loss": 2.2595, "step": 6663 }, { "epoch": 0.16, "learning_rate": 1.9069664658496864e-05, "loss": 2.1553, "step": 6664 }, { "epoch": 0.16, "learning_rate": 1.9069328370259484e-05, "loss": 2.1563, "step": 6665 }, { "epoch": 0.16, "learning_rate": 1.9068992024220322e-05, "loss": 2.218, "step": 6666 }, { "epoch": 0.16, "learning_rate": 1.906865562038152e-05, "loss": 2.2816, "step": 6667 }, { "epoch": 0.16, "learning_rate": 1.906831915874522e-05, "loss": 2.2086, "step": 6668 }, { "epoch": 0.16, "learning_rate": 1.9067982639313565e-05, "loss": 2.2375, "step": 6669 }, { "epoch": 0.16, "learning_rate": 1.9067646062088704e-05, "loss": 2.259, "step": 6670 }, { "epoch": 0.16, "learning_rate": 1.9067309427072776e-05, "loss": 2.2066, "step": 6671 }, { "epoch": 0.16, "learning_rate": 1.906697273426793e-05, "loss": 2.3181, "step": 6672 }, { "epoch": 0.16, "learning_rate": 1.9066635983676316e-05, "loss": 2.2273, "step": 6673 }, { "epoch": 0.16, "learning_rate": 1.9066299175300073e-05, "loss": 2.4654, "step": 6674 }, { "epoch": 0.16, "learning_rate": 1.9065962309141353e-05, "loss": 2.3513, "step": 6675 }, { "epoch": 0.16, "learning_rate": 1.90656253852023e-05, "loss": 2.2022, "step": 6676 }, { "epoch": 0.16, "learning_rate": 1.906528840348506e-05, "loss": 2.093, "step": 6677 }, { "epoch": 0.16, "learning_rate": 1.9064951363991783e-05, "loss": 2.0941, "step": 6678 }, { "epoch": 0.16, "learning_rate": 1.906461426672462e-05, "loss": 2.197, "step": 6679 }, { "epoch": 0.16, "learning_rate": 1.9064277111685712e-05, "loss": 2.3925, "step": 6680 }, { "epoch": 0.16, "learning_rate": 1.9063939898877214e-05, "loss": 2.1822, "step": 6681 }, { "epoch": 0.16, "learning_rate": 1.9063602628301275e-05, "loss": 2.2382, "step": 6682 }, { "epoch": 0.16, "learning_rate": 1.906326529996004e-05, "loss": 2.283, "step": 6683 }, { "epoch": 0.16, "learning_rate": 1.906292791385566e-05, "loss": 2.2078, "step": 6684 }, { "epoch": 0.16, "learning_rate": 1.9062590469990294e-05, "loss": 2.3321, "step": 6685 }, { "epoch": 0.16, "learning_rate": 1.906225296836608e-05, "loss": 2.2094, "step": 6686 }, { "epoch": 0.16, "learning_rate": 1.9061915408985175e-05, "loss": 2.2206, "step": 6687 }, { "epoch": 0.16, "learning_rate": 1.906157779184973e-05, "loss": 2.1563, "step": 6688 }, { "epoch": 0.16, "learning_rate": 1.9061240116961897e-05, "loss": 2.1802, "step": 6689 }, { "epoch": 0.16, "learning_rate": 1.9060902384323827e-05, "loss": 2.4228, "step": 6690 }, { "epoch": 0.16, "learning_rate": 1.9060564593937676e-05, "loss": 2.0969, "step": 6691 }, { "epoch": 0.16, "learning_rate": 1.9060226745805594e-05, "loss": 2.2223, "step": 6692 }, { "epoch": 0.16, "learning_rate": 1.905988883992973e-05, "loss": 2.1042, "step": 6693 }, { "epoch": 0.16, "learning_rate": 1.9059550876312243e-05, "loss": 2.3805, "step": 6694 }, { "epoch": 0.17, "learning_rate": 1.905921285495529e-05, "loss": 2.4195, "step": 6695 }, { "epoch": 0.17, "learning_rate": 1.9058874775861018e-05, "loss": 2.362, "step": 6696 }, { "epoch": 0.17, "learning_rate": 1.9058536639031585e-05, "loss": 2.2866, "step": 6697 }, { "epoch": 0.17, "learning_rate": 1.9058198444469146e-05, "loss": 2.1011, "step": 6698 }, { "epoch": 0.17, "learning_rate": 1.9057860192175855e-05, "loss": 2.3138, "step": 6699 }, { "epoch": 0.17, "learning_rate": 1.905752188215387e-05, "loss": 2.1878, "step": 6700 }, { "epoch": 0.17, "learning_rate": 1.905718351440535e-05, "loss": 2.2722, "step": 6701 }, { "epoch": 0.17, "learning_rate": 1.9056845088932442e-05, "loss": 2.3874, "step": 6702 }, { "epoch": 0.17, "learning_rate": 1.905650660573731e-05, "loss": 2.1198, "step": 6703 }, { "epoch": 0.17, "learning_rate": 1.905616806482211e-05, "loss": 2.2166, "step": 6704 }, { "epoch": 0.17, "learning_rate": 1.9055829466189e-05, "loss": 2.2075, "step": 6705 }, { "epoch": 0.17, "learning_rate": 1.9055490809840144e-05, "loss": 2.2092, "step": 6706 }, { "epoch": 0.17, "learning_rate": 1.9055152095777687e-05, "loss": 2.2869, "step": 6707 }, { "epoch": 0.17, "learning_rate": 1.9054813324003795e-05, "loss": 2.2146, "step": 6708 }, { "epoch": 0.17, "learning_rate": 1.905447449452063e-05, "loss": 2.256, "step": 6709 }, { "epoch": 0.17, "learning_rate": 1.9054135607330345e-05, "loss": 2.3106, "step": 6710 }, { "epoch": 0.17, "learning_rate": 1.9053796662435106e-05, "loss": 2.1142, "step": 6711 }, { "epoch": 0.17, "learning_rate": 1.9053457659837066e-05, "loss": 2.3288, "step": 6712 }, { "epoch": 0.17, "learning_rate": 1.9053118599538395e-05, "loss": 2.2447, "step": 6713 }, { "epoch": 0.17, "learning_rate": 1.9052779481541247e-05, "loss": 2.213, "step": 6714 }, { "epoch": 0.17, "learning_rate": 1.905244030584778e-05, "loss": 2.316, "step": 6715 }, { "epoch": 0.17, "learning_rate": 1.9052101072460168e-05, "loss": 2.2089, "step": 6716 }, { "epoch": 0.17, "learning_rate": 1.9051761781380565e-05, "loss": 2.1702, "step": 6717 }, { "epoch": 0.17, "learning_rate": 1.905142243261113e-05, "loss": 2.1765, "step": 6718 }, { "epoch": 0.17, "learning_rate": 1.9051083026154034e-05, "loss": 2.2598, "step": 6719 }, { "epoch": 0.17, "learning_rate": 1.9050743562011434e-05, "loss": 2.1495, "step": 6720 }, { "epoch": 0.17, "learning_rate": 1.9050404040185497e-05, "loss": 2.3651, "step": 6721 }, { "epoch": 0.17, "learning_rate": 1.9050064460678385e-05, "loss": 2.2194, "step": 6722 }, { "epoch": 0.17, "learning_rate": 1.9049724823492262e-05, "loss": 2.0192, "step": 6723 }, { "epoch": 0.17, "learning_rate": 1.9049385128629295e-05, "loss": 2.1956, "step": 6724 }, { "epoch": 0.17, "learning_rate": 1.9049045376091648e-05, "loss": 2.2433, "step": 6725 }, { "epoch": 0.17, "learning_rate": 1.9048705565881485e-05, "loss": 2.1926, "step": 6726 }, { "epoch": 0.17, "learning_rate": 1.9048365698000973e-05, "loss": 2.3257, "step": 6727 }, { "epoch": 0.17, "learning_rate": 1.9048025772452275e-05, "loss": 2.3105, "step": 6728 }, { "epoch": 0.17, "learning_rate": 1.9047685789237565e-05, "loss": 2.2853, "step": 6729 }, { "epoch": 0.17, "learning_rate": 1.9047345748359002e-05, "loss": 2.3221, "step": 6730 }, { "epoch": 0.17, "learning_rate": 1.9047005649818753e-05, "loss": 2.2208, "step": 6731 }, { "epoch": 0.17, "learning_rate": 1.9046665493618993e-05, "loss": 2.1798, "step": 6732 }, { "epoch": 0.17, "learning_rate": 1.9046325279761887e-05, "loss": 2.1367, "step": 6733 }, { "epoch": 0.17, "learning_rate": 1.9045985008249595e-05, "loss": 2.2719, "step": 6734 }, { "epoch": 0.17, "learning_rate": 1.9045644679084298e-05, "loss": 2.2621, "step": 6735 }, { "epoch": 0.17, "learning_rate": 1.904530429226816e-05, "loss": 2.2816, "step": 6736 }, { "epoch": 0.17, "learning_rate": 1.9044963847803348e-05, "loss": 2.1003, "step": 6737 }, { "epoch": 0.17, "learning_rate": 1.9044623345692032e-05, "loss": 2.3397, "step": 6738 }, { "epoch": 0.17, "learning_rate": 1.9044282785936387e-05, "loss": 2.2786, "step": 6739 }, { "epoch": 0.17, "learning_rate": 1.9043942168538578e-05, "loss": 2.1764, "step": 6740 }, { "epoch": 0.17, "learning_rate": 1.904360149350078e-05, "loss": 2.2495, "step": 6741 }, { "epoch": 0.17, "learning_rate": 1.904326076082516e-05, "loss": 2.1811, "step": 6742 }, { "epoch": 0.17, "learning_rate": 1.9042919970513897e-05, "loss": 2.3027, "step": 6743 }, { "epoch": 0.17, "learning_rate": 1.9042579122569153e-05, "loss": 2.2162, "step": 6744 }, { "epoch": 0.17, "learning_rate": 1.904223821699311e-05, "loss": 2.4422, "step": 6745 }, { "epoch": 0.17, "learning_rate": 1.9041897253787934e-05, "loss": 2.1227, "step": 6746 }, { "epoch": 0.17, "learning_rate": 1.90415562329558e-05, "loss": 2.2364, "step": 6747 }, { "epoch": 0.17, "learning_rate": 1.9041215154498878e-05, "loss": 2.3947, "step": 6748 }, { "epoch": 0.17, "learning_rate": 1.904087401841935e-05, "loss": 2.4083, "step": 6749 }, { "epoch": 0.17, "learning_rate": 1.904053282471938e-05, "loss": 2.3186, "step": 6750 }, { "epoch": 0.17, "learning_rate": 1.9040191573401156e-05, "loss": 2.1061, "step": 6751 }, { "epoch": 0.17, "learning_rate": 1.903985026446684e-05, "loss": 2.274, "step": 6752 }, { "epoch": 0.17, "learning_rate": 1.9039508897918612e-05, "loss": 2.4538, "step": 6753 }, { "epoch": 0.17, "learning_rate": 1.903916747375865e-05, "loss": 2.2493, "step": 6754 }, { "epoch": 0.17, "learning_rate": 1.9038825991989127e-05, "loss": 2.0763, "step": 6755 }, { "epoch": 0.17, "learning_rate": 1.9038484452612222e-05, "loss": 2.2446, "step": 6756 }, { "epoch": 0.17, "learning_rate": 1.903814285563011e-05, "loss": 2.3427, "step": 6757 }, { "epoch": 0.17, "learning_rate": 1.9037801201044966e-05, "loss": 2.3149, "step": 6758 }, { "epoch": 0.17, "learning_rate": 1.903745948885897e-05, "loss": 2.1868, "step": 6759 }, { "epoch": 0.17, "learning_rate": 1.90371177190743e-05, "loss": 2.2332, "step": 6760 }, { "epoch": 0.17, "learning_rate": 1.903677589169313e-05, "loss": 2.2823, "step": 6761 }, { "epoch": 0.17, "learning_rate": 1.9036434006717648e-05, "loss": 2.2981, "step": 6762 }, { "epoch": 0.17, "learning_rate": 1.9036092064150024e-05, "loss": 2.2171, "step": 6763 }, { "epoch": 0.17, "learning_rate": 1.9035750063992443e-05, "loss": 2.2607, "step": 6764 }, { "epoch": 0.17, "learning_rate": 1.903540800624708e-05, "loss": 2.2349, "step": 6765 }, { "epoch": 0.17, "learning_rate": 1.903506589091612e-05, "loss": 2.1139, "step": 6766 }, { "epoch": 0.17, "learning_rate": 1.9034723718001734e-05, "loss": 2.0982, "step": 6767 }, { "epoch": 0.17, "learning_rate": 1.9034381487506116e-05, "loss": 2.0715, "step": 6768 }, { "epoch": 0.17, "learning_rate": 1.9034039199431435e-05, "loss": 2.2662, "step": 6769 }, { "epoch": 0.17, "learning_rate": 1.9033696853779883e-05, "loss": 2.2929, "step": 6770 }, { "epoch": 0.17, "learning_rate": 1.9033354450553637e-05, "loss": 2.3233, "step": 6771 }, { "epoch": 0.17, "learning_rate": 1.9033011989754877e-05, "loss": 2.3097, "step": 6772 }, { "epoch": 0.17, "learning_rate": 1.903266947138579e-05, "loss": 2.3384, "step": 6773 }, { "epoch": 0.17, "learning_rate": 1.9032326895448553e-05, "loss": 2.1285, "step": 6774 }, { "epoch": 0.17, "learning_rate": 1.9031984261945355e-05, "loss": 2.1659, "step": 6775 }, { "epoch": 0.17, "learning_rate": 1.9031641570878378e-05, "loss": 2.2049, "step": 6776 }, { "epoch": 0.17, "learning_rate": 1.9031298822249807e-05, "loss": 2.1789, "step": 6777 }, { "epoch": 0.17, "learning_rate": 1.9030956016061822e-05, "loss": 2.2138, "step": 6778 }, { "epoch": 0.17, "learning_rate": 1.9030613152316615e-05, "loss": 2.1542, "step": 6779 }, { "epoch": 0.17, "learning_rate": 1.9030270231016365e-05, "loss": 2.4243, "step": 6780 }, { "epoch": 0.17, "learning_rate": 1.9029927252163263e-05, "loss": 2.1662, "step": 6781 }, { "epoch": 0.17, "learning_rate": 1.902958421575949e-05, "loss": 2.166, "step": 6782 }, { "epoch": 0.17, "learning_rate": 1.902924112180723e-05, "loss": 2.3747, "step": 6783 }, { "epoch": 0.17, "learning_rate": 1.902889797030868e-05, "loss": 2.139, "step": 6784 }, { "epoch": 0.17, "learning_rate": 1.902855476126602e-05, "loss": 2.1174, "step": 6785 }, { "epoch": 0.17, "learning_rate": 1.9028211494681436e-05, "loss": 2.2312, "step": 6786 }, { "epoch": 0.17, "learning_rate": 1.902786817055712e-05, "loss": 2.0985, "step": 6787 }, { "epoch": 0.17, "learning_rate": 1.9027524788895255e-05, "loss": 2.3414, "step": 6788 }, { "epoch": 0.17, "learning_rate": 1.9027181349698034e-05, "loss": 2.2287, "step": 6789 }, { "epoch": 0.17, "learning_rate": 1.9026837852967646e-05, "loss": 2.1188, "step": 6790 }, { "epoch": 0.17, "learning_rate": 1.9026494298706276e-05, "loss": 2.3105, "step": 6791 }, { "epoch": 0.17, "learning_rate": 1.9026150686916118e-05, "loss": 1.9763, "step": 6792 }, { "epoch": 0.17, "learning_rate": 1.9025807017599362e-05, "loss": 2.2878, "step": 6793 }, { "epoch": 0.17, "learning_rate": 1.9025463290758193e-05, "loss": 2.1311, "step": 6794 }, { "epoch": 0.17, "learning_rate": 1.9025119506394806e-05, "loss": 2.2324, "step": 6795 }, { "epoch": 0.17, "learning_rate": 1.902477566451139e-05, "loss": 2.1731, "step": 6796 }, { "epoch": 0.17, "learning_rate": 1.9024431765110138e-05, "loss": 2.3824, "step": 6797 }, { "epoch": 0.17, "learning_rate": 1.9024087808193245e-05, "loss": 2.2648, "step": 6798 }, { "epoch": 0.17, "learning_rate": 1.9023743793762896e-05, "loss": 2.2217, "step": 6799 }, { "epoch": 0.17, "learning_rate": 1.9023399721821287e-05, "loss": 2.1537, "step": 6800 }, { "epoch": 0.17, "learning_rate": 1.9023055592370612e-05, "loss": 2.1644, "step": 6801 }, { "epoch": 0.17, "learning_rate": 1.9022711405413062e-05, "loss": 2.1179, "step": 6802 }, { "epoch": 0.17, "learning_rate": 1.9022367160950835e-05, "loss": 2.3967, "step": 6803 }, { "epoch": 0.17, "learning_rate": 1.902202285898612e-05, "loss": 2.1702, "step": 6804 }, { "epoch": 0.17, "learning_rate": 1.902167849952111e-05, "loss": 2.145, "step": 6805 }, { "epoch": 0.17, "learning_rate": 1.9021334082558004e-05, "loss": 2.2274, "step": 6806 }, { "epoch": 0.17, "learning_rate": 1.9020989608099e-05, "loss": 2.3083, "step": 6807 }, { "epoch": 0.17, "learning_rate": 1.9020645076146285e-05, "loss": 2.2915, "step": 6808 }, { "epoch": 0.17, "learning_rate": 1.9020300486702063e-05, "loss": 2.1636, "step": 6809 }, { "epoch": 0.17, "learning_rate": 1.9019955839768526e-05, "loss": 2.2651, "step": 6810 }, { "epoch": 0.17, "learning_rate": 1.9019611135347866e-05, "loss": 2.1694, "step": 6811 }, { "epoch": 0.17, "learning_rate": 1.901926637344229e-05, "loss": 2.3058, "step": 6812 }, { "epoch": 0.17, "learning_rate": 1.9018921554053988e-05, "loss": 2.2374, "step": 6813 }, { "epoch": 0.17, "learning_rate": 1.9018576677185157e-05, "loss": 2.2543, "step": 6814 }, { "epoch": 0.17, "learning_rate": 1.9018231742838003e-05, "loss": 2.2505, "step": 6815 }, { "epoch": 0.17, "learning_rate": 1.9017886751014717e-05, "loss": 2.2193, "step": 6816 }, { "epoch": 0.17, "learning_rate": 1.90175417017175e-05, "loss": 2.3286, "step": 6817 }, { "epoch": 0.17, "learning_rate": 1.9017196594948553e-05, "loss": 2.202, "step": 6818 }, { "epoch": 0.17, "learning_rate": 1.9016851430710067e-05, "loss": 2.3105, "step": 6819 }, { "epoch": 0.17, "learning_rate": 1.9016506209004255e-05, "loss": 2.3297, "step": 6820 }, { "epoch": 0.17, "learning_rate": 1.9016160929833308e-05, "loss": 2.0475, "step": 6821 }, { "epoch": 0.17, "learning_rate": 1.9015815593199432e-05, "loss": 2.2815, "step": 6822 }, { "epoch": 0.17, "learning_rate": 1.901547019910482e-05, "loss": 2.1502, "step": 6823 }, { "epoch": 0.17, "learning_rate": 1.901512474755168e-05, "loss": 2.2931, "step": 6824 }, { "epoch": 0.17, "learning_rate": 1.9014779238542214e-05, "loss": 2.1572, "step": 6825 }, { "epoch": 0.17, "learning_rate": 1.901443367207862e-05, "loss": 2.2718, "step": 6826 }, { "epoch": 0.17, "learning_rate": 1.9014088048163105e-05, "loss": 2.1462, "step": 6827 }, { "epoch": 0.17, "learning_rate": 1.9013742366797866e-05, "loss": 2.3406, "step": 6828 }, { "epoch": 0.17, "learning_rate": 1.901339662798511e-05, "loss": 2.3548, "step": 6829 }, { "epoch": 0.17, "learning_rate": 1.901305083172704e-05, "loss": 2.2899, "step": 6830 }, { "epoch": 0.17, "learning_rate": 1.9012704978025863e-05, "loss": 2.2334, "step": 6831 }, { "epoch": 0.17, "learning_rate": 1.9012359066883777e-05, "loss": 2.1702, "step": 6832 }, { "epoch": 0.17, "learning_rate": 1.901201309830299e-05, "loss": 2.0329, "step": 6833 }, { "epoch": 0.17, "learning_rate": 1.901166707228571e-05, "loss": 2.4147, "step": 6834 }, { "epoch": 0.17, "learning_rate": 1.9011320988834133e-05, "loss": 2.1946, "step": 6835 }, { "epoch": 0.17, "learning_rate": 1.9010974847950474e-05, "loss": 2.2013, "step": 6836 }, { "epoch": 0.17, "learning_rate": 1.9010628649636934e-05, "loss": 2.2646, "step": 6837 }, { "epoch": 0.17, "learning_rate": 1.9010282393895725e-05, "loss": 2.2431, "step": 6838 }, { "epoch": 0.17, "learning_rate": 1.9009936080729044e-05, "loss": 2.2334, "step": 6839 }, { "epoch": 0.17, "learning_rate": 1.900958971013911e-05, "loss": 2.3264, "step": 6840 }, { "epoch": 0.17, "learning_rate": 1.900924328212812e-05, "loss": 2.1268, "step": 6841 }, { "epoch": 0.17, "learning_rate": 1.9008896796698288e-05, "loss": 2.2124, "step": 6842 }, { "epoch": 0.17, "learning_rate": 1.9008550253851823e-05, "loss": 2.332, "step": 6843 }, { "epoch": 0.17, "learning_rate": 1.9008203653590927e-05, "loss": 2.2617, "step": 6844 }, { "epoch": 0.17, "learning_rate": 1.900785699591782e-05, "loss": 2.2737, "step": 6845 }, { "epoch": 0.17, "learning_rate": 1.90075102808347e-05, "loss": 2.4013, "step": 6846 }, { "epoch": 0.17, "learning_rate": 1.9007163508343783e-05, "loss": 2.2911, "step": 6847 }, { "epoch": 0.17, "learning_rate": 1.9006816678447278e-05, "loss": 2.1873, "step": 6848 }, { "epoch": 0.17, "learning_rate": 1.9006469791147394e-05, "loss": 2.06, "step": 6849 }, { "epoch": 0.17, "learning_rate": 1.900612284644634e-05, "loss": 2.2325, "step": 6850 }, { "epoch": 0.17, "learning_rate": 1.9005775844346333e-05, "loss": 2.0587, "step": 6851 }, { "epoch": 0.17, "learning_rate": 1.9005428784849585e-05, "loss": 2.2983, "step": 6852 }, { "epoch": 0.17, "learning_rate": 1.90050816679583e-05, "loss": 2.2791, "step": 6853 }, { "epoch": 0.17, "learning_rate": 1.9004734493674696e-05, "loss": 2.2236, "step": 6854 }, { "epoch": 0.17, "learning_rate": 1.9004387262000984e-05, "loss": 2.2807, "step": 6855 }, { "epoch": 0.17, "learning_rate": 1.900403997293938e-05, "loss": 2.1995, "step": 6856 }, { "epoch": 0.17, "learning_rate": 1.900369262649209e-05, "loss": 2.2751, "step": 6857 }, { "epoch": 0.17, "learning_rate": 1.9003345222661338e-05, "loss": 2.2279, "step": 6858 }, { "epoch": 0.17, "learning_rate": 1.900299776144933e-05, "loss": 2.2647, "step": 6859 }, { "epoch": 0.17, "learning_rate": 1.9002650242858285e-05, "loss": 2.2054, "step": 6860 }, { "epoch": 0.17, "learning_rate": 1.9002302666890417e-05, "loss": 2.0755, "step": 6861 }, { "epoch": 0.17, "learning_rate": 1.9001955033547935e-05, "loss": 2.0581, "step": 6862 }, { "epoch": 0.17, "learning_rate": 1.9001607342833065e-05, "loss": 2.2988, "step": 6863 }, { "epoch": 0.17, "learning_rate": 1.9001259594748018e-05, "loss": 2.0779, "step": 6864 }, { "epoch": 0.17, "learning_rate": 1.9000911789295004e-05, "loss": 2.2007, "step": 6865 }, { "epoch": 0.17, "learning_rate": 1.9000563926476253e-05, "loss": 2.2798, "step": 6866 }, { "epoch": 0.17, "learning_rate": 1.9000216006293972e-05, "loss": 2.2574, "step": 6867 }, { "epoch": 0.17, "learning_rate": 1.8999868028750382e-05, "loss": 2.1367, "step": 6868 }, { "epoch": 0.17, "learning_rate": 1.89995199938477e-05, "loss": 2.101, "step": 6869 }, { "epoch": 0.17, "learning_rate": 1.899917190158814e-05, "loss": 1.9804, "step": 6870 }, { "epoch": 0.17, "learning_rate": 1.899882375197393e-05, "loss": 2.1958, "step": 6871 }, { "epoch": 0.17, "learning_rate": 1.8998475545007283e-05, "loss": 2.1715, "step": 6872 }, { "epoch": 0.17, "learning_rate": 1.8998127280690416e-05, "loss": 2.3363, "step": 6873 }, { "epoch": 0.17, "learning_rate": 1.8997778959025556e-05, "loss": 2.2615, "step": 6874 }, { "epoch": 0.17, "learning_rate": 1.8997430580014912e-05, "loss": 2.166, "step": 6875 }, { "epoch": 0.17, "learning_rate": 1.8997082143660717e-05, "loss": 2.0251, "step": 6876 }, { "epoch": 0.17, "learning_rate": 1.8996733649965183e-05, "loss": 2.2088, "step": 6877 }, { "epoch": 0.17, "learning_rate": 1.8996385098930533e-05, "loss": 2.1936, "step": 6878 }, { "epoch": 0.17, "learning_rate": 1.899603649055899e-05, "loss": 2.1264, "step": 6879 }, { "epoch": 0.17, "learning_rate": 1.8995687824852772e-05, "loss": 2.1741, "step": 6880 }, { "epoch": 0.17, "learning_rate": 1.8995339101814107e-05, "loss": 2.2885, "step": 6881 }, { "epoch": 0.17, "learning_rate": 1.8994990321445215e-05, "loss": 2.2149, "step": 6882 }, { "epoch": 0.17, "learning_rate": 1.8994641483748314e-05, "loss": 2.1525, "step": 6883 }, { "epoch": 0.17, "learning_rate": 1.8994292588725636e-05, "loss": 2.3086, "step": 6884 }, { "epoch": 0.17, "learning_rate": 1.8993943636379396e-05, "loss": 2.2234, "step": 6885 }, { "epoch": 0.17, "learning_rate": 1.8993594626711825e-05, "loss": 2.1485, "step": 6886 }, { "epoch": 0.17, "learning_rate": 1.8993245559725144e-05, "loss": 2.1404, "step": 6887 }, { "epoch": 0.17, "learning_rate": 1.899289643542158e-05, "loss": 2.2493, "step": 6888 }, { "epoch": 0.17, "learning_rate": 1.8992547253803357e-05, "loss": 2.3688, "step": 6889 }, { "epoch": 0.17, "learning_rate": 1.8992198014872694e-05, "loss": 2.1447, "step": 6890 }, { "epoch": 0.17, "learning_rate": 1.8991848718631827e-05, "loss": 2.3511, "step": 6891 }, { "epoch": 0.17, "learning_rate": 1.8991499365082977e-05, "loss": 2.028, "step": 6892 }, { "epoch": 0.17, "learning_rate": 1.899114995422837e-05, "loss": 2.3188, "step": 6893 }, { "epoch": 0.17, "learning_rate": 1.8990800486070234e-05, "loss": 1.9832, "step": 6894 }, { "epoch": 0.17, "learning_rate": 1.89904509606108e-05, "loss": 2.2304, "step": 6895 }, { "epoch": 0.17, "learning_rate": 1.899010137785229e-05, "loss": 2.354, "step": 6896 }, { "epoch": 0.17, "learning_rate": 1.898975173779693e-05, "loss": 2.1547, "step": 6897 }, { "epoch": 0.17, "learning_rate": 1.898940204044696e-05, "loss": 2.1176, "step": 6898 }, { "epoch": 0.17, "learning_rate": 1.8989052285804596e-05, "loss": 2.1199, "step": 6899 }, { "epoch": 0.17, "learning_rate": 1.8988702473872073e-05, "loss": 2.4279, "step": 6900 }, { "epoch": 0.17, "learning_rate": 1.898835260465162e-05, "loss": 2.3976, "step": 6901 }, { "epoch": 0.17, "learning_rate": 1.898800267814547e-05, "loss": 2.2271, "step": 6902 }, { "epoch": 0.17, "learning_rate": 1.8987652694355848e-05, "loss": 2.1396, "step": 6903 }, { "epoch": 0.17, "learning_rate": 1.8987302653284985e-05, "loss": 2.1098, "step": 6904 }, { "epoch": 0.17, "learning_rate": 1.8986952554935114e-05, "loss": 2.3818, "step": 6905 }, { "epoch": 0.17, "learning_rate": 1.8986602399308466e-05, "loss": 2.0305, "step": 6906 }, { "epoch": 0.17, "learning_rate": 1.898625218640727e-05, "loss": 2.1469, "step": 6907 }, { "epoch": 0.17, "learning_rate": 1.8985901916233764e-05, "loss": 2.0331, "step": 6908 }, { "epoch": 0.17, "learning_rate": 1.8985551588790172e-05, "loss": 2.4167, "step": 6909 }, { "epoch": 0.17, "learning_rate": 1.8985201204078736e-05, "loss": 2.2367, "step": 6910 }, { "epoch": 0.17, "learning_rate": 1.898485076210168e-05, "loss": 2.2065, "step": 6911 }, { "epoch": 0.17, "learning_rate": 1.8984500262861247e-05, "loss": 2.2459, "step": 6912 }, { "epoch": 0.17, "learning_rate": 1.898414970635966e-05, "loss": 2.1424, "step": 6913 }, { "epoch": 0.17, "learning_rate": 1.8983799092599166e-05, "loss": 2.1793, "step": 6914 }, { "epoch": 0.17, "learning_rate": 1.898344842158199e-05, "loss": 2.2782, "step": 6915 }, { "epoch": 0.17, "learning_rate": 1.898309769331037e-05, "loss": 2.1394, "step": 6916 }, { "epoch": 0.17, "learning_rate": 1.8982746907786537e-05, "loss": 2.2506, "step": 6917 }, { "epoch": 0.17, "learning_rate": 1.8982396065012735e-05, "loss": 2.2924, "step": 6918 }, { "epoch": 0.17, "learning_rate": 1.8982045164991195e-05, "loss": 2.198, "step": 6919 }, { "epoch": 0.17, "learning_rate": 1.898169420772415e-05, "loss": 2.2441, "step": 6920 }, { "epoch": 0.17, "learning_rate": 1.8981343193213845e-05, "loss": 2.2073, "step": 6921 }, { "epoch": 0.17, "learning_rate": 1.898099212146251e-05, "loss": 2.2717, "step": 6922 }, { "epoch": 0.17, "learning_rate": 1.898064099247239e-05, "loss": 2.3185, "step": 6923 }, { "epoch": 0.17, "learning_rate": 1.8980289806245714e-05, "loss": 2.1666, "step": 6924 }, { "epoch": 0.17, "learning_rate": 1.8979938562784724e-05, "loss": 2.4079, "step": 6925 }, { "epoch": 0.17, "learning_rate": 1.8979587262091663e-05, "loss": 2.1368, "step": 6926 }, { "epoch": 0.17, "learning_rate": 1.8979235904168764e-05, "loss": 2.1619, "step": 6927 }, { "epoch": 0.17, "learning_rate": 1.8978884489018265e-05, "loss": 2.2793, "step": 6928 }, { "epoch": 0.17, "learning_rate": 1.8978533016642413e-05, "loss": 2.1587, "step": 6929 }, { "epoch": 0.17, "learning_rate": 1.8978181487043445e-05, "loss": 2.361, "step": 6930 }, { "epoch": 0.17, "learning_rate": 1.8977829900223597e-05, "loss": 2.1759, "step": 6931 }, { "epoch": 0.17, "learning_rate": 1.8977478256185117e-05, "loss": 2.2821, "step": 6932 }, { "epoch": 0.17, "learning_rate": 1.897712655493024e-05, "loss": 2.2637, "step": 6933 }, { "epoch": 0.17, "learning_rate": 1.897677479646121e-05, "loss": 2.1724, "step": 6934 }, { "epoch": 0.17, "learning_rate": 1.8976422980780266e-05, "loss": 2.205, "step": 6935 }, { "epoch": 0.17, "learning_rate": 1.897607110788966e-05, "loss": 2.1746, "step": 6936 }, { "epoch": 0.17, "learning_rate": 1.8975719177791623e-05, "loss": 2.256, "step": 6937 }, { "epoch": 0.17, "learning_rate": 1.8975367190488404e-05, "loss": 2.2224, "step": 6938 }, { "epoch": 0.17, "learning_rate": 1.8975015145982244e-05, "loss": 2.1188, "step": 6939 }, { "epoch": 0.17, "learning_rate": 1.8974663044275386e-05, "loss": 2.1325, "step": 6940 }, { "epoch": 0.17, "learning_rate": 1.8974310885370078e-05, "loss": 2.1972, "step": 6941 }, { "epoch": 0.17, "learning_rate": 1.8973958669268562e-05, "loss": 2.2223, "step": 6942 }, { "epoch": 0.17, "learning_rate": 1.897360639597308e-05, "loss": 2.2284, "step": 6943 }, { "epoch": 0.17, "learning_rate": 1.8973254065485883e-05, "loss": 2.2085, "step": 6944 }, { "epoch": 0.17, "learning_rate": 1.8972901677809214e-05, "loss": 2.3012, "step": 6945 }, { "epoch": 0.17, "learning_rate": 1.8972549232945315e-05, "loss": 2.1975, "step": 6946 }, { "epoch": 0.17, "learning_rate": 1.897219673089644e-05, "loss": 2.351, "step": 6947 }, { "epoch": 0.17, "learning_rate": 1.897184417166483e-05, "loss": 2.4927, "step": 6948 }, { "epoch": 0.17, "learning_rate": 1.897149155525273e-05, "loss": 2.3903, "step": 6949 }, { "epoch": 0.17, "learning_rate": 1.8971138881662394e-05, "loss": 2.2587, "step": 6950 }, { "epoch": 0.17, "learning_rate": 1.8970786150896066e-05, "loss": 2.3579, "step": 6951 }, { "epoch": 0.17, "learning_rate": 1.8970433362955997e-05, "loss": 2.2128, "step": 6952 }, { "epoch": 0.17, "learning_rate": 1.8970080517844426e-05, "loss": 2.1009, "step": 6953 }, { "epoch": 0.17, "learning_rate": 1.896972761556361e-05, "loss": 2.2712, "step": 6954 }, { "epoch": 0.17, "learning_rate": 1.89693746561158e-05, "loss": 2.3177, "step": 6955 }, { "epoch": 0.17, "learning_rate": 1.8969021639503243e-05, "loss": 2.214, "step": 6956 }, { "epoch": 0.17, "learning_rate": 1.896866856572819e-05, "loss": 2.0516, "step": 6957 }, { "epoch": 0.17, "learning_rate": 1.896831543479288e-05, "loss": 2.0111, "step": 6958 }, { "epoch": 0.17, "learning_rate": 1.896796224669958e-05, "loss": 2.3453, "step": 6959 }, { "epoch": 0.17, "learning_rate": 1.8967609001450533e-05, "loss": 2.4375, "step": 6960 }, { "epoch": 0.17, "learning_rate": 1.896725569904799e-05, "loss": 2.2835, "step": 6961 }, { "epoch": 0.17, "learning_rate": 1.896690233949421e-05, "loss": 2.1332, "step": 6962 }, { "epoch": 0.17, "learning_rate": 1.896654892279143e-05, "loss": 2.0731, "step": 6963 }, { "epoch": 0.17, "learning_rate": 1.8966195448941917e-05, "loss": 2.3602, "step": 6964 }, { "epoch": 0.17, "learning_rate": 1.8965841917947917e-05, "loss": 2.245, "step": 6965 }, { "epoch": 0.17, "learning_rate": 1.896548832981169e-05, "loss": 2.186, "step": 6966 }, { "epoch": 0.17, "learning_rate": 1.8965134684535476e-05, "loss": 2.2172, "step": 6967 }, { "epoch": 0.17, "learning_rate": 1.896478098212154e-05, "loss": 2.1222, "step": 6968 }, { "epoch": 0.17, "learning_rate": 1.8964427222572132e-05, "loss": 2.3057, "step": 6969 }, { "epoch": 0.17, "learning_rate": 1.8964073405889508e-05, "loss": 2.3161, "step": 6970 }, { "epoch": 0.17, "learning_rate": 1.8963719532075928e-05, "loss": 2.2448, "step": 6971 }, { "epoch": 0.17, "learning_rate": 1.8963365601133638e-05, "loss": 2.2058, "step": 6972 }, { "epoch": 0.17, "learning_rate": 1.8963011613064898e-05, "loss": 2.0366, "step": 6973 }, { "epoch": 0.17, "learning_rate": 1.896265756787196e-05, "loss": 2.2368, "step": 6974 }, { "epoch": 0.17, "learning_rate": 1.8962303465557092e-05, "loss": 2.2938, "step": 6975 }, { "epoch": 0.17, "learning_rate": 1.896194930612254e-05, "loss": 2.2359, "step": 6976 }, { "epoch": 0.17, "learning_rate": 1.8961595089570567e-05, "loss": 2.3835, "step": 6977 }, { "epoch": 0.17, "learning_rate": 1.8961240815903426e-05, "loss": 2.286, "step": 6978 }, { "epoch": 0.17, "learning_rate": 1.8960886485123374e-05, "loss": 2.228, "step": 6979 }, { "epoch": 0.17, "learning_rate": 1.8960532097232677e-05, "loss": 2.1732, "step": 6980 }, { "epoch": 0.17, "learning_rate": 1.8960177652233586e-05, "loss": 2.155, "step": 6981 }, { "epoch": 0.17, "learning_rate": 1.8959823150128364e-05, "loss": 2.3199, "step": 6982 }, { "epoch": 0.17, "learning_rate": 1.8959468590919267e-05, "loss": 2.1725, "step": 6983 }, { "epoch": 0.17, "learning_rate": 1.895911397460856e-05, "loss": 2.2847, "step": 6984 }, { "epoch": 0.17, "learning_rate": 1.8958759301198496e-05, "loss": 2.2565, "step": 6985 }, { "epoch": 0.17, "learning_rate": 1.895840457069134e-05, "loss": 2.3399, "step": 6986 }, { "epoch": 0.17, "learning_rate": 1.8958049783089357e-05, "loss": 2.1693, "step": 6987 }, { "epoch": 0.17, "learning_rate": 1.89576949383948e-05, "loss": 2.2912, "step": 6988 }, { "epoch": 0.17, "learning_rate": 1.8957340036609935e-05, "loss": 2.3627, "step": 6989 }, { "epoch": 0.17, "learning_rate": 1.895698507773702e-05, "loss": 2.283, "step": 6990 }, { "epoch": 0.17, "learning_rate": 1.895663006177832e-05, "loss": 2.1745, "step": 6991 }, { "epoch": 0.17, "learning_rate": 1.8956274988736103e-05, "loss": 2.2337, "step": 6992 }, { "epoch": 0.17, "learning_rate": 1.895591985861262e-05, "loss": 2.1468, "step": 6993 }, { "epoch": 0.17, "learning_rate": 1.8955564671410146e-05, "loss": 2.1653, "step": 6994 }, { "epoch": 0.17, "learning_rate": 1.895520942713094e-05, "loss": 2.3934, "step": 6995 }, { "epoch": 0.17, "learning_rate": 1.8954854125777265e-05, "loss": 2.4452, "step": 6996 }, { "epoch": 0.17, "learning_rate": 1.8954498767351386e-05, "loss": 2.1673, "step": 6997 }, { "epoch": 0.17, "learning_rate": 1.8954143351855565e-05, "loss": 2.1574, "step": 6998 }, { "epoch": 0.17, "learning_rate": 1.8953787879292074e-05, "loss": 2.3092, "step": 6999 }, { "epoch": 0.17, "learning_rate": 1.8953432349663174e-05, "loss": 2.189, "step": 7000 }, { "epoch": 0.17, "learning_rate": 1.8953076762971132e-05, "loss": 2.2147, "step": 7001 }, { "epoch": 0.17, "learning_rate": 1.8952721119218217e-05, "loss": 2.3179, "step": 7002 }, { "epoch": 0.17, "learning_rate": 1.8952365418406686e-05, "loss": 2.3913, "step": 7003 }, { "epoch": 0.17, "learning_rate": 1.895200966053882e-05, "loss": 2.3837, "step": 7004 }, { "epoch": 0.17, "learning_rate": 1.8951653845616872e-05, "loss": 2.2755, "step": 7005 }, { "epoch": 0.17, "learning_rate": 1.895129797364312e-05, "loss": 2.1825, "step": 7006 }, { "epoch": 0.17, "learning_rate": 1.895094204461983e-05, "loss": 2.1264, "step": 7007 }, { "epoch": 0.17, "learning_rate": 1.8950586058549268e-05, "loss": 2.1827, "step": 7008 }, { "epoch": 0.17, "learning_rate": 1.8950230015433705e-05, "loss": 2.1273, "step": 7009 }, { "epoch": 0.17, "learning_rate": 1.8949873915275406e-05, "loss": 2.4036, "step": 7010 }, { "epoch": 0.17, "learning_rate": 1.894951775807665e-05, "loss": 2.2921, "step": 7011 }, { "epoch": 0.17, "learning_rate": 1.8949161543839697e-05, "loss": 2.3121, "step": 7012 }, { "epoch": 0.17, "learning_rate": 1.894880527256682e-05, "loss": 2.077, "step": 7013 }, { "epoch": 0.17, "learning_rate": 1.8948448944260293e-05, "loss": 2.3486, "step": 7014 }, { "epoch": 0.17, "learning_rate": 1.8948092558922385e-05, "loss": 2.432, "step": 7015 }, { "epoch": 0.17, "learning_rate": 1.8947736116555366e-05, "loss": 2.2076, "step": 7016 }, { "epoch": 0.17, "learning_rate": 1.8947379617161507e-05, "loss": 2.2147, "step": 7017 }, { "epoch": 0.17, "learning_rate": 1.8947023060743084e-05, "loss": 2.2344, "step": 7018 }, { "epoch": 0.17, "learning_rate": 1.8946666447302368e-05, "loss": 2.2522, "step": 7019 }, { "epoch": 0.17, "learning_rate": 1.894630977684163e-05, "loss": 2.3542, "step": 7020 }, { "epoch": 0.17, "learning_rate": 1.8945953049363145e-05, "loss": 2.1959, "step": 7021 }, { "epoch": 0.17, "learning_rate": 1.894559626486918e-05, "loss": 2.1823, "step": 7022 }, { "epoch": 0.17, "learning_rate": 1.8945239423362022e-05, "loss": 2.2663, "step": 7023 }, { "epoch": 0.17, "learning_rate": 1.8944882524843935e-05, "loss": 1.9867, "step": 7024 }, { "epoch": 0.17, "learning_rate": 1.8944525569317196e-05, "loss": 2.1397, "step": 7025 }, { "epoch": 0.17, "learning_rate": 1.8944168556784084e-05, "loss": 2.399, "step": 7026 }, { "epoch": 0.17, "learning_rate": 1.894381148724687e-05, "loss": 2.2814, "step": 7027 }, { "epoch": 0.17, "learning_rate": 1.894345436070783e-05, "loss": 2.1259, "step": 7028 }, { "epoch": 0.17, "learning_rate": 1.894309717716924e-05, "loss": 2.2431, "step": 7029 }, { "epoch": 0.17, "learning_rate": 1.8942739936633377e-05, "loss": 2.0841, "step": 7030 }, { "epoch": 0.17, "learning_rate": 1.8942382639102518e-05, "loss": 2.2924, "step": 7031 }, { "epoch": 0.17, "learning_rate": 1.894202528457894e-05, "loss": 2.2722, "step": 7032 }, { "epoch": 0.17, "learning_rate": 1.8941667873064924e-05, "loss": 2.0816, "step": 7033 }, { "epoch": 0.17, "learning_rate": 1.894131040456274e-05, "loss": 2.3153, "step": 7034 }, { "epoch": 0.17, "learning_rate": 1.8940952879074675e-05, "loss": 2.1695, "step": 7035 }, { "epoch": 0.17, "learning_rate": 1.8940595296603e-05, "loss": 2.3692, "step": 7036 }, { "epoch": 0.17, "learning_rate": 1.8940237657149997e-05, "loss": 2.1207, "step": 7037 }, { "epoch": 0.17, "learning_rate": 1.893987996071795e-05, "loss": 2.2909, "step": 7038 }, { "epoch": 0.17, "learning_rate": 1.8939522207309133e-05, "loss": 2.3003, "step": 7039 }, { "epoch": 0.17, "learning_rate": 1.8939164396925827e-05, "loss": 2.3478, "step": 7040 }, { "epoch": 0.17, "learning_rate": 1.8938806529570313e-05, "loss": 2.2962, "step": 7041 }, { "epoch": 0.17, "learning_rate": 1.893844860524487e-05, "loss": 2.2166, "step": 7042 }, { "epoch": 0.17, "learning_rate": 1.8938090623951784e-05, "loss": 2.19, "step": 7043 }, { "epoch": 0.17, "learning_rate": 1.8937732585693333e-05, "loss": 2.2553, "step": 7044 }, { "epoch": 0.17, "learning_rate": 1.89373744904718e-05, "loss": 2.1877, "step": 7045 }, { "epoch": 0.17, "learning_rate": 1.8937016338289467e-05, "loss": 2.3361, "step": 7046 }, { "epoch": 0.17, "learning_rate": 1.8936658129148614e-05, "loss": 2.4589, "step": 7047 }, { "epoch": 0.17, "learning_rate": 1.8936299863051527e-05, "loss": 2.1732, "step": 7048 }, { "epoch": 0.17, "learning_rate": 1.893594154000049e-05, "loss": 2.0493, "step": 7049 }, { "epoch": 0.17, "learning_rate": 1.8935583159997787e-05, "loss": 2.1431, "step": 7050 }, { "epoch": 0.17, "learning_rate": 1.8935224723045698e-05, "loss": 2.2116, "step": 7051 }, { "epoch": 0.17, "learning_rate": 1.8934866229146513e-05, "loss": 2.2149, "step": 7052 }, { "epoch": 0.17, "learning_rate": 1.8934507678302506e-05, "loss": 2.265, "step": 7053 }, { "epoch": 0.17, "learning_rate": 1.893414907051598e-05, "loss": 2.3021, "step": 7054 }, { "epoch": 0.17, "learning_rate": 1.8933790405789203e-05, "loss": 2.2162, "step": 7055 }, { "epoch": 0.17, "learning_rate": 1.893343168412447e-05, "loss": 2.2718, "step": 7056 }, { "epoch": 0.17, "learning_rate": 1.893307290552407e-05, "loss": 2.1815, "step": 7057 }, { "epoch": 0.17, "learning_rate": 1.893271406999028e-05, "loss": 2.2701, "step": 7058 }, { "epoch": 0.17, "learning_rate": 1.8932355177525396e-05, "loss": 2.068, "step": 7059 }, { "epoch": 0.17, "learning_rate": 1.8931996228131695e-05, "loss": 2.0832, "step": 7060 }, { "epoch": 0.17, "learning_rate": 1.8931637221811476e-05, "loss": 2.2251, "step": 7061 }, { "epoch": 0.17, "learning_rate": 1.893127815856702e-05, "loss": 2.0908, "step": 7062 }, { "epoch": 0.17, "learning_rate": 1.8930919038400622e-05, "loss": 2.2528, "step": 7063 }, { "epoch": 0.17, "learning_rate": 1.8930559861314562e-05, "loss": 2.3506, "step": 7064 }, { "epoch": 0.17, "learning_rate": 1.8930200627311138e-05, "loss": 2.1836, "step": 7065 }, { "epoch": 0.17, "learning_rate": 1.892984133639263e-05, "loss": 2.0154, "step": 7066 }, { "epoch": 0.17, "learning_rate": 1.8929481988561335e-05, "loss": 2.1296, "step": 7067 }, { "epoch": 0.17, "learning_rate": 1.8929122583819544e-05, "loss": 2.0556, "step": 7068 }, { "epoch": 0.17, "learning_rate": 1.8928763122169544e-05, "loss": 2.2131, "step": 7069 }, { "epoch": 0.17, "learning_rate": 1.8928403603613623e-05, "loss": 2.1577, "step": 7070 }, { "epoch": 0.17, "learning_rate": 1.8928044028154077e-05, "loss": 2.2855, "step": 7071 }, { "epoch": 0.17, "learning_rate": 1.89276843957932e-05, "loss": 2.0979, "step": 7072 }, { "epoch": 0.17, "learning_rate": 1.892732470653328e-05, "loss": 2.2875, "step": 7073 }, { "epoch": 0.17, "learning_rate": 1.892696496037661e-05, "loss": 2.097, "step": 7074 }, { "epoch": 0.17, "learning_rate": 1.892660515732548e-05, "loss": 2.1028, "step": 7075 }, { "epoch": 0.17, "learning_rate": 1.892624529738219e-05, "loss": 2.414, "step": 7076 }, { "epoch": 0.17, "learning_rate": 1.892588538054903e-05, "loss": 2.2529, "step": 7077 }, { "epoch": 0.17, "learning_rate": 1.892552540682829e-05, "loss": 2.1772, "step": 7078 }, { "epoch": 0.17, "learning_rate": 1.892516537622227e-05, "loss": 2.1402, "step": 7079 }, { "epoch": 0.17, "learning_rate": 1.8924805288733266e-05, "loss": 2.0682, "step": 7080 }, { "epoch": 0.17, "learning_rate": 1.8924445144363567e-05, "loss": 2.1655, "step": 7081 }, { "epoch": 0.17, "learning_rate": 1.892408494311547e-05, "loss": 2.1988, "step": 7082 }, { "epoch": 0.17, "learning_rate": 1.892372468499127e-05, "loss": 2.2237, "step": 7083 }, { "epoch": 0.17, "learning_rate": 1.8923364369993265e-05, "loss": 2.2489, "step": 7084 }, { "epoch": 0.17, "learning_rate": 1.892300399812375e-05, "loss": 2.2324, "step": 7085 }, { "epoch": 0.17, "learning_rate": 1.8922643569385025e-05, "loss": 2.3305, "step": 7086 }, { "epoch": 0.17, "learning_rate": 1.8922283083779382e-05, "loss": 2.2073, "step": 7087 }, { "epoch": 0.17, "learning_rate": 1.8921922541309127e-05, "loss": 2.2749, "step": 7088 }, { "epoch": 0.17, "learning_rate": 1.8921561941976544e-05, "loss": 2.0589, "step": 7089 }, { "epoch": 0.17, "learning_rate": 1.8921201285783946e-05, "loss": 2.1629, "step": 7090 }, { "epoch": 0.17, "learning_rate": 1.8920840572733622e-05, "loss": 2.236, "step": 7091 }, { "epoch": 0.17, "learning_rate": 1.8920479802827872e-05, "loss": 2.4667, "step": 7092 }, { "epoch": 0.17, "learning_rate": 1.8920118976069e-05, "loss": 2.2373, "step": 7093 }, { "epoch": 0.17, "learning_rate": 1.89197580924593e-05, "loss": 2.1763, "step": 7094 }, { "epoch": 0.17, "learning_rate": 1.8919397152001078e-05, "loss": 2.2561, "step": 7095 }, { "epoch": 0.17, "learning_rate": 1.891903615469663e-05, "loss": 2.3062, "step": 7096 }, { "epoch": 0.17, "learning_rate": 1.8918675100548256e-05, "loss": 2.2802, "step": 7097 }, { "epoch": 0.17, "learning_rate": 1.8918313989558262e-05, "loss": 2.2186, "step": 7098 }, { "epoch": 0.17, "learning_rate": 1.8917952821728945e-05, "loss": 1.9938, "step": 7099 }, { "epoch": 0.18, "learning_rate": 1.891759159706261e-05, "loss": 2.154, "step": 7100 }, { "epoch": 0.18, "learning_rate": 1.8917230315561553e-05, "loss": 2.1998, "step": 7101 }, { "epoch": 0.18, "learning_rate": 1.8916868977228085e-05, "loss": 2.1408, "step": 7102 }, { "epoch": 0.18, "learning_rate": 1.8916507582064507e-05, "loss": 2.3585, "step": 7103 }, { "epoch": 0.18, "learning_rate": 1.891614613007312e-05, "loss": 2.2341, "step": 7104 }, { "epoch": 0.18, "learning_rate": 1.8915784621256223e-05, "loss": 2.4199, "step": 7105 }, { "epoch": 0.18, "learning_rate": 1.891542305561613e-05, "loss": 2.1524, "step": 7106 }, { "epoch": 0.18, "learning_rate": 1.8915061433155137e-05, "loss": 2.3084, "step": 7107 }, { "epoch": 0.18, "learning_rate": 1.8914699753875553e-05, "loss": 2.1768, "step": 7108 }, { "epoch": 0.18, "learning_rate": 1.8914338017779682e-05, "loss": 2.4392, "step": 7109 }, { "epoch": 0.18, "learning_rate": 1.8913976224869836e-05, "loss": 2.3033, "step": 7110 }, { "epoch": 0.18, "learning_rate": 1.8913614375148307e-05, "loss": 2.332, "step": 7111 }, { "epoch": 0.18, "learning_rate": 1.891325246861741e-05, "loss": 2.2181, "step": 7112 }, { "epoch": 0.18, "learning_rate": 1.8912890505279452e-05, "loss": 2.3905, "step": 7113 }, { "epoch": 0.18, "learning_rate": 1.8912528485136737e-05, "loss": 2.2915, "step": 7114 }, { "epoch": 0.18, "learning_rate": 1.8912166408191576e-05, "loss": 2.1962, "step": 7115 }, { "epoch": 0.18, "learning_rate": 1.891180427444627e-05, "loss": 2.2612, "step": 7116 }, { "epoch": 0.18, "learning_rate": 1.8911442083903135e-05, "loss": 2.3005, "step": 7117 }, { "epoch": 0.18, "learning_rate": 1.891107983656447e-05, "loss": 2.1429, "step": 7118 }, { "epoch": 0.18, "learning_rate": 1.8910717532432593e-05, "loss": 2.2199, "step": 7119 }, { "epoch": 0.18, "learning_rate": 1.8910355171509807e-05, "loss": 2.1566, "step": 7120 }, { "epoch": 0.18, "learning_rate": 1.8909992753798426e-05, "loss": 2.2435, "step": 7121 }, { "epoch": 0.18, "learning_rate": 1.890963027930076e-05, "loss": 2.3122, "step": 7122 }, { "epoch": 0.18, "learning_rate": 1.890926774801911e-05, "loss": 2.2032, "step": 7123 }, { "epoch": 0.18, "learning_rate": 1.89089051599558e-05, "loss": 2.1659, "step": 7124 }, { "epoch": 0.18, "learning_rate": 1.8908542515113128e-05, "loss": 2.2498, "step": 7125 }, { "epoch": 0.18, "learning_rate": 1.8908179813493416e-05, "loss": 2.2079, "step": 7126 }, { "epoch": 0.18, "learning_rate": 1.890781705509897e-05, "loss": 2.1392, "step": 7127 }, { "epoch": 0.18, "learning_rate": 1.89074542399321e-05, "loss": 2.2167, "step": 7128 }, { "epoch": 0.18, "learning_rate": 1.8907091367995123e-05, "loss": 2.3856, "step": 7129 }, { "epoch": 0.18, "learning_rate": 1.890672843929035e-05, "loss": 2.1171, "step": 7130 }, { "epoch": 0.18, "learning_rate": 1.8906365453820097e-05, "loss": 2.4268, "step": 7131 }, { "epoch": 0.18, "learning_rate": 1.890600241158667e-05, "loss": 2.2982, "step": 7132 }, { "epoch": 0.18, "learning_rate": 1.890563931259239e-05, "loss": 2.2397, "step": 7133 }, { "epoch": 0.18, "learning_rate": 1.890527615683957e-05, "loss": 2.3548, "step": 7134 }, { "epoch": 0.18, "learning_rate": 1.890491294433052e-05, "loss": 2.2677, "step": 7135 }, { "epoch": 0.18, "learning_rate": 1.890454967506756e-05, "loss": 2.1684, "step": 7136 }, { "epoch": 0.18, "learning_rate": 1.8904186349052997e-05, "loss": 2.297, "step": 7137 }, { "epoch": 0.18, "learning_rate": 1.890382296628916e-05, "loss": 2.1539, "step": 7138 }, { "epoch": 0.18, "learning_rate": 1.8903459526778352e-05, "loss": 2.4535, "step": 7139 }, { "epoch": 0.18, "learning_rate": 1.8903096030522898e-05, "loss": 2.2968, "step": 7140 }, { "epoch": 0.18, "learning_rate": 1.8902732477525112e-05, "loss": 2.201, "step": 7141 }, { "epoch": 0.18, "learning_rate": 1.8902368867787307e-05, "loss": 2.2197, "step": 7142 }, { "epoch": 0.18, "learning_rate": 1.8902005201311807e-05, "loss": 2.1924, "step": 7143 }, { "epoch": 0.18, "learning_rate": 1.8901641478100928e-05, "loss": 2.1958, "step": 7144 }, { "epoch": 0.18, "learning_rate": 1.8901277698156985e-05, "loss": 2.052, "step": 7145 }, { "epoch": 0.18, "learning_rate": 1.89009138614823e-05, "loss": 2.0304, "step": 7146 }, { "epoch": 0.18, "learning_rate": 1.8900549968079187e-05, "loss": 2.3234, "step": 7147 }, { "epoch": 0.18, "learning_rate": 1.890018601794997e-05, "loss": 2.3893, "step": 7148 }, { "epoch": 0.18, "learning_rate": 1.8899822011096967e-05, "loss": 2.1895, "step": 7149 }, { "epoch": 0.18, "learning_rate": 1.88994579475225e-05, "loss": 2.4358, "step": 7150 }, { "epoch": 0.18, "learning_rate": 1.8899093827228884e-05, "loss": 2.3061, "step": 7151 }, { "epoch": 0.18, "learning_rate": 1.8898729650218444e-05, "loss": 2.2361, "step": 7152 }, { "epoch": 0.18, "learning_rate": 1.88983654164935e-05, "loss": 2.2278, "step": 7153 }, { "epoch": 0.18, "learning_rate": 1.8898001126056372e-05, "loss": 2.3626, "step": 7154 }, { "epoch": 0.18, "learning_rate": 1.8897636778909386e-05, "loss": 2.3061, "step": 7155 }, { "epoch": 0.18, "learning_rate": 1.889727237505486e-05, "loss": 2.3056, "step": 7156 }, { "epoch": 0.18, "learning_rate": 1.8896907914495118e-05, "loss": 2.2727, "step": 7157 }, { "epoch": 0.18, "learning_rate": 1.889654339723248e-05, "loss": 2.2785, "step": 7158 }, { "epoch": 0.18, "learning_rate": 1.8896178823269276e-05, "loss": 2.3079, "step": 7159 }, { "epoch": 0.18, "learning_rate": 1.889581419260782e-05, "loss": 2.2346, "step": 7160 }, { "epoch": 0.18, "learning_rate": 1.8895449505250446e-05, "loss": 2.3544, "step": 7161 }, { "epoch": 0.18, "learning_rate": 1.8895084761199468e-05, "loss": 2.2454, "step": 7162 }, { "epoch": 0.18, "learning_rate": 1.889471996045722e-05, "loss": 2.2429, "step": 7163 }, { "epoch": 0.18, "learning_rate": 1.8894355103026024e-05, "loss": 2.1684, "step": 7164 }, { "epoch": 0.18, "learning_rate": 1.8893990188908203e-05, "loss": 2.1664, "step": 7165 }, { "epoch": 0.18, "learning_rate": 1.889362521810608e-05, "loss": 2.3465, "step": 7166 }, { "epoch": 0.18, "learning_rate": 1.889326019062199e-05, "loss": 2.269, "step": 7167 }, { "epoch": 0.18, "learning_rate": 1.8892895106458255e-05, "loss": 2.2227, "step": 7168 }, { "epoch": 0.18, "learning_rate": 1.88925299656172e-05, "loss": 2.3574, "step": 7169 }, { "epoch": 0.18, "learning_rate": 1.8892164768101152e-05, "loss": 2.1528, "step": 7170 }, { "epoch": 0.18, "learning_rate": 1.889179951391244e-05, "loss": 2.2865, "step": 7171 }, { "epoch": 0.18, "learning_rate": 1.8891434203053394e-05, "loss": 2.2419, "step": 7172 }, { "epoch": 0.18, "learning_rate": 1.889106883552634e-05, "loss": 2.1241, "step": 7173 }, { "epoch": 0.18, "learning_rate": 1.8890703411333606e-05, "loss": 2.411, "step": 7174 }, { "epoch": 0.18, "learning_rate": 1.889033793047752e-05, "loss": 2.254, "step": 7175 }, { "epoch": 0.18, "learning_rate": 1.8889972392960418e-05, "loss": 2.4071, "step": 7176 }, { "epoch": 0.18, "learning_rate": 1.8889606798784622e-05, "loss": 2.1247, "step": 7177 }, { "epoch": 0.18, "learning_rate": 1.8889241147952462e-05, "loss": 2.3706, "step": 7178 }, { "epoch": 0.18, "learning_rate": 1.8888875440466273e-05, "loss": 2.1532, "step": 7179 }, { "epoch": 0.18, "learning_rate": 1.8888509676328387e-05, "loss": 2.0884, "step": 7180 }, { "epoch": 0.18, "learning_rate": 1.8888143855541126e-05, "loss": 2.2837, "step": 7181 }, { "epoch": 0.18, "learning_rate": 1.8887777978106836e-05, "loss": 2.2181, "step": 7182 }, { "epoch": 0.18, "learning_rate": 1.888741204402783e-05, "loss": 2.2115, "step": 7183 }, { "epoch": 0.18, "learning_rate": 1.8887046053306463e-05, "loss": 2.1739, "step": 7184 }, { "epoch": 0.18, "learning_rate": 1.8886680005945045e-05, "loss": 2.2396, "step": 7185 }, { "epoch": 0.18, "learning_rate": 1.8886313901945925e-05, "loss": 2.32, "step": 7186 }, { "epoch": 0.18, "learning_rate": 1.888594774131143e-05, "loss": 2.1731, "step": 7187 }, { "epoch": 0.18, "learning_rate": 1.8885581524043894e-05, "loss": 2.2371, "step": 7188 }, { "epoch": 0.18, "learning_rate": 1.888521525014565e-05, "loss": 2.3818, "step": 7189 }, { "epoch": 0.18, "learning_rate": 1.8884848919619033e-05, "loss": 2.1453, "step": 7190 }, { "epoch": 0.18, "learning_rate": 1.888448253246638e-05, "loss": 2.3304, "step": 7191 }, { "epoch": 0.18, "learning_rate": 1.8884116088690023e-05, "loss": 2.2582, "step": 7192 }, { "epoch": 0.18, "learning_rate": 1.88837495882923e-05, "loss": 2.2831, "step": 7193 }, { "epoch": 0.18, "learning_rate": 1.8883383031275543e-05, "loss": 2.3296, "step": 7194 }, { "epoch": 0.18, "learning_rate": 1.8883016417642096e-05, "loss": 2.2981, "step": 7195 }, { "epoch": 0.18, "learning_rate": 1.888264974739429e-05, "loss": 2.3143, "step": 7196 }, { "epoch": 0.18, "learning_rate": 1.8882283020534458e-05, "loss": 2.2715, "step": 7197 }, { "epoch": 0.18, "learning_rate": 1.8881916237064942e-05, "loss": 2.1992, "step": 7198 }, { "epoch": 0.18, "learning_rate": 1.8881549396988082e-05, "loss": 2.1722, "step": 7199 }, { "epoch": 0.18, "learning_rate": 1.8881182500306213e-05, "loss": 2.2055, "step": 7200 }, { "epoch": 0.18, "learning_rate": 1.888081554702167e-05, "loss": 2.291, "step": 7201 }, { "epoch": 0.18, "learning_rate": 1.8880448537136803e-05, "loss": 2.3471, "step": 7202 }, { "epoch": 0.18, "learning_rate": 1.8880081470653935e-05, "loss": 2.1187, "step": 7203 }, { "epoch": 0.18, "learning_rate": 1.887971434757542e-05, "loss": 2.2285, "step": 7204 }, { "epoch": 0.18, "learning_rate": 1.8879347167903586e-05, "loss": 2.1559, "step": 7205 }, { "epoch": 0.18, "learning_rate": 1.8878979931640784e-05, "loss": 2.3447, "step": 7206 }, { "epoch": 0.18, "learning_rate": 1.887861263878935e-05, "loss": 2.2314, "step": 7207 }, { "epoch": 0.18, "learning_rate": 1.887824528935162e-05, "loss": 2.2195, "step": 7208 }, { "epoch": 0.18, "learning_rate": 1.887787788332994e-05, "loss": 2.2204, "step": 7209 }, { "epoch": 0.18, "learning_rate": 1.8877510420726653e-05, "loss": 2.1976, "step": 7210 }, { "epoch": 0.18, "learning_rate": 1.88771429015441e-05, "loss": 2.0971, "step": 7211 }, { "epoch": 0.18, "learning_rate": 1.887677532578462e-05, "loss": 2.1288, "step": 7212 }, { "epoch": 0.18, "learning_rate": 1.8876407693450556e-05, "loss": 2.2009, "step": 7213 }, { "epoch": 0.18, "learning_rate": 1.8876040004544257e-05, "loss": 2.3376, "step": 7214 }, { "epoch": 0.18, "learning_rate": 1.8875672259068062e-05, "loss": 2.5665, "step": 7215 }, { "epoch": 0.18, "learning_rate": 1.8875304457024314e-05, "loss": 2.1469, "step": 7216 }, { "epoch": 0.18, "learning_rate": 1.8874936598415356e-05, "loss": 1.9946, "step": 7217 }, { "epoch": 0.18, "learning_rate": 1.8874568683243538e-05, "loss": 2.1869, "step": 7218 }, { "epoch": 0.18, "learning_rate": 1.8874200711511202e-05, "loss": 2.2832, "step": 7219 }, { "epoch": 0.18, "learning_rate": 1.8873832683220688e-05, "loss": 2.1542, "step": 7220 }, { "epoch": 0.18, "learning_rate": 1.8873464598374353e-05, "loss": 2.3825, "step": 7221 }, { "epoch": 0.18, "learning_rate": 1.8873096456974532e-05, "loss": 2.1103, "step": 7222 }, { "epoch": 0.18, "learning_rate": 1.8872728259023576e-05, "loss": 2.2248, "step": 7223 }, { "epoch": 0.18, "learning_rate": 1.8872360004523833e-05, "loss": 2.0353, "step": 7224 }, { "epoch": 0.18, "learning_rate": 1.8871991693477648e-05, "loss": 2.4077, "step": 7225 }, { "epoch": 0.18, "learning_rate": 1.887162332588737e-05, "loss": 2.3132, "step": 7226 }, { "epoch": 0.18, "learning_rate": 1.8871254901755342e-05, "loss": 2.269, "step": 7227 }, { "epoch": 0.18, "learning_rate": 1.8870886421083918e-05, "loss": 2.2392, "step": 7228 }, { "epoch": 0.18, "learning_rate": 1.8870517883875443e-05, "loss": 2.3284, "step": 7229 }, { "epoch": 0.18, "learning_rate": 1.8870149290132266e-05, "loss": 2.2951, "step": 7230 }, { "epoch": 0.18, "learning_rate": 1.8869780639856737e-05, "loss": 2.2142, "step": 7231 }, { "epoch": 0.18, "learning_rate": 1.8869411933051206e-05, "loss": 2.2264, "step": 7232 }, { "epoch": 0.18, "learning_rate": 1.8869043169718023e-05, "loss": 2.4202, "step": 7233 }, { "epoch": 0.18, "learning_rate": 1.8868674349859535e-05, "loss": 2.3131, "step": 7234 }, { "epoch": 0.18, "learning_rate": 1.8868305473478096e-05, "loss": 2.1866, "step": 7235 }, { "epoch": 0.18, "learning_rate": 1.886793654057606e-05, "loss": 2.2344, "step": 7236 }, { "epoch": 0.18, "learning_rate": 1.886756755115577e-05, "loss": 2.1529, "step": 7237 }, { "epoch": 0.18, "learning_rate": 1.8867198505219583e-05, "loss": 2.405, "step": 7238 }, { "epoch": 0.18, "learning_rate": 1.886682940276985e-05, "loss": 2.2402, "step": 7239 }, { "epoch": 0.18, "learning_rate": 1.8866460243808927e-05, "loss": 2.2502, "step": 7240 }, { "epoch": 0.18, "learning_rate": 1.886609102833916e-05, "loss": 2.3441, "step": 7241 }, { "epoch": 0.18, "learning_rate": 1.8865721756362907e-05, "loss": 2.3299, "step": 7242 }, { "epoch": 0.18, "learning_rate": 1.8865352427882518e-05, "loss": 2.2367, "step": 7243 }, { "epoch": 0.18, "learning_rate": 1.8864983042900352e-05, "loss": 2.2498, "step": 7244 }, { "epoch": 0.18, "learning_rate": 1.8864613601418757e-05, "loss": 2.173, "step": 7245 }, { "epoch": 0.18, "learning_rate": 1.886424410344009e-05, "loss": 2.2483, "step": 7246 }, { "epoch": 0.18, "learning_rate": 1.8863874548966708e-05, "loss": 2.3195, "step": 7247 }, { "epoch": 0.18, "learning_rate": 1.8863504938000966e-05, "loss": 2.1435, "step": 7248 }, { "epoch": 0.18, "learning_rate": 1.8863135270545218e-05, "loss": 2.3351, "step": 7249 }, { "epoch": 0.18, "learning_rate": 1.886276554660182e-05, "loss": 2.1537, "step": 7250 }, { "epoch": 0.18, "learning_rate": 1.886239576617313e-05, "loss": 2.2925, "step": 7251 }, { "epoch": 0.18, "learning_rate": 1.88620259292615e-05, "loss": 2.3029, "step": 7252 }, { "epoch": 0.18, "learning_rate": 1.8861656035869296e-05, "loss": 2.173, "step": 7253 }, { "epoch": 0.18, "learning_rate": 1.8861286085998866e-05, "loss": 1.9558, "step": 7254 }, { "epoch": 0.18, "learning_rate": 1.8860916079652574e-05, "loss": 2.2901, "step": 7255 }, { "epoch": 0.18, "learning_rate": 1.8860546016832777e-05, "loss": 2.282, "step": 7256 }, { "epoch": 0.18, "learning_rate": 1.886017589754183e-05, "loss": 2.3852, "step": 7257 }, { "epoch": 0.18, "learning_rate": 1.8859805721782096e-05, "loss": 2.1253, "step": 7258 }, { "epoch": 0.18, "learning_rate": 1.8859435489555933e-05, "loss": 2.2402, "step": 7259 }, { "epoch": 0.18, "learning_rate": 1.88590652008657e-05, "loss": 2.3347, "step": 7260 }, { "epoch": 0.18, "learning_rate": 1.8858694855713757e-05, "loss": 2.1549, "step": 7261 }, { "epoch": 0.18, "learning_rate": 1.8858324454102465e-05, "loss": 2.3076, "step": 7262 }, { "epoch": 0.18, "learning_rate": 1.8857953996034184e-05, "loss": 2.4118, "step": 7263 }, { "epoch": 0.18, "learning_rate": 1.8857583481511275e-05, "loss": 2.2478, "step": 7264 }, { "epoch": 0.18, "learning_rate": 1.8857212910536104e-05, "loss": 2.1943, "step": 7265 }, { "epoch": 0.18, "learning_rate": 1.8856842283111025e-05, "loss": 2.2455, "step": 7266 }, { "epoch": 0.18, "learning_rate": 1.8856471599238404e-05, "loss": 2.1165, "step": 7267 }, { "epoch": 0.18, "learning_rate": 1.8856100858920605e-05, "loss": 2.3937, "step": 7268 }, { "epoch": 0.18, "learning_rate": 1.8855730062159986e-05, "loss": 2.0724, "step": 7269 }, { "epoch": 0.18, "learning_rate": 1.8855359208958915e-05, "loss": 2.3577, "step": 7270 }, { "epoch": 0.18, "learning_rate": 1.8854988299319755e-05, "loss": 2.1315, "step": 7271 }, { "epoch": 0.18, "learning_rate": 1.8854617333244866e-05, "loss": 2.163, "step": 7272 }, { "epoch": 0.18, "learning_rate": 1.885424631073662e-05, "loss": 2.2691, "step": 7273 }, { "epoch": 0.18, "learning_rate": 1.885387523179737e-05, "loss": 2.1594, "step": 7274 }, { "epoch": 0.18, "learning_rate": 1.8853504096429493e-05, "loss": 2.2278, "step": 7275 }, { "epoch": 0.18, "learning_rate": 1.8853132904635347e-05, "loss": 2.1944, "step": 7276 }, { "epoch": 0.18, "learning_rate": 1.88527616564173e-05, "loss": 2.1749, "step": 7277 }, { "epoch": 0.18, "learning_rate": 1.8852390351777718e-05, "loss": 2.2379, "step": 7278 }, { "epoch": 0.18, "learning_rate": 1.885201899071897e-05, "loss": 2.379, "step": 7279 }, { "epoch": 0.18, "learning_rate": 1.8851647573243418e-05, "loss": 2.37, "step": 7280 }, { "epoch": 0.18, "learning_rate": 1.885127609935343e-05, "loss": 2.1015, "step": 7281 }, { "epoch": 0.18, "learning_rate": 1.8850904569051374e-05, "loss": 2.2694, "step": 7282 }, { "epoch": 0.18, "learning_rate": 1.8850532982339623e-05, "loss": 2.1993, "step": 7283 }, { "epoch": 0.18, "learning_rate": 1.885016133922054e-05, "loss": 2.4222, "step": 7284 }, { "epoch": 0.18, "learning_rate": 1.884978963969649e-05, "loss": 2.3578, "step": 7285 }, { "epoch": 0.18, "learning_rate": 1.8849417883769853e-05, "loss": 2.3411, "step": 7286 }, { "epoch": 0.18, "learning_rate": 1.8849046071442988e-05, "loss": 2.3162, "step": 7287 }, { "epoch": 0.18, "learning_rate": 1.8848674202718268e-05, "loss": 2.2582, "step": 7288 }, { "epoch": 0.18, "learning_rate": 1.8848302277598063e-05, "loss": 2.2046, "step": 7289 }, { "epoch": 0.18, "learning_rate": 1.8847930296084746e-05, "loss": 2.1579, "step": 7290 }, { "epoch": 0.18, "learning_rate": 1.8847558258180688e-05, "loss": 2.2175, "step": 7291 }, { "epoch": 0.18, "learning_rate": 1.8847186163888254e-05, "loss": 2.2114, "step": 7292 }, { "epoch": 0.18, "learning_rate": 1.884681401320982e-05, "loss": 2.2173, "step": 7293 }, { "epoch": 0.18, "learning_rate": 1.884644180614776e-05, "loss": 1.9411, "step": 7294 }, { "epoch": 0.18, "learning_rate": 1.884606954270444e-05, "loss": 2.2447, "step": 7295 }, { "epoch": 0.18, "learning_rate": 1.884569722288224e-05, "loss": 2.1587, "step": 7296 }, { "epoch": 0.18, "learning_rate": 1.8845324846683522e-05, "loss": 2.2227, "step": 7297 }, { "epoch": 0.18, "learning_rate": 1.8844952414110673e-05, "loss": 2.1405, "step": 7298 }, { "epoch": 0.18, "learning_rate": 1.8844579925166056e-05, "loss": 2.283, "step": 7299 }, { "epoch": 0.18, "learning_rate": 1.884420737985205e-05, "loss": 2.1463, "step": 7300 }, { "epoch": 0.18, "learning_rate": 1.884383477817103e-05, "loss": 2.1712, "step": 7301 }, { "epoch": 0.18, "learning_rate": 1.8843462120125365e-05, "loss": 2.3264, "step": 7302 }, { "epoch": 0.18, "learning_rate": 1.8843089405717437e-05, "loss": 2.2026, "step": 7303 }, { "epoch": 0.18, "learning_rate": 1.8842716634949618e-05, "loss": 2.3192, "step": 7304 }, { "epoch": 0.18, "learning_rate": 1.8842343807824285e-05, "loss": 2.151, "step": 7305 }, { "epoch": 0.18, "learning_rate": 1.884197092434381e-05, "loss": 2.2715, "step": 7306 }, { "epoch": 0.18, "learning_rate": 1.8841597984510573e-05, "loss": 2.3362, "step": 7307 }, { "epoch": 0.18, "learning_rate": 1.8841224988326954e-05, "loss": 2.2852, "step": 7308 }, { "epoch": 0.18, "learning_rate": 1.8840851935795323e-05, "loss": 2.3301, "step": 7309 }, { "epoch": 0.18, "learning_rate": 1.8840478826918066e-05, "loss": 2.2658, "step": 7310 }, { "epoch": 0.18, "learning_rate": 1.884010566169755e-05, "loss": 2.1501, "step": 7311 }, { "epoch": 0.18, "learning_rate": 1.8839732440136167e-05, "loss": 2.2581, "step": 7312 }, { "epoch": 0.18, "learning_rate": 1.8839359162236285e-05, "loss": 2.2243, "step": 7313 }, { "epoch": 0.18, "learning_rate": 1.8838985828000286e-05, "loss": 2.1005, "step": 7314 }, { "epoch": 0.18, "learning_rate": 1.883861243743055e-05, "loss": 2.0702, "step": 7315 }, { "epoch": 0.18, "learning_rate": 1.8838238990529455e-05, "loss": 2.2387, "step": 7316 }, { "epoch": 0.18, "learning_rate": 1.883786548729939e-05, "loss": 2.241, "step": 7317 }, { "epoch": 0.18, "learning_rate": 1.8837491927742718e-05, "loss": 2.0768, "step": 7318 }, { "epoch": 0.18, "learning_rate": 1.8837118311861836e-05, "loss": 2.2465, "step": 7319 }, { "epoch": 0.18, "learning_rate": 1.8836744639659118e-05, "loss": 2.3863, "step": 7320 }, { "epoch": 0.18, "learning_rate": 1.8836370911136942e-05, "loss": 2.2753, "step": 7321 }, { "epoch": 0.18, "learning_rate": 1.8835997126297697e-05, "loss": 2.1716, "step": 7322 }, { "epoch": 0.18, "learning_rate": 1.8835623285143768e-05, "loss": 2.3497, "step": 7323 }, { "epoch": 0.18, "learning_rate": 1.8835249387677527e-05, "loss": 2.1546, "step": 7324 }, { "epoch": 0.18, "learning_rate": 1.883487543390136e-05, "loss": 2.2938, "step": 7325 }, { "epoch": 0.18, "learning_rate": 1.8834501423817658e-05, "loss": 2.1533, "step": 7326 }, { "epoch": 0.18, "learning_rate": 1.88341273574288e-05, "loss": 2.1031, "step": 7327 }, { "epoch": 0.18, "learning_rate": 1.8833753234737167e-05, "loss": 2.1986, "step": 7328 }, { "epoch": 0.18, "learning_rate": 1.8833379055745147e-05, "loss": 2.1423, "step": 7329 }, { "epoch": 0.18, "learning_rate": 1.8833004820455122e-05, "loss": 2.0833, "step": 7330 }, { "epoch": 0.18, "learning_rate": 1.8832630528869478e-05, "loss": 2.2517, "step": 7331 }, { "epoch": 0.18, "learning_rate": 1.8832256180990604e-05, "loss": 2.0317, "step": 7332 }, { "epoch": 0.18, "learning_rate": 1.8831881776820878e-05, "loss": 2.3841, "step": 7333 }, { "epoch": 0.18, "learning_rate": 1.8831507316362696e-05, "loss": 2.2289, "step": 7334 }, { "epoch": 0.18, "learning_rate": 1.883113279961844e-05, "loss": 2.2923, "step": 7335 }, { "epoch": 0.18, "learning_rate": 1.8830758226590493e-05, "loss": 2.389, "step": 7336 }, { "epoch": 0.18, "learning_rate": 1.883038359728125e-05, "loss": 2.1843, "step": 7337 }, { "epoch": 0.18, "learning_rate": 1.883000891169309e-05, "loss": 2.1853, "step": 7338 }, { "epoch": 0.18, "learning_rate": 1.882963416982841e-05, "loss": 2.1879, "step": 7339 }, { "epoch": 0.18, "learning_rate": 1.882925937168959e-05, "loss": 2.2992, "step": 7340 }, { "epoch": 0.18, "learning_rate": 1.8828884517279028e-05, "loss": 2.2521, "step": 7341 }, { "epoch": 0.18, "learning_rate": 1.8828509606599104e-05, "loss": 2.1744, "step": 7342 }, { "epoch": 0.18, "learning_rate": 1.882813463965221e-05, "loss": 2.0871, "step": 7343 }, { "epoch": 0.18, "learning_rate": 1.882775961644074e-05, "loss": 2.2063, "step": 7344 }, { "epoch": 0.18, "learning_rate": 1.882738453696708e-05, "loss": 2.2845, "step": 7345 }, { "epoch": 0.18, "learning_rate": 1.882700940123362e-05, "loss": 2.1954, "step": 7346 }, { "epoch": 0.18, "learning_rate": 1.8826634209242756e-05, "loss": 2.1928, "step": 7347 }, { "epoch": 0.18, "learning_rate": 1.882625896099687e-05, "loss": 2.3067, "step": 7348 }, { "epoch": 0.18, "learning_rate": 1.8825883656498363e-05, "loss": 2.2689, "step": 7349 }, { "epoch": 0.18, "learning_rate": 1.8825508295749624e-05, "loss": 2.1652, "step": 7350 }, { "epoch": 0.18, "learning_rate": 1.8825132878753045e-05, "loss": 2.2555, "step": 7351 }, { "epoch": 0.18, "learning_rate": 1.8824757405511015e-05, "loss": 2.1361, "step": 7352 }, { "epoch": 0.18, "learning_rate": 1.8824381876025933e-05, "loss": 2.2506, "step": 7353 }, { "epoch": 0.18, "learning_rate": 1.8824006290300188e-05, "loss": 2.3009, "step": 7354 }, { "epoch": 0.18, "learning_rate": 1.8823630648336175e-05, "loss": 2.3464, "step": 7355 }, { "epoch": 0.18, "learning_rate": 1.882325495013629e-05, "loss": 2.3188, "step": 7356 }, { "epoch": 0.18, "learning_rate": 1.8822879195702922e-05, "loss": 2.2448, "step": 7357 }, { "epoch": 0.18, "learning_rate": 1.8822503385038474e-05, "loss": 2.0531, "step": 7358 }, { "epoch": 0.18, "learning_rate": 1.8822127518145335e-05, "loss": 2.1565, "step": 7359 }, { "epoch": 0.18, "learning_rate": 1.88217515950259e-05, "loss": 2.1868, "step": 7360 }, { "epoch": 0.18, "learning_rate": 1.882137561568257e-05, "loss": 2.2188, "step": 7361 }, { "epoch": 0.18, "learning_rate": 1.882099958011774e-05, "loss": 2.2619, "step": 7362 }, { "epoch": 0.18, "learning_rate": 1.88206234883338e-05, "loss": 2.3974, "step": 7363 }, { "epoch": 0.18, "learning_rate": 1.8820247340333154e-05, "loss": 2.2531, "step": 7364 }, { "epoch": 0.18, "learning_rate": 1.8819871136118198e-05, "loss": 2.2782, "step": 7365 }, { "epoch": 0.18, "learning_rate": 1.881949487569133e-05, "loss": 2.2523, "step": 7366 }, { "epoch": 0.18, "learning_rate": 1.8819118559054945e-05, "loss": 2.0687, "step": 7367 }, { "epoch": 0.18, "learning_rate": 1.8818742186211443e-05, "loss": 2.1684, "step": 7368 }, { "epoch": 0.18, "learning_rate": 1.8818365757163226e-05, "loss": 2.2349, "step": 7369 }, { "epoch": 0.18, "learning_rate": 1.8817989271912685e-05, "loss": 2.2612, "step": 7370 }, { "epoch": 0.18, "learning_rate": 1.881761273046223e-05, "loss": 2.1883, "step": 7371 }, { "epoch": 0.18, "learning_rate": 1.8817236132814253e-05, "loss": 2.1926, "step": 7372 }, { "epoch": 0.18, "learning_rate": 1.881685947897116e-05, "loss": 2.3795, "step": 7373 }, { "epoch": 0.18, "learning_rate": 1.8816482768935343e-05, "loss": 2.0895, "step": 7374 }, { "epoch": 0.18, "learning_rate": 1.8816106002709212e-05, "loss": 2.271, "step": 7375 }, { "epoch": 0.18, "learning_rate": 1.8815729180295164e-05, "loss": 2.2773, "step": 7376 }, { "epoch": 0.18, "learning_rate": 1.88153523016956e-05, "loss": 2.1859, "step": 7377 }, { "epoch": 0.18, "learning_rate": 1.881497536691292e-05, "loss": 2.2193, "step": 7378 }, { "epoch": 0.18, "learning_rate": 1.881459837594953e-05, "loss": 2.3676, "step": 7379 }, { "epoch": 0.18, "learning_rate": 1.8814221328807837e-05, "loss": 2.2112, "step": 7380 }, { "epoch": 0.18, "learning_rate": 1.8813844225490236e-05, "loss": 2.352, "step": 7381 }, { "epoch": 0.18, "learning_rate": 1.8813467065999135e-05, "loss": 2.037, "step": 7382 }, { "epoch": 0.18, "learning_rate": 1.8813089850336933e-05, "loss": 2.2308, "step": 7383 }, { "epoch": 0.18, "learning_rate": 1.881271257850604e-05, "loss": 2.1367, "step": 7384 }, { "epoch": 0.18, "learning_rate": 1.8812335250508854e-05, "loss": 2.1942, "step": 7385 }, { "epoch": 0.18, "learning_rate": 1.8811957866347787e-05, "loss": 2.3512, "step": 7386 }, { "epoch": 0.18, "learning_rate": 1.8811580426025243e-05, "loss": 2.2357, "step": 7387 }, { "epoch": 0.18, "learning_rate": 1.8811202929543622e-05, "loss": 2.2346, "step": 7388 }, { "epoch": 0.18, "learning_rate": 1.8810825376905333e-05, "loss": 2.187, "step": 7389 }, { "epoch": 0.18, "learning_rate": 1.8810447768112784e-05, "loss": 2.2118, "step": 7390 }, { "epoch": 0.18, "learning_rate": 1.8810070103168375e-05, "loss": 2.1189, "step": 7391 }, { "epoch": 0.18, "learning_rate": 1.8809692382074525e-05, "loss": 2.3107, "step": 7392 }, { "epoch": 0.18, "learning_rate": 1.8809314604833632e-05, "loss": 2.1499, "step": 7393 }, { "epoch": 0.18, "learning_rate": 1.8808936771448105e-05, "loss": 2.2642, "step": 7394 }, { "epoch": 0.18, "learning_rate": 1.8808558881920357e-05, "loss": 2.2572, "step": 7395 }, { "epoch": 0.18, "learning_rate": 1.880818093625279e-05, "loss": 2.0911, "step": 7396 }, { "epoch": 0.18, "learning_rate": 1.8807802934447813e-05, "loss": 2.1809, "step": 7397 }, { "epoch": 0.18, "learning_rate": 1.8807424876507842e-05, "loss": 2.2621, "step": 7398 }, { "epoch": 0.18, "learning_rate": 1.8807046762435278e-05, "loss": 2.3822, "step": 7399 }, { "epoch": 0.18, "learning_rate": 1.8806668592232538e-05, "loss": 2.1351, "step": 7400 }, { "epoch": 0.18, "learning_rate": 1.8806290365902024e-05, "loss": 2.2064, "step": 7401 }, { "epoch": 0.18, "learning_rate": 1.8805912083446157e-05, "loss": 2.2607, "step": 7402 }, { "epoch": 0.18, "learning_rate": 1.880553374486734e-05, "loss": 2.2516, "step": 7403 }, { "epoch": 0.18, "learning_rate": 1.880515535016799e-05, "loss": 2.1199, "step": 7404 }, { "epoch": 0.18, "learning_rate": 1.880477689935051e-05, "loss": 2.2551, "step": 7405 }, { "epoch": 0.18, "learning_rate": 1.8804398392417324e-05, "loss": 2.2541, "step": 7406 }, { "epoch": 0.18, "learning_rate": 1.8804019829370833e-05, "loss": 2.2728, "step": 7407 }, { "epoch": 0.18, "learning_rate": 1.8803641210213455e-05, "loss": 2.1696, "step": 7408 }, { "epoch": 0.18, "learning_rate": 1.88032625349476e-05, "loss": 2.3097, "step": 7409 }, { "epoch": 0.18, "learning_rate": 1.8802883803575687e-05, "loss": 2.2503, "step": 7410 }, { "epoch": 0.18, "learning_rate": 1.880250501610013e-05, "loss": 1.9255, "step": 7411 }, { "epoch": 0.18, "learning_rate": 1.8802126172523337e-05, "loss": 2.0505, "step": 7412 }, { "epoch": 0.18, "learning_rate": 1.8801747272847722e-05, "loss": 2.3205, "step": 7413 }, { "epoch": 0.18, "learning_rate": 1.8801368317075706e-05, "loss": 2.1532, "step": 7414 }, { "epoch": 0.18, "learning_rate": 1.8800989305209706e-05, "loss": 2.2113, "step": 7415 }, { "epoch": 0.18, "learning_rate": 1.8800610237252128e-05, "loss": 2.0751, "step": 7416 }, { "epoch": 0.18, "learning_rate": 1.8800231113205392e-05, "loss": 2.2032, "step": 7417 }, { "epoch": 0.18, "learning_rate": 1.8799851933071917e-05, "loss": 2.2912, "step": 7418 }, { "epoch": 0.18, "learning_rate": 1.879947269685412e-05, "loss": 2.0517, "step": 7419 }, { "epoch": 0.18, "learning_rate": 1.8799093404554413e-05, "loss": 2.2932, "step": 7420 }, { "epoch": 0.18, "learning_rate": 1.8798714056175217e-05, "loss": 2.2958, "step": 7421 }, { "epoch": 0.18, "learning_rate": 1.879833465171895e-05, "loss": 2.0792, "step": 7422 }, { "epoch": 0.18, "learning_rate": 1.8797955191188027e-05, "loss": 2.2419, "step": 7423 }, { "epoch": 0.18, "learning_rate": 1.879757567458487e-05, "loss": 1.9803, "step": 7424 }, { "epoch": 0.18, "learning_rate": 1.8797196101911893e-05, "loss": 2.2006, "step": 7425 }, { "epoch": 0.18, "learning_rate": 1.8796816473171523e-05, "loss": 2.3201, "step": 7426 }, { "epoch": 0.18, "learning_rate": 1.879643678836617e-05, "loss": 2.3494, "step": 7427 }, { "epoch": 0.18, "learning_rate": 1.879605704749826e-05, "loss": 2.2179, "step": 7428 }, { "epoch": 0.18, "learning_rate": 1.8795677250570216e-05, "loss": 2.1479, "step": 7429 }, { "epoch": 0.18, "learning_rate": 1.879529739758445e-05, "loss": 2.1543, "step": 7430 }, { "epoch": 0.18, "learning_rate": 1.8794917488543388e-05, "loss": 2.2863, "step": 7431 }, { "epoch": 0.18, "learning_rate": 1.8794537523449454e-05, "loss": 2.2523, "step": 7432 }, { "epoch": 0.18, "learning_rate": 1.8794157502305062e-05, "loss": 2.3011, "step": 7433 }, { "epoch": 0.18, "learning_rate": 1.879377742511264e-05, "loss": 2.2443, "step": 7434 }, { "epoch": 0.18, "learning_rate": 1.8793397291874608e-05, "loss": 2.3371, "step": 7435 }, { "epoch": 0.18, "learning_rate": 1.879301710259339e-05, "loss": 2.2711, "step": 7436 }, { "epoch": 0.18, "learning_rate": 1.879263685727141e-05, "loss": 2.1785, "step": 7437 }, { "epoch": 0.18, "learning_rate": 1.8792256555911087e-05, "loss": 2.1486, "step": 7438 }, { "epoch": 0.18, "learning_rate": 1.879187619851485e-05, "loss": 2.2381, "step": 7439 }, { "epoch": 0.18, "learning_rate": 1.879149578508512e-05, "loss": 2.2234, "step": 7440 }, { "epoch": 0.18, "learning_rate": 1.8791115315624322e-05, "loss": 2.3538, "step": 7441 }, { "epoch": 0.18, "learning_rate": 1.879073479013488e-05, "loss": 2.2423, "step": 7442 }, { "epoch": 0.18, "learning_rate": 1.8790354208619224e-05, "loss": 2.2339, "step": 7443 }, { "epoch": 0.18, "learning_rate": 1.878997357107977e-05, "loss": 2.3725, "step": 7444 }, { "epoch": 0.18, "learning_rate": 1.8789592877518956e-05, "loss": 2.0411, "step": 7445 }, { "epoch": 0.18, "learning_rate": 1.87892121279392e-05, "loss": 2.133, "step": 7446 }, { "epoch": 0.18, "learning_rate": 1.8788831322342932e-05, "loss": 2.2249, "step": 7447 }, { "epoch": 0.18, "learning_rate": 1.8788450460732575e-05, "loss": 2.2034, "step": 7448 }, { "epoch": 0.18, "learning_rate": 1.8788069543110562e-05, "loss": 2.2706, "step": 7449 }, { "epoch": 0.18, "learning_rate": 1.8787688569479316e-05, "loss": 2.3648, "step": 7450 }, { "epoch": 0.18, "learning_rate": 1.8787307539841265e-05, "loss": 2.1534, "step": 7451 }, { "epoch": 0.18, "learning_rate": 1.8786926454198842e-05, "loss": 2.1816, "step": 7452 }, { "epoch": 0.18, "learning_rate": 1.8786545312554473e-05, "loss": 2.2809, "step": 7453 }, { "epoch": 0.18, "learning_rate": 1.8786164114910585e-05, "loss": 2.2741, "step": 7454 }, { "epoch": 0.18, "learning_rate": 1.8785782861269612e-05, "loss": 2.2535, "step": 7455 }, { "epoch": 0.18, "learning_rate": 1.8785401551633982e-05, "loss": 2.3065, "step": 7456 }, { "epoch": 0.18, "learning_rate": 1.8785020186006123e-05, "loss": 2.2106, "step": 7457 }, { "epoch": 0.18, "learning_rate": 1.878463876438847e-05, "loss": 1.994, "step": 7458 }, { "epoch": 0.18, "learning_rate": 1.878425728678345e-05, "loss": 2.303, "step": 7459 }, { "epoch": 0.18, "learning_rate": 1.8783875753193494e-05, "loss": 2.2442, "step": 7460 }, { "epoch": 0.18, "learning_rate": 1.8783494163621037e-05, "loss": 2.2778, "step": 7461 }, { "epoch": 0.18, "learning_rate": 1.8783112518068508e-05, "loss": 2.1905, "step": 7462 }, { "epoch": 0.18, "learning_rate": 1.8782730816538343e-05, "loss": 2.266, "step": 7463 }, { "epoch": 0.18, "learning_rate": 1.878234905903297e-05, "loss": 2.1939, "step": 7464 }, { "epoch": 0.18, "learning_rate": 1.8781967245554825e-05, "loss": 2.1204, "step": 7465 }, { "epoch": 0.18, "learning_rate": 1.878158537610634e-05, "loss": 2.2934, "step": 7466 }, { "epoch": 0.18, "learning_rate": 1.8781203450689952e-05, "loss": 2.2969, "step": 7467 }, { "epoch": 0.18, "learning_rate": 1.878082146930809e-05, "loss": 2.0363, "step": 7468 }, { "epoch": 0.18, "learning_rate": 1.8780439431963192e-05, "loss": 2.1069, "step": 7469 }, { "epoch": 0.18, "learning_rate": 1.8780057338657693e-05, "loss": 2.3157, "step": 7470 }, { "epoch": 0.18, "learning_rate": 1.8779675189394027e-05, "loss": 2.1779, "step": 7471 }, { "epoch": 0.18, "learning_rate": 1.877929298417463e-05, "loss": 2.2038, "step": 7472 }, { "epoch": 0.18, "learning_rate": 1.8778910723001938e-05, "loss": 2.2479, "step": 7473 }, { "epoch": 0.18, "learning_rate": 1.8778528405878384e-05, "loss": 2.1915, "step": 7474 }, { "epoch": 0.18, "learning_rate": 1.877814603280641e-05, "loss": 2.1496, "step": 7475 }, { "epoch": 0.18, "learning_rate": 1.877776360378845e-05, "loss": 2.2449, "step": 7476 }, { "epoch": 0.18, "learning_rate": 1.8777381118826944e-05, "loss": 2.2565, "step": 7477 }, { "epoch": 0.18, "learning_rate": 1.8776998577924327e-05, "loss": 2.364, "step": 7478 }, { "epoch": 0.18, "learning_rate": 1.8776615981083037e-05, "loss": 2.1544, "step": 7479 }, { "epoch": 0.18, "learning_rate": 1.877623332830551e-05, "loss": 2.2722, "step": 7480 }, { "epoch": 0.18, "learning_rate": 1.877585061959419e-05, "loss": 2.2803, "step": 7481 }, { "epoch": 0.18, "learning_rate": 1.8775467854951516e-05, "loss": 2.3203, "step": 7482 }, { "epoch": 0.18, "learning_rate": 1.8775085034379924e-05, "loss": 2.1243, "step": 7483 }, { "epoch": 0.18, "learning_rate": 1.8774702157881856e-05, "loss": 2.0903, "step": 7484 }, { "epoch": 0.18, "learning_rate": 1.8774319225459753e-05, "loss": 2.1225, "step": 7485 }, { "epoch": 0.18, "learning_rate": 1.8773936237116054e-05, "loss": 2.3388, "step": 7486 }, { "epoch": 0.18, "learning_rate": 1.8773553192853197e-05, "loss": 2.238, "step": 7487 }, { "epoch": 0.18, "learning_rate": 1.8773170092673627e-05, "loss": 2.3596, "step": 7488 }, { "epoch": 0.18, "learning_rate": 1.8772786936579786e-05, "loss": 2.1942, "step": 7489 }, { "epoch": 0.18, "learning_rate": 1.8772403724574113e-05, "loss": 2.2384, "step": 7490 }, { "epoch": 0.18, "learning_rate": 1.8772020456659054e-05, "loss": 2.334, "step": 7491 }, { "epoch": 0.18, "learning_rate": 1.877163713283705e-05, "loss": 2.2018, "step": 7492 }, { "epoch": 0.18, "learning_rate": 1.877125375311054e-05, "loss": 2.2197, "step": 7493 }, { "epoch": 0.18, "learning_rate": 1.877087031748198e-05, "loss": 2.1314, "step": 7494 }, { "epoch": 0.18, "learning_rate": 1.8770486825953798e-05, "loss": 2.2376, "step": 7495 }, { "epoch": 0.18, "learning_rate": 1.8770103278528448e-05, "loss": 2.3122, "step": 7496 }, { "epoch": 0.18, "learning_rate": 1.876971967520837e-05, "loss": 2.1559, "step": 7497 }, { "epoch": 0.18, "learning_rate": 1.8769336015996012e-05, "loss": 2.1034, "step": 7498 }, { "epoch": 0.18, "learning_rate": 1.8768952300893816e-05, "loss": 2.2013, "step": 7499 }, { "epoch": 0.18, "learning_rate": 1.8768568529904233e-05, "loss": 2.2509, "step": 7500 }, { "epoch": 0.18, "learning_rate": 1.8768184703029703e-05, "loss": 2.3256, "step": 7501 }, { "epoch": 0.18, "learning_rate": 1.8767800820272673e-05, "loss": 2.3365, "step": 7502 }, { "epoch": 0.18, "learning_rate": 1.876741688163559e-05, "loss": 2.1216, "step": 7503 }, { "epoch": 0.18, "learning_rate": 1.8767032887120905e-05, "loss": 2.1099, "step": 7504 }, { "epoch": 0.18, "learning_rate": 1.8766648836731064e-05, "loss": 2.0252, "step": 7505 }, { "epoch": 0.19, "learning_rate": 1.876626473046851e-05, "loss": 1.9974, "step": 7506 }, { "epoch": 0.19, "learning_rate": 1.8765880568335694e-05, "loss": 2.2827, "step": 7507 }, { "epoch": 0.19, "learning_rate": 1.8765496350335064e-05, "loss": 2.2422, "step": 7508 }, { "epoch": 0.19, "learning_rate": 1.876511207646907e-05, "loss": 2.2282, "step": 7509 }, { "epoch": 0.19, "learning_rate": 1.8764727746740157e-05, "loss": 2.2414, "step": 7510 }, { "epoch": 0.19, "learning_rate": 1.8764343361150783e-05, "loss": 2.0798, "step": 7511 }, { "epoch": 0.19, "learning_rate": 1.876395891970339e-05, "loss": 2.1969, "step": 7512 }, { "epoch": 0.19, "learning_rate": 1.8763574422400433e-05, "loss": 2.266, "step": 7513 }, { "epoch": 0.19, "learning_rate": 1.876318986924436e-05, "loss": 2.1617, "step": 7514 }, { "epoch": 0.19, "learning_rate": 1.876280526023762e-05, "loss": 2.1754, "step": 7515 }, { "epoch": 0.19, "learning_rate": 1.8762420595382667e-05, "loss": 2.2577, "step": 7516 }, { "epoch": 0.19, "learning_rate": 1.8762035874681953e-05, "loss": 2.2095, "step": 7517 }, { "epoch": 0.19, "learning_rate": 1.8761651098137928e-05, "loss": 2.3703, "step": 7518 }, { "epoch": 0.19, "learning_rate": 1.8761266265753043e-05, "loss": 2.5165, "step": 7519 }, { "epoch": 0.19, "learning_rate": 1.8760881377529757e-05, "loss": 2.1482, "step": 7520 }, { "epoch": 0.19, "learning_rate": 1.8760496433470514e-05, "loss": 2.3282, "step": 7521 }, { "epoch": 0.19, "learning_rate": 1.8760111433577775e-05, "loss": 2.288, "step": 7522 }, { "epoch": 0.19, "learning_rate": 1.8759726377853992e-05, "loss": 2.316, "step": 7523 }, { "epoch": 0.19, "learning_rate": 1.8759341266301616e-05, "loss": 2.1796, "step": 7524 }, { "epoch": 0.19, "learning_rate": 1.875895609892311e-05, "loss": 2.2261, "step": 7525 }, { "epoch": 0.19, "learning_rate": 1.8758570875720915e-05, "loss": 2.2396, "step": 7526 }, { "epoch": 0.19, "learning_rate": 1.8758185596697494e-05, "loss": 2.4355, "step": 7527 }, { "epoch": 0.19, "learning_rate": 1.8757800261855304e-05, "loss": 2.049, "step": 7528 }, { "epoch": 0.19, "learning_rate": 1.8757414871196798e-05, "loss": 2.2476, "step": 7529 }, { "epoch": 0.19, "learning_rate": 1.8757029424724434e-05, "loss": 2.245, "step": 7530 }, { "epoch": 0.19, "learning_rate": 1.8756643922440665e-05, "loss": 2.2541, "step": 7531 }, { "epoch": 0.19, "learning_rate": 1.8756258364347953e-05, "loss": 2.2644, "step": 7532 }, { "epoch": 0.19, "learning_rate": 1.8755872750448753e-05, "loss": 2.2185, "step": 7533 }, { "epoch": 0.19, "learning_rate": 1.875548708074552e-05, "loss": 2.2999, "step": 7534 }, { "epoch": 0.19, "learning_rate": 1.8755101355240718e-05, "loss": 2.4805, "step": 7535 }, { "epoch": 0.19, "learning_rate": 1.8754715573936802e-05, "loss": 2.2009, "step": 7536 }, { "epoch": 0.19, "learning_rate": 1.8754329736836228e-05, "loss": 2.2389, "step": 7537 }, { "epoch": 0.19, "learning_rate": 1.875394384394146e-05, "loss": 2.2632, "step": 7538 }, { "epoch": 0.19, "learning_rate": 1.875355789525495e-05, "loss": 2.1051, "step": 7539 }, { "epoch": 0.19, "learning_rate": 1.875317189077917e-05, "loss": 2.1281, "step": 7540 }, { "epoch": 0.19, "learning_rate": 1.8752785830516567e-05, "loss": 2.1984, "step": 7541 }, { "epoch": 0.19, "learning_rate": 1.8752399714469612e-05, "loss": 2.2704, "step": 7542 }, { "epoch": 0.19, "learning_rate": 1.875201354264076e-05, "loss": 2.2624, "step": 7543 }, { "epoch": 0.19, "learning_rate": 1.875162731503247e-05, "loss": 2.3373, "step": 7544 }, { "epoch": 0.19, "learning_rate": 1.875124103164721e-05, "loss": 2.2342, "step": 7545 }, { "epoch": 0.19, "learning_rate": 1.875085469248744e-05, "loss": 2.2366, "step": 7546 }, { "epoch": 0.19, "learning_rate": 1.875046829755562e-05, "loss": 2.212, "step": 7547 }, { "epoch": 0.19, "learning_rate": 1.8750081846854213e-05, "loss": 2.2023, "step": 7548 }, { "epoch": 0.19, "learning_rate": 1.8749695340385683e-05, "loss": 2.4373, "step": 7549 }, { "epoch": 0.19, "learning_rate": 1.8749308778152494e-05, "loss": 2.3118, "step": 7550 }, { "epoch": 0.19, "learning_rate": 1.8748922160157108e-05, "loss": 2.3259, "step": 7551 }, { "epoch": 0.19, "learning_rate": 1.874853548640199e-05, "loss": 2.1801, "step": 7552 }, { "epoch": 0.19, "learning_rate": 1.8748148756889606e-05, "loss": 2.168, "step": 7553 }, { "epoch": 0.19, "learning_rate": 1.8747761971622417e-05, "loss": 2.4643, "step": 7554 }, { "epoch": 0.19, "learning_rate": 1.8747375130602893e-05, "loss": 2.2308, "step": 7555 }, { "epoch": 0.19, "learning_rate": 1.8746988233833493e-05, "loss": 2.0621, "step": 7556 }, { "epoch": 0.19, "learning_rate": 1.874660128131669e-05, "loss": 2.4386, "step": 7557 }, { "epoch": 0.19, "learning_rate": 1.874621427305494e-05, "loss": 2.1669, "step": 7558 }, { "epoch": 0.19, "learning_rate": 1.8745827209050723e-05, "loss": 2.1236, "step": 7559 }, { "epoch": 0.19, "learning_rate": 1.8745440089306497e-05, "loss": 2.2324, "step": 7560 }, { "epoch": 0.19, "learning_rate": 1.874505291382473e-05, "loss": 2.1469, "step": 7561 }, { "epoch": 0.19, "learning_rate": 1.8744665682607892e-05, "loss": 2.0832, "step": 7562 }, { "epoch": 0.19, "learning_rate": 1.8744278395658447e-05, "loss": 2.0736, "step": 7563 }, { "epoch": 0.19, "learning_rate": 1.8743891052978872e-05, "loss": 2.318, "step": 7564 }, { "epoch": 0.19, "learning_rate": 1.8743503654571628e-05, "loss": 2.1445, "step": 7565 }, { "epoch": 0.19, "learning_rate": 1.8743116200439184e-05, "loss": 2.231, "step": 7566 }, { "epoch": 0.19, "learning_rate": 1.874272869058401e-05, "loss": 2.3103, "step": 7567 }, { "epoch": 0.19, "learning_rate": 1.8742341125008575e-05, "loss": 2.2872, "step": 7568 }, { "epoch": 0.19, "learning_rate": 1.8741953503715358e-05, "loss": 2.1642, "step": 7569 }, { "epoch": 0.19, "learning_rate": 1.8741565826706817e-05, "loss": 2.2939, "step": 7570 }, { "epoch": 0.19, "learning_rate": 1.874117809398543e-05, "loss": 2.2994, "step": 7571 }, { "epoch": 0.19, "learning_rate": 1.8740790305553666e-05, "loss": 2.1834, "step": 7572 }, { "epoch": 0.19, "learning_rate": 1.8740402461413998e-05, "loss": 2.2627, "step": 7573 }, { "epoch": 0.19, "learning_rate": 1.8740014561568897e-05, "loss": 2.1487, "step": 7574 }, { "epoch": 0.19, "learning_rate": 1.8739626606020832e-05, "loss": 2.0656, "step": 7575 }, { "epoch": 0.19, "learning_rate": 1.873923859477228e-05, "loss": 2.2983, "step": 7576 }, { "epoch": 0.19, "learning_rate": 1.8738850527825712e-05, "loss": 2.0466, "step": 7577 }, { "epoch": 0.19, "learning_rate": 1.8738462405183604e-05, "loss": 2.0215, "step": 7578 }, { "epoch": 0.19, "learning_rate": 1.8738074226848425e-05, "loss": 2.2431, "step": 7579 }, { "epoch": 0.19, "learning_rate": 1.873768599282265e-05, "loss": 2.3554, "step": 7580 }, { "epoch": 0.19, "learning_rate": 1.8737297703108757e-05, "loss": 2.0945, "step": 7581 }, { "epoch": 0.19, "learning_rate": 1.8736909357709215e-05, "loss": 2.1652, "step": 7582 }, { "epoch": 0.19, "learning_rate": 1.8736520956626506e-05, "loss": 2.4496, "step": 7583 }, { "epoch": 0.19, "learning_rate": 1.8736132499863103e-05, "loss": 2.2359, "step": 7584 }, { "epoch": 0.19, "learning_rate": 1.8735743987421477e-05, "loss": 2.3703, "step": 7585 }, { "epoch": 0.19, "learning_rate": 1.8735355419304108e-05, "loss": 2.1145, "step": 7586 }, { "epoch": 0.19, "learning_rate": 1.8734966795513476e-05, "loss": 2.2412, "step": 7587 }, { "epoch": 0.19, "learning_rate": 1.873457811605205e-05, "loss": 2.2449, "step": 7588 }, { "epoch": 0.19, "learning_rate": 1.873418938092231e-05, "loss": 2.2161, "step": 7589 }, { "epoch": 0.19, "learning_rate": 1.873380059012674e-05, "loss": 2.2267, "step": 7590 }, { "epoch": 0.19, "learning_rate": 1.8733411743667812e-05, "loss": 2.0995, "step": 7591 }, { "epoch": 0.19, "learning_rate": 1.8733022841548e-05, "loss": 2.1697, "step": 7592 }, { "epoch": 0.19, "learning_rate": 1.8732633883769787e-05, "loss": 2.2157, "step": 7593 }, { "epoch": 0.19, "learning_rate": 1.8732244870335657e-05, "loss": 2.1007, "step": 7594 }, { "epoch": 0.19, "learning_rate": 1.8731855801248085e-05, "loss": 2.3084, "step": 7595 }, { "epoch": 0.19, "learning_rate": 1.873146667650955e-05, "loss": 2.2309, "step": 7596 }, { "epoch": 0.19, "learning_rate": 1.873107749612253e-05, "loss": 2.2121, "step": 7597 }, { "epoch": 0.19, "learning_rate": 1.8730688260089507e-05, "loss": 2.2079, "step": 7598 }, { "epoch": 0.19, "learning_rate": 1.8730298968412962e-05, "loss": 2.129, "step": 7599 }, { "epoch": 0.19, "learning_rate": 1.8729909621095382e-05, "loss": 2.1501, "step": 7600 }, { "epoch": 0.19, "learning_rate": 1.872952021813924e-05, "loss": 2.1783, "step": 7601 }, { "epoch": 0.19, "learning_rate": 1.872913075954702e-05, "loss": 2.1855, "step": 7602 }, { "epoch": 0.19, "learning_rate": 1.8728741245321204e-05, "loss": 2.1822, "step": 7603 }, { "epoch": 0.19, "learning_rate": 1.8728351675464273e-05, "loss": 2.3561, "step": 7604 }, { "epoch": 0.19, "learning_rate": 1.872796204997872e-05, "loss": 2.0929, "step": 7605 }, { "epoch": 0.19, "learning_rate": 1.8727572368867016e-05, "loss": 2.2052, "step": 7606 }, { "epoch": 0.19, "learning_rate": 1.8727182632131646e-05, "loss": 2.1264, "step": 7607 }, { "epoch": 0.19, "learning_rate": 1.8726792839775105e-05, "loss": 2.1792, "step": 7608 }, { "epoch": 0.19, "learning_rate": 1.8726402991799863e-05, "loss": 2.272, "step": 7609 }, { "epoch": 0.19, "learning_rate": 1.872601308820841e-05, "loss": 2.3137, "step": 7610 }, { "epoch": 0.19, "learning_rate": 1.8725623129003235e-05, "loss": 2.1427, "step": 7611 }, { "epoch": 0.19, "learning_rate": 1.872523311418682e-05, "loss": 2.3792, "step": 7612 }, { "epoch": 0.19, "learning_rate": 1.872484304376165e-05, "loss": 2.2263, "step": 7613 }, { "epoch": 0.19, "learning_rate": 1.8724452917730212e-05, "loss": 2.1307, "step": 7614 }, { "epoch": 0.19, "learning_rate": 1.8724062736094994e-05, "loss": 2.2881, "step": 7615 }, { "epoch": 0.19, "learning_rate": 1.8723672498858477e-05, "loss": 2.357, "step": 7616 }, { "epoch": 0.19, "learning_rate": 1.8723282206023158e-05, "loss": 2.0573, "step": 7617 }, { "epoch": 0.19, "learning_rate": 1.8722891857591516e-05, "loss": 2.3535, "step": 7618 }, { "epoch": 0.19, "learning_rate": 1.872250145356604e-05, "loss": 2.2138, "step": 7619 }, { "epoch": 0.19, "learning_rate": 1.8722110993949222e-05, "loss": 2.247, "step": 7620 }, { "epoch": 0.19, "learning_rate": 1.8721720478743546e-05, "loss": 2.1805, "step": 7621 }, { "epoch": 0.19, "learning_rate": 1.8721329907951505e-05, "loss": 2.1415, "step": 7622 }, { "epoch": 0.19, "learning_rate": 1.8720939281575584e-05, "loss": 2.2862, "step": 7623 }, { "epoch": 0.19, "learning_rate": 1.8720548599618275e-05, "loss": 2.239, "step": 7624 }, { "epoch": 0.19, "learning_rate": 1.872015786208207e-05, "loss": 2.2118, "step": 7625 }, { "epoch": 0.19, "learning_rate": 1.8719767068969457e-05, "loss": 2.1472, "step": 7626 }, { "epoch": 0.19, "learning_rate": 1.8719376220282926e-05, "loss": 2.1004, "step": 7627 }, { "epoch": 0.19, "learning_rate": 1.8718985316024968e-05, "loss": 2.0268, "step": 7628 }, { "epoch": 0.19, "learning_rate": 1.8718594356198077e-05, "loss": 2.2641, "step": 7629 }, { "epoch": 0.19, "learning_rate": 1.871820334080474e-05, "loss": 2.1178, "step": 7630 }, { "epoch": 0.19, "learning_rate": 1.8717812269847457e-05, "loss": 2.3311, "step": 7631 }, { "epoch": 0.19, "learning_rate": 1.871742114332871e-05, "loss": 2.2423, "step": 7632 }, { "epoch": 0.19, "learning_rate": 1.8717029961251e-05, "loss": 2.1249, "step": 7633 }, { "epoch": 0.19, "learning_rate": 1.8716638723616815e-05, "loss": 2.2718, "step": 7634 }, { "epoch": 0.19, "learning_rate": 1.8716247430428653e-05, "loss": 2.2067, "step": 7635 }, { "epoch": 0.19, "learning_rate": 1.8715856081689007e-05, "loss": 2.079, "step": 7636 }, { "epoch": 0.19, "learning_rate": 1.8715464677400365e-05, "loss": 2.0968, "step": 7637 }, { "epoch": 0.19, "learning_rate": 1.871507321756523e-05, "loss": 2.1706, "step": 7638 }, { "epoch": 0.19, "learning_rate": 1.8714681702186092e-05, "loss": 2.3535, "step": 7639 }, { "epoch": 0.19, "learning_rate": 1.871429013126545e-05, "loss": 2.1921, "step": 7640 }, { "epoch": 0.19, "learning_rate": 1.8713898504805794e-05, "loss": 2.3124, "step": 7641 }, { "epoch": 0.19, "learning_rate": 1.871350682280962e-05, "loss": 2.1488, "step": 7642 }, { "epoch": 0.19, "learning_rate": 1.8713115085279433e-05, "loss": 2.1383, "step": 7643 }, { "epoch": 0.19, "learning_rate": 1.871272329221772e-05, "loss": 2.0778, "step": 7644 }, { "epoch": 0.19, "learning_rate": 1.8712331443626986e-05, "loss": 2.1572, "step": 7645 }, { "epoch": 0.19, "learning_rate": 1.8711939539509724e-05, "loss": 2.1922, "step": 7646 }, { "epoch": 0.19, "learning_rate": 1.871154757986843e-05, "loss": 2.1934, "step": 7647 }, { "epoch": 0.19, "learning_rate": 1.8711155564705605e-05, "loss": 2.2278, "step": 7648 }, { "epoch": 0.19, "learning_rate": 1.8710763494023746e-05, "loss": 2.0674, "step": 7649 }, { "epoch": 0.19, "learning_rate": 1.8710371367825355e-05, "loss": 2.1812, "step": 7650 }, { "epoch": 0.19, "learning_rate": 1.8709979186112927e-05, "loss": 2.2594, "step": 7651 }, { "epoch": 0.19, "learning_rate": 1.8709586948888964e-05, "loss": 2.1226, "step": 7652 }, { "epoch": 0.19, "learning_rate": 1.8709194656155964e-05, "loss": 2.2627, "step": 7653 }, { "epoch": 0.19, "learning_rate": 1.8708802307916424e-05, "loss": 2.201, "step": 7654 }, { "epoch": 0.19, "learning_rate": 1.8708409904172855e-05, "loss": 2.323, "step": 7655 }, { "epoch": 0.19, "learning_rate": 1.870801744492775e-05, "loss": 2.2449, "step": 7656 }, { "epoch": 0.19, "learning_rate": 1.870762493018361e-05, "loss": 2.2774, "step": 7657 }, { "epoch": 0.19, "learning_rate": 1.870723235994294e-05, "loss": 2.2472, "step": 7658 }, { "epoch": 0.19, "learning_rate": 1.8706839734208242e-05, "loss": 2.1176, "step": 7659 }, { "epoch": 0.19, "learning_rate": 1.8706447052982015e-05, "loss": 2.205, "step": 7660 }, { "epoch": 0.19, "learning_rate": 1.8706054316266766e-05, "loss": 2.2925, "step": 7661 }, { "epoch": 0.19, "learning_rate": 1.8705661524064993e-05, "loss": 2.2391, "step": 7662 }, { "epoch": 0.19, "learning_rate": 1.87052686763792e-05, "loss": 2.1486, "step": 7663 }, { "epoch": 0.19, "learning_rate": 1.8704875773211898e-05, "loss": 2.38, "step": 7664 }, { "epoch": 0.19, "learning_rate": 1.8704482814565582e-05, "loss": 2.1633, "step": 7665 }, { "epoch": 0.19, "learning_rate": 1.8704089800442763e-05, "loss": 2.0104, "step": 7666 }, { "epoch": 0.19, "learning_rate": 1.8703696730845944e-05, "loss": 2.2356, "step": 7667 }, { "epoch": 0.19, "learning_rate": 1.8703303605777624e-05, "loss": 2.186, "step": 7668 }, { "epoch": 0.19, "learning_rate": 1.870291042524032e-05, "loss": 2.322, "step": 7669 }, { "epoch": 0.19, "learning_rate": 1.8702517189236528e-05, "loss": 2.1857, "step": 7670 }, { "epoch": 0.19, "learning_rate": 1.870212389776876e-05, "loss": 2.3435, "step": 7671 }, { "epoch": 0.19, "learning_rate": 1.8701730550839518e-05, "loss": 2.3327, "step": 7672 }, { "epoch": 0.19, "learning_rate": 1.8701337148451314e-05, "loss": 2.3159, "step": 7673 }, { "epoch": 0.19, "learning_rate": 1.870094369060665e-05, "loss": 2.3036, "step": 7674 }, { "epoch": 0.19, "learning_rate": 1.870055017730804e-05, "loss": 2.1049, "step": 7675 }, { "epoch": 0.19, "learning_rate": 1.870015660855798e-05, "loss": 2.078, "step": 7676 }, { "epoch": 0.19, "learning_rate": 1.8699762984358997e-05, "loss": 2.3491, "step": 7677 }, { "epoch": 0.19, "learning_rate": 1.8699369304713585e-05, "loss": 2.2722, "step": 7678 }, { "epoch": 0.19, "learning_rate": 1.8698975569624256e-05, "loss": 2.1114, "step": 7679 }, { "epoch": 0.19, "learning_rate": 1.869858177909352e-05, "loss": 2.4222, "step": 7680 }, { "epoch": 0.19, "learning_rate": 1.869818793312389e-05, "loss": 2.2589, "step": 7681 }, { "epoch": 0.19, "learning_rate": 1.8697794031717873e-05, "loss": 2.3337, "step": 7682 }, { "epoch": 0.19, "learning_rate": 1.8697400074877982e-05, "loss": 2.0034, "step": 7683 }, { "epoch": 0.19, "learning_rate": 1.8697006062606723e-05, "loss": 2.3034, "step": 7684 }, { "epoch": 0.19, "learning_rate": 1.8696611994906614e-05, "loss": 2.2992, "step": 7685 }, { "epoch": 0.19, "learning_rate": 1.869621787178016e-05, "loss": 2.1076, "step": 7686 }, { "epoch": 0.19, "learning_rate": 1.8695823693229875e-05, "loss": 2.0874, "step": 7687 }, { "epoch": 0.19, "learning_rate": 1.869542945925827e-05, "loss": 2.3913, "step": 7688 }, { "epoch": 0.19, "learning_rate": 1.869503516986786e-05, "loss": 2.2557, "step": 7689 }, { "epoch": 0.19, "learning_rate": 1.8694640825061158e-05, "loss": 2.2436, "step": 7690 }, { "epoch": 0.19, "learning_rate": 1.8694246424840678e-05, "loss": 2.1305, "step": 7691 }, { "epoch": 0.19, "learning_rate": 1.8693851969208932e-05, "loss": 2.1342, "step": 7692 }, { "epoch": 0.19, "learning_rate": 1.869345745816843e-05, "loss": 2.2422, "step": 7693 }, { "epoch": 0.19, "learning_rate": 1.8693062891721696e-05, "loss": 2.3878, "step": 7694 }, { "epoch": 0.19, "learning_rate": 1.8692668269871237e-05, "loss": 2.2823, "step": 7695 }, { "epoch": 0.19, "learning_rate": 1.869227359261957e-05, "loss": 2.3456, "step": 7696 }, { "epoch": 0.19, "learning_rate": 1.869187885996921e-05, "loss": 2.3155, "step": 7697 }, { "epoch": 0.19, "learning_rate": 1.8691484071922677e-05, "loss": 2.1171, "step": 7698 }, { "epoch": 0.19, "learning_rate": 1.869108922848248e-05, "loss": 2.2769, "step": 7699 }, { "epoch": 0.19, "learning_rate": 1.869069432965114e-05, "loss": 2.2474, "step": 7700 }, { "epoch": 0.19, "learning_rate": 1.8690299375431172e-05, "loss": 2.1231, "step": 7701 }, { "epoch": 0.19, "learning_rate": 1.8689904365825097e-05, "loss": 2.3499, "step": 7702 }, { "epoch": 0.19, "learning_rate": 1.8689509300835426e-05, "loss": 2.016, "step": 7703 }, { "epoch": 0.19, "learning_rate": 1.868911418046468e-05, "loss": 2.213, "step": 7704 }, { "epoch": 0.19, "learning_rate": 1.868871900471538e-05, "loss": 2.1579, "step": 7705 }, { "epoch": 0.19, "learning_rate": 1.8688323773590042e-05, "loss": 2.2444, "step": 7706 }, { "epoch": 0.19, "learning_rate": 1.8687928487091186e-05, "loss": 2.2408, "step": 7707 }, { "epoch": 0.19, "learning_rate": 1.868753314522133e-05, "loss": 2.2405, "step": 7708 }, { "epoch": 0.19, "learning_rate": 1.8687137747982992e-05, "loss": 2.1235, "step": 7709 }, { "epoch": 0.19, "learning_rate": 1.8686742295378695e-05, "loss": 2.3206, "step": 7710 }, { "epoch": 0.19, "learning_rate": 1.868634678741096e-05, "loss": 2.4825, "step": 7711 }, { "epoch": 0.19, "learning_rate": 1.8685951224082305e-05, "loss": 2.1543, "step": 7712 }, { "epoch": 0.19, "learning_rate": 1.868555560539525e-05, "loss": 2.3818, "step": 7713 }, { "epoch": 0.19, "learning_rate": 1.868515993135232e-05, "loss": 2.1912, "step": 7714 }, { "epoch": 0.19, "learning_rate": 1.8684764201956035e-05, "loss": 2.2312, "step": 7715 }, { "epoch": 0.19, "learning_rate": 1.8684368417208917e-05, "loss": 2.2985, "step": 7716 }, { "epoch": 0.19, "learning_rate": 1.868397257711349e-05, "loss": 2.1527, "step": 7717 }, { "epoch": 0.19, "learning_rate": 1.868357668167228e-05, "loss": 2.2754, "step": 7718 }, { "epoch": 0.19, "learning_rate": 1.86831807308878e-05, "loss": 2.2673, "step": 7719 }, { "epoch": 0.19, "learning_rate": 1.8682784724762582e-05, "loss": 2.3183, "step": 7720 }, { "epoch": 0.19, "learning_rate": 1.8682388663299148e-05, "loss": 2.2, "step": 7721 }, { "epoch": 0.19, "learning_rate": 1.868199254650002e-05, "loss": 2.1083, "step": 7722 }, { "epoch": 0.19, "learning_rate": 1.8681596374367722e-05, "loss": 2.2934, "step": 7723 }, { "epoch": 0.19, "learning_rate": 1.8681200146904786e-05, "loss": 2.1242, "step": 7724 }, { "epoch": 0.19, "learning_rate": 1.8680803864113726e-05, "loss": 2.3538, "step": 7725 }, { "epoch": 0.19, "learning_rate": 1.8680407525997083e-05, "loss": 2.1535, "step": 7726 }, { "epoch": 0.19, "learning_rate": 1.8680011132557364e-05, "loss": 2.2885, "step": 7727 }, { "epoch": 0.19, "learning_rate": 1.8679614683797113e-05, "loss": 2.2156, "step": 7728 }, { "epoch": 0.19, "learning_rate": 1.8679218179718844e-05, "loss": 2.2126, "step": 7729 }, { "epoch": 0.19, "learning_rate": 1.8678821620325093e-05, "loss": 2.2349, "step": 7730 }, { "epoch": 0.19, "learning_rate": 1.8678425005618382e-05, "loss": 2.2612, "step": 7731 }, { "epoch": 0.19, "learning_rate": 1.867802833560124e-05, "loss": 2.2273, "step": 7732 }, { "epoch": 0.19, "learning_rate": 1.8677631610276194e-05, "loss": 2.2292, "step": 7733 }, { "epoch": 0.19, "learning_rate": 1.8677234829645778e-05, "loss": 2.0787, "step": 7734 }, { "epoch": 0.19, "learning_rate": 1.867683799371251e-05, "loss": 2.111, "step": 7735 }, { "epoch": 0.19, "learning_rate": 1.867644110247893e-05, "loss": 2.228, "step": 7736 }, { "epoch": 0.19, "learning_rate": 1.867604415594756e-05, "loss": 2.6536, "step": 7737 }, { "epoch": 0.19, "learning_rate": 1.8675647154120938e-05, "loss": 2.2767, "step": 7738 }, { "epoch": 0.19, "learning_rate": 1.8675250097001587e-05, "loss": 2.1772, "step": 7739 }, { "epoch": 0.19, "learning_rate": 1.867485298459204e-05, "loss": 2.3035, "step": 7740 }, { "epoch": 0.19, "learning_rate": 1.8674455816894826e-05, "loss": 2.161, "step": 7741 }, { "epoch": 0.19, "learning_rate": 1.8674058593912482e-05, "loss": 2.1708, "step": 7742 }, { "epoch": 0.19, "learning_rate": 1.8673661315647533e-05, "loss": 2.3355, "step": 7743 }, { "epoch": 0.19, "learning_rate": 1.8673263982102513e-05, "loss": 2.2655, "step": 7744 }, { "epoch": 0.19, "learning_rate": 1.8672866593279958e-05, "loss": 2.3103, "step": 7745 }, { "epoch": 0.19, "learning_rate": 1.8672469149182397e-05, "loss": 2.2194, "step": 7746 }, { "epoch": 0.19, "learning_rate": 1.8672071649812365e-05, "loss": 2.2946, "step": 7747 }, { "epoch": 0.19, "learning_rate": 1.8671674095172393e-05, "loss": 2.2485, "step": 7748 }, { "epoch": 0.19, "learning_rate": 1.8671276485265016e-05, "loss": 2.2754, "step": 7749 }, { "epoch": 0.19, "learning_rate": 1.8670878820092767e-05, "loss": 2.1114, "step": 7750 }, { "epoch": 0.19, "learning_rate": 1.867048109965818e-05, "loss": 2.1525, "step": 7751 }, { "epoch": 0.19, "learning_rate": 1.8670083323963793e-05, "loss": 2.2989, "step": 7752 }, { "epoch": 0.19, "learning_rate": 1.8669685493012143e-05, "loss": 2.1539, "step": 7753 }, { "epoch": 0.19, "learning_rate": 1.8669287606805756e-05, "loss": 2.1972, "step": 7754 }, { "epoch": 0.19, "learning_rate": 1.866888966534718e-05, "loss": 2.1324, "step": 7755 }, { "epoch": 0.19, "learning_rate": 1.866849166863894e-05, "loss": 2.318, "step": 7756 }, { "epoch": 0.19, "learning_rate": 1.866809361668358e-05, "loss": 2.2878, "step": 7757 }, { "epoch": 0.19, "learning_rate": 1.8667695509483633e-05, "loss": 2.2175, "step": 7758 }, { "epoch": 0.19, "learning_rate": 1.8667297347041638e-05, "loss": 2.1194, "step": 7759 }, { "epoch": 0.19, "learning_rate": 1.8666899129360135e-05, "loss": 2.2934, "step": 7760 }, { "epoch": 0.19, "learning_rate": 1.8666500856441655e-05, "loss": 2.1268, "step": 7761 }, { "epoch": 0.19, "learning_rate": 1.8666102528288746e-05, "loss": 2.0168, "step": 7762 }, { "epoch": 0.19, "learning_rate": 1.8665704144903936e-05, "loss": 2.3187, "step": 7763 }, { "epoch": 0.19, "learning_rate": 1.8665305706289772e-05, "loss": 2.347, "step": 7764 }, { "epoch": 0.19, "learning_rate": 1.8664907212448793e-05, "loss": 2.1884, "step": 7765 }, { "epoch": 0.19, "learning_rate": 1.8664508663383535e-05, "loss": 2.3255, "step": 7766 }, { "epoch": 0.19, "learning_rate": 1.8664110059096537e-05, "loss": 2.103, "step": 7767 }, { "epoch": 0.19, "learning_rate": 1.8663711399590346e-05, "loss": 2.1498, "step": 7768 }, { "epoch": 0.19, "learning_rate": 1.8663312684867498e-05, "loss": 2.1388, "step": 7769 }, { "epoch": 0.19, "learning_rate": 1.8662913914930532e-05, "loss": 2.1193, "step": 7770 }, { "epoch": 0.19, "learning_rate": 1.8662515089781993e-05, "loss": 2.201, "step": 7771 }, { "epoch": 0.19, "learning_rate": 1.8662116209424426e-05, "loss": 2.1995, "step": 7772 }, { "epoch": 0.19, "learning_rate": 1.8661717273860366e-05, "loss": 2.2257, "step": 7773 }, { "epoch": 0.19, "learning_rate": 1.866131828309236e-05, "loss": 2.2548, "step": 7774 }, { "epoch": 0.19, "learning_rate": 1.866091923712295e-05, "loss": 2.2526, "step": 7775 }, { "epoch": 0.19, "learning_rate": 1.866052013595468e-05, "loss": 2.2053, "step": 7776 }, { "epoch": 0.19, "learning_rate": 1.866012097959009e-05, "loss": 2.2361, "step": 7777 }, { "epoch": 0.19, "learning_rate": 1.865972176803173e-05, "loss": 2.2793, "step": 7778 }, { "epoch": 0.19, "learning_rate": 1.865932250128214e-05, "loss": 2.1283, "step": 7779 }, { "epoch": 0.19, "learning_rate": 1.8658923179343866e-05, "loss": 2.2787, "step": 7780 }, { "epoch": 0.19, "learning_rate": 1.8658523802219453e-05, "loss": 2.1473, "step": 7781 }, { "epoch": 0.19, "learning_rate": 1.8658124369911444e-05, "loss": 2.3091, "step": 7782 }, { "epoch": 0.19, "learning_rate": 1.865772488242239e-05, "loss": 2.3732, "step": 7783 }, { "epoch": 0.19, "learning_rate": 1.865732533975483e-05, "loss": 2.3676, "step": 7784 }, { "epoch": 0.19, "learning_rate": 1.865692574191132e-05, "loss": 2.0304, "step": 7785 }, { "epoch": 0.19, "learning_rate": 1.8656526088894395e-05, "loss": 2.2011, "step": 7786 }, { "epoch": 0.19, "learning_rate": 1.865612638070661e-05, "loss": 2.2403, "step": 7787 }, { "epoch": 0.19, "learning_rate": 1.865572661735051e-05, "loss": 2.14, "step": 7788 }, { "epoch": 0.19, "learning_rate": 1.8655326798828646e-05, "loss": 2.0912, "step": 7789 }, { "epoch": 0.19, "learning_rate": 1.8654926925143562e-05, "loss": 2.0932, "step": 7790 }, { "epoch": 0.19, "learning_rate": 1.8654526996297808e-05, "loss": 2.3443, "step": 7791 }, { "epoch": 0.19, "learning_rate": 1.8654127012293933e-05, "loss": 2.1261, "step": 7792 }, { "epoch": 0.19, "learning_rate": 1.8653726973134488e-05, "loss": 2.0753, "step": 7793 }, { "epoch": 0.19, "learning_rate": 1.865332687882202e-05, "loss": 2.2665, "step": 7794 }, { "epoch": 0.19, "learning_rate": 1.8652926729359078e-05, "loss": 2.1079, "step": 7795 }, { "epoch": 0.19, "learning_rate": 1.8652526524748213e-05, "loss": 2.1156, "step": 7796 }, { "epoch": 0.19, "learning_rate": 1.8652126264991982e-05, "loss": 2.3433, "step": 7797 }, { "epoch": 0.19, "learning_rate": 1.8651725950092925e-05, "loss": 2.1349, "step": 7798 }, { "epoch": 0.19, "learning_rate": 1.8651325580053604e-05, "loss": 2.054, "step": 7799 }, { "epoch": 0.19, "learning_rate": 1.865092515487656e-05, "loss": 2.1882, "step": 7800 }, { "epoch": 0.19, "learning_rate": 1.8650524674564353e-05, "loss": 2.1769, "step": 7801 }, { "epoch": 0.19, "learning_rate": 1.865012413911954e-05, "loss": 2.0661, "step": 7802 }, { "epoch": 0.19, "learning_rate": 1.8649723548544657e-05, "loss": 2.2628, "step": 7803 }, { "epoch": 0.19, "learning_rate": 1.864932290284227e-05, "loss": 2.2553, "step": 7804 }, { "epoch": 0.19, "learning_rate": 1.8648922202014932e-05, "loss": 2.274, "step": 7805 }, { "epoch": 0.19, "learning_rate": 1.8648521446065192e-05, "loss": 2.2978, "step": 7806 }, { "epoch": 0.19, "learning_rate": 1.8648120634995606e-05, "loss": 2.0768, "step": 7807 }, { "epoch": 0.19, "learning_rate": 1.8647719768808733e-05, "loss": 2.2607, "step": 7808 }, { "epoch": 0.19, "learning_rate": 1.864731884750712e-05, "loss": 2.1596, "step": 7809 }, { "epoch": 0.19, "learning_rate": 1.8646917871093327e-05, "loss": 2.2058, "step": 7810 }, { "epoch": 0.19, "learning_rate": 1.8646516839569905e-05, "loss": 2.1534, "step": 7811 }, { "epoch": 0.19, "learning_rate": 1.8646115752939417e-05, "loss": 2.3646, "step": 7812 }, { "epoch": 0.19, "learning_rate": 1.8645714611204414e-05, "loss": 2.205, "step": 7813 }, { "epoch": 0.19, "learning_rate": 1.8645313414367455e-05, "loss": 2.2334, "step": 7814 }, { "epoch": 0.19, "learning_rate": 1.8644912162431098e-05, "loss": 2.1854, "step": 7815 }, { "epoch": 0.19, "learning_rate": 1.8644510855397894e-05, "loss": 2.2827, "step": 7816 }, { "epoch": 0.19, "learning_rate": 1.864410949327041e-05, "loss": 2.1845, "step": 7817 }, { "epoch": 0.19, "learning_rate": 1.8643708076051196e-05, "loss": 2.0887, "step": 7818 }, { "epoch": 0.19, "learning_rate": 1.8643306603742816e-05, "loss": 2.2055, "step": 7819 }, { "epoch": 0.19, "learning_rate": 1.8642905076347824e-05, "loss": 2.262, "step": 7820 }, { "epoch": 0.19, "learning_rate": 1.8642503493868785e-05, "loss": 2.4218, "step": 7821 }, { "epoch": 0.19, "learning_rate": 1.864210185630825e-05, "loss": 2.2079, "step": 7822 }, { "epoch": 0.19, "learning_rate": 1.8641700163668783e-05, "loss": 2.2804, "step": 7823 }, { "epoch": 0.19, "learning_rate": 1.864129841595295e-05, "loss": 2.2024, "step": 7824 }, { "epoch": 0.19, "learning_rate": 1.86408966131633e-05, "loss": 2.2853, "step": 7825 }, { "epoch": 0.19, "learning_rate": 1.8640494755302402e-05, "loss": 2.1939, "step": 7826 }, { "epoch": 0.19, "learning_rate": 1.8640092842372815e-05, "loss": 2.0963, "step": 7827 }, { "epoch": 0.19, "learning_rate": 1.8639690874377105e-05, "loss": 1.9987, "step": 7828 }, { "epoch": 0.19, "learning_rate": 1.8639288851317826e-05, "loss": 2.273, "step": 7829 }, { "epoch": 0.19, "learning_rate": 1.8638886773197543e-05, "loss": 2.3139, "step": 7830 }, { "epoch": 0.19, "learning_rate": 1.863848464001882e-05, "loss": 2.1043, "step": 7831 }, { "epoch": 0.19, "learning_rate": 1.8638082451784217e-05, "loss": 2.1262, "step": 7832 }, { "epoch": 0.19, "learning_rate": 1.8637680208496304e-05, "loss": 2.1545, "step": 7833 }, { "epoch": 0.19, "learning_rate": 1.863727791015764e-05, "loss": 2.1213, "step": 7834 }, { "epoch": 0.19, "learning_rate": 1.8636875556770785e-05, "loss": 2.3985, "step": 7835 }, { "epoch": 0.19, "learning_rate": 1.863647314833831e-05, "loss": 2.2623, "step": 7836 }, { "epoch": 0.19, "learning_rate": 1.8636070684862778e-05, "loss": 2.1979, "step": 7837 }, { "epoch": 0.19, "learning_rate": 1.8635668166346748e-05, "loss": 2.034, "step": 7838 }, { "epoch": 0.19, "learning_rate": 1.8635265592792795e-05, "loss": 1.9887, "step": 7839 }, { "epoch": 0.19, "learning_rate": 1.863486296420348e-05, "loss": 2.1414, "step": 7840 }, { "epoch": 0.19, "learning_rate": 1.8634460280581364e-05, "loss": 2.288, "step": 7841 }, { "epoch": 0.19, "learning_rate": 1.8634057541929027e-05, "loss": 2.1557, "step": 7842 }, { "epoch": 0.19, "learning_rate": 1.8633654748249022e-05, "loss": 2.2936, "step": 7843 }, { "epoch": 0.19, "learning_rate": 1.8633251899543924e-05, "loss": 2.1745, "step": 7844 }, { "epoch": 0.19, "learning_rate": 1.86328489958163e-05, "loss": 2.1509, "step": 7845 }, { "epoch": 0.19, "learning_rate": 1.8632446037068714e-05, "loss": 2.3661, "step": 7846 }, { "epoch": 0.19, "learning_rate": 1.8632043023303735e-05, "loss": 2.3308, "step": 7847 }, { "epoch": 0.19, "learning_rate": 1.8631639954523932e-05, "loss": 2.0624, "step": 7848 }, { "epoch": 0.19, "learning_rate": 1.8631236830731877e-05, "loss": 2.3721, "step": 7849 }, { "epoch": 0.19, "learning_rate": 1.8630833651930134e-05, "loss": 2.2631, "step": 7850 }, { "epoch": 0.19, "learning_rate": 1.863043041812128e-05, "loss": 2.2973, "step": 7851 }, { "epoch": 0.19, "learning_rate": 1.8630027129307874e-05, "loss": 2.3433, "step": 7852 }, { "epoch": 0.19, "learning_rate": 1.8629623785492495e-05, "loss": 2.4087, "step": 7853 }, { "epoch": 0.19, "learning_rate": 1.8629220386677716e-05, "loss": 2.2842, "step": 7854 }, { "epoch": 0.19, "learning_rate": 1.86288169328661e-05, "loss": 2.1168, "step": 7855 }, { "epoch": 0.19, "learning_rate": 1.862841342406022e-05, "loss": 2.4163, "step": 7856 }, { "epoch": 0.19, "learning_rate": 1.8628009860262653e-05, "loss": 2.1895, "step": 7857 }, { "epoch": 0.19, "learning_rate": 1.862760624147596e-05, "loss": 2.0453, "step": 7858 }, { "epoch": 0.19, "learning_rate": 1.862720256770273e-05, "loss": 2.1259, "step": 7859 }, { "epoch": 0.19, "learning_rate": 1.8626798838945525e-05, "loss": 2.4684, "step": 7860 }, { "epoch": 0.19, "learning_rate": 1.8626395055206915e-05, "loss": 2.2635, "step": 7861 }, { "epoch": 0.19, "learning_rate": 1.8625991216489478e-05, "loss": 2.4334, "step": 7862 }, { "epoch": 0.19, "learning_rate": 1.862558732279579e-05, "loss": 2.3559, "step": 7863 }, { "epoch": 0.19, "learning_rate": 1.8625183374128425e-05, "loss": 2.1177, "step": 7864 }, { "epoch": 0.19, "learning_rate": 1.8624779370489952e-05, "loss": 2.2217, "step": 7865 }, { "epoch": 0.19, "learning_rate": 1.862437531188295e-05, "loss": 2.2391, "step": 7866 }, { "epoch": 0.19, "learning_rate": 1.8623971198309997e-05, "loss": 2.252, "step": 7867 }, { "epoch": 0.19, "learning_rate": 1.862356702977366e-05, "loss": 2.079, "step": 7868 }, { "epoch": 0.19, "learning_rate": 1.862316280627652e-05, "loss": 2.1614, "step": 7869 }, { "epoch": 0.19, "learning_rate": 1.862275852782116e-05, "loss": 2.1969, "step": 7870 }, { "epoch": 0.19, "learning_rate": 1.8622354194410145e-05, "loss": 2.3046, "step": 7871 }, { "epoch": 0.19, "learning_rate": 1.862194980604606e-05, "loss": 2.3339, "step": 7872 }, { "epoch": 0.19, "learning_rate": 1.8621545362731473e-05, "loss": 2.2365, "step": 7873 }, { "epoch": 0.19, "learning_rate": 1.862114086446897e-05, "loss": 2.347, "step": 7874 }, { "epoch": 0.19, "learning_rate": 1.862073631126113e-05, "loss": 2.3149, "step": 7875 }, { "epoch": 0.19, "learning_rate": 1.8620331703110523e-05, "loss": 2.2012, "step": 7876 }, { "epoch": 0.19, "learning_rate": 1.8619927040019737e-05, "loss": 2.2513, "step": 7877 }, { "epoch": 0.19, "learning_rate": 1.8619522321991346e-05, "loss": 2.1003, "step": 7878 }, { "epoch": 0.19, "learning_rate": 1.861911754902793e-05, "loss": 2.1102, "step": 7879 }, { "epoch": 0.19, "learning_rate": 1.8618712721132065e-05, "loss": 2.1881, "step": 7880 }, { "epoch": 0.19, "learning_rate": 1.8618307838306343e-05, "loss": 2.1534, "step": 7881 }, { "epoch": 0.19, "learning_rate": 1.861790290055333e-05, "loss": 2.1756, "step": 7882 }, { "epoch": 0.19, "learning_rate": 1.8617497907875616e-05, "loss": 2.1429, "step": 7883 }, { "epoch": 0.19, "learning_rate": 1.8617092860275782e-05, "loss": 2.1788, "step": 7884 }, { "epoch": 0.19, "learning_rate": 1.86166877577564e-05, "loss": 2.2202, "step": 7885 }, { "epoch": 0.19, "learning_rate": 1.8616282600320064e-05, "loss": 2.1202, "step": 7886 }, { "epoch": 0.19, "learning_rate": 1.861587738796935e-05, "loss": 2.1361, "step": 7887 }, { "epoch": 0.19, "learning_rate": 1.8615472120706842e-05, "loss": 2.2383, "step": 7888 }, { "epoch": 0.19, "learning_rate": 1.861506679853512e-05, "loss": 2.1319, "step": 7889 }, { "epoch": 0.19, "learning_rate": 1.8614661421456774e-05, "loss": 2.224, "step": 7890 }, { "epoch": 0.19, "learning_rate": 1.8614255989474378e-05, "loss": 2.0943, "step": 7891 }, { "epoch": 0.19, "learning_rate": 1.8613850502590524e-05, "loss": 2.2709, "step": 7892 }, { "epoch": 0.19, "learning_rate": 1.8613444960807796e-05, "loss": 2.2258, "step": 7893 }, { "epoch": 0.19, "learning_rate": 1.8613039364128774e-05, "loss": 2.1817, "step": 7894 }, { "epoch": 0.19, "learning_rate": 1.8612633712556044e-05, "loss": 2.2693, "step": 7895 }, { "epoch": 0.19, "learning_rate": 1.861222800609219e-05, "loss": 2.048, "step": 7896 }, { "epoch": 0.19, "learning_rate": 1.8611822244739806e-05, "loss": 2.2316, "step": 7897 }, { "epoch": 0.19, "learning_rate": 1.8611416428501472e-05, "loss": 2.4349, "step": 7898 }, { "epoch": 0.19, "learning_rate": 1.861101055737977e-05, "loss": 2.1626, "step": 7899 }, { "epoch": 0.19, "learning_rate": 1.8610604631377297e-05, "loss": 2.2809, "step": 7900 }, { "epoch": 0.19, "learning_rate": 1.8610198650496632e-05, "loss": 2.1227, "step": 7901 }, { "epoch": 0.19, "learning_rate": 1.8609792614740366e-05, "loss": 2.0777, "step": 7902 }, { "epoch": 0.19, "learning_rate": 1.8609386524111082e-05, "loss": 2.2941, "step": 7903 }, { "epoch": 0.19, "learning_rate": 1.8608980378611377e-05, "loss": 2.0794, "step": 7904 }, { "epoch": 0.19, "learning_rate": 1.8608574178243836e-05, "loss": 2.2118, "step": 7905 }, { "epoch": 0.19, "learning_rate": 1.860816792301104e-05, "loss": 2.2671, "step": 7906 }, { "epoch": 0.19, "learning_rate": 1.8607761612915587e-05, "loss": 2.2803, "step": 7907 }, { "epoch": 0.19, "learning_rate": 1.8607355247960063e-05, "loss": 2.1431, "step": 7908 }, { "epoch": 0.19, "learning_rate": 1.860694882814706e-05, "loss": 2.4373, "step": 7909 }, { "epoch": 0.19, "learning_rate": 1.860654235347917e-05, "loss": 2.2379, "step": 7910 }, { "epoch": 0.19, "learning_rate": 1.8606135823958977e-05, "loss": 2.3254, "step": 7911 }, { "epoch": 0.2, "learning_rate": 1.8605729239589076e-05, "loss": 2.1345, "step": 7912 }, { "epoch": 0.2, "learning_rate": 1.860532260037206e-05, "loss": 2.1824, "step": 7913 }, { "epoch": 0.2, "learning_rate": 1.8604915906310518e-05, "loss": 2.3972, "step": 7914 }, { "epoch": 0.2, "learning_rate": 1.860450915740704e-05, "loss": 2.2533, "step": 7915 }, { "epoch": 0.2, "learning_rate": 1.8604102353664224e-05, "loss": 2.1562, "step": 7916 }, { "epoch": 0.2, "learning_rate": 1.8603695495084658e-05, "loss": 2.2199, "step": 7917 }, { "epoch": 0.2, "learning_rate": 1.860328858167094e-05, "loss": 2.1822, "step": 7918 }, { "epoch": 0.2, "learning_rate": 1.8602881613425657e-05, "loss": 2.1495, "step": 7919 }, { "epoch": 0.2, "learning_rate": 1.860247459035141e-05, "loss": 2.3171, "step": 7920 }, { "epoch": 0.2, "learning_rate": 1.860206751245078e-05, "loss": 2.2276, "step": 7921 }, { "epoch": 0.2, "learning_rate": 1.860166037972638e-05, "loss": 2.2061, "step": 7922 }, { "epoch": 0.2, "learning_rate": 1.860125319218079e-05, "loss": 2.3063, "step": 7923 }, { "epoch": 0.2, "learning_rate": 1.8600845949816614e-05, "loss": 2.291, "step": 7924 }, { "epoch": 0.2, "learning_rate": 1.8600438652636443e-05, "loss": 2.2777, "step": 7925 }, { "epoch": 0.2, "learning_rate": 1.860003130064287e-05, "loss": 2.1615, "step": 7926 }, { "epoch": 0.2, "learning_rate": 1.8599623893838497e-05, "loss": 2.3598, "step": 7927 }, { "epoch": 0.2, "learning_rate": 1.8599216432225917e-05, "loss": 2.1459, "step": 7928 }, { "epoch": 0.2, "learning_rate": 1.859880891580773e-05, "loss": 2.1115, "step": 7929 }, { "epoch": 0.2, "learning_rate": 1.8598401344586533e-05, "loss": 2.1683, "step": 7930 }, { "epoch": 0.2, "learning_rate": 1.859799371856492e-05, "loss": 2.0917, "step": 7931 }, { "epoch": 0.2, "learning_rate": 1.859758603774549e-05, "loss": 2.2552, "step": 7932 }, { "epoch": 0.2, "learning_rate": 1.8597178302130844e-05, "loss": 2.2698, "step": 7933 }, { "epoch": 0.2, "learning_rate": 1.859677051172358e-05, "loss": 2.4059, "step": 7934 }, { "epoch": 0.2, "learning_rate": 1.8596362666526292e-05, "loss": 2.2003, "step": 7935 }, { "epoch": 0.2, "learning_rate": 1.8595954766541585e-05, "loss": 1.9616, "step": 7936 }, { "epoch": 0.2, "learning_rate": 1.859554681177206e-05, "loss": 2.0638, "step": 7937 }, { "epoch": 0.2, "learning_rate": 1.859513880222031e-05, "loss": 2.2449, "step": 7938 }, { "epoch": 0.2, "learning_rate": 1.8594730737888942e-05, "loss": 2.3046, "step": 7939 }, { "epoch": 0.2, "learning_rate": 1.8594322618780552e-05, "loss": 2.1512, "step": 7940 }, { "epoch": 0.2, "learning_rate": 1.8593914444897743e-05, "loss": 2.4276, "step": 7941 }, { "epoch": 0.2, "learning_rate": 1.8593506216243118e-05, "loss": 2.2907, "step": 7942 }, { "epoch": 0.2, "learning_rate": 1.859309793281928e-05, "loss": 2.0795, "step": 7943 }, { "epoch": 0.2, "learning_rate": 1.8592689594628824e-05, "loss": 2.2684, "step": 7944 }, { "epoch": 0.2, "learning_rate": 1.8592281201674357e-05, "loss": 2.2245, "step": 7945 }, { "epoch": 0.2, "learning_rate": 1.8591872753958484e-05, "loss": 2.3382, "step": 7946 }, { "epoch": 0.2, "learning_rate": 1.8591464251483806e-05, "loss": 2.1354, "step": 7947 }, { "epoch": 0.2, "learning_rate": 1.8591055694252923e-05, "loss": 2.3555, "step": 7948 }, { "epoch": 0.2, "learning_rate": 1.8590647082268447e-05, "loss": 2.1547, "step": 7949 }, { "epoch": 0.2, "learning_rate": 1.8590238415532976e-05, "loss": 2.1712, "step": 7950 }, { "epoch": 0.2, "learning_rate": 1.8589829694049116e-05, "loss": 2.3701, "step": 7951 }, { "epoch": 0.2, "learning_rate": 1.858942091781947e-05, "loss": 2.1857, "step": 7952 }, { "epoch": 0.2, "learning_rate": 1.8589012086846648e-05, "loss": 2.0883, "step": 7953 }, { "epoch": 0.2, "learning_rate": 1.8588603201133254e-05, "loss": 2.3029, "step": 7954 }, { "epoch": 0.2, "learning_rate": 1.858819426068189e-05, "loss": 2.3643, "step": 7955 }, { "epoch": 0.2, "learning_rate": 1.8587785265495163e-05, "loss": 2.1736, "step": 7956 }, { "epoch": 0.2, "learning_rate": 1.8587376215575684e-05, "loss": 2.3285, "step": 7957 }, { "epoch": 0.2, "learning_rate": 1.858696711092606e-05, "loss": 2.2592, "step": 7958 }, { "epoch": 0.2, "learning_rate": 1.8586557951548893e-05, "loss": 2.0834, "step": 7959 }, { "epoch": 0.2, "learning_rate": 1.8586148737446795e-05, "loss": 2.2563, "step": 7960 }, { "epoch": 0.2, "learning_rate": 1.8585739468622375e-05, "loss": 2.2988, "step": 7961 }, { "epoch": 0.2, "learning_rate": 1.858533014507824e-05, "loss": 2.1873, "step": 7962 }, { "epoch": 0.2, "learning_rate": 1.8584920766816993e-05, "loss": 2.195, "step": 7963 }, { "epoch": 0.2, "learning_rate": 1.8584511333841253e-05, "loss": 2.0171, "step": 7964 }, { "epoch": 0.2, "learning_rate": 1.8584101846153625e-05, "loss": 2.1592, "step": 7965 }, { "epoch": 0.2, "learning_rate": 1.8583692303756712e-05, "loss": 2.2581, "step": 7966 }, { "epoch": 0.2, "learning_rate": 1.8583282706653136e-05, "loss": 2.265, "step": 7967 }, { "epoch": 0.2, "learning_rate": 1.85828730548455e-05, "loss": 2.1109, "step": 7968 }, { "epoch": 0.2, "learning_rate": 1.858246334833642e-05, "loss": 2.3034, "step": 7969 }, { "epoch": 0.2, "learning_rate": 1.85820535871285e-05, "loss": 2.1729, "step": 7970 }, { "epoch": 0.2, "learning_rate": 1.8581643771224362e-05, "loss": 2.2246, "step": 7971 }, { "epoch": 0.2, "learning_rate": 1.8581233900626607e-05, "loss": 2.1618, "step": 7972 }, { "epoch": 0.2, "learning_rate": 1.858082397533785e-05, "loss": 2.242, "step": 7973 }, { "epoch": 0.2, "learning_rate": 1.8580413995360707e-05, "loss": 2.2715, "step": 7974 }, { "epoch": 0.2, "learning_rate": 1.858000396069779e-05, "loss": 2.3689, "step": 7975 }, { "epoch": 0.2, "learning_rate": 1.8579593871351712e-05, "loss": 2.3345, "step": 7976 }, { "epoch": 0.2, "learning_rate": 1.8579183727325084e-05, "loss": 2.1542, "step": 7977 }, { "epoch": 0.2, "learning_rate": 1.8578773528620525e-05, "loss": 2.2839, "step": 7978 }, { "epoch": 0.2, "learning_rate": 1.8578363275240646e-05, "loss": 2.2544, "step": 7979 }, { "epoch": 0.2, "learning_rate": 1.857795296718806e-05, "loss": 2.1709, "step": 7980 }, { "epoch": 0.2, "learning_rate": 1.8577542604465386e-05, "loss": 2.3836, "step": 7981 }, { "epoch": 0.2, "learning_rate": 1.857713218707524e-05, "loss": 2.1907, "step": 7982 }, { "epoch": 0.2, "learning_rate": 1.8576721715020232e-05, "loss": 2.1997, "step": 7983 }, { "epoch": 0.2, "learning_rate": 1.8576311188302984e-05, "loss": 2.1713, "step": 7984 }, { "epoch": 0.2, "learning_rate": 1.8575900606926106e-05, "loss": 2.1663, "step": 7985 }, { "epoch": 0.2, "learning_rate": 1.857548997089222e-05, "loss": 2.08, "step": 7986 }, { "epoch": 0.2, "learning_rate": 1.8575079280203944e-05, "loss": 2.2827, "step": 7987 }, { "epoch": 0.2, "learning_rate": 1.857466853486389e-05, "loss": 2.1167, "step": 7988 }, { "epoch": 0.2, "learning_rate": 1.857425773487468e-05, "loss": 2.156, "step": 7989 }, { "epoch": 0.2, "learning_rate": 1.8573846880238933e-05, "loss": 2.1409, "step": 7990 }, { "epoch": 0.2, "learning_rate": 1.8573435970959265e-05, "loss": 2.0934, "step": 7991 }, { "epoch": 0.2, "learning_rate": 1.8573025007038297e-05, "loss": 2.1148, "step": 7992 }, { "epoch": 0.2, "learning_rate": 1.8572613988478645e-05, "loss": 2.2824, "step": 7993 }, { "epoch": 0.2, "learning_rate": 1.8572202915282932e-05, "loss": 2.2665, "step": 7994 }, { "epoch": 0.2, "learning_rate": 1.8571791787453774e-05, "loss": 2.0676, "step": 7995 }, { "epoch": 0.2, "learning_rate": 1.8571380604993793e-05, "loss": 2.1834, "step": 7996 }, { "epoch": 0.2, "learning_rate": 1.857096936790561e-05, "loss": 2.2714, "step": 7997 }, { "epoch": 0.2, "learning_rate": 1.8570558076191847e-05, "loss": 2.2286, "step": 7998 }, { "epoch": 0.2, "learning_rate": 1.8570146729855124e-05, "loss": 2.2768, "step": 7999 }, { "epoch": 0.2, "learning_rate": 1.8569735328898064e-05, "loss": 2.2478, "step": 8000 }, { "epoch": 0.2, "learning_rate": 1.856932387332329e-05, "loss": 2.1557, "step": 8001 }, { "epoch": 0.2, "learning_rate": 1.8568912363133416e-05, "loss": 2.2978, "step": 8002 }, { "epoch": 0.2, "learning_rate": 1.8568500798331074e-05, "loss": 2.2258, "step": 8003 }, { "epoch": 0.2, "learning_rate": 1.8568089178918887e-05, "loss": 2.293, "step": 8004 }, { "epoch": 0.2, "learning_rate": 1.8567677504899473e-05, "loss": 2.2751, "step": 8005 }, { "epoch": 0.2, "learning_rate": 1.856726577627546e-05, "loss": 2.216, "step": 8006 }, { "epoch": 0.2, "learning_rate": 1.856685399304947e-05, "loss": 2.2095, "step": 8007 }, { "epoch": 0.2, "learning_rate": 1.856644215522412e-05, "loss": 2.0472, "step": 8008 }, { "epoch": 0.2, "learning_rate": 1.8566030262802052e-05, "loss": 2.0966, "step": 8009 }, { "epoch": 0.2, "learning_rate": 1.8565618315785877e-05, "loss": 2.1857, "step": 8010 }, { "epoch": 0.2, "learning_rate": 1.8565206314178225e-05, "loss": 2.2113, "step": 8011 }, { "epoch": 0.2, "learning_rate": 1.8564794257981725e-05, "loss": 2.0128, "step": 8012 }, { "epoch": 0.2, "learning_rate": 1.8564382147198997e-05, "loss": 2.1777, "step": 8013 }, { "epoch": 0.2, "learning_rate": 1.8563969981832673e-05, "loss": 2.2165, "step": 8014 }, { "epoch": 0.2, "learning_rate": 1.8563557761885372e-05, "loss": 2.2352, "step": 8015 }, { "epoch": 0.2, "learning_rate": 1.8563145487359734e-05, "loss": 2.1691, "step": 8016 }, { "epoch": 0.2, "learning_rate": 1.8562733158258373e-05, "loss": 2.3055, "step": 8017 }, { "epoch": 0.2, "learning_rate": 1.8562320774583927e-05, "loss": 2.2951, "step": 8018 }, { "epoch": 0.2, "learning_rate": 1.8561908336339017e-05, "loss": 2.2828, "step": 8019 }, { "epoch": 0.2, "learning_rate": 1.856149584352628e-05, "loss": 2.033, "step": 8020 }, { "epoch": 0.2, "learning_rate": 1.8561083296148337e-05, "loss": 2.0065, "step": 8021 }, { "epoch": 0.2, "learning_rate": 1.8560670694207822e-05, "loss": 2.0092, "step": 8022 }, { "epoch": 0.2, "learning_rate": 1.8560258037707356e-05, "loss": 2.3367, "step": 8023 }, { "epoch": 0.2, "learning_rate": 1.8559845326649584e-05, "loss": 2.3192, "step": 8024 }, { "epoch": 0.2, "learning_rate": 1.8559432561037122e-05, "loss": 2.2387, "step": 8025 }, { "epoch": 0.2, "learning_rate": 1.855901974087261e-05, "loss": 2.2582, "step": 8026 }, { "epoch": 0.2, "learning_rate": 1.855860686615868e-05, "loss": 2.226, "step": 8027 }, { "epoch": 0.2, "learning_rate": 1.855819393689795e-05, "loss": 2.1557, "step": 8028 }, { "epoch": 0.2, "learning_rate": 1.855778095309307e-05, "loss": 2.2577, "step": 8029 }, { "epoch": 0.2, "learning_rate": 1.8557367914746654e-05, "loss": 2.2688, "step": 8030 }, { "epoch": 0.2, "learning_rate": 1.855695482186135e-05, "loss": 2.2797, "step": 8031 }, { "epoch": 0.2, "learning_rate": 1.8556541674439782e-05, "loss": 2.4119, "step": 8032 }, { "epoch": 0.2, "learning_rate": 1.8556128472484587e-05, "loss": 2.2871, "step": 8033 }, { "epoch": 0.2, "learning_rate": 1.8555715215998396e-05, "loss": 2.3739, "step": 8034 }, { "epoch": 0.2, "learning_rate": 1.8555301904983842e-05, "loss": 2.2889, "step": 8035 }, { "epoch": 0.2, "learning_rate": 1.8554888539443566e-05, "loss": 2.003, "step": 8036 }, { "epoch": 0.2, "learning_rate": 1.855447511938019e-05, "loss": 2.1136, "step": 8037 }, { "epoch": 0.2, "learning_rate": 1.8554061644796363e-05, "loss": 2.2122, "step": 8038 }, { "epoch": 0.2, "learning_rate": 1.8553648115694712e-05, "loss": 2.1888, "step": 8039 }, { "epoch": 0.2, "learning_rate": 1.855323453207787e-05, "loss": 2.2897, "step": 8040 }, { "epoch": 0.2, "learning_rate": 1.855282089394848e-05, "loss": 2.2666, "step": 8041 }, { "epoch": 0.2, "learning_rate": 1.8552407201309175e-05, "loss": 2.0717, "step": 8042 }, { "epoch": 0.2, "learning_rate": 1.8551993454162592e-05, "loss": 2.4431, "step": 8043 }, { "epoch": 0.2, "learning_rate": 1.8551579652511366e-05, "loss": 2.1861, "step": 8044 }, { "epoch": 0.2, "learning_rate": 1.8551165796358138e-05, "loss": 2.1578, "step": 8045 }, { "epoch": 0.2, "learning_rate": 1.8550751885705542e-05, "loss": 2.1828, "step": 8046 }, { "epoch": 0.2, "learning_rate": 1.855033792055622e-05, "loss": 2.1888, "step": 8047 }, { "epoch": 0.2, "learning_rate": 1.8549923900912804e-05, "loss": 2.271, "step": 8048 }, { "epoch": 0.2, "learning_rate": 1.854950982677794e-05, "loss": 2.1489, "step": 8049 }, { "epoch": 0.2, "learning_rate": 1.8549095698154264e-05, "loss": 2.1631, "step": 8050 }, { "epoch": 0.2, "learning_rate": 1.8548681515044413e-05, "loss": 2.1719, "step": 8051 }, { "epoch": 0.2, "learning_rate": 1.854826727745103e-05, "loss": 2.1727, "step": 8052 }, { "epoch": 0.2, "learning_rate": 1.8547852985376753e-05, "loss": 2.2808, "step": 8053 }, { "epoch": 0.2, "learning_rate": 1.854743863882422e-05, "loss": 2.3919, "step": 8054 }, { "epoch": 0.2, "learning_rate": 1.854702423779608e-05, "loss": 2.1837, "step": 8055 }, { "epoch": 0.2, "learning_rate": 1.854660978229497e-05, "loss": 2.2518, "step": 8056 }, { "epoch": 0.2, "learning_rate": 1.8546195272323527e-05, "loss": 2.0163, "step": 8057 }, { "epoch": 0.2, "learning_rate": 1.8545780707884397e-05, "loss": 2.2962, "step": 8058 }, { "epoch": 0.2, "learning_rate": 1.8545366088980222e-05, "loss": 2.2014, "step": 8059 }, { "epoch": 0.2, "learning_rate": 1.8544951415613645e-05, "loss": 2.1232, "step": 8060 }, { "epoch": 0.2, "learning_rate": 1.8544536687787308e-05, "loss": 2.1918, "step": 8061 }, { "epoch": 0.2, "learning_rate": 1.8544121905503854e-05, "loss": 2.3534, "step": 8062 }, { "epoch": 0.2, "learning_rate": 1.8543707068765925e-05, "loss": 2.3188, "step": 8063 }, { "epoch": 0.2, "learning_rate": 1.854329217757617e-05, "loss": 2.1877, "step": 8064 }, { "epoch": 0.2, "learning_rate": 1.8542877231937227e-05, "loss": 1.9904, "step": 8065 }, { "epoch": 0.2, "learning_rate": 1.8542462231851742e-05, "loss": 2.1147, "step": 8066 }, { "epoch": 0.2, "learning_rate": 1.854204717732236e-05, "loss": 2.2998, "step": 8067 }, { "epoch": 0.2, "learning_rate": 1.8541632068351733e-05, "loss": 2.3887, "step": 8068 }, { "epoch": 0.2, "learning_rate": 1.8541216904942498e-05, "loss": 2.3762, "step": 8069 }, { "epoch": 0.2, "learning_rate": 1.8540801687097304e-05, "loss": 2.0755, "step": 8070 }, { "epoch": 0.2, "learning_rate": 1.85403864148188e-05, "loss": 2.2117, "step": 8071 }, { "epoch": 0.2, "learning_rate": 1.8539971088109627e-05, "loss": 2.174, "step": 8072 }, { "epoch": 0.2, "learning_rate": 1.853955570697244e-05, "loss": 2.0895, "step": 8073 }, { "epoch": 0.2, "learning_rate": 1.8539140271409878e-05, "loss": 2.2275, "step": 8074 }, { "epoch": 0.2, "learning_rate": 1.8538724781424592e-05, "loss": 2.0258, "step": 8075 }, { "epoch": 0.2, "learning_rate": 1.853830923701923e-05, "loss": 2.4846, "step": 8076 }, { "epoch": 0.2, "learning_rate": 1.853789363819644e-05, "loss": 2.2029, "step": 8077 }, { "epoch": 0.2, "learning_rate": 1.8537477984958873e-05, "loss": 2.1806, "step": 8078 }, { "epoch": 0.2, "learning_rate": 1.8537062277309177e-05, "loss": 2.296, "step": 8079 }, { "epoch": 0.2, "learning_rate": 1.853664651525e-05, "loss": 2.328, "step": 8080 }, { "epoch": 0.2, "learning_rate": 1.8536230698783993e-05, "loss": 2.1478, "step": 8081 }, { "epoch": 0.2, "learning_rate": 1.8535814827913806e-05, "loss": 2.077, "step": 8082 }, { "epoch": 0.2, "learning_rate": 1.8535398902642087e-05, "loss": 2.2289, "step": 8083 }, { "epoch": 0.2, "learning_rate": 1.8534982922971493e-05, "loss": 2.2941, "step": 8084 }, { "epoch": 0.2, "learning_rate": 1.853456688890467e-05, "loss": 2.0234, "step": 8085 }, { "epoch": 0.2, "learning_rate": 1.853415080044427e-05, "loss": 2.1975, "step": 8086 }, { "epoch": 0.2, "learning_rate": 1.853373465759295e-05, "loss": 2.2327, "step": 8087 }, { "epoch": 0.2, "learning_rate": 1.853331846035335e-05, "loss": 2.2042, "step": 8088 }, { "epoch": 0.2, "learning_rate": 1.8532902208728138e-05, "loss": 2.2003, "step": 8089 }, { "epoch": 0.2, "learning_rate": 1.8532485902719958e-05, "loss": 2.1444, "step": 8090 }, { "epoch": 0.2, "learning_rate": 1.8532069542331462e-05, "loss": 2.1325, "step": 8091 }, { "epoch": 0.2, "learning_rate": 1.853165312756531e-05, "loss": 2.1628, "step": 8092 }, { "epoch": 0.2, "learning_rate": 1.8531236658424146e-05, "loss": 2.1107, "step": 8093 }, { "epoch": 0.2, "learning_rate": 1.853082013491063e-05, "loss": 2.3863, "step": 8094 }, { "epoch": 0.2, "learning_rate": 1.8530403557027426e-05, "loss": 2.12, "step": 8095 }, { "epoch": 0.2, "learning_rate": 1.8529986924777174e-05, "loss": 2.2488, "step": 8096 }, { "epoch": 0.2, "learning_rate": 1.8529570238162538e-05, "loss": 2.1707, "step": 8097 }, { "epoch": 0.2, "learning_rate": 1.8529153497186168e-05, "loss": 2.0, "step": 8098 }, { "epoch": 0.2, "learning_rate": 1.8528736701850727e-05, "loss": 2.1739, "step": 8099 }, { "epoch": 0.2, "learning_rate": 1.8528319852158864e-05, "loss": 2.2772, "step": 8100 }, { "epoch": 0.2, "learning_rate": 1.8527902948113243e-05, "loss": 2.1519, "step": 8101 }, { "epoch": 0.2, "learning_rate": 1.8527485989716515e-05, "loss": 2.1287, "step": 8102 }, { "epoch": 0.2, "learning_rate": 1.8527068976971342e-05, "loss": 2.1737, "step": 8103 }, { "epoch": 0.2, "learning_rate": 1.8526651909880375e-05, "loss": 2.2303, "step": 8104 }, { "epoch": 0.2, "learning_rate": 1.852623478844628e-05, "loss": 2.3526, "step": 8105 }, { "epoch": 0.2, "learning_rate": 1.8525817612671712e-05, "loss": 2.1729, "step": 8106 }, { "epoch": 0.2, "learning_rate": 1.8525400382559328e-05, "loss": 2.2759, "step": 8107 }, { "epoch": 0.2, "learning_rate": 1.8524983098111792e-05, "loss": 2.1689, "step": 8108 }, { "epoch": 0.2, "learning_rate": 1.852456575933176e-05, "loss": 2.3393, "step": 8109 }, { "epoch": 0.2, "learning_rate": 1.852414836622189e-05, "loss": 2.2487, "step": 8110 }, { "epoch": 0.2, "learning_rate": 1.8523730918784846e-05, "loss": 2.2608, "step": 8111 }, { "epoch": 0.2, "learning_rate": 1.8523313417023287e-05, "loss": 2.2542, "step": 8112 }, { "epoch": 0.2, "learning_rate": 1.852289586093987e-05, "loss": 2.0475, "step": 8113 }, { "epoch": 0.2, "learning_rate": 1.8522478250537266e-05, "loss": 2.2485, "step": 8114 }, { "epoch": 0.2, "eval_loss": 1.8662521839141846, "eval_runtime": 95.095, "eval_samples_per_second": 618.823, "eval_steps_per_second": 4.837, "step": 8114 }, { "epoch": 0.2, "learning_rate": 1.8522060585818128e-05, "loss": 2.1362, "step": 8115 }, { "epoch": 0.2, "learning_rate": 1.8521642866785123e-05, "loss": 2.2015, "step": 8116 }, { "epoch": 0.2, "learning_rate": 1.8521225093440907e-05, "loss": 2.3489, "step": 8117 }, { "epoch": 0.2, "learning_rate": 1.852080726578815e-05, "loss": 2.1931, "step": 8118 }, { "epoch": 0.2, "learning_rate": 1.852038938382951e-05, "loss": 2.1897, "step": 8119 }, { "epoch": 0.2, "learning_rate": 1.8519971447567652e-05, "loss": 2.1126, "step": 8120 }, { "epoch": 0.2, "learning_rate": 1.851955345700524e-05, "loss": 2.1272, "step": 8121 }, { "epoch": 0.2, "learning_rate": 1.8519135412144934e-05, "loss": 2.1844, "step": 8122 }, { "epoch": 0.2, "learning_rate": 1.8518717312989403e-05, "loss": 2.2588, "step": 8123 }, { "epoch": 0.2, "learning_rate": 1.8518299159541314e-05, "loss": 2.3677, "step": 8124 }, { "epoch": 0.2, "learning_rate": 1.8517880951803323e-05, "loss": 2.2971, "step": 8125 }, { "epoch": 0.2, "learning_rate": 1.851746268977811e-05, "loss": 2.2646, "step": 8126 }, { "epoch": 0.2, "learning_rate": 1.8517044373468322e-05, "loss": 2.2698, "step": 8127 }, { "epoch": 0.2, "learning_rate": 1.851662600287664e-05, "loss": 2.2275, "step": 8128 }, { "epoch": 0.2, "learning_rate": 1.851620757800572e-05, "loss": 2.2289, "step": 8129 }, { "epoch": 0.2, "learning_rate": 1.8515789098858238e-05, "loss": 2.1699, "step": 8130 }, { "epoch": 0.2, "learning_rate": 1.8515370565436854e-05, "loss": 2.2563, "step": 8131 }, { "epoch": 0.2, "learning_rate": 1.851495197774424e-05, "loss": 2.1163, "step": 8132 }, { "epoch": 0.2, "learning_rate": 1.851453333578306e-05, "loss": 2.1875, "step": 8133 }, { "epoch": 0.2, "learning_rate": 1.8514114639555985e-05, "loss": 2.3497, "step": 8134 }, { "epoch": 0.2, "learning_rate": 1.8513695889065685e-05, "loss": 2.185, "step": 8135 }, { "epoch": 0.2, "learning_rate": 1.8513277084314827e-05, "loss": 2.054, "step": 8136 }, { "epoch": 0.2, "learning_rate": 1.8512858225306077e-05, "loss": 2.1653, "step": 8137 }, { "epoch": 0.2, "learning_rate": 1.851243931204211e-05, "loss": 2.1407, "step": 8138 }, { "epoch": 0.2, "learning_rate": 1.851202034452559e-05, "loss": 2.0763, "step": 8139 }, { "epoch": 0.2, "learning_rate": 1.8511601322759194e-05, "loss": 2.1302, "step": 8140 }, { "epoch": 0.2, "learning_rate": 1.8511182246745587e-05, "loss": 2.265, "step": 8141 }, { "epoch": 0.2, "learning_rate": 1.851076311648744e-05, "loss": 2.3869, "step": 8142 }, { "epoch": 0.2, "learning_rate": 1.851034393198743e-05, "loss": 2.2093, "step": 8143 }, { "epoch": 0.2, "learning_rate": 1.850992469324822e-05, "loss": 2.2701, "step": 8144 }, { "epoch": 0.2, "learning_rate": 1.850950540027249e-05, "loss": 2.2922, "step": 8145 }, { "epoch": 0.2, "learning_rate": 1.8509086053062908e-05, "loss": 2.0505, "step": 8146 }, { "epoch": 0.2, "learning_rate": 1.8508666651622146e-05, "loss": 2.2629, "step": 8147 }, { "epoch": 0.2, "learning_rate": 1.8508247195952882e-05, "loss": 2.208, "step": 8148 }, { "epoch": 0.2, "learning_rate": 1.8507827686057784e-05, "loss": 2.0627, "step": 8149 }, { "epoch": 0.2, "learning_rate": 1.8507408121939527e-05, "loss": 2.1076, "step": 8150 }, { "epoch": 0.2, "learning_rate": 1.8506988503600784e-05, "loss": 2.1168, "step": 8151 }, { "epoch": 0.2, "learning_rate": 1.8506568831044234e-05, "loss": 2.2098, "step": 8152 }, { "epoch": 0.2, "learning_rate": 1.8506149104272547e-05, "loss": 2.2063, "step": 8153 }, { "epoch": 0.2, "learning_rate": 1.85057293232884e-05, "loss": 2.1811, "step": 8154 }, { "epoch": 0.2, "learning_rate": 1.850530948809447e-05, "loss": 2.2494, "step": 8155 }, { "epoch": 0.2, "learning_rate": 1.850488959869343e-05, "loss": 2.1396, "step": 8156 }, { "epoch": 0.2, "learning_rate": 1.8504469655087957e-05, "loss": 2.2457, "step": 8157 }, { "epoch": 0.2, "learning_rate": 1.8504049657280726e-05, "loss": 2.145, "step": 8158 }, { "epoch": 0.2, "learning_rate": 1.8503629605274412e-05, "loss": 2.3834, "step": 8159 }, { "epoch": 0.2, "learning_rate": 1.8503209499071698e-05, "loss": 2.1535, "step": 8160 }, { "epoch": 0.2, "learning_rate": 1.8502789338675263e-05, "loss": 2.1317, "step": 8161 }, { "epoch": 0.2, "learning_rate": 1.850236912408778e-05, "loss": 2.2563, "step": 8162 }, { "epoch": 0.2, "learning_rate": 1.8501948855311923e-05, "loss": 2.151, "step": 8163 }, { "epoch": 0.2, "learning_rate": 1.8501528532350374e-05, "loss": 2.2004, "step": 8164 }, { "epoch": 0.2, "learning_rate": 1.8501108155205818e-05, "loss": 2.1479, "step": 8165 }, { "epoch": 0.2, "learning_rate": 1.8500687723880928e-05, "loss": 2.0431, "step": 8166 }, { "epoch": 0.2, "learning_rate": 1.8500267238378385e-05, "loss": 2.2088, "step": 8167 }, { "epoch": 0.2, "learning_rate": 1.8499846698700867e-05, "loss": 2.2906, "step": 8168 }, { "epoch": 0.2, "learning_rate": 1.8499426104851056e-05, "loss": 2.3197, "step": 8169 }, { "epoch": 0.2, "learning_rate": 1.8499005456831635e-05, "loss": 2.121, "step": 8170 }, { "epoch": 0.2, "learning_rate": 1.849858475464528e-05, "loss": 2.2183, "step": 8171 }, { "epoch": 0.2, "learning_rate": 1.849816399829468e-05, "loss": 2.359, "step": 8172 }, { "epoch": 0.2, "learning_rate": 1.8497743187782503e-05, "loss": 2.2047, "step": 8173 }, { "epoch": 0.2, "learning_rate": 1.8497322323111445e-05, "loss": 2.2618, "step": 8174 }, { "epoch": 0.2, "learning_rate": 1.8496901404284182e-05, "loss": 2.0259, "step": 8175 }, { "epoch": 0.2, "learning_rate": 1.8496480431303397e-05, "loss": 2.319, "step": 8176 }, { "epoch": 0.2, "learning_rate": 1.849605940417177e-05, "loss": 2.2409, "step": 8177 }, { "epoch": 0.2, "learning_rate": 1.8495638322891992e-05, "loss": 2.3201, "step": 8178 }, { "epoch": 0.2, "learning_rate": 1.849521718746674e-05, "loss": 1.9414, "step": 8179 }, { "epoch": 0.2, "learning_rate": 1.84947959978987e-05, "loss": 2.3001, "step": 8180 }, { "epoch": 0.2, "learning_rate": 1.849437475419056e-05, "loss": 2.0818, "step": 8181 }, { "epoch": 0.2, "learning_rate": 1.8493953456344998e-05, "loss": 1.9916, "step": 8182 }, { "epoch": 0.2, "learning_rate": 1.8493532104364704e-05, "loss": 2.2538, "step": 8183 }, { "epoch": 0.2, "learning_rate": 1.8493110698252364e-05, "loss": 2.1852, "step": 8184 }, { "epoch": 0.2, "learning_rate": 1.849268923801066e-05, "loss": 2.5757, "step": 8185 }, { "epoch": 0.2, "learning_rate": 1.8492267723642275e-05, "loss": 2.114, "step": 8186 }, { "epoch": 0.2, "learning_rate": 1.8491846155149907e-05, "loss": 2.0299, "step": 8187 }, { "epoch": 0.2, "learning_rate": 1.8491424532536234e-05, "loss": 2.1251, "step": 8188 }, { "epoch": 0.2, "learning_rate": 1.8491002855803945e-05, "loss": 2.17, "step": 8189 }, { "epoch": 0.2, "learning_rate": 1.8490581124955727e-05, "loss": 2.1778, "step": 8190 }, { "epoch": 0.2, "learning_rate": 1.849015933999427e-05, "loss": 2.0803, "step": 8191 }, { "epoch": 0.2, "learning_rate": 1.848973750092226e-05, "loss": 1.9995, "step": 8192 }, { "epoch": 0.2, "learning_rate": 1.8489315607742383e-05, "loss": 2.3265, "step": 8193 }, { "epoch": 0.2, "learning_rate": 1.8488893660457335e-05, "loss": 2.3086, "step": 8194 }, { "epoch": 0.2, "learning_rate": 1.84884716590698e-05, "loss": 2.3283, "step": 8195 }, { "epoch": 0.2, "learning_rate": 1.848804960358247e-05, "loss": 2.4277, "step": 8196 }, { "epoch": 0.2, "learning_rate": 1.848762749399803e-05, "loss": 2.2377, "step": 8197 }, { "epoch": 0.2, "learning_rate": 1.848720533031918e-05, "loss": 2.2862, "step": 8198 }, { "epoch": 0.2, "learning_rate": 1.84867831125486e-05, "loss": 2.1868, "step": 8199 }, { "epoch": 0.2, "learning_rate": 1.8486360840688983e-05, "loss": 2.1065, "step": 8200 }, { "epoch": 0.2, "learning_rate": 1.848593851474303e-05, "loss": 2.3654, "step": 8201 }, { "epoch": 0.2, "learning_rate": 1.848551613471342e-05, "loss": 2.0088, "step": 8202 }, { "epoch": 0.2, "learning_rate": 1.8485093700602853e-05, "loss": 2.1674, "step": 8203 }, { "epoch": 0.2, "learning_rate": 1.8484671212414013e-05, "loss": 2.2673, "step": 8204 }, { "epoch": 0.2, "learning_rate": 1.8484248670149604e-05, "loss": 2.0847, "step": 8205 }, { "epoch": 0.2, "learning_rate": 1.8483826073812313e-05, "loss": 2.2114, "step": 8206 }, { "epoch": 0.2, "learning_rate": 1.8483403423404833e-05, "loss": 2.0356, "step": 8207 }, { "epoch": 0.2, "learning_rate": 1.8482980718929857e-05, "loss": 2.296, "step": 8208 }, { "epoch": 0.2, "learning_rate": 1.8482557960390083e-05, "loss": 2.0747, "step": 8209 }, { "epoch": 0.2, "learning_rate": 1.84821351477882e-05, "loss": 2.2318, "step": 8210 }, { "epoch": 0.2, "learning_rate": 1.8481712281126907e-05, "loss": 2.5196, "step": 8211 }, { "epoch": 0.2, "learning_rate": 1.84812893604089e-05, "loss": 2.3209, "step": 8212 }, { "epoch": 0.2, "learning_rate": 1.8480866385636867e-05, "loss": 1.98, "step": 8213 }, { "epoch": 0.2, "learning_rate": 1.848044335681351e-05, "loss": 2.2355, "step": 8214 }, { "epoch": 0.2, "learning_rate": 1.8480020273941525e-05, "loss": 2.1545, "step": 8215 }, { "epoch": 0.2, "learning_rate": 1.8479597137023605e-05, "loss": 2.1552, "step": 8216 }, { "epoch": 0.2, "learning_rate": 1.8479173946062453e-05, "loss": 2.2155, "step": 8217 }, { "epoch": 0.2, "learning_rate": 1.847875070106076e-05, "loss": 2.1616, "step": 8218 }, { "epoch": 0.2, "learning_rate": 1.8478327402021223e-05, "loss": 2.2044, "step": 8219 }, { "epoch": 0.2, "learning_rate": 1.8477904048946547e-05, "loss": 2.3006, "step": 8220 }, { "epoch": 0.2, "learning_rate": 1.847748064183942e-05, "loss": 2.2924, "step": 8221 }, { "epoch": 0.2, "learning_rate": 1.8477057180702552e-05, "loss": 2.4644, "step": 8222 }, { "epoch": 0.2, "learning_rate": 1.8476633665538634e-05, "loss": 2.215, "step": 8223 }, { "epoch": 0.2, "learning_rate": 1.8476210096350365e-05, "loss": 2.2741, "step": 8224 }, { "epoch": 0.2, "learning_rate": 1.847578647314045e-05, "loss": 2.174, "step": 8225 }, { "epoch": 0.2, "learning_rate": 1.8475362795911584e-05, "loss": 2.261, "step": 8226 }, { "epoch": 0.2, "learning_rate": 1.847493906466647e-05, "loss": 2.2401, "step": 8227 }, { "epoch": 0.2, "learning_rate": 1.8474515279407806e-05, "loss": 2.065, "step": 8228 }, { "epoch": 0.2, "learning_rate": 1.8474091440138298e-05, "loss": 2.0657, "step": 8229 }, { "epoch": 0.2, "learning_rate": 1.8473667546860638e-05, "loss": 2.3263, "step": 8230 }, { "epoch": 0.2, "learning_rate": 1.8473243599577536e-05, "loss": 2.0935, "step": 8231 }, { "epoch": 0.2, "learning_rate": 1.8472819598291692e-05, "loss": 2.2873, "step": 8232 }, { "epoch": 0.2, "learning_rate": 1.8472395543005805e-05, "loss": 2.2081, "step": 8233 }, { "epoch": 0.2, "learning_rate": 1.8471971433722583e-05, "loss": 2.3134, "step": 8234 }, { "epoch": 0.2, "learning_rate": 1.8471547270444724e-05, "loss": 2.2295, "step": 8235 }, { "epoch": 0.2, "learning_rate": 1.8471123053174932e-05, "loss": 2.1262, "step": 8236 }, { "epoch": 0.2, "learning_rate": 1.8470698781915918e-05, "loss": 2.2396, "step": 8237 }, { "epoch": 0.2, "learning_rate": 1.8470274456670374e-05, "loss": 2.1687, "step": 8238 }, { "epoch": 0.2, "learning_rate": 1.8469850077441015e-05, "loss": 2.1996, "step": 8239 }, { "epoch": 0.2, "learning_rate": 1.8469425644230537e-05, "loss": 1.9692, "step": 8240 }, { "epoch": 0.2, "learning_rate": 1.846900115704165e-05, "loss": 2.2801, "step": 8241 }, { "epoch": 0.2, "learning_rate": 1.8468576615877065e-05, "loss": 2.1306, "step": 8242 }, { "epoch": 0.2, "learning_rate": 1.8468152020739472e-05, "loss": 2.3055, "step": 8243 }, { "epoch": 0.2, "learning_rate": 1.8467727371631596e-05, "loss": 2.3892, "step": 8244 }, { "epoch": 0.2, "learning_rate": 1.8467302668556126e-05, "loss": 2.2739, "step": 8245 }, { "epoch": 0.2, "learning_rate": 1.8466877911515778e-05, "loss": 2.077, "step": 8246 }, { "epoch": 0.2, "learning_rate": 1.846645310051326e-05, "loss": 2.2384, "step": 8247 }, { "epoch": 0.2, "learning_rate": 1.8466028235551276e-05, "loss": 2.0748, "step": 8248 }, { "epoch": 0.2, "learning_rate": 1.8465603316632537e-05, "loss": 2.3427, "step": 8249 }, { "epoch": 0.2, "learning_rate": 1.846517834375975e-05, "loss": 2.2349, "step": 8250 }, { "epoch": 0.2, "learning_rate": 1.846475331693562e-05, "loss": 2.08, "step": 8251 }, { "epoch": 0.2, "learning_rate": 1.8464328236162857e-05, "loss": 1.9768, "step": 8252 }, { "epoch": 0.2, "learning_rate": 1.8463903101444175e-05, "loss": 2.0075, "step": 8253 }, { "epoch": 0.2, "learning_rate": 1.8463477912782278e-05, "loss": 2.1385, "step": 8254 }, { "epoch": 0.2, "learning_rate": 1.846305267017988e-05, "loss": 2.3098, "step": 8255 }, { "epoch": 0.2, "learning_rate": 1.846262737363969e-05, "loss": 2.1356, "step": 8256 }, { "epoch": 0.2, "learning_rate": 1.8462202023164416e-05, "loss": 2.1705, "step": 8257 }, { "epoch": 0.2, "learning_rate": 1.846177661875677e-05, "loss": 2.1344, "step": 8258 }, { "epoch": 0.2, "learning_rate": 1.8461351160419467e-05, "loss": 2.2234, "step": 8259 }, { "epoch": 0.2, "learning_rate": 1.846092564815521e-05, "loss": 2.1422, "step": 8260 }, { "epoch": 0.2, "learning_rate": 1.8460500081966725e-05, "loss": 2.2432, "step": 8261 }, { "epoch": 0.2, "learning_rate": 1.8460074461856707e-05, "loss": 2.0772, "step": 8262 }, { "epoch": 0.2, "learning_rate": 1.8459648787827882e-05, "loss": 2.1149, "step": 8263 }, { "epoch": 0.2, "learning_rate": 1.8459223059882957e-05, "loss": 2.0821, "step": 8264 }, { "epoch": 0.2, "learning_rate": 1.845879727802465e-05, "loss": 2.2848, "step": 8265 }, { "epoch": 0.2, "learning_rate": 1.8458371442255668e-05, "loss": 2.3061, "step": 8266 }, { "epoch": 0.2, "learning_rate": 1.845794555257873e-05, "loss": 2.1476, "step": 8267 }, { "epoch": 0.2, "learning_rate": 1.8457519608996546e-05, "loss": 2.2052, "step": 8268 }, { "epoch": 0.2, "learning_rate": 1.8457093611511835e-05, "loss": 2.2961, "step": 8269 }, { "epoch": 0.2, "learning_rate": 1.8456667560127308e-05, "loss": 2.2681, "step": 8270 }, { "epoch": 0.2, "learning_rate": 1.8456241454845685e-05, "loss": 2.2144, "step": 8271 }, { "epoch": 0.2, "learning_rate": 1.845581529566968e-05, "loss": 2.2501, "step": 8272 }, { "epoch": 0.2, "learning_rate": 1.845538908260201e-05, "loss": 2.1569, "step": 8273 }, { "epoch": 0.2, "learning_rate": 1.8454962815645384e-05, "loss": 2.1262, "step": 8274 }, { "epoch": 0.2, "learning_rate": 1.845453649480253e-05, "loss": 2.2608, "step": 8275 }, { "epoch": 0.2, "learning_rate": 1.8454110120076157e-05, "loss": 2.2425, "step": 8276 }, { "epoch": 0.2, "learning_rate": 1.8453683691468986e-05, "loss": 2.1947, "step": 8277 }, { "epoch": 0.2, "learning_rate": 1.8453257208983734e-05, "loss": 2.1242, "step": 8278 }, { "epoch": 0.2, "learning_rate": 1.845283067262312e-05, "loss": 2.2914, "step": 8279 }, { "epoch": 0.2, "learning_rate": 1.845240408238986e-05, "loss": 2.2237, "step": 8280 }, { "epoch": 0.2, "learning_rate": 1.8451977438286674e-05, "loss": 2.2371, "step": 8281 }, { "epoch": 0.2, "learning_rate": 1.845155074031628e-05, "loss": 2.4315, "step": 8282 }, { "epoch": 0.2, "learning_rate": 1.84511239884814e-05, "loss": 2.2921, "step": 8283 }, { "epoch": 0.2, "learning_rate": 1.8450697182784754e-05, "loss": 2.1955, "step": 8284 }, { "epoch": 0.2, "learning_rate": 1.8450270323229063e-05, "loss": 2.1888, "step": 8285 }, { "epoch": 0.2, "learning_rate": 1.8449843409817042e-05, "loss": 2.2391, "step": 8286 }, { "epoch": 0.2, "learning_rate": 1.8449416442551412e-05, "loss": 2.1868, "step": 8287 }, { "epoch": 0.2, "learning_rate": 1.8448989421434903e-05, "loss": 2.2368, "step": 8288 }, { "epoch": 0.2, "learning_rate": 1.844856234647023e-05, "loss": 2.2184, "step": 8289 }, { "epoch": 0.2, "learning_rate": 1.8448135217660115e-05, "loss": 2.1862, "step": 8290 }, { "epoch": 0.2, "learning_rate": 1.844770803500728e-05, "loss": 2.0994, "step": 8291 }, { "epoch": 0.2, "learning_rate": 1.844728079851445e-05, "loss": 2.1539, "step": 8292 }, { "epoch": 0.2, "learning_rate": 1.8446853508184345e-05, "loss": 2.2488, "step": 8293 }, { "epoch": 0.2, "learning_rate": 1.844642616401969e-05, "loss": 2.262, "step": 8294 }, { "epoch": 0.2, "learning_rate": 1.844599876602321e-05, "loss": 2.3485, "step": 8295 }, { "epoch": 0.2, "learning_rate": 1.8445571314197628e-05, "loss": 2.0634, "step": 8296 }, { "epoch": 0.2, "learning_rate": 1.844514380854567e-05, "loss": 2.1479, "step": 8297 }, { "epoch": 0.2, "learning_rate": 1.8444716249070055e-05, "loss": 2.2036, "step": 8298 }, { "epoch": 0.2, "learning_rate": 1.8444288635773507e-05, "loss": 1.9388, "step": 8299 }, { "epoch": 0.2, "learning_rate": 1.8443860968658763e-05, "loss": 2.3158, "step": 8300 }, { "epoch": 0.2, "learning_rate": 1.844343324772854e-05, "loss": 2.1032, "step": 8301 }, { "epoch": 0.2, "learning_rate": 1.844300547298556e-05, "loss": 2.088, "step": 8302 }, { "epoch": 0.2, "learning_rate": 1.844257764443256e-05, "loss": 2.2741, "step": 8303 }, { "epoch": 0.2, "learning_rate": 1.8442149762072258e-05, "loss": 2.111, "step": 8304 }, { "epoch": 0.2, "learning_rate": 1.8441721825907387e-05, "loss": 2.3934, "step": 8305 }, { "epoch": 0.2, "learning_rate": 1.844129383594067e-05, "loss": 2.3046, "step": 8306 }, { "epoch": 0.2, "learning_rate": 1.8440865792174837e-05, "loss": 2.1776, "step": 8307 }, { "epoch": 0.2, "learning_rate": 1.8440437694612616e-05, "loss": 2.1715, "step": 8308 }, { "epoch": 0.2, "learning_rate": 1.844000954325673e-05, "loss": 1.9836, "step": 8309 }, { "epoch": 0.2, "learning_rate": 1.843958133810992e-05, "loss": 2.1158, "step": 8310 }, { "epoch": 0.2, "learning_rate": 1.8439153079174907e-05, "loss": 1.9857, "step": 8311 }, { "epoch": 0.2, "learning_rate": 1.8438724766454418e-05, "loss": 2.2198, "step": 8312 }, { "epoch": 0.2, "learning_rate": 1.8438296399951185e-05, "loss": 2.2088, "step": 8313 }, { "epoch": 0.2, "learning_rate": 1.8437867979667944e-05, "loss": 2.1317, "step": 8314 }, { "epoch": 0.2, "learning_rate": 1.8437439505607414e-05, "loss": 2.1327, "step": 8315 }, { "epoch": 0.2, "learning_rate": 1.8437010977772338e-05, "loss": 2.3506, "step": 8316 }, { "epoch": 0.21, "learning_rate": 1.8436582396165437e-05, "loss": 2.0054, "step": 8317 }, { "epoch": 0.21, "learning_rate": 1.843615376078945e-05, "loss": 2.2004, "step": 8318 }, { "epoch": 0.21, "learning_rate": 1.8435725071647107e-05, "loss": 2.2726, "step": 8319 }, { "epoch": 0.21, "learning_rate": 1.8435296328741136e-05, "loss": 1.98, "step": 8320 }, { "epoch": 0.21, "learning_rate": 1.843486753207427e-05, "loss": 2.3239, "step": 8321 }, { "epoch": 0.21, "learning_rate": 1.843443868164925e-05, "loss": 2.1822, "step": 8322 }, { "epoch": 0.21, "learning_rate": 1.8434009777468806e-05, "loss": 2.1059, "step": 8323 }, { "epoch": 0.21, "learning_rate": 1.843358081953566e-05, "loss": 2.1226, "step": 8324 }, { "epoch": 0.21, "learning_rate": 1.8433151807852562e-05, "loss": 2.2434, "step": 8325 }, { "epoch": 0.21, "learning_rate": 1.8432722742422234e-05, "loss": 2.1875, "step": 8326 }, { "epoch": 0.21, "learning_rate": 1.843229362324742e-05, "loss": 2.29, "step": 8327 }, { "epoch": 0.21, "learning_rate": 1.843186445033085e-05, "loss": 1.9771, "step": 8328 }, { "epoch": 0.21, "learning_rate": 1.843143522367526e-05, "loss": 2.1072, "step": 8329 }, { "epoch": 0.21, "learning_rate": 1.8431005943283385e-05, "loss": 2.1612, "step": 8330 }, { "epoch": 0.21, "learning_rate": 1.843057660915796e-05, "loss": 2.0963, "step": 8331 }, { "epoch": 0.21, "learning_rate": 1.8430147221301727e-05, "loss": 1.9513, "step": 8332 }, { "epoch": 0.21, "learning_rate": 1.8429717779717417e-05, "loss": 2.2617, "step": 8333 }, { "epoch": 0.21, "learning_rate": 1.842928828440777e-05, "loss": 2.1454, "step": 8334 }, { "epoch": 0.21, "learning_rate": 1.8428858735375517e-05, "loss": 2.2987, "step": 8335 }, { "epoch": 0.21, "learning_rate": 1.8428429132623398e-05, "loss": 2.1706, "step": 8336 }, { "epoch": 0.21, "learning_rate": 1.8427999476154158e-05, "loss": 2.3158, "step": 8337 }, { "epoch": 0.21, "learning_rate": 1.8427569765970534e-05, "loss": 2.2935, "step": 8338 }, { "epoch": 0.21, "learning_rate": 1.8427140002075256e-05, "loss": 2.2203, "step": 8339 }, { "epoch": 0.21, "learning_rate": 1.8426710184471072e-05, "loss": 2.0434, "step": 8340 }, { "epoch": 0.21, "learning_rate": 1.8426280313160717e-05, "loss": 2.2547, "step": 8341 }, { "epoch": 0.21, "learning_rate": 1.842585038814693e-05, "loss": 2.2606, "step": 8342 }, { "epoch": 0.21, "learning_rate": 1.8425420409432453e-05, "loss": 2.1021, "step": 8343 }, { "epoch": 0.21, "learning_rate": 1.8424990377020026e-05, "loss": 2.2607, "step": 8344 }, { "epoch": 0.21, "learning_rate": 1.842456029091239e-05, "loss": 2.2104, "step": 8345 }, { "epoch": 0.21, "learning_rate": 1.8424130151112285e-05, "loss": 2.2328, "step": 8346 }, { "epoch": 0.21, "learning_rate": 1.8423699957622455e-05, "loss": 2.2341, "step": 8347 }, { "epoch": 0.21, "learning_rate": 1.842326971044564e-05, "loss": 2.4094, "step": 8348 }, { "epoch": 0.21, "learning_rate": 1.842283940958458e-05, "loss": 2.2006, "step": 8349 }, { "epoch": 0.21, "learning_rate": 1.842240905504202e-05, "loss": 2.3338, "step": 8350 }, { "epoch": 0.21, "learning_rate": 1.8421978646820705e-05, "loss": 2.1726, "step": 8351 }, { "epoch": 0.21, "learning_rate": 1.8421548184923374e-05, "loss": 2.051, "step": 8352 }, { "epoch": 0.21, "learning_rate": 1.842111766935277e-05, "loss": 2.1809, "step": 8353 }, { "epoch": 0.21, "learning_rate": 1.842068710011164e-05, "loss": 2.1887, "step": 8354 }, { "epoch": 0.21, "learning_rate": 1.8420256477202726e-05, "loss": 2.0837, "step": 8355 }, { "epoch": 0.21, "learning_rate": 1.8419825800628774e-05, "loss": 2.074, "step": 8356 }, { "epoch": 0.21, "learning_rate": 1.841939507039253e-05, "loss": 2.0179, "step": 8357 }, { "epoch": 0.21, "learning_rate": 1.8418964286496734e-05, "loss": 2.3825, "step": 8358 }, { "epoch": 0.21, "learning_rate": 1.8418533448944135e-05, "loss": 2.2232, "step": 8359 }, { "epoch": 0.21, "learning_rate": 1.841810255773748e-05, "loss": 2.2174, "step": 8360 }, { "epoch": 0.21, "learning_rate": 1.8417671612879517e-05, "loss": 2.2965, "step": 8361 }, { "epoch": 0.21, "learning_rate": 1.8417240614372986e-05, "loss": 2.2505, "step": 8362 }, { "epoch": 0.21, "learning_rate": 1.841680956222064e-05, "loss": 2.2612, "step": 8363 }, { "epoch": 0.21, "learning_rate": 1.8416378456425223e-05, "loss": 2.2634, "step": 8364 }, { "epoch": 0.21, "learning_rate": 1.8415947296989483e-05, "loss": 2.1829, "step": 8365 }, { "epoch": 0.21, "learning_rate": 1.8415516083916168e-05, "loss": 2.2207, "step": 8366 }, { "epoch": 0.21, "learning_rate": 1.8415084817208025e-05, "loss": 2.1541, "step": 8367 }, { "epoch": 0.21, "learning_rate": 1.8414653496867806e-05, "loss": 2.1979, "step": 8368 }, { "epoch": 0.21, "learning_rate": 1.8414222122898257e-05, "loss": 2.2776, "step": 8369 }, { "epoch": 0.21, "learning_rate": 1.841379069530213e-05, "loss": 2.0212, "step": 8370 }, { "epoch": 0.21, "learning_rate": 1.841335921408217e-05, "loss": 2.2439, "step": 8371 }, { "epoch": 0.21, "learning_rate": 1.8412927679241133e-05, "loss": 2.0973, "step": 8372 }, { "epoch": 0.21, "learning_rate": 1.8412496090781765e-05, "loss": 2.311, "step": 8373 }, { "epoch": 0.21, "learning_rate": 1.8412064448706818e-05, "loss": 2.2324, "step": 8374 }, { "epoch": 0.21, "learning_rate": 1.841163275301904e-05, "loss": 2.0523, "step": 8375 }, { "epoch": 0.21, "learning_rate": 1.841120100372119e-05, "loss": 2.1727, "step": 8376 }, { "epoch": 0.21, "learning_rate": 1.841076920081601e-05, "loss": 2.1777, "step": 8377 }, { "epoch": 0.21, "learning_rate": 1.8410337344306262e-05, "loss": 2.2089, "step": 8378 }, { "epoch": 0.21, "learning_rate": 1.8409905434194688e-05, "loss": 2.2168, "step": 8379 }, { "epoch": 0.21, "learning_rate": 1.840947347048405e-05, "loss": 2.3105, "step": 8380 }, { "epoch": 0.21, "learning_rate": 1.8409041453177096e-05, "loss": 2.3875, "step": 8381 }, { "epoch": 0.21, "learning_rate": 1.840860938227658e-05, "loss": 2.1172, "step": 8382 }, { "epoch": 0.21, "learning_rate": 1.8408177257785256e-05, "loss": 2.1236, "step": 8383 }, { "epoch": 0.21, "learning_rate": 1.8407745079705878e-05, "loss": 2.1447, "step": 8384 }, { "epoch": 0.21, "learning_rate": 1.84073128480412e-05, "loss": 2.2426, "step": 8385 }, { "epoch": 0.21, "learning_rate": 1.840688056279398e-05, "loss": 2.1349, "step": 8386 }, { "epoch": 0.21, "learning_rate": 1.840644822396697e-05, "loss": 2.218, "step": 8387 }, { "epoch": 0.21, "learning_rate": 1.840601583156292e-05, "loss": 2.0861, "step": 8388 }, { "epoch": 0.21, "learning_rate": 1.84055833855846e-05, "loss": 2.3665, "step": 8389 }, { "epoch": 0.21, "learning_rate": 1.8405150886034753e-05, "loss": 2.1466, "step": 8390 }, { "epoch": 0.21, "learning_rate": 1.840471833291614e-05, "loss": 2.1015, "step": 8391 }, { "epoch": 0.21, "learning_rate": 1.840428572623152e-05, "loss": 2.0722, "step": 8392 }, { "epoch": 0.21, "learning_rate": 1.8403853065983646e-05, "loss": 2.1934, "step": 8393 }, { "epoch": 0.21, "learning_rate": 1.840342035217528e-05, "loss": 2.1037, "step": 8394 }, { "epoch": 0.21, "learning_rate": 1.8402987584809176e-05, "loss": 2.2483, "step": 8395 }, { "epoch": 0.21, "learning_rate": 1.8402554763888095e-05, "loss": 2.0841, "step": 8396 }, { "epoch": 0.21, "learning_rate": 1.8402121889414795e-05, "loss": 2.0645, "step": 8397 }, { "epoch": 0.21, "learning_rate": 1.8401688961392032e-05, "loss": 2.1072, "step": 8398 }, { "epoch": 0.21, "learning_rate": 1.8401255979822566e-05, "loss": 2.4413, "step": 8399 }, { "epoch": 0.21, "learning_rate": 1.840082294470916e-05, "loss": 2.2594, "step": 8400 }, { "epoch": 0.21, "learning_rate": 1.8400389856054574e-05, "loss": 2.1616, "step": 8401 }, { "epoch": 0.21, "learning_rate": 1.8399956713861563e-05, "loss": 2.1199, "step": 8402 }, { "epoch": 0.21, "learning_rate": 1.8399523518132888e-05, "loss": 2.2806, "step": 8403 }, { "epoch": 0.21, "learning_rate": 1.8399090268871314e-05, "loss": 2.2473, "step": 8404 }, { "epoch": 0.21, "learning_rate": 1.8398656966079603e-05, "loss": 2.1305, "step": 8405 }, { "epoch": 0.21, "learning_rate": 1.8398223609760512e-05, "loss": 2.1477, "step": 8406 }, { "epoch": 0.21, "learning_rate": 1.8397790199916806e-05, "loss": 2.0987, "step": 8407 }, { "epoch": 0.21, "learning_rate": 1.8397356736551246e-05, "loss": 2.2373, "step": 8408 }, { "epoch": 0.21, "learning_rate": 1.8396923219666597e-05, "loss": 2.0997, "step": 8409 }, { "epoch": 0.21, "learning_rate": 1.8396489649265613e-05, "loss": 2.0934, "step": 8410 }, { "epoch": 0.21, "learning_rate": 1.839605602535107e-05, "loss": 2.2874, "step": 8411 }, { "epoch": 0.21, "learning_rate": 1.8395622347925726e-05, "loss": 2.2861, "step": 8412 }, { "epoch": 0.21, "learning_rate": 1.839518861699234e-05, "loss": 2.1492, "step": 8413 }, { "epoch": 0.21, "learning_rate": 1.8394754832553684e-05, "loss": 2.1865, "step": 8414 }, { "epoch": 0.21, "learning_rate": 1.8394320994612523e-05, "loss": 2.0485, "step": 8415 }, { "epoch": 0.21, "learning_rate": 1.8393887103171613e-05, "loss": 2.0187, "step": 8416 }, { "epoch": 0.21, "learning_rate": 1.8393453158233728e-05, "loss": 2.1291, "step": 8417 }, { "epoch": 0.21, "learning_rate": 1.839301915980163e-05, "loss": 2.0303, "step": 8418 }, { "epoch": 0.21, "learning_rate": 1.8392585107878085e-05, "loss": 2.1855, "step": 8419 }, { "epoch": 0.21, "learning_rate": 1.839215100246586e-05, "loss": 2.2506, "step": 8420 }, { "epoch": 0.21, "learning_rate": 1.8391716843567724e-05, "loss": 2.1527, "step": 8421 }, { "epoch": 0.21, "learning_rate": 1.839128263118644e-05, "loss": 2.2268, "step": 8422 }, { "epoch": 0.21, "learning_rate": 1.839084836532478e-05, "loss": 2.0831, "step": 8423 }, { "epoch": 0.21, "learning_rate": 1.8390414045985504e-05, "loss": 2.1656, "step": 8424 }, { "epoch": 0.21, "learning_rate": 1.838997967317139e-05, "loss": 2.1548, "step": 8425 }, { "epoch": 0.21, "learning_rate": 1.8389545246885195e-05, "loss": 2.2184, "step": 8426 }, { "epoch": 0.21, "learning_rate": 1.8389110767129697e-05, "loss": 2.2178, "step": 8427 }, { "epoch": 0.21, "learning_rate": 1.8388676233907664e-05, "loss": 2.2195, "step": 8428 }, { "epoch": 0.21, "learning_rate": 1.8388241647221863e-05, "loss": 2.3475, "step": 8429 }, { "epoch": 0.21, "learning_rate": 1.8387807007075062e-05, "loss": 2.2644, "step": 8430 }, { "epoch": 0.21, "learning_rate": 1.8387372313470034e-05, "loss": 2.3038, "step": 8431 }, { "epoch": 0.21, "learning_rate": 1.838693756640955e-05, "loss": 2.1768, "step": 8432 }, { "epoch": 0.21, "learning_rate": 1.8386502765896383e-05, "loss": 2.2335, "step": 8433 }, { "epoch": 0.21, "learning_rate": 1.8386067911933296e-05, "loss": 2.2699, "step": 8434 }, { "epoch": 0.21, "learning_rate": 1.838563300452307e-05, "loss": 2.1319, "step": 8435 }, { "epoch": 0.21, "learning_rate": 1.8385198043668468e-05, "loss": 2.3694, "step": 8436 }, { "epoch": 0.21, "learning_rate": 1.8384763029372266e-05, "loss": 2.061, "step": 8437 }, { "epoch": 0.21, "learning_rate": 1.8384327961637237e-05, "loss": 2.155, "step": 8438 }, { "epoch": 0.21, "learning_rate": 1.8383892840466155e-05, "loss": 2.1859, "step": 8439 }, { "epoch": 0.21, "learning_rate": 1.838345766586179e-05, "loss": 2.2663, "step": 8440 }, { "epoch": 0.21, "learning_rate": 1.8383022437826916e-05, "loss": 2.2029, "step": 8441 }, { "epoch": 0.21, "learning_rate": 1.8382587156364312e-05, "loss": 2.3108, "step": 8442 }, { "epoch": 0.21, "learning_rate": 1.8382151821476748e-05, "loss": 2.4264, "step": 8443 }, { "epoch": 0.21, "learning_rate": 1.8381716433166995e-05, "loss": 2.3755, "step": 8444 }, { "epoch": 0.21, "learning_rate": 1.8381280991437833e-05, "loss": 2.2703, "step": 8445 }, { "epoch": 0.21, "learning_rate": 1.838084549629204e-05, "loss": 2.2281, "step": 8446 }, { "epoch": 0.21, "learning_rate": 1.8380409947732384e-05, "loss": 2.2271, "step": 8447 }, { "epoch": 0.21, "learning_rate": 1.8379974345761643e-05, "loss": 2.2657, "step": 8448 }, { "epoch": 0.21, "learning_rate": 1.8379538690382596e-05, "loss": 2.3319, "step": 8449 }, { "epoch": 0.21, "learning_rate": 1.837910298159802e-05, "loss": 2.2785, "step": 8450 }, { "epoch": 0.21, "learning_rate": 1.8378667219410685e-05, "loss": 2.3363, "step": 8451 }, { "epoch": 0.21, "learning_rate": 1.8378231403823375e-05, "loss": 2.2525, "step": 8452 }, { "epoch": 0.21, "learning_rate": 1.837779553483887e-05, "loss": 2.4106, "step": 8453 }, { "epoch": 0.21, "learning_rate": 1.8377359612459938e-05, "loss": 1.9985, "step": 8454 }, { "epoch": 0.21, "learning_rate": 1.837692363668937e-05, "loss": 2.2603, "step": 8455 }, { "epoch": 0.21, "learning_rate": 1.8376487607529933e-05, "loss": 2.2161, "step": 8456 }, { "epoch": 0.21, "learning_rate": 1.837605152498441e-05, "loss": 2.1726, "step": 8457 }, { "epoch": 0.21, "learning_rate": 1.8375615389055587e-05, "loss": 2.1381, "step": 8458 }, { "epoch": 0.21, "learning_rate": 1.8375179199746228e-05, "loss": 2.1123, "step": 8459 }, { "epoch": 0.21, "learning_rate": 1.837474295705913e-05, "loss": 2.1999, "step": 8460 }, { "epoch": 0.21, "learning_rate": 1.8374306660997062e-05, "loss": 2.2735, "step": 8461 }, { "epoch": 0.21, "learning_rate": 1.8373870311562813e-05, "loss": 2.2268, "step": 8462 }, { "epoch": 0.21, "learning_rate": 1.8373433908759155e-05, "loss": 1.9426, "step": 8463 }, { "epoch": 0.21, "learning_rate": 1.8372997452588874e-05, "loss": 2.1608, "step": 8464 }, { "epoch": 0.21, "learning_rate": 1.8372560943054755e-05, "loss": 2.2586, "step": 8465 }, { "epoch": 0.21, "learning_rate": 1.8372124380159577e-05, "loss": 2.2369, "step": 8466 }, { "epoch": 0.21, "learning_rate": 1.8371687763906116e-05, "loss": 2.094, "step": 8467 }, { "epoch": 0.21, "learning_rate": 1.8371251094297163e-05, "loss": 2.2293, "step": 8468 }, { "epoch": 0.21, "learning_rate": 1.8370814371335502e-05, "loss": 2.3583, "step": 8469 }, { "epoch": 0.21, "learning_rate": 1.8370377595023908e-05, "loss": 2.3881, "step": 8470 }, { "epoch": 0.21, "learning_rate": 1.8369940765365176e-05, "loss": 2.2824, "step": 8471 }, { "epoch": 0.21, "learning_rate": 1.836950388236208e-05, "loss": 2.2021, "step": 8472 }, { "epoch": 0.21, "learning_rate": 1.8369066946017408e-05, "loss": 2.2443, "step": 8473 }, { "epoch": 0.21, "learning_rate": 1.8368629956333945e-05, "loss": 2.1049, "step": 8474 }, { "epoch": 0.21, "learning_rate": 1.8368192913314478e-05, "loss": 2.2232, "step": 8475 }, { "epoch": 0.21, "learning_rate": 1.8367755816961788e-05, "loss": 2.0491, "step": 8476 }, { "epoch": 0.21, "learning_rate": 1.8367318667278663e-05, "loss": 2.2335, "step": 8477 }, { "epoch": 0.21, "learning_rate": 1.8366881464267893e-05, "loss": 2.2137, "step": 8478 }, { "epoch": 0.21, "learning_rate": 1.8366444207932258e-05, "loss": 2.2927, "step": 8479 }, { "epoch": 0.21, "learning_rate": 1.8366006898274547e-05, "loss": 2.18, "step": 8480 }, { "epoch": 0.21, "learning_rate": 1.836556953529755e-05, "loss": 2.2587, "step": 8481 }, { "epoch": 0.21, "learning_rate": 1.836513211900405e-05, "loss": 2.1379, "step": 8482 }, { "epoch": 0.21, "learning_rate": 1.8364694649396838e-05, "loss": 2.215, "step": 8483 }, { "epoch": 0.21, "learning_rate": 1.83642571264787e-05, "loss": 2.0915, "step": 8484 }, { "epoch": 0.21, "learning_rate": 1.8363819550252424e-05, "loss": 2.1792, "step": 8485 }, { "epoch": 0.21, "learning_rate": 1.8363381920720805e-05, "loss": 2.198, "step": 8486 }, { "epoch": 0.21, "learning_rate": 1.8362944237886625e-05, "loss": 2.1921, "step": 8487 }, { "epoch": 0.21, "learning_rate": 1.8362506501752674e-05, "loss": 2.1684, "step": 8488 }, { "epoch": 0.21, "learning_rate": 1.8362068712321746e-05, "loss": 2.0718, "step": 8489 }, { "epoch": 0.21, "learning_rate": 1.8361630869596627e-05, "loss": 2.3357, "step": 8490 }, { "epoch": 0.21, "learning_rate": 1.8361192973580115e-05, "loss": 2.1108, "step": 8491 }, { "epoch": 0.21, "learning_rate": 1.836075502427499e-05, "loss": 2.1781, "step": 8492 }, { "epoch": 0.21, "learning_rate": 1.836031702168405e-05, "loss": 2.1247, "step": 8493 }, { "epoch": 0.21, "learning_rate": 1.8359878965810084e-05, "loss": 2.1852, "step": 8494 }, { "epoch": 0.21, "learning_rate": 1.8359440856655887e-05, "loss": 2.1814, "step": 8495 }, { "epoch": 0.21, "learning_rate": 1.8359002694224247e-05, "loss": 2.3932, "step": 8496 }, { "epoch": 0.21, "learning_rate": 1.835856447851796e-05, "loss": 2.1853, "step": 8497 }, { "epoch": 0.21, "learning_rate": 1.8358126209539818e-05, "loss": 2.1336, "step": 8498 }, { "epoch": 0.21, "learning_rate": 1.8357687887292613e-05, "loss": 2.2964, "step": 8499 }, { "epoch": 0.21, "learning_rate": 1.8357249511779138e-05, "loss": 2.1551, "step": 8500 }, { "epoch": 0.21, "learning_rate": 1.8356811083002192e-05, "loss": 2.2981, "step": 8501 }, { "epoch": 0.21, "learning_rate": 1.8356372600964563e-05, "loss": 2.3667, "step": 8502 }, { "epoch": 0.21, "learning_rate": 1.835593406566905e-05, "loss": 2.2117, "step": 8503 }, { "epoch": 0.21, "learning_rate": 1.8355495477118443e-05, "loss": 2.2155, "step": 8504 }, { "epoch": 0.21, "learning_rate": 1.835505683531554e-05, "loss": 2.2823, "step": 8505 }, { "epoch": 0.21, "learning_rate": 1.835461814026314e-05, "loss": 2.3159, "step": 8506 }, { "epoch": 0.21, "learning_rate": 1.8354179391964036e-05, "loss": 2.0874, "step": 8507 }, { "epoch": 0.21, "learning_rate": 1.8353740590421022e-05, "loss": 2.2421, "step": 8508 }, { "epoch": 0.21, "learning_rate": 1.83533017356369e-05, "loss": 2.0555, "step": 8509 }, { "epoch": 0.21, "learning_rate": 1.835286282761446e-05, "loss": 2.3205, "step": 8510 }, { "epoch": 0.21, "learning_rate": 1.8352423866356507e-05, "loss": 2.0619, "step": 8511 }, { "epoch": 0.21, "learning_rate": 1.8351984851865832e-05, "loss": 2.1802, "step": 8512 }, { "epoch": 0.21, "learning_rate": 1.835154578414524e-05, "loss": 2.1413, "step": 8513 }, { "epoch": 0.21, "learning_rate": 1.835110666319752e-05, "loss": 2.0315, "step": 8514 }, { "epoch": 0.21, "learning_rate": 1.8350667489025473e-05, "loss": 2.3845, "step": 8515 }, { "epoch": 0.21, "learning_rate": 1.8350228261631905e-05, "loss": 2.0881, "step": 8516 }, { "epoch": 0.21, "learning_rate": 1.834978898101961e-05, "loss": 2.2645, "step": 8517 }, { "epoch": 0.21, "learning_rate": 1.834934964719139e-05, "loss": 2.239, "step": 8518 }, { "epoch": 0.21, "learning_rate": 1.8348910260150043e-05, "loss": 1.9076, "step": 8519 }, { "epoch": 0.21, "learning_rate": 1.8348470819898372e-05, "loss": 2.1591, "step": 8520 }, { "epoch": 0.21, "learning_rate": 1.8348031326439172e-05, "loss": 2.178, "step": 8521 }, { "epoch": 0.21, "learning_rate": 1.834759177977525e-05, "loss": 2.257, "step": 8522 }, { "epoch": 0.21, "learning_rate": 1.8347152179909403e-05, "loss": 2.2986, "step": 8523 }, { "epoch": 0.21, "learning_rate": 1.834671252684444e-05, "loss": 2.1833, "step": 8524 }, { "epoch": 0.21, "learning_rate": 1.8346272820583152e-05, "loss": 2.1617, "step": 8525 }, { "epoch": 0.21, "learning_rate": 1.8345833061128352e-05, "loss": 2.0852, "step": 8526 }, { "epoch": 0.21, "learning_rate": 1.8345393248482834e-05, "loss": 2.0938, "step": 8527 }, { "epoch": 0.21, "learning_rate": 1.834495338264941e-05, "loss": 2.2636, "step": 8528 }, { "epoch": 0.21, "learning_rate": 1.8344513463630876e-05, "loss": 2.1517, "step": 8529 }, { "epoch": 0.21, "learning_rate": 1.8344073491430036e-05, "loss": 2.2622, "step": 8530 }, { "epoch": 0.21, "learning_rate": 1.83436334660497e-05, "loss": 2.0156, "step": 8531 }, { "epoch": 0.21, "learning_rate": 1.834319338749267e-05, "loss": 2.1975, "step": 8532 }, { "epoch": 0.21, "learning_rate": 1.8342753255761746e-05, "loss": 2.3429, "step": 8533 }, { "epoch": 0.21, "learning_rate": 1.8342313070859735e-05, "loss": 2.2173, "step": 8534 }, { "epoch": 0.21, "learning_rate": 1.8341872832789446e-05, "loss": 2.2853, "step": 8535 }, { "epoch": 0.21, "learning_rate": 1.8341432541553684e-05, "loss": 2.1855, "step": 8536 }, { "epoch": 0.21, "learning_rate": 1.8340992197155258e-05, "loss": 2.319, "step": 8537 }, { "epoch": 0.21, "learning_rate": 1.8340551799596965e-05, "loss": 2.2428, "step": 8538 }, { "epoch": 0.21, "learning_rate": 1.8340111348881617e-05, "loss": 2.0, "step": 8539 }, { "epoch": 0.21, "learning_rate": 1.833967084501202e-05, "loss": 2.2385, "step": 8540 }, { "epoch": 0.21, "learning_rate": 1.8339230287990986e-05, "loss": 2.2062, "step": 8541 }, { "epoch": 0.21, "learning_rate": 1.8338789677821322e-05, "loss": 2.2343, "step": 8542 }, { "epoch": 0.21, "learning_rate": 1.8338349014505833e-05, "loss": 2.1652, "step": 8543 }, { "epoch": 0.21, "learning_rate": 1.8337908298047324e-05, "loss": 2.2392, "step": 8544 }, { "epoch": 0.21, "learning_rate": 1.833746752844861e-05, "loss": 2.0938, "step": 8545 }, { "epoch": 0.21, "learning_rate": 1.83370267057125e-05, "loss": 2.1252, "step": 8546 }, { "epoch": 0.21, "learning_rate": 1.83365858298418e-05, "loss": 2.131, "step": 8547 }, { "epoch": 0.21, "learning_rate": 1.8336144900839322e-05, "loss": 2.3884, "step": 8548 }, { "epoch": 0.21, "learning_rate": 1.833570391870788e-05, "loss": 2.3556, "step": 8549 }, { "epoch": 0.21, "learning_rate": 1.8335262883450275e-05, "loss": 2.2044, "step": 8550 }, { "epoch": 0.21, "learning_rate": 1.8334821795069326e-05, "loss": 2.1419, "step": 8551 }, { "epoch": 0.21, "learning_rate": 1.8334380653567838e-05, "loss": 2.0928, "step": 8552 }, { "epoch": 0.21, "learning_rate": 1.8333939458948632e-05, "loss": 2.226, "step": 8553 }, { "epoch": 0.21, "learning_rate": 1.833349821121451e-05, "loss": 2.1479, "step": 8554 }, { "epoch": 0.21, "learning_rate": 1.8333056910368286e-05, "loss": 2.238, "step": 8555 }, { "epoch": 0.21, "learning_rate": 1.8332615556412778e-05, "loss": 2.2926, "step": 8556 }, { "epoch": 0.21, "learning_rate": 1.8332174149350797e-05, "loss": 2.2969, "step": 8557 }, { "epoch": 0.21, "learning_rate": 1.8331732689185153e-05, "loss": 2.2647, "step": 8558 }, { "epoch": 0.21, "learning_rate": 1.8331291175918657e-05, "loss": 2.1669, "step": 8559 }, { "epoch": 0.21, "learning_rate": 1.8330849609554133e-05, "loss": 1.9812, "step": 8560 }, { "epoch": 0.21, "learning_rate": 1.833040799009439e-05, "loss": 2.2608, "step": 8561 }, { "epoch": 0.21, "learning_rate": 1.832996631754224e-05, "loss": 2.2106, "step": 8562 }, { "epoch": 0.21, "learning_rate": 1.8329524591900498e-05, "loss": 2.0247, "step": 8563 }, { "epoch": 0.21, "learning_rate": 1.832908281317198e-05, "loss": 2.2666, "step": 8564 }, { "epoch": 0.21, "learning_rate": 1.832864098135951e-05, "loss": 2.3092, "step": 8565 }, { "epoch": 0.21, "learning_rate": 1.832819909646589e-05, "loss": 2.0417, "step": 8566 }, { "epoch": 0.21, "learning_rate": 1.8327757158493945e-05, "loss": 2.186, "step": 8567 }, { "epoch": 0.21, "learning_rate": 1.8327315167446492e-05, "loss": 2.113, "step": 8568 }, { "epoch": 0.21, "learning_rate": 1.8326873123326342e-05, "loss": 2.2989, "step": 8569 }, { "epoch": 0.21, "learning_rate": 1.832643102613632e-05, "loss": 2.169, "step": 8570 }, { "epoch": 0.21, "learning_rate": 1.8325988875879238e-05, "loss": 2.155, "step": 8571 }, { "epoch": 0.21, "learning_rate": 1.8325546672557914e-05, "loss": 2.1524, "step": 8572 }, { "epoch": 0.21, "learning_rate": 1.832510441617517e-05, "loss": 2.0183, "step": 8573 }, { "epoch": 0.21, "learning_rate": 1.8324662106733822e-05, "loss": 2.156, "step": 8574 }, { "epoch": 0.21, "learning_rate": 1.8324219744236685e-05, "loss": 2.2371, "step": 8575 }, { "epoch": 0.21, "learning_rate": 1.832377732868659e-05, "loss": 2.3917, "step": 8576 }, { "epoch": 0.21, "learning_rate": 1.8323334860086346e-05, "loss": 2.1562, "step": 8577 }, { "epoch": 0.21, "learning_rate": 1.8322892338438775e-05, "loss": 2.2337, "step": 8578 }, { "epoch": 0.21, "learning_rate": 1.83224497637467e-05, "loss": 2.1545, "step": 8579 }, { "epoch": 0.21, "learning_rate": 1.8322007136012943e-05, "loss": 2.288, "step": 8580 }, { "epoch": 0.21, "learning_rate": 1.832156445524032e-05, "loss": 2.1686, "step": 8581 }, { "epoch": 0.21, "learning_rate": 1.8321121721431653e-05, "loss": 2.3411, "step": 8582 }, { "epoch": 0.21, "learning_rate": 1.832067893458977e-05, "loss": 2.1483, "step": 8583 }, { "epoch": 0.21, "learning_rate": 1.8320236094717484e-05, "loss": 2.1308, "step": 8584 }, { "epoch": 0.21, "learning_rate": 1.8319793201817625e-05, "loss": 2.0686, "step": 8585 }, { "epoch": 0.21, "learning_rate": 1.831935025589301e-05, "loss": 2.2066, "step": 8586 }, { "epoch": 0.21, "learning_rate": 1.8318907256946466e-05, "loss": 2.0871, "step": 8587 }, { "epoch": 0.21, "learning_rate": 1.8318464204980817e-05, "loss": 2.2908, "step": 8588 }, { "epoch": 0.21, "learning_rate": 1.831802109999888e-05, "loss": 2.0986, "step": 8589 }, { "epoch": 0.21, "learning_rate": 1.8317577942003487e-05, "loss": 1.996, "step": 8590 }, { "epoch": 0.21, "learning_rate": 1.8317134730997457e-05, "loss": 2.0701, "step": 8591 }, { "epoch": 0.21, "learning_rate": 1.8316691466983622e-05, "loss": 2.0857, "step": 8592 }, { "epoch": 0.21, "learning_rate": 1.83162481499648e-05, "loss": 2.1114, "step": 8593 }, { "epoch": 0.21, "learning_rate": 1.8315804779943814e-05, "loss": 2.1748, "step": 8594 }, { "epoch": 0.21, "learning_rate": 1.8315361356923496e-05, "loss": 2.2924, "step": 8595 }, { "epoch": 0.21, "learning_rate": 1.8314917880906672e-05, "loss": 2.1725, "step": 8596 }, { "epoch": 0.21, "learning_rate": 1.8314474351896166e-05, "loss": 2.2838, "step": 8597 }, { "epoch": 0.21, "learning_rate": 1.8314030769894805e-05, "loss": 2.124, "step": 8598 }, { "epoch": 0.21, "learning_rate": 1.8313587134905418e-05, "loss": 2.2727, "step": 8599 }, { "epoch": 0.21, "learning_rate": 1.831314344693083e-05, "loss": 2.2463, "step": 8600 }, { "epoch": 0.21, "learning_rate": 1.831269970597387e-05, "loss": 2.1477, "step": 8601 }, { "epoch": 0.21, "learning_rate": 1.8312255912037364e-05, "loss": 2.3184, "step": 8602 }, { "epoch": 0.21, "learning_rate": 1.8311812065124144e-05, "loss": 2.3118, "step": 8603 }, { "epoch": 0.21, "learning_rate": 1.8311368165237034e-05, "loss": 2.2043, "step": 8604 }, { "epoch": 0.21, "learning_rate": 1.8310924212378868e-05, "loss": 2.2072, "step": 8605 }, { "epoch": 0.21, "learning_rate": 1.8310480206552473e-05, "loss": 2.1418, "step": 8606 }, { "epoch": 0.21, "learning_rate": 1.831003614776068e-05, "loss": 2.095, "step": 8607 }, { "epoch": 0.21, "learning_rate": 1.830959203600632e-05, "loss": 2.146, "step": 8608 }, { "epoch": 0.21, "learning_rate": 1.8309147871292217e-05, "loss": 2.0902, "step": 8609 }, { "epoch": 0.21, "learning_rate": 1.830870365362121e-05, "loss": 2.2485, "step": 8610 }, { "epoch": 0.21, "learning_rate": 1.830825938299613e-05, "loss": 2.159, "step": 8611 }, { "epoch": 0.21, "learning_rate": 1.8307815059419804e-05, "loss": 2.3192, "step": 8612 }, { "epoch": 0.21, "learning_rate": 1.8307370682895064e-05, "loss": 2.2759, "step": 8613 }, { "epoch": 0.21, "learning_rate": 1.830692625342474e-05, "loss": 2.1005, "step": 8614 }, { "epoch": 0.21, "learning_rate": 1.8306481771011672e-05, "loss": 2.1183, "step": 8615 }, { "epoch": 0.21, "learning_rate": 1.8306037235658686e-05, "loss": 2.1584, "step": 8616 }, { "epoch": 0.21, "learning_rate": 1.830559264736862e-05, "loss": 2.1876, "step": 8617 }, { "epoch": 0.21, "learning_rate": 1.8305148006144306e-05, "loss": 2.2705, "step": 8618 }, { "epoch": 0.21, "learning_rate": 1.8304703311988573e-05, "loss": 2.1348, "step": 8619 }, { "epoch": 0.21, "learning_rate": 1.8304258564904262e-05, "loss": 2.2629, "step": 8620 }, { "epoch": 0.21, "learning_rate": 1.830381376489421e-05, "loss": 2.1053, "step": 8621 }, { "epoch": 0.21, "learning_rate": 1.8303368911961235e-05, "loss": 2.018, "step": 8622 }, { "epoch": 0.21, "learning_rate": 1.8302924006108195e-05, "loss": 2.2143, "step": 8623 }, { "epoch": 0.21, "learning_rate": 1.830247904733791e-05, "loss": 2.1987, "step": 8624 }, { "epoch": 0.21, "learning_rate": 1.8302034035653218e-05, "loss": 2.1556, "step": 8625 }, { "epoch": 0.21, "learning_rate": 1.830158897105696e-05, "loss": 2.1681, "step": 8626 }, { "epoch": 0.21, "learning_rate": 1.830114385355197e-05, "loss": 2.0829, "step": 8627 }, { "epoch": 0.21, "learning_rate": 1.8300698683141088e-05, "loss": 2.0785, "step": 8628 }, { "epoch": 0.21, "learning_rate": 1.8300253459827143e-05, "loss": 2.1114, "step": 8629 }, { "epoch": 0.21, "learning_rate": 1.8299808183612978e-05, "loss": 2.0504, "step": 8630 }, { "epoch": 0.21, "learning_rate": 1.8299362854501436e-05, "loss": 2.1361, "step": 8631 }, { "epoch": 0.21, "learning_rate": 1.8298917472495346e-05, "loss": 2.2495, "step": 8632 }, { "epoch": 0.21, "learning_rate": 1.829847203759755e-05, "loss": 2.1872, "step": 8633 }, { "epoch": 0.21, "learning_rate": 1.829802654981089e-05, "loss": 2.2649, "step": 8634 }, { "epoch": 0.21, "learning_rate": 1.82975810091382e-05, "loss": 2.2021, "step": 8635 }, { "epoch": 0.21, "learning_rate": 1.829713541558232e-05, "loss": 2.1077, "step": 8636 }, { "epoch": 0.21, "learning_rate": 1.8296689769146096e-05, "loss": 2.1664, "step": 8637 }, { "epoch": 0.21, "learning_rate": 1.8296244069832364e-05, "loss": 2.136, "step": 8638 }, { "epoch": 0.21, "learning_rate": 1.8295798317643965e-05, "loss": 2.2278, "step": 8639 }, { "epoch": 0.21, "learning_rate": 1.8295352512583736e-05, "loss": 2.1686, "step": 8640 }, { "epoch": 0.21, "learning_rate": 1.8294906654654526e-05, "loss": 2.4001, "step": 8641 }, { "epoch": 0.21, "learning_rate": 1.8294460743859174e-05, "loss": 2.2817, "step": 8642 }, { "epoch": 0.21, "learning_rate": 1.8294014780200516e-05, "loss": 2.0684, "step": 8643 }, { "epoch": 0.21, "learning_rate": 1.82935687636814e-05, "loss": 2.2574, "step": 8644 }, { "epoch": 0.21, "learning_rate": 1.8293122694304667e-05, "loss": 2.1819, "step": 8645 }, { "epoch": 0.21, "learning_rate": 1.8292676572073164e-05, "loss": 2.2448, "step": 8646 }, { "epoch": 0.21, "learning_rate": 1.829223039698973e-05, "loss": 2.1162, "step": 8647 }, { "epoch": 0.21, "learning_rate": 1.8291784169057208e-05, "loss": 2.2393, "step": 8648 }, { "epoch": 0.21, "learning_rate": 1.829133788827844e-05, "loss": 2.2829, "step": 8649 }, { "epoch": 0.21, "learning_rate": 1.829089155465628e-05, "loss": 2.3385, "step": 8650 }, { "epoch": 0.21, "learning_rate": 1.829044516819356e-05, "loss": 2.3275, "step": 8651 }, { "epoch": 0.21, "learning_rate": 1.8289998728893136e-05, "loss": 2.2884, "step": 8652 }, { "epoch": 0.21, "learning_rate": 1.8289552236757846e-05, "loss": 2.1546, "step": 8653 }, { "epoch": 0.21, "learning_rate": 1.828910569179054e-05, "loss": 2.1668, "step": 8654 }, { "epoch": 0.21, "learning_rate": 1.828865909399406e-05, "loss": 2.2898, "step": 8655 }, { "epoch": 0.21, "learning_rate": 1.8288212443371253e-05, "loss": 2.1181, "step": 8656 }, { "epoch": 0.21, "learning_rate": 1.8287765739924973e-05, "loss": 2.1781, "step": 8657 }, { "epoch": 0.21, "learning_rate": 1.8287318983658057e-05, "loss": 2.2337, "step": 8658 }, { "epoch": 0.21, "learning_rate": 1.8286872174573354e-05, "loss": 2.1647, "step": 8659 }, { "epoch": 0.21, "learning_rate": 1.828642531267372e-05, "loss": 2.3274, "step": 8660 }, { "epoch": 0.21, "learning_rate": 1.828597839796199e-05, "loss": 2.3124, "step": 8661 }, { "epoch": 0.21, "learning_rate": 1.8285531430441023e-05, "loss": 2.0064, "step": 8662 }, { "epoch": 0.21, "learning_rate": 1.8285084410113666e-05, "loss": 2.1364, "step": 8663 }, { "epoch": 0.21, "learning_rate": 1.8284637336982763e-05, "loss": 2.0659, "step": 8664 }, { "epoch": 0.21, "learning_rate": 1.8284190211051167e-05, "loss": 2.217, "step": 8665 }, { "epoch": 0.21, "learning_rate": 1.8283743032321727e-05, "loss": 2.2921, "step": 8666 }, { "epoch": 0.21, "learning_rate": 1.8283295800797293e-05, "loss": 2.3252, "step": 8667 }, { "epoch": 0.21, "learning_rate": 1.8282848516480717e-05, "loss": 2.1054, "step": 8668 }, { "epoch": 0.21, "learning_rate": 1.8282401179374848e-05, "loss": 2.1759, "step": 8669 }, { "epoch": 0.21, "learning_rate": 1.8281953789482534e-05, "loss": 2.1399, "step": 8670 }, { "epoch": 0.21, "learning_rate": 1.8281506346806633e-05, "loss": 2.2111, "step": 8671 }, { "epoch": 0.21, "learning_rate": 1.828105885134999e-05, "loss": 2.0421, "step": 8672 }, { "epoch": 0.21, "learning_rate": 1.8280611303115462e-05, "loss": 2.1913, "step": 8673 }, { "epoch": 0.21, "learning_rate": 1.8280163702105902e-05, "loss": 2.2419, "step": 8674 }, { "epoch": 0.21, "learning_rate": 1.8279716048324156e-05, "loss": 2.1879, "step": 8675 }, { "epoch": 0.21, "learning_rate": 1.8279268341773083e-05, "loss": 2.1808, "step": 8676 }, { "epoch": 0.21, "learning_rate": 1.8278820582455535e-05, "loss": 2.2065, "step": 8677 }, { "epoch": 0.21, "learning_rate": 1.827837277037437e-05, "loss": 2.254, "step": 8678 }, { "epoch": 0.21, "learning_rate": 1.827792490553243e-05, "loss": 2.3993, "step": 8679 }, { "epoch": 0.21, "learning_rate": 1.827747698793258e-05, "loss": 2.3182, "step": 8680 }, { "epoch": 0.21, "learning_rate": 1.8277029017577672e-05, "loss": 2.2243, "step": 8681 }, { "epoch": 0.21, "learning_rate": 1.8276580994470557e-05, "loss": 2.2253, "step": 8682 }, { "epoch": 0.21, "learning_rate": 1.82761329186141e-05, "loss": 2.3306, "step": 8683 }, { "epoch": 0.21, "learning_rate": 1.8275684790011148e-05, "loss": 2.1716, "step": 8684 }, { "epoch": 0.21, "learning_rate": 1.8275236608664556e-05, "loss": 2.204, "step": 8685 }, { "epoch": 0.21, "learning_rate": 1.827478837457719e-05, "loss": 2.1435, "step": 8686 }, { "epoch": 0.21, "learning_rate": 1.82743400877519e-05, "loss": 2.1651, "step": 8687 }, { "epoch": 0.21, "learning_rate": 1.827389174819154e-05, "loss": 2.0936, "step": 8688 }, { "epoch": 0.21, "learning_rate": 1.827344335589898e-05, "loss": 2.3965, "step": 8689 }, { "epoch": 0.21, "learning_rate": 1.8272994910877063e-05, "loss": 2.2357, "step": 8690 }, { "epoch": 0.21, "learning_rate": 1.8272546413128652e-05, "loss": 2.0748, "step": 8691 }, { "epoch": 0.21, "learning_rate": 1.827209786265661e-05, "loss": 1.9853, "step": 8692 }, { "epoch": 0.21, "learning_rate": 1.827164925946379e-05, "loss": 2.1835, "step": 8693 }, { "epoch": 0.21, "learning_rate": 1.8271200603553058e-05, "loss": 2.2685, "step": 8694 }, { "epoch": 0.21, "learning_rate": 1.8270751894927264e-05, "loss": 2.1852, "step": 8695 }, { "epoch": 0.21, "learning_rate": 1.8270303133589275e-05, "loss": 2.3622, "step": 8696 }, { "epoch": 0.21, "learning_rate": 1.826985431954195e-05, "loss": 2.2416, "step": 8697 }, { "epoch": 0.21, "learning_rate": 1.826940545278815e-05, "loss": 2.2402, "step": 8698 }, { "epoch": 0.21, "learning_rate": 1.8268956533330728e-05, "loss": 2.2749, "step": 8699 }, { "epoch": 0.21, "learning_rate": 1.8268507561172554e-05, "loss": 2.2502, "step": 8700 }, { "epoch": 0.21, "learning_rate": 1.826805853631649e-05, "loss": 2.2298, "step": 8701 }, { "epoch": 0.21, "learning_rate": 1.826760945876539e-05, "loss": 2.1843, "step": 8702 }, { "epoch": 0.21, "learning_rate": 1.8267160328522122e-05, "loss": 2.1037, "step": 8703 }, { "epoch": 0.21, "learning_rate": 1.8266711145589547e-05, "loss": 2.1378, "step": 8704 }, { "epoch": 0.21, "learning_rate": 1.826626190997053e-05, "loss": 2.327, "step": 8705 }, { "epoch": 0.21, "learning_rate": 1.826581262166793e-05, "loss": 2.2126, "step": 8706 }, { "epoch": 0.21, "learning_rate": 1.8265363280684613e-05, "loss": 2.1601, "step": 8707 }, { "epoch": 0.21, "learning_rate": 1.826491388702344e-05, "loss": 2.1691, "step": 8708 }, { "epoch": 0.21, "learning_rate": 1.826446444068728e-05, "loss": 2.1529, "step": 8709 }, { "epoch": 0.21, "learning_rate": 1.8264014941678994e-05, "loss": 2.1839, "step": 8710 }, { "epoch": 0.21, "learning_rate": 1.8263565390001447e-05, "loss": 2.2443, "step": 8711 }, { "epoch": 0.21, "learning_rate": 1.8263115785657505e-05, "loss": 2.3528, "step": 8712 }, { "epoch": 0.21, "learning_rate": 1.826266612865003e-05, "loss": 2.2127, "step": 8713 }, { "epoch": 0.21, "learning_rate": 1.8262216418981896e-05, "loss": 2.2007, "step": 8714 }, { "epoch": 0.21, "learning_rate": 1.8261766656655962e-05, "loss": 2.1356, "step": 8715 }, { "epoch": 0.21, "learning_rate": 1.82613168416751e-05, "loss": 2.1928, "step": 8716 }, { "epoch": 0.21, "learning_rate": 1.8260866974042174e-05, "loss": 2.1431, "step": 8717 }, { "epoch": 0.21, "learning_rate": 1.8260417053760045e-05, "loss": 2.2859, "step": 8718 }, { "epoch": 0.21, "learning_rate": 1.825996708083159e-05, "loss": 2.2029, "step": 8719 }, { "epoch": 0.21, "learning_rate": 1.825951705525967e-05, "loss": 2.2827, "step": 8720 }, { "epoch": 0.21, "learning_rate": 1.8259066977047158e-05, "loss": 2.1026, "step": 8721 }, { "epoch": 0.21, "learning_rate": 1.8258616846196918e-05, "loss": 2.2385, "step": 8722 }, { "epoch": 0.22, "learning_rate": 1.8258166662711823e-05, "loss": 2.0091, "step": 8723 }, { "epoch": 0.22, "learning_rate": 1.8257716426594744e-05, "loss": 2.0887, "step": 8724 }, { "epoch": 0.22, "learning_rate": 1.8257266137848547e-05, "loss": 2.2953, "step": 8725 }, { "epoch": 0.22, "learning_rate": 1.8256815796476097e-05, "loss": 2.0987, "step": 8726 }, { "epoch": 0.22, "learning_rate": 1.8256365402480273e-05, "loss": 2.3392, "step": 8727 }, { "epoch": 0.22, "learning_rate": 1.825591495586394e-05, "loss": 2.2615, "step": 8728 }, { "epoch": 0.22, "learning_rate": 1.825546445662997e-05, "loss": 2.2546, "step": 8729 }, { "epoch": 0.22, "learning_rate": 1.8255013904781237e-05, "loss": 2.2537, "step": 8730 }, { "epoch": 0.22, "learning_rate": 1.8254563300320605e-05, "loss": 2.1285, "step": 8731 }, { "epoch": 0.22, "learning_rate": 1.8254112643250957e-05, "loss": 2.132, "step": 8732 }, { "epoch": 0.22, "learning_rate": 1.8253661933575155e-05, "loss": 2.3738, "step": 8733 }, { "epoch": 0.22, "learning_rate": 1.8253211171296076e-05, "loss": 2.1848, "step": 8734 }, { "epoch": 0.22, "learning_rate": 1.8252760356416597e-05, "loss": 2.1085, "step": 8735 }, { "epoch": 0.22, "learning_rate": 1.825230948893958e-05, "loss": 2.2159, "step": 8736 }, { "epoch": 0.22, "learning_rate": 1.825185856886791e-05, "loss": 2.1851, "step": 8737 }, { "epoch": 0.22, "learning_rate": 1.8251407596204457e-05, "loss": 2.2006, "step": 8738 }, { "epoch": 0.22, "learning_rate": 1.825095657095209e-05, "loss": 2.1073, "step": 8739 }, { "epoch": 0.22, "learning_rate": 1.825050549311369e-05, "loss": 2.1703, "step": 8740 }, { "epoch": 0.22, "learning_rate": 1.825005436269213e-05, "loss": 2.1569, "step": 8741 }, { "epoch": 0.22, "learning_rate": 1.8249603179690285e-05, "loss": 2.1532, "step": 8742 }, { "epoch": 0.22, "learning_rate": 1.824915194411103e-05, "loss": 2.3432, "step": 8743 }, { "epoch": 0.22, "learning_rate": 1.8248700655957243e-05, "loss": 2.2222, "step": 8744 }, { "epoch": 0.22, "learning_rate": 1.8248249315231795e-05, "loss": 2.2509, "step": 8745 }, { "epoch": 0.22, "learning_rate": 1.8247797921937572e-05, "loss": 2.1459, "step": 8746 }, { "epoch": 0.22, "learning_rate": 1.824734647607744e-05, "loss": 2.1083, "step": 8747 }, { "epoch": 0.22, "learning_rate": 1.8246894977654284e-05, "loss": 2.1367, "step": 8748 }, { "epoch": 0.22, "learning_rate": 1.8246443426670976e-05, "loss": 2.2323, "step": 8749 }, { "epoch": 0.22, "learning_rate": 1.82459918231304e-05, "loss": 2.0548, "step": 8750 }, { "epoch": 0.22, "learning_rate": 1.824554016703543e-05, "loss": 2.3093, "step": 8751 }, { "epoch": 0.22, "learning_rate": 1.824508845838894e-05, "loss": 2.3531, "step": 8752 }, { "epoch": 0.22, "learning_rate": 1.8244636697193818e-05, "loss": 2.1906, "step": 8753 }, { "epoch": 0.22, "learning_rate": 1.824418488345294e-05, "loss": 2.1118, "step": 8754 }, { "epoch": 0.22, "learning_rate": 1.824373301716918e-05, "loss": 2.1544, "step": 8755 }, { "epoch": 0.22, "learning_rate": 1.8243281098345435e-05, "loss": 2.172, "step": 8756 }, { "epoch": 0.22, "learning_rate": 1.8242829126984564e-05, "loss": 2.1615, "step": 8757 }, { "epoch": 0.22, "learning_rate": 1.8242377103089457e-05, "loss": 2.1372, "step": 8758 }, { "epoch": 0.22, "learning_rate": 1.8241925026662995e-05, "loss": 2.3264, "step": 8759 }, { "epoch": 0.22, "learning_rate": 1.824147289770806e-05, "loss": 2.1375, "step": 8760 }, { "epoch": 0.22, "learning_rate": 1.8241020716227535e-05, "loss": 2.1953, "step": 8761 }, { "epoch": 0.22, "learning_rate": 1.8240568482224295e-05, "loss": 2.2087, "step": 8762 }, { "epoch": 0.22, "learning_rate": 1.824011619570123e-05, "loss": 2.2418, "step": 8763 }, { "epoch": 0.22, "learning_rate": 1.8239663856661214e-05, "loss": 2.2326, "step": 8764 }, { "epoch": 0.22, "learning_rate": 1.823921146510714e-05, "loss": 2.0955, "step": 8765 }, { "epoch": 0.22, "learning_rate": 1.8238759021041884e-05, "loss": 2.1638, "step": 8766 }, { "epoch": 0.22, "learning_rate": 1.823830652446833e-05, "loss": 2.0208, "step": 8767 }, { "epoch": 0.22, "learning_rate": 1.823785397538937e-05, "loss": 2.2903, "step": 8768 }, { "epoch": 0.22, "learning_rate": 1.8237401373807876e-05, "loss": 2.0977, "step": 8769 }, { "epoch": 0.22, "learning_rate": 1.8236948719726744e-05, "loss": 2.2386, "step": 8770 }, { "epoch": 0.22, "learning_rate": 1.8236496013148848e-05, "loss": 2.3024, "step": 8771 }, { "epoch": 0.22, "learning_rate": 1.823604325407708e-05, "loss": 2.1514, "step": 8772 }, { "epoch": 0.22, "learning_rate": 1.823559044251433e-05, "loss": 2.338, "step": 8773 }, { "epoch": 0.22, "learning_rate": 1.823513757846347e-05, "loss": 2.058, "step": 8774 }, { "epoch": 0.22, "learning_rate": 1.8234684661927398e-05, "loss": 2.1535, "step": 8775 }, { "epoch": 0.22, "learning_rate": 1.8234231692908998e-05, "loss": 2.3653, "step": 8776 }, { "epoch": 0.22, "learning_rate": 1.8233778671411153e-05, "loss": 2.1974, "step": 8777 }, { "epoch": 0.22, "learning_rate": 1.8233325597436758e-05, "loss": 2.1891, "step": 8778 }, { "epoch": 0.22, "learning_rate": 1.8232872470988692e-05, "loss": 2.1233, "step": 8779 }, { "epoch": 0.22, "learning_rate": 1.8232419292069846e-05, "loss": 2.264, "step": 8780 }, { "epoch": 0.22, "learning_rate": 1.8231966060683112e-05, "loss": 2.383, "step": 8781 }, { "epoch": 0.22, "learning_rate": 1.8231512776831374e-05, "loss": 2.1187, "step": 8782 }, { "epoch": 0.22, "learning_rate": 1.823105944051752e-05, "loss": 2.1922, "step": 8783 }, { "epoch": 0.22, "learning_rate": 1.8230606051744444e-05, "loss": 2.2516, "step": 8784 }, { "epoch": 0.22, "learning_rate": 1.8230152610515035e-05, "loss": 2.3007, "step": 8785 }, { "epoch": 0.22, "learning_rate": 1.822969911683218e-05, "loss": 2.236, "step": 8786 }, { "epoch": 0.22, "learning_rate": 1.822924557069877e-05, "loss": 2.2301, "step": 8787 }, { "epoch": 0.22, "learning_rate": 1.8228791972117697e-05, "loss": 2.3634, "step": 8788 }, { "epoch": 0.22, "learning_rate": 1.822833832109185e-05, "loss": 2.0968, "step": 8789 }, { "epoch": 0.22, "learning_rate": 1.8227884617624124e-05, "loss": 2.361, "step": 8790 }, { "epoch": 0.22, "learning_rate": 1.8227430861717405e-05, "loss": 2.0187, "step": 8791 }, { "epoch": 0.22, "learning_rate": 1.822697705337459e-05, "loss": 2.3228, "step": 8792 }, { "epoch": 0.22, "learning_rate": 1.8226523192598567e-05, "loss": 2.1378, "step": 8793 }, { "epoch": 0.22, "learning_rate": 1.8226069279392232e-05, "loss": 2.3788, "step": 8794 }, { "epoch": 0.22, "learning_rate": 1.8225615313758478e-05, "loss": 2.174, "step": 8795 }, { "epoch": 0.22, "learning_rate": 1.8225161295700194e-05, "loss": 2.1993, "step": 8796 }, { "epoch": 0.22, "learning_rate": 1.8224707225220278e-05, "loss": 2.2868, "step": 8797 }, { "epoch": 0.22, "learning_rate": 1.8224253102321623e-05, "loss": 2.0842, "step": 8798 }, { "epoch": 0.22, "learning_rate": 1.8223798927007123e-05, "loss": 2.1799, "step": 8799 }, { "epoch": 0.22, "learning_rate": 1.822334469927967e-05, "loss": 2.3344, "step": 8800 }, { "epoch": 0.22, "learning_rate": 1.8222890419142162e-05, "loss": 2.2219, "step": 8801 }, { "epoch": 0.22, "learning_rate": 1.8222436086597496e-05, "loss": 2.1715, "step": 8802 }, { "epoch": 0.22, "learning_rate": 1.8221981701648562e-05, "loss": 2.1194, "step": 8803 }, { "epoch": 0.22, "learning_rate": 1.8221527264298263e-05, "loss": 2.2381, "step": 8804 }, { "epoch": 0.22, "learning_rate": 1.822107277454949e-05, "loss": 2.1721, "step": 8805 }, { "epoch": 0.22, "learning_rate": 1.8220618232405138e-05, "loss": 2.1988, "step": 8806 }, { "epoch": 0.22, "learning_rate": 1.8220163637868113e-05, "loss": 2.0223, "step": 8807 }, { "epoch": 0.22, "learning_rate": 1.8219708990941304e-05, "loss": 2.2135, "step": 8808 }, { "epoch": 0.22, "learning_rate": 1.8219254291627607e-05, "loss": 2.0939, "step": 8809 }, { "epoch": 0.22, "learning_rate": 1.8218799539929927e-05, "loss": 2.3631, "step": 8810 }, { "epoch": 0.22, "learning_rate": 1.8218344735851157e-05, "loss": 2.3138, "step": 8811 }, { "epoch": 0.22, "learning_rate": 1.82178898793942e-05, "loss": 2.3018, "step": 8812 }, { "epoch": 0.22, "learning_rate": 1.821743497056195e-05, "loss": 2.1113, "step": 8813 }, { "epoch": 0.22, "learning_rate": 1.8216980009357312e-05, "loss": 2.269, "step": 8814 }, { "epoch": 0.22, "learning_rate": 1.821652499578318e-05, "loss": 2.3273, "step": 8815 }, { "epoch": 0.22, "learning_rate": 1.8216069929842458e-05, "loss": 2.2584, "step": 8816 }, { "epoch": 0.22, "learning_rate": 1.8215614811538046e-05, "loss": 2.1112, "step": 8817 }, { "epoch": 0.22, "learning_rate": 1.8215159640872842e-05, "loss": 2.2562, "step": 8818 }, { "epoch": 0.22, "learning_rate": 1.821470441784975e-05, "loss": 2.3231, "step": 8819 }, { "epoch": 0.22, "learning_rate": 1.8214249142471667e-05, "loss": 2.2577, "step": 8820 }, { "epoch": 0.22, "learning_rate": 1.82137938147415e-05, "loss": 2.0107, "step": 8821 }, { "epoch": 0.22, "learning_rate": 1.8213338434662144e-05, "loss": 2.1436, "step": 8822 }, { "epoch": 0.22, "learning_rate": 1.821288300223651e-05, "loss": 2.2565, "step": 8823 }, { "epoch": 0.22, "learning_rate": 1.8212427517467494e-05, "loss": 2.192, "step": 8824 }, { "epoch": 0.22, "learning_rate": 1.8211971980358002e-05, "loss": 2.0942, "step": 8825 }, { "epoch": 0.22, "learning_rate": 1.8211516390910934e-05, "loss": 2.2766, "step": 8826 }, { "epoch": 0.22, "learning_rate": 1.82110607491292e-05, "loss": 2.1888, "step": 8827 }, { "epoch": 0.22, "learning_rate": 1.8210605055015693e-05, "loss": 2.1398, "step": 8828 }, { "epoch": 0.22, "learning_rate": 1.821014930857333e-05, "loss": 2.1627, "step": 8829 }, { "epoch": 0.22, "learning_rate": 1.820969350980501e-05, "loss": 2.1957, "step": 8830 }, { "epoch": 0.22, "learning_rate": 1.8209237658713636e-05, "loss": 2.0491, "step": 8831 }, { "epoch": 0.22, "learning_rate": 1.8208781755302114e-05, "loss": 2.3413, "step": 8832 }, { "epoch": 0.22, "learning_rate": 1.8208325799573354e-05, "loss": 2.2781, "step": 8833 }, { "epoch": 0.22, "learning_rate": 1.8207869791530257e-05, "loss": 2.0859, "step": 8834 }, { "epoch": 0.22, "learning_rate": 1.820741373117573e-05, "loss": 2.341, "step": 8835 }, { "epoch": 0.22, "learning_rate": 1.8206957618512683e-05, "loss": 2.2077, "step": 8836 }, { "epoch": 0.22, "learning_rate": 1.8206501453544017e-05, "loss": 2.2703, "step": 8837 }, { "epoch": 0.22, "learning_rate": 1.8206045236272646e-05, "loss": 2.3215, "step": 8838 }, { "epoch": 0.22, "learning_rate": 1.8205588966701472e-05, "loss": 2.1603, "step": 8839 }, { "epoch": 0.22, "learning_rate": 1.8205132644833407e-05, "loss": 2.2839, "step": 8840 }, { "epoch": 0.22, "learning_rate": 1.8204676270671357e-05, "loss": 2.3102, "step": 8841 }, { "epoch": 0.22, "learning_rate": 1.8204219844218233e-05, "loss": 2.1466, "step": 8842 }, { "epoch": 0.22, "learning_rate": 1.820376336547694e-05, "loss": 2.2808, "step": 8843 }, { "epoch": 0.22, "learning_rate": 1.820330683445039e-05, "loss": 2.0472, "step": 8844 }, { "epoch": 0.22, "learning_rate": 1.8202850251141492e-05, "loss": 2.2045, "step": 8845 }, { "epoch": 0.22, "learning_rate": 1.820239361555316e-05, "loss": 2.3388, "step": 8846 }, { "epoch": 0.22, "learning_rate": 1.8201936927688296e-05, "loss": 2.1422, "step": 8847 }, { "epoch": 0.22, "learning_rate": 1.8201480187549815e-05, "loss": 2.2824, "step": 8848 }, { "epoch": 0.22, "learning_rate": 1.820102339514063e-05, "loss": 2.3415, "step": 8849 }, { "epoch": 0.22, "learning_rate": 1.820056655046365e-05, "loss": 2.1229, "step": 8850 }, { "epoch": 0.22, "learning_rate": 1.8200109653521786e-05, "loss": 2.1313, "step": 8851 }, { "epoch": 0.22, "learning_rate": 1.8199652704317948e-05, "loss": 2.2994, "step": 8852 }, { "epoch": 0.22, "learning_rate": 1.8199195702855058e-05, "loss": 2.3123, "step": 8853 }, { "epoch": 0.22, "learning_rate": 1.8198738649136015e-05, "loss": 2.1108, "step": 8854 }, { "epoch": 0.22, "learning_rate": 1.8198281543163743e-05, "loss": 2.2487, "step": 8855 }, { "epoch": 0.22, "learning_rate": 1.819782438494115e-05, "loss": 2.1828, "step": 8856 }, { "epoch": 0.22, "learning_rate": 1.819736717447115e-05, "loss": 2.1117, "step": 8857 }, { "epoch": 0.22, "learning_rate": 1.8196909911756656e-05, "loss": 2.2019, "step": 8858 }, { "epoch": 0.22, "learning_rate": 1.8196452596800586e-05, "loss": 2.2057, "step": 8859 }, { "epoch": 0.22, "learning_rate": 1.8195995229605854e-05, "loss": 2.242, "step": 8860 }, { "epoch": 0.22, "learning_rate": 1.819553781017537e-05, "loss": 2.0952, "step": 8861 }, { "epoch": 0.22, "learning_rate": 1.8195080338512055e-05, "loss": 2.3579, "step": 8862 }, { "epoch": 0.22, "learning_rate": 1.819462281461882e-05, "loss": 2.181, "step": 8863 }, { "epoch": 0.22, "learning_rate": 1.8194165238498586e-05, "loss": 2.3052, "step": 8864 }, { "epoch": 0.22, "learning_rate": 1.8193707610154265e-05, "loss": 2.1383, "step": 8865 }, { "epoch": 0.22, "learning_rate": 1.8193249929588774e-05, "loss": 2.3027, "step": 8866 }, { "epoch": 0.22, "learning_rate": 1.8192792196805033e-05, "loss": 2.1122, "step": 8867 }, { "epoch": 0.22, "learning_rate": 1.819233441180596e-05, "loss": 2.2132, "step": 8868 }, { "epoch": 0.22, "learning_rate": 1.8191876574594465e-05, "loss": 2.2162, "step": 8869 }, { "epoch": 0.22, "learning_rate": 1.819141868517347e-05, "loss": 2.2189, "step": 8870 }, { "epoch": 0.22, "learning_rate": 1.8190960743545897e-05, "loss": 2.3367, "step": 8871 }, { "epoch": 0.22, "learning_rate": 1.8190502749714666e-05, "loss": 2.2695, "step": 8872 }, { "epoch": 0.22, "learning_rate": 1.8190044703682684e-05, "loss": 2.1529, "step": 8873 }, { "epoch": 0.22, "learning_rate": 1.8189586605452884e-05, "loss": 2.1104, "step": 8874 }, { "epoch": 0.22, "learning_rate": 1.8189128455028178e-05, "loss": 2.1338, "step": 8875 }, { "epoch": 0.22, "learning_rate": 1.8188670252411484e-05, "loss": 2.3135, "step": 8876 }, { "epoch": 0.22, "learning_rate": 1.818821199760573e-05, "loss": 2.3628, "step": 8877 }, { "epoch": 0.22, "learning_rate": 1.8187753690613833e-05, "loss": 2.1876, "step": 8878 }, { "epoch": 0.22, "learning_rate": 1.818729533143871e-05, "loss": 2.2175, "step": 8879 }, { "epoch": 0.22, "learning_rate": 1.8186836920083286e-05, "loss": 2.2185, "step": 8880 }, { "epoch": 0.22, "learning_rate": 1.8186378456550483e-05, "loss": 2.3445, "step": 8881 }, { "epoch": 0.22, "learning_rate": 1.8185919940843225e-05, "loss": 2.2384, "step": 8882 }, { "epoch": 0.22, "learning_rate": 1.8185461372964426e-05, "loss": 2.2727, "step": 8883 }, { "epoch": 0.22, "learning_rate": 1.818500275291702e-05, "loss": 2.0314, "step": 8884 }, { "epoch": 0.22, "learning_rate": 1.8184544080703922e-05, "loss": 2.4187, "step": 8885 }, { "epoch": 0.22, "learning_rate": 1.8184085356328055e-05, "loss": 2.2555, "step": 8886 }, { "epoch": 0.22, "learning_rate": 1.818362657979235e-05, "loss": 2.1826, "step": 8887 }, { "epoch": 0.22, "learning_rate": 1.8183167751099722e-05, "loss": 2.195, "step": 8888 }, { "epoch": 0.22, "learning_rate": 1.8182708870253098e-05, "loss": 2.1467, "step": 8889 }, { "epoch": 0.22, "learning_rate": 1.8182249937255405e-05, "loss": 2.0928, "step": 8890 }, { "epoch": 0.22, "learning_rate": 1.8181790952109568e-05, "loss": 2.2609, "step": 8891 }, { "epoch": 0.22, "learning_rate": 1.818133191481851e-05, "loss": 2.3232, "step": 8892 }, { "epoch": 0.22, "learning_rate": 1.8180872825385158e-05, "loss": 1.9956, "step": 8893 }, { "epoch": 0.22, "learning_rate": 1.8180413683812434e-05, "loss": 2.148, "step": 8894 }, { "epoch": 0.22, "learning_rate": 1.8179954490103275e-05, "loss": 2.1876, "step": 8895 }, { "epoch": 0.22, "learning_rate": 1.8179495244260593e-05, "loss": 2.1617, "step": 8896 }, { "epoch": 0.22, "learning_rate": 1.8179035946287325e-05, "loss": 2.3554, "step": 8897 }, { "epoch": 0.22, "learning_rate": 1.8178576596186393e-05, "loss": 2.0108, "step": 8898 }, { "epoch": 0.22, "learning_rate": 1.817811719396073e-05, "loss": 2.3272, "step": 8899 }, { "epoch": 0.22, "learning_rate": 1.8177657739613265e-05, "loss": 2.3376, "step": 8900 }, { "epoch": 0.22, "learning_rate": 1.8177198233146917e-05, "loss": 2.0911, "step": 8901 }, { "epoch": 0.22, "learning_rate": 1.8176738674564617e-05, "loss": 2.2801, "step": 8902 }, { "epoch": 0.22, "learning_rate": 1.81762790638693e-05, "loss": 2.1374, "step": 8903 }, { "epoch": 0.22, "learning_rate": 1.8175819401063895e-05, "loss": 2.1014, "step": 8904 }, { "epoch": 0.22, "learning_rate": 1.8175359686151324e-05, "loss": 2.3486, "step": 8905 }, { "epoch": 0.22, "learning_rate": 1.8174899919134525e-05, "loss": 2.1353, "step": 8906 }, { "epoch": 0.22, "learning_rate": 1.817444010001642e-05, "loss": 2.31, "step": 8907 }, { "epoch": 0.22, "learning_rate": 1.8173980228799948e-05, "loss": 2.2119, "step": 8908 }, { "epoch": 0.22, "learning_rate": 1.8173520305488036e-05, "loss": 2.2921, "step": 8909 }, { "epoch": 0.22, "learning_rate": 1.8173060330083614e-05, "loss": 2.1594, "step": 8910 }, { "epoch": 0.22, "learning_rate": 1.8172600302589614e-05, "loss": 2.3509, "step": 8911 }, { "epoch": 0.22, "learning_rate": 1.817214022300897e-05, "loss": 2.1624, "step": 8912 }, { "epoch": 0.22, "learning_rate": 1.8171680091344615e-05, "loss": 2.1979, "step": 8913 }, { "epoch": 0.22, "learning_rate": 1.8171219907599477e-05, "loss": 2.2094, "step": 8914 }, { "epoch": 0.22, "learning_rate": 1.817075967177649e-05, "loss": 2.0762, "step": 8915 }, { "epoch": 0.22, "learning_rate": 1.8170299383878596e-05, "loss": 2.3751, "step": 8916 }, { "epoch": 0.22, "learning_rate": 1.8169839043908718e-05, "loss": 2.011, "step": 8917 }, { "epoch": 0.22, "learning_rate": 1.8169378651869793e-05, "loss": 2.2022, "step": 8918 }, { "epoch": 0.22, "learning_rate": 1.8168918207764755e-05, "loss": 2.2296, "step": 8919 }, { "epoch": 0.22, "learning_rate": 1.8168457711596536e-05, "loss": 2.1025, "step": 8920 }, { "epoch": 0.22, "learning_rate": 1.816799716336808e-05, "loss": 2.0655, "step": 8921 }, { "epoch": 0.22, "learning_rate": 1.816753656308231e-05, "loss": 2.1872, "step": 8922 }, { "epoch": 0.22, "learning_rate": 1.8167075910742174e-05, "loss": 2.372, "step": 8923 }, { "epoch": 0.22, "learning_rate": 1.8166615206350596e-05, "loss": 2.1487, "step": 8924 }, { "epoch": 0.22, "learning_rate": 1.816615444991052e-05, "loss": 2.3746, "step": 8925 }, { "epoch": 0.22, "learning_rate": 1.8165693641424883e-05, "loss": 2.1331, "step": 8926 }, { "epoch": 0.22, "learning_rate": 1.8165232780896616e-05, "loss": 2.0587, "step": 8927 }, { "epoch": 0.22, "learning_rate": 1.816477186832866e-05, "loss": 2.2497, "step": 8928 }, { "epoch": 0.22, "learning_rate": 1.8164310903723953e-05, "loss": 2.2757, "step": 8929 }, { "epoch": 0.22, "learning_rate": 1.8163849887085433e-05, "loss": 2.2317, "step": 8930 }, { "epoch": 0.22, "learning_rate": 1.8163388818416034e-05, "loss": 2.356, "step": 8931 }, { "epoch": 0.22, "learning_rate": 1.8162927697718703e-05, "loss": 2.1414, "step": 8932 }, { "epoch": 0.22, "learning_rate": 1.816246652499637e-05, "loss": 2.0628, "step": 8933 }, { "epoch": 0.22, "learning_rate": 1.8162005300251974e-05, "loss": 2.1113, "step": 8934 }, { "epoch": 0.22, "learning_rate": 1.8161544023488463e-05, "loss": 2.3782, "step": 8935 }, { "epoch": 0.22, "learning_rate": 1.816108269470877e-05, "loss": 2.121, "step": 8936 }, { "epoch": 0.22, "learning_rate": 1.816062131391584e-05, "loss": 2.0779, "step": 8937 }, { "epoch": 0.22, "learning_rate": 1.816015988111261e-05, "loss": 2.2274, "step": 8938 }, { "epoch": 0.22, "learning_rate": 1.8159698396302023e-05, "loss": 2.0412, "step": 8939 }, { "epoch": 0.22, "learning_rate": 1.8159236859487017e-05, "loss": 2.2405, "step": 8940 }, { "epoch": 0.22, "learning_rate": 1.8158775270670535e-05, "loss": 2.2622, "step": 8941 }, { "epoch": 0.22, "learning_rate": 1.815831362985552e-05, "loss": 2.2495, "step": 8942 }, { "epoch": 0.22, "learning_rate": 1.8157851937044913e-05, "loss": 2.1458, "step": 8943 }, { "epoch": 0.22, "learning_rate": 1.815739019224166e-05, "loss": 2.1758, "step": 8944 }, { "epoch": 0.22, "learning_rate": 1.8156928395448697e-05, "loss": 2.2359, "step": 8945 }, { "epoch": 0.22, "learning_rate": 1.8156466546668974e-05, "loss": 2.2135, "step": 8946 }, { "epoch": 0.22, "learning_rate": 1.815600464590543e-05, "loss": 2.1857, "step": 8947 }, { "epoch": 0.22, "learning_rate": 1.815554269316101e-05, "loss": 2.0876, "step": 8948 }, { "epoch": 0.22, "learning_rate": 1.815508068843866e-05, "loss": 2.2814, "step": 8949 }, { "epoch": 0.22, "learning_rate": 1.8154618631741323e-05, "loss": 2.1526, "step": 8950 }, { "epoch": 0.22, "learning_rate": 1.8154156523071938e-05, "loss": 2.2408, "step": 8951 }, { "epoch": 0.22, "learning_rate": 1.8153694362433464e-05, "loss": 2.3607, "step": 8952 }, { "epoch": 0.22, "learning_rate": 1.8153232149828838e-05, "loss": 2.1543, "step": 8953 }, { "epoch": 0.22, "learning_rate": 1.8152769885261003e-05, "loss": 2.1167, "step": 8954 }, { "epoch": 0.22, "learning_rate": 1.8152307568732908e-05, "loss": 2.2549, "step": 8955 }, { "epoch": 0.22, "learning_rate": 1.8151845200247504e-05, "loss": 2.2911, "step": 8956 }, { "epoch": 0.22, "learning_rate": 1.8151382779807735e-05, "loss": 2.2044, "step": 8957 }, { "epoch": 0.22, "learning_rate": 1.815092030741654e-05, "loss": 2.0465, "step": 8958 }, { "epoch": 0.22, "learning_rate": 1.815045778307688e-05, "loss": 2.2085, "step": 8959 }, { "epoch": 0.22, "learning_rate": 1.8149995206791697e-05, "loss": 2.3098, "step": 8960 }, { "epoch": 0.22, "learning_rate": 1.8149532578563934e-05, "loss": 2.1219, "step": 8961 }, { "epoch": 0.22, "learning_rate": 1.8149069898396545e-05, "loss": 2.1483, "step": 8962 }, { "epoch": 0.22, "learning_rate": 1.8148607166292476e-05, "loss": 2.1105, "step": 8963 }, { "epoch": 0.22, "learning_rate": 1.8148144382254684e-05, "loss": 2.1933, "step": 8964 }, { "epoch": 0.22, "learning_rate": 1.814768154628611e-05, "loss": 2.0835, "step": 8965 }, { "epoch": 0.22, "learning_rate": 1.81472186583897e-05, "loss": 2.211, "step": 8966 }, { "epoch": 0.22, "learning_rate": 1.8146755718568418e-05, "loss": 2.0898, "step": 8967 }, { "epoch": 0.22, "learning_rate": 1.8146292726825204e-05, "loss": 2.0148, "step": 8968 }, { "epoch": 0.22, "learning_rate": 1.8145829683163013e-05, "loss": 2.1733, "step": 8969 }, { "epoch": 0.22, "learning_rate": 1.8145366587584795e-05, "loss": 1.9973, "step": 8970 }, { "epoch": 0.22, "learning_rate": 1.81449034400935e-05, "loss": 2.1358, "step": 8971 }, { "epoch": 0.22, "learning_rate": 1.814444024069208e-05, "loss": 2.256, "step": 8972 }, { "epoch": 0.22, "learning_rate": 1.814397698938349e-05, "loss": 2.0898, "step": 8973 }, { "epoch": 0.22, "learning_rate": 1.814351368617068e-05, "loss": 2.1943, "step": 8974 }, { "epoch": 0.22, "learning_rate": 1.8143050331056602e-05, "loss": 2.0282, "step": 8975 }, { "epoch": 0.22, "learning_rate": 1.8142586924044214e-05, "loss": 2.1086, "step": 8976 }, { "epoch": 0.22, "learning_rate": 1.8142123465136464e-05, "loss": 2.195, "step": 8977 }, { "epoch": 0.22, "learning_rate": 1.8141659954336305e-05, "loss": 2.1544, "step": 8978 }, { "epoch": 0.22, "learning_rate": 1.8141196391646697e-05, "loss": 2.1967, "step": 8979 }, { "epoch": 0.22, "learning_rate": 1.8140732777070592e-05, "loss": 2.357, "step": 8980 }, { "epoch": 0.22, "learning_rate": 1.814026911061094e-05, "loss": 2.1219, "step": 8981 }, { "epoch": 0.22, "learning_rate": 1.8139805392270705e-05, "loss": 2.2137, "step": 8982 }, { "epoch": 0.22, "learning_rate": 1.8139341622052834e-05, "loss": 2.3013, "step": 8983 }, { "epoch": 0.22, "learning_rate": 1.8138877799960288e-05, "loss": 2.3193, "step": 8984 }, { "epoch": 0.22, "learning_rate": 1.8138413925996023e-05, "loss": 2.2294, "step": 8985 }, { "epoch": 0.22, "learning_rate": 1.813795000016299e-05, "loss": 2.2304, "step": 8986 }, { "epoch": 0.22, "learning_rate": 1.813748602246415e-05, "loss": 2.1344, "step": 8987 }, { "epoch": 0.22, "learning_rate": 1.8137021992902463e-05, "loss": 2.1264, "step": 8988 }, { "epoch": 0.22, "learning_rate": 1.813655791148088e-05, "loss": 2.2831, "step": 8989 }, { "epoch": 0.22, "learning_rate": 1.813609377820236e-05, "loss": 2.1969, "step": 8990 }, { "epoch": 0.22, "learning_rate": 1.8135629593069866e-05, "loss": 2.1959, "step": 8991 }, { "epoch": 0.22, "learning_rate": 1.8135165356086356e-05, "loss": 2.1575, "step": 8992 }, { "epoch": 0.22, "learning_rate": 1.8134701067254782e-05, "loss": 2.1239, "step": 8993 }, { "epoch": 0.22, "learning_rate": 1.8134236726578108e-05, "loss": 2.2459, "step": 8994 }, { "epoch": 0.22, "learning_rate": 1.8133772334059294e-05, "loss": 2.3109, "step": 8995 }, { "epoch": 0.22, "learning_rate": 1.8133307889701294e-05, "loss": 2.3193, "step": 8996 }, { "epoch": 0.22, "learning_rate": 1.8132843393507077e-05, "loss": 2.2704, "step": 8997 }, { "epoch": 0.22, "learning_rate": 1.8132378845479596e-05, "loss": 2.314, "step": 8998 }, { "epoch": 0.22, "learning_rate": 1.8131914245621814e-05, "loss": 2.1982, "step": 8999 }, { "epoch": 0.22, "learning_rate": 1.8131449593936695e-05, "loss": 1.9962, "step": 9000 }, { "epoch": 0.22, "learning_rate": 1.8130984890427193e-05, "loss": 2.2864, "step": 9001 }, { "epoch": 0.22, "learning_rate": 1.813052013509628e-05, "loss": 2.1536, "step": 9002 }, { "epoch": 0.22, "learning_rate": 1.8130055327946913e-05, "loss": 2.3927, "step": 9003 }, { "epoch": 0.22, "learning_rate": 1.8129590468982047e-05, "loss": 2.2592, "step": 9004 }, { "epoch": 0.22, "learning_rate": 1.8129125558204657e-05, "loss": 2.0937, "step": 9005 }, { "epoch": 0.22, "learning_rate": 1.81286605956177e-05, "loss": 2.2542, "step": 9006 }, { "epoch": 0.22, "learning_rate": 1.812819558122414e-05, "loss": 2.207, "step": 9007 }, { "epoch": 0.22, "learning_rate": 1.8127730515026936e-05, "loss": 2.2065, "step": 9008 }, { "epoch": 0.22, "learning_rate": 1.812726539702906e-05, "loss": 2.1536, "step": 9009 }, { "epoch": 0.22, "learning_rate": 1.8126800227233475e-05, "loss": 2.1054, "step": 9010 }, { "epoch": 0.22, "learning_rate": 1.8126335005643146e-05, "loss": 2.1547, "step": 9011 }, { "epoch": 0.22, "learning_rate": 1.812586973226103e-05, "loss": 2.3354, "step": 9012 }, { "epoch": 0.22, "learning_rate": 1.81254044070901e-05, "loss": 2.3054, "step": 9013 }, { "epoch": 0.22, "learning_rate": 1.812493903013332e-05, "loss": 2.2839, "step": 9014 }, { "epoch": 0.22, "learning_rate": 1.8124473601393655e-05, "loss": 2.2038, "step": 9015 }, { "epoch": 0.22, "learning_rate": 1.812400812087407e-05, "loss": 2.0376, "step": 9016 }, { "epoch": 0.22, "learning_rate": 1.812354258857754e-05, "loss": 2.0838, "step": 9017 }, { "epoch": 0.22, "learning_rate": 1.812307700450702e-05, "loss": 2.2028, "step": 9018 }, { "epoch": 0.22, "learning_rate": 1.8122611368665485e-05, "loss": 2.1866, "step": 9019 }, { "epoch": 0.22, "learning_rate": 1.8122145681055903e-05, "loss": 2.2658, "step": 9020 }, { "epoch": 0.22, "learning_rate": 1.8121679941681238e-05, "loss": 2.3587, "step": 9021 }, { "epoch": 0.22, "learning_rate": 1.8121214150544457e-05, "loss": 2.1417, "step": 9022 }, { "epoch": 0.22, "learning_rate": 1.8120748307648535e-05, "loss": 2.2932, "step": 9023 }, { "epoch": 0.22, "learning_rate": 1.8120282412996434e-05, "loss": 2.4284, "step": 9024 }, { "epoch": 0.22, "learning_rate": 1.811981646659113e-05, "loss": 2.2555, "step": 9025 }, { "epoch": 0.22, "learning_rate": 1.8119350468435588e-05, "loss": 2.1339, "step": 9026 }, { "epoch": 0.22, "learning_rate": 1.8118884418532782e-05, "loss": 2.3371, "step": 9027 }, { "epoch": 0.22, "learning_rate": 1.8118418316885676e-05, "loss": 2.2126, "step": 9028 }, { "epoch": 0.22, "learning_rate": 1.8117952163497248e-05, "loss": 2.1416, "step": 9029 }, { "epoch": 0.22, "learning_rate": 1.811748595837046e-05, "loss": 2.2796, "step": 9030 }, { "epoch": 0.22, "learning_rate": 1.8117019701508294e-05, "loss": 2.2374, "step": 9031 }, { "epoch": 0.22, "learning_rate": 1.8116553392913715e-05, "loss": 2.1951, "step": 9032 }, { "epoch": 0.22, "learning_rate": 1.8116087032589696e-05, "loss": 2.2278, "step": 9033 }, { "epoch": 0.22, "learning_rate": 1.8115620620539208e-05, "loss": 2.2716, "step": 9034 }, { "epoch": 0.22, "learning_rate": 1.8115154156765226e-05, "loss": 2.3797, "step": 9035 }, { "epoch": 0.22, "learning_rate": 1.811468764127072e-05, "loss": 2.0989, "step": 9036 }, { "epoch": 0.22, "learning_rate": 1.811422107405867e-05, "loss": 2.3186, "step": 9037 }, { "epoch": 0.22, "learning_rate": 1.811375445513204e-05, "loss": 2.4493, "step": 9038 }, { "epoch": 0.22, "learning_rate": 1.811328778449381e-05, "loss": 2.2162, "step": 9039 }, { "epoch": 0.22, "learning_rate": 1.811282106214695e-05, "loss": 1.9791, "step": 9040 }, { "epoch": 0.22, "learning_rate": 1.8112354288094446e-05, "loss": 2.2964, "step": 9041 }, { "epoch": 0.22, "learning_rate": 1.8111887462339256e-05, "loss": 2.1127, "step": 9042 }, { "epoch": 0.22, "learning_rate": 1.8111420584884367e-05, "loss": 2.2361, "step": 9043 }, { "epoch": 0.22, "learning_rate": 1.8110953655732753e-05, "loss": 2.1377, "step": 9044 }, { "epoch": 0.22, "learning_rate": 1.8110486674887386e-05, "loss": 2.2898, "step": 9045 }, { "epoch": 0.22, "learning_rate": 1.8110019642351244e-05, "loss": 2.2599, "step": 9046 }, { "epoch": 0.22, "learning_rate": 1.81095525581273e-05, "loss": 2.023, "step": 9047 }, { "epoch": 0.22, "learning_rate": 1.8109085422218543e-05, "loss": 2.1639, "step": 9048 }, { "epoch": 0.22, "learning_rate": 1.8108618234627937e-05, "loss": 2.4415, "step": 9049 }, { "epoch": 0.22, "learning_rate": 1.8108150995358466e-05, "loss": 2.2164, "step": 9050 }, { "epoch": 0.22, "learning_rate": 1.8107683704413104e-05, "loss": 2.2928, "step": 9051 }, { "epoch": 0.22, "learning_rate": 1.810721636179483e-05, "loss": 2.1487, "step": 9052 }, { "epoch": 0.22, "learning_rate": 1.8106748967506628e-05, "loss": 1.9861, "step": 9053 }, { "epoch": 0.22, "learning_rate": 1.8106281521551475e-05, "loss": 2.1573, "step": 9054 }, { "epoch": 0.22, "learning_rate": 1.8105814023932344e-05, "loss": 2.1522, "step": 9055 }, { "epoch": 0.22, "learning_rate": 1.810534647465222e-05, "loss": 2.1466, "step": 9056 }, { "epoch": 0.22, "learning_rate": 1.8104878873714082e-05, "loss": 2.1269, "step": 9057 }, { "epoch": 0.22, "learning_rate": 1.8104411221120908e-05, "loss": 2.2544, "step": 9058 }, { "epoch": 0.22, "learning_rate": 1.810394351687568e-05, "loss": 2.106, "step": 9059 }, { "epoch": 0.22, "learning_rate": 1.8103475760981378e-05, "loss": 2.1943, "step": 9060 }, { "epoch": 0.22, "learning_rate": 1.810300795344099e-05, "loss": 2.1513, "step": 9061 }, { "epoch": 0.22, "learning_rate": 1.8102540094257485e-05, "loss": 2.3119, "step": 9062 }, { "epoch": 0.22, "learning_rate": 1.8102072183433856e-05, "loss": 2.2764, "step": 9063 }, { "epoch": 0.22, "learning_rate": 1.810160422097308e-05, "loss": 2.3061, "step": 9064 }, { "epoch": 0.22, "learning_rate": 1.8101136206878135e-05, "loss": 2.1414, "step": 9065 }, { "epoch": 0.22, "learning_rate": 1.8100668141152013e-05, "loss": 2.4151, "step": 9066 }, { "epoch": 0.22, "learning_rate": 1.8100200023797692e-05, "loss": 2.0418, "step": 9067 }, { "epoch": 0.22, "learning_rate": 1.809973185481816e-05, "loss": 2.3002, "step": 9068 }, { "epoch": 0.22, "learning_rate": 1.8099263634216394e-05, "loss": 2.2378, "step": 9069 }, { "epoch": 0.22, "learning_rate": 1.809879536199538e-05, "loss": 2.3452, "step": 9070 }, { "epoch": 0.22, "learning_rate": 1.8098327038158106e-05, "loss": 2.275, "step": 9071 }, { "epoch": 0.22, "learning_rate": 1.8097858662707556e-05, "loss": 2.0414, "step": 9072 }, { "epoch": 0.22, "learning_rate": 1.8097390235646707e-05, "loss": 2.1477, "step": 9073 }, { "epoch": 0.22, "learning_rate": 1.809692175697856e-05, "loss": 2.2479, "step": 9074 }, { "epoch": 0.22, "learning_rate": 1.8096453226706088e-05, "loss": 2.2376, "step": 9075 }, { "epoch": 0.22, "learning_rate": 1.809598464483228e-05, "loss": 2.2134, "step": 9076 }, { "epoch": 0.22, "learning_rate": 1.8095516011360125e-05, "loss": 2.0772, "step": 9077 }, { "epoch": 0.22, "learning_rate": 1.8095047326292606e-05, "loss": 2.1995, "step": 9078 }, { "epoch": 0.22, "learning_rate": 1.8094578589632714e-05, "loss": 2.1125, "step": 9079 }, { "epoch": 0.22, "learning_rate": 1.8094109801383436e-05, "loss": 2.1483, "step": 9080 }, { "epoch": 0.22, "learning_rate": 1.8093640961547758e-05, "loss": 2.0421, "step": 9081 }, { "epoch": 0.22, "learning_rate": 1.8093172070128665e-05, "loss": 2.2835, "step": 9082 }, { "epoch": 0.22, "learning_rate": 1.8092703127129154e-05, "loss": 2.164, "step": 9083 }, { "epoch": 0.22, "learning_rate": 1.8092234132552205e-05, "loss": 2.2269, "step": 9084 }, { "epoch": 0.22, "learning_rate": 1.8091765086400813e-05, "loss": 2.2324, "step": 9085 }, { "epoch": 0.22, "learning_rate": 1.8091295988677964e-05, "loss": 2.285, "step": 9086 }, { "epoch": 0.22, "learning_rate": 1.8090826839386646e-05, "loss": 2.1978, "step": 9087 }, { "epoch": 0.22, "learning_rate": 1.8090357638529856e-05, "loss": 2.2496, "step": 9088 }, { "epoch": 0.22, "learning_rate": 1.808988838611058e-05, "loss": 2.2428, "step": 9089 }, { "epoch": 0.22, "learning_rate": 1.8089419082131806e-05, "loss": 2.2336, "step": 9090 }, { "epoch": 0.22, "learning_rate": 1.808894972659653e-05, "loss": 2.5044, "step": 9091 }, { "epoch": 0.22, "learning_rate": 1.808848031950774e-05, "loss": 2.1597, "step": 9092 }, { "epoch": 0.22, "learning_rate": 1.8088010860868432e-05, "loss": 2.1514, "step": 9093 }, { "epoch": 0.22, "learning_rate": 1.808754135068159e-05, "loss": 2.3258, "step": 9094 }, { "epoch": 0.22, "learning_rate": 1.8087071788950213e-05, "loss": 2.15, "step": 9095 }, { "epoch": 0.22, "learning_rate": 1.8086602175677296e-05, "loss": 2.1094, "step": 9096 }, { "epoch": 0.22, "learning_rate": 1.8086132510865825e-05, "loss": 2.2959, "step": 9097 }, { "epoch": 0.22, "learning_rate": 1.8085662794518793e-05, "loss": 2.254, "step": 9098 }, { "epoch": 0.22, "learning_rate": 1.80851930266392e-05, "loss": 2.1396, "step": 9099 }, { "epoch": 0.22, "learning_rate": 1.808472320723004e-05, "loss": 2.1281, "step": 9100 }, { "epoch": 0.22, "learning_rate": 1.8084253336294303e-05, "loss": 2.1266, "step": 9101 }, { "epoch": 0.22, "learning_rate": 1.8083783413834983e-05, "loss": 2.2487, "step": 9102 }, { "epoch": 0.22, "learning_rate": 1.8083313439855078e-05, "loss": 2.207, "step": 9103 }, { "epoch": 0.22, "learning_rate": 1.808284341435758e-05, "loss": 1.9935, "step": 9104 }, { "epoch": 0.22, "learning_rate": 1.808237333734549e-05, "loss": 2.1682, "step": 9105 }, { "epoch": 0.22, "learning_rate": 1.80819032088218e-05, "loss": 2.0665, "step": 9106 }, { "epoch": 0.22, "learning_rate": 1.8081433028789504e-05, "loss": 2.265, "step": 9107 }, { "epoch": 0.22, "learning_rate": 1.8080962797251606e-05, "loss": 2.3378, "step": 9108 }, { "epoch": 0.22, "learning_rate": 1.8080492514211096e-05, "loss": 2.1502, "step": 9109 }, { "epoch": 0.22, "learning_rate": 1.8080022179670972e-05, "loss": 2.1676, "step": 9110 }, { "epoch": 0.22, "learning_rate": 1.8079551793634237e-05, "loss": 2.1936, "step": 9111 }, { "epoch": 0.22, "learning_rate": 1.8079081356103882e-05, "loss": 2.2374, "step": 9112 }, { "epoch": 0.22, "learning_rate": 1.8078610867082912e-05, "loss": 2.1862, "step": 9113 }, { "epoch": 0.22, "learning_rate": 1.807814032657432e-05, "loss": 2.2113, "step": 9114 }, { "epoch": 0.22, "learning_rate": 1.8077669734581107e-05, "loss": 2.2046, "step": 9115 }, { "epoch": 0.22, "learning_rate": 1.8077199091106275e-05, "loss": 1.9997, "step": 9116 }, { "epoch": 0.22, "learning_rate": 1.807672839615282e-05, "loss": 2.3749, "step": 9117 }, { "epoch": 0.22, "learning_rate": 1.8076257649723738e-05, "loss": 2.2977, "step": 9118 }, { "epoch": 0.22, "learning_rate": 1.807578685182204e-05, "loss": 2.211, "step": 9119 }, { "epoch": 0.22, "learning_rate": 1.807531600245072e-05, "loss": 2.2168, "step": 9120 }, { "epoch": 0.22, "learning_rate": 1.8074845101612778e-05, "loss": 2.2881, "step": 9121 }, { "epoch": 0.22, "learning_rate": 1.8074374149311217e-05, "loss": 2.1296, "step": 9122 }, { "epoch": 0.22, "learning_rate": 1.8073903145549037e-05, "loss": 2.1593, "step": 9123 }, { "epoch": 0.22, "learning_rate": 1.8073432090329243e-05, "loss": 2.2671, "step": 9124 }, { "epoch": 0.22, "learning_rate": 1.8072960983654834e-05, "loss": 2.1857, "step": 9125 }, { "epoch": 0.22, "learning_rate": 1.807248982552881e-05, "loss": 2.2288, "step": 9126 }, { "epoch": 0.22, "learning_rate": 1.8072018615954182e-05, "loss": 2.2211, "step": 9127 }, { "epoch": 0.22, "learning_rate": 1.807154735493395e-05, "loss": 2.2332, "step": 9128 }, { "epoch": 0.23, "learning_rate": 1.8071076042471114e-05, "loss": 2.1282, "step": 9129 }, { "epoch": 0.23, "learning_rate": 1.8070604678568682e-05, "loss": 2.2078, "step": 9130 }, { "epoch": 0.23, "learning_rate": 1.8070133263229656e-05, "loss": 2.2289, "step": 9131 }, { "epoch": 0.23, "learning_rate": 1.8069661796457038e-05, "loss": 2.2076, "step": 9132 }, { "epoch": 0.23, "learning_rate": 1.8069190278253836e-05, "loss": 2.1862, "step": 9133 }, { "epoch": 0.23, "learning_rate": 1.8068718708623055e-05, "loss": 2.2057, "step": 9134 }, { "epoch": 0.23, "learning_rate": 1.8068247087567702e-05, "loss": 2.2126, "step": 9135 }, { "epoch": 0.23, "learning_rate": 1.806777541509078e-05, "loss": 2.2918, "step": 9136 }, { "epoch": 0.23, "learning_rate": 1.8067303691195298e-05, "loss": 2.1911, "step": 9137 }, { "epoch": 0.23, "learning_rate": 1.8066831915884257e-05, "loss": 2.2313, "step": 9138 }, { "epoch": 0.23, "learning_rate": 1.8066360089160668e-05, "loss": 2.2528, "step": 9139 }, { "epoch": 0.23, "learning_rate": 1.806588821102754e-05, "loss": 2.4862, "step": 9140 }, { "epoch": 0.23, "learning_rate": 1.8065416281487875e-05, "loss": 2.2655, "step": 9141 }, { "epoch": 0.23, "learning_rate": 1.8064944300544685e-05, "loss": 2.2122, "step": 9142 }, { "epoch": 0.23, "learning_rate": 1.8064472268200975e-05, "loss": 2.0429, "step": 9143 }, { "epoch": 0.23, "learning_rate": 1.8064000184459756e-05, "loss": 2.1174, "step": 9144 }, { "epoch": 0.23, "learning_rate": 1.8063528049324037e-05, "loss": 2.1923, "step": 9145 }, { "epoch": 0.23, "learning_rate": 1.8063055862796827e-05, "loss": 2.1785, "step": 9146 }, { "epoch": 0.23, "learning_rate": 1.806258362488113e-05, "loss": 2.1597, "step": 9147 }, { "epoch": 0.23, "learning_rate": 1.806211133557996e-05, "loss": 2.2546, "step": 9148 }, { "epoch": 0.23, "learning_rate": 1.8061638994896332e-05, "loss": 2.1663, "step": 9149 }, { "epoch": 0.23, "learning_rate": 1.806116660283325e-05, "loss": 2.3015, "step": 9150 }, { "epoch": 0.23, "learning_rate": 1.8060694159393727e-05, "loss": 2.1719, "step": 9151 }, { "epoch": 0.23, "learning_rate": 1.806022166458077e-05, "loss": 2.2799, "step": 9152 }, { "epoch": 0.23, "learning_rate": 1.8059749118397395e-05, "loss": 2.112, "step": 9153 }, { "epoch": 0.23, "learning_rate": 1.805927652084661e-05, "loss": 2.1696, "step": 9154 }, { "epoch": 0.23, "learning_rate": 1.8058803871931436e-05, "loss": 2.1394, "step": 9155 }, { "epoch": 0.23, "learning_rate": 1.8058331171654873e-05, "loss": 2.2407, "step": 9156 }, { "epoch": 0.23, "learning_rate": 1.805785842001994e-05, "loss": 2.212, "step": 9157 }, { "epoch": 0.23, "learning_rate": 1.805738561702965e-05, "loss": 1.9604, "step": 9158 }, { "epoch": 0.23, "learning_rate": 1.805691276268701e-05, "loss": 2.1437, "step": 9159 }, { "epoch": 0.23, "learning_rate": 1.805643985699505e-05, "loss": 2.0847, "step": 9160 }, { "epoch": 0.23, "learning_rate": 1.8055966899956767e-05, "loss": 2.0863, "step": 9161 }, { "epoch": 0.23, "learning_rate": 1.8055493891575185e-05, "loss": 2.116, "step": 9162 }, { "epoch": 0.23, "learning_rate": 1.805502083185331e-05, "loss": 2.3208, "step": 9163 }, { "epoch": 0.23, "learning_rate": 1.8054547720794165e-05, "loss": 2.1832, "step": 9164 }, { "epoch": 0.23, "learning_rate": 1.8054074558400764e-05, "loss": 2.065, "step": 9165 }, { "epoch": 0.23, "learning_rate": 1.805360134467612e-05, "loss": 2.2669, "step": 9166 }, { "epoch": 0.23, "learning_rate": 1.805312807962325e-05, "loss": 2.1524, "step": 9167 }, { "epoch": 0.23, "learning_rate": 1.8052654763245168e-05, "loss": 2.1424, "step": 9168 }, { "epoch": 0.23, "learning_rate": 1.8052181395544896e-05, "loss": 2.4287, "step": 9169 }, { "epoch": 0.23, "learning_rate": 1.8051707976525445e-05, "loss": 2.2542, "step": 9170 }, { "epoch": 0.23, "learning_rate": 1.8051234506189838e-05, "loss": 2.1478, "step": 9171 }, { "epoch": 0.23, "learning_rate": 1.805076098454109e-05, "loss": 2.1585, "step": 9172 }, { "epoch": 0.23, "learning_rate": 1.8050287411582217e-05, "loss": 2.288, "step": 9173 }, { "epoch": 0.23, "learning_rate": 1.8049813787316237e-05, "loss": 2.22, "step": 9174 }, { "epoch": 0.23, "learning_rate": 1.8049340111746173e-05, "loss": 2.1434, "step": 9175 }, { "epoch": 0.23, "learning_rate": 1.804886638487504e-05, "loss": 2.2026, "step": 9176 }, { "epoch": 0.23, "learning_rate": 1.804839260670586e-05, "loss": 2.131, "step": 9177 }, { "epoch": 0.23, "learning_rate": 1.804791877724165e-05, "loss": 2.1483, "step": 9178 }, { "epoch": 0.23, "learning_rate": 1.804744489648543e-05, "loss": 2.212, "step": 9179 }, { "epoch": 0.23, "learning_rate": 1.804697096444022e-05, "loss": 2.2557, "step": 9180 }, { "epoch": 0.23, "learning_rate": 1.8046496981109044e-05, "loss": 2.1722, "step": 9181 }, { "epoch": 0.23, "learning_rate": 1.804602294649492e-05, "loss": 2.3684, "step": 9182 }, { "epoch": 0.23, "learning_rate": 1.8045548860600873e-05, "loss": 2.3712, "step": 9183 }, { "epoch": 0.23, "learning_rate": 1.8045074723429915e-05, "loss": 2.3142, "step": 9184 }, { "epoch": 0.23, "learning_rate": 1.8044600534985078e-05, "loss": 2.2493, "step": 9185 }, { "epoch": 0.23, "learning_rate": 1.804412629526938e-05, "loss": 2.2708, "step": 9186 }, { "epoch": 0.23, "learning_rate": 1.8043652004285844e-05, "loss": 2.1027, "step": 9187 }, { "epoch": 0.23, "learning_rate": 1.8043177662037486e-05, "loss": 2.3406, "step": 9188 }, { "epoch": 0.23, "learning_rate": 1.804270326852734e-05, "loss": 2.3338, "step": 9189 }, { "epoch": 0.23, "learning_rate": 1.8042228823758428e-05, "loss": 2.1959, "step": 9190 }, { "epoch": 0.23, "learning_rate": 1.804175432773377e-05, "loss": 2.3148, "step": 9191 }, { "epoch": 0.23, "learning_rate": 1.8041279780456387e-05, "loss": 2.3193, "step": 9192 }, { "epoch": 0.23, "learning_rate": 1.804080518192931e-05, "loss": 2.1756, "step": 9193 }, { "epoch": 0.23, "learning_rate": 1.804033053215556e-05, "loss": 2.0528, "step": 9194 }, { "epoch": 0.23, "learning_rate": 1.8039855831138167e-05, "loss": 2.1951, "step": 9195 }, { "epoch": 0.23, "learning_rate": 1.803938107888015e-05, "loss": 2.0839, "step": 9196 }, { "epoch": 0.23, "learning_rate": 1.8038906275384537e-05, "loss": 2.2649, "step": 9197 }, { "epoch": 0.23, "learning_rate": 1.803843142065435e-05, "loss": 2.1648, "step": 9198 }, { "epoch": 0.23, "learning_rate": 1.8037956514692627e-05, "loss": 2.15, "step": 9199 }, { "epoch": 0.23, "learning_rate": 1.8037481557502388e-05, "loss": 2.2258, "step": 9200 }, { "epoch": 0.23, "learning_rate": 1.8037006549086655e-05, "loss": 2.1661, "step": 9201 }, { "epoch": 0.23, "learning_rate": 1.8036531489448462e-05, "loss": 2.0867, "step": 9202 }, { "epoch": 0.23, "learning_rate": 1.8036056378590837e-05, "loss": 2.0874, "step": 9203 }, { "epoch": 0.23, "learning_rate": 1.8035581216516802e-05, "loss": 2.1485, "step": 9204 }, { "epoch": 0.23, "learning_rate": 1.8035106003229394e-05, "loss": 1.9683, "step": 9205 }, { "epoch": 0.23, "learning_rate": 1.803463073873163e-05, "loss": 2.2073, "step": 9206 }, { "epoch": 0.23, "learning_rate": 1.803415542302655e-05, "loss": 2.3429, "step": 9207 }, { "epoch": 0.23, "learning_rate": 1.8033680056117178e-05, "loss": 2.1315, "step": 9208 }, { "epoch": 0.23, "learning_rate": 1.8033204638006547e-05, "loss": 2.1844, "step": 9209 }, { "epoch": 0.23, "learning_rate": 1.803272916869768e-05, "loss": 2.0956, "step": 9210 }, { "epoch": 0.23, "learning_rate": 1.8032253648193618e-05, "loss": 2.1384, "step": 9211 }, { "epoch": 0.23, "learning_rate": 1.8031778076497387e-05, "loss": 2.0946, "step": 9212 }, { "epoch": 0.23, "learning_rate": 1.8031302453612012e-05, "loss": 2.153, "step": 9213 }, { "epoch": 0.23, "learning_rate": 1.803082677954053e-05, "loss": 2.3368, "step": 9214 }, { "epoch": 0.23, "learning_rate": 1.803035105428597e-05, "loss": 2.0721, "step": 9215 }, { "epoch": 0.23, "learning_rate": 1.802987527785137e-05, "loss": 2.1187, "step": 9216 }, { "epoch": 0.23, "learning_rate": 1.8029399450239754e-05, "loss": 2.2357, "step": 9217 }, { "epoch": 0.23, "learning_rate": 1.8028923571454165e-05, "loss": 2.2266, "step": 9218 }, { "epoch": 0.23, "learning_rate": 1.8028447641497625e-05, "loss": 2.1401, "step": 9219 }, { "epoch": 0.23, "learning_rate": 1.802797166037317e-05, "loss": 2.2153, "step": 9220 }, { "epoch": 0.23, "learning_rate": 1.802749562808384e-05, "loss": 2.1998, "step": 9221 }, { "epoch": 0.23, "learning_rate": 1.8027019544632662e-05, "loss": 2.0731, "step": 9222 }, { "epoch": 0.23, "learning_rate": 1.802654341002267e-05, "loss": 2.4513, "step": 9223 }, { "epoch": 0.23, "learning_rate": 1.8026067224256906e-05, "loss": 2.1605, "step": 9224 }, { "epoch": 0.23, "learning_rate": 1.8025590987338393e-05, "loss": 2.204, "step": 9225 }, { "epoch": 0.23, "learning_rate": 1.802511469927018e-05, "loss": 2.2678, "step": 9226 }, { "epoch": 0.23, "learning_rate": 1.8024638360055292e-05, "loss": 2.0425, "step": 9227 }, { "epoch": 0.23, "learning_rate": 1.8024161969696768e-05, "loss": 2.2052, "step": 9228 }, { "epoch": 0.23, "learning_rate": 1.8023685528197647e-05, "loss": 1.8843, "step": 9229 }, { "epoch": 0.23, "learning_rate": 1.8023209035560966e-05, "loss": 2.2185, "step": 9230 }, { "epoch": 0.23, "learning_rate": 1.8022732491789755e-05, "loss": 2.0689, "step": 9231 }, { "epoch": 0.23, "learning_rate": 1.8022255896887056e-05, "loss": 2.3112, "step": 9232 }, { "epoch": 0.23, "learning_rate": 1.8021779250855905e-05, "loss": 2.1142, "step": 9233 }, { "epoch": 0.23, "learning_rate": 1.802130255369934e-05, "loss": 2.3608, "step": 9234 }, { "epoch": 0.23, "learning_rate": 1.8020825805420404e-05, "loss": 2.2061, "step": 9235 }, { "epoch": 0.23, "learning_rate": 1.8020349006022128e-05, "loss": 2.2788, "step": 9236 }, { "epoch": 0.23, "learning_rate": 1.8019872155507555e-05, "loss": 2.1874, "step": 9237 }, { "epoch": 0.23, "learning_rate": 1.8019395253879724e-05, "loss": 2.1993, "step": 9238 }, { "epoch": 0.23, "learning_rate": 1.801891830114167e-05, "loss": 2.1571, "step": 9239 }, { "epoch": 0.23, "learning_rate": 1.8018441297296437e-05, "loss": 2.0685, "step": 9240 }, { "epoch": 0.23, "learning_rate": 1.8017964242347067e-05, "loss": 2.0636, "step": 9241 }, { "epoch": 0.23, "learning_rate": 1.80174871362966e-05, "loss": 2.1339, "step": 9242 }, { "epoch": 0.23, "learning_rate": 1.801700997914807e-05, "loss": 2.2702, "step": 9243 }, { "epoch": 0.23, "learning_rate": 1.8016532770904523e-05, "loss": 2.1082, "step": 9244 }, { "epoch": 0.23, "learning_rate": 1.8016055511569003e-05, "loss": 2.2275, "step": 9245 }, { "epoch": 0.23, "learning_rate": 1.8015578201144545e-05, "loss": 2.2769, "step": 9246 }, { "epoch": 0.23, "learning_rate": 1.80151008396342e-05, "loss": 2.0917, "step": 9247 }, { "epoch": 0.23, "learning_rate": 1.8014623427041003e-05, "loss": 1.9849, "step": 9248 }, { "epoch": 0.23, "learning_rate": 1.8014145963368e-05, "loss": 2.1105, "step": 9249 }, { "epoch": 0.23, "learning_rate": 1.801366844861823e-05, "loss": 2.2405, "step": 9250 }, { "epoch": 0.23, "learning_rate": 1.801319088279474e-05, "loss": 2.0909, "step": 9251 }, { "epoch": 0.23, "learning_rate": 1.8012713265900576e-05, "loss": 2.0865, "step": 9252 }, { "epoch": 0.23, "learning_rate": 1.8012235597938778e-05, "loss": 2.1057, "step": 9253 }, { "epoch": 0.23, "learning_rate": 1.801175787891239e-05, "loss": 2.1972, "step": 9254 }, { "epoch": 0.23, "learning_rate": 1.801128010882446e-05, "loss": 1.9423, "step": 9255 }, { "epoch": 0.23, "learning_rate": 1.801080228767803e-05, "loss": 2.2136, "step": 9256 }, { "epoch": 0.23, "learning_rate": 1.8010324415476144e-05, "loss": 2.1036, "step": 9257 }, { "epoch": 0.23, "learning_rate": 1.800984649222185e-05, "loss": 2.1886, "step": 9258 }, { "epoch": 0.23, "learning_rate": 1.80093685179182e-05, "loss": 2.0287, "step": 9259 }, { "epoch": 0.23, "learning_rate": 1.8008890492568228e-05, "loss": 2.2278, "step": 9260 }, { "epoch": 0.23, "learning_rate": 1.8008412416174993e-05, "loss": 2.2003, "step": 9261 }, { "epoch": 0.23, "learning_rate": 1.800793428874153e-05, "loss": 2.2468, "step": 9262 }, { "epoch": 0.23, "learning_rate": 1.8007456110270896e-05, "loss": 2.3242, "step": 9263 }, { "epoch": 0.23, "learning_rate": 1.8006977880766132e-05, "loss": 2.3667, "step": 9264 }, { "epoch": 0.23, "learning_rate": 1.800649960023029e-05, "loss": 2.1385, "step": 9265 }, { "epoch": 0.23, "learning_rate": 1.8006021268666417e-05, "loss": 2.1741, "step": 9266 }, { "epoch": 0.23, "learning_rate": 1.8005542886077562e-05, "loss": 2.1434, "step": 9267 }, { "epoch": 0.23, "learning_rate": 1.800506445246677e-05, "loss": 2.1866, "step": 9268 }, { "epoch": 0.23, "learning_rate": 1.8004585967837097e-05, "loss": 2.1929, "step": 9269 }, { "epoch": 0.23, "learning_rate": 1.8004107432191587e-05, "loss": 2.3099, "step": 9270 }, { "epoch": 0.23, "learning_rate": 1.800362884553329e-05, "loss": 2.312, "step": 9271 }, { "epoch": 0.23, "learning_rate": 1.800315020786526e-05, "loss": 2.1854, "step": 9272 }, { "epoch": 0.23, "learning_rate": 1.8002671519190546e-05, "loss": 2.2559, "step": 9273 }, { "epoch": 0.23, "learning_rate": 1.8002192779512197e-05, "loss": 2.2093, "step": 9274 }, { "epoch": 0.23, "learning_rate": 1.800171398883327e-05, "loss": 2.174, "step": 9275 }, { "epoch": 0.23, "learning_rate": 1.8001235147156805e-05, "loss": 2.0665, "step": 9276 }, { "epoch": 0.23, "learning_rate": 1.800075625448587e-05, "loss": 2.2451, "step": 9277 }, { "epoch": 0.23, "learning_rate": 1.80002773108235e-05, "loss": 2.0005, "step": 9278 }, { "epoch": 0.23, "learning_rate": 1.7999798316172762e-05, "loss": 2.1139, "step": 9279 }, { "epoch": 0.23, "learning_rate": 1.7999319270536693e-05, "loss": 2.1175, "step": 9280 }, { "epoch": 0.23, "learning_rate": 1.7998840173918365e-05, "loss": 2.1975, "step": 9281 }, { "epoch": 0.23, "learning_rate": 1.7998361026320817e-05, "loss": 2.079, "step": 9282 }, { "epoch": 0.23, "learning_rate": 1.799788182774711e-05, "loss": 2.1701, "step": 9283 }, { "epoch": 0.23, "learning_rate": 1.7997402578200292e-05, "loss": 2.3279, "step": 9284 }, { "epoch": 0.23, "learning_rate": 1.799692327768342e-05, "loss": 2.368, "step": 9285 }, { "epoch": 0.23, "learning_rate": 1.7996443926199557e-05, "loss": 2.2583, "step": 9286 }, { "epoch": 0.23, "learning_rate": 1.7995964523751747e-05, "loss": 2.1424, "step": 9287 }, { "epoch": 0.23, "learning_rate": 1.7995485070343047e-05, "loss": 2.1237, "step": 9288 }, { "epoch": 0.23, "learning_rate": 1.799500556597652e-05, "loss": 2.0732, "step": 9289 }, { "epoch": 0.23, "learning_rate": 1.799452601065521e-05, "loss": 2.1295, "step": 9290 }, { "epoch": 0.23, "learning_rate": 1.7994046404382186e-05, "loss": 2.2352, "step": 9291 }, { "epoch": 0.23, "learning_rate": 1.79935667471605e-05, "loss": 2.1932, "step": 9292 }, { "epoch": 0.23, "learning_rate": 1.7993087038993206e-05, "loss": 2.2405, "step": 9293 }, { "epoch": 0.23, "learning_rate": 1.7992607279883362e-05, "loss": 2.4039, "step": 9294 }, { "epoch": 0.23, "learning_rate": 1.7992127469834028e-05, "loss": 2.3341, "step": 9295 }, { "epoch": 0.23, "learning_rate": 1.7991647608848263e-05, "loss": 2.2453, "step": 9296 }, { "epoch": 0.23, "learning_rate": 1.799116769692912e-05, "loss": 2.0817, "step": 9297 }, { "epoch": 0.23, "learning_rate": 1.7990687734079665e-05, "loss": 2.142, "step": 9298 }, { "epoch": 0.23, "learning_rate": 1.799020772030295e-05, "loss": 2.0974, "step": 9299 }, { "epoch": 0.23, "learning_rate": 1.7989727655602036e-05, "loss": 2.1491, "step": 9300 }, { "epoch": 0.23, "learning_rate": 1.798924753997999e-05, "loss": 2.0775, "step": 9301 }, { "epoch": 0.23, "learning_rate": 1.7988767373439858e-05, "loss": 2.1598, "step": 9302 }, { "epoch": 0.23, "learning_rate": 1.798828715598471e-05, "loss": 2.1629, "step": 9303 }, { "epoch": 0.23, "learning_rate": 1.798780688761761e-05, "loss": 2.104, "step": 9304 }, { "epoch": 0.23, "learning_rate": 1.798732656834161e-05, "loss": 2.3526, "step": 9305 }, { "epoch": 0.23, "learning_rate": 1.7986846198159775e-05, "loss": 2.3481, "step": 9306 }, { "epoch": 0.23, "learning_rate": 1.7986365777075163e-05, "loss": 2.2352, "step": 9307 }, { "epoch": 0.23, "learning_rate": 1.7985885305090845e-05, "loss": 2.3587, "step": 9308 }, { "epoch": 0.23, "learning_rate": 1.7985404782209875e-05, "loss": 2.3158, "step": 9309 }, { "epoch": 0.23, "learning_rate": 1.7984924208435317e-05, "loss": 2.2495, "step": 9310 }, { "epoch": 0.23, "learning_rate": 1.7984443583770236e-05, "loss": 2.2225, "step": 9311 }, { "epoch": 0.23, "learning_rate": 1.798396290821769e-05, "loss": 2.3473, "step": 9312 }, { "epoch": 0.23, "learning_rate": 1.798348218178075e-05, "loss": 2.1369, "step": 9313 }, { "epoch": 0.23, "learning_rate": 1.7983001404462477e-05, "loss": 2.2236, "step": 9314 }, { "epoch": 0.23, "learning_rate": 1.7982520576265937e-05, "loss": 2.0242, "step": 9315 }, { "epoch": 0.23, "learning_rate": 1.7982039697194186e-05, "loss": 2.1579, "step": 9316 }, { "epoch": 0.23, "learning_rate": 1.7981558767250296e-05, "loss": 2.2874, "step": 9317 }, { "epoch": 0.23, "learning_rate": 1.7981077786437335e-05, "loss": 2.2071, "step": 9318 }, { "epoch": 0.23, "learning_rate": 1.798059675475836e-05, "loss": 2.2219, "step": 9319 }, { "epoch": 0.23, "learning_rate": 1.798011567221644e-05, "loss": 2.0592, "step": 9320 }, { "epoch": 0.23, "learning_rate": 1.797963453881465e-05, "loss": 2.2232, "step": 9321 }, { "epoch": 0.23, "learning_rate": 1.797915335455604e-05, "loss": 2.1802, "step": 9322 }, { "epoch": 0.23, "learning_rate": 1.7978672119443688e-05, "loss": 2.0223, "step": 9323 }, { "epoch": 0.23, "learning_rate": 1.797819083348066e-05, "loss": 2.2489, "step": 9324 }, { "epoch": 0.23, "learning_rate": 1.7977709496670024e-05, "loss": 2.2208, "step": 9325 }, { "epoch": 0.23, "learning_rate": 1.7977228109014842e-05, "loss": 2.33, "step": 9326 }, { "epoch": 0.23, "learning_rate": 1.7976746670518185e-05, "loss": 2.1807, "step": 9327 }, { "epoch": 0.23, "learning_rate": 1.7976265181183123e-05, "loss": 2.2947, "step": 9328 }, { "epoch": 0.23, "learning_rate": 1.7975783641012723e-05, "loss": 2.2562, "step": 9329 }, { "epoch": 0.23, "learning_rate": 1.7975302050010058e-05, "loss": 2.2424, "step": 9330 }, { "epoch": 0.23, "learning_rate": 1.7974820408178192e-05, "loss": 2.0485, "step": 9331 }, { "epoch": 0.23, "learning_rate": 1.7974338715520195e-05, "loss": 2.2962, "step": 9332 }, { "epoch": 0.23, "learning_rate": 1.7973856972039137e-05, "loss": 2.1618, "step": 9333 }, { "epoch": 0.23, "learning_rate": 1.797337517773809e-05, "loss": 2.2577, "step": 9334 }, { "epoch": 0.23, "learning_rate": 1.797289333262013e-05, "loss": 2.1678, "step": 9335 }, { "epoch": 0.23, "learning_rate": 1.7972411436688318e-05, "loss": 2.1982, "step": 9336 }, { "epoch": 0.23, "learning_rate": 1.7971929489945733e-05, "loss": 2.1704, "step": 9337 }, { "epoch": 0.23, "learning_rate": 1.7971447492395442e-05, "loss": 1.9626, "step": 9338 }, { "epoch": 0.23, "learning_rate": 1.7970965444040514e-05, "loss": 2.4036, "step": 9339 }, { "epoch": 0.23, "learning_rate": 1.797048334488403e-05, "loss": 2.2334, "step": 9340 }, { "epoch": 0.23, "learning_rate": 1.797000119492906e-05, "loss": 2.1325, "step": 9341 }, { "epoch": 0.23, "learning_rate": 1.796951899417867e-05, "loss": 2.2996, "step": 9342 }, { "epoch": 0.23, "learning_rate": 1.7969036742635942e-05, "loss": 2.1457, "step": 9343 }, { "epoch": 0.23, "learning_rate": 1.796855444030395e-05, "loss": 2.4015, "step": 9344 }, { "epoch": 0.23, "learning_rate": 1.7968072087185754e-05, "loss": 2.2934, "step": 9345 }, { "epoch": 0.23, "learning_rate": 1.7967589683284443e-05, "loss": 2.0453, "step": 9346 }, { "epoch": 0.23, "learning_rate": 1.7967107228603085e-05, "loss": 2.2902, "step": 9347 }, { "epoch": 0.23, "learning_rate": 1.7966624723144757e-05, "loss": 2.2412, "step": 9348 }, { "epoch": 0.23, "learning_rate": 1.7966142166912534e-05, "loss": 2.1486, "step": 9349 }, { "epoch": 0.23, "learning_rate": 1.796565955990949e-05, "loss": 2.265, "step": 9350 }, { "epoch": 0.23, "learning_rate": 1.7965176902138698e-05, "loss": 2.0002, "step": 9351 }, { "epoch": 0.23, "learning_rate": 1.7964694193603246e-05, "loss": 2.0455, "step": 9352 }, { "epoch": 0.23, "learning_rate": 1.7964211434306196e-05, "loss": 2.2294, "step": 9353 }, { "epoch": 0.23, "learning_rate": 1.7963728624250633e-05, "loss": 2.375, "step": 9354 }, { "epoch": 0.23, "learning_rate": 1.7963245763439632e-05, "loss": 2.222, "step": 9355 }, { "epoch": 0.23, "learning_rate": 1.796276285187627e-05, "loss": 2.3164, "step": 9356 }, { "epoch": 0.23, "learning_rate": 1.796227988956363e-05, "loss": 2.1723, "step": 9357 }, { "epoch": 0.23, "learning_rate": 1.7961796876504778e-05, "loss": 2.1727, "step": 9358 }, { "epoch": 0.23, "learning_rate": 1.7961313812702803e-05, "loss": 2.1756, "step": 9359 }, { "epoch": 0.23, "learning_rate": 1.796083069816078e-05, "loss": 2.2735, "step": 9360 }, { "epoch": 0.23, "learning_rate": 1.7960347532881793e-05, "loss": 2.357, "step": 9361 }, { "epoch": 0.23, "learning_rate": 1.7959864316868912e-05, "loss": 2.2258, "step": 9362 }, { "epoch": 0.23, "learning_rate": 1.7959381050125222e-05, "loss": 2.0218, "step": 9363 }, { "epoch": 0.23, "learning_rate": 1.7958897732653803e-05, "loss": 2.349, "step": 9364 }, { "epoch": 0.23, "learning_rate": 1.7958414364457737e-05, "loss": 2.1339, "step": 9365 }, { "epoch": 0.23, "learning_rate": 1.7957930945540102e-05, "loss": 2.1032, "step": 9366 }, { "epoch": 0.23, "learning_rate": 1.7957447475903978e-05, "loss": 2.3074, "step": 9367 }, { "epoch": 0.23, "learning_rate": 1.795696395555245e-05, "loss": 1.9548, "step": 9368 }, { "epoch": 0.23, "learning_rate": 1.7956480384488596e-05, "loss": 1.9776, "step": 9369 }, { "epoch": 0.23, "learning_rate": 1.79559967627155e-05, "loss": 2.2155, "step": 9370 }, { "epoch": 0.23, "learning_rate": 1.7955513090236246e-05, "loss": 2.197, "step": 9371 }, { "epoch": 0.23, "learning_rate": 1.795502936705391e-05, "loss": 2.1883, "step": 9372 }, { "epoch": 0.23, "learning_rate": 1.7954545593171585e-05, "loss": 2.1641, "step": 9373 }, { "epoch": 0.23, "learning_rate": 1.7954061768592345e-05, "loss": 2.1308, "step": 9374 }, { "epoch": 0.23, "learning_rate": 1.7953577893319277e-05, "loss": 2.2053, "step": 9375 }, { "epoch": 0.23, "learning_rate": 1.7953093967355465e-05, "loss": 2.2453, "step": 9376 }, { "epoch": 0.23, "learning_rate": 1.7952609990703995e-05, "loss": 2.2938, "step": 9377 }, { "epoch": 0.23, "learning_rate": 1.795212596336795e-05, "loss": 2.1489, "step": 9378 }, { "epoch": 0.23, "learning_rate": 1.7951641885350412e-05, "loss": 2.2123, "step": 9379 }, { "epoch": 0.23, "learning_rate": 1.795115775665447e-05, "loss": 2.2719, "step": 9380 }, { "epoch": 0.23, "learning_rate": 1.795067357728321e-05, "loss": 2.1491, "step": 9381 }, { "epoch": 0.23, "learning_rate": 1.7950189347239716e-05, "loss": 2.1662, "step": 9382 }, { "epoch": 0.23, "learning_rate": 1.7949705066527075e-05, "loss": 2.0324, "step": 9383 }, { "epoch": 0.23, "learning_rate": 1.794922073514837e-05, "loss": 2.1, "step": 9384 }, { "epoch": 0.23, "learning_rate": 1.7948736353106692e-05, "loss": 2.2893, "step": 9385 }, { "epoch": 0.23, "learning_rate": 1.7948251920405128e-05, "loss": 2.3309, "step": 9386 }, { "epoch": 0.23, "learning_rate": 1.7947767437046762e-05, "loss": 2.0375, "step": 9387 }, { "epoch": 0.23, "learning_rate": 1.7947282903034687e-05, "loss": 2.1022, "step": 9388 }, { "epoch": 0.23, "learning_rate": 1.7946798318371987e-05, "loss": 2.1261, "step": 9389 }, { "epoch": 0.23, "learning_rate": 1.7946313683061748e-05, "loss": 2.1261, "step": 9390 }, { "epoch": 0.23, "learning_rate": 1.7945828997107067e-05, "loss": 2.2241, "step": 9391 }, { "epoch": 0.23, "learning_rate": 1.7945344260511027e-05, "loss": 2.2435, "step": 9392 }, { "epoch": 0.23, "learning_rate": 1.7944859473276718e-05, "loss": 2.1079, "step": 9393 }, { "epoch": 0.23, "learning_rate": 1.794437463540723e-05, "loss": 2.2687, "step": 9394 }, { "epoch": 0.23, "learning_rate": 1.7943889746905653e-05, "loss": 2.0507, "step": 9395 }, { "epoch": 0.23, "learning_rate": 1.7943404807775084e-05, "loss": 2.1255, "step": 9396 }, { "epoch": 0.23, "learning_rate": 1.79429198180186e-05, "loss": 2.1997, "step": 9397 }, { "epoch": 0.23, "learning_rate": 1.79424347776393e-05, "loss": 2.1353, "step": 9398 }, { "epoch": 0.23, "learning_rate": 1.7941949686640278e-05, "loss": 2.2039, "step": 9399 }, { "epoch": 0.23, "learning_rate": 1.794146454502462e-05, "loss": 2.3001, "step": 9400 }, { "epoch": 0.23, "learning_rate": 1.7940979352795422e-05, "loss": 2.2318, "step": 9401 }, { "epoch": 0.23, "learning_rate": 1.7940494109955775e-05, "loss": 2.3788, "step": 9402 }, { "epoch": 0.23, "learning_rate": 1.794000881650877e-05, "loss": 2.1431, "step": 9403 }, { "epoch": 0.23, "learning_rate": 1.79395234724575e-05, "loss": 2.1011, "step": 9404 }, { "epoch": 0.23, "learning_rate": 1.7939038077805065e-05, "loss": 2.302, "step": 9405 }, { "epoch": 0.23, "learning_rate": 1.7938552632554545e-05, "loss": 2.1399, "step": 9406 }, { "epoch": 0.23, "learning_rate": 1.793806713670905e-05, "loss": 2.1216, "step": 9407 }, { "epoch": 0.23, "learning_rate": 1.793758159027166e-05, "loss": 2.1736, "step": 9408 }, { "epoch": 0.23, "learning_rate": 1.7937095993245482e-05, "loss": 2.2908, "step": 9409 }, { "epoch": 0.23, "learning_rate": 1.79366103456336e-05, "loss": 2.2471, "step": 9410 }, { "epoch": 0.23, "learning_rate": 1.7936124647439115e-05, "loss": 2.1393, "step": 9411 }, { "epoch": 0.23, "learning_rate": 1.793563889866512e-05, "loss": 2.2814, "step": 9412 }, { "epoch": 0.23, "learning_rate": 1.7935153099314713e-05, "loss": 2.1345, "step": 9413 }, { "epoch": 0.23, "learning_rate": 1.7934667249390992e-05, "loss": 2.2599, "step": 9414 }, { "epoch": 0.23, "learning_rate": 1.7934181348897045e-05, "loss": 2.1842, "step": 9415 }, { "epoch": 0.23, "learning_rate": 1.7933695397835977e-05, "loss": 2.3023, "step": 9416 }, { "epoch": 0.23, "learning_rate": 1.7933209396210887e-05, "loss": 2.3315, "step": 9417 }, { "epoch": 0.23, "learning_rate": 1.7932723344024862e-05, "loss": 2.0818, "step": 9418 }, { "epoch": 0.23, "learning_rate": 1.793223724128101e-05, "loss": 2.3104, "step": 9419 }, { "epoch": 0.23, "learning_rate": 1.7931751087982423e-05, "loss": 2.2217, "step": 9420 }, { "epoch": 0.23, "learning_rate": 1.79312648841322e-05, "loss": 2.05, "step": 9421 }, { "epoch": 0.23, "learning_rate": 1.793077862973345e-05, "loss": 2.1542, "step": 9422 }, { "epoch": 0.23, "learning_rate": 1.7930292324789252e-05, "loss": 2.157, "step": 9423 }, { "epoch": 0.23, "learning_rate": 1.7929805969302722e-05, "loss": 2.3245, "step": 9424 }, { "epoch": 0.23, "learning_rate": 1.7929319563276954e-05, "loss": 1.9949, "step": 9425 }, { "epoch": 0.23, "learning_rate": 1.7928833106715046e-05, "loss": 2.222, "step": 9426 }, { "epoch": 0.23, "learning_rate": 1.7928346599620104e-05, "loss": 2.0979, "step": 9427 }, { "epoch": 0.23, "learning_rate": 1.792786004199522e-05, "loss": 2.2114, "step": 9428 }, { "epoch": 0.23, "learning_rate": 1.7927373433843506e-05, "loss": 2.0234, "step": 9429 }, { "epoch": 0.23, "learning_rate": 1.7926886775168054e-05, "loss": 2.0573, "step": 9430 }, { "epoch": 0.23, "learning_rate": 1.7926400065971973e-05, "loss": 2.1719, "step": 9431 }, { "epoch": 0.23, "learning_rate": 1.792591330625836e-05, "loss": 2.0235, "step": 9432 }, { "epoch": 0.23, "learning_rate": 1.7925426496030318e-05, "loss": 2.0936, "step": 9433 }, { "epoch": 0.23, "learning_rate": 1.792493963529095e-05, "loss": 2.1835, "step": 9434 }, { "epoch": 0.23, "learning_rate": 1.792445272404336e-05, "loss": 2.29, "step": 9435 }, { "epoch": 0.23, "learning_rate": 1.7923965762290646e-05, "loss": 2.1261, "step": 9436 }, { "epoch": 0.23, "learning_rate": 1.792347875003592e-05, "loss": 2.1321, "step": 9437 }, { "epoch": 0.23, "learning_rate": 1.7922991687282282e-05, "loss": 2.0013, "step": 9438 }, { "epoch": 0.23, "learning_rate": 1.7922504574032837e-05, "loss": 2.0621, "step": 9439 }, { "epoch": 0.23, "learning_rate": 1.7922017410290683e-05, "loss": 2.0693, "step": 9440 }, { "epoch": 0.23, "learning_rate": 1.7921530196058935e-05, "loss": 2.1137, "step": 9441 }, { "epoch": 0.23, "learning_rate": 1.7921042931340693e-05, "loss": 2.1571, "step": 9442 }, { "epoch": 0.23, "learning_rate": 1.792055561613906e-05, "loss": 2.1875, "step": 9443 }, { "epoch": 0.23, "learning_rate": 1.7920068250457147e-05, "loss": 2.1595, "step": 9444 }, { "epoch": 0.23, "learning_rate": 1.7919580834298057e-05, "loss": 2.1874, "step": 9445 }, { "epoch": 0.23, "learning_rate": 1.79190933676649e-05, "loss": 2.1121, "step": 9446 }, { "epoch": 0.23, "learning_rate": 1.7918605850560777e-05, "loss": 2.3, "step": 9447 }, { "epoch": 0.23, "learning_rate": 1.79181182829888e-05, "loss": 2.3012, "step": 9448 }, { "epoch": 0.23, "learning_rate": 1.7917630664952077e-05, "loss": 2.1037, "step": 9449 }, { "epoch": 0.23, "learning_rate": 1.7917142996453713e-05, "loss": 2.1086, "step": 9450 }, { "epoch": 0.23, "learning_rate": 1.791665527749681e-05, "loss": 2.1533, "step": 9451 }, { "epoch": 0.23, "learning_rate": 1.7916167508084486e-05, "loss": 2.1194, "step": 9452 }, { "epoch": 0.23, "learning_rate": 1.791567968821985e-05, "loss": 2.1508, "step": 9453 }, { "epoch": 0.23, "learning_rate": 1.7915191817906005e-05, "loss": 2.0721, "step": 9454 }, { "epoch": 0.23, "learning_rate": 1.7914703897146067e-05, "loss": 2.2194, "step": 9455 }, { "epoch": 0.23, "learning_rate": 1.791421592594314e-05, "loss": 2.1539, "step": 9456 }, { "epoch": 0.23, "learning_rate": 1.791372790430033e-05, "loss": 2.2505, "step": 9457 }, { "epoch": 0.23, "learning_rate": 1.791323983222076e-05, "loss": 2.1973, "step": 9458 }, { "epoch": 0.23, "learning_rate": 1.791275170970753e-05, "loss": 2.1932, "step": 9459 }, { "epoch": 0.23, "learning_rate": 1.791226353676376e-05, "loss": 2.1453, "step": 9460 }, { "epoch": 0.23, "learning_rate": 1.791177531339255e-05, "loss": 2.251, "step": 9461 }, { "epoch": 0.23, "learning_rate": 1.791128703959702e-05, "loss": 2.1622, "step": 9462 }, { "epoch": 0.23, "learning_rate": 1.791079871538028e-05, "loss": 2.0591, "step": 9463 }, { "epoch": 0.23, "learning_rate": 1.7910310340745438e-05, "loss": 2.192, "step": 9464 }, { "epoch": 0.23, "learning_rate": 1.7909821915695613e-05, "loss": 2.1539, "step": 9465 }, { "epoch": 0.23, "learning_rate": 1.7909333440233917e-05, "loss": 2.256, "step": 9466 }, { "epoch": 0.23, "learning_rate": 1.7908844914363458e-05, "loss": 2.4399, "step": 9467 }, { "epoch": 0.23, "learning_rate": 1.7908356338087352e-05, "loss": 2.0754, "step": 9468 }, { "epoch": 0.23, "learning_rate": 1.790786771140872e-05, "loss": 2.209, "step": 9469 }, { "epoch": 0.23, "learning_rate": 1.7907379034330662e-05, "loss": 2.1292, "step": 9470 }, { "epoch": 0.23, "learning_rate": 1.7906890306856305e-05, "loss": 2.241, "step": 9471 }, { "epoch": 0.23, "learning_rate": 1.790640152898876e-05, "loss": 2.056, "step": 9472 }, { "epoch": 0.23, "learning_rate": 1.7905912700731143e-05, "loss": 2.3311, "step": 9473 }, { "epoch": 0.23, "learning_rate": 1.7905423822086562e-05, "loss": 2.1935, "step": 9474 }, { "epoch": 0.23, "learning_rate": 1.7904934893058144e-05, "loss": 2.2169, "step": 9475 }, { "epoch": 0.23, "learning_rate": 1.7904445913649e-05, "loss": 2.2091, "step": 9476 }, { "epoch": 0.23, "learning_rate": 1.7903956883862244e-05, "loss": 2.0219, "step": 9477 }, { "epoch": 0.23, "learning_rate": 1.7903467803700993e-05, "loss": 2.2347, "step": 9478 }, { "epoch": 0.23, "learning_rate": 1.7902978673168367e-05, "loss": 2.236, "step": 9479 }, { "epoch": 0.23, "learning_rate": 1.7902489492267485e-05, "loss": 2.147, "step": 9480 }, { "epoch": 0.23, "learning_rate": 1.790200026100146e-05, "loss": 2.1861, "step": 9481 }, { "epoch": 0.23, "learning_rate": 1.7901510979373414e-05, "loss": 2.2814, "step": 9482 }, { "epoch": 0.23, "learning_rate": 1.790102164738646e-05, "loss": 2.1881, "step": 9483 }, { "epoch": 0.23, "learning_rate": 1.790053226504372e-05, "loss": 2.166, "step": 9484 }, { "epoch": 0.23, "learning_rate": 1.790004283234832e-05, "loss": 2.229, "step": 9485 }, { "epoch": 0.23, "learning_rate": 1.7899553349303365e-05, "loss": 2.3049, "step": 9486 }, { "epoch": 0.23, "learning_rate": 1.7899063815911986e-05, "loss": 2.1645, "step": 9487 }, { "epoch": 0.23, "learning_rate": 1.7898574232177298e-05, "loss": 2.3368, "step": 9488 }, { "epoch": 0.23, "learning_rate": 1.7898084598102422e-05, "loss": 2.1023, "step": 9489 }, { "epoch": 0.23, "learning_rate": 1.789759491369048e-05, "loss": 2.2758, "step": 9490 }, { "epoch": 0.23, "learning_rate": 1.7897105178944593e-05, "loss": 2.1307, "step": 9491 }, { "epoch": 0.23, "learning_rate": 1.789661539386788e-05, "loss": 2.1589, "step": 9492 }, { "epoch": 0.23, "learning_rate": 1.789612555846346e-05, "loss": 2.1654, "step": 9493 }, { "epoch": 0.23, "learning_rate": 1.7895635672734462e-05, "loss": 2.1761, "step": 9494 }, { "epoch": 0.23, "learning_rate": 1.7895145736684006e-05, "loss": 2.0745, "step": 9495 }, { "epoch": 0.23, "learning_rate": 1.7894655750315212e-05, "loss": 2.1386, "step": 9496 }, { "epoch": 0.23, "learning_rate": 1.78941657136312e-05, "loss": 2.0989, "step": 9497 }, { "epoch": 0.23, "learning_rate": 1.7893675626635106e-05, "loss": 2.232, "step": 9498 }, { "epoch": 0.23, "learning_rate": 1.7893185489330038e-05, "loss": 2.0805, "step": 9499 }, { "epoch": 0.23, "learning_rate": 1.7892695301719128e-05, "loss": 2.1782, "step": 9500 }, { "epoch": 0.23, "learning_rate": 1.78922050638055e-05, "loss": 2.2279, "step": 9501 }, { "epoch": 0.23, "learning_rate": 1.7891714775592276e-05, "loss": 2.2516, "step": 9502 }, { "epoch": 0.23, "learning_rate": 1.7891224437082582e-05, "loss": 2.0291, "step": 9503 }, { "epoch": 0.23, "learning_rate": 1.789073404827954e-05, "loss": 2.1913, "step": 9504 }, { "epoch": 0.23, "learning_rate": 1.7890243609186283e-05, "loss": 2.1334, "step": 9505 }, { "epoch": 0.23, "learning_rate": 1.7889753119805933e-05, "loss": 2.019, "step": 9506 }, { "epoch": 0.23, "learning_rate": 1.7889262580141615e-05, "loss": 2.1597, "step": 9507 }, { "epoch": 0.23, "learning_rate": 1.7888771990196448e-05, "loss": 2.2249, "step": 9508 }, { "epoch": 0.23, "learning_rate": 1.7888281349973574e-05, "loss": 2.2904, "step": 9509 }, { "epoch": 0.23, "learning_rate": 1.788779065947611e-05, "loss": 2.1738, "step": 9510 }, { "epoch": 0.23, "learning_rate": 1.7887299918707182e-05, "loss": 2.1632, "step": 9511 }, { "epoch": 0.23, "learning_rate": 1.7886809127669926e-05, "loss": 2.1841, "step": 9512 }, { "epoch": 0.23, "learning_rate": 1.7886318286367465e-05, "loss": 2.1391, "step": 9513 }, { "epoch": 0.23, "learning_rate": 1.7885827394802925e-05, "loss": 2.267, "step": 9514 }, { "epoch": 0.23, "learning_rate": 1.788533645297944e-05, "loss": 2.1455, "step": 9515 }, { "epoch": 0.23, "learning_rate": 1.7884845460900132e-05, "loss": 2.2972, "step": 9516 }, { "epoch": 0.23, "learning_rate": 1.7884354418568136e-05, "loss": 2.2055, "step": 9517 }, { "epoch": 0.23, "learning_rate": 1.788386332598658e-05, "loss": 2.2911, "step": 9518 }, { "epoch": 0.23, "learning_rate": 1.7883372183158595e-05, "loss": 2.3754, "step": 9519 }, { "epoch": 0.23, "learning_rate": 1.788288099008731e-05, "loss": 2.2501, "step": 9520 }, { "epoch": 0.23, "learning_rate": 1.7882389746775854e-05, "loss": 2.1845, "step": 9521 }, { "epoch": 0.23, "learning_rate": 1.788189845322736e-05, "loss": 2.1569, "step": 9522 }, { "epoch": 0.23, "learning_rate": 1.7881407109444958e-05, "loss": 2.2098, "step": 9523 }, { "epoch": 0.23, "learning_rate": 1.788091571543178e-05, "loss": 2.259, "step": 9524 }, { "epoch": 0.23, "learning_rate": 1.7880424271190962e-05, "loss": 2.2705, "step": 9525 }, { "epoch": 0.23, "learning_rate": 1.7879932776725627e-05, "loss": 2.2206, "step": 9526 }, { "epoch": 0.23, "learning_rate": 1.7879441232038912e-05, "loss": 2.028, "step": 9527 }, { "epoch": 0.23, "learning_rate": 1.7878949637133954e-05, "loss": 2.2868, "step": 9528 }, { "epoch": 0.23, "learning_rate": 1.787845799201388e-05, "loss": 2.1706, "step": 9529 }, { "epoch": 0.23, "learning_rate": 1.7877966296681825e-05, "loss": 2.0982, "step": 9530 }, { "epoch": 0.23, "learning_rate": 1.7877474551140925e-05, "loss": 2.1168, "step": 9531 }, { "epoch": 0.23, "learning_rate": 1.7876982755394312e-05, "loss": 2.1715, "step": 9532 }, { "epoch": 0.23, "learning_rate": 1.787649090944512e-05, "loss": 2.0822, "step": 9533 }, { "epoch": 0.24, "learning_rate": 1.7875999013296486e-05, "loss": 2.0239, "step": 9534 }, { "epoch": 0.24, "learning_rate": 1.7875507066951542e-05, "loss": 2.0988, "step": 9535 }, { "epoch": 0.24, "learning_rate": 1.7875015070413427e-05, "loss": 2.2183, "step": 9536 }, { "epoch": 0.24, "learning_rate": 1.7874523023685273e-05, "loss": 2.0743, "step": 9537 }, { "epoch": 0.24, "learning_rate": 1.7874030926770217e-05, "loss": 2.1591, "step": 9538 }, { "epoch": 0.24, "learning_rate": 1.7873538779671397e-05, "loss": 2.2762, "step": 9539 }, { "epoch": 0.24, "learning_rate": 1.787304658239195e-05, "loss": 2.1985, "step": 9540 }, { "epoch": 0.24, "learning_rate": 1.7872554334935008e-05, "loss": 2.0135, "step": 9541 }, { "epoch": 0.24, "learning_rate": 1.787206203730371e-05, "loss": 2.1794, "step": 9542 }, { "epoch": 0.24, "learning_rate": 1.7871569689501203e-05, "loss": 2.0371, "step": 9543 }, { "epoch": 0.24, "learning_rate": 1.787107729153061e-05, "loss": 2.184, "step": 9544 }, { "epoch": 0.24, "learning_rate": 1.787058484339508e-05, "loss": 1.9195, "step": 9545 }, { "epoch": 0.24, "learning_rate": 1.787009234509775e-05, "loss": 2.2138, "step": 9546 }, { "epoch": 0.24, "learning_rate": 1.786959979664175e-05, "loss": 2.0468, "step": 9547 }, { "epoch": 0.24, "learning_rate": 1.786910719803023e-05, "loss": 2.4102, "step": 9548 }, { "epoch": 0.24, "learning_rate": 1.7868614549266323e-05, "loss": 2.0866, "step": 9549 }, { "epoch": 0.24, "learning_rate": 1.7868121850353174e-05, "loss": 2.1236, "step": 9550 }, { "epoch": 0.24, "learning_rate": 1.786762910129392e-05, "loss": 2.137, "step": 9551 }, { "epoch": 0.24, "learning_rate": 1.7867136302091697e-05, "loss": 2.1469, "step": 9552 }, { "epoch": 0.24, "learning_rate": 1.7866643452749658e-05, "loss": 2.2279, "step": 9553 }, { "epoch": 0.24, "learning_rate": 1.7866150553270932e-05, "loss": 2.1044, "step": 9554 }, { "epoch": 0.24, "learning_rate": 1.7865657603658663e-05, "loss": 2.1381, "step": 9555 }, { "epoch": 0.24, "learning_rate": 1.7865164603916e-05, "loss": 2.2017, "step": 9556 }, { "epoch": 0.24, "learning_rate": 1.7864671554046076e-05, "loss": 2.2739, "step": 9557 }, { "epoch": 0.24, "learning_rate": 1.786417845405204e-05, "loss": 2.323, "step": 9558 }, { "epoch": 0.24, "learning_rate": 1.786368530393703e-05, "loss": 2.166, "step": 9559 }, { "epoch": 0.24, "learning_rate": 1.7863192103704192e-05, "loss": 2.2887, "step": 9560 }, { "epoch": 0.24, "learning_rate": 1.7862698853356667e-05, "loss": 2.2249, "step": 9561 }, { "epoch": 0.24, "learning_rate": 1.78622055528976e-05, "loss": 2.1158, "step": 9562 }, { "epoch": 0.24, "learning_rate": 1.7861712202330135e-05, "loss": 2.1731, "step": 9563 }, { "epoch": 0.24, "learning_rate": 1.786121880165742e-05, "loss": 2.1066, "step": 9564 }, { "epoch": 0.24, "learning_rate": 1.7860725350882587e-05, "loss": 2.2023, "step": 9565 }, { "epoch": 0.24, "learning_rate": 1.7860231850008795e-05, "loss": 2.2125, "step": 9566 }, { "epoch": 0.24, "learning_rate": 1.7859738299039183e-05, "loss": 2.1716, "step": 9567 }, { "epoch": 0.24, "learning_rate": 1.7859244697976897e-05, "loss": 2.1643, "step": 9568 }, { "epoch": 0.24, "learning_rate": 1.7858751046825088e-05, "loss": 2.2075, "step": 9569 }, { "epoch": 0.24, "learning_rate": 1.7858257345586892e-05, "loss": 2.0694, "step": 9570 }, { "epoch": 0.24, "learning_rate": 1.7857763594265464e-05, "loss": 2.2361, "step": 9571 }, { "epoch": 0.24, "learning_rate": 1.7857269792863945e-05, "loss": 2.0899, "step": 9572 }, { "epoch": 0.24, "learning_rate": 1.7856775941385485e-05, "loss": 2.2243, "step": 9573 }, { "epoch": 0.24, "learning_rate": 1.7856282039833233e-05, "loss": 2.0068, "step": 9574 }, { "epoch": 0.24, "learning_rate": 1.7855788088210336e-05, "loss": 2.2169, "step": 9575 }, { "epoch": 0.24, "learning_rate": 1.7855294086519942e-05, "loss": 2.0286, "step": 9576 }, { "epoch": 0.24, "learning_rate": 1.7854800034765197e-05, "loss": 2.2181, "step": 9577 }, { "epoch": 0.24, "learning_rate": 1.785430593294925e-05, "loss": 2.2425, "step": 9578 }, { "epoch": 0.24, "learning_rate": 1.7853811781075256e-05, "loss": 2.0284, "step": 9579 }, { "epoch": 0.24, "learning_rate": 1.7853317579146354e-05, "loss": 2.1605, "step": 9580 }, { "epoch": 0.24, "learning_rate": 1.7852823327165706e-05, "loss": 2.2978, "step": 9581 }, { "epoch": 0.24, "learning_rate": 1.7852329025136453e-05, "loss": 2.2646, "step": 9582 }, { "epoch": 0.24, "learning_rate": 1.785183467306175e-05, "loss": 2.1511, "step": 9583 }, { "epoch": 0.24, "learning_rate": 1.785134027094474e-05, "loss": 2.1311, "step": 9584 }, { "epoch": 0.24, "learning_rate": 1.7850845818788585e-05, "loss": 2.0593, "step": 9585 }, { "epoch": 0.24, "learning_rate": 1.785035131659643e-05, "loss": 1.9978, "step": 9586 }, { "epoch": 0.24, "learning_rate": 1.784985676437143e-05, "loss": 2.3249, "step": 9587 }, { "epoch": 0.24, "learning_rate": 1.7849362162116727e-05, "loss": 2.0584, "step": 9588 }, { "epoch": 0.24, "learning_rate": 1.784886750983549e-05, "loss": 2.1997, "step": 9589 }, { "epoch": 0.24, "learning_rate": 1.7848372807530858e-05, "loss": 2.0948, "step": 9590 }, { "epoch": 0.24, "learning_rate": 1.7847878055205988e-05, "loss": 2.0809, "step": 9591 }, { "epoch": 0.24, "learning_rate": 1.7847383252864036e-05, "loss": 2.0845, "step": 9592 }, { "epoch": 0.24, "learning_rate": 1.784688840050815e-05, "loss": 2.1878, "step": 9593 }, { "epoch": 0.24, "learning_rate": 1.784639349814149e-05, "loss": 2.2393, "step": 9594 }, { "epoch": 0.24, "learning_rate": 1.7845898545767207e-05, "loss": 2.0315, "step": 9595 }, { "epoch": 0.24, "learning_rate": 1.7845403543388453e-05, "loss": 2.0554, "step": 9596 }, { "epoch": 0.24, "learning_rate": 1.7844908491008385e-05, "loss": 2.1155, "step": 9597 }, { "epoch": 0.24, "learning_rate": 1.784441338863016e-05, "loss": 2.2168, "step": 9598 }, { "epoch": 0.24, "learning_rate": 1.7843918236256933e-05, "loss": 2.0917, "step": 9599 }, { "epoch": 0.24, "learning_rate": 1.784342303389186e-05, "loss": 2.106, "step": 9600 }, { "epoch": 0.24, "learning_rate": 1.7842927781538095e-05, "loss": 2.3021, "step": 9601 }, { "epoch": 0.24, "learning_rate": 1.7842432479198795e-05, "loss": 2.3328, "step": 9602 }, { "epoch": 0.24, "learning_rate": 1.7841937126877117e-05, "loss": 1.9915, "step": 9603 }, { "epoch": 0.24, "learning_rate": 1.7841441724576215e-05, "loss": 2.2323, "step": 9604 }, { "epoch": 0.24, "learning_rate": 1.7840946272299256e-05, "loss": 2.0962, "step": 9605 }, { "epoch": 0.24, "learning_rate": 1.7840450770049385e-05, "loss": 2.2719, "step": 9606 }, { "epoch": 0.24, "learning_rate": 1.783995521782977e-05, "loss": 2.161, "step": 9607 }, { "epoch": 0.24, "learning_rate": 1.7839459615643565e-05, "loss": 2.1279, "step": 9608 }, { "epoch": 0.24, "learning_rate": 1.7838963963493925e-05, "loss": 2.1513, "step": 9609 }, { "epoch": 0.24, "learning_rate": 1.7838468261384018e-05, "loss": 2.1215, "step": 9610 }, { "epoch": 0.24, "learning_rate": 1.7837972509316996e-05, "loss": 2.067, "step": 9611 }, { "epoch": 0.24, "learning_rate": 1.7837476707296022e-05, "loss": 2.0501, "step": 9612 }, { "epoch": 0.24, "learning_rate": 1.7836980855324252e-05, "loss": 2.1212, "step": 9613 }, { "epoch": 0.24, "learning_rate": 1.783648495340485e-05, "loss": 2.2242, "step": 9614 }, { "epoch": 0.24, "learning_rate": 1.7835989001540976e-05, "loss": 2.3416, "step": 9615 }, { "epoch": 0.24, "learning_rate": 1.7835492999735793e-05, "loss": 2.0977, "step": 9616 }, { "epoch": 0.24, "learning_rate": 1.7834996947992458e-05, "loss": 2.0329, "step": 9617 }, { "epoch": 0.24, "learning_rate": 1.7834500846314135e-05, "loss": 2.2844, "step": 9618 }, { "epoch": 0.24, "learning_rate": 1.7834004694703982e-05, "loss": 2.1965, "step": 9619 }, { "epoch": 0.24, "learning_rate": 1.7833508493165167e-05, "loss": 2.0147, "step": 9620 }, { "epoch": 0.24, "learning_rate": 1.7833012241700848e-05, "loss": 2.1862, "step": 9621 }, { "epoch": 0.24, "learning_rate": 1.783251594031419e-05, "loss": 2.2196, "step": 9622 }, { "epoch": 0.24, "learning_rate": 1.7832019589008355e-05, "loss": 2.1545, "step": 9623 }, { "epoch": 0.24, "learning_rate": 1.7831523187786503e-05, "loss": 2.1921, "step": 9624 }, { "epoch": 0.24, "learning_rate": 1.7831026736651805e-05, "loss": 2.008, "step": 9625 }, { "epoch": 0.24, "learning_rate": 1.7830530235607422e-05, "loss": 1.9805, "step": 9626 }, { "epoch": 0.24, "learning_rate": 1.783003368465652e-05, "loss": 2.3204, "step": 9627 }, { "epoch": 0.24, "learning_rate": 1.782953708380226e-05, "loss": 2.2016, "step": 9628 }, { "epoch": 0.24, "learning_rate": 1.7829040433047808e-05, "loss": 2.2863, "step": 9629 }, { "epoch": 0.24, "learning_rate": 1.7828543732396326e-05, "loss": 2.2131, "step": 9630 }, { "epoch": 0.24, "learning_rate": 1.782804698185099e-05, "loss": 2.0999, "step": 9631 }, { "epoch": 0.24, "learning_rate": 1.7827550181414954e-05, "loss": 2.1813, "step": 9632 }, { "epoch": 0.24, "learning_rate": 1.782705333109139e-05, "loss": 2.072, "step": 9633 }, { "epoch": 0.24, "learning_rate": 1.7826556430883467e-05, "loss": 2.391, "step": 9634 }, { "epoch": 0.24, "learning_rate": 1.7826059480794352e-05, "loss": 2.0575, "step": 9635 }, { "epoch": 0.24, "learning_rate": 1.7825562480827204e-05, "loss": 2.1697, "step": 9636 }, { "epoch": 0.24, "learning_rate": 1.78250654309852e-05, "loss": 2.2577, "step": 9637 }, { "epoch": 0.24, "learning_rate": 1.7824568331271504e-05, "loss": 2.24, "step": 9638 }, { "epoch": 0.24, "learning_rate": 1.782407118168928e-05, "loss": 2.3395, "step": 9639 }, { "epoch": 0.24, "learning_rate": 1.7823573982241706e-05, "loss": 2.0807, "step": 9640 }, { "epoch": 0.24, "learning_rate": 1.782307673293194e-05, "loss": 2.1889, "step": 9641 }, { "epoch": 0.24, "learning_rate": 1.782257943376316e-05, "loss": 1.9846, "step": 9642 }, { "epoch": 0.24, "learning_rate": 1.7822082084738532e-05, "loss": 2.1692, "step": 9643 }, { "epoch": 0.24, "learning_rate": 1.7821584685861224e-05, "loss": 2.2469, "step": 9644 }, { "epoch": 0.24, "learning_rate": 1.7821087237134408e-05, "loss": 2.0286, "step": 9645 }, { "epoch": 0.24, "learning_rate": 1.7820589738561254e-05, "loss": 2.1406, "step": 9646 }, { "epoch": 0.24, "learning_rate": 1.7820092190144935e-05, "loss": 2.0418, "step": 9647 }, { "epoch": 0.24, "learning_rate": 1.7819594591888617e-05, "loss": 2.1734, "step": 9648 }, { "epoch": 0.24, "learning_rate": 1.7819096943795476e-05, "loss": 2.1858, "step": 9649 }, { "epoch": 0.24, "learning_rate": 1.7818599245868684e-05, "loss": 2.2218, "step": 9650 }, { "epoch": 0.24, "learning_rate": 1.781810149811141e-05, "loss": 2.3482, "step": 9651 }, { "epoch": 0.24, "learning_rate": 1.7817603700526824e-05, "loss": 2.0184, "step": 9652 }, { "epoch": 0.24, "learning_rate": 1.7817105853118103e-05, "loss": 2.1281, "step": 9653 }, { "epoch": 0.24, "learning_rate": 1.781660795588842e-05, "loss": 2.1548, "step": 9654 }, { "epoch": 0.24, "learning_rate": 1.781611000884095e-05, "loss": 2.1294, "step": 9655 }, { "epoch": 0.24, "learning_rate": 1.781561201197886e-05, "loss": 2.2453, "step": 9656 }, { "epoch": 0.24, "learning_rate": 1.7815113965305327e-05, "loss": 1.9446, "step": 9657 }, { "epoch": 0.24, "learning_rate": 1.781461586882353e-05, "loss": 2.0812, "step": 9658 }, { "epoch": 0.24, "learning_rate": 1.7814117722536637e-05, "loss": 2.0733, "step": 9659 }, { "epoch": 0.24, "learning_rate": 1.7813619526447823e-05, "loss": 2.1338, "step": 9660 }, { "epoch": 0.24, "learning_rate": 1.7813121280560268e-05, "loss": 2.168, "step": 9661 }, { "epoch": 0.24, "learning_rate": 1.7812622984877144e-05, "loss": 2.2551, "step": 9662 }, { "epoch": 0.24, "learning_rate": 1.781212463940163e-05, "loss": 2.1081, "step": 9663 }, { "epoch": 0.24, "learning_rate": 1.7811626244136896e-05, "loss": 2.129, "step": 9664 }, { "epoch": 0.24, "learning_rate": 1.7811127799086122e-05, "loss": 1.97, "step": 9665 }, { "epoch": 0.24, "learning_rate": 1.7810629304252488e-05, "loss": 2.2436, "step": 9666 }, { "epoch": 0.24, "learning_rate": 1.7810130759639167e-05, "loss": 2.0109, "step": 9667 }, { "epoch": 0.24, "learning_rate": 1.7809632165249335e-05, "loss": 2.2422, "step": 9668 }, { "epoch": 0.24, "learning_rate": 1.7809133521086177e-05, "loss": 2.1859, "step": 9669 }, { "epoch": 0.24, "learning_rate": 1.780863482715286e-05, "loss": 2.2754, "step": 9670 }, { "epoch": 0.24, "learning_rate": 1.7808136083452578e-05, "loss": 2.2811, "step": 9671 }, { "epoch": 0.24, "learning_rate": 1.7807637289988492e-05, "loss": 2.0838, "step": 9672 }, { "epoch": 0.24, "learning_rate": 1.780713844676379e-05, "loss": 2.3259, "step": 9673 }, { "epoch": 0.24, "learning_rate": 1.7806639553781654e-05, "loss": 2.1969, "step": 9674 }, { "epoch": 0.24, "learning_rate": 1.7806140611045257e-05, "loss": 2.0906, "step": 9675 }, { "epoch": 0.24, "learning_rate": 1.780564161855778e-05, "loss": 2.1566, "step": 9676 }, { "epoch": 0.24, "learning_rate": 1.7805142576322408e-05, "loss": 2.1244, "step": 9677 }, { "epoch": 0.24, "learning_rate": 1.780464348434232e-05, "loss": 2.071, "step": 9678 }, { "epoch": 0.24, "learning_rate": 1.7804144342620695e-05, "loss": 2.358, "step": 9679 }, { "epoch": 0.24, "learning_rate": 1.7803645151160714e-05, "loss": 2.1987, "step": 9680 }, { "epoch": 0.24, "learning_rate": 1.780314590996556e-05, "loss": 2.1918, "step": 9681 }, { "epoch": 0.24, "learning_rate": 1.780264661903841e-05, "loss": 2.1644, "step": 9682 }, { "epoch": 0.24, "learning_rate": 1.7802147278382456e-05, "loss": 2.2891, "step": 9683 }, { "epoch": 0.24, "learning_rate": 1.780164788800087e-05, "loss": 2.167, "step": 9684 }, { "epoch": 0.24, "learning_rate": 1.7801148447896843e-05, "loss": 1.9706, "step": 9685 }, { "epoch": 0.24, "learning_rate": 1.7800648958073554e-05, "loss": 2.1161, "step": 9686 }, { "epoch": 0.24, "learning_rate": 1.7800149418534185e-05, "loss": 2.0743, "step": 9687 }, { "epoch": 0.24, "learning_rate": 1.7799649829281924e-05, "loss": 2.242, "step": 9688 }, { "epoch": 0.24, "learning_rate": 1.7799150190319952e-05, "loss": 2.1149, "step": 9689 }, { "epoch": 0.24, "learning_rate": 1.7798650501651454e-05, "loss": 2.25, "step": 9690 }, { "epoch": 0.24, "learning_rate": 1.7798150763279615e-05, "loss": 2.2954, "step": 9691 }, { "epoch": 0.24, "learning_rate": 1.779765097520762e-05, "loss": 2.2769, "step": 9692 }, { "epoch": 0.24, "learning_rate": 1.7797151137438654e-05, "loss": 2.1595, "step": 9693 }, { "epoch": 0.24, "learning_rate": 1.7796651249975904e-05, "loss": 2.2145, "step": 9694 }, { "epoch": 0.24, "learning_rate": 1.7796151312822556e-05, "loss": 2.257, "step": 9695 }, { "epoch": 0.24, "learning_rate": 1.779565132598179e-05, "loss": 2.3355, "step": 9696 }, { "epoch": 0.24, "learning_rate": 1.77951512894568e-05, "loss": 2.232, "step": 9697 }, { "epoch": 0.24, "learning_rate": 1.7794651203250774e-05, "loss": 1.9958, "step": 9698 }, { "epoch": 0.24, "learning_rate": 1.779415106736689e-05, "loss": 2.2684, "step": 9699 }, { "epoch": 0.24, "learning_rate": 1.7793650881808344e-05, "loss": 2.2446, "step": 9700 }, { "epoch": 0.24, "learning_rate": 1.7793150646578322e-05, "loss": 2.0382, "step": 9701 }, { "epoch": 0.24, "learning_rate": 1.779265036168001e-05, "loss": 2.1346, "step": 9702 }, { "epoch": 0.24, "learning_rate": 1.7792150027116596e-05, "loss": 2.0748, "step": 9703 }, { "epoch": 0.24, "learning_rate": 1.779164964289127e-05, "loss": 2.2976, "step": 9704 }, { "epoch": 0.24, "learning_rate": 1.7791149209007227e-05, "loss": 2.0572, "step": 9705 }, { "epoch": 0.24, "learning_rate": 1.7790648725467646e-05, "loss": 2.1716, "step": 9706 }, { "epoch": 0.24, "learning_rate": 1.7790148192275722e-05, "loss": 2.3047, "step": 9707 }, { "epoch": 0.24, "learning_rate": 1.7789647609434647e-05, "loss": 2.0546, "step": 9708 }, { "epoch": 0.24, "learning_rate": 1.778914697694761e-05, "loss": 2.1556, "step": 9709 }, { "epoch": 0.24, "learning_rate": 1.7788646294817797e-05, "loss": 1.8626, "step": 9710 }, { "epoch": 0.24, "learning_rate": 1.7788145563048407e-05, "loss": 2.2316, "step": 9711 }, { "epoch": 0.24, "learning_rate": 1.7787644781642626e-05, "loss": 2.3817, "step": 9712 }, { "epoch": 0.24, "learning_rate": 1.7787143950603646e-05, "loss": 2.3208, "step": 9713 }, { "epoch": 0.24, "learning_rate": 1.7786643069934663e-05, "loss": 2.1188, "step": 9714 }, { "epoch": 0.24, "learning_rate": 1.778614213963886e-05, "loss": 2.0056, "step": 9715 }, { "epoch": 0.24, "learning_rate": 1.778564115971944e-05, "loss": 2.1685, "step": 9716 }, { "epoch": 0.24, "learning_rate": 1.7785140130179592e-05, "loss": 2.0465, "step": 9717 }, { "epoch": 0.24, "learning_rate": 1.7784639051022506e-05, "loss": 2.2208, "step": 9718 }, { "epoch": 0.24, "learning_rate": 1.778413792225138e-05, "loss": 2.1421, "step": 9719 }, { "epoch": 0.24, "learning_rate": 1.778363674386941e-05, "loss": 2.4492, "step": 9720 }, { "epoch": 0.24, "learning_rate": 1.778313551587978e-05, "loss": 2.1054, "step": 9721 }, { "epoch": 0.24, "learning_rate": 1.778263423828569e-05, "loss": 2.2675, "step": 9722 }, { "epoch": 0.24, "learning_rate": 1.7782132911090344e-05, "loss": 2.3043, "step": 9723 }, { "epoch": 0.24, "learning_rate": 1.7781631534296925e-05, "loss": 2.2377, "step": 9724 }, { "epoch": 0.24, "learning_rate": 1.778113010790863e-05, "loss": 2.0212, "step": 9725 }, { "epoch": 0.24, "learning_rate": 1.7780628631928656e-05, "loss": 2.2606, "step": 9726 }, { "epoch": 0.24, "learning_rate": 1.7780127106360203e-05, "loss": 2.213, "step": 9727 }, { "epoch": 0.24, "learning_rate": 1.7779625531206464e-05, "loss": 2.2942, "step": 9728 }, { "epoch": 0.24, "learning_rate": 1.7779123906470635e-05, "loss": 2.2221, "step": 9729 }, { "epoch": 0.24, "learning_rate": 1.7778622232155918e-05, "loss": 2.4328, "step": 9730 }, { "epoch": 0.24, "learning_rate": 1.7778120508265505e-05, "loss": 2.0852, "step": 9731 }, { "epoch": 0.24, "learning_rate": 1.7777618734802593e-05, "loss": 2.2075, "step": 9732 }, { "epoch": 0.24, "learning_rate": 1.777711691177038e-05, "loss": 2.3008, "step": 9733 }, { "epoch": 0.24, "learning_rate": 1.7776615039172075e-05, "loss": 2.2275, "step": 9734 }, { "epoch": 0.24, "learning_rate": 1.777611311701086e-05, "loss": 2.1194, "step": 9735 }, { "epoch": 0.24, "learning_rate": 1.7775611145289945e-05, "loss": 2.1552, "step": 9736 }, { "epoch": 0.24, "learning_rate": 1.7775109124012525e-05, "loss": 2.3363, "step": 9737 }, { "epoch": 0.24, "learning_rate": 1.77746070531818e-05, "loss": 2.0551, "step": 9738 }, { "epoch": 0.24, "learning_rate": 1.7774104932800977e-05, "loss": 2.1339, "step": 9739 }, { "epoch": 0.24, "learning_rate": 1.7773602762873242e-05, "loss": 2.0113, "step": 9740 }, { "epoch": 0.24, "learning_rate": 1.7773100543401805e-05, "loss": 2.0082, "step": 9741 }, { "epoch": 0.24, "learning_rate": 1.7772598274389867e-05, "loss": 2.2972, "step": 9742 }, { "epoch": 0.24, "learning_rate": 1.7772095955840626e-05, "loss": 2.1559, "step": 9743 }, { "epoch": 0.24, "learning_rate": 1.7771593587757286e-05, "loss": 2.2007, "step": 9744 }, { "epoch": 0.24, "learning_rate": 1.7771091170143046e-05, "loss": 2.0911, "step": 9745 }, { "epoch": 0.24, "learning_rate": 1.777058870300111e-05, "loss": 2.1357, "step": 9746 }, { "epoch": 0.24, "learning_rate": 1.7770086186334678e-05, "loss": 2.2618, "step": 9747 }, { "epoch": 0.24, "learning_rate": 1.7769583620146956e-05, "loss": 2.226, "step": 9748 }, { "epoch": 0.24, "learning_rate": 1.7769081004441147e-05, "loss": 2.1003, "step": 9749 }, { "epoch": 0.24, "learning_rate": 1.776857833922045e-05, "loss": 2.0496, "step": 9750 }, { "epoch": 0.24, "learning_rate": 1.7768075624488076e-05, "loss": 2.1805, "step": 9751 }, { "epoch": 0.24, "learning_rate": 1.776757286024722e-05, "loss": 2.0744, "step": 9752 }, { "epoch": 0.24, "learning_rate": 1.7767070046501095e-05, "loss": 2.1853, "step": 9753 }, { "epoch": 0.24, "learning_rate": 1.7766567183252895e-05, "loss": 2.124, "step": 9754 }, { "epoch": 0.24, "learning_rate": 1.7766064270505837e-05, "loss": 2.1313, "step": 9755 }, { "epoch": 0.24, "learning_rate": 1.776556130826312e-05, "loss": 2.271, "step": 9756 }, { "epoch": 0.24, "learning_rate": 1.7765058296527948e-05, "loss": 2.1388, "step": 9757 }, { "epoch": 0.24, "learning_rate": 1.776455523530353e-05, "loss": 2.1018, "step": 9758 }, { "epoch": 0.24, "learning_rate": 1.7764052124593068e-05, "loss": 2.1741, "step": 9759 }, { "epoch": 0.24, "learning_rate": 1.7763548964399775e-05, "loss": 2.4006, "step": 9760 }, { "epoch": 0.24, "learning_rate": 1.7763045754726853e-05, "loss": 2.4441, "step": 9761 }, { "epoch": 0.24, "learning_rate": 1.7762542495577513e-05, "loss": 2.147, "step": 9762 }, { "epoch": 0.24, "learning_rate": 1.7762039186954954e-05, "loss": 2.1382, "step": 9763 }, { "epoch": 0.24, "learning_rate": 1.7761535828862398e-05, "loss": 2.1101, "step": 9764 }, { "epoch": 0.24, "learning_rate": 1.776103242130304e-05, "loss": 2.1416, "step": 9765 }, { "epoch": 0.24, "learning_rate": 1.7760528964280092e-05, "loss": 2.1949, "step": 9766 }, { "epoch": 0.24, "learning_rate": 1.776002545779677e-05, "loss": 2.1529, "step": 9767 }, { "epoch": 0.24, "learning_rate": 1.775952190185627e-05, "loss": 2.2095, "step": 9768 }, { "epoch": 0.24, "learning_rate": 1.775901829646181e-05, "loss": 2.2474, "step": 9769 }, { "epoch": 0.24, "learning_rate": 1.7758514641616596e-05, "loss": 2.3606, "step": 9770 }, { "epoch": 0.24, "learning_rate": 1.7758010937323843e-05, "loss": 2.0895, "step": 9771 }, { "epoch": 0.24, "learning_rate": 1.7757507183586758e-05, "loss": 2.1036, "step": 9772 }, { "epoch": 0.24, "learning_rate": 1.7757003380408552e-05, "loss": 2.1509, "step": 9773 }, { "epoch": 0.24, "learning_rate": 1.7756499527792432e-05, "loss": 2.0413, "step": 9774 }, { "epoch": 0.24, "learning_rate": 1.7755995625741616e-05, "loss": 2.1129, "step": 9775 }, { "epoch": 0.24, "learning_rate": 1.7755491674259312e-05, "loss": 2.1655, "step": 9776 }, { "epoch": 0.24, "learning_rate": 1.775498767334873e-05, "loss": 2.0525, "step": 9777 }, { "epoch": 0.24, "learning_rate": 1.7754483623013086e-05, "loss": 2.0508, "step": 9778 }, { "epoch": 0.24, "learning_rate": 1.7753979523255588e-05, "loss": 2.2125, "step": 9779 }, { "epoch": 0.24, "learning_rate": 1.7753475374079457e-05, "loss": 2.2207, "step": 9780 }, { "epoch": 0.24, "learning_rate": 1.7752971175487898e-05, "loss": 2.1302, "step": 9781 }, { "epoch": 0.24, "learning_rate": 1.7752466927484127e-05, "loss": 2.2174, "step": 9782 }, { "epoch": 0.24, "learning_rate": 1.7751962630071355e-05, "loss": 2.0485, "step": 9783 }, { "epoch": 0.24, "learning_rate": 1.77514582832528e-05, "loss": 2.183, "step": 9784 }, { "epoch": 0.24, "learning_rate": 1.7750953887031677e-05, "loss": 2.1768, "step": 9785 }, { "epoch": 0.24, "learning_rate": 1.77504494414112e-05, "loss": 2.125, "step": 9786 }, { "epoch": 0.24, "learning_rate": 1.7749944946394582e-05, "loss": 2.027, "step": 9787 }, { "epoch": 0.24, "learning_rate": 1.7749440401985038e-05, "loss": 2.2127, "step": 9788 }, { "epoch": 0.24, "learning_rate": 1.7748935808185785e-05, "loss": 2.1267, "step": 9789 }, { "epoch": 0.24, "learning_rate": 1.774843116500004e-05, "loss": 2.1528, "step": 9790 }, { "epoch": 0.24, "learning_rate": 1.7747926472431015e-05, "loss": 2.0521, "step": 9791 }, { "epoch": 0.24, "learning_rate": 1.7747421730481935e-05, "loss": 2.0843, "step": 9792 }, { "epoch": 0.24, "learning_rate": 1.7746916939156008e-05, "loss": 2.1889, "step": 9793 }, { "epoch": 0.24, "learning_rate": 1.7746412098456453e-05, "loss": 2.181, "step": 9794 }, { "epoch": 0.24, "learning_rate": 1.774590720838649e-05, "loss": 1.9367, "step": 9795 }, { "epoch": 0.24, "learning_rate": 1.7745402268949342e-05, "loss": 2.1938, "step": 9796 }, { "epoch": 0.24, "learning_rate": 1.7744897280148214e-05, "loss": 2.168, "step": 9797 }, { "epoch": 0.24, "learning_rate": 1.7744392241986336e-05, "loss": 2.1589, "step": 9798 }, { "epoch": 0.24, "learning_rate": 1.774388715446692e-05, "loss": 2.269, "step": 9799 }, { "epoch": 0.24, "learning_rate": 1.774338201759319e-05, "loss": 2.2504, "step": 9800 }, { "epoch": 0.24, "learning_rate": 1.7742876831368357e-05, "loss": 2.2562, "step": 9801 }, { "epoch": 0.24, "learning_rate": 1.7742371595795652e-05, "loss": 1.9776, "step": 9802 }, { "epoch": 0.24, "learning_rate": 1.7741866310878287e-05, "loss": 2.1128, "step": 9803 }, { "epoch": 0.24, "learning_rate": 1.7741360976619488e-05, "loss": 2.1078, "step": 9804 }, { "epoch": 0.24, "learning_rate": 1.7740855593022473e-05, "loss": 2.0718, "step": 9805 }, { "epoch": 0.24, "learning_rate": 1.7740350160090457e-05, "loss": 2.0952, "step": 9806 }, { "epoch": 0.24, "learning_rate": 1.773984467782667e-05, "loss": 2.4334, "step": 9807 }, { "epoch": 0.24, "learning_rate": 1.7739339146234334e-05, "loss": 2.2593, "step": 9808 }, { "epoch": 0.24, "learning_rate": 1.7738833565316664e-05, "loss": 2.1459, "step": 9809 }, { "epoch": 0.24, "learning_rate": 1.7738327935076885e-05, "loss": 2.0978, "step": 9810 }, { "epoch": 0.24, "learning_rate": 1.7737822255518224e-05, "loss": 1.9903, "step": 9811 }, { "epoch": 0.24, "learning_rate": 1.7737316526643895e-05, "loss": 2.1991, "step": 9812 }, { "epoch": 0.24, "learning_rate": 1.773681074845713e-05, "loss": 2.0489, "step": 9813 }, { "epoch": 0.24, "learning_rate": 1.773630492096115e-05, "loss": 2.0674, "step": 9814 }, { "epoch": 0.24, "learning_rate": 1.7735799044159173e-05, "loss": 2.3256, "step": 9815 }, { "epoch": 0.24, "learning_rate": 1.7735293118054428e-05, "loss": 2.1081, "step": 9816 }, { "epoch": 0.24, "learning_rate": 1.7734787142650143e-05, "loss": 2.1288, "step": 9817 }, { "epoch": 0.24, "learning_rate": 1.7734281117949533e-05, "loss": 2.1457, "step": 9818 }, { "epoch": 0.24, "learning_rate": 1.7733775043955834e-05, "loss": 2.1715, "step": 9819 }, { "epoch": 0.24, "learning_rate": 1.7733268920672265e-05, "loss": 2.1009, "step": 9820 }, { "epoch": 0.24, "learning_rate": 1.7732762748102054e-05, "loss": 2.0893, "step": 9821 }, { "epoch": 0.24, "learning_rate": 1.7732256526248423e-05, "loss": 2.1759, "step": 9822 }, { "epoch": 0.24, "learning_rate": 1.7731750255114603e-05, "loss": 2.2853, "step": 9823 }, { "epoch": 0.24, "learning_rate": 1.773124393470382e-05, "loss": 2.2039, "step": 9824 }, { "epoch": 0.24, "learning_rate": 1.7730737565019298e-05, "loss": 2.1845, "step": 9825 }, { "epoch": 0.24, "learning_rate": 1.773023114606427e-05, "loss": 2.1508, "step": 9826 }, { "epoch": 0.24, "learning_rate": 1.7729724677841954e-05, "loss": 2.1075, "step": 9827 }, { "epoch": 0.24, "learning_rate": 1.7729218160355587e-05, "loss": 2.066, "step": 9828 }, { "epoch": 0.24, "learning_rate": 1.7728711593608393e-05, "loss": 2.1427, "step": 9829 }, { "epoch": 0.24, "learning_rate": 1.77282049776036e-05, "loss": 2.3061, "step": 9830 }, { "epoch": 0.24, "learning_rate": 1.772769831234444e-05, "loss": 2.1056, "step": 9831 }, { "epoch": 0.24, "learning_rate": 1.772719159783414e-05, "loss": 2.0783, "step": 9832 }, { "epoch": 0.24, "learning_rate": 1.7726684834075932e-05, "loss": 2.2044, "step": 9833 }, { "epoch": 0.24, "learning_rate": 1.7726178021073038e-05, "loss": 2.2132, "step": 9834 }, { "epoch": 0.24, "learning_rate": 1.7725671158828696e-05, "loss": 2.2201, "step": 9835 }, { "epoch": 0.24, "learning_rate": 1.7725164247346136e-05, "loss": 2.2061, "step": 9836 }, { "epoch": 0.24, "learning_rate": 1.7724657286628586e-05, "loss": 2.3359, "step": 9837 }, { "epoch": 0.24, "learning_rate": 1.7724150276679276e-05, "loss": 2.2795, "step": 9838 }, { "epoch": 0.24, "learning_rate": 1.772364321750144e-05, "loss": 2.199, "step": 9839 }, { "epoch": 0.24, "learning_rate": 1.772313610909831e-05, "loss": 2.0729, "step": 9840 }, { "epoch": 0.24, "learning_rate": 1.7722628951473115e-05, "loss": 2.186, "step": 9841 }, { "epoch": 0.24, "learning_rate": 1.7722121744629088e-05, "loss": 2.2234, "step": 9842 }, { "epoch": 0.24, "learning_rate": 1.7721614488569466e-05, "loss": 2.2079, "step": 9843 }, { "epoch": 0.24, "learning_rate": 1.7721107183297474e-05, "loss": 2.1812, "step": 9844 }, { "epoch": 0.24, "learning_rate": 1.7720599828816353e-05, "loss": 1.9783, "step": 9845 }, { "epoch": 0.24, "learning_rate": 1.772009242512933e-05, "loss": 2.4196, "step": 9846 }, { "epoch": 0.24, "learning_rate": 1.7719584972239645e-05, "loss": 2.2808, "step": 9847 }, { "epoch": 0.24, "learning_rate": 1.7719077470150528e-05, "loss": 2.1043, "step": 9848 }, { "epoch": 0.24, "learning_rate": 1.7718569918865216e-05, "loss": 2.3104, "step": 9849 }, { "epoch": 0.24, "learning_rate": 1.7718062318386944e-05, "loss": 2.2034, "step": 9850 }, { "epoch": 0.24, "learning_rate": 1.771755466871894e-05, "loss": 2.2368, "step": 9851 }, { "epoch": 0.24, "learning_rate": 1.771704696986445e-05, "loss": 2.2013, "step": 9852 }, { "epoch": 0.24, "learning_rate": 1.7716539221826703e-05, "loss": 2.3712, "step": 9853 }, { "epoch": 0.24, "learning_rate": 1.771603142460894e-05, "loss": 2.1608, "step": 9854 }, { "epoch": 0.24, "learning_rate": 1.771552357821439e-05, "loss": 2.1387, "step": 9855 }, { "epoch": 0.24, "learning_rate": 1.7715015682646292e-05, "loss": 2.1671, "step": 9856 }, { "epoch": 0.24, "learning_rate": 1.771450773790789e-05, "loss": 2.1505, "step": 9857 }, { "epoch": 0.24, "learning_rate": 1.771399974400241e-05, "loss": 2.2741, "step": 9858 }, { "epoch": 0.24, "learning_rate": 1.7713491700933104e-05, "loss": 2.1882, "step": 9859 }, { "epoch": 0.24, "learning_rate": 1.7712983608703197e-05, "loss": 2.196, "step": 9860 }, { "epoch": 0.24, "learning_rate": 1.771247546731593e-05, "loss": 2.1976, "step": 9861 }, { "epoch": 0.24, "learning_rate": 1.7711967276774547e-05, "loss": 2.0128, "step": 9862 }, { "epoch": 0.24, "learning_rate": 1.7711459037082282e-05, "loss": 2.2529, "step": 9863 }, { "epoch": 0.24, "learning_rate": 1.7710950748242373e-05, "loss": 2.1561, "step": 9864 }, { "epoch": 0.24, "learning_rate": 1.7710442410258064e-05, "loss": 2.2334, "step": 9865 }, { "epoch": 0.24, "learning_rate": 1.7709934023132593e-05, "loss": 1.9996, "step": 9866 }, { "epoch": 0.24, "learning_rate": 1.7709425586869202e-05, "loss": 2.173, "step": 9867 }, { "epoch": 0.24, "learning_rate": 1.7708917101471127e-05, "loss": 2.2908, "step": 9868 }, { "epoch": 0.24, "learning_rate": 1.7708408566941608e-05, "loss": 2.0453, "step": 9869 }, { "epoch": 0.24, "learning_rate": 1.7707899983283893e-05, "loss": 2.029, "step": 9870 }, { "epoch": 0.24, "learning_rate": 1.770739135050122e-05, "loss": 2.011, "step": 9871 }, { "epoch": 0.24, "learning_rate": 1.7706882668596826e-05, "loss": 2.0265, "step": 9872 }, { "epoch": 0.24, "learning_rate": 1.7706373937573962e-05, "loss": 2.2767, "step": 9873 }, { "epoch": 0.24, "learning_rate": 1.7705865157435863e-05, "loss": 2.106, "step": 9874 }, { "epoch": 0.24, "learning_rate": 1.7705356328185773e-05, "loss": 2.1709, "step": 9875 }, { "epoch": 0.24, "learning_rate": 1.7704847449826942e-05, "loss": 2.1322, "step": 9876 }, { "epoch": 0.24, "learning_rate": 1.77043385223626e-05, "loss": 2.1279, "step": 9877 }, { "epoch": 0.24, "learning_rate": 1.7703829545796e-05, "loss": 2.1314, "step": 9878 }, { "epoch": 0.24, "learning_rate": 1.770332052013039e-05, "loss": 2.196, "step": 9879 }, { "epoch": 0.24, "learning_rate": 1.7702811445368998e-05, "loss": 2.1155, "step": 9880 }, { "epoch": 0.24, "learning_rate": 1.7702302321515083e-05, "loss": 2.1886, "step": 9881 }, { "epoch": 0.24, "learning_rate": 1.7701793148571884e-05, "loss": 2.1412, "step": 9882 }, { "epoch": 0.24, "learning_rate": 1.770128392654265e-05, "loss": 2.3353, "step": 9883 }, { "epoch": 0.24, "learning_rate": 1.7700774655430622e-05, "loss": 2.0497, "step": 9884 }, { "epoch": 0.24, "learning_rate": 1.7700265335239047e-05, "loss": 2.3085, "step": 9885 }, { "epoch": 0.24, "learning_rate": 1.7699755965971172e-05, "loss": 2.1279, "step": 9886 }, { "epoch": 0.24, "learning_rate": 1.7699246547630244e-05, "loss": 2.2345, "step": 9887 }, { "epoch": 0.24, "learning_rate": 1.7698737080219507e-05, "loss": 2.2268, "step": 9888 }, { "epoch": 0.24, "learning_rate": 1.769822756374221e-05, "loss": 2.0959, "step": 9889 }, { "epoch": 0.24, "learning_rate": 1.76977179982016e-05, "loss": 2.346, "step": 9890 }, { "epoch": 0.24, "learning_rate": 1.7697208383600926e-05, "loss": 2.2473, "step": 9891 }, { "epoch": 0.24, "learning_rate": 1.769669871994343e-05, "loss": 2.2721, "step": 9892 }, { "epoch": 0.24, "learning_rate": 1.769618900723237e-05, "loss": 2.1689, "step": 9893 }, { "epoch": 0.24, "learning_rate": 1.7695679245470988e-05, "loss": 2.2887, "step": 9894 }, { "epoch": 0.24, "learning_rate": 1.769516943466253e-05, "loss": 2.0369, "step": 9895 }, { "epoch": 0.24, "learning_rate": 1.7694659574810256e-05, "loss": 2.2024, "step": 9896 }, { "epoch": 0.24, "learning_rate": 1.7694149665917403e-05, "loss": 2.1482, "step": 9897 }, { "epoch": 0.24, "learning_rate": 1.769363970798723e-05, "loss": 2.1851, "step": 9898 }, { "epoch": 0.24, "learning_rate": 1.7693129701022984e-05, "loss": 2.1428, "step": 9899 }, { "epoch": 0.24, "learning_rate": 1.7692619645027918e-05, "loss": 2.2744, "step": 9900 }, { "epoch": 0.24, "learning_rate": 1.7692109540005277e-05, "loss": 2.1703, "step": 9901 }, { "epoch": 0.24, "learning_rate": 1.7691599385958314e-05, "loss": 2.2423, "step": 9902 }, { "epoch": 0.24, "learning_rate": 1.769108918289028e-05, "loss": 2.0435, "step": 9903 }, { "epoch": 0.24, "learning_rate": 1.769057893080443e-05, "loss": 2.1978, "step": 9904 }, { "epoch": 0.24, "learning_rate": 1.7690068629704016e-05, "loss": 2.1376, "step": 9905 }, { "epoch": 0.24, "learning_rate": 1.768955827959229e-05, "loss": 2.1879, "step": 9906 }, { "epoch": 0.24, "learning_rate": 1.7689047880472498e-05, "loss": 2.3482, "step": 9907 }, { "epoch": 0.24, "learning_rate": 1.7688537432347903e-05, "loss": 2.191, "step": 9908 }, { "epoch": 0.24, "learning_rate": 1.768802693522175e-05, "loss": 2.073, "step": 9909 }, { "epoch": 0.24, "learning_rate": 1.7687516389097297e-05, "loss": 2.1533, "step": 9910 }, { "epoch": 0.24, "learning_rate": 1.7687005793977797e-05, "loss": 2.173, "step": 9911 }, { "epoch": 0.24, "learning_rate": 1.7686495149866502e-05, "loss": 2.329, "step": 9912 }, { "epoch": 0.24, "learning_rate": 1.768598445676667e-05, "loss": 2.1172, "step": 9913 }, { "epoch": 0.24, "learning_rate": 1.7685473714681554e-05, "loss": 2.1276, "step": 9914 }, { "epoch": 0.24, "learning_rate": 1.768496292361441e-05, "loss": 2.1105, "step": 9915 }, { "epoch": 0.24, "learning_rate": 1.7684452083568493e-05, "loss": 2.0228, "step": 9916 }, { "epoch": 0.24, "learning_rate": 1.7683941194547057e-05, "loss": 1.9678, "step": 9917 }, { "epoch": 0.24, "learning_rate": 1.768343025655336e-05, "loss": 2.1719, "step": 9918 }, { "epoch": 0.24, "learning_rate": 1.768291926959066e-05, "loss": 2.2154, "step": 9919 }, { "epoch": 0.24, "learning_rate": 1.768240823366221e-05, "loss": 2.0197, "step": 9920 }, { "epoch": 0.24, "learning_rate": 1.768189714877127e-05, "loss": 2.198, "step": 9921 }, { "epoch": 0.24, "learning_rate": 1.7681386014921097e-05, "loss": 2.0947, "step": 9922 }, { "epoch": 0.24, "learning_rate": 1.7680874832114943e-05, "loss": 2.0666, "step": 9923 }, { "epoch": 0.24, "learning_rate": 1.7680363600356075e-05, "loss": 2.2054, "step": 9924 }, { "epoch": 0.24, "learning_rate": 1.7679852319647742e-05, "loss": 2.1923, "step": 9925 }, { "epoch": 0.24, "learning_rate": 1.767934098999321e-05, "loss": 2.2636, "step": 9926 }, { "epoch": 0.24, "learning_rate": 1.7678829611395737e-05, "loss": 2.1616, "step": 9927 }, { "epoch": 0.24, "learning_rate": 1.767831818385858e-05, "loss": 2.1073, "step": 9928 }, { "epoch": 0.24, "learning_rate": 1.7677806707384995e-05, "loss": 2.0825, "step": 9929 }, { "epoch": 0.24, "learning_rate": 1.7677295181978253e-05, "loss": 2.0976, "step": 9930 }, { "epoch": 0.24, "learning_rate": 1.7676783607641602e-05, "loss": 2.2689, "step": 9931 }, { "epoch": 0.24, "learning_rate": 1.767627198437831e-05, "loss": 2.1792, "step": 9932 }, { "epoch": 0.24, "learning_rate": 1.767576031219163e-05, "loss": 2.0823, "step": 9933 }, { "epoch": 0.24, "learning_rate": 1.767524859108483e-05, "loss": 2.1806, "step": 9934 }, { "epoch": 0.24, "learning_rate": 1.7674736821061173e-05, "loss": 2.2507, "step": 9935 }, { "epoch": 0.24, "learning_rate": 1.7674225002123914e-05, "loss": 2.0057, "step": 9936 }, { "epoch": 0.24, "learning_rate": 1.7673713134276317e-05, "loss": 2.2851, "step": 9937 }, { "epoch": 0.24, "learning_rate": 1.7673201217521647e-05, "loss": 2.0627, "step": 9938 }, { "epoch": 0.24, "learning_rate": 1.7672689251863168e-05, "loss": 2.2986, "step": 9939 }, { "epoch": 0.25, "learning_rate": 1.7672177237304134e-05, "loss": 2.3382, "step": 9940 }, { "epoch": 0.25, "learning_rate": 1.7671665173847817e-05, "loss": 2.2543, "step": 9941 }, { "epoch": 0.25, "learning_rate": 1.7671153061497478e-05, "loss": 2.1151, "step": 9942 }, { "epoch": 0.25, "learning_rate": 1.767064090025638e-05, "loss": 2.1313, "step": 9943 }, { "epoch": 0.25, "learning_rate": 1.767012869012779e-05, "loss": 2.2186, "step": 9944 }, { "epoch": 0.25, "learning_rate": 1.7669616431114967e-05, "loss": 2.2774, "step": 9945 }, { "epoch": 0.25, "learning_rate": 1.766910412322118e-05, "loss": 2.1125, "step": 9946 }, { "epoch": 0.25, "learning_rate": 1.7668591766449696e-05, "loss": 2.1463, "step": 9947 }, { "epoch": 0.25, "learning_rate": 1.7668079360803774e-05, "loss": 2.1059, "step": 9948 }, { "epoch": 0.25, "learning_rate": 1.7667566906286687e-05, "loss": 2.2242, "step": 9949 }, { "epoch": 0.25, "learning_rate": 1.7667054402901694e-05, "loss": 2.0936, "step": 9950 }, { "epoch": 0.25, "learning_rate": 1.7666541850652065e-05, "loss": 2.2778, "step": 9951 }, { "epoch": 0.25, "learning_rate": 1.7666029249541066e-05, "loss": 2.209, "step": 9952 }, { "epoch": 0.25, "learning_rate": 1.7665516599571966e-05, "loss": 2.0813, "step": 9953 }, { "epoch": 0.25, "learning_rate": 1.766500390074803e-05, "loss": 2.2416, "step": 9954 }, { "epoch": 0.25, "learning_rate": 1.7664491153072523e-05, "loss": 2.0546, "step": 9955 }, { "epoch": 0.25, "learning_rate": 1.766397835654872e-05, "loss": 2.1574, "step": 9956 }, { "epoch": 0.25, "learning_rate": 1.7663465511179886e-05, "loss": 2.1669, "step": 9957 }, { "epoch": 0.25, "learning_rate": 1.7662952616969287e-05, "loss": 2.1876, "step": 9958 }, { "epoch": 0.25, "learning_rate": 1.7662439673920194e-05, "loss": 2.2623, "step": 9959 }, { "epoch": 0.25, "learning_rate": 1.7661926682035875e-05, "loss": 2.0645, "step": 9960 }, { "epoch": 0.25, "learning_rate": 1.76614136413196e-05, "loss": 2.2604, "step": 9961 }, { "epoch": 0.25, "learning_rate": 1.7660900551774642e-05, "loss": 2.3138, "step": 9962 }, { "epoch": 0.25, "learning_rate": 1.7660387413404265e-05, "loss": 2.215, "step": 9963 }, { "epoch": 0.25, "learning_rate": 1.7659874226211745e-05, "loss": 2.3373, "step": 9964 }, { "epoch": 0.25, "learning_rate": 1.765936099020035e-05, "loss": 2.2912, "step": 9965 }, { "epoch": 0.25, "learning_rate": 1.7658847705373348e-05, "loss": 2.2234, "step": 9966 }, { "epoch": 0.25, "learning_rate": 1.765833437173402e-05, "loss": 2.2809, "step": 9967 }, { "epoch": 0.25, "learning_rate": 1.7657820989285626e-05, "loss": 2.2272, "step": 9968 }, { "epoch": 0.25, "learning_rate": 1.7657307558031442e-05, "loss": 2.2329, "step": 9969 }, { "epoch": 0.25, "learning_rate": 1.7656794077974747e-05, "loss": 2.1611, "step": 9970 }, { "epoch": 0.25, "learning_rate": 1.7656280549118803e-05, "loss": 2.1371, "step": 9971 }, { "epoch": 0.25, "learning_rate": 1.765576697146689e-05, "loss": 2.2991, "step": 9972 }, { "epoch": 0.25, "learning_rate": 1.765525334502228e-05, "loss": 1.9735, "step": 9973 }, { "epoch": 0.25, "learning_rate": 1.7654739669788242e-05, "loss": 2.0669, "step": 9974 }, { "epoch": 0.25, "learning_rate": 1.7654225945768057e-05, "loss": 2.2259, "step": 9975 }, { "epoch": 0.25, "learning_rate": 1.7653712172964995e-05, "loss": 2.0968, "step": 9976 }, { "epoch": 0.25, "learning_rate": 1.765319835138233e-05, "loss": 2.4145, "step": 9977 }, { "epoch": 0.25, "learning_rate": 1.765268448102334e-05, "loss": 2.2425, "step": 9978 }, { "epoch": 0.25, "learning_rate": 1.7652170561891298e-05, "loss": 2.1239, "step": 9979 }, { "epoch": 0.25, "learning_rate": 1.7651656593989476e-05, "loss": 2.0762, "step": 9980 }, { "epoch": 0.25, "learning_rate": 1.7651142577321156e-05, "loss": 2.1044, "step": 9981 }, { "epoch": 0.25, "learning_rate": 1.7650628511889613e-05, "loss": 2.0705, "step": 9982 }, { "epoch": 0.25, "learning_rate": 1.7650114397698114e-05, "loss": 2.1915, "step": 9983 }, { "epoch": 0.25, "learning_rate": 1.7649600234749948e-05, "loss": 2.2533, "step": 9984 }, { "epoch": 0.25, "learning_rate": 1.7649086023048385e-05, "loss": 2.1249, "step": 9985 }, { "epoch": 0.25, "learning_rate": 1.7648571762596705e-05, "loss": 2.0335, "step": 9986 }, { "epoch": 0.25, "learning_rate": 1.7648057453398184e-05, "loss": 1.9675, "step": 9987 }, { "epoch": 0.25, "learning_rate": 1.76475430954561e-05, "loss": 2.0243, "step": 9988 }, { "epoch": 0.25, "learning_rate": 1.764702868877373e-05, "loss": 2.2134, "step": 9989 }, { "epoch": 0.25, "learning_rate": 1.7646514233354358e-05, "loss": 2.2932, "step": 9990 }, { "epoch": 0.25, "learning_rate": 1.7645999729201256e-05, "loss": 2.2514, "step": 9991 }, { "epoch": 0.25, "learning_rate": 1.764548517631771e-05, "loss": 2.0569, "step": 9992 }, { "epoch": 0.25, "learning_rate": 1.764497057470699e-05, "loss": 2.1715, "step": 9993 }, { "epoch": 0.25, "learning_rate": 1.7644455924372382e-05, "loss": 2.3075, "step": 9994 }, { "epoch": 0.25, "learning_rate": 1.764394122531717e-05, "loss": 2.3577, "step": 9995 }, { "epoch": 0.25, "learning_rate": 1.764342647754462e-05, "loss": 2.2103, "step": 9996 }, { "epoch": 0.25, "learning_rate": 1.7642911681058034e-05, "loss": 2.055, "step": 9997 }, { "epoch": 0.25, "learning_rate": 1.7642396835860672e-05, "loss": 2.2155, "step": 9998 }, { "epoch": 0.25, "learning_rate": 1.7641881941955833e-05, "loss": 2.0889, "step": 9999 }, { "epoch": 0.25, "learning_rate": 1.764136699934678e-05, "loss": 2.1609, "step": 10000 }, { "epoch": 0.25, "learning_rate": 1.764085200803681e-05, "loss": 2.2409, "step": 10001 }, { "epoch": 0.25, "learning_rate": 1.76403369680292e-05, "loss": 2.0643, "step": 10002 }, { "epoch": 0.25, "learning_rate": 1.763982187932723e-05, "loss": 2.0362, "step": 10003 }, { "epoch": 0.25, "learning_rate": 1.7639306741934192e-05, "loss": 2.0401, "step": 10004 }, { "epoch": 0.25, "learning_rate": 1.7638791555853357e-05, "loss": 2.0644, "step": 10005 }, { "epoch": 0.25, "learning_rate": 1.7638276321088016e-05, "loss": 2.1589, "step": 10006 }, { "epoch": 0.25, "learning_rate": 1.763776103764145e-05, "loss": 2.1796, "step": 10007 }, { "epoch": 0.25, "learning_rate": 1.7637245705516945e-05, "loss": 2.1429, "step": 10008 }, { "epoch": 0.25, "learning_rate": 1.763673032471778e-05, "loss": 2.1567, "step": 10009 }, { "epoch": 0.25, "learning_rate": 1.763621489524725e-05, "loss": 2.109, "step": 10010 }, { "epoch": 0.25, "learning_rate": 1.7635699417108627e-05, "loss": 2.1253, "step": 10011 }, { "epoch": 0.25, "learning_rate": 1.763518389030521e-05, "loss": 2.0456, "step": 10012 }, { "epoch": 0.25, "learning_rate": 1.7634668314840274e-05, "loss": 2.1371, "step": 10013 }, { "epoch": 0.25, "learning_rate": 1.763415269071711e-05, "loss": 2.1493, "step": 10014 }, { "epoch": 0.25, "learning_rate": 1.7633637017939e-05, "loss": 2.3095, "step": 10015 }, { "epoch": 0.25, "learning_rate": 1.763312129650924e-05, "loss": 2.2042, "step": 10016 }, { "epoch": 0.25, "learning_rate": 1.7632605526431104e-05, "loss": 2.0831, "step": 10017 }, { "epoch": 0.25, "learning_rate": 1.763208970770789e-05, "loss": 2.0357, "step": 10018 }, { "epoch": 0.25, "learning_rate": 1.7631573840342876e-05, "loss": 2.2122, "step": 10019 }, { "epoch": 0.25, "learning_rate": 1.7631057924339358e-05, "loss": 2.128, "step": 10020 }, { "epoch": 0.25, "learning_rate": 1.7630541959700624e-05, "loss": 2.1381, "step": 10021 }, { "epoch": 0.25, "learning_rate": 1.7630025946429957e-05, "loss": 2.3139, "step": 10022 }, { "epoch": 0.25, "learning_rate": 1.7629509884530644e-05, "loss": 2.271, "step": 10023 }, { "epoch": 0.25, "learning_rate": 1.7628993774005986e-05, "loss": 2.1067, "step": 10024 }, { "epoch": 0.25, "learning_rate": 1.7628477614859258e-05, "loss": 2.2094, "step": 10025 }, { "epoch": 0.25, "learning_rate": 1.7627961407093756e-05, "loss": 2.3802, "step": 10026 }, { "epoch": 0.25, "learning_rate": 1.7627445150712775e-05, "loss": 2.127, "step": 10027 }, { "epoch": 0.25, "learning_rate": 1.7626928845719598e-05, "loss": 2.1664, "step": 10028 }, { "epoch": 0.25, "learning_rate": 1.7626412492117517e-05, "loss": 2.2684, "step": 10029 }, { "epoch": 0.25, "learning_rate": 1.7625896089909826e-05, "loss": 1.9763, "step": 10030 }, { "epoch": 0.25, "learning_rate": 1.7625379639099808e-05, "loss": 2.107, "step": 10031 }, { "epoch": 0.25, "learning_rate": 1.7624863139690766e-05, "loss": 2.1705, "step": 10032 }, { "epoch": 0.25, "learning_rate": 1.7624346591685984e-05, "loss": 2.3163, "step": 10033 }, { "epoch": 0.25, "learning_rate": 1.7623829995088758e-05, "loss": 2.0956, "step": 10034 }, { "epoch": 0.25, "learning_rate": 1.762331334990238e-05, "loss": 2.247, "step": 10035 }, { "epoch": 0.25, "learning_rate": 1.762279665613014e-05, "loss": 2.3487, "step": 10036 }, { "epoch": 0.25, "learning_rate": 1.7622279913775333e-05, "loss": 2.2181, "step": 10037 }, { "epoch": 0.25, "learning_rate": 1.7621763122841255e-05, "loss": 2.0178, "step": 10038 }, { "epoch": 0.25, "learning_rate": 1.7621246283331192e-05, "loss": 2.1024, "step": 10039 }, { "epoch": 0.25, "learning_rate": 1.7620729395248445e-05, "loss": 2.1071, "step": 10040 }, { "epoch": 0.25, "learning_rate": 1.7620212458596303e-05, "loss": 2.1632, "step": 10041 }, { "epoch": 0.25, "learning_rate": 1.761969547337807e-05, "loss": 2.1768, "step": 10042 }, { "epoch": 0.25, "learning_rate": 1.761917843959703e-05, "loss": 2.2188, "step": 10043 }, { "epoch": 0.25, "learning_rate": 1.7618661357256482e-05, "loss": 2.2475, "step": 10044 }, { "epoch": 0.25, "learning_rate": 1.7618144226359723e-05, "loss": 2.1754, "step": 10045 }, { "epoch": 0.25, "learning_rate": 1.761762704691005e-05, "loss": 2.1207, "step": 10046 }, { "epoch": 0.25, "learning_rate": 1.7617109818910754e-05, "loss": 2.245, "step": 10047 }, { "epoch": 0.25, "learning_rate": 1.7616592542365142e-05, "loss": 2.2983, "step": 10048 }, { "epoch": 0.25, "learning_rate": 1.7616075217276494e-05, "loss": 2.2142, "step": 10049 }, { "epoch": 0.25, "learning_rate": 1.7615557843648123e-05, "loss": 2.2667, "step": 10050 }, { "epoch": 0.25, "learning_rate": 1.761504042148332e-05, "loss": 2.1583, "step": 10051 }, { "epoch": 0.25, "learning_rate": 1.761452295078538e-05, "loss": 2.1733, "step": 10052 }, { "epoch": 0.25, "learning_rate": 1.7614005431557606e-05, "loss": 2.1466, "step": 10053 }, { "epoch": 0.25, "learning_rate": 1.7613487863803293e-05, "loss": 2.2139, "step": 10054 }, { "epoch": 0.25, "learning_rate": 1.7612970247525738e-05, "loss": 2.2714, "step": 10055 }, { "epoch": 0.25, "learning_rate": 1.761245258272825e-05, "loss": 2.0702, "step": 10056 }, { "epoch": 0.25, "learning_rate": 1.7611934869414112e-05, "loss": 2.0066, "step": 10057 }, { "epoch": 0.25, "learning_rate": 1.7611417107586637e-05, "loss": 2.2995, "step": 10058 }, { "epoch": 0.25, "learning_rate": 1.761089929724912e-05, "loss": 2.0423, "step": 10059 }, { "epoch": 0.25, "learning_rate": 1.761038143840486e-05, "loss": 2.0582, "step": 10060 }, { "epoch": 0.25, "learning_rate": 1.760986353105716e-05, "loss": 2.119, "step": 10061 }, { "epoch": 0.25, "learning_rate": 1.760934557520932e-05, "loss": 2.087, "step": 10062 }, { "epoch": 0.25, "learning_rate": 1.760882757086464e-05, "loss": 2.1343, "step": 10063 }, { "epoch": 0.25, "learning_rate": 1.7608309518026425e-05, "loss": 2.0643, "step": 10064 }, { "epoch": 0.25, "learning_rate": 1.7607791416697973e-05, "loss": 2.0924, "step": 10065 }, { "epoch": 0.25, "learning_rate": 1.7607273266882585e-05, "loss": 2.1149, "step": 10066 }, { "epoch": 0.25, "learning_rate": 1.7606755068583566e-05, "loss": 2.0894, "step": 10067 }, { "epoch": 0.25, "learning_rate": 1.760623682180422e-05, "loss": 2.139, "step": 10068 }, { "epoch": 0.25, "learning_rate": 1.7605718526547843e-05, "loss": 2.1336, "step": 10069 }, { "epoch": 0.25, "learning_rate": 1.760520018281775e-05, "loss": 2.1991, "step": 10070 }, { "epoch": 0.25, "learning_rate": 1.760468179061723e-05, "loss": 2.022, "step": 10071 }, { "epoch": 0.25, "learning_rate": 1.76041633499496e-05, "loss": 2.1682, "step": 10072 }, { "epoch": 0.25, "learning_rate": 1.760364486081816e-05, "loss": 2.0398, "step": 10073 }, { "epoch": 0.25, "learning_rate": 1.760312632322621e-05, "loss": 2.1115, "step": 10074 }, { "epoch": 0.25, "learning_rate": 1.760260773717706e-05, "loss": 1.9952, "step": 10075 }, { "epoch": 0.25, "learning_rate": 1.7602089102674012e-05, "loss": 2.1847, "step": 10076 }, { "epoch": 0.25, "learning_rate": 1.7601570419720372e-05, "loss": 2.2326, "step": 10077 }, { "epoch": 0.25, "learning_rate": 1.7601051688319447e-05, "loss": 2.1414, "step": 10078 }, { "epoch": 0.25, "learning_rate": 1.7600532908474545e-05, "loss": 2.1644, "step": 10079 }, { "epoch": 0.25, "learning_rate": 1.7600014080188964e-05, "loss": 2.1378, "step": 10080 }, { "epoch": 0.25, "learning_rate": 1.759949520346602e-05, "loss": 2.2346, "step": 10081 }, { "epoch": 0.25, "learning_rate": 1.759897627830902e-05, "loss": 2.1349, "step": 10082 }, { "epoch": 0.25, "learning_rate": 1.7598457304721265e-05, "loss": 2.277, "step": 10083 }, { "epoch": 0.25, "learning_rate": 1.7597938282706062e-05, "loss": 2.2347, "step": 10084 }, { "epoch": 0.25, "learning_rate": 1.7597419212266725e-05, "loss": 2.2579, "step": 10085 }, { "epoch": 0.25, "learning_rate": 1.7596900093406563e-05, "loss": 2.3106, "step": 10086 }, { "epoch": 0.25, "learning_rate": 1.7596380926128875e-05, "loss": 2.3374, "step": 10087 }, { "epoch": 0.25, "learning_rate": 1.7595861710436977e-05, "loss": 2.2443, "step": 10088 }, { "epoch": 0.25, "learning_rate": 1.759534244633418e-05, "loss": 2.1928, "step": 10089 }, { "epoch": 0.25, "learning_rate": 1.7594823133823788e-05, "loss": 2.1662, "step": 10090 }, { "epoch": 0.25, "learning_rate": 1.7594303772909114e-05, "loss": 1.9967, "step": 10091 }, { "epoch": 0.25, "learning_rate": 1.7593784363593467e-05, "loss": 2.1432, "step": 10092 }, { "epoch": 0.25, "learning_rate": 1.7593264905880156e-05, "loss": 2.2583, "step": 10093 }, { "epoch": 0.25, "learning_rate": 1.7592745399772495e-05, "loss": 2.2039, "step": 10094 }, { "epoch": 0.25, "learning_rate": 1.7592225845273794e-05, "loss": 2.071, "step": 10095 }, { "epoch": 0.25, "learning_rate": 1.759170624238736e-05, "loss": 2.2431, "step": 10096 }, { "epoch": 0.25, "learning_rate": 1.759118659111651e-05, "loss": 2.0575, "step": 10097 }, { "epoch": 0.25, "learning_rate": 1.7590666891464553e-05, "loss": 2.2142, "step": 10098 }, { "epoch": 0.25, "learning_rate": 1.7590147143434806e-05, "loss": 2.1724, "step": 10099 }, { "epoch": 0.25, "learning_rate": 1.7589627347030577e-05, "loss": 2.1425, "step": 10100 }, { "epoch": 0.25, "learning_rate": 1.7589107502255172e-05, "loss": 2.2429, "step": 10101 }, { "epoch": 0.25, "learning_rate": 1.758858760911192e-05, "loss": 2.4118, "step": 10102 }, { "epoch": 0.25, "learning_rate": 1.7588067667604123e-05, "loss": 2.1371, "step": 10103 }, { "epoch": 0.25, "learning_rate": 1.75875476777351e-05, "loss": 2.3169, "step": 10104 }, { "epoch": 0.25, "learning_rate": 1.7587027639508157e-05, "loss": 2.0326, "step": 10105 }, { "epoch": 0.25, "learning_rate": 1.7586507552926622e-05, "loss": 2.2077, "step": 10106 }, { "epoch": 0.25, "learning_rate": 1.7585987417993794e-05, "loss": 2.1293, "step": 10107 }, { "epoch": 0.25, "learning_rate": 1.7585467234713002e-05, "loss": 2.2122, "step": 10108 }, { "epoch": 0.25, "learning_rate": 1.7584947003087554e-05, "loss": 2.1506, "step": 10109 }, { "epoch": 0.25, "learning_rate": 1.7584426723120767e-05, "loss": 2.2272, "step": 10110 }, { "epoch": 0.25, "learning_rate": 1.7583906394815957e-05, "loss": 2.1814, "step": 10111 }, { "epoch": 0.25, "learning_rate": 1.758338601817644e-05, "loss": 2.1141, "step": 10112 }, { "epoch": 0.25, "learning_rate": 1.758286559320553e-05, "loss": 2.2132, "step": 10113 }, { "epoch": 0.25, "learning_rate": 1.758234511990655e-05, "loss": 2.0423, "step": 10114 }, { "epoch": 0.25, "learning_rate": 1.758182459828281e-05, "loss": 2.2686, "step": 10115 }, { "epoch": 0.25, "learning_rate": 1.7581304028337634e-05, "loss": 2.3644, "step": 10116 }, { "epoch": 0.25, "learning_rate": 1.7580783410074335e-05, "loss": 2.0488, "step": 10117 }, { "epoch": 0.25, "learning_rate": 1.7580262743496232e-05, "loss": 2.123, "step": 10118 }, { "epoch": 0.25, "learning_rate": 1.7579742028606646e-05, "loss": 2.1751, "step": 10119 }, { "epoch": 0.25, "learning_rate": 1.7579221265408894e-05, "loss": 2.0285, "step": 10120 }, { "epoch": 0.25, "learning_rate": 1.7578700453906293e-05, "loss": 2.0807, "step": 10121 }, { "epoch": 0.25, "learning_rate": 1.7578179594102167e-05, "loss": 2.0913, "step": 10122 }, { "epoch": 0.25, "learning_rate": 1.7577658685999832e-05, "loss": 2.2094, "step": 10123 }, { "epoch": 0.25, "learning_rate": 1.7577137729602606e-05, "loss": 2.1602, "step": 10124 }, { "epoch": 0.25, "learning_rate": 1.7576616724913816e-05, "loss": 2.2747, "step": 10125 }, { "epoch": 0.25, "learning_rate": 1.7576095671936775e-05, "loss": 2.2372, "step": 10126 }, { "epoch": 0.25, "learning_rate": 1.7575574570674808e-05, "loss": 2.1866, "step": 10127 }, { "epoch": 0.25, "learning_rate": 1.7575053421131238e-05, "loss": 1.985, "step": 10128 }, { "epoch": 0.25, "learning_rate": 1.757453222330938e-05, "loss": 2.2897, "step": 10129 }, { "epoch": 0.25, "learning_rate": 1.7574010977212562e-05, "loss": 2.1451, "step": 10130 }, { "epoch": 0.25, "learning_rate": 1.7573489682844105e-05, "loss": 2.1986, "step": 10131 }, { "epoch": 0.25, "learning_rate": 1.757296834020733e-05, "loss": 2.2635, "step": 10132 }, { "epoch": 0.25, "learning_rate": 1.7572446949305556e-05, "loss": 2.1626, "step": 10133 }, { "epoch": 0.25, "learning_rate": 1.757192551014211e-05, "loss": 2.0619, "step": 10134 }, { "epoch": 0.25, "learning_rate": 1.757140402272032e-05, "loss": 2.1376, "step": 10135 }, { "epoch": 0.25, "learning_rate": 1.75708824870435e-05, "loss": 2.2168, "step": 10136 }, { "epoch": 0.25, "learning_rate": 1.7570360903114983e-05, "loss": 2.0709, "step": 10137 }, { "epoch": 0.25, "learning_rate": 1.7569839270938083e-05, "loss": 2.0137, "step": 10138 }, { "epoch": 0.25, "learning_rate": 1.7569317590516135e-05, "loss": 2.1339, "step": 10139 }, { "epoch": 0.25, "learning_rate": 1.756879586185246e-05, "loss": 2.0875, "step": 10140 }, { "epoch": 0.25, "learning_rate": 1.756827408495038e-05, "loss": 2.1892, "step": 10141 }, { "epoch": 0.25, "learning_rate": 1.7567752259813225e-05, "loss": 2.3161, "step": 10142 }, { "epoch": 0.25, "learning_rate": 1.7567230386444317e-05, "loss": 2.2347, "step": 10143 }, { "epoch": 0.25, "learning_rate": 1.7566708464846983e-05, "loss": 2.1013, "step": 10144 }, { "epoch": 0.25, "learning_rate": 1.756618649502455e-05, "loss": 2.2083, "step": 10145 }, { "epoch": 0.25, "learning_rate": 1.7565664476980345e-05, "loss": 2.1909, "step": 10146 }, { "epoch": 0.25, "learning_rate": 1.7565142410717694e-05, "loss": 2.1357, "step": 10147 }, { "epoch": 0.25, "learning_rate": 1.7564620296239927e-05, "loss": 2.1711, "step": 10148 }, { "epoch": 0.25, "learning_rate": 1.7564098133550368e-05, "loss": 2.0706, "step": 10149 }, { "epoch": 0.25, "learning_rate": 1.7563575922652346e-05, "loss": 2.1716, "step": 10150 }, { "epoch": 0.25, "learning_rate": 1.756305366354919e-05, "loss": 2.2624, "step": 10151 }, { "epoch": 0.25, "learning_rate": 1.756253135624423e-05, "loss": 2.0997, "step": 10152 }, { "epoch": 0.25, "learning_rate": 1.7562009000740795e-05, "loss": 2.2357, "step": 10153 }, { "epoch": 0.25, "learning_rate": 1.7561486597042208e-05, "loss": 2.2188, "step": 10154 }, { "epoch": 0.25, "learning_rate": 1.7560964145151803e-05, "loss": 2.1874, "step": 10155 }, { "epoch": 0.25, "learning_rate": 1.756044164507291e-05, "loss": 2.1593, "step": 10156 }, { "epoch": 0.25, "learning_rate": 1.7559919096808862e-05, "loss": 2.1522, "step": 10157 }, { "epoch": 0.25, "learning_rate": 1.7559396500362985e-05, "loss": 1.9975, "step": 10158 }, { "epoch": 0.25, "learning_rate": 1.7558873855738605e-05, "loss": 2.1322, "step": 10159 }, { "epoch": 0.25, "learning_rate": 1.7558351162939063e-05, "loss": 2.1056, "step": 10160 }, { "epoch": 0.25, "learning_rate": 1.7557828421967687e-05, "loss": 2.005, "step": 10161 }, { "epoch": 0.25, "learning_rate": 1.7557305632827804e-05, "loss": 1.9685, "step": 10162 }, { "epoch": 0.25, "learning_rate": 1.755678279552275e-05, "loss": 2.2458, "step": 10163 }, { "epoch": 0.25, "learning_rate": 1.755625991005586e-05, "loss": 2.2219, "step": 10164 }, { "epoch": 0.25, "learning_rate": 1.7555736976430458e-05, "loss": 2.2706, "step": 10165 }, { "epoch": 0.25, "learning_rate": 1.755521399464989e-05, "loss": 2.0405, "step": 10166 }, { "epoch": 0.25, "learning_rate": 1.755469096471747e-05, "loss": 2.1076, "step": 10167 }, { "epoch": 0.25, "learning_rate": 1.7554167886636547e-05, "loss": 2.0666, "step": 10168 }, { "epoch": 0.25, "learning_rate": 1.7553644760410453e-05, "loss": 2.1016, "step": 10169 }, { "epoch": 0.25, "learning_rate": 1.7553121586042517e-05, "loss": 1.9564, "step": 10170 }, { "epoch": 0.25, "learning_rate": 1.7552598363536075e-05, "loss": 2.1288, "step": 10171 }, { "epoch": 0.25, "learning_rate": 1.7552075092894465e-05, "loss": 2.0975, "step": 10172 }, { "epoch": 0.25, "learning_rate": 1.7551551774121018e-05, "loss": 2.2176, "step": 10173 }, { "epoch": 0.25, "learning_rate": 1.7551028407219067e-05, "loss": 2.1737, "step": 10174 }, { "epoch": 0.25, "learning_rate": 1.7550504992191957e-05, "loss": 2.1633, "step": 10175 }, { "epoch": 0.25, "learning_rate": 1.7549981529043014e-05, "loss": 2.2818, "step": 10176 }, { "epoch": 0.25, "learning_rate": 1.754945801777558e-05, "loss": 2.2547, "step": 10177 }, { "epoch": 0.25, "learning_rate": 1.754893445839299e-05, "loss": 2.0499, "step": 10178 }, { "epoch": 0.25, "learning_rate": 1.7548410850898582e-05, "loss": 2.3588, "step": 10179 }, { "epoch": 0.25, "learning_rate": 1.754788719529569e-05, "loss": 2.1231, "step": 10180 }, { "epoch": 0.25, "learning_rate": 1.754736349158765e-05, "loss": 2.0823, "step": 10181 }, { "epoch": 0.25, "learning_rate": 1.7546839739777806e-05, "loss": 2.1013, "step": 10182 }, { "epoch": 0.25, "learning_rate": 1.7546315939869492e-05, "loss": 2.1973, "step": 10183 }, { "epoch": 0.25, "learning_rate": 1.7545792091866048e-05, "loss": 2.1912, "step": 10184 }, { "epoch": 0.25, "learning_rate": 1.7545268195770814e-05, "loss": 2.1922, "step": 10185 }, { "epoch": 0.25, "learning_rate": 1.7544744251587124e-05, "loss": 2.2476, "step": 10186 }, { "epoch": 0.25, "learning_rate": 1.754422025931832e-05, "loss": 2.1066, "step": 10187 }, { "epoch": 0.25, "learning_rate": 1.7543696218967743e-05, "loss": 2.2713, "step": 10188 }, { "epoch": 0.25, "learning_rate": 1.754317213053873e-05, "loss": 2.1786, "step": 10189 }, { "epoch": 0.25, "learning_rate": 1.7542647994034623e-05, "loss": 2.3029, "step": 10190 }, { "epoch": 0.25, "learning_rate": 1.7542123809458764e-05, "loss": 2.0966, "step": 10191 }, { "epoch": 0.25, "learning_rate": 1.754159957681449e-05, "loss": 2.1662, "step": 10192 }, { "epoch": 0.25, "learning_rate": 1.7541075296105146e-05, "loss": 1.9978, "step": 10193 }, { "epoch": 0.25, "learning_rate": 1.7540550967334068e-05, "loss": 2.1942, "step": 10194 }, { "epoch": 0.25, "learning_rate": 1.7540026590504607e-05, "loss": 2.2779, "step": 10195 }, { "epoch": 0.25, "learning_rate": 1.7539502165620097e-05, "loss": 2.0965, "step": 10196 }, { "epoch": 0.25, "learning_rate": 1.753897769268388e-05, "loss": 2.0607, "step": 10197 }, { "epoch": 0.25, "learning_rate": 1.7538453171699304e-05, "loss": 2.1896, "step": 10198 }, { "epoch": 0.25, "learning_rate": 1.753792860266971e-05, "loss": 2.2012, "step": 10199 }, { "epoch": 0.25, "learning_rate": 1.7537403985598438e-05, "loss": 2.267, "step": 10200 }, { "epoch": 0.25, "learning_rate": 1.7536879320488835e-05, "loss": 2.0169, "step": 10201 }, { "epoch": 0.25, "learning_rate": 1.7536354607344244e-05, "loss": 2.3026, "step": 10202 }, { "epoch": 0.25, "learning_rate": 1.753582984616801e-05, "loss": 2.1136, "step": 10203 }, { "epoch": 0.25, "learning_rate": 1.7535305036963476e-05, "loss": 2.2345, "step": 10204 }, { "epoch": 0.25, "learning_rate": 1.7534780179733987e-05, "loss": 2.2974, "step": 10205 }, { "epoch": 0.25, "learning_rate": 1.7534255274482886e-05, "loss": 2.2065, "step": 10206 }, { "epoch": 0.25, "learning_rate": 1.753373032121352e-05, "loss": 2.2578, "step": 10207 }, { "epoch": 0.25, "learning_rate": 1.753320531992924e-05, "loss": 2.0856, "step": 10208 }, { "epoch": 0.25, "learning_rate": 1.7532680270633387e-05, "loss": 2.0408, "step": 10209 }, { "epoch": 0.25, "learning_rate": 1.753215517332931e-05, "loss": 2.0694, "step": 10210 }, { "epoch": 0.25, "learning_rate": 1.7531630028020347e-05, "loss": 2.2881, "step": 10211 }, { "epoch": 0.25, "learning_rate": 1.7531104834709854e-05, "loss": 2.144, "step": 10212 }, { "epoch": 0.25, "learning_rate": 1.7530579593401176e-05, "loss": 2.2129, "step": 10213 }, { "epoch": 0.25, "learning_rate": 1.753005430409766e-05, "loss": 2.309, "step": 10214 }, { "epoch": 0.25, "learning_rate": 1.7529528966802656e-05, "loss": 2.1574, "step": 10215 }, { "epoch": 0.25, "learning_rate": 1.752900358151951e-05, "loss": 2.2821, "step": 10216 }, { "epoch": 0.25, "learning_rate": 1.7528478148251565e-05, "loss": 2.1446, "step": 10217 }, { "epoch": 0.25, "learning_rate": 1.752795266700218e-05, "loss": 2.0768, "step": 10218 }, { "epoch": 0.25, "learning_rate": 1.75274271377747e-05, "loss": 2.1498, "step": 10219 }, { "epoch": 0.25, "learning_rate": 1.7526901560572472e-05, "loss": 2.2477, "step": 10220 }, { "epoch": 0.25, "learning_rate": 1.7526375935398847e-05, "loss": 2.1342, "step": 10221 }, { "epoch": 0.25, "learning_rate": 1.7525850262257177e-05, "loss": 2.0571, "step": 10222 }, { "epoch": 0.25, "learning_rate": 1.752532454115081e-05, "loss": 2.0977, "step": 10223 }, { "epoch": 0.25, "learning_rate": 1.7524798772083094e-05, "loss": 1.95, "step": 10224 }, { "epoch": 0.25, "learning_rate": 1.7524272955057388e-05, "loss": 2.2078, "step": 10225 }, { "epoch": 0.25, "learning_rate": 1.7523747090077036e-05, "loss": 2.3567, "step": 10226 }, { "epoch": 0.25, "learning_rate": 1.7523221177145393e-05, "loss": 2.2676, "step": 10227 }, { "epoch": 0.25, "learning_rate": 1.7522695216265808e-05, "loss": 2.1543, "step": 10228 }, { "epoch": 0.25, "learning_rate": 1.7522169207441637e-05, "loss": 2.2636, "step": 10229 }, { "epoch": 0.25, "learning_rate": 1.7521643150676227e-05, "loss": 2.0454, "step": 10230 }, { "epoch": 0.25, "learning_rate": 1.7521117045972937e-05, "loss": 2.111, "step": 10231 }, { "epoch": 0.25, "learning_rate": 1.7520590893335118e-05, "loss": 2.1698, "step": 10232 }, { "epoch": 0.25, "learning_rate": 1.752006469276612e-05, "loss": 2.0282, "step": 10233 }, { "epoch": 0.25, "learning_rate": 1.7519538444269297e-05, "loss": 2.0396, "step": 10234 }, { "epoch": 0.25, "learning_rate": 1.7519012147848007e-05, "loss": 2.281, "step": 10235 }, { "epoch": 0.25, "learning_rate": 1.7518485803505605e-05, "loss": 2.108, "step": 10236 }, { "epoch": 0.25, "learning_rate": 1.751795941124544e-05, "loss": 2.2099, "step": 10237 }, { "epoch": 0.25, "learning_rate": 1.751743297107087e-05, "loss": 2.0655, "step": 10238 }, { "epoch": 0.25, "learning_rate": 1.751690648298525e-05, "loss": 2.2125, "step": 10239 }, { "epoch": 0.25, "learning_rate": 1.7516379946991935e-05, "loss": 2.3409, "step": 10240 }, { "epoch": 0.25, "learning_rate": 1.751585336309428e-05, "loss": 2.1847, "step": 10241 }, { "epoch": 0.25, "learning_rate": 1.7515326731295642e-05, "loss": 2.1758, "step": 10242 }, { "epoch": 0.25, "learning_rate": 1.751480005159938e-05, "loss": 2.2552, "step": 10243 }, { "epoch": 0.25, "learning_rate": 1.7514273324008848e-05, "loss": 2.0245, "step": 10244 }, { "epoch": 0.25, "learning_rate": 1.7513746548527404e-05, "loss": 2.1906, "step": 10245 }, { "epoch": 0.25, "learning_rate": 1.75132197251584e-05, "loss": 2.0643, "step": 10246 }, { "epoch": 0.25, "learning_rate": 1.7512692853905204e-05, "loss": 2.186, "step": 10247 }, { "epoch": 0.25, "learning_rate": 1.7512165934771163e-05, "loss": 2.2184, "step": 10248 }, { "epoch": 0.25, "learning_rate": 1.7511638967759643e-05, "loss": 2.2224, "step": 10249 }, { "epoch": 0.25, "learning_rate": 1.7511111952874e-05, "loss": 2.3368, "step": 10250 }, { "epoch": 0.25, "learning_rate": 1.751058489011759e-05, "loss": 2.31, "step": 10251 }, { "epoch": 0.25, "learning_rate": 1.7510057779493782e-05, "loss": 2.1991, "step": 10252 }, { "epoch": 0.25, "learning_rate": 1.7509530621005923e-05, "loss": 2.0048, "step": 10253 }, { "epoch": 0.25, "learning_rate": 1.7509003414657376e-05, "loss": 2.1703, "step": 10254 }, { "epoch": 0.25, "learning_rate": 1.7508476160451508e-05, "loss": 2.1744, "step": 10255 }, { "epoch": 0.25, "learning_rate": 1.750794885839167e-05, "loss": 2.1135, "step": 10256 }, { "epoch": 0.25, "learning_rate": 1.750742150848123e-05, "loss": 2.1521, "step": 10257 }, { "epoch": 0.25, "learning_rate": 1.7506894110723547e-05, "loss": 2.0661, "step": 10258 }, { "epoch": 0.25, "learning_rate": 1.750636666512198e-05, "loss": 2.1255, "step": 10259 }, { "epoch": 0.25, "learning_rate": 1.750583917167989e-05, "loss": 2.1666, "step": 10260 }, { "epoch": 0.25, "learning_rate": 1.7505311630400642e-05, "loss": 2.2146, "step": 10261 }, { "epoch": 0.25, "learning_rate": 1.75047840412876e-05, "loss": 2.3699, "step": 10262 }, { "epoch": 0.25, "learning_rate": 1.750425640434412e-05, "loss": 2.2482, "step": 10263 }, { "epoch": 0.25, "learning_rate": 1.750372871957357e-05, "loss": 2.2126, "step": 10264 }, { "epoch": 0.25, "learning_rate": 1.7503200986979313e-05, "loss": 2.1304, "step": 10265 }, { "epoch": 0.25, "learning_rate": 1.750267320656471e-05, "loss": 2.1754, "step": 10266 }, { "epoch": 0.25, "learning_rate": 1.7502145378333123e-05, "loss": 2.0777, "step": 10267 }, { "epoch": 0.25, "learning_rate": 1.750161750228792e-05, "loss": 2.2421, "step": 10268 }, { "epoch": 0.25, "learning_rate": 1.7501089578432465e-05, "loss": 2.0652, "step": 10269 }, { "epoch": 0.25, "learning_rate": 1.750056160677012e-05, "loss": 2.0772, "step": 10270 }, { "epoch": 0.25, "learning_rate": 1.7500033587304252e-05, "loss": 2.2435, "step": 10271 }, { "epoch": 0.25, "learning_rate": 1.7499505520038227e-05, "loss": 2.2872, "step": 10272 }, { "epoch": 0.25, "learning_rate": 1.7498977404975407e-05, "loss": 2.2178, "step": 10273 }, { "epoch": 0.25, "learning_rate": 1.7498449242119163e-05, "loss": 2.2194, "step": 10274 }, { "epoch": 0.25, "learning_rate": 1.749792103147286e-05, "loss": 2.1934, "step": 10275 }, { "epoch": 0.25, "learning_rate": 1.7497392773039857e-05, "loss": 2.2515, "step": 10276 }, { "epoch": 0.25, "learning_rate": 1.7496864466823527e-05, "loss": 2.1806, "step": 10277 }, { "epoch": 0.25, "learning_rate": 1.749633611282724e-05, "loss": 2.0787, "step": 10278 }, { "epoch": 0.25, "learning_rate": 1.749580771105436e-05, "loss": 2.2502, "step": 10279 }, { "epoch": 0.25, "learning_rate": 1.749527926150825e-05, "loss": 2.0716, "step": 10280 }, { "epoch": 0.25, "learning_rate": 1.749475076419229e-05, "loss": 2.108, "step": 10281 }, { "epoch": 0.25, "learning_rate": 1.7494222219109835e-05, "loss": 2.0849, "step": 10282 }, { "epoch": 0.25, "learning_rate": 1.749369362626426e-05, "loss": 2.2161, "step": 10283 }, { "epoch": 0.25, "learning_rate": 1.7493164985658937e-05, "loss": 2.1956, "step": 10284 }, { "epoch": 0.25, "learning_rate": 1.749263629729723e-05, "loss": 2.312, "step": 10285 }, { "epoch": 0.25, "learning_rate": 1.7492107561182506e-05, "loss": 2.1112, "step": 10286 }, { "epoch": 0.25, "learning_rate": 1.7491578777318145e-05, "loss": 2.2306, "step": 10287 }, { "epoch": 0.25, "learning_rate": 1.7491049945707508e-05, "loss": 2.109, "step": 10288 }, { "epoch": 0.25, "learning_rate": 1.7490521066353968e-05, "loss": 2.0135, "step": 10289 }, { "epoch": 0.25, "learning_rate": 1.74899921392609e-05, "loss": 2.2516, "step": 10290 }, { "epoch": 0.25, "learning_rate": 1.7489463164431665e-05, "loss": 2.1782, "step": 10291 }, { "epoch": 0.25, "learning_rate": 1.7488934141869647e-05, "loss": 2.1888, "step": 10292 }, { "epoch": 0.25, "learning_rate": 1.7488405071578208e-05, "loss": 2.1326, "step": 10293 }, { "epoch": 0.25, "learning_rate": 1.748787595356072e-05, "loss": 2.1168, "step": 10294 }, { "epoch": 0.25, "learning_rate": 1.7487346787820564e-05, "loss": 2.101, "step": 10295 }, { "epoch": 0.25, "learning_rate": 1.7486817574361105e-05, "loss": 1.9126, "step": 10296 }, { "epoch": 0.25, "learning_rate": 1.7486288313185715e-05, "loss": 2.0926, "step": 10297 }, { "epoch": 0.25, "learning_rate": 1.748575900429777e-05, "loss": 2.2464, "step": 10298 }, { "epoch": 0.25, "learning_rate": 1.7485229647700646e-05, "loss": 2.2095, "step": 10299 }, { "epoch": 0.25, "learning_rate": 1.7484700243397715e-05, "loss": 2.2034, "step": 10300 }, { "epoch": 0.25, "learning_rate": 1.7484170791392348e-05, "loss": 2.2179, "step": 10301 }, { "epoch": 0.25, "learning_rate": 1.748364129168792e-05, "loss": 2.0788, "step": 10302 }, { "epoch": 0.25, "learning_rate": 1.7483111744287812e-05, "loss": 2.2669, "step": 10303 }, { "epoch": 0.25, "learning_rate": 1.748258214919539e-05, "loss": 2.2017, "step": 10304 }, { "epoch": 0.25, "learning_rate": 1.7482052506414035e-05, "loss": 2.1106, "step": 10305 }, { "epoch": 0.25, "learning_rate": 1.7481522815947118e-05, "loss": 2.2363, "step": 10306 }, { "epoch": 0.25, "learning_rate": 1.748099307779802e-05, "loss": 2.2246, "step": 10307 }, { "epoch": 0.25, "learning_rate": 1.7480463291970114e-05, "loss": 2.2843, "step": 10308 }, { "epoch": 0.25, "learning_rate": 1.7479933458466783e-05, "loss": 2.0927, "step": 10309 }, { "epoch": 0.25, "learning_rate": 1.7479403577291394e-05, "loss": 2.1932, "step": 10310 }, { "epoch": 0.25, "learning_rate": 1.747887364844733e-05, "loss": 2.1911, "step": 10311 }, { "epoch": 0.25, "learning_rate": 1.7478343671937965e-05, "loss": 2.0924, "step": 10312 }, { "epoch": 0.25, "learning_rate": 1.747781364776668e-05, "loss": 2.1778, "step": 10313 }, { "epoch": 0.25, "learning_rate": 1.747728357593685e-05, "loss": 2.1698, "step": 10314 }, { "epoch": 0.25, "learning_rate": 1.747675345645186e-05, "loss": 2.1615, "step": 10315 }, { "epoch": 0.25, "learning_rate": 1.7476223289315077e-05, "loss": 2.2006, "step": 10316 }, { "epoch": 0.25, "learning_rate": 1.7475693074529894e-05, "loss": 2.1492, "step": 10317 }, { "epoch": 0.25, "learning_rate": 1.7475162812099676e-05, "loss": 2.1171, "step": 10318 }, { "epoch": 0.25, "learning_rate": 1.7474632502027813e-05, "loss": 2.1486, "step": 10319 }, { "epoch": 0.25, "learning_rate": 1.7474102144317678e-05, "loss": 1.9759, "step": 10320 }, { "epoch": 0.25, "learning_rate": 1.7473571738972657e-05, "loss": 2.2531, "step": 10321 }, { "epoch": 0.25, "learning_rate": 1.7473041285996127e-05, "loss": 2.261, "step": 10322 }, { "epoch": 0.25, "learning_rate": 1.747251078539147e-05, "loss": 2.1709, "step": 10323 }, { "epoch": 0.25, "learning_rate": 1.7471980237162066e-05, "loss": 2.252, "step": 10324 }, { "epoch": 0.25, "learning_rate": 1.7471449641311298e-05, "loss": 2.2251, "step": 10325 }, { "epoch": 0.25, "learning_rate": 1.7470918997842544e-05, "loss": 2.0495, "step": 10326 }, { "epoch": 0.25, "learning_rate": 1.747038830675919e-05, "loss": 2.3054, "step": 10327 }, { "epoch": 0.25, "learning_rate": 1.7469857568064617e-05, "loss": 2.1495, "step": 10328 }, { "epoch": 0.25, "learning_rate": 1.7469326781762208e-05, "loss": 2.1154, "step": 10329 }, { "epoch": 0.25, "learning_rate": 1.7468795947855344e-05, "loss": 2.1406, "step": 10330 }, { "epoch": 0.25, "learning_rate": 1.746826506634741e-05, "loss": 2.3021, "step": 10331 }, { "epoch": 0.25, "learning_rate": 1.746773413724179e-05, "loss": 2.2258, "step": 10332 }, { "epoch": 0.25, "learning_rate": 1.7467203160541864e-05, "loss": 2.2541, "step": 10333 }, { "epoch": 0.25, "learning_rate": 1.7466672136251023e-05, "loss": 2.0656, "step": 10334 }, { "epoch": 0.25, "learning_rate": 1.746614106437264e-05, "loss": 2.1612, "step": 10335 }, { "epoch": 0.25, "learning_rate": 1.7465609944910114e-05, "loss": 2.3095, "step": 10336 }, { "epoch": 0.25, "learning_rate": 1.7465078777866818e-05, "loss": 2.0503, "step": 10337 }, { "epoch": 0.25, "learning_rate": 1.7464547563246146e-05, "loss": 2.3453, "step": 10338 }, { "epoch": 0.25, "learning_rate": 1.746401630105148e-05, "loss": 2.0878, "step": 10339 }, { "epoch": 0.25, "learning_rate": 1.74634849912862e-05, "loss": 2.2823, "step": 10340 }, { "epoch": 0.25, "learning_rate": 1.74629536339537e-05, "loss": 2.0707, "step": 10341 }, { "epoch": 0.25, "learning_rate": 1.7462422229057368e-05, "loss": 2.1421, "step": 10342 }, { "epoch": 0.25, "learning_rate": 1.7461890776600585e-05, "loss": 2.0524, "step": 10343 }, { "epoch": 0.25, "learning_rate": 1.7461359276586743e-05, "loss": 2.0393, "step": 10344 }, { "epoch": 0.25, "learning_rate": 1.7460827729019222e-05, "loss": 2.1371, "step": 10345 }, { "epoch": 0.26, "learning_rate": 1.7460296133901416e-05, "loss": 2.1032, "step": 10346 }, { "epoch": 0.26, "learning_rate": 1.7459764491236714e-05, "loss": 2.1609, "step": 10347 }, { "epoch": 0.26, "learning_rate": 1.74592328010285e-05, "loss": 2.2174, "step": 10348 }, { "epoch": 0.26, "learning_rate": 1.7458701063280163e-05, "loss": 2.174, "step": 10349 }, { "epoch": 0.26, "learning_rate": 1.7458169277995095e-05, "loss": 2.3386, "step": 10350 }, { "epoch": 0.26, "learning_rate": 1.7457637445176683e-05, "loss": 2.0576, "step": 10351 }, { "epoch": 0.26, "learning_rate": 1.7457105564828318e-05, "loss": 2.3047, "step": 10352 }, { "epoch": 0.26, "learning_rate": 1.745657363695339e-05, "loss": 2.1301, "step": 10353 }, { "epoch": 0.26, "learning_rate": 1.745604166155529e-05, "loss": 2.2432, "step": 10354 }, { "epoch": 0.26, "learning_rate": 1.74555096386374e-05, "loss": 2.0557, "step": 10355 }, { "epoch": 0.26, "learning_rate": 1.745497756820312e-05, "loss": 2.1639, "step": 10356 }, { "epoch": 0.26, "learning_rate": 1.745444545025584e-05, "loss": 2.1225, "step": 10357 }, { "epoch": 0.26, "learning_rate": 1.745391328479895e-05, "loss": 2.1086, "step": 10358 }, { "epoch": 0.26, "learning_rate": 1.745338107183584e-05, "loss": 2.1856, "step": 10359 }, { "epoch": 0.26, "learning_rate": 1.7452848811369904e-05, "loss": 2.1343, "step": 10360 }, { "epoch": 0.26, "learning_rate": 1.745231650340453e-05, "loss": 2.1198, "step": 10361 }, { "epoch": 0.26, "learning_rate": 1.745178414794312e-05, "loss": 2.17, "step": 10362 }, { "epoch": 0.26, "learning_rate": 1.745125174498906e-05, "loss": 1.9592, "step": 10363 }, { "epoch": 0.26, "learning_rate": 1.7450719294545747e-05, "loss": 2.2974, "step": 10364 }, { "epoch": 0.26, "learning_rate": 1.7450186796616566e-05, "loss": 2.1263, "step": 10365 }, { "epoch": 0.26, "learning_rate": 1.7449654251204918e-05, "loss": 2.1779, "step": 10366 }, { "epoch": 0.26, "learning_rate": 1.7449121658314196e-05, "loss": 2.0255, "step": 10367 }, { "epoch": 0.26, "learning_rate": 1.7448589017947796e-05, "loss": 2.1319, "step": 10368 }, { "epoch": 0.26, "learning_rate": 1.744805633010911e-05, "loss": 2.2084, "step": 10369 }, { "epoch": 0.26, "learning_rate": 1.7447523594801533e-05, "loss": 2.264, "step": 10370 }, { "epoch": 0.26, "learning_rate": 1.744699081202846e-05, "loss": 2.2951, "step": 10371 }, { "epoch": 0.26, "learning_rate": 1.744645798179329e-05, "loss": 2.3434, "step": 10372 }, { "epoch": 0.26, "learning_rate": 1.7445925104099414e-05, "loss": 2.1997, "step": 10373 }, { "epoch": 0.26, "learning_rate": 1.7445392178950234e-05, "loss": 2.092, "step": 10374 }, { "epoch": 0.26, "learning_rate": 1.7444859206349143e-05, "loss": 2.1732, "step": 10375 }, { "epoch": 0.26, "learning_rate": 1.7444326186299533e-05, "loss": 2.0446, "step": 10376 }, { "epoch": 0.26, "learning_rate": 1.7443793118804812e-05, "loss": 2.1371, "step": 10377 }, { "epoch": 0.26, "learning_rate": 1.744326000386837e-05, "loss": 2.2308, "step": 10378 }, { "epoch": 0.26, "learning_rate": 1.7442726841493605e-05, "loss": 2.1289, "step": 10379 }, { "epoch": 0.26, "learning_rate": 1.7442193631683916e-05, "loss": 2.044, "step": 10380 }, { "epoch": 0.26, "learning_rate": 1.7441660374442705e-05, "loss": 2.2348, "step": 10381 }, { "epoch": 0.26, "learning_rate": 1.7441127069773362e-05, "loss": 2.1678, "step": 10382 }, { "epoch": 0.26, "learning_rate": 1.7440593717679294e-05, "loss": 2.0159, "step": 10383 }, { "epoch": 0.26, "learning_rate": 1.7440060318163898e-05, "loss": 2.1878, "step": 10384 }, { "epoch": 0.26, "learning_rate": 1.7439526871230574e-05, "loss": 2.2578, "step": 10385 }, { "epoch": 0.26, "learning_rate": 1.743899337688272e-05, "loss": 2.0752, "step": 10386 }, { "epoch": 0.26, "learning_rate": 1.7438459835123736e-05, "loss": 2.0751, "step": 10387 }, { "epoch": 0.26, "learning_rate": 1.743792624595702e-05, "loss": 2.1969, "step": 10388 }, { "epoch": 0.26, "learning_rate": 1.743739260938598e-05, "loss": 2.2043, "step": 10389 }, { "epoch": 0.26, "learning_rate": 1.7436858925414016e-05, "loss": 2.1636, "step": 10390 }, { "epoch": 0.26, "learning_rate": 1.7436325194044524e-05, "loss": 2.0768, "step": 10391 }, { "epoch": 0.26, "learning_rate": 1.7435791415280906e-05, "loss": 2.1756, "step": 10392 }, { "epoch": 0.26, "learning_rate": 1.7435257589126568e-05, "loss": 2.3347, "step": 10393 }, { "epoch": 0.26, "learning_rate": 1.743472371558491e-05, "loss": 2.0428, "step": 10394 }, { "epoch": 0.26, "learning_rate": 1.7434189794659335e-05, "loss": 2.1926, "step": 10395 }, { "epoch": 0.26, "learning_rate": 1.7433655826353243e-05, "loss": 2.2708, "step": 10396 }, { "epoch": 0.26, "learning_rate": 1.743312181067004e-05, "loss": 2.1092, "step": 10397 }, { "epoch": 0.26, "learning_rate": 1.7432587747613134e-05, "loss": 2.1282, "step": 10398 }, { "epoch": 0.26, "learning_rate": 1.7432053637185922e-05, "loss": 2.1513, "step": 10399 }, { "epoch": 0.26, "learning_rate": 1.7431519479391808e-05, "loss": 2.2038, "step": 10400 }, { "epoch": 0.26, "learning_rate": 1.74309852742342e-05, "loss": 2.1968, "step": 10401 }, { "epoch": 0.26, "learning_rate": 1.7430451021716503e-05, "loss": 2.2804, "step": 10402 }, { "epoch": 0.26, "learning_rate": 1.7429916721842116e-05, "loss": 2.1802, "step": 10403 }, { "epoch": 0.26, "learning_rate": 1.7429382374614453e-05, "loss": 2.1258, "step": 10404 }, { "epoch": 0.26, "learning_rate": 1.742884798003691e-05, "loss": 2.15, "step": 10405 }, { "epoch": 0.26, "learning_rate": 1.7428313538112902e-05, "loss": 1.9766, "step": 10406 }, { "epoch": 0.26, "learning_rate": 1.742777904884583e-05, "loss": 1.9284, "step": 10407 }, { "epoch": 0.26, "learning_rate": 1.74272445122391e-05, "loss": 2.0146, "step": 10408 }, { "epoch": 0.26, "learning_rate": 1.7426709928296124e-05, "loss": 2.1701, "step": 10409 }, { "epoch": 0.26, "learning_rate": 1.74261752970203e-05, "loss": 2.1789, "step": 10410 }, { "epoch": 0.26, "learning_rate": 1.7425640618415042e-05, "loss": 2.2623, "step": 10411 }, { "epoch": 0.26, "learning_rate": 1.742510589248376e-05, "loss": 2.3533, "step": 10412 }, { "epoch": 0.26, "learning_rate": 1.7424571119229855e-05, "loss": 2.1484, "step": 10413 }, { "epoch": 0.26, "learning_rate": 1.742403629865674e-05, "loss": 2.2286, "step": 10414 }, { "epoch": 0.26, "learning_rate": 1.742350143076782e-05, "loss": 2.164, "step": 10415 }, { "epoch": 0.26, "learning_rate": 1.742296651556651e-05, "loss": 2.1892, "step": 10416 }, { "epoch": 0.26, "learning_rate": 1.7422431553056217e-05, "loss": 2.0375, "step": 10417 }, { "epoch": 0.26, "learning_rate": 1.742189654324034e-05, "loss": 2.0968, "step": 10418 }, { "epoch": 0.26, "learning_rate": 1.7421361486122307e-05, "loss": 2.1647, "step": 10419 }, { "epoch": 0.26, "learning_rate": 1.7420826381705513e-05, "loss": 2.1639, "step": 10420 }, { "epoch": 0.26, "learning_rate": 1.7420291229993374e-05, "loss": 2.1829, "step": 10421 }, { "epoch": 0.26, "learning_rate": 1.7419756030989302e-05, "loss": 2.0654, "step": 10422 }, { "epoch": 0.26, "learning_rate": 1.7419220784696712e-05, "loss": 2.2372, "step": 10423 }, { "epoch": 0.26, "learning_rate": 1.7418685491119002e-05, "loss": 2.2138, "step": 10424 }, { "epoch": 0.26, "learning_rate": 1.7418150150259595e-05, "loss": 2.3396, "step": 10425 }, { "epoch": 0.26, "learning_rate": 1.74176147621219e-05, "loss": 2.248, "step": 10426 }, { "epoch": 0.26, "learning_rate": 1.741707932670933e-05, "loss": 1.9776, "step": 10427 }, { "epoch": 0.26, "learning_rate": 1.7416543844025292e-05, "loss": 2.3802, "step": 10428 }, { "epoch": 0.26, "learning_rate": 1.7416008314073206e-05, "loss": 2.2773, "step": 10429 }, { "epoch": 0.26, "learning_rate": 1.741547273685648e-05, "loss": 2.2018, "step": 10430 }, { "epoch": 0.26, "learning_rate": 1.7414937112378534e-05, "loss": 2.0991, "step": 10431 }, { "epoch": 0.26, "learning_rate": 1.7414401440642775e-05, "loss": 2.2245, "step": 10432 }, { "epoch": 0.26, "learning_rate": 1.741386572165262e-05, "loss": 2.1813, "step": 10433 }, { "epoch": 0.26, "learning_rate": 1.741332995541148e-05, "loss": 2.0865, "step": 10434 }, { "epoch": 0.26, "learning_rate": 1.7412794141922777e-05, "loss": 2.306, "step": 10435 }, { "epoch": 0.26, "learning_rate": 1.7412258281189918e-05, "loss": 2.2878, "step": 10436 }, { "epoch": 0.26, "learning_rate": 1.741172237321632e-05, "loss": 2.2506, "step": 10437 }, { "epoch": 0.26, "learning_rate": 1.7411186418005405e-05, "loss": 2.2512, "step": 10438 }, { "epoch": 0.26, "learning_rate": 1.7410650415560577e-05, "loss": 2.2352, "step": 10439 }, { "epoch": 0.26, "learning_rate": 1.7410114365885263e-05, "loss": 1.9536, "step": 10440 }, { "epoch": 0.26, "learning_rate": 1.7409578268982875e-05, "loss": 2.1453, "step": 10441 }, { "epoch": 0.26, "learning_rate": 1.740904212485683e-05, "loss": 2.2419, "step": 10442 }, { "epoch": 0.26, "learning_rate": 1.7408505933510544e-05, "loss": 2.1819, "step": 10443 }, { "epoch": 0.26, "learning_rate": 1.7407969694947435e-05, "loss": 2.2246, "step": 10444 }, { "epoch": 0.26, "learning_rate": 1.7407433409170923e-05, "loss": 2.1678, "step": 10445 }, { "epoch": 0.26, "learning_rate": 1.7406897076184423e-05, "loss": 2.1022, "step": 10446 }, { "epoch": 0.26, "learning_rate": 1.7406360695991354e-05, "loss": 2.0735, "step": 10447 }, { "epoch": 0.26, "learning_rate": 1.7405824268595134e-05, "loss": 2.0579, "step": 10448 }, { "epoch": 0.26, "learning_rate": 1.740528779399918e-05, "loss": 2.0901, "step": 10449 }, { "epoch": 0.26, "learning_rate": 1.740475127220692e-05, "loss": 2.1265, "step": 10450 }, { "epoch": 0.26, "learning_rate": 1.740421470322176e-05, "loss": 2.2597, "step": 10451 }, { "epoch": 0.26, "learning_rate": 1.740367808704713e-05, "loss": 2.2838, "step": 10452 }, { "epoch": 0.26, "learning_rate": 1.7403141423686448e-05, "loss": 2.3105, "step": 10453 }, { "epoch": 0.26, "learning_rate": 1.7402604713143133e-05, "loss": 2.2285, "step": 10454 }, { "epoch": 0.26, "learning_rate": 1.7402067955420605e-05, "loss": 2.1713, "step": 10455 }, { "epoch": 0.26, "learning_rate": 1.7401531150522285e-05, "loss": 2.3106, "step": 10456 }, { "epoch": 0.26, "learning_rate": 1.7400994298451593e-05, "loss": 2.2132, "step": 10457 }, { "epoch": 0.26, "learning_rate": 1.7400457399211954e-05, "loss": 2.129, "step": 10458 }, { "epoch": 0.26, "learning_rate": 1.739992045280679e-05, "loss": 2.231, "step": 10459 }, { "epoch": 0.26, "learning_rate": 1.739938345923952e-05, "loss": 2.137, "step": 10460 }, { "epoch": 0.26, "learning_rate": 1.7398846418513568e-05, "loss": 2.1789, "step": 10461 }, { "epoch": 0.26, "learning_rate": 1.7398309330632356e-05, "loss": 2.2204, "step": 10462 }, { "epoch": 0.26, "learning_rate": 1.739777219559931e-05, "loss": 2.3373, "step": 10463 }, { "epoch": 0.26, "learning_rate": 1.7397235013417848e-05, "loss": 2.2146, "step": 10464 }, { "epoch": 0.26, "learning_rate": 1.7396697784091397e-05, "loss": 2.2176, "step": 10465 }, { "epoch": 0.26, "learning_rate": 1.7396160507623384e-05, "loss": 2.1911, "step": 10466 }, { "epoch": 0.26, "learning_rate": 1.7395623184017226e-05, "loss": 2.1378, "step": 10467 }, { "epoch": 0.26, "learning_rate": 1.739508581327635e-05, "loss": 2.0293, "step": 10468 }, { "epoch": 0.26, "learning_rate": 1.7394548395404184e-05, "loss": 2.1303, "step": 10469 }, { "epoch": 0.26, "learning_rate": 1.739401093040415e-05, "loss": 2.3067, "step": 10470 }, { "epoch": 0.26, "learning_rate": 1.7393473418279677e-05, "loss": 2.1212, "step": 10471 }, { "epoch": 0.26, "learning_rate": 1.739293585903419e-05, "loss": 2.205, "step": 10472 }, { "epoch": 0.26, "learning_rate": 1.739239825267111e-05, "loss": 2.382, "step": 10473 }, { "epoch": 0.26, "learning_rate": 1.739186059919387e-05, "loss": 2.1298, "step": 10474 }, { "epoch": 0.26, "learning_rate": 1.739132289860589e-05, "loss": 2.1383, "step": 10475 }, { "epoch": 0.26, "learning_rate": 1.7390785150910604e-05, "loss": 2.1028, "step": 10476 }, { "epoch": 0.26, "learning_rate": 1.7390247356111435e-05, "loss": 2.2114, "step": 10477 }, { "epoch": 0.26, "learning_rate": 1.7389709514211807e-05, "loss": 2.2047, "step": 10478 }, { "epoch": 0.26, "learning_rate": 1.7389171625215154e-05, "loss": 2.1225, "step": 10479 }, { "epoch": 0.26, "learning_rate": 1.7388633689124903e-05, "loss": 2.3289, "step": 10480 }, { "epoch": 0.26, "learning_rate": 1.7388095705944482e-05, "loss": 2.1981, "step": 10481 }, { "epoch": 0.26, "learning_rate": 1.738755767567732e-05, "loss": 2.1924, "step": 10482 }, { "epoch": 0.26, "learning_rate": 1.7387019598326845e-05, "loss": 2.1576, "step": 10483 }, { "epoch": 0.26, "learning_rate": 1.7386481473896487e-05, "loss": 2.1901, "step": 10484 }, { "epoch": 0.26, "learning_rate": 1.7385943302389675e-05, "loss": 2.2064, "step": 10485 }, { "epoch": 0.26, "learning_rate": 1.7385405083809836e-05, "loss": 2.3363, "step": 10486 }, { "epoch": 0.26, "learning_rate": 1.7384866818160407e-05, "loss": 2.0166, "step": 10487 }, { "epoch": 0.26, "learning_rate": 1.7384328505444816e-05, "loss": 2.0806, "step": 10488 }, { "epoch": 0.26, "learning_rate": 1.738379014566649e-05, "loss": 2.227, "step": 10489 }, { "epoch": 0.26, "learning_rate": 1.7383251738828865e-05, "loss": 2.3189, "step": 10490 }, { "epoch": 0.26, "learning_rate": 1.7382713284935372e-05, "loss": 2.3835, "step": 10491 }, { "epoch": 0.26, "learning_rate": 1.7382174783989437e-05, "loss": 2.2901, "step": 10492 }, { "epoch": 0.26, "learning_rate": 1.73816362359945e-05, "loss": 2.0974, "step": 10493 }, { "epoch": 0.26, "learning_rate": 1.738109764095399e-05, "loss": 2.0943, "step": 10494 }, { "epoch": 0.26, "learning_rate": 1.7380558998871334e-05, "loss": 2.1737, "step": 10495 }, { "epoch": 0.26, "learning_rate": 1.7380020309749975e-05, "loss": 2.1311, "step": 10496 }, { "epoch": 0.26, "learning_rate": 1.737948157359334e-05, "loss": 2.1521, "step": 10497 }, { "epoch": 0.26, "learning_rate": 1.7378942790404863e-05, "loss": 2.2624, "step": 10498 }, { "epoch": 0.26, "learning_rate": 1.737840396018798e-05, "loss": 2.1666, "step": 10499 }, { "epoch": 0.26, "learning_rate": 1.7377865082946128e-05, "loss": 2.1741, "step": 10500 }, { "epoch": 0.26, "learning_rate": 1.7377326158682734e-05, "loss": 2.2633, "step": 10501 }, { "epoch": 0.26, "learning_rate": 1.7376787187401232e-05, "loss": 2.1994, "step": 10502 }, { "epoch": 0.26, "learning_rate": 1.737624816910507e-05, "loss": 2.1497, "step": 10503 }, { "epoch": 0.26, "learning_rate": 1.7375709103797665e-05, "loss": 2.115, "step": 10504 }, { "epoch": 0.26, "learning_rate": 1.7375169991482468e-05, "loss": 2.0927, "step": 10505 }, { "epoch": 0.26, "learning_rate": 1.7374630832162908e-05, "loss": 2.0702, "step": 10506 }, { "epoch": 0.26, "learning_rate": 1.737409162584242e-05, "loss": 2.3705, "step": 10507 }, { "epoch": 0.26, "learning_rate": 1.7373552372524448e-05, "loss": 2.0627, "step": 10508 }, { "epoch": 0.26, "learning_rate": 1.7373013072212422e-05, "loss": 2.0143, "step": 10509 }, { "epoch": 0.26, "learning_rate": 1.7372473724909776e-05, "loss": 2.0397, "step": 10510 }, { "epoch": 0.26, "learning_rate": 1.737193433061996e-05, "loss": 2.1828, "step": 10511 }, { "epoch": 0.26, "learning_rate": 1.73713948893464e-05, "loss": 2.1273, "step": 10512 }, { "epoch": 0.26, "learning_rate": 1.737085540109254e-05, "loss": 2.2878, "step": 10513 }, { "epoch": 0.26, "learning_rate": 1.7370315865861812e-05, "loss": 2.3371, "step": 10514 }, { "epoch": 0.26, "learning_rate": 1.736977628365766e-05, "loss": 2.1725, "step": 10515 }, { "epoch": 0.26, "learning_rate": 1.7369236654483525e-05, "loss": 2.3462, "step": 10516 }, { "epoch": 0.26, "learning_rate": 1.7368696978342845e-05, "loss": 2.0419, "step": 10517 }, { "epoch": 0.26, "learning_rate": 1.7368157255239057e-05, "loss": 2.3492, "step": 10518 }, { "epoch": 0.26, "learning_rate": 1.7367617485175596e-05, "loss": 2.1848, "step": 10519 }, { "epoch": 0.26, "learning_rate": 1.736707766815591e-05, "loss": 2.1952, "step": 10520 }, { "epoch": 0.26, "learning_rate": 1.736653780418344e-05, "loss": 2.1865, "step": 10521 }, { "epoch": 0.26, "learning_rate": 1.7365997893261625e-05, "loss": 2.2185, "step": 10522 }, { "epoch": 0.26, "learning_rate": 1.73654579353939e-05, "loss": 2.2373, "step": 10523 }, { "epoch": 0.26, "learning_rate": 1.7364917930583715e-05, "loss": 2.0276, "step": 10524 }, { "epoch": 0.26, "learning_rate": 1.7364377878834508e-05, "loss": 2.181, "step": 10525 }, { "epoch": 0.26, "learning_rate": 1.7363837780149717e-05, "loss": 2.0703, "step": 10526 }, { "epoch": 0.26, "learning_rate": 1.736329763453279e-05, "loss": 2.3365, "step": 10527 }, { "epoch": 0.26, "learning_rate": 1.7362757441987167e-05, "loss": 2.1079, "step": 10528 }, { "epoch": 0.26, "learning_rate": 1.7362217202516295e-05, "loss": 2.2072, "step": 10529 }, { "epoch": 0.26, "learning_rate": 1.736167691612361e-05, "loss": 2.2287, "step": 10530 }, { "epoch": 0.26, "learning_rate": 1.736113658281256e-05, "loss": 2.2253, "step": 10531 }, { "epoch": 0.26, "learning_rate": 1.7360596202586585e-05, "loss": 2.2236, "step": 10532 }, { "epoch": 0.26, "learning_rate": 1.7360055775449133e-05, "loss": 2.1233, "step": 10533 }, { "epoch": 0.26, "learning_rate": 1.735951530140365e-05, "loss": 2.1928, "step": 10534 }, { "epoch": 0.26, "learning_rate": 1.7358974780453576e-05, "loss": 2.1455, "step": 10535 }, { "epoch": 0.26, "learning_rate": 1.7358434212602355e-05, "loss": 2.0551, "step": 10536 }, { "epoch": 0.26, "learning_rate": 1.7357893597853438e-05, "loss": 2.0689, "step": 10537 }, { "epoch": 0.26, "learning_rate": 1.7357352936210266e-05, "loss": 2.0345, "step": 10538 }, { "epoch": 0.26, "learning_rate": 1.7356812227676285e-05, "loss": 2.2021, "step": 10539 }, { "epoch": 0.26, "learning_rate": 1.735627147225494e-05, "loss": 2.1408, "step": 10540 }, { "epoch": 0.26, "learning_rate": 1.7355730669949684e-05, "loss": 2.0479, "step": 10541 }, { "epoch": 0.26, "learning_rate": 1.735518982076396e-05, "loss": 2.1616, "step": 10542 }, { "epoch": 0.26, "learning_rate": 1.7354648924701213e-05, "loss": 2.3326, "step": 10543 }, { "epoch": 0.26, "learning_rate": 1.735410798176489e-05, "loss": 2.1096, "step": 10544 }, { "epoch": 0.26, "learning_rate": 1.7353566991958438e-05, "loss": 2.3119, "step": 10545 }, { "epoch": 0.26, "learning_rate": 1.7353025955285313e-05, "loss": 2.0483, "step": 10546 }, { "epoch": 0.26, "learning_rate": 1.7352484871748955e-05, "loss": 2.173, "step": 10547 }, { "epoch": 0.26, "learning_rate": 1.7351943741352814e-05, "loss": 2.2459, "step": 10548 }, { "epoch": 0.26, "learning_rate": 1.7351402564100338e-05, "loss": 2.0322, "step": 10549 }, { "epoch": 0.26, "learning_rate": 1.735086133999498e-05, "loss": 2.0029, "step": 10550 }, { "epoch": 0.26, "learning_rate": 1.735032006904019e-05, "loss": 2.3336, "step": 10551 }, { "epoch": 0.26, "learning_rate": 1.7349778751239407e-05, "loss": 2.1729, "step": 10552 }, { "epoch": 0.26, "learning_rate": 1.7349237386596094e-05, "loss": 2.0691, "step": 10553 }, { "epoch": 0.26, "learning_rate": 1.7348695975113695e-05, "loss": 2.2275, "step": 10554 }, { "epoch": 0.26, "learning_rate": 1.7348154516795664e-05, "loss": 2.2599, "step": 10555 }, { "epoch": 0.26, "learning_rate": 1.7347613011645448e-05, "loss": 2.1553, "step": 10556 }, { "epoch": 0.26, "learning_rate": 1.7347071459666497e-05, "loss": 2.2085, "step": 10557 }, { "epoch": 0.26, "learning_rate": 1.7346529860862268e-05, "loss": 2.2033, "step": 10558 }, { "epoch": 0.26, "learning_rate": 1.734598821523621e-05, "loss": 2.1344, "step": 10559 }, { "epoch": 0.26, "learning_rate": 1.7345446522791775e-05, "loss": 2.1851, "step": 10560 }, { "epoch": 0.26, "learning_rate": 1.7344904783532417e-05, "loss": 2.1131, "step": 10561 }, { "epoch": 0.26, "learning_rate": 1.7344362997461583e-05, "loss": 2.0129, "step": 10562 }, { "epoch": 0.26, "learning_rate": 1.7343821164582732e-05, "loss": 2.2208, "step": 10563 }, { "epoch": 0.26, "learning_rate": 1.7343279284899318e-05, "loss": 2.2407, "step": 10564 }, { "epoch": 0.26, "learning_rate": 1.734273735841479e-05, "loss": 2.2757, "step": 10565 }, { "epoch": 0.26, "learning_rate": 1.7342195385132604e-05, "loss": 2.1004, "step": 10566 }, { "epoch": 0.26, "learning_rate": 1.7341653365056213e-05, "loss": 2.4397, "step": 10567 }, { "epoch": 0.26, "learning_rate": 1.734111129818907e-05, "loss": 2.1481, "step": 10568 }, { "epoch": 0.26, "learning_rate": 1.734056918453464e-05, "loss": 2.3071, "step": 10569 }, { "epoch": 0.26, "learning_rate": 1.7340027024096367e-05, "loss": 2.1383, "step": 10570 }, { "epoch": 0.26, "learning_rate": 1.7339484816877705e-05, "loss": 2.1389, "step": 10571 }, { "epoch": 0.26, "learning_rate": 1.733894256288212e-05, "loss": 2.3493, "step": 10572 }, { "epoch": 0.26, "learning_rate": 1.733840026211306e-05, "loss": 2.1013, "step": 10573 }, { "epoch": 0.26, "learning_rate": 1.7337857914573987e-05, "loss": 2.2808, "step": 10574 }, { "epoch": 0.26, "learning_rate": 1.7337315520268346e-05, "loss": 2.0922, "step": 10575 }, { "epoch": 0.26, "learning_rate": 1.733677307919961e-05, "loss": 2.0615, "step": 10576 }, { "epoch": 0.26, "learning_rate": 1.7336230591371225e-05, "loss": 2.0624, "step": 10577 }, { "epoch": 0.26, "learning_rate": 1.7335688056786652e-05, "loss": 2.0754, "step": 10578 }, { "epoch": 0.26, "learning_rate": 1.733514547544935e-05, "loss": 2.1602, "step": 10579 }, { "epoch": 0.26, "learning_rate": 1.7334602847362772e-05, "loss": 2.2648, "step": 10580 }, { "epoch": 0.26, "learning_rate": 1.7334060172530384e-05, "loss": 2.1197, "step": 10581 }, { "epoch": 0.26, "learning_rate": 1.7333517450955638e-05, "loss": 2.3448, "step": 10582 }, { "epoch": 0.26, "learning_rate": 1.7332974682641996e-05, "loss": 2.197, "step": 10583 }, { "epoch": 0.26, "learning_rate": 1.7332431867592915e-05, "loss": 2.1873, "step": 10584 }, { "epoch": 0.26, "learning_rate": 1.7331889005811858e-05, "loss": 2.2631, "step": 10585 }, { "epoch": 0.26, "learning_rate": 1.7331346097302283e-05, "loss": 2.0737, "step": 10586 }, { "epoch": 0.26, "learning_rate": 1.733080314206765e-05, "loss": 2.2575, "step": 10587 }, { "epoch": 0.26, "learning_rate": 1.733026014011142e-05, "loss": 1.983, "step": 10588 }, { "epoch": 0.26, "learning_rate": 1.732971709143705e-05, "loss": 2.2002, "step": 10589 }, { "epoch": 0.26, "learning_rate": 1.7329173996048004e-05, "loss": 2.0842, "step": 10590 }, { "epoch": 0.26, "learning_rate": 1.732863085394775e-05, "loss": 1.9183, "step": 10591 }, { "epoch": 0.26, "learning_rate": 1.7328087665139736e-05, "loss": 2.1811, "step": 10592 }, { "epoch": 0.26, "learning_rate": 1.7327544429627437e-05, "loss": 2.0424, "step": 10593 }, { "epoch": 0.26, "learning_rate": 1.7327001147414305e-05, "loss": 2.2066, "step": 10594 }, { "epoch": 0.26, "learning_rate": 1.732645781850381e-05, "loss": 2.2132, "step": 10595 }, { "epoch": 0.26, "learning_rate": 1.732591444289941e-05, "loss": 2.2318, "step": 10596 }, { "epoch": 0.26, "learning_rate": 1.732537102060457e-05, "loss": 2.2746, "step": 10597 }, { "epoch": 0.26, "learning_rate": 1.7324827551622754e-05, "loss": 2.0965, "step": 10598 }, { "epoch": 0.26, "learning_rate": 1.7324284035957425e-05, "loss": 2.2712, "step": 10599 }, { "epoch": 0.26, "learning_rate": 1.7323740473612046e-05, "loss": 2.325, "step": 10600 }, { "epoch": 0.26, "learning_rate": 1.7323196864590083e-05, "loss": 2.0815, "step": 10601 }, { "epoch": 0.26, "learning_rate": 1.7322653208895e-05, "loss": 2.308, "step": 10602 }, { "epoch": 0.26, "learning_rate": 1.732210950653026e-05, "loss": 2.0978, "step": 10603 }, { "epoch": 0.26, "learning_rate": 1.732156575749933e-05, "loss": 2.3006, "step": 10604 }, { "epoch": 0.26, "learning_rate": 1.7321021961805678e-05, "loss": 2.1747, "step": 10605 }, { "epoch": 0.26, "learning_rate": 1.7320478119452765e-05, "loss": 2.4571, "step": 10606 }, { "epoch": 0.26, "learning_rate": 1.731993423044406e-05, "loss": 2.1846, "step": 10607 }, { "epoch": 0.26, "learning_rate": 1.731939029478303e-05, "loss": 2.2476, "step": 10608 }, { "epoch": 0.26, "learning_rate": 1.7318846312473138e-05, "loss": 2.2776, "step": 10609 }, { "epoch": 0.26, "learning_rate": 1.731830228351785e-05, "loss": 2.1753, "step": 10610 }, { "epoch": 0.26, "learning_rate": 1.7317758207920643e-05, "loss": 2.0231, "step": 10611 }, { "epoch": 0.26, "learning_rate": 1.7317214085684978e-05, "loss": 2.2598, "step": 10612 }, { "epoch": 0.26, "learning_rate": 1.7316669916814318e-05, "loss": 2.1516, "step": 10613 }, { "epoch": 0.26, "learning_rate": 1.731612570131214e-05, "loss": 2.3443, "step": 10614 }, { "epoch": 0.26, "learning_rate": 1.7315581439181906e-05, "loss": 2.0876, "step": 10615 }, { "epoch": 0.26, "learning_rate": 1.731503713042709e-05, "loss": 2.1419, "step": 10616 }, { "epoch": 0.26, "learning_rate": 1.731449277505116e-05, "loss": 2.2552, "step": 10617 }, { "epoch": 0.26, "learning_rate": 1.731394837305758e-05, "loss": 1.9943, "step": 10618 }, { "epoch": 0.26, "learning_rate": 1.7313403924449825e-05, "loss": 1.9906, "step": 10619 }, { "epoch": 0.26, "learning_rate": 1.7312859429231364e-05, "loss": 2.2786, "step": 10620 }, { "epoch": 0.26, "learning_rate": 1.7312314887405667e-05, "loss": 2.1022, "step": 10621 }, { "epoch": 0.26, "learning_rate": 1.7311770298976203e-05, "loss": 2.0678, "step": 10622 }, { "epoch": 0.26, "learning_rate": 1.7311225663946443e-05, "loss": 2.0375, "step": 10623 }, { "epoch": 0.26, "learning_rate": 1.731068098231986e-05, "loss": 2.1533, "step": 10624 }, { "epoch": 0.26, "learning_rate": 1.7310136254099924e-05, "loss": 2.0951, "step": 10625 }, { "epoch": 0.26, "learning_rate": 1.730959147929011e-05, "loss": 2.2602, "step": 10626 }, { "epoch": 0.26, "learning_rate": 1.7309046657893884e-05, "loss": 2.0775, "step": 10627 }, { "epoch": 0.26, "learning_rate": 1.7308501789914725e-05, "loss": 2.0627, "step": 10628 }, { "epoch": 0.26, "learning_rate": 1.73079568753561e-05, "loss": 2.1767, "step": 10629 }, { "epoch": 0.26, "learning_rate": 1.7307411914221484e-05, "loss": 2.1397, "step": 10630 }, { "epoch": 0.26, "learning_rate": 1.730686690651435e-05, "loss": 2.2968, "step": 10631 }, { "epoch": 0.26, "learning_rate": 1.7306321852238176e-05, "loss": 2.31, "step": 10632 }, { "epoch": 0.26, "learning_rate": 1.730577675139643e-05, "loss": 2.1269, "step": 10633 }, { "epoch": 0.26, "learning_rate": 1.7305231603992587e-05, "loss": 2.0856, "step": 10634 }, { "epoch": 0.26, "learning_rate": 1.7304686410030122e-05, "loss": 2.264, "step": 10635 }, { "epoch": 0.26, "learning_rate": 1.730414116951251e-05, "loss": 2.226, "step": 10636 }, { "epoch": 0.26, "learning_rate": 1.730359588244323e-05, "loss": 2.2269, "step": 10637 }, { "epoch": 0.26, "learning_rate": 1.7303050548825745e-05, "loss": 2.132, "step": 10638 }, { "epoch": 0.26, "learning_rate": 1.7302505168663544e-05, "loss": 2.2674, "step": 10639 }, { "epoch": 0.26, "learning_rate": 1.73019597419601e-05, "loss": 2.2943, "step": 10640 }, { "epoch": 0.26, "learning_rate": 1.730141426871888e-05, "loss": 2.1443, "step": 10641 }, { "epoch": 0.26, "learning_rate": 1.7300868748943372e-05, "loss": 2.2646, "step": 10642 }, { "epoch": 0.26, "learning_rate": 1.7300323182637048e-05, "loss": 2.1499, "step": 10643 }, { "epoch": 0.26, "learning_rate": 1.7299777569803384e-05, "loss": 2.0947, "step": 10644 }, { "epoch": 0.26, "learning_rate": 1.7299231910445862e-05, "loss": 2.1929, "step": 10645 }, { "epoch": 0.26, "learning_rate": 1.7298686204567953e-05, "loss": 2.1379, "step": 10646 }, { "epoch": 0.26, "learning_rate": 1.729814045217314e-05, "loss": 2.215, "step": 10647 }, { "epoch": 0.26, "learning_rate": 1.72975946532649e-05, "loss": 2.2165, "step": 10648 }, { "epoch": 0.26, "learning_rate": 1.7297048807846708e-05, "loss": 2.2022, "step": 10649 }, { "epoch": 0.26, "learning_rate": 1.7296502915922047e-05, "loss": 2.1526, "step": 10650 }, { "epoch": 0.26, "learning_rate": 1.7295956977494396e-05, "loss": 2.1898, "step": 10651 }, { "epoch": 0.26, "learning_rate": 1.7295410992567232e-05, "loss": 2.4894, "step": 10652 }, { "epoch": 0.26, "learning_rate": 1.729486496114404e-05, "loss": 2.1493, "step": 10653 }, { "epoch": 0.26, "learning_rate": 1.7294318883228292e-05, "loss": 2.1296, "step": 10654 }, { "epoch": 0.26, "learning_rate": 1.7293772758823472e-05, "loss": 2.086, "step": 10655 }, { "epoch": 0.26, "learning_rate": 1.7293226587933067e-05, "loss": 2.1241, "step": 10656 }, { "epoch": 0.26, "learning_rate": 1.7292680370560546e-05, "loss": 2.1655, "step": 10657 }, { "epoch": 0.26, "learning_rate": 1.72921341067094e-05, "loss": 2.0557, "step": 10658 }, { "epoch": 0.26, "learning_rate": 1.7291587796383104e-05, "loss": 2.1825, "step": 10659 }, { "epoch": 0.26, "learning_rate": 1.7291041439585146e-05, "loss": 2.2771, "step": 10660 }, { "epoch": 0.26, "learning_rate": 1.7290495036319004e-05, "loss": 2.2195, "step": 10661 }, { "epoch": 0.26, "learning_rate": 1.728994858658816e-05, "loss": 2.1297, "step": 10662 }, { "epoch": 0.26, "learning_rate": 1.7289402090396098e-05, "loss": 2.1494, "step": 10663 }, { "epoch": 0.26, "learning_rate": 1.7288855547746303e-05, "loss": 2.2995, "step": 10664 }, { "epoch": 0.26, "learning_rate": 1.7288308958642254e-05, "loss": 2.3003, "step": 10665 }, { "epoch": 0.26, "learning_rate": 1.728776232308744e-05, "loss": 2.1598, "step": 10666 }, { "epoch": 0.26, "learning_rate": 1.7287215641085337e-05, "loss": 2.2375, "step": 10667 }, { "epoch": 0.26, "learning_rate": 1.7286668912639438e-05, "loss": 2.1369, "step": 10668 }, { "epoch": 0.26, "learning_rate": 1.728612213775322e-05, "loss": 2.2051, "step": 10669 }, { "epoch": 0.26, "learning_rate": 1.728557531643017e-05, "loss": 2.2485, "step": 10670 }, { "epoch": 0.26, "learning_rate": 1.728502844867378e-05, "loss": 2.0195, "step": 10671 }, { "epoch": 0.26, "learning_rate": 1.728448153448753e-05, "loss": 2.2257, "step": 10672 }, { "epoch": 0.26, "learning_rate": 1.72839345738749e-05, "loss": 1.9748, "step": 10673 }, { "epoch": 0.26, "learning_rate": 1.728338756683938e-05, "loss": 2.1409, "step": 10674 }, { "epoch": 0.26, "learning_rate": 1.728284051338446e-05, "loss": 2.2565, "step": 10675 }, { "epoch": 0.26, "learning_rate": 1.7282293413513627e-05, "loss": 2.1981, "step": 10676 }, { "epoch": 0.26, "learning_rate": 1.728174626723036e-05, "loss": 1.9803, "step": 10677 }, { "epoch": 0.26, "learning_rate": 1.7281199074538154e-05, "loss": 2.1262, "step": 10678 }, { "epoch": 0.26, "learning_rate": 1.7280651835440493e-05, "loss": 2.3637, "step": 10679 }, { "epoch": 0.26, "learning_rate": 1.7280104549940862e-05, "loss": 2.1203, "step": 10680 }, { "epoch": 0.26, "learning_rate": 1.7279557218042756e-05, "loss": 2.2725, "step": 10681 }, { "epoch": 0.26, "learning_rate": 1.727900983974966e-05, "loss": 2.3206, "step": 10682 }, { "epoch": 0.26, "learning_rate": 1.727846241506506e-05, "loss": 2.2729, "step": 10683 }, { "epoch": 0.26, "learning_rate": 1.7277914943992447e-05, "loss": 2.1946, "step": 10684 }, { "epoch": 0.26, "learning_rate": 1.727736742653531e-05, "loss": 2.1806, "step": 10685 }, { "epoch": 0.26, "learning_rate": 1.727681986269714e-05, "loss": 2.2954, "step": 10686 }, { "epoch": 0.26, "learning_rate": 1.7276272252481425e-05, "loss": 2.0933, "step": 10687 }, { "epoch": 0.26, "learning_rate": 1.727572459589166e-05, "loss": 2.0581, "step": 10688 }, { "epoch": 0.26, "learning_rate": 1.7275176892931326e-05, "loss": 2.2277, "step": 10689 }, { "epoch": 0.26, "learning_rate": 1.7274629143603914e-05, "loss": 2.1493, "step": 10690 }, { "epoch": 0.26, "learning_rate": 1.727408134791293e-05, "loss": 2.1275, "step": 10691 }, { "epoch": 0.26, "learning_rate": 1.727353350586185e-05, "loss": 2.3037, "step": 10692 }, { "epoch": 0.26, "learning_rate": 1.7272985617454172e-05, "loss": 2.1309, "step": 10693 }, { "epoch": 0.26, "learning_rate": 1.7272437682693382e-05, "loss": 2.1628, "step": 10694 }, { "epoch": 0.26, "learning_rate": 1.7271889701582984e-05, "loss": 2.2515, "step": 10695 }, { "epoch": 0.26, "learning_rate": 1.7271341674126456e-05, "loss": 2.1424, "step": 10696 }, { "epoch": 0.26, "learning_rate": 1.72707936003273e-05, "loss": 2.1951, "step": 10697 }, { "epoch": 0.26, "learning_rate": 1.727024548018901e-05, "loss": 2.113, "step": 10698 }, { "epoch": 0.26, "learning_rate": 1.7269697313715074e-05, "loss": 2.1139, "step": 10699 }, { "epoch": 0.26, "learning_rate": 1.7269149100908988e-05, "loss": 2.082, "step": 10700 }, { "epoch": 0.26, "learning_rate": 1.7268600841774244e-05, "loss": 2.0668, "step": 10701 }, { "epoch": 0.26, "learning_rate": 1.7268052536314344e-05, "loss": 2.2636, "step": 10702 }, { "epoch": 0.26, "learning_rate": 1.726750418453277e-05, "loss": 2.1844, "step": 10703 }, { "epoch": 0.26, "learning_rate": 1.7266955786433025e-05, "loss": 2.14, "step": 10704 }, { "epoch": 0.26, "learning_rate": 1.7266407342018606e-05, "loss": 2.0678, "step": 10705 }, { "epoch": 0.26, "learning_rate": 1.7265858851293003e-05, "loss": 2.0141, "step": 10706 }, { "epoch": 0.26, "learning_rate": 1.7265310314259713e-05, "loss": 2.0899, "step": 10707 }, { "epoch": 0.26, "learning_rate": 1.726476173092223e-05, "loss": 2.3149, "step": 10708 }, { "epoch": 0.26, "learning_rate": 1.7264213101284058e-05, "loss": 1.8839, "step": 10709 }, { "epoch": 0.26, "learning_rate": 1.7263664425348684e-05, "loss": 2.2255, "step": 10710 }, { "epoch": 0.26, "learning_rate": 1.7263115703119612e-05, "loss": 2.1658, "step": 10711 }, { "epoch": 0.26, "learning_rate": 1.7262566934600338e-05, "loss": 2.1514, "step": 10712 }, { "epoch": 0.26, "learning_rate": 1.7262018119794356e-05, "loss": 2.1654, "step": 10713 }, { "epoch": 0.26, "learning_rate": 1.726146925870517e-05, "loss": 2.2434, "step": 10714 }, { "epoch": 0.26, "learning_rate": 1.7260920351336267e-05, "loss": 2.2986, "step": 10715 }, { "epoch": 0.26, "learning_rate": 1.726037139769116e-05, "loss": 2.2958, "step": 10716 }, { "epoch": 0.26, "learning_rate": 1.7259822397773334e-05, "loss": 2.1936, "step": 10717 }, { "epoch": 0.26, "learning_rate": 1.7259273351586296e-05, "loss": 2.2136, "step": 10718 }, { "epoch": 0.26, "learning_rate": 1.7258724259133542e-05, "loss": 2.2316, "step": 10719 }, { "epoch": 0.26, "learning_rate": 1.7258175120418573e-05, "loss": 2.0552, "step": 10720 }, { "epoch": 0.26, "learning_rate": 1.7257625935444888e-05, "loss": 2.0925, "step": 10721 }, { "epoch": 0.26, "learning_rate": 1.725707670421599e-05, "loss": 2.176, "step": 10722 }, { "epoch": 0.26, "learning_rate": 1.7256527426735375e-05, "loss": 2.1175, "step": 10723 }, { "epoch": 0.26, "learning_rate": 1.7255978103006547e-05, "loss": 2.2233, "step": 10724 }, { "epoch": 0.26, "learning_rate": 1.725542873303301e-05, "loss": 2.1691, "step": 10725 }, { "epoch": 0.26, "learning_rate": 1.7254879316818253e-05, "loss": 2.0127, "step": 10726 }, { "epoch": 0.26, "learning_rate": 1.725432985436579e-05, "loss": 2.1033, "step": 10727 }, { "epoch": 0.26, "learning_rate": 1.7253780345679122e-05, "loss": 2.2673, "step": 10728 }, { "epoch": 0.26, "learning_rate": 1.725323079076174e-05, "loss": 2.214, "step": 10729 }, { "epoch": 0.26, "learning_rate": 1.725268118961716e-05, "loss": 2.2305, "step": 10730 }, { "epoch": 0.26, "learning_rate": 1.725213154224888e-05, "loss": 2.1337, "step": 10731 }, { "epoch": 0.26, "learning_rate": 1.72515818486604e-05, "loss": 2.175, "step": 10732 }, { "epoch": 0.26, "learning_rate": 1.7251032108855225e-05, "loss": 2.1201, "step": 10733 }, { "epoch": 0.26, "learning_rate": 1.725048232283686e-05, "loss": 2.2653, "step": 10734 }, { "epoch": 0.26, "learning_rate": 1.7249932490608808e-05, "loss": 2.4026, "step": 10735 }, { "epoch": 0.26, "learning_rate": 1.7249382612174576e-05, "loss": 2.0138, "step": 10736 }, { "epoch": 0.26, "learning_rate": 1.7248832687537667e-05, "loss": 2.2726, "step": 10737 }, { "epoch": 0.26, "learning_rate": 1.724828271670158e-05, "loss": 1.9616, "step": 10738 }, { "epoch": 0.26, "learning_rate": 1.724773269966983e-05, "loss": 2.2196, "step": 10739 }, { "epoch": 0.26, "learning_rate": 1.7247182636445914e-05, "loss": 2.3043, "step": 10740 }, { "epoch": 0.26, "learning_rate": 1.724663252703334e-05, "loss": 2.1473, "step": 10741 }, { "epoch": 0.26, "learning_rate": 1.7246082371435618e-05, "loss": 2.1413, "step": 10742 }, { "epoch": 0.26, "learning_rate": 1.724553216965625e-05, "loss": 2.135, "step": 10743 }, { "epoch": 0.26, "learning_rate": 1.7244981921698745e-05, "loss": 2.1501, "step": 10744 }, { "epoch": 0.26, "learning_rate": 1.724443162756661e-05, "loss": 1.9669, "step": 10745 }, { "epoch": 0.26, "learning_rate": 1.7243881287263354e-05, "loss": 2.121, "step": 10746 }, { "epoch": 0.26, "learning_rate": 1.724333090079248e-05, "loss": 2.0615, "step": 10747 }, { "epoch": 0.26, "learning_rate": 1.7242780468157494e-05, "loss": 2.1826, "step": 10748 }, { "epoch": 0.26, "learning_rate": 1.724222998936191e-05, "loss": 2.1906, "step": 10749 }, { "epoch": 0.26, "learning_rate": 1.7241679464409236e-05, "loss": 2.3952, "step": 10750 }, { "epoch": 0.26, "learning_rate": 1.7241128893302976e-05, "loss": 2.0942, "step": 10751 }, { "epoch": 0.27, "learning_rate": 1.7240578276046644e-05, "loss": 2.2013, "step": 10752 }, { "epoch": 0.27, "learning_rate": 1.7240027612643743e-05, "loss": 2.1407, "step": 10753 }, { "epoch": 0.27, "learning_rate": 1.723947690309779e-05, "loss": 2.2081, "step": 10754 }, { "epoch": 0.27, "learning_rate": 1.7238926147412288e-05, "loss": 2.047, "step": 10755 }, { "epoch": 0.27, "learning_rate": 1.7238375345590753e-05, "loss": 2.095, "step": 10756 }, { "epoch": 0.27, "learning_rate": 1.7237824497636697e-05, "loss": 2.0859, "step": 10757 }, { "epoch": 0.27, "learning_rate": 1.7237273603553623e-05, "loss": 2.1389, "step": 10758 }, { "epoch": 0.27, "learning_rate": 1.7236722663345044e-05, "loss": 2.0135, "step": 10759 }, { "epoch": 0.27, "learning_rate": 1.7236171677014475e-05, "loss": 2.1603, "step": 10760 }, { "epoch": 0.27, "learning_rate": 1.7235620644565428e-05, "loss": 2.3319, "step": 10761 }, { "epoch": 0.27, "learning_rate": 1.723506956600141e-05, "loss": 2.026, "step": 10762 }, { "epoch": 0.27, "learning_rate": 1.7234518441325936e-05, "loss": 1.995, "step": 10763 }, { "epoch": 0.27, "learning_rate": 1.7233967270542518e-05, "loss": 2.0253, "step": 10764 }, { "epoch": 0.27, "learning_rate": 1.723341605365467e-05, "loss": 2.0182, "step": 10765 }, { "epoch": 0.27, "learning_rate": 1.72328647906659e-05, "loss": 2.2404, "step": 10766 }, { "epoch": 0.27, "learning_rate": 1.723231348157973e-05, "loss": 2.1677, "step": 10767 }, { "epoch": 0.27, "learning_rate": 1.723176212639967e-05, "loss": 2.1241, "step": 10768 }, { "epoch": 0.27, "learning_rate": 1.7231210725129232e-05, "loss": 2.1764, "step": 10769 }, { "epoch": 0.27, "learning_rate": 1.723065927777193e-05, "loss": 2.4301, "step": 10770 }, { "epoch": 0.27, "learning_rate": 1.723010778433128e-05, "loss": 2.1558, "step": 10771 }, { "epoch": 0.27, "learning_rate": 1.7229556244810802e-05, "loss": 2.1868, "step": 10772 }, { "epoch": 0.27, "learning_rate": 1.7229004659213998e-05, "loss": 2.2775, "step": 10773 }, { "epoch": 0.27, "learning_rate": 1.7228453027544398e-05, "loss": 2.2718, "step": 10774 }, { "epoch": 0.27, "learning_rate": 1.7227901349805506e-05, "loss": 2.3311, "step": 10775 }, { "epoch": 0.27, "learning_rate": 1.7227349626000846e-05, "loss": 2.0594, "step": 10776 }, { "epoch": 0.27, "learning_rate": 1.722679785613393e-05, "loss": 2.1588, "step": 10777 }, { "epoch": 0.27, "learning_rate": 1.7226246040208278e-05, "loss": 2.0759, "step": 10778 }, { "epoch": 0.27, "learning_rate": 1.7225694178227405e-05, "loss": 2.1071, "step": 10779 }, { "epoch": 0.27, "learning_rate": 1.7225142270194827e-05, "loss": 2.2726, "step": 10780 }, { "epoch": 0.27, "learning_rate": 1.722459031611407e-05, "loss": 2.2082, "step": 10781 }, { "epoch": 0.27, "learning_rate": 1.7224038315988633e-05, "loss": 2.0719, "step": 10782 }, { "epoch": 0.27, "learning_rate": 1.722348626982205e-05, "loss": 2.0211, "step": 10783 }, { "epoch": 0.27, "learning_rate": 1.7222934177617836e-05, "loss": 2.0644, "step": 10784 }, { "epoch": 0.27, "learning_rate": 1.7222382039379507e-05, "loss": 2.1583, "step": 10785 }, { "epoch": 0.27, "learning_rate": 1.7221829855110587e-05, "loss": 1.856, "step": 10786 }, { "epoch": 0.27, "learning_rate": 1.7221277624814586e-05, "loss": 2.2512, "step": 10787 }, { "epoch": 0.27, "learning_rate": 1.7220725348495034e-05, "loss": 2.1774, "step": 10788 }, { "epoch": 0.27, "learning_rate": 1.7220173026155443e-05, "loss": 2.0689, "step": 10789 }, { "epoch": 0.27, "learning_rate": 1.721962065779934e-05, "loss": 2.072, "step": 10790 }, { "epoch": 0.27, "learning_rate": 1.721906824343024e-05, "loss": 2.2075, "step": 10791 }, { "epoch": 0.27, "learning_rate": 1.7218515783051665e-05, "loss": 2.2063, "step": 10792 }, { "epoch": 0.27, "learning_rate": 1.7217963276667136e-05, "loss": 2.2833, "step": 10793 }, { "epoch": 0.27, "learning_rate": 1.7217410724280177e-05, "loss": 2.1195, "step": 10794 }, { "epoch": 0.27, "learning_rate": 1.7216858125894303e-05, "loss": 2.2043, "step": 10795 }, { "epoch": 0.27, "learning_rate": 1.7216305481513047e-05, "loss": 1.9921, "step": 10796 }, { "epoch": 0.27, "learning_rate": 1.721575279113992e-05, "loss": 2.114, "step": 10797 }, { "epoch": 0.27, "learning_rate": 1.7215200054778448e-05, "loss": 2.2252, "step": 10798 }, { "epoch": 0.27, "learning_rate": 1.7214647272432157e-05, "loss": 2.2108, "step": 10799 }, { "epoch": 0.27, "learning_rate": 1.7214094444104567e-05, "loss": 2.083, "step": 10800 }, { "epoch": 0.27, "learning_rate": 1.7213541569799202e-05, "loss": 2.1236, "step": 10801 }, { "epoch": 0.27, "learning_rate": 1.7212988649519585e-05, "loss": 2.0837, "step": 10802 }, { "epoch": 0.27, "learning_rate": 1.7212435683269244e-05, "loss": 2.0903, "step": 10803 }, { "epoch": 0.27, "learning_rate": 1.7211882671051696e-05, "loss": 2.3475, "step": 10804 }, { "epoch": 0.27, "learning_rate": 1.7211329612870474e-05, "loss": 2.2292, "step": 10805 }, { "epoch": 0.27, "learning_rate": 1.721077650872909e-05, "loss": 2.0454, "step": 10806 }, { "epoch": 0.27, "learning_rate": 1.7210223358631086e-05, "loss": 2.499, "step": 10807 }, { "epoch": 0.27, "learning_rate": 1.7209670162579975e-05, "loss": 2.2649, "step": 10808 }, { "epoch": 0.27, "learning_rate": 1.7209116920579285e-05, "loss": 2.133, "step": 10809 }, { "epoch": 0.27, "learning_rate": 1.7208563632632546e-05, "loss": 2.278, "step": 10810 }, { "epoch": 0.27, "learning_rate": 1.7208010298743282e-05, "loss": 2.2832, "step": 10811 }, { "epoch": 0.27, "learning_rate": 1.720745691891502e-05, "loss": 1.9769, "step": 10812 }, { "epoch": 0.27, "learning_rate": 1.7206903493151284e-05, "loss": 2.2083, "step": 10813 }, { "epoch": 0.27, "learning_rate": 1.7206350021455604e-05, "loss": 2.2231, "step": 10814 }, { "epoch": 0.27, "learning_rate": 1.7205796503831507e-05, "loss": 2.3316, "step": 10815 }, { "epoch": 0.27, "learning_rate": 1.7205242940282517e-05, "loss": 2.239, "step": 10816 }, { "epoch": 0.27, "learning_rate": 1.720468933081217e-05, "loss": 2.2347, "step": 10817 }, { "epoch": 0.27, "learning_rate": 1.7204135675423993e-05, "loss": 2.1295, "step": 10818 }, { "epoch": 0.27, "learning_rate": 1.7203581974121507e-05, "loss": 2.2063, "step": 10819 }, { "epoch": 0.27, "learning_rate": 1.7203028226908245e-05, "loss": 2.2045, "step": 10820 }, { "epoch": 0.27, "learning_rate": 1.720247443378774e-05, "loss": 2.1444, "step": 10821 }, { "epoch": 0.27, "learning_rate": 1.7201920594763513e-05, "loss": 2.135, "step": 10822 }, { "epoch": 0.27, "learning_rate": 1.7201366709839103e-05, "loss": 2.193, "step": 10823 }, { "epoch": 0.27, "learning_rate": 1.7200812779018034e-05, "loss": 2.1694, "step": 10824 }, { "epoch": 0.27, "learning_rate": 1.7200258802303838e-05, "loss": 2.3078, "step": 10825 }, { "epoch": 0.27, "learning_rate": 1.719970477970005e-05, "loss": 2.0954, "step": 10826 }, { "epoch": 0.27, "learning_rate": 1.7199150711210194e-05, "loss": 2.0443, "step": 10827 }, { "epoch": 0.27, "learning_rate": 1.7198596596837806e-05, "loss": 2.1318, "step": 10828 }, { "epoch": 0.27, "learning_rate": 1.7198042436586418e-05, "loss": 2.0959, "step": 10829 }, { "epoch": 0.27, "learning_rate": 1.7197488230459555e-05, "loss": 2.2884, "step": 10830 }, { "epoch": 0.27, "learning_rate": 1.7196933978460756e-05, "loss": 2.1349, "step": 10831 }, { "epoch": 0.27, "learning_rate": 1.719637968059355e-05, "loss": 2.1315, "step": 10832 }, { "epoch": 0.27, "learning_rate": 1.7195825336861473e-05, "loss": 2.2988, "step": 10833 }, { "epoch": 0.27, "learning_rate": 1.7195270947268054e-05, "loss": 2.2716, "step": 10834 }, { "epoch": 0.27, "learning_rate": 1.7194716511816833e-05, "loss": 2.1768, "step": 10835 }, { "epoch": 0.27, "learning_rate": 1.7194162030511334e-05, "loss": 2.195, "step": 10836 }, { "epoch": 0.27, "learning_rate": 1.71936075033551e-05, "loss": 2.2721, "step": 10837 }, { "epoch": 0.27, "learning_rate": 1.7193052930351657e-05, "loss": 2.0512, "step": 10838 }, { "epoch": 0.27, "learning_rate": 1.7192498311504545e-05, "loss": 2.1591, "step": 10839 }, { "epoch": 0.27, "learning_rate": 1.7191943646817298e-05, "loss": 2.0974, "step": 10840 }, { "epoch": 0.27, "learning_rate": 1.719138893629345e-05, "loss": 2.1547, "step": 10841 }, { "epoch": 0.27, "learning_rate": 1.7190834179936535e-05, "loss": 2.1064, "step": 10842 }, { "epoch": 0.27, "learning_rate": 1.7190279377750094e-05, "loss": 1.9981, "step": 10843 }, { "epoch": 0.27, "learning_rate": 1.7189724529737656e-05, "loss": 2.2781, "step": 10844 }, { "epoch": 0.27, "learning_rate": 1.718916963590276e-05, "loss": 2.1774, "step": 10845 }, { "epoch": 0.27, "learning_rate": 1.7188614696248948e-05, "loss": 2.0436, "step": 10846 }, { "epoch": 0.27, "learning_rate": 1.7188059710779746e-05, "loss": 2.0905, "step": 10847 }, { "epoch": 0.27, "learning_rate": 1.7187504679498702e-05, "loss": 2.1572, "step": 10848 }, { "epoch": 0.27, "learning_rate": 1.7186949602409345e-05, "loss": 2.1134, "step": 10849 }, { "epoch": 0.27, "learning_rate": 1.7186394479515217e-05, "loss": 2.2914, "step": 10850 }, { "epoch": 0.27, "learning_rate": 1.7185839310819855e-05, "loss": 2.0094, "step": 10851 }, { "epoch": 0.27, "learning_rate": 1.7185284096326795e-05, "loss": 2.2302, "step": 10852 }, { "epoch": 0.27, "learning_rate": 1.718472883603958e-05, "loss": 2.1466, "step": 10853 }, { "epoch": 0.27, "learning_rate": 1.718417352996175e-05, "loss": 2.1323, "step": 10854 }, { "epoch": 0.27, "learning_rate": 1.7183618178096834e-05, "loss": 2.1133, "step": 10855 }, { "epoch": 0.27, "learning_rate": 1.7183062780448383e-05, "loss": 2.1937, "step": 10856 }, { "epoch": 0.27, "learning_rate": 1.7182507337019928e-05, "loss": 2.2283, "step": 10857 }, { "epoch": 0.27, "learning_rate": 1.7181951847815018e-05, "loss": 2.0936, "step": 10858 }, { "epoch": 0.27, "learning_rate": 1.7181396312837186e-05, "loss": 2.0565, "step": 10859 }, { "epoch": 0.27, "learning_rate": 1.7180840732089976e-05, "loss": 2.1455, "step": 10860 }, { "epoch": 0.27, "learning_rate": 1.7180285105576925e-05, "loss": 2.2234, "step": 10861 }, { "epoch": 0.27, "learning_rate": 1.717972943330158e-05, "loss": 2.1884, "step": 10862 }, { "epoch": 0.27, "learning_rate": 1.717917371526748e-05, "loss": 2.1651, "step": 10863 }, { "epoch": 0.27, "learning_rate": 1.7178617951478163e-05, "loss": 1.9935, "step": 10864 }, { "epoch": 0.27, "learning_rate": 1.717806214193718e-05, "loss": 2.2738, "step": 10865 }, { "epoch": 0.27, "learning_rate": 1.717750628664806e-05, "loss": 1.9833, "step": 10866 }, { "epoch": 0.27, "learning_rate": 1.717695038561436e-05, "loss": 2.1329, "step": 10867 }, { "epoch": 0.27, "learning_rate": 1.7176394438839612e-05, "loss": 1.9194, "step": 10868 }, { "epoch": 0.27, "learning_rate": 1.7175838446327365e-05, "loss": 2.2508, "step": 10869 }, { "epoch": 0.27, "learning_rate": 1.717528240808116e-05, "loss": 2.1815, "step": 10870 }, { "epoch": 0.27, "learning_rate": 1.717472632410454e-05, "loss": 1.9865, "step": 10871 }, { "epoch": 0.27, "learning_rate": 1.7174170194401058e-05, "loss": 2.2397, "step": 10872 }, { "epoch": 0.27, "learning_rate": 1.717361401897425e-05, "loss": 2.2381, "step": 10873 }, { "epoch": 0.27, "learning_rate": 1.7173057797827657e-05, "loss": 2.1794, "step": 10874 }, { "epoch": 0.27, "learning_rate": 1.717250153096483e-05, "loss": 2.1943, "step": 10875 }, { "epoch": 0.27, "learning_rate": 1.7171945218389313e-05, "loss": 2.085, "step": 10876 }, { "epoch": 0.27, "learning_rate": 1.7171388860104655e-05, "loss": 2.2387, "step": 10877 }, { "epoch": 0.27, "learning_rate": 1.71708324561144e-05, "loss": 2.1557, "step": 10878 }, { "epoch": 0.27, "learning_rate": 1.717027600642209e-05, "loss": 2.2304, "step": 10879 }, { "epoch": 0.27, "learning_rate": 1.7169719511031272e-05, "loss": 2.215, "step": 10880 }, { "epoch": 0.27, "learning_rate": 1.71691629699455e-05, "loss": 2.2437, "step": 10881 }, { "epoch": 0.27, "learning_rate": 1.716860638316831e-05, "loss": 2.1142, "step": 10882 }, { "epoch": 0.27, "learning_rate": 1.716804975070326e-05, "loss": 2.1348, "step": 10883 }, { "epoch": 0.27, "learning_rate": 1.716749307255389e-05, "loss": 2.153, "step": 10884 }, { "epoch": 0.27, "learning_rate": 1.7166936348723754e-05, "loss": 2.08, "step": 10885 }, { "epoch": 0.27, "learning_rate": 1.7166379579216394e-05, "loss": 2.0744, "step": 10886 }, { "epoch": 0.27, "learning_rate": 1.7165822764035365e-05, "loss": 2.1386, "step": 10887 }, { "epoch": 0.27, "learning_rate": 1.716526590318421e-05, "loss": 2.2239, "step": 10888 }, { "epoch": 0.27, "learning_rate": 1.7164708996666478e-05, "loss": 2.1408, "step": 10889 }, { "epoch": 0.27, "learning_rate": 1.7164152044485726e-05, "loss": 2.4119, "step": 10890 }, { "epoch": 0.27, "learning_rate": 1.7163595046645494e-05, "loss": 2.3469, "step": 10891 }, { "epoch": 0.27, "learning_rate": 1.716303800314934e-05, "loss": 2.1631, "step": 10892 }, { "epoch": 0.27, "learning_rate": 1.7162480914000804e-05, "loss": 1.8635, "step": 10893 }, { "epoch": 0.27, "learning_rate": 1.716192377920345e-05, "loss": 2.2678, "step": 10894 }, { "epoch": 0.27, "learning_rate": 1.716136659876082e-05, "loss": 2.2582, "step": 10895 }, { "epoch": 0.27, "learning_rate": 1.716080937267646e-05, "loss": 2.1603, "step": 10896 }, { "epoch": 0.27, "learning_rate": 1.7160252100953938e-05, "loss": 2.1759, "step": 10897 }, { "epoch": 0.27, "learning_rate": 1.715969478359679e-05, "loss": 2.128, "step": 10898 }, { "epoch": 0.27, "learning_rate": 1.7159137420608577e-05, "loss": 2.2588, "step": 10899 }, { "epoch": 0.27, "learning_rate": 1.7158580011992845e-05, "loss": 2.0716, "step": 10900 }, { "epoch": 0.27, "learning_rate": 1.7158022557753155e-05, "loss": 2.1016, "step": 10901 }, { "epoch": 0.27, "learning_rate": 1.715746505789305e-05, "loss": 2.0961, "step": 10902 }, { "epoch": 0.27, "learning_rate": 1.715690751241609e-05, "loss": 2.0898, "step": 10903 }, { "epoch": 0.27, "learning_rate": 1.7156349921325822e-05, "loss": 2.1488, "step": 10904 }, { "epoch": 0.27, "learning_rate": 1.7155792284625805e-05, "loss": 2.1991, "step": 10905 }, { "epoch": 0.27, "learning_rate": 1.7155234602319594e-05, "loss": 2.1484, "step": 10906 }, { "epoch": 0.27, "learning_rate": 1.715467687441074e-05, "loss": 2.0155, "step": 10907 }, { "epoch": 0.27, "learning_rate": 1.71541191009028e-05, "loss": 2.1368, "step": 10908 }, { "epoch": 0.27, "learning_rate": 1.7153561281799325e-05, "loss": 2.1183, "step": 10909 }, { "epoch": 0.27, "learning_rate": 1.7153003417103873e-05, "loss": 2.1426, "step": 10910 }, { "epoch": 0.27, "learning_rate": 1.7152445506820004e-05, "loss": 2.2879, "step": 10911 }, { "epoch": 0.27, "learning_rate": 1.7151887550951265e-05, "loss": 2.3164, "step": 10912 }, { "epoch": 0.27, "learning_rate": 1.7151329549501216e-05, "loss": 2.287, "step": 10913 }, { "epoch": 0.27, "learning_rate": 1.715077150247341e-05, "loss": 2.2823, "step": 10914 }, { "epoch": 0.27, "learning_rate": 1.715021340987141e-05, "loss": 2.2422, "step": 10915 }, { "epoch": 0.27, "learning_rate": 1.714965527169877e-05, "loss": 2.1711, "step": 10916 }, { "epoch": 0.27, "learning_rate": 1.714909708795905e-05, "loss": 2.2054, "step": 10917 }, { "epoch": 0.27, "learning_rate": 1.71485388586558e-05, "loss": 2.1992, "step": 10918 }, { "epoch": 0.27, "learning_rate": 1.7147980583792584e-05, "loss": 2.2389, "step": 10919 }, { "epoch": 0.27, "learning_rate": 1.7147422263372953e-05, "loss": 2.176, "step": 10920 }, { "epoch": 0.27, "learning_rate": 1.714686389740048e-05, "loss": 2.0872, "step": 10921 }, { "epoch": 0.27, "learning_rate": 1.7146305485878708e-05, "loss": 1.9751, "step": 10922 }, { "epoch": 0.27, "learning_rate": 1.7145747028811205e-05, "loss": 2.2643, "step": 10923 }, { "epoch": 0.27, "learning_rate": 1.7145188526201523e-05, "loss": 1.9913, "step": 10924 }, { "epoch": 0.27, "learning_rate": 1.714462997805323e-05, "loss": 2.1329, "step": 10925 }, { "epoch": 0.27, "learning_rate": 1.7144071384369882e-05, "loss": 2.1896, "step": 10926 }, { "epoch": 0.27, "learning_rate": 1.714351274515504e-05, "loss": 2.2123, "step": 10927 }, { "epoch": 0.27, "learning_rate": 1.714295406041226e-05, "loss": 1.9826, "step": 10928 }, { "epoch": 0.27, "learning_rate": 1.7142395330145105e-05, "loss": 2.2203, "step": 10929 }, { "epoch": 0.27, "learning_rate": 1.7141836554357142e-05, "loss": 2.2284, "step": 10930 }, { "epoch": 0.27, "learning_rate": 1.7141277733051923e-05, "loss": 2.2007, "step": 10931 }, { "epoch": 0.27, "learning_rate": 1.714071886623301e-05, "loss": 1.933, "step": 10932 }, { "epoch": 0.27, "learning_rate": 1.7140159953903975e-05, "loss": 2.2272, "step": 10933 }, { "epoch": 0.27, "learning_rate": 1.7139600996068375e-05, "loss": 2.1899, "step": 10934 }, { "epoch": 0.27, "learning_rate": 1.713904199272977e-05, "loss": 2.1126, "step": 10935 }, { "epoch": 0.27, "learning_rate": 1.713848294389172e-05, "loss": 2.2947, "step": 10936 }, { "epoch": 0.27, "learning_rate": 1.7137923849557794e-05, "loss": 2.2487, "step": 10937 }, { "epoch": 0.27, "learning_rate": 1.713736470973155e-05, "loss": 2.0158, "step": 10938 }, { "epoch": 0.27, "learning_rate": 1.713680552441656e-05, "loss": 2.1923, "step": 10939 }, { "epoch": 0.27, "learning_rate": 1.713624629361638e-05, "loss": 1.9308, "step": 10940 }, { "epoch": 0.27, "learning_rate": 1.713568701733458e-05, "loss": 2.1802, "step": 10941 }, { "epoch": 0.27, "learning_rate": 1.7135127695574717e-05, "loss": 2.2291, "step": 10942 }, { "epoch": 0.27, "learning_rate": 1.713456832834036e-05, "loss": 2.2357, "step": 10943 }, { "epoch": 0.27, "learning_rate": 1.7134008915635075e-05, "loss": 2.0949, "step": 10944 }, { "epoch": 0.27, "learning_rate": 1.713344945746243e-05, "loss": 2.0314, "step": 10945 }, { "epoch": 0.27, "learning_rate": 1.7132889953825984e-05, "loss": 2.2849, "step": 10946 }, { "epoch": 0.27, "learning_rate": 1.7132330404729304e-05, "loss": 2.3543, "step": 10947 }, { "epoch": 0.27, "learning_rate": 1.7131770810175957e-05, "loss": 2.0993, "step": 10948 }, { "epoch": 0.27, "learning_rate": 1.713121117016951e-05, "loss": 2.1936, "step": 10949 }, { "epoch": 0.27, "learning_rate": 1.7130651484713538e-05, "loss": 2.2133, "step": 10950 }, { "epoch": 0.27, "learning_rate": 1.7130091753811594e-05, "loss": 2.0755, "step": 10951 }, { "epoch": 0.27, "learning_rate": 1.712953197746725e-05, "loss": 2.1128, "step": 10952 }, { "epoch": 0.27, "learning_rate": 1.712897215568408e-05, "loss": 2.0213, "step": 10953 }, { "epoch": 0.27, "learning_rate": 1.7128412288465645e-05, "loss": 2.1933, "step": 10954 }, { "epoch": 0.27, "learning_rate": 1.7127852375815516e-05, "loss": 2.3305, "step": 10955 }, { "epoch": 0.27, "learning_rate": 1.712729241773726e-05, "loss": 2.289, "step": 10956 }, { "epoch": 0.27, "learning_rate": 1.7126732414234447e-05, "loss": 2.0756, "step": 10957 }, { "epoch": 0.27, "learning_rate": 1.7126172365310645e-05, "loss": 2.0867, "step": 10958 }, { "epoch": 0.27, "learning_rate": 1.7125612270969428e-05, "loss": 2.1244, "step": 10959 }, { "epoch": 0.27, "learning_rate": 1.7125052131214357e-05, "loss": 2.1769, "step": 10960 }, { "epoch": 0.27, "learning_rate": 1.712449194604901e-05, "loss": 2.0945, "step": 10961 }, { "epoch": 0.27, "learning_rate": 1.7123931715476952e-05, "loss": 2.1711, "step": 10962 }, { "epoch": 0.27, "learning_rate": 1.7123371439501754e-05, "loss": 2.2553, "step": 10963 }, { "epoch": 0.27, "learning_rate": 1.7122811118126993e-05, "loss": 2.2774, "step": 10964 }, { "epoch": 0.27, "learning_rate": 1.7122250751356233e-05, "loss": 2.2014, "step": 10965 }, { "epoch": 0.27, "learning_rate": 1.7121690339193048e-05, "loss": 2.359, "step": 10966 }, { "epoch": 0.27, "learning_rate": 1.7121129881641006e-05, "loss": 2.1133, "step": 10967 }, { "epoch": 0.27, "learning_rate": 1.712056937870369e-05, "loss": 2.1116, "step": 10968 }, { "epoch": 0.27, "learning_rate": 1.712000883038466e-05, "loss": 2.0551, "step": 10969 }, { "epoch": 0.27, "learning_rate": 1.7119448236687496e-05, "loss": 2.0302, "step": 10970 }, { "epoch": 0.27, "learning_rate": 1.7118887597615765e-05, "loss": 2.2922, "step": 10971 }, { "epoch": 0.27, "learning_rate": 1.7118326913173047e-05, "loss": 2.3642, "step": 10972 }, { "epoch": 0.27, "learning_rate": 1.711776618336291e-05, "loss": 2.2624, "step": 10973 }, { "epoch": 0.27, "learning_rate": 1.7117205408188928e-05, "loss": 2.1464, "step": 10974 }, { "epoch": 0.27, "learning_rate": 1.711664458765468e-05, "loss": 2.1426, "step": 10975 }, { "epoch": 0.27, "learning_rate": 1.711608372176373e-05, "loss": 2.0996, "step": 10976 }, { "epoch": 0.27, "learning_rate": 1.7115522810519663e-05, "loss": 2.024, "step": 10977 }, { "epoch": 0.27, "learning_rate": 1.7114961853926052e-05, "loss": 2.2142, "step": 10978 }, { "epoch": 0.27, "learning_rate": 1.711440085198647e-05, "loss": 2.246, "step": 10979 }, { "epoch": 0.27, "learning_rate": 1.7113839804704494e-05, "loss": 2.1048, "step": 10980 }, { "epoch": 0.27, "learning_rate": 1.7113278712083696e-05, "loss": 2.2163, "step": 10981 }, { "epoch": 0.27, "learning_rate": 1.7112717574127656e-05, "loss": 2.0956, "step": 10982 }, { "epoch": 0.27, "learning_rate": 1.711215639083995e-05, "loss": 2.067, "step": 10983 }, { "epoch": 0.27, "learning_rate": 1.711159516222415e-05, "loss": 2.1925, "step": 10984 }, { "epoch": 0.27, "learning_rate": 1.7111033888283838e-05, "loss": 2.1309, "step": 10985 }, { "epoch": 0.27, "learning_rate": 1.711047256902259e-05, "loss": 2.2564, "step": 10986 }, { "epoch": 0.27, "learning_rate": 1.7109911204443986e-05, "loss": 2.1808, "step": 10987 }, { "epoch": 0.27, "learning_rate": 1.71093497945516e-05, "loss": 2.2487, "step": 10988 }, { "epoch": 0.27, "learning_rate": 1.7108788339349004e-05, "loss": 2.0253, "step": 10989 }, { "epoch": 0.27, "learning_rate": 1.710822683883979e-05, "loss": 2.1921, "step": 10990 }, { "epoch": 0.27, "learning_rate": 1.7107665293027528e-05, "loss": 2.1259, "step": 10991 }, { "epoch": 0.27, "learning_rate": 1.71071037019158e-05, "loss": 2.0529, "step": 10992 }, { "epoch": 0.27, "learning_rate": 1.710654206550818e-05, "loss": 2.1034, "step": 10993 }, { "epoch": 0.27, "learning_rate": 1.7105980383808257e-05, "loss": 2.0229, "step": 10994 }, { "epoch": 0.27, "learning_rate": 1.71054186568196e-05, "loss": 2.3077, "step": 10995 }, { "epoch": 0.27, "learning_rate": 1.71048568845458e-05, "loss": 2.1956, "step": 10996 }, { "epoch": 0.27, "learning_rate": 1.7104295066990428e-05, "loss": 2.2937, "step": 10997 }, { "epoch": 0.27, "learning_rate": 1.7103733204157065e-05, "loss": 2.1704, "step": 10998 }, { "epoch": 0.27, "learning_rate": 1.71031712960493e-05, "loss": 2.1406, "step": 10999 }, { "epoch": 0.27, "learning_rate": 1.710260934267071e-05, "loss": 2.0648, "step": 11000 }, { "epoch": 0.27, "learning_rate": 1.7102047344024876e-05, "loss": 2.1291, "step": 11001 }, { "epoch": 0.27, "learning_rate": 1.7101485300115375e-05, "loss": 2.2419, "step": 11002 }, { "epoch": 0.27, "learning_rate": 1.71009232109458e-05, "loss": 2.2742, "step": 11003 }, { "epoch": 0.27, "learning_rate": 1.7100361076519723e-05, "loss": 2.2143, "step": 11004 }, { "epoch": 0.27, "learning_rate": 1.7099798896840734e-05, "loss": 2.1237, "step": 11005 }, { "epoch": 0.27, "learning_rate": 1.7099236671912408e-05, "loss": 2.2342, "step": 11006 }, { "epoch": 0.27, "learning_rate": 1.7098674401738337e-05, "loss": 2.2405, "step": 11007 }, { "epoch": 0.27, "learning_rate": 1.70981120863221e-05, "loss": 2.0341, "step": 11008 }, { "epoch": 0.27, "learning_rate": 1.7097549725667282e-05, "loss": 2.2034, "step": 11009 }, { "epoch": 0.27, "learning_rate": 1.7096987319777466e-05, "loss": 2.1645, "step": 11010 }, { "epoch": 0.27, "learning_rate": 1.7096424868656235e-05, "loss": 2.1514, "step": 11011 }, { "epoch": 0.27, "learning_rate": 1.709586237230718e-05, "loss": 2.0998, "step": 11012 }, { "epoch": 0.27, "learning_rate": 1.709529983073388e-05, "loss": 2.2145, "step": 11013 }, { "epoch": 0.27, "learning_rate": 1.709473724393992e-05, "loss": 2.045, "step": 11014 }, { "epoch": 0.27, "learning_rate": 1.7094174611928885e-05, "loss": 2.1684, "step": 11015 }, { "epoch": 0.27, "learning_rate": 1.709361193470437e-05, "loss": 2.0221, "step": 11016 }, { "epoch": 0.27, "learning_rate": 1.709304921226995e-05, "loss": 2.1537, "step": 11017 }, { "epoch": 0.27, "learning_rate": 1.7092486444629218e-05, "loss": 2.1986, "step": 11018 }, { "epoch": 0.27, "learning_rate": 1.7091923631785756e-05, "loss": 2.0614, "step": 11019 }, { "epoch": 0.27, "learning_rate": 1.7091360773743155e-05, "loss": 2.022, "step": 11020 }, { "epoch": 0.27, "learning_rate": 1.7090797870505e-05, "loss": 2.2267, "step": 11021 }, { "epoch": 0.27, "learning_rate": 1.7090234922074883e-05, "loss": 2.0947, "step": 11022 }, { "epoch": 0.27, "learning_rate": 1.7089671928456385e-05, "loss": 1.9958, "step": 11023 }, { "epoch": 0.27, "learning_rate": 1.7089108889653098e-05, "loss": 1.9919, "step": 11024 }, { "epoch": 0.27, "learning_rate": 1.708854580566861e-05, "loss": 2.2254, "step": 11025 }, { "epoch": 0.27, "learning_rate": 1.708798267650651e-05, "loss": 2.1724, "step": 11026 }, { "epoch": 0.27, "learning_rate": 1.7087419502170385e-05, "loss": 2.1467, "step": 11027 }, { "epoch": 0.27, "learning_rate": 1.7086856282663826e-05, "loss": 2.0512, "step": 11028 }, { "epoch": 0.27, "learning_rate": 1.708629301799042e-05, "loss": 2.237, "step": 11029 }, { "epoch": 0.27, "learning_rate": 1.7085729708153765e-05, "loss": 2.1573, "step": 11030 }, { "epoch": 0.27, "learning_rate": 1.7085166353157442e-05, "loss": 2.2011, "step": 11031 }, { "epoch": 0.27, "learning_rate": 1.7084602953005044e-05, "loss": 2.2006, "step": 11032 }, { "epoch": 0.27, "learning_rate": 1.7084039507700162e-05, "loss": 2.154, "step": 11033 }, { "epoch": 0.27, "learning_rate": 1.708347601724639e-05, "loss": 1.9002, "step": 11034 }, { "epoch": 0.27, "learning_rate": 1.7082912481647316e-05, "loss": 2.1809, "step": 11035 }, { "epoch": 0.27, "learning_rate": 1.7082348900906535e-05, "loss": 2.225, "step": 11036 }, { "epoch": 0.27, "learning_rate": 1.7081785275027632e-05, "loss": 2.292, "step": 11037 }, { "epoch": 0.27, "learning_rate": 1.7081221604014206e-05, "loss": 1.9424, "step": 11038 }, { "epoch": 0.27, "learning_rate": 1.708065788786985e-05, "loss": 2.1465, "step": 11039 }, { "epoch": 0.27, "learning_rate": 1.7080094126598147e-05, "loss": 2.2159, "step": 11040 }, { "epoch": 0.27, "learning_rate": 1.70795303202027e-05, "loss": 2.198, "step": 11041 }, { "epoch": 0.27, "learning_rate": 1.7078966468687098e-05, "loss": 1.8536, "step": 11042 }, { "epoch": 0.27, "learning_rate": 1.707840257205494e-05, "loss": 2.1473, "step": 11043 }, { "epoch": 0.27, "learning_rate": 1.707783863030981e-05, "loss": 1.9989, "step": 11044 }, { "epoch": 0.27, "learning_rate": 1.7077274643455313e-05, "loss": 2.1197, "step": 11045 }, { "epoch": 0.27, "learning_rate": 1.7076710611495032e-05, "loss": 1.9906, "step": 11046 }, { "epoch": 0.27, "learning_rate": 1.7076146534432575e-05, "loss": 2.2191, "step": 11047 }, { "epoch": 0.27, "learning_rate": 1.707558241227153e-05, "loss": 2.1633, "step": 11048 }, { "epoch": 0.27, "learning_rate": 1.7075018245015485e-05, "loss": 2.1548, "step": 11049 }, { "epoch": 0.27, "learning_rate": 1.707445403266805e-05, "loss": 1.9836, "step": 11050 }, { "epoch": 0.27, "learning_rate": 1.707388977523281e-05, "loss": 2.1621, "step": 11051 }, { "epoch": 0.27, "learning_rate": 1.7073325472713363e-05, "loss": 2.2246, "step": 11052 }, { "epoch": 0.27, "learning_rate": 1.7072761125113312e-05, "loss": 2.2182, "step": 11053 }, { "epoch": 0.27, "learning_rate": 1.7072196732436252e-05, "loss": 1.9996, "step": 11054 }, { "epoch": 0.27, "learning_rate": 1.7071632294685773e-05, "loss": 2.3108, "step": 11055 }, { "epoch": 0.27, "learning_rate": 1.7071067811865477e-05, "loss": 2.0598, "step": 11056 }, { "epoch": 0.27, "learning_rate": 1.7070503283978963e-05, "loss": 2.1167, "step": 11057 }, { "epoch": 0.27, "learning_rate": 1.7069938711029825e-05, "loss": 2.2714, "step": 11058 }, { "epoch": 0.27, "learning_rate": 1.7069374093021665e-05, "loss": 2.1859, "step": 11059 }, { "epoch": 0.27, "learning_rate": 1.706880942995808e-05, "loss": 2.0764, "step": 11060 }, { "epoch": 0.27, "learning_rate": 1.706824472184267e-05, "loss": 2.1247, "step": 11061 }, { "epoch": 0.27, "learning_rate": 1.706767996867903e-05, "loss": 2.1901, "step": 11062 }, { "epoch": 0.27, "learning_rate": 1.7067115170470767e-05, "loss": 2.2425, "step": 11063 }, { "epoch": 0.27, "learning_rate": 1.7066550327221472e-05, "loss": 1.9843, "step": 11064 }, { "epoch": 0.27, "learning_rate": 1.7065985438934748e-05, "loss": 2.2482, "step": 11065 }, { "epoch": 0.27, "learning_rate": 1.7065420505614203e-05, "loss": 2.0688, "step": 11066 }, { "epoch": 0.27, "learning_rate": 1.7064855527263424e-05, "loss": 2.1913, "step": 11067 }, { "epoch": 0.27, "learning_rate": 1.7064290503886017e-05, "loss": 2.3027, "step": 11068 }, { "epoch": 0.27, "learning_rate": 1.7063725435485587e-05, "loss": 1.9918, "step": 11069 }, { "epoch": 0.27, "learning_rate": 1.7063160322065735e-05, "loss": 2.2695, "step": 11070 }, { "epoch": 0.27, "learning_rate": 1.706259516363006e-05, "loss": 2.0588, "step": 11071 }, { "epoch": 0.27, "learning_rate": 1.7062029960182163e-05, "loss": 2.254, "step": 11072 }, { "epoch": 0.27, "learning_rate": 1.7061464711725645e-05, "loss": 2.0819, "step": 11073 }, { "epoch": 0.27, "learning_rate": 1.7060899418264116e-05, "loss": 2.1792, "step": 11074 }, { "epoch": 0.27, "learning_rate": 1.706033407980117e-05, "loss": 2.0549, "step": 11075 }, { "epoch": 0.27, "learning_rate": 1.7059768696340414e-05, "loss": 2.1441, "step": 11076 }, { "epoch": 0.27, "learning_rate": 1.7059203267885455e-05, "loss": 2.0935, "step": 11077 }, { "epoch": 0.27, "learning_rate": 1.705863779443989e-05, "loss": 2.285, "step": 11078 }, { "epoch": 0.27, "learning_rate": 1.7058072276007328e-05, "loss": 2.2796, "step": 11079 }, { "epoch": 0.27, "learning_rate": 1.7057506712591373e-05, "loss": 2.0573, "step": 11080 }, { "epoch": 0.27, "learning_rate": 1.7056941104195625e-05, "loss": 2.2805, "step": 11081 }, { "epoch": 0.27, "learning_rate": 1.705637545082369e-05, "loss": 2.158, "step": 11082 }, { "epoch": 0.27, "learning_rate": 1.7055809752479176e-05, "loss": 2.0505, "step": 11083 }, { "epoch": 0.27, "learning_rate": 1.7055244009165685e-05, "loss": 1.9874, "step": 11084 }, { "epoch": 0.27, "learning_rate": 1.705467822088683e-05, "loss": 2.2816, "step": 11085 }, { "epoch": 0.27, "learning_rate": 1.705411238764621e-05, "loss": 2.0403, "step": 11086 }, { "epoch": 0.27, "learning_rate": 1.705354650944743e-05, "loss": 2.0505, "step": 11087 }, { "epoch": 0.27, "learning_rate": 1.7052980586294098e-05, "loss": 2.1876, "step": 11088 }, { "epoch": 0.27, "learning_rate": 1.7052414618189827e-05, "loss": 2.1599, "step": 11089 }, { "epoch": 0.27, "learning_rate": 1.7051848605138216e-05, "loss": 2.1272, "step": 11090 }, { "epoch": 0.27, "learning_rate": 1.705128254714288e-05, "loss": 2.1408, "step": 11091 }, { "epoch": 0.27, "learning_rate": 1.705071644420742e-05, "loss": 2.0262, "step": 11092 }, { "epoch": 0.27, "learning_rate": 1.7050150296335447e-05, "loss": 2.3109, "step": 11093 }, { "epoch": 0.27, "learning_rate": 1.7049584103530567e-05, "loss": 2.0115, "step": 11094 }, { "epoch": 0.27, "learning_rate": 1.704901786579639e-05, "loss": 2.0973, "step": 11095 }, { "epoch": 0.27, "learning_rate": 1.7048451583136526e-05, "loss": 2.2112, "step": 11096 }, { "epoch": 0.27, "learning_rate": 1.7047885255554582e-05, "loss": 2.0448, "step": 11097 }, { "epoch": 0.27, "learning_rate": 1.7047318883054166e-05, "loss": 2.1987, "step": 11098 }, { "epoch": 0.27, "learning_rate": 1.7046752465638893e-05, "loss": 1.9917, "step": 11099 }, { "epoch": 0.27, "learning_rate": 1.704618600331237e-05, "loss": 2.134, "step": 11100 }, { "epoch": 0.27, "learning_rate": 1.704561949607821e-05, "loss": 2.1529, "step": 11101 }, { "epoch": 0.27, "learning_rate": 1.7045052943940017e-05, "loss": 2.2963, "step": 11102 }, { "epoch": 0.27, "learning_rate": 1.704448634690141e-05, "loss": 2.2353, "step": 11103 }, { "epoch": 0.27, "learning_rate": 1.704391970496599e-05, "loss": 2.1338, "step": 11104 }, { "epoch": 0.27, "learning_rate": 1.7043353018137376e-05, "loss": 2.0514, "step": 11105 }, { "epoch": 0.27, "learning_rate": 1.704278628641918e-05, "loss": 2.1142, "step": 11106 }, { "epoch": 0.27, "learning_rate": 1.7042219509815012e-05, "loss": 2.1978, "step": 11107 }, { "epoch": 0.27, "learning_rate": 1.7041652688328482e-05, "loss": 2.108, "step": 11108 }, { "epoch": 0.27, "learning_rate": 1.7041085821963202e-05, "loss": 2.0551, "step": 11109 }, { "epoch": 0.27, "learning_rate": 1.7040518910722794e-05, "loss": 2.3799, "step": 11110 }, { "epoch": 0.27, "learning_rate": 1.703995195461086e-05, "loss": 2.234, "step": 11111 }, { "epoch": 0.27, "learning_rate": 1.7039384953631018e-05, "loss": 2.1901, "step": 11112 }, { "epoch": 0.27, "learning_rate": 1.7038817907786884e-05, "loss": 2.337, "step": 11113 }, { "epoch": 0.27, "learning_rate": 1.7038250817082068e-05, "loss": 2.2819, "step": 11114 }, { "epoch": 0.27, "learning_rate": 1.7037683681520187e-05, "loss": 2.2306, "step": 11115 }, { "epoch": 0.27, "learning_rate": 1.7037116501104855e-05, "loss": 2.1278, "step": 11116 }, { "epoch": 0.27, "learning_rate": 1.7036549275839684e-05, "loss": 2.205, "step": 11117 }, { "epoch": 0.27, "learning_rate": 1.7035982005728292e-05, "loss": 2.0179, "step": 11118 }, { "epoch": 0.27, "learning_rate": 1.7035414690774296e-05, "loss": 2.1113, "step": 11119 }, { "epoch": 0.27, "learning_rate": 1.7034847330981306e-05, "loss": 2.3081, "step": 11120 }, { "epoch": 0.27, "learning_rate": 1.703427992635294e-05, "loss": 2.2046, "step": 11121 }, { "epoch": 0.27, "learning_rate": 1.703371247689282e-05, "loss": 2.2101, "step": 11122 }, { "epoch": 0.27, "learning_rate": 1.7033144982604557e-05, "loss": 2.135, "step": 11123 }, { "epoch": 0.27, "learning_rate": 1.7032577443491768e-05, "loss": 2.1901, "step": 11124 }, { "epoch": 0.27, "learning_rate": 1.703200985955807e-05, "loss": 2.1162, "step": 11125 }, { "epoch": 0.27, "learning_rate": 1.7031442230807085e-05, "loss": 2.0626, "step": 11126 }, { "epoch": 0.27, "learning_rate": 1.7030874557242426e-05, "loss": 2.28, "step": 11127 }, { "epoch": 0.27, "learning_rate": 1.703030683886771e-05, "loss": 2.1728, "step": 11128 }, { "epoch": 0.27, "learning_rate": 1.702973907568656e-05, "loss": 2.0825, "step": 11129 }, { "epoch": 0.27, "learning_rate": 1.702917126770259e-05, "loss": 2.1975, "step": 11130 }, { "epoch": 0.27, "learning_rate": 1.702860341491942e-05, "loss": 2.0843, "step": 11131 }, { "epoch": 0.27, "learning_rate": 1.7028035517340673e-05, "loss": 2.0318, "step": 11132 }, { "epoch": 0.27, "learning_rate": 1.702746757496996e-05, "loss": 2.1969, "step": 11133 }, { "epoch": 0.27, "learning_rate": 1.7026899587810913e-05, "loss": 2.2151, "step": 11134 }, { "epoch": 0.27, "learning_rate": 1.7026331555867145e-05, "loss": 2.2091, "step": 11135 }, { "epoch": 0.27, "learning_rate": 1.702576347914227e-05, "loss": 1.9981, "step": 11136 }, { "epoch": 0.27, "learning_rate": 1.7025195357639917e-05, "loss": 2.241, "step": 11137 }, { "epoch": 0.27, "learning_rate": 1.7024627191363706e-05, "loss": 2.28, "step": 11138 }, { "epoch": 0.27, "learning_rate": 1.7024058980317255e-05, "loss": 2.175, "step": 11139 }, { "epoch": 0.27, "learning_rate": 1.7023490724504187e-05, "loss": 2.2361, "step": 11140 }, { "epoch": 0.27, "learning_rate": 1.7022922423928124e-05, "loss": 2.2653, "step": 11141 }, { "epoch": 0.27, "learning_rate": 1.702235407859269e-05, "loss": 2.1293, "step": 11142 }, { "epoch": 0.27, "learning_rate": 1.70217856885015e-05, "loss": 2.0326, "step": 11143 }, { "epoch": 0.27, "learning_rate": 1.7021217253658187e-05, "loss": 2.3191, "step": 11144 }, { "epoch": 0.27, "learning_rate": 1.7020648774066367e-05, "loss": 2.3792, "step": 11145 }, { "epoch": 0.27, "learning_rate": 1.7020080249729662e-05, "loss": 2.1745, "step": 11146 }, { "epoch": 0.27, "learning_rate": 1.7019511680651697e-05, "loss": 2.3515, "step": 11147 }, { "epoch": 0.27, "learning_rate": 1.70189430668361e-05, "loss": 1.9946, "step": 11148 }, { "epoch": 0.27, "learning_rate": 1.701837440828649e-05, "loss": 2.3115, "step": 11149 }, { "epoch": 0.27, "learning_rate": 1.7017805705006493e-05, "loss": 2.2419, "step": 11150 }, { "epoch": 0.27, "learning_rate": 1.7017236956999734e-05, "loss": 2.1664, "step": 11151 }, { "epoch": 0.27, "learning_rate": 1.7016668164269834e-05, "loss": 2.1224, "step": 11152 }, { "epoch": 0.27, "learning_rate": 1.7016099326820424e-05, "loss": 2.1245, "step": 11153 }, { "epoch": 0.27, "learning_rate": 1.7015530444655125e-05, "loss": 2.1697, "step": 11154 }, { "epoch": 0.27, "learning_rate": 1.7014961517777565e-05, "loss": 2.0707, "step": 11155 }, { "epoch": 0.27, "learning_rate": 1.7014392546191367e-05, "loss": 1.9273, "step": 11156 }, { "epoch": 0.28, "learning_rate": 1.7013823529900162e-05, "loss": 2.1972, "step": 11157 }, { "epoch": 0.28, "learning_rate": 1.7013254468907573e-05, "loss": 2.2959, "step": 11158 }, { "epoch": 0.28, "learning_rate": 1.7012685363217228e-05, "loss": 2.2076, "step": 11159 }, { "epoch": 0.28, "learning_rate": 1.7012116212832752e-05, "loss": 1.9738, "step": 11160 }, { "epoch": 0.28, "learning_rate": 1.7011547017757775e-05, "loss": 1.9536, "step": 11161 }, { "epoch": 0.28, "learning_rate": 1.7010977777995927e-05, "loss": 2.2134, "step": 11162 }, { "epoch": 0.28, "learning_rate": 1.701040849355083e-05, "loss": 2.1151, "step": 11163 }, { "epoch": 0.28, "learning_rate": 1.7009839164426115e-05, "loss": 2.1936, "step": 11164 }, { "epoch": 0.28, "learning_rate": 1.700926979062541e-05, "loss": 2.2022, "step": 11165 }, { "epoch": 0.28, "learning_rate": 1.7008700372152347e-05, "loss": 2.2132, "step": 11166 }, { "epoch": 0.28, "learning_rate": 1.700813090901055e-05, "loss": 2.0378, "step": 11167 }, { "epoch": 0.28, "learning_rate": 1.700756140120365e-05, "loss": 2.2477, "step": 11168 }, { "epoch": 0.28, "learning_rate": 1.700699184873528e-05, "loss": 2.2761, "step": 11169 }, { "epoch": 0.28, "learning_rate": 1.7006422251609067e-05, "loss": 2.0547, "step": 11170 }, { "epoch": 0.28, "learning_rate": 1.7005852609828643e-05, "loss": 2.0977, "step": 11171 }, { "epoch": 0.28, "learning_rate": 1.7005282923397635e-05, "loss": 1.9758, "step": 11172 }, { "epoch": 0.28, "learning_rate": 1.7004713192319677e-05, "loss": 2.1646, "step": 11173 }, { "epoch": 0.28, "learning_rate": 1.7004143416598396e-05, "loss": 2.0662, "step": 11174 }, { "epoch": 0.28, "learning_rate": 1.700357359623743e-05, "loss": 2.2147, "step": 11175 }, { "epoch": 0.28, "learning_rate": 1.7003003731240407e-05, "loss": 2.0886, "step": 11176 }, { "epoch": 0.28, "learning_rate": 1.7002433821610956e-05, "loss": 2.2163, "step": 11177 }, { "epoch": 0.28, "learning_rate": 1.7001863867352715e-05, "loss": 2.2402, "step": 11178 }, { "epoch": 0.28, "learning_rate": 1.700129386846931e-05, "loss": 2.3585, "step": 11179 }, { "epoch": 0.28, "learning_rate": 1.7000723824964382e-05, "loss": 2.2003, "step": 11180 }, { "epoch": 0.28, "learning_rate": 1.700015373684156e-05, "loss": 2.0826, "step": 11181 }, { "epoch": 0.28, "learning_rate": 1.6999583604104477e-05, "loss": 2.2261, "step": 11182 }, { "epoch": 0.28, "learning_rate": 1.699901342675676e-05, "loss": 2.2488, "step": 11183 }, { "epoch": 0.28, "learning_rate": 1.6998443204802057e-05, "loss": 2.2203, "step": 11184 }, { "epoch": 0.28, "learning_rate": 1.6997872938243988e-05, "loss": 2.0539, "step": 11185 }, { "epoch": 0.28, "learning_rate": 1.6997302627086202e-05, "loss": 2.0269, "step": 11186 }, { "epoch": 0.28, "learning_rate": 1.699673227133232e-05, "loss": 2.2169, "step": 11187 }, { "epoch": 0.28, "learning_rate": 1.6996161870985987e-05, "loss": 2.2038, "step": 11188 }, { "epoch": 0.28, "learning_rate": 1.6995591426050833e-05, "loss": 2.0629, "step": 11189 }, { "epoch": 0.28, "learning_rate": 1.6995020936530493e-05, "loss": 2.1984, "step": 11190 }, { "epoch": 0.28, "learning_rate": 1.6994450402428607e-05, "loss": 2.1539, "step": 11191 }, { "epoch": 0.28, "learning_rate": 1.699387982374881e-05, "loss": 2.2142, "step": 11192 }, { "epoch": 0.28, "learning_rate": 1.6993309200494735e-05, "loss": 2.132, "step": 11193 }, { "epoch": 0.28, "learning_rate": 1.6992738532670023e-05, "loss": 2.042, "step": 11194 }, { "epoch": 0.28, "learning_rate": 1.699216782027831e-05, "loss": 1.9764, "step": 11195 }, { "epoch": 0.28, "learning_rate": 1.6991597063323228e-05, "loss": 2.0992, "step": 11196 }, { "epoch": 0.28, "learning_rate": 1.6991026261808426e-05, "loss": 2.2219, "step": 11197 }, { "epoch": 0.28, "learning_rate": 1.6990455415737536e-05, "loss": 2.075, "step": 11198 }, { "epoch": 0.28, "learning_rate": 1.6989884525114187e-05, "loss": 2.191, "step": 11199 }, { "epoch": 0.28, "learning_rate": 1.6989313589942035e-05, "loss": 2.128, "step": 11200 }, { "epoch": 0.28, "learning_rate": 1.6988742610224704e-05, "loss": 2.0713, "step": 11201 }, { "epoch": 0.28, "learning_rate": 1.698817158596584e-05, "loss": 2.254, "step": 11202 }, { "epoch": 0.28, "learning_rate": 1.698760051716908e-05, "loss": 2.1123, "step": 11203 }, { "epoch": 0.28, "learning_rate": 1.698702940383807e-05, "loss": 2.1469, "step": 11204 }, { "epoch": 0.28, "learning_rate": 1.698645824597644e-05, "loss": 2.017, "step": 11205 }, { "epoch": 0.28, "learning_rate": 1.6985887043587835e-05, "loss": 2.2276, "step": 11206 }, { "epoch": 0.28, "learning_rate": 1.69853157966759e-05, "loss": 2.1671, "step": 11207 }, { "epoch": 0.28, "learning_rate": 1.6984744505244264e-05, "loss": 2.087, "step": 11208 }, { "epoch": 0.28, "learning_rate": 1.698417316929658e-05, "loss": 2.0287, "step": 11209 }, { "epoch": 0.28, "learning_rate": 1.6983601788836482e-05, "loss": 2.1563, "step": 11210 }, { "epoch": 0.28, "learning_rate": 1.6983030363867616e-05, "loss": 2.1928, "step": 11211 }, { "epoch": 0.28, "learning_rate": 1.698245889439362e-05, "loss": 1.9993, "step": 11212 }, { "epoch": 0.28, "learning_rate": 1.6981887380418135e-05, "loss": 2.0316, "step": 11213 }, { "epoch": 0.28, "learning_rate": 1.698131582194481e-05, "loss": 2.045, "step": 11214 }, { "epoch": 0.28, "learning_rate": 1.6980744218977284e-05, "loss": 2.1518, "step": 11215 }, { "epoch": 0.28, "learning_rate": 1.6980172571519202e-05, "loss": 2.1811, "step": 11216 }, { "epoch": 0.28, "learning_rate": 1.6979600879574204e-05, "loss": 2.1791, "step": 11217 }, { "epoch": 0.28, "learning_rate": 1.697902914314593e-05, "loss": 2.0727, "step": 11218 }, { "epoch": 0.28, "learning_rate": 1.6978457362238033e-05, "loss": 2.1221, "step": 11219 }, { "epoch": 0.28, "learning_rate": 1.6977885536854155e-05, "loss": 2.2445, "step": 11220 }, { "epoch": 0.28, "learning_rate": 1.6977313666997937e-05, "loss": 2.1389, "step": 11221 }, { "epoch": 0.28, "learning_rate": 1.697674175267302e-05, "loss": 2.0468, "step": 11222 }, { "epoch": 0.28, "learning_rate": 1.697616979388306e-05, "loss": 2.1895, "step": 11223 }, { "epoch": 0.28, "learning_rate": 1.6975597790631696e-05, "loss": 1.9243, "step": 11224 }, { "epoch": 0.28, "learning_rate": 1.6975025742922574e-05, "loss": 2.1427, "step": 11225 }, { "epoch": 0.28, "learning_rate": 1.6974453650759343e-05, "loss": 2.2689, "step": 11226 }, { "epoch": 0.28, "learning_rate": 1.697388151414564e-05, "loss": 2.1752, "step": 11227 }, { "epoch": 0.28, "learning_rate": 1.697330933308512e-05, "loss": 2.116, "step": 11228 }, { "epoch": 0.28, "learning_rate": 1.6972737107581426e-05, "loss": 2.0495, "step": 11229 }, { "epoch": 0.28, "learning_rate": 1.6972164837638208e-05, "loss": 2.15, "step": 11230 }, { "epoch": 0.28, "learning_rate": 1.697159252325911e-05, "loss": 2.102, "step": 11231 }, { "epoch": 0.28, "learning_rate": 1.697102016444778e-05, "loss": 2.2189, "step": 11232 }, { "epoch": 0.28, "learning_rate": 1.6970447761207867e-05, "loss": 2.2039, "step": 11233 }, { "epoch": 0.28, "learning_rate": 1.696987531354302e-05, "loss": 2.1189, "step": 11234 }, { "epoch": 0.28, "learning_rate": 1.6969302821456885e-05, "loss": 2.1626, "step": 11235 }, { "epoch": 0.28, "learning_rate": 1.6968730284953113e-05, "loss": 2.3065, "step": 11236 }, { "epoch": 0.28, "learning_rate": 1.696815770403535e-05, "loss": 2.1459, "step": 11237 }, { "epoch": 0.28, "learning_rate": 1.6967585078707246e-05, "loss": 2.0431, "step": 11238 }, { "epoch": 0.28, "learning_rate": 1.6967012408972455e-05, "loss": 2.1152, "step": 11239 }, { "epoch": 0.28, "learning_rate": 1.6966439694834626e-05, "loss": 2.1486, "step": 11240 }, { "epoch": 0.28, "learning_rate": 1.6965866936297403e-05, "loss": 2.0459, "step": 11241 }, { "epoch": 0.28, "learning_rate": 1.6965294133364438e-05, "loss": 2.1944, "step": 11242 }, { "epoch": 0.28, "learning_rate": 1.696472128603939e-05, "loss": 2.2754, "step": 11243 }, { "epoch": 0.28, "learning_rate": 1.6964148394325896e-05, "loss": 2.2637, "step": 11244 }, { "epoch": 0.28, "learning_rate": 1.6963575458227616e-05, "loss": 2.1386, "step": 11245 }, { "epoch": 0.28, "learning_rate": 1.6963002477748204e-05, "loss": 2.048, "step": 11246 }, { "epoch": 0.28, "learning_rate": 1.6962429452891303e-05, "loss": 2.077, "step": 11247 }, { "epoch": 0.28, "learning_rate": 1.6961856383660572e-05, "loss": 2.1383, "step": 11248 }, { "epoch": 0.28, "learning_rate": 1.6961283270059667e-05, "loss": 2.1609, "step": 11249 }, { "epoch": 0.28, "learning_rate": 1.696071011209223e-05, "loss": 2.4392, "step": 11250 }, { "epoch": 0.28, "learning_rate": 1.6960136909761916e-05, "loss": 2.1843, "step": 11251 }, { "epoch": 0.28, "learning_rate": 1.695956366307239e-05, "loss": 1.9829, "step": 11252 }, { "epoch": 0.28, "learning_rate": 1.695899037202729e-05, "loss": 2.1356, "step": 11253 }, { "epoch": 0.28, "learning_rate": 1.695841703663028e-05, "loss": 2.0568, "step": 11254 }, { "epoch": 0.28, "learning_rate": 1.6957843656885008e-05, "loss": 2.3152, "step": 11255 }, { "epoch": 0.28, "learning_rate": 1.695727023279513e-05, "loss": 2.237, "step": 11256 }, { "epoch": 0.28, "learning_rate": 1.6956696764364302e-05, "loss": 2.2168, "step": 11257 }, { "epoch": 0.28, "learning_rate": 1.695612325159618e-05, "loss": 2.0842, "step": 11258 }, { "epoch": 0.28, "learning_rate": 1.6955549694494417e-05, "loss": 2.1516, "step": 11259 }, { "epoch": 0.28, "learning_rate": 1.695497609306267e-05, "loss": 2.1922, "step": 11260 }, { "epoch": 0.28, "learning_rate": 1.695440244730459e-05, "loss": 2.1524, "step": 11261 }, { "epoch": 0.28, "learning_rate": 1.695382875722384e-05, "loss": 2.0469, "step": 11262 }, { "epoch": 0.28, "learning_rate": 1.6953255022824075e-05, "loss": 2.0556, "step": 11263 }, { "epoch": 0.28, "learning_rate": 1.6952681244108945e-05, "loss": 2.1292, "step": 11264 }, { "epoch": 0.28, "learning_rate": 1.6952107421082117e-05, "loss": 2.0233, "step": 11265 }, { "epoch": 0.28, "learning_rate": 1.695153355374724e-05, "loss": 2.3058, "step": 11266 }, { "epoch": 0.28, "learning_rate": 1.6950959642107975e-05, "loss": 2.0462, "step": 11267 }, { "epoch": 0.28, "learning_rate": 1.6950385686167975e-05, "loss": 2.2235, "step": 11268 }, { "epoch": 0.28, "learning_rate": 1.6949811685930904e-05, "loss": 1.9168, "step": 11269 }, { "epoch": 0.28, "learning_rate": 1.6949237641400424e-05, "loss": 2.0564, "step": 11270 }, { "epoch": 0.28, "learning_rate": 1.694866355258018e-05, "loss": 1.9594, "step": 11271 }, { "epoch": 0.28, "learning_rate": 1.6948089419473844e-05, "loss": 2.1077, "step": 11272 }, { "epoch": 0.28, "learning_rate": 1.6947515242085067e-05, "loss": 2.2209, "step": 11273 }, { "epoch": 0.28, "learning_rate": 1.694694102041751e-05, "loss": 2.1333, "step": 11274 }, { "epoch": 0.28, "learning_rate": 1.694636675447484e-05, "loss": 2.094, "step": 11275 }, { "epoch": 0.28, "learning_rate": 1.6945792444260705e-05, "loss": 2.3238, "step": 11276 }, { "epoch": 0.28, "learning_rate": 1.694521808977877e-05, "loss": 2.2333, "step": 11277 }, { "epoch": 0.28, "learning_rate": 1.6944643691032702e-05, "loss": 1.9829, "step": 11278 }, { "epoch": 0.28, "learning_rate": 1.6944069248026153e-05, "loss": 2.0597, "step": 11279 }, { "epoch": 0.28, "learning_rate": 1.694349476076279e-05, "loss": 2.301, "step": 11280 }, { "epoch": 0.28, "learning_rate": 1.694292022924627e-05, "loss": 2.2172, "step": 11281 }, { "epoch": 0.28, "learning_rate": 1.6942345653480256e-05, "loss": 2.0197, "step": 11282 }, { "epoch": 0.28, "learning_rate": 1.694177103346841e-05, "loss": 2.1886, "step": 11283 }, { "epoch": 0.28, "learning_rate": 1.6941196369214403e-05, "loss": 2.2383, "step": 11284 }, { "epoch": 0.28, "learning_rate": 1.6940621660721883e-05, "loss": 2.0494, "step": 11285 }, { "epoch": 0.28, "learning_rate": 1.694004690799452e-05, "loss": 2.1519, "step": 11286 }, { "epoch": 0.28, "learning_rate": 1.6939472111035976e-05, "loss": 2.2442, "step": 11287 }, { "epoch": 0.28, "learning_rate": 1.6938897269849914e-05, "loss": 2.0572, "step": 11288 }, { "epoch": 0.28, "learning_rate": 1.693832238444e-05, "loss": 1.8935, "step": 11289 }, { "epoch": 0.28, "learning_rate": 1.69377474548099e-05, "loss": 2.0879, "step": 11290 }, { "epoch": 0.28, "learning_rate": 1.6937172480963266e-05, "loss": 1.863, "step": 11291 }, { "epoch": 0.28, "learning_rate": 1.6936597462903775e-05, "loss": 2.3185, "step": 11292 }, { "epoch": 0.28, "learning_rate": 1.693602240063509e-05, "loss": 2.2487, "step": 11293 }, { "epoch": 0.28, "learning_rate": 1.6935447294160874e-05, "loss": 2.3248, "step": 11294 }, { "epoch": 0.28, "learning_rate": 1.693487214348479e-05, "loss": 2.2148, "step": 11295 }, { "epoch": 0.28, "learning_rate": 1.6934296948610504e-05, "loss": 1.9234, "step": 11296 }, { "epoch": 0.28, "learning_rate": 1.6933721709541686e-05, "loss": 2.1848, "step": 11297 }, { "epoch": 0.28, "learning_rate": 1.6933146426282e-05, "loss": 2.1519, "step": 11298 }, { "epoch": 0.28, "learning_rate": 1.6932571098835113e-05, "loss": 2.0951, "step": 11299 }, { "epoch": 0.28, "learning_rate": 1.693199572720469e-05, "loss": 2.1884, "step": 11300 }, { "epoch": 0.28, "learning_rate": 1.6931420311394397e-05, "loss": 2.1885, "step": 11301 }, { "epoch": 0.28, "learning_rate": 1.6930844851407907e-05, "loss": 2.186, "step": 11302 }, { "epoch": 0.28, "learning_rate": 1.6930269347248883e-05, "loss": 2.3438, "step": 11303 }, { "epoch": 0.28, "learning_rate": 1.6929693798920994e-05, "loss": 2.1244, "step": 11304 }, { "epoch": 0.28, "learning_rate": 1.692911820642791e-05, "loss": 1.8868, "step": 11305 }, { "epoch": 0.28, "learning_rate": 1.6928542569773294e-05, "loss": 2.0616, "step": 11306 }, { "epoch": 0.28, "learning_rate": 1.6927966888960822e-05, "loss": 2.0462, "step": 11307 }, { "epoch": 0.28, "learning_rate": 1.6927391163994156e-05, "loss": 2.1905, "step": 11308 }, { "epoch": 0.28, "learning_rate": 1.6926815394876968e-05, "loss": 2.1356, "step": 11309 }, { "epoch": 0.28, "learning_rate": 1.6926239581612928e-05, "loss": 2.2797, "step": 11310 }, { "epoch": 0.28, "learning_rate": 1.6925663724205708e-05, "loss": 2.242, "step": 11311 }, { "epoch": 0.28, "learning_rate": 1.6925087822658977e-05, "loss": 1.9607, "step": 11312 }, { "epoch": 0.28, "learning_rate": 1.6924511876976403e-05, "loss": 2.153, "step": 11313 }, { "epoch": 0.28, "learning_rate": 1.6923935887161656e-05, "loss": 2.1576, "step": 11314 }, { "epoch": 0.28, "learning_rate": 1.692335985321841e-05, "loss": 2.3149, "step": 11315 }, { "epoch": 0.28, "learning_rate": 1.692278377515034e-05, "loss": 2.1058, "step": 11316 }, { "epoch": 0.28, "learning_rate": 1.6922207652961112e-05, "loss": 2.2699, "step": 11317 }, { "epoch": 0.28, "learning_rate": 1.6921631486654396e-05, "loss": 2.0496, "step": 11318 }, { "epoch": 0.28, "learning_rate": 1.6921055276233866e-05, "loss": 2.2485, "step": 11319 }, { "epoch": 0.28, "learning_rate": 1.6920479021703194e-05, "loss": 2.0604, "step": 11320 }, { "epoch": 0.28, "learning_rate": 1.691990272306606e-05, "loss": 2.0444, "step": 11321 }, { "epoch": 0.28, "learning_rate": 1.6919326380326125e-05, "loss": 2.0201, "step": 11322 }, { "epoch": 0.28, "learning_rate": 1.6918749993487073e-05, "loss": 2.0811, "step": 11323 }, { "epoch": 0.28, "learning_rate": 1.6918173562552568e-05, "loss": 2.1815, "step": 11324 }, { "epoch": 0.28, "learning_rate": 1.691759708752629e-05, "loss": 2.2053, "step": 11325 }, { "epoch": 0.28, "learning_rate": 1.6917020568411914e-05, "loss": 2.1311, "step": 11326 }, { "epoch": 0.28, "learning_rate": 1.6916444005213108e-05, "loss": 2.1366, "step": 11327 }, { "epoch": 0.28, "learning_rate": 1.6915867397933553e-05, "loss": 2.2271, "step": 11328 }, { "epoch": 0.28, "learning_rate": 1.691529074657692e-05, "loss": 2.0276, "step": 11329 }, { "epoch": 0.28, "learning_rate": 1.6914714051146886e-05, "loss": 2.0331, "step": 11330 }, { "epoch": 0.28, "learning_rate": 1.6914137311647124e-05, "loss": 2.1442, "step": 11331 }, { "epoch": 0.28, "learning_rate": 1.6913560528081313e-05, "loss": 2.3235, "step": 11332 }, { "epoch": 0.28, "learning_rate": 1.6912983700453126e-05, "loss": 2.1318, "step": 11333 }, { "epoch": 0.28, "learning_rate": 1.6912406828766242e-05, "loss": 2.081, "step": 11334 }, { "epoch": 0.28, "learning_rate": 1.691182991302434e-05, "loss": 2.3038, "step": 11335 }, { "epoch": 0.28, "learning_rate": 1.6911252953231086e-05, "loss": 2.2759, "step": 11336 }, { "epoch": 0.28, "learning_rate": 1.6910675949390173e-05, "loss": 2.0584, "step": 11337 }, { "epoch": 0.28, "learning_rate": 1.6910098901505267e-05, "loss": 2.0643, "step": 11338 }, { "epoch": 0.28, "learning_rate": 1.6909521809580046e-05, "loss": 2.1276, "step": 11339 }, { "epoch": 0.28, "learning_rate": 1.690894467361819e-05, "loss": 2.2004, "step": 11340 }, { "epoch": 0.28, "learning_rate": 1.6908367493623387e-05, "loss": 1.9729, "step": 11341 }, { "epoch": 0.28, "learning_rate": 1.69077902695993e-05, "loss": 2.1898, "step": 11342 }, { "epoch": 0.28, "learning_rate": 1.6907213001549615e-05, "loss": 2.0373, "step": 11343 }, { "epoch": 0.28, "learning_rate": 1.690663568947801e-05, "loss": 2.1842, "step": 11344 }, { "epoch": 0.28, "learning_rate": 1.6906058333388166e-05, "loss": 2.308, "step": 11345 }, { "epoch": 0.28, "learning_rate": 1.690548093328376e-05, "loss": 2.0591, "step": 11346 }, { "epoch": 0.28, "learning_rate": 1.6904903489168476e-05, "loss": 2.102, "step": 11347 }, { "epoch": 0.28, "learning_rate": 1.690432600104599e-05, "loss": 2.0903, "step": 11348 }, { "epoch": 0.28, "learning_rate": 1.6903748468919984e-05, "loss": 2.1453, "step": 11349 }, { "epoch": 0.28, "learning_rate": 1.690317089279414e-05, "loss": 2.2163, "step": 11350 }, { "epoch": 0.28, "learning_rate": 1.6902593272672137e-05, "loss": 2.3266, "step": 11351 }, { "epoch": 0.28, "learning_rate": 1.690201560855766e-05, "loss": 2.0952, "step": 11352 }, { "epoch": 0.28, "learning_rate": 1.6901437900454385e-05, "loss": 2.2047, "step": 11353 }, { "epoch": 0.28, "learning_rate": 1.6900860148366e-05, "loss": 2.0963, "step": 11354 }, { "epoch": 0.28, "learning_rate": 1.6900282352296184e-05, "loss": 2.0484, "step": 11355 }, { "epoch": 0.28, "learning_rate": 1.6899704512248617e-05, "loss": 2.2794, "step": 11356 }, { "epoch": 0.28, "learning_rate": 1.6899126628226986e-05, "loss": 2.2647, "step": 11357 }, { "epoch": 0.28, "learning_rate": 1.6898548700234977e-05, "loss": 1.9755, "step": 11358 }, { "epoch": 0.28, "learning_rate": 1.6897970728276263e-05, "loss": 2.1758, "step": 11359 }, { "epoch": 0.28, "learning_rate": 1.6897392712354536e-05, "loss": 2.0522, "step": 11360 }, { "epoch": 0.28, "learning_rate": 1.689681465247348e-05, "loss": 2.2215, "step": 11361 }, { "epoch": 0.28, "learning_rate": 1.6896236548636773e-05, "loss": 2.2279, "step": 11362 }, { "epoch": 0.28, "learning_rate": 1.6895658400848107e-05, "loss": 1.9151, "step": 11363 }, { "epoch": 0.28, "learning_rate": 1.6895080209111156e-05, "loss": 2.3021, "step": 11364 }, { "epoch": 0.28, "learning_rate": 1.6894501973429614e-05, "loss": 2.1725, "step": 11365 }, { "epoch": 0.28, "learning_rate": 1.6893923693807168e-05, "loss": 2.0124, "step": 11366 }, { "epoch": 0.28, "learning_rate": 1.68933453702475e-05, "loss": 2.1291, "step": 11367 }, { "epoch": 0.28, "learning_rate": 1.689276700275429e-05, "loss": 2.0732, "step": 11368 }, { "epoch": 0.28, "learning_rate": 1.6892188591331232e-05, "loss": 2.2508, "step": 11369 }, { "epoch": 0.28, "learning_rate": 1.6891610135982012e-05, "loss": 2.2428, "step": 11370 }, { "epoch": 0.28, "learning_rate": 1.6891031636710312e-05, "loss": 1.9959, "step": 11371 }, { "epoch": 0.28, "learning_rate": 1.6890453093519825e-05, "loss": 2.3489, "step": 11372 }, { "epoch": 0.28, "learning_rate": 1.688987450641423e-05, "loss": 2.1436, "step": 11373 }, { "epoch": 0.28, "learning_rate": 1.6889295875397222e-05, "loss": 2.2138, "step": 11374 }, { "epoch": 0.28, "learning_rate": 1.6888717200472488e-05, "loss": 2.0291, "step": 11375 }, { "epoch": 0.28, "learning_rate": 1.6888138481643716e-05, "loss": 2.4262, "step": 11376 }, { "epoch": 0.28, "learning_rate": 1.6887559718914584e-05, "loss": 2.2508, "step": 11377 }, { "epoch": 0.28, "learning_rate": 1.6886980912288797e-05, "loss": 2.1885, "step": 11378 }, { "epoch": 0.28, "learning_rate": 1.6886402061770034e-05, "loss": 2.0333, "step": 11379 }, { "epoch": 0.28, "learning_rate": 1.6885823167361984e-05, "loss": 2.0749, "step": 11380 }, { "epoch": 0.28, "learning_rate": 1.6885244229068344e-05, "loss": 2.0493, "step": 11381 }, { "epoch": 0.28, "learning_rate": 1.6884665246892796e-05, "loss": 2.1047, "step": 11382 }, { "epoch": 0.28, "learning_rate": 1.688408622083903e-05, "loss": 2.1841, "step": 11383 }, { "epoch": 0.28, "learning_rate": 1.6883507150910742e-05, "loss": 1.9266, "step": 11384 }, { "epoch": 0.28, "learning_rate": 1.6882928037111624e-05, "loss": 2.1269, "step": 11385 }, { "epoch": 0.28, "learning_rate": 1.6882348879445355e-05, "loss": 2.1502, "step": 11386 }, { "epoch": 0.28, "learning_rate": 1.6881769677915637e-05, "loss": 2.1553, "step": 11387 }, { "epoch": 0.28, "learning_rate": 1.688119043252616e-05, "loss": 2.1807, "step": 11388 }, { "epoch": 0.28, "learning_rate": 1.688061114328061e-05, "loss": 2.1974, "step": 11389 }, { "epoch": 0.28, "learning_rate": 1.6880031810182683e-05, "loss": 2.0645, "step": 11390 }, { "epoch": 0.28, "learning_rate": 1.6879452433236077e-05, "loss": 2.2314, "step": 11391 }, { "epoch": 0.28, "learning_rate": 1.687887301244447e-05, "loss": 2.0701, "step": 11392 }, { "epoch": 0.28, "learning_rate": 1.6878293547811567e-05, "loss": 2.0179, "step": 11393 }, { "epoch": 0.28, "learning_rate": 1.687771403934106e-05, "loss": 2.0319, "step": 11394 }, { "epoch": 0.28, "learning_rate": 1.687713448703664e-05, "loss": 2.25, "step": 11395 }, { "epoch": 0.28, "learning_rate": 1.6876554890901998e-05, "loss": 2.2228, "step": 11396 }, { "epoch": 0.28, "learning_rate": 1.6875975250940833e-05, "loss": 2.1319, "step": 11397 }, { "epoch": 0.28, "learning_rate": 1.6875395567156837e-05, "loss": 2.1437, "step": 11398 }, { "epoch": 0.28, "learning_rate": 1.68748158395537e-05, "loss": 2.1412, "step": 11399 }, { "epoch": 0.28, "learning_rate": 1.6874236068135127e-05, "loss": 1.9745, "step": 11400 }, { "epoch": 0.28, "learning_rate": 1.6873656252904807e-05, "loss": 2.1129, "step": 11401 }, { "epoch": 0.28, "learning_rate": 1.687307639386643e-05, "loss": 2.0002, "step": 11402 }, { "epoch": 0.28, "learning_rate": 1.6872496491023702e-05, "loss": 1.9976, "step": 11403 }, { "epoch": 0.28, "learning_rate": 1.6871916544380314e-05, "loss": 2.0498, "step": 11404 }, { "epoch": 0.28, "learning_rate": 1.6871336553939958e-05, "loss": 2.3366, "step": 11405 }, { "epoch": 0.28, "learning_rate": 1.6870756519706342e-05, "loss": 2.2211, "step": 11406 }, { "epoch": 0.28, "learning_rate": 1.687017644168315e-05, "loss": 2.2584, "step": 11407 }, { "epoch": 0.28, "learning_rate": 1.6869596319874086e-05, "loss": 2.2208, "step": 11408 }, { "epoch": 0.28, "learning_rate": 1.6869016154282848e-05, "loss": 2.0424, "step": 11409 }, { "epoch": 0.28, "learning_rate": 1.686843594491313e-05, "loss": 2.1907, "step": 11410 }, { "epoch": 0.28, "learning_rate": 1.6867855691768632e-05, "loss": 2.1108, "step": 11411 }, { "epoch": 0.28, "learning_rate": 1.686727539485305e-05, "loss": 2.0949, "step": 11412 }, { "epoch": 0.28, "learning_rate": 1.686669505417008e-05, "loss": 2.0404, "step": 11413 }, { "epoch": 0.28, "learning_rate": 1.686611466972343e-05, "loss": 2.163, "step": 11414 }, { "epoch": 0.28, "learning_rate": 1.6865534241516797e-05, "loss": 2.1164, "step": 11415 }, { "epoch": 0.28, "learning_rate": 1.686495376955387e-05, "loss": 2.1601, "step": 11416 }, { "epoch": 0.28, "learning_rate": 1.6864373253838356e-05, "loss": 2.2828, "step": 11417 }, { "epoch": 0.28, "learning_rate": 1.686379269437396e-05, "loss": 2.1136, "step": 11418 }, { "epoch": 0.28, "learning_rate": 1.686321209116437e-05, "loss": 2.085, "step": 11419 }, { "epoch": 0.28, "learning_rate": 1.68626314442133e-05, "loss": 2.2485, "step": 11420 }, { "epoch": 0.28, "learning_rate": 1.6862050753524436e-05, "loss": 2.1701, "step": 11421 }, { "epoch": 0.28, "learning_rate": 1.6861470019101486e-05, "loss": 2.1776, "step": 11422 }, { "epoch": 0.28, "learning_rate": 1.6860889240948157e-05, "loss": 2.2003, "step": 11423 }, { "epoch": 0.28, "learning_rate": 1.686030841906814e-05, "loss": 2.1155, "step": 11424 }, { "epoch": 0.28, "learning_rate": 1.6859727553465146e-05, "loss": 2.0159, "step": 11425 }, { "epoch": 0.28, "learning_rate": 1.6859146644142873e-05, "loss": 2.0594, "step": 11426 }, { "epoch": 0.28, "learning_rate": 1.685856569110502e-05, "loss": 2.272, "step": 11427 }, { "epoch": 0.28, "learning_rate": 1.6857984694355292e-05, "loss": 2.1163, "step": 11428 }, { "epoch": 0.28, "learning_rate": 1.6857403653897394e-05, "loss": 1.941, "step": 11429 }, { "epoch": 0.28, "learning_rate": 1.685682256973503e-05, "loss": 2.1919, "step": 11430 }, { "epoch": 0.28, "learning_rate": 1.6856241441871896e-05, "loss": 2.1129, "step": 11431 }, { "epoch": 0.28, "learning_rate": 1.6855660270311706e-05, "loss": 2.2932, "step": 11432 }, { "epoch": 0.28, "learning_rate": 1.6855079055058158e-05, "loss": 2.0778, "step": 11433 }, { "epoch": 0.28, "learning_rate": 1.6854497796114955e-05, "loss": 2.1609, "step": 11434 }, { "epoch": 0.28, "learning_rate": 1.6853916493485805e-05, "loss": 2.0023, "step": 11435 }, { "epoch": 0.28, "learning_rate": 1.6853335147174413e-05, "loss": 2.126, "step": 11436 }, { "epoch": 0.28, "learning_rate": 1.685275375718448e-05, "loss": 2.1672, "step": 11437 }, { "epoch": 0.28, "learning_rate": 1.6852172323519716e-05, "loss": 2.1804, "step": 11438 }, { "epoch": 0.28, "learning_rate": 1.6851590846183823e-05, "loss": 2.0983, "step": 11439 }, { "epoch": 0.28, "learning_rate": 1.685100932518051e-05, "loss": 2.1163, "step": 11440 }, { "epoch": 0.28, "learning_rate": 1.6850427760513483e-05, "loss": 2.1614, "step": 11441 }, { "epoch": 0.28, "learning_rate": 1.6849846152186442e-05, "loss": 2.1805, "step": 11442 }, { "epoch": 0.28, "learning_rate": 1.6849264500203103e-05, "loss": 2.1779, "step": 11443 }, { "epoch": 0.28, "learning_rate": 1.6848682804567175e-05, "loss": 1.9486, "step": 11444 }, { "epoch": 0.28, "learning_rate": 1.684810106528235e-05, "loss": 2.1389, "step": 11445 }, { "epoch": 0.28, "learning_rate": 1.684751928235235e-05, "loss": 2.1968, "step": 11446 }, { "epoch": 0.28, "learning_rate": 1.6846937455780878e-05, "loss": 2.1602, "step": 11447 }, { "epoch": 0.28, "learning_rate": 1.684635558557164e-05, "loss": 2.136, "step": 11448 }, { "epoch": 0.28, "learning_rate": 1.6845773671728345e-05, "loss": 2.1123, "step": 11449 }, { "epoch": 0.28, "learning_rate": 1.684519171425471e-05, "loss": 2.1184, "step": 11450 }, { "epoch": 0.28, "learning_rate": 1.6844609713154432e-05, "loss": 2.1417, "step": 11451 }, { "epoch": 0.28, "learning_rate": 1.684402766843123e-05, "loss": 2.25, "step": 11452 }, { "epoch": 0.28, "learning_rate": 1.6843445580088805e-05, "loss": 1.9418, "step": 11453 }, { "epoch": 0.28, "learning_rate": 1.684286344813087e-05, "loss": 1.9676, "step": 11454 }, { "epoch": 0.28, "learning_rate": 1.6842281272561143e-05, "loss": 2.1554, "step": 11455 }, { "epoch": 0.28, "learning_rate": 1.6841699053383322e-05, "loss": 2.1353, "step": 11456 }, { "epoch": 0.28, "learning_rate": 1.6841116790601126e-05, "loss": 2.1129, "step": 11457 }, { "epoch": 0.28, "learning_rate": 1.6840534484218262e-05, "loss": 2.0992, "step": 11458 }, { "epoch": 0.28, "learning_rate": 1.6839952134238446e-05, "loss": 2.2048, "step": 11459 }, { "epoch": 0.28, "learning_rate": 1.6839369740665377e-05, "loss": 2.0821, "step": 11460 }, { "epoch": 0.28, "learning_rate": 1.6838787303502784e-05, "loss": 2.2094, "step": 11461 }, { "epoch": 0.28, "learning_rate": 1.683820482275437e-05, "loss": 2.0856, "step": 11462 }, { "epoch": 0.28, "learning_rate": 1.6837622298423846e-05, "loss": 2.093, "step": 11463 }, { "epoch": 0.28, "learning_rate": 1.6837039730514927e-05, "loss": 2.2264, "step": 11464 }, { "epoch": 0.28, "learning_rate": 1.6836457119031324e-05, "loss": 2.3544, "step": 11465 }, { "epoch": 0.28, "learning_rate": 1.6835874463976755e-05, "loss": 2.1918, "step": 11466 }, { "epoch": 0.28, "learning_rate": 1.683529176535493e-05, "loss": 2.0786, "step": 11467 }, { "epoch": 0.28, "learning_rate": 1.683470902316956e-05, "loss": 2.2742, "step": 11468 }, { "epoch": 0.28, "learning_rate": 1.6834126237424365e-05, "loss": 2.07, "step": 11469 }, { "epoch": 0.28, "learning_rate": 1.6833543408123058e-05, "loss": 2.143, "step": 11470 }, { "epoch": 0.28, "learning_rate": 1.6832960535269346e-05, "loss": 2.1415, "step": 11471 }, { "epoch": 0.28, "learning_rate": 1.6832377618866953e-05, "loss": 2.1451, "step": 11472 }, { "epoch": 0.28, "learning_rate": 1.6831794658919595e-05, "loss": 2.3429, "step": 11473 }, { "epoch": 0.28, "learning_rate": 1.6831211655430977e-05, "loss": 2.1512, "step": 11474 }, { "epoch": 0.28, "learning_rate": 1.683062860840482e-05, "loss": 2.2165, "step": 11475 }, { "epoch": 0.28, "learning_rate": 1.6830045517844844e-05, "loss": 2.1568, "step": 11476 }, { "epoch": 0.28, "learning_rate": 1.6829462383754763e-05, "loss": 1.9999, "step": 11477 }, { "epoch": 0.28, "learning_rate": 1.6828879206138292e-05, "loss": 2.2424, "step": 11478 }, { "epoch": 0.28, "learning_rate": 1.6828295984999147e-05, "loss": 1.9678, "step": 11479 }, { "epoch": 0.28, "learning_rate": 1.6827712720341048e-05, "loss": 2.2427, "step": 11480 }, { "epoch": 0.28, "learning_rate": 1.682712941216771e-05, "loss": 2.011, "step": 11481 }, { "epoch": 0.28, "learning_rate": 1.682654606048285e-05, "loss": 2.1197, "step": 11482 }, { "epoch": 0.28, "learning_rate": 1.682596266529019e-05, "loss": 2.1703, "step": 11483 }, { "epoch": 0.28, "learning_rate": 1.6825379226593442e-05, "loss": 2.1364, "step": 11484 }, { "epoch": 0.28, "learning_rate": 1.682479574439633e-05, "loss": 2.0209, "step": 11485 }, { "epoch": 0.28, "learning_rate": 1.6824212218702572e-05, "loss": 2.119, "step": 11486 }, { "epoch": 0.28, "learning_rate": 1.6823628649515885e-05, "loss": 2.1216, "step": 11487 }, { "epoch": 0.28, "learning_rate": 1.6823045036839988e-05, "loss": 2.2226, "step": 11488 }, { "epoch": 0.28, "learning_rate": 1.68224613806786e-05, "loss": 2.1138, "step": 11489 }, { "epoch": 0.28, "learning_rate": 1.6821877681035445e-05, "loss": 2.1236, "step": 11490 }, { "epoch": 0.28, "learning_rate": 1.6821293937914238e-05, "loss": 2.145, "step": 11491 }, { "epoch": 0.28, "learning_rate": 1.6820710151318703e-05, "loss": 2.2179, "step": 11492 }, { "epoch": 0.28, "learning_rate": 1.6820126321252556e-05, "loss": 2.2786, "step": 11493 }, { "epoch": 0.28, "learning_rate": 1.6819542447719524e-05, "loss": 2.0335, "step": 11494 }, { "epoch": 0.28, "learning_rate": 1.6818958530723328e-05, "loss": 2.19, "step": 11495 }, { "epoch": 0.28, "learning_rate": 1.6818374570267683e-05, "loss": 2.0147, "step": 11496 }, { "epoch": 0.28, "learning_rate": 1.681779056635632e-05, "loss": 2.0189, "step": 11497 }, { "epoch": 0.28, "learning_rate": 1.681720651899295e-05, "loss": 2.1818, "step": 11498 }, { "epoch": 0.28, "learning_rate": 1.6816622428181305e-05, "loss": 2.0213, "step": 11499 }, { "epoch": 0.28, "learning_rate": 1.6816038293925102e-05, "loss": 2.2825, "step": 11500 }, { "epoch": 0.28, "learning_rate": 1.6815454116228068e-05, "loss": 2.0537, "step": 11501 }, { "epoch": 0.28, "learning_rate": 1.6814869895093918e-05, "loss": 2.1229, "step": 11502 }, { "epoch": 0.28, "learning_rate": 1.6814285630526388e-05, "loss": 2.0661, "step": 11503 }, { "epoch": 0.28, "learning_rate": 1.681370132252919e-05, "loss": 2.1555, "step": 11504 }, { "epoch": 0.28, "learning_rate": 1.681311697110606e-05, "loss": 2.1408, "step": 11505 }, { "epoch": 0.28, "learning_rate": 1.6812532576260708e-05, "loss": 2.1953, "step": 11506 }, { "epoch": 0.28, "learning_rate": 1.681194813799687e-05, "loss": 2.2678, "step": 11507 }, { "epoch": 0.28, "learning_rate": 1.6811363656318263e-05, "loss": 1.8939, "step": 11508 }, { "epoch": 0.28, "learning_rate": 1.6810779131228617e-05, "loss": 2.1736, "step": 11509 }, { "epoch": 0.28, "learning_rate": 1.6810194562731655e-05, "loss": 2.2185, "step": 11510 }, { "epoch": 0.28, "learning_rate": 1.6809609950831106e-05, "loss": 2.0366, "step": 11511 }, { "epoch": 0.28, "learning_rate": 1.6809025295530693e-05, "loss": 2.2235, "step": 11512 }, { "epoch": 0.28, "learning_rate": 1.6808440596834142e-05, "loss": 2.1788, "step": 11513 }, { "epoch": 0.28, "learning_rate": 1.680785585474518e-05, "loss": 2.1949, "step": 11514 }, { "epoch": 0.28, "learning_rate": 1.680727106926754e-05, "loss": 2.0714, "step": 11515 }, { "epoch": 0.28, "learning_rate": 1.6806686240404933e-05, "loss": 2.2536, "step": 11516 }, { "epoch": 0.28, "learning_rate": 1.6806101368161103e-05, "loss": 2.1536, "step": 11517 }, { "epoch": 0.28, "learning_rate": 1.680551645253977e-05, "loss": 1.9991, "step": 11518 }, { "epoch": 0.28, "learning_rate": 1.6804931493544662e-05, "loss": 1.9565, "step": 11519 }, { "epoch": 0.28, "learning_rate": 1.6804346491179505e-05, "loss": 2.0604, "step": 11520 }, { "epoch": 0.28, "learning_rate": 1.6803761445448037e-05, "loss": 2.2037, "step": 11521 }, { "epoch": 0.28, "learning_rate": 1.6803176356353974e-05, "loss": 2.2001, "step": 11522 }, { "epoch": 0.28, "learning_rate": 1.6802591223901053e-05, "loss": 2.1665, "step": 11523 }, { "epoch": 0.28, "learning_rate": 1.6802006048093002e-05, "loss": 2.0999, "step": 11524 }, { "epoch": 0.28, "learning_rate": 1.680142082893355e-05, "loss": 2.1418, "step": 11525 }, { "epoch": 0.28, "learning_rate": 1.6800835566426425e-05, "loss": 1.9646, "step": 11526 }, { "epoch": 0.28, "learning_rate": 1.680025026057536e-05, "loss": 2.0619, "step": 11527 }, { "epoch": 0.28, "learning_rate": 1.6799664911384083e-05, "loss": 2.1886, "step": 11528 }, { "epoch": 0.28, "learning_rate": 1.6799079518856326e-05, "loss": 2.1232, "step": 11529 }, { "epoch": 0.28, "learning_rate": 1.679849408299582e-05, "loss": 2.2049, "step": 11530 }, { "epoch": 0.28, "learning_rate": 1.6797908603806297e-05, "loss": 2.0141, "step": 11531 }, { "epoch": 0.28, "learning_rate": 1.6797323081291482e-05, "loss": 2.255, "step": 11532 }, { "epoch": 0.28, "learning_rate": 1.6796737515455116e-05, "loss": 2.3031, "step": 11533 }, { "epoch": 0.28, "learning_rate": 1.6796151906300923e-05, "loss": 2.2123, "step": 11534 }, { "epoch": 0.28, "learning_rate": 1.6795566253832646e-05, "loss": 2.1618, "step": 11535 }, { "epoch": 0.28, "learning_rate": 1.6794980558054006e-05, "loss": 2.2252, "step": 11536 }, { "epoch": 0.28, "learning_rate": 1.679439481896874e-05, "loss": 2.1449, "step": 11537 }, { "epoch": 0.28, "learning_rate": 1.679380903658058e-05, "loss": 2.0742, "step": 11538 }, { "epoch": 0.28, "learning_rate": 1.6793223210893267e-05, "loss": 2.1012, "step": 11539 }, { "epoch": 0.28, "learning_rate": 1.679263734191052e-05, "loss": 2.1955, "step": 11540 }, { "epoch": 0.28, "learning_rate": 1.679205142963609e-05, "loss": 2.1148, "step": 11541 }, { "epoch": 0.28, "learning_rate": 1.67914654740737e-05, "loss": 2.0906, "step": 11542 }, { "epoch": 0.28, "learning_rate": 1.6790879475227088e-05, "loss": 2.2993, "step": 11543 }, { "epoch": 0.28, "learning_rate": 1.6790293433099986e-05, "loss": 2.0855, "step": 11544 }, { "epoch": 0.28, "learning_rate": 1.6789707347696134e-05, "loss": 2.1665, "step": 11545 }, { "epoch": 0.28, "learning_rate": 1.678912121901926e-05, "loss": 2.2729, "step": 11546 }, { "epoch": 0.28, "learning_rate": 1.6788535047073103e-05, "loss": 2.1346, "step": 11547 }, { "epoch": 0.28, "learning_rate": 1.678794883186141e-05, "loss": 2.1941, "step": 11548 }, { "epoch": 0.28, "learning_rate": 1.67873625733879e-05, "loss": 2.1964, "step": 11549 }, { "epoch": 0.28, "learning_rate": 1.678677627165631e-05, "loss": 2.1229, "step": 11550 }, { "epoch": 0.28, "learning_rate": 1.678618992667039e-05, "loss": 2.1632, "step": 11551 }, { "epoch": 0.28, "learning_rate": 1.678560353843387e-05, "loss": 2.1294, "step": 11552 }, { "epoch": 0.28, "learning_rate": 1.678501710695049e-05, "loss": 2.182, "step": 11553 }, { "epoch": 0.28, "learning_rate": 1.678443063222398e-05, "loss": 2.0557, "step": 11554 }, { "epoch": 0.28, "learning_rate": 1.6783844114258084e-05, "loss": 2.2814, "step": 11555 }, { "epoch": 0.28, "learning_rate": 1.6783257553056537e-05, "loss": 2.0338, "step": 11556 }, { "epoch": 0.28, "learning_rate": 1.6782670948623083e-05, "loss": 2.0126, "step": 11557 }, { "epoch": 0.28, "learning_rate": 1.678208430096145e-05, "loss": 2.2824, "step": 11558 }, { "epoch": 0.28, "learning_rate": 1.6781497610075392e-05, "loss": 2.1308, "step": 11559 }, { "epoch": 0.28, "learning_rate": 1.6780910875968634e-05, "loss": 2.2232, "step": 11560 }, { "epoch": 0.28, "learning_rate": 1.6780324098644923e-05, "loss": 2.0781, "step": 11561 }, { "epoch": 0.28, "learning_rate": 1.6779737278107997e-05, "loss": 2.3856, "step": 11562 }, { "epoch": 0.29, "learning_rate": 1.6779150414361596e-05, "loss": 2.2079, "step": 11563 }, { "epoch": 0.29, "learning_rate": 1.677856350740946e-05, "loss": 2.2142, "step": 11564 }, { "epoch": 0.29, "learning_rate": 1.677797655725533e-05, "loss": 2.2467, "step": 11565 }, { "epoch": 0.29, "learning_rate": 1.6777389563902945e-05, "loss": 2.1204, "step": 11566 }, { "epoch": 0.29, "learning_rate": 1.677680252735605e-05, "loss": 1.9924, "step": 11567 }, { "epoch": 0.29, "learning_rate": 1.677621544761838e-05, "loss": 1.9364, "step": 11568 }, { "epoch": 0.29, "learning_rate": 1.6775628324693684e-05, "loss": 1.9928, "step": 11569 }, { "epoch": 0.29, "learning_rate": 1.67750411585857e-05, "loss": 2.0129, "step": 11570 }, { "epoch": 0.29, "learning_rate": 1.6774453949298174e-05, "loss": 2.1746, "step": 11571 }, { "epoch": 0.29, "learning_rate": 1.6773866696834838e-05, "loss": 2.1977, "step": 11572 }, { "epoch": 0.29, "learning_rate": 1.6773279401199446e-05, "loss": 2.1453, "step": 11573 }, { "epoch": 0.29, "learning_rate": 1.6772692062395738e-05, "loss": 2.1784, "step": 11574 }, { "epoch": 0.29, "learning_rate": 1.6772104680427458e-05, "loss": 2.1514, "step": 11575 }, { "epoch": 0.29, "learning_rate": 1.6771517255298343e-05, "loss": 2.1851, "step": 11576 }, { "epoch": 0.29, "learning_rate": 1.6770929787012143e-05, "loss": 2.1824, "step": 11577 }, { "epoch": 0.29, "learning_rate": 1.67703422755726e-05, "loss": 2.3096, "step": 11578 }, { "epoch": 0.29, "learning_rate": 1.6769754720983465e-05, "loss": 2.1526, "step": 11579 }, { "epoch": 0.29, "learning_rate": 1.6769167123248473e-05, "loss": 2.1678, "step": 11580 }, { "epoch": 0.29, "learning_rate": 1.676857948237137e-05, "loss": 1.9853, "step": 11581 }, { "epoch": 0.29, "learning_rate": 1.6767991798355907e-05, "loss": 2.1098, "step": 11582 }, { "epoch": 0.29, "learning_rate": 1.6767404071205827e-05, "loss": 2.0556, "step": 11583 }, { "epoch": 0.29, "learning_rate": 1.6766816300924874e-05, "loss": 2.1827, "step": 11584 }, { "epoch": 0.29, "learning_rate": 1.6766228487516798e-05, "loss": 2.0749, "step": 11585 }, { "epoch": 0.29, "learning_rate": 1.676564063098534e-05, "loss": 2.0451, "step": 11586 }, { "epoch": 0.29, "learning_rate": 1.6765052731334252e-05, "loss": 2.1696, "step": 11587 }, { "epoch": 0.29, "learning_rate": 1.6764464788567277e-05, "loss": 2.2002, "step": 11588 }, { "epoch": 0.29, "learning_rate": 1.6763876802688165e-05, "loss": 2.1621, "step": 11589 }, { "epoch": 0.29, "learning_rate": 1.676328877370066e-05, "loss": 2.0085, "step": 11590 }, { "epoch": 0.29, "learning_rate": 1.676270070160851e-05, "loss": 2.1616, "step": 11591 }, { "epoch": 0.29, "learning_rate": 1.6762112586415468e-05, "loss": 2.139, "step": 11592 }, { "epoch": 0.29, "learning_rate": 1.676152442812528e-05, "loss": 2.1038, "step": 11593 }, { "epoch": 0.29, "learning_rate": 1.6760936226741686e-05, "loss": 2.0158, "step": 11594 }, { "epoch": 0.29, "learning_rate": 1.6760347982268447e-05, "loss": 2.2133, "step": 11595 }, { "epoch": 0.29, "learning_rate": 1.6759759694709304e-05, "loss": 2.1636, "step": 11596 }, { "epoch": 0.29, "learning_rate": 1.6759171364068016e-05, "loss": 2.0795, "step": 11597 }, { "epoch": 0.29, "learning_rate": 1.6758582990348323e-05, "loss": 1.9918, "step": 11598 }, { "epoch": 0.29, "learning_rate": 1.6757994573553976e-05, "loss": 2.1417, "step": 11599 }, { "epoch": 0.29, "learning_rate": 1.675740611368873e-05, "loss": 2.1937, "step": 11600 }, { "epoch": 0.29, "learning_rate": 1.675681761075633e-05, "loss": 2.2565, "step": 11601 }, { "epoch": 0.29, "learning_rate": 1.675622906476053e-05, "loss": 2.18, "step": 11602 }, { "epoch": 0.29, "learning_rate": 1.6755640475705083e-05, "loss": 2.1521, "step": 11603 }, { "epoch": 0.29, "learning_rate": 1.6755051843593733e-05, "loss": 2.1005, "step": 11604 }, { "epoch": 0.29, "learning_rate": 1.675446316843024e-05, "loss": 2.1383, "step": 11605 }, { "epoch": 0.29, "learning_rate": 1.675387445021835e-05, "loss": 2.0008, "step": 11606 }, { "epoch": 0.29, "learning_rate": 1.6753285688961824e-05, "loss": 1.997, "step": 11607 }, { "epoch": 0.29, "learning_rate": 1.6752696884664397e-05, "loss": 2.1718, "step": 11608 }, { "epoch": 0.29, "learning_rate": 1.6752108037329837e-05, "loss": 2.0591, "step": 11609 }, { "epoch": 0.29, "learning_rate": 1.6751519146961893e-05, "loss": 2.2976, "step": 11610 }, { "epoch": 0.29, "learning_rate": 1.6750930213564316e-05, "loss": 2.3195, "step": 11611 }, { "epoch": 0.29, "learning_rate": 1.675034123714086e-05, "loss": 2.0359, "step": 11612 }, { "epoch": 0.29, "learning_rate": 1.674975221769528e-05, "loss": 2.208, "step": 11613 }, { "epoch": 0.29, "learning_rate": 1.674916315523133e-05, "loss": 1.916, "step": 11614 }, { "epoch": 0.29, "learning_rate": 1.6748574049752764e-05, "loss": 2.1458, "step": 11615 }, { "epoch": 0.29, "learning_rate": 1.6747984901263333e-05, "loss": 2.2751, "step": 11616 }, { "epoch": 0.29, "learning_rate": 1.6747395709766797e-05, "loss": 2.3404, "step": 11617 }, { "epoch": 0.29, "learning_rate": 1.6746806475266914e-05, "loss": 2.2459, "step": 11618 }, { "epoch": 0.29, "learning_rate": 1.674621719776743e-05, "loss": 2.1297, "step": 11619 }, { "epoch": 0.29, "learning_rate": 1.6745627877272103e-05, "loss": 2.2218, "step": 11620 }, { "epoch": 0.29, "learning_rate": 1.6745038513784694e-05, "loss": 2.1258, "step": 11621 }, { "epoch": 0.29, "learning_rate": 1.6744449107308957e-05, "loss": 2.1237, "step": 11622 }, { "epoch": 0.29, "learning_rate": 1.6743859657848645e-05, "loss": 2.2081, "step": 11623 }, { "epoch": 0.29, "learning_rate": 1.6743270165407524e-05, "loss": 2.099, "step": 11624 }, { "epoch": 0.29, "learning_rate": 1.674268062998934e-05, "loss": 2.0228, "step": 11625 }, { "epoch": 0.29, "learning_rate": 1.6742091051597855e-05, "loss": 2.1969, "step": 11626 }, { "epoch": 0.29, "learning_rate": 1.6741501430236825e-05, "loss": 2.0525, "step": 11627 }, { "epoch": 0.29, "learning_rate": 1.6740911765910014e-05, "loss": 2.1529, "step": 11628 }, { "epoch": 0.29, "learning_rate": 1.674032205862117e-05, "loss": 2.2208, "step": 11629 }, { "epoch": 0.29, "learning_rate": 1.673973230837406e-05, "loss": 2.0728, "step": 11630 }, { "epoch": 0.29, "learning_rate": 1.673914251517244e-05, "loss": 2.174, "step": 11631 }, { "epoch": 0.29, "learning_rate": 1.673855267902007e-05, "loss": 2.2764, "step": 11632 }, { "epoch": 0.29, "learning_rate": 1.6737962799920704e-05, "loss": 2.1464, "step": 11633 }, { "epoch": 0.29, "learning_rate": 1.6737372877878106e-05, "loss": 2.1169, "step": 11634 }, { "epoch": 0.29, "learning_rate": 1.6736782912896037e-05, "loss": 2.1553, "step": 11635 }, { "epoch": 0.29, "learning_rate": 1.6736192904978253e-05, "loss": 2.1197, "step": 11636 }, { "epoch": 0.29, "learning_rate": 1.673560285412852e-05, "loss": 2.1207, "step": 11637 }, { "epoch": 0.29, "learning_rate": 1.673501276035059e-05, "loss": 2.2298, "step": 11638 }, { "epoch": 0.29, "learning_rate": 1.6734422623648228e-05, "loss": 2.2647, "step": 11639 }, { "epoch": 0.29, "learning_rate": 1.6733832444025204e-05, "loss": 2.0862, "step": 11640 }, { "epoch": 0.29, "learning_rate": 1.6733242221485264e-05, "loss": 1.9844, "step": 11641 }, { "epoch": 0.29, "learning_rate": 1.673265195603218e-05, "loss": 2.2312, "step": 11642 }, { "epoch": 0.29, "learning_rate": 1.673206164766971e-05, "loss": 2.1931, "step": 11643 }, { "epoch": 0.29, "learning_rate": 1.6731471296401617e-05, "loss": 2.0863, "step": 11644 }, { "epoch": 0.29, "learning_rate": 1.6730880902231667e-05, "loss": 1.9886, "step": 11645 }, { "epoch": 0.29, "learning_rate": 1.6730290465163617e-05, "loss": 2.2031, "step": 11646 }, { "epoch": 0.29, "learning_rate": 1.672969998520123e-05, "loss": 2.1267, "step": 11647 }, { "epoch": 0.29, "learning_rate": 1.6729109462348276e-05, "loss": 2.0822, "step": 11648 }, { "epoch": 0.29, "learning_rate": 1.672851889660851e-05, "loss": 2.2262, "step": 11649 }, { "epoch": 0.29, "learning_rate": 1.6727928287985703e-05, "loss": 2.1673, "step": 11650 }, { "epoch": 0.29, "learning_rate": 1.672733763648362e-05, "loss": 2.0963, "step": 11651 }, { "epoch": 0.29, "learning_rate": 1.6726746942106015e-05, "loss": 2.0531, "step": 11652 }, { "epoch": 0.29, "learning_rate": 1.6726156204856665e-05, "loss": 2.1796, "step": 11653 }, { "epoch": 0.29, "learning_rate": 1.6725565424739324e-05, "loss": 2.2029, "step": 11654 }, { "epoch": 0.29, "learning_rate": 1.6724974601757768e-05, "loss": 2.1043, "step": 11655 }, { "epoch": 0.29, "learning_rate": 1.6724383735915756e-05, "loss": 2.1483, "step": 11656 }, { "epoch": 0.29, "learning_rate": 1.6723792827217054e-05, "loss": 2.182, "step": 11657 }, { "epoch": 0.29, "learning_rate": 1.6723201875665428e-05, "loss": 2.0437, "step": 11658 }, { "epoch": 0.29, "learning_rate": 1.6722610881264648e-05, "loss": 2.0414, "step": 11659 }, { "epoch": 0.29, "learning_rate": 1.6722019844018474e-05, "loss": 2.0368, "step": 11660 }, { "epoch": 0.29, "learning_rate": 1.6721428763930682e-05, "loss": 2.167, "step": 11661 }, { "epoch": 0.29, "learning_rate": 1.6720837641005033e-05, "loss": 2.2113, "step": 11662 }, { "epoch": 0.29, "learning_rate": 1.6720246475245293e-05, "loss": 2.1296, "step": 11663 }, { "epoch": 0.29, "learning_rate": 1.6719655266655234e-05, "loss": 2.0671, "step": 11664 }, { "epoch": 0.29, "learning_rate": 1.6719064015238623e-05, "loss": 2.1191, "step": 11665 }, { "epoch": 0.29, "learning_rate": 1.6718472720999225e-05, "loss": 2.185, "step": 11666 }, { "epoch": 0.29, "learning_rate": 1.671788138394081e-05, "loss": 2.2398, "step": 11667 }, { "epoch": 0.29, "learning_rate": 1.6717290004067147e-05, "loss": 2.1505, "step": 11668 }, { "epoch": 0.29, "learning_rate": 1.6716698581382012e-05, "loss": 2.2334, "step": 11669 }, { "epoch": 0.29, "learning_rate": 1.671610711588916e-05, "loss": 2.207, "step": 11670 }, { "epoch": 0.29, "learning_rate": 1.6715515607592377e-05, "loss": 2.3795, "step": 11671 }, { "epoch": 0.29, "learning_rate": 1.6714924056495418e-05, "loss": 2.0096, "step": 11672 }, { "epoch": 0.29, "learning_rate": 1.671433246260206e-05, "loss": 2.1534, "step": 11673 }, { "epoch": 0.29, "learning_rate": 1.6713740825916075e-05, "loss": 2.2398, "step": 11674 }, { "epoch": 0.29, "learning_rate": 1.6713149146441234e-05, "loss": 2.2103, "step": 11675 }, { "epoch": 0.29, "learning_rate": 1.6712557424181305e-05, "loss": 2.0148, "step": 11676 }, { "epoch": 0.29, "learning_rate": 1.6711965659140056e-05, "loss": 1.9314, "step": 11677 }, { "epoch": 0.29, "learning_rate": 1.6711373851321265e-05, "loss": 2.0812, "step": 11678 }, { "epoch": 0.29, "learning_rate": 1.67107820007287e-05, "loss": 2.048, "step": 11679 }, { "epoch": 0.29, "learning_rate": 1.6710190107366136e-05, "loss": 2.1609, "step": 11680 }, { "epoch": 0.29, "learning_rate": 1.6709598171237344e-05, "loss": 2.2149, "step": 11681 }, { "epoch": 0.29, "learning_rate": 1.6709006192346097e-05, "loss": 2.1022, "step": 11682 }, { "epoch": 0.29, "learning_rate": 1.6708414170696166e-05, "loss": 2.031, "step": 11683 }, { "epoch": 0.29, "learning_rate": 1.6707822106291325e-05, "loss": 2.1495, "step": 11684 }, { "epoch": 0.29, "learning_rate": 1.670722999913535e-05, "loss": 1.9654, "step": 11685 }, { "epoch": 0.29, "learning_rate": 1.6706637849232006e-05, "loss": 2.1272, "step": 11686 }, { "epoch": 0.29, "learning_rate": 1.6706045656585076e-05, "loss": 2.0189, "step": 11687 }, { "epoch": 0.29, "learning_rate": 1.6705453421198333e-05, "loss": 2.2151, "step": 11688 }, { "epoch": 0.29, "learning_rate": 1.670486114307555e-05, "loss": 2.4158, "step": 11689 }, { "epoch": 0.29, "learning_rate": 1.6704268822220504e-05, "loss": 2.2134, "step": 11690 }, { "epoch": 0.29, "learning_rate": 1.6703676458636965e-05, "loss": 2.2051, "step": 11691 }, { "epoch": 0.29, "learning_rate": 1.6703084052328712e-05, "loss": 2.1869, "step": 11692 }, { "epoch": 0.29, "learning_rate": 1.670249160329952e-05, "loss": 2.2599, "step": 11693 }, { "epoch": 0.29, "learning_rate": 1.6701899111553162e-05, "loss": 2.181, "step": 11694 }, { "epoch": 0.29, "learning_rate": 1.6701306577093422e-05, "loss": 2.1528, "step": 11695 }, { "epoch": 0.29, "learning_rate": 1.670071399992407e-05, "loss": 2.1578, "step": 11696 }, { "epoch": 0.29, "learning_rate": 1.670012138004888e-05, "loss": 2.146, "step": 11697 }, { "epoch": 0.29, "learning_rate": 1.669952871747164e-05, "loss": 2.1525, "step": 11698 }, { "epoch": 0.29, "learning_rate": 1.6698936012196114e-05, "loss": 2.0727, "step": 11699 }, { "epoch": 0.29, "learning_rate": 1.6698343264226087e-05, "loss": 1.9813, "step": 11700 }, { "epoch": 0.29, "learning_rate": 1.6697750473565338e-05, "loss": 2.1508, "step": 11701 }, { "epoch": 0.29, "learning_rate": 1.6697157640217638e-05, "loss": 2.1718, "step": 11702 }, { "epoch": 0.29, "learning_rate": 1.6696564764186772e-05, "loss": 2.0226, "step": 11703 }, { "epoch": 0.29, "learning_rate": 1.669597184547652e-05, "loss": 1.9295, "step": 11704 }, { "epoch": 0.29, "learning_rate": 1.669537888409065e-05, "loss": 2.3095, "step": 11705 }, { "epoch": 0.29, "learning_rate": 1.6694785880032955e-05, "loss": 2.0316, "step": 11706 }, { "epoch": 0.29, "learning_rate": 1.6694192833307204e-05, "loss": 2.2183, "step": 11707 }, { "epoch": 0.29, "learning_rate": 1.6693599743917182e-05, "loss": 1.9708, "step": 11708 }, { "epoch": 0.29, "learning_rate": 1.6693006611866668e-05, "loss": 2.2375, "step": 11709 }, { "epoch": 0.29, "learning_rate": 1.6692413437159442e-05, "loss": 2.0776, "step": 11710 }, { "epoch": 0.29, "learning_rate": 1.6691820219799283e-05, "loss": 2.1171, "step": 11711 }, { "epoch": 0.29, "learning_rate": 1.669122695978997e-05, "loss": 2.3093, "step": 11712 }, { "epoch": 0.29, "learning_rate": 1.6690633657135292e-05, "loss": 2.2665, "step": 11713 }, { "epoch": 0.29, "learning_rate": 1.6690040311839026e-05, "loss": 2.1182, "step": 11714 }, { "epoch": 0.29, "learning_rate": 1.6689446923904947e-05, "loss": 1.9865, "step": 11715 }, { "epoch": 0.29, "learning_rate": 1.6688853493336847e-05, "loss": 2.2477, "step": 11716 }, { "epoch": 0.29, "learning_rate": 1.66882600201385e-05, "loss": 2.1171, "step": 11717 }, { "epoch": 0.29, "learning_rate": 1.6687666504313696e-05, "loss": 2.0815, "step": 11718 }, { "epoch": 0.29, "learning_rate": 1.6687072945866213e-05, "loss": 2.3415, "step": 11719 }, { "epoch": 0.29, "learning_rate": 1.6686479344799838e-05, "loss": 2.123, "step": 11720 }, { "epoch": 0.29, "learning_rate": 1.6685885701118344e-05, "loss": 2.2686, "step": 11721 }, { "epoch": 0.29, "learning_rate": 1.6685292014825525e-05, "loss": 2.2675, "step": 11722 }, { "epoch": 0.29, "learning_rate": 1.6684698285925164e-05, "loss": 2.1076, "step": 11723 }, { "epoch": 0.29, "learning_rate": 1.668410451442104e-05, "loss": 2.026, "step": 11724 }, { "epoch": 0.29, "learning_rate": 1.668351070031694e-05, "loss": 2.1164, "step": 11725 }, { "epoch": 0.29, "learning_rate": 1.6682916843616646e-05, "loss": 2.2568, "step": 11726 }, { "epoch": 0.29, "learning_rate": 1.668232294432395e-05, "loss": 2.1801, "step": 11727 }, { "epoch": 0.29, "learning_rate": 1.6681729002442625e-05, "loss": 2.0265, "step": 11728 }, { "epoch": 0.29, "learning_rate": 1.668113501797647e-05, "loss": 2.1942, "step": 11729 }, { "epoch": 0.29, "learning_rate": 1.6680540990929263e-05, "loss": 2.1001, "step": 11730 }, { "epoch": 0.29, "learning_rate": 1.667994692130479e-05, "loss": 2.0535, "step": 11731 }, { "epoch": 0.29, "learning_rate": 1.6679352809106837e-05, "loss": 2.2153, "step": 11732 }, { "epoch": 0.29, "learning_rate": 1.6678758654339195e-05, "loss": 2.2068, "step": 11733 }, { "epoch": 0.29, "learning_rate": 1.6678164457005643e-05, "loss": 2.1631, "step": 11734 }, { "epoch": 0.29, "learning_rate": 1.6677570217109976e-05, "loss": 1.9984, "step": 11735 }, { "epoch": 0.29, "learning_rate": 1.6676975934655976e-05, "loss": 2.0085, "step": 11736 }, { "epoch": 0.29, "learning_rate": 1.6676381609647437e-05, "loss": 2.0804, "step": 11737 }, { "epoch": 0.29, "learning_rate": 1.6675787242088137e-05, "loss": 2.176, "step": 11738 }, { "epoch": 0.29, "learning_rate": 1.6675192831981874e-05, "loss": 2.0042, "step": 11739 }, { "epoch": 0.29, "learning_rate": 1.6674598379332425e-05, "loss": 2.2619, "step": 11740 }, { "epoch": 0.29, "learning_rate": 1.6674003884143587e-05, "loss": 2.1403, "step": 11741 }, { "epoch": 0.29, "learning_rate": 1.6673409346419155e-05, "loss": 2.1537, "step": 11742 }, { "epoch": 0.29, "learning_rate": 1.6672814766162902e-05, "loss": 2.1103, "step": 11743 }, { "epoch": 0.29, "learning_rate": 1.6672220143378626e-05, "loss": 2.1364, "step": 11744 }, { "epoch": 0.29, "learning_rate": 1.667162547807012e-05, "loss": 2.0951, "step": 11745 }, { "epoch": 0.29, "learning_rate": 1.6671030770241173e-05, "loss": 2.0639, "step": 11746 }, { "epoch": 0.29, "learning_rate": 1.667043601989557e-05, "loss": 2.1234, "step": 11747 }, { "epoch": 0.29, "learning_rate": 1.66698412270371e-05, "loss": 2.1703, "step": 11748 }, { "epoch": 0.29, "learning_rate": 1.666924639166957e-05, "loss": 2.1636, "step": 11749 }, { "epoch": 0.29, "learning_rate": 1.6668651513796746e-05, "loss": 2.1226, "step": 11750 }, { "epoch": 0.29, "learning_rate": 1.666805659342244e-05, "loss": 2.2774, "step": 11751 }, { "epoch": 0.29, "learning_rate": 1.6667461630550434e-05, "loss": 2.1039, "step": 11752 }, { "epoch": 0.29, "learning_rate": 1.6666866625184524e-05, "loss": 2.1557, "step": 11753 }, { "epoch": 0.29, "learning_rate": 1.66662715773285e-05, "loss": 2.2463, "step": 11754 }, { "epoch": 0.29, "learning_rate": 1.666567648698615e-05, "loss": 2.0819, "step": 11755 }, { "epoch": 0.29, "learning_rate": 1.6665081354161276e-05, "loss": 2.1514, "step": 11756 }, { "epoch": 0.29, "learning_rate": 1.6664486178857666e-05, "loss": 2.2227, "step": 11757 }, { "epoch": 0.29, "learning_rate": 1.6663890961079117e-05, "loss": 2.1713, "step": 11758 }, { "epoch": 0.29, "learning_rate": 1.6663295700829416e-05, "loss": 2.1291, "step": 11759 }, { "epoch": 0.29, "learning_rate": 1.666270039811236e-05, "loss": 2.2195, "step": 11760 }, { "epoch": 0.29, "learning_rate": 1.666210505293174e-05, "loss": 2.123, "step": 11761 }, { "epoch": 0.29, "learning_rate": 1.6661509665291357e-05, "loss": 2.3502, "step": 11762 }, { "epoch": 0.29, "learning_rate": 1.6660914235195e-05, "loss": 2.1477, "step": 11763 }, { "epoch": 0.29, "learning_rate": 1.666031876264647e-05, "loss": 2.3274, "step": 11764 }, { "epoch": 0.29, "learning_rate": 1.6659723247649552e-05, "loss": 2.0814, "step": 11765 }, { "epoch": 0.29, "learning_rate": 1.665912769020805e-05, "loss": 2.2515, "step": 11766 }, { "epoch": 0.29, "learning_rate": 1.6658532090325758e-05, "loss": 2.068, "step": 11767 }, { "epoch": 0.29, "learning_rate": 1.665793644800647e-05, "loss": 2.277, "step": 11768 }, { "epoch": 0.29, "learning_rate": 1.6657340763253986e-05, "loss": 2.2498, "step": 11769 }, { "epoch": 0.29, "learning_rate": 1.6656745036072096e-05, "loss": 2.2605, "step": 11770 }, { "epoch": 0.29, "learning_rate": 1.6656149266464604e-05, "loss": 2.087, "step": 11771 }, { "epoch": 0.29, "learning_rate": 1.66555534544353e-05, "loss": 2.1057, "step": 11772 }, { "epoch": 0.29, "learning_rate": 1.6654957599987987e-05, "loss": 2.2006, "step": 11773 }, { "epoch": 0.29, "learning_rate": 1.6654361703126462e-05, "loss": 2.1376, "step": 11774 }, { "epoch": 0.29, "learning_rate": 1.6653765763854516e-05, "loss": 2.0059, "step": 11775 }, { "epoch": 0.29, "learning_rate": 1.6653169782175957e-05, "loss": 2.2404, "step": 11776 }, { "epoch": 0.29, "learning_rate": 1.6652573758094576e-05, "loss": 2.0593, "step": 11777 }, { "epoch": 0.29, "learning_rate": 1.6651977691614175e-05, "loss": 2.2987, "step": 11778 }, { "epoch": 0.29, "learning_rate": 1.6651381582738555e-05, "loss": 1.9405, "step": 11779 }, { "epoch": 0.29, "learning_rate": 1.6650785431471508e-05, "loss": 2.3484, "step": 11780 }, { "epoch": 0.29, "learning_rate": 1.665018923781684e-05, "loss": 2.133, "step": 11781 }, { "epoch": 0.29, "learning_rate": 1.6649593001778354e-05, "loss": 2.24, "step": 11782 }, { "epoch": 0.29, "learning_rate": 1.6648996723359837e-05, "loss": 2.3152, "step": 11783 }, { "epoch": 0.29, "learning_rate": 1.66484004025651e-05, "loss": 2.112, "step": 11784 }, { "epoch": 0.29, "learning_rate": 1.6647804039397943e-05, "loss": 2.0173, "step": 11785 }, { "epoch": 0.29, "learning_rate": 1.6647207633862162e-05, "loss": 2.0265, "step": 11786 }, { "epoch": 0.29, "learning_rate": 1.6646611185961558e-05, "loss": 2.0617, "step": 11787 }, { "epoch": 0.29, "learning_rate": 1.6646014695699938e-05, "loss": 1.9303, "step": 11788 }, { "epoch": 0.29, "learning_rate": 1.66454181630811e-05, "loss": 1.9709, "step": 11789 }, { "epoch": 0.29, "learning_rate": 1.6644821588108847e-05, "loss": 2.2949, "step": 11790 }, { "epoch": 0.29, "learning_rate": 1.664422497078698e-05, "loss": 2.2451, "step": 11791 }, { "epoch": 0.29, "learning_rate": 1.6643628311119302e-05, "loss": 2.2465, "step": 11792 }, { "epoch": 0.29, "learning_rate": 1.6643031609109617e-05, "loss": 2.005, "step": 11793 }, { "epoch": 0.29, "learning_rate": 1.6642434864761724e-05, "loss": 2.0844, "step": 11794 }, { "epoch": 0.29, "learning_rate": 1.6641838078079427e-05, "loss": 2.1063, "step": 11795 }, { "epoch": 0.29, "learning_rate": 1.6641241249066536e-05, "loss": 2.1483, "step": 11796 }, { "epoch": 0.29, "learning_rate": 1.664064437772685e-05, "loss": 1.9785, "step": 11797 }, { "epoch": 0.29, "learning_rate": 1.6640047464064173e-05, "loss": 2.1305, "step": 11798 }, { "epoch": 0.29, "learning_rate": 1.6639450508082307e-05, "loss": 2.1058, "step": 11799 }, { "epoch": 0.29, "learning_rate": 1.663885350978506e-05, "loss": 2.2293, "step": 11800 }, { "epoch": 0.29, "learning_rate": 1.6638256469176238e-05, "loss": 2.0408, "step": 11801 }, { "epoch": 0.29, "learning_rate": 1.6637659386259644e-05, "loss": 2.0927, "step": 11802 }, { "epoch": 0.29, "learning_rate": 1.663706226103908e-05, "loss": 2.1271, "step": 11803 }, { "epoch": 0.29, "learning_rate": 1.663646509351836e-05, "loss": 2.2132, "step": 11804 }, { "epoch": 0.29, "learning_rate": 1.6635867883701282e-05, "loss": 2.2879, "step": 11805 }, { "epoch": 0.29, "learning_rate": 1.6635270631591656e-05, "loss": 2.0487, "step": 11806 }, { "epoch": 0.29, "learning_rate": 1.663467333719329e-05, "loss": 2.2339, "step": 11807 }, { "epoch": 0.29, "learning_rate": 1.6634076000509987e-05, "loss": 2.0975, "step": 11808 }, { "epoch": 0.29, "learning_rate": 1.6633478621545552e-05, "loss": 2.238, "step": 11809 }, { "epoch": 0.29, "learning_rate": 1.66328812003038e-05, "loss": 2.1741, "step": 11810 }, { "epoch": 0.29, "learning_rate": 1.6632283736788536e-05, "loss": 2.0895, "step": 11811 }, { "epoch": 0.29, "learning_rate": 1.6631686231003564e-05, "loss": 2.2117, "step": 11812 }, { "epoch": 0.29, "learning_rate": 1.6631088682952695e-05, "loss": 2.1672, "step": 11813 }, { "epoch": 0.29, "learning_rate": 1.6630491092639734e-05, "loss": 1.9993, "step": 11814 }, { "epoch": 0.29, "learning_rate": 1.66298934600685e-05, "loss": 2.152, "step": 11815 }, { "epoch": 0.29, "learning_rate": 1.6629295785242788e-05, "loss": 2.098, "step": 11816 }, { "epoch": 0.29, "learning_rate": 1.6628698068166416e-05, "loss": 2.052, "step": 11817 }, { "epoch": 0.29, "learning_rate": 1.6628100308843186e-05, "loss": 1.9112, "step": 11818 }, { "epoch": 0.29, "learning_rate": 1.6627502507276917e-05, "loss": 2.2296, "step": 11819 }, { "epoch": 0.29, "learning_rate": 1.6626904663471413e-05, "loss": 2.0762, "step": 11820 }, { "epoch": 0.29, "learning_rate": 1.6626306777430488e-05, "loss": 2.1782, "step": 11821 }, { "epoch": 0.29, "learning_rate": 1.662570884915795e-05, "loss": 2.0559, "step": 11822 }, { "epoch": 0.29, "learning_rate": 1.6625110878657612e-05, "loss": 2.0574, "step": 11823 }, { "epoch": 0.29, "learning_rate": 1.662451286593328e-05, "loss": 2.0796, "step": 11824 }, { "epoch": 0.29, "learning_rate": 1.662391481098877e-05, "loss": 2.2168, "step": 11825 }, { "epoch": 0.29, "learning_rate": 1.6623316713827896e-05, "loss": 2.2367, "step": 11826 }, { "epoch": 0.29, "learning_rate": 1.662271857445446e-05, "loss": 2.0147, "step": 11827 }, { "epoch": 0.29, "learning_rate": 1.662212039287228e-05, "loss": 2.2848, "step": 11828 }, { "epoch": 0.29, "learning_rate": 1.6621522169085175e-05, "loss": 2.1996, "step": 11829 }, { "epoch": 0.29, "learning_rate": 1.6620923903096945e-05, "loss": 2.1134, "step": 11830 }, { "epoch": 0.29, "learning_rate": 1.6620325594911414e-05, "loss": 2.0784, "step": 11831 }, { "epoch": 0.29, "learning_rate": 1.6619727244532388e-05, "loss": 2.0042, "step": 11832 }, { "epoch": 0.29, "learning_rate": 1.6619128851963683e-05, "loss": 2.1714, "step": 11833 }, { "epoch": 0.29, "learning_rate": 1.661853041720911e-05, "loss": 2.2075, "step": 11834 }, { "epoch": 0.29, "learning_rate": 1.661793194027249e-05, "loss": 2.1419, "step": 11835 }, { "epoch": 0.29, "learning_rate": 1.6617333421157634e-05, "loss": 2.1344, "step": 11836 }, { "epoch": 0.29, "learning_rate": 1.6616734859868352e-05, "loss": 2.1919, "step": 11837 }, { "epoch": 0.29, "learning_rate": 1.6616136256408464e-05, "loss": 2.1465, "step": 11838 }, { "epoch": 0.29, "learning_rate": 1.6615537610781785e-05, "loss": 2.1737, "step": 11839 }, { "epoch": 0.29, "learning_rate": 1.6614938922992124e-05, "loss": 2.1506, "step": 11840 }, { "epoch": 0.29, "learning_rate": 1.6614340193043305e-05, "loss": 2.1834, "step": 11841 }, { "epoch": 0.29, "learning_rate": 1.661374142093914e-05, "loss": 2.1705, "step": 11842 }, { "epoch": 0.29, "learning_rate": 1.6613142606683445e-05, "loss": 2.215, "step": 11843 }, { "epoch": 0.29, "learning_rate": 1.6612543750280038e-05, "loss": 2.0979, "step": 11844 }, { "epoch": 0.29, "learning_rate": 1.6611944851732733e-05, "loss": 2.1089, "step": 11845 }, { "epoch": 0.29, "learning_rate": 1.6611345911045353e-05, "loss": 2.1153, "step": 11846 }, { "epoch": 0.29, "learning_rate": 1.6610746928221705e-05, "loss": 2.2247, "step": 11847 }, { "epoch": 0.29, "learning_rate": 1.6610147903265615e-05, "loss": 2.0875, "step": 11848 }, { "epoch": 0.29, "learning_rate": 1.66095488361809e-05, "loss": 2.1783, "step": 11849 }, { "epoch": 0.29, "learning_rate": 1.6608949726971374e-05, "loss": 2.1868, "step": 11850 }, { "epoch": 0.29, "learning_rate": 1.6608350575640862e-05, "loss": 1.9428, "step": 11851 }, { "epoch": 0.29, "learning_rate": 1.660775138219317e-05, "loss": 2.1017, "step": 11852 }, { "epoch": 0.29, "learning_rate": 1.660715214663213e-05, "loss": 2.0378, "step": 11853 }, { "epoch": 0.29, "learning_rate": 1.6606552868961556e-05, "loss": 2.0612, "step": 11854 }, { "epoch": 0.29, "learning_rate": 1.660595354918527e-05, "loss": 2.1453, "step": 11855 }, { "epoch": 0.29, "learning_rate": 1.6605354187307085e-05, "loss": 2.1474, "step": 11856 }, { "epoch": 0.29, "learning_rate": 1.6604754783330826e-05, "loss": 1.9127, "step": 11857 }, { "epoch": 0.29, "learning_rate": 1.6604155337260316e-05, "loss": 2.2189, "step": 11858 }, { "epoch": 0.29, "learning_rate": 1.6603555849099367e-05, "loss": 2.1591, "step": 11859 }, { "epoch": 0.29, "learning_rate": 1.6602956318851807e-05, "loss": 2.3453, "step": 11860 }, { "epoch": 0.29, "learning_rate": 1.6602356746521455e-05, "loss": 2.0204, "step": 11861 }, { "epoch": 0.29, "learning_rate": 1.660175713211213e-05, "loss": 2.0329, "step": 11862 }, { "epoch": 0.29, "learning_rate": 1.6601157475627654e-05, "loss": 1.8976, "step": 11863 }, { "epoch": 0.29, "learning_rate": 1.660055777707185e-05, "loss": 2.2537, "step": 11864 }, { "epoch": 0.29, "learning_rate": 1.6599958036448548e-05, "loss": 2.1187, "step": 11865 }, { "epoch": 0.29, "learning_rate": 1.6599358253761555e-05, "loss": 2.1436, "step": 11866 }, { "epoch": 0.29, "learning_rate": 1.65987584290147e-05, "loss": 2.1564, "step": 11867 }, { "epoch": 0.29, "learning_rate": 1.659815856221181e-05, "loss": 2.1799, "step": 11868 }, { "epoch": 0.29, "learning_rate": 1.659755865335671e-05, "loss": 2.0172, "step": 11869 }, { "epoch": 0.29, "learning_rate": 1.6596958702453214e-05, "loss": 2.0956, "step": 11870 }, { "epoch": 0.29, "learning_rate": 1.659635870950515e-05, "loss": 2.2504, "step": 11871 }, { "epoch": 0.29, "learning_rate": 1.659575867451634e-05, "loss": 2.2559, "step": 11872 }, { "epoch": 0.29, "learning_rate": 1.6595158597490612e-05, "loss": 2.1102, "step": 11873 }, { "epoch": 0.29, "learning_rate": 1.6594558478431792e-05, "loss": 2.1714, "step": 11874 }, { "epoch": 0.29, "learning_rate": 1.6593958317343698e-05, "loss": 2.0165, "step": 11875 }, { "epoch": 0.29, "learning_rate": 1.659335811423016e-05, "loss": 2.3324, "step": 11876 }, { "epoch": 0.29, "learning_rate": 1.6592757869095005e-05, "loss": 2.1755, "step": 11877 }, { "epoch": 0.29, "learning_rate": 1.6592157581942056e-05, "loss": 2.0722, "step": 11878 }, { "epoch": 0.29, "learning_rate": 1.6591557252775134e-05, "loss": 2.0506, "step": 11879 }, { "epoch": 0.29, "learning_rate": 1.6590956881598072e-05, "loss": 2.0818, "step": 11880 }, { "epoch": 0.29, "learning_rate": 1.659035646841469e-05, "loss": 2.1448, "step": 11881 }, { "epoch": 0.29, "learning_rate": 1.6589756013228823e-05, "loss": 2.2021, "step": 11882 }, { "epoch": 0.29, "learning_rate": 1.658915551604429e-05, "loss": 2.1957, "step": 11883 }, { "epoch": 0.29, "learning_rate": 1.6588554976864924e-05, "loss": 2.2248, "step": 11884 }, { "epoch": 0.29, "learning_rate": 1.658795439569455e-05, "loss": 2.103, "step": 11885 }, { "epoch": 0.29, "learning_rate": 1.6587353772536994e-05, "loss": 2.0409, "step": 11886 }, { "epoch": 0.29, "learning_rate": 1.6586753107396087e-05, "loss": 2.039, "step": 11887 }, { "epoch": 0.29, "learning_rate": 1.6586152400275657e-05, "loss": 2.0349, "step": 11888 }, { "epoch": 0.29, "learning_rate": 1.658555165117953e-05, "loss": 2.029, "step": 11889 }, { "epoch": 0.29, "learning_rate": 1.6584950860111534e-05, "loss": 1.9244, "step": 11890 }, { "epoch": 0.29, "learning_rate": 1.6584350027075503e-05, "loss": 2.2147, "step": 11891 }, { "epoch": 0.29, "learning_rate": 1.658374915207526e-05, "loss": 2.283, "step": 11892 }, { "epoch": 0.29, "learning_rate": 1.6583148235114644e-05, "loss": 2.1962, "step": 11893 }, { "epoch": 0.29, "learning_rate": 1.6582547276197476e-05, "loss": 2.045, "step": 11894 }, { "epoch": 0.29, "learning_rate": 1.6581946275327586e-05, "loss": 2.2423, "step": 11895 }, { "epoch": 0.29, "learning_rate": 1.6581345232508812e-05, "loss": 2.1154, "step": 11896 }, { "epoch": 0.29, "learning_rate": 1.6580744147744978e-05, "loss": 2.0586, "step": 11897 }, { "epoch": 0.29, "learning_rate": 1.6580143021039917e-05, "loss": 2.0539, "step": 11898 }, { "epoch": 0.29, "learning_rate": 1.657954185239746e-05, "loss": 2.3204, "step": 11899 }, { "epoch": 0.29, "learning_rate": 1.657894064182144e-05, "loss": 2.0431, "step": 11900 }, { "epoch": 0.29, "learning_rate": 1.6578339389315686e-05, "loss": 1.9322, "step": 11901 }, { "epoch": 0.29, "learning_rate": 1.657773809488403e-05, "loss": 2.2039, "step": 11902 }, { "epoch": 0.29, "learning_rate": 1.657713675853031e-05, "loss": 2.1056, "step": 11903 }, { "epoch": 0.29, "learning_rate": 1.657653538025835e-05, "loss": 2.1388, "step": 11904 }, { "epoch": 0.29, "learning_rate": 1.6575933960071986e-05, "loss": 2.2992, "step": 11905 }, { "epoch": 0.29, "learning_rate": 1.6575332497975055e-05, "loss": 2.1319, "step": 11906 }, { "epoch": 0.29, "learning_rate": 1.6574730993971382e-05, "loss": 2.0707, "step": 11907 }, { "epoch": 0.29, "learning_rate": 1.657412944806481e-05, "loss": 2.1604, "step": 11908 }, { "epoch": 0.29, "learning_rate": 1.6573527860259167e-05, "loss": 2.2737, "step": 11909 }, { "epoch": 0.29, "learning_rate": 1.6572926230558292e-05, "loss": 2.3714, "step": 11910 }, { "epoch": 0.29, "learning_rate": 1.6572324558966013e-05, "loss": 2.1005, "step": 11911 }, { "epoch": 0.29, "learning_rate": 1.657172284548617e-05, "loss": 2.2024, "step": 11912 }, { "epoch": 0.29, "learning_rate": 1.6571121090122593e-05, "loss": 2.1172, "step": 11913 }, { "epoch": 0.29, "learning_rate": 1.6570519292879123e-05, "loss": 2.0246, "step": 11914 }, { "epoch": 0.29, "learning_rate": 1.6569917453759587e-05, "loss": 2.2767, "step": 11915 }, { "epoch": 0.29, "learning_rate": 1.656931557276783e-05, "loss": 2.1778, "step": 11916 }, { "epoch": 0.29, "learning_rate": 1.6568713649907684e-05, "loss": 2.1098, "step": 11917 }, { "epoch": 0.29, "learning_rate": 1.6568111685182984e-05, "loss": 2.3069, "step": 11918 }, { "epoch": 0.29, "learning_rate": 1.6567509678597567e-05, "loss": 2.127, "step": 11919 }, { "epoch": 0.29, "learning_rate": 1.6566907630155272e-05, "loss": 2.324, "step": 11920 }, { "epoch": 0.29, "learning_rate": 1.6566305539859933e-05, "loss": 2.1346, "step": 11921 }, { "epoch": 0.29, "learning_rate": 1.6565703407715394e-05, "loss": 2.1216, "step": 11922 }, { "epoch": 0.29, "learning_rate": 1.6565101233725482e-05, "loss": 2.0828, "step": 11923 }, { "epoch": 0.29, "learning_rate": 1.6564499017894043e-05, "loss": 2.1123, "step": 11924 }, { "epoch": 0.29, "learning_rate": 1.656389676022491e-05, "loss": 2.0538, "step": 11925 }, { "epoch": 0.29, "learning_rate": 1.6563294460721926e-05, "loss": 2.2557, "step": 11926 }, { "epoch": 0.29, "learning_rate": 1.656269211938893e-05, "loss": 2.086, "step": 11927 }, { "epoch": 0.29, "learning_rate": 1.656208973622975e-05, "loss": 2.1719, "step": 11928 }, { "epoch": 0.29, "learning_rate": 1.6561487311248238e-05, "loss": 2.2051, "step": 11929 }, { "epoch": 0.29, "learning_rate": 1.6560884844448232e-05, "loss": 2.042, "step": 11930 }, { "epoch": 0.29, "learning_rate": 1.6560282335833566e-05, "loss": 2.1627, "step": 11931 }, { "epoch": 0.29, "learning_rate": 1.6559679785408084e-05, "loss": 2.2349, "step": 11932 }, { "epoch": 0.29, "learning_rate": 1.6559077193175624e-05, "loss": 2.1879, "step": 11933 }, { "epoch": 0.29, "learning_rate": 1.6558474559140028e-05, "loss": 2.1043, "step": 11934 }, { "epoch": 0.29, "learning_rate": 1.6557871883305134e-05, "loss": 2.1306, "step": 11935 }, { "epoch": 0.29, "learning_rate": 1.655726916567479e-05, "loss": 2.0132, "step": 11936 }, { "epoch": 0.29, "learning_rate": 1.6556666406252828e-05, "loss": 2.0803, "step": 11937 }, { "epoch": 0.29, "learning_rate": 1.65560636050431e-05, "loss": 2.1541, "step": 11938 }, { "epoch": 0.29, "learning_rate": 1.6555460762049435e-05, "loss": 2.1569, "step": 11939 }, { "epoch": 0.29, "learning_rate": 1.6554857877275685e-05, "loss": 2.2369, "step": 11940 }, { "epoch": 0.29, "learning_rate": 1.655425495072569e-05, "loss": 2.12, "step": 11941 }, { "epoch": 0.29, "learning_rate": 1.6553651982403293e-05, "loss": 2.1869, "step": 11942 }, { "epoch": 0.29, "learning_rate": 1.6553048972312336e-05, "loss": 2.0508, "step": 11943 }, { "epoch": 0.29, "learning_rate": 1.655244592045666e-05, "loss": 2.042, "step": 11944 }, { "epoch": 0.29, "learning_rate": 1.6551842826840113e-05, "loss": 2.2081, "step": 11945 }, { "epoch": 0.29, "learning_rate": 1.6551239691466536e-05, "loss": 2.039, "step": 11946 }, { "epoch": 0.29, "learning_rate": 1.6550636514339777e-05, "loss": 2.0311, "step": 11947 }, { "epoch": 0.29, "learning_rate": 1.655003329546367e-05, "loss": 1.9732, "step": 11948 }, { "epoch": 0.29, "learning_rate": 1.654943003484207e-05, "loss": 2.1278, "step": 11949 }, { "epoch": 0.29, "learning_rate": 1.6548826732478817e-05, "loss": 2.0692, "step": 11950 }, { "epoch": 0.29, "learning_rate": 1.6548223388377757e-05, "loss": 2.0944, "step": 11951 }, { "epoch": 0.29, "learning_rate": 1.6547620002542737e-05, "loss": 2.147, "step": 11952 }, { "epoch": 0.29, "learning_rate": 1.6547016574977603e-05, "loss": 2.2522, "step": 11953 }, { "epoch": 0.29, "learning_rate": 1.654641310568619e-05, "loss": 2.2208, "step": 11954 }, { "epoch": 0.29, "learning_rate": 1.654580959467236e-05, "loss": 2.0972, "step": 11955 }, { "epoch": 0.29, "learning_rate": 1.654520604193995e-05, "loss": 2.2741, "step": 11956 }, { "epoch": 0.29, "learning_rate": 1.654460244749281e-05, "loss": 2.1398, "step": 11957 }, { "epoch": 0.29, "learning_rate": 1.6543998811334785e-05, "loss": 2.0755, "step": 11958 }, { "epoch": 0.29, "learning_rate": 1.6543395133469724e-05, "loss": 2.1048, "step": 11959 }, { "epoch": 0.29, "learning_rate": 1.654279141390147e-05, "loss": 1.8466, "step": 11960 }, { "epoch": 0.29, "learning_rate": 1.6542187652633878e-05, "loss": 2.1746, "step": 11961 }, { "epoch": 0.29, "learning_rate": 1.6541583849670792e-05, "loss": 2.1794, "step": 11962 }, { "epoch": 0.29, "learning_rate": 1.654098000501606e-05, "loss": 2.0248, "step": 11963 }, { "epoch": 0.29, "learning_rate": 1.6540376118673527e-05, "loss": 2.0864, "step": 11964 }, { "epoch": 0.29, "learning_rate": 1.653977219064705e-05, "loss": 2.0578, "step": 11965 }, { "epoch": 0.29, "learning_rate": 1.6539168220940472e-05, "loss": 2.2876, "step": 11966 }, { "epoch": 0.29, "learning_rate": 1.6538564209557645e-05, "loss": 2.1958, "step": 11967 }, { "epoch": 0.29, "learning_rate": 1.6537960156502416e-05, "loss": 2.2922, "step": 11968 }, { "epoch": 0.3, "learning_rate": 1.6537356061778634e-05, "loss": 2.0433, "step": 11969 }, { "epoch": 0.3, "learning_rate": 1.6536751925390152e-05, "loss": 2.1475, "step": 11970 }, { "epoch": 0.3, "learning_rate": 1.6536147747340826e-05, "loss": 2.1176, "step": 11971 }, { "epoch": 0.3, "learning_rate": 1.6535543527634494e-05, "loss": 2.113, "step": 11972 }, { "epoch": 0.3, "learning_rate": 1.6534939266275015e-05, "loss": 2.1538, "step": 11973 }, { "epoch": 0.3, "learning_rate": 1.653433496326624e-05, "loss": 2.1264, "step": 11974 }, { "epoch": 0.3, "learning_rate": 1.6533730618612017e-05, "loss": 2.2578, "step": 11975 }, { "epoch": 0.3, "learning_rate": 1.6533126232316198e-05, "loss": 2.2016, "step": 11976 }, { "epoch": 0.3, "learning_rate": 1.6532521804382638e-05, "loss": 2.1456, "step": 11977 }, { "epoch": 0.3, "learning_rate": 1.653191733481519e-05, "loss": 2.1164, "step": 11978 }, { "epoch": 0.3, "learning_rate": 1.65313128236177e-05, "loss": 2.3279, "step": 11979 }, { "epoch": 0.3, "learning_rate": 1.653070827079403e-05, "loss": 2.1649, "step": 11980 }, { "epoch": 0.3, "learning_rate": 1.6530103676348024e-05, "loss": 2.0644, "step": 11981 }, { "epoch": 0.3, "learning_rate": 1.6529499040283537e-05, "loss": 2.28, "step": 11982 }, { "epoch": 0.3, "learning_rate": 1.6528894362604427e-05, "loss": 2.1412, "step": 11983 }, { "epoch": 0.3, "learning_rate": 1.652828964331455e-05, "loss": 2.1693, "step": 11984 }, { "epoch": 0.3, "learning_rate": 1.652768488241775e-05, "loss": 2.1142, "step": 11985 }, { "epoch": 0.3, "learning_rate": 1.652708007991789e-05, "loss": 2.3111, "step": 11986 }, { "epoch": 0.3, "learning_rate": 1.652647523581882e-05, "loss": 2.1232, "step": 11987 }, { "epoch": 0.3, "learning_rate": 1.6525870350124394e-05, "loss": 2.0309, "step": 11988 }, { "epoch": 0.3, "learning_rate": 1.652526542283847e-05, "loss": 2.2389, "step": 11989 }, { "epoch": 0.3, "learning_rate": 1.6524660453964908e-05, "loss": 1.9756, "step": 11990 }, { "epoch": 0.3, "learning_rate": 1.6524055443507556e-05, "loss": 2.0217, "step": 11991 }, { "epoch": 0.3, "learning_rate": 1.652345039147027e-05, "loss": 2.3686, "step": 11992 }, { "epoch": 0.3, "learning_rate": 1.6522845297856913e-05, "loss": 2.1507, "step": 11993 }, { "epoch": 0.3, "learning_rate": 1.6522240162671334e-05, "loss": 2.2223, "step": 11994 }, { "epoch": 0.3, "learning_rate": 1.6521634985917393e-05, "loss": 2.1821, "step": 11995 }, { "epoch": 0.3, "learning_rate": 1.6521029767598952e-05, "loss": 2.2749, "step": 11996 }, { "epoch": 0.3, "learning_rate": 1.6520424507719857e-05, "loss": 2.137, "step": 11997 }, { "epoch": 0.3, "learning_rate": 1.6519819206283974e-05, "loss": 2.3472, "step": 11998 }, { "epoch": 0.3, "learning_rate": 1.6519213863295158e-05, "loss": 2.094, "step": 11999 }, { "epoch": 0.3, "learning_rate": 1.651860847875727e-05, "loss": 2.0822, "step": 12000 }, { "epoch": 0.3, "learning_rate": 1.6518003052674162e-05, "loss": 2.2344, "step": 12001 }, { "epoch": 0.3, "learning_rate": 1.6517397585049696e-05, "loss": 2.2939, "step": 12002 }, { "epoch": 0.3, "learning_rate": 1.6516792075887734e-05, "loss": 2.0627, "step": 12003 }, { "epoch": 0.3, "learning_rate": 1.651618652519213e-05, "loss": 2.1181, "step": 12004 }, { "epoch": 0.3, "learning_rate": 1.651558093296675e-05, "loss": 2.1169, "step": 12005 }, { "epoch": 0.3, "learning_rate": 1.6514975299215444e-05, "loss": 2.1734, "step": 12006 }, { "epoch": 0.3, "learning_rate": 1.651436962394208e-05, "loss": 2.1749, "step": 12007 }, { "epoch": 0.3, "learning_rate": 1.6513763907150514e-05, "loss": 2.0507, "step": 12008 }, { "epoch": 0.3, "learning_rate": 1.6513158148844612e-05, "loss": 2.0693, "step": 12009 }, { "epoch": 0.3, "learning_rate": 1.6512552349028227e-05, "loss": 2.1209, "step": 12010 }, { "epoch": 0.3, "learning_rate": 1.6511946507705226e-05, "loss": 2.1047, "step": 12011 }, { "epoch": 0.3, "learning_rate": 1.6511340624879463e-05, "loss": 2.0516, "step": 12012 }, { "epoch": 0.3, "learning_rate": 1.651073470055481e-05, "loss": 1.993, "step": 12013 }, { "epoch": 0.3, "learning_rate": 1.651012873473512e-05, "loss": 2.1642, "step": 12014 }, { "epoch": 0.3, "learning_rate": 1.650952272742426e-05, "loss": 2.1351, "step": 12015 }, { "epoch": 0.3, "learning_rate": 1.6508916678626085e-05, "loss": 2.2267, "step": 12016 }, { "epoch": 0.3, "learning_rate": 1.6508310588344464e-05, "loss": 2.1918, "step": 12017 }, { "epoch": 0.3, "learning_rate": 1.6507704456583263e-05, "loss": 1.9562, "step": 12018 }, { "epoch": 0.3, "learning_rate": 1.650709828334634e-05, "loss": 2.1951, "step": 12019 }, { "epoch": 0.3, "learning_rate": 1.6506492068637556e-05, "loss": 2.1054, "step": 12020 }, { "epoch": 0.3, "learning_rate": 1.650588581246078e-05, "loss": 2.1022, "step": 12021 }, { "epoch": 0.3, "learning_rate": 1.650527951481987e-05, "loss": 2.0892, "step": 12022 }, { "epoch": 0.3, "learning_rate": 1.6504673175718697e-05, "loss": 2.0932, "step": 12023 }, { "epoch": 0.3, "learning_rate": 1.650406679516112e-05, "loss": 2.0384, "step": 12024 }, { "epoch": 0.3, "learning_rate": 1.6503460373151006e-05, "loss": 2.1137, "step": 12025 }, { "epoch": 0.3, "learning_rate": 1.6502853909692222e-05, "loss": 2.2117, "step": 12026 }, { "epoch": 0.3, "learning_rate": 1.650224740478863e-05, "loss": 2.1238, "step": 12027 }, { "epoch": 0.3, "learning_rate": 1.6501640858444098e-05, "loss": 2.0746, "step": 12028 }, { "epoch": 0.3, "learning_rate": 1.6501034270662487e-05, "loss": 2.2588, "step": 12029 }, { "epoch": 0.3, "learning_rate": 1.6500427641447666e-05, "loss": 2.1257, "step": 12030 }, { "epoch": 0.3, "learning_rate": 1.64998209708035e-05, "loss": 2.1911, "step": 12031 }, { "epoch": 0.3, "learning_rate": 1.649921425873386e-05, "loss": 1.925, "step": 12032 }, { "epoch": 0.3, "learning_rate": 1.6498607505242608e-05, "loss": 2.2431, "step": 12033 }, { "epoch": 0.3, "learning_rate": 1.6498000710333614e-05, "loss": 2.0767, "step": 12034 }, { "epoch": 0.3, "learning_rate": 1.649739387401074e-05, "loss": 2.1852, "step": 12035 }, { "epoch": 0.3, "learning_rate": 1.649678699627786e-05, "loss": 2.0496, "step": 12036 }, { "epoch": 0.3, "learning_rate": 1.6496180077138838e-05, "loss": 2.0616, "step": 12037 }, { "epoch": 0.3, "learning_rate": 1.6495573116597546e-05, "loss": 2.1654, "step": 12038 }, { "epoch": 0.3, "learning_rate": 1.649496611465785e-05, "loss": 2.2122, "step": 12039 }, { "epoch": 0.3, "learning_rate": 1.6494359071323613e-05, "loss": 2.2273, "step": 12040 }, { "epoch": 0.3, "learning_rate": 1.6493751986598716e-05, "loss": 2.072, "step": 12041 }, { "epoch": 0.3, "learning_rate": 1.6493144860487015e-05, "loss": 2.046, "step": 12042 }, { "epoch": 0.3, "learning_rate": 1.649253769299239e-05, "loss": 2.0721, "step": 12043 }, { "epoch": 0.3, "learning_rate": 1.6491930484118705e-05, "loss": 2.0083, "step": 12044 }, { "epoch": 0.3, "learning_rate": 1.649132323386983e-05, "loss": 2.1196, "step": 12045 }, { "epoch": 0.3, "learning_rate": 1.6490715942249638e-05, "loss": 2.1818, "step": 12046 }, { "epoch": 0.3, "learning_rate": 1.6490108609262e-05, "loss": 2.1928, "step": 12047 }, { "epoch": 0.3, "learning_rate": 1.648950123491078e-05, "loss": 2.0554, "step": 12048 }, { "epoch": 0.3, "learning_rate": 1.648889381919986e-05, "loss": 2.1199, "step": 12049 }, { "epoch": 0.3, "learning_rate": 1.6488286362133098e-05, "loss": 2.0732, "step": 12050 }, { "epoch": 0.3, "learning_rate": 1.6487678863714378e-05, "loss": 2.1867, "step": 12051 }, { "epoch": 0.3, "learning_rate": 1.6487071323947565e-05, "loss": 2.1031, "step": 12052 }, { "epoch": 0.3, "learning_rate": 1.6486463742836532e-05, "loss": 2.0249, "step": 12053 }, { "epoch": 0.3, "learning_rate": 1.648585612038515e-05, "loss": 2.2127, "step": 12054 }, { "epoch": 0.3, "learning_rate": 1.6485248456597295e-05, "loss": 2.1948, "step": 12055 }, { "epoch": 0.3, "learning_rate": 1.648464075147684e-05, "loss": 2.1297, "step": 12056 }, { "epoch": 0.3, "learning_rate": 1.6484033005027652e-05, "loss": 2.251, "step": 12057 }, { "epoch": 0.3, "learning_rate": 1.648342521725361e-05, "loss": 2.1876, "step": 12058 }, { "epoch": 0.3, "learning_rate": 1.6482817388158588e-05, "loss": 2.0504, "step": 12059 }, { "epoch": 0.3, "learning_rate": 1.6482209517746457e-05, "loss": 2.143, "step": 12060 }, { "epoch": 0.3, "learning_rate": 1.648160160602109e-05, "loss": 2.1849, "step": 12061 }, { "epoch": 0.3, "learning_rate": 1.6480993652986367e-05, "loss": 2.1287, "step": 12062 }, { "epoch": 0.3, "learning_rate": 1.648038565864616e-05, "loss": 2.0278, "step": 12063 }, { "epoch": 0.3, "learning_rate": 1.6479777623004342e-05, "loss": 2.2531, "step": 12064 }, { "epoch": 0.3, "learning_rate": 1.647916954606479e-05, "loss": 2.1279, "step": 12065 }, { "epoch": 0.3, "learning_rate": 1.6478561427831373e-05, "loss": 2.1773, "step": 12066 }, { "epoch": 0.3, "learning_rate": 1.647795326830798e-05, "loss": 2.0987, "step": 12067 }, { "epoch": 0.3, "learning_rate": 1.647734506749848e-05, "loss": 2.2177, "step": 12068 }, { "epoch": 0.3, "learning_rate": 1.6476736825406745e-05, "loss": 2.1532, "step": 12069 }, { "epoch": 0.3, "learning_rate": 1.6476128542036657e-05, "loss": 2.0095, "step": 12070 }, { "epoch": 0.3, "learning_rate": 1.647552021739209e-05, "loss": 2.1621, "step": 12071 }, { "epoch": 0.3, "learning_rate": 1.6474911851476923e-05, "loss": 2.2006, "step": 12072 }, { "epoch": 0.3, "learning_rate": 1.6474303444295036e-05, "loss": 2.1499, "step": 12073 }, { "epoch": 0.3, "learning_rate": 1.64736949958503e-05, "loss": 2.1904, "step": 12074 }, { "epoch": 0.3, "learning_rate": 1.6473086506146592e-05, "loss": 2.2489, "step": 12075 }, { "epoch": 0.3, "learning_rate": 1.64724779751878e-05, "loss": 2.231, "step": 12076 }, { "epoch": 0.3, "learning_rate": 1.6471869402977797e-05, "loss": 2.1194, "step": 12077 }, { "epoch": 0.3, "learning_rate": 1.647126078952046e-05, "loss": 2.1075, "step": 12078 }, { "epoch": 0.3, "learning_rate": 1.647065213481967e-05, "loss": 2.1166, "step": 12079 }, { "epoch": 0.3, "learning_rate": 1.64700434388793e-05, "loss": 2.0195, "step": 12080 }, { "epoch": 0.3, "learning_rate": 1.646943470170324e-05, "loss": 2.1308, "step": 12081 }, { "epoch": 0.3, "learning_rate": 1.6468825923295365e-05, "loss": 2.033, "step": 12082 }, { "epoch": 0.3, "learning_rate": 1.6468217103659552e-05, "loss": 2.2622, "step": 12083 }, { "epoch": 0.3, "learning_rate": 1.646760824279968e-05, "loss": 2.1504, "step": 12084 }, { "epoch": 0.3, "learning_rate": 1.6466999340719637e-05, "loss": 1.9725, "step": 12085 }, { "epoch": 0.3, "learning_rate": 1.64663903974233e-05, "loss": 2.3325, "step": 12086 }, { "epoch": 0.3, "learning_rate": 1.646578141291455e-05, "loss": 2.3172, "step": 12087 }, { "epoch": 0.3, "learning_rate": 1.6465172387197266e-05, "loss": 2.208, "step": 12088 }, { "epoch": 0.3, "learning_rate": 1.646456332027533e-05, "loss": 2.3761, "step": 12089 }, { "epoch": 0.3, "learning_rate": 1.646395421215263e-05, "loss": 2.1465, "step": 12090 }, { "epoch": 0.3, "learning_rate": 1.6463345062833038e-05, "loss": 2.2049, "step": 12091 }, { "epoch": 0.3, "learning_rate": 1.6462735872320445e-05, "loss": 2.1579, "step": 12092 }, { "epoch": 0.3, "learning_rate": 1.646212664061873e-05, "loss": 2.0045, "step": 12093 }, { "epoch": 0.3, "learning_rate": 1.6461517367731773e-05, "loss": 2.1359, "step": 12094 }, { "epoch": 0.3, "learning_rate": 1.6460908053663463e-05, "loss": 2.1086, "step": 12095 }, { "epoch": 0.3, "learning_rate": 1.646029869841768e-05, "loss": 2.2297, "step": 12096 }, { "epoch": 0.3, "learning_rate": 1.645968930199831e-05, "loss": 2.1674, "step": 12097 }, { "epoch": 0.3, "learning_rate": 1.645907986440923e-05, "loss": 2.0895, "step": 12098 }, { "epoch": 0.3, "learning_rate": 1.6458470385654334e-05, "loss": 2.0142, "step": 12099 }, { "epoch": 0.3, "learning_rate": 1.64578608657375e-05, "loss": 2.1683, "step": 12100 }, { "epoch": 0.3, "learning_rate": 1.645725130466261e-05, "loss": 2.144, "step": 12101 }, { "epoch": 0.3, "learning_rate": 1.6456641702433557e-05, "loss": 2.2349, "step": 12102 }, { "epoch": 0.3, "learning_rate": 1.645603205905422e-05, "loss": 2.0635, "step": 12103 }, { "epoch": 0.3, "learning_rate": 1.6455422374528488e-05, "loss": 2.1682, "step": 12104 }, { "epoch": 0.3, "learning_rate": 1.6454812648860248e-05, "loss": 2.316, "step": 12105 }, { "epoch": 0.3, "learning_rate": 1.645420288205338e-05, "loss": 2.2405, "step": 12106 }, { "epoch": 0.3, "learning_rate": 1.6453593074111777e-05, "loss": 2.2031, "step": 12107 }, { "epoch": 0.3, "learning_rate": 1.645298322503932e-05, "loss": 2.0576, "step": 12108 }, { "epoch": 0.3, "learning_rate": 1.6452373334839898e-05, "loss": 2.1102, "step": 12109 }, { "epoch": 0.3, "learning_rate": 1.64517634035174e-05, "loss": 2.2063, "step": 12110 }, { "epoch": 0.3, "learning_rate": 1.6451153431075708e-05, "loss": 2.1335, "step": 12111 }, { "epoch": 0.3, "learning_rate": 1.6450543417518713e-05, "loss": 2.0787, "step": 12112 }, { "epoch": 0.3, "learning_rate": 1.6449933362850307e-05, "loss": 2.0752, "step": 12113 }, { "epoch": 0.3, "learning_rate": 1.6449323267074365e-05, "loss": 2.2258, "step": 12114 }, { "epoch": 0.3, "learning_rate": 1.644871313019479e-05, "loss": 2.0262, "step": 12115 }, { "epoch": 0.3, "learning_rate": 1.6448102952215467e-05, "loss": 2.0249, "step": 12116 }, { "epoch": 0.3, "learning_rate": 1.644749273314028e-05, "loss": 2.1862, "step": 12117 }, { "epoch": 0.3, "learning_rate": 1.644688247297312e-05, "loss": 2.1735, "step": 12118 }, { "epoch": 0.3, "learning_rate": 1.6446272171717877e-05, "loss": 2.0363, "step": 12119 }, { "epoch": 0.3, "learning_rate": 1.644566182937844e-05, "loss": 2.2128, "step": 12120 }, { "epoch": 0.3, "learning_rate": 1.6445051445958702e-05, "loss": 1.9751, "step": 12121 }, { "epoch": 0.3, "learning_rate": 1.6444441021462548e-05, "loss": 2.1622, "step": 12122 }, { "epoch": 0.3, "learning_rate": 1.6443830555893875e-05, "loss": 2.0968, "step": 12123 }, { "epoch": 0.3, "learning_rate": 1.6443220049256565e-05, "loss": 2.1093, "step": 12124 }, { "epoch": 0.3, "learning_rate": 1.644260950155452e-05, "loss": 1.9985, "step": 12125 }, { "epoch": 0.3, "learning_rate": 1.6441998912791616e-05, "loss": 2.165, "step": 12126 }, { "epoch": 0.3, "learning_rate": 1.644138828297176e-05, "loss": 2.191, "step": 12127 }, { "epoch": 0.3, "learning_rate": 1.6440777612098836e-05, "loss": 2.1093, "step": 12128 }, { "epoch": 0.3, "learning_rate": 1.644016690017674e-05, "loss": 2.0104, "step": 12129 }, { "epoch": 0.3, "learning_rate": 1.6439556147209356e-05, "loss": 2.2244, "step": 12130 }, { "epoch": 0.3, "learning_rate": 1.6438945353200585e-05, "loss": 1.8931, "step": 12131 }, { "epoch": 0.3, "learning_rate": 1.643833451815432e-05, "loss": 2.0181, "step": 12132 }, { "epoch": 0.3, "learning_rate": 1.6437723642074447e-05, "loss": 2.1106, "step": 12133 }, { "epoch": 0.3, "learning_rate": 1.6437112724964863e-05, "loss": 2.0632, "step": 12134 }, { "epoch": 0.3, "learning_rate": 1.643650176682946e-05, "loss": 2.1344, "step": 12135 }, { "epoch": 0.3, "learning_rate": 1.643589076767214e-05, "loss": 2.0273, "step": 12136 }, { "epoch": 0.3, "learning_rate": 1.643527972749679e-05, "loss": 1.9734, "step": 12137 }, { "epoch": 0.3, "learning_rate": 1.64346686463073e-05, "loss": 2.1838, "step": 12138 }, { "epoch": 0.3, "learning_rate": 1.643405752410757e-05, "loss": 2.0617, "step": 12139 }, { "epoch": 0.3, "learning_rate": 1.6433446360901498e-05, "loss": 2.2879, "step": 12140 }, { "epoch": 0.3, "learning_rate": 1.6432835156692976e-05, "loss": 2.0911, "step": 12141 }, { "epoch": 0.3, "learning_rate": 1.6432223911485895e-05, "loss": 2.1686, "step": 12142 }, { "epoch": 0.3, "learning_rate": 1.643161262528416e-05, "loss": 2.2064, "step": 12143 }, { "epoch": 0.3, "learning_rate": 1.643100129809166e-05, "loss": 2.1419, "step": 12144 }, { "epoch": 0.3, "learning_rate": 1.6430389929912293e-05, "loss": 2.0884, "step": 12145 }, { "epoch": 0.3, "learning_rate": 1.6429778520749953e-05, "loss": 2.2118, "step": 12146 }, { "epoch": 0.3, "learning_rate": 1.642916707060854e-05, "loss": 2.1202, "step": 12147 }, { "epoch": 0.3, "learning_rate": 1.642855557949195e-05, "loss": 2.1714, "step": 12148 }, { "epoch": 0.3, "learning_rate": 1.642794404740408e-05, "loss": 2.1102, "step": 12149 }, { "epoch": 0.3, "learning_rate": 1.642733247434883e-05, "loss": 1.9079, "step": 12150 }, { "epoch": 0.3, "learning_rate": 1.6426720860330095e-05, "loss": 2.1742, "step": 12151 }, { "epoch": 0.3, "learning_rate": 1.642610920535177e-05, "loss": 2.0752, "step": 12152 }, { "epoch": 0.3, "learning_rate": 1.642549750941776e-05, "loss": 2.1314, "step": 12153 }, { "epoch": 0.3, "learning_rate": 1.642488577253196e-05, "loss": 2.1368, "step": 12154 }, { "epoch": 0.3, "learning_rate": 1.642427399469827e-05, "loss": 2.1885, "step": 12155 }, { "epoch": 0.3, "learning_rate": 1.6423662175920587e-05, "loss": 2.2701, "step": 12156 }, { "epoch": 0.3, "learning_rate": 1.6423050316202812e-05, "loss": 2.0493, "step": 12157 }, { "epoch": 0.3, "learning_rate": 1.6422438415548842e-05, "loss": 2.1447, "step": 12158 }, { "epoch": 0.3, "learning_rate": 1.642182647396258e-05, "loss": 2.1044, "step": 12159 }, { "epoch": 0.3, "learning_rate": 1.642121449144793e-05, "loss": 2.1165, "step": 12160 }, { "epoch": 0.3, "learning_rate": 1.6420602468008782e-05, "loss": 2.0455, "step": 12161 }, { "epoch": 0.3, "learning_rate": 1.6419990403649043e-05, "loss": 2.3565, "step": 12162 }, { "epoch": 0.3, "learning_rate": 1.641937829837261e-05, "loss": 2.2097, "step": 12163 }, { "epoch": 0.3, "learning_rate": 1.6418766152183394e-05, "loss": 1.9583, "step": 12164 }, { "epoch": 0.3, "learning_rate": 1.6418153965085287e-05, "loss": 2.075, "step": 12165 }, { "epoch": 0.3, "learning_rate": 1.641754173708219e-05, "loss": 2.1778, "step": 12166 }, { "epoch": 0.3, "learning_rate": 1.641692946817801e-05, "loss": 2.1371, "step": 12167 }, { "epoch": 0.3, "learning_rate": 1.641631715837665e-05, "loss": 2.1254, "step": 12168 }, { "epoch": 0.3, "learning_rate": 1.6415704807682006e-05, "loss": 2.1726, "step": 12169 }, { "epoch": 0.3, "learning_rate": 1.6415092416097986e-05, "loss": 2.1173, "step": 12170 }, { "epoch": 0.3, "learning_rate": 1.641447998362849e-05, "loss": 2.1425, "step": 12171 }, { "epoch": 0.3, "eval_loss": 1.812556266784668, "eval_runtime": 94.6937, "eval_samples_per_second": 621.446, "eval_steps_per_second": 4.858, "step": 12171 }, { "epoch": 0.3, "learning_rate": 1.6413867510277427e-05, "loss": 2.213, "step": 12172 }, { "epoch": 0.3, "learning_rate": 1.641325499604869e-05, "loss": 2.0133, "step": 12173 }, { "epoch": 0.3, "learning_rate": 1.6412642440946193e-05, "loss": 2.3265, "step": 12174 }, { "epoch": 0.3, "learning_rate": 1.6412029844973835e-05, "loss": 2.2787, "step": 12175 }, { "epoch": 0.3, "learning_rate": 1.6411417208135518e-05, "loss": 2.0876, "step": 12176 }, { "epoch": 0.3, "learning_rate": 1.6410804530435155e-05, "loss": 2.1298, "step": 12177 }, { "epoch": 0.3, "learning_rate": 1.6410191811876644e-05, "loss": 2.0198, "step": 12178 }, { "epoch": 0.3, "learning_rate": 1.6409579052463888e-05, "loss": 2.0964, "step": 12179 }, { "epoch": 0.3, "learning_rate": 1.64089662522008e-05, "loss": 2.1315, "step": 12180 }, { "epoch": 0.3, "learning_rate": 1.6408353411091278e-05, "loss": 2.1128, "step": 12181 }, { "epoch": 0.3, "learning_rate": 1.6407740529139234e-05, "loss": 2.1414, "step": 12182 }, { "epoch": 0.3, "learning_rate": 1.640712760634857e-05, "loss": 2.2436, "step": 12183 }, { "epoch": 0.3, "learning_rate": 1.6406514642723197e-05, "loss": 2.1202, "step": 12184 }, { "epoch": 0.3, "learning_rate": 1.6405901638267018e-05, "loss": 2.0804, "step": 12185 }, { "epoch": 0.3, "learning_rate": 1.6405288592983935e-05, "loss": 2.0055, "step": 12186 }, { "epoch": 0.3, "learning_rate": 1.6404675506877862e-05, "loss": 2.0026, "step": 12187 }, { "epoch": 0.3, "learning_rate": 1.640406237995271e-05, "loss": 2.293, "step": 12188 }, { "epoch": 0.3, "learning_rate": 1.6403449212212375e-05, "loss": 2.1595, "step": 12189 }, { "epoch": 0.3, "learning_rate": 1.6402836003660775e-05, "loss": 2.1116, "step": 12190 }, { "epoch": 0.3, "learning_rate": 1.640222275430181e-05, "loss": 2.0318, "step": 12191 }, { "epoch": 0.3, "learning_rate": 1.6401609464139398e-05, "loss": 2.2153, "step": 12192 }, { "epoch": 0.3, "learning_rate": 1.6400996133177442e-05, "loss": 2.079, "step": 12193 }, { "epoch": 0.3, "learning_rate": 1.640038276141985e-05, "loss": 2.2048, "step": 12194 }, { "epoch": 0.3, "learning_rate": 1.6399769348870537e-05, "loss": 2.2183, "step": 12195 }, { "epoch": 0.3, "learning_rate": 1.6399155895533404e-05, "loss": 2.2754, "step": 12196 }, { "epoch": 0.3, "learning_rate": 1.6398542401412367e-05, "loss": 2.1323, "step": 12197 }, { "epoch": 0.3, "learning_rate": 1.639792886651133e-05, "loss": 1.9955, "step": 12198 }, { "epoch": 0.3, "learning_rate": 1.639731529083421e-05, "loss": 2.1676, "step": 12199 }, { "epoch": 0.3, "learning_rate": 1.6396701674384913e-05, "loss": 2.1825, "step": 12200 }, { "epoch": 0.3, "learning_rate": 1.6396088017167356e-05, "loss": 1.9752, "step": 12201 }, { "epoch": 0.3, "learning_rate": 1.6395474319185445e-05, "loss": 2.2322, "step": 12202 }, { "epoch": 0.3, "learning_rate": 1.6394860580443088e-05, "loss": 2.188, "step": 12203 }, { "epoch": 0.3, "learning_rate": 1.6394246800944206e-05, "loss": 1.9458, "step": 12204 }, { "epoch": 0.3, "learning_rate": 1.63936329806927e-05, "loss": 2.17, "step": 12205 }, { "epoch": 0.3, "learning_rate": 1.6393019119692487e-05, "loss": 1.9537, "step": 12206 }, { "epoch": 0.3, "learning_rate": 1.6392405217947486e-05, "loss": 2.2614, "step": 12207 }, { "epoch": 0.3, "learning_rate": 1.63917912754616e-05, "loss": 2.0222, "step": 12208 }, { "epoch": 0.3, "learning_rate": 1.6391177292238744e-05, "loss": 2.1243, "step": 12209 }, { "epoch": 0.3, "learning_rate": 1.639056326828283e-05, "loss": 2.0952, "step": 12210 }, { "epoch": 0.3, "learning_rate": 1.6389949203597777e-05, "loss": 2.0975, "step": 12211 }, { "epoch": 0.3, "learning_rate": 1.63893350981875e-05, "loss": 2.1979, "step": 12212 }, { "epoch": 0.3, "learning_rate": 1.6388720952055897e-05, "loss": 2.2086, "step": 12213 }, { "epoch": 0.3, "learning_rate": 1.63881067652069e-05, "loss": 2.1305, "step": 12214 }, { "epoch": 0.3, "learning_rate": 1.638749253764442e-05, "loss": 2.4092, "step": 12215 }, { "epoch": 0.3, "learning_rate": 1.638687826937236e-05, "loss": 2.1935, "step": 12216 }, { "epoch": 0.3, "learning_rate": 1.638626396039465e-05, "loss": 2.3481, "step": 12217 }, { "epoch": 0.3, "learning_rate": 1.63856496107152e-05, "loss": 1.9869, "step": 12218 }, { "epoch": 0.3, "learning_rate": 1.638503522033792e-05, "loss": 1.9578, "step": 12219 }, { "epoch": 0.3, "learning_rate": 1.638442078926673e-05, "loss": 2.1632, "step": 12220 }, { "epoch": 0.3, "learning_rate": 1.6383806317505544e-05, "loss": 1.8798, "step": 12221 }, { "epoch": 0.3, "learning_rate": 1.6383191805058282e-05, "loss": 2.3023, "step": 12222 }, { "epoch": 0.3, "learning_rate": 1.6382577251928857e-05, "loss": 2.0355, "step": 12223 }, { "epoch": 0.3, "learning_rate": 1.638196265812119e-05, "loss": 2.0987, "step": 12224 }, { "epoch": 0.3, "learning_rate": 1.6381348023639195e-05, "loss": 2.1623, "step": 12225 }, { "epoch": 0.3, "learning_rate": 1.6380733348486787e-05, "loss": 2.169, "step": 12226 }, { "epoch": 0.3, "learning_rate": 1.6380118632667884e-05, "loss": 2.2973, "step": 12227 }, { "epoch": 0.3, "learning_rate": 1.6379503876186415e-05, "loss": 1.9764, "step": 12228 }, { "epoch": 0.3, "learning_rate": 1.637888907904628e-05, "loss": 2.2563, "step": 12229 }, { "epoch": 0.3, "learning_rate": 1.6378274241251407e-05, "loss": 2.1312, "step": 12230 }, { "epoch": 0.3, "learning_rate": 1.6377659362805714e-05, "loss": 2.1003, "step": 12231 }, { "epoch": 0.3, "learning_rate": 1.637704444371312e-05, "loss": 2.1288, "step": 12232 }, { "epoch": 0.3, "learning_rate": 1.6376429483977547e-05, "loss": 2.253, "step": 12233 }, { "epoch": 0.3, "learning_rate": 1.637581448360291e-05, "loss": 2.0964, "step": 12234 }, { "epoch": 0.3, "learning_rate": 1.6375199442593125e-05, "loss": 2.1067, "step": 12235 }, { "epoch": 0.3, "learning_rate": 1.637458436095212e-05, "loss": 2.0859, "step": 12236 }, { "epoch": 0.3, "learning_rate": 1.637396923868381e-05, "loss": 2.2537, "step": 12237 }, { "epoch": 0.3, "learning_rate": 1.6373354075792117e-05, "loss": 2.2161, "step": 12238 }, { "epoch": 0.3, "learning_rate": 1.637273887228096e-05, "loss": 2.1585, "step": 12239 }, { "epoch": 0.3, "learning_rate": 1.6372123628154265e-05, "loss": 1.8644, "step": 12240 }, { "epoch": 0.3, "learning_rate": 1.637150834341595e-05, "loss": 2.1339, "step": 12241 }, { "epoch": 0.3, "learning_rate": 1.637089301806993e-05, "loss": 2.0321, "step": 12242 }, { "epoch": 0.3, "learning_rate": 1.6370277652120138e-05, "loss": 2.1864, "step": 12243 }, { "epoch": 0.3, "learning_rate": 1.6369662245570487e-05, "loss": 2.1719, "step": 12244 }, { "epoch": 0.3, "learning_rate": 1.6369046798424902e-05, "loss": 2.1712, "step": 12245 }, { "epoch": 0.3, "learning_rate": 1.636843131068731e-05, "loss": 2.1666, "step": 12246 }, { "epoch": 0.3, "learning_rate": 1.636781578236163e-05, "loss": 2.1081, "step": 12247 }, { "epoch": 0.3, "learning_rate": 1.6367200213451782e-05, "loss": 2.1483, "step": 12248 }, { "epoch": 0.3, "learning_rate": 1.636658460396169e-05, "loss": 2.2722, "step": 12249 }, { "epoch": 0.3, "learning_rate": 1.6365968953895286e-05, "loss": 2.1919, "step": 12250 }, { "epoch": 0.3, "learning_rate": 1.6365353263256484e-05, "loss": 2.1733, "step": 12251 }, { "epoch": 0.3, "learning_rate": 1.636473753204921e-05, "loss": 2.2529, "step": 12252 }, { "epoch": 0.3, "learning_rate": 1.6364121760277392e-05, "loss": 2.3086, "step": 12253 }, { "epoch": 0.3, "learning_rate": 1.636350594794495e-05, "loss": 2.2169, "step": 12254 }, { "epoch": 0.3, "learning_rate": 1.636289009505581e-05, "loss": 2.1406, "step": 12255 }, { "epoch": 0.3, "learning_rate": 1.6362274201613902e-05, "loss": 2.1219, "step": 12256 }, { "epoch": 0.3, "learning_rate": 1.6361658267623144e-05, "loss": 2.1913, "step": 12257 }, { "epoch": 0.3, "learning_rate": 1.6361042293087463e-05, "loss": 2.2156, "step": 12258 }, { "epoch": 0.3, "learning_rate": 1.6360426278010787e-05, "loss": 2.2249, "step": 12259 }, { "epoch": 0.3, "learning_rate": 1.6359810222397045e-05, "loss": 2.1374, "step": 12260 }, { "epoch": 0.3, "learning_rate": 1.635919412625016e-05, "loss": 2.2815, "step": 12261 }, { "epoch": 0.3, "learning_rate": 1.6358577989574054e-05, "loss": 2.1233, "step": 12262 }, { "epoch": 0.3, "learning_rate": 1.635796181237266e-05, "loss": 2.0724, "step": 12263 }, { "epoch": 0.3, "learning_rate": 1.6357345594649902e-05, "loss": 2.038, "step": 12264 }, { "epoch": 0.3, "learning_rate": 1.6356729336409714e-05, "loss": 2.1148, "step": 12265 }, { "epoch": 0.3, "learning_rate": 1.6356113037656015e-05, "loss": 2.0337, "step": 12266 }, { "epoch": 0.3, "learning_rate": 1.6355496698392738e-05, "loss": 2.1923, "step": 12267 }, { "epoch": 0.3, "learning_rate": 1.6354880318623803e-05, "loss": 1.9256, "step": 12268 }, { "epoch": 0.3, "learning_rate": 1.635426389835315e-05, "loss": 2.2131, "step": 12269 }, { "epoch": 0.3, "learning_rate": 1.6353647437584702e-05, "loss": 2.2079, "step": 12270 }, { "epoch": 0.3, "learning_rate": 1.635303093632239e-05, "loss": 2.2257, "step": 12271 }, { "epoch": 0.3, "learning_rate": 1.6352414394570137e-05, "loss": 2.2137, "step": 12272 }, { "epoch": 0.3, "learning_rate": 1.635179781233188e-05, "loss": 2.1922, "step": 12273 }, { "epoch": 0.3, "learning_rate": 1.6351181189611546e-05, "loss": 2.2754, "step": 12274 }, { "epoch": 0.3, "learning_rate": 1.635056452641306e-05, "loss": 2.272, "step": 12275 }, { "epoch": 0.3, "learning_rate": 1.6349947822740364e-05, "loss": 2.2037, "step": 12276 }, { "epoch": 0.3, "learning_rate": 1.6349331078597374e-05, "loss": 2.0663, "step": 12277 }, { "epoch": 0.3, "learning_rate": 1.634871429398803e-05, "loss": 2.1201, "step": 12278 }, { "epoch": 0.3, "learning_rate": 1.6348097468916263e-05, "loss": 2.0172, "step": 12279 }, { "epoch": 0.3, "learning_rate": 1.6347480603386e-05, "loss": 2.2695, "step": 12280 }, { "epoch": 0.3, "learning_rate": 1.6346863697401175e-05, "loss": 1.9321, "step": 12281 }, { "epoch": 0.3, "learning_rate": 1.634624675096572e-05, "loss": 2.1045, "step": 12282 }, { "epoch": 0.3, "learning_rate": 1.6345629764083567e-05, "loss": 2.0461, "step": 12283 }, { "epoch": 0.3, "learning_rate": 1.6345012736758646e-05, "loss": 1.9836, "step": 12284 }, { "epoch": 0.3, "learning_rate": 1.634439566899489e-05, "loss": 2.1737, "step": 12285 }, { "epoch": 0.3, "learning_rate": 1.6343778560796236e-05, "loss": 2.0536, "step": 12286 }, { "epoch": 0.3, "learning_rate": 1.634316141216661e-05, "loss": 2.025, "step": 12287 }, { "epoch": 0.3, "learning_rate": 1.6342544223109954e-05, "loss": 2.1443, "step": 12288 }, { "epoch": 0.3, "learning_rate": 1.6341926993630194e-05, "loss": 2.01, "step": 12289 }, { "epoch": 0.3, "learning_rate": 1.6341309723731266e-05, "loss": 2.2866, "step": 12290 }, { "epoch": 0.3, "learning_rate": 1.6340692413417106e-05, "loss": 1.9988, "step": 12291 }, { "epoch": 0.3, "learning_rate": 1.6340075062691644e-05, "loss": 2.2693, "step": 12292 }, { "epoch": 0.3, "learning_rate": 1.6339457671558818e-05, "loss": 2.0312, "step": 12293 }, { "epoch": 0.3, "learning_rate": 1.6338840240022564e-05, "loss": 2.0855, "step": 12294 }, { "epoch": 0.3, "learning_rate": 1.633822276808682e-05, "loss": 2.2733, "step": 12295 }, { "epoch": 0.3, "learning_rate": 1.6337605255755508e-05, "loss": 2.2392, "step": 12296 }, { "epoch": 0.3, "learning_rate": 1.6336987703032576e-05, "loss": 2.3354, "step": 12297 }, { "epoch": 0.3, "learning_rate": 1.6336370109921955e-05, "loss": 2.0327, "step": 12298 }, { "epoch": 0.3, "learning_rate": 1.6335752476427585e-05, "loss": 2.0951, "step": 12299 }, { "epoch": 0.3, "learning_rate": 1.63351348025534e-05, "loss": 2.3344, "step": 12300 }, { "epoch": 0.3, "learning_rate": 1.633451708830333e-05, "loss": 2.0128, "step": 12301 }, { "epoch": 0.3, "learning_rate": 1.6333899333681325e-05, "loss": 2.1378, "step": 12302 }, { "epoch": 0.3, "learning_rate": 1.6333281538691314e-05, "loss": 2.2208, "step": 12303 }, { "epoch": 0.3, "learning_rate": 1.6332663703337236e-05, "loss": 2.2069, "step": 12304 }, { "epoch": 0.3, "learning_rate": 1.6332045827623027e-05, "loss": 2.1389, "step": 12305 }, { "epoch": 0.3, "learning_rate": 1.6331427911552627e-05, "loss": 2.0841, "step": 12306 }, { "epoch": 0.3, "learning_rate": 1.633080995512997e-05, "loss": 2.0854, "step": 12307 }, { "epoch": 0.3, "learning_rate": 1.6330191958359006e-05, "loss": 2.2429, "step": 12308 }, { "epoch": 0.3, "learning_rate": 1.632957392124366e-05, "loss": 2.1714, "step": 12309 }, { "epoch": 0.3, "learning_rate": 1.632895584378788e-05, "loss": 1.9631, "step": 12310 }, { "epoch": 0.3, "learning_rate": 1.6328337725995597e-05, "loss": 2.2554, "step": 12311 }, { "epoch": 0.3, "learning_rate": 1.6327719567870757e-05, "loss": 2.0968, "step": 12312 }, { "epoch": 0.3, "learning_rate": 1.63271013694173e-05, "loss": 2.3622, "step": 12313 }, { "epoch": 0.3, "learning_rate": 1.6326483130639162e-05, "loss": 2.0882, "step": 12314 }, { "epoch": 0.3, "learning_rate": 1.6325864851540288e-05, "loss": 2.0075, "step": 12315 }, { "epoch": 0.3, "learning_rate": 1.6325246532124614e-05, "loss": 2.1211, "step": 12316 }, { "epoch": 0.3, "learning_rate": 1.6324628172396083e-05, "loss": 1.9956, "step": 12317 }, { "epoch": 0.3, "learning_rate": 1.6324009772358634e-05, "loss": 2.1421, "step": 12318 }, { "epoch": 0.3, "learning_rate": 1.6323391332016213e-05, "loss": 2.0388, "step": 12319 }, { "epoch": 0.3, "learning_rate": 1.6322772851372753e-05, "loss": 2.1688, "step": 12320 }, { "epoch": 0.3, "learning_rate": 1.6322154330432207e-05, "loss": 2.1271, "step": 12321 }, { "epoch": 0.3, "learning_rate": 1.632153576919851e-05, "loss": 2.1828, "step": 12322 }, { "epoch": 0.3, "learning_rate": 1.63209171676756e-05, "loss": 2.1453, "step": 12323 }, { "epoch": 0.3, "learning_rate": 1.632029852586743e-05, "loss": 2.1394, "step": 12324 }, { "epoch": 0.3, "learning_rate": 1.6319679843777937e-05, "loss": 2.2119, "step": 12325 }, { "epoch": 0.3, "learning_rate": 1.6319061121411064e-05, "loss": 2.1708, "step": 12326 }, { "epoch": 0.3, "learning_rate": 1.6318442358770757e-05, "loss": 1.8989, "step": 12327 }, { "epoch": 0.3, "learning_rate": 1.6317823555860958e-05, "loss": 2.0297, "step": 12328 }, { "epoch": 0.3, "learning_rate": 1.631720471268561e-05, "loss": 2.1544, "step": 12329 }, { "epoch": 0.3, "learning_rate": 1.631658582924865e-05, "loss": 2.1785, "step": 12330 }, { "epoch": 0.3, "learning_rate": 1.6315966905554038e-05, "loss": 2.1536, "step": 12331 }, { "epoch": 0.3, "learning_rate": 1.6315347941605712e-05, "loss": 2.0026, "step": 12332 }, { "epoch": 0.3, "learning_rate": 1.631472893740761e-05, "loss": 2.0465, "step": 12333 }, { "epoch": 0.3, "learning_rate": 1.6314109892963686e-05, "loss": 2.0437, "step": 12334 }, { "epoch": 0.3, "learning_rate": 1.6313490808277885e-05, "loss": 2.0593, "step": 12335 }, { "epoch": 0.3, "learning_rate": 1.6312871683354142e-05, "loss": 2.1198, "step": 12336 }, { "epoch": 0.3, "learning_rate": 1.6312252518196416e-05, "loss": 2.2146, "step": 12337 }, { "epoch": 0.3, "learning_rate": 1.6311633312808642e-05, "loss": 2.1809, "step": 12338 }, { "epoch": 0.3, "learning_rate": 1.631101406719478e-05, "loss": 2.106, "step": 12339 }, { "epoch": 0.3, "learning_rate": 1.6310394781358762e-05, "loss": 2.168, "step": 12340 }, { "epoch": 0.3, "learning_rate": 1.6309775455304545e-05, "loss": 2.1821, "step": 12341 }, { "epoch": 0.3, "learning_rate": 1.630915608903607e-05, "loss": 2.1413, "step": 12342 }, { "epoch": 0.3, "learning_rate": 1.630853668255729e-05, "loss": 2.0359, "step": 12343 }, { "epoch": 0.3, "learning_rate": 1.6307917235872147e-05, "loss": 2.2123, "step": 12344 }, { "epoch": 0.3, "learning_rate": 1.6307297748984595e-05, "loss": 2.0659, "step": 12345 }, { "epoch": 0.3, "learning_rate": 1.6306678221898574e-05, "loss": 2.085, "step": 12346 }, { "epoch": 0.3, "learning_rate": 1.630605865461804e-05, "loss": 2.1866, "step": 12347 }, { "epoch": 0.3, "learning_rate": 1.6305439047146943e-05, "loss": 2.1732, "step": 12348 }, { "epoch": 0.3, "learning_rate": 1.6304819399489224e-05, "loss": 2.2662, "step": 12349 }, { "epoch": 0.3, "learning_rate": 1.6304199711648838e-05, "loss": 1.9901, "step": 12350 }, { "epoch": 0.3, "learning_rate": 1.6303579983629733e-05, "loss": 2.2182, "step": 12351 }, { "epoch": 0.3, "learning_rate": 1.6302960215435858e-05, "loss": 2.1312, "step": 12352 }, { "epoch": 0.3, "learning_rate": 1.6302340407071164e-05, "loss": 2.0766, "step": 12353 }, { "epoch": 0.3, "learning_rate": 1.63017205585396e-05, "loss": 2.1311, "step": 12354 }, { "epoch": 0.3, "learning_rate": 1.6301100669845112e-05, "loss": 1.9048, "step": 12355 }, { "epoch": 0.3, "learning_rate": 1.6300480740991662e-05, "loss": 2.1317, "step": 12356 }, { "epoch": 0.3, "learning_rate": 1.6299860771983196e-05, "loss": 2.1226, "step": 12357 }, { "epoch": 0.3, "learning_rate": 1.629924076282366e-05, "loss": 2.0859, "step": 12358 }, { "epoch": 0.3, "learning_rate": 1.6298620713517012e-05, "loss": 2.288, "step": 12359 }, { "epoch": 0.3, "learning_rate": 1.62980006240672e-05, "loss": 1.9755, "step": 12360 }, { "epoch": 0.3, "learning_rate": 1.629738049447818e-05, "loss": 2.2686, "step": 12361 }, { "epoch": 0.3, "learning_rate": 1.6296760324753903e-05, "loss": 2.0649, "step": 12362 }, { "epoch": 0.3, "learning_rate": 1.6296140114898316e-05, "loss": 2.1656, "step": 12363 }, { "epoch": 0.3, "learning_rate": 1.629551986491538e-05, "loss": 2.1912, "step": 12364 }, { "epoch": 0.3, "learning_rate": 1.629489957480904e-05, "loss": 2.2184, "step": 12365 }, { "epoch": 0.3, "learning_rate": 1.6294279244583257e-05, "loss": 2.2865, "step": 12366 }, { "epoch": 0.3, "learning_rate": 1.629365887424198e-05, "loss": 2.3573, "step": 12367 }, { "epoch": 0.3, "learning_rate": 1.6293038463789165e-05, "loss": 2.0032, "step": 12368 }, { "epoch": 0.3, "learning_rate": 1.6292418013228767e-05, "loss": 2.0923, "step": 12369 }, { "epoch": 0.3, "learning_rate": 1.6291797522564733e-05, "loss": 2.1654, "step": 12370 }, { "epoch": 0.3, "learning_rate": 1.6291176991801024e-05, "loss": 2.0875, "step": 12371 }, { "epoch": 0.3, "learning_rate": 1.6290556420941598e-05, "loss": 2.1172, "step": 12372 }, { "epoch": 0.3, "learning_rate": 1.62899358099904e-05, "loss": 2.1822, "step": 12373 }, { "epoch": 0.31, "learning_rate": 1.62893151589514e-05, "loss": 2.2748, "step": 12374 }, { "epoch": 0.31, "learning_rate": 1.6288694467828534e-05, "loss": 2.061, "step": 12375 }, { "epoch": 0.31, "learning_rate": 1.6288073736625777e-05, "loss": 2.0785, "step": 12376 }, { "epoch": 0.31, "learning_rate": 1.6287452965347073e-05, "loss": 2.2147, "step": 12377 }, { "epoch": 0.31, "learning_rate": 1.6286832153996382e-05, "loss": 2.2757, "step": 12378 }, { "epoch": 0.31, "learning_rate": 1.6286211302577665e-05, "loss": 2.2992, "step": 12379 }, { "epoch": 0.31, "learning_rate": 1.628559041109487e-05, "loss": 2.0397, "step": 12380 }, { "epoch": 0.31, "learning_rate": 1.628496947955196e-05, "loss": 1.9305, "step": 12381 }, { "epoch": 0.31, "learning_rate": 1.6284348507952893e-05, "loss": 2.2034, "step": 12382 }, { "epoch": 0.31, "learning_rate": 1.6283727496301624e-05, "loss": 2.1103, "step": 12383 }, { "epoch": 0.31, "learning_rate": 1.6283106444602112e-05, "loss": 2.2987, "step": 12384 }, { "epoch": 0.31, "learning_rate": 1.6282485352858314e-05, "loss": 2.2221, "step": 12385 }, { "epoch": 0.31, "learning_rate": 1.628186422107419e-05, "loss": 2.1511, "step": 12386 }, { "epoch": 0.31, "learning_rate": 1.62812430492537e-05, "loss": 2.1344, "step": 12387 }, { "epoch": 0.31, "learning_rate": 1.62806218374008e-05, "loss": 2.0301, "step": 12388 }, { "epoch": 0.31, "learning_rate": 1.6280000585519446e-05, "loss": 2.2249, "step": 12389 }, { "epoch": 0.31, "learning_rate": 1.6279379293613605e-05, "loss": 2.096, "step": 12390 }, { "epoch": 0.31, "learning_rate": 1.6278757961687235e-05, "loss": 1.9616, "step": 12391 }, { "epoch": 0.31, "learning_rate": 1.6278136589744292e-05, "loss": 2.2684, "step": 12392 }, { "epoch": 0.31, "learning_rate": 1.6277515177788742e-05, "loss": 2.0991, "step": 12393 }, { "epoch": 0.31, "learning_rate": 1.627689372582454e-05, "loss": 2.1451, "step": 12394 }, { "epoch": 0.31, "learning_rate": 1.6276272233855645e-05, "loss": 2.0854, "step": 12395 }, { "epoch": 0.31, "learning_rate": 1.627565070188603e-05, "loss": 2.2304, "step": 12396 }, { "epoch": 0.31, "learning_rate": 1.627502912991964e-05, "loss": 2.1099, "step": 12397 }, { "epoch": 0.31, "learning_rate": 1.6274407517960448e-05, "loss": 2.3194, "step": 12398 }, { "epoch": 0.31, "learning_rate": 1.6273785866012407e-05, "loss": 2.1198, "step": 12399 }, { "epoch": 0.31, "learning_rate": 1.627316417407949e-05, "loss": 2.2194, "step": 12400 }, { "epoch": 0.31, "learning_rate": 1.6272542442165654e-05, "loss": 2.2889, "step": 12401 }, { "epoch": 0.31, "learning_rate": 1.6271920670274858e-05, "loss": 2.107, "step": 12402 }, { "epoch": 0.31, "learning_rate": 1.627129885841107e-05, "loss": 2.0578, "step": 12403 }, { "epoch": 0.31, "learning_rate": 1.6270677006578245e-05, "loss": 1.8926, "step": 12404 }, { "epoch": 0.31, "learning_rate": 1.6270055114780356e-05, "loss": 2.0334, "step": 12405 }, { "epoch": 0.31, "learning_rate": 1.6269433183021364e-05, "loss": 2.2737, "step": 12406 }, { "epoch": 0.31, "learning_rate": 1.626881121130523e-05, "loss": 2.1041, "step": 12407 }, { "epoch": 0.31, "learning_rate": 1.626818919963592e-05, "loss": 2.2712, "step": 12408 }, { "epoch": 0.31, "learning_rate": 1.626756714801739e-05, "loss": 2.0013, "step": 12409 }, { "epoch": 0.31, "learning_rate": 1.6266945056453623e-05, "loss": 2.2482, "step": 12410 }, { "epoch": 0.31, "learning_rate": 1.6266322924948564e-05, "loss": 2.058, "step": 12411 }, { "epoch": 0.31, "learning_rate": 1.6265700753506194e-05, "loss": 2.1634, "step": 12412 }, { "epoch": 0.31, "learning_rate": 1.6265078542130466e-05, "loss": 1.9524, "step": 12413 }, { "epoch": 0.31, "learning_rate": 1.6264456290825353e-05, "loss": 2.0597, "step": 12414 }, { "epoch": 0.31, "learning_rate": 1.6263833999594818e-05, "loss": 1.9724, "step": 12415 }, { "epoch": 0.31, "learning_rate": 1.6263211668442827e-05, "loss": 2.1795, "step": 12416 }, { "epoch": 0.31, "learning_rate": 1.6262589297373348e-05, "loss": 2.1892, "step": 12417 }, { "epoch": 0.31, "learning_rate": 1.6261966886390345e-05, "loss": 2.219, "step": 12418 }, { "epoch": 0.31, "learning_rate": 1.626134443549779e-05, "loss": 2.169, "step": 12419 }, { "epoch": 0.31, "learning_rate": 1.6260721944699643e-05, "loss": 2.0708, "step": 12420 }, { "epoch": 0.31, "learning_rate": 1.6260099413999877e-05, "loss": 2.2198, "step": 12421 }, { "epoch": 0.31, "learning_rate": 1.6259476843402456e-05, "loss": 2.1446, "step": 12422 }, { "epoch": 0.31, "learning_rate": 1.6258854232911345e-05, "loss": 2.1003, "step": 12423 }, { "epoch": 0.31, "learning_rate": 1.6258231582530522e-05, "loss": 2.1191, "step": 12424 }, { "epoch": 0.31, "learning_rate": 1.625760889226395e-05, "loss": 2.1811, "step": 12425 }, { "epoch": 0.31, "learning_rate": 1.6256986162115596e-05, "loss": 1.9385, "step": 12426 }, { "epoch": 0.31, "learning_rate": 1.6256363392089428e-05, "loss": 2.1724, "step": 12427 }, { "epoch": 0.31, "learning_rate": 1.6255740582189418e-05, "loss": 2.2536, "step": 12428 }, { "epoch": 0.31, "learning_rate": 1.6255117732419533e-05, "loss": 2.1932, "step": 12429 }, { "epoch": 0.31, "learning_rate": 1.6254494842783747e-05, "loss": 1.9602, "step": 12430 }, { "epoch": 0.31, "learning_rate": 1.625387191328603e-05, "loss": 2.1704, "step": 12431 }, { "epoch": 0.31, "learning_rate": 1.6253248943930345e-05, "loss": 2.2884, "step": 12432 }, { "epoch": 0.31, "learning_rate": 1.6252625934720664e-05, "loss": 2.0395, "step": 12433 }, { "epoch": 0.31, "learning_rate": 1.6252002885660965e-05, "loss": 2.142, "step": 12434 }, { "epoch": 0.31, "learning_rate": 1.6251379796755214e-05, "loss": 2.1752, "step": 12435 }, { "epoch": 0.31, "learning_rate": 1.625075666800738e-05, "loss": 2.0222, "step": 12436 }, { "epoch": 0.31, "learning_rate": 1.6250133499421436e-05, "loss": 2.1573, "step": 12437 }, { "epoch": 0.31, "learning_rate": 1.6249510291001355e-05, "loss": 2.3039, "step": 12438 }, { "epoch": 0.31, "learning_rate": 1.6248887042751112e-05, "loss": 2.0976, "step": 12439 }, { "epoch": 0.31, "learning_rate": 1.624826375467467e-05, "loss": 2.1142, "step": 12440 }, { "epoch": 0.31, "learning_rate": 1.6247640426776012e-05, "loss": 2.1366, "step": 12441 }, { "epoch": 0.31, "learning_rate": 1.6247017059059103e-05, "loss": 2.1895, "step": 12442 }, { "epoch": 0.31, "learning_rate": 1.6246393651527916e-05, "loss": 2.0609, "step": 12443 }, { "epoch": 0.31, "learning_rate": 1.6245770204186427e-05, "loss": 2.3369, "step": 12444 }, { "epoch": 0.31, "learning_rate": 1.6245146717038615e-05, "loss": 2.0635, "step": 12445 }, { "epoch": 0.31, "learning_rate": 1.624452319008844e-05, "loss": 2.2155, "step": 12446 }, { "epoch": 0.31, "learning_rate": 1.624389962333989e-05, "loss": 2.0125, "step": 12447 }, { "epoch": 0.31, "learning_rate": 1.624327601679693e-05, "loss": 2.1304, "step": 12448 }, { "epoch": 0.31, "learning_rate": 1.624265237046354e-05, "loss": 1.9609, "step": 12449 }, { "epoch": 0.31, "learning_rate": 1.624202868434369e-05, "loss": 2.1659, "step": 12450 }, { "epoch": 0.31, "learning_rate": 1.6241404958441352e-05, "loss": 2.1242, "step": 12451 }, { "epoch": 0.31, "learning_rate": 1.6240781192760514e-05, "loss": 2.0546, "step": 12452 }, { "epoch": 0.31, "learning_rate": 1.624015738730514e-05, "loss": 2.2452, "step": 12453 }, { "epoch": 0.31, "learning_rate": 1.6239533542079212e-05, "loss": 2.2458, "step": 12454 }, { "epoch": 0.31, "learning_rate": 1.62389096570867e-05, "loss": 2.2559, "step": 12455 }, { "epoch": 0.31, "learning_rate": 1.6238285732331583e-05, "loss": 2.2029, "step": 12456 }, { "epoch": 0.31, "learning_rate": 1.6237661767817842e-05, "loss": 2.1073, "step": 12457 }, { "epoch": 0.31, "learning_rate": 1.6237037763549446e-05, "loss": 2.162, "step": 12458 }, { "epoch": 0.31, "learning_rate": 1.623641371953038e-05, "loss": 2.1068, "step": 12459 }, { "epoch": 0.31, "learning_rate": 1.6235789635764612e-05, "loss": 2.2994, "step": 12460 }, { "epoch": 0.31, "learning_rate": 1.623516551225613e-05, "loss": 2.1572, "step": 12461 }, { "epoch": 0.31, "learning_rate": 1.6234541349008898e-05, "loss": 2.1977, "step": 12462 }, { "epoch": 0.31, "learning_rate": 1.623391714602691e-05, "loss": 2.1891, "step": 12463 }, { "epoch": 0.31, "learning_rate": 1.6233292903314133e-05, "loss": 2.384, "step": 12464 }, { "epoch": 0.31, "learning_rate": 1.623266862087455e-05, "loss": 2.1781, "step": 12465 }, { "epoch": 0.31, "learning_rate": 1.623204429871214e-05, "loss": 1.9613, "step": 12466 }, { "epoch": 0.31, "learning_rate": 1.6231419936830875e-05, "loss": 2.1642, "step": 12467 }, { "epoch": 0.31, "learning_rate": 1.6230795535234746e-05, "loss": 2.0419, "step": 12468 }, { "epoch": 0.31, "learning_rate": 1.6230171093927724e-05, "loss": 1.9844, "step": 12469 }, { "epoch": 0.31, "learning_rate": 1.622954661291379e-05, "loss": 2.1801, "step": 12470 }, { "epoch": 0.31, "learning_rate": 1.6228922092196927e-05, "loss": 2.1012, "step": 12471 }, { "epoch": 0.31, "learning_rate": 1.6228297531781115e-05, "loss": 2.1502, "step": 12472 }, { "epoch": 0.31, "learning_rate": 1.622767293167033e-05, "loss": 2.2165, "step": 12473 }, { "epoch": 0.31, "learning_rate": 1.622704829186856e-05, "loss": 2.2417, "step": 12474 }, { "epoch": 0.31, "learning_rate": 1.6226423612379776e-05, "loss": 2.0439, "step": 12475 }, { "epoch": 0.31, "learning_rate": 1.622579889320797e-05, "loss": 2.0157, "step": 12476 }, { "epoch": 0.31, "learning_rate": 1.622517413435712e-05, "loss": 2.1949, "step": 12477 }, { "epoch": 0.31, "learning_rate": 1.62245493358312e-05, "loss": 2.0706, "step": 12478 }, { "epoch": 0.31, "learning_rate": 1.6223924497634206e-05, "loss": 2.1274, "step": 12479 }, { "epoch": 0.31, "learning_rate": 1.6223299619770106e-05, "loss": 2.0505, "step": 12480 }, { "epoch": 0.31, "learning_rate": 1.6222674702242892e-05, "loss": 2.165, "step": 12481 }, { "epoch": 0.31, "learning_rate": 1.6222049745056546e-05, "loss": 2.0172, "step": 12482 }, { "epoch": 0.31, "learning_rate": 1.622142474821505e-05, "loss": 2.1678, "step": 12483 }, { "epoch": 0.31, "learning_rate": 1.6220799711722385e-05, "loss": 1.9758, "step": 12484 }, { "epoch": 0.31, "learning_rate": 1.6220174635582536e-05, "loss": 2.1333, "step": 12485 }, { "epoch": 0.31, "learning_rate": 1.6219549519799485e-05, "loss": 2.219, "step": 12486 }, { "epoch": 0.31, "learning_rate": 1.621892436437722e-05, "loss": 2.1584, "step": 12487 }, { "epoch": 0.31, "learning_rate": 1.6218299169319722e-05, "loss": 2.1552, "step": 12488 }, { "epoch": 0.31, "learning_rate": 1.621767393463098e-05, "loss": 1.995, "step": 12489 }, { "epoch": 0.31, "learning_rate": 1.6217048660314975e-05, "loss": 2.0007, "step": 12490 }, { "epoch": 0.31, "learning_rate": 1.6216423346375688e-05, "loss": 2.0665, "step": 12491 }, { "epoch": 0.31, "learning_rate": 1.6215797992817113e-05, "loss": 2.0753, "step": 12492 }, { "epoch": 0.31, "learning_rate": 1.6215172599643233e-05, "loss": 2.1064, "step": 12493 }, { "epoch": 0.31, "learning_rate": 1.6214547166858034e-05, "loss": 2.0465, "step": 12494 }, { "epoch": 0.31, "learning_rate": 1.6213921694465496e-05, "loss": 2.1078, "step": 12495 }, { "epoch": 0.31, "learning_rate": 1.621329618246961e-05, "loss": 2.0541, "step": 12496 }, { "epoch": 0.31, "learning_rate": 1.6212670630874364e-05, "loss": 2.1491, "step": 12497 }, { "epoch": 0.31, "learning_rate": 1.6212045039683744e-05, "loss": 2.0375, "step": 12498 }, { "epoch": 0.31, "learning_rate": 1.6211419408901737e-05, "loss": 2.192, "step": 12499 }, { "epoch": 0.31, "learning_rate": 1.6210793738532328e-05, "loss": 2.2918, "step": 12500 }, { "epoch": 0.31, "learning_rate": 1.6210168028579507e-05, "loss": 2.165, "step": 12501 }, { "epoch": 0.31, "learning_rate": 1.6209542279047262e-05, "loss": 2.0791, "step": 12502 }, { "epoch": 0.31, "learning_rate": 1.620891648993958e-05, "loss": 2.0191, "step": 12503 }, { "epoch": 0.31, "learning_rate": 1.620829066126045e-05, "loss": 2.0564, "step": 12504 }, { "epoch": 0.31, "learning_rate": 1.620766479301386e-05, "loss": 2.1528, "step": 12505 }, { "epoch": 0.31, "learning_rate": 1.6207038885203802e-05, "loss": 1.973, "step": 12506 }, { "epoch": 0.31, "learning_rate": 1.620641293783426e-05, "loss": 2.1983, "step": 12507 }, { "epoch": 0.31, "learning_rate": 1.6205786950909224e-05, "loss": 2.0764, "step": 12508 }, { "epoch": 0.31, "learning_rate": 1.620516092443269e-05, "loss": 2.164, "step": 12509 }, { "epoch": 0.31, "learning_rate": 1.6204534858408637e-05, "loss": 2.2415, "step": 12510 }, { "epoch": 0.31, "learning_rate": 1.6203908752841065e-05, "loss": 2.1465, "step": 12511 }, { "epoch": 0.31, "learning_rate": 1.620328260773396e-05, "loss": 2.005, "step": 12512 }, { "epoch": 0.31, "learning_rate": 1.6202656423091317e-05, "loss": 2.0795, "step": 12513 }, { "epoch": 0.31, "learning_rate": 1.620203019891712e-05, "loss": 2.1315, "step": 12514 }, { "epoch": 0.31, "learning_rate": 1.620140393521536e-05, "loss": 2.1386, "step": 12515 }, { "epoch": 0.31, "learning_rate": 1.6200777631990038e-05, "loss": 2.0104, "step": 12516 }, { "epoch": 0.31, "learning_rate": 1.6200151289245133e-05, "loss": 2.2576, "step": 12517 }, { "epoch": 0.31, "learning_rate": 1.6199524906984647e-05, "loss": 2.0934, "step": 12518 }, { "epoch": 0.31, "learning_rate": 1.619889848521257e-05, "loss": 2.0199, "step": 12519 }, { "epoch": 0.31, "learning_rate": 1.619827202393289e-05, "loss": 2.1132, "step": 12520 }, { "epoch": 0.31, "learning_rate": 1.6197645523149604e-05, "loss": 2.186, "step": 12521 }, { "epoch": 0.31, "learning_rate": 1.61970189828667e-05, "loss": 2.1698, "step": 12522 }, { "epoch": 0.31, "learning_rate": 1.619639240308818e-05, "loss": 2.0663, "step": 12523 }, { "epoch": 0.31, "learning_rate": 1.619576578381803e-05, "loss": 2.0857, "step": 12524 }, { "epoch": 0.31, "learning_rate": 1.619513912506024e-05, "loss": 2.0089, "step": 12525 }, { "epoch": 0.31, "learning_rate": 1.6194512426818816e-05, "loss": 2.2667, "step": 12526 }, { "epoch": 0.31, "learning_rate": 1.6193885689097745e-05, "loss": 2.115, "step": 12527 }, { "epoch": 0.31, "learning_rate": 1.6193258911901022e-05, "loss": 2.0942, "step": 12528 }, { "epoch": 0.31, "learning_rate": 1.6192632095232638e-05, "loss": 1.8878, "step": 12529 }, { "epoch": 0.31, "learning_rate": 1.619200523909659e-05, "loss": 2.2038, "step": 12530 }, { "epoch": 0.31, "learning_rate": 1.6191378343496884e-05, "loss": 1.8465, "step": 12531 }, { "epoch": 0.31, "learning_rate": 1.61907514084375e-05, "loss": 2.0661, "step": 12532 }, { "epoch": 0.31, "learning_rate": 1.619012443392244e-05, "loss": 1.9416, "step": 12533 }, { "epoch": 0.31, "learning_rate": 1.6189497419955696e-05, "loss": 2.1562, "step": 12534 }, { "epoch": 0.31, "learning_rate": 1.6188870366541275e-05, "loss": 2.075, "step": 12535 }, { "epoch": 0.31, "learning_rate": 1.6188243273683162e-05, "loss": 2.0779, "step": 12536 }, { "epoch": 0.31, "learning_rate": 1.6187616141385358e-05, "loss": 2.2178, "step": 12537 }, { "epoch": 0.31, "learning_rate": 1.6186988969651862e-05, "loss": 2.1287, "step": 12538 }, { "epoch": 0.31, "learning_rate": 1.618636175848667e-05, "loss": 2.1507, "step": 12539 }, { "epoch": 0.31, "learning_rate": 1.6185734507893772e-05, "loss": 2.1534, "step": 12540 }, { "epoch": 0.31, "learning_rate": 1.6185107217877177e-05, "loss": 2.1028, "step": 12541 }, { "epoch": 0.31, "learning_rate": 1.6184479888440878e-05, "loss": 2.2918, "step": 12542 }, { "epoch": 0.31, "learning_rate": 1.6183852519588873e-05, "loss": 2.153, "step": 12543 }, { "epoch": 0.31, "learning_rate": 1.618322511132516e-05, "loss": 2.2343, "step": 12544 }, { "epoch": 0.31, "learning_rate": 1.6182597663653738e-05, "loss": 2.0042, "step": 12545 }, { "epoch": 0.31, "learning_rate": 1.618197017657861e-05, "loss": 2.2435, "step": 12546 }, { "epoch": 0.31, "learning_rate": 1.6181342650103766e-05, "loss": 2.1328, "step": 12547 }, { "epoch": 0.31, "learning_rate": 1.6180715084233215e-05, "loss": 2.197, "step": 12548 }, { "epoch": 0.31, "learning_rate": 1.6180087478970947e-05, "loss": 1.8983, "step": 12549 }, { "epoch": 0.31, "learning_rate": 1.6179459834320973e-05, "loss": 2.2611, "step": 12550 }, { "epoch": 0.31, "learning_rate": 1.6178832150287286e-05, "loss": 2.0898, "step": 12551 }, { "epoch": 0.31, "learning_rate": 1.6178204426873887e-05, "loss": 1.9865, "step": 12552 }, { "epoch": 0.31, "learning_rate": 1.6177576664084778e-05, "loss": 2.2547, "step": 12553 }, { "epoch": 0.31, "learning_rate": 1.617694886192396e-05, "loss": 2.1549, "step": 12554 }, { "epoch": 0.31, "learning_rate": 1.6176321020395436e-05, "loss": 2.1423, "step": 12555 }, { "epoch": 0.31, "learning_rate": 1.6175693139503202e-05, "loss": 2.1674, "step": 12556 }, { "epoch": 0.31, "learning_rate": 1.6175065219251264e-05, "loss": 2.074, "step": 12557 }, { "epoch": 0.31, "learning_rate": 1.617443725964362e-05, "loss": 1.9832, "step": 12558 }, { "epoch": 0.31, "learning_rate": 1.617380926068428e-05, "loss": 1.9042, "step": 12559 }, { "epoch": 0.31, "learning_rate": 1.6173181222377243e-05, "loss": 2.1585, "step": 12560 }, { "epoch": 0.31, "learning_rate": 1.617255314472651e-05, "loss": 2.2025, "step": 12561 }, { "epoch": 0.31, "learning_rate": 1.617192502773608e-05, "loss": 2.1465, "step": 12562 }, { "epoch": 0.31, "learning_rate": 1.617129687140996e-05, "loss": 2.2326, "step": 12563 }, { "epoch": 0.31, "learning_rate": 1.617066867575216e-05, "loss": 2.1138, "step": 12564 }, { "epoch": 0.31, "learning_rate": 1.617004044076667e-05, "loss": 2.204, "step": 12565 }, { "epoch": 0.31, "learning_rate": 1.6169412166457506e-05, "loss": 2.2208, "step": 12566 }, { "epoch": 0.31, "learning_rate": 1.616878385282867e-05, "loss": 2.3753, "step": 12567 }, { "epoch": 0.31, "learning_rate": 1.6168155499884157e-05, "loss": 2.0235, "step": 12568 }, { "epoch": 0.31, "learning_rate": 1.6167527107627985e-05, "loss": 1.8742, "step": 12569 }, { "epoch": 0.31, "learning_rate": 1.616689867606415e-05, "loss": 2.2354, "step": 12570 }, { "epoch": 0.31, "learning_rate": 1.6166270205196658e-05, "loss": 2.2323, "step": 12571 }, { "epoch": 0.31, "learning_rate": 1.6165641695029523e-05, "loss": 2.1228, "step": 12572 }, { "epoch": 0.31, "learning_rate": 1.6165013145566736e-05, "loss": 2.0605, "step": 12573 }, { "epoch": 0.31, "learning_rate": 1.6164384556812317e-05, "loss": 2.1606, "step": 12574 }, { "epoch": 0.31, "learning_rate": 1.6163755928770264e-05, "loss": 2.2787, "step": 12575 }, { "epoch": 0.31, "learning_rate": 1.6163127261444588e-05, "loss": 2.0809, "step": 12576 }, { "epoch": 0.31, "learning_rate": 1.616249855483929e-05, "loss": 2.2232, "step": 12577 }, { "epoch": 0.31, "learning_rate": 1.6161869808958383e-05, "loss": 2.3068, "step": 12578 }, { "epoch": 0.31, "learning_rate": 1.616124102380587e-05, "loss": 2.1544, "step": 12579 }, { "epoch": 0.31, "learning_rate": 1.616061219938576e-05, "loss": 2.1381, "step": 12580 }, { "epoch": 0.31, "learning_rate": 1.6159983335702065e-05, "loss": 2.2678, "step": 12581 }, { "epoch": 0.31, "learning_rate": 1.6159354432758786e-05, "loss": 2.0642, "step": 12582 }, { "epoch": 0.31, "learning_rate": 1.615872549055993e-05, "loss": 2.067, "step": 12583 }, { "epoch": 0.31, "learning_rate": 1.6158096509109515e-05, "loss": 2.1064, "step": 12584 }, { "epoch": 0.31, "learning_rate": 1.6157467488411543e-05, "loss": 2.27, "step": 12585 }, { "epoch": 0.31, "learning_rate": 1.6156838428470022e-05, "loss": 1.9941, "step": 12586 }, { "epoch": 0.31, "learning_rate": 1.6156209329288962e-05, "loss": 2.1309, "step": 12587 }, { "epoch": 0.31, "learning_rate": 1.6155580190872376e-05, "loss": 2.1679, "step": 12588 }, { "epoch": 0.31, "learning_rate": 1.615495101322427e-05, "loss": 2.1606, "step": 12589 }, { "epoch": 0.31, "learning_rate": 1.615432179634866e-05, "loss": 2.1094, "step": 12590 }, { "epoch": 0.31, "learning_rate": 1.615369254024955e-05, "loss": 2.0526, "step": 12591 }, { "epoch": 0.31, "learning_rate": 1.6153063244930946e-05, "loss": 2.2364, "step": 12592 }, { "epoch": 0.31, "learning_rate": 1.6152433910396868e-05, "loss": 2.1788, "step": 12593 }, { "epoch": 0.31, "learning_rate": 1.6151804536651328e-05, "loss": 2.1787, "step": 12594 }, { "epoch": 0.31, "learning_rate": 1.6151175123698327e-05, "loss": 1.9769, "step": 12595 }, { "epoch": 0.31, "learning_rate": 1.6150545671541883e-05, "loss": 2.1396, "step": 12596 }, { "epoch": 0.31, "learning_rate": 1.614991618018601e-05, "loss": 2.0767, "step": 12597 }, { "epoch": 0.31, "learning_rate": 1.6149286649634714e-05, "loss": 1.9886, "step": 12598 }, { "epoch": 0.31, "learning_rate": 1.6148657079892015e-05, "loss": 2.1161, "step": 12599 }, { "epoch": 0.31, "learning_rate": 1.614802747096192e-05, "loss": 2.0818, "step": 12600 }, { "epoch": 0.31, "learning_rate": 1.6147397822848438e-05, "loss": 2.224, "step": 12601 }, { "epoch": 0.31, "learning_rate": 1.6146768135555587e-05, "loss": 2.1476, "step": 12602 }, { "epoch": 0.31, "learning_rate": 1.614613840908738e-05, "loss": 2.3361, "step": 12603 }, { "epoch": 0.31, "learning_rate": 1.614550864344783e-05, "loss": 2.0092, "step": 12604 }, { "epoch": 0.31, "learning_rate": 1.6144878838640953e-05, "loss": 2.156, "step": 12605 }, { "epoch": 0.31, "learning_rate": 1.6144248994670758e-05, "loss": 2.1112, "step": 12606 }, { "epoch": 0.31, "learning_rate": 1.6143619111541264e-05, "loss": 2.1521, "step": 12607 }, { "epoch": 0.31, "learning_rate": 1.614298918925648e-05, "loss": 2.1034, "step": 12608 }, { "epoch": 0.31, "learning_rate": 1.614235922782043e-05, "loss": 2.1243, "step": 12609 }, { "epoch": 0.31, "learning_rate": 1.6141729227237114e-05, "loss": 2.1976, "step": 12610 }, { "epoch": 0.31, "learning_rate": 1.614109918751056e-05, "loss": 2.265, "step": 12611 }, { "epoch": 0.31, "learning_rate": 1.6140469108644784e-05, "loss": 2.1485, "step": 12612 }, { "epoch": 0.31, "learning_rate": 1.6139838990643794e-05, "loss": 2.0131, "step": 12613 }, { "epoch": 0.31, "learning_rate": 1.6139208833511606e-05, "loss": 2.3138, "step": 12614 }, { "epoch": 0.31, "learning_rate": 1.6138578637252242e-05, "loss": 2.2237, "step": 12615 }, { "epoch": 0.31, "learning_rate": 1.6137948401869718e-05, "loss": 2.2177, "step": 12616 }, { "epoch": 0.31, "learning_rate": 1.6137318127368048e-05, "loss": 2.1112, "step": 12617 }, { "epoch": 0.31, "learning_rate": 1.6136687813751243e-05, "loss": 2.1181, "step": 12618 }, { "epoch": 0.31, "learning_rate": 1.6136057461023335e-05, "loss": 2.0696, "step": 12619 }, { "epoch": 0.31, "learning_rate": 1.6135427069188333e-05, "loss": 2.1514, "step": 12620 }, { "epoch": 0.31, "learning_rate": 1.6134796638250248e-05, "loss": 2.2604, "step": 12621 }, { "epoch": 0.31, "learning_rate": 1.613416616821311e-05, "loss": 2.0866, "step": 12622 }, { "epoch": 0.31, "learning_rate": 1.613353565908093e-05, "loss": 2.1489, "step": 12623 }, { "epoch": 0.31, "learning_rate": 1.6132905110857727e-05, "loss": 2.2334, "step": 12624 }, { "epoch": 0.31, "learning_rate": 1.6132274523547528e-05, "loss": 2.2084, "step": 12625 }, { "epoch": 0.31, "learning_rate": 1.6131643897154337e-05, "loss": 2.2741, "step": 12626 }, { "epoch": 0.31, "learning_rate": 1.613101323168218e-05, "loss": 2.1395, "step": 12627 }, { "epoch": 0.31, "learning_rate": 1.6130382527135086e-05, "loss": 2.2811, "step": 12628 }, { "epoch": 0.31, "learning_rate": 1.612975178351706e-05, "loss": 2.1401, "step": 12629 }, { "epoch": 0.31, "learning_rate": 1.612912100083213e-05, "loss": 2.1676, "step": 12630 }, { "epoch": 0.31, "learning_rate": 1.6128490179084316e-05, "loss": 2.1039, "step": 12631 }, { "epoch": 0.31, "learning_rate": 1.6127859318277634e-05, "loss": 2.1508, "step": 12632 }, { "epoch": 0.31, "learning_rate": 1.612722841841611e-05, "loss": 1.8857, "step": 12633 }, { "epoch": 0.31, "learning_rate": 1.6126597479503763e-05, "loss": 2.0228, "step": 12634 }, { "epoch": 0.31, "learning_rate": 1.612596650154461e-05, "loss": 2.129, "step": 12635 }, { "epoch": 0.31, "learning_rate": 1.612533548454268e-05, "loss": 2.1627, "step": 12636 }, { "epoch": 0.31, "learning_rate": 1.612470442850199e-05, "loss": 2.0213, "step": 12637 }, { "epoch": 0.31, "learning_rate": 1.612407333342656e-05, "loss": 2.0332, "step": 12638 }, { "epoch": 0.31, "learning_rate": 1.612344219932042e-05, "loss": 2.2981, "step": 12639 }, { "epoch": 0.31, "learning_rate": 1.6122811026187583e-05, "loss": 2.1206, "step": 12640 }, { "epoch": 0.31, "learning_rate": 1.6122179814032075e-05, "loss": 2.0632, "step": 12641 }, { "epoch": 0.31, "learning_rate": 1.612154856285792e-05, "loss": 2.148, "step": 12642 }, { "epoch": 0.31, "learning_rate": 1.6120917272669145e-05, "loss": 2.0805, "step": 12643 }, { "epoch": 0.31, "learning_rate": 1.6120285943469767e-05, "loss": 2.1725, "step": 12644 }, { "epoch": 0.31, "learning_rate": 1.6119654575263814e-05, "loss": 2.261, "step": 12645 }, { "epoch": 0.31, "learning_rate": 1.6119023168055307e-05, "loss": 2.1939, "step": 12646 }, { "epoch": 0.31, "learning_rate": 1.6118391721848268e-05, "loss": 2.2116, "step": 12647 }, { "epoch": 0.31, "learning_rate": 1.611776023664673e-05, "loss": 2.2051, "step": 12648 }, { "epoch": 0.31, "learning_rate": 1.6117128712454708e-05, "loss": 1.9129, "step": 12649 }, { "epoch": 0.31, "learning_rate": 1.6116497149276234e-05, "loss": 2.1671, "step": 12650 }, { "epoch": 0.31, "learning_rate": 1.611586554711533e-05, "loss": 2.0606, "step": 12651 }, { "epoch": 0.31, "learning_rate": 1.611523390597602e-05, "loss": 2.1749, "step": 12652 }, { "epoch": 0.31, "learning_rate": 1.6114602225862334e-05, "loss": 1.9995, "step": 12653 }, { "epoch": 0.31, "learning_rate": 1.6113970506778294e-05, "loss": 2.2094, "step": 12654 }, { "epoch": 0.31, "learning_rate": 1.6113338748727924e-05, "loss": 2.1933, "step": 12655 }, { "epoch": 0.31, "learning_rate": 1.6112706951715258e-05, "loss": 2.0827, "step": 12656 }, { "epoch": 0.31, "learning_rate": 1.6112075115744317e-05, "loss": 2.0855, "step": 12657 }, { "epoch": 0.31, "learning_rate": 1.611144324081913e-05, "loss": 2.113, "step": 12658 }, { "epoch": 0.31, "learning_rate": 1.6110811326943725e-05, "loss": 2.1295, "step": 12659 }, { "epoch": 0.31, "learning_rate": 1.6110179374122126e-05, "loss": 1.8239, "step": 12660 }, { "epoch": 0.31, "learning_rate": 1.6109547382358364e-05, "loss": 2.0333, "step": 12661 }, { "epoch": 0.31, "learning_rate": 1.6108915351656464e-05, "loss": 2.0964, "step": 12662 }, { "epoch": 0.31, "learning_rate": 1.6108283282020456e-05, "loss": 2.2109, "step": 12663 }, { "epoch": 0.31, "learning_rate": 1.6107651173454368e-05, "loss": 2.1058, "step": 12664 }, { "epoch": 0.31, "learning_rate": 1.610701902596223e-05, "loss": 2.0704, "step": 12665 }, { "epoch": 0.31, "learning_rate": 1.6106386839548067e-05, "loss": 2.2413, "step": 12666 }, { "epoch": 0.31, "learning_rate": 1.610575461421591e-05, "loss": 2.1003, "step": 12667 }, { "epoch": 0.31, "learning_rate": 1.6105122349969795e-05, "loss": 2.1284, "step": 12668 }, { "epoch": 0.31, "learning_rate": 1.610449004681374e-05, "loss": 2.3243, "step": 12669 }, { "epoch": 0.31, "learning_rate": 1.6103857704751784e-05, "loss": 2.0913, "step": 12670 }, { "epoch": 0.31, "learning_rate": 1.610322532378795e-05, "loss": 2.1051, "step": 12671 }, { "epoch": 0.31, "learning_rate": 1.6102592903926275e-05, "loss": 2.2295, "step": 12672 }, { "epoch": 0.31, "learning_rate": 1.610196044517079e-05, "loss": 2.1532, "step": 12673 }, { "epoch": 0.31, "learning_rate": 1.6101327947525514e-05, "loss": 2.0725, "step": 12674 }, { "epoch": 0.31, "learning_rate": 1.610069541099449e-05, "loss": 2.1386, "step": 12675 }, { "epoch": 0.31, "learning_rate": 1.6100062835581748e-05, "loss": 2.2247, "step": 12676 }, { "epoch": 0.31, "learning_rate": 1.6099430221291317e-05, "loss": 2.1215, "step": 12677 }, { "epoch": 0.31, "learning_rate": 1.6098797568127228e-05, "loss": 2.0425, "step": 12678 }, { "epoch": 0.31, "learning_rate": 1.6098164876093515e-05, "loss": 2.0976, "step": 12679 }, { "epoch": 0.31, "learning_rate": 1.609753214519421e-05, "loss": 2.1344, "step": 12680 }, { "epoch": 0.31, "learning_rate": 1.6096899375433346e-05, "loss": 2.1086, "step": 12681 }, { "epoch": 0.31, "learning_rate": 1.6096266566814954e-05, "loss": 2.0821, "step": 12682 }, { "epoch": 0.31, "learning_rate": 1.6095633719343072e-05, "loss": 2.0658, "step": 12683 }, { "epoch": 0.31, "learning_rate": 1.6095000833021726e-05, "loss": 1.9922, "step": 12684 }, { "epoch": 0.31, "learning_rate": 1.6094367907854954e-05, "loss": 2.0517, "step": 12685 }, { "epoch": 0.31, "learning_rate": 1.609373494384679e-05, "loss": 2.1824, "step": 12686 }, { "epoch": 0.31, "learning_rate": 1.609310194100127e-05, "loss": 2.0547, "step": 12687 }, { "epoch": 0.31, "learning_rate": 1.6092468899322424e-05, "loss": 2.1223, "step": 12688 }, { "epoch": 0.31, "learning_rate": 1.6091835818814285e-05, "loss": 2.1554, "step": 12689 }, { "epoch": 0.31, "learning_rate": 1.6091202699480894e-05, "loss": 2.0624, "step": 12690 }, { "epoch": 0.31, "learning_rate": 1.6090569541326285e-05, "loss": 2.2753, "step": 12691 }, { "epoch": 0.31, "learning_rate": 1.6089936344354487e-05, "loss": 1.9871, "step": 12692 }, { "epoch": 0.31, "learning_rate": 1.608930310856954e-05, "loss": 2.1372, "step": 12693 }, { "epoch": 0.31, "learning_rate": 1.6088669833975483e-05, "loss": 2.1762, "step": 12694 }, { "epoch": 0.31, "learning_rate": 1.6088036520576347e-05, "loss": 1.982, "step": 12695 }, { "epoch": 0.31, "learning_rate": 1.6087403168376173e-05, "loss": 2.2646, "step": 12696 }, { "epoch": 0.31, "learning_rate": 1.6086769777378994e-05, "loss": 1.9056, "step": 12697 }, { "epoch": 0.31, "learning_rate": 1.6086136347588844e-05, "loss": 2.1664, "step": 12698 }, { "epoch": 0.31, "learning_rate": 1.608550287900977e-05, "loss": 2.1709, "step": 12699 }, { "epoch": 0.31, "learning_rate": 1.6084869371645796e-05, "loss": 2.0472, "step": 12700 }, { "epoch": 0.31, "learning_rate": 1.608423582550097e-05, "loss": 2.1471, "step": 12701 }, { "epoch": 0.31, "learning_rate": 1.6083602240579324e-05, "loss": 2.1897, "step": 12702 }, { "epoch": 0.31, "learning_rate": 1.6082968616884898e-05, "loss": 2.1874, "step": 12703 }, { "epoch": 0.31, "learning_rate": 1.6082334954421736e-05, "loss": 2.1307, "step": 12704 }, { "epoch": 0.31, "learning_rate": 1.6081701253193864e-05, "loss": 2.2075, "step": 12705 }, { "epoch": 0.31, "learning_rate": 1.6081067513205334e-05, "loss": 2.1048, "step": 12706 }, { "epoch": 0.31, "learning_rate": 1.6080433734460178e-05, "loss": 2.2533, "step": 12707 }, { "epoch": 0.31, "learning_rate": 1.6079799916962434e-05, "loss": 2.118, "step": 12708 }, { "epoch": 0.31, "learning_rate": 1.6079166060716142e-05, "loss": 2.2423, "step": 12709 }, { "epoch": 0.31, "learning_rate": 1.6078532165725346e-05, "loss": 2.0616, "step": 12710 }, { "epoch": 0.31, "learning_rate": 1.6077898231994086e-05, "loss": 2.1489, "step": 12711 }, { "epoch": 0.31, "learning_rate": 1.60772642595264e-05, "loss": 2.0728, "step": 12712 }, { "epoch": 0.31, "learning_rate": 1.6076630248326323e-05, "loss": 2.0724, "step": 12713 }, { "epoch": 0.31, "learning_rate": 1.6075996198397904e-05, "loss": 2.3165, "step": 12714 }, { "epoch": 0.31, "learning_rate": 1.607536210974518e-05, "loss": 2.0663, "step": 12715 }, { "epoch": 0.31, "learning_rate": 1.6074727982372194e-05, "loss": 2.2592, "step": 12716 }, { "epoch": 0.31, "learning_rate": 1.6074093816282988e-05, "loss": 2.0345, "step": 12717 }, { "epoch": 0.31, "learning_rate": 1.6073459611481605e-05, "loss": 1.7983, "step": 12718 }, { "epoch": 0.31, "learning_rate": 1.607282536797208e-05, "loss": 2.0455, "step": 12719 }, { "epoch": 0.31, "learning_rate": 1.607219108575846e-05, "loss": 2.1456, "step": 12720 }, { "epoch": 0.31, "learning_rate": 1.607155676484479e-05, "loss": 2.1109, "step": 12721 }, { "epoch": 0.31, "learning_rate": 1.607092240523511e-05, "loss": 2.0847, "step": 12722 }, { "epoch": 0.31, "learning_rate": 1.607028800693346e-05, "loss": 2.0554, "step": 12723 }, { "epoch": 0.31, "learning_rate": 1.606965356994389e-05, "loss": 2.0622, "step": 12724 }, { "epoch": 0.31, "learning_rate": 1.6069019094270437e-05, "loss": 2.1332, "step": 12725 }, { "epoch": 0.31, "learning_rate": 1.606838457991715e-05, "loss": 2.1092, "step": 12726 }, { "epoch": 0.31, "learning_rate": 1.606775002688807e-05, "loss": 2.0688, "step": 12727 }, { "epoch": 0.31, "learning_rate": 1.6067115435187238e-05, "loss": 1.9793, "step": 12728 }, { "epoch": 0.31, "learning_rate": 1.6066480804818704e-05, "loss": 2.1103, "step": 12729 }, { "epoch": 0.31, "learning_rate": 1.6065846135786514e-05, "loss": 1.9934, "step": 12730 }, { "epoch": 0.31, "learning_rate": 1.6065211428094707e-05, "loss": 2.1111, "step": 12731 }, { "epoch": 0.31, "learning_rate": 1.6064576681747332e-05, "loss": 2.1122, "step": 12732 }, { "epoch": 0.31, "learning_rate": 1.606394189674843e-05, "loss": 1.9968, "step": 12733 }, { "epoch": 0.31, "learning_rate": 1.6063307073102055e-05, "loss": 1.8483, "step": 12734 }, { "epoch": 0.31, "learning_rate": 1.6062672210812248e-05, "loss": 2.0444, "step": 12735 }, { "epoch": 0.31, "learning_rate": 1.606203730988305e-05, "loss": 2.1136, "step": 12736 }, { "epoch": 0.31, "learning_rate": 1.6061402370318516e-05, "loss": 2.2846, "step": 12737 }, { "epoch": 0.31, "learning_rate": 1.606076739212269e-05, "loss": 1.9866, "step": 12738 }, { "epoch": 0.31, "learning_rate": 1.6060132375299618e-05, "loss": 2.1442, "step": 12739 }, { "epoch": 0.31, "learning_rate": 1.6059497319853345e-05, "loss": 2.0249, "step": 12740 }, { "epoch": 0.31, "learning_rate": 1.6058862225787924e-05, "loss": 2.2124, "step": 12741 }, { "epoch": 0.31, "learning_rate": 1.60582270931074e-05, "loss": 2.0907, "step": 12742 }, { "epoch": 0.31, "learning_rate": 1.6057591921815815e-05, "loss": 2.1258, "step": 12743 }, { "epoch": 0.31, "learning_rate": 1.6056956711917225e-05, "loss": 2.1334, "step": 12744 }, { "epoch": 0.31, "learning_rate": 1.6056321463415678e-05, "loss": 2.3032, "step": 12745 }, { "epoch": 0.31, "learning_rate": 1.6055686176315225e-05, "loss": 1.954, "step": 12746 }, { "epoch": 0.31, "learning_rate": 1.60550508506199e-05, "loss": 2.1125, "step": 12747 }, { "epoch": 0.31, "learning_rate": 1.605441548633377e-05, "loss": 2.1076, "step": 12748 }, { "epoch": 0.31, "learning_rate": 1.6053780083460874e-05, "loss": 2.3149, "step": 12749 }, { "epoch": 0.31, "learning_rate": 1.6053144642005266e-05, "loss": 1.9153, "step": 12750 }, { "epoch": 0.31, "learning_rate": 1.6052509161970996e-05, "loss": 2.2894, "step": 12751 }, { "epoch": 0.31, "learning_rate": 1.605187364336211e-05, "loss": 2.067, "step": 12752 }, { "epoch": 0.31, "learning_rate": 1.6051238086182662e-05, "loss": 2.1928, "step": 12753 }, { "epoch": 0.31, "learning_rate": 1.6050602490436702e-05, "loss": 2.2671, "step": 12754 }, { "epoch": 0.31, "learning_rate": 1.604996685612828e-05, "loss": 2.019, "step": 12755 }, { "epoch": 0.31, "learning_rate": 1.6049331183261447e-05, "loss": 2.2248, "step": 12756 }, { "epoch": 0.31, "learning_rate": 1.6048695471840253e-05, "loss": 2.1839, "step": 12757 }, { "epoch": 0.31, "learning_rate": 1.6048059721868756e-05, "loss": 2.2405, "step": 12758 }, { "epoch": 0.31, "learning_rate": 1.6047423933351e-05, "loss": 2.1696, "step": 12759 }, { "epoch": 0.31, "learning_rate": 1.604678810629104e-05, "loss": 2.0354, "step": 12760 }, { "epoch": 0.31, "learning_rate": 1.6046152240692934e-05, "loss": 2.2333, "step": 12761 }, { "epoch": 0.31, "learning_rate": 1.6045516336560726e-05, "loss": 2.1059, "step": 12762 }, { "epoch": 0.31, "learning_rate": 1.604488039389847e-05, "loss": 2.2874, "step": 12763 }, { "epoch": 0.31, "learning_rate": 1.6044244412710225e-05, "loss": 2.1023, "step": 12764 }, { "epoch": 0.31, "learning_rate": 1.604360839300004e-05, "loss": 2.1607, "step": 12765 }, { "epoch": 0.31, "learning_rate": 1.6042972334771967e-05, "loss": 1.9814, "step": 12766 }, { "epoch": 0.31, "learning_rate": 1.6042336238030064e-05, "loss": 2.0902, "step": 12767 }, { "epoch": 0.31, "learning_rate": 1.604170010277838e-05, "loss": 2.1568, "step": 12768 }, { "epoch": 0.31, "learning_rate": 1.604106392902098e-05, "loss": 2.0671, "step": 12769 }, { "epoch": 0.31, "learning_rate": 1.6040427716761903e-05, "loss": 2.1136, "step": 12770 }, { "epoch": 0.31, "learning_rate": 1.6039791466005216e-05, "loss": 2.1693, "step": 12771 }, { "epoch": 0.31, "learning_rate": 1.6039155176754967e-05, "loss": 2.1625, "step": 12772 }, { "epoch": 0.31, "learning_rate": 1.6038518849015213e-05, "loss": 2.0718, "step": 12773 }, { "epoch": 0.31, "learning_rate": 1.603788248279001e-05, "loss": 2.0994, "step": 12774 }, { "epoch": 0.31, "learning_rate": 1.603724607808342e-05, "loss": 2.2271, "step": 12775 }, { "epoch": 0.31, "learning_rate": 1.6036609634899488e-05, "loss": 2.084, "step": 12776 }, { "epoch": 0.31, "learning_rate": 1.603597315324228e-05, "loss": 2.2409, "step": 12777 }, { "epoch": 0.31, "learning_rate": 1.6035336633115844e-05, "loss": 2.3068, "step": 12778 }, { "epoch": 0.31, "learning_rate": 1.6034700074524243e-05, "loss": 2.1677, "step": 12779 }, { "epoch": 0.32, "learning_rate": 1.6034063477471532e-05, "loss": 2.2263, "step": 12780 }, { "epoch": 0.32, "learning_rate": 1.6033426841961764e-05, "loss": 2.1409, "step": 12781 }, { "epoch": 0.32, "learning_rate": 1.6032790167999003e-05, "loss": 2.2761, "step": 12782 }, { "epoch": 0.32, "learning_rate": 1.6032153455587306e-05, "loss": 2.0949, "step": 12783 }, { "epoch": 0.32, "learning_rate": 1.603151670473073e-05, "loss": 2.2182, "step": 12784 }, { "epoch": 0.32, "learning_rate": 1.603087991543333e-05, "loss": 2.1221, "step": 12785 }, { "epoch": 0.32, "learning_rate": 1.6030243087699164e-05, "loss": 2.1606, "step": 12786 }, { "epoch": 0.32, "learning_rate": 1.6029606221532295e-05, "loss": 2.1233, "step": 12787 }, { "epoch": 0.32, "learning_rate": 1.6028969316936785e-05, "loss": 1.903, "step": 12788 }, { "epoch": 0.32, "learning_rate": 1.6028332373916685e-05, "loss": 2.1404, "step": 12789 }, { "epoch": 0.32, "learning_rate": 1.6027695392476055e-05, "loss": 1.9903, "step": 12790 }, { "epoch": 0.32, "learning_rate": 1.6027058372618968e-05, "loss": 2.058, "step": 12791 }, { "epoch": 0.32, "learning_rate": 1.6026421314349465e-05, "loss": 2.0722, "step": 12792 }, { "epoch": 0.32, "learning_rate": 1.602578421767162e-05, "loss": 2.1993, "step": 12793 }, { "epoch": 0.32, "learning_rate": 1.602514708258949e-05, "loss": 2.2622, "step": 12794 }, { "epoch": 0.32, "learning_rate": 1.6024509909107123e-05, "loss": 2.0945, "step": 12795 }, { "epoch": 0.32, "learning_rate": 1.60238726972286e-05, "loss": 2.1293, "step": 12796 }, { "epoch": 0.32, "learning_rate": 1.6023235446957973e-05, "loss": 2.1147, "step": 12797 }, { "epoch": 0.32, "learning_rate": 1.6022598158299303e-05, "loss": 2.0724, "step": 12798 }, { "epoch": 0.32, "learning_rate": 1.602196083125665e-05, "loss": 2.0637, "step": 12799 }, { "epoch": 0.32, "learning_rate": 1.602132346583408e-05, "loss": 1.9484, "step": 12800 }, { "epoch": 0.32, "learning_rate": 1.6020686062035652e-05, "loss": 2.1639, "step": 12801 }, { "epoch": 0.32, "learning_rate": 1.602004861986543e-05, "loss": 2.1802, "step": 12802 }, { "epoch": 0.32, "learning_rate": 1.601941113932748e-05, "loss": 2.1363, "step": 12803 }, { "epoch": 0.32, "learning_rate": 1.601877362042586e-05, "loss": 2.0515, "step": 12804 }, { "epoch": 0.32, "learning_rate": 1.601813606316463e-05, "loss": 2.0101, "step": 12805 }, { "epoch": 0.32, "learning_rate": 1.601749846754786e-05, "loss": 2.1492, "step": 12806 }, { "epoch": 0.32, "learning_rate": 1.601686083357961e-05, "loss": 2.1171, "step": 12807 }, { "epoch": 0.32, "learning_rate": 1.6016223161263947e-05, "loss": 2.1566, "step": 12808 }, { "epoch": 0.32, "learning_rate": 1.601558545060493e-05, "loss": 2.224, "step": 12809 }, { "epoch": 0.32, "learning_rate": 1.601494770160663e-05, "loss": 2.023, "step": 12810 }, { "epoch": 0.32, "learning_rate": 1.6014309914273107e-05, "loss": 2.0865, "step": 12811 }, { "epoch": 0.32, "learning_rate": 1.6013672088608427e-05, "loss": 2.1613, "step": 12812 }, { "epoch": 0.32, "learning_rate": 1.6013034224616655e-05, "loss": 2.1413, "step": 12813 }, { "epoch": 0.32, "learning_rate": 1.601239632230186e-05, "loss": 2.102, "step": 12814 }, { "epoch": 0.32, "learning_rate": 1.6011758381668096e-05, "loss": 2.0565, "step": 12815 }, { "epoch": 0.32, "learning_rate": 1.6011120402719438e-05, "loss": 2.1029, "step": 12816 }, { "epoch": 0.32, "learning_rate": 1.6010482385459957e-05, "loss": 2.0767, "step": 12817 }, { "epoch": 0.32, "learning_rate": 1.6009844329893706e-05, "loss": 2.0743, "step": 12818 }, { "epoch": 0.32, "learning_rate": 1.6009206236024764e-05, "loss": 2.0886, "step": 12819 }, { "epoch": 0.32, "learning_rate": 1.600856810385719e-05, "loss": 2.1721, "step": 12820 }, { "epoch": 0.32, "learning_rate": 1.6007929933395053e-05, "loss": 2.1875, "step": 12821 }, { "epoch": 0.32, "learning_rate": 1.600729172464242e-05, "loss": 2.1976, "step": 12822 }, { "epoch": 0.32, "learning_rate": 1.600665347760336e-05, "loss": 2.1221, "step": 12823 }, { "epoch": 0.32, "learning_rate": 1.600601519228194e-05, "loss": 1.9398, "step": 12824 }, { "epoch": 0.32, "learning_rate": 1.6005376868682227e-05, "loss": 2.1958, "step": 12825 }, { "epoch": 0.32, "learning_rate": 1.600473850680829e-05, "loss": 2.1437, "step": 12826 }, { "epoch": 0.32, "learning_rate": 1.60041001066642e-05, "loss": 2.245, "step": 12827 }, { "epoch": 0.32, "learning_rate": 1.6003461668254023e-05, "loss": 2.0715, "step": 12828 }, { "epoch": 0.32, "learning_rate": 1.6002823191581826e-05, "loss": 2.2918, "step": 12829 }, { "epoch": 0.32, "learning_rate": 1.600218467665168e-05, "loss": 2.1629, "step": 12830 }, { "epoch": 0.32, "learning_rate": 1.600154612346766e-05, "loss": 2.0945, "step": 12831 }, { "epoch": 0.32, "learning_rate": 1.6000907532033826e-05, "loss": 2.1617, "step": 12832 }, { "epoch": 0.32, "learning_rate": 1.6000268902354257e-05, "loss": 2.1528, "step": 12833 }, { "epoch": 0.32, "learning_rate": 1.5999630234433015e-05, "loss": 2.1655, "step": 12834 }, { "epoch": 0.32, "learning_rate": 1.5998991528274174e-05, "loss": 2.1458, "step": 12835 }, { "epoch": 0.32, "learning_rate": 1.5998352783881807e-05, "loss": 2.2054, "step": 12836 }, { "epoch": 0.32, "learning_rate": 1.5997714001259985e-05, "loss": 2.1122, "step": 12837 }, { "epoch": 0.32, "learning_rate": 1.599707518041277e-05, "loss": 2.0738, "step": 12838 }, { "epoch": 0.32, "learning_rate": 1.5996436321344246e-05, "loss": 2.1358, "step": 12839 }, { "epoch": 0.32, "learning_rate": 1.599579742405848e-05, "loss": 1.9927, "step": 12840 }, { "epoch": 0.32, "learning_rate": 1.5995158488559538e-05, "loss": 2.0046, "step": 12841 }, { "epoch": 0.32, "learning_rate": 1.59945195148515e-05, "loss": 2.0526, "step": 12842 }, { "epoch": 0.32, "learning_rate": 1.5993880502938438e-05, "loss": 2.0541, "step": 12843 }, { "epoch": 0.32, "learning_rate": 1.599324145282442e-05, "loss": 2.1181, "step": 12844 }, { "epoch": 0.32, "learning_rate": 1.599260236451352e-05, "loss": 2.3109, "step": 12845 }, { "epoch": 0.32, "learning_rate": 1.5991963238009815e-05, "loss": 2.0353, "step": 12846 }, { "epoch": 0.32, "learning_rate": 1.5991324073317372e-05, "loss": 2.0183, "step": 12847 }, { "epoch": 0.32, "learning_rate": 1.5990684870440268e-05, "loss": 2.1827, "step": 12848 }, { "epoch": 0.32, "learning_rate": 1.599004562938258e-05, "loss": 2.1556, "step": 12849 }, { "epoch": 0.32, "learning_rate": 1.598940635014838e-05, "loss": 2.2225, "step": 12850 }, { "epoch": 0.32, "learning_rate": 1.5988767032741738e-05, "loss": 2.126, "step": 12851 }, { "epoch": 0.32, "learning_rate": 1.5988127677166735e-05, "loss": 2.2616, "step": 12852 }, { "epoch": 0.32, "learning_rate": 1.598748828342744e-05, "loss": 2.0262, "step": 12853 }, { "epoch": 0.32, "learning_rate": 1.5986848851527933e-05, "loss": 2.1987, "step": 12854 }, { "epoch": 0.32, "learning_rate": 1.5986209381472283e-05, "loss": 2.1982, "step": 12855 }, { "epoch": 0.32, "learning_rate": 1.5985569873264575e-05, "loss": 2.1367, "step": 12856 }, { "epoch": 0.32, "learning_rate": 1.598493032690888e-05, "loss": 2.084, "step": 12857 }, { "epoch": 0.32, "learning_rate": 1.598429074240927e-05, "loss": 1.9914, "step": 12858 }, { "epoch": 0.32, "learning_rate": 1.598365111976983e-05, "loss": 1.9287, "step": 12859 }, { "epoch": 0.32, "learning_rate": 1.598301145899462e-05, "loss": 2.0186, "step": 12860 }, { "epoch": 0.32, "learning_rate": 1.598237176008774e-05, "loss": 2.1259, "step": 12861 }, { "epoch": 0.32, "learning_rate": 1.5981732023053253e-05, "loss": 1.8084, "step": 12862 }, { "epoch": 0.32, "learning_rate": 1.5981092247895234e-05, "loss": 2.2636, "step": 12863 }, { "epoch": 0.32, "learning_rate": 1.5980452434617765e-05, "loss": 2.1212, "step": 12864 }, { "epoch": 0.32, "learning_rate": 1.5979812583224927e-05, "loss": 2.2439, "step": 12865 }, { "epoch": 0.32, "learning_rate": 1.5979172693720795e-05, "loss": 1.9899, "step": 12866 }, { "epoch": 0.32, "learning_rate": 1.5978532766109446e-05, "loss": 1.9424, "step": 12867 }, { "epoch": 0.32, "learning_rate": 1.5977892800394957e-05, "loss": 2.1713, "step": 12868 }, { "epoch": 0.32, "learning_rate": 1.5977252796581414e-05, "loss": 2.0822, "step": 12869 }, { "epoch": 0.32, "learning_rate": 1.5976612754672886e-05, "loss": 2.0354, "step": 12870 }, { "epoch": 0.32, "learning_rate": 1.597597267467346e-05, "loss": 2.1282, "step": 12871 }, { "epoch": 0.32, "learning_rate": 1.597533255658721e-05, "loss": 2.0642, "step": 12872 }, { "epoch": 0.32, "learning_rate": 1.597469240041822e-05, "loss": 2.2567, "step": 12873 }, { "epoch": 0.32, "learning_rate": 1.597405220617057e-05, "loss": 2.017, "step": 12874 }, { "epoch": 0.32, "learning_rate": 1.5973411973848337e-05, "loss": 1.9789, "step": 12875 }, { "epoch": 0.32, "learning_rate": 1.5972771703455602e-05, "loss": 2.1072, "step": 12876 }, { "epoch": 0.32, "learning_rate": 1.5972131394996445e-05, "loss": 2.2552, "step": 12877 }, { "epoch": 0.32, "learning_rate": 1.597149104847495e-05, "loss": 1.9818, "step": 12878 }, { "epoch": 0.32, "learning_rate": 1.5970850663895198e-05, "loss": 2.2491, "step": 12879 }, { "epoch": 0.32, "learning_rate": 1.5970210241261268e-05, "loss": 2.0543, "step": 12880 }, { "epoch": 0.32, "learning_rate": 1.596956978057724e-05, "loss": 2.2121, "step": 12881 }, { "epoch": 0.32, "learning_rate": 1.59689292818472e-05, "loss": 2.179, "step": 12882 }, { "epoch": 0.32, "learning_rate": 1.596828874507523e-05, "loss": 2.0658, "step": 12883 }, { "epoch": 0.32, "learning_rate": 1.596764817026541e-05, "loss": 2.2855, "step": 12884 }, { "epoch": 0.32, "learning_rate": 1.5967007557421822e-05, "loss": 1.983, "step": 12885 }, { "epoch": 0.32, "learning_rate": 1.5966366906548552e-05, "loss": 2.033, "step": 12886 }, { "epoch": 0.32, "learning_rate": 1.5965726217649683e-05, "loss": 2.0729, "step": 12887 }, { "epoch": 0.32, "learning_rate": 1.5965085490729293e-05, "loss": 2.0654, "step": 12888 }, { "epoch": 0.32, "learning_rate": 1.596444472579147e-05, "loss": 2.1816, "step": 12889 }, { "epoch": 0.32, "learning_rate": 1.5963803922840298e-05, "loss": 1.9665, "step": 12890 }, { "epoch": 0.32, "learning_rate": 1.596316308187986e-05, "loss": 2.2966, "step": 12891 }, { "epoch": 0.32, "learning_rate": 1.596252220291424e-05, "loss": 2.0173, "step": 12892 }, { "epoch": 0.32, "learning_rate": 1.5961881285947523e-05, "loss": 2.0459, "step": 12893 }, { "epoch": 0.32, "learning_rate": 1.5961240330983794e-05, "loss": 2.0901, "step": 12894 }, { "epoch": 0.32, "learning_rate": 1.5960599338027137e-05, "loss": 2.1379, "step": 12895 }, { "epoch": 0.32, "learning_rate": 1.595995830708164e-05, "loss": 2.2458, "step": 12896 }, { "epoch": 0.32, "learning_rate": 1.595931723815138e-05, "loss": 2.1787, "step": 12897 }, { "epoch": 0.32, "learning_rate": 1.5958676131240455e-05, "loss": 2.2038, "step": 12898 }, { "epoch": 0.32, "learning_rate": 1.5958034986352945e-05, "loss": 2.0927, "step": 12899 }, { "epoch": 0.32, "learning_rate": 1.5957393803492935e-05, "loss": 1.8909, "step": 12900 }, { "epoch": 0.32, "learning_rate": 1.5956752582664513e-05, "loss": 2.1961, "step": 12901 }, { "epoch": 0.32, "learning_rate": 1.5956111323871763e-05, "loss": 2.0797, "step": 12902 }, { "epoch": 0.32, "learning_rate": 1.595547002711878e-05, "loss": 2.1446, "step": 12903 }, { "epoch": 0.32, "learning_rate": 1.5954828692409643e-05, "loss": 2.1066, "step": 12904 }, { "epoch": 0.32, "learning_rate": 1.595418731974844e-05, "loss": 2.1154, "step": 12905 }, { "epoch": 0.32, "learning_rate": 1.5953545909139262e-05, "loss": 2.2496, "step": 12906 }, { "epoch": 0.32, "learning_rate": 1.59529044605862e-05, "loss": 2.0854, "step": 12907 }, { "epoch": 0.32, "learning_rate": 1.595226297409333e-05, "loss": 2.004, "step": 12908 }, { "epoch": 0.32, "learning_rate": 1.5951621449664756e-05, "loss": 2.127, "step": 12909 }, { "epoch": 0.32, "learning_rate": 1.5950979887304556e-05, "loss": 2.2756, "step": 12910 }, { "epoch": 0.32, "learning_rate": 1.595033828701682e-05, "loss": 1.9985, "step": 12911 }, { "epoch": 0.32, "learning_rate": 1.5949696648805638e-05, "loss": 2.1458, "step": 12912 }, { "epoch": 0.32, "learning_rate": 1.5949054972675103e-05, "loss": 2.2838, "step": 12913 }, { "epoch": 0.32, "learning_rate": 1.5948413258629303e-05, "loss": 1.9851, "step": 12914 }, { "epoch": 0.32, "learning_rate": 1.5947771506672325e-05, "loss": 2.2673, "step": 12915 }, { "epoch": 0.32, "learning_rate": 1.594712971680826e-05, "loss": 2.1714, "step": 12916 }, { "epoch": 0.32, "learning_rate": 1.59464878890412e-05, "loss": 2.0439, "step": 12917 }, { "epoch": 0.32, "learning_rate": 1.5945846023375234e-05, "loss": 2.1105, "step": 12918 }, { "epoch": 0.32, "learning_rate": 1.5945204119814456e-05, "loss": 2.0886, "step": 12919 }, { "epoch": 0.32, "learning_rate": 1.5944562178362956e-05, "loss": 1.9599, "step": 12920 }, { "epoch": 0.32, "learning_rate": 1.5943920199024817e-05, "loss": 2.1745, "step": 12921 }, { "epoch": 0.32, "learning_rate": 1.5943278181804143e-05, "loss": 1.9853, "step": 12922 }, { "epoch": 0.32, "learning_rate": 1.594263612670502e-05, "loss": 2.1825, "step": 12923 }, { "epoch": 0.32, "learning_rate": 1.594199403373154e-05, "loss": 2.2113, "step": 12924 }, { "epoch": 0.32, "learning_rate": 1.5941351902887793e-05, "loss": 2.155, "step": 12925 }, { "epoch": 0.32, "learning_rate": 1.5940709734177878e-05, "loss": 2.1485, "step": 12926 }, { "epoch": 0.32, "learning_rate": 1.594006752760588e-05, "loss": 2.0544, "step": 12927 }, { "epoch": 0.32, "learning_rate": 1.59394252831759e-05, "loss": 2.0481, "step": 12928 }, { "epoch": 0.32, "learning_rate": 1.593878300089202e-05, "loss": 2.0583, "step": 12929 }, { "epoch": 0.32, "learning_rate": 1.593814068075835e-05, "loss": 2.2775, "step": 12930 }, { "epoch": 0.32, "learning_rate": 1.593749832277897e-05, "loss": 2.2838, "step": 12931 }, { "epoch": 0.32, "learning_rate": 1.5936855926957977e-05, "loss": 2.0056, "step": 12932 }, { "epoch": 0.32, "learning_rate": 1.5936213493299467e-05, "loss": 2.1464, "step": 12933 }, { "epoch": 0.32, "learning_rate": 1.5935571021807535e-05, "loss": 2.1918, "step": 12934 }, { "epoch": 0.32, "learning_rate": 1.5934928512486272e-05, "loss": 2.2073, "step": 12935 }, { "epoch": 0.32, "learning_rate": 1.5934285965339777e-05, "loss": 1.9341, "step": 12936 }, { "epoch": 0.32, "learning_rate": 1.5933643380372143e-05, "loss": 2.164, "step": 12937 }, { "epoch": 0.32, "learning_rate": 1.5933000757587465e-05, "loss": 2.1378, "step": 12938 }, { "epoch": 0.32, "learning_rate": 1.5932358096989845e-05, "loss": 1.9994, "step": 12939 }, { "epoch": 0.32, "learning_rate": 1.5931715398583367e-05, "loss": 2.2027, "step": 12940 }, { "epoch": 0.32, "learning_rate": 1.5931072662372135e-05, "loss": 2.2058, "step": 12941 }, { "epoch": 0.32, "learning_rate": 1.593042988836025e-05, "loss": 2.1959, "step": 12942 }, { "epoch": 0.32, "learning_rate": 1.5929787076551794e-05, "loss": 2.1472, "step": 12943 }, { "epoch": 0.32, "learning_rate": 1.5929144226950877e-05, "loss": 2.0129, "step": 12944 }, { "epoch": 0.32, "learning_rate": 1.592850133956159e-05, "loss": 2.2634, "step": 12945 }, { "epoch": 0.32, "learning_rate": 1.592785841438804e-05, "loss": 2.1793, "step": 12946 }, { "epoch": 0.32, "learning_rate": 1.5927215451434307e-05, "loss": 2.0525, "step": 12947 }, { "epoch": 0.32, "learning_rate": 1.59265724507045e-05, "loss": 2.3087, "step": 12948 }, { "epoch": 0.32, "learning_rate": 1.5925929412202718e-05, "loss": 2.2, "step": 12949 }, { "epoch": 0.32, "learning_rate": 1.5925286335933053e-05, "loss": 2.1431, "step": 12950 }, { "epoch": 0.32, "learning_rate": 1.592464322189961e-05, "loss": 2.0988, "step": 12951 }, { "epoch": 0.32, "learning_rate": 1.5924000070106483e-05, "loss": 1.9858, "step": 12952 }, { "epoch": 0.32, "learning_rate": 1.5923356880557776e-05, "loss": 2.1861, "step": 12953 }, { "epoch": 0.32, "learning_rate": 1.5922713653257584e-05, "loss": 2.2036, "step": 12954 }, { "epoch": 0.32, "learning_rate": 1.5922070388210005e-05, "loss": 2.0129, "step": 12955 }, { "epoch": 0.32, "learning_rate": 1.5921427085419143e-05, "loss": 2.1794, "step": 12956 }, { "epoch": 0.32, "learning_rate": 1.5920783744889097e-05, "loss": 2.189, "step": 12957 }, { "epoch": 0.32, "learning_rate": 1.5920140366623967e-05, "loss": 2.3093, "step": 12958 }, { "epoch": 0.32, "learning_rate": 1.5919496950627856e-05, "loss": 2.3118, "step": 12959 }, { "epoch": 0.32, "learning_rate": 1.5918853496904857e-05, "loss": 2.1232, "step": 12960 }, { "epoch": 0.32, "learning_rate": 1.5918210005459077e-05, "loss": 1.9155, "step": 12961 }, { "epoch": 0.32, "learning_rate": 1.5917566476294617e-05, "loss": 2.0812, "step": 12962 }, { "epoch": 0.32, "learning_rate": 1.591692290941558e-05, "loss": 2.0514, "step": 12963 }, { "epoch": 0.32, "learning_rate": 1.5916279304826058e-05, "loss": 2.0188, "step": 12964 }, { "epoch": 0.32, "learning_rate": 1.5915635662530164e-05, "loss": 1.9792, "step": 12965 }, { "epoch": 0.32, "learning_rate": 1.5914991982531996e-05, "loss": 2.1196, "step": 12966 }, { "epoch": 0.32, "learning_rate": 1.5914348264835658e-05, "loss": 2.2254, "step": 12967 }, { "epoch": 0.32, "learning_rate": 1.591370450944525e-05, "loss": 2.0328, "step": 12968 }, { "epoch": 0.32, "learning_rate": 1.5913060716364873e-05, "loss": 2.0317, "step": 12969 }, { "epoch": 0.32, "learning_rate": 1.5912416885598633e-05, "loss": 2.2192, "step": 12970 }, { "epoch": 0.32, "learning_rate": 1.5911773017150637e-05, "loss": 2.1705, "step": 12971 }, { "epoch": 0.32, "learning_rate": 1.5911129111024983e-05, "loss": 2.0009, "step": 12972 }, { "epoch": 0.32, "learning_rate": 1.591048516722578e-05, "loss": 2.2249, "step": 12973 }, { "epoch": 0.32, "learning_rate": 1.590984118575712e-05, "loss": 2.1241, "step": 12974 }, { "epoch": 0.32, "learning_rate": 1.5909197166623126e-05, "loss": 2.1914, "step": 12975 }, { "epoch": 0.32, "learning_rate": 1.5908553109827888e-05, "loss": 2.1249, "step": 12976 }, { "epoch": 0.32, "learning_rate": 1.590790901537552e-05, "loss": 1.9256, "step": 12977 }, { "epoch": 0.32, "learning_rate": 1.590726488327012e-05, "loss": 2.1195, "step": 12978 }, { "epoch": 0.32, "learning_rate": 1.5906620713515793e-05, "loss": 2.1889, "step": 12979 }, { "epoch": 0.32, "learning_rate": 1.590597650611665e-05, "loss": 2.1179, "step": 12980 }, { "epoch": 0.32, "learning_rate": 1.590533226107679e-05, "loss": 2.1271, "step": 12981 }, { "epoch": 0.32, "learning_rate": 1.590468797840033e-05, "loss": 2.226, "step": 12982 }, { "epoch": 0.32, "learning_rate": 1.5904043658091362e-05, "loss": 2.1972, "step": 12983 }, { "epoch": 0.32, "learning_rate": 1.5903399300154004e-05, "loss": 2.0987, "step": 12984 }, { "epoch": 0.32, "learning_rate": 1.590275490459236e-05, "loss": 2.226, "step": 12985 }, { "epoch": 0.32, "learning_rate": 1.590211047141053e-05, "loss": 2.1892, "step": 12986 }, { "epoch": 0.32, "learning_rate": 1.5901466000612632e-05, "loss": 2.1361, "step": 12987 }, { "epoch": 0.32, "learning_rate": 1.5900821492202764e-05, "loss": 2.1924, "step": 12988 }, { "epoch": 0.32, "learning_rate": 1.5900176946185037e-05, "loss": 2.1219, "step": 12989 }, { "epoch": 0.32, "learning_rate": 1.5899532362563563e-05, "loss": 2.0904, "step": 12990 }, { "epoch": 0.32, "learning_rate": 1.5898887741342445e-05, "loss": 2.026, "step": 12991 }, { "epoch": 0.32, "learning_rate": 1.5898243082525794e-05, "loss": 2.1378, "step": 12992 }, { "epoch": 0.32, "learning_rate": 1.589759838611772e-05, "loss": 2.1245, "step": 12993 }, { "epoch": 0.32, "learning_rate": 1.5896953652122323e-05, "loss": 2.1144, "step": 12994 }, { "epoch": 0.32, "learning_rate": 1.5896308880543723e-05, "loss": 1.9374, "step": 12995 }, { "epoch": 0.32, "learning_rate": 1.589566407138602e-05, "loss": 2.0247, "step": 12996 }, { "epoch": 0.32, "learning_rate": 1.5895019224653336e-05, "loss": 2.252, "step": 12997 }, { "epoch": 0.32, "learning_rate": 1.5894374340349768e-05, "loss": 2.1319, "step": 12998 }, { "epoch": 0.32, "learning_rate": 1.5893729418479432e-05, "loss": 2.0581, "step": 12999 }, { "epoch": 0.32, "learning_rate": 1.589308445904644e-05, "loss": 1.9784, "step": 13000 }, { "epoch": 0.32, "learning_rate": 1.58924394620549e-05, "loss": 2.0877, "step": 13001 }, { "epoch": 0.32, "learning_rate": 1.5891794427508926e-05, "loss": 2.2221, "step": 13002 }, { "epoch": 0.32, "learning_rate": 1.5891149355412617e-05, "loss": 2.1314, "step": 13003 }, { "epoch": 0.32, "learning_rate": 1.58905042457701e-05, "loss": 2.0088, "step": 13004 }, { "epoch": 0.32, "learning_rate": 1.588985909858548e-05, "loss": 2.1198, "step": 13005 }, { "epoch": 0.32, "learning_rate": 1.5889213913862867e-05, "loss": 2.085, "step": 13006 }, { "epoch": 0.32, "learning_rate": 1.5888568691606374e-05, "loss": 2.1186, "step": 13007 }, { "epoch": 0.32, "learning_rate": 1.588792343182011e-05, "loss": 1.977, "step": 13008 }, { "epoch": 0.32, "learning_rate": 1.5887278134508198e-05, "loss": 2.0914, "step": 13009 }, { "epoch": 0.32, "learning_rate": 1.588663279967474e-05, "loss": 2.0613, "step": 13010 }, { "epoch": 0.32, "learning_rate": 1.5885987427323857e-05, "loss": 2.2059, "step": 13011 }, { "epoch": 0.32, "learning_rate": 1.5885342017459653e-05, "loss": 2.1064, "step": 13012 }, { "epoch": 0.32, "learning_rate": 1.5884696570086248e-05, "loss": 2.4661, "step": 13013 }, { "epoch": 0.32, "learning_rate": 1.5884051085207756e-05, "loss": 2.1544, "step": 13014 }, { "epoch": 0.32, "learning_rate": 1.5883405562828285e-05, "loss": 2.134, "step": 13015 }, { "epoch": 0.32, "learning_rate": 1.5882760002951954e-05, "loss": 2.2612, "step": 13016 }, { "epoch": 0.32, "learning_rate": 1.588211440558288e-05, "loss": 2.1718, "step": 13017 }, { "epoch": 0.32, "learning_rate": 1.5881468770725168e-05, "loss": 1.9591, "step": 13018 }, { "epoch": 0.32, "learning_rate": 1.5880823098382943e-05, "loss": 2.2691, "step": 13019 }, { "epoch": 0.32, "learning_rate": 1.5880177388560313e-05, "loss": 2.15, "step": 13020 }, { "epoch": 0.32, "learning_rate": 1.58795316412614e-05, "loss": 2.0646, "step": 13021 }, { "epoch": 0.32, "learning_rate": 1.5878885856490314e-05, "loss": 2.0661, "step": 13022 }, { "epoch": 0.32, "learning_rate": 1.5878240034251172e-05, "loss": 2.1854, "step": 13023 }, { "epoch": 0.32, "learning_rate": 1.5877594174548092e-05, "loss": 1.994, "step": 13024 }, { "epoch": 0.32, "learning_rate": 1.5876948277385186e-05, "loss": 2.2305, "step": 13025 }, { "epoch": 0.32, "learning_rate": 1.5876302342766578e-05, "loss": 2.3073, "step": 13026 }, { "epoch": 0.32, "learning_rate": 1.5875656370696374e-05, "loss": 2.1699, "step": 13027 }, { "epoch": 0.32, "learning_rate": 1.58750103611787e-05, "loss": 2.268, "step": 13028 }, { "epoch": 0.32, "learning_rate": 1.5874364314217673e-05, "loss": 2.0907, "step": 13029 }, { "epoch": 0.32, "learning_rate": 1.587371822981741e-05, "loss": 2.1176, "step": 13030 }, { "epoch": 0.32, "learning_rate": 1.5873072107982016e-05, "loss": 2.0055, "step": 13031 }, { "epoch": 0.32, "learning_rate": 1.5872425948715626e-05, "loss": 2.0754, "step": 13032 }, { "epoch": 0.32, "learning_rate": 1.5871779752022355e-05, "loss": 2.1563, "step": 13033 }, { "epoch": 0.32, "learning_rate": 1.5871133517906313e-05, "loss": 2.1875, "step": 13034 }, { "epoch": 0.32, "learning_rate": 1.5870487246371626e-05, "loss": 2.1535, "step": 13035 }, { "epoch": 0.32, "learning_rate": 1.586984093742241e-05, "loss": 2.1785, "step": 13036 }, { "epoch": 0.32, "learning_rate": 1.5869194591062783e-05, "loss": 2.1415, "step": 13037 }, { "epoch": 0.32, "learning_rate": 1.5868548207296867e-05, "loss": 2.052, "step": 13038 }, { "epoch": 0.32, "learning_rate": 1.5867901786128785e-05, "loss": 2.0182, "step": 13039 }, { "epoch": 0.32, "learning_rate": 1.5867255327562647e-05, "loss": 2.1025, "step": 13040 }, { "epoch": 0.32, "learning_rate": 1.5866608831602582e-05, "loss": 2.0273, "step": 13041 }, { "epoch": 0.32, "learning_rate": 1.5865962298252708e-05, "loss": 2.1282, "step": 13042 }, { "epoch": 0.32, "learning_rate": 1.586531572751714e-05, "loss": 2.1603, "step": 13043 }, { "epoch": 0.32, "learning_rate": 1.5864669119400006e-05, "loss": 2.0947, "step": 13044 }, { "epoch": 0.32, "learning_rate": 1.5864022473905426e-05, "loss": 2.0566, "step": 13045 }, { "epoch": 0.32, "learning_rate": 1.5863375791037516e-05, "loss": 2.1492, "step": 13046 }, { "epoch": 0.32, "learning_rate": 1.5862729070800403e-05, "loss": 2.1557, "step": 13047 }, { "epoch": 0.32, "learning_rate": 1.5862082313198207e-05, "loss": 2.1501, "step": 13048 }, { "epoch": 0.32, "learning_rate": 1.586143551823505e-05, "loss": 2.2893, "step": 13049 }, { "epoch": 0.32, "learning_rate": 1.5860788685915056e-05, "loss": 2.0111, "step": 13050 }, { "epoch": 0.32, "learning_rate": 1.5860141816242347e-05, "loss": 2.0637, "step": 13051 }, { "epoch": 0.32, "learning_rate": 1.585949490922104e-05, "loss": 2.041, "step": 13052 }, { "epoch": 0.32, "learning_rate": 1.5858847964855264e-05, "loss": 2.0673, "step": 13053 }, { "epoch": 0.32, "learning_rate": 1.5858200983149143e-05, "loss": 2.3754, "step": 13054 }, { "epoch": 0.32, "learning_rate": 1.5857553964106795e-05, "loss": 2.0633, "step": 13055 }, { "epoch": 0.32, "learning_rate": 1.585690690773235e-05, "loss": 2.0578, "step": 13056 }, { "epoch": 0.32, "learning_rate": 1.5856259814029925e-05, "loss": 2.2561, "step": 13057 }, { "epoch": 0.32, "learning_rate": 1.585561268300365e-05, "loss": 2.0453, "step": 13058 }, { "epoch": 0.32, "learning_rate": 1.5854965514657643e-05, "loss": 2.2595, "step": 13059 }, { "epoch": 0.32, "learning_rate": 1.5854318308996038e-05, "loss": 2.2269, "step": 13060 }, { "epoch": 0.32, "learning_rate": 1.585367106602295e-05, "loss": 2.0623, "step": 13061 }, { "epoch": 0.32, "learning_rate": 1.5853023785742512e-05, "loss": 2.1662, "step": 13062 }, { "epoch": 0.32, "learning_rate": 1.5852376468158842e-05, "loss": 2.169, "step": 13063 }, { "epoch": 0.32, "learning_rate": 1.5851729113276073e-05, "loss": 2.005, "step": 13064 }, { "epoch": 0.32, "learning_rate": 1.5851081721098328e-05, "loss": 2.2006, "step": 13065 }, { "epoch": 0.32, "learning_rate": 1.585043429162973e-05, "loss": 2.371, "step": 13066 }, { "epoch": 0.32, "learning_rate": 1.584978682487441e-05, "loss": 2.2159, "step": 13067 }, { "epoch": 0.32, "learning_rate": 1.584913932083649e-05, "loss": 2.0984, "step": 13068 }, { "epoch": 0.32, "learning_rate": 1.58484917795201e-05, "loss": 2.0356, "step": 13069 }, { "epoch": 0.32, "learning_rate": 1.5847844200929366e-05, "loss": 2.0784, "step": 13070 }, { "epoch": 0.32, "learning_rate": 1.5847196585068413e-05, "loss": 2.0507, "step": 13071 }, { "epoch": 0.32, "learning_rate": 1.5846548931941376e-05, "loss": 2.2483, "step": 13072 }, { "epoch": 0.32, "learning_rate": 1.5845901241552373e-05, "loss": 2.2215, "step": 13073 }, { "epoch": 0.32, "learning_rate": 1.5845253513905535e-05, "loss": 2.0592, "step": 13074 }, { "epoch": 0.32, "learning_rate": 1.5844605749004995e-05, "loss": 2.2402, "step": 13075 }, { "epoch": 0.32, "learning_rate": 1.5843957946854874e-05, "loss": 2.1419, "step": 13076 }, { "epoch": 0.32, "learning_rate": 1.5843310107459308e-05, "loss": 2.198, "step": 13077 }, { "epoch": 0.32, "learning_rate": 1.584266223082242e-05, "loss": 2.0461, "step": 13078 }, { "epoch": 0.32, "learning_rate": 1.5842014316948345e-05, "loss": 2.0517, "step": 13079 }, { "epoch": 0.32, "learning_rate": 1.5841366365841205e-05, "loss": 2.2801, "step": 13080 }, { "epoch": 0.32, "learning_rate": 1.5840718377505137e-05, "loss": 2.3103, "step": 13081 }, { "epoch": 0.32, "learning_rate": 1.5840070351944265e-05, "loss": 2.1063, "step": 13082 }, { "epoch": 0.32, "learning_rate": 1.5839422289162723e-05, "loss": 2.0978, "step": 13083 }, { "epoch": 0.32, "learning_rate": 1.583877418916464e-05, "loss": 2.032, "step": 13084 }, { "epoch": 0.32, "learning_rate": 1.5838126051954143e-05, "loss": 2.1824, "step": 13085 }, { "epoch": 0.32, "learning_rate": 1.583747787753537e-05, "loss": 2.1935, "step": 13086 }, { "epoch": 0.32, "learning_rate": 1.5836829665912444e-05, "loss": 2.2406, "step": 13087 }, { "epoch": 0.32, "learning_rate": 1.5836181417089503e-05, "loss": 2.2577, "step": 13088 }, { "epoch": 0.32, "learning_rate": 1.5835533131070676e-05, "loss": 2.1467, "step": 13089 }, { "epoch": 0.32, "learning_rate": 1.5834884807860095e-05, "loss": 2.1742, "step": 13090 }, { "epoch": 0.32, "learning_rate": 1.583423644746189e-05, "loss": 2.1922, "step": 13091 }, { "epoch": 0.32, "learning_rate": 1.5833588049880196e-05, "loss": 2.1639, "step": 13092 }, { "epoch": 0.32, "learning_rate": 1.5832939615119144e-05, "loss": 2.1543, "step": 13093 }, { "epoch": 0.32, "learning_rate": 1.5832291143182866e-05, "loss": 1.9576, "step": 13094 }, { "epoch": 0.32, "learning_rate": 1.5831642634075496e-05, "loss": 2.1828, "step": 13095 }, { "epoch": 0.32, "learning_rate": 1.5830994087801167e-05, "loss": 1.9523, "step": 13096 }, { "epoch": 0.32, "learning_rate": 1.583034550436401e-05, "loss": 2.01, "step": 13097 }, { "epoch": 0.32, "learning_rate": 1.5829696883768167e-05, "loss": 2.1741, "step": 13098 }, { "epoch": 0.32, "learning_rate": 1.5829048226017764e-05, "loss": 2.1642, "step": 13099 }, { "epoch": 0.32, "learning_rate": 1.5828399531116932e-05, "loss": 1.8491, "step": 13100 }, { "epoch": 0.32, "learning_rate": 1.5827750799069816e-05, "loss": 2.0707, "step": 13101 }, { "epoch": 0.32, "learning_rate": 1.5827102029880546e-05, "loss": 2.061, "step": 13102 }, { "epoch": 0.32, "learning_rate": 1.582645322355325e-05, "loss": 2.0691, "step": 13103 }, { "epoch": 0.32, "learning_rate": 1.5825804380092074e-05, "loss": 1.9432, "step": 13104 }, { "epoch": 0.32, "learning_rate": 1.5825155499501142e-05, "loss": 2.2327, "step": 13105 }, { "epoch": 0.32, "learning_rate": 1.58245065817846e-05, "loss": 2.2465, "step": 13106 }, { "epoch": 0.32, "learning_rate": 1.5823857626946583e-05, "loss": 1.9846, "step": 13107 }, { "epoch": 0.32, "learning_rate": 1.582320863499122e-05, "loss": 2.4091, "step": 13108 }, { "epoch": 0.32, "learning_rate": 1.5822559605922648e-05, "loss": 2.2591, "step": 13109 }, { "epoch": 0.32, "learning_rate": 1.5821910539745006e-05, "loss": 1.9628, "step": 13110 }, { "epoch": 0.32, "learning_rate": 1.5821261436462434e-05, "loss": 2.0699, "step": 13111 }, { "epoch": 0.32, "learning_rate": 1.5820612296079064e-05, "loss": 2.2192, "step": 13112 }, { "epoch": 0.32, "learning_rate": 1.581996311859904e-05, "loss": 2.0342, "step": 13113 }, { "epoch": 0.32, "learning_rate": 1.581931390402649e-05, "loss": 2.0053, "step": 13114 }, { "epoch": 0.32, "learning_rate": 1.581866465236555e-05, "loss": 2.2871, "step": 13115 }, { "epoch": 0.32, "learning_rate": 1.5818015363620373e-05, "loss": 2.0033, "step": 13116 }, { "epoch": 0.32, "learning_rate": 1.5817366037795087e-05, "loss": 2.1059, "step": 13117 }, { "epoch": 0.32, "learning_rate": 1.5816716674893827e-05, "loss": 2.0923, "step": 13118 }, { "epoch": 0.32, "learning_rate": 1.5816067274920737e-05, "loss": 2.1607, "step": 13119 }, { "epoch": 0.32, "learning_rate": 1.5815417837879958e-05, "loss": 2.0673, "step": 13120 }, { "epoch": 0.32, "learning_rate": 1.5814768363775625e-05, "loss": 2.0844, "step": 13121 }, { "epoch": 0.32, "learning_rate": 1.5814118852611876e-05, "loss": 2.1673, "step": 13122 }, { "epoch": 0.32, "learning_rate": 1.5813469304392855e-05, "loss": 2.2073, "step": 13123 }, { "epoch": 0.32, "learning_rate": 1.58128197191227e-05, "loss": 2.1826, "step": 13124 }, { "epoch": 0.32, "learning_rate": 1.5812170096805548e-05, "loss": 2.1012, "step": 13125 }, { "epoch": 0.32, "learning_rate": 1.5811520437445545e-05, "loss": 2.0636, "step": 13126 }, { "epoch": 0.32, "learning_rate": 1.5810870741046828e-05, "loss": 2.1194, "step": 13127 }, { "epoch": 0.32, "learning_rate": 1.5810221007613536e-05, "loss": 2.1847, "step": 13128 }, { "epoch": 0.32, "learning_rate": 1.580957123714981e-05, "loss": 2.0177, "step": 13129 }, { "epoch": 0.32, "learning_rate": 1.5808921429659795e-05, "loss": 2.0642, "step": 13130 }, { "epoch": 0.32, "learning_rate": 1.5808271585147636e-05, "loss": 2.0943, "step": 13131 }, { "epoch": 0.32, "learning_rate": 1.580762170361746e-05, "loss": 2.0857, "step": 13132 }, { "epoch": 0.32, "learning_rate": 1.5806971785073426e-05, "loss": 2.2273, "step": 13133 }, { "epoch": 0.32, "learning_rate": 1.5806321829519666e-05, "loss": 2.1443, "step": 13134 }, { "epoch": 0.32, "learning_rate": 1.5805671836960324e-05, "loss": 2.1843, "step": 13135 }, { "epoch": 0.32, "learning_rate": 1.5805021807399543e-05, "loss": 2.1017, "step": 13136 }, { "epoch": 0.32, "learning_rate": 1.5804371740841468e-05, "loss": 2.027, "step": 13137 }, { "epoch": 0.32, "learning_rate": 1.580372163729024e-05, "loss": 2.313, "step": 13138 }, { "epoch": 0.32, "learning_rate": 1.580307149675e-05, "loss": 2.1924, "step": 13139 }, { "epoch": 0.32, "learning_rate": 1.5802421319224895e-05, "loss": 2.1198, "step": 13140 }, { "epoch": 0.32, "learning_rate": 1.580177110471907e-05, "loss": 2.0403, "step": 13141 }, { "epoch": 0.32, "learning_rate": 1.5801120853236666e-05, "loss": 2.1258, "step": 13142 }, { "epoch": 0.32, "learning_rate": 1.5800470564781827e-05, "loss": 2.2081, "step": 13143 }, { "epoch": 0.32, "learning_rate": 1.57998202393587e-05, "loss": 2.0599, "step": 13144 }, { "epoch": 0.32, "learning_rate": 1.5799169876971426e-05, "loss": 2.1478, "step": 13145 }, { "epoch": 0.32, "learning_rate": 1.579851947762416e-05, "loss": 2.1279, "step": 13146 }, { "epoch": 0.32, "learning_rate": 1.5797869041321034e-05, "loss": 2.0871, "step": 13147 }, { "epoch": 0.32, "learning_rate": 1.5797218568066198e-05, "loss": 1.9985, "step": 13148 }, { "epoch": 0.32, "learning_rate": 1.5796568057863802e-05, "loss": 1.9309, "step": 13149 }, { "epoch": 0.32, "learning_rate": 1.5795917510717986e-05, "loss": 2.0554, "step": 13150 }, { "epoch": 0.32, "learning_rate": 1.57952669266329e-05, "loss": 2.1747, "step": 13151 }, { "epoch": 0.32, "learning_rate": 1.5794616305612692e-05, "loss": 2.169, "step": 13152 }, { "epoch": 0.32, "learning_rate": 1.57939656476615e-05, "loss": 2.403, "step": 13153 }, { "epoch": 0.32, "learning_rate": 1.579331495278348e-05, "loss": 2.0353, "step": 13154 }, { "epoch": 0.32, "learning_rate": 1.5792664220982778e-05, "loss": 2.1503, "step": 13155 }, { "epoch": 0.32, "learning_rate": 1.5792013452263538e-05, "loss": 2.1597, "step": 13156 }, { "epoch": 0.32, "learning_rate": 1.5791362646629907e-05, "loss": 2.0719, "step": 13157 }, { "epoch": 0.32, "learning_rate": 1.5790711804086033e-05, "loss": 2.1079, "step": 13158 }, { "epoch": 0.32, "learning_rate": 1.5790060924636067e-05, "loss": 2.2105, "step": 13159 }, { "epoch": 0.32, "learning_rate": 1.5789410008284157e-05, "loss": 2.0205, "step": 13160 }, { "epoch": 0.32, "learning_rate": 1.5788759055034453e-05, "loss": 2.3029, "step": 13161 }, { "epoch": 0.32, "learning_rate": 1.5788108064891097e-05, "loss": 2.1644, "step": 13162 }, { "epoch": 0.32, "learning_rate": 1.578745703785824e-05, "loss": 2.0469, "step": 13163 }, { "epoch": 0.32, "learning_rate": 1.5786805973940038e-05, "loss": 2.1459, "step": 13164 }, { "epoch": 0.32, "learning_rate": 1.5786154873140636e-05, "loss": 2.2083, "step": 13165 }, { "epoch": 0.32, "learning_rate": 1.5785503735464178e-05, "loss": 2.0675, "step": 13166 }, { "epoch": 0.32, "learning_rate": 1.5784852560914824e-05, "loss": 2.1421, "step": 13167 }, { "epoch": 0.32, "learning_rate": 1.5784201349496714e-05, "loss": 2.1304, "step": 13168 }, { "epoch": 0.32, "learning_rate": 1.578355010121401e-05, "loss": 2.1346, "step": 13169 }, { "epoch": 0.32, "learning_rate": 1.5782898816070856e-05, "loss": 2.1638, "step": 13170 }, { "epoch": 0.32, "learning_rate": 1.5782247494071402e-05, "loss": 2.0596, "step": 13171 }, { "epoch": 0.32, "learning_rate": 1.5781596135219798e-05, "loss": 1.9837, "step": 13172 }, { "epoch": 0.32, "learning_rate": 1.5780944739520197e-05, "loss": 1.8144, "step": 13173 }, { "epoch": 0.32, "learning_rate": 1.5780293306976754e-05, "loss": 2.0828, "step": 13174 }, { "epoch": 0.32, "learning_rate": 1.5779641837593618e-05, "loss": 2.2239, "step": 13175 }, { "epoch": 0.32, "learning_rate": 1.577899033137494e-05, "loss": 2.1404, "step": 13176 }, { "epoch": 0.32, "learning_rate": 1.5778338788324875e-05, "loss": 2.0459, "step": 13177 }, { "epoch": 0.32, "learning_rate": 1.5777687208447572e-05, "loss": 2.0444, "step": 13178 }, { "epoch": 0.32, "learning_rate": 1.577703559174719e-05, "loss": 2.2091, "step": 13179 }, { "epoch": 0.32, "learning_rate": 1.577638393822787e-05, "loss": 2.2258, "step": 13180 }, { "epoch": 0.32, "learning_rate": 1.5775732247893777e-05, "loss": 2.0266, "step": 13181 }, { "epoch": 0.32, "learning_rate": 1.577508052074906e-05, "loss": 2.0792, "step": 13182 }, { "epoch": 0.32, "learning_rate": 1.577442875679787e-05, "loss": 1.9929, "step": 13183 }, { "epoch": 0.32, "learning_rate": 1.5773776956044365e-05, "loss": 2.2597, "step": 13184 }, { "epoch": 0.32, "learning_rate": 1.57731251184927e-05, "loss": 1.988, "step": 13185 }, { "epoch": 0.33, "learning_rate": 1.5772473244147026e-05, "loss": 2.0732, "step": 13186 }, { "epoch": 0.33, "learning_rate": 1.57718213330115e-05, "loss": 2.1481, "step": 13187 }, { "epoch": 0.33, "learning_rate": 1.5771169385090273e-05, "loss": 2.107, "step": 13188 }, { "epoch": 0.33, "learning_rate": 1.5770517400387504e-05, "loss": 2.183, "step": 13189 }, { "epoch": 0.33, "learning_rate": 1.576986537890735e-05, "loss": 2.063, "step": 13190 }, { "epoch": 0.33, "learning_rate": 1.576921332065396e-05, "loss": 2.2101, "step": 13191 }, { "epoch": 0.33, "learning_rate": 1.5768561225631495e-05, "loss": 2.189, "step": 13192 }, { "epoch": 0.33, "learning_rate": 1.5767909093844108e-05, "loss": 2.0144, "step": 13193 }, { "epoch": 0.33, "learning_rate": 1.5767256925295956e-05, "loss": 2.0538, "step": 13194 }, { "epoch": 0.33, "learning_rate": 1.57666047199912e-05, "loss": 2.1751, "step": 13195 }, { "epoch": 0.33, "learning_rate": 1.576595247793399e-05, "loss": 2.2007, "step": 13196 }, { "epoch": 0.33, "learning_rate": 1.5765300199128484e-05, "loss": 2.1833, "step": 13197 }, { "epoch": 0.33, "learning_rate": 1.576464788357884e-05, "loss": 1.9672, "step": 13198 }, { "epoch": 0.33, "learning_rate": 1.5763995531289222e-05, "loss": 2.2339, "step": 13199 }, { "epoch": 0.33, "learning_rate": 1.576334314226378e-05, "loss": 2.1536, "step": 13200 }, { "epoch": 0.33, "learning_rate": 1.576269071650667e-05, "loss": 2.1672, "step": 13201 }, { "epoch": 0.33, "learning_rate": 1.5762038254022057e-05, "loss": 2.1088, "step": 13202 }, { "epoch": 0.33, "learning_rate": 1.57613857548141e-05, "loss": 2.0937, "step": 13203 }, { "epoch": 0.33, "learning_rate": 1.576073321888695e-05, "loss": 2.2242, "step": 13204 }, { "epoch": 0.33, "learning_rate": 1.5760080646244773e-05, "loss": 2.1383, "step": 13205 }, { "epoch": 0.33, "learning_rate": 1.575942803689172e-05, "loss": 1.9839, "step": 13206 }, { "epoch": 0.33, "learning_rate": 1.5758775390831956e-05, "loss": 2.1845, "step": 13207 }, { "epoch": 0.33, "learning_rate": 1.5758122708069644e-05, "loss": 1.9945, "step": 13208 }, { "epoch": 0.33, "learning_rate": 1.575746998860894e-05, "loss": 2.1607, "step": 13209 }, { "epoch": 0.33, "learning_rate": 1.5756817232453997e-05, "loss": 2.1838, "step": 13210 }, { "epoch": 0.33, "learning_rate": 1.5756164439608986e-05, "loss": 2.3053, "step": 13211 }, { "epoch": 0.33, "learning_rate": 1.575551161007806e-05, "loss": 2.0632, "step": 13212 }, { "epoch": 0.33, "learning_rate": 1.5754858743865387e-05, "loss": 1.947, "step": 13213 }, { "epoch": 0.33, "learning_rate": 1.575420584097512e-05, "loss": 2.0638, "step": 13214 }, { "epoch": 0.33, "learning_rate": 1.5753552901411425e-05, "loss": 1.9571, "step": 13215 }, { "epoch": 0.33, "learning_rate": 1.5752899925178466e-05, "loss": 2.3438, "step": 13216 }, { "epoch": 0.33, "learning_rate": 1.5752246912280397e-05, "loss": 2.0963, "step": 13217 }, { "epoch": 0.33, "learning_rate": 1.5751593862721388e-05, "loss": 2.0318, "step": 13218 }, { "epoch": 0.33, "learning_rate": 1.575094077650559e-05, "loss": 1.9712, "step": 13219 }, { "epoch": 0.33, "learning_rate": 1.5750287653637174e-05, "loss": 1.9831, "step": 13220 }, { "epoch": 0.33, "learning_rate": 1.5749634494120304e-05, "loss": 2.2345, "step": 13221 }, { "epoch": 0.33, "learning_rate": 1.574898129795914e-05, "loss": 2.0295, "step": 13222 }, { "epoch": 0.33, "learning_rate": 1.5748328065157843e-05, "loss": 2.0387, "step": 13223 }, { "epoch": 0.33, "learning_rate": 1.5747674795720576e-05, "loss": 1.9684, "step": 13224 }, { "epoch": 0.33, "learning_rate": 1.574702148965151e-05, "loss": 2.0081, "step": 13225 }, { "epoch": 0.33, "learning_rate": 1.5746368146954797e-05, "loss": 2.2262, "step": 13226 }, { "epoch": 0.33, "learning_rate": 1.574571476763461e-05, "loss": 2.0279, "step": 13227 }, { "epoch": 0.33, "learning_rate": 1.574506135169511e-05, "loss": 2.105, "step": 13228 }, { "epoch": 0.33, "learning_rate": 1.5744407899140463e-05, "loss": 2.2236, "step": 13229 }, { "epoch": 0.33, "learning_rate": 1.574375440997483e-05, "loss": 1.9023, "step": 13230 }, { "epoch": 0.33, "learning_rate": 1.574310088420238e-05, "loss": 2.0974, "step": 13231 }, { "epoch": 0.33, "learning_rate": 1.5742447321827277e-05, "loss": 2.0383, "step": 13232 }, { "epoch": 0.33, "learning_rate": 1.5741793722853682e-05, "loss": 1.9859, "step": 13233 }, { "epoch": 0.33, "learning_rate": 1.574114008728577e-05, "loss": 1.8481, "step": 13234 }, { "epoch": 0.33, "learning_rate": 1.5740486415127694e-05, "loss": 2.084, "step": 13235 }, { "epoch": 0.33, "learning_rate": 1.5739832706383633e-05, "loss": 2.2385, "step": 13236 }, { "epoch": 0.33, "learning_rate": 1.5739178961057748e-05, "loss": 2.2051, "step": 13237 }, { "epoch": 0.33, "learning_rate": 1.5738525179154204e-05, "loss": 2.1522, "step": 13238 }, { "epoch": 0.33, "learning_rate": 1.5737871360677167e-05, "loss": 2.0466, "step": 13239 }, { "epoch": 0.33, "learning_rate": 1.5737217505630808e-05, "loss": 2.0043, "step": 13240 }, { "epoch": 0.33, "learning_rate": 1.5736563614019293e-05, "loss": 2.2177, "step": 13241 }, { "epoch": 0.33, "learning_rate": 1.5735909685846783e-05, "loss": 2.0291, "step": 13242 }, { "epoch": 0.33, "learning_rate": 1.5735255721117456e-05, "loss": 2.0814, "step": 13243 }, { "epoch": 0.33, "learning_rate": 1.5734601719835473e-05, "loss": 2.1112, "step": 13244 }, { "epoch": 0.33, "learning_rate": 1.5733947682005007e-05, "loss": 2.1866, "step": 13245 }, { "epoch": 0.33, "learning_rate": 1.5733293607630223e-05, "loss": 2.1404, "step": 13246 }, { "epoch": 0.33, "learning_rate": 1.5732639496715288e-05, "loss": 2.1853, "step": 13247 }, { "epoch": 0.33, "learning_rate": 1.573198534926438e-05, "loss": 2.0997, "step": 13248 }, { "epoch": 0.33, "learning_rate": 1.5731331165281652e-05, "loss": 1.9154, "step": 13249 }, { "epoch": 0.33, "learning_rate": 1.573067694477129e-05, "loss": 1.9893, "step": 13250 }, { "epoch": 0.33, "learning_rate": 1.573002268773745e-05, "loss": 2.2342, "step": 13251 }, { "epoch": 0.33, "learning_rate": 1.5729368394184314e-05, "loss": 2.1523, "step": 13252 }, { "epoch": 0.33, "learning_rate": 1.5728714064116046e-05, "loss": 2.1168, "step": 13253 }, { "epoch": 0.33, "learning_rate": 1.5728059697536813e-05, "loss": 2.215, "step": 13254 }, { "epoch": 0.33, "learning_rate": 1.572740529445079e-05, "loss": 2.0556, "step": 13255 }, { "epoch": 0.33, "learning_rate": 1.5726750854862144e-05, "loss": 2.0985, "step": 13256 }, { "epoch": 0.33, "learning_rate": 1.5726096378775052e-05, "loss": 2.1925, "step": 13257 }, { "epoch": 0.33, "learning_rate": 1.5725441866193682e-05, "loss": 2.1905, "step": 13258 }, { "epoch": 0.33, "learning_rate": 1.5724787317122203e-05, "loss": 2.1931, "step": 13259 }, { "epoch": 0.33, "learning_rate": 1.572413273156479e-05, "loss": 2.0862, "step": 13260 }, { "epoch": 0.33, "learning_rate": 1.5723478109525615e-05, "loss": 2.0384, "step": 13261 }, { "epoch": 0.33, "learning_rate": 1.5722823451008843e-05, "loss": 2.1913, "step": 13262 }, { "epoch": 0.33, "learning_rate": 1.572216875601866e-05, "loss": 2.2078, "step": 13263 }, { "epoch": 0.33, "learning_rate": 1.5721514024559222e-05, "loss": 2.0566, "step": 13264 }, { "epoch": 0.33, "learning_rate": 1.572085925663472e-05, "loss": 2.0773, "step": 13265 }, { "epoch": 0.33, "learning_rate": 1.5720204452249312e-05, "loss": 2.1091, "step": 13266 }, { "epoch": 0.33, "learning_rate": 1.5719549611407177e-05, "loss": 2.211, "step": 13267 }, { "epoch": 0.33, "learning_rate": 1.5718894734112488e-05, "loss": 2.0737, "step": 13268 }, { "epoch": 0.33, "learning_rate": 1.571823982036942e-05, "loss": 2.2252, "step": 13269 }, { "epoch": 0.33, "learning_rate": 1.5717584870182147e-05, "loss": 2.1909, "step": 13270 }, { "epoch": 0.33, "learning_rate": 1.5716929883554844e-05, "loss": 2.1082, "step": 13271 }, { "epoch": 0.33, "learning_rate": 1.571627486049168e-05, "loss": 2.0962, "step": 13272 }, { "epoch": 0.33, "learning_rate": 1.5715619800996836e-05, "loss": 2.1448, "step": 13273 }, { "epoch": 0.33, "learning_rate": 1.571496470507448e-05, "loss": 2.0826, "step": 13274 }, { "epoch": 0.33, "learning_rate": 1.5714309572728796e-05, "loss": 1.9974, "step": 13275 }, { "epoch": 0.33, "learning_rate": 1.5713654403963957e-05, "loss": 2.097, "step": 13276 }, { "epoch": 0.33, "learning_rate": 1.571299919878413e-05, "loss": 2.0063, "step": 13277 }, { "epoch": 0.33, "learning_rate": 1.5712343957193502e-05, "loss": 2.186, "step": 13278 }, { "epoch": 0.33, "learning_rate": 1.571168867919624e-05, "loss": 1.9151, "step": 13279 }, { "epoch": 0.33, "learning_rate": 1.5711033364796527e-05, "loss": 2.0198, "step": 13280 }, { "epoch": 0.33, "learning_rate": 1.571037801399854e-05, "loss": 2.2434, "step": 13281 }, { "epoch": 0.33, "learning_rate": 1.570972262680645e-05, "loss": 2.1334, "step": 13282 }, { "epoch": 0.33, "learning_rate": 1.570906720322444e-05, "loss": 2.242, "step": 13283 }, { "epoch": 0.33, "learning_rate": 1.570841174325668e-05, "loss": 2.1208, "step": 13284 }, { "epoch": 0.33, "learning_rate": 1.570775624690735e-05, "loss": 1.9309, "step": 13285 }, { "epoch": 0.33, "learning_rate": 1.5707100714180638e-05, "loss": 1.9966, "step": 13286 }, { "epoch": 0.33, "learning_rate": 1.5706445145080707e-05, "loss": 2.1544, "step": 13287 }, { "epoch": 0.33, "learning_rate": 1.570578953961174e-05, "loss": 2.1546, "step": 13288 }, { "epoch": 0.33, "learning_rate": 1.570513389777792e-05, "loss": 2.1721, "step": 13289 }, { "epoch": 0.33, "learning_rate": 1.570447821958342e-05, "loss": 2.1013, "step": 13290 }, { "epoch": 0.33, "learning_rate": 1.570382250503242e-05, "loss": 2.3276, "step": 13291 }, { "epoch": 0.33, "learning_rate": 1.5703166754129106e-05, "loss": 2.0898, "step": 13292 }, { "epoch": 0.33, "learning_rate": 1.5702510966877648e-05, "loss": 2.1763, "step": 13293 }, { "epoch": 0.33, "learning_rate": 1.5701855143282233e-05, "loss": 2.1327, "step": 13294 }, { "epoch": 0.33, "learning_rate": 1.5701199283347034e-05, "loss": 2.1432, "step": 13295 }, { "epoch": 0.33, "learning_rate": 1.5700543387076233e-05, "loss": 1.9914, "step": 13296 }, { "epoch": 0.33, "learning_rate": 1.5699887454474013e-05, "loss": 2.1357, "step": 13297 }, { "epoch": 0.33, "learning_rate": 1.569923148554455e-05, "loss": 2.2417, "step": 13298 }, { "epoch": 0.33, "learning_rate": 1.569857548029203e-05, "loss": 2.1976, "step": 13299 }, { "epoch": 0.33, "learning_rate": 1.5697919438720633e-05, "loss": 1.9329, "step": 13300 }, { "epoch": 0.33, "learning_rate": 1.5697263360834533e-05, "loss": 2.0483, "step": 13301 }, { "epoch": 0.33, "learning_rate": 1.569660724663792e-05, "loss": 1.9666, "step": 13302 }, { "epoch": 0.33, "learning_rate": 1.5695951096134973e-05, "loss": 2.1748, "step": 13303 }, { "epoch": 0.33, "learning_rate": 1.5695294909329875e-05, "loss": 1.9986, "step": 13304 }, { "epoch": 0.33, "learning_rate": 1.5694638686226803e-05, "loss": 2.1065, "step": 13305 }, { "epoch": 0.33, "learning_rate": 1.5693982426829944e-05, "loss": 1.9771, "step": 13306 }, { "epoch": 0.33, "learning_rate": 1.569332613114348e-05, "loss": 2.1555, "step": 13307 }, { "epoch": 0.33, "learning_rate": 1.5692669799171593e-05, "loss": 2.1783, "step": 13308 }, { "epoch": 0.33, "learning_rate": 1.5692013430918468e-05, "loss": 2.2351, "step": 13309 }, { "epoch": 0.33, "learning_rate": 1.5691357026388284e-05, "loss": 2.2472, "step": 13310 }, { "epoch": 0.33, "learning_rate": 1.569070058558523e-05, "loss": 2.0676, "step": 13311 }, { "epoch": 0.33, "learning_rate": 1.5690044108513483e-05, "loss": 2.0905, "step": 13312 }, { "epoch": 0.33, "learning_rate": 1.568938759517723e-05, "loss": 2.1153, "step": 13313 }, { "epoch": 0.33, "learning_rate": 1.5688731045580658e-05, "loss": 2.065, "step": 13314 }, { "epoch": 0.33, "learning_rate": 1.5688074459727947e-05, "loss": 2.2284, "step": 13315 }, { "epoch": 0.33, "learning_rate": 1.5687417837623287e-05, "loss": 2.1729, "step": 13316 }, { "epoch": 0.33, "learning_rate": 1.568676117927086e-05, "loss": 2.2552, "step": 13317 }, { "epoch": 0.33, "learning_rate": 1.5686104484674847e-05, "loss": 2.1015, "step": 13318 }, { "epoch": 0.33, "learning_rate": 1.568544775383944e-05, "loss": 2.0742, "step": 13319 }, { "epoch": 0.33, "learning_rate": 1.5684790986768818e-05, "loss": 2.1305, "step": 13320 }, { "epoch": 0.33, "learning_rate": 1.5684134183467175e-05, "loss": 2.1704, "step": 13321 }, { "epoch": 0.33, "learning_rate": 1.568347734393869e-05, "loss": 2.3277, "step": 13322 }, { "epoch": 0.33, "learning_rate": 1.568282046818755e-05, "loss": 2.1331, "step": 13323 }, { "epoch": 0.33, "learning_rate": 1.5682163556217947e-05, "loss": 2.124, "step": 13324 }, { "epoch": 0.33, "learning_rate": 1.568150660803406e-05, "loss": 2.0172, "step": 13325 }, { "epoch": 0.33, "learning_rate": 1.5680849623640084e-05, "loss": 2.1003, "step": 13326 }, { "epoch": 0.33, "learning_rate": 1.5680192603040198e-05, "loss": 2.2345, "step": 13327 }, { "epoch": 0.33, "learning_rate": 1.5679535546238595e-05, "loss": 2.0711, "step": 13328 }, { "epoch": 0.33, "learning_rate": 1.5678878453239463e-05, "loss": 2.1935, "step": 13329 }, { "epoch": 0.33, "learning_rate": 1.5678221324046985e-05, "loss": 2.1237, "step": 13330 }, { "epoch": 0.33, "learning_rate": 1.5677564158665356e-05, "loss": 2.1968, "step": 13331 }, { "epoch": 0.33, "learning_rate": 1.5676906957098755e-05, "loss": 1.8893, "step": 13332 }, { "epoch": 0.33, "learning_rate": 1.567624971935138e-05, "loss": 2.112, "step": 13333 }, { "epoch": 0.33, "learning_rate": 1.5675592445427414e-05, "loss": 2.1873, "step": 13334 }, { "epoch": 0.33, "learning_rate": 1.5674935135331047e-05, "loss": 2.2007, "step": 13335 }, { "epoch": 0.33, "learning_rate": 1.567427778906647e-05, "loss": 2.1321, "step": 13336 }, { "epoch": 0.33, "learning_rate": 1.567362040663787e-05, "loss": 2.0666, "step": 13337 }, { "epoch": 0.33, "learning_rate": 1.5672962988049442e-05, "loss": 1.9563, "step": 13338 }, { "epoch": 0.33, "learning_rate": 1.567230553330537e-05, "loss": 2.0992, "step": 13339 }, { "epoch": 0.33, "learning_rate": 1.5671648042409847e-05, "loss": 1.9473, "step": 13340 }, { "epoch": 0.33, "learning_rate": 1.5670990515367065e-05, "loss": 2.077, "step": 13341 }, { "epoch": 0.33, "learning_rate": 1.5670332952181207e-05, "loss": 2.0741, "step": 13342 }, { "epoch": 0.33, "learning_rate": 1.5669675352856472e-05, "loss": 1.9697, "step": 13343 }, { "epoch": 0.33, "learning_rate": 1.5669017717397043e-05, "loss": 2.2027, "step": 13344 }, { "epoch": 0.33, "learning_rate": 1.5668360045807123e-05, "loss": 2.1371, "step": 13345 }, { "epoch": 0.33, "learning_rate": 1.5667702338090895e-05, "loss": 2.1322, "step": 13346 }, { "epoch": 0.33, "learning_rate": 1.5667044594252554e-05, "loss": 2.2216, "step": 13347 }, { "epoch": 0.33, "learning_rate": 1.566638681429629e-05, "loss": 1.9707, "step": 13348 }, { "epoch": 0.33, "learning_rate": 1.5665728998226297e-05, "loss": 2.1685, "step": 13349 }, { "epoch": 0.33, "learning_rate": 1.5665071146046767e-05, "loss": 1.9301, "step": 13350 }, { "epoch": 0.33, "learning_rate": 1.566441325776189e-05, "loss": 2.0533, "step": 13351 }, { "epoch": 0.33, "learning_rate": 1.5663755333375862e-05, "loss": 2.1936, "step": 13352 }, { "epoch": 0.33, "learning_rate": 1.5663097372892876e-05, "loss": 2.3045, "step": 13353 }, { "epoch": 0.33, "learning_rate": 1.5662439376317126e-05, "loss": 1.8722, "step": 13354 }, { "epoch": 0.33, "learning_rate": 1.5661781343652804e-05, "loss": 2.2274, "step": 13355 }, { "epoch": 0.33, "learning_rate": 1.5661123274904104e-05, "loss": 2.0603, "step": 13356 }, { "epoch": 0.33, "learning_rate": 1.5660465170075218e-05, "loss": 2.121, "step": 13357 }, { "epoch": 0.33, "learning_rate": 1.565980702917035e-05, "loss": 2.1222, "step": 13358 }, { "epoch": 0.33, "learning_rate": 1.565914885219368e-05, "loss": 2.2045, "step": 13359 }, { "epoch": 0.33, "learning_rate": 1.5658490639149414e-05, "loss": 2.1084, "step": 13360 }, { "epoch": 0.33, "learning_rate": 1.565783239004174e-05, "loss": 2.0078, "step": 13361 }, { "epoch": 0.33, "learning_rate": 1.5657174104874858e-05, "loss": 2.0953, "step": 13362 }, { "epoch": 0.33, "learning_rate": 1.5656515783652964e-05, "loss": 2.0566, "step": 13363 }, { "epoch": 0.33, "learning_rate": 1.565585742638025e-05, "loss": 2.1354, "step": 13364 }, { "epoch": 0.33, "learning_rate": 1.565519903306091e-05, "loss": 2.161, "step": 13365 }, { "epoch": 0.33, "learning_rate": 1.565454060369915e-05, "loss": 2.1177, "step": 13366 }, { "epoch": 0.33, "learning_rate": 1.5653882138299155e-05, "loss": 2.2676, "step": 13367 }, { "epoch": 0.33, "learning_rate": 1.565322363686513e-05, "loss": 2.2033, "step": 13368 }, { "epoch": 0.33, "learning_rate": 1.5652565099401265e-05, "loss": 2.3815, "step": 13369 }, { "epoch": 0.33, "learning_rate": 1.565190652591176e-05, "loss": 2.1139, "step": 13370 }, { "epoch": 0.33, "learning_rate": 1.5651247916400814e-05, "loss": 2.1213, "step": 13371 }, { "epoch": 0.33, "learning_rate": 1.5650589270872628e-05, "loss": 2.1501, "step": 13372 }, { "epoch": 0.33, "learning_rate": 1.564993058933139e-05, "loss": 2.0278, "step": 13373 }, { "epoch": 0.33, "learning_rate": 1.5649271871781305e-05, "loss": 2.3089, "step": 13374 }, { "epoch": 0.33, "learning_rate": 1.564861311822657e-05, "loss": 2.1802, "step": 13375 }, { "epoch": 0.33, "learning_rate": 1.564795432867138e-05, "loss": 2.1335, "step": 13376 }, { "epoch": 0.33, "learning_rate": 1.5647295503119937e-05, "loss": 2.1122, "step": 13377 }, { "epoch": 0.33, "learning_rate": 1.564663664157644e-05, "loss": 2.017, "step": 13378 }, { "epoch": 0.33, "learning_rate": 1.564597774404509e-05, "loss": 2.3262, "step": 13379 }, { "epoch": 0.33, "learning_rate": 1.5645318810530083e-05, "loss": 2.011, "step": 13380 }, { "epoch": 0.33, "learning_rate": 1.5644659841035616e-05, "loss": 2.0458, "step": 13381 }, { "epoch": 0.33, "learning_rate": 1.5644000835565894e-05, "loss": 2.143, "step": 13382 }, { "epoch": 0.33, "learning_rate": 1.5643341794125117e-05, "loss": 2.1537, "step": 13383 }, { "epoch": 0.33, "learning_rate": 1.564268271671748e-05, "loss": 2.172, "step": 13384 }, { "epoch": 0.33, "learning_rate": 1.564202360334719e-05, "loss": 2.2175, "step": 13385 }, { "epoch": 0.33, "learning_rate": 1.5641364454018442e-05, "loss": 2.0215, "step": 13386 }, { "epoch": 0.33, "learning_rate": 1.5640705268735444e-05, "loss": 2.1318, "step": 13387 }, { "epoch": 0.33, "learning_rate": 1.5640046047502392e-05, "loss": 2.0725, "step": 13388 }, { "epoch": 0.33, "learning_rate": 1.5639386790323488e-05, "loss": 2.1017, "step": 13389 }, { "epoch": 0.33, "learning_rate": 1.5638727497202933e-05, "loss": 2.0582, "step": 13390 }, { "epoch": 0.33, "learning_rate": 1.563806816814493e-05, "loss": 2.0503, "step": 13391 }, { "epoch": 0.33, "learning_rate": 1.5637408803153685e-05, "loss": 1.9612, "step": 13392 }, { "epoch": 0.33, "learning_rate": 1.5636749402233395e-05, "loss": 1.9464, "step": 13393 }, { "epoch": 0.33, "learning_rate": 1.5636089965388262e-05, "loss": 2.2208, "step": 13394 }, { "epoch": 0.33, "learning_rate": 1.563543049262249e-05, "loss": 2.2675, "step": 13395 }, { "epoch": 0.33, "learning_rate": 1.5634770983940285e-05, "loss": 2.0996, "step": 13396 }, { "epoch": 0.33, "learning_rate": 1.563411143934585e-05, "loss": 2.2238, "step": 13397 }, { "epoch": 0.33, "learning_rate": 1.5633451858843386e-05, "loss": 2.1928, "step": 13398 }, { "epoch": 0.33, "learning_rate": 1.5632792242437094e-05, "loss": 1.9985, "step": 13399 }, { "epoch": 0.33, "learning_rate": 1.5632132590131185e-05, "loss": 2.122, "step": 13400 }, { "epoch": 0.33, "learning_rate": 1.5631472901929855e-05, "loss": 2.2119, "step": 13401 }, { "epoch": 0.33, "learning_rate": 1.5630813177837317e-05, "loss": 1.9979, "step": 13402 }, { "epoch": 0.33, "learning_rate": 1.563015341785777e-05, "loss": 2.0407, "step": 13403 }, { "epoch": 0.33, "learning_rate": 1.5629493621995423e-05, "loss": 2.0753, "step": 13404 }, { "epoch": 0.33, "learning_rate": 1.5628833790254474e-05, "loss": 2.2735, "step": 13405 }, { "epoch": 0.33, "learning_rate": 1.5628173922639135e-05, "loss": 1.8811, "step": 13406 }, { "epoch": 0.33, "learning_rate": 1.562751401915361e-05, "loss": 2.0799, "step": 13407 }, { "epoch": 0.33, "learning_rate": 1.5626854079802107e-05, "loss": 1.8985, "step": 13408 }, { "epoch": 0.33, "learning_rate": 1.5626194104588823e-05, "loss": 2.0685, "step": 13409 }, { "epoch": 0.33, "learning_rate": 1.5625534093517975e-05, "loss": 2.205, "step": 13410 }, { "epoch": 0.33, "learning_rate": 1.5624874046593764e-05, "loss": 2.2016, "step": 13411 }, { "epoch": 0.33, "learning_rate": 1.5624213963820397e-05, "loss": 2.119, "step": 13412 }, { "epoch": 0.33, "learning_rate": 1.562355384520208e-05, "loss": 2.065, "step": 13413 }, { "epoch": 0.33, "learning_rate": 1.5622893690743025e-05, "loss": 2.2515, "step": 13414 }, { "epoch": 0.33, "learning_rate": 1.5622233500447436e-05, "loss": 1.9457, "step": 13415 }, { "epoch": 0.33, "learning_rate": 1.5621573274319517e-05, "loss": 2.1351, "step": 13416 }, { "epoch": 0.33, "learning_rate": 1.562091301236348e-05, "loss": 2.3537, "step": 13417 }, { "epoch": 0.33, "learning_rate": 1.5620252714583536e-05, "loss": 2.1119, "step": 13418 }, { "epoch": 0.33, "learning_rate": 1.5619592380983887e-05, "loss": 2.2163, "step": 13419 }, { "epoch": 0.33, "learning_rate": 1.5618932011568747e-05, "loss": 2.1861, "step": 13420 }, { "epoch": 0.33, "learning_rate": 1.5618271606342318e-05, "loss": 2.2383, "step": 13421 }, { "epoch": 0.33, "learning_rate": 1.5617611165308818e-05, "loss": 2.1513, "step": 13422 }, { "epoch": 0.33, "learning_rate": 1.5616950688472448e-05, "loss": 1.9521, "step": 13423 }, { "epoch": 0.33, "learning_rate": 1.561629017583742e-05, "loss": 2.0236, "step": 13424 }, { "epoch": 0.33, "learning_rate": 1.5615629627407946e-05, "loss": 2.0583, "step": 13425 }, { "epoch": 0.33, "learning_rate": 1.5614969043188235e-05, "loss": 2.27, "step": 13426 }, { "epoch": 0.33, "learning_rate": 1.5614308423182498e-05, "loss": 2.0834, "step": 13427 }, { "epoch": 0.33, "learning_rate": 1.561364776739494e-05, "loss": 2.1212, "step": 13428 }, { "epoch": 0.33, "learning_rate": 1.5612987075829776e-05, "loss": 2.0277, "step": 13429 }, { "epoch": 0.33, "learning_rate": 1.561232634849122e-05, "loss": 2.1512, "step": 13430 }, { "epoch": 0.33, "learning_rate": 1.5611665585383476e-05, "loss": 2.0628, "step": 13431 }, { "epoch": 0.33, "learning_rate": 1.561100478651076e-05, "loss": 2.1443, "step": 13432 }, { "epoch": 0.33, "learning_rate": 1.561034395187728e-05, "loss": 2.1687, "step": 13433 }, { "epoch": 0.33, "learning_rate": 1.5609683081487247e-05, "loss": 2.0359, "step": 13434 }, { "epoch": 0.33, "learning_rate": 1.560902217534488e-05, "loss": 2.167, "step": 13435 }, { "epoch": 0.33, "learning_rate": 1.5608361233454384e-05, "loss": 2.0322, "step": 13436 }, { "epoch": 0.33, "learning_rate": 1.560770025581998e-05, "loss": 2.1952, "step": 13437 }, { "epoch": 0.33, "learning_rate": 1.560703924244587e-05, "loss": 2.0592, "step": 13438 }, { "epoch": 0.33, "learning_rate": 1.560637819333627e-05, "loss": 2.183, "step": 13439 }, { "epoch": 0.33, "learning_rate": 1.56057171084954e-05, "loss": 2.1022, "step": 13440 }, { "epoch": 0.33, "learning_rate": 1.5605055987927463e-05, "loss": 2.2113, "step": 13441 }, { "epoch": 0.33, "learning_rate": 1.560439483163668e-05, "loss": 2.2128, "step": 13442 }, { "epoch": 0.33, "learning_rate": 1.5603733639627262e-05, "loss": 2.0455, "step": 13443 }, { "epoch": 0.33, "learning_rate": 1.560307241190342e-05, "loss": 2.1057, "step": 13444 }, { "epoch": 0.33, "learning_rate": 1.5602411148469377e-05, "loss": 2.2024, "step": 13445 }, { "epoch": 0.33, "learning_rate": 1.560174984932934e-05, "loss": 2.0455, "step": 13446 }, { "epoch": 0.33, "learning_rate": 1.5601088514487527e-05, "loss": 2.0154, "step": 13447 }, { "epoch": 0.33, "learning_rate": 1.5600427143948148e-05, "loss": 1.9931, "step": 13448 }, { "epoch": 0.33, "learning_rate": 1.559976573771542e-05, "loss": 2.3673, "step": 13449 }, { "epoch": 0.33, "learning_rate": 1.5599104295793567e-05, "loss": 1.9352, "step": 13450 }, { "epoch": 0.33, "learning_rate": 1.5598442818186794e-05, "loss": 2.1155, "step": 13451 }, { "epoch": 0.33, "learning_rate": 1.559778130489932e-05, "loss": 2.1014, "step": 13452 }, { "epoch": 0.33, "learning_rate": 1.559711975593536e-05, "loss": 2.04, "step": 13453 }, { "epoch": 0.33, "learning_rate": 1.5596458171299133e-05, "loss": 2.1882, "step": 13454 }, { "epoch": 0.33, "learning_rate": 1.5595796550994857e-05, "loss": 2.0803, "step": 13455 }, { "epoch": 0.33, "learning_rate": 1.5595134895026744e-05, "loss": 2.0812, "step": 13456 }, { "epoch": 0.33, "learning_rate": 1.5594473203399015e-05, "loss": 2.0282, "step": 13457 }, { "epoch": 0.33, "learning_rate": 1.559381147611588e-05, "loss": 2.0071, "step": 13458 }, { "epoch": 0.33, "learning_rate": 1.5593149713181564e-05, "loss": 2.1032, "step": 13459 }, { "epoch": 0.33, "learning_rate": 1.5592487914600286e-05, "loss": 2.1876, "step": 13460 }, { "epoch": 0.33, "learning_rate": 1.5591826080376257e-05, "loss": 2.013, "step": 13461 }, { "epoch": 0.33, "learning_rate": 1.5591164210513697e-05, "loss": 2.0369, "step": 13462 }, { "epoch": 0.33, "learning_rate": 1.5590502305016823e-05, "loss": 2.0462, "step": 13463 }, { "epoch": 0.33, "learning_rate": 1.558984036388986e-05, "loss": 2.1255, "step": 13464 }, { "epoch": 0.33, "learning_rate": 1.558917838713702e-05, "loss": 2.056, "step": 13465 }, { "epoch": 0.33, "learning_rate": 1.558851637476253e-05, "loss": 2.0471, "step": 13466 }, { "epoch": 0.33, "learning_rate": 1.55878543267706e-05, "loss": 2.3172, "step": 13467 }, { "epoch": 0.33, "learning_rate": 1.5587192243165453e-05, "loss": 2.1782, "step": 13468 }, { "epoch": 0.33, "learning_rate": 1.5586530123951308e-05, "loss": 1.9677, "step": 13469 }, { "epoch": 0.33, "learning_rate": 1.558586796913239e-05, "loss": 2.03, "step": 13470 }, { "epoch": 0.33, "learning_rate": 1.558520577871291e-05, "loss": 2.0447, "step": 13471 }, { "epoch": 0.33, "learning_rate": 1.5584543552697096e-05, "loss": 2.1434, "step": 13472 }, { "epoch": 0.33, "learning_rate": 1.5583881291089167e-05, "loss": 2.1079, "step": 13473 }, { "epoch": 0.33, "learning_rate": 1.5583218993893342e-05, "loss": 2.1006, "step": 13474 }, { "epoch": 0.33, "learning_rate": 1.5582556661113842e-05, "loss": 2.2299, "step": 13475 }, { "epoch": 0.33, "learning_rate": 1.5581894292754893e-05, "loss": 2.2386, "step": 13476 }, { "epoch": 0.33, "learning_rate": 1.558123188882071e-05, "loss": 2.3256, "step": 13477 }, { "epoch": 0.33, "learning_rate": 1.558056944931551e-05, "loss": 1.9658, "step": 13478 }, { "epoch": 0.33, "learning_rate": 1.5579906974243533e-05, "loss": 1.9107, "step": 13479 }, { "epoch": 0.33, "learning_rate": 1.5579244463608984e-05, "loss": 1.996, "step": 13480 }, { "epoch": 0.33, "learning_rate": 1.5578581917416095e-05, "loss": 2.1181, "step": 13481 }, { "epoch": 0.33, "learning_rate": 1.557791933566908e-05, "loss": 2.1862, "step": 13482 }, { "epoch": 0.33, "learning_rate": 1.5577256718372177e-05, "loss": 1.9808, "step": 13483 }, { "epoch": 0.33, "learning_rate": 1.5576594065529592e-05, "loss": 2.2686, "step": 13484 }, { "epoch": 0.33, "learning_rate": 1.5575931377145563e-05, "loss": 1.9484, "step": 13485 }, { "epoch": 0.33, "learning_rate": 1.55752686532243e-05, "loss": 2.1232, "step": 13486 }, { "epoch": 0.33, "learning_rate": 1.5574605893770032e-05, "loss": 2.1743, "step": 13487 }, { "epoch": 0.33, "learning_rate": 1.5573943098786984e-05, "loss": 2.0735, "step": 13488 }, { "epoch": 0.33, "learning_rate": 1.5573280268279384e-05, "loss": 2.038, "step": 13489 }, { "epoch": 0.33, "learning_rate": 1.5572617402251448e-05, "loss": 2.2986, "step": 13490 }, { "epoch": 0.33, "learning_rate": 1.557195450070741e-05, "loss": 2.3067, "step": 13491 }, { "epoch": 0.33, "learning_rate": 1.5571291563651488e-05, "loss": 2.141, "step": 13492 }, { "epoch": 0.33, "learning_rate": 1.5570628591087905e-05, "loss": 1.9575, "step": 13493 }, { "epoch": 0.33, "learning_rate": 1.5569965583020898e-05, "loss": 2.1075, "step": 13494 }, { "epoch": 0.33, "learning_rate": 1.5569302539454683e-05, "loss": 2.2277, "step": 13495 }, { "epoch": 0.33, "learning_rate": 1.5568639460393485e-05, "loss": 2.014, "step": 13496 }, { "epoch": 0.33, "learning_rate": 1.5567976345841535e-05, "loss": 1.9732, "step": 13497 }, { "epoch": 0.33, "learning_rate": 1.5567313195803055e-05, "loss": 2.0875, "step": 13498 }, { "epoch": 0.33, "learning_rate": 1.556665001028228e-05, "loss": 2.093, "step": 13499 }, { "epoch": 0.33, "learning_rate": 1.556598678928342e-05, "loss": 2.0955, "step": 13500 }, { "epoch": 0.33, "learning_rate": 1.556532353281072e-05, "loss": 2.0704, "step": 13501 }, { "epoch": 0.33, "learning_rate": 1.55646602408684e-05, "loss": 2.2277, "step": 13502 }, { "epoch": 0.33, "learning_rate": 1.5563996913460684e-05, "loss": 1.9032, "step": 13503 }, { "epoch": 0.33, "learning_rate": 1.5563333550591803e-05, "loss": 2.3064, "step": 13504 }, { "epoch": 0.33, "learning_rate": 1.5562670152265984e-05, "loss": 2.1867, "step": 13505 }, { "epoch": 0.33, "learning_rate": 1.5562006718487455e-05, "loss": 2.0955, "step": 13506 }, { "epoch": 0.33, "learning_rate": 1.5561343249260445e-05, "loss": 2.087, "step": 13507 }, { "epoch": 0.33, "learning_rate": 1.5560679744589182e-05, "loss": 2.1629, "step": 13508 }, { "epoch": 0.33, "learning_rate": 1.5560016204477894e-05, "loss": 1.9898, "step": 13509 }, { "epoch": 0.33, "learning_rate": 1.5559352628930814e-05, "loss": 2.0669, "step": 13510 }, { "epoch": 0.33, "learning_rate": 1.5558689017952166e-05, "loss": 2.0396, "step": 13511 }, { "epoch": 0.33, "learning_rate": 1.555802537154618e-05, "loss": 2.1811, "step": 13512 }, { "epoch": 0.33, "learning_rate": 1.5557361689717087e-05, "loss": 2.1295, "step": 13513 }, { "epoch": 0.33, "learning_rate": 1.555669797246912e-05, "loss": 2.068, "step": 13514 }, { "epoch": 0.33, "learning_rate": 1.55560342198065e-05, "loss": 2.0993, "step": 13515 }, { "epoch": 0.33, "learning_rate": 1.555537043173347e-05, "loss": 2.0184, "step": 13516 }, { "epoch": 0.33, "learning_rate": 1.555470660825425e-05, "loss": 2.0116, "step": 13517 }, { "epoch": 0.33, "learning_rate": 1.5554042749373076e-05, "loss": 2.1558, "step": 13518 }, { "epoch": 0.33, "learning_rate": 1.5553378855094176e-05, "loss": 2.2572, "step": 13519 }, { "epoch": 0.33, "learning_rate": 1.5552714925421787e-05, "loss": 2.0227, "step": 13520 }, { "epoch": 0.33, "learning_rate": 1.555205096036013e-05, "loss": 2.1908, "step": 13521 }, { "epoch": 0.33, "learning_rate": 1.5551386959913447e-05, "loss": 2.1772, "step": 13522 }, { "epoch": 0.33, "learning_rate": 1.5550722924085963e-05, "loss": 2.082, "step": 13523 }, { "epoch": 0.33, "learning_rate": 1.5550058852881917e-05, "loss": 2.1815, "step": 13524 }, { "epoch": 0.33, "learning_rate": 1.554939474630553e-05, "loss": 2.0799, "step": 13525 }, { "epoch": 0.33, "learning_rate": 1.5548730604361048e-05, "loss": 2.1341, "step": 13526 }, { "epoch": 0.33, "learning_rate": 1.55480664270527e-05, "loss": 1.9212, "step": 13527 }, { "epoch": 0.33, "learning_rate": 1.554740221438471e-05, "loss": 2.2502, "step": 13528 }, { "epoch": 0.33, "learning_rate": 1.5546737966361322e-05, "loss": 2.1357, "step": 13529 }, { "epoch": 0.33, "learning_rate": 1.5546073682986765e-05, "loss": 2.0564, "step": 13530 }, { "epoch": 0.33, "learning_rate": 1.554540936426527e-05, "loss": 2.1577, "step": 13531 }, { "epoch": 0.33, "learning_rate": 1.5544745010201075e-05, "loss": 2.2382, "step": 13532 }, { "epoch": 0.33, "learning_rate": 1.5544080620798414e-05, "loss": 2.1826, "step": 13533 }, { "epoch": 0.33, "learning_rate": 1.5543416196061522e-05, "loss": 1.9598, "step": 13534 }, { "epoch": 0.33, "learning_rate": 1.5542751735994628e-05, "loss": 1.9177, "step": 13535 }, { "epoch": 0.33, "learning_rate": 1.5542087240601975e-05, "loss": 2.1005, "step": 13536 }, { "epoch": 0.33, "learning_rate": 1.554142270988779e-05, "loss": 2.0538, "step": 13537 }, { "epoch": 0.33, "learning_rate": 1.5540758143856315e-05, "loss": 2.0831, "step": 13538 }, { "epoch": 0.33, "learning_rate": 1.554009354251178e-05, "loss": 2.1479, "step": 13539 }, { "epoch": 0.33, "learning_rate": 1.5539428905858426e-05, "loss": 2.1085, "step": 13540 }, { "epoch": 0.33, "learning_rate": 1.5538764233900484e-05, "loss": 2.1238, "step": 13541 }, { "epoch": 0.33, "learning_rate": 1.5538099526642196e-05, "loss": 2.1483, "step": 13542 }, { "epoch": 0.33, "learning_rate": 1.5537434784087792e-05, "loss": 1.9867, "step": 13543 }, { "epoch": 0.33, "learning_rate": 1.5536770006241516e-05, "loss": 2.019, "step": 13544 }, { "epoch": 0.33, "learning_rate": 1.5536105193107593e-05, "loss": 2.1247, "step": 13545 }, { "epoch": 0.33, "learning_rate": 1.5535440344690273e-05, "loss": 1.8891, "step": 13546 }, { "epoch": 0.33, "learning_rate": 1.5534775460993786e-05, "loss": 2.112, "step": 13547 }, { "epoch": 0.33, "learning_rate": 1.553411054202237e-05, "loss": 2.0585, "step": 13548 }, { "epoch": 0.33, "learning_rate": 1.5533445587780268e-05, "loss": 2.1884, "step": 13549 }, { "epoch": 0.33, "learning_rate": 1.553278059827171e-05, "loss": 2.0105, "step": 13550 }, { "epoch": 0.33, "learning_rate": 1.5532115573500937e-05, "loss": 2.0949, "step": 13551 }, { "epoch": 0.33, "learning_rate": 1.553145051347219e-05, "loss": 2.1857, "step": 13552 }, { "epoch": 0.33, "learning_rate": 1.553078541818971e-05, "loss": 1.9748, "step": 13553 }, { "epoch": 0.33, "learning_rate": 1.5530120287657732e-05, "loss": 1.9896, "step": 13554 }, { "epoch": 0.33, "learning_rate": 1.552945512188049e-05, "loss": 2.1733, "step": 13555 }, { "epoch": 0.33, "learning_rate": 1.5528789920862232e-05, "loss": 2.1502, "step": 13556 }, { "epoch": 0.33, "learning_rate": 1.5528124684607194e-05, "loss": 2.1521, "step": 13557 }, { "epoch": 0.33, "learning_rate": 1.5527459413119613e-05, "loss": 2.0434, "step": 13558 }, { "epoch": 0.33, "learning_rate": 1.5526794106403735e-05, "loss": 2.2486, "step": 13559 }, { "epoch": 0.33, "learning_rate": 1.5526128764463796e-05, "loss": 2.1076, "step": 13560 }, { "epoch": 0.33, "learning_rate": 1.5525463387304038e-05, "loss": 2.0811, "step": 13561 }, { "epoch": 0.33, "learning_rate": 1.55247979749287e-05, "loss": 2.0452, "step": 13562 }, { "epoch": 0.33, "learning_rate": 1.5524132527342025e-05, "loss": 2.1219, "step": 13563 }, { "epoch": 0.33, "learning_rate": 1.552346704454825e-05, "loss": 2.0787, "step": 13564 }, { "epoch": 0.33, "learning_rate": 1.5522801526551624e-05, "loss": 2.0062, "step": 13565 }, { "epoch": 0.33, "learning_rate": 1.552213597335638e-05, "loss": 2.0798, "step": 13566 }, { "epoch": 0.33, "learning_rate": 1.552147038496677e-05, "loss": 2.18, "step": 13567 }, { "epoch": 0.33, "learning_rate": 1.5520804761387023e-05, "loss": 2.0972, "step": 13568 }, { "epoch": 0.33, "learning_rate": 1.552013910262139e-05, "loss": 2.0319, "step": 13569 }, { "epoch": 0.33, "learning_rate": 1.5519473408674114e-05, "loss": 2.1921, "step": 13570 }, { "epoch": 0.33, "learning_rate": 1.551880767954943e-05, "loss": 2.0412, "step": 13571 }, { "epoch": 0.33, "learning_rate": 1.5518141915251594e-05, "loss": 2.2054, "step": 13572 }, { "epoch": 0.33, "learning_rate": 1.5517476115784836e-05, "loss": 2.0587, "step": 13573 }, { "epoch": 0.33, "learning_rate": 1.5516810281153404e-05, "loss": 2.2285, "step": 13574 }, { "epoch": 0.33, "learning_rate": 1.5516144411361543e-05, "loss": 1.9621, "step": 13575 }, { "epoch": 0.33, "learning_rate": 1.5515478506413498e-05, "loss": 2.182, "step": 13576 }, { "epoch": 0.33, "learning_rate": 1.551481256631351e-05, "loss": 2.1358, "step": 13577 }, { "epoch": 0.33, "learning_rate": 1.5514146591065825e-05, "loss": 2.1024, "step": 13578 }, { "epoch": 0.33, "learning_rate": 1.5513480580674684e-05, "loss": 2.0737, "step": 13579 }, { "epoch": 0.33, "learning_rate": 1.5512814535144337e-05, "loss": 2.0249, "step": 13580 }, { "epoch": 0.33, "learning_rate": 1.5512148454479026e-05, "loss": 2.2402, "step": 13581 }, { "epoch": 0.33, "learning_rate": 1.5511482338682996e-05, "loss": 2.1178, "step": 13582 }, { "epoch": 0.33, "learning_rate": 1.5510816187760493e-05, "loss": 2.1077, "step": 13583 }, { "epoch": 0.33, "learning_rate": 1.5510150001715763e-05, "loss": 2.1847, "step": 13584 }, { "epoch": 0.33, "learning_rate": 1.550948378055305e-05, "loss": 2.1381, "step": 13585 }, { "epoch": 0.33, "learning_rate": 1.5508817524276602e-05, "loss": 2.2297, "step": 13586 }, { "epoch": 0.33, "learning_rate": 1.5508151232890666e-05, "loss": 2.1346, "step": 13587 }, { "epoch": 0.33, "learning_rate": 1.5507484906399487e-05, "loss": 2.1147, "step": 13588 }, { "epoch": 0.33, "learning_rate": 1.550681854480731e-05, "loss": 1.9586, "step": 13589 }, { "epoch": 0.33, "learning_rate": 1.5506152148118388e-05, "loss": 2.0943, "step": 13590 }, { "epoch": 0.34, "learning_rate": 1.550548571633696e-05, "loss": 2.0527, "step": 13591 }, { "epoch": 0.34, "learning_rate": 1.550481924946728e-05, "loss": 1.8088, "step": 13592 }, { "epoch": 0.34, "learning_rate": 1.5504152747513588e-05, "loss": 2.1978, "step": 13593 }, { "epoch": 0.34, "learning_rate": 1.5503486210480142e-05, "loss": 2.2202, "step": 13594 }, { "epoch": 0.34, "learning_rate": 1.550281963837118e-05, "loss": 2.1151, "step": 13595 }, { "epoch": 0.34, "learning_rate": 1.550215303119096e-05, "loss": 1.9454, "step": 13596 }, { "epoch": 0.34, "learning_rate": 1.550148638894372e-05, "loss": 2.0825, "step": 13597 }, { "epoch": 0.34, "learning_rate": 1.550081971163372e-05, "loss": 2.0036, "step": 13598 }, { "epoch": 0.34, "learning_rate": 1.55001529992652e-05, "loss": 2.0315, "step": 13599 }, { "epoch": 0.34, "learning_rate": 1.5499486251842416e-05, "loss": 1.9635, "step": 13600 }, { "epoch": 0.34, "learning_rate": 1.549881946936961e-05, "loss": 2.2402, "step": 13601 }, { "epoch": 0.34, "learning_rate": 1.549815265185104e-05, "loss": 2.0405, "step": 13602 }, { "epoch": 0.34, "learning_rate": 1.5497485799290946e-05, "loss": 1.9528, "step": 13603 }, { "epoch": 0.34, "learning_rate": 1.5496818911693585e-05, "loss": 2.1823, "step": 13604 }, { "epoch": 0.34, "learning_rate": 1.549615198906321e-05, "loss": 2.1151, "step": 13605 }, { "epoch": 0.34, "learning_rate": 1.549548503140406e-05, "loss": 1.9603, "step": 13606 }, { "epoch": 0.34, "learning_rate": 1.54948180387204e-05, "loss": 2.1534, "step": 13607 }, { "epoch": 0.34, "learning_rate": 1.549415101101647e-05, "loss": 2.0569, "step": 13608 }, { "epoch": 0.34, "learning_rate": 1.5493483948296522e-05, "loss": 2.2388, "step": 13609 }, { "epoch": 0.34, "learning_rate": 1.5492816850564816e-05, "loss": 2.1705, "step": 13610 }, { "epoch": 0.34, "learning_rate": 1.5492149717825597e-05, "loss": 2.168, "step": 13611 }, { "epoch": 0.34, "learning_rate": 1.5491482550083118e-05, "loss": 2.1026, "step": 13612 }, { "epoch": 0.34, "learning_rate": 1.549081534734163e-05, "loss": 2.095, "step": 13613 }, { "epoch": 0.34, "learning_rate": 1.5490148109605386e-05, "loss": 2.3007, "step": 13614 }, { "epoch": 0.34, "learning_rate": 1.5489480836878642e-05, "loss": 1.8923, "step": 13615 }, { "epoch": 0.34, "learning_rate": 1.5488813529165645e-05, "loss": 2.009, "step": 13616 }, { "epoch": 0.34, "learning_rate": 1.5488146186470652e-05, "loss": 2.1924, "step": 13617 }, { "epoch": 0.34, "learning_rate": 1.548747880879791e-05, "loss": 2.2224, "step": 13618 }, { "epoch": 0.34, "learning_rate": 1.548681139615168e-05, "loss": 1.9712, "step": 13619 }, { "epoch": 0.34, "learning_rate": 1.5486143948536217e-05, "loss": 2.0582, "step": 13620 }, { "epoch": 0.34, "learning_rate": 1.548547646595577e-05, "loss": 2.168, "step": 13621 }, { "epoch": 0.34, "learning_rate": 1.5484808948414588e-05, "loss": 2.1167, "step": 13622 }, { "epoch": 0.34, "learning_rate": 1.5484141395916937e-05, "loss": 2.0335, "step": 13623 }, { "epoch": 0.34, "learning_rate": 1.5483473808467064e-05, "loss": 2.1255, "step": 13624 }, { "epoch": 0.34, "learning_rate": 1.5482806186069225e-05, "loss": 2.3389, "step": 13625 }, { "epoch": 0.34, "learning_rate": 1.548213852872768e-05, "loss": 2.1103, "step": 13626 }, { "epoch": 0.34, "learning_rate": 1.548147083644667e-05, "loss": 2.2827, "step": 13627 }, { "epoch": 0.34, "learning_rate": 1.5480803109230463e-05, "loss": 2.1659, "step": 13628 }, { "epoch": 0.34, "learning_rate": 1.5480135347083317e-05, "loss": 1.9736, "step": 13629 }, { "epoch": 0.34, "learning_rate": 1.5479467550009477e-05, "loss": 2.0617, "step": 13630 }, { "epoch": 0.34, "learning_rate": 1.547879971801321e-05, "loss": 2.1369, "step": 13631 }, { "epoch": 0.34, "learning_rate": 1.5478131851098758e-05, "loss": 2.237, "step": 13632 }, { "epoch": 0.34, "learning_rate": 1.5477463949270394e-05, "loss": 2.0379, "step": 13633 }, { "epoch": 0.34, "learning_rate": 1.5476796012532364e-05, "loss": 2.1898, "step": 13634 }, { "epoch": 0.34, "learning_rate": 1.547612804088893e-05, "loss": 2.0069, "step": 13635 }, { "epoch": 0.34, "learning_rate": 1.547546003434435e-05, "loss": 1.9997, "step": 13636 }, { "epoch": 0.34, "learning_rate": 1.547479199290287e-05, "loss": 2.2205, "step": 13637 }, { "epoch": 0.34, "learning_rate": 1.5474123916568763e-05, "loss": 2.1201, "step": 13638 }, { "epoch": 0.34, "learning_rate": 1.547345580534628e-05, "loss": 2.2091, "step": 13639 }, { "epoch": 0.34, "learning_rate": 1.5472787659239678e-05, "loss": 2.0714, "step": 13640 }, { "epoch": 0.34, "learning_rate": 1.547211947825322e-05, "loss": 2.0112, "step": 13641 }, { "epoch": 0.34, "learning_rate": 1.5471451262391154e-05, "loss": 2.095, "step": 13642 }, { "epoch": 0.34, "learning_rate": 1.5470783011657753e-05, "loss": 2.1823, "step": 13643 }, { "epoch": 0.34, "learning_rate": 1.5470114726057267e-05, "loss": 2.0548, "step": 13644 }, { "epoch": 0.34, "learning_rate": 1.5469446405593958e-05, "loss": 2.0594, "step": 13645 }, { "epoch": 0.34, "learning_rate": 1.5468778050272083e-05, "loss": 2.1043, "step": 13646 }, { "epoch": 0.34, "learning_rate": 1.54681096600959e-05, "loss": 1.9861, "step": 13647 }, { "epoch": 0.34, "learning_rate": 1.5467441235069677e-05, "loss": 2.2657, "step": 13648 }, { "epoch": 0.34, "learning_rate": 1.5466772775197668e-05, "loss": 2.2307, "step": 13649 }, { "epoch": 0.34, "learning_rate": 1.5466104280484134e-05, "loss": 2.1036, "step": 13650 }, { "epoch": 0.34, "learning_rate": 1.5465435750933338e-05, "loss": 2.0484, "step": 13651 }, { "epoch": 0.34, "learning_rate": 1.5464767186549538e-05, "loss": 2.0349, "step": 13652 }, { "epoch": 0.34, "learning_rate": 1.5464098587336994e-05, "loss": 1.9016, "step": 13653 }, { "epoch": 0.34, "learning_rate": 1.546342995329997e-05, "loss": 2.1666, "step": 13654 }, { "epoch": 0.34, "learning_rate": 1.5462761284442727e-05, "loss": 2.0201, "step": 13655 }, { "epoch": 0.34, "learning_rate": 1.5462092580769526e-05, "loss": 2.2088, "step": 13656 }, { "epoch": 0.34, "learning_rate": 1.546142384228463e-05, "loss": 2.0688, "step": 13657 }, { "epoch": 0.34, "learning_rate": 1.54607550689923e-05, "loss": 2.1374, "step": 13658 }, { "epoch": 0.34, "learning_rate": 1.5460086260896793e-05, "loss": 2.1073, "step": 13659 }, { "epoch": 0.34, "learning_rate": 1.5459417418002384e-05, "loss": 2.1498, "step": 13660 }, { "epoch": 0.34, "learning_rate": 1.5458748540313324e-05, "loss": 2.1811, "step": 13661 }, { "epoch": 0.34, "learning_rate": 1.545807962783388e-05, "loss": 2.1184, "step": 13662 }, { "epoch": 0.34, "learning_rate": 1.545741068056832e-05, "loss": 2.0987, "step": 13663 }, { "epoch": 0.34, "learning_rate": 1.54567416985209e-05, "loss": 2.1887, "step": 13664 }, { "epoch": 0.34, "learning_rate": 1.545607268169589e-05, "loss": 2.1434, "step": 13665 }, { "epoch": 0.34, "learning_rate": 1.5455403630097546e-05, "loss": 2.0437, "step": 13666 }, { "epoch": 0.34, "learning_rate": 1.545473454373014e-05, "loss": 2.0809, "step": 13667 }, { "epoch": 0.34, "learning_rate": 1.5454065422597933e-05, "loss": 2.0006, "step": 13668 }, { "epoch": 0.34, "learning_rate": 1.545339626670519e-05, "loss": 2.1095, "step": 13669 }, { "epoch": 0.34, "learning_rate": 1.5452727076056175e-05, "loss": 1.9759, "step": 13670 }, { "epoch": 0.34, "learning_rate": 1.545205785065515e-05, "loss": 2.1676, "step": 13671 }, { "epoch": 0.34, "learning_rate": 1.545138859050639e-05, "loss": 2.0761, "step": 13672 }, { "epoch": 0.34, "learning_rate": 1.545071929561415e-05, "loss": 1.9918, "step": 13673 }, { "epoch": 0.34, "learning_rate": 1.54500499659827e-05, "loss": 2.1092, "step": 13674 }, { "epoch": 0.34, "learning_rate": 1.5449380601616305e-05, "loss": 2.1574, "step": 13675 }, { "epoch": 0.34, "learning_rate": 1.5448711202519232e-05, "loss": 2.1265, "step": 13676 }, { "epoch": 0.34, "learning_rate": 1.5448041768695746e-05, "loss": 2.082, "step": 13677 }, { "epoch": 0.34, "learning_rate": 1.5447372300150117e-05, "loss": 2.015, "step": 13678 }, { "epoch": 0.34, "learning_rate": 1.544670279688661e-05, "loss": 2.1786, "step": 13679 }, { "epoch": 0.34, "learning_rate": 1.5446033258909486e-05, "loss": 2.0153, "step": 13680 }, { "epoch": 0.34, "learning_rate": 1.5445363686223015e-05, "loss": 2.0758, "step": 13681 }, { "epoch": 0.34, "learning_rate": 1.544469407883147e-05, "loss": 2.1729, "step": 13682 }, { "epoch": 0.34, "learning_rate": 1.544402443673912e-05, "loss": 2.1571, "step": 13683 }, { "epoch": 0.34, "learning_rate": 1.5443354759950225e-05, "loss": 2.0419, "step": 13684 }, { "epoch": 0.34, "learning_rate": 1.5442685048469054e-05, "loss": 2.1588, "step": 13685 }, { "epoch": 0.34, "learning_rate": 1.544201530229988e-05, "loss": 2.0354, "step": 13686 }, { "epoch": 0.34, "learning_rate": 1.5441345521446968e-05, "loss": 2.0127, "step": 13687 }, { "epoch": 0.34, "learning_rate": 1.544067570591459e-05, "loss": 1.9974, "step": 13688 }, { "epoch": 0.34, "learning_rate": 1.544000585570701e-05, "loss": 2.0665, "step": 13689 }, { "epoch": 0.34, "learning_rate": 1.54393359708285e-05, "loss": 2.0209, "step": 13690 }, { "epoch": 0.34, "learning_rate": 1.5438666051283327e-05, "loss": 2.1266, "step": 13691 }, { "epoch": 0.34, "learning_rate": 1.543799609707577e-05, "loss": 2.038, "step": 13692 }, { "epoch": 0.34, "learning_rate": 1.5437326108210085e-05, "loss": 2.2377, "step": 13693 }, { "epoch": 0.34, "learning_rate": 1.5436656084690552e-05, "loss": 2.1712, "step": 13694 }, { "epoch": 0.34, "learning_rate": 1.5435986026521436e-05, "loss": 2.0391, "step": 13695 }, { "epoch": 0.34, "learning_rate": 1.543531593370701e-05, "loss": 2.1582, "step": 13696 }, { "epoch": 0.34, "learning_rate": 1.5434645806251546e-05, "loss": 2.1554, "step": 13697 }, { "epoch": 0.34, "learning_rate": 1.543397564415931e-05, "loss": 2.2448, "step": 13698 }, { "epoch": 0.34, "learning_rate": 1.543330544743458e-05, "loss": 2.1582, "step": 13699 }, { "epoch": 0.34, "learning_rate": 1.543263521608162e-05, "loss": 2.0729, "step": 13700 }, { "epoch": 0.34, "learning_rate": 1.5431964950104706e-05, "loss": 2.0846, "step": 13701 }, { "epoch": 0.34, "learning_rate": 1.5431294649508113e-05, "loss": 1.8486, "step": 13702 }, { "epoch": 0.34, "learning_rate": 1.5430624314296105e-05, "loss": 2.2225, "step": 13703 }, { "epoch": 0.34, "learning_rate": 1.5429953944472962e-05, "loss": 2.0861, "step": 13704 }, { "epoch": 0.34, "learning_rate": 1.5429283540042947e-05, "loss": 1.9906, "step": 13705 }, { "epoch": 0.34, "learning_rate": 1.5428613101010345e-05, "loss": 2.082, "step": 13706 }, { "epoch": 0.34, "learning_rate": 1.5427942627379418e-05, "loss": 2.0547, "step": 13707 }, { "epoch": 0.34, "learning_rate": 1.542727211915445e-05, "loss": 2.141, "step": 13708 }, { "epoch": 0.34, "learning_rate": 1.54266015763397e-05, "loss": 1.8934, "step": 13709 }, { "epoch": 0.34, "learning_rate": 1.5425930998939456e-05, "loss": 1.963, "step": 13710 }, { "epoch": 0.34, "learning_rate": 1.5425260386957983e-05, "loss": 1.9861, "step": 13711 }, { "epoch": 0.34, "learning_rate": 1.5424589740399557e-05, "loss": 2.1971, "step": 13712 }, { "epoch": 0.34, "learning_rate": 1.5423919059268456e-05, "loss": 1.9523, "step": 13713 }, { "epoch": 0.34, "learning_rate": 1.5423248343568947e-05, "loss": 2.0164, "step": 13714 }, { "epoch": 0.34, "learning_rate": 1.542257759330531e-05, "loss": 2.0079, "step": 13715 }, { "epoch": 0.34, "learning_rate": 1.542190680848182e-05, "loss": 2.2128, "step": 13716 }, { "epoch": 0.34, "learning_rate": 1.542123598910275e-05, "loss": 1.983, "step": 13717 }, { "epoch": 0.34, "learning_rate": 1.5420565135172378e-05, "loss": 2.1438, "step": 13718 }, { "epoch": 0.34, "learning_rate": 1.5419894246694976e-05, "loss": 2.153, "step": 13719 }, { "epoch": 0.34, "learning_rate": 1.5419223323674822e-05, "loss": 2.2222, "step": 13720 }, { "epoch": 0.34, "learning_rate": 1.5418552366116192e-05, "loss": 1.9712, "step": 13721 }, { "epoch": 0.34, "learning_rate": 1.541788137402336e-05, "loss": 2.2272, "step": 13722 }, { "epoch": 0.34, "learning_rate": 1.5417210347400605e-05, "loss": 2.2203, "step": 13723 }, { "epoch": 0.34, "learning_rate": 1.5416539286252207e-05, "loss": 2.2931, "step": 13724 }, { "epoch": 0.34, "learning_rate": 1.5415868190582432e-05, "loss": 2.0661, "step": 13725 }, { "epoch": 0.34, "learning_rate": 1.5415197060395573e-05, "loss": 1.8932, "step": 13726 }, { "epoch": 0.34, "learning_rate": 1.5414525895695893e-05, "loss": 2.1155, "step": 13727 }, { "epoch": 0.34, "learning_rate": 1.5413854696487677e-05, "loss": 2.1806, "step": 13728 }, { "epoch": 0.34, "learning_rate": 1.5413183462775197e-05, "loss": 1.9937, "step": 13729 }, { "epoch": 0.34, "learning_rate": 1.5412512194562738e-05, "loss": 2.1938, "step": 13730 }, { "epoch": 0.34, "learning_rate": 1.5411840891854573e-05, "loss": 2.0538, "step": 13731 }, { "epoch": 0.34, "learning_rate": 1.541116955465498e-05, "loss": 2.1505, "step": 13732 }, { "epoch": 0.34, "learning_rate": 1.5410498182968246e-05, "loss": 2.0119, "step": 13733 }, { "epoch": 0.34, "learning_rate": 1.540982677679864e-05, "loss": 2.0484, "step": 13734 }, { "epoch": 0.34, "learning_rate": 1.5409155336150444e-05, "loss": 2.1197, "step": 13735 }, { "epoch": 0.34, "learning_rate": 1.5408483861027944e-05, "loss": 2.2569, "step": 13736 }, { "epoch": 0.34, "learning_rate": 1.5407812351435408e-05, "loss": 2.0936, "step": 13737 }, { "epoch": 0.34, "learning_rate": 1.540714080737712e-05, "loss": 2.2204, "step": 13738 }, { "epoch": 0.34, "learning_rate": 1.5406469228857366e-05, "loss": 2.1252, "step": 13739 }, { "epoch": 0.34, "learning_rate": 1.5405797615880423e-05, "loss": 2.291, "step": 13740 }, { "epoch": 0.34, "learning_rate": 1.5405125968450568e-05, "loss": 2.1845, "step": 13741 }, { "epoch": 0.34, "learning_rate": 1.5404454286572082e-05, "loss": 2.0227, "step": 13742 }, { "epoch": 0.34, "learning_rate": 1.540378257024925e-05, "loss": 2.1291, "step": 13743 }, { "epoch": 0.34, "learning_rate": 1.5403110819486347e-05, "loss": 2.3212, "step": 13744 }, { "epoch": 0.34, "learning_rate": 1.5402439034287663e-05, "loss": 2.2095, "step": 13745 }, { "epoch": 0.34, "learning_rate": 1.5401767214657473e-05, "loss": 2.0746, "step": 13746 }, { "epoch": 0.34, "learning_rate": 1.540109536060006e-05, "loss": 2.0316, "step": 13747 }, { "epoch": 0.34, "learning_rate": 1.5400423472119704e-05, "loss": 2.0036, "step": 13748 }, { "epoch": 0.34, "learning_rate": 1.5399751549220687e-05, "loss": 2.17, "step": 13749 }, { "epoch": 0.34, "learning_rate": 1.53990795919073e-05, "loss": 2.0614, "step": 13750 }, { "epoch": 0.34, "learning_rate": 1.5398407600183817e-05, "loss": 1.9276, "step": 13751 }, { "epoch": 0.34, "learning_rate": 1.5397735574054522e-05, "loss": 1.9362, "step": 13752 }, { "epoch": 0.34, "learning_rate": 1.5397063513523696e-05, "loss": 2.1722, "step": 13753 }, { "epoch": 0.34, "learning_rate": 1.539639141859563e-05, "loss": 1.9473, "step": 13754 }, { "epoch": 0.34, "learning_rate": 1.5395719289274604e-05, "loss": 2.0594, "step": 13755 }, { "epoch": 0.34, "learning_rate": 1.53950471255649e-05, "loss": 2.3411, "step": 13756 }, { "epoch": 0.34, "learning_rate": 1.53943749274708e-05, "loss": 2.0857, "step": 13757 }, { "epoch": 0.34, "learning_rate": 1.5393702694996588e-05, "loss": 2.2573, "step": 13758 }, { "epoch": 0.34, "learning_rate": 1.5393030428146558e-05, "loss": 2.0114, "step": 13759 }, { "epoch": 0.34, "learning_rate": 1.5392358126924984e-05, "loss": 2.085, "step": 13760 }, { "epoch": 0.34, "learning_rate": 1.5391685791336153e-05, "loss": 2.1395, "step": 13761 }, { "epoch": 0.34, "learning_rate": 1.5391013421384354e-05, "loss": 1.9994, "step": 13762 }, { "epoch": 0.34, "learning_rate": 1.5390341017073866e-05, "loss": 2.1666, "step": 13763 }, { "epoch": 0.34, "learning_rate": 1.5389668578408983e-05, "loss": 1.9191, "step": 13764 }, { "epoch": 0.34, "learning_rate": 1.5388996105393982e-05, "loss": 2.0949, "step": 13765 }, { "epoch": 0.34, "learning_rate": 1.538832359803315e-05, "loss": 1.9912, "step": 13766 }, { "epoch": 0.34, "learning_rate": 1.5387651056330782e-05, "loss": 2.1707, "step": 13767 }, { "epoch": 0.34, "learning_rate": 1.5386978480291154e-05, "loss": 2.0775, "step": 13768 }, { "epoch": 0.34, "learning_rate": 1.538630586991856e-05, "loss": 2.0836, "step": 13769 }, { "epoch": 0.34, "learning_rate": 1.538563322521728e-05, "loss": 2.0616, "step": 13770 }, { "epoch": 0.34, "learning_rate": 1.5384960546191608e-05, "loss": 2.361, "step": 13771 }, { "epoch": 0.34, "learning_rate": 1.5384287832845822e-05, "loss": 1.9504, "step": 13772 }, { "epoch": 0.34, "learning_rate": 1.5383615085184217e-05, "loss": 2.1563, "step": 13773 }, { "epoch": 0.34, "learning_rate": 1.5382942303211083e-05, "loss": 1.9174, "step": 13774 }, { "epoch": 0.34, "learning_rate": 1.53822694869307e-05, "loss": 2.2438, "step": 13775 }, { "epoch": 0.34, "learning_rate": 1.5381596636347357e-05, "loss": 2.1252, "step": 13776 }, { "epoch": 0.34, "learning_rate": 1.5380923751465347e-05, "loss": 2.2306, "step": 13777 }, { "epoch": 0.34, "learning_rate": 1.5380250832288955e-05, "loss": 2.1667, "step": 13778 }, { "epoch": 0.34, "learning_rate": 1.5379577878822475e-05, "loss": 2.2252, "step": 13779 }, { "epoch": 0.34, "learning_rate": 1.537890489107019e-05, "loss": 2.0686, "step": 13780 }, { "epoch": 0.34, "learning_rate": 1.5378231869036393e-05, "loss": 2.1889, "step": 13781 }, { "epoch": 0.34, "learning_rate": 1.5377558812725367e-05, "loss": 2.0116, "step": 13782 }, { "epoch": 0.34, "learning_rate": 1.5376885722141413e-05, "loss": 2.2203, "step": 13783 }, { "epoch": 0.34, "learning_rate": 1.5376212597288806e-05, "loss": 2.2996, "step": 13784 }, { "epoch": 0.34, "learning_rate": 1.537553943817185e-05, "loss": 2.105, "step": 13785 }, { "epoch": 0.34, "learning_rate": 1.5374866244794827e-05, "loss": 2.1385, "step": 13786 }, { "epoch": 0.34, "learning_rate": 1.537419301716203e-05, "loss": 2.0513, "step": 13787 }, { "epoch": 0.34, "learning_rate": 1.5373519755277747e-05, "loss": 2.0491, "step": 13788 }, { "epoch": 0.34, "learning_rate": 1.5372846459146276e-05, "loss": 2.1705, "step": 13789 }, { "epoch": 0.34, "learning_rate": 1.5372173128771903e-05, "loss": 2.3173, "step": 13790 }, { "epoch": 0.34, "learning_rate": 1.537149976415892e-05, "loss": 2.346, "step": 13791 }, { "epoch": 0.34, "learning_rate": 1.5370826365311615e-05, "loss": 2.3194, "step": 13792 }, { "epoch": 0.34, "learning_rate": 1.5370152932234283e-05, "loss": 2.186, "step": 13793 }, { "epoch": 0.34, "learning_rate": 1.536947946493122e-05, "loss": 2.1359, "step": 13794 }, { "epoch": 0.34, "learning_rate": 1.5368805963406714e-05, "loss": 1.9304, "step": 13795 }, { "epoch": 0.34, "learning_rate": 1.536813242766506e-05, "loss": 2.0877, "step": 13796 }, { "epoch": 0.34, "learning_rate": 1.5367458857710543e-05, "loss": 2.1785, "step": 13797 }, { "epoch": 0.34, "learning_rate": 1.5366785253547463e-05, "loss": 2.1533, "step": 13798 }, { "epoch": 0.34, "learning_rate": 1.5366111615180113e-05, "loss": 2.3415, "step": 13799 }, { "epoch": 0.34, "learning_rate": 1.5365437942612784e-05, "loss": 2.112, "step": 13800 }, { "epoch": 0.34, "learning_rate": 1.536476423584977e-05, "loss": 2.0672, "step": 13801 }, { "epoch": 0.34, "learning_rate": 1.5364090494895364e-05, "loss": 2.1777, "step": 13802 }, { "epoch": 0.34, "learning_rate": 1.5363416719753866e-05, "loss": 2.0543, "step": 13803 }, { "epoch": 0.34, "learning_rate": 1.536274291042956e-05, "loss": 2.0424, "step": 13804 }, { "epoch": 0.34, "learning_rate": 1.536206906692675e-05, "loss": 2.1172, "step": 13805 }, { "epoch": 0.34, "learning_rate": 1.5361395189249723e-05, "loss": 2.0785, "step": 13806 }, { "epoch": 0.34, "learning_rate": 1.5360721277402783e-05, "loss": 2.1248, "step": 13807 }, { "epoch": 0.34, "learning_rate": 1.5360047331390216e-05, "loss": 2.1043, "step": 13808 }, { "epoch": 0.34, "learning_rate": 1.535937335121632e-05, "loss": 2.1136, "step": 13809 }, { "epoch": 0.34, "learning_rate": 1.535869933688539e-05, "loss": 2.0693, "step": 13810 }, { "epoch": 0.34, "learning_rate": 1.535802528840172e-05, "loss": 2.0695, "step": 13811 }, { "epoch": 0.34, "learning_rate": 1.5357351205769615e-05, "loss": 1.9701, "step": 13812 }, { "epoch": 0.34, "learning_rate": 1.5356677088993363e-05, "loss": 2.11, "step": 13813 }, { "epoch": 0.34, "learning_rate": 1.5356002938077265e-05, "loss": 2.2157, "step": 13814 }, { "epoch": 0.34, "learning_rate": 1.535532875302561e-05, "loss": 2.1039, "step": 13815 }, { "epoch": 0.34, "learning_rate": 1.53546545338427e-05, "loss": 2.0783, "step": 13816 }, { "epoch": 0.34, "learning_rate": 1.535398028053283e-05, "loss": 2.1125, "step": 13817 }, { "epoch": 0.34, "learning_rate": 1.5353305993100304e-05, "loss": 1.964, "step": 13818 }, { "epoch": 0.34, "learning_rate": 1.5352631671549412e-05, "loss": 2.2121, "step": 13819 }, { "epoch": 0.34, "learning_rate": 1.5351957315884455e-05, "loss": 2.1504, "step": 13820 }, { "epoch": 0.34, "learning_rate": 1.5351282926109725e-05, "loss": 2.0091, "step": 13821 }, { "epoch": 0.34, "learning_rate": 1.5350608502229528e-05, "loss": 2.2153, "step": 13822 }, { "epoch": 0.34, "learning_rate": 1.534993404424816e-05, "loss": 2.1159, "step": 13823 }, { "epoch": 0.34, "learning_rate": 1.5349259552169923e-05, "loss": 2.0765, "step": 13824 }, { "epoch": 0.34, "learning_rate": 1.5348585025999107e-05, "loss": 2.2705, "step": 13825 }, { "epoch": 0.34, "learning_rate": 1.5347910465740016e-05, "loss": 2.0036, "step": 13826 }, { "epoch": 0.34, "learning_rate": 1.534723587139695e-05, "loss": 2.1103, "step": 13827 }, { "epoch": 0.34, "learning_rate": 1.53465612429742e-05, "loss": 1.9622, "step": 13828 }, { "epoch": 0.34, "learning_rate": 1.534588658047608e-05, "loss": 2.2067, "step": 13829 }, { "epoch": 0.34, "learning_rate": 1.534521188390688e-05, "loss": 2.127, "step": 13830 }, { "epoch": 0.34, "learning_rate": 1.5344537153270906e-05, "loss": 1.9785, "step": 13831 }, { "epoch": 0.34, "learning_rate": 1.534386238857245e-05, "loss": 2.0284, "step": 13832 }, { "epoch": 0.34, "learning_rate": 1.5343187589815825e-05, "loss": 2.171, "step": 13833 }, { "epoch": 0.34, "learning_rate": 1.5342512757005322e-05, "loss": 2.0897, "step": 13834 }, { "epoch": 0.34, "learning_rate": 1.5341837890145238e-05, "loss": 2.1917, "step": 13835 }, { "epoch": 0.34, "learning_rate": 1.5341162989239883e-05, "loss": 2.2798, "step": 13836 }, { "epoch": 0.34, "learning_rate": 1.5340488054293555e-05, "loss": 2.2414, "step": 13837 }, { "epoch": 0.34, "learning_rate": 1.533981308531056e-05, "loss": 2.1766, "step": 13838 }, { "epoch": 0.34, "learning_rate": 1.5339138082295194e-05, "loss": 2.2413, "step": 13839 }, { "epoch": 0.34, "learning_rate": 1.533846304525176e-05, "loss": 2.2676, "step": 13840 }, { "epoch": 0.34, "learning_rate": 1.5337787974184565e-05, "loss": 2.1674, "step": 13841 }, { "epoch": 0.34, "learning_rate": 1.5337112869097907e-05, "loss": 2.0146, "step": 13842 }, { "epoch": 0.34, "learning_rate": 1.5336437729996086e-05, "loss": 2.0211, "step": 13843 }, { "epoch": 0.34, "learning_rate": 1.533576255688341e-05, "loss": 1.9487, "step": 13844 }, { "epoch": 0.34, "learning_rate": 1.533508734976418e-05, "loss": 2.0446, "step": 13845 }, { "epoch": 0.34, "learning_rate": 1.5334412108642695e-05, "loss": 2.1468, "step": 13846 }, { "epoch": 0.34, "learning_rate": 1.533373683352327e-05, "loss": 2.14, "step": 13847 }, { "epoch": 0.34, "learning_rate": 1.53330615244102e-05, "loss": 2.2815, "step": 13848 }, { "epoch": 0.34, "learning_rate": 1.5332386181307793e-05, "loss": 2.1035, "step": 13849 }, { "epoch": 0.34, "learning_rate": 1.5331710804220346e-05, "loss": 2.1301, "step": 13850 }, { "epoch": 0.34, "learning_rate": 1.533103539315217e-05, "loss": 2.1113, "step": 13851 }, { "epoch": 0.34, "learning_rate": 1.5330359948107572e-05, "loss": 2.0242, "step": 13852 }, { "epoch": 0.34, "learning_rate": 1.532968446909085e-05, "loss": 2.1532, "step": 13853 }, { "epoch": 0.34, "learning_rate": 1.5329008956106312e-05, "loss": 2.1921, "step": 13854 }, { "epoch": 0.34, "learning_rate": 1.5328333409158263e-05, "loss": 2.1279, "step": 13855 }, { "epoch": 0.34, "learning_rate": 1.532765782825101e-05, "loss": 1.9563, "step": 13856 }, { "epoch": 0.34, "learning_rate": 1.5326982213388858e-05, "loss": 2.1402, "step": 13857 }, { "epoch": 0.34, "learning_rate": 1.532630656457611e-05, "loss": 2.1323, "step": 13858 }, { "epoch": 0.34, "learning_rate": 1.5325630881817077e-05, "loss": 2.0217, "step": 13859 }, { "epoch": 0.34, "learning_rate": 1.532495516511606e-05, "loss": 2.1906, "step": 13860 }, { "epoch": 0.34, "learning_rate": 1.532427941447737e-05, "loss": 2.1984, "step": 13861 }, { "epoch": 0.34, "learning_rate": 1.5323603629905313e-05, "loss": 2.0263, "step": 13862 }, { "epoch": 0.34, "learning_rate": 1.5322927811404193e-05, "loss": 1.9992, "step": 13863 }, { "epoch": 0.34, "learning_rate": 1.532225195897832e-05, "loss": 2.075, "step": 13864 }, { "epoch": 0.34, "learning_rate": 1.5321576072632e-05, "loss": 2.2212, "step": 13865 }, { "epoch": 0.34, "learning_rate": 1.5320900152369543e-05, "loss": 2.116, "step": 13866 }, { "epoch": 0.34, "learning_rate": 1.5320224198195256e-05, "loss": 2.078, "step": 13867 }, { "epoch": 0.34, "learning_rate": 1.5319548210113447e-05, "loss": 2.0095, "step": 13868 }, { "epoch": 0.34, "learning_rate": 1.531887218812842e-05, "loss": 2.1153, "step": 13869 }, { "epoch": 0.34, "learning_rate": 1.5318196132244487e-05, "loss": 2.128, "step": 13870 }, { "epoch": 0.34, "learning_rate": 1.531752004246596e-05, "loss": 2.1027, "step": 13871 }, { "epoch": 0.34, "learning_rate": 1.5316843918797143e-05, "loss": 2.1854, "step": 13872 }, { "epoch": 0.34, "learning_rate": 1.5316167761242347e-05, "loss": 2.0685, "step": 13873 }, { "epoch": 0.34, "learning_rate": 1.5315491569805882e-05, "loss": 2.0777, "step": 13874 }, { "epoch": 0.34, "learning_rate": 1.531481534449206e-05, "loss": 2.2115, "step": 13875 }, { "epoch": 0.34, "learning_rate": 1.5314139085305182e-05, "loss": 2.1105, "step": 13876 }, { "epoch": 0.34, "learning_rate": 1.5313462792249568e-05, "loss": 2.1363, "step": 13877 }, { "epoch": 0.34, "learning_rate": 1.5312786465329525e-05, "loss": 2.2087, "step": 13878 }, { "epoch": 0.34, "learning_rate": 1.531211010454936e-05, "loss": 2.0987, "step": 13879 }, { "epoch": 0.34, "learning_rate": 1.5311433709913384e-05, "loss": 2.0938, "step": 13880 }, { "epoch": 0.34, "learning_rate": 1.5310757281425912e-05, "loss": 2.3234, "step": 13881 }, { "epoch": 0.34, "learning_rate": 1.5310080819091254e-05, "loss": 2.159, "step": 13882 }, { "epoch": 0.34, "learning_rate": 1.530940432291372e-05, "loss": 2.081, "step": 13883 }, { "epoch": 0.34, "learning_rate": 1.5308727792897623e-05, "loss": 2.0539, "step": 13884 }, { "epoch": 0.34, "learning_rate": 1.5308051229047273e-05, "loss": 2.1377, "step": 13885 }, { "epoch": 0.34, "learning_rate": 1.5307374631366982e-05, "loss": 2.1667, "step": 13886 }, { "epoch": 0.34, "learning_rate": 1.5306697999861064e-05, "loss": 2.0802, "step": 13887 }, { "epoch": 0.34, "learning_rate": 1.5306021334533826e-05, "loss": 2.1748, "step": 13888 }, { "epoch": 0.34, "learning_rate": 1.5305344635389593e-05, "loss": 2.0172, "step": 13889 }, { "epoch": 0.34, "learning_rate": 1.5304667902432662e-05, "loss": 2.1769, "step": 13890 }, { "epoch": 0.34, "learning_rate": 1.5303991135667356e-05, "loss": 2.1721, "step": 13891 }, { "epoch": 0.34, "learning_rate": 1.530331433509799e-05, "loss": 2.1493, "step": 13892 }, { "epoch": 0.34, "learning_rate": 1.5302637500728866e-05, "loss": 2.215, "step": 13893 }, { "epoch": 0.34, "learning_rate": 1.530196063256431e-05, "loss": 2.172, "step": 13894 }, { "epoch": 0.34, "learning_rate": 1.530128373060863e-05, "loss": 2.1655, "step": 13895 }, { "epoch": 0.34, "learning_rate": 1.5300606794866143e-05, "loss": 2.2794, "step": 13896 }, { "epoch": 0.34, "learning_rate": 1.5299929825341157e-05, "loss": 2.2699, "step": 13897 }, { "epoch": 0.34, "learning_rate": 1.5299252822037994e-05, "loss": 2.0748, "step": 13898 }, { "epoch": 0.34, "learning_rate": 1.5298575784960968e-05, "loss": 2.1622, "step": 13899 }, { "epoch": 0.34, "learning_rate": 1.5297898714114388e-05, "loss": 1.9105, "step": 13900 }, { "epoch": 0.34, "learning_rate": 1.5297221609502575e-05, "loss": 2.1682, "step": 13901 }, { "epoch": 0.34, "learning_rate": 1.529654447112984e-05, "loss": 2.1802, "step": 13902 }, { "epoch": 0.34, "learning_rate": 1.5295867299000503e-05, "loss": 2.1018, "step": 13903 }, { "epoch": 0.34, "learning_rate": 1.529519009311888e-05, "loss": 2.0659, "step": 13904 }, { "epoch": 0.34, "learning_rate": 1.529451285348928e-05, "loss": 2.0416, "step": 13905 }, { "epoch": 0.34, "learning_rate": 1.5293835580116025e-05, "loss": 1.8873, "step": 13906 }, { "epoch": 0.34, "learning_rate": 1.5293158273003435e-05, "loss": 2.1825, "step": 13907 }, { "epoch": 0.34, "learning_rate": 1.5292480932155816e-05, "loss": 2.0445, "step": 13908 }, { "epoch": 0.34, "learning_rate": 1.52918035575775e-05, "loss": 2.2304, "step": 13909 }, { "epoch": 0.34, "learning_rate": 1.5291126149272787e-05, "loss": 2.1119, "step": 13910 }, { "epoch": 0.34, "learning_rate": 1.529044870724601e-05, "loss": 1.9611, "step": 13911 }, { "epoch": 0.34, "learning_rate": 1.5289771231501475e-05, "loss": 2.1549, "step": 13912 }, { "epoch": 0.34, "learning_rate": 1.52890937220435e-05, "loss": 2.1767, "step": 13913 }, { "epoch": 0.34, "learning_rate": 1.5288416178876412e-05, "loss": 2.1038, "step": 13914 }, { "epoch": 0.34, "learning_rate": 1.5287738602004523e-05, "loss": 1.9632, "step": 13915 }, { "epoch": 0.34, "learning_rate": 1.528706099143215e-05, "loss": 1.9267, "step": 13916 }, { "epoch": 0.34, "learning_rate": 1.528638334716362e-05, "loss": 2.0302, "step": 13917 }, { "epoch": 0.34, "learning_rate": 1.5285705669203244e-05, "loss": 2.1554, "step": 13918 }, { "epoch": 0.34, "learning_rate": 1.5285027957555345e-05, "loss": 2.1504, "step": 13919 }, { "epoch": 0.34, "learning_rate": 1.5284350212224236e-05, "loss": 1.8535, "step": 13920 }, { "epoch": 0.34, "learning_rate": 1.5283672433214246e-05, "loss": 2.2516, "step": 13921 }, { "epoch": 0.34, "learning_rate": 1.5282994620529686e-05, "loss": 2.1248, "step": 13922 }, { "epoch": 0.34, "learning_rate": 1.5282316774174878e-05, "loss": 2.1636, "step": 13923 }, { "epoch": 0.34, "learning_rate": 1.528163889415415e-05, "loss": 2.1309, "step": 13924 }, { "epoch": 0.34, "learning_rate": 1.528096098047181e-05, "loss": 2.0394, "step": 13925 }, { "epoch": 0.34, "learning_rate": 1.528028303313219e-05, "loss": 2.2234, "step": 13926 }, { "epoch": 0.34, "learning_rate": 1.52796050521396e-05, "loss": 2.217, "step": 13927 }, { "epoch": 0.34, "learning_rate": 1.5278927037498366e-05, "loss": 2.102, "step": 13928 }, { "epoch": 0.34, "learning_rate": 1.5278248989212815e-05, "loss": 2.0785, "step": 13929 }, { "epoch": 0.34, "learning_rate": 1.527757090728726e-05, "loss": 2.0123, "step": 13930 }, { "epoch": 0.34, "learning_rate": 1.527689279172603e-05, "loss": 2.0448, "step": 13931 }, { "epoch": 0.34, "learning_rate": 1.5276214642533437e-05, "loss": 2.1217, "step": 13932 }, { "epoch": 0.34, "learning_rate": 1.527553645971381e-05, "loss": 2.0925, "step": 13933 }, { "epoch": 0.34, "learning_rate": 1.527485824327147e-05, "loss": 2.0318, "step": 13934 }, { "epoch": 0.34, "learning_rate": 1.5274179993210742e-05, "loss": 2.0147, "step": 13935 }, { "epoch": 0.34, "learning_rate": 1.5273501709535944e-05, "loss": 2.2584, "step": 13936 }, { "epoch": 0.34, "learning_rate": 1.5272823392251397e-05, "loss": 2.2595, "step": 13937 }, { "epoch": 0.34, "learning_rate": 1.5272145041361434e-05, "loss": 2.1213, "step": 13938 }, { "epoch": 0.34, "learning_rate": 1.527146665687037e-05, "loss": 2.0881, "step": 13939 }, { "epoch": 0.34, "learning_rate": 1.527078823878253e-05, "loss": 2.0455, "step": 13940 }, { "epoch": 0.34, "learning_rate": 1.5270109787102242e-05, "loss": 2.1346, "step": 13941 }, { "epoch": 0.34, "learning_rate": 1.5269431301833826e-05, "loss": 2.2264, "step": 13942 }, { "epoch": 0.34, "learning_rate": 1.5268752782981607e-05, "loss": 2.1272, "step": 13943 }, { "epoch": 0.34, "learning_rate": 1.5268074230549906e-05, "loss": 2.0793, "step": 13944 }, { "epoch": 0.34, "learning_rate": 1.526739564454306e-05, "loss": 2.1815, "step": 13945 }, { "epoch": 0.34, "learning_rate": 1.5266717024965375e-05, "loss": 2.2329, "step": 13946 }, { "epoch": 0.34, "learning_rate": 1.526603837182119e-05, "loss": 1.9652, "step": 13947 }, { "epoch": 0.34, "learning_rate": 1.5265359685114826e-05, "loss": 1.9329, "step": 13948 }, { "epoch": 0.34, "learning_rate": 1.526468096485061e-05, "loss": 2.0383, "step": 13949 }, { "epoch": 0.34, "learning_rate": 1.5264002211032865e-05, "loss": 2.1695, "step": 13950 }, { "epoch": 0.34, "learning_rate": 1.526332342366592e-05, "loss": 2.1659, "step": 13951 }, { "epoch": 0.34, "learning_rate": 1.52626446027541e-05, "loss": 2.1075, "step": 13952 }, { "epoch": 0.34, "learning_rate": 1.5261965748301725e-05, "loss": 2.0711, "step": 13953 }, { "epoch": 0.34, "learning_rate": 1.5261286860313134e-05, "loss": 2.0641, "step": 13954 }, { "epoch": 0.34, "learning_rate": 1.5260607938792643e-05, "loss": 2.062, "step": 13955 }, { "epoch": 0.34, "learning_rate": 1.5259928983744586e-05, "loss": 2.155, "step": 13956 }, { "epoch": 0.34, "learning_rate": 1.525924999517329e-05, "loss": 2.2699, "step": 13957 }, { "epoch": 0.34, "learning_rate": 1.5258570973083076e-05, "loss": 2.0692, "step": 13958 }, { "epoch": 0.34, "learning_rate": 1.5257891917478274e-05, "loss": 2.0807, "step": 13959 }, { "epoch": 0.34, "learning_rate": 1.5257212828363214e-05, "loss": 1.9959, "step": 13960 }, { "epoch": 0.34, "learning_rate": 1.5256533705742225e-05, "loss": 2.2651, "step": 13961 }, { "epoch": 0.34, "learning_rate": 1.5255854549619634e-05, "loss": 2.1702, "step": 13962 }, { "epoch": 0.34, "learning_rate": 1.525517535999977e-05, "loss": 2.1224, "step": 13963 }, { "epoch": 0.34, "learning_rate": 1.525449613688696e-05, "loss": 2.2141, "step": 13964 }, { "epoch": 0.34, "learning_rate": 1.5253816880285533e-05, "loss": 2.1206, "step": 13965 }, { "epoch": 0.34, "learning_rate": 1.5253137590199818e-05, "loss": 1.9153, "step": 13966 }, { "epoch": 0.34, "learning_rate": 1.5252458266634147e-05, "loss": 2.0523, "step": 13967 }, { "epoch": 0.34, "learning_rate": 1.5251778909592844e-05, "loss": 1.9893, "step": 13968 }, { "epoch": 0.34, "learning_rate": 1.5251099519080246e-05, "loss": 2.2068, "step": 13969 }, { "epoch": 0.34, "learning_rate": 1.525042009510068e-05, "loss": 2.0889, "step": 13970 }, { "epoch": 0.34, "learning_rate": 1.5249740637658473e-05, "loss": 2.0382, "step": 13971 }, { "epoch": 0.34, "learning_rate": 1.5249061146757961e-05, "loss": 1.9367, "step": 13972 }, { "epoch": 0.34, "learning_rate": 1.524838162240347e-05, "loss": 2.2534, "step": 13973 }, { "epoch": 0.34, "learning_rate": 1.5247702064599334e-05, "loss": 2.1366, "step": 13974 }, { "epoch": 0.34, "learning_rate": 1.524702247334988e-05, "loss": 2.1184, "step": 13975 }, { "epoch": 0.34, "learning_rate": 1.5246342848659444e-05, "loss": 2.0924, "step": 13976 }, { "epoch": 0.34, "learning_rate": 1.5245663190532351e-05, "loss": 2.1772, "step": 13977 }, { "epoch": 0.34, "learning_rate": 1.5244983498972942e-05, "loss": 1.9036, "step": 13978 }, { "epoch": 0.34, "learning_rate": 1.5244303773985545e-05, "loss": 1.9584, "step": 13979 }, { "epoch": 0.34, "learning_rate": 1.5243624015574488e-05, "loss": 1.9918, "step": 13980 }, { "epoch": 0.34, "learning_rate": 1.5242944223744105e-05, "loss": 2.0638, "step": 13981 }, { "epoch": 0.34, "learning_rate": 1.5242264398498732e-05, "loss": 2.079, "step": 13982 }, { "epoch": 0.34, "learning_rate": 1.52415845398427e-05, "loss": 1.9163, "step": 13983 }, { "epoch": 0.34, "learning_rate": 1.5240904647780341e-05, "loss": 2.0068, "step": 13984 }, { "epoch": 0.34, "learning_rate": 1.5240224722315988e-05, "loss": 2.2603, "step": 13985 }, { "epoch": 0.34, "learning_rate": 1.5239544763453973e-05, "loss": 2.2726, "step": 13986 }, { "epoch": 0.34, "learning_rate": 1.5238864771198634e-05, "loss": 2.0884, "step": 13987 }, { "epoch": 0.34, "learning_rate": 1.5238184745554306e-05, "loss": 2.0125, "step": 13988 }, { "epoch": 0.34, "learning_rate": 1.5237504686525315e-05, "loss": 2.1285, "step": 13989 }, { "epoch": 0.34, "learning_rate": 1.5236824594115999e-05, "loss": 2.1922, "step": 13990 }, { "epoch": 0.34, "learning_rate": 1.5236144468330697e-05, "loss": 2.136, "step": 13991 }, { "epoch": 0.34, "learning_rate": 1.5235464309173738e-05, "loss": 2.2889, "step": 13992 }, { "epoch": 0.34, "learning_rate": 1.5234784116649459e-05, "loss": 2.1642, "step": 13993 }, { "epoch": 0.34, "learning_rate": 1.5234103890762196e-05, "loss": 2.1423, "step": 13994 }, { "epoch": 0.34, "learning_rate": 1.523342363151628e-05, "loss": 2.2482, "step": 13995 }, { "epoch": 0.34, "learning_rate": 1.523274333891605e-05, "loss": 2.141, "step": 13996 }, { "epoch": 0.35, "learning_rate": 1.5232063012965843e-05, "loss": 2.0157, "step": 13997 }, { "epoch": 0.35, "learning_rate": 1.5231382653669995e-05, "loss": 2.0424, "step": 13998 }, { "epoch": 0.35, "learning_rate": 1.5230702261032839e-05, "loss": 1.9691, "step": 13999 }, { "epoch": 0.35, "learning_rate": 1.523002183505871e-05, "loss": 1.9797, "step": 14000 }, { "epoch": 0.35, "learning_rate": 1.522934137575195e-05, "loss": 2.1571, "step": 14001 }, { "epoch": 0.35, "learning_rate": 1.5228660883116897e-05, "loss": 2.1933, "step": 14002 }, { "epoch": 0.35, "learning_rate": 1.5227980357157877e-05, "loss": 2.0972, "step": 14003 }, { "epoch": 0.35, "learning_rate": 1.5227299797879239e-05, "loss": 2.0207, "step": 14004 }, { "epoch": 0.35, "learning_rate": 1.5226619205285314e-05, "loss": 2.1166, "step": 14005 }, { "epoch": 0.35, "learning_rate": 1.5225938579380444e-05, "loss": 2.1045, "step": 14006 }, { "epoch": 0.35, "learning_rate": 1.522525792016896e-05, "loss": 2.0611, "step": 14007 }, { "epoch": 0.35, "learning_rate": 1.5224577227655211e-05, "loss": 2.138, "step": 14008 }, { "epoch": 0.35, "learning_rate": 1.5223896501843522e-05, "loss": 2.1169, "step": 14009 }, { "epoch": 0.35, "learning_rate": 1.5223215742738239e-05, "loss": 2.0669, "step": 14010 }, { "epoch": 0.35, "learning_rate": 1.5222534950343706e-05, "loss": 2.1761, "step": 14011 }, { "epoch": 0.35, "learning_rate": 1.522185412466425e-05, "loss": 2.0439, "step": 14012 }, { "epoch": 0.35, "learning_rate": 1.5221173265704217e-05, "loss": 2.1589, "step": 14013 }, { "epoch": 0.35, "learning_rate": 1.5220492373467944e-05, "loss": 2.1901, "step": 14014 }, { "epoch": 0.35, "learning_rate": 1.5219811447959774e-05, "loss": 2.2403, "step": 14015 }, { "epoch": 0.35, "learning_rate": 1.5219130489184046e-05, "loss": 2.18, "step": 14016 }, { "epoch": 0.35, "learning_rate": 1.5218449497145096e-05, "loss": 2.1022, "step": 14017 }, { "epoch": 0.35, "learning_rate": 1.521776847184727e-05, "loss": 2.2261, "step": 14018 }, { "epoch": 0.35, "learning_rate": 1.52170874132949e-05, "loss": 2.0015, "step": 14019 }, { "epoch": 0.35, "learning_rate": 1.5216406321492335e-05, "loss": 1.9974, "step": 14020 }, { "epoch": 0.35, "learning_rate": 1.5215725196443909e-05, "loss": 2.0281, "step": 14021 }, { "epoch": 0.35, "learning_rate": 1.521504403815397e-05, "loss": 2.2531, "step": 14022 }, { "epoch": 0.35, "learning_rate": 1.5214362846626855e-05, "loss": 2.0439, "step": 14023 }, { "epoch": 0.35, "learning_rate": 1.5213681621866907e-05, "loss": 1.9984, "step": 14024 }, { "epoch": 0.35, "learning_rate": 1.5213000363878466e-05, "loss": 2.075, "step": 14025 }, { "epoch": 0.35, "learning_rate": 1.5212319072665876e-05, "loss": 2.1263, "step": 14026 }, { "epoch": 0.35, "learning_rate": 1.5211637748233477e-05, "loss": 1.9969, "step": 14027 }, { "epoch": 0.35, "learning_rate": 1.5210956390585612e-05, "loss": 2.2799, "step": 14028 }, { "epoch": 0.35, "learning_rate": 1.5210274999726622e-05, "loss": 2.2712, "step": 14029 }, { "epoch": 0.35, "learning_rate": 1.5209593575660852e-05, "loss": 2.0888, "step": 14030 }, { "epoch": 0.35, "learning_rate": 1.5208912118392647e-05, "loss": 2.0391, "step": 14031 }, { "epoch": 0.35, "learning_rate": 1.5208230627926349e-05, "loss": 2.0605, "step": 14032 }, { "epoch": 0.35, "learning_rate": 1.5207549104266295e-05, "loss": 2.1171, "step": 14033 }, { "epoch": 0.35, "learning_rate": 1.5206867547416834e-05, "loss": 2.2151, "step": 14034 }, { "epoch": 0.35, "learning_rate": 1.5206185957382312e-05, "loss": 2.1232, "step": 14035 }, { "epoch": 0.35, "learning_rate": 1.5205504334167072e-05, "loss": 2.1644, "step": 14036 }, { "epoch": 0.35, "learning_rate": 1.5204822677775453e-05, "loss": 2.0693, "step": 14037 }, { "epoch": 0.35, "learning_rate": 1.5204140988211801e-05, "loss": 2.2222, "step": 14038 }, { "epoch": 0.35, "learning_rate": 1.5203459265480465e-05, "loss": 2.0669, "step": 14039 }, { "epoch": 0.35, "learning_rate": 1.5202777509585787e-05, "loss": 2.0692, "step": 14040 }, { "epoch": 0.35, "learning_rate": 1.5202095720532113e-05, "loss": 2.22, "step": 14041 }, { "epoch": 0.35, "learning_rate": 1.5201413898323788e-05, "loss": 2.061, "step": 14042 }, { "epoch": 0.35, "learning_rate": 1.5200732042965154e-05, "loss": 2.1104, "step": 14043 }, { "epoch": 0.35, "learning_rate": 1.5200050154460565e-05, "loss": 2.0239, "step": 14044 }, { "epoch": 0.35, "learning_rate": 1.5199368232814359e-05, "loss": 2.0827, "step": 14045 }, { "epoch": 0.35, "learning_rate": 1.5198686278030885e-05, "loss": 2.0288, "step": 14046 }, { "epoch": 0.35, "learning_rate": 1.5198004290114487e-05, "loss": 2.1612, "step": 14047 }, { "epoch": 0.35, "learning_rate": 1.5197322269069514e-05, "loss": 2.1807, "step": 14048 }, { "epoch": 0.35, "learning_rate": 1.5196640214900316e-05, "loss": 2.1706, "step": 14049 }, { "epoch": 0.35, "learning_rate": 1.5195958127611236e-05, "loss": 2.0277, "step": 14050 }, { "epoch": 0.35, "learning_rate": 1.5195276007206618e-05, "loss": 2.0332, "step": 14051 }, { "epoch": 0.35, "learning_rate": 1.5194593853690815e-05, "loss": 2.2653, "step": 14052 }, { "epoch": 0.35, "learning_rate": 1.5193911667068172e-05, "loss": 2.1775, "step": 14053 }, { "epoch": 0.35, "learning_rate": 1.5193229447343041e-05, "loss": 2.1648, "step": 14054 }, { "epoch": 0.35, "learning_rate": 1.5192547194519762e-05, "loss": 2.2756, "step": 14055 }, { "epoch": 0.35, "learning_rate": 1.5191864908602687e-05, "loss": 2.1095, "step": 14056 }, { "epoch": 0.35, "learning_rate": 1.5191182589596167e-05, "loss": 2.2588, "step": 14057 }, { "epoch": 0.35, "learning_rate": 1.5190500237504547e-05, "loss": 2.0714, "step": 14058 }, { "epoch": 0.35, "learning_rate": 1.5189817852332178e-05, "loss": 1.9618, "step": 14059 }, { "epoch": 0.35, "learning_rate": 1.518913543408341e-05, "loss": 1.8407, "step": 14060 }, { "epoch": 0.35, "learning_rate": 1.5188452982762592e-05, "loss": 2.0508, "step": 14061 }, { "epoch": 0.35, "learning_rate": 1.5187770498374071e-05, "loss": 2.1328, "step": 14062 }, { "epoch": 0.35, "learning_rate": 1.5187087980922195e-05, "loss": 2.1213, "step": 14063 }, { "epoch": 0.35, "learning_rate": 1.518640543041132e-05, "loss": 2.0108, "step": 14064 }, { "epoch": 0.35, "learning_rate": 1.518572284684579e-05, "loss": 1.9956, "step": 14065 }, { "epoch": 0.35, "learning_rate": 1.5185040230229961e-05, "loss": 2.1826, "step": 14066 }, { "epoch": 0.35, "learning_rate": 1.5184357580568178e-05, "loss": 2.2015, "step": 14067 }, { "epoch": 0.35, "learning_rate": 1.5183674897864798e-05, "loss": 2.2356, "step": 14068 }, { "epoch": 0.35, "learning_rate": 1.5182992182124167e-05, "loss": 2.2027, "step": 14069 }, { "epoch": 0.35, "learning_rate": 1.518230943335064e-05, "loss": 2.0837, "step": 14070 }, { "epoch": 0.35, "learning_rate": 1.5181626651548565e-05, "loss": 2.1683, "step": 14071 }, { "epoch": 0.35, "learning_rate": 1.5180943836722293e-05, "loss": 2.0474, "step": 14072 }, { "epoch": 0.35, "learning_rate": 1.5180260988876176e-05, "loss": 2.1994, "step": 14073 }, { "epoch": 0.35, "learning_rate": 1.5179578108014569e-05, "loss": 2.1432, "step": 14074 }, { "epoch": 0.35, "learning_rate": 1.5178895194141824e-05, "loss": 2.0992, "step": 14075 }, { "epoch": 0.35, "learning_rate": 1.5178212247262295e-05, "loss": 2.0092, "step": 14076 }, { "epoch": 0.35, "learning_rate": 1.5177529267380327e-05, "loss": 2.2143, "step": 14077 }, { "epoch": 0.35, "learning_rate": 1.517684625450028e-05, "loss": 2.0715, "step": 14078 }, { "epoch": 0.35, "learning_rate": 1.5176163208626503e-05, "loss": 2.0595, "step": 14079 }, { "epoch": 0.35, "learning_rate": 1.5175480129763353e-05, "loss": 2.1304, "step": 14080 }, { "epoch": 0.35, "learning_rate": 1.5174797017915181e-05, "loss": 2.1326, "step": 14081 }, { "epoch": 0.35, "learning_rate": 1.517411387308634e-05, "loss": 2.1604, "step": 14082 }, { "epoch": 0.35, "learning_rate": 1.5173430695281184e-05, "loss": 1.9465, "step": 14083 }, { "epoch": 0.35, "learning_rate": 1.517274748450407e-05, "loss": 1.9868, "step": 14084 }, { "epoch": 0.35, "learning_rate": 1.5172064240759352e-05, "loss": 2.1078, "step": 14085 }, { "epoch": 0.35, "learning_rate": 1.5171380964051383e-05, "loss": 2.0766, "step": 14086 }, { "epoch": 0.35, "learning_rate": 1.5170697654384514e-05, "loss": 1.9772, "step": 14087 }, { "epoch": 0.35, "learning_rate": 1.5170014311763108e-05, "loss": 1.9857, "step": 14088 }, { "epoch": 0.35, "learning_rate": 1.5169330936191516e-05, "loss": 2.1691, "step": 14089 }, { "epoch": 0.35, "learning_rate": 1.516864752767409e-05, "loss": 2.0802, "step": 14090 }, { "epoch": 0.35, "learning_rate": 1.5167964086215189e-05, "loss": 2.0861, "step": 14091 }, { "epoch": 0.35, "learning_rate": 1.5167280611819169e-05, "loss": 2.1072, "step": 14092 }, { "epoch": 0.35, "learning_rate": 1.5166597104490386e-05, "loss": 2.2389, "step": 14093 }, { "epoch": 0.35, "learning_rate": 1.5165913564233199e-05, "loss": 1.9941, "step": 14094 }, { "epoch": 0.35, "learning_rate": 1.5165229991051958e-05, "loss": 2.0658, "step": 14095 }, { "epoch": 0.35, "learning_rate": 1.5164546384951022e-05, "loss": 2.0949, "step": 14096 }, { "epoch": 0.35, "learning_rate": 1.5163862745934752e-05, "loss": 2.1665, "step": 14097 }, { "epoch": 0.35, "learning_rate": 1.5163179074007501e-05, "loss": 2.1912, "step": 14098 }, { "epoch": 0.35, "learning_rate": 1.5162495369173629e-05, "loss": 2.0331, "step": 14099 }, { "epoch": 0.35, "learning_rate": 1.5161811631437486e-05, "loss": 2.2734, "step": 14100 }, { "epoch": 0.35, "learning_rate": 1.5161127860803438e-05, "loss": 2.2234, "step": 14101 }, { "epoch": 0.35, "learning_rate": 1.5160444057275841e-05, "loss": 2.0558, "step": 14102 }, { "epoch": 0.35, "learning_rate": 1.5159760220859054e-05, "loss": 2.1074, "step": 14103 }, { "epoch": 0.35, "learning_rate": 1.5159076351557431e-05, "loss": 2.1254, "step": 14104 }, { "epoch": 0.35, "learning_rate": 1.5158392449375334e-05, "loss": 2.0745, "step": 14105 }, { "epoch": 0.35, "learning_rate": 1.515770851431712e-05, "loss": 2.2118, "step": 14106 }, { "epoch": 0.35, "learning_rate": 1.5157024546387149e-05, "loss": 2.1155, "step": 14107 }, { "epoch": 0.35, "learning_rate": 1.5156340545589783e-05, "loss": 2.2235, "step": 14108 }, { "epoch": 0.35, "learning_rate": 1.5155656511929374e-05, "loss": 2.1519, "step": 14109 }, { "epoch": 0.35, "learning_rate": 1.5154972445410288e-05, "loss": 2.1945, "step": 14110 }, { "epoch": 0.35, "learning_rate": 1.5154288346036881e-05, "loss": 2.3318, "step": 14111 }, { "epoch": 0.35, "learning_rate": 1.5153604213813516e-05, "loss": 2.1702, "step": 14112 }, { "epoch": 0.35, "learning_rate": 1.5152920048744552e-05, "loss": 2.1963, "step": 14113 }, { "epoch": 0.35, "learning_rate": 1.515223585083435e-05, "loss": 2.1104, "step": 14114 }, { "epoch": 0.35, "learning_rate": 1.5151551620087267e-05, "loss": 2.02, "step": 14115 }, { "epoch": 0.35, "learning_rate": 1.5150867356507667e-05, "loss": 2.2546, "step": 14116 }, { "epoch": 0.35, "learning_rate": 1.5150183060099915e-05, "loss": 2.1422, "step": 14117 }, { "epoch": 0.35, "learning_rate": 1.5149498730868364e-05, "loss": 2.1423, "step": 14118 }, { "epoch": 0.35, "learning_rate": 1.514881436881738e-05, "loss": 2.1476, "step": 14119 }, { "epoch": 0.35, "learning_rate": 1.5148129973951325e-05, "loss": 2.2143, "step": 14120 }, { "epoch": 0.35, "learning_rate": 1.514744554627456e-05, "loss": 2.0828, "step": 14121 }, { "epoch": 0.35, "learning_rate": 1.5146761085791446e-05, "loss": 2.1014, "step": 14122 }, { "epoch": 0.35, "learning_rate": 1.5146076592506346e-05, "loss": 1.8932, "step": 14123 }, { "epoch": 0.35, "learning_rate": 1.5145392066423625e-05, "loss": 2.1811, "step": 14124 }, { "epoch": 0.35, "learning_rate": 1.5144707507547641e-05, "loss": 2.0561, "step": 14125 }, { "epoch": 0.35, "learning_rate": 1.5144022915882761e-05, "loss": 2.0952, "step": 14126 }, { "epoch": 0.35, "learning_rate": 1.5143338291433348e-05, "loss": 1.9181, "step": 14127 }, { "epoch": 0.35, "learning_rate": 1.514265363420376e-05, "loss": 2.1606, "step": 14128 }, { "epoch": 0.35, "learning_rate": 1.5141968944198366e-05, "loss": 2.0452, "step": 14129 }, { "epoch": 0.35, "learning_rate": 1.5141284221421529e-05, "loss": 2.3028, "step": 14130 }, { "epoch": 0.35, "learning_rate": 1.5140599465877612e-05, "loss": 2.0136, "step": 14131 }, { "epoch": 0.35, "learning_rate": 1.513991467757098e-05, "loss": 2.3167, "step": 14132 }, { "epoch": 0.35, "learning_rate": 1.5139229856505998e-05, "loss": 2.1407, "step": 14133 }, { "epoch": 0.35, "learning_rate": 1.5138545002687028e-05, "loss": 2.1319, "step": 14134 }, { "epoch": 0.35, "learning_rate": 1.5137860116118432e-05, "loss": 1.883, "step": 14135 }, { "epoch": 0.35, "learning_rate": 1.5137175196804581e-05, "loss": 2.1091, "step": 14136 }, { "epoch": 0.35, "learning_rate": 1.5136490244749842e-05, "loss": 2.06, "step": 14137 }, { "epoch": 0.35, "learning_rate": 1.5135805259958574e-05, "loss": 2.1874, "step": 14138 }, { "epoch": 0.35, "learning_rate": 1.5135120242435148e-05, "loss": 2.0697, "step": 14139 }, { "epoch": 0.35, "learning_rate": 1.5134435192183923e-05, "loss": 2.0712, "step": 14140 }, { "epoch": 0.35, "learning_rate": 1.5133750109209273e-05, "loss": 2.179, "step": 14141 }, { "epoch": 0.35, "learning_rate": 1.5133064993515558e-05, "loss": 2.0998, "step": 14142 }, { "epoch": 0.35, "learning_rate": 1.513237984510715e-05, "loss": 1.9736, "step": 14143 }, { "epoch": 0.35, "learning_rate": 1.5131694663988407e-05, "loss": 2.1989, "step": 14144 }, { "epoch": 0.35, "learning_rate": 1.5131009450163706e-05, "loss": 2.2148, "step": 14145 }, { "epoch": 0.35, "learning_rate": 1.5130324203637407e-05, "loss": 1.8966, "step": 14146 }, { "epoch": 0.35, "learning_rate": 1.5129638924413883e-05, "loss": 2.121, "step": 14147 }, { "epoch": 0.35, "learning_rate": 1.5128953612497497e-05, "loss": 2.0781, "step": 14148 }, { "epoch": 0.35, "learning_rate": 1.5128268267892621e-05, "loss": 2.0574, "step": 14149 }, { "epoch": 0.35, "learning_rate": 1.5127582890603614e-05, "loss": 2.1365, "step": 14150 }, { "epoch": 0.35, "learning_rate": 1.5126897480634854e-05, "loss": 2.1842, "step": 14151 }, { "epoch": 0.35, "learning_rate": 1.5126212037990705e-05, "loss": 2.1164, "step": 14152 }, { "epoch": 0.35, "learning_rate": 1.5125526562675536e-05, "loss": 2.1801, "step": 14153 }, { "epoch": 0.35, "learning_rate": 1.5124841054693714e-05, "loss": 2.1727, "step": 14154 }, { "epoch": 0.35, "learning_rate": 1.5124155514049612e-05, "loss": 1.9549, "step": 14155 }, { "epoch": 0.35, "learning_rate": 1.5123469940747599e-05, "loss": 2.0106, "step": 14156 }, { "epoch": 0.35, "learning_rate": 1.512278433479204e-05, "loss": 2.2053, "step": 14157 }, { "epoch": 0.35, "learning_rate": 1.5122098696187308e-05, "loss": 2.004, "step": 14158 }, { "epoch": 0.35, "learning_rate": 1.512141302493777e-05, "loss": 1.9564, "step": 14159 }, { "epoch": 0.35, "learning_rate": 1.51207273210478e-05, "loss": 2.2095, "step": 14160 }, { "epoch": 0.35, "learning_rate": 1.5120041584521765e-05, "loss": 1.9937, "step": 14161 }, { "epoch": 0.35, "learning_rate": 1.5119355815364034e-05, "loss": 2.1464, "step": 14162 }, { "epoch": 0.35, "learning_rate": 1.5118670013578983e-05, "loss": 2.1879, "step": 14163 }, { "epoch": 0.35, "learning_rate": 1.5117984179170976e-05, "loss": 2.0707, "step": 14164 }, { "epoch": 0.35, "learning_rate": 1.5117298312144394e-05, "loss": 2.0068, "step": 14165 }, { "epoch": 0.35, "learning_rate": 1.51166124125036e-05, "loss": 1.9729, "step": 14166 }, { "epoch": 0.35, "learning_rate": 1.5115926480252965e-05, "loss": 2.1099, "step": 14167 }, { "epoch": 0.35, "learning_rate": 1.5115240515396865e-05, "loss": 2.0811, "step": 14168 }, { "epoch": 0.35, "learning_rate": 1.511455451793967e-05, "loss": 2.1239, "step": 14169 }, { "epoch": 0.35, "learning_rate": 1.5113868487885755e-05, "loss": 2.0321, "step": 14170 }, { "epoch": 0.35, "learning_rate": 1.5113182425239483e-05, "loss": 2.213, "step": 14171 }, { "epoch": 0.35, "learning_rate": 1.511249633000524e-05, "loss": 2.2222, "step": 14172 }, { "epoch": 0.35, "learning_rate": 1.5111810202187389e-05, "loss": 2.2371, "step": 14173 }, { "epoch": 0.35, "learning_rate": 1.5111124041790302e-05, "loss": 2.1018, "step": 14174 }, { "epoch": 0.35, "learning_rate": 1.511043784881836e-05, "loss": 2.1602, "step": 14175 }, { "epoch": 0.35, "learning_rate": 1.5109751623275935e-05, "loss": 2.0892, "step": 14176 }, { "epoch": 0.35, "learning_rate": 1.5109065365167393e-05, "loss": 2.2837, "step": 14177 }, { "epoch": 0.35, "learning_rate": 1.5108379074497112e-05, "loss": 2.0412, "step": 14178 }, { "epoch": 0.35, "learning_rate": 1.510769275126947e-05, "loss": 2.167, "step": 14179 }, { "epoch": 0.35, "learning_rate": 1.5107006395488832e-05, "loss": 2.0954, "step": 14180 }, { "epoch": 0.35, "learning_rate": 1.5106320007159583e-05, "loss": 2.1133, "step": 14181 }, { "epoch": 0.35, "learning_rate": 1.510563358628609e-05, "loss": 2.1102, "step": 14182 }, { "epoch": 0.35, "learning_rate": 1.5104947132872732e-05, "loss": 2.1849, "step": 14183 }, { "epoch": 0.35, "learning_rate": 1.5104260646923883e-05, "loss": 2.055, "step": 14184 }, { "epoch": 0.35, "learning_rate": 1.5103574128443916e-05, "loss": 2.1506, "step": 14185 }, { "epoch": 0.35, "learning_rate": 1.5102887577437209e-05, "loss": 1.9468, "step": 14186 }, { "epoch": 0.35, "learning_rate": 1.5102200993908134e-05, "loss": 2.0362, "step": 14187 }, { "epoch": 0.35, "learning_rate": 1.510151437786107e-05, "loss": 2.0863, "step": 14188 }, { "epoch": 0.35, "learning_rate": 1.510082772930039e-05, "loss": 2.1805, "step": 14189 }, { "epoch": 0.35, "learning_rate": 1.5100141048230476e-05, "loss": 2.1127, "step": 14190 }, { "epoch": 0.35, "learning_rate": 1.5099454334655703e-05, "loss": 2.0898, "step": 14191 }, { "epoch": 0.35, "learning_rate": 1.5098767588580444e-05, "loss": 2.1523, "step": 14192 }, { "epoch": 0.35, "learning_rate": 1.5098080810009075e-05, "loss": 2.0217, "step": 14193 }, { "epoch": 0.35, "learning_rate": 1.5097393998945977e-05, "loss": 2.3161, "step": 14194 }, { "epoch": 0.35, "learning_rate": 1.509670715539553e-05, "loss": 2.1497, "step": 14195 }, { "epoch": 0.35, "learning_rate": 1.50960202793621e-05, "loss": 2.1358, "step": 14196 }, { "epoch": 0.35, "learning_rate": 1.5095333370850075e-05, "loss": 2.1368, "step": 14197 }, { "epoch": 0.35, "learning_rate": 1.509464642986383e-05, "loss": 2.1287, "step": 14198 }, { "epoch": 0.35, "learning_rate": 1.5093959456407743e-05, "loss": 2.3446, "step": 14199 }, { "epoch": 0.35, "learning_rate": 1.5093272450486196e-05, "loss": 2.0035, "step": 14200 }, { "epoch": 0.35, "learning_rate": 1.509258541210356e-05, "loss": 1.9837, "step": 14201 }, { "epoch": 0.35, "learning_rate": 1.509189834126422e-05, "loss": 2.151, "step": 14202 }, { "epoch": 0.35, "learning_rate": 1.5091211237972547e-05, "loss": 2.313, "step": 14203 }, { "epoch": 0.35, "learning_rate": 1.509052410223293e-05, "loss": 2.2493, "step": 14204 }, { "epoch": 0.35, "learning_rate": 1.5089836934049742e-05, "loss": 2.1779, "step": 14205 }, { "epoch": 0.35, "learning_rate": 1.5089149733427367e-05, "loss": 2.0655, "step": 14206 }, { "epoch": 0.35, "learning_rate": 1.5088462500370181e-05, "loss": 2.2443, "step": 14207 }, { "epoch": 0.35, "learning_rate": 1.5087775234882565e-05, "loss": 2.044, "step": 14208 }, { "epoch": 0.35, "learning_rate": 1.5087087936968901e-05, "loss": 2.0604, "step": 14209 }, { "epoch": 0.35, "learning_rate": 1.5086400606633564e-05, "loss": 2.1989, "step": 14210 }, { "epoch": 0.35, "learning_rate": 1.5085713243880942e-05, "loss": 2.1384, "step": 14211 }, { "epoch": 0.35, "learning_rate": 1.508502584871541e-05, "loss": 2.0935, "step": 14212 }, { "epoch": 0.35, "learning_rate": 1.5084338421141352e-05, "loss": 2.0887, "step": 14213 }, { "epoch": 0.35, "learning_rate": 1.5083650961163145e-05, "loss": 2.2442, "step": 14214 }, { "epoch": 0.35, "learning_rate": 1.5082963468785179e-05, "loss": 2.0176, "step": 14215 }, { "epoch": 0.35, "learning_rate": 1.5082275944011828e-05, "loss": 2.0858, "step": 14216 }, { "epoch": 0.35, "learning_rate": 1.5081588386847474e-05, "loss": 2.224, "step": 14217 }, { "epoch": 0.35, "learning_rate": 1.5080900797296506e-05, "loss": 2.1754, "step": 14218 }, { "epoch": 0.35, "learning_rate": 1.5080213175363298e-05, "loss": 2.2665, "step": 14219 }, { "epoch": 0.35, "learning_rate": 1.5079525521052238e-05, "loss": 2.0673, "step": 14220 }, { "epoch": 0.35, "learning_rate": 1.5078837834367706e-05, "loss": 2.1239, "step": 14221 }, { "epoch": 0.35, "learning_rate": 1.5078150115314084e-05, "loss": 2.1519, "step": 14222 }, { "epoch": 0.35, "learning_rate": 1.5077462363895756e-05, "loss": 2.0739, "step": 14223 }, { "epoch": 0.35, "learning_rate": 1.5076774580117106e-05, "loss": 2.0592, "step": 14224 }, { "epoch": 0.35, "learning_rate": 1.5076086763982517e-05, "loss": 1.9522, "step": 14225 }, { "epoch": 0.35, "learning_rate": 1.5075398915496373e-05, "loss": 2.102, "step": 14226 }, { "epoch": 0.35, "learning_rate": 1.5074711034663056e-05, "loss": 2.2162, "step": 14227 }, { "epoch": 0.35, "learning_rate": 1.5074023121486959e-05, "loss": 1.9555, "step": 14228 }, { "epoch": 0.35, "learning_rate": 1.5073335175972453e-05, "loss": 2.1832, "step": 14229 }, { "epoch": 0.35, "learning_rate": 1.5072647198123931e-05, "loss": 1.9514, "step": 14230 }, { "epoch": 0.35, "learning_rate": 1.5071959187945771e-05, "loss": 1.9923, "step": 14231 }, { "epoch": 0.35, "learning_rate": 1.5071271145442365e-05, "loss": 2.1697, "step": 14232 }, { "epoch": 0.35, "learning_rate": 1.5070583070618095e-05, "loss": 2.2008, "step": 14233 }, { "epoch": 0.35, "learning_rate": 1.5069894963477348e-05, "loss": 2.2776, "step": 14234 }, { "epoch": 0.35, "learning_rate": 1.5069206824024508e-05, "loss": 2.0911, "step": 14235 }, { "epoch": 0.35, "learning_rate": 1.506851865226396e-05, "loss": 2.0138, "step": 14236 }, { "epoch": 0.35, "learning_rate": 1.5067830448200086e-05, "loss": 2.3322, "step": 14237 }, { "epoch": 0.35, "learning_rate": 1.5067142211837285e-05, "loss": 2.1918, "step": 14238 }, { "epoch": 0.35, "learning_rate": 1.5066453943179933e-05, "loss": 2.153, "step": 14239 }, { "epoch": 0.35, "learning_rate": 1.5065765642232415e-05, "loss": 2.0906, "step": 14240 }, { "epoch": 0.35, "learning_rate": 1.5065077308999128e-05, "loss": 2.087, "step": 14241 }, { "epoch": 0.35, "learning_rate": 1.5064388943484446e-05, "loss": 1.9517, "step": 14242 }, { "epoch": 0.35, "learning_rate": 1.5063700545692768e-05, "loss": 2.0015, "step": 14243 }, { "epoch": 0.35, "learning_rate": 1.5063012115628475e-05, "loss": 2.0994, "step": 14244 }, { "epoch": 0.35, "learning_rate": 1.5062323653295956e-05, "loss": 2.0829, "step": 14245 }, { "epoch": 0.35, "learning_rate": 1.5061635158699597e-05, "loss": 2.0674, "step": 14246 }, { "epoch": 0.35, "learning_rate": 1.5060946631843787e-05, "loss": 2.0696, "step": 14247 }, { "epoch": 0.35, "learning_rate": 1.5060258072732919e-05, "loss": 2.0471, "step": 14248 }, { "epoch": 0.35, "learning_rate": 1.5059569481371373e-05, "loss": 2.0674, "step": 14249 }, { "epoch": 0.35, "learning_rate": 1.5058880857763543e-05, "loss": 2.0563, "step": 14250 }, { "epoch": 0.35, "learning_rate": 1.5058192201913816e-05, "loss": 2.1342, "step": 14251 }, { "epoch": 0.35, "learning_rate": 1.5057503513826584e-05, "loss": 2.2328, "step": 14252 }, { "epoch": 0.35, "learning_rate": 1.5056814793506233e-05, "loss": 2.1492, "step": 14253 }, { "epoch": 0.35, "learning_rate": 1.5056126040957152e-05, "loss": 2.2383, "step": 14254 }, { "epoch": 0.35, "learning_rate": 1.5055437256183734e-05, "loss": 2.1366, "step": 14255 }, { "epoch": 0.35, "learning_rate": 1.5054748439190366e-05, "loss": 2.2819, "step": 14256 }, { "epoch": 0.35, "learning_rate": 1.5054059589981438e-05, "loss": 2.026, "step": 14257 }, { "epoch": 0.35, "learning_rate": 1.505337070856134e-05, "loss": 2.0318, "step": 14258 }, { "epoch": 0.35, "learning_rate": 1.5052681794934465e-05, "loss": 2.0544, "step": 14259 }, { "epoch": 0.35, "learning_rate": 1.5051992849105204e-05, "loss": 2.075, "step": 14260 }, { "epoch": 0.35, "learning_rate": 1.5051303871077945e-05, "loss": 2.0882, "step": 14261 }, { "epoch": 0.35, "learning_rate": 1.505061486085708e-05, "loss": 2.0292, "step": 14262 }, { "epoch": 0.35, "learning_rate": 1.5049925818447001e-05, "loss": 2.0255, "step": 14263 }, { "epoch": 0.35, "learning_rate": 1.5049236743852098e-05, "loss": 2.0, "step": 14264 }, { "epoch": 0.35, "learning_rate": 1.5048547637076766e-05, "loss": 2.2284, "step": 14265 }, { "epoch": 0.35, "learning_rate": 1.5047858498125391e-05, "loss": 2.1347, "step": 14266 }, { "epoch": 0.35, "learning_rate": 1.504716932700237e-05, "loss": 2.1804, "step": 14267 }, { "epoch": 0.35, "learning_rate": 1.5046480123712095e-05, "loss": 2.0808, "step": 14268 }, { "epoch": 0.35, "learning_rate": 1.5045790888258959e-05, "loss": 2.1515, "step": 14269 }, { "epoch": 0.35, "learning_rate": 1.5045101620647353e-05, "loss": 2.2166, "step": 14270 }, { "epoch": 0.35, "learning_rate": 1.5044412320881667e-05, "loss": 2.0663, "step": 14271 }, { "epoch": 0.35, "learning_rate": 1.50437229889663e-05, "loss": 2.361, "step": 14272 }, { "epoch": 0.35, "learning_rate": 1.5043033624905643e-05, "loss": 2.0025, "step": 14273 }, { "epoch": 0.35, "learning_rate": 1.5042344228704088e-05, "loss": 2.0512, "step": 14274 }, { "epoch": 0.35, "learning_rate": 1.504165480036603e-05, "loss": 2.1058, "step": 14275 }, { "epoch": 0.35, "learning_rate": 1.5040965339895861e-05, "loss": 2.0905, "step": 14276 }, { "epoch": 0.35, "learning_rate": 1.5040275847297981e-05, "loss": 2.1795, "step": 14277 }, { "epoch": 0.35, "learning_rate": 1.5039586322576778e-05, "loss": 2.1159, "step": 14278 }, { "epoch": 0.35, "learning_rate": 1.5038896765736653e-05, "loss": 2.0687, "step": 14279 }, { "epoch": 0.35, "learning_rate": 1.5038207176781988e-05, "loss": 2.345, "step": 14280 }, { "epoch": 0.35, "learning_rate": 1.5037517555717195e-05, "loss": 1.901, "step": 14281 }, { "epoch": 0.35, "learning_rate": 1.5036827902546659e-05, "loss": 2.1573, "step": 14282 }, { "epoch": 0.35, "learning_rate": 1.5036138217274778e-05, "loss": 2.1644, "step": 14283 }, { "epoch": 0.35, "learning_rate": 1.5035448499905943e-05, "loss": 1.9524, "step": 14284 }, { "epoch": 0.35, "learning_rate": 1.5034758750444556e-05, "loss": 2.1828, "step": 14285 }, { "epoch": 0.35, "learning_rate": 1.503406896889501e-05, "loss": 2.1531, "step": 14286 }, { "epoch": 0.35, "learning_rate": 1.5033379155261703e-05, "loss": 2.0555, "step": 14287 }, { "epoch": 0.35, "learning_rate": 1.5032689309549032e-05, "loss": 2.1601, "step": 14288 }, { "epoch": 0.35, "learning_rate": 1.503199943176139e-05, "loss": 2.0395, "step": 14289 }, { "epoch": 0.35, "learning_rate": 1.5031309521903175e-05, "loss": 1.9087, "step": 14290 }, { "epoch": 0.35, "learning_rate": 1.5030619579978787e-05, "loss": 2.1421, "step": 14291 }, { "epoch": 0.35, "learning_rate": 1.502992960599262e-05, "loss": 2.18, "step": 14292 }, { "epoch": 0.35, "learning_rate": 1.5029239599949068e-05, "loss": 2.2955, "step": 14293 }, { "epoch": 0.35, "learning_rate": 1.5028549561852539e-05, "loss": 2.0908, "step": 14294 }, { "epoch": 0.35, "learning_rate": 1.5027859491707422e-05, "loss": 1.9868, "step": 14295 }, { "epoch": 0.35, "learning_rate": 1.5027169389518119e-05, "loss": 2.0579, "step": 14296 }, { "epoch": 0.35, "learning_rate": 1.5026479255289025e-05, "loss": 1.8764, "step": 14297 }, { "epoch": 0.35, "learning_rate": 1.5025789089024544e-05, "loss": 2.1392, "step": 14298 }, { "epoch": 0.35, "learning_rate": 1.5025098890729072e-05, "loss": 2.273, "step": 14299 }, { "epoch": 0.35, "learning_rate": 1.5024408660407001e-05, "loss": 2.0757, "step": 14300 }, { "epoch": 0.35, "learning_rate": 1.5023718398062743e-05, "loss": 2.1442, "step": 14301 }, { "epoch": 0.35, "learning_rate": 1.5023028103700687e-05, "loss": 2.2045, "step": 14302 }, { "epoch": 0.35, "learning_rate": 1.5022337777325237e-05, "loss": 2.1711, "step": 14303 }, { "epoch": 0.35, "learning_rate": 1.5021647418940789e-05, "loss": 2.1061, "step": 14304 }, { "epoch": 0.35, "learning_rate": 1.5020957028551747e-05, "loss": 1.9802, "step": 14305 }, { "epoch": 0.35, "learning_rate": 1.5020266606162513e-05, "loss": 2.1137, "step": 14306 }, { "epoch": 0.35, "learning_rate": 1.501957615177748e-05, "loss": 2.2506, "step": 14307 }, { "epoch": 0.35, "learning_rate": 1.5018885665401055e-05, "loss": 2.1344, "step": 14308 }, { "epoch": 0.35, "learning_rate": 1.5018195147037633e-05, "loss": 2.171, "step": 14309 }, { "epoch": 0.35, "learning_rate": 1.5017504596691618e-05, "loss": 2.045, "step": 14310 }, { "epoch": 0.35, "learning_rate": 1.5016814014367413e-05, "loss": 2.0855, "step": 14311 }, { "epoch": 0.35, "learning_rate": 1.5016123400069417e-05, "loss": 1.9374, "step": 14312 }, { "epoch": 0.35, "learning_rate": 1.5015432753802033e-05, "loss": 2.1081, "step": 14313 }, { "epoch": 0.35, "learning_rate": 1.5014742075569659e-05, "loss": 2.2371, "step": 14314 }, { "epoch": 0.35, "learning_rate": 1.5014051365376703e-05, "loss": 2.1677, "step": 14315 }, { "epoch": 0.35, "learning_rate": 1.5013360623227559e-05, "loss": 1.929, "step": 14316 }, { "epoch": 0.35, "learning_rate": 1.5012669849126638e-05, "loss": 1.9642, "step": 14317 }, { "epoch": 0.35, "learning_rate": 1.5011979043078338e-05, "loss": 2.156, "step": 14318 }, { "epoch": 0.35, "learning_rate": 1.501128820508706e-05, "loss": 2.103, "step": 14319 }, { "epoch": 0.35, "learning_rate": 1.5010597335157209e-05, "loss": 2.2812, "step": 14320 }, { "epoch": 0.35, "learning_rate": 1.5009906433293189e-05, "loss": 2.0811, "step": 14321 }, { "epoch": 0.35, "learning_rate": 1.5009215499499403e-05, "loss": 2.0339, "step": 14322 }, { "epoch": 0.35, "learning_rate": 1.5008524533780256e-05, "loss": 1.9493, "step": 14323 }, { "epoch": 0.35, "learning_rate": 1.5007833536140147e-05, "loss": 2.0362, "step": 14324 }, { "epoch": 0.35, "learning_rate": 1.5007142506583483e-05, "loss": 2.2813, "step": 14325 }, { "epoch": 0.35, "learning_rate": 1.5006451445114669e-05, "loss": 2.2593, "step": 14326 }, { "epoch": 0.35, "learning_rate": 1.500576035173811e-05, "loss": 2.1859, "step": 14327 }, { "epoch": 0.35, "learning_rate": 1.5005069226458207e-05, "loss": 2.0335, "step": 14328 }, { "epoch": 0.35, "learning_rate": 1.5004378069279363e-05, "loss": 2.0604, "step": 14329 }, { "epoch": 0.35, "learning_rate": 1.5003686880205992e-05, "loss": 1.9411, "step": 14330 }, { "epoch": 0.35, "learning_rate": 1.5002995659242493e-05, "loss": 2.0375, "step": 14331 }, { "epoch": 0.35, "learning_rate": 1.5002304406393275e-05, "loss": 1.9817, "step": 14332 }, { "epoch": 0.35, "learning_rate": 1.5001613121662736e-05, "loss": 2.0547, "step": 14333 }, { "epoch": 0.35, "learning_rate": 1.5000921805055288e-05, "loss": 2.2803, "step": 14334 }, { "epoch": 0.35, "learning_rate": 1.5000230456575336e-05, "loss": 2.0087, "step": 14335 }, { "epoch": 0.35, "learning_rate": 1.4999539076227288e-05, "loss": 2.0909, "step": 14336 }, { "epoch": 0.35, "learning_rate": 1.4998847664015543e-05, "loss": 2.2887, "step": 14337 }, { "epoch": 0.35, "learning_rate": 1.4998156219944515e-05, "loss": 2.0895, "step": 14338 }, { "epoch": 0.35, "learning_rate": 1.4997464744018608e-05, "loss": 2.2084, "step": 14339 }, { "epoch": 0.35, "learning_rate": 1.4996773236242232e-05, "loss": 2.1922, "step": 14340 }, { "epoch": 0.35, "learning_rate": 1.4996081696619791e-05, "loss": 2.0853, "step": 14341 }, { "epoch": 0.35, "learning_rate": 1.4995390125155694e-05, "loss": 2.2323, "step": 14342 }, { "epoch": 0.35, "learning_rate": 1.4994698521854344e-05, "loss": 1.9168, "step": 14343 }, { "epoch": 0.35, "learning_rate": 1.4994006886720156e-05, "loss": 2.2404, "step": 14344 }, { "epoch": 0.35, "learning_rate": 1.4993315219757533e-05, "loss": 2.0105, "step": 14345 }, { "epoch": 0.35, "learning_rate": 1.4992623520970886e-05, "loss": 2.0385, "step": 14346 }, { "epoch": 0.35, "learning_rate": 1.4991931790364623e-05, "loss": 2.1415, "step": 14347 }, { "epoch": 0.35, "learning_rate": 1.4991240027943148e-05, "loss": 1.9517, "step": 14348 }, { "epoch": 0.35, "learning_rate": 1.4990548233710876e-05, "loss": 2.2475, "step": 14349 }, { "epoch": 0.35, "learning_rate": 1.4989856407672214e-05, "loss": 2.1219, "step": 14350 }, { "epoch": 0.35, "learning_rate": 1.4989164549831572e-05, "loss": 2.0874, "step": 14351 }, { "epoch": 0.35, "learning_rate": 1.4988472660193358e-05, "loss": 2.0137, "step": 14352 }, { "epoch": 0.35, "learning_rate": 1.4987780738761982e-05, "loss": 2.0176, "step": 14353 }, { "epoch": 0.35, "learning_rate": 1.4987088785541852e-05, "loss": 1.8922, "step": 14354 }, { "epoch": 0.35, "learning_rate": 1.498639680053738e-05, "loss": 2.1485, "step": 14355 }, { "epoch": 0.35, "learning_rate": 1.4985704783752976e-05, "loss": 2.2684, "step": 14356 }, { "epoch": 0.35, "learning_rate": 1.4985012735193054e-05, "loss": 2.2201, "step": 14357 }, { "epoch": 0.35, "learning_rate": 1.4984320654862018e-05, "loss": 2.0416, "step": 14358 }, { "epoch": 0.35, "learning_rate": 1.4983628542764286e-05, "loss": 2.1243, "step": 14359 }, { "epoch": 0.35, "learning_rate": 1.4982936398904261e-05, "loss": 2.1694, "step": 14360 }, { "epoch": 0.35, "learning_rate": 1.4982244223286363e-05, "loss": 2.0915, "step": 14361 }, { "epoch": 0.35, "learning_rate": 1.4981552015914993e-05, "loss": 2.0065, "step": 14362 }, { "epoch": 0.35, "learning_rate": 1.4980859776794572e-05, "loss": 2.1335, "step": 14363 }, { "epoch": 0.35, "learning_rate": 1.4980167505929506e-05, "loss": 2.2335, "step": 14364 }, { "epoch": 0.35, "learning_rate": 1.4979475203324212e-05, "loss": 2.2493, "step": 14365 }, { "epoch": 0.35, "learning_rate": 1.49787828689831e-05, "loss": 1.9814, "step": 14366 }, { "epoch": 0.35, "learning_rate": 1.497809050291058e-05, "loss": 2.0832, "step": 14367 }, { "epoch": 0.35, "learning_rate": 1.4977398105111069e-05, "loss": 2.1504, "step": 14368 }, { "epoch": 0.35, "learning_rate": 1.4976705675588976e-05, "loss": 2.0887, "step": 14369 }, { "epoch": 0.35, "learning_rate": 1.4976013214348719e-05, "loss": 2.0373, "step": 14370 }, { "epoch": 0.35, "learning_rate": 1.4975320721394707e-05, "loss": 2.1448, "step": 14371 }, { "epoch": 0.35, "learning_rate": 1.4974628196731353e-05, "loss": 2.117, "step": 14372 }, { "epoch": 0.35, "learning_rate": 1.4973935640363071e-05, "loss": 2.0936, "step": 14373 }, { "epoch": 0.35, "learning_rate": 1.497324305229428e-05, "loss": 2.1718, "step": 14374 }, { "epoch": 0.35, "learning_rate": 1.4972550432529386e-05, "loss": 1.9318, "step": 14375 }, { "epoch": 0.35, "learning_rate": 1.4971857781072812e-05, "loss": 2.1159, "step": 14376 }, { "epoch": 0.35, "learning_rate": 1.4971165097928966e-05, "loss": 2.0708, "step": 14377 }, { "epoch": 0.35, "learning_rate": 1.4970472383102265e-05, "loss": 2.1295, "step": 14378 }, { "epoch": 0.35, "learning_rate": 1.4969779636597123e-05, "loss": 2.2624, "step": 14379 }, { "epoch": 0.35, "learning_rate": 1.4969086858417958e-05, "loss": 2.0869, "step": 14380 }, { "epoch": 0.35, "learning_rate": 1.496839404856918e-05, "loss": 1.993, "step": 14381 }, { "epoch": 0.35, "learning_rate": 1.4967701207055207e-05, "loss": 2.2994, "step": 14382 }, { "epoch": 0.35, "learning_rate": 1.496700833388046e-05, "loss": 2.3596, "step": 14383 }, { "epoch": 0.35, "learning_rate": 1.4966315429049347e-05, "loss": 2.2044, "step": 14384 }, { "epoch": 0.35, "learning_rate": 1.4965622492566287e-05, "loss": 2.0691, "step": 14385 }, { "epoch": 0.35, "learning_rate": 1.4964929524435698e-05, "loss": 2.0042, "step": 14386 }, { "epoch": 0.35, "learning_rate": 1.4964236524661994e-05, "loss": 2.1069, "step": 14387 }, { "epoch": 0.35, "learning_rate": 1.4963543493249595e-05, "loss": 2.057, "step": 14388 }, { "epoch": 0.35, "learning_rate": 1.4962850430202913e-05, "loss": 2.1489, "step": 14389 }, { "epoch": 0.35, "learning_rate": 1.4962157335526369e-05, "loss": 2.2529, "step": 14390 }, { "epoch": 0.35, "learning_rate": 1.4961464209224379e-05, "loss": 2.2299, "step": 14391 }, { "epoch": 0.35, "learning_rate": 1.496077105130136e-05, "loss": 2.2441, "step": 14392 }, { "epoch": 0.35, "learning_rate": 1.4960077861761732e-05, "loss": 2.0066, "step": 14393 }, { "epoch": 0.35, "learning_rate": 1.495938464060991e-05, "loss": 2.2218, "step": 14394 }, { "epoch": 0.35, "learning_rate": 1.4958691387850315e-05, "loss": 2.0162, "step": 14395 }, { "epoch": 0.35, "learning_rate": 1.4957998103487361e-05, "loss": 1.9949, "step": 14396 }, { "epoch": 0.35, "learning_rate": 1.4957304787525473e-05, "loss": 2.0382, "step": 14397 }, { "epoch": 0.35, "learning_rate": 1.4956611439969064e-05, "loss": 2.2024, "step": 14398 }, { "epoch": 0.35, "learning_rate": 1.4955918060822554e-05, "loss": 2.0011, "step": 14399 }, { "epoch": 0.35, "learning_rate": 1.4955224650090365e-05, "loss": 2.15, "step": 14400 }, { "epoch": 0.35, "learning_rate": 1.4954531207776915e-05, "loss": 2.2366, "step": 14401 }, { "epoch": 0.35, "learning_rate": 1.495383773388662e-05, "loss": 2.255, "step": 14402 }, { "epoch": 0.36, "learning_rate": 1.4953144228423906e-05, "loss": 2.1165, "step": 14403 }, { "epoch": 0.36, "learning_rate": 1.4952450691393187e-05, "loss": 2.0863, "step": 14404 }, { "epoch": 0.36, "learning_rate": 1.4951757122798888e-05, "loss": 2.0326, "step": 14405 }, { "epoch": 0.36, "learning_rate": 1.4951063522645426e-05, "loss": 2.0763, "step": 14406 }, { "epoch": 0.36, "learning_rate": 1.4950369890937225e-05, "loss": 2.059, "step": 14407 }, { "epoch": 0.36, "learning_rate": 1.4949676227678698e-05, "loss": 2.1189, "step": 14408 }, { "epoch": 0.36, "learning_rate": 1.4948982532874274e-05, "loss": 2.0576, "step": 14409 }, { "epoch": 0.36, "learning_rate": 1.4948288806528373e-05, "loss": 2.0789, "step": 14410 }, { "epoch": 0.36, "learning_rate": 1.4947595048645414e-05, "loss": 2.1058, "step": 14411 }, { "epoch": 0.36, "learning_rate": 1.4946901259229817e-05, "loss": 2.1535, "step": 14412 }, { "epoch": 0.36, "learning_rate": 1.4946207438286011e-05, "loss": 2.0797, "step": 14413 }, { "epoch": 0.36, "learning_rate": 1.494551358581841e-05, "loss": 2.1137, "step": 14414 }, { "epoch": 0.36, "learning_rate": 1.4944819701831439e-05, "loss": 2.0832, "step": 14415 }, { "epoch": 0.36, "learning_rate": 1.4944125786329521e-05, "loss": 2.2354, "step": 14416 }, { "epoch": 0.36, "learning_rate": 1.4943431839317078e-05, "loss": 2.098, "step": 14417 }, { "epoch": 0.36, "learning_rate": 1.4942737860798533e-05, "loss": 2.0008, "step": 14418 }, { "epoch": 0.36, "learning_rate": 1.494204385077831e-05, "loss": 2.1627, "step": 14419 }, { "epoch": 0.36, "learning_rate": 1.4941349809260827e-05, "loss": 2.0323, "step": 14420 }, { "epoch": 0.36, "learning_rate": 1.4940655736250515e-05, "loss": 2.0262, "step": 14421 }, { "epoch": 0.36, "learning_rate": 1.4939961631751793e-05, "loss": 2.1124, "step": 14422 }, { "epoch": 0.36, "learning_rate": 1.4939267495769088e-05, "loss": 2.1811, "step": 14423 }, { "epoch": 0.36, "learning_rate": 1.493857332830682e-05, "loss": 2.2138, "step": 14424 }, { "epoch": 0.36, "learning_rate": 1.4937879129369413e-05, "loss": 2.1798, "step": 14425 }, { "epoch": 0.36, "learning_rate": 1.4937184898961293e-05, "loss": 2.1983, "step": 14426 }, { "epoch": 0.36, "learning_rate": 1.4936490637086886e-05, "loss": 2.0795, "step": 14427 }, { "epoch": 0.36, "learning_rate": 1.4935796343750615e-05, "loss": 2.1223, "step": 14428 }, { "epoch": 0.36, "learning_rate": 1.4935102018956907e-05, "loss": 2.062, "step": 14429 }, { "epoch": 0.36, "learning_rate": 1.4934407662710184e-05, "loss": 2.3069, "step": 14430 }, { "epoch": 0.36, "learning_rate": 1.4933713275014872e-05, "loss": 2.0809, "step": 14431 }, { "epoch": 0.36, "learning_rate": 1.49330188558754e-05, "loss": 2.0826, "step": 14432 }, { "epoch": 0.36, "learning_rate": 1.493232440529619e-05, "loss": 2.2434, "step": 14433 }, { "epoch": 0.36, "learning_rate": 1.4931629923281668e-05, "loss": 2.2297, "step": 14434 }, { "epoch": 0.36, "learning_rate": 1.493093540983626e-05, "loss": 1.9707, "step": 14435 }, { "epoch": 0.36, "learning_rate": 1.4930240864964398e-05, "loss": 2.1439, "step": 14436 }, { "epoch": 0.36, "learning_rate": 1.4929546288670503e-05, "loss": 2.3259, "step": 14437 }, { "epoch": 0.36, "learning_rate": 1.4928851680959002e-05, "loss": 2.1295, "step": 14438 }, { "epoch": 0.36, "learning_rate": 1.4928157041834323e-05, "loss": 2.0789, "step": 14439 }, { "epoch": 0.36, "learning_rate": 1.4927462371300894e-05, "loss": 2.1689, "step": 14440 }, { "epoch": 0.36, "learning_rate": 1.492676766936314e-05, "loss": 2.0431, "step": 14441 }, { "epoch": 0.36, "learning_rate": 1.492607293602549e-05, "loss": 2.2028, "step": 14442 }, { "epoch": 0.36, "learning_rate": 1.4925378171292372e-05, "loss": 2.0584, "step": 14443 }, { "epoch": 0.36, "learning_rate": 1.4924683375168216e-05, "loss": 2.0352, "step": 14444 }, { "epoch": 0.36, "learning_rate": 1.4923988547657445e-05, "loss": 2.2229, "step": 14445 }, { "epoch": 0.36, "learning_rate": 1.4923293688764493e-05, "loss": 2.0893, "step": 14446 }, { "epoch": 0.36, "learning_rate": 1.4922598798493783e-05, "loss": 2.1364, "step": 14447 }, { "epoch": 0.36, "learning_rate": 1.492190387684975e-05, "loss": 2.078, "step": 14448 }, { "epoch": 0.36, "learning_rate": 1.4921208923836817e-05, "loss": 2.0829, "step": 14449 }, { "epoch": 0.36, "learning_rate": 1.4920513939459415e-05, "loss": 1.9948, "step": 14450 }, { "epoch": 0.36, "learning_rate": 1.4919818923721979e-05, "loss": 2.0654, "step": 14451 }, { "epoch": 0.36, "learning_rate": 1.4919123876628928e-05, "loss": 2.0308, "step": 14452 }, { "epoch": 0.36, "learning_rate": 1.4918428798184699e-05, "loss": 2.0572, "step": 14453 }, { "epoch": 0.36, "learning_rate": 1.491773368839372e-05, "loss": 2.2433, "step": 14454 }, { "epoch": 0.36, "learning_rate": 1.4917038547260423e-05, "loss": 2.1418, "step": 14455 }, { "epoch": 0.36, "learning_rate": 1.4916343374789237e-05, "loss": 2.2166, "step": 14456 }, { "epoch": 0.36, "learning_rate": 1.4915648170984593e-05, "loss": 2.1118, "step": 14457 }, { "epoch": 0.36, "learning_rate": 1.4914952935850922e-05, "loss": 2.1415, "step": 14458 }, { "epoch": 0.36, "learning_rate": 1.4914257669392652e-05, "loss": 2.2563, "step": 14459 }, { "epoch": 0.36, "learning_rate": 1.4913562371614216e-05, "loss": 2.0146, "step": 14460 }, { "epoch": 0.36, "learning_rate": 1.4912867042520045e-05, "loss": 2.0344, "step": 14461 }, { "epoch": 0.36, "learning_rate": 1.4912171682114574e-05, "loss": 2.0776, "step": 14462 }, { "epoch": 0.36, "learning_rate": 1.4911476290402231e-05, "loss": 2.1732, "step": 14463 }, { "epoch": 0.36, "learning_rate": 1.4910780867387449e-05, "loss": 2.1077, "step": 14464 }, { "epoch": 0.36, "learning_rate": 1.491008541307466e-05, "loss": 2.1836, "step": 14465 }, { "epoch": 0.36, "learning_rate": 1.4909389927468297e-05, "loss": 2.0866, "step": 14466 }, { "epoch": 0.36, "learning_rate": 1.4908694410572792e-05, "loss": 2.1896, "step": 14467 }, { "epoch": 0.36, "learning_rate": 1.4907998862392577e-05, "loss": 1.9954, "step": 14468 }, { "epoch": 0.36, "learning_rate": 1.4907303282932086e-05, "loss": 2.1558, "step": 14469 }, { "epoch": 0.36, "learning_rate": 1.4906607672195751e-05, "loss": 2.0672, "step": 14470 }, { "epoch": 0.36, "learning_rate": 1.4905912030188007e-05, "loss": 2.0608, "step": 14471 }, { "epoch": 0.36, "learning_rate": 1.490521635691329e-05, "loss": 2.0358, "step": 14472 }, { "epoch": 0.36, "learning_rate": 1.4904520652376029e-05, "loss": 2.0099, "step": 14473 }, { "epoch": 0.36, "learning_rate": 1.4903824916580656e-05, "loss": 2.1717, "step": 14474 }, { "epoch": 0.36, "learning_rate": 1.4903129149531613e-05, "loss": 1.8418, "step": 14475 }, { "epoch": 0.36, "learning_rate": 1.4902433351233328e-05, "loss": 2.209, "step": 14476 }, { "epoch": 0.36, "learning_rate": 1.4901737521690236e-05, "loss": 2.0033, "step": 14477 }, { "epoch": 0.36, "learning_rate": 1.4901041660906777e-05, "loss": 2.2718, "step": 14478 }, { "epoch": 0.36, "learning_rate": 1.4900345768887378e-05, "loss": 2.061, "step": 14479 }, { "epoch": 0.36, "learning_rate": 1.489964984563648e-05, "loss": 2.1431, "step": 14480 }, { "epoch": 0.36, "learning_rate": 1.489895389115852e-05, "loss": 2.0689, "step": 14481 }, { "epoch": 0.36, "learning_rate": 1.4898257905457929e-05, "loss": 2.0754, "step": 14482 }, { "epoch": 0.36, "learning_rate": 1.489756188853914e-05, "loss": 2.2665, "step": 14483 }, { "epoch": 0.36, "learning_rate": 1.4896865840406597e-05, "loss": 2.0537, "step": 14484 }, { "epoch": 0.36, "learning_rate": 1.4896169761064734e-05, "loss": 2.2706, "step": 14485 }, { "epoch": 0.36, "learning_rate": 1.4895473650517979e-05, "loss": 2.2787, "step": 14486 }, { "epoch": 0.36, "learning_rate": 1.4894777508770779e-05, "loss": 1.958, "step": 14487 }, { "epoch": 0.36, "learning_rate": 1.4894081335827566e-05, "loss": 2.0423, "step": 14488 }, { "epoch": 0.36, "learning_rate": 1.4893385131692779e-05, "loss": 2.2967, "step": 14489 }, { "epoch": 0.36, "learning_rate": 1.4892688896370856e-05, "loss": 2.0116, "step": 14490 }, { "epoch": 0.36, "learning_rate": 1.4891992629866228e-05, "loss": 1.9805, "step": 14491 }, { "epoch": 0.36, "learning_rate": 1.4891296332183338e-05, "loss": 2.1583, "step": 14492 }, { "epoch": 0.36, "learning_rate": 1.4890600003326623e-05, "loss": 2.0069, "step": 14493 }, { "epoch": 0.36, "learning_rate": 1.4889903643300518e-05, "loss": 1.9831, "step": 14494 }, { "epoch": 0.36, "learning_rate": 1.4889207252109466e-05, "loss": 2.072, "step": 14495 }, { "epoch": 0.36, "learning_rate": 1.4888510829757902e-05, "loss": 2.0519, "step": 14496 }, { "epoch": 0.36, "learning_rate": 1.4887814376250267e-05, "loss": 2.1046, "step": 14497 }, { "epoch": 0.36, "learning_rate": 1.4887117891590993e-05, "loss": 2.2281, "step": 14498 }, { "epoch": 0.36, "learning_rate": 1.488642137578453e-05, "loss": 2.1348, "step": 14499 }, { "epoch": 0.36, "learning_rate": 1.4885724828835308e-05, "loss": 2.1722, "step": 14500 }, { "epoch": 0.36, "learning_rate": 1.488502825074777e-05, "loss": 1.9703, "step": 14501 }, { "epoch": 0.36, "learning_rate": 1.4884331641526358e-05, "loss": 2.1614, "step": 14502 }, { "epoch": 0.36, "learning_rate": 1.4883635001175506e-05, "loss": 2.1964, "step": 14503 }, { "epoch": 0.36, "learning_rate": 1.4882938329699652e-05, "loss": 2.1091, "step": 14504 }, { "epoch": 0.36, "learning_rate": 1.4882241627103246e-05, "loss": 2.3159, "step": 14505 }, { "epoch": 0.36, "learning_rate": 1.4881544893390725e-05, "loss": 2.029, "step": 14506 }, { "epoch": 0.36, "learning_rate": 1.4880848128566524e-05, "loss": 2.1386, "step": 14507 }, { "epoch": 0.36, "learning_rate": 1.4880151332635086e-05, "loss": 2.1148, "step": 14508 }, { "epoch": 0.36, "learning_rate": 1.4879454505600857e-05, "loss": 2.3085, "step": 14509 }, { "epoch": 0.36, "learning_rate": 1.4878757647468274e-05, "loss": 2.0647, "step": 14510 }, { "epoch": 0.36, "learning_rate": 1.4878060758241776e-05, "loss": 1.9969, "step": 14511 }, { "epoch": 0.36, "learning_rate": 1.4877363837925806e-05, "loss": 2.24, "step": 14512 }, { "epoch": 0.36, "learning_rate": 1.487666688652481e-05, "loss": 2.2035, "step": 14513 }, { "epoch": 0.36, "learning_rate": 1.4875969904043226e-05, "loss": 2.2189, "step": 14514 }, { "epoch": 0.36, "learning_rate": 1.4875272890485497e-05, "loss": 2.1766, "step": 14515 }, { "epoch": 0.36, "learning_rate": 1.4874575845856063e-05, "loss": 2.0745, "step": 14516 }, { "epoch": 0.36, "learning_rate": 1.4873878770159369e-05, "loss": 2.2348, "step": 14517 }, { "epoch": 0.36, "learning_rate": 1.487318166339986e-05, "loss": 2.1381, "step": 14518 }, { "epoch": 0.36, "learning_rate": 1.4872484525581973e-05, "loss": 2.1453, "step": 14519 }, { "epoch": 0.36, "learning_rate": 1.4871787356710158e-05, "loss": 2.0395, "step": 14520 }, { "epoch": 0.36, "learning_rate": 1.487109015678885e-05, "loss": 2.1408, "step": 14521 }, { "epoch": 0.36, "learning_rate": 1.4870392925822499e-05, "loss": 2.0402, "step": 14522 }, { "epoch": 0.36, "learning_rate": 1.4869695663815547e-05, "loss": 2.1021, "step": 14523 }, { "epoch": 0.36, "learning_rate": 1.4868998370772439e-05, "loss": 2.1942, "step": 14524 }, { "epoch": 0.36, "learning_rate": 1.4868301046697615e-05, "loss": 2.0083, "step": 14525 }, { "epoch": 0.36, "learning_rate": 1.4867603691595523e-05, "loss": 2.1259, "step": 14526 }, { "epoch": 0.36, "learning_rate": 1.4866906305470606e-05, "loss": 2.1865, "step": 14527 }, { "epoch": 0.36, "learning_rate": 1.486620888832731e-05, "loss": 2.2327, "step": 14528 }, { "epoch": 0.36, "learning_rate": 1.4865511440170077e-05, "loss": 2.1537, "step": 14529 }, { "epoch": 0.36, "learning_rate": 1.4864813961003356e-05, "loss": 2.0521, "step": 14530 }, { "epoch": 0.36, "learning_rate": 1.486411645083159e-05, "loss": 2.0925, "step": 14531 }, { "epoch": 0.36, "learning_rate": 1.486341890965922e-05, "loss": 2.1861, "step": 14532 }, { "epoch": 0.36, "learning_rate": 1.4862721337490701e-05, "loss": 2.028, "step": 14533 }, { "epoch": 0.36, "learning_rate": 1.4862023734330476e-05, "loss": 2.0482, "step": 14534 }, { "epoch": 0.36, "learning_rate": 1.4861326100182987e-05, "loss": 2.149, "step": 14535 }, { "epoch": 0.36, "learning_rate": 1.4860628435052682e-05, "loss": 2.1341, "step": 14536 }, { "epoch": 0.36, "learning_rate": 1.4859930738944006e-05, "loss": 2.1906, "step": 14537 }, { "epoch": 0.36, "learning_rate": 1.4859233011861411e-05, "loss": 2.0949, "step": 14538 }, { "epoch": 0.36, "learning_rate": 1.4858535253809339e-05, "loss": 2.0586, "step": 14539 }, { "epoch": 0.36, "learning_rate": 1.485783746479224e-05, "loss": 2.1282, "step": 14540 }, { "epoch": 0.36, "learning_rate": 1.4857139644814555e-05, "loss": 2.0694, "step": 14541 }, { "epoch": 0.36, "learning_rate": 1.485644179388074e-05, "loss": 2.0938, "step": 14542 }, { "epoch": 0.36, "learning_rate": 1.485574391199524e-05, "loss": 2.0919, "step": 14543 }, { "epoch": 0.36, "learning_rate": 1.4855045999162501e-05, "loss": 1.9733, "step": 14544 }, { "epoch": 0.36, "learning_rate": 1.4854348055386972e-05, "loss": 2.0402, "step": 14545 }, { "epoch": 0.36, "learning_rate": 1.4853650080673096e-05, "loss": 2.1814, "step": 14546 }, { "epoch": 0.36, "learning_rate": 1.485295207502533e-05, "loss": 2.1309, "step": 14547 }, { "epoch": 0.36, "learning_rate": 1.4852254038448117e-05, "loss": 2.0052, "step": 14548 }, { "epoch": 0.36, "learning_rate": 1.4851555970945911e-05, "loss": 2.0853, "step": 14549 }, { "epoch": 0.36, "learning_rate": 1.4850857872523156e-05, "loss": 2.0586, "step": 14550 }, { "epoch": 0.36, "learning_rate": 1.4850159743184302e-05, "loss": 2.2136, "step": 14551 }, { "epoch": 0.36, "learning_rate": 1.4849461582933803e-05, "loss": 2.1816, "step": 14552 }, { "epoch": 0.36, "learning_rate": 1.4848763391776102e-05, "loss": 2.0269, "step": 14553 }, { "epoch": 0.36, "learning_rate": 1.4848065169715653e-05, "loss": 2.1152, "step": 14554 }, { "epoch": 0.36, "learning_rate": 1.4847366916756905e-05, "loss": 2.0399, "step": 14555 }, { "epoch": 0.36, "learning_rate": 1.4846668632904304e-05, "loss": 2.1687, "step": 14556 }, { "epoch": 0.36, "learning_rate": 1.4845970318162307e-05, "loss": 2.0103, "step": 14557 }, { "epoch": 0.36, "learning_rate": 1.4845271972535362e-05, "loss": 2.1213, "step": 14558 }, { "epoch": 0.36, "learning_rate": 1.484457359602792e-05, "loss": 1.9571, "step": 14559 }, { "epoch": 0.36, "learning_rate": 1.4843875188644429e-05, "loss": 1.9449, "step": 14560 }, { "epoch": 0.36, "learning_rate": 1.4843176750389346e-05, "loss": 2.2344, "step": 14561 }, { "epoch": 0.36, "learning_rate": 1.4842478281267118e-05, "loss": 2.0887, "step": 14562 }, { "epoch": 0.36, "learning_rate": 1.48417797812822e-05, "loss": 1.8763, "step": 14563 }, { "epoch": 0.36, "learning_rate": 1.484108125043904e-05, "loss": 2.0275, "step": 14564 }, { "epoch": 0.36, "learning_rate": 1.484038268874209e-05, "loss": 2.1025, "step": 14565 }, { "epoch": 0.36, "learning_rate": 1.4839684096195803e-05, "loss": 2.0494, "step": 14566 }, { "epoch": 0.36, "learning_rate": 1.4838985472804635e-05, "loss": 1.9478, "step": 14567 }, { "epoch": 0.36, "learning_rate": 1.4838286818573035e-05, "loss": 2.1327, "step": 14568 }, { "epoch": 0.36, "learning_rate": 1.4837588133505456e-05, "loss": 2.1805, "step": 14569 }, { "epoch": 0.36, "learning_rate": 1.483688941760635e-05, "loss": 2.1117, "step": 14570 }, { "epoch": 0.36, "learning_rate": 1.4836190670880175e-05, "loss": 2.1129, "step": 14571 }, { "epoch": 0.36, "learning_rate": 1.4835491893331378e-05, "loss": 2.1576, "step": 14572 }, { "epoch": 0.36, "learning_rate": 1.4834793084964416e-05, "loss": 1.9615, "step": 14573 }, { "epoch": 0.36, "learning_rate": 1.4834094245783741e-05, "loss": 2.2134, "step": 14574 }, { "epoch": 0.36, "learning_rate": 1.4833395375793806e-05, "loss": 2.1335, "step": 14575 }, { "epoch": 0.36, "learning_rate": 1.483269647499907e-05, "loss": 1.9774, "step": 14576 }, { "epoch": 0.36, "learning_rate": 1.4831997543403987e-05, "loss": 2.0603, "step": 14577 }, { "epoch": 0.36, "learning_rate": 1.4831298581013007e-05, "loss": 1.9688, "step": 14578 }, { "epoch": 0.36, "learning_rate": 1.4830599587830585e-05, "loss": 2.2047, "step": 14579 }, { "epoch": 0.36, "learning_rate": 1.4829900563861176e-05, "loss": 2.0138, "step": 14580 }, { "epoch": 0.36, "learning_rate": 1.4829201509109244e-05, "loss": 1.9992, "step": 14581 }, { "epoch": 0.36, "learning_rate": 1.4828502423579231e-05, "loss": 2.2499, "step": 14582 }, { "epoch": 0.36, "learning_rate": 1.4827803307275598e-05, "loss": 2.2873, "step": 14583 }, { "epoch": 0.36, "learning_rate": 1.4827104160202804e-05, "loss": 2.0877, "step": 14584 }, { "epoch": 0.36, "learning_rate": 1.4826404982365299e-05, "loss": 2.2135, "step": 14585 }, { "epoch": 0.36, "learning_rate": 1.4825705773767545e-05, "loss": 2.1349, "step": 14586 }, { "epoch": 0.36, "learning_rate": 1.4825006534413996e-05, "loss": 1.996, "step": 14587 }, { "epoch": 0.36, "learning_rate": 1.4824307264309105e-05, "loss": 2.1409, "step": 14588 }, { "epoch": 0.36, "learning_rate": 1.4823607963457337e-05, "loss": 1.9681, "step": 14589 }, { "epoch": 0.36, "learning_rate": 1.4822908631863135e-05, "loss": 1.8659, "step": 14590 }, { "epoch": 0.36, "learning_rate": 1.4822209269530971e-05, "loss": 2.1423, "step": 14591 }, { "epoch": 0.36, "learning_rate": 1.4821509876465294e-05, "loss": 2.127, "step": 14592 }, { "epoch": 0.36, "learning_rate": 1.4820810452670564e-05, "loss": 2.0802, "step": 14593 }, { "epoch": 0.36, "learning_rate": 1.4820110998151238e-05, "loss": 2.0791, "step": 14594 }, { "epoch": 0.36, "learning_rate": 1.481941151291177e-05, "loss": 2.0985, "step": 14595 }, { "epoch": 0.36, "learning_rate": 1.4818711996956625e-05, "loss": 2.1419, "step": 14596 }, { "epoch": 0.36, "learning_rate": 1.481801245029026e-05, "loss": 2.0674, "step": 14597 }, { "epoch": 0.36, "learning_rate": 1.481731287291713e-05, "loss": 1.9857, "step": 14598 }, { "epoch": 0.36, "learning_rate": 1.4816613264841692e-05, "loss": 2.0975, "step": 14599 }, { "epoch": 0.36, "learning_rate": 1.4815913626068411e-05, "loss": 2.3326, "step": 14600 }, { "epoch": 0.36, "learning_rate": 1.4815213956601742e-05, "loss": 2.2852, "step": 14601 }, { "epoch": 0.36, "learning_rate": 1.4814514256446147e-05, "loss": 2.0874, "step": 14602 }, { "epoch": 0.36, "learning_rate": 1.4813814525606082e-05, "loss": 1.9959, "step": 14603 }, { "epoch": 0.36, "learning_rate": 1.4813114764086008e-05, "loss": 2.2289, "step": 14604 }, { "epoch": 0.36, "learning_rate": 1.4812414971890385e-05, "loss": 2.0381, "step": 14605 }, { "epoch": 0.36, "learning_rate": 1.4811715149023675e-05, "loss": 2.1526, "step": 14606 }, { "epoch": 0.36, "learning_rate": 1.4811015295490336e-05, "loss": 2.0573, "step": 14607 }, { "epoch": 0.36, "learning_rate": 1.4810315411294829e-05, "loss": 2.0482, "step": 14608 }, { "epoch": 0.36, "learning_rate": 1.4809615496441612e-05, "loss": 2.0899, "step": 14609 }, { "epoch": 0.36, "learning_rate": 1.4808915550935146e-05, "loss": 1.9996, "step": 14610 }, { "epoch": 0.36, "learning_rate": 1.4808215574779898e-05, "loss": 2.1, "step": 14611 }, { "epoch": 0.36, "learning_rate": 1.4807515567980324e-05, "loss": 2.1081, "step": 14612 }, { "epoch": 0.36, "learning_rate": 1.4806815530540889e-05, "loss": 2.0188, "step": 14613 }, { "epoch": 0.36, "learning_rate": 1.4806115462466047e-05, "loss": 2.0559, "step": 14614 }, { "epoch": 0.36, "learning_rate": 1.480541536376027e-05, "loss": 2.1977, "step": 14615 }, { "epoch": 0.36, "learning_rate": 1.4804715234428012e-05, "loss": 1.9743, "step": 14616 }, { "epoch": 0.36, "learning_rate": 1.4804015074473739e-05, "loss": 2.1103, "step": 14617 }, { "epoch": 0.36, "learning_rate": 1.480331488390191e-05, "loss": 2.249, "step": 14618 }, { "epoch": 0.36, "learning_rate": 1.4802614662716992e-05, "loss": 2.0319, "step": 14619 }, { "epoch": 0.36, "learning_rate": 1.4801914410923446e-05, "loss": 1.9818, "step": 14620 }, { "epoch": 0.36, "learning_rate": 1.4801214128525733e-05, "loss": 2.2832, "step": 14621 }, { "epoch": 0.36, "learning_rate": 1.480051381552832e-05, "loss": 2.1716, "step": 14622 }, { "epoch": 0.36, "learning_rate": 1.4799813471935666e-05, "loss": 2.1258, "step": 14623 }, { "epoch": 0.36, "learning_rate": 1.4799113097752238e-05, "loss": 1.9877, "step": 14624 }, { "epoch": 0.36, "learning_rate": 1.4798412692982496e-05, "loss": 1.9975, "step": 14625 }, { "epoch": 0.36, "learning_rate": 1.4797712257630909e-05, "loss": 2.3152, "step": 14626 }, { "epoch": 0.36, "learning_rate": 1.4797011791701935e-05, "loss": 2.1383, "step": 14627 }, { "epoch": 0.36, "learning_rate": 1.4796311295200043e-05, "loss": 2.0707, "step": 14628 }, { "epoch": 0.36, "learning_rate": 1.4795610768129697e-05, "loss": 2.1373, "step": 14629 }, { "epoch": 0.36, "learning_rate": 1.4794910210495362e-05, "loss": 2.1927, "step": 14630 }, { "epoch": 0.36, "learning_rate": 1.4794209622301499e-05, "loss": 2.1461, "step": 14631 }, { "epoch": 0.36, "learning_rate": 1.4793509003552578e-05, "loss": 2.0524, "step": 14632 }, { "epoch": 0.36, "learning_rate": 1.4792808354253058e-05, "loss": 1.9287, "step": 14633 }, { "epoch": 0.36, "learning_rate": 1.4792107674407412e-05, "loss": 2.0966, "step": 14634 }, { "epoch": 0.36, "learning_rate": 1.4791406964020102e-05, "loss": 2.0488, "step": 14635 }, { "epoch": 0.36, "learning_rate": 1.479070622309559e-05, "loss": 2.1324, "step": 14636 }, { "epoch": 0.36, "learning_rate": 1.479000545163835e-05, "loss": 2.0385, "step": 14637 }, { "epoch": 0.36, "learning_rate": 1.4789304649652842e-05, "loss": 1.9486, "step": 14638 }, { "epoch": 0.36, "learning_rate": 1.4788603817143536e-05, "loss": 2.0262, "step": 14639 }, { "epoch": 0.36, "learning_rate": 1.4787902954114898e-05, "loss": 2.0107, "step": 14640 }, { "epoch": 0.36, "learning_rate": 1.4787202060571396e-05, "loss": 2.0888, "step": 14641 }, { "epoch": 0.36, "learning_rate": 1.478650113651749e-05, "loss": 2.3745, "step": 14642 }, { "epoch": 0.36, "learning_rate": 1.4785800181957654e-05, "loss": 2.1751, "step": 14643 }, { "epoch": 0.36, "learning_rate": 1.4785099196896355e-05, "loss": 2.0871, "step": 14644 }, { "epoch": 0.36, "learning_rate": 1.4784398181338058e-05, "loss": 2.0118, "step": 14645 }, { "epoch": 0.36, "learning_rate": 1.4783697135287234e-05, "loss": 2.1282, "step": 14646 }, { "epoch": 0.36, "learning_rate": 1.4782996058748348e-05, "loss": 2.1132, "step": 14647 }, { "epoch": 0.36, "learning_rate": 1.4782294951725868e-05, "loss": 2.065, "step": 14648 }, { "epoch": 0.36, "learning_rate": 1.4781593814224268e-05, "loss": 2.1871, "step": 14649 }, { "epoch": 0.36, "learning_rate": 1.4780892646248011e-05, "loss": 2.1507, "step": 14650 }, { "epoch": 0.36, "learning_rate": 1.4780191447801565e-05, "loss": 2.1542, "step": 14651 }, { "epoch": 0.36, "learning_rate": 1.4779490218889403e-05, "loss": 2.1768, "step": 14652 }, { "epoch": 0.36, "learning_rate": 1.477878895951599e-05, "loss": 2.1753, "step": 14653 }, { "epoch": 0.36, "learning_rate": 1.47780876696858e-05, "loss": 2.2709, "step": 14654 }, { "epoch": 0.36, "learning_rate": 1.4777386349403298e-05, "loss": 2.0797, "step": 14655 }, { "epoch": 0.36, "learning_rate": 1.4776684998672956e-05, "loss": 2.0411, "step": 14656 }, { "epoch": 0.36, "learning_rate": 1.4775983617499245e-05, "loss": 2.0834, "step": 14657 }, { "epoch": 0.36, "learning_rate": 1.4775282205886635e-05, "loss": 2.1537, "step": 14658 }, { "epoch": 0.36, "learning_rate": 1.4774580763839595e-05, "loss": 2.1224, "step": 14659 }, { "epoch": 0.36, "learning_rate": 1.4773879291362593e-05, "loss": 2.0799, "step": 14660 }, { "epoch": 0.36, "learning_rate": 1.4773177788460107e-05, "loss": 2.0253, "step": 14661 }, { "epoch": 0.36, "learning_rate": 1.4772476255136599e-05, "loss": 2.0401, "step": 14662 }, { "epoch": 0.36, "learning_rate": 1.4771774691396546e-05, "loss": 2.1538, "step": 14663 }, { "epoch": 0.36, "learning_rate": 1.4771073097244416e-05, "loss": 2.1098, "step": 14664 }, { "epoch": 0.36, "learning_rate": 1.4770371472684685e-05, "loss": 2.2159, "step": 14665 }, { "epoch": 0.36, "learning_rate": 1.4769669817721823e-05, "loss": 2.0203, "step": 14666 }, { "epoch": 0.36, "learning_rate": 1.4768968132360297e-05, "loss": 2.0959, "step": 14667 }, { "epoch": 0.36, "learning_rate": 1.4768266416604586e-05, "loss": 2.0785, "step": 14668 }, { "epoch": 0.36, "learning_rate": 1.476756467045916e-05, "loss": 2.1416, "step": 14669 }, { "epoch": 0.36, "learning_rate": 1.4766862893928488e-05, "loss": 1.9395, "step": 14670 }, { "epoch": 0.36, "learning_rate": 1.4766161087017046e-05, "loss": 2.0387, "step": 14671 }, { "epoch": 0.36, "learning_rate": 1.4765459249729307e-05, "loss": 2.1661, "step": 14672 }, { "epoch": 0.36, "learning_rate": 1.476475738206974e-05, "loss": 2.2254, "step": 14673 }, { "epoch": 0.36, "learning_rate": 1.4764055484042827e-05, "loss": 2.0939, "step": 14674 }, { "epoch": 0.36, "learning_rate": 1.4763353555653032e-05, "loss": 2.0987, "step": 14675 }, { "epoch": 0.36, "learning_rate": 1.4762651596904836e-05, "loss": 1.986, "step": 14676 }, { "epoch": 0.36, "learning_rate": 1.4761949607802705e-05, "loss": 2.0727, "step": 14677 }, { "epoch": 0.36, "learning_rate": 1.476124758835112e-05, "loss": 2.0735, "step": 14678 }, { "epoch": 0.36, "learning_rate": 1.4760545538554551e-05, "loss": 1.9947, "step": 14679 }, { "epoch": 0.36, "learning_rate": 1.4759843458417473e-05, "loss": 2.1559, "step": 14680 }, { "epoch": 0.36, "learning_rate": 1.4759141347944363e-05, "loss": 2.2123, "step": 14681 }, { "epoch": 0.36, "learning_rate": 1.4758439207139694e-05, "loss": 2.0444, "step": 14682 }, { "epoch": 0.36, "learning_rate": 1.475773703600794e-05, "loss": 1.9812, "step": 14683 }, { "epoch": 0.36, "learning_rate": 1.4757034834553581e-05, "loss": 2.0888, "step": 14684 }, { "epoch": 0.36, "learning_rate": 1.4756332602781087e-05, "loss": 2.001, "step": 14685 }, { "epoch": 0.36, "learning_rate": 1.4755630340694932e-05, "loss": 2.0939, "step": 14686 }, { "epoch": 0.36, "learning_rate": 1.4754928048299597e-05, "loss": 2.3136, "step": 14687 }, { "epoch": 0.36, "learning_rate": 1.4754225725599558e-05, "loss": 2.0917, "step": 14688 }, { "epoch": 0.36, "learning_rate": 1.4753523372599285e-05, "loss": 2.1491, "step": 14689 }, { "epoch": 0.36, "learning_rate": 1.4752820989303263e-05, "loss": 2.1856, "step": 14690 }, { "epoch": 0.36, "learning_rate": 1.4752118575715961e-05, "loss": 2.238, "step": 14691 }, { "epoch": 0.36, "learning_rate": 1.475141613184186e-05, "loss": 2.1823, "step": 14692 }, { "epoch": 0.36, "learning_rate": 1.4750713657685435e-05, "loss": 2.1167, "step": 14693 }, { "epoch": 0.36, "learning_rate": 1.4750011153251165e-05, "loss": 2.1165, "step": 14694 }, { "epoch": 0.36, "learning_rate": 1.4749308618543526e-05, "loss": 2.1253, "step": 14695 }, { "epoch": 0.36, "learning_rate": 1.4748606053566992e-05, "loss": 2.1069, "step": 14696 }, { "epoch": 0.36, "learning_rate": 1.4747903458326048e-05, "loss": 2.1793, "step": 14697 }, { "epoch": 0.36, "learning_rate": 1.4747200832825164e-05, "loss": 2.1171, "step": 14698 }, { "epoch": 0.36, "learning_rate": 1.4746498177068824e-05, "loss": 2.1614, "step": 14699 }, { "epoch": 0.36, "learning_rate": 1.4745795491061505e-05, "loss": 2.0449, "step": 14700 }, { "epoch": 0.36, "learning_rate": 1.4745092774807683e-05, "loss": 2.004, "step": 14701 }, { "epoch": 0.36, "learning_rate": 1.474439002831184e-05, "loss": 2.1011, "step": 14702 }, { "epoch": 0.36, "learning_rate": 1.4743687251578455e-05, "loss": 2.1211, "step": 14703 }, { "epoch": 0.36, "learning_rate": 1.4742984444612003e-05, "loss": 2.1677, "step": 14704 }, { "epoch": 0.36, "learning_rate": 1.4742281607416963e-05, "loss": 2.0929, "step": 14705 }, { "epoch": 0.36, "learning_rate": 1.4741578739997821e-05, "loss": 2.0701, "step": 14706 }, { "epoch": 0.36, "learning_rate": 1.474087584235905e-05, "loss": 2.1536, "step": 14707 }, { "epoch": 0.36, "learning_rate": 1.4740172914505134e-05, "loss": 2.1043, "step": 14708 }, { "epoch": 0.36, "learning_rate": 1.4739469956440551e-05, "loss": 1.9832, "step": 14709 }, { "epoch": 0.36, "learning_rate": 1.4738766968169779e-05, "loss": 1.99, "step": 14710 }, { "epoch": 0.36, "learning_rate": 1.47380639496973e-05, "loss": 2.057, "step": 14711 }, { "epoch": 0.36, "learning_rate": 1.4737360901027602e-05, "loss": 2.1544, "step": 14712 }, { "epoch": 0.36, "learning_rate": 1.4736657822165154e-05, "loss": 2.3031, "step": 14713 }, { "epoch": 0.36, "learning_rate": 1.4735954713114441e-05, "loss": 2.1073, "step": 14714 }, { "epoch": 0.36, "learning_rate": 1.473525157387995e-05, "loss": 2.2036, "step": 14715 }, { "epoch": 0.36, "learning_rate": 1.4734548404466154e-05, "loss": 2.0866, "step": 14716 }, { "epoch": 0.36, "learning_rate": 1.4733845204877537e-05, "loss": 2.1192, "step": 14717 }, { "epoch": 0.36, "learning_rate": 1.4733141975118586e-05, "loss": 2.0998, "step": 14718 }, { "epoch": 0.36, "learning_rate": 1.4732438715193775e-05, "loss": 2.1086, "step": 14719 }, { "epoch": 0.36, "learning_rate": 1.4731735425107592e-05, "loss": 2.0072, "step": 14720 }, { "epoch": 0.36, "learning_rate": 1.4731032104864516e-05, "loss": 2.115, "step": 14721 }, { "epoch": 0.36, "learning_rate": 1.473032875446903e-05, "loss": 2.1355, "step": 14722 }, { "epoch": 0.36, "learning_rate": 1.472962537392562e-05, "loss": 2.1291, "step": 14723 }, { "epoch": 0.36, "learning_rate": 1.4728921963238762e-05, "loss": 2.1431, "step": 14724 }, { "epoch": 0.36, "learning_rate": 1.4728218522412945e-05, "loss": 2.1699, "step": 14725 }, { "epoch": 0.36, "learning_rate": 1.4727515051452652e-05, "loss": 2.1035, "step": 14726 }, { "epoch": 0.36, "learning_rate": 1.4726811550362364e-05, "loss": 2.0823, "step": 14727 }, { "epoch": 0.36, "learning_rate": 1.4726108019146566e-05, "loss": 2.1952, "step": 14728 }, { "epoch": 0.36, "learning_rate": 1.4725404457809742e-05, "loss": 2.0394, "step": 14729 }, { "epoch": 0.36, "learning_rate": 1.4724700866356372e-05, "loss": 2.2976, "step": 14730 }, { "epoch": 0.36, "learning_rate": 1.4723997244790946e-05, "loss": 2.0488, "step": 14731 }, { "epoch": 0.36, "learning_rate": 1.472329359311795e-05, "loss": 2.2167, "step": 14732 }, { "epoch": 0.36, "learning_rate": 1.4722589911341856e-05, "loss": 2.2072, "step": 14733 }, { "epoch": 0.36, "learning_rate": 1.4721886199467164e-05, "loss": 2.1327, "step": 14734 }, { "epoch": 0.36, "learning_rate": 1.472118245749835e-05, "loss": 2.123, "step": 14735 }, { "epoch": 0.36, "learning_rate": 1.47204786854399e-05, "loss": 2.0751, "step": 14736 }, { "epoch": 0.36, "learning_rate": 1.4719774883296305e-05, "loss": 2.106, "step": 14737 }, { "epoch": 0.36, "learning_rate": 1.4719071051072045e-05, "loss": 2.0581, "step": 14738 }, { "epoch": 0.36, "learning_rate": 1.4718367188771607e-05, "loss": 2.0766, "step": 14739 }, { "epoch": 0.36, "learning_rate": 1.4717663296399476e-05, "loss": 2.0744, "step": 14740 }, { "epoch": 0.36, "learning_rate": 1.4716959373960143e-05, "loss": 2.1138, "step": 14741 }, { "epoch": 0.36, "learning_rate": 1.4716255421458086e-05, "loss": 2.2663, "step": 14742 }, { "epoch": 0.36, "learning_rate": 1.4715551438897799e-05, "loss": 1.9745, "step": 14743 }, { "epoch": 0.36, "learning_rate": 1.4714847426283765e-05, "loss": 2.1228, "step": 14744 }, { "epoch": 0.36, "learning_rate": 1.4714143383620471e-05, "loss": 2.0799, "step": 14745 }, { "epoch": 0.36, "learning_rate": 1.4713439310912407e-05, "loss": 2.2179, "step": 14746 }, { "epoch": 0.36, "learning_rate": 1.4712735208164057e-05, "loss": 2.0056, "step": 14747 }, { "epoch": 0.36, "learning_rate": 1.471203107537991e-05, "loss": 2.2359, "step": 14748 }, { "epoch": 0.36, "learning_rate": 1.4711326912564455e-05, "loss": 2.1604, "step": 14749 }, { "epoch": 0.36, "learning_rate": 1.4710622719722177e-05, "loss": 2.2236, "step": 14750 }, { "epoch": 0.36, "learning_rate": 1.4709918496857562e-05, "loss": 2.0577, "step": 14751 }, { "epoch": 0.36, "learning_rate": 1.4709214243975106e-05, "loss": 2.0801, "step": 14752 }, { "epoch": 0.36, "learning_rate": 1.470850996107929e-05, "loss": 2.1321, "step": 14753 }, { "epoch": 0.36, "learning_rate": 1.4707805648174608e-05, "loss": 1.9995, "step": 14754 }, { "epoch": 0.36, "learning_rate": 1.4707101305265546e-05, "loss": 2.1369, "step": 14755 }, { "epoch": 0.36, "learning_rate": 1.4706396932356593e-05, "loss": 2.1421, "step": 14756 }, { "epoch": 0.36, "learning_rate": 1.4705692529452238e-05, "loss": 2.0367, "step": 14757 }, { "epoch": 0.36, "learning_rate": 1.470498809655697e-05, "loss": 2.07, "step": 14758 }, { "epoch": 0.36, "learning_rate": 1.4704283633675281e-05, "loss": 2.0988, "step": 14759 }, { "epoch": 0.36, "learning_rate": 1.4703579140811657e-05, "loss": 2.0466, "step": 14760 }, { "epoch": 0.36, "learning_rate": 1.4702874617970593e-05, "loss": 2.0997, "step": 14761 }, { "epoch": 0.36, "learning_rate": 1.4702170065156575e-05, "loss": 2.1583, "step": 14762 }, { "epoch": 0.36, "learning_rate": 1.4701465482374097e-05, "loss": 2.1503, "step": 14763 }, { "epoch": 0.36, "learning_rate": 1.4700760869627643e-05, "loss": 2.0133, "step": 14764 }, { "epoch": 0.36, "learning_rate": 1.4700056226921714e-05, "loss": 2.1441, "step": 14765 }, { "epoch": 0.36, "learning_rate": 1.4699351554260793e-05, "loss": 2.0762, "step": 14766 }, { "epoch": 0.36, "learning_rate": 1.469864685164937e-05, "loss": 1.9465, "step": 14767 }, { "epoch": 0.36, "learning_rate": 1.4697942119091942e-05, "loss": 2.1151, "step": 14768 }, { "epoch": 0.36, "learning_rate": 1.4697237356592995e-05, "loss": 2.1321, "step": 14769 }, { "epoch": 0.36, "learning_rate": 1.4696532564157026e-05, "loss": 1.9317, "step": 14770 }, { "epoch": 0.36, "learning_rate": 1.4695827741788524e-05, "loss": 2.04, "step": 14771 }, { "epoch": 0.36, "learning_rate": 1.4695122889491984e-05, "loss": 1.9007, "step": 14772 }, { "epoch": 0.36, "learning_rate": 1.4694418007271893e-05, "loss": 1.8212, "step": 14773 }, { "epoch": 0.36, "learning_rate": 1.4693713095132744e-05, "loss": 2.0943, "step": 14774 }, { "epoch": 0.36, "learning_rate": 1.4693008153079037e-05, "loss": 2.1222, "step": 14775 }, { "epoch": 0.36, "learning_rate": 1.4692303181115254e-05, "loss": 1.8997, "step": 14776 }, { "epoch": 0.36, "learning_rate": 1.4691598179245898e-05, "loss": 2.0408, "step": 14777 }, { "epoch": 0.36, "learning_rate": 1.4690893147475454e-05, "loss": 1.9977, "step": 14778 }, { "epoch": 0.36, "learning_rate": 1.4690188085808423e-05, "loss": 2.0289, "step": 14779 }, { "epoch": 0.36, "learning_rate": 1.4689482994249294e-05, "loss": 2.1331, "step": 14780 }, { "epoch": 0.36, "learning_rate": 1.4688777872802562e-05, "loss": 2.0591, "step": 14781 }, { "epoch": 0.36, "learning_rate": 1.468807272147272e-05, "loss": 2.1135, "step": 14782 }, { "epoch": 0.36, "learning_rate": 1.468736754026426e-05, "loss": 2.2374, "step": 14783 }, { "epoch": 0.36, "learning_rate": 1.4686662329181683e-05, "loss": 1.8844, "step": 14784 }, { "epoch": 0.36, "learning_rate": 1.4685957088229476e-05, "loss": 2.1458, "step": 14785 }, { "epoch": 0.36, "learning_rate": 1.4685251817412143e-05, "loss": 2.1949, "step": 14786 }, { "epoch": 0.36, "learning_rate": 1.4684546516734168e-05, "loss": 2.0886, "step": 14787 }, { "epoch": 0.36, "learning_rate": 1.4683841186200055e-05, "loss": 1.9675, "step": 14788 }, { "epoch": 0.36, "learning_rate": 1.4683135825814292e-05, "loss": 1.9947, "step": 14789 }, { "epoch": 0.36, "learning_rate": 1.4682430435581381e-05, "loss": 1.9949, "step": 14790 }, { "epoch": 0.36, "learning_rate": 1.4681725015505814e-05, "loss": 2.1513, "step": 14791 }, { "epoch": 0.36, "learning_rate": 1.4681019565592089e-05, "loss": 2.1071, "step": 14792 }, { "epoch": 0.36, "learning_rate": 1.4680314085844698e-05, "loss": 1.9611, "step": 14793 }, { "epoch": 0.36, "learning_rate": 1.467960857626814e-05, "loss": 2.2055, "step": 14794 }, { "epoch": 0.36, "learning_rate": 1.4678903036866913e-05, "loss": 2.2142, "step": 14795 }, { "epoch": 0.36, "learning_rate": 1.4678197467645514e-05, "loss": 2.2182, "step": 14796 }, { "epoch": 0.36, "learning_rate": 1.4677491868608438e-05, "loss": 2.1024, "step": 14797 }, { "epoch": 0.36, "learning_rate": 1.4676786239760179e-05, "loss": 2.169, "step": 14798 }, { "epoch": 0.36, "learning_rate": 1.467608058110524e-05, "loss": 1.9874, "step": 14799 }, { "epoch": 0.36, "learning_rate": 1.4675374892648112e-05, "loss": 2.2422, "step": 14800 }, { "epoch": 0.36, "learning_rate": 1.4674669174393298e-05, "loss": 2.1752, "step": 14801 }, { "epoch": 0.36, "learning_rate": 1.4673963426345292e-05, "loss": 2.0724, "step": 14802 }, { "epoch": 0.36, "learning_rate": 1.4673257648508597e-05, "loss": 2.1316, "step": 14803 }, { "epoch": 0.36, "learning_rate": 1.4672551840887708e-05, "loss": 2.1434, "step": 14804 }, { "epoch": 0.36, "learning_rate": 1.4671846003487123e-05, "loss": 2.2447, "step": 14805 }, { "epoch": 0.36, "learning_rate": 1.4671140136311339e-05, "loss": 2.2689, "step": 14806 }, { "epoch": 0.36, "learning_rate": 1.4670434239364857e-05, "loss": 2.0585, "step": 14807 }, { "epoch": 0.36, "learning_rate": 1.4669728312652176e-05, "loss": 2.1529, "step": 14808 }, { "epoch": 0.37, "learning_rate": 1.4669022356177796e-05, "loss": 2.1111, "step": 14809 }, { "epoch": 0.37, "learning_rate": 1.4668316369946216e-05, "loss": 2.22, "step": 14810 }, { "epoch": 0.37, "learning_rate": 1.466761035396193e-05, "loss": 2.1047, "step": 14811 }, { "epoch": 0.37, "learning_rate": 1.4666904308229444e-05, "loss": 2.2538, "step": 14812 }, { "epoch": 0.37, "learning_rate": 1.4666198232753254e-05, "loss": 2.173, "step": 14813 }, { "epoch": 0.37, "learning_rate": 1.4665492127537868e-05, "loss": 2.2027, "step": 14814 }, { "epoch": 0.37, "learning_rate": 1.4664785992587776e-05, "loss": 2.0369, "step": 14815 }, { "epoch": 0.37, "learning_rate": 1.4664079827907485e-05, "loss": 2.1709, "step": 14816 }, { "epoch": 0.37, "learning_rate": 1.466337363350149e-05, "loss": 2.0954, "step": 14817 }, { "epoch": 0.37, "learning_rate": 1.4662667409374295e-05, "loss": 2.0853, "step": 14818 }, { "epoch": 0.37, "learning_rate": 1.4661961155530404e-05, "loss": 2.1512, "step": 14819 }, { "epoch": 0.37, "learning_rate": 1.4661254871974312e-05, "loss": 2.2673, "step": 14820 }, { "epoch": 0.37, "learning_rate": 1.4660548558710527e-05, "loss": 1.9997, "step": 14821 }, { "epoch": 0.37, "learning_rate": 1.4659842215743545e-05, "loss": 2.0402, "step": 14822 }, { "epoch": 0.37, "learning_rate": 1.4659135843077872e-05, "loss": 1.9929, "step": 14823 }, { "epoch": 0.37, "learning_rate": 1.4658429440718008e-05, "loss": 1.9523, "step": 14824 }, { "epoch": 0.37, "learning_rate": 1.4657723008668452e-05, "loss": 2.0412, "step": 14825 }, { "epoch": 0.37, "learning_rate": 1.4657016546933713e-05, "loss": 1.9604, "step": 14826 }, { "epoch": 0.37, "learning_rate": 1.4656310055518286e-05, "loss": 2.1108, "step": 14827 }, { "epoch": 0.37, "learning_rate": 1.465560353442668e-05, "loss": 2.2865, "step": 14828 }, { "epoch": 0.37, "learning_rate": 1.4654896983663393e-05, "loss": 1.9691, "step": 14829 }, { "epoch": 0.37, "learning_rate": 1.4654190403232935e-05, "loss": 2.0528, "step": 14830 }, { "epoch": 0.37, "learning_rate": 1.4653483793139801e-05, "loss": 2.0704, "step": 14831 }, { "epoch": 0.37, "learning_rate": 1.46527771533885e-05, "loss": 2.0636, "step": 14832 }, { "epoch": 0.37, "learning_rate": 1.4652070483983534e-05, "loss": 2.0592, "step": 14833 }, { "epoch": 0.37, "learning_rate": 1.4651363784929407e-05, "loss": 2.1458, "step": 14834 }, { "epoch": 0.37, "learning_rate": 1.4650657056230622e-05, "loss": 2.2299, "step": 14835 }, { "epoch": 0.37, "learning_rate": 1.4649950297891683e-05, "loss": 2.202, "step": 14836 }, { "epoch": 0.37, "learning_rate": 1.4649243509917097e-05, "loss": 2.098, "step": 14837 }, { "epoch": 0.37, "learning_rate": 1.4648536692311364e-05, "loss": 2.0368, "step": 14838 }, { "epoch": 0.37, "learning_rate": 1.4647829845078996e-05, "loss": 2.0997, "step": 14839 }, { "epoch": 0.37, "learning_rate": 1.4647122968224493e-05, "loss": 2.1342, "step": 14840 }, { "epoch": 0.37, "learning_rate": 1.4646416061752357e-05, "loss": 1.9745, "step": 14841 }, { "epoch": 0.37, "learning_rate": 1.46457091256671e-05, "loss": 2.1165, "step": 14842 }, { "epoch": 0.37, "learning_rate": 1.4645002159973226e-05, "loss": 2.1165, "step": 14843 }, { "epoch": 0.37, "learning_rate": 1.4644295164675237e-05, "loss": 2.0869, "step": 14844 }, { "epoch": 0.37, "learning_rate": 1.4643588139777645e-05, "loss": 2.0299, "step": 14845 }, { "epoch": 0.37, "learning_rate": 1.4642881085284949e-05, "loss": 2.2014, "step": 14846 }, { "epoch": 0.37, "learning_rate": 1.4642174001201656e-05, "loss": 2.1204, "step": 14847 }, { "epoch": 0.37, "learning_rate": 1.464146688753228e-05, "loss": 2.115, "step": 14848 }, { "epoch": 0.37, "learning_rate": 1.4640759744281324e-05, "loss": 2.0562, "step": 14849 }, { "epoch": 0.37, "learning_rate": 1.4640052571453291e-05, "loss": 2.2416, "step": 14850 }, { "epoch": 0.37, "learning_rate": 1.4639345369052691e-05, "loss": 2.1318, "step": 14851 }, { "epoch": 0.37, "learning_rate": 1.4638638137084032e-05, "loss": 2.0332, "step": 14852 }, { "epoch": 0.37, "learning_rate": 1.463793087555182e-05, "loss": 2.1097, "step": 14853 }, { "epoch": 0.37, "learning_rate": 1.4637223584460563e-05, "loss": 2.0713, "step": 14854 }, { "epoch": 0.37, "learning_rate": 1.463651626381477e-05, "loss": 2.1817, "step": 14855 }, { "epoch": 0.37, "learning_rate": 1.4635808913618942e-05, "loss": 2.1891, "step": 14856 }, { "epoch": 0.37, "learning_rate": 1.46351015338776e-05, "loss": 2.0813, "step": 14857 }, { "epoch": 0.37, "learning_rate": 1.4634394124595243e-05, "loss": 2.2118, "step": 14858 }, { "epoch": 0.37, "learning_rate": 1.4633686685776382e-05, "loss": 2.1431, "step": 14859 }, { "epoch": 0.37, "learning_rate": 1.4632979217425529e-05, "loss": 2.2152, "step": 14860 }, { "epoch": 0.37, "learning_rate": 1.4632271719547183e-05, "loss": 2.0511, "step": 14861 }, { "epoch": 0.37, "learning_rate": 1.4631564192145864e-05, "loss": 2.1983, "step": 14862 }, { "epoch": 0.37, "learning_rate": 1.4630856635226076e-05, "loss": 2.1642, "step": 14863 }, { "epoch": 0.37, "learning_rate": 1.463014904879233e-05, "loss": 2.1848, "step": 14864 }, { "epoch": 0.37, "learning_rate": 1.4629441432849131e-05, "loss": 2.0506, "step": 14865 }, { "epoch": 0.37, "learning_rate": 1.4628733787400999e-05, "loss": 2.1872, "step": 14866 }, { "epoch": 0.37, "learning_rate": 1.4628026112452439e-05, "loss": 2.0603, "step": 14867 }, { "epoch": 0.37, "learning_rate": 1.4627318408007955e-05, "loss": 2.2005, "step": 14868 }, { "epoch": 0.37, "learning_rate": 1.4626610674072068e-05, "loss": 2.0894, "step": 14869 }, { "epoch": 0.37, "learning_rate": 1.462590291064928e-05, "loss": 2.3544, "step": 14870 }, { "epoch": 0.37, "learning_rate": 1.4625195117744105e-05, "loss": 2.2363, "step": 14871 }, { "epoch": 0.37, "learning_rate": 1.4624487295361056e-05, "loss": 2.2183, "step": 14872 }, { "epoch": 0.37, "learning_rate": 1.462377944350464e-05, "loss": 2.0197, "step": 14873 }, { "epoch": 0.37, "learning_rate": 1.4623071562179372e-05, "loss": 1.9905, "step": 14874 }, { "epoch": 0.37, "learning_rate": 1.4622363651389761e-05, "loss": 2.1514, "step": 14875 }, { "epoch": 0.37, "learning_rate": 1.462165571114032e-05, "loss": 2.0209, "step": 14876 }, { "epoch": 0.37, "learning_rate": 1.4620947741435565e-05, "loss": 2.149, "step": 14877 }, { "epoch": 0.37, "learning_rate": 1.4620239742280003e-05, "loss": 2.0496, "step": 14878 }, { "epoch": 0.37, "learning_rate": 1.4619531713678146e-05, "loss": 2.1604, "step": 14879 }, { "epoch": 0.37, "learning_rate": 1.4618823655634507e-05, "loss": 1.9742, "step": 14880 }, { "epoch": 0.37, "learning_rate": 1.4618115568153601e-05, "loss": 2.2353, "step": 14881 }, { "epoch": 0.37, "learning_rate": 1.4617407451239937e-05, "loss": 2.1983, "step": 14882 }, { "epoch": 0.37, "learning_rate": 1.4616699304898035e-05, "loss": 2.2191, "step": 14883 }, { "epoch": 0.37, "learning_rate": 1.4615991129132401e-05, "loss": 2.0919, "step": 14884 }, { "epoch": 0.37, "learning_rate": 1.4615282923947548e-05, "loss": 2.0999, "step": 14885 }, { "epoch": 0.37, "learning_rate": 1.4614574689347997e-05, "loss": 2.1594, "step": 14886 }, { "epoch": 0.37, "learning_rate": 1.461386642533826e-05, "loss": 1.9808, "step": 14887 }, { "epoch": 0.37, "learning_rate": 1.4613158131922847e-05, "loss": 2.0696, "step": 14888 }, { "epoch": 0.37, "learning_rate": 1.4612449809106273e-05, "loss": 2.0806, "step": 14889 }, { "epoch": 0.37, "learning_rate": 1.4611741456893051e-05, "loss": 1.9632, "step": 14890 }, { "epoch": 0.37, "learning_rate": 1.4611033075287697e-05, "loss": 1.9813, "step": 14891 }, { "epoch": 0.37, "learning_rate": 1.461032466429473e-05, "loss": 2.0579, "step": 14892 }, { "epoch": 0.37, "learning_rate": 1.4609616223918664e-05, "loss": 2.0864, "step": 14893 }, { "epoch": 0.37, "learning_rate": 1.4608907754164005e-05, "loss": 2.1859, "step": 14894 }, { "epoch": 0.37, "learning_rate": 1.4608199255035279e-05, "loss": 1.9623, "step": 14895 }, { "epoch": 0.37, "learning_rate": 1.4607490726536998e-05, "loss": 2.1155, "step": 14896 }, { "epoch": 0.37, "learning_rate": 1.4606782168673675e-05, "loss": 2.2409, "step": 14897 }, { "epoch": 0.37, "learning_rate": 1.4606073581449829e-05, "loss": 1.9726, "step": 14898 }, { "epoch": 0.37, "learning_rate": 1.4605364964869973e-05, "loss": 2.0701, "step": 14899 }, { "epoch": 0.37, "learning_rate": 1.4604656318938624e-05, "loss": 2.1706, "step": 14900 }, { "epoch": 0.37, "learning_rate": 1.4603947643660304e-05, "loss": 2.0574, "step": 14901 }, { "epoch": 0.37, "learning_rate": 1.4603238939039523e-05, "loss": 2.0923, "step": 14902 }, { "epoch": 0.37, "learning_rate": 1.4602530205080803e-05, "loss": 2.1275, "step": 14903 }, { "epoch": 0.37, "learning_rate": 1.4601821441788655e-05, "loss": 2.1445, "step": 14904 }, { "epoch": 0.37, "learning_rate": 1.4601112649167598e-05, "loss": 2.0668, "step": 14905 }, { "epoch": 0.37, "learning_rate": 1.4600403827222154e-05, "loss": 2.398, "step": 14906 }, { "epoch": 0.37, "learning_rate": 1.4599694975956837e-05, "loss": 2.2018, "step": 14907 }, { "epoch": 0.37, "learning_rate": 1.4598986095376163e-05, "loss": 1.9483, "step": 14908 }, { "epoch": 0.37, "learning_rate": 1.4598277185484651e-05, "loss": 2.1042, "step": 14909 }, { "epoch": 0.37, "learning_rate": 1.4597568246286821e-05, "loss": 2.0145, "step": 14910 }, { "epoch": 0.37, "learning_rate": 1.4596859277787191e-05, "loss": 2.1081, "step": 14911 }, { "epoch": 0.37, "learning_rate": 1.459615027999028e-05, "loss": 2.2224, "step": 14912 }, { "epoch": 0.37, "learning_rate": 1.4595441252900604e-05, "loss": 2.1658, "step": 14913 }, { "epoch": 0.37, "learning_rate": 1.4594732196522683e-05, "loss": 2.0119, "step": 14914 }, { "epoch": 0.37, "learning_rate": 1.4594023110861038e-05, "loss": 2.0863, "step": 14915 }, { "epoch": 0.37, "learning_rate": 1.4593313995920184e-05, "loss": 2.0337, "step": 14916 }, { "epoch": 0.37, "learning_rate": 1.4592604851704643e-05, "loss": 2.1256, "step": 14917 }, { "epoch": 0.37, "learning_rate": 1.4591895678218937e-05, "loss": 2.0328, "step": 14918 }, { "epoch": 0.37, "learning_rate": 1.459118647546758e-05, "loss": 2.1973, "step": 14919 }, { "epoch": 0.37, "learning_rate": 1.4590477243455098e-05, "loss": 2.2472, "step": 14920 }, { "epoch": 0.37, "learning_rate": 1.4589767982186009e-05, "loss": 2.0776, "step": 14921 }, { "epoch": 0.37, "learning_rate": 1.4589058691664832e-05, "loss": 2.0849, "step": 14922 }, { "epoch": 0.37, "learning_rate": 1.4588349371896092e-05, "loss": 2.0951, "step": 14923 }, { "epoch": 0.37, "learning_rate": 1.45876400228843e-05, "loss": 2.099, "step": 14924 }, { "epoch": 0.37, "learning_rate": 1.4586930644633985e-05, "loss": 2.0392, "step": 14925 }, { "epoch": 0.37, "learning_rate": 1.4586221237149664e-05, "loss": 2.1318, "step": 14926 }, { "epoch": 0.37, "learning_rate": 1.4585511800435865e-05, "loss": 2.2085, "step": 14927 }, { "epoch": 0.37, "learning_rate": 1.45848023344971e-05, "loss": 2.0131, "step": 14928 }, { "epoch": 0.37, "learning_rate": 1.4584092839337902e-05, "loss": 2.0278, "step": 14929 }, { "epoch": 0.37, "learning_rate": 1.4583383314962784e-05, "loss": 2.0206, "step": 14930 }, { "epoch": 0.37, "learning_rate": 1.4582673761376269e-05, "loss": 1.9982, "step": 14931 }, { "epoch": 0.37, "learning_rate": 1.4581964178582882e-05, "loss": 2.0522, "step": 14932 }, { "epoch": 0.37, "learning_rate": 1.4581254566587141e-05, "loss": 2.2625, "step": 14933 }, { "epoch": 0.37, "learning_rate": 1.4580544925393575e-05, "loss": 2.0958, "step": 14934 }, { "epoch": 0.37, "learning_rate": 1.4579835255006702e-05, "loss": 2.1442, "step": 14935 }, { "epoch": 0.37, "learning_rate": 1.4579125555431048e-05, "loss": 2.1171, "step": 14936 }, { "epoch": 0.37, "learning_rate": 1.4578415826671133e-05, "loss": 2.1913, "step": 14937 }, { "epoch": 0.37, "learning_rate": 1.4577706068731481e-05, "loss": 2.1875, "step": 14938 }, { "epoch": 0.37, "learning_rate": 1.4576996281616617e-05, "loss": 2.1887, "step": 14939 }, { "epoch": 0.37, "learning_rate": 1.4576286465331066e-05, "loss": 2.2411, "step": 14940 }, { "epoch": 0.37, "learning_rate": 1.457557661987935e-05, "loss": 1.8952, "step": 14941 }, { "epoch": 0.37, "learning_rate": 1.4574866745265991e-05, "loss": 2.0151, "step": 14942 }, { "epoch": 0.37, "learning_rate": 1.4574156841495515e-05, "loss": 2.2938, "step": 14943 }, { "epoch": 0.37, "learning_rate": 1.4573446908572448e-05, "loss": 2.2158, "step": 14944 }, { "epoch": 0.37, "learning_rate": 1.4572736946501314e-05, "loss": 2.1456, "step": 14945 }, { "epoch": 0.37, "learning_rate": 1.4572026955286636e-05, "loss": 2.082, "step": 14946 }, { "epoch": 0.37, "learning_rate": 1.4571316934932942e-05, "loss": 2.0583, "step": 14947 }, { "epoch": 0.37, "learning_rate": 1.4570606885444752e-05, "loss": 2.077, "step": 14948 }, { "epoch": 0.37, "learning_rate": 1.4569896806826597e-05, "loss": 2.193, "step": 14949 }, { "epoch": 0.37, "learning_rate": 1.4569186699083003e-05, "loss": 2.0566, "step": 14950 }, { "epoch": 0.37, "learning_rate": 1.456847656221849e-05, "loss": 2.1796, "step": 14951 }, { "epoch": 0.37, "learning_rate": 1.4567766396237586e-05, "loss": 2.2708, "step": 14952 }, { "epoch": 0.37, "learning_rate": 1.4567056201144819e-05, "loss": 1.9386, "step": 14953 }, { "epoch": 0.37, "learning_rate": 1.4566345976944716e-05, "loss": 2.0579, "step": 14954 }, { "epoch": 0.37, "learning_rate": 1.45656357236418e-05, "loss": 2.029, "step": 14955 }, { "epoch": 0.37, "learning_rate": 1.4564925441240601e-05, "loss": 1.9949, "step": 14956 }, { "epoch": 0.37, "learning_rate": 1.4564215129745645e-05, "loss": 2.1379, "step": 14957 }, { "epoch": 0.37, "learning_rate": 1.4563504789161455e-05, "loss": 2.2205, "step": 14958 }, { "epoch": 0.37, "learning_rate": 1.4562794419492568e-05, "loss": 2.0411, "step": 14959 }, { "epoch": 0.37, "learning_rate": 1.4562084020743502e-05, "loss": 2.1057, "step": 14960 }, { "epoch": 0.37, "learning_rate": 1.4561373592918786e-05, "loss": 2.1504, "step": 14961 }, { "epoch": 0.37, "learning_rate": 1.456066313602295e-05, "loss": 2.203, "step": 14962 }, { "epoch": 0.37, "learning_rate": 1.4559952650060523e-05, "loss": 2.1513, "step": 14963 }, { "epoch": 0.37, "learning_rate": 1.4559242135036031e-05, "loss": 2.1357, "step": 14964 }, { "epoch": 0.37, "learning_rate": 1.4558531590954004e-05, "loss": 1.9733, "step": 14965 }, { "epoch": 0.37, "learning_rate": 1.455782101781897e-05, "loss": 2.0693, "step": 14966 }, { "epoch": 0.37, "learning_rate": 1.4557110415635456e-05, "loss": 2.1264, "step": 14967 }, { "epoch": 0.37, "learning_rate": 1.4556399784407992e-05, "loss": 1.9552, "step": 14968 }, { "epoch": 0.37, "learning_rate": 1.4555689124141108e-05, "loss": 2.0778, "step": 14969 }, { "epoch": 0.37, "learning_rate": 1.455497843483933e-05, "loss": 2.1134, "step": 14970 }, { "epoch": 0.37, "learning_rate": 1.4554267716507195e-05, "loss": 1.9847, "step": 14971 }, { "epoch": 0.37, "learning_rate": 1.4553556969149223e-05, "loss": 2.0914, "step": 14972 }, { "epoch": 0.37, "learning_rate": 1.4552846192769952e-05, "loss": 2.186, "step": 14973 }, { "epoch": 0.37, "learning_rate": 1.4552135387373906e-05, "loss": 2.2314, "step": 14974 }, { "epoch": 0.37, "learning_rate": 1.455142455296562e-05, "loss": 2.167, "step": 14975 }, { "epoch": 0.37, "learning_rate": 1.455071368954962e-05, "loss": 1.9453, "step": 14976 }, { "epoch": 0.37, "learning_rate": 1.455000279713044e-05, "loss": 2.0607, "step": 14977 }, { "epoch": 0.37, "learning_rate": 1.4549291875712607e-05, "loss": 2.1025, "step": 14978 }, { "epoch": 0.37, "learning_rate": 1.4548580925300657e-05, "loss": 1.8142, "step": 14979 }, { "epoch": 0.37, "learning_rate": 1.4547869945899118e-05, "loss": 2.1897, "step": 14980 }, { "epoch": 0.37, "learning_rate": 1.4547158937512519e-05, "loss": 2.0456, "step": 14981 }, { "epoch": 0.37, "learning_rate": 1.4546447900145398e-05, "loss": 2.1026, "step": 14982 }, { "epoch": 0.37, "learning_rate": 1.454573683380228e-05, "loss": 2.1099, "step": 14983 }, { "epoch": 0.37, "learning_rate": 1.4545025738487702e-05, "loss": 1.9224, "step": 14984 }, { "epoch": 0.37, "learning_rate": 1.4544314614206195e-05, "loss": 2.1418, "step": 14985 }, { "epoch": 0.37, "learning_rate": 1.4543603460962286e-05, "loss": 2.1502, "step": 14986 }, { "epoch": 0.37, "learning_rate": 1.4542892278760512e-05, "loss": 2.1114, "step": 14987 }, { "epoch": 0.37, "learning_rate": 1.4542181067605405e-05, "loss": 2.0771, "step": 14988 }, { "epoch": 0.37, "learning_rate": 1.45414698275015e-05, "loss": 2.2009, "step": 14989 }, { "epoch": 0.37, "learning_rate": 1.4540758558453327e-05, "loss": 2.1622, "step": 14990 }, { "epoch": 0.37, "learning_rate": 1.454004726046542e-05, "loss": 2.174, "step": 14991 }, { "epoch": 0.37, "learning_rate": 1.4539335933542312e-05, "loss": 2.2632, "step": 14992 }, { "epoch": 0.37, "learning_rate": 1.4538624577688536e-05, "loss": 2.0893, "step": 14993 }, { "epoch": 0.37, "learning_rate": 1.4537913192908629e-05, "loss": 2.019, "step": 14994 }, { "epoch": 0.37, "learning_rate": 1.453720177920712e-05, "loss": 2.11, "step": 14995 }, { "epoch": 0.37, "learning_rate": 1.4536490336588545e-05, "loss": 2.0715, "step": 14996 }, { "epoch": 0.37, "learning_rate": 1.453577886505744e-05, "loss": 2.2934, "step": 14997 }, { "epoch": 0.37, "learning_rate": 1.4535067364618338e-05, "loss": 2.2707, "step": 14998 }, { "epoch": 0.37, "learning_rate": 1.4534355835275774e-05, "loss": 1.9347, "step": 14999 }, { "epoch": 0.37, "learning_rate": 1.4533644277034282e-05, "loss": 1.9939, "step": 15000 }, { "epoch": 0.37, "learning_rate": 1.4532932689898398e-05, "loss": 2.1628, "step": 15001 }, { "epoch": 0.37, "learning_rate": 1.4532221073872656e-05, "loss": 2.0876, "step": 15002 }, { "epoch": 0.37, "learning_rate": 1.4531509428961592e-05, "loss": 2.0941, "step": 15003 }, { "epoch": 0.37, "learning_rate": 1.453079775516974e-05, "loss": 2.0969, "step": 15004 }, { "epoch": 0.37, "learning_rate": 1.4530086052501642e-05, "loss": 2.2087, "step": 15005 }, { "epoch": 0.37, "learning_rate": 1.4529374320961825e-05, "loss": 2.1687, "step": 15006 }, { "epoch": 0.37, "learning_rate": 1.4528662560554832e-05, "loss": 2.0601, "step": 15007 }, { "epoch": 0.37, "learning_rate": 1.4527950771285195e-05, "loss": 2.0376, "step": 15008 }, { "epoch": 0.37, "learning_rate": 1.4527238953157454e-05, "loss": 2.0925, "step": 15009 }, { "epoch": 0.37, "learning_rate": 1.4526527106176142e-05, "loss": 2.2422, "step": 15010 }, { "epoch": 0.37, "learning_rate": 1.4525815230345799e-05, "loss": 2.1341, "step": 15011 }, { "epoch": 0.37, "learning_rate": 1.4525103325670959e-05, "loss": 1.9785, "step": 15012 }, { "epoch": 0.37, "learning_rate": 1.4524391392156161e-05, "loss": 2.1273, "step": 15013 }, { "epoch": 0.37, "learning_rate": 1.4523679429805944e-05, "loss": 2.2026, "step": 15014 }, { "epoch": 0.37, "learning_rate": 1.4522967438624842e-05, "loss": 2.0775, "step": 15015 }, { "epoch": 0.37, "learning_rate": 1.4522255418617394e-05, "loss": 1.9479, "step": 15016 }, { "epoch": 0.37, "learning_rate": 1.452154336978814e-05, "loss": 2.2395, "step": 15017 }, { "epoch": 0.37, "learning_rate": 1.4520831292141618e-05, "loss": 1.9687, "step": 15018 }, { "epoch": 0.37, "learning_rate": 1.4520119185682364e-05, "loss": 2.1333, "step": 15019 }, { "epoch": 0.37, "learning_rate": 1.4519407050414916e-05, "loss": 2.0093, "step": 15020 }, { "epoch": 0.37, "learning_rate": 1.4518694886343813e-05, "loss": 2.1093, "step": 15021 }, { "epoch": 0.37, "learning_rate": 1.4517982693473599e-05, "loss": 2.0325, "step": 15022 }, { "epoch": 0.37, "learning_rate": 1.4517270471808804e-05, "loss": 2.0091, "step": 15023 }, { "epoch": 0.37, "learning_rate": 1.4516558221353976e-05, "loss": 2.0641, "step": 15024 }, { "epoch": 0.37, "learning_rate": 1.4515845942113646e-05, "loss": 2.1057, "step": 15025 }, { "epoch": 0.37, "learning_rate": 1.4515133634092362e-05, "loss": 2.0726, "step": 15026 }, { "epoch": 0.37, "learning_rate": 1.4514421297294662e-05, "loss": 2.0573, "step": 15027 }, { "epoch": 0.37, "learning_rate": 1.451370893172508e-05, "loss": 2.0209, "step": 15028 }, { "epoch": 0.37, "learning_rate": 1.4512996537388162e-05, "loss": 2.1518, "step": 15029 }, { "epoch": 0.37, "learning_rate": 1.4512284114288444e-05, "loss": 2.0555, "step": 15030 }, { "epoch": 0.37, "learning_rate": 1.451157166243047e-05, "loss": 2.1839, "step": 15031 }, { "epoch": 0.37, "learning_rate": 1.4510859181818778e-05, "loss": 1.9413, "step": 15032 }, { "epoch": 0.37, "learning_rate": 1.4510146672457913e-05, "loss": 1.9663, "step": 15033 }, { "epoch": 0.37, "learning_rate": 1.4509434134352414e-05, "loss": 2.0757, "step": 15034 }, { "epoch": 0.37, "learning_rate": 1.4508721567506818e-05, "loss": 1.9895, "step": 15035 }, { "epoch": 0.37, "learning_rate": 1.4508008971925671e-05, "loss": 2.0137, "step": 15036 }, { "epoch": 0.37, "learning_rate": 1.4507296347613517e-05, "loss": 2.1497, "step": 15037 }, { "epoch": 0.37, "learning_rate": 1.4506583694574894e-05, "loss": 2.1973, "step": 15038 }, { "epoch": 0.37, "learning_rate": 1.4505871012814341e-05, "loss": 2.2564, "step": 15039 }, { "epoch": 0.37, "learning_rate": 1.4505158302336406e-05, "loss": 2.014, "step": 15040 }, { "epoch": 0.37, "learning_rate": 1.4504445563145626e-05, "loss": 2.0412, "step": 15041 }, { "epoch": 0.37, "learning_rate": 1.450373279524655e-05, "loss": 2.13, "step": 15042 }, { "epoch": 0.37, "learning_rate": 1.4503019998643717e-05, "loss": 2.195, "step": 15043 }, { "epoch": 0.37, "learning_rate": 1.4502307173341668e-05, "loss": 1.9431, "step": 15044 }, { "epoch": 0.37, "learning_rate": 1.4501594319344949e-05, "loss": 2.1176, "step": 15045 }, { "epoch": 0.37, "learning_rate": 1.4500881436658103e-05, "loss": 2.0413, "step": 15046 }, { "epoch": 0.37, "learning_rate": 1.4500168525285674e-05, "loss": 2.1708, "step": 15047 }, { "epoch": 0.37, "learning_rate": 1.4499455585232201e-05, "loss": 2.111, "step": 15048 }, { "epoch": 0.37, "learning_rate": 1.4498742616502233e-05, "loss": 2.0039, "step": 15049 }, { "epoch": 0.37, "learning_rate": 1.4498029619100311e-05, "loss": 1.8805, "step": 15050 }, { "epoch": 0.37, "learning_rate": 1.4497316593030983e-05, "loss": 1.9861, "step": 15051 }, { "epoch": 0.37, "learning_rate": 1.449660353829879e-05, "loss": 2.0671, "step": 15052 }, { "epoch": 0.37, "learning_rate": 1.4495890454908278e-05, "loss": 2.0747, "step": 15053 }, { "epoch": 0.37, "learning_rate": 1.4495177342863987e-05, "loss": 1.9908, "step": 15054 }, { "epoch": 0.37, "learning_rate": 1.449446420217047e-05, "loss": 2.1364, "step": 15055 }, { "epoch": 0.37, "learning_rate": 1.4493751032832267e-05, "loss": 2.1764, "step": 15056 }, { "epoch": 0.37, "learning_rate": 1.4493037834853923e-05, "loss": 2.0799, "step": 15057 }, { "epoch": 0.37, "learning_rate": 1.4492324608239984e-05, "loss": 2.2623, "step": 15058 }, { "epoch": 0.37, "learning_rate": 1.4491611352994996e-05, "loss": 2.031, "step": 15059 }, { "epoch": 0.37, "learning_rate": 1.4490898069123506e-05, "loss": 2.0073, "step": 15060 }, { "epoch": 0.37, "learning_rate": 1.4490184756630061e-05, "loss": 2.1316, "step": 15061 }, { "epoch": 0.37, "learning_rate": 1.4489471415519205e-05, "loss": 2.0829, "step": 15062 }, { "epoch": 0.37, "learning_rate": 1.4488758045795486e-05, "loss": 2.0448, "step": 15063 }, { "epoch": 0.37, "learning_rate": 1.4488044647463445e-05, "loss": 2.2167, "step": 15064 }, { "epoch": 0.37, "learning_rate": 1.4487331220527634e-05, "loss": 2.0761, "step": 15065 }, { "epoch": 0.37, "learning_rate": 1.4486617764992596e-05, "loss": 2.0542, "step": 15066 }, { "epoch": 0.37, "learning_rate": 1.4485904280862885e-05, "loss": 2.0175, "step": 15067 }, { "epoch": 0.37, "learning_rate": 1.4485190768143045e-05, "loss": 2.0553, "step": 15068 }, { "epoch": 0.37, "learning_rate": 1.448447722683762e-05, "loss": 2.1008, "step": 15069 }, { "epoch": 0.37, "learning_rate": 1.4483763656951162e-05, "loss": 2.1923, "step": 15070 }, { "epoch": 0.37, "learning_rate": 1.4483050058488215e-05, "loss": 1.9069, "step": 15071 }, { "epoch": 0.37, "learning_rate": 1.448233643145333e-05, "loss": 2.0955, "step": 15072 }, { "epoch": 0.37, "learning_rate": 1.4481622775851053e-05, "loss": 2.1225, "step": 15073 }, { "epoch": 0.37, "learning_rate": 1.4480909091685935e-05, "loss": 2.0862, "step": 15074 }, { "epoch": 0.37, "learning_rate": 1.448019537896252e-05, "loss": 1.9833, "step": 15075 }, { "epoch": 0.37, "learning_rate": 1.4479481637685363e-05, "loss": 1.8965, "step": 15076 }, { "epoch": 0.37, "learning_rate": 1.4478767867859011e-05, "loss": 2.1601, "step": 15077 }, { "epoch": 0.37, "learning_rate": 1.4478054069488008e-05, "loss": 2.1063, "step": 15078 }, { "epoch": 0.37, "learning_rate": 1.4477340242576911e-05, "loss": 2.0336, "step": 15079 }, { "epoch": 0.37, "learning_rate": 1.4476626387130263e-05, "loss": 2.1954, "step": 15080 }, { "epoch": 0.37, "learning_rate": 1.4475912503152618e-05, "loss": 2.0228, "step": 15081 }, { "epoch": 0.37, "learning_rate": 1.4475198590648523e-05, "loss": 2.1513, "step": 15082 }, { "epoch": 0.37, "learning_rate": 1.4474484649622527e-05, "loss": 2.0962, "step": 15083 }, { "epoch": 0.37, "learning_rate": 1.4473770680079182e-05, "loss": 2.0499, "step": 15084 }, { "epoch": 0.37, "learning_rate": 1.4473056682023042e-05, "loss": 2.1055, "step": 15085 }, { "epoch": 0.37, "learning_rate": 1.4472342655458654e-05, "loss": 2.045, "step": 15086 }, { "epoch": 0.37, "learning_rate": 1.4471628600390565e-05, "loss": 2.0634, "step": 15087 }, { "epoch": 0.37, "learning_rate": 1.4470914516823332e-05, "loss": 2.1943, "step": 15088 }, { "epoch": 0.37, "learning_rate": 1.4470200404761503e-05, "loss": 2.0709, "step": 15089 }, { "epoch": 0.37, "learning_rate": 1.4469486264209632e-05, "loss": 1.9441, "step": 15090 }, { "epoch": 0.37, "learning_rate": 1.4468772095172267e-05, "loss": 2.074, "step": 15091 }, { "epoch": 0.37, "learning_rate": 1.4468057897653959e-05, "loss": 2.1641, "step": 15092 }, { "epoch": 0.37, "learning_rate": 1.4467343671659264e-05, "loss": 2.0255, "step": 15093 }, { "epoch": 0.37, "learning_rate": 1.4466629417192733e-05, "loss": 2.1408, "step": 15094 }, { "epoch": 0.37, "learning_rate": 1.4465915134258914e-05, "loss": 2.2213, "step": 15095 }, { "epoch": 0.37, "learning_rate": 1.4465200822862366e-05, "loss": 1.844, "step": 15096 }, { "epoch": 0.37, "learning_rate": 1.4464486483007636e-05, "loss": 2.0239, "step": 15097 }, { "epoch": 0.37, "learning_rate": 1.446377211469928e-05, "loss": 2.0151, "step": 15098 }, { "epoch": 0.37, "learning_rate": 1.446305771794185e-05, "loss": 2.128, "step": 15099 }, { "epoch": 0.37, "learning_rate": 1.4462343292739895e-05, "loss": 1.9945, "step": 15100 }, { "epoch": 0.37, "learning_rate": 1.4461628839097974e-05, "loss": 2.2025, "step": 15101 }, { "epoch": 0.37, "learning_rate": 1.4460914357020637e-05, "loss": 2.1551, "step": 15102 }, { "epoch": 0.37, "learning_rate": 1.4460199846512442e-05, "loss": 1.9992, "step": 15103 }, { "epoch": 0.37, "learning_rate": 1.4459485307577938e-05, "loss": 2.0882, "step": 15104 }, { "epoch": 0.37, "learning_rate": 1.4458770740221683e-05, "loss": 1.9398, "step": 15105 }, { "epoch": 0.37, "learning_rate": 1.4458056144448228e-05, "loss": 2.1421, "step": 15106 }, { "epoch": 0.37, "learning_rate": 1.4457341520262127e-05, "loss": 2.2272, "step": 15107 }, { "epoch": 0.37, "learning_rate": 1.4456626867667938e-05, "loss": 1.9375, "step": 15108 }, { "epoch": 0.37, "learning_rate": 1.4455912186670211e-05, "loss": 2.29, "step": 15109 }, { "epoch": 0.37, "learning_rate": 1.4455197477273505e-05, "loss": 2.0454, "step": 15110 }, { "epoch": 0.37, "learning_rate": 1.4454482739482375e-05, "loss": 2.1746, "step": 15111 }, { "epoch": 0.37, "learning_rate": 1.4453767973301372e-05, "loss": 2.2436, "step": 15112 }, { "epoch": 0.37, "learning_rate": 1.4453053178735055e-05, "loss": 1.961, "step": 15113 }, { "epoch": 0.37, "learning_rate": 1.4452338355787978e-05, "loss": 2.2219, "step": 15114 }, { "epoch": 0.37, "learning_rate": 1.44516235044647e-05, "loss": 2.1152, "step": 15115 }, { "epoch": 0.37, "learning_rate": 1.4450908624769775e-05, "loss": 1.8569, "step": 15116 }, { "epoch": 0.37, "learning_rate": 1.4450193716707757e-05, "loss": 2.0431, "step": 15117 }, { "epoch": 0.37, "learning_rate": 1.4449478780283203e-05, "loss": 2.1492, "step": 15118 }, { "epoch": 0.37, "learning_rate": 1.4448763815500672e-05, "loss": 2.0728, "step": 15119 }, { "epoch": 0.37, "learning_rate": 1.444804882236472e-05, "loss": 2.3547, "step": 15120 }, { "epoch": 0.37, "learning_rate": 1.4447333800879904e-05, "loss": 2.069, "step": 15121 }, { "epoch": 0.37, "learning_rate": 1.4446618751050778e-05, "loss": 2.0105, "step": 15122 }, { "epoch": 0.37, "learning_rate": 1.4445903672881904e-05, "loss": 2.0348, "step": 15123 }, { "epoch": 0.37, "learning_rate": 1.4445188566377837e-05, "loss": 2.1522, "step": 15124 }, { "epoch": 0.37, "learning_rate": 1.4444473431543134e-05, "loss": 2.1421, "step": 15125 }, { "epoch": 0.37, "learning_rate": 1.4443758268382353e-05, "loss": 1.9467, "step": 15126 }, { "epoch": 0.37, "learning_rate": 1.4443043076900053e-05, "loss": 2.0724, "step": 15127 }, { "epoch": 0.37, "learning_rate": 1.4442327857100789e-05, "loss": 2.2276, "step": 15128 }, { "epoch": 0.37, "learning_rate": 1.4441612608989123e-05, "loss": 2.042, "step": 15129 }, { "epoch": 0.37, "learning_rate": 1.4440897332569614e-05, "loss": 2.0192, "step": 15130 }, { "epoch": 0.37, "learning_rate": 1.4440182027846816e-05, "loss": 2.0682, "step": 15131 }, { "epoch": 0.37, "learning_rate": 1.4439466694825296e-05, "loss": 2.0286, "step": 15132 }, { "epoch": 0.37, "learning_rate": 1.4438751333509605e-05, "loss": 1.9311, "step": 15133 }, { "epoch": 0.37, "learning_rate": 1.4438035943904304e-05, "loss": 1.9664, "step": 15134 }, { "epoch": 0.37, "learning_rate": 1.4437320526013957e-05, "loss": 2.131, "step": 15135 }, { "epoch": 0.37, "learning_rate": 1.4436605079843115e-05, "loss": 2.0256, "step": 15136 }, { "epoch": 0.37, "learning_rate": 1.4435889605396347e-05, "loss": 2.098, "step": 15137 }, { "epoch": 0.37, "learning_rate": 1.4435174102678207e-05, "loss": 2.0678, "step": 15138 }, { "epoch": 0.37, "learning_rate": 1.443445857169326e-05, "loss": 2.1077, "step": 15139 }, { "epoch": 0.37, "learning_rate": 1.4433743012446061e-05, "loss": 2.2532, "step": 15140 }, { "epoch": 0.37, "learning_rate": 1.4433027424941172e-05, "loss": 2.1417, "step": 15141 }, { "epoch": 0.37, "learning_rate": 1.4432311809183157e-05, "loss": 1.9598, "step": 15142 }, { "epoch": 0.37, "learning_rate": 1.4431596165176573e-05, "loss": 2.1501, "step": 15143 }, { "epoch": 0.37, "learning_rate": 1.4430880492925985e-05, "loss": 2.0455, "step": 15144 }, { "epoch": 0.37, "learning_rate": 1.4430164792435947e-05, "loss": 2.2049, "step": 15145 }, { "epoch": 0.37, "learning_rate": 1.4429449063711025e-05, "loss": 2.0788, "step": 15146 }, { "epoch": 0.37, "learning_rate": 1.4428733306755784e-05, "loss": 1.9554, "step": 15147 }, { "epoch": 0.37, "learning_rate": 1.4428017521574782e-05, "loss": 2.1497, "step": 15148 }, { "epoch": 0.37, "learning_rate": 1.4427301708172583e-05, "loss": 2.0045, "step": 15149 }, { "epoch": 0.37, "learning_rate": 1.4426585866553744e-05, "loss": 2.1492, "step": 15150 }, { "epoch": 0.37, "learning_rate": 1.442586999672283e-05, "loss": 2.1205, "step": 15151 }, { "epoch": 0.37, "learning_rate": 1.4425154098684408e-05, "loss": 1.9812, "step": 15152 }, { "epoch": 0.37, "learning_rate": 1.4424438172443034e-05, "loss": 2.2153, "step": 15153 }, { "epoch": 0.37, "learning_rate": 1.4423722218003272e-05, "loss": 2.1975, "step": 15154 }, { "epoch": 0.37, "learning_rate": 1.4423006235369692e-05, "loss": 2.0317, "step": 15155 }, { "epoch": 0.37, "learning_rate": 1.4422290224546847e-05, "loss": 2.0646, "step": 15156 }, { "epoch": 0.37, "learning_rate": 1.4421574185539306e-05, "loss": 2.0655, "step": 15157 }, { "epoch": 0.37, "learning_rate": 1.4420858118351636e-05, "loss": 2.1792, "step": 15158 }, { "epoch": 0.37, "learning_rate": 1.4420142022988392e-05, "loss": 2.0687, "step": 15159 }, { "epoch": 0.37, "learning_rate": 1.4419425899454144e-05, "loss": 2.0466, "step": 15160 }, { "epoch": 0.37, "learning_rate": 1.4418709747753451e-05, "loss": 2.199, "step": 15161 }, { "epoch": 0.37, "learning_rate": 1.4417993567890887e-05, "loss": 2.2177, "step": 15162 }, { "epoch": 0.37, "learning_rate": 1.4417277359871005e-05, "loss": 2.238, "step": 15163 }, { "epoch": 0.37, "learning_rate": 1.4416561123698378e-05, "loss": 2.0779, "step": 15164 }, { "epoch": 0.37, "learning_rate": 1.4415844859377565e-05, "loss": 2.2447, "step": 15165 }, { "epoch": 0.37, "learning_rate": 1.4415128566913136e-05, "loss": 2.1323, "step": 15166 }, { "epoch": 0.37, "learning_rate": 1.4414412246309653e-05, "loss": 2.2815, "step": 15167 }, { "epoch": 0.37, "learning_rate": 1.4413695897571685e-05, "loss": 2.0811, "step": 15168 }, { "epoch": 0.37, "learning_rate": 1.4412979520703792e-05, "loss": 2.0211, "step": 15169 }, { "epoch": 0.37, "learning_rate": 1.4412263115710538e-05, "loss": 2.2475, "step": 15170 }, { "epoch": 0.37, "learning_rate": 1.44115466825965e-05, "loss": 1.8778, "step": 15171 }, { "epoch": 0.37, "learning_rate": 1.4410830221366233e-05, "loss": 2.2551, "step": 15172 }, { "epoch": 0.37, "learning_rate": 1.441011373202431e-05, "loss": 2.1064, "step": 15173 }, { "epoch": 0.37, "learning_rate": 1.4409397214575295e-05, "loss": 2.1075, "step": 15174 }, { "epoch": 0.37, "learning_rate": 1.4408680669023754e-05, "loss": 1.9092, "step": 15175 }, { "epoch": 0.37, "learning_rate": 1.4407964095374255e-05, "loss": 2.0385, "step": 15176 }, { "epoch": 0.37, "learning_rate": 1.4407247493631362e-05, "loss": 2.2225, "step": 15177 }, { "epoch": 0.37, "learning_rate": 1.4406530863799648e-05, "loss": 2.0253, "step": 15178 }, { "epoch": 0.37, "learning_rate": 1.4405814205883676e-05, "loss": 2.1428, "step": 15179 }, { "epoch": 0.37, "learning_rate": 1.440509751988801e-05, "loss": 2.255, "step": 15180 }, { "epoch": 0.37, "learning_rate": 1.4404380805817224e-05, "loss": 2.0986, "step": 15181 }, { "epoch": 0.37, "learning_rate": 1.4403664063675887e-05, "loss": 1.9795, "step": 15182 }, { "epoch": 0.37, "learning_rate": 1.4402947293468562e-05, "loss": 2.0665, "step": 15183 }, { "epoch": 0.37, "learning_rate": 1.4402230495199816e-05, "loss": 1.9866, "step": 15184 }, { "epoch": 0.37, "learning_rate": 1.4401513668874224e-05, "loss": 2.0952, "step": 15185 }, { "epoch": 0.37, "learning_rate": 1.440079681449635e-05, "loss": 1.8758, "step": 15186 }, { "epoch": 0.37, "learning_rate": 1.4400079932070763e-05, "loss": 2.1562, "step": 15187 }, { "epoch": 0.37, "learning_rate": 1.4399363021602036e-05, "loss": 2.0408, "step": 15188 }, { "epoch": 0.37, "learning_rate": 1.4398646083094729e-05, "loss": 2.0536, "step": 15189 }, { "epoch": 0.37, "learning_rate": 1.4397929116553417e-05, "loss": 2.2333, "step": 15190 }, { "epoch": 0.37, "learning_rate": 1.4397212121982676e-05, "loss": 2.0714, "step": 15191 }, { "epoch": 0.37, "learning_rate": 1.4396495099387065e-05, "loss": 2.2682, "step": 15192 }, { "epoch": 0.37, "learning_rate": 1.4395778048771158e-05, "loss": 2.3394, "step": 15193 }, { "epoch": 0.37, "learning_rate": 1.4395060970139524e-05, "loss": 2.1602, "step": 15194 }, { "epoch": 0.37, "learning_rate": 1.4394343863496738e-05, "loss": 2.0705, "step": 15195 }, { "epoch": 0.37, "learning_rate": 1.4393626728847365e-05, "loss": 2.0957, "step": 15196 }, { "epoch": 0.37, "learning_rate": 1.4392909566195976e-05, "loss": 2.1275, "step": 15197 }, { "epoch": 0.37, "learning_rate": 1.4392192375547143e-05, "loss": 2.0149, "step": 15198 }, { "epoch": 0.37, "learning_rate": 1.4391475156905434e-05, "loss": 1.9745, "step": 15199 }, { "epoch": 0.37, "learning_rate": 1.4390757910275426e-05, "loss": 2.0284, "step": 15200 }, { "epoch": 0.37, "learning_rate": 1.4390040635661686e-05, "loss": 2.1546, "step": 15201 }, { "epoch": 0.37, "learning_rate": 1.4389323333068787e-05, "loss": 2.1004, "step": 15202 }, { "epoch": 0.37, "learning_rate": 1.4388606002501301e-05, "loss": 2.0297, "step": 15203 }, { "epoch": 0.37, "learning_rate": 1.4387888643963797e-05, "loss": 2.0691, "step": 15204 }, { "epoch": 0.37, "learning_rate": 1.4387171257460847e-05, "loss": 1.9239, "step": 15205 }, { "epoch": 0.37, "learning_rate": 1.4386453842997028e-05, "loss": 2.1612, "step": 15206 }, { "epoch": 0.37, "learning_rate": 1.4385736400576906e-05, "loss": 2.0045, "step": 15207 }, { "epoch": 0.37, "learning_rate": 1.4385018930205062e-05, "loss": 2.0308, "step": 15208 }, { "epoch": 0.37, "learning_rate": 1.4384301431886057e-05, "loss": 2.0186, "step": 15209 }, { "epoch": 0.37, "learning_rate": 1.4383583905624474e-05, "loss": 2.0322, "step": 15210 }, { "epoch": 0.37, "learning_rate": 1.4382866351424882e-05, "loss": 2.0622, "step": 15211 }, { "epoch": 0.37, "learning_rate": 1.4382148769291852e-05, "loss": 2.2784, "step": 15212 }, { "epoch": 0.37, "learning_rate": 1.4381431159229964e-05, "loss": 1.9109, "step": 15213 }, { "epoch": 0.38, "learning_rate": 1.4380713521243782e-05, "loss": 2.0627, "step": 15214 }, { "epoch": 0.38, "learning_rate": 1.437999585533789e-05, "loss": 2.0757, "step": 15215 }, { "epoch": 0.38, "learning_rate": 1.4379278161516851e-05, "loss": 2.1526, "step": 15216 }, { "epoch": 0.38, "learning_rate": 1.4378560439785251e-05, "loss": 2.0461, "step": 15217 }, { "epoch": 0.38, "learning_rate": 1.437784269014766e-05, "loss": 2.1685, "step": 15218 }, { "epoch": 0.38, "learning_rate": 1.4377124912608645e-05, "loss": 2.0462, "step": 15219 }, { "epoch": 0.38, "learning_rate": 1.4376407107172788e-05, "loss": 2.0356, "step": 15220 }, { "epoch": 0.38, "learning_rate": 1.4375689273844664e-05, "loss": 2.1083, "step": 15221 }, { "epoch": 0.38, "learning_rate": 1.4374971412628847e-05, "loss": 2.2145, "step": 15222 }, { "epoch": 0.38, "learning_rate": 1.4374253523529906e-05, "loss": 1.9817, "step": 15223 }, { "epoch": 0.38, "learning_rate": 1.4373535606552428e-05, "loss": 2.0798, "step": 15224 }, { "epoch": 0.38, "learning_rate": 1.4372817661700977e-05, "loss": 2.1688, "step": 15225 }, { "epoch": 0.38, "learning_rate": 1.4372099688980136e-05, "loss": 2.1096, "step": 15226 }, { "epoch": 0.38, "learning_rate": 1.4371381688394482e-05, "loss": 2.0181, "step": 15227 }, { "epoch": 0.38, "learning_rate": 1.4370663659948582e-05, "loss": 2.0737, "step": 15228 }, { "epoch": 0.38, "learning_rate": 1.4369945603647025e-05, "loss": 1.9169, "step": 15229 }, { "epoch": 0.38, "learning_rate": 1.4369227519494378e-05, "loss": 2.1651, "step": 15230 }, { "epoch": 0.38, "learning_rate": 1.436850940749522e-05, "loss": 2.0566, "step": 15231 }, { "epoch": 0.38, "learning_rate": 1.4367791267654126e-05, "loss": 2.1995, "step": 15232 }, { "epoch": 0.38, "learning_rate": 1.4367073099975674e-05, "loss": 2.1504, "step": 15233 }, { "epoch": 0.38, "learning_rate": 1.4366354904464444e-05, "loss": 1.9713, "step": 15234 }, { "epoch": 0.38, "learning_rate": 1.436563668112501e-05, "loss": 2.0425, "step": 15235 }, { "epoch": 0.38, "learning_rate": 1.4364918429961953e-05, "loss": 2.1717, "step": 15236 }, { "epoch": 0.38, "learning_rate": 1.4364200150979847e-05, "loss": 2.2204, "step": 15237 }, { "epoch": 0.38, "learning_rate": 1.4363481844183271e-05, "loss": 2.0533, "step": 15238 }, { "epoch": 0.38, "learning_rate": 1.4362763509576804e-05, "loss": 2.091, "step": 15239 }, { "epoch": 0.38, "learning_rate": 1.4362045147165024e-05, "loss": 2.0667, "step": 15240 }, { "epoch": 0.38, "learning_rate": 1.436132675695251e-05, "loss": 2.3322, "step": 15241 }, { "epoch": 0.38, "learning_rate": 1.4360608338943835e-05, "loss": 2.1646, "step": 15242 }, { "epoch": 0.38, "learning_rate": 1.4359889893143582e-05, "loss": 2.0144, "step": 15243 }, { "epoch": 0.38, "learning_rate": 1.4359171419556332e-05, "loss": 2.0973, "step": 15244 }, { "epoch": 0.38, "learning_rate": 1.4358452918186663e-05, "loss": 1.9092, "step": 15245 }, { "epoch": 0.38, "learning_rate": 1.4357734389039152e-05, "loss": 2.0405, "step": 15246 }, { "epoch": 0.38, "learning_rate": 1.4357015832118382e-05, "loss": 2.0643, "step": 15247 }, { "epoch": 0.38, "learning_rate": 1.4356297247428926e-05, "loss": 2.1467, "step": 15248 }, { "epoch": 0.38, "learning_rate": 1.4355578634975369e-05, "loss": 2.0732, "step": 15249 }, { "epoch": 0.38, "learning_rate": 1.4354859994762291e-05, "loss": 2.1456, "step": 15250 }, { "epoch": 0.38, "learning_rate": 1.435414132679427e-05, "loss": 2.2413, "step": 15251 }, { "epoch": 0.38, "learning_rate": 1.4353422631075886e-05, "loss": 2.231, "step": 15252 }, { "epoch": 0.38, "learning_rate": 1.4352703907611724e-05, "loss": 2.0021, "step": 15253 }, { "epoch": 0.38, "learning_rate": 1.4351985156406361e-05, "loss": 1.9338, "step": 15254 }, { "epoch": 0.38, "learning_rate": 1.435126637746438e-05, "loss": 2.0004, "step": 15255 }, { "epoch": 0.38, "learning_rate": 1.4350547570790356e-05, "loss": 2.0386, "step": 15256 }, { "epoch": 0.38, "learning_rate": 1.4349828736388875e-05, "loss": 1.975, "step": 15257 }, { "epoch": 0.38, "learning_rate": 1.4349109874264522e-05, "loss": 2.1129, "step": 15258 }, { "epoch": 0.38, "learning_rate": 1.4348390984421872e-05, "loss": 2.1674, "step": 15259 }, { "epoch": 0.38, "learning_rate": 1.4347672066865508e-05, "loss": 2.1692, "step": 15260 }, { "epoch": 0.38, "learning_rate": 1.4346953121600014e-05, "loss": 1.9471, "step": 15261 }, { "epoch": 0.38, "learning_rate": 1.4346234148629969e-05, "loss": 1.9915, "step": 15262 }, { "epoch": 0.38, "learning_rate": 1.434551514795996e-05, "loss": 2.095, "step": 15263 }, { "epoch": 0.38, "learning_rate": 1.4344796119594568e-05, "loss": 2.1125, "step": 15264 }, { "epoch": 0.38, "learning_rate": 1.4344077063538373e-05, "loss": 2.032, "step": 15265 }, { "epoch": 0.38, "learning_rate": 1.434335797979596e-05, "loss": 2.1775, "step": 15266 }, { "epoch": 0.38, "learning_rate": 1.4342638868371909e-05, "loss": 2.1614, "step": 15267 }, { "epoch": 0.38, "learning_rate": 1.4341919729270806e-05, "loss": 2.0646, "step": 15268 }, { "epoch": 0.38, "learning_rate": 1.4341200562497232e-05, "loss": 2.2543, "step": 15269 }, { "epoch": 0.38, "learning_rate": 1.4340481368055776e-05, "loss": 2.2216, "step": 15270 }, { "epoch": 0.38, "learning_rate": 1.4339762145951018e-05, "loss": 2.0769, "step": 15271 }, { "epoch": 0.38, "learning_rate": 1.4339042896187536e-05, "loss": 2.0906, "step": 15272 }, { "epoch": 0.38, "learning_rate": 1.4338323618769922e-05, "loss": 2.0248, "step": 15273 }, { "epoch": 0.38, "learning_rate": 1.4337604313702759e-05, "loss": 1.9448, "step": 15274 }, { "epoch": 0.38, "learning_rate": 1.4336884980990631e-05, "loss": 1.948, "step": 15275 }, { "epoch": 0.38, "learning_rate": 1.4336165620638121e-05, "loss": 2.0994, "step": 15276 }, { "epoch": 0.38, "learning_rate": 1.4335446232649814e-05, "loss": 2.1719, "step": 15277 }, { "epoch": 0.38, "learning_rate": 1.4334726817030293e-05, "loss": 1.9826, "step": 15278 }, { "epoch": 0.38, "learning_rate": 1.4334007373784147e-05, "loss": 2.1405, "step": 15279 }, { "epoch": 0.38, "learning_rate": 1.4333287902915963e-05, "loss": 2.1827, "step": 15280 }, { "epoch": 0.38, "learning_rate": 1.4332568404430319e-05, "loss": 2.0881, "step": 15281 }, { "epoch": 0.38, "learning_rate": 1.4331848878331804e-05, "loss": 1.9848, "step": 15282 }, { "epoch": 0.38, "learning_rate": 1.4331129324625006e-05, "loss": 2.0392, "step": 15283 }, { "epoch": 0.38, "learning_rate": 1.433040974331451e-05, "loss": 1.9908, "step": 15284 }, { "epoch": 0.38, "learning_rate": 1.4329690134404901e-05, "loss": 2.08, "step": 15285 }, { "epoch": 0.38, "learning_rate": 1.4328970497900765e-05, "loss": 2.1143, "step": 15286 }, { "epoch": 0.38, "learning_rate": 1.432825083380669e-05, "loss": 1.9938, "step": 15287 }, { "epoch": 0.38, "learning_rate": 1.432753114212726e-05, "loss": 1.9556, "step": 15288 }, { "epoch": 0.38, "learning_rate": 1.4326811422867067e-05, "loss": 2.0082, "step": 15289 }, { "epoch": 0.38, "learning_rate": 1.4326091676030694e-05, "loss": 2.1852, "step": 15290 }, { "epoch": 0.38, "learning_rate": 1.4325371901622727e-05, "loss": 2.0015, "step": 15291 }, { "epoch": 0.38, "learning_rate": 1.4324652099647755e-05, "loss": 2.1138, "step": 15292 }, { "epoch": 0.38, "learning_rate": 1.432393227011037e-05, "loss": 1.943, "step": 15293 }, { "epoch": 0.38, "learning_rate": 1.432321241301515e-05, "loss": 2.2255, "step": 15294 }, { "epoch": 0.38, "learning_rate": 1.4322492528366691e-05, "loss": 2.0056, "step": 15295 }, { "epoch": 0.38, "learning_rate": 1.4321772616169578e-05, "loss": 2.1186, "step": 15296 }, { "epoch": 0.38, "learning_rate": 1.4321052676428402e-05, "loss": 2.221, "step": 15297 }, { "epoch": 0.38, "learning_rate": 1.4320332709147749e-05, "loss": 2.2313, "step": 15298 }, { "epoch": 0.38, "learning_rate": 1.4319612714332206e-05, "loss": 2.113, "step": 15299 }, { "epoch": 0.38, "learning_rate": 1.4318892691986364e-05, "loss": 2.1567, "step": 15300 }, { "epoch": 0.38, "learning_rate": 1.431817264211481e-05, "loss": 2.2373, "step": 15301 }, { "epoch": 0.38, "learning_rate": 1.4317452564722135e-05, "loss": 2.1592, "step": 15302 }, { "epoch": 0.38, "learning_rate": 1.4316732459812927e-05, "loss": 2.142, "step": 15303 }, { "epoch": 0.38, "learning_rate": 1.4316012327391777e-05, "loss": 2.1124, "step": 15304 }, { "epoch": 0.38, "learning_rate": 1.4315292167463275e-05, "loss": 2.2034, "step": 15305 }, { "epoch": 0.38, "learning_rate": 1.4314571980032008e-05, "loss": 2.0075, "step": 15306 }, { "epoch": 0.38, "learning_rate": 1.4313851765102569e-05, "loss": 2.0359, "step": 15307 }, { "epoch": 0.38, "learning_rate": 1.4313131522679546e-05, "loss": 2.1745, "step": 15308 }, { "epoch": 0.38, "learning_rate": 1.4312411252767532e-05, "loss": 2.2204, "step": 15309 }, { "epoch": 0.38, "learning_rate": 1.4311690955371114e-05, "loss": 2.1696, "step": 15310 }, { "epoch": 0.38, "learning_rate": 1.4310970630494885e-05, "loss": 1.9874, "step": 15311 }, { "epoch": 0.38, "learning_rate": 1.4310250278143433e-05, "loss": 1.955, "step": 15312 }, { "epoch": 0.38, "learning_rate": 1.4309529898321355e-05, "loss": 2.149, "step": 15313 }, { "epoch": 0.38, "learning_rate": 1.4308809491033238e-05, "loss": 2.2441, "step": 15314 }, { "epoch": 0.38, "learning_rate": 1.4308089056283672e-05, "loss": 2.0796, "step": 15315 }, { "epoch": 0.38, "learning_rate": 1.4307368594077251e-05, "loss": 2.1469, "step": 15316 }, { "epoch": 0.38, "learning_rate": 1.4306648104418568e-05, "loss": 2.1255, "step": 15317 }, { "epoch": 0.38, "learning_rate": 1.4305927587312212e-05, "loss": 2.1316, "step": 15318 }, { "epoch": 0.38, "learning_rate": 1.4305207042762776e-05, "loss": 2.1061, "step": 15319 }, { "epoch": 0.38, "learning_rate": 1.4304486470774852e-05, "loss": 2.0375, "step": 15320 }, { "epoch": 0.38, "learning_rate": 1.4303765871353033e-05, "loss": 2.0695, "step": 15321 }, { "epoch": 0.38, "learning_rate": 1.430304524450191e-05, "loss": 1.866, "step": 15322 }, { "epoch": 0.38, "learning_rate": 1.430232459022608e-05, "loss": 2.0262, "step": 15323 }, { "epoch": 0.38, "learning_rate": 1.4301603908530135e-05, "loss": 2.2595, "step": 15324 }, { "epoch": 0.38, "learning_rate": 1.430088319941866e-05, "loss": 2.1612, "step": 15325 }, { "epoch": 0.38, "learning_rate": 1.430016246289626e-05, "loss": 1.9635, "step": 15326 }, { "epoch": 0.38, "learning_rate": 1.4299441698967522e-05, "loss": 2.1849, "step": 15327 }, { "epoch": 0.38, "learning_rate": 1.4298720907637042e-05, "loss": 2.0358, "step": 15328 }, { "epoch": 0.38, "learning_rate": 1.4298000088909412e-05, "loss": 2.1546, "step": 15329 }, { "epoch": 0.38, "learning_rate": 1.4297279242789227e-05, "loss": 2.1534, "step": 15330 }, { "epoch": 0.38, "learning_rate": 1.4296558369281077e-05, "loss": 1.9351, "step": 15331 }, { "epoch": 0.38, "learning_rate": 1.4295837468389566e-05, "loss": 2.2503, "step": 15332 }, { "epoch": 0.38, "learning_rate": 1.429511654011928e-05, "loss": 2.074, "step": 15333 }, { "epoch": 0.38, "learning_rate": 1.4294395584474819e-05, "loss": 1.9149, "step": 15334 }, { "epoch": 0.38, "learning_rate": 1.429367460146077e-05, "loss": 1.9946, "step": 15335 }, { "epoch": 0.38, "learning_rate": 1.4292953591081737e-05, "loss": 2.2078, "step": 15336 }, { "epoch": 0.38, "learning_rate": 1.4292232553342314e-05, "loss": 2.1524, "step": 15337 }, { "epoch": 0.38, "learning_rate": 1.429151148824709e-05, "loss": 2.035, "step": 15338 }, { "epoch": 0.38, "learning_rate": 1.4290790395800666e-05, "loss": 2.081, "step": 15339 }, { "epoch": 0.38, "learning_rate": 1.4290069276007637e-05, "loss": 2.005, "step": 15340 }, { "epoch": 0.38, "learning_rate": 1.4289348128872597e-05, "loss": 2.1351, "step": 15341 }, { "epoch": 0.38, "learning_rate": 1.4288626954400147e-05, "loss": 2.0466, "step": 15342 }, { "epoch": 0.38, "learning_rate": 1.428790575259488e-05, "loss": 2.0661, "step": 15343 }, { "epoch": 0.38, "learning_rate": 1.4287184523461388e-05, "loss": 2.0449, "step": 15344 }, { "epoch": 0.38, "learning_rate": 1.4286463267004273e-05, "loss": 2.0327, "step": 15345 }, { "epoch": 0.38, "learning_rate": 1.4285741983228132e-05, "loss": 2.1251, "step": 15346 }, { "epoch": 0.38, "learning_rate": 1.428502067213756e-05, "loss": 2.1274, "step": 15347 }, { "epoch": 0.38, "learning_rate": 1.4284299333737155e-05, "loss": 2.188, "step": 15348 }, { "epoch": 0.38, "learning_rate": 1.4283577968031513e-05, "loss": 2.1283, "step": 15349 }, { "epoch": 0.38, "learning_rate": 1.4282856575025235e-05, "loss": 1.9295, "step": 15350 }, { "epoch": 0.38, "learning_rate": 1.4282135154722917e-05, "loss": 2.2926, "step": 15351 }, { "epoch": 0.38, "learning_rate": 1.4281413707129153e-05, "loss": 2.0447, "step": 15352 }, { "epoch": 0.38, "learning_rate": 1.4280692232248545e-05, "loss": 2.2316, "step": 15353 }, { "epoch": 0.38, "learning_rate": 1.427997073008569e-05, "loss": 2.1047, "step": 15354 }, { "epoch": 0.38, "learning_rate": 1.427924920064519e-05, "loss": 2.162, "step": 15355 }, { "epoch": 0.38, "learning_rate": 1.4278527643931634e-05, "loss": 1.9294, "step": 15356 }, { "epoch": 0.38, "learning_rate": 1.4277806059949633e-05, "loss": 2.2026, "step": 15357 }, { "epoch": 0.38, "learning_rate": 1.4277084448703778e-05, "loss": 2.1077, "step": 15358 }, { "epoch": 0.38, "learning_rate": 1.4276362810198668e-05, "loss": 1.9452, "step": 15359 }, { "epoch": 0.38, "learning_rate": 1.4275641144438907e-05, "loss": 2.0538, "step": 15360 }, { "epoch": 0.38, "learning_rate": 1.427491945142909e-05, "loss": 2.1583, "step": 15361 }, { "epoch": 0.38, "learning_rate": 1.427419773117382e-05, "loss": 2.0215, "step": 15362 }, { "epoch": 0.38, "learning_rate": 1.4273475983677693e-05, "loss": 2.1381, "step": 15363 }, { "epoch": 0.38, "learning_rate": 1.4272754208945309e-05, "loss": 2.3476, "step": 15364 }, { "epoch": 0.38, "learning_rate": 1.4272032406981273e-05, "loss": 2.1368, "step": 15365 }, { "epoch": 0.38, "learning_rate": 1.4271310577790183e-05, "loss": 2.0975, "step": 15366 }, { "epoch": 0.38, "learning_rate": 1.4270588721376637e-05, "loss": 2.2209, "step": 15367 }, { "epoch": 0.38, "learning_rate": 1.4269866837745237e-05, "loss": 1.9573, "step": 15368 }, { "epoch": 0.38, "learning_rate": 1.4269144926900584e-05, "loss": 2.0576, "step": 15369 }, { "epoch": 0.38, "learning_rate": 1.4268422988847282e-05, "loss": 2.3273, "step": 15370 }, { "epoch": 0.38, "learning_rate": 1.4267701023589925e-05, "loss": 2.1256, "step": 15371 }, { "epoch": 0.38, "learning_rate": 1.4266979031133123e-05, "loss": 1.9873, "step": 15372 }, { "epoch": 0.38, "learning_rate": 1.4266257011481467e-05, "loss": 2.1676, "step": 15373 }, { "epoch": 0.38, "learning_rate": 1.426553496463957e-05, "loss": 2.0246, "step": 15374 }, { "epoch": 0.38, "learning_rate": 1.4264812890612027e-05, "loss": 1.9894, "step": 15375 }, { "epoch": 0.38, "learning_rate": 1.4264090789403442e-05, "loss": 1.9669, "step": 15376 }, { "epoch": 0.38, "learning_rate": 1.4263368661018417e-05, "loss": 2.1501, "step": 15377 }, { "epoch": 0.38, "learning_rate": 1.4262646505461553e-05, "loss": 2.0842, "step": 15378 }, { "epoch": 0.38, "learning_rate": 1.4261924322737455e-05, "loss": 2.1112, "step": 15379 }, { "epoch": 0.38, "learning_rate": 1.4261202112850723e-05, "loss": 2.298, "step": 15380 }, { "epoch": 0.38, "learning_rate": 1.4260479875805963e-05, "loss": 2.2333, "step": 15381 }, { "epoch": 0.38, "learning_rate": 1.4259757611607774e-05, "loss": 2.1331, "step": 15382 }, { "epoch": 0.38, "learning_rate": 1.4259035320260761e-05, "loss": 2.0693, "step": 15383 }, { "epoch": 0.38, "learning_rate": 1.425831300176953e-05, "loss": 2.1061, "step": 15384 }, { "epoch": 0.38, "learning_rate": 1.4257590656138682e-05, "loss": 2.082, "step": 15385 }, { "epoch": 0.38, "learning_rate": 1.4256868283372822e-05, "loss": 2.0836, "step": 15386 }, { "epoch": 0.38, "learning_rate": 1.4256145883476555e-05, "loss": 2.0664, "step": 15387 }, { "epoch": 0.38, "learning_rate": 1.4255423456454478e-05, "loss": 2.0973, "step": 15388 }, { "epoch": 0.38, "learning_rate": 1.4254701002311204e-05, "loss": 1.9288, "step": 15389 }, { "epoch": 0.38, "learning_rate": 1.4253978521051334e-05, "loss": 2.0116, "step": 15390 }, { "epoch": 0.38, "learning_rate": 1.425325601267947e-05, "loss": 2.0502, "step": 15391 }, { "epoch": 0.38, "learning_rate": 1.4252533477200224e-05, "loss": 2.2328, "step": 15392 }, { "epoch": 0.38, "learning_rate": 1.425181091461819e-05, "loss": 2.0856, "step": 15393 }, { "epoch": 0.38, "learning_rate": 1.4251088324937983e-05, "loss": 2.0589, "step": 15394 }, { "epoch": 0.38, "learning_rate": 1.4250365708164204e-05, "loss": 2.056, "step": 15395 }, { "epoch": 0.38, "learning_rate": 1.4249643064301462e-05, "loss": 2.1397, "step": 15396 }, { "epoch": 0.38, "learning_rate": 1.4248920393354357e-05, "loss": 2.0707, "step": 15397 }, { "epoch": 0.38, "learning_rate": 1.4248197695327499e-05, "loss": 2.1479, "step": 15398 }, { "epoch": 0.38, "learning_rate": 1.4247474970225493e-05, "loss": 2.0567, "step": 15399 }, { "epoch": 0.38, "learning_rate": 1.424675221805294e-05, "loss": 2.1171, "step": 15400 }, { "epoch": 0.38, "learning_rate": 1.4246029438814458e-05, "loss": 2.058, "step": 15401 }, { "epoch": 0.38, "learning_rate": 1.4245306632514639e-05, "loss": 2.2173, "step": 15402 }, { "epoch": 0.38, "learning_rate": 1.4244583799158102e-05, "loss": 1.9566, "step": 15403 }, { "epoch": 0.38, "learning_rate": 1.424386093874945e-05, "loss": 1.9479, "step": 15404 }, { "epoch": 0.38, "learning_rate": 1.4243138051293288e-05, "loss": 2.1284, "step": 15405 }, { "epoch": 0.38, "learning_rate": 1.4242415136794226e-05, "loss": 2.2957, "step": 15406 }, { "epoch": 0.38, "learning_rate": 1.4241692195256867e-05, "loss": 2.0101, "step": 15407 }, { "epoch": 0.38, "learning_rate": 1.4240969226685824e-05, "loss": 2.0766, "step": 15408 }, { "epoch": 0.38, "learning_rate": 1.42402462310857e-05, "loss": 2.0982, "step": 15409 }, { "epoch": 0.38, "learning_rate": 1.4239523208461106e-05, "loss": 1.9935, "step": 15410 }, { "epoch": 0.38, "learning_rate": 1.4238800158816648e-05, "loss": 2.231, "step": 15411 }, { "epoch": 0.38, "learning_rate": 1.4238077082156933e-05, "loss": 2.1062, "step": 15412 }, { "epoch": 0.38, "learning_rate": 1.4237353978486577e-05, "loss": 2.2545, "step": 15413 }, { "epoch": 0.38, "learning_rate": 1.423663084781018e-05, "loss": 2.0997, "step": 15414 }, { "epoch": 0.38, "learning_rate": 1.4235907690132355e-05, "loss": 1.9247, "step": 15415 }, { "epoch": 0.38, "learning_rate": 1.4235184505457711e-05, "loss": 2.0807, "step": 15416 }, { "epoch": 0.38, "learning_rate": 1.4234461293790851e-05, "loss": 2.1575, "step": 15417 }, { "epoch": 0.38, "learning_rate": 1.4233738055136393e-05, "loss": 2.0851, "step": 15418 }, { "epoch": 0.38, "learning_rate": 1.4233014789498943e-05, "loss": 2.0818, "step": 15419 }, { "epoch": 0.38, "learning_rate": 1.4232291496883108e-05, "loss": 2.1525, "step": 15420 }, { "epoch": 0.38, "learning_rate": 1.4231568177293501e-05, "loss": 2.1448, "step": 15421 }, { "epoch": 0.38, "learning_rate": 1.4230844830734732e-05, "loss": 2.0137, "step": 15422 }, { "epoch": 0.38, "learning_rate": 1.423012145721141e-05, "loss": 2.1188, "step": 15423 }, { "epoch": 0.38, "learning_rate": 1.4229398056728145e-05, "loss": 2.0677, "step": 15424 }, { "epoch": 0.38, "learning_rate": 1.4228674629289549e-05, "loss": 2.1928, "step": 15425 }, { "epoch": 0.38, "learning_rate": 1.4227951174900228e-05, "loss": 2.1139, "step": 15426 }, { "epoch": 0.38, "learning_rate": 1.4227227693564798e-05, "loss": 2.2151, "step": 15427 }, { "epoch": 0.38, "learning_rate": 1.4226504185287869e-05, "loss": 2.0346, "step": 15428 }, { "epoch": 0.38, "learning_rate": 1.422578065007405e-05, "loss": 2.1072, "step": 15429 }, { "epoch": 0.38, "learning_rate": 1.4225057087927959e-05, "loss": 1.971, "step": 15430 }, { "epoch": 0.38, "learning_rate": 1.4224333498854195e-05, "loss": 1.9648, "step": 15431 }, { "epoch": 0.38, "learning_rate": 1.4223609882857381e-05, "loss": 2.1755, "step": 15432 }, { "epoch": 0.38, "learning_rate": 1.4222886239942124e-05, "loss": 2.2587, "step": 15433 }, { "epoch": 0.38, "learning_rate": 1.4222162570113039e-05, "loss": 1.9488, "step": 15434 }, { "epoch": 0.38, "learning_rate": 1.4221438873374731e-05, "loss": 2.0771, "step": 15435 }, { "epoch": 0.38, "learning_rate": 1.4220715149731819e-05, "loss": 2.2267, "step": 15436 }, { "epoch": 0.38, "learning_rate": 1.4219991399188915e-05, "loss": 2.0209, "step": 15437 }, { "epoch": 0.38, "learning_rate": 1.4219267621750633e-05, "loss": 2.0366, "step": 15438 }, { "epoch": 0.38, "learning_rate": 1.4218543817421579e-05, "loss": 2.1084, "step": 15439 }, { "epoch": 0.38, "learning_rate": 1.4217819986206374e-05, "loss": 2.2335, "step": 15440 }, { "epoch": 0.38, "learning_rate": 1.4217096128109624e-05, "loss": 2.0732, "step": 15441 }, { "epoch": 0.38, "learning_rate": 1.4216372243135949e-05, "loss": 2.0975, "step": 15442 }, { "epoch": 0.38, "learning_rate": 1.4215648331289959e-05, "loss": 2.1593, "step": 15443 }, { "epoch": 0.38, "learning_rate": 1.4214924392576268e-05, "loss": 2.1706, "step": 15444 }, { "epoch": 0.38, "learning_rate": 1.4214200426999492e-05, "loss": 2.2256, "step": 15445 }, { "epoch": 0.38, "learning_rate": 1.421347643456424e-05, "loss": 1.9748, "step": 15446 }, { "epoch": 0.38, "learning_rate": 1.421275241527513e-05, "loss": 2.1669, "step": 15447 }, { "epoch": 0.38, "learning_rate": 1.421202836913678e-05, "loss": 1.9729, "step": 15448 }, { "epoch": 0.38, "learning_rate": 1.4211304296153796e-05, "loss": 1.9298, "step": 15449 }, { "epoch": 0.38, "learning_rate": 1.4210580196330802e-05, "loss": 2.0231, "step": 15450 }, { "epoch": 0.38, "learning_rate": 1.4209856069672403e-05, "loss": 2.0525, "step": 15451 }, { "epoch": 0.38, "learning_rate": 1.420913191618322e-05, "loss": 2.1089, "step": 15452 }, { "epoch": 0.38, "learning_rate": 1.4208407735867869e-05, "loss": 2.1129, "step": 15453 }, { "epoch": 0.38, "learning_rate": 1.4207683528730964e-05, "loss": 2.1816, "step": 15454 }, { "epoch": 0.38, "learning_rate": 1.4206959294777123e-05, "loss": 2.0154, "step": 15455 }, { "epoch": 0.38, "learning_rate": 1.4206235034010955e-05, "loss": 2.0779, "step": 15456 }, { "epoch": 0.38, "learning_rate": 1.4205510746437084e-05, "loss": 1.9897, "step": 15457 }, { "epoch": 0.38, "learning_rate": 1.420478643206012e-05, "loss": 1.9319, "step": 15458 }, { "epoch": 0.38, "learning_rate": 1.4204062090884685e-05, "loss": 2.1817, "step": 15459 }, { "epoch": 0.38, "learning_rate": 1.4203337722915386e-05, "loss": 2.1918, "step": 15460 }, { "epoch": 0.38, "learning_rate": 1.4202613328156853e-05, "loss": 1.9123, "step": 15461 }, { "epoch": 0.38, "learning_rate": 1.420188890661369e-05, "loss": 1.9993, "step": 15462 }, { "epoch": 0.38, "learning_rate": 1.4201164458290524e-05, "loss": 2.0932, "step": 15463 }, { "epoch": 0.38, "learning_rate": 1.4200439983191964e-05, "loss": 1.9797, "step": 15464 }, { "epoch": 0.38, "learning_rate": 1.4199715481322633e-05, "loss": 2.0951, "step": 15465 }, { "epoch": 0.38, "learning_rate": 1.4198990952687148e-05, "loss": 2.2513, "step": 15466 }, { "epoch": 0.38, "learning_rate": 1.4198266397290126e-05, "loss": 2.1725, "step": 15467 }, { "epoch": 0.38, "learning_rate": 1.4197541815136181e-05, "loss": 2.1088, "step": 15468 }, { "epoch": 0.38, "learning_rate": 1.4196817206229937e-05, "loss": 1.904, "step": 15469 }, { "epoch": 0.38, "learning_rate": 1.4196092570576006e-05, "loss": 2.1025, "step": 15470 }, { "epoch": 0.38, "learning_rate": 1.419536790817901e-05, "loss": 2.0755, "step": 15471 }, { "epoch": 0.38, "learning_rate": 1.4194643219043566e-05, "loss": 2.2155, "step": 15472 }, { "epoch": 0.38, "learning_rate": 1.4193918503174297e-05, "loss": 2.1505, "step": 15473 }, { "epoch": 0.38, "learning_rate": 1.4193193760575819e-05, "loss": 2.0375, "step": 15474 }, { "epoch": 0.38, "learning_rate": 1.4192468991252745e-05, "loss": 2.0909, "step": 15475 }, { "epoch": 0.38, "learning_rate": 1.4191744195209707e-05, "loss": 1.824, "step": 15476 }, { "epoch": 0.38, "learning_rate": 1.4191019372451314e-05, "loss": 2.064, "step": 15477 }, { "epoch": 0.38, "learning_rate": 1.419029452298219e-05, "loss": 2.0236, "step": 15478 }, { "epoch": 0.38, "learning_rate": 1.4189569646806951e-05, "loss": 1.9748, "step": 15479 }, { "epoch": 0.38, "learning_rate": 1.4188844743930219e-05, "loss": 2.1489, "step": 15480 }, { "epoch": 0.38, "learning_rate": 1.4188119814356616e-05, "loss": 1.9807, "step": 15481 }, { "epoch": 0.38, "learning_rate": 1.4187394858090762e-05, "loss": 2.0666, "step": 15482 }, { "epoch": 0.38, "learning_rate": 1.4186669875137273e-05, "loss": 2.2071, "step": 15483 }, { "epoch": 0.38, "learning_rate": 1.4185944865500774e-05, "loss": 2.0797, "step": 15484 }, { "epoch": 0.38, "learning_rate": 1.4185219829185883e-05, "loss": 1.9433, "step": 15485 }, { "epoch": 0.38, "learning_rate": 1.4184494766197223e-05, "loss": 2.0568, "step": 15486 }, { "epoch": 0.38, "learning_rate": 1.4183769676539416e-05, "loss": 2.0587, "step": 15487 }, { "epoch": 0.38, "learning_rate": 1.4183044560217078e-05, "loss": 2.2815, "step": 15488 }, { "epoch": 0.38, "learning_rate": 1.4182319417234835e-05, "loss": 2.2176, "step": 15489 }, { "epoch": 0.38, "learning_rate": 1.4181594247597307e-05, "loss": 2.0325, "step": 15490 }, { "epoch": 0.38, "learning_rate": 1.418086905130912e-05, "loss": 2.0636, "step": 15491 }, { "epoch": 0.38, "learning_rate": 1.4180143828374888e-05, "loss": 2.1134, "step": 15492 }, { "epoch": 0.38, "learning_rate": 1.417941857879924e-05, "loss": 2.2103, "step": 15493 }, { "epoch": 0.38, "learning_rate": 1.4178693302586792e-05, "loss": 2.0934, "step": 15494 }, { "epoch": 0.38, "learning_rate": 1.4177967999742171e-05, "loss": 1.8858, "step": 15495 }, { "epoch": 0.38, "learning_rate": 1.4177242670269999e-05, "loss": 2.0103, "step": 15496 }, { "epoch": 0.38, "learning_rate": 1.4176517314174895e-05, "loss": 2.0175, "step": 15497 }, { "epoch": 0.38, "learning_rate": 1.4175791931461491e-05, "loss": 2.1799, "step": 15498 }, { "epoch": 0.38, "learning_rate": 1.41750665221344e-05, "loss": 1.9702, "step": 15499 }, { "epoch": 0.38, "learning_rate": 1.4174341086198248e-05, "loss": 2.0332, "step": 15500 }, { "epoch": 0.38, "learning_rate": 1.4173615623657666e-05, "loss": 2.0427, "step": 15501 }, { "epoch": 0.38, "learning_rate": 1.4172890134517267e-05, "loss": 2.0451, "step": 15502 }, { "epoch": 0.38, "learning_rate": 1.417216461878168e-05, "loss": 1.9374, "step": 15503 }, { "epoch": 0.38, "learning_rate": 1.417143907645553e-05, "loss": 2.0973, "step": 15504 }, { "epoch": 0.38, "learning_rate": 1.4170713507543437e-05, "loss": 2.2347, "step": 15505 }, { "epoch": 0.38, "learning_rate": 1.4169987912050025e-05, "loss": 2.2318, "step": 15506 }, { "epoch": 0.38, "learning_rate": 1.4169262289979926e-05, "loss": 2.2816, "step": 15507 }, { "epoch": 0.38, "learning_rate": 1.4168536641337757e-05, "loss": 2.0539, "step": 15508 }, { "epoch": 0.38, "learning_rate": 1.4167810966128145e-05, "loss": 2.061, "step": 15509 }, { "epoch": 0.38, "learning_rate": 1.416708526435572e-05, "loss": 2.0302, "step": 15510 }, { "epoch": 0.38, "learning_rate": 1.41663595360251e-05, "loss": 2.1801, "step": 15511 }, { "epoch": 0.38, "learning_rate": 1.4165633781140913e-05, "loss": 2.0249, "step": 15512 }, { "epoch": 0.38, "learning_rate": 1.4164907999707783e-05, "loss": 2.1206, "step": 15513 }, { "epoch": 0.38, "learning_rate": 1.4164182191730337e-05, "loss": 2.1624, "step": 15514 }, { "epoch": 0.38, "learning_rate": 1.4163456357213202e-05, "loss": 2.2726, "step": 15515 }, { "epoch": 0.38, "learning_rate": 1.4162730496161002e-05, "loss": 2.0754, "step": 15516 }, { "epoch": 0.38, "learning_rate": 1.4162004608578365e-05, "loss": 1.9036, "step": 15517 }, { "epoch": 0.38, "learning_rate": 1.4161278694469916e-05, "loss": 2.1172, "step": 15518 }, { "epoch": 0.38, "learning_rate": 1.4160552753840282e-05, "loss": 2.0779, "step": 15519 }, { "epoch": 0.38, "learning_rate": 1.4159826786694088e-05, "loss": 2.0653, "step": 15520 }, { "epoch": 0.38, "learning_rate": 1.4159100793035965e-05, "loss": 2.1513, "step": 15521 }, { "epoch": 0.38, "learning_rate": 1.4158374772870534e-05, "loss": 2.2581, "step": 15522 }, { "epoch": 0.38, "learning_rate": 1.4157648726202426e-05, "loss": 2.0315, "step": 15523 }, { "epoch": 0.38, "learning_rate": 1.4156922653036266e-05, "loss": 2.0127, "step": 15524 }, { "epoch": 0.38, "learning_rate": 1.4156196553376688e-05, "loss": 1.9456, "step": 15525 }, { "epoch": 0.38, "learning_rate": 1.4155470427228312e-05, "loss": 2.1516, "step": 15526 }, { "epoch": 0.38, "learning_rate": 1.4154744274595771e-05, "loss": 2.0719, "step": 15527 }, { "epoch": 0.38, "learning_rate": 1.4154018095483688e-05, "loss": 2.2013, "step": 15528 }, { "epoch": 0.38, "learning_rate": 1.4153291889896696e-05, "loss": 2.1544, "step": 15529 }, { "epoch": 0.38, "learning_rate": 1.415256565783942e-05, "loss": 2.1673, "step": 15530 }, { "epoch": 0.38, "learning_rate": 1.4151839399316492e-05, "loss": 2.038, "step": 15531 }, { "epoch": 0.38, "learning_rate": 1.4151113114332536e-05, "loss": 1.9672, "step": 15532 }, { "epoch": 0.38, "learning_rate": 1.4150386802892183e-05, "loss": 2.0721, "step": 15533 }, { "epoch": 0.38, "learning_rate": 1.4149660465000061e-05, "loss": 2.0689, "step": 15534 }, { "epoch": 0.38, "learning_rate": 1.4148934100660805e-05, "loss": 2.0534, "step": 15535 }, { "epoch": 0.38, "learning_rate": 1.4148207709879038e-05, "loss": 2.0286, "step": 15536 }, { "epoch": 0.38, "learning_rate": 1.4147481292659391e-05, "loss": 2.0241, "step": 15537 }, { "epoch": 0.38, "learning_rate": 1.4146754849006494e-05, "loss": 2.3475, "step": 15538 }, { "epoch": 0.38, "learning_rate": 1.4146028378924976e-05, "loss": 2.0803, "step": 15539 }, { "epoch": 0.38, "learning_rate": 1.414530188241947e-05, "loss": 2.0475, "step": 15540 }, { "epoch": 0.38, "learning_rate": 1.4144575359494601e-05, "loss": 2.1421, "step": 15541 }, { "epoch": 0.38, "learning_rate": 1.4143848810155008e-05, "loss": 2.0214, "step": 15542 }, { "epoch": 0.38, "learning_rate": 1.4143122234405311e-05, "loss": 2.1683, "step": 15543 }, { "epoch": 0.38, "learning_rate": 1.4142395632250147e-05, "loss": 2.1019, "step": 15544 }, { "epoch": 0.38, "learning_rate": 1.4141669003694147e-05, "loss": 2.12, "step": 15545 }, { "epoch": 0.38, "learning_rate": 1.4140942348741942e-05, "loss": 2.0718, "step": 15546 }, { "epoch": 0.38, "learning_rate": 1.4140215667398159e-05, "loss": 2.1676, "step": 15547 }, { "epoch": 0.38, "learning_rate": 1.4139488959667434e-05, "loss": 2.2303, "step": 15548 }, { "epoch": 0.38, "learning_rate": 1.4138762225554396e-05, "loss": 2.1407, "step": 15549 }, { "epoch": 0.38, "learning_rate": 1.4138035465063675e-05, "loss": 1.9775, "step": 15550 }, { "epoch": 0.38, "learning_rate": 1.4137308678199909e-05, "loss": 2.125, "step": 15551 }, { "epoch": 0.38, "learning_rate": 1.4136581864967724e-05, "loss": 2.0692, "step": 15552 }, { "epoch": 0.38, "learning_rate": 1.4135855025371759e-05, "loss": 2.0068, "step": 15553 }, { "epoch": 0.38, "learning_rate": 1.4135128159416638e-05, "loss": 2.2919, "step": 15554 }, { "epoch": 0.38, "learning_rate": 1.4134401267107e-05, "loss": 2.2387, "step": 15555 }, { "epoch": 0.38, "learning_rate": 1.4133674348447474e-05, "loss": 2.2069, "step": 15556 }, { "epoch": 0.38, "learning_rate": 1.4132947403442694e-05, "loss": 2.0933, "step": 15557 }, { "epoch": 0.38, "learning_rate": 1.4132220432097295e-05, "loss": 2.0349, "step": 15558 }, { "epoch": 0.38, "learning_rate": 1.4131493434415907e-05, "loss": 2.229, "step": 15559 }, { "epoch": 0.38, "learning_rate": 1.4130766410403164e-05, "loss": 2.0983, "step": 15560 }, { "epoch": 0.38, "learning_rate": 1.4130039360063703e-05, "loss": 2.1532, "step": 15561 }, { "epoch": 0.38, "learning_rate": 1.4129312283402152e-05, "loss": 2.0544, "step": 15562 }, { "epoch": 0.38, "learning_rate": 1.4128585180423154e-05, "loss": 1.9324, "step": 15563 }, { "epoch": 0.38, "learning_rate": 1.4127858051131331e-05, "loss": 1.922, "step": 15564 }, { "epoch": 0.38, "learning_rate": 1.4127130895531327e-05, "loss": 1.9833, "step": 15565 }, { "epoch": 0.38, "learning_rate": 1.412640371362777e-05, "loss": 2.1235, "step": 15566 }, { "epoch": 0.38, "learning_rate": 1.41256765054253e-05, "loss": 2.0475, "step": 15567 }, { "epoch": 0.38, "learning_rate": 1.4124949270928547e-05, "loss": 2.2542, "step": 15568 }, { "epoch": 0.38, "learning_rate": 1.412422201014215e-05, "loss": 1.9247, "step": 15569 }, { "epoch": 0.38, "learning_rate": 1.412349472307074e-05, "loss": 2.1812, "step": 15570 }, { "epoch": 0.38, "learning_rate": 1.4122767409718959e-05, "loss": 2.1246, "step": 15571 }, { "epoch": 0.38, "learning_rate": 1.4122040070091432e-05, "loss": 2.0029, "step": 15572 }, { "epoch": 0.38, "learning_rate": 1.4121312704192802e-05, "loss": 2.0339, "step": 15573 }, { "epoch": 0.38, "learning_rate": 1.4120585312027704e-05, "loss": 2.0784, "step": 15574 }, { "epoch": 0.38, "learning_rate": 1.4119857893600773e-05, "loss": 2.0208, "step": 15575 }, { "epoch": 0.38, "learning_rate": 1.4119130448916644e-05, "loss": 2.1345, "step": 15576 }, { "epoch": 0.38, "learning_rate": 1.4118402977979954e-05, "loss": 1.9023, "step": 15577 }, { "epoch": 0.38, "learning_rate": 1.411767548079534e-05, "loss": 2.151, "step": 15578 }, { "epoch": 0.38, "learning_rate": 1.411694795736744e-05, "loss": 2.0705, "step": 15579 }, { "epoch": 0.38, "learning_rate": 1.411622040770089e-05, "loss": 2.0823, "step": 15580 }, { "epoch": 0.38, "learning_rate": 1.4115492831800322e-05, "loss": 2.0952, "step": 15581 }, { "epoch": 0.38, "learning_rate": 1.4114765229670378e-05, "loss": 2.1206, "step": 15582 }, { "epoch": 0.38, "learning_rate": 1.4114037601315696e-05, "loss": 2.1237, "step": 15583 }, { "epoch": 0.38, "learning_rate": 1.4113309946740908e-05, "loss": 2.0729, "step": 15584 }, { "epoch": 0.38, "learning_rate": 1.4112582265950658e-05, "loss": 2.1741, "step": 15585 }, { "epoch": 0.38, "learning_rate": 1.411185455894958e-05, "loss": 2.141, "step": 15586 }, { "epoch": 0.38, "learning_rate": 1.4111126825742314e-05, "loss": 2.1547, "step": 15587 }, { "epoch": 0.38, "learning_rate": 1.4110399066333498e-05, "loss": 2.0606, "step": 15588 }, { "epoch": 0.38, "learning_rate": 1.4109671280727768e-05, "loss": 1.9505, "step": 15589 }, { "epoch": 0.38, "learning_rate": 1.4108943468929763e-05, "loss": 2.2937, "step": 15590 }, { "epoch": 0.38, "learning_rate": 1.4108215630944122e-05, "loss": 2.0682, "step": 15591 }, { "epoch": 0.38, "learning_rate": 1.4107487766775486e-05, "loss": 2.2856, "step": 15592 }, { "epoch": 0.38, "learning_rate": 1.4106759876428488e-05, "loss": 2.1719, "step": 15593 }, { "epoch": 0.38, "learning_rate": 1.4106031959907773e-05, "loss": 2.1104, "step": 15594 }, { "epoch": 0.38, "learning_rate": 1.410530401721798e-05, "loss": 2.0623, "step": 15595 }, { "epoch": 0.38, "learning_rate": 1.4104576048363745e-05, "loss": 2.0814, "step": 15596 }, { "epoch": 0.38, "learning_rate": 1.410384805334971e-05, "loss": 2.0233, "step": 15597 }, { "epoch": 0.38, "learning_rate": 1.4103120032180513e-05, "loss": 2.2028, "step": 15598 }, { "epoch": 0.38, "learning_rate": 1.4102391984860794e-05, "loss": 1.8572, "step": 15599 }, { "epoch": 0.38, "learning_rate": 1.4101663911395198e-05, "loss": 2.1375, "step": 15600 }, { "epoch": 0.38, "learning_rate": 1.4100935811788357e-05, "loss": 2.2146, "step": 15601 }, { "epoch": 0.38, "learning_rate": 1.4100207686044915e-05, "loss": 2.1093, "step": 15602 }, { "epoch": 0.38, "learning_rate": 1.4099479534169516e-05, "loss": 2.2036, "step": 15603 }, { "epoch": 0.38, "learning_rate": 1.4098751356166797e-05, "loss": 2.1512, "step": 15604 }, { "epoch": 0.38, "learning_rate": 1.4098023152041398e-05, "loss": 2.0876, "step": 15605 }, { "epoch": 0.38, "learning_rate": 1.4097294921797965e-05, "loss": 2.0661, "step": 15606 }, { "epoch": 0.38, "learning_rate": 1.4096566665441135e-05, "loss": 2.0744, "step": 15607 }, { "epoch": 0.38, "learning_rate": 1.4095838382975553e-05, "loss": 2.0442, "step": 15608 }, { "epoch": 0.38, "learning_rate": 1.4095110074405855e-05, "loss": 2.1915, "step": 15609 }, { "epoch": 0.38, "learning_rate": 1.4094381739736687e-05, "loss": 1.9688, "step": 15610 }, { "epoch": 0.38, "learning_rate": 1.4093653378972686e-05, "loss": 2.0305, "step": 15611 }, { "epoch": 0.38, "learning_rate": 1.4092924992118504e-05, "loss": 2.1558, "step": 15612 }, { "epoch": 0.38, "learning_rate": 1.4092196579178774e-05, "loss": 2.0709, "step": 15613 }, { "epoch": 0.38, "learning_rate": 1.4091468140158145e-05, "loss": 2.0083, "step": 15614 }, { "epoch": 0.38, "learning_rate": 1.4090739675061251e-05, "loss": 1.9639, "step": 15615 }, { "epoch": 0.38, "learning_rate": 1.4090011183892744e-05, "loss": 2.1281, "step": 15616 }, { "epoch": 0.38, "learning_rate": 1.408928266665726e-05, "loss": 2.0916, "step": 15617 }, { "epoch": 0.38, "learning_rate": 1.4088554123359448e-05, "loss": 1.9339, "step": 15618 }, { "epoch": 0.38, "learning_rate": 1.4087825554003947e-05, "loss": 2.0498, "step": 15619 }, { "epoch": 0.39, "learning_rate": 1.40870969585954e-05, "loss": 2.1873, "step": 15620 }, { "epoch": 0.39, "learning_rate": 1.4086368337138451e-05, "loss": 2.1728, "step": 15621 }, { "epoch": 0.39, "learning_rate": 1.4085639689637747e-05, "loss": 2.3174, "step": 15622 }, { "epoch": 0.39, "learning_rate": 1.408491101609793e-05, "loss": 2.0275, "step": 15623 }, { "epoch": 0.39, "learning_rate": 1.4084182316523646e-05, "loss": 2.064, "step": 15624 }, { "epoch": 0.39, "learning_rate": 1.4083453590919533e-05, "loss": 2.1141, "step": 15625 }, { "epoch": 0.39, "learning_rate": 1.4082724839290243e-05, "loss": 2.0615, "step": 15626 }, { "epoch": 0.39, "learning_rate": 1.4081996061640415e-05, "loss": 2.1171, "step": 15627 }, { "epoch": 0.39, "learning_rate": 1.4081267257974694e-05, "loss": 2.0545, "step": 15628 }, { "epoch": 0.39, "learning_rate": 1.408053842829773e-05, "loss": 2.1228, "step": 15629 }, { "epoch": 0.39, "learning_rate": 1.4079809572614162e-05, "loss": 2.1181, "step": 15630 }, { "epoch": 0.39, "learning_rate": 1.407908069092864e-05, "loss": 2.1674, "step": 15631 }, { "epoch": 0.39, "learning_rate": 1.4078351783245808e-05, "loss": 2.0082, "step": 15632 }, { "epoch": 0.39, "learning_rate": 1.407762284957031e-05, "loss": 1.9743, "step": 15633 }, { "epoch": 0.39, "learning_rate": 1.4076893889906794e-05, "loss": 2.0497, "step": 15634 }, { "epoch": 0.39, "learning_rate": 1.4076164904259902e-05, "loss": 1.9869, "step": 15635 }, { "epoch": 0.39, "learning_rate": 1.4075435892634284e-05, "loss": 2.1024, "step": 15636 }, { "epoch": 0.39, "learning_rate": 1.4074706855034584e-05, "loss": 2.058, "step": 15637 }, { "epoch": 0.39, "learning_rate": 1.4073977791465451e-05, "loss": 2.1564, "step": 15638 }, { "epoch": 0.39, "learning_rate": 1.4073248701931528e-05, "loss": 1.9388, "step": 15639 }, { "epoch": 0.39, "learning_rate": 1.4072519586437465e-05, "loss": 2.2123, "step": 15640 }, { "epoch": 0.39, "learning_rate": 1.4071790444987907e-05, "loss": 2.0863, "step": 15641 }, { "epoch": 0.39, "learning_rate": 1.4071061277587503e-05, "loss": 2.0638, "step": 15642 }, { "epoch": 0.39, "learning_rate": 1.4070332084240898e-05, "loss": 2.1719, "step": 15643 }, { "epoch": 0.39, "learning_rate": 1.4069602864952738e-05, "loss": 2.0691, "step": 15644 }, { "epoch": 0.39, "learning_rate": 1.4068873619727676e-05, "loss": 2.0897, "step": 15645 }, { "epoch": 0.39, "learning_rate": 1.4068144348570352e-05, "loss": 2.0303, "step": 15646 }, { "epoch": 0.39, "learning_rate": 1.4067415051485422e-05, "loss": 2.0797, "step": 15647 }, { "epoch": 0.39, "learning_rate": 1.406668572847753e-05, "loss": 1.999, "step": 15648 }, { "epoch": 0.39, "learning_rate": 1.4065956379551322e-05, "loss": 2.1648, "step": 15649 }, { "epoch": 0.39, "learning_rate": 1.4065227004711452e-05, "loss": 2.0962, "step": 15650 }, { "epoch": 0.39, "learning_rate": 1.4064497603962566e-05, "loss": 2.0507, "step": 15651 }, { "epoch": 0.39, "learning_rate": 1.4063768177309308e-05, "loss": 2.0511, "step": 15652 }, { "epoch": 0.39, "learning_rate": 1.4063038724756338e-05, "loss": 2.055, "step": 15653 }, { "epoch": 0.39, "learning_rate": 1.406230924630829e-05, "loss": 2.0259, "step": 15654 }, { "epoch": 0.39, "learning_rate": 1.4061579741969823e-05, "loss": 2.0822, "step": 15655 }, { "epoch": 0.39, "learning_rate": 1.406085021174559e-05, "loss": 2.0901, "step": 15656 }, { "epoch": 0.39, "learning_rate": 1.4060120655640231e-05, "loss": 2.0596, "step": 15657 }, { "epoch": 0.39, "learning_rate": 1.4059391073658402e-05, "loss": 1.8994, "step": 15658 }, { "epoch": 0.39, "learning_rate": 1.4058661465804747e-05, "loss": 2.1247, "step": 15659 }, { "epoch": 0.39, "learning_rate": 1.4057931832083925e-05, "loss": 1.9905, "step": 15660 }, { "epoch": 0.39, "learning_rate": 1.4057202172500578e-05, "loss": 2.3157, "step": 15661 }, { "epoch": 0.39, "learning_rate": 1.405647248705936e-05, "loss": 2.1057, "step": 15662 }, { "epoch": 0.39, "learning_rate": 1.4055742775764918e-05, "loss": 2.1025, "step": 15663 }, { "epoch": 0.39, "learning_rate": 1.4055013038621908e-05, "loss": 2.2188, "step": 15664 }, { "epoch": 0.39, "learning_rate": 1.4054283275634979e-05, "loss": 2.1022, "step": 15665 }, { "epoch": 0.39, "learning_rate": 1.4053553486808781e-05, "loss": 2.0688, "step": 15666 }, { "epoch": 0.39, "learning_rate": 1.4052823672147965e-05, "loss": 2.1287, "step": 15667 }, { "epoch": 0.39, "learning_rate": 1.4052093831657185e-05, "loss": 2.0135, "step": 15668 }, { "epoch": 0.39, "learning_rate": 1.405136396534109e-05, "loss": 2.0837, "step": 15669 }, { "epoch": 0.39, "learning_rate": 1.405063407320433e-05, "loss": 2.1269, "step": 15670 }, { "epoch": 0.39, "learning_rate": 1.4049904155251561e-05, "loss": 2.1504, "step": 15671 }, { "epoch": 0.39, "learning_rate": 1.404917421148743e-05, "loss": 2.0231, "step": 15672 }, { "epoch": 0.39, "learning_rate": 1.4048444241916594e-05, "loss": 2.0079, "step": 15673 }, { "epoch": 0.39, "learning_rate": 1.4047714246543704e-05, "loss": 2.0487, "step": 15674 }, { "epoch": 0.39, "learning_rate": 1.4046984225373413e-05, "loss": 2.0662, "step": 15675 }, { "epoch": 0.39, "learning_rate": 1.4046254178410372e-05, "loss": 2.0144, "step": 15676 }, { "epoch": 0.39, "learning_rate": 1.4045524105659234e-05, "loss": 1.8729, "step": 15677 }, { "epoch": 0.39, "learning_rate": 1.4044794007124652e-05, "loss": 2.1556, "step": 15678 }, { "epoch": 0.39, "learning_rate": 1.404406388281128e-05, "loss": 1.9527, "step": 15679 }, { "epoch": 0.39, "learning_rate": 1.4043333732723773e-05, "loss": 2.1304, "step": 15680 }, { "epoch": 0.39, "learning_rate": 1.4042603556866778e-05, "loss": 2.0874, "step": 15681 }, { "epoch": 0.39, "learning_rate": 1.4041873355244957e-05, "loss": 2.0677, "step": 15682 }, { "epoch": 0.39, "learning_rate": 1.4041143127862955e-05, "loss": 1.9838, "step": 15683 }, { "epoch": 0.39, "learning_rate": 1.4040412874725436e-05, "loss": 2.2289, "step": 15684 }, { "epoch": 0.39, "learning_rate": 1.4039682595837047e-05, "loss": 2.0685, "step": 15685 }, { "epoch": 0.39, "learning_rate": 1.4038952291202446e-05, "loss": 2.0941, "step": 15686 }, { "epoch": 0.39, "learning_rate": 1.4038221960826285e-05, "loss": 2.0035, "step": 15687 }, { "epoch": 0.39, "learning_rate": 1.4037491604713217e-05, "loss": 2.0024, "step": 15688 }, { "epoch": 0.39, "learning_rate": 1.4036761222867903e-05, "loss": 2.0376, "step": 15689 }, { "epoch": 0.39, "learning_rate": 1.403603081529499e-05, "loss": 1.9487, "step": 15690 }, { "epoch": 0.39, "learning_rate": 1.403530038199914e-05, "loss": 2.1709, "step": 15691 }, { "epoch": 0.39, "learning_rate": 1.4034569922985006e-05, "loss": 2.0524, "step": 15692 }, { "epoch": 0.39, "learning_rate": 1.4033839438257242e-05, "loss": 2.0634, "step": 15693 }, { "epoch": 0.39, "learning_rate": 1.4033108927820505e-05, "loss": 2.0172, "step": 15694 }, { "epoch": 0.39, "learning_rate": 1.403237839167945e-05, "loss": 2.194, "step": 15695 }, { "epoch": 0.39, "learning_rate": 1.4031647829838736e-05, "loss": 2.0525, "step": 15696 }, { "epoch": 0.39, "learning_rate": 1.403091724230301e-05, "loss": 1.9869, "step": 15697 }, { "epoch": 0.39, "learning_rate": 1.403018662907694e-05, "loss": 2.1095, "step": 15698 }, { "epoch": 0.39, "learning_rate": 1.4029455990165176e-05, "loss": 2.1864, "step": 15699 }, { "epoch": 0.39, "learning_rate": 1.4028725325572377e-05, "loss": 2.1748, "step": 15700 }, { "epoch": 0.39, "learning_rate": 1.40279946353032e-05, "loss": 2.1562, "step": 15701 }, { "epoch": 0.39, "learning_rate": 1.4027263919362294e-05, "loss": 1.8107, "step": 15702 }, { "epoch": 0.39, "learning_rate": 1.4026533177754328e-05, "loss": 2.1232, "step": 15703 }, { "epoch": 0.39, "learning_rate": 1.4025802410483953e-05, "loss": 2.1361, "step": 15704 }, { "epoch": 0.39, "learning_rate": 1.4025071617555828e-05, "loss": 2.1406, "step": 15705 }, { "epoch": 0.39, "learning_rate": 1.402434079897461e-05, "loss": 2.1443, "step": 15706 }, { "epoch": 0.39, "learning_rate": 1.4023609954744952e-05, "loss": 2.1928, "step": 15707 }, { "epoch": 0.39, "learning_rate": 1.4022879084871519e-05, "loss": 2.1206, "step": 15708 }, { "epoch": 0.39, "learning_rate": 1.4022148189358968e-05, "loss": 2.0758, "step": 15709 }, { "epoch": 0.39, "learning_rate": 1.4021417268211957e-05, "loss": 2.1953, "step": 15710 }, { "epoch": 0.39, "learning_rate": 1.4020686321435142e-05, "loss": 2.1692, "step": 15711 }, { "epoch": 0.39, "learning_rate": 1.401995534903318e-05, "loss": 1.9827, "step": 15712 }, { "epoch": 0.39, "learning_rate": 1.4019224351010735e-05, "loss": 2.0608, "step": 15713 }, { "epoch": 0.39, "learning_rate": 1.4018493327372465e-05, "loss": 2.0237, "step": 15714 }, { "epoch": 0.39, "learning_rate": 1.4017762278123024e-05, "loss": 2.1101, "step": 15715 }, { "epoch": 0.39, "learning_rate": 1.4017031203267075e-05, "loss": 2.1123, "step": 15716 }, { "epoch": 0.39, "learning_rate": 1.4016300102809278e-05, "loss": 1.9413, "step": 15717 }, { "epoch": 0.39, "learning_rate": 1.4015568976754292e-05, "loss": 2.0949, "step": 15718 }, { "epoch": 0.39, "learning_rate": 1.4014837825106776e-05, "loss": 1.8451, "step": 15719 }, { "epoch": 0.39, "learning_rate": 1.4014106647871391e-05, "loss": 2.1717, "step": 15720 }, { "epoch": 0.39, "learning_rate": 1.4013375445052795e-05, "loss": 2.2139, "step": 15721 }, { "epoch": 0.39, "learning_rate": 1.4012644216655648e-05, "loss": 2.1102, "step": 15722 }, { "epoch": 0.39, "learning_rate": 1.4011912962684615e-05, "loss": 2.1252, "step": 15723 }, { "epoch": 0.39, "learning_rate": 1.4011181683144353e-05, "loss": 2.1769, "step": 15724 }, { "epoch": 0.39, "learning_rate": 1.401045037803952e-05, "loss": 2.1417, "step": 15725 }, { "epoch": 0.39, "learning_rate": 1.4009719047374779e-05, "loss": 2.1005, "step": 15726 }, { "epoch": 0.39, "learning_rate": 1.4008987691154796e-05, "loss": 2.1116, "step": 15727 }, { "epoch": 0.39, "learning_rate": 1.4008256309384228e-05, "loss": 1.966, "step": 15728 }, { "epoch": 0.39, "learning_rate": 1.4007524902067732e-05, "loss": 2.0622, "step": 15729 }, { "epoch": 0.39, "learning_rate": 1.4006793469209978e-05, "loss": 2.208, "step": 15730 }, { "epoch": 0.39, "learning_rate": 1.400606201081562e-05, "loss": 2.1687, "step": 15731 }, { "epoch": 0.39, "learning_rate": 1.4005330526889325e-05, "loss": 2.1798, "step": 15732 }, { "epoch": 0.39, "learning_rate": 1.4004599017435753e-05, "loss": 2.028, "step": 15733 }, { "epoch": 0.39, "learning_rate": 1.4003867482459565e-05, "loss": 2.0798, "step": 15734 }, { "epoch": 0.39, "learning_rate": 1.4003135921965428e-05, "loss": 2.0393, "step": 15735 }, { "epoch": 0.39, "learning_rate": 1.4002404335957998e-05, "loss": 2.1469, "step": 15736 }, { "epoch": 0.39, "learning_rate": 1.4001672724441944e-05, "loss": 2.04, "step": 15737 }, { "epoch": 0.39, "learning_rate": 1.4000941087421924e-05, "loss": 2.0618, "step": 15738 }, { "epoch": 0.39, "learning_rate": 1.4000209424902602e-05, "loss": 1.9866, "step": 15739 }, { "epoch": 0.39, "learning_rate": 1.3999477736888643e-05, "loss": 2.114, "step": 15740 }, { "epoch": 0.39, "learning_rate": 1.3998746023384706e-05, "loss": 1.9288, "step": 15741 }, { "epoch": 0.39, "learning_rate": 1.399801428439546e-05, "loss": 2.0653, "step": 15742 }, { "epoch": 0.39, "learning_rate": 1.3997282519925565e-05, "loss": 2.1855, "step": 15743 }, { "epoch": 0.39, "learning_rate": 1.3996550729979688e-05, "loss": 1.8341, "step": 15744 }, { "epoch": 0.39, "learning_rate": 1.399581891456249e-05, "loss": 2.053, "step": 15745 }, { "epoch": 0.39, "learning_rate": 1.3995087073678633e-05, "loss": 2.1534, "step": 15746 }, { "epoch": 0.39, "learning_rate": 1.3994355207332786e-05, "loss": 2.0729, "step": 15747 }, { "epoch": 0.39, "learning_rate": 1.3993623315529613e-05, "loss": 1.9864, "step": 15748 }, { "epoch": 0.39, "learning_rate": 1.3992891398273775e-05, "loss": 1.9652, "step": 15749 }, { "epoch": 0.39, "learning_rate": 1.3992159455569943e-05, "loss": 1.9426, "step": 15750 }, { "epoch": 0.39, "learning_rate": 1.399142748742277e-05, "loss": 1.9185, "step": 15751 }, { "epoch": 0.39, "learning_rate": 1.3990695493836935e-05, "loss": 2.0442, "step": 15752 }, { "epoch": 0.39, "learning_rate": 1.3989963474817094e-05, "loss": 1.9951, "step": 15753 }, { "epoch": 0.39, "learning_rate": 1.3989231430367918e-05, "loss": 1.9507, "step": 15754 }, { "epoch": 0.39, "learning_rate": 1.3988499360494069e-05, "loss": 1.9913, "step": 15755 }, { "epoch": 0.39, "learning_rate": 1.3987767265200213e-05, "loss": 2.1543, "step": 15756 }, { "epoch": 0.39, "learning_rate": 1.398703514449102e-05, "loss": 2.0214, "step": 15757 }, { "epoch": 0.39, "learning_rate": 1.3986302998371149e-05, "loss": 2.054, "step": 15758 }, { "epoch": 0.39, "learning_rate": 1.3985570826845272e-05, "loss": 2.0854, "step": 15759 }, { "epoch": 0.39, "learning_rate": 1.398483862991805e-05, "loss": 2.1568, "step": 15760 }, { "epoch": 0.39, "learning_rate": 1.3984106407594154e-05, "loss": 2.2077, "step": 15761 }, { "epoch": 0.39, "learning_rate": 1.3983374159878251e-05, "loss": 2.1331, "step": 15762 }, { "epoch": 0.39, "learning_rate": 1.3982641886775006e-05, "loss": 1.9423, "step": 15763 }, { "epoch": 0.39, "learning_rate": 1.3981909588289087e-05, "loss": 2.0955, "step": 15764 }, { "epoch": 0.39, "learning_rate": 1.398117726442516e-05, "loss": 1.9845, "step": 15765 }, { "epoch": 0.39, "learning_rate": 1.3980444915187893e-05, "loss": 2.1021, "step": 15766 }, { "epoch": 0.39, "learning_rate": 1.3979712540581952e-05, "loss": 2.1087, "step": 15767 }, { "epoch": 0.39, "learning_rate": 1.3978980140612009e-05, "loss": 2.0033, "step": 15768 }, { "epoch": 0.39, "learning_rate": 1.3978247715282725e-05, "loss": 2.009, "step": 15769 }, { "epoch": 0.39, "learning_rate": 1.3977515264598772e-05, "loss": 2.2389, "step": 15770 }, { "epoch": 0.39, "learning_rate": 1.397678278856482e-05, "loss": 2.0962, "step": 15771 }, { "epoch": 0.39, "learning_rate": 1.3976050287185535e-05, "loss": 2.1942, "step": 15772 }, { "epoch": 0.39, "learning_rate": 1.3975317760465585e-05, "loss": 2.1617, "step": 15773 }, { "epoch": 0.39, "learning_rate": 1.3974585208409642e-05, "loss": 2.0005, "step": 15774 }, { "epoch": 0.39, "learning_rate": 1.3973852631022368e-05, "loss": 2.2853, "step": 15775 }, { "epoch": 0.39, "learning_rate": 1.397312002830844e-05, "loss": 2.2288, "step": 15776 }, { "epoch": 0.39, "learning_rate": 1.3972387400272522e-05, "loss": 2.2694, "step": 15777 }, { "epoch": 0.39, "learning_rate": 1.3971654746919284e-05, "loss": 1.9107, "step": 15778 }, { "epoch": 0.39, "learning_rate": 1.3970922068253397e-05, "loss": 2.0227, "step": 15779 }, { "epoch": 0.39, "learning_rate": 1.3970189364279526e-05, "loss": 2.1545, "step": 15780 }, { "epoch": 0.39, "learning_rate": 1.3969456635002347e-05, "loss": 2.1166, "step": 15781 }, { "epoch": 0.39, "learning_rate": 1.3968723880426529e-05, "loss": 2.1217, "step": 15782 }, { "epoch": 0.39, "learning_rate": 1.3967991100556739e-05, "loss": 2.2323, "step": 15783 }, { "epoch": 0.39, "learning_rate": 1.3967258295397646e-05, "loss": 2.0457, "step": 15784 }, { "epoch": 0.39, "learning_rate": 1.3966525464953926e-05, "loss": 1.9731, "step": 15785 }, { "epoch": 0.39, "learning_rate": 1.3965792609230247e-05, "loss": 2.2552, "step": 15786 }, { "epoch": 0.39, "learning_rate": 1.3965059728231276e-05, "loss": 2.1032, "step": 15787 }, { "epoch": 0.39, "learning_rate": 1.396432682196169e-05, "loss": 1.9629, "step": 15788 }, { "epoch": 0.39, "learning_rate": 1.3963593890426154e-05, "loss": 2.0536, "step": 15789 }, { "epoch": 0.39, "learning_rate": 1.3962860933629347e-05, "loss": 2.0726, "step": 15790 }, { "epoch": 0.39, "learning_rate": 1.3962127951575934e-05, "loss": 2.0532, "step": 15791 }, { "epoch": 0.39, "learning_rate": 1.396139494427059e-05, "loss": 2.0316, "step": 15792 }, { "epoch": 0.39, "learning_rate": 1.3960661911717984e-05, "loss": 2.0399, "step": 15793 }, { "epoch": 0.39, "learning_rate": 1.3959928853922786e-05, "loss": 2.1024, "step": 15794 }, { "epoch": 0.39, "learning_rate": 1.3959195770889672e-05, "loss": 2.2637, "step": 15795 }, { "epoch": 0.39, "learning_rate": 1.3958462662623313e-05, "loss": 2.1672, "step": 15796 }, { "epoch": 0.39, "learning_rate": 1.3957729529128383e-05, "loss": 2.2257, "step": 15797 }, { "epoch": 0.39, "learning_rate": 1.3956996370409552e-05, "loss": 2.1252, "step": 15798 }, { "epoch": 0.39, "learning_rate": 1.3956263186471496e-05, "loss": 1.7125, "step": 15799 }, { "epoch": 0.39, "learning_rate": 1.3955529977318883e-05, "loss": 2.0714, "step": 15800 }, { "epoch": 0.39, "learning_rate": 1.395479674295639e-05, "loss": 2.1188, "step": 15801 }, { "epoch": 0.39, "learning_rate": 1.3954063483388687e-05, "loss": 2.1864, "step": 15802 }, { "epoch": 0.39, "learning_rate": 1.3953330198620447e-05, "loss": 2.1098, "step": 15803 }, { "epoch": 0.39, "learning_rate": 1.3952596888656346e-05, "loss": 2.1009, "step": 15804 }, { "epoch": 0.39, "learning_rate": 1.3951863553501059e-05, "loss": 2.299, "step": 15805 }, { "epoch": 0.39, "learning_rate": 1.3951130193159256e-05, "loss": 2.1358, "step": 15806 }, { "epoch": 0.39, "learning_rate": 1.3950396807635615e-05, "loss": 2.1338, "step": 15807 }, { "epoch": 0.39, "learning_rate": 1.3949663396934808e-05, "loss": 2.0463, "step": 15808 }, { "epoch": 0.39, "learning_rate": 1.3948929961061506e-05, "loss": 2.1838, "step": 15809 }, { "epoch": 0.39, "learning_rate": 1.3948196500020388e-05, "loss": 2.172, "step": 15810 }, { "epoch": 0.39, "learning_rate": 1.3947463013816128e-05, "loss": 1.9767, "step": 15811 }, { "epoch": 0.39, "learning_rate": 1.3946729502453398e-05, "loss": 2.0015, "step": 15812 }, { "epoch": 0.39, "learning_rate": 1.3945995965936874e-05, "loss": 2.0181, "step": 15813 }, { "epoch": 0.39, "learning_rate": 1.394526240427123e-05, "loss": 2.0773, "step": 15814 }, { "epoch": 0.39, "learning_rate": 1.394452881746115e-05, "loss": 2.0078, "step": 15815 }, { "epoch": 0.39, "learning_rate": 1.3943795205511298e-05, "loss": 2.0804, "step": 15816 }, { "epoch": 0.39, "learning_rate": 1.3943061568426356e-05, "loss": 2.0711, "step": 15817 }, { "epoch": 0.39, "learning_rate": 1.3942327906210993e-05, "loss": 1.8471, "step": 15818 }, { "epoch": 0.39, "learning_rate": 1.3941594218869894e-05, "loss": 2.1286, "step": 15819 }, { "epoch": 0.39, "learning_rate": 1.3940860506407729e-05, "loss": 2.1861, "step": 15820 }, { "epoch": 0.39, "learning_rate": 1.3940126768829174e-05, "loss": 2.0801, "step": 15821 }, { "epoch": 0.39, "learning_rate": 1.3939393006138908e-05, "loss": 2.2348, "step": 15822 }, { "epoch": 0.39, "learning_rate": 1.3938659218341608e-05, "loss": 2.0288, "step": 15823 }, { "epoch": 0.39, "learning_rate": 1.3937925405441948e-05, "loss": 2.1267, "step": 15824 }, { "epoch": 0.39, "learning_rate": 1.3937191567444608e-05, "loss": 2.0553, "step": 15825 }, { "epoch": 0.39, "learning_rate": 1.393645770435426e-05, "loss": 1.9485, "step": 15826 }, { "epoch": 0.39, "learning_rate": 1.3935723816175588e-05, "loss": 2.137, "step": 15827 }, { "epoch": 0.39, "learning_rate": 1.393498990291326e-05, "loss": 2.0232, "step": 15828 }, { "epoch": 0.39, "learning_rate": 1.3934255964571965e-05, "loss": 2.1516, "step": 15829 }, { "epoch": 0.39, "learning_rate": 1.3933522001156373e-05, "loss": 1.9777, "step": 15830 }, { "epoch": 0.39, "learning_rate": 1.393278801267116e-05, "loss": 1.9858, "step": 15831 }, { "epoch": 0.39, "learning_rate": 1.393205399912101e-05, "loss": 2.131, "step": 15832 }, { "epoch": 0.39, "learning_rate": 1.3931319960510597e-05, "loss": 2.1225, "step": 15833 }, { "epoch": 0.39, "learning_rate": 1.3930585896844603e-05, "loss": 1.9254, "step": 15834 }, { "epoch": 0.39, "learning_rate": 1.3929851808127704e-05, "loss": 2.2448, "step": 15835 }, { "epoch": 0.39, "learning_rate": 1.3929117694364576e-05, "loss": 2.1099, "step": 15836 }, { "epoch": 0.39, "learning_rate": 1.3928383555559903e-05, "loss": 2.009, "step": 15837 }, { "epoch": 0.39, "learning_rate": 1.3927649391718361e-05, "loss": 1.9636, "step": 15838 }, { "epoch": 0.39, "learning_rate": 1.392691520284463e-05, "loss": 2.2416, "step": 15839 }, { "epoch": 0.39, "learning_rate": 1.3926180988943383e-05, "loss": 2.182, "step": 15840 }, { "epoch": 0.39, "learning_rate": 1.392544675001931e-05, "loss": 2.0416, "step": 15841 }, { "epoch": 0.39, "learning_rate": 1.3924712486077086e-05, "loss": 2.174, "step": 15842 }, { "epoch": 0.39, "learning_rate": 1.3923978197121391e-05, "loss": 2.01, "step": 15843 }, { "epoch": 0.39, "learning_rate": 1.3923243883156904e-05, "loss": 2.043, "step": 15844 }, { "epoch": 0.39, "learning_rate": 1.3922509544188301e-05, "loss": 2.1844, "step": 15845 }, { "epoch": 0.39, "learning_rate": 1.3921775180220272e-05, "loss": 1.9585, "step": 15846 }, { "epoch": 0.39, "learning_rate": 1.3921040791257485e-05, "loss": 2.0334, "step": 15847 }, { "epoch": 0.39, "learning_rate": 1.3920306377304634e-05, "loss": 2.1642, "step": 15848 }, { "epoch": 0.39, "learning_rate": 1.3919571938366388e-05, "loss": 2.2003, "step": 15849 }, { "epoch": 0.39, "learning_rate": 1.3918837474447436e-05, "loss": 2.2159, "step": 15850 }, { "epoch": 0.39, "learning_rate": 1.3918102985552455e-05, "loss": 2.032, "step": 15851 }, { "epoch": 0.39, "learning_rate": 1.3917368471686123e-05, "loss": 1.9051, "step": 15852 }, { "epoch": 0.39, "learning_rate": 1.391663393285313e-05, "loss": 1.9389, "step": 15853 }, { "epoch": 0.39, "learning_rate": 1.391589936905815e-05, "loss": 2.1361, "step": 15854 }, { "epoch": 0.39, "learning_rate": 1.391516478030587e-05, "loss": 1.9761, "step": 15855 }, { "epoch": 0.39, "learning_rate": 1.3914430166600964e-05, "loss": 1.8679, "step": 15856 }, { "epoch": 0.39, "learning_rate": 1.3913695527948123e-05, "loss": 2.0851, "step": 15857 }, { "epoch": 0.39, "learning_rate": 1.3912960864352023e-05, "loss": 2.0021, "step": 15858 }, { "epoch": 0.39, "learning_rate": 1.391222617581735e-05, "loss": 2.1473, "step": 15859 }, { "epoch": 0.39, "learning_rate": 1.3911491462348783e-05, "loss": 2.2598, "step": 15860 }, { "epoch": 0.39, "learning_rate": 1.3910756723951008e-05, "loss": 2.011, "step": 15861 }, { "epoch": 0.39, "learning_rate": 1.3910021960628703e-05, "loss": 2.1021, "step": 15862 }, { "epoch": 0.39, "learning_rate": 1.3909287172386557e-05, "loss": 2.2932, "step": 15863 }, { "epoch": 0.39, "learning_rate": 1.3908552359229248e-05, "loss": 2.0611, "step": 15864 }, { "epoch": 0.39, "learning_rate": 1.390781752116146e-05, "loss": 2.1818, "step": 15865 }, { "epoch": 0.39, "learning_rate": 1.390708265818788e-05, "loss": 2.0781, "step": 15866 }, { "epoch": 0.39, "learning_rate": 1.3906347770313185e-05, "loss": 2.1182, "step": 15867 }, { "epoch": 0.39, "learning_rate": 1.3905612857542065e-05, "loss": 2.1887, "step": 15868 }, { "epoch": 0.39, "learning_rate": 1.3904877919879205e-05, "loss": 2.2011, "step": 15869 }, { "epoch": 0.39, "learning_rate": 1.3904142957329283e-05, "loss": 2.1126, "step": 15870 }, { "epoch": 0.39, "learning_rate": 1.3903407969896984e-05, "loss": 2.0493, "step": 15871 }, { "epoch": 0.39, "learning_rate": 1.3902672957586993e-05, "loss": 2.0723, "step": 15872 }, { "epoch": 0.39, "learning_rate": 1.3901937920403998e-05, "loss": 2.0738, "step": 15873 }, { "epoch": 0.39, "learning_rate": 1.390120285835268e-05, "loss": 2.088, "step": 15874 }, { "epoch": 0.39, "learning_rate": 1.3900467771437725e-05, "loss": 2.0286, "step": 15875 }, { "epoch": 0.39, "learning_rate": 1.3899732659663816e-05, "loss": 2.1348, "step": 15876 }, { "epoch": 0.39, "learning_rate": 1.3898997523035642e-05, "loss": 2.1925, "step": 15877 }, { "epoch": 0.39, "learning_rate": 1.3898262361557885e-05, "loss": 2.0818, "step": 15878 }, { "epoch": 0.39, "learning_rate": 1.3897527175235234e-05, "loss": 2.1377, "step": 15879 }, { "epoch": 0.39, "learning_rate": 1.3896791964072371e-05, "loss": 2.0709, "step": 15880 }, { "epoch": 0.39, "learning_rate": 1.3896056728073984e-05, "loss": 2.0943, "step": 15881 }, { "epoch": 0.39, "learning_rate": 1.3895321467244756e-05, "loss": 1.9674, "step": 15882 }, { "epoch": 0.39, "learning_rate": 1.3894586181589374e-05, "loss": 2.1514, "step": 15883 }, { "epoch": 0.39, "learning_rate": 1.3893850871112524e-05, "loss": 2.2465, "step": 15884 }, { "epoch": 0.39, "learning_rate": 1.3893115535818897e-05, "loss": 1.9366, "step": 15885 }, { "epoch": 0.39, "learning_rate": 1.3892380175713175e-05, "loss": 2.2893, "step": 15886 }, { "epoch": 0.39, "learning_rate": 1.3891644790800042e-05, "loss": 2.0629, "step": 15887 }, { "epoch": 0.39, "learning_rate": 1.3890909381084195e-05, "loss": 1.8974, "step": 15888 }, { "epoch": 0.39, "learning_rate": 1.389017394657031e-05, "loss": 2.1642, "step": 15889 }, { "epoch": 0.39, "learning_rate": 1.3889438487263082e-05, "loss": 2.14, "step": 15890 }, { "epoch": 0.39, "learning_rate": 1.388870300316719e-05, "loss": 2.1585, "step": 15891 }, { "epoch": 0.39, "learning_rate": 1.388796749428733e-05, "loss": 2.1487, "step": 15892 }, { "epoch": 0.39, "learning_rate": 1.3887231960628185e-05, "loss": 1.9822, "step": 15893 }, { "epoch": 0.39, "learning_rate": 1.3886496402194444e-05, "loss": 2.0261, "step": 15894 }, { "epoch": 0.39, "learning_rate": 1.3885760818990797e-05, "loss": 2.2456, "step": 15895 }, { "epoch": 0.39, "learning_rate": 1.3885025211021925e-05, "loss": 2.0196, "step": 15896 }, { "epoch": 0.39, "learning_rate": 1.3884289578292527e-05, "loss": 2.3556, "step": 15897 }, { "epoch": 0.39, "learning_rate": 1.3883553920807283e-05, "loss": 2.1543, "step": 15898 }, { "epoch": 0.39, "learning_rate": 1.3882818238570885e-05, "loss": 2.115, "step": 15899 }, { "epoch": 0.39, "learning_rate": 1.388208253158802e-05, "loss": 1.8479, "step": 15900 }, { "epoch": 0.39, "learning_rate": 1.3881346799863376e-05, "loss": 1.963, "step": 15901 }, { "epoch": 0.39, "learning_rate": 1.3880611043401645e-05, "loss": 2.1912, "step": 15902 }, { "epoch": 0.39, "learning_rate": 1.3879875262207518e-05, "loss": 2.092, "step": 15903 }, { "epoch": 0.39, "learning_rate": 1.387913945628568e-05, "loss": 2.0325, "step": 15904 }, { "epoch": 0.39, "learning_rate": 1.3878403625640821e-05, "loss": 1.9439, "step": 15905 }, { "epoch": 0.39, "learning_rate": 1.3877667770277633e-05, "loss": 2.2334, "step": 15906 }, { "epoch": 0.39, "learning_rate": 1.3876931890200805e-05, "loss": 1.9958, "step": 15907 }, { "epoch": 0.39, "learning_rate": 1.3876195985415026e-05, "loss": 1.9584, "step": 15908 }, { "epoch": 0.39, "learning_rate": 1.3875460055924985e-05, "loss": 2.1019, "step": 15909 }, { "epoch": 0.39, "learning_rate": 1.3874724101735376e-05, "loss": 2.1401, "step": 15910 }, { "epoch": 0.39, "learning_rate": 1.3873988122850886e-05, "loss": 2.1093, "step": 15911 }, { "epoch": 0.39, "learning_rate": 1.387325211927621e-05, "loss": 2.2059, "step": 15912 }, { "epoch": 0.39, "learning_rate": 1.3872516091016035e-05, "loss": 1.8393, "step": 15913 }, { "epoch": 0.39, "learning_rate": 1.3871780038075051e-05, "loss": 2.0021, "step": 15914 }, { "epoch": 0.39, "learning_rate": 1.3871043960457953e-05, "loss": 2.057, "step": 15915 }, { "epoch": 0.39, "learning_rate": 1.3870307858169429e-05, "loss": 1.9475, "step": 15916 }, { "epoch": 0.39, "learning_rate": 1.3869571731214172e-05, "loss": 2.0383, "step": 15917 }, { "epoch": 0.39, "learning_rate": 1.3868835579596872e-05, "loss": 2.0684, "step": 15918 }, { "epoch": 0.39, "learning_rate": 1.3868099403322226e-05, "loss": 2.0654, "step": 15919 }, { "epoch": 0.39, "learning_rate": 1.3867363202394919e-05, "loss": 2.1789, "step": 15920 }, { "epoch": 0.39, "learning_rate": 1.3866626976819647e-05, "loss": 2.0637, "step": 15921 }, { "epoch": 0.39, "learning_rate": 1.3865890726601101e-05, "loss": 1.9872, "step": 15922 }, { "epoch": 0.39, "learning_rate": 1.3865154451743973e-05, "loss": 1.9282, "step": 15923 }, { "epoch": 0.39, "learning_rate": 1.386441815225296e-05, "loss": 2.105, "step": 15924 }, { "epoch": 0.39, "learning_rate": 1.3863681828132744e-05, "loss": 2.0454, "step": 15925 }, { "epoch": 0.39, "learning_rate": 1.3862945479388028e-05, "loss": 2.0664, "step": 15926 }, { "epoch": 0.39, "learning_rate": 1.38622091060235e-05, "loss": 1.9945, "step": 15927 }, { "epoch": 0.39, "learning_rate": 1.3861472708043858e-05, "loss": 2.0846, "step": 15928 }, { "epoch": 0.39, "learning_rate": 1.386073628545379e-05, "loss": 2.0475, "step": 15929 }, { "epoch": 0.39, "learning_rate": 1.385999983825799e-05, "loss": 1.9066, "step": 15930 }, { "epoch": 0.39, "learning_rate": 1.3859263366461156e-05, "loss": 2.11, "step": 15931 }, { "epoch": 0.39, "learning_rate": 1.3858526870067977e-05, "loss": 2.0288, "step": 15932 }, { "epoch": 0.39, "learning_rate": 1.3857790349083149e-05, "loss": 2.0238, "step": 15933 }, { "epoch": 0.39, "learning_rate": 1.3857053803511367e-05, "loss": 2.2695, "step": 15934 }, { "epoch": 0.39, "learning_rate": 1.3856317233357324e-05, "loss": 2.0424, "step": 15935 }, { "epoch": 0.39, "learning_rate": 1.385558063862571e-05, "loss": 2.2777, "step": 15936 }, { "epoch": 0.39, "learning_rate": 1.385484401932123e-05, "loss": 1.8807, "step": 15937 }, { "epoch": 0.39, "learning_rate": 1.3854107375448571e-05, "loss": 2.0049, "step": 15938 }, { "epoch": 0.39, "learning_rate": 1.3853370707012428e-05, "loss": 2.0385, "step": 15939 }, { "epoch": 0.39, "learning_rate": 1.3852634014017499e-05, "loss": 1.9796, "step": 15940 }, { "epoch": 0.39, "learning_rate": 1.385189729646848e-05, "loss": 2.085, "step": 15941 }, { "epoch": 0.39, "learning_rate": 1.3851160554370063e-05, "loss": 2.0057, "step": 15942 }, { "epoch": 0.39, "learning_rate": 1.3850423787726943e-05, "loss": 2.0883, "step": 15943 }, { "epoch": 0.39, "learning_rate": 1.384968699654382e-05, "loss": 1.9645, "step": 15944 }, { "epoch": 0.39, "learning_rate": 1.3848950180825382e-05, "loss": 2.1679, "step": 15945 }, { "epoch": 0.39, "learning_rate": 1.3848213340576333e-05, "loss": 2.0759, "step": 15946 }, { "epoch": 0.39, "learning_rate": 1.384747647580137e-05, "loss": 2.1584, "step": 15947 }, { "epoch": 0.39, "learning_rate": 1.3846739586505182e-05, "loss": 2.1701, "step": 15948 }, { "epoch": 0.39, "learning_rate": 1.3846002672692467e-05, "loss": 2.0322, "step": 15949 }, { "epoch": 0.39, "learning_rate": 1.3845265734367928e-05, "loss": 1.9341, "step": 15950 }, { "epoch": 0.39, "learning_rate": 1.3844528771536256e-05, "loss": 2.0839, "step": 15951 }, { "epoch": 0.39, "learning_rate": 1.3843791784202149e-05, "loss": 2.1003, "step": 15952 }, { "epoch": 0.39, "learning_rate": 1.3843054772370304e-05, "loss": 2.1111, "step": 15953 }, { "epoch": 0.39, "learning_rate": 1.3842317736045417e-05, "loss": 2.0394, "step": 15954 }, { "epoch": 0.39, "learning_rate": 1.3841580675232189e-05, "loss": 2.0894, "step": 15955 }, { "epoch": 0.39, "learning_rate": 1.3840843589935316e-05, "loss": 2.2528, "step": 15956 }, { "epoch": 0.39, "learning_rate": 1.3840106480159494e-05, "loss": 2.237, "step": 15957 }, { "epoch": 0.39, "learning_rate": 1.3839369345909421e-05, "loss": 2.1667, "step": 15958 }, { "epoch": 0.39, "learning_rate": 1.3838632187189796e-05, "loss": 2.1662, "step": 15959 }, { "epoch": 0.39, "learning_rate": 1.3837895004005319e-05, "loss": 2.0748, "step": 15960 }, { "epoch": 0.39, "learning_rate": 1.3837157796360684e-05, "loss": 2.0653, "step": 15961 }, { "epoch": 0.39, "learning_rate": 1.3836420564260593e-05, "loss": 1.96, "step": 15962 }, { "epoch": 0.39, "learning_rate": 1.3835683307709742e-05, "loss": 2.0836, "step": 15963 }, { "epoch": 0.39, "learning_rate": 1.3834946026712835e-05, "loss": 2.2377, "step": 15964 }, { "epoch": 0.39, "learning_rate": 1.3834208721274564e-05, "loss": 2.1702, "step": 15965 }, { "epoch": 0.39, "learning_rate": 1.3833471391399633e-05, "loss": 2.0885, "step": 15966 }, { "epoch": 0.39, "learning_rate": 1.383273403709274e-05, "loss": 2.0638, "step": 15967 }, { "epoch": 0.39, "learning_rate": 1.3831996658358582e-05, "loss": 2.205, "step": 15968 }, { "epoch": 0.39, "learning_rate": 1.383125925520186e-05, "loss": 2.0988, "step": 15969 }, { "epoch": 0.39, "learning_rate": 1.3830521827627278e-05, "loss": 2.1177, "step": 15970 }, { "epoch": 0.39, "learning_rate": 1.3829784375639528e-05, "loss": 2.0515, "step": 15971 }, { "epoch": 0.39, "learning_rate": 1.3829046899243316e-05, "loss": 2.1088, "step": 15972 }, { "epoch": 0.39, "learning_rate": 1.382830939844334e-05, "loss": 2.0243, "step": 15973 }, { "epoch": 0.39, "learning_rate": 1.38275718732443e-05, "loss": 2.0786, "step": 15974 }, { "epoch": 0.39, "learning_rate": 1.3826834323650899e-05, "loss": 1.9764, "step": 15975 }, { "epoch": 0.39, "learning_rate": 1.3826096749667835e-05, "loss": 2.0648, "step": 15976 }, { "epoch": 0.39, "learning_rate": 1.3825359151299811e-05, "loss": 2.071, "step": 15977 }, { "epoch": 0.39, "learning_rate": 1.382462152855152e-05, "loss": 2.0775, "step": 15978 }, { "epoch": 0.39, "learning_rate": 1.3823883881427676e-05, "loss": 2.1062, "step": 15979 }, { "epoch": 0.39, "learning_rate": 1.3823146209932971e-05, "loss": 1.8747, "step": 15980 }, { "epoch": 0.39, "learning_rate": 1.382240851407211e-05, "loss": 1.8944, "step": 15981 }, { "epoch": 0.39, "learning_rate": 1.3821670793849795e-05, "loss": 2.209, "step": 15982 }, { "epoch": 0.39, "learning_rate": 1.3820933049270724e-05, "loss": 2.0601, "step": 15983 }, { "epoch": 0.39, "learning_rate": 1.3820195280339603e-05, "loss": 2.301, "step": 15984 }, { "epoch": 0.39, "learning_rate": 1.3819457487061134e-05, "loss": 2.0954, "step": 15985 }, { "epoch": 0.39, "learning_rate": 1.3818719669440017e-05, "loss": 1.9737, "step": 15986 }, { "epoch": 0.39, "learning_rate": 1.3817981827480955e-05, "loss": 2.1189, "step": 15987 }, { "epoch": 0.39, "learning_rate": 1.381724396118865e-05, "loss": 1.9658, "step": 15988 }, { "epoch": 0.39, "learning_rate": 1.3816506070567801e-05, "loss": 1.9754, "step": 15989 }, { "epoch": 0.39, "learning_rate": 1.3815768155623122e-05, "loss": 1.9964, "step": 15990 }, { "epoch": 0.39, "learning_rate": 1.3815030216359306e-05, "loss": 2.1697, "step": 15991 }, { "epoch": 0.39, "learning_rate": 1.3814292252781058e-05, "loss": 1.8433, "step": 15992 }, { "epoch": 0.39, "learning_rate": 1.3813554264893085e-05, "loss": 1.966, "step": 15993 }, { "epoch": 0.39, "learning_rate": 1.3812816252700087e-05, "loss": 2.0339, "step": 15994 }, { "epoch": 0.39, "learning_rate": 1.3812078216206768e-05, "loss": 1.9919, "step": 15995 }, { "epoch": 0.39, "learning_rate": 1.3811340155417834e-05, "loss": 2.1175, "step": 15996 }, { "epoch": 0.39, "learning_rate": 1.3810602070337984e-05, "loss": 2.1487, "step": 15997 }, { "epoch": 0.39, "learning_rate": 1.3809863960971925e-05, "loss": 1.9453, "step": 15998 }, { "epoch": 0.39, "learning_rate": 1.3809125827324364e-05, "loss": 2.1647, "step": 15999 }, { "epoch": 0.39, "learning_rate": 1.38083876694e-05, "loss": 2.1527, "step": 16000 }, { "epoch": 0.39, "learning_rate": 1.3807649487203545e-05, "loss": 2.0898, "step": 16001 }, { "epoch": 0.39, "learning_rate": 1.3806911280739694e-05, "loss": 2.0036, "step": 16002 }, { "epoch": 0.39, "learning_rate": 1.3806173050013158e-05, "loss": 2.0313, "step": 16003 }, { "epoch": 0.39, "learning_rate": 1.3805434795028643e-05, "loss": 2.0118, "step": 16004 }, { "epoch": 0.39, "learning_rate": 1.3804696515790851e-05, "loss": 2.1883, "step": 16005 }, { "epoch": 0.39, "learning_rate": 1.3803958212304487e-05, "loss": 2.1095, "step": 16006 }, { "epoch": 0.39, "learning_rate": 1.3803219884574258e-05, "loss": 2.0619, "step": 16007 }, { "epoch": 0.39, "learning_rate": 1.3802481532604868e-05, "loss": 1.9561, "step": 16008 }, { "epoch": 0.39, "learning_rate": 1.3801743156401027e-05, "loss": 2.1441, "step": 16009 }, { "epoch": 0.39, "learning_rate": 1.3801004755967435e-05, "loss": 1.9503, "step": 16010 }, { "epoch": 0.39, "learning_rate": 1.3800266331308802e-05, "loss": 2.1646, "step": 16011 }, { "epoch": 0.39, "learning_rate": 1.3799527882429834e-05, "loss": 1.9732, "step": 16012 }, { "epoch": 0.39, "learning_rate": 1.3798789409335235e-05, "loss": 2.1374, "step": 16013 }, { "epoch": 0.39, "learning_rate": 1.3798050912029716e-05, "loss": 1.975, "step": 16014 }, { "epoch": 0.39, "learning_rate": 1.3797312390517978e-05, "loss": 2.2261, "step": 16015 }, { "epoch": 0.39, "learning_rate": 1.3796573844804733e-05, "loss": 2.1891, "step": 16016 }, { "epoch": 0.39, "learning_rate": 1.3795835274894682e-05, "loss": 2.2032, "step": 16017 }, { "epoch": 0.39, "learning_rate": 1.3795096680792538e-05, "loss": 2.0639, "step": 16018 }, { "epoch": 0.39, "learning_rate": 1.3794358062503008e-05, "loss": 1.986, "step": 16019 }, { "epoch": 0.39, "learning_rate": 1.3793619420030796e-05, "loss": 2.0624, "step": 16020 }, { "epoch": 0.39, "learning_rate": 1.3792880753380609e-05, "loss": 2.1254, "step": 16021 }, { "epoch": 0.39, "learning_rate": 1.3792142062557159e-05, "loss": 2.0219, "step": 16022 }, { "epoch": 0.39, "learning_rate": 1.3791403347565152e-05, "loss": 2.1505, "step": 16023 }, { "epoch": 0.39, "learning_rate": 1.3790664608409295e-05, "loss": 2.0097, "step": 16024 }, { "epoch": 0.39, "learning_rate": 1.3789925845094297e-05, "loss": 2.291, "step": 16025 }, { "epoch": 0.4, "learning_rate": 1.3789187057624864e-05, "loss": 2.1456, "step": 16026 }, { "epoch": 0.4, "learning_rate": 1.3788448246005713e-05, "loss": 1.951, "step": 16027 }, { "epoch": 0.4, "learning_rate": 1.3787709410241543e-05, "loss": 2.1845, "step": 16028 }, { "epoch": 0.4, "learning_rate": 1.3786970550337069e-05, "loss": 2.2994, "step": 16029 }, { "epoch": 0.4, "learning_rate": 1.3786231666296996e-05, "loss": 2.2046, "step": 16030 }, { "epoch": 0.4, "learning_rate": 1.3785492758126031e-05, "loss": 1.9813, "step": 16031 }, { "epoch": 0.4, "learning_rate": 1.3784753825828894e-05, "loss": 2.0697, "step": 16032 }, { "epoch": 0.4, "learning_rate": 1.3784014869410281e-05, "loss": 1.9671, "step": 16033 }, { "epoch": 0.4, "learning_rate": 1.3783275888874912e-05, "loss": 2.0092, "step": 16034 }, { "epoch": 0.4, "learning_rate": 1.3782536884227491e-05, "loss": 1.9966, "step": 16035 }, { "epoch": 0.4, "learning_rate": 1.378179785547273e-05, "loss": 1.941, "step": 16036 }, { "epoch": 0.4, "learning_rate": 1.3781058802615338e-05, "loss": 2.1716, "step": 16037 }, { "epoch": 0.4, "learning_rate": 1.378031972566003e-05, "loss": 2.1579, "step": 16038 }, { "epoch": 0.4, "learning_rate": 1.377958062461151e-05, "loss": 2.183, "step": 16039 }, { "epoch": 0.4, "learning_rate": 1.3778841499474492e-05, "loss": 2.1129, "step": 16040 }, { "epoch": 0.4, "learning_rate": 1.3778102350253684e-05, "loss": 1.9302, "step": 16041 }, { "epoch": 0.4, "learning_rate": 1.3777363176953795e-05, "loss": 2.1645, "step": 16042 }, { "epoch": 0.4, "learning_rate": 1.3776623979579544e-05, "loss": 2.1736, "step": 16043 }, { "epoch": 0.4, "learning_rate": 1.3775884758135638e-05, "loss": 2.3005, "step": 16044 }, { "epoch": 0.4, "learning_rate": 1.3775145512626789e-05, "loss": 2.0182, "step": 16045 }, { "epoch": 0.4, "learning_rate": 1.3774406243057703e-05, "loss": 2.0036, "step": 16046 }, { "epoch": 0.4, "learning_rate": 1.3773666949433098e-05, "loss": 2.0205, "step": 16047 }, { "epoch": 0.4, "learning_rate": 1.3772927631757684e-05, "loss": 1.8909, "step": 16048 }, { "epoch": 0.4, "learning_rate": 1.3772188290036174e-05, "loss": 2.1425, "step": 16049 }, { "epoch": 0.4, "learning_rate": 1.3771448924273277e-05, "loss": 1.9353, "step": 16050 }, { "epoch": 0.4, "learning_rate": 1.3770709534473705e-05, "loss": 2.1733, "step": 16051 }, { "epoch": 0.4, "learning_rate": 1.3769970120642176e-05, "loss": 2.0286, "step": 16052 }, { "epoch": 0.4, "learning_rate": 1.3769230682783397e-05, "loss": 2.1455, "step": 16053 }, { "epoch": 0.4, "learning_rate": 1.3768491220902082e-05, "loss": 2.1066, "step": 16054 }, { "epoch": 0.4, "learning_rate": 1.3767751735002945e-05, "loss": 2.0925, "step": 16055 }, { "epoch": 0.4, "learning_rate": 1.37670122250907e-05, "loss": 2.1432, "step": 16056 }, { "epoch": 0.4, "learning_rate": 1.3766272691170058e-05, "loss": 2.1045, "step": 16057 }, { "epoch": 0.4, "learning_rate": 1.376553313324573e-05, "loss": 2.0082, "step": 16058 }, { "epoch": 0.4, "learning_rate": 1.3764793551322432e-05, "loss": 2.2934, "step": 16059 }, { "epoch": 0.4, "learning_rate": 1.3764053945404878e-05, "loss": 2.1467, "step": 16060 }, { "epoch": 0.4, "learning_rate": 1.3763314315497785e-05, "loss": 2.1278, "step": 16061 }, { "epoch": 0.4, "learning_rate": 1.3762574661605861e-05, "loss": 2.1165, "step": 16062 }, { "epoch": 0.4, "learning_rate": 1.3761834983733823e-05, "loss": 1.9047, "step": 16063 }, { "epoch": 0.4, "learning_rate": 1.3761095281886387e-05, "loss": 1.9514, "step": 16064 }, { "epoch": 0.4, "learning_rate": 1.3760355556068261e-05, "loss": 1.8854, "step": 16065 }, { "epoch": 0.4, "learning_rate": 1.3759615806284166e-05, "loss": 1.8213, "step": 16066 }, { "epoch": 0.4, "learning_rate": 1.3758876032538815e-05, "loss": 1.9754, "step": 16067 }, { "epoch": 0.4, "learning_rate": 1.3758136234836918e-05, "loss": 2.0261, "step": 16068 }, { "epoch": 0.4, "learning_rate": 1.3757396413183197e-05, "loss": 2.0727, "step": 16069 }, { "epoch": 0.4, "learning_rate": 1.3756656567582364e-05, "loss": 1.9232, "step": 16070 }, { "epoch": 0.4, "learning_rate": 1.3755916698039134e-05, "loss": 2.0773, "step": 16071 }, { "epoch": 0.4, "learning_rate": 1.3755176804558224e-05, "loss": 2.114, "step": 16072 }, { "epoch": 0.4, "learning_rate": 1.3754436887144348e-05, "loss": 2.1774, "step": 16073 }, { "epoch": 0.4, "learning_rate": 1.3753696945802224e-05, "loss": 2.0695, "step": 16074 }, { "epoch": 0.4, "learning_rate": 1.3752956980536563e-05, "loss": 2.0854, "step": 16075 }, { "epoch": 0.4, "learning_rate": 1.3752216991352083e-05, "loss": 2.2958, "step": 16076 }, { "epoch": 0.4, "learning_rate": 1.3751476978253504e-05, "loss": 1.9983, "step": 16077 }, { "epoch": 0.4, "learning_rate": 1.375073694124554e-05, "loss": 2.1419, "step": 16078 }, { "epoch": 0.4, "learning_rate": 1.3749996880332904e-05, "loss": 2.1124, "step": 16079 }, { "epoch": 0.4, "learning_rate": 1.3749256795520319e-05, "loss": 2.1675, "step": 16080 }, { "epoch": 0.4, "learning_rate": 1.3748516686812497e-05, "loss": 2.192, "step": 16081 }, { "epoch": 0.4, "learning_rate": 1.3747776554214159e-05, "loss": 1.9861, "step": 16082 }, { "epoch": 0.4, "learning_rate": 1.3747036397730017e-05, "loss": 1.9428, "step": 16083 }, { "epoch": 0.4, "learning_rate": 1.374629621736479e-05, "loss": 2.1271, "step": 16084 }, { "epoch": 0.4, "learning_rate": 1.37455560131232e-05, "loss": 2.0758, "step": 16085 }, { "epoch": 0.4, "learning_rate": 1.3744815785009954e-05, "loss": 2.066, "step": 16086 }, { "epoch": 0.4, "learning_rate": 1.3744075533029783e-05, "loss": 2.025, "step": 16087 }, { "epoch": 0.4, "learning_rate": 1.3743335257187397e-05, "loss": 1.9619, "step": 16088 }, { "epoch": 0.4, "learning_rate": 1.3742594957487515e-05, "loss": 1.9506, "step": 16089 }, { "epoch": 0.4, "learning_rate": 1.3741854633934857e-05, "loss": 1.973, "step": 16090 }, { "epoch": 0.4, "learning_rate": 1.374111428653414e-05, "loss": 2.1959, "step": 16091 }, { "epoch": 0.4, "learning_rate": 1.3740373915290082e-05, "loss": 2.0379, "step": 16092 }, { "epoch": 0.4, "learning_rate": 1.37396335202074e-05, "loss": 2.1492, "step": 16093 }, { "epoch": 0.4, "learning_rate": 1.3738893101290818e-05, "loss": 2.092, "step": 16094 }, { "epoch": 0.4, "learning_rate": 1.3738152658545047e-05, "loss": 2.0682, "step": 16095 }, { "epoch": 0.4, "learning_rate": 1.3737412191974814e-05, "loss": 2.0734, "step": 16096 }, { "epoch": 0.4, "learning_rate": 1.3736671701584838e-05, "loss": 2.2084, "step": 16097 }, { "epoch": 0.4, "learning_rate": 1.3735931187379834e-05, "loss": 1.9537, "step": 16098 }, { "epoch": 0.4, "learning_rate": 1.3735190649364519e-05, "loss": 2.0635, "step": 16099 }, { "epoch": 0.4, "learning_rate": 1.373445008754362e-05, "loss": 2.08, "step": 16100 }, { "epoch": 0.4, "learning_rate": 1.3733709501921855e-05, "loss": 1.9403, "step": 16101 }, { "epoch": 0.4, "learning_rate": 1.3732968892503937e-05, "loss": 2.0681, "step": 16102 }, { "epoch": 0.4, "learning_rate": 1.3732228259294597e-05, "loss": 2.07, "step": 16103 }, { "epoch": 0.4, "learning_rate": 1.3731487602298547e-05, "loss": 2.1288, "step": 16104 }, { "epoch": 0.4, "learning_rate": 1.3730746921520513e-05, "loss": 2.028, "step": 16105 }, { "epoch": 0.4, "learning_rate": 1.3730006216965213e-05, "loss": 2.1816, "step": 16106 }, { "epoch": 0.4, "learning_rate": 1.3729265488637366e-05, "loss": 2.2074, "step": 16107 }, { "epoch": 0.4, "learning_rate": 1.3728524736541697e-05, "loss": 1.8462, "step": 16108 }, { "epoch": 0.4, "learning_rate": 1.3727783960682922e-05, "loss": 2.2859, "step": 16109 }, { "epoch": 0.4, "learning_rate": 1.3727043161065765e-05, "loss": 2.1934, "step": 16110 }, { "epoch": 0.4, "learning_rate": 1.3726302337694948e-05, "loss": 2.16, "step": 16111 }, { "epoch": 0.4, "learning_rate": 1.3725561490575192e-05, "loss": 2.0672, "step": 16112 }, { "epoch": 0.4, "learning_rate": 1.372482061971122e-05, "loss": 2.0109, "step": 16113 }, { "epoch": 0.4, "learning_rate": 1.372407972510775e-05, "loss": 2.0418, "step": 16114 }, { "epoch": 0.4, "learning_rate": 1.3723338806769509e-05, "loss": 2.0093, "step": 16115 }, { "epoch": 0.4, "learning_rate": 1.3722597864701214e-05, "loss": 2.1384, "step": 16116 }, { "epoch": 0.4, "learning_rate": 1.372185689890759e-05, "loss": 2.0597, "step": 16117 }, { "epoch": 0.4, "learning_rate": 1.3721115909393359e-05, "loss": 2.1926, "step": 16118 }, { "epoch": 0.4, "learning_rate": 1.3720374896163244e-05, "loss": 2.0947, "step": 16119 }, { "epoch": 0.4, "learning_rate": 1.3719633859221965e-05, "loss": 2.1343, "step": 16120 }, { "epoch": 0.4, "learning_rate": 1.3718892798574253e-05, "loss": 2.0808, "step": 16121 }, { "epoch": 0.4, "learning_rate": 1.3718151714224821e-05, "loss": 2.156, "step": 16122 }, { "epoch": 0.4, "learning_rate": 1.3717410606178396e-05, "loss": 2.0416, "step": 16123 }, { "epoch": 0.4, "learning_rate": 1.3716669474439704e-05, "loss": 2.0248, "step": 16124 }, { "epoch": 0.4, "learning_rate": 1.3715928319013465e-05, "loss": 1.9705, "step": 16125 }, { "epoch": 0.4, "learning_rate": 1.3715187139904406e-05, "loss": 2.0982, "step": 16126 }, { "epoch": 0.4, "learning_rate": 1.3714445937117246e-05, "loss": 2.158, "step": 16127 }, { "epoch": 0.4, "learning_rate": 1.3713704710656712e-05, "loss": 1.9602, "step": 16128 }, { "epoch": 0.4, "learning_rate": 1.3712963460527529e-05, "loss": 1.9985, "step": 16129 }, { "epoch": 0.4, "learning_rate": 1.3712222186734416e-05, "loss": 2.0425, "step": 16130 }, { "epoch": 0.4, "learning_rate": 1.3711480889282105e-05, "loss": 2.1734, "step": 16131 }, { "epoch": 0.4, "learning_rate": 1.3710739568175316e-05, "loss": 2.1093, "step": 16132 }, { "epoch": 0.4, "learning_rate": 1.3709998223418772e-05, "loss": 2.081, "step": 16133 }, { "epoch": 0.4, "learning_rate": 1.3709256855017204e-05, "loss": 2.0029, "step": 16134 }, { "epoch": 0.4, "learning_rate": 1.3708515462975331e-05, "loss": 2.0095, "step": 16135 }, { "epoch": 0.4, "learning_rate": 1.3707774047297884e-05, "loss": 2.0072, "step": 16136 }, { "epoch": 0.4, "learning_rate": 1.3707032607989579e-05, "loss": 2.2462, "step": 16137 }, { "epoch": 0.4, "learning_rate": 1.3706291145055151e-05, "loss": 2.2261, "step": 16138 }, { "epoch": 0.4, "learning_rate": 1.3705549658499317e-05, "loss": 2.2208, "step": 16139 }, { "epoch": 0.4, "learning_rate": 1.370480814832681e-05, "loss": 2.0785, "step": 16140 }, { "epoch": 0.4, "learning_rate": 1.3704066614542358e-05, "loss": 2.0295, "step": 16141 }, { "epoch": 0.4, "learning_rate": 1.3703325057150676e-05, "loss": 2.0719, "step": 16142 }, { "epoch": 0.4, "learning_rate": 1.3702583476156498e-05, "loss": 2.1067, "step": 16143 }, { "epoch": 0.4, "learning_rate": 1.3701841871564555e-05, "loss": 1.9876, "step": 16144 }, { "epoch": 0.4, "learning_rate": 1.370110024337956e-05, "loss": 1.9637, "step": 16145 }, { "epoch": 0.4, "learning_rate": 1.3700358591606249e-05, "loss": 2.0399, "step": 16146 }, { "epoch": 0.4, "learning_rate": 1.3699616916249348e-05, "loss": 2.0099, "step": 16147 }, { "epoch": 0.4, "learning_rate": 1.369887521731358e-05, "loss": 2.3323, "step": 16148 }, { "epoch": 0.4, "learning_rate": 1.3698133494803677e-05, "loss": 2.0419, "step": 16149 }, { "epoch": 0.4, "learning_rate": 1.3697391748724364e-05, "loss": 2.0425, "step": 16150 }, { "epoch": 0.4, "learning_rate": 1.369664997908037e-05, "loss": 2.0978, "step": 16151 }, { "epoch": 0.4, "learning_rate": 1.3695908185876416e-05, "loss": 2.04, "step": 16152 }, { "epoch": 0.4, "learning_rate": 1.3695166369117239e-05, "loss": 1.9769, "step": 16153 }, { "epoch": 0.4, "learning_rate": 1.3694424528807561e-05, "loss": 2.0159, "step": 16154 }, { "epoch": 0.4, "learning_rate": 1.369368266495211e-05, "loss": 2.1394, "step": 16155 }, { "epoch": 0.4, "learning_rate": 1.3692940777555619e-05, "loss": 2.1407, "step": 16156 }, { "epoch": 0.4, "learning_rate": 1.369219886662281e-05, "loss": 2.0004, "step": 16157 }, { "epoch": 0.4, "learning_rate": 1.3691456932158416e-05, "loss": 2.2294, "step": 16158 }, { "epoch": 0.4, "learning_rate": 1.3690714974167166e-05, "loss": 2.0388, "step": 16159 }, { "epoch": 0.4, "learning_rate": 1.3689972992653784e-05, "loss": 2.1998, "step": 16160 }, { "epoch": 0.4, "learning_rate": 1.3689230987623003e-05, "loss": 2.0462, "step": 16161 }, { "epoch": 0.4, "learning_rate": 1.3688488959079546e-05, "loss": 2.1269, "step": 16162 }, { "epoch": 0.4, "learning_rate": 1.3687746907028152e-05, "loss": 2.1265, "step": 16163 }, { "epoch": 0.4, "learning_rate": 1.3687004831473542e-05, "loss": 1.9991, "step": 16164 }, { "epoch": 0.4, "learning_rate": 1.368626273242045e-05, "loss": 2.0569, "step": 16165 }, { "epoch": 0.4, "learning_rate": 1.3685520609873606e-05, "loss": 2.0021, "step": 16166 }, { "epoch": 0.4, "learning_rate": 1.3684778463837735e-05, "loss": 2.1042, "step": 16167 }, { "epoch": 0.4, "learning_rate": 1.3684036294317572e-05, "loss": 2.1953, "step": 16168 }, { "epoch": 0.4, "learning_rate": 1.3683294101317846e-05, "loss": 1.9811, "step": 16169 }, { "epoch": 0.4, "learning_rate": 1.3682551884843288e-05, "loss": 2.0002, "step": 16170 }, { "epoch": 0.4, "learning_rate": 1.368180964489862e-05, "loss": 2.1783, "step": 16171 }, { "epoch": 0.4, "learning_rate": 1.3681067381488584e-05, "loss": 2.1941, "step": 16172 }, { "epoch": 0.4, "learning_rate": 1.3680325094617904e-05, "loss": 2.0453, "step": 16173 }, { "epoch": 0.4, "learning_rate": 1.3679582784291314e-05, "loss": 2.0488, "step": 16174 }, { "epoch": 0.4, "learning_rate": 1.3678840450513544e-05, "loss": 2.0435, "step": 16175 }, { "epoch": 0.4, "learning_rate": 1.3678098093289322e-05, "loss": 1.9723, "step": 16176 }, { "epoch": 0.4, "learning_rate": 1.3677355712623387e-05, "loss": 2.121, "step": 16177 }, { "epoch": 0.4, "learning_rate": 1.3676613308520463e-05, "loss": 2.0562, "step": 16178 }, { "epoch": 0.4, "learning_rate": 1.3675870880985284e-05, "loss": 2.1402, "step": 16179 }, { "epoch": 0.4, "learning_rate": 1.3675128430022583e-05, "loss": 2.2376, "step": 16180 }, { "epoch": 0.4, "learning_rate": 1.367438595563709e-05, "loss": 2.2191, "step": 16181 }, { "epoch": 0.4, "learning_rate": 1.3673643457833536e-05, "loss": 1.9917, "step": 16182 }, { "epoch": 0.4, "learning_rate": 1.3672900936616657e-05, "loss": 2.1548, "step": 16183 }, { "epoch": 0.4, "learning_rate": 1.3672158391991186e-05, "loss": 1.992, "step": 16184 }, { "epoch": 0.4, "learning_rate": 1.367141582396185e-05, "loss": 1.9809, "step": 16185 }, { "epoch": 0.4, "learning_rate": 1.3670673232533384e-05, "loss": 2.1049, "step": 16186 }, { "epoch": 0.4, "learning_rate": 1.3669930617710523e-05, "loss": 2.1641, "step": 16187 }, { "epoch": 0.4, "learning_rate": 1.3669187979497998e-05, "loss": 2.1088, "step": 16188 }, { "epoch": 0.4, "learning_rate": 1.3668445317900543e-05, "loss": 2.2049, "step": 16189 }, { "epoch": 0.4, "learning_rate": 1.3667702632922886e-05, "loss": 2.0846, "step": 16190 }, { "epoch": 0.4, "learning_rate": 1.3666959924569767e-05, "loss": 2.0518, "step": 16191 }, { "epoch": 0.4, "learning_rate": 1.3666217192845918e-05, "loss": 2.1298, "step": 16192 }, { "epoch": 0.4, "learning_rate": 1.3665474437756072e-05, "loss": 2.1943, "step": 16193 }, { "epoch": 0.4, "learning_rate": 1.3664731659304967e-05, "loss": 2.148, "step": 16194 }, { "epoch": 0.4, "learning_rate": 1.3663988857497328e-05, "loss": 2.1267, "step": 16195 }, { "epoch": 0.4, "learning_rate": 1.3663246032337894e-05, "loss": 2.0975, "step": 16196 }, { "epoch": 0.4, "learning_rate": 1.36625031838314e-05, "loss": 1.963, "step": 16197 }, { "epoch": 0.4, "learning_rate": 1.3661760311982581e-05, "loss": 2.091, "step": 16198 }, { "epoch": 0.4, "learning_rate": 1.3661017416796167e-05, "loss": 2.1165, "step": 16199 }, { "epoch": 0.4, "learning_rate": 1.3660274498276898e-05, "loss": 1.9892, "step": 16200 }, { "epoch": 0.4, "learning_rate": 1.3659531556429506e-05, "loss": 2.312, "step": 16201 }, { "epoch": 0.4, "learning_rate": 1.3658788591258727e-05, "loss": 2.1155, "step": 16202 }, { "epoch": 0.4, "learning_rate": 1.3658045602769296e-05, "loss": 2.0648, "step": 16203 }, { "epoch": 0.4, "learning_rate": 1.365730259096595e-05, "loss": 2.117, "step": 16204 }, { "epoch": 0.4, "learning_rate": 1.3656559555853422e-05, "loss": 2.144, "step": 16205 }, { "epoch": 0.4, "learning_rate": 1.3655816497436447e-05, "loss": 2.0466, "step": 16206 }, { "epoch": 0.4, "learning_rate": 1.3655073415719764e-05, "loss": 2.2735, "step": 16207 }, { "epoch": 0.4, "learning_rate": 1.3654330310708103e-05, "loss": 2.2076, "step": 16208 }, { "epoch": 0.4, "learning_rate": 1.3653587182406208e-05, "loss": 2.0229, "step": 16209 }, { "epoch": 0.4, "learning_rate": 1.3652844030818809e-05, "loss": 2.0717, "step": 16210 }, { "epoch": 0.4, "learning_rate": 1.3652100855950645e-05, "loss": 1.9888, "step": 16211 }, { "epoch": 0.4, "learning_rate": 1.3651357657806453e-05, "loss": 1.9589, "step": 16212 }, { "epoch": 0.4, "learning_rate": 1.365061443639097e-05, "loss": 2.0514, "step": 16213 }, { "epoch": 0.4, "learning_rate": 1.364987119170893e-05, "loss": 1.9612, "step": 16214 }, { "epoch": 0.4, "learning_rate": 1.3649127923765069e-05, "loss": 2.255, "step": 16215 }, { "epoch": 0.4, "learning_rate": 1.3648384632564131e-05, "loss": 2.1468, "step": 16216 }, { "epoch": 0.4, "learning_rate": 1.3647641318110845e-05, "loss": 2.0746, "step": 16217 }, { "epoch": 0.4, "learning_rate": 1.3646897980409955e-05, "loss": 2.12, "step": 16218 }, { "epoch": 0.4, "learning_rate": 1.3646154619466195e-05, "loss": 2.1278, "step": 16219 }, { "epoch": 0.4, "learning_rate": 1.3645411235284299e-05, "loss": 2.2413, "step": 16220 }, { "epoch": 0.4, "learning_rate": 1.3644667827869013e-05, "loss": 2.2284, "step": 16221 }, { "epoch": 0.4, "learning_rate": 1.3643924397225073e-05, "loss": 2.0313, "step": 16222 }, { "epoch": 0.4, "learning_rate": 1.3643180943357213e-05, "loss": 2.1397, "step": 16223 }, { "epoch": 0.4, "learning_rate": 1.3642437466270175e-05, "loss": 1.8807, "step": 16224 }, { "epoch": 0.4, "learning_rate": 1.3641693965968692e-05, "loss": 2.1213, "step": 16225 }, { "epoch": 0.4, "learning_rate": 1.364095044245751e-05, "loss": 2.0351, "step": 16226 }, { "epoch": 0.4, "learning_rate": 1.3640206895741363e-05, "loss": 2.1204, "step": 16227 }, { "epoch": 0.4, "learning_rate": 1.3639463325824993e-05, "loss": 2.0352, "step": 16228 }, { "epoch": 0.4, "eval_loss": 1.7716996669769287, "eval_runtime": 94.7642, "eval_samples_per_second": 620.984, "eval_steps_per_second": 4.854, "step": 16228 }, { "epoch": 0.4, "learning_rate": 1.3638719732713134e-05, "loss": 2.1458, "step": 16229 }, { "epoch": 0.4, "learning_rate": 1.3637976116410531e-05, "loss": 2.0558, "step": 16230 }, { "epoch": 0.4, "learning_rate": 1.363723247692192e-05, "loss": 2.0648, "step": 16231 }, { "epoch": 0.4, "learning_rate": 1.3636488814252043e-05, "loss": 2.0463, "step": 16232 }, { "epoch": 0.4, "learning_rate": 1.3635745128405635e-05, "loss": 2.0808, "step": 16233 }, { "epoch": 0.4, "learning_rate": 1.3635001419387439e-05, "loss": 2.1666, "step": 16234 }, { "epoch": 0.4, "learning_rate": 1.3634257687202192e-05, "loss": 2.1953, "step": 16235 }, { "epoch": 0.4, "learning_rate": 1.363351393185464e-05, "loss": 2.058, "step": 16236 }, { "epoch": 0.4, "learning_rate": 1.3632770153349519e-05, "loss": 2.1031, "step": 16237 }, { "epoch": 0.4, "learning_rate": 1.3632026351691572e-05, "loss": 1.973, "step": 16238 }, { "epoch": 0.4, "learning_rate": 1.3631282526885532e-05, "loss": 2.0614, "step": 16239 }, { "epoch": 0.4, "learning_rate": 1.3630538678936147e-05, "loss": 1.9863, "step": 16240 }, { "epoch": 0.4, "learning_rate": 1.3629794807848158e-05, "loss": 2.1619, "step": 16241 }, { "epoch": 0.4, "learning_rate": 1.3629050913626304e-05, "loss": 2.0086, "step": 16242 }, { "epoch": 0.4, "learning_rate": 1.3628306996275322e-05, "loss": 2.1772, "step": 16243 }, { "epoch": 0.4, "learning_rate": 1.362756305579996e-05, "loss": 2.0702, "step": 16244 }, { "epoch": 0.4, "learning_rate": 1.3626819092204956e-05, "loss": 2.2873, "step": 16245 }, { "epoch": 0.4, "learning_rate": 1.3626075105495051e-05, "loss": 1.9804, "step": 16246 }, { "epoch": 0.4, "learning_rate": 1.362533109567499e-05, "loss": 2.0754, "step": 16247 }, { "epoch": 0.4, "learning_rate": 1.3624587062749513e-05, "loss": 1.9528, "step": 16248 }, { "epoch": 0.4, "learning_rate": 1.3623843006723357e-05, "loss": 1.8719, "step": 16249 }, { "epoch": 0.4, "learning_rate": 1.362309892760127e-05, "loss": 2.1511, "step": 16250 }, { "epoch": 0.4, "learning_rate": 1.3622354825387995e-05, "loss": 2.1454, "step": 16251 }, { "epoch": 0.4, "learning_rate": 1.362161070008827e-05, "loss": 1.967, "step": 16252 }, { "epoch": 0.4, "learning_rate": 1.3620866551706842e-05, "loss": 2.2342, "step": 16253 }, { "epoch": 0.4, "learning_rate": 1.3620122380248449e-05, "loss": 2.0973, "step": 16254 }, { "epoch": 0.4, "learning_rate": 1.3619378185717837e-05, "loss": 2.148, "step": 16255 }, { "epoch": 0.4, "learning_rate": 1.361863396811975e-05, "loss": 2.1039, "step": 16256 }, { "epoch": 0.4, "learning_rate": 1.3617889727458928e-05, "loss": 1.945, "step": 16257 }, { "epoch": 0.4, "learning_rate": 1.3617145463740115e-05, "loss": 2.0317, "step": 16258 }, { "epoch": 0.4, "learning_rate": 1.3616401176968058e-05, "loss": 2.1888, "step": 16259 }, { "epoch": 0.4, "learning_rate": 1.3615656867147495e-05, "loss": 2.1918, "step": 16260 }, { "epoch": 0.4, "learning_rate": 1.361491253428317e-05, "loss": 2.0975, "step": 16261 }, { "epoch": 0.4, "learning_rate": 1.3614168178379835e-05, "loss": 2.0615, "step": 16262 }, { "epoch": 0.4, "learning_rate": 1.3613423799442225e-05, "loss": 2.1377, "step": 16263 }, { "epoch": 0.4, "learning_rate": 1.361267939747509e-05, "loss": 1.8501, "step": 16264 }, { "epoch": 0.4, "learning_rate": 1.3611934972483169e-05, "loss": 2.0559, "step": 16265 }, { "epoch": 0.4, "learning_rate": 1.3611190524471214e-05, "loss": 2.1883, "step": 16266 }, { "epoch": 0.4, "learning_rate": 1.3610446053443962e-05, "loss": 2.1864, "step": 16267 }, { "epoch": 0.4, "learning_rate": 1.3609701559406158e-05, "loss": 2.0445, "step": 16268 }, { "epoch": 0.4, "learning_rate": 1.3608957042362554e-05, "loss": 2.2308, "step": 16269 }, { "epoch": 0.4, "learning_rate": 1.3608212502317886e-05, "loss": 2.0893, "step": 16270 }, { "epoch": 0.4, "learning_rate": 1.3607467939276907e-05, "loss": 2.0299, "step": 16271 }, { "epoch": 0.4, "learning_rate": 1.360672335324436e-05, "loss": 2.1699, "step": 16272 }, { "epoch": 0.4, "learning_rate": 1.3605978744224985e-05, "loss": 2.0004, "step": 16273 }, { "epoch": 0.4, "learning_rate": 1.3605234112223537e-05, "loss": 2.0958, "step": 16274 }, { "epoch": 0.4, "learning_rate": 1.3604489457244756e-05, "loss": 1.936, "step": 16275 }, { "epoch": 0.4, "learning_rate": 1.360374477929339e-05, "loss": 2.0168, "step": 16276 }, { "epoch": 0.4, "learning_rate": 1.3603000078374185e-05, "loss": 1.9076, "step": 16277 }, { "epoch": 0.4, "learning_rate": 1.3602255354491883e-05, "loss": 2.1732, "step": 16278 }, { "epoch": 0.4, "learning_rate": 1.3601510607651234e-05, "loss": 2.1915, "step": 16279 }, { "epoch": 0.4, "learning_rate": 1.3600765837856985e-05, "loss": 2.0947, "step": 16280 }, { "epoch": 0.4, "learning_rate": 1.3600021045113882e-05, "loss": 2.1529, "step": 16281 }, { "epoch": 0.4, "learning_rate": 1.3599276229426673e-05, "loss": 2.1186, "step": 16282 }, { "epoch": 0.4, "learning_rate": 1.35985313908001e-05, "loss": 2.1328, "step": 16283 }, { "epoch": 0.4, "learning_rate": 1.3597786529238917e-05, "loss": 1.9095, "step": 16284 }, { "epoch": 0.4, "learning_rate": 1.3597041644747865e-05, "loss": 2.1257, "step": 16285 }, { "epoch": 0.4, "learning_rate": 1.3596296737331697e-05, "loss": 2.2011, "step": 16286 }, { "epoch": 0.4, "learning_rate": 1.3595551806995157e-05, "loss": 2.1527, "step": 16287 }, { "epoch": 0.4, "learning_rate": 1.3594806853742992e-05, "loss": 1.9716, "step": 16288 }, { "epoch": 0.4, "learning_rate": 1.3594061877579953e-05, "loss": 2.2168, "step": 16289 }, { "epoch": 0.4, "learning_rate": 1.359331687851079e-05, "loss": 2.0804, "step": 16290 }, { "epoch": 0.4, "learning_rate": 1.3592571856540241e-05, "loss": 2.0515, "step": 16291 }, { "epoch": 0.4, "learning_rate": 1.3591826811673064e-05, "loss": 2.04, "step": 16292 }, { "epoch": 0.4, "learning_rate": 1.3591081743914005e-05, "loss": 2.1322, "step": 16293 }, { "epoch": 0.4, "learning_rate": 1.359033665326781e-05, "loss": 2.0292, "step": 16294 }, { "epoch": 0.4, "learning_rate": 1.3589591539739233e-05, "loss": 2.1538, "step": 16295 }, { "epoch": 0.4, "learning_rate": 1.3588846403333014e-05, "loss": 1.9164, "step": 16296 }, { "epoch": 0.4, "learning_rate": 1.3588101244053908e-05, "loss": 2.13, "step": 16297 }, { "epoch": 0.4, "learning_rate": 1.3587356061906666e-05, "loss": 2.117, "step": 16298 }, { "epoch": 0.4, "learning_rate": 1.3586610856896036e-05, "loss": 1.855, "step": 16299 }, { "epoch": 0.4, "learning_rate": 1.3585865629026764e-05, "loss": 2.042, "step": 16300 }, { "epoch": 0.4, "learning_rate": 1.3585120378303603e-05, "loss": 1.974, "step": 16301 }, { "epoch": 0.4, "learning_rate": 1.35843751047313e-05, "loss": 1.9238, "step": 16302 }, { "epoch": 0.4, "learning_rate": 1.3583629808314606e-05, "loss": 2.0135, "step": 16303 }, { "epoch": 0.4, "learning_rate": 1.3582884489058272e-05, "loss": 2.0633, "step": 16304 }, { "epoch": 0.4, "learning_rate": 1.3582139146967047e-05, "loss": 2.114, "step": 16305 }, { "epoch": 0.4, "learning_rate": 1.3581393782045682e-05, "loss": 2.209, "step": 16306 }, { "epoch": 0.4, "learning_rate": 1.3580648394298927e-05, "loss": 2.0803, "step": 16307 }, { "epoch": 0.4, "learning_rate": 1.3579902983731535e-05, "loss": 1.8781, "step": 16308 }, { "epoch": 0.4, "learning_rate": 1.3579157550348252e-05, "loss": 2.1105, "step": 16309 }, { "epoch": 0.4, "learning_rate": 1.3578412094153834e-05, "loss": 2.2157, "step": 16310 }, { "epoch": 0.4, "learning_rate": 1.3577666615153029e-05, "loss": 2.152, "step": 16311 }, { "epoch": 0.4, "learning_rate": 1.3576921113350587e-05, "loss": 2.1715, "step": 16312 }, { "epoch": 0.4, "learning_rate": 1.3576175588751262e-05, "loss": 2.4097, "step": 16313 }, { "epoch": 0.4, "learning_rate": 1.3575430041359803e-05, "loss": 1.9773, "step": 16314 }, { "epoch": 0.4, "learning_rate": 1.3574684471180966e-05, "loss": 2.0245, "step": 16315 }, { "epoch": 0.4, "learning_rate": 1.3573938878219497e-05, "loss": 2.1295, "step": 16316 }, { "epoch": 0.4, "learning_rate": 1.3573193262480154e-05, "loss": 2.14, "step": 16317 }, { "epoch": 0.4, "learning_rate": 1.3572447623967684e-05, "loss": 1.9358, "step": 16318 }, { "epoch": 0.4, "learning_rate": 1.3571701962686843e-05, "loss": 2.0008, "step": 16319 }, { "epoch": 0.4, "learning_rate": 1.357095627864238e-05, "loss": 1.9417, "step": 16320 }, { "epoch": 0.4, "learning_rate": 1.3570210571839046e-05, "loss": 2.0955, "step": 16321 }, { "epoch": 0.4, "learning_rate": 1.3569464842281602e-05, "loss": 2.054, "step": 16322 }, { "epoch": 0.4, "learning_rate": 1.356871908997479e-05, "loss": 2.0641, "step": 16323 }, { "epoch": 0.4, "learning_rate": 1.3567973314923371e-05, "loss": 1.855, "step": 16324 }, { "epoch": 0.4, "learning_rate": 1.3567227517132095e-05, "loss": 2.0068, "step": 16325 }, { "epoch": 0.4, "learning_rate": 1.3566481696605716e-05, "loss": 2.1645, "step": 16326 }, { "epoch": 0.4, "learning_rate": 1.3565735853348988e-05, "loss": 2.1833, "step": 16327 }, { "epoch": 0.4, "learning_rate": 1.3564989987366662e-05, "loss": 2.3241, "step": 16328 }, { "epoch": 0.4, "learning_rate": 1.3564244098663492e-05, "loss": 2.1668, "step": 16329 }, { "epoch": 0.4, "learning_rate": 1.3563498187244235e-05, "loss": 2.0457, "step": 16330 }, { "epoch": 0.4, "learning_rate": 1.356275225311364e-05, "loss": 2.0539, "step": 16331 }, { "epoch": 0.4, "learning_rate": 1.3562006296276465e-05, "loss": 2.2656, "step": 16332 }, { "epoch": 0.4, "learning_rate": 1.3561260316737461e-05, "loss": 2.1562, "step": 16333 }, { "epoch": 0.4, "learning_rate": 1.3560514314501387e-05, "loss": 2.1097, "step": 16334 }, { "epoch": 0.4, "learning_rate": 1.3559768289572996e-05, "loss": 2.201, "step": 16335 }, { "epoch": 0.4, "learning_rate": 1.3559022241957039e-05, "loss": 1.8338, "step": 16336 }, { "epoch": 0.4, "learning_rate": 1.3558276171658275e-05, "loss": 2.1888, "step": 16337 }, { "epoch": 0.4, "learning_rate": 1.3557530078681459e-05, "loss": 2.1143, "step": 16338 }, { "epoch": 0.4, "learning_rate": 1.3556783963031342e-05, "loss": 2.0779, "step": 16339 }, { "epoch": 0.4, "learning_rate": 1.355603782471268e-05, "loss": 2.1247, "step": 16340 }, { "epoch": 0.4, "learning_rate": 1.3555291663730233e-05, "loss": 2.1127, "step": 16341 }, { "epoch": 0.4, "learning_rate": 1.3554545480088752e-05, "loss": 2.1469, "step": 16342 }, { "epoch": 0.4, "learning_rate": 1.3553799273792996e-05, "loss": 2.1323, "step": 16343 }, { "epoch": 0.4, "learning_rate": 1.3553053044847717e-05, "loss": 2.052, "step": 16344 }, { "epoch": 0.4, "learning_rate": 1.3552306793257675e-05, "loss": 2.0484, "step": 16345 }, { "epoch": 0.4, "learning_rate": 1.3551560519027622e-05, "loss": 2.1859, "step": 16346 }, { "epoch": 0.4, "learning_rate": 1.3550814222162318e-05, "loss": 2.0721, "step": 16347 }, { "epoch": 0.4, "learning_rate": 1.3550067902666519e-05, "loss": 2.1565, "step": 16348 }, { "epoch": 0.4, "learning_rate": 1.3549321560544976e-05, "loss": 2.0993, "step": 16349 }, { "epoch": 0.4, "learning_rate": 1.3548575195802451e-05, "loss": 2.1296, "step": 16350 }, { "epoch": 0.4, "learning_rate": 1.3547828808443704e-05, "loss": 1.9629, "step": 16351 }, { "epoch": 0.4, "learning_rate": 1.3547082398473486e-05, "loss": 2.0632, "step": 16352 }, { "epoch": 0.4, "learning_rate": 1.3546335965896556e-05, "loss": 2.0409, "step": 16353 }, { "epoch": 0.4, "learning_rate": 1.3545589510717671e-05, "loss": 2.0551, "step": 16354 }, { "epoch": 0.4, "learning_rate": 1.3544843032941586e-05, "loss": 2.1324, "step": 16355 }, { "epoch": 0.4, "learning_rate": 1.3544096532573064e-05, "loss": 2.0595, "step": 16356 }, { "epoch": 0.4, "learning_rate": 1.354335000961686e-05, "loss": 2.0584, "step": 16357 }, { "epoch": 0.4, "learning_rate": 1.3542603464077728e-05, "loss": 2.1027, "step": 16358 }, { "epoch": 0.4, "learning_rate": 1.3541856895960433e-05, "loss": 2.1708, "step": 16359 }, { "epoch": 0.4, "learning_rate": 1.354111030526973e-05, "loss": 2.1605, "step": 16360 }, { "epoch": 0.4, "learning_rate": 1.3540363692010374e-05, "loss": 1.9906, "step": 16361 }, { "epoch": 0.4, "learning_rate": 1.3539617056187129e-05, "loss": 1.9253, "step": 16362 }, { "epoch": 0.4, "learning_rate": 1.3538870397804753e-05, "loss": 2.06, "step": 16363 }, { "epoch": 0.4, "learning_rate": 1.3538123716868e-05, "loss": 1.8627, "step": 16364 }, { "epoch": 0.4, "learning_rate": 1.353737701338163e-05, "loss": 1.9593, "step": 16365 }, { "epoch": 0.4, "learning_rate": 1.3536630287350405e-05, "loss": 2.1872, "step": 16366 }, { "epoch": 0.4, "learning_rate": 1.3535883538779083e-05, "loss": 2.1159, "step": 16367 }, { "epoch": 0.4, "learning_rate": 1.3535136767672423e-05, "loss": 2.0818, "step": 16368 }, { "epoch": 0.4, "learning_rate": 1.3534389974035182e-05, "loss": 1.9356, "step": 16369 }, { "epoch": 0.4, "learning_rate": 1.3533643157872125e-05, "loss": 2.2117, "step": 16370 }, { "epoch": 0.4, "learning_rate": 1.3532896319188006e-05, "loss": 1.8976, "step": 16371 }, { "epoch": 0.4, "learning_rate": 1.3532149457987588e-05, "loss": 1.9343, "step": 16372 }, { "epoch": 0.4, "learning_rate": 1.3531402574275634e-05, "loss": 2.1068, "step": 16373 }, { "epoch": 0.4, "learning_rate": 1.3530655668056896e-05, "loss": 2.1452, "step": 16374 }, { "epoch": 0.4, "learning_rate": 1.3529908739336141e-05, "loss": 2.0584, "step": 16375 }, { "epoch": 0.4, "learning_rate": 1.3529161788118125e-05, "loss": 1.9444, "step": 16376 }, { "epoch": 0.4, "learning_rate": 1.3528414814407614e-05, "loss": 2.1155, "step": 16377 }, { "epoch": 0.4, "learning_rate": 1.3527667818209365e-05, "loss": 2.1495, "step": 16378 }, { "epoch": 0.4, "learning_rate": 1.3526920799528137e-05, "loss": 1.9219, "step": 16379 }, { "epoch": 0.4, "learning_rate": 1.3526173758368695e-05, "loss": 2.2064, "step": 16380 }, { "epoch": 0.4, "learning_rate": 1.3525426694735799e-05, "loss": 2.1935, "step": 16381 }, { "epoch": 0.4, "learning_rate": 1.352467960863421e-05, "loss": 2.0133, "step": 16382 }, { "epoch": 0.4, "learning_rate": 1.3523932500068686e-05, "loss": 1.9703, "step": 16383 }, { "epoch": 0.4, "learning_rate": 1.3523185369043997e-05, "loss": 2.0073, "step": 16384 }, { "epoch": 0.4, "learning_rate": 1.3522438215564893e-05, "loss": 2.0893, "step": 16385 }, { "epoch": 0.4, "learning_rate": 1.3521691039636146e-05, "loss": 2.1233, "step": 16386 }, { "epoch": 0.4, "learning_rate": 1.3520943841262516e-05, "loss": 2.1935, "step": 16387 }, { "epoch": 0.4, "learning_rate": 1.3520196620448763e-05, "loss": 2.1413, "step": 16388 }, { "epoch": 0.4, "learning_rate": 1.3519449377199645e-05, "loss": 2.1653, "step": 16389 }, { "epoch": 0.4, "learning_rate": 1.3518702111519933e-05, "loss": 2.1936, "step": 16390 }, { "epoch": 0.4, "learning_rate": 1.3517954823414385e-05, "loss": 2.0008, "step": 16391 }, { "epoch": 0.4, "learning_rate": 1.3517207512887764e-05, "loss": 2.2017, "step": 16392 }, { "epoch": 0.4, "learning_rate": 1.3516460179944835e-05, "loss": 2.2013, "step": 16393 }, { "epoch": 0.4, "learning_rate": 1.3515712824590357e-05, "loss": 1.9399, "step": 16394 }, { "epoch": 0.4, "learning_rate": 1.3514965446829097e-05, "loss": 2.0928, "step": 16395 }, { "epoch": 0.4, "learning_rate": 1.3514218046665818e-05, "loss": 2.0534, "step": 16396 }, { "epoch": 0.4, "learning_rate": 1.351347062410528e-05, "loss": 2.1176, "step": 16397 }, { "epoch": 0.4, "learning_rate": 1.3512723179152249e-05, "loss": 2.1801, "step": 16398 }, { "epoch": 0.4, "learning_rate": 1.3511975711811486e-05, "loss": 2.2205, "step": 16399 }, { "epoch": 0.4, "learning_rate": 1.3511228222087762e-05, "loss": 2.33, "step": 16400 }, { "epoch": 0.4, "learning_rate": 1.3510480709985834e-05, "loss": 2.0602, "step": 16401 }, { "epoch": 0.4, "learning_rate": 1.3509733175510469e-05, "loss": 2.0796, "step": 16402 }, { "epoch": 0.4, "learning_rate": 1.350898561866643e-05, "loss": 2.0171, "step": 16403 }, { "epoch": 0.4, "learning_rate": 1.350823803945848e-05, "loss": 2.0919, "step": 16404 }, { "epoch": 0.4, "learning_rate": 1.350749043789139e-05, "loss": 2.1029, "step": 16405 }, { "epoch": 0.4, "learning_rate": 1.350674281396992e-05, "loss": 2.2728, "step": 16406 }, { "epoch": 0.4, "learning_rate": 1.3505995167698833e-05, "loss": 1.9505, "step": 16407 }, { "epoch": 0.4, "learning_rate": 1.3505247499082896e-05, "loss": 2.2174, "step": 16408 }, { "epoch": 0.4, "learning_rate": 1.3504499808126876e-05, "loss": 1.9772, "step": 16409 }, { "epoch": 0.4, "learning_rate": 1.3503752094835533e-05, "loss": 2.0001, "step": 16410 }, { "epoch": 0.4, "learning_rate": 1.350300435921364e-05, "loss": 2.133, "step": 16411 }, { "epoch": 0.4, "learning_rate": 1.3502256601265959e-05, "loss": 2.2339, "step": 16412 }, { "epoch": 0.4, "learning_rate": 1.3501508820997251e-05, "loss": 2.2261, "step": 16413 }, { "epoch": 0.4, "learning_rate": 1.3500761018412292e-05, "loss": 2.1177, "step": 16414 }, { "epoch": 0.4, "learning_rate": 1.3500013193515839e-05, "loss": 2.1965, "step": 16415 }, { "epoch": 0.4, "learning_rate": 1.3499265346312662e-05, "loss": 2.1875, "step": 16416 }, { "epoch": 0.4, "learning_rate": 1.3498517476807527e-05, "loss": 2.0024, "step": 16417 }, { "epoch": 0.4, "learning_rate": 1.3497769585005199e-05, "loss": 2.1126, "step": 16418 }, { "epoch": 0.4, "learning_rate": 1.3497021670910444e-05, "loss": 1.9657, "step": 16419 }, { "epoch": 0.4, "learning_rate": 1.3496273734528032e-05, "loss": 2.1295, "step": 16420 }, { "epoch": 0.4, "learning_rate": 1.3495525775862728e-05, "loss": 2.0542, "step": 16421 }, { "epoch": 0.4, "learning_rate": 1.34947777949193e-05, "loss": 2.0469, "step": 16422 }, { "epoch": 0.4, "learning_rate": 1.3494029791702511e-05, "loss": 2.081, "step": 16423 }, { "epoch": 0.4, "learning_rate": 1.3493281766217134e-05, "loss": 2.0937, "step": 16424 }, { "epoch": 0.4, "learning_rate": 1.3492533718467935e-05, "loss": 2.0956, "step": 16425 }, { "epoch": 0.4, "learning_rate": 1.3491785648459679e-05, "loss": 2.0309, "step": 16426 }, { "epoch": 0.4, "learning_rate": 1.3491037556197135e-05, "loss": 2.0432, "step": 16427 }, { "epoch": 0.4, "learning_rate": 1.3490289441685071e-05, "loss": 2.1799, "step": 16428 }, { "epoch": 0.4, "learning_rate": 1.3489541304928253e-05, "loss": 1.9543, "step": 16429 }, { "epoch": 0.4, "learning_rate": 1.3488793145931454e-05, "loss": 1.8659, "step": 16430 }, { "epoch": 0.41, "learning_rate": 1.3488044964699438e-05, "loss": 1.9581, "step": 16431 }, { "epoch": 0.41, "learning_rate": 1.3487296761236976e-05, "loss": 2.068, "step": 16432 }, { "epoch": 0.41, "learning_rate": 1.3486548535548834e-05, "loss": 2.1612, "step": 16433 }, { "epoch": 0.41, "learning_rate": 1.3485800287639784e-05, "loss": 2.1135, "step": 16434 }, { "epoch": 0.41, "learning_rate": 1.3485052017514591e-05, "loss": 2.2638, "step": 16435 }, { "epoch": 0.41, "learning_rate": 1.3484303725178026e-05, "loss": 2.1449, "step": 16436 }, { "epoch": 0.41, "learning_rate": 1.3483555410634857e-05, "loss": 2.0812, "step": 16437 }, { "epoch": 0.41, "learning_rate": 1.3482807073889852e-05, "loss": 1.8398, "step": 16438 }, { "epoch": 0.41, "learning_rate": 1.3482058714947788e-05, "loss": 2.0656, "step": 16439 }, { "epoch": 0.41, "learning_rate": 1.3481310333813426e-05, "loss": 2.0591, "step": 16440 }, { "epoch": 0.41, "learning_rate": 1.3480561930491538e-05, "loss": 2.1528, "step": 16441 }, { "epoch": 0.41, "learning_rate": 1.3479813504986896e-05, "loss": 1.9356, "step": 16442 }, { "epoch": 0.41, "learning_rate": 1.3479065057304268e-05, "loss": 1.8807, "step": 16443 }, { "epoch": 0.41, "learning_rate": 1.3478316587448426e-05, "loss": 2.1679, "step": 16444 }, { "epoch": 0.41, "learning_rate": 1.3477568095424137e-05, "loss": 2.0148, "step": 16445 }, { "epoch": 0.41, "learning_rate": 1.3476819581236173e-05, "loss": 2.0518, "step": 16446 }, { "epoch": 0.41, "learning_rate": 1.3476071044889306e-05, "loss": 2.0964, "step": 16447 }, { "epoch": 0.41, "learning_rate": 1.3475322486388302e-05, "loss": 2.1233, "step": 16448 }, { "epoch": 0.41, "learning_rate": 1.347457390573794e-05, "loss": 2.004, "step": 16449 }, { "epoch": 0.41, "learning_rate": 1.3473825302942985e-05, "loss": 2.0834, "step": 16450 }, { "epoch": 0.41, "learning_rate": 1.347307667800821e-05, "loss": 2.1518, "step": 16451 }, { "epoch": 0.41, "learning_rate": 1.3472328030938381e-05, "loss": 2.0246, "step": 16452 }, { "epoch": 0.41, "learning_rate": 1.3471579361738279e-05, "loss": 2.1408, "step": 16453 }, { "epoch": 0.41, "learning_rate": 1.3470830670412666e-05, "loss": 2.2135, "step": 16454 }, { "epoch": 0.41, "learning_rate": 1.347008195696632e-05, "loss": 2.1287, "step": 16455 }, { "epoch": 0.41, "learning_rate": 1.3469333221404013e-05, "loss": 1.9794, "step": 16456 }, { "epoch": 0.41, "learning_rate": 1.3468584463730511e-05, "loss": 2.0866, "step": 16457 }, { "epoch": 0.41, "learning_rate": 1.3467835683950591e-05, "loss": 2.0793, "step": 16458 }, { "epoch": 0.41, "learning_rate": 1.3467086882069027e-05, "loss": 2.1954, "step": 16459 }, { "epoch": 0.41, "learning_rate": 1.3466338058090586e-05, "loss": 2.0895, "step": 16460 }, { "epoch": 0.41, "learning_rate": 1.3465589212020045e-05, "loss": 2.0066, "step": 16461 }, { "epoch": 0.41, "learning_rate": 1.3464840343862171e-05, "loss": 2.2678, "step": 16462 }, { "epoch": 0.41, "learning_rate": 1.346409145362174e-05, "loss": 2.1537, "step": 16463 }, { "epoch": 0.41, "learning_rate": 1.3463342541303529e-05, "loss": 1.9418, "step": 16464 }, { "epoch": 0.41, "learning_rate": 1.3462593606912307e-05, "loss": 2.1533, "step": 16465 }, { "epoch": 0.41, "learning_rate": 1.3461844650452842e-05, "loss": 2.0641, "step": 16466 }, { "epoch": 0.41, "learning_rate": 1.3461095671929917e-05, "loss": 2.1003, "step": 16467 }, { "epoch": 0.41, "learning_rate": 1.3460346671348303e-05, "loss": 2.1616, "step": 16468 }, { "epoch": 0.41, "learning_rate": 1.3459597648712772e-05, "loss": 1.9749, "step": 16469 }, { "epoch": 0.41, "learning_rate": 1.3458848604028096e-05, "loss": 2.0643, "step": 16470 }, { "epoch": 0.41, "learning_rate": 1.345809953729905e-05, "loss": 2.2895, "step": 16471 }, { "epoch": 0.41, "learning_rate": 1.3457350448530406e-05, "loss": 2.0157, "step": 16472 }, { "epoch": 0.41, "learning_rate": 1.3456601337726944e-05, "loss": 2.1319, "step": 16473 }, { "epoch": 0.41, "learning_rate": 1.3455852204893438e-05, "loss": 1.992, "step": 16474 }, { "epoch": 0.41, "learning_rate": 1.3455103050034657e-05, "loss": 2.1289, "step": 16475 }, { "epoch": 0.41, "learning_rate": 1.3454353873155376e-05, "loss": 2.287, "step": 16476 }, { "epoch": 0.41, "learning_rate": 1.3453604674260376e-05, "loss": 2.0355, "step": 16477 }, { "epoch": 0.41, "learning_rate": 1.3452855453354427e-05, "loss": 2.0405, "step": 16478 }, { "epoch": 0.41, "learning_rate": 1.3452106210442304e-05, "loss": 2.2035, "step": 16479 }, { "epoch": 0.41, "learning_rate": 1.3451356945528782e-05, "loss": 2.2674, "step": 16480 }, { "epoch": 0.41, "learning_rate": 1.3450607658618637e-05, "loss": 2.1894, "step": 16481 }, { "epoch": 0.41, "learning_rate": 1.3449858349716648e-05, "loss": 2.2194, "step": 16482 }, { "epoch": 0.41, "learning_rate": 1.3449109018827587e-05, "loss": 2.1651, "step": 16483 }, { "epoch": 0.41, "learning_rate": 1.344835966595623e-05, "loss": 2.1501, "step": 16484 }, { "epoch": 0.41, "learning_rate": 1.3447610291107354e-05, "loss": 2.1049, "step": 16485 }, { "epoch": 0.41, "learning_rate": 1.344686089428573e-05, "loss": 2.194, "step": 16486 }, { "epoch": 0.41, "learning_rate": 1.344611147549614e-05, "loss": 2.1514, "step": 16487 }, { "epoch": 0.41, "learning_rate": 1.344536203474336e-05, "loss": 2.2438, "step": 16488 }, { "epoch": 0.41, "learning_rate": 1.3444612572032162e-05, "loss": 1.8794, "step": 16489 }, { "epoch": 0.41, "learning_rate": 1.3443863087367327e-05, "loss": 1.9202, "step": 16490 }, { "epoch": 0.41, "learning_rate": 1.3443113580753631e-05, "loss": 2.1158, "step": 16491 }, { "epoch": 0.41, "learning_rate": 1.344236405219585e-05, "loss": 2.1886, "step": 16492 }, { "epoch": 0.41, "learning_rate": 1.3441614501698763e-05, "loss": 2.0205, "step": 16493 }, { "epoch": 0.41, "learning_rate": 1.3440864929267143e-05, "loss": 2.0504, "step": 16494 }, { "epoch": 0.41, "learning_rate": 1.3440115334905768e-05, "loss": 2.0421, "step": 16495 }, { "epoch": 0.41, "learning_rate": 1.343936571861942e-05, "loss": 2.1367, "step": 16496 }, { "epoch": 0.41, "learning_rate": 1.3438616080412872e-05, "loss": 2.0119, "step": 16497 }, { "epoch": 0.41, "learning_rate": 1.3437866420290902e-05, "loss": 1.9872, "step": 16498 }, { "epoch": 0.41, "learning_rate": 1.3437116738258288e-05, "loss": 1.8304, "step": 16499 }, { "epoch": 0.41, "learning_rate": 1.343636703431981e-05, "loss": 1.9342, "step": 16500 }, { "epoch": 0.41, "learning_rate": 1.3435617308480247e-05, "loss": 2.1742, "step": 16501 }, { "epoch": 0.41, "learning_rate": 1.3434867560744375e-05, "loss": 2.2229, "step": 16502 }, { "epoch": 0.41, "learning_rate": 1.3434117791116972e-05, "loss": 2.1146, "step": 16503 }, { "epoch": 0.41, "learning_rate": 1.3433367999602818e-05, "loss": 2.0869, "step": 16504 }, { "epoch": 0.41, "learning_rate": 1.3432618186206687e-05, "loss": 2.0874, "step": 16505 }, { "epoch": 0.41, "learning_rate": 1.3431868350933367e-05, "loss": 2.0756, "step": 16506 }, { "epoch": 0.41, "learning_rate": 1.3431118493787626e-05, "loss": 2.2012, "step": 16507 }, { "epoch": 0.41, "learning_rate": 1.3430368614774252e-05, "loss": 2.0541, "step": 16508 }, { "epoch": 0.41, "learning_rate": 1.3429618713898023e-05, "loss": 2.0109, "step": 16509 }, { "epoch": 0.41, "learning_rate": 1.3428868791163711e-05, "loss": 2.1214, "step": 16510 }, { "epoch": 0.41, "learning_rate": 1.3428118846576104e-05, "loss": 2.1804, "step": 16511 }, { "epoch": 0.41, "learning_rate": 1.3427368880139979e-05, "loss": 2.0792, "step": 16512 }, { "epoch": 0.41, "learning_rate": 1.3426618891860115e-05, "loss": 1.9935, "step": 16513 }, { "epoch": 0.41, "learning_rate": 1.342586888174129e-05, "loss": 1.9261, "step": 16514 }, { "epoch": 0.41, "learning_rate": 1.3425118849788288e-05, "loss": 2.0088, "step": 16515 }, { "epoch": 0.41, "learning_rate": 1.3424368796005884e-05, "loss": 2.0902, "step": 16516 }, { "epoch": 0.41, "learning_rate": 1.3423618720398866e-05, "loss": 2.0347, "step": 16517 }, { "epoch": 0.41, "learning_rate": 1.342286862297201e-05, "loss": 2.2089, "step": 16518 }, { "epoch": 0.41, "learning_rate": 1.3422118503730095e-05, "loss": 2.2337, "step": 16519 }, { "epoch": 0.41, "learning_rate": 1.3421368362677903e-05, "loss": 2.1205, "step": 16520 }, { "epoch": 0.41, "learning_rate": 1.3420618199820216e-05, "loss": 2.3262, "step": 16521 }, { "epoch": 0.41, "learning_rate": 1.3419868015161814e-05, "loss": 2.0852, "step": 16522 }, { "epoch": 0.41, "learning_rate": 1.3419117808707483e-05, "loss": 2.0915, "step": 16523 }, { "epoch": 0.41, "learning_rate": 1.3418367580461994e-05, "loss": 2.0064, "step": 16524 }, { "epoch": 0.41, "learning_rate": 1.3417617330430134e-05, "loss": 2.1678, "step": 16525 }, { "epoch": 0.41, "learning_rate": 1.3416867058616689e-05, "loss": 2.0876, "step": 16526 }, { "epoch": 0.41, "learning_rate": 1.3416116765026437e-05, "loss": 2.1546, "step": 16527 }, { "epoch": 0.41, "learning_rate": 1.3415366449664158e-05, "loss": 2.0875, "step": 16528 }, { "epoch": 0.41, "learning_rate": 1.3414616112534633e-05, "loss": 2.0638, "step": 16529 }, { "epoch": 0.41, "learning_rate": 1.3413865753642652e-05, "loss": 2.0924, "step": 16530 }, { "epoch": 0.41, "learning_rate": 1.341311537299299e-05, "loss": 2.1179, "step": 16531 }, { "epoch": 0.41, "learning_rate": 1.341236497059043e-05, "loss": 2.0154, "step": 16532 }, { "epoch": 0.41, "learning_rate": 1.3411614546439755e-05, "loss": 2.24, "step": 16533 }, { "epoch": 0.41, "learning_rate": 1.341086410054575e-05, "loss": 2.1358, "step": 16534 }, { "epoch": 0.41, "learning_rate": 1.3410113632913197e-05, "loss": 2.077, "step": 16535 }, { "epoch": 0.41, "learning_rate": 1.3409363143546878e-05, "loss": 2.1224, "step": 16536 }, { "epoch": 0.41, "learning_rate": 1.3408612632451579e-05, "loss": 1.9476, "step": 16537 }, { "epoch": 0.41, "learning_rate": 1.3407862099632078e-05, "loss": 2.0962, "step": 16538 }, { "epoch": 0.41, "learning_rate": 1.3407111545093163e-05, "loss": 2.1783, "step": 16539 }, { "epoch": 0.41, "learning_rate": 1.3406360968839615e-05, "loss": 2.1524, "step": 16540 }, { "epoch": 0.41, "learning_rate": 1.3405610370876218e-05, "loss": 1.9889, "step": 16541 }, { "epoch": 0.41, "learning_rate": 1.3404859751207755e-05, "loss": 2.1702, "step": 16542 }, { "epoch": 0.41, "learning_rate": 1.3404109109839015e-05, "loss": 2.1488, "step": 16543 }, { "epoch": 0.41, "learning_rate": 1.3403358446774775e-05, "loss": 2.0655, "step": 16544 }, { "epoch": 0.41, "learning_rate": 1.3402607762019823e-05, "loss": 2.0746, "step": 16545 }, { "epoch": 0.41, "learning_rate": 1.3401857055578945e-05, "loss": 1.8717, "step": 16546 }, { "epoch": 0.41, "learning_rate": 1.3401106327456921e-05, "loss": 2.0597, "step": 16547 }, { "epoch": 0.41, "learning_rate": 1.340035557765854e-05, "loss": 2.0718, "step": 16548 }, { "epoch": 0.41, "learning_rate": 1.3399604806188582e-05, "loss": 2.1941, "step": 16549 }, { "epoch": 0.41, "learning_rate": 1.3398854013051836e-05, "loss": 2.0615, "step": 16550 }, { "epoch": 0.41, "learning_rate": 1.3398103198253085e-05, "loss": 2.3642, "step": 16551 }, { "epoch": 0.41, "learning_rate": 1.3397352361797118e-05, "loss": 2.0025, "step": 16552 }, { "epoch": 0.41, "learning_rate": 1.3396601503688714e-05, "loss": 1.9957, "step": 16553 }, { "epoch": 0.41, "learning_rate": 1.3395850623932663e-05, "loss": 2.0374, "step": 16554 }, { "epoch": 0.41, "learning_rate": 1.3395099722533748e-05, "loss": 2.1813, "step": 16555 }, { "epoch": 0.41, "learning_rate": 1.339434879949676e-05, "loss": 2.0346, "step": 16556 }, { "epoch": 0.41, "learning_rate": 1.3393597854826479e-05, "loss": 2.1533, "step": 16557 }, { "epoch": 0.41, "learning_rate": 1.3392846888527688e-05, "loss": 2.0977, "step": 16558 }, { "epoch": 0.41, "learning_rate": 1.3392095900605184e-05, "loss": 1.9775, "step": 16559 }, { "epoch": 0.41, "learning_rate": 1.3391344891063745e-05, "loss": 2.0314, "step": 16560 }, { "epoch": 0.41, "learning_rate": 1.339059385990816e-05, "loss": 2.13, "step": 16561 }, { "epoch": 0.41, "learning_rate": 1.3389842807143216e-05, "loss": 2.11, "step": 16562 }, { "epoch": 0.41, "learning_rate": 1.3389091732773695e-05, "loss": 2.1549, "step": 16563 }, { "epoch": 0.41, "learning_rate": 1.3388340636804396e-05, "loss": 1.9251, "step": 16564 }, { "epoch": 0.41, "learning_rate": 1.3387589519240092e-05, "loss": 2.0975, "step": 16565 }, { "epoch": 0.41, "learning_rate": 1.3386838380085578e-05, "loss": 2.249, "step": 16566 }, { "epoch": 0.41, "learning_rate": 1.3386087219345641e-05, "loss": 1.948, "step": 16567 }, { "epoch": 0.41, "learning_rate": 1.3385336037025063e-05, "loss": 2.2186, "step": 16568 }, { "epoch": 0.41, "learning_rate": 1.3384584833128634e-05, "loss": 2.2618, "step": 16569 }, { "epoch": 0.41, "learning_rate": 1.3383833607661146e-05, "loss": 2.0277, "step": 16570 }, { "epoch": 0.41, "learning_rate": 1.3383082360627384e-05, "loss": 1.9724, "step": 16571 }, { "epoch": 0.41, "learning_rate": 1.3382331092032134e-05, "loss": 2.1982, "step": 16572 }, { "epoch": 0.41, "learning_rate": 1.3381579801880187e-05, "loss": 2.1078, "step": 16573 }, { "epoch": 0.41, "learning_rate": 1.3380828490176327e-05, "loss": 2.0838, "step": 16574 }, { "epoch": 0.41, "learning_rate": 1.338007715692535e-05, "loss": 1.9096, "step": 16575 }, { "epoch": 0.41, "learning_rate": 1.3379325802132037e-05, "loss": 2.1339, "step": 16576 }, { "epoch": 0.41, "learning_rate": 1.3378574425801179e-05, "loss": 2.1042, "step": 16577 }, { "epoch": 0.41, "learning_rate": 1.3377823027937565e-05, "loss": 2.0259, "step": 16578 }, { "epoch": 0.41, "learning_rate": 1.3377071608545986e-05, "loss": 2.1714, "step": 16579 }, { "epoch": 0.41, "learning_rate": 1.3376320167631228e-05, "loss": 2.0768, "step": 16580 }, { "epoch": 0.41, "learning_rate": 1.337556870519808e-05, "loss": 2.1305, "step": 16581 }, { "epoch": 0.41, "learning_rate": 1.3374817221251336e-05, "loss": 2.008, "step": 16582 }, { "epoch": 0.41, "learning_rate": 1.3374065715795777e-05, "loss": 1.8554, "step": 16583 }, { "epoch": 0.41, "learning_rate": 1.3373314188836202e-05, "loss": 2.1838, "step": 16584 }, { "epoch": 0.41, "learning_rate": 1.3372562640377395e-05, "loss": 2.0588, "step": 16585 }, { "epoch": 0.41, "learning_rate": 1.3371811070424145e-05, "loss": 2.0766, "step": 16586 }, { "epoch": 0.41, "learning_rate": 1.3371059478981246e-05, "loss": 2.0258, "step": 16587 }, { "epoch": 0.41, "learning_rate": 1.3370307866053488e-05, "loss": 2.1164, "step": 16588 }, { "epoch": 0.41, "learning_rate": 1.3369556231645657e-05, "loss": 2.1157, "step": 16589 }, { "epoch": 0.41, "learning_rate": 1.3368804575762549e-05, "loss": 2.2169, "step": 16590 }, { "epoch": 0.41, "learning_rate": 1.336805289840895e-05, "loss": 2.2239, "step": 16591 }, { "epoch": 0.41, "learning_rate": 1.336730119958965e-05, "loss": 2.0471, "step": 16592 }, { "epoch": 0.41, "learning_rate": 1.3366549479309445e-05, "loss": 2.1482, "step": 16593 }, { "epoch": 0.41, "learning_rate": 1.3365797737573123e-05, "loss": 1.8721, "step": 16594 }, { "epoch": 0.41, "learning_rate": 1.3365045974385474e-05, "loss": 2.1008, "step": 16595 }, { "epoch": 0.41, "learning_rate": 1.336429418975129e-05, "loss": 2.1608, "step": 16596 }, { "epoch": 0.41, "learning_rate": 1.336354238367536e-05, "loss": 2.1312, "step": 16597 }, { "epoch": 0.41, "learning_rate": 1.3362790556162483e-05, "loss": 2.1445, "step": 16598 }, { "epoch": 0.41, "learning_rate": 1.3362038707217446e-05, "loss": 2.1265, "step": 16599 }, { "epoch": 0.41, "learning_rate": 1.336128683684504e-05, "loss": 2.141, "step": 16600 }, { "epoch": 0.41, "learning_rate": 1.3360534945050057e-05, "loss": 1.9938, "step": 16601 }, { "epoch": 0.41, "learning_rate": 1.3359783031837287e-05, "loss": 2.0463, "step": 16602 }, { "epoch": 0.41, "learning_rate": 1.3359031097211527e-05, "loss": 2.0708, "step": 16603 }, { "epoch": 0.41, "learning_rate": 1.3358279141177568e-05, "loss": 2.0561, "step": 16604 }, { "epoch": 0.41, "learning_rate": 1.3357527163740201e-05, "loss": 1.9929, "step": 16605 }, { "epoch": 0.41, "learning_rate": 1.3356775164904219e-05, "loss": 2.0614, "step": 16606 }, { "epoch": 0.41, "learning_rate": 1.3356023144674413e-05, "loss": 2.3081, "step": 16607 }, { "epoch": 0.41, "learning_rate": 1.3355271103055581e-05, "loss": 2.215, "step": 16608 }, { "epoch": 0.41, "learning_rate": 1.3354519040052511e-05, "loss": 2.0431, "step": 16609 }, { "epoch": 0.41, "learning_rate": 1.335376695567e-05, "loss": 2.0774, "step": 16610 }, { "epoch": 0.41, "learning_rate": 1.3353014849912838e-05, "loss": 2.0728, "step": 16611 }, { "epoch": 0.41, "learning_rate": 1.335226272278582e-05, "loss": 2.176, "step": 16612 }, { "epoch": 0.41, "learning_rate": 1.3351510574293735e-05, "loss": 2.2625, "step": 16613 }, { "epoch": 0.41, "learning_rate": 1.3350758404441386e-05, "loss": 2.1965, "step": 16614 }, { "epoch": 0.41, "learning_rate": 1.3350006213233562e-05, "loss": 1.999, "step": 16615 }, { "epoch": 0.41, "learning_rate": 1.3349254000675053e-05, "loss": 1.9898, "step": 16616 }, { "epoch": 0.41, "learning_rate": 1.3348501766770659e-05, "loss": 2.0165, "step": 16617 }, { "epoch": 0.41, "learning_rate": 1.3347749511525172e-05, "loss": 2.0717, "step": 16618 }, { "epoch": 0.41, "learning_rate": 1.3346997234943388e-05, "loss": 1.9321, "step": 16619 }, { "epoch": 0.41, "learning_rate": 1.3346244937030094e-05, "loss": 2.1042, "step": 16620 }, { "epoch": 0.41, "learning_rate": 1.3345492617790096e-05, "loss": 2.0589, "step": 16621 }, { "epoch": 0.41, "learning_rate": 1.334474027722818e-05, "loss": 2.088, "step": 16622 }, { "epoch": 0.41, "learning_rate": 1.3343987915349149e-05, "loss": 2.1653, "step": 16623 }, { "epoch": 0.41, "learning_rate": 1.334323553215779e-05, "loss": 2.1024, "step": 16624 }, { "epoch": 0.41, "learning_rate": 1.3342483127658904e-05, "loss": 2.1342, "step": 16625 }, { "epoch": 0.41, "learning_rate": 1.3341730701857279e-05, "loss": 2.0249, "step": 16626 }, { "epoch": 0.41, "learning_rate": 1.3340978254757717e-05, "loss": 2.0606, "step": 16627 }, { "epoch": 0.41, "learning_rate": 1.3340225786365014e-05, "loss": 2.2065, "step": 16628 }, { "epoch": 0.41, "learning_rate": 1.3339473296683958e-05, "loss": 2.0222, "step": 16629 }, { "epoch": 0.41, "learning_rate": 1.3338720785719357e-05, "loss": 2.0192, "step": 16630 }, { "epoch": 0.41, "learning_rate": 1.3337968253475994e-05, "loss": 1.9939, "step": 16631 }, { "epoch": 0.41, "learning_rate": 1.3337215699958677e-05, "loss": 2.1102, "step": 16632 }, { "epoch": 0.41, "learning_rate": 1.3336463125172195e-05, "loss": 2.1301, "step": 16633 }, { "epoch": 0.41, "learning_rate": 1.3335710529121346e-05, "loss": 1.8986, "step": 16634 }, { "epoch": 0.41, "learning_rate": 1.3334957911810928e-05, "loss": 2.0723, "step": 16635 }, { "epoch": 0.41, "learning_rate": 1.3334205273245733e-05, "loss": 2.0342, "step": 16636 }, { "epoch": 0.41, "learning_rate": 1.3333452613430565e-05, "loss": 2.0506, "step": 16637 }, { "epoch": 0.41, "learning_rate": 1.3332699932370217e-05, "loss": 2.0828, "step": 16638 }, { "epoch": 0.41, "learning_rate": 1.3331947230069482e-05, "loss": 2.0134, "step": 16639 }, { "epoch": 0.41, "learning_rate": 1.3331194506533165e-05, "loss": 2.0015, "step": 16640 }, { "epoch": 0.41, "learning_rate": 1.3330441761766058e-05, "loss": 1.9731, "step": 16641 }, { "epoch": 0.41, "learning_rate": 1.332968899577296e-05, "loss": 1.9028, "step": 16642 }, { "epoch": 0.41, "learning_rate": 1.3328936208558672e-05, "loss": 1.9222, "step": 16643 }, { "epoch": 0.41, "learning_rate": 1.3328183400127987e-05, "loss": 2.0322, "step": 16644 }, { "epoch": 0.41, "learning_rate": 1.3327430570485705e-05, "loss": 2.0952, "step": 16645 }, { "epoch": 0.41, "learning_rate": 1.3326677719636623e-05, "loss": 2.0096, "step": 16646 }, { "epoch": 0.41, "learning_rate": 1.332592484758554e-05, "loss": 1.9866, "step": 16647 }, { "epoch": 0.41, "learning_rate": 1.3325171954337253e-05, "loss": 1.9379, "step": 16648 }, { "epoch": 0.41, "learning_rate": 1.332441903989656e-05, "loss": 2.0549, "step": 16649 }, { "epoch": 0.41, "learning_rate": 1.3323666104268265e-05, "loss": 2.1565, "step": 16650 }, { "epoch": 0.41, "learning_rate": 1.332291314745716e-05, "loss": 1.9836, "step": 16651 }, { "epoch": 0.41, "learning_rate": 1.3322160169468048e-05, "loss": 2.2482, "step": 16652 }, { "epoch": 0.41, "learning_rate": 1.3321407170305727e-05, "loss": 2.0763, "step": 16653 }, { "epoch": 0.41, "learning_rate": 1.3320654149974997e-05, "loss": 2.1167, "step": 16654 }, { "epoch": 0.41, "learning_rate": 1.3319901108480653e-05, "loss": 2.0762, "step": 16655 }, { "epoch": 0.41, "learning_rate": 1.3319148045827497e-05, "loss": 2.1, "step": 16656 }, { "epoch": 0.41, "learning_rate": 1.331839496202033e-05, "loss": 1.9887, "step": 16657 }, { "epoch": 0.41, "learning_rate": 1.331764185706395e-05, "loss": 2.0161, "step": 16658 }, { "epoch": 0.41, "learning_rate": 1.331688873096316e-05, "loss": 2.1357, "step": 16659 }, { "epoch": 0.41, "learning_rate": 1.3316135583722754e-05, "loss": 2.2184, "step": 16660 }, { "epoch": 0.41, "learning_rate": 1.3315382415347536e-05, "loss": 2.1544, "step": 16661 }, { "epoch": 0.41, "learning_rate": 1.3314629225842306e-05, "loss": 2.102, "step": 16662 }, { "epoch": 0.41, "learning_rate": 1.3313876015211864e-05, "loss": 2.1401, "step": 16663 }, { "epoch": 0.41, "learning_rate": 1.3313122783461008e-05, "loss": 2.0193, "step": 16664 }, { "epoch": 0.41, "learning_rate": 1.3312369530594542e-05, "loss": 2.0264, "step": 16665 }, { "epoch": 0.41, "learning_rate": 1.3311616256617266e-05, "loss": 2.1105, "step": 16666 }, { "epoch": 0.41, "learning_rate": 1.3310862961533981e-05, "loss": 1.9473, "step": 16667 }, { "epoch": 0.41, "learning_rate": 1.3310109645349487e-05, "loss": 1.8753, "step": 16668 }, { "epoch": 0.41, "learning_rate": 1.3309356308068586e-05, "loss": 2.0121, "step": 16669 }, { "epoch": 0.41, "learning_rate": 1.3308602949696077e-05, "loss": 1.9227, "step": 16670 }, { "epoch": 0.41, "learning_rate": 1.3307849570236763e-05, "loss": 2.0246, "step": 16671 }, { "epoch": 0.41, "learning_rate": 1.3307096169695449e-05, "loss": 2.1454, "step": 16672 }, { "epoch": 0.41, "learning_rate": 1.330634274807693e-05, "loss": 2.1664, "step": 16673 }, { "epoch": 0.41, "learning_rate": 1.3305589305386013e-05, "loss": 2.1134, "step": 16674 }, { "epoch": 0.41, "learning_rate": 1.3304835841627496e-05, "loss": 1.9479, "step": 16675 }, { "epoch": 0.41, "learning_rate": 1.3304082356806186e-05, "loss": 2.0592, "step": 16676 }, { "epoch": 0.41, "learning_rate": 1.330332885092688e-05, "loss": 1.9676, "step": 16677 }, { "epoch": 0.41, "learning_rate": 1.3302575323994385e-05, "loss": 2.033, "step": 16678 }, { "epoch": 0.41, "learning_rate": 1.3301821776013496e-05, "loss": 2.1241, "step": 16679 }, { "epoch": 0.41, "learning_rate": 1.3301068206989026e-05, "loss": 2.1779, "step": 16680 }, { "epoch": 0.41, "learning_rate": 1.3300314616925771e-05, "loss": 2.1392, "step": 16681 }, { "epoch": 0.41, "learning_rate": 1.3299561005828534e-05, "loss": 2.1492, "step": 16682 }, { "epoch": 0.41, "learning_rate": 1.3298807373702123e-05, "loss": 2.0825, "step": 16683 }, { "epoch": 0.41, "learning_rate": 1.3298053720551333e-05, "loss": 2.0685, "step": 16684 }, { "epoch": 0.41, "learning_rate": 1.3297300046380977e-05, "loss": 1.9795, "step": 16685 }, { "epoch": 0.41, "learning_rate": 1.329654635119585e-05, "loss": 2.0159, "step": 16686 }, { "epoch": 0.41, "learning_rate": 1.329579263500076e-05, "loss": 2.0624, "step": 16687 }, { "epoch": 0.41, "learning_rate": 1.329503889780051e-05, "loss": 2.3212, "step": 16688 }, { "epoch": 0.41, "learning_rate": 1.3294285139599898e-05, "loss": 2.203, "step": 16689 }, { "epoch": 0.41, "learning_rate": 1.329353136040374e-05, "loss": 2.1714, "step": 16690 }, { "epoch": 0.41, "learning_rate": 1.329277756021683e-05, "loss": 1.9949, "step": 16691 }, { "epoch": 0.41, "learning_rate": 1.3292023739043975e-05, "loss": 2.0245, "step": 16692 }, { "epoch": 0.41, "learning_rate": 1.3291269896889982e-05, "loss": 2.0227, "step": 16693 }, { "epoch": 0.41, "learning_rate": 1.3290516033759651e-05, "loss": 2.0578, "step": 16694 }, { "epoch": 0.41, "learning_rate": 1.328976214965779e-05, "loss": 1.8802, "step": 16695 }, { "epoch": 0.41, "learning_rate": 1.3289008244589206e-05, "loss": 2.1274, "step": 16696 }, { "epoch": 0.41, "learning_rate": 1.3288254318558697e-05, "loss": 2.1219, "step": 16697 }, { "epoch": 0.41, "learning_rate": 1.3287500371571073e-05, "loss": 2.0383, "step": 16698 }, { "epoch": 0.41, "learning_rate": 1.3286746403631137e-05, "loss": 2.1156, "step": 16699 }, { "epoch": 0.41, "learning_rate": 1.3285992414743692e-05, "loss": 2.1615, "step": 16700 }, { "epoch": 0.41, "learning_rate": 1.328523840491355e-05, "loss": 1.9583, "step": 16701 }, { "epoch": 0.41, "learning_rate": 1.3284484374145514e-05, "loss": 2.0327, "step": 16702 }, { "epoch": 0.41, "learning_rate": 1.3283730322444386e-05, "loss": 2.1281, "step": 16703 }, { "epoch": 0.41, "learning_rate": 1.3282976249814976e-05, "loss": 2.0865, "step": 16704 }, { "epoch": 0.41, "learning_rate": 1.3282222156262088e-05, "loss": 2.1089, "step": 16705 }, { "epoch": 0.41, "learning_rate": 1.3281468041790531e-05, "loss": 2.123, "step": 16706 }, { "epoch": 0.41, "learning_rate": 1.3280713906405107e-05, "loss": 2.1149, "step": 16707 }, { "epoch": 0.41, "learning_rate": 1.3279959750110622e-05, "loss": 2.0203, "step": 16708 }, { "epoch": 0.41, "learning_rate": 1.3279205572911884e-05, "loss": 2.0814, "step": 16709 }, { "epoch": 0.41, "learning_rate": 1.3278451374813703e-05, "loss": 2.0781, "step": 16710 }, { "epoch": 0.41, "learning_rate": 1.3277697155820884e-05, "loss": 2.1349, "step": 16711 }, { "epoch": 0.41, "learning_rate": 1.3276942915938232e-05, "loss": 2.1096, "step": 16712 }, { "epoch": 0.41, "learning_rate": 1.3276188655170553e-05, "loss": 1.879, "step": 16713 }, { "epoch": 0.41, "learning_rate": 1.3275434373522657e-05, "loss": 2.0945, "step": 16714 }, { "epoch": 0.41, "learning_rate": 1.327468007099935e-05, "loss": 1.9491, "step": 16715 }, { "epoch": 0.41, "learning_rate": 1.3273925747605443e-05, "loss": 1.9367, "step": 16716 }, { "epoch": 0.41, "learning_rate": 1.3273171403345737e-05, "loss": 1.9355, "step": 16717 }, { "epoch": 0.41, "learning_rate": 1.327241703822504e-05, "loss": 2.1895, "step": 16718 }, { "epoch": 0.41, "learning_rate": 1.3271662652248168e-05, "loss": 2.1576, "step": 16719 }, { "epoch": 0.41, "learning_rate": 1.3270908245419923e-05, "loss": 2.0776, "step": 16720 }, { "epoch": 0.41, "learning_rate": 1.3270153817745113e-05, "loss": 2.0771, "step": 16721 }, { "epoch": 0.41, "learning_rate": 1.3269399369228549e-05, "loss": 2.0453, "step": 16722 }, { "epoch": 0.41, "learning_rate": 1.3268644899875032e-05, "loss": 2.0836, "step": 16723 }, { "epoch": 0.41, "learning_rate": 1.3267890409689379e-05, "loss": 2.0305, "step": 16724 }, { "epoch": 0.41, "learning_rate": 1.3267135898676395e-05, "loss": 2.1366, "step": 16725 }, { "epoch": 0.41, "learning_rate": 1.326638136684089e-05, "loss": 2.128, "step": 16726 }, { "epoch": 0.41, "learning_rate": 1.3265626814187673e-05, "loss": 2.0449, "step": 16727 }, { "epoch": 0.41, "learning_rate": 1.3264872240721548e-05, "loss": 2.0804, "step": 16728 }, { "epoch": 0.41, "learning_rate": 1.3264117646447331e-05, "loss": 1.9397, "step": 16729 }, { "epoch": 0.41, "learning_rate": 1.3263363031369831e-05, "loss": 2.1361, "step": 16730 }, { "epoch": 0.41, "learning_rate": 1.3262608395493851e-05, "loss": 1.8217, "step": 16731 }, { "epoch": 0.41, "learning_rate": 1.3261853738824204e-05, "loss": 2.2113, "step": 16732 }, { "epoch": 0.41, "learning_rate": 1.3261099061365705e-05, "loss": 2.1759, "step": 16733 }, { "epoch": 0.41, "learning_rate": 1.3260344363123156e-05, "loss": 1.9642, "step": 16734 }, { "epoch": 0.41, "learning_rate": 1.3259589644101367e-05, "loss": 2.1567, "step": 16735 }, { "epoch": 0.41, "learning_rate": 1.3258834904305155e-05, "loss": 1.9942, "step": 16736 }, { "epoch": 0.41, "learning_rate": 1.3258080143739322e-05, "loss": 2.0362, "step": 16737 }, { "epoch": 0.41, "learning_rate": 1.3257325362408685e-05, "loss": 2.1795, "step": 16738 }, { "epoch": 0.41, "learning_rate": 1.3256570560318057e-05, "loss": 2.2083, "step": 16739 }, { "epoch": 0.41, "learning_rate": 1.3255815737472238e-05, "loss": 2.2002, "step": 16740 }, { "epoch": 0.41, "learning_rate": 1.3255060893876045e-05, "loss": 2.1294, "step": 16741 }, { "epoch": 0.41, "learning_rate": 1.3254306029534285e-05, "loss": 1.9968, "step": 16742 }, { "epoch": 0.41, "learning_rate": 1.3253551144451778e-05, "loss": 2.1274, "step": 16743 }, { "epoch": 0.41, "learning_rate": 1.3252796238633324e-05, "loss": 2.2161, "step": 16744 }, { "epoch": 0.41, "learning_rate": 1.3252041312083743e-05, "loss": 2.1282, "step": 16745 }, { "epoch": 0.41, "learning_rate": 1.3251286364807842e-05, "loss": 2.0413, "step": 16746 }, { "epoch": 0.41, "learning_rate": 1.3250531396810434e-05, "loss": 2.2486, "step": 16747 }, { "epoch": 0.41, "learning_rate": 1.324977640809633e-05, "loss": 2.1104, "step": 16748 }, { "epoch": 0.41, "learning_rate": 1.3249021398670341e-05, "loss": 2.0855, "step": 16749 }, { "epoch": 0.41, "learning_rate": 1.324826636853728e-05, "loss": 1.852, "step": 16750 }, { "epoch": 0.41, "learning_rate": 1.324751131770196e-05, "loss": 2.1569, "step": 16751 }, { "epoch": 0.41, "learning_rate": 1.3246756246169187e-05, "loss": 2.1343, "step": 16752 }, { "epoch": 0.41, "learning_rate": 1.3246001153943784e-05, "loss": 2.0814, "step": 16753 }, { "epoch": 0.41, "learning_rate": 1.3245246041030554e-05, "loss": 1.9334, "step": 16754 }, { "epoch": 0.41, "learning_rate": 1.3244490907434317e-05, "loss": 1.9921, "step": 16755 }, { "epoch": 0.41, "learning_rate": 1.3243735753159883e-05, "loss": 1.9734, "step": 16756 }, { "epoch": 0.41, "learning_rate": 1.324298057821206e-05, "loss": 2.0303, "step": 16757 }, { "epoch": 0.41, "learning_rate": 1.3242225382595667e-05, "loss": 2.0177, "step": 16758 }, { "epoch": 0.41, "learning_rate": 1.3241470166315514e-05, "loss": 1.863, "step": 16759 }, { "epoch": 0.41, "learning_rate": 1.3240714929376417e-05, "loss": 2.0175, "step": 16760 }, { "epoch": 0.41, "learning_rate": 1.3239959671783184e-05, "loss": 2.1503, "step": 16761 }, { "epoch": 0.41, "learning_rate": 1.3239204393540634e-05, "loss": 2.0951, "step": 16762 }, { "epoch": 0.41, "learning_rate": 1.323844909465358e-05, "loss": 2.0774, "step": 16763 }, { "epoch": 0.41, "learning_rate": 1.3237693775126834e-05, "loss": 2.0389, "step": 16764 }, { "epoch": 0.41, "learning_rate": 1.3236938434965212e-05, "loss": 2.0548, "step": 16765 }, { "epoch": 0.41, "learning_rate": 1.3236183074173524e-05, "loss": 2.1404, "step": 16766 }, { "epoch": 0.41, "learning_rate": 1.3235427692756586e-05, "loss": 1.9948, "step": 16767 }, { "epoch": 0.41, "learning_rate": 1.3234672290719215e-05, "loss": 2.0478, "step": 16768 }, { "epoch": 0.41, "learning_rate": 1.3233916868066223e-05, "loss": 2.1277, "step": 16769 }, { "epoch": 0.41, "learning_rate": 1.3233161424802426e-05, "loss": 1.971, "step": 16770 }, { "epoch": 0.41, "learning_rate": 1.3232405960932633e-05, "loss": 1.9567, "step": 16771 }, { "epoch": 0.41, "learning_rate": 1.3231650476461666e-05, "loss": 2.1113, "step": 16772 }, { "epoch": 0.41, "learning_rate": 1.3230894971394339e-05, "loss": 2.0192, "step": 16773 }, { "epoch": 0.41, "learning_rate": 1.3230139445735466e-05, "loss": 1.9485, "step": 16774 }, { "epoch": 0.41, "learning_rate": 1.3229383899489861e-05, "loss": 2.0961, "step": 16775 }, { "epoch": 0.41, "learning_rate": 1.3228628332662336e-05, "loss": 2.1988, "step": 16776 }, { "epoch": 0.41, "learning_rate": 1.3227872745257715e-05, "loss": 2.1891, "step": 16777 }, { "epoch": 0.41, "learning_rate": 1.3227117137280807e-05, "loss": 2.0023, "step": 16778 }, { "epoch": 0.41, "learning_rate": 1.3226361508736429e-05, "loss": 2.2973, "step": 16779 }, { "epoch": 0.41, "learning_rate": 1.32256058596294e-05, "loss": 1.8341, "step": 16780 }, { "epoch": 0.41, "learning_rate": 1.322485018996453e-05, "loss": 1.8137, "step": 16781 }, { "epoch": 0.41, "learning_rate": 1.3224094499746643e-05, "loss": 2.0617, "step": 16782 }, { "epoch": 0.41, "learning_rate": 1.3223338788980551e-05, "loss": 2.1362, "step": 16783 }, { "epoch": 0.41, "learning_rate": 1.3222583057671068e-05, "loss": 2.0373, "step": 16784 }, { "epoch": 0.41, "learning_rate": 1.3221827305823016e-05, "loss": 2.2374, "step": 16785 }, { "epoch": 0.41, "learning_rate": 1.3221071533441203e-05, "loss": 2.1136, "step": 16786 }, { "epoch": 0.41, "learning_rate": 1.3220315740530455e-05, "loss": 2.0741, "step": 16787 }, { "epoch": 0.41, "learning_rate": 1.3219559927095586e-05, "loss": 2.0707, "step": 16788 }, { "epoch": 0.41, "learning_rate": 1.321880409314141e-05, "loss": 2.0082, "step": 16789 }, { "epoch": 0.41, "learning_rate": 1.3218048238672747e-05, "loss": 2.0492, "step": 16790 }, { "epoch": 0.41, "learning_rate": 1.3217292363694415e-05, "loss": 2.0051, "step": 16791 }, { "epoch": 0.41, "learning_rate": 1.3216536468211232e-05, "loss": 1.9787, "step": 16792 }, { "epoch": 0.41, "learning_rate": 1.3215780552228012e-05, "loss": 1.902, "step": 16793 }, { "epoch": 0.41, "learning_rate": 1.3215024615749576e-05, "loss": 2.0983, "step": 16794 }, { "epoch": 0.41, "learning_rate": 1.3214268658780738e-05, "loss": 2.0711, "step": 16795 }, { "epoch": 0.41, "learning_rate": 1.3213512681326319e-05, "loss": 1.9949, "step": 16796 }, { "epoch": 0.41, "learning_rate": 1.3212756683391135e-05, "loss": 2.0499, "step": 16797 }, { "epoch": 0.41, "learning_rate": 1.3212000664980008e-05, "loss": 2.0334, "step": 16798 }, { "epoch": 0.41, "learning_rate": 1.3211244626097754e-05, "loss": 2.0018, "step": 16799 }, { "epoch": 0.41, "learning_rate": 1.3210488566749188e-05, "loss": 1.9201, "step": 16800 }, { "epoch": 0.41, "learning_rate": 1.3209732486939137e-05, "loss": 2.0721, "step": 16801 }, { "epoch": 0.41, "learning_rate": 1.3208976386672413e-05, "loss": 2.0527, "step": 16802 }, { "epoch": 0.41, "learning_rate": 1.3208220265953839e-05, "loss": 2.1129, "step": 16803 }, { "epoch": 0.41, "learning_rate": 1.320746412478823e-05, "loss": 2.0647, "step": 16804 }, { "epoch": 0.41, "learning_rate": 1.3206707963180403e-05, "loss": 2.1051, "step": 16805 }, { "epoch": 0.41, "learning_rate": 1.3205951781135183e-05, "loss": 1.7937, "step": 16806 }, { "epoch": 0.41, "learning_rate": 1.320519557865739e-05, "loss": 2.2264, "step": 16807 }, { "epoch": 0.41, "learning_rate": 1.320443935575184e-05, "loss": 2.0078, "step": 16808 }, { "epoch": 0.41, "learning_rate": 1.3203683112423353e-05, "loss": 1.8714, "step": 16809 }, { "epoch": 0.41, "learning_rate": 1.320292684867675e-05, "loss": 1.9853, "step": 16810 }, { "epoch": 0.41, "learning_rate": 1.3202170564516853e-05, "loss": 2.192, "step": 16811 }, { "epoch": 0.41, "learning_rate": 1.3201414259948476e-05, "loss": 2.0174, "step": 16812 }, { "epoch": 0.41, "learning_rate": 1.3200657934976442e-05, "loss": 2.0582, "step": 16813 }, { "epoch": 0.41, "learning_rate": 1.3199901589605572e-05, "loss": 2.1235, "step": 16814 }, { "epoch": 0.41, "learning_rate": 1.3199145223840688e-05, "loss": 2.0694, "step": 16815 }, { "epoch": 0.41, "learning_rate": 1.3198388837686609e-05, "loss": 1.9806, "step": 16816 }, { "epoch": 0.41, "learning_rate": 1.3197632431148155e-05, "loss": 1.9883, "step": 16817 }, { "epoch": 0.41, "learning_rate": 1.3196876004230149e-05, "loss": 2.0172, "step": 16818 }, { "epoch": 0.41, "learning_rate": 1.3196119556937408e-05, "loss": 2.0767, "step": 16819 }, { "epoch": 0.41, "learning_rate": 1.3195363089274755e-05, "loss": 2.012, "step": 16820 }, { "epoch": 0.41, "learning_rate": 1.3194606601247016e-05, "loss": 1.9427, "step": 16821 }, { "epoch": 0.41, "learning_rate": 1.3193850092859005e-05, "loss": 2.0614, "step": 16822 }, { "epoch": 0.41, "learning_rate": 1.3193093564115544e-05, "loss": 2.0583, "step": 16823 }, { "epoch": 0.41, "learning_rate": 1.3192337015021459e-05, "loss": 2.0786, "step": 16824 }, { "epoch": 0.41, "learning_rate": 1.3191580445581571e-05, "loss": 2.2195, "step": 16825 }, { "epoch": 0.41, "learning_rate": 1.3190823855800701e-05, "loss": 1.9722, "step": 16826 }, { "epoch": 0.41, "learning_rate": 1.319006724568367e-05, "loss": 2.0715, "step": 16827 }, { "epoch": 0.41, "learning_rate": 1.3189310615235302e-05, "loss": 2.147, "step": 16828 }, { "epoch": 0.41, "learning_rate": 1.3188553964460415e-05, "loss": 1.9976, "step": 16829 }, { "epoch": 0.41, "learning_rate": 1.3187797293363838e-05, "loss": 1.8925, "step": 16830 }, { "epoch": 0.41, "learning_rate": 1.318704060195039e-05, "loss": 2.1434, "step": 16831 }, { "epoch": 0.41, "learning_rate": 1.318628389022489e-05, "loss": 2.083, "step": 16832 }, { "epoch": 0.41, "learning_rate": 1.3185527158192166e-05, "loss": 2.11, "step": 16833 }, { "epoch": 0.41, "learning_rate": 1.3184770405857039e-05, "loss": 2.1179, "step": 16834 }, { "epoch": 0.41, "learning_rate": 1.3184013633224333e-05, "loss": 2.0857, "step": 16835 }, { "epoch": 0.41, "learning_rate": 1.3183256840298872e-05, "loss": 2.0708, "step": 16836 }, { "epoch": 0.42, "learning_rate": 1.3182500027085476e-05, "loss": 2.0466, "step": 16837 }, { "epoch": 0.42, "learning_rate": 1.3181743193588972e-05, "loss": 2.1366, "step": 16838 }, { "epoch": 0.42, "learning_rate": 1.3180986339814177e-05, "loss": 1.826, "step": 16839 }, { "epoch": 0.42, "learning_rate": 1.3180229465765924e-05, "loss": 2.0446, "step": 16840 }, { "epoch": 0.42, "learning_rate": 1.317947257144903e-05, "loss": 2.1495, "step": 16841 }, { "epoch": 0.42, "learning_rate": 1.3178715656868322e-05, "loss": 1.9005, "step": 16842 }, { "epoch": 0.42, "learning_rate": 1.3177958722028623e-05, "loss": 2.0619, "step": 16843 }, { "epoch": 0.42, "learning_rate": 1.3177201766934758e-05, "loss": 2.2134, "step": 16844 }, { "epoch": 0.42, "learning_rate": 1.317644479159155e-05, "loss": 2.2191, "step": 16845 }, { "epoch": 0.42, "learning_rate": 1.3175687796003824e-05, "loss": 2.1645, "step": 16846 }, { "epoch": 0.42, "learning_rate": 1.3174930780176407e-05, "loss": 2.0586, "step": 16847 }, { "epoch": 0.42, "learning_rate": 1.317417374411412e-05, "loss": 2.093, "step": 16848 }, { "epoch": 0.42, "learning_rate": 1.3173416687821787e-05, "loss": 2.0937, "step": 16849 }, { "epoch": 0.42, "learning_rate": 1.3172659611304237e-05, "loss": 2.1951, "step": 16850 }, { "epoch": 0.42, "learning_rate": 1.3171902514566292e-05, "loss": 2.0932, "step": 16851 }, { "epoch": 0.42, "learning_rate": 1.3171145397612781e-05, "loss": 2.1429, "step": 16852 }, { "epoch": 0.42, "learning_rate": 1.3170388260448525e-05, "loss": 2.1801, "step": 16853 }, { "epoch": 0.42, "learning_rate": 1.3169631103078353e-05, "loss": 1.9813, "step": 16854 }, { "epoch": 0.42, "learning_rate": 1.3168873925507089e-05, "loss": 1.9681, "step": 16855 }, { "epoch": 0.42, "learning_rate": 1.316811672773956e-05, "loss": 2.1069, "step": 16856 }, { "epoch": 0.42, "learning_rate": 1.316735950978059e-05, "loss": 2.1482, "step": 16857 }, { "epoch": 0.42, "learning_rate": 1.3166602271635003e-05, "loss": 2.0624, "step": 16858 }, { "epoch": 0.42, "learning_rate": 1.3165845013307628e-05, "loss": 2.0545, "step": 16859 }, { "epoch": 0.42, "learning_rate": 1.3165087734803292e-05, "loss": 2.0986, "step": 16860 }, { "epoch": 0.42, "learning_rate": 1.3164330436126821e-05, "loss": 2.0956, "step": 16861 }, { "epoch": 0.42, "learning_rate": 1.3163573117283041e-05, "loss": 2.0394, "step": 16862 }, { "epoch": 0.42, "learning_rate": 1.3162815778276779e-05, "loss": 2.1741, "step": 16863 }, { "epoch": 0.42, "learning_rate": 1.316205841911286e-05, "loss": 2.0761, "step": 16864 }, { "epoch": 0.42, "learning_rate": 1.3161301039796113e-05, "loss": 2.2223, "step": 16865 }, { "epoch": 0.42, "learning_rate": 1.3160543640331365e-05, "loss": 2.0092, "step": 16866 }, { "epoch": 0.42, "learning_rate": 1.315978622072344e-05, "loss": 1.9941, "step": 16867 }, { "epoch": 0.42, "learning_rate": 1.3159028780977168e-05, "loss": 1.9843, "step": 16868 }, { "epoch": 0.42, "learning_rate": 1.3158271321097376e-05, "loss": 1.9721, "step": 16869 }, { "epoch": 0.42, "learning_rate": 1.3157513841088893e-05, "loss": 2.0609, "step": 16870 }, { "epoch": 0.42, "learning_rate": 1.3156756340956547e-05, "loss": 2.1549, "step": 16871 }, { "epoch": 0.42, "learning_rate": 1.3155998820705163e-05, "loss": 1.9902, "step": 16872 }, { "epoch": 0.42, "learning_rate": 1.3155241280339565e-05, "loss": 1.8928, "step": 16873 }, { "epoch": 0.42, "learning_rate": 1.3154483719864592e-05, "loss": 2.2387, "step": 16874 }, { "epoch": 0.42, "learning_rate": 1.3153726139285066e-05, "loss": 1.967, "step": 16875 }, { "epoch": 0.42, "learning_rate": 1.315296853860581e-05, "loss": 2.1355, "step": 16876 }, { "epoch": 0.42, "learning_rate": 1.3152210917831662e-05, "loss": 1.9477, "step": 16877 }, { "epoch": 0.42, "learning_rate": 1.3151453276967447e-05, "loss": 2.1271, "step": 16878 }, { "epoch": 0.42, "learning_rate": 1.3150695616017991e-05, "loss": 1.9913, "step": 16879 }, { "epoch": 0.42, "learning_rate": 1.3149937934988129e-05, "loss": 1.8989, "step": 16880 }, { "epoch": 0.42, "learning_rate": 1.3149180233882683e-05, "loss": 1.9396, "step": 16881 }, { "epoch": 0.42, "learning_rate": 1.3148422512706486e-05, "loss": 2.2222, "step": 16882 }, { "epoch": 0.42, "learning_rate": 1.3147664771464369e-05, "loss": 2.1483, "step": 16883 }, { "epoch": 0.42, "learning_rate": 1.3146907010161153e-05, "loss": 2.2135, "step": 16884 }, { "epoch": 0.42, "learning_rate": 1.3146149228801677e-05, "loss": 2.191, "step": 16885 }, { "epoch": 0.42, "learning_rate": 1.3145391427390765e-05, "loss": 1.9793, "step": 16886 }, { "epoch": 0.42, "learning_rate": 1.3144633605933245e-05, "loss": 2.1495, "step": 16887 }, { "epoch": 0.42, "learning_rate": 1.3143875764433958e-05, "loss": 2.0208, "step": 16888 }, { "epoch": 0.42, "learning_rate": 1.3143117902897722e-05, "loss": 2.0124, "step": 16889 }, { "epoch": 0.42, "learning_rate": 1.3142360021329372e-05, "loss": 2.2659, "step": 16890 }, { "epoch": 0.42, "learning_rate": 1.3141602119733738e-05, "loss": 2.1636, "step": 16891 }, { "epoch": 0.42, "learning_rate": 1.3140844198115645e-05, "loss": 1.8356, "step": 16892 }, { "epoch": 0.42, "learning_rate": 1.3140086256479933e-05, "loss": 1.9241, "step": 16893 }, { "epoch": 0.42, "learning_rate": 1.3139328294831426e-05, "loss": 2.0203, "step": 16894 }, { "epoch": 0.42, "learning_rate": 1.3138570313174958e-05, "loss": 2.2632, "step": 16895 }, { "epoch": 0.42, "learning_rate": 1.3137812311515361e-05, "loss": 2.1356, "step": 16896 }, { "epoch": 0.42, "learning_rate": 1.3137054289857459e-05, "loss": 2.0784, "step": 16897 }, { "epoch": 0.42, "learning_rate": 1.313629624820609e-05, "loss": 1.8735, "step": 16898 }, { "epoch": 0.42, "learning_rate": 1.3135538186566083e-05, "loss": 2.0825, "step": 16899 }, { "epoch": 0.42, "learning_rate": 1.3134780104942271e-05, "loss": 1.9737, "step": 16900 }, { "epoch": 0.42, "learning_rate": 1.313402200333948e-05, "loss": 2.0429, "step": 16901 }, { "epoch": 0.42, "learning_rate": 1.3133263881762547e-05, "loss": 1.9337, "step": 16902 }, { "epoch": 0.42, "learning_rate": 1.3132505740216301e-05, "loss": 1.913, "step": 16903 }, { "epoch": 0.42, "learning_rate": 1.3131747578705578e-05, "loss": 1.8982, "step": 16904 }, { "epoch": 0.42, "learning_rate": 1.3130989397235205e-05, "loss": 2.1247, "step": 16905 }, { "epoch": 0.42, "learning_rate": 1.3130231195810019e-05, "loss": 2.0237, "step": 16906 }, { "epoch": 0.42, "learning_rate": 1.3129472974434845e-05, "loss": 2.0549, "step": 16907 }, { "epoch": 0.42, "learning_rate": 1.3128714733114522e-05, "loss": 2.2572, "step": 16908 }, { "epoch": 0.42, "learning_rate": 1.3127956471853883e-05, "loss": 2.0824, "step": 16909 }, { "epoch": 0.42, "learning_rate": 1.3127198190657753e-05, "loss": 1.9526, "step": 16910 }, { "epoch": 0.42, "learning_rate": 1.3126439889530972e-05, "loss": 1.9412, "step": 16911 }, { "epoch": 0.42, "learning_rate": 1.312568156847837e-05, "loss": 1.9606, "step": 16912 }, { "epoch": 0.42, "learning_rate": 1.3124923227504782e-05, "loss": 2.1266, "step": 16913 }, { "epoch": 0.42, "learning_rate": 1.3124164866615038e-05, "loss": 2.0652, "step": 16914 }, { "epoch": 0.42, "learning_rate": 1.3123406485813978e-05, "loss": 2.233, "step": 16915 }, { "epoch": 0.42, "learning_rate": 1.3122648085106425e-05, "loss": 2.098, "step": 16916 }, { "epoch": 0.42, "learning_rate": 1.312188966449722e-05, "loss": 1.9741, "step": 16917 }, { "epoch": 0.42, "learning_rate": 1.3121131223991197e-05, "loss": 1.9942, "step": 16918 }, { "epoch": 0.42, "learning_rate": 1.3120372763593182e-05, "loss": 2.0049, "step": 16919 }, { "epoch": 0.42, "learning_rate": 1.3119614283308022e-05, "loss": 1.9858, "step": 16920 }, { "epoch": 0.42, "learning_rate": 1.3118855783140538e-05, "loss": 2.2915, "step": 16921 }, { "epoch": 0.42, "learning_rate": 1.3118097263095573e-05, "loss": 1.9782, "step": 16922 }, { "epoch": 0.42, "learning_rate": 1.3117338723177956e-05, "loss": 2.1109, "step": 16923 }, { "epoch": 0.42, "learning_rate": 1.3116580163392524e-05, "loss": 2.1503, "step": 16924 }, { "epoch": 0.42, "learning_rate": 1.311582158374411e-05, "loss": 2.0371, "step": 16925 }, { "epoch": 0.42, "learning_rate": 1.3115062984237552e-05, "loss": 1.9807, "step": 16926 }, { "epoch": 0.42, "learning_rate": 1.3114304364877679e-05, "loss": 2.1683, "step": 16927 }, { "epoch": 0.42, "learning_rate": 1.3113545725669332e-05, "loss": 2.2006, "step": 16928 }, { "epoch": 0.42, "learning_rate": 1.3112787066617343e-05, "loss": 2.2346, "step": 16929 }, { "epoch": 0.42, "learning_rate": 1.3112028387726547e-05, "loss": 2.2412, "step": 16930 }, { "epoch": 0.42, "learning_rate": 1.3111269689001778e-05, "loss": 1.9865, "step": 16931 }, { "epoch": 0.42, "learning_rate": 1.3110510970447875e-05, "loss": 2.0668, "step": 16932 }, { "epoch": 0.42, "learning_rate": 1.3109752232069675e-05, "loss": 1.9578, "step": 16933 }, { "epoch": 0.42, "learning_rate": 1.310899347387201e-05, "loss": 2.1705, "step": 16934 }, { "epoch": 0.42, "learning_rate": 1.3108234695859713e-05, "loss": 2.0445, "step": 16935 }, { "epoch": 0.42, "learning_rate": 1.3107475898037625e-05, "loss": 2.0977, "step": 16936 }, { "epoch": 0.42, "learning_rate": 1.310671708041058e-05, "loss": 2.0139, "step": 16937 }, { "epoch": 0.42, "learning_rate": 1.3105958242983411e-05, "loss": 2.0571, "step": 16938 }, { "epoch": 0.42, "learning_rate": 1.3105199385760963e-05, "loss": 2.1014, "step": 16939 }, { "epoch": 0.42, "learning_rate": 1.3104440508748067e-05, "loss": 2.1801, "step": 16940 }, { "epoch": 0.42, "learning_rate": 1.3103681611949557e-05, "loss": 2.2282, "step": 16941 }, { "epoch": 0.42, "learning_rate": 1.3102922695370275e-05, "loss": 2.1283, "step": 16942 }, { "epoch": 0.42, "learning_rate": 1.3102163759015057e-05, "loss": 2.0892, "step": 16943 }, { "epoch": 0.42, "learning_rate": 1.3101404802888738e-05, "loss": 2.1258, "step": 16944 }, { "epoch": 0.42, "learning_rate": 1.310064582699615e-05, "loss": 2.1381, "step": 16945 }, { "epoch": 0.42, "learning_rate": 1.3099886831342141e-05, "loss": 2.1154, "step": 16946 }, { "epoch": 0.42, "learning_rate": 1.3099127815931541e-05, "loss": 1.9896, "step": 16947 }, { "epoch": 0.42, "learning_rate": 1.3098368780769191e-05, "loss": 2.128, "step": 16948 }, { "epoch": 0.42, "learning_rate": 1.3097609725859925e-05, "loss": 2.2026, "step": 16949 }, { "epoch": 0.42, "learning_rate": 1.3096850651208584e-05, "loss": 2.1256, "step": 16950 }, { "epoch": 0.42, "learning_rate": 1.3096091556820003e-05, "loss": 2.0982, "step": 16951 }, { "epoch": 0.42, "learning_rate": 1.3095332442699024e-05, "loss": 2.2419, "step": 16952 }, { "epoch": 0.42, "learning_rate": 1.3094573308850482e-05, "loss": 2.0257, "step": 16953 }, { "epoch": 0.42, "learning_rate": 1.3093814155279215e-05, "loss": 2.0316, "step": 16954 }, { "epoch": 0.42, "learning_rate": 1.3093054981990064e-05, "loss": 2.116, "step": 16955 }, { "epoch": 0.42, "learning_rate": 1.3092295788987861e-05, "loss": 2.1642, "step": 16956 }, { "epoch": 0.42, "learning_rate": 1.3091536576277453e-05, "loss": 2.0865, "step": 16957 }, { "epoch": 0.42, "learning_rate": 1.3090777343863674e-05, "loss": 1.9752, "step": 16958 }, { "epoch": 0.42, "learning_rate": 1.3090018091751364e-05, "loss": 2.1899, "step": 16959 }, { "epoch": 0.42, "learning_rate": 1.308925881994536e-05, "loss": 2.3395, "step": 16960 }, { "epoch": 0.42, "learning_rate": 1.3088499528450504e-05, "loss": 1.9807, "step": 16961 }, { "epoch": 0.42, "learning_rate": 1.3087740217271636e-05, "loss": 1.9717, "step": 16962 }, { "epoch": 0.42, "learning_rate": 1.308698088641359e-05, "loss": 1.9565, "step": 16963 }, { "epoch": 0.42, "learning_rate": 1.3086221535881212e-05, "loss": 2.0343, "step": 16964 }, { "epoch": 0.42, "learning_rate": 1.3085462165679334e-05, "loss": 2.3615, "step": 16965 }, { "epoch": 0.42, "learning_rate": 1.3084702775812805e-05, "loss": 1.99, "step": 16966 }, { "epoch": 0.42, "learning_rate": 1.3083943366286458e-05, "loss": 2.3139, "step": 16967 }, { "epoch": 0.42, "learning_rate": 1.3083183937105135e-05, "loss": 2.2954, "step": 16968 }, { "epoch": 0.42, "learning_rate": 1.3082424488273673e-05, "loss": 2.1167, "step": 16969 }, { "epoch": 0.42, "learning_rate": 1.3081665019796922e-05, "loss": 1.9075, "step": 16970 }, { "epoch": 0.42, "learning_rate": 1.308090553167971e-05, "loss": 1.902, "step": 16971 }, { "epoch": 0.42, "learning_rate": 1.3080146023926883e-05, "loss": 1.9963, "step": 16972 }, { "epoch": 0.42, "learning_rate": 1.3079386496543282e-05, "loss": 2.1789, "step": 16973 }, { "epoch": 0.42, "learning_rate": 1.3078626949533747e-05, "loss": 1.9941, "step": 16974 }, { "epoch": 0.42, "learning_rate": 1.3077867382903123e-05, "loss": 2.1496, "step": 16975 }, { "epoch": 0.42, "learning_rate": 1.3077107796656245e-05, "loss": 1.985, "step": 16976 }, { "epoch": 0.42, "learning_rate": 1.3076348190797956e-05, "loss": 1.9188, "step": 16977 }, { "epoch": 0.42, "learning_rate": 1.3075588565333099e-05, "loss": 2.1642, "step": 16978 }, { "epoch": 0.42, "learning_rate": 1.3074828920266508e-05, "loss": 2.0125, "step": 16979 }, { "epoch": 0.42, "learning_rate": 1.3074069255603036e-05, "loss": 2.1317, "step": 16980 }, { "epoch": 0.42, "learning_rate": 1.3073309571347516e-05, "loss": 2.0882, "step": 16981 }, { "epoch": 0.42, "learning_rate": 1.3072549867504796e-05, "loss": 2.0971, "step": 16982 }, { "epoch": 0.42, "learning_rate": 1.307179014407971e-05, "loss": 2.0836, "step": 16983 }, { "epoch": 0.42, "learning_rate": 1.3071030401077107e-05, "loss": 1.9151, "step": 16984 }, { "epoch": 0.42, "learning_rate": 1.3070270638501825e-05, "loss": 2.0842, "step": 16985 }, { "epoch": 0.42, "learning_rate": 1.3069510856358709e-05, "loss": 1.9326, "step": 16986 }, { "epoch": 0.42, "learning_rate": 1.3068751054652599e-05, "loss": 1.9296, "step": 16987 }, { "epoch": 0.42, "learning_rate": 1.3067991233388343e-05, "loss": 2.2278, "step": 16988 }, { "epoch": 0.42, "learning_rate": 1.3067231392570771e-05, "loss": 1.9915, "step": 16989 }, { "epoch": 0.42, "learning_rate": 1.3066471532204736e-05, "loss": 1.996, "step": 16990 }, { "epoch": 0.42, "learning_rate": 1.3065711652295081e-05, "loss": 2.1289, "step": 16991 }, { "epoch": 0.42, "learning_rate": 1.3064951752846646e-05, "loss": 2.0901, "step": 16992 }, { "epoch": 0.42, "learning_rate": 1.3064191833864275e-05, "loss": 2.0786, "step": 16993 }, { "epoch": 0.42, "learning_rate": 1.306343189535281e-05, "loss": 2.1023, "step": 16994 }, { "epoch": 0.42, "learning_rate": 1.3062671937317095e-05, "loss": 2.0759, "step": 16995 }, { "epoch": 0.42, "learning_rate": 1.3061911959761977e-05, "loss": 1.9506, "step": 16996 }, { "epoch": 0.42, "learning_rate": 1.3061151962692293e-05, "loss": 2.0996, "step": 16997 }, { "epoch": 0.42, "learning_rate": 1.306039194611289e-05, "loss": 2.0616, "step": 16998 }, { "epoch": 0.42, "learning_rate": 1.305963191002861e-05, "loss": 1.9268, "step": 16999 }, { "epoch": 0.42, "learning_rate": 1.30588718544443e-05, "loss": 2.1738, "step": 17000 }, { "epoch": 0.42, "learning_rate": 1.3058111779364806e-05, "loss": 2.1265, "step": 17001 }, { "epoch": 0.42, "learning_rate": 1.3057351684794966e-05, "loss": 1.9291, "step": 17002 }, { "epoch": 0.42, "learning_rate": 1.3056591570739628e-05, "loss": 2.061, "step": 17003 }, { "epoch": 0.42, "learning_rate": 1.3055831437203636e-05, "loss": 2.2732, "step": 17004 }, { "epoch": 0.42, "learning_rate": 1.3055071284191835e-05, "loss": 2.1734, "step": 17005 }, { "epoch": 0.42, "learning_rate": 1.3054311111709068e-05, "loss": 2.0337, "step": 17006 }, { "epoch": 0.42, "learning_rate": 1.305355091976018e-05, "loss": 2.1263, "step": 17007 }, { "epoch": 0.42, "learning_rate": 1.3052790708350016e-05, "loss": 2.0422, "step": 17008 }, { "epoch": 0.42, "learning_rate": 1.3052030477483425e-05, "loss": 2.0638, "step": 17009 }, { "epoch": 0.42, "learning_rate": 1.3051270227165249e-05, "loss": 2.164, "step": 17010 }, { "epoch": 0.42, "learning_rate": 1.3050509957400332e-05, "loss": 2.1274, "step": 17011 }, { "epoch": 0.42, "learning_rate": 1.3049749668193523e-05, "loss": 1.9908, "step": 17012 }, { "epoch": 0.42, "learning_rate": 1.3048989359549662e-05, "loss": 2.1688, "step": 17013 }, { "epoch": 0.42, "learning_rate": 1.3048229031473602e-05, "loss": 2.1075, "step": 17014 }, { "epoch": 0.42, "learning_rate": 1.3047468683970183e-05, "loss": 2.0232, "step": 17015 }, { "epoch": 0.42, "learning_rate": 1.3046708317044253e-05, "loss": 2.2228, "step": 17016 }, { "epoch": 0.42, "learning_rate": 1.3045947930700658e-05, "loss": 2.244, "step": 17017 }, { "epoch": 0.42, "learning_rate": 1.3045187524944242e-05, "loss": 1.9986, "step": 17018 }, { "epoch": 0.42, "learning_rate": 1.3044427099779857e-05, "loss": 2.04, "step": 17019 }, { "epoch": 0.42, "learning_rate": 1.3043666655212345e-05, "loss": 2.1993, "step": 17020 }, { "epoch": 0.42, "learning_rate": 1.3042906191246555e-05, "loss": 2.0577, "step": 17021 }, { "epoch": 0.42, "learning_rate": 1.304214570788733e-05, "loss": 2.0333, "step": 17022 }, { "epoch": 0.42, "learning_rate": 1.3041385205139519e-05, "loss": 2.1551, "step": 17023 }, { "epoch": 0.42, "learning_rate": 1.3040624683007969e-05, "loss": 1.9671, "step": 17024 }, { "epoch": 0.42, "learning_rate": 1.3039864141497525e-05, "loss": 1.9692, "step": 17025 }, { "epoch": 0.42, "learning_rate": 1.303910358061304e-05, "loss": 1.9357, "step": 17026 }, { "epoch": 0.42, "learning_rate": 1.3038343000359355e-05, "loss": 2.0068, "step": 17027 }, { "epoch": 0.42, "learning_rate": 1.303758240074132e-05, "loss": 2.0763, "step": 17028 }, { "epoch": 0.42, "learning_rate": 1.3036821781763784e-05, "loss": 2.0484, "step": 17029 }, { "epoch": 0.42, "learning_rate": 1.3036061143431593e-05, "loss": 2.1417, "step": 17030 }, { "epoch": 0.42, "learning_rate": 1.3035300485749596e-05, "loss": 2.1221, "step": 17031 }, { "epoch": 0.42, "learning_rate": 1.3034539808722637e-05, "loss": 2.1538, "step": 17032 }, { "epoch": 0.42, "learning_rate": 1.3033779112355568e-05, "loss": 2.0218, "step": 17033 }, { "epoch": 0.42, "learning_rate": 1.3033018396653235e-05, "loss": 2.244, "step": 17034 }, { "epoch": 0.42, "learning_rate": 1.3032257661620489e-05, "loss": 2.1572, "step": 17035 }, { "epoch": 0.42, "learning_rate": 1.3031496907262175e-05, "loss": 1.9034, "step": 17036 }, { "epoch": 0.42, "learning_rate": 1.3030736133583144e-05, "loss": 2.0156, "step": 17037 }, { "epoch": 0.42, "learning_rate": 1.3029975340588245e-05, "loss": 1.9539, "step": 17038 }, { "epoch": 0.42, "learning_rate": 1.3029214528282325e-05, "loss": 2.1704, "step": 17039 }, { "epoch": 0.42, "learning_rate": 1.3028453696670236e-05, "loss": 2.0979, "step": 17040 }, { "epoch": 0.42, "learning_rate": 1.3027692845756823e-05, "loss": 2.2483, "step": 17041 }, { "epoch": 0.42, "learning_rate": 1.3026931975546933e-05, "loss": 2.0142, "step": 17042 }, { "epoch": 0.42, "learning_rate": 1.302617108604542e-05, "loss": 2.027, "step": 17043 }, { "epoch": 0.42, "learning_rate": 1.3025410177257136e-05, "loss": 2.1627, "step": 17044 }, { "epoch": 0.42, "learning_rate": 1.3024649249186927e-05, "loss": 2.173, "step": 17045 }, { "epoch": 0.42, "learning_rate": 1.3023888301839641e-05, "loss": 2.0385, "step": 17046 }, { "epoch": 0.42, "learning_rate": 1.3023127335220129e-05, "loss": 1.9, "step": 17047 }, { "epoch": 0.42, "learning_rate": 1.3022366349333241e-05, "loss": 2.0106, "step": 17048 }, { "epoch": 0.42, "learning_rate": 1.302160534418383e-05, "loss": 2.1903, "step": 17049 }, { "epoch": 0.42, "learning_rate": 1.3020844319776741e-05, "loss": 2.0174, "step": 17050 }, { "epoch": 0.42, "learning_rate": 1.3020083276116828e-05, "loss": 1.9535, "step": 17051 }, { "epoch": 0.42, "learning_rate": 1.3019322213208938e-05, "loss": 2.1147, "step": 17052 }, { "epoch": 0.42, "learning_rate": 1.3018561131057926e-05, "loss": 2.048, "step": 17053 }, { "epoch": 0.42, "learning_rate": 1.301780002966864e-05, "loss": 2.1857, "step": 17054 }, { "epoch": 0.42, "learning_rate": 1.301703890904593e-05, "loss": 2.0905, "step": 17055 }, { "epoch": 0.42, "learning_rate": 1.3016277769194648e-05, "loss": 2.0975, "step": 17056 }, { "epoch": 0.42, "learning_rate": 1.3015516610119643e-05, "loss": 2.1502, "step": 17057 }, { "epoch": 0.42, "learning_rate": 1.3014755431825772e-05, "loss": 1.9998, "step": 17058 }, { "epoch": 0.42, "learning_rate": 1.301399423431788e-05, "loss": 2.0994, "step": 17059 }, { "epoch": 0.42, "learning_rate": 1.3013233017600819e-05, "loss": 2.008, "step": 17060 }, { "epoch": 0.42, "learning_rate": 1.3012471781679441e-05, "loss": 2.0026, "step": 17061 }, { "epoch": 0.42, "learning_rate": 1.3011710526558601e-05, "loss": 2.1297, "step": 17062 }, { "epoch": 0.42, "learning_rate": 1.301094925224315e-05, "loss": 1.967, "step": 17063 }, { "epoch": 0.42, "learning_rate": 1.3010187958737937e-05, "loss": 1.975, "step": 17064 }, { "epoch": 0.42, "learning_rate": 1.3009426646047816e-05, "loss": 2.1581, "step": 17065 }, { "epoch": 0.42, "learning_rate": 1.3008665314177633e-05, "loss": 1.9782, "step": 17066 }, { "epoch": 0.42, "learning_rate": 1.3007903963132251e-05, "loss": 1.967, "step": 17067 }, { "epoch": 0.42, "learning_rate": 1.3007142592916515e-05, "loss": 2.1118, "step": 17068 }, { "epoch": 0.42, "learning_rate": 1.300638120353528e-05, "loss": 2.0972, "step": 17069 }, { "epoch": 0.42, "learning_rate": 1.3005619794993396e-05, "loss": 2.1656, "step": 17070 }, { "epoch": 0.42, "learning_rate": 1.3004858367295717e-05, "loss": 2.0696, "step": 17071 }, { "epoch": 0.42, "learning_rate": 1.3004096920447097e-05, "loss": 2.0484, "step": 17072 }, { "epoch": 0.42, "learning_rate": 1.300333545445239e-05, "loss": 1.9756, "step": 17073 }, { "epoch": 0.42, "learning_rate": 1.3002573969316447e-05, "loss": 2.0545, "step": 17074 }, { "epoch": 0.42, "learning_rate": 1.3001812465044123e-05, "loss": 2.0508, "step": 17075 }, { "epoch": 0.42, "learning_rate": 1.3001050941640268e-05, "loss": 2.1468, "step": 17076 }, { "epoch": 0.42, "learning_rate": 1.3000289399109737e-05, "loss": 2.0228, "step": 17077 }, { "epoch": 0.42, "learning_rate": 1.2999527837457384e-05, "loss": 2.1059, "step": 17078 }, { "epoch": 0.42, "learning_rate": 1.2998766256688065e-05, "loss": 2.2736, "step": 17079 }, { "epoch": 0.42, "learning_rate": 1.299800465680663e-05, "loss": 2.2439, "step": 17080 }, { "epoch": 0.42, "learning_rate": 1.299724303781793e-05, "loss": 2.0571, "step": 17081 }, { "epoch": 0.42, "learning_rate": 1.2996481399726828e-05, "loss": 2.1111, "step": 17082 }, { "epoch": 0.42, "learning_rate": 1.2995719742538172e-05, "loss": 2.0645, "step": 17083 }, { "epoch": 0.42, "learning_rate": 1.2994958066256821e-05, "loss": 2.1453, "step": 17084 }, { "epoch": 0.42, "learning_rate": 1.2994196370887624e-05, "loss": 1.9661, "step": 17085 }, { "epoch": 0.42, "learning_rate": 1.2993434656435435e-05, "loss": 2.0218, "step": 17086 }, { "epoch": 0.42, "learning_rate": 1.2992672922905114e-05, "loss": 2.1588, "step": 17087 }, { "epoch": 0.42, "learning_rate": 1.2991911170301512e-05, "loss": 2.0282, "step": 17088 }, { "epoch": 0.42, "learning_rate": 1.2991149398629488e-05, "loss": 2.1802, "step": 17089 }, { "epoch": 0.42, "learning_rate": 1.299038760789389e-05, "loss": 2.033, "step": 17090 }, { "epoch": 0.42, "learning_rate": 1.298962579809958e-05, "loss": 2.1282, "step": 17091 }, { "epoch": 0.42, "learning_rate": 1.298886396925141e-05, "loss": 2.0849, "step": 17092 }, { "epoch": 0.42, "learning_rate": 1.2988102121354237e-05, "loss": 2.1616, "step": 17093 }, { "epoch": 0.42, "learning_rate": 1.2987340254412915e-05, "loss": 2.1781, "step": 17094 }, { "epoch": 0.42, "learning_rate": 1.2986578368432297e-05, "loss": 2.1305, "step": 17095 }, { "epoch": 0.42, "learning_rate": 1.2985816463417242e-05, "loss": 2.0299, "step": 17096 }, { "epoch": 0.42, "learning_rate": 1.2985054539372608e-05, "loss": 2.173, "step": 17097 }, { "epoch": 0.42, "learning_rate": 1.298429259630325e-05, "loss": 2.0413, "step": 17098 }, { "epoch": 0.42, "learning_rate": 1.2983530634214021e-05, "loss": 2.1001, "step": 17099 }, { "epoch": 0.42, "learning_rate": 1.2982768653109776e-05, "loss": 2.0964, "step": 17100 }, { "epoch": 0.42, "learning_rate": 1.2982006652995378e-05, "loss": 2.1625, "step": 17101 }, { "epoch": 0.42, "learning_rate": 1.298124463387568e-05, "loss": 2.2692, "step": 17102 }, { "epoch": 0.42, "learning_rate": 1.2980482595755537e-05, "loss": 2.0896, "step": 17103 }, { "epoch": 0.42, "learning_rate": 1.2979720538639807e-05, "loss": 2.0081, "step": 17104 }, { "epoch": 0.42, "learning_rate": 1.2978958462533346e-05, "loss": 2.0099, "step": 17105 }, { "epoch": 0.42, "learning_rate": 1.2978196367441013e-05, "loss": 2.2049, "step": 17106 }, { "epoch": 0.42, "learning_rate": 1.2977434253367665e-05, "loss": 2.0013, "step": 17107 }, { "epoch": 0.42, "learning_rate": 1.297667212031816e-05, "loss": 2.1017, "step": 17108 }, { "epoch": 0.42, "learning_rate": 1.297590996829735e-05, "loss": 2.0603, "step": 17109 }, { "epoch": 0.42, "learning_rate": 1.2975147797310096e-05, "loss": 1.9065, "step": 17110 }, { "epoch": 0.42, "learning_rate": 1.2974385607361258e-05, "loss": 2.0831, "step": 17111 }, { "epoch": 0.42, "learning_rate": 1.2973623398455688e-05, "loss": 2.0846, "step": 17112 }, { "epoch": 0.42, "learning_rate": 1.2972861170598247e-05, "loss": 2.066, "step": 17113 }, { "epoch": 0.42, "learning_rate": 1.2972098923793797e-05, "loss": 2.148, "step": 17114 }, { "epoch": 0.42, "learning_rate": 1.2971336658047187e-05, "loss": 2.2187, "step": 17115 }, { "epoch": 0.42, "learning_rate": 1.2970574373363283e-05, "loss": 2.0598, "step": 17116 }, { "epoch": 0.42, "learning_rate": 1.2969812069746941e-05, "loss": 2.01, "step": 17117 }, { "epoch": 0.42, "learning_rate": 1.296904974720302e-05, "loss": 2.0988, "step": 17118 }, { "epoch": 0.42, "learning_rate": 1.2968287405736373e-05, "loss": 2.1035, "step": 17119 }, { "epoch": 0.42, "learning_rate": 1.2967525045351866e-05, "loss": 1.9386, "step": 17120 }, { "epoch": 0.42, "learning_rate": 1.2966762666054353e-05, "loss": 2.0234, "step": 17121 }, { "epoch": 0.42, "learning_rate": 1.2966000267848695e-05, "loss": 2.1048, "step": 17122 }, { "epoch": 0.42, "learning_rate": 1.2965237850739754e-05, "loss": 2.1163, "step": 17123 }, { "epoch": 0.42, "learning_rate": 1.296447541473238e-05, "loss": 1.9944, "step": 17124 }, { "epoch": 0.42, "learning_rate": 1.2963712959831443e-05, "loss": 2.0579, "step": 17125 }, { "epoch": 0.42, "learning_rate": 1.2962950486041797e-05, "loss": 2.0351, "step": 17126 }, { "epoch": 0.42, "learning_rate": 1.2962187993368301e-05, "loss": 2.0266, "step": 17127 }, { "epoch": 0.42, "learning_rate": 1.2961425481815818e-05, "loss": 2.1114, "step": 17128 }, { "epoch": 0.42, "learning_rate": 1.2960662951389201e-05, "loss": 2.0879, "step": 17129 }, { "epoch": 0.42, "learning_rate": 1.2959900402093316e-05, "loss": 2.1167, "step": 17130 }, { "epoch": 0.42, "learning_rate": 1.2959137833933019e-05, "loss": 2.1517, "step": 17131 }, { "epoch": 0.42, "learning_rate": 1.2958375246913175e-05, "loss": 2.0602, "step": 17132 }, { "epoch": 0.42, "learning_rate": 1.295761264103864e-05, "loss": 2.0812, "step": 17133 }, { "epoch": 0.42, "learning_rate": 1.2956850016314276e-05, "loss": 2.1887, "step": 17134 }, { "epoch": 0.42, "learning_rate": 1.2956087372744945e-05, "loss": 2.1824, "step": 17135 }, { "epoch": 0.42, "learning_rate": 1.2955324710335504e-05, "loss": 1.983, "step": 17136 }, { "epoch": 0.42, "learning_rate": 1.2954562029090819e-05, "loss": 2.1982, "step": 17137 }, { "epoch": 0.42, "learning_rate": 1.2953799329015744e-05, "loss": 2.0744, "step": 17138 }, { "epoch": 0.42, "learning_rate": 1.2953036610115141e-05, "loss": 1.955, "step": 17139 }, { "epoch": 0.42, "learning_rate": 1.2952273872393875e-05, "loss": 2.0924, "step": 17140 }, { "epoch": 0.42, "learning_rate": 1.2951511115856807e-05, "loss": 1.8737, "step": 17141 }, { "epoch": 0.42, "learning_rate": 1.2950748340508796e-05, "loss": 1.9974, "step": 17142 }, { "epoch": 0.42, "learning_rate": 1.2949985546354706e-05, "loss": 2.0704, "step": 17143 }, { "epoch": 0.42, "learning_rate": 1.2949222733399393e-05, "loss": 2.283, "step": 17144 }, { "epoch": 0.42, "learning_rate": 1.2948459901647725e-05, "loss": 2.1866, "step": 17145 }, { "epoch": 0.42, "learning_rate": 1.2947697051104562e-05, "loss": 2.0606, "step": 17146 }, { "epoch": 0.42, "learning_rate": 1.2946934181774764e-05, "loss": 2.1428, "step": 17147 }, { "epoch": 0.42, "learning_rate": 1.2946171293663193e-05, "loss": 2.0224, "step": 17148 }, { "epoch": 0.42, "learning_rate": 1.2945408386774712e-05, "loss": 2.1201, "step": 17149 }, { "epoch": 0.42, "learning_rate": 1.2944645461114185e-05, "loss": 2.177, "step": 17150 }, { "epoch": 0.42, "learning_rate": 1.2943882516686473e-05, "loss": 1.9471, "step": 17151 }, { "epoch": 0.42, "learning_rate": 1.2943119553496437e-05, "loss": 2.0826, "step": 17152 }, { "epoch": 0.42, "learning_rate": 1.2942356571548942e-05, "loss": 2.0461, "step": 17153 }, { "epoch": 0.42, "learning_rate": 1.294159357084885e-05, "loss": 1.955, "step": 17154 }, { "epoch": 0.42, "learning_rate": 1.2940830551401022e-05, "loss": 1.9399, "step": 17155 }, { "epoch": 0.42, "learning_rate": 1.2940067513210325e-05, "loss": 2.0228, "step": 17156 }, { "epoch": 0.42, "learning_rate": 1.2939304456281616e-05, "loss": 2.0771, "step": 17157 }, { "epoch": 0.42, "learning_rate": 1.2938541380619763e-05, "loss": 2.0634, "step": 17158 }, { "epoch": 0.42, "learning_rate": 1.293777828622963e-05, "loss": 2.0413, "step": 17159 }, { "epoch": 0.42, "learning_rate": 1.2937015173116078e-05, "loss": 2.0627, "step": 17160 }, { "epoch": 0.42, "learning_rate": 1.293625204128397e-05, "loss": 2.0828, "step": 17161 }, { "epoch": 0.42, "learning_rate": 1.2935488890738172e-05, "loss": 1.9549, "step": 17162 }, { "epoch": 0.42, "learning_rate": 1.2934725721483545e-05, "loss": 2.0809, "step": 17163 }, { "epoch": 0.42, "learning_rate": 1.2933962533524957e-05, "loss": 2.0575, "step": 17164 }, { "epoch": 0.42, "learning_rate": 1.2933199326867269e-05, "loss": 2.019, "step": 17165 }, { "epoch": 0.42, "learning_rate": 1.2932436101515344e-05, "loss": 2.0304, "step": 17166 }, { "epoch": 0.42, "learning_rate": 1.293167285747405e-05, "loss": 2.0902, "step": 17167 }, { "epoch": 0.42, "learning_rate": 1.2930909594748246e-05, "loss": 1.937, "step": 17168 }, { "epoch": 0.42, "learning_rate": 1.2930146313342804e-05, "loss": 2.017, "step": 17169 }, { "epoch": 0.42, "learning_rate": 1.2929383013262585e-05, "loss": 2.1446, "step": 17170 }, { "epoch": 0.42, "learning_rate": 1.2928619694512451e-05, "loss": 1.9352, "step": 17171 }, { "epoch": 0.42, "learning_rate": 1.2927856357097271e-05, "loss": 2.0525, "step": 17172 }, { "epoch": 0.42, "learning_rate": 1.2927093001021906e-05, "loss": 2.2244, "step": 17173 }, { "epoch": 0.42, "learning_rate": 1.2926329626291226e-05, "loss": 2.1808, "step": 17174 }, { "epoch": 0.42, "learning_rate": 1.292556623291009e-05, "loss": 2.0011, "step": 17175 }, { "epoch": 0.42, "learning_rate": 1.292480282088337e-05, "loss": 2.0243, "step": 17176 }, { "epoch": 0.42, "learning_rate": 1.2924039390215928e-05, "loss": 2.043, "step": 17177 }, { "epoch": 0.42, "learning_rate": 1.292327594091263e-05, "loss": 2.1844, "step": 17178 }, { "epoch": 0.42, "learning_rate": 1.292251247297834e-05, "loss": 2.0273, "step": 17179 }, { "epoch": 0.42, "learning_rate": 1.292174898641793e-05, "loss": 2.1702, "step": 17180 }, { "epoch": 0.42, "learning_rate": 1.292098548123626e-05, "loss": 1.9731, "step": 17181 }, { "epoch": 0.42, "learning_rate": 1.2920221957438192e-05, "loss": 2.0644, "step": 17182 }, { "epoch": 0.42, "learning_rate": 1.2919458415028604e-05, "loss": 2.0245, "step": 17183 }, { "epoch": 0.42, "learning_rate": 1.2918694854012356e-05, "loss": 2.141, "step": 17184 }, { "epoch": 0.42, "learning_rate": 1.2917931274394311e-05, "loss": 2.1196, "step": 17185 }, { "epoch": 0.42, "learning_rate": 1.2917167676179341e-05, "loss": 2.1346, "step": 17186 }, { "epoch": 0.42, "learning_rate": 1.291640405937231e-05, "loss": 2.0074, "step": 17187 }, { "epoch": 0.42, "learning_rate": 1.2915640423978086e-05, "loss": 1.8743, "step": 17188 }, { "epoch": 0.42, "learning_rate": 1.2914876770001537e-05, "loss": 2.0704, "step": 17189 }, { "epoch": 0.42, "learning_rate": 1.2914113097447527e-05, "loss": 2.2088, "step": 17190 }, { "epoch": 0.42, "learning_rate": 1.2913349406320923e-05, "loss": 2.0054, "step": 17191 }, { "epoch": 0.42, "learning_rate": 1.2912585696626594e-05, "loss": 2.0589, "step": 17192 }, { "epoch": 0.42, "learning_rate": 1.2911821968369407e-05, "loss": 1.9666, "step": 17193 }, { "epoch": 0.42, "learning_rate": 1.2911058221554233e-05, "loss": 2.089, "step": 17194 }, { "epoch": 0.42, "learning_rate": 1.2910294456185934e-05, "loss": 2.0506, "step": 17195 }, { "epoch": 0.42, "learning_rate": 1.2909530672269382e-05, "loss": 2.023, "step": 17196 }, { "epoch": 0.42, "learning_rate": 1.2908766869809439e-05, "loss": 2.0305, "step": 17197 }, { "epoch": 0.42, "learning_rate": 1.2908003048810979e-05, "loss": 2.0988, "step": 17198 }, { "epoch": 0.42, "learning_rate": 1.290723920927887e-05, "loss": 2.0902, "step": 17199 }, { "epoch": 0.42, "learning_rate": 1.2906475351217974e-05, "loss": 2.0265, "step": 17200 }, { "epoch": 0.42, "learning_rate": 1.2905711474633166e-05, "loss": 2.0421, "step": 17201 }, { "epoch": 0.42, "learning_rate": 1.290494757952931e-05, "loss": 2.112, "step": 17202 }, { "epoch": 0.42, "learning_rate": 1.2904183665911277e-05, "loss": 2.0239, "step": 17203 }, { "epoch": 0.42, "learning_rate": 1.2903419733783937e-05, "loss": 2.0038, "step": 17204 }, { "epoch": 0.42, "learning_rate": 1.2902655783152156e-05, "loss": 2.0262, "step": 17205 }, { "epoch": 0.42, "learning_rate": 1.2901891814020802e-05, "loss": 1.9374, "step": 17206 }, { "epoch": 0.42, "learning_rate": 1.2901127826394745e-05, "loss": 1.9351, "step": 17207 }, { "epoch": 0.42, "learning_rate": 1.2900363820278859e-05, "loss": 2.1735, "step": 17208 }, { "epoch": 0.42, "learning_rate": 1.2899599795678005e-05, "loss": 2.0048, "step": 17209 }, { "epoch": 0.42, "learning_rate": 1.289883575259706e-05, "loss": 1.9727, "step": 17210 }, { "epoch": 0.42, "learning_rate": 1.2898071691040888e-05, "loss": 1.9314, "step": 17211 }, { "epoch": 0.42, "learning_rate": 1.289730761101436e-05, "loss": 2.1543, "step": 17212 }, { "epoch": 0.42, "learning_rate": 1.2896543512522349e-05, "loss": 2.0678, "step": 17213 }, { "epoch": 0.42, "learning_rate": 1.2895779395569721e-05, "loss": 1.9711, "step": 17214 }, { "epoch": 0.42, "learning_rate": 1.2895015260161347e-05, "loss": 2.0896, "step": 17215 }, { "epoch": 0.42, "learning_rate": 1.2894251106302097e-05, "loss": 2.0046, "step": 17216 }, { "epoch": 0.42, "learning_rate": 1.2893486933996844e-05, "loss": 2.0189, "step": 17217 }, { "epoch": 0.42, "learning_rate": 1.289272274325045e-05, "loss": 2.1316, "step": 17218 }, { "epoch": 0.42, "learning_rate": 1.2891958534067797e-05, "loss": 2.0943, "step": 17219 }, { "epoch": 0.42, "learning_rate": 1.2891194306453747e-05, "loss": 2.0638, "step": 17220 }, { "epoch": 0.42, "learning_rate": 1.2890430060413173e-05, "loss": 2.106, "step": 17221 }, { "epoch": 0.42, "learning_rate": 1.2889665795950947e-05, "loss": 2.1378, "step": 17222 }, { "epoch": 0.42, "learning_rate": 1.2888901513071941e-05, "loss": 2.1125, "step": 17223 }, { "epoch": 0.42, "learning_rate": 1.2888137211781024e-05, "loss": 1.9125, "step": 17224 }, { "epoch": 0.42, "learning_rate": 1.2887372892083065e-05, "loss": 2.0597, "step": 17225 }, { "epoch": 0.42, "learning_rate": 1.2886608553982937e-05, "loss": 2.1651, "step": 17226 }, { "epoch": 0.42, "learning_rate": 1.2885844197485513e-05, "loss": 2.1417, "step": 17227 }, { "epoch": 0.42, "learning_rate": 1.2885079822595664e-05, "loss": 2.0528, "step": 17228 }, { "epoch": 0.42, "learning_rate": 1.2884315429318264e-05, "loss": 2.1863, "step": 17229 }, { "epoch": 0.42, "learning_rate": 1.2883551017658178e-05, "loss": 2.0512, "step": 17230 }, { "epoch": 0.42, "learning_rate": 1.2882786587620283e-05, "loss": 2.2213, "step": 17231 }, { "epoch": 0.42, "learning_rate": 1.288202213920945e-05, "loss": 2.2107, "step": 17232 }, { "epoch": 0.42, "learning_rate": 1.288125767243055e-05, "loss": 2.0681, "step": 17233 }, { "epoch": 0.42, "learning_rate": 1.2880493187288457e-05, "loss": 2.1104, "step": 17234 }, { "epoch": 0.42, "learning_rate": 1.287972868378804e-05, "loss": 2.1713, "step": 17235 }, { "epoch": 0.42, "learning_rate": 1.2878964161934175e-05, "loss": 2.0106, "step": 17236 }, { "epoch": 0.42, "learning_rate": 1.2878199621731731e-05, "loss": 2.0856, "step": 17237 }, { "epoch": 0.42, "learning_rate": 1.2877435063185585e-05, "loss": 2.1966, "step": 17238 }, { "epoch": 0.42, "learning_rate": 1.2876670486300608e-05, "loss": 2.0651, "step": 17239 }, { "epoch": 0.42, "learning_rate": 1.2875905891081671e-05, "loss": 2.0741, "step": 17240 }, { "epoch": 0.42, "learning_rate": 1.2875141277533649e-05, "loss": 1.9793, "step": 17241 }, { "epoch": 0.42, "learning_rate": 1.2874376645661416e-05, "loss": 2.142, "step": 17242 }, { "epoch": 0.43, "learning_rate": 1.2873611995469844e-05, "loss": 2.0727, "step": 17243 }, { "epoch": 0.43, "learning_rate": 1.2872847326963803e-05, "loss": 2.0999, "step": 17244 }, { "epoch": 0.43, "learning_rate": 1.2872082640148171e-05, "loss": 2.0837, "step": 17245 }, { "epoch": 0.43, "learning_rate": 1.287131793502782e-05, "loss": 2.0531, "step": 17246 }, { "epoch": 0.43, "learning_rate": 1.2870553211607626e-05, "loss": 1.8856, "step": 17247 }, { "epoch": 0.43, "learning_rate": 1.2869788469892461e-05, "loss": 1.9557, "step": 17248 }, { "epoch": 0.43, "learning_rate": 1.2869023709887198e-05, "loss": 2.0007, "step": 17249 }, { "epoch": 0.43, "learning_rate": 1.286825893159671e-05, "loss": 1.9784, "step": 17250 }, { "epoch": 0.43, "learning_rate": 1.2867494135025878e-05, "loss": 1.8505, "step": 17251 }, { "epoch": 0.43, "learning_rate": 1.2866729320179568e-05, "loss": 2.0298, "step": 17252 }, { "epoch": 0.43, "learning_rate": 1.2865964487062657e-05, "loss": 2.1642, "step": 17253 }, { "epoch": 0.43, "learning_rate": 1.2865199635680021e-05, "loss": 2.062, "step": 17254 }, { "epoch": 0.43, "learning_rate": 1.2864434766036533e-05, "loss": 2.1377, "step": 17255 }, { "epoch": 0.43, "learning_rate": 1.286366987813707e-05, "loss": 1.9098, "step": 17256 }, { "epoch": 0.43, "learning_rate": 1.286290497198651e-05, "loss": 1.9733, "step": 17257 }, { "epoch": 0.43, "learning_rate": 1.2862140047589717e-05, "loss": 1.9305, "step": 17258 }, { "epoch": 0.43, "learning_rate": 1.2861375104951577e-05, "loss": 2.1676, "step": 17259 }, { "epoch": 0.43, "learning_rate": 1.2860610144076958e-05, "loss": 1.9923, "step": 17260 }, { "epoch": 0.43, "learning_rate": 1.285984516497074e-05, "loss": 2.0998, "step": 17261 }, { "epoch": 0.43, "learning_rate": 1.2859080167637794e-05, "loss": 2.068, "step": 17262 }, { "epoch": 0.43, "learning_rate": 1.2858315152083003e-05, "loss": 2.1445, "step": 17263 }, { "epoch": 0.43, "learning_rate": 1.2857550118311235e-05, "loss": 2.0053, "step": 17264 }, { "epoch": 0.43, "learning_rate": 1.2856785066327368e-05, "loss": 2.1541, "step": 17265 }, { "epoch": 0.43, "learning_rate": 1.2856019996136281e-05, "loss": 2.0378, "step": 17266 }, { "epoch": 0.43, "learning_rate": 1.2855254907742849e-05, "loss": 1.966, "step": 17267 }, { "epoch": 0.43, "learning_rate": 1.2854489801151946e-05, "loss": 2.1798, "step": 17268 }, { "epoch": 0.43, "learning_rate": 1.2853724676368447e-05, "loss": 2.0594, "step": 17269 }, { "epoch": 0.43, "learning_rate": 1.2852959533397231e-05, "loss": 2.1714, "step": 17270 }, { "epoch": 0.43, "learning_rate": 1.2852194372243175e-05, "loss": 1.917, "step": 17271 }, { "epoch": 0.43, "learning_rate": 1.2851429192911154e-05, "loss": 2.05, "step": 17272 }, { "epoch": 0.43, "learning_rate": 1.2850663995406048e-05, "loss": 1.8679, "step": 17273 }, { "epoch": 0.43, "learning_rate": 1.2849898779732729e-05, "loss": 1.9241, "step": 17274 }, { "epoch": 0.43, "learning_rate": 1.2849133545896076e-05, "loss": 1.981, "step": 17275 }, { "epoch": 0.43, "learning_rate": 1.2848368293900968e-05, "loss": 2.1581, "step": 17276 }, { "epoch": 0.43, "learning_rate": 1.284760302375228e-05, "loss": 1.9401, "step": 17277 }, { "epoch": 0.43, "learning_rate": 1.2846837735454891e-05, "loss": 2.0886, "step": 17278 }, { "epoch": 0.43, "learning_rate": 1.2846072429013674e-05, "loss": 2.1156, "step": 17279 }, { "epoch": 0.43, "learning_rate": 1.284530710443351e-05, "loss": 2.0489, "step": 17280 }, { "epoch": 0.43, "learning_rate": 1.284454176171928e-05, "loss": 2.0076, "step": 17281 }, { "epoch": 0.43, "learning_rate": 1.2843776400875859e-05, "loss": 1.974, "step": 17282 }, { "epoch": 0.43, "learning_rate": 1.2843011021908123e-05, "loss": 2.1283, "step": 17283 }, { "epoch": 0.43, "learning_rate": 1.2842245624820949e-05, "loss": 2.1322, "step": 17284 }, { "epoch": 0.43, "learning_rate": 1.2841480209619219e-05, "loss": 1.8935, "step": 17285 }, { "epoch": 0.43, "learning_rate": 1.2840714776307813e-05, "loss": 2.176, "step": 17286 }, { "epoch": 0.43, "learning_rate": 1.28399493248916e-05, "loss": 2.0868, "step": 17287 }, { "epoch": 0.43, "learning_rate": 1.2839183855375467e-05, "loss": 2.2072, "step": 17288 }, { "epoch": 0.43, "learning_rate": 1.2838418367764289e-05, "loss": 1.9662, "step": 17289 }, { "epoch": 0.43, "learning_rate": 1.2837652862062947e-05, "loss": 2.0557, "step": 17290 }, { "epoch": 0.43, "learning_rate": 1.2836887338276318e-05, "loss": 2.0369, "step": 17291 }, { "epoch": 0.43, "learning_rate": 1.2836121796409283e-05, "loss": 2.1289, "step": 17292 }, { "epoch": 0.43, "learning_rate": 1.2835356236466719e-05, "loss": 2.0324, "step": 17293 }, { "epoch": 0.43, "learning_rate": 1.2834590658453503e-05, "loss": 1.8511, "step": 17294 }, { "epoch": 0.43, "learning_rate": 1.2833825062374521e-05, "loss": 1.984, "step": 17295 }, { "epoch": 0.43, "learning_rate": 1.2833059448234645e-05, "loss": 2.157, "step": 17296 }, { "epoch": 0.43, "learning_rate": 1.2832293816038758e-05, "loss": 2.0401, "step": 17297 }, { "epoch": 0.43, "learning_rate": 1.283152816579174e-05, "loss": 2.0442, "step": 17298 }, { "epoch": 0.43, "learning_rate": 1.283076249749847e-05, "loss": 2.087, "step": 17299 }, { "epoch": 0.43, "learning_rate": 1.2829996811163831e-05, "loss": 1.8723, "step": 17300 }, { "epoch": 0.43, "learning_rate": 1.2829231106792698e-05, "loss": 2.1274, "step": 17301 }, { "epoch": 0.43, "learning_rate": 1.2828465384389954e-05, "loss": 2.0721, "step": 17302 }, { "epoch": 0.43, "learning_rate": 1.2827699643960476e-05, "loss": 1.9559, "step": 17303 }, { "epoch": 0.43, "learning_rate": 1.2826933885509149e-05, "loss": 1.9563, "step": 17304 }, { "epoch": 0.43, "learning_rate": 1.282616810904085e-05, "loss": 1.98, "step": 17305 }, { "epoch": 0.43, "learning_rate": 1.282540231456046e-05, "loss": 2.0323, "step": 17306 }, { "epoch": 0.43, "learning_rate": 1.2824636502072863e-05, "loss": 2.2066, "step": 17307 }, { "epoch": 0.43, "learning_rate": 1.2823870671582932e-05, "loss": 1.979, "step": 17308 }, { "epoch": 0.43, "learning_rate": 1.2823104823095558e-05, "loss": 1.9894, "step": 17309 }, { "epoch": 0.43, "learning_rate": 1.2822338956615617e-05, "loss": 2.3762, "step": 17310 }, { "epoch": 0.43, "learning_rate": 1.2821573072147988e-05, "loss": 1.9466, "step": 17311 }, { "epoch": 0.43, "learning_rate": 1.2820807169697554e-05, "loss": 2.103, "step": 17312 }, { "epoch": 0.43, "learning_rate": 1.2820041249269196e-05, "loss": 2.1044, "step": 17313 }, { "epoch": 0.43, "learning_rate": 1.2819275310867798e-05, "loss": 2.1945, "step": 17314 }, { "epoch": 0.43, "learning_rate": 1.2818509354498236e-05, "loss": 2.112, "step": 17315 }, { "epoch": 0.43, "learning_rate": 1.2817743380165399e-05, "loss": 2.2154, "step": 17316 }, { "epoch": 0.43, "learning_rate": 1.2816977387874166e-05, "loss": 2.0184, "step": 17317 }, { "epoch": 0.43, "learning_rate": 1.2816211377629415e-05, "loss": 1.9052, "step": 17318 }, { "epoch": 0.43, "learning_rate": 1.2815445349436033e-05, "loss": 2.1634, "step": 17319 }, { "epoch": 0.43, "learning_rate": 1.28146793032989e-05, "loss": 2.1214, "step": 17320 }, { "epoch": 0.43, "learning_rate": 1.2813913239222899e-05, "loss": 2.0161, "step": 17321 }, { "epoch": 0.43, "learning_rate": 1.2813147157212913e-05, "loss": 2.1405, "step": 17322 }, { "epoch": 0.43, "learning_rate": 1.281238105727382e-05, "loss": 2.0571, "step": 17323 }, { "epoch": 0.43, "learning_rate": 1.2811614939410504e-05, "loss": 1.9996, "step": 17324 }, { "epoch": 0.43, "learning_rate": 1.2810848803627856e-05, "loss": 2.0415, "step": 17325 }, { "epoch": 0.43, "learning_rate": 1.281008264993075e-05, "loss": 2.1524, "step": 17326 }, { "epoch": 0.43, "learning_rate": 1.2809316478324069e-05, "loss": 2.1458, "step": 17327 }, { "epoch": 0.43, "learning_rate": 1.2808550288812702e-05, "loss": 2.0131, "step": 17328 }, { "epoch": 0.43, "learning_rate": 1.2807784081401529e-05, "loss": 2.1279, "step": 17329 }, { "epoch": 0.43, "learning_rate": 1.2807017856095432e-05, "loss": 2.1604, "step": 17330 }, { "epoch": 0.43, "learning_rate": 1.2806251612899294e-05, "loss": 2.021, "step": 17331 }, { "epoch": 0.43, "learning_rate": 1.2805485351818001e-05, "loss": 2.0687, "step": 17332 }, { "epoch": 0.43, "learning_rate": 1.2804719072856433e-05, "loss": 1.9883, "step": 17333 }, { "epoch": 0.43, "learning_rate": 1.2803952776019479e-05, "loss": 2.0663, "step": 17334 }, { "epoch": 0.43, "learning_rate": 1.2803186461312022e-05, "loss": 2.0434, "step": 17335 }, { "epoch": 0.43, "learning_rate": 1.2802420128738945e-05, "loss": 2.1431, "step": 17336 }, { "epoch": 0.43, "learning_rate": 1.2801653778305125e-05, "loss": 2.1289, "step": 17337 }, { "epoch": 0.43, "learning_rate": 1.2800887410015457e-05, "loss": 2.1052, "step": 17338 }, { "epoch": 0.43, "learning_rate": 1.2800121023874818e-05, "loss": 1.9106, "step": 17339 }, { "epoch": 0.43, "learning_rate": 1.2799354619888098e-05, "loss": 2.141, "step": 17340 }, { "epoch": 0.43, "learning_rate": 1.2798588198060178e-05, "loss": 2.1537, "step": 17341 }, { "epoch": 0.43, "learning_rate": 1.279782175839594e-05, "loss": 2.1467, "step": 17342 }, { "epoch": 0.43, "learning_rate": 1.2797055300900276e-05, "loss": 1.9128, "step": 17343 }, { "epoch": 0.43, "learning_rate": 1.2796288825578067e-05, "loss": 1.9534, "step": 17344 }, { "epoch": 0.43, "learning_rate": 1.2795522332434197e-05, "loss": 2.0974, "step": 17345 }, { "epoch": 0.43, "learning_rate": 1.2794755821473555e-05, "loss": 2.0444, "step": 17346 }, { "epoch": 0.43, "learning_rate": 1.2793989292701017e-05, "loss": 2.1331, "step": 17347 }, { "epoch": 0.43, "learning_rate": 1.2793222746121479e-05, "loss": 2.0375, "step": 17348 }, { "epoch": 0.43, "learning_rate": 1.2792456181739822e-05, "loss": 2.0384, "step": 17349 }, { "epoch": 0.43, "learning_rate": 1.279168959956093e-05, "loss": 2.0373, "step": 17350 }, { "epoch": 0.43, "learning_rate": 1.2790922999589692e-05, "loss": 1.8967, "step": 17351 }, { "epoch": 0.43, "learning_rate": 1.2790156381830992e-05, "loss": 2.0282, "step": 17352 }, { "epoch": 0.43, "learning_rate": 1.2789389746289715e-05, "loss": 1.9652, "step": 17353 }, { "epoch": 0.43, "learning_rate": 1.2788623092970749e-05, "loss": 1.9139, "step": 17354 }, { "epoch": 0.43, "learning_rate": 1.278785642187898e-05, "loss": 2.0473, "step": 17355 }, { "epoch": 0.43, "learning_rate": 1.2787089733019294e-05, "loss": 1.8821, "step": 17356 }, { "epoch": 0.43, "learning_rate": 1.2786323026396575e-05, "loss": 2.1528, "step": 17357 }, { "epoch": 0.43, "learning_rate": 1.2785556302015711e-05, "loss": 2.0505, "step": 17358 }, { "epoch": 0.43, "learning_rate": 1.278478955988159e-05, "loss": 1.9197, "step": 17359 }, { "epoch": 0.43, "learning_rate": 1.2784022799999098e-05, "loss": 2.1747, "step": 17360 }, { "epoch": 0.43, "learning_rate": 1.2783256022373118e-05, "loss": 2.0725, "step": 17361 }, { "epoch": 0.43, "learning_rate": 1.2782489227008545e-05, "loss": 2.0605, "step": 17362 }, { "epoch": 0.43, "learning_rate": 1.278172241391026e-05, "loss": 2.0175, "step": 17363 }, { "epoch": 0.43, "learning_rate": 1.2780955583083153e-05, "loss": 2.3591, "step": 17364 }, { "epoch": 0.43, "learning_rate": 1.2780188734532107e-05, "loss": 2.1139, "step": 17365 }, { "epoch": 0.43, "learning_rate": 1.2779421868262011e-05, "loss": 2.0128, "step": 17366 }, { "epoch": 0.43, "learning_rate": 1.2778654984277756e-05, "loss": 1.9185, "step": 17367 }, { "epoch": 0.43, "learning_rate": 1.2777888082584227e-05, "loss": 2.019, "step": 17368 }, { "epoch": 0.43, "learning_rate": 1.277712116318631e-05, "loss": 1.9723, "step": 17369 }, { "epoch": 0.43, "learning_rate": 1.2776354226088899e-05, "loss": 2.0333, "step": 17370 }, { "epoch": 0.43, "learning_rate": 1.2775587271296872e-05, "loss": 2.1843, "step": 17371 }, { "epoch": 0.43, "learning_rate": 1.277482029881513e-05, "loss": 2.1732, "step": 17372 }, { "epoch": 0.43, "learning_rate": 1.2774053308648548e-05, "loss": 2.1425, "step": 17373 }, { "epoch": 0.43, "learning_rate": 1.2773286300802023e-05, "loss": 2.0684, "step": 17374 }, { "epoch": 0.43, "learning_rate": 1.2772519275280441e-05, "loss": 2.0557, "step": 17375 }, { "epoch": 0.43, "learning_rate": 1.2771752232088688e-05, "loss": 1.7929, "step": 17376 }, { "epoch": 0.43, "learning_rate": 1.2770985171231653e-05, "loss": 2.1125, "step": 17377 }, { "epoch": 0.43, "learning_rate": 1.277021809271423e-05, "loss": 2.0638, "step": 17378 }, { "epoch": 0.43, "learning_rate": 1.2769450996541303e-05, "loss": 1.9149, "step": 17379 }, { "epoch": 0.43, "learning_rate": 1.2768683882717765e-05, "loss": 1.9639, "step": 17380 }, { "epoch": 0.43, "learning_rate": 1.2767916751248499e-05, "loss": 2.0535, "step": 17381 }, { "epoch": 0.43, "learning_rate": 1.27671496021384e-05, "loss": 2.1149, "step": 17382 }, { "epoch": 0.43, "learning_rate": 1.2766382435392353e-05, "loss": 1.9313, "step": 17383 }, { "epoch": 0.43, "learning_rate": 1.276561525101525e-05, "loss": 2.0485, "step": 17384 }, { "epoch": 0.43, "learning_rate": 1.2764848049011977e-05, "loss": 2.0061, "step": 17385 }, { "epoch": 0.43, "learning_rate": 1.2764080829387429e-05, "loss": 1.9184, "step": 17386 }, { "epoch": 0.43, "learning_rate": 1.2763313592146494e-05, "loss": 2.2257, "step": 17387 }, { "epoch": 0.43, "learning_rate": 1.276254633729406e-05, "loss": 2.2778, "step": 17388 }, { "epoch": 0.43, "learning_rate": 1.2761779064835018e-05, "loss": 2.1087, "step": 17389 }, { "epoch": 0.43, "learning_rate": 1.2761011774774254e-05, "loss": 2.0122, "step": 17390 }, { "epoch": 0.43, "learning_rate": 1.2760244467116668e-05, "loss": 2.017, "step": 17391 }, { "epoch": 0.43, "learning_rate": 1.2759477141867142e-05, "loss": 1.9831, "step": 17392 }, { "epoch": 0.43, "learning_rate": 1.275870979903057e-05, "loss": 1.9692, "step": 17393 }, { "epoch": 0.43, "learning_rate": 1.2757942438611841e-05, "loss": 1.8967, "step": 17394 }, { "epoch": 0.43, "learning_rate": 1.2757175060615843e-05, "loss": 2.0625, "step": 17395 }, { "epoch": 0.43, "learning_rate": 1.2756407665047474e-05, "loss": 2.2466, "step": 17396 }, { "epoch": 0.43, "learning_rate": 1.2755640251911618e-05, "loss": 2.0383, "step": 17397 }, { "epoch": 0.43, "learning_rate": 1.2754872821213171e-05, "loss": 2.2137, "step": 17398 }, { "epoch": 0.43, "learning_rate": 1.2754105372957022e-05, "loss": 2.1442, "step": 17399 }, { "epoch": 0.43, "learning_rate": 1.2753337907148058e-05, "loss": 1.8951, "step": 17400 }, { "epoch": 0.43, "learning_rate": 1.2752570423791177e-05, "loss": 2.133, "step": 17401 }, { "epoch": 0.43, "learning_rate": 1.2751802922891267e-05, "loss": 1.8755, "step": 17402 }, { "epoch": 0.43, "learning_rate": 1.275103540445322e-05, "loss": 2.0959, "step": 17403 }, { "epoch": 0.43, "learning_rate": 1.275026786848193e-05, "loss": 2.1004, "step": 17404 }, { "epoch": 0.43, "learning_rate": 1.2749500314982282e-05, "loss": 1.9084, "step": 17405 }, { "epoch": 0.43, "learning_rate": 1.2748732743959176e-05, "loss": 2.1131, "step": 17406 }, { "epoch": 0.43, "learning_rate": 1.2747965155417501e-05, "loss": 2.1935, "step": 17407 }, { "epoch": 0.43, "learning_rate": 1.2747197549362146e-05, "loss": 2.0589, "step": 17408 }, { "epoch": 0.43, "learning_rate": 1.2746429925798009e-05, "loss": 1.9773, "step": 17409 }, { "epoch": 0.43, "learning_rate": 1.2745662284729976e-05, "loss": 1.9581, "step": 17410 }, { "epoch": 0.43, "learning_rate": 1.2744894626162942e-05, "loss": 2.1233, "step": 17411 }, { "epoch": 0.43, "learning_rate": 1.27441269501018e-05, "loss": 2.028, "step": 17412 }, { "epoch": 0.43, "learning_rate": 1.2743359256551447e-05, "loss": 2.107, "step": 17413 }, { "epoch": 0.43, "learning_rate": 1.2742591545516766e-05, "loss": 1.8639, "step": 17414 }, { "epoch": 0.43, "learning_rate": 1.2741823817002658e-05, "loss": 2.0526, "step": 17415 }, { "epoch": 0.43, "learning_rate": 1.2741056071014014e-05, "loss": 1.988, "step": 17416 }, { "epoch": 0.43, "learning_rate": 1.2740288307555728e-05, "loss": 2.0601, "step": 17417 }, { "epoch": 0.43, "learning_rate": 1.273952052663269e-05, "loss": 1.9223, "step": 17418 }, { "epoch": 0.43, "learning_rate": 1.2738752728249793e-05, "loss": 2.007, "step": 17419 }, { "epoch": 0.43, "learning_rate": 1.2737984912411934e-05, "loss": 1.9027, "step": 17420 }, { "epoch": 0.43, "learning_rate": 1.2737217079124001e-05, "loss": 2.1718, "step": 17421 }, { "epoch": 0.43, "learning_rate": 1.2736449228390897e-05, "loss": 1.895, "step": 17422 }, { "epoch": 0.43, "learning_rate": 1.2735681360217507e-05, "loss": 2.0953, "step": 17423 }, { "epoch": 0.43, "learning_rate": 1.2734913474608728e-05, "loss": 1.9496, "step": 17424 }, { "epoch": 0.43, "learning_rate": 1.2734145571569456e-05, "loss": 2.1283, "step": 17425 }, { "epoch": 0.43, "learning_rate": 1.2733377651104583e-05, "loss": 2.0985, "step": 17426 }, { "epoch": 0.43, "learning_rate": 1.2732609713219004e-05, "loss": 1.9677, "step": 17427 }, { "epoch": 0.43, "learning_rate": 1.273184175791761e-05, "loss": 2.2024, "step": 17428 }, { "epoch": 0.43, "learning_rate": 1.27310737852053e-05, "loss": 1.9697, "step": 17429 }, { "epoch": 0.43, "learning_rate": 1.2730305795086966e-05, "loss": 1.8742, "step": 17430 }, { "epoch": 0.43, "learning_rate": 1.2729537787567506e-05, "loss": 2.015, "step": 17431 }, { "epoch": 0.43, "learning_rate": 1.2728769762651809e-05, "loss": 2.0315, "step": 17432 }, { "epoch": 0.43, "learning_rate": 1.2728001720344776e-05, "loss": 1.9371, "step": 17433 }, { "epoch": 0.43, "learning_rate": 1.2727233660651295e-05, "loss": 2.1308, "step": 17434 }, { "epoch": 0.43, "learning_rate": 1.2726465583576268e-05, "loss": 1.9913, "step": 17435 }, { "epoch": 0.43, "learning_rate": 1.2725697489124589e-05, "loss": 1.9266, "step": 17436 }, { "epoch": 0.43, "learning_rate": 1.272492937730115e-05, "loss": 2.0168, "step": 17437 }, { "epoch": 0.43, "learning_rate": 1.2724161248110844e-05, "loss": 2.0742, "step": 17438 }, { "epoch": 0.43, "learning_rate": 1.2723393101558573e-05, "loss": 2.0684, "step": 17439 }, { "epoch": 0.43, "learning_rate": 1.2722624937649233e-05, "loss": 1.9962, "step": 17440 }, { "epoch": 0.43, "learning_rate": 1.2721856756387715e-05, "loss": 2.0271, "step": 17441 }, { "epoch": 0.43, "learning_rate": 1.2721088557778917e-05, "loss": 1.9059, "step": 17442 }, { "epoch": 0.43, "learning_rate": 1.2720320341827734e-05, "loss": 2.0154, "step": 17443 }, { "epoch": 0.43, "learning_rate": 1.2719552108539064e-05, "loss": 2.0613, "step": 17444 }, { "epoch": 0.43, "learning_rate": 1.2718783857917803e-05, "loss": 1.8233, "step": 17445 }, { "epoch": 0.43, "learning_rate": 1.2718015589968846e-05, "loss": 2.1898, "step": 17446 }, { "epoch": 0.43, "learning_rate": 1.2717247304697087e-05, "loss": 1.9749, "step": 17447 }, { "epoch": 0.43, "learning_rate": 1.2716479002107427e-05, "loss": 1.9252, "step": 17448 }, { "epoch": 0.43, "learning_rate": 1.2715710682204762e-05, "loss": 2.035, "step": 17449 }, { "epoch": 0.43, "learning_rate": 1.2714942344993988e-05, "loss": 2.1635, "step": 17450 }, { "epoch": 0.43, "learning_rate": 1.2714173990480003e-05, "loss": 1.8846, "step": 17451 }, { "epoch": 0.43, "learning_rate": 1.27134056186677e-05, "loss": 2.1745, "step": 17452 }, { "epoch": 0.43, "learning_rate": 1.2712637229561979e-05, "loss": 2.0143, "step": 17453 }, { "epoch": 0.43, "learning_rate": 1.2711868823167737e-05, "loss": 2.0014, "step": 17454 }, { "epoch": 0.43, "learning_rate": 1.2711100399489873e-05, "loss": 2.1322, "step": 17455 }, { "epoch": 0.43, "learning_rate": 1.271033195853328e-05, "loss": 2.1484, "step": 17456 }, { "epoch": 0.43, "learning_rate": 1.2709563500302861e-05, "loss": 2.0024, "step": 17457 }, { "epoch": 0.43, "learning_rate": 1.2708795024803506e-05, "loss": 1.9111, "step": 17458 }, { "epoch": 0.43, "learning_rate": 1.2708026532040122e-05, "loss": 2.0604, "step": 17459 }, { "epoch": 0.43, "learning_rate": 1.2707258022017603e-05, "loss": 2.1506, "step": 17460 }, { "epoch": 0.43, "learning_rate": 1.2706489494740844e-05, "loss": 2.0021, "step": 17461 }, { "epoch": 0.43, "learning_rate": 1.2705720950214748e-05, "loss": 2.2048, "step": 17462 }, { "epoch": 0.43, "learning_rate": 1.2704952388444207e-05, "loss": 2.153, "step": 17463 }, { "epoch": 0.43, "learning_rate": 1.2704183809434123e-05, "loss": 2.018, "step": 17464 }, { "epoch": 0.43, "learning_rate": 1.2703415213189396e-05, "loss": 2.0821, "step": 17465 }, { "epoch": 0.43, "learning_rate": 1.2702646599714922e-05, "loss": 2.0052, "step": 17466 }, { "epoch": 0.43, "learning_rate": 1.2701877969015604e-05, "loss": 2.327, "step": 17467 }, { "epoch": 0.43, "learning_rate": 1.2701109321096332e-05, "loss": 2.1987, "step": 17468 }, { "epoch": 0.43, "learning_rate": 1.2700340655962013e-05, "loss": 1.9982, "step": 17469 }, { "epoch": 0.43, "learning_rate": 1.2699571973617542e-05, "loss": 1.9406, "step": 17470 }, { "epoch": 0.43, "learning_rate": 1.269880327406782e-05, "loss": 2.0895, "step": 17471 }, { "epoch": 0.43, "learning_rate": 1.2698034557317743e-05, "loss": 2.0064, "step": 17472 }, { "epoch": 0.43, "learning_rate": 1.2697265823372216e-05, "loss": 2.0195, "step": 17473 }, { "epoch": 0.43, "learning_rate": 1.269649707223613e-05, "loss": 2.0462, "step": 17474 }, { "epoch": 0.43, "learning_rate": 1.2695728303914394e-05, "loss": 2.2507, "step": 17475 }, { "epoch": 0.43, "learning_rate": 1.26949595184119e-05, "loss": 2.0233, "step": 17476 }, { "epoch": 0.43, "learning_rate": 1.2694190715733552e-05, "loss": 2.0231, "step": 17477 }, { "epoch": 0.43, "learning_rate": 1.269342189588425e-05, "loss": 2.0499, "step": 17478 }, { "epoch": 0.43, "learning_rate": 1.2692653058868891e-05, "loss": 1.9227, "step": 17479 }, { "epoch": 0.43, "learning_rate": 1.2691884204692377e-05, "loss": 2.0286, "step": 17480 }, { "epoch": 0.43, "learning_rate": 1.2691115333359607e-05, "loss": 2.0644, "step": 17481 }, { "epoch": 0.43, "learning_rate": 1.2690346444875484e-05, "loss": 2.0722, "step": 17482 }, { "epoch": 0.43, "learning_rate": 1.2689577539244901e-05, "loss": 2.0698, "step": 17483 }, { "epoch": 0.43, "learning_rate": 1.268880861647277e-05, "loss": 2.0902, "step": 17484 }, { "epoch": 0.43, "learning_rate": 1.2688039676563983e-05, "loss": 2.0767, "step": 17485 }, { "epoch": 0.43, "learning_rate": 1.2687270719523444e-05, "loss": 2.1838, "step": 17486 }, { "epoch": 0.43, "learning_rate": 1.268650174535605e-05, "loss": 1.8457, "step": 17487 }, { "epoch": 0.43, "learning_rate": 1.2685732754066706e-05, "loss": 2.167, "step": 17488 }, { "epoch": 0.43, "learning_rate": 1.2684963745660315e-05, "loss": 2.0798, "step": 17489 }, { "epoch": 0.43, "learning_rate": 1.2684194720141772e-05, "loss": 2.0282, "step": 17490 }, { "epoch": 0.43, "learning_rate": 1.2683425677515982e-05, "loss": 2.1199, "step": 17491 }, { "epoch": 0.43, "learning_rate": 1.2682656617787843e-05, "loss": 2.156, "step": 17492 }, { "epoch": 0.43, "learning_rate": 1.2681887540962263e-05, "loss": 1.9366, "step": 17493 }, { "epoch": 0.43, "learning_rate": 1.268111844704414e-05, "loss": 2.077, "step": 17494 }, { "epoch": 0.43, "learning_rate": 1.2680349336038375e-05, "loss": 2.1441, "step": 17495 }, { "epoch": 0.43, "learning_rate": 1.2679580207949868e-05, "loss": 1.9271, "step": 17496 }, { "epoch": 0.43, "learning_rate": 1.2678811062783525e-05, "loss": 1.9763, "step": 17497 }, { "epoch": 0.43, "learning_rate": 1.2678041900544245e-05, "loss": 1.9558, "step": 17498 }, { "epoch": 0.43, "learning_rate": 1.2677272721236931e-05, "loss": 1.9806, "step": 17499 }, { "epoch": 0.43, "learning_rate": 1.2676503524866486e-05, "loss": 1.8865, "step": 17500 }, { "epoch": 0.43, "learning_rate": 1.2675734311437814e-05, "loss": 2.0037, "step": 17501 }, { "epoch": 0.43, "learning_rate": 1.267496508095581e-05, "loss": 1.9646, "step": 17502 }, { "epoch": 0.43, "learning_rate": 1.2674195833425386e-05, "loss": 2.1955, "step": 17503 }, { "epoch": 0.43, "learning_rate": 1.267342656885144e-05, "loss": 2.0286, "step": 17504 }, { "epoch": 0.43, "learning_rate": 1.2672657287238875e-05, "loss": 2.0677, "step": 17505 }, { "epoch": 0.43, "learning_rate": 1.2671887988592593e-05, "loss": 1.9822, "step": 17506 }, { "epoch": 0.43, "learning_rate": 1.2671118672917499e-05, "loss": 1.9933, "step": 17507 }, { "epoch": 0.43, "learning_rate": 1.2670349340218493e-05, "loss": 2.2696, "step": 17508 }, { "epoch": 0.43, "learning_rate": 1.2669579990500482e-05, "loss": 2.0457, "step": 17509 }, { "epoch": 0.43, "learning_rate": 1.2668810623768368e-05, "loss": 2.241, "step": 17510 }, { "epoch": 0.43, "learning_rate": 1.2668041240027055e-05, "loss": 1.8878, "step": 17511 }, { "epoch": 0.43, "learning_rate": 1.2667271839281446e-05, "loss": 2.1289, "step": 17512 }, { "epoch": 0.43, "learning_rate": 1.2666502421536441e-05, "loss": 1.871, "step": 17513 }, { "epoch": 0.43, "learning_rate": 1.2665732986796952e-05, "loss": 2.1507, "step": 17514 }, { "epoch": 0.43, "learning_rate": 1.2664963535067876e-05, "loss": 2.0135, "step": 17515 }, { "epoch": 0.43, "learning_rate": 1.2664194066354117e-05, "loss": 2.0779, "step": 17516 }, { "epoch": 0.43, "learning_rate": 1.266342458066058e-05, "loss": 2.1279, "step": 17517 }, { "epoch": 0.43, "learning_rate": 1.2662655077992175e-05, "loss": 2.0953, "step": 17518 }, { "epoch": 0.43, "learning_rate": 1.2661885558353797e-05, "loss": 1.986, "step": 17519 }, { "epoch": 0.43, "learning_rate": 1.2661116021750356e-05, "loss": 2.0937, "step": 17520 }, { "epoch": 0.43, "learning_rate": 1.2660346468186757e-05, "loss": 2.009, "step": 17521 }, { "epoch": 0.43, "learning_rate": 1.2659576897667902e-05, "loss": 2.0951, "step": 17522 }, { "epoch": 0.43, "learning_rate": 1.2658807310198697e-05, "loss": 2.0923, "step": 17523 }, { "epoch": 0.43, "learning_rate": 1.2658037705784048e-05, "loss": 1.9423, "step": 17524 }, { "epoch": 0.43, "learning_rate": 1.2657268084428855e-05, "loss": 2.0812, "step": 17525 }, { "epoch": 0.43, "learning_rate": 1.2656498446138026e-05, "loss": 2.0816, "step": 17526 }, { "epoch": 0.43, "learning_rate": 1.265572879091647e-05, "loss": 2.0232, "step": 17527 }, { "epoch": 0.43, "learning_rate": 1.2654959118769087e-05, "loss": 2.2764, "step": 17528 }, { "epoch": 0.43, "learning_rate": 1.2654189429700788e-05, "loss": 2.0243, "step": 17529 }, { "epoch": 0.43, "learning_rate": 1.2653419723716472e-05, "loss": 2.1487, "step": 17530 }, { "epoch": 0.43, "learning_rate": 1.2652650000821046e-05, "loss": 2.0129, "step": 17531 }, { "epoch": 0.43, "learning_rate": 1.2651880261019418e-05, "loss": 2.1089, "step": 17532 }, { "epoch": 0.43, "learning_rate": 1.2651110504316495e-05, "loss": 2.1936, "step": 17533 }, { "epoch": 0.43, "learning_rate": 1.2650340730717176e-05, "loss": 2.007, "step": 17534 }, { "epoch": 0.43, "learning_rate": 1.2649570940226376e-05, "loss": 2.115, "step": 17535 }, { "epoch": 0.43, "learning_rate": 1.2648801132848996e-05, "loss": 2.0757, "step": 17536 }, { "epoch": 0.43, "learning_rate": 1.2648031308589944e-05, "loss": 2.0028, "step": 17537 }, { "epoch": 0.43, "learning_rate": 1.2647261467454124e-05, "loss": 1.8498, "step": 17538 }, { "epoch": 0.43, "learning_rate": 1.2646491609446445e-05, "loss": 2.0173, "step": 17539 }, { "epoch": 0.43, "learning_rate": 1.264572173457181e-05, "loss": 2.0434, "step": 17540 }, { "epoch": 0.43, "learning_rate": 1.2644951842835131e-05, "loss": 2.0832, "step": 17541 }, { "epoch": 0.43, "learning_rate": 1.2644181934241311e-05, "loss": 2.0086, "step": 17542 }, { "epoch": 0.43, "learning_rate": 1.2643412008795258e-05, "loss": 2.1929, "step": 17543 }, { "epoch": 0.43, "learning_rate": 1.2642642066501878e-05, "loss": 1.9788, "step": 17544 }, { "epoch": 0.43, "learning_rate": 1.2641872107366078e-05, "loss": 1.9722, "step": 17545 }, { "epoch": 0.43, "learning_rate": 1.2641102131392768e-05, "loss": 2.0956, "step": 17546 }, { "epoch": 0.43, "learning_rate": 1.2640332138586854e-05, "loss": 2.0571, "step": 17547 }, { "epoch": 0.43, "learning_rate": 1.2639562128953242e-05, "loss": 1.8672, "step": 17548 }, { "epoch": 0.43, "learning_rate": 1.2638792102496841e-05, "loss": 2.1816, "step": 17549 }, { "epoch": 0.43, "learning_rate": 1.2638022059222557e-05, "loss": 2.1133, "step": 17550 }, { "epoch": 0.43, "learning_rate": 1.2637251999135298e-05, "loss": 2.0552, "step": 17551 }, { "epoch": 0.43, "learning_rate": 1.263648192223997e-05, "loss": 2.1364, "step": 17552 }, { "epoch": 0.43, "learning_rate": 1.263571182854149e-05, "loss": 2.074, "step": 17553 }, { "epoch": 0.43, "learning_rate": 1.2634941718044758e-05, "loss": 2.1602, "step": 17554 }, { "epoch": 0.43, "learning_rate": 1.263417159075468e-05, "loss": 2.0812, "step": 17555 }, { "epoch": 0.43, "learning_rate": 1.263340144667617e-05, "loss": 1.9831, "step": 17556 }, { "epoch": 0.43, "learning_rate": 1.2632631285814136e-05, "loss": 2.0762, "step": 17557 }, { "epoch": 0.43, "learning_rate": 1.2631861108173485e-05, "loss": 2.0217, "step": 17558 }, { "epoch": 0.43, "learning_rate": 1.2631090913759122e-05, "loss": 1.9613, "step": 17559 }, { "epoch": 0.43, "learning_rate": 1.2630320702575962e-05, "loss": 2.052, "step": 17560 }, { "epoch": 0.43, "learning_rate": 1.2629550474628908e-05, "loss": 2.1338, "step": 17561 }, { "epoch": 0.43, "learning_rate": 1.2628780229922874e-05, "loss": 2.0869, "step": 17562 }, { "epoch": 0.43, "learning_rate": 1.2628009968462769e-05, "loss": 2.107, "step": 17563 }, { "epoch": 0.43, "learning_rate": 1.2627239690253496e-05, "loss": 2.1629, "step": 17564 }, { "epoch": 0.43, "learning_rate": 1.262646939529997e-05, "loss": 2.1514, "step": 17565 }, { "epoch": 0.43, "learning_rate": 1.26256990836071e-05, "loss": 2.0432, "step": 17566 }, { "epoch": 0.43, "learning_rate": 1.2624928755179793e-05, "loss": 2.1335, "step": 17567 }, { "epoch": 0.43, "learning_rate": 1.262415841002296e-05, "loss": 1.9683, "step": 17568 }, { "epoch": 0.43, "learning_rate": 1.2623388048141508e-05, "loss": 2.0917, "step": 17569 }, { "epoch": 0.43, "learning_rate": 1.2622617669540349e-05, "loss": 2.0428, "step": 17570 }, { "epoch": 0.43, "learning_rate": 1.2621847274224396e-05, "loss": 2.005, "step": 17571 }, { "epoch": 0.43, "learning_rate": 1.2621076862198555e-05, "loss": 1.9284, "step": 17572 }, { "epoch": 0.43, "learning_rate": 1.2620306433467737e-05, "loss": 2.2229, "step": 17573 }, { "epoch": 0.43, "learning_rate": 1.261953598803685e-05, "loss": 2.0642, "step": 17574 }, { "epoch": 0.43, "learning_rate": 1.261876552591081e-05, "loss": 2.0674, "step": 17575 }, { "epoch": 0.43, "learning_rate": 1.261799504709452e-05, "loss": 2.0553, "step": 17576 }, { "epoch": 0.43, "learning_rate": 1.2617224551592898e-05, "loss": 2.1515, "step": 17577 }, { "epoch": 0.43, "learning_rate": 1.2616454039410848e-05, "loss": 2.1005, "step": 17578 }, { "epoch": 0.43, "learning_rate": 1.2615683510553285e-05, "loss": 1.9845, "step": 17579 }, { "epoch": 0.43, "learning_rate": 1.261491296502512e-05, "loss": 2.0936, "step": 17580 }, { "epoch": 0.43, "learning_rate": 1.261414240283126e-05, "loss": 1.9867, "step": 17581 }, { "epoch": 0.43, "learning_rate": 1.2613371823976622e-05, "loss": 2.0686, "step": 17582 }, { "epoch": 0.43, "learning_rate": 1.261260122846611e-05, "loss": 1.9992, "step": 17583 }, { "epoch": 0.43, "learning_rate": 1.2611830616304639e-05, "loss": 2.0374, "step": 17584 }, { "epoch": 0.43, "learning_rate": 1.261105998749712e-05, "loss": 1.9758, "step": 17585 }, { "epoch": 0.43, "learning_rate": 1.2610289342048469e-05, "loss": 2.0534, "step": 17586 }, { "epoch": 0.43, "learning_rate": 1.2609518679963589e-05, "loss": 1.968, "step": 17587 }, { "epoch": 0.43, "learning_rate": 1.26087480012474e-05, "loss": 2.0856, "step": 17588 }, { "epoch": 0.43, "learning_rate": 1.2607977305904804e-05, "loss": 2.0722, "step": 17589 }, { "epoch": 0.43, "learning_rate": 1.2607206593940724e-05, "loss": 2.0377, "step": 17590 }, { "epoch": 0.43, "learning_rate": 1.2606435865360064e-05, "loss": 1.9298, "step": 17591 }, { "epoch": 0.43, "learning_rate": 1.2605665120167742e-05, "loss": 1.9772, "step": 17592 }, { "epoch": 0.43, "learning_rate": 1.2604894358368665e-05, "loss": 2.1662, "step": 17593 }, { "epoch": 0.43, "learning_rate": 1.2604123579967746e-05, "loss": 1.9666, "step": 17594 }, { "epoch": 0.43, "learning_rate": 1.26033527849699e-05, "loss": 2.0936, "step": 17595 }, { "epoch": 0.43, "learning_rate": 1.2602581973380037e-05, "loss": 2.0102, "step": 17596 }, { "epoch": 0.43, "learning_rate": 1.2601811145203074e-05, "loss": 2.1572, "step": 17597 }, { "epoch": 0.43, "learning_rate": 1.2601040300443916e-05, "loss": 2.0506, "step": 17598 }, { "epoch": 0.43, "learning_rate": 1.2600269439107484e-05, "loss": 1.9937, "step": 17599 }, { "epoch": 0.43, "learning_rate": 1.2599498561198689e-05, "loss": 1.995, "step": 17600 }, { "epoch": 0.43, "learning_rate": 1.259872766672244e-05, "loss": 2.2132, "step": 17601 }, { "epoch": 0.43, "learning_rate": 1.2597956755683654e-05, "loss": 1.7426, "step": 17602 }, { "epoch": 0.43, "learning_rate": 1.259718582808724e-05, "loss": 1.9827, "step": 17603 }, { "epoch": 0.43, "learning_rate": 1.2596414883938117e-05, "loss": 2.0778, "step": 17604 }, { "epoch": 0.43, "learning_rate": 1.2595643923241194e-05, "loss": 2.0309, "step": 17605 }, { "epoch": 0.43, "learning_rate": 1.2594872946001389e-05, "loss": 2.1794, "step": 17606 }, { "epoch": 0.43, "learning_rate": 1.2594101952223612e-05, "loss": 2.1211, "step": 17607 }, { "epoch": 0.43, "learning_rate": 1.2593330941912776e-05, "loss": 1.9827, "step": 17608 }, { "epoch": 0.43, "learning_rate": 1.25925599150738e-05, "loss": 2.1766, "step": 17609 }, { "epoch": 0.43, "learning_rate": 1.2591788871711592e-05, "loss": 2.0292, "step": 17610 }, { "epoch": 0.43, "learning_rate": 1.259101781183107e-05, "loss": 2.1609, "step": 17611 }, { "epoch": 0.43, "learning_rate": 1.2590246735437149e-05, "loss": 2.1259, "step": 17612 }, { "epoch": 0.43, "learning_rate": 1.2589475642534737e-05, "loss": 2.1547, "step": 17613 }, { "epoch": 0.43, "learning_rate": 1.2588704533128752e-05, "loss": 2.0574, "step": 17614 }, { "epoch": 0.43, "learning_rate": 1.2587933407224116e-05, "loss": 2.03, "step": 17615 }, { "epoch": 0.43, "learning_rate": 1.2587162264825733e-05, "loss": 1.8933, "step": 17616 }, { "epoch": 0.43, "learning_rate": 1.2586391105938523e-05, "loss": 2.0458, "step": 17617 }, { "epoch": 0.43, "learning_rate": 1.2585619930567398e-05, "loss": 1.8922, "step": 17618 }, { "epoch": 0.43, "learning_rate": 1.2584848738717277e-05, "loss": 2.1203, "step": 17619 }, { "epoch": 0.43, "learning_rate": 1.258407753039307e-05, "loss": 2.0556, "step": 17620 }, { "epoch": 0.43, "learning_rate": 1.2583306305599696e-05, "loss": 2.0291, "step": 17621 }, { "epoch": 0.43, "learning_rate": 1.2582535064342067e-05, "loss": 2.0919, "step": 17622 }, { "epoch": 0.43, "learning_rate": 1.2581763806625102e-05, "loss": 1.9474, "step": 17623 }, { "epoch": 0.43, "learning_rate": 1.2580992532453715e-05, "loss": 2.1225, "step": 17624 }, { "epoch": 0.43, "learning_rate": 1.2580221241832824e-05, "loss": 2.1288, "step": 17625 }, { "epoch": 0.43, "learning_rate": 1.257944993476734e-05, "loss": 2.0694, "step": 17626 }, { "epoch": 0.43, "learning_rate": 1.2578678611262179e-05, "loss": 2.1359, "step": 17627 }, { "epoch": 0.43, "learning_rate": 1.257790727132226e-05, "loss": 2.1268, "step": 17628 }, { "epoch": 0.43, "learning_rate": 1.25771359149525e-05, "loss": 2.0586, "step": 17629 }, { "epoch": 0.43, "learning_rate": 1.2576364542157812e-05, "loss": 1.9719, "step": 17630 }, { "epoch": 0.43, "learning_rate": 1.2575593152943111e-05, "loss": 2.1171, "step": 17631 }, { "epoch": 0.43, "learning_rate": 1.2574821747313318e-05, "loss": 1.8477, "step": 17632 }, { "epoch": 0.43, "learning_rate": 1.2574050325273346e-05, "loss": 2.1573, "step": 17633 }, { "epoch": 0.43, "learning_rate": 1.257327888682811e-05, "loss": 2.0721, "step": 17634 }, { "epoch": 0.43, "learning_rate": 1.2572507431982532e-05, "loss": 2.1644, "step": 17635 }, { "epoch": 0.43, "learning_rate": 1.2571735960741528e-05, "loss": 2.1422, "step": 17636 }, { "epoch": 0.43, "learning_rate": 1.2570964473110007e-05, "loss": 1.976, "step": 17637 }, { "epoch": 0.43, "learning_rate": 1.2570192969092896e-05, "loss": 1.9363, "step": 17638 }, { "epoch": 0.43, "learning_rate": 1.2569421448695107e-05, "loss": 2.1397, "step": 17639 }, { "epoch": 0.43, "learning_rate": 1.2568649911921555e-05, "loss": 2.1141, "step": 17640 }, { "epoch": 0.43, "learning_rate": 1.2567878358777161e-05, "loss": 2.0523, "step": 17641 }, { "epoch": 0.43, "learning_rate": 1.2567106789266844e-05, "loss": 2.3414, "step": 17642 }, { "epoch": 0.43, "learning_rate": 1.2566335203395514e-05, "loss": 1.9381, "step": 17643 }, { "epoch": 0.43, "learning_rate": 1.2565563601168098e-05, "loss": 1.907, "step": 17644 }, { "epoch": 0.43, "learning_rate": 1.2564791982589509e-05, "loss": 1.9359, "step": 17645 }, { "epoch": 0.43, "learning_rate": 1.2564020347664663e-05, "loss": 2.1678, "step": 17646 }, { "epoch": 0.43, "learning_rate": 1.256324869639848e-05, "loss": 2.2409, "step": 17647 }, { "epoch": 0.44, "learning_rate": 1.2562477028795877e-05, "loss": 1.9369, "step": 17648 }, { "epoch": 0.44, "learning_rate": 1.2561705344861772e-05, "loss": 2.015, "step": 17649 }, { "epoch": 0.44, "learning_rate": 1.2560933644601088e-05, "loss": 2.0207, "step": 17650 }, { "epoch": 0.44, "learning_rate": 1.2560161928018735e-05, "loss": 1.9557, "step": 17651 }, { "epoch": 0.44, "learning_rate": 1.2559390195119638e-05, "loss": 2.1165, "step": 17652 }, { "epoch": 0.44, "learning_rate": 1.2558618445908712e-05, "loss": 2.1294, "step": 17653 }, { "epoch": 0.44, "learning_rate": 1.255784668039088e-05, "loss": 2.0459, "step": 17654 }, { "epoch": 0.44, "learning_rate": 1.2557074898571054e-05, "loss": 1.9191, "step": 17655 }, { "epoch": 0.44, "learning_rate": 1.2556303100454157e-05, "loss": 2.0981, "step": 17656 }, { "epoch": 0.44, "learning_rate": 1.2555531286045108e-05, "loss": 2.1235, "step": 17657 }, { "epoch": 0.44, "learning_rate": 1.2554759455348824e-05, "loss": 2.1061, "step": 17658 }, { "epoch": 0.44, "learning_rate": 1.2553987608370227e-05, "loss": 2.0382, "step": 17659 }, { "epoch": 0.44, "learning_rate": 1.2553215745114233e-05, "loss": 1.8981, "step": 17660 }, { "epoch": 0.44, "learning_rate": 1.2552443865585764e-05, "loss": 1.9563, "step": 17661 }, { "epoch": 0.44, "learning_rate": 1.2551671969789739e-05, "loss": 2.0517, "step": 17662 }, { "epoch": 0.44, "learning_rate": 1.2550900057731078e-05, "loss": 2.0532, "step": 17663 }, { "epoch": 0.44, "learning_rate": 1.2550128129414698e-05, "loss": 2.0596, "step": 17664 }, { "epoch": 0.44, "learning_rate": 1.254935618484552e-05, "loss": 2.1846, "step": 17665 }, { "epoch": 0.44, "learning_rate": 1.2548584224028465e-05, "loss": 2.0776, "step": 17666 }, { "epoch": 0.44, "learning_rate": 1.254781224696845e-05, "loss": 2.0096, "step": 17667 }, { "epoch": 0.44, "learning_rate": 1.25470402536704e-05, "loss": 2.1082, "step": 17668 }, { "epoch": 0.44, "learning_rate": 1.254626824413923e-05, "loss": 1.9478, "step": 17669 }, { "epoch": 0.44, "learning_rate": 1.2545496218379867e-05, "loss": 2.0104, "step": 17670 }, { "epoch": 0.44, "learning_rate": 1.2544724176397223e-05, "loss": 1.9763, "step": 17671 }, { "epoch": 0.44, "learning_rate": 1.2543952118196223e-05, "loss": 2.0189, "step": 17672 }, { "epoch": 0.44, "learning_rate": 1.2543180043781789e-05, "loss": 1.9594, "step": 17673 }, { "epoch": 0.44, "learning_rate": 1.254240795315884e-05, "loss": 1.8907, "step": 17674 }, { "epoch": 0.44, "learning_rate": 1.2541635846332291e-05, "loss": 2.0458, "step": 17675 }, { "epoch": 0.44, "learning_rate": 1.2540863723307073e-05, "loss": 1.9038, "step": 17676 }, { "epoch": 0.44, "learning_rate": 1.25400915840881e-05, "loss": 2.0777, "step": 17677 }, { "epoch": 0.44, "learning_rate": 1.25393194286803e-05, "loss": 1.9581, "step": 17678 }, { "epoch": 0.44, "learning_rate": 1.253854725708859e-05, "loss": 1.989, "step": 17679 }, { "epoch": 0.44, "learning_rate": 1.2537775069317885e-05, "loss": 1.9927, "step": 17680 }, { "epoch": 0.44, "learning_rate": 1.2537002865373116e-05, "loss": 1.9178, "step": 17681 }, { "epoch": 0.44, "learning_rate": 1.2536230645259202e-05, "loss": 2.0745, "step": 17682 }, { "epoch": 0.44, "learning_rate": 1.2535458408981062e-05, "loss": 2.0155, "step": 17683 }, { "epoch": 0.44, "learning_rate": 1.253468615654362e-05, "loss": 2.194, "step": 17684 }, { "epoch": 0.44, "learning_rate": 1.2533913887951794e-05, "loss": 1.9822, "step": 17685 }, { "epoch": 0.44, "learning_rate": 1.2533141603210513e-05, "loss": 2.1721, "step": 17686 }, { "epoch": 0.44, "learning_rate": 1.2532369302324696e-05, "loss": 1.9995, "step": 17687 }, { "epoch": 0.44, "learning_rate": 1.2531596985299261e-05, "loss": 2.1717, "step": 17688 }, { "epoch": 0.44, "learning_rate": 1.2530824652139136e-05, "loss": 1.9787, "step": 17689 }, { "epoch": 0.44, "learning_rate": 1.2530052302849237e-05, "loss": 2.0489, "step": 17690 }, { "epoch": 0.44, "learning_rate": 1.2529279937434493e-05, "loss": 2.1254, "step": 17691 }, { "epoch": 0.44, "learning_rate": 1.2528507555899823e-05, "loss": 2.0128, "step": 17692 }, { "epoch": 0.44, "learning_rate": 1.252773515825015e-05, "loss": 2.0039, "step": 17693 }, { "epoch": 0.44, "learning_rate": 1.2526962744490398e-05, "loss": 2.1413, "step": 17694 }, { "epoch": 0.44, "learning_rate": 1.2526190314625485e-05, "loss": 2.0674, "step": 17695 }, { "epoch": 0.44, "learning_rate": 1.2525417868660343e-05, "loss": 2.0393, "step": 17696 }, { "epoch": 0.44, "learning_rate": 1.2524645406599889e-05, "loss": 1.9541, "step": 17697 }, { "epoch": 0.44, "learning_rate": 1.2523872928449045e-05, "loss": 2.0488, "step": 17698 }, { "epoch": 0.44, "learning_rate": 1.2523100434212738e-05, "loss": 2.0782, "step": 17699 }, { "epoch": 0.44, "learning_rate": 1.2522327923895889e-05, "loss": 2.1688, "step": 17700 }, { "epoch": 0.44, "learning_rate": 1.2521555397503422e-05, "loss": 2.2579, "step": 17701 }, { "epoch": 0.44, "learning_rate": 1.252078285504026e-05, "loss": 2.0903, "step": 17702 }, { "epoch": 0.44, "learning_rate": 1.252001029651133e-05, "loss": 2.0481, "step": 17703 }, { "epoch": 0.44, "learning_rate": 1.2519237721921551e-05, "loss": 1.9626, "step": 17704 }, { "epoch": 0.44, "learning_rate": 1.2518465131275848e-05, "loss": 2.0851, "step": 17705 }, { "epoch": 0.44, "learning_rate": 1.2517692524579148e-05, "loss": 2.1898, "step": 17706 }, { "epoch": 0.44, "learning_rate": 1.2516919901836373e-05, "loss": 2.0566, "step": 17707 }, { "epoch": 0.44, "learning_rate": 1.2516147263052447e-05, "loss": 2.1617, "step": 17708 }, { "epoch": 0.44, "learning_rate": 1.2515374608232294e-05, "loss": 2.0203, "step": 17709 }, { "epoch": 0.44, "learning_rate": 1.2514601937380839e-05, "loss": 2.2106, "step": 17710 }, { "epoch": 0.44, "learning_rate": 1.2513829250503004e-05, "loss": 2.0716, "step": 17711 }, { "epoch": 0.44, "learning_rate": 1.2513056547603719e-05, "loss": 1.9496, "step": 17712 }, { "epoch": 0.44, "learning_rate": 1.2512283828687905e-05, "loss": 2.1695, "step": 17713 }, { "epoch": 0.44, "learning_rate": 1.2511511093760486e-05, "loss": 2.0679, "step": 17714 }, { "epoch": 0.44, "learning_rate": 1.2510738342826391e-05, "loss": 1.9873, "step": 17715 }, { "epoch": 0.44, "learning_rate": 1.2509965575890543e-05, "loss": 1.9003, "step": 17716 }, { "epoch": 0.44, "learning_rate": 1.2509192792957864e-05, "loss": 1.9442, "step": 17717 }, { "epoch": 0.44, "learning_rate": 1.2508419994033281e-05, "loss": 1.8589, "step": 17718 }, { "epoch": 0.44, "learning_rate": 1.2507647179121722e-05, "loss": 2.0474, "step": 17719 }, { "epoch": 0.44, "learning_rate": 1.2506874348228109e-05, "loss": 2.3253, "step": 17720 }, { "epoch": 0.44, "learning_rate": 1.2506101501357368e-05, "loss": 1.9985, "step": 17721 }, { "epoch": 0.44, "learning_rate": 1.2505328638514427e-05, "loss": 1.8712, "step": 17722 }, { "epoch": 0.44, "learning_rate": 1.250455575970421e-05, "loss": 1.9823, "step": 17723 }, { "epoch": 0.44, "learning_rate": 1.2503782864931642e-05, "loss": 1.9639, "step": 17724 }, { "epoch": 0.44, "learning_rate": 1.250300995420165e-05, "loss": 2.0414, "step": 17725 }, { "epoch": 0.44, "learning_rate": 1.2502237027519162e-05, "loss": 2.0751, "step": 17726 }, { "epoch": 0.44, "learning_rate": 1.2501464084889099e-05, "loss": 1.9996, "step": 17727 }, { "epoch": 0.44, "learning_rate": 1.2500691126316391e-05, "loss": 2.1425, "step": 17728 }, { "epoch": 0.44, "learning_rate": 1.2499918151805963e-05, "loss": 2.0838, "step": 17729 }, { "epoch": 0.44, "learning_rate": 1.2499145161362745e-05, "loss": 1.952, "step": 17730 }, { "epoch": 0.44, "learning_rate": 1.2498372154991658e-05, "loss": 2.2432, "step": 17731 }, { "epoch": 0.44, "learning_rate": 1.2497599132697631e-05, "loss": 2.0639, "step": 17732 }, { "epoch": 0.44, "learning_rate": 1.2496826094485592e-05, "loss": 1.8811, "step": 17733 }, { "epoch": 0.44, "learning_rate": 1.2496053040360465e-05, "loss": 2.1039, "step": 17734 }, { "epoch": 0.44, "learning_rate": 1.2495279970327178e-05, "loss": 1.9147, "step": 17735 }, { "epoch": 0.44, "learning_rate": 1.2494506884390663e-05, "loss": 2.0511, "step": 17736 }, { "epoch": 0.44, "learning_rate": 1.2493733782555838e-05, "loss": 1.9854, "step": 17737 }, { "epoch": 0.44, "learning_rate": 1.2492960664827635e-05, "loss": 1.9287, "step": 17738 }, { "epoch": 0.44, "learning_rate": 1.2492187531210982e-05, "loss": 2.0489, "step": 17739 }, { "epoch": 0.44, "learning_rate": 1.2491414381710807e-05, "loss": 1.8736, "step": 17740 }, { "epoch": 0.44, "learning_rate": 1.2490641216332038e-05, "loss": 2.171, "step": 17741 }, { "epoch": 0.44, "learning_rate": 1.2489868035079598e-05, "loss": 1.9646, "step": 17742 }, { "epoch": 0.44, "learning_rate": 1.2489094837958415e-05, "loss": 1.8991, "step": 17743 }, { "epoch": 0.44, "learning_rate": 1.2488321624973426e-05, "loss": 2.0287, "step": 17744 }, { "epoch": 0.44, "learning_rate": 1.2487548396129547e-05, "loss": 1.9869, "step": 17745 }, { "epoch": 0.44, "learning_rate": 1.2486775151431714e-05, "loss": 1.969, "step": 17746 }, { "epoch": 0.44, "learning_rate": 1.248600189088485e-05, "loss": 2.1118, "step": 17747 }, { "epoch": 0.44, "learning_rate": 1.2485228614493887e-05, "loss": 2.0999, "step": 17748 }, { "epoch": 0.44, "learning_rate": 1.2484455322263753e-05, "loss": 1.9955, "step": 17749 }, { "epoch": 0.44, "learning_rate": 1.2483682014199377e-05, "loss": 2.0741, "step": 17750 }, { "epoch": 0.44, "learning_rate": 1.2482908690305684e-05, "loss": 1.958, "step": 17751 }, { "epoch": 0.44, "learning_rate": 1.2482135350587607e-05, "loss": 1.9738, "step": 17752 }, { "epoch": 0.44, "learning_rate": 1.2481361995050068e-05, "loss": 2.0943, "step": 17753 }, { "epoch": 0.44, "learning_rate": 1.2480588623698004e-05, "loss": 2.1573, "step": 17754 }, { "epoch": 0.44, "learning_rate": 1.2479815236536337e-05, "loss": 2.0399, "step": 17755 }, { "epoch": 0.44, "learning_rate": 1.2479041833570003e-05, "loss": 2.277, "step": 17756 }, { "epoch": 0.44, "learning_rate": 1.2478268414803926e-05, "loss": 2.0273, "step": 17757 }, { "epoch": 0.44, "learning_rate": 1.2477494980243036e-05, "loss": 2.0719, "step": 17758 }, { "epoch": 0.44, "learning_rate": 1.2476721529892264e-05, "loss": 1.9788, "step": 17759 }, { "epoch": 0.44, "learning_rate": 1.247594806375654e-05, "loss": 2.242, "step": 17760 }, { "epoch": 0.44, "learning_rate": 1.2475174581840792e-05, "loss": 1.9783, "step": 17761 }, { "epoch": 0.44, "learning_rate": 1.2474401084149945e-05, "loss": 2.013, "step": 17762 }, { "epoch": 0.44, "learning_rate": 1.247362757068894e-05, "loss": 2.2404, "step": 17763 }, { "epoch": 0.44, "learning_rate": 1.2472854041462694e-05, "loss": 1.9731, "step": 17764 }, { "epoch": 0.44, "learning_rate": 1.2472080496476149e-05, "loss": 2.036, "step": 17765 }, { "epoch": 0.44, "learning_rate": 1.2471306935734226e-05, "loss": 2.149, "step": 17766 }, { "epoch": 0.44, "learning_rate": 1.247053335924186e-05, "loss": 2.1144, "step": 17767 }, { "epoch": 0.44, "learning_rate": 1.246975976700398e-05, "loss": 2.0139, "step": 17768 }, { "epoch": 0.44, "learning_rate": 1.2468986159025513e-05, "loss": 2.1534, "step": 17769 }, { "epoch": 0.44, "learning_rate": 1.2468212535311397e-05, "loss": 2.0863, "step": 17770 }, { "epoch": 0.44, "learning_rate": 1.2467438895866554e-05, "loss": 2.0946, "step": 17771 }, { "epoch": 0.44, "learning_rate": 1.2466665240695921e-05, "loss": 1.9354, "step": 17772 }, { "epoch": 0.44, "learning_rate": 1.2465891569804428e-05, "loss": 1.8944, "step": 17773 }, { "epoch": 0.44, "learning_rate": 1.2465117883197002e-05, "loss": 2.0864, "step": 17774 }, { "epoch": 0.44, "learning_rate": 1.2464344180878577e-05, "loss": 1.9797, "step": 17775 }, { "epoch": 0.44, "learning_rate": 1.2463570462854085e-05, "loss": 2.0628, "step": 17776 }, { "epoch": 0.44, "learning_rate": 1.2462796729128452e-05, "loss": 2.1506, "step": 17777 }, { "epoch": 0.44, "learning_rate": 1.2462022979706617e-05, "loss": 2.123, "step": 17778 }, { "epoch": 0.44, "learning_rate": 1.2461249214593504e-05, "loss": 2.0053, "step": 17779 }, { "epoch": 0.44, "learning_rate": 1.2460475433794048e-05, "loss": 1.8629, "step": 17780 }, { "epoch": 0.44, "learning_rate": 1.2459701637313181e-05, "loss": 2.072, "step": 17781 }, { "epoch": 0.44, "learning_rate": 1.2458927825155835e-05, "loss": 2.0829, "step": 17782 }, { "epoch": 0.44, "learning_rate": 1.245815399732694e-05, "loss": 2.0662, "step": 17783 }, { "epoch": 0.44, "learning_rate": 1.2457380153831428e-05, "loss": 2.1007, "step": 17784 }, { "epoch": 0.44, "learning_rate": 1.2456606294674231e-05, "loss": 1.9641, "step": 17785 }, { "epoch": 0.44, "learning_rate": 1.2455832419860286e-05, "loss": 2.0369, "step": 17786 }, { "epoch": 0.44, "learning_rate": 1.2455058529394512e-05, "loss": 2.117, "step": 17787 }, { "epoch": 0.44, "learning_rate": 1.2454284623281855e-05, "loss": 2.1388, "step": 17788 }, { "epoch": 0.44, "learning_rate": 1.2453510701527243e-05, "loss": 2.104, "step": 17789 }, { "epoch": 0.44, "learning_rate": 1.2452736764135607e-05, "loss": 1.9557, "step": 17790 }, { "epoch": 0.44, "learning_rate": 1.2451962811111879e-05, "loss": 2.0523, "step": 17791 }, { "epoch": 0.44, "learning_rate": 1.2451188842460993e-05, "loss": 2.0363, "step": 17792 }, { "epoch": 0.44, "learning_rate": 1.2450414858187885e-05, "loss": 2.0638, "step": 17793 }, { "epoch": 0.44, "learning_rate": 1.2449640858297482e-05, "loss": 2.3348, "step": 17794 }, { "epoch": 0.44, "learning_rate": 1.244886684279472e-05, "loss": 2.0167, "step": 17795 }, { "epoch": 0.44, "learning_rate": 1.2448092811684531e-05, "loss": 2.2551, "step": 17796 }, { "epoch": 0.44, "learning_rate": 1.2447318764971848e-05, "loss": 2.0204, "step": 17797 }, { "epoch": 0.44, "learning_rate": 1.2446544702661605e-05, "loss": 1.9646, "step": 17798 }, { "epoch": 0.44, "learning_rate": 1.2445770624758735e-05, "loss": 2.0238, "step": 17799 }, { "epoch": 0.44, "learning_rate": 1.2444996531268175e-05, "loss": 2.1603, "step": 17800 }, { "epoch": 0.44, "learning_rate": 1.2444222422194852e-05, "loss": 2.3089, "step": 17801 }, { "epoch": 0.44, "learning_rate": 1.2443448297543706e-05, "loss": 2.1566, "step": 17802 }, { "epoch": 0.44, "learning_rate": 1.2442674157319666e-05, "loss": 1.8682, "step": 17803 }, { "epoch": 0.44, "learning_rate": 1.2441900001527667e-05, "loss": 1.9382, "step": 17804 }, { "epoch": 0.44, "learning_rate": 1.2441125830172645e-05, "loss": 1.9246, "step": 17805 }, { "epoch": 0.44, "learning_rate": 1.2440351643259529e-05, "loss": 2.1514, "step": 17806 }, { "epoch": 0.44, "learning_rate": 1.2439577440793258e-05, "loss": 2.1472, "step": 17807 }, { "epoch": 0.44, "learning_rate": 1.2438803222778766e-05, "loss": 1.9606, "step": 17808 }, { "epoch": 0.44, "learning_rate": 1.2438028989220987e-05, "loss": 2.062, "step": 17809 }, { "epoch": 0.44, "learning_rate": 1.2437254740124856e-05, "loss": 2.0412, "step": 17810 }, { "epoch": 0.44, "learning_rate": 1.2436480475495303e-05, "loss": 1.9644, "step": 17811 }, { "epoch": 0.44, "learning_rate": 1.2435706195337266e-05, "loss": 2.0227, "step": 17812 }, { "epoch": 0.44, "learning_rate": 1.2434931899655682e-05, "loss": 1.967, "step": 17813 }, { "epoch": 0.44, "learning_rate": 1.2434157588455484e-05, "loss": 2.0896, "step": 17814 }, { "epoch": 0.44, "learning_rate": 1.2433383261741602e-05, "loss": 1.8882, "step": 17815 }, { "epoch": 0.44, "learning_rate": 1.2432608919518978e-05, "loss": 2.0459, "step": 17816 }, { "epoch": 0.44, "learning_rate": 1.2431834561792544e-05, "loss": 1.9959, "step": 17817 }, { "epoch": 0.44, "learning_rate": 1.243106018856724e-05, "loss": 2.0755, "step": 17818 }, { "epoch": 0.44, "learning_rate": 1.2430285799847992e-05, "loss": 2.029, "step": 17819 }, { "epoch": 0.44, "learning_rate": 1.2429511395639744e-05, "loss": 2.1033, "step": 17820 }, { "epoch": 0.44, "learning_rate": 1.2428736975947423e-05, "loss": 1.889, "step": 17821 }, { "epoch": 0.44, "learning_rate": 1.2427962540775976e-05, "loss": 2.0479, "step": 17822 }, { "epoch": 0.44, "learning_rate": 1.242718809013033e-05, "loss": 2.153, "step": 17823 }, { "epoch": 0.44, "learning_rate": 1.242641362401542e-05, "loss": 2.1999, "step": 17824 }, { "epoch": 0.44, "learning_rate": 1.242563914243619e-05, "loss": 2.1094, "step": 17825 }, { "epoch": 0.44, "learning_rate": 1.2424864645397567e-05, "loss": 2.1278, "step": 17826 }, { "epoch": 0.44, "learning_rate": 1.2424090132904497e-05, "loss": 1.9236, "step": 17827 }, { "epoch": 0.44, "learning_rate": 1.2423315604961908e-05, "loss": 2.1888, "step": 17828 }, { "epoch": 0.44, "learning_rate": 1.2422541061574737e-05, "loss": 2.0804, "step": 17829 }, { "epoch": 0.44, "learning_rate": 1.2421766502747923e-05, "loss": 2.1022, "step": 17830 }, { "epoch": 0.44, "learning_rate": 1.2420991928486404e-05, "loss": 2.0288, "step": 17831 }, { "epoch": 0.44, "learning_rate": 1.2420217338795113e-05, "loss": 2.1505, "step": 17832 }, { "epoch": 0.44, "learning_rate": 1.2419442733678988e-05, "loss": 1.9521, "step": 17833 }, { "epoch": 0.44, "learning_rate": 1.2418668113142966e-05, "loss": 2.1993, "step": 17834 }, { "epoch": 0.44, "learning_rate": 1.2417893477191982e-05, "loss": 2.0604, "step": 17835 }, { "epoch": 0.44, "learning_rate": 1.2417118825830978e-05, "loss": 2.1004, "step": 17836 }, { "epoch": 0.44, "learning_rate": 1.2416344159064888e-05, "loss": 2.033, "step": 17837 }, { "epoch": 0.44, "learning_rate": 1.2415569476898649e-05, "loss": 1.972, "step": 17838 }, { "epoch": 0.44, "learning_rate": 1.2414794779337199e-05, "loss": 1.8736, "step": 17839 }, { "epoch": 0.44, "learning_rate": 1.241402006638547e-05, "loss": 2.0762, "step": 17840 }, { "epoch": 0.44, "learning_rate": 1.241324533804841e-05, "loss": 2.0463, "step": 17841 }, { "epoch": 0.44, "learning_rate": 1.2412470594330947e-05, "loss": 2.0584, "step": 17842 }, { "epoch": 0.44, "learning_rate": 1.2411695835238025e-05, "loss": 2.1324, "step": 17843 }, { "epoch": 0.44, "learning_rate": 1.2410921060774581e-05, "loss": 2.1782, "step": 17844 }, { "epoch": 0.44, "learning_rate": 1.2410146270945547e-05, "loss": 1.9067, "step": 17845 }, { "epoch": 0.44, "learning_rate": 1.2409371465755868e-05, "loss": 2.236, "step": 17846 }, { "epoch": 0.44, "learning_rate": 1.2408596645210482e-05, "loss": 1.9749, "step": 17847 }, { "epoch": 0.44, "learning_rate": 1.2407821809314322e-05, "loss": 1.9316, "step": 17848 }, { "epoch": 0.44, "learning_rate": 1.2407046958072328e-05, "loss": 2.0145, "step": 17849 }, { "epoch": 0.44, "learning_rate": 1.240627209148944e-05, "loss": 2.086, "step": 17850 }, { "epoch": 0.44, "learning_rate": 1.2405497209570597e-05, "loss": 2.12, "step": 17851 }, { "epoch": 0.44, "learning_rate": 1.2404722312320735e-05, "loss": 2.0196, "step": 17852 }, { "epoch": 0.44, "learning_rate": 1.2403947399744794e-05, "loss": 2.1096, "step": 17853 }, { "epoch": 0.44, "learning_rate": 1.2403172471847715e-05, "loss": 2.0604, "step": 17854 }, { "epoch": 0.44, "learning_rate": 1.2402397528634431e-05, "loss": 2.1619, "step": 17855 }, { "epoch": 0.44, "learning_rate": 1.2401622570109888e-05, "loss": 2.1034, "step": 17856 }, { "epoch": 0.44, "learning_rate": 1.240084759627902e-05, "loss": 2.0761, "step": 17857 }, { "epoch": 0.44, "learning_rate": 1.2400072607146769e-05, "loss": 2.1108, "step": 17858 }, { "epoch": 0.44, "learning_rate": 1.239929760271807e-05, "loss": 2.1547, "step": 17859 }, { "epoch": 0.44, "learning_rate": 1.2398522582997867e-05, "loss": 2.0576, "step": 17860 }, { "epoch": 0.44, "learning_rate": 1.23977475479911e-05, "loss": 2.0231, "step": 17861 }, { "epoch": 0.44, "learning_rate": 1.2396972497702704e-05, "loss": 2.0908, "step": 17862 }, { "epoch": 0.44, "learning_rate": 1.2396197432137624e-05, "loss": 1.9952, "step": 17863 }, { "epoch": 0.44, "learning_rate": 1.2395422351300793e-05, "loss": 2.0971, "step": 17864 }, { "epoch": 0.44, "learning_rate": 1.2394647255197158e-05, "loss": 2.2361, "step": 17865 }, { "epoch": 0.44, "learning_rate": 1.2393872143831653e-05, "loss": 2.1571, "step": 17866 }, { "epoch": 0.44, "learning_rate": 1.2393097017209223e-05, "loss": 1.9612, "step": 17867 }, { "epoch": 0.44, "learning_rate": 1.2392321875334806e-05, "loss": 2.0404, "step": 17868 }, { "epoch": 0.44, "learning_rate": 1.2391546718213338e-05, "loss": 2.0443, "step": 17869 }, { "epoch": 0.44, "learning_rate": 1.2390771545849767e-05, "loss": 1.974, "step": 17870 }, { "epoch": 0.44, "learning_rate": 1.238999635824903e-05, "loss": 2.1534, "step": 17871 }, { "epoch": 0.44, "learning_rate": 1.2389221155416067e-05, "loss": 1.9231, "step": 17872 }, { "epoch": 0.44, "learning_rate": 1.2388445937355819e-05, "loss": 1.9564, "step": 17873 }, { "epoch": 0.44, "learning_rate": 1.2387670704073225e-05, "loss": 2.0804, "step": 17874 }, { "epoch": 0.44, "learning_rate": 1.238689545557323e-05, "loss": 1.9789, "step": 17875 }, { "epoch": 0.44, "learning_rate": 1.2386120191860772e-05, "loss": 2.0077, "step": 17876 }, { "epoch": 0.44, "learning_rate": 1.238534491294079e-05, "loss": 2.1305, "step": 17877 }, { "epoch": 0.44, "learning_rate": 1.238456961881823e-05, "loss": 2.0741, "step": 17878 }, { "epoch": 0.44, "learning_rate": 1.2383794309498028e-05, "loss": 2.0798, "step": 17879 }, { "epoch": 0.44, "learning_rate": 1.2383018984985128e-05, "loss": 2.0031, "step": 17880 }, { "epoch": 0.44, "learning_rate": 1.2382243645284476e-05, "loss": 1.9554, "step": 17881 }, { "epoch": 0.44, "learning_rate": 1.2381468290401006e-05, "loss": 1.9401, "step": 17882 }, { "epoch": 0.44, "learning_rate": 1.2380692920339662e-05, "loss": 2.0015, "step": 17883 }, { "epoch": 0.44, "learning_rate": 1.2379917535105385e-05, "loss": 2.0797, "step": 17884 }, { "epoch": 0.44, "learning_rate": 1.237914213470312e-05, "loss": 1.9775, "step": 17885 }, { "epoch": 0.44, "learning_rate": 1.2378366719137806e-05, "loss": 1.985, "step": 17886 }, { "epoch": 0.44, "learning_rate": 1.2377591288414385e-05, "loss": 2.0982, "step": 17887 }, { "epoch": 0.44, "learning_rate": 1.2376815842537802e-05, "loss": 2.3346, "step": 17888 }, { "epoch": 0.44, "learning_rate": 1.2376040381512997e-05, "loss": 1.9712, "step": 17889 }, { "epoch": 0.44, "learning_rate": 1.2375264905344913e-05, "loss": 1.9268, "step": 17890 }, { "epoch": 0.44, "learning_rate": 1.2374489414038488e-05, "loss": 2.0695, "step": 17891 }, { "epoch": 0.44, "learning_rate": 1.2373713907598673e-05, "loss": 2.1592, "step": 17892 }, { "epoch": 0.44, "learning_rate": 1.2372938386030399e-05, "loss": 2.1432, "step": 17893 }, { "epoch": 0.44, "learning_rate": 1.2372162849338623e-05, "loss": 1.969, "step": 17894 }, { "epoch": 0.44, "learning_rate": 1.2371387297528274e-05, "loss": 2.1118, "step": 17895 }, { "epoch": 0.44, "learning_rate": 1.2370611730604303e-05, "loss": 1.9789, "step": 17896 }, { "epoch": 0.44, "learning_rate": 1.2369836148571652e-05, "loss": 2.1428, "step": 17897 }, { "epoch": 0.44, "learning_rate": 1.2369060551435261e-05, "loss": 2.0428, "step": 17898 }, { "epoch": 0.44, "learning_rate": 1.2368284939200077e-05, "loss": 2.1045, "step": 17899 }, { "epoch": 0.44, "learning_rate": 1.2367509311871042e-05, "loss": 2.0684, "step": 17900 }, { "epoch": 0.44, "learning_rate": 1.2366733669453096e-05, "loss": 2.1276, "step": 17901 }, { "epoch": 0.44, "learning_rate": 1.2365958011951188e-05, "loss": 1.9115, "step": 17902 }, { "epoch": 0.44, "learning_rate": 1.2365182339370254e-05, "loss": 1.9136, "step": 17903 }, { "epoch": 0.44, "learning_rate": 1.2364406651715243e-05, "loss": 1.9715, "step": 17904 }, { "epoch": 0.44, "learning_rate": 1.2363630948991101e-05, "loss": 1.9594, "step": 17905 }, { "epoch": 0.44, "learning_rate": 1.2362855231202768e-05, "loss": 1.9566, "step": 17906 }, { "epoch": 0.44, "learning_rate": 1.2362079498355191e-05, "loss": 2.2163, "step": 17907 }, { "epoch": 0.44, "learning_rate": 1.2361303750453307e-05, "loss": 2.1957, "step": 17908 }, { "epoch": 0.44, "learning_rate": 1.2360527987502064e-05, "loss": 2.0483, "step": 17909 }, { "epoch": 0.44, "learning_rate": 1.235975220950641e-05, "loss": 1.9869, "step": 17910 }, { "epoch": 0.44, "learning_rate": 1.2358976416471286e-05, "loss": 2.0354, "step": 17911 }, { "epoch": 0.44, "learning_rate": 1.2358200608401636e-05, "loss": 2.1097, "step": 17912 }, { "epoch": 0.44, "learning_rate": 1.2357424785302403e-05, "loss": 2.1992, "step": 17913 }, { "epoch": 0.44, "learning_rate": 1.2356648947178536e-05, "loss": 2.0015, "step": 17914 }, { "epoch": 0.44, "learning_rate": 1.2355873094034978e-05, "loss": 1.9191, "step": 17915 }, { "epoch": 0.44, "learning_rate": 1.2355097225876672e-05, "loss": 2.0907, "step": 17916 }, { "epoch": 0.44, "learning_rate": 1.2354321342708562e-05, "loss": 2.0957, "step": 17917 }, { "epoch": 0.44, "learning_rate": 1.2353545444535597e-05, "loss": 2.1134, "step": 17918 }, { "epoch": 0.44, "learning_rate": 1.2352769531362718e-05, "loss": 2.0212, "step": 17919 }, { "epoch": 0.44, "learning_rate": 1.2351993603194875e-05, "loss": 2.2776, "step": 17920 }, { "epoch": 0.44, "learning_rate": 1.2351217660037007e-05, "loss": 2.1443, "step": 17921 }, { "epoch": 0.44, "learning_rate": 1.2350441701894062e-05, "loss": 1.9746, "step": 17922 }, { "epoch": 0.44, "learning_rate": 1.234966572877099e-05, "loss": 2.0553, "step": 17923 }, { "epoch": 0.44, "learning_rate": 1.2348889740672732e-05, "loss": 1.9209, "step": 17924 }, { "epoch": 0.44, "learning_rate": 1.2348113737604232e-05, "loss": 2.1215, "step": 17925 }, { "epoch": 0.44, "learning_rate": 1.234733771957044e-05, "loss": 1.9179, "step": 17926 }, { "epoch": 0.44, "learning_rate": 1.2346561686576297e-05, "loss": 2.1291, "step": 17927 }, { "epoch": 0.44, "learning_rate": 1.2345785638626752e-05, "loss": 2.0358, "step": 17928 }, { "epoch": 0.44, "learning_rate": 1.2345009575726753e-05, "loss": 2.0275, "step": 17929 }, { "epoch": 0.44, "learning_rate": 1.2344233497881243e-05, "loss": 2.0525, "step": 17930 }, { "epoch": 0.44, "learning_rate": 1.234345740509517e-05, "loss": 1.9898, "step": 17931 }, { "epoch": 0.44, "learning_rate": 1.2342681297373475e-05, "loss": 2.1609, "step": 17932 }, { "epoch": 0.44, "learning_rate": 1.2341905174721111e-05, "loss": 2.1603, "step": 17933 }, { "epoch": 0.44, "learning_rate": 1.2341129037143025e-05, "loss": 1.9596, "step": 17934 }, { "epoch": 0.44, "learning_rate": 1.234035288464416e-05, "loss": 2.1293, "step": 17935 }, { "epoch": 0.44, "learning_rate": 1.233957671722946e-05, "loss": 1.984, "step": 17936 }, { "epoch": 0.44, "learning_rate": 1.2338800534903876e-05, "loss": 1.9753, "step": 17937 }, { "epoch": 0.44, "learning_rate": 1.2338024337672356e-05, "loss": 2.0517, "step": 17938 }, { "epoch": 0.44, "learning_rate": 1.2337248125539843e-05, "loss": 2.1504, "step": 17939 }, { "epoch": 0.44, "learning_rate": 1.2336471898511287e-05, "loss": 2.0454, "step": 17940 }, { "epoch": 0.44, "learning_rate": 1.2335695656591634e-05, "loss": 2.0595, "step": 17941 }, { "epoch": 0.44, "learning_rate": 1.2334919399785829e-05, "loss": 1.9404, "step": 17942 }, { "epoch": 0.44, "learning_rate": 1.2334143128098826e-05, "loss": 2.2079, "step": 17943 }, { "epoch": 0.44, "learning_rate": 1.2333366841535566e-05, "loss": 2.0779, "step": 17944 }, { "epoch": 0.44, "learning_rate": 1.2332590540101e-05, "loss": 1.895, "step": 17945 }, { "epoch": 0.44, "learning_rate": 1.2331814223800072e-05, "loss": 2.1114, "step": 17946 }, { "epoch": 0.44, "learning_rate": 1.2331037892637734e-05, "loss": 2.0554, "step": 17947 }, { "epoch": 0.44, "learning_rate": 1.2330261546618929e-05, "loss": 2.1754, "step": 17948 }, { "epoch": 0.44, "learning_rate": 1.2329485185748611e-05, "loss": 1.9916, "step": 17949 }, { "epoch": 0.44, "learning_rate": 1.2328708810031724e-05, "loss": 2.0902, "step": 17950 }, { "epoch": 0.44, "learning_rate": 1.2327932419473216e-05, "loss": 2.2133, "step": 17951 }, { "epoch": 0.44, "learning_rate": 1.2327156014078037e-05, "loss": 2.0985, "step": 17952 }, { "epoch": 0.44, "learning_rate": 1.2326379593851134e-05, "loss": 1.9315, "step": 17953 }, { "epoch": 0.44, "learning_rate": 1.2325603158797457e-05, "loss": 2.0276, "step": 17954 }, { "epoch": 0.44, "learning_rate": 1.2324826708921954e-05, "loss": 1.9096, "step": 17955 }, { "epoch": 0.44, "learning_rate": 1.2324050244229566e-05, "loss": 2.1623, "step": 17956 }, { "epoch": 0.44, "learning_rate": 1.2323273764725252e-05, "loss": 2.1998, "step": 17957 }, { "epoch": 0.44, "learning_rate": 1.232249727041396e-05, "loss": 2.0636, "step": 17958 }, { "epoch": 0.44, "learning_rate": 1.2321720761300634e-05, "loss": 1.8463, "step": 17959 }, { "epoch": 0.44, "learning_rate": 1.2320944237390226e-05, "loss": 2.0464, "step": 17960 }, { "epoch": 0.44, "learning_rate": 1.2320167698687682e-05, "loss": 2.2536, "step": 17961 }, { "epoch": 0.44, "learning_rate": 1.2319391145197954e-05, "loss": 2.0517, "step": 17962 }, { "epoch": 0.44, "learning_rate": 1.2318614576925991e-05, "loss": 2.0677, "step": 17963 }, { "epoch": 0.44, "learning_rate": 1.2317837993876743e-05, "loss": 2.1798, "step": 17964 }, { "epoch": 0.44, "learning_rate": 1.2317061396055155e-05, "loss": 2.1518, "step": 17965 }, { "epoch": 0.44, "learning_rate": 1.231628478346618e-05, "loss": 2.0794, "step": 17966 }, { "epoch": 0.44, "learning_rate": 1.2315508156114767e-05, "loss": 2.0605, "step": 17967 }, { "epoch": 0.44, "learning_rate": 1.2314731514005866e-05, "loss": 2.2276, "step": 17968 }, { "epoch": 0.44, "learning_rate": 1.231395485714443e-05, "loss": 2.1095, "step": 17969 }, { "epoch": 0.44, "learning_rate": 1.2313178185535404e-05, "loss": 2.0543, "step": 17970 }, { "epoch": 0.44, "learning_rate": 1.2312401499183736e-05, "loss": 1.8859, "step": 17971 }, { "epoch": 0.44, "learning_rate": 1.2311624798094385e-05, "loss": 1.9806, "step": 17972 }, { "epoch": 0.44, "learning_rate": 1.2310848082272294e-05, "loss": 2.0254, "step": 17973 }, { "epoch": 0.44, "learning_rate": 1.2310071351722412e-05, "loss": 2.0752, "step": 17974 }, { "epoch": 0.44, "learning_rate": 1.2309294606449698e-05, "loss": 2.141, "step": 17975 }, { "epoch": 0.44, "learning_rate": 1.2308517846459092e-05, "loss": 2.0688, "step": 17976 }, { "epoch": 0.44, "learning_rate": 1.230774107175555e-05, "loss": 2.2047, "step": 17977 }, { "epoch": 0.44, "learning_rate": 1.2306964282344026e-05, "loss": 1.989, "step": 17978 }, { "epoch": 0.44, "learning_rate": 1.2306187478229463e-05, "loss": 1.9687, "step": 17979 }, { "epoch": 0.44, "learning_rate": 1.2305410659416816e-05, "loss": 1.9891, "step": 17980 }, { "epoch": 0.44, "learning_rate": 1.2304633825911037e-05, "loss": 2.0251, "step": 17981 }, { "epoch": 0.44, "learning_rate": 1.2303856977717077e-05, "loss": 2.0443, "step": 17982 }, { "epoch": 0.44, "learning_rate": 1.2303080114839881e-05, "loss": 2.0693, "step": 17983 }, { "epoch": 0.44, "learning_rate": 1.2302303237284407e-05, "loss": 2.013, "step": 17984 }, { "epoch": 0.44, "learning_rate": 1.2301526345055605e-05, "loss": 2.0623, "step": 17985 }, { "epoch": 0.44, "learning_rate": 1.2300749438158425e-05, "loss": 1.9577, "step": 17986 }, { "epoch": 0.44, "learning_rate": 1.229997251659782e-05, "loss": 1.9765, "step": 17987 }, { "epoch": 0.44, "learning_rate": 1.229919558037874e-05, "loss": 1.9683, "step": 17988 }, { "epoch": 0.44, "learning_rate": 1.229841862950614e-05, "loss": 1.8671, "step": 17989 }, { "epoch": 0.44, "learning_rate": 1.2297641663984963e-05, "loss": 2.0676, "step": 17990 }, { "epoch": 0.44, "learning_rate": 1.2296864683820172e-05, "loss": 1.8568, "step": 17991 }, { "epoch": 0.44, "learning_rate": 1.229608768901671e-05, "loss": 1.9804, "step": 17992 }, { "epoch": 0.44, "learning_rate": 1.2295310679579536e-05, "loss": 2.0939, "step": 17993 }, { "epoch": 0.44, "learning_rate": 1.2294533655513598e-05, "loss": 2.1658, "step": 17994 }, { "epoch": 0.44, "learning_rate": 1.2293756616823846e-05, "loss": 2.1116, "step": 17995 }, { "epoch": 0.44, "learning_rate": 1.2292979563515241e-05, "loss": 2.002, "step": 17996 }, { "epoch": 0.44, "learning_rate": 1.229220249559273e-05, "loss": 2.1136, "step": 17997 }, { "epoch": 0.44, "learning_rate": 1.2291425413061262e-05, "loss": 2.086, "step": 17998 }, { "epoch": 0.44, "learning_rate": 1.2290648315925793e-05, "loss": 1.9991, "step": 17999 }, { "epoch": 0.44, "learning_rate": 1.2289871204191278e-05, "loss": 2.18, "step": 18000 }, { "epoch": 0.44, "learning_rate": 1.2289094077862664e-05, "loss": 2.0003, "step": 18001 }, { "epoch": 0.44, "learning_rate": 1.2288316936944911e-05, "loss": 2.1485, "step": 18002 }, { "epoch": 0.44, "learning_rate": 1.2287539781442967e-05, "loss": 2.2452, "step": 18003 }, { "epoch": 0.44, "learning_rate": 1.228676261136179e-05, "loss": 2.0001, "step": 18004 }, { "epoch": 0.44, "learning_rate": 1.2285985426706323e-05, "loss": 2.0812, "step": 18005 }, { "epoch": 0.44, "learning_rate": 1.228520822748153e-05, "loss": 1.8981, "step": 18006 }, { "epoch": 0.44, "learning_rate": 1.2284431013692362e-05, "loss": 2.0912, "step": 18007 }, { "epoch": 0.44, "learning_rate": 1.2283653785343765e-05, "loss": 2.0954, "step": 18008 }, { "epoch": 0.44, "learning_rate": 1.2282876542440704e-05, "loss": 1.9871, "step": 18009 }, { "epoch": 0.44, "learning_rate": 1.2282099284988122e-05, "loss": 1.9702, "step": 18010 }, { "epoch": 0.44, "learning_rate": 1.2281322012990982e-05, "loss": 2.0409, "step": 18011 }, { "epoch": 0.44, "learning_rate": 1.2280544726454232e-05, "loss": 2.2119, "step": 18012 }, { "epoch": 0.44, "learning_rate": 1.2279767425382826e-05, "loss": 1.9845, "step": 18013 }, { "epoch": 0.44, "learning_rate": 1.227899010978172e-05, "loss": 2.0479, "step": 18014 }, { "epoch": 0.44, "learning_rate": 1.2278212779655867e-05, "loss": 2.1924, "step": 18015 }, { "epoch": 0.44, "learning_rate": 1.227743543501022e-05, "loss": 2.1164, "step": 18016 }, { "epoch": 0.44, "learning_rate": 1.2276658075849736e-05, "loss": 2.0949, "step": 18017 }, { "epoch": 0.44, "learning_rate": 1.227588070217937e-05, "loss": 2.274, "step": 18018 }, { "epoch": 0.44, "learning_rate": 1.227510331400407e-05, "loss": 1.9721, "step": 18019 }, { "epoch": 0.44, "learning_rate": 1.22743259113288e-05, "loss": 1.9541, "step": 18020 }, { "epoch": 0.44, "learning_rate": 1.227354849415851e-05, "loss": 1.9789, "step": 18021 }, { "epoch": 0.44, "learning_rate": 1.2272771062498153e-05, "loss": 2.0515, "step": 18022 }, { "epoch": 0.44, "learning_rate": 1.2271993616352683e-05, "loss": 2.0405, "step": 18023 }, { "epoch": 0.44, "learning_rate": 1.2271216155727059e-05, "loss": 2.2424, "step": 18024 }, { "epoch": 0.44, "learning_rate": 1.2270438680626236e-05, "loss": 2.0384, "step": 18025 }, { "epoch": 0.44, "learning_rate": 1.2269661191055165e-05, "loss": 2.1427, "step": 18026 }, { "epoch": 0.44, "learning_rate": 1.2268883687018805e-05, "loss": 1.9352, "step": 18027 }, { "epoch": 0.44, "learning_rate": 1.226810616852211e-05, "loss": 2.0997, "step": 18028 }, { "epoch": 0.44, "learning_rate": 1.2267328635570032e-05, "loss": 2.1193, "step": 18029 }, { "epoch": 0.44, "learning_rate": 1.2266551088167534e-05, "loss": 2.12, "step": 18030 }, { "epoch": 0.44, "learning_rate": 1.2265773526319565e-05, "loss": 2.2415, "step": 18031 }, { "epoch": 0.44, "learning_rate": 1.2264995950031085e-05, "loss": 1.9646, "step": 18032 }, { "epoch": 0.44, "learning_rate": 1.2264218359307046e-05, "loss": 2.099, "step": 18033 }, { "epoch": 0.44, "learning_rate": 1.2263440754152405e-05, "loss": 1.9634, "step": 18034 }, { "epoch": 0.44, "learning_rate": 1.226266313457212e-05, "loss": 2.0961, "step": 18035 }, { "epoch": 0.44, "learning_rate": 1.2261885500571143e-05, "loss": 2.1245, "step": 18036 }, { "epoch": 0.44, "learning_rate": 1.2261107852154435e-05, "loss": 2.1575, "step": 18037 }, { "epoch": 0.44, "learning_rate": 1.2260330189326945e-05, "loss": 2.0372, "step": 18038 }, { "epoch": 0.44, "learning_rate": 1.225955251209364e-05, "loss": 2.0102, "step": 18039 }, { "epoch": 0.44, "learning_rate": 1.225877482045947e-05, "loss": 1.9486, "step": 18040 }, { "epoch": 0.44, "learning_rate": 1.225799711442939e-05, "loss": 2.1462, "step": 18041 }, { "epoch": 0.44, "learning_rate": 1.225721939400836e-05, "loss": 2.0139, "step": 18042 }, { "epoch": 0.44, "learning_rate": 1.2256441659201332e-05, "loss": 1.9741, "step": 18043 }, { "epoch": 0.44, "learning_rate": 1.225566391001327e-05, "loss": 2.0834, "step": 18044 }, { "epoch": 0.44, "learning_rate": 1.2254886146449124e-05, "loss": 2.1107, "step": 18045 }, { "epoch": 0.44, "learning_rate": 1.2254108368513853e-05, "loss": 2.0246, "step": 18046 }, { "epoch": 0.44, "learning_rate": 1.2253330576212416e-05, "loss": 1.8134, "step": 18047 }, { "epoch": 0.44, "learning_rate": 1.2252552769549769e-05, "loss": 2.0323, "step": 18048 }, { "epoch": 0.44, "learning_rate": 1.225177494853087e-05, "loss": 2.1449, "step": 18049 }, { "epoch": 0.44, "learning_rate": 1.2250997113160676e-05, "loss": 2.0819, "step": 18050 }, { "epoch": 0.44, "learning_rate": 1.2250219263444144e-05, "loss": 2.0834, "step": 18051 }, { "epoch": 0.44, "learning_rate": 1.2249441399386228e-05, "loss": 2.055, "step": 18052 }, { "epoch": 0.44, "learning_rate": 1.224866352099189e-05, "loss": 1.9948, "step": 18053 }, { "epoch": 0.45, "learning_rate": 1.2247885628266088e-05, "loss": 2.0705, "step": 18054 }, { "epoch": 0.45, "learning_rate": 1.2247107721213776e-05, "loss": 1.9712, "step": 18055 }, { "epoch": 0.45, "learning_rate": 1.2246329799839918e-05, "loss": 2.1514, "step": 18056 }, { "epoch": 0.45, "learning_rate": 1.2245551864149465e-05, "loss": 2.0469, "step": 18057 }, { "epoch": 0.45, "learning_rate": 1.2244773914147379e-05, "loss": 1.9985, "step": 18058 }, { "epoch": 0.45, "learning_rate": 1.2243995949838618e-05, "loss": 2.0931, "step": 18059 }, { "epoch": 0.45, "learning_rate": 1.2243217971228138e-05, "loss": 2.0202, "step": 18060 }, { "epoch": 0.45, "learning_rate": 1.2242439978320899e-05, "loss": 1.8965, "step": 18061 }, { "epoch": 0.45, "learning_rate": 1.2241661971121861e-05, "loss": 1.8185, "step": 18062 }, { "epoch": 0.45, "learning_rate": 1.224088394963598e-05, "loss": 2.2169, "step": 18063 }, { "epoch": 0.45, "learning_rate": 1.2240105913868214e-05, "loss": 2.2256, "step": 18064 }, { "epoch": 0.45, "learning_rate": 1.2239327863823524e-05, "loss": 1.9332, "step": 18065 }, { "epoch": 0.45, "learning_rate": 1.223854979950687e-05, "loss": 2.13, "step": 18066 }, { "epoch": 0.45, "learning_rate": 1.2237771720923203e-05, "loss": 2.1053, "step": 18067 }, { "epoch": 0.45, "learning_rate": 1.2236993628077494e-05, "loss": 2.0467, "step": 18068 }, { "epoch": 0.45, "learning_rate": 1.2236215520974693e-05, "loss": 2.0763, "step": 18069 }, { "epoch": 0.45, "learning_rate": 1.223543739961976e-05, "loss": 2.1157, "step": 18070 }, { "epoch": 0.45, "learning_rate": 1.2234659264017657e-05, "loss": 1.9366, "step": 18071 }, { "epoch": 0.45, "learning_rate": 1.223388111417334e-05, "loss": 2.0012, "step": 18072 }, { "epoch": 0.45, "learning_rate": 1.2233102950091775e-05, "loss": 2.0462, "step": 18073 }, { "epoch": 0.45, "learning_rate": 1.2232324771777914e-05, "loss": 2.099, "step": 18074 }, { "epoch": 0.45, "learning_rate": 1.2231546579236724e-05, "loss": 1.982, "step": 18075 }, { "epoch": 0.45, "learning_rate": 1.223076837247316e-05, "loss": 1.9623, "step": 18076 }, { "epoch": 0.45, "learning_rate": 1.2229990151492175e-05, "loss": 2.225, "step": 18077 }, { "epoch": 0.45, "learning_rate": 1.2229211916298743e-05, "loss": 2.0857, "step": 18078 }, { "epoch": 0.45, "learning_rate": 1.2228433666897811e-05, "loss": 2.0789, "step": 18079 }, { "epoch": 0.45, "learning_rate": 1.222765540329435e-05, "loss": 2.0694, "step": 18080 }, { "epoch": 0.45, "learning_rate": 1.2226877125493315e-05, "loss": 1.9242, "step": 18081 }, { "epoch": 0.45, "learning_rate": 1.2226098833499664e-05, "loss": 1.9621, "step": 18082 }, { "epoch": 0.45, "learning_rate": 1.222532052731836e-05, "loss": 2.1155, "step": 18083 }, { "epoch": 0.45, "learning_rate": 1.2224542206954367e-05, "loss": 2.0167, "step": 18084 }, { "epoch": 0.45, "learning_rate": 1.2223763872412638e-05, "loss": 1.993, "step": 18085 }, { "epoch": 0.45, "learning_rate": 1.2222985523698139e-05, "loss": 2.0957, "step": 18086 }, { "epoch": 0.45, "learning_rate": 1.2222207160815824e-05, "loss": 1.9982, "step": 18087 }, { "epoch": 0.45, "learning_rate": 1.2221428783770661e-05, "loss": 1.86, "step": 18088 }, { "epoch": 0.45, "learning_rate": 1.222065039256761e-05, "loss": 2.1382, "step": 18089 }, { "epoch": 0.45, "learning_rate": 1.2219871987211632e-05, "loss": 2.0108, "step": 18090 }, { "epoch": 0.45, "learning_rate": 1.2219093567707682e-05, "loss": 1.9889, "step": 18091 }, { "epoch": 0.45, "learning_rate": 1.2218315134060727e-05, "loss": 2.0023, "step": 18092 }, { "epoch": 0.45, "learning_rate": 1.2217536686275728e-05, "loss": 2.0233, "step": 18093 }, { "epoch": 0.45, "learning_rate": 1.2216758224357645e-05, "loss": 2.0398, "step": 18094 }, { "epoch": 0.45, "learning_rate": 1.2215979748311439e-05, "loss": 2.0309, "step": 18095 }, { "epoch": 0.45, "learning_rate": 1.2215201258142072e-05, "loss": 1.9226, "step": 18096 }, { "epoch": 0.45, "learning_rate": 1.2214422753854503e-05, "loss": 2.0867, "step": 18097 }, { "epoch": 0.45, "learning_rate": 1.2213644235453699e-05, "loss": 2.0163, "step": 18098 }, { "epoch": 0.45, "learning_rate": 1.2212865702944619e-05, "loss": 2.083, "step": 18099 }, { "epoch": 0.45, "learning_rate": 1.2212087156332225e-05, "loss": 2.1178, "step": 18100 }, { "epoch": 0.45, "learning_rate": 1.2211308595621474e-05, "loss": 1.9787, "step": 18101 }, { "epoch": 0.45, "learning_rate": 1.2210530020817337e-05, "loss": 2.0112, "step": 18102 }, { "epoch": 0.45, "learning_rate": 1.2209751431924772e-05, "loss": 1.9448, "step": 18103 }, { "epoch": 0.45, "learning_rate": 1.2208972828948743e-05, "loss": 2.1268, "step": 18104 }, { "epoch": 0.45, "learning_rate": 1.2208194211894204e-05, "loss": 2.0221, "step": 18105 }, { "epoch": 0.45, "learning_rate": 1.2207415580766125e-05, "loss": 2.2688, "step": 18106 }, { "epoch": 0.45, "learning_rate": 1.2206636935569469e-05, "loss": 2.087, "step": 18107 }, { "epoch": 0.45, "learning_rate": 1.2205858276309198e-05, "loss": 1.9871, "step": 18108 }, { "epoch": 0.45, "learning_rate": 1.2205079602990274e-05, "loss": 2.0644, "step": 18109 }, { "epoch": 0.45, "learning_rate": 1.2204300915617657e-05, "loss": 1.9896, "step": 18110 }, { "epoch": 0.45, "learning_rate": 1.220352221419631e-05, "loss": 1.906, "step": 18111 }, { "epoch": 0.45, "learning_rate": 1.22027434987312e-05, "loss": 2.0958, "step": 18112 }, { "epoch": 0.45, "learning_rate": 1.220196476922729e-05, "loss": 2.0336, "step": 18113 }, { "epoch": 0.45, "learning_rate": 1.2201186025689535e-05, "loss": 2.3162, "step": 18114 }, { "epoch": 0.45, "learning_rate": 1.220040726812291e-05, "loss": 1.9936, "step": 18115 }, { "epoch": 0.45, "learning_rate": 1.2199628496532368e-05, "loss": 1.8982, "step": 18116 }, { "epoch": 0.45, "learning_rate": 1.219884971092288e-05, "loss": 2.2163, "step": 18117 }, { "epoch": 0.45, "learning_rate": 1.2198070911299404e-05, "loss": 2.1471, "step": 18118 }, { "epoch": 0.45, "learning_rate": 1.2197292097666908e-05, "loss": 1.8882, "step": 18119 }, { "epoch": 0.45, "learning_rate": 1.2196513270030354e-05, "loss": 2.0574, "step": 18120 }, { "epoch": 0.45, "learning_rate": 1.2195734428394701e-05, "loss": 2.0516, "step": 18121 }, { "epoch": 0.45, "learning_rate": 1.2194955572764922e-05, "loss": 2.055, "step": 18122 }, { "epoch": 0.45, "learning_rate": 1.219417670314597e-05, "loss": 2.1503, "step": 18123 }, { "epoch": 0.45, "learning_rate": 1.2193397819542818e-05, "loss": 2.0539, "step": 18124 }, { "epoch": 0.45, "learning_rate": 1.2192618921960426e-05, "loss": 1.8564, "step": 18125 }, { "epoch": 0.45, "learning_rate": 1.2191840010403762e-05, "loss": 1.9717, "step": 18126 }, { "epoch": 0.45, "learning_rate": 1.2191061084877784e-05, "loss": 1.9506, "step": 18127 }, { "epoch": 0.45, "learning_rate": 1.2190282145387462e-05, "loss": 2.2411, "step": 18128 }, { "epoch": 0.45, "learning_rate": 1.2189503191937756e-05, "loss": 2.0352, "step": 18129 }, { "epoch": 0.45, "learning_rate": 1.2188724224533633e-05, "loss": 2.166, "step": 18130 }, { "epoch": 0.45, "learning_rate": 1.218794524318006e-05, "loss": 2.1445, "step": 18131 }, { "epoch": 0.45, "learning_rate": 1.2187166247881994e-05, "loss": 1.7671, "step": 18132 }, { "epoch": 0.45, "learning_rate": 1.2186387238644408e-05, "loss": 1.9986, "step": 18133 }, { "epoch": 0.45, "learning_rate": 1.2185608215472264e-05, "loss": 2.1261, "step": 18134 }, { "epoch": 0.45, "learning_rate": 1.2184829178370522e-05, "loss": 2.1153, "step": 18135 }, { "epoch": 0.45, "learning_rate": 1.2184050127344159e-05, "loss": 2.039, "step": 18136 }, { "epoch": 0.45, "learning_rate": 1.2183271062398127e-05, "loss": 2.0757, "step": 18137 }, { "epoch": 0.45, "learning_rate": 1.2182491983537402e-05, "loss": 2.1368, "step": 18138 }, { "epoch": 0.45, "learning_rate": 1.218171289076694e-05, "loss": 2.1244, "step": 18139 }, { "epoch": 0.45, "learning_rate": 1.218093378409171e-05, "loss": 2.0231, "step": 18140 }, { "epoch": 0.45, "learning_rate": 1.2180154663516681e-05, "loss": 2.0257, "step": 18141 }, { "epoch": 0.45, "learning_rate": 1.2179375529046815e-05, "loss": 1.919, "step": 18142 }, { "epoch": 0.45, "learning_rate": 1.217859638068708e-05, "loss": 2.0517, "step": 18143 }, { "epoch": 0.45, "learning_rate": 1.2177817218442438e-05, "loss": 2.0862, "step": 18144 }, { "epoch": 0.45, "learning_rate": 1.2177038042317857e-05, "loss": 2.0227, "step": 18145 }, { "epoch": 0.45, "learning_rate": 1.2176258852318305e-05, "loss": 2.1486, "step": 18146 }, { "epoch": 0.45, "learning_rate": 1.2175479648448747e-05, "loss": 2.1857, "step": 18147 }, { "epoch": 0.45, "learning_rate": 1.2174700430714145e-05, "loss": 2.1209, "step": 18148 }, { "epoch": 0.45, "learning_rate": 1.217392119911947e-05, "loss": 1.8503, "step": 18149 }, { "epoch": 0.45, "learning_rate": 1.2173141953669685e-05, "loss": 2.1436, "step": 18150 }, { "epoch": 0.45, "learning_rate": 1.2172362694369761e-05, "loss": 2.002, "step": 18151 }, { "epoch": 0.45, "learning_rate": 1.2171583421224659e-05, "loss": 2.1916, "step": 18152 }, { "epoch": 0.45, "learning_rate": 1.2170804134239352e-05, "loss": 2.0453, "step": 18153 }, { "epoch": 0.45, "learning_rate": 1.21700248334188e-05, "loss": 2.0935, "step": 18154 }, { "epoch": 0.45, "learning_rate": 1.2169245518767973e-05, "loss": 2.0323, "step": 18155 }, { "epoch": 0.45, "learning_rate": 1.2168466190291836e-05, "loss": 2.1274, "step": 18156 }, { "epoch": 0.45, "learning_rate": 1.2167686847995362e-05, "loss": 2.1735, "step": 18157 }, { "epoch": 0.45, "learning_rate": 1.2166907491883508e-05, "loss": 2.0692, "step": 18158 }, { "epoch": 0.45, "learning_rate": 1.2166128121961247e-05, "loss": 2.0316, "step": 18159 }, { "epoch": 0.45, "learning_rate": 1.2165348738233547e-05, "loss": 1.9853, "step": 18160 }, { "epoch": 0.45, "learning_rate": 1.2164569340705375e-05, "loss": 2.1339, "step": 18161 }, { "epoch": 0.45, "learning_rate": 1.2163789929381696e-05, "loss": 2.1243, "step": 18162 }, { "epoch": 0.45, "learning_rate": 1.2163010504267478e-05, "loss": 2.2267, "step": 18163 }, { "epoch": 0.45, "learning_rate": 1.2162231065367689e-05, "loss": 1.9778, "step": 18164 }, { "epoch": 0.45, "learning_rate": 1.2161451612687298e-05, "loss": 1.8607, "step": 18165 }, { "epoch": 0.45, "learning_rate": 1.2160672146231272e-05, "loss": 2.0082, "step": 18166 }, { "epoch": 0.45, "learning_rate": 1.2159892666004573e-05, "loss": 1.972, "step": 18167 }, { "epoch": 0.45, "learning_rate": 1.215911317201218e-05, "loss": 2.0328, "step": 18168 }, { "epoch": 0.45, "learning_rate": 1.2158333664259052e-05, "loss": 2.0156, "step": 18169 }, { "epoch": 0.45, "learning_rate": 1.2157554142750164e-05, "loss": 2.0562, "step": 18170 }, { "epoch": 0.45, "learning_rate": 1.2156774607490479e-05, "loss": 2.0909, "step": 18171 }, { "epoch": 0.45, "learning_rate": 1.2155995058484963e-05, "loss": 2.0145, "step": 18172 }, { "epoch": 0.45, "learning_rate": 1.2155215495738593e-05, "loss": 2.1249, "step": 18173 }, { "epoch": 0.45, "learning_rate": 1.2154435919256327e-05, "loss": 2.0437, "step": 18174 }, { "epoch": 0.45, "learning_rate": 1.2153656329043142e-05, "loss": 2.1174, "step": 18175 }, { "epoch": 0.45, "learning_rate": 1.2152876725104e-05, "loss": 1.9459, "step": 18176 }, { "epoch": 0.45, "learning_rate": 1.2152097107443878e-05, "loss": 2.1254, "step": 18177 }, { "epoch": 0.45, "learning_rate": 1.215131747606774e-05, "loss": 1.9422, "step": 18178 }, { "epoch": 0.45, "learning_rate": 1.215053783098055e-05, "loss": 1.9566, "step": 18179 }, { "epoch": 0.45, "learning_rate": 1.2149758172187282e-05, "loss": 2.1791, "step": 18180 }, { "epoch": 0.45, "learning_rate": 1.2148978499692908e-05, "loss": 2.0578, "step": 18181 }, { "epoch": 0.45, "learning_rate": 1.2148198813502392e-05, "loss": 2.0462, "step": 18182 }, { "epoch": 0.45, "learning_rate": 1.2147419113620705e-05, "loss": 2.0646, "step": 18183 }, { "epoch": 0.45, "learning_rate": 1.2146639400052816e-05, "loss": 1.989, "step": 18184 }, { "epoch": 0.45, "learning_rate": 1.2145859672803692e-05, "loss": 2.3049, "step": 18185 }, { "epoch": 0.45, "learning_rate": 1.2145079931878309e-05, "loss": 1.9738, "step": 18186 }, { "epoch": 0.45, "learning_rate": 1.2144300177281631e-05, "loss": 2.0946, "step": 18187 }, { "epoch": 0.45, "learning_rate": 1.2143520409018627e-05, "loss": 1.9547, "step": 18188 }, { "epoch": 0.45, "learning_rate": 1.2142740627094274e-05, "loss": 2.1176, "step": 18189 }, { "epoch": 0.45, "learning_rate": 1.2141960831513532e-05, "loss": 2.0587, "step": 18190 }, { "epoch": 0.45, "learning_rate": 1.214118102228138e-05, "loss": 2.2295, "step": 18191 }, { "epoch": 0.45, "learning_rate": 1.2140401199402783e-05, "loss": 1.9628, "step": 18192 }, { "epoch": 0.45, "learning_rate": 1.2139621362882708e-05, "loss": 2.0312, "step": 18193 }, { "epoch": 0.45, "learning_rate": 1.2138841512726128e-05, "loss": 2.1992, "step": 18194 }, { "epoch": 0.45, "learning_rate": 1.2138061648938018e-05, "loss": 2.0219, "step": 18195 }, { "epoch": 0.45, "learning_rate": 1.2137281771523341e-05, "loss": 2.0547, "step": 18196 }, { "epoch": 0.45, "learning_rate": 1.2136501880487075e-05, "loss": 1.9267, "step": 18197 }, { "epoch": 0.45, "learning_rate": 1.2135721975834182e-05, "loss": 1.7725, "step": 18198 }, { "epoch": 0.45, "learning_rate": 1.2134942057569639e-05, "loss": 2.0096, "step": 18199 }, { "epoch": 0.45, "learning_rate": 1.2134162125698415e-05, "loss": 2.0122, "step": 18200 }, { "epoch": 0.45, "learning_rate": 1.213338218022548e-05, "loss": 2.1811, "step": 18201 }, { "epoch": 0.45, "learning_rate": 1.2132602221155803e-05, "loss": 2.198, "step": 18202 }, { "epoch": 0.45, "learning_rate": 1.2131822248494355e-05, "loss": 2.0911, "step": 18203 }, { "epoch": 0.45, "learning_rate": 1.2131042262246116e-05, "loss": 2.0989, "step": 18204 }, { "epoch": 0.45, "learning_rate": 1.2130262262416046e-05, "loss": 2.175, "step": 18205 }, { "epoch": 0.45, "learning_rate": 1.2129482249009122e-05, "loss": 2.0787, "step": 18206 }, { "epoch": 0.45, "learning_rate": 1.2128702222030312e-05, "loss": 2.2458, "step": 18207 }, { "epoch": 0.45, "learning_rate": 1.2127922181484588e-05, "loss": 2.0157, "step": 18208 }, { "epoch": 0.45, "learning_rate": 1.2127142127376926e-05, "loss": 1.993, "step": 18209 }, { "epoch": 0.45, "learning_rate": 1.2126362059712293e-05, "loss": 2.0225, "step": 18210 }, { "epoch": 0.45, "learning_rate": 1.212558197849566e-05, "loss": 1.9891, "step": 18211 }, { "epoch": 0.45, "learning_rate": 1.2124801883732e-05, "loss": 2.0983, "step": 18212 }, { "epoch": 0.45, "learning_rate": 1.2124021775426286e-05, "loss": 1.9368, "step": 18213 }, { "epoch": 0.45, "learning_rate": 1.212324165358349e-05, "loss": 2.208, "step": 18214 }, { "epoch": 0.45, "learning_rate": 1.2122461518208583e-05, "loss": 2.014, "step": 18215 }, { "epoch": 0.45, "learning_rate": 1.2121681369306537e-05, "loss": 2.1236, "step": 18216 }, { "epoch": 0.45, "learning_rate": 1.2120901206882323e-05, "loss": 2.2875, "step": 18217 }, { "epoch": 0.45, "learning_rate": 1.2120121030940914e-05, "loss": 2.2493, "step": 18218 }, { "epoch": 0.45, "learning_rate": 1.2119340841487286e-05, "loss": 2.0404, "step": 18219 }, { "epoch": 0.45, "learning_rate": 1.2118560638526404e-05, "loss": 2.0419, "step": 18220 }, { "epoch": 0.45, "learning_rate": 1.2117780422063248e-05, "loss": 1.858, "step": 18221 }, { "epoch": 0.45, "learning_rate": 1.2117000192102784e-05, "loss": 2.0286, "step": 18222 }, { "epoch": 0.45, "learning_rate": 1.211621994864999e-05, "loss": 1.9253, "step": 18223 }, { "epoch": 0.45, "learning_rate": 1.2115439691709837e-05, "loss": 2.0834, "step": 18224 }, { "epoch": 0.45, "learning_rate": 1.2114659421287299e-05, "loss": 2.0114, "step": 18225 }, { "epoch": 0.45, "learning_rate": 1.2113879137387343e-05, "loss": 2.0466, "step": 18226 }, { "epoch": 0.45, "learning_rate": 1.2113098840014948e-05, "loss": 2.1527, "step": 18227 }, { "epoch": 0.45, "learning_rate": 1.2112318529175087e-05, "loss": 2.1247, "step": 18228 }, { "epoch": 0.45, "learning_rate": 1.2111538204872728e-05, "loss": 2.2461, "step": 18229 }, { "epoch": 0.45, "learning_rate": 1.2110757867112853e-05, "loss": 2.0011, "step": 18230 }, { "epoch": 0.45, "learning_rate": 1.2109977515900426e-05, "loss": 1.9262, "step": 18231 }, { "epoch": 0.45, "learning_rate": 1.2109197151240426e-05, "loss": 1.9953, "step": 18232 }, { "epoch": 0.45, "learning_rate": 1.2108416773137825e-05, "loss": 2.0342, "step": 18233 }, { "epoch": 0.45, "learning_rate": 1.2107636381597598e-05, "loss": 2.1345, "step": 18234 }, { "epoch": 0.45, "learning_rate": 1.2106855976624719e-05, "loss": 2.0365, "step": 18235 }, { "epoch": 0.45, "learning_rate": 1.2106075558224156e-05, "loss": 2.1064, "step": 18236 }, { "epoch": 0.45, "learning_rate": 1.2105295126400889e-05, "loss": 2.0145, "step": 18237 }, { "epoch": 0.45, "learning_rate": 1.2104514681159889e-05, "loss": 2.0355, "step": 18238 }, { "epoch": 0.45, "learning_rate": 1.2103734222506133e-05, "loss": 2.0913, "step": 18239 }, { "epoch": 0.45, "learning_rate": 1.2102953750444593e-05, "loss": 2.1496, "step": 18240 }, { "epoch": 0.45, "learning_rate": 1.2102173264980244e-05, "loss": 2.0491, "step": 18241 }, { "epoch": 0.45, "learning_rate": 1.2101392766118057e-05, "loss": 2.1137, "step": 18242 }, { "epoch": 0.45, "learning_rate": 1.2100612253863011e-05, "loss": 1.9726, "step": 18243 }, { "epoch": 0.45, "learning_rate": 1.209983172822008e-05, "loss": 2.1544, "step": 18244 }, { "epoch": 0.45, "learning_rate": 1.2099051189194235e-05, "loss": 2.2828, "step": 18245 }, { "epoch": 0.45, "learning_rate": 1.2098270636790452e-05, "loss": 2.1359, "step": 18246 }, { "epoch": 0.45, "learning_rate": 1.2097490071013707e-05, "loss": 1.968, "step": 18247 }, { "epoch": 0.45, "learning_rate": 1.2096709491868975e-05, "loss": 2.1298, "step": 18248 }, { "epoch": 0.45, "learning_rate": 1.2095928899361229e-05, "loss": 2.0201, "step": 18249 }, { "epoch": 0.45, "learning_rate": 1.2095148293495448e-05, "loss": 2.0951, "step": 18250 }, { "epoch": 0.45, "learning_rate": 1.2094367674276599e-05, "loss": 2.0273, "step": 18251 }, { "epoch": 0.45, "learning_rate": 1.2093587041709668e-05, "loss": 2.0946, "step": 18252 }, { "epoch": 0.45, "learning_rate": 1.2092806395799621e-05, "loss": 2.1122, "step": 18253 }, { "epoch": 0.45, "learning_rate": 1.2092025736551439e-05, "loss": 2.0511, "step": 18254 }, { "epoch": 0.45, "learning_rate": 1.2091245063970092e-05, "loss": 2.1572, "step": 18255 }, { "epoch": 0.45, "learning_rate": 1.2090464378060558e-05, "loss": 1.7956, "step": 18256 }, { "epoch": 0.45, "learning_rate": 1.2089683678827817e-05, "loss": 2.1308, "step": 18257 }, { "epoch": 0.45, "learning_rate": 1.208890296627684e-05, "loss": 1.9324, "step": 18258 }, { "epoch": 0.45, "learning_rate": 1.2088122240412603e-05, "loss": 2.0523, "step": 18259 }, { "epoch": 0.45, "learning_rate": 1.2087341501240084e-05, "loss": 2.2829, "step": 18260 }, { "epoch": 0.45, "learning_rate": 1.2086560748764255e-05, "loss": 2.0073, "step": 18261 }, { "epoch": 0.45, "learning_rate": 1.2085779982990095e-05, "loss": 2.0505, "step": 18262 }, { "epoch": 0.45, "learning_rate": 1.2084999203922583e-05, "loss": 2.0933, "step": 18263 }, { "epoch": 0.45, "learning_rate": 1.2084218411566686e-05, "loss": 1.7623, "step": 18264 }, { "epoch": 0.45, "learning_rate": 1.2083437605927389e-05, "loss": 2.17, "step": 18265 }, { "epoch": 0.45, "learning_rate": 1.2082656787009663e-05, "loss": 1.963, "step": 18266 }, { "epoch": 0.45, "learning_rate": 1.208187595481849e-05, "loss": 2.0294, "step": 18267 }, { "epoch": 0.45, "learning_rate": 1.208109510935884e-05, "loss": 1.7389, "step": 18268 }, { "epoch": 0.45, "learning_rate": 1.2080314250635694e-05, "loss": 1.9345, "step": 18269 }, { "epoch": 0.45, "learning_rate": 1.207953337865403e-05, "loss": 2.1362, "step": 18270 }, { "epoch": 0.45, "learning_rate": 1.2078752493418817e-05, "loss": 2.158, "step": 18271 }, { "epoch": 0.45, "learning_rate": 1.207797159493504e-05, "loss": 2.0241, "step": 18272 }, { "epoch": 0.45, "learning_rate": 1.2077190683207669e-05, "loss": 2.0807, "step": 18273 }, { "epoch": 0.45, "learning_rate": 1.207640975824169e-05, "loss": 2.0575, "step": 18274 }, { "epoch": 0.45, "learning_rate": 1.2075628820042071e-05, "loss": 2.1824, "step": 18275 }, { "epoch": 0.45, "learning_rate": 1.2074847868613795e-05, "loss": 1.8683, "step": 18276 }, { "epoch": 0.45, "learning_rate": 1.2074066903961839e-05, "loss": 1.9595, "step": 18277 }, { "epoch": 0.45, "learning_rate": 1.2073285926091176e-05, "loss": 2.2243, "step": 18278 }, { "epoch": 0.45, "learning_rate": 1.2072504935006788e-05, "loss": 2.2248, "step": 18279 }, { "epoch": 0.45, "learning_rate": 1.2071723930713647e-05, "loss": 2.0492, "step": 18280 }, { "epoch": 0.45, "learning_rate": 1.2070942913216739e-05, "loss": 1.9562, "step": 18281 }, { "epoch": 0.45, "learning_rate": 1.2070161882521034e-05, "loss": 1.9534, "step": 18282 }, { "epoch": 0.45, "learning_rate": 1.2069380838631512e-05, "loss": 2.233, "step": 18283 }, { "epoch": 0.45, "learning_rate": 1.2068599781553156e-05, "loss": 1.9929, "step": 18284 }, { "epoch": 0.45, "learning_rate": 1.2067818711290934e-05, "loss": 2.1088, "step": 18285 }, { "epoch": 0.45, "learning_rate": 1.2067037627849834e-05, "loss": 2.0716, "step": 18286 }, { "epoch": 0.45, "learning_rate": 1.2066256531234827e-05, "loss": 2.0585, "step": 18287 }, { "epoch": 0.45, "learning_rate": 1.2065475421450894e-05, "loss": 2.1117, "step": 18288 }, { "epoch": 0.45, "learning_rate": 1.2064694298503013e-05, "loss": 2.1193, "step": 18289 }, { "epoch": 0.45, "learning_rate": 1.2063913162396164e-05, "loss": 2.1328, "step": 18290 }, { "epoch": 0.45, "learning_rate": 1.2063132013135318e-05, "loss": 2.0562, "step": 18291 }, { "epoch": 0.45, "learning_rate": 1.2062350850725467e-05, "loss": 2.0585, "step": 18292 }, { "epoch": 0.45, "learning_rate": 1.2061569675171578e-05, "loss": 2.1962, "step": 18293 }, { "epoch": 0.45, "learning_rate": 1.2060788486478637e-05, "loss": 2.1594, "step": 18294 }, { "epoch": 0.45, "learning_rate": 1.2060007284651614e-05, "loss": 2.0282, "step": 18295 }, { "epoch": 0.45, "learning_rate": 1.2059226069695498e-05, "loss": 2.1124, "step": 18296 }, { "epoch": 0.45, "learning_rate": 1.205844484161526e-05, "loss": 1.9474, "step": 18297 }, { "epoch": 0.45, "learning_rate": 1.2057663600415883e-05, "loss": 1.9034, "step": 18298 }, { "epoch": 0.45, "learning_rate": 1.2056882346102345e-05, "loss": 2.0135, "step": 18299 }, { "epoch": 0.45, "learning_rate": 1.2056101078679626e-05, "loss": 1.8791, "step": 18300 }, { "epoch": 0.45, "learning_rate": 1.2055319798152705e-05, "loss": 1.9965, "step": 18301 }, { "epoch": 0.45, "learning_rate": 1.2054538504526561e-05, "loss": 2.1044, "step": 18302 }, { "epoch": 0.45, "learning_rate": 1.2053757197806175e-05, "loss": 1.8419, "step": 18303 }, { "epoch": 0.45, "learning_rate": 1.2052975877996523e-05, "loss": 2.045, "step": 18304 }, { "epoch": 0.45, "learning_rate": 1.2052194545102586e-05, "loss": 2.0371, "step": 18305 }, { "epoch": 0.45, "learning_rate": 1.2051413199129347e-05, "loss": 2.0384, "step": 18306 }, { "epoch": 0.45, "learning_rate": 1.205063184008178e-05, "loss": 1.9847, "step": 18307 }, { "epoch": 0.45, "learning_rate": 1.2049850467964873e-05, "loss": 2.2843, "step": 18308 }, { "epoch": 0.45, "learning_rate": 1.2049069082783596e-05, "loss": 2.0465, "step": 18309 }, { "epoch": 0.45, "learning_rate": 1.2048287684542935e-05, "loss": 2.0964, "step": 18310 }, { "epoch": 0.45, "learning_rate": 1.2047506273247871e-05, "loss": 2.2548, "step": 18311 }, { "epoch": 0.45, "learning_rate": 1.2046724848903382e-05, "loss": 2.1421, "step": 18312 }, { "epoch": 0.45, "learning_rate": 1.2045943411514447e-05, "loss": 2.1797, "step": 18313 }, { "epoch": 0.45, "learning_rate": 1.2045161961086047e-05, "loss": 2.0425, "step": 18314 }, { "epoch": 0.45, "learning_rate": 1.2044380497623166e-05, "loss": 2.305, "step": 18315 }, { "epoch": 0.45, "learning_rate": 1.2043599021130778e-05, "loss": 2.0546, "step": 18316 }, { "epoch": 0.45, "learning_rate": 1.204281753161387e-05, "loss": 2.0724, "step": 18317 }, { "epoch": 0.45, "learning_rate": 1.2042036029077422e-05, "loss": 2.1182, "step": 18318 }, { "epoch": 0.45, "learning_rate": 1.204125451352641e-05, "loss": 2.0749, "step": 18319 }, { "epoch": 0.45, "learning_rate": 1.2040472984965818e-05, "loss": 2.0842, "step": 18320 }, { "epoch": 0.45, "learning_rate": 1.203969144340063e-05, "loss": 2.0626, "step": 18321 }, { "epoch": 0.45, "learning_rate": 1.2038909888835821e-05, "loss": 1.9901, "step": 18322 }, { "epoch": 0.45, "learning_rate": 1.2038128321276377e-05, "loss": 2.0935, "step": 18323 }, { "epoch": 0.45, "learning_rate": 1.2037346740727273e-05, "loss": 1.9992, "step": 18324 }, { "epoch": 0.45, "learning_rate": 1.2036565147193494e-05, "loss": 2.0565, "step": 18325 }, { "epoch": 0.45, "learning_rate": 1.2035783540680027e-05, "loss": 2.1047, "step": 18326 }, { "epoch": 0.45, "learning_rate": 1.2035001921191844e-05, "loss": 1.9825, "step": 18327 }, { "epoch": 0.45, "learning_rate": 1.2034220288733934e-05, "loss": 2.0579, "step": 18328 }, { "epoch": 0.45, "learning_rate": 1.203343864331127e-05, "loss": 2.0027, "step": 18329 }, { "epoch": 0.45, "learning_rate": 1.2032656984928843e-05, "loss": 2.094, "step": 18330 }, { "epoch": 0.45, "learning_rate": 1.203187531359163e-05, "loss": 2.1025, "step": 18331 }, { "epoch": 0.45, "learning_rate": 1.2031093629304614e-05, "loss": 2.236, "step": 18332 }, { "epoch": 0.45, "learning_rate": 1.2030311932072774e-05, "loss": 2.2884, "step": 18333 }, { "epoch": 0.45, "learning_rate": 1.2029530221901095e-05, "loss": 2.1001, "step": 18334 }, { "epoch": 0.45, "learning_rate": 1.2028748498794558e-05, "loss": 2.0557, "step": 18335 }, { "epoch": 0.45, "learning_rate": 1.2027966762758147e-05, "loss": 2.0443, "step": 18336 }, { "epoch": 0.45, "learning_rate": 1.2027185013796843e-05, "loss": 2.0961, "step": 18337 }, { "epoch": 0.45, "learning_rate": 1.2026403251915625e-05, "loss": 2.1501, "step": 18338 }, { "epoch": 0.45, "learning_rate": 1.2025621477119485e-05, "loss": 1.929, "step": 18339 }, { "epoch": 0.45, "learning_rate": 1.2024839689413394e-05, "loss": 1.9438, "step": 18340 }, { "epoch": 0.45, "learning_rate": 1.2024057888802342e-05, "loss": 2.0411, "step": 18341 }, { "epoch": 0.45, "learning_rate": 1.2023276075291308e-05, "loss": 2.08, "step": 18342 }, { "epoch": 0.45, "learning_rate": 1.2022494248885279e-05, "loss": 1.9613, "step": 18343 }, { "epoch": 0.45, "learning_rate": 1.2021712409589231e-05, "loss": 2.0777, "step": 18344 }, { "epoch": 0.45, "learning_rate": 1.2020930557408155e-05, "loss": 1.9532, "step": 18345 }, { "epoch": 0.45, "learning_rate": 1.2020148692347026e-05, "loss": 2.174, "step": 18346 }, { "epoch": 0.45, "learning_rate": 1.2019366814410836e-05, "loss": 2.1865, "step": 18347 }, { "epoch": 0.45, "learning_rate": 1.201858492360456e-05, "loss": 1.944, "step": 18348 }, { "epoch": 0.45, "learning_rate": 1.2017803019933183e-05, "loss": 1.9469, "step": 18349 }, { "epoch": 0.45, "learning_rate": 1.2017021103401694e-05, "loss": 2.0303, "step": 18350 }, { "epoch": 0.45, "learning_rate": 1.2016239174015067e-05, "loss": 2.1498, "step": 18351 }, { "epoch": 0.45, "learning_rate": 1.2015457231778294e-05, "loss": 2.2434, "step": 18352 }, { "epoch": 0.45, "learning_rate": 1.2014675276696352e-05, "loss": 2.055, "step": 18353 }, { "epoch": 0.45, "learning_rate": 1.2013893308774234e-05, "loss": 2.0134, "step": 18354 }, { "epoch": 0.45, "learning_rate": 1.2013111328016913e-05, "loss": 2.2071, "step": 18355 }, { "epoch": 0.45, "learning_rate": 1.201232933442938e-05, "loss": 1.987, "step": 18356 }, { "epoch": 0.45, "learning_rate": 1.2011547328016614e-05, "loss": 1.9721, "step": 18357 }, { "epoch": 0.45, "learning_rate": 1.2010765308783602e-05, "loss": 1.9979, "step": 18358 }, { "epoch": 0.45, "learning_rate": 1.2009983276735329e-05, "loss": 1.9515, "step": 18359 }, { "epoch": 0.45, "learning_rate": 1.2009201231876775e-05, "loss": 2.1499, "step": 18360 }, { "epoch": 0.45, "learning_rate": 1.200841917421293e-05, "loss": 2.117, "step": 18361 }, { "epoch": 0.45, "learning_rate": 1.200763710374877e-05, "loss": 2.121, "step": 18362 }, { "epoch": 0.45, "learning_rate": 1.2006855020489287e-05, "loss": 2.1042, "step": 18363 }, { "epoch": 0.45, "learning_rate": 1.2006072924439465e-05, "loss": 2.0799, "step": 18364 }, { "epoch": 0.45, "learning_rate": 1.2005290815604287e-05, "loss": 1.9767, "step": 18365 }, { "epoch": 0.45, "learning_rate": 1.2004508693988734e-05, "loss": 1.9321, "step": 18366 }, { "epoch": 0.45, "learning_rate": 1.2003726559597795e-05, "loss": 1.971, "step": 18367 }, { "epoch": 0.45, "learning_rate": 1.2002944412436455e-05, "loss": 2.2157, "step": 18368 }, { "epoch": 0.45, "learning_rate": 1.2002162252509693e-05, "loss": 1.9027, "step": 18369 }, { "epoch": 0.45, "learning_rate": 1.2001380079822501e-05, "loss": 2.1531, "step": 18370 }, { "epoch": 0.45, "learning_rate": 1.2000597894379865e-05, "loss": 2.0352, "step": 18371 }, { "epoch": 0.45, "learning_rate": 1.1999815696186763e-05, "loss": 2.1385, "step": 18372 }, { "epoch": 0.45, "learning_rate": 1.1999033485248184e-05, "loss": 2.093, "step": 18373 }, { "epoch": 0.45, "learning_rate": 1.1998251261569113e-05, "loss": 2.0352, "step": 18374 }, { "epoch": 0.45, "learning_rate": 1.1997469025154537e-05, "loss": 2.2053, "step": 18375 }, { "epoch": 0.45, "learning_rate": 1.1996686776009441e-05, "loss": 1.899, "step": 18376 }, { "epoch": 0.45, "learning_rate": 1.1995904514138808e-05, "loss": 1.9946, "step": 18377 }, { "epoch": 0.45, "learning_rate": 1.1995122239547623e-05, "loss": 1.9709, "step": 18378 }, { "epoch": 0.45, "learning_rate": 1.1994339952240876e-05, "loss": 1.9915, "step": 18379 }, { "epoch": 0.45, "learning_rate": 1.199355765222355e-05, "loss": 1.9196, "step": 18380 }, { "epoch": 0.45, "learning_rate": 1.1992775339500632e-05, "loss": 2.0474, "step": 18381 }, { "epoch": 0.45, "learning_rate": 1.1991993014077105e-05, "loss": 1.9333, "step": 18382 }, { "epoch": 0.45, "learning_rate": 1.199121067595796e-05, "loss": 2.0977, "step": 18383 }, { "epoch": 0.45, "learning_rate": 1.199042832514818e-05, "loss": 2.1243, "step": 18384 }, { "epoch": 0.45, "learning_rate": 1.1989645961652752e-05, "loss": 2.1295, "step": 18385 }, { "epoch": 0.45, "learning_rate": 1.198886358547666e-05, "loss": 2.0641, "step": 18386 }, { "epoch": 0.45, "learning_rate": 1.1988081196624893e-05, "loss": 1.8752, "step": 18387 }, { "epoch": 0.45, "learning_rate": 1.1987298795102437e-05, "loss": 2.0736, "step": 18388 }, { "epoch": 0.45, "learning_rate": 1.1986516380914279e-05, "loss": 1.9186, "step": 18389 }, { "epoch": 0.45, "learning_rate": 1.1985733954065404e-05, "loss": 2.0017, "step": 18390 }, { "epoch": 0.45, "learning_rate": 1.1984951514560798e-05, "loss": 2.0273, "step": 18391 }, { "epoch": 0.45, "learning_rate": 1.198416906240545e-05, "loss": 2.0456, "step": 18392 }, { "epoch": 0.45, "learning_rate": 1.1983386597604348e-05, "loss": 1.9812, "step": 18393 }, { "epoch": 0.45, "learning_rate": 1.1982604120162476e-05, "loss": 2.0977, "step": 18394 }, { "epoch": 0.45, "learning_rate": 1.198182163008482e-05, "loss": 1.9279, "step": 18395 }, { "epoch": 0.45, "learning_rate": 1.1981039127376368e-05, "loss": 1.9922, "step": 18396 }, { "epoch": 0.45, "learning_rate": 1.1980256612042109e-05, "loss": 1.9679, "step": 18397 }, { "epoch": 0.45, "learning_rate": 1.1979474084087032e-05, "loss": 2.2242, "step": 18398 }, { "epoch": 0.45, "learning_rate": 1.197869154351612e-05, "loss": 2.0154, "step": 18399 }, { "epoch": 0.45, "learning_rate": 1.1977908990334364e-05, "loss": 2.0889, "step": 18400 }, { "epoch": 0.45, "learning_rate": 1.1977126424546746e-05, "loss": 1.913, "step": 18401 }, { "epoch": 0.45, "learning_rate": 1.1976343846158259e-05, "loss": 1.9266, "step": 18402 }, { "epoch": 0.45, "learning_rate": 1.1975561255173887e-05, "loss": 2.0556, "step": 18403 }, { "epoch": 0.45, "learning_rate": 1.1974778651598618e-05, "loss": 2.0593, "step": 18404 }, { "epoch": 0.45, "learning_rate": 1.1973996035437443e-05, "loss": 2.1565, "step": 18405 }, { "epoch": 0.45, "learning_rate": 1.1973213406695348e-05, "loss": 1.9733, "step": 18406 }, { "epoch": 0.45, "learning_rate": 1.197243076537732e-05, "loss": 2.1836, "step": 18407 }, { "epoch": 0.45, "learning_rate": 1.197164811148835e-05, "loss": 2.0401, "step": 18408 }, { "epoch": 0.45, "learning_rate": 1.1970865445033423e-05, "loss": 1.9679, "step": 18409 }, { "epoch": 0.45, "learning_rate": 1.197008276601753e-05, "loss": 2.1404, "step": 18410 }, { "epoch": 0.45, "learning_rate": 1.1969300074445654e-05, "loss": 1.9356, "step": 18411 }, { "epoch": 0.45, "learning_rate": 1.196851737032279e-05, "loss": 1.9606, "step": 18412 }, { "epoch": 0.45, "learning_rate": 1.196773465365392e-05, "loss": 2.0898, "step": 18413 }, { "epoch": 0.45, "learning_rate": 1.1966951924444037e-05, "loss": 2.0161, "step": 18414 }, { "epoch": 0.45, "learning_rate": 1.1966169182698131e-05, "loss": 2.15, "step": 18415 }, { "epoch": 0.45, "learning_rate": 1.1965386428421185e-05, "loss": 2.0937, "step": 18416 }, { "epoch": 0.45, "learning_rate": 1.1964603661618191e-05, "loss": 2.0504, "step": 18417 }, { "epoch": 0.45, "learning_rate": 1.196382088229414e-05, "loss": 2.0034, "step": 18418 }, { "epoch": 0.45, "learning_rate": 1.1963038090454018e-05, "loss": 2.0263, "step": 18419 }, { "epoch": 0.45, "learning_rate": 1.196225528610281e-05, "loss": 2.1557, "step": 18420 }, { "epoch": 0.45, "learning_rate": 1.1961472469245515e-05, "loss": 2.2385, "step": 18421 }, { "epoch": 0.45, "learning_rate": 1.1960689639887114e-05, "loss": 2.0837, "step": 18422 }, { "epoch": 0.45, "learning_rate": 1.19599067980326e-05, "loss": 1.974, "step": 18423 }, { "epoch": 0.45, "learning_rate": 1.195912394368696e-05, "loss": 2.1872, "step": 18424 }, { "epoch": 0.45, "learning_rate": 1.1958341076855185e-05, "loss": 2.1065, "step": 18425 }, { "epoch": 0.45, "learning_rate": 1.1957558197542265e-05, "loss": 2.1167, "step": 18426 }, { "epoch": 0.45, "learning_rate": 1.195677530575319e-05, "loss": 2.1128, "step": 18427 }, { "epoch": 0.45, "learning_rate": 1.1955992401492946e-05, "loss": 1.9829, "step": 18428 }, { "epoch": 0.45, "learning_rate": 1.1955209484766527e-05, "loss": 2.0324, "step": 18429 }, { "epoch": 0.45, "learning_rate": 1.1954426555578917e-05, "loss": 2.08, "step": 18430 }, { "epoch": 0.45, "learning_rate": 1.195364361393511e-05, "loss": 1.9318, "step": 18431 }, { "epoch": 0.45, "learning_rate": 1.19528606598401e-05, "loss": 1.844, "step": 18432 }, { "epoch": 0.45, "learning_rate": 1.1952077693298872e-05, "loss": 2.0968, "step": 18433 }, { "epoch": 0.45, "learning_rate": 1.1951294714316415e-05, "loss": 2.0764, "step": 18434 }, { "epoch": 0.45, "learning_rate": 1.1950511722897718e-05, "loss": 2.0535, "step": 18435 }, { "epoch": 0.45, "learning_rate": 1.1949728719047777e-05, "loss": 1.9771, "step": 18436 }, { "epoch": 0.45, "learning_rate": 1.1948945702771581e-05, "loss": 2.1171, "step": 18437 }, { "epoch": 0.45, "learning_rate": 1.1948162674074118e-05, "loss": 2.116, "step": 18438 }, { "epoch": 0.45, "learning_rate": 1.1947379632960377e-05, "loss": 2.1797, "step": 18439 }, { "epoch": 0.45, "learning_rate": 1.1946596579435352e-05, "loss": 2.1052, "step": 18440 }, { "epoch": 0.45, "learning_rate": 1.1945813513504035e-05, "loss": 2.0636, "step": 18441 }, { "epoch": 0.45, "learning_rate": 1.1945030435171412e-05, "loss": 2.0618, "step": 18442 }, { "epoch": 0.45, "learning_rate": 1.1944247344442477e-05, "loss": 2.1529, "step": 18443 }, { "epoch": 0.45, "learning_rate": 1.1943464241322221e-05, "loss": 2.1218, "step": 18444 }, { "epoch": 0.45, "learning_rate": 1.1942681125815631e-05, "loss": 2.1784, "step": 18445 }, { "epoch": 0.45, "learning_rate": 1.1941897997927705e-05, "loss": 2.2711, "step": 18446 }, { "epoch": 0.45, "learning_rate": 1.1941114857663427e-05, "loss": 2.1725, "step": 18447 }, { "epoch": 0.45, "learning_rate": 1.1940331705027792e-05, "loss": 2.061, "step": 18448 }, { "epoch": 0.45, "learning_rate": 1.1939548540025792e-05, "loss": 1.977, "step": 18449 }, { "epoch": 0.45, "learning_rate": 1.1938765362662414e-05, "loss": 2.0084, "step": 18450 }, { "epoch": 0.45, "learning_rate": 1.1937982172942656e-05, "loss": 1.853, "step": 18451 }, { "epoch": 0.45, "learning_rate": 1.1937198970871506e-05, "loss": 2.0337, "step": 18452 }, { "epoch": 0.45, "learning_rate": 1.1936415756453953e-05, "loss": 2.017, "step": 18453 }, { "epoch": 0.45, "learning_rate": 1.1935632529694993e-05, "loss": 2.0136, "step": 18454 }, { "epoch": 0.45, "learning_rate": 1.1934849290599617e-05, "loss": 2.137, "step": 18455 }, { "epoch": 0.45, "learning_rate": 1.1934066039172816e-05, "loss": 1.9657, "step": 18456 }, { "epoch": 0.45, "learning_rate": 1.1933282775419578e-05, "loss": 1.9458, "step": 18457 }, { "epoch": 0.45, "learning_rate": 1.1932499499344903e-05, "loss": 2.2021, "step": 18458 }, { "epoch": 0.45, "learning_rate": 1.1931716210953776e-05, "loss": 1.9283, "step": 18459 }, { "epoch": 0.46, "learning_rate": 1.1930932910251194e-05, "loss": 1.9766, "step": 18460 }, { "epoch": 0.46, "learning_rate": 1.1930149597242149e-05, "loss": 1.9514, "step": 18461 }, { "epoch": 0.46, "learning_rate": 1.1929366271931628e-05, "loss": 2.1933, "step": 18462 }, { "epoch": 0.46, "learning_rate": 1.1928582934324631e-05, "loss": 1.8845, "step": 18463 }, { "epoch": 0.46, "learning_rate": 1.192779958442614e-05, "loss": 1.8593, "step": 18464 }, { "epoch": 0.46, "learning_rate": 1.192701622224116e-05, "loss": 2.0833, "step": 18465 }, { "epoch": 0.46, "learning_rate": 1.1926232847774674e-05, "loss": 1.8363, "step": 18466 }, { "epoch": 0.46, "learning_rate": 1.1925449461031682e-05, "loss": 1.9801, "step": 18467 }, { "epoch": 0.46, "learning_rate": 1.1924666062017169e-05, "loss": 1.9622, "step": 18468 }, { "epoch": 0.46, "learning_rate": 1.1923882650736133e-05, "loss": 2.1332, "step": 18469 }, { "epoch": 0.46, "learning_rate": 1.1923099227193569e-05, "loss": 2.303, "step": 18470 }, { "epoch": 0.46, "learning_rate": 1.1922315791394465e-05, "loss": 1.996, "step": 18471 }, { "epoch": 0.46, "learning_rate": 1.1921532343343816e-05, "loss": 2.0202, "step": 18472 }, { "epoch": 0.46, "learning_rate": 1.1920748883046615e-05, "loss": 1.9878, "step": 18473 }, { "epoch": 0.46, "learning_rate": 1.1919965410507854e-05, "loss": 1.9615, "step": 18474 }, { "epoch": 0.46, "learning_rate": 1.1919181925732527e-05, "loss": 2.1443, "step": 18475 }, { "epoch": 0.46, "learning_rate": 1.1918398428725632e-05, "loss": 1.9816, "step": 18476 }, { "epoch": 0.46, "learning_rate": 1.1917614919492158e-05, "loss": 2.1456, "step": 18477 }, { "epoch": 0.46, "learning_rate": 1.1916831398037098e-05, "loss": 2.0834, "step": 18478 }, { "epoch": 0.46, "learning_rate": 1.1916047864365445e-05, "loss": 2.1103, "step": 18479 }, { "epoch": 0.46, "learning_rate": 1.1915264318482195e-05, "loss": 2.0447, "step": 18480 }, { "epoch": 0.46, "learning_rate": 1.1914480760392344e-05, "loss": 2.015, "step": 18481 }, { "epoch": 0.46, "learning_rate": 1.1913697190100884e-05, "loss": 2.1161, "step": 18482 }, { "epoch": 0.46, "learning_rate": 1.1912913607612804e-05, "loss": 2.086, "step": 18483 }, { "epoch": 0.46, "learning_rate": 1.1912130012933103e-05, "loss": 2.0529, "step": 18484 }, { "epoch": 0.46, "learning_rate": 1.1911346406066775e-05, "loss": 1.9444, "step": 18485 }, { "epoch": 0.46, "learning_rate": 1.1910562787018815e-05, "loss": 1.9465, "step": 18486 }, { "epoch": 0.46, "learning_rate": 1.1909779155794216e-05, "loss": 1.9523, "step": 18487 }, { "epoch": 0.46, "learning_rate": 1.1908995512397969e-05, "loss": 2.1185, "step": 18488 }, { "epoch": 0.46, "learning_rate": 1.1908211856835074e-05, "loss": 2.145, "step": 18489 }, { "epoch": 0.46, "learning_rate": 1.1907428189110524e-05, "loss": 2.215, "step": 18490 }, { "epoch": 0.46, "learning_rate": 1.190664450922931e-05, "loss": 2.0311, "step": 18491 }, { "epoch": 0.46, "learning_rate": 1.1905860817196428e-05, "loss": 1.8791, "step": 18492 }, { "epoch": 0.46, "learning_rate": 1.1905077113016875e-05, "loss": 1.9922, "step": 18493 }, { "epoch": 0.46, "learning_rate": 1.1904293396695646e-05, "loss": 2.0736, "step": 18494 }, { "epoch": 0.46, "learning_rate": 1.1903509668237736e-05, "loss": 1.9421, "step": 18495 }, { "epoch": 0.46, "learning_rate": 1.1902725927648136e-05, "loss": 2.0229, "step": 18496 }, { "epoch": 0.46, "learning_rate": 1.1901942174931845e-05, "loss": 2.1266, "step": 18497 }, { "epoch": 0.46, "learning_rate": 1.1901158410093853e-05, "loss": 2.0646, "step": 18498 }, { "epoch": 0.46, "learning_rate": 1.1900374633139165e-05, "loss": 2.1696, "step": 18499 }, { "epoch": 0.46, "learning_rate": 1.1899590844072767e-05, "loss": 2.1495, "step": 18500 }, { "epoch": 0.46, "learning_rate": 1.1898807042899657e-05, "loss": 2.1857, "step": 18501 }, { "epoch": 0.46, "learning_rate": 1.1898023229624831e-05, "loss": 2.0858, "step": 18502 }, { "epoch": 0.46, "learning_rate": 1.1897239404253283e-05, "loss": 2.1292, "step": 18503 }, { "epoch": 0.46, "learning_rate": 1.1896455566790014e-05, "loss": 1.968, "step": 18504 }, { "epoch": 0.46, "learning_rate": 1.1895671717240011e-05, "loss": 1.9789, "step": 18505 }, { "epoch": 0.46, "learning_rate": 1.189488785560828e-05, "loss": 2.0613, "step": 18506 }, { "epoch": 0.46, "learning_rate": 1.1894103981899807e-05, "loss": 2.0436, "step": 18507 }, { "epoch": 0.46, "learning_rate": 1.1893320096119591e-05, "loss": 2.1439, "step": 18508 }, { "epoch": 0.46, "learning_rate": 1.1892536198272632e-05, "loss": 1.9261, "step": 18509 }, { "epoch": 0.46, "learning_rate": 1.1891752288363921e-05, "loss": 2.0229, "step": 18510 }, { "epoch": 0.46, "learning_rate": 1.1890968366398456e-05, "loss": 2.0709, "step": 18511 }, { "epoch": 0.46, "learning_rate": 1.1890184432381236e-05, "loss": 2.0205, "step": 18512 }, { "epoch": 0.46, "learning_rate": 1.1889400486317253e-05, "loss": 2.0554, "step": 18513 }, { "epoch": 0.46, "learning_rate": 1.1888616528211505e-05, "loss": 2.0928, "step": 18514 }, { "epoch": 0.46, "learning_rate": 1.188783255806899e-05, "loss": 1.8883, "step": 18515 }, { "epoch": 0.46, "learning_rate": 1.1887048575894702e-05, "loss": 1.9632, "step": 18516 }, { "epoch": 0.46, "learning_rate": 1.1886264581693635e-05, "loss": 2.2203, "step": 18517 }, { "epoch": 0.46, "learning_rate": 1.1885480575470793e-05, "loss": 2.0905, "step": 18518 }, { "epoch": 0.46, "learning_rate": 1.1884696557231167e-05, "loss": 2.0097, "step": 18519 }, { "epoch": 0.46, "learning_rate": 1.1883912526979757e-05, "loss": 1.896, "step": 18520 }, { "epoch": 0.46, "learning_rate": 1.1883128484721558e-05, "loss": 1.9929, "step": 18521 }, { "epoch": 0.46, "learning_rate": 1.1882344430461568e-05, "loss": 1.9123, "step": 18522 }, { "epoch": 0.46, "learning_rate": 1.1881560364204783e-05, "loss": 2.0544, "step": 18523 }, { "epoch": 0.46, "learning_rate": 1.18807762859562e-05, "loss": 2.0732, "step": 18524 }, { "epoch": 0.46, "learning_rate": 1.1879992195720818e-05, "loss": 2.1605, "step": 18525 }, { "epoch": 0.46, "learning_rate": 1.1879208093503632e-05, "loss": 2.017, "step": 18526 }, { "epoch": 0.46, "learning_rate": 1.1878423979309642e-05, "loss": 2.0637, "step": 18527 }, { "epoch": 0.46, "learning_rate": 1.1877639853143839e-05, "loss": 2.183, "step": 18528 }, { "epoch": 0.46, "learning_rate": 1.1876855715011231e-05, "loss": 1.875, "step": 18529 }, { "epoch": 0.46, "learning_rate": 1.1876071564916807e-05, "loss": 2.0465, "step": 18530 }, { "epoch": 0.46, "learning_rate": 1.1875287402865569e-05, "loss": 1.9894, "step": 18531 }, { "epoch": 0.46, "learning_rate": 1.1874503228862513e-05, "loss": 2.1643, "step": 18532 }, { "epoch": 0.46, "learning_rate": 1.1873719042912637e-05, "loss": 2.1712, "step": 18533 }, { "epoch": 0.46, "learning_rate": 1.1872934845020938e-05, "loss": 2.0357, "step": 18534 }, { "epoch": 0.46, "learning_rate": 1.1872150635192414e-05, "loss": 1.9557, "step": 18535 }, { "epoch": 0.46, "learning_rate": 1.1871366413432067e-05, "loss": 1.7874, "step": 18536 }, { "epoch": 0.46, "learning_rate": 1.1870582179744888e-05, "loss": 2.0893, "step": 18537 }, { "epoch": 0.46, "learning_rate": 1.1869797934135882e-05, "loss": 2.1122, "step": 18538 }, { "epoch": 0.46, "learning_rate": 1.1869013676610044e-05, "loss": 2.0772, "step": 18539 }, { "epoch": 0.46, "learning_rate": 1.1868229407172374e-05, "loss": 1.9675, "step": 18540 }, { "epoch": 0.46, "learning_rate": 1.1867445125827865e-05, "loss": 2.0009, "step": 18541 }, { "epoch": 0.46, "learning_rate": 1.1866660832581525e-05, "loss": 2.071, "step": 18542 }, { "epoch": 0.46, "learning_rate": 1.1865876527438345e-05, "loss": 2.0841, "step": 18543 }, { "epoch": 0.46, "learning_rate": 1.1865092210403325e-05, "loss": 2.1316, "step": 18544 }, { "epoch": 0.46, "learning_rate": 1.1864307881481462e-05, "loss": 2.044, "step": 18545 }, { "epoch": 0.46, "learning_rate": 1.186352354067776e-05, "loss": 2.065, "step": 18546 }, { "epoch": 0.46, "learning_rate": 1.1862739187997217e-05, "loss": 2.2833, "step": 18547 }, { "epoch": 0.46, "learning_rate": 1.1861954823444828e-05, "loss": 2.0035, "step": 18548 }, { "epoch": 0.46, "learning_rate": 1.1861170447025596e-05, "loss": 2.1224, "step": 18549 }, { "epoch": 0.46, "learning_rate": 1.1860386058744517e-05, "loss": 2.046, "step": 18550 }, { "epoch": 0.46, "learning_rate": 1.185960165860659e-05, "loss": 1.8942, "step": 18551 }, { "epoch": 0.46, "learning_rate": 1.1858817246616818e-05, "loss": 1.9928, "step": 18552 }, { "epoch": 0.46, "learning_rate": 1.1858032822780197e-05, "loss": 2.1265, "step": 18553 }, { "epoch": 0.46, "learning_rate": 1.1857248387101724e-05, "loss": 1.8541, "step": 18554 }, { "epoch": 0.46, "learning_rate": 1.1856463939586407e-05, "loss": 2.0064, "step": 18555 }, { "epoch": 0.46, "learning_rate": 1.1855679480239238e-05, "loss": 2.0929, "step": 18556 }, { "epoch": 0.46, "learning_rate": 1.1854895009065216e-05, "loss": 2.0021, "step": 18557 }, { "epoch": 0.46, "learning_rate": 1.1854110526069349e-05, "loss": 1.9454, "step": 18558 }, { "epoch": 0.46, "learning_rate": 1.1853326031256629e-05, "loss": 1.9599, "step": 18559 }, { "epoch": 0.46, "learning_rate": 1.185254152463206e-05, "loss": 2.025, "step": 18560 }, { "epoch": 0.46, "learning_rate": 1.1851757006200637e-05, "loss": 2.2205, "step": 18561 }, { "epoch": 0.46, "learning_rate": 1.1850972475967363e-05, "loss": 2.081, "step": 18562 }, { "epoch": 0.46, "learning_rate": 1.185018793393724e-05, "loss": 2.0537, "step": 18563 }, { "epoch": 0.46, "learning_rate": 1.1849403380115265e-05, "loss": 2.1291, "step": 18564 }, { "epoch": 0.46, "learning_rate": 1.1848618814506443e-05, "loss": 2.0836, "step": 18565 }, { "epoch": 0.46, "learning_rate": 1.1847834237115767e-05, "loss": 2.0489, "step": 18566 }, { "epoch": 0.46, "learning_rate": 1.1847049647948244e-05, "loss": 2.0395, "step": 18567 }, { "epoch": 0.46, "learning_rate": 1.184626504700887e-05, "loss": 1.917, "step": 18568 }, { "epoch": 0.46, "learning_rate": 1.1845480434302648e-05, "loss": 2.0898, "step": 18569 }, { "epoch": 0.46, "learning_rate": 1.1844695809834577e-05, "loss": 2.0311, "step": 18570 }, { "epoch": 0.46, "learning_rate": 1.1843911173609659e-05, "loss": 2.0013, "step": 18571 }, { "epoch": 0.46, "learning_rate": 1.1843126525632892e-05, "loss": 1.9523, "step": 18572 }, { "epoch": 0.46, "learning_rate": 1.1842341865909282e-05, "loss": 2.1511, "step": 18573 }, { "epoch": 0.46, "learning_rate": 1.1841557194443824e-05, "loss": 2.1103, "step": 18574 }, { "epoch": 0.46, "learning_rate": 1.1840772511241523e-05, "loss": 2.2245, "step": 18575 }, { "epoch": 0.46, "learning_rate": 1.183998781630738e-05, "loss": 2.1701, "step": 18576 }, { "epoch": 0.46, "learning_rate": 1.1839203109646396e-05, "loss": 2.1683, "step": 18577 }, { "epoch": 0.46, "learning_rate": 1.183841839126357e-05, "loss": 2.0507, "step": 18578 }, { "epoch": 0.46, "learning_rate": 1.1837633661163903e-05, "loss": 2.0678, "step": 18579 }, { "epoch": 0.46, "learning_rate": 1.18368489193524e-05, "loss": 2.2262, "step": 18580 }, { "epoch": 0.46, "learning_rate": 1.1836064165834059e-05, "loss": 2.1105, "step": 18581 }, { "epoch": 0.46, "learning_rate": 1.1835279400613881e-05, "loss": 2.2805, "step": 18582 }, { "epoch": 0.46, "learning_rate": 1.1834494623696873e-05, "loss": 1.8918, "step": 18583 }, { "epoch": 0.46, "learning_rate": 1.1833709835088032e-05, "loss": 2.0723, "step": 18584 }, { "epoch": 0.46, "learning_rate": 1.1832925034792358e-05, "loss": 1.9155, "step": 18585 }, { "epoch": 0.46, "learning_rate": 1.1832140222814857e-05, "loss": 1.9432, "step": 18586 }, { "epoch": 0.46, "learning_rate": 1.1831355399160531e-05, "loss": 2.1033, "step": 18587 }, { "epoch": 0.46, "learning_rate": 1.1830570563834376e-05, "loss": 1.9252, "step": 18588 }, { "epoch": 0.46, "learning_rate": 1.1829785716841402e-05, "loss": 2.1786, "step": 18589 }, { "epoch": 0.46, "learning_rate": 1.1829000858186606e-05, "loss": 2.1491, "step": 18590 }, { "epoch": 0.46, "learning_rate": 1.1828215987874991e-05, "loss": 1.9475, "step": 18591 }, { "epoch": 0.46, "learning_rate": 1.1827431105911563e-05, "loss": 2.1233, "step": 18592 }, { "epoch": 0.46, "learning_rate": 1.1826646212301317e-05, "loss": 2.0022, "step": 18593 }, { "epoch": 0.46, "learning_rate": 1.1825861307049262e-05, "loss": 2.1337, "step": 18594 }, { "epoch": 0.46, "learning_rate": 1.1825076390160395e-05, "loss": 2.0966, "step": 18595 }, { "epoch": 0.46, "learning_rate": 1.1824291461639722e-05, "loss": 1.9724, "step": 18596 }, { "epoch": 0.46, "learning_rate": 1.1823506521492245e-05, "loss": 2.2673, "step": 18597 }, { "epoch": 0.46, "learning_rate": 1.182272156972297e-05, "loss": 2.1476, "step": 18598 }, { "epoch": 0.46, "learning_rate": 1.1821936606336892e-05, "loss": 1.9832, "step": 18599 }, { "epoch": 0.46, "learning_rate": 1.182115163133902e-05, "loss": 2.0274, "step": 18600 }, { "epoch": 0.46, "learning_rate": 1.1820366644734357e-05, "loss": 1.9864, "step": 18601 }, { "epoch": 0.46, "learning_rate": 1.1819581646527905e-05, "loss": 2.0757, "step": 18602 }, { "epoch": 0.46, "learning_rate": 1.1818796636724663e-05, "loss": 2.0969, "step": 18603 }, { "epoch": 0.46, "learning_rate": 1.1818011615329637e-05, "loss": 1.9821, "step": 18604 }, { "epoch": 0.46, "learning_rate": 1.1817226582347833e-05, "loss": 2.2886, "step": 18605 }, { "epoch": 0.46, "learning_rate": 1.1816441537784249e-05, "loss": 2.0422, "step": 18606 }, { "epoch": 0.46, "learning_rate": 1.1815656481643893e-05, "loss": 1.9595, "step": 18607 }, { "epoch": 0.46, "learning_rate": 1.1814871413931766e-05, "loss": 2.0242, "step": 18608 }, { "epoch": 0.46, "learning_rate": 1.1814086334652873e-05, "loss": 1.9533, "step": 18609 }, { "epoch": 0.46, "learning_rate": 1.1813301243812214e-05, "loss": 1.9249, "step": 18610 }, { "epoch": 0.46, "learning_rate": 1.1812516141414798e-05, "loss": 2.1649, "step": 18611 }, { "epoch": 0.46, "learning_rate": 1.1811731027465627e-05, "loss": 1.9192, "step": 18612 }, { "epoch": 0.46, "learning_rate": 1.1810945901969702e-05, "loss": 2.0044, "step": 18613 }, { "epoch": 0.46, "learning_rate": 1.1810160764932025e-05, "loss": 2.1411, "step": 18614 }, { "epoch": 0.46, "learning_rate": 1.1809375616357606e-05, "loss": 2.0712, "step": 18615 }, { "epoch": 0.46, "learning_rate": 1.1808590456251448e-05, "loss": 2.1168, "step": 18616 }, { "epoch": 0.46, "learning_rate": 1.1807805284618555e-05, "loss": 2.0264, "step": 18617 }, { "epoch": 0.46, "learning_rate": 1.1807020101463929e-05, "loss": 2.0445, "step": 18618 }, { "epoch": 0.46, "learning_rate": 1.180623490679257e-05, "loss": 2.0808, "step": 18619 }, { "epoch": 0.46, "learning_rate": 1.1805449700609492e-05, "loss": 2.031, "step": 18620 }, { "epoch": 0.46, "learning_rate": 1.1804664482919694e-05, "loss": 2.0353, "step": 18621 }, { "epoch": 0.46, "learning_rate": 1.180387925372818e-05, "loss": 2.1287, "step": 18622 }, { "epoch": 0.46, "learning_rate": 1.1803094013039958e-05, "loss": 2.1681, "step": 18623 }, { "epoch": 0.46, "learning_rate": 1.1802308760860026e-05, "loss": 2.1272, "step": 18624 }, { "epoch": 0.46, "learning_rate": 1.1801523497193397e-05, "loss": 1.9998, "step": 18625 }, { "epoch": 0.46, "learning_rate": 1.1800738222045072e-05, "loss": 2.0468, "step": 18626 }, { "epoch": 0.46, "learning_rate": 1.1799952935420054e-05, "loss": 2.1683, "step": 18627 }, { "epoch": 0.46, "learning_rate": 1.1799167637323348e-05, "loss": 2.1017, "step": 18628 }, { "epoch": 0.46, "learning_rate": 1.1798382327759961e-05, "loss": 2.0988, "step": 18629 }, { "epoch": 0.46, "learning_rate": 1.1797597006734898e-05, "loss": 2.1038, "step": 18630 }, { "epoch": 0.46, "learning_rate": 1.1796811674253163e-05, "loss": 2.0718, "step": 18631 }, { "epoch": 0.46, "learning_rate": 1.179602633031976e-05, "loss": 2.0751, "step": 18632 }, { "epoch": 0.46, "learning_rate": 1.17952409749397e-05, "loss": 1.987, "step": 18633 }, { "epoch": 0.46, "learning_rate": 1.1794455608117981e-05, "loss": 2.0873, "step": 18634 }, { "epoch": 0.46, "learning_rate": 1.1793670229859614e-05, "loss": 1.9159, "step": 18635 }, { "epoch": 0.46, "learning_rate": 1.17928848401696e-05, "loss": 2.061, "step": 18636 }, { "epoch": 0.46, "learning_rate": 1.179209943905295e-05, "loss": 2.2004, "step": 18637 }, { "epoch": 0.46, "learning_rate": 1.179131402651466e-05, "loss": 2.1166, "step": 18638 }, { "epoch": 0.46, "learning_rate": 1.1790528602559745e-05, "loss": 1.9985, "step": 18639 }, { "epoch": 0.46, "learning_rate": 1.178974316719321e-05, "loss": 1.9621, "step": 18640 }, { "epoch": 0.46, "learning_rate": 1.1788957720420056e-05, "loss": 2.1519, "step": 18641 }, { "epoch": 0.46, "learning_rate": 1.1788172262245292e-05, "loss": 2.0179, "step": 18642 }, { "epoch": 0.46, "learning_rate": 1.1787386792673924e-05, "loss": 2.0581, "step": 18643 }, { "epoch": 0.46, "learning_rate": 1.1786601311710956e-05, "loss": 2.0386, "step": 18644 }, { "epoch": 0.46, "learning_rate": 1.1785815819361397e-05, "loss": 2.0653, "step": 18645 }, { "epoch": 0.46, "learning_rate": 1.1785030315630253e-05, "loss": 2.1147, "step": 18646 }, { "epoch": 0.46, "learning_rate": 1.178424480052253e-05, "loss": 1.9327, "step": 18647 }, { "epoch": 0.46, "learning_rate": 1.1783459274043227e-05, "loss": 2.1005, "step": 18648 }, { "epoch": 0.46, "learning_rate": 1.178267373619736e-05, "loss": 2.024, "step": 18649 }, { "epoch": 0.46, "learning_rate": 1.1781888186989933e-05, "loss": 2.1726, "step": 18650 }, { "epoch": 0.46, "learning_rate": 1.1781102626425954e-05, "loss": 1.9632, "step": 18651 }, { "epoch": 0.46, "learning_rate": 1.1780317054510425e-05, "loss": 1.8836, "step": 18652 }, { "epoch": 0.46, "learning_rate": 1.1779531471248355e-05, "loss": 2.1548, "step": 18653 }, { "epoch": 0.46, "learning_rate": 1.1778745876644753e-05, "loss": 2.1933, "step": 18654 }, { "epoch": 0.46, "learning_rate": 1.1777960270704622e-05, "loss": 2.1786, "step": 18655 }, { "epoch": 0.46, "learning_rate": 1.1777174653432973e-05, "loss": 1.9567, "step": 18656 }, { "epoch": 0.46, "learning_rate": 1.1776389024834807e-05, "loss": 2.0708, "step": 18657 }, { "epoch": 0.46, "learning_rate": 1.1775603384915137e-05, "loss": 2.1115, "step": 18658 }, { "epoch": 0.46, "learning_rate": 1.1774817733678967e-05, "loss": 1.9831, "step": 18659 }, { "epoch": 0.46, "learning_rate": 1.1774032071131306e-05, "loss": 2.0157, "step": 18660 }, { "epoch": 0.46, "learning_rate": 1.1773246397277163e-05, "loss": 2.0162, "step": 18661 }, { "epoch": 0.46, "learning_rate": 1.177246071212154e-05, "loss": 1.9226, "step": 18662 }, { "epoch": 0.46, "learning_rate": 1.1771675015669447e-05, "loss": 1.9246, "step": 18663 }, { "epoch": 0.46, "learning_rate": 1.1770889307925893e-05, "loss": 2.0235, "step": 18664 }, { "epoch": 0.46, "learning_rate": 1.1770103588895885e-05, "loss": 2.1816, "step": 18665 }, { "epoch": 0.46, "learning_rate": 1.1769317858584428e-05, "loss": 2.1201, "step": 18666 }, { "epoch": 0.46, "learning_rate": 1.1768532116996531e-05, "loss": 1.8621, "step": 18667 }, { "epoch": 0.46, "learning_rate": 1.1767746364137202e-05, "loss": 2.1666, "step": 18668 }, { "epoch": 0.46, "learning_rate": 1.1766960600011451e-05, "loss": 2.0641, "step": 18669 }, { "epoch": 0.46, "learning_rate": 1.1766174824624286e-05, "loss": 2.032, "step": 18670 }, { "epoch": 0.46, "learning_rate": 1.1765389037980713e-05, "loss": 2.1062, "step": 18671 }, { "epoch": 0.46, "learning_rate": 1.1764603240085737e-05, "loss": 1.9221, "step": 18672 }, { "epoch": 0.46, "learning_rate": 1.1763817430944374e-05, "loss": 2.0822, "step": 18673 }, { "epoch": 0.46, "learning_rate": 1.1763031610561627e-05, "loss": 2.154, "step": 18674 }, { "epoch": 0.46, "learning_rate": 1.1762245778942503e-05, "loss": 2.0923, "step": 18675 }, { "epoch": 0.46, "learning_rate": 1.176145993609201e-05, "loss": 2.0657, "step": 18676 }, { "epoch": 0.46, "learning_rate": 1.176067408201516e-05, "loss": 2.134, "step": 18677 }, { "epoch": 0.46, "learning_rate": 1.1759888216716962e-05, "loss": 2.1303, "step": 18678 }, { "epoch": 0.46, "learning_rate": 1.1759102340202424e-05, "loss": 2.1629, "step": 18679 }, { "epoch": 0.46, "learning_rate": 1.1758316452476553e-05, "loss": 1.9789, "step": 18680 }, { "epoch": 0.46, "learning_rate": 1.1757530553544357e-05, "loss": 2.1711, "step": 18681 }, { "epoch": 0.46, "learning_rate": 1.1756744643410844e-05, "loss": 2.1897, "step": 18682 }, { "epoch": 0.46, "learning_rate": 1.175595872208103e-05, "loss": 2.1092, "step": 18683 }, { "epoch": 0.46, "learning_rate": 1.1755172789559914e-05, "loss": 2.1126, "step": 18684 }, { "epoch": 0.46, "learning_rate": 1.1754386845852512e-05, "loss": 2.039, "step": 18685 }, { "epoch": 0.46, "learning_rate": 1.1753600890963832e-05, "loss": 1.9401, "step": 18686 }, { "epoch": 0.46, "learning_rate": 1.1752814924898879e-05, "loss": 2.0586, "step": 18687 }, { "epoch": 0.46, "learning_rate": 1.1752028947662666e-05, "loss": 1.9458, "step": 18688 }, { "epoch": 0.46, "learning_rate": 1.1751242959260205e-05, "loss": 1.8687, "step": 18689 }, { "epoch": 0.46, "learning_rate": 1.17504569596965e-05, "loss": 2.0262, "step": 18690 }, { "epoch": 0.46, "learning_rate": 1.1749670948976559e-05, "loss": 1.9769, "step": 18691 }, { "epoch": 0.46, "learning_rate": 1.17488849271054e-05, "loss": 1.814, "step": 18692 }, { "epoch": 0.46, "learning_rate": 1.1748098894088024e-05, "loss": 2.029, "step": 18693 }, { "epoch": 0.46, "learning_rate": 1.1747312849929444e-05, "loss": 2.1407, "step": 18694 }, { "epoch": 0.46, "learning_rate": 1.1746526794634672e-05, "loss": 1.9204, "step": 18695 }, { "epoch": 0.46, "learning_rate": 1.1745740728208712e-05, "loss": 2.0782, "step": 18696 }, { "epoch": 0.46, "learning_rate": 1.174495465065658e-05, "loss": 2.1041, "step": 18697 }, { "epoch": 0.46, "learning_rate": 1.1744168561983285e-05, "loss": 1.9804, "step": 18698 }, { "epoch": 0.46, "learning_rate": 1.1743382462193835e-05, "loss": 2.0517, "step": 18699 }, { "epoch": 0.46, "learning_rate": 1.1742596351293238e-05, "loss": 1.9933, "step": 18700 }, { "epoch": 0.46, "learning_rate": 1.1741810229286507e-05, "loss": 2.034, "step": 18701 }, { "epoch": 0.46, "learning_rate": 1.1741024096178652e-05, "loss": 2.118, "step": 18702 }, { "epoch": 0.46, "learning_rate": 1.1740237951974682e-05, "loss": 2.017, "step": 18703 }, { "epoch": 0.46, "learning_rate": 1.1739451796679611e-05, "loss": 1.9618, "step": 18704 }, { "epoch": 0.46, "learning_rate": 1.1738665630298447e-05, "loss": 2.1593, "step": 18705 }, { "epoch": 0.46, "learning_rate": 1.1737879452836196e-05, "loss": 2.0986, "step": 18706 }, { "epoch": 0.46, "learning_rate": 1.1737093264297879e-05, "loss": 1.9407, "step": 18707 }, { "epoch": 0.46, "learning_rate": 1.1736307064688496e-05, "loss": 1.9215, "step": 18708 }, { "epoch": 0.46, "learning_rate": 1.1735520854013066e-05, "loss": 1.8541, "step": 18709 }, { "epoch": 0.46, "learning_rate": 1.1734734632276594e-05, "loss": 2.0647, "step": 18710 }, { "epoch": 0.46, "learning_rate": 1.173394839948409e-05, "loss": 2.0991, "step": 18711 }, { "epoch": 0.46, "learning_rate": 1.173316215564057e-05, "loss": 2.0821, "step": 18712 }, { "epoch": 0.46, "learning_rate": 1.1732375900751043e-05, "loss": 1.975, "step": 18713 }, { "epoch": 0.46, "learning_rate": 1.1731589634820519e-05, "loss": 1.9692, "step": 18714 }, { "epoch": 0.46, "learning_rate": 1.1730803357854013e-05, "loss": 2.0638, "step": 18715 }, { "epoch": 0.46, "learning_rate": 1.173001706985653e-05, "loss": 2.0904, "step": 18716 }, { "epoch": 0.46, "learning_rate": 1.1729230770833084e-05, "loss": 2.1878, "step": 18717 }, { "epoch": 0.46, "learning_rate": 1.1728444460788688e-05, "loss": 2.124, "step": 18718 }, { "epoch": 0.46, "learning_rate": 1.1727658139728355e-05, "loss": 2.1398, "step": 18719 }, { "epoch": 0.46, "learning_rate": 1.1726871807657088e-05, "loss": 2.0033, "step": 18720 }, { "epoch": 0.46, "learning_rate": 1.1726085464579906e-05, "loss": 1.9737, "step": 18721 }, { "epoch": 0.46, "learning_rate": 1.172529911050182e-05, "loss": 1.9877, "step": 18722 }, { "epoch": 0.46, "learning_rate": 1.172451274542784e-05, "loss": 2.0135, "step": 18723 }, { "epoch": 0.46, "learning_rate": 1.172372636936298e-05, "loss": 2.2151, "step": 18724 }, { "epoch": 0.46, "learning_rate": 1.1722939982312246e-05, "loss": 1.9598, "step": 18725 }, { "epoch": 0.46, "learning_rate": 1.172215358428066e-05, "loss": 1.999, "step": 18726 }, { "epoch": 0.46, "learning_rate": 1.1721367175273227e-05, "loss": 2.1155, "step": 18727 }, { "epoch": 0.46, "learning_rate": 1.1720580755294957e-05, "loss": 2.0206, "step": 18728 }, { "epoch": 0.46, "learning_rate": 1.1719794324350863e-05, "loss": 2.1443, "step": 18729 }, { "epoch": 0.46, "learning_rate": 1.1719007882445962e-05, "loss": 2.0329, "step": 18730 }, { "epoch": 0.46, "learning_rate": 1.1718221429585263e-05, "loss": 2.0792, "step": 18731 }, { "epoch": 0.46, "learning_rate": 1.1717434965773782e-05, "loss": 2.1464, "step": 18732 }, { "epoch": 0.46, "learning_rate": 1.1716648491016524e-05, "loss": 2.1136, "step": 18733 }, { "epoch": 0.46, "learning_rate": 1.171586200531851e-05, "loss": 2.0606, "step": 18734 }, { "epoch": 0.46, "learning_rate": 1.1715075508684743e-05, "loss": 2.0109, "step": 18735 }, { "epoch": 0.46, "learning_rate": 1.1714289001120244e-05, "loss": 1.9205, "step": 18736 }, { "epoch": 0.46, "learning_rate": 1.1713502482630023e-05, "loss": 2.1309, "step": 18737 }, { "epoch": 0.46, "learning_rate": 1.1712715953219091e-05, "loss": 2.0541, "step": 18738 }, { "epoch": 0.46, "learning_rate": 1.1711929412892464e-05, "loss": 2.0752, "step": 18739 }, { "epoch": 0.46, "learning_rate": 1.1711142861655149e-05, "loss": 2.0064, "step": 18740 }, { "epoch": 0.46, "learning_rate": 1.1710356299512167e-05, "loss": 1.8684, "step": 18741 }, { "epoch": 0.46, "learning_rate": 1.1709569726468526e-05, "loss": 2.0227, "step": 18742 }, { "epoch": 0.46, "learning_rate": 1.170878314252924e-05, "loss": 2.098, "step": 18743 }, { "epoch": 0.46, "learning_rate": 1.1707996547699323e-05, "loss": 2.1642, "step": 18744 }, { "epoch": 0.46, "learning_rate": 1.1707209941983784e-05, "loss": 2.1095, "step": 18745 }, { "epoch": 0.46, "learning_rate": 1.1706423325387642e-05, "loss": 2.2185, "step": 18746 }, { "epoch": 0.46, "learning_rate": 1.1705636697915906e-05, "loss": 2.1564, "step": 18747 }, { "epoch": 0.46, "learning_rate": 1.1704850059573594e-05, "loss": 2.0502, "step": 18748 }, { "epoch": 0.46, "learning_rate": 1.1704063410365714e-05, "loss": 2.2171, "step": 18749 }, { "epoch": 0.46, "learning_rate": 1.1703276750297285e-05, "loss": 2.0396, "step": 18750 }, { "epoch": 0.46, "learning_rate": 1.1702490079373319e-05, "loss": 2.027, "step": 18751 }, { "epoch": 0.46, "learning_rate": 1.1701703397598827e-05, "loss": 2.2598, "step": 18752 }, { "epoch": 0.46, "learning_rate": 1.1700916704978826e-05, "loss": 2.0381, "step": 18753 }, { "epoch": 0.46, "learning_rate": 1.1700130001518329e-05, "loss": 2.133, "step": 18754 }, { "epoch": 0.46, "learning_rate": 1.1699343287222346e-05, "loss": 2.1144, "step": 18755 }, { "epoch": 0.46, "learning_rate": 1.1698556562095897e-05, "loss": 2.042, "step": 18756 }, { "epoch": 0.46, "learning_rate": 1.1697769826143994e-05, "loss": 2.1146, "step": 18757 }, { "epoch": 0.46, "learning_rate": 1.1696983079371651e-05, "loss": 1.9901, "step": 18758 }, { "epoch": 0.46, "learning_rate": 1.1696196321783879e-05, "loss": 1.9041, "step": 18759 }, { "epoch": 0.46, "learning_rate": 1.1695409553385697e-05, "loss": 2.05, "step": 18760 }, { "epoch": 0.46, "learning_rate": 1.1694622774182118e-05, "loss": 2.1525, "step": 18761 }, { "epoch": 0.46, "learning_rate": 1.1693835984178155e-05, "loss": 2.2523, "step": 18762 }, { "epoch": 0.46, "learning_rate": 1.1693049183378824e-05, "loss": 1.8422, "step": 18763 }, { "epoch": 0.46, "learning_rate": 1.1692262371789136e-05, "loss": 1.9976, "step": 18764 }, { "epoch": 0.46, "learning_rate": 1.169147554941411e-05, "loss": 2.1694, "step": 18765 }, { "epoch": 0.46, "learning_rate": 1.169068871625876e-05, "loss": 2.0302, "step": 18766 }, { "epoch": 0.46, "learning_rate": 1.16899018723281e-05, "loss": 2.0891, "step": 18767 }, { "epoch": 0.46, "learning_rate": 1.1689115017627144e-05, "loss": 1.8976, "step": 18768 }, { "epoch": 0.46, "learning_rate": 1.1688328152160907e-05, "loss": 2.2355, "step": 18769 }, { "epoch": 0.46, "learning_rate": 1.1687541275934405e-05, "loss": 1.8716, "step": 18770 }, { "epoch": 0.46, "learning_rate": 1.1686754388952653e-05, "loss": 2.3229, "step": 18771 }, { "epoch": 0.46, "learning_rate": 1.1685967491220667e-05, "loss": 2.1912, "step": 18772 }, { "epoch": 0.46, "learning_rate": 1.1685180582743457e-05, "loss": 2.0901, "step": 18773 }, { "epoch": 0.46, "learning_rate": 1.1684393663526041e-05, "loss": 1.895, "step": 18774 }, { "epoch": 0.46, "learning_rate": 1.1683606733573439e-05, "loss": 2.0599, "step": 18775 }, { "epoch": 0.46, "learning_rate": 1.1682819792890664e-05, "loss": 2.0728, "step": 18776 }, { "epoch": 0.46, "learning_rate": 1.1682032841482727e-05, "loss": 1.9209, "step": 18777 }, { "epoch": 0.46, "learning_rate": 1.1681245879354646e-05, "loss": 2.1427, "step": 18778 }, { "epoch": 0.46, "learning_rate": 1.1680458906511438e-05, "loss": 1.914, "step": 18779 }, { "epoch": 0.46, "learning_rate": 1.1679671922958118e-05, "loss": 2.0835, "step": 18780 }, { "epoch": 0.46, "learning_rate": 1.1678884928699702e-05, "loss": 2.0715, "step": 18781 }, { "epoch": 0.46, "learning_rate": 1.1678097923741206e-05, "loss": 2.0749, "step": 18782 }, { "epoch": 0.46, "learning_rate": 1.167731090808764e-05, "loss": 2.0806, "step": 18783 }, { "epoch": 0.46, "learning_rate": 1.167652388174403e-05, "loss": 1.976, "step": 18784 }, { "epoch": 0.46, "learning_rate": 1.1675736844715386e-05, "loss": 2.0846, "step": 18785 }, { "epoch": 0.46, "learning_rate": 1.1674949797006725e-05, "loss": 2.2716, "step": 18786 }, { "epoch": 0.46, "learning_rate": 1.1674162738623066e-05, "loss": 2.213, "step": 18787 }, { "epoch": 0.46, "learning_rate": 1.1673375669569415e-05, "loss": 1.8511, "step": 18788 }, { "epoch": 0.46, "learning_rate": 1.1672588589850801e-05, "loss": 2.08, "step": 18789 }, { "epoch": 0.46, "learning_rate": 1.1671801499472236e-05, "loss": 1.9832, "step": 18790 }, { "epoch": 0.46, "learning_rate": 1.1671014398438729e-05, "loss": 2.1642, "step": 18791 }, { "epoch": 0.46, "learning_rate": 1.1670227286755308e-05, "loss": 1.8963, "step": 18792 }, { "epoch": 0.46, "learning_rate": 1.166944016442698e-05, "loss": 2.0397, "step": 18793 }, { "epoch": 0.46, "learning_rate": 1.166865303145877e-05, "loss": 2.0682, "step": 18794 }, { "epoch": 0.46, "learning_rate": 1.166786588785569e-05, "loss": 1.9047, "step": 18795 }, { "epoch": 0.46, "learning_rate": 1.1667078733622754e-05, "loss": 2.1475, "step": 18796 }, { "epoch": 0.46, "learning_rate": 1.1666291568764985e-05, "loss": 2.088, "step": 18797 }, { "epoch": 0.46, "learning_rate": 1.1665504393287395e-05, "loss": 2.0403, "step": 18798 }, { "epoch": 0.46, "learning_rate": 1.1664717207195003e-05, "loss": 1.9385, "step": 18799 }, { "epoch": 0.46, "learning_rate": 1.1663930010492826e-05, "loss": 1.9813, "step": 18800 }, { "epoch": 0.46, "learning_rate": 1.166314280318588e-05, "loss": 2.1394, "step": 18801 }, { "epoch": 0.46, "learning_rate": 1.1662355585279186e-05, "loss": 2.0438, "step": 18802 }, { "epoch": 0.46, "learning_rate": 1.1661568356777753e-05, "loss": 2.0109, "step": 18803 }, { "epoch": 0.46, "learning_rate": 1.1660781117686609e-05, "loss": 2.0629, "step": 18804 }, { "epoch": 0.46, "learning_rate": 1.1659993868010761e-05, "loss": 2.1784, "step": 18805 }, { "epoch": 0.46, "learning_rate": 1.1659206607755233e-05, "loss": 2.1289, "step": 18806 }, { "epoch": 0.46, "learning_rate": 1.165841933692504e-05, "loss": 2.0377, "step": 18807 }, { "epoch": 0.46, "learning_rate": 1.1657632055525201e-05, "loss": 1.9862, "step": 18808 }, { "epoch": 0.46, "learning_rate": 1.165684476356073e-05, "loss": 2.1576, "step": 18809 }, { "epoch": 0.46, "learning_rate": 1.1656057461036651e-05, "loss": 2.154, "step": 18810 }, { "epoch": 0.46, "learning_rate": 1.1655270147957978e-05, "loss": 2.1058, "step": 18811 }, { "epoch": 0.46, "learning_rate": 1.1654482824329727e-05, "loss": 2.0726, "step": 18812 }, { "epoch": 0.46, "learning_rate": 1.165369549015692e-05, "loss": 1.9514, "step": 18813 }, { "epoch": 0.46, "learning_rate": 1.165290814544457e-05, "loss": 2.1235, "step": 18814 }, { "epoch": 0.46, "learning_rate": 1.16521207901977e-05, "loss": 2.0104, "step": 18815 }, { "epoch": 0.46, "learning_rate": 1.1651333424421325e-05, "loss": 2.0456, "step": 18816 }, { "epoch": 0.46, "learning_rate": 1.1650546048120464e-05, "loss": 2.0528, "step": 18817 }, { "epoch": 0.46, "learning_rate": 1.1649758661300133e-05, "loss": 2.0612, "step": 18818 }, { "epoch": 0.46, "learning_rate": 1.1648971263965354e-05, "loss": 1.8057, "step": 18819 }, { "epoch": 0.46, "learning_rate": 1.1648183856121145e-05, "loss": 1.9833, "step": 18820 }, { "epoch": 0.46, "learning_rate": 1.1647396437772523e-05, "loss": 1.9586, "step": 18821 }, { "epoch": 0.46, "learning_rate": 1.1646609008924506e-05, "loss": 2.003, "step": 18822 }, { "epoch": 0.46, "learning_rate": 1.1645821569582112e-05, "loss": 2.0751, "step": 18823 }, { "epoch": 0.46, "learning_rate": 1.1645034119750363e-05, "loss": 1.8629, "step": 18824 }, { "epoch": 0.46, "learning_rate": 1.1644246659434272e-05, "loss": 1.9987, "step": 18825 }, { "epoch": 0.46, "learning_rate": 1.1643459188638863e-05, "loss": 2.1068, "step": 18826 }, { "epoch": 0.46, "learning_rate": 1.1642671707369153e-05, "loss": 2.124, "step": 18827 }, { "epoch": 0.46, "learning_rate": 1.164188421563016e-05, "loss": 2.1362, "step": 18828 }, { "epoch": 0.46, "learning_rate": 1.1641096713426907e-05, "loss": 2.032, "step": 18829 }, { "epoch": 0.46, "learning_rate": 1.164030920076441e-05, "loss": 2.0435, "step": 18830 }, { "epoch": 0.46, "learning_rate": 1.1639521677647684e-05, "loss": 2.1201, "step": 18831 }, { "epoch": 0.46, "learning_rate": 1.1638734144081753e-05, "loss": 1.943, "step": 18832 }, { "epoch": 0.46, "learning_rate": 1.1637946600071638e-05, "loss": 2.0787, "step": 18833 }, { "epoch": 0.46, "learning_rate": 1.1637159045622352e-05, "loss": 2.1014, "step": 18834 }, { "epoch": 0.46, "learning_rate": 1.163637148073892e-05, "loss": 2.143, "step": 18835 }, { "epoch": 0.46, "learning_rate": 1.1635583905426358e-05, "loss": 1.98, "step": 18836 }, { "epoch": 0.46, "learning_rate": 1.1634796319689687e-05, "loss": 1.9124, "step": 18837 }, { "epoch": 0.46, "learning_rate": 1.163400872353393e-05, "loss": 2.1065, "step": 18838 }, { "epoch": 0.46, "learning_rate": 1.16332211169641e-05, "loss": 2.015, "step": 18839 }, { "epoch": 0.46, "learning_rate": 1.1632433499985218e-05, "loss": 2.0572, "step": 18840 }, { "epoch": 0.46, "learning_rate": 1.1631645872602306e-05, "loss": 1.8998, "step": 18841 }, { "epoch": 0.46, "learning_rate": 1.1630858234820383e-05, "loss": 2.0447, "step": 18842 }, { "epoch": 0.46, "learning_rate": 1.163007058664447e-05, "loss": 2.0215, "step": 18843 }, { "epoch": 0.46, "learning_rate": 1.1629282928079585e-05, "loss": 1.8837, "step": 18844 }, { "epoch": 0.46, "learning_rate": 1.1628495259130752e-05, "loss": 2.1009, "step": 18845 }, { "epoch": 0.46, "learning_rate": 1.1627707579802983e-05, "loss": 1.9247, "step": 18846 }, { "epoch": 0.46, "learning_rate": 1.1626919890101307e-05, "loss": 2.0383, "step": 18847 }, { "epoch": 0.46, "learning_rate": 1.1626132190030738e-05, "loss": 2.0577, "step": 18848 }, { "epoch": 0.46, "learning_rate": 1.1625344479596301e-05, "loss": 1.9019, "step": 18849 }, { "epoch": 0.46, "learning_rate": 1.1624556758803014e-05, "loss": 2.1049, "step": 18850 }, { "epoch": 0.46, "learning_rate": 1.1623769027655896e-05, "loss": 2.1375, "step": 18851 }, { "epoch": 0.46, "learning_rate": 1.162298128615997e-05, "loss": 2.061, "step": 18852 }, { "epoch": 0.46, "learning_rate": 1.162219353432025e-05, "loss": 1.9748, "step": 18853 }, { "epoch": 0.46, "learning_rate": 1.1621405772141767e-05, "loss": 1.8966, "step": 18854 }, { "epoch": 0.46, "learning_rate": 1.1620617999629538e-05, "loss": 2.06, "step": 18855 }, { "epoch": 0.46, "learning_rate": 1.1619830216788576e-05, "loss": 2.0314, "step": 18856 }, { "epoch": 0.46, "learning_rate": 1.1619042423623915e-05, "loss": 2.156, "step": 18857 }, { "epoch": 0.46, "learning_rate": 1.1618254620140565e-05, "loss": 1.8791, "step": 18858 }, { "epoch": 0.46, "learning_rate": 1.1617466806343553e-05, "loss": 2.0777, "step": 18859 }, { "epoch": 0.46, "learning_rate": 1.1616678982237896e-05, "loss": 2.0044, "step": 18860 }, { "epoch": 0.46, "learning_rate": 1.161589114782862e-05, "loss": 1.9795, "step": 18861 }, { "epoch": 0.46, "learning_rate": 1.1615103303120738e-05, "loss": 2.0524, "step": 18862 }, { "epoch": 0.46, "learning_rate": 1.161431544811928e-05, "loss": 2.0752, "step": 18863 }, { "epoch": 0.46, "learning_rate": 1.1613527582829265e-05, "loss": 2.0662, "step": 18864 }, { "epoch": 0.46, "learning_rate": 1.1612739707255709e-05, "loss": 2.0461, "step": 18865 }, { "epoch": 0.47, "learning_rate": 1.161195182140364e-05, "loss": 2.0354, "step": 18866 }, { "epoch": 0.47, "learning_rate": 1.1611163925278077e-05, "loss": 2.0882, "step": 18867 }, { "epoch": 0.47, "learning_rate": 1.161037601888404e-05, "loss": 1.9719, "step": 18868 }, { "epoch": 0.47, "learning_rate": 1.160958810222655e-05, "loss": 2.0351, "step": 18869 }, { "epoch": 0.47, "learning_rate": 1.1608800175310632e-05, "loss": 2.0813, "step": 18870 }, { "epoch": 0.47, "learning_rate": 1.1608012238141306e-05, "loss": 2.0011, "step": 18871 }, { "epoch": 0.47, "learning_rate": 1.1607224290723595e-05, "loss": 2.1367, "step": 18872 }, { "epoch": 0.47, "learning_rate": 1.1606436333062521e-05, "loss": 1.8928, "step": 18873 }, { "epoch": 0.47, "learning_rate": 1.1605648365163104e-05, "loss": 2.1156, "step": 18874 }, { "epoch": 0.47, "learning_rate": 1.1604860387030364e-05, "loss": 2.136, "step": 18875 }, { "epoch": 0.47, "learning_rate": 1.1604072398669328e-05, "loss": 1.9737, "step": 18876 }, { "epoch": 0.47, "learning_rate": 1.1603284400085016e-05, "loss": 1.9556, "step": 18877 }, { "epoch": 0.47, "learning_rate": 1.160249639128245e-05, "loss": 2.1077, "step": 18878 }, { "epoch": 0.47, "learning_rate": 1.1601708372266652e-05, "loss": 1.8751, "step": 18879 }, { "epoch": 0.47, "learning_rate": 1.160092034304264e-05, "loss": 2.0974, "step": 18880 }, { "epoch": 0.47, "learning_rate": 1.1600132303615448e-05, "loss": 2.0355, "step": 18881 }, { "epoch": 0.47, "learning_rate": 1.1599344253990088e-05, "loss": 2.1776, "step": 18882 }, { "epoch": 0.47, "learning_rate": 1.1598556194171589e-05, "loss": 2.0686, "step": 18883 }, { "epoch": 0.47, "learning_rate": 1.1597768124164967e-05, "loss": 2.2063, "step": 18884 }, { "epoch": 0.47, "learning_rate": 1.1596980043975248e-05, "loss": 1.9715, "step": 18885 }, { "epoch": 0.47, "learning_rate": 1.1596191953607456e-05, "loss": 1.8774, "step": 18886 }, { "epoch": 0.47, "learning_rate": 1.1595403853066611e-05, "loss": 2.187, "step": 18887 }, { "epoch": 0.47, "learning_rate": 1.1594615742357737e-05, "loss": 2.0404, "step": 18888 }, { "epoch": 0.47, "learning_rate": 1.159382762148586e-05, "loss": 2.1747, "step": 18889 }, { "epoch": 0.47, "learning_rate": 1.1593039490455997e-05, "loss": 1.9061, "step": 18890 }, { "epoch": 0.47, "learning_rate": 1.1592251349273179e-05, "loss": 2.0601, "step": 18891 }, { "epoch": 0.47, "learning_rate": 1.159146319794242e-05, "loss": 2.0368, "step": 18892 }, { "epoch": 0.47, "learning_rate": 1.159067503646875e-05, "loss": 2.0666, "step": 18893 }, { "epoch": 0.47, "learning_rate": 1.1589886864857188e-05, "loss": 2.0971, "step": 18894 }, { "epoch": 0.47, "learning_rate": 1.1589098683112758e-05, "loss": 2.0661, "step": 18895 }, { "epoch": 0.47, "learning_rate": 1.1588310491240484e-05, "loss": 1.9814, "step": 18896 }, { "epoch": 0.47, "learning_rate": 1.1587522289245393e-05, "loss": 2.1188, "step": 18897 }, { "epoch": 0.47, "learning_rate": 1.1586734077132502e-05, "loss": 2.0351, "step": 18898 }, { "epoch": 0.47, "learning_rate": 1.1585945854906837e-05, "loss": 2.0423, "step": 18899 }, { "epoch": 0.47, "learning_rate": 1.1585157622573427e-05, "loss": 2.0729, "step": 18900 }, { "epoch": 0.47, "learning_rate": 1.1584369380137286e-05, "loss": 2.1069, "step": 18901 }, { "epoch": 0.47, "learning_rate": 1.1583581127603445e-05, "loss": 2.0799, "step": 18902 }, { "epoch": 0.47, "learning_rate": 1.1582792864976925e-05, "loss": 1.8539, "step": 18903 }, { "epoch": 0.47, "learning_rate": 1.158200459226275e-05, "loss": 1.9512, "step": 18904 }, { "epoch": 0.47, "learning_rate": 1.1581216309465941e-05, "loss": 2.0768, "step": 18905 }, { "epoch": 0.47, "learning_rate": 1.1580428016591529e-05, "loss": 2.0305, "step": 18906 }, { "epoch": 0.47, "learning_rate": 1.1579639713644535e-05, "loss": 2.0125, "step": 18907 }, { "epoch": 0.47, "learning_rate": 1.1578851400629982e-05, "loss": 2.1296, "step": 18908 }, { "epoch": 0.47, "learning_rate": 1.1578063077552892e-05, "loss": 2.1002, "step": 18909 }, { "epoch": 0.47, "learning_rate": 1.1577274744418294e-05, "loss": 2.0402, "step": 18910 }, { "epoch": 0.47, "learning_rate": 1.1576486401231208e-05, "loss": 2.0029, "step": 18911 }, { "epoch": 0.47, "learning_rate": 1.1575698047996664e-05, "loss": 2.0141, "step": 18912 }, { "epoch": 0.47, "learning_rate": 1.1574909684719677e-05, "loss": 2.1457, "step": 18913 }, { "epoch": 0.47, "learning_rate": 1.1574121311405282e-05, "loss": 2.0441, "step": 18914 }, { "epoch": 0.47, "learning_rate": 1.1573332928058498e-05, "loss": 1.8129, "step": 18915 }, { "epoch": 0.47, "learning_rate": 1.1572544534684349e-05, "loss": 2.0657, "step": 18916 }, { "epoch": 0.47, "learning_rate": 1.1571756131287863e-05, "loss": 2.117, "step": 18917 }, { "epoch": 0.47, "learning_rate": 1.1570967717874065e-05, "loss": 1.918, "step": 18918 }, { "epoch": 0.47, "learning_rate": 1.1570179294447972e-05, "loss": 2.1513, "step": 18919 }, { "epoch": 0.47, "learning_rate": 1.1569390861014618e-05, "loss": 2.0815, "step": 18920 }, { "epoch": 0.47, "learning_rate": 1.1568602417579027e-05, "loss": 1.9807, "step": 18921 }, { "epoch": 0.47, "learning_rate": 1.1567813964146215e-05, "loss": 1.8892, "step": 18922 }, { "epoch": 0.47, "learning_rate": 1.1567025500721219e-05, "loss": 1.92, "step": 18923 }, { "epoch": 0.47, "learning_rate": 1.1566237027309054e-05, "loss": 2.1152, "step": 18924 }, { "epoch": 0.47, "learning_rate": 1.1565448543914757e-05, "loss": 2.1602, "step": 18925 }, { "epoch": 0.47, "learning_rate": 1.1564660050543342e-05, "loss": 1.9202, "step": 18926 }, { "epoch": 0.47, "learning_rate": 1.156387154719984e-05, "loss": 1.9648, "step": 18927 }, { "epoch": 0.47, "learning_rate": 1.1563083033889273e-05, "loss": 2.104, "step": 18928 }, { "epoch": 0.47, "learning_rate": 1.156229451061667e-05, "loss": 2.013, "step": 18929 }, { "epoch": 0.47, "learning_rate": 1.1561505977387055e-05, "loss": 2.032, "step": 18930 }, { "epoch": 0.47, "learning_rate": 1.1560717434205453e-05, "loss": 1.9952, "step": 18931 }, { "epoch": 0.47, "learning_rate": 1.155992888107689e-05, "loss": 2.0109, "step": 18932 }, { "epoch": 0.47, "learning_rate": 1.1559140318006389e-05, "loss": 1.9322, "step": 18933 }, { "epoch": 0.47, "learning_rate": 1.1558351744998984e-05, "loss": 2.0054, "step": 18934 }, { "epoch": 0.47, "learning_rate": 1.1557563162059693e-05, "loss": 2.2203, "step": 18935 }, { "epoch": 0.47, "learning_rate": 1.1556774569193547e-05, "loss": 2.0748, "step": 18936 }, { "epoch": 0.47, "learning_rate": 1.1555985966405568e-05, "loss": 2.0483, "step": 18937 }, { "epoch": 0.47, "learning_rate": 1.155519735370078e-05, "loss": 2.1503, "step": 18938 }, { "epoch": 0.47, "learning_rate": 1.1554408731084214e-05, "loss": 2.1099, "step": 18939 }, { "epoch": 0.47, "learning_rate": 1.1553620098560896e-05, "loss": 1.9764, "step": 18940 }, { "epoch": 0.47, "learning_rate": 1.1552831456135848e-05, "loss": 2.0531, "step": 18941 }, { "epoch": 0.47, "learning_rate": 1.1552042803814103e-05, "loss": 1.9478, "step": 18942 }, { "epoch": 0.47, "learning_rate": 1.1551254141600681e-05, "loss": 1.9973, "step": 18943 }, { "epoch": 0.47, "learning_rate": 1.1550465469500611e-05, "loss": 2.1193, "step": 18944 }, { "epoch": 0.47, "learning_rate": 1.1549676787518922e-05, "loss": 2.0296, "step": 18945 }, { "epoch": 0.47, "learning_rate": 1.1548888095660636e-05, "loss": 2.0179, "step": 18946 }, { "epoch": 0.47, "learning_rate": 1.1548099393930781e-05, "loss": 2.0879, "step": 18947 }, { "epoch": 0.47, "learning_rate": 1.1547310682334383e-05, "loss": 1.9848, "step": 18948 }, { "epoch": 0.47, "learning_rate": 1.1546521960876469e-05, "loss": 2.1369, "step": 18949 }, { "epoch": 0.47, "learning_rate": 1.1545733229562069e-05, "loss": 2.0411, "step": 18950 }, { "epoch": 0.47, "learning_rate": 1.1544944488396208e-05, "loss": 2.1075, "step": 18951 }, { "epoch": 0.47, "learning_rate": 1.1544155737383911e-05, "loss": 2.1433, "step": 18952 }, { "epoch": 0.47, "learning_rate": 1.1543366976530204e-05, "loss": 2.2051, "step": 18953 }, { "epoch": 0.47, "learning_rate": 1.1542578205840121e-05, "loss": 2.1493, "step": 18954 }, { "epoch": 0.47, "learning_rate": 1.1541789425318682e-05, "loss": 2.0012, "step": 18955 }, { "epoch": 0.47, "learning_rate": 1.1541000634970915e-05, "loss": 1.8295, "step": 18956 }, { "epoch": 0.47, "learning_rate": 1.1540211834801851e-05, "loss": 2.071, "step": 18957 }, { "epoch": 0.47, "learning_rate": 1.1539423024816511e-05, "loss": 1.945, "step": 18958 }, { "epoch": 0.47, "learning_rate": 1.1538634205019932e-05, "loss": 2.0742, "step": 18959 }, { "epoch": 0.47, "learning_rate": 1.1537845375417133e-05, "loss": 2.1289, "step": 18960 }, { "epoch": 0.47, "learning_rate": 1.1537056536013145e-05, "loss": 2.0177, "step": 18961 }, { "epoch": 0.47, "learning_rate": 1.1536267686812994e-05, "loss": 1.999, "step": 18962 }, { "epoch": 0.47, "learning_rate": 1.153547882782171e-05, "loss": 2.0435, "step": 18963 }, { "epoch": 0.47, "learning_rate": 1.1534689959044318e-05, "loss": 2.0049, "step": 18964 }, { "epoch": 0.47, "learning_rate": 1.1533901080485847e-05, "loss": 2.2286, "step": 18965 }, { "epoch": 0.47, "learning_rate": 1.1533112192151324e-05, "loss": 2.1616, "step": 18966 }, { "epoch": 0.47, "learning_rate": 1.1532323294045776e-05, "loss": 1.9727, "step": 18967 }, { "epoch": 0.47, "learning_rate": 1.1531534386174233e-05, "loss": 2.0004, "step": 18968 }, { "epoch": 0.47, "learning_rate": 1.1530745468541724e-05, "loss": 1.9307, "step": 18969 }, { "epoch": 0.47, "learning_rate": 1.1529956541153276e-05, "loss": 2.2117, "step": 18970 }, { "epoch": 0.47, "learning_rate": 1.1529167604013914e-05, "loss": 2.0306, "step": 18971 }, { "epoch": 0.47, "learning_rate": 1.1528378657128666e-05, "loss": 2.14, "step": 18972 }, { "epoch": 0.47, "learning_rate": 1.1527589700502569e-05, "loss": 2.1405, "step": 18973 }, { "epoch": 0.47, "learning_rate": 1.1526800734140641e-05, "loss": 2.2267, "step": 18974 }, { "epoch": 0.47, "learning_rate": 1.1526011758047914e-05, "loss": 1.8874, "step": 18975 }, { "epoch": 0.47, "learning_rate": 1.1525222772229419e-05, "loss": 2.036, "step": 18976 }, { "epoch": 0.47, "learning_rate": 1.1524433776690181e-05, "loss": 2.156, "step": 18977 }, { "epoch": 0.47, "learning_rate": 1.152364477143523e-05, "loss": 2.0231, "step": 18978 }, { "epoch": 0.47, "learning_rate": 1.1522855756469595e-05, "loss": 1.9867, "step": 18979 }, { "epoch": 0.47, "learning_rate": 1.1522066731798302e-05, "loss": 1.9833, "step": 18980 }, { "epoch": 0.47, "learning_rate": 1.1521277697426385e-05, "loss": 1.9636, "step": 18981 }, { "epoch": 0.47, "learning_rate": 1.1520488653358865e-05, "loss": 1.9335, "step": 18982 }, { "epoch": 0.47, "learning_rate": 1.151969959960078e-05, "loss": 2.0605, "step": 18983 }, { "epoch": 0.47, "learning_rate": 1.1518910536157151e-05, "loss": 1.963, "step": 18984 }, { "epoch": 0.47, "learning_rate": 1.1518121463033012e-05, "loss": 2.1576, "step": 18985 }, { "epoch": 0.47, "learning_rate": 1.1517332380233386e-05, "loss": 2.1176, "step": 18986 }, { "epoch": 0.47, "learning_rate": 1.1516543287763311e-05, "loss": 1.9646, "step": 18987 }, { "epoch": 0.47, "learning_rate": 1.151575418562781e-05, "loss": 1.9779, "step": 18988 }, { "epoch": 0.47, "learning_rate": 1.1514965073831915e-05, "loss": 2.232, "step": 18989 }, { "epoch": 0.47, "learning_rate": 1.1514175952380653e-05, "loss": 1.8371, "step": 18990 }, { "epoch": 0.47, "learning_rate": 1.1513386821279052e-05, "loss": 2.1144, "step": 18991 }, { "epoch": 0.47, "learning_rate": 1.1512597680532146e-05, "loss": 1.9301, "step": 18992 }, { "epoch": 0.47, "learning_rate": 1.151180853014496e-05, "loss": 1.9447, "step": 18993 }, { "epoch": 0.47, "learning_rate": 1.1511019370122528e-05, "loss": 2.1777, "step": 18994 }, { "epoch": 0.47, "learning_rate": 1.1510230200469877e-05, "loss": 2.0572, "step": 18995 }, { "epoch": 0.47, "learning_rate": 1.1509441021192033e-05, "loss": 2.0726, "step": 18996 }, { "epoch": 0.47, "learning_rate": 1.1508651832294033e-05, "loss": 2.1238, "step": 18997 }, { "epoch": 0.47, "learning_rate": 1.1507862633780902e-05, "loss": 2.04, "step": 18998 }, { "epoch": 0.47, "learning_rate": 1.1507073425657673e-05, "loss": 2.0602, "step": 18999 }, { "epoch": 0.47, "learning_rate": 1.1506284207929372e-05, "loss": 2.006, "step": 19000 }, { "epoch": 0.47, "learning_rate": 1.1505494980601031e-05, "loss": 2.2174, "step": 19001 }, { "epoch": 0.47, "learning_rate": 1.1504705743677678e-05, "loss": 2.1177, "step": 19002 }, { "epoch": 0.47, "learning_rate": 1.1503916497164347e-05, "loss": 2.1244, "step": 19003 }, { "epoch": 0.47, "learning_rate": 1.1503127241066067e-05, "loss": 1.9423, "step": 19004 }, { "epoch": 0.47, "learning_rate": 1.1502337975387868e-05, "loss": 2.0332, "step": 19005 }, { "epoch": 0.47, "learning_rate": 1.1501548700134774e-05, "loss": 1.961, "step": 19006 }, { "epoch": 0.47, "learning_rate": 1.1500759415311826e-05, "loss": 2.0931, "step": 19007 }, { "epoch": 0.47, "learning_rate": 1.149997012092405e-05, "loss": 2.0304, "step": 19008 }, { "epoch": 0.47, "learning_rate": 1.1499180816976475e-05, "loss": 2.2138, "step": 19009 }, { "epoch": 0.47, "learning_rate": 1.1498391503474126e-05, "loss": 2.1243, "step": 19010 }, { "epoch": 0.47, "learning_rate": 1.1497602180422043e-05, "loss": 2.0412, "step": 19011 }, { "epoch": 0.47, "learning_rate": 1.1496812847825256e-05, "loss": 2.055, "step": 19012 }, { "epoch": 0.47, "learning_rate": 1.149602350568879e-05, "loss": 2.0452, "step": 19013 }, { "epoch": 0.47, "learning_rate": 1.1495234154017682e-05, "loss": 2.0789, "step": 19014 }, { "epoch": 0.47, "learning_rate": 1.1494444792816957e-05, "loss": 2.2185, "step": 19015 }, { "epoch": 0.47, "learning_rate": 1.1493655422091647e-05, "loss": 2.008, "step": 19016 }, { "epoch": 0.47, "learning_rate": 1.1492866041846789e-05, "loss": 2.032, "step": 19017 }, { "epoch": 0.47, "learning_rate": 1.1492076652087405e-05, "loss": 2.1045, "step": 19018 }, { "epoch": 0.47, "learning_rate": 1.149128725281853e-05, "loss": 1.9983, "step": 19019 }, { "epoch": 0.47, "learning_rate": 1.1490497844045194e-05, "loss": 2.1237, "step": 19020 }, { "epoch": 0.47, "learning_rate": 1.1489708425772434e-05, "loss": 2.1604, "step": 19021 }, { "epoch": 0.47, "learning_rate": 1.1488918998005274e-05, "loss": 1.8754, "step": 19022 }, { "epoch": 0.47, "learning_rate": 1.148812956074875e-05, "loss": 1.9104, "step": 19023 }, { "epoch": 0.47, "learning_rate": 1.1487340114007892e-05, "loss": 1.8695, "step": 19024 }, { "epoch": 0.47, "learning_rate": 1.1486550657787727e-05, "loss": 1.9665, "step": 19025 }, { "epoch": 0.47, "learning_rate": 1.1485761192093292e-05, "loss": 1.9897, "step": 19026 }, { "epoch": 0.47, "learning_rate": 1.1484971716929618e-05, "loss": 1.9317, "step": 19027 }, { "epoch": 0.47, "learning_rate": 1.1484182232301733e-05, "loss": 2.0035, "step": 19028 }, { "epoch": 0.47, "learning_rate": 1.1483392738214674e-05, "loss": 2.0205, "step": 19029 }, { "epoch": 0.47, "learning_rate": 1.1482603234673464e-05, "loss": 2.1418, "step": 19030 }, { "epoch": 0.47, "learning_rate": 1.1481813721683146e-05, "loss": 2.0995, "step": 19031 }, { "epoch": 0.47, "learning_rate": 1.1481024199248744e-05, "loss": 1.9838, "step": 19032 }, { "epoch": 0.47, "learning_rate": 1.1480234667375294e-05, "loss": 2.0373, "step": 19033 }, { "epoch": 0.47, "learning_rate": 1.1479445126067825e-05, "loss": 2.0254, "step": 19034 }, { "epoch": 0.47, "learning_rate": 1.147865557533137e-05, "loss": 1.912, "step": 19035 }, { "epoch": 0.47, "learning_rate": 1.1477866015170959e-05, "loss": 2.0549, "step": 19036 }, { "epoch": 0.47, "learning_rate": 1.1477076445591628e-05, "loss": 2.0292, "step": 19037 }, { "epoch": 0.47, "learning_rate": 1.1476286866598408e-05, "loss": 1.7885, "step": 19038 }, { "epoch": 0.47, "learning_rate": 1.147549727819633e-05, "loss": 1.9563, "step": 19039 }, { "epoch": 0.47, "learning_rate": 1.1474707680390427e-05, "loss": 2.0921, "step": 19040 }, { "epoch": 0.47, "learning_rate": 1.1473918073185732e-05, "loss": 2.0249, "step": 19041 }, { "epoch": 0.47, "learning_rate": 1.1473128456587276e-05, "loss": 2.0856, "step": 19042 }, { "epoch": 0.47, "learning_rate": 1.1472338830600094e-05, "loss": 2.2823, "step": 19043 }, { "epoch": 0.47, "learning_rate": 1.1471549195229215e-05, "loss": 2.1051, "step": 19044 }, { "epoch": 0.47, "learning_rate": 1.1470759550479673e-05, "loss": 2.079, "step": 19045 }, { "epoch": 0.47, "learning_rate": 1.1469969896356501e-05, "loss": 2.1085, "step": 19046 }, { "epoch": 0.47, "learning_rate": 1.1469180232864735e-05, "loss": 2.1174, "step": 19047 }, { "epoch": 0.47, "learning_rate": 1.1468390560009402e-05, "loss": 2.1879, "step": 19048 }, { "epoch": 0.47, "learning_rate": 1.1467600877795536e-05, "loss": 2.0323, "step": 19049 }, { "epoch": 0.47, "learning_rate": 1.1466811186228174e-05, "loss": 2.0563, "step": 19050 }, { "epoch": 0.47, "learning_rate": 1.1466021485312346e-05, "loss": 1.9262, "step": 19051 }, { "epoch": 0.47, "learning_rate": 1.1465231775053085e-05, "loss": 1.8018, "step": 19052 }, { "epoch": 0.47, "learning_rate": 1.1464442055455425e-05, "loss": 1.9751, "step": 19053 }, { "epoch": 0.47, "learning_rate": 1.1463652326524397e-05, "loss": 1.9411, "step": 19054 }, { "epoch": 0.47, "learning_rate": 1.1462862588265033e-05, "loss": 2.1195, "step": 19055 }, { "epoch": 0.47, "learning_rate": 1.1462072840682374e-05, "loss": 2.141, "step": 19056 }, { "epoch": 0.47, "learning_rate": 1.1461283083781447e-05, "loss": 1.8911, "step": 19057 }, { "epoch": 0.47, "learning_rate": 1.1460493317567288e-05, "loss": 1.9926, "step": 19058 }, { "epoch": 0.47, "learning_rate": 1.1459703542044926e-05, "loss": 2.2454, "step": 19059 }, { "epoch": 0.47, "learning_rate": 1.1458913757219398e-05, "loss": 1.8994, "step": 19060 }, { "epoch": 0.47, "learning_rate": 1.145812396309574e-05, "loss": 2.0534, "step": 19061 }, { "epoch": 0.47, "learning_rate": 1.1457334159678981e-05, "loss": 2.0126, "step": 19062 }, { "epoch": 0.47, "learning_rate": 1.1456544346974157e-05, "loss": 1.895, "step": 19063 }, { "epoch": 0.47, "learning_rate": 1.1455754524986298e-05, "loss": 1.9958, "step": 19064 }, { "epoch": 0.47, "learning_rate": 1.1454964693720446e-05, "loss": 1.9832, "step": 19065 }, { "epoch": 0.47, "learning_rate": 1.1454174853181627e-05, "loss": 1.8272, "step": 19066 }, { "epoch": 0.47, "learning_rate": 1.145338500337488e-05, "loss": 1.8662, "step": 19067 }, { "epoch": 0.47, "learning_rate": 1.1452595144305235e-05, "loss": 2.0649, "step": 19068 }, { "epoch": 0.47, "learning_rate": 1.1451805275977725e-05, "loss": 2.1582, "step": 19069 }, { "epoch": 0.47, "learning_rate": 1.1451015398397392e-05, "loss": 2.0729, "step": 19070 }, { "epoch": 0.47, "learning_rate": 1.1450225511569261e-05, "loss": 1.8709, "step": 19071 }, { "epoch": 0.47, "learning_rate": 1.1449435615498371e-05, "loss": 2.0072, "step": 19072 }, { "epoch": 0.47, "learning_rate": 1.1448645710189756e-05, "loss": 2.3091, "step": 19073 }, { "epoch": 0.47, "learning_rate": 1.144785579564845e-05, "loss": 2.0367, "step": 19074 }, { "epoch": 0.47, "learning_rate": 1.1447065871879488e-05, "loss": 1.9452, "step": 19075 }, { "epoch": 0.47, "learning_rate": 1.1446275938887905e-05, "loss": 1.9794, "step": 19076 }, { "epoch": 0.47, "learning_rate": 1.1445485996678732e-05, "loss": 2.0836, "step": 19077 }, { "epoch": 0.47, "learning_rate": 1.1444696045257004e-05, "loss": 2.1364, "step": 19078 }, { "epoch": 0.47, "learning_rate": 1.144390608462776e-05, "loss": 1.9387, "step": 19079 }, { "epoch": 0.47, "learning_rate": 1.1443116114796032e-05, "loss": 1.9924, "step": 19080 }, { "epoch": 0.47, "learning_rate": 1.1442326135766852e-05, "loss": 1.9786, "step": 19081 }, { "epoch": 0.47, "learning_rate": 1.1441536147545261e-05, "loss": 2.0544, "step": 19082 }, { "epoch": 0.47, "learning_rate": 1.1440746150136286e-05, "loss": 1.9543, "step": 19083 }, { "epoch": 0.47, "learning_rate": 1.1439956143544968e-05, "loss": 2.2045, "step": 19084 }, { "epoch": 0.47, "learning_rate": 1.1439166127776342e-05, "loss": 2.0567, "step": 19085 }, { "epoch": 0.47, "learning_rate": 1.1438376102835443e-05, "loss": 2.0185, "step": 19086 }, { "epoch": 0.47, "learning_rate": 1.14375860687273e-05, "loss": 2.0663, "step": 19087 }, { "epoch": 0.47, "learning_rate": 1.1436796025456953e-05, "loss": 1.9063, "step": 19088 }, { "epoch": 0.47, "learning_rate": 1.1436005973029438e-05, "loss": 1.8694, "step": 19089 }, { "epoch": 0.47, "learning_rate": 1.1435215911449789e-05, "loss": 1.9685, "step": 19090 }, { "epoch": 0.47, "learning_rate": 1.143442584072304e-05, "loss": 2.0936, "step": 19091 }, { "epoch": 0.47, "learning_rate": 1.1433635760854229e-05, "loss": 2.0387, "step": 19092 }, { "epoch": 0.47, "learning_rate": 1.143284567184839e-05, "loss": 2.1728, "step": 19093 }, { "epoch": 0.47, "learning_rate": 1.1432055573710559e-05, "loss": 2.0119, "step": 19094 }, { "epoch": 0.47, "learning_rate": 1.1431265466445771e-05, "loss": 2.0657, "step": 19095 }, { "epoch": 0.47, "learning_rate": 1.1430475350059061e-05, "loss": 2.1977, "step": 19096 }, { "epoch": 0.47, "learning_rate": 1.1429685224555465e-05, "loss": 2.1215, "step": 19097 }, { "epoch": 0.47, "learning_rate": 1.142889508994002e-05, "loss": 1.9925, "step": 19098 }, { "epoch": 0.47, "learning_rate": 1.142810494621776e-05, "loss": 2.0142, "step": 19099 }, { "epoch": 0.47, "learning_rate": 1.1427314793393722e-05, "loss": 2.1652, "step": 19100 }, { "epoch": 0.47, "learning_rate": 1.1426524631472944e-05, "loss": 2.0028, "step": 19101 }, { "epoch": 0.47, "learning_rate": 1.1425734460460455e-05, "loss": 1.8927, "step": 19102 }, { "epoch": 0.47, "learning_rate": 1.1424944280361299e-05, "loss": 2.0731, "step": 19103 }, { "epoch": 0.47, "learning_rate": 1.1424154091180509e-05, "loss": 1.8724, "step": 19104 }, { "epoch": 0.47, "learning_rate": 1.142336389292312e-05, "loss": 2.1265, "step": 19105 }, { "epoch": 0.47, "learning_rate": 1.1422573685594167e-05, "loss": 1.9746, "step": 19106 }, { "epoch": 0.47, "learning_rate": 1.1421783469198693e-05, "loss": 2.0216, "step": 19107 }, { "epoch": 0.47, "learning_rate": 1.1420993243741725e-05, "loss": 2.08, "step": 19108 }, { "epoch": 0.47, "learning_rate": 1.1420203009228305e-05, "loss": 2.0094, "step": 19109 }, { "epoch": 0.47, "learning_rate": 1.1419412765663472e-05, "loss": 2.0703, "step": 19110 }, { "epoch": 0.47, "learning_rate": 1.1418622513052255e-05, "loss": 2.0025, "step": 19111 }, { "epoch": 0.47, "learning_rate": 1.1417832251399694e-05, "loss": 2.008, "step": 19112 }, { "epoch": 0.47, "learning_rate": 1.141704198071083e-05, "loss": 1.8542, "step": 19113 }, { "epoch": 0.47, "learning_rate": 1.1416251700990693e-05, "loss": 1.9552, "step": 19114 }, { "epoch": 0.47, "learning_rate": 1.1415461412244321e-05, "loss": 2.0875, "step": 19115 }, { "epoch": 0.47, "learning_rate": 1.1414671114476754e-05, "loss": 2.2924, "step": 19116 }, { "epoch": 0.47, "learning_rate": 1.1413880807693025e-05, "loss": 1.9042, "step": 19117 }, { "epoch": 0.47, "learning_rate": 1.1413090491898175e-05, "loss": 2.0607, "step": 19118 }, { "epoch": 0.47, "learning_rate": 1.1412300167097241e-05, "loss": 2.1527, "step": 19119 }, { "epoch": 0.47, "learning_rate": 1.1411509833295256e-05, "loss": 2.091, "step": 19120 }, { "epoch": 0.47, "learning_rate": 1.1410719490497257e-05, "loss": 2.0049, "step": 19121 }, { "epoch": 0.47, "learning_rate": 1.1409929138708282e-05, "loss": 1.9016, "step": 19122 }, { "epoch": 0.47, "learning_rate": 1.1409138777933373e-05, "loss": 2.025, "step": 19123 }, { "epoch": 0.47, "learning_rate": 1.140834840817756e-05, "loss": 2.0919, "step": 19124 }, { "epoch": 0.47, "learning_rate": 1.1407558029445887e-05, "loss": 2.2145, "step": 19125 }, { "epoch": 0.47, "learning_rate": 1.1406767641743386e-05, "loss": 1.9848, "step": 19126 }, { "epoch": 0.47, "learning_rate": 1.1405977245075093e-05, "loss": 1.889, "step": 19127 }, { "epoch": 0.47, "learning_rate": 1.1405186839446053e-05, "loss": 2.0193, "step": 19128 }, { "epoch": 0.47, "learning_rate": 1.1404396424861299e-05, "loss": 1.983, "step": 19129 }, { "epoch": 0.47, "learning_rate": 1.1403606001325871e-05, "loss": 2.0697, "step": 19130 }, { "epoch": 0.47, "learning_rate": 1.1402815568844798e-05, "loss": 1.9626, "step": 19131 }, { "epoch": 0.47, "learning_rate": 1.140202512742313e-05, "loss": 1.9919, "step": 19132 }, { "epoch": 0.47, "learning_rate": 1.1401234677065896e-05, "loss": 2.172, "step": 19133 }, { "epoch": 0.47, "learning_rate": 1.1400444217778138e-05, "loss": 2.0598, "step": 19134 }, { "epoch": 0.47, "learning_rate": 1.1399653749564893e-05, "loss": 2.1063, "step": 19135 }, { "epoch": 0.47, "learning_rate": 1.1398863272431197e-05, "loss": 2.2156, "step": 19136 }, { "epoch": 0.47, "learning_rate": 1.139807278638209e-05, "loss": 2.0814, "step": 19137 }, { "epoch": 0.47, "learning_rate": 1.1397282291422611e-05, "loss": 1.919, "step": 19138 }, { "epoch": 0.47, "learning_rate": 1.1396491787557797e-05, "loss": 1.9976, "step": 19139 }, { "epoch": 0.47, "learning_rate": 1.1395701274792685e-05, "loss": 2.0926, "step": 19140 }, { "epoch": 0.47, "learning_rate": 1.1394910753132312e-05, "loss": 2.0894, "step": 19141 }, { "epoch": 0.47, "learning_rate": 1.139412022258172e-05, "loss": 1.9426, "step": 19142 }, { "epoch": 0.47, "learning_rate": 1.1393329683145944e-05, "loss": 1.9778, "step": 19143 }, { "epoch": 0.47, "learning_rate": 1.1392539134830025e-05, "loss": 1.9979, "step": 19144 }, { "epoch": 0.47, "learning_rate": 1.1391748577638998e-05, "loss": 1.8369, "step": 19145 }, { "epoch": 0.47, "learning_rate": 1.1390958011577907e-05, "loss": 1.9196, "step": 19146 }, { "epoch": 0.47, "learning_rate": 1.1390167436651787e-05, "loss": 2.1999, "step": 19147 }, { "epoch": 0.47, "learning_rate": 1.1389376852865675e-05, "loss": 1.8804, "step": 19148 }, { "epoch": 0.47, "learning_rate": 1.1388586260224615e-05, "loss": 2.2339, "step": 19149 }, { "epoch": 0.47, "learning_rate": 1.1387795658733638e-05, "loss": 2.1039, "step": 19150 }, { "epoch": 0.47, "learning_rate": 1.1387005048397787e-05, "loss": 2.0619, "step": 19151 }, { "epoch": 0.47, "learning_rate": 1.1386214429222102e-05, "loss": 2.031, "step": 19152 }, { "epoch": 0.47, "learning_rate": 1.1385423801211623e-05, "loss": 1.8514, "step": 19153 }, { "epoch": 0.47, "learning_rate": 1.1384633164371387e-05, "loss": 2.0535, "step": 19154 }, { "epoch": 0.47, "learning_rate": 1.138384251870643e-05, "loss": 2.0445, "step": 19155 }, { "epoch": 0.47, "learning_rate": 1.1383051864221792e-05, "loss": 2.1135, "step": 19156 }, { "epoch": 0.47, "learning_rate": 1.1382261200922515e-05, "loss": 2.1225, "step": 19157 }, { "epoch": 0.47, "learning_rate": 1.1381470528813639e-05, "loss": 2.0204, "step": 19158 }, { "epoch": 0.47, "learning_rate": 1.1380679847900197e-05, "loss": 1.9673, "step": 19159 }, { "epoch": 0.47, "learning_rate": 1.1379889158187239e-05, "loss": 1.8312, "step": 19160 }, { "epoch": 0.47, "learning_rate": 1.1379098459679792e-05, "loss": 2.0773, "step": 19161 }, { "epoch": 0.47, "learning_rate": 1.1378307752382902e-05, "loss": 2.093, "step": 19162 }, { "epoch": 0.47, "learning_rate": 1.137751703630161e-05, "loss": 1.9735, "step": 19163 }, { "epoch": 0.47, "learning_rate": 1.1376726311440954e-05, "loss": 1.9232, "step": 19164 }, { "epoch": 0.47, "learning_rate": 1.1375935577805968e-05, "loss": 2.1838, "step": 19165 }, { "epoch": 0.47, "learning_rate": 1.1375144835401697e-05, "loss": 2.2741, "step": 19166 }, { "epoch": 0.47, "learning_rate": 1.1374354084233183e-05, "loss": 1.9692, "step": 19167 }, { "epoch": 0.47, "learning_rate": 1.1373563324305459e-05, "loss": 2.18, "step": 19168 }, { "epoch": 0.47, "learning_rate": 1.137277255562357e-05, "loss": 1.8528, "step": 19169 }, { "epoch": 0.47, "learning_rate": 1.1371981778192554e-05, "loss": 2.1495, "step": 19170 }, { "epoch": 0.47, "learning_rate": 1.137119099201745e-05, "loss": 2.028, "step": 19171 }, { "epoch": 0.47, "learning_rate": 1.1370400197103302e-05, "loss": 2.1744, "step": 19172 }, { "epoch": 0.47, "learning_rate": 1.1369609393455146e-05, "loss": 2.1362, "step": 19173 }, { "epoch": 0.47, "learning_rate": 1.136881858107802e-05, "loss": 2.1322, "step": 19174 }, { "epoch": 0.47, "learning_rate": 1.136802775997697e-05, "loss": 2.0129, "step": 19175 }, { "epoch": 0.47, "learning_rate": 1.1367236930157032e-05, "loss": 1.9645, "step": 19176 }, { "epoch": 0.47, "learning_rate": 1.1366446091623247e-05, "loss": 2.1134, "step": 19177 }, { "epoch": 0.47, "learning_rate": 1.1365655244380656e-05, "loss": 1.9551, "step": 19178 }, { "epoch": 0.47, "learning_rate": 1.13648643884343e-05, "loss": 2.0244, "step": 19179 }, { "epoch": 0.47, "learning_rate": 1.1364073523789218e-05, "loss": 2.0789, "step": 19180 }, { "epoch": 0.47, "learning_rate": 1.1363282650450452e-05, "loss": 2.0525, "step": 19181 }, { "epoch": 0.47, "learning_rate": 1.136249176842304e-05, "loss": 2.0845, "step": 19182 }, { "epoch": 0.47, "learning_rate": 1.1361700877712025e-05, "loss": 2.1041, "step": 19183 }, { "epoch": 0.47, "learning_rate": 1.1360909978322447e-05, "loss": 1.9647, "step": 19184 }, { "epoch": 0.47, "learning_rate": 1.1360119070259344e-05, "loss": 1.9962, "step": 19185 }, { "epoch": 0.47, "learning_rate": 1.1359328153527757e-05, "loss": 2.117, "step": 19186 }, { "epoch": 0.47, "learning_rate": 1.1358537228132733e-05, "loss": 1.9497, "step": 19187 }, { "epoch": 0.47, "learning_rate": 1.1357746294079308e-05, "loss": 2.0402, "step": 19188 }, { "epoch": 0.47, "learning_rate": 1.1356955351372523e-05, "loss": 1.9981, "step": 19189 }, { "epoch": 0.47, "learning_rate": 1.1356164400017418e-05, "loss": 2.1754, "step": 19190 }, { "epoch": 0.47, "learning_rate": 1.1355373440019034e-05, "loss": 2.1873, "step": 19191 }, { "epoch": 0.47, "learning_rate": 1.1354582471382418e-05, "loss": 2.0852, "step": 19192 }, { "epoch": 0.47, "learning_rate": 1.1353791494112604e-05, "loss": 1.8971, "step": 19193 }, { "epoch": 0.47, "learning_rate": 1.1353000508214635e-05, "loss": 2.1862, "step": 19194 }, { "epoch": 0.47, "learning_rate": 1.1352209513693554e-05, "loss": 2.0735, "step": 19195 }, { "epoch": 0.47, "learning_rate": 1.13514185105544e-05, "loss": 1.9361, "step": 19196 }, { "epoch": 0.47, "learning_rate": 1.1350627498802215e-05, "loss": 2.1221, "step": 19197 }, { "epoch": 0.47, "learning_rate": 1.1349836478442042e-05, "loss": 1.9342, "step": 19198 }, { "epoch": 0.47, "learning_rate": 1.1349045449478918e-05, "loss": 1.8735, "step": 19199 }, { "epoch": 0.47, "learning_rate": 1.1348254411917894e-05, "loss": 2.0087, "step": 19200 }, { "epoch": 0.47, "learning_rate": 1.1347463365764002e-05, "loss": 2.1085, "step": 19201 }, { "epoch": 0.47, "learning_rate": 1.1346672311022288e-05, "loss": 1.998, "step": 19202 }, { "epoch": 0.47, "learning_rate": 1.134588124769779e-05, "loss": 2.1028, "step": 19203 }, { "epoch": 0.47, "learning_rate": 1.1345090175795552e-05, "loss": 2.1015, "step": 19204 }, { "epoch": 0.47, "learning_rate": 1.1344299095320617e-05, "loss": 2.1742, "step": 19205 }, { "epoch": 0.47, "learning_rate": 1.1343508006278028e-05, "loss": 2.1036, "step": 19206 }, { "epoch": 0.47, "learning_rate": 1.1342716908672822e-05, "loss": 1.9271, "step": 19207 }, { "epoch": 0.47, "learning_rate": 1.1341925802510045e-05, "loss": 2.1958, "step": 19208 }, { "epoch": 0.47, "learning_rate": 1.1341134687794738e-05, "loss": 2.0077, "step": 19209 }, { "epoch": 0.47, "learning_rate": 1.134034356453194e-05, "loss": 1.9455, "step": 19210 }, { "epoch": 0.47, "learning_rate": 1.13395524327267e-05, "loss": 1.9161, "step": 19211 }, { "epoch": 0.47, "learning_rate": 1.1338761292384051e-05, "loss": 1.8609, "step": 19212 }, { "epoch": 0.47, "learning_rate": 1.133797014350904e-05, "loss": 2.0327, "step": 19213 }, { "epoch": 0.47, "learning_rate": 1.1337178986106711e-05, "loss": 1.9115, "step": 19214 }, { "epoch": 0.47, "learning_rate": 1.1336387820182106e-05, "loss": 2.1613, "step": 19215 }, { "epoch": 0.47, "learning_rate": 1.1335596645740266e-05, "loss": 2.0997, "step": 19216 }, { "epoch": 0.47, "learning_rate": 1.1334805462786231e-05, "loss": 2.0869, "step": 19217 }, { "epoch": 0.47, "learning_rate": 1.1334014271325046e-05, "loss": 1.9801, "step": 19218 }, { "epoch": 0.47, "learning_rate": 1.1333223071361755e-05, "loss": 2.0891, "step": 19219 }, { "epoch": 0.47, "learning_rate": 1.1332431862901398e-05, "loss": 1.9285, "step": 19220 }, { "epoch": 0.47, "learning_rate": 1.1331640645949017e-05, "loss": 1.9685, "step": 19221 }, { "epoch": 0.47, "learning_rate": 1.1330849420509658e-05, "loss": 2.0067, "step": 19222 }, { "epoch": 0.47, "learning_rate": 1.133005818658836e-05, "loss": 2.1108, "step": 19223 }, { "epoch": 0.47, "learning_rate": 1.132926694419017e-05, "loss": 1.9894, "step": 19224 }, { "epoch": 0.47, "learning_rate": 1.1328475693320128e-05, "loss": 2.2294, "step": 19225 }, { "epoch": 0.47, "learning_rate": 1.132768443398328e-05, "loss": 2.2185, "step": 19226 }, { "epoch": 0.47, "learning_rate": 1.1326893166184663e-05, "loss": 2.1554, "step": 19227 }, { "epoch": 0.47, "learning_rate": 1.1326101889929322e-05, "loss": 2.1084, "step": 19228 }, { "epoch": 0.47, "learning_rate": 1.1325310605222306e-05, "loss": 2.1503, "step": 19229 }, { "epoch": 0.47, "learning_rate": 1.1324519312068649e-05, "loss": 1.8574, "step": 19230 }, { "epoch": 0.47, "learning_rate": 1.1323728010473403e-05, "loss": 1.9291, "step": 19231 }, { "epoch": 0.47, "learning_rate": 1.1322936700441605e-05, "loss": 1.9394, "step": 19232 }, { "epoch": 0.47, "learning_rate": 1.1322145381978297e-05, "loss": 1.9538, "step": 19233 }, { "epoch": 0.47, "learning_rate": 1.1321354055088531e-05, "loss": 2.0179, "step": 19234 }, { "epoch": 0.47, "learning_rate": 1.1320562719777344e-05, "loss": 2.0295, "step": 19235 }, { "epoch": 0.47, "learning_rate": 1.131977137604978e-05, "loss": 2.0038, "step": 19236 }, { "epoch": 0.47, "learning_rate": 1.1318980023910884e-05, "loss": 2.1214, "step": 19237 }, { "epoch": 0.47, "learning_rate": 1.1318188663365694e-05, "loss": 1.9387, "step": 19238 }, { "epoch": 0.47, "learning_rate": 1.1317397294419258e-05, "loss": 1.9629, "step": 19239 }, { "epoch": 0.47, "learning_rate": 1.1316605917076624e-05, "loss": 1.9776, "step": 19240 }, { "epoch": 0.47, "learning_rate": 1.131581453134283e-05, "loss": 2.0375, "step": 19241 }, { "epoch": 0.47, "learning_rate": 1.1315023137222923e-05, "loss": 2.224, "step": 19242 }, { "epoch": 0.47, "learning_rate": 1.131423173472194e-05, "loss": 2.0465, "step": 19243 }, { "epoch": 0.47, "learning_rate": 1.1313440323844935e-05, "loss": 1.8334, "step": 19244 }, { "epoch": 0.47, "learning_rate": 1.1312648904596946e-05, "loss": 2.1498, "step": 19245 }, { "epoch": 0.47, "learning_rate": 1.1311857476983018e-05, "loss": 1.8777, "step": 19246 }, { "epoch": 0.47, "learning_rate": 1.131106604100819e-05, "loss": 1.9025, "step": 19247 }, { "epoch": 0.47, "learning_rate": 1.1310274596677513e-05, "loss": 2.0836, "step": 19248 }, { "epoch": 0.47, "learning_rate": 1.1309483143996032e-05, "loss": 2.0462, "step": 19249 }, { "epoch": 0.47, "learning_rate": 1.1308691682968787e-05, "loss": 1.8887, "step": 19250 }, { "epoch": 0.47, "learning_rate": 1.1307900213600822e-05, "loss": 2.0763, "step": 19251 }, { "epoch": 0.47, "learning_rate": 1.130710873589718e-05, "loss": 2.1577, "step": 19252 }, { "epoch": 0.47, "learning_rate": 1.1306317249862916e-05, "loss": 2.049, "step": 19253 }, { "epoch": 0.47, "learning_rate": 1.1305525755503062e-05, "loss": 2.2315, "step": 19254 }, { "epoch": 0.47, "learning_rate": 1.1304734252822667e-05, "loss": 2.1705, "step": 19255 }, { "epoch": 0.47, "learning_rate": 1.1303942741826773e-05, "loss": 2.0091, "step": 19256 }, { "epoch": 0.47, "learning_rate": 1.1303151222520427e-05, "loss": 1.9825, "step": 19257 }, { "epoch": 0.47, "learning_rate": 1.1302359694908678e-05, "loss": 1.8571, "step": 19258 }, { "epoch": 0.47, "learning_rate": 1.1301568158996562e-05, "loss": 1.8394, "step": 19259 }, { "epoch": 0.47, "learning_rate": 1.1300776614789134e-05, "loss": 2.1198, "step": 19260 }, { "epoch": 0.47, "learning_rate": 1.1299985062291427e-05, "loss": 1.9596, "step": 19261 }, { "epoch": 0.47, "learning_rate": 1.1299193501508492e-05, "loss": 2.0109, "step": 19262 }, { "epoch": 0.47, "learning_rate": 1.1298401932445376e-05, "loss": 2.2357, "step": 19263 }, { "epoch": 0.47, "learning_rate": 1.129761035510712e-05, "loss": 2.0803, "step": 19264 }, { "epoch": 0.47, "learning_rate": 1.129681876949877e-05, "loss": 2.0663, "step": 19265 }, { "epoch": 0.47, "learning_rate": 1.1296027175625372e-05, "loss": 1.8936, "step": 19266 }, { "epoch": 0.47, "learning_rate": 1.1295235573491969e-05, "loss": 1.8769, "step": 19267 }, { "epoch": 0.47, "learning_rate": 1.129444396310361e-05, "loss": 1.8321, "step": 19268 }, { "epoch": 0.47, "learning_rate": 1.1293652344465335e-05, "loss": 2.0249, "step": 19269 }, { "epoch": 0.47, "learning_rate": 1.1292860717582195e-05, "loss": 2.3312, "step": 19270 }, { "epoch": 0.48, "learning_rate": 1.1292069082459232e-05, "loss": 2.1103, "step": 19271 }, { "epoch": 0.48, "learning_rate": 1.1291277439101489e-05, "loss": 2.1635, "step": 19272 }, { "epoch": 0.48, "learning_rate": 1.1290485787514016e-05, "loss": 1.9919, "step": 19273 }, { "epoch": 0.48, "learning_rate": 1.1289694127701853e-05, "loss": 2.0711, "step": 19274 }, { "epoch": 0.48, "learning_rate": 1.1288902459670054e-05, "loss": 2.1276, "step": 19275 }, { "epoch": 0.48, "learning_rate": 1.1288110783423658e-05, "loss": 2.129, "step": 19276 }, { "epoch": 0.48, "learning_rate": 1.128731909896771e-05, "loss": 2.073, "step": 19277 }, { "epoch": 0.48, "learning_rate": 1.1286527406307261e-05, "loss": 2.0695, "step": 19278 }, { "epoch": 0.48, "learning_rate": 1.1285735705447353e-05, "loss": 2.1497, "step": 19279 }, { "epoch": 0.48, "learning_rate": 1.1284943996393032e-05, "loss": 1.868, "step": 19280 }, { "epoch": 0.48, "learning_rate": 1.1284152279149343e-05, "loss": 1.9328, "step": 19281 }, { "epoch": 0.48, "learning_rate": 1.1283360553721335e-05, "loss": 2.0842, "step": 19282 }, { "epoch": 0.48, "learning_rate": 1.1282568820114049e-05, "loss": 2.0772, "step": 19283 }, { "epoch": 0.48, "learning_rate": 1.1281777078332536e-05, "loss": 2.149, "step": 19284 }, { "epoch": 0.48, "learning_rate": 1.128098532838184e-05, "loss": 1.988, "step": 19285 }, { "epoch": 0.48, "learning_rate": 1.1280193570267007e-05, "loss": 1.9348, "step": 19286 }, { "epoch": 0.48, "learning_rate": 1.1279401803993082e-05, "loss": 2.2318, "step": 19287 }, { "epoch": 0.48, "learning_rate": 1.1278610029565114e-05, "loss": 2.1073, "step": 19288 }, { "epoch": 0.48, "learning_rate": 1.1277818246988148e-05, "loss": 1.9504, "step": 19289 }, { "epoch": 0.48, "learning_rate": 1.1277026456267228e-05, "loss": 1.9871, "step": 19290 }, { "epoch": 0.48, "learning_rate": 1.1276234657407404e-05, "loss": 2.1684, "step": 19291 }, { "epoch": 0.48, "learning_rate": 1.1275442850413715e-05, "loss": 1.9482, "step": 19292 }, { "epoch": 0.48, "learning_rate": 1.127465103529122e-05, "loss": 2.0142, "step": 19293 }, { "epoch": 0.48, "learning_rate": 1.1273859212044958e-05, "loss": 2.0135, "step": 19294 }, { "epoch": 0.48, "learning_rate": 1.1273067380679975e-05, "loss": 1.9958, "step": 19295 }, { "epoch": 0.48, "learning_rate": 1.1272275541201316e-05, "loss": 2.0973, "step": 19296 }, { "epoch": 0.48, "learning_rate": 1.127148369361403e-05, "loss": 2.0266, "step": 19297 }, { "epoch": 0.48, "learning_rate": 1.127069183792317e-05, "loss": 2.1375, "step": 19298 }, { "epoch": 0.48, "learning_rate": 1.1269899974133774e-05, "loss": 2.0579, "step": 19299 }, { "epoch": 0.48, "learning_rate": 1.126910810225089e-05, "loss": 2.1814, "step": 19300 }, { "epoch": 0.48, "learning_rate": 1.1268316222279566e-05, "loss": 2.052, "step": 19301 }, { "epoch": 0.48, "learning_rate": 1.1267524334224848e-05, "loss": 2.0117, "step": 19302 }, { "epoch": 0.48, "learning_rate": 1.1266732438091789e-05, "loss": 2.2026, "step": 19303 }, { "epoch": 0.48, "learning_rate": 1.1265940533885432e-05, "loss": 1.8857, "step": 19304 }, { "epoch": 0.48, "learning_rate": 1.126514862161082e-05, "loss": 1.9392, "step": 19305 }, { "epoch": 0.48, "learning_rate": 1.1264356701273003e-05, "loss": 2.1821, "step": 19306 }, { "epoch": 0.48, "learning_rate": 1.1263564772877032e-05, "loss": 1.9266, "step": 19307 }, { "epoch": 0.48, "learning_rate": 1.1262772836427949e-05, "loss": 1.9792, "step": 19308 }, { "epoch": 0.48, "learning_rate": 1.1261980891930801e-05, "loss": 2.0117, "step": 19309 }, { "epoch": 0.48, "learning_rate": 1.1261188939390637e-05, "loss": 2.149, "step": 19310 }, { "epoch": 0.48, "learning_rate": 1.126039697881251e-05, "loss": 2.2142, "step": 19311 }, { "epoch": 0.48, "learning_rate": 1.125960501020146e-05, "loss": 2.0301, "step": 19312 }, { "epoch": 0.48, "learning_rate": 1.1258813033562537e-05, "loss": 1.9201, "step": 19313 }, { "epoch": 0.48, "learning_rate": 1.125802104890079e-05, "loss": 1.9671, "step": 19314 }, { "epoch": 0.48, "learning_rate": 1.125722905622126e-05, "loss": 2.0824, "step": 19315 }, { "epoch": 0.48, "learning_rate": 1.1256437055529004e-05, "loss": 2.0426, "step": 19316 }, { "epoch": 0.48, "learning_rate": 1.1255645046829063e-05, "loss": 2.0085, "step": 19317 }, { "epoch": 0.48, "learning_rate": 1.1254853030126486e-05, "loss": 1.9222, "step": 19318 }, { "epoch": 0.48, "learning_rate": 1.1254061005426325e-05, "loss": 2.195, "step": 19319 }, { "epoch": 0.48, "learning_rate": 1.125326897273362e-05, "loss": 2.1098, "step": 19320 }, { "epoch": 0.48, "learning_rate": 1.1252476932053426e-05, "loss": 1.8909, "step": 19321 }, { "epoch": 0.48, "learning_rate": 1.125168488339079e-05, "loss": 2.165, "step": 19322 }, { "epoch": 0.48, "learning_rate": 1.1250892826750756e-05, "loss": 2.0457, "step": 19323 }, { "epoch": 0.48, "learning_rate": 1.1250100762138376e-05, "loss": 2.0268, "step": 19324 }, { "epoch": 0.48, "learning_rate": 1.1249308689558695e-05, "loss": 2.1413, "step": 19325 }, { "epoch": 0.48, "learning_rate": 1.1248516609016765e-05, "loss": 2.0762, "step": 19326 }, { "epoch": 0.48, "learning_rate": 1.1247724520517626e-05, "loss": 2.1185, "step": 19327 }, { "epoch": 0.48, "learning_rate": 1.1246932424066337e-05, "loss": 1.9667, "step": 19328 }, { "epoch": 0.48, "learning_rate": 1.1246140319667942e-05, "loss": 2.0258, "step": 19329 }, { "epoch": 0.48, "learning_rate": 1.1245348207327484e-05, "loss": 2.0437, "step": 19330 }, { "epoch": 0.48, "learning_rate": 1.124455608705002e-05, "loss": 2.0963, "step": 19331 }, { "epoch": 0.48, "learning_rate": 1.1243763958840594e-05, "loss": 2.0805, "step": 19332 }, { "epoch": 0.48, "learning_rate": 1.1242971822704253e-05, "loss": 1.9542, "step": 19333 }, { "epoch": 0.48, "learning_rate": 1.1242179678646048e-05, "loss": 2.2164, "step": 19334 }, { "epoch": 0.48, "learning_rate": 1.1241387526671031e-05, "loss": 1.8105, "step": 19335 }, { "epoch": 0.48, "learning_rate": 1.1240595366784241e-05, "loss": 2.0515, "step": 19336 }, { "epoch": 0.48, "learning_rate": 1.1239803198990735e-05, "loss": 1.7599, "step": 19337 }, { "epoch": 0.48, "learning_rate": 1.1239011023295561e-05, "loss": 2.055, "step": 19338 }, { "epoch": 0.48, "learning_rate": 1.1238218839703762e-05, "loss": 2.1969, "step": 19339 }, { "epoch": 0.48, "learning_rate": 1.1237426648220392e-05, "loss": 1.9551, "step": 19340 }, { "epoch": 0.48, "learning_rate": 1.1236634448850504e-05, "loss": 2.1137, "step": 19341 }, { "epoch": 0.48, "learning_rate": 1.1235842241599135e-05, "loss": 2.0166, "step": 19342 }, { "epoch": 0.48, "learning_rate": 1.1235050026471343e-05, "loss": 1.9881, "step": 19343 }, { "epoch": 0.48, "learning_rate": 1.1234257803472175e-05, "loss": 1.8355, "step": 19344 }, { "epoch": 0.48, "learning_rate": 1.1233465572606677e-05, "loss": 2.1354, "step": 19345 }, { "epoch": 0.48, "learning_rate": 1.1232673333879904e-05, "loss": 2.136, "step": 19346 }, { "epoch": 0.48, "learning_rate": 1.1231881087296903e-05, "loss": 2.0767, "step": 19347 }, { "epoch": 0.48, "learning_rate": 1.1231088832862722e-05, "loss": 2.0976, "step": 19348 }, { "epoch": 0.48, "learning_rate": 1.1230296570582407e-05, "loss": 1.9072, "step": 19349 }, { "epoch": 0.48, "learning_rate": 1.1229504300461014e-05, "loss": 2.0749, "step": 19350 }, { "epoch": 0.48, "learning_rate": 1.122871202250359e-05, "loss": 1.9391, "step": 19351 }, { "epoch": 0.48, "learning_rate": 1.1227919736715185e-05, "loss": 2.158, "step": 19352 }, { "epoch": 0.48, "learning_rate": 1.1227127443100843e-05, "loss": 1.9376, "step": 19353 }, { "epoch": 0.48, "learning_rate": 1.1226335141665619e-05, "loss": 1.9016, "step": 19354 }, { "epoch": 0.48, "learning_rate": 1.1225542832414562e-05, "loss": 2.1835, "step": 19355 }, { "epoch": 0.48, "learning_rate": 1.1224750515352722e-05, "loss": 2.0028, "step": 19356 }, { "epoch": 0.48, "learning_rate": 1.122395819048515e-05, "loss": 2.0876, "step": 19357 }, { "epoch": 0.48, "learning_rate": 1.1223165857816892e-05, "loss": 2.0588, "step": 19358 }, { "epoch": 0.48, "learning_rate": 1.1222373517352995e-05, "loss": 2.067, "step": 19359 }, { "epoch": 0.48, "learning_rate": 1.122158116909852e-05, "loss": 2.1974, "step": 19360 }, { "epoch": 0.48, "learning_rate": 1.1220788813058505e-05, "loss": 2.0186, "step": 19361 }, { "epoch": 0.48, "learning_rate": 1.1219996449238006e-05, "loss": 1.9925, "step": 19362 }, { "epoch": 0.48, "learning_rate": 1.1219204077642074e-05, "loss": 1.9929, "step": 19363 }, { "epoch": 0.48, "learning_rate": 1.1218411698275753e-05, "loss": 1.8642, "step": 19364 }, { "epoch": 0.48, "learning_rate": 1.1217619311144102e-05, "loss": 2.0455, "step": 19365 }, { "epoch": 0.48, "learning_rate": 1.1216826916252163e-05, "loss": 2.3158, "step": 19366 }, { "epoch": 0.48, "learning_rate": 1.1216034513604994e-05, "loss": 2.1595, "step": 19367 }, { "epoch": 0.48, "learning_rate": 1.1215242103207635e-05, "loss": 2.0199, "step": 19368 }, { "epoch": 0.48, "learning_rate": 1.1214449685065145e-05, "loss": 2.2407, "step": 19369 }, { "epoch": 0.48, "learning_rate": 1.1213657259182573e-05, "loss": 2.0532, "step": 19370 }, { "epoch": 0.48, "learning_rate": 1.1212864825564964e-05, "loss": 2.2031, "step": 19371 }, { "epoch": 0.48, "learning_rate": 1.1212072384217374e-05, "loss": 1.947, "step": 19372 }, { "epoch": 0.48, "learning_rate": 1.1211279935144851e-05, "loss": 2.0429, "step": 19373 }, { "epoch": 0.48, "learning_rate": 1.1210487478352448e-05, "loss": 2.0341, "step": 19374 }, { "epoch": 0.48, "learning_rate": 1.1209695013845214e-05, "loss": 2.1385, "step": 19375 }, { "epoch": 0.48, "learning_rate": 1.1208902541628199e-05, "loss": 2.0064, "step": 19376 }, { "epoch": 0.48, "learning_rate": 1.1208110061706455e-05, "loss": 2.0807, "step": 19377 }, { "epoch": 0.48, "learning_rate": 1.1207317574085028e-05, "loss": 2.0345, "step": 19378 }, { "epoch": 0.48, "learning_rate": 1.1206525078768975e-05, "loss": 2.0625, "step": 19379 }, { "epoch": 0.48, "learning_rate": 1.1205732575763344e-05, "loss": 2.0515, "step": 19380 }, { "epoch": 0.48, "learning_rate": 1.1204940065073187e-05, "loss": 2.1043, "step": 19381 }, { "epoch": 0.48, "learning_rate": 1.1204147546703554e-05, "loss": 2.0185, "step": 19382 }, { "epoch": 0.48, "learning_rate": 1.1203355020659495e-05, "loss": 2.0049, "step": 19383 }, { "epoch": 0.48, "learning_rate": 1.1202562486946064e-05, "loss": 1.9609, "step": 19384 }, { "epoch": 0.48, "learning_rate": 1.120176994556831e-05, "loss": 1.7727, "step": 19385 }, { "epoch": 0.48, "learning_rate": 1.1200977396531285e-05, "loss": 1.872, "step": 19386 }, { "epoch": 0.48, "learning_rate": 1.1200184839840038e-05, "loss": 2.0312, "step": 19387 }, { "epoch": 0.48, "learning_rate": 1.1199392275499626e-05, "loss": 1.9055, "step": 19388 }, { "epoch": 0.48, "learning_rate": 1.119859970351509e-05, "loss": 2.1051, "step": 19389 }, { "epoch": 0.48, "learning_rate": 1.1197807123891491e-05, "loss": 2.2086, "step": 19390 }, { "epoch": 0.48, "learning_rate": 1.1197014536633876e-05, "loss": 1.8783, "step": 19391 }, { "epoch": 0.48, "learning_rate": 1.1196221941747301e-05, "loss": 2.0345, "step": 19392 }, { "epoch": 0.48, "learning_rate": 1.1195429339236808e-05, "loss": 1.9651, "step": 19393 }, { "epoch": 0.48, "learning_rate": 1.1194636729107458e-05, "loss": 2.0077, "step": 19394 }, { "epoch": 0.48, "learning_rate": 1.1193844111364299e-05, "loss": 1.9371, "step": 19395 }, { "epoch": 0.48, "learning_rate": 1.1193051486012378e-05, "loss": 1.9876, "step": 19396 }, { "epoch": 0.48, "learning_rate": 1.1192258853056754e-05, "loss": 2.065, "step": 19397 }, { "epoch": 0.48, "learning_rate": 1.1191466212502475e-05, "loss": 2.077, "step": 19398 }, { "epoch": 0.48, "learning_rate": 1.1190673564354596e-05, "loss": 2.132, "step": 19399 }, { "epoch": 0.48, "learning_rate": 1.1189880908618165e-05, "loss": 2.0634, "step": 19400 }, { "epoch": 0.48, "learning_rate": 1.1189088245298234e-05, "loss": 1.9697, "step": 19401 }, { "epoch": 0.48, "learning_rate": 1.1188295574399857e-05, "loss": 1.9957, "step": 19402 }, { "epoch": 0.48, "learning_rate": 1.1187502895928086e-05, "loss": 2.1624, "step": 19403 }, { "epoch": 0.48, "learning_rate": 1.1186710209887973e-05, "loss": 1.9359, "step": 19404 }, { "epoch": 0.48, "learning_rate": 1.1185917516284563e-05, "loss": 2.0358, "step": 19405 }, { "epoch": 0.48, "learning_rate": 1.118512481512292e-05, "loss": 1.9369, "step": 19406 }, { "epoch": 0.48, "learning_rate": 1.1184332106408087e-05, "loss": 1.9976, "step": 19407 }, { "epoch": 0.48, "learning_rate": 1.1183539390145121e-05, "loss": 2.1114, "step": 19408 }, { "epoch": 0.48, "learning_rate": 1.1182746666339074e-05, "loss": 1.9605, "step": 19409 }, { "epoch": 0.48, "learning_rate": 1.1181953934994998e-05, "loss": 2.0226, "step": 19410 }, { "epoch": 0.48, "learning_rate": 1.1181161196117942e-05, "loss": 1.9936, "step": 19411 }, { "epoch": 0.48, "learning_rate": 1.118036844971296e-05, "loss": 2.0843, "step": 19412 }, { "epoch": 0.48, "learning_rate": 1.1179575695785108e-05, "loss": 2.0992, "step": 19413 }, { "epoch": 0.48, "learning_rate": 1.1178782934339431e-05, "loss": 2.0688, "step": 19414 }, { "epoch": 0.48, "learning_rate": 1.1177990165380992e-05, "loss": 2.1883, "step": 19415 }, { "epoch": 0.48, "learning_rate": 1.1177197388914836e-05, "loss": 2.1582, "step": 19416 }, { "epoch": 0.48, "learning_rate": 1.1176404604946014e-05, "loss": 1.9198, "step": 19417 }, { "epoch": 0.48, "learning_rate": 1.1175611813479586e-05, "loss": 2.0859, "step": 19418 }, { "epoch": 0.48, "learning_rate": 1.11748190145206e-05, "loss": 2.1297, "step": 19419 }, { "epoch": 0.48, "learning_rate": 1.1174026208074109e-05, "loss": 1.9601, "step": 19420 }, { "epoch": 0.48, "learning_rate": 1.1173233394145166e-05, "loss": 1.8904, "step": 19421 }, { "epoch": 0.48, "learning_rate": 1.1172440572738823e-05, "loss": 2.1417, "step": 19422 }, { "epoch": 0.48, "learning_rate": 1.1171647743860135e-05, "loss": 2.0741, "step": 19423 }, { "epoch": 0.48, "learning_rate": 1.1170854907514155e-05, "loss": 2.1941, "step": 19424 }, { "epoch": 0.48, "learning_rate": 1.1170062063705935e-05, "loss": 2.2155, "step": 19425 }, { "epoch": 0.48, "learning_rate": 1.116926921244053e-05, "loss": 2.0695, "step": 19426 }, { "epoch": 0.48, "learning_rate": 1.1168476353722987e-05, "loss": 1.8866, "step": 19427 }, { "epoch": 0.48, "learning_rate": 1.1167683487558368e-05, "loss": 2.0406, "step": 19428 }, { "epoch": 0.48, "learning_rate": 1.1166890613951717e-05, "loss": 1.9914, "step": 19429 }, { "epoch": 0.48, "learning_rate": 1.1166097732908097e-05, "loss": 2.0276, "step": 19430 }, { "epoch": 0.48, "learning_rate": 1.116530484443255e-05, "loss": 2.0687, "step": 19431 }, { "epoch": 0.48, "learning_rate": 1.1164511948530139e-05, "loss": 2.0816, "step": 19432 }, { "epoch": 0.48, "learning_rate": 1.1163719045205914e-05, "loss": 2.0546, "step": 19433 }, { "epoch": 0.48, "learning_rate": 1.116292613446493e-05, "loss": 1.7674, "step": 19434 }, { "epoch": 0.48, "learning_rate": 1.1162133216312235e-05, "loss": 2.129, "step": 19435 }, { "epoch": 0.48, "learning_rate": 1.1161340290752887e-05, "loss": 2.1179, "step": 19436 }, { "epoch": 0.48, "learning_rate": 1.116054735779194e-05, "loss": 1.8784, "step": 19437 }, { "epoch": 0.48, "learning_rate": 1.1159754417434447e-05, "loss": 1.9686, "step": 19438 }, { "epoch": 0.48, "learning_rate": 1.115896146968546e-05, "loss": 2.0284, "step": 19439 }, { "epoch": 0.48, "learning_rate": 1.1158168514550034e-05, "loss": 2.0583, "step": 19440 }, { "epoch": 0.48, "learning_rate": 1.1157375552033223e-05, "loss": 2.042, "step": 19441 }, { "epoch": 0.48, "learning_rate": 1.1156582582140078e-05, "loss": 1.9982, "step": 19442 }, { "epoch": 0.48, "learning_rate": 1.1155789604875658e-05, "loss": 2.0488, "step": 19443 }, { "epoch": 0.48, "learning_rate": 1.1154996620245015e-05, "loss": 2.1755, "step": 19444 }, { "epoch": 0.48, "learning_rate": 1.1154203628253203e-05, "loss": 2.1044, "step": 19445 }, { "epoch": 0.48, "learning_rate": 1.115341062890527e-05, "loss": 2.1795, "step": 19446 }, { "epoch": 0.48, "learning_rate": 1.1152617622206278e-05, "loss": 2.2733, "step": 19447 }, { "epoch": 0.48, "learning_rate": 1.1151824608161278e-05, "loss": 1.8961, "step": 19448 }, { "epoch": 0.48, "learning_rate": 1.1151031586775325e-05, "loss": 2.0101, "step": 19449 }, { "epoch": 0.48, "learning_rate": 1.1150238558053474e-05, "loss": 1.9641, "step": 19450 }, { "epoch": 0.48, "learning_rate": 1.1149445522000774e-05, "loss": 1.9757, "step": 19451 }, { "epoch": 0.48, "learning_rate": 1.1148652478622287e-05, "loss": 2.1454, "step": 19452 }, { "epoch": 0.48, "learning_rate": 1.1147859427923063e-05, "loss": 2.0349, "step": 19453 }, { "epoch": 0.48, "learning_rate": 1.1147066369908156e-05, "loss": 2.1776, "step": 19454 }, { "epoch": 0.48, "learning_rate": 1.114627330458262e-05, "loss": 2.2492, "step": 19455 }, { "epoch": 0.48, "learning_rate": 1.1145480231951511e-05, "loss": 2.1428, "step": 19456 }, { "epoch": 0.48, "learning_rate": 1.1144687152019884e-05, "loss": 2.0239, "step": 19457 }, { "epoch": 0.48, "learning_rate": 1.1143894064792791e-05, "loss": 1.9678, "step": 19458 }, { "epoch": 0.48, "learning_rate": 1.1143100970275289e-05, "loss": 2.0176, "step": 19459 }, { "epoch": 0.48, "learning_rate": 1.1142307868472431e-05, "loss": 2.0446, "step": 19460 }, { "epoch": 0.48, "learning_rate": 1.1141514759389276e-05, "loss": 1.9934, "step": 19461 }, { "epoch": 0.48, "learning_rate": 1.1140721643030873e-05, "loss": 2.1335, "step": 19462 }, { "epoch": 0.48, "learning_rate": 1.1139928519402282e-05, "loss": 2.0324, "step": 19463 }, { "epoch": 0.48, "learning_rate": 1.1139135388508553e-05, "loss": 1.9762, "step": 19464 }, { "epoch": 0.48, "learning_rate": 1.1138342250354741e-05, "loss": 2.0522, "step": 19465 }, { "epoch": 0.48, "learning_rate": 1.1137549104945905e-05, "loss": 1.8283, "step": 19466 }, { "epoch": 0.48, "learning_rate": 1.1136755952287098e-05, "loss": 2.1773, "step": 19467 }, { "epoch": 0.48, "learning_rate": 1.1135962792383375e-05, "loss": 2.0616, "step": 19468 }, { "epoch": 0.48, "learning_rate": 1.113516962523979e-05, "loss": 1.9817, "step": 19469 }, { "epoch": 0.48, "learning_rate": 1.11343764508614e-05, "loss": 2.0944, "step": 19470 }, { "epoch": 0.48, "learning_rate": 1.113358326925326e-05, "loss": 2.0298, "step": 19471 }, { "epoch": 0.48, "learning_rate": 1.1132790080420422e-05, "loss": 2.2004, "step": 19472 }, { "epoch": 0.48, "learning_rate": 1.1131996884367946e-05, "loss": 1.9985, "step": 19473 }, { "epoch": 0.48, "learning_rate": 1.1131203681100885e-05, "loss": 1.8941, "step": 19474 }, { "epoch": 0.48, "learning_rate": 1.113041047062429e-05, "loss": 2.1576, "step": 19475 }, { "epoch": 0.48, "learning_rate": 1.1129617252943224e-05, "loss": 2.0201, "step": 19476 }, { "epoch": 0.48, "learning_rate": 1.1128824028062738e-05, "loss": 2.1605, "step": 19477 }, { "epoch": 0.48, "learning_rate": 1.112803079598789e-05, "loss": 2.0747, "step": 19478 }, { "epoch": 0.48, "learning_rate": 1.1127237556723735e-05, "loss": 2.083, "step": 19479 }, { "epoch": 0.48, "learning_rate": 1.1126444310275324e-05, "loss": 2.039, "step": 19480 }, { "epoch": 0.48, "learning_rate": 1.1125651056647719e-05, "loss": 1.9244, "step": 19481 }, { "epoch": 0.48, "learning_rate": 1.1124857795845974e-05, "loss": 1.9581, "step": 19482 }, { "epoch": 0.48, "learning_rate": 1.1124064527875143e-05, "loss": 2.1441, "step": 19483 }, { "epoch": 0.48, "learning_rate": 1.1123271252740279e-05, "loss": 2.0633, "step": 19484 }, { "epoch": 0.48, "learning_rate": 1.1122477970446443e-05, "loss": 1.9566, "step": 19485 }, { "epoch": 0.48, "learning_rate": 1.112168468099869e-05, "loss": 2.0593, "step": 19486 }, { "epoch": 0.48, "learning_rate": 1.1120891384402077e-05, "loss": 1.9818, "step": 19487 }, { "epoch": 0.48, "learning_rate": 1.1120098080661657e-05, "loss": 2.1613, "step": 19488 }, { "epoch": 0.48, "learning_rate": 1.1119304769782483e-05, "loss": 2.2255, "step": 19489 }, { "epoch": 0.48, "learning_rate": 1.1118511451769617e-05, "loss": 1.9125, "step": 19490 }, { "epoch": 0.48, "learning_rate": 1.1117718126628117e-05, "loss": 1.8999, "step": 19491 }, { "epoch": 0.48, "learning_rate": 1.1116924794363032e-05, "loss": 2.0251, "step": 19492 }, { "epoch": 0.48, "learning_rate": 1.111613145497942e-05, "loss": 2.0483, "step": 19493 }, { "epoch": 0.48, "learning_rate": 1.1115338108482336e-05, "loss": 2.0875, "step": 19494 }, { "epoch": 0.48, "learning_rate": 1.1114544754876844e-05, "loss": 2.0747, "step": 19495 }, { "epoch": 0.48, "learning_rate": 1.1113751394167993e-05, "loss": 2.0385, "step": 19496 }, { "epoch": 0.48, "learning_rate": 1.1112958026360843e-05, "loss": 1.9939, "step": 19497 }, { "epoch": 0.48, "learning_rate": 1.1112164651460448e-05, "loss": 2.045, "step": 19498 }, { "epoch": 0.48, "learning_rate": 1.1111371269471864e-05, "loss": 2.1198, "step": 19499 }, { "epoch": 0.48, "learning_rate": 1.111057788040015e-05, "loss": 1.8689, "step": 19500 }, { "epoch": 0.48, "learning_rate": 1.1109784484250362e-05, "loss": 2.083, "step": 19501 }, { "epoch": 0.48, "learning_rate": 1.1108991081027553e-05, "loss": 2.1688, "step": 19502 }, { "epoch": 0.48, "learning_rate": 1.1108197670736785e-05, "loss": 2.0143, "step": 19503 }, { "epoch": 0.48, "learning_rate": 1.110740425338311e-05, "loss": 1.9359, "step": 19504 }, { "epoch": 0.48, "learning_rate": 1.110661082897159e-05, "loss": 2.2043, "step": 19505 }, { "epoch": 0.48, "learning_rate": 1.1105817397507274e-05, "loss": 1.9126, "step": 19506 }, { "epoch": 0.48, "learning_rate": 1.1105023958995228e-05, "loss": 1.7963, "step": 19507 }, { "epoch": 0.48, "learning_rate": 1.1104230513440503e-05, "loss": 2.0365, "step": 19508 }, { "epoch": 0.48, "learning_rate": 1.1103437060848154e-05, "loss": 2.0827, "step": 19509 }, { "epoch": 0.48, "learning_rate": 1.1102643601223244e-05, "loss": 1.9452, "step": 19510 }, { "epoch": 0.48, "learning_rate": 1.1101850134570824e-05, "loss": 2.0164, "step": 19511 }, { "epoch": 0.48, "learning_rate": 1.1101056660895957e-05, "loss": 1.9135, "step": 19512 }, { "epoch": 0.48, "learning_rate": 1.1100263180203696e-05, "loss": 2.0353, "step": 19513 }, { "epoch": 0.48, "learning_rate": 1.1099469692499099e-05, "loss": 2.0577, "step": 19514 }, { "epoch": 0.48, "learning_rate": 1.1098676197787227e-05, "loss": 1.9606, "step": 19515 }, { "epoch": 0.48, "learning_rate": 1.1097882696073128e-05, "loss": 1.9895, "step": 19516 }, { "epoch": 0.48, "learning_rate": 1.109708918736187e-05, "loss": 1.908, "step": 19517 }, { "epoch": 0.48, "learning_rate": 1.10962956716585e-05, "loss": 1.9865, "step": 19518 }, { "epoch": 0.48, "learning_rate": 1.1095502148968082e-05, "loss": 2.0483, "step": 19519 }, { "epoch": 0.48, "learning_rate": 1.109470861929567e-05, "loss": 2.159, "step": 19520 }, { "epoch": 0.48, "learning_rate": 1.1093915082646328e-05, "loss": 2.2129, "step": 19521 }, { "epoch": 0.48, "learning_rate": 1.1093121539025106e-05, "loss": 1.9944, "step": 19522 }, { "epoch": 0.48, "learning_rate": 1.1092327988437064e-05, "loss": 1.9386, "step": 19523 }, { "epoch": 0.48, "learning_rate": 1.1091534430887261e-05, "loss": 2.1844, "step": 19524 }, { "epoch": 0.48, "learning_rate": 1.1090740866380752e-05, "loss": 2.049, "step": 19525 }, { "epoch": 0.48, "learning_rate": 1.1089947294922599e-05, "loss": 2.0492, "step": 19526 }, { "epoch": 0.48, "learning_rate": 1.1089153716517854e-05, "loss": 2.0499, "step": 19527 }, { "epoch": 0.48, "learning_rate": 1.1088360131171576e-05, "loss": 1.9151, "step": 19528 }, { "epoch": 0.48, "learning_rate": 1.1087566538888823e-05, "loss": 2.1426, "step": 19529 }, { "epoch": 0.48, "learning_rate": 1.1086772939674659e-05, "loss": 2.1072, "step": 19530 }, { "epoch": 0.48, "learning_rate": 1.1085979333534135e-05, "loss": 2.1939, "step": 19531 }, { "epoch": 0.48, "learning_rate": 1.108518572047231e-05, "loss": 2.1614, "step": 19532 }, { "epoch": 0.48, "learning_rate": 1.1084392100494244e-05, "loss": 2.0079, "step": 19533 }, { "epoch": 0.48, "learning_rate": 1.1083598473604992e-05, "loss": 2.0544, "step": 19534 }, { "epoch": 0.48, "learning_rate": 1.1082804839809616e-05, "loss": 1.9511, "step": 19535 }, { "epoch": 0.48, "learning_rate": 1.1082011199113172e-05, "loss": 2.1078, "step": 19536 }, { "epoch": 0.48, "learning_rate": 1.1081217551520716e-05, "loss": 2.1157, "step": 19537 }, { "epoch": 0.48, "learning_rate": 1.1080423897037306e-05, "loss": 2.1639, "step": 19538 }, { "epoch": 0.48, "learning_rate": 1.1079630235668007e-05, "loss": 1.9946, "step": 19539 }, { "epoch": 0.48, "learning_rate": 1.107883656741787e-05, "loss": 2.0888, "step": 19540 }, { "epoch": 0.48, "learning_rate": 1.1078042892291959e-05, "loss": 2.0207, "step": 19541 }, { "epoch": 0.48, "learning_rate": 1.1077249210295328e-05, "loss": 2.0543, "step": 19542 }, { "epoch": 0.48, "learning_rate": 1.1076455521433033e-05, "loss": 1.8823, "step": 19543 }, { "epoch": 0.48, "learning_rate": 1.107566182571014e-05, "loss": 2.03, "step": 19544 }, { "epoch": 0.48, "learning_rate": 1.1074868123131703e-05, "loss": 2.1119, "step": 19545 }, { "epoch": 0.48, "learning_rate": 1.107407441370278e-05, "loss": 2.0673, "step": 19546 }, { "epoch": 0.48, "learning_rate": 1.107328069742843e-05, "loss": 1.9139, "step": 19547 }, { "epoch": 0.48, "learning_rate": 1.1072486974313716e-05, "loss": 2.1142, "step": 19548 }, { "epoch": 0.48, "learning_rate": 1.1071693244363693e-05, "loss": 1.9036, "step": 19549 }, { "epoch": 0.48, "learning_rate": 1.1070899507583416e-05, "loss": 2.141, "step": 19550 }, { "epoch": 0.48, "learning_rate": 1.1070105763977951e-05, "loss": 2.0611, "step": 19551 }, { "epoch": 0.48, "learning_rate": 1.106931201355235e-05, "loss": 2.1608, "step": 19552 }, { "epoch": 0.48, "learning_rate": 1.1068518256311677e-05, "loss": 2.0944, "step": 19553 }, { "epoch": 0.48, "learning_rate": 1.106772449226099e-05, "loss": 1.9284, "step": 19554 }, { "epoch": 0.48, "learning_rate": 1.1066930721405342e-05, "loss": 2.0315, "step": 19555 }, { "epoch": 0.48, "learning_rate": 1.1066136943749801e-05, "loss": 1.7716, "step": 19556 }, { "epoch": 0.48, "learning_rate": 1.106534315929942e-05, "loss": 1.9795, "step": 19557 }, { "epoch": 0.48, "learning_rate": 1.1064549368059261e-05, "loss": 2.084, "step": 19558 }, { "epoch": 0.48, "learning_rate": 1.1063755570034382e-05, "loss": 1.9933, "step": 19559 }, { "epoch": 0.48, "learning_rate": 1.1062961765229843e-05, "loss": 2.1642, "step": 19560 }, { "epoch": 0.48, "learning_rate": 1.1062167953650702e-05, "loss": 1.966, "step": 19561 }, { "epoch": 0.48, "learning_rate": 1.1061374135302017e-05, "loss": 2.0553, "step": 19562 }, { "epoch": 0.48, "learning_rate": 1.1060580310188848e-05, "loss": 2.01, "step": 19563 }, { "epoch": 0.48, "learning_rate": 1.1059786478316255e-05, "loss": 2.1464, "step": 19564 }, { "epoch": 0.48, "learning_rate": 1.1058992639689299e-05, "loss": 1.9482, "step": 19565 }, { "epoch": 0.48, "learning_rate": 1.105819879431304e-05, "loss": 1.9634, "step": 19566 }, { "epoch": 0.48, "learning_rate": 1.1057404942192531e-05, "loss": 2.0866, "step": 19567 }, { "epoch": 0.48, "learning_rate": 1.1056611083332838e-05, "loss": 2.1077, "step": 19568 }, { "epoch": 0.48, "learning_rate": 1.1055817217739018e-05, "loss": 2.0491, "step": 19569 }, { "epoch": 0.48, "learning_rate": 1.1055023345416132e-05, "loss": 2.062, "step": 19570 }, { "epoch": 0.48, "learning_rate": 1.1054229466369235e-05, "loss": 2.0674, "step": 19571 }, { "epoch": 0.48, "learning_rate": 1.1053435580603391e-05, "loss": 2.0853, "step": 19572 }, { "epoch": 0.48, "learning_rate": 1.105264168812366e-05, "loss": 2.232, "step": 19573 }, { "epoch": 0.48, "learning_rate": 1.10518477889351e-05, "loss": 2.0969, "step": 19574 }, { "epoch": 0.48, "learning_rate": 1.105105388304277e-05, "loss": 2.0578, "step": 19575 }, { "epoch": 0.48, "learning_rate": 1.1050259970451734e-05, "loss": 1.9678, "step": 19576 }, { "epoch": 0.48, "learning_rate": 1.1049466051167046e-05, "loss": 2.2348, "step": 19577 }, { "epoch": 0.48, "learning_rate": 1.1048672125193772e-05, "loss": 2.2539, "step": 19578 }, { "epoch": 0.48, "learning_rate": 1.1047878192536967e-05, "loss": 2.027, "step": 19579 }, { "epoch": 0.48, "learning_rate": 1.1047084253201695e-05, "loss": 2.0616, "step": 19580 }, { "epoch": 0.48, "learning_rate": 1.1046290307193008e-05, "loss": 1.9808, "step": 19581 }, { "epoch": 0.48, "learning_rate": 1.1045496354515974e-05, "loss": 2.0436, "step": 19582 }, { "epoch": 0.48, "learning_rate": 1.1044702395175654e-05, "loss": 2.0616, "step": 19583 }, { "epoch": 0.48, "learning_rate": 1.1043908429177103e-05, "loss": 1.8606, "step": 19584 }, { "epoch": 0.48, "learning_rate": 1.1043114456525386e-05, "loss": 2.2356, "step": 19585 }, { "epoch": 0.48, "learning_rate": 1.1042320477225556e-05, "loss": 2.0099, "step": 19586 }, { "epoch": 0.48, "learning_rate": 1.1041526491282681e-05, "loss": 2.0869, "step": 19587 }, { "epoch": 0.48, "learning_rate": 1.104073249870182e-05, "loss": 1.8849, "step": 19588 }, { "epoch": 0.48, "learning_rate": 1.103993849948803e-05, "loss": 2.0265, "step": 19589 }, { "epoch": 0.48, "learning_rate": 1.1039144493646373e-05, "loss": 2.0029, "step": 19590 }, { "epoch": 0.48, "learning_rate": 1.1038350481181905e-05, "loss": 2.0566, "step": 19591 }, { "epoch": 0.48, "learning_rate": 1.1037556462099697e-05, "loss": 2.0072, "step": 19592 }, { "epoch": 0.48, "learning_rate": 1.1036762436404802e-05, "loss": 2.0638, "step": 19593 }, { "epoch": 0.48, "learning_rate": 1.1035968404102282e-05, "loss": 1.9805, "step": 19594 }, { "epoch": 0.48, "learning_rate": 1.10351743651972e-05, "loss": 2.0072, "step": 19595 }, { "epoch": 0.48, "learning_rate": 1.103438031969461e-05, "loss": 2.1413, "step": 19596 }, { "epoch": 0.48, "learning_rate": 1.1033586267599579e-05, "loss": 1.9996, "step": 19597 }, { "epoch": 0.48, "learning_rate": 1.1032792208917165e-05, "loss": 1.88, "step": 19598 }, { "epoch": 0.48, "learning_rate": 1.1031998143652429e-05, "loss": 2.006, "step": 19599 }, { "epoch": 0.48, "learning_rate": 1.1031204071810435e-05, "loss": 1.9873, "step": 19600 }, { "epoch": 0.48, "learning_rate": 1.1030409993396239e-05, "loss": 1.8843, "step": 19601 }, { "epoch": 0.48, "learning_rate": 1.1029615908414905e-05, "loss": 1.9064, "step": 19602 }, { "epoch": 0.48, "learning_rate": 1.1028821816871496e-05, "loss": 2.1203, "step": 19603 }, { "epoch": 0.48, "learning_rate": 1.1028027718771069e-05, "loss": 2.0308, "step": 19604 }, { "epoch": 0.48, "learning_rate": 1.1027233614118684e-05, "loss": 2.0101, "step": 19605 }, { "epoch": 0.48, "learning_rate": 1.1026439502919405e-05, "loss": 2.0486, "step": 19606 }, { "epoch": 0.48, "learning_rate": 1.1025645385178297e-05, "loss": 1.921, "step": 19607 }, { "epoch": 0.48, "learning_rate": 1.1024851260900411e-05, "loss": 1.9619, "step": 19608 }, { "epoch": 0.48, "learning_rate": 1.1024057130090817e-05, "loss": 1.8671, "step": 19609 }, { "epoch": 0.48, "learning_rate": 1.102326299275457e-05, "loss": 2.0983, "step": 19610 }, { "epoch": 0.48, "learning_rate": 1.1022468848896739e-05, "loss": 2.0453, "step": 19611 }, { "epoch": 0.48, "learning_rate": 1.1021674698522378e-05, "loss": 2.0624, "step": 19612 }, { "epoch": 0.48, "learning_rate": 1.1020880541636553e-05, "loss": 1.9632, "step": 19613 }, { "epoch": 0.48, "learning_rate": 1.1020086378244324e-05, "loss": 1.9664, "step": 19614 }, { "epoch": 0.48, "learning_rate": 1.101929220835075e-05, "loss": 2.0322, "step": 19615 }, { "epoch": 0.48, "learning_rate": 1.1018498031960893e-05, "loss": 2.1034, "step": 19616 }, { "epoch": 0.48, "learning_rate": 1.1017703849079817e-05, "loss": 2.1902, "step": 19617 }, { "epoch": 0.48, "learning_rate": 1.1016909659712584e-05, "loss": 1.8821, "step": 19618 }, { "epoch": 0.48, "learning_rate": 1.1016115463864257e-05, "loss": 2.0957, "step": 19619 }, { "epoch": 0.48, "learning_rate": 1.1015321261539891e-05, "loss": 2.1035, "step": 19620 }, { "epoch": 0.48, "learning_rate": 1.1014527052744552e-05, "loss": 2.0472, "step": 19621 }, { "epoch": 0.48, "learning_rate": 1.1013732837483305e-05, "loss": 1.9826, "step": 19622 }, { "epoch": 0.48, "learning_rate": 1.1012938615761206e-05, "loss": 2.2015, "step": 19623 }, { "epoch": 0.48, "learning_rate": 1.1012144387583316e-05, "loss": 1.9509, "step": 19624 }, { "epoch": 0.48, "learning_rate": 1.1011350152954705e-05, "loss": 2.0988, "step": 19625 }, { "epoch": 0.48, "learning_rate": 1.1010555911880425e-05, "loss": 2.1511, "step": 19626 }, { "epoch": 0.48, "learning_rate": 1.1009761664365546e-05, "loss": 2.0772, "step": 19627 }, { "epoch": 0.48, "learning_rate": 1.1008967410415128e-05, "loss": 1.9545, "step": 19628 }, { "epoch": 0.48, "learning_rate": 1.1008173150034231e-05, "loss": 2.0924, "step": 19629 }, { "epoch": 0.48, "learning_rate": 1.1007378883227914e-05, "loss": 2.0072, "step": 19630 }, { "epoch": 0.48, "learning_rate": 1.100658461000125e-05, "loss": 1.9369, "step": 19631 }, { "epoch": 0.48, "learning_rate": 1.100579033035929e-05, "loss": 2.0212, "step": 19632 }, { "epoch": 0.48, "learning_rate": 1.1004996044307099e-05, "loss": 1.8134, "step": 19633 }, { "epoch": 0.48, "learning_rate": 1.1004201751849744e-05, "loss": 2.0391, "step": 19634 }, { "epoch": 0.48, "learning_rate": 1.1003407452992281e-05, "loss": 2.1064, "step": 19635 }, { "epoch": 0.48, "learning_rate": 1.1002613147739777e-05, "loss": 2.2112, "step": 19636 }, { "epoch": 0.48, "learning_rate": 1.1001818836097294e-05, "loss": 2.0663, "step": 19637 }, { "epoch": 0.48, "learning_rate": 1.1001024518069892e-05, "loss": 2.077, "step": 19638 }, { "epoch": 0.48, "learning_rate": 1.1000230193662634e-05, "loss": 1.925, "step": 19639 }, { "epoch": 0.48, "learning_rate": 1.0999435862880584e-05, "loss": 1.9513, "step": 19640 }, { "epoch": 0.48, "learning_rate": 1.0998641525728802e-05, "loss": 1.962, "step": 19641 }, { "epoch": 0.48, "learning_rate": 1.0997847182212354e-05, "loss": 1.886, "step": 19642 }, { "epoch": 0.48, "learning_rate": 1.0997052832336298e-05, "loss": 2.0142, "step": 19643 }, { "epoch": 0.48, "learning_rate": 1.0996258476105698e-05, "loss": 2.0685, "step": 19644 }, { "epoch": 0.48, "learning_rate": 1.0995464113525622e-05, "loss": 2.2595, "step": 19645 }, { "epoch": 0.48, "learning_rate": 1.099466974460113e-05, "loss": 2.0251, "step": 19646 }, { "epoch": 0.48, "learning_rate": 1.0993875369337282e-05, "loss": 2.0759, "step": 19647 }, { "epoch": 0.48, "learning_rate": 1.0993080987739142e-05, "loss": 2.0591, "step": 19648 }, { "epoch": 0.48, "learning_rate": 1.0992286599811772e-05, "loss": 2.0863, "step": 19649 }, { "epoch": 0.48, "learning_rate": 1.0991492205560237e-05, "loss": 1.9334, "step": 19650 }, { "epoch": 0.48, "learning_rate": 1.0990697804989601e-05, "loss": 2.13, "step": 19651 }, { "epoch": 0.48, "learning_rate": 1.0989903398104922e-05, "loss": 2.0936, "step": 19652 }, { "epoch": 0.48, "learning_rate": 1.0989108984911268e-05, "loss": 2.0996, "step": 19653 }, { "epoch": 0.48, "learning_rate": 1.0988314565413696e-05, "loss": 2.0921, "step": 19654 }, { "epoch": 0.48, "learning_rate": 1.0987520139617278e-05, "loss": 2.1056, "step": 19655 }, { "epoch": 0.48, "learning_rate": 1.0986725707527071e-05, "loss": 1.9418, "step": 19656 }, { "epoch": 0.48, "learning_rate": 1.098593126914814e-05, "loss": 1.8995, "step": 19657 }, { "epoch": 0.48, "learning_rate": 1.0985136824485548e-05, "loss": 2.1456, "step": 19658 }, { "epoch": 0.48, "learning_rate": 1.0984342373544356e-05, "loss": 1.8152, "step": 19659 }, { "epoch": 0.48, "learning_rate": 1.098354791632963e-05, "loss": 2.0192, "step": 19660 }, { "epoch": 0.48, "learning_rate": 1.0982753452846433e-05, "loss": 1.9952, "step": 19661 }, { "epoch": 0.48, "learning_rate": 1.0981958983099828e-05, "loss": 1.9857, "step": 19662 }, { "epoch": 0.48, "learning_rate": 1.0981164507094879e-05, "loss": 1.979, "step": 19663 }, { "epoch": 0.48, "learning_rate": 1.0980370024836646e-05, "loss": 2.094, "step": 19664 }, { "epoch": 0.48, "learning_rate": 1.0979575536330199e-05, "loss": 2.035, "step": 19665 }, { "epoch": 0.48, "learning_rate": 1.0978781041580597e-05, "loss": 2.1216, "step": 19666 }, { "epoch": 0.48, "learning_rate": 1.0977986540592904e-05, "loss": 2.2275, "step": 19667 }, { "epoch": 0.48, "learning_rate": 1.0977192033372182e-05, "loss": 2.0517, "step": 19668 }, { "epoch": 0.48, "learning_rate": 1.0976397519923498e-05, "loss": 1.9427, "step": 19669 }, { "epoch": 0.48, "learning_rate": 1.0975603000251915e-05, "loss": 2.0714, "step": 19670 }, { "epoch": 0.48, "learning_rate": 1.0974808474362497e-05, "loss": 2.0161, "step": 19671 }, { "epoch": 0.48, "learning_rate": 1.0974013942260307e-05, "loss": 1.8187, "step": 19672 }, { "epoch": 0.48, "learning_rate": 1.0973219403950406e-05, "loss": 2.0682, "step": 19673 }, { "epoch": 0.48, "learning_rate": 1.0972424859437863e-05, "loss": 2.0658, "step": 19674 }, { "epoch": 0.48, "learning_rate": 1.0971630308727738e-05, "loss": 1.805, "step": 19675 }, { "epoch": 0.48, "learning_rate": 1.09708357518251e-05, "loss": 2.002, "step": 19676 }, { "epoch": 0.49, "learning_rate": 1.0970041188735003e-05, "loss": 2.0498, "step": 19677 }, { "epoch": 0.49, "learning_rate": 1.0969246619462524e-05, "loss": 2.046, "step": 19678 }, { "epoch": 0.49, "learning_rate": 1.0968452044012713e-05, "loss": 1.9574, "step": 19679 }, { "epoch": 0.49, "learning_rate": 1.0967657462390648e-05, "loss": 1.9423, "step": 19680 }, { "epoch": 0.49, "learning_rate": 1.0966862874601386e-05, "loss": 1.9825, "step": 19681 }, { "epoch": 0.49, "learning_rate": 1.096606828064999e-05, "loss": 2.039, "step": 19682 }, { "epoch": 0.49, "learning_rate": 1.0965273680541526e-05, "loss": 1.8287, "step": 19683 }, { "epoch": 0.49, "learning_rate": 1.0964479074281058e-05, "loss": 1.9413, "step": 19684 }, { "epoch": 0.49, "learning_rate": 1.0963684461873652e-05, "loss": 2.0486, "step": 19685 }, { "epoch": 0.49, "learning_rate": 1.0962889843324369e-05, "loss": 1.953, "step": 19686 }, { "epoch": 0.49, "learning_rate": 1.0962095218638276e-05, "loss": 2.1968, "step": 19687 }, { "epoch": 0.49, "learning_rate": 1.0961300587820436e-05, "loss": 2.082, "step": 19688 }, { "epoch": 0.49, "learning_rate": 1.0960505950875915e-05, "loss": 1.9663, "step": 19689 }, { "epoch": 0.49, "learning_rate": 1.0959711307809777e-05, "loss": 2.0027, "step": 19690 }, { "epoch": 0.49, "learning_rate": 1.0958916658627088e-05, "loss": 1.9854, "step": 19691 }, { "epoch": 0.49, "learning_rate": 1.0958122003332906e-05, "loss": 1.9223, "step": 19692 }, { "epoch": 0.49, "learning_rate": 1.0957327341932303e-05, "loss": 2.1105, "step": 19693 }, { "epoch": 0.49, "learning_rate": 1.095653267443034e-05, "loss": 2.021, "step": 19694 }, { "epoch": 0.49, "learning_rate": 1.0955738000832081e-05, "loss": 2.013, "step": 19695 }, { "epoch": 0.49, "learning_rate": 1.0954943321142594e-05, "loss": 2.1611, "step": 19696 }, { "epoch": 0.49, "learning_rate": 1.095414863536694e-05, "loss": 2.0854, "step": 19697 }, { "epoch": 0.49, "learning_rate": 1.095335394351019e-05, "loss": 1.9987, "step": 19698 }, { "epoch": 0.49, "learning_rate": 1.0952559245577403e-05, "loss": 2.0292, "step": 19699 }, { "epoch": 0.49, "learning_rate": 1.0951764541573646e-05, "loss": 1.924, "step": 19700 }, { "epoch": 0.49, "learning_rate": 1.0950969831503983e-05, "loss": 2.0487, "step": 19701 }, { "epoch": 0.49, "learning_rate": 1.0950175115373477e-05, "loss": 2.1546, "step": 19702 }, { "epoch": 0.49, "learning_rate": 1.09493803931872e-05, "loss": 2.0678, "step": 19703 }, { "epoch": 0.49, "learning_rate": 1.0948585664950206e-05, "loss": 1.8392, "step": 19704 }, { "epoch": 0.49, "learning_rate": 1.0947790930667572e-05, "loss": 2.0223, "step": 19705 }, { "epoch": 0.49, "learning_rate": 1.0946996190344357e-05, "loss": 2.0141, "step": 19706 }, { "epoch": 0.49, "learning_rate": 1.0946201443985623e-05, "loss": 2.0012, "step": 19707 }, { "epoch": 0.49, "learning_rate": 1.094540669159644e-05, "loss": 2.1506, "step": 19708 }, { "epoch": 0.49, "learning_rate": 1.0944611933181875e-05, "loss": 1.9515, "step": 19709 }, { "epoch": 0.49, "learning_rate": 1.0943817168746988e-05, "loss": 2.1453, "step": 19710 }, { "epoch": 0.49, "learning_rate": 1.0943022398296847e-05, "loss": 2.0086, "step": 19711 }, { "epoch": 0.49, "learning_rate": 1.0942227621836516e-05, "loss": 2.041, "step": 19712 }, { "epoch": 0.49, "learning_rate": 1.094143283937106e-05, "loss": 2.01, "step": 19713 }, { "epoch": 0.49, "learning_rate": 1.0940638050905549e-05, "loss": 1.9152, "step": 19714 }, { "epoch": 0.49, "learning_rate": 1.0939843256445046e-05, "loss": 1.996, "step": 19715 }, { "epoch": 0.49, "learning_rate": 1.0939048455994615e-05, "loss": 2.067, "step": 19716 }, { "epoch": 0.49, "learning_rate": 1.0938253649559318e-05, "loss": 2.0075, "step": 19717 }, { "epoch": 0.49, "learning_rate": 1.0937458837144227e-05, "loss": 2.102, "step": 19718 }, { "epoch": 0.49, "learning_rate": 1.0936664018754409e-05, "loss": 2.0193, "step": 19719 }, { "epoch": 0.49, "learning_rate": 1.0935869194394922e-05, "loss": 2.0335, "step": 19720 }, { "epoch": 0.49, "learning_rate": 1.0935074364070834e-05, "loss": 2.3107, "step": 19721 }, { "epoch": 0.49, "learning_rate": 1.0934279527787214e-05, "loss": 1.9685, "step": 19722 }, { "epoch": 0.49, "learning_rate": 1.0933484685549127e-05, "loss": 2.0469, "step": 19723 }, { "epoch": 0.49, "learning_rate": 1.0932689837361638e-05, "loss": 1.9722, "step": 19724 }, { "epoch": 0.49, "learning_rate": 1.0931894983229814e-05, "loss": 1.9221, "step": 19725 }, { "epoch": 0.49, "learning_rate": 1.0931100123158717e-05, "loss": 1.8465, "step": 19726 }, { "epoch": 0.49, "learning_rate": 1.0930305257153417e-05, "loss": 2.0929, "step": 19727 }, { "epoch": 0.49, "learning_rate": 1.092951038521898e-05, "loss": 2.0766, "step": 19728 }, { "epoch": 0.49, "learning_rate": 1.0928715507360467e-05, "loss": 2.128, "step": 19729 }, { "epoch": 0.49, "learning_rate": 1.0927920623582948e-05, "loss": 2.0376, "step": 19730 }, { "epoch": 0.49, "learning_rate": 1.0927125733891488e-05, "loss": 2.0831, "step": 19731 }, { "epoch": 0.49, "learning_rate": 1.0926330838291155e-05, "loss": 1.9525, "step": 19732 }, { "epoch": 0.49, "learning_rate": 1.0925535936787014e-05, "loss": 2.0074, "step": 19733 }, { "epoch": 0.49, "learning_rate": 1.0924741029384131e-05, "loss": 2.0951, "step": 19734 }, { "epoch": 0.49, "learning_rate": 1.0923946116087571e-05, "loss": 2.0586, "step": 19735 }, { "epoch": 0.49, "learning_rate": 1.0923151196902402e-05, "loss": 2.2542, "step": 19736 }, { "epoch": 0.49, "learning_rate": 1.092235627183369e-05, "loss": 1.9829, "step": 19737 }, { "epoch": 0.49, "learning_rate": 1.09215613408865e-05, "loss": 1.8686, "step": 19738 }, { "epoch": 0.49, "learning_rate": 1.09207664040659e-05, "loss": 1.973, "step": 19739 }, { "epoch": 0.49, "learning_rate": 1.0919971461376955e-05, "loss": 1.9505, "step": 19740 }, { "epoch": 0.49, "learning_rate": 1.0919176512824729e-05, "loss": 1.976, "step": 19741 }, { "epoch": 0.49, "learning_rate": 1.0918381558414295e-05, "loss": 2.1561, "step": 19742 }, { "epoch": 0.49, "learning_rate": 1.0917586598150719e-05, "loss": 1.8325, "step": 19743 }, { "epoch": 0.49, "learning_rate": 1.091679163203906e-05, "loss": 1.9537, "step": 19744 }, { "epoch": 0.49, "learning_rate": 1.0915996660084391e-05, "loss": 2.0033, "step": 19745 }, { "epoch": 0.49, "learning_rate": 1.0915201682291774e-05, "loss": 2.0251, "step": 19746 }, { "epoch": 0.49, "learning_rate": 1.0914406698666282e-05, "loss": 1.9256, "step": 19747 }, { "epoch": 0.49, "learning_rate": 1.0913611709212974e-05, "loss": 2.0601, "step": 19748 }, { "epoch": 0.49, "learning_rate": 1.0912816713936924e-05, "loss": 2.1452, "step": 19749 }, { "epoch": 0.49, "learning_rate": 1.0912021712843197e-05, "loss": 1.8917, "step": 19750 }, { "epoch": 0.49, "learning_rate": 1.0911226705936854e-05, "loss": 1.9229, "step": 19751 }, { "epoch": 0.49, "learning_rate": 1.0910431693222972e-05, "loss": 2.0605, "step": 19752 }, { "epoch": 0.49, "learning_rate": 1.0909636674706608e-05, "loss": 2.1831, "step": 19753 }, { "epoch": 0.49, "learning_rate": 1.0908841650392834e-05, "loss": 1.987, "step": 19754 }, { "epoch": 0.49, "learning_rate": 1.0908046620286717e-05, "loss": 2.0428, "step": 19755 }, { "epoch": 0.49, "learning_rate": 1.0907251584393322e-05, "loss": 1.9379, "step": 19756 }, { "epoch": 0.49, "learning_rate": 1.0906456542717714e-05, "loss": 2.3187, "step": 19757 }, { "epoch": 0.49, "learning_rate": 1.0905661495264968e-05, "loss": 2.0352, "step": 19758 }, { "epoch": 0.49, "learning_rate": 1.0904866442040142e-05, "loss": 2.0347, "step": 19759 }, { "epoch": 0.49, "learning_rate": 1.090407138304831e-05, "loss": 2.1621, "step": 19760 }, { "epoch": 0.49, "learning_rate": 1.0903276318294536e-05, "loss": 2.0017, "step": 19761 }, { "epoch": 0.49, "learning_rate": 1.0902481247783887e-05, "loss": 2.0937, "step": 19762 }, { "epoch": 0.49, "learning_rate": 1.0901686171521434e-05, "loss": 1.9991, "step": 19763 }, { "epoch": 0.49, "learning_rate": 1.0900891089512238e-05, "loss": 2.0404, "step": 19764 }, { "epoch": 0.49, "learning_rate": 1.090009600176137e-05, "loss": 2.0355, "step": 19765 }, { "epoch": 0.49, "learning_rate": 1.0899300908273893e-05, "loss": 1.8154, "step": 19766 }, { "epoch": 0.49, "learning_rate": 1.0898505809054884e-05, "loss": 1.9771, "step": 19767 }, { "epoch": 0.49, "learning_rate": 1.0897710704109403e-05, "loss": 2.2981, "step": 19768 }, { "epoch": 0.49, "learning_rate": 1.0896915593442521e-05, "loss": 2.0894, "step": 19769 }, { "epoch": 0.49, "learning_rate": 1.0896120477059298e-05, "loss": 2.0522, "step": 19770 }, { "epoch": 0.49, "learning_rate": 1.0895325354964812e-05, "loss": 1.8396, "step": 19771 }, { "epoch": 0.49, "learning_rate": 1.0894530227164125e-05, "loss": 2.0604, "step": 19772 }, { "epoch": 0.49, "learning_rate": 1.0893735093662306e-05, "loss": 2.1811, "step": 19773 }, { "epoch": 0.49, "learning_rate": 1.0892939954464418e-05, "loss": 2.115, "step": 19774 }, { "epoch": 0.49, "learning_rate": 1.0892144809575535e-05, "loss": 1.9015, "step": 19775 }, { "epoch": 0.49, "learning_rate": 1.0891349659000723e-05, "loss": 2.1699, "step": 19776 }, { "epoch": 0.49, "learning_rate": 1.089055450274505e-05, "loss": 2.061, "step": 19777 }, { "epoch": 0.49, "learning_rate": 1.0889759340813581e-05, "loss": 1.9321, "step": 19778 }, { "epoch": 0.49, "learning_rate": 1.0888964173211389e-05, "loss": 2.0423, "step": 19779 }, { "epoch": 0.49, "learning_rate": 1.0888168999943534e-05, "loss": 2.0705, "step": 19780 }, { "epoch": 0.49, "learning_rate": 1.0887373821015093e-05, "loss": 2.1001, "step": 19781 }, { "epoch": 0.49, "learning_rate": 1.0886578636431127e-05, "loss": 2.0189, "step": 19782 }, { "epoch": 0.49, "learning_rate": 1.0885783446196704e-05, "loss": 1.9119, "step": 19783 }, { "epoch": 0.49, "learning_rate": 1.0884988250316897e-05, "loss": 2.0813, "step": 19784 }, { "epoch": 0.49, "learning_rate": 1.0884193048796771e-05, "loss": 2.0703, "step": 19785 }, { "epoch": 0.49, "learning_rate": 1.0883397841641396e-05, "loss": 2.2789, "step": 19786 }, { "epoch": 0.49, "learning_rate": 1.0882602628855839e-05, "loss": 1.9778, "step": 19787 }, { "epoch": 0.49, "learning_rate": 1.0881807410445167e-05, "loss": 2.1043, "step": 19788 }, { "epoch": 0.49, "learning_rate": 1.0881012186414448e-05, "loss": 2.2838, "step": 19789 }, { "epoch": 0.49, "learning_rate": 1.0880216956768752e-05, "loss": 2.1357, "step": 19790 }, { "epoch": 0.49, "learning_rate": 1.0879421721513146e-05, "loss": 2.0348, "step": 19791 }, { "epoch": 0.49, "learning_rate": 1.0878626480652699e-05, "loss": 2.1536, "step": 19792 }, { "epoch": 0.49, "learning_rate": 1.087783123419248e-05, "loss": 2.0246, "step": 19793 }, { "epoch": 0.49, "learning_rate": 1.0877035982137554e-05, "loss": 1.8876, "step": 19794 }, { "epoch": 0.49, "learning_rate": 1.0876240724492993e-05, "loss": 2.0313, "step": 19795 }, { "epoch": 0.49, "learning_rate": 1.0875445461263866e-05, "loss": 2.1784, "step": 19796 }, { "epoch": 0.49, "learning_rate": 1.0874650192455238e-05, "loss": 2.0923, "step": 19797 }, { "epoch": 0.49, "learning_rate": 1.0873854918072183e-05, "loss": 2.1006, "step": 19798 }, { "epoch": 0.49, "learning_rate": 1.087305963811976e-05, "loss": 2.1357, "step": 19799 }, { "epoch": 0.49, "learning_rate": 1.0872264352603045e-05, "loss": 2.0658, "step": 19800 }, { "epoch": 0.49, "learning_rate": 1.0871469061527107e-05, "loss": 2.2869, "step": 19801 }, { "epoch": 0.49, "learning_rate": 1.0870673764897011e-05, "loss": 1.9792, "step": 19802 }, { "epoch": 0.49, "learning_rate": 1.0869878462717828e-05, "loss": 2.1385, "step": 19803 }, { "epoch": 0.49, "learning_rate": 1.0869083154994626e-05, "loss": 2.1348, "step": 19804 }, { "epoch": 0.49, "learning_rate": 1.0868287841732474e-05, "loss": 1.9574, "step": 19805 }, { "epoch": 0.49, "learning_rate": 1.0867492522936443e-05, "loss": 2.0763, "step": 19806 }, { "epoch": 0.49, "learning_rate": 1.0866697198611598e-05, "loss": 2.1127, "step": 19807 }, { "epoch": 0.49, "learning_rate": 1.0865901868763006e-05, "loss": 2.0973, "step": 19808 }, { "epoch": 0.49, "learning_rate": 1.0865106533395743e-05, "loss": 2.1051, "step": 19809 }, { "epoch": 0.49, "learning_rate": 1.0864311192514868e-05, "loss": 2.131, "step": 19810 }, { "epoch": 0.49, "learning_rate": 1.0863515846125463e-05, "loss": 2.225, "step": 19811 }, { "epoch": 0.49, "learning_rate": 1.0862720494232588e-05, "loss": 1.9669, "step": 19812 }, { "epoch": 0.49, "learning_rate": 1.0861925136841314e-05, "loss": 2.1093, "step": 19813 }, { "epoch": 0.49, "learning_rate": 1.086112977395671e-05, "loss": 2.0572, "step": 19814 }, { "epoch": 0.49, "learning_rate": 1.0860334405583846e-05, "loss": 2.0564, "step": 19815 }, { "epoch": 0.49, "learning_rate": 1.085953903172779e-05, "loss": 2.1235, "step": 19816 }, { "epoch": 0.49, "learning_rate": 1.0858743652393613e-05, "loss": 1.9916, "step": 19817 }, { "epoch": 0.49, "learning_rate": 1.085794826758638e-05, "loss": 1.9874, "step": 19818 }, { "epoch": 0.49, "learning_rate": 1.0857152877311164e-05, "loss": 2.0264, "step": 19819 }, { "epoch": 0.49, "learning_rate": 1.0856357481573034e-05, "loss": 2.0243, "step": 19820 }, { "epoch": 0.49, "learning_rate": 1.085556208037706e-05, "loss": 2.0037, "step": 19821 }, { "epoch": 0.49, "learning_rate": 1.0854766673728309e-05, "loss": 1.9572, "step": 19822 }, { "epoch": 0.49, "learning_rate": 1.085397126163185e-05, "loss": 2.0794, "step": 19823 }, { "epoch": 0.49, "learning_rate": 1.0853175844092757e-05, "loss": 2.2179, "step": 19824 }, { "epoch": 0.49, "learning_rate": 1.0852380421116094e-05, "loss": 2.0984, "step": 19825 }, { "epoch": 0.49, "learning_rate": 1.0851584992706933e-05, "loss": 1.9143, "step": 19826 }, { "epoch": 0.49, "learning_rate": 1.0850789558870342e-05, "loss": 2.0721, "step": 19827 }, { "epoch": 0.49, "learning_rate": 1.084999411961139e-05, "loss": 2.0999, "step": 19828 }, { "epoch": 0.49, "learning_rate": 1.0849198674935155e-05, "loss": 2.0444, "step": 19829 }, { "epoch": 0.49, "learning_rate": 1.0848403224846698e-05, "loss": 1.9136, "step": 19830 }, { "epoch": 0.49, "learning_rate": 1.084760776935109e-05, "loss": 2.117, "step": 19831 }, { "epoch": 0.49, "learning_rate": 1.0846812308453401e-05, "loss": 2.1645, "step": 19832 }, { "epoch": 0.49, "learning_rate": 1.0846016842158699e-05, "loss": 2.0567, "step": 19833 }, { "epoch": 0.49, "learning_rate": 1.0845221370472058e-05, "loss": 2.0441, "step": 19834 }, { "epoch": 0.49, "learning_rate": 1.0844425893398549e-05, "loss": 2.0252, "step": 19835 }, { "epoch": 0.49, "learning_rate": 1.0843630410943233e-05, "loss": 1.9539, "step": 19836 }, { "epoch": 0.49, "learning_rate": 1.084283492311119e-05, "loss": 1.9782, "step": 19837 }, { "epoch": 0.49, "learning_rate": 1.084203942990748e-05, "loss": 2.0225, "step": 19838 }, { "epoch": 0.49, "learning_rate": 1.0841243931337184e-05, "loss": 1.9609, "step": 19839 }, { "epoch": 0.49, "learning_rate": 1.0840448427405363e-05, "loss": 2.0221, "step": 19840 }, { "epoch": 0.49, "learning_rate": 1.0839652918117092e-05, "loss": 2.0311, "step": 19841 }, { "epoch": 0.49, "learning_rate": 1.083885740347744e-05, "loss": 2.2379, "step": 19842 }, { "epoch": 0.49, "learning_rate": 1.0838061883491473e-05, "loss": 2.1947, "step": 19843 }, { "epoch": 0.49, "learning_rate": 1.0837266358164267e-05, "loss": 1.7386, "step": 19844 }, { "epoch": 0.49, "learning_rate": 1.083647082750089e-05, "loss": 2.0175, "step": 19845 }, { "epoch": 0.49, "learning_rate": 1.083567529150641e-05, "loss": 1.856, "step": 19846 }, { "epoch": 0.49, "learning_rate": 1.0834879750185898e-05, "loss": 1.9764, "step": 19847 }, { "epoch": 0.49, "learning_rate": 1.0834084203544428e-05, "loss": 1.974, "step": 19848 }, { "epoch": 0.49, "learning_rate": 1.0833288651587068e-05, "loss": 2.0074, "step": 19849 }, { "epoch": 0.49, "learning_rate": 1.0832493094318888e-05, "loss": 2.2203, "step": 19850 }, { "epoch": 0.49, "learning_rate": 1.0831697531744955e-05, "loss": 1.9699, "step": 19851 }, { "epoch": 0.49, "learning_rate": 1.0830901963870344e-05, "loss": 2.0475, "step": 19852 }, { "epoch": 0.49, "learning_rate": 1.0830106390700126e-05, "loss": 2.1945, "step": 19853 }, { "epoch": 0.49, "learning_rate": 1.0829310812239367e-05, "loss": 2.1999, "step": 19854 }, { "epoch": 0.49, "learning_rate": 1.0828515228493139e-05, "loss": 1.9498, "step": 19855 }, { "epoch": 0.49, "learning_rate": 1.0827719639466516e-05, "loss": 2.0978, "step": 19856 }, { "epoch": 0.49, "learning_rate": 1.0826924045164564e-05, "loss": 2.1003, "step": 19857 }, { "epoch": 0.49, "learning_rate": 1.0826128445592359e-05, "loss": 2.1171, "step": 19858 }, { "epoch": 0.49, "learning_rate": 1.0825332840754965e-05, "loss": 2.108, "step": 19859 }, { "epoch": 0.49, "learning_rate": 1.0824537230657459e-05, "loss": 2.0006, "step": 19860 }, { "epoch": 0.49, "learning_rate": 1.0823741615304906e-05, "loss": 1.9963, "step": 19861 }, { "epoch": 0.49, "learning_rate": 1.0822945994702378e-05, "loss": 1.9733, "step": 19862 }, { "epoch": 0.49, "learning_rate": 1.0822150368854947e-05, "loss": 2.0444, "step": 19863 }, { "epoch": 0.49, "learning_rate": 1.0821354737767686e-05, "loss": 1.8882, "step": 19864 }, { "epoch": 0.49, "learning_rate": 1.0820559101445664e-05, "loss": 2.0326, "step": 19865 }, { "epoch": 0.49, "learning_rate": 1.081976345989395e-05, "loss": 2.2098, "step": 19866 }, { "epoch": 0.49, "learning_rate": 1.0818967813117615e-05, "loss": 1.8583, "step": 19867 }, { "epoch": 0.49, "learning_rate": 1.0818172161121736e-05, "loss": 2.0136, "step": 19868 }, { "epoch": 0.49, "learning_rate": 1.0817376503911376e-05, "loss": 2.1371, "step": 19869 }, { "epoch": 0.49, "learning_rate": 1.0816580841491611e-05, "loss": 1.9944, "step": 19870 }, { "epoch": 0.49, "learning_rate": 1.0815785173867505e-05, "loss": 1.9383, "step": 19871 }, { "epoch": 0.49, "learning_rate": 1.0814989501044138e-05, "loss": 2.1713, "step": 19872 }, { "epoch": 0.49, "learning_rate": 1.0814193823026578e-05, "loss": 1.9854, "step": 19873 }, { "epoch": 0.49, "learning_rate": 1.0813398139819893e-05, "loss": 1.8861, "step": 19874 }, { "epoch": 0.49, "learning_rate": 1.081260245142916e-05, "loss": 2.0618, "step": 19875 }, { "epoch": 0.49, "learning_rate": 1.0811806757859441e-05, "loss": 1.9552, "step": 19876 }, { "epoch": 0.49, "learning_rate": 1.0811011059115819e-05, "loss": 1.7555, "step": 19877 }, { "epoch": 0.49, "learning_rate": 1.0810215355203358e-05, "loss": 2.054, "step": 19878 }, { "epoch": 0.49, "learning_rate": 1.0809419646127131e-05, "loss": 2.134, "step": 19879 }, { "epoch": 0.49, "learning_rate": 1.0808623931892204e-05, "loss": 1.8972, "step": 19880 }, { "epoch": 0.49, "learning_rate": 1.0807828212503655e-05, "loss": 1.9419, "step": 19881 }, { "epoch": 0.49, "learning_rate": 1.0807032487966556e-05, "loss": 1.8033, "step": 19882 }, { "epoch": 0.49, "learning_rate": 1.0806236758285975e-05, "loss": 2.0534, "step": 19883 }, { "epoch": 0.49, "learning_rate": 1.0805441023466986e-05, "loss": 1.8618, "step": 19884 }, { "epoch": 0.49, "learning_rate": 1.0804645283514656e-05, "loss": 1.888, "step": 19885 }, { "epoch": 0.49, "learning_rate": 1.0803849538434061e-05, "loss": 2.1678, "step": 19886 }, { "epoch": 0.49, "learning_rate": 1.0803053788230271e-05, "loss": 1.9894, "step": 19887 }, { "epoch": 0.49, "learning_rate": 1.0802258032908357e-05, "loss": 2.1594, "step": 19888 }, { "epoch": 0.49, "learning_rate": 1.0801462272473388e-05, "loss": 1.9779, "step": 19889 }, { "epoch": 0.49, "learning_rate": 1.0800666506930444e-05, "loss": 1.9236, "step": 19890 }, { "epoch": 0.49, "learning_rate": 1.0799870736284588e-05, "loss": 1.9485, "step": 19891 }, { "epoch": 0.49, "learning_rate": 1.0799074960540897e-05, "loss": 1.9671, "step": 19892 }, { "epoch": 0.49, "learning_rate": 1.079827917970444e-05, "loss": 2.1353, "step": 19893 }, { "epoch": 0.49, "learning_rate": 1.0797483393780292e-05, "loss": 2.126, "step": 19894 }, { "epoch": 0.49, "learning_rate": 1.0796687602773522e-05, "loss": 2.0476, "step": 19895 }, { "epoch": 0.49, "learning_rate": 1.0795891806689198e-05, "loss": 2.1449, "step": 19896 }, { "epoch": 0.49, "learning_rate": 1.07950960055324e-05, "loss": 1.9714, "step": 19897 }, { "epoch": 0.49, "learning_rate": 1.0794300199308192e-05, "loss": 2.1437, "step": 19898 }, { "epoch": 0.49, "learning_rate": 1.0793504388021655e-05, "loss": 1.9167, "step": 19899 }, { "epoch": 0.49, "learning_rate": 1.0792708571677852e-05, "loss": 2.0177, "step": 19900 }, { "epoch": 0.49, "learning_rate": 1.0791912750281861e-05, "loss": 1.9435, "step": 19901 }, { "epoch": 0.49, "learning_rate": 1.0791116923838752e-05, "loss": 2.0462, "step": 19902 }, { "epoch": 0.49, "learning_rate": 1.0790321092353598e-05, "loss": 1.9858, "step": 19903 }, { "epoch": 0.49, "learning_rate": 1.0789525255831467e-05, "loss": 2.2022, "step": 19904 }, { "epoch": 0.49, "learning_rate": 1.0788729414277435e-05, "loss": 2.1472, "step": 19905 }, { "epoch": 0.49, "learning_rate": 1.0787933567696576e-05, "loss": 2.1112, "step": 19906 }, { "epoch": 0.49, "learning_rate": 1.0787137716093956e-05, "loss": 2.1364, "step": 19907 }, { "epoch": 0.49, "learning_rate": 1.0786341859474654e-05, "loss": 1.931, "step": 19908 }, { "epoch": 0.49, "learning_rate": 1.0785545997843738e-05, "loss": 1.918, "step": 19909 }, { "epoch": 0.49, "learning_rate": 1.0784750131206281e-05, "loss": 2.0337, "step": 19910 }, { "epoch": 0.49, "learning_rate": 1.0783954259567356e-05, "loss": 2.0901, "step": 19911 }, { "epoch": 0.49, "learning_rate": 1.0783158382932036e-05, "loss": 2.0964, "step": 19912 }, { "epoch": 0.49, "learning_rate": 1.078236250130539e-05, "loss": 1.952, "step": 19913 }, { "epoch": 0.49, "learning_rate": 1.0781566614692493e-05, "loss": 1.9326, "step": 19914 }, { "epoch": 0.49, "learning_rate": 1.078077072309842e-05, "loss": 2.0598, "step": 19915 }, { "epoch": 0.49, "learning_rate": 1.0779974826528237e-05, "loss": 2.038, "step": 19916 }, { "epoch": 0.49, "learning_rate": 1.0779178924987024e-05, "loss": 2.0652, "step": 19917 }, { "epoch": 0.49, "learning_rate": 1.0778383018479848e-05, "loss": 1.8833, "step": 19918 }, { "epoch": 0.49, "learning_rate": 1.0777587107011784e-05, "loss": 2.0154, "step": 19919 }, { "epoch": 0.49, "learning_rate": 1.0776791190587902e-05, "loss": 1.9473, "step": 19920 }, { "epoch": 0.49, "learning_rate": 1.0775995269213279e-05, "loss": 1.9272, "step": 19921 }, { "epoch": 0.49, "learning_rate": 1.0775199342892986e-05, "loss": 2.0067, "step": 19922 }, { "epoch": 0.49, "learning_rate": 1.0774403411632091e-05, "loss": 2.1714, "step": 19923 }, { "epoch": 0.49, "learning_rate": 1.0773607475435672e-05, "loss": 2.0942, "step": 19924 }, { "epoch": 0.49, "learning_rate": 1.07728115343088e-05, "loss": 2.1253, "step": 19925 }, { "epoch": 0.49, "learning_rate": 1.0772015588256552e-05, "loss": 2.0163, "step": 19926 }, { "epoch": 0.49, "learning_rate": 1.0771219637283994e-05, "loss": 1.9366, "step": 19927 }, { "epoch": 0.49, "learning_rate": 1.0770423681396206e-05, "loss": 2.0368, "step": 19928 }, { "epoch": 0.49, "learning_rate": 1.0769627720598252e-05, "loss": 2.0577, "step": 19929 }, { "epoch": 0.49, "learning_rate": 1.076883175489521e-05, "loss": 2.1396, "step": 19930 }, { "epoch": 0.49, "learning_rate": 1.0768035784292155e-05, "loss": 1.95, "step": 19931 }, { "epoch": 0.49, "learning_rate": 1.0767239808794154e-05, "loss": 2.133, "step": 19932 }, { "epoch": 0.49, "learning_rate": 1.0766443828406288e-05, "loss": 2.1435, "step": 19933 }, { "epoch": 0.49, "learning_rate": 1.0765647843133622e-05, "loss": 1.9886, "step": 19934 }, { "epoch": 0.49, "learning_rate": 1.0764851852981235e-05, "loss": 2.1489, "step": 19935 }, { "epoch": 0.49, "learning_rate": 1.0764055857954198e-05, "loss": 2.0925, "step": 19936 }, { "epoch": 0.49, "learning_rate": 1.0763259858057584e-05, "loss": 1.9768, "step": 19937 }, { "epoch": 0.49, "learning_rate": 1.0762463853296467e-05, "loss": 2.0416, "step": 19938 }, { "epoch": 0.49, "learning_rate": 1.0761667843675916e-05, "loss": 1.8358, "step": 19939 }, { "epoch": 0.49, "learning_rate": 1.0760871829201011e-05, "loss": 2.1411, "step": 19940 }, { "epoch": 0.49, "learning_rate": 1.076007580987682e-05, "loss": 2.1016, "step": 19941 }, { "epoch": 0.49, "learning_rate": 1.0759279785708419e-05, "loss": 1.8624, "step": 19942 }, { "epoch": 0.49, "learning_rate": 1.0758483756700879e-05, "loss": 1.9107, "step": 19943 }, { "epoch": 0.49, "learning_rate": 1.0757687722859276e-05, "loss": 1.9468, "step": 19944 }, { "epoch": 0.49, "learning_rate": 1.0756891684188682e-05, "loss": 2.0423, "step": 19945 }, { "epoch": 0.49, "learning_rate": 1.0756095640694173e-05, "loss": 2.1292, "step": 19946 }, { "epoch": 0.49, "learning_rate": 1.0755299592380817e-05, "loss": 2.031, "step": 19947 }, { "epoch": 0.49, "learning_rate": 1.0754503539253693e-05, "loss": 2.1393, "step": 19948 }, { "epoch": 0.49, "learning_rate": 1.075370748131787e-05, "loss": 2.0182, "step": 19949 }, { "epoch": 0.49, "learning_rate": 1.0752911418578424e-05, "loss": 2.011, "step": 19950 }, { "epoch": 0.49, "learning_rate": 1.0752115351040426e-05, "loss": 2.0772, "step": 19951 }, { "epoch": 0.49, "learning_rate": 1.0751319278708956e-05, "loss": 1.8975, "step": 19952 }, { "epoch": 0.49, "learning_rate": 1.0750523201589082e-05, "loss": 2.064, "step": 19953 }, { "epoch": 0.49, "learning_rate": 1.0749727119685878e-05, "loss": 2.186, "step": 19954 }, { "epoch": 0.49, "learning_rate": 1.074893103300442e-05, "loss": 2.1673, "step": 19955 }, { "epoch": 0.49, "learning_rate": 1.0748134941549779e-05, "loss": 1.8181, "step": 19956 }, { "epoch": 0.49, "learning_rate": 1.0747338845327032e-05, "loss": 1.943, "step": 19957 }, { "epoch": 0.49, "learning_rate": 1.0746542744341249e-05, "loss": 2.0727, "step": 19958 }, { "epoch": 0.49, "learning_rate": 1.0745746638597507e-05, "loss": 1.9562, "step": 19959 }, { "epoch": 0.49, "learning_rate": 1.0744950528100876e-05, "loss": 1.965, "step": 19960 }, { "epoch": 0.49, "learning_rate": 1.0744154412856436e-05, "loss": 1.917, "step": 19961 }, { "epoch": 0.49, "learning_rate": 1.0743358292869258e-05, "loss": 1.9815, "step": 19962 }, { "epoch": 0.49, "learning_rate": 1.0742562168144411e-05, "loss": 2.2172, "step": 19963 }, { "epoch": 0.49, "learning_rate": 1.0741766038686976e-05, "loss": 2.1705, "step": 19964 }, { "epoch": 0.49, "learning_rate": 1.0740969904502025e-05, "loss": 1.8765, "step": 19965 }, { "epoch": 0.49, "learning_rate": 1.0740173765594633e-05, "loss": 1.9826, "step": 19966 }, { "epoch": 0.49, "learning_rate": 1.0739377621969867e-05, "loss": 2.1389, "step": 19967 }, { "epoch": 0.49, "learning_rate": 1.073858147363281e-05, "loss": 1.7983, "step": 19968 }, { "epoch": 0.49, "learning_rate": 1.0737785320588531e-05, "loss": 2.0985, "step": 19969 }, { "epoch": 0.49, "learning_rate": 1.0736989162842107e-05, "loss": 2.1008, "step": 19970 }, { "epoch": 0.49, "learning_rate": 1.0736193000398612e-05, "loss": 2.0781, "step": 19971 }, { "epoch": 0.49, "learning_rate": 1.0735396833263119e-05, "loss": 1.9913, "step": 19972 }, { "epoch": 0.49, "learning_rate": 1.0734600661440699e-05, "loss": 2.0657, "step": 19973 }, { "epoch": 0.49, "learning_rate": 1.0733804484936432e-05, "loss": 2.0864, "step": 19974 }, { "epoch": 0.49, "learning_rate": 1.073300830375539e-05, "loss": 2.0538, "step": 19975 }, { "epoch": 0.49, "learning_rate": 1.0732212117902646e-05, "loss": 2.0995, "step": 19976 }, { "epoch": 0.49, "learning_rate": 1.0731415927383275e-05, "loss": 2.0018, "step": 19977 }, { "epoch": 0.49, "learning_rate": 1.0730619732202351e-05, "loss": 2.007, "step": 19978 }, { "epoch": 0.49, "learning_rate": 1.0729823532364954e-05, "loss": 1.997, "step": 19979 }, { "epoch": 0.49, "learning_rate": 1.0729027327876151e-05, "loss": 1.9707, "step": 19980 }, { "epoch": 0.49, "learning_rate": 1.0728231118741022e-05, "loss": 1.9762, "step": 19981 }, { "epoch": 0.49, "learning_rate": 1.0727434904964635e-05, "loss": 2.0577, "step": 19982 }, { "epoch": 0.49, "learning_rate": 1.072663868655207e-05, "loss": 2.0166, "step": 19983 }, { "epoch": 0.49, "learning_rate": 1.0725842463508399e-05, "loss": 1.9958, "step": 19984 }, { "epoch": 0.49, "learning_rate": 1.0725046235838696e-05, "loss": 2.2764, "step": 19985 }, { "epoch": 0.49, "learning_rate": 1.072425000354804e-05, "loss": 2.0696, "step": 19986 }, { "epoch": 0.49, "learning_rate": 1.0723453766641502e-05, "loss": 1.9765, "step": 19987 }, { "epoch": 0.49, "learning_rate": 1.0722657525124156e-05, "loss": 2.1046, "step": 19988 }, { "epoch": 0.49, "learning_rate": 1.072186127900108e-05, "loss": 1.9466, "step": 19989 }, { "epoch": 0.49, "learning_rate": 1.0721065028277347e-05, "loss": 1.9027, "step": 19990 }, { "epoch": 0.49, "learning_rate": 1.0720268772958031e-05, "loss": 1.963, "step": 19991 }, { "epoch": 0.49, "learning_rate": 1.0719472513048203e-05, "loss": 2.11, "step": 19992 }, { "epoch": 0.49, "learning_rate": 1.0718676248552949e-05, "loss": 1.9325, "step": 19993 }, { "epoch": 0.49, "learning_rate": 1.0717879979477332e-05, "loss": 2.0724, "step": 19994 }, { "epoch": 0.49, "learning_rate": 1.0717083705826434e-05, "loss": 1.937, "step": 19995 }, { "epoch": 0.49, "learning_rate": 1.071628742760533e-05, "loss": 1.8883, "step": 19996 }, { "epoch": 0.49, "learning_rate": 1.0715491144819088e-05, "loss": 1.9435, "step": 19997 }, { "epoch": 0.49, "learning_rate": 1.0714694857472792e-05, "loss": 2.0626, "step": 19998 }, { "epoch": 0.49, "learning_rate": 1.0713898565571512e-05, "loss": 2.1793, "step": 19999 }, { "epoch": 0.49, "learning_rate": 1.0713102269120324e-05, "loss": 1.9984, "step": 20000 }, { "epoch": 0.49, "learning_rate": 1.07123059681243e-05, "loss": 1.8317, "step": 20001 }, { "epoch": 0.49, "learning_rate": 1.071150966258852e-05, "loss": 2.0474, "step": 20002 }, { "epoch": 0.49, "learning_rate": 1.0710713352518056e-05, "loss": 1.9947, "step": 20003 }, { "epoch": 0.49, "learning_rate": 1.0709917037917984e-05, "loss": 2.2237, "step": 20004 }, { "epoch": 0.49, "learning_rate": 1.0709120718793382e-05, "loss": 2.1157, "step": 20005 }, { "epoch": 0.49, "learning_rate": 1.0708324395149323e-05, "loss": 1.9145, "step": 20006 }, { "epoch": 0.49, "learning_rate": 1.0707528066990877e-05, "loss": 2.0912, "step": 20007 }, { "epoch": 0.49, "learning_rate": 1.0706731734323128e-05, "loss": 2.0535, "step": 20008 }, { "epoch": 0.49, "learning_rate": 1.0705935397151146e-05, "loss": 2.0882, "step": 20009 }, { "epoch": 0.49, "learning_rate": 1.070513905548001e-05, "loss": 2.066, "step": 20010 }, { "epoch": 0.49, "learning_rate": 1.0704342709314787e-05, "loss": 2.1077, "step": 20011 }, { "epoch": 0.49, "learning_rate": 1.0703546358660561e-05, "loss": 2.1613, "step": 20012 }, { "epoch": 0.49, "learning_rate": 1.0702750003522409e-05, "loss": 2.0008, "step": 20013 }, { "epoch": 0.49, "learning_rate": 1.0701953643905398e-05, "loss": 2.1404, "step": 20014 }, { "epoch": 0.49, "learning_rate": 1.070115727981461e-05, "loss": 2.0429, "step": 20015 }, { "epoch": 0.49, "learning_rate": 1.0700360911255118e-05, "loss": 2.0513, "step": 20016 }, { "epoch": 0.49, "learning_rate": 1.0699564538231994e-05, "loss": 2.0892, "step": 20017 }, { "epoch": 0.49, "learning_rate": 1.0698768160750322e-05, "loss": 1.9708, "step": 20018 }, { "epoch": 0.49, "learning_rate": 1.069797177881517e-05, "loss": 2.0443, "step": 20019 }, { "epoch": 0.49, "learning_rate": 1.0697175392431616e-05, "loss": 2.0241, "step": 20020 }, { "epoch": 0.49, "learning_rate": 1.0696379001604735e-05, "loss": 2.033, "step": 20021 }, { "epoch": 0.49, "learning_rate": 1.0695582606339608e-05, "loss": 2.13, "step": 20022 }, { "epoch": 0.49, "learning_rate": 1.0694786206641304e-05, "loss": 2.0841, "step": 20023 }, { "epoch": 0.49, "learning_rate": 1.06939898025149e-05, "loss": 2.0136, "step": 20024 }, { "epoch": 0.49, "learning_rate": 1.0693193393965478e-05, "loss": 2.2619, "step": 20025 }, { "epoch": 0.49, "learning_rate": 1.0692396980998101e-05, "loss": 2.1735, "step": 20026 }, { "epoch": 0.49, "learning_rate": 1.0691600563617857e-05, "loss": 1.9993, "step": 20027 }, { "epoch": 0.49, "learning_rate": 1.0690804141829818e-05, "loss": 2.0485, "step": 20028 }, { "epoch": 0.49, "learning_rate": 1.0690007715639055e-05, "loss": 1.9817, "step": 20029 }, { "epoch": 0.49, "learning_rate": 1.0689211285050652e-05, "loss": 2.0265, "step": 20030 }, { "epoch": 0.49, "learning_rate": 1.0688414850069676e-05, "loss": 2.0445, "step": 20031 }, { "epoch": 0.49, "learning_rate": 1.0687618410701212e-05, "loss": 1.8924, "step": 20032 }, { "epoch": 0.49, "learning_rate": 1.0686821966950332e-05, "loss": 2.1599, "step": 20033 }, { "epoch": 0.49, "learning_rate": 1.0686025518822111e-05, "loss": 2.0794, "step": 20034 }, { "epoch": 0.49, "learning_rate": 1.0685229066321624e-05, "loss": 2.2231, "step": 20035 }, { "epoch": 0.49, "learning_rate": 1.0684432609453953e-05, "loss": 2.0599, "step": 20036 }, { "epoch": 0.49, "learning_rate": 1.0683636148224166e-05, "loss": 2.0252, "step": 20037 }, { "epoch": 0.49, "learning_rate": 1.0682839682637343e-05, "loss": 2.1153, "step": 20038 }, { "epoch": 0.49, "learning_rate": 1.068204321269856e-05, "loss": 2.0598, "step": 20039 }, { "epoch": 0.49, "learning_rate": 1.0681246738412896e-05, "loss": 2.0227, "step": 20040 }, { "epoch": 0.49, "learning_rate": 1.0680450259785422e-05, "loss": 1.9563, "step": 20041 }, { "epoch": 0.49, "learning_rate": 1.067965377682122e-05, "loss": 2.0045, "step": 20042 }, { "epoch": 0.49, "learning_rate": 1.067885728952536e-05, "loss": 1.9581, "step": 20043 }, { "epoch": 0.49, "learning_rate": 1.0678060797902923e-05, "loss": 2.0923, "step": 20044 }, { "epoch": 0.49, "learning_rate": 1.0677264301958982e-05, "loss": 2.0255, "step": 20045 }, { "epoch": 0.49, "learning_rate": 1.0676467801698616e-05, "loss": 1.9765, "step": 20046 }, { "epoch": 0.49, "learning_rate": 1.0675671297126897e-05, "loss": 2.0872, "step": 20047 }, { "epoch": 0.49, "learning_rate": 1.067487478824891e-05, "loss": 2.097, "step": 20048 }, { "epoch": 0.49, "learning_rate": 1.0674078275069724e-05, "loss": 1.8699, "step": 20049 }, { "epoch": 0.49, "learning_rate": 1.0673281757594416e-05, "loss": 1.9682, "step": 20050 }, { "epoch": 0.49, "learning_rate": 1.0672485235828064e-05, "loss": 1.9295, "step": 20051 }, { "epoch": 0.49, "learning_rate": 1.0671688709775748e-05, "loss": 2.0789, "step": 20052 }, { "epoch": 0.49, "learning_rate": 1.0670892179442539e-05, "loss": 1.9076, "step": 20053 }, { "epoch": 0.49, "learning_rate": 1.0670095644833514e-05, "loss": 2.0602, "step": 20054 }, { "epoch": 0.49, "learning_rate": 1.0669299105953749e-05, "loss": 1.9887, "step": 20055 }, { "epoch": 0.49, "learning_rate": 1.0668502562808325e-05, "loss": 1.9575, "step": 20056 }, { "epoch": 0.49, "learning_rate": 1.066770601540232e-05, "loss": 2.1472, "step": 20057 }, { "epoch": 0.49, "learning_rate": 1.0666909463740802e-05, "loss": 2.1568, "step": 20058 }, { "epoch": 0.49, "learning_rate": 1.0666112907828856e-05, "loss": 2.1317, "step": 20059 }, { "epoch": 0.49, "learning_rate": 1.0665316347671553e-05, "loss": 2.185, "step": 20060 }, { "epoch": 0.49, "learning_rate": 1.0664519783273972e-05, "loss": 1.9718, "step": 20061 }, { "epoch": 0.49, "learning_rate": 1.0663723214641194e-05, "loss": 1.9615, "step": 20062 }, { "epoch": 0.49, "learning_rate": 1.0662926641778289e-05, "loss": 2.2431, "step": 20063 }, { "epoch": 0.49, "learning_rate": 1.0662130064690333e-05, "loss": 1.8039, "step": 20064 }, { "epoch": 0.49, "learning_rate": 1.0661333483382409e-05, "loss": 2.0564, "step": 20065 }, { "epoch": 0.49, "learning_rate": 1.066053689785959e-05, "loss": 2.0834, "step": 20066 }, { "epoch": 0.49, "learning_rate": 1.0659740308126957e-05, "loss": 2.12, "step": 20067 }, { "epoch": 0.49, "learning_rate": 1.0658943714189585e-05, "loss": 1.8853, "step": 20068 }, { "epoch": 0.49, "learning_rate": 1.0658147116052546e-05, "loss": 1.9719, "step": 20069 }, { "epoch": 0.49, "learning_rate": 1.0657350513720921e-05, "loss": 1.9833, "step": 20070 }, { "epoch": 0.49, "learning_rate": 1.065655390719979e-05, "loss": 1.9532, "step": 20071 }, { "epoch": 0.49, "learning_rate": 1.0655757296494227e-05, "loss": 2.1019, "step": 20072 }, { "epoch": 0.49, "learning_rate": 1.0654960681609306e-05, "loss": 1.9835, "step": 20073 }, { "epoch": 0.49, "learning_rate": 1.0654164062550107e-05, "loss": 2.0883, "step": 20074 }, { "epoch": 0.49, "learning_rate": 1.0653367439321709e-05, "loss": 2.0711, "step": 20075 }, { "epoch": 0.49, "learning_rate": 1.0652570811929188e-05, "loss": 1.9987, "step": 20076 }, { "epoch": 0.49, "learning_rate": 1.0651774180377618e-05, "loss": 1.9014, "step": 20077 }, { "epoch": 0.49, "learning_rate": 1.0650977544672083e-05, "loss": 1.9228, "step": 20078 }, { "epoch": 0.49, "learning_rate": 1.0650180904817652e-05, "loss": 1.8848, "step": 20079 }, { "epoch": 0.49, "learning_rate": 1.0649384260819408e-05, "loss": 1.9827, "step": 20080 }, { "epoch": 0.49, "learning_rate": 1.0648587612682427e-05, "loss": 2.089, "step": 20081 }, { "epoch": 0.49, "learning_rate": 1.0647790960411783e-05, "loss": 1.9123, "step": 20082 }, { "epoch": 0.5, "learning_rate": 1.064699430401256e-05, "loss": 2.0924, "step": 20083 }, { "epoch": 0.5, "learning_rate": 1.0646197643489825e-05, "loss": 2.0345, "step": 20084 }, { "epoch": 0.5, "learning_rate": 1.064540097884867e-05, "loss": 1.9341, "step": 20085 }, { "epoch": 0.5, "learning_rate": 1.064460431009416e-05, "loss": 2.1204, "step": 20086 }, { "epoch": 0.5, "learning_rate": 1.0643807637231377e-05, "loss": 2.1754, "step": 20087 }, { "epoch": 0.5, "learning_rate": 1.0643010960265401e-05, "loss": 2.2149, "step": 20088 }, { "epoch": 0.5, "learning_rate": 1.0642214279201302e-05, "loss": 2.0888, "step": 20089 }, { "epoch": 0.5, "learning_rate": 1.0641417594044163e-05, "loss": 2.0053, "step": 20090 }, { "epoch": 0.5, "learning_rate": 1.0640620904799062e-05, "loss": 2.0437, "step": 20091 }, { "epoch": 0.5, "learning_rate": 1.0639824211471074e-05, "loss": 1.9387, "step": 20092 }, { "epoch": 0.5, "learning_rate": 1.0639027514065281e-05, "loss": 2.1806, "step": 20093 }, { "epoch": 0.5, "learning_rate": 1.0638230812586751e-05, "loss": 2.0819, "step": 20094 }, { "epoch": 0.5, "learning_rate": 1.0637434107040574e-05, "loss": 1.985, "step": 20095 }, { "epoch": 0.5, "learning_rate": 1.0636637397431822e-05, "loss": 2.07, "step": 20096 }, { "epoch": 0.5, "learning_rate": 1.063584068376557e-05, "loss": 2.0222, "step": 20097 }, { "epoch": 0.5, "learning_rate": 1.0635043966046899e-05, "loss": 1.9731, "step": 20098 }, { "epoch": 0.5, "learning_rate": 1.0634247244280884e-05, "loss": 2.1378, "step": 20099 }, { "epoch": 0.5, "learning_rate": 1.0633450518472603e-05, "loss": 2.1403, "step": 20100 }, { "epoch": 0.5, "learning_rate": 1.063265378862714e-05, "loss": 2.0231, "step": 20101 }, { "epoch": 0.5, "learning_rate": 1.0631857054749567e-05, "loss": 2.057, "step": 20102 }, { "epoch": 0.5, "learning_rate": 1.0631060316844964e-05, "loss": 1.9964, "step": 20103 }, { "epoch": 0.5, "learning_rate": 1.0630263574918403e-05, "loss": 1.9605, "step": 20104 }, { "epoch": 0.5, "learning_rate": 1.062946682897497e-05, "loss": 1.938, "step": 20105 }, { "epoch": 0.5, "learning_rate": 1.0628670079019743e-05, "loss": 1.9556, "step": 20106 }, { "epoch": 0.5, "learning_rate": 1.0627873325057794e-05, "loss": 1.982, "step": 20107 }, { "epoch": 0.5, "learning_rate": 1.06270765670942e-05, "loss": 1.8984, "step": 20108 }, { "epoch": 0.5, "learning_rate": 1.0626279805134045e-05, "loss": 1.9853, "step": 20109 }, { "epoch": 0.5, "learning_rate": 1.0625483039182408e-05, "loss": 2.2318, "step": 20110 }, { "epoch": 0.5, "learning_rate": 1.0624686269244362e-05, "loss": 2.2048, "step": 20111 }, { "epoch": 0.5, "learning_rate": 1.0623889495324986e-05, "loss": 2.1286, "step": 20112 }, { "epoch": 0.5, "learning_rate": 1.0623092717429357e-05, "loss": 2.0538, "step": 20113 }, { "epoch": 0.5, "learning_rate": 1.0622295935562559e-05, "loss": 1.9923, "step": 20114 }, { "epoch": 0.5, "learning_rate": 1.0621499149729666e-05, "loss": 1.9204, "step": 20115 }, { "epoch": 0.5, "learning_rate": 1.0620702359935755e-05, "loss": 1.9326, "step": 20116 }, { "epoch": 0.5, "learning_rate": 1.0619905566185904e-05, "loss": 1.9553, "step": 20117 }, { "epoch": 0.5, "learning_rate": 1.0619108768485193e-05, "loss": 1.8748, "step": 20118 }, { "epoch": 0.5, "learning_rate": 1.0618311966838702e-05, "loss": 1.9284, "step": 20119 }, { "epoch": 0.5, "learning_rate": 1.0617515161251508e-05, "loss": 2.2224, "step": 20120 }, { "epoch": 0.5, "learning_rate": 1.0616718351728687e-05, "loss": 2.0146, "step": 20121 }, { "epoch": 0.5, "learning_rate": 1.061592153827532e-05, "loss": 1.9846, "step": 20122 }, { "epoch": 0.5, "learning_rate": 1.0615124720896482e-05, "loss": 2.0456, "step": 20123 }, { "epoch": 0.5, "learning_rate": 1.0614327899597257e-05, "loss": 1.9427, "step": 20124 }, { "epoch": 0.5, "learning_rate": 1.0613531074382719e-05, "loss": 1.8962, "step": 20125 }, { "epoch": 0.5, "learning_rate": 1.0612734245257945e-05, "loss": 1.89, "step": 20126 }, { "epoch": 0.5, "learning_rate": 1.061193741222802e-05, "loss": 2.043, "step": 20127 }, { "epoch": 0.5, "learning_rate": 1.0611140575298014e-05, "loss": 1.9998, "step": 20128 }, { "epoch": 0.5, "learning_rate": 1.0610343734473012e-05, "loss": 1.9979, "step": 20129 }, { "epoch": 0.5, "learning_rate": 1.0609546889758092e-05, "loss": 1.9748, "step": 20130 }, { "epoch": 0.5, "learning_rate": 1.0608750041158331e-05, "loss": 1.9612, "step": 20131 }, { "epoch": 0.5, "learning_rate": 1.0607953188678805e-05, "loss": 2.0707, "step": 20132 }, { "epoch": 0.5, "learning_rate": 1.0607156332324595e-05, "loss": 2.1053, "step": 20133 }, { "epoch": 0.5, "learning_rate": 1.0606359472100783e-05, "loss": 2.0287, "step": 20134 }, { "epoch": 0.5, "learning_rate": 1.060556260801244e-05, "loss": 1.9168, "step": 20135 }, { "epoch": 0.5, "learning_rate": 1.0604765740064656e-05, "loss": 2.1933, "step": 20136 }, { "epoch": 0.5, "learning_rate": 1.0603968868262498e-05, "loss": 2.1267, "step": 20137 }, { "epoch": 0.5, "learning_rate": 1.060317199261105e-05, "loss": 2.2179, "step": 20138 }, { "epoch": 0.5, "learning_rate": 1.060237511311539e-05, "loss": 2.1647, "step": 20139 }, { "epoch": 0.5, "learning_rate": 1.0601578229780598e-05, "loss": 1.9039, "step": 20140 }, { "epoch": 0.5, "learning_rate": 1.0600781342611752e-05, "loss": 1.9941, "step": 20141 }, { "epoch": 0.5, "learning_rate": 1.059998445161393e-05, "loss": 2.0428, "step": 20142 }, { "epoch": 0.5, "learning_rate": 1.0599187556792212e-05, "loss": 2.1011, "step": 20143 }, { "epoch": 0.5, "learning_rate": 1.0598390658151675e-05, "loss": 2.1274, "step": 20144 }, { "epoch": 0.5, "learning_rate": 1.0597593755697402e-05, "loss": 1.9024, "step": 20145 }, { "epoch": 0.5, "learning_rate": 1.059679684943447e-05, "loss": 2.1214, "step": 20146 }, { "epoch": 0.5, "learning_rate": 1.0595999939367953e-05, "loss": 2.1104, "step": 20147 }, { "epoch": 0.5, "learning_rate": 1.0595203025502935e-05, "loss": 1.9664, "step": 20148 }, { "epoch": 0.5, "learning_rate": 1.0594406107844497e-05, "loss": 2.1748, "step": 20149 }, { "epoch": 0.5, "learning_rate": 1.0593609186397716e-05, "loss": 1.9231, "step": 20150 }, { "epoch": 0.5, "learning_rate": 1.0592812261167668e-05, "loss": 2.0765, "step": 20151 }, { "epoch": 0.5, "learning_rate": 1.0592015332159432e-05, "loss": 2.0549, "step": 20152 }, { "epoch": 0.5, "learning_rate": 1.0591218399378092e-05, "loss": 1.9832, "step": 20153 }, { "epoch": 0.5, "learning_rate": 1.0590421462828723e-05, "loss": 1.9738, "step": 20154 }, { "epoch": 0.5, "learning_rate": 1.0589624522516409e-05, "loss": 1.8876, "step": 20155 }, { "epoch": 0.5, "learning_rate": 1.0588827578446223e-05, "loss": 2.092, "step": 20156 }, { "epoch": 0.5, "learning_rate": 1.0588030630623246e-05, "loss": 2.0753, "step": 20157 }, { "epoch": 0.5, "learning_rate": 1.058723367905256e-05, "loss": 1.9748, "step": 20158 }, { "epoch": 0.5, "learning_rate": 1.0586436723739242e-05, "loss": 1.9631, "step": 20159 }, { "epoch": 0.5, "learning_rate": 1.0585639764688374e-05, "loss": 2.109, "step": 20160 }, { "epoch": 0.5, "learning_rate": 1.0584842801905026e-05, "loss": 1.9854, "step": 20161 }, { "epoch": 0.5, "learning_rate": 1.0584045835394288e-05, "loss": 2.0673, "step": 20162 }, { "epoch": 0.5, "learning_rate": 1.0583248865161238e-05, "loss": 1.8695, "step": 20163 }, { "epoch": 0.5, "learning_rate": 1.0582451891210952e-05, "loss": 2.0038, "step": 20164 }, { "epoch": 0.5, "learning_rate": 1.058165491354851e-05, "loss": 2.1629, "step": 20165 }, { "epoch": 0.5, "learning_rate": 1.0580857932178993e-05, "loss": 1.9805, "step": 20166 }, { "epoch": 0.5, "learning_rate": 1.0580060947107474e-05, "loss": 2.213, "step": 20167 }, { "epoch": 0.5, "learning_rate": 1.057926395833904e-05, "loss": 2.1259, "step": 20168 }, { "epoch": 0.5, "learning_rate": 1.057846696587877e-05, "loss": 2.1271, "step": 20169 }, { "epoch": 0.5, "learning_rate": 1.0577669969731738e-05, "loss": 2.0396, "step": 20170 }, { "epoch": 0.5, "learning_rate": 1.0576872969903029e-05, "loss": 1.972, "step": 20171 }, { "epoch": 0.5, "learning_rate": 1.0576075966397721e-05, "loss": 2.0669, "step": 20172 }, { "epoch": 0.5, "learning_rate": 1.0575278959220894e-05, "loss": 1.8467, "step": 20173 }, { "epoch": 0.5, "learning_rate": 1.0574481948377625e-05, "loss": 2.1016, "step": 20174 }, { "epoch": 0.5, "learning_rate": 1.0573684933872999e-05, "loss": 2.01, "step": 20175 }, { "epoch": 0.5, "learning_rate": 1.0572887915712087e-05, "loss": 1.9821, "step": 20176 }, { "epoch": 0.5, "learning_rate": 1.0572090893899976e-05, "loss": 2.0588, "step": 20177 }, { "epoch": 0.5, "learning_rate": 1.0571293868441743e-05, "loss": 1.9163, "step": 20178 }, { "epoch": 0.5, "learning_rate": 1.0570496839342468e-05, "loss": 2.1283, "step": 20179 }, { "epoch": 0.5, "learning_rate": 1.056969980660723e-05, "loss": 2.2025, "step": 20180 }, { "epoch": 0.5, "learning_rate": 1.056890277024111e-05, "loss": 2.0407, "step": 20181 }, { "epoch": 0.5, "learning_rate": 1.0568105730249188e-05, "loss": 2.1843, "step": 20182 }, { "epoch": 0.5, "learning_rate": 1.0567308686636542e-05, "loss": 1.9392, "step": 20183 }, { "epoch": 0.5, "learning_rate": 1.0566511639408255e-05, "loss": 2.0402, "step": 20184 }, { "epoch": 0.5, "learning_rate": 1.0565714588569403e-05, "loss": 2.0681, "step": 20185 }, { "epoch": 0.5, "learning_rate": 1.0564917534125066e-05, "loss": 1.9793, "step": 20186 }, { "epoch": 0.5, "learning_rate": 1.0564120476080327e-05, "loss": 2.1775, "step": 20187 }, { "epoch": 0.5, "learning_rate": 1.0563323414440264e-05, "loss": 1.9853, "step": 20188 }, { "epoch": 0.5, "learning_rate": 1.0562526349209956e-05, "loss": 2.0677, "step": 20189 }, { "epoch": 0.5, "learning_rate": 1.0561729280394489e-05, "loss": 2.0445, "step": 20190 }, { "epoch": 0.5, "learning_rate": 1.0560932207998934e-05, "loss": 2.1044, "step": 20191 }, { "epoch": 0.5, "learning_rate": 1.0560135132028375e-05, "loss": 1.8343, "step": 20192 }, { "epoch": 0.5, "learning_rate": 1.0559338052487895e-05, "loss": 1.9681, "step": 20193 }, { "epoch": 0.5, "learning_rate": 1.0558540969382569e-05, "loss": 1.9438, "step": 20194 }, { "epoch": 0.5, "learning_rate": 1.055774388271748e-05, "loss": 2.0728, "step": 20195 }, { "epoch": 0.5, "learning_rate": 1.0556946792497708e-05, "loss": 2.1128, "step": 20196 }, { "epoch": 0.5, "learning_rate": 1.0556149698728332e-05, "loss": 2.2239, "step": 20197 }, { "epoch": 0.5, "learning_rate": 1.0555352601414432e-05, "loss": 2.2891, "step": 20198 }, { "epoch": 0.5, "learning_rate": 1.0554555500561092e-05, "loss": 2.0199, "step": 20199 }, { "epoch": 0.5, "learning_rate": 1.0553758396173385e-05, "loss": 2.066, "step": 20200 }, { "epoch": 0.5, "learning_rate": 1.0552961288256398e-05, "loss": 2.0254, "step": 20201 }, { "epoch": 0.5, "learning_rate": 1.0552164176815209e-05, "loss": 1.9883, "step": 20202 }, { "epoch": 0.5, "learning_rate": 1.0551367061854896e-05, "loss": 1.9656, "step": 20203 }, { "epoch": 0.5, "learning_rate": 1.0550569943380542e-05, "loss": 1.9734, "step": 20204 }, { "epoch": 0.5, "learning_rate": 1.0549772821397226e-05, "loss": 2.0453, "step": 20205 }, { "epoch": 0.5, "learning_rate": 1.0548975695910027e-05, "loss": 2.2483, "step": 20206 }, { "epoch": 0.5, "learning_rate": 1.0548178566924027e-05, "loss": 2.0803, "step": 20207 }, { "epoch": 0.5, "learning_rate": 1.054738143444431e-05, "loss": 2.0685, "step": 20208 }, { "epoch": 0.5, "learning_rate": 1.0546584298475953e-05, "loss": 2.0365, "step": 20209 }, { "epoch": 0.5, "learning_rate": 1.054578715902403e-05, "loss": 2.1735, "step": 20210 }, { "epoch": 0.5, "learning_rate": 1.0544990016093632e-05, "loss": 1.9386, "step": 20211 }, { "epoch": 0.5, "learning_rate": 1.0544192869689835e-05, "loss": 2.0204, "step": 20212 }, { "epoch": 0.5, "learning_rate": 1.0543395719817717e-05, "loss": 2.078, "step": 20213 }, { "epoch": 0.5, "learning_rate": 1.0542598566482365e-05, "loss": 1.8993, "step": 20214 }, { "epoch": 0.5, "learning_rate": 1.0541801409688853e-05, "loss": 1.9228, "step": 20215 }, { "epoch": 0.5, "learning_rate": 1.0541004249442264e-05, "loss": 2.097, "step": 20216 }, { "epoch": 0.5, "learning_rate": 1.054020708574768e-05, "loss": 2.101, "step": 20217 }, { "epoch": 0.5, "learning_rate": 1.053940991861018e-05, "loss": 1.9651, "step": 20218 }, { "epoch": 0.5, "learning_rate": 1.0538612748034844e-05, "loss": 2.0027, "step": 20219 }, { "epoch": 0.5, "learning_rate": 1.0537815574026751e-05, "loss": 2.0378, "step": 20220 }, { "epoch": 0.5, "learning_rate": 1.0537018396590987e-05, "loss": 2.191, "step": 20221 }, { "epoch": 0.5, "learning_rate": 1.0536221215732629e-05, "loss": 2.0833, "step": 20222 }, { "epoch": 0.5, "learning_rate": 1.0535424031456759e-05, "loss": 1.9914, "step": 20223 }, { "epoch": 0.5, "learning_rate": 1.0534626843768459e-05, "loss": 2.0494, "step": 20224 }, { "epoch": 0.5, "learning_rate": 1.0533829652672803e-05, "loss": 2.0123, "step": 20225 }, { "epoch": 0.5, "learning_rate": 1.053303245817488e-05, "loss": 2.0209, "step": 20226 }, { "epoch": 0.5, "learning_rate": 1.0532235260279768e-05, "loss": 2.1293, "step": 20227 }, { "epoch": 0.5, "learning_rate": 1.0531438058992548e-05, "loss": 2.1065, "step": 20228 }, { "epoch": 0.5, "learning_rate": 1.0530640854318294e-05, "loss": 2.0159, "step": 20229 }, { "epoch": 0.5, "learning_rate": 1.05298436462621e-05, "loss": 2.0907, "step": 20230 }, { "epoch": 0.5, "learning_rate": 1.0529046434829033e-05, "loss": 2.0666, "step": 20231 }, { "epoch": 0.5, "learning_rate": 1.0528249220024185e-05, "loss": 2.1441, "step": 20232 }, { "epoch": 0.5, "learning_rate": 1.0527452001852635e-05, "loss": 2.1296, "step": 20233 }, { "epoch": 0.5, "learning_rate": 1.0526654780319456e-05, "loss": 2.0514, "step": 20234 }, { "epoch": 0.5, "learning_rate": 1.0525857555429737e-05, "loss": 1.8464, "step": 20235 }, { "epoch": 0.5, "learning_rate": 1.0525060327188557e-05, "loss": 2.0666, "step": 20236 }, { "epoch": 0.5, "learning_rate": 1.0524263095600998e-05, "loss": 1.952, "step": 20237 }, { "epoch": 0.5, "learning_rate": 1.0523465860672137e-05, "loss": 2.0768, "step": 20238 }, { "epoch": 0.5, "learning_rate": 1.0522668622407054e-05, "loss": 2.0332, "step": 20239 }, { "epoch": 0.5, "learning_rate": 1.052187138081084e-05, "loss": 2.0487, "step": 20240 }, { "epoch": 0.5, "learning_rate": 1.0521074135888564e-05, "loss": 2.0638, "step": 20241 }, { "epoch": 0.5, "learning_rate": 1.0520276887645318e-05, "loss": 2.2483, "step": 20242 }, { "epoch": 0.5, "learning_rate": 1.0519479636086176e-05, "loss": 1.911, "step": 20243 }, { "epoch": 0.5, "learning_rate": 1.0518682381216218e-05, "loss": 2.1791, "step": 20244 }, { "epoch": 0.5, "learning_rate": 1.0517885123040531e-05, "loss": 1.8812, "step": 20245 }, { "epoch": 0.5, "learning_rate": 1.0517087861564195e-05, "loss": 2.0276, "step": 20246 }, { "epoch": 0.5, "learning_rate": 1.0516290596792286e-05, "loss": 2.0583, "step": 20247 }, { "epoch": 0.5, "learning_rate": 1.051549332872989e-05, "loss": 2.052, "step": 20248 }, { "epoch": 0.5, "learning_rate": 1.0514696057382088e-05, "loss": 1.8869, "step": 20249 }, { "epoch": 0.5, "learning_rate": 1.051389878275396e-05, "loss": 1.782, "step": 20250 }, { "epoch": 0.5, "learning_rate": 1.0513101504850587e-05, "loss": 1.9963, "step": 20251 }, { "epoch": 0.5, "learning_rate": 1.0512304223677053e-05, "loss": 2.0952, "step": 20252 }, { "epoch": 0.5, "learning_rate": 1.0511506939238434e-05, "loss": 2.3086, "step": 20253 }, { "epoch": 0.5, "learning_rate": 1.0510709651539814e-05, "loss": 1.9875, "step": 20254 }, { "epoch": 0.5, "learning_rate": 1.0509912360586278e-05, "loss": 2.0107, "step": 20255 }, { "epoch": 0.5, "learning_rate": 1.0509115066382904e-05, "loss": 1.9388, "step": 20256 }, { "epoch": 0.5, "learning_rate": 1.050831776893477e-05, "loss": 2.0371, "step": 20257 }, { "epoch": 0.5, "learning_rate": 1.0507520468246966e-05, "loss": 1.9111, "step": 20258 }, { "epoch": 0.5, "learning_rate": 1.0506723164324566e-05, "loss": 1.8446, "step": 20259 }, { "epoch": 0.5, "learning_rate": 1.0505925857172654e-05, "loss": 1.9059, "step": 20260 }, { "epoch": 0.5, "learning_rate": 1.0505128546796316e-05, "loss": 1.9086, "step": 20261 }, { "epoch": 0.5, "learning_rate": 1.0504331233200624e-05, "loss": 1.9768, "step": 20262 }, { "epoch": 0.5, "learning_rate": 1.0503533916390665e-05, "loss": 2.0632, "step": 20263 }, { "epoch": 0.5, "learning_rate": 1.0502736596371523e-05, "loss": 1.9873, "step": 20264 }, { "epoch": 0.5, "learning_rate": 1.0501939273148276e-05, "loss": 2.209, "step": 20265 }, { "epoch": 0.5, "learning_rate": 1.0501141946726003e-05, "loss": 2.1332, "step": 20266 }, { "epoch": 0.5, "learning_rate": 1.0500344617109795e-05, "loss": 2.2232, "step": 20267 }, { "epoch": 0.5, "learning_rate": 1.0499547284304724e-05, "loss": 1.9511, "step": 20268 }, { "epoch": 0.5, "learning_rate": 1.0498749948315874e-05, "loss": 2.2256, "step": 20269 }, { "epoch": 0.5, "learning_rate": 1.0497952609148332e-05, "loss": 2.2961, "step": 20270 }, { "epoch": 0.5, "learning_rate": 1.0497155266807176e-05, "loss": 1.9963, "step": 20271 }, { "epoch": 0.5, "learning_rate": 1.0496357921297485e-05, "loss": 2.063, "step": 20272 }, { "epoch": 0.5, "learning_rate": 1.049556057262434e-05, "loss": 1.9453, "step": 20273 }, { "epoch": 0.5, "learning_rate": 1.0494763220792831e-05, "loss": 2.1629, "step": 20274 }, { "epoch": 0.5, "learning_rate": 1.0493965865808031e-05, "loss": 1.9881, "step": 20275 }, { "epoch": 0.5, "learning_rate": 1.0493168507675028e-05, "loss": 2.2104, "step": 20276 }, { "epoch": 0.5, "learning_rate": 1.0492371146398903e-05, "loss": 2.0581, "step": 20277 }, { "epoch": 0.5, "learning_rate": 1.0491573781984734e-05, "loss": 2.0689, "step": 20278 }, { "epoch": 0.5, "learning_rate": 1.0490776414437604e-05, "loss": 2.0349, "step": 20279 }, { "epoch": 0.5, "learning_rate": 1.04899790437626e-05, "loss": 2.0536, "step": 20280 }, { "epoch": 0.5, "learning_rate": 1.0489181669964798e-05, "loss": 2.0613, "step": 20281 }, { "epoch": 0.5, "learning_rate": 1.048838429304928e-05, "loss": 2.1145, "step": 20282 }, { "epoch": 0.5, "learning_rate": 1.048758691302113e-05, "loss": 2.0746, "step": 20283 }, { "epoch": 0.5, "learning_rate": 1.048678952988543e-05, "loss": 2.0825, "step": 20284 }, { "epoch": 0.5, "learning_rate": 1.0485992143647263e-05, "loss": 1.9663, "step": 20285 }, { "epoch": 0.5, "eval_loss": 1.739554762840271, "eval_runtime": 94.7213, "eval_samples_per_second": 621.265, "eval_steps_per_second": 4.856, "step": 20285 }, { "epoch": 0.5, "learning_rate": 1.048519475431171e-05, "loss": 2.1425, "step": 20286 }, { "epoch": 0.5, "learning_rate": 1.0484397361883848e-05, "loss": 1.9259, "step": 20287 }, { "epoch": 0.5, "learning_rate": 1.048359996636877e-05, "loss": 2.1206, "step": 20288 }, { "epoch": 0.5, "learning_rate": 1.048280256777155e-05, "loss": 2.1656, "step": 20289 }, { "epoch": 0.5, "learning_rate": 1.0482005166097272e-05, "loss": 2.0825, "step": 20290 }, { "epoch": 0.5, "learning_rate": 1.0481207761351017e-05, "loss": 1.882, "step": 20291 }, { "epoch": 0.5, "learning_rate": 1.0480410353537867e-05, "loss": 2.02, "step": 20292 }, { "epoch": 0.5, "learning_rate": 1.0479612942662904e-05, "loss": 2.0608, "step": 20293 }, { "epoch": 0.5, "learning_rate": 1.0478815528731215e-05, "loss": 1.9441, "step": 20294 }, { "epoch": 0.5, "learning_rate": 1.0478018111747879e-05, "loss": 2.124, "step": 20295 }, { "epoch": 0.5, "learning_rate": 1.0477220691717977e-05, "loss": 2.0531, "step": 20296 }, { "epoch": 0.5, "learning_rate": 1.0476423268646589e-05, "loss": 1.9073, "step": 20297 }, { "epoch": 0.5, "learning_rate": 1.0475625842538804e-05, "loss": 1.9876, "step": 20298 }, { "epoch": 0.5, "learning_rate": 1.04748284133997e-05, "loss": 2.0985, "step": 20299 }, { "epoch": 0.5, "learning_rate": 1.0474030981234358e-05, "loss": 1.9103, "step": 20300 }, { "epoch": 0.5, "learning_rate": 1.0473233546047864e-05, "loss": 2.1448, "step": 20301 }, { "epoch": 0.5, "learning_rate": 1.0472436107845294e-05, "loss": 2.0099, "step": 20302 }, { "epoch": 0.5, "learning_rate": 1.0471638666631738e-05, "loss": 2.1669, "step": 20303 }, { "epoch": 0.5, "learning_rate": 1.0470841222412275e-05, "loss": 2.1214, "step": 20304 }, { "epoch": 0.5, "learning_rate": 1.0470043775191989e-05, "loss": 2.0683, "step": 20305 }, { "epoch": 0.5, "learning_rate": 1.0469246324975961e-05, "loss": 1.9231, "step": 20306 }, { "epoch": 0.5, "learning_rate": 1.046844887176927e-05, "loss": 2.0792, "step": 20307 }, { "epoch": 0.5, "learning_rate": 1.0467651415577004e-05, "loss": 1.9984, "step": 20308 }, { "epoch": 0.5, "learning_rate": 1.0466853956404243e-05, "loss": 2.0532, "step": 20309 }, { "epoch": 0.5, "learning_rate": 1.0466056494256067e-05, "loss": 2.0414, "step": 20310 }, { "epoch": 0.5, "learning_rate": 1.0465259029137565e-05, "loss": 1.9092, "step": 20311 }, { "epoch": 0.5, "learning_rate": 1.0464461561053812e-05, "loss": 2.0219, "step": 20312 }, { "epoch": 0.5, "learning_rate": 1.0463664090009894e-05, "loss": 1.9302, "step": 20313 }, { "epoch": 0.5, "learning_rate": 1.0462866616010899e-05, "loss": 2.0208, "step": 20314 }, { "epoch": 0.5, "learning_rate": 1.0462069139061901e-05, "loss": 2.0761, "step": 20315 }, { "epoch": 0.5, "learning_rate": 1.0461271659167987e-05, "loss": 2.1954, "step": 20316 }, { "epoch": 0.5, "learning_rate": 1.0460474176334235e-05, "loss": 2.113, "step": 20317 }, { "epoch": 0.5, "learning_rate": 1.0459676690565732e-05, "loss": 2.0748, "step": 20318 }, { "epoch": 0.5, "learning_rate": 1.045887920186756e-05, "loss": 1.8986, "step": 20319 }, { "epoch": 0.5, "learning_rate": 1.04580817102448e-05, "loss": 2.0314, "step": 20320 }, { "epoch": 0.5, "learning_rate": 1.0457284215702539e-05, "loss": 2.0162, "step": 20321 }, { "epoch": 0.5, "learning_rate": 1.0456486718245856e-05, "loss": 1.8895, "step": 20322 }, { "epoch": 0.5, "learning_rate": 1.0455689217879834e-05, "loss": 2.0152, "step": 20323 }, { "epoch": 0.5, "learning_rate": 1.0454891714609557e-05, "loss": 2.207, "step": 20324 }, { "epoch": 0.5, "learning_rate": 1.0454094208440106e-05, "loss": 2.1781, "step": 20325 }, { "epoch": 0.5, "learning_rate": 1.0453296699376562e-05, "loss": 2.1335, "step": 20326 }, { "epoch": 0.5, "learning_rate": 1.0452499187424015e-05, "loss": 1.9585, "step": 20327 }, { "epoch": 0.5, "learning_rate": 1.0451701672587538e-05, "loss": 1.9678, "step": 20328 }, { "epoch": 0.5, "learning_rate": 1.0450904154872223e-05, "loss": 1.9914, "step": 20329 }, { "epoch": 0.5, "learning_rate": 1.0450106634283148e-05, "loss": 2.161, "step": 20330 }, { "epoch": 0.5, "learning_rate": 1.0449309110825396e-05, "loss": 2.0178, "step": 20331 }, { "epoch": 0.5, "learning_rate": 1.0448511584504054e-05, "loss": 1.9041, "step": 20332 }, { "epoch": 0.5, "learning_rate": 1.0447714055324199e-05, "loss": 2.1059, "step": 20333 }, { "epoch": 0.5, "learning_rate": 1.0446916523290916e-05, "loss": 2.143, "step": 20334 }, { "epoch": 0.5, "learning_rate": 1.0446118988409288e-05, "loss": 2.0849, "step": 20335 }, { "epoch": 0.5, "learning_rate": 1.0445321450684399e-05, "loss": 2.0271, "step": 20336 }, { "epoch": 0.5, "learning_rate": 1.044452391012133e-05, "loss": 1.7823, "step": 20337 }, { "epoch": 0.5, "learning_rate": 1.0443726366725167e-05, "loss": 1.9687, "step": 20338 }, { "epoch": 0.5, "learning_rate": 1.044292882050099e-05, "loss": 1.9244, "step": 20339 }, { "epoch": 0.5, "learning_rate": 1.0442131271453884e-05, "loss": 2.0747, "step": 20340 }, { "epoch": 0.5, "learning_rate": 1.044133371958893e-05, "loss": 2.0403, "step": 20341 }, { "epoch": 0.5, "learning_rate": 1.0440536164911214e-05, "loss": 1.9759, "step": 20342 }, { "epoch": 0.5, "learning_rate": 1.0439738607425817e-05, "loss": 1.7271, "step": 20343 }, { "epoch": 0.5, "learning_rate": 1.0438941047137825e-05, "loss": 2.0821, "step": 20344 }, { "epoch": 0.5, "learning_rate": 1.0438143484052314e-05, "loss": 1.9572, "step": 20345 }, { "epoch": 0.5, "learning_rate": 1.0437345918174373e-05, "loss": 1.9094, "step": 20346 }, { "epoch": 0.5, "learning_rate": 1.0436548349509086e-05, "loss": 1.8393, "step": 20347 }, { "epoch": 0.5, "learning_rate": 1.0435750778061534e-05, "loss": 2.0949, "step": 20348 }, { "epoch": 0.5, "learning_rate": 1.0434953203836799e-05, "loss": 1.8195, "step": 20349 }, { "epoch": 0.5, "learning_rate": 1.0434155626839966e-05, "loss": 2.0435, "step": 20350 }, { "epoch": 0.5, "learning_rate": 1.0433358047076117e-05, "loss": 2.1693, "step": 20351 }, { "epoch": 0.5, "learning_rate": 1.0432560464550337e-05, "loss": 1.9218, "step": 20352 }, { "epoch": 0.5, "learning_rate": 1.0431762879267708e-05, "loss": 1.9666, "step": 20353 }, { "epoch": 0.5, "learning_rate": 1.0430965291233312e-05, "loss": 1.9806, "step": 20354 }, { "epoch": 0.5, "learning_rate": 1.0430167700452234e-05, "loss": 2.0001, "step": 20355 }, { "epoch": 0.5, "learning_rate": 1.042937010692956e-05, "loss": 2.0224, "step": 20356 }, { "epoch": 0.5, "learning_rate": 1.0428572510670367e-05, "loss": 2.0472, "step": 20357 }, { "epoch": 0.5, "learning_rate": 1.0427774911679744e-05, "loss": 2.1011, "step": 20358 }, { "epoch": 0.5, "learning_rate": 1.0426977309962772e-05, "loss": 2.1985, "step": 20359 }, { "epoch": 0.5, "learning_rate": 1.0426179705524533e-05, "loss": 1.8685, "step": 20360 }, { "epoch": 0.5, "learning_rate": 1.0425382098370111e-05, "loss": 2.1514, "step": 20361 }, { "epoch": 0.5, "learning_rate": 1.0424584488504595e-05, "loss": 1.8626, "step": 20362 }, { "epoch": 0.5, "learning_rate": 1.0423786875933058e-05, "loss": 2.1266, "step": 20363 }, { "epoch": 0.5, "learning_rate": 1.0422989260660591e-05, "loss": 1.9988, "step": 20364 }, { "epoch": 0.5, "learning_rate": 1.0422191642692277e-05, "loss": 1.9805, "step": 20365 }, { "epoch": 0.5, "learning_rate": 1.0421394022033197e-05, "loss": 2.0677, "step": 20366 }, { "epoch": 0.5, "learning_rate": 1.0420596398688436e-05, "loss": 2.0064, "step": 20367 }, { "epoch": 0.5, "learning_rate": 1.0419798772663076e-05, "loss": 2.1113, "step": 20368 }, { "epoch": 0.5, "learning_rate": 1.0419001143962203e-05, "loss": 2.1236, "step": 20369 }, { "epoch": 0.5, "learning_rate": 1.0418203512590898e-05, "loss": 1.9528, "step": 20370 }, { "epoch": 0.5, "learning_rate": 1.0417405878554248e-05, "loss": 2.0603, "step": 20371 }, { "epoch": 0.5, "learning_rate": 1.041660824185733e-05, "loss": 2.0828, "step": 20372 }, { "epoch": 0.5, "learning_rate": 1.0415810602505237e-05, "loss": 2.0709, "step": 20373 }, { "epoch": 0.5, "learning_rate": 1.0415012960503045e-05, "loss": 1.7134, "step": 20374 }, { "epoch": 0.5, "learning_rate": 1.0414215315855839e-05, "loss": 2.0096, "step": 20375 }, { "epoch": 0.5, "learning_rate": 1.0413417668568707e-05, "loss": 2.254, "step": 20376 }, { "epoch": 0.5, "learning_rate": 1.0412620018646726e-05, "loss": 2.0666, "step": 20377 }, { "epoch": 0.5, "learning_rate": 1.0411822366094986e-05, "loss": 2.0833, "step": 20378 }, { "epoch": 0.5, "learning_rate": 1.0411024710918564e-05, "loss": 2.0592, "step": 20379 }, { "epoch": 0.5, "learning_rate": 1.0410227053122552e-05, "loss": 2.0373, "step": 20380 }, { "epoch": 0.5, "learning_rate": 1.0409429392712025e-05, "loss": 2.0088, "step": 20381 }, { "epoch": 0.5, "learning_rate": 1.0408631729692076e-05, "loss": 1.881, "step": 20382 }, { "epoch": 0.5, "learning_rate": 1.0407834064067782e-05, "loss": 2.1407, "step": 20383 }, { "epoch": 0.5, "learning_rate": 1.0407036395844226e-05, "loss": 1.9787, "step": 20384 }, { "epoch": 0.5, "learning_rate": 1.0406238725026497e-05, "loss": 2.0379, "step": 20385 }, { "epoch": 0.5, "learning_rate": 1.0405441051619676e-05, "loss": 1.9761, "step": 20386 }, { "epoch": 0.5, "learning_rate": 1.0404643375628847e-05, "loss": 2.0693, "step": 20387 }, { "epoch": 0.5, "learning_rate": 1.0403845697059094e-05, "loss": 2.1437, "step": 20388 }, { "epoch": 0.5, "learning_rate": 1.0403048015915498e-05, "loss": 2.149, "step": 20389 }, { "epoch": 0.5, "learning_rate": 1.0402250332203147e-05, "loss": 1.9678, "step": 20390 }, { "epoch": 0.5, "learning_rate": 1.0401452645927126e-05, "loss": 2.1364, "step": 20391 }, { "epoch": 0.5, "learning_rate": 1.0400654957092515e-05, "loss": 1.9878, "step": 20392 }, { "epoch": 0.5, "learning_rate": 1.03998572657044e-05, "loss": 2.0513, "step": 20393 }, { "epoch": 0.5, "learning_rate": 1.0399059571767863e-05, "loss": 2.2651, "step": 20394 }, { "epoch": 0.5, "learning_rate": 1.039826187528799e-05, "loss": 2.0391, "step": 20395 }, { "epoch": 0.5, "learning_rate": 1.0397464176269865e-05, "loss": 2.0117, "step": 20396 }, { "epoch": 0.5, "learning_rate": 1.039666647471857e-05, "loss": 2.0166, "step": 20397 }, { "epoch": 0.5, "learning_rate": 1.039586877063919e-05, "loss": 2.1583, "step": 20398 }, { "epoch": 0.5, "learning_rate": 1.0395071064036808e-05, "loss": 2.0833, "step": 20399 }, { "epoch": 0.5, "learning_rate": 1.0394273354916511e-05, "loss": 2.06, "step": 20400 }, { "epoch": 0.5, "learning_rate": 1.0393475643283385e-05, "loss": 2.0365, "step": 20401 }, { "epoch": 0.5, "learning_rate": 1.0392677929142507e-05, "loss": 2.0335, "step": 20402 }, { "epoch": 0.5, "learning_rate": 1.0391880212498965e-05, "loss": 1.998, "step": 20403 }, { "epoch": 0.5, "learning_rate": 1.039108249335784e-05, "loss": 1.9846, "step": 20404 }, { "epoch": 0.5, "learning_rate": 1.0390284771724224e-05, "loss": 1.8881, "step": 20405 }, { "epoch": 0.5, "learning_rate": 1.0389487047603193e-05, "loss": 2.0795, "step": 20406 }, { "epoch": 0.5, "learning_rate": 1.0388689320999835e-05, "loss": 2.1181, "step": 20407 }, { "epoch": 0.5, "learning_rate": 1.038789159191923e-05, "loss": 2.0215, "step": 20408 }, { "epoch": 0.5, "learning_rate": 1.0387093860366469e-05, "loss": 1.9484, "step": 20409 }, { "epoch": 0.5, "learning_rate": 1.0386296126346631e-05, "loss": 2.007, "step": 20410 }, { "epoch": 0.5, "learning_rate": 1.0385498389864805e-05, "loss": 2.0656, "step": 20411 }, { "epoch": 0.5, "learning_rate": 1.0384700650926069e-05, "loss": 1.971, "step": 20412 }, { "epoch": 0.5, "learning_rate": 1.038390290953551e-05, "loss": 1.909, "step": 20413 }, { "epoch": 0.5, "learning_rate": 1.0383105165698216e-05, "loss": 1.996, "step": 20414 }, { "epoch": 0.5, "learning_rate": 1.0382307419419267e-05, "loss": 2.0235, "step": 20415 }, { "epoch": 0.5, "learning_rate": 1.0381509670703744e-05, "loss": 2.0756, "step": 20416 }, { "epoch": 0.5, "learning_rate": 1.038071191955674e-05, "loss": 2.0703, "step": 20417 }, { "epoch": 0.5, "learning_rate": 1.0379914165983332e-05, "loss": 2.0084, "step": 20418 }, { "epoch": 0.5, "learning_rate": 1.037911640998861e-05, "loss": 2.0638, "step": 20419 }, { "epoch": 0.5, "learning_rate": 1.0378318651577655e-05, "loss": 1.9976, "step": 20420 }, { "epoch": 0.5, "learning_rate": 1.0377520890755553e-05, "loss": 1.8308, "step": 20421 }, { "epoch": 0.5, "learning_rate": 1.0376723127527385e-05, "loss": 2.0443, "step": 20422 }, { "epoch": 0.5, "learning_rate": 1.0375925361898235e-05, "loss": 2.0402, "step": 20423 }, { "epoch": 0.5, "learning_rate": 1.0375127593873196e-05, "loss": 2.1121, "step": 20424 }, { "epoch": 0.5, "learning_rate": 1.0374329823457343e-05, "loss": 1.9862, "step": 20425 }, { "epoch": 0.5, "learning_rate": 1.0373532050655766e-05, "loss": 1.9948, "step": 20426 }, { "epoch": 0.5, "learning_rate": 1.0372734275473545e-05, "loss": 2.0529, "step": 20427 }, { "epoch": 0.5, "learning_rate": 1.0371936497915768e-05, "loss": 1.8264, "step": 20428 }, { "epoch": 0.5, "learning_rate": 1.037113871798752e-05, "loss": 2.0301, "step": 20429 }, { "epoch": 0.5, "learning_rate": 1.0370340935693884e-05, "loss": 2.0433, "step": 20430 }, { "epoch": 0.5, "learning_rate": 1.0369543151039944e-05, "loss": 2.019, "step": 20431 }, { "epoch": 0.5, "learning_rate": 1.0368745364030784e-05, "loss": 1.9971, "step": 20432 }, { "epoch": 0.5, "learning_rate": 1.0367947574671492e-05, "loss": 2.0033, "step": 20433 }, { "epoch": 0.5, "learning_rate": 1.0367149782967145e-05, "loss": 2.0497, "step": 20434 }, { "epoch": 0.5, "learning_rate": 1.0366351988922838e-05, "loss": 1.9754, "step": 20435 }, { "epoch": 0.5, "learning_rate": 1.036555419254365e-05, "loss": 2.014, "step": 20436 }, { "epoch": 0.5, "learning_rate": 1.0364756393834663e-05, "loss": 2.1724, "step": 20437 }, { "epoch": 0.5, "learning_rate": 1.0363958592800966e-05, "loss": 2.1062, "step": 20438 }, { "epoch": 0.5, "learning_rate": 1.0363160789447644e-05, "loss": 1.8828, "step": 20439 }, { "epoch": 0.5, "learning_rate": 1.036236298377978e-05, "loss": 2.0558, "step": 20440 }, { "epoch": 0.5, "learning_rate": 1.0361565175802456e-05, "loss": 1.9512, "step": 20441 }, { "epoch": 0.5, "learning_rate": 1.0360767365520758e-05, "loss": 1.975, "step": 20442 }, { "epoch": 0.5, "learning_rate": 1.0359969552939775e-05, "loss": 2.0262, "step": 20443 }, { "epoch": 0.5, "learning_rate": 1.035917173806459e-05, "loss": 2.0831, "step": 20444 }, { "epoch": 0.5, "learning_rate": 1.0358373920900283e-05, "loss": 2.114, "step": 20445 }, { "epoch": 0.5, "learning_rate": 1.0357576101451945e-05, "loss": 2.0013, "step": 20446 }, { "epoch": 0.5, "learning_rate": 1.0356778279724656e-05, "loss": 2.0817, "step": 20447 }, { "epoch": 0.5, "learning_rate": 1.0355980455723505e-05, "loss": 2.1805, "step": 20448 }, { "epoch": 0.5, "learning_rate": 1.0355182629453573e-05, "loss": 1.9988, "step": 20449 }, { "epoch": 0.5, "learning_rate": 1.0354384800919948e-05, "loss": 1.9199, "step": 20450 }, { "epoch": 0.5, "learning_rate": 1.0353586970127709e-05, "loss": 2.0948, "step": 20451 }, { "epoch": 0.5, "learning_rate": 1.0352789137081947e-05, "loss": 2.0379, "step": 20452 }, { "epoch": 0.5, "learning_rate": 1.0351991301787745e-05, "loss": 1.9052, "step": 20453 }, { "epoch": 0.5, "learning_rate": 1.0351193464250189e-05, "loss": 1.9809, "step": 20454 }, { "epoch": 0.5, "learning_rate": 1.0350395624474362e-05, "loss": 2.226, "step": 20455 }, { "epoch": 0.5, "learning_rate": 1.034959778246535e-05, "loss": 1.9305, "step": 20456 }, { "epoch": 0.5, "learning_rate": 1.0348799938228235e-05, "loss": 2.0834, "step": 20457 }, { "epoch": 0.5, "learning_rate": 1.0348002091768107e-05, "loss": 1.9704, "step": 20458 }, { "epoch": 0.5, "learning_rate": 1.0347204243090047e-05, "loss": 1.8936, "step": 20459 }, { "epoch": 0.5, "learning_rate": 1.0346406392199143e-05, "loss": 2.0494, "step": 20460 }, { "epoch": 0.5, "learning_rate": 1.0345608539100475e-05, "loss": 2.1776, "step": 20461 }, { "epoch": 0.5, "learning_rate": 1.0344810683799133e-05, "loss": 1.9452, "step": 20462 }, { "epoch": 0.5, "learning_rate": 1.03440128263002e-05, "loss": 1.8241, "step": 20463 }, { "epoch": 0.5, "learning_rate": 1.034321496660876e-05, "loss": 2.1363, "step": 20464 }, { "epoch": 0.5, "learning_rate": 1.0342417104729902e-05, "loss": 2.0649, "step": 20465 }, { "epoch": 0.5, "learning_rate": 1.0341619240668706e-05, "loss": 2.1081, "step": 20466 }, { "epoch": 0.5, "learning_rate": 1.034082137443026e-05, "loss": 2.0235, "step": 20467 }, { "epoch": 0.5, "learning_rate": 1.034002350601965e-05, "loss": 1.9828, "step": 20468 }, { "epoch": 0.5, "learning_rate": 1.0339225635441956e-05, "loss": 2.0858, "step": 20469 }, { "epoch": 0.5, "learning_rate": 1.0338427762702267e-05, "loss": 1.9993, "step": 20470 }, { "epoch": 0.5, "learning_rate": 1.0337629887805668e-05, "loss": 2.0194, "step": 20471 }, { "epoch": 0.5, "learning_rate": 1.0336832010757245e-05, "loss": 2.1005, "step": 20472 }, { "epoch": 0.5, "learning_rate": 1.0336034131562081e-05, "loss": 2.1758, "step": 20473 }, { "epoch": 0.5, "learning_rate": 1.0335236250225263e-05, "loss": 2.033, "step": 20474 }, { "epoch": 0.5, "learning_rate": 1.0334438366751876e-05, "loss": 2.0531, "step": 20475 }, { "epoch": 0.5, "learning_rate": 1.0333640481147001e-05, "loss": 2.0245, "step": 20476 }, { "epoch": 0.5, "learning_rate": 1.0332842593415728e-05, "loss": 2.017, "step": 20477 }, { "epoch": 0.5, "learning_rate": 1.0332044703563138e-05, "loss": 2.1251, "step": 20478 }, { "epoch": 0.5, "learning_rate": 1.0331246811594321e-05, "loss": 2.0814, "step": 20479 }, { "epoch": 0.5, "learning_rate": 1.0330448917514364e-05, "loss": 2.143, "step": 20480 }, { "epoch": 0.5, "learning_rate": 1.0329651021328342e-05, "loss": 2.0804, "step": 20481 }, { "epoch": 0.5, "learning_rate": 1.0328853123041351e-05, "loss": 1.9352, "step": 20482 }, { "epoch": 0.5, "learning_rate": 1.032805522265847e-05, "loss": 2.0584, "step": 20483 }, { "epoch": 0.5, "learning_rate": 1.0327257320184789e-05, "loss": 1.8516, "step": 20484 }, { "epoch": 0.5, "learning_rate": 1.0326459415625386e-05, "loss": 2.1262, "step": 20485 }, { "epoch": 0.5, "learning_rate": 1.0325661508985354e-05, "loss": 1.9565, "step": 20486 }, { "epoch": 0.5, "learning_rate": 1.0324863600269772e-05, "loss": 2.0536, "step": 20487 }, { "epoch": 0.51, "learning_rate": 1.032406568948373e-05, "loss": 2.1712, "step": 20488 }, { "epoch": 0.51, "learning_rate": 1.0323267776632313e-05, "loss": 2.1154, "step": 20489 }, { "epoch": 0.51, "learning_rate": 1.0322469861720605e-05, "loss": 1.9448, "step": 20490 }, { "epoch": 0.51, "learning_rate": 1.0321671944753689e-05, "loss": 1.9877, "step": 20491 }, { "epoch": 0.51, "learning_rate": 1.0320874025736656e-05, "loss": 1.9526, "step": 20492 }, { "epoch": 0.51, "learning_rate": 1.0320076104674587e-05, "loss": 1.8514, "step": 20493 }, { "epoch": 0.51, "learning_rate": 1.0319278181572567e-05, "loss": 2.2071, "step": 20494 }, { "epoch": 0.51, "learning_rate": 1.0318480256435686e-05, "loss": 2.0603, "step": 20495 }, { "epoch": 0.51, "learning_rate": 1.0317682329269024e-05, "loss": 1.9832, "step": 20496 }, { "epoch": 0.51, "learning_rate": 1.0316884400077668e-05, "loss": 2.1299, "step": 20497 }, { "epoch": 0.51, "learning_rate": 1.0316086468866709e-05, "loss": 1.997, "step": 20498 }, { "epoch": 0.51, "learning_rate": 1.0315288535641227e-05, "loss": 1.8543, "step": 20499 }, { "epoch": 0.51, "learning_rate": 1.0314490600406305e-05, "loss": 2.0285, "step": 20500 }, { "epoch": 0.51, "learning_rate": 1.0313692663167033e-05, "loss": 2.1158, "step": 20501 }, { "epoch": 0.51, "learning_rate": 1.0312894723928497e-05, "loss": 1.8562, "step": 20502 }, { "epoch": 0.51, "learning_rate": 1.031209678269578e-05, "loss": 2.1596, "step": 20503 }, { "epoch": 0.51, "learning_rate": 1.031129883947397e-05, "loss": 2.1964, "step": 20504 }, { "epoch": 0.51, "learning_rate": 1.0310500894268147e-05, "loss": 2.0012, "step": 20505 }, { "epoch": 0.51, "learning_rate": 1.0309702947083404e-05, "loss": 2.0069, "step": 20506 }, { "epoch": 0.51, "learning_rate": 1.0308904997924823e-05, "loss": 2.0802, "step": 20507 }, { "epoch": 0.51, "learning_rate": 1.0308107046797489e-05, "loss": 1.9046, "step": 20508 }, { "epoch": 0.51, "learning_rate": 1.0307309093706491e-05, "loss": 2.1777, "step": 20509 }, { "epoch": 0.51, "learning_rate": 1.0306511138656907e-05, "loss": 2.0726, "step": 20510 }, { "epoch": 0.51, "learning_rate": 1.0305713181653831e-05, "loss": 2.0803, "step": 20511 }, { "epoch": 0.51, "learning_rate": 1.0304915222702343e-05, "loss": 2.0971, "step": 20512 }, { "epoch": 0.51, "learning_rate": 1.0304117261807533e-05, "loss": 2.0441, "step": 20513 }, { "epoch": 0.51, "learning_rate": 1.0303319298974486e-05, "loss": 2.0216, "step": 20514 }, { "epoch": 0.51, "learning_rate": 1.0302521334208282e-05, "loss": 1.9642, "step": 20515 }, { "epoch": 0.51, "learning_rate": 1.0301723367514014e-05, "loss": 2.12, "step": 20516 }, { "epoch": 0.51, "learning_rate": 1.0300925398896765e-05, "loss": 2.0424, "step": 20517 }, { "epoch": 0.51, "learning_rate": 1.030012742836162e-05, "loss": 2.0321, "step": 20518 }, { "epoch": 0.51, "learning_rate": 1.0299329455913665e-05, "loss": 2.067, "step": 20519 }, { "epoch": 0.51, "learning_rate": 1.0298531481557983e-05, "loss": 2.1738, "step": 20520 }, { "epoch": 0.51, "learning_rate": 1.0297733505299663e-05, "loss": 2.209, "step": 20521 }, { "epoch": 0.51, "learning_rate": 1.0296935527143793e-05, "loss": 1.9556, "step": 20522 }, { "epoch": 0.51, "learning_rate": 1.0296137547095454e-05, "loss": 2.0307, "step": 20523 }, { "epoch": 0.51, "learning_rate": 1.0295339565159735e-05, "loss": 1.9206, "step": 20524 }, { "epoch": 0.51, "learning_rate": 1.0294541581341722e-05, "loss": 2.0785, "step": 20525 }, { "epoch": 0.51, "learning_rate": 1.02937435956465e-05, "loss": 2.0079, "step": 20526 }, { "epoch": 0.51, "learning_rate": 1.0292945608079153e-05, "loss": 1.9454, "step": 20527 }, { "epoch": 0.51, "learning_rate": 1.0292147618644768e-05, "loss": 2.0452, "step": 20528 }, { "epoch": 0.51, "learning_rate": 1.0291349627348428e-05, "loss": 2.0058, "step": 20529 }, { "epoch": 0.51, "learning_rate": 1.0290551634195225e-05, "loss": 2.0413, "step": 20530 }, { "epoch": 0.51, "learning_rate": 1.0289753639190241e-05, "loss": 2.0875, "step": 20531 }, { "epoch": 0.51, "learning_rate": 1.0288955642338565e-05, "loss": 2.0997, "step": 20532 }, { "epoch": 0.51, "learning_rate": 1.0288157643645278e-05, "loss": 2.1341, "step": 20533 }, { "epoch": 0.51, "learning_rate": 1.028735964311547e-05, "loss": 1.8703, "step": 20534 }, { "epoch": 0.51, "learning_rate": 1.0286561640754222e-05, "loss": 1.9971, "step": 20535 }, { "epoch": 0.51, "learning_rate": 1.0285763636566628e-05, "loss": 2.0985, "step": 20536 }, { "epoch": 0.51, "learning_rate": 1.0284965630557767e-05, "loss": 2.1008, "step": 20537 }, { "epoch": 0.51, "learning_rate": 1.0284167622732726e-05, "loss": 2.1143, "step": 20538 }, { "epoch": 0.51, "learning_rate": 1.0283369613096592e-05, "loss": 2.1582, "step": 20539 }, { "epoch": 0.51, "learning_rate": 1.0282571601654453e-05, "loss": 2.081, "step": 20540 }, { "epoch": 0.51, "learning_rate": 1.0281773588411393e-05, "loss": 2.001, "step": 20541 }, { "epoch": 0.51, "learning_rate": 1.0280975573372498e-05, "loss": 1.9806, "step": 20542 }, { "epoch": 0.51, "learning_rate": 1.0280177556542855e-05, "loss": 1.9787, "step": 20543 }, { "epoch": 0.51, "learning_rate": 1.0279379537927546e-05, "loss": 1.9838, "step": 20544 }, { "epoch": 0.51, "learning_rate": 1.027858151753166e-05, "loss": 1.9709, "step": 20545 }, { "epoch": 0.51, "learning_rate": 1.0277783495360287e-05, "loss": 2.0999, "step": 20546 }, { "epoch": 0.51, "learning_rate": 1.0276985471418505e-05, "loss": 2.135, "step": 20547 }, { "epoch": 0.51, "learning_rate": 1.0276187445711408e-05, "loss": 2.0058, "step": 20548 }, { "epoch": 0.51, "learning_rate": 1.0275389418244078e-05, "loss": 1.9325, "step": 20549 }, { "epoch": 0.51, "learning_rate": 1.02745913890216e-05, "loss": 2.1863, "step": 20550 }, { "epoch": 0.51, "learning_rate": 1.0273793358049062e-05, "loss": 1.9229, "step": 20551 }, { "epoch": 0.51, "learning_rate": 1.0272995325331551e-05, "loss": 1.8147, "step": 20552 }, { "epoch": 0.51, "learning_rate": 1.0272197290874151e-05, "loss": 2.1052, "step": 20553 }, { "epoch": 0.51, "learning_rate": 1.0271399254681947e-05, "loss": 2.0741, "step": 20554 }, { "epoch": 0.51, "learning_rate": 1.027060121676003e-05, "loss": 1.9386, "step": 20555 }, { "epoch": 0.51, "learning_rate": 1.0269803177113479e-05, "loss": 2.0567, "step": 20556 }, { "epoch": 0.51, "learning_rate": 1.026900513574739e-05, "loss": 2.0018, "step": 20557 }, { "epoch": 0.51, "learning_rate": 1.026820709266684e-05, "loss": 2.2971, "step": 20558 }, { "epoch": 0.51, "learning_rate": 1.026740904787692e-05, "loss": 2.1365, "step": 20559 }, { "epoch": 0.51, "learning_rate": 1.0266611001382716e-05, "loss": 2.0919, "step": 20560 }, { "epoch": 0.51, "learning_rate": 1.0265812953189312e-05, "loss": 2.0744, "step": 20561 }, { "epoch": 0.51, "learning_rate": 1.0265014903301796e-05, "loss": 2.0044, "step": 20562 }, { "epoch": 0.51, "learning_rate": 1.026421685172525e-05, "loss": 1.9302, "step": 20563 }, { "epoch": 0.51, "learning_rate": 1.0263418798464767e-05, "loss": 2.0408, "step": 20564 }, { "epoch": 0.51, "learning_rate": 1.0262620743525432e-05, "loss": 2.1007, "step": 20565 }, { "epoch": 0.51, "learning_rate": 1.0261822686912327e-05, "loss": 1.9061, "step": 20566 }, { "epoch": 0.51, "learning_rate": 1.0261024628630543e-05, "loss": 1.7727, "step": 20567 }, { "epoch": 0.51, "learning_rate": 1.026022656868516e-05, "loss": 2.1196, "step": 20568 }, { "epoch": 0.51, "learning_rate": 1.025942850708127e-05, "loss": 1.9005, "step": 20569 }, { "epoch": 0.51, "learning_rate": 1.025863044382396e-05, "loss": 2.0148, "step": 20570 }, { "epoch": 0.51, "learning_rate": 1.0257832378918311e-05, "loss": 1.9229, "step": 20571 }, { "epoch": 0.51, "learning_rate": 1.0257034312369413e-05, "loss": 1.8574, "step": 20572 }, { "epoch": 0.51, "learning_rate": 1.0256236244182352e-05, "loss": 2.2053, "step": 20573 }, { "epoch": 0.51, "learning_rate": 1.0255438174362214e-05, "loss": 2.032, "step": 20574 }, { "epoch": 0.51, "learning_rate": 1.0254640102914084e-05, "loss": 2.0015, "step": 20575 }, { "epoch": 0.51, "learning_rate": 1.0253842029843051e-05, "loss": 1.8462, "step": 20576 }, { "epoch": 0.51, "learning_rate": 1.02530439551542e-05, "loss": 2.1506, "step": 20577 }, { "epoch": 0.51, "learning_rate": 1.0252245878852615e-05, "loss": 1.9958, "step": 20578 }, { "epoch": 0.51, "learning_rate": 1.0251447800943388e-05, "loss": 2.0157, "step": 20579 }, { "epoch": 0.51, "learning_rate": 1.0250649721431602e-05, "loss": 2.0506, "step": 20580 }, { "epoch": 0.51, "learning_rate": 1.0249851640322343e-05, "loss": 2.0511, "step": 20581 }, { "epoch": 0.51, "learning_rate": 1.0249053557620696e-05, "loss": 1.9073, "step": 20582 }, { "epoch": 0.51, "learning_rate": 1.024825547333175e-05, "loss": 2.0802, "step": 20583 }, { "epoch": 0.51, "learning_rate": 1.0247457387460594e-05, "loss": 2.09, "step": 20584 }, { "epoch": 0.51, "learning_rate": 1.024665930001231e-05, "loss": 2.059, "step": 20585 }, { "epoch": 0.51, "learning_rate": 1.0245861210991985e-05, "loss": 2.0054, "step": 20586 }, { "epoch": 0.51, "learning_rate": 1.0245063120404705e-05, "loss": 1.964, "step": 20587 }, { "epoch": 0.51, "learning_rate": 1.0244265028255563e-05, "loss": 1.9324, "step": 20588 }, { "epoch": 0.51, "learning_rate": 1.0243466934549638e-05, "loss": 2.15, "step": 20589 }, { "epoch": 0.51, "learning_rate": 1.0242668839292018e-05, "loss": 2.1429, "step": 20590 }, { "epoch": 0.51, "learning_rate": 1.024187074248779e-05, "loss": 1.9265, "step": 20591 }, { "epoch": 0.51, "learning_rate": 1.0241072644142038e-05, "loss": 1.9813, "step": 20592 }, { "epoch": 0.51, "learning_rate": 1.0240274544259856e-05, "loss": 1.9311, "step": 20593 }, { "epoch": 0.51, "learning_rate": 1.0239476442846326e-05, "loss": 2.0361, "step": 20594 }, { "epoch": 0.51, "learning_rate": 1.0238678339906533e-05, "loss": 1.9635, "step": 20595 }, { "epoch": 0.51, "learning_rate": 1.0237880235445567e-05, "loss": 2.068, "step": 20596 }, { "epoch": 0.51, "learning_rate": 1.023708212946851e-05, "loss": 2.0145, "step": 20597 }, { "epoch": 0.51, "learning_rate": 1.0236284021980454e-05, "loss": 2.1399, "step": 20598 }, { "epoch": 0.51, "learning_rate": 1.0235485912986483e-05, "loss": 1.9837, "step": 20599 }, { "epoch": 0.51, "learning_rate": 1.0234687802491679e-05, "loss": 2.0463, "step": 20600 }, { "epoch": 0.51, "learning_rate": 1.0233889690501137e-05, "loss": 1.9677, "step": 20601 }, { "epoch": 0.51, "learning_rate": 1.0233091577019937e-05, "loss": 2.1369, "step": 20602 }, { "epoch": 0.51, "learning_rate": 1.023229346205317e-05, "loss": 2.081, "step": 20603 }, { "epoch": 0.51, "learning_rate": 1.0231495345605924e-05, "loss": 2.0508, "step": 20604 }, { "epoch": 0.51, "learning_rate": 1.0230697227683282e-05, "loss": 1.9162, "step": 20605 }, { "epoch": 0.51, "learning_rate": 1.0229899108290329e-05, "loss": 1.8781, "step": 20606 }, { "epoch": 0.51, "learning_rate": 1.0229100987432155e-05, "loss": 2.0979, "step": 20607 }, { "epoch": 0.51, "learning_rate": 1.0228302865113843e-05, "loss": 1.928, "step": 20608 }, { "epoch": 0.51, "learning_rate": 1.0227504741340485e-05, "loss": 2.1, "step": 20609 }, { "epoch": 0.51, "learning_rate": 1.0226706616117166e-05, "loss": 1.9176, "step": 20610 }, { "epoch": 0.51, "learning_rate": 1.0225908489448973e-05, "loss": 1.8483, "step": 20611 }, { "epoch": 0.51, "learning_rate": 1.0225110361340986e-05, "loss": 2.0795, "step": 20612 }, { "epoch": 0.51, "learning_rate": 1.0224312231798302e-05, "loss": 1.9294, "step": 20613 }, { "epoch": 0.51, "learning_rate": 1.0223514100826002e-05, "loss": 1.8773, "step": 20614 }, { "epoch": 0.51, "learning_rate": 1.0222715968429176e-05, "loss": 2.0749, "step": 20615 }, { "epoch": 0.51, "learning_rate": 1.0221917834612904e-05, "loss": 2.0662, "step": 20616 }, { "epoch": 0.51, "learning_rate": 1.022111969938228e-05, "loss": 1.9882, "step": 20617 }, { "epoch": 0.51, "learning_rate": 1.0220321562742387e-05, "loss": 2.0099, "step": 20618 }, { "epoch": 0.51, "learning_rate": 1.0219523424698312e-05, "loss": 2.3111, "step": 20619 }, { "epoch": 0.51, "learning_rate": 1.0218725285255146e-05, "loss": 2.1808, "step": 20620 }, { "epoch": 0.51, "learning_rate": 1.021792714441797e-05, "loss": 2.1375, "step": 20621 }, { "epoch": 0.51, "learning_rate": 1.0217129002191874e-05, "loss": 2.0108, "step": 20622 }, { "epoch": 0.51, "learning_rate": 1.0216330858581945e-05, "loss": 2.0477, "step": 20623 }, { "epoch": 0.51, "learning_rate": 1.0215532713593267e-05, "loss": 2.0283, "step": 20624 }, { "epoch": 0.51, "learning_rate": 1.0214734567230931e-05, "loss": 2.0243, "step": 20625 }, { "epoch": 0.51, "learning_rate": 1.0213936419500019e-05, "loss": 2.1346, "step": 20626 }, { "epoch": 0.51, "learning_rate": 1.021313827040562e-05, "loss": 2.1232, "step": 20627 }, { "epoch": 0.51, "learning_rate": 1.0212340119952823e-05, "loss": 2.0988, "step": 20628 }, { "epoch": 0.51, "learning_rate": 1.0211541968146716e-05, "loss": 2.0791, "step": 20629 }, { "epoch": 0.51, "learning_rate": 1.021074381499238e-05, "loss": 2.1158, "step": 20630 }, { "epoch": 0.51, "learning_rate": 1.0209945660494905e-05, "loss": 2.0353, "step": 20631 }, { "epoch": 0.51, "learning_rate": 1.0209147504659378e-05, "loss": 1.8547, "step": 20632 }, { "epoch": 0.51, "learning_rate": 1.0208349347490887e-05, "loss": 1.9969, "step": 20633 }, { "epoch": 0.51, "learning_rate": 1.0207551188994517e-05, "loss": 1.9782, "step": 20634 }, { "epoch": 0.51, "learning_rate": 1.0206753029175355e-05, "loss": 2.0457, "step": 20635 }, { "epoch": 0.51, "learning_rate": 1.0205954868038488e-05, "loss": 2.0507, "step": 20636 }, { "epoch": 0.51, "learning_rate": 1.0205156705589006e-05, "loss": 1.9597, "step": 20637 }, { "epoch": 0.51, "learning_rate": 1.020435854183199e-05, "loss": 1.9614, "step": 20638 }, { "epoch": 0.51, "learning_rate": 1.0203560376772534e-05, "loss": 1.9492, "step": 20639 }, { "epoch": 0.51, "learning_rate": 1.0202762210415723e-05, "loss": 2.0154, "step": 20640 }, { "epoch": 0.51, "learning_rate": 1.0201964042766637e-05, "loss": 2.0846, "step": 20641 }, { "epoch": 0.51, "learning_rate": 1.020116587383037e-05, "loss": 2.1197, "step": 20642 }, { "epoch": 0.51, "learning_rate": 1.0200367703612008e-05, "loss": 1.9597, "step": 20643 }, { "epoch": 0.51, "learning_rate": 1.0199569532116638e-05, "loss": 1.9464, "step": 20644 }, { "epoch": 0.51, "learning_rate": 1.0198771359349342e-05, "loss": 1.9707, "step": 20645 }, { "epoch": 0.51, "learning_rate": 1.0197973185315216e-05, "loss": 2.0763, "step": 20646 }, { "epoch": 0.51, "learning_rate": 1.0197175010019342e-05, "loss": 1.9339, "step": 20647 }, { "epoch": 0.51, "learning_rate": 1.0196376833466807e-05, "loss": 2.1116, "step": 20648 }, { "epoch": 0.51, "learning_rate": 1.01955786556627e-05, "loss": 1.8526, "step": 20649 }, { "epoch": 0.51, "learning_rate": 1.0194780476612103e-05, "loss": 1.7979, "step": 20650 }, { "epoch": 0.51, "learning_rate": 1.019398229632011e-05, "loss": 2.0095, "step": 20651 }, { "epoch": 0.51, "learning_rate": 1.0193184114791801e-05, "loss": 1.9901, "step": 20652 }, { "epoch": 0.51, "learning_rate": 1.0192385932032268e-05, "loss": 1.874, "step": 20653 }, { "epoch": 0.51, "learning_rate": 1.0191587748046597e-05, "loss": 1.968, "step": 20654 }, { "epoch": 0.51, "learning_rate": 1.0190789562839874e-05, "loss": 2.1314, "step": 20655 }, { "epoch": 0.51, "learning_rate": 1.0189991376417187e-05, "loss": 1.9769, "step": 20656 }, { "epoch": 0.51, "learning_rate": 1.0189193188783627e-05, "loss": 1.931, "step": 20657 }, { "epoch": 0.51, "learning_rate": 1.0188394999944274e-05, "loss": 1.9904, "step": 20658 }, { "epoch": 0.51, "learning_rate": 1.0187596809904218e-05, "loss": 2.0928, "step": 20659 }, { "epoch": 0.51, "learning_rate": 1.0186798618668548e-05, "loss": 2.1005, "step": 20660 }, { "epoch": 0.51, "learning_rate": 1.0186000426242348e-05, "loss": 2.1031, "step": 20661 }, { "epoch": 0.51, "learning_rate": 1.0185202232630705e-05, "loss": 2.1534, "step": 20662 }, { "epoch": 0.51, "learning_rate": 1.0184404037838711e-05, "loss": 2.0496, "step": 20663 }, { "epoch": 0.51, "learning_rate": 1.018360584187145e-05, "loss": 1.9187, "step": 20664 }, { "epoch": 0.51, "learning_rate": 1.0182807644734004e-05, "loss": 2.1482, "step": 20665 }, { "epoch": 0.51, "learning_rate": 1.018200944643147e-05, "loss": 1.9798, "step": 20666 }, { "epoch": 0.51, "learning_rate": 1.0181211246968932e-05, "loss": 1.9674, "step": 20667 }, { "epoch": 0.51, "learning_rate": 1.0180413046351475e-05, "loss": 2.0241, "step": 20668 }, { "epoch": 0.51, "learning_rate": 1.0179614844584185e-05, "loss": 2.0314, "step": 20669 }, { "epoch": 0.51, "learning_rate": 1.0178816641672149e-05, "loss": 2.1891, "step": 20670 }, { "epoch": 0.51, "learning_rate": 1.0178018437620459e-05, "loss": 2.1272, "step": 20671 }, { "epoch": 0.51, "learning_rate": 1.0177220232434197e-05, "loss": 2.1248, "step": 20672 }, { "epoch": 0.51, "learning_rate": 1.0176422026118457e-05, "loss": 1.7389, "step": 20673 }, { "epoch": 0.51, "learning_rate": 1.0175623818678317e-05, "loss": 1.9084, "step": 20674 }, { "epoch": 0.51, "learning_rate": 1.0174825610118872e-05, "loss": 2.0579, "step": 20675 }, { "epoch": 0.51, "learning_rate": 1.0174027400445207e-05, "loss": 1.9695, "step": 20676 }, { "epoch": 0.51, "learning_rate": 1.017322918966241e-05, "loss": 2.1646, "step": 20677 }, { "epoch": 0.51, "learning_rate": 1.0172430977775564e-05, "loss": 2.2072, "step": 20678 }, { "epoch": 0.51, "learning_rate": 1.017163276478976e-05, "loss": 2.0081, "step": 20679 }, { "epoch": 0.51, "learning_rate": 1.0170834550710082e-05, "loss": 2.014, "step": 20680 }, { "epoch": 0.51, "learning_rate": 1.0170036335541624e-05, "loss": 2.0936, "step": 20681 }, { "epoch": 0.51, "learning_rate": 1.0169238119289469e-05, "loss": 2.0317, "step": 20682 }, { "epoch": 0.51, "learning_rate": 1.0168439901958703e-05, "loss": 2.2654, "step": 20683 }, { "epoch": 0.51, "learning_rate": 1.0167641683554412e-05, "loss": 1.9036, "step": 20684 }, { "epoch": 0.51, "learning_rate": 1.0166843464081688e-05, "loss": 2.0491, "step": 20685 }, { "epoch": 0.51, "learning_rate": 1.0166045243545617e-05, "loss": 1.9018, "step": 20686 }, { "epoch": 0.51, "learning_rate": 1.0165247021951285e-05, "loss": 2.1452, "step": 20687 }, { "epoch": 0.51, "learning_rate": 1.016444879930378e-05, "loss": 1.905, "step": 20688 }, { "epoch": 0.51, "learning_rate": 1.0163650575608185e-05, "loss": 2.0883, "step": 20689 }, { "epoch": 0.51, "learning_rate": 1.01628523508696e-05, "loss": 1.9432, "step": 20690 }, { "epoch": 0.51, "learning_rate": 1.01620541250931e-05, "loss": 1.9711, "step": 20691 }, { "epoch": 0.51, "learning_rate": 1.0161255898283774e-05, "loss": 2.0837, "step": 20692 }, { "epoch": 0.51, "learning_rate": 1.0160457670446715e-05, "loss": 1.8801, "step": 20693 }, { "epoch": 0.51, "learning_rate": 1.0159659441587005e-05, "loss": 1.8089, "step": 20694 }, { "epoch": 0.51, "learning_rate": 1.0158861211709733e-05, "loss": 2.0725, "step": 20695 }, { "epoch": 0.51, "learning_rate": 1.0158062980819987e-05, "loss": 1.9823, "step": 20696 }, { "epoch": 0.51, "learning_rate": 1.0157264748922852e-05, "loss": 2.1793, "step": 20697 }, { "epoch": 0.51, "learning_rate": 1.0156466516023422e-05, "loss": 1.9655, "step": 20698 }, { "epoch": 0.51, "learning_rate": 1.0155668282126775e-05, "loss": 2.0637, "step": 20699 }, { "epoch": 0.51, "learning_rate": 1.0154870047238007e-05, "loss": 1.9987, "step": 20700 }, { "epoch": 0.51, "learning_rate": 1.0154071811362203e-05, "loss": 2.0835, "step": 20701 }, { "epoch": 0.51, "learning_rate": 1.0153273574504448e-05, "loss": 1.9546, "step": 20702 }, { "epoch": 0.51, "learning_rate": 1.0152475336669828e-05, "loss": 1.9019, "step": 20703 }, { "epoch": 0.51, "learning_rate": 1.0151677097863436e-05, "loss": 2.1686, "step": 20704 }, { "epoch": 0.51, "learning_rate": 1.0150878858090354e-05, "loss": 2.1382, "step": 20705 }, { "epoch": 0.51, "learning_rate": 1.015008061735567e-05, "loss": 1.9184, "step": 20706 }, { "epoch": 0.51, "learning_rate": 1.0149282375664478e-05, "loss": 1.8344, "step": 20707 }, { "epoch": 0.51, "learning_rate": 1.0148484133021858e-05, "loss": 2.196, "step": 20708 }, { "epoch": 0.51, "learning_rate": 1.0147685889432901e-05, "loss": 2.2108, "step": 20709 }, { "epoch": 0.51, "learning_rate": 1.0146887644902697e-05, "loss": 2.2151, "step": 20710 }, { "epoch": 0.51, "learning_rate": 1.0146089399436326e-05, "loss": 2.0287, "step": 20711 }, { "epoch": 0.51, "learning_rate": 1.014529115303888e-05, "loss": 2.3089, "step": 20712 }, { "epoch": 0.51, "learning_rate": 1.0144492905715446e-05, "loss": 1.9001, "step": 20713 }, { "epoch": 0.51, "learning_rate": 1.0143694657471115e-05, "loss": 2.0445, "step": 20714 }, { "epoch": 0.51, "learning_rate": 1.0142896408310965e-05, "loss": 1.9058, "step": 20715 }, { "epoch": 0.51, "learning_rate": 1.0142098158240095e-05, "loss": 2.0218, "step": 20716 }, { "epoch": 0.51, "learning_rate": 1.0141299907263586e-05, "loss": 2.011, "step": 20717 }, { "epoch": 0.51, "learning_rate": 1.0140501655386523e-05, "loss": 1.8504, "step": 20718 }, { "epoch": 0.51, "learning_rate": 1.0139703402614003e-05, "loss": 2.0683, "step": 20719 }, { "epoch": 0.51, "learning_rate": 1.0138905148951104e-05, "loss": 1.9564, "step": 20720 }, { "epoch": 0.51, "learning_rate": 1.013810689440292e-05, "loss": 1.9813, "step": 20721 }, { "epoch": 0.51, "learning_rate": 1.0137308638974532e-05, "loss": 2.0158, "step": 20722 }, { "epoch": 0.51, "learning_rate": 1.0136510382671035e-05, "loss": 2.07, "step": 20723 }, { "epoch": 0.51, "learning_rate": 1.0135712125497507e-05, "loss": 1.8497, "step": 20724 }, { "epoch": 0.51, "learning_rate": 1.0134913867459046e-05, "loss": 1.9622, "step": 20725 }, { "epoch": 0.51, "learning_rate": 1.0134115608560736e-05, "loss": 1.8825, "step": 20726 }, { "epoch": 0.51, "learning_rate": 1.0133317348807663e-05, "loss": 2.0974, "step": 20727 }, { "epoch": 0.51, "learning_rate": 1.013251908820491e-05, "loss": 1.9218, "step": 20728 }, { "epoch": 0.51, "learning_rate": 1.0131720826757577e-05, "loss": 1.9917, "step": 20729 }, { "epoch": 0.51, "learning_rate": 1.013092256447074e-05, "loss": 2.0925, "step": 20730 }, { "epoch": 0.51, "learning_rate": 1.0130124301349491e-05, "loss": 2.0254, "step": 20731 }, { "epoch": 0.51, "learning_rate": 1.012932603739892e-05, "loss": 2.0525, "step": 20732 }, { "epoch": 0.51, "learning_rate": 1.0128527772624107e-05, "loss": 2.1781, "step": 20733 }, { "epoch": 0.51, "learning_rate": 1.0127729507030148e-05, "loss": 2.0027, "step": 20734 }, { "epoch": 0.51, "learning_rate": 1.0126931240622127e-05, "loss": 2.0433, "step": 20735 }, { "epoch": 0.51, "learning_rate": 1.0126132973405132e-05, "loss": 1.9726, "step": 20736 }, { "epoch": 0.51, "learning_rate": 1.0125334705384248e-05, "loss": 1.7735, "step": 20737 }, { "epoch": 0.51, "learning_rate": 1.0124536436564568e-05, "loss": 2.0046, "step": 20738 }, { "epoch": 0.51, "learning_rate": 1.0123738166951178e-05, "loss": 2.0667, "step": 20739 }, { "epoch": 0.51, "learning_rate": 1.012293989654916e-05, "loss": 1.9478, "step": 20740 }, { "epoch": 0.51, "learning_rate": 1.0122141625363611e-05, "loss": 1.9703, "step": 20741 }, { "epoch": 0.51, "learning_rate": 1.0121343353399608e-05, "loss": 1.8792, "step": 20742 }, { "epoch": 0.51, "learning_rate": 1.0120545080662247e-05, "loss": 2.1619, "step": 20743 }, { "epoch": 0.51, "learning_rate": 1.0119746807156615e-05, "loss": 2.1472, "step": 20744 }, { "epoch": 0.51, "learning_rate": 1.0118948532887796e-05, "loss": 2.1197, "step": 20745 }, { "epoch": 0.51, "learning_rate": 1.0118150257860879e-05, "loss": 2.0844, "step": 20746 }, { "epoch": 0.51, "learning_rate": 1.0117351982080951e-05, "loss": 1.9424, "step": 20747 }, { "epoch": 0.51, "learning_rate": 1.0116553705553102e-05, "loss": 2.0184, "step": 20748 }, { "epoch": 0.51, "learning_rate": 1.0115755428282419e-05, "loss": 2.0454, "step": 20749 }, { "epoch": 0.51, "learning_rate": 1.0114957150273985e-05, "loss": 1.872, "step": 20750 }, { "epoch": 0.51, "learning_rate": 1.0114158871532895e-05, "loss": 2.2299, "step": 20751 }, { "epoch": 0.51, "learning_rate": 1.011336059206423e-05, "loss": 2.0461, "step": 20752 }, { "epoch": 0.51, "learning_rate": 1.0112562311873084e-05, "loss": 1.9622, "step": 20753 }, { "epoch": 0.51, "learning_rate": 1.0111764030964543e-05, "loss": 2.1893, "step": 20754 }, { "epoch": 0.51, "learning_rate": 1.0110965749343693e-05, "loss": 1.8628, "step": 20755 }, { "epoch": 0.51, "learning_rate": 1.0110167467015622e-05, "loss": 2.1666, "step": 20756 }, { "epoch": 0.51, "learning_rate": 1.0109369183985414e-05, "loss": 2.0902, "step": 20757 }, { "epoch": 0.51, "learning_rate": 1.0108570900258163e-05, "loss": 2.0316, "step": 20758 }, { "epoch": 0.51, "learning_rate": 1.0107772615838953e-05, "loss": 1.9902, "step": 20759 }, { "epoch": 0.51, "learning_rate": 1.0106974330732875e-05, "loss": 2.1201, "step": 20760 }, { "epoch": 0.51, "learning_rate": 1.0106176044945012e-05, "loss": 2.1018, "step": 20761 }, { "epoch": 0.51, "learning_rate": 1.0105377758480457e-05, "loss": 1.9945, "step": 20762 }, { "epoch": 0.51, "learning_rate": 1.0104579471344296e-05, "loss": 1.9013, "step": 20763 }, { "epoch": 0.51, "learning_rate": 1.0103781183541617e-05, "loss": 2.1838, "step": 20764 }, { "epoch": 0.51, "learning_rate": 1.0102982895077505e-05, "loss": 2.0773, "step": 20765 }, { "epoch": 0.51, "learning_rate": 1.0102184605957046e-05, "loss": 2.0846, "step": 20766 }, { "epoch": 0.51, "learning_rate": 1.0101386316185335e-05, "loss": 2.0447, "step": 20767 }, { "epoch": 0.51, "learning_rate": 1.0100588025767452e-05, "loss": 2.0417, "step": 20768 }, { "epoch": 0.51, "learning_rate": 1.0099789734708494e-05, "loss": 2.0141, "step": 20769 }, { "epoch": 0.51, "learning_rate": 1.0098991443013542e-05, "loss": 2.0338, "step": 20770 }, { "epoch": 0.51, "learning_rate": 1.0098193150687683e-05, "loss": 2.117, "step": 20771 }, { "epoch": 0.51, "learning_rate": 1.009739485773601e-05, "loss": 1.9769, "step": 20772 }, { "epoch": 0.51, "learning_rate": 1.0096596564163606e-05, "loss": 2.0714, "step": 20773 }, { "epoch": 0.51, "learning_rate": 1.009579826997556e-05, "loss": 2.0074, "step": 20774 }, { "epoch": 0.51, "learning_rate": 1.0094999975176962e-05, "loss": 1.8721, "step": 20775 }, { "epoch": 0.51, "learning_rate": 1.0094201679772897e-05, "loss": 2.1818, "step": 20776 }, { "epoch": 0.51, "learning_rate": 1.0093403383768454e-05, "loss": 2.2016, "step": 20777 }, { "epoch": 0.51, "learning_rate": 1.009260508716872e-05, "loss": 2.1117, "step": 20778 }, { "epoch": 0.51, "learning_rate": 1.0091806789978789e-05, "loss": 2.1158, "step": 20779 }, { "epoch": 0.51, "learning_rate": 1.0091008492203738e-05, "loss": 2.1296, "step": 20780 }, { "epoch": 0.51, "learning_rate": 1.0090210193848661e-05, "loss": 2.0532, "step": 20781 }, { "epoch": 0.51, "learning_rate": 1.0089411894918646e-05, "loss": 2.0877, "step": 20782 }, { "epoch": 0.51, "learning_rate": 1.0088613595418782e-05, "loss": 2.0515, "step": 20783 }, { "epoch": 0.51, "learning_rate": 1.0087815295354149e-05, "loss": 2.0245, "step": 20784 }, { "epoch": 0.51, "learning_rate": 1.0087016994729844e-05, "loss": 1.9056, "step": 20785 }, { "epoch": 0.51, "learning_rate": 1.0086218693550951e-05, "loss": 1.9243, "step": 20786 }, { "epoch": 0.51, "learning_rate": 1.0085420391822559e-05, "loss": 2.1049, "step": 20787 }, { "epoch": 0.51, "learning_rate": 1.0084622089549753e-05, "loss": 2.1338, "step": 20788 }, { "epoch": 0.51, "learning_rate": 1.0083823786737625e-05, "loss": 2.0379, "step": 20789 }, { "epoch": 0.51, "learning_rate": 1.008302548339126e-05, "loss": 2.0388, "step": 20790 }, { "epoch": 0.51, "learning_rate": 1.0082227179515747e-05, "loss": 1.9282, "step": 20791 }, { "epoch": 0.51, "learning_rate": 1.0081428875116172e-05, "loss": 2.0757, "step": 20792 }, { "epoch": 0.51, "learning_rate": 1.0080630570197623e-05, "loss": 2.0471, "step": 20793 }, { "epoch": 0.51, "learning_rate": 1.0079832264765193e-05, "loss": 2.0112, "step": 20794 }, { "epoch": 0.51, "learning_rate": 1.007903395882396e-05, "loss": 2.1444, "step": 20795 }, { "epoch": 0.51, "learning_rate": 1.0078235652379025e-05, "loss": 1.9754, "step": 20796 }, { "epoch": 0.51, "learning_rate": 1.0077437345435466e-05, "loss": 1.8043, "step": 20797 }, { "epoch": 0.51, "learning_rate": 1.0076639037998372e-05, "loss": 2.1097, "step": 20798 }, { "epoch": 0.51, "learning_rate": 1.0075840730072834e-05, "loss": 2.0052, "step": 20799 }, { "epoch": 0.51, "learning_rate": 1.0075042421663937e-05, "loss": 2.0846, "step": 20800 }, { "epoch": 0.51, "learning_rate": 1.0074244112776772e-05, "loss": 2.0564, "step": 20801 }, { "epoch": 0.51, "learning_rate": 1.0073445803416421e-05, "loss": 1.9275, "step": 20802 }, { "epoch": 0.51, "learning_rate": 1.0072647493587982e-05, "loss": 2.0141, "step": 20803 }, { "epoch": 0.51, "learning_rate": 1.0071849183296534e-05, "loss": 1.8824, "step": 20804 }, { "epoch": 0.51, "learning_rate": 1.0071050872547166e-05, "loss": 1.9828, "step": 20805 }, { "epoch": 0.51, "learning_rate": 1.0070252561344971e-05, "loss": 2.0009, "step": 20806 }, { "epoch": 0.51, "learning_rate": 1.0069454249695031e-05, "loss": 2.0133, "step": 20807 }, { "epoch": 0.51, "learning_rate": 1.0068655937602439e-05, "loss": 2.059, "step": 20808 }, { "epoch": 0.51, "learning_rate": 1.006785762507228e-05, "loss": 2.0906, "step": 20809 }, { "epoch": 0.51, "learning_rate": 1.0067059312109637e-05, "loss": 1.9979, "step": 20810 }, { "epoch": 0.51, "learning_rate": 1.0066260998719605e-05, "loss": 2.0599, "step": 20811 }, { "epoch": 0.51, "learning_rate": 1.0065462684907275e-05, "loss": 1.9429, "step": 20812 }, { "epoch": 0.51, "learning_rate": 1.0064664370677725e-05, "loss": 2.0286, "step": 20813 }, { "epoch": 0.51, "learning_rate": 1.0063866056036051e-05, "loss": 1.9525, "step": 20814 }, { "epoch": 0.51, "learning_rate": 1.0063067740987336e-05, "loss": 2.0112, "step": 20815 }, { "epoch": 0.51, "learning_rate": 1.006226942553667e-05, "loss": 2.0218, "step": 20816 }, { "epoch": 0.51, "learning_rate": 1.0061471109689143e-05, "loss": 1.9709, "step": 20817 }, { "epoch": 0.51, "learning_rate": 1.006067279344984e-05, "loss": 1.9625, "step": 20818 }, { "epoch": 0.51, "learning_rate": 1.0059874476823846e-05, "loss": 2.121, "step": 20819 }, { "epoch": 0.51, "learning_rate": 1.0059076159816253e-05, "loss": 2.1801, "step": 20820 }, { "epoch": 0.51, "learning_rate": 1.0058277842432151e-05, "loss": 2.1374, "step": 20821 }, { "epoch": 0.51, "learning_rate": 1.0057479524676625e-05, "loss": 2.0054, "step": 20822 }, { "epoch": 0.51, "learning_rate": 1.0056681206554764e-05, "loss": 2.0774, "step": 20823 }, { "epoch": 0.51, "learning_rate": 1.005588288807165e-05, "loss": 1.9021, "step": 20824 }, { "epoch": 0.51, "learning_rate": 1.0055084569232383e-05, "loss": 2.034, "step": 20825 }, { "epoch": 0.51, "learning_rate": 1.005428625004204e-05, "loss": 1.9958, "step": 20826 }, { "epoch": 0.51, "learning_rate": 1.0053487930505714e-05, "loss": 1.993, "step": 20827 }, { "epoch": 0.51, "learning_rate": 1.0052689610628492e-05, "loss": 1.9929, "step": 20828 }, { "epoch": 0.51, "learning_rate": 1.0051891290415461e-05, "loss": 2.0689, "step": 20829 }, { "epoch": 0.51, "learning_rate": 1.0051092969871709e-05, "loss": 1.9305, "step": 20830 }, { "epoch": 0.51, "learning_rate": 1.005029464900233e-05, "loss": 1.9741, "step": 20831 }, { "epoch": 0.51, "learning_rate": 1.0049496327812403e-05, "loss": 1.8413, "step": 20832 }, { "epoch": 0.51, "learning_rate": 1.0048698006307021e-05, "loss": 1.8846, "step": 20833 }, { "epoch": 0.51, "learning_rate": 1.004789968449127e-05, "loss": 2.1179, "step": 20834 }, { "epoch": 0.51, "learning_rate": 1.0047101362370238e-05, "loss": 1.9718, "step": 20835 }, { "epoch": 0.51, "learning_rate": 1.0046303039949015e-05, "loss": 1.7862, "step": 20836 }, { "epoch": 0.51, "learning_rate": 1.0045504717232687e-05, "loss": 1.9431, "step": 20837 }, { "epoch": 0.51, "learning_rate": 1.0044706394226342e-05, "loss": 1.9546, "step": 20838 }, { "epoch": 0.51, "learning_rate": 1.004390807093507e-05, "loss": 2.2223, "step": 20839 }, { "epoch": 0.51, "learning_rate": 1.0043109747363954e-05, "loss": 1.994, "step": 20840 }, { "epoch": 0.51, "learning_rate": 1.004231142351809e-05, "loss": 1.9676, "step": 20841 }, { "epoch": 0.51, "learning_rate": 1.004151309940256e-05, "loss": 1.9764, "step": 20842 }, { "epoch": 0.51, "learning_rate": 1.0040714775022455e-05, "loss": 2.0406, "step": 20843 }, { "epoch": 0.51, "learning_rate": 1.0039916450382858e-05, "loss": 2.1136, "step": 20844 }, { "epoch": 0.51, "learning_rate": 1.0039118125488862e-05, "loss": 1.9919, "step": 20845 }, { "epoch": 0.51, "learning_rate": 1.0038319800345552e-05, "loss": 2.1324, "step": 20846 }, { "epoch": 0.51, "learning_rate": 1.003752147495802e-05, "loss": 1.8615, "step": 20847 }, { "epoch": 0.51, "learning_rate": 1.003672314933135e-05, "loss": 2.0143, "step": 20848 }, { "epoch": 0.51, "learning_rate": 1.0035924823470633e-05, "loss": 1.9219, "step": 20849 }, { "epoch": 0.51, "learning_rate": 1.0035126497380954e-05, "loss": 1.9826, "step": 20850 }, { "epoch": 0.51, "learning_rate": 1.0034328171067404e-05, "loss": 2.1198, "step": 20851 }, { "epoch": 0.51, "learning_rate": 1.0033529844535068e-05, "loss": 2.0009, "step": 20852 }, { "epoch": 0.51, "learning_rate": 1.0032731517789034e-05, "loss": 1.8823, "step": 20853 }, { "epoch": 0.51, "learning_rate": 1.0031933190834393e-05, "loss": 1.9772, "step": 20854 }, { "epoch": 0.51, "learning_rate": 1.003113486367623e-05, "loss": 2.0535, "step": 20855 }, { "epoch": 0.51, "learning_rate": 1.0030336536319635e-05, "loss": 1.9229, "step": 20856 }, { "epoch": 0.51, "learning_rate": 1.0029538208769698e-05, "loss": 2.0013, "step": 20857 }, { "epoch": 0.51, "learning_rate": 1.0028739881031502e-05, "loss": 1.9412, "step": 20858 }, { "epoch": 0.51, "learning_rate": 1.0027941553110136e-05, "loss": 1.908, "step": 20859 }, { "epoch": 0.51, "learning_rate": 1.0027143225010694e-05, "loss": 2.034, "step": 20860 }, { "epoch": 0.51, "learning_rate": 1.0026344896738256e-05, "loss": 2.0382, "step": 20861 }, { "epoch": 0.51, "learning_rate": 1.0025546568297915e-05, "loss": 1.9234, "step": 20862 }, { "epoch": 0.51, "learning_rate": 1.0024748239694754e-05, "loss": 1.9363, "step": 20863 }, { "epoch": 0.51, "learning_rate": 1.0023949910933866e-05, "loss": 1.811, "step": 20864 }, { "epoch": 0.51, "learning_rate": 1.0023151582020338e-05, "loss": 1.9348, "step": 20865 }, { "epoch": 0.51, "learning_rate": 1.0022353252959258e-05, "loss": 1.9858, "step": 20866 }, { "epoch": 0.51, "learning_rate": 1.0021554923755714e-05, "loss": 2.0761, "step": 20867 }, { "epoch": 0.51, "learning_rate": 1.0020756594414792e-05, "loss": 2.2289, "step": 20868 }, { "epoch": 0.51, "learning_rate": 1.0019958264941581e-05, "loss": 2.1182, "step": 20869 }, { "epoch": 0.51, "learning_rate": 1.0019159935341172e-05, "loss": 2.0605, "step": 20870 }, { "epoch": 0.51, "learning_rate": 1.001836160561865e-05, "loss": 1.9842, "step": 20871 }, { "epoch": 0.51, "learning_rate": 1.0017563275779099e-05, "loss": 2.0047, "step": 20872 }, { "epoch": 0.51, "learning_rate": 1.0016764945827613e-05, "loss": 2.0963, "step": 20873 }, { "epoch": 0.51, "learning_rate": 1.0015966615769282e-05, "loss": 1.9854, "step": 20874 }, { "epoch": 0.51, "learning_rate": 1.0015168285609189e-05, "loss": 1.9062, "step": 20875 }, { "epoch": 0.51, "learning_rate": 1.0014369955352424e-05, "loss": 2.0943, "step": 20876 }, { "epoch": 0.51, "learning_rate": 1.0013571625004074e-05, "loss": 1.97, "step": 20877 }, { "epoch": 0.51, "learning_rate": 1.0012773294569227e-05, "loss": 2.0585, "step": 20878 }, { "epoch": 0.51, "learning_rate": 1.0011974964052974e-05, "loss": 1.8792, "step": 20879 }, { "epoch": 0.51, "learning_rate": 1.0011176633460397e-05, "loss": 1.9478, "step": 20880 }, { "epoch": 0.51, "learning_rate": 1.0010378302796589e-05, "loss": 2.0391, "step": 20881 }, { "epoch": 0.51, "learning_rate": 1.0009579972066636e-05, "loss": 1.9161, "step": 20882 }, { "epoch": 0.51, "learning_rate": 1.0008781641275628e-05, "loss": 2.1166, "step": 20883 }, { "epoch": 0.51, "learning_rate": 1.0007983310428652e-05, "loss": 2.0881, "step": 20884 }, { "epoch": 0.51, "learning_rate": 1.0007184979530797e-05, "loss": 1.9105, "step": 20885 }, { "epoch": 0.51, "learning_rate": 1.0006386648587149e-05, "loss": 1.8486, "step": 20886 }, { "epoch": 0.51, "learning_rate": 1.0005588317602795e-05, "loss": 1.9567, "step": 20887 }, { "epoch": 0.51, "learning_rate": 1.0004789986582826e-05, "loss": 1.9168, "step": 20888 }, { "epoch": 0.51, "learning_rate": 1.000399165553233e-05, "loss": 2.1644, "step": 20889 }, { "epoch": 0.51, "learning_rate": 1.000319332445639e-05, "loss": 1.8496, "step": 20890 }, { "epoch": 0.51, "learning_rate": 1.0002394993360103e-05, "loss": 2.0466, "step": 20891 }, { "epoch": 0.51, "learning_rate": 1.0001596662248547e-05, "loss": 1.9471, "step": 20892 }, { "epoch": 0.51, "learning_rate": 1.0000798331126818e-05, "loss": 2.1357, "step": 20893 }, { "epoch": 0.52, "learning_rate": 1e-05, "loss": 1.9662, "step": 20894 }, { "epoch": 0.52, "learning_rate": 9.999201668873182e-06, "loss": 2.0316, "step": 20895 }, { "epoch": 0.52, "learning_rate": 9.998403337751456e-06, "loss": 2.0199, "step": 20896 }, { "epoch": 0.52, "learning_rate": 9.9976050066399e-06, "loss": 1.9669, "step": 20897 }, { "epoch": 0.52, "learning_rate": 9.996806675543613e-06, "loss": 1.8876, "step": 20898 }, { "epoch": 0.52, "learning_rate": 9.996008344467672e-06, "loss": 1.93, "step": 20899 }, { "epoch": 0.52, "learning_rate": 9.995210013417174e-06, "loss": 2.1812, "step": 20900 }, { "epoch": 0.52, "learning_rate": 9.994411682397209e-06, "loss": 1.9992, "step": 20901 }, { "epoch": 0.52, "learning_rate": 9.993613351412852e-06, "loss": 2.1831, "step": 20902 }, { "epoch": 0.52, "learning_rate": 9.992815020469208e-06, "loss": 2.0771, "step": 20903 }, { "epoch": 0.52, "learning_rate": 9.99201668957135e-06, "loss": 1.9166, "step": 20904 }, { "epoch": 0.52, "learning_rate": 9.991218358724372e-06, "loss": 2.1178, "step": 20905 }, { "epoch": 0.52, "learning_rate": 9.990420027933366e-06, "loss": 2.0677, "step": 20906 }, { "epoch": 0.52, "learning_rate": 9.989621697203413e-06, "loss": 1.7885, "step": 20907 }, { "epoch": 0.52, "learning_rate": 9.988823366539606e-06, "loss": 1.8678, "step": 20908 }, { "epoch": 0.52, "learning_rate": 9.98802503594703e-06, "loss": 1.9427, "step": 20909 }, { "epoch": 0.52, "learning_rate": 9.987226705430774e-06, "loss": 1.9637, "step": 20910 }, { "epoch": 0.52, "learning_rate": 9.98642837499593e-06, "loss": 1.9464, "step": 20911 }, { "epoch": 0.52, "learning_rate": 9.985630044647578e-06, "loss": 1.9915, "step": 20912 }, { "epoch": 0.52, "learning_rate": 9.984831714390816e-06, "loss": 1.9167, "step": 20913 }, { "epoch": 0.52, "learning_rate": 9.984033384230721e-06, "loss": 2.1023, "step": 20914 }, { "epoch": 0.52, "learning_rate": 9.983235054172387e-06, "loss": 2.0886, "step": 20915 }, { "epoch": 0.52, "learning_rate": 9.982436724220905e-06, "loss": 2.0428, "step": 20916 }, { "epoch": 0.52, "learning_rate": 9.981638394381354e-06, "loss": 1.9414, "step": 20917 }, { "epoch": 0.52, "learning_rate": 9.980840064658835e-06, "loss": 2.0118, "step": 20918 }, { "epoch": 0.52, "learning_rate": 9.980041735058422e-06, "loss": 1.7813, "step": 20919 }, { "epoch": 0.52, "learning_rate": 9.97924340558521e-06, "loss": 2.198, "step": 20920 }, { "epoch": 0.52, "learning_rate": 9.97844507624429e-06, "loss": 1.9592, "step": 20921 }, { "epoch": 0.52, "learning_rate": 9.977646747040745e-06, "loss": 2.0254, "step": 20922 }, { "epoch": 0.52, "learning_rate": 9.976848417979664e-06, "loss": 2.0539, "step": 20923 }, { "epoch": 0.52, "learning_rate": 9.976050089066137e-06, "loss": 2.0175, "step": 20924 }, { "epoch": 0.52, "learning_rate": 9.975251760305247e-06, "loss": 2.0071, "step": 20925 }, { "epoch": 0.52, "learning_rate": 9.97445343170209e-06, "loss": 1.9802, "step": 20926 }, { "epoch": 0.52, "learning_rate": 9.973655103261747e-06, "loss": 2.1257, "step": 20927 }, { "epoch": 0.52, "learning_rate": 9.972856774989307e-06, "loss": 1.9683, "step": 20928 }, { "epoch": 0.52, "learning_rate": 9.972058446889866e-06, "loss": 2.0923, "step": 20929 }, { "epoch": 0.52, "learning_rate": 9.9712601189685e-06, "loss": 2.0179, "step": 20930 }, { "epoch": 0.52, "learning_rate": 9.970461791230307e-06, "loss": 1.845, "step": 20931 }, { "epoch": 0.52, "learning_rate": 9.969663463680366e-06, "loss": 2.1243, "step": 20932 }, { "epoch": 0.52, "learning_rate": 9.96886513632377e-06, "loss": 1.9324, "step": 20933 }, { "epoch": 0.52, "learning_rate": 9.968066809165608e-06, "loss": 2.1674, "step": 20934 }, { "epoch": 0.52, "learning_rate": 9.967268482210967e-06, "loss": 1.9793, "step": 20935 }, { "epoch": 0.52, "learning_rate": 9.966470155464937e-06, "loss": 1.9878, "step": 20936 }, { "epoch": 0.52, "learning_rate": 9.9656718289326e-06, "loss": 2.1351, "step": 20937 }, { "epoch": 0.52, "learning_rate": 9.964873502619045e-06, "loss": 2.2438, "step": 20938 }, { "epoch": 0.52, "learning_rate": 9.96407517652937e-06, "loss": 2.0239, "step": 20939 }, { "epoch": 0.52, "learning_rate": 9.96327685066865e-06, "loss": 2.1564, "step": 20940 }, { "epoch": 0.52, "learning_rate": 9.962478525041981e-06, "loss": 1.8054, "step": 20941 }, { "epoch": 0.52, "learning_rate": 9.96168019965445e-06, "loss": 2.1833, "step": 20942 }, { "epoch": 0.52, "learning_rate": 9.96088187451114e-06, "loss": 2.0176, "step": 20943 }, { "epoch": 0.52, "learning_rate": 9.960083549617145e-06, "loss": 2.1065, "step": 20944 }, { "epoch": 0.52, "learning_rate": 9.959285224977548e-06, "loss": 1.8446, "step": 20945 }, { "epoch": 0.52, "learning_rate": 9.958486900597444e-06, "loss": 2.1721, "step": 20946 }, { "epoch": 0.52, "learning_rate": 9.957688576481911e-06, "loss": 2.0718, "step": 20947 }, { "epoch": 0.52, "learning_rate": 9.956890252636046e-06, "loss": 1.9613, "step": 20948 }, { "epoch": 0.52, "learning_rate": 9.956091929064935e-06, "loss": 2.0601, "step": 20949 }, { "epoch": 0.52, "learning_rate": 9.955293605773661e-06, "loss": 2.1145, "step": 20950 }, { "epoch": 0.52, "learning_rate": 9.954495282767317e-06, "loss": 1.9152, "step": 20951 }, { "epoch": 0.52, "learning_rate": 9.953696960050988e-06, "loss": 1.9853, "step": 20952 }, { "epoch": 0.52, "learning_rate": 9.952898637629762e-06, "loss": 1.89, "step": 20953 }, { "epoch": 0.52, "learning_rate": 9.952100315508733e-06, "loss": 1.9303, "step": 20954 }, { "epoch": 0.52, "learning_rate": 9.95130199369298e-06, "loss": 1.9197, "step": 20955 }, { "epoch": 0.52, "learning_rate": 9.950503672187602e-06, "loss": 2.0237, "step": 20956 }, { "epoch": 0.52, "learning_rate": 9.949705350997674e-06, "loss": 2.0402, "step": 20957 }, { "epoch": 0.52, "learning_rate": 9.948907030128289e-06, "loss": 2.0784, "step": 20958 }, { "epoch": 0.52, "learning_rate": 9.948108709584542e-06, "loss": 1.9146, "step": 20959 }, { "epoch": 0.52, "learning_rate": 9.947310389371511e-06, "loss": 1.9924, "step": 20960 }, { "epoch": 0.52, "learning_rate": 9.94651206949429e-06, "loss": 2.0875, "step": 20961 }, { "epoch": 0.52, "learning_rate": 9.945713749957963e-06, "loss": 1.8477, "step": 20962 }, { "epoch": 0.52, "learning_rate": 9.944915430767619e-06, "loss": 1.9911, "step": 20963 }, { "epoch": 0.52, "learning_rate": 9.944117111928352e-06, "loss": 1.9895, "step": 20964 }, { "epoch": 0.52, "learning_rate": 9.94331879344524e-06, "loss": 2.0215, "step": 20965 }, { "epoch": 0.52, "learning_rate": 9.94252047532338e-06, "loss": 1.9195, "step": 20966 }, { "epoch": 0.52, "learning_rate": 9.941722157567852e-06, "loss": 2.1216, "step": 20967 }, { "epoch": 0.52, "learning_rate": 9.940923840183747e-06, "loss": 1.9361, "step": 20968 }, { "epoch": 0.52, "learning_rate": 9.940125523176158e-06, "loss": 2.1438, "step": 20969 }, { "epoch": 0.52, "learning_rate": 9.939327206550164e-06, "loss": 2.0032, "step": 20970 }, { "epoch": 0.52, "learning_rate": 9.938528890310862e-06, "loss": 1.8877, "step": 20971 }, { "epoch": 0.52, "learning_rate": 9.937730574463332e-06, "loss": 2.2871, "step": 20972 }, { "epoch": 0.52, "learning_rate": 9.936932259012666e-06, "loss": 1.9629, "step": 20973 }, { "epoch": 0.52, "learning_rate": 9.936133943963952e-06, "loss": 2.1155, "step": 20974 }, { "epoch": 0.52, "learning_rate": 9.935335629322277e-06, "loss": 1.9968, "step": 20975 }, { "epoch": 0.52, "learning_rate": 9.93453731509273e-06, "loss": 2.1596, "step": 20976 }, { "epoch": 0.52, "learning_rate": 9.933739001280396e-06, "loss": 1.9639, "step": 20977 }, { "epoch": 0.52, "learning_rate": 9.932940687890366e-06, "loss": 1.9882, "step": 20978 }, { "epoch": 0.52, "learning_rate": 9.932142374927727e-06, "loss": 2.1799, "step": 20979 }, { "epoch": 0.52, "learning_rate": 9.931344062397564e-06, "loss": 2.1384, "step": 20980 }, { "epoch": 0.52, "learning_rate": 9.930545750304974e-06, "loss": 2.0251, "step": 20981 }, { "epoch": 0.52, "learning_rate": 9.929747438655034e-06, "loss": 1.9169, "step": 20982 }, { "epoch": 0.52, "learning_rate": 9.928949127452836e-06, "loss": 2.0526, "step": 20983 }, { "epoch": 0.52, "learning_rate": 9.928150816703471e-06, "loss": 2.0483, "step": 20984 }, { "epoch": 0.52, "learning_rate": 9.927352506412022e-06, "loss": 2.0688, "step": 20985 }, { "epoch": 0.52, "learning_rate": 9.926554196583577e-06, "loss": 2.1071, "step": 20986 }, { "epoch": 0.52, "learning_rate": 9.92575588722323e-06, "loss": 2.0618, "step": 20987 }, { "epoch": 0.52, "learning_rate": 9.924957578336063e-06, "loss": 2.2235, "step": 20988 }, { "epoch": 0.52, "learning_rate": 9.924159269927171e-06, "loss": 2.059, "step": 20989 }, { "epoch": 0.52, "learning_rate": 9.92336096200163e-06, "loss": 1.9436, "step": 20990 }, { "epoch": 0.52, "learning_rate": 9.922562654564536e-06, "loss": 1.9337, "step": 20991 }, { "epoch": 0.52, "learning_rate": 9.921764347620978e-06, "loss": 2.1251, "step": 20992 }, { "epoch": 0.52, "learning_rate": 9.92096604117604e-06, "loss": 1.9969, "step": 20993 }, { "epoch": 0.52, "learning_rate": 9.92016773523481e-06, "loss": 1.8317, "step": 20994 }, { "epoch": 0.52, "learning_rate": 9.919369429802378e-06, "loss": 2.1299, "step": 20995 }, { "epoch": 0.52, "learning_rate": 9.918571124883832e-06, "loss": 2.0619, "step": 20996 }, { "epoch": 0.52, "learning_rate": 9.917772820484257e-06, "loss": 2.1249, "step": 20997 }, { "epoch": 0.52, "learning_rate": 9.916974516608741e-06, "loss": 1.9806, "step": 20998 }, { "epoch": 0.52, "learning_rate": 9.91617621326238e-06, "loss": 2.003, "step": 20999 }, { "epoch": 0.52, "learning_rate": 9.915377910450249e-06, "loss": 2.0663, "step": 21000 }, { "epoch": 0.52, "learning_rate": 9.914579608177443e-06, "loss": 1.9226, "step": 21001 }, { "epoch": 0.52, "learning_rate": 9.913781306449052e-06, "loss": 1.8482, "step": 21002 }, { "epoch": 0.52, "learning_rate": 9.912983005270158e-06, "loss": 1.9336, "step": 21003 }, { "epoch": 0.52, "learning_rate": 9.912184704645854e-06, "loss": 2.061, "step": 21004 }, { "epoch": 0.52, "learning_rate": 9.911386404581223e-06, "loss": 2.3741, "step": 21005 }, { "epoch": 0.52, "learning_rate": 9.910588105081354e-06, "loss": 1.871, "step": 21006 }, { "epoch": 0.52, "learning_rate": 9.909789806151342e-06, "loss": 2.1064, "step": 21007 }, { "epoch": 0.52, "learning_rate": 9.908991507796264e-06, "loss": 1.9903, "step": 21008 }, { "epoch": 0.52, "learning_rate": 9.908193210021218e-06, "loss": 2.1273, "step": 21009 }, { "epoch": 0.52, "learning_rate": 9.907394912831281e-06, "loss": 2.1826, "step": 21010 }, { "epoch": 0.52, "learning_rate": 9.906596616231546e-06, "loss": 1.964, "step": 21011 }, { "epoch": 0.52, "learning_rate": 9.905798320227105e-06, "loss": 2.0171, "step": 21012 }, { "epoch": 0.52, "learning_rate": 9.905000024823042e-06, "loss": 1.9581, "step": 21013 }, { "epoch": 0.52, "learning_rate": 9.904201730024443e-06, "loss": 1.7834, "step": 21014 }, { "epoch": 0.52, "learning_rate": 9.903403435836397e-06, "loss": 2.0881, "step": 21015 }, { "epoch": 0.52, "learning_rate": 9.902605142263992e-06, "loss": 1.8563, "step": 21016 }, { "epoch": 0.52, "learning_rate": 9.90180684931232e-06, "loss": 2.0568, "step": 21017 }, { "epoch": 0.52, "learning_rate": 9.901008556986461e-06, "loss": 1.9935, "step": 21018 }, { "epoch": 0.52, "learning_rate": 9.900210265291511e-06, "loss": 2.0561, "step": 21019 }, { "epoch": 0.52, "learning_rate": 9.89941197423255e-06, "loss": 1.9709, "step": 21020 }, { "epoch": 0.52, "learning_rate": 9.898613683814668e-06, "loss": 1.9163, "step": 21021 }, { "epoch": 0.52, "learning_rate": 9.897815394042957e-06, "loss": 2.0251, "step": 21022 }, { "epoch": 0.52, "learning_rate": 9.8970171049225e-06, "loss": 1.9625, "step": 21023 }, { "epoch": 0.52, "learning_rate": 9.89621881645839e-06, "loss": 2.0608, "step": 21024 }, { "epoch": 0.52, "learning_rate": 9.895420528655707e-06, "loss": 1.932, "step": 21025 }, { "epoch": 0.52, "learning_rate": 9.894622241519543e-06, "loss": 2.0869, "step": 21026 }, { "epoch": 0.52, "learning_rate": 9.89382395505499e-06, "loss": 1.91, "step": 21027 }, { "epoch": 0.52, "learning_rate": 9.893025669267129e-06, "loss": 2.075, "step": 21028 }, { "epoch": 0.52, "learning_rate": 9.89222738416105e-06, "loss": 2.0913, "step": 21029 }, { "epoch": 0.52, "learning_rate": 9.891429099741839e-06, "loss": 2.169, "step": 21030 }, { "epoch": 0.52, "learning_rate": 9.890630816014589e-06, "loss": 2.0462, "step": 21031 }, { "epoch": 0.52, "learning_rate": 9.889832532984385e-06, "loss": 2.149, "step": 21032 }, { "epoch": 0.52, "learning_rate": 9.889034250656312e-06, "loss": 2.0202, "step": 21033 }, { "epoch": 0.52, "learning_rate": 9.888235969035462e-06, "loss": 2.2152, "step": 21034 }, { "epoch": 0.52, "learning_rate": 9.887437688126918e-06, "loss": 2.032, "step": 21035 }, { "epoch": 0.52, "learning_rate": 9.88663940793577e-06, "loss": 1.8833, "step": 21036 }, { "epoch": 0.52, "learning_rate": 9.88584112846711e-06, "loss": 2.0045, "step": 21037 }, { "epoch": 0.52, "learning_rate": 9.885042849726018e-06, "loss": 2.0302, "step": 21038 }, { "epoch": 0.52, "learning_rate": 9.884244571717588e-06, "loss": 2.0457, "step": 21039 }, { "epoch": 0.52, "learning_rate": 9.883446294446901e-06, "loss": 2.006, "step": 21040 }, { "epoch": 0.52, "learning_rate": 9.88264801791905e-06, "loss": 1.9667, "step": 21041 }, { "epoch": 0.52, "learning_rate": 9.881849742139126e-06, "loss": 2.0846, "step": 21042 }, { "epoch": 0.52, "learning_rate": 9.881051467112208e-06, "loss": 2.2019, "step": 21043 }, { "epoch": 0.52, "learning_rate": 9.880253192843387e-06, "loss": 1.8787, "step": 21044 }, { "epoch": 0.52, "learning_rate": 9.879454919337755e-06, "loss": 2.0849, "step": 21045 }, { "epoch": 0.52, "learning_rate": 9.878656646600394e-06, "loss": 1.918, "step": 21046 }, { "epoch": 0.52, "learning_rate": 9.877858374636394e-06, "loss": 1.9666, "step": 21047 }, { "epoch": 0.52, "learning_rate": 9.877060103450842e-06, "loss": 1.7966, "step": 21048 }, { "epoch": 0.52, "learning_rate": 9.876261833048825e-06, "loss": 1.8011, "step": 21049 }, { "epoch": 0.52, "learning_rate": 9.875463563435433e-06, "loss": 1.8948, "step": 21050 }, { "epoch": 0.52, "learning_rate": 9.874665294615752e-06, "loss": 1.7357, "step": 21051 }, { "epoch": 0.52, "learning_rate": 9.873867026594873e-06, "loss": 1.892, "step": 21052 }, { "epoch": 0.52, "learning_rate": 9.873068759377874e-06, "loss": 1.9684, "step": 21053 }, { "epoch": 0.52, "learning_rate": 9.872270492969853e-06, "loss": 2.0954, "step": 21054 }, { "epoch": 0.52, "learning_rate": 9.871472227375897e-06, "loss": 2.0284, "step": 21055 }, { "epoch": 0.52, "learning_rate": 9.870673962601085e-06, "loss": 2.0221, "step": 21056 }, { "epoch": 0.52, "learning_rate": 9.869875698650514e-06, "loss": 1.9534, "step": 21057 }, { "epoch": 0.52, "learning_rate": 9.869077435529263e-06, "loss": 2.1506, "step": 21058 }, { "epoch": 0.52, "learning_rate": 9.868279173242425e-06, "loss": 2.0208, "step": 21059 }, { "epoch": 0.52, "learning_rate": 9.867480911795093e-06, "loss": 2.0381, "step": 21060 }, { "epoch": 0.52, "learning_rate": 9.86668265119234e-06, "loss": 2.0093, "step": 21061 }, { "epoch": 0.52, "learning_rate": 9.86588439143927e-06, "loss": 2.0321, "step": 21062 }, { "epoch": 0.52, "learning_rate": 9.865086132540956e-06, "loss": 1.9671, "step": 21063 }, { "epoch": 0.52, "learning_rate": 9.864287874502493e-06, "loss": 2.0461, "step": 21064 }, { "epoch": 0.52, "learning_rate": 9.86348961732897e-06, "loss": 1.9965, "step": 21065 }, { "epoch": 0.52, "learning_rate": 9.862691361025471e-06, "loss": 2.0822, "step": 21066 }, { "epoch": 0.52, "learning_rate": 9.861893105597086e-06, "loss": 2.1991, "step": 21067 }, { "epoch": 0.52, "learning_rate": 9.861094851048898e-06, "loss": 1.9735, "step": 21068 }, { "epoch": 0.52, "learning_rate": 9.860296597385999e-06, "loss": 1.9629, "step": 21069 }, { "epoch": 0.52, "learning_rate": 9.859498344613478e-06, "loss": 2.0426, "step": 21070 }, { "epoch": 0.52, "learning_rate": 9.858700092736417e-06, "loss": 1.9475, "step": 21071 }, { "epoch": 0.52, "learning_rate": 9.85790184175991e-06, "loss": 1.9948, "step": 21072 }, { "epoch": 0.52, "learning_rate": 9.857103591689037e-06, "loss": 2.1386, "step": 21073 }, { "epoch": 0.52, "learning_rate": 9.856305342528888e-06, "loss": 1.9009, "step": 21074 }, { "epoch": 0.52, "learning_rate": 9.855507094284557e-06, "loss": 2.166, "step": 21075 }, { "epoch": 0.52, "learning_rate": 9.854708846961121e-06, "loss": 2.075, "step": 21076 }, { "epoch": 0.52, "learning_rate": 9.85391060056368e-06, "loss": 2.0612, "step": 21077 }, { "epoch": 0.52, "learning_rate": 9.853112355097308e-06, "loss": 2.0062, "step": 21078 }, { "epoch": 0.52, "learning_rate": 9.852314110567099e-06, "loss": 1.9084, "step": 21079 }, { "epoch": 0.52, "learning_rate": 9.851515866978145e-06, "loss": 2.0504, "step": 21080 }, { "epoch": 0.52, "learning_rate": 9.850717624335524e-06, "loss": 2.153, "step": 21081 }, { "epoch": 0.52, "learning_rate": 9.849919382644331e-06, "loss": 2.1008, "step": 21082 }, { "epoch": 0.52, "learning_rate": 9.849121141909648e-06, "loss": 2.024, "step": 21083 }, { "epoch": 0.52, "learning_rate": 9.848322902136568e-06, "loss": 2.062, "step": 21084 }, { "epoch": 0.52, "learning_rate": 9.847524663330174e-06, "loss": 2.124, "step": 21085 }, { "epoch": 0.52, "learning_rate": 9.846726425495555e-06, "loss": 2.1272, "step": 21086 }, { "epoch": 0.52, "learning_rate": 9.845928188637802e-06, "loss": 2.2066, "step": 21087 }, { "epoch": 0.52, "learning_rate": 9.845129952761994e-06, "loss": 2.0976, "step": 21088 }, { "epoch": 0.52, "learning_rate": 9.844331717873225e-06, "loss": 1.9386, "step": 21089 }, { "epoch": 0.52, "learning_rate": 9.843533483976583e-06, "loss": 1.9941, "step": 21090 }, { "epoch": 0.52, "learning_rate": 9.84273525107715e-06, "loss": 1.9966, "step": 21091 }, { "epoch": 0.52, "learning_rate": 9.841937019180019e-06, "loss": 1.8847, "step": 21092 }, { "epoch": 0.52, "learning_rate": 9.841138788290269e-06, "loss": 1.83, "step": 21093 }, { "epoch": 0.52, "learning_rate": 9.840340558412997e-06, "loss": 2.0759, "step": 21094 }, { "epoch": 0.52, "learning_rate": 9.83954232955329e-06, "loss": 2.0855, "step": 21095 }, { "epoch": 0.52, "learning_rate": 9.838744101716227e-06, "loss": 1.8459, "step": 21096 }, { "epoch": 0.52, "learning_rate": 9.837945874906907e-06, "loss": 1.9984, "step": 21097 }, { "epoch": 0.52, "learning_rate": 9.837147649130404e-06, "loss": 2.0461, "step": 21098 }, { "epoch": 0.52, "learning_rate": 9.836349424391813e-06, "loss": 1.8097, "step": 21099 }, { "epoch": 0.52, "learning_rate": 9.835551200696222e-06, "loss": 1.9831, "step": 21100 }, { "epoch": 0.52, "learning_rate": 9.834752978048717e-06, "loss": 2.0677, "step": 21101 }, { "epoch": 0.52, "learning_rate": 9.833954756454385e-06, "loss": 2.001, "step": 21102 }, { "epoch": 0.52, "learning_rate": 9.833156535918314e-06, "loss": 1.9997, "step": 21103 }, { "epoch": 0.52, "learning_rate": 9.832358316445588e-06, "loss": 2.2222, "step": 21104 }, { "epoch": 0.52, "learning_rate": 9.831560098041302e-06, "loss": 2.0483, "step": 21105 }, { "epoch": 0.52, "learning_rate": 9.830761880710534e-06, "loss": 2.0009, "step": 21106 }, { "epoch": 0.52, "learning_rate": 9.829963664458376e-06, "loss": 2.1095, "step": 21107 }, { "epoch": 0.52, "learning_rate": 9.82916544928992e-06, "loss": 2.1189, "step": 21108 }, { "epoch": 0.52, "learning_rate": 9.828367235210242e-06, "loss": 2.105, "step": 21109 }, { "epoch": 0.52, "learning_rate": 9.82756902222444e-06, "loss": 1.8767, "step": 21110 }, { "epoch": 0.52, "learning_rate": 9.826770810337593e-06, "loss": 2.079, "step": 21111 }, { "epoch": 0.52, "learning_rate": 9.825972599554791e-06, "loss": 2.0918, "step": 21112 }, { "epoch": 0.52, "learning_rate": 9.82517438988113e-06, "loss": 1.9476, "step": 21113 }, { "epoch": 0.52, "learning_rate": 9.824376181321683e-06, "loss": 2.0274, "step": 21114 }, { "epoch": 0.52, "learning_rate": 9.823577973881548e-06, "loss": 2.0307, "step": 21115 }, { "epoch": 0.52, "learning_rate": 9.822779767565805e-06, "loss": 2.1584, "step": 21116 }, { "epoch": 0.52, "learning_rate": 9.821981562379543e-06, "loss": 1.8825, "step": 21117 }, { "epoch": 0.52, "learning_rate": 9.821183358327854e-06, "loss": 1.9511, "step": 21118 }, { "epoch": 0.52, "learning_rate": 9.820385155415818e-06, "loss": 2.0968, "step": 21119 }, { "epoch": 0.52, "learning_rate": 9.81958695364853e-06, "loss": 2.086, "step": 21120 }, { "epoch": 0.52, "learning_rate": 9.818788753031071e-06, "loss": 2.068, "step": 21121 }, { "epoch": 0.52, "learning_rate": 9.817990553568529e-06, "loss": 1.9463, "step": 21122 }, { "epoch": 0.52, "learning_rate": 9.817192355265997e-06, "loss": 1.9959, "step": 21123 }, { "epoch": 0.52, "learning_rate": 9.816394158128554e-06, "loss": 1.9545, "step": 21124 }, { "epoch": 0.52, "learning_rate": 9.815595962161295e-06, "loss": 1.8731, "step": 21125 }, { "epoch": 0.52, "learning_rate": 9.814797767369297e-06, "loss": 2.0096, "step": 21126 }, { "epoch": 0.52, "learning_rate": 9.813999573757655e-06, "loss": 2.0298, "step": 21127 }, { "epoch": 0.52, "learning_rate": 9.813201381331457e-06, "loss": 1.8397, "step": 21128 }, { "epoch": 0.52, "learning_rate": 9.812403190095784e-06, "loss": 2.0969, "step": 21129 }, { "epoch": 0.52, "learning_rate": 9.811605000055731e-06, "loss": 2.0181, "step": 21130 }, { "epoch": 0.52, "learning_rate": 9.810806811216377e-06, "loss": 2.1478, "step": 21131 }, { "epoch": 0.52, "learning_rate": 9.810008623582813e-06, "loss": 1.9014, "step": 21132 }, { "epoch": 0.52, "learning_rate": 9.80921043716013e-06, "loss": 1.8246, "step": 21133 }, { "epoch": 0.52, "learning_rate": 9.808412251953405e-06, "loss": 2.0771, "step": 21134 }, { "epoch": 0.52, "learning_rate": 9.807614067967735e-06, "loss": 1.8008, "step": 21135 }, { "epoch": 0.52, "learning_rate": 9.8068158852082e-06, "loss": 1.9696, "step": 21136 }, { "epoch": 0.52, "learning_rate": 9.806017703679892e-06, "loss": 1.9978, "step": 21137 }, { "epoch": 0.52, "learning_rate": 9.8052195233879e-06, "loss": 2.0187, "step": 21138 }, { "epoch": 0.52, "learning_rate": 9.804421344337304e-06, "loss": 1.9659, "step": 21139 }, { "epoch": 0.52, "learning_rate": 9.803623166533198e-06, "loss": 1.8691, "step": 21140 }, { "epoch": 0.52, "learning_rate": 9.802824989980661e-06, "loss": 1.9602, "step": 21141 }, { "epoch": 0.52, "learning_rate": 9.802026814684784e-06, "loss": 1.9091, "step": 21142 }, { "epoch": 0.52, "learning_rate": 9.80122864065066e-06, "loss": 1.9284, "step": 21143 }, { "epoch": 0.52, "learning_rate": 9.800430467883365e-06, "loss": 2.0565, "step": 21144 }, { "epoch": 0.52, "learning_rate": 9.799632296387997e-06, "loss": 1.9878, "step": 21145 }, { "epoch": 0.52, "learning_rate": 9.798834126169631e-06, "loss": 2.04, "step": 21146 }, { "epoch": 0.52, "learning_rate": 9.798035957233364e-06, "loss": 1.9651, "step": 21147 }, { "epoch": 0.52, "learning_rate": 9.797237789584282e-06, "loss": 2.1102, "step": 21148 }, { "epoch": 0.52, "learning_rate": 9.796439623227467e-06, "loss": 2.0465, "step": 21149 }, { "epoch": 0.52, "learning_rate": 9.795641458168012e-06, "loss": 1.9607, "step": 21150 }, { "epoch": 0.52, "learning_rate": 9.794843294410998e-06, "loss": 2.0656, "step": 21151 }, { "epoch": 0.52, "learning_rate": 9.794045131961512e-06, "loss": 2.1855, "step": 21152 }, { "epoch": 0.52, "learning_rate": 9.793246970824646e-06, "loss": 2.1437, "step": 21153 }, { "epoch": 0.52, "learning_rate": 9.792448811005484e-06, "loss": 2.0217, "step": 21154 }, { "epoch": 0.52, "learning_rate": 9.791650652509117e-06, "loss": 2.0963, "step": 21155 }, { "epoch": 0.52, "learning_rate": 9.790852495340624e-06, "loss": 2.245, "step": 21156 }, { "epoch": 0.52, "learning_rate": 9.790054339505097e-06, "loss": 2.0008, "step": 21157 }, { "epoch": 0.52, "learning_rate": 9.789256185007624e-06, "loss": 2.0486, "step": 21158 }, { "epoch": 0.52, "learning_rate": 9.788458031853287e-06, "loss": 1.9739, "step": 21159 }, { "epoch": 0.52, "learning_rate": 9.78765988004718e-06, "loss": 1.9984, "step": 21160 }, { "epoch": 0.52, "learning_rate": 9.786861729594383e-06, "loss": 1.9922, "step": 21161 }, { "epoch": 0.52, "learning_rate": 9.786063580499983e-06, "loss": 2.1949, "step": 21162 }, { "epoch": 0.52, "learning_rate": 9.785265432769074e-06, "loss": 1.9485, "step": 21163 }, { "epoch": 0.52, "learning_rate": 9.784467286406734e-06, "loss": 2.1023, "step": 21164 }, { "epoch": 0.52, "learning_rate": 9.783669141418055e-06, "loss": 2.1732, "step": 21165 }, { "epoch": 0.52, "learning_rate": 9.78287099780813e-06, "loss": 2.009, "step": 21166 }, { "epoch": 0.52, "learning_rate": 9.782072855582031e-06, "loss": 1.9238, "step": 21167 }, { "epoch": 0.52, "learning_rate": 9.781274714744859e-06, "loss": 2.0637, "step": 21168 }, { "epoch": 0.52, "learning_rate": 9.78047657530169e-06, "loss": 2.043, "step": 21169 }, { "epoch": 0.52, "learning_rate": 9.779678437257615e-06, "loss": 2.067, "step": 21170 }, { "epoch": 0.52, "learning_rate": 9.778880300617723e-06, "loss": 1.9378, "step": 21171 }, { "epoch": 0.52, "learning_rate": 9.778082165387097e-06, "loss": 1.9594, "step": 21172 }, { "epoch": 0.52, "learning_rate": 9.777284031570829e-06, "loss": 1.9019, "step": 21173 }, { "epoch": 0.52, "learning_rate": 9.776485899174e-06, "loss": 1.9626, "step": 21174 }, { "epoch": 0.52, "learning_rate": 9.775687768201698e-06, "loss": 1.972, "step": 21175 }, { "epoch": 0.52, "learning_rate": 9.774889638659017e-06, "loss": 1.9627, "step": 21176 }, { "epoch": 0.52, "learning_rate": 9.774091510551032e-06, "loss": 1.9821, "step": 21177 }, { "epoch": 0.52, "learning_rate": 9.773293383882839e-06, "loss": 1.9889, "step": 21178 }, { "epoch": 0.52, "learning_rate": 9.772495258659518e-06, "loss": 2.1231, "step": 21179 }, { "epoch": 0.52, "learning_rate": 9.771697134886158e-06, "loss": 1.9985, "step": 21180 }, { "epoch": 0.52, "learning_rate": 9.77089901256785e-06, "loss": 2.1462, "step": 21181 }, { "epoch": 0.52, "learning_rate": 9.770100891709673e-06, "loss": 2.093, "step": 21182 }, { "epoch": 0.52, "learning_rate": 9.769302772316725e-06, "loss": 1.9898, "step": 21183 }, { "epoch": 0.52, "learning_rate": 9.76850465439408e-06, "loss": 2.0384, "step": 21184 }, { "epoch": 0.52, "learning_rate": 9.767706537946828e-06, "loss": 2.0618, "step": 21185 }, { "epoch": 0.52, "learning_rate": 9.766908422980064e-06, "loss": 2.1111, "step": 21186 }, { "epoch": 0.52, "learning_rate": 9.766110309498865e-06, "loss": 1.9779, "step": 21187 }, { "epoch": 0.52, "learning_rate": 9.765312197508323e-06, "loss": 2.0378, "step": 21188 }, { "epoch": 0.52, "learning_rate": 9.76451408701352e-06, "loss": 1.9222, "step": 21189 }, { "epoch": 0.52, "learning_rate": 9.763715978019547e-06, "loss": 1.9425, "step": 21190 }, { "epoch": 0.52, "learning_rate": 9.762917870531494e-06, "loss": 2.0305, "step": 21191 }, { "epoch": 0.52, "learning_rate": 9.762119764554437e-06, "loss": 2.0847, "step": 21192 }, { "epoch": 0.52, "learning_rate": 9.761321660093472e-06, "loss": 2.036, "step": 21193 }, { "epoch": 0.52, "learning_rate": 9.760523557153677e-06, "loss": 2.1718, "step": 21194 }, { "epoch": 0.52, "learning_rate": 9.759725455740144e-06, "loss": 2.0297, "step": 21195 }, { "epoch": 0.52, "learning_rate": 9.758927355857965e-06, "loss": 2.2272, "step": 21196 }, { "epoch": 0.52, "learning_rate": 9.758129257512214e-06, "loss": 1.9438, "step": 21197 }, { "epoch": 0.52, "learning_rate": 9.757331160707988e-06, "loss": 2.0358, "step": 21198 }, { "epoch": 0.52, "learning_rate": 9.756533065450367e-06, "loss": 2.1446, "step": 21199 }, { "epoch": 0.52, "learning_rate": 9.755734971744439e-06, "loss": 2.0481, "step": 21200 }, { "epoch": 0.52, "learning_rate": 9.754936879595296e-06, "loss": 2.0531, "step": 21201 }, { "epoch": 0.52, "learning_rate": 9.754138789008017e-06, "loss": 1.8944, "step": 21202 }, { "epoch": 0.52, "learning_rate": 9.753340699987695e-06, "loss": 2.001, "step": 21203 }, { "epoch": 0.52, "learning_rate": 9.75254261253941e-06, "loss": 2.2086, "step": 21204 }, { "epoch": 0.52, "learning_rate": 9.75174452666825e-06, "loss": 2.0052, "step": 21205 }, { "epoch": 0.52, "learning_rate": 9.750946442379307e-06, "loss": 2.0549, "step": 21206 }, { "epoch": 0.52, "learning_rate": 9.750148359677658e-06, "loss": 1.8862, "step": 21207 }, { "epoch": 0.52, "learning_rate": 9.749350278568403e-06, "loss": 2.144, "step": 21208 }, { "epoch": 0.52, "learning_rate": 9.748552199056614e-06, "loss": 2.0023, "step": 21209 }, { "epoch": 0.52, "learning_rate": 9.747754121147385e-06, "loss": 2.0725, "step": 21210 }, { "epoch": 0.52, "learning_rate": 9.746956044845805e-06, "loss": 1.9735, "step": 21211 }, { "epoch": 0.52, "learning_rate": 9.746157970156952e-06, "loss": 2.0964, "step": 21212 }, { "epoch": 0.52, "learning_rate": 9.74535989708592e-06, "loss": 1.896, "step": 21213 }, { "epoch": 0.52, "learning_rate": 9.744561825637791e-06, "loss": 2.021, "step": 21214 }, { "epoch": 0.52, "learning_rate": 9.743763755817652e-06, "loss": 2.1288, "step": 21215 }, { "epoch": 0.52, "learning_rate": 9.74296568763059e-06, "loss": 2.0462, "step": 21216 }, { "epoch": 0.52, "learning_rate": 9.74216762108169e-06, "loss": 2.093, "step": 21217 }, { "epoch": 0.52, "learning_rate": 9.741369556176046e-06, "loss": 2.0958, "step": 21218 }, { "epoch": 0.52, "learning_rate": 9.740571492918732e-06, "loss": 1.9553, "step": 21219 }, { "epoch": 0.52, "learning_rate": 9.739773431314842e-06, "loss": 1.9488, "step": 21220 }, { "epoch": 0.52, "learning_rate": 9.738975371369464e-06, "loss": 2.0317, "step": 21221 }, { "epoch": 0.52, "learning_rate": 9.738177313087675e-06, "loss": 2.2083, "step": 21222 }, { "epoch": 0.52, "learning_rate": 9.73737925647457e-06, "loss": 2.0569, "step": 21223 }, { "epoch": 0.52, "learning_rate": 9.736581201535234e-06, "loss": 2.0062, "step": 21224 }, { "epoch": 0.52, "learning_rate": 9.735783148274748e-06, "loss": 2.0771, "step": 21225 }, { "epoch": 0.52, "learning_rate": 9.73498509669821e-06, "loss": 2.0453, "step": 21226 }, { "epoch": 0.52, "learning_rate": 9.734187046810692e-06, "loss": 1.9476, "step": 21227 }, { "epoch": 0.52, "learning_rate": 9.733388998617285e-06, "loss": 2.0471, "step": 21228 }, { "epoch": 0.52, "learning_rate": 9.732590952123083e-06, "loss": 1.9408, "step": 21229 }, { "epoch": 0.52, "learning_rate": 9.731792907333162e-06, "loss": 2.0897, "step": 21230 }, { "epoch": 0.52, "learning_rate": 9.730994864252613e-06, "loss": 2.1154, "step": 21231 }, { "epoch": 0.52, "learning_rate": 9.730196822886523e-06, "loss": 1.9826, "step": 21232 }, { "epoch": 0.52, "learning_rate": 9.729398783239972e-06, "loss": 2.1851, "step": 21233 }, { "epoch": 0.52, "learning_rate": 9.728600745318055e-06, "loss": 2.0569, "step": 21234 }, { "epoch": 0.52, "learning_rate": 9.727802709125852e-06, "loss": 2.1157, "step": 21235 }, { "epoch": 0.52, "learning_rate": 9.727004674668454e-06, "loss": 1.9921, "step": 21236 }, { "epoch": 0.52, "learning_rate": 9.726206641950941e-06, "loss": 1.8804, "step": 21237 }, { "epoch": 0.52, "learning_rate": 9.725408610978402e-06, "loss": 1.8328, "step": 21238 }, { "epoch": 0.52, "learning_rate": 9.724610581755927e-06, "loss": 2.0519, "step": 21239 }, { "epoch": 0.52, "learning_rate": 9.723812554288594e-06, "loss": 1.9595, "step": 21240 }, { "epoch": 0.52, "learning_rate": 9.723014528581497e-06, "loss": 2.0026, "step": 21241 }, { "epoch": 0.52, "learning_rate": 9.722216504639716e-06, "loss": 2.2119, "step": 21242 }, { "epoch": 0.52, "learning_rate": 9.72141848246834e-06, "loss": 2.2142, "step": 21243 }, { "epoch": 0.52, "learning_rate": 9.720620462072457e-06, "loss": 2.1215, "step": 21244 }, { "epoch": 0.52, "learning_rate": 9.719822443457149e-06, "loss": 2.0071, "step": 21245 }, { "epoch": 0.52, "learning_rate": 9.719024426627507e-06, "loss": 2.0141, "step": 21246 }, { "epoch": 0.52, "learning_rate": 9.71822641158861e-06, "loss": 2.1382, "step": 21247 }, { "epoch": 0.52, "learning_rate": 9.717428398345549e-06, "loss": 1.9411, "step": 21248 }, { "epoch": 0.52, "learning_rate": 9.716630386903412e-06, "loss": 2.0, "step": 21249 }, { "epoch": 0.52, "learning_rate": 9.715832377267277e-06, "loss": 1.8068, "step": 21250 }, { "epoch": 0.52, "learning_rate": 9.715034369442238e-06, "loss": 1.9373, "step": 21251 }, { "epoch": 0.52, "learning_rate": 9.714236363433375e-06, "loss": 1.9689, "step": 21252 }, { "epoch": 0.52, "learning_rate": 9.713438359245777e-06, "loss": 2.0479, "step": 21253 }, { "epoch": 0.52, "learning_rate": 9.712640356884536e-06, "loss": 2.0422, "step": 21254 }, { "epoch": 0.52, "learning_rate": 9.711842356354725e-06, "loss": 1.8119, "step": 21255 }, { "epoch": 0.52, "learning_rate": 9.711044357661442e-06, "loss": 1.9196, "step": 21256 }, { "epoch": 0.52, "learning_rate": 9.710246360809762e-06, "loss": 1.9688, "step": 21257 }, { "epoch": 0.52, "learning_rate": 9.709448365804777e-06, "loss": 2.0945, "step": 21258 }, { "epoch": 0.52, "learning_rate": 9.708650372651573e-06, "loss": 2.0006, "step": 21259 }, { "epoch": 0.52, "learning_rate": 9.707852381355235e-06, "loss": 1.9312, "step": 21260 }, { "epoch": 0.52, "learning_rate": 9.707054391920852e-06, "loss": 1.9858, "step": 21261 }, { "epoch": 0.52, "learning_rate": 9.706256404353505e-06, "loss": 2.1089, "step": 21262 }, { "epoch": 0.52, "learning_rate": 9.70545841865828e-06, "loss": 2.096, "step": 21263 }, { "epoch": 0.52, "learning_rate": 9.704660434840267e-06, "loss": 1.9476, "step": 21264 }, { "epoch": 0.52, "learning_rate": 9.703862452904547e-06, "loss": 1.9396, "step": 21265 }, { "epoch": 0.52, "learning_rate": 9.70306447285621e-06, "loss": 1.8831, "step": 21266 }, { "epoch": 0.52, "learning_rate": 9.70226649470034e-06, "loss": 1.885, "step": 21267 }, { "epoch": 0.52, "learning_rate": 9.70146851844202e-06, "loss": 2.1618, "step": 21268 }, { "epoch": 0.52, "learning_rate": 9.700670544086341e-06, "loss": 1.8222, "step": 21269 }, { "epoch": 0.52, "learning_rate": 9.699872571638384e-06, "loss": 2.0098, "step": 21270 }, { "epoch": 0.52, "learning_rate": 9.69907460110324e-06, "loss": 2.0543, "step": 21271 }, { "epoch": 0.52, "learning_rate": 9.69827663248599e-06, "loss": 2.0283, "step": 21272 }, { "epoch": 0.52, "learning_rate": 9.697478665791718e-06, "loss": 2.0566, "step": 21273 }, { "epoch": 0.52, "learning_rate": 9.696680701025521e-06, "loss": 2.2054, "step": 21274 }, { "epoch": 0.52, "learning_rate": 9.69588273819247e-06, "loss": 1.9944, "step": 21275 }, { "epoch": 0.52, "learning_rate": 9.695084777297659e-06, "loss": 2.0331, "step": 21276 }, { "epoch": 0.52, "learning_rate": 9.694286818346172e-06, "loss": 2.1208, "step": 21277 }, { "epoch": 0.52, "learning_rate": 9.693488861343093e-06, "loss": 1.9673, "step": 21278 }, { "epoch": 0.52, "learning_rate": 9.692690906293514e-06, "loss": 1.9551, "step": 21279 }, { "epoch": 0.52, "learning_rate": 9.691892953202513e-06, "loss": 2.0335, "step": 21280 }, { "epoch": 0.52, "learning_rate": 9.691095002075179e-06, "loss": 2.0259, "step": 21281 }, { "epoch": 0.52, "learning_rate": 9.690297052916599e-06, "loss": 2.1696, "step": 21282 }, { "epoch": 0.52, "learning_rate": 9.689499105731855e-06, "loss": 1.8787, "step": 21283 }, { "epoch": 0.52, "learning_rate": 9.688701160526035e-06, "loss": 2.248, "step": 21284 }, { "epoch": 0.52, "learning_rate": 9.687903217304225e-06, "loss": 2.1223, "step": 21285 }, { "epoch": 0.52, "learning_rate": 9.687105276071505e-06, "loss": 1.9295, "step": 21286 }, { "epoch": 0.52, "learning_rate": 9.68630733683297e-06, "loss": 1.9255, "step": 21287 }, { "epoch": 0.52, "learning_rate": 9.685509399593697e-06, "loss": 2.0186, "step": 21288 }, { "epoch": 0.52, "learning_rate": 9.684711464358778e-06, "loss": 2.1505, "step": 21289 }, { "epoch": 0.52, "learning_rate": 9.683913531133295e-06, "loss": 2.0589, "step": 21290 }, { "epoch": 0.52, "learning_rate": 9.683115599922332e-06, "loss": 1.8205, "step": 21291 }, { "epoch": 0.52, "learning_rate": 9.682317670730981e-06, "loss": 2.0619, "step": 21292 }, { "epoch": 0.52, "learning_rate": 9.681519743564318e-06, "loss": 2.1117, "step": 21293 }, { "epoch": 0.52, "learning_rate": 9.680721818427436e-06, "loss": 2.0942, "step": 21294 }, { "epoch": 0.52, "learning_rate": 9.679923895325416e-06, "loss": 2.0254, "step": 21295 }, { "epoch": 0.52, "learning_rate": 9.679125974263344e-06, "loss": 1.9, "step": 21296 }, { "epoch": 0.52, "learning_rate": 9.678328055246313e-06, "loss": 2.0561, "step": 21297 }, { "epoch": 0.52, "learning_rate": 9.677530138279398e-06, "loss": 1.8558, "step": 21298 }, { "epoch": 0.52, "learning_rate": 9.676732223367692e-06, "loss": 2.0531, "step": 21299 }, { "epoch": 0.53, "learning_rate": 9.675934310516273e-06, "loss": 2.1081, "step": 21300 }, { "epoch": 0.53, "learning_rate": 9.67513639973023e-06, "loss": 1.9612, "step": 21301 }, { "epoch": 0.53, "learning_rate": 9.67433849101465e-06, "loss": 2.115, "step": 21302 }, { "epoch": 0.53, "learning_rate": 9.673540584374616e-06, "loss": 2.2114, "step": 21303 }, { "epoch": 0.53, "learning_rate": 9.672742679815218e-06, "loss": 2.0287, "step": 21304 }, { "epoch": 0.53, "learning_rate": 9.671944777341533e-06, "loss": 2.0215, "step": 21305 }, { "epoch": 0.53, "learning_rate": 9.67114687695865e-06, "loss": 2.0617, "step": 21306 }, { "epoch": 0.53, "learning_rate": 9.670348978671662e-06, "loss": 1.9037, "step": 21307 }, { "epoch": 0.53, "learning_rate": 9.66955108248564e-06, "loss": 1.9834, "step": 21308 }, { "epoch": 0.53, "learning_rate": 9.668753188405684e-06, "loss": 2.0744, "step": 21309 }, { "epoch": 0.53, "learning_rate": 9.667955296436864e-06, "loss": 1.8539, "step": 21310 }, { "epoch": 0.53, "learning_rate": 9.667157406584275e-06, "loss": 2.0448, "step": 21311 }, { "epoch": 0.53, "learning_rate": 9.666359518853004e-06, "loss": 2.0011, "step": 21312 }, { "epoch": 0.53, "learning_rate": 9.665561633248128e-06, "loss": 2.01, "step": 21313 }, { "epoch": 0.53, "learning_rate": 9.664763749774742e-06, "loss": 2.0605, "step": 21314 }, { "epoch": 0.53, "learning_rate": 9.66396586843792e-06, "loss": 2.0815, "step": 21315 }, { "epoch": 0.53, "learning_rate": 9.663167989242756e-06, "loss": 1.9714, "step": 21316 }, { "epoch": 0.53, "learning_rate": 9.662370112194334e-06, "loss": 1.9812, "step": 21317 }, { "epoch": 0.53, "learning_rate": 9.661572237297735e-06, "loss": 1.8119, "step": 21318 }, { "epoch": 0.53, "learning_rate": 9.660774364558047e-06, "loss": 2.0553, "step": 21319 }, { "epoch": 0.53, "learning_rate": 9.659976493980353e-06, "loss": 2.1549, "step": 21320 }, { "epoch": 0.53, "learning_rate": 9.659178625569743e-06, "loss": 2.1553, "step": 21321 }, { "epoch": 0.53, "learning_rate": 9.658380759331297e-06, "loss": 2.0168, "step": 21322 }, { "epoch": 0.53, "learning_rate": 9.6575828952701e-06, "loss": 2.0264, "step": 21323 }, { "epoch": 0.53, "learning_rate": 9.656785033391243e-06, "loss": 2.1179, "step": 21324 }, { "epoch": 0.53, "learning_rate": 9.655987173699803e-06, "loss": 1.9242, "step": 21325 }, { "epoch": 0.53, "learning_rate": 9.655189316200869e-06, "loss": 2.2399, "step": 21326 }, { "epoch": 0.53, "learning_rate": 9.65439146089953e-06, "loss": 2.0007, "step": 21327 }, { "epoch": 0.53, "learning_rate": 9.65359360780086e-06, "loss": 2.0017, "step": 21328 }, { "epoch": 0.53, "learning_rate": 9.652795756909956e-06, "loss": 1.9791, "step": 21329 }, { "epoch": 0.53, "learning_rate": 9.651997908231894e-06, "loss": 1.9728, "step": 21330 }, { "epoch": 0.53, "learning_rate": 9.651200061771765e-06, "loss": 1.8339, "step": 21331 }, { "epoch": 0.53, "learning_rate": 9.650402217534655e-06, "loss": 1.9969, "step": 21332 }, { "epoch": 0.53, "learning_rate": 9.649604375525641e-06, "loss": 1.9925, "step": 21333 }, { "epoch": 0.53, "learning_rate": 9.648806535749816e-06, "loss": 1.9428, "step": 21334 }, { "epoch": 0.53, "learning_rate": 9.648008698212259e-06, "loss": 2.1647, "step": 21335 }, { "epoch": 0.53, "learning_rate": 9.647210862918054e-06, "loss": 2.1281, "step": 21336 }, { "epoch": 0.53, "learning_rate": 9.646413029872294e-06, "loss": 1.8285, "step": 21337 }, { "epoch": 0.53, "learning_rate": 9.645615199080057e-06, "loss": 2.0941, "step": 21338 }, { "epoch": 0.53, "learning_rate": 9.64481737054643e-06, "loss": 1.9695, "step": 21339 }, { "epoch": 0.53, "learning_rate": 9.644019544276499e-06, "loss": 1.9387, "step": 21340 }, { "epoch": 0.53, "learning_rate": 9.643221720275346e-06, "loss": 1.9766, "step": 21341 }, { "epoch": 0.53, "learning_rate": 9.64242389854806e-06, "loss": 2.1034, "step": 21342 }, { "epoch": 0.53, "learning_rate": 9.64162607909972e-06, "loss": 2.1177, "step": 21343 }, { "epoch": 0.53, "learning_rate": 9.640828261935412e-06, "loss": 2.1052, "step": 21344 }, { "epoch": 0.53, "learning_rate": 9.640030447060228e-06, "loss": 2.1619, "step": 21345 }, { "epoch": 0.53, "learning_rate": 9.639232634479243e-06, "loss": 1.9971, "step": 21346 }, { "epoch": 0.53, "learning_rate": 9.63843482419755e-06, "loss": 1.9793, "step": 21347 }, { "epoch": 0.53, "learning_rate": 9.637637016220225e-06, "loss": 2.2446, "step": 21348 }, { "epoch": 0.53, "learning_rate": 9.636839210552357e-06, "loss": 2.1005, "step": 21349 }, { "epoch": 0.53, "learning_rate": 9.636041407199037e-06, "loss": 1.8459, "step": 21350 }, { "epoch": 0.53, "learning_rate": 9.635243606165339e-06, "loss": 2.0214, "step": 21351 }, { "epoch": 0.53, "learning_rate": 9.634445807456356e-06, "loss": 1.9399, "step": 21352 }, { "epoch": 0.53, "learning_rate": 9.633648011077167e-06, "loss": 2.1073, "step": 21353 }, { "epoch": 0.53, "learning_rate": 9.632850217032855e-06, "loss": 1.9674, "step": 21354 }, { "epoch": 0.53, "learning_rate": 9.632052425328512e-06, "loss": 2.2376, "step": 21355 }, { "epoch": 0.53, "learning_rate": 9.63125463596922e-06, "loss": 2.0682, "step": 21356 }, { "epoch": 0.53, "learning_rate": 9.63045684896006e-06, "loss": 2.0808, "step": 21357 }, { "epoch": 0.53, "learning_rate": 9.62965906430612e-06, "loss": 1.9657, "step": 21358 }, { "epoch": 0.53, "learning_rate": 9.628861282012481e-06, "loss": 2.0596, "step": 21359 }, { "epoch": 0.53, "learning_rate": 9.628063502084235e-06, "loss": 1.8684, "step": 21360 }, { "epoch": 0.53, "learning_rate": 9.627265724526456e-06, "loss": 2.3165, "step": 21361 }, { "epoch": 0.53, "learning_rate": 9.62646794934424e-06, "loss": 1.9126, "step": 21362 }, { "epoch": 0.53, "learning_rate": 9.62567017654266e-06, "loss": 2.0391, "step": 21363 }, { "epoch": 0.53, "learning_rate": 9.624872406126808e-06, "loss": 1.9506, "step": 21364 }, { "epoch": 0.53, "learning_rate": 9.624074638101766e-06, "loss": 2.1311, "step": 21365 }, { "epoch": 0.53, "learning_rate": 9.623276872472617e-06, "loss": 1.9147, "step": 21366 }, { "epoch": 0.53, "learning_rate": 9.622479109244454e-06, "loss": 2.2963, "step": 21367 }, { "epoch": 0.53, "learning_rate": 9.621681348422349e-06, "loss": 2.1151, "step": 21368 }, { "epoch": 0.53, "learning_rate": 9.620883590011392e-06, "loss": 1.9425, "step": 21369 }, { "epoch": 0.53, "learning_rate": 9.620085834016672e-06, "loss": 1.7539, "step": 21370 }, { "epoch": 0.53, "learning_rate": 9.619288080443263e-06, "loss": 2.0083, "step": 21371 }, { "epoch": 0.53, "learning_rate": 9.618490329296258e-06, "loss": 2.0503, "step": 21372 }, { "epoch": 0.53, "learning_rate": 9.617692580580737e-06, "loss": 1.9018, "step": 21373 }, { "epoch": 0.53, "learning_rate": 9.616894834301788e-06, "loss": 2.1313, "step": 21374 }, { "epoch": 0.53, "learning_rate": 9.616097090464492e-06, "loss": 1.9921, "step": 21375 }, { "epoch": 0.53, "learning_rate": 9.615299349073933e-06, "loss": 1.8308, "step": 21376 }, { "epoch": 0.53, "learning_rate": 9.614501610135201e-06, "loss": 1.9431, "step": 21377 }, { "epoch": 0.53, "learning_rate": 9.61370387365337e-06, "loss": 2.1871, "step": 21378 }, { "epoch": 0.53, "learning_rate": 9.612906139633533e-06, "loss": 2.0296, "step": 21379 }, { "epoch": 0.53, "learning_rate": 9.612108408080773e-06, "loss": 2.1372, "step": 21380 }, { "epoch": 0.53, "learning_rate": 9.61131067900017e-06, "loss": 2.0139, "step": 21381 }, { "epoch": 0.53, "learning_rate": 9.610512952396812e-06, "loss": 1.8699, "step": 21382 }, { "epoch": 0.53, "learning_rate": 9.60971522827578e-06, "loss": 1.909, "step": 21383 }, { "epoch": 0.53, "learning_rate": 9.60891750664216e-06, "loss": 2.0703, "step": 21384 }, { "epoch": 0.53, "learning_rate": 9.60811978750104e-06, "loss": 2.0087, "step": 21385 }, { "epoch": 0.53, "learning_rate": 9.607322070857496e-06, "loss": 2.0853, "step": 21386 }, { "epoch": 0.53, "learning_rate": 9.606524356716622e-06, "loss": 2.0889, "step": 21387 }, { "epoch": 0.53, "learning_rate": 9.60572664508349e-06, "loss": 2.0524, "step": 21388 }, { "epoch": 0.53, "learning_rate": 9.604928935963192e-06, "loss": 2.0186, "step": 21389 }, { "epoch": 0.53, "learning_rate": 9.604131229360813e-06, "loss": 1.9043, "step": 21390 }, { "epoch": 0.53, "learning_rate": 9.603333525281433e-06, "loss": 2.0487, "step": 21391 }, { "epoch": 0.53, "learning_rate": 9.602535823730142e-06, "loss": 2.045, "step": 21392 }, { "epoch": 0.53, "learning_rate": 9.601738124712014e-06, "loss": 1.8767, "step": 21393 }, { "epoch": 0.53, "learning_rate": 9.600940428232139e-06, "loss": 1.8766, "step": 21394 }, { "epoch": 0.53, "learning_rate": 9.600142734295606e-06, "loss": 2.0151, "step": 21395 }, { "epoch": 0.53, "learning_rate": 9.599345042907489e-06, "loss": 2.2015, "step": 21396 }, { "epoch": 0.53, "learning_rate": 9.59854735407288e-06, "loss": 2.1443, "step": 21397 }, { "epoch": 0.53, "learning_rate": 9.597749667796856e-06, "loss": 2.0263, "step": 21398 }, { "epoch": 0.53, "learning_rate": 9.596951984084503e-06, "loss": 2.0802, "step": 21399 }, { "epoch": 0.53, "learning_rate": 9.59615430294091e-06, "loss": 1.8134, "step": 21400 }, { "epoch": 0.53, "learning_rate": 9.595356624371156e-06, "loss": 2.0266, "step": 21401 }, { "epoch": 0.53, "learning_rate": 9.594558948380326e-06, "loss": 1.9231, "step": 21402 }, { "epoch": 0.53, "learning_rate": 9.593761274973506e-06, "loss": 1.8492, "step": 21403 }, { "epoch": 0.53, "learning_rate": 9.592963604155775e-06, "loss": 2.2097, "step": 21404 }, { "epoch": 0.53, "learning_rate": 9.592165935932223e-06, "loss": 1.8212, "step": 21405 }, { "epoch": 0.53, "learning_rate": 9.591368270307927e-06, "loss": 2.1367, "step": 21406 }, { "epoch": 0.53, "learning_rate": 9.590570607287975e-06, "loss": 1.9744, "step": 21407 }, { "epoch": 0.53, "learning_rate": 9.58977294687745e-06, "loss": 1.9379, "step": 21408 }, { "epoch": 0.53, "learning_rate": 9.588975289081434e-06, "loss": 2.0069, "step": 21409 }, { "epoch": 0.53, "learning_rate": 9.588177633905019e-06, "loss": 2.1127, "step": 21410 }, { "epoch": 0.53, "learning_rate": 9.587379981353275e-06, "loss": 2.0412, "step": 21411 }, { "epoch": 0.53, "learning_rate": 9.586582331431295e-06, "loss": 1.8227, "step": 21412 }, { "epoch": 0.53, "learning_rate": 9.585784684144163e-06, "loss": 2.0105, "step": 21413 }, { "epoch": 0.53, "learning_rate": 9.584987039496958e-06, "loss": 2.0752, "step": 21414 }, { "epoch": 0.53, "learning_rate": 9.584189397494768e-06, "loss": 2.2979, "step": 21415 }, { "epoch": 0.53, "learning_rate": 9.583391758142671e-06, "loss": 2.0394, "step": 21416 }, { "epoch": 0.53, "learning_rate": 9.582594121445755e-06, "loss": 2.0415, "step": 21417 }, { "epoch": 0.53, "learning_rate": 9.581796487409103e-06, "loss": 1.9546, "step": 21418 }, { "epoch": 0.53, "learning_rate": 9.580998856037798e-06, "loss": 1.8607, "step": 21419 }, { "epoch": 0.53, "learning_rate": 9.580201227336927e-06, "loss": 2.068, "step": 21420 }, { "epoch": 0.53, "learning_rate": 9.579403601311566e-06, "loss": 2.1351, "step": 21421 }, { "epoch": 0.53, "learning_rate": 9.578605977966804e-06, "loss": 2.0897, "step": 21422 }, { "epoch": 0.53, "learning_rate": 9.577808357307727e-06, "loss": 2.0225, "step": 21423 }, { "epoch": 0.53, "learning_rate": 9.57701073933941e-06, "loss": 2.084, "step": 21424 }, { "epoch": 0.53, "learning_rate": 9.576213124066946e-06, "loss": 2.0849, "step": 21425 }, { "epoch": 0.53, "learning_rate": 9.575415511495408e-06, "loss": 1.9502, "step": 21426 }, { "epoch": 0.53, "learning_rate": 9.574617901629889e-06, "loss": 1.9801, "step": 21427 }, { "epoch": 0.53, "learning_rate": 9.57382029447547e-06, "loss": 1.9695, "step": 21428 }, { "epoch": 0.53, "learning_rate": 9.57302269003723e-06, "loss": 1.9261, "step": 21429 }, { "epoch": 0.53, "learning_rate": 9.57222508832026e-06, "loss": 2.032, "step": 21430 }, { "epoch": 0.53, "learning_rate": 9.571427489329636e-06, "loss": 1.9845, "step": 21431 }, { "epoch": 0.53, "learning_rate": 9.570629893070442e-06, "loss": 2.0176, "step": 21432 }, { "epoch": 0.53, "learning_rate": 9.56983229954777e-06, "loss": 1.9474, "step": 21433 }, { "epoch": 0.53, "learning_rate": 9.56903470876669e-06, "loss": 2.093, "step": 21434 }, { "epoch": 0.53, "learning_rate": 9.568237120732297e-06, "loss": 2.0371, "step": 21435 }, { "epoch": 0.53, "learning_rate": 9.567439535449664e-06, "loss": 2.0682, "step": 21436 }, { "epoch": 0.53, "learning_rate": 9.566641952923883e-06, "loss": 2.0232, "step": 21437 }, { "epoch": 0.53, "learning_rate": 9.565844373160037e-06, "loss": 1.8414, "step": 21438 }, { "epoch": 0.53, "learning_rate": 9.565046796163203e-06, "loss": 1.9726, "step": 21439 }, { "epoch": 0.53, "learning_rate": 9.564249221938471e-06, "loss": 1.9873, "step": 21440 }, { "epoch": 0.53, "learning_rate": 9.563451650490917e-06, "loss": 2.2353, "step": 21441 }, { "epoch": 0.53, "learning_rate": 9.562654081825627e-06, "loss": 2.0483, "step": 21442 }, { "epoch": 0.53, "learning_rate": 9.561856515947689e-06, "loss": 2.1544, "step": 21443 }, { "epoch": 0.53, "learning_rate": 9.561058952862178e-06, "loss": 2.027, "step": 21444 }, { "epoch": 0.53, "learning_rate": 9.560261392574188e-06, "loss": 1.9208, "step": 21445 }, { "epoch": 0.53, "learning_rate": 9.559463835088788e-06, "loss": 2.1832, "step": 21446 }, { "epoch": 0.53, "learning_rate": 9.558666280411071e-06, "loss": 2.015, "step": 21447 }, { "epoch": 0.53, "learning_rate": 9.55786872854612e-06, "loss": 1.9088, "step": 21448 }, { "epoch": 0.53, "learning_rate": 9.557071179499013e-06, "loss": 1.9943, "step": 21449 }, { "epoch": 0.53, "learning_rate": 9.556273633274837e-06, "loss": 1.9682, "step": 21450 }, { "epoch": 0.53, "learning_rate": 9.555476089878674e-06, "loss": 2.0357, "step": 21451 }, { "epoch": 0.53, "learning_rate": 9.554678549315605e-06, "loss": 1.9743, "step": 21452 }, { "epoch": 0.53, "learning_rate": 9.553881011590717e-06, "loss": 2.0365, "step": 21453 }, { "epoch": 0.53, "learning_rate": 9.553083476709088e-06, "loss": 2.22, "step": 21454 }, { "epoch": 0.53, "learning_rate": 9.552285944675808e-06, "loss": 2.0301, "step": 21455 }, { "epoch": 0.53, "learning_rate": 9.551488415495951e-06, "loss": 2.0986, "step": 21456 }, { "epoch": 0.53, "learning_rate": 9.550690889174604e-06, "loss": 1.7889, "step": 21457 }, { "epoch": 0.53, "learning_rate": 9.549893365716855e-06, "loss": 1.9994, "step": 21458 }, { "epoch": 0.53, "learning_rate": 9.54909584512778e-06, "loss": 2.016, "step": 21459 }, { "epoch": 0.53, "learning_rate": 9.548298327412462e-06, "loss": 1.9755, "step": 21460 }, { "epoch": 0.53, "learning_rate": 9.547500812575989e-06, "loss": 2.0546, "step": 21461 }, { "epoch": 0.53, "learning_rate": 9.546703300623438e-06, "loss": 2.064, "step": 21462 }, { "epoch": 0.53, "learning_rate": 9.545905791559899e-06, "loss": 1.8961, "step": 21463 }, { "epoch": 0.53, "learning_rate": 9.545108285390446e-06, "loss": 2.0906, "step": 21464 }, { "epoch": 0.53, "learning_rate": 9.544310782120167e-06, "loss": 1.8067, "step": 21465 }, { "epoch": 0.53, "learning_rate": 9.543513281754148e-06, "loss": 2.02, "step": 21466 }, { "epoch": 0.53, "learning_rate": 9.542715784297463e-06, "loss": 2.2431, "step": 21467 }, { "epoch": 0.53, "learning_rate": 9.541918289755203e-06, "loss": 2.1118, "step": 21468 }, { "epoch": 0.53, "learning_rate": 9.541120798132443e-06, "loss": 2.1016, "step": 21469 }, { "epoch": 0.53, "learning_rate": 9.54032330943427e-06, "loss": 2.1866, "step": 21470 }, { "epoch": 0.53, "learning_rate": 9.539525823665769e-06, "loss": 1.993, "step": 21471 }, { "epoch": 0.53, "learning_rate": 9.538728340832018e-06, "loss": 1.97, "step": 21472 }, { "epoch": 0.53, "learning_rate": 9.537930860938105e-06, "loss": 2.0539, "step": 21473 }, { "epoch": 0.53, "learning_rate": 9.537133383989104e-06, "loss": 2.0661, "step": 21474 }, { "epoch": 0.53, "learning_rate": 9.536335909990105e-06, "loss": 2.1344, "step": 21475 }, { "epoch": 0.53, "learning_rate": 9.535538438946191e-06, "loss": 2.0529, "step": 21476 }, { "epoch": 0.53, "learning_rate": 9.534740970862438e-06, "loss": 1.9683, "step": 21477 }, { "epoch": 0.53, "learning_rate": 9.533943505743936e-06, "loss": 1.9368, "step": 21478 }, { "epoch": 0.53, "learning_rate": 9.53314604359576e-06, "loss": 2.0411, "step": 21479 }, { "epoch": 0.53, "learning_rate": 9.532348584422997e-06, "loss": 1.9386, "step": 21480 }, { "epoch": 0.53, "learning_rate": 9.531551128230732e-06, "loss": 1.9799, "step": 21481 }, { "epoch": 0.53, "learning_rate": 9.530753675024042e-06, "loss": 1.9216, "step": 21482 }, { "epoch": 0.53, "learning_rate": 9.529956224808016e-06, "loss": 1.9396, "step": 21483 }, { "epoch": 0.53, "learning_rate": 9.529158777587727e-06, "loss": 2.201, "step": 21484 }, { "epoch": 0.53, "learning_rate": 9.528361333368261e-06, "loss": 2.0775, "step": 21485 }, { "epoch": 0.53, "learning_rate": 9.527563892154708e-06, "loss": 2.1569, "step": 21486 }, { "epoch": 0.53, "learning_rate": 9.526766453952141e-06, "loss": 2.0875, "step": 21487 }, { "epoch": 0.53, "learning_rate": 9.525969018765646e-06, "loss": 2.0187, "step": 21488 }, { "epoch": 0.53, "learning_rate": 9.525171586600303e-06, "loss": 1.963, "step": 21489 }, { "epoch": 0.53, "learning_rate": 9.524374157461198e-06, "loss": 2.125, "step": 21490 }, { "epoch": 0.53, "learning_rate": 9.523576731353413e-06, "loss": 1.8756, "step": 21491 }, { "epoch": 0.53, "learning_rate": 9.522779308282026e-06, "loss": 1.8714, "step": 21492 }, { "epoch": 0.53, "learning_rate": 9.521981888252126e-06, "loss": 2.0121, "step": 21493 }, { "epoch": 0.53, "learning_rate": 9.521184471268787e-06, "loss": 2.0334, "step": 21494 }, { "epoch": 0.53, "learning_rate": 9.520387057337096e-06, "loss": 2.0753, "step": 21495 }, { "epoch": 0.53, "learning_rate": 9.519589646462136e-06, "loss": 1.9556, "step": 21496 }, { "epoch": 0.53, "learning_rate": 9.518792238648986e-06, "loss": 2.1251, "step": 21497 }, { "epoch": 0.53, "learning_rate": 9.517994833902733e-06, "loss": 1.9958, "step": 21498 }, { "epoch": 0.53, "learning_rate": 9.517197432228453e-06, "loss": 2.0385, "step": 21499 }, { "epoch": 0.53, "learning_rate": 9.51640003363123e-06, "loss": 2.0488, "step": 21500 }, { "epoch": 0.53, "learning_rate": 9.515602638116153e-06, "loss": 1.9386, "step": 21501 }, { "epoch": 0.53, "learning_rate": 9.514805245688294e-06, "loss": 2.0361, "step": 21502 }, { "epoch": 0.53, "learning_rate": 9.51400785635274e-06, "loss": 1.9423, "step": 21503 }, { "epoch": 0.53, "learning_rate": 9.513210470114573e-06, "loss": 2.0724, "step": 21504 }, { "epoch": 0.53, "learning_rate": 9.512413086978872e-06, "loss": 1.9949, "step": 21505 }, { "epoch": 0.53, "learning_rate": 9.511615706950723e-06, "loss": 2.0289, "step": 21506 }, { "epoch": 0.53, "learning_rate": 9.510818330035205e-06, "loss": 2.1075, "step": 21507 }, { "epoch": 0.53, "learning_rate": 9.510020956237405e-06, "loss": 2.0195, "step": 21508 }, { "epoch": 0.53, "learning_rate": 9.509223585562398e-06, "loss": 2.1402, "step": 21509 }, { "epoch": 0.53, "learning_rate": 9.508426218015267e-06, "loss": 1.9976, "step": 21510 }, { "epoch": 0.53, "learning_rate": 9.507628853601102e-06, "loss": 1.9516, "step": 21511 }, { "epoch": 0.53, "learning_rate": 9.506831492324973e-06, "loss": 2.218, "step": 21512 }, { "epoch": 0.53, "learning_rate": 9.50603413419197e-06, "loss": 2.0831, "step": 21513 }, { "epoch": 0.53, "learning_rate": 9.505236779207172e-06, "loss": 1.8493, "step": 21514 }, { "epoch": 0.53, "learning_rate": 9.504439427375659e-06, "loss": 1.8542, "step": 21515 }, { "epoch": 0.53, "learning_rate": 9.50364207870252e-06, "loss": 1.9713, "step": 21516 }, { "epoch": 0.53, "learning_rate": 9.502844733192829e-06, "loss": 2.1737, "step": 21517 }, { "epoch": 0.53, "learning_rate": 9.502047390851668e-06, "loss": 2.0551, "step": 21518 }, { "epoch": 0.53, "learning_rate": 9.501250051684128e-06, "loss": 1.9476, "step": 21519 }, { "epoch": 0.53, "learning_rate": 9.500452715695278e-06, "loss": 2.1095, "step": 21520 }, { "epoch": 0.53, "learning_rate": 9.499655382890209e-06, "loss": 2.1997, "step": 21521 }, { "epoch": 0.53, "learning_rate": 9.498858053273998e-06, "loss": 2.0321, "step": 21522 }, { "epoch": 0.53, "learning_rate": 9.498060726851727e-06, "loss": 2.0693, "step": 21523 }, { "epoch": 0.53, "learning_rate": 9.49726340362848e-06, "loss": 1.8591, "step": 21524 }, { "epoch": 0.53, "learning_rate": 9.496466083609335e-06, "loss": 1.8736, "step": 21525 }, { "epoch": 0.53, "learning_rate": 9.495668766799381e-06, "loss": 2.1901, "step": 21526 }, { "epoch": 0.53, "learning_rate": 9.49487145320369e-06, "loss": 2.0122, "step": 21527 }, { "epoch": 0.53, "learning_rate": 9.494074142827346e-06, "loss": 2.0475, "step": 21528 }, { "epoch": 0.53, "learning_rate": 9.493276835675439e-06, "loss": 1.948, "step": 21529 }, { "epoch": 0.53, "learning_rate": 9.492479531753037e-06, "loss": 1.9701, "step": 21530 }, { "epoch": 0.53, "learning_rate": 9.491682231065231e-06, "loss": 1.9581, "step": 21531 }, { "epoch": 0.53, "learning_rate": 9.4908849336171e-06, "loss": 1.9177, "step": 21532 }, { "epoch": 0.53, "learning_rate": 9.490087639413724e-06, "loss": 1.9641, "step": 21533 }, { "epoch": 0.53, "learning_rate": 9.489290348460188e-06, "loss": 1.9191, "step": 21534 }, { "epoch": 0.53, "learning_rate": 9.48849306076157e-06, "loss": 2.1265, "step": 21535 }, { "epoch": 0.53, "learning_rate": 9.487695776322954e-06, "loss": 2.0226, "step": 21536 }, { "epoch": 0.53, "learning_rate": 9.486898495149417e-06, "loss": 2.1252, "step": 21537 }, { "epoch": 0.53, "learning_rate": 9.486101217246041e-06, "loss": 2.1732, "step": 21538 }, { "epoch": 0.53, "learning_rate": 9.485303942617914e-06, "loss": 1.9901, "step": 21539 }, { "epoch": 0.53, "learning_rate": 9.484506671270113e-06, "loss": 2.0454, "step": 21540 }, { "epoch": 0.53, "learning_rate": 9.483709403207717e-06, "loss": 2.0519, "step": 21541 }, { "epoch": 0.53, "learning_rate": 9.482912138435809e-06, "loss": 2.0051, "step": 21542 }, { "epoch": 0.53, "learning_rate": 9.482114876959469e-06, "loss": 2.0689, "step": 21543 }, { "epoch": 0.53, "learning_rate": 9.481317618783784e-06, "loss": 2.1389, "step": 21544 }, { "epoch": 0.53, "learning_rate": 9.480520363913826e-06, "loss": 2.0452, "step": 21545 }, { "epoch": 0.53, "learning_rate": 9.479723112354687e-06, "loss": 1.9453, "step": 21546 }, { "epoch": 0.53, "learning_rate": 9.478925864111437e-06, "loss": 1.9436, "step": 21547 }, { "epoch": 0.53, "learning_rate": 9.478128619189163e-06, "loss": 1.9612, "step": 21548 }, { "epoch": 0.53, "learning_rate": 9.477331377592947e-06, "loss": 1.8558, "step": 21549 }, { "epoch": 0.53, "learning_rate": 9.476534139327866e-06, "loss": 2.0564, "step": 21550 }, { "epoch": 0.53, "learning_rate": 9.475736904399007e-06, "loss": 2.2521, "step": 21551 }, { "epoch": 0.53, "learning_rate": 9.474939672811444e-06, "loss": 1.9297, "step": 21552 }, { "epoch": 0.53, "learning_rate": 9.474142444570263e-06, "loss": 2.0628, "step": 21553 }, { "epoch": 0.53, "learning_rate": 9.473345219680547e-06, "loss": 2.0219, "step": 21554 }, { "epoch": 0.53, "learning_rate": 9.472547998147369e-06, "loss": 1.8416, "step": 21555 }, { "epoch": 0.53, "learning_rate": 9.471750779975817e-06, "loss": 2.0104, "step": 21556 }, { "epoch": 0.53, "learning_rate": 9.47095356517097e-06, "loss": 2.102, "step": 21557 }, { "epoch": 0.53, "learning_rate": 9.470156353737905e-06, "loss": 2.0207, "step": 21558 }, { "epoch": 0.53, "learning_rate": 9.469359145681707e-06, "loss": 2.0561, "step": 21559 }, { "epoch": 0.53, "learning_rate": 9.468561941007457e-06, "loss": 2.0037, "step": 21560 }, { "epoch": 0.53, "learning_rate": 9.467764739720238e-06, "loss": 1.934, "step": 21561 }, { "epoch": 0.53, "learning_rate": 9.466967541825122e-06, "loss": 2.0346, "step": 21562 }, { "epoch": 0.53, "learning_rate": 9.466170347327199e-06, "loss": 1.9699, "step": 21563 }, { "epoch": 0.53, "learning_rate": 9.465373156231548e-06, "loss": 1.9615, "step": 21564 }, { "epoch": 0.53, "learning_rate": 9.464575968543243e-06, "loss": 1.9016, "step": 21565 }, { "epoch": 0.53, "learning_rate": 9.463778784267374e-06, "loss": 1.9916, "step": 21566 }, { "epoch": 0.53, "learning_rate": 9.462981603409015e-06, "loss": 2.1082, "step": 21567 }, { "epoch": 0.53, "learning_rate": 9.462184425973249e-06, "loss": 2.08, "step": 21568 }, { "epoch": 0.53, "learning_rate": 9.461387251965161e-06, "loss": 2.0028, "step": 21569 }, { "epoch": 0.53, "learning_rate": 9.460590081389824e-06, "loss": 2.0158, "step": 21570 }, { "epoch": 0.53, "learning_rate": 9.459792914252325e-06, "loss": 2.0804, "step": 21571 }, { "epoch": 0.53, "learning_rate": 9.45899575055774e-06, "loss": 2.1699, "step": 21572 }, { "epoch": 0.53, "learning_rate": 9.458198590311149e-06, "loss": 1.9791, "step": 21573 }, { "epoch": 0.53, "learning_rate": 9.457401433517639e-06, "loss": 1.9727, "step": 21574 }, { "epoch": 0.53, "learning_rate": 9.456604280182285e-06, "loss": 2.1591, "step": 21575 }, { "epoch": 0.53, "learning_rate": 9.455807130310167e-06, "loss": 2.0081, "step": 21576 }, { "epoch": 0.53, "learning_rate": 9.45500998390637e-06, "loss": 2.0717, "step": 21577 }, { "epoch": 0.53, "learning_rate": 9.45421284097597e-06, "loss": 2.1083, "step": 21578 }, { "epoch": 0.53, "learning_rate": 9.453415701524053e-06, "loss": 2.0818, "step": 21579 }, { "epoch": 0.53, "learning_rate": 9.452618565555692e-06, "loss": 2.0359, "step": 21580 }, { "epoch": 0.53, "learning_rate": 9.451821433075971e-06, "loss": 2.2948, "step": 21581 }, { "epoch": 0.53, "learning_rate": 9.451024304089977e-06, "loss": 2.0721, "step": 21582 }, { "epoch": 0.53, "learning_rate": 9.450227178602777e-06, "loss": 2.0427, "step": 21583 }, { "epoch": 0.53, "learning_rate": 9.449430056619462e-06, "loss": 1.9129, "step": 21584 }, { "epoch": 0.53, "learning_rate": 9.448632938145106e-06, "loss": 2.1172, "step": 21585 }, { "epoch": 0.53, "learning_rate": 9.447835823184793e-06, "loss": 2.0437, "step": 21586 }, { "epoch": 0.53, "learning_rate": 9.447038711743605e-06, "loss": 2.1652, "step": 21587 }, { "epoch": 0.53, "learning_rate": 9.446241603826617e-06, "loss": 2.0593, "step": 21588 }, { "epoch": 0.53, "learning_rate": 9.445444499438915e-06, "loss": 2.0491, "step": 21589 }, { "epoch": 0.53, "learning_rate": 9.44464739858557e-06, "loss": 1.9744, "step": 21590 }, { "epoch": 0.53, "learning_rate": 9.44385030127167e-06, "loss": 2.1201, "step": 21591 }, { "epoch": 0.53, "learning_rate": 9.443053207502294e-06, "loss": 2.0399, "step": 21592 }, { "epoch": 0.53, "learning_rate": 9.442256117282524e-06, "loss": 1.985, "step": 21593 }, { "epoch": 0.53, "learning_rate": 9.441459030617434e-06, "loss": 1.9898, "step": 21594 }, { "epoch": 0.53, "learning_rate": 9.440661947512108e-06, "loss": 2.152, "step": 21595 }, { "epoch": 0.53, "learning_rate": 9.439864867971626e-06, "loss": 1.9302, "step": 21596 }, { "epoch": 0.53, "learning_rate": 9.439067792001072e-06, "loss": 1.9959, "step": 21597 }, { "epoch": 0.53, "learning_rate": 9.438270719605516e-06, "loss": 1.9865, "step": 21598 }, { "epoch": 0.53, "learning_rate": 9.437473650790049e-06, "loss": 2.2186, "step": 21599 }, { "epoch": 0.53, "learning_rate": 9.43667658555974e-06, "loss": 1.8823, "step": 21600 }, { "epoch": 0.53, "learning_rate": 9.435879523919675e-06, "loss": 2.0303, "step": 21601 }, { "epoch": 0.53, "learning_rate": 9.435082465874938e-06, "loss": 2.1009, "step": 21602 }, { "epoch": 0.53, "learning_rate": 9.4342854114306e-06, "loss": 1.9548, "step": 21603 }, { "epoch": 0.53, "learning_rate": 9.433488360591752e-06, "loss": 2.1371, "step": 21604 }, { "epoch": 0.53, "learning_rate": 9.432691313363461e-06, "loss": 1.9349, "step": 21605 }, { "epoch": 0.53, "learning_rate": 9.431894269750814e-06, "loss": 2.0845, "step": 21606 }, { "epoch": 0.53, "learning_rate": 9.431097229758893e-06, "loss": 1.8981, "step": 21607 }, { "epoch": 0.53, "learning_rate": 9.430300193392774e-06, "loss": 1.8865, "step": 21608 }, { "epoch": 0.53, "learning_rate": 9.429503160657535e-06, "loss": 1.9985, "step": 21609 }, { "epoch": 0.53, "learning_rate": 9.42870613155826e-06, "loss": 1.8735, "step": 21610 }, { "epoch": 0.53, "learning_rate": 9.427909106100028e-06, "loss": 1.844, "step": 21611 }, { "epoch": 0.53, "learning_rate": 9.427112084287916e-06, "loss": 2.0284, "step": 21612 }, { "epoch": 0.53, "learning_rate": 9.426315066127004e-06, "loss": 2.0883, "step": 21613 }, { "epoch": 0.53, "learning_rate": 9.425518051622378e-06, "loss": 2.0626, "step": 21614 }, { "epoch": 0.53, "learning_rate": 9.42472104077911e-06, "loss": 2.1011, "step": 21615 }, { "epoch": 0.53, "learning_rate": 9.423924033602278e-06, "loss": 2.0563, "step": 21616 }, { "epoch": 0.53, "learning_rate": 9.423127030096973e-06, "loss": 1.9871, "step": 21617 }, { "epoch": 0.53, "learning_rate": 9.422330030268263e-06, "loss": 2.0865, "step": 21618 }, { "epoch": 0.53, "learning_rate": 9.421533034121235e-06, "loss": 2.0615, "step": 21619 }, { "epoch": 0.53, "learning_rate": 9.420736041660961e-06, "loss": 1.8769, "step": 21620 }, { "epoch": 0.53, "learning_rate": 9.419939052892528e-06, "loss": 1.9487, "step": 21621 }, { "epoch": 0.53, "learning_rate": 9.419142067821014e-06, "loss": 2.0814, "step": 21622 }, { "epoch": 0.53, "learning_rate": 9.418345086451494e-06, "loss": 1.9832, "step": 21623 }, { "epoch": 0.53, "learning_rate": 9.417548108789055e-06, "loss": 1.9391, "step": 21624 }, { "epoch": 0.53, "learning_rate": 9.416751134838765e-06, "loss": 2.1036, "step": 21625 }, { "epoch": 0.53, "learning_rate": 9.415954164605712e-06, "loss": 2.1326, "step": 21626 }, { "epoch": 0.53, "learning_rate": 9.415157198094975e-06, "loss": 2.2528, "step": 21627 }, { "epoch": 0.53, "learning_rate": 9.414360235311631e-06, "loss": 2.2088, "step": 21628 }, { "epoch": 0.53, "learning_rate": 9.413563276260763e-06, "loss": 1.9792, "step": 21629 }, { "epoch": 0.53, "learning_rate": 9.412766320947442e-06, "loss": 2.1386, "step": 21630 }, { "epoch": 0.53, "learning_rate": 9.411969369376756e-06, "loss": 1.9792, "step": 21631 }, { "epoch": 0.53, "learning_rate": 9.411172421553782e-06, "loss": 1.9368, "step": 21632 }, { "epoch": 0.53, "learning_rate": 9.410375477483596e-06, "loss": 2.0787, "step": 21633 }, { "epoch": 0.53, "learning_rate": 9.40957853717128e-06, "loss": 1.946, "step": 21634 }, { "epoch": 0.53, "learning_rate": 9.408781600621912e-06, "loss": 2.0259, "step": 21635 }, { "epoch": 0.53, "learning_rate": 9.407984667840571e-06, "loss": 2.1302, "step": 21636 }, { "epoch": 0.53, "learning_rate": 9.407187738832337e-06, "loss": 1.863, "step": 21637 }, { "epoch": 0.53, "learning_rate": 9.406390813602288e-06, "loss": 2.0085, "step": 21638 }, { "epoch": 0.53, "learning_rate": 9.405593892155503e-06, "loss": 1.87, "step": 21639 }, { "epoch": 0.53, "learning_rate": 9.404796974497066e-06, "loss": 2.0572, "step": 21640 }, { "epoch": 0.53, "learning_rate": 9.404000060632048e-06, "loss": 1.9171, "step": 21641 }, { "epoch": 0.53, "learning_rate": 9.403203150565536e-06, "loss": 1.8618, "step": 21642 }, { "epoch": 0.53, "learning_rate": 9.4024062443026e-06, "loss": 1.9995, "step": 21643 }, { "epoch": 0.53, "learning_rate": 9.401609341848324e-06, "loss": 1.9554, "step": 21644 }, { "epoch": 0.53, "learning_rate": 9.40081244320779e-06, "loss": 2.0543, "step": 21645 }, { "epoch": 0.53, "learning_rate": 9.400015548386072e-06, "loss": 1.8266, "step": 21646 }, { "epoch": 0.53, "learning_rate": 9.399218657388251e-06, "loss": 2.1068, "step": 21647 }, { "epoch": 0.53, "learning_rate": 9.398421770219404e-06, "loss": 2.0366, "step": 21648 }, { "epoch": 0.53, "learning_rate": 9.39762488688461e-06, "loss": 2.2151, "step": 21649 }, { "epoch": 0.53, "learning_rate": 9.396828007388954e-06, "loss": 1.8933, "step": 21650 }, { "epoch": 0.53, "learning_rate": 9.396031131737506e-06, "loss": 1.981, "step": 21651 }, { "epoch": 0.53, "learning_rate": 9.39523425993535e-06, "loss": 2.1984, "step": 21652 }, { "epoch": 0.53, "learning_rate": 9.394437391987562e-06, "loss": 2.0307, "step": 21653 }, { "epoch": 0.53, "learning_rate": 9.393640527899219e-06, "loss": 2.1196, "step": 21654 }, { "epoch": 0.53, "learning_rate": 9.392843667675406e-06, "loss": 2.0157, "step": 21655 }, { "epoch": 0.53, "learning_rate": 9.392046811321196e-06, "loss": 2.0988, "step": 21656 }, { "epoch": 0.53, "learning_rate": 9.391249958841676e-06, "loss": 2.0726, "step": 21657 }, { "epoch": 0.53, "learning_rate": 9.390453110241911e-06, "loss": 1.9823, "step": 21658 }, { "epoch": 0.53, "learning_rate": 9.38965626552699e-06, "loss": 1.9706, "step": 21659 }, { "epoch": 0.53, "learning_rate": 9.38885942470199e-06, "loss": 1.9043, "step": 21660 }, { "epoch": 0.53, "learning_rate": 9.388062587771985e-06, "loss": 2.0573, "step": 21661 }, { "epoch": 0.53, "learning_rate": 9.387265754742058e-06, "loss": 2.1068, "step": 21662 }, { "epoch": 0.53, "learning_rate": 9.386468925617284e-06, "loss": 1.9095, "step": 21663 }, { "epoch": 0.53, "learning_rate": 9.385672100402746e-06, "loss": 2.0296, "step": 21664 }, { "epoch": 0.53, "learning_rate": 9.38487527910352e-06, "loss": 1.9938, "step": 21665 }, { "epoch": 0.53, "learning_rate": 9.384078461724683e-06, "loss": 2.0575, "step": 21666 }, { "epoch": 0.53, "learning_rate": 9.383281648271318e-06, "loss": 2.1574, "step": 21667 }, { "epoch": 0.53, "learning_rate": 9.382484838748496e-06, "loss": 1.9283, "step": 21668 }, { "epoch": 0.53, "learning_rate": 9.381688033161298e-06, "loss": 2.0735, "step": 21669 }, { "epoch": 0.53, "learning_rate": 9.38089123151481e-06, "loss": 1.9756, "step": 21670 }, { "epoch": 0.53, "learning_rate": 9.3800944338141e-06, "loss": 2.0062, "step": 21671 }, { "epoch": 0.53, "learning_rate": 9.37929764006425e-06, "loss": 1.9486, "step": 21672 }, { "epoch": 0.53, "learning_rate": 9.378500850270336e-06, "loss": 1.9606, "step": 21673 }, { "epoch": 0.53, "learning_rate": 9.377704064437441e-06, "loss": 1.9911, "step": 21674 }, { "epoch": 0.53, "learning_rate": 9.376907282570644e-06, "loss": 2.0315, "step": 21675 }, { "epoch": 0.53, "learning_rate": 9.376110504675017e-06, "loss": 2.1896, "step": 21676 }, { "epoch": 0.53, "learning_rate": 9.375313730755643e-06, "loss": 1.9757, "step": 21677 }, { "epoch": 0.53, "learning_rate": 9.374516960817595e-06, "loss": 2.0943, "step": 21678 }, { "epoch": 0.53, "learning_rate": 9.373720194865955e-06, "loss": 1.8652, "step": 21679 }, { "epoch": 0.53, "learning_rate": 9.372923432905801e-06, "loss": 1.987, "step": 21680 }, { "epoch": 0.53, "learning_rate": 9.37212667494221e-06, "loss": 2.1997, "step": 21681 }, { "epoch": 0.53, "learning_rate": 9.371329920980264e-06, "loss": 2.0308, "step": 21682 }, { "epoch": 0.53, "learning_rate": 9.370533171025032e-06, "loss": 2.1758, "step": 21683 }, { "epoch": 0.53, "learning_rate": 9.369736425081597e-06, "loss": 1.975, "step": 21684 }, { "epoch": 0.53, "learning_rate": 9.368939683155043e-06, "loss": 2.0307, "step": 21685 }, { "epoch": 0.53, "learning_rate": 9.368142945250437e-06, "loss": 1.8107, "step": 21686 }, { "epoch": 0.53, "learning_rate": 9.367346211372866e-06, "loss": 1.8813, "step": 21687 }, { "epoch": 0.53, "learning_rate": 9.3665494815274e-06, "loss": 2.0578, "step": 21688 }, { "epoch": 0.53, "learning_rate": 9.365752755719119e-06, "loss": 2.0342, "step": 21689 }, { "epoch": 0.53, "learning_rate": 9.364956033953106e-06, "loss": 2.0371, "step": 21690 }, { "epoch": 0.53, "learning_rate": 9.364159316234434e-06, "loss": 1.914, "step": 21691 }, { "epoch": 0.53, "learning_rate": 9.363362602568185e-06, "loss": 1.9818, "step": 21692 }, { "epoch": 0.53, "learning_rate": 9.36256589295943e-06, "loss": 1.9063, "step": 21693 }, { "epoch": 0.53, "learning_rate": 9.361769187413247e-06, "loss": 1.8089, "step": 21694 }, { "epoch": 0.53, "learning_rate": 9.360972485934725e-06, "loss": 2.1002, "step": 21695 }, { "epoch": 0.53, "learning_rate": 9.360175788528928e-06, "loss": 1.8624, "step": 21696 }, { "epoch": 0.53, "learning_rate": 9.35937909520094e-06, "loss": 1.9878, "step": 21697 }, { "epoch": 0.53, "learning_rate": 9.358582405955839e-06, "loss": 2.2218, "step": 21698 }, { "epoch": 0.53, "learning_rate": 9.357785720798698e-06, "loss": 2.0683, "step": 21699 }, { "epoch": 0.53, "learning_rate": 9.356989039734606e-06, "loss": 2.0659, "step": 21700 }, { "epoch": 0.53, "learning_rate": 9.356192362768624e-06, "loss": 1.9103, "step": 21701 }, { "epoch": 0.53, "learning_rate": 9.35539568990584e-06, "loss": 2.1418, "step": 21702 }, { "epoch": 0.53, "learning_rate": 9.354599021151333e-06, "loss": 2.0827, "step": 21703 }, { "epoch": 0.53, "learning_rate": 9.353802356510173e-06, "loss": 2.0037, "step": 21704 }, { "epoch": 0.54, "learning_rate": 9.353005695987446e-06, "loss": 1.9757, "step": 21705 }, { "epoch": 0.54, "learning_rate": 9.352209039588219e-06, "loss": 2.1965, "step": 21706 }, { "epoch": 0.54, "learning_rate": 9.351412387317575e-06, "loss": 1.9649, "step": 21707 }, { "epoch": 0.54, "learning_rate": 9.350615739180595e-06, "loss": 2.0661, "step": 21708 }, { "epoch": 0.54, "learning_rate": 9.34981909518235e-06, "loss": 2.0352, "step": 21709 }, { "epoch": 0.54, "learning_rate": 9.349022455327922e-06, "loss": 2.1815, "step": 21710 }, { "epoch": 0.54, "learning_rate": 9.348225819622383e-06, "loss": 1.9123, "step": 21711 }, { "epoch": 0.54, "learning_rate": 9.347429188070814e-06, "loss": 1.7885, "step": 21712 }, { "epoch": 0.54, "learning_rate": 9.346632560678295e-06, "loss": 2.0559, "step": 21713 }, { "epoch": 0.54, "learning_rate": 9.345835937449895e-06, "loss": 2.2116, "step": 21714 }, { "epoch": 0.54, "learning_rate": 9.345039318390699e-06, "loss": 2.041, "step": 21715 }, { "epoch": 0.54, "learning_rate": 9.344242703505777e-06, "loss": 1.8533, "step": 21716 }, { "epoch": 0.54, "learning_rate": 9.343446092800212e-06, "loss": 2.0854, "step": 21717 }, { "epoch": 0.54, "learning_rate": 9.34264948627908e-06, "loss": 1.9677, "step": 21718 }, { "epoch": 0.54, "learning_rate": 9.341852883947456e-06, "loss": 1.8965, "step": 21719 }, { "epoch": 0.54, "learning_rate": 9.341056285810422e-06, "loss": 1.9694, "step": 21720 }, { "epoch": 0.54, "learning_rate": 9.340259691873046e-06, "loss": 1.8694, "step": 21721 }, { "epoch": 0.54, "learning_rate": 9.33946310214041e-06, "loss": 1.9431, "step": 21722 }, { "epoch": 0.54, "learning_rate": 9.338666516617595e-06, "loss": 2.1797, "step": 21723 }, { "epoch": 0.54, "learning_rate": 9.337869935309668e-06, "loss": 2.0821, "step": 21724 }, { "epoch": 0.54, "learning_rate": 9.337073358221718e-06, "loss": 2.132, "step": 21725 }, { "epoch": 0.54, "learning_rate": 9.33627678535881e-06, "loss": 2.0284, "step": 21726 }, { "epoch": 0.54, "learning_rate": 9.335480216726027e-06, "loss": 2.1747, "step": 21727 }, { "epoch": 0.54, "learning_rate": 9.33468365232845e-06, "loss": 1.907, "step": 21728 }, { "epoch": 0.54, "learning_rate": 9.333887092171147e-06, "loss": 1.8892, "step": 21729 }, { "epoch": 0.54, "learning_rate": 9.333090536259201e-06, "loss": 1.9909, "step": 21730 }, { "epoch": 0.54, "learning_rate": 9.332293984597684e-06, "loss": 2.171, "step": 21731 }, { "epoch": 0.54, "learning_rate": 9.331497437191674e-06, "loss": 1.8425, "step": 21732 }, { "epoch": 0.54, "learning_rate": 9.330700894046253e-06, "loss": 1.8538, "step": 21733 }, { "epoch": 0.54, "learning_rate": 9.329904355166488e-06, "loss": 2.0197, "step": 21734 }, { "epoch": 0.54, "learning_rate": 9.329107820557466e-06, "loss": 2.084, "step": 21735 }, { "epoch": 0.54, "learning_rate": 9.328311290224257e-06, "loss": 2.1329, "step": 21736 }, { "epoch": 0.54, "learning_rate": 9.327514764171937e-06, "loss": 2.0823, "step": 21737 }, { "epoch": 0.54, "learning_rate": 9.326718242405587e-06, "loss": 1.911, "step": 21738 }, { "epoch": 0.54, "learning_rate": 9.325921724930279e-06, "loss": 2.0089, "step": 21739 }, { "epoch": 0.54, "learning_rate": 9.325125211751095e-06, "loss": 1.9134, "step": 21740 }, { "epoch": 0.54, "learning_rate": 9.324328702873105e-06, "loss": 1.9002, "step": 21741 }, { "epoch": 0.54, "learning_rate": 9.323532198301388e-06, "loss": 2.1405, "step": 21742 }, { "epoch": 0.54, "learning_rate": 9.322735698041021e-06, "loss": 2.0747, "step": 21743 }, { "epoch": 0.54, "learning_rate": 9.32193920209708e-06, "loss": 2.0023, "step": 21744 }, { "epoch": 0.54, "learning_rate": 9.321142710474644e-06, "loss": 2.0551, "step": 21745 }, { "epoch": 0.54, "learning_rate": 9.320346223178784e-06, "loss": 2.0739, "step": 21746 }, { "epoch": 0.54, "learning_rate": 9.319549740214578e-06, "loss": 2.0296, "step": 21747 }, { "epoch": 0.54, "learning_rate": 9.318753261587108e-06, "loss": 1.9241, "step": 21748 }, { "epoch": 0.54, "learning_rate": 9.317956787301442e-06, "loss": 1.9141, "step": 21749 }, { "epoch": 0.54, "learning_rate": 9.31716031736266e-06, "loss": 1.9485, "step": 21750 }, { "epoch": 0.54, "learning_rate": 9.316363851775837e-06, "loss": 2.0135, "step": 21751 }, { "epoch": 0.54, "learning_rate": 9.315567390546049e-06, "loss": 2.1045, "step": 21752 }, { "epoch": 0.54, "learning_rate": 9.314770933678379e-06, "loss": 2.0489, "step": 21753 }, { "epoch": 0.54, "learning_rate": 9.313974481177892e-06, "loss": 1.9325, "step": 21754 }, { "epoch": 0.54, "learning_rate": 9.31317803304967e-06, "loss": 2.1358, "step": 21755 }, { "epoch": 0.54, "learning_rate": 9.312381589298791e-06, "loss": 1.9075, "step": 21756 }, { "epoch": 0.54, "learning_rate": 9.311585149930324e-06, "loss": 2.1054, "step": 21757 }, { "epoch": 0.54, "learning_rate": 9.310788714949353e-06, "loss": 1.9812, "step": 21758 }, { "epoch": 0.54, "learning_rate": 9.309992284360946e-06, "loss": 1.988, "step": 21759 }, { "epoch": 0.54, "learning_rate": 9.309195858170186e-06, "loss": 2.0936, "step": 21760 }, { "epoch": 0.54, "learning_rate": 9.308399436382145e-06, "loss": 2.0544, "step": 21761 }, { "epoch": 0.54, "learning_rate": 9.307603019001899e-06, "loss": 2.1049, "step": 21762 }, { "epoch": 0.54, "learning_rate": 9.306806606034528e-06, "loss": 2.1414, "step": 21763 }, { "epoch": 0.54, "learning_rate": 9.3060101974851e-06, "loss": 2.1655, "step": 21764 }, { "epoch": 0.54, "learning_rate": 9.305213793358696e-06, "loss": 1.9335, "step": 21765 }, { "epoch": 0.54, "learning_rate": 9.304417393660395e-06, "loss": 1.9959, "step": 21766 }, { "epoch": 0.54, "learning_rate": 9.303620998395265e-06, "loss": 2.2484, "step": 21767 }, { "epoch": 0.54, "learning_rate": 9.302824607568385e-06, "loss": 2.1237, "step": 21768 }, { "epoch": 0.54, "learning_rate": 9.302028221184833e-06, "loss": 2.0228, "step": 21769 }, { "epoch": 0.54, "learning_rate": 9.30123183924968e-06, "loss": 1.9849, "step": 21770 }, { "epoch": 0.54, "learning_rate": 9.300435461768008e-06, "loss": 2.0614, "step": 21771 }, { "epoch": 0.54, "learning_rate": 9.299639088744885e-06, "loss": 1.9334, "step": 21772 }, { "epoch": 0.54, "learning_rate": 9.298842720185394e-06, "loss": 1.8626, "step": 21773 }, { "epoch": 0.54, "learning_rate": 9.298046356094605e-06, "loss": 1.8452, "step": 21774 }, { "epoch": 0.54, "learning_rate": 9.297249996477593e-06, "loss": 2.0707, "step": 21775 }, { "epoch": 0.54, "learning_rate": 9.29645364133944e-06, "loss": 1.9184, "step": 21776 }, { "epoch": 0.54, "learning_rate": 9.295657290685215e-06, "loss": 2.0186, "step": 21777 }, { "epoch": 0.54, "learning_rate": 9.294860944519996e-06, "loss": 1.95, "step": 21778 }, { "epoch": 0.54, "learning_rate": 9.294064602848856e-06, "loss": 2.2547, "step": 21779 }, { "epoch": 0.54, "learning_rate": 9.293268265676872e-06, "loss": 2.0652, "step": 21780 }, { "epoch": 0.54, "learning_rate": 9.292471933009124e-06, "loss": 1.9958, "step": 21781 }, { "epoch": 0.54, "learning_rate": 9.29167560485068e-06, "loss": 2.0881, "step": 21782 }, { "epoch": 0.54, "learning_rate": 9.290879281206623e-06, "loss": 2.0793, "step": 21783 }, { "epoch": 0.54, "learning_rate": 9.290082962082017e-06, "loss": 2.0227, "step": 21784 }, { "epoch": 0.54, "learning_rate": 9.289286647481945e-06, "loss": 2.0362, "step": 21785 }, { "epoch": 0.54, "learning_rate": 9.288490337411482e-06, "loss": 1.9264, "step": 21786 }, { "epoch": 0.54, "learning_rate": 9.287694031875701e-06, "loss": 1.9977, "step": 21787 }, { "epoch": 0.54, "learning_rate": 9.286897730879681e-06, "loss": 2.239, "step": 21788 }, { "epoch": 0.54, "learning_rate": 9.286101434428491e-06, "loss": 1.9508, "step": 21789 }, { "epoch": 0.54, "learning_rate": 9.28530514252721e-06, "loss": 2.1211, "step": 21790 }, { "epoch": 0.54, "learning_rate": 9.284508855180914e-06, "loss": 1.8539, "step": 21791 }, { "epoch": 0.54, "learning_rate": 9.283712572394672e-06, "loss": 2.025, "step": 21792 }, { "epoch": 0.54, "learning_rate": 9.282916294173568e-06, "loss": 2.0092, "step": 21793 }, { "epoch": 0.54, "learning_rate": 9.282120020522672e-06, "loss": 2.0159, "step": 21794 }, { "epoch": 0.54, "learning_rate": 9.281323751447055e-06, "loss": 1.7895, "step": 21795 }, { "epoch": 0.54, "learning_rate": 9.280527486951799e-06, "loss": 1.8803, "step": 21796 }, { "epoch": 0.54, "learning_rate": 9.279731227041972e-06, "loss": 1.8999, "step": 21797 }, { "epoch": 0.54, "learning_rate": 9.278934971722658e-06, "loss": 2.1495, "step": 21798 }, { "epoch": 0.54, "learning_rate": 9.278138720998925e-06, "loss": 2.1105, "step": 21799 }, { "epoch": 0.54, "learning_rate": 9.277342474875844e-06, "loss": 2.0639, "step": 21800 }, { "epoch": 0.54, "learning_rate": 9.276546233358503e-06, "loss": 2.1728, "step": 21801 }, { "epoch": 0.54, "learning_rate": 9.275749996451962e-06, "loss": 2.0549, "step": 21802 }, { "epoch": 0.54, "learning_rate": 9.274953764161307e-06, "loss": 2.0392, "step": 21803 }, { "epoch": 0.54, "learning_rate": 9.274157536491603e-06, "loss": 1.9837, "step": 21804 }, { "epoch": 0.54, "learning_rate": 9.273361313447932e-06, "loss": 2.03, "step": 21805 }, { "epoch": 0.54, "learning_rate": 9.272565095035368e-06, "loss": 2.167, "step": 21806 }, { "epoch": 0.54, "learning_rate": 9.271768881258983e-06, "loss": 1.9293, "step": 21807 }, { "epoch": 0.54, "learning_rate": 9.270972672123854e-06, "loss": 2.021, "step": 21808 }, { "epoch": 0.54, "learning_rate": 9.27017646763505e-06, "loss": 2.0678, "step": 21809 }, { "epoch": 0.54, "learning_rate": 9.269380267797647e-06, "loss": 1.9312, "step": 21810 }, { "epoch": 0.54, "learning_rate": 9.268584072616727e-06, "loss": 2.1313, "step": 21811 }, { "epoch": 0.54, "learning_rate": 9.267787882097357e-06, "loss": 1.9238, "step": 21812 }, { "epoch": 0.54, "learning_rate": 9.266991696244613e-06, "loss": 2.1117, "step": 21813 }, { "epoch": 0.54, "learning_rate": 9.26619551506357e-06, "loss": 2.231, "step": 21814 }, { "epoch": 0.54, "learning_rate": 9.265399338559303e-06, "loss": 1.9892, "step": 21815 }, { "epoch": 0.54, "learning_rate": 9.264603166736888e-06, "loss": 1.9734, "step": 21816 }, { "epoch": 0.54, "learning_rate": 9.263806999601392e-06, "loss": 1.8956, "step": 21817 }, { "epoch": 0.54, "learning_rate": 9.263010837157893e-06, "loss": 2.0189, "step": 21818 }, { "epoch": 0.54, "learning_rate": 9.262214679411472e-06, "loss": 1.9318, "step": 21819 }, { "epoch": 0.54, "learning_rate": 9.261418526367192e-06, "loss": 2.196, "step": 21820 }, { "epoch": 0.54, "learning_rate": 9.260622378030134e-06, "loss": 1.8651, "step": 21821 }, { "epoch": 0.54, "learning_rate": 9.259826234405372e-06, "loss": 1.883, "step": 21822 }, { "epoch": 0.54, "learning_rate": 9.259030095497975e-06, "loss": 2.2, "step": 21823 }, { "epoch": 0.54, "learning_rate": 9.258233961313026e-06, "loss": 1.9979, "step": 21824 }, { "epoch": 0.54, "learning_rate": 9.257437831855589e-06, "loss": 2.0185, "step": 21825 }, { "epoch": 0.54, "learning_rate": 9.256641707130747e-06, "loss": 1.9901, "step": 21826 }, { "epoch": 0.54, "learning_rate": 9.255845587143568e-06, "loss": 2.0658, "step": 21827 }, { "epoch": 0.54, "learning_rate": 9.255049471899125e-06, "loss": 2.0318, "step": 21828 }, { "epoch": 0.54, "learning_rate": 9.254253361402496e-06, "loss": 2.0461, "step": 21829 }, { "epoch": 0.54, "learning_rate": 9.253457255658755e-06, "loss": 1.9872, "step": 21830 }, { "epoch": 0.54, "learning_rate": 9.252661154672973e-06, "loss": 1.9277, "step": 21831 }, { "epoch": 0.54, "learning_rate": 9.251865058450223e-06, "loss": 1.7548, "step": 21832 }, { "epoch": 0.54, "learning_rate": 9.251068966995582e-06, "loss": 2.0457, "step": 21833 }, { "epoch": 0.54, "learning_rate": 9.250272880314126e-06, "loss": 2.1178, "step": 21834 }, { "epoch": 0.54, "learning_rate": 9.249476798410921e-06, "loss": 2.0299, "step": 21835 }, { "epoch": 0.54, "learning_rate": 9.24868072129105e-06, "loss": 2.1519, "step": 21836 }, { "epoch": 0.54, "learning_rate": 9.247884648959576e-06, "loss": 2.0193, "step": 21837 }, { "epoch": 0.54, "learning_rate": 9.24708858142158e-06, "loss": 1.9231, "step": 21838 }, { "epoch": 0.54, "learning_rate": 9.246292518682134e-06, "loss": 1.8994, "step": 21839 }, { "epoch": 0.54, "learning_rate": 9.24549646074631e-06, "loss": 2.0139, "step": 21840 }, { "epoch": 0.54, "learning_rate": 9.244700407619186e-06, "loss": 1.9917, "step": 21841 }, { "epoch": 0.54, "learning_rate": 9.24390435930583e-06, "loss": 2.133, "step": 21842 }, { "epoch": 0.54, "learning_rate": 9.243108315811318e-06, "loss": 2.0138, "step": 21843 }, { "epoch": 0.54, "learning_rate": 9.242312277140728e-06, "loss": 2.0701, "step": 21844 }, { "epoch": 0.54, "learning_rate": 9.241516243299123e-06, "loss": 1.9672, "step": 21845 }, { "epoch": 0.54, "learning_rate": 9.240720214291584e-06, "loss": 2.0288, "step": 21846 }, { "epoch": 0.54, "learning_rate": 9.239924190123182e-06, "loss": 1.9116, "step": 21847 }, { "epoch": 0.54, "learning_rate": 9.239128170798992e-06, "loss": 1.8389, "step": 21848 }, { "epoch": 0.54, "learning_rate": 9.238332156324086e-06, "loss": 1.872, "step": 21849 }, { "epoch": 0.54, "learning_rate": 9.237536146703536e-06, "loss": 2.1055, "step": 21850 }, { "epoch": 0.54, "learning_rate": 9.23674014194242e-06, "loss": 2.0091, "step": 21851 }, { "epoch": 0.54, "learning_rate": 9.235944142045805e-06, "loss": 2.1138, "step": 21852 }, { "epoch": 0.54, "learning_rate": 9.235148147018765e-06, "loss": 2.2809, "step": 21853 }, { "epoch": 0.54, "learning_rate": 9.234352156866381e-06, "loss": 2.0874, "step": 21854 }, { "epoch": 0.54, "learning_rate": 9.233556171593715e-06, "loss": 1.9129, "step": 21855 }, { "epoch": 0.54, "learning_rate": 9.232760191205848e-06, "loss": 2.0346, "step": 21856 }, { "epoch": 0.54, "learning_rate": 9.231964215707849e-06, "loss": 1.9214, "step": 21857 }, { "epoch": 0.54, "learning_rate": 9.231168245104792e-06, "loss": 1.8737, "step": 21858 }, { "epoch": 0.54, "learning_rate": 9.230372279401753e-06, "loss": 2.2629, "step": 21859 }, { "epoch": 0.54, "learning_rate": 9.2295763186038e-06, "loss": 1.9599, "step": 21860 }, { "epoch": 0.54, "learning_rate": 9.22878036271601e-06, "loss": 2.0091, "step": 21861 }, { "epoch": 0.54, "learning_rate": 9.227984411743451e-06, "loss": 1.9119, "step": 21862 }, { "epoch": 0.54, "learning_rate": 9.227188465691199e-06, "loss": 2.174, "step": 21863 }, { "epoch": 0.54, "learning_rate": 9.22639252456433e-06, "loss": 1.9174, "step": 21864 }, { "epoch": 0.54, "learning_rate": 9.225596588367912e-06, "loss": 2.1051, "step": 21865 }, { "epoch": 0.54, "learning_rate": 9.22480065710702e-06, "loss": 2.0881, "step": 21866 }, { "epoch": 0.54, "learning_rate": 9.224004730786724e-06, "loss": 2.0102, "step": 21867 }, { "epoch": 0.54, "learning_rate": 9.2232088094121e-06, "loss": 1.8535, "step": 21868 }, { "epoch": 0.54, "learning_rate": 9.222412892988221e-06, "loss": 2.2052, "step": 21869 }, { "epoch": 0.54, "learning_rate": 9.221616981520155e-06, "loss": 2.0117, "step": 21870 }, { "epoch": 0.54, "learning_rate": 9.220821075012983e-06, "loss": 1.7696, "step": 21871 }, { "epoch": 0.54, "learning_rate": 9.220025173471766e-06, "loss": 1.9119, "step": 21872 }, { "epoch": 0.54, "learning_rate": 9.219229276901583e-06, "loss": 2.0025, "step": 21873 }, { "epoch": 0.54, "learning_rate": 9.218433385307509e-06, "loss": 1.9822, "step": 21874 }, { "epoch": 0.54, "learning_rate": 9.217637498694611e-06, "loss": 1.9817, "step": 21875 }, { "epoch": 0.54, "learning_rate": 9.216841617067966e-06, "loss": 2.0766, "step": 21876 }, { "epoch": 0.54, "learning_rate": 9.216045740432647e-06, "loss": 2.0365, "step": 21877 }, { "epoch": 0.54, "learning_rate": 9.21524986879372e-06, "loss": 2.0805, "step": 21878 }, { "epoch": 0.54, "learning_rate": 9.214454002156267e-06, "loss": 1.7961, "step": 21879 }, { "epoch": 0.54, "learning_rate": 9.213658140525349e-06, "loss": 2.1672, "step": 21880 }, { "epoch": 0.54, "learning_rate": 9.212862283906044e-06, "loss": 2.063, "step": 21881 }, { "epoch": 0.54, "learning_rate": 9.212066432303427e-06, "loss": 1.8656, "step": 21882 }, { "epoch": 0.54, "learning_rate": 9.211270585722567e-06, "loss": 2.0384, "step": 21883 }, { "epoch": 0.54, "learning_rate": 9.210474744168538e-06, "loss": 1.8665, "step": 21884 }, { "epoch": 0.54, "learning_rate": 9.209678907646407e-06, "loss": 2.0113, "step": 21885 }, { "epoch": 0.54, "learning_rate": 9.20888307616125e-06, "loss": 2.038, "step": 21886 }, { "epoch": 0.54, "learning_rate": 9.208087249718142e-06, "loss": 2.1397, "step": 21887 }, { "epoch": 0.54, "learning_rate": 9.20729142832215e-06, "loss": 1.9253, "step": 21888 }, { "epoch": 0.54, "learning_rate": 9.20649561197835e-06, "loss": 1.9993, "step": 21889 }, { "epoch": 0.54, "learning_rate": 9.205699800691812e-06, "loss": 2.0013, "step": 21890 }, { "epoch": 0.54, "learning_rate": 9.204903994467604e-06, "loss": 1.8756, "step": 21891 }, { "epoch": 0.54, "learning_rate": 9.204108193310805e-06, "loss": 1.919, "step": 21892 }, { "epoch": 0.54, "learning_rate": 9.203312397226482e-06, "loss": 1.9887, "step": 21893 }, { "epoch": 0.54, "learning_rate": 9.202516606219714e-06, "loss": 2.0635, "step": 21894 }, { "epoch": 0.54, "learning_rate": 9.201720820295563e-06, "loss": 1.918, "step": 21895 }, { "epoch": 0.54, "learning_rate": 9.200925039459103e-06, "loss": 1.874, "step": 21896 }, { "epoch": 0.54, "learning_rate": 9.200129263715415e-06, "loss": 1.9814, "step": 21897 }, { "epoch": 0.54, "learning_rate": 9.19933349306956e-06, "loss": 2.1107, "step": 21898 }, { "epoch": 0.54, "learning_rate": 9.198537727526613e-06, "loss": 2.146, "step": 21899 }, { "epoch": 0.54, "learning_rate": 9.197741967091645e-06, "loss": 1.8809, "step": 21900 }, { "epoch": 0.54, "learning_rate": 9.196946211769734e-06, "loss": 1.9844, "step": 21901 }, { "epoch": 0.54, "learning_rate": 9.196150461565942e-06, "loss": 1.915, "step": 21902 }, { "epoch": 0.54, "learning_rate": 9.195354716485345e-06, "loss": 1.9935, "step": 21903 }, { "epoch": 0.54, "learning_rate": 9.19455897653302e-06, "loss": 2.135, "step": 21904 }, { "epoch": 0.54, "learning_rate": 9.193763241714028e-06, "loss": 2.0612, "step": 21905 }, { "epoch": 0.54, "learning_rate": 9.192967512033446e-06, "loss": 2.098, "step": 21906 }, { "epoch": 0.54, "learning_rate": 9.192171787496349e-06, "loss": 2.0034, "step": 21907 }, { "epoch": 0.54, "learning_rate": 9.191376068107797e-06, "loss": 2.0267, "step": 21908 }, { "epoch": 0.54, "learning_rate": 9.190580353872876e-06, "loss": 2.0645, "step": 21909 }, { "epoch": 0.54, "learning_rate": 9.189784644796646e-06, "loss": 1.9471, "step": 21910 }, { "epoch": 0.54, "learning_rate": 9.188988940884181e-06, "loss": 1.946, "step": 21911 }, { "epoch": 0.54, "learning_rate": 9.18819324214056e-06, "loss": 2.0534, "step": 21912 }, { "epoch": 0.54, "learning_rate": 9.187397548570844e-06, "loss": 1.8961, "step": 21913 }, { "epoch": 0.54, "learning_rate": 9.18660186018011e-06, "loss": 2.2161, "step": 21914 }, { "epoch": 0.54, "learning_rate": 9.185806176973427e-06, "loss": 1.9355, "step": 21915 }, { "epoch": 0.54, "learning_rate": 9.185010498955864e-06, "loss": 1.8638, "step": 21916 }, { "epoch": 0.54, "learning_rate": 9.184214826132497e-06, "loss": 1.9555, "step": 21917 }, { "epoch": 0.54, "learning_rate": 9.183419158508394e-06, "loss": 2.0297, "step": 21918 }, { "epoch": 0.54, "learning_rate": 9.182623496088629e-06, "loss": 2.0269, "step": 21919 }, { "epoch": 0.54, "learning_rate": 9.181827838878268e-06, "loss": 1.8477, "step": 21920 }, { "epoch": 0.54, "learning_rate": 9.181032186882383e-06, "loss": 1.9913, "step": 21921 }, { "epoch": 0.54, "learning_rate": 9.180236540106053e-06, "loss": 2.1534, "step": 21922 }, { "epoch": 0.54, "learning_rate": 9.179440898554338e-06, "loss": 1.8529, "step": 21923 }, { "epoch": 0.54, "learning_rate": 9.178645262232319e-06, "loss": 1.8984, "step": 21924 }, { "epoch": 0.54, "learning_rate": 9.177849631145055e-06, "loss": 2.0303, "step": 21925 }, { "epoch": 0.54, "learning_rate": 9.177054005297625e-06, "loss": 1.9849, "step": 21926 }, { "epoch": 0.54, "learning_rate": 9.176258384695098e-06, "loss": 2.1881, "step": 21927 }, { "epoch": 0.54, "learning_rate": 9.175462769342545e-06, "loss": 2.007, "step": 21928 }, { "epoch": 0.54, "learning_rate": 9.174667159245039e-06, "loss": 1.9014, "step": 21929 }, { "epoch": 0.54, "learning_rate": 9.173871554407646e-06, "loss": 2.2111, "step": 21930 }, { "epoch": 0.54, "learning_rate": 9.173075954835435e-06, "loss": 2.1126, "step": 21931 }, { "epoch": 0.54, "learning_rate": 9.172280360533488e-06, "loss": 2.0326, "step": 21932 }, { "epoch": 0.54, "learning_rate": 9.171484771506863e-06, "loss": 2.0564, "step": 21933 }, { "epoch": 0.54, "learning_rate": 9.170689187760635e-06, "loss": 1.8858, "step": 21934 }, { "epoch": 0.54, "learning_rate": 9.169893609299879e-06, "loss": 2.1077, "step": 21935 }, { "epoch": 0.54, "learning_rate": 9.169098036129656e-06, "loss": 2.0624, "step": 21936 }, { "epoch": 0.54, "learning_rate": 9.168302468255048e-06, "loss": 1.9632, "step": 21937 }, { "epoch": 0.54, "learning_rate": 9.167506905681117e-06, "loss": 1.975, "step": 21938 }, { "epoch": 0.54, "learning_rate": 9.166711348412934e-06, "loss": 1.9318, "step": 21939 }, { "epoch": 0.54, "learning_rate": 9.165915796455574e-06, "loss": 1.8332, "step": 21940 }, { "epoch": 0.54, "learning_rate": 9.165120249814102e-06, "loss": 1.9554, "step": 21941 }, { "epoch": 0.54, "learning_rate": 9.164324708493596e-06, "loss": 1.8859, "step": 21942 }, { "epoch": 0.54, "learning_rate": 9.163529172499114e-06, "loss": 2.2071, "step": 21943 }, { "epoch": 0.54, "learning_rate": 9.162733641835734e-06, "loss": 1.9538, "step": 21944 }, { "epoch": 0.54, "learning_rate": 9.16193811650853e-06, "loss": 2.1075, "step": 21945 }, { "epoch": 0.54, "learning_rate": 9.161142596522563e-06, "loss": 2.1532, "step": 21946 }, { "epoch": 0.54, "learning_rate": 9.160347081882912e-06, "loss": 2.1212, "step": 21947 }, { "epoch": 0.54, "learning_rate": 9.159551572594639e-06, "loss": 2.1722, "step": 21948 }, { "epoch": 0.54, "learning_rate": 9.158756068662817e-06, "loss": 2.0239, "step": 21949 }, { "epoch": 0.54, "learning_rate": 9.157960570092522e-06, "loss": 1.9187, "step": 21950 }, { "epoch": 0.54, "learning_rate": 9.157165076888815e-06, "loss": 1.9926, "step": 21951 }, { "epoch": 0.54, "learning_rate": 9.156369589056769e-06, "loss": 2.0607, "step": 21952 }, { "epoch": 0.54, "learning_rate": 9.155574106601454e-06, "loss": 1.828, "step": 21953 }, { "epoch": 0.54, "learning_rate": 9.154778629527943e-06, "loss": 2.0729, "step": 21954 }, { "epoch": 0.54, "learning_rate": 9.153983157841302e-06, "loss": 1.9915, "step": 21955 }, { "epoch": 0.54, "learning_rate": 9.153187691546602e-06, "loss": 2.11, "step": 21956 }, { "epoch": 0.54, "learning_rate": 9.152392230648915e-06, "loss": 2.0366, "step": 21957 }, { "epoch": 0.54, "learning_rate": 9.151596775153306e-06, "loss": 2.0513, "step": 21958 }, { "epoch": 0.54, "learning_rate": 9.150801325064846e-06, "loss": 2.0197, "step": 21959 }, { "epoch": 0.54, "learning_rate": 9.150005880388611e-06, "loss": 1.9179, "step": 21960 }, { "epoch": 0.54, "learning_rate": 9.149210441129661e-06, "loss": 1.9118, "step": 21961 }, { "epoch": 0.54, "learning_rate": 9.148415007293072e-06, "loss": 1.9825, "step": 21962 }, { "epoch": 0.54, "learning_rate": 9.14761957888391e-06, "loss": 1.8771, "step": 21963 }, { "epoch": 0.54, "learning_rate": 9.146824155907245e-06, "loss": 1.925, "step": 21964 }, { "epoch": 0.54, "learning_rate": 9.146028738368153e-06, "loss": 1.9627, "step": 21965 }, { "epoch": 0.54, "learning_rate": 9.145233326271694e-06, "loss": 1.8708, "step": 21966 }, { "epoch": 0.54, "learning_rate": 9.144437919622945e-06, "loss": 2.1418, "step": 21967 }, { "epoch": 0.54, "learning_rate": 9.143642518426968e-06, "loss": 2.224, "step": 21968 }, { "epoch": 0.54, "learning_rate": 9.142847122688836e-06, "loss": 2.0666, "step": 21969 }, { "epoch": 0.54, "learning_rate": 9.142051732413622e-06, "loss": 1.8967, "step": 21970 }, { "epoch": 0.54, "learning_rate": 9.141256347606389e-06, "loss": 1.9607, "step": 21971 }, { "epoch": 0.54, "learning_rate": 9.140460968272214e-06, "loss": 2.16, "step": 21972 }, { "epoch": 0.54, "learning_rate": 9.139665594416156e-06, "loss": 1.9221, "step": 21973 }, { "epoch": 0.54, "learning_rate": 9.138870226043291e-06, "loss": 1.9779, "step": 21974 }, { "epoch": 0.54, "learning_rate": 9.13807486315869e-06, "loss": 2.0377, "step": 21975 }, { "epoch": 0.54, "learning_rate": 9.137279505767414e-06, "loss": 1.8532, "step": 21976 }, { "epoch": 0.54, "learning_rate": 9.136484153874542e-06, "loss": 1.9969, "step": 21977 }, { "epoch": 0.54, "learning_rate": 9.135688807485133e-06, "loss": 2.1909, "step": 21978 }, { "epoch": 0.54, "learning_rate": 9.13489346660426e-06, "loss": 1.986, "step": 21979 }, { "epoch": 0.54, "learning_rate": 9.134098131236997e-06, "loss": 1.9866, "step": 21980 }, { "epoch": 0.54, "learning_rate": 9.133302801388407e-06, "loss": 2.1217, "step": 21981 }, { "epoch": 0.54, "learning_rate": 9.132507477063564e-06, "loss": 2.0462, "step": 21982 }, { "epoch": 0.54, "learning_rate": 9.131712158267529e-06, "loss": 2.1114, "step": 21983 }, { "epoch": 0.54, "learning_rate": 9.130916845005375e-06, "loss": 1.9392, "step": 21984 }, { "epoch": 0.54, "learning_rate": 9.130121537282177e-06, "loss": 1.939, "step": 21985 }, { "epoch": 0.54, "learning_rate": 9.12932623510299e-06, "loss": 2.0949, "step": 21986 }, { "epoch": 0.54, "learning_rate": 9.128530938472898e-06, "loss": 1.8862, "step": 21987 }, { "epoch": 0.54, "learning_rate": 9.127735647396956e-06, "loss": 1.9606, "step": 21988 }, { "epoch": 0.54, "learning_rate": 9.126940361880241e-06, "loss": 2.0909, "step": 21989 }, { "epoch": 0.54, "learning_rate": 9.126145081927824e-06, "loss": 1.7928, "step": 21990 }, { "epoch": 0.54, "learning_rate": 9.125349807544764e-06, "loss": 1.9246, "step": 21991 }, { "epoch": 0.54, "learning_rate": 9.124554538736134e-06, "loss": 2.0142, "step": 21992 }, { "epoch": 0.54, "learning_rate": 9.123759275507009e-06, "loss": 2.0526, "step": 21993 }, { "epoch": 0.54, "learning_rate": 9.122964017862448e-06, "loss": 2.1279, "step": 21994 }, { "epoch": 0.54, "learning_rate": 9.122168765807527e-06, "loss": 2.0269, "step": 21995 }, { "epoch": 0.54, "learning_rate": 9.121373519347304e-06, "loss": 2.1747, "step": 21996 }, { "epoch": 0.54, "learning_rate": 9.120578278486856e-06, "loss": 2.005, "step": 21997 }, { "epoch": 0.54, "learning_rate": 9.119783043231251e-06, "loss": 2.0092, "step": 21998 }, { "epoch": 0.54, "learning_rate": 9.118987813585554e-06, "loss": 1.8872, "step": 21999 }, { "epoch": 0.54, "learning_rate": 9.118192589554837e-06, "loss": 1.956, "step": 22000 }, { "epoch": 0.54, "learning_rate": 9.117397371144164e-06, "loss": 1.9388, "step": 22001 }, { "epoch": 0.54, "learning_rate": 9.116602158358604e-06, "loss": 2.0169, "step": 22002 }, { "epoch": 0.54, "learning_rate": 9.11580695120323e-06, "loss": 2.0368, "step": 22003 }, { "epoch": 0.54, "learning_rate": 9.115011749683105e-06, "loss": 1.9462, "step": 22004 }, { "epoch": 0.54, "learning_rate": 9.114216553803297e-06, "loss": 2.1368, "step": 22005 }, { "epoch": 0.54, "learning_rate": 9.113421363568876e-06, "loss": 2.0135, "step": 22006 }, { "epoch": 0.54, "learning_rate": 9.112626178984908e-06, "loss": 2.1237, "step": 22007 }, { "epoch": 0.54, "learning_rate": 9.111831000056468e-06, "loss": 1.9098, "step": 22008 }, { "epoch": 0.54, "learning_rate": 9.111035826788615e-06, "loss": 2.0971, "step": 22009 }, { "epoch": 0.54, "learning_rate": 9.110240659186422e-06, "loss": 2.0051, "step": 22010 }, { "epoch": 0.54, "learning_rate": 9.109445497254953e-06, "loss": 2.1213, "step": 22011 }, { "epoch": 0.54, "learning_rate": 9.108650340999277e-06, "loss": 2.0171, "step": 22012 }, { "epoch": 0.54, "learning_rate": 9.107855190424468e-06, "loss": 1.8973, "step": 22013 }, { "epoch": 0.54, "learning_rate": 9.107060045535583e-06, "loss": 1.9996, "step": 22014 }, { "epoch": 0.54, "learning_rate": 9.1062649063377e-06, "loss": 2.1642, "step": 22015 }, { "epoch": 0.54, "learning_rate": 9.105469772835878e-06, "loss": 2.0477, "step": 22016 }, { "epoch": 0.54, "learning_rate": 9.104674645035188e-06, "loss": 1.8438, "step": 22017 }, { "epoch": 0.54, "learning_rate": 9.103879522940703e-06, "loss": 1.8403, "step": 22018 }, { "epoch": 0.54, "learning_rate": 9.103084406557484e-06, "loss": 1.9243, "step": 22019 }, { "epoch": 0.54, "learning_rate": 9.102289295890602e-06, "loss": 2.0908, "step": 22020 }, { "epoch": 0.54, "learning_rate": 9.101494190945119e-06, "loss": 1.9075, "step": 22021 }, { "epoch": 0.54, "learning_rate": 9.100699091726106e-06, "loss": 2.1281, "step": 22022 }, { "epoch": 0.54, "learning_rate": 9.099903998238634e-06, "loss": 1.9806, "step": 22023 }, { "epoch": 0.54, "learning_rate": 9.099108910487766e-06, "loss": 2.0242, "step": 22024 }, { "epoch": 0.54, "learning_rate": 9.098313828478571e-06, "loss": 2.0354, "step": 22025 }, { "epoch": 0.54, "learning_rate": 9.097518752216115e-06, "loss": 1.9116, "step": 22026 }, { "epoch": 0.54, "learning_rate": 9.096723681705465e-06, "loss": 2.0086, "step": 22027 }, { "epoch": 0.54, "learning_rate": 9.095928616951694e-06, "loss": 2.1514, "step": 22028 }, { "epoch": 0.54, "learning_rate": 9.09513355795986e-06, "loss": 2.1536, "step": 22029 }, { "epoch": 0.54, "learning_rate": 9.094338504735037e-06, "loss": 1.9467, "step": 22030 }, { "epoch": 0.54, "learning_rate": 9.09354345728229e-06, "loss": 1.8861, "step": 22031 }, { "epoch": 0.54, "learning_rate": 9.092748415606683e-06, "loss": 1.9929, "step": 22032 }, { "epoch": 0.54, "learning_rate": 9.091953379713288e-06, "loss": 2.0108, "step": 22033 }, { "epoch": 0.54, "learning_rate": 9.091158349607167e-06, "loss": 1.9518, "step": 22034 }, { "epoch": 0.54, "learning_rate": 9.090363325293397e-06, "loss": 1.9164, "step": 22035 }, { "epoch": 0.54, "learning_rate": 9.089568306777032e-06, "loss": 1.9393, "step": 22036 }, { "epoch": 0.54, "learning_rate": 9.088773294063145e-06, "loss": 1.9765, "step": 22037 }, { "epoch": 0.54, "learning_rate": 9.087978287156808e-06, "loss": 1.9251, "step": 22038 }, { "epoch": 0.54, "learning_rate": 9.087183286063077e-06, "loss": 2.0156, "step": 22039 }, { "epoch": 0.54, "learning_rate": 9.086388290787028e-06, "loss": 1.9916, "step": 22040 }, { "epoch": 0.54, "learning_rate": 9.08559330133372e-06, "loss": 2.0666, "step": 22041 }, { "epoch": 0.54, "learning_rate": 9.084798317708226e-06, "loss": 1.8848, "step": 22042 }, { "epoch": 0.54, "learning_rate": 9.084003339915614e-06, "loss": 2.0079, "step": 22043 }, { "epoch": 0.54, "learning_rate": 9.083208367960943e-06, "loss": 2.062, "step": 22044 }, { "epoch": 0.54, "learning_rate": 9.082413401849288e-06, "loss": 2.1454, "step": 22045 }, { "epoch": 0.54, "learning_rate": 9.081618441585706e-06, "loss": 1.782, "step": 22046 }, { "epoch": 0.54, "learning_rate": 9.080823487175271e-06, "loss": 2.0618, "step": 22047 }, { "epoch": 0.54, "learning_rate": 9.08002853862305e-06, "loss": 1.8275, "step": 22048 }, { "epoch": 0.54, "learning_rate": 9.079233595934106e-06, "loss": 2.0061, "step": 22049 }, { "epoch": 0.54, "learning_rate": 9.078438659113502e-06, "loss": 1.9216, "step": 22050 }, { "epoch": 0.54, "learning_rate": 9.077643728166314e-06, "loss": 1.7721, "step": 22051 }, { "epoch": 0.54, "learning_rate": 9.0768488030976e-06, "loss": 2.0911, "step": 22052 }, { "epoch": 0.54, "learning_rate": 9.076053883912432e-06, "loss": 2.0495, "step": 22053 }, { "epoch": 0.54, "learning_rate": 9.075258970615872e-06, "loss": 1.9743, "step": 22054 }, { "epoch": 0.54, "learning_rate": 9.074464063212987e-06, "loss": 2.0159, "step": 22055 }, { "epoch": 0.54, "learning_rate": 9.073669161708848e-06, "loss": 2.0444, "step": 22056 }, { "epoch": 0.54, "learning_rate": 9.072874266108514e-06, "loss": 1.9929, "step": 22057 }, { "epoch": 0.54, "learning_rate": 9.072079376417054e-06, "loss": 2.1131, "step": 22058 }, { "epoch": 0.54, "learning_rate": 9.071284492639534e-06, "loss": 2.0229, "step": 22059 }, { "epoch": 0.54, "learning_rate": 9.070489614781022e-06, "loss": 2.0385, "step": 22060 }, { "epoch": 0.54, "learning_rate": 9.069694742846586e-06, "loss": 2.0341, "step": 22061 }, { "epoch": 0.54, "learning_rate": 9.068899876841285e-06, "loss": 2.1102, "step": 22062 }, { "epoch": 0.54, "learning_rate": 9.06810501677019e-06, "loss": 2.0494, "step": 22063 }, { "epoch": 0.54, "learning_rate": 9.067310162638364e-06, "loss": 1.9637, "step": 22064 }, { "epoch": 0.54, "learning_rate": 9.066515314450873e-06, "loss": 1.8946, "step": 22065 }, { "epoch": 0.54, "learning_rate": 9.065720472212789e-06, "loss": 1.9314, "step": 22066 }, { "epoch": 0.54, "learning_rate": 9.064925635929167e-06, "loss": 2.0785, "step": 22067 }, { "epoch": 0.54, "learning_rate": 9.064130805605083e-06, "loss": 1.9021, "step": 22068 }, { "epoch": 0.54, "learning_rate": 9.063335981245596e-06, "loss": 2.1459, "step": 22069 }, { "epoch": 0.54, "learning_rate": 9.062541162855773e-06, "loss": 1.9707, "step": 22070 }, { "epoch": 0.54, "learning_rate": 9.061746350440686e-06, "loss": 2.0815, "step": 22071 }, { "epoch": 0.54, "learning_rate": 9.06095154400539e-06, "loss": 1.9561, "step": 22072 }, { "epoch": 0.54, "learning_rate": 9.06015674355496e-06, "loss": 2.1228, "step": 22073 }, { "epoch": 0.54, "learning_rate": 9.059361949094453e-06, "loss": 2.1591, "step": 22074 }, { "epoch": 0.54, "learning_rate": 9.05856716062894e-06, "loss": 2.0403, "step": 22075 }, { "epoch": 0.54, "learning_rate": 9.057772378163487e-06, "loss": 1.9865, "step": 22076 }, { "epoch": 0.54, "learning_rate": 9.056977601703154e-06, "loss": 2.0839, "step": 22077 }, { "epoch": 0.54, "learning_rate": 9.056182831253017e-06, "loss": 1.9515, "step": 22078 }, { "epoch": 0.54, "learning_rate": 9.055388066818128e-06, "loss": 1.9558, "step": 22079 }, { "epoch": 0.54, "learning_rate": 9.05459330840356e-06, "loss": 1.9325, "step": 22080 }, { "epoch": 0.54, "learning_rate": 9.05379855601438e-06, "loss": 1.9716, "step": 22081 }, { "epoch": 0.54, "learning_rate": 9.053003809655648e-06, "loss": 2.1201, "step": 22082 }, { "epoch": 0.54, "learning_rate": 9.052209069332431e-06, "loss": 1.9184, "step": 22083 }, { "epoch": 0.54, "learning_rate": 9.051414335049796e-06, "loss": 2.0973, "step": 22084 }, { "epoch": 0.54, "learning_rate": 9.050619606812804e-06, "loss": 2.0647, "step": 22085 }, { "epoch": 0.54, "learning_rate": 9.049824884626524e-06, "loss": 2.1274, "step": 22086 }, { "epoch": 0.54, "learning_rate": 9.049030168496019e-06, "loss": 2.0178, "step": 22087 }, { "epoch": 0.54, "learning_rate": 9.048235458426357e-06, "loss": 2.0423, "step": 22088 }, { "epoch": 0.54, "learning_rate": 9.047440754422599e-06, "loss": 2.049, "step": 22089 }, { "epoch": 0.54, "learning_rate": 9.04664605648981e-06, "loss": 1.9167, "step": 22090 }, { "epoch": 0.54, "learning_rate": 9.04585136463306e-06, "loss": 2.0943, "step": 22091 }, { "epoch": 0.54, "learning_rate": 9.045056678857407e-06, "loss": 2.1668, "step": 22092 }, { "epoch": 0.54, "learning_rate": 9.04426199916792e-06, "loss": 1.934, "step": 22093 }, { "epoch": 0.54, "learning_rate": 9.043467325569663e-06, "loss": 2.0185, "step": 22094 }, { "epoch": 0.54, "learning_rate": 9.042672658067699e-06, "loss": 2.0848, "step": 22095 }, { "epoch": 0.54, "learning_rate": 9.041877996667099e-06, "loss": 1.956, "step": 22096 }, { "epoch": 0.54, "learning_rate": 9.041083341372917e-06, "loss": 2.0542, "step": 22097 }, { "epoch": 0.54, "learning_rate": 9.040288692190228e-06, "loss": 2.0569, "step": 22098 }, { "epoch": 0.54, "learning_rate": 9.039494049124088e-06, "loss": 2.0473, "step": 22099 }, { "epoch": 0.54, "learning_rate": 9.038699412179565e-06, "loss": 2.0919, "step": 22100 }, { "epoch": 0.54, "learning_rate": 9.037904781361725e-06, "loss": 1.9517, "step": 22101 }, { "epoch": 0.54, "learning_rate": 9.037110156675634e-06, "loss": 2.2125, "step": 22102 }, { "epoch": 0.54, "learning_rate": 9.036315538126353e-06, "loss": 2.0579, "step": 22103 }, { "epoch": 0.54, "learning_rate": 9.035520925718945e-06, "loss": 1.9527, "step": 22104 }, { "epoch": 0.54, "learning_rate": 9.034726319458476e-06, "loss": 2.1111, "step": 22105 }, { "epoch": 0.54, "learning_rate": 9.033931719350015e-06, "loss": 2.0197, "step": 22106 }, { "epoch": 0.54, "learning_rate": 9.033137125398618e-06, "loss": 2.0196, "step": 22107 }, { "epoch": 0.54, "learning_rate": 9.032342537609358e-06, "loss": 2.0687, "step": 22108 }, { "epoch": 0.54, "learning_rate": 9.031547955987288e-06, "loss": 1.8546, "step": 22109 }, { "epoch": 0.54, "learning_rate": 9.03075338053748e-06, "loss": 1.9597, "step": 22110 }, { "epoch": 0.55, "learning_rate": 9.029958811264999e-06, "loss": 1.8718, "step": 22111 }, { "epoch": 0.55, "learning_rate": 9.029164248174904e-06, "loss": 2.0512, "step": 22112 }, { "epoch": 0.55, "learning_rate": 9.028369691272262e-06, "loss": 2.1682, "step": 22113 }, { "epoch": 0.55, "learning_rate": 9.02757514056214e-06, "loss": 1.8921, "step": 22114 }, { "epoch": 0.55, "learning_rate": 9.026780596049596e-06, "loss": 2.1321, "step": 22115 }, { "epoch": 0.55, "learning_rate": 9.025986057739698e-06, "loss": 2.1023, "step": 22116 }, { "epoch": 0.55, "learning_rate": 9.025191525637507e-06, "loss": 2.1152, "step": 22117 }, { "epoch": 0.55, "learning_rate": 9.024396999748087e-06, "loss": 2.1288, "step": 22118 }, { "epoch": 0.55, "learning_rate": 9.023602480076504e-06, "loss": 1.9775, "step": 22119 }, { "epoch": 0.55, "learning_rate": 9.02280796662782e-06, "loss": 2.0037, "step": 22120 }, { "epoch": 0.55, "learning_rate": 9.022013459407101e-06, "loss": 1.9036, "step": 22121 }, { "epoch": 0.55, "learning_rate": 9.021218958419406e-06, "loss": 2.1431, "step": 22122 }, { "epoch": 0.55, "learning_rate": 9.020424463669803e-06, "loss": 2.069, "step": 22123 }, { "epoch": 0.55, "learning_rate": 9.019629975163357e-06, "loss": 2.111, "step": 22124 }, { "epoch": 0.55, "learning_rate": 9.018835492905125e-06, "loss": 2.0686, "step": 22125 }, { "epoch": 0.55, "learning_rate": 9.018041016900177e-06, "loss": 1.8999, "step": 22126 }, { "epoch": 0.55, "learning_rate": 9.017246547153572e-06, "loss": 1.978, "step": 22127 }, { "epoch": 0.55, "learning_rate": 9.016452083670371e-06, "loss": 1.9437, "step": 22128 }, { "epoch": 0.55, "learning_rate": 9.015657626455647e-06, "loss": 1.8547, "step": 22129 }, { "epoch": 0.55, "learning_rate": 9.014863175514454e-06, "loss": 2.0936, "step": 22130 }, { "epoch": 0.55, "learning_rate": 9.014068730851865e-06, "loss": 2.0395, "step": 22131 }, { "epoch": 0.55, "learning_rate": 9.01327429247293e-06, "loss": 1.9434, "step": 22132 }, { "epoch": 0.55, "learning_rate": 9.012479860382722e-06, "loss": 1.9992, "step": 22133 }, { "epoch": 0.55, "learning_rate": 9.011685434586306e-06, "loss": 2.0217, "step": 22134 }, { "epoch": 0.55, "learning_rate": 9.010891015088736e-06, "loss": 2.1526, "step": 22135 }, { "epoch": 0.55, "learning_rate": 9.010096601895083e-06, "loss": 2.0399, "step": 22136 }, { "epoch": 0.55, "learning_rate": 9.009302195010404e-06, "loss": 1.9775, "step": 22137 }, { "epoch": 0.55, "learning_rate": 9.008507794439766e-06, "loss": 2.2846, "step": 22138 }, { "epoch": 0.55, "learning_rate": 9.007713400188232e-06, "loss": 2.0621, "step": 22139 }, { "epoch": 0.55, "learning_rate": 9.00691901226086e-06, "loss": 2.0623, "step": 22140 }, { "epoch": 0.55, "learning_rate": 9.006124630662723e-06, "loss": 1.9993, "step": 22141 }, { "epoch": 0.55, "learning_rate": 9.005330255398874e-06, "loss": 2.0219, "step": 22142 }, { "epoch": 0.55, "learning_rate": 9.004535886474376e-06, "loss": 1.8998, "step": 22143 }, { "epoch": 0.55, "learning_rate": 9.003741523894303e-06, "loss": 1.904, "step": 22144 }, { "epoch": 0.55, "learning_rate": 9.002947167663704e-06, "loss": 1.8753, "step": 22145 }, { "epoch": 0.55, "learning_rate": 9.002152817787651e-06, "loss": 2.019, "step": 22146 }, { "epoch": 0.55, "learning_rate": 9.0013584742712e-06, "loss": 2.2338, "step": 22147 }, { "epoch": 0.55, "learning_rate": 9.000564137119418e-06, "loss": 1.8766, "step": 22148 }, { "epoch": 0.55, "learning_rate": 8.999769806337371e-06, "loss": 2.0342, "step": 22149 }, { "epoch": 0.55, "learning_rate": 8.998975481930112e-06, "loss": 2.0191, "step": 22150 }, { "epoch": 0.55, "learning_rate": 8.998181163902711e-06, "loss": 1.9731, "step": 22151 }, { "epoch": 0.55, "learning_rate": 8.997386852260226e-06, "loss": 1.8548, "step": 22152 }, { "epoch": 0.55, "learning_rate": 8.99659254700772e-06, "loss": 1.907, "step": 22153 }, { "epoch": 0.55, "learning_rate": 8.99579824815026e-06, "loss": 1.8149, "step": 22154 }, { "epoch": 0.55, "learning_rate": 8.995003955692903e-06, "loss": 1.9715, "step": 22155 }, { "epoch": 0.55, "learning_rate": 8.994209669640715e-06, "loss": 1.9219, "step": 22156 }, { "epoch": 0.55, "learning_rate": 8.993415389998753e-06, "loss": 1.923, "step": 22157 }, { "epoch": 0.55, "learning_rate": 8.992621116772084e-06, "loss": 2.0119, "step": 22158 }, { "epoch": 0.55, "learning_rate": 8.991826849965774e-06, "loss": 1.9996, "step": 22159 }, { "epoch": 0.55, "learning_rate": 8.991032589584874e-06, "loss": 1.8477, "step": 22160 }, { "epoch": 0.55, "learning_rate": 8.990238335634459e-06, "loss": 2.2184, "step": 22161 }, { "epoch": 0.55, "learning_rate": 8.989444088119578e-06, "loss": 1.897, "step": 22162 }, { "epoch": 0.55, "learning_rate": 8.988649847045298e-06, "loss": 1.9003, "step": 22163 }, { "epoch": 0.55, "learning_rate": 8.987855612416685e-06, "loss": 1.9591, "step": 22164 }, { "epoch": 0.55, "learning_rate": 8.987061384238798e-06, "loss": 2.2015, "step": 22165 }, { "epoch": 0.55, "learning_rate": 8.9862671625167e-06, "loss": 2.0197, "step": 22166 }, { "epoch": 0.55, "learning_rate": 8.98547294725545e-06, "loss": 1.9251, "step": 22167 }, { "epoch": 0.55, "learning_rate": 8.98467873846011e-06, "loss": 1.7832, "step": 22168 }, { "epoch": 0.55, "learning_rate": 8.983884536135748e-06, "loss": 2.1735, "step": 22169 }, { "epoch": 0.55, "learning_rate": 8.983090340287417e-06, "loss": 2.0992, "step": 22170 }, { "epoch": 0.55, "learning_rate": 8.982296150920183e-06, "loss": 2.0253, "step": 22171 }, { "epoch": 0.55, "learning_rate": 8.98150196803911e-06, "loss": 2.0741, "step": 22172 }, { "epoch": 0.55, "learning_rate": 8.980707791649255e-06, "loss": 2.052, "step": 22173 }, { "epoch": 0.55, "learning_rate": 8.97991362175568e-06, "loss": 2.1541, "step": 22174 }, { "epoch": 0.55, "learning_rate": 8.97911945836345e-06, "loss": 2.0621, "step": 22175 }, { "epoch": 0.55, "learning_rate": 8.978325301477622e-06, "loss": 2.1319, "step": 22176 }, { "epoch": 0.55, "learning_rate": 8.977531151103264e-06, "loss": 1.8745, "step": 22177 }, { "epoch": 0.55, "learning_rate": 8.97673700724543e-06, "loss": 1.8738, "step": 22178 }, { "epoch": 0.55, "learning_rate": 8.975942869909188e-06, "loss": 2.1905, "step": 22179 }, { "epoch": 0.55, "learning_rate": 8.975148739099592e-06, "loss": 1.9268, "step": 22180 }, { "epoch": 0.55, "learning_rate": 8.974354614821707e-06, "loss": 1.9468, "step": 22181 }, { "epoch": 0.55, "learning_rate": 8.973560497080596e-06, "loss": 1.9997, "step": 22182 }, { "epoch": 0.55, "learning_rate": 8.972766385881316e-06, "loss": 1.9644, "step": 22183 }, { "epoch": 0.55, "learning_rate": 8.971972281228935e-06, "loss": 2.0103, "step": 22184 }, { "epoch": 0.55, "learning_rate": 8.971178183128506e-06, "loss": 2.1851, "step": 22185 }, { "epoch": 0.55, "learning_rate": 8.970384091585094e-06, "loss": 1.9422, "step": 22186 }, { "epoch": 0.55, "learning_rate": 8.969590006603763e-06, "loss": 1.8871, "step": 22187 }, { "epoch": 0.55, "learning_rate": 8.968795928189567e-06, "loss": 1.6821, "step": 22188 }, { "epoch": 0.55, "learning_rate": 8.968001856347573e-06, "loss": 2.1779, "step": 22189 }, { "epoch": 0.55, "learning_rate": 8.967207791082837e-06, "loss": 2.0757, "step": 22190 }, { "epoch": 0.55, "learning_rate": 8.966413732400426e-06, "loss": 2.0174, "step": 22191 }, { "epoch": 0.55, "learning_rate": 8.965619680305394e-06, "loss": 2.0143, "step": 22192 }, { "epoch": 0.55, "learning_rate": 8.964825634802806e-06, "loss": 2.0459, "step": 22193 }, { "epoch": 0.55, "learning_rate": 8.964031595897723e-06, "loss": 2.0362, "step": 22194 }, { "epoch": 0.55, "learning_rate": 8.963237563595202e-06, "loss": 2.141, "step": 22195 }, { "epoch": 0.55, "learning_rate": 8.962443537900304e-06, "loss": 1.7634, "step": 22196 }, { "epoch": 0.55, "learning_rate": 8.961649518818097e-06, "loss": 2.0525, "step": 22197 }, { "epoch": 0.55, "learning_rate": 8.960855506353632e-06, "loss": 1.821, "step": 22198 }, { "epoch": 0.55, "learning_rate": 8.960061500511974e-06, "loss": 2.0444, "step": 22199 }, { "epoch": 0.55, "learning_rate": 8.959267501298183e-06, "loss": 2.1528, "step": 22200 }, { "epoch": 0.55, "learning_rate": 8.958473508717317e-06, "loss": 1.9989, "step": 22201 }, { "epoch": 0.55, "learning_rate": 8.957679522774445e-06, "loss": 1.9505, "step": 22202 }, { "epoch": 0.55, "learning_rate": 8.956885543474617e-06, "loss": 2.0506, "step": 22203 }, { "epoch": 0.55, "learning_rate": 8.956091570822902e-06, "loss": 2.0943, "step": 22204 }, { "epoch": 0.55, "learning_rate": 8.95529760482435e-06, "loss": 1.9648, "step": 22205 }, { "epoch": 0.55, "learning_rate": 8.954503645484026e-06, "loss": 1.989, "step": 22206 }, { "epoch": 0.55, "learning_rate": 8.953709692806994e-06, "loss": 1.9787, "step": 22207 }, { "epoch": 0.55, "learning_rate": 8.95291574679831e-06, "loss": 2.0479, "step": 22208 }, { "epoch": 0.55, "learning_rate": 8.952121807463038e-06, "loss": 1.9998, "step": 22209 }, { "epoch": 0.55, "learning_rate": 8.951327874806232e-06, "loss": 1.8699, "step": 22210 }, { "epoch": 0.55, "learning_rate": 8.950533948832955e-06, "loss": 1.9665, "step": 22211 }, { "epoch": 0.55, "learning_rate": 8.949740029548271e-06, "loss": 2.0117, "step": 22212 }, { "epoch": 0.55, "learning_rate": 8.948946116957231e-06, "loss": 2.0046, "step": 22213 }, { "epoch": 0.55, "learning_rate": 8.948152211064905e-06, "loss": 1.9813, "step": 22214 }, { "epoch": 0.55, "learning_rate": 8.947358311876344e-06, "loss": 2.0021, "step": 22215 }, { "epoch": 0.55, "learning_rate": 8.946564419396612e-06, "loss": 2.0497, "step": 22216 }, { "epoch": 0.55, "learning_rate": 8.945770533630768e-06, "loss": 1.934, "step": 22217 }, { "epoch": 0.55, "learning_rate": 8.944976654583873e-06, "loss": 1.9893, "step": 22218 }, { "epoch": 0.55, "learning_rate": 8.944182782260987e-06, "loss": 1.9942, "step": 22219 }, { "epoch": 0.55, "learning_rate": 8.943388916667165e-06, "loss": 2.1083, "step": 22220 }, { "epoch": 0.55, "learning_rate": 8.94259505780747e-06, "loss": 2.0408, "step": 22221 }, { "epoch": 0.55, "learning_rate": 8.941801205686966e-06, "loss": 2.0007, "step": 22222 }, { "epoch": 0.55, "learning_rate": 8.941007360310703e-06, "loss": 2.0366, "step": 22223 }, { "epoch": 0.55, "learning_rate": 8.940213521683747e-06, "loss": 1.9861, "step": 22224 }, { "epoch": 0.55, "learning_rate": 8.939419689811154e-06, "loss": 2.1127, "step": 22225 }, { "epoch": 0.55, "learning_rate": 8.938625864697985e-06, "loss": 1.9741, "step": 22226 }, { "epoch": 0.55, "learning_rate": 8.937832046349304e-06, "loss": 1.9678, "step": 22227 }, { "epoch": 0.55, "learning_rate": 8.93703823477016e-06, "loss": 1.9877, "step": 22228 }, { "epoch": 0.55, "learning_rate": 8.936244429965618e-06, "loss": 1.9099, "step": 22229 }, { "epoch": 0.55, "learning_rate": 8.935450631940742e-06, "loss": 1.9603, "step": 22230 }, { "epoch": 0.55, "learning_rate": 8.93465684070058e-06, "loss": 2.0545, "step": 22231 }, { "epoch": 0.55, "learning_rate": 8.933863056250204e-06, "loss": 2.0652, "step": 22232 }, { "epoch": 0.55, "learning_rate": 8.93306927859466e-06, "loss": 1.8661, "step": 22233 }, { "epoch": 0.55, "learning_rate": 8.932275507739014e-06, "loss": 2.0817, "step": 22234 }, { "epoch": 0.55, "learning_rate": 8.931481743688326e-06, "loss": 2.0241, "step": 22235 }, { "epoch": 0.55, "learning_rate": 8.93068798644765e-06, "loss": 2.1529, "step": 22236 }, { "epoch": 0.55, "learning_rate": 8.929894236022054e-06, "loss": 1.9623, "step": 22237 }, { "epoch": 0.55, "learning_rate": 8.929100492416585e-06, "loss": 2.1301, "step": 22238 }, { "epoch": 0.55, "learning_rate": 8.928306755636309e-06, "loss": 2.0001, "step": 22239 }, { "epoch": 0.55, "learning_rate": 8.927513025686287e-06, "loss": 1.97, "step": 22240 }, { "epoch": 0.55, "learning_rate": 8.926719302571569e-06, "loss": 1.9691, "step": 22241 }, { "epoch": 0.55, "learning_rate": 8.925925586297221e-06, "loss": 2.1246, "step": 22242 }, { "epoch": 0.55, "learning_rate": 8.925131876868299e-06, "loss": 1.8805, "step": 22243 }, { "epoch": 0.55, "learning_rate": 8.924338174289862e-06, "loss": 2.0508, "step": 22244 }, { "epoch": 0.55, "learning_rate": 8.923544478566968e-06, "loss": 1.9752, "step": 22245 }, { "epoch": 0.55, "learning_rate": 8.922750789704675e-06, "loss": 2.079, "step": 22246 }, { "epoch": 0.55, "learning_rate": 8.921957107708047e-06, "loss": 1.9153, "step": 22247 }, { "epoch": 0.55, "learning_rate": 8.921163432582132e-06, "loss": 1.9332, "step": 22248 }, { "epoch": 0.55, "learning_rate": 8.920369764331995e-06, "loss": 2.0339, "step": 22249 }, { "epoch": 0.55, "learning_rate": 8.919576102962697e-06, "loss": 1.8402, "step": 22250 }, { "epoch": 0.55, "learning_rate": 8.918782448479288e-06, "loss": 1.9555, "step": 22251 }, { "epoch": 0.55, "learning_rate": 8.917988800886833e-06, "loss": 2.0393, "step": 22252 }, { "epoch": 0.55, "learning_rate": 8.917195160190386e-06, "loss": 1.9859, "step": 22253 }, { "epoch": 0.55, "learning_rate": 8.916401526395008e-06, "loss": 1.94, "step": 22254 }, { "epoch": 0.55, "learning_rate": 8.91560789950576e-06, "loss": 1.8226, "step": 22255 }, { "epoch": 0.55, "learning_rate": 8.914814279527691e-06, "loss": 2.0183, "step": 22256 }, { "epoch": 0.55, "learning_rate": 8.91402066646587e-06, "loss": 1.8831, "step": 22257 }, { "epoch": 0.55, "learning_rate": 8.913227060325344e-06, "loss": 1.9555, "step": 22258 }, { "epoch": 0.55, "learning_rate": 8.912433461111177e-06, "loss": 2.0732, "step": 22259 }, { "epoch": 0.55, "learning_rate": 8.911639868828426e-06, "loss": 2.0294, "step": 22260 }, { "epoch": 0.55, "learning_rate": 8.91084628348215e-06, "loss": 2.0759, "step": 22261 }, { "epoch": 0.55, "learning_rate": 8.910052705077408e-06, "loss": 1.8941, "step": 22262 }, { "epoch": 0.55, "learning_rate": 8.90925913361925e-06, "loss": 2.0717, "step": 22263 }, { "epoch": 0.55, "learning_rate": 8.90846556911274e-06, "loss": 1.9472, "step": 22264 }, { "epoch": 0.55, "learning_rate": 8.90767201156294e-06, "loss": 2.1019, "step": 22265 }, { "epoch": 0.55, "learning_rate": 8.906878460974896e-06, "loss": 1.9103, "step": 22266 }, { "epoch": 0.55, "learning_rate": 8.906084917353677e-06, "loss": 2.0223, "step": 22267 }, { "epoch": 0.55, "learning_rate": 8.905291380704331e-06, "loss": 2.1457, "step": 22268 }, { "epoch": 0.55, "learning_rate": 8.90449785103192e-06, "loss": 1.6634, "step": 22269 }, { "epoch": 0.55, "learning_rate": 8.903704328341504e-06, "loss": 1.8593, "step": 22270 }, { "epoch": 0.55, "learning_rate": 8.902910812638134e-06, "loss": 2.1124, "step": 22271 }, { "epoch": 0.55, "learning_rate": 8.902117303926876e-06, "loss": 2.0327, "step": 22272 }, { "epoch": 0.55, "learning_rate": 8.901323802212778e-06, "loss": 1.9877, "step": 22273 }, { "epoch": 0.55, "learning_rate": 8.900530307500901e-06, "loss": 1.958, "step": 22274 }, { "epoch": 0.55, "learning_rate": 8.899736819796307e-06, "loss": 2.0062, "step": 22275 }, { "epoch": 0.55, "learning_rate": 8.898943339104045e-06, "loss": 1.9546, "step": 22276 }, { "epoch": 0.55, "learning_rate": 8.898149865429178e-06, "loss": 2.1042, "step": 22277 }, { "epoch": 0.55, "learning_rate": 8.897356398776759e-06, "loss": 2.0858, "step": 22278 }, { "epoch": 0.55, "learning_rate": 8.896562939151846e-06, "loss": 2.1008, "step": 22279 }, { "epoch": 0.55, "learning_rate": 8.895769486559502e-06, "loss": 1.8297, "step": 22280 }, { "epoch": 0.55, "learning_rate": 8.894976041004775e-06, "loss": 2.155, "step": 22281 }, { "epoch": 0.55, "learning_rate": 8.894182602492729e-06, "loss": 2.1024, "step": 22282 }, { "epoch": 0.55, "learning_rate": 8.893389171028415e-06, "loss": 1.9971, "step": 22283 }, { "epoch": 0.55, "learning_rate": 8.892595746616891e-06, "loss": 1.9748, "step": 22284 }, { "epoch": 0.55, "learning_rate": 8.89180232926322e-06, "loss": 1.9653, "step": 22285 }, { "epoch": 0.55, "learning_rate": 8.89100891897245e-06, "loss": 1.9894, "step": 22286 }, { "epoch": 0.55, "learning_rate": 8.89021551574964e-06, "loss": 1.9006, "step": 22287 }, { "epoch": 0.55, "learning_rate": 8.889422119599851e-06, "loss": 2.0874, "step": 22288 }, { "epoch": 0.55, "learning_rate": 8.888628730528136e-06, "loss": 1.9863, "step": 22289 }, { "epoch": 0.55, "learning_rate": 8.887835348539555e-06, "loss": 1.8329, "step": 22290 }, { "epoch": 0.55, "learning_rate": 8.887041973639159e-06, "loss": 2.0922, "step": 22291 }, { "epoch": 0.55, "learning_rate": 8.886248605832005e-06, "loss": 1.9429, "step": 22292 }, { "epoch": 0.55, "learning_rate": 8.88545524512316e-06, "loss": 2.1551, "step": 22293 }, { "epoch": 0.55, "learning_rate": 8.884661891517664e-06, "loss": 1.9926, "step": 22294 }, { "epoch": 0.55, "learning_rate": 8.883868545020584e-06, "loss": 2.0496, "step": 22295 }, { "epoch": 0.55, "learning_rate": 8.883075205636972e-06, "loss": 2.1204, "step": 22296 }, { "epoch": 0.55, "learning_rate": 8.882281873371885e-06, "loss": 1.8755, "step": 22297 }, { "epoch": 0.55, "learning_rate": 8.881488548230384e-06, "loss": 1.9207, "step": 22298 }, { "epoch": 0.55, "learning_rate": 8.880695230217518e-06, "loss": 2.0272, "step": 22299 }, { "epoch": 0.55, "learning_rate": 8.87990191933835e-06, "loss": 1.9452, "step": 22300 }, { "epoch": 0.55, "learning_rate": 8.879108615597926e-06, "loss": 2.0981, "step": 22301 }, { "epoch": 0.55, "learning_rate": 8.87831531900131e-06, "loss": 1.9628, "step": 22302 }, { "epoch": 0.55, "learning_rate": 8.87752202955356e-06, "loss": 1.988, "step": 22303 }, { "epoch": 0.55, "learning_rate": 8.876728747259723e-06, "loss": 1.9908, "step": 22304 }, { "epoch": 0.55, "learning_rate": 8.875935472124862e-06, "loss": 2.0089, "step": 22305 }, { "epoch": 0.55, "learning_rate": 8.87514220415403e-06, "loss": 1.973, "step": 22306 }, { "epoch": 0.55, "learning_rate": 8.874348943352281e-06, "loss": 1.9986, "step": 22307 }, { "epoch": 0.55, "learning_rate": 8.87355568972468e-06, "loss": 2.0398, "step": 22308 }, { "epoch": 0.55, "learning_rate": 8.872762443276268e-06, "loss": 2.165, "step": 22309 }, { "epoch": 0.55, "learning_rate": 8.871969204012115e-06, "loss": 1.9694, "step": 22310 }, { "epoch": 0.55, "learning_rate": 8.871175971937265e-06, "loss": 1.9089, "step": 22311 }, { "epoch": 0.55, "learning_rate": 8.870382747056777e-06, "loss": 1.9932, "step": 22312 }, { "epoch": 0.55, "learning_rate": 8.869589529375711e-06, "loss": 1.8955, "step": 22313 }, { "epoch": 0.55, "learning_rate": 8.868796318899119e-06, "loss": 1.833, "step": 22314 }, { "epoch": 0.55, "learning_rate": 8.86800311563206e-06, "loss": 2.0566, "step": 22315 }, { "epoch": 0.55, "learning_rate": 8.86720991957958e-06, "loss": 1.9028, "step": 22316 }, { "epoch": 0.55, "learning_rate": 8.866416730746742e-06, "loss": 1.9997, "step": 22317 }, { "epoch": 0.55, "learning_rate": 8.865623549138603e-06, "loss": 2.0565, "step": 22318 }, { "epoch": 0.55, "learning_rate": 8.864830374760211e-06, "loss": 2.0678, "step": 22319 }, { "epoch": 0.55, "learning_rate": 8.864037207616628e-06, "loss": 1.9652, "step": 22320 }, { "epoch": 0.55, "learning_rate": 8.863244047712904e-06, "loss": 2.1145, "step": 22321 }, { "epoch": 0.55, "learning_rate": 8.862450895054096e-06, "loss": 1.999, "step": 22322 }, { "epoch": 0.55, "learning_rate": 8.86165774964526e-06, "loss": 2.0175, "step": 22323 }, { "epoch": 0.55, "learning_rate": 8.860864611491448e-06, "loss": 2.105, "step": 22324 }, { "epoch": 0.55, "learning_rate": 8.860071480597723e-06, "loss": 2.0549, "step": 22325 }, { "epoch": 0.55, "learning_rate": 8.859278356969128e-06, "loss": 1.9883, "step": 22326 }, { "epoch": 0.55, "learning_rate": 8.858485240610724e-06, "loss": 2.03, "step": 22327 }, { "epoch": 0.55, "learning_rate": 8.85769213152757e-06, "loss": 1.8865, "step": 22328 }, { "epoch": 0.55, "learning_rate": 8.856899029724714e-06, "loss": 1.9402, "step": 22329 }, { "epoch": 0.55, "learning_rate": 8.856105935207214e-06, "loss": 2.1069, "step": 22330 }, { "epoch": 0.55, "learning_rate": 8.855312847980119e-06, "loss": 2.0857, "step": 22331 }, { "epoch": 0.55, "learning_rate": 8.85451976804849e-06, "loss": 2.0089, "step": 22332 }, { "epoch": 0.55, "learning_rate": 8.853726695417384e-06, "loss": 1.9257, "step": 22333 }, { "epoch": 0.55, "learning_rate": 8.852933630091848e-06, "loss": 1.9876, "step": 22334 }, { "epoch": 0.55, "learning_rate": 8.852140572076944e-06, "loss": 1.752, "step": 22335 }, { "epoch": 0.55, "learning_rate": 8.851347521377717e-06, "loss": 1.9842, "step": 22336 }, { "epoch": 0.55, "learning_rate": 8.850554477999227e-06, "loss": 2.1137, "step": 22337 }, { "epoch": 0.55, "learning_rate": 8.84976144194653e-06, "loss": 1.8999, "step": 22338 }, { "epoch": 0.55, "learning_rate": 8.848968413224678e-06, "loss": 1.9845, "step": 22339 }, { "epoch": 0.55, "learning_rate": 8.848175391838725e-06, "loss": 1.9658, "step": 22340 }, { "epoch": 0.55, "learning_rate": 8.847382377793723e-06, "loss": 2.0198, "step": 22341 }, { "epoch": 0.55, "learning_rate": 8.84658937109473e-06, "loss": 2.0073, "step": 22342 }, { "epoch": 0.55, "learning_rate": 8.845796371746802e-06, "loss": 1.9448, "step": 22343 }, { "epoch": 0.55, "learning_rate": 8.845003379754988e-06, "loss": 2.0149, "step": 22344 }, { "epoch": 0.55, "learning_rate": 8.844210395124346e-06, "loss": 1.8694, "step": 22345 }, { "epoch": 0.55, "learning_rate": 8.843417417859924e-06, "loss": 1.9727, "step": 22346 }, { "epoch": 0.55, "learning_rate": 8.84262444796678e-06, "loss": 1.956, "step": 22347 }, { "epoch": 0.55, "learning_rate": 8.84183148544997e-06, "loss": 2.071, "step": 22348 }, { "epoch": 0.55, "learning_rate": 8.841038530314543e-06, "loss": 2.2396, "step": 22349 }, { "epoch": 0.55, "learning_rate": 8.840245582565555e-06, "loss": 1.9651, "step": 22350 }, { "epoch": 0.55, "learning_rate": 8.839452642208062e-06, "loss": 1.9624, "step": 22351 }, { "epoch": 0.55, "learning_rate": 8.838659709247114e-06, "loss": 2.0543, "step": 22352 }, { "epoch": 0.55, "learning_rate": 8.83786678368777e-06, "loss": 1.8795, "step": 22353 }, { "epoch": 0.55, "learning_rate": 8.837073865535076e-06, "loss": 2.1104, "step": 22354 }, { "epoch": 0.55, "learning_rate": 8.836280954794087e-06, "loss": 2.0292, "step": 22355 }, { "epoch": 0.55, "learning_rate": 8.835488051469865e-06, "loss": 2.0728, "step": 22356 }, { "epoch": 0.55, "learning_rate": 8.834695155567452e-06, "loss": 1.9754, "step": 22357 }, { "epoch": 0.55, "learning_rate": 8.83390226709191e-06, "loss": 2.1867, "step": 22358 }, { "epoch": 0.55, "learning_rate": 8.833109386048284e-06, "loss": 2.0681, "step": 22359 }, { "epoch": 0.55, "learning_rate": 8.832316512441634e-06, "loss": 1.9798, "step": 22360 }, { "epoch": 0.55, "learning_rate": 8.831523646277016e-06, "loss": 2.0394, "step": 22361 }, { "epoch": 0.55, "learning_rate": 8.830730787559475e-06, "loss": 2.1072, "step": 22362 }, { "epoch": 0.55, "learning_rate": 8.82993793629407e-06, "loss": 2.1141, "step": 22363 }, { "epoch": 0.55, "learning_rate": 8.829145092485848e-06, "loss": 1.9248, "step": 22364 }, { "epoch": 0.55, "learning_rate": 8.828352256139867e-06, "loss": 2.0516, "step": 22365 }, { "epoch": 0.55, "learning_rate": 8.82755942726118e-06, "loss": 1.9199, "step": 22366 }, { "epoch": 0.55, "learning_rate": 8.826766605854836e-06, "loss": 1.9089, "step": 22367 }, { "epoch": 0.55, "learning_rate": 8.825973791925898e-06, "loss": 2.0221, "step": 22368 }, { "epoch": 0.55, "learning_rate": 8.825180985479403e-06, "loss": 1.8541, "step": 22369 }, { "epoch": 0.55, "learning_rate": 8.824388186520416e-06, "loss": 2.0179, "step": 22370 }, { "epoch": 0.55, "learning_rate": 8.823595395053989e-06, "loss": 2.1065, "step": 22371 }, { "epoch": 0.55, "learning_rate": 8.822802611085169e-06, "loss": 2.0131, "step": 22372 }, { "epoch": 0.55, "learning_rate": 8.822009834619011e-06, "loss": 1.9929, "step": 22373 }, { "epoch": 0.55, "learning_rate": 8.82121706566057e-06, "loss": 1.9962, "step": 22374 }, { "epoch": 0.55, "learning_rate": 8.820424304214895e-06, "loss": 1.9489, "step": 22375 }, { "epoch": 0.55, "learning_rate": 8.819631550287042e-06, "loss": 2.0774, "step": 22376 }, { "epoch": 0.55, "learning_rate": 8.81883880388206e-06, "loss": 2.0982, "step": 22377 }, { "epoch": 0.55, "learning_rate": 8.818046065005008e-06, "loss": 2.0984, "step": 22378 }, { "epoch": 0.55, "learning_rate": 8.817253333660927e-06, "loss": 1.9655, "step": 22379 }, { "epoch": 0.55, "learning_rate": 8.816460609854879e-06, "loss": 1.9502, "step": 22380 }, { "epoch": 0.55, "learning_rate": 8.815667893591916e-06, "loss": 2.0075, "step": 22381 }, { "epoch": 0.55, "learning_rate": 8.814875184877083e-06, "loss": 1.9565, "step": 22382 }, { "epoch": 0.55, "learning_rate": 8.814082483715439e-06, "loss": 2.0827, "step": 22383 }, { "epoch": 0.55, "learning_rate": 8.813289790112032e-06, "loss": 2.0624, "step": 22384 }, { "epoch": 0.55, "learning_rate": 8.812497104071915e-06, "loss": 2.0511, "step": 22385 }, { "epoch": 0.55, "learning_rate": 8.811704425600146e-06, "loss": 1.8471, "step": 22386 }, { "epoch": 0.55, "learning_rate": 8.810911754701768e-06, "loss": 1.9666, "step": 22387 }, { "epoch": 0.55, "learning_rate": 8.81011909138184e-06, "loss": 2.1222, "step": 22388 }, { "epoch": 0.55, "learning_rate": 8.809326435645408e-06, "loss": 2.1055, "step": 22389 }, { "epoch": 0.55, "learning_rate": 8.808533787497525e-06, "loss": 2.0855, "step": 22390 }, { "epoch": 0.55, "learning_rate": 8.807741146943248e-06, "loss": 1.9821, "step": 22391 }, { "epoch": 0.55, "learning_rate": 8.806948513987624e-06, "loss": 1.8016, "step": 22392 }, { "epoch": 0.55, "learning_rate": 8.806155888635708e-06, "loss": 1.907, "step": 22393 }, { "epoch": 0.55, "learning_rate": 8.805363270892545e-06, "loss": 2.0083, "step": 22394 }, { "epoch": 0.55, "learning_rate": 8.804570660763192e-06, "loss": 2.074, "step": 22395 }, { "epoch": 0.55, "learning_rate": 8.803778058252706e-06, "loss": 1.9404, "step": 22396 }, { "epoch": 0.55, "learning_rate": 8.802985463366125e-06, "loss": 2.1267, "step": 22397 }, { "epoch": 0.55, "learning_rate": 8.802192876108514e-06, "loss": 1.9858, "step": 22398 }, { "epoch": 0.55, "learning_rate": 8.801400296484913e-06, "loss": 2.1204, "step": 22399 }, { "epoch": 0.55, "learning_rate": 8.800607724500378e-06, "loss": 1.9127, "step": 22400 }, { "epoch": 0.55, "learning_rate": 8.799815160159964e-06, "loss": 2.0102, "step": 22401 }, { "epoch": 0.55, "learning_rate": 8.799022603468717e-06, "loss": 2.0238, "step": 22402 }, { "epoch": 0.55, "learning_rate": 8.798230054431694e-06, "loss": 1.8818, "step": 22403 }, { "epoch": 0.55, "learning_rate": 8.797437513053939e-06, "loss": 2.0681, "step": 22404 }, { "epoch": 0.55, "learning_rate": 8.796644979340507e-06, "loss": 1.9643, "step": 22405 }, { "epoch": 0.55, "learning_rate": 8.79585245329645e-06, "loss": 2.103, "step": 22406 }, { "epoch": 0.55, "learning_rate": 8.795059934926816e-06, "loss": 2.1063, "step": 22407 }, { "epoch": 0.55, "learning_rate": 8.794267424236658e-06, "loss": 1.9257, "step": 22408 }, { "epoch": 0.55, "learning_rate": 8.793474921231028e-06, "loss": 1.9537, "step": 22409 }, { "epoch": 0.55, "learning_rate": 8.792682425914975e-06, "loss": 1.9006, "step": 22410 }, { "epoch": 0.55, "learning_rate": 8.791889938293552e-06, "loss": 1.9116, "step": 22411 }, { "epoch": 0.55, "learning_rate": 8.791097458371805e-06, "loss": 1.9333, "step": 22412 }, { "epoch": 0.55, "learning_rate": 8.790304986154787e-06, "loss": 1.9535, "step": 22413 }, { "epoch": 0.55, "learning_rate": 8.789512521647555e-06, "loss": 2.0652, "step": 22414 }, { "epoch": 0.55, "learning_rate": 8.78872006485515e-06, "loss": 2.0816, "step": 22415 }, { "epoch": 0.55, "learning_rate": 8.78792761578263e-06, "loss": 1.9107, "step": 22416 }, { "epoch": 0.55, "learning_rate": 8.787135174435038e-06, "loss": 1.9419, "step": 22417 }, { "epoch": 0.55, "learning_rate": 8.78634274081743e-06, "loss": 1.9497, "step": 22418 }, { "epoch": 0.55, "learning_rate": 8.785550314934857e-06, "loss": 1.9795, "step": 22419 }, { "epoch": 0.55, "learning_rate": 8.784757896792365e-06, "loss": 2.0032, "step": 22420 }, { "epoch": 0.55, "learning_rate": 8.783965486395013e-06, "loss": 2.1201, "step": 22421 }, { "epoch": 0.55, "learning_rate": 8.783173083747839e-06, "loss": 1.889, "step": 22422 }, { "epoch": 0.55, "learning_rate": 8.7823806888559e-06, "loss": 1.9135, "step": 22423 }, { "epoch": 0.55, "learning_rate": 8.781588301724248e-06, "loss": 2.086, "step": 22424 }, { "epoch": 0.55, "learning_rate": 8.780795922357929e-06, "loss": 1.9529, "step": 22425 }, { "epoch": 0.55, "learning_rate": 8.780003550761997e-06, "loss": 2.0983, "step": 22426 }, { "epoch": 0.55, "learning_rate": 8.779211186941497e-06, "loss": 1.987, "step": 22427 }, { "epoch": 0.55, "learning_rate": 8.778418830901486e-06, "loss": 1.7227, "step": 22428 }, { "epoch": 0.55, "learning_rate": 8.777626482647007e-06, "loss": 1.9517, "step": 22429 }, { "epoch": 0.55, "learning_rate": 8.776834142183113e-06, "loss": 1.9555, "step": 22430 }, { "epoch": 0.55, "learning_rate": 8.776041809514855e-06, "loss": 1.8771, "step": 22431 }, { "epoch": 0.55, "learning_rate": 8.77524948464728e-06, "loss": 2.0554, "step": 22432 }, { "epoch": 0.55, "learning_rate": 8.774457167585438e-06, "loss": 1.985, "step": 22433 }, { "epoch": 0.55, "learning_rate": 8.773664858334385e-06, "loss": 1.9623, "step": 22434 }, { "epoch": 0.55, "learning_rate": 8.77287255689916e-06, "loss": 1.9502, "step": 22435 }, { "epoch": 0.55, "learning_rate": 8.772080263284822e-06, "loss": 2.1865, "step": 22436 }, { "epoch": 0.55, "learning_rate": 8.771287977496412e-06, "loss": 2.0673, "step": 22437 }, { "epoch": 0.55, "learning_rate": 8.770495699538986e-06, "loss": 2.1457, "step": 22438 }, { "epoch": 0.55, "learning_rate": 8.769703429417595e-06, "loss": 2.0213, "step": 22439 }, { "epoch": 0.55, "learning_rate": 8.768911167137282e-06, "loss": 1.9705, "step": 22440 }, { "epoch": 0.55, "learning_rate": 8.768118912703103e-06, "loss": 2.0537, "step": 22441 }, { "epoch": 0.55, "learning_rate": 8.767326666120099e-06, "loss": 2.043, "step": 22442 }, { "epoch": 0.55, "learning_rate": 8.766534427393323e-06, "loss": 1.9337, "step": 22443 }, { "epoch": 0.55, "learning_rate": 8.765742196527829e-06, "loss": 1.9609, "step": 22444 }, { "epoch": 0.55, "learning_rate": 8.764949973528658e-06, "loss": 1.9818, "step": 22445 }, { "epoch": 0.55, "learning_rate": 8.764157758400869e-06, "loss": 1.891, "step": 22446 }, { "epoch": 0.55, "learning_rate": 8.763365551149501e-06, "loss": 2.1056, "step": 22447 }, { "epoch": 0.55, "learning_rate": 8.762573351779608e-06, "loss": 2.1517, "step": 22448 }, { "epoch": 0.55, "learning_rate": 8.761781160296241e-06, "loss": 2.0688, "step": 22449 }, { "epoch": 0.55, "learning_rate": 8.760988976704444e-06, "loss": 2.1542, "step": 22450 }, { "epoch": 0.55, "learning_rate": 8.76019680100927e-06, "loss": 2.0063, "step": 22451 }, { "epoch": 0.55, "learning_rate": 8.759404633215762e-06, "loss": 1.9715, "step": 22452 }, { "epoch": 0.55, "learning_rate": 8.758612473328974e-06, "loss": 1.9874, "step": 22453 }, { "epoch": 0.55, "learning_rate": 8.757820321353954e-06, "loss": 2.0724, "step": 22454 }, { "epoch": 0.55, "learning_rate": 8.757028177295749e-06, "loss": 2.1832, "step": 22455 }, { "epoch": 0.55, "learning_rate": 8.756236041159412e-06, "loss": 2.2018, "step": 22456 }, { "epoch": 0.55, "learning_rate": 8.755443912949983e-06, "loss": 2.0696, "step": 22457 }, { "epoch": 0.55, "learning_rate": 8.754651792672516e-06, "loss": 1.9868, "step": 22458 }, { "epoch": 0.55, "learning_rate": 8.753859680332064e-06, "loss": 1.9989, "step": 22459 }, { "epoch": 0.55, "learning_rate": 8.753067575933667e-06, "loss": 1.8274, "step": 22460 }, { "epoch": 0.55, "learning_rate": 8.752275479482376e-06, "loss": 2.2549, "step": 22461 }, { "epoch": 0.55, "learning_rate": 8.75148339098324e-06, "loss": 2.0921, "step": 22462 }, { "epoch": 0.55, "learning_rate": 8.750691310441308e-06, "loss": 1.9959, "step": 22463 }, { "epoch": 0.55, "learning_rate": 8.749899237861629e-06, "loss": 2.1223, "step": 22464 }, { "epoch": 0.55, "learning_rate": 8.749107173249246e-06, "loss": 1.8984, "step": 22465 }, { "epoch": 0.55, "learning_rate": 8.748315116609211e-06, "loss": 2.0352, "step": 22466 }, { "epoch": 0.55, "learning_rate": 8.747523067946576e-06, "loss": 2.1071, "step": 22467 }, { "epoch": 0.55, "learning_rate": 8.74673102726638e-06, "loss": 2.1026, "step": 22468 }, { "epoch": 0.55, "learning_rate": 8.745938994573682e-06, "loss": 1.9775, "step": 22469 }, { "epoch": 0.55, "learning_rate": 8.745146969873517e-06, "loss": 1.9967, "step": 22470 }, { "epoch": 0.55, "learning_rate": 8.744354953170939e-06, "loss": 2.0005, "step": 22471 }, { "epoch": 0.55, "learning_rate": 8.743562944471001e-06, "loss": 2.0583, "step": 22472 }, { "epoch": 0.55, "learning_rate": 8.742770943778741e-06, "loss": 2.0205, "step": 22473 }, { "epoch": 0.55, "learning_rate": 8.741978951099217e-06, "loss": 2.0439, "step": 22474 }, { "epoch": 0.55, "learning_rate": 8.741186966437466e-06, "loss": 1.9776, "step": 22475 }, { "epoch": 0.55, "learning_rate": 8.74039498979854e-06, "loss": 2.0276, "step": 22476 }, { "epoch": 0.55, "learning_rate": 8.739603021187494e-06, "loss": 2.0187, "step": 22477 }, { "epoch": 0.55, "learning_rate": 8.738811060609363e-06, "loss": 2.0232, "step": 22478 }, { "epoch": 0.55, "learning_rate": 8.738019108069202e-06, "loss": 2.0603, "step": 22479 }, { "epoch": 0.55, "learning_rate": 8.737227163572056e-06, "loss": 1.926, "step": 22480 }, { "epoch": 0.55, "learning_rate": 8.736435227122973e-06, "loss": 2.0535, "step": 22481 }, { "epoch": 0.55, "learning_rate": 8.735643298727e-06, "loss": 2.2352, "step": 22482 }, { "epoch": 0.55, "learning_rate": 8.734851378389183e-06, "loss": 2.0629, "step": 22483 }, { "epoch": 0.55, "learning_rate": 8.734059466114575e-06, "loss": 2.0666, "step": 22484 }, { "epoch": 0.55, "learning_rate": 8.733267561908213e-06, "loss": 1.8552, "step": 22485 }, { "epoch": 0.55, "learning_rate": 8.73247566577515e-06, "loss": 1.8696, "step": 22486 }, { "epoch": 0.55, "learning_rate": 8.73168377772044e-06, "loss": 1.9617, "step": 22487 }, { "epoch": 0.55, "learning_rate": 8.730891897749115e-06, "loss": 1.8478, "step": 22488 }, { "epoch": 0.55, "learning_rate": 8.730100025866231e-06, "loss": 1.9521, "step": 22489 }, { "epoch": 0.55, "learning_rate": 8.729308162076834e-06, "loss": 2.0461, "step": 22490 }, { "epoch": 0.55, "learning_rate": 8.728516306385968e-06, "loss": 2.1639, "step": 22491 }, { "epoch": 0.55, "learning_rate": 8.727724458798688e-06, "loss": 2.1018, "step": 22492 }, { "epoch": 0.55, "learning_rate": 8.726932619320029e-06, "loss": 2.281, "step": 22493 }, { "epoch": 0.55, "learning_rate": 8.726140787955049e-06, "loss": 1.8692, "step": 22494 }, { "epoch": 0.55, "learning_rate": 8.725348964708782e-06, "loss": 1.9795, "step": 22495 }, { "epoch": 0.55, "learning_rate": 8.724557149586283e-06, "loss": 1.9176, "step": 22496 }, { "epoch": 0.55, "learning_rate": 8.723765342592601e-06, "loss": 1.9016, "step": 22497 }, { "epoch": 0.55, "learning_rate": 8.722973543732774e-06, "loss": 2.0718, "step": 22498 }, { "epoch": 0.55, "learning_rate": 8.722181753011859e-06, "loss": 1.906, "step": 22499 }, { "epoch": 0.55, "learning_rate": 8.721389970434888e-06, "loss": 1.9957, "step": 22500 }, { "epoch": 0.55, "learning_rate": 8.72059819600692e-06, "loss": 1.9535, "step": 22501 }, { "epoch": 0.55, "learning_rate": 8.719806429732998e-06, "loss": 1.9525, "step": 22502 }, { "epoch": 0.55, "learning_rate": 8.719014671618162e-06, "loss": 1.973, "step": 22503 }, { "epoch": 0.55, "learning_rate": 8.718222921667469e-06, "loss": 1.9698, "step": 22504 }, { "epoch": 0.55, "learning_rate": 8.717431179885954e-06, "loss": 2.0015, "step": 22505 }, { "epoch": 0.55, "learning_rate": 8.716639446278669e-06, "loss": 1.8579, "step": 22506 }, { "epoch": 0.55, "learning_rate": 8.71584772085066e-06, "loss": 1.9205, "step": 22507 }, { "epoch": 0.55, "learning_rate": 8.71505600360697e-06, "loss": 1.8335, "step": 22508 }, { "epoch": 0.55, "learning_rate": 8.714264294552652e-06, "loss": 2.1485, "step": 22509 }, { "epoch": 0.55, "learning_rate": 8.713472593692742e-06, "loss": 2.1368, "step": 22510 }, { "epoch": 0.55, "learning_rate": 8.71268090103229e-06, "loss": 1.9914, "step": 22511 }, { "epoch": 0.55, "learning_rate": 8.711889216576347e-06, "loss": 2.1771, "step": 22512 }, { "epoch": 0.55, "learning_rate": 8.71109754032995e-06, "loss": 2.0399, "step": 22513 }, { "epoch": 0.55, "learning_rate": 8.710305872298148e-06, "loss": 2.0503, "step": 22514 }, { "epoch": 0.55, "learning_rate": 8.709514212485987e-06, "loss": 2.042, "step": 22515 }, { "epoch": 0.55, "learning_rate": 8.708722560898513e-06, "loss": 2.1045, "step": 22516 }, { "epoch": 0.56, "learning_rate": 8.707930917540773e-06, "loss": 2.0228, "step": 22517 }, { "epoch": 0.56, "learning_rate": 8.707139282417808e-06, "loss": 2.0857, "step": 22518 }, { "epoch": 0.56, "learning_rate": 8.70634765553467e-06, "loss": 1.7584, "step": 22519 }, { "epoch": 0.56, "learning_rate": 8.705556036896394e-06, "loss": 2.1278, "step": 22520 }, { "epoch": 0.56, "learning_rate": 8.704764426508033e-06, "loss": 2.1374, "step": 22521 }, { "epoch": 0.56, "learning_rate": 8.703972824374633e-06, "loss": 2.1017, "step": 22522 }, { "epoch": 0.56, "learning_rate": 8.703181230501232e-06, "loss": 2.0563, "step": 22523 }, { "epoch": 0.56, "learning_rate": 8.702389644892883e-06, "loss": 1.9119, "step": 22524 }, { "epoch": 0.56, "learning_rate": 8.701598067554626e-06, "loss": 1.8542, "step": 22525 }, { "epoch": 0.56, "learning_rate": 8.700806498491508e-06, "loss": 2.063, "step": 22526 }, { "epoch": 0.56, "learning_rate": 8.700014937708578e-06, "loss": 2.1244, "step": 22527 }, { "epoch": 0.56, "learning_rate": 8.699223385210871e-06, "loss": 2.026, "step": 22528 }, { "epoch": 0.56, "learning_rate": 8.698431841003436e-06, "loss": 2.1241, "step": 22529 }, { "epoch": 0.56, "learning_rate": 8.697640305091326e-06, "loss": 2.0146, "step": 22530 }, { "epoch": 0.56, "learning_rate": 8.696848777479573e-06, "loss": 2.0148, "step": 22531 }, { "epoch": 0.56, "learning_rate": 8.696057258173229e-06, "loss": 1.9622, "step": 22532 }, { "epoch": 0.56, "learning_rate": 8.695265747177336e-06, "loss": 2.0769, "step": 22533 }, { "epoch": 0.56, "learning_rate": 8.69447424449694e-06, "loss": 1.7188, "step": 22534 }, { "epoch": 0.56, "learning_rate": 8.69368275013709e-06, "loss": 1.9736, "step": 22535 }, { "epoch": 0.56, "learning_rate": 8.692891264102818e-06, "loss": 1.9736, "step": 22536 }, { "epoch": 0.56, "learning_rate": 8.692099786399183e-06, "loss": 2.0706, "step": 22537 }, { "epoch": 0.56, "learning_rate": 8.691308317031218e-06, "loss": 2.0299, "step": 22538 }, { "epoch": 0.56, "learning_rate": 8.69051685600397e-06, "loss": 1.9909, "step": 22539 }, { "epoch": 0.56, "learning_rate": 8.689725403322489e-06, "loss": 2.1137, "step": 22540 }, { "epoch": 0.56, "learning_rate": 8.688933958991812e-06, "loss": 1.8888, "step": 22541 }, { "epoch": 0.56, "learning_rate": 8.688142523016987e-06, "loss": 2.1147, "step": 22542 }, { "epoch": 0.56, "learning_rate": 8.687351095403057e-06, "loss": 1.9593, "step": 22543 }, { "epoch": 0.56, "learning_rate": 8.686559676155065e-06, "loss": 1.9325, "step": 22544 }, { "epoch": 0.56, "learning_rate": 8.685768265278061e-06, "loss": 1.9568, "step": 22545 }, { "epoch": 0.56, "learning_rate": 8.68497686277708e-06, "loss": 2.0553, "step": 22546 }, { "epoch": 0.56, "learning_rate": 8.684185468657173e-06, "loss": 2.028, "step": 22547 }, { "epoch": 0.56, "learning_rate": 8.68339408292338e-06, "loss": 2.2524, "step": 22548 }, { "epoch": 0.56, "learning_rate": 8.682602705580742e-06, "loss": 2.1162, "step": 22549 }, { "epoch": 0.56, "learning_rate": 8.68181133663431e-06, "loss": 1.839, "step": 22550 }, { "epoch": 0.56, "learning_rate": 8.68101997608912e-06, "loss": 2.1585, "step": 22551 }, { "epoch": 0.56, "learning_rate": 8.680228623950226e-06, "loss": 1.9573, "step": 22552 }, { "epoch": 0.56, "learning_rate": 8.67943728022266e-06, "loss": 1.7243, "step": 22553 }, { "epoch": 0.56, "learning_rate": 8.67864594491147e-06, "loss": 2.0517, "step": 22554 }, { "epoch": 0.56, "learning_rate": 8.677854618021704e-06, "loss": 1.9739, "step": 22555 }, { "epoch": 0.56, "learning_rate": 8.677063299558398e-06, "loss": 2.1299, "step": 22556 }, { "epoch": 0.56, "learning_rate": 8.676271989526604e-06, "loss": 2.1009, "step": 22557 }, { "epoch": 0.56, "learning_rate": 8.675480687931354e-06, "loss": 2.0344, "step": 22558 }, { "epoch": 0.56, "learning_rate": 8.674689394777698e-06, "loss": 1.9534, "step": 22559 }, { "epoch": 0.56, "learning_rate": 8.67389811007068e-06, "loss": 1.9213, "step": 22560 }, { "epoch": 0.56, "learning_rate": 8.673106833815338e-06, "loss": 1.9609, "step": 22561 }, { "epoch": 0.56, "learning_rate": 8.672315566016727e-06, "loss": 1.9923, "step": 22562 }, { "epoch": 0.56, "learning_rate": 8.671524306679873e-06, "loss": 1.9112, "step": 22563 }, { "epoch": 0.56, "learning_rate": 8.67073305580983e-06, "loss": 1.8273, "step": 22564 }, { "epoch": 0.56, "learning_rate": 8.669941813411642e-06, "loss": 1.9543, "step": 22565 }, { "epoch": 0.56, "learning_rate": 8.669150579490343e-06, "loss": 1.9981, "step": 22566 }, { "epoch": 0.56, "learning_rate": 8.668359354050986e-06, "loss": 2.1696, "step": 22567 }, { "epoch": 0.56, "learning_rate": 8.667568137098605e-06, "loss": 1.9237, "step": 22568 }, { "epoch": 0.56, "learning_rate": 8.666776928638247e-06, "loss": 2.0164, "step": 22569 }, { "epoch": 0.56, "learning_rate": 8.665985728674957e-06, "loss": 1.9207, "step": 22570 }, { "epoch": 0.56, "learning_rate": 8.665194537213772e-06, "loss": 2.1307, "step": 22571 }, { "epoch": 0.56, "learning_rate": 8.66440335425974e-06, "loss": 2.1458, "step": 22572 }, { "epoch": 0.56, "learning_rate": 8.663612179817897e-06, "loss": 1.8761, "step": 22573 }, { "epoch": 0.56, "learning_rate": 8.66282101389329e-06, "loss": 1.8861, "step": 22574 }, { "epoch": 0.56, "learning_rate": 8.662029856490963e-06, "loss": 1.9664, "step": 22575 }, { "epoch": 0.56, "learning_rate": 8.661238707615954e-06, "loss": 1.9333, "step": 22576 }, { "epoch": 0.56, "learning_rate": 8.660447567273307e-06, "loss": 1.9993, "step": 22577 }, { "epoch": 0.56, "learning_rate": 8.659656435468062e-06, "loss": 2.0221, "step": 22578 }, { "epoch": 0.56, "learning_rate": 8.658865312205264e-06, "loss": 1.8865, "step": 22579 }, { "epoch": 0.56, "learning_rate": 8.658074197489958e-06, "loss": 2.0785, "step": 22580 }, { "epoch": 0.56, "learning_rate": 8.65728309132718e-06, "loss": 2.046, "step": 22581 }, { "epoch": 0.56, "learning_rate": 8.656491993721979e-06, "loss": 1.9263, "step": 22582 }, { "epoch": 0.56, "learning_rate": 8.655700904679385e-06, "loss": 2.0184, "step": 22583 }, { "epoch": 0.56, "learning_rate": 8.65490982420445e-06, "loss": 2.0272, "step": 22584 }, { "epoch": 0.56, "learning_rate": 8.654118752302214e-06, "loss": 2.0251, "step": 22585 }, { "epoch": 0.56, "learning_rate": 8.653327688977715e-06, "loss": 2.0934, "step": 22586 }, { "epoch": 0.56, "learning_rate": 8.652536634235998e-06, "loss": 2.0177, "step": 22587 }, { "epoch": 0.56, "learning_rate": 8.651745588082109e-06, "loss": 2.1729, "step": 22588 }, { "epoch": 0.56, "learning_rate": 8.65095455052108e-06, "loss": 2.0212, "step": 22589 }, { "epoch": 0.56, "learning_rate": 8.650163521557962e-06, "loss": 2.0433, "step": 22590 }, { "epoch": 0.56, "learning_rate": 8.649372501197788e-06, "loss": 1.9538, "step": 22591 }, { "epoch": 0.56, "learning_rate": 8.648581489445602e-06, "loss": 2.034, "step": 22592 }, { "epoch": 0.56, "learning_rate": 8.647790486306451e-06, "loss": 2.0756, "step": 22593 }, { "epoch": 0.56, "learning_rate": 8.646999491785368e-06, "loss": 2.112, "step": 22594 }, { "epoch": 0.56, "learning_rate": 8.6462085058874e-06, "loss": 1.953, "step": 22595 }, { "epoch": 0.56, "learning_rate": 8.645417528617584e-06, "loss": 1.9082, "step": 22596 }, { "epoch": 0.56, "learning_rate": 8.644626559980964e-06, "loss": 1.8941, "step": 22597 }, { "epoch": 0.56, "learning_rate": 8.643835599982586e-06, "loss": 1.7782, "step": 22598 }, { "epoch": 0.56, "learning_rate": 8.64304464862748e-06, "loss": 1.9499, "step": 22599 }, { "epoch": 0.56, "learning_rate": 8.642253705920697e-06, "loss": 1.9549, "step": 22600 }, { "epoch": 0.56, "learning_rate": 8.64146277186727e-06, "loss": 1.8251, "step": 22601 }, { "epoch": 0.56, "learning_rate": 8.640671846472244e-06, "loss": 1.9232, "step": 22602 }, { "epoch": 0.56, "learning_rate": 8.63988092974066e-06, "loss": 1.8646, "step": 22603 }, { "epoch": 0.56, "learning_rate": 8.639090021677556e-06, "loss": 2.1913, "step": 22604 }, { "epoch": 0.56, "learning_rate": 8.63829912228798e-06, "loss": 2.063, "step": 22605 }, { "epoch": 0.56, "learning_rate": 8.637508231576963e-06, "loss": 1.898, "step": 22606 }, { "epoch": 0.56, "learning_rate": 8.63671734954955e-06, "loss": 1.8514, "step": 22607 }, { "epoch": 0.56, "learning_rate": 8.635926476210786e-06, "loss": 2.1709, "step": 22608 }, { "epoch": 0.56, "learning_rate": 8.635135611565701e-06, "loss": 1.9368, "step": 22609 }, { "epoch": 0.56, "learning_rate": 8.634344755619346e-06, "loss": 2.0517, "step": 22610 }, { "epoch": 0.56, "learning_rate": 8.633553908376757e-06, "loss": 1.9973, "step": 22611 }, { "epoch": 0.56, "learning_rate": 8.632763069842971e-06, "loss": 2.0859, "step": 22612 }, { "epoch": 0.56, "learning_rate": 8.631972240023035e-06, "loss": 2.0185, "step": 22613 }, { "epoch": 0.56, "learning_rate": 8.631181418921982e-06, "loss": 1.9227, "step": 22614 }, { "epoch": 0.56, "learning_rate": 8.630390606544861e-06, "loss": 2.1066, "step": 22615 }, { "epoch": 0.56, "learning_rate": 8.629599802896702e-06, "loss": 1.9909, "step": 22616 }, { "epoch": 0.56, "learning_rate": 8.628809007982551e-06, "loss": 2.0005, "step": 22617 }, { "epoch": 0.56, "learning_rate": 8.628018221807451e-06, "loss": 2.1302, "step": 22618 }, { "epoch": 0.56, "learning_rate": 8.627227444376432e-06, "loss": 2.2242, "step": 22619 }, { "epoch": 0.56, "learning_rate": 8.626436675694544e-06, "loss": 2.1916, "step": 22620 }, { "epoch": 0.56, "learning_rate": 8.62564591576682e-06, "loss": 1.9264, "step": 22621 }, { "epoch": 0.56, "learning_rate": 8.624855164598303e-06, "loss": 1.8606, "step": 22622 }, { "epoch": 0.56, "learning_rate": 8.624064422194036e-06, "loss": 1.7952, "step": 22623 }, { "epoch": 0.56, "learning_rate": 8.623273688559051e-06, "loss": 2.0438, "step": 22624 }, { "epoch": 0.56, "learning_rate": 8.622482963698395e-06, "loss": 2.0941, "step": 22625 }, { "epoch": 0.56, "learning_rate": 8.6216922476171e-06, "loss": 1.9888, "step": 22626 }, { "epoch": 0.56, "learning_rate": 8.62090154032021e-06, "loss": 1.8135, "step": 22627 }, { "epoch": 0.56, "learning_rate": 8.620110841812766e-06, "loss": 2.2287, "step": 22628 }, { "epoch": 0.56, "learning_rate": 8.619320152099804e-06, "loss": 1.9851, "step": 22629 }, { "epoch": 0.56, "learning_rate": 8.618529471186365e-06, "loss": 1.9301, "step": 22630 }, { "epoch": 0.56, "learning_rate": 8.617738799077487e-06, "loss": 1.8732, "step": 22631 }, { "epoch": 0.56, "learning_rate": 8.61694813577821e-06, "loss": 2.0136, "step": 22632 }, { "epoch": 0.56, "learning_rate": 8.616157481293577e-06, "loss": 1.9544, "step": 22633 }, { "epoch": 0.56, "learning_rate": 8.615366835628619e-06, "loss": 1.9451, "step": 22634 }, { "epoch": 0.56, "learning_rate": 8.614576198788382e-06, "loss": 1.8601, "step": 22635 }, { "epoch": 0.56, "learning_rate": 8.613785570777901e-06, "loss": 2.0834, "step": 22636 }, { "epoch": 0.56, "learning_rate": 8.612994951602215e-06, "loss": 1.9666, "step": 22637 }, { "epoch": 0.56, "learning_rate": 8.612204341266365e-06, "loss": 1.8917, "step": 22638 }, { "epoch": 0.56, "learning_rate": 8.611413739775389e-06, "loss": 2.0418, "step": 22639 }, { "epoch": 0.56, "learning_rate": 8.610623147134329e-06, "loss": 2.0736, "step": 22640 }, { "epoch": 0.56, "learning_rate": 8.609832563348218e-06, "loss": 2.0903, "step": 22641 }, { "epoch": 0.56, "learning_rate": 8.609041988422095e-06, "loss": 1.9546, "step": 22642 }, { "epoch": 0.56, "learning_rate": 8.608251422361005e-06, "loss": 1.928, "step": 22643 }, { "epoch": 0.56, "learning_rate": 8.607460865169978e-06, "loss": 2.1187, "step": 22644 }, { "epoch": 0.56, "learning_rate": 8.606670316854058e-06, "loss": 2.1269, "step": 22645 }, { "epoch": 0.56, "learning_rate": 8.605879777418283e-06, "loss": 1.9685, "step": 22646 }, { "epoch": 0.56, "learning_rate": 8.605089246867691e-06, "loss": 1.8613, "step": 22647 }, { "epoch": 0.56, "learning_rate": 8.60429872520732e-06, "loss": 1.9077, "step": 22648 }, { "epoch": 0.56, "learning_rate": 8.603508212442206e-06, "loss": 2.0357, "step": 22649 }, { "epoch": 0.56, "learning_rate": 8.602717708577389e-06, "loss": 2.0299, "step": 22650 }, { "epoch": 0.56, "learning_rate": 8.601927213617911e-06, "loss": 1.9227, "step": 22651 }, { "epoch": 0.56, "learning_rate": 8.601136727568804e-06, "loss": 2.0577, "step": 22652 }, { "epoch": 0.56, "learning_rate": 8.600346250435112e-06, "loss": 1.9707, "step": 22653 }, { "epoch": 0.56, "learning_rate": 8.599555782221864e-06, "loss": 1.9751, "step": 22654 }, { "epoch": 0.56, "learning_rate": 8.598765322934105e-06, "loss": 2.0939, "step": 22655 }, { "epoch": 0.56, "learning_rate": 8.597974872576873e-06, "loss": 2.025, "step": 22656 }, { "epoch": 0.56, "learning_rate": 8.5971844311552e-06, "loss": 1.9659, "step": 22657 }, { "epoch": 0.56, "learning_rate": 8.596393998674136e-06, "loss": 2.0035, "step": 22658 }, { "epoch": 0.56, "learning_rate": 8.595603575138703e-06, "loss": 2.0305, "step": 22659 }, { "epoch": 0.56, "learning_rate": 8.594813160553948e-06, "loss": 2.2, "step": 22660 }, { "epoch": 0.56, "learning_rate": 8.594022754924909e-06, "loss": 2.0244, "step": 22661 }, { "epoch": 0.56, "learning_rate": 8.593232358256617e-06, "loss": 2.0126, "step": 22662 }, { "epoch": 0.56, "learning_rate": 8.592441970554117e-06, "loss": 2.0296, "step": 22663 }, { "epoch": 0.56, "learning_rate": 8.591651591822442e-06, "loss": 2.0349, "step": 22664 }, { "epoch": 0.56, "learning_rate": 8.59086122206663e-06, "loss": 1.9327, "step": 22665 }, { "epoch": 0.56, "learning_rate": 8.59007086129172e-06, "loss": 2.0581, "step": 22666 }, { "epoch": 0.56, "learning_rate": 8.589280509502745e-06, "loss": 1.9106, "step": 22667 }, { "epoch": 0.56, "learning_rate": 8.588490166704751e-06, "loss": 2.1072, "step": 22668 }, { "epoch": 0.56, "learning_rate": 8.587699832902762e-06, "loss": 1.8324, "step": 22669 }, { "epoch": 0.56, "learning_rate": 8.586909508101824e-06, "loss": 2.0408, "step": 22670 }, { "epoch": 0.56, "learning_rate": 8.586119192306977e-06, "loss": 1.9302, "step": 22671 }, { "epoch": 0.56, "learning_rate": 8.585328885523248e-06, "loss": 1.9356, "step": 22672 }, { "epoch": 0.56, "learning_rate": 8.584538587755682e-06, "loss": 2.002, "step": 22673 }, { "epoch": 0.56, "learning_rate": 8.583748299009309e-06, "loss": 1.9033, "step": 22674 }, { "epoch": 0.56, "learning_rate": 8.582958019289172e-06, "loss": 1.9625, "step": 22675 }, { "epoch": 0.56, "learning_rate": 8.582167748600308e-06, "loss": 1.9898, "step": 22676 }, { "epoch": 0.56, "learning_rate": 8.581377486947746e-06, "loss": 1.9582, "step": 22677 }, { "epoch": 0.56, "learning_rate": 8.580587234336534e-06, "loss": 2.0406, "step": 22678 }, { "epoch": 0.56, "learning_rate": 8.579796990771697e-06, "loss": 2.012, "step": 22679 }, { "epoch": 0.56, "learning_rate": 8.579006756258276e-06, "loss": 2.0831, "step": 22680 }, { "epoch": 0.56, "learning_rate": 8.578216530801312e-06, "loss": 2.1522, "step": 22681 }, { "epoch": 0.56, "learning_rate": 8.577426314405834e-06, "loss": 1.9878, "step": 22682 }, { "epoch": 0.56, "learning_rate": 8.576636107076884e-06, "loss": 2.0079, "step": 22683 }, { "epoch": 0.56, "learning_rate": 8.575845908819493e-06, "loss": 2.0328, "step": 22684 }, { "epoch": 0.56, "learning_rate": 8.575055719638701e-06, "loss": 1.9318, "step": 22685 }, { "epoch": 0.56, "learning_rate": 8.574265539539548e-06, "loss": 1.8653, "step": 22686 }, { "epoch": 0.56, "learning_rate": 8.57347536852706e-06, "loss": 1.959, "step": 22687 }, { "epoch": 0.56, "learning_rate": 8.572685206606283e-06, "loss": 2.031, "step": 22688 }, { "epoch": 0.56, "learning_rate": 8.571895053782245e-06, "loss": 1.9572, "step": 22689 }, { "epoch": 0.56, "learning_rate": 8.571104910059984e-06, "loss": 2.0828, "step": 22690 }, { "epoch": 0.56, "learning_rate": 8.57031477544454e-06, "loss": 2.0224, "step": 22691 }, { "epoch": 0.56, "learning_rate": 8.56952464994094e-06, "loss": 2.1954, "step": 22692 }, { "epoch": 0.56, "learning_rate": 8.568734533554235e-06, "loss": 2.2163, "step": 22693 }, { "epoch": 0.56, "learning_rate": 8.567944426289444e-06, "loss": 1.979, "step": 22694 }, { "epoch": 0.56, "learning_rate": 8.567154328151612e-06, "loss": 2.0634, "step": 22695 }, { "epoch": 0.56, "learning_rate": 8.566364239145775e-06, "loss": 2.0116, "step": 22696 }, { "epoch": 0.56, "learning_rate": 8.565574159276962e-06, "loss": 1.9798, "step": 22697 }, { "epoch": 0.56, "learning_rate": 8.564784088550216e-06, "loss": 1.9548, "step": 22698 }, { "epoch": 0.56, "learning_rate": 8.563994026970564e-06, "loss": 2.1568, "step": 22699 }, { "epoch": 0.56, "learning_rate": 8.563203974543049e-06, "loss": 2.0953, "step": 22700 }, { "epoch": 0.56, "learning_rate": 8.562413931272703e-06, "loss": 1.9242, "step": 22701 }, { "epoch": 0.56, "learning_rate": 8.56162389716456e-06, "loss": 2.0539, "step": 22702 }, { "epoch": 0.56, "learning_rate": 8.560833872223658e-06, "loss": 1.9018, "step": 22703 }, { "epoch": 0.56, "learning_rate": 8.560043856455034e-06, "loss": 1.9399, "step": 22704 }, { "epoch": 0.56, "learning_rate": 8.559253849863715e-06, "loss": 1.9221, "step": 22705 }, { "epoch": 0.56, "learning_rate": 8.558463852454746e-06, "loss": 2.2671, "step": 22706 }, { "epoch": 0.56, "learning_rate": 8.557673864233151e-06, "loss": 1.9045, "step": 22707 }, { "epoch": 0.56, "learning_rate": 8.556883885203972e-06, "loss": 2.0466, "step": 22708 }, { "epoch": 0.56, "learning_rate": 8.556093915372243e-06, "loss": 1.9246, "step": 22709 }, { "epoch": 0.56, "learning_rate": 8.555303954742996e-06, "loss": 1.9371, "step": 22710 }, { "epoch": 0.56, "learning_rate": 8.554514003321273e-06, "loss": 2.0668, "step": 22711 }, { "epoch": 0.56, "learning_rate": 8.553724061112099e-06, "loss": 2.0093, "step": 22712 }, { "epoch": 0.56, "learning_rate": 8.552934128120512e-06, "loss": 1.9369, "step": 22713 }, { "epoch": 0.56, "learning_rate": 8.552144204351553e-06, "loss": 1.9571, "step": 22714 }, { "epoch": 0.56, "learning_rate": 8.551354289810246e-06, "loss": 1.8724, "step": 22715 }, { "epoch": 0.56, "learning_rate": 8.55056438450163e-06, "loss": 1.977, "step": 22716 }, { "epoch": 0.56, "learning_rate": 8.54977448843074e-06, "loss": 2.1599, "step": 22717 }, { "epoch": 0.56, "learning_rate": 8.548984601602612e-06, "loss": 1.9843, "step": 22718 }, { "epoch": 0.56, "learning_rate": 8.548194724022277e-06, "loss": 1.9076, "step": 22719 }, { "epoch": 0.56, "learning_rate": 8.547404855694767e-06, "loss": 2.0797, "step": 22720 }, { "epoch": 0.56, "learning_rate": 8.546614996625126e-06, "loss": 1.7896, "step": 22721 }, { "epoch": 0.56, "learning_rate": 8.545825146818376e-06, "loss": 2.0912, "step": 22722 }, { "epoch": 0.56, "learning_rate": 8.545035306279556e-06, "loss": 2.1715, "step": 22723 }, { "epoch": 0.56, "learning_rate": 8.544245475013705e-06, "loss": 1.8193, "step": 22724 }, { "epoch": 0.56, "learning_rate": 8.543455653025847e-06, "loss": 1.9353, "step": 22725 }, { "epoch": 0.56, "learning_rate": 8.542665840321022e-06, "loss": 2.1377, "step": 22726 }, { "epoch": 0.56, "learning_rate": 8.541876036904261e-06, "loss": 2.073, "step": 22727 }, { "epoch": 0.56, "learning_rate": 8.5410862427806e-06, "loss": 2.0348, "step": 22728 }, { "epoch": 0.56, "learning_rate": 8.540296457955077e-06, "loss": 1.9359, "step": 22729 }, { "epoch": 0.56, "learning_rate": 8.539506682432716e-06, "loss": 2.1315, "step": 22730 }, { "epoch": 0.56, "learning_rate": 8.538716916218558e-06, "loss": 2.0486, "step": 22731 }, { "epoch": 0.56, "learning_rate": 8.537927159317628e-06, "loss": 2.1026, "step": 22732 }, { "epoch": 0.56, "learning_rate": 8.537137411734967e-06, "loss": 1.9409, "step": 22733 }, { "epoch": 0.56, "learning_rate": 8.536347673475607e-06, "loss": 1.9893, "step": 22734 }, { "epoch": 0.56, "learning_rate": 8.535557944544579e-06, "loss": 2.0661, "step": 22735 }, { "epoch": 0.56, "learning_rate": 8.53476822494692e-06, "loss": 2.1568, "step": 22736 }, { "epoch": 0.56, "learning_rate": 8.533978514687657e-06, "loss": 2.0324, "step": 22737 }, { "epoch": 0.56, "learning_rate": 8.533188813771827e-06, "loss": 2.0834, "step": 22738 }, { "epoch": 0.56, "learning_rate": 8.532399122204466e-06, "loss": 1.9376, "step": 22739 }, { "epoch": 0.56, "learning_rate": 8.5316094399906e-06, "loss": 2.0577, "step": 22740 }, { "epoch": 0.56, "learning_rate": 8.53081976713527e-06, "loss": 1.8716, "step": 22741 }, { "epoch": 0.56, "learning_rate": 8.5300301036435e-06, "loss": 2.1971, "step": 22742 }, { "epoch": 0.56, "learning_rate": 8.529240449520328e-06, "loss": 1.9073, "step": 22743 }, { "epoch": 0.56, "learning_rate": 8.528450804770789e-06, "loss": 1.9513, "step": 22744 }, { "epoch": 0.56, "learning_rate": 8.527661169399908e-06, "loss": 2.0676, "step": 22745 }, { "epoch": 0.56, "learning_rate": 8.526871543412727e-06, "loss": 2.0589, "step": 22746 }, { "epoch": 0.56, "learning_rate": 8.526081926814272e-06, "loss": 2.1276, "step": 22747 }, { "epoch": 0.56, "learning_rate": 8.525292319609574e-06, "loss": 1.9497, "step": 22748 }, { "epoch": 0.56, "learning_rate": 8.524502721803673e-06, "loss": 1.784, "step": 22749 }, { "epoch": 0.56, "learning_rate": 8.523713133401594e-06, "loss": 2.0256, "step": 22750 }, { "epoch": 0.56, "learning_rate": 8.522923554408376e-06, "loss": 2.0551, "step": 22751 }, { "epoch": 0.56, "learning_rate": 8.522133984829043e-06, "loss": 2.1154, "step": 22752 }, { "epoch": 0.56, "learning_rate": 8.521344424668635e-06, "loss": 2.0756, "step": 22753 }, { "epoch": 0.56, "learning_rate": 8.52055487393218e-06, "loss": 1.9794, "step": 22754 }, { "epoch": 0.56, "learning_rate": 8.519765332624709e-06, "loss": 1.9951, "step": 22755 }, { "epoch": 0.56, "learning_rate": 8.518975800751261e-06, "loss": 1.911, "step": 22756 }, { "epoch": 0.56, "learning_rate": 8.518186278316857e-06, "loss": 2.09, "step": 22757 }, { "epoch": 0.56, "learning_rate": 8.517396765326536e-06, "loss": 2.0208, "step": 22758 }, { "epoch": 0.56, "learning_rate": 8.516607261785332e-06, "loss": 1.8671, "step": 22759 }, { "epoch": 0.56, "learning_rate": 8.51581776769827e-06, "loss": 1.9203, "step": 22760 }, { "epoch": 0.56, "learning_rate": 8.515028283070385e-06, "loss": 1.882, "step": 22761 }, { "epoch": 0.56, "learning_rate": 8.51423880790671e-06, "loss": 1.915, "step": 22762 }, { "epoch": 0.56, "learning_rate": 8.513449342212273e-06, "loss": 1.9613, "step": 22763 }, { "epoch": 0.56, "learning_rate": 8.512659885992114e-06, "loss": 2.0382, "step": 22764 }, { "epoch": 0.56, "learning_rate": 8.511870439251253e-06, "loss": 2.0397, "step": 22765 }, { "epoch": 0.56, "learning_rate": 8.511081001994726e-06, "loss": 1.9153, "step": 22766 }, { "epoch": 0.56, "learning_rate": 8.51029157422757e-06, "loss": 2.0444, "step": 22767 }, { "epoch": 0.56, "learning_rate": 8.509502155954806e-06, "loss": 1.9408, "step": 22768 }, { "epoch": 0.56, "learning_rate": 8.508712747181473e-06, "loss": 2.0778, "step": 22769 }, { "epoch": 0.56, "learning_rate": 8.507923347912598e-06, "loss": 1.9643, "step": 22770 }, { "epoch": 0.56, "learning_rate": 8.507133958153215e-06, "loss": 1.9557, "step": 22771 }, { "epoch": 0.56, "learning_rate": 8.506344577908356e-06, "loss": 2.0326, "step": 22772 }, { "epoch": 0.56, "learning_rate": 8.505555207183045e-06, "loss": 2.0035, "step": 22773 }, { "epoch": 0.56, "learning_rate": 8.504765845982323e-06, "loss": 2.0553, "step": 22774 }, { "epoch": 0.56, "learning_rate": 8.50397649431121e-06, "loss": 2.1665, "step": 22775 }, { "epoch": 0.56, "learning_rate": 8.503187152174746e-06, "loss": 2.1107, "step": 22776 }, { "epoch": 0.56, "learning_rate": 8.502397819577958e-06, "loss": 2.0054, "step": 22777 }, { "epoch": 0.56, "learning_rate": 8.501608496525876e-06, "loss": 2.0547, "step": 22778 }, { "epoch": 0.56, "learning_rate": 8.500819183023532e-06, "loss": 1.9727, "step": 22779 }, { "epoch": 0.56, "learning_rate": 8.500029879075954e-06, "loss": 1.8187, "step": 22780 }, { "epoch": 0.56, "learning_rate": 8.499240584688174e-06, "loss": 1.8309, "step": 22781 }, { "epoch": 0.56, "learning_rate": 8.498451299865228e-06, "loss": 2.0732, "step": 22782 }, { "epoch": 0.56, "learning_rate": 8.497662024612136e-06, "loss": 1.8407, "step": 22783 }, { "epoch": 0.56, "learning_rate": 8.496872758933938e-06, "loss": 1.9805, "step": 22784 }, { "epoch": 0.56, "learning_rate": 8.496083502835654e-06, "loss": 1.7698, "step": 22785 }, { "epoch": 0.56, "learning_rate": 8.495294256322322e-06, "loss": 2.2529, "step": 22786 }, { "epoch": 0.56, "learning_rate": 8.494505019398972e-06, "loss": 1.89, "step": 22787 }, { "epoch": 0.56, "learning_rate": 8.49371579207063e-06, "loss": 1.9312, "step": 22788 }, { "epoch": 0.56, "learning_rate": 8.492926574342332e-06, "loss": 2.0938, "step": 22789 }, { "epoch": 0.56, "learning_rate": 8.4921373662191e-06, "loss": 2.0345, "step": 22790 }, { "epoch": 0.56, "learning_rate": 8.491348167705968e-06, "loss": 2.0567, "step": 22791 }, { "epoch": 0.56, "learning_rate": 8.49055897880797e-06, "loss": 2.054, "step": 22792 }, { "epoch": 0.56, "learning_rate": 8.489769799530128e-06, "loss": 2.0701, "step": 22793 }, { "epoch": 0.56, "learning_rate": 8.488980629877477e-06, "loss": 2.0718, "step": 22794 }, { "epoch": 0.56, "learning_rate": 8.488191469855042e-06, "loss": 2.1398, "step": 22795 }, { "epoch": 0.56, "learning_rate": 8.487402319467856e-06, "loss": 2.1548, "step": 22796 }, { "epoch": 0.56, "learning_rate": 8.48661317872095e-06, "loss": 1.9748, "step": 22797 }, { "epoch": 0.56, "learning_rate": 8.48582404761935e-06, "loss": 1.9516, "step": 22798 }, { "epoch": 0.56, "learning_rate": 8.48503492616809e-06, "loss": 2.2295, "step": 22799 }, { "epoch": 0.56, "learning_rate": 8.484245814372191e-06, "loss": 1.9004, "step": 22800 }, { "epoch": 0.56, "learning_rate": 8.483456712236689e-06, "loss": 2.0658, "step": 22801 }, { "epoch": 0.56, "learning_rate": 8.482667619766615e-06, "loss": 1.9482, "step": 22802 }, { "epoch": 0.56, "learning_rate": 8.481878536966992e-06, "loss": 2.134, "step": 22803 }, { "epoch": 0.56, "learning_rate": 8.481089463842852e-06, "loss": 1.9506, "step": 22804 }, { "epoch": 0.56, "learning_rate": 8.480300400399222e-06, "loss": 1.9815, "step": 22805 }, { "epoch": 0.56, "learning_rate": 8.479511346641135e-06, "loss": 1.9493, "step": 22806 }, { "epoch": 0.56, "learning_rate": 8.47872230257362e-06, "loss": 1.9778, "step": 22807 }, { "epoch": 0.56, "learning_rate": 8.4779332682017e-06, "loss": 2.1387, "step": 22808 }, { "epoch": 0.56, "learning_rate": 8.477144243530411e-06, "loss": 2.0689, "step": 22809 }, { "epoch": 0.56, "learning_rate": 8.476355228564772e-06, "loss": 1.9876, "step": 22810 }, { "epoch": 0.56, "learning_rate": 8.47556622330982e-06, "loss": 2.0617, "step": 22811 }, { "epoch": 0.56, "learning_rate": 8.474777227770586e-06, "loss": 2.0324, "step": 22812 }, { "epoch": 0.56, "learning_rate": 8.473988241952088e-06, "loss": 1.9911, "step": 22813 }, { "epoch": 0.56, "learning_rate": 8.473199265859364e-06, "loss": 1.9829, "step": 22814 }, { "epoch": 0.56, "learning_rate": 8.472410299497434e-06, "loss": 1.9737, "step": 22815 }, { "epoch": 0.56, "learning_rate": 8.471621342871334e-06, "loss": 1.9979, "step": 22816 }, { "epoch": 0.56, "learning_rate": 8.470832395986091e-06, "loss": 2.009, "step": 22817 }, { "epoch": 0.56, "learning_rate": 8.470043458846729e-06, "loss": 1.921, "step": 22818 }, { "epoch": 0.56, "learning_rate": 8.469254531458281e-06, "loss": 1.9583, "step": 22819 }, { "epoch": 0.56, "learning_rate": 8.468465613825769e-06, "loss": 1.9478, "step": 22820 }, { "epoch": 0.56, "learning_rate": 8.467676705954226e-06, "loss": 2.0278, "step": 22821 }, { "epoch": 0.56, "learning_rate": 8.466887807848681e-06, "loss": 1.9869, "step": 22822 }, { "epoch": 0.56, "learning_rate": 8.466098919514156e-06, "loss": 2.0101, "step": 22823 }, { "epoch": 0.56, "learning_rate": 8.465310040955684e-06, "loss": 1.8482, "step": 22824 }, { "epoch": 0.56, "learning_rate": 8.464521172178294e-06, "loss": 1.8439, "step": 22825 }, { "epoch": 0.56, "learning_rate": 8.463732313187008e-06, "loss": 2.0481, "step": 22826 }, { "epoch": 0.56, "learning_rate": 8.462943463986858e-06, "loss": 2.0151, "step": 22827 }, { "epoch": 0.56, "learning_rate": 8.46215462458287e-06, "loss": 1.9966, "step": 22828 }, { "epoch": 0.56, "learning_rate": 8.46136579498007e-06, "loss": 2.0374, "step": 22829 }, { "epoch": 0.56, "learning_rate": 8.46057697518349e-06, "loss": 2.0854, "step": 22830 }, { "epoch": 0.56, "learning_rate": 8.459788165198152e-06, "loss": 2.1276, "step": 22831 }, { "epoch": 0.56, "learning_rate": 8.458999365029088e-06, "loss": 2.0267, "step": 22832 }, { "epoch": 0.56, "learning_rate": 8.458210574681321e-06, "loss": 2.1126, "step": 22833 }, { "epoch": 0.56, "learning_rate": 8.45742179415988e-06, "loss": 2.0778, "step": 22834 }, { "epoch": 0.56, "learning_rate": 8.456633023469797e-06, "loss": 1.9018, "step": 22835 }, { "epoch": 0.56, "learning_rate": 8.455844262616092e-06, "loss": 1.8484, "step": 22836 }, { "epoch": 0.56, "learning_rate": 8.455055511603799e-06, "loss": 2.1728, "step": 22837 }, { "epoch": 0.56, "learning_rate": 8.454266770437935e-06, "loss": 2.0091, "step": 22838 }, { "epoch": 0.56, "learning_rate": 8.453478039123531e-06, "loss": 1.9339, "step": 22839 }, { "epoch": 0.56, "learning_rate": 8.45268931766562e-06, "loss": 2.1087, "step": 22840 }, { "epoch": 0.56, "learning_rate": 8.451900606069222e-06, "loss": 2.027, "step": 22841 }, { "epoch": 0.56, "learning_rate": 8.451111904339369e-06, "loss": 1.9347, "step": 22842 }, { "epoch": 0.56, "learning_rate": 8.450323212481082e-06, "loss": 1.834, "step": 22843 }, { "epoch": 0.56, "learning_rate": 8.449534530499388e-06, "loss": 2.05, "step": 22844 }, { "epoch": 0.56, "learning_rate": 8.448745858399322e-06, "loss": 1.869, "step": 22845 }, { "epoch": 0.56, "learning_rate": 8.447957196185899e-06, "loss": 1.9876, "step": 22846 }, { "epoch": 0.56, "learning_rate": 8.447168543864153e-06, "loss": 2.0205, "step": 22847 }, { "epoch": 0.56, "learning_rate": 8.446379901439109e-06, "loss": 1.8695, "step": 22848 }, { "epoch": 0.56, "learning_rate": 8.445591268915788e-06, "loss": 1.994, "step": 22849 }, { "epoch": 0.56, "learning_rate": 8.444802646299223e-06, "loss": 2.0206, "step": 22850 }, { "epoch": 0.56, "learning_rate": 8.444014033594437e-06, "loss": 2.1647, "step": 22851 }, { "epoch": 0.56, "learning_rate": 8.443225430806458e-06, "loss": 1.9376, "step": 22852 }, { "epoch": 0.56, "learning_rate": 8.442436837940309e-06, "loss": 1.9889, "step": 22853 }, { "epoch": 0.56, "learning_rate": 8.441648255001017e-06, "loss": 2.17, "step": 22854 }, { "epoch": 0.56, "learning_rate": 8.440859681993613e-06, "loss": 2.0166, "step": 22855 }, { "epoch": 0.56, "learning_rate": 8.440071118923114e-06, "loss": 1.9361, "step": 22856 }, { "epoch": 0.56, "learning_rate": 8.43928256579455e-06, "loss": 1.982, "step": 22857 }, { "epoch": 0.56, "learning_rate": 8.438494022612948e-06, "loss": 2.0304, "step": 22858 }, { "epoch": 0.56, "learning_rate": 8.43770548938333e-06, "loss": 2.1298, "step": 22859 }, { "epoch": 0.56, "learning_rate": 8.43691696611073e-06, "loss": 1.9231, "step": 22860 }, { "epoch": 0.56, "learning_rate": 8.436128452800163e-06, "loss": 2.0541, "step": 22861 }, { "epoch": 0.56, "learning_rate": 8.435339949456663e-06, "loss": 2.0284, "step": 22862 }, { "epoch": 0.56, "learning_rate": 8.434551456085248e-06, "loss": 2.1159, "step": 22863 }, { "epoch": 0.56, "learning_rate": 8.433762972690944e-06, "loss": 1.943, "step": 22864 }, { "epoch": 0.56, "learning_rate": 8.432974499278786e-06, "loss": 2.0783, "step": 22865 }, { "epoch": 0.56, "learning_rate": 8.432186035853787e-06, "loss": 1.8932, "step": 22866 }, { "epoch": 0.56, "learning_rate": 8.43139758242098e-06, "loss": 2.1587, "step": 22867 }, { "epoch": 0.56, "learning_rate": 8.430609138985384e-06, "loss": 1.9768, "step": 22868 }, { "epoch": 0.56, "learning_rate": 8.42982070555203e-06, "loss": 1.9369, "step": 22869 }, { "epoch": 0.56, "learning_rate": 8.429032282125942e-06, "loss": 1.7966, "step": 22870 }, { "epoch": 0.56, "learning_rate": 8.42824386871214e-06, "loss": 2.0145, "step": 22871 }, { "epoch": 0.56, "learning_rate": 8.427455465315654e-06, "loss": 2.1023, "step": 22872 }, { "epoch": 0.56, "learning_rate": 8.426667071941505e-06, "loss": 2.0791, "step": 22873 }, { "epoch": 0.56, "learning_rate": 8.42587868859472e-06, "loss": 1.9586, "step": 22874 }, { "epoch": 0.56, "learning_rate": 8.425090315280324e-06, "loss": 2.0012, "step": 22875 }, { "epoch": 0.56, "learning_rate": 8.42430195200334e-06, "loss": 2.0065, "step": 22876 }, { "epoch": 0.56, "learning_rate": 8.423513598768795e-06, "loss": 1.8554, "step": 22877 }, { "epoch": 0.56, "learning_rate": 8.422725255581709e-06, "loss": 2.1296, "step": 22878 }, { "epoch": 0.56, "learning_rate": 8.42193692244711e-06, "loss": 2.0345, "step": 22879 }, { "epoch": 0.56, "learning_rate": 8.421148599370023e-06, "loss": 2.0526, "step": 22880 }, { "epoch": 0.56, "learning_rate": 8.420360286355467e-06, "loss": 1.991, "step": 22881 }, { "epoch": 0.56, "learning_rate": 8.41957198340847e-06, "loss": 1.9338, "step": 22882 }, { "epoch": 0.56, "learning_rate": 8.41878369053406e-06, "loss": 2.0077, "step": 22883 }, { "epoch": 0.56, "learning_rate": 8.417995407737254e-06, "loss": 2.0623, "step": 22884 }, { "epoch": 0.56, "learning_rate": 8.417207135023078e-06, "loss": 1.9498, "step": 22885 }, { "epoch": 0.56, "learning_rate": 8.416418872396557e-06, "loss": 1.9143, "step": 22886 }, { "epoch": 0.56, "learning_rate": 8.415630619862714e-06, "loss": 2.0443, "step": 22887 }, { "epoch": 0.56, "learning_rate": 8.414842377426578e-06, "loss": 2.0957, "step": 22888 }, { "epoch": 0.56, "learning_rate": 8.414054145093163e-06, "loss": 1.8125, "step": 22889 }, { "epoch": 0.56, "learning_rate": 8.413265922867503e-06, "loss": 2.014, "step": 22890 }, { "epoch": 0.56, "learning_rate": 8.41247771075461e-06, "loss": 2.0662, "step": 22891 }, { "epoch": 0.56, "learning_rate": 8.411689508759516e-06, "loss": 2.0088, "step": 22892 }, { "epoch": 0.56, "learning_rate": 8.410901316887244e-06, "loss": 1.9504, "step": 22893 }, { "epoch": 0.56, "learning_rate": 8.410113135142814e-06, "loss": 1.9438, "step": 22894 }, { "epoch": 0.56, "learning_rate": 8.409324963531255e-06, "loss": 1.9688, "step": 22895 }, { "epoch": 0.56, "learning_rate": 8.408536802057582e-06, "loss": 2.0201, "step": 22896 }, { "epoch": 0.56, "learning_rate": 8.407748650726823e-06, "loss": 2.188, "step": 22897 }, { "epoch": 0.56, "learning_rate": 8.406960509544004e-06, "loss": 2.082, "step": 22898 }, { "epoch": 0.56, "learning_rate": 8.406172378514141e-06, "loss": 2.0669, "step": 22899 }, { "epoch": 0.56, "learning_rate": 8.405384257642264e-06, "loss": 1.896, "step": 22900 }, { "epoch": 0.56, "learning_rate": 8.404596146933392e-06, "loss": 2.0563, "step": 22901 }, { "epoch": 0.56, "learning_rate": 8.403808046392548e-06, "loss": 2.0445, "step": 22902 }, { "epoch": 0.56, "learning_rate": 8.403019956024755e-06, "loss": 2.0043, "step": 22903 }, { "epoch": 0.56, "learning_rate": 8.402231875835036e-06, "loss": 1.8852, "step": 22904 }, { "epoch": 0.56, "learning_rate": 8.401443805828418e-06, "loss": 2.068, "step": 22905 }, { "epoch": 0.56, "learning_rate": 8.400655746009914e-06, "loss": 1.9915, "step": 22906 }, { "epoch": 0.56, "learning_rate": 8.399867696384553e-06, "loss": 2.0314, "step": 22907 }, { "epoch": 0.56, "learning_rate": 8.399079656957361e-06, "loss": 2.0151, "step": 22908 }, { "epoch": 0.56, "learning_rate": 8.398291627733353e-06, "loss": 2.0607, "step": 22909 }, { "epoch": 0.56, "learning_rate": 8.397503608717555e-06, "loss": 1.8889, "step": 22910 }, { "epoch": 0.56, "learning_rate": 8.396715599914986e-06, "loss": 1.9209, "step": 22911 }, { "epoch": 0.56, "learning_rate": 8.395927601330672e-06, "loss": 2.1488, "step": 22912 }, { "epoch": 0.56, "learning_rate": 8.395139612969638e-06, "loss": 1.8827, "step": 22913 }, { "epoch": 0.56, "learning_rate": 8.3943516348369e-06, "loss": 2.0855, "step": 22914 }, { "epoch": 0.56, "learning_rate": 8.393563666937484e-06, "loss": 1.8771, "step": 22915 }, { "epoch": 0.56, "learning_rate": 8.392775709276407e-06, "loss": 1.9438, "step": 22916 }, { "epoch": 0.56, "learning_rate": 8.391987761858693e-06, "loss": 1.9687, "step": 22917 }, { "epoch": 0.56, "learning_rate": 8.39119982468937e-06, "loss": 2.0135, "step": 22918 }, { "epoch": 0.56, "learning_rate": 8.390411897773452e-06, "loss": 2.0368, "step": 22919 }, { "epoch": 0.56, "learning_rate": 8.389623981115965e-06, "loss": 2.0681, "step": 22920 }, { "epoch": 0.56, "learning_rate": 8.388836074721926e-06, "loss": 2.0281, "step": 22921 }, { "epoch": 0.56, "learning_rate": 8.388048178596361e-06, "loss": 2.0096, "step": 22922 }, { "epoch": 0.57, "learning_rate": 8.387260292744294e-06, "loss": 2.0781, "step": 22923 }, { "epoch": 0.57, "learning_rate": 8.386472417170738e-06, "loss": 1.9712, "step": 22924 }, { "epoch": 0.57, "learning_rate": 8.385684551880723e-06, "loss": 2.0948, "step": 22925 }, { "epoch": 0.57, "learning_rate": 8.384896696879265e-06, "loss": 1.9834, "step": 22926 }, { "epoch": 0.57, "learning_rate": 8.384108852171385e-06, "loss": 2.0448, "step": 22927 }, { "epoch": 0.57, "learning_rate": 8.383321017762105e-06, "loss": 2.1172, "step": 22928 }, { "epoch": 0.57, "learning_rate": 8.38253319365645e-06, "loss": 1.98, "step": 22929 }, { "epoch": 0.57, "learning_rate": 8.381745379859438e-06, "loss": 1.9924, "step": 22930 }, { "epoch": 0.57, "learning_rate": 8.380957576376089e-06, "loss": 1.8641, "step": 22931 }, { "epoch": 0.57, "learning_rate": 8.380169783211422e-06, "loss": 2.0684, "step": 22932 }, { "epoch": 0.57, "learning_rate": 8.379382000370469e-06, "loss": 2.0585, "step": 22933 }, { "epoch": 0.57, "learning_rate": 8.378594227858236e-06, "loss": 2.1412, "step": 22934 }, { "epoch": 0.57, "learning_rate": 8.377806465679752e-06, "loss": 2.0769, "step": 22935 }, { "epoch": 0.57, "learning_rate": 8.377018713840036e-06, "loss": 2.0534, "step": 22936 }, { "epoch": 0.57, "learning_rate": 8.376230972344108e-06, "loss": 2.1144, "step": 22937 }, { "epoch": 0.57, "learning_rate": 8.375443241196991e-06, "loss": 2.0427, "step": 22938 }, { "epoch": 0.57, "learning_rate": 8.3746555204037e-06, "loss": 1.8933, "step": 22939 }, { "epoch": 0.57, "learning_rate": 8.37386780996926e-06, "loss": 1.9817, "step": 22940 }, { "epoch": 0.57, "learning_rate": 8.373080109898696e-06, "loss": 2.0235, "step": 22941 }, { "epoch": 0.57, "learning_rate": 8.372292420197017e-06, "loss": 2.087, "step": 22942 }, { "epoch": 0.57, "learning_rate": 8.371504740869253e-06, "loss": 1.9694, "step": 22943 }, { "epoch": 0.57, "learning_rate": 8.370717071920416e-06, "loss": 2.0376, "step": 22944 }, { "epoch": 0.57, "learning_rate": 8.369929413355531e-06, "loss": 2.0791, "step": 22945 }, { "epoch": 0.57, "learning_rate": 8.369141765179618e-06, "loss": 2.0754, "step": 22946 }, { "epoch": 0.57, "learning_rate": 8.368354127397694e-06, "loss": 1.9654, "step": 22947 }, { "epoch": 0.57, "learning_rate": 8.367566500014785e-06, "loss": 1.8992, "step": 22948 }, { "epoch": 0.57, "learning_rate": 8.366778883035904e-06, "loss": 2.0198, "step": 22949 }, { "epoch": 0.57, "learning_rate": 8.365991276466072e-06, "loss": 2.0265, "step": 22950 }, { "epoch": 0.57, "learning_rate": 8.365203680310315e-06, "loss": 1.9641, "step": 22951 }, { "epoch": 0.57, "learning_rate": 8.364416094573644e-06, "loss": 2.1266, "step": 22952 }, { "epoch": 0.57, "learning_rate": 8.363628519261082e-06, "loss": 2.1459, "step": 22953 }, { "epoch": 0.57, "learning_rate": 8.36284095437765e-06, "loss": 1.9178, "step": 22954 }, { "epoch": 0.57, "learning_rate": 8.362053399928365e-06, "loss": 2.0448, "step": 22955 }, { "epoch": 0.57, "learning_rate": 8.361265855918248e-06, "loss": 1.9869, "step": 22956 }, { "epoch": 0.57, "learning_rate": 8.360478322352317e-06, "loss": 2.0851, "step": 22957 }, { "epoch": 0.57, "learning_rate": 8.359690799235596e-06, "loss": 1.9421, "step": 22958 }, { "epoch": 0.57, "learning_rate": 8.358903286573095e-06, "loss": 1.9824, "step": 22959 }, { "epoch": 0.57, "learning_rate": 8.358115784369838e-06, "loss": 2.0188, "step": 22960 }, { "epoch": 0.57, "learning_rate": 8.357328292630849e-06, "loss": 2.1048, "step": 22961 }, { "epoch": 0.57, "learning_rate": 8.356540811361139e-06, "loss": 1.9312, "step": 22962 }, { "epoch": 0.57, "learning_rate": 8.355753340565731e-06, "loss": 2.1521, "step": 22963 }, { "epoch": 0.57, "learning_rate": 8.35496588024964e-06, "loss": 2.2463, "step": 22964 }, { "epoch": 0.57, "learning_rate": 8.354178430417888e-06, "loss": 2.0048, "step": 22965 }, { "epoch": 0.57, "learning_rate": 8.353390991075499e-06, "loss": 2.1208, "step": 22966 }, { "epoch": 0.57, "learning_rate": 8.352603562227479e-06, "loss": 1.9297, "step": 22967 }, { "epoch": 0.57, "learning_rate": 8.351816143878859e-06, "loss": 2.1631, "step": 22968 }, { "epoch": 0.57, "learning_rate": 8.351028736034648e-06, "loss": 2.0358, "step": 22969 }, { "epoch": 0.57, "learning_rate": 8.350241338699869e-06, "loss": 2.0518, "step": 22970 }, { "epoch": 0.57, "learning_rate": 8.34945395187954e-06, "loss": 1.9703, "step": 22971 }, { "epoch": 0.57, "learning_rate": 8.348666575578678e-06, "loss": 1.9976, "step": 22972 }, { "epoch": 0.57, "learning_rate": 8.347879209802305e-06, "loss": 1.9132, "step": 22973 }, { "epoch": 0.57, "learning_rate": 8.347091854555432e-06, "loss": 2.1942, "step": 22974 }, { "epoch": 0.57, "learning_rate": 8.346304509843081e-06, "loss": 1.9841, "step": 22975 }, { "epoch": 0.57, "learning_rate": 8.345517175670276e-06, "loss": 1.9431, "step": 22976 }, { "epoch": 0.57, "learning_rate": 8.344729852042023e-06, "loss": 1.9837, "step": 22977 }, { "epoch": 0.57, "learning_rate": 8.343942538963354e-06, "loss": 2.0172, "step": 22978 }, { "epoch": 0.57, "learning_rate": 8.343155236439271e-06, "loss": 1.9415, "step": 22979 }, { "epoch": 0.57, "learning_rate": 8.3423679444748e-06, "loss": 2.0888, "step": 22980 }, { "epoch": 0.57, "learning_rate": 8.341580663074962e-06, "loss": 1.9689, "step": 22981 }, { "epoch": 0.57, "learning_rate": 8.340793392244768e-06, "loss": 1.9862, "step": 22982 }, { "epoch": 0.57, "learning_rate": 8.340006131989244e-06, "loss": 2.0103, "step": 22983 }, { "epoch": 0.57, "learning_rate": 8.339218882313396e-06, "loss": 2.0372, "step": 22984 }, { "epoch": 0.57, "learning_rate": 8.338431643222247e-06, "loss": 1.8455, "step": 22985 }, { "epoch": 0.57, "learning_rate": 8.33764441472082e-06, "loss": 2.0223, "step": 22986 }, { "epoch": 0.57, "learning_rate": 8.336857196814121e-06, "loss": 2.0515, "step": 22987 }, { "epoch": 0.57, "learning_rate": 8.336069989507179e-06, "loss": 1.9538, "step": 22988 }, { "epoch": 0.57, "learning_rate": 8.335282792804999e-06, "loss": 2.1371, "step": 22989 }, { "epoch": 0.57, "learning_rate": 8.334495606712608e-06, "loss": 1.9081, "step": 22990 }, { "epoch": 0.57, "learning_rate": 8.333708431235019e-06, "loss": 1.9658, "step": 22991 }, { "epoch": 0.57, "learning_rate": 8.332921266377247e-06, "loss": 2.0794, "step": 22992 }, { "epoch": 0.57, "learning_rate": 8.332134112144317e-06, "loss": 1.9995, "step": 22993 }, { "epoch": 0.57, "learning_rate": 8.331346968541235e-06, "loss": 1.9546, "step": 22994 }, { "epoch": 0.57, "learning_rate": 8.33055983557302e-06, "loss": 1.9702, "step": 22995 }, { "epoch": 0.57, "learning_rate": 8.329772713244698e-06, "loss": 2.034, "step": 22996 }, { "epoch": 0.57, "learning_rate": 8.328985601561273e-06, "loss": 1.9977, "step": 22997 }, { "epoch": 0.57, "learning_rate": 8.32819850052777e-06, "loss": 2.0907, "step": 22998 }, { "epoch": 0.57, "learning_rate": 8.327411410149202e-06, "loss": 1.9801, "step": 22999 }, { "epoch": 0.57, "learning_rate": 8.326624330430585e-06, "loss": 1.8752, "step": 23000 }, { "epoch": 0.57, "learning_rate": 8.325837261376941e-06, "loss": 1.997, "step": 23001 }, { "epoch": 0.57, "learning_rate": 8.325050202993276e-06, "loss": 2.0265, "step": 23002 }, { "epoch": 0.57, "learning_rate": 8.324263155284614e-06, "loss": 1.9276, "step": 23003 }, { "epoch": 0.57, "learning_rate": 8.323476118255973e-06, "loss": 2.0239, "step": 23004 }, { "epoch": 0.57, "learning_rate": 8.322689091912359e-06, "loss": 2.0407, "step": 23005 }, { "epoch": 0.57, "learning_rate": 8.3219020762588e-06, "loss": 2.0022, "step": 23006 }, { "epoch": 0.57, "learning_rate": 8.3211150713003e-06, "loss": 2.1413, "step": 23007 }, { "epoch": 0.57, "learning_rate": 8.320328077041884e-06, "loss": 1.9238, "step": 23008 }, { "epoch": 0.57, "learning_rate": 8.319541093488564e-06, "loss": 2.0476, "step": 23009 }, { "epoch": 0.57, "learning_rate": 8.318754120645356e-06, "loss": 1.9355, "step": 23010 }, { "epoch": 0.57, "learning_rate": 8.317967158517278e-06, "loss": 1.9231, "step": 23011 }, { "epoch": 0.57, "learning_rate": 8.317180207109341e-06, "loss": 1.878, "step": 23012 }, { "epoch": 0.57, "learning_rate": 8.316393266426561e-06, "loss": 1.9126, "step": 23013 }, { "epoch": 0.57, "learning_rate": 8.315606336473962e-06, "loss": 2.046, "step": 23014 }, { "epoch": 0.57, "learning_rate": 8.314819417256547e-06, "loss": 2.0229, "step": 23015 }, { "epoch": 0.57, "learning_rate": 8.31403250877934e-06, "loss": 1.903, "step": 23016 }, { "epoch": 0.57, "learning_rate": 8.313245611047349e-06, "loss": 1.9792, "step": 23017 }, { "epoch": 0.57, "learning_rate": 8.312458724065595e-06, "loss": 1.9524, "step": 23018 }, { "epoch": 0.57, "learning_rate": 8.311671847839097e-06, "loss": 2.0291, "step": 23019 }, { "epoch": 0.57, "learning_rate": 8.310884982372859e-06, "loss": 1.9938, "step": 23020 }, { "epoch": 0.57, "learning_rate": 8.310098127671905e-06, "loss": 2.1777, "step": 23021 }, { "epoch": 0.57, "learning_rate": 8.309311283741243e-06, "loss": 2.0725, "step": 23022 }, { "epoch": 0.57, "learning_rate": 8.30852445058589e-06, "loss": 1.8779, "step": 23023 }, { "epoch": 0.57, "learning_rate": 8.307737628210865e-06, "loss": 1.9716, "step": 23024 }, { "epoch": 0.57, "learning_rate": 8.306950816621178e-06, "loss": 2.1306, "step": 23025 }, { "epoch": 0.57, "learning_rate": 8.306164015821849e-06, "loss": 1.947, "step": 23026 }, { "epoch": 0.57, "learning_rate": 8.305377225817885e-06, "loss": 1.9969, "step": 23027 }, { "epoch": 0.57, "learning_rate": 8.304590446614304e-06, "loss": 2.0461, "step": 23028 }, { "epoch": 0.57, "learning_rate": 8.303803678216125e-06, "loss": 2.033, "step": 23029 }, { "epoch": 0.57, "learning_rate": 8.303016920628352e-06, "loss": 2.0636, "step": 23030 }, { "epoch": 0.57, "learning_rate": 8.302230173856011e-06, "loss": 1.9754, "step": 23031 }, { "epoch": 0.57, "learning_rate": 8.301443437904107e-06, "loss": 1.8403, "step": 23032 }, { "epoch": 0.57, "learning_rate": 8.300656712777656e-06, "loss": 2.0755, "step": 23033 }, { "epoch": 0.57, "learning_rate": 8.299869998481676e-06, "loss": 1.975, "step": 23034 }, { "epoch": 0.57, "learning_rate": 8.299083295021176e-06, "loss": 1.9658, "step": 23035 }, { "epoch": 0.57, "learning_rate": 8.298296602401178e-06, "loss": 1.8792, "step": 23036 }, { "epoch": 0.57, "learning_rate": 8.297509920626684e-06, "loss": 2.1308, "step": 23037 }, { "epoch": 0.57, "learning_rate": 8.296723249702715e-06, "loss": 1.9629, "step": 23038 }, { "epoch": 0.57, "learning_rate": 8.295936589634288e-06, "loss": 1.827, "step": 23039 }, { "epoch": 0.57, "learning_rate": 8.295149940426409e-06, "loss": 1.9484, "step": 23040 }, { "epoch": 0.57, "learning_rate": 8.294363302084096e-06, "loss": 2.0981, "step": 23041 }, { "epoch": 0.57, "learning_rate": 8.293576674612361e-06, "loss": 1.9763, "step": 23042 }, { "epoch": 0.57, "learning_rate": 8.292790058016216e-06, "loss": 2.208, "step": 23043 }, { "epoch": 0.57, "learning_rate": 8.292003452300684e-06, "loss": 2.0989, "step": 23044 }, { "epoch": 0.57, "learning_rate": 8.291216857470763e-06, "loss": 1.9911, "step": 23045 }, { "epoch": 0.57, "learning_rate": 8.290430273531479e-06, "loss": 1.9083, "step": 23046 }, { "epoch": 0.57, "learning_rate": 8.289643700487837e-06, "loss": 1.9164, "step": 23047 }, { "epoch": 0.57, "learning_rate": 8.28885713834485e-06, "loss": 1.8184, "step": 23048 }, { "epoch": 0.57, "learning_rate": 8.288070587107542e-06, "loss": 2.0304, "step": 23049 }, { "epoch": 0.57, "learning_rate": 8.287284046780912e-06, "loss": 1.8162, "step": 23050 }, { "epoch": 0.57, "learning_rate": 8.28649751736998e-06, "loss": 2.0781, "step": 23051 }, { "epoch": 0.57, "learning_rate": 8.285710998879758e-06, "loss": 2.1663, "step": 23052 }, { "epoch": 0.57, "learning_rate": 8.284924491315256e-06, "loss": 1.8873, "step": 23053 }, { "epoch": 0.57, "learning_rate": 8.284137994681495e-06, "loss": 2.0625, "step": 23054 }, { "epoch": 0.57, "learning_rate": 8.283351508983477e-06, "loss": 2.0333, "step": 23055 }, { "epoch": 0.57, "learning_rate": 8.282565034226225e-06, "loss": 1.7929, "step": 23056 }, { "epoch": 0.57, "learning_rate": 8.281778570414739e-06, "loss": 1.9908, "step": 23057 }, { "epoch": 0.57, "learning_rate": 8.280992117554038e-06, "loss": 2.0662, "step": 23058 }, { "epoch": 0.57, "learning_rate": 8.280205675649138e-06, "loss": 2.0044, "step": 23059 }, { "epoch": 0.57, "learning_rate": 8.279419244705046e-06, "loss": 1.944, "step": 23060 }, { "epoch": 0.57, "learning_rate": 8.278632824726778e-06, "loss": 2.0119, "step": 23061 }, { "epoch": 0.57, "learning_rate": 8.277846415719344e-06, "loss": 1.8541, "step": 23062 }, { "epoch": 0.57, "learning_rate": 8.277060017687752e-06, "loss": 2.0399, "step": 23063 }, { "epoch": 0.57, "learning_rate": 8.276273630637024e-06, "loss": 2.0584, "step": 23064 }, { "epoch": 0.57, "learning_rate": 8.275487254572161e-06, "loss": 1.9226, "step": 23065 }, { "epoch": 0.57, "learning_rate": 8.27470088949818e-06, "loss": 2.0417, "step": 23066 }, { "epoch": 0.57, "learning_rate": 8.273914535420097e-06, "loss": 1.8207, "step": 23067 }, { "epoch": 0.57, "learning_rate": 8.273128192342914e-06, "loss": 2.0231, "step": 23068 }, { "epoch": 0.57, "learning_rate": 8.272341860271651e-06, "loss": 1.9344, "step": 23069 }, { "epoch": 0.57, "learning_rate": 8.271555539211314e-06, "loss": 2.0719, "step": 23070 }, { "epoch": 0.57, "learning_rate": 8.270769229166916e-06, "loss": 1.9334, "step": 23071 }, { "epoch": 0.57, "learning_rate": 8.269982930143474e-06, "loss": 2.017, "step": 23072 }, { "epoch": 0.57, "learning_rate": 8.269196642145992e-06, "loss": 1.9269, "step": 23073 }, { "epoch": 0.57, "learning_rate": 8.268410365179486e-06, "loss": 1.9071, "step": 23074 }, { "epoch": 0.57, "learning_rate": 8.26762409924896e-06, "loss": 2.0872, "step": 23075 }, { "epoch": 0.57, "learning_rate": 8.266837844359431e-06, "loss": 2.0298, "step": 23076 }, { "epoch": 0.57, "learning_rate": 8.266051600515913e-06, "loss": 1.9206, "step": 23077 }, { "epoch": 0.57, "learning_rate": 8.265265367723409e-06, "loss": 1.949, "step": 23078 }, { "epoch": 0.57, "learning_rate": 8.26447914598694e-06, "loss": 2.1074, "step": 23079 }, { "epoch": 0.57, "learning_rate": 8.263692935311507e-06, "loss": 2.1648, "step": 23080 }, { "epoch": 0.57, "learning_rate": 8.262906735702123e-06, "loss": 2.0678, "step": 23081 }, { "epoch": 0.57, "learning_rate": 8.262120547163805e-06, "loss": 2.0087, "step": 23082 }, { "epoch": 0.57, "learning_rate": 8.261334369701556e-06, "loss": 2.0648, "step": 23083 }, { "epoch": 0.57, "learning_rate": 8.260548203320394e-06, "loss": 2.1822, "step": 23084 }, { "epoch": 0.57, "learning_rate": 8.25976204802532e-06, "loss": 1.9906, "step": 23085 }, { "epoch": 0.57, "learning_rate": 8.25897590382135e-06, "loss": 1.9922, "step": 23086 }, { "epoch": 0.57, "learning_rate": 8.258189770713496e-06, "loss": 1.8864, "step": 23087 }, { "epoch": 0.57, "learning_rate": 8.257403648706764e-06, "loss": 2.0416, "step": 23088 }, { "epoch": 0.57, "learning_rate": 8.256617537806171e-06, "loss": 1.8889, "step": 23089 }, { "epoch": 0.57, "learning_rate": 8.255831438016717e-06, "loss": 1.9188, "step": 23090 }, { "epoch": 0.57, "learning_rate": 8.25504534934342e-06, "loss": 1.903, "step": 23091 }, { "epoch": 0.57, "learning_rate": 8.25425927179129e-06, "loss": 1.9724, "step": 23092 }, { "epoch": 0.57, "learning_rate": 8.253473205365331e-06, "loss": 1.8617, "step": 23093 }, { "epoch": 0.57, "learning_rate": 8.252687150070558e-06, "loss": 2.1504, "step": 23094 }, { "epoch": 0.57, "learning_rate": 8.251901105911977e-06, "loss": 1.9645, "step": 23095 }, { "epoch": 0.57, "learning_rate": 8.251115072894602e-06, "loss": 1.7861, "step": 23096 }, { "epoch": 0.57, "learning_rate": 8.250329051023444e-06, "loss": 1.9626, "step": 23097 }, { "epoch": 0.57, "learning_rate": 8.249543040303504e-06, "loss": 2.0053, "step": 23098 }, { "epoch": 0.57, "learning_rate": 8.248757040739802e-06, "loss": 2.0817, "step": 23099 }, { "epoch": 0.57, "learning_rate": 8.247971052337336e-06, "loss": 1.9749, "step": 23100 }, { "epoch": 0.57, "learning_rate": 8.247185075101123e-06, "loss": 1.9152, "step": 23101 }, { "epoch": 0.57, "learning_rate": 8.246399109036175e-06, "loss": 2.0019, "step": 23102 }, { "epoch": 0.57, "learning_rate": 8.24561315414749e-06, "loss": 1.951, "step": 23103 }, { "epoch": 0.57, "learning_rate": 8.244827210440089e-06, "loss": 2.0495, "step": 23104 }, { "epoch": 0.57, "learning_rate": 8.244041277918974e-06, "loss": 1.9618, "step": 23105 }, { "epoch": 0.57, "learning_rate": 8.243255356589156e-06, "loss": 1.9064, "step": 23106 }, { "epoch": 0.57, "learning_rate": 8.242469446455648e-06, "loss": 2.1107, "step": 23107 }, { "epoch": 0.57, "learning_rate": 8.241683547523452e-06, "loss": 2.2822, "step": 23108 }, { "epoch": 0.57, "learning_rate": 8.240897659797583e-06, "loss": 2.0858, "step": 23109 }, { "epoch": 0.57, "learning_rate": 8.24011178328304e-06, "loss": 1.8456, "step": 23110 }, { "epoch": 0.57, "learning_rate": 8.239325917984841e-06, "loss": 2.0404, "step": 23111 }, { "epoch": 0.57, "learning_rate": 8.238540063907992e-06, "loss": 1.8741, "step": 23112 }, { "epoch": 0.57, "learning_rate": 8.237754221057502e-06, "loss": 1.9329, "step": 23113 }, { "epoch": 0.57, "learning_rate": 8.23696838943838e-06, "loss": 1.9687, "step": 23114 }, { "epoch": 0.57, "learning_rate": 8.23618256905563e-06, "loss": 2.0433, "step": 23115 }, { "epoch": 0.57, "learning_rate": 8.235396759914263e-06, "loss": 1.9697, "step": 23116 }, { "epoch": 0.57, "learning_rate": 8.23461096201929e-06, "loss": 2.1096, "step": 23117 }, { "epoch": 0.57, "learning_rate": 8.233825175375716e-06, "loss": 2.0153, "step": 23118 }, { "epoch": 0.57, "learning_rate": 8.233039399988549e-06, "loss": 1.8641, "step": 23119 }, { "epoch": 0.57, "learning_rate": 8.232253635862801e-06, "loss": 1.7848, "step": 23120 }, { "epoch": 0.57, "learning_rate": 8.231467883003472e-06, "loss": 1.9408, "step": 23121 }, { "epoch": 0.57, "learning_rate": 8.230682141415577e-06, "loss": 2.0803, "step": 23122 }, { "epoch": 0.57, "learning_rate": 8.229896411104119e-06, "loss": 2.0349, "step": 23123 }, { "epoch": 0.57, "learning_rate": 8.229110692074109e-06, "loss": 1.919, "step": 23124 }, { "epoch": 0.57, "learning_rate": 8.228324984330557e-06, "loss": 2.0081, "step": 23125 }, { "epoch": 0.57, "learning_rate": 8.227539287878462e-06, "loss": 1.8773, "step": 23126 }, { "epoch": 0.57, "learning_rate": 8.226753602722843e-06, "loss": 2.0416, "step": 23127 }, { "epoch": 0.57, "learning_rate": 8.225967928868697e-06, "loss": 2.0531, "step": 23128 }, { "epoch": 0.57, "learning_rate": 8.225182266321034e-06, "loss": 1.9576, "step": 23129 }, { "epoch": 0.57, "learning_rate": 8.224396615084865e-06, "loss": 1.8453, "step": 23130 }, { "epoch": 0.57, "learning_rate": 8.223610975165194e-06, "loss": 2.0316, "step": 23131 }, { "epoch": 0.57, "learning_rate": 8.222825346567032e-06, "loss": 2.0447, "step": 23132 }, { "epoch": 0.57, "learning_rate": 8.222039729295381e-06, "loss": 1.9358, "step": 23133 }, { "epoch": 0.57, "learning_rate": 8.221254123355249e-06, "loss": 2.0746, "step": 23134 }, { "epoch": 0.57, "learning_rate": 8.220468528751648e-06, "loss": 2.1226, "step": 23135 }, { "epoch": 0.57, "learning_rate": 8.219682945489578e-06, "loss": 2.0888, "step": 23136 }, { "epoch": 0.57, "learning_rate": 8.21889737357405e-06, "loss": 1.8776, "step": 23137 }, { "epoch": 0.57, "learning_rate": 8.21811181301007e-06, "loss": 2.1171, "step": 23138 }, { "epoch": 0.57, "learning_rate": 8.217326263802641e-06, "loss": 1.9046, "step": 23139 }, { "epoch": 0.57, "learning_rate": 8.216540725956775e-06, "loss": 1.9887, "step": 23140 }, { "epoch": 0.57, "learning_rate": 8.215755199477476e-06, "loss": 1.9312, "step": 23141 }, { "epoch": 0.57, "learning_rate": 8.214969684369752e-06, "loss": 1.8351, "step": 23142 }, { "epoch": 0.57, "learning_rate": 8.214184180638605e-06, "loss": 2.097, "step": 23143 }, { "epoch": 0.57, "learning_rate": 8.213398688289045e-06, "loss": 2.081, "step": 23144 }, { "epoch": 0.57, "learning_rate": 8.212613207326082e-06, "loss": 2.0022, "step": 23145 }, { "epoch": 0.57, "learning_rate": 8.211827737754711e-06, "loss": 2.1546, "step": 23146 }, { "epoch": 0.57, "learning_rate": 8.211042279579948e-06, "loss": 2.0114, "step": 23147 }, { "epoch": 0.57, "learning_rate": 8.210256832806793e-06, "loss": 2.11, "step": 23148 }, { "epoch": 0.57, "learning_rate": 8.209471397440253e-06, "loss": 1.9624, "step": 23149 }, { "epoch": 0.57, "learning_rate": 8.208685973485343e-06, "loss": 2.1655, "step": 23150 }, { "epoch": 0.57, "learning_rate": 8.207900560947056e-06, "loss": 1.9327, "step": 23151 }, { "epoch": 0.57, "learning_rate": 8.207115159830405e-06, "loss": 1.9659, "step": 23152 }, { "epoch": 0.57, "learning_rate": 8.20632977014039e-06, "loss": 2.0777, "step": 23153 }, { "epoch": 0.57, "learning_rate": 8.20554439188202e-06, "loss": 1.9208, "step": 23154 }, { "epoch": 0.57, "learning_rate": 8.204759025060303e-06, "loss": 1.9965, "step": 23155 }, { "epoch": 0.57, "learning_rate": 8.203973669680241e-06, "loss": 2.1887, "step": 23156 }, { "epoch": 0.57, "learning_rate": 8.20318832574684e-06, "loss": 2.1911, "step": 23157 }, { "epoch": 0.57, "learning_rate": 8.202402993265103e-06, "loss": 1.8957, "step": 23158 }, { "epoch": 0.57, "learning_rate": 8.201617672240039e-06, "loss": 2.056, "step": 23159 }, { "epoch": 0.57, "learning_rate": 8.200832362676655e-06, "loss": 1.9681, "step": 23160 }, { "epoch": 0.57, "learning_rate": 8.20004706457995e-06, "loss": 2.1639, "step": 23161 }, { "epoch": 0.57, "learning_rate": 8.199261777954935e-06, "loss": 2.0226, "step": 23162 }, { "epoch": 0.57, "learning_rate": 8.198476502806607e-06, "loss": 2.0952, "step": 23163 }, { "epoch": 0.57, "learning_rate": 8.197691239139974e-06, "loss": 2.0946, "step": 23164 }, { "epoch": 0.57, "learning_rate": 8.196905986960045e-06, "loss": 1.9997, "step": 23165 }, { "epoch": 0.57, "learning_rate": 8.196120746271821e-06, "loss": 1.8574, "step": 23166 }, { "epoch": 0.57, "learning_rate": 8.19533551708031e-06, "loss": 1.9971, "step": 23167 }, { "epoch": 0.57, "learning_rate": 8.19455029939051e-06, "loss": 2.0281, "step": 23168 }, { "epoch": 0.57, "learning_rate": 8.19376509320743e-06, "loss": 2.1107, "step": 23169 }, { "epoch": 0.57, "learning_rate": 8.192979898536078e-06, "loss": 1.9911, "step": 23170 }, { "epoch": 0.57, "learning_rate": 8.192194715381448e-06, "loss": 1.9336, "step": 23171 }, { "epoch": 0.57, "learning_rate": 8.191409543748554e-06, "loss": 2.1156, "step": 23172 }, { "epoch": 0.57, "learning_rate": 8.190624383642396e-06, "loss": 1.8625, "step": 23173 }, { "epoch": 0.57, "learning_rate": 8.189839235067976e-06, "loss": 2.0772, "step": 23174 }, { "epoch": 0.57, "learning_rate": 8.189054098030303e-06, "loss": 2.1575, "step": 23175 }, { "epoch": 0.57, "learning_rate": 8.188268972534377e-06, "loss": 1.9945, "step": 23176 }, { "epoch": 0.57, "learning_rate": 8.187483858585202e-06, "loss": 1.9976, "step": 23177 }, { "epoch": 0.57, "learning_rate": 8.186698756187788e-06, "loss": 1.9217, "step": 23178 }, { "epoch": 0.57, "learning_rate": 8.185913665347129e-06, "loss": 2.0183, "step": 23179 }, { "epoch": 0.57, "learning_rate": 8.185128586068238e-06, "loss": 1.8705, "step": 23180 }, { "epoch": 0.57, "learning_rate": 8.18434351835611e-06, "loss": 1.9332, "step": 23181 }, { "epoch": 0.57, "learning_rate": 8.183558462215753e-06, "loss": 2.0422, "step": 23182 }, { "epoch": 0.57, "learning_rate": 8.18277341765217e-06, "loss": 2.1007, "step": 23183 }, { "epoch": 0.57, "learning_rate": 8.181988384670363e-06, "loss": 1.9897, "step": 23184 }, { "epoch": 0.57, "learning_rate": 8.181203363275342e-06, "loss": 1.9993, "step": 23185 }, { "epoch": 0.57, "learning_rate": 8.180418353472099e-06, "loss": 2.0218, "step": 23186 }, { "epoch": 0.57, "learning_rate": 8.179633355265644e-06, "loss": 1.9753, "step": 23187 }, { "epoch": 0.57, "learning_rate": 8.178848368660982e-06, "loss": 1.8818, "step": 23188 }, { "epoch": 0.57, "learning_rate": 8.178063393663108e-06, "loss": 2.0339, "step": 23189 }, { "epoch": 0.57, "learning_rate": 8.177278430277033e-06, "loss": 2.0063, "step": 23190 }, { "epoch": 0.57, "learning_rate": 8.176493478507756e-06, "loss": 2.0403, "step": 23191 }, { "epoch": 0.57, "learning_rate": 8.17570853836028e-06, "loss": 2.0722, "step": 23192 }, { "epoch": 0.57, "learning_rate": 8.174923609839609e-06, "loss": 1.9695, "step": 23193 }, { "epoch": 0.57, "learning_rate": 8.174138692950742e-06, "loss": 1.9259, "step": 23194 }, { "epoch": 0.57, "learning_rate": 8.173353787698688e-06, "loss": 1.9164, "step": 23195 }, { "epoch": 0.57, "learning_rate": 8.172568894088442e-06, "loss": 1.8637, "step": 23196 }, { "epoch": 0.57, "learning_rate": 8.171784012125009e-06, "loss": 1.8804, "step": 23197 }, { "epoch": 0.57, "learning_rate": 8.170999141813398e-06, "loss": 1.8169, "step": 23198 }, { "epoch": 0.57, "learning_rate": 8.170214283158601e-06, "loss": 2.0227, "step": 23199 }, { "epoch": 0.57, "learning_rate": 8.169429436165626e-06, "loss": 1.9424, "step": 23200 }, { "epoch": 0.57, "learning_rate": 8.168644600839472e-06, "loss": 2.1714, "step": 23201 }, { "epoch": 0.57, "learning_rate": 8.167859777185143e-06, "loss": 2.0941, "step": 23202 }, { "epoch": 0.57, "learning_rate": 8.167074965207646e-06, "loss": 2.1363, "step": 23203 }, { "epoch": 0.57, "learning_rate": 8.166290164911971e-06, "loss": 2.1458, "step": 23204 }, { "epoch": 0.57, "learning_rate": 8.165505376303132e-06, "loss": 2.0712, "step": 23205 }, { "epoch": 0.57, "learning_rate": 8.16472059938612e-06, "loss": 2.0116, "step": 23206 }, { "epoch": 0.57, "learning_rate": 8.163935834165943e-06, "loss": 2.0722, "step": 23207 }, { "epoch": 0.57, "learning_rate": 8.163151080647604e-06, "loss": 1.832, "step": 23208 }, { "epoch": 0.57, "learning_rate": 8.1623663388361e-06, "loss": 2.1455, "step": 23209 }, { "epoch": 0.57, "learning_rate": 8.161581608736436e-06, "loss": 2.0423, "step": 23210 }, { "epoch": 0.57, "learning_rate": 8.160796890353608e-06, "loss": 2.032, "step": 23211 }, { "epoch": 0.57, "learning_rate": 8.160012183692621e-06, "loss": 1.9363, "step": 23212 }, { "epoch": 0.57, "learning_rate": 8.159227488758479e-06, "loss": 2.0677, "step": 23213 }, { "epoch": 0.57, "learning_rate": 8.158442805556178e-06, "loss": 1.9945, "step": 23214 }, { "epoch": 0.57, "learning_rate": 8.157658134090725e-06, "loss": 2.0744, "step": 23215 }, { "epoch": 0.57, "learning_rate": 8.15687347436711e-06, "loss": 2.0454, "step": 23216 }, { "epoch": 0.57, "learning_rate": 8.156088826390345e-06, "loss": 2.05, "step": 23217 }, { "epoch": 0.57, "learning_rate": 8.155304190165428e-06, "loss": 1.9334, "step": 23218 }, { "epoch": 0.57, "learning_rate": 8.154519565697354e-06, "loss": 2.2005, "step": 23219 }, { "epoch": 0.57, "learning_rate": 8.153734952991135e-06, "loss": 1.8805, "step": 23220 }, { "epoch": 0.57, "learning_rate": 8.152950352051761e-06, "loss": 1.951, "step": 23221 }, { "epoch": 0.57, "learning_rate": 8.152165762884234e-06, "loss": 1.9506, "step": 23222 }, { "epoch": 0.57, "learning_rate": 8.151381185493562e-06, "loss": 1.9373, "step": 23223 }, { "epoch": 0.57, "learning_rate": 8.150596619884737e-06, "loss": 2.0803, "step": 23224 }, { "epoch": 0.57, "learning_rate": 8.149812066062763e-06, "loss": 1.93, "step": 23225 }, { "epoch": 0.57, "learning_rate": 8.149027524032638e-06, "loss": 1.9963, "step": 23226 }, { "epoch": 0.57, "learning_rate": 8.148242993799367e-06, "loss": 2.0362, "step": 23227 }, { "epoch": 0.57, "learning_rate": 8.147458475367946e-06, "loss": 1.9885, "step": 23228 }, { "epoch": 0.57, "learning_rate": 8.146673968743373e-06, "loss": 2.0386, "step": 23229 }, { "epoch": 0.57, "learning_rate": 8.145889473930656e-06, "loss": 1.7797, "step": 23230 }, { "epoch": 0.57, "learning_rate": 8.145104990934785e-06, "loss": 2.0958, "step": 23231 }, { "epoch": 0.57, "learning_rate": 8.144320519760764e-06, "loss": 1.9812, "step": 23232 }, { "epoch": 0.57, "learning_rate": 8.143536060413598e-06, "loss": 2.0305, "step": 23233 }, { "epoch": 0.57, "learning_rate": 8.142751612898277e-06, "loss": 1.9276, "step": 23234 }, { "epoch": 0.57, "learning_rate": 8.141967177219807e-06, "loss": 1.885, "step": 23235 }, { "epoch": 0.57, "learning_rate": 8.141182753383186e-06, "loss": 1.925, "step": 23236 }, { "epoch": 0.57, "learning_rate": 8.140398341393411e-06, "loss": 1.8292, "step": 23237 }, { "epoch": 0.57, "learning_rate": 8.139613941255488e-06, "loss": 1.9136, "step": 23238 }, { "epoch": 0.57, "learning_rate": 8.138829552974408e-06, "loss": 1.9726, "step": 23239 }, { "epoch": 0.57, "learning_rate": 8.138045176555172e-06, "loss": 1.8738, "step": 23240 }, { "epoch": 0.57, "learning_rate": 8.137260812002786e-06, "loss": 1.8714, "step": 23241 }, { "epoch": 0.57, "learning_rate": 8.13647645932224e-06, "loss": 2.0679, "step": 23242 }, { "epoch": 0.57, "learning_rate": 8.13569211851854e-06, "loss": 1.9617, "step": 23243 }, { "epoch": 0.57, "learning_rate": 8.134907789596678e-06, "loss": 1.9738, "step": 23244 }, { "epoch": 0.57, "learning_rate": 8.134123472561658e-06, "loss": 2.0229, "step": 23245 }, { "epoch": 0.57, "learning_rate": 8.13333916741848e-06, "loss": 1.9429, "step": 23246 }, { "epoch": 0.57, "learning_rate": 8.132554874172135e-06, "loss": 1.8133, "step": 23247 }, { "epoch": 0.57, "learning_rate": 8.131770592827632e-06, "loss": 1.8644, "step": 23248 }, { "epoch": 0.57, "learning_rate": 8.130986323389958e-06, "loss": 2.0082, "step": 23249 }, { "epoch": 0.57, "learning_rate": 8.13020206586412e-06, "loss": 2.1128, "step": 23250 }, { "epoch": 0.57, "learning_rate": 8.129417820255114e-06, "loss": 2.108, "step": 23251 }, { "epoch": 0.57, "learning_rate": 8.128633586567936e-06, "loss": 1.9425, "step": 23252 }, { "epoch": 0.57, "learning_rate": 8.127849364807588e-06, "loss": 1.9694, "step": 23253 }, { "epoch": 0.57, "learning_rate": 8.127065154979065e-06, "loss": 2.0518, "step": 23254 }, { "epoch": 0.57, "learning_rate": 8.126280957087365e-06, "loss": 2.1774, "step": 23255 }, { "epoch": 0.57, "learning_rate": 8.125496771137492e-06, "loss": 1.9572, "step": 23256 }, { "epoch": 0.57, "learning_rate": 8.124712597134433e-06, "loss": 2.0451, "step": 23257 }, { "epoch": 0.57, "learning_rate": 8.123928435083197e-06, "loss": 2.1757, "step": 23258 }, { "epoch": 0.57, "learning_rate": 8.123144284988772e-06, "loss": 1.8948, "step": 23259 }, { "epoch": 0.57, "learning_rate": 8.122360146856161e-06, "loss": 1.8372, "step": 23260 }, { "epoch": 0.57, "learning_rate": 8.121576020690363e-06, "loss": 2.0116, "step": 23261 }, { "epoch": 0.57, "learning_rate": 8.120791906496372e-06, "loss": 2.1397, "step": 23262 }, { "epoch": 0.57, "learning_rate": 8.120007804279186e-06, "loss": 1.917, "step": 23263 }, { "epoch": 0.57, "learning_rate": 8.119223714043802e-06, "loss": 2.1335, "step": 23264 }, { "epoch": 0.57, "learning_rate": 8.118439635795219e-06, "loss": 2.065, "step": 23265 }, { "epoch": 0.57, "learning_rate": 8.117655569538435e-06, "loss": 2.0069, "step": 23266 }, { "epoch": 0.57, "learning_rate": 8.116871515278444e-06, "loss": 2.0064, "step": 23267 }, { "epoch": 0.57, "learning_rate": 8.116087473020248e-06, "loss": 2.0201, "step": 23268 }, { "epoch": 0.57, "learning_rate": 8.115303442768836e-06, "loss": 2.0035, "step": 23269 }, { "epoch": 0.57, "learning_rate": 8.114519424529209e-06, "loss": 1.947, "step": 23270 }, { "epoch": 0.57, "learning_rate": 8.113735418306366e-06, "loss": 1.9119, "step": 23271 }, { "epoch": 0.57, "learning_rate": 8.112951424105301e-06, "loss": 1.9173, "step": 23272 }, { "epoch": 0.57, "learning_rate": 8.112167441931015e-06, "loss": 1.9927, "step": 23273 }, { "epoch": 0.57, "learning_rate": 8.111383471788497e-06, "loss": 1.7977, "step": 23274 }, { "epoch": 0.57, "learning_rate": 8.110599513682749e-06, "loss": 2.1431, "step": 23275 }, { "epoch": 0.57, "learning_rate": 8.109815567618768e-06, "loss": 1.956, "step": 23276 }, { "epoch": 0.57, "learning_rate": 8.109031633601546e-06, "loss": 1.9668, "step": 23277 }, { "epoch": 0.57, "learning_rate": 8.10824771163608e-06, "loss": 2.0395, "step": 23278 }, { "epoch": 0.57, "learning_rate": 8.107463801727371e-06, "loss": 1.9012, "step": 23279 }, { "epoch": 0.57, "learning_rate": 8.10667990388041e-06, "loss": 1.991, "step": 23280 }, { "epoch": 0.57, "learning_rate": 8.105896018100198e-06, "loss": 1.8034, "step": 23281 }, { "epoch": 0.57, "learning_rate": 8.105112144391725e-06, "loss": 2.0994, "step": 23282 }, { "epoch": 0.57, "learning_rate": 8.104328282759992e-06, "loss": 2.0012, "step": 23283 }, { "epoch": 0.57, "learning_rate": 8.103544433209991e-06, "loss": 1.9957, "step": 23284 }, { "epoch": 0.57, "learning_rate": 8.102760595746719e-06, "loss": 2.1218, "step": 23285 }, { "epoch": 0.57, "learning_rate": 8.101976770375175e-06, "loss": 2.1809, "step": 23286 }, { "epoch": 0.57, "learning_rate": 8.101192957100348e-06, "loss": 1.9417, "step": 23287 }, { "epoch": 0.57, "learning_rate": 8.100409155927238e-06, "loss": 2.0679, "step": 23288 }, { "epoch": 0.57, "learning_rate": 8.09962536686084e-06, "loss": 1.9032, "step": 23289 }, { "epoch": 0.57, "learning_rate": 8.098841589906145e-06, "loss": 1.9491, "step": 23290 }, { "epoch": 0.57, "learning_rate": 8.09805782506816e-06, "loss": 1.9069, "step": 23291 }, { "epoch": 0.57, "learning_rate": 8.097274072351865e-06, "loss": 2.1214, "step": 23292 }, { "epoch": 0.57, "learning_rate": 8.09649033176227e-06, "loss": 1.9637, "step": 23293 }, { "epoch": 0.57, "learning_rate": 8.095706603304356e-06, "loss": 1.9749, "step": 23294 }, { "epoch": 0.57, "learning_rate": 8.094922886983125e-06, "loss": 1.9904, "step": 23295 }, { "epoch": 0.57, "learning_rate": 8.094139182803574e-06, "loss": 1.9774, "step": 23296 }, { "epoch": 0.57, "learning_rate": 8.093355490770693e-06, "loss": 1.9318, "step": 23297 }, { "epoch": 0.57, "learning_rate": 8.09257181088948e-06, "loss": 2.0662, "step": 23298 }, { "epoch": 0.57, "learning_rate": 8.09178814316493e-06, "loss": 1.9192, "step": 23299 }, { "epoch": 0.57, "learning_rate": 8.091004487602033e-06, "loss": 1.9684, "step": 23300 }, { "epoch": 0.57, "learning_rate": 8.090220844205789e-06, "loss": 2.0543, "step": 23301 }, { "epoch": 0.57, "learning_rate": 8.089437212981187e-06, "loss": 1.9064, "step": 23302 }, { "epoch": 0.57, "learning_rate": 8.088653593933225e-06, "loss": 2.0308, "step": 23303 }, { "epoch": 0.57, "learning_rate": 8.0878699870669e-06, "loss": 2.0736, "step": 23304 }, { "epoch": 0.57, "learning_rate": 8.087086392387199e-06, "loss": 1.9175, "step": 23305 }, { "epoch": 0.57, "learning_rate": 8.08630280989912e-06, "loss": 2.1293, "step": 23306 }, { "epoch": 0.57, "learning_rate": 8.085519239607658e-06, "loss": 1.9768, "step": 23307 }, { "epoch": 0.57, "learning_rate": 8.084735681517803e-06, "loss": 1.8902, "step": 23308 }, { "epoch": 0.57, "learning_rate": 8.083952135634559e-06, "loss": 1.9703, "step": 23309 }, { "epoch": 0.57, "learning_rate": 8.083168601962905e-06, "loss": 2.1514, "step": 23310 }, { "epoch": 0.57, "learning_rate": 8.082385080507849e-06, "loss": 1.9638, "step": 23311 }, { "epoch": 0.57, "learning_rate": 8.081601571274371e-06, "loss": 1.8931, "step": 23312 }, { "epoch": 0.57, "learning_rate": 8.080818074267474e-06, "loss": 1.9892, "step": 23313 }, { "epoch": 0.57, "learning_rate": 8.080034589492148e-06, "loss": 2.1156, "step": 23314 }, { "epoch": 0.57, "learning_rate": 8.079251116953386e-06, "loss": 1.8945, "step": 23315 }, { "epoch": 0.57, "learning_rate": 8.078467656656189e-06, "loss": 2.0357, "step": 23316 }, { "epoch": 0.57, "learning_rate": 8.077684208605538e-06, "loss": 2.0146, "step": 23317 }, { "epoch": 0.57, "learning_rate": 8.076900772806432e-06, "loss": 1.9631, "step": 23318 }, { "epoch": 0.57, "learning_rate": 8.076117349263868e-06, "loss": 1.9794, "step": 23319 }, { "epoch": 0.57, "learning_rate": 8.075333937982833e-06, "loss": 1.9252, "step": 23320 }, { "epoch": 0.57, "learning_rate": 8.074550538968324e-06, "loss": 2.104, "step": 23321 }, { "epoch": 0.57, "learning_rate": 8.07376715222533e-06, "loss": 2.1975, "step": 23322 }, { "epoch": 0.57, "learning_rate": 8.072983777758843e-06, "loss": 1.9939, "step": 23323 }, { "epoch": 0.57, "learning_rate": 8.072200415573861e-06, "loss": 1.7806, "step": 23324 }, { "epoch": 0.57, "learning_rate": 8.071417065675372e-06, "loss": 2.0819, "step": 23325 }, { "epoch": 0.57, "learning_rate": 8.070633728068375e-06, "loss": 2.1035, "step": 23326 }, { "epoch": 0.57, "learning_rate": 8.069850402757855e-06, "loss": 2.0572, "step": 23327 }, { "epoch": 0.58, "learning_rate": 8.069067089748806e-06, "loss": 1.9323, "step": 23328 }, { "epoch": 0.58, "learning_rate": 8.068283789046226e-06, "loss": 2.2456, "step": 23329 }, { "epoch": 0.58, "learning_rate": 8.067500500655099e-06, "loss": 1.8191, "step": 23330 }, { "epoch": 0.58, "learning_rate": 8.066717224580423e-06, "loss": 1.8803, "step": 23331 }, { "epoch": 0.58, "learning_rate": 8.065933960827187e-06, "loss": 1.909, "step": 23332 }, { "epoch": 0.58, "learning_rate": 8.065150709400383e-06, "loss": 2.1918, "step": 23333 }, { "epoch": 0.58, "learning_rate": 8.064367470305008e-06, "loss": 2.0013, "step": 23334 }, { "epoch": 0.58, "learning_rate": 8.063584243546048e-06, "loss": 2.0277, "step": 23335 }, { "epoch": 0.58, "learning_rate": 8.0628010291285e-06, "loss": 1.8569, "step": 23336 }, { "epoch": 0.58, "learning_rate": 8.062017827057347e-06, "loss": 2.0698, "step": 23337 }, { "epoch": 0.58, "learning_rate": 8.061234637337586e-06, "loss": 1.9724, "step": 23338 }, { "epoch": 0.58, "learning_rate": 8.060451459974213e-06, "loss": 2.013, "step": 23339 }, { "epoch": 0.58, "learning_rate": 8.059668294972211e-06, "loss": 1.956, "step": 23340 }, { "epoch": 0.58, "learning_rate": 8.058885142336578e-06, "loss": 1.9449, "step": 23341 }, { "epoch": 0.58, "learning_rate": 8.058102002072299e-06, "loss": 1.8779, "step": 23342 }, { "epoch": 0.58, "learning_rate": 8.05731887418437e-06, "loss": 1.9787, "step": 23343 }, { "epoch": 0.58, "learning_rate": 8.056535758677784e-06, "loss": 1.9012, "step": 23344 }, { "epoch": 0.58, "learning_rate": 8.055752655557525e-06, "loss": 2.0105, "step": 23345 }, { "epoch": 0.58, "learning_rate": 8.054969564828593e-06, "loss": 1.8941, "step": 23346 }, { "epoch": 0.58, "learning_rate": 8.05418648649597e-06, "loss": 2.0389, "step": 23347 }, { "epoch": 0.58, "learning_rate": 8.05340342056465e-06, "loss": 1.8851, "step": 23348 }, { "epoch": 0.58, "learning_rate": 8.052620367039625e-06, "loss": 2.0955, "step": 23349 }, { "epoch": 0.58, "learning_rate": 8.051837325925884e-06, "loss": 2.0061, "step": 23350 }, { "epoch": 0.58, "learning_rate": 8.051054297228424e-06, "loss": 1.9804, "step": 23351 }, { "epoch": 0.58, "learning_rate": 8.050271280952225e-06, "loss": 1.9535, "step": 23352 }, { "epoch": 0.58, "learning_rate": 8.049488277102282e-06, "loss": 1.9925, "step": 23353 }, { "epoch": 0.58, "learning_rate": 8.048705285683591e-06, "loss": 1.972, "step": 23354 }, { "epoch": 0.58, "learning_rate": 8.047922306701133e-06, "loss": 1.977, "step": 23355 }, { "epoch": 0.58, "learning_rate": 8.0471393401599e-06, "loss": 2.003, "step": 23356 }, { "epoch": 0.58, "learning_rate": 8.046356386064891e-06, "loss": 2.1642, "step": 23357 }, { "epoch": 0.58, "learning_rate": 8.045573444421085e-06, "loss": 2.0684, "step": 23358 }, { "epoch": 0.58, "learning_rate": 8.044790515233478e-06, "loss": 1.9601, "step": 23359 }, { "epoch": 0.58, "learning_rate": 8.044007598507057e-06, "loss": 2.0314, "step": 23360 }, { "epoch": 0.58, "learning_rate": 8.043224694246811e-06, "loss": 2.1301, "step": 23361 }, { "epoch": 0.58, "learning_rate": 8.042441802457738e-06, "loss": 1.9735, "step": 23362 }, { "epoch": 0.58, "learning_rate": 8.041658923144815e-06, "loss": 1.9367, "step": 23363 }, { "epoch": 0.58, "learning_rate": 8.040876056313045e-06, "loss": 2.0367, "step": 23364 }, { "epoch": 0.58, "learning_rate": 8.040093201967404e-06, "loss": 1.8712, "step": 23365 }, { "epoch": 0.58, "learning_rate": 8.039310360112888e-06, "loss": 1.9318, "step": 23366 }, { "epoch": 0.58, "learning_rate": 8.038527530754489e-06, "loss": 1.9326, "step": 23367 }, { "epoch": 0.58, "learning_rate": 8.03774471389719e-06, "loss": 2.0768, "step": 23368 }, { "epoch": 0.58, "learning_rate": 8.036961909545989e-06, "loss": 2.091, "step": 23369 }, { "epoch": 0.58, "learning_rate": 8.036179117705864e-06, "loss": 1.9773, "step": 23370 }, { "epoch": 0.58, "learning_rate": 8.035396338381809e-06, "loss": 2.0575, "step": 23371 }, { "epoch": 0.58, "learning_rate": 8.034613571578819e-06, "loss": 2.0192, "step": 23372 }, { "epoch": 0.58, "learning_rate": 8.033830817301874e-06, "loss": 1.9391, "step": 23373 }, { "epoch": 0.58, "learning_rate": 8.033048075555966e-06, "loss": 2.217, "step": 23374 }, { "epoch": 0.58, "learning_rate": 8.032265346346084e-06, "loss": 1.9208, "step": 23375 }, { "epoch": 0.58, "learning_rate": 8.031482629677214e-06, "loss": 2.1063, "step": 23376 }, { "epoch": 0.58, "learning_rate": 8.030699925554348e-06, "loss": 2.0479, "step": 23377 }, { "epoch": 0.58, "learning_rate": 8.029917233982473e-06, "loss": 1.9393, "step": 23378 }, { "epoch": 0.58, "learning_rate": 8.02913455496658e-06, "loss": 2.0074, "step": 23379 }, { "epoch": 0.58, "learning_rate": 8.028351888511651e-06, "loss": 2.0579, "step": 23380 }, { "epoch": 0.58, "learning_rate": 8.02756923462268e-06, "loss": 1.9422, "step": 23381 }, { "epoch": 0.58, "learning_rate": 8.026786593304656e-06, "loss": 1.9403, "step": 23382 }, { "epoch": 0.58, "learning_rate": 8.026003964562559e-06, "loss": 2.0739, "step": 23383 }, { "epoch": 0.58, "learning_rate": 8.025221348401383e-06, "loss": 2.0797, "step": 23384 }, { "epoch": 0.58, "learning_rate": 8.024438744826116e-06, "loss": 1.931, "step": 23385 }, { "epoch": 0.58, "learning_rate": 8.023656153841743e-06, "loss": 2.0728, "step": 23386 }, { "epoch": 0.58, "learning_rate": 8.022873575453258e-06, "loss": 2.072, "step": 23387 }, { "epoch": 0.58, "learning_rate": 8.02209100966564e-06, "loss": 1.9678, "step": 23388 }, { "epoch": 0.58, "learning_rate": 8.021308456483884e-06, "loss": 2.0608, "step": 23389 }, { "epoch": 0.58, "learning_rate": 8.020525915912971e-06, "loss": 2.1069, "step": 23390 }, { "epoch": 0.58, "learning_rate": 8.01974338795789e-06, "loss": 1.9722, "step": 23391 }, { "epoch": 0.58, "learning_rate": 8.018960872623635e-06, "loss": 1.9901, "step": 23392 }, { "epoch": 0.58, "learning_rate": 8.018178369915184e-06, "loss": 1.944, "step": 23393 }, { "epoch": 0.58, "learning_rate": 8.017395879837529e-06, "loss": 2.0201, "step": 23394 }, { "epoch": 0.58, "learning_rate": 8.016613402395654e-06, "loss": 1.8144, "step": 23395 }, { "epoch": 0.58, "learning_rate": 8.01583093759455e-06, "loss": 2.0435, "step": 23396 }, { "epoch": 0.58, "learning_rate": 8.015048485439205e-06, "loss": 2.158, "step": 23397 }, { "epoch": 0.58, "learning_rate": 8.0142660459346e-06, "loss": 2.1392, "step": 23398 }, { "epoch": 0.58, "learning_rate": 8.013483619085726e-06, "loss": 1.9268, "step": 23399 }, { "epoch": 0.58, "learning_rate": 8.012701204897566e-06, "loss": 1.9746, "step": 23400 }, { "epoch": 0.58, "learning_rate": 8.011918803375109e-06, "loss": 2.1599, "step": 23401 }, { "epoch": 0.58, "learning_rate": 8.011136414523342e-06, "loss": 1.8742, "step": 23402 }, { "epoch": 0.58, "learning_rate": 8.01035403834725e-06, "loss": 1.9305, "step": 23403 }, { "epoch": 0.58, "learning_rate": 8.009571674851824e-06, "loss": 2.0231, "step": 23404 }, { "epoch": 0.58, "learning_rate": 8.008789324042043e-06, "loss": 2.0804, "step": 23405 }, { "epoch": 0.58, "learning_rate": 8.008006985922895e-06, "loss": 2.0686, "step": 23406 }, { "epoch": 0.58, "learning_rate": 8.007224660499373e-06, "loss": 2.0135, "step": 23407 }, { "epoch": 0.58, "learning_rate": 8.006442347776453e-06, "loss": 2.0129, "step": 23408 }, { "epoch": 0.58, "learning_rate": 8.005660047759127e-06, "loss": 2.0436, "step": 23409 }, { "epoch": 0.58, "learning_rate": 8.00487776045238e-06, "loss": 2.006, "step": 23410 }, { "epoch": 0.58, "learning_rate": 8.004095485861197e-06, "loss": 1.9604, "step": 23411 }, { "epoch": 0.58, "learning_rate": 8.003313223990564e-06, "loss": 1.9917, "step": 23412 }, { "epoch": 0.58, "learning_rate": 8.002530974845464e-06, "loss": 1.9234, "step": 23413 }, { "epoch": 0.58, "learning_rate": 8.001748738430885e-06, "loss": 2.1097, "step": 23414 }, { "epoch": 0.58, "learning_rate": 8.000966514751818e-06, "loss": 1.8902, "step": 23415 }, { "epoch": 0.58, "learning_rate": 8.000184303813238e-06, "loss": 2.0082, "step": 23416 }, { "epoch": 0.58, "learning_rate": 7.99940210562014e-06, "loss": 2.1496, "step": 23417 }, { "epoch": 0.58, "learning_rate": 7.9986199201775e-06, "loss": 1.9761, "step": 23418 }, { "epoch": 0.58, "learning_rate": 7.997837747490307e-06, "loss": 1.9914, "step": 23419 }, { "epoch": 0.58, "learning_rate": 7.997055587563549e-06, "loss": 2.0686, "step": 23420 }, { "epoch": 0.58, "learning_rate": 7.996273440402205e-06, "loss": 1.9646, "step": 23421 }, { "epoch": 0.58, "learning_rate": 7.99549130601127e-06, "loss": 1.9433, "step": 23422 }, { "epoch": 0.58, "learning_rate": 7.994709184395716e-06, "loss": 2.0639, "step": 23423 }, { "epoch": 0.58, "learning_rate": 7.993927075560535e-06, "loss": 2.1875, "step": 23424 }, { "epoch": 0.58, "learning_rate": 7.993144979510715e-06, "loss": 2.0451, "step": 23425 }, { "epoch": 0.58, "learning_rate": 7.99236289625123e-06, "loss": 2.1096, "step": 23426 }, { "epoch": 0.58, "learning_rate": 7.991580825787074e-06, "loss": 2.1683, "step": 23427 }, { "epoch": 0.58, "learning_rate": 7.990798768123228e-06, "loss": 2.0161, "step": 23428 }, { "epoch": 0.58, "learning_rate": 7.990016723264674e-06, "loss": 1.9308, "step": 23429 }, { "epoch": 0.58, "learning_rate": 7.9892346912164e-06, "loss": 2.0348, "step": 23430 }, { "epoch": 0.58, "learning_rate": 7.988452671983387e-06, "loss": 2.0337, "step": 23431 }, { "epoch": 0.58, "learning_rate": 7.987670665570625e-06, "loss": 2.0136, "step": 23432 }, { "epoch": 0.58, "learning_rate": 7.986888671983089e-06, "loss": 2.0784, "step": 23433 }, { "epoch": 0.58, "learning_rate": 7.986106691225768e-06, "loss": 1.9314, "step": 23434 }, { "epoch": 0.58, "learning_rate": 7.98532472330365e-06, "loss": 1.993, "step": 23435 }, { "epoch": 0.58, "learning_rate": 7.984542768221708e-06, "loss": 1.928, "step": 23436 }, { "epoch": 0.58, "learning_rate": 7.983760825984936e-06, "loss": 2.0095, "step": 23437 }, { "epoch": 0.58, "learning_rate": 7.98297889659831e-06, "loss": 1.976, "step": 23438 }, { "epoch": 0.58, "learning_rate": 7.982196980066816e-06, "loss": 1.905, "step": 23439 }, { "epoch": 0.58, "learning_rate": 7.981415076395445e-06, "loss": 2.1263, "step": 23440 }, { "epoch": 0.58, "learning_rate": 7.980633185589167e-06, "loss": 1.8068, "step": 23441 }, { "epoch": 0.58, "learning_rate": 7.979851307652977e-06, "loss": 1.9634, "step": 23442 }, { "epoch": 0.58, "learning_rate": 7.979069442591848e-06, "loss": 1.8124, "step": 23443 }, { "epoch": 0.58, "learning_rate": 7.978287590410769e-06, "loss": 1.9416, "step": 23444 }, { "epoch": 0.58, "learning_rate": 7.977505751114724e-06, "loss": 1.9827, "step": 23445 }, { "epoch": 0.58, "learning_rate": 7.976723924708695e-06, "loss": 1.8709, "step": 23446 }, { "epoch": 0.58, "learning_rate": 7.975942111197661e-06, "loss": 1.9477, "step": 23447 }, { "epoch": 0.58, "learning_rate": 7.975160310586609e-06, "loss": 2.0488, "step": 23448 }, { "epoch": 0.58, "learning_rate": 7.974378522880517e-06, "loss": 1.931, "step": 23449 }, { "epoch": 0.58, "learning_rate": 7.973596748084376e-06, "loss": 2.077, "step": 23450 }, { "epoch": 0.58, "learning_rate": 7.97281498620316e-06, "loss": 1.8669, "step": 23451 }, { "epoch": 0.58, "learning_rate": 7.972033237241858e-06, "loss": 1.8949, "step": 23452 }, { "epoch": 0.58, "learning_rate": 7.971251501205447e-06, "loss": 2.1406, "step": 23453 }, { "epoch": 0.58, "learning_rate": 7.970469778098908e-06, "loss": 1.9987, "step": 23454 }, { "epoch": 0.58, "learning_rate": 7.969688067927231e-06, "loss": 2.0155, "step": 23455 }, { "epoch": 0.58, "learning_rate": 7.96890637069539e-06, "loss": 1.8855, "step": 23456 }, { "epoch": 0.58, "learning_rate": 7.968124686408377e-06, "loss": 1.9444, "step": 23457 }, { "epoch": 0.58, "learning_rate": 7.96734301507116e-06, "loss": 1.8758, "step": 23458 }, { "epoch": 0.58, "learning_rate": 7.96656135668873e-06, "loss": 2.063, "step": 23459 }, { "epoch": 0.58, "learning_rate": 7.965779711266071e-06, "loss": 2.1054, "step": 23460 }, { "epoch": 0.58, "learning_rate": 7.964998078808158e-06, "loss": 2.0927, "step": 23461 }, { "epoch": 0.58, "learning_rate": 7.964216459319976e-06, "loss": 2.0665, "step": 23462 }, { "epoch": 0.58, "learning_rate": 7.963434852806507e-06, "loss": 2.1576, "step": 23463 }, { "epoch": 0.58, "learning_rate": 7.962653259272728e-06, "loss": 2.0013, "step": 23464 }, { "epoch": 0.58, "learning_rate": 7.961871678723628e-06, "loss": 2.0822, "step": 23465 }, { "epoch": 0.58, "learning_rate": 7.96109011116418e-06, "loss": 2.067, "step": 23466 }, { "epoch": 0.58, "learning_rate": 7.960308556599376e-06, "loss": 2.0969, "step": 23467 }, { "epoch": 0.58, "learning_rate": 7.959527015034184e-06, "loss": 2.0156, "step": 23468 }, { "epoch": 0.58, "learning_rate": 7.95874548647359e-06, "loss": 1.8487, "step": 23469 }, { "epoch": 0.58, "learning_rate": 7.957963970922583e-06, "loss": 2.0792, "step": 23470 }, { "epoch": 0.58, "learning_rate": 7.957182468386131e-06, "loss": 2.0339, "step": 23471 }, { "epoch": 0.58, "learning_rate": 7.956400978869225e-06, "loss": 1.6843, "step": 23472 }, { "epoch": 0.58, "learning_rate": 7.955619502376836e-06, "loss": 2.1486, "step": 23473 }, { "epoch": 0.58, "learning_rate": 7.954838038913953e-06, "loss": 1.9977, "step": 23474 }, { "epoch": 0.58, "learning_rate": 7.954056588485558e-06, "loss": 2.2014, "step": 23475 }, { "epoch": 0.58, "learning_rate": 7.953275151096623e-06, "loss": 1.9346, "step": 23476 }, { "epoch": 0.58, "learning_rate": 7.95249372675213e-06, "loss": 2.0121, "step": 23477 }, { "epoch": 0.58, "learning_rate": 7.951712315457067e-06, "loss": 1.847, "step": 23478 }, { "epoch": 0.58, "learning_rate": 7.950930917216406e-06, "loss": 1.7413, "step": 23479 }, { "epoch": 0.58, "learning_rate": 7.950149532035132e-06, "loss": 2.0702, "step": 23480 }, { "epoch": 0.58, "learning_rate": 7.949368159918223e-06, "loss": 1.9779, "step": 23481 }, { "epoch": 0.58, "learning_rate": 7.948586800870655e-06, "loss": 2.1241, "step": 23482 }, { "epoch": 0.58, "learning_rate": 7.947805454897415e-06, "loss": 1.9485, "step": 23483 }, { "epoch": 0.58, "learning_rate": 7.947024122003479e-06, "loss": 2.1825, "step": 23484 }, { "epoch": 0.58, "learning_rate": 7.94624280219383e-06, "loss": 2.0818, "step": 23485 }, { "epoch": 0.58, "learning_rate": 7.94546149547344e-06, "loss": 2.0554, "step": 23486 }, { "epoch": 0.58, "learning_rate": 7.944680201847296e-06, "loss": 1.9488, "step": 23487 }, { "epoch": 0.58, "learning_rate": 7.943898921320378e-06, "loss": 1.9512, "step": 23488 }, { "epoch": 0.58, "learning_rate": 7.943117653897657e-06, "loss": 2.137, "step": 23489 }, { "epoch": 0.58, "learning_rate": 7.94233639958412e-06, "loss": 1.9349, "step": 23490 }, { "epoch": 0.58, "learning_rate": 7.941555158384742e-06, "loss": 1.9509, "step": 23491 }, { "epoch": 0.58, "learning_rate": 7.940773930304504e-06, "loss": 2.0784, "step": 23492 }, { "epoch": 0.58, "learning_rate": 7.939992715348387e-06, "loss": 1.9588, "step": 23493 }, { "epoch": 0.58, "learning_rate": 7.939211513521366e-06, "loss": 1.9543, "step": 23494 }, { "epoch": 0.58, "learning_rate": 7.938430324828425e-06, "loss": 1.949, "step": 23495 }, { "epoch": 0.58, "learning_rate": 7.937649149274536e-06, "loss": 2.0174, "step": 23496 }, { "epoch": 0.58, "learning_rate": 7.93686798686468e-06, "loss": 2.1354, "step": 23497 }, { "epoch": 0.58, "learning_rate": 7.936086837603841e-06, "loss": 1.9603, "step": 23498 }, { "epoch": 0.58, "learning_rate": 7.93530570149699e-06, "loss": 2.0384, "step": 23499 }, { "epoch": 0.58, "learning_rate": 7.93452457854911e-06, "loss": 1.9537, "step": 23500 }, { "epoch": 0.58, "learning_rate": 7.933743468765176e-06, "loss": 2.0078, "step": 23501 }, { "epoch": 0.58, "learning_rate": 7.932962372150167e-06, "loss": 1.8125, "step": 23502 }, { "epoch": 0.58, "learning_rate": 7.932181288709069e-06, "loss": 2.1203, "step": 23503 }, { "epoch": 0.58, "learning_rate": 7.931400218446848e-06, "loss": 1.997, "step": 23504 }, { "epoch": 0.58, "learning_rate": 7.930619161368491e-06, "loss": 1.9213, "step": 23505 }, { "epoch": 0.58, "learning_rate": 7.929838117478969e-06, "loss": 2.1174, "step": 23506 }, { "epoch": 0.58, "learning_rate": 7.929057086783264e-06, "loss": 1.9549, "step": 23507 }, { "epoch": 0.58, "learning_rate": 7.928276069286354e-06, "loss": 1.9594, "step": 23508 }, { "epoch": 0.58, "learning_rate": 7.927495064993215e-06, "loss": 2.1727, "step": 23509 }, { "epoch": 0.58, "learning_rate": 7.926714073908827e-06, "loss": 2.0618, "step": 23510 }, { "epoch": 0.58, "learning_rate": 7.925933096038166e-06, "loss": 1.8183, "step": 23511 }, { "epoch": 0.58, "learning_rate": 7.925152131386206e-06, "loss": 2.0821, "step": 23512 }, { "epoch": 0.58, "learning_rate": 7.924371179957932e-06, "loss": 2.0173, "step": 23513 }, { "epoch": 0.58, "learning_rate": 7.923590241758313e-06, "loss": 2.051, "step": 23514 }, { "epoch": 0.58, "learning_rate": 7.922809316792333e-06, "loss": 2.1131, "step": 23515 }, { "epoch": 0.58, "learning_rate": 7.922028405064963e-06, "loss": 1.9572, "step": 23516 }, { "epoch": 0.58, "learning_rate": 7.921247506581186e-06, "loss": 2.0963, "step": 23517 }, { "epoch": 0.58, "learning_rate": 7.920466621345975e-06, "loss": 2.0369, "step": 23518 }, { "epoch": 0.58, "learning_rate": 7.919685749364308e-06, "loss": 2.1305, "step": 23519 }, { "epoch": 0.58, "learning_rate": 7.918904890641165e-06, "loss": 2.0691, "step": 23520 }, { "epoch": 0.58, "learning_rate": 7.918124045181515e-06, "loss": 2.0362, "step": 23521 }, { "epoch": 0.58, "learning_rate": 7.917343212990338e-06, "loss": 1.8941, "step": 23522 }, { "epoch": 0.58, "learning_rate": 7.916562394072615e-06, "loss": 1.9489, "step": 23523 }, { "epoch": 0.58, "learning_rate": 7.915781588433317e-06, "loss": 2.0355, "step": 23524 }, { "epoch": 0.58, "learning_rate": 7.915000796077423e-06, "loss": 1.8191, "step": 23525 }, { "epoch": 0.58, "learning_rate": 7.914220017009907e-06, "loss": 2.0714, "step": 23526 }, { "epoch": 0.58, "learning_rate": 7.913439251235747e-06, "loss": 1.9722, "step": 23527 }, { "epoch": 0.58, "learning_rate": 7.91265849875992e-06, "loss": 2.1234, "step": 23528 }, { "epoch": 0.58, "learning_rate": 7.911877759587399e-06, "loss": 2.0021, "step": 23529 }, { "epoch": 0.58, "learning_rate": 7.911097033723165e-06, "loss": 1.9216, "step": 23530 }, { "epoch": 0.58, "learning_rate": 7.910316321172186e-06, "loss": 2.0168, "step": 23531 }, { "epoch": 0.58, "learning_rate": 7.909535621939442e-06, "loss": 2.0178, "step": 23532 }, { "epoch": 0.58, "learning_rate": 7.908754936029912e-06, "loss": 1.9367, "step": 23533 }, { "epoch": 0.58, "learning_rate": 7.907974263448565e-06, "loss": 1.9487, "step": 23534 }, { "epoch": 0.58, "learning_rate": 7.907193604200382e-06, "loss": 1.9773, "step": 23535 }, { "epoch": 0.58, "learning_rate": 7.906412958290337e-06, "loss": 1.9726, "step": 23536 }, { "epoch": 0.58, "learning_rate": 7.905632325723401e-06, "loss": 1.9978, "step": 23537 }, { "epoch": 0.58, "learning_rate": 7.904851706504557e-06, "loss": 1.9338, "step": 23538 }, { "epoch": 0.58, "learning_rate": 7.904071100638773e-06, "loss": 2.0898, "step": 23539 }, { "epoch": 0.58, "learning_rate": 7.903290508131027e-06, "loss": 2.0417, "step": 23540 }, { "epoch": 0.58, "learning_rate": 7.902509928986298e-06, "loss": 1.988, "step": 23541 }, { "epoch": 0.58, "learning_rate": 7.901729363209551e-06, "loss": 1.9139, "step": 23542 }, { "epoch": 0.58, "learning_rate": 7.90094881080577e-06, "loss": 2.1413, "step": 23543 }, { "epoch": 0.58, "learning_rate": 7.900168271779923e-06, "loss": 2.1054, "step": 23544 }, { "epoch": 0.58, "learning_rate": 7.89938774613699e-06, "loss": 1.9637, "step": 23545 }, { "epoch": 0.58, "learning_rate": 7.898607233881945e-06, "loss": 1.9669, "step": 23546 }, { "epoch": 0.58, "learning_rate": 7.897826735019759e-06, "loss": 2.1067, "step": 23547 }, { "epoch": 0.58, "learning_rate": 7.897046249555412e-06, "loss": 1.888, "step": 23548 }, { "epoch": 0.58, "learning_rate": 7.89626577749387e-06, "loss": 1.9046, "step": 23549 }, { "epoch": 0.58, "learning_rate": 7.895485318840111e-06, "loss": 1.9715, "step": 23550 }, { "epoch": 0.58, "learning_rate": 7.894704873599114e-06, "loss": 1.9635, "step": 23551 }, { "epoch": 0.58, "learning_rate": 7.893924441775847e-06, "loss": 2.1228, "step": 23552 }, { "epoch": 0.58, "learning_rate": 7.893144023375288e-06, "loss": 2.244, "step": 23553 }, { "epoch": 0.58, "learning_rate": 7.892363618402405e-06, "loss": 2.0489, "step": 23554 }, { "epoch": 0.58, "learning_rate": 7.891583226862175e-06, "loss": 1.9668, "step": 23555 }, { "epoch": 0.58, "learning_rate": 7.890802848759578e-06, "loss": 2.0441, "step": 23556 }, { "epoch": 0.58, "learning_rate": 7.890022484099576e-06, "loss": 2.0997, "step": 23557 }, { "epoch": 0.58, "learning_rate": 7.889242132887154e-06, "loss": 1.935, "step": 23558 }, { "epoch": 0.58, "learning_rate": 7.888461795127274e-06, "loss": 2.0665, "step": 23559 }, { "epoch": 0.58, "learning_rate": 7.887681470824916e-06, "loss": 2.0135, "step": 23560 }, { "epoch": 0.58, "learning_rate": 7.886901159985056e-06, "loss": 2.0055, "step": 23561 }, { "epoch": 0.58, "learning_rate": 7.886120862612658e-06, "loss": 1.8838, "step": 23562 }, { "epoch": 0.58, "learning_rate": 7.885340578712708e-06, "loss": 1.9472, "step": 23563 }, { "epoch": 0.58, "learning_rate": 7.884560308290165e-06, "loss": 2.1199, "step": 23564 }, { "epoch": 0.58, "learning_rate": 7.883780051350011e-06, "loss": 1.942, "step": 23565 }, { "epoch": 0.58, "learning_rate": 7.88299980789722e-06, "loss": 1.9772, "step": 23566 }, { "epoch": 0.58, "learning_rate": 7.882219577936756e-06, "loss": 2.0006, "step": 23567 }, { "epoch": 0.58, "learning_rate": 7.881439361473598e-06, "loss": 1.8271, "step": 23568 }, { "epoch": 0.58, "learning_rate": 7.880659158512718e-06, "loss": 1.9561, "step": 23569 }, { "epoch": 0.58, "learning_rate": 7.879878969059087e-06, "loss": 2.0338, "step": 23570 }, { "epoch": 0.58, "learning_rate": 7.87909879311768e-06, "loss": 1.9676, "step": 23571 }, { "epoch": 0.58, "learning_rate": 7.878318630693466e-06, "loss": 1.8769, "step": 23572 }, { "epoch": 0.58, "learning_rate": 7.877538481791423e-06, "loss": 1.9963, "step": 23573 }, { "epoch": 0.58, "learning_rate": 7.876758346416513e-06, "loss": 1.9156, "step": 23574 }, { "epoch": 0.58, "learning_rate": 7.875978224573715e-06, "loss": 2.1889, "step": 23575 }, { "epoch": 0.58, "learning_rate": 7.875198116268003e-06, "loss": 2.0941, "step": 23576 }, { "epoch": 0.58, "learning_rate": 7.874418021504343e-06, "loss": 2.0023, "step": 23577 }, { "epoch": 0.58, "learning_rate": 7.873637940287712e-06, "loss": 2.0371, "step": 23578 }, { "epoch": 0.58, "learning_rate": 7.872857872623076e-06, "loss": 1.8904, "step": 23579 }, { "epoch": 0.58, "learning_rate": 7.87207781851541e-06, "loss": 1.9597, "step": 23580 }, { "epoch": 0.58, "learning_rate": 7.871297777969691e-06, "loss": 1.9392, "step": 23581 }, { "epoch": 0.58, "learning_rate": 7.870517750990882e-06, "loss": 1.9563, "step": 23582 }, { "epoch": 0.58, "learning_rate": 7.869737737583959e-06, "loss": 2.1235, "step": 23583 }, { "epoch": 0.58, "learning_rate": 7.868957737753889e-06, "loss": 2.0785, "step": 23584 }, { "epoch": 0.58, "learning_rate": 7.868177751505645e-06, "loss": 1.9502, "step": 23585 }, { "epoch": 0.58, "learning_rate": 7.8673977788442e-06, "loss": 2.1344, "step": 23586 }, { "epoch": 0.58, "learning_rate": 7.866617819774524e-06, "loss": 2.0356, "step": 23587 }, { "epoch": 0.58, "learning_rate": 7.86583787430159e-06, "loss": 2.0357, "step": 23588 }, { "epoch": 0.58, "learning_rate": 7.865057942430364e-06, "loss": 1.8098, "step": 23589 }, { "epoch": 0.58, "learning_rate": 7.86427802416582e-06, "loss": 1.9309, "step": 23590 }, { "epoch": 0.58, "learning_rate": 7.863498119512932e-06, "loss": 1.9912, "step": 23591 }, { "epoch": 0.58, "learning_rate": 7.86271822847666e-06, "loss": 1.9655, "step": 23592 }, { "epoch": 0.58, "learning_rate": 7.861938351061984e-06, "loss": 2.0087, "step": 23593 }, { "epoch": 0.58, "learning_rate": 7.861158487273875e-06, "loss": 1.9653, "step": 23594 }, { "epoch": 0.58, "learning_rate": 7.860378637117297e-06, "loss": 1.9622, "step": 23595 }, { "epoch": 0.58, "learning_rate": 7.859598800597223e-06, "loss": 2.0227, "step": 23596 }, { "epoch": 0.58, "learning_rate": 7.858818977718622e-06, "loss": 2.0423, "step": 23597 }, { "epoch": 0.58, "learning_rate": 7.858039168486466e-06, "loss": 2.0581, "step": 23598 }, { "epoch": 0.58, "learning_rate": 7.85725937290573e-06, "loss": 2.0279, "step": 23599 }, { "epoch": 0.58, "learning_rate": 7.856479590981371e-06, "loss": 1.8901, "step": 23600 }, { "epoch": 0.58, "learning_rate": 7.855699822718374e-06, "loss": 2.0062, "step": 23601 }, { "epoch": 0.58, "learning_rate": 7.854920068121694e-06, "loss": 2.0144, "step": 23602 }, { "epoch": 0.58, "learning_rate": 7.854140327196308e-06, "loss": 1.9346, "step": 23603 }, { "epoch": 0.58, "learning_rate": 7.853360599947186e-06, "loss": 1.9046, "step": 23604 }, { "epoch": 0.58, "learning_rate": 7.852580886379297e-06, "loss": 2.0751, "step": 23605 }, { "epoch": 0.58, "learning_rate": 7.851801186497611e-06, "loss": 2.0038, "step": 23606 }, { "epoch": 0.58, "learning_rate": 7.851021500307094e-06, "loss": 1.9403, "step": 23607 }, { "epoch": 0.58, "learning_rate": 7.850241827812716e-06, "loss": 1.9456, "step": 23608 }, { "epoch": 0.58, "learning_rate": 7.849462169019454e-06, "loss": 1.9491, "step": 23609 }, { "epoch": 0.58, "learning_rate": 7.848682523932266e-06, "loss": 1.8485, "step": 23610 }, { "epoch": 0.58, "learning_rate": 7.847902892556127e-06, "loss": 2.0417, "step": 23611 }, { "epoch": 0.58, "learning_rate": 7.847123274896001e-06, "loss": 1.8851, "step": 23612 }, { "epoch": 0.58, "learning_rate": 7.84634367095686e-06, "loss": 2.1331, "step": 23613 }, { "epoch": 0.58, "learning_rate": 7.845564080743675e-06, "loss": 1.9069, "step": 23614 }, { "epoch": 0.58, "learning_rate": 7.844784504261412e-06, "loss": 1.9281, "step": 23615 }, { "epoch": 0.58, "learning_rate": 7.844004941515042e-06, "loss": 2.0181, "step": 23616 }, { "epoch": 0.58, "learning_rate": 7.843225392509526e-06, "loss": 1.7118, "step": 23617 }, { "epoch": 0.58, "learning_rate": 7.842445857249838e-06, "loss": 1.9488, "step": 23618 }, { "epoch": 0.58, "learning_rate": 7.84166633574095e-06, "loss": 1.9255, "step": 23619 }, { "epoch": 0.58, "learning_rate": 7.840886827987823e-06, "loss": 1.9649, "step": 23620 }, { "epoch": 0.58, "learning_rate": 7.840107333995429e-06, "loss": 1.9509, "step": 23621 }, { "epoch": 0.58, "learning_rate": 7.839327853768733e-06, "loss": 2.0946, "step": 23622 }, { "epoch": 0.58, "learning_rate": 7.838548387312704e-06, "loss": 2.0002, "step": 23623 }, { "epoch": 0.58, "learning_rate": 7.837768934632314e-06, "loss": 2.0462, "step": 23624 }, { "epoch": 0.58, "learning_rate": 7.836989495732524e-06, "loss": 2.0715, "step": 23625 }, { "epoch": 0.58, "learning_rate": 7.83621007061831e-06, "loss": 2.1151, "step": 23626 }, { "epoch": 0.58, "learning_rate": 7.83543065929463e-06, "loss": 1.9763, "step": 23627 }, { "epoch": 0.58, "learning_rate": 7.834651261766455e-06, "loss": 2.0505, "step": 23628 }, { "epoch": 0.58, "learning_rate": 7.833871878038756e-06, "loss": 2.0238, "step": 23629 }, { "epoch": 0.58, "learning_rate": 7.833092508116496e-06, "loss": 1.9965, "step": 23630 }, { "epoch": 0.58, "learning_rate": 7.832313152004645e-06, "loss": 2.1012, "step": 23631 }, { "epoch": 0.58, "learning_rate": 7.831533809708166e-06, "loss": 1.7646, "step": 23632 }, { "epoch": 0.58, "learning_rate": 7.830754481232029e-06, "loss": 2.0785, "step": 23633 }, { "epoch": 0.58, "learning_rate": 7.829975166581204e-06, "loss": 1.9465, "step": 23634 }, { "epoch": 0.58, "learning_rate": 7.82919586576065e-06, "loss": 2.0361, "step": 23635 }, { "epoch": 0.58, "learning_rate": 7.828416578775344e-06, "loss": 1.9982, "step": 23636 }, { "epoch": 0.58, "learning_rate": 7.827637305630242e-06, "loss": 1.9045, "step": 23637 }, { "epoch": 0.58, "learning_rate": 7.826858046330315e-06, "loss": 2.0773, "step": 23638 }, { "epoch": 0.58, "learning_rate": 7.826078800880532e-06, "loss": 1.9997, "step": 23639 }, { "epoch": 0.58, "learning_rate": 7.825299569285857e-06, "loss": 1.891, "step": 23640 }, { "epoch": 0.58, "learning_rate": 7.82452035155126e-06, "loss": 1.9098, "step": 23641 }, { "epoch": 0.58, "learning_rate": 7.823741147681697e-06, "loss": 1.9179, "step": 23642 }, { "epoch": 0.58, "learning_rate": 7.822961957682143e-06, "loss": 2.0729, "step": 23643 }, { "epoch": 0.58, "learning_rate": 7.822182781557565e-06, "loss": 1.595, "step": 23644 }, { "epoch": 0.58, "learning_rate": 7.821403619312924e-06, "loss": 1.7682, "step": 23645 }, { "epoch": 0.58, "learning_rate": 7.820624470953188e-06, "loss": 2.0005, "step": 23646 }, { "epoch": 0.58, "learning_rate": 7.819845336483322e-06, "loss": 2.0829, "step": 23647 }, { "epoch": 0.58, "learning_rate": 7.819066215908291e-06, "loss": 2.0252, "step": 23648 }, { "epoch": 0.58, "learning_rate": 7.818287109233064e-06, "loss": 2.1701, "step": 23649 }, { "epoch": 0.58, "learning_rate": 7.817508016462602e-06, "loss": 1.9895, "step": 23650 }, { "epoch": 0.58, "learning_rate": 7.816728937601871e-06, "loss": 2.0145, "step": 23651 }, { "epoch": 0.58, "learning_rate": 7.815949872655846e-06, "loss": 1.8968, "step": 23652 }, { "epoch": 0.58, "learning_rate": 7.815170821629476e-06, "loss": 2.0139, "step": 23653 }, { "epoch": 0.58, "learning_rate": 7.814391784527741e-06, "loss": 1.9298, "step": 23654 }, { "epoch": 0.58, "learning_rate": 7.813612761355594e-06, "loss": 1.9391, "step": 23655 }, { "epoch": 0.58, "learning_rate": 7.812833752118006e-06, "loss": 1.9687, "step": 23656 }, { "epoch": 0.58, "learning_rate": 7.812054756819944e-06, "loss": 2.0637, "step": 23657 }, { "epoch": 0.58, "learning_rate": 7.811275775466367e-06, "loss": 2.033, "step": 23658 }, { "epoch": 0.58, "learning_rate": 7.810496808062247e-06, "loss": 2.0669, "step": 23659 }, { "epoch": 0.58, "learning_rate": 7.809717854612541e-06, "loss": 2.1315, "step": 23660 }, { "epoch": 0.58, "learning_rate": 7.808938915122216e-06, "loss": 2.0665, "step": 23661 }, { "epoch": 0.58, "learning_rate": 7.808159989596243e-06, "loss": 1.8004, "step": 23662 }, { "epoch": 0.58, "learning_rate": 7.807381078039574e-06, "loss": 1.9099, "step": 23663 }, { "epoch": 0.58, "learning_rate": 7.806602180457186e-06, "loss": 2.1561, "step": 23664 }, { "epoch": 0.58, "learning_rate": 7.805823296854033e-06, "loss": 1.9539, "step": 23665 }, { "epoch": 0.58, "learning_rate": 7.805044427235081e-06, "loss": 2.11, "step": 23666 }, { "epoch": 0.58, "learning_rate": 7.8042655716053e-06, "loss": 1.8534, "step": 23667 }, { "epoch": 0.58, "learning_rate": 7.803486729969649e-06, "loss": 1.9638, "step": 23668 }, { "epoch": 0.58, "learning_rate": 7.802707902333096e-06, "loss": 1.8208, "step": 23669 }, { "epoch": 0.58, "learning_rate": 7.801929088700597e-06, "loss": 2.1014, "step": 23670 }, { "epoch": 0.58, "learning_rate": 7.801150289077121e-06, "loss": 2.0183, "step": 23671 }, { "epoch": 0.58, "learning_rate": 7.800371503467634e-06, "loss": 2.1071, "step": 23672 }, { "epoch": 0.58, "learning_rate": 7.799592731877092e-06, "loss": 1.9805, "step": 23673 }, { "epoch": 0.58, "learning_rate": 7.798813974310467e-06, "loss": 1.8772, "step": 23674 }, { "epoch": 0.58, "learning_rate": 7.798035230772713e-06, "loss": 1.8457, "step": 23675 }, { "epoch": 0.58, "learning_rate": 7.7972565012688e-06, "loss": 2.0514, "step": 23676 }, { "epoch": 0.58, "learning_rate": 7.796477785803693e-06, "loss": 1.9507, "step": 23677 }, { "epoch": 0.58, "learning_rate": 7.795699084382347e-06, "loss": 2.0447, "step": 23678 }, { "epoch": 0.58, "learning_rate": 7.794920397009731e-06, "loss": 1.9917, "step": 23679 }, { "epoch": 0.58, "learning_rate": 7.794141723690804e-06, "loss": 2.0051, "step": 23680 }, { "epoch": 0.58, "learning_rate": 7.79336306443053e-06, "loss": 2.0559, "step": 23681 }, { "epoch": 0.58, "learning_rate": 7.792584419233877e-06, "loss": 2.0007, "step": 23682 }, { "epoch": 0.58, "learning_rate": 7.7918057881058e-06, "loss": 2.0864, "step": 23683 }, { "epoch": 0.58, "learning_rate": 7.791027171051264e-06, "loss": 2.1154, "step": 23684 }, { "epoch": 0.58, "learning_rate": 7.79024856807523e-06, "loss": 2.0372, "step": 23685 }, { "epoch": 0.58, "learning_rate": 7.789469979182663e-06, "loss": 2.0877, "step": 23686 }, { "epoch": 0.58, "learning_rate": 7.788691404378527e-06, "loss": 2.1234, "step": 23687 }, { "epoch": 0.58, "learning_rate": 7.78791284366778e-06, "loss": 2.0236, "step": 23688 }, { "epoch": 0.58, "learning_rate": 7.787134297055386e-06, "loss": 1.918, "step": 23689 }, { "epoch": 0.58, "learning_rate": 7.786355764546305e-06, "loss": 1.9301, "step": 23690 }, { "epoch": 0.58, "learning_rate": 7.785577246145499e-06, "loss": 1.8636, "step": 23691 }, { "epoch": 0.58, "learning_rate": 7.784798741857932e-06, "loss": 2.0093, "step": 23692 }, { "epoch": 0.58, "learning_rate": 7.784020251688563e-06, "loss": 2.0912, "step": 23693 }, { "epoch": 0.58, "learning_rate": 7.78324177564236e-06, "loss": 1.9126, "step": 23694 }, { "epoch": 0.58, "learning_rate": 7.782463313724274e-06, "loss": 2.2388, "step": 23695 }, { "epoch": 0.58, "learning_rate": 7.781684865939275e-06, "loss": 1.7297, "step": 23696 }, { "epoch": 0.58, "learning_rate": 7.780906432292323e-06, "loss": 1.992, "step": 23697 }, { "epoch": 0.58, "learning_rate": 7.780128012788373e-06, "loss": 2.0111, "step": 23698 }, { "epoch": 0.58, "learning_rate": 7.779349607432393e-06, "loss": 1.8823, "step": 23699 }, { "epoch": 0.58, "learning_rate": 7.778571216229342e-06, "loss": 1.9533, "step": 23700 }, { "epoch": 0.58, "learning_rate": 7.777792839184179e-06, "loss": 2.0699, "step": 23701 }, { "epoch": 0.58, "learning_rate": 7.777014476301868e-06, "loss": 1.9453, "step": 23702 }, { "epoch": 0.58, "learning_rate": 7.776236127587365e-06, "loss": 1.9345, "step": 23703 }, { "epoch": 0.58, "learning_rate": 7.77545779304564e-06, "loss": 2.1288, "step": 23704 }, { "epoch": 0.58, "learning_rate": 7.774679472681642e-06, "loss": 1.9549, "step": 23705 }, { "epoch": 0.58, "learning_rate": 7.773901166500338e-06, "loss": 2.1368, "step": 23706 }, { "epoch": 0.58, "learning_rate": 7.77312287450669e-06, "loss": 2.1076, "step": 23707 }, { "epoch": 0.58, "learning_rate": 7.772344596705652e-06, "loss": 1.9965, "step": 23708 }, { "epoch": 0.58, "learning_rate": 7.77156633310219e-06, "loss": 1.8967, "step": 23709 }, { "epoch": 0.58, "learning_rate": 7.770788083701262e-06, "loss": 2.0503, "step": 23710 }, { "epoch": 0.58, "learning_rate": 7.770009848507824e-06, "loss": 1.7365, "step": 23711 }, { "epoch": 0.58, "learning_rate": 7.769231627526847e-06, "loss": 1.9094, "step": 23712 }, { "epoch": 0.58, "learning_rate": 7.76845342076328e-06, "loss": 1.9205, "step": 23713 }, { "epoch": 0.58, "learning_rate": 7.767675228222084e-06, "loss": 1.9837, "step": 23714 }, { "epoch": 0.58, "learning_rate": 7.766897049908228e-06, "loss": 1.9967, "step": 23715 }, { "epoch": 0.58, "learning_rate": 7.76611888582666e-06, "loss": 2.1217, "step": 23716 }, { "epoch": 0.58, "learning_rate": 7.765340735982345e-06, "loss": 2.0349, "step": 23717 }, { "epoch": 0.58, "learning_rate": 7.764562600380242e-06, "loss": 2.01, "step": 23718 }, { "epoch": 0.58, "learning_rate": 7.76378447902531e-06, "loss": 1.895, "step": 23719 }, { "epoch": 0.58, "learning_rate": 7.76300637192251e-06, "loss": 1.9458, "step": 23720 }, { "epoch": 0.58, "learning_rate": 7.762228279076795e-06, "loss": 2.0114, "step": 23721 }, { "epoch": 0.58, "learning_rate": 7.761450200493134e-06, "loss": 1.7946, "step": 23722 }, { "epoch": 0.58, "learning_rate": 7.760672136176477e-06, "loss": 2.009, "step": 23723 }, { "epoch": 0.58, "learning_rate": 7.759894086131786e-06, "loss": 1.9712, "step": 23724 }, { "epoch": 0.58, "learning_rate": 7.759116050364024e-06, "loss": 2.0537, "step": 23725 }, { "epoch": 0.58, "learning_rate": 7.75833802887814e-06, "loss": 1.9202, "step": 23726 }, { "epoch": 0.58, "learning_rate": 7.757560021679103e-06, "loss": 2.0343, "step": 23727 }, { "epoch": 0.58, "learning_rate": 7.756782028771863e-06, "loss": 2.1129, "step": 23728 }, { "epoch": 0.58, "learning_rate": 7.756004050161384e-06, "loss": 1.8392, "step": 23729 }, { "epoch": 0.58, "learning_rate": 7.755226085852625e-06, "loss": 2.1286, "step": 23730 }, { "epoch": 0.58, "learning_rate": 7.754448135850537e-06, "loss": 1.8551, "step": 23731 }, { "epoch": 0.58, "learning_rate": 7.753670200160089e-06, "loss": 1.8908, "step": 23732 }, { "epoch": 0.58, "learning_rate": 7.752892278786225e-06, "loss": 1.9368, "step": 23733 }, { "epoch": 0.59, "learning_rate": 7.752114371733914e-06, "loss": 1.9764, "step": 23734 }, { "epoch": 0.59, "learning_rate": 7.751336479008113e-06, "loss": 1.9044, "step": 23735 }, { "epoch": 0.59, "learning_rate": 7.750558600613775e-06, "loss": 2.1544, "step": 23736 }, { "epoch": 0.59, "learning_rate": 7.749780736555861e-06, "loss": 2.0675, "step": 23737 }, { "epoch": 0.59, "learning_rate": 7.749002886839327e-06, "loss": 2.0142, "step": 23738 }, { "epoch": 0.59, "learning_rate": 7.74822505146913e-06, "loss": 2.0803, "step": 23739 }, { "epoch": 0.59, "learning_rate": 7.747447230450233e-06, "loss": 1.7448, "step": 23740 }, { "epoch": 0.59, "learning_rate": 7.746669423787585e-06, "loss": 1.8596, "step": 23741 }, { "epoch": 0.59, "learning_rate": 7.745891631486153e-06, "loss": 1.9873, "step": 23742 }, { "epoch": 0.59, "learning_rate": 7.745113853550881e-06, "loss": 2.0299, "step": 23743 }, { "epoch": 0.59, "learning_rate": 7.744336089986735e-06, "loss": 2.0001, "step": 23744 }, { "epoch": 0.59, "learning_rate": 7.743558340798672e-06, "loss": 2.0169, "step": 23745 }, { "epoch": 0.59, "learning_rate": 7.742780605991643e-06, "loss": 2.0817, "step": 23746 }, { "epoch": 0.59, "learning_rate": 7.742002885570616e-06, "loss": 1.8941, "step": 23747 }, { "epoch": 0.59, "learning_rate": 7.741225179540535e-06, "loss": 2.0785, "step": 23748 }, { "epoch": 0.59, "learning_rate": 7.740447487906361e-06, "loss": 2.0566, "step": 23749 }, { "epoch": 0.59, "learning_rate": 7.739669810673057e-06, "loss": 2.0485, "step": 23750 }, { "epoch": 0.59, "learning_rate": 7.738892147845569e-06, "loss": 2.0041, "step": 23751 }, { "epoch": 0.59, "learning_rate": 7.738114499428862e-06, "loss": 2.003, "step": 23752 }, { "epoch": 0.59, "learning_rate": 7.737336865427884e-06, "loss": 1.9573, "step": 23753 }, { "epoch": 0.59, "learning_rate": 7.736559245847599e-06, "loss": 2.0046, "step": 23754 }, { "epoch": 0.59, "learning_rate": 7.73578164069296e-06, "loss": 2.0941, "step": 23755 }, { "epoch": 0.59, "learning_rate": 7.735004049968919e-06, "loss": 1.964, "step": 23756 }, { "epoch": 0.59, "learning_rate": 7.734226473680439e-06, "loss": 2.0705, "step": 23757 }, { "epoch": 0.59, "learning_rate": 7.73344891183247e-06, "loss": 2.0555, "step": 23758 }, { "epoch": 0.59, "learning_rate": 7.73267136442997e-06, "loss": 1.912, "step": 23759 }, { "epoch": 0.59, "learning_rate": 7.731893831477896e-06, "loss": 2.0345, "step": 23760 }, { "epoch": 0.59, "learning_rate": 7.731116312981198e-06, "loss": 2.1228, "step": 23761 }, { "epoch": 0.59, "learning_rate": 7.730338808944839e-06, "loss": 2.0606, "step": 23762 }, { "epoch": 0.59, "learning_rate": 7.729561319373766e-06, "loss": 1.8387, "step": 23763 }, { "epoch": 0.59, "learning_rate": 7.728783844272941e-06, "loss": 2.184, "step": 23764 }, { "epoch": 0.59, "learning_rate": 7.72800638364732e-06, "loss": 2.0624, "step": 23765 }, { "epoch": 0.59, "learning_rate": 7.72722893750185e-06, "loss": 2.0888, "step": 23766 }, { "epoch": 0.59, "learning_rate": 7.726451505841495e-06, "loss": 2.0141, "step": 23767 }, { "epoch": 0.59, "learning_rate": 7.725674088671202e-06, "loss": 1.9531, "step": 23768 }, { "epoch": 0.59, "learning_rate": 7.724896685995928e-06, "loss": 1.9909, "step": 23769 }, { "epoch": 0.59, "learning_rate": 7.724119297820633e-06, "loss": 1.9133, "step": 23770 }, { "epoch": 0.59, "learning_rate": 7.723341924150266e-06, "loss": 1.9654, "step": 23771 }, { "epoch": 0.59, "learning_rate": 7.722564564989781e-06, "loss": 2.0023, "step": 23772 }, { "epoch": 0.59, "learning_rate": 7.721787220344136e-06, "loss": 1.8659, "step": 23773 }, { "epoch": 0.59, "learning_rate": 7.721009890218282e-06, "loss": 2.1124, "step": 23774 }, { "epoch": 0.59, "learning_rate": 7.720232574617179e-06, "loss": 1.9461, "step": 23775 }, { "epoch": 0.59, "learning_rate": 7.719455273545772e-06, "loss": 1.9504, "step": 23776 }, { "epoch": 0.59, "learning_rate": 7.71867798700902e-06, "loss": 2.0273, "step": 23777 }, { "epoch": 0.59, "learning_rate": 7.717900715011881e-06, "loss": 1.9131, "step": 23778 }, { "epoch": 0.59, "learning_rate": 7.7171234575593e-06, "loss": 2.0653, "step": 23779 }, { "epoch": 0.59, "learning_rate": 7.716346214656237e-06, "loss": 1.9318, "step": 23780 }, { "epoch": 0.59, "learning_rate": 7.715568986307641e-06, "loss": 2.1141, "step": 23781 }, { "epoch": 0.59, "learning_rate": 7.71479177251847e-06, "loss": 2.0785, "step": 23782 }, { "epoch": 0.59, "learning_rate": 7.714014573293679e-06, "loss": 2.0101, "step": 23783 }, { "epoch": 0.59, "learning_rate": 7.713237388638214e-06, "loss": 2.2853, "step": 23784 }, { "epoch": 0.59, "learning_rate": 7.712460218557037e-06, "loss": 2.2188, "step": 23785 }, { "epoch": 0.59, "learning_rate": 7.71168306305509e-06, "loss": 1.8456, "step": 23786 }, { "epoch": 0.59, "learning_rate": 7.710905922137336e-06, "loss": 1.9452, "step": 23787 }, { "epoch": 0.59, "learning_rate": 7.710128795808726e-06, "loss": 1.9082, "step": 23788 }, { "epoch": 0.59, "learning_rate": 7.70935168407421e-06, "loss": 1.8678, "step": 23789 }, { "epoch": 0.59, "learning_rate": 7.708574586938743e-06, "loss": 2.0561, "step": 23790 }, { "epoch": 0.59, "learning_rate": 7.707797504407276e-06, "loss": 2.2656, "step": 23791 }, { "epoch": 0.59, "learning_rate": 7.70702043648476e-06, "loss": 1.9099, "step": 23792 }, { "epoch": 0.59, "learning_rate": 7.706243383176155e-06, "loss": 1.8784, "step": 23793 }, { "epoch": 0.59, "learning_rate": 7.705466344486405e-06, "loss": 2.0424, "step": 23794 }, { "epoch": 0.59, "learning_rate": 7.70468932042047e-06, "loss": 1.9708, "step": 23795 }, { "epoch": 0.59, "learning_rate": 7.703912310983292e-06, "loss": 2.1097, "step": 23796 }, { "epoch": 0.59, "learning_rate": 7.703135316179832e-06, "loss": 2.1605, "step": 23797 }, { "epoch": 0.59, "learning_rate": 7.702358336015039e-06, "loss": 1.8679, "step": 23798 }, { "epoch": 0.59, "learning_rate": 7.701581370493863e-06, "loss": 1.8018, "step": 23799 }, { "epoch": 0.59, "learning_rate": 7.700804419621264e-06, "loss": 1.8454, "step": 23800 }, { "epoch": 0.59, "learning_rate": 7.700027483402182e-06, "loss": 1.9412, "step": 23801 }, { "epoch": 0.59, "learning_rate": 7.699250561841574e-06, "loss": 1.8844, "step": 23802 }, { "epoch": 0.59, "learning_rate": 7.698473654944398e-06, "loss": 1.9132, "step": 23803 }, { "epoch": 0.59, "learning_rate": 7.697696762715595e-06, "loss": 2.2002, "step": 23804 }, { "epoch": 0.59, "learning_rate": 7.69691988516012e-06, "loss": 1.9113, "step": 23805 }, { "epoch": 0.59, "learning_rate": 7.696143022282926e-06, "loss": 1.8661, "step": 23806 }, { "epoch": 0.59, "learning_rate": 7.695366174088965e-06, "loss": 2.0471, "step": 23807 }, { "epoch": 0.59, "learning_rate": 7.694589340583185e-06, "loss": 1.8798, "step": 23808 }, { "epoch": 0.59, "learning_rate": 7.693812521770539e-06, "loss": 2.0226, "step": 23809 }, { "epoch": 0.59, "learning_rate": 7.69303571765598e-06, "loss": 1.9185, "step": 23810 }, { "epoch": 0.59, "learning_rate": 7.692258928244451e-06, "loss": 2.0659, "step": 23811 }, { "epoch": 0.59, "learning_rate": 7.69148215354091e-06, "loss": 1.9305, "step": 23812 }, { "epoch": 0.59, "learning_rate": 7.690705393550309e-06, "loss": 1.9237, "step": 23813 }, { "epoch": 0.59, "learning_rate": 7.68992864827759e-06, "loss": 1.9994, "step": 23814 }, { "epoch": 0.59, "learning_rate": 7.689151917727711e-06, "loss": 2.2089, "step": 23815 }, { "epoch": 0.59, "learning_rate": 7.688375201905618e-06, "loss": 2.1211, "step": 23816 }, { "epoch": 0.59, "learning_rate": 7.687598500816262e-06, "loss": 2.0401, "step": 23817 }, { "epoch": 0.59, "learning_rate": 7.6868218144646e-06, "loss": 1.8667, "step": 23818 }, { "epoch": 0.59, "learning_rate": 7.686045142855573e-06, "loss": 2.1407, "step": 23819 }, { "epoch": 0.59, "learning_rate": 7.685268485994137e-06, "loss": 1.8048, "step": 23820 }, { "epoch": 0.59, "learning_rate": 7.684491843885235e-06, "loss": 1.9344, "step": 23821 }, { "epoch": 0.59, "learning_rate": 7.683715216533822e-06, "loss": 1.9978, "step": 23822 }, { "epoch": 0.59, "learning_rate": 7.68293860394485e-06, "loss": 2.0025, "step": 23823 }, { "epoch": 0.59, "learning_rate": 7.682162006123262e-06, "loss": 1.9926, "step": 23824 }, { "epoch": 0.59, "learning_rate": 7.681385423074014e-06, "loss": 1.7713, "step": 23825 }, { "epoch": 0.59, "learning_rate": 7.68060885480205e-06, "loss": 1.9823, "step": 23826 }, { "epoch": 0.59, "learning_rate": 7.67983230131232e-06, "loss": 1.916, "step": 23827 }, { "epoch": 0.59, "learning_rate": 7.679055762609779e-06, "loss": 1.8659, "step": 23828 }, { "epoch": 0.59, "learning_rate": 7.678279238699368e-06, "loss": 1.8456, "step": 23829 }, { "epoch": 0.59, "learning_rate": 7.677502729586042e-06, "loss": 1.9039, "step": 23830 }, { "epoch": 0.59, "learning_rate": 7.676726235274751e-06, "loss": 1.9396, "step": 23831 }, { "epoch": 0.59, "learning_rate": 7.675949755770436e-06, "loss": 2.0495, "step": 23832 }, { "epoch": 0.59, "learning_rate": 7.675173291078053e-06, "loss": 1.9816, "step": 23833 }, { "epoch": 0.59, "learning_rate": 7.674396841202546e-06, "loss": 1.9871, "step": 23834 }, { "epoch": 0.59, "learning_rate": 7.673620406148866e-06, "loss": 1.8635, "step": 23835 }, { "epoch": 0.59, "learning_rate": 7.672843985921966e-06, "loss": 2.1066, "step": 23836 }, { "epoch": 0.59, "learning_rate": 7.672067580526785e-06, "loss": 2.0536, "step": 23837 }, { "epoch": 0.59, "learning_rate": 7.671291189968281e-06, "loss": 2.0189, "step": 23838 }, { "epoch": 0.59, "learning_rate": 7.670514814251392e-06, "loss": 2.0554, "step": 23839 }, { "epoch": 0.59, "learning_rate": 7.669738453381071e-06, "loss": 2.0237, "step": 23840 }, { "epoch": 0.59, "learning_rate": 7.66896210736227e-06, "loss": 2.0134, "step": 23841 }, { "epoch": 0.59, "learning_rate": 7.668185776199928e-06, "loss": 1.9876, "step": 23842 }, { "epoch": 0.59, "learning_rate": 7.667409459899005e-06, "loss": 2.0206, "step": 23843 }, { "epoch": 0.59, "learning_rate": 7.666633158464436e-06, "loss": 1.921, "step": 23844 }, { "epoch": 0.59, "learning_rate": 7.665856871901176e-06, "loss": 2.0301, "step": 23845 }, { "epoch": 0.59, "learning_rate": 7.665080600214173e-06, "loss": 1.98, "step": 23846 }, { "epoch": 0.59, "learning_rate": 7.664304343408368e-06, "loss": 1.9814, "step": 23847 }, { "epoch": 0.59, "learning_rate": 7.663528101488718e-06, "loss": 1.9866, "step": 23848 }, { "epoch": 0.59, "learning_rate": 7.66275187446016e-06, "loss": 2.0083, "step": 23849 }, { "epoch": 0.59, "learning_rate": 7.661975662327647e-06, "loss": 1.9174, "step": 23850 }, { "epoch": 0.59, "learning_rate": 7.661199465096126e-06, "loss": 1.962, "step": 23851 }, { "epoch": 0.59, "learning_rate": 7.660423282770542e-06, "loss": 1.8885, "step": 23852 }, { "epoch": 0.59, "learning_rate": 7.659647115355846e-06, "loss": 2.0564, "step": 23853 }, { "epoch": 0.59, "learning_rate": 7.658870962856978e-06, "loss": 1.9985, "step": 23854 }, { "epoch": 0.59, "learning_rate": 7.658094825278887e-06, "loss": 1.9038, "step": 23855 }, { "epoch": 0.59, "learning_rate": 7.657318702626528e-06, "loss": 2.0703, "step": 23856 }, { "epoch": 0.59, "learning_rate": 7.656542594904834e-06, "loss": 2.0096, "step": 23857 }, { "epoch": 0.59, "learning_rate": 7.65576650211876e-06, "loss": 1.9987, "step": 23858 }, { "epoch": 0.59, "learning_rate": 7.654990424273248e-06, "loss": 2.0834, "step": 23859 }, { "epoch": 0.59, "learning_rate": 7.65421436137325e-06, "loss": 2.0723, "step": 23860 }, { "epoch": 0.59, "learning_rate": 7.653438313423706e-06, "loss": 2.1289, "step": 23861 }, { "epoch": 0.59, "learning_rate": 7.652662280429563e-06, "loss": 1.9353, "step": 23862 }, { "epoch": 0.59, "learning_rate": 7.651886262395773e-06, "loss": 2.0373, "step": 23863 }, { "epoch": 0.59, "learning_rate": 7.651110259327273e-06, "loss": 1.8973, "step": 23864 }, { "epoch": 0.59, "learning_rate": 7.650334271229011e-06, "loss": 2.0221, "step": 23865 }, { "epoch": 0.59, "learning_rate": 7.64955829810594e-06, "loss": 2.0996, "step": 23866 }, { "epoch": 0.59, "learning_rate": 7.648782339962995e-06, "loss": 1.921, "step": 23867 }, { "epoch": 0.59, "learning_rate": 7.64800639680513e-06, "loss": 2.1183, "step": 23868 }, { "epoch": 0.59, "learning_rate": 7.647230468637284e-06, "loss": 2.106, "step": 23869 }, { "epoch": 0.59, "learning_rate": 7.646454555464404e-06, "loss": 2.0322, "step": 23870 }, { "epoch": 0.59, "learning_rate": 7.645678657291441e-06, "loss": 1.9972, "step": 23871 }, { "epoch": 0.59, "learning_rate": 7.644902774123332e-06, "loss": 2.0464, "step": 23872 }, { "epoch": 0.59, "learning_rate": 7.644126905965029e-06, "loss": 1.9555, "step": 23873 }, { "epoch": 0.59, "learning_rate": 7.643351052821467e-06, "loss": 2.0259, "step": 23874 }, { "epoch": 0.59, "learning_rate": 7.642575214697599e-06, "loss": 2.0936, "step": 23875 }, { "epoch": 0.59, "learning_rate": 7.641799391598367e-06, "loss": 1.9614, "step": 23876 }, { "epoch": 0.59, "learning_rate": 7.641023583528715e-06, "loss": 2.0664, "step": 23877 }, { "epoch": 0.59, "learning_rate": 7.640247790493593e-06, "loss": 1.8623, "step": 23878 }, { "epoch": 0.59, "learning_rate": 7.639472012497938e-06, "loss": 2.0511, "step": 23879 }, { "epoch": 0.59, "learning_rate": 7.638696249546695e-06, "loss": 1.9203, "step": 23880 }, { "epoch": 0.59, "learning_rate": 7.637920501644816e-06, "loss": 2.0535, "step": 23881 }, { "epoch": 0.59, "learning_rate": 7.637144768797234e-06, "loss": 1.972, "step": 23882 }, { "epoch": 0.59, "learning_rate": 7.636369051008902e-06, "loss": 2.0944, "step": 23883 }, { "epoch": 0.59, "learning_rate": 7.635593348284758e-06, "loss": 1.9707, "step": 23884 }, { "epoch": 0.59, "learning_rate": 7.634817660629747e-06, "loss": 1.956, "step": 23885 }, { "epoch": 0.59, "learning_rate": 7.634041988048817e-06, "loss": 1.9593, "step": 23886 }, { "epoch": 0.59, "learning_rate": 7.633266330546905e-06, "loss": 1.922, "step": 23887 }, { "epoch": 0.59, "learning_rate": 7.63249068812896e-06, "loss": 2.0121, "step": 23888 }, { "epoch": 0.59, "learning_rate": 7.631715060799926e-06, "loss": 2.0925, "step": 23889 }, { "epoch": 0.59, "learning_rate": 7.630939448564739e-06, "loss": 2.0027, "step": 23890 }, { "epoch": 0.59, "learning_rate": 7.630163851428353e-06, "loss": 2.0052, "step": 23891 }, { "epoch": 0.59, "learning_rate": 7.629388269395699e-06, "loss": 1.973, "step": 23892 }, { "epoch": 0.59, "learning_rate": 7.628612702471727e-06, "loss": 1.8854, "step": 23893 }, { "epoch": 0.59, "learning_rate": 7.6278371506613815e-06, "loss": 2.0832, "step": 23894 }, { "epoch": 0.59, "learning_rate": 7.627061613969599e-06, "loss": 2.2898, "step": 23895 }, { "epoch": 0.59, "learning_rate": 7.626286092401333e-06, "loss": 1.9788, "step": 23896 }, { "epoch": 0.59, "learning_rate": 7.625510585961513e-06, "loss": 2.0662, "step": 23897 }, { "epoch": 0.59, "learning_rate": 7.624735094655089e-06, "loss": 2.0841, "step": 23898 }, { "epoch": 0.59, "learning_rate": 7.623959618487007e-06, "loss": 2.0594, "step": 23899 }, { "epoch": 0.59, "learning_rate": 7.623184157462199e-06, "loss": 2.1041, "step": 23900 }, { "epoch": 0.59, "learning_rate": 7.622408711585618e-06, "loss": 1.9848, "step": 23901 }, { "epoch": 0.59, "learning_rate": 7.621633280862197e-06, "loss": 2.0693, "step": 23902 }, { "epoch": 0.59, "learning_rate": 7.620857865296883e-06, "loss": 1.8551, "step": 23903 }, { "epoch": 0.59, "learning_rate": 7.620082464894617e-06, "loss": 2.0016, "step": 23904 }, { "epoch": 0.59, "learning_rate": 7.619307079660341e-06, "loss": 2.0312, "step": 23905 }, { "epoch": 0.59, "learning_rate": 7.618531709598999e-06, "loss": 2.1165, "step": 23906 }, { "epoch": 0.59, "learning_rate": 7.617756354715528e-06, "loss": 2.0393, "step": 23907 }, { "epoch": 0.59, "learning_rate": 7.616981015014872e-06, "loss": 2.1417, "step": 23908 }, { "epoch": 0.59, "learning_rate": 7.616205690501976e-06, "loss": 1.8315, "step": 23909 }, { "epoch": 0.59, "learning_rate": 7.615430381181774e-06, "loss": 2.0296, "step": 23910 }, { "epoch": 0.59, "learning_rate": 7.614655087059214e-06, "loss": 1.9897, "step": 23911 }, { "epoch": 0.59, "learning_rate": 7.6138798081392316e-06, "loss": 2.1176, "step": 23912 }, { "epoch": 0.59, "learning_rate": 7.613104544426771e-06, "loss": 2.0583, "step": 23913 }, { "epoch": 0.59, "learning_rate": 7.612329295926777e-06, "loss": 1.989, "step": 23914 }, { "epoch": 0.59, "learning_rate": 7.611554062644183e-06, "loss": 1.9985, "step": 23915 }, { "epoch": 0.59, "learning_rate": 7.610778844583937e-06, "loss": 1.9069, "step": 23916 }, { "epoch": 0.59, "learning_rate": 7.610003641750973e-06, "loss": 1.9502, "step": 23917 }, { "epoch": 0.59, "learning_rate": 7.609228454150232e-06, "loss": 1.9191, "step": 23918 }, { "epoch": 0.59, "learning_rate": 7.608453281786664e-06, "loss": 1.9665, "step": 23919 }, { "epoch": 0.59, "learning_rate": 7.607678124665198e-06, "loss": 1.7935, "step": 23920 }, { "epoch": 0.59, "learning_rate": 7.606902982790781e-06, "loss": 2.0635, "step": 23921 }, { "epoch": 0.59, "learning_rate": 7.606127856168348e-06, "loss": 2.1216, "step": 23922 }, { "epoch": 0.59, "learning_rate": 7.605352744802843e-06, "loss": 2.0297, "step": 23923 }, { "epoch": 0.59, "learning_rate": 7.60457764869921e-06, "loss": 2.0569, "step": 23924 }, { "epoch": 0.59, "learning_rate": 7.603802567862379e-06, "loss": 1.9956, "step": 23925 }, { "epoch": 0.59, "learning_rate": 7.6030275022973e-06, "loss": 1.8616, "step": 23926 }, { "epoch": 0.59, "learning_rate": 7.602252452008904e-06, "loss": 1.8563, "step": 23927 }, { "epoch": 0.59, "learning_rate": 7.601477417002134e-06, "loss": 1.929, "step": 23928 }, { "epoch": 0.59, "learning_rate": 7.600702397281932e-06, "loss": 2.112, "step": 23929 }, { "epoch": 0.59, "learning_rate": 7.599927392853235e-06, "loss": 1.8986, "step": 23930 }, { "epoch": 0.59, "learning_rate": 7.599152403720985e-06, "loss": 2.0099, "step": 23931 }, { "epoch": 0.59, "learning_rate": 7.598377429890116e-06, "loss": 1.9905, "step": 23932 }, { "epoch": 0.59, "learning_rate": 7.59760247136557e-06, "loss": 2.1579, "step": 23933 }, { "epoch": 0.59, "learning_rate": 7.59682752815229e-06, "loss": 1.8553, "step": 23934 }, { "epoch": 0.59, "learning_rate": 7.5960526002552085e-06, "loss": 1.9314, "step": 23935 }, { "epoch": 0.59, "learning_rate": 7.595277687679269e-06, "loss": 1.8623, "step": 23936 }, { "epoch": 0.59, "learning_rate": 7.594502790429408e-06, "loss": 1.8442, "step": 23937 }, { "epoch": 0.59, "learning_rate": 7.593727908510562e-06, "loss": 1.9086, "step": 23938 }, { "epoch": 0.59, "learning_rate": 7.592953041927676e-06, "loss": 2.0711, "step": 23939 }, { "epoch": 0.59, "learning_rate": 7.592178190685681e-06, "loss": 2.0315, "step": 23940 }, { "epoch": 0.59, "learning_rate": 7.591403354789524e-06, "loss": 2.1557, "step": 23941 }, { "epoch": 0.59, "learning_rate": 7.590628534244133e-06, "loss": 2.0366, "step": 23942 }, { "epoch": 0.59, "learning_rate": 7.5898537290544525e-06, "loss": 1.8549, "step": 23943 }, { "epoch": 0.59, "learning_rate": 7.589078939225425e-06, "loss": 1.7814, "step": 23944 }, { "epoch": 0.59, "learning_rate": 7.588304164761977e-06, "loss": 2.0283, "step": 23945 }, { "epoch": 0.59, "learning_rate": 7.5875294056690565e-06, "loss": 2.0398, "step": 23946 }, { "epoch": 0.59, "learning_rate": 7.586754661951593e-06, "loss": 1.9023, "step": 23947 }, { "epoch": 0.59, "learning_rate": 7.58597993361453e-06, "loss": 1.8144, "step": 23948 }, { "epoch": 0.59, "learning_rate": 7.585205220662807e-06, "loss": 1.966, "step": 23949 }, { "epoch": 0.59, "learning_rate": 7.584430523101354e-06, "loss": 1.928, "step": 23950 }, { "epoch": 0.59, "learning_rate": 7.583655840935113e-06, "loss": 2.0289, "step": 23951 }, { "epoch": 0.59, "learning_rate": 7.582881174169026e-06, "loss": 2.0474, "step": 23952 }, { "epoch": 0.59, "learning_rate": 7.582106522808018e-06, "loss": 2.1246, "step": 23953 }, { "epoch": 0.59, "learning_rate": 7.581331886857038e-06, "loss": 1.9735, "step": 23954 }, { "epoch": 0.59, "learning_rate": 7.580557266321016e-06, "loss": 1.9703, "step": 23955 }, { "epoch": 0.59, "learning_rate": 7.57978266120489e-06, "loss": 2.0427, "step": 23956 }, { "epoch": 0.59, "learning_rate": 7.579008071513599e-06, "loss": 2.0099, "step": 23957 }, { "epoch": 0.59, "learning_rate": 7.578233497252077e-06, "loss": 1.9866, "step": 23958 }, { "epoch": 0.59, "learning_rate": 7.577458938425267e-06, "loss": 1.9468, "step": 23959 }, { "epoch": 0.59, "learning_rate": 7.576684395038096e-06, "loss": 2.0202, "step": 23960 }, { "epoch": 0.59, "learning_rate": 7.575909867095505e-06, "loss": 1.9481, "step": 23961 }, { "epoch": 0.59, "learning_rate": 7.575135354602434e-06, "loss": 1.9461, "step": 23962 }, { "epoch": 0.59, "learning_rate": 7.574360857563813e-06, "loss": 2.0062, "step": 23963 }, { "epoch": 0.59, "learning_rate": 7.573586375984582e-06, "loss": 2.041, "step": 23964 }, { "epoch": 0.59, "learning_rate": 7.572811909869673e-06, "loss": 2.05, "step": 23965 }, { "epoch": 0.59, "learning_rate": 7.572037459224025e-06, "loss": 1.9899, "step": 23966 }, { "epoch": 0.59, "learning_rate": 7.571263024052578e-06, "loss": 1.9647, "step": 23967 }, { "epoch": 0.59, "learning_rate": 7.5704886043602596e-06, "loss": 1.9836, "step": 23968 }, { "epoch": 0.59, "learning_rate": 7.5697142001520115e-06, "loss": 1.9597, "step": 23969 }, { "epoch": 0.59, "learning_rate": 7.568939811432765e-06, "loss": 2.0555, "step": 23970 }, { "epoch": 0.59, "learning_rate": 7.568165438207455e-06, "loss": 2.1178, "step": 23971 }, { "epoch": 0.59, "learning_rate": 7.567391080481025e-06, "loss": 2.0059, "step": 23972 }, { "epoch": 0.59, "learning_rate": 7.5666167382584e-06, "loss": 1.8893, "step": 23973 }, { "epoch": 0.59, "learning_rate": 7.565842411544521e-06, "loss": 2.1983, "step": 23974 }, { "epoch": 0.59, "learning_rate": 7.56506810034432e-06, "loss": 2.1077, "step": 23975 }, { "epoch": 0.59, "learning_rate": 7.5642938046627335e-06, "loss": 1.818, "step": 23976 }, { "epoch": 0.59, "learning_rate": 7.563519524504701e-06, "loss": 2.2072, "step": 23977 }, { "epoch": 0.59, "learning_rate": 7.562745259875148e-06, "loss": 2.0139, "step": 23978 }, { "epoch": 0.59, "learning_rate": 7.561971010779018e-06, "loss": 2.0777, "step": 23979 }, { "epoch": 0.59, "learning_rate": 7.561196777221236e-06, "loss": 1.9834, "step": 23980 }, { "epoch": 0.59, "learning_rate": 7.560422559206743e-06, "loss": 1.984, "step": 23981 }, { "epoch": 0.59, "learning_rate": 7.559648356740473e-06, "loss": 1.9508, "step": 23982 }, { "epoch": 0.59, "learning_rate": 7.5588741698273584e-06, "loss": 1.9273, "step": 23983 }, { "epoch": 0.59, "learning_rate": 7.558099998472338e-06, "loss": 1.807, "step": 23984 }, { "epoch": 0.59, "learning_rate": 7.5573258426803385e-06, "loss": 2.055, "step": 23985 }, { "epoch": 0.59, "learning_rate": 7.556551702456296e-06, "loss": 2.0714, "step": 23986 }, { "epoch": 0.59, "learning_rate": 7.555777577805151e-06, "loss": 1.987, "step": 23987 }, { "epoch": 0.59, "learning_rate": 7.555003468731827e-06, "loss": 2.0817, "step": 23988 }, { "epoch": 0.59, "learning_rate": 7.554229375241266e-06, "loss": 2.1644, "step": 23989 }, { "epoch": 0.59, "learning_rate": 7.553455297338398e-06, "loss": 1.8999, "step": 23990 }, { "epoch": 0.59, "learning_rate": 7.552681235028154e-06, "loss": 1.9788, "step": 23991 }, { "epoch": 0.59, "learning_rate": 7.551907188315473e-06, "loss": 1.9919, "step": 23992 }, { "epoch": 0.59, "learning_rate": 7.551133157205283e-06, "loss": 2.0921, "step": 23993 }, { "epoch": 0.59, "learning_rate": 7.550359141702524e-06, "loss": 1.809, "step": 23994 }, { "epoch": 0.59, "learning_rate": 7.549585141812119e-06, "loss": 1.966, "step": 23995 }, { "epoch": 0.59, "learning_rate": 7.548811157539006e-06, "loss": 2.1183, "step": 23996 }, { "epoch": 0.59, "learning_rate": 7.548037188888124e-06, "loss": 1.9389, "step": 23997 }, { "epoch": 0.59, "learning_rate": 7.547263235864396e-06, "loss": 1.89, "step": 23998 }, { "epoch": 0.59, "learning_rate": 7.54648929847276e-06, "loss": 1.9397, "step": 23999 }, { "epoch": 0.59, "learning_rate": 7.545715376718145e-06, "loss": 1.7971, "step": 24000 }, { "epoch": 0.59, "learning_rate": 7.5449414706054865e-06, "loss": 1.8974, "step": 24001 }, { "epoch": 0.59, "learning_rate": 7.544167580139721e-06, "loss": 2.115, "step": 24002 }, { "epoch": 0.59, "learning_rate": 7.54339370532577e-06, "loss": 2.123, "step": 24003 }, { "epoch": 0.59, "learning_rate": 7.542619846168576e-06, "loss": 2.1023, "step": 24004 }, { "epoch": 0.59, "learning_rate": 7.5418460026730636e-06, "loss": 1.9772, "step": 24005 }, { "epoch": 0.59, "learning_rate": 7.541072174844166e-06, "loss": 2.0721, "step": 24006 }, { "epoch": 0.59, "learning_rate": 7.5402983626868194e-06, "loss": 1.9707, "step": 24007 }, { "epoch": 0.59, "learning_rate": 7.5395245662059535e-06, "loss": 2.221, "step": 24008 }, { "epoch": 0.59, "learning_rate": 7.5387507854064964e-06, "loss": 2.029, "step": 24009 }, { "epoch": 0.59, "learning_rate": 7.5379770202933865e-06, "loss": 2.1139, "step": 24010 }, { "epoch": 0.59, "learning_rate": 7.5372032708715484e-06, "loss": 2.0391, "step": 24011 }, { "epoch": 0.59, "learning_rate": 7.5364295371459195e-06, "loss": 1.9305, "step": 24012 }, { "epoch": 0.59, "learning_rate": 7.535655819121426e-06, "loss": 1.8939, "step": 24013 }, { "epoch": 0.59, "learning_rate": 7.5348821168029995e-06, "loss": 2.0563, "step": 24014 }, { "epoch": 0.59, "learning_rate": 7.534108430195577e-06, "loss": 1.9558, "step": 24015 }, { "epoch": 0.59, "learning_rate": 7.5333347593040805e-06, "loss": 1.8653, "step": 24016 }, { "epoch": 0.59, "learning_rate": 7.5325611041334475e-06, "loss": 1.9985, "step": 24017 }, { "epoch": 0.59, "learning_rate": 7.5317874646886055e-06, "loss": 2.0426, "step": 24018 }, { "epoch": 0.59, "learning_rate": 7.531013840974486e-06, "loss": 1.8037, "step": 24019 }, { "epoch": 0.59, "learning_rate": 7.530240232996024e-06, "loss": 2.0559, "step": 24020 }, { "epoch": 0.59, "learning_rate": 7.529466640758142e-06, "loss": 1.9309, "step": 24021 }, { "epoch": 0.59, "learning_rate": 7.5286930642657785e-06, "loss": 1.9875, "step": 24022 }, { "epoch": 0.59, "learning_rate": 7.527919503523855e-06, "loss": 1.9569, "step": 24023 }, { "epoch": 0.59, "learning_rate": 7.527145958537306e-06, "loss": 1.9677, "step": 24024 }, { "epoch": 0.59, "learning_rate": 7.526372429311064e-06, "loss": 2.1377, "step": 24025 }, { "epoch": 0.59, "learning_rate": 7.525598915850057e-06, "loss": 1.9909, "step": 24026 }, { "epoch": 0.59, "learning_rate": 7.524825418159214e-06, "loss": 1.9596, "step": 24027 }, { "epoch": 0.59, "learning_rate": 7.5240519362434625e-06, "loss": 1.8526, "step": 24028 }, { "epoch": 0.59, "learning_rate": 7.5232784701077355e-06, "loss": 1.8798, "step": 24029 }, { "epoch": 0.59, "learning_rate": 7.522505019756967e-06, "loss": 1.8988, "step": 24030 }, { "epoch": 0.59, "learning_rate": 7.521731585196076e-06, "loss": 2.1394, "step": 24031 }, { "epoch": 0.59, "learning_rate": 7.520958166430001e-06, "loss": 2.1816, "step": 24032 }, { "epoch": 0.59, "learning_rate": 7.5201847634636645e-06, "loss": 2.0559, "step": 24033 }, { "epoch": 0.59, "learning_rate": 7.5194113763019996e-06, "loss": 2.0254, "step": 24034 }, { "epoch": 0.59, "learning_rate": 7.5186380049499344e-06, "loss": 2.1933, "step": 24035 }, { "epoch": 0.59, "learning_rate": 7.517864649412396e-06, "loss": 1.9542, "step": 24036 }, { "epoch": 0.59, "learning_rate": 7.51709130969432e-06, "loss": 2.0239, "step": 24037 }, { "epoch": 0.59, "learning_rate": 7.516317985800626e-06, "loss": 2.004, "step": 24038 }, { "epoch": 0.59, "learning_rate": 7.5155446777362474e-06, "loss": 2.0678, "step": 24039 }, { "epoch": 0.59, "learning_rate": 7.514771385506115e-06, "loss": 1.912, "step": 24040 }, { "epoch": 0.59, "learning_rate": 7.513998109115152e-06, "loss": 2.108, "step": 24041 }, { "epoch": 0.59, "learning_rate": 7.51322484856829e-06, "loss": 1.9292, "step": 24042 }, { "epoch": 0.59, "learning_rate": 7.512451603870455e-06, "loss": 1.9369, "step": 24043 }, { "epoch": 0.59, "learning_rate": 7.511678375026579e-06, "loss": 1.8786, "step": 24044 }, { "epoch": 0.59, "learning_rate": 7.510905162041586e-06, "loss": 1.9033, "step": 24045 }, { "epoch": 0.59, "learning_rate": 7.510131964920405e-06, "loss": 1.982, "step": 24046 }, { "epoch": 0.59, "learning_rate": 7.509358783667968e-06, "loss": 1.9434, "step": 24047 }, { "epoch": 0.59, "learning_rate": 7.5085856182891945e-06, "loss": 1.9187, "step": 24048 }, { "epoch": 0.59, "learning_rate": 7.507812468789018e-06, "loss": 1.9292, "step": 24049 }, { "epoch": 0.59, "learning_rate": 7.507039335172368e-06, "loss": 2.0504, "step": 24050 }, { "epoch": 0.59, "learning_rate": 7.506266217444166e-06, "loss": 1.9994, "step": 24051 }, { "epoch": 0.59, "learning_rate": 7.505493115609343e-06, "loss": 2.0764, "step": 24052 }, { "epoch": 0.59, "learning_rate": 7.504720029672823e-06, "loss": 1.8598, "step": 24053 }, { "epoch": 0.59, "learning_rate": 7.503946959639536e-06, "loss": 1.9901, "step": 24054 }, { "epoch": 0.59, "learning_rate": 7.5031739055144125e-06, "loss": 1.9262, "step": 24055 }, { "epoch": 0.59, "learning_rate": 7.502400867302371e-06, "loss": 1.942, "step": 24056 }, { "epoch": 0.59, "learning_rate": 7.501627845008348e-06, "loss": 2.0108, "step": 24057 }, { "epoch": 0.59, "learning_rate": 7.50085483863726e-06, "loss": 1.9391, "step": 24058 }, { "epoch": 0.59, "learning_rate": 7.500081848194037e-06, "loss": 2.0164, "step": 24059 }, { "epoch": 0.59, "learning_rate": 7.4993088736836105e-06, "loss": 1.9941, "step": 24060 }, { "epoch": 0.59, "learning_rate": 7.498535915110905e-06, "loss": 2.0229, "step": 24061 }, { "epoch": 0.59, "learning_rate": 7.497762972480843e-06, "loss": 1.8454, "step": 24062 }, { "epoch": 0.59, "learning_rate": 7.496990045798351e-06, "loss": 2.0045, "step": 24063 }, { "epoch": 0.59, "learning_rate": 7.4962171350683584e-06, "loss": 2.0428, "step": 24064 }, { "epoch": 0.59, "learning_rate": 7.495444240295795e-06, "loss": 1.9571, "step": 24065 }, { "epoch": 0.59, "learning_rate": 7.4946713614855745e-06, "loss": 2.0194, "step": 24066 }, { "epoch": 0.59, "learning_rate": 7.493898498642632e-06, "loss": 1.9193, "step": 24067 }, { "epoch": 0.59, "learning_rate": 7.4931256517718955e-06, "loss": 2.0511, "step": 24068 }, { "epoch": 0.59, "learning_rate": 7.492352820878282e-06, "loss": 1.9162, "step": 24069 }, { "epoch": 0.59, "learning_rate": 7.491580005966722e-06, "loss": 2.0449, "step": 24070 }, { "epoch": 0.59, "learning_rate": 7.490807207042139e-06, "loss": 1.9738, "step": 24071 }, { "epoch": 0.59, "learning_rate": 7.490034424109458e-06, "loss": 2.0445, "step": 24072 }, { "epoch": 0.59, "learning_rate": 7.489261657173612e-06, "loss": 1.9132, "step": 24073 }, { "epoch": 0.59, "learning_rate": 7.488488906239513e-06, "loss": 1.9169, "step": 24074 }, { "epoch": 0.59, "learning_rate": 7.487716171312098e-06, "loss": 1.9686, "step": 24075 }, { "epoch": 0.59, "learning_rate": 7.486943452396283e-06, "loss": 2.1486, "step": 24076 }, { "epoch": 0.59, "learning_rate": 7.486170749496996e-06, "loss": 1.842, "step": 24077 }, { "epoch": 0.59, "learning_rate": 7.4853980626191645e-06, "loss": 1.9392, "step": 24078 }, { "epoch": 0.59, "learning_rate": 7.484625391767711e-06, "loss": 2.0097, "step": 24079 }, { "epoch": 0.59, "learning_rate": 7.483852736947558e-06, "loss": 1.9637, "step": 24080 }, { "epoch": 0.59, "learning_rate": 7.48308009816363e-06, "loss": 2.0701, "step": 24081 }, { "epoch": 0.59, "learning_rate": 7.482307475420854e-06, "loss": 1.942, "step": 24082 }, { "epoch": 0.59, "learning_rate": 7.481534868724155e-06, "loss": 1.9681, "step": 24083 }, { "epoch": 0.59, "learning_rate": 7.4807622780784515e-06, "loss": 1.9501, "step": 24084 }, { "epoch": 0.59, "learning_rate": 7.479989703488676e-06, "loss": 1.9415, "step": 24085 }, { "epoch": 0.59, "learning_rate": 7.479217144959743e-06, "loss": 2.0625, "step": 24086 }, { "epoch": 0.59, "learning_rate": 7.47844460249658e-06, "loss": 1.959, "step": 24087 }, { "epoch": 0.59, "learning_rate": 7.477672076104114e-06, "loss": 1.7009, "step": 24088 }, { "epoch": 0.59, "learning_rate": 7.476899565787263e-06, "loss": 2.066, "step": 24089 }, { "epoch": 0.59, "learning_rate": 7.4761270715509584e-06, "loss": 1.9799, "step": 24090 }, { "epoch": 0.59, "learning_rate": 7.475354593400115e-06, "loss": 1.9495, "step": 24091 }, { "epoch": 0.59, "learning_rate": 7.474582131339659e-06, "loss": 1.8859, "step": 24092 }, { "epoch": 0.59, "learning_rate": 7.473809685374517e-06, "loss": 2.0025, "step": 24093 }, { "epoch": 0.59, "learning_rate": 7.473037255509605e-06, "loss": 2.0987, "step": 24094 }, { "epoch": 0.59, "learning_rate": 7.472264841749853e-06, "loss": 1.794, "step": 24095 }, { "epoch": 0.59, "learning_rate": 7.471492444100179e-06, "loss": 1.9852, "step": 24096 }, { "epoch": 0.59, "learning_rate": 7.470720062565509e-06, "loss": 1.9664, "step": 24097 }, { "epoch": 0.59, "learning_rate": 7.469947697150766e-06, "loss": 2.0692, "step": 24098 }, { "epoch": 0.59, "learning_rate": 7.4691753478608676e-06, "loss": 1.9374, "step": 24099 }, { "epoch": 0.59, "learning_rate": 7.468403014700743e-06, "loss": 2.103, "step": 24100 }, { "epoch": 0.59, "learning_rate": 7.467630697675307e-06, "loss": 2.14, "step": 24101 }, { "epoch": 0.59, "learning_rate": 7.466858396789487e-06, "loss": 1.9475, "step": 24102 }, { "epoch": 0.59, "learning_rate": 7.466086112048207e-06, "loss": 1.9008, "step": 24103 }, { "epoch": 0.59, "learning_rate": 7.465313843456383e-06, "loss": 1.9883, "step": 24104 }, { "epoch": 0.59, "learning_rate": 7.464541591018941e-06, "loss": 2.0072, "step": 24105 }, { "epoch": 0.59, "learning_rate": 7.4637693547408e-06, "loss": 2.1015, "step": 24106 }, { "epoch": 0.59, "learning_rate": 7.4629971346268835e-06, "loss": 1.9323, "step": 24107 }, { "epoch": 0.59, "learning_rate": 7.462224930682117e-06, "loss": 1.9401, "step": 24108 }, { "epoch": 0.59, "learning_rate": 7.461452742911415e-06, "loss": 1.8357, "step": 24109 }, { "epoch": 0.59, "learning_rate": 7.460680571319705e-06, "loss": 2.036, "step": 24110 }, { "epoch": 0.59, "learning_rate": 7.4599084159119005e-06, "loss": 1.897, "step": 24111 }, { "epoch": 0.59, "learning_rate": 7.459136276692929e-06, "loss": 1.774, "step": 24112 }, { "epoch": 0.59, "learning_rate": 7.45836415366771e-06, "loss": 1.9605, "step": 24113 }, { "epoch": 0.59, "learning_rate": 7.457592046841164e-06, "loss": 2.1434, "step": 24114 }, { "epoch": 0.59, "learning_rate": 7.456819956218216e-06, "loss": 1.8322, "step": 24115 }, { "epoch": 0.59, "learning_rate": 7.45604788180378e-06, "loss": 2.0415, "step": 24116 }, { "epoch": 0.59, "learning_rate": 7.455275823602779e-06, "loss": 2.2077, "step": 24117 }, { "epoch": 0.59, "learning_rate": 7.4545037816201395e-06, "loss": 1.9476, "step": 24118 }, { "epoch": 0.59, "learning_rate": 7.453731755860771e-06, "loss": 2.1229, "step": 24119 }, { "epoch": 0.59, "learning_rate": 7.452959746329607e-06, "loss": 1.998, "step": 24120 }, { "epoch": 0.59, "learning_rate": 7.452187753031553e-06, "loss": 2.0257, "step": 24121 }, { "epoch": 0.59, "learning_rate": 7.451415775971539e-06, "loss": 1.9207, "step": 24122 }, { "epoch": 0.59, "learning_rate": 7.450643815154483e-06, "loss": 2.1594, "step": 24123 }, { "epoch": 0.59, "learning_rate": 7.449871870585306e-06, "loss": 1.8898, "step": 24124 }, { "epoch": 0.59, "learning_rate": 7.449099942268924e-06, "loss": 1.9534, "step": 24125 }, { "epoch": 0.59, "learning_rate": 7.448328030210264e-06, "loss": 1.9124, "step": 24126 }, { "epoch": 0.59, "learning_rate": 7.447556134414238e-06, "loss": 1.992, "step": 24127 }, { "epoch": 0.59, "learning_rate": 7.446784254885771e-06, "loss": 1.8165, "step": 24128 }, { "epoch": 0.59, "learning_rate": 7.446012391629777e-06, "loss": 2.0129, "step": 24129 }, { "epoch": 0.59, "learning_rate": 7.445240544651178e-06, "loss": 1.9485, "step": 24130 }, { "epoch": 0.59, "learning_rate": 7.444468713954896e-06, "loss": 1.9825, "step": 24131 }, { "epoch": 0.59, "learning_rate": 7.443696899545844e-06, "loss": 2.03, "step": 24132 }, { "epoch": 0.59, "learning_rate": 7.442925101428951e-06, "loss": 2.0162, "step": 24133 }, { "epoch": 0.59, "learning_rate": 7.4421533196091245e-06, "loss": 1.9201, "step": 24134 }, { "epoch": 0.59, "learning_rate": 7.441381554091288e-06, "loss": 1.7878, "step": 24135 }, { "epoch": 0.59, "learning_rate": 7.440609804880365e-06, "loss": 2.1243, "step": 24136 }, { "epoch": 0.59, "learning_rate": 7.4398380719812665e-06, "loss": 1.9813, "step": 24137 }, { "epoch": 0.59, "learning_rate": 7.439066355398918e-06, "loss": 2.0405, "step": 24138 }, { "epoch": 0.59, "learning_rate": 7.438294655138231e-06, "loss": 2.0049, "step": 24139 }, { "epoch": 0.6, "learning_rate": 7.4375229712041256e-06, "loss": 1.9462, "step": 24140 }, { "epoch": 0.6, "learning_rate": 7.436751303601524e-06, "loss": 1.989, "step": 24141 }, { "epoch": 0.6, "learning_rate": 7.4359796523353386e-06, "loss": 1.896, "step": 24142 }, { "epoch": 0.6, "learning_rate": 7.435208017410496e-06, "loss": 1.8006, "step": 24143 }, { "epoch": 0.6, "learning_rate": 7.4344363988319045e-06, "loss": 2.132, "step": 24144 }, { "epoch": 0.6, "learning_rate": 7.433664796604485e-06, "loss": 1.904, "step": 24145 }, { "epoch": 0.6, "learning_rate": 7.4328932107331605e-06, "loss": 1.9533, "step": 24146 }, { "epoch": 0.6, "learning_rate": 7.43212164122284e-06, "loss": 2.1331, "step": 24147 }, { "epoch": 0.6, "learning_rate": 7.431350088078449e-06, "loss": 2.0864, "step": 24148 }, { "epoch": 0.6, "learning_rate": 7.4305785513048966e-06, "loss": 2.0922, "step": 24149 }, { "epoch": 0.6, "learning_rate": 7.429807030907104e-06, "loss": 2.12, "step": 24150 }, { "epoch": 0.6, "learning_rate": 7.429035526889995e-06, "loss": 2.0653, "step": 24151 }, { "epoch": 0.6, "learning_rate": 7.428264039258475e-06, "loss": 2.0526, "step": 24152 }, { "epoch": 0.6, "learning_rate": 7.4274925680174715e-06, "loss": 2.0621, "step": 24153 }, { "epoch": 0.6, "learning_rate": 7.426721113171891e-06, "loss": 1.8766, "step": 24154 }, { "epoch": 0.6, "learning_rate": 7.425949674726656e-06, "loss": 1.9269, "step": 24155 }, { "epoch": 0.6, "learning_rate": 7.425178252686687e-06, "loss": 2.1349, "step": 24156 }, { "epoch": 0.6, "learning_rate": 7.4244068470568905e-06, "loss": 2.0375, "step": 24157 }, { "epoch": 0.6, "learning_rate": 7.4236354578421936e-06, "loss": 1.9666, "step": 24158 }, { "epoch": 0.6, "learning_rate": 7.422864085047503e-06, "loss": 1.8474, "step": 24159 }, { "epoch": 0.6, "learning_rate": 7.4220927286777385e-06, "loss": 2.1107, "step": 24160 }, { "epoch": 0.6, "learning_rate": 7.421321388737823e-06, "loss": 2.0369, "step": 24161 }, { "epoch": 0.6, "learning_rate": 7.420550065232663e-06, "loss": 1.9179, "step": 24162 }, { "epoch": 0.6, "learning_rate": 7.419778758167182e-06, "loss": 2.0163, "step": 24163 }, { "epoch": 0.6, "learning_rate": 7.419007467546287e-06, "loss": 1.8303, "step": 24164 }, { "epoch": 0.6, "learning_rate": 7.418236193374899e-06, "loss": 2.1249, "step": 24165 }, { "epoch": 0.6, "learning_rate": 7.417464935657935e-06, "loss": 2.1522, "step": 24166 }, { "epoch": 0.6, "learning_rate": 7.416693694400305e-06, "loss": 1.875, "step": 24167 }, { "epoch": 0.6, "learning_rate": 7.415922469606935e-06, "loss": 2.023, "step": 24168 }, { "epoch": 0.6, "learning_rate": 7.415151261282727e-06, "loss": 1.9761, "step": 24169 }, { "epoch": 0.6, "learning_rate": 7.414380069432603e-06, "loss": 1.8475, "step": 24170 }, { "epoch": 0.6, "learning_rate": 7.413608894061482e-06, "loss": 2.1124, "step": 24171 }, { "epoch": 0.6, "learning_rate": 7.41283773517427e-06, "loss": 1.8941, "step": 24172 }, { "epoch": 0.6, "learning_rate": 7.41206659277589e-06, "loss": 2.1562, "step": 24173 }, { "epoch": 0.6, "learning_rate": 7.411295466871249e-06, "loss": 1.8507, "step": 24174 }, { "epoch": 0.6, "learning_rate": 7.410524357465265e-06, "loss": 1.965, "step": 24175 }, { "epoch": 0.6, "learning_rate": 7.409753264562858e-06, "loss": 1.9437, "step": 24176 }, { "epoch": 0.6, "learning_rate": 7.408982188168931e-06, "loss": 1.7815, "step": 24177 }, { "epoch": 0.6, "learning_rate": 7.408211128288413e-06, "loss": 1.9277, "step": 24178 }, { "epoch": 0.6, "learning_rate": 7.407440084926205e-06, "loss": 1.9944, "step": 24179 }, { "epoch": 0.6, "learning_rate": 7.406669058087226e-06, "loss": 2.0802, "step": 24180 }, { "epoch": 0.6, "learning_rate": 7.4058980477763945e-06, "loss": 2.0801, "step": 24181 }, { "epoch": 0.6, "learning_rate": 7.405127053998616e-06, "loss": 2.0842, "step": 24182 }, { "epoch": 0.6, "learning_rate": 7.404356076758809e-06, "loss": 2.0146, "step": 24183 }, { "epoch": 0.6, "learning_rate": 7.4035851160618854e-06, "loss": 1.9638, "step": 24184 }, { "epoch": 0.6, "learning_rate": 7.40281417191276e-06, "loss": 1.9958, "step": 24185 }, { "epoch": 0.6, "learning_rate": 7.40204324431635e-06, "loss": 2.1061, "step": 24186 }, { "epoch": 0.6, "learning_rate": 7.401272333277563e-06, "loss": 2.0417, "step": 24187 }, { "epoch": 0.6, "learning_rate": 7.400501438801313e-06, "loss": 2.0951, "step": 24188 }, { "epoch": 0.6, "learning_rate": 7.399730560892517e-06, "loss": 1.9001, "step": 24189 }, { "epoch": 0.6, "learning_rate": 7.398959699556083e-06, "loss": 2.0565, "step": 24190 }, { "epoch": 0.6, "learning_rate": 7.398188854796932e-06, "loss": 2.0534, "step": 24191 }, { "epoch": 0.6, "learning_rate": 7.397418026619965e-06, "loss": 1.9931, "step": 24192 }, { "epoch": 0.6, "learning_rate": 7.396647215030101e-06, "loss": 1.9731, "step": 24193 }, { "epoch": 0.6, "learning_rate": 7.395876420032256e-06, "loss": 2.0104, "step": 24194 }, { "epoch": 0.6, "learning_rate": 7.395105641631337e-06, "loss": 1.9185, "step": 24195 }, { "epoch": 0.6, "learning_rate": 7.394334879832264e-06, "loss": 2.0606, "step": 24196 }, { "epoch": 0.6, "learning_rate": 7.393564134639936e-06, "loss": 2.1717, "step": 24197 }, { "epoch": 0.6, "learning_rate": 7.392793406059278e-06, "loss": 1.9159, "step": 24198 }, { "epoch": 0.6, "learning_rate": 7.392022694095198e-06, "loss": 1.9338, "step": 24199 }, { "epoch": 0.6, "learning_rate": 7.3912519987526045e-06, "loss": 2.0136, "step": 24200 }, { "epoch": 0.6, "learning_rate": 7.390481320036414e-06, "loss": 1.8864, "step": 24201 }, { "epoch": 0.6, "learning_rate": 7.3897106579515345e-06, "loss": 1.8851, "step": 24202 }, { "epoch": 0.6, "learning_rate": 7.388940012502879e-06, "loss": 2.0638, "step": 24203 }, { "epoch": 0.6, "learning_rate": 7.3881693836953635e-06, "loss": 2.1105, "step": 24204 }, { "epoch": 0.6, "learning_rate": 7.387398771533894e-06, "loss": 1.912, "step": 24205 }, { "epoch": 0.6, "learning_rate": 7.386628176023385e-06, "loss": 1.8765, "step": 24206 }, { "epoch": 0.6, "learning_rate": 7.385857597168743e-06, "loss": 2.0198, "step": 24207 }, { "epoch": 0.6, "learning_rate": 7.385087034974882e-06, "loss": 1.7653, "step": 24208 }, { "epoch": 0.6, "learning_rate": 7.384316489446719e-06, "loss": 2.0074, "step": 24209 }, { "epoch": 0.6, "learning_rate": 7.383545960589155e-06, "loss": 2.1381, "step": 24210 }, { "epoch": 0.6, "learning_rate": 7.382775448407107e-06, "loss": 2.1119, "step": 24211 }, { "epoch": 0.6, "learning_rate": 7.382004952905482e-06, "loss": 2.2131, "step": 24212 }, { "epoch": 0.6, "learning_rate": 7.381234474089192e-06, "loss": 1.8923, "step": 24213 }, { "epoch": 0.6, "learning_rate": 7.3804640119631534e-06, "loss": 1.9748, "step": 24214 }, { "epoch": 0.6, "learning_rate": 7.379693566532267e-06, "loss": 1.9565, "step": 24215 }, { "epoch": 0.6, "learning_rate": 7.37892313780145e-06, "loss": 2.2218, "step": 24216 }, { "epoch": 0.6, "learning_rate": 7.378152725775607e-06, "loss": 2.0511, "step": 24217 }, { "epoch": 0.6, "learning_rate": 7.377382330459652e-06, "loss": 2.2547, "step": 24218 }, { "epoch": 0.6, "learning_rate": 7.376611951858495e-06, "loss": 1.9796, "step": 24219 }, { "epoch": 0.6, "learning_rate": 7.375841589977042e-06, "loss": 1.9879, "step": 24220 }, { "epoch": 0.6, "learning_rate": 7.375071244820211e-06, "loss": 2.0896, "step": 24221 }, { "epoch": 0.6, "learning_rate": 7.374300916392902e-06, "loss": 2.0637, "step": 24222 }, { "epoch": 0.6, "learning_rate": 7.37353060470003e-06, "loss": 1.9785, "step": 24223 }, { "epoch": 0.6, "learning_rate": 7.372760309746507e-06, "loss": 2.064, "step": 24224 }, { "epoch": 0.6, "learning_rate": 7.371990031537235e-06, "loss": 2.0117, "step": 24225 }, { "epoch": 0.6, "learning_rate": 7.3712197700771285e-06, "loss": 1.8547, "step": 24226 }, { "epoch": 0.6, "learning_rate": 7.370449525371095e-06, "loss": 2.2253, "step": 24227 }, { "epoch": 0.6, "learning_rate": 7.369679297424041e-06, "loss": 2.0642, "step": 24228 }, { "epoch": 0.6, "learning_rate": 7.368909086240881e-06, "loss": 1.8942, "step": 24229 }, { "epoch": 0.6, "learning_rate": 7.368138891826518e-06, "loss": 2.0023, "step": 24230 }, { "epoch": 0.6, "learning_rate": 7.3673687141858684e-06, "loss": 1.9586, "step": 24231 }, { "epoch": 0.6, "learning_rate": 7.366598553323832e-06, "loss": 1.973, "step": 24232 }, { "epoch": 0.6, "learning_rate": 7.365828409245322e-06, "loss": 1.9697, "step": 24233 }, { "epoch": 0.6, "learning_rate": 7.365058281955247e-06, "loss": 2.1471, "step": 24234 }, { "epoch": 0.6, "learning_rate": 7.364288171458512e-06, "loss": 2.2319, "step": 24235 }, { "epoch": 0.6, "learning_rate": 7.36351807776003e-06, "loss": 2.0687, "step": 24236 }, { "epoch": 0.6, "learning_rate": 7.362748000864703e-06, "loss": 2.1121, "step": 24237 }, { "epoch": 0.6, "learning_rate": 7.361977940777445e-06, "loss": 1.7495, "step": 24238 }, { "epoch": 0.6, "learning_rate": 7.361207897503164e-06, "loss": 1.9004, "step": 24239 }, { "epoch": 0.6, "learning_rate": 7.360437871046759e-06, "loss": 1.8742, "step": 24240 }, { "epoch": 0.6, "learning_rate": 7.3596678614131516e-06, "loss": 2.0077, "step": 24241 }, { "epoch": 0.6, "learning_rate": 7.358897868607234e-06, "loss": 1.9463, "step": 24242 }, { "epoch": 0.6, "learning_rate": 7.358127892633922e-06, "loss": 2.026, "step": 24243 }, { "epoch": 0.6, "learning_rate": 7.357357933498125e-06, "loss": 2.0028, "step": 24244 }, { "epoch": 0.6, "learning_rate": 7.356587991204747e-06, "loss": 1.7502, "step": 24245 }, { "epoch": 0.6, "learning_rate": 7.3558180657586905e-06, "loss": 2.1092, "step": 24246 }, { "epoch": 0.6, "learning_rate": 7.355048157164872e-06, "loss": 1.9385, "step": 24247 }, { "epoch": 0.6, "learning_rate": 7.35427826542819e-06, "loss": 1.9691, "step": 24248 }, { "epoch": 0.6, "learning_rate": 7.353508390553561e-06, "loss": 2.1783, "step": 24249 }, { "epoch": 0.6, "learning_rate": 7.352738532545879e-06, "loss": 2.1001, "step": 24250 }, { "epoch": 0.6, "learning_rate": 7.351968691410059e-06, "loss": 2.1712, "step": 24251 }, { "epoch": 0.6, "learning_rate": 7.3511988671510084e-06, "loss": 2.0381, "step": 24252 }, { "epoch": 0.6, "learning_rate": 7.350429059773626e-06, "loss": 1.8245, "step": 24253 }, { "epoch": 0.6, "learning_rate": 7.349659269282825e-06, "loss": 1.8198, "step": 24254 }, { "epoch": 0.6, "learning_rate": 7.348889495683509e-06, "loss": 2.1762, "step": 24255 }, { "epoch": 0.6, "learning_rate": 7.3481197389805815e-06, "loss": 2.0026, "step": 24256 }, { "epoch": 0.6, "learning_rate": 7.347349999178957e-06, "loss": 1.9804, "step": 24257 }, { "epoch": 0.6, "learning_rate": 7.34658027628353e-06, "loss": 2.0544, "step": 24258 }, { "epoch": 0.6, "learning_rate": 7.345810570299218e-06, "loss": 2.1208, "step": 24259 }, { "epoch": 0.6, "learning_rate": 7.345040881230913e-06, "loss": 1.8999, "step": 24260 }, { "epoch": 0.6, "learning_rate": 7.344271209083531e-06, "loss": 1.9991, "step": 24261 }, { "epoch": 0.6, "learning_rate": 7.3435015538619756e-06, "loss": 2.0544, "step": 24262 }, { "epoch": 0.6, "learning_rate": 7.342731915571148e-06, "loss": 2.025, "step": 24263 }, { "epoch": 0.6, "learning_rate": 7.341962294215958e-06, "loss": 2.0096, "step": 24264 }, { "epoch": 0.6, "learning_rate": 7.341192689801305e-06, "loss": 1.9731, "step": 24265 }, { "epoch": 0.6, "learning_rate": 7.340423102332098e-06, "loss": 1.869, "step": 24266 }, { "epoch": 0.6, "learning_rate": 7.339653531813247e-06, "loss": 2.0289, "step": 24267 }, { "epoch": 0.6, "learning_rate": 7.338883978249645e-06, "loss": 1.9373, "step": 24268 }, { "epoch": 0.6, "learning_rate": 7.338114441646207e-06, "loss": 2.0655, "step": 24269 }, { "epoch": 0.6, "learning_rate": 7.337344922007829e-06, "loss": 2.0252, "step": 24270 }, { "epoch": 0.6, "learning_rate": 7.33657541933942e-06, "loss": 1.9948, "step": 24271 }, { "epoch": 0.6, "learning_rate": 7.335805933645885e-06, "loss": 1.8126, "step": 24272 }, { "epoch": 0.6, "learning_rate": 7.335036464932126e-06, "loss": 1.942, "step": 24273 }, { "epoch": 0.6, "learning_rate": 7.334267013203053e-06, "loss": 1.9289, "step": 24274 }, { "epoch": 0.6, "learning_rate": 7.33349757846356e-06, "loss": 1.9148, "step": 24275 }, { "epoch": 0.6, "learning_rate": 7.3327281607185565e-06, "loss": 1.9471, "step": 24276 }, { "epoch": 0.6, "learning_rate": 7.331958759972948e-06, "loss": 1.8671, "step": 24277 }, { "epoch": 0.6, "learning_rate": 7.331189376231634e-06, "loss": 1.8359, "step": 24278 }, { "epoch": 0.6, "learning_rate": 7.330420009499519e-06, "loss": 1.8162, "step": 24279 }, { "epoch": 0.6, "learning_rate": 7.32965065978151e-06, "loss": 2.0219, "step": 24280 }, { "epoch": 0.6, "learning_rate": 7.328881327082504e-06, "loss": 2.0563, "step": 24281 }, { "epoch": 0.6, "learning_rate": 7.328112011407411e-06, "loss": 2.0061, "step": 24282 }, { "epoch": 0.6, "learning_rate": 7.327342712761128e-06, "loss": 1.9779, "step": 24283 }, { "epoch": 0.6, "learning_rate": 7.326573431148565e-06, "loss": 2.1001, "step": 24284 }, { "epoch": 0.6, "learning_rate": 7.325804166574617e-06, "loss": 1.9591, "step": 24285 }, { "epoch": 0.6, "learning_rate": 7.325034919044191e-06, "loss": 1.8841, "step": 24286 }, { "epoch": 0.6, "learning_rate": 7.324265688562193e-06, "loss": 1.9765, "step": 24287 }, { "epoch": 0.6, "learning_rate": 7.323496475133517e-06, "loss": 1.955, "step": 24288 }, { "epoch": 0.6, "learning_rate": 7.322727278763074e-06, "loss": 2.0694, "step": 24289 }, { "epoch": 0.6, "learning_rate": 7.321958099455757e-06, "loss": 2.016, "step": 24290 }, { "epoch": 0.6, "learning_rate": 7.3211889372164765e-06, "loss": 1.9375, "step": 24291 }, { "epoch": 0.6, "learning_rate": 7.320419792050135e-06, "loss": 1.9975, "step": 24292 }, { "epoch": 0.6, "learning_rate": 7.319650663961628e-06, "loss": 2.1671, "step": 24293 }, { "epoch": 0.6, "learning_rate": 7.318881552955865e-06, "loss": 1.9681, "step": 24294 }, { "epoch": 0.6, "learning_rate": 7.3181124590377386e-06, "loss": 1.7673, "step": 24295 }, { "epoch": 0.6, "learning_rate": 7.317343382212157e-06, "loss": 1.9227, "step": 24296 }, { "epoch": 0.6, "learning_rate": 7.316574322484021e-06, "loss": 1.9936, "step": 24297 }, { "epoch": 0.6, "learning_rate": 7.315805279858232e-06, "loss": 1.9018, "step": 24298 }, { "epoch": 0.6, "learning_rate": 7.3150362543396906e-06, "loss": 2.0665, "step": 24299 }, { "epoch": 0.6, "learning_rate": 7.314267245933295e-06, "loss": 1.9232, "step": 24300 }, { "epoch": 0.6, "learning_rate": 7.313498254643952e-06, "loss": 1.829, "step": 24301 }, { "epoch": 0.6, "learning_rate": 7.3127292804765626e-06, "loss": 2.0108, "step": 24302 }, { "epoch": 0.6, "learning_rate": 7.31196032343602e-06, "loss": 1.8442, "step": 24303 }, { "epoch": 0.6, "learning_rate": 7.311191383527232e-06, "loss": 2.0036, "step": 24304 }, { "epoch": 0.6, "learning_rate": 7.310422460755101e-06, "loss": 1.9644, "step": 24305 }, { "epoch": 0.6, "learning_rate": 7.3096535551245205e-06, "loss": 1.8276, "step": 24306 }, { "epoch": 0.6, "learning_rate": 7.308884666640396e-06, "loss": 1.9901, "step": 24307 }, { "epoch": 0.6, "learning_rate": 7.308115795307626e-06, "loss": 1.9545, "step": 24308 }, { "epoch": 0.6, "learning_rate": 7.307346941131111e-06, "loss": 2.0795, "step": 24309 }, { "epoch": 0.6, "learning_rate": 7.306578104115754e-06, "loss": 1.9866, "step": 24310 }, { "epoch": 0.6, "learning_rate": 7.305809284266448e-06, "loss": 1.9093, "step": 24311 }, { "epoch": 0.6, "learning_rate": 7.305040481588104e-06, "loss": 2.1833, "step": 24312 }, { "epoch": 0.6, "learning_rate": 7.30427169608561e-06, "loss": 2.0547, "step": 24313 }, { "epoch": 0.6, "learning_rate": 7.30350292776387e-06, "loss": 2.0272, "step": 24314 }, { "epoch": 0.6, "learning_rate": 7.3027341766277885e-06, "loss": 2.1902, "step": 24315 }, { "epoch": 0.6, "learning_rate": 7.301965442682259e-06, "loss": 1.994, "step": 24316 }, { "epoch": 0.6, "learning_rate": 7.301196725932185e-06, "loss": 2.0438, "step": 24317 }, { "epoch": 0.6, "learning_rate": 7.30042802638246e-06, "loss": 2.0608, "step": 24318 }, { "epoch": 0.6, "learning_rate": 7.299659344037988e-06, "loss": 1.9735, "step": 24319 }, { "epoch": 0.6, "learning_rate": 7.2988906789036715e-06, "loss": 2.1267, "step": 24320 }, { "epoch": 0.6, "learning_rate": 7.298122030984401e-06, "loss": 2.0404, "step": 24321 }, { "epoch": 0.6, "learning_rate": 7.297353400285082e-06, "loss": 1.8223, "step": 24322 }, { "epoch": 0.6, "learning_rate": 7.296584786810607e-06, "loss": 1.9948, "step": 24323 }, { "epoch": 0.6, "learning_rate": 7.295816190565878e-06, "loss": 1.9296, "step": 24324 }, { "epoch": 0.6, "learning_rate": 7.295047611555796e-06, "loss": 2.0622, "step": 24325 }, { "epoch": 0.6, "learning_rate": 7.294279049785256e-06, "loss": 2.0314, "step": 24326 }, { "epoch": 0.6, "learning_rate": 7.293510505259161e-06, "loss": 1.9631, "step": 24327 }, { "epoch": 0.6, "learning_rate": 7.2927419779824004e-06, "loss": 1.9056, "step": 24328 }, { "epoch": 0.6, "learning_rate": 7.291973467959878e-06, "loss": 1.8314, "step": 24329 }, { "epoch": 0.6, "learning_rate": 7.291204975196495e-06, "loss": 2.0135, "step": 24330 }, { "epoch": 0.6, "learning_rate": 7.290436499697142e-06, "loss": 1.9951, "step": 24331 }, { "epoch": 0.6, "learning_rate": 7.289668041466722e-06, "loss": 1.977, "step": 24332 }, { "epoch": 0.6, "learning_rate": 7.288899600510129e-06, "loss": 1.8614, "step": 24333 }, { "epoch": 0.6, "learning_rate": 7.288131176832265e-06, "loss": 2.0641, "step": 24334 }, { "epoch": 0.6, "learning_rate": 7.287362770438025e-06, "loss": 2.0207, "step": 24335 }, { "epoch": 0.6, "learning_rate": 7.286594381332302e-06, "loss": 2.0299, "step": 24336 }, { "epoch": 0.6, "learning_rate": 7.285826009520003e-06, "loss": 1.902, "step": 24337 }, { "epoch": 0.6, "learning_rate": 7.285057655006014e-06, "loss": 1.8934, "step": 24338 }, { "epoch": 0.6, "learning_rate": 7.2842893177952385e-06, "loss": 1.9164, "step": 24339 }, { "epoch": 0.6, "learning_rate": 7.283520997892576e-06, "loss": 2.0731, "step": 24340 }, { "epoch": 0.6, "learning_rate": 7.282752695302915e-06, "loss": 2.1504, "step": 24341 }, { "epoch": 0.6, "learning_rate": 7.281984410031159e-06, "loss": 1.8722, "step": 24342 }, { "epoch": 0.6, "eval_loss": 1.710910439491272, "eval_runtime": 94.8658, "eval_samples_per_second": 620.319, "eval_steps_per_second": 4.849, "step": 24342 }, { "epoch": 0.6, "learning_rate": 7.281216142082199e-06, "loss": 1.9304, "step": 24343 }, { "epoch": 0.6, "learning_rate": 7.280447891460936e-06, "loss": 1.989, "step": 24344 }, { "epoch": 0.6, "learning_rate": 7.279679658172269e-06, "loss": 2.1171, "step": 24345 }, { "epoch": 0.6, "learning_rate": 7.278911442221085e-06, "loss": 1.9313, "step": 24346 }, { "epoch": 0.6, "learning_rate": 7.27814324361229e-06, "loss": 1.8287, "step": 24347 }, { "epoch": 0.6, "learning_rate": 7.277375062350771e-06, "loss": 2.0958, "step": 24348 }, { "epoch": 0.6, "learning_rate": 7.276606898441427e-06, "loss": 1.8642, "step": 24349 }, { "epoch": 0.6, "learning_rate": 7.275838751889158e-06, "loss": 2.0286, "step": 24350 }, { "epoch": 0.6, "learning_rate": 7.275070622698854e-06, "loss": 1.9621, "step": 24351 }, { "epoch": 0.6, "learning_rate": 7.274302510875417e-06, "loss": 1.8217, "step": 24352 }, { "epoch": 0.6, "learning_rate": 7.273534416423734e-06, "loss": 2.0708, "step": 24353 }, { "epoch": 0.6, "learning_rate": 7.272766339348705e-06, "loss": 2.0468, "step": 24354 }, { "epoch": 0.6, "learning_rate": 7.271998279655229e-06, "loss": 2.0466, "step": 24355 }, { "epoch": 0.6, "learning_rate": 7.271230237348193e-06, "loss": 1.8038, "step": 24356 }, { "epoch": 0.6, "learning_rate": 7.2704622124325e-06, "loss": 2.1368, "step": 24357 }, { "epoch": 0.6, "learning_rate": 7.269694204913036e-06, "loss": 1.9823, "step": 24358 }, { "epoch": 0.6, "learning_rate": 7.268926214794703e-06, "loss": 1.9905, "step": 24359 }, { "epoch": 0.6, "learning_rate": 7.268158242082393e-06, "loss": 1.7919, "step": 24360 }, { "epoch": 0.6, "learning_rate": 7.267390286780999e-06, "loss": 1.9788, "step": 24361 }, { "epoch": 0.6, "learning_rate": 7.266622348895418e-06, "loss": 2.0142, "step": 24362 }, { "epoch": 0.6, "learning_rate": 7.2658544284305465e-06, "loss": 2.1661, "step": 24363 }, { "epoch": 0.6, "learning_rate": 7.265086525391272e-06, "loss": 1.9456, "step": 24364 }, { "epoch": 0.6, "learning_rate": 7.264318639782497e-06, "loss": 2.0582, "step": 24365 }, { "epoch": 0.6, "learning_rate": 7.263550771609107e-06, "loss": 1.9602, "step": 24366 }, { "epoch": 0.6, "learning_rate": 7.262782920875999e-06, "loss": 1.9728, "step": 24367 }, { "epoch": 0.6, "learning_rate": 7.26201508758807e-06, "loss": 2.0477, "step": 24368 }, { "epoch": 0.6, "learning_rate": 7.261247271750211e-06, "loss": 1.9022, "step": 24369 }, { "epoch": 0.6, "learning_rate": 7.260479473367316e-06, "loss": 2.0595, "step": 24370 }, { "epoch": 0.6, "learning_rate": 7.259711692444275e-06, "loss": 1.8528, "step": 24371 }, { "epoch": 0.6, "learning_rate": 7.2589439289859855e-06, "loss": 1.9559, "step": 24372 }, { "epoch": 0.6, "learning_rate": 7.258176182997343e-06, "loss": 2.0744, "step": 24373 }, { "epoch": 0.6, "learning_rate": 7.2574084544832346e-06, "loss": 1.81, "step": 24374 }, { "epoch": 0.6, "learning_rate": 7.2566407434485575e-06, "loss": 2.0789, "step": 24375 }, { "epoch": 0.6, "learning_rate": 7.255873049898201e-06, "loss": 1.9406, "step": 24376 }, { "epoch": 0.6, "learning_rate": 7.255105373837059e-06, "loss": 1.8542, "step": 24377 }, { "epoch": 0.6, "learning_rate": 7.254337715270028e-06, "loss": 1.8391, "step": 24378 }, { "epoch": 0.6, "learning_rate": 7.2535700742019945e-06, "loss": 1.951, "step": 24379 }, { "epoch": 0.6, "learning_rate": 7.252802450637857e-06, "loss": 1.9616, "step": 24380 }, { "epoch": 0.6, "learning_rate": 7.252034844582503e-06, "loss": 2.1086, "step": 24381 }, { "epoch": 0.6, "learning_rate": 7.251267256040824e-06, "loss": 1.9555, "step": 24382 }, { "epoch": 0.6, "learning_rate": 7.25049968501772e-06, "loss": 1.9184, "step": 24383 }, { "epoch": 0.6, "learning_rate": 7.249732131518075e-06, "loss": 1.928, "step": 24384 }, { "epoch": 0.6, "learning_rate": 7.2489645955467835e-06, "loss": 1.8426, "step": 24385 }, { "epoch": 0.6, "learning_rate": 7.248197077108734e-06, "loss": 1.905, "step": 24386 }, { "epoch": 0.6, "learning_rate": 7.247429576208825e-06, "loss": 1.8314, "step": 24387 }, { "epoch": 0.6, "learning_rate": 7.246662092851945e-06, "loss": 1.964, "step": 24388 }, { "epoch": 0.6, "learning_rate": 7.245894627042982e-06, "loss": 2.0695, "step": 24389 }, { "epoch": 0.6, "learning_rate": 7.245127178786834e-06, "loss": 2.0988, "step": 24390 }, { "epoch": 0.6, "learning_rate": 7.244359748088384e-06, "loss": 1.9798, "step": 24391 }, { "epoch": 0.6, "learning_rate": 7.2435923349525274e-06, "loss": 1.9527, "step": 24392 }, { "epoch": 0.6, "learning_rate": 7.24282493938416e-06, "loss": 1.9992, "step": 24393 }, { "epoch": 0.6, "learning_rate": 7.242057561388162e-06, "loss": 2.3266, "step": 24394 }, { "epoch": 0.6, "learning_rate": 7.241290200969434e-06, "loss": 2.1613, "step": 24395 }, { "epoch": 0.6, "learning_rate": 7.240522858132859e-06, "loss": 1.9976, "step": 24396 }, { "epoch": 0.6, "learning_rate": 7.239755532883332e-06, "loss": 1.954, "step": 24397 }, { "epoch": 0.6, "learning_rate": 7.2389882252257474e-06, "loss": 1.912, "step": 24398 }, { "epoch": 0.6, "learning_rate": 7.238220935164986e-06, "loss": 1.9719, "step": 24399 }, { "epoch": 0.6, "learning_rate": 7.237453662705946e-06, "loss": 1.9473, "step": 24400 }, { "epoch": 0.6, "learning_rate": 7.23668640785351e-06, "loss": 1.9337, "step": 24401 }, { "epoch": 0.6, "learning_rate": 7.235919170612572e-06, "loss": 2.2319, "step": 24402 }, { "epoch": 0.6, "learning_rate": 7.235151950988024e-06, "loss": 1.9851, "step": 24403 }, { "epoch": 0.6, "learning_rate": 7.234384748984753e-06, "loss": 2.0145, "step": 24404 }, { "epoch": 0.6, "learning_rate": 7.2336175646076515e-06, "loss": 1.958, "step": 24405 }, { "epoch": 0.6, "learning_rate": 7.232850397861604e-06, "loss": 1.9379, "step": 24406 }, { "epoch": 0.6, "learning_rate": 7.232083248751503e-06, "loss": 1.9932, "step": 24407 }, { "epoch": 0.6, "learning_rate": 7.231316117282239e-06, "loss": 1.9059, "step": 24408 }, { "epoch": 0.6, "learning_rate": 7.230549003458698e-06, "loss": 1.9375, "step": 24409 }, { "epoch": 0.6, "learning_rate": 7.229781907285773e-06, "loss": 1.808, "step": 24410 }, { "epoch": 0.6, "learning_rate": 7.229014828768348e-06, "loss": 2.0375, "step": 24411 }, { "epoch": 0.6, "learning_rate": 7.228247767911315e-06, "loss": 1.9691, "step": 24412 }, { "epoch": 0.6, "learning_rate": 7.2274807247195644e-06, "loss": 2.0148, "step": 24413 }, { "epoch": 0.6, "learning_rate": 7.226713699197979e-06, "loss": 2.1589, "step": 24414 }, { "epoch": 0.6, "learning_rate": 7.225946691351457e-06, "loss": 1.9045, "step": 24415 }, { "epoch": 0.6, "learning_rate": 7.225179701184875e-06, "loss": 2.0242, "step": 24416 }, { "epoch": 0.6, "learning_rate": 7.224412728703127e-06, "loss": 2.0638, "step": 24417 }, { "epoch": 0.6, "learning_rate": 7.2236457739111064e-06, "loss": 1.9486, "step": 24418 }, { "epoch": 0.6, "learning_rate": 7.222878836813692e-06, "loss": 1.8825, "step": 24419 }, { "epoch": 0.6, "learning_rate": 7.222111917415777e-06, "loss": 2.0671, "step": 24420 }, { "epoch": 0.6, "learning_rate": 7.2213450157222465e-06, "loss": 2.0131, "step": 24421 }, { "epoch": 0.6, "learning_rate": 7.22057813173799e-06, "loss": 1.979, "step": 24422 }, { "epoch": 0.6, "learning_rate": 7.219811265467897e-06, "loss": 1.9604, "step": 24423 }, { "epoch": 0.6, "learning_rate": 7.219044416916851e-06, "loss": 1.9741, "step": 24424 }, { "epoch": 0.6, "learning_rate": 7.218277586089741e-06, "loss": 2.0186, "step": 24425 }, { "epoch": 0.6, "learning_rate": 7.2175107729914585e-06, "loss": 1.9946, "step": 24426 }, { "epoch": 0.6, "learning_rate": 7.216743977626882e-06, "loss": 1.9232, "step": 24427 }, { "epoch": 0.6, "learning_rate": 7.2159772000009075e-06, "loss": 1.8701, "step": 24428 }, { "epoch": 0.6, "learning_rate": 7.215210440118414e-06, "loss": 2.1355, "step": 24429 }, { "epoch": 0.6, "learning_rate": 7.21444369798429e-06, "loss": 2.0481, "step": 24430 }, { "epoch": 0.6, "learning_rate": 7.213676973603428e-06, "loss": 2.0577, "step": 24431 }, { "epoch": 0.6, "learning_rate": 7.212910266980708e-06, "loss": 2.1612, "step": 24432 }, { "epoch": 0.6, "learning_rate": 7.212143578121024e-06, "loss": 1.9379, "step": 24433 }, { "epoch": 0.6, "learning_rate": 7.211376907029253e-06, "loss": 2.0377, "step": 24434 }, { "epoch": 0.6, "learning_rate": 7.210610253710285e-06, "loss": 1.8997, "step": 24435 }, { "epoch": 0.6, "learning_rate": 7.209843618169012e-06, "loss": 2.1421, "step": 24436 }, { "epoch": 0.6, "learning_rate": 7.209077000410309e-06, "loss": 2.0279, "step": 24437 }, { "epoch": 0.6, "learning_rate": 7.208310400439073e-06, "loss": 2.0734, "step": 24438 }, { "epoch": 0.6, "learning_rate": 7.207543818260181e-06, "loss": 2.1105, "step": 24439 }, { "epoch": 0.6, "learning_rate": 7.206777253878521e-06, "loss": 2.0509, "step": 24440 }, { "epoch": 0.6, "learning_rate": 7.206010707298985e-06, "loss": 1.9498, "step": 24441 }, { "epoch": 0.6, "learning_rate": 7.20524417852645e-06, "loss": 1.8994, "step": 24442 }, { "epoch": 0.6, "learning_rate": 7.204477667565808e-06, "loss": 1.9847, "step": 24443 }, { "epoch": 0.6, "learning_rate": 7.203711174421937e-06, "loss": 1.9851, "step": 24444 }, { "epoch": 0.6, "learning_rate": 7.202944699099725e-06, "loss": 1.9959, "step": 24445 }, { "epoch": 0.6, "learning_rate": 7.202178241604063e-06, "loss": 1.875, "step": 24446 }, { "epoch": 0.6, "learning_rate": 7.201411801939826e-06, "loss": 2.1218, "step": 24447 }, { "epoch": 0.6, "learning_rate": 7.200645380111906e-06, "loss": 2.2012, "step": 24448 }, { "epoch": 0.6, "learning_rate": 7.199878976125183e-06, "loss": 1.9437, "step": 24449 }, { "epoch": 0.6, "learning_rate": 7.1991125899845446e-06, "loss": 2.2068, "step": 24450 }, { "epoch": 0.6, "learning_rate": 7.198346221694879e-06, "loss": 2.0308, "step": 24451 }, { "epoch": 0.6, "learning_rate": 7.19757987126106e-06, "loss": 1.9319, "step": 24452 }, { "epoch": 0.6, "learning_rate": 7.196813538687983e-06, "loss": 1.8355, "step": 24453 }, { "epoch": 0.6, "learning_rate": 7.196047223980522e-06, "loss": 2.0469, "step": 24454 }, { "epoch": 0.6, "learning_rate": 7.1952809271435665e-06, "loss": 1.8847, "step": 24455 }, { "epoch": 0.6, "learning_rate": 7.194514648182003e-06, "loss": 1.981, "step": 24456 }, { "epoch": 0.6, "learning_rate": 7.193748387100708e-06, "loss": 2.1273, "step": 24457 }, { "epoch": 0.6, "learning_rate": 7.192982143904574e-06, "loss": 1.8861, "step": 24458 }, { "epoch": 0.6, "learning_rate": 7.192215918598475e-06, "loss": 1.9727, "step": 24459 }, { "epoch": 0.6, "learning_rate": 7.191449711187299e-06, "loss": 1.8777, "step": 24460 }, { "epoch": 0.6, "learning_rate": 7.190683521675934e-06, "loss": 2.1128, "step": 24461 }, { "epoch": 0.6, "learning_rate": 7.1899173500692534e-06, "loss": 1.796, "step": 24462 }, { "epoch": 0.6, "learning_rate": 7.189151196372148e-06, "loss": 1.9242, "step": 24463 }, { "epoch": 0.6, "learning_rate": 7.188385060589498e-06, "loss": 1.9971, "step": 24464 }, { "epoch": 0.6, "learning_rate": 7.187618942726184e-06, "loss": 1.9656, "step": 24465 }, { "epoch": 0.6, "learning_rate": 7.186852842787093e-06, "loss": 2.0624, "step": 24466 }, { "epoch": 0.6, "learning_rate": 7.186086760777104e-06, "loss": 1.9732, "step": 24467 }, { "epoch": 0.6, "learning_rate": 7.185320696701105e-06, "loss": 2.1016, "step": 24468 }, { "epoch": 0.6, "learning_rate": 7.184554650563971e-06, "loss": 1.9207, "step": 24469 }, { "epoch": 0.6, "learning_rate": 7.183788622370586e-06, "loss": 2.1451, "step": 24470 }, { "epoch": 0.6, "learning_rate": 7.183022612125839e-06, "loss": 2.0784, "step": 24471 }, { "epoch": 0.6, "learning_rate": 7.182256619834604e-06, "loss": 2.0539, "step": 24472 }, { "epoch": 0.6, "learning_rate": 7.181490645501765e-06, "loss": 1.9214, "step": 24473 }, { "epoch": 0.6, "learning_rate": 7.1807246891322055e-06, "loss": 1.9969, "step": 24474 }, { "epoch": 0.6, "learning_rate": 7.179958750730804e-06, "loss": 2.1346, "step": 24475 }, { "epoch": 0.6, "learning_rate": 7.17919283030245e-06, "loss": 2.0071, "step": 24476 }, { "epoch": 0.6, "learning_rate": 7.178426927852016e-06, "loss": 2.022, "step": 24477 }, { "epoch": 0.6, "learning_rate": 7.17766104338439e-06, "loss": 2.0011, "step": 24478 }, { "epoch": 0.6, "learning_rate": 7.176895176904445e-06, "loss": 2.0363, "step": 24479 }, { "epoch": 0.6, "learning_rate": 7.1761293284170675e-06, "loss": 1.8095, "step": 24480 }, { "epoch": 0.6, "learning_rate": 7.175363497927143e-06, "loss": 1.9365, "step": 24481 }, { "epoch": 0.6, "learning_rate": 7.174597685439541e-06, "loss": 1.8952, "step": 24482 }, { "epoch": 0.6, "learning_rate": 7.1738318909591506e-06, "loss": 1.9499, "step": 24483 }, { "epoch": 0.6, "learning_rate": 7.173066114490854e-06, "loss": 1.9559, "step": 24484 }, { "epoch": 0.6, "learning_rate": 7.172300356039524e-06, "loss": 1.9748, "step": 24485 }, { "epoch": 0.6, "learning_rate": 7.171534615610051e-06, "loss": 2.0701, "step": 24486 }, { "epoch": 0.6, "learning_rate": 7.170768893207304e-06, "loss": 1.8882, "step": 24487 }, { "epoch": 0.6, "learning_rate": 7.17000318883617e-06, "loss": 1.9412, "step": 24488 }, { "epoch": 0.6, "learning_rate": 7.169237502501531e-06, "loss": 2.0182, "step": 24489 }, { "epoch": 0.6, "learning_rate": 7.168471834208262e-06, "loss": 1.9083, "step": 24490 }, { "epoch": 0.6, "learning_rate": 7.167706183961245e-06, "loss": 2.0408, "step": 24491 }, { "epoch": 0.6, "learning_rate": 7.166940551765357e-06, "loss": 1.8004, "step": 24492 }, { "epoch": 0.6, "learning_rate": 7.166174937625481e-06, "loss": 2.0239, "step": 24493 }, { "epoch": 0.6, "learning_rate": 7.165409341546499e-06, "loss": 2.0224, "step": 24494 }, { "epoch": 0.6, "learning_rate": 7.164643763533283e-06, "loss": 2.1555, "step": 24495 }, { "epoch": 0.6, "learning_rate": 7.1638782035907216e-06, "loss": 2.1606, "step": 24496 }, { "epoch": 0.6, "learning_rate": 7.163112661723683e-06, "loss": 1.8374, "step": 24497 }, { "epoch": 0.6, "learning_rate": 7.162347137937053e-06, "loss": 1.9911, "step": 24498 }, { "epoch": 0.6, "learning_rate": 7.1615816322357145e-06, "loss": 2.0214, "step": 24499 }, { "epoch": 0.6, "learning_rate": 7.160816144624536e-06, "loss": 1.9031, "step": 24500 }, { "epoch": 0.6, "learning_rate": 7.1600506751084035e-06, "loss": 2.0141, "step": 24501 }, { "epoch": 0.6, "learning_rate": 7.159285223692192e-06, "loss": 2.0489, "step": 24502 }, { "epoch": 0.6, "learning_rate": 7.1585197903807804e-06, "loss": 1.8145, "step": 24503 }, { "epoch": 0.6, "learning_rate": 7.157754375179053e-06, "loss": 2.0543, "step": 24504 }, { "epoch": 0.6, "learning_rate": 7.156988978091879e-06, "loss": 1.8658, "step": 24505 }, { "epoch": 0.6, "learning_rate": 7.156223599124146e-06, "loss": 2.1504, "step": 24506 }, { "epoch": 0.6, "learning_rate": 7.155458238280722e-06, "loss": 1.8891, "step": 24507 }, { "epoch": 0.6, "learning_rate": 7.1546928955664885e-06, "loss": 2.0785, "step": 24508 }, { "epoch": 0.6, "learning_rate": 7.153927570986328e-06, "loss": 2.0099, "step": 24509 }, { "epoch": 0.6, "learning_rate": 7.153162264545111e-06, "loss": 1.9497, "step": 24510 }, { "epoch": 0.6, "learning_rate": 7.1523969762477245e-06, "loss": 2.0444, "step": 24511 }, { "epoch": 0.6, "learning_rate": 7.151631706099035e-06, "loss": 1.8849, "step": 24512 }, { "epoch": 0.6, "learning_rate": 7.150866454103924e-06, "loss": 1.8984, "step": 24513 }, { "epoch": 0.6, "learning_rate": 7.150101220267276e-06, "loss": 2.0691, "step": 24514 }, { "epoch": 0.6, "learning_rate": 7.149336004593956e-06, "loss": 1.965, "step": 24515 }, { "epoch": 0.6, "learning_rate": 7.148570807088848e-06, "loss": 2.0827, "step": 24516 }, { "epoch": 0.6, "learning_rate": 7.14780562775683e-06, "loss": 2.0088, "step": 24517 }, { "epoch": 0.6, "learning_rate": 7.147040466602772e-06, "loss": 2.034, "step": 24518 }, { "epoch": 0.6, "learning_rate": 7.146275323631557e-06, "loss": 2.0343, "step": 24519 }, { "epoch": 0.6, "learning_rate": 7.145510198848058e-06, "loss": 2.0042, "step": 24520 }, { "epoch": 0.6, "learning_rate": 7.144745092257157e-06, "loss": 2.0713, "step": 24521 }, { "epoch": 0.6, "learning_rate": 7.143980003863722e-06, "loss": 2.1043, "step": 24522 }, { "epoch": 0.6, "learning_rate": 7.143214933672632e-06, "loss": 1.9961, "step": 24523 }, { "epoch": 0.6, "learning_rate": 7.142449881688769e-06, "loss": 2.1725, "step": 24524 }, { "epoch": 0.6, "learning_rate": 7.141684847917001e-06, "loss": 2.1254, "step": 24525 }, { "epoch": 0.6, "learning_rate": 7.140919832362208e-06, "loss": 2.3183, "step": 24526 }, { "epoch": 0.6, "learning_rate": 7.140154835029263e-06, "loss": 2.0, "step": 24527 }, { "epoch": 0.6, "learning_rate": 7.139389855923042e-06, "loss": 2.1145, "step": 24528 }, { "epoch": 0.6, "learning_rate": 7.138624895048427e-06, "loss": 1.9295, "step": 24529 }, { "epoch": 0.6, "learning_rate": 7.137859952410284e-06, "loss": 1.8634, "step": 24530 }, { "epoch": 0.6, "learning_rate": 7.137095028013497e-06, "loss": 1.9892, "step": 24531 }, { "epoch": 0.6, "learning_rate": 7.136330121862931e-06, "loss": 1.9563, "step": 24532 }, { "epoch": 0.6, "learning_rate": 7.135565233963466e-06, "loss": 1.7951, "step": 24533 }, { "epoch": 0.6, "learning_rate": 7.13480036431998e-06, "loss": 2.0235, "step": 24534 }, { "epoch": 0.6, "learning_rate": 7.134035512937346e-06, "loss": 1.9895, "step": 24535 }, { "epoch": 0.6, "learning_rate": 7.133270679820436e-06, "loss": 2.1017, "step": 24536 }, { "epoch": 0.6, "learning_rate": 7.132505864974126e-06, "loss": 1.9573, "step": 24537 }, { "epoch": 0.6, "learning_rate": 7.1317410684032895e-06, "loss": 1.9751, "step": 24538 }, { "epoch": 0.6, "learning_rate": 7.1309762901128055e-06, "loss": 2.0642, "step": 24539 }, { "epoch": 0.6, "learning_rate": 7.130211530107541e-06, "loss": 1.874, "step": 24540 }, { "epoch": 0.6, "learning_rate": 7.129446788392373e-06, "loss": 2.1394, "step": 24541 }, { "epoch": 0.6, "learning_rate": 7.128682064972182e-06, "loss": 1.9353, "step": 24542 }, { "epoch": 0.6, "learning_rate": 7.1279173598518305e-06, "loss": 1.8426, "step": 24543 }, { "epoch": 0.6, "learning_rate": 7.1271526730362e-06, "loss": 1.8559, "step": 24544 }, { "epoch": 0.61, "learning_rate": 7.126388004530161e-06, "loss": 2.1755, "step": 24545 }, { "epoch": 0.61, "learning_rate": 7.125623354338585e-06, "loss": 2.1183, "step": 24546 }, { "epoch": 0.61, "learning_rate": 7.1248587224663545e-06, "loss": 1.9286, "step": 24547 }, { "epoch": 0.61, "learning_rate": 7.1240941089183314e-06, "loss": 1.9797, "step": 24548 }, { "epoch": 0.61, "learning_rate": 7.1233295136993975e-06, "loss": 2.0357, "step": 24549 }, { "epoch": 0.61, "learning_rate": 7.122564936814418e-06, "loss": 2.0901, "step": 24550 }, { "epoch": 0.61, "learning_rate": 7.121800378268271e-06, "loss": 2.0209, "step": 24551 }, { "epoch": 0.61, "learning_rate": 7.121035838065829e-06, "loss": 2.0613, "step": 24552 }, { "epoch": 0.61, "learning_rate": 7.120271316211964e-06, "loss": 1.8655, "step": 24553 }, { "epoch": 0.61, "learning_rate": 7.119506812711548e-06, "loss": 2.1337, "step": 24554 }, { "epoch": 0.61, "learning_rate": 7.118742327569453e-06, "loss": 1.9728, "step": 24555 }, { "epoch": 0.61, "learning_rate": 7.117977860790551e-06, "loss": 2.06, "step": 24556 }, { "epoch": 0.61, "learning_rate": 7.11721341237972e-06, "loss": 1.7726, "step": 24557 }, { "epoch": 0.61, "learning_rate": 7.116448982341824e-06, "loss": 1.8761, "step": 24558 }, { "epoch": 0.61, "learning_rate": 7.115684570681741e-06, "loss": 2.0693, "step": 24559 }, { "epoch": 0.61, "learning_rate": 7.114920177404336e-06, "loss": 1.9183, "step": 24560 }, { "epoch": 0.61, "learning_rate": 7.114155802514486e-06, "loss": 2.0088, "step": 24561 }, { "epoch": 0.61, "learning_rate": 7.1133914460170635e-06, "loss": 2.1335, "step": 24562 }, { "epoch": 0.61, "learning_rate": 7.1126271079169365e-06, "loss": 2.1796, "step": 24563 }, { "epoch": 0.61, "learning_rate": 7.1118627882189815e-06, "loss": 1.8601, "step": 24564 }, { "epoch": 0.61, "learning_rate": 7.111098486928061e-06, "loss": 2.0256, "step": 24565 }, { "epoch": 0.61, "learning_rate": 7.110334204049052e-06, "loss": 2.0195, "step": 24566 }, { "epoch": 0.61, "learning_rate": 7.109569939586829e-06, "loss": 1.9961, "step": 24567 }, { "epoch": 0.61, "learning_rate": 7.108805693546255e-06, "loss": 1.864, "step": 24568 }, { "epoch": 0.61, "learning_rate": 7.1080414659322075e-06, "loss": 2.0172, "step": 24569 }, { "epoch": 0.61, "learning_rate": 7.1072772567495515e-06, "loss": 1.9846, "step": 24570 }, { "epoch": 0.61, "learning_rate": 7.10651306600316e-06, "loss": 1.8335, "step": 24571 }, { "epoch": 0.61, "learning_rate": 7.105748893697906e-06, "loss": 1.9432, "step": 24572 }, { "epoch": 0.61, "learning_rate": 7.104984739838654e-06, "loss": 1.9233, "step": 24573 }, { "epoch": 0.61, "learning_rate": 7.104220604430283e-06, "loss": 2.1446, "step": 24574 }, { "epoch": 0.61, "learning_rate": 7.103456487477654e-06, "loss": 2.0056, "step": 24575 }, { "epoch": 0.61, "learning_rate": 7.10269238898564e-06, "loss": 2.1562, "step": 24576 }, { "epoch": 0.61, "learning_rate": 7.101928308959116e-06, "loss": 2.1941, "step": 24577 }, { "epoch": 0.61, "learning_rate": 7.101164247402943e-06, "loss": 2.1281, "step": 24578 }, { "epoch": 0.61, "learning_rate": 7.100400204321998e-06, "loss": 2.0731, "step": 24579 }, { "epoch": 0.61, "learning_rate": 7.099636179721143e-06, "loss": 2.0657, "step": 24580 }, { "epoch": 0.61, "learning_rate": 7.098872173605254e-06, "loss": 2.0492, "step": 24581 }, { "epoch": 0.61, "learning_rate": 7.0981081859792025e-06, "loss": 1.9329, "step": 24582 }, { "epoch": 0.61, "learning_rate": 7.097344216847848e-06, "loss": 1.8676, "step": 24583 }, { "epoch": 0.61, "learning_rate": 7.096580266216069e-06, "loss": 2.0501, "step": 24584 }, { "epoch": 0.61, "learning_rate": 7.095816334088725e-06, "loss": 1.9821, "step": 24585 }, { "epoch": 0.61, "learning_rate": 7.095052420470692e-06, "loss": 2.0914, "step": 24586 }, { "epoch": 0.61, "learning_rate": 7.094288525366837e-06, "loss": 1.9195, "step": 24587 }, { "epoch": 0.61, "learning_rate": 7.093524648782029e-06, "loss": 1.7475, "step": 24588 }, { "epoch": 0.61, "learning_rate": 7.092760790721136e-06, "loss": 2.1502, "step": 24589 }, { "epoch": 0.61, "learning_rate": 7.091996951189023e-06, "loss": 2.0106, "step": 24590 }, { "epoch": 0.61, "learning_rate": 7.091233130190561e-06, "loss": 1.9922, "step": 24591 }, { "epoch": 0.61, "learning_rate": 7.090469327730622e-06, "loss": 2.0706, "step": 24592 }, { "epoch": 0.61, "learning_rate": 7.0897055438140685e-06, "loss": 1.9318, "step": 24593 }, { "epoch": 0.61, "learning_rate": 7.088941778445773e-06, "loss": 2.1227, "step": 24594 }, { "epoch": 0.61, "learning_rate": 7.088178031630595e-06, "loss": 2.081, "step": 24595 }, { "epoch": 0.61, "learning_rate": 7.087414303373408e-06, "loss": 1.9658, "step": 24596 }, { "epoch": 0.61, "learning_rate": 7.086650593679081e-06, "loss": 2.0866, "step": 24597 }, { "epoch": 0.61, "learning_rate": 7.085886902552477e-06, "loss": 1.9826, "step": 24598 }, { "epoch": 0.61, "learning_rate": 7.085123229998465e-06, "loss": 1.8472, "step": 24599 }, { "epoch": 0.61, "learning_rate": 7.084359576021917e-06, "loss": 1.9245, "step": 24600 }, { "epoch": 0.61, "learning_rate": 7.083595940627692e-06, "loss": 1.9983, "step": 24601 }, { "epoch": 0.61, "learning_rate": 7.082832323820663e-06, "loss": 2.1622, "step": 24602 }, { "epoch": 0.61, "learning_rate": 7.0820687256056915e-06, "loss": 1.9533, "step": 24603 }, { "epoch": 0.61, "learning_rate": 7.081305145987647e-06, "loss": 2.06, "step": 24604 }, { "epoch": 0.61, "learning_rate": 7.080541584971399e-06, "loss": 1.9469, "step": 24605 }, { "epoch": 0.61, "learning_rate": 7.079778042561809e-06, "loss": 1.938, "step": 24606 }, { "epoch": 0.61, "learning_rate": 7.079014518763746e-06, "loss": 2.1324, "step": 24607 }, { "epoch": 0.61, "learning_rate": 7.078251013582074e-06, "loss": 1.9704, "step": 24608 }, { "epoch": 0.61, "learning_rate": 7.077487527021658e-06, "loss": 1.7605, "step": 24609 }, { "epoch": 0.61, "learning_rate": 7.076724059087373e-06, "loss": 1.8699, "step": 24610 }, { "epoch": 0.61, "learning_rate": 7.075960609784074e-06, "loss": 1.8251, "step": 24611 }, { "epoch": 0.61, "learning_rate": 7.075197179116634e-06, "loss": 2.0011, "step": 24612 }, { "epoch": 0.61, "learning_rate": 7.074433767089912e-06, "loss": 2.0382, "step": 24613 }, { "epoch": 0.61, "learning_rate": 7.0736703737087766e-06, "loss": 1.9005, "step": 24614 }, { "epoch": 0.61, "learning_rate": 7.072906998978096e-06, "loss": 1.9361, "step": 24615 }, { "epoch": 0.61, "learning_rate": 7.0721436429027314e-06, "loss": 2.05, "step": 24616 }, { "epoch": 0.61, "learning_rate": 7.071380305487553e-06, "loss": 1.9592, "step": 24617 }, { "epoch": 0.61, "learning_rate": 7.070616986737418e-06, "loss": 2.0291, "step": 24618 }, { "epoch": 0.61, "learning_rate": 7.069853686657196e-06, "loss": 2.0215, "step": 24619 }, { "epoch": 0.61, "learning_rate": 7.069090405251755e-06, "loss": 1.8396, "step": 24620 }, { "epoch": 0.61, "learning_rate": 7.0683271425259525e-06, "loss": 1.9799, "step": 24621 }, { "epoch": 0.61, "learning_rate": 7.0675638984846596e-06, "loss": 2.017, "step": 24622 }, { "epoch": 0.61, "learning_rate": 7.066800673132734e-06, "loss": 2.0869, "step": 24623 }, { "epoch": 0.61, "learning_rate": 7.066037466475046e-06, "loss": 1.969, "step": 24624 }, { "epoch": 0.61, "learning_rate": 7.065274278516456e-06, "loss": 2.0352, "step": 24625 }, { "epoch": 0.61, "learning_rate": 7.064511109261829e-06, "loss": 2.01, "step": 24626 }, { "epoch": 0.61, "learning_rate": 7.063747958716034e-06, "loss": 1.9339, "step": 24627 }, { "epoch": 0.61, "learning_rate": 7.062984826883925e-06, "loss": 2.046, "step": 24628 }, { "epoch": 0.61, "learning_rate": 7.062221713770372e-06, "loss": 1.9278, "step": 24629 }, { "epoch": 0.61, "learning_rate": 7.061458619380241e-06, "loss": 2.048, "step": 24630 }, { "epoch": 0.61, "learning_rate": 7.060695543718387e-06, "loss": 2.1667, "step": 24631 }, { "epoch": 0.61, "learning_rate": 7.05993248678968e-06, "loss": 1.9333, "step": 24632 }, { "epoch": 0.61, "learning_rate": 7.05916944859898e-06, "loss": 1.7943, "step": 24633 }, { "epoch": 0.61, "learning_rate": 7.0584064291511526e-06, "loss": 2.0321, "step": 24634 }, { "epoch": 0.61, "learning_rate": 7.057643428451062e-06, "loss": 2.0288, "step": 24635 }, { "epoch": 0.61, "learning_rate": 7.056880446503564e-06, "loss": 1.9462, "step": 24636 }, { "epoch": 0.61, "learning_rate": 7.056117483313533e-06, "loss": 1.9345, "step": 24637 }, { "epoch": 0.61, "learning_rate": 7.055354538885819e-06, "loss": 1.837, "step": 24638 }, { "epoch": 0.61, "learning_rate": 7.05459161322529e-06, "loss": 1.8658, "step": 24639 }, { "epoch": 0.61, "learning_rate": 7.05382870633681e-06, "loss": 1.8784, "step": 24640 }, { "epoch": 0.61, "learning_rate": 7.0530658182252385e-06, "loss": 2.0765, "step": 24641 }, { "epoch": 0.61, "learning_rate": 7.052302948895443e-06, "loss": 1.9148, "step": 24642 }, { "epoch": 0.61, "learning_rate": 7.051540098352277e-06, "loss": 1.8269, "step": 24643 }, { "epoch": 0.61, "learning_rate": 7.050777266600608e-06, "loss": 2.0426, "step": 24644 }, { "epoch": 0.61, "learning_rate": 7.050014453645299e-06, "loss": 1.9005, "step": 24645 }, { "epoch": 0.61, "learning_rate": 7.049251659491207e-06, "loss": 2.0796, "step": 24646 }, { "epoch": 0.61, "learning_rate": 7.048488884143199e-06, "loss": 2.0894, "step": 24647 }, { "epoch": 0.61, "learning_rate": 7.047726127606128e-06, "loss": 2.026, "step": 24648 }, { "epoch": 0.61, "learning_rate": 7.046963389884862e-06, "loss": 1.9488, "step": 24649 }, { "epoch": 0.61, "learning_rate": 7.046200670984262e-06, "loss": 1.9574, "step": 24650 }, { "epoch": 0.61, "learning_rate": 7.0454379709091855e-06, "loss": 1.9789, "step": 24651 }, { "epoch": 0.61, "learning_rate": 7.0446752896645e-06, "loss": 1.9393, "step": 24652 }, { "epoch": 0.61, "learning_rate": 7.043912627255058e-06, "loss": 1.9837, "step": 24653 }, { "epoch": 0.61, "learning_rate": 7.043149983685724e-06, "loss": 1.8145, "step": 24654 }, { "epoch": 0.61, "learning_rate": 7.042387358961364e-06, "loss": 2.0766, "step": 24655 }, { "epoch": 0.61, "learning_rate": 7.0416247530868285e-06, "loss": 2.0114, "step": 24656 }, { "epoch": 0.61, "learning_rate": 7.040862166066984e-06, "loss": 1.9221, "step": 24657 }, { "epoch": 0.61, "learning_rate": 7.040099597906688e-06, "loss": 2.0223, "step": 24658 }, { "epoch": 0.61, "learning_rate": 7.039337048610803e-06, "loss": 2.0171, "step": 24659 }, { "epoch": 0.61, "learning_rate": 7.038574518184189e-06, "loss": 2.1177, "step": 24660 }, { "epoch": 0.61, "learning_rate": 7.037812006631701e-06, "loss": 2.0458, "step": 24661 }, { "epoch": 0.61, "learning_rate": 7.037049513958204e-06, "loss": 2.0074, "step": 24662 }, { "epoch": 0.61, "learning_rate": 7.036287040168559e-06, "loss": 2.2081, "step": 24663 }, { "epoch": 0.61, "learning_rate": 7.035524585267619e-06, "loss": 2.1501, "step": 24664 }, { "epoch": 0.61, "learning_rate": 7.034762149260252e-06, "loss": 2.0502, "step": 24665 }, { "epoch": 0.61, "learning_rate": 7.0339997321513065e-06, "loss": 2.071, "step": 24666 }, { "epoch": 0.61, "learning_rate": 7.033237333945648e-06, "loss": 2.1126, "step": 24667 }, { "epoch": 0.61, "learning_rate": 7.0324749546481365e-06, "loss": 1.9997, "step": 24668 }, { "epoch": 0.61, "learning_rate": 7.031712594263628e-06, "loss": 1.9495, "step": 24669 }, { "epoch": 0.61, "learning_rate": 7.0309502527969865e-06, "loss": 2.1096, "step": 24670 }, { "epoch": 0.61, "learning_rate": 7.0301879302530605e-06, "loss": 1.9991, "step": 24671 }, { "epoch": 0.61, "learning_rate": 7.029425626636717e-06, "loss": 1.7747, "step": 24672 }, { "epoch": 0.61, "learning_rate": 7.028663341952815e-06, "loss": 2.0023, "step": 24673 }, { "epoch": 0.61, "learning_rate": 7.027901076206207e-06, "loss": 2.1126, "step": 24674 }, { "epoch": 0.61, "learning_rate": 7.027138829401754e-06, "loss": 1.9988, "step": 24675 }, { "epoch": 0.61, "learning_rate": 7.026376601544314e-06, "loss": 1.9331, "step": 24676 }, { "epoch": 0.61, "learning_rate": 7.0256143926387465e-06, "loss": 1.994, "step": 24677 }, { "epoch": 0.61, "learning_rate": 7.024852202689907e-06, "loss": 1.8522, "step": 24678 }, { "epoch": 0.61, "learning_rate": 7.024090031702653e-06, "loss": 1.9484, "step": 24679 }, { "epoch": 0.61, "learning_rate": 7.0233278796818465e-06, "loss": 2.0752, "step": 24680 }, { "epoch": 0.61, "learning_rate": 7.022565746632337e-06, "loss": 1.9864, "step": 24681 }, { "epoch": 0.61, "learning_rate": 7.0218036325589866e-06, "loss": 1.8795, "step": 24682 }, { "epoch": 0.61, "learning_rate": 7.021041537466657e-06, "loss": 2.0839, "step": 24683 }, { "epoch": 0.61, "learning_rate": 7.020279461360195e-06, "loss": 2.0934, "step": 24684 }, { "epoch": 0.61, "learning_rate": 7.019517404244467e-06, "loss": 1.9343, "step": 24685 }, { "epoch": 0.61, "learning_rate": 7.018755366124322e-06, "loss": 1.9193, "step": 24686 }, { "epoch": 0.61, "learning_rate": 7.0179933470046215e-06, "loss": 2.0142, "step": 24687 }, { "epoch": 0.61, "learning_rate": 7.017231346890225e-06, "loss": 1.7292, "step": 24688 }, { "epoch": 0.61, "learning_rate": 7.0164693657859824e-06, "loss": 1.9461, "step": 24689 }, { "epoch": 0.61, "learning_rate": 7.015707403696757e-06, "loss": 1.9919, "step": 24690 }, { "epoch": 0.61, "learning_rate": 7.014945460627394e-06, "loss": 2.0456, "step": 24691 }, { "epoch": 0.61, "learning_rate": 7.014183536582758e-06, "loss": 2.1468, "step": 24692 }, { "epoch": 0.61, "learning_rate": 7.013421631567706e-06, "loss": 1.771, "step": 24693 }, { "epoch": 0.61, "learning_rate": 7.012659745587089e-06, "loss": 2.0397, "step": 24694 }, { "epoch": 0.61, "learning_rate": 7.0118978786457685e-06, "loss": 1.9651, "step": 24695 }, { "epoch": 0.61, "learning_rate": 7.0111360307485935e-06, "loss": 1.8254, "step": 24696 }, { "epoch": 0.61, "learning_rate": 7.01037420190042e-06, "loss": 2.1843, "step": 24697 }, { "epoch": 0.61, "learning_rate": 7.0096123921061136e-06, "loss": 1.8543, "step": 24698 }, { "epoch": 0.61, "learning_rate": 7.008850601370516e-06, "loss": 1.8264, "step": 24699 }, { "epoch": 0.61, "learning_rate": 7.008088829698492e-06, "loss": 2.0602, "step": 24700 }, { "epoch": 0.61, "learning_rate": 7.007327077094889e-06, "loss": 1.9762, "step": 24701 }, { "epoch": 0.61, "learning_rate": 7.006565343564566e-06, "loss": 1.8613, "step": 24702 }, { "epoch": 0.61, "learning_rate": 7.005803629112381e-06, "loss": 2.0318, "step": 24703 }, { "epoch": 0.61, "learning_rate": 7.005041933743182e-06, "loss": 2.0579, "step": 24704 }, { "epoch": 0.61, "learning_rate": 7.004280257461831e-06, "loss": 1.9213, "step": 24705 }, { "epoch": 0.61, "learning_rate": 7.003518600273174e-06, "loss": 1.979, "step": 24706 }, { "epoch": 0.61, "learning_rate": 7.00275696218207e-06, "loss": 1.8497, "step": 24707 }, { "epoch": 0.61, "learning_rate": 7.0019953431933765e-06, "loss": 1.8246, "step": 24708 }, { "epoch": 0.61, "learning_rate": 7.0012337433119395e-06, "loss": 2.1013, "step": 24709 }, { "epoch": 0.61, "learning_rate": 7.000472162542618e-06, "loss": 2.0399, "step": 24710 }, { "epoch": 0.61, "learning_rate": 6.999710600890265e-06, "loss": 1.9878, "step": 24711 }, { "epoch": 0.61, "learning_rate": 6.998949058359733e-06, "loss": 2.0885, "step": 24712 }, { "epoch": 0.61, "learning_rate": 6.99818753495588e-06, "loss": 2.0342, "step": 24713 }, { "epoch": 0.61, "learning_rate": 6.997426030683554e-06, "loss": 1.8134, "step": 24714 }, { "epoch": 0.61, "learning_rate": 6.9966645455476135e-06, "loss": 1.8776, "step": 24715 }, { "epoch": 0.61, "learning_rate": 6.995903079552903e-06, "loss": 2.0216, "step": 24716 }, { "epoch": 0.61, "learning_rate": 6.995141632704283e-06, "loss": 2.052, "step": 24717 }, { "epoch": 0.61, "learning_rate": 6.994380205006609e-06, "loss": 2.1804, "step": 24718 }, { "epoch": 0.61, "learning_rate": 6.993618796464725e-06, "loss": 1.9733, "step": 24719 }, { "epoch": 0.61, "learning_rate": 6.9928574070834876e-06, "loss": 1.9675, "step": 24720 }, { "epoch": 0.61, "learning_rate": 6.992096036867753e-06, "loss": 2.0896, "step": 24721 }, { "epoch": 0.61, "learning_rate": 6.991334685822366e-06, "loss": 1.9054, "step": 24722 }, { "epoch": 0.61, "learning_rate": 6.99057335395219e-06, "loss": 2.0402, "step": 24723 }, { "epoch": 0.61, "learning_rate": 6.989812041262067e-06, "loss": 1.9099, "step": 24724 }, { "epoch": 0.61, "learning_rate": 6.989050747756852e-06, "loss": 2.142, "step": 24725 }, { "epoch": 0.61, "learning_rate": 6.988289473441401e-06, "loss": 1.9621, "step": 24726 }, { "epoch": 0.61, "learning_rate": 6.9875282183205606e-06, "loss": 2.0303, "step": 24727 }, { "epoch": 0.61, "learning_rate": 6.986766982399184e-06, "loss": 2.0227, "step": 24728 }, { "epoch": 0.61, "learning_rate": 6.9860057656821234e-06, "loss": 2.0253, "step": 24729 }, { "epoch": 0.61, "learning_rate": 6.98524456817423e-06, "loss": 1.8581, "step": 24730 }, { "epoch": 0.61, "learning_rate": 6.984483389880359e-06, "loss": 1.9392, "step": 24731 }, { "epoch": 0.61, "learning_rate": 6.983722230805354e-06, "loss": 2.04, "step": 24732 }, { "epoch": 0.61, "learning_rate": 6.982961090954074e-06, "loss": 1.925, "step": 24733 }, { "epoch": 0.61, "learning_rate": 6.982199970331364e-06, "loss": 1.8875, "step": 24734 }, { "epoch": 0.61, "learning_rate": 6.981438868942076e-06, "loss": 2.0854, "step": 24735 }, { "epoch": 0.61, "learning_rate": 6.980677786791066e-06, "loss": 1.9494, "step": 24736 }, { "epoch": 0.61, "learning_rate": 6.9799167238831755e-06, "loss": 1.9725, "step": 24737 }, { "epoch": 0.61, "learning_rate": 6.979155680223262e-06, "loss": 2.0772, "step": 24738 }, { "epoch": 0.61, "learning_rate": 6.978394655816172e-06, "loss": 2.0248, "step": 24739 }, { "epoch": 0.61, "learning_rate": 6.977633650666758e-06, "loss": 1.8755, "step": 24740 }, { "epoch": 0.61, "learning_rate": 6.976872664779873e-06, "loss": 1.9746, "step": 24741 }, { "epoch": 0.61, "learning_rate": 6.976111698160362e-06, "loss": 1.8727, "step": 24742 }, { "epoch": 0.61, "learning_rate": 6.975350750813078e-06, "loss": 2.0238, "step": 24743 }, { "epoch": 0.61, "learning_rate": 6.9745898227428655e-06, "loss": 2.0888, "step": 24744 }, { "epoch": 0.61, "learning_rate": 6.973828913954579e-06, "loss": 1.9689, "step": 24745 }, { "epoch": 0.61, "learning_rate": 6.973068024453069e-06, "loss": 1.9521, "step": 24746 }, { "epoch": 0.61, "learning_rate": 6.972307154243181e-06, "loss": 2.1106, "step": 24747 }, { "epoch": 0.61, "learning_rate": 6.97154630332977e-06, "loss": 2.0883, "step": 24748 }, { "epoch": 0.61, "learning_rate": 6.970785471717677e-06, "loss": 1.9213, "step": 24749 }, { "epoch": 0.61, "learning_rate": 6.970024659411756e-06, "loss": 2.0839, "step": 24750 }, { "epoch": 0.61, "learning_rate": 6.9692638664168574e-06, "loss": 2.1523, "step": 24751 }, { "epoch": 0.61, "learning_rate": 6.968503092737826e-06, "loss": 1.9116, "step": 24752 }, { "epoch": 0.61, "learning_rate": 6.967742338379514e-06, "loss": 2.0739, "step": 24753 }, { "epoch": 0.61, "learning_rate": 6.9669816033467675e-06, "loss": 2.0326, "step": 24754 }, { "epoch": 0.61, "learning_rate": 6.966220887644435e-06, "loss": 2.0116, "step": 24755 }, { "epoch": 0.61, "learning_rate": 6.965460191277367e-06, "loss": 1.9092, "step": 24756 }, { "epoch": 0.61, "learning_rate": 6.964699514250407e-06, "loss": 1.9338, "step": 24757 }, { "epoch": 0.61, "learning_rate": 6.963938856568411e-06, "loss": 2.0529, "step": 24758 }, { "epoch": 0.61, "learning_rate": 6.963178218236217e-06, "loss": 2.0033, "step": 24759 }, { "epoch": 0.61, "learning_rate": 6.962417599258679e-06, "loss": 2.0627, "step": 24760 }, { "epoch": 0.61, "learning_rate": 6.961656999640649e-06, "loss": 1.9353, "step": 24761 }, { "epoch": 0.61, "learning_rate": 6.960896419386962e-06, "loss": 2.0515, "step": 24762 }, { "epoch": 0.61, "learning_rate": 6.960135858502477e-06, "loss": 2.1618, "step": 24763 }, { "epoch": 0.61, "learning_rate": 6.9593753169920345e-06, "loss": 1.796, "step": 24764 }, { "epoch": 0.61, "learning_rate": 6.958614794860483e-06, "loss": 1.8982, "step": 24765 }, { "epoch": 0.61, "learning_rate": 6.957854292112675e-06, "loss": 1.8971, "step": 24766 }, { "epoch": 0.61, "learning_rate": 6.957093808753449e-06, "loss": 2.0128, "step": 24767 }, { "epoch": 0.61, "learning_rate": 6.95633334478766e-06, "loss": 1.8646, "step": 24768 }, { "epoch": 0.61, "learning_rate": 6.955572900220147e-06, "loss": 2.083, "step": 24769 }, { "epoch": 0.61, "learning_rate": 6.954812475055759e-06, "loss": 2.0778, "step": 24770 }, { "epoch": 0.61, "learning_rate": 6.954052069299347e-06, "loss": 2.0186, "step": 24771 }, { "epoch": 0.61, "learning_rate": 6.953291682955751e-06, "loss": 2.1971, "step": 24772 }, { "epoch": 0.61, "learning_rate": 6.952531316029823e-06, "loss": 2.1023, "step": 24773 }, { "epoch": 0.61, "learning_rate": 6.951770968526401e-06, "loss": 1.9816, "step": 24774 }, { "epoch": 0.61, "learning_rate": 6.9510106404503374e-06, "loss": 1.9142, "step": 24775 }, { "epoch": 0.61, "learning_rate": 6.950250331806482e-06, "loss": 1.9692, "step": 24776 }, { "epoch": 0.61, "learning_rate": 6.949490042599669e-06, "loss": 2.1041, "step": 24777 }, { "epoch": 0.61, "learning_rate": 6.948729772834753e-06, "loss": 1.912, "step": 24778 }, { "epoch": 0.61, "learning_rate": 6.947969522516578e-06, "loss": 1.9692, "step": 24779 }, { "epoch": 0.61, "learning_rate": 6.947209291649984e-06, "loss": 2.0793, "step": 24780 }, { "epoch": 0.61, "learning_rate": 6.946449080239823e-06, "loss": 1.9417, "step": 24781 }, { "epoch": 0.61, "learning_rate": 6.945688888290936e-06, "loss": 2.0637, "step": 24782 }, { "epoch": 0.61, "learning_rate": 6.944928715808167e-06, "loss": 1.9367, "step": 24783 }, { "epoch": 0.61, "learning_rate": 6.944168562796367e-06, "loss": 1.8669, "step": 24784 }, { "epoch": 0.61, "learning_rate": 6.943408429260374e-06, "loss": 1.8934, "step": 24785 }, { "epoch": 0.61, "learning_rate": 6.942648315205038e-06, "loss": 1.794, "step": 24786 }, { "epoch": 0.61, "learning_rate": 6.941888220635198e-06, "loss": 2.0466, "step": 24787 }, { "epoch": 0.61, "learning_rate": 6.9411281455557e-06, "loss": 1.9714, "step": 24788 }, { "epoch": 0.61, "learning_rate": 6.940368089971395e-06, "loss": 2.0049, "step": 24789 }, { "epoch": 0.61, "learning_rate": 6.9396080538871144e-06, "loss": 1.9525, "step": 24790 }, { "epoch": 0.61, "learning_rate": 6.938848037307713e-06, "loss": 2.0074, "step": 24791 }, { "epoch": 0.61, "learning_rate": 6.938088040238027e-06, "loss": 1.8479, "step": 24792 }, { "epoch": 0.61, "learning_rate": 6.937328062682904e-06, "loss": 1.8103, "step": 24793 }, { "epoch": 0.61, "learning_rate": 6.936568104647193e-06, "loss": 2.0365, "step": 24794 }, { "epoch": 0.61, "learning_rate": 6.935808166135727e-06, "loss": 1.9815, "step": 24795 }, { "epoch": 0.61, "learning_rate": 6.935048247153358e-06, "loss": 1.8603, "step": 24796 }, { "epoch": 0.61, "learning_rate": 6.93428834770492e-06, "loss": 1.8998, "step": 24797 }, { "epoch": 0.61, "learning_rate": 6.933528467795264e-06, "loss": 2.0215, "step": 24798 }, { "epoch": 0.61, "learning_rate": 6.932768607429232e-06, "loss": 1.9136, "step": 24799 }, { "epoch": 0.61, "learning_rate": 6.932008766611662e-06, "loss": 1.9132, "step": 24800 }, { "epoch": 0.61, "learning_rate": 6.9312489453474044e-06, "loss": 1.9866, "step": 24801 }, { "epoch": 0.61, "learning_rate": 6.930489143641294e-06, "loss": 2.1065, "step": 24802 }, { "epoch": 0.61, "learning_rate": 6.929729361498176e-06, "loss": 1.9764, "step": 24803 }, { "epoch": 0.61, "learning_rate": 6.928969598922897e-06, "loss": 2.0296, "step": 24804 }, { "epoch": 0.61, "learning_rate": 6.928209855920292e-06, "loss": 1.8931, "step": 24805 }, { "epoch": 0.61, "learning_rate": 6.927450132495208e-06, "loss": 2.0634, "step": 24806 }, { "epoch": 0.61, "learning_rate": 6.926690428652487e-06, "loss": 1.9121, "step": 24807 }, { "epoch": 0.61, "learning_rate": 6.925930744396967e-06, "loss": 1.8464, "step": 24808 }, { "epoch": 0.61, "learning_rate": 6.9251710797334924e-06, "loss": 2.115, "step": 24809 }, { "epoch": 0.61, "learning_rate": 6.9244114346669045e-06, "loss": 1.9228, "step": 24810 }, { "epoch": 0.61, "learning_rate": 6.923651809202049e-06, "loss": 2.0507, "step": 24811 }, { "epoch": 0.61, "learning_rate": 6.922892203343758e-06, "loss": 2.174, "step": 24812 }, { "epoch": 0.61, "learning_rate": 6.922132617096878e-06, "loss": 1.9224, "step": 24813 }, { "epoch": 0.61, "learning_rate": 6.921373050466255e-06, "loss": 1.9401, "step": 24814 }, { "epoch": 0.61, "learning_rate": 6.920613503456719e-06, "loss": 2.1221, "step": 24815 }, { "epoch": 0.61, "learning_rate": 6.91985397607312e-06, "loss": 1.8961, "step": 24816 }, { "epoch": 0.61, "learning_rate": 6.919094468320293e-06, "loss": 1.9538, "step": 24817 }, { "epoch": 0.61, "learning_rate": 6.918334980203081e-06, "loss": 1.8778, "step": 24818 }, { "epoch": 0.61, "learning_rate": 6.9175755117263285e-06, "loss": 2.0489, "step": 24819 }, { "epoch": 0.61, "learning_rate": 6.916816062894868e-06, "loss": 1.9646, "step": 24820 }, { "epoch": 0.61, "learning_rate": 6.916056633713547e-06, "loss": 2.0081, "step": 24821 }, { "epoch": 0.61, "learning_rate": 6.915297224187198e-06, "loss": 2.1225, "step": 24822 }, { "epoch": 0.61, "learning_rate": 6.914537834320666e-06, "loss": 1.9188, "step": 24823 }, { "epoch": 0.61, "learning_rate": 6.9137784641187926e-06, "loss": 1.9322, "step": 24824 }, { "epoch": 0.61, "learning_rate": 6.913019113586412e-06, "loss": 1.9536, "step": 24825 }, { "epoch": 0.61, "learning_rate": 6.912259782728369e-06, "loss": 2.0038, "step": 24826 }, { "epoch": 0.61, "learning_rate": 6.911500471549497e-06, "loss": 1.7955, "step": 24827 }, { "epoch": 0.61, "learning_rate": 6.910741180054641e-06, "loss": 2.0866, "step": 24828 }, { "epoch": 0.61, "learning_rate": 6.90998190824864e-06, "loss": 1.9244, "step": 24829 }, { "epoch": 0.61, "learning_rate": 6.90922265613633e-06, "loss": 1.9068, "step": 24830 }, { "epoch": 0.61, "learning_rate": 6.908463423722552e-06, "loss": 1.9498, "step": 24831 }, { "epoch": 0.61, "learning_rate": 6.907704211012143e-06, "loss": 1.9719, "step": 24832 }, { "epoch": 0.61, "learning_rate": 6.90694501800994e-06, "loss": 1.9549, "step": 24833 }, { "epoch": 0.61, "learning_rate": 6.9061858447207886e-06, "loss": 2.1033, "step": 24834 }, { "epoch": 0.61, "learning_rate": 6.90542669114952e-06, "loss": 1.9307, "step": 24835 }, { "epoch": 0.61, "learning_rate": 6.904667557300977e-06, "loss": 1.8982, "step": 24836 }, { "epoch": 0.61, "learning_rate": 6.903908443179998e-06, "loss": 1.9577, "step": 24837 }, { "epoch": 0.61, "learning_rate": 6.903149348791418e-06, "loss": 2.0123, "step": 24838 }, { "epoch": 0.61, "learning_rate": 6.902390274140079e-06, "loss": 1.8622, "step": 24839 }, { "epoch": 0.61, "learning_rate": 6.901631219230813e-06, "loss": 2.0787, "step": 24840 }, { "epoch": 0.61, "learning_rate": 6.900872184068461e-06, "loss": 1.9642, "step": 24841 }, { "epoch": 0.61, "learning_rate": 6.900113168657861e-06, "loss": 2.1213, "step": 24842 }, { "epoch": 0.61, "learning_rate": 6.8993541730038515e-06, "loss": 2.0436, "step": 24843 }, { "epoch": 0.61, "learning_rate": 6.8985951971112685e-06, "loss": 2.0383, "step": 24844 }, { "epoch": 0.61, "learning_rate": 6.897836240984946e-06, "loss": 1.9151, "step": 24845 }, { "epoch": 0.61, "learning_rate": 6.897077304629724e-06, "loss": 1.8476, "step": 24846 }, { "epoch": 0.61, "learning_rate": 6.896318388050445e-06, "loss": 1.879, "step": 24847 }, { "epoch": 0.61, "learning_rate": 6.895559491251935e-06, "loss": 1.9345, "step": 24848 }, { "epoch": 0.61, "learning_rate": 6.8948006142390415e-06, "loss": 1.9512, "step": 24849 }, { "epoch": 0.61, "learning_rate": 6.894041757016589e-06, "loss": 2.1135, "step": 24850 }, { "epoch": 0.61, "learning_rate": 6.893282919589423e-06, "loss": 1.847, "step": 24851 }, { "epoch": 0.61, "learning_rate": 6.892524101962379e-06, "loss": 1.9483, "step": 24852 }, { "epoch": 0.61, "learning_rate": 6.89176530414029e-06, "loss": 1.8848, "step": 24853 }, { "epoch": 0.61, "learning_rate": 6.891006526127997e-06, "loss": 2.1715, "step": 24854 }, { "epoch": 0.61, "learning_rate": 6.890247767930328e-06, "loss": 2.0603, "step": 24855 }, { "epoch": 0.61, "learning_rate": 6.889489029552123e-06, "loss": 1.839, "step": 24856 }, { "epoch": 0.61, "learning_rate": 6.888730310998224e-06, "loss": 1.9395, "step": 24857 }, { "epoch": 0.61, "learning_rate": 6.887971612273456e-06, "loss": 1.9581, "step": 24858 }, { "epoch": 0.61, "learning_rate": 6.887212933382661e-06, "loss": 1.9681, "step": 24859 }, { "epoch": 0.61, "learning_rate": 6.8864542743306715e-06, "loss": 2.0249, "step": 24860 }, { "epoch": 0.61, "learning_rate": 6.885695635122323e-06, "loss": 1.9117, "step": 24861 }, { "epoch": 0.61, "learning_rate": 6.884937015762452e-06, "loss": 1.9482, "step": 24862 }, { "epoch": 0.61, "learning_rate": 6.884178416255891e-06, "loss": 1.9049, "step": 24863 }, { "epoch": 0.61, "learning_rate": 6.8834198366074814e-06, "loss": 2.0551, "step": 24864 }, { "epoch": 0.61, "learning_rate": 6.882661276822047e-06, "loss": 1.995, "step": 24865 }, { "epoch": 0.61, "learning_rate": 6.881902736904429e-06, "loss": 1.8978, "step": 24866 }, { "epoch": 0.61, "learning_rate": 6.881144216859465e-06, "loss": 2.0394, "step": 24867 }, { "epoch": 0.61, "learning_rate": 6.8803857166919815e-06, "loss": 1.9913, "step": 24868 }, { "epoch": 0.61, "learning_rate": 6.879627236406818e-06, "loss": 2.0438, "step": 24869 }, { "epoch": 0.61, "learning_rate": 6.8788687760088055e-06, "loss": 2.0118, "step": 24870 }, { "epoch": 0.61, "learning_rate": 6.878110335502779e-06, "loss": 2.025, "step": 24871 }, { "epoch": 0.61, "learning_rate": 6.877351914893578e-06, "loss": 2.0021, "step": 24872 }, { "epoch": 0.61, "learning_rate": 6.876593514186025e-06, "loss": 2.0111, "step": 24873 }, { "epoch": 0.61, "learning_rate": 6.875835133384965e-06, "loss": 1.9193, "step": 24874 }, { "epoch": 0.61, "learning_rate": 6.875076772495222e-06, "loss": 1.9743, "step": 24875 }, { "epoch": 0.61, "learning_rate": 6.874318431521632e-06, "loss": 2.1023, "step": 24876 }, { "epoch": 0.61, "learning_rate": 6.873560110469031e-06, "loss": 1.8679, "step": 24877 }, { "epoch": 0.61, "learning_rate": 6.872801809342251e-06, "loss": 1.9015, "step": 24878 }, { "epoch": 0.61, "learning_rate": 6.872043528146124e-06, "loss": 1.8731, "step": 24879 }, { "epoch": 0.61, "learning_rate": 6.871285266885481e-06, "loss": 2.0156, "step": 24880 }, { "epoch": 0.61, "learning_rate": 6.870527025565157e-06, "loss": 2.0542, "step": 24881 }, { "epoch": 0.61, "learning_rate": 6.869768804189987e-06, "loss": 1.8827, "step": 24882 }, { "epoch": 0.61, "learning_rate": 6.869010602764797e-06, "loss": 2.0146, "step": 24883 }, { "epoch": 0.61, "learning_rate": 6.868252421294428e-06, "loss": 1.8493, "step": 24884 }, { "epoch": 0.61, "learning_rate": 6.8674942597837015e-06, "loss": 1.8539, "step": 24885 }, { "epoch": 0.61, "learning_rate": 6.866736118237456e-06, "loss": 1.9194, "step": 24886 }, { "epoch": 0.61, "learning_rate": 6.8659779966605224e-06, "loss": 1.9774, "step": 24887 }, { "epoch": 0.61, "learning_rate": 6.865219895057732e-06, "loss": 2.0994, "step": 24888 }, { "epoch": 0.61, "learning_rate": 6.86446181343392e-06, "loss": 1.8951, "step": 24889 }, { "epoch": 0.61, "learning_rate": 6.863703751793912e-06, "loss": 1.802, "step": 24890 }, { "epoch": 0.61, "learning_rate": 6.862945710142541e-06, "loss": 1.9708, "step": 24891 }, { "epoch": 0.61, "learning_rate": 6.862187688484644e-06, "loss": 2.0135, "step": 24892 }, { "epoch": 0.61, "learning_rate": 6.861429686825043e-06, "loss": 1.9882, "step": 24893 }, { "epoch": 0.61, "learning_rate": 6.860671705168575e-06, "loss": 2.0257, "step": 24894 }, { "epoch": 0.61, "learning_rate": 6.859913743520068e-06, "loss": 1.9193, "step": 24895 }, { "epoch": 0.61, "learning_rate": 6.859155801884356e-06, "loss": 2.0008, "step": 24896 }, { "epoch": 0.61, "learning_rate": 6.8583978802662675e-06, "loss": 1.8744, "step": 24897 }, { "epoch": 0.61, "learning_rate": 6.857639978670632e-06, "loss": 1.9696, "step": 24898 }, { "epoch": 0.61, "learning_rate": 6.85688209710228e-06, "loss": 2.0068, "step": 24899 }, { "epoch": 0.61, "learning_rate": 6.856124235566046e-06, "loss": 2.0736, "step": 24900 }, { "epoch": 0.61, "learning_rate": 6.855366394066753e-06, "loss": 1.9902, "step": 24901 }, { "epoch": 0.61, "learning_rate": 6.85460857260924e-06, "loss": 1.871, "step": 24902 }, { "epoch": 0.61, "learning_rate": 6.853850771198327e-06, "loss": 1.9951, "step": 24903 }, { "epoch": 0.61, "learning_rate": 6.8530929898388484e-06, "loss": 1.8998, "step": 24904 }, { "epoch": 0.61, "learning_rate": 6.852335228535636e-06, "loss": 2.0538, "step": 24905 }, { "epoch": 0.61, "learning_rate": 6.851577487293515e-06, "loss": 1.9911, "step": 24906 }, { "epoch": 0.61, "learning_rate": 6.8508197661173205e-06, "loss": 1.9531, "step": 24907 }, { "epoch": 0.61, "learning_rate": 6.8500620650118735e-06, "loss": 1.9859, "step": 24908 }, { "epoch": 0.61, "learning_rate": 6.849304383982007e-06, "loss": 1.9298, "step": 24909 }, { "epoch": 0.61, "learning_rate": 6.8485467230325565e-06, "loss": 1.9833, "step": 24910 }, { "epoch": 0.61, "learning_rate": 6.847789082168339e-06, "loss": 1.9722, "step": 24911 }, { "epoch": 0.61, "learning_rate": 6.847031461394192e-06, "loss": 2.0832, "step": 24912 }, { "epoch": 0.61, "learning_rate": 6.846273860714938e-06, "loss": 2.1895, "step": 24913 }, { "epoch": 0.61, "learning_rate": 6.8455162801354115e-06, "loss": 1.9213, "step": 24914 }, { "epoch": 0.61, "learning_rate": 6.844758719660436e-06, "loss": 2.0672, "step": 24915 }, { "epoch": 0.61, "learning_rate": 6.844001179294841e-06, "loss": 2.2727, "step": 24916 }, { "epoch": 0.61, "learning_rate": 6.84324365904346e-06, "loss": 2.0363, "step": 24917 }, { "epoch": 0.61, "learning_rate": 6.842486158911109e-06, "loss": 1.8756, "step": 24918 }, { "epoch": 0.61, "learning_rate": 6.841728678902624e-06, "loss": 1.9505, "step": 24919 }, { "epoch": 0.61, "learning_rate": 6.840971219022836e-06, "loss": 1.8664, "step": 24920 }, { "epoch": 0.61, "learning_rate": 6.840213779276563e-06, "loss": 1.7638, "step": 24921 }, { "epoch": 0.61, "learning_rate": 6.839456359668641e-06, "loss": 1.8769, "step": 24922 }, { "epoch": 0.61, "learning_rate": 6.83869896020389e-06, "loss": 1.9579, "step": 24923 }, { "epoch": 0.61, "learning_rate": 6.8379415808871415e-06, "loss": 1.9265, "step": 24924 }, { "epoch": 0.61, "learning_rate": 6.837184221723225e-06, "loss": 2.1057, "step": 24925 }, { "epoch": 0.61, "learning_rate": 6.836426882716961e-06, "loss": 1.942, "step": 24926 }, { "epoch": 0.61, "learning_rate": 6.835669563873184e-06, "loss": 2.0284, "step": 24927 }, { "epoch": 0.61, "learning_rate": 6.834912265196712e-06, "loss": 2.0684, "step": 24928 }, { "epoch": 0.61, "learning_rate": 6.834154986692374e-06, "loss": 1.9237, "step": 24929 }, { "epoch": 0.61, "learning_rate": 6.833397728365001e-06, "loss": 2.0555, "step": 24930 }, { "epoch": 0.61, "learning_rate": 6.8326404902194134e-06, "loss": 2.1115, "step": 24931 }, { "epoch": 0.61, "learning_rate": 6.831883272260444e-06, "loss": 1.8279, "step": 24932 }, { "epoch": 0.61, "learning_rate": 6.831126074492913e-06, "loss": 1.8696, "step": 24933 }, { "epoch": 0.61, "learning_rate": 6.830368896921646e-06, "loss": 2.1947, "step": 24934 }, { "epoch": 0.61, "learning_rate": 6.829611739551477e-06, "loss": 1.9667, "step": 24935 }, { "epoch": 0.61, "learning_rate": 6.8288546023872205e-06, "loss": 2.1187, "step": 24936 }, { "epoch": 0.61, "learning_rate": 6.828097485433711e-06, "loss": 2.1109, "step": 24937 }, { "epoch": 0.61, "learning_rate": 6.827340388695767e-06, "loss": 2.0229, "step": 24938 }, { "epoch": 0.61, "learning_rate": 6.826583312178215e-06, "loss": 1.9914, "step": 24939 }, { "epoch": 0.61, "learning_rate": 6.8258262558858855e-06, "loss": 2.1425, "step": 24940 }, { "epoch": 0.61, "learning_rate": 6.825069219823597e-06, "loss": 2.0286, "step": 24941 }, { "epoch": 0.61, "learning_rate": 6.82431220399618e-06, "loss": 2.083, "step": 24942 }, { "epoch": 0.61, "learning_rate": 6.823555208408452e-06, "loss": 2.0622, "step": 24943 }, { "epoch": 0.61, "learning_rate": 6.822798233065244e-06, "loss": 2.02, "step": 24944 }, { "epoch": 0.61, "learning_rate": 6.8220412779713805e-06, "loss": 1.9471, "step": 24945 }, { "epoch": 0.61, "learning_rate": 6.82128434313168e-06, "loss": 1.9661, "step": 24946 }, { "epoch": 0.61, "learning_rate": 6.820527428550974e-06, "loss": 2.0755, "step": 24947 }, { "epoch": 0.61, "learning_rate": 6.819770534234078e-06, "loss": 1.8389, "step": 24948 }, { "epoch": 0.61, "learning_rate": 6.819013660185822e-06, "loss": 2.1288, "step": 24949 }, { "epoch": 0.61, "learning_rate": 6.8182568064110336e-06, "loss": 1.807, "step": 24950 }, { "epoch": 0.62, "learning_rate": 6.817499972914526e-06, "loss": 2.0011, "step": 24951 }, { "epoch": 0.62, "learning_rate": 6.816743159701134e-06, "loss": 2.039, "step": 24952 }, { "epoch": 0.62, "learning_rate": 6.81598636677567e-06, "loss": 1.9894, "step": 24953 }, { "epoch": 0.62, "learning_rate": 6.815229594142962e-06, "loss": 1.9074, "step": 24954 }, { "epoch": 0.62, "learning_rate": 6.814472841807839e-06, "loss": 1.9663, "step": 24955 }, { "epoch": 0.62, "learning_rate": 6.8137161097751135e-06, "loss": 2.0691, "step": 24956 }, { "epoch": 0.62, "learning_rate": 6.812959398049614e-06, "loss": 1.9662, "step": 24957 }, { "epoch": 0.62, "learning_rate": 6.812202706636165e-06, "loss": 1.9968, "step": 24958 }, { "epoch": 0.62, "learning_rate": 6.811446035539585e-06, "loss": 1.8868, "step": 24959 }, { "epoch": 0.62, "learning_rate": 6.8106893847647025e-06, "loss": 1.9124, "step": 24960 }, { "epoch": 0.62, "learning_rate": 6.809932754316333e-06, "loss": 1.8309, "step": 24961 }, { "epoch": 0.62, "learning_rate": 6.8091761441993e-06, "loss": 1.9796, "step": 24962 }, { "epoch": 0.62, "learning_rate": 6.808419554418432e-06, "loss": 2.0308, "step": 24963 }, { "epoch": 0.62, "learning_rate": 6.807662984978543e-06, "loss": 1.9667, "step": 24964 }, { "epoch": 0.62, "learning_rate": 6.806906435884458e-06, "loss": 2.0961, "step": 24965 }, { "epoch": 0.62, "learning_rate": 6.806149907140999e-06, "loss": 2.0778, "step": 24966 }, { "epoch": 0.62, "learning_rate": 6.805393398752989e-06, "loss": 1.9514, "step": 24967 }, { "epoch": 0.62, "learning_rate": 6.804636910725248e-06, "loss": 2.0848, "step": 24968 }, { "epoch": 0.62, "learning_rate": 6.803880443062594e-06, "loss": 1.8736, "step": 24969 }, { "epoch": 0.62, "learning_rate": 6.803123995769857e-06, "loss": 2.0633, "step": 24970 }, { "epoch": 0.62, "learning_rate": 6.802367568851849e-06, "loss": 2.0872, "step": 24971 }, { "epoch": 0.62, "learning_rate": 6.801611162313393e-06, "loss": 1.9744, "step": 24972 }, { "epoch": 0.62, "learning_rate": 6.800854776159316e-06, "loss": 2.094, "step": 24973 }, { "epoch": 0.62, "learning_rate": 6.8000984103944305e-06, "loss": 1.8793, "step": 24974 }, { "epoch": 0.62, "learning_rate": 6.799342065023562e-06, "loss": 2.0443, "step": 24975 }, { "epoch": 0.62, "learning_rate": 6.7985857400515275e-06, "loss": 1.961, "step": 24976 }, { "epoch": 0.62, "learning_rate": 6.79782943548315e-06, "loss": 1.9352, "step": 24977 }, { "epoch": 0.62, "learning_rate": 6.7970731513232526e-06, "loss": 1.9357, "step": 24978 }, { "epoch": 0.62, "learning_rate": 6.796316887576649e-06, "loss": 2.0304, "step": 24979 }, { "epoch": 0.62, "learning_rate": 6.795560644248165e-06, "loss": 2.0193, "step": 24980 }, { "epoch": 0.62, "learning_rate": 6.794804421342613e-06, "loss": 2.0326, "step": 24981 }, { "epoch": 0.62, "learning_rate": 6.794048218864817e-06, "loss": 1.8475, "step": 24982 }, { "epoch": 0.62, "learning_rate": 6.7932920368195986e-06, "loss": 1.9478, "step": 24983 }, { "epoch": 0.62, "learning_rate": 6.792535875211774e-06, "loss": 1.9795, "step": 24984 }, { "epoch": 0.62, "learning_rate": 6.791779734046167e-06, "loss": 1.9967, "step": 24985 }, { "epoch": 0.62, "learning_rate": 6.791023613327589e-06, "loss": 1.8237, "step": 24986 }, { "epoch": 0.62, "learning_rate": 6.790267513060863e-06, "loss": 1.8956, "step": 24987 }, { "epoch": 0.62, "learning_rate": 6.789511433250813e-06, "loss": 1.8864, "step": 24988 }, { "epoch": 0.62, "learning_rate": 6.7887553739022476e-06, "loss": 1.8409, "step": 24989 }, { "epoch": 0.62, "learning_rate": 6.787999335019997e-06, "loss": 1.9168, "step": 24990 }, { "epoch": 0.62, "learning_rate": 6.787243316608867e-06, "loss": 1.9632, "step": 24991 }, { "epoch": 0.62, "learning_rate": 6.7864873186736835e-06, "loss": 1.8191, "step": 24992 }, { "epoch": 0.62, "learning_rate": 6.785731341219265e-06, "loss": 1.932, "step": 24993 }, { "epoch": 0.62, "learning_rate": 6.784975384250428e-06, "loss": 1.8816, "step": 24994 }, { "epoch": 0.62, "learning_rate": 6.784219447771991e-06, "loss": 2.0038, "step": 24995 }, { "epoch": 0.62, "learning_rate": 6.78346353178877e-06, "loss": 2.1928, "step": 24996 }, { "epoch": 0.62, "learning_rate": 6.7827076363055845e-06, "loss": 1.8221, "step": 24997 }, { "epoch": 0.62, "learning_rate": 6.781951761327255e-06, "loss": 2.0492, "step": 24998 }, { "epoch": 0.62, "learning_rate": 6.7811959068585904e-06, "loss": 1.8977, "step": 24999 }, { "epoch": 0.62, "learning_rate": 6.780440072904417e-06, "loss": 1.9988, "step": 25000 }, { "epoch": 0.62, "learning_rate": 6.779684259469545e-06, "loss": 1.9079, "step": 25001 }, { "epoch": 0.62, "learning_rate": 6.778928466558796e-06, "loss": 2.0715, "step": 25002 }, { "epoch": 0.62, "learning_rate": 6.77817269417699e-06, "loss": 1.9819, "step": 25003 }, { "epoch": 0.62, "learning_rate": 6.777416942328934e-06, "loss": 1.9338, "step": 25004 }, { "epoch": 0.62, "learning_rate": 6.776661211019455e-06, "loss": 1.9805, "step": 25005 }, { "epoch": 0.62, "learning_rate": 6.775905500253359e-06, "loss": 2.0678, "step": 25006 }, { "epoch": 0.62, "learning_rate": 6.775149810035469e-06, "loss": 1.9181, "step": 25007 }, { "epoch": 0.62, "learning_rate": 6.774394140370605e-06, "loss": 1.8326, "step": 25008 }, { "epoch": 0.62, "learning_rate": 6.773638491263574e-06, "loss": 1.9991, "step": 25009 }, { "epoch": 0.62, "learning_rate": 6.772882862719197e-06, "loss": 2.0545, "step": 25010 }, { "epoch": 0.62, "learning_rate": 6.772127254742289e-06, "loss": 2.0222, "step": 25011 }, { "epoch": 0.62, "learning_rate": 6.771371667337665e-06, "loss": 1.7663, "step": 25012 }, { "epoch": 0.62, "learning_rate": 6.7706161005101455e-06, "loss": 2.0477, "step": 25013 }, { "epoch": 0.62, "learning_rate": 6.769860554264538e-06, "loss": 1.8934, "step": 25014 }, { "epoch": 0.62, "learning_rate": 6.769105028605662e-06, "loss": 1.8772, "step": 25015 }, { "epoch": 0.62, "learning_rate": 6.768349523538335e-06, "loss": 2.0052, "step": 25016 }, { "epoch": 0.62, "learning_rate": 6.767594039067367e-06, "loss": 2.0927, "step": 25017 }, { "epoch": 0.62, "learning_rate": 6.766838575197578e-06, "loss": 2.0515, "step": 25018 }, { "epoch": 0.62, "learning_rate": 6.766083131933779e-06, "loss": 2.0534, "step": 25019 }, { "epoch": 0.62, "learning_rate": 6.7653277092807845e-06, "loss": 1.9167, "step": 25020 }, { "epoch": 0.62, "learning_rate": 6.764572307243415e-06, "loss": 2.0523, "step": 25021 }, { "epoch": 0.62, "learning_rate": 6.763816925826479e-06, "loss": 1.9847, "step": 25022 }, { "epoch": 0.62, "learning_rate": 6.763061565034794e-06, "loss": 1.9361, "step": 25023 }, { "epoch": 0.62, "learning_rate": 6.762306224873168e-06, "loss": 1.925, "step": 25024 }, { "epoch": 0.62, "learning_rate": 6.761550905346421e-06, "loss": 1.8513, "step": 25025 }, { "epoch": 0.62, "learning_rate": 6.76079560645937e-06, "loss": 2.0115, "step": 25026 }, { "epoch": 0.62, "learning_rate": 6.760040328216818e-06, "loss": 2.0553, "step": 25027 }, { "epoch": 0.62, "learning_rate": 6.759285070623588e-06, "loss": 1.9602, "step": 25028 }, { "epoch": 0.62, "learning_rate": 6.758529833684488e-06, "loss": 1.9865, "step": 25029 }, { "epoch": 0.62, "learning_rate": 6.757774617404334e-06, "loss": 2.1576, "step": 25030 }, { "epoch": 0.62, "learning_rate": 6.757019421787944e-06, "loss": 1.9766, "step": 25031 }, { "epoch": 0.62, "learning_rate": 6.7562642468401204e-06, "loss": 1.9999, "step": 25032 }, { "epoch": 0.62, "learning_rate": 6.755509092565686e-06, "loss": 1.9287, "step": 25033 }, { "epoch": 0.62, "learning_rate": 6.754753958969446e-06, "loss": 2.1611, "step": 25034 }, { "epoch": 0.62, "learning_rate": 6.753998846056218e-06, "loss": 2.0635, "step": 25035 }, { "epoch": 0.62, "learning_rate": 6.753243753830814e-06, "loss": 1.9117, "step": 25036 }, { "epoch": 0.62, "learning_rate": 6.752488682298044e-06, "loss": 1.9688, "step": 25037 }, { "epoch": 0.62, "learning_rate": 6.7517336314627246e-06, "loss": 2.0892, "step": 25038 }, { "epoch": 0.62, "learning_rate": 6.750978601329663e-06, "loss": 1.9295, "step": 25039 }, { "epoch": 0.62, "learning_rate": 6.750223591903673e-06, "loss": 2.068, "step": 25040 }, { "epoch": 0.62, "learning_rate": 6.74946860318957e-06, "loss": 2.0622, "step": 25041 }, { "epoch": 0.62, "learning_rate": 6.748713635192161e-06, "loss": 1.9855, "step": 25042 }, { "epoch": 0.62, "learning_rate": 6.74795868791626e-06, "loss": 1.9343, "step": 25043 }, { "epoch": 0.62, "learning_rate": 6.747203761366679e-06, "loss": 1.8694, "step": 25044 }, { "epoch": 0.62, "learning_rate": 6.746448855548225e-06, "loss": 2.0267, "step": 25045 }, { "epoch": 0.62, "learning_rate": 6.745693970465717e-06, "loss": 1.8869, "step": 25046 }, { "epoch": 0.62, "learning_rate": 6.744939106123958e-06, "loss": 1.9938, "step": 25047 }, { "epoch": 0.62, "learning_rate": 6.744184262527767e-06, "loss": 1.9938, "step": 25048 }, { "epoch": 0.62, "learning_rate": 6.743429439681947e-06, "loss": 2.0941, "step": 25049 }, { "epoch": 0.62, "learning_rate": 6.742674637591313e-06, "loss": 2.0666, "step": 25050 }, { "epoch": 0.62, "learning_rate": 6.741919856260679e-06, "loss": 1.9757, "step": 25051 }, { "epoch": 0.62, "learning_rate": 6.741165095694848e-06, "loss": 2.0237, "step": 25052 }, { "epoch": 0.62, "learning_rate": 6.7404103558986345e-06, "loss": 1.9675, "step": 25053 }, { "epoch": 0.62, "learning_rate": 6.739655636876847e-06, "loss": 1.8873, "step": 25054 }, { "epoch": 0.62, "learning_rate": 6.738900938634296e-06, "loss": 1.975, "step": 25055 }, { "epoch": 0.62, "learning_rate": 6.738146261175797e-06, "loss": 2.0898, "step": 25056 }, { "epoch": 0.62, "learning_rate": 6.73739160450615e-06, "loss": 1.9948, "step": 25057 }, { "epoch": 0.62, "learning_rate": 6.736636968630175e-06, "loss": 2.107, "step": 25058 }, { "epoch": 0.62, "learning_rate": 6.73588235355267e-06, "loss": 2.0639, "step": 25059 }, { "epoch": 0.62, "learning_rate": 6.7351277592784525e-06, "loss": 2.1106, "step": 25060 }, { "epoch": 0.62, "learning_rate": 6.734373185812331e-06, "loss": 1.9011, "step": 25061 }, { "epoch": 0.62, "learning_rate": 6.733618633159114e-06, "loss": 1.8537, "step": 25062 }, { "epoch": 0.62, "learning_rate": 6.7328641013236086e-06, "loss": 2.0748, "step": 25063 }, { "epoch": 0.62, "learning_rate": 6.732109590310623e-06, "loss": 1.9409, "step": 25064 }, { "epoch": 0.62, "learning_rate": 6.7313551001249685e-06, "loss": 1.8569, "step": 25065 }, { "epoch": 0.62, "learning_rate": 6.730600630771458e-06, "loss": 1.9527, "step": 25066 }, { "epoch": 0.62, "learning_rate": 6.7298461822548895e-06, "loss": 1.9345, "step": 25067 }, { "epoch": 0.62, "learning_rate": 6.729091754580081e-06, "loss": 1.9374, "step": 25068 }, { "epoch": 0.62, "learning_rate": 6.7283373477518346e-06, "loss": 1.9176, "step": 25069 }, { "epoch": 0.62, "learning_rate": 6.72758296177496e-06, "loss": 1.9602, "step": 25070 }, { "epoch": 0.62, "learning_rate": 6.7268285966542666e-06, "loss": 1.9031, "step": 25071 }, { "epoch": 0.62, "learning_rate": 6.72607425239456e-06, "loss": 1.9576, "step": 25072 }, { "epoch": 0.62, "learning_rate": 6.725319929000648e-06, "loss": 1.9434, "step": 25073 }, { "epoch": 0.62, "learning_rate": 6.724565626477345e-06, "loss": 1.9217, "step": 25074 }, { "epoch": 0.62, "learning_rate": 6.723811344829447e-06, "loss": 1.8753, "step": 25075 }, { "epoch": 0.62, "learning_rate": 6.723057084061772e-06, "loss": 1.9724, "step": 25076 }, { "epoch": 0.62, "learning_rate": 6.722302844179118e-06, "loss": 1.9469, "step": 25077 }, { "epoch": 0.62, "learning_rate": 6.721548625186296e-06, "loss": 1.9307, "step": 25078 }, { "epoch": 0.62, "learning_rate": 6.720794427088118e-06, "loss": 1.9087, "step": 25079 }, { "epoch": 0.62, "learning_rate": 6.720040249889381e-06, "loss": 2.0734, "step": 25080 }, { "epoch": 0.62, "learning_rate": 6.719286093594899e-06, "loss": 1.9434, "step": 25081 }, { "epoch": 0.62, "learning_rate": 6.718531958209473e-06, "loss": 2.0503, "step": 25082 }, { "epoch": 0.62, "learning_rate": 6.717777843737912e-06, "loss": 1.7643, "step": 25083 }, { "epoch": 0.62, "learning_rate": 6.717023750185027e-06, "loss": 1.856, "step": 25084 }, { "epoch": 0.62, "learning_rate": 6.716269677555616e-06, "loss": 1.8221, "step": 25085 }, { "epoch": 0.62, "learning_rate": 6.715515625854492e-06, "loss": 1.9208, "step": 25086 }, { "epoch": 0.62, "learning_rate": 6.714761595086452e-06, "loss": 1.9828, "step": 25087 }, { "epoch": 0.62, "learning_rate": 6.714007585256308e-06, "loss": 1.9982, "step": 25088 }, { "epoch": 0.62, "learning_rate": 6.713253596368867e-06, "loss": 1.9819, "step": 25089 }, { "epoch": 0.62, "learning_rate": 6.71249962842893e-06, "loss": 2.0281, "step": 25090 }, { "epoch": 0.62, "learning_rate": 6.711745681441308e-06, "loss": 1.9993, "step": 25091 }, { "epoch": 0.62, "learning_rate": 6.710991755410799e-06, "loss": 1.8919, "step": 25092 }, { "epoch": 0.62, "learning_rate": 6.710237850342209e-06, "loss": 1.8884, "step": 25093 }, { "epoch": 0.62, "learning_rate": 6.709483966240351e-06, "loss": 1.832, "step": 25094 }, { "epoch": 0.62, "learning_rate": 6.7087301031100205e-06, "loss": 2.0892, "step": 25095 }, { "epoch": 0.62, "learning_rate": 6.707976260956027e-06, "loss": 1.9818, "step": 25096 }, { "epoch": 0.62, "learning_rate": 6.707222439783174e-06, "loss": 2.0846, "step": 25097 }, { "epoch": 0.62, "learning_rate": 6.706468639596263e-06, "loss": 2.051, "step": 25098 }, { "epoch": 0.62, "learning_rate": 6.705714860400103e-06, "loss": 2.077, "step": 25099 }, { "epoch": 0.62, "learning_rate": 6.704961102199494e-06, "loss": 1.9551, "step": 25100 }, { "epoch": 0.62, "learning_rate": 6.704207364999245e-06, "loss": 1.9906, "step": 25101 }, { "epoch": 0.62, "learning_rate": 6.703453648804153e-06, "loss": 1.8506, "step": 25102 }, { "epoch": 0.62, "learning_rate": 6.702699953619025e-06, "loss": 1.8822, "step": 25103 }, { "epoch": 0.62, "learning_rate": 6.70194627944867e-06, "loss": 1.9716, "step": 25104 }, { "epoch": 0.62, "learning_rate": 6.701192626297881e-06, "loss": 1.8458, "step": 25105 }, { "epoch": 0.62, "learning_rate": 6.7004389941714675e-06, "loss": 2.1636, "step": 25106 }, { "epoch": 0.62, "learning_rate": 6.69968538307423e-06, "loss": 1.9769, "step": 25107 }, { "epoch": 0.62, "learning_rate": 6.698931793010974e-06, "loss": 1.956, "step": 25108 }, { "epoch": 0.62, "learning_rate": 6.698178223986505e-06, "loss": 1.9836, "step": 25109 }, { "epoch": 0.62, "learning_rate": 6.69742467600562e-06, "loss": 2.0454, "step": 25110 }, { "epoch": 0.62, "learning_rate": 6.696671149073126e-06, "loss": 1.9972, "step": 25111 }, { "epoch": 0.62, "learning_rate": 6.695917643193819e-06, "loss": 2.1571, "step": 25112 }, { "epoch": 0.62, "learning_rate": 6.695164158372506e-06, "loss": 2.0865, "step": 25113 }, { "epoch": 0.62, "learning_rate": 6.694410694613991e-06, "loss": 1.9047, "step": 25114 }, { "epoch": 0.62, "learning_rate": 6.693657251923074e-06, "loss": 2.0063, "step": 25115 }, { "epoch": 0.62, "learning_rate": 6.692903830304557e-06, "loss": 2.0661, "step": 25116 }, { "epoch": 0.62, "learning_rate": 6.692150429763239e-06, "loss": 1.9668, "step": 25117 }, { "epoch": 0.62, "learning_rate": 6.691397050303925e-06, "loss": 1.757, "step": 25118 }, { "epoch": 0.62, "learning_rate": 6.690643691931419e-06, "loss": 1.9259, "step": 25119 }, { "epoch": 0.62, "learning_rate": 6.689890354650516e-06, "loss": 2.1342, "step": 25120 }, { "epoch": 0.62, "learning_rate": 6.689137038466025e-06, "loss": 1.9, "step": 25121 }, { "epoch": 0.62, "learning_rate": 6.688383743382737e-06, "loss": 2.0371, "step": 25122 }, { "epoch": 0.62, "learning_rate": 6.68763046940546e-06, "loss": 1.7522, "step": 25123 }, { "epoch": 0.62, "learning_rate": 6.686877216538994e-06, "loss": 1.8668, "step": 25124 }, { "epoch": 0.62, "learning_rate": 6.686123984788138e-06, "loss": 2.1265, "step": 25125 }, { "epoch": 0.62, "learning_rate": 6.685370774157698e-06, "loss": 1.8466, "step": 25126 }, { "epoch": 0.62, "learning_rate": 6.684617584652467e-06, "loss": 2.0802, "step": 25127 }, { "epoch": 0.62, "learning_rate": 6.6838644162772485e-06, "loss": 1.9699, "step": 25128 }, { "epoch": 0.62, "learning_rate": 6.683111269036846e-06, "loss": 1.9824, "step": 25129 }, { "epoch": 0.62, "learning_rate": 6.682358142936052e-06, "loss": 1.9202, "step": 25130 }, { "epoch": 0.62, "learning_rate": 6.6816050379796724e-06, "loss": 1.9086, "step": 25131 }, { "epoch": 0.62, "learning_rate": 6.680851954172505e-06, "loss": 1.9075, "step": 25132 }, { "epoch": 0.62, "learning_rate": 6.680098891519351e-06, "loss": 1.9845, "step": 25133 }, { "epoch": 0.62, "learning_rate": 6.679345850025008e-06, "loss": 1.9733, "step": 25134 }, { "epoch": 0.62, "learning_rate": 6.678592829694275e-06, "loss": 1.9817, "step": 25135 }, { "epoch": 0.62, "learning_rate": 6.677839830531952e-06, "loss": 2.051, "step": 25136 }, { "epoch": 0.62, "learning_rate": 6.677086852542842e-06, "loss": 2.047, "step": 25137 }, { "epoch": 0.62, "learning_rate": 6.676333895731737e-06, "loss": 2.0637, "step": 25138 }, { "epoch": 0.62, "learning_rate": 6.675580960103443e-06, "loss": 2.0749, "step": 25139 }, { "epoch": 0.62, "learning_rate": 6.674828045662752e-06, "loss": 2.0314, "step": 25140 }, { "epoch": 0.62, "learning_rate": 6.674075152414463e-06, "loss": 1.9046, "step": 25141 }, { "epoch": 0.62, "learning_rate": 6.673322280363381e-06, "loss": 2.028, "step": 25142 }, { "epoch": 0.62, "learning_rate": 6.6725694295142975e-06, "loss": 1.9774, "step": 25143 }, { "epoch": 0.62, "learning_rate": 6.671816599872018e-06, "loss": 1.7861, "step": 25144 }, { "epoch": 0.62, "learning_rate": 6.6710637914413304e-06, "loss": 2.0283, "step": 25145 }, { "epoch": 0.62, "learning_rate": 6.6703110042270396e-06, "loss": 1.9805, "step": 25146 }, { "epoch": 0.62, "learning_rate": 6.669558238233946e-06, "loss": 2.1197, "step": 25147 }, { "epoch": 0.62, "learning_rate": 6.668805493466838e-06, "loss": 1.8165, "step": 25148 }, { "epoch": 0.62, "learning_rate": 6.6680527699305206e-06, "loss": 1.9379, "step": 25149 }, { "epoch": 0.62, "learning_rate": 6.667300067629787e-06, "loss": 1.7944, "step": 25150 }, { "epoch": 0.62, "learning_rate": 6.666547386569438e-06, "loss": 1.9621, "step": 25151 }, { "epoch": 0.62, "learning_rate": 6.665794726754269e-06, "loss": 1.9838, "step": 25152 }, { "epoch": 0.62, "learning_rate": 6.6650420881890755e-06, "loss": 2.049, "step": 25153 }, { "epoch": 0.62, "learning_rate": 6.664289470878658e-06, "loss": 1.8633, "step": 25154 }, { "epoch": 0.62, "learning_rate": 6.663536874827807e-06, "loss": 1.9782, "step": 25155 }, { "epoch": 0.62, "learning_rate": 6.662784300041325e-06, "loss": 2.056, "step": 25156 }, { "epoch": 0.62, "learning_rate": 6.662031746524008e-06, "loss": 1.9174, "step": 25157 }, { "epoch": 0.62, "learning_rate": 6.661279214280647e-06, "loss": 1.9539, "step": 25158 }, { "epoch": 0.62, "learning_rate": 6.660526703316044e-06, "loss": 1.7904, "step": 25159 }, { "epoch": 0.62, "learning_rate": 6.659774213634991e-06, "loss": 2.0019, "step": 25160 }, { "epoch": 0.62, "learning_rate": 6.659021745242284e-06, "loss": 1.9289, "step": 25161 }, { "epoch": 0.62, "learning_rate": 6.6582692981427255e-06, "loss": 1.9909, "step": 25162 }, { "epoch": 0.62, "learning_rate": 6.657516872341101e-06, "loss": 1.9482, "step": 25163 }, { "epoch": 0.62, "learning_rate": 6.656764467842216e-06, "loss": 1.8975, "step": 25164 }, { "epoch": 0.62, "learning_rate": 6.6560120846508556e-06, "loss": 1.9293, "step": 25165 }, { "epoch": 0.62, "learning_rate": 6.65525972277182e-06, "loss": 1.9016, "step": 25166 }, { "epoch": 0.62, "learning_rate": 6.654507382209906e-06, "loss": 1.9151, "step": 25167 }, { "epoch": 0.62, "learning_rate": 6.6537550629699065e-06, "loss": 2.249, "step": 25168 }, { "epoch": 0.62, "learning_rate": 6.653002765056618e-06, "loss": 2.0899, "step": 25169 }, { "epoch": 0.62, "learning_rate": 6.6522504884748295e-06, "loss": 2.25, "step": 25170 }, { "epoch": 0.62, "learning_rate": 6.6514982332293405e-06, "loss": 1.9927, "step": 25171 }, { "epoch": 0.62, "learning_rate": 6.65074599932495e-06, "loss": 1.9358, "step": 25172 }, { "epoch": 0.62, "learning_rate": 6.649993786766441e-06, "loss": 1.9857, "step": 25173 }, { "epoch": 0.62, "learning_rate": 6.649241595558617e-06, "loss": 1.9196, "step": 25174 }, { "epoch": 0.62, "learning_rate": 6.648489425706266e-06, "loss": 1.9551, "step": 25175 }, { "epoch": 0.62, "learning_rate": 6.647737277214184e-06, "loss": 2.106, "step": 25176 }, { "epoch": 0.62, "learning_rate": 6.646985150087166e-06, "loss": 1.956, "step": 25177 }, { "epoch": 0.62, "learning_rate": 6.6462330443300015e-06, "loss": 1.8466, "step": 25178 }, { "epoch": 0.62, "learning_rate": 6.645480959947492e-06, "loss": 1.9853, "step": 25179 }, { "epoch": 0.62, "learning_rate": 6.644728896944421e-06, "loss": 1.9296, "step": 25180 }, { "epoch": 0.62, "learning_rate": 6.643976855325587e-06, "loss": 2.1681, "step": 25181 }, { "epoch": 0.62, "learning_rate": 6.643224835095785e-06, "loss": 1.9212, "step": 25182 }, { "epoch": 0.62, "learning_rate": 6.6424728362598024e-06, "loss": 1.8772, "step": 25183 }, { "epoch": 0.62, "learning_rate": 6.6417208588224355e-06, "loss": 2.059, "step": 25184 }, { "epoch": 0.62, "learning_rate": 6.640968902788474e-06, "loss": 1.9755, "step": 25185 }, { "epoch": 0.62, "learning_rate": 6.640216968162716e-06, "loss": 2.1255, "step": 25186 }, { "epoch": 0.62, "learning_rate": 6.6394650549499485e-06, "loss": 2.0735, "step": 25187 }, { "epoch": 0.62, "learning_rate": 6.638713163154964e-06, "loss": 2.167, "step": 25188 }, { "epoch": 0.62, "learning_rate": 6.63796129278256e-06, "loss": 1.9187, "step": 25189 }, { "epoch": 0.62, "learning_rate": 6.63720944383752e-06, "loss": 2.1091, "step": 25190 }, { "epoch": 0.62, "learning_rate": 6.63645761632464e-06, "loss": 1.908, "step": 25191 }, { "epoch": 0.62, "learning_rate": 6.635705810248716e-06, "loss": 1.9868, "step": 25192 }, { "epoch": 0.62, "learning_rate": 6.6349540256145304e-06, "loss": 1.8328, "step": 25193 }, { "epoch": 0.62, "learning_rate": 6.634202262426881e-06, "loss": 2.0063, "step": 25194 }, { "epoch": 0.62, "learning_rate": 6.633450520690558e-06, "loss": 1.8607, "step": 25195 }, { "epoch": 0.62, "learning_rate": 6.632698800410351e-06, "loss": 2.0063, "step": 25196 }, { "epoch": 0.62, "learning_rate": 6.631947101591055e-06, "loss": 1.8745, "step": 25197 }, { "epoch": 0.62, "learning_rate": 6.631195424237455e-06, "loss": 2.065, "step": 25198 }, { "epoch": 0.62, "learning_rate": 6.630443768354343e-06, "loss": 2.0511, "step": 25199 }, { "epoch": 0.62, "learning_rate": 6.629692133946517e-06, "loss": 1.9616, "step": 25200 }, { "epoch": 0.62, "learning_rate": 6.628940521018755e-06, "loss": 2.0084, "step": 25201 }, { "epoch": 0.62, "learning_rate": 6.628188929575857e-06, "loss": 2.0243, "step": 25202 }, { "epoch": 0.62, "learning_rate": 6.627437359622608e-06, "loss": 1.9517, "step": 25203 }, { "epoch": 0.62, "learning_rate": 6.626685811163801e-06, "loss": 2.1313, "step": 25204 }, { "epoch": 0.62, "learning_rate": 6.625934284204225e-06, "loss": 1.8975, "step": 25205 }, { "epoch": 0.62, "learning_rate": 6.625182778748667e-06, "loss": 2.117, "step": 25206 }, { "epoch": 0.62, "learning_rate": 6.624431294801923e-06, "loss": 2.1247, "step": 25207 }, { "epoch": 0.62, "learning_rate": 6.623679832368776e-06, "loss": 1.9458, "step": 25208 }, { "epoch": 0.62, "learning_rate": 6.6229283914540155e-06, "loss": 2.0178, "step": 25209 }, { "epoch": 0.62, "learning_rate": 6.622176972062439e-06, "loss": 1.9994, "step": 25210 }, { "epoch": 0.62, "learning_rate": 6.621425574198823e-06, "loss": 1.9509, "step": 25211 }, { "epoch": 0.62, "learning_rate": 6.620674197867967e-06, "loss": 2.0011, "step": 25212 }, { "epoch": 0.62, "learning_rate": 6.619922843074653e-06, "loss": 1.9073, "step": 25213 }, { "epoch": 0.62, "learning_rate": 6.619171509823672e-06, "loss": 2.2669, "step": 25214 }, { "epoch": 0.62, "learning_rate": 6.618420198119817e-06, "loss": 1.7518, "step": 25215 }, { "epoch": 0.62, "learning_rate": 6.617668907967868e-06, "loss": 1.9439, "step": 25216 }, { "epoch": 0.62, "learning_rate": 6.616917639372622e-06, "loss": 2.0107, "step": 25217 }, { "epoch": 0.62, "learning_rate": 6.616166392338857e-06, "loss": 1.9648, "step": 25218 }, { "epoch": 0.62, "learning_rate": 6.615415166871366e-06, "loss": 1.9124, "step": 25219 }, { "epoch": 0.62, "learning_rate": 6.614663962974941e-06, "loss": 1.8528, "step": 25220 }, { "epoch": 0.62, "learning_rate": 6.613912780654363e-06, "loss": 1.8491, "step": 25221 }, { "epoch": 0.62, "learning_rate": 6.6131616199144266e-06, "loss": 2.0521, "step": 25222 }, { "epoch": 0.62, "learning_rate": 6.61241048075991e-06, "loss": 2.0733, "step": 25223 }, { "epoch": 0.62, "learning_rate": 6.611659363195606e-06, "loss": 2.1664, "step": 25224 }, { "epoch": 0.62, "learning_rate": 6.610908267226306e-06, "loss": 2.0416, "step": 25225 }, { "epoch": 0.62, "learning_rate": 6.610157192856788e-06, "loss": 1.8124, "step": 25226 }, { "epoch": 0.62, "learning_rate": 6.609406140091845e-06, "loss": 1.9598, "step": 25227 }, { "epoch": 0.62, "learning_rate": 6.6086551089362595e-06, "loss": 1.9755, "step": 25228 }, { "epoch": 0.62, "learning_rate": 6.6079040993948194e-06, "loss": 1.925, "step": 25229 }, { "epoch": 0.62, "learning_rate": 6.607153111472313e-06, "loss": 1.947, "step": 25230 }, { "epoch": 0.62, "learning_rate": 6.606402145173525e-06, "loss": 1.8258, "step": 25231 }, { "epoch": 0.62, "learning_rate": 6.605651200503246e-06, "loss": 2.16, "step": 25232 }, { "epoch": 0.62, "learning_rate": 6.604900277466253e-06, "loss": 2.0723, "step": 25233 }, { "epoch": 0.62, "learning_rate": 6.604149376067338e-06, "loss": 1.9735, "step": 25234 }, { "epoch": 0.62, "learning_rate": 6.6033984963112895e-06, "loss": 1.99, "step": 25235 }, { "epoch": 0.62, "learning_rate": 6.602647638202886e-06, "loss": 1.9538, "step": 25236 }, { "epoch": 0.62, "learning_rate": 6.601896801746916e-06, "loss": 1.9967, "step": 25237 }, { "epoch": 0.62, "learning_rate": 6.601145986948165e-06, "loss": 2.1892, "step": 25238 }, { "epoch": 0.62, "learning_rate": 6.600395193811417e-06, "loss": 2.0562, "step": 25239 }, { "epoch": 0.62, "learning_rate": 6.599644422341464e-06, "loss": 1.926, "step": 25240 }, { "epoch": 0.62, "learning_rate": 6.598893672543081e-06, "loss": 1.8873, "step": 25241 }, { "epoch": 0.62, "learning_rate": 6.5981429444210596e-06, "loss": 2.0501, "step": 25242 }, { "epoch": 0.62, "learning_rate": 6.597392237980179e-06, "loss": 2.0125, "step": 25243 }, { "epoch": 0.62, "learning_rate": 6.596641553225227e-06, "loss": 1.9832, "step": 25244 }, { "epoch": 0.62, "learning_rate": 6.59589089016099e-06, "loss": 2.1119, "step": 25245 }, { "epoch": 0.62, "learning_rate": 6.595140248792247e-06, "loss": 2.0313, "step": 25246 }, { "epoch": 0.62, "learning_rate": 6.594389629123785e-06, "loss": 1.9798, "step": 25247 }, { "epoch": 0.62, "learning_rate": 6.593639031160388e-06, "loss": 2.0963, "step": 25248 }, { "epoch": 0.62, "learning_rate": 6.592888454906838e-06, "loss": 1.9103, "step": 25249 }, { "epoch": 0.62, "learning_rate": 6.5921379003679255e-06, "loss": 1.9553, "step": 25250 }, { "epoch": 0.62, "learning_rate": 6.591387367548423e-06, "loss": 1.9398, "step": 25251 }, { "epoch": 0.62, "learning_rate": 6.590636856453121e-06, "loss": 1.9714, "step": 25252 }, { "epoch": 0.62, "learning_rate": 6.589886367086805e-06, "loss": 1.9621, "step": 25253 }, { "epoch": 0.62, "learning_rate": 6.589135899454251e-06, "loss": 2.0726, "step": 25254 }, { "epoch": 0.62, "learning_rate": 6.588385453560246e-06, "loss": 2.0067, "step": 25255 }, { "epoch": 0.62, "learning_rate": 6.587635029409573e-06, "loss": 1.9703, "step": 25256 }, { "epoch": 0.62, "learning_rate": 6.586884627007013e-06, "loss": 2.0314, "step": 25257 }, { "epoch": 0.62, "learning_rate": 6.586134246357352e-06, "loss": 2.0489, "step": 25258 }, { "epoch": 0.62, "learning_rate": 6.585383887465366e-06, "loss": 1.9969, "step": 25259 }, { "epoch": 0.62, "learning_rate": 6.584633550335846e-06, "loss": 1.9348, "step": 25260 }, { "epoch": 0.62, "learning_rate": 6.583883234973567e-06, "loss": 1.7373, "step": 25261 }, { "epoch": 0.62, "learning_rate": 6.5831329413833126e-06, "loss": 2.136, "step": 25262 }, { "epoch": 0.62, "learning_rate": 6.582382669569868e-06, "loss": 1.8561, "step": 25263 }, { "epoch": 0.62, "learning_rate": 6.581632419538009e-06, "loss": 2.2182, "step": 25264 }, { "epoch": 0.62, "learning_rate": 6.580882191292524e-06, "loss": 1.9259, "step": 25265 }, { "epoch": 0.62, "learning_rate": 6.580131984838187e-06, "loss": 2.1159, "step": 25266 }, { "epoch": 0.62, "learning_rate": 6.5793818001797845e-06, "loss": 2.0482, "step": 25267 }, { "epoch": 0.62, "learning_rate": 6.578631637322101e-06, "loss": 1.9806, "step": 25268 }, { "epoch": 0.62, "learning_rate": 6.577881496269908e-06, "loss": 2.1536, "step": 25269 }, { "epoch": 0.62, "learning_rate": 6.577131377027997e-06, "loss": 1.9985, "step": 25270 }, { "epoch": 0.62, "learning_rate": 6.576381279601137e-06, "loss": 1.9811, "step": 25271 }, { "epoch": 0.62, "learning_rate": 6.575631203994116e-06, "loss": 1.959, "step": 25272 }, { "epoch": 0.62, "learning_rate": 6.574881150211716e-06, "loss": 1.9911, "step": 25273 }, { "epoch": 0.62, "learning_rate": 6.574131118258711e-06, "loss": 1.9621, "step": 25274 }, { "epoch": 0.62, "learning_rate": 6.57338110813989e-06, "loss": 1.9985, "step": 25275 }, { "epoch": 0.62, "learning_rate": 6.572631119860024e-06, "loss": 2.0675, "step": 25276 }, { "epoch": 0.62, "learning_rate": 6.571881153423895e-06, "loss": 1.9939, "step": 25277 }, { "epoch": 0.62, "learning_rate": 6.571131208836292e-06, "loss": 2.0552, "step": 25278 }, { "epoch": 0.62, "learning_rate": 6.570381286101981e-06, "loss": 2.1883, "step": 25279 }, { "epoch": 0.62, "learning_rate": 6.5696313852257525e-06, "loss": 2.0155, "step": 25280 }, { "epoch": 0.62, "learning_rate": 6.568881506212377e-06, "loss": 1.9089, "step": 25281 }, { "epoch": 0.62, "learning_rate": 6.568131649066637e-06, "loss": 1.9727, "step": 25282 }, { "epoch": 0.62, "learning_rate": 6.567381813793314e-06, "loss": 1.8776, "step": 25283 }, { "epoch": 0.62, "learning_rate": 6.566632000397185e-06, "loss": 2.0667, "step": 25284 }, { "epoch": 0.62, "learning_rate": 6.565882208883032e-06, "loss": 1.9797, "step": 25285 }, { "epoch": 0.62, "learning_rate": 6.565132439255628e-06, "loss": 1.8821, "step": 25286 }, { "epoch": 0.62, "learning_rate": 6.564382691519753e-06, "loss": 2.0626, "step": 25287 }, { "epoch": 0.62, "learning_rate": 6.563632965680191e-06, "loss": 1.8611, "step": 25288 }, { "epoch": 0.62, "learning_rate": 6.5628832617417125e-06, "loss": 2.0954, "step": 25289 }, { "epoch": 0.62, "learning_rate": 6.562133579709102e-06, "loss": 1.9129, "step": 25290 }, { "epoch": 0.62, "learning_rate": 6.561383919587131e-06, "loss": 2.0183, "step": 25291 }, { "epoch": 0.62, "learning_rate": 6.5606342813805826e-06, "loss": 1.9574, "step": 25292 }, { "epoch": 0.62, "learning_rate": 6.559884665094235e-06, "loss": 1.9637, "step": 25293 }, { "epoch": 0.62, "learning_rate": 6.5591350707328605e-06, "loss": 1.9199, "step": 25294 }, { "epoch": 0.62, "learning_rate": 6.558385498301243e-06, "loss": 1.913, "step": 25295 }, { "epoch": 0.62, "learning_rate": 6.5576359478041525e-06, "loss": 1.9474, "step": 25296 }, { "epoch": 0.62, "learning_rate": 6.556886419246369e-06, "loss": 2.1671, "step": 25297 }, { "epoch": 0.62, "learning_rate": 6.556136912632676e-06, "loss": 1.8417, "step": 25298 }, { "epoch": 0.62, "learning_rate": 6.555387427967841e-06, "loss": 1.9708, "step": 25299 }, { "epoch": 0.62, "learning_rate": 6.554637965256645e-06, "loss": 2.1274, "step": 25300 }, { "epoch": 0.62, "learning_rate": 6.553888524503862e-06, "loss": 1.9287, "step": 25301 }, { "epoch": 0.62, "learning_rate": 6.553139105714271e-06, "loss": 1.9068, "step": 25302 }, { "epoch": 0.62, "learning_rate": 6.552389708892652e-06, "loss": 2.1628, "step": 25303 }, { "epoch": 0.62, "learning_rate": 6.551640334043774e-06, "loss": 1.9419, "step": 25304 }, { "epoch": 0.62, "learning_rate": 6.550890981172418e-06, "loss": 2.0793, "step": 25305 }, { "epoch": 0.62, "learning_rate": 6.550141650283355e-06, "loss": 1.9378, "step": 25306 }, { "epoch": 0.62, "learning_rate": 6.5493923413813634e-06, "loss": 1.9417, "step": 25307 }, { "epoch": 0.62, "learning_rate": 6.54864305447122e-06, "loss": 2.0982, "step": 25308 }, { "epoch": 0.62, "learning_rate": 6.547893789557697e-06, "loss": 1.7725, "step": 25309 }, { "epoch": 0.62, "learning_rate": 6.547144546645574e-06, "loss": 1.997, "step": 25310 }, { "epoch": 0.62, "learning_rate": 6.546395325739627e-06, "loss": 1.967, "step": 25311 }, { "epoch": 0.62, "learning_rate": 6.5456461268446234e-06, "loss": 2.1268, "step": 25312 }, { "epoch": 0.62, "learning_rate": 6.544896949965348e-06, "loss": 1.8473, "step": 25313 }, { "epoch": 0.62, "learning_rate": 6.544147795106566e-06, "loss": 2.0295, "step": 25314 }, { "epoch": 0.62, "learning_rate": 6.543398662273055e-06, "loss": 1.9756, "step": 25315 }, { "epoch": 0.62, "learning_rate": 6.542649551469596e-06, "loss": 1.9261, "step": 25316 }, { "epoch": 0.62, "learning_rate": 6.541900462700954e-06, "loss": 1.9543, "step": 25317 }, { "epoch": 0.62, "learning_rate": 6.5411513959719094e-06, "loss": 1.8736, "step": 25318 }, { "epoch": 0.62, "learning_rate": 6.540402351287231e-06, "loss": 2.0888, "step": 25319 }, { "epoch": 0.62, "learning_rate": 6.539653328651697e-06, "loss": 1.8396, "step": 25320 }, { "epoch": 0.62, "learning_rate": 6.538904328070083e-06, "loss": 1.8075, "step": 25321 }, { "epoch": 0.62, "learning_rate": 6.538155349547157e-06, "loss": 2.0099, "step": 25322 }, { "epoch": 0.62, "learning_rate": 6.537406393087699e-06, "loss": 2.0109, "step": 25323 }, { "epoch": 0.62, "learning_rate": 6.536657458696475e-06, "loss": 2.0713, "step": 25324 }, { "epoch": 0.62, "learning_rate": 6.53590854637826e-06, "loss": 1.9145, "step": 25325 }, { "epoch": 0.62, "learning_rate": 6.5351596561378325e-06, "loss": 1.9105, "step": 25326 }, { "epoch": 0.62, "learning_rate": 6.534410787979959e-06, "loss": 1.8985, "step": 25327 }, { "epoch": 0.62, "learning_rate": 6.5336619419094175e-06, "loss": 1.8759, "step": 25328 }, { "epoch": 0.62, "learning_rate": 6.532913117930975e-06, "loss": 1.9073, "step": 25329 }, { "epoch": 0.62, "learning_rate": 6.532164316049407e-06, "loss": 2.0411, "step": 25330 }, { "epoch": 0.62, "learning_rate": 6.531415536269492e-06, "loss": 1.9343, "step": 25331 }, { "epoch": 0.62, "learning_rate": 6.53066677859599e-06, "loss": 1.9655, "step": 25332 }, { "epoch": 0.62, "learning_rate": 6.529918043033682e-06, "loss": 1.6918, "step": 25333 }, { "epoch": 0.62, "learning_rate": 6.529169329587337e-06, "loss": 1.933, "step": 25334 }, { "epoch": 0.62, "learning_rate": 6.5284206382617244e-06, "loss": 2.117, "step": 25335 }, { "epoch": 0.62, "learning_rate": 6.5276719690616204e-06, "loss": 1.9945, "step": 25336 }, { "epoch": 0.62, "learning_rate": 6.526923321991794e-06, "loss": 1.8986, "step": 25337 }, { "epoch": 0.62, "learning_rate": 6.526174697057019e-06, "loss": 1.9058, "step": 25338 }, { "epoch": 0.62, "learning_rate": 6.525426094262063e-06, "loss": 2.0137, "step": 25339 }, { "epoch": 0.62, "learning_rate": 6.524677513611697e-06, "loss": 2.0051, "step": 25340 }, { "epoch": 0.62, "learning_rate": 6.523928955110699e-06, "loss": 2.0614, "step": 25341 }, { "epoch": 0.62, "learning_rate": 6.52318041876383e-06, "loss": 1.9625, "step": 25342 }, { "epoch": 0.62, "learning_rate": 6.522431904575867e-06, "loss": 1.9451, "step": 25343 }, { "epoch": 0.62, "learning_rate": 6.521683412551578e-06, "loss": 2.0816, "step": 25344 }, { "epoch": 0.62, "learning_rate": 6.520934942695732e-06, "loss": 1.9997, "step": 25345 }, { "epoch": 0.62, "learning_rate": 6.5201864950131075e-06, "loss": 2.0256, "step": 25346 }, { "epoch": 0.62, "learning_rate": 6.5194380695084634e-06, "loss": 2.0323, "step": 25347 }, { "epoch": 0.62, "learning_rate": 6.51868966618658e-06, "loss": 1.9091, "step": 25348 }, { "epoch": 0.62, "learning_rate": 6.517941285052216e-06, "loss": 1.9418, "step": 25349 }, { "epoch": 0.62, "learning_rate": 6.517192926110147e-06, "loss": 1.9234, "step": 25350 }, { "epoch": 0.62, "learning_rate": 6.516444589365147e-06, "loss": 1.8635, "step": 25351 }, { "epoch": 0.62, "learning_rate": 6.5156962748219785e-06, "loss": 1.8945, "step": 25352 }, { "epoch": 0.62, "learning_rate": 6.514947982485414e-06, "loss": 2.0425, "step": 25353 }, { "epoch": 0.62, "learning_rate": 6.514199712360221e-06, "loss": 1.8531, "step": 25354 }, { "epoch": 0.62, "learning_rate": 6.513451464451167e-06, "loss": 1.8722, "step": 25355 }, { "epoch": 0.62, "learning_rate": 6.512703238763027e-06, "loss": 1.9323, "step": 25356 }, { "epoch": 0.63, "learning_rate": 6.511955035300563e-06, "loss": 1.8562, "step": 25357 }, { "epoch": 0.63, "learning_rate": 6.5112068540685505e-06, "loss": 1.9121, "step": 25358 }, { "epoch": 0.63, "learning_rate": 6.510458695071748e-06, "loss": 1.9881, "step": 25359 }, { "epoch": 0.63, "learning_rate": 6.509710558314932e-06, "loss": 1.9796, "step": 25360 }, { "epoch": 0.63, "learning_rate": 6.508962443802869e-06, "loss": 1.9912, "step": 25361 }, { "epoch": 0.63, "learning_rate": 6.508214351540323e-06, "loss": 1.969, "step": 25362 }, { "epoch": 0.63, "learning_rate": 6.50746628153207e-06, "loss": 2.0413, "step": 25363 }, { "epoch": 0.63, "learning_rate": 6.506718233782868e-06, "loss": 2.2139, "step": 25364 }, { "epoch": 0.63, "learning_rate": 6.505970208297489e-06, "loss": 2.005, "step": 25365 }, { "epoch": 0.63, "learning_rate": 6.505222205080705e-06, "loss": 1.9487, "step": 25366 }, { "epoch": 0.63, "learning_rate": 6.504474224137275e-06, "loss": 2.024, "step": 25367 }, { "epoch": 0.63, "learning_rate": 6.503726265471971e-06, "loss": 1.9144, "step": 25368 }, { "epoch": 0.63, "learning_rate": 6.50297832908956e-06, "loss": 1.8981, "step": 25369 }, { "epoch": 0.63, "learning_rate": 6.502230414994806e-06, "loss": 1.7938, "step": 25370 }, { "epoch": 0.63, "learning_rate": 6.501482523192477e-06, "loss": 1.8478, "step": 25371 }, { "epoch": 0.63, "learning_rate": 6.500734653687341e-06, "loss": 2.1093, "step": 25372 }, { "epoch": 0.63, "learning_rate": 6.499986806484162e-06, "loss": 2.0446, "step": 25373 }, { "epoch": 0.63, "learning_rate": 6.499238981587712e-06, "loss": 2.0852, "step": 25374 }, { "epoch": 0.63, "learning_rate": 6.498491179002749e-06, "loss": 1.8423, "step": 25375 }, { "epoch": 0.63, "learning_rate": 6.497743398734047e-06, "loss": 1.9265, "step": 25376 }, { "epoch": 0.63, "learning_rate": 6.496995640786362e-06, "loss": 1.9592, "step": 25377 }, { "epoch": 0.63, "learning_rate": 6.496247905164465e-06, "loss": 2.121, "step": 25378 }, { "epoch": 0.63, "learning_rate": 6.495500191873128e-06, "loss": 2.0214, "step": 25379 }, { "epoch": 0.63, "learning_rate": 6.4947525009171056e-06, "loss": 2.0473, "step": 25380 }, { "epoch": 0.63, "learning_rate": 6.494004832301171e-06, "loss": 2.0995, "step": 25381 }, { "epoch": 0.63, "learning_rate": 6.4932571860300846e-06, "loss": 2.0806, "step": 25382 }, { "epoch": 0.63, "learning_rate": 6.492509562108611e-06, "loss": 1.8294, "step": 25383 }, { "epoch": 0.63, "learning_rate": 6.49176196054152e-06, "loss": 1.9861, "step": 25384 }, { "epoch": 0.63, "learning_rate": 6.4910143813335715e-06, "loss": 1.8782, "step": 25385 }, { "epoch": 0.63, "learning_rate": 6.490266824489534e-06, "loss": 2.0578, "step": 25386 }, { "epoch": 0.63, "learning_rate": 6.489519290014169e-06, "loss": 1.9684, "step": 25387 }, { "epoch": 0.63, "learning_rate": 6.488771777912239e-06, "loss": 1.9748, "step": 25388 }, { "epoch": 0.63, "learning_rate": 6.488024288188514e-06, "loss": 2.0145, "step": 25389 }, { "epoch": 0.63, "learning_rate": 6.487276820847752e-06, "loss": 1.8585, "step": 25390 }, { "epoch": 0.63, "learning_rate": 6.486529375894724e-06, "loss": 2.1442, "step": 25391 }, { "epoch": 0.63, "learning_rate": 6.485781953334186e-06, "loss": 1.9566, "step": 25392 }, { "epoch": 0.63, "learning_rate": 6.485034553170903e-06, "loss": 2.1035, "step": 25393 }, { "epoch": 0.63, "learning_rate": 6.484287175409646e-06, "loss": 1.9104, "step": 25394 }, { "epoch": 0.63, "learning_rate": 6.483539820055168e-06, "loss": 2.1152, "step": 25395 }, { "epoch": 0.63, "learning_rate": 6.482792487112238e-06, "loss": 2.0431, "step": 25396 }, { "epoch": 0.63, "learning_rate": 6.482045176585616e-06, "loss": 1.9919, "step": 25397 }, { "epoch": 0.63, "learning_rate": 6.481297888480067e-06, "loss": 1.9729, "step": 25398 }, { "epoch": 0.63, "learning_rate": 6.480550622800358e-06, "loss": 1.9044, "step": 25399 }, { "epoch": 0.63, "learning_rate": 6.479803379551241e-06, "loss": 1.8753, "step": 25400 }, { "epoch": 0.63, "learning_rate": 6.479056158737489e-06, "loss": 1.8841, "step": 25401 }, { "epoch": 0.63, "learning_rate": 6.478308960363857e-06, "loss": 1.838, "step": 25402 }, { "epoch": 0.63, "learning_rate": 6.477561784435108e-06, "loss": 2.0238, "step": 25403 }, { "epoch": 0.63, "learning_rate": 6.4768146309560074e-06, "loss": 1.9031, "step": 25404 }, { "epoch": 0.63, "learning_rate": 6.476067499931317e-06, "loss": 2.096, "step": 25405 }, { "epoch": 0.63, "learning_rate": 6.475320391365796e-06, "loss": 1.9897, "step": 25406 }, { "epoch": 0.63, "learning_rate": 6.474573305264205e-06, "loss": 2.0422, "step": 25407 }, { "epoch": 0.63, "learning_rate": 6.473826241631306e-06, "loss": 1.9003, "step": 25408 }, { "epoch": 0.63, "learning_rate": 6.473079200471866e-06, "loss": 2.0321, "step": 25409 }, { "epoch": 0.63, "learning_rate": 6.472332181790638e-06, "loss": 1.8709, "step": 25410 }, { "epoch": 0.63, "learning_rate": 6.47158518559239e-06, "loss": 1.9799, "step": 25411 }, { "epoch": 0.63, "learning_rate": 6.470838211881878e-06, "loss": 2.0211, "step": 25412 }, { "epoch": 0.63, "learning_rate": 6.470091260663862e-06, "loss": 1.9524, "step": 25413 }, { "epoch": 0.63, "learning_rate": 6.469344331943106e-06, "loss": 2.0577, "step": 25414 }, { "epoch": 0.63, "learning_rate": 6.468597425724369e-06, "loss": 1.9783, "step": 25415 }, { "epoch": 0.63, "learning_rate": 6.467850542012415e-06, "loss": 2.0336, "step": 25416 }, { "epoch": 0.63, "learning_rate": 6.467103680811997e-06, "loss": 1.7971, "step": 25417 }, { "epoch": 0.63, "learning_rate": 6.466356842127877e-06, "loss": 1.8636, "step": 25418 }, { "epoch": 0.63, "learning_rate": 6.465610025964821e-06, "loss": 1.9424, "step": 25419 }, { "epoch": 0.63, "learning_rate": 6.464863232327581e-06, "loss": 1.9117, "step": 25420 }, { "epoch": 0.63, "learning_rate": 6.464116461220921e-06, "loss": 2.096, "step": 25421 }, { "epoch": 0.63, "learning_rate": 6.463369712649597e-06, "loss": 2.1705, "step": 25422 }, { "epoch": 0.63, "learning_rate": 6.462622986618373e-06, "loss": 1.922, "step": 25423 }, { "epoch": 0.63, "learning_rate": 6.4618762831320045e-06, "loss": 2.1633, "step": 25424 }, { "epoch": 0.63, "learning_rate": 6.461129602195251e-06, "loss": 2.0029, "step": 25425 }, { "epoch": 0.63, "learning_rate": 6.460382943812875e-06, "loss": 1.9451, "step": 25426 }, { "epoch": 0.63, "learning_rate": 6.459636307989627e-06, "loss": 1.9642, "step": 25427 }, { "epoch": 0.63, "learning_rate": 6.458889694730272e-06, "loss": 1.9408, "step": 25428 }, { "epoch": 0.63, "learning_rate": 6.458143104039571e-06, "loss": 2.0914, "step": 25429 }, { "epoch": 0.63, "learning_rate": 6.457396535922273e-06, "loss": 1.9663, "step": 25430 }, { "epoch": 0.63, "learning_rate": 6.456649990383144e-06, "loss": 1.9726, "step": 25431 }, { "epoch": 0.63, "learning_rate": 6.455903467426938e-06, "loss": 2.2708, "step": 25432 }, { "epoch": 0.63, "learning_rate": 6.455156967058414e-06, "loss": 1.8635, "step": 25433 }, { "epoch": 0.63, "learning_rate": 6.4544104892823345e-06, "loss": 1.9778, "step": 25434 }, { "epoch": 0.63, "learning_rate": 6.453664034103448e-06, "loss": 1.8803, "step": 25435 }, { "epoch": 0.63, "learning_rate": 6.452917601526515e-06, "loss": 2.1365, "step": 25436 }, { "epoch": 0.63, "learning_rate": 6.452171191556299e-06, "loss": 1.783, "step": 25437 }, { "epoch": 0.63, "learning_rate": 6.451424804197548e-06, "loss": 1.882, "step": 25438 }, { "epoch": 0.63, "learning_rate": 6.4506784394550246e-06, "loss": 2.0062, "step": 25439 }, { "epoch": 0.63, "learning_rate": 6.449932097333484e-06, "loss": 2.0411, "step": 25440 }, { "epoch": 0.63, "learning_rate": 6.449185777837683e-06, "loss": 1.9411, "step": 25441 }, { "epoch": 0.63, "learning_rate": 6.4484394809723794e-06, "loss": 1.9953, "step": 25442 }, { "epoch": 0.63, "learning_rate": 6.447693206742327e-06, "loss": 2.005, "step": 25443 }, { "epoch": 0.63, "learning_rate": 6.446946955152287e-06, "loss": 1.8482, "step": 25444 }, { "epoch": 0.63, "learning_rate": 6.446200726207008e-06, "loss": 1.9514, "step": 25445 }, { "epoch": 0.63, "learning_rate": 6.445454519911249e-06, "loss": 1.7825, "step": 25446 }, { "epoch": 0.63, "learning_rate": 6.444708336269772e-06, "loss": 1.9982, "step": 25447 }, { "epoch": 0.63, "learning_rate": 6.443962175287321e-06, "loss": 2.2117, "step": 25448 }, { "epoch": 0.63, "learning_rate": 6.443216036968662e-06, "loss": 2.2395, "step": 25449 }, { "epoch": 0.63, "learning_rate": 6.442469921318545e-06, "loss": 1.9246, "step": 25450 }, { "epoch": 0.63, "learning_rate": 6.4417238283417246e-06, "loss": 1.8835, "step": 25451 }, { "epoch": 0.63, "learning_rate": 6.440977758042963e-06, "loss": 1.9954, "step": 25452 }, { "epoch": 0.63, "learning_rate": 6.440231710427006e-06, "loss": 1.8123, "step": 25453 }, { "epoch": 0.63, "learning_rate": 6.4394856854986165e-06, "loss": 2.0887, "step": 25454 }, { "epoch": 0.63, "learning_rate": 6.43873968326254e-06, "loss": 1.9991, "step": 25455 }, { "epoch": 0.63, "learning_rate": 6.437993703723537e-06, "loss": 2.0022, "step": 25456 }, { "epoch": 0.63, "learning_rate": 6.437247746886363e-06, "loss": 2.1141, "step": 25457 }, { "epoch": 0.63, "learning_rate": 6.436501812755769e-06, "loss": 1.9531, "step": 25458 }, { "epoch": 0.63, "learning_rate": 6.435755901336512e-06, "loss": 1.9533, "step": 25459 }, { "epoch": 0.63, "learning_rate": 6.435010012633342e-06, "loss": 2.0299, "step": 25460 }, { "epoch": 0.63, "learning_rate": 6.434264146651015e-06, "loss": 1.8683, "step": 25461 }, { "epoch": 0.63, "learning_rate": 6.433518303394287e-06, "loss": 1.9188, "step": 25462 }, { "epoch": 0.63, "learning_rate": 6.432772482867906e-06, "loss": 1.9509, "step": 25463 }, { "epoch": 0.63, "learning_rate": 6.432026685076633e-06, "loss": 1.9171, "step": 25464 }, { "epoch": 0.63, "learning_rate": 6.431280910025213e-06, "loss": 1.7667, "step": 25465 }, { "epoch": 0.63, "learning_rate": 6.430535157718402e-06, "loss": 1.9854, "step": 25466 }, { "epoch": 0.63, "learning_rate": 6.429789428160955e-06, "loss": 1.8824, "step": 25467 }, { "epoch": 0.63, "learning_rate": 6.429043721357623e-06, "loss": 1.8407, "step": 25468 }, { "epoch": 0.63, "learning_rate": 6.428298037313163e-06, "loss": 2.0755, "step": 25469 }, { "epoch": 0.63, "learning_rate": 6.427552376032318e-06, "loss": 1.8744, "step": 25470 }, { "epoch": 0.63, "learning_rate": 6.426806737519847e-06, "loss": 2.1425, "step": 25471 }, { "epoch": 0.63, "learning_rate": 6.4260611217805045e-06, "loss": 2.1308, "step": 25472 }, { "epoch": 0.63, "learning_rate": 6.425315528819036e-06, "loss": 2.0436, "step": 25473 }, { "epoch": 0.63, "learning_rate": 6.4245699586401985e-06, "loss": 2.1289, "step": 25474 }, { "epoch": 0.63, "learning_rate": 6.4238244112487405e-06, "loss": 2.0251, "step": 25475 }, { "epoch": 0.63, "learning_rate": 6.423078886649416e-06, "loss": 1.9089, "step": 25476 }, { "epoch": 0.63, "learning_rate": 6.422333384846975e-06, "loss": 1.9714, "step": 25477 }, { "epoch": 0.63, "learning_rate": 6.421587905846168e-06, "loss": 2.0372, "step": 25478 }, { "epoch": 0.63, "learning_rate": 6.420842449651752e-06, "loss": 1.8683, "step": 25479 }, { "epoch": 0.63, "learning_rate": 6.420097016268469e-06, "loss": 1.9739, "step": 25480 }, { "epoch": 0.63, "learning_rate": 6.419351605701074e-06, "loss": 1.9982, "step": 25481 }, { "epoch": 0.63, "learning_rate": 6.418606217954322e-06, "loss": 2.0605, "step": 25482 }, { "epoch": 0.63, "learning_rate": 6.417860853032956e-06, "loss": 1.9417, "step": 25483 }, { "epoch": 0.63, "learning_rate": 6.417115510941732e-06, "loss": 1.9134, "step": 25484 }, { "epoch": 0.63, "learning_rate": 6.416370191685397e-06, "loss": 1.9632, "step": 25485 }, { "epoch": 0.63, "learning_rate": 6.415624895268702e-06, "loss": 2.0335, "step": 25486 }, { "epoch": 0.63, "learning_rate": 6.4148796216964015e-06, "loss": 1.8553, "step": 25487 }, { "epoch": 0.63, "learning_rate": 6.414134370973238e-06, "loss": 1.9189, "step": 25488 }, { "epoch": 0.63, "learning_rate": 6.413389143103965e-06, "loss": 1.9187, "step": 25489 }, { "epoch": 0.63, "learning_rate": 6.4126439380933356e-06, "loss": 2.0288, "step": 25490 }, { "epoch": 0.63, "learning_rate": 6.411898755946091e-06, "loss": 1.8818, "step": 25491 }, { "epoch": 0.63, "learning_rate": 6.411153596666988e-06, "loss": 2.0238, "step": 25492 }, { "epoch": 0.63, "learning_rate": 6.410408460260771e-06, "loss": 1.8824, "step": 25493 }, { "epoch": 0.63, "learning_rate": 6.409663346732191e-06, "loss": 2.0931, "step": 25494 }, { "epoch": 0.63, "learning_rate": 6.4089182560859975e-06, "loss": 1.9136, "step": 25495 }, { "epoch": 0.63, "learning_rate": 6.408173188326936e-06, "loss": 1.9025, "step": 25496 }, { "epoch": 0.63, "learning_rate": 6.407428143459762e-06, "loss": 2.0206, "step": 25497 }, { "epoch": 0.63, "learning_rate": 6.406683121489214e-06, "loss": 2.1203, "step": 25498 }, { "epoch": 0.63, "learning_rate": 6.405938122420045e-06, "loss": 1.9416, "step": 25499 }, { "epoch": 0.63, "learning_rate": 6.40519314625701e-06, "loss": 2.0492, "step": 25500 }, { "epoch": 0.63, "learning_rate": 6.404448193004845e-06, "loss": 1.9606, "step": 25501 }, { "epoch": 0.63, "learning_rate": 6.403703262668306e-06, "loss": 2.252, "step": 25502 }, { "epoch": 0.63, "learning_rate": 6.402958355252136e-06, "loss": 1.984, "step": 25503 }, { "epoch": 0.63, "learning_rate": 6.402213470761085e-06, "loss": 1.9984, "step": 25504 }, { "epoch": 0.63, "learning_rate": 6.4014686091999036e-06, "loss": 1.8755, "step": 25505 }, { "epoch": 0.63, "learning_rate": 6.400723770573331e-06, "loss": 1.9686, "step": 25506 }, { "epoch": 0.63, "learning_rate": 6.399978954886124e-06, "loss": 1.8475, "step": 25507 }, { "epoch": 0.63, "learning_rate": 6.399234162143019e-06, "loss": 2.0493, "step": 25508 }, { "epoch": 0.63, "learning_rate": 6.398489392348768e-06, "loss": 1.736, "step": 25509 }, { "epoch": 0.63, "learning_rate": 6.397744645508121e-06, "loss": 1.8733, "step": 25510 }, { "epoch": 0.63, "learning_rate": 6.396999921625819e-06, "loss": 2.0028, "step": 25511 }, { "epoch": 0.63, "learning_rate": 6.396255220706614e-06, "loss": 1.9358, "step": 25512 }, { "epoch": 0.63, "learning_rate": 6.395510542755245e-06, "loss": 1.9846, "step": 25513 }, { "epoch": 0.63, "learning_rate": 6.394765887776463e-06, "loss": 1.8415, "step": 25514 }, { "epoch": 0.63, "learning_rate": 6.3940212557750155e-06, "loss": 2.0421, "step": 25515 }, { "epoch": 0.63, "learning_rate": 6.393276646755644e-06, "loss": 1.9351, "step": 25516 }, { "epoch": 0.63, "learning_rate": 6.392532060723098e-06, "loss": 1.9383, "step": 25517 }, { "epoch": 0.63, "learning_rate": 6.391787497682116e-06, "loss": 1.8034, "step": 25518 }, { "epoch": 0.63, "learning_rate": 6.39104295763745e-06, "loss": 2.0431, "step": 25519 }, { "epoch": 0.63, "learning_rate": 6.390298440593845e-06, "loss": 2.0977, "step": 25520 }, { "epoch": 0.63, "learning_rate": 6.389553946556042e-06, "loss": 1.9566, "step": 25521 }, { "epoch": 0.63, "learning_rate": 6.388809475528792e-06, "loss": 2.0213, "step": 25522 }, { "epoch": 0.63, "learning_rate": 6.388065027516832e-06, "loss": 1.8457, "step": 25523 }, { "epoch": 0.63, "learning_rate": 6.387320602524911e-06, "loss": 2.0848, "step": 25524 }, { "epoch": 0.63, "learning_rate": 6.386576200557777e-06, "loss": 1.876, "step": 25525 }, { "epoch": 0.63, "learning_rate": 6.385831821620168e-06, "loss": 1.942, "step": 25526 }, { "epoch": 0.63, "learning_rate": 6.38508746571683e-06, "loss": 1.8997, "step": 25527 }, { "epoch": 0.63, "learning_rate": 6.3843431328525075e-06, "loss": 1.9101, "step": 25528 }, { "epoch": 0.63, "learning_rate": 6.383598823031945e-06, "loss": 1.8704, "step": 25529 }, { "epoch": 0.63, "learning_rate": 6.382854536259888e-06, "loss": 2.0871, "step": 25530 }, { "epoch": 0.63, "learning_rate": 6.382110272541075e-06, "loss": 1.9362, "step": 25531 }, { "epoch": 0.63, "learning_rate": 6.381366031880255e-06, "loss": 2.0331, "step": 25532 }, { "epoch": 0.63, "learning_rate": 6.380621814282166e-06, "loss": 1.9685, "step": 25533 }, { "epoch": 0.63, "learning_rate": 6.379877619751554e-06, "loss": 1.8737, "step": 25534 }, { "epoch": 0.63, "learning_rate": 6.379133448293163e-06, "loss": 1.9524, "step": 25535 }, { "epoch": 0.63, "learning_rate": 6.378389299911734e-06, "loss": 1.9484, "step": 25536 }, { "epoch": 0.63, "learning_rate": 6.37764517461201e-06, "loss": 1.8984, "step": 25537 }, { "epoch": 0.63, "learning_rate": 6.376901072398731e-06, "loss": 1.9594, "step": 25538 }, { "epoch": 0.63, "learning_rate": 6.3761569932766446e-06, "loss": 2.0686, "step": 25539 }, { "epoch": 0.63, "learning_rate": 6.375412937250494e-06, "loss": 2.0425, "step": 25540 }, { "epoch": 0.63, "learning_rate": 6.3746689043250125e-06, "loss": 1.7688, "step": 25541 }, { "epoch": 0.63, "learning_rate": 6.373924894504953e-06, "loss": 2.0925, "step": 25542 }, { "epoch": 0.63, "learning_rate": 6.373180907795047e-06, "loss": 1.8309, "step": 25543 }, { "epoch": 0.63, "learning_rate": 6.3724369442000424e-06, "loss": 2.1431, "step": 25544 }, { "epoch": 0.63, "learning_rate": 6.3716930037246795e-06, "loss": 1.9116, "step": 25545 }, { "epoch": 0.63, "learning_rate": 6.370949086373699e-06, "loss": 2.0046, "step": 25546 }, { "epoch": 0.63, "learning_rate": 6.3702051921518415e-06, "loss": 1.9631, "step": 25547 }, { "epoch": 0.63, "learning_rate": 6.369461321063855e-06, "loss": 2.0468, "step": 25548 }, { "epoch": 0.63, "learning_rate": 6.36871747311447e-06, "loss": 1.9761, "step": 25549 }, { "epoch": 0.63, "learning_rate": 6.367973648308434e-06, "loss": 1.887, "step": 25550 }, { "epoch": 0.63, "learning_rate": 6.367229846650483e-06, "loss": 1.7832, "step": 25551 }, { "epoch": 0.63, "learning_rate": 6.366486068145361e-06, "loss": 2.0395, "step": 25552 }, { "epoch": 0.63, "learning_rate": 6.36574231279781e-06, "loss": 2.1299, "step": 25553 }, { "epoch": 0.63, "learning_rate": 6.3649985806125645e-06, "loss": 2.0404, "step": 25554 }, { "epoch": 0.63, "learning_rate": 6.364254871594369e-06, "loss": 2.1222, "step": 25555 }, { "epoch": 0.63, "learning_rate": 6.36351118574796e-06, "loss": 1.9464, "step": 25556 }, { "epoch": 0.63, "learning_rate": 6.36276752307808e-06, "loss": 2.1274, "step": 25557 }, { "epoch": 0.63, "learning_rate": 6.362023883589472e-06, "loss": 2.046, "step": 25558 }, { "epoch": 0.63, "learning_rate": 6.361280267286866e-06, "loss": 2.0952, "step": 25559 }, { "epoch": 0.63, "learning_rate": 6.360536674175012e-06, "loss": 2.0704, "step": 25560 }, { "epoch": 0.63, "learning_rate": 6.359793104258639e-06, "loss": 1.9923, "step": 25561 }, { "epoch": 0.63, "learning_rate": 6.359049557542491e-06, "loss": 2.031, "step": 25562 }, { "epoch": 0.63, "learning_rate": 6.35830603403131e-06, "loss": 2.1683, "step": 25563 }, { "epoch": 0.63, "learning_rate": 6.357562533729828e-06, "loss": 1.8687, "step": 25564 }, { "epoch": 0.63, "learning_rate": 6.356819056642791e-06, "loss": 2.1092, "step": 25565 }, { "epoch": 0.63, "learning_rate": 6.35607560277493e-06, "loss": 2.0631, "step": 25566 }, { "epoch": 0.63, "learning_rate": 6.355332172130987e-06, "loss": 2.0821, "step": 25567 }, { "epoch": 0.63, "learning_rate": 6.3545887647157035e-06, "loss": 1.9716, "step": 25568 }, { "epoch": 0.63, "learning_rate": 6.353845380533809e-06, "loss": 2.0578, "step": 25569 }, { "epoch": 0.63, "learning_rate": 6.353102019590052e-06, "loss": 1.975, "step": 25570 }, { "epoch": 0.63, "learning_rate": 6.352358681889159e-06, "loss": 1.8717, "step": 25571 }, { "epoch": 0.63, "learning_rate": 6.351615367435873e-06, "loss": 2.1398, "step": 25572 }, { "epoch": 0.63, "learning_rate": 6.350872076234933e-06, "loss": 1.853, "step": 25573 }, { "epoch": 0.63, "learning_rate": 6.350128808291073e-06, "loss": 1.9984, "step": 25574 }, { "epoch": 0.63, "learning_rate": 6.349385563609036e-06, "loss": 1.926, "step": 25575 }, { "epoch": 0.63, "learning_rate": 6.34864234219355e-06, "loss": 1.9786, "step": 25576 }, { "epoch": 0.63, "learning_rate": 6.347899144049355e-06, "loss": 2.0607, "step": 25577 }, { "epoch": 0.63, "learning_rate": 6.3471559691811946e-06, "loss": 2.0598, "step": 25578 }, { "epoch": 0.63, "learning_rate": 6.346412817593795e-06, "loss": 2.0072, "step": 25579 }, { "epoch": 0.63, "learning_rate": 6.3456696892919e-06, "loss": 2.1695, "step": 25580 }, { "epoch": 0.63, "learning_rate": 6.34492658428024e-06, "loss": 2.0217, "step": 25581 }, { "epoch": 0.63, "learning_rate": 6.3441835025635546e-06, "loss": 1.9027, "step": 25582 }, { "epoch": 0.63, "learning_rate": 6.343440444146582e-06, "loss": 2.0191, "step": 25583 }, { "epoch": 0.63, "learning_rate": 6.342697409034052e-06, "loss": 2.0726, "step": 25584 }, { "epoch": 0.63, "learning_rate": 6.341954397230707e-06, "loss": 1.8985, "step": 25585 }, { "epoch": 0.63, "learning_rate": 6.341211408741274e-06, "loss": 2.0003, "step": 25586 }, { "epoch": 0.63, "learning_rate": 6.340468443570496e-06, "loss": 1.9737, "step": 25587 }, { "epoch": 0.63, "learning_rate": 6.339725501723107e-06, "loss": 1.9812, "step": 25588 }, { "epoch": 0.63, "learning_rate": 6.338982583203837e-06, "loss": 1.9569, "step": 25589 }, { "epoch": 0.63, "learning_rate": 6.338239688017425e-06, "loss": 2.0435, "step": 25590 }, { "epoch": 0.63, "learning_rate": 6.337496816168602e-06, "loss": 1.8315, "step": 25591 }, { "epoch": 0.63, "learning_rate": 6.336753967662108e-06, "loss": 1.9486, "step": 25592 }, { "epoch": 0.63, "learning_rate": 6.336011142502677e-06, "loss": 1.8497, "step": 25593 }, { "epoch": 0.63, "learning_rate": 6.335268340695038e-06, "loss": 2.0671, "step": 25594 }, { "epoch": 0.63, "learning_rate": 6.334525562243931e-06, "loss": 1.7609, "step": 25595 }, { "epoch": 0.63, "learning_rate": 6.333782807154084e-06, "loss": 1.9683, "step": 25596 }, { "epoch": 0.63, "learning_rate": 6.333040075430234e-06, "loss": 2.0371, "step": 25597 }, { "epoch": 0.63, "learning_rate": 6.332297367077116e-06, "loss": 1.8241, "step": 25598 }, { "epoch": 0.63, "learning_rate": 6.331554682099461e-06, "loss": 1.9657, "step": 25599 }, { "epoch": 0.63, "learning_rate": 6.330812020502007e-06, "loss": 2.2956, "step": 25600 }, { "epoch": 0.63, "learning_rate": 6.330069382289481e-06, "loss": 1.9891, "step": 25601 }, { "epoch": 0.63, "learning_rate": 6.329326767466617e-06, "loss": 1.9714, "step": 25602 }, { "epoch": 0.63, "learning_rate": 6.328584176038155e-06, "loss": 2.0122, "step": 25603 }, { "epoch": 0.63, "learning_rate": 6.3278416080088176e-06, "loss": 2.0867, "step": 25604 }, { "epoch": 0.63, "learning_rate": 6.327099063383344e-06, "loss": 1.8765, "step": 25605 }, { "epoch": 0.63, "learning_rate": 6.326356542166466e-06, "loss": 2.0507, "step": 25606 }, { "epoch": 0.63, "learning_rate": 6.3256140443629136e-06, "loss": 2.0328, "step": 25607 }, { "epoch": 0.63, "learning_rate": 6.32487156997742e-06, "loss": 2.0228, "step": 25608 }, { "epoch": 0.63, "learning_rate": 6.324129119014717e-06, "loss": 2.031, "step": 25609 }, { "epoch": 0.63, "learning_rate": 6.323386691479538e-06, "loss": 1.9867, "step": 25610 }, { "epoch": 0.63, "learning_rate": 6.322644287376617e-06, "loss": 1.9094, "step": 25611 }, { "epoch": 0.63, "learning_rate": 6.321901906710678e-06, "loss": 1.9628, "step": 25612 }, { "epoch": 0.63, "learning_rate": 6.321159549486461e-06, "loss": 2.016, "step": 25613 }, { "epoch": 0.63, "learning_rate": 6.3204172157086896e-06, "loss": 2.049, "step": 25614 }, { "epoch": 0.63, "learning_rate": 6.319674905382098e-06, "loss": 1.9166, "step": 25615 }, { "epoch": 0.63, "learning_rate": 6.31893261851142e-06, "loss": 1.942, "step": 25616 }, { "epoch": 0.63, "learning_rate": 6.318190355101381e-06, "loss": 2.1286, "step": 25617 }, { "epoch": 0.63, "learning_rate": 6.317448115156719e-06, "loss": 1.9724, "step": 25618 }, { "epoch": 0.63, "learning_rate": 6.316705898682156e-06, "loss": 2.1219, "step": 25619 }, { "epoch": 0.63, "learning_rate": 6.315963705682428e-06, "loss": 1.9545, "step": 25620 }, { "epoch": 0.63, "learning_rate": 6.315221536162267e-06, "loss": 1.8848, "step": 25621 }, { "epoch": 0.63, "learning_rate": 6.314479390126395e-06, "loss": 1.9177, "step": 25622 }, { "epoch": 0.63, "learning_rate": 6.313737267579552e-06, "loss": 2.0699, "step": 25623 }, { "epoch": 0.63, "learning_rate": 6.31299516852646e-06, "loss": 2.0551, "step": 25624 }, { "epoch": 0.63, "learning_rate": 6.31225309297185e-06, "loss": 2.0002, "step": 25625 }, { "epoch": 0.63, "learning_rate": 6.311511040920455e-06, "loss": 1.8922, "step": 25626 }, { "epoch": 0.63, "learning_rate": 6.310769012377e-06, "loss": 1.9117, "step": 25627 }, { "epoch": 0.63, "learning_rate": 6.310027007346221e-06, "loss": 2.1613, "step": 25628 }, { "epoch": 0.63, "learning_rate": 6.3092850258328385e-06, "loss": 1.8198, "step": 25629 }, { "epoch": 0.63, "learning_rate": 6.3085430678415834e-06, "loss": 1.9267, "step": 25630 }, { "epoch": 0.63, "learning_rate": 6.307801133377193e-06, "loss": 2.0505, "step": 25631 }, { "epoch": 0.63, "learning_rate": 6.307059222444384e-06, "loss": 1.8872, "step": 25632 }, { "epoch": 0.63, "learning_rate": 6.306317335047891e-06, "loss": 1.8256, "step": 25633 }, { "epoch": 0.63, "learning_rate": 6.305575471192441e-06, "loss": 2.0364, "step": 25634 }, { "epoch": 0.63, "learning_rate": 6.304833630882762e-06, "loss": 2.1624, "step": 25635 }, { "epoch": 0.63, "learning_rate": 6.304091814123586e-06, "loss": 2.0059, "step": 25636 }, { "epoch": 0.63, "learning_rate": 6.303350020919634e-06, "loss": 2.0796, "step": 25637 }, { "epoch": 0.63, "learning_rate": 6.3026082512756414e-06, "loss": 1.9543, "step": 25638 }, { "epoch": 0.63, "learning_rate": 6.301866505196326e-06, "loss": 2.0219, "step": 25639 }, { "epoch": 0.63, "learning_rate": 6.301124782686421e-06, "loss": 2.0224, "step": 25640 }, { "epoch": 0.63, "learning_rate": 6.300383083750655e-06, "loss": 1.9136, "step": 25641 }, { "epoch": 0.63, "learning_rate": 6.299641408393754e-06, "loss": 1.9222, "step": 25642 }, { "epoch": 0.63, "learning_rate": 6.298899756620444e-06, "loss": 2.0741, "step": 25643 }, { "epoch": 0.63, "learning_rate": 6.29815812843545e-06, "loss": 1.9136, "step": 25644 }, { "epoch": 0.63, "learning_rate": 6.2974165238435e-06, "loss": 2.0034, "step": 25645 }, { "epoch": 0.63, "learning_rate": 6.296674942849326e-06, "loss": 1.9079, "step": 25646 }, { "epoch": 0.63, "learning_rate": 6.2959333854576456e-06, "loss": 1.8224, "step": 25647 }, { "epoch": 0.63, "learning_rate": 6.295191851673193e-06, "loss": 2.0746, "step": 25648 }, { "epoch": 0.63, "learning_rate": 6.294450341500684e-06, "loss": 2.0451, "step": 25649 }, { "epoch": 0.63, "learning_rate": 6.293708854944853e-06, "loss": 1.8211, "step": 25650 }, { "epoch": 0.63, "learning_rate": 6.292967392010424e-06, "loss": 2.0249, "step": 25651 }, { "epoch": 0.63, "learning_rate": 6.29222595270212e-06, "loss": 1.9071, "step": 25652 }, { "epoch": 0.63, "learning_rate": 6.291484537024672e-06, "loss": 1.9645, "step": 25653 }, { "epoch": 0.63, "learning_rate": 6.290743144982799e-06, "loss": 1.9419, "step": 25654 }, { "epoch": 0.63, "learning_rate": 6.2900017765812285e-06, "loss": 1.9333, "step": 25655 }, { "epoch": 0.63, "learning_rate": 6.2892604318246884e-06, "loss": 1.9807, "step": 25656 }, { "epoch": 0.63, "learning_rate": 6.288519110717899e-06, "loss": 1.8819, "step": 25657 }, { "epoch": 0.63, "learning_rate": 6.287777813265586e-06, "loss": 2.1395, "step": 25658 }, { "epoch": 0.63, "learning_rate": 6.287036539472476e-06, "loss": 1.8625, "step": 25659 }, { "epoch": 0.63, "learning_rate": 6.286295289343292e-06, "loss": 1.984, "step": 25660 }, { "epoch": 0.63, "learning_rate": 6.285554062882758e-06, "loss": 1.9043, "step": 25661 }, { "epoch": 0.63, "learning_rate": 6.284812860095597e-06, "loss": 2.0882, "step": 25662 }, { "epoch": 0.63, "learning_rate": 6.284071680986539e-06, "loss": 2.0444, "step": 25663 }, { "epoch": 0.63, "learning_rate": 6.283330525560298e-06, "loss": 1.7595, "step": 25664 }, { "epoch": 0.63, "learning_rate": 6.282589393821604e-06, "loss": 1.9736, "step": 25665 }, { "epoch": 0.63, "learning_rate": 6.281848285775183e-06, "loss": 1.9354, "step": 25666 }, { "epoch": 0.63, "learning_rate": 6.281107201425751e-06, "loss": 1.9251, "step": 25667 }, { "epoch": 0.63, "learning_rate": 6.280366140778033e-06, "loss": 2.0243, "step": 25668 }, { "epoch": 0.63, "learning_rate": 6.2796251038367585e-06, "loss": 1.9615, "step": 25669 }, { "epoch": 0.63, "learning_rate": 6.278884090606641e-06, "loss": 2.0336, "step": 25670 }, { "epoch": 0.63, "learning_rate": 6.2781431010924135e-06, "loss": 1.8377, "step": 25671 }, { "epoch": 0.63, "learning_rate": 6.277402135298789e-06, "loss": 1.979, "step": 25672 }, { "epoch": 0.63, "learning_rate": 6.2766611932304935e-06, "loss": 1.9826, "step": 25673 }, { "epoch": 0.63, "learning_rate": 6.2759202748922535e-06, "loss": 2.0231, "step": 25674 }, { "epoch": 0.63, "learning_rate": 6.275179380288783e-06, "loss": 1.9848, "step": 25675 }, { "epoch": 0.63, "learning_rate": 6.274438509424809e-06, "loss": 1.9909, "step": 25676 }, { "epoch": 0.63, "learning_rate": 6.273697662305055e-06, "loss": 2.0994, "step": 25677 }, { "epoch": 0.63, "learning_rate": 6.272956838934237e-06, "loss": 1.9809, "step": 25678 }, { "epoch": 0.63, "learning_rate": 6.272216039317082e-06, "loss": 2.0273, "step": 25679 }, { "epoch": 0.63, "learning_rate": 6.271475263458306e-06, "loss": 2.062, "step": 25680 }, { "epoch": 0.63, "learning_rate": 6.270734511362638e-06, "loss": 2.0548, "step": 25681 }, { "epoch": 0.63, "learning_rate": 6.269993783034791e-06, "loss": 2.0925, "step": 25682 }, { "epoch": 0.63, "learning_rate": 6.269253078479488e-06, "loss": 2.0283, "step": 25683 }, { "epoch": 0.63, "learning_rate": 6.268512397701456e-06, "loss": 1.995, "step": 25684 }, { "epoch": 0.63, "learning_rate": 6.267771740705405e-06, "loss": 2.101, "step": 25685 }, { "epoch": 0.63, "learning_rate": 6.267031107496065e-06, "loss": 1.8537, "step": 25686 }, { "epoch": 0.63, "learning_rate": 6.2662904980781495e-06, "loss": 2.0158, "step": 25687 }, { "epoch": 0.63, "learning_rate": 6.26554991245638e-06, "loss": 1.921, "step": 25688 }, { "epoch": 0.63, "learning_rate": 6.264809350635483e-06, "loss": 2.0004, "step": 25689 }, { "epoch": 0.63, "learning_rate": 6.26406881262017e-06, "loss": 1.8314, "step": 25690 }, { "epoch": 0.63, "learning_rate": 6.263328298415168e-06, "loss": 2.01, "step": 25691 }, { "epoch": 0.63, "learning_rate": 6.262587808025186e-06, "loss": 1.8949, "step": 25692 }, { "epoch": 0.63, "learning_rate": 6.261847341454952e-06, "loss": 1.6904, "step": 25693 }, { "epoch": 0.63, "learning_rate": 6.261106898709186e-06, "loss": 1.9857, "step": 25694 }, { "epoch": 0.63, "learning_rate": 6.260366479792602e-06, "loss": 2.0257, "step": 25695 }, { "epoch": 0.63, "learning_rate": 6.259626084709922e-06, "loss": 1.9897, "step": 25696 }, { "epoch": 0.63, "learning_rate": 6.2588857134658635e-06, "loss": 1.9622, "step": 25697 }, { "epoch": 0.63, "learning_rate": 6.258145366065143e-06, "loss": 1.9684, "step": 25698 }, { "epoch": 0.63, "learning_rate": 6.257405042512488e-06, "loss": 2.1292, "step": 25699 }, { "epoch": 0.63, "learning_rate": 6.2566647428126036e-06, "loss": 1.9293, "step": 25700 }, { "epoch": 0.63, "learning_rate": 6.2559244669702205e-06, "loss": 2.002, "step": 25701 }, { "epoch": 0.63, "learning_rate": 6.255184214990046e-06, "loss": 1.9076, "step": 25702 }, { "epoch": 0.63, "learning_rate": 6.2544439868768035e-06, "loss": 2.1012, "step": 25703 }, { "epoch": 0.63, "learning_rate": 6.253703782635212e-06, "loss": 1.9336, "step": 25704 }, { "epoch": 0.63, "learning_rate": 6.252963602269986e-06, "loss": 2.0366, "step": 25705 }, { "epoch": 0.63, "learning_rate": 6.252223445785848e-06, "loss": 2.0792, "step": 25706 }, { "epoch": 0.63, "learning_rate": 6.251483313187505e-06, "loss": 2.0629, "step": 25707 }, { "epoch": 0.63, "learning_rate": 6.250743204479684e-06, "loss": 1.9244, "step": 25708 }, { "epoch": 0.63, "learning_rate": 6.250003119667099e-06, "loss": 1.8697, "step": 25709 }, { "epoch": 0.63, "learning_rate": 6.249263058754464e-06, "loss": 2.0144, "step": 25710 }, { "epoch": 0.63, "learning_rate": 6.2485230217465e-06, "loss": 1.9618, "step": 25711 }, { "epoch": 0.63, "learning_rate": 6.247783008647918e-06, "loss": 2.0718, "step": 25712 }, { "epoch": 0.63, "learning_rate": 6.247043019463441e-06, "loss": 2.109, "step": 25713 }, { "epoch": 0.63, "learning_rate": 6.246303054197782e-06, "loss": 2.087, "step": 25714 }, { "epoch": 0.63, "learning_rate": 6.245563112855654e-06, "loss": 1.8972, "step": 25715 }, { "epoch": 0.63, "learning_rate": 6.2448231954417805e-06, "loss": 2.053, "step": 25716 }, { "epoch": 0.63, "learning_rate": 6.244083301960869e-06, "loss": 1.9715, "step": 25717 }, { "epoch": 0.63, "learning_rate": 6.243343432417636e-06, "loss": 1.9637, "step": 25718 }, { "epoch": 0.63, "learning_rate": 6.242603586816807e-06, "loss": 1.9511, "step": 25719 }, { "epoch": 0.63, "learning_rate": 6.241863765163083e-06, "loss": 1.9035, "step": 25720 }, { "epoch": 0.63, "learning_rate": 6.2411239674611904e-06, "loss": 1.894, "step": 25721 }, { "epoch": 0.63, "learning_rate": 6.240384193715836e-06, "loss": 1.8969, "step": 25722 }, { "epoch": 0.63, "learning_rate": 6.2396444439317395e-06, "loss": 2.1202, "step": 25723 }, { "epoch": 0.63, "learning_rate": 6.2389047181136185e-06, "loss": 2.1464, "step": 25724 }, { "epoch": 0.63, "learning_rate": 6.2381650162661775e-06, "loss": 2.0177, "step": 25725 }, { "epoch": 0.63, "learning_rate": 6.237425338394139e-06, "loss": 1.9868, "step": 25726 }, { "epoch": 0.63, "learning_rate": 6.236685684502219e-06, "loss": 1.8282, "step": 25727 }, { "epoch": 0.63, "learning_rate": 6.235946054595121e-06, "loss": 2.0334, "step": 25728 }, { "epoch": 0.63, "learning_rate": 6.235206448677571e-06, "loss": 1.9812, "step": 25729 }, { "epoch": 0.63, "learning_rate": 6.234466866754272e-06, "loss": 1.933, "step": 25730 }, { "epoch": 0.63, "learning_rate": 6.233727308829946e-06, "loss": 1.889, "step": 25731 }, { "epoch": 0.63, "learning_rate": 6.2329877749093045e-06, "loss": 2.1024, "step": 25732 }, { "epoch": 0.63, "learning_rate": 6.232248264997056e-06, "loss": 2.0813, "step": 25733 }, { "epoch": 0.63, "learning_rate": 6.231508779097922e-06, "loss": 2.0479, "step": 25734 }, { "epoch": 0.63, "learning_rate": 6.230769317216606e-06, "loss": 1.9084, "step": 25735 }, { "epoch": 0.63, "learning_rate": 6.230029879357826e-06, "loss": 2.007, "step": 25736 }, { "epoch": 0.63, "learning_rate": 6.229290465526299e-06, "loss": 2.0201, "step": 25737 }, { "epoch": 0.63, "learning_rate": 6.228551075726728e-06, "loss": 2.0616, "step": 25738 }, { "epoch": 0.63, "learning_rate": 6.227811709963831e-06, "loss": 2.0337, "step": 25739 }, { "epoch": 0.63, "learning_rate": 6.227072368242317e-06, "loss": 1.9399, "step": 25740 }, { "epoch": 0.63, "learning_rate": 6.226333050566902e-06, "loss": 1.9913, "step": 25741 }, { "epoch": 0.63, "learning_rate": 6.225593756942301e-06, "loss": 1.9414, "step": 25742 }, { "epoch": 0.63, "learning_rate": 6.2248544873732155e-06, "loss": 2.1248, "step": 25743 }, { "epoch": 0.63, "learning_rate": 6.224115241864366e-06, "loss": 1.8974, "step": 25744 }, { "epoch": 0.63, "learning_rate": 6.223376020420457e-06, "loss": 2.0107, "step": 25745 }, { "epoch": 0.63, "learning_rate": 6.222636823046205e-06, "loss": 2.0368, "step": 25746 }, { "epoch": 0.63, "learning_rate": 6.22189764974632e-06, "loss": 1.8824, "step": 25747 }, { "epoch": 0.63, "learning_rate": 6.221158500525512e-06, "loss": 2.2049, "step": 25748 }, { "epoch": 0.63, "learning_rate": 6.2204193753884945e-06, "loss": 1.9016, "step": 25749 }, { "epoch": 0.63, "learning_rate": 6.219680274339973e-06, "loss": 2.0471, "step": 25750 }, { "epoch": 0.63, "learning_rate": 6.218941197384661e-06, "loss": 1.9252, "step": 25751 }, { "epoch": 0.63, "learning_rate": 6.218202144527272e-06, "loss": 2.2431, "step": 25752 }, { "epoch": 0.63, "learning_rate": 6.2174631157725105e-06, "loss": 1.8898, "step": 25753 }, { "epoch": 0.63, "learning_rate": 6.216724111125094e-06, "loss": 2.0079, "step": 25754 }, { "epoch": 0.63, "learning_rate": 6.215985130589721e-06, "loss": 1.9514, "step": 25755 }, { "epoch": 0.63, "learning_rate": 6.21524617417111e-06, "loss": 2.1259, "step": 25756 }, { "epoch": 0.63, "learning_rate": 6.21450724187397e-06, "loss": 1.8908, "step": 25757 }, { "epoch": 0.63, "learning_rate": 6.213768333703008e-06, "loss": 1.9459, "step": 25758 }, { "epoch": 0.63, "learning_rate": 6.213029449662937e-06, "loss": 1.8931, "step": 25759 }, { "epoch": 0.63, "learning_rate": 6.21229058975846e-06, "loss": 1.9984, "step": 25760 }, { "epoch": 0.63, "learning_rate": 6.211551753994289e-06, "loss": 1.9883, "step": 25761 }, { "epoch": 0.64, "learning_rate": 6.210812942375137e-06, "loss": 1.9298, "step": 25762 }, { "epoch": 0.64, "learning_rate": 6.210074154905705e-06, "loss": 2.056, "step": 25763 }, { "epoch": 0.64, "learning_rate": 6.20933539159071e-06, "loss": 1.819, "step": 25764 }, { "epoch": 0.64, "learning_rate": 6.208596652434851e-06, "loss": 1.9179, "step": 25765 }, { "epoch": 0.64, "learning_rate": 6.207857937442844e-06, "loss": 1.9375, "step": 25766 }, { "epoch": 0.64, "learning_rate": 6.2071192466193945e-06, "loss": 2.0437, "step": 25767 }, { "epoch": 0.64, "learning_rate": 6.206380579969207e-06, "loss": 2.0808, "step": 25768 }, { "epoch": 0.64, "learning_rate": 6.205641937496998e-06, "loss": 1.9312, "step": 25769 }, { "epoch": 0.64, "learning_rate": 6.204903319207465e-06, "loss": 1.9788, "step": 25770 }, { "epoch": 0.64, "learning_rate": 6.20416472510532e-06, "loss": 2.0041, "step": 25771 }, { "epoch": 0.64, "learning_rate": 6.203426155195272e-06, "loss": 2.0906, "step": 25772 }, { "epoch": 0.64, "learning_rate": 6.2026876094820255e-06, "loss": 1.8632, "step": 25773 }, { "epoch": 0.64, "learning_rate": 6.201949087970288e-06, "loss": 1.9065, "step": 25774 }, { "epoch": 0.64, "learning_rate": 6.2012105906647655e-06, "loss": 1.9218, "step": 25775 }, { "epoch": 0.64, "learning_rate": 6.200472117570166e-06, "loss": 2.0052, "step": 25776 }, { "epoch": 0.64, "learning_rate": 6.199733668691202e-06, "loss": 2.1226, "step": 25777 }, { "epoch": 0.64, "learning_rate": 6.198995244032566e-06, "loss": 2.0217, "step": 25778 }, { "epoch": 0.64, "learning_rate": 6.198256843598978e-06, "loss": 2.1368, "step": 25779 }, { "epoch": 0.64, "learning_rate": 6.197518467395135e-06, "loss": 1.7835, "step": 25780 }, { "epoch": 0.64, "learning_rate": 6.196780115425745e-06, "loss": 1.9269, "step": 25781 }, { "epoch": 0.64, "learning_rate": 6.196041787695517e-06, "loss": 1.9494, "step": 25782 }, { "epoch": 0.64, "learning_rate": 6.195303484209152e-06, "loss": 1.8616, "step": 25783 }, { "epoch": 0.64, "learning_rate": 6.194565204971359e-06, "loss": 1.8921, "step": 25784 }, { "epoch": 0.64, "learning_rate": 6.1938269499868435e-06, "loss": 1.9965, "step": 25785 }, { "epoch": 0.64, "learning_rate": 6.193088719260307e-06, "loss": 1.9891, "step": 25786 }, { "epoch": 0.64, "learning_rate": 6.192350512796461e-06, "loss": 1.9004, "step": 25787 }, { "epoch": 0.64, "learning_rate": 6.191612330600001e-06, "loss": 1.9679, "step": 25788 }, { "epoch": 0.64, "learning_rate": 6.190874172675639e-06, "loss": 1.94, "step": 25789 }, { "epoch": 0.64, "learning_rate": 6.1901360390280785e-06, "loss": 1.9066, "step": 25790 }, { "epoch": 0.64, "learning_rate": 6.18939792966202e-06, "loss": 2.0737, "step": 25791 }, { "epoch": 0.64, "learning_rate": 6.188659844582172e-06, "loss": 1.973, "step": 25792 }, { "epoch": 0.64, "learning_rate": 6.187921783793234e-06, "loss": 2.0049, "step": 25793 }, { "epoch": 0.64, "learning_rate": 6.187183747299914e-06, "loss": 1.7997, "step": 25794 }, { "epoch": 0.64, "learning_rate": 6.186445735106919e-06, "loss": 1.9421, "step": 25795 }, { "epoch": 0.64, "learning_rate": 6.185707747218943e-06, "loss": 1.9159, "step": 25796 }, { "epoch": 0.64, "learning_rate": 6.1849697836406985e-06, "loss": 1.7995, "step": 25797 }, { "epoch": 0.64, "learning_rate": 6.184231844376882e-06, "loss": 1.9792, "step": 25798 }, { "epoch": 0.64, "learning_rate": 6.183493929432198e-06, "loss": 2.0491, "step": 25799 }, { "epoch": 0.64, "learning_rate": 6.182756038811354e-06, "loss": 2.0214, "step": 25800 }, { "epoch": 0.64, "learning_rate": 6.182018172519048e-06, "loss": 2.0823, "step": 25801 }, { "epoch": 0.64, "learning_rate": 6.181280330559988e-06, "loss": 1.9046, "step": 25802 }, { "epoch": 0.64, "learning_rate": 6.180542512938869e-06, "loss": 1.9188, "step": 25803 }, { "epoch": 0.64, "learning_rate": 6.179804719660396e-06, "loss": 2.0718, "step": 25804 }, { "epoch": 0.64, "learning_rate": 6.179066950729278e-06, "loss": 1.913, "step": 25805 }, { "epoch": 0.64, "learning_rate": 6.178329206150208e-06, "loss": 2.011, "step": 25806 }, { "epoch": 0.64, "learning_rate": 6.177591485927894e-06, "loss": 2.0545, "step": 25807 }, { "epoch": 0.64, "learning_rate": 6.176853790067032e-06, "loss": 2.0974, "step": 25808 }, { "epoch": 0.64, "learning_rate": 6.176116118572326e-06, "loss": 1.9178, "step": 25809 }, { "epoch": 0.64, "learning_rate": 6.175378471448481e-06, "loss": 1.9639, "step": 25810 }, { "epoch": 0.64, "learning_rate": 6.174640848700193e-06, "loss": 1.8292, "step": 25811 }, { "epoch": 0.64, "learning_rate": 6.173903250332169e-06, "loss": 1.8038, "step": 25812 }, { "epoch": 0.64, "learning_rate": 6.173165676349103e-06, "loss": 2.0731, "step": 25813 }, { "epoch": 0.64, "learning_rate": 6.172428126755699e-06, "loss": 1.9487, "step": 25814 }, { "epoch": 0.64, "learning_rate": 6.171690601556664e-06, "loss": 1.9433, "step": 25815 }, { "epoch": 0.64, "learning_rate": 6.1709531007566854e-06, "loss": 1.9699, "step": 25816 }, { "epoch": 0.64, "learning_rate": 6.170215624360474e-06, "loss": 2.1256, "step": 25817 }, { "epoch": 0.64, "learning_rate": 6.169478172372725e-06, "loss": 1.7819, "step": 25818 }, { "epoch": 0.64, "learning_rate": 6.168740744798138e-06, "loss": 1.892, "step": 25819 }, { "epoch": 0.64, "learning_rate": 6.16800334164142e-06, "loss": 2.0949, "step": 25820 }, { "epoch": 0.64, "learning_rate": 6.1672659629072614e-06, "loss": 1.9736, "step": 25821 }, { "epoch": 0.64, "learning_rate": 6.166528608600371e-06, "loss": 1.8661, "step": 25822 }, { "epoch": 0.64, "learning_rate": 6.165791278725437e-06, "loss": 1.9672, "step": 25823 }, { "epoch": 0.64, "learning_rate": 6.165053973287167e-06, "loss": 2.0538, "step": 25824 }, { "epoch": 0.64, "learning_rate": 6.16431669229026e-06, "loss": 1.9994, "step": 25825 }, { "epoch": 0.64, "learning_rate": 6.1635794357394096e-06, "loss": 1.932, "step": 25826 }, { "epoch": 0.64, "learning_rate": 6.162842203639319e-06, "loss": 1.9302, "step": 25827 }, { "epoch": 0.64, "learning_rate": 6.162104995994685e-06, "loss": 2.0518, "step": 25828 }, { "epoch": 0.64, "learning_rate": 6.161367812810205e-06, "loss": 1.9532, "step": 25829 }, { "epoch": 0.64, "learning_rate": 6.160630654090584e-06, "loss": 2.1638, "step": 25830 }, { "epoch": 0.64, "learning_rate": 6.15989351984051e-06, "loss": 2.1397, "step": 25831 }, { "epoch": 0.64, "learning_rate": 6.159156410064691e-06, "loss": 1.9711, "step": 25832 }, { "epoch": 0.64, "learning_rate": 6.158419324767815e-06, "loss": 1.9139, "step": 25833 }, { "epoch": 0.64, "learning_rate": 6.157682263954585e-06, "loss": 1.9583, "step": 25834 }, { "epoch": 0.64, "learning_rate": 6.1569452276297005e-06, "loss": 1.7897, "step": 25835 }, { "epoch": 0.64, "learning_rate": 6.1562082157978545e-06, "loss": 1.955, "step": 25836 }, { "epoch": 0.64, "learning_rate": 6.15547122846375e-06, "loss": 1.7672, "step": 25837 }, { "epoch": 0.64, "learning_rate": 6.154734265632075e-06, "loss": 2.0572, "step": 25838 }, { "epoch": 0.64, "learning_rate": 6.153997327307532e-06, "loss": 1.9958, "step": 25839 }, { "epoch": 0.64, "learning_rate": 6.153260413494823e-06, "loss": 1.9375, "step": 25840 }, { "epoch": 0.64, "learning_rate": 6.1525235241986345e-06, "loss": 2.029, "step": 25841 }, { "epoch": 0.64, "learning_rate": 6.151786659423668e-06, "loss": 1.8535, "step": 25842 }, { "epoch": 0.64, "learning_rate": 6.151049819174621e-06, "loss": 1.9295, "step": 25843 }, { "epoch": 0.64, "learning_rate": 6.150313003456185e-06, "loss": 2.1114, "step": 25844 }, { "epoch": 0.64, "learning_rate": 6.149576212273061e-06, "loss": 1.927, "step": 25845 }, { "epoch": 0.64, "learning_rate": 6.14883944562994e-06, "loss": 1.9661, "step": 25846 }, { "epoch": 0.64, "learning_rate": 6.148102703531522e-06, "loss": 1.7608, "step": 25847 }, { "epoch": 0.64, "learning_rate": 6.147365985982503e-06, "loss": 1.9481, "step": 25848 }, { "epoch": 0.64, "learning_rate": 6.146629292987572e-06, "loss": 2.0337, "step": 25849 }, { "epoch": 0.64, "learning_rate": 6.145892624551434e-06, "loss": 1.9428, "step": 25850 }, { "epoch": 0.64, "learning_rate": 6.145155980678773e-06, "loss": 1.9647, "step": 25851 }, { "epoch": 0.64, "learning_rate": 6.14441936137429e-06, "loss": 2.0046, "step": 25852 }, { "epoch": 0.64, "learning_rate": 6.14368276664268e-06, "loss": 1.8928, "step": 25853 }, { "epoch": 0.64, "learning_rate": 6.142946196488635e-06, "loss": 1.9265, "step": 25854 }, { "epoch": 0.64, "learning_rate": 6.142209650916855e-06, "loss": 1.9655, "step": 25855 }, { "epoch": 0.64, "learning_rate": 6.141473129932025e-06, "loss": 1.8439, "step": 25856 }, { "epoch": 0.64, "learning_rate": 6.1407366335388465e-06, "loss": 1.939, "step": 25857 }, { "epoch": 0.64, "learning_rate": 6.1400001617420124e-06, "loss": 2.0485, "step": 25858 }, { "epoch": 0.64, "learning_rate": 6.139263714546213e-06, "loss": 2.0556, "step": 25859 }, { "epoch": 0.64, "learning_rate": 6.138527291956146e-06, "loss": 2.0176, "step": 25860 }, { "epoch": 0.64, "learning_rate": 6.137790893976503e-06, "loss": 2.0862, "step": 25861 }, { "epoch": 0.64, "learning_rate": 6.137054520611974e-06, "loss": 2.1101, "step": 25862 }, { "epoch": 0.64, "learning_rate": 6.136318171867259e-06, "loss": 2.0395, "step": 25863 }, { "epoch": 0.64, "learning_rate": 6.135581847747044e-06, "loss": 2.2181, "step": 25864 }, { "epoch": 0.64, "learning_rate": 6.13484554825603e-06, "loss": 2.1086, "step": 25865 }, { "epoch": 0.64, "learning_rate": 6.134109273398901e-06, "loss": 2.0425, "step": 25866 }, { "epoch": 0.64, "learning_rate": 6.133373023180353e-06, "loss": 1.9105, "step": 25867 }, { "epoch": 0.64, "learning_rate": 6.132636797605085e-06, "loss": 1.8611, "step": 25868 }, { "epoch": 0.64, "learning_rate": 6.131900596677777e-06, "loss": 2.0588, "step": 25869 }, { "epoch": 0.64, "learning_rate": 6.131164420403128e-06, "loss": 2.0247, "step": 25870 }, { "epoch": 0.64, "learning_rate": 6.130428268785828e-06, "loss": 2.0232, "step": 25871 }, { "epoch": 0.64, "learning_rate": 6.129692141830571e-06, "loss": 2.0409, "step": 25872 }, { "epoch": 0.64, "learning_rate": 6.1289560395420515e-06, "loss": 2.0822, "step": 25873 }, { "epoch": 0.64, "learning_rate": 6.128219961924951e-06, "loss": 1.8648, "step": 25874 }, { "epoch": 0.64, "learning_rate": 6.127483908983971e-06, "loss": 1.9675, "step": 25875 }, { "epoch": 0.64, "learning_rate": 6.126747880723793e-06, "loss": 2.1068, "step": 25876 }, { "epoch": 0.64, "learning_rate": 6.126011877149114e-06, "loss": 2.1238, "step": 25877 }, { "epoch": 0.64, "learning_rate": 6.125275898264628e-06, "loss": 1.9429, "step": 25878 }, { "epoch": 0.64, "learning_rate": 6.124539944075018e-06, "loss": 2.0051, "step": 25879 }, { "epoch": 0.64, "learning_rate": 6.123804014584979e-06, "loss": 2.0126, "step": 25880 }, { "epoch": 0.64, "learning_rate": 6.123068109799199e-06, "loss": 1.9308, "step": 25881 }, { "epoch": 0.64, "learning_rate": 6.122332229722367e-06, "loss": 2.0571, "step": 25882 }, { "epoch": 0.64, "learning_rate": 6.121596374359182e-06, "loss": 1.993, "step": 25883 }, { "epoch": 0.64, "learning_rate": 6.120860543714322e-06, "loss": 1.9458, "step": 25884 }, { "epoch": 0.64, "learning_rate": 6.120124737792488e-06, "loss": 1.76, "step": 25885 }, { "epoch": 0.64, "learning_rate": 6.119388956598358e-06, "loss": 2.0979, "step": 25886 }, { "epoch": 0.64, "learning_rate": 6.1186532001366264e-06, "loss": 1.9742, "step": 25887 }, { "epoch": 0.64, "learning_rate": 6.117917468411985e-06, "loss": 1.972, "step": 25888 }, { "epoch": 0.64, "learning_rate": 6.117181761429119e-06, "loss": 1.8382, "step": 25889 }, { "epoch": 0.64, "learning_rate": 6.1164460791927216e-06, "loss": 2.0038, "step": 25890 }, { "epoch": 0.64, "learning_rate": 6.115710421707477e-06, "loss": 2.1526, "step": 25891 }, { "epoch": 0.64, "learning_rate": 6.114974788978074e-06, "loss": 1.9159, "step": 25892 }, { "epoch": 0.64, "learning_rate": 6.1142391810092085e-06, "loss": 1.9446, "step": 25893 }, { "epoch": 0.64, "learning_rate": 6.113503597805558e-06, "loss": 1.9301, "step": 25894 }, { "epoch": 0.64, "learning_rate": 6.112768039371818e-06, "loss": 1.8768, "step": 25895 }, { "epoch": 0.64, "learning_rate": 6.112032505712674e-06, "loss": 1.9837, "step": 25896 }, { "epoch": 0.64, "learning_rate": 6.111296996832812e-06, "loss": 2.1293, "step": 25897 }, { "epoch": 0.64, "learning_rate": 6.1105615127369235e-06, "loss": 2.0403, "step": 25898 }, { "epoch": 0.64, "learning_rate": 6.109826053429691e-06, "loss": 1.9466, "step": 25899 }, { "epoch": 0.64, "learning_rate": 6.10909061891581e-06, "loss": 1.9701, "step": 25900 }, { "epoch": 0.64, "learning_rate": 6.1083552091999585e-06, "loss": 1.8304, "step": 25901 }, { "epoch": 0.64, "learning_rate": 6.107619824286828e-06, "loss": 1.9442, "step": 25902 }, { "epoch": 0.64, "learning_rate": 6.106884464181108e-06, "loss": 1.9403, "step": 25903 }, { "epoch": 0.64, "learning_rate": 6.106149128887477e-06, "loss": 2.0838, "step": 25904 }, { "epoch": 0.64, "learning_rate": 6.105413818410628e-06, "loss": 1.9926, "step": 25905 }, { "epoch": 0.64, "learning_rate": 6.104678532755248e-06, "loss": 1.9123, "step": 25906 }, { "epoch": 0.64, "learning_rate": 6.103943271926018e-06, "loss": 2.063, "step": 25907 }, { "epoch": 0.64, "learning_rate": 6.103208035927632e-06, "loss": 2.0676, "step": 25908 }, { "epoch": 0.64, "learning_rate": 6.102472824764767e-06, "loss": 1.996, "step": 25909 }, { "epoch": 0.64, "learning_rate": 6.101737638442114e-06, "loss": 2.0056, "step": 25910 }, { "epoch": 0.64, "learning_rate": 6.1010024769643595e-06, "loss": 1.9263, "step": 25911 }, { "epoch": 0.64, "learning_rate": 6.100267340336184e-06, "loss": 1.9379, "step": 25912 }, { "epoch": 0.64, "learning_rate": 6.099532228562277e-06, "loss": 1.9106, "step": 25913 }, { "epoch": 0.64, "learning_rate": 6.098797141647323e-06, "loss": 1.9869, "step": 25914 }, { "epoch": 0.64, "learning_rate": 6.098062079596004e-06, "loss": 1.8825, "step": 25915 }, { "epoch": 0.64, "learning_rate": 6.0973270424130096e-06, "loss": 2.0253, "step": 25916 }, { "epoch": 0.64, "learning_rate": 6.096592030103019e-06, "loss": 1.9822, "step": 25917 }, { "epoch": 0.64, "learning_rate": 6.095857042670723e-06, "loss": 2.0731, "step": 25918 }, { "epoch": 0.64, "learning_rate": 6.095122080120799e-06, "loss": 1.9589, "step": 25919 }, { "epoch": 0.64, "learning_rate": 6.094387142457934e-06, "loss": 1.9674, "step": 25920 }, { "epoch": 0.64, "learning_rate": 6.0936522296868175e-06, "loss": 1.948, "step": 25921 }, { "epoch": 0.64, "learning_rate": 6.092917341812123e-06, "loss": 1.7708, "step": 25922 }, { "epoch": 0.64, "learning_rate": 6.092182478838543e-06, "loss": 2.0031, "step": 25923 }, { "epoch": 0.64, "learning_rate": 6.091447640770756e-06, "loss": 2.048, "step": 25924 }, { "epoch": 0.64, "learning_rate": 6.090712827613445e-06, "loss": 2.0433, "step": 25925 }, { "epoch": 0.64, "learning_rate": 6.0899780393713e-06, "loss": 1.9105, "step": 25926 }, { "epoch": 0.64, "learning_rate": 6.089243276048995e-06, "loss": 1.9635, "step": 25927 }, { "epoch": 0.64, "learning_rate": 6.088508537651221e-06, "loss": 1.9408, "step": 25928 }, { "epoch": 0.64, "learning_rate": 6.0877738241826545e-06, "loss": 2.0781, "step": 25929 }, { "epoch": 0.64, "learning_rate": 6.087039135647978e-06, "loss": 1.9272, "step": 25930 }, { "epoch": 0.64, "learning_rate": 6.08630447205188e-06, "loss": 2.035, "step": 25931 }, { "epoch": 0.64, "learning_rate": 6.085569833399039e-06, "loss": 2.0232, "step": 25932 }, { "epoch": 0.64, "learning_rate": 6.0848352196941364e-06, "loss": 2.0257, "step": 25933 }, { "epoch": 0.64, "learning_rate": 6.084100630941851e-06, "loss": 1.931, "step": 25934 }, { "epoch": 0.64, "learning_rate": 6.083366067146872e-06, "loss": 1.7735, "step": 25935 }, { "epoch": 0.64, "learning_rate": 6.082631528313879e-06, "loss": 2.0778, "step": 25936 }, { "epoch": 0.64, "learning_rate": 6.081897014447549e-06, "loss": 1.9934, "step": 25937 }, { "epoch": 0.64, "learning_rate": 6.081162525552569e-06, "loss": 1.8892, "step": 25938 }, { "epoch": 0.64, "learning_rate": 6.080428061633616e-06, "loss": 2.1488, "step": 25939 }, { "epoch": 0.64, "learning_rate": 6.0796936226953695e-06, "loss": 2.1029, "step": 25940 }, { "epoch": 0.64, "learning_rate": 6.078959208742517e-06, "loss": 1.878, "step": 25941 }, { "epoch": 0.64, "learning_rate": 6.078224819779732e-06, "loss": 2.0935, "step": 25942 }, { "epoch": 0.64, "learning_rate": 6.077490455811703e-06, "loss": 1.8892, "step": 25943 }, { "epoch": 0.64, "learning_rate": 6.0767561168431015e-06, "loss": 2.049, "step": 25944 }, { "epoch": 0.64, "learning_rate": 6.076021802878611e-06, "loss": 2.0179, "step": 25945 }, { "epoch": 0.64, "learning_rate": 6.075287513922919e-06, "loss": 2.0373, "step": 25946 }, { "epoch": 0.64, "learning_rate": 6.074553249980692e-06, "loss": 1.8908, "step": 25947 }, { "epoch": 0.64, "learning_rate": 6.073819011056619e-06, "loss": 1.8968, "step": 25948 }, { "epoch": 0.64, "learning_rate": 6.073084797155376e-06, "loss": 2.1202, "step": 25949 }, { "epoch": 0.64, "learning_rate": 6.072350608281643e-06, "loss": 1.9413, "step": 25950 }, { "epoch": 0.64, "learning_rate": 6.0716164444401e-06, "loss": 2.0624, "step": 25951 }, { "epoch": 0.64, "learning_rate": 6.070882305635426e-06, "loss": 1.8388, "step": 25952 }, { "epoch": 0.64, "learning_rate": 6.070148191872303e-06, "loss": 2.0476, "step": 25953 }, { "epoch": 0.64, "learning_rate": 6.0694141031554e-06, "loss": 2.1654, "step": 25954 }, { "epoch": 0.64, "learning_rate": 6.068680039489404e-06, "loss": 2.0265, "step": 25955 }, { "epoch": 0.64, "learning_rate": 6.067946000878995e-06, "loss": 2.0242, "step": 25956 }, { "epoch": 0.64, "learning_rate": 6.067211987328842e-06, "loss": 2.2768, "step": 25957 }, { "epoch": 0.64, "learning_rate": 6.0664779988436316e-06, "loss": 2.1427, "step": 25958 }, { "epoch": 0.64, "learning_rate": 6.065744035428039e-06, "loss": 1.8284, "step": 25959 }, { "epoch": 0.64, "learning_rate": 6.065010097086739e-06, "loss": 1.9605, "step": 25960 }, { "epoch": 0.64, "learning_rate": 6.0642761838244165e-06, "loss": 1.9433, "step": 25961 }, { "epoch": 0.64, "learning_rate": 6.063542295645741e-06, "loss": 1.9306, "step": 25962 }, { "epoch": 0.64, "learning_rate": 6.0628084325553935e-06, "loss": 2.0452, "step": 25963 }, { "epoch": 0.64, "learning_rate": 6.062074594558054e-06, "loss": 1.9398, "step": 25964 }, { "epoch": 0.64, "learning_rate": 6.061340781658394e-06, "loss": 1.9741, "step": 25965 }, { "epoch": 0.64, "learning_rate": 6.060606993861093e-06, "loss": 1.9162, "step": 25966 }, { "epoch": 0.64, "learning_rate": 6.059873231170829e-06, "loss": 1.9374, "step": 25967 }, { "epoch": 0.64, "learning_rate": 6.059139493592274e-06, "loss": 2.1132, "step": 25968 }, { "epoch": 0.64, "learning_rate": 6.05840578113011e-06, "loss": 1.7729, "step": 25969 }, { "epoch": 0.64, "learning_rate": 6.057672093789007e-06, "loss": 1.8785, "step": 25970 }, { "epoch": 0.64, "learning_rate": 6.056938431573649e-06, "loss": 1.9622, "step": 25971 }, { "epoch": 0.64, "learning_rate": 6.056204794488705e-06, "loss": 1.8708, "step": 25972 }, { "epoch": 0.64, "learning_rate": 6.0554711825388524e-06, "loss": 2.1967, "step": 25973 }, { "epoch": 0.64, "learning_rate": 6.05473759572877e-06, "loss": 1.974, "step": 25974 }, { "epoch": 0.64, "learning_rate": 6.054004034063128e-06, "loss": 2.0939, "step": 25975 }, { "epoch": 0.64, "learning_rate": 6.053270497546606e-06, "loss": 1.8475, "step": 25976 }, { "epoch": 0.64, "learning_rate": 6.052536986183874e-06, "loss": 1.9752, "step": 25977 }, { "epoch": 0.64, "learning_rate": 6.0518034999796126e-06, "loss": 2.0005, "step": 25978 }, { "epoch": 0.64, "learning_rate": 6.051070038938497e-06, "loss": 2.0449, "step": 25979 }, { "epoch": 0.64, "learning_rate": 6.050336603065195e-06, "loss": 1.983, "step": 25980 }, { "epoch": 0.64, "learning_rate": 6.049603192364389e-06, "loss": 2.0827, "step": 25981 }, { "epoch": 0.64, "learning_rate": 6.048869806840745e-06, "loss": 2.1777, "step": 25982 }, { "epoch": 0.64, "learning_rate": 6.048136446498942e-06, "loss": 2.072, "step": 25983 }, { "epoch": 0.64, "learning_rate": 6.047403111343655e-06, "loss": 2.0382, "step": 25984 }, { "epoch": 0.64, "learning_rate": 6.0466698013795556e-06, "loss": 2.0299, "step": 25985 }, { "epoch": 0.64, "learning_rate": 6.045936516611318e-06, "loss": 1.9957, "step": 25986 }, { "epoch": 0.64, "learning_rate": 6.045203257043615e-06, "loss": 1.9848, "step": 25987 }, { "epoch": 0.64, "learning_rate": 6.044470022681119e-06, "loss": 2.0304, "step": 25988 }, { "epoch": 0.64, "learning_rate": 6.0437368135285095e-06, "loss": 2.1211, "step": 25989 }, { "epoch": 0.64, "learning_rate": 6.04300362959045e-06, "loss": 2.022, "step": 25990 }, { "epoch": 0.64, "learning_rate": 6.042270470871621e-06, "loss": 1.9137, "step": 25991 }, { "epoch": 0.64, "learning_rate": 6.041537337376688e-06, "loss": 1.9054, "step": 25992 }, { "epoch": 0.64, "learning_rate": 6.04080422911033e-06, "loss": 1.9016, "step": 25993 }, { "epoch": 0.64, "learning_rate": 6.040071146077217e-06, "loss": 1.9069, "step": 25994 }, { "epoch": 0.64, "learning_rate": 6.0393380882820205e-06, "loss": 2.1007, "step": 25995 }, { "epoch": 0.64, "learning_rate": 6.038605055729416e-06, "loss": 2.0221, "step": 25996 }, { "epoch": 0.64, "learning_rate": 6.037872048424068e-06, "loss": 2.0516, "step": 25997 }, { "epoch": 0.64, "learning_rate": 6.037139066370654e-06, "loss": 1.7215, "step": 25998 }, { "epoch": 0.64, "learning_rate": 6.0364061095738466e-06, "loss": 1.853, "step": 25999 }, { "epoch": 0.64, "learning_rate": 6.0356731780383126e-06, "loss": 2.0963, "step": 26000 }, { "epoch": 0.64, "learning_rate": 6.034940271768725e-06, "loss": 1.9426, "step": 26001 }, { "epoch": 0.64, "learning_rate": 6.034207390769755e-06, "loss": 1.9774, "step": 26002 }, { "epoch": 0.64, "learning_rate": 6.033474535046076e-06, "loss": 1.9888, "step": 26003 }, { "epoch": 0.64, "learning_rate": 6.032741704602356e-06, "loss": 2.1531, "step": 26004 }, { "epoch": 0.64, "learning_rate": 6.032008899443264e-06, "loss": 2.0518, "step": 26005 }, { "epoch": 0.64, "learning_rate": 6.031276119573477e-06, "loss": 2.2406, "step": 26006 }, { "epoch": 0.64, "learning_rate": 6.030543364997655e-06, "loss": 2.0496, "step": 26007 }, { "epoch": 0.64, "learning_rate": 6.029810635720474e-06, "loss": 1.985, "step": 26008 }, { "epoch": 0.64, "learning_rate": 6.02907793174661e-06, "loss": 1.956, "step": 26009 }, { "epoch": 0.64, "learning_rate": 6.028345253080721e-06, "loss": 1.9343, "step": 26010 }, { "epoch": 0.64, "learning_rate": 6.027612599727483e-06, "loss": 2.0003, "step": 26011 }, { "epoch": 0.64, "learning_rate": 6.026879971691564e-06, "loss": 1.8128, "step": 26012 }, { "epoch": 0.64, "learning_rate": 6.0261473689776315e-06, "loss": 1.9033, "step": 26013 }, { "epoch": 0.64, "learning_rate": 6.025414791590362e-06, "loss": 1.7928, "step": 26014 }, { "epoch": 0.64, "learning_rate": 6.0246822395344165e-06, "loss": 2.0118, "step": 26015 }, { "epoch": 0.64, "learning_rate": 6.0239497128144696e-06, "loss": 1.963, "step": 26016 }, { "epoch": 0.64, "learning_rate": 6.0232172114351836e-06, "loss": 1.9166, "step": 26017 }, { "epoch": 0.64, "learning_rate": 6.022484735401229e-06, "loss": 2.0086, "step": 26018 }, { "epoch": 0.64, "learning_rate": 6.021752284717279e-06, "loss": 1.9845, "step": 26019 }, { "epoch": 0.64, "learning_rate": 6.021019859387995e-06, "loss": 2.0123, "step": 26020 }, { "epoch": 0.64, "learning_rate": 6.020287459418051e-06, "loss": 2.006, "step": 26021 }, { "epoch": 0.64, "learning_rate": 6.019555084812112e-06, "loss": 1.93, "step": 26022 }, { "epoch": 0.64, "learning_rate": 6.018822735574843e-06, "loss": 1.8873, "step": 26023 }, { "epoch": 0.64, "learning_rate": 6.018090411710917e-06, "loss": 1.8552, "step": 26024 }, { "epoch": 0.64, "learning_rate": 6.017358113224996e-06, "loss": 1.9442, "step": 26025 }, { "epoch": 0.64, "learning_rate": 6.016625840121751e-06, "loss": 1.9306, "step": 26026 }, { "epoch": 0.64, "learning_rate": 6.015893592405849e-06, "loss": 2.0503, "step": 26027 }, { "epoch": 0.64, "learning_rate": 6.015161370081952e-06, "loss": 1.954, "step": 26028 }, { "epoch": 0.64, "learning_rate": 6.014429173154734e-06, "loss": 1.8715, "step": 26029 }, { "epoch": 0.64, "learning_rate": 6.0136970016288535e-06, "loss": 1.8743, "step": 26030 }, { "epoch": 0.64, "learning_rate": 6.012964855508983e-06, "loss": 2.1177, "step": 26031 }, { "epoch": 0.64, "learning_rate": 6.01223273479979e-06, "loss": 2.1104, "step": 26032 }, { "epoch": 0.64, "learning_rate": 6.011500639505933e-06, "loss": 1.9177, "step": 26033 }, { "epoch": 0.64, "learning_rate": 6.010768569632086e-06, "loss": 1.8192, "step": 26034 }, { "epoch": 0.64, "learning_rate": 6.010036525182908e-06, "loss": 2.0143, "step": 26035 }, { "epoch": 0.64, "learning_rate": 6.009304506163068e-06, "loss": 2.0098, "step": 26036 }, { "epoch": 0.64, "learning_rate": 6.008572512577231e-06, "loss": 2.0708, "step": 26037 }, { "epoch": 0.64, "learning_rate": 6.0078405444300615e-06, "loss": 2.0671, "step": 26038 }, { "epoch": 0.64, "learning_rate": 6.007108601726228e-06, "loss": 2.1652, "step": 26039 }, { "epoch": 0.64, "learning_rate": 6.00637668447039e-06, "loss": 1.9595, "step": 26040 }, { "epoch": 0.64, "learning_rate": 6.005644792667213e-06, "loss": 1.9974, "step": 26041 }, { "epoch": 0.64, "learning_rate": 6.00491292632137e-06, "loss": 1.8581, "step": 26042 }, { "epoch": 0.64, "learning_rate": 6.004181085437514e-06, "loss": 1.9278, "step": 26043 }, { "epoch": 0.64, "learning_rate": 6.003449270020317e-06, "loss": 2.0337, "step": 26044 }, { "epoch": 0.64, "learning_rate": 6.002717480074438e-06, "loss": 2.1668, "step": 26045 }, { "epoch": 0.64, "learning_rate": 6.001985715604541e-06, "loss": 2.0618, "step": 26046 }, { "epoch": 0.64, "learning_rate": 6.001253976615295e-06, "loss": 1.941, "step": 26047 }, { "epoch": 0.64, "learning_rate": 6.000522263111359e-06, "loss": 1.8424, "step": 26048 }, { "epoch": 0.64, "learning_rate": 5.999790575097402e-06, "loss": 1.9596, "step": 26049 }, { "epoch": 0.64, "learning_rate": 5.9990589125780796e-06, "loss": 1.8524, "step": 26050 }, { "epoch": 0.64, "learning_rate": 5.998327275558058e-06, "loss": 1.992, "step": 26051 }, { "epoch": 0.64, "learning_rate": 5.997595664042004e-06, "loss": 1.9631, "step": 26052 }, { "epoch": 0.64, "learning_rate": 5.996864078034575e-06, "loss": 1.8666, "step": 26053 }, { "epoch": 0.64, "learning_rate": 5.996132517540436e-06, "loss": 1.9466, "step": 26054 }, { "epoch": 0.64, "learning_rate": 5.995400982564248e-06, "loss": 1.9763, "step": 26055 }, { "epoch": 0.64, "learning_rate": 5.994669473110675e-06, "loss": 2.1871, "step": 26056 }, { "epoch": 0.64, "learning_rate": 5.993937989184383e-06, "loss": 1.8608, "step": 26057 }, { "epoch": 0.64, "learning_rate": 5.993206530790025e-06, "loss": 1.9718, "step": 26058 }, { "epoch": 0.64, "learning_rate": 5.992475097932271e-06, "loss": 1.9625, "step": 26059 }, { "epoch": 0.64, "learning_rate": 5.9917436906157765e-06, "loss": 1.8869, "step": 26060 }, { "epoch": 0.64, "learning_rate": 5.991012308845205e-06, "loss": 1.9943, "step": 26061 }, { "epoch": 0.64, "learning_rate": 5.9902809526252225e-06, "loss": 1.9354, "step": 26062 }, { "epoch": 0.64, "learning_rate": 5.989549621960483e-06, "loss": 1.9411, "step": 26063 }, { "epoch": 0.64, "learning_rate": 5.9888183168556525e-06, "loss": 1.9483, "step": 26064 }, { "epoch": 0.64, "learning_rate": 5.988087037315387e-06, "loss": 1.9443, "step": 26065 }, { "epoch": 0.64, "learning_rate": 5.987355783344351e-06, "loss": 1.9405, "step": 26066 }, { "epoch": 0.64, "learning_rate": 5.986624554947209e-06, "loss": 1.9133, "step": 26067 }, { "epoch": 0.64, "learning_rate": 5.985893352128611e-06, "loss": 2.026, "step": 26068 }, { "epoch": 0.64, "learning_rate": 5.9851621748932285e-06, "loss": 1.9987, "step": 26069 }, { "epoch": 0.64, "learning_rate": 5.984431023245712e-06, "loss": 2.0603, "step": 26070 }, { "epoch": 0.64, "learning_rate": 5.983699897190724e-06, "loss": 2.0275, "step": 26071 }, { "epoch": 0.64, "learning_rate": 5.982968796732927e-06, "loss": 1.9999, "step": 26072 }, { "epoch": 0.64, "learning_rate": 5.982237721876978e-06, "loss": 2.1014, "step": 26073 }, { "epoch": 0.64, "learning_rate": 5.981506672627542e-06, "loss": 2.0345, "step": 26074 }, { "epoch": 0.64, "learning_rate": 5.980775648989268e-06, "loss": 2.105, "step": 26075 }, { "epoch": 0.64, "learning_rate": 5.980044650966821e-06, "loss": 2.009, "step": 26076 }, { "epoch": 0.64, "learning_rate": 5.979313678564865e-06, "loss": 2.1254, "step": 26077 }, { "epoch": 0.64, "learning_rate": 5.978582731788046e-06, "loss": 1.9343, "step": 26078 }, { "epoch": 0.64, "learning_rate": 5.9778518106410365e-06, "loss": 2.0397, "step": 26079 }, { "epoch": 0.64, "learning_rate": 5.977120915128483e-06, "loss": 1.9451, "step": 26080 }, { "epoch": 0.64, "learning_rate": 5.9763900452550495e-06, "loss": 1.9186, "step": 26081 }, { "epoch": 0.64, "learning_rate": 5.975659201025396e-06, "loss": 2.1253, "step": 26082 }, { "epoch": 0.64, "learning_rate": 5.974928382444175e-06, "loss": 2.0264, "step": 26083 }, { "epoch": 0.64, "learning_rate": 5.974197589516047e-06, "loss": 1.9422, "step": 26084 }, { "epoch": 0.64, "learning_rate": 5.973466822245675e-06, "loss": 1.9214, "step": 26085 }, { "epoch": 0.64, "learning_rate": 5.9727360806377066e-06, "loss": 1.7497, "step": 26086 }, { "epoch": 0.64, "learning_rate": 5.972005364696807e-06, "loss": 2.0375, "step": 26087 }, { "epoch": 0.64, "learning_rate": 5.971274674427625e-06, "loss": 1.821, "step": 26088 }, { "epoch": 0.64, "learning_rate": 5.970544009834824e-06, "loss": 2.0254, "step": 26089 }, { "epoch": 0.64, "learning_rate": 5.969813370923061e-06, "loss": 2.0161, "step": 26090 }, { "epoch": 0.64, "learning_rate": 5.969082757696989e-06, "loss": 2.0515, "step": 26091 }, { "epoch": 0.64, "learning_rate": 5.96835217016127e-06, "loss": 1.9691, "step": 26092 }, { "epoch": 0.64, "learning_rate": 5.967621608320553e-06, "loss": 1.957, "step": 26093 }, { "epoch": 0.64, "learning_rate": 5.966891072179496e-06, "loss": 1.8622, "step": 26094 }, { "epoch": 0.64, "learning_rate": 5.966160561742762e-06, "loss": 1.972, "step": 26095 }, { "epoch": 0.64, "learning_rate": 5.965430077014996e-06, "loss": 1.967, "step": 26096 }, { "epoch": 0.64, "learning_rate": 5.964699618000864e-06, "loss": 1.8873, "step": 26097 }, { "epoch": 0.64, "learning_rate": 5.963969184705012e-06, "loss": 1.9128, "step": 26098 }, { "epoch": 0.64, "learning_rate": 5.963238777132101e-06, "loss": 1.7489, "step": 26099 }, { "epoch": 0.64, "learning_rate": 5.962508395286786e-06, "loss": 1.9011, "step": 26100 }, { "epoch": 0.64, "learning_rate": 5.961778039173718e-06, "loss": 2.1327, "step": 26101 }, { "epoch": 0.64, "learning_rate": 5.961047708797559e-06, "loss": 2.1044, "step": 26102 }, { "epoch": 0.64, "learning_rate": 5.960317404162954e-06, "loss": 2.0226, "step": 26103 }, { "epoch": 0.64, "learning_rate": 5.959587125274564e-06, "loss": 1.9308, "step": 26104 }, { "epoch": 0.64, "learning_rate": 5.958856872137047e-06, "loss": 1.9578, "step": 26105 }, { "epoch": 0.64, "learning_rate": 5.958126644755047e-06, "loss": 1.8734, "step": 26106 }, { "epoch": 0.64, "learning_rate": 5.957396443133225e-06, "loss": 1.9864, "step": 26107 }, { "epoch": 0.64, "learning_rate": 5.956666267276231e-06, "loss": 1.9378, "step": 26108 }, { "epoch": 0.64, "learning_rate": 5.955936117188721e-06, "loss": 1.9682, "step": 26109 }, { "epoch": 0.64, "learning_rate": 5.955205992875352e-06, "loss": 1.8683, "step": 26110 }, { "epoch": 0.64, "learning_rate": 5.954475894340768e-06, "loss": 2.0678, "step": 26111 }, { "epoch": 0.64, "learning_rate": 5.953745821589633e-06, "loss": 2.0149, "step": 26112 }, { "epoch": 0.64, "learning_rate": 5.9530157746265895e-06, "loss": 2.0347, "step": 26113 }, { "epoch": 0.64, "learning_rate": 5.952285753456296e-06, "loss": 2.0328, "step": 26114 }, { "epoch": 0.64, "learning_rate": 5.95155575808341e-06, "loss": 1.8606, "step": 26115 }, { "epoch": 0.64, "learning_rate": 5.950825788512572e-06, "loss": 1.9844, "step": 26116 }, { "epoch": 0.64, "learning_rate": 5.950095844748444e-06, "loss": 1.9514, "step": 26117 }, { "epoch": 0.64, "learning_rate": 5.949365926795672e-06, "loss": 2.0203, "step": 26118 }, { "epoch": 0.64, "learning_rate": 5.948636034658912e-06, "loss": 1.9783, "step": 26119 }, { "epoch": 0.64, "learning_rate": 5.9479061683428186e-06, "loss": 2.0225, "step": 26120 }, { "epoch": 0.64, "learning_rate": 5.947176327852035e-06, "loss": 1.9853, "step": 26121 }, { "epoch": 0.64, "learning_rate": 5.946446513191223e-06, "loss": 1.8705, "step": 26122 }, { "epoch": 0.64, "learning_rate": 5.9457167243650235e-06, "loss": 1.8976, "step": 26123 }, { "epoch": 0.64, "learning_rate": 5.944986961378093e-06, "loss": 2.1074, "step": 26124 }, { "epoch": 0.64, "learning_rate": 5.944257224235084e-06, "loss": 1.9666, "step": 26125 }, { "epoch": 0.64, "learning_rate": 5.943527512940643e-06, "loss": 2.0147, "step": 26126 }, { "epoch": 0.64, "learning_rate": 5.942797827499427e-06, "loss": 1.8924, "step": 26127 }, { "epoch": 0.64, "learning_rate": 5.942068167916079e-06, "loss": 2.0032, "step": 26128 }, { "epoch": 0.64, "learning_rate": 5.941338534195253e-06, "loss": 2.0867, "step": 26129 }, { "epoch": 0.64, "learning_rate": 5.940608926341603e-06, "loss": 2.1364, "step": 26130 }, { "epoch": 0.64, "learning_rate": 5.9398793443597714e-06, "loss": 1.991, "step": 26131 }, { "epoch": 0.64, "learning_rate": 5.939149788254415e-06, "loss": 2.0232, "step": 26132 }, { "epoch": 0.64, "learning_rate": 5.938420258030178e-06, "loss": 1.8286, "step": 26133 }, { "epoch": 0.64, "learning_rate": 5.937690753691711e-06, "loss": 2.0748, "step": 26134 }, { "epoch": 0.64, "learning_rate": 5.936961275243669e-06, "loss": 2.0179, "step": 26135 }, { "epoch": 0.64, "learning_rate": 5.936231822690692e-06, "loss": 2.0693, "step": 26136 }, { "epoch": 0.64, "learning_rate": 5.9355023960374405e-06, "loss": 1.8374, "step": 26137 }, { "epoch": 0.64, "learning_rate": 5.934772995288551e-06, "loss": 1.9666, "step": 26138 }, { "epoch": 0.64, "learning_rate": 5.934043620448677e-06, "loss": 1.9712, "step": 26139 }, { "epoch": 0.64, "learning_rate": 5.933314271522475e-06, "loss": 1.9754, "step": 26140 }, { "epoch": 0.64, "learning_rate": 5.932584948514582e-06, "loss": 2.0019, "step": 26141 }, { "epoch": 0.64, "learning_rate": 5.931855651429647e-06, "loss": 2.0542, "step": 26142 }, { "epoch": 0.64, "learning_rate": 5.931126380272327e-06, "loss": 1.9663, "step": 26143 }, { "epoch": 0.64, "learning_rate": 5.930397135047263e-06, "loss": 2.1668, "step": 26144 }, { "epoch": 0.64, "learning_rate": 5.9296679157591075e-06, "loss": 2.0493, "step": 26145 }, { "epoch": 0.64, "learning_rate": 5.9289387224125006e-06, "loss": 1.997, "step": 26146 }, { "epoch": 0.64, "learning_rate": 5.928209555012093e-06, "loss": 2.16, "step": 26147 }, { "epoch": 0.64, "learning_rate": 5.927480413562539e-06, "loss": 1.955, "step": 26148 }, { "epoch": 0.64, "learning_rate": 5.926751298068474e-06, "loss": 1.9372, "step": 26149 }, { "epoch": 0.64, "learning_rate": 5.926022208534552e-06, "loss": 2.0105, "step": 26150 }, { "epoch": 0.64, "learning_rate": 5.925293144965418e-06, "loss": 2.0334, "step": 26151 }, { "epoch": 0.64, "learning_rate": 5.924564107365718e-06, "loss": 1.8309, "step": 26152 }, { "epoch": 0.64, "learning_rate": 5.9238350957401e-06, "loss": 2.005, "step": 26153 }, { "epoch": 0.64, "learning_rate": 5.923106110093209e-06, "loss": 1.8914, "step": 26154 }, { "epoch": 0.64, "learning_rate": 5.9223771504296945e-06, "loss": 1.8785, "step": 26155 }, { "epoch": 0.64, "learning_rate": 5.9216482167541945e-06, "loss": 1.8848, "step": 26156 }, { "epoch": 0.64, "learning_rate": 5.92091930907136e-06, "loss": 2.0931, "step": 26157 }, { "epoch": 0.64, "learning_rate": 5.920190427385841e-06, "loss": 1.9379, "step": 26158 }, { "epoch": 0.64, "learning_rate": 5.919461571702273e-06, "loss": 1.9859, "step": 26159 }, { "epoch": 0.64, "learning_rate": 5.918732742025308e-06, "loss": 1.8727, "step": 26160 }, { "epoch": 0.64, "learning_rate": 5.918003938359589e-06, "loss": 1.9122, "step": 26161 }, { "epoch": 0.64, "learning_rate": 5.917275160709759e-06, "loss": 1.7017, "step": 26162 }, { "epoch": 0.64, "learning_rate": 5.91654640908047e-06, "loss": 1.9047, "step": 26163 }, { "epoch": 0.64, "learning_rate": 5.915817683476357e-06, "loss": 2.1735, "step": 26164 }, { "epoch": 0.64, "learning_rate": 5.915088983902073e-06, "loss": 2.1569, "step": 26165 }, { "epoch": 0.64, "learning_rate": 5.914360310362255e-06, "loss": 1.9593, "step": 26166 }, { "epoch": 0.64, "learning_rate": 5.91363166286155e-06, "loss": 1.9992, "step": 26167 }, { "epoch": 0.65, "learning_rate": 5.912903041404605e-06, "loss": 1.8, "step": 26168 }, { "epoch": 0.65, "learning_rate": 5.912174445996058e-06, "loss": 1.9069, "step": 26169 }, { "epoch": 0.65, "learning_rate": 5.911445876640556e-06, "loss": 2.0339, "step": 26170 }, { "epoch": 0.65, "learning_rate": 5.910717333342742e-06, "loss": 1.8671, "step": 26171 }, { "epoch": 0.65, "learning_rate": 5.9099888161072575e-06, "loss": 2.0287, "step": 26172 }, { "epoch": 0.65, "learning_rate": 5.909260324938751e-06, "loss": 1.9627, "step": 26173 }, { "epoch": 0.65, "learning_rate": 5.908531859841858e-06, "loss": 1.9653, "step": 26174 }, { "epoch": 0.65, "learning_rate": 5.9078034208212295e-06, "loss": 1.9422, "step": 26175 }, { "epoch": 0.65, "learning_rate": 5.907075007881499e-06, "loss": 2.03, "step": 26176 }, { "epoch": 0.65, "learning_rate": 5.906346621027313e-06, "loss": 1.8029, "step": 26177 }, { "epoch": 0.65, "learning_rate": 5.905618260263316e-06, "loss": 2.103, "step": 26178 }, { "epoch": 0.65, "learning_rate": 5.904889925594146e-06, "loss": 1.8646, "step": 26179 }, { "epoch": 0.65, "learning_rate": 5.904161617024452e-06, "loss": 1.8278, "step": 26180 }, { "epoch": 0.65, "learning_rate": 5.9034333345588665e-06, "loss": 1.9996, "step": 26181 }, { "epoch": 0.65, "learning_rate": 5.902705078202035e-06, "loss": 1.9345, "step": 26182 }, { "epoch": 0.65, "learning_rate": 5.901976847958604e-06, "loss": 2.0843, "step": 26183 }, { "epoch": 0.65, "learning_rate": 5.901248643833206e-06, "loss": 1.9383, "step": 26184 }, { "epoch": 0.65, "learning_rate": 5.900520465830487e-06, "loss": 2.0406, "step": 26185 }, { "epoch": 0.65, "learning_rate": 5.899792313955088e-06, "loss": 2.0544, "step": 26186 }, { "epoch": 0.65, "learning_rate": 5.8990641882116465e-06, "loss": 1.867, "step": 26187 }, { "epoch": 0.65, "learning_rate": 5.898336088604808e-06, "loss": 1.9979, "step": 26188 }, { "epoch": 0.65, "learning_rate": 5.8976080151392065e-06, "loss": 1.9196, "step": 26189 }, { "epoch": 0.65, "learning_rate": 5.896879967819492e-06, "loss": 1.855, "step": 26190 }, { "epoch": 0.65, "learning_rate": 5.896151946650294e-06, "loss": 2.0626, "step": 26191 }, { "epoch": 0.65, "learning_rate": 5.895423951636257e-06, "loss": 1.999, "step": 26192 }, { "epoch": 0.65, "learning_rate": 5.894695982782024e-06, "loss": 2.2743, "step": 26193 }, { "epoch": 0.65, "learning_rate": 5.893968040092228e-06, "loss": 1.932, "step": 26194 }, { "epoch": 0.65, "learning_rate": 5.893240123571516e-06, "loss": 1.936, "step": 26195 }, { "epoch": 0.65, "learning_rate": 5.892512233224518e-06, "loss": 1.9215, "step": 26196 }, { "epoch": 0.65, "learning_rate": 5.891784369055878e-06, "loss": 2.1629, "step": 26197 }, { "epoch": 0.65, "learning_rate": 5.89105653107024e-06, "loss": 1.9247, "step": 26198 }, { "epoch": 0.65, "learning_rate": 5.8903287192722345e-06, "loss": 1.8843, "step": 26199 }, { "epoch": 0.65, "learning_rate": 5.889600933666504e-06, "loss": 1.9158, "step": 26200 }, { "epoch": 0.65, "learning_rate": 5.888873174257688e-06, "loss": 2.0003, "step": 26201 }, { "epoch": 0.65, "learning_rate": 5.888145441050421e-06, "loss": 2.0625, "step": 26202 }, { "epoch": 0.65, "learning_rate": 5.887417734049343e-06, "loss": 1.9125, "step": 26203 }, { "epoch": 0.65, "learning_rate": 5.886690053259094e-06, "loss": 2.0502, "step": 26204 }, { "epoch": 0.65, "learning_rate": 5.885962398684307e-06, "loss": 1.9253, "step": 26205 }, { "epoch": 0.65, "learning_rate": 5.885234770329624e-06, "loss": 2.0062, "step": 26206 }, { "epoch": 0.65, "learning_rate": 5.88450716819968e-06, "loss": 1.9735, "step": 26207 }, { "epoch": 0.65, "learning_rate": 5.883779592299117e-06, "loss": 1.9608, "step": 26208 }, { "epoch": 0.65, "learning_rate": 5.8830520426325624e-06, "loss": 1.9192, "step": 26209 }, { "epoch": 0.65, "learning_rate": 5.882324519204661e-06, "loss": 1.7992, "step": 26210 }, { "epoch": 0.65, "learning_rate": 5.881597022020049e-06, "loss": 2.047, "step": 26211 }, { "epoch": 0.65, "learning_rate": 5.880869551083359e-06, "loss": 1.9108, "step": 26212 }, { "epoch": 0.65, "learning_rate": 5.880142106399231e-06, "loss": 2.0123, "step": 26213 }, { "epoch": 0.65, "learning_rate": 5.879414687972298e-06, "loss": 1.8122, "step": 26214 }, { "epoch": 0.65, "learning_rate": 5.878687295807197e-06, "loss": 1.8793, "step": 26215 }, { "epoch": 0.65, "learning_rate": 5.877959929908571e-06, "loss": 2.0331, "step": 26216 }, { "epoch": 0.65, "learning_rate": 5.877232590281046e-06, "loss": 1.9347, "step": 26217 }, { "epoch": 0.65, "learning_rate": 5.876505276929263e-06, "loss": 2.0887, "step": 26218 }, { "epoch": 0.65, "learning_rate": 5.875777989857854e-06, "loss": 1.8648, "step": 26219 }, { "epoch": 0.65, "learning_rate": 5.875050729071454e-06, "loss": 1.9009, "step": 26220 }, { "epoch": 0.65, "learning_rate": 5.874323494574703e-06, "loss": 2.1548, "step": 26221 }, { "epoch": 0.65, "learning_rate": 5.8735962863722314e-06, "loss": 1.933, "step": 26222 }, { "epoch": 0.65, "learning_rate": 5.872869104468677e-06, "loss": 1.8717, "step": 26223 }, { "epoch": 0.65, "learning_rate": 5.872141948868671e-06, "loss": 1.9806, "step": 26224 }, { "epoch": 0.65, "learning_rate": 5.871414819576848e-06, "loss": 1.6634, "step": 26225 }, { "epoch": 0.65, "learning_rate": 5.87068771659785e-06, "loss": 2.0109, "step": 26226 }, { "epoch": 0.65, "learning_rate": 5.869960639936298e-06, "loss": 1.9987, "step": 26227 }, { "epoch": 0.65, "learning_rate": 5.86923358959684e-06, "loss": 2.0061, "step": 26228 }, { "epoch": 0.65, "learning_rate": 5.868506565584097e-06, "loss": 2.0174, "step": 26229 }, { "epoch": 0.65, "learning_rate": 5.867779567902707e-06, "loss": 1.9492, "step": 26230 }, { "epoch": 0.65, "learning_rate": 5.8670525965573076e-06, "loss": 2.0065, "step": 26231 }, { "epoch": 0.65, "learning_rate": 5.866325651552528e-06, "loss": 2.0527, "step": 26232 }, { "epoch": 0.65, "learning_rate": 5.865598732893005e-06, "loss": 2.1301, "step": 26233 }, { "epoch": 0.65, "learning_rate": 5.864871840583364e-06, "loss": 1.9507, "step": 26234 }, { "epoch": 0.65, "learning_rate": 5.864144974628243e-06, "loss": 1.9464, "step": 26235 }, { "epoch": 0.65, "learning_rate": 5.863418135032279e-06, "loss": 2.0011, "step": 26236 }, { "epoch": 0.65, "learning_rate": 5.862691321800094e-06, "loss": 1.7823, "step": 26237 }, { "epoch": 0.65, "learning_rate": 5.8619645349363266e-06, "loss": 1.8891, "step": 26238 }, { "epoch": 0.65, "learning_rate": 5.861237774445607e-06, "loss": 1.9194, "step": 26239 }, { "epoch": 0.65, "learning_rate": 5.860511040332571e-06, "loss": 2.0835, "step": 26240 }, { "epoch": 0.65, "learning_rate": 5.859784332601845e-06, "loss": 2.0565, "step": 26241 }, { "epoch": 0.65, "learning_rate": 5.859057651258062e-06, "loss": 2.1229, "step": 26242 }, { "epoch": 0.65, "learning_rate": 5.8583309963058565e-06, "loss": 1.9963, "step": 26243 }, { "epoch": 0.65, "learning_rate": 5.857604367749855e-06, "loss": 1.9843, "step": 26244 }, { "epoch": 0.65, "learning_rate": 5.85687776559469e-06, "loss": 1.8666, "step": 26245 }, { "epoch": 0.65, "learning_rate": 5.856151189844998e-06, "loss": 1.9901, "step": 26246 }, { "epoch": 0.65, "learning_rate": 5.8554246405054e-06, "loss": 1.9647, "step": 26247 }, { "epoch": 0.65, "learning_rate": 5.854698117580535e-06, "loss": 2.0208, "step": 26248 }, { "epoch": 0.65, "learning_rate": 5.853971621075026e-06, "loss": 1.9927, "step": 26249 }, { "epoch": 0.65, "learning_rate": 5.853245150993508e-06, "loss": 1.9217, "step": 26250 }, { "epoch": 0.65, "learning_rate": 5.852518707340613e-06, "loss": 1.9725, "step": 26251 }, { "epoch": 0.65, "learning_rate": 5.851792290120966e-06, "loss": 2.11, "step": 26252 }, { "epoch": 0.65, "learning_rate": 5.851065899339201e-06, "loss": 2.0291, "step": 26253 }, { "epoch": 0.65, "learning_rate": 5.85033953499994e-06, "loss": 1.9871, "step": 26254 }, { "epoch": 0.65, "learning_rate": 5.84961319710782e-06, "loss": 2.1221, "step": 26255 }, { "epoch": 0.65, "learning_rate": 5.8488868856674685e-06, "loss": 2.1444, "step": 26256 }, { "epoch": 0.65, "learning_rate": 5.848160600683511e-06, "loss": 2.1297, "step": 26257 }, { "epoch": 0.65, "learning_rate": 5.847434342160582e-06, "loss": 2.1386, "step": 26258 }, { "epoch": 0.65, "learning_rate": 5.846708110103308e-06, "loss": 1.8904, "step": 26259 }, { "epoch": 0.65, "learning_rate": 5.845981904516312e-06, "loss": 1.9826, "step": 26260 }, { "epoch": 0.65, "learning_rate": 5.845255725404233e-06, "loss": 1.8875, "step": 26261 }, { "epoch": 0.65, "learning_rate": 5.844529572771689e-06, "loss": 1.947, "step": 26262 }, { "epoch": 0.65, "learning_rate": 5.843803446623313e-06, "loss": 1.9547, "step": 26263 }, { "epoch": 0.65, "learning_rate": 5.843077346963734e-06, "loss": 1.9075, "step": 26264 }, { "epoch": 0.65, "learning_rate": 5.842351273797576e-06, "loss": 1.7709, "step": 26265 }, { "epoch": 0.65, "learning_rate": 5.841625227129469e-06, "loss": 2.0793, "step": 26266 }, { "epoch": 0.65, "learning_rate": 5.840899206964039e-06, "loss": 2.088, "step": 26267 }, { "epoch": 0.65, "learning_rate": 5.840173213305912e-06, "loss": 1.8371, "step": 26268 }, { "epoch": 0.65, "learning_rate": 5.839447246159722e-06, "loss": 1.9482, "step": 26269 }, { "epoch": 0.65, "learning_rate": 5.838721305530086e-06, "loss": 2.1183, "step": 26270 }, { "epoch": 0.65, "learning_rate": 5.837995391421639e-06, "loss": 1.9011, "step": 26271 }, { "epoch": 0.65, "learning_rate": 5.8372695038390005e-06, "loss": 2.1165, "step": 26272 }, { "epoch": 0.65, "learning_rate": 5.836543642786799e-06, "loss": 1.9924, "step": 26273 }, { "epoch": 0.65, "learning_rate": 5.835817808269666e-06, "loss": 2.0076, "step": 26274 }, { "epoch": 0.65, "learning_rate": 5.83509200029222e-06, "loss": 2.0087, "step": 26275 }, { "epoch": 0.65, "learning_rate": 5.8343662188590935e-06, "loss": 2.0148, "step": 26276 }, { "epoch": 0.65, "learning_rate": 5.8336404639749056e-06, "loss": 1.9724, "step": 26277 }, { "epoch": 0.65, "learning_rate": 5.832914735644284e-06, "loss": 1.9065, "step": 26278 }, { "epoch": 0.65, "learning_rate": 5.832189033871854e-06, "loss": 1.9811, "step": 26279 }, { "epoch": 0.65, "learning_rate": 5.831463358662242e-06, "loss": 2.1165, "step": 26280 }, { "epoch": 0.65, "learning_rate": 5.830737710020076e-06, "loss": 1.9211, "step": 26281 }, { "epoch": 0.65, "learning_rate": 5.830012087949974e-06, "loss": 2.0236, "step": 26282 }, { "epoch": 0.65, "learning_rate": 5.829286492456565e-06, "loss": 1.8042, "step": 26283 }, { "epoch": 0.65, "learning_rate": 5.8285609235444755e-06, "loss": 1.8249, "step": 26284 }, { "epoch": 0.65, "learning_rate": 5.827835381218321e-06, "loss": 1.9551, "step": 26285 }, { "epoch": 0.65, "learning_rate": 5.827109865482737e-06, "loss": 2.1733, "step": 26286 }, { "epoch": 0.65, "learning_rate": 5.8263843763423376e-06, "loss": 2.059, "step": 26287 }, { "epoch": 0.65, "learning_rate": 5.825658913801751e-06, "loss": 1.8438, "step": 26288 }, { "epoch": 0.65, "learning_rate": 5.824933477865604e-06, "loss": 1.9692, "step": 26289 }, { "epoch": 0.65, "learning_rate": 5.824208068538512e-06, "loss": 2.0018, "step": 26290 }, { "epoch": 0.65, "learning_rate": 5.823482685825108e-06, "loss": 1.9706, "step": 26291 }, { "epoch": 0.65, "learning_rate": 5.822757329730005e-06, "loss": 2.0111, "step": 26292 }, { "epoch": 0.65, "learning_rate": 5.822032000257831e-06, "loss": 2.0424, "step": 26293 }, { "epoch": 0.65, "learning_rate": 5.821306697413214e-06, "loss": 1.8986, "step": 26294 }, { "epoch": 0.65, "learning_rate": 5.820581421200766e-06, "loss": 1.9902, "step": 26295 }, { "epoch": 0.65, "learning_rate": 5.819856171625115e-06, "loss": 1.8444, "step": 26296 }, { "epoch": 0.65, "learning_rate": 5.819130948690882e-06, "loss": 1.9992, "step": 26297 }, { "epoch": 0.65, "learning_rate": 5.81840575240269e-06, "loss": 2.0615, "step": 26298 }, { "epoch": 0.65, "learning_rate": 5.817680582765166e-06, "loss": 1.9742, "step": 26299 }, { "epoch": 0.65, "learning_rate": 5.816955439782922e-06, "loss": 1.9338, "step": 26300 }, { "epoch": 0.65, "learning_rate": 5.816230323460588e-06, "loss": 1.8974, "step": 26301 }, { "epoch": 0.65, "learning_rate": 5.815505233802778e-06, "loss": 1.869, "step": 26302 }, { "epoch": 0.65, "learning_rate": 5.814780170814117e-06, "loss": 2.1181, "step": 26303 }, { "epoch": 0.65, "learning_rate": 5.81405513449923e-06, "loss": 1.9359, "step": 26304 }, { "epoch": 0.65, "learning_rate": 5.81333012486273e-06, "loss": 2.002, "step": 26305 }, { "epoch": 0.65, "learning_rate": 5.812605141909244e-06, "loss": 2.0711, "step": 26306 }, { "epoch": 0.65, "learning_rate": 5.811880185643387e-06, "loss": 2.0966, "step": 26307 }, { "epoch": 0.65, "learning_rate": 5.811155256069783e-06, "loss": 1.9902, "step": 26308 }, { "epoch": 0.65, "learning_rate": 5.810430353193055e-06, "loss": 2.0505, "step": 26309 }, { "epoch": 0.65, "learning_rate": 5.8097054770178155e-06, "loss": 2.1311, "step": 26310 }, { "epoch": 0.65, "learning_rate": 5.808980627548693e-06, "loss": 1.9501, "step": 26311 }, { "epoch": 0.65, "learning_rate": 5.808255804790298e-06, "loss": 1.8794, "step": 26312 }, { "epoch": 0.65, "learning_rate": 5.8075310087472556e-06, "loss": 2.0008, "step": 26313 }, { "epoch": 0.65, "learning_rate": 5.806806239424184e-06, "loss": 2.03, "step": 26314 }, { "epoch": 0.65, "learning_rate": 5.806081496825704e-06, "loss": 2.0411, "step": 26315 }, { "epoch": 0.65, "learning_rate": 5.8053567809564356e-06, "loss": 1.8246, "step": 26316 }, { "epoch": 0.65, "learning_rate": 5.804632091820992e-06, "loss": 1.7765, "step": 26317 }, { "epoch": 0.65, "learning_rate": 5.803907429423996e-06, "loss": 2.0228, "step": 26318 }, { "epoch": 0.65, "learning_rate": 5.803182793770068e-06, "loss": 1.7301, "step": 26319 }, { "epoch": 0.65, "learning_rate": 5.802458184863822e-06, "loss": 1.9921, "step": 26320 }, { "epoch": 0.65, "learning_rate": 5.801733602709876e-06, "loss": 1.8718, "step": 26321 }, { "epoch": 0.65, "learning_rate": 5.8010090473128545e-06, "loss": 2.0104, "step": 26322 }, { "epoch": 0.65, "learning_rate": 5.800284518677367e-06, "loss": 1.887, "step": 26323 }, { "epoch": 0.65, "learning_rate": 5.799560016808039e-06, "loss": 1.8963, "step": 26324 }, { "epoch": 0.65, "learning_rate": 5.79883554170948e-06, "loss": 2.0647, "step": 26325 }, { "epoch": 0.65, "learning_rate": 5.79811109338631e-06, "loss": 2.0415, "step": 26326 }, { "epoch": 0.65, "learning_rate": 5.797386671843153e-06, "loss": 1.8602, "step": 26327 }, { "epoch": 0.65, "learning_rate": 5.7966622770846145e-06, "loss": 1.9126, "step": 26328 }, { "epoch": 0.65, "learning_rate": 5.795937909115322e-06, "loss": 2.0264, "step": 26329 }, { "epoch": 0.65, "learning_rate": 5.795213567939884e-06, "loss": 1.9938, "step": 26330 }, { "epoch": 0.65, "learning_rate": 5.7944892535629195e-06, "loss": 1.9852, "step": 26331 }, { "epoch": 0.65, "learning_rate": 5.793764965989046e-06, "loss": 1.9547, "step": 26332 }, { "epoch": 0.65, "learning_rate": 5.793040705222879e-06, "loss": 1.8788, "step": 26333 }, { "epoch": 0.65, "learning_rate": 5.792316471269037e-06, "loss": 2.0865, "step": 26334 }, { "epoch": 0.65, "learning_rate": 5.791592264132131e-06, "loss": 1.9504, "step": 26335 }, { "epoch": 0.65, "learning_rate": 5.790868083816779e-06, "loss": 2.0086, "step": 26336 }, { "epoch": 0.65, "learning_rate": 5.790143930327601e-06, "loss": 1.8214, "step": 26337 }, { "epoch": 0.65, "learning_rate": 5.7894198036692026e-06, "loss": 2.1174, "step": 26338 }, { "epoch": 0.65, "learning_rate": 5.7886957038462074e-06, "loss": 1.9668, "step": 26339 }, { "epoch": 0.65, "learning_rate": 5.787971630863224e-06, "loss": 2.0332, "step": 26340 }, { "epoch": 0.65, "learning_rate": 5.78724758472487e-06, "loss": 2.0879, "step": 26341 }, { "epoch": 0.65, "learning_rate": 5.786523565435764e-06, "loss": 1.7873, "step": 26342 }, { "epoch": 0.65, "learning_rate": 5.785799573000512e-06, "loss": 2.1806, "step": 26343 }, { "epoch": 0.65, "learning_rate": 5.785075607423737e-06, "loss": 1.7595, "step": 26344 }, { "epoch": 0.65, "learning_rate": 5.7843516687100445e-06, "loss": 1.9818, "step": 26345 }, { "epoch": 0.65, "learning_rate": 5.783627756864053e-06, "loss": 2.0241, "step": 26346 }, { "epoch": 0.65, "learning_rate": 5.782903871890379e-06, "loss": 1.975, "step": 26347 }, { "epoch": 0.65, "learning_rate": 5.78218001379363e-06, "loss": 1.7977, "step": 26348 }, { "epoch": 0.65, "learning_rate": 5.781456182578422e-06, "loss": 1.9334, "step": 26349 }, { "epoch": 0.65, "learning_rate": 5.78073237824937e-06, "loss": 1.9923, "step": 26350 }, { "epoch": 0.65, "learning_rate": 5.780008600811084e-06, "loss": 1.9558, "step": 26351 }, { "epoch": 0.65, "learning_rate": 5.779284850268182e-06, "loss": 2.1218, "step": 26352 }, { "epoch": 0.65, "learning_rate": 5.778561126625269e-06, "loss": 2.084, "step": 26353 }, { "epoch": 0.65, "learning_rate": 5.7778374298869664e-06, "loss": 1.9745, "step": 26354 }, { "epoch": 0.65, "learning_rate": 5.777113760057878e-06, "loss": 2.0455, "step": 26355 }, { "epoch": 0.65, "learning_rate": 5.776390117142621e-06, "loss": 1.9625, "step": 26356 }, { "epoch": 0.65, "learning_rate": 5.7756665011458065e-06, "loss": 1.9272, "step": 26357 }, { "epoch": 0.65, "learning_rate": 5.7749429120720454e-06, "loss": 2.2452, "step": 26358 }, { "epoch": 0.65, "learning_rate": 5.774219349925954e-06, "loss": 1.9235, "step": 26359 }, { "epoch": 0.65, "learning_rate": 5.773495814712134e-06, "loss": 1.9641, "step": 26360 }, { "epoch": 0.65, "learning_rate": 5.772772306435204e-06, "loss": 2.0925, "step": 26361 }, { "epoch": 0.65, "learning_rate": 5.772048825099777e-06, "loss": 1.984, "step": 26362 }, { "epoch": 0.65, "learning_rate": 5.771325370710457e-06, "loss": 1.9691, "step": 26363 }, { "epoch": 0.65, "learning_rate": 5.7706019432718605e-06, "loss": 1.904, "step": 26364 }, { "epoch": 0.65, "learning_rate": 5.769878542788595e-06, "loss": 1.9097, "step": 26365 }, { "epoch": 0.65, "learning_rate": 5.7691551692652705e-06, "loss": 1.8736, "step": 26366 }, { "epoch": 0.65, "learning_rate": 5.768431822706499e-06, "loss": 1.9393, "step": 26367 }, { "epoch": 0.65, "learning_rate": 5.767708503116891e-06, "loss": 2.0136, "step": 26368 }, { "epoch": 0.65, "learning_rate": 5.766985210501062e-06, "loss": 1.8069, "step": 26369 }, { "epoch": 0.65, "learning_rate": 5.766261944863608e-06, "loss": 2.1341, "step": 26370 }, { "epoch": 0.65, "learning_rate": 5.765538706209148e-06, "loss": 1.9136, "step": 26371 }, { "epoch": 0.65, "learning_rate": 5.764815494542294e-06, "loss": 1.951, "step": 26372 }, { "epoch": 0.65, "learning_rate": 5.764092309867647e-06, "loss": 1.9181, "step": 26373 }, { "epoch": 0.65, "learning_rate": 5.763369152189824e-06, "loss": 1.8738, "step": 26374 }, { "epoch": 0.65, "learning_rate": 5.762646021513426e-06, "loss": 1.9247, "step": 26375 }, { "epoch": 0.65, "learning_rate": 5.761922917843066e-06, "loss": 2.0391, "step": 26376 }, { "epoch": 0.65, "learning_rate": 5.761199841183356e-06, "loss": 1.8885, "step": 26377 }, { "epoch": 0.65, "learning_rate": 5.760476791538898e-06, "loss": 2.0157, "step": 26378 }, { "epoch": 0.65, "learning_rate": 5.759753768914302e-06, "loss": 1.7784, "step": 26379 }, { "epoch": 0.65, "learning_rate": 5.759030773314181e-06, "loss": 1.9743, "step": 26380 }, { "epoch": 0.65, "learning_rate": 5.758307804743135e-06, "loss": 2.1361, "step": 26381 }, { "epoch": 0.65, "learning_rate": 5.75758486320578e-06, "loss": 1.8509, "step": 26382 }, { "epoch": 0.65, "learning_rate": 5.7568619487067155e-06, "loss": 2.0001, "step": 26383 }, { "epoch": 0.65, "learning_rate": 5.756139061250552e-06, "loss": 2.0411, "step": 26384 }, { "epoch": 0.65, "learning_rate": 5.755416200841897e-06, "loss": 1.9742, "step": 26385 }, { "epoch": 0.65, "learning_rate": 5.754693367485359e-06, "loss": 1.9866, "step": 26386 }, { "epoch": 0.65, "learning_rate": 5.7539705611855465e-06, "loss": 1.9966, "step": 26387 }, { "epoch": 0.65, "learning_rate": 5.75324778194706e-06, "loss": 1.8813, "step": 26388 }, { "epoch": 0.65, "learning_rate": 5.75252502977451e-06, "loss": 1.9081, "step": 26389 }, { "epoch": 0.65, "learning_rate": 5.751802304672505e-06, "loss": 1.8598, "step": 26390 }, { "epoch": 0.65, "learning_rate": 5.751079606645644e-06, "loss": 1.9045, "step": 26391 }, { "epoch": 0.65, "learning_rate": 5.750356935698542e-06, "loss": 1.9167, "step": 26392 }, { "epoch": 0.65, "learning_rate": 5.749634291835797e-06, "loss": 2.0299, "step": 26393 }, { "epoch": 0.65, "learning_rate": 5.748911675062017e-06, "loss": 1.9376, "step": 26394 }, { "epoch": 0.65, "learning_rate": 5.748189085381812e-06, "loss": 1.9013, "step": 26395 }, { "epoch": 0.65, "learning_rate": 5.74746652279978e-06, "loss": 2.001, "step": 26396 }, { "epoch": 0.65, "learning_rate": 5.746743987320534e-06, "loss": 1.7929, "step": 26397 }, { "epoch": 0.65, "learning_rate": 5.746021478948671e-06, "loss": 2.0657, "step": 26398 }, { "epoch": 0.65, "learning_rate": 5.745298997688798e-06, "loss": 2.0161, "step": 26399 }, { "epoch": 0.65, "learning_rate": 5.744576543545526e-06, "loss": 1.965, "step": 26400 }, { "epoch": 0.65, "learning_rate": 5.7438541165234504e-06, "loss": 2.1075, "step": 26401 }, { "epoch": 0.65, "learning_rate": 5.743131716627179e-06, "loss": 2.0349, "step": 26402 }, { "epoch": 0.65, "learning_rate": 5.742409343861317e-06, "loss": 2.0178, "step": 26403 }, { "epoch": 0.65, "learning_rate": 5.741686998230469e-06, "loss": 1.9876, "step": 26404 }, { "epoch": 0.65, "learning_rate": 5.7409646797392404e-06, "loss": 2.0857, "step": 26405 }, { "epoch": 0.65, "learning_rate": 5.740242388392228e-06, "loss": 1.8174, "step": 26406 }, { "epoch": 0.65, "learning_rate": 5.739520124194042e-06, "loss": 1.8043, "step": 26407 }, { "epoch": 0.65, "learning_rate": 5.738797887149279e-06, "loss": 2.034, "step": 26408 }, { "epoch": 0.65, "learning_rate": 5.738075677262547e-06, "loss": 2.0394, "step": 26409 }, { "epoch": 0.65, "learning_rate": 5.73735349453845e-06, "loss": 1.9177, "step": 26410 }, { "epoch": 0.65, "learning_rate": 5.736631338981585e-06, "loss": 1.8444, "step": 26411 }, { "epoch": 0.65, "learning_rate": 5.735909210596562e-06, "loss": 2.0241, "step": 26412 }, { "epoch": 0.65, "learning_rate": 5.735187109387976e-06, "loss": 1.923, "step": 26413 }, { "epoch": 0.65, "learning_rate": 5.734465035360431e-06, "loss": 1.9863, "step": 26414 }, { "epoch": 0.65, "learning_rate": 5.733742988518536e-06, "loss": 2.172, "step": 26415 }, { "epoch": 0.65, "learning_rate": 5.733020968866882e-06, "loss": 1.9427, "step": 26416 }, { "epoch": 0.65, "learning_rate": 5.73229897641008e-06, "loss": 1.9535, "step": 26417 }, { "epoch": 0.65, "learning_rate": 5.731577011152724e-06, "loss": 1.8867, "step": 26418 }, { "epoch": 0.65, "learning_rate": 5.730855073099418e-06, "loss": 1.9665, "step": 26419 }, { "epoch": 0.65, "learning_rate": 5.730133162254765e-06, "loss": 1.9121, "step": 26420 }, { "epoch": 0.65, "learning_rate": 5.729411278623365e-06, "loss": 1.8148, "step": 26421 }, { "epoch": 0.65, "learning_rate": 5.728689422209821e-06, "loss": 1.9475, "step": 26422 }, { "epoch": 0.65, "learning_rate": 5.7279675930187284e-06, "loss": 1.8932, "step": 26423 }, { "epoch": 0.65, "learning_rate": 5.727245791054689e-06, "loss": 2.0038, "step": 26424 }, { "epoch": 0.65, "learning_rate": 5.726524016322311e-06, "loss": 1.9, "step": 26425 }, { "epoch": 0.65, "learning_rate": 5.725802268826184e-06, "loss": 1.9412, "step": 26426 }, { "epoch": 0.65, "learning_rate": 5.7250805485709135e-06, "loss": 2.0409, "step": 26427 }, { "epoch": 0.65, "learning_rate": 5.724358855561096e-06, "loss": 2.1091, "step": 26428 }, { "epoch": 0.65, "learning_rate": 5.723637189801332e-06, "loss": 1.907, "step": 26429 }, { "epoch": 0.65, "learning_rate": 5.722915551296226e-06, "loss": 1.9557, "step": 26430 }, { "epoch": 0.65, "learning_rate": 5.722193940050371e-06, "loss": 1.7888, "step": 26431 }, { "epoch": 0.65, "learning_rate": 5.721472356068369e-06, "loss": 1.7754, "step": 26432 }, { "epoch": 0.65, "learning_rate": 5.7207507993548165e-06, "loss": 2.0065, "step": 26433 }, { "epoch": 0.65, "learning_rate": 5.720029269914312e-06, "loss": 1.891, "step": 26434 }, { "epoch": 0.65, "learning_rate": 5.719307767751461e-06, "loss": 2.1654, "step": 26435 }, { "epoch": 0.65, "learning_rate": 5.718586292870851e-06, "loss": 2.0334, "step": 26436 }, { "epoch": 0.65, "learning_rate": 5.7178648452770875e-06, "loss": 1.9676, "step": 26437 }, { "epoch": 0.65, "learning_rate": 5.717143424974766e-06, "loss": 1.9092, "step": 26438 }, { "epoch": 0.65, "learning_rate": 5.716422031968485e-06, "loss": 2.0149, "step": 26439 }, { "epoch": 0.65, "learning_rate": 5.715700666262847e-06, "loss": 2.0835, "step": 26440 }, { "epoch": 0.65, "learning_rate": 5.714979327862441e-06, "loss": 1.8707, "step": 26441 }, { "epoch": 0.65, "learning_rate": 5.714258016771868e-06, "loss": 2.0099, "step": 26442 }, { "epoch": 0.65, "learning_rate": 5.713536732995729e-06, "loss": 1.9671, "step": 26443 }, { "epoch": 0.65, "learning_rate": 5.712815476538613e-06, "loss": 1.8749, "step": 26444 }, { "epoch": 0.65, "learning_rate": 5.7120942474051264e-06, "loss": 1.9783, "step": 26445 }, { "epoch": 0.65, "learning_rate": 5.711373045599855e-06, "loss": 1.8833, "step": 26446 }, { "epoch": 0.65, "learning_rate": 5.710651871127402e-06, "loss": 1.9829, "step": 26447 }, { "epoch": 0.65, "learning_rate": 5.709930723992365e-06, "loss": 1.9114, "step": 26448 }, { "epoch": 0.65, "learning_rate": 5.709209604199335e-06, "loss": 1.9791, "step": 26449 }, { "epoch": 0.65, "learning_rate": 5.708488511752914e-06, "loss": 1.9661, "step": 26450 }, { "epoch": 0.65, "learning_rate": 5.70776744665769e-06, "loss": 1.9785, "step": 26451 }, { "epoch": 0.65, "learning_rate": 5.7070464089182646e-06, "loss": 1.8787, "step": 26452 }, { "epoch": 0.65, "learning_rate": 5.706325398539234e-06, "loss": 1.9367, "step": 26453 }, { "epoch": 0.65, "learning_rate": 5.705604415525188e-06, "loss": 1.9689, "step": 26454 }, { "epoch": 0.65, "learning_rate": 5.7048834598807224e-06, "loss": 1.8989, "step": 26455 }, { "epoch": 0.65, "learning_rate": 5.7041625316104356e-06, "loss": 1.8892, "step": 26456 }, { "epoch": 0.65, "learning_rate": 5.703441630718921e-06, "loss": 1.961, "step": 26457 }, { "epoch": 0.65, "learning_rate": 5.702720757210777e-06, "loss": 1.9682, "step": 26458 }, { "epoch": 0.65, "learning_rate": 5.70199991109059e-06, "loss": 1.9102, "step": 26459 }, { "epoch": 0.65, "learning_rate": 5.701279092362961e-06, "loss": 1.9178, "step": 26460 }, { "epoch": 0.65, "learning_rate": 5.70055830103248e-06, "loss": 1.9894, "step": 26461 }, { "epoch": 0.65, "learning_rate": 5.6998375371037405e-06, "loss": 1.9297, "step": 26462 }, { "epoch": 0.65, "learning_rate": 5.699116800581341e-06, "loss": 1.9721, "step": 26463 }, { "epoch": 0.65, "learning_rate": 5.6983960914698685e-06, "loss": 1.9818, "step": 26464 }, { "epoch": 0.65, "learning_rate": 5.6976754097739236e-06, "loss": 1.8457, "step": 26465 }, { "epoch": 0.65, "learning_rate": 5.696954755498092e-06, "loss": 2.0528, "step": 26466 }, { "epoch": 0.65, "learning_rate": 5.696234128646969e-06, "loss": 1.9718, "step": 26467 }, { "epoch": 0.65, "learning_rate": 5.695513529225154e-06, "loss": 2.0388, "step": 26468 }, { "epoch": 0.65, "learning_rate": 5.694792957237228e-06, "loss": 2.0184, "step": 26469 }, { "epoch": 0.65, "learning_rate": 5.694072412687795e-06, "loss": 1.8529, "step": 26470 }, { "epoch": 0.65, "learning_rate": 5.693351895581437e-06, "loss": 1.9412, "step": 26471 }, { "epoch": 0.65, "learning_rate": 5.692631405922752e-06, "loss": 1.9162, "step": 26472 }, { "epoch": 0.65, "learning_rate": 5.69191094371633e-06, "loss": 1.7277, "step": 26473 }, { "epoch": 0.65, "learning_rate": 5.691190508966763e-06, "loss": 1.997, "step": 26474 }, { "epoch": 0.65, "learning_rate": 5.690470101678647e-06, "loss": 1.9631, "step": 26475 }, { "epoch": 0.65, "learning_rate": 5.6897497218565665e-06, "loss": 1.8656, "step": 26476 }, { "epoch": 0.65, "learning_rate": 5.689029369505115e-06, "loss": 1.9363, "step": 26477 }, { "epoch": 0.65, "learning_rate": 5.688309044628888e-06, "loss": 1.8408, "step": 26478 }, { "epoch": 0.65, "learning_rate": 5.687588747232469e-06, "loss": 1.9707, "step": 26479 }, { "epoch": 0.65, "learning_rate": 5.686868477320458e-06, "loss": 1.9824, "step": 26480 }, { "epoch": 0.65, "learning_rate": 5.686148234897433e-06, "loss": 1.9161, "step": 26481 }, { "epoch": 0.65, "learning_rate": 5.685428019967993e-06, "loss": 2.0635, "step": 26482 }, { "epoch": 0.65, "learning_rate": 5.684707832536729e-06, "loss": 1.8837, "step": 26483 }, { "epoch": 0.65, "learning_rate": 5.683987672608226e-06, "loss": 1.8802, "step": 26484 }, { "epoch": 0.65, "learning_rate": 5.683267540187078e-06, "loss": 2.042, "step": 26485 }, { "epoch": 0.65, "learning_rate": 5.68254743527787e-06, "loss": 2.0325, "step": 26486 }, { "epoch": 0.65, "learning_rate": 5.681827357885195e-06, "loss": 1.9021, "step": 26487 }, { "epoch": 0.65, "learning_rate": 5.6811073080136435e-06, "loss": 1.9199, "step": 26488 }, { "epoch": 0.65, "learning_rate": 5.6803872856677984e-06, "loss": 1.9714, "step": 26489 }, { "epoch": 0.65, "learning_rate": 5.6796672908522555e-06, "loss": 1.8989, "step": 26490 }, { "epoch": 0.65, "learning_rate": 5.6789473235716e-06, "loss": 1.8455, "step": 26491 }, { "epoch": 0.65, "learning_rate": 5.67822738383042e-06, "loss": 2.1286, "step": 26492 }, { "epoch": 0.65, "learning_rate": 5.67750747163331e-06, "loss": 2.0358, "step": 26493 }, { "epoch": 0.65, "learning_rate": 5.6767875869848495e-06, "loss": 1.7764, "step": 26494 }, { "epoch": 0.65, "learning_rate": 5.676067729889632e-06, "loss": 1.815, "step": 26495 }, { "epoch": 0.65, "learning_rate": 5.675347900352246e-06, "loss": 1.9759, "step": 26496 }, { "epoch": 0.65, "learning_rate": 5.674628098377274e-06, "loss": 2.0935, "step": 26497 }, { "epoch": 0.65, "learning_rate": 5.6739083239693104e-06, "loss": 1.6948, "step": 26498 }, { "epoch": 0.65, "learning_rate": 5.673188577132936e-06, "loss": 2.0385, "step": 26499 }, { "epoch": 0.65, "learning_rate": 5.672468857872739e-06, "loss": 2.1153, "step": 26500 }, { "epoch": 0.65, "learning_rate": 5.671749166193314e-06, "loss": 2.1411, "step": 26501 }, { "epoch": 0.65, "learning_rate": 5.671029502099237e-06, "loss": 1.9724, "step": 26502 }, { "epoch": 0.65, "learning_rate": 5.670309865595105e-06, "loss": 1.9323, "step": 26503 }, { "epoch": 0.65, "learning_rate": 5.669590256685493e-06, "loss": 1.8627, "step": 26504 }, { "epoch": 0.65, "learning_rate": 5.6688706753749955e-06, "loss": 1.9562, "step": 26505 }, { "epoch": 0.65, "learning_rate": 5.668151121668201e-06, "loss": 2.0627, "step": 26506 }, { "epoch": 0.65, "learning_rate": 5.667431595569686e-06, "loss": 1.9747, "step": 26507 }, { "epoch": 0.65, "learning_rate": 5.666712097084041e-06, "loss": 1.9821, "step": 26508 }, { "epoch": 0.65, "learning_rate": 5.6659926262158526e-06, "loss": 2.1105, "step": 26509 }, { "epoch": 0.65, "learning_rate": 5.665273182969706e-06, "loss": 2.1159, "step": 26510 }, { "epoch": 0.65, "learning_rate": 5.664553767350189e-06, "loss": 1.8904, "step": 26511 }, { "epoch": 0.65, "learning_rate": 5.6638343793618815e-06, "loss": 2.0722, "step": 26512 }, { "epoch": 0.65, "learning_rate": 5.663115019009373e-06, "loss": 1.9472, "step": 26513 }, { "epoch": 0.65, "learning_rate": 5.662395686297242e-06, "loss": 1.9451, "step": 26514 }, { "epoch": 0.65, "learning_rate": 5.661676381230077e-06, "loss": 2.1139, "step": 26515 }, { "epoch": 0.65, "learning_rate": 5.660957103812467e-06, "loss": 2.0208, "step": 26516 }, { "epoch": 0.65, "learning_rate": 5.660237854048986e-06, "loss": 2.0537, "step": 26517 }, { "epoch": 0.65, "learning_rate": 5.65951863194423e-06, "loss": 1.9194, "step": 26518 }, { "epoch": 0.65, "learning_rate": 5.65879943750277e-06, "loss": 2.0257, "step": 26519 }, { "epoch": 0.65, "learning_rate": 5.658080270729196e-06, "loss": 1.9871, "step": 26520 }, { "epoch": 0.65, "learning_rate": 5.657361131628096e-06, "loss": 2.0487, "step": 26521 }, { "epoch": 0.65, "learning_rate": 5.656642020204045e-06, "loss": 1.9119, "step": 26522 }, { "epoch": 0.65, "learning_rate": 5.655922936461633e-06, "loss": 2.0182, "step": 26523 }, { "epoch": 0.65, "learning_rate": 5.655203880405438e-06, "loss": 2.099, "step": 26524 }, { "epoch": 0.65, "learning_rate": 5.654484852040042e-06, "loss": 1.8326, "step": 26525 }, { "epoch": 0.65, "learning_rate": 5.653765851370031e-06, "loss": 2.0125, "step": 26526 }, { "epoch": 0.65, "learning_rate": 5.653046878399987e-06, "loss": 1.9164, "step": 26527 }, { "epoch": 0.65, "learning_rate": 5.652327933134496e-06, "loss": 2.2038, "step": 26528 }, { "epoch": 0.65, "learning_rate": 5.651609015578131e-06, "loss": 2.0693, "step": 26529 }, { "epoch": 0.65, "learning_rate": 5.65089012573548e-06, "loss": 2.0001, "step": 26530 }, { "epoch": 0.65, "learning_rate": 5.650171263611126e-06, "loss": 1.9527, "step": 26531 }, { "epoch": 0.65, "learning_rate": 5.649452429209646e-06, "loss": 1.9899, "step": 26532 }, { "epoch": 0.65, "learning_rate": 5.648733622535627e-06, "loss": 2.0458, "step": 26533 }, { "epoch": 0.65, "learning_rate": 5.648014843593641e-06, "loss": 1.8777, "step": 26534 }, { "epoch": 0.65, "learning_rate": 5.647296092388277e-06, "loss": 2.0168, "step": 26535 }, { "epoch": 0.65, "learning_rate": 5.646577368924116e-06, "loss": 1.9197, "step": 26536 }, { "epoch": 0.65, "learning_rate": 5.645858673205733e-06, "loss": 1.999, "step": 26537 }, { "epoch": 0.65, "learning_rate": 5.645140005237714e-06, "loss": 1.987, "step": 26538 }, { "epoch": 0.65, "learning_rate": 5.644421365024635e-06, "loss": 1.8908, "step": 26539 }, { "epoch": 0.65, "learning_rate": 5.643702752571078e-06, "loss": 1.9545, "step": 26540 }, { "epoch": 0.65, "learning_rate": 5.6429841678816265e-06, "loss": 2.1283, "step": 26541 }, { "epoch": 0.65, "learning_rate": 5.642265610960852e-06, "loss": 1.8892, "step": 26542 }, { "epoch": 0.65, "learning_rate": 5.641547081813339e-06, "loss": 1.896, "step": 26543 }, { "epoch": 0.65, "learning_rate": 5.6408285804436676e-06, "loss": 1.9406, "step": 26544 }, { "epoch": 0.65, "learning_rate": 5.640110106856416e-06, "loss": 1.9112, "step": 26545 }, { "epoch": 0.65, "learning_rate": 5.639391661056168e-06, "loss": 1.9776, "step": 26546 }, { "epoch": 0.65, "learning_rate": 5.6386732430474935e-06, "loss": 2.0044, "step": 26547 }, { "epoch": 0.65, "learning_rate": 5.637954852834979e-06, "loss": 2.0268, "step": 26548 }, { "epoch": 0.65, "learning_rate": 5.637236490423197e-06, "loss": 1.9567, "step": 26549 }, { "epoch": 0.65, "learning_rate": 5.636518155816728e-06, "loss": 1.9259, "step": 26550 }, { "epoch": 0.65, "learning_rate": 5.635799849020156e-06, "loss": 1.8371, "step": 26551 }, { "epoch": 0.65, "learning_rate": 5.635081570038049e-06, "loss": 1.9229, "step": 26552 }, { "epoch": 0.65, "learning_rate": 5.6343633188749934e-06, "loss": 2.0093, "step": 26553 }, { "epoch": 0.65, "learning_rate": 5.633645095535559e-06, "loss": 2.1895, "step": 26554 }, { "epoch": 0.65, "learning_rate": 5.632926900024327e-06, "loss": 2.0718, "step": 26555 }, { "epoch": 0.65, "learning_rate": 5.632208732345881e-06, "loss": 1.9445, "step": 26556 }, { "epoch": 0.65, "learning_rate": 5.631490592504786e-06, "loss": 1.9809, "step": 26557 }, { "epoch": 0.65, "learning_rate": 5.630772480505627e-06, "loss": 1.9153, "step": 26558 }, { "epoch": 0.65, "learning_rate": 5.6300543963529805e-06, "loss": 2.1372, "step": 26559 }, { "epoch": 0.65, "learning_rate": 5.629336340051419e-06, "loss": 2.0665, "step": 26560 }, { "epoch": 0.65, "learning_rate": 5.628618311605521e-06, "loss": 1.9507, "step": 26561 }, { "epoch": 0.65, "learning_rate": 5.627900311019863e-06, "loss": 1.9717, "step": 26562 }, { "epoch": 0.65, "learning_rate": 5.627182338299022e-06, "loss": 1.994, "step": 26563 }, { "epoch": 0.65, "learning_rate": 5.626464393447576e-06, "loss": 2.1569, "step": 26564 }, { "epoch": 0.65, "learning_rate": 5.625746476470095e-06, "loss": 1.968, "step": 26565 }, { "epoch": 0.65, "learning_rate": 5.62502858737116e-06, "loss": 2.1155, "step": 26566 }, { "epoch": 0.65, "learning_rate": 5.62431072615534e-06, "loss": 1.919, "step": 26567 }, { "epoch": 0.65, "learning_rate": 5.623592892827212e-06, "loss": 1.9422, "step": 26568 }, { "epoch": 0.65, "learning_rate": 5.62287508739136e-06, "loss": 1.9284, "step": 26569 }, { "epoch": 0.65, "learning_rate": 5.622157309852345e-06, "loss": 2.0095, "step": 26570 }, { "epoch": 0.65, "learning_rate": 5.621439560214753e-06, "loss": 2.0322, "step": 26571 }, { "epoch": 0.65, "learning_rate": 5.620721838483149e-06, "loss": 2.061, "step": 26572 }, { "epoch": 0.65, "learning_rate": 5.620004144662113e-06, "loss": 1.984, "step": 26573 }, { "epoch": 0.66, "learning_rate": 5.619286478756222e-06, "loss": 2.0412, "step": 26574 }, { "epoch": 0.66, "learning_rate": 5.618568840770041e-06, "loss": 1.924, "step": 26575 }, { "epoch": 0.66, "learning_rate": 5.617851230708149e-06, "loss": 1.8918, "step": 26576 }, { "epoch": 0.66, "learning_rate": 5.6171336485751236e-06, "loss": 1.9087, "step": 26577 }, { "epoch": 0.66, "learning_rate": 5.61641609437553e-06, "loss": 1.9206, "step": 26578 }, { "epoch": 0.66, "learning_rate": 5.615698568113944e-06, "loss": 2.1248, "step": 26579 }, { "epoch": 0.66, "learning_rate": 5.614981069794941e-06, "loss": 2.0335, "step": 26580 }, { "epoch": 0.66, "learning_rate": 5.614263599423095e-06, "loss": 1.993, "step": 26581 }, { "epoch": 0.66, "learning_rate": 5.613546157002973e-06, "loss": 1.9618, "step": 26582 }, { "epoch": 0.66, "learning_rate": 5.612828742539153e-06, "loss": 1.8473, "step": 26583 }, { "epoch": 0.66, "learning_rate": 5.612111356036207e-06, "loss": 2.0532, "step": 26584 }, { "epoch": 0.66, "learning_rate": 5.611393997498702e-06, "loss": 2.0614, "step": 26585 }, { "epoch": 0.66, "learning_rate": 5.610676666931216e-06, "loss": 1.9365, "step": 26586 }, { "epoch": 0.66, "learning_rate": 5.609959364338315e-06, "loss": 1.879, "step": 26587 }, { "epoch": 0.66, "learning_rate": 5.609242089724575e-06, "loss": 1.9631, "step": 26588 }, { "epoch": 0.66, "learning_rate": 5.60852484309457e-06, "loss": 1.8439, "step": 26589 }, { "epoch": 0.66, "learning_rate": 5.6078076244528615e-06, "loss": 1.9335, "step": 26590 }, { "epoch": 0.66, "learning_rate": 5.60709043380403e-06, "loss": 1.864, "step": 26591 }, { "epoch": 0.66, "learning_rate": 5.606373271152641e-06, "loss": 1.9488, "step": 26592 }, { "epoch": 0.66, "learning_rate": 5.605656136503266e-06, "loss": 1.9257, "step": 26593 }, { "epoch": 0.66, "learning_rate": 5.604939029860476e-06, "loss": 2.0345, "step": 26594 }, { "epoch": 0.66, "learning_rate": 5.604221951228846e-06, "loss": 1.8428, "step": 26595 }, { "epoch": 0.66, "learning_rate": 5.603504900612937e-06, "loss": 2.0363, "step": 26596 }, { "epoch": 0.66, "learning_rate": 5.602787878017326e-06, "loss": 2.0935, "step": 26597 }, { "epoch": 0.66, "learning_rate": 5.60207088344658e-06, "loss": 1.9905, "step": 26598 }, { "epoch": 0.66, "learning_rate": 5.6013539169052745e-06, "loss": 1.8601, "step": 26599 }, { "epoch": 0.66, "learning_rate": 5.6006369783979686e-06, "loss": 2.0049, "step": 26600 }, { "epoch": 0.66, "learning_rate": 5.599920067929241e-06, "loss": 1.8827, "step": 26601 }, { "epoch": 0.66, "learning_rate": 5.599203185503654e-06, "loss": 1.9553, "step": 26602 }, { "epoch": 0.66, "learning_rate": 5.598486331125779e-06, "loss": 2.2855, "step": 26603 }, { "epoch": 0.66, "learning_rate": 5.597769504800188e-06, "loss": 1.973, "step": 26604 }, { "epoch": 0.66, "learning_rate": 5.597052706531443e-06, "loss": 1.874, "step": 26605 }, { "epoch": 0.66, "learning_rate": 5.59633593632412e-06, "loss": 2.0477, "step": 26606 }, { "epoch": 0.66, "learning_rate": 5.59561919418278e-06, "loss": 1.8703, "step": 26607 }, { "epoch": 0.66, "learning_rate": 5.594902480111992e-06, "loss": 2.0471, "step": 26608 }, { "epoch": 0.66, "learning_rate": 5.594185794116333e-06, "loss": 1.9167, "step": 26609 }, { "epoch": 0.66, "learning_rate": 5.593469136200358e-06, "loss": 1.8823, "step": 26610 }, { "epoch": 0.66, "learning_rate": 5.592752506368639e-06, "loss": 2.0264, "step": 26611 }, { "epoch": 0.66, "learning_rate": 5.592035904625748e-06, "loss": 1.9297, "step": 26612 }, { "epoch": 0.66, "learning_rate": 5.59131933097625e-06, "loss": 1.9943, "step": 26613 }, { "epoch": 0.66, "learning_rate": 5.590602785424708e-06, "loss": 1.9767, "step": 26614 }, { "epoch": 0.66, "learning_rate": 5.58988626797569e-06, "loss": 1.9835, "step": 26615 }, { "epoch": 0.66, "learning_rate": 5.589169778633765e-06, "loss": 1.9807, "step": 26616 }, { "epoch": 0.66, "learning_rate": 5.588453317403503e-06, "loss": 1.9707, "step": 26617 }, { "epoch": 0.66, "learning_rate": 5.587736884289461e-06, "loss": 1.9237, "step": 26618 }, { "epoch": 0.66, "learning_rate": 5.587020479296214e-06, "loss": 2.0569, "step": 26619 }, { "epoch": 0.66, "learning_rate": 5.586304102428319e-06, "loss": 1.9301, "step": 26620 }, { "epoch": 0.66, "learning_rate": 5.5855877536903465e-06, "loss": 2.1113, "step": 26621 }, { "epoch": 0.66, "learning_rate": 5.584871433086867e-06, "loss": 2.0129, "step": 26622 }, { "epoch": 0.66, "learning_rate": 5.5841551406224355e-06, "loss": 2.0579, "step": 26623 }, { "epoch": 0.66, "learning_rate": 5.5834388763016265e-06, "loss": 1.9384, "step": 26624 }, { "epoch": 0.66, "learning_rate": 5.582722640128997e-06, "loss": 1.8686, "step": 26625 }, { "epoch": 0.66, "learning_rate": 5.582006432109116e-06, "loss": 1.9068, "step": 26626 }, { "epoch": 0.66, "learning_rate": 5.581290252246552e-06, "loss": 1.9154, "step": 26627 }, { "epoch": 0.66, "learning_rate": 5.58057410054586e-06, "loss": 1.9443, "step": 26628 }, { "epoch": 0.66, "learning_rate": 5.579857977011611e-06, "loss": 1.8284, "step": 26629 }, { "epoch": 0.66, "learning_rate": 5.579141881648371e-06, "loss": 1.9415, "step": 26630 }, { "epoch": 0.66, "learning_rate": 5.578425814460695e-06, "loss": 1.9993, "step": 26631 }, { "epoch": 0.66, "learning_rate": 5.577709775453155e-06, "loss": 1.9531, "step": 26632 }, { "epoch": 0.66, "learning_rate": 5.576993764630309e-06, "loss": 1.9964, "step": 26633 }, { "epoch": 0.66, "learning_rate": 5.576277781996729e-06, "loss": 2.0687, "step": 26634 }, { "epoch": 0.66, "learning_rate": 5.5755618275569675e-06, "loss": 2.0538, "step": 26635 }, { "epoch": 0.66, "learning_rate": 5.574845901315593e-06, "loss": 2.0646, "step": 26636 }, { "epoch": 0.66, "learning_rate": 5.5741300032771714e-06, "loss": 2.1913, "step": 26637 }, { "epoch": 0.66, "learning_rate": 5.573414133446258e-06, "loss": 1.8987, "step": 26638 }, { "epoch": 0.66, "learning_rate": 5.572698291827423e-06, "loss": 2.0097, "step": 26639 }, { "epoch": 0.66, "learning_rate": 5.571982478425219e-06, "loss": 1.9267, "step": 26640 }, { "epoch": 0.66, "learning_rate": 5.571266693244216e-06, "loss": 2.005, "step": 26641 }, { "epoch": 0.66, "learning_rate": 5.570550936288977e-06, "loss": 1.8607, "step": 26642 }, { "epoch": 0.66, "learning_rate": 5.569835207564056e-06, "loss": 2.0013, "step": 26643 }, { "epoch": 0.66, "learning_rate": 5.5691195070740225e-06, "loss": 1.8361, "step": 26644 }, { "epoch": 0.66, "learning_rate": 5.5684038348234304e-06, "loss": 1.9661, "step": 26645 }, { "epoch": 0.66, "learning_rate": 5.567688190816845e-06, "loss": 2.1081, "step": 26646 }, { "epoch": 0.66, "learning_rate": 5.5669725750588285e-06, "loss": 2.0064, "step": 26647 }, { "epoch": 0.66, "learning_rate": 5.566256987553943e-06, "loss": 2.005, "step": 26648 }, { "epoch": 0.66, "learning_rate": 5.565541428306743e-06, "loss": 2.0649, "step": 26649 }, { "epoch": 0.66, "learning_rate": 5.564825897321793e-06, "loss": 1.9774, "step": 26650 }, { "epoch": 0.66, "learning_rate": 5.5641103946036525e-06, "loss": 1.8351, "step": 26651 }, { "epoch": 0.66, "learning_rate": 5.563394920156886e-06, "loss": 1.9752, "step": 26652 }, { "epoch": 0.66, "learning_rate": 5.562679473986046e-06, "loss": 2.0576, "step": 26653 }, { "epoch": 0.66, "learning_rate": 5.561964056095699e-06, "loss": 2.1266, "step": 26654 }, { "epoch": 0.66, "learning_rate": 5.561248666490398e-06, "loss": 2.0136, "step": 26655 }, { "epoch": 0.66, "learning_rate": 5.560533305174706e-06, "loss": 1.8229, "step": 26656 }, { "epoch": 0.66, "learning_rate": 5.559817972153185e-06, "loss": 1.9414, "step": 26657 }, { "epoch": 0.66, "learning_rate": 5.559102667430388e-06, "loss": 2.1195, "step": 26658 }, { "epoch": 0.66, "learning_rate": 5.558387391010881e-06, "loss": 2.1165, "step": 26659 }, { "epoch": 0.66, "learning_rate": 5.557672142899215e-06, "loss": 2.0473, "step": 26660 }, { "epoch": 0.66, "learning_rate": 5.556956923099951e-06, "loss": 1.871, "step": 26661 }, { "epoch": 0.66, "learning_rate": 5.556241731617653e-06, "loss": 1.8269, "step": 26662 }, { "epoch": 0.66, "learning_rate": 5.55552656845687e-06, "loss": 1.8547, "step": 26663 }, { "epoch": 0.66, "learning_rate": 5.554811433622166e-06, "loss": 1.9387, "step": 26664 }, { "epoch": 0.66, "learning_rate": 5.554096327118097e-06, "loss": 1.9705, "step": 26665 }, { "epoch": 0.66, "learning_rate": 5.553381248949225e-06, "loss": 1.986, "step": 26666 }, { "epoch": 0.66, "learning_rate": 5.552666199120098e-06, "loss": 1.789, "step": 26667 }, { "epoch": 0.66, "learning_rate": 5.5519511776352796e-06, "loss": 1.9535, "step": 26668 }, { "epoch": 0.66, "learning_rate": 5.5512361844993304e-06, "loss": 1.9037, "step": 26669 }, { "epoch": 0.66, "learning_rate": 5.550521219716798e-06, "loss": 2.044, "step": 26670 }, { "epoch": 0.66, "learning_rate": 5.549806283292245e-06, "loss": 2.0578, "step": 26671 }, { "epoch": 0.66, "learning_rate": 5.549091375230229e-06, "loss": 1.9733, "step": 26672 }, { "epoch": 0.66, "learning_rate": 5.548376495535301e-06, "loss": 2.214, "step": 26673 }, { "epoch": 0.66, "learning_rate": 5.547661644212021e-06, "loss": 1.789, "step": 26674 }, { "epoch": 0.66, "learning_rate": 5.546946821264949e-06, "loss": 1.9485, "step": 26675 }, { "epoch": 0.66, "learning_rate": 5.5462320266986305e-06, "loss": 2.0587, "step": 26676 }, { "epoch": 0.66, "learning_rate": 5.545517260517632e-06, "loss": 2.023, "step": 26677 }, { "epoch": 0.66, "learning_rate": 5.544802522726498e-06, "loss": 2.076, "step": 26678 }, { "epoch": 0.66, "learning_rate": 5.54408781332979e-06, "loss": 2.0082, "step": 26679 }, { "epoch": 0.66, "learning_rate": 5.543373132332068e-06, "loss": 1.9853, "step": 26680 }, { "epoch": 0.66, "learning_rate": 5.542658479737877e-06, "loss": 1.9107, "step": 26681 }, { "epoch": 0.66, "learning_rate": 5.541943855551775e-06, "loss": 1.9718, "step": 26682 }, { "epoch": 0.66, "learning_rate": 5.541229259778318e-06, "loss": 2.0389, "step": 26683 }, { "epoch": 0.66, "learning_rate": 5.540514692422063e-06, "loss": 2.0489, "step": 26684 }, { "epoch": 0.66, "learning_rate": 5.539800153487559e-06, "loss": 2.0085, "step": 26685 }, { "epoch": 0.66, "learning_rate": 5.539085642979362e-06, "loss": 2.0118, "step": 26686 }, { "epoch": 0.66, "learning_rate": 5.538371160902029e-06, "loss": 1.7932, "step": 26687 }, { "epoch": 0.66, "learning_rate": 5.537656707260106e-06, "loss": 1.9561, "step": 26688 }, { "epoch": 0.66, "learning_rate": 5.536942282058153e-06, "loss": 2.1984, "step": 26689 }, { "epoch": 0.66, "learning_rate": 5.5362278853007245e-06, "loss": 2.2098, "step": 26690 }, { "epoch": 0.66, "learning_rate": 5.535513516992366e-06, "loss": 2.1325, "step": 26691 }, { "epoch": 0.66, "learning_rate": 5.534799177137638e-06, "loss": 2.0022, "step": 26692 }, { "epoch": 0.66, "learning_rate": 5.5340848657410875e-06, "loss": 2.081, "step": 26693 }, { "epoch": 0.66, "learning_rate": 5.5333705828072695e-06, "loss": 2.003, "step": 26694 }, { "epoch": 0.66, "learning_rate": 5.53265632834074e-06, "loss": 2.0668, "step": 26695 }, { "epoch": 0.66, "learning_rate": 5.531942102346044e-06, "loss": 2.0504, "step": 26696 }, { "epoch": 0.66, "learning_rate": 5.531227904827741e-06, "loss": 1.9489, "step": 26697 }, { "epoch": 0.66, "learning_rate": 5.5305137357903726e-06, "loss": 2.065, "step": 26698 }, { "epoch": 0.66, "learning_rate": 5.5297995952384985e-06, "loss": 1.8111, "step": 26699 }, { "epoch": 0.66, "learning_rate": 5.529085483176669e-06, "loss": 1.8944, "step": 26700 }, { "epoch": 0.66, "learning_rate": 5.528371399609435e-06, "loss": 1.9985, "step": 26701 }, { "epoch": 0.66, "learning_rate": 5.5276573445413505e-06, "loss": 1.9177, "step": 26702 }, { "epoch": 0.66, "learning_rate": 5.526943317976959e-06, "loss": 1.7713, "step": 26703 }, { "epoch": 0.66, "learning_rate": 5.526229319920816e-06, "loss": 2.1104, "step": 26704 }, { "epoch": 0.66, "learning_rate": 5.525515350377475e-06, "loss": 2.1792, "step": 26705 }, { "epoch": 0.66, "learning_rate": 5.524801409351479e-06, "loss": 1.7864, "step": 26706 }, { "epoch": 0.66, "learning_rate": 5.524087496847387e-06, "loss": 1.9753, "step": 26707 }, { "epoch": 0.66, "learning_rate": 5.523373612869739e-06, "loss": 2.211, "step": 26708 }, { "epoch": 0.66, "learning_rate": 5.52265975742309e-06, "loss": 2.0855, "step": 26709 }, { "epoch": 0.66, "learning_rate": 5.521945930511994e-06, "loss": 1.9324, "step": 26710 }, { "epoch": 0.66, "learning_rate": 5.521232132140991e-06, "loss": 1.9006, "step": 26711 }, { "epoch": 0.66, "learning_rate": 5.52051836231464e-06, "loss": 1.9278, "step": 26712 }, { "epoch": 0.66, "learning_rate": 5.519804621037481e-06, "loss": 1.9572, "step": 26713 }, { "epoch": 0.66, "learning_rate": 5.519090908314066e-06, "loss": 2.0597, "step": 26714 }, { "epoch": 0.66, "learning_rate": 5.5183772241489515e-06, "loss": 1.9003, "step": 26715 }, { "epoch": 0.66, "learning_rate": 5.517663568546674e-06, "loss": 1.9487, "step": 26716 }, { "epoch": 0.66, "learning_rate": 5.516949941511788e-06, "loss": 1.8157, "step": 26717 }, { "epoch": 0.66, "learning_rate": 5.51623634304884e-06, "loss": 1.9033, "step": 26718 }, { "epoch": 0.66, "learning_rate": 5.5155227731623824e-06, "loss": 1.9094, "step": 26719 }, { "epoch": 0.66, "learning_rate": 5.514809231856958e-06, "loss": 2.033, "step": 26720 }, { "epoch": 0.66, "learning_rate": 5.514095719137115e-06, "loss": 2.085, "step": 26721 }, { "epoch": 0.66, "learning_rate": 5.513382235007405e-06, "loss": 1.939, "step": 26722 }, { "epoch": 0.66, "learning_rate": 5.512668779472368e-06, "loss": 2.037, "step": 26723 }, { "epoch": 0.66, "learning_rate": 5.511955352536558e-06, "loss": 2.0109, "step": 26724 }, { "epoch": 0.66, "learning_rate": 5.5112419542045205e-06, "loss": 2.027, "step": 26725 }, { "epoch": 0.66, "learning_rate": 5.510528584480797e-06, "loss": 1.925, "step": 26726 }, { "epoch": 0.66, "learning_rate": 5.509815243369942e-06, "loss": 1.6692, "step": 26727 }, { "epoch": 0.66, "learning_rate": 5.509101930876495e-06, "loss": 2.0515, "step": 26728 }, { "epoch": 0.66, "learning_rate": 5.508388647005004e-06, "loss": 1.9831, "step": 26729 }, { "epoch": 0.66, "learning_rate": 5.50767539176002e-06, "loss": 1.9656, "step": 26730 }, { "epoch": 0.66, "learning_rate": 5.506962165146082e-06, "loss": 2.1346, "step": 26731 }, { "epoch": 0.66, "learning_rate": 5.506248967167736e-06, "loss": 2.0029, "step": 26732 }, { "epoch": 0.66, "learning_rate": 5.505535797829535e-06, "loss": 2.0825, "step": 26733 }, { "epoch": 0.66, "learning_rate": 5.504822657136015e-06, "loss": 2.0875, "step": 26734 }, { "epoch": 0.66, "learning_rate": 5.504109545091725e-06, "loss": 1.8249, "step": 26735 }, { "epoch": 0.66, "learning_rate": 5.5033964617012114e-06, "loss": 1.9105, "step": 26736 }, { "epoch": 0.66, "learning_rate": 5.50268340696902e-06, "loss": 1.8415, "step": 26737 }, { "epoch": 0.66, "learning_rate": 5.501970380899689e-06, "loss": 1.9771, "step": 26738 }, { "epoch": 0.66, "learning_rate": 5.501257383497767e-06, "loss": 2.0101, "step": 26739 }, { "epoch": 0.66, "learning_rate": 5.500544414767801e-06, "loss": 1.9788, "step": 26740 }, { "epoch": 0.66, "learning_rate": 5.49983147471433e-06, "loss": 1.9225, "step": 26741 }, { "epoch": 0.66, "learning_rate": 5.499118563341897e-06, "loss": 1.8401, "step": 26742 }, { "epoch": 0.66, "learning_rate": 5.498405680655054e-06, "loss": 1.9692, "step": 26743 }, { "epoch": 0.66, "learning_rate": 5.497692826658334e-06, "loss": 1.9407, "step": 26744 }, { "epoch": 0.66, "learning_rate": 5.496980001356288e-06, "loss": 1.9001, "step": 26745 }, { "epoch": 0.66, "learning_rate": 5.496267204753453e-06, "loss": 2.0467, "step": 26746 }, { "epoch": 0.66, "learning_rate": 5.495554436854375e-06, "loss": 1.8723, "step": 26747 }, { "epoch": 0.66, "learning_rate": 5.4948416976636e-06, "loss": 1.8671, "step": 26748 }, { "epoch": 0.66, "learning_rate": 5.494128987185663e-06, "loss": 1.8657, "step": 26749 }, { "epoch": 0.66, "learning_rate": 5.493416305425115e-06, "loss": 2.03, "step": 26750 }, { "epoch": 0.66, "learning_rate": 5.4927036523864885e-06, "loss": 1.8033, "step": 26751 }, { "epoch": 0.66, "learning_rate": 5.491991028074331e-06, "loss": 1.8032, "step": 26752 }, { "epoch": 0.66, "learning_rate": 5.491278432493183e-06, "loss": 1.9722, "step": 26753 }, { "epoch": 0.66, "learning_rate": 5.490565865647588e-06, "loss": 2.0896, "step": 26754 }, { "epoch": 0.66, "learning_rate": 5.48985332754209e-06, "loss": 1.9561, "step": 26755 }, { "epoch": 0.66, "learning_rate": 5.489140818181222e-06, "loss": 2.1331, "step": 26756 }, { "epoch": 0.66, "learning_rate": 5.488428337569531e-06, "loss": 1.8605, "step": 26757 }, { "epoch": 0.66, "learning_rate": 5.48771588571156e-06, "loss": 2.0112, "step": 26758 }, { "epoch": 0.66, "learning_rate": 5.487003462611842e-06, "loss": 1.9768, "step": 26759 }, { "epoch": 0.66, "learning_rate": 5.4862910682749245e-06, "loss": 1.7939, "step": 26760 }, { "epoch": 0.66, "learning_rate": 5.4855787027053435e-06, "loss": 1.9148, "step": 26761 }, { "epoch": 0.66, "learning_rate": 5.484866365907639e-06, "loss": 2.0533, "step": 26762 }, { "epoch": 0.66, "learning_rate": 5.484154057886356e-06, "loss": 1.995, "step": 26763 }, { "epoch": 0.66, "learning_rate": 5.483441778646028e-06, "loss": 2.0096, "step": 26764 }, { "epoch": 0.66, "learning_rate": 5.4827295281912e-06, "loss": 1.9161, "step": 26765 }, { "epoch": 0.66, "learning_rate": 5.482017306526406e-06, "loss": 2.2114, "step": 26766 }, { "epoch": 0.66, "learning_rate": 5.481305113656188e-06, "loss": 1.8549, "step": 26767 }, { "epoch": 0.66, "learning_rate": 5.48059294958509e-06, "loss": 1.9093, "step": 26768 }, { "epoch": 0.66, "learning_rate": 5.479880814317641e-06, "loss": 2.0184, "step": 26769 }, { "epoch": 0.66, "learning_rate": 5.479168707858384e-06, "loss": 2.0304, "step": 26770 }, { "epoch": 0.66, "learning_rate": 5.478456630211859e-06, "loss": 1.9704, "step": 26771 }, { "epoch": 0.66, "learning_rate": 5.477744581382604e-06, "loss": 2.0899, "step": 26772 }, { "epoch": 0.66, "learning_rate": 5.4770325613751595e-06, "loss": 2.0334, "step": 26773 }, { "epoch": 0.66, "learning_rate": 5.476320570194057e-06, "loss": 2.0006, "step": 26774 }, { "epoch": 0.66, "learning_rate": 5.475608607843841e-06, "loss": 2.135, "step": 26775 }, { "epoch": 0.66, "learning_rate": 5.474896674329042e-06, "loss": 1.9616, "step": 26776 }, { "epoch": 0.66, "learning_rate": 5.474184769654203e-06, "loss": 1.9475, "step": 26777 }, { "epoch": 0.66, "learning_rate": 5.4734728938238615e-06, "loss": 2.0001, "step": 26778 }, { "epoch": 0.66, "learning_rate": 5.472761046842548e-06, "loss": 2.0145, "step": 26779 }, { "epoch": 0.66, "learning_rate": 5.472049228714808e-06, "loss": 2.0504, "step": 26780 }, { "epoch": 0.66, "learning_rate": 5.471337439445171e-06, "loss": 1.9894, "step": 26781 }, { "epoch": 0.66, "learning_rate": 5.470625679038175e-06, "loss": 2.0472, "step": 26782 }, { "epoch": 0.66, "learning_rate": 5.4699139474983635e-06, "loss": 1.9379, "step": 26783 }, { "epoch": 0.66, "learning_rate": 5.469202244830261e-06, "loss": 2.0166, "step": 26784 }, { "epoch": 0.66, "learning_rate": 5.468490571038414e-06, "loss": 1.9323, "step": 26785 }, { "epoch": 0.66, "learning_rate": 5.467778926127349e-06, "loss": 1.9369, "step": 26786 }, { "epoch": 0.66, "learning_rate": 5.467067310101606e-06, "loss": 2.0047, "step": 26787 }, { "epoch": 0.66, "learning_rate": 5.466355722965721e-06, "loss": 1.9491, "step": 26788 }, { "epoch": 0.66, "learning_rate": 5.465644164724226e-06, "loss": 1.9217, "step": 26789 }, { "epoch": 0.66, "learning_rate": 5.464932635381665e-06, "loss": 2.0932, "step": 26790 }, { "epoch": 0.66, "learning_rate": 5.464221134942562e-06, "loss": 2.2476, "step": 26791 }, { "epoch": 0.66, "learning_rate": 5.463509663411455e-06, "loss": 2.0537, "step": 26792 }, { "epoch": 0.66, "learning_rate": 5.462798220792884e-06, "loss": 1.8946, "step": 26793 }, { "epoch": 0.66, "learning_rate": 5.4620868070913735e-06, "loss": 1.9481, "step": 26794 }, { "epoch": 0.66, "learning_rate": 5.461375422311463e-06, "loss": 1.8992, "step": 26795 }, { "epoch": 0.66, "learning_rate": 5.460664066457691e-06, "loss": 1.9923, "step": 26796 }, { "epoch": 0.66, "learning_rate": 5.459952739534582e-06, "loss": 1.9192, "step": 26797 }, { "epoch": 0.66, "learning_rate": 5.459241441546677e-06, "loss": 2.1194, "step": 26798 }, { "epoch": 0.66, "learning_rate": 5.458530172498502e-06, "loss": 2.077, "step": 26799 }, { "epoch": 0.66, "learning_rate": 5.4578189323945955e-06, "loss": 1.953, "step": 26800 }, { "epoch": 0.66, "learning_rate": 5.457107721239492e-06, "loss": 2.02, "step": 26801 }, { "epoch": 0.66, "learning_rate": 5.456396539037719e-06, "loss": 1.7749, "step": 26802 }, { "epoch": 0.66, "learning_rate": 5.455685385793813e-06, "loss": 2.1067, "step": 26803 }, { "epoch": 0.66, "learning_rate": 5.4549742615123025e-06, "loss": 1.9915, "step": 26804 }, { "epoch": 0.66, "learning_rate": 5.454263166197722e-06, "loss": 1.9961, "step": 26805 }, { "epoch": 0.66, "learning_rate": 5.453552099854604e-06, "loss": 1.9728, "step": 26806 }, { "epoch": 0.66, "learning_rate": 5.452841062487479e-06, "loss": 1.922, "step": 26807 }, { "epoch": 0.66, "learning_rate": 5.452130054100886e-06, "loss": 2.0679, "step": 26808 }, { "epoch": 0.66, "learning_rate": 5.451419074699345e-06, "loss": 1.879, "step": 26809 }, { "epoch": 0.66, "learning_rate": 5.450708124287392e-06, "loss": 2.144, "step": 26810 }, { "epoch": 0.66, "learning_rate": 5.449997202869563e-06, "loss": 1.9684, "step": 26811 }, { "epoch": 0.66, "learning_rate": 5.449286310450381e-06, "loss": 1.9036, "step": 26812 }, { "epoch": 0.66, "learning_rate": 5.4485754470343855e-06, "loss": 1.9267, "step": 26813 }, { "epoch": 0.66, "learning_rate": 5.447864612626096e-06, "loss": 1.9962, "step": 26814 }, { "epoch": 0.66, "learning_rate": 5.44715380723005e-06, "loss": 2.0893, "step": 26815 }, { "epoch": 0.66, "learning_rate": 5.44644303085078e-06, "loss": 2.0366, "step": 26816 }, { "epoch": 0.66, "learning_rate": 5.4457322834928086e-06, "loss": 1.9891, "step": 26817 }, { "epoch": 0.66, "learning_rate": 5.445021565160673e-06, "loss": 1.893, "step": 26818 }, { "epoch": 0.66, "learning_rate": 5.444310875858896e-06, "loss": 1.9974, "step": 26819 }, { "epoch": 0.66, "learning_rate": 5.443600215592011e-06, "loss": 1.8625, "step": 26820 }, { "epoch": 0.66, "learning_rate": 5.44288958436455e-06, "loss": 2.0622, "step": 26821 }, { "epoch": 0.66, "learning_rate": 5.442178982181034e-06, "loss": 1.9811, "step": 26822 }, { "epoch": 0.66, "learning_rate": 5.441468409045998e-06, "loss": 1.9623, "step": 26823 }, { "epoch": 0.66, "learning_rate": 5.440757864963969e-06, "loss": 2.0045, "step": 26824 }, { "epoch": 0.66, "learning_rate": 5.4400473499394766e-06, "loss": 1.8609, "step": 26825 }, { "epoch": 0.66, "learning_rate": 5.439336863977052e-06, "loss": 1.8927, "step": 26826 }, { "epoch": 0.66, "learning_rate": 5.438626407081216e-06, "loss": 1.989, "step": 26827 }, { "epoch": 0.66, "learning_rate": 5.4379159792565024e-06, "loss": 1.9804, "step": 26828 }, { "epoch": 0.66, "learning_rate": 5.437205580507437e-06, "loss": 1.8589, "step": 26829 }, { "epoch": 0.66, "learning_rate": 5.436495210838544e-06, "loss": 2.0211, "step": 26830 }, { "epoch": 0.66, "learning_rate": 5.435784870254359e-06, "loss": 1.8697, "step": 26831 }, { "epoch": 0.66, "learning_rate": 5.4350745587594016e-06, "loss": 1.9004, "step": 26832 }, { "epoch": 0.66, "learning_rate": 5.434364276358204e-06, "loss": 1.9773, "step": 26833 }, { "epoch": 0.66, "learning_rate": 5.433654023055288e-06, "loss": 1.8856, "step": 26834 }, { "epoch": 0.66, "learning_rate": 5.432943798855182e-06, "loss": 1.9144, "step": 26835 }, { "epoch": 0.66, "learning_rate": 5.432233603762419e-06, "loss": 1.9893, "step": 26836 }, { "epoch": 0.66, "learning_rate": 5.431523437781515e-06, "loss": 2.0044, "step": 26837 }, { "epoch": 0.66, "learning_rate": 5.430813300917005e-06, "loss": 1.9271, "step": 26838 }, { "epoch": 0.66, "learning_rate": 5.430103193173406e-06, "loss": 1.9684, "step": 26839 }, { "epoch": 0.66, "learning_rate": 5.429393114555251e-06, "loss": 1.9703, "step": 26840 }, { "epoch": 0.66, "learning_rate": 5.428683065067062e-06, "loss": 2.0069, "step": 26841 }, { "epoch": 0.66, "learning_rate": 5.427973044713364e-06, "loss": 1.9357, "step": 26842 }, { "epoch": 0.66, "learning_rate": 5.42726305349869e-06, "loss": 2.0533, "step": 26843 }, { "epoch": 0.66, "learning_rate": 5.4265530914275535e-06, "loss": 2.0609, "step": 26844 }, { "epoch": 0.66, "learning_rate": 5.425843158504484e-06, "loss": 2.0084, "step": 26845 }, { "epoch": 0.66, "learning_rate": 5.425133254734012e-06, "loss": 2.0353, "step": 26846 }, { "epoch": 0.66, "learning_rate": 5.424423380120654e-06, "loss": 2.006, "step": 26847 }, { "epoch": 0.66, "learning_rate": 5.423713534668939e-06, "loss": 2.0232, "step": 26848 }, { "epoch": 0.66, "learning_rate": 5.423003718383384e-06, "loss": 2.0779, "step": 26849 }, { "epoch": 0.66, "learning_rate": 5.4222939312685195e-06, "loss": 1.8213, "step": 26850 }, { "epoch": 0.66, "learning_rate": 5.421584173328871e-06, "loss": 2.2363, "step": 26851 }, { "epoch": 0.66, "learning_rate": 5.4208744445689555e-06, "loss": 1.9134, "step": 26852 }, { "epoch": 0.66, "learning_rate": 5.4201647449933e-06, "loss": 2.0851, "step": 26853 }, { "epoch": 0.66, "learning_rate": 5.419455074606429e-06, "loss": 2.0766, "step": 26854 }, { "epoch": 0.66, "learning_rate": 5.418745433412859e-06, "loss": 1.9167, "step": 26855 }, { "epoch": 0.66, "learning_rate": 5.418035821417125e-06, "loss": 2.1045, "step": 26856 }, { "epoch": 0.66, "learning_rate": 5.417326238623735e-06, "loss": 1.924, "step": 26857 }, { "epoch": 0.66, "learning_rate": 5.416616685037219e-06, "loss": 2.0072, "step": 26858 }, { "epoch": 0.66, "learning_rate": 5.415907160662099e-06, "loss": 1.8029, "step": 26859 }, { "epoch": 0.66, "learning_rate": 5.415197665502896e-06, "loss": 2.0503, "step": 26860 }, { "epoch": 0.66, "learning_rate": 5.414488199564137e-06, "loss": 2.0418, "step": 26861 }, { "epoch": 0.66, "learning_rate": 5.413778762850334e-06, "loss": 2.0186, "step": 26862 }, { "epoch": 0.66, "learning_rate": 5.4130693553660154e-06, "loss": 1.9989, "step": 26863 }, { "epoch": 0.66, "learning_rate": 5.412359977115704e-06, "loss": 1.7107, "step": 26864 }, { "epoch": 0.66, "learning_rate": 5.4116506281039126e-06, "loss": 2.0022, "step": 26865 }, { "epoch": 0.66, "learning_rate": 5.4109413083351714e-06, "loss": 2.002, "step": 26866 }, { "epoch": 0.66, "learning_rate": 5.4102320178139935e-06, "loss": 2.0724, "step": 26867 }, { "epoch": 0.66, "learning_rate": 5.409522756544902e-06, "loss": 2.0281, "step": 26868 }, { "epoch": 0.66, "learning_rate": 5.408813524532421e-06, "loss": 1.9825, "step": 26869 }, { "epoch": 0.66, "learning_rate": 5.408104321781066e-06, "loss": 2.0944, "step": 26870 }, { "epoch": 0.66, "learning_rate": 5.407395148295361e-06, "loss": 2.0069, "step": 26871 }, { "epoch": 0.66, "learning_rate": 5.40668600407982e-06, "loss": 1.9654, "step": 26872 }, { "epoch": 0.66, "learning_rate": 5.405976889138967e-06, "loss": 2.073, "step": 26873 }, { "epoch": 0.66, "learning_rate": 5.405267803477322e-06, "loss": 1.9966, "step": 26874 }, { "epoch": 0.66, "learning_rate": 5.4045587470994e-06, "loss": 1.893, "step": 26875 }, { "epoch": 0.66, "learning_rate": 5.403849720009722e-06, "loss": 2.0275, "step": 26876 }, { "epoch": 0.66, "learning_rate": 5.403140722212809e-06, "loss": 1.9534, "step": 26877 }, { "epoch": 0.66, "learning_rate": 5.402431753713178e-06, "loss": 1.8418, "step": 26878 }, { "epoch": 0.66, "learning_rate": 5.40172281451535e-06, "loss": 2.0561, "step": 26879 }, { "epoch": 0.66, "learning_rate": 5.401013904623838e-06, "loss": 1.9893, "step": 26880 }, { "epoch": 0.66, "learning_rate": 5.4003050240431676e-06, "loss": 1.7851, "step": 26881 }, { "epoch": 0.66, "learning_rate": 5.399596172777848e-06, "loss": 2.0044, "step": 26882 }, { "epoch": 0.66, "learning_rate": 5.3988873508324015e-06, "loss": 1.9592, "step": 26883 }, { "epoch": 0.66, "learning_rate": 5.398178558211349e-06, "loss": 1.5279, "step": 26884 }, { "epoch": 0.66, "learning_rate": 5.397469794919201e-06, "loss": 2.0465, "step": 26885 }, { "epoch": 0.66, "learning_rate": 5.396761060960481e-06, "loss": 1.8621, "step": 26886 }, { "epoch": 0.66, "learning_rate": 5.396052356339699e-06, "loss": 1.9839, "step": 26887 }, { "epoch": 0.66, "learning_rate": 5.395343681061375e-06, "loss": 1.9364, "step": 26888 }, { "epoch": 0.66, "learning_rate": 5.3946350351300315e-06, "loss": 1.8229, "step": 26889 }, { "epoch": 0.66, "learning_rate": 5.393926418550176e-06, "loss": 2.0602, "step": 26890 }, { "epoch": 0.66, "learning_rate": 5.393217831326332e-06, "loss": 1.9301, "step": 26891 }, { "epoch": 0.66, "learning_rate": 5.392509273463008e-06, "loss": 1.9607, "step": 26892 }, { "epoch": 0.66, "learning_rate": 5.391800744964724e-06, "loss": 2.0342, "step": 26893 }, { "epoch": 0.66, "learning_rate": 5.391092245835996e-06, "loss": 1.7517, "step": 26894 }, { "epoch": 0.66, "learning_rate": 5.3903837760813386e-06, "loss": 2.0786, "step": 26895 }, { "epoch": 0.66, "learning_rate": 5.389675335705271e-06, "loss": 1.9793, "step": 26896 }, { "epoch": 0.66, "learning_rate": 5.3889669247123025e-06, "loss": 2.0554, "step": 26897 }, { "epoch": 0.66, "learning_rate": 5.38825854310695e-06, "loss": 2.067, "step": 26898 }, { "epoch": 0.66, "learning_rate": 5.387550190893732e-06, "loss": 2.0622, "step": 26899 }, { "epoch": 0.66, "learning_rate": 5.386841868077158e-06, "loss": 1.9261, "step": 26900 }, { "epoch": 0.66, "learning_rate": 5.3861335746617445e-06, "loss": 2.102, "step": 26901 }, { "epoch": 0.66, "learning_rate": 5.3854253106520034e-06, "loss": 1.9655, "step": 26902 }, { "epoch": 0.66, "learning_rate": 5.38471707605245e-06, "loss": 1.9554, "step": 26903 }, { "epoch": 0.66, "learning_rate": 5.384008870867604e-06, "loss": 1.8938, "step": 26904 }, { "epoch": 0.66, "learning_rate": 5.3833006951019695e-06, "loss": 1.9241, "step": 26905 }, { "epoch": 0.66, "learning_rate": 5.382592548760067e-06, "loss": 1.906, "step": 26906 }, { "epoch": 0.66, "learning_rate": 5.3818844318464045e-06, "loss": 1.8858, "step": 26907 }, { "epoch": 0.66, "learning_rate": 5.381176344365496e-06, "loss": 2.1117, "step": 26908 }, { "epoch": 0.66, "learning_rate": 5.380468286321861e-06, "loss": 1.9557, "step": 26909 }, { "epoch": 0.66, "learning_rate": 5.3797602577200024e-06, "loss": 1.9678, "step": 26910 }, { "epoch": 0.66, "learning_rate": 5.3790522585644375e-06, "loss": 2.0077, "step": 26911 }, { "epoch": 0.66, "learning_rate": 5.378344288859678e-06, "loss": 1.979, "step": 26912 }, { "epoch": 0.66, "learning_rate": 5.377636348610238e-06, "loss": 1.9774, "step": 26913 }, { "epoch": 0.66, "learning_rate": 5.376928437820631e-06, "loss": 2.096, "step": 26914 }, { "epoch": 0.66, "learning_rate": 5.376220556495362e-06, "loss": 2.1499, "step": 26915 }, { "epoch": 0.66, "learning_rate": 5.375512704638946e-06, "loss": 1.9682, "step": 26916 }, { "epoch": 0.66, "learning_rate": 5.374804882255898e-06, "loss": 1.9267, "step": 26917 }, { "epoch": 0.66, "learning_rate": 5.374097089350722e-06, "loss": 2.0012, "step": 26918 }, { "epoch": 0.66, "learning_rate": 5.3733893259279375e-06, "loss": 1.9492, "step": 26919 }, { "epoch": 0.66, "learning_rate": 5.372681591992046e-06, "loss": 1.9159, "step": 26920 }, { "epoch": 0.66, "learning_rate": 5.371973887547564e-06, "loss": 1.858, "step": 26921 }, { "epoch": 0.66, "learning_rate": 5.371266212599003e-06, "loss": 1.8888, "step": 26922 }, { "epoch": 0.66, "learning_rate": 5.370558567150868e-06, "loss": 2.0424, "step": 26923 }, { "epoch": 0.66, "learning_rate": 5.369850951207676e-06, "loss": 1.8989, "step": 26924 }, { "epoch": 0.66, "learning_rate": 5.369143364773927e-06, "loss": 1.963, "step": 26925 }, { "epoch": 0.66, "learning_rate": 5.368435807854138e-06, "loss": 1.9111, "step": 26926 }, { "epoch": 0.66, "learning_rate": 5.367728280452821e-06, "loss": 1.9574, "step": 26927 }, { "epoch": 0.66, "learning_rate": 5.367020782574477e-06, "loss": 1.9664, "step": 26928 }, { "epoch": 0.66, "learning_rate": 5.366313314223619e-06, "loss": 1.9059, "step": 26929 }, { "epoch": 0.66, "learning_rate": 5.3656058754047566e-06, "loss": 2.0504, "step": 26930 }, { "epoch": 0.66, "learning_rate": 5.364898466122399e-06, "loss": 1.9515, "step": 26931 }, { "epoch": 0.66, "learning_rate": 5.364191086381057e-06, "loss": 1.9729, "step": 26932 }, { "epoch": 0.66, "learning_rate": 5.363483736185233e-06, "loss": 1.9282, "step": 26933 }, { "epoch": 0.66, "learning_rate": 5.362776415539442e-06, "loss": 2.0779, "step": 26934 }, { "epoch": 0.66, "learning_rate": 5.362069124448183e-06, "loss": 2.1872, "step": 26935 }, { "epoch": 0.66, "learning_rate": 5.3613618629159695e-06, "loss": 1.9264, "step": 26936 }, { "epoch": 0.66, "learning_rate": 5.3606546309473125e-06, "loss": 2.2207, "step": 26937 }, { "epoch": 0.66, "learning_rate": 5.359947428546711e-06, "loss": 1.8432, "step": 26938 }, { "epoch": 0.66, "learning_rate": 5.3592402557186815e-06, "loss": 1.9822, "step": 26939 }, { "epoch": 0.66, "learning_rate": 5.358533112467722e-06, "loss": 1.8493, "step": 26940 }, { "epoch": 0.66, "learning_rate": 5.357825998798344e-06, "loss": 2.0133, "step": 26941 }, { "epoch": 0.66, "learning_rate": 5.357118914715057e-06, "loss": 1.9567, "step": 26942 }, { "epoch": 0.66, "learning_rate": 5.356411860222362e-06, "loss": 1.9104, "step": 26943 }, { "epoch": 0.66, "learning_rate": 5.355704835324769e-06, "loss": 1.8624, "step": 26944 }, { "epoch": 0.66, "learning_rate": 5.354997840026779e-06, "loss": 1.9488, "step": 26945 }, { "epoch": 0.66, "learning_rate": 5.354290874332903e-06, "loss": 2.1067, "step": 26946 }, { "epoch": 0.66, "learning_rate": 5.353583938247644e-06, "loss": 2.1227, "step": 26947 }, { "epoch": 0.66, "learning_rate": 5.35287703177551e-06, "loss": 2.0417, "step": 26948 }, { "epoch": 0.66, "learning_rate": 5.352170154921008e-06, "loss": 2.0027, "step": 26949 }, { "epoch": 0.66, "learning_rate": 5.351463307688636e-06, "loss": 1.9453, "step": 26950 }, { "epoch": 0.66, "learning_rate": 5.350756490082904e-06, "loss": 1.9355, "step": 26951 }, { "epoch": 0.66, "learning_rate": 5.35004970210832e-06, "loss": 1.8162, "step": 26952 }, { "epoch": 0.66, "learning_rate": 5.34934294376938e-06, "loss": 2.0162, "step": 26953 }, { "epoch": 0.66, "learning_rate": 5.348636215070598e-06, "loss": 1.9178, "step": 26954 }, { "epoch": 0.66, "learning_rate": 5.347929516016469e-06, "loss": 2.0277, "step": 26955 }, { "epoch": 0.66, "learning_rate": 5.347222846611501e-06, "loss": 1.8827, "step": 26956 }, { "epoch": 0.66, "learning_rate": 5.346516206860202e-06, "loss": 2.019, "step": 26957 }, { "epoch": 0.66, "learning_rate": 5.345809596767069e-06, "loss": 1.8187, "step": 26958 }, { "epoch": 0.66, "learning_rate": 5.34510301633661e-06, "loss": 1.9474, "step": 26959 }, { "epoch": 0.66, "learning_rate": 5.344396465573325e-06, "loss": 2.1072, "step": 26960 }, { "epoch": 0.66, "learning_rate": 5.343689944481716e-06, "loss": 2.1231, "step": 26961 }, { "epoch": 0.66, "learning_rate": 5.342983453066294e-06, "loss": 1.8046, "step": 26962 }, { "epoch": 0.66, "learning_rate": 5.342276991331552e-06, "loss": 1.9599, "step": 26963 }, { "epoch": 0.66, "learning_rate": 5.341570559281997e-06, "loss": 1.9432, "step": 26964 }, { "epoch": 0.66, "learning_rate": 5.34086415692213e-06, "loss": 1.9292, "step": 26965 }, { "epoch": 0.66, "learning_rate": 5.340157784256455e-06, "loss": 1.9322, "step": 26966 }, { "epoch": 0.66, "learning_rate": 5.339451441289475e-06, "loss": 1.9942, "step": 26967 }, { "epoch": 0.66, "learning_rate": 5.338745128025688e-06, "loss": 2.0268, "step": 26968 }, { "epoch": 0.66, "learning_rate": 5.338038844469595e-06, "loss": 1.9339, "step": 26969 }, { "epoch": 0.66, "learning_rate": 5.337332590625705e-06, "loss": 1.991, "step": 26970 }, { "epoch": 0.66, "learning_rate": 5.336626366498512e-06, "loss": 2.0894, "step": 26971 }, { "epoch": 0.66, "learning_rate": 5.335920172092521e-06, "loss": 1.9549, "step": 26972 }, { "epoch": 0.66, "learning_rate": 5.335214007412226e-06, "loss": 1.922, "step": 26973 }, { "epoch": 0.66, "learning_rate": 5.334507872462134e-06, "loss": 1.8411, "step": 26974 }, { "epoch": 0.66, "learning_rate": 5.3338017672467464e-06, "loss": 1.8759, "step": 26975 }, { "epoch": 0.66, "learning_rate": 5.333095691770558e-06, "loss": 1.9906, "step": 26976 }, { "epoch": 0.66, "learning_rate": 5.332389646038074e-06, "loss": 1.8674, "step": 26977 }, { "epoch": 0.66, "learning_rate": 5.33168363005379e-06, "loss": 1.9372, "step": 26978 }, { "epoch": 0.66, "learning_rate": 5.330977643822206e-06, "loss": 2.1547, "step": 26979 }, { "epoch": 0.67, "learning_rate": 5.330271687347828e-06, "loss": 2.1016, "step": 26980 }, { "epoch": 0.67, "learning_rate": 5.329565760635146e-06, "loss": 1.9535, "step": 26981 }, { "epoch": 0.67, "learning_rate": 5.328859863688664e-06, "loss": 1.9716, "step": 26982 }, { "epoch": 0.67, "learning_rate": 5.328153996512879e-06, "loss": 1.8691, "step": 26983 }, { "epoch": 0.67, "learning_rate": 5.327448159112292e-06, "loss": 1.9752, "step": 26984 }, { "epoch": 0.67, "learning_rate": 5.3267423514914055e-06, "loss": 2.1341, "step": 26985 }, { "epoch": 0.67, "learning_rate": 5.326036573654707e-06, "loss": 2.0453, "step": 26986 }, { "epoch": 0.67, "learning_rate": 5.325330825606706e-06, "loss": 2.0212, "step": 26987 }, { "epoch": 0.67, "learning_rate": 5.324625107351891e-06, "loss": 2.041, "step": 26988 }, { "epoch": 0.67, "learning_rate": 5.323919418894763e-06, "loss": 1.8903, "step": 26989 }, { "epoch": 0.67, "learning_rate": 5.323213760239825e-06, "loss": 1.9682, "step": 26990 }, { "epoch": 0.67, "learning_rate": 5.322508131391566e-06, "loss": 1.8749, "step": 26991 }, { "epoch": 0.67, "learning_rate": 5.321802532354491e-06, "loss": 1.9154, "step": 26992 }, { "epoch": 0.67, "learning_rate": 5.321096963133087e-06, "loss": 2.14, "step": 26993 }, { "epoch": 0.67, "learning_rate": 5.320391423731861e-06, "loss": 1.852, "step": 26994 }, { "epoch": 0.67, "learning_rate": 5.319685914155306e-06, "loss": 2.1439, "step": 26995 }, { "epoch": 0.67, "learning_rate": 5.318980434407915e-06, "loss": 1.8269, "step": 26996 }, { "epoch": 0.67, "learning_rate": 5.3182749844941915e-06, "loss": 1.8882, "step": 26997 }, { "epoch": 0.67, "learning_rate": 5.317569564418623e-06, "loss": 2.0752, "step": 26998 }, { "epoch": 0.67, "learning_rate": 5.31686417418571e-06, "loss": 2.0688, "step": 26999 }, { "epoch": 0.67, "learning_rate": 5.316158813799948e-06, "loss": 1.9603, "step": 27000 }, { "epoch": 0.67, "learning_rate": 5.315453483265831e-06, "loss": 1.8982, "step": 27001 }, { "epoch": 0.67, "learning_rate": 5.314748182587861e-06, "loss": 1.9976, "step": 27002 }, { "epoch": 0.67, "learning_rate": 5.314042911770523e-06, "loss": 2.0956, "step": 27003 }, { "epoch": 0.67, "learning_rate": 5.313337670818317e-06, "loss": 1.8469, "step": 27004 }, { "epoch": 0.67, "learning_rate": 5.312632459735741e-06, "loss": 2.0311, "step": 27005 }, { "epoch": 0.67, "learning_rate": 5.311927278527282e-06, "loss": 2.2283, "step": 27006 }, { "epoch": 0.67, "learning_rate": 5.311222127197443e-06, "loss": 1.9424, "step": 27007 }, { "epoch": 0.67, "learning_rate": 5.310517005750708e-06, "loss": 1.8531, "step": 27008 }, { "epoch": 0.67, "learning_rate": 5.309811914191578e-06, "loss": 2.0446, "step": 27009 }, { "epoch": 0.67, "learning_rate": 5.3091068525245484e-06, "loss": 1.8859, "step": 27010 }, { "epoch": 0.67, "learning_rate": 5.308401820754105e-06, "loss": 1.9104, "step": 27011 }, { "epoch": 0.67, "learning_rate": 5.30769681888475e-06, "loss": 1.966, "step": 27012 }, { "epoch": 0.67, "learning_rate": 5.306991846920968e-06, "loss": 2.1046, "step": 27013 }, { "epoch": 0.67, "learning_rate": 5.306286904867257e-06, "loss": 1.8721, "step": 27014 }, { "epoch": 0.67, "learning_rate": 5.305581992728114e-06, "loss": 1.8965, "step": 27015 }, { "epoch": 0.67, "learning_rate": 5.304877110508022e-06, "loss": 1.8429, "step": 27016 }, { "epoch": 0.67, "learning_rate": 5.304172258211478e-06, "loss": 2.0078, "step": 27017 }, { "epoch": 0.67, "learning_rate": 5.303467435842975e-06, "loss": 1.978, "step": 27018 }, { "epoch": 0.67, "learning_rate": 5.302762643407003e-06, "loss": 1.8969, "step": 27019 }, { "epoch": 0.67, "learning_rate": 5.302057880908062e-06, "loss": 1.9839, "step": 27020 }, { "epoch": 0.67, "learning_rate": 5.301353148350632e-06, "loss": 1.9534, "step": 27021 }, { "epoch": 0.67, "learning_rate": 5.300648445739212e-06, "loss": 1.9673, "step": 27022 }, { "epoch": 0.67, "learning_rate": 5.299943773078289e-06, "loss": 1.8335, "step": 27023 }, { "epoch": 0.67, "learning_rate": 5.299239130372356e-06, "loss": 1.9014, "step": 27024 }, { "epoch": 0.67, "learning_rate": 5.298534517625907e-06, "loss": 2.0034, "step": 27025 }, { "epoch": 0.67, "learning_rate": 5.297829934843426e-06, "loss": 1.8414, "step": 27026 }, { "epoch": 0.67, "learning_rate": 5.297125382029411e-06, "loss": 1.9672, "step": 27027 }, { "epoch": 0.67, "learning_rate": 5.296420859188345e-06, "loss": 1.9248, "step": 27028 }, { "epoch": 0.67, "learning_rate": 5.295716366324721e-06, "loss": 2.1442, "step": 27029 }, { "epoch": 0.67, "learning_rate": 5.2950119034430344e-06, "loss": 2.0228, "step": 27030 }, { "epoch": 0.67, "learning_rate": 5.294307470547767e-06, "loss": 1.9731, "step": 27031 }, { "epoch": 0.67, "learning_rate": 5.293603067643411e-06, "loss": 1.9694, "step": 27032 }, { "epoch": 0.67, "learning_rate": 5.292898694734461e-06, "loss": 1.9488, "step": 27033 }, { "epoch": 0.67, "learning_rate": 5.292194351825396e-06, "loss": 1.9683, "step": 27034 }, { "epoch": 0.67, "learning_rate": 5.291490038920711e-06, "loss": 2.0118, "step": 27035 }, { "epoch": 0.67, "learning_rate": 5.290785756024896e-06, "loss": 2.0007, "step": 27036 }, { "epoch": 0.67, "learning_rate": 5.290081503142437e-06, "loss": 1.8954, "step": 27037 }, { "epoch": 0.67, "learning_rate": 5.289377280277826e-06, "loss": 1.8782, "step": 27038 }, { "epoch": 0.67, "learning_rate": 5.288673087435547e-06, "loss": 1.8583, "step": 27039 }, { "epoch": 0.67, "learning_rate": 5.287968924620093e-06, "loss": 2.1506, "step": 27040 }, { "epoch": 0.67, "learning_rate": 5.2872647918359445e-06, "loss": 1.8564, "step": 27041 }, { "epoch": 0.67, "learning_rate": 5.286560689087595e-06, "loss": 2.057, "step": 27042 }, { "epoch": 0.67, "learning_rate": 5.285856616379531e-06, "loss": 1.8827, "step": 27043 }, { "epoch": 0.67, "learning_rate": 5.285152573716237e-06, "loss": 2.0724, "step": 27044 }, { "epoch": 0.67, "learning_rate": 5.2844485611022064e-06, "loss": 1.894, "step": 27045 }, { "epoch": 0.67, "learning_rate": 5.283744578541917e-06, "loss": 2.0665, "step": 27046 }, { "epoch": 0.67, "learning_rate": 5.283040626039861e-06, "loss": 2.0508, "step": 27047 }, { "epoch": 0.67, "learning_rate": 5.282336703600528e-06, "loss": 2.0047, "step": 27048 }, { "epoch": 0.67, "learning_rate": 5.281632811228397e-06, "loss": 2.0226, "step": 27049 }, { "epoch": 0.67, "learning_rate": 5.280928948927961e-06, "loss": 2.0653, "step": 27050 }, { "epoch": 0.67, "learning_rate": 5.280225116703699e-06, "loss": 2.0877, "step": 27051 }, { "epoch": 0.67, "learning_rate": 5.279521314560101e-06, "loss": 2.0335, "step": 27052 }, { "epoch": 0.67, "learning_rate": 5.2788175425016515e-06, "loss": 2.0585, "step": 27053 }, { "epoch": 0.67, "learning_rate": 5.278113800532838e-06, "loss": 1.7943, "step": 27054 }, { "epoch": 0.67, "learning_rate": 5.277410088658146e-06, "loss": 1.9994, "step": 27055 }, { "epoch": 0.67, "learning_rate": 5.276706406882055e-06, "loss": 1.9423, "step": 27056 }, { "epoch": 0.67, "learning_rate": 5.276002755209053e-06, "loss": 2.0298, "step": 27057 }, { "epoch": 0.67, "learning_rate": 5.27529913364363e-06, "loss": 1.9937, "step": 27058 }, { "epoch": 0.67, "learning_rate": 5.274595542190262e-06, "loss": 1.9998, "step": 27059 }, { "epoch": 0.67, "learning_rate": 5.273891980853438e-06, "loss": 2.0019, "step": 27060 }, { "epoch": 0.67, "learning_rate": 5.2731884496376385e-06, "loss": 1.9238, "step": 27061 }, { "epoch": 0.67, "learning_rate": 5.272484948547349e-06, "loss": 1.9036, "step": 27062 }, { "epoch": 0.67, "learning_rate": 5.271781477587058e-06, "loss": 1.8155, "step": 27063 }, { "epoch": 0.67, "learning_rate": 5.27107803676124e-06, "loss": 2.1479, "step": 27064 }, { "epoch": 0.67, "learning_rate": 5.270374626074387e-06, "loss": 1.9574, "step": 27065 }, { "epoch": 0.67, "learning_rate": 5.269671245530974e-06, "loss": 1.9994, "step": 27066 }, { "epoch": 0.67, "learning_rate": 5.268967895135487e-06, "loss": 1.9047, "step": 27067 }, { "epoch": 0.67, "learning_rate": 5.268264574892414e-06, "loss": 1.9696, "step": 27068 }, { "epoch": 0.67, "learning_rate": 5.267561284806229e-06, "loss": 1.9687, "step": 27069 }, { "epoch": 0.67, "learning_rate": 5.266858024881418e-06, "loss": 1.9093, "step": 27070 }, { "epoch": 0.67, "learning_rate": 5.2661547951224624e-06, "loss": 1.8679, "step": 27071 }, { "epoch": 0.67, "learning_rate": 5.2654515955338456e-06, "loss": 1.9042, "step": 27072 }, { "epoch": 0.67, "learning_rate": 5.264748426120054e-06, "loss": 1.9677, "step": 27073 }, { "epoch": 0.67, "learning_rate": 5.264045286885559e-06, "loss": 1.9743, "step": 27074 }, { "epoch": 0.67, "learning_rate": 5.26334217783485e-06, "loss": 2.1349, "step": 27075 }, { "epoch": 0.67, "learning_rate": 5.262639098972402e-06, "loss": 1.9176, "step": 27076 }, { "epoch": 0.67, "learning_rate": 5.261936050302699e-06, "loss": 1.9105, "step": 27077 }, { "epoch": 0.67, "learning_rate": 5.261233031830224e-06, "loss": 2.0196, "step": 27078 }, { "epoch": 0.67, "learning_rate": 5.260530043559453e-06, "loss": 2.0367, "step": 27079 }, { "epoch": 0.67, "learning_rate": 5.2598270854948704e-06, "loss": 1.7374, "step": 27080 }, { "epoch": 0.67, "learning_rate": 5.259124157640954e-06, "loss": 1.9039, "step": 27081 }, { "epoch": 0.67, "learning_rate": 5.258421260002182e-06, "loss": 2.0499, "step": 27082 }, { "epoch": 0.67, "learning_rate": 5.25771839258304e-06, "loss": 2.0775, "step": 27083 }, { "epoch": 0.67, "learning_rate": 5.257015555388001e-06, "loss": 2.1555, "step": 27084 }, { "epoch": 0.67, "learning_rate": 5.256312748421548e-06, "loss": 1.897, "step": 27085 }, { "epoch": 0.67, "learning_rate": 5.2556099716881634e-06, "loss": 2.0308, "step": 27086 }, { "epoch": 0.67, "learning_rate": 5.2549072251923185e-06, "loss": 1.8236, "step": 27087 }, { "epoch": 0.67, "learning_rate": 5.254204508938497e-06, "loss": 1.834, "step": 27088 }, { "epoch": 0.67, "learning_rate": 5.253501822931175e-06, "loss": 1.9104, "step": 27089 }, { "epoch": 0.67, "learning_rate": 5.252799167174834e-06, "loss": 1.9964, "step": 27090 }, { "epoch": 0.67, "learning_rate": 5.2520965416739554e-06, "loss": 2.0252, "step": 27091 }, { "epoch": 0.67, "learning_rate": 5.2513939464330076e-06, "loss": 2.0057, "step": 27092 }, { "epoch": 0.67, "learning_rate": 5.250691381456478e-06, "loss": 1.8257, "step": 27093 }, { "epoch": 0.67, "learning_rate": 5.2499888467488365e-06, "loss": 1.8555, "step": 27094 }, { "epoch": 0.67, "learning_rate": 5.249286342314566e-06, "loss": 1.8676, "step": 27095 }, { "epoch": 0.67, "learning_rate": 5.248583868158142e-06, "loss": 2.0108, "step": 27096 }, { "epoch": 0.67, "learning_rate": 5.24788142428404e-06, "loss": 2.026, "step": 27097 }, { "epoch": 0.67, "learning_rate": 5.247179010696742e-06, "loss": 1.9293, "step": 27098 }, { "epoch": 0.67, "learning_rate": 5.2464766274007166e-06, "loss": 1.8584, "step": 27099 }, { "epoch": 0.67, "learning_rate": 5.245774274400444e-06, "loss": 1.8768, "step": 27100 }, { "epoch": 0.67, "learning_rate": 5.245071951700407e-06, "loss": 1.9219, "step": 27101 }, { "epoch": 0.67, "learning_rate": 5.244369659305071e-06, "loss": 1.9935, "step": 27102 }, { "epoch": 0.67, "learning_rate": 5.243667397218917e-06, "loss": 1.8566, "step": 27103 }, { "epoch": 0.67, "learning_rate": 5.2429651654464255e-06, "loss": 1.8366, "step": 27104 }, { "epoch": 0.67, "learning_rate": 5.242262963992062e-06, "loss": 2.1013, "step": 27105 }, { "epoch": 0.67, "learning_rate": 5.241560792860308e-06, "loss": 1.9221, "step": 27106 }, { "epoch": 0.67, "learning_rate": 5.240858652055637e-06, "loss": 2.119, "step": 27107 }, { "epoch": 0.67, "learning_rate": 5.240156541582529e-06, "loss": 1.88, "step": 27108 }, { "epoch": 0.67, "learning_rate": 5.2394544614454515e-06, "loss": 2.1459, "step": 27109 }, { "epoch": 0.67, "learning_rate": 5.238752411648882e-06, "loss": 1.8935, "step": 27110 }, { "epoch": 0.67, "learning_rate": 5.238050392197298e-06, "loss": 1.9501, "step": 27111 }, { "epoch": 0.67, "learning_rate": 5.237348403095167e-06, "loss": 1.8228, "step": 27112 }, { "epoch": 0.67, "learning_rate": 5.236646444346971e-06, "loss": 2.027, "step": 27113 }, { "epoch": 0.67, "learning_rate": 5.235944515957177e-06, "loss": 1.9221, "step": 27114 }, { "epoch": 0.67, "learning_rate": 5.235242617930258e-06, "loss": 1.8177, "step": 27115 }, { "epoch": 0.67, "learning_rate": 5.234540750270697e-06, "loss": 1.8991, "step": 27116 }, { "epoch": 0.67, "learning_rate": 5.233838912982956e-06, "loss": 1.828, "step": 27117 }, { "epoch": 0.67, "learning_rate": 5.233137106071517e-06, "loss": 2.0606, "step": 27118 }, { "epoch": 0.67, "learning_rate": 5.232435329540844e-06, "loss": 1.9204, "step": 27119 }, { "epoch": 0.67, "learning_rate": 5.231733583395415e-06, "loss": 2.0619, "step": 27120 }, { "epoch": 0.67, "learning_rate": 5.231031867639702e-06, "loss": 1.9927, "step": 27121 }, { "epoch": 0.67, "learning_rate": 5.23033018227818e-06, "loss": 2.039, "step": 27122 }, { "epoch": 0.67, "learning_rate": 5.229628527315316e-06, "loss": 1.7627, "step": 27123 }, { "epoch": 0.67, "learning_rate": 5.2289269027555835e-06, "loss": 2.0106, "step": 27124 }, { "epoch": 0.67, "learning_rate": 5.228225308603454e-06, "loss": 1.9128, "step": 27125 }, { "epoch": 0.67, "learning_rate": 5.227523744863403e-06, "loss": 1.849, "step": 27126 }, { "epoch": 0.67, "learning_rate": 5.226822211539897e-06, "loss": 2.0297, "step": 27127 }, { "epoch": 0.67, "learning_rate": 5.22612070863741e-06, "loss": 1.9754, "step": 27128 }, { "epoch": 0.67, "learning_rate": 5.225419236160409e-06, "loss": 1.7663, "step": 27129 }, { "epoch": 0.67, "learning_rate": 5.224717794113366e-06, "loss": 2.076, "step": 27130 }, { "epoch": 0.67, "learning_rate": 5.224016382500758e-06, "loss": 2.2027, "step": 27131 }, { "epoch": 0.67, "learning_rate": 5.223315001327045e-06, "loss": 2.0552, "step": 27132 }, { "epoch": 0.67, "learning_rate": 5.222613650596706e-06, "loss": 1.8886, "step": 27133 }, { "epoch": 0.67, "learning_rate": 5.221912330314205e-06, "loss": 2.0387, "step": 27134 }, { "epoch": 0.67, "learning_rate": 5.221211040484012e-06, "loss": 2.0236, "step": 27135 }, { "epoch": 0.67, "learning_rate": 5.220509781110603e-06, "loss": 1.8093, "step": 27136 }, { "epoch": 0.67, "learning_rate": 5.219808552198439e-06, "loss": 1.9319, "step": 27137 }, { "epoch": 0.67, "learning_rate": 5.219107353751993e-06, "loss": 1.9006, "step": 27138 }, { "epoch": 0.67, "learning_rate": 5.218406185775737e-06, "loss": 1.9488, "step": 27139 }, { "epoch": 0.67, "learning_rate": 5.2177050482741335e-06, "loss": 1.905, "step": 27140 }, { "epoch": 0.67, "learning_rate": 5.2170039412516535e-06, "loss": 1.8601, "step": 27141 }, { "epoch": 0.67, "learning_rate": 5.216302864712767e-06, "loss": 1.7735, "step": 27142 }, { "epoch": 0.67, "learning_rate": 5.215601818661944e-06, "loss": 1.9592, "step": 27143 }, { "epoch": 0.67, "learning_rate": 5.214900803103647e-06, "loss": 2.0973, "step": 27144 }, { "epoch": 0.67, "learning_rate": 5.214199818042347e-06, "loss": 1.9542, "step": 27145 }, { "epoch": 0.67, "learning_rate": 5.213498863482513e-06, "loss": 1.82, "step": 27146 }, { "epoch": 0.67, "learning_rate": 5.2127979394286085e-06, "loss": 1.8904, "step": 27147 }, { "epoch": 0.67, "learning_rate": 5.212097045885103e-06, "loss": 1.8856, "step": 27148 }, { "epoch": 0.67, "learning_rate": 5.211396182856466e-06, "loss": 1.9358, "step": 27149 }, { "epoch": 0.67, "learning_rate": 5.210695350347159e-06, "loss": 1.898, "step": 27150 }, { "epoch": 0.67, "learning_rate": 5.2099945483616545e-06, "loss": 2.0465, "step": 27151 }, { "epoch": 0.67, "learning_rate": 5.2092937769044115e-06, "loss": 1.9339, "step": 27152 }, { "epoch": 0.67, "learning_rate": 5.208593035979901e-06, "loss": 2.0195, "step": 27153 }, { "epoch": 0.67, "learning_rate": 5.2078923255925924e-06, "loss": 1.9684, "step": 27154 }, { "epoch": 0.67, "learning_rate": 5.207191645746945e-06, "loss": 2.0946, "step": 27155 }, { "epoch": 0.67, "learning_rate": 5.2064909964474265e-06, "loss": 2.0057, "step": 27156 }, { "epoch": 0.67, "learning_rate": 5.205790377698505e-06, "loss": 1.9085, "step": 27157 }, { "epoch": 0.67, "learning_rate": 5.205089789504642e-06, "loss": 1.9756, "step": 27158 }, { "epoch": 0.67, "learning_rate": 5.204389231870304e-06, "loss": 1.9152, "step": 27159 }, { "epoch": 0.67, "learning_rate": 5.203688704799956e-06, "loss": 1.8835, "step": 27160 }, { "epoch": 0.67, "learning_rate": 5.202988208298068e-06, "loss": 1.9952, "step": 27161 }, { "epoch": 0.67, "learning_rate": 5.2022877423690945e-06, "loss": 1.8903, "step": 27162 }, { "epoch": 0.67, "learning_rate": 5.201587307017504e-06, "loss": 2.0318, "step": 27163 }, { "epoch": 0.67, "learning_rate": 5.200886902247766e-06, "loss": 2.0381, "step": 27164 }, { "epoch": 0.67, "learning_rate": 5.200186528064336e-06, "loss": 1.9572, "step": 27165 }, { "epoch": 0.67, "learning_rate": 5.1994861844716845e-06, "loss": 1.917, "step": 27166 }, { "epoch": 0.67, "learning_rate": 5.19878587147427e-06, "loss": 1.8155, "step": 27167 }, { "epoch": 0.67, "learning_rate": 5.198085589076556e-06, "loss": 2.0153, "step": 27168 }, { "epoch": 0.67, "learning_rate": 5.1973853372830116e-06, "loss": 1.9603, "step": 27169 }, { "epoch": 0.67, "learning_rate": 5.196685116098093e-06, "loss": 2.0236, "step": 27170 }, { "epoch": 0.67, "learning_rate": 5.195984925526268e-06, "loss": 1.9287, "step": 27171 }, { "epoch": 0.67, "learning_rate": 5.195284765571993e-06, "loss": 1.9725, "step": 27172 }, { "epoch": 0.67, "learning_rate": 5.194584636239735e-06, "loss": 1.8874, "step": 27173 }, { "epoch": 0.67, "learning_rate": 5.193884537533953e-06, "loss": 1.8935, "step": 27174 }, { "epoch": 0.67, "learning_rate": 5.193184469459117e-06, "loss": 1.8111, "step": 27175 }, { "epoch": 0.67, "learning_rate": 5.1924844320196775e-06, "loss": 2.0095, "step": 27176 }, { "epoch": 0.67, "learning_rate": 5.191784425220103e-06, "loss": 1.8529, "step": 27177 }, { "epoch": 0.67, "learning_rate": 5.191084449064853e-06, "loss": 1.8563, "step": 27178 }, { "epoch": 0.67, "learning_rate": 5.190384503558393e-06, "loss": 1.9212, "step": 27179 }, { "epoch": 0.67, "learning_rate": 5.189684588705175e-06, "loss": 1.9979, "step": 27180 }, { "epoch": 0.67, "learning_rate": 5.18898470450967e-06, "loss": 2.0567, "step": 27181 }, { "epoch": 0.67, "learning_rate": 5.188284850976328e-06, "loss": 1.9883, "step": 27182 }, { "epoch": 0.67, "learning_rate": 5.187585028109615e-06, "loss": 1.9275, "step": 27183 }, { "epoch": 0.67, "learning_rate": 5.186885235913995e-06, "loss": 1.9747, "step": 27184 }, { "epoch": 0.67, "learning_rate": 5.186185474393921e-06, "loss": 1.9908, "step": 27185 }, { "epoch": 0.67, "learning_rate": 5.185485743553859e-06, "loss": 1.885, "step": 27186 }, { "epoch": 0.67, "learning_rate": 5.184786043398261e-06, "loss": 2.0331, "step": 27187 }, { "epoch": 0.67, "learning_rate": 5.184086373931592e-06, "loss": 2.2506, "step": 27188 }, { "epoch": 0.67, "learning_rate": 5.183386735158312e-06, "loss": 1.9563, "step": 27189 }, { "epoch": 0.67, "learning_rate": 5.1826871270828745e-06, "loss": 1.9122, "step": 27190 }, { "epoch": 0.67, "learning_rate": 5.1819875497097435e-06, "loss": 2.0614, "step": 27191 }, { "epoch": 0.67, "learning_rate": 5.1812880030433745e-06, "loss": 1.9926, "step": 27192 }, { "epoch": 0.67, "learning_rate": 5.1805884870882315e-06, "loss": 2.1138, "step": 27193 }, { "epoch": 0.67, "learning_rate": 5.1798890018487655e-06, "loss": 1.962, "step": 27194 }, { "epoch": 0.67, "learning_rate": 5.179189547329437e-06, "loss": 2.0757, "step": 27195 }, { "epoch": 0.67, "learning_rate": 5.178490123534709e-06, "loss": 1.8944, "step": 27196 }, { "epoch": 0.67, "learning_rate": 5.177790730469031e-06, "loss": 1.7614, "step": 27197 }, { "epoch": 0.67, "learning_rate": 5.1770913681368636e-06, "loss": 2.1064, "step": 27198 }, { "epoch": 0.67, "learning_rate": 5.176392036542669e-06, "loss": 1.8717, "step": 27199 }, { "epoch": 0.67, "learning_rate": 5.175692735690896e-06, "loss": 2.0187, "step": 27200 }, { "epoch": 0.67, "learning_rate": 5.174993465586009e-06, "loss": 1.8624, "step": 27201 }, { "epoch": 0.67, "learning_rate": 5.174294226232457e-06, "loss": 2.094, "step": 27202 }, { "epoch": 0.67, "learning_rate": 5.1735950176347e-06, "loss": 1.8982, "step": 27203 }, { "epoch": 0.67, "learning_rate": 5.172895839797201e-06, "loss": 2.0475, "step": 27204 }, { "epoch": 0.67, "learning_rate": 5.172196692724405e-06, "loss": 1.8701, "step": 27205 }, { "epoch": 0.67, "learning_rate": 5.171497576420772e-06, "loss": 1.8443, "step": 27206 }, { "epoch": 0.67, "learning_rate": 5.170798490890763e-06, "loss": 2.0531, "step": 27207 }, { "epoch": 0.67, "learning_rate": 5.170099436138824e-06, "loss": 2.2426, "step": 27208 }, { "epoch": 0.67, "learning_rate": 5.169400412169416e-06, "loss": 2.034, "step": 27209 }, { "epoch": 0.67, "learning_rate": 5.168701418986994e-06, "loss": 1.8582, "step": 27210 }, { "epoch": 0.67, "learning_rate": 5.168002456596016e-06, "loss": 1.9435, "step": 27211 }, { "epoch": 0.67, "learning_rate": 5.167303525000929e-06, "loss": 1.8996, "step": 27212 }, { "epoch": 0.67, "learning_rate": 5.166604624206192e-06, "loss": 1.8648, "step": 27213 }, { "epoch": 0.67, "learning_rate": 5.165905754216263e-06, "loss": 1.8774, "step": 27214 }, { "epoch": 0.67, "learning_rate": 5.165206915035586e-06, "loss": 1.9572, "step": 27215 }, { "epoch": 0.67, "learning_rate": 5.164508106668623e-06, "loss": 2.1493, "step": 27216 }, { "epoch": 0.67, "learning_rate": 5.163809329119829e-06, "loss": 1.8317, "step": 27217 }, { "epoch": 0.67, "learning_rate": 5.16311058239365e-06, "loss": 2.0665, "step": 27218 }, { "epoch": 0.67, "learning_rate": 5.162411866494547e-06, "loss": 2.0798, "step": 27219 }, { "epoch": 0.67, "learning_rate": 5.1617131814269685e-06, "loss": 1.9157, "step": 27220 }, { "epoch": 0.67, "learning_rate": 5.161014527195366e-06, "loss": 1.9888, "step": 27221 }, { "epoch": 0.67, "learning_rate": 5.1603159038041995e-06, "loss": 2.1492, "step": 27222 }, { "epoch": 0.67, "learning_rate": 5.1596173112579126e-06, "loss": 2.334, "step": 27223 }, { "epoch": 0.67, "learning_rate": 5.158918749560966e-06, "loss": 2.1011, "step": 27224 }, { "epoch": 0.67, "learning_rate": 5.158220218717804e-06, "loss": 2.0384, "step": 27225 }, { "epoch": 0.67, "learning_rate": 5.157521718732884e-06, "loss": 2.0797, "step": 27226 }, { "epoch": 0.67, "learning_rate": 5.156823249610654e-06, "loss": 2.0351, "step": 27227 }, { "epoch": 0.67, "learning_rate": 5.156124811355574e-06, "loss": 1.913, "step": 27228 }, { "epoch": 0.67, "learning_rate": 5.155426403972084e-06, "loss": 1.8744, "step": 27229 }, { "epoch": 0.67, "learning_rate": 5.154728027464639e-06, "loss": 1.9414, "step": 27230 }, { "epoch": 0.67, "learning_rate": 5.154029681837693e-06, "loss": 1.8992, "step": 27231 }, { "epoch": 0.67, "learning_rate": 5.153331367095699e-06, "loss": 2.1118, "step": 27232 }, { "epoch": 0.67, "learning_rate": 5.1526330832431e-06, "loss": 2.0028, "step": 27233 }, { "epoch": 0.67, "learning_rate": 5.151934830284353e-06, "loss": 1.8026, "step": 27234 }, { "epoch": 0.67, "learning_rate": 5.151236608223902e-06, "loss": 2.0492, "step": 27235 }, { "epoch": 0.67, "learning_rate": 5.1505384170662e-06, "loss": 2.0316, "step": 27236 }, { "epoch": 0.67, "learning_rate": 5.149840256815699e-06, "loss": 1.9191, "step": 27237 }, { "epoch": 0.67, "learning_rate": 5.149142127476845e-06, "loss": 2.0819, "step": 27238 }, { "epoch": 0.67, "learning_rate": 5.148444029054094e-06, "loss": 2.0693, "step": 27239 }, { "epoch": 0.67, "learning_rate": 5.147745961551884e-06, "loss": 1.9287, "step": 27240 }, { "epoch": 0.67, "learning_rate": 5.1470479249746704e-06, "loss": 1.9561, "step": 27241 }, { "epoch": 0.67, "learning_rate": 5.146349919326907e-06, "loss": 2.1078, "step": 27242 }, { "epoch": 0.67, "learning_rate": 5.145651944613034e-06, "loss": 1.9305, "step": 27243 }, { "epoch": 0.67, "learning_rate": 5.144954000837502e-06, "loss": 2.0941, "step": 27244 }, { "epoch": 0.67, "learning_rate": 5.144256088004761e-06, "loss": 1.9389, "step": 27245 }, { "epoch": 0.67, "learning_rate": 5.143558206119261e-06, "loss": 2.0839, "step": 27246 }, { "epoch": 0.67, "learning_rate": 5.142860355185445e-06, "loss": 2.0099, "step": 27247 }, { "epoch": 0.67, "learning_rate": 5.142162535207762e-06, "loss": 1.9233, "step": 27248 }, { "epoch": 0.67, "learning_rate": 5.141464746190665e-06, "loss": 1.9511, "step": 27249 }, { "epoch": 0.67, "learning_rate": 5.140766988138592e-06, "loss": 1.9485, "step": 27250 }, { "epoch": 0.67, "learning_rate": 5.1400692610559934e-06, "loss": 1.8812, "step": 27251 }, { "epoch": 0.67, "learning_rate": 5.139371564947322e-06, "loss": 1.8636, "step": 27252 }, { "epoch": 0.67, "learning_rate": 5.138673899817017e-06, "loss": 2.0306, "step": 27253 }, { "epoch": 0.67, "learning_rate": 5.13797626566953e-06, "loss": 2.0934, "step": 27254 }, { "epoch": 0.67, "learning_rate": 5.137278662509302e-06, "loss": 2.1434, "step": 27255 }, { "epoch": 0.67, "learning_rate": 5.13658109034078e-06, "loss": 2.0183, "step": 27256 }, { "epoch": 0.67, "learning_rate": 5.135883549168417e-06, "loss": 1.9145, "step": 27257 }, { "epoch": 0.67, "learning_rate": 5.135186038996649e-06, "loss": 1.8161, "step": 27258 }, { "epoch": 0.67, "learning_rate": 5.134488559829929e-06, "loss": 2.0007, "step": 27259 }, { "epoch": 0.67, "learning_rate": 5.133791111672696e-06, "loss": 1.8734, "step": 27260 }, { "epoch": 0.67, "learning_rate": 5.133093694529398e-06, "loss": 1.9005, "step": 27261 }, { "epoch": 0.67, "learning_rate": 5.132396308404479e-06, "loss": 1.8281, "step": 27262 }, { "epoch": 0.67, "learning_rate": 5.131698953302387e-06, "loss": 1.9373, "step": 27263 }, { "epoch": 0.67, "learning_rate": 5.131001629227565e-06, "loss": 2.0206, "step": 27264 }, { "epoch": 0.67, "learning_rate": 5.130304336184455e-06, "loss": 2.0466, "step": 27265 }, { "epoch": 0.67, "learning_rate": 5.129607074177501e-06, "loss": 1.7758, "step": 27266 }, { "epoch": 0.67, "learning_rate": 5.128909843211151e-06, "loss": 1.843, "step": 27267 }, { "epoch": 0.67, "learning_rate": 5.128212643289845e-06, "loss": 1.9407, "step": 27268 }, { "epoch": 0.67, "learning_rate": 5.127515474418026e-06, "loss": 1.9409, "step": 27269 }, { "epoch": 0.67, "learning_rate": 5.126818336600143e-06, "loss": 1.9112, "step": 27270 }, { "epoch": 0.67, "learning_rate": 5.126121229840631e-06, "loss": 1.986, "step": 27271 }, { "epoch": 0.67, "learning_rate": 5.125424154143941e-06, "loss": 1.7699, "step": 27272 }, { "epoch": 0.67, "learning_rate": 5.124727109514507e-06, "loss": 2.1711, "step": 27273 }, { "epoch": 0.67, "learning_rate": 5.124030095956775e-06, "loss": 1.9918, "step": 27274 }, { "epoch": 0.67, "learning_rate": 5.123333113475193e-06, "loss": 1.7499, "step": 27275 }, { "epoch": 0.67, "learning_rate": 5.1226361620741946e-06, "loss": 1.871, "step": 27276 }, { "epoch": 0.67, "learning_rate": 5.1219392417582296e-06, "loss": 1.9231, "step": 27277 }, { "epoch": 0.67, "learning_rate": 5.121242352531732e-06, "loss": 1.7163, "step": 27278 }, { "epoch": 0.67, "learning_rate": 5.120545494399145e-06, "loss": 2.1873, "step": 27279 }, { "epoch": 0.67, "learning_rate": 5.119848667364913e-06, "loss": 2.0216, "step": 27280 }, { "epoch": 0.67, "learning_rate": 5.119151871433476e-06, "loss": 1.9634, "step": 27281 }, { "epoch": 0.67, "learning_rate": 5.118455106609279e-06, "loss": 1.7999, "step": 27282 }, { "epoch": 0.67, "learning_rate": 5.117758372896753e-06, "loss": 1.9688, "step": 27283 }, { "epoch": 0.67, "learning_rate": 5.117061670300345e-06, "loss": 2.0391, "step": 27284 }, { "epoch": 0.67, "learning_rate": 5.116364998824499e-06, "loss": 1.983, "step": 27285 }, { "epoch": 0.67, "learning_rate": 5.1156683584736445e-06, "loss": 1.8435, "step": 27286 }, { "epoch": 0.67, "learning_rate": 5.114971749252233e-06, "loss": 1.9536, "step": 27287 }, { "epoch": 0.67, "learning_rate": 5.114275171164693e-06, "loss": 1.9221, "step": 27288 }, { "epoch": 0.67, "learning_rate": 5.113578624215472e-06, "loss": 2.0293, "step": 27289 }, { "epoch": 0.67, "learning_rate": 5.112882108409007e-06, "loss": 2.1444, "step": 27290 }, { "epoch": 0.67, "learning_rate": 5.112185623749737e-06, "loss": 2.0644, "step": 27291 }, { "epoch": 0.67, "learning_rate": 5.111489170242104e-06, "loss": 1.9471, "step": 27292 }, { "epoch": 0.67, "learning_rate": 5.110792747890537e-06, "loss": 2.0617, "step": 27293 }, { "epoch": 0.67, "learning_rate": 5.1100963566994835e-06, "loss": 1.9859, "step": 27294 }, { "epoch": 0.67, "learning_rate": 5.109399996673383e-06, "loss": 1.9002, "step": 27295 }, { "epoch": 0.67, "learning_rate": 5.108703667816666e-06, "loss": 1.7581, "step": 27296 }, { "epoch": 0.67, "learning_rate": 5.108007370133774e-06, "loss": 1.8196, "step": 27297 }, { "epoch": 0.67, "learning_rate": 5.107311103629147e-06, "loss": 1.8968, "step": 27298 }, { "epoch": 0.67, "learning_rate": 5.106614868307219e-06, "loss": 1.9496, "step": 27299 }, { "epoch": 0.67, "learning_rate": 5.1059186641724335e-06, "loss": 1.7995, "step": 27300 }, { "epoch": 0.67, "learning_rate": 5.10522249122922e-06, "loss": 1.8982, "step": 27301 }, { "epoch": 0.67, "learning_rate": 5.104526349482023e-06, "loss": 1.9605, "step": 27302 }, { "epoch": 0.67, "learning_rate": 5.10383023893527e-06, "loss": 1.9681, "step": 27303 }, { "epoch": 0.67, "learning_rate": 5.103134159593404e-06, "loss": 1.8926, "step": 27304 }, { "epoch": 0.67, "learning_rate": 5.102438111460861e-06, "loss": 1.8821, "step": 27305 }, { "epoch": 0.67, "learning_rate": 5.101742094542075e-06, "loss": 1.9079, "step": 27306 }, { "epoch": 0.67, "learning_rate": 5.101046108841486e-06, "loss": 2.0334, "step": 27307 }, { "epoch": 0.67, "learning_rate": 5.100350154363522e-06, "loss": 2.1555, "step": 27308 }, { "epoch": 0.67, "learning_rate": 5.0996542311126225e-06, "loss": 2.0963, "step": 27309 }, { "epoch": 0.67, "learning_rate": 5.09895833909323e-06, "loss": 1.9457, "step": 27310 }, { "epoch": 0.67, "learning_rate": 5.098262478309766e-06, "loss": 1.9304, "step": 27311 }, { "epoch": 0.67, "learning_rate": 5.097566648766679e-06, "loss": 1.9631, "step": 27312 }, { "epoch": 0.67, "learning_rate": 5.0968708504683925e-06, "loss": 1.9263, "step": 27313 }, { "epoch": 0.67, "learning_rate": 5.0961750834193465e-06, "loss": 1.909, "step": 27314 }, { "epoch": 0.67, "learning_rate": 5.095479347623975e-06, "loss": 1.944, "step": 27315 }, { "epoch": 0.67, "learning_rate": 5.094783643086712e-06, "loss": 1.9968, "step": 27316 }, { "epoch": 0.67, "learning_rate": 5.094087969811994e-06, "loss": 2.0127, "step": 27317 }, { "epoch": 0.67, "learning_rate": 5.093392327804249e-06, "loss": 1.9804, "step": 27318 }, { "epoch": 0.67, "learning_rate": 5.0926967170679145e-06, "loss": 1.9101, "step": 27319 }, { "epoch": 0.67, "learning_rate": 5.092001137607427e-06, "loss": 1.9849, "step": 27320 }, { "epoch": 0.67, "learning_rate": 5.09130558942721e-06, "loss": 1.9581, "step": 27321 }, { "epoch": 0.67, "learning_rate": 5.090610072531707e-06, "loss": 2.1111, "step": 27322 }, { "epoch": 0.67, "learning_rate": 5.0899145869253415e-06, "loss": 2.0913, "step": 27323 }, { "epoch": 0.67, "learning_rate": 5.089219132612553e-06, "loss": 2.1064, "step": 27324 }, { "epoch": 0.67, "learning_rate": 5.088523709597772e-06, "loss": 1.8255, "step": 27325 }, { "epoch": 0.67, "learning_rate": 5.087828317885428e-06, "loss": 2.1263, "step": 27326 }, { "epoch": 0.67, "learning_rate": 5.087132957479954e-06, "loss": 1.8904, "step": 27327 }, { "epoch": 0.67, "learning_rate": 5.086437628385788e-06, "loss": 1.9582, "step": 27328 }, { "epoch": 0.67, "learning_rate": 5.085742330607352e-06, "loss": 1.9525, "step": 27329 }, { "epoch": 0.67, "learning_rate": 5.085047064149085e-06, "loss": 2.0461, "step": 27330 }, { "epoch": 0.67, "learning_rate": 5.0843518290154105e-06, "loss": 1.9525, "step": 27331 }, { "epoch": 0.67, "learning_rate": 5.0836566252107645e-06, "loss": 1.9908, "step": 27332 }, { "epoch": 0.67, "learning_rate": 5.0829614527395764e-06, "loss": 1.6989, "step": 27333 }, { "epoch": 0.67, "learning_rate": 5.082266311606279e-06, "loss": 1.8709, "step": 27334 }, { "epoch": 0.67, "learning_rate": 5.081571201815303e-06, "loss": 1.9846, "step": 27335 }, { "epoch": 0.67, "learning_rate": 5.080876123371073e-06, "loss": 1.8903, "step": 27336 }, { "epoch": 0.67, "learning_rate": 5.080181076278024e-06, "loss": 2.1562, "step": 27337 }, { "epoch": 0.67, "learning_rate": 5.079486060540586e-06, "loss": 2.124, "step": 27338 }, { "epoch": 0.67, "learning_rate": 5.078791076163185e-06, "loss": 1.9909, "step": 27339 }, { "epoch": 0.67, "learning_rate": 5.078096123150255e-06, "loss": 1.9162, "step": 27340 }, { "epoch": 0.67, "learning_rate": 5.077401201506218e-06, "loss": 1.7192, "step": 27341 }, { "epoch": 0.67, "learning_rate": 5.076706311235509e-06, "loss": 2.0165, "step": 27342 }, { "epoch": 0.67, "learning_rate": 5.076011452342558e-06, "loss": 1.9169, "step": 27343 }, { "epoch": 0.67, "learning_rate": 5.075316624831788e-06, "loss": 1.837, "step": 27344 }, { "epoch": 0.67, "learning_rate": 5.074621828707632e-06, "loss": 1.7319, "step": 27345 }, { "epoch": 0.67, "learning_rate": 5.073927063974513e-06, "loss": 1.944, "step": 27346 }, { "epoch": 0.67, "learning_rate": 5.073232330636864e-06, "loss": 1.8656, "step": 27347 }, { "epoch": 0.67, "learning_rate": 5.072537628699112e-06, "loss": 1.9383, "step": 27348 }, { "epoch": 0.67, "learning_rate": 5.07184295816568e-06, "loss": 1.9014, "step": 27349 }, { "epoch": 0.67, "learning_rate": 5.0711483190410016e-06, "loss": 1.9683, "step": 27350 }, { "epoch": 0.67, "learning_rate": 5.0704537113294995e-06, "loss": 1.9508, "step": 27351 }, { "epoch": 0.67, "learning_rate": 5.0697591350356015e-06, "loss": 1.8716, "step": 27352 }, { "epoch": 0.67, "learning_rate": 5.069064590163739e-06, "loss": 1.9727, "step": 27353 }, { "epoch": 0.67, "learning_rate": 5.068370076718333e-06, "loss": 1.9593, "step": 27354 }, { "epoch": 0.67, "learning_rate": 5.067675594703814e-06, "loss": 1.9307, "step": 27355 }, { "epoch": 0.67, "learning_rate": 5.066981144124603e-06, "loss": 1.9164, "step": 27356 }, { "epoch": 0.67, "learning_rate": 5.066286724985128e-06, "loss": 1.9859, "step": 27357 }, { "epoch": 0.67, "learning_rate": 5.06559233728982e-06, "loss": 2.0858, "step": 27358 }, { "epoch": 0.67, "learning_rate": 5.064897981043095e-06, "loss": 2.0067, "step": 27359 }, { "epoch": 0.67, "learning_rate": 5.064203656249389e-06, "loss": 2.1816, "step": 27360 }, { "epoch": 0.67, "learning_rate": 5.063509362913117e-06, "loss": 2.0893, "step": 27361 }, { "epoch": 0.67, "learning_rate": 5.062815101038707e-06, "loss": 1.9377, "step": 27362 }, { "epoch": 0.67, "learning_rate": 5.062120870630591e-06, "loss": 1.9317, "step": 27363 }, { "epoch": 0.67, "learning_rate": 5.061426671693184e-06, "loss": 2.1529, "step": 27364 }, { "epoch": 0.67, "learning_rate": 5.060732504230919e-06, "loss": 1.9184, "step": 27365 }, { "epoch": 0.67, "learning_rate": 5.06003836824821e-06, "loss": 1.8923, "step": 27366 }, { "epoch": 0.67, "learning_rate": 5.059344263749486e-06, "loss": 2.205, "step": 27367 }, { "epoch": 0.67, "learning_rate": 5.058650190739173e-06, "loss": 1.9555, "step": 27368 }, { "epoch": 0.67, "learning_rate": 5.057956149221691e-06, "loss": 1.8651, "step": 27369 }, { "epoch": 0.67, "learning_rate": 5.05726213920147e-06, "loss": 1.8892, "step": 27370 }, { "epoch": 0.67, "learning_rate": 5.056568160682924e-06, "loss": 1.8462, "step": 27371 }, { "epoch": 0.67, "learning_rate": 5.05587421367048e-06, "loss": 2.1632, "step": 27372 }, { "epoch": 0.67, "learning_rate": 5.055180298168564e-06, "loss": 2.0884, "step": 27373 }, { "epoch": 0.67, "learning_rate": 5.054486414181592e-06, "loss": 2.007, "step": 27374 }, { "epoch": 0.67, "learning_rate": 5.053792561713995e-06, "loss": 2.1147, "step": 27375 }, { "epoch": 0.67, "learning_rate": 5.053098740770183e-06, "loss": 1.9393, "step": 27376 }, { "epoch": 0.67, "learning_rate": 5.052404951354588e-06, "loss": 1.789, "step": 27377 }, { "epoch": 0.67, "learning_rate": 5.051711193471631e-06, "loss": 1.8328, "step": 27378 }, { "epoch": 0.67, "learning_rate": 5.051017467125727e-06, "loss": 1.9592, "step": 27379 }, { "epoch": 0.67, "learning_rate": 5.050323772321306e-06, "loss": 1.9523, "step": 27380 }, { "epoch": 0.67, "learning_rate": 5.049630109062782e-06, "loss": 2.0237, "step": 27381 }, { "epoch": 0.67, "learning_rate": 5.048936477354577e-06, "loss": 2.1712, "step": 27382 }, { "epoch": 0.67, "learning_rate": 5.048242877201117e-06, "loss": 1.9792, "step": 27383 }, { "epoch": 0.67, "learning_rate": 5.047549308606816e-06, "loss": 1.7846, "step": 27384 }, { "epoch": 0.68, "learning_rate": 5.046855771576097e-06, "loss": 1.951, "step": 27385 }, { "epoch": 0.68, "learning_rate": 5.046162266113381e-06, "loss": 2.0111, "step": 27386 }, { "epoch": 0.68, "learning_rate": 5.045468792223085e-06, "loss": 1.9004, "step": 27387 }, { "epoch": 0.68, "learning_rate": 5.044775349909636e-06, "loss": 1.8536, "step": 27388 }, { "epoch": 0.68, "learning_rate": 5.044081939177445e-06, "loss": 1.9272, "step": 27389 }, { "epoch": 0.68, "learning_rate": 5.043388560030935e-06, "loss": 2.079, "step": 27390 }, { "epoch": 0.68, "learning_rate": 5.042695212474529e-06, "loss": 1.9048, "step": 27391 }, { "epoch": 0.68, "learning_rate": 5.042001896512638e-06, "loss": 2.0875, "step": 27392 }, { "epoch": 0.68, "learning_rate": 5.041308612149689e-06, "loss": 2.0284, "step": 27393 }, { "epoch": 0.68, "learning_rate": 5.040615359390091e-06, "loss": 1.8198, "step": 27394 }, { "epoch": 0.68, "learning_rate": 5.039922138238269e-06, "loss": 1.9322, "step": 27395 }, { "epoch": 0.68, "learning_rate": 5.0392289486986425e-06, "loss": 2.0242, "step": 27396 }, { "epoch": 0.68, "learning_rate": 5.038535790775623e-06, "loss": 1.9995, "step": 27397 }, { "epoch": 0.68, "learning_rate": 5.0378426644736355e-06, "loss": 1.8476, "step": 27398 }, { "epoch": 0.68, "learning_rate": 5.03714956979709e-06, "loss": 1.9066, "step": 27399 }, { "epoch": 0.68, "learning_rate": 5.036456506750409e-06, "loss": 2.1598, "step": 27400 }, { "epoch": 0.68, "learning_rate": 5.03576347533801e-06, "loss": 2.0241, "step": 27401 }, { "epoch": 0.68, "learning_rate": 5.035070475564305e-06, "loss": 1.9205, "step": 27402 }, { "epoch": 0.68, "learning_rate": 5.034377507433714e-06, "loss": 2.0743, "step": 27403 }, { "epoch": 0.68, "learning_rate": 5.033684570950654e-06, "loss": 2.2359, "step": 27404 }, { "epoch": 0.68, "learning_rate": 5.03299166611954e-06, "loss": 1.9942, "step": 27405 }, { "epoch": 0.68, "learning_rate": 5.032298792944793e-06, "loss": 1.9113, "step": 27406 }, { "epoch": 0.68, "learning_rate": 5.03160595143082e-06, "loss": 1.9428, "step": 27407 }, { "epoch": 0.68, "learning_rate": 5.030913141582047e-06, "loss": 1.8867, "step": 27408 }, { "epoch": 0.68, "learning_rate": 5.030220363402878e-06, "loss": 2.0009, "step": 27409 }, { "epoch": 0.68, "learning_rate": 5.029527616897736e-06, "loss": 1.9139, "step": 27410 }, { "epoch": 0.68, "learning_rate": 5.028834902071038e-06, "loss": 1.9042, "step": 27411 }, { "epoch": 0.68, "learning_rate": 5.02814221892719e-06, "loss": 1.9575, "step": 27412 }, { "epoch": 0.68, "learning_rate": 5.027449567470617e-06, "loss": 2.0588, "step": 27413 }, { "epoch": 0.68, "learning_rate": 5.026756947705725e-06, "loss": 1.9365, "step": 27414 }, { "epoch": 0.68, "learning_rate": 5.026064359636931e-06, "loss": 1.9528, "step": 27415 }, { "epoch": 0.68, "learning_rate": 5.025371803268653e-06, "loss": 2.1367, "step": 27416 }, { "epoch": 0.68, "learning_rate": 5.024679278605298e-06, "loss": 2.0628, "step": 27417 }, { "epoch": 0.68, "learning_rate": 5.023986785651287e-06, "loss": 1.9565, "step": 27418 }, { "epoch": 0.68, "learning_rate": 5.023294324411028e-06, "loss": 1.9382, "step": 27419 }, { "epoch": 0.68, "learning_rate": 5.022601894888934e-06, "loss": 1.9771, "step": 27420 }, { "epoch": 0.68, "learning_rate": 5.0219094970894215e-06, "loss": 1.7413, "step": 27421 }, { "epoch": 0.68, "learning_rate": 5.0212171310169e-06, "loss": 1.9916, "step": 27422 }, { "epoch": 0.68, "learning_rate": 5.020524796675789e-06, "loss": 1.8925, "step": 27423 }, { "epoch": 0.68, "learning_rate": 5.0198324940704944e-06, "loss": 2.2095, "step": 27424 }, { "epoch": 0.68, "learning_rate": 5.01914022320543e-06, "loss": 1.931, "step": 27425 }, { "epoch": 0.68, "learning_rate": 5.01844798408501e-06, "loss": 1.7717, "step": 27426 }, { "epoch": 0.68, "learning_rate": 5.017755776713642e-06, "loss": 2.0039, "step": 27427 }, { "epoch": 0.68, "learning_rate": 5.017063601095743e-06, "loss": 1.9375, "step": 27428 }, { "epoch": 0.68, "learning_rate": 5.016371457235718e-06, "loss": 1.8679, "step": 27429 }, { "epoch": 0.68, "learning_rate": 5.0156793451379826e-06, "loss": 1.9469, "step": 27430 }, { "epoch": 0.68, "learning_rate": 5.0149872648069495e-06, "loss": 2.0973, "step": 27431 }, { "epoch": 0.68, "learning_rate": 5.0142952162470245e-06, "loss": 1.867, "step": 27432 }, { "epoch": 0.68, "learning_rate": 5.013603199462624e-06, "loss": 2.0925, "step": 27433 }, { "epoch": 0.68, "learning_rate": 5.012911214458152e-06, "loss": 1.9198, "step": 27434 }, { "epoch": 0.68, "learning_rate": 5.0122192612380225e-06, "loss": 2.1479, "step": 27435 }, { "epoch": 0.68, "learning_rate": 5.011527339806648e-06, "loss": 1.9916, "step": 27436 }, { "epoch": 0.68, "learning_rate": 5.010835450168432e-06, "loss": 2.0262, "step": 27437 }, { "epoch": 0.68, "learning_rate": 5.010143592327788e-06, "loss": 1.9827, "step": 27438 }, { "epoch": 0.68, "learning_rate": 5.009451766289124e-06, "loss": 2.0868, "step": 27439 }, { "epoch": 0.68, "learning_rate": 5.0087599720568514e-06, "loss": 2.1411, "step": 27440 }, { "epoch": 0.68, "learning_rate": 5.008068209635382e-06, "loss": 2.0334, "step": 27441 }, { "epoch": 0.68, "learning_rate": 5.0073764790291155e-06, "loss": 1.9016, "step": 27442 }, { "epoch": 0.68, "learning_rate": 5.006684780242467e-06, "loss": 2.0799, "step": 27443 }, { "epoch": 0.68, "learning_rate": 5.005993113279847e-06, "loss": 1.9168, "step": 27444 }, { "epoch": 0.68, "learning_rate": 5.0053014781456564e-06, "loss": 1.9893, "step": 27445 }, { "epoch": 0.68, "learning_rate": 5.004609874844311e-06, "loss": 2.0731, "step": 27446 }, { "epoch": 0.68, "learning_rate": 5.003918303380211e-06, "loss": 1.7554, "step": 27447 }, { "epoch": 0.68, "learning_rate": 5.003226763757769e-06, "loss": 1.8894, "step": 27448 }, { "epoch": 0.68, "learning_rate": 5.002535255981393e-06, "loss": 1.9516, "step": 27449 }, { "epoch": 0.68, "learning_rate": 5.001843780055487e-06, "loss": 1.9569, "step": 27450 }, { "epoch": 0.68, "learning_rate": 5.0011523359844616e-06, "loss": 1.98, "step": 27451 }, { "epoch": 0.68, "learning_rate": 5.000460923772719e-06, "loss": 1.9922, "step": 27452 }, { "epoch": 0.68, "learning_rate": 4.999769543424666e-06, "loss": 1.8258, "step": 27453 }, { "epoch": 0.68, "learning_rate": 4.999078194944717e-06, "loss": 2.011, "step": 27454 }, { "epoch": 0.68, "learning_rate": 4.998386878337267e-06, "loss": 2.0357, "step": 27455 }, { "epoch": 0.68, "learning_rate": 4.997695593606728e-06, "loss": 1.8246, "step": 27456 }, { "epoch": 0.68, "learning_rate": 4.9970043407575065e-06, "loss": 2.0152, "step": 27457 }, { "epoch": 0.68, "learning_rate": 4.996313119794005e-06, "loss": 1.7789, "step": 27458 }, { "epoch": 0.68, "learning_rate": 4.995621930720635e-06, "loss": 2.054, "step": 27459 }, { "epoch": 0.68, "learning_rate": 4.994930773541794e-06, "loss": 1.9938, "step": 27460 }, { "epoch": 0.68, "learning_rate": 4.9942396482618935e-06, "loss": 1.7999, "step": 27461 }, { "epoch": 0.68, "learning_rate": 4.993548554885333e-06, "loss": 2.0038, "step": 27462 }, { "epoch": 0.68, "learning_rate": 4.992857493416517e-06, "loss": 1.8837, "step": 27463 }, { "epoch": 0.68, "learning_rate": 4.9921664638598564e-06, "loss": 1.9985, "step": 27464 }, { "epoch": 0.68, "learning_rate": 4.991475466219748e-06, "loss": 1.9741, "step": 27465 }, { "epoch": 0.68, "learning_rate": 4.9907845005006005e-06, "loss": 2.0538, "step": 27466 }, { "epoch": 0.68, "learning_rate": 4.990093566706814e-06, "loss": 1.8184, "step": 27467 }, { "epoch": 0.68, "learning_rate": 4.989402664842793e-06, "loss": 1.9168, "step": 27468 }, { "epoch": 0.68, "learning_rate": 4.988711794912945e-06, "loss": 2.0883, "step": 27469 }, { "epoch": 0.68, "learning_rate": 4.988020956921667e-06, "loss": 1.953, "step": 27470 }, { "epoch": 0.68, "learning_rate": 4.987330150873369e-06, "loss": 2.0066, "step": 27471 }, { "epoch": 0.68, "learning_rate": 4.9866393767724444e-06, "loss": 1.8779, "step": 27472 }, { "epoch": 0.68, "learning_rate": 4.985948634623301e-06, "loss": 2.0134, "step": 27473 }, { "epoch": 0.68, "learning_rate": 4.9852579244303425e-06, "loss": 1.9868, "step": 27474 }, { "epoch": 0.68, "learning_rate": 4.984567246197967e-06, "loss": 1.9402, "step": 27475 }, { "epoch": 0.68, "learning_rate": 4.983876599930585e-06, "loss": 1.8584, "step": 27476 }, { "epoch": 0.68, "learning_rate": 4.983185985632587e-06, "loss": 1.8733, "step": 27477 }, { "epoch": 0.68, "learning_rate": 4.982495403308381e-06, "loss": 1.9705, "step": 27478 }, { "epoch": 0.68, "learning_rate": 4.98180485296237e-06, "loss": 2.1052, "step": 27479 }, { "epoch": 0.68, "learning_rate": 4.981114334598947e-06, "loss": 2.0581, "step": 27480 }, { "epoch": 0.68, "learning_rate": 4.980423848222524e-06, "loss": 2.0266, "step": 27481 }, { "epoch": 0.68, "learning_rate": 4.9797333938374905e-06, "loss": 2.0818, "step": 27482 }, { "epoch": 0.68, "learning_rate": 4.979042971448253e-06, "loss": 1.8643, "step": 27483 }, { "epoch": 0.68, "learning_rate": 4.978352581059214e-06, "loss": 1.9428, "step": 27484 }, { "epoch": 0.68, "learning_rate": 4.977662222674766e-06, "loss": 2.0906, "step": 27485 }, { "epoch": 0.68, "learning_rate": 4.976971896299318e-06, "loss": 1.8567, "step": 27486 }, { "epoch": 0.68, "learning_rate": 4.976281601937262e-06, "loss": 2.0435, "step": 27487 }, { "epoch": 0.68, "learning_rate": 4.975591339593e-06, "loss": 1.8395, "step": 27488 }, { "epoch": 0.68, "learning_rate": 4.9749011092709355e-06, "loss": 1.7791, "step": 27489 }, { "epoch": 0.68, "learning_rate": 4.97421091097546e-06, "loss": 2.0654, "step": 27490 }, { "epoch": 0.68, "learning_rate": 4.9735207447109755e-06, "loss": 2.0248, "step": 27491 }, { "epoch": 0.68, "learning_rate": 4.972830610481882e-06, "loss": 2.0454, "step": 27492 }, { "epoch": 0.68, "learning_rate": 4.972140508292578e-06, "loss": 1.7985, "step": 27493 }, { "epoch": 0.68, "learning_rate": 4.971450438147464e-06, "loss": 2.0818, "step": 27494 }, { "epoch": 0.68, "learning_rate": 4.9707604000509315e-06, "loss": 2.078, "step": 27495 }, { "epoch": 0.68, "learning_rate": 4.970070394007386e-06, "loss": 2.0235, "step": 27496 }, { "epoch": 0.68, "learning_rate": 4.969380420021217e-06, "loss": 1.8708, "step": 27497 }, { "epoch": 0.68, "learning_rate": 4.968690478096826e-06, "loss": 1.8517, "step": 27498 }, { "epoch": 0.68, "learning_rate": 4.968000568238614e-06, "loss": 1.9467, "step": 27499 }, { "epoch": 0.68, "learning_rate": 4.967310690450972e-06, "loss": 1.9312, "step": 27500 }, { "epoch": 0.68, "learning_rate": 4.966620844738302e-06, "loss": 2.017, "step": 27501 }, { "epoch": 0.68, "learning_rate": 4.965931031104992e-06, "loss": 1.9514, "step": 27502 }, { "epoch": 0.68, "learning_rate": 4.965241249555446e-06, "loss": 1.9585, "step": 27503 }, { "epoch": 0.68, "learning_rate": 4.964551500094061e-06, "loss": 1.8943, "step": 27504 }, { "epoch": 0.68, "learning_rate": 4.963861782725228e-06, "loss": 1.884, "step": 27505 }, { "epoch": 0.68, "learning_rate": 4.963172097453344e-06, "loss": 1.9711, "step": 27506 }, { "epoch": 0.68, "learning_rate": 4.962482444282809e-06, "loss": 2.0428, "step": 27507 }, { "epoch": 0.68, "learning_rate": 4.961792823218012e-06, "loss": 2.0725, "step": 27508 }, { "epoch": 0.68, "learning_rate": 4.961103234263351e-06, "loss": 1.9863, "step": 27509 }, { "epoch": 0.68, "learning_rate": 4.960413677423221e-06, "loss": 1.9266, "step": 27510 }, { "epoch": 0.68, "learning_rate": 4.959724152702018e-06, "loss": 2.271, "step": 27511 }, { "epoch": 0.68, "learning_rate": 4.9590346601041396e-06, "loss": 2.0709, "step": 27512 }, { "epoch": 0.68, "learning_rate": 4.95834519963397e-06, "loss": 2.0498, "step": 27513 }, { "epoch": 0.68, "learning_rate": 4.957655771295915e-06, "loss": 1.8491, "step": 27514 }, { "epoch": 0.68, "learning_rate": 4.95696637509436e-06, "loss": 1.9563, "step": 27515 }, { "epoch": 0.68, "learning_rate": 4.9562770110337e-06, "loss": 1.8161, "step": 27516 }, { "epoch": 0.68, "learning_rate": 4.955587679118335e-06, "loss": 2.053, "step": 27517 }, { "epoch": 0.68, "learning_rate": 4.954898379352649e-06, "loss": 1.9695, "step": 27518 }, { "epoch": 0.68, "learning_rate": 4.954209111741045e-06, "loss": 1.8492, "step": 27519 }, { "epoch": 0.68, "learning_rate": 4.953519876287906e-06, "loss": 1.9824, "step": 27520 }, { "epoch": 0.68, "learning_rate": 4.952830672997629e-06, "loss": 1.9425, "step": 27521 }, { "epoch": 0.68, "learning_rate": 4.952141501874612e-06, "loss": 1.9274, "step": 27522 }, { "epoch": 0.68, "learning_rate": 4.951452362923238e-06, "loss": 2.0534, "step": 27523 }, { "epoch": 0.68, "learning_rate": 4.950763256147907e-06, "loss": 1.9406, "step": 27524 }, { "epoch": 0.68, "learning_rate": 4.950074181553003e-06, "loss": 1.9771, "step": 27525 }, { "epoch": 0.68, "learning_rate": 4.949385139142922e-06, "loss": 1.9473, "step": 27526 }, { "epoch": 0.68, "learning_rate": 4.948696128922057e-06, "loss": 2.0162, "step": 27527 }, { "epoch": 0.68, "learning_rate": 4.948007150894796e-06, "loss": 2.1006, "step": 27528 }, { "epoch": 0.68, "learning_rate": 4.9473182050655365e-06, "loss": 1.902, "step": 27529 }, { "epoch": 0.68, "learning_rate": 4.94662929143866e-06, "loss": 2.1153, "step": 27530 }, { "epoch": 0.68, "learning_rate": 4.945940410018563e-06, "loss": 2.0385, "step": 27531 }, { "epoch": 0.68, "learning_rate": 4.9452515608096385e-06, "loss": 1.9563, "step": 27532 }, { "epoch": 0.68, "learning_rate": 4.944562743816268e-06, "loss": 2.0396, "step": 27533 }, { "epoch": 0.68, "learning_rate": 4.943873959042852e-06, "loss": 1.9647, "step": 27534 }, { "epoch": 0.68, "learning_rate": 4.9431852064937705e-06, "loss": 1.8218, "step": 27535 }, { "epoch": 0.68, "learning_rate": 4.942496486173418e-06, "loss": 2.062, "step": 27536 }, { "epoch": 0.68, "learning_rate": 4.941807798086187e-06, "loss": 2.0147, "step": 27537 }, { "epoch": 0.68, "learning_rate": 4.941119142236459e-06, "loss": 2.049, "step": 27538 }, { "epoch": 0.68, "learning_rate": 4.9404305186286316e-06, "loss": 1.9734, "step": 27539 }, { "epoch": 0.68, "learning_rate": 4.939741927267087e-06, "loss": 1.9964, "step": 27540 }, { "epoch": 0.68, "learning_rate": 4.939053368156215e-06, "loss": 2.0864, "step": 27541 }, { "epoch": 0.68, "learning_rate": 4.9383648413004086e-06, "loss": 1.9698, "step": 27542 }, { "epoch": 0.68, "learning_rate": 4.937676346704049e-06, "loss": 1.9534, "step": 27543 }, { "epoch": 0.68, "learning_rate": 4.936987884371528e-06, "loss": 2.1707, "step": 27544 }, { "epoch": 0.68, "learning_rate": 4.936299454307233e-06, "loss": 1.9017, "step": 27545 }, { "epoch": 0.68, "learning_rate": 4.935611056515551e-06, "loss": 2.1308, "step": 27546 }, { "epoch": 0.68, "learning_rate": 4.934922691000875e-06, "loss": 1.7183, "step": 27547 }, { "epoch": 0.68, "learning_rate": 4.934234357767585e-06, "loss": 1.9817, "step": 27548 }, { "epoch": 0.68, "learning_rate": 4.933546056820072e-06, "loss": 1.9277, "step": 27549 }, { "epoch": 0.68, "learning_rate": 4.932857788162718e-06, "loss": 2.1984, "step": 27550 }, { "epoch": 0.68, "learning_rate": 4.932169551799913e-06, "loss": 1.9819, "step": 27551 }, { "epoch": 0.68, "learning_rate": 4.931481347736046e-06, "loss": 1.9817, "step": 27552 }, { "epoch": 0.68, "learning_rate": 4.930793175975497e-06, "loss": 1.9332, "step": 27553 }, { "epoch": 0.68, "learning_rate": 4.930105036522658e-06, "loss": 1.9879, "step": 27554 }, { "epoch": 0.68, "learning_rate": 4.929416929381908e-06, "loss": 1.8939, "step": 27555 }, { "epoch": 0.68, "learning_rate": 4.928728854557637e-06, "loss": 1.9587, "step": 27556 }, { "epoch": 0.68, "learning_rate": 4.928040812054234e-06, "loss": 1.8689, "step": 27557 }, { "epoch": 0.68, "learning_rate": 4.927352801876075e-06, "loss": 1.8279, "step": 27558 }, { "epoch": 0.68, "learning_rate": 4.926664824027555e-06, "loss": 1.9281, "step": 27559 }, { "epoch": 0.68, "learning_rate": 4.925976878513048e-06, "loss": 1.9799, "step": 27560 }, { "epoch": 0.68, "learning_rate": 4.925288965336945e-06, "loss": 1.9573, "step": 27561 }, { "epoch": 0.68, "learning_rate": 4.924601084503628e-06, "loss": 2.0247, "step": 27562 }, { "epoch": 0.68, "learning_rate": 4.923913236017483e-06, "loss": 1.8993, "step": 27563 }, { "epoch": 0.68, "learning_rate": 4.9232254198828935e-06, "loss": 2.1226, "step": 27564 }, { "epoch": 0.68, "learning_rate": 4.9225376361042465e-06, "loss": 1.9492, "step": 27565 }, { "epoch": 0.68, "learning_rate": 4.921849884685919e-06, "loss": 2.0081, "step": 27566 }, { "epoch": 0.68, "learning_rate": 4.921162165632299e-06, "loss": 2.0611, "step": 27567 }, { "epoch": 0.68, "learning_rate": 4.920474478947765e-06, "loss": 1.8639, "step": 27568 }, { "epoch": 0.68, "learning_rate": 4.919786824636702e-06, "loss": 1.9792, "step": 27569 }, { "epoch": 0.68, "learning_rate": 4.919099202703498e-06, "loss": 1.9308, "step": 27570 }, { "epoch": 0.68, "learning_rate": 4.918411613152526e-06, "loss": 2.1193, "step": 27571 }, { "epoch": 0.68, "learning_rate": 4.917724055988177e-06, "loss": 1.9053, "step": 27572 }, { "epoch": 0.68, "learning_rate": 4.917036531214824e-06, "loss": 1.9359, "step": 27573 }, { "epoch": 0.68, "learning_rate": 4.916349038836854e-06, "loss": 1.9009, "step": 27574 }, { "epoch": 0.68, "learning_rate": 4.915661578858653e-06, "loss": 2.0335, "step": 27575 }, { "epoch": 0.68, "learning_rate": 4.914974151284594e-06, "loss": 1.9224, "step": 27576 }, { "epoch": 0.68, "learning_rate": 4.914286756119064e-06, "loss": 2.1248, "step": 27577 }, { "epoch": 0.68, "learning_rate": 4.91359939336644e-06, "loss": 2.02, "step": 27578 }, { "epoch": 0.68, "learning_rate": 4.912912063031103e-06, "loss": 2.0415, "step": 27579 }, { "epoch": 0.68, "learning_rate": 4.912224765117437e-06, "loss": 1.9267, "step": 27580 }, { "epoch": 0.68, "learning_rate": 4.911537499629819e-06, "loss": 2.0863, "step": 27581 }, { "epoch": 0.68, "learning_rate": 4.9108502665726354e-06, "loss": 2.0208, "step": 27582 }, { "epoch": 0.68, "learning_rate": 4.9101630659502585e-06, "loss": 1.909, "step": 27583 }, { "epoch": 0.68, "learning_rate": 4.90947589776707e-06, "loss": 1.835, "step": 27584 }, { "epoch": 0.68, "learning_rate": 4.908788762027454e-06, "loss": 2.0472, "step": 27585 }, { "epoch": 0.68, "learning_rate": 4.908101658735784e-06, "loss": 1.9799, "step": 27586 }, { "epoch": 0.68, "learning_rate": 4.907414587896444e-06, "loss": 1.9959, "step": 27587 }, { "epoch": 0.68, "learning_rate": 4.9067275495138075e-06, "loss": 1.8431, "step": 27588 }, { "epoch": 0.68, "learning_rate": 4.906040543592257e-06, "loss": 1.8206, "step": 27589 }, { "epoch": 0.68, "learning_rate": 4.905353570136172e-06, "loss": 2.0007, "step": 27590 }, { "epoch": 0.68, "learning_rate": 4.904666629149927e-06, "loss": 2.1085, "step": 27591 }, { "epoch": 0.68, "learning_rate": 4.903979720637905e-06, "loss": 1.8648, "step": 27592 }, { "epoch": 0.68, "learning_rate": 4.903292844604477e-06, "loss": 1.9063, "step": 27593 }, { "epoch": 0.68, "learning_rate": 4.902606001054024e-06, "loss": 2.0785, "step": 27594 }, { "epoch": 0.68, "learning_rate": 4.901919189990929e-06, "loss": 2.0394, "step": 27595 }, { "epoch": 0.68, "learning_rate": 4.9012324114195605e-06, "loss": 1.9585, "step": 27596 }, { "epoch": 0.68, "learning_rate": 4.900545665344299e-06, "loss": 1.8933, "step": 27597 }, { "epoch": 0.68, "learning_rate": 4.899858951769523e-06, "loss": 2.0233, "step": 27598 }, { "epoch": 0.68, "learning_rate": 4.899172270699607e-06, "loss": 2.232, "step": 27599 }, { "epoch": 0.68, "learning_rate": 4.898485622138932e-06, "loss": 1.9146, "step": 27600 }, { "epoch": 0.68, "learning_rate": 4.897799006091868e-06, "loss": 2.0086, "step": 27601 }, { "epoch": 0.68, "learning_rate": 4.897112422562796e-06, "loss": 2.0169, "step": 27602 }, { "epoch": 0.68, "learning_rate": 4.896425871556088e-06, "loss": 1.7561, "step": 27603 }, { "epoch": 0.68, "learning_rate": 4.895739353076119e-06, "loss": 1.9375, "step": 27604 }, { "epoch": 0.68, "learning_rate": 4.895052867127271e-06, "loss": 1.8692, "step": 27605 }, { "epoch": 0.68, "learning_rate": 4.894366413713912e-06, "loss": 1.8776, "step": 27606 }, { "epoch": 0.68, "learning_rate": 4.893679992840421e-06, "loss": 2.0826, "step": 27607 }, { "epoch": 0.68, "learning_rate": 4.89299360451117e-06, "loss": 1.9132, "step": 27608 }, { "epoch": 0.68, "learning_rate": 4.8923072487305335e-06, "loss": 1.8944, "step": 27609 }, { "epoch": 0.68, "learning_rate": 4.891620925502892e-06, "loss": 2.0275, "step": 27610 }, { "epoch": 0.68, "learning_rate": 4.8909346348326115e-06, "loss": 2.0133, "step": 27611 }, { "epoch": 0.68, "learning_rate": 4.8902483767240695e-06, "loss": 2.0883, "step": 27612 }, { "epoch": 0.68, "learning_rate": 4.889562151181644e-06, "loss": 2.0345, "step": 27613 }, { "epoch": 0.68, "learning_rate": 4.8888759582097e-06, "loss": 2.0401, "step": 27614 }, { "epoch": 0.68, "learning_rate": 4.888189797812615e-06, "loss": 1.8497, "step": 27615 }, { "epoch": 0.68, "learning_rate": 4.887503669994762e-06, "loss": 2.0799, "step": 27616 }, { "epoch": 0.68, "learning_rate": 4.8868175747605195e-06, "loss": 1.9652, "step": 27617 }, { "epoch": 0.68, "learning_rate": 4.886131512114249e-06, "loss": 1.8709, "step": 27618 }, { "epoch": 0.68, "learning_rate": 4.885445482060331e-06, "loss": 1.9094, "step": 27619 }, { "epoch": 0.68, "learning_rate": 4.8847594846031385e-06, "loss": 1.8191, "step": 27620 }, { "epoch": 0.68, "learning_rate": 4.884073519747038e-06, "loss": 1.8564, "step": 27621 }, { "epoch": 0.68, "learning_rate": 4.883387587496403e-06, "loss": 1.866, "step": 27622 }, { "epoch": 0.68, "learning_rate": 4.88270168785561e-06, "loss": 2.021, "step": 27623 }, { "epoch": 0.68, "learning_rate": 4.8820158208290236e-06, "loss": 1.858, "step": 27624 }, { "epoch": 0.68, "learning_rate": 4.881329986421023e-06, "loss": 1.7281, "step": 27625 }, { "epoch": 0.68, "learning_rate": 4.880644184635969e-06, "loss": 1.9385, "step": 27626 }, { "epoch": 0.68, "learning_rate": 4.8799584154782385e-06, "loss": 1.9513, "step": 27627 }, { "epoch": 0.68, "learning_rate": 4.879272678952206e-06, "loss": 1.8462, "step": 27628 }, { "epoch": 0.68, "learning_rate": 4.878586975062234e-06, "loss": 1.9547, "step": 27629 }, { "epoch": 0.68, "learning_rate": 4.877901303812699e-06, "loss": 1.8157, "step": 27630 }, { "epoch": 0.68, "learning_rate": 4.877215665207965e-06, "loss": 1.8755, "step": 27631 }, { "epoch": 0.68, "learning_rate": 4.8765300592524045e-06, "loss": 2.114, "step": 27632 }, { "epoch": 0.68, "learning_rate": 4.875844485950388e-06, "loss": 1.9905, "step": 27633 }, { "epoch": 0.68, "learning_rate": 4.8751589453062845e-06, "loss": 2.0933, "step": 27634 }, { "epoch": 0.68, "learning_rate": 4.874473437324468e-06, "loss": 1.977, "step": 27635 }, { "epoch": 0.68, "learning_rate": 4.873787962009296e-06, "loss": 1.8658, "step": 27636 }, { "epoch": 0.68, "learning_rate": 4.873102519365146e-06, "loss": 2.1719, "step": 27637 }, { "epoch": 0.68, "learning_rate": 4.872417109396388e-06, "loss": 2.0819, "step": 27638 }, { "epoch": 0.68, "learning_rate": 4.871731732107383e-06, "loss": 1.8919, "step": 27639 }, { "epoch": 0.68, "learning_rate": 4.871046387502507e-06, "loss": 2.0255, "step": 27640 }, { "epoch": 0.68, "learning_rate": 4.87036107558612e-06, "loss": 1.7445, "step": 27641 }, { "epoch": 0.68, "learning_rate": 4.869675796362593e-06, "loss": 1.9176, "step": 27642 }, { "epoch": 0.68, "learning_rate": 4.8689905498362975e-06, "loss": 1.8836, "step": 27643 }, { "epoch": 0.68, "learning_rate": 4.868305336011595e-06, "loss": 1.9978, "step": 27644 }, { "epoch": 0.68, "learning_rate": 4.867620154892858e-06, "loss": 2.0754, "step": 27645 }, { "epoch": 0.68, "learning_rate": 4.866935006484447e-06, "loss": 2.0291, "step": 27646 }, { "epoch": 0.68, "learning_rate": 4.866249890790731e-06, "loss": 1.8234, "step": 27647 }, { "epoch": 0.68, "learning_rate": 4.865564807816082e-06, "loss": 1.943, "step": 27648 }, { "epoch": 0.68, "learning_rate": 4.864879757564857e-06, "loss": 2.0702, "step": 27649 }, { "epoch": 0.68, "learning_rate": 4.864194740041428e-06, "loss": 1.781, "step": 27650 }, { "epoch": 0.68, "learning_rate": 4.86350975525016e-06, "loss": 1.9571, "step": 27651 }, { "epoch": 0.68, "learning_rate": 4.862824803195417e-06, "loss": 2.0769, "step": 27652 }, { "epoch": 0.68, "learning_rate": 4.862139883881569e-06, "loss": 1.9161, "step": 27653 }, { "epoch": 0.68, "learning_rate": 4.861454997312975e-06, "loss": 2.0795, "step": 27654 }, { "epoch": 0.68, "learning_rate": 4.860770143494008e-06, "loss": 1.9807, "step": 27655 }, { "epoch": 0.68, "learning_rate": 4.860085322429022e-06, "loss": 2.0916, "step": 27656 }, { "epoch": 0.68, "learning_rate": 4.859400534122388e-06, "loss": 2.074, "step": 27657 }, { "epoch": 0.68, "learning_rate": 4.8587157785784735e-06, "loss": 1.9007, "step": 27658 }, { "epoch": 0.68, "learning_rate": 4.858031055801635e-06, "loss": 1.928, "step": 27659 }, { "epoch": 0.68, "learning_rate": 4.857346365796244e-06, "loss": 2.0208, "step": 27660 }, { "epoch": 0.68, "learning_rate": 4.856661708566657e-06, "loss": 2.141, "step": 27661 }, { "epoch": 0.68, "learning_rate": 4.8559770841172404e-06, "loss": 1.8535, "step": 27662 }, { "epoch": 0.68, "learning_rate": 4.855292492452364e-06, "loss": 1.8936, "step": 27663 }, { "epoch": 0.68, "learning_rate": 4.854607933576379e-06, "loss": 1.9523, "step": 27664 }, { "epoch": 0.68, "learning_rate": 4.853923407493656e-06, "loss": 1.8651, "step": 27665 }, { "epoch": 0.68, "learning_rate": 4.853238914208559e-06, "loss": 1.8804, "step": 27666 }, { "epoch": 0.68, "learning_rate": 4.8525544537254445e-06, "loss": 1.7959, "step": 27667 }, { "epoch": 0.68, "learning_rate": 4.851870026048678e-06, "loss": 1.8991, "step": 27668 }, { "epoch": 0.68, "learning_rate": 4.8511856311826215e-06, "loss": 2.0118, "step": 27669 }, { "epoch": 0.68, "learning_rate": 4.85050126913164e-06, "loss": 1.9419, "step": 27670 }, { "epoch": 0.68, "learning_rate": 4.849816939900088e-06, "loss": 1.7359, "step": 27671 }, { "epoch": 0.68, "learning_rate": 4.849132643492333e-06, "loss": 1.8771, "step": 27672 }, { "epoch": 0.68, "learning_rate": 4.848448379912736e-06, "loss": 2.0436, "step": 27673 }, { "epoch": 0.68, "learning_rate": 4.847764149165653e-06, "loss": 1.9316, "step": 27674 }, { "epoch": 0.68, "learning_rate": 4.847079951255454e-06, "loss": 2.0568, "step": 27675 }, { "epoch": 0.68, "learning_rate": 4.8463957861864875e-06, "loss": 1.9844, "step": 27676 }, { "epoch": 0.68, "learning_rate": 4.845711653963121e-06, "loss": 1.958, "step": 27677 }, { "epoch": 0.68, "learning_rate": 4.845027554589717e-06, "loss": 2.0071, "step": 27678 }, { "epoch": 0.68, "learning_rate": 4.8443434880706296e-06, "loss": 1.8633, "step": 27679 }, { "epoch": 0.68, "learning_rate": 4.843659454410221e-06, "loss": 1.9873, "step": 27680 }, { "epoch": 0.68, "learning_rate": 4.8429754536128544e-06, "loss": 2.0689, "step": 27681 }, { "epoch": 0.68, "learning_rate": 4.842291485682883e-06, "loss": 1.9821, "step": 27682 }, { "epoch": 0.68, "learning_rate": 4.841607550624668e-06, "loss": 1.9686, "step": 27683 }, { "epoch": 0.68, "learning_rate": 4.840923648442574e-06, "loss": 2.1203, "step": 27684 }, { "epoch": 0.68, "learning_rate": 4.840239779140949e-06, "loss": 2.0873, "step": 27685 }, { "epoch": 0.68, "learning_rate": 4.839555942724159e-06, "loss": 1.8365, "step": 27686 }, { "epoch": 0.68, "learning_rate": 4.838872139196561e-06, "loss": 1.9939, "step": 27687 }, { "epoch": 0.68, "learning_rate": 4.838188368562516e-06, "loss": 1.9532, "step": 27688 }, { "epoch": 0.68, "learning_rate": 4.837504630826375e-06, "loss": 1.9886, "step": 27689 }, { "epoch": 0.68, "learning_rate": 4.836820925992499e-06, "loss": 2.2813, "step": 27690 }, { "epoch": 0.68, "learning_rate": 4.83613725406525e-06, "loss": 1.804, "step": 27691 }, { "epoch": 0.68, "learning_rate": 4.8354536150489785e-06, "loss": 1.9401, "step": 27692 }, { "epoch": 0.68, "learning_rate": 4.834770008948046e-06, "loss": 1.9498, "step": 27693 }, { "epoch": 0.68, "learning_rate": 4.8340864357668046e-06, "loss": 2.0083, "step": 27694 }, { "epoch": 0.68, "learning_rate": 4.833402895509615e-06, "loss": 1.9157, "step": 27695 }, { "epoch": 0.68, "learning_rate": 4.832719388180835e-06, "loss": 2.0723, "step": 27696 }, { "epoch": 0.68, "learning_rate": 4.832035913784814e-06, "loss": 1.8879, "step": 27697 }, { "epoch": 0.68, "learning_rate": 4.831352472325917e-06, "loss": 2.115, "step": 27698 }, { "epoch": 0.68, "learning_rate": 4.830669063808491e-06, "loss": 2.2703, "step": 27699 }, { "epoch": 0.68, "learning_rate": 4.829985688236896e-06, "loss": 1.9864, "step": 27700 }, { "epoch": 0.68, "learning_rate": 4.829302345615486e-06, "loss": 1.8375, "step": 27701 }, { "epoch": 0.68, "learning_rate": 4.828619035948623e-06, "loss": 1.9554, "step": 27702 }, { "epoch": 0.68, "learning_rate": 4.82793575924065e-06, "loss": 1.7979, "step": 27703 }, { "epoch": 0.68, "learning_rate": 4.82725251549593e-06, "loss": 2.0053, "step": 27704 }, { "epoch": 0.68, "learning_rate": 4.826569304718814e-06, "loss": 1.938, "step": 27705 }, { "epoch": 0.68, "learning_rate": 4.825886126913663e-06, "loss": 1.884, "step": 27706 }, { "epoch": 0.68, "learning_rate": 4.825202982084821e-06, "loss": 2.063, "step": 27707 }, { "epoch": 0.68, "learning_rate": 4.824519870236651e-06, "loss": 2.0196, "step": 27708 }, { "epoch": 0.68, "learning_rate": 4.8238367913735e-06, "loss": 1.8729, "step": 27709 }, { "epoch": 0.68, "learning_rate": 4.8231537454997216e-06, "loss": 1.903, "step": 27710 }, { "epoch": 0.68, "learning_rate": 4.822470732619676e-06, "loss": 1.8351, "step": 27711 }, { "epoch": 0.68, "learning_rate": 4.821787752737709e-06, "loss": 1.9604, "step": 27712 }, { "epoch": 0.68, "learning_rate": 4.821104805858179e-06, "loss": 1.9526, "step": 27713 }, { "epoch": 0.68, "learning_rate": 4.8204218919854326e-06, "loss": 1.9961, "step": 27714 }, { "epoch": 0.68, "learning_rate": 4.819739011123825e-06, "loss": 1.8733, "step": 27715 }, { "epoch": 0.68, "learning_rate": 4.819056163277713e-06, "loss": 1.8603, "step": 27716 }, { "epoch": 0.68, "learning_rate": 4.81837334845144e-06, "loss": 2.0529, "step": 27717 }, { "epoch": 0.68, "learning_rate": 4.817690566649364e-06, "loss": 2.0433, "step": 27718 }, { "epoch": 0.68, "learning_rate": 4.817007817875833e-06, "loss": 1.9171, "step": 27719 }, { "epoch": 0.68, "learning_rate": 4.816325102135204e-06, "loss": 1.9603, "step": 27720 }, { "epoch": 0.68, "learning_rate": 4.8156424194318216e-06, "loss": 1.9119, "step": 27721 }, { "epoch": 0.68, "learning_rate": 4.81495976977004e-06, "loss": 1.8992, "step": 27722 }, { "epoch": 0.68, "learning_rate": 4.814277153154213e-06, "loss": 1.9942, "step": 27723 }, { "epoch": 0.68, "learning_rate": 4.813594569588683e-06, "loss": 1.9461, "step": 27724 }, { "epoch": 0.68, "learning_rate": 4.812912019077806e-06, "loss": 1.8107, "step": 27725 }, { "epoch": 0.68, "learning_rate": 4.812229501625935e-06, "loss": 1.9356, "step": 27726 }, { "epoch": 0.68, "learning_rate": 4.811547017237411e-06, "loss": 1.8741, "step": 27727 }, { "epoch": 0.68, "learning_rate": 4.810864565916594e-06, "loss": 1.8376, "step": 27728 }, { "epoch": 0.68, "learning_rate": 4.8101821476678235e-06, "loss": 1.8209, "step": 27729 }, { "epoch": 0.68, "learning_rate": 4.809499762495453e-06, "loss": 2.0124, "step": 27730 }, { "epoch": 0.68, "learning_rate": 4.808817410403838e-06, "loss": 1.8466, "step": 27731 }, { "epoch": 0.68, "learning_rate": 4.808135091397315e-06, "loss": 1.9253, "step": 27732 }, { "epoch": 0.68, "learning_rate": 4.807452805480245e-06, "loss": 1.8051, "step": 27733 }, { "epoch": 0.68, "learning_rate": 4.806770552656965e-06, "loss": 1.8175, "step": 27734 }, { "epoch": 0.68, "learning_rate": 4.8060883329318295e-06, "loss": 1.9681, "step": 27735 }, { "epoch": 0.68, "learning_rate": 4.805406146309186e-06, "loss": 1.8759, "step": 27736 }, { "epoch": 0.68, "learning_rate": 4.804723992793385e-06, "loss": 1.9197, "step": 27737 }, { "epoch": 0.68, "learning_rate": 4.804041872388767e-06, "loss": 1.8055, "step": 27738 }, { "epoch": 0.68, "learning_rate": 4.8033597850996845e-06, "loss": 1.9217, "step": 27739 }, { "epoch": 0.68, "learning_rate": 4.8026777309304836e-06, "loss": 1.9272, "step": 27740 }, { "epoch": 0.68, "learning_rate": 4.801995709885515e-06, "loss": 2.0248, "step": 27741 }, { "epoch": 0.68, "learning_rate": 4.801313721969118e-06, "loss": 1.9489, "step": 27742 }, { "epoch": 0.68, "learning_rate": 4.800631767185642e-06, "loss": 1.914, "step": 27743 }, { "epoch": 0.68, "learning_rate": 4.799949845539439e-06, "loss": 1.912, "step": 27744 }, { "epoch": 0.68, "learning_rate": 4.799267957034846e-06, "loss": 1.972, "step": 27745 }, { "epoch": 0.68, "learning_rate": 4.798586101676216e-06, "loss": 2.059, "step": 27746 }, { "epoch": 0.68, "learning_rate": 4.7979042794678885e-06, "loss": 2.0484, "step": 27747 }, { "epoch": 0.68, "learning_rate": 4.797222490414214e-06, "loss": 2.0794, "step": 27748 }, { "epoch": 0.68, "learning_rate": 4.796540734519538e-06, "loss": 2.0964, "step": 27749 }, { "epoch": 0.68, "learning_rate": 4.795859011788201e-06, "loss": 1.9711, "step": 27750 }, { "epoch": 0.68, "learning_rate": 4.7951773222245535e-06, "loss": 1.834, "step": 27751 }, { "epoch": 0.68, "learning_rate": 4.794495665832934e-06, "loss": 2.0569, "step": 27752 }, { "epoch": 0.68, "learning_rate": 4.793814042617691e-06, "loss": 1.9357, "step": 27753 }, { "epoch": 0.68, "learning_rate": 4.7931324525831655e-06, "loss": 1.9795, "step": 27754 }, { "epoch": 0.68, "learning_rate": 4.792450895733709e-06, "loss": 2.0946, "step": 27755 }, { "epoch": 0.68, "learning_rate": 4.791769372073656e-06, "loss": 2.0228, "step": 27756 }, { "epoch": 0.68, "learning_rate": 4.791087881607353e-06, "loss": 1.9166, "step": 27757 }, { "epoch": 0.68, "learning_rate": 4.7904064243391456e-06, "loss": 2.0351, "step": 27758 }, { "epoch": 0.68, "learning_rate": 4.78972500027338e-06, "loss": 2.0893, "step": 27759 }, { "epoch": 0.68, "learning_rate": 4.78904360941439e-06, "loss": 2.0792, "step": 27760 }, { "epoch": 0.68, "learning_rate": 4.788362251766528e-06, "loss": 1.8105, "step": 27761 }, { "epoch": 0.68, "learning_rate": 4.787680927334127e-06, "loss": 1.8874, "step": 27762 }, { "epoch": 0.68, "learning_rate": 4.786999636121536e-06, "loss": 1.8122, "step": 27763 }, { "epoch": 0.68, "learning_rate": 4.786318378133097e-06, "loss": 1.9896, "step": 27764 }, { "epoch": 0.68, "learning_rate": 4.785637153373147e-06, "loss": 1.9895, "step": 27765 }, { "epoch": 0.68, "learning_rate": 4.784955961846034e-06, "loss": 1.965, "step": 27766 }, { "epoch": 0.68, "learning_rate": 4.7842748035560935e-06, "loss": 2.2379, "step": 27767 }, { "epoch": 0.68, "learning_rate": 4.783593678507669e-06, "loss": 2.0711, "step": 27768 }, { "epoch": 0.68, "learning_rate": 4.782912586705105e-06, "loss": 1.9125, "step": 27769 }, { "epoch": 0.68, "learning_rate": 4.782231528152737e-06, "loss": 1.9763, "step": 27770 }, { "epoch": 0.68, "learning_rate": 4.781550502854906e-06, "loss": 2.0324, "step": 27771 }, { "epoch": 0.68, "learning_rate": 4.780869510815955e-06, "loss": 1.9651, "step": 27772 }, { "epoch": 0.68, "learning_rate": 4.780188552040228e-06, "loss": 2.0036, "step": 27773 }, { "epoch": 0.68, "learning_rate": 4.779507626532056e-06, "loss": 2.0415, "step": 27774 }, { "epoch": 0.68, "learning_rate": 4.7788267342957836e-06, "loss": 2.0769, "step": 27775 }, { "epoch": 0.68, "learning_rate": 4.778145875335753e-06, "loss": 1.9051, "step": 27776 }, { "epoch": 0.68, "learning_rate": 4.777465049656298e-06, "loss": 1.931, "step": 27777 }, { "epoch": 0.68, "learning_rate": 4.776784257261759e-06, "loss": 2.014, "step": 27778 }, { "epoch": 0.68, "learning_rate": 4.7761034981564805e-06, "loss": 2.1691, "step": 27779 }, { "epoch": 0.68, "learning_rate": 4.775422772344793e-06, "loss": 2.0082, "step": 27780 }, { "epoch": 0.68, "learning_rate": 4.774742079831043e-06, "loss": 1.8272, "step": 27781 }, { "epoch": 0.68, "learning_rate": 4.7740614206195605e-06, "loss": 1.8794, "step": 27782 }, { "epoch": 0.68, "learning_rate": 4.773380794714687e-06, "loss": 1.9344, "step": 27783 }, { "epoch": 0.68, "learning_rate": 4.772700202120765e-06, "loss": 1.7906, "step": 27784 }, { "epoch": 0.68, "learning_rate": 4.772019642842125e-06, "loss": 2.0412, "step": 27785 }, { "epoch": 0.68, "learning_rate": 4.77133911688311e-06, "loss": 1.9068, "step": 27786 }, { "epoch": 0.68, "learning_rate": 4.7706586242480525e-06, "loss": 1.9044, "step": 27787 }, { "epoch": 0.68, "learning_rate": 4.769978164941292e-06, "loss": 1.9927, "step": 27788 }, { "epoch": 0.68, "learning_rate": 4.769297738967164e-06, "loss": 1.9019, "step": 27789 }, { "epoch": 0.68, "learning_rate": 4.768617346330005e-06, "loss": 1.7706, "step": 27790 }, { "epoch": 0.69, "learning_rate": 4.7679369870341584e-06, "loss": 1.957, "step": 27791 }, { "epoch": 0.69, "learning_rate": 4.7672566610839514e-06, "loss": 1.9903, "step": 27792 }, { "epoch": 0.69, "learning_rate": 4.766576368483721e-06, "loss": 2.072, "step": 27793 }, { "epoch": 0.69, "learning_rate": 4.7658961092378085e-06, "loss": 2.0649, "step": 27794 }, { "epoch": 0.69, "learning_rate": 4.7652158833505436e-06, "loss": 1.9317, "step": 27795 }, { "epoch": 0.69, "learning_rate": 4.764535690826267e-06, "loss": 1.8626, "step": 27796 }, { "epoch": 0.69, "learning_rate": 4.763855531669307e-06, "loss": 1.9125, "step": 27797 }, { "epoch": 0.69, "learning_rate": 4.763175405884002e-06, "loss": 1.6749, "step": 27798 }, { "epoch": 0.69, "learning_rate": 4.76249531347469e-06, "loss": 2.0117, "step": 27799 }, { "epoch": 0.69, "learning_rate": 4.761815254445699e-06, "loss": 2.1731, "step": 27800 }, { "epoch": 0.69, "learning_rate": 4.7611352288013655e-06, "loss": 2.0458, "step": 27801 }, { "epoch": 0.69, "learning_rate": 4.76045523654603e-06, "loss": 1.8339, "step": 27802 }, { "epoch": 0.69, "learning_rate": 4.759775277684015e-06, "loss": 1.9841, "step": 27803 }, { "epoch": 0.69, "learning_rate": 4.759095352219666e-06, "loss": 2.0342, "step": 27804 }, { "epoch": 0.69, "learning_rate": 4.758415460157305e-06, "loss": 2.0032, "step": 27805 }, { "epoch": 0.69, "learning_rate": 4.757735601501271e-06, "loss": 2.0135, "step": 27806 }, { "epoch": 0.69, "learning_rate": 4.757055776255896e-06, "loss": 1.9163, "step": 27807 }, { "epoch": 0.69, "learning_rate": 4.756375984425513e-06, "loss": 1.8709, "step": 27808 }, { "epoch": 0.69, "learning_rate": 4.755696226014459e-06, "loss": 1.9862, "step": 27809 }, { "epoch": 0.69, "learning_rate": 4.7550165010270574e-06, "loss": 1.8369, "step": 27810 }, { "epoch": 0.69, "learning_rate": 4.754336809467646e-06, "loss": 1.8572, "step": 27811 }, { "epoch": 0.69, "learning_rate": 4.75365715134056e-06, "loss": 1.922, "step": 27812 }, { "epoch": 0.69, "learning_rate": 4.752977526650121e-06, "loss": 2.1642, "step": 27813 }, { "epoch": 0.69, "learning_rate": 4.75229793540067e-06, "loss": 2.0072, "step": 27814 }, { "epoch": 0.69, "learning_rate": 4.751618377596532e-06, "loss": 2.0745, "step": 27815 }, { "epoch": 0.69, "learning_rate": 4.75093885324204e-06, "loss": 2.0806, "step": 27816 }, { "epoch": 0.69, "learning_rate": 4.7502593623415294e-06, "loss": 1.9693, "step": 27817 }, { "epoch": 0.69, "learning_rate": 4.749579904899322e-06, "loss": 1.8922, "step": 27818 }, { "epoch": 0.69, "learning_rate": 4.748900480919758e-06, "loss": 1.9352, "step": 27819 }, { "epoch": 0.69, "learning_rate": 4.748221090407158e-06, "loss": 1.9747, "step": 27820 }, { "epoch": 0.69, "learning_rate": 4.747541733365857e-06, "loss": 1.9757, "step": 27821 }, { "epoch": 0.69, "learning_rate": 4.746862409800187e-06, "loss": 2.0131, "step": 27822 }, { "epoch": 0.69, "learning_rate": 4.746183119714471e-06, "loss": 1.8899, "step": 27823 }, { "epoch": 0.69, "learning_rate": 4.745503863113043e-06, "loss": 1.9675, "step": 27824 }, { "epoch": 0.69, "learning_rate": 4.7448246400002316e-06, "loss": 2.0373, "step": 27825 }, { "epoch": 0.69, "learning_rate": 4.744145450380369e-06, "loss": 1.9426, "step": 27826 }, { "epoch": 0.69, "learning_rate": 4.743466294257776e-06, "loss": 1.9521, "step": 27827 }, { "epoch": 0.69, "learning_rate": 4.742787171636785e-06, "loss": 1.8255, "step": 27828 }, { "epoch": 0.69, "learning_rate": 4.742108082521729e-06, "loss": 1.871, "step": 27829 }, { "epoch": 0.69, "learning_rate": 4.741429026916927e-06, "loss": 2.0802, "step": 27830 }, { "epoch": 0.69, "learning_rate": 4.740750004826713e-06, "loss": 1.8555, "step": 27831 }, { "epoch": 0.69, "learning_rate": 4.7400710162554165e-06, "loss": 1.9998, "step": 27832 }, { "epoch": 0.69, "learning_rate": 4.739392061207358e-06, "loss": 2.0176, "step": 27833 }, { "epoch": 0.69, "learning_rate": 4.738713139686871e-06, "loss": 1.9734, "step": 27834 }, { "epoch": 0.69, "learning_rate": 4.738034251698277e-06, "loss": 1.9886, "step": 27835 }, { "epoch": 0.69, "learning_rate": 4.737355397245905e-06, "loss": 2.0062, "step": 27836 }, { "epoch": 0.69, "learning_rate": 4.736676576334086e-06, "loss": 2.0719, "step": 27837 }, { "epoch": 0.69, "learning_rate": 4.735997788967138e-06, "loss": 2.009, "step": 27838 }, { "epoch": 0.69, "learning_rate": 4.735319035149396e-06, "loss": 2.0175, "step": 27839 }, { "epoch": 0.69, "learning_rate": 4.734640314885178e-06, "loss": 1.9565, "step": 27840 }, { "epoch": 0.69, "learning_rate": 4.733961628178814e-06, "loss": 2.003, "step": 27841 }, { "epoch": 0.69, "learning_rate": 4.7332829750346265e-06, "loss": 1.7505, "step": 27842 }, { "epoch": 0.69, "learning_rate": 4.732604355456945e-06, "loss": 2.0706, "step": 27843 }, { "epoch": 0.69, "learning_rate": 4.731925769450096e-06, "loss": 1.9632, "step": 27844 }, { "epoch": 0.69, "learning_rate": 4.731247217018395e-06, "loss": 2.0722, "step": 27845 }, { "epoch": 0.69, "learning_rate": 4.730568698166175e-06, "loss": 1.8735, "step": 27846 }, { "epoch": 0.69, "learning_rate": 4.729890212897761e-06, "loss": 1.8564, "step": 27847 }, { "epoch": 0.69, "learning_rate": 4.729211761217469e-06, "loss": 1.907, "step": 27848 }, { "epoch": 0.69, "learning_rate": 4.728533343129633e-06, "loss": 1.9861, "step": 27849 }, { "epoch": 0.69, "learning_rate": 4.727854958638569e-06, "loss": 2.2225, "step": 27850 }, { "epoch": 0.69, "learning_rate": 4.727176607748602e-06, "loss": 1.935, "step": 27851 }, { "epoch": 0.69, "learning_rate": 4.726498290464062e-06, "loss": 2.029, "step": 27852 }, { "epoch": 0.69, "learning_rate": 4.725820006789262e-06, "loss": 2.0085, "step": 27853 }, { "epoch": 0.69, "learning_rate": 4.725141756728535e-06, "loss": 1.8747, "step": 27854 }, { "epoch": 0.69, "learning_rate": 4.7244635402861935e-06, "loss": 2.1215, "step": 27855 }, { "epoch": 0.69, "learning_rate": 4.723785357466566e-06, "loss": 1.8143, "step": 27856 }, { "epoch": 0.69, "learning_rate": 4.723107208273977e-06, "loss": 1.8874, "step": 27857 }, { "epoch": 0.69, "learning_rate": 4.722429092712743e-06, "loss": 1.7489, "step": 27858 }, { "epoch": 0.69, "learning_rate": 4.721751010787187e-06, "loss": 1.8411, "step": 27859 }, { "epoch": 0.69, "learning_rate": 4.721072962501633e-06, "loss": 1.9731, "step": 27860 }, { "epoch": 0.69, "learning_rate": 4.720394947860399e-06, "loss": 2.0827, "step": 27861 }, { "epoch": 0.69, "learning_rate": 4.719716966867815e-06, "loss": 2.0075, "step": 27862 }, { "epoch": 0.69, "learning_rate": 4.719039019528191e-06, "loss": 2.0308, "step": 27863 }, { "epoch": 0.69, "learning_rate": 4.718361105845851e-06, "loss": 1.903, "step": 27864 }, { "epoch": 0.69, "learning_rate": 4.717683225825123e-06, "loss": 1.8021, "step": 27865 }, { "epoch": 0.69, "learning_rate": 4.717005379470315e-06, "loss": 1.9941, "step": 27866 }, { "epoch": 0.69, "learning_rate": 4.716327566785759e-06, "loss": 1.9067, "step": 27867 }, { "epoch": 0.69, "learning_rate": 4.715649787775766e-06, "loss": 2.0219, "step": 27868 }, { "epoch": 0.69, "learning_rate": 4.714972042444657e-06, "loss": 2.0247, "step": 27869 }, { "epoch": 0.69, "learning_rate": 4.7142943307967585e-06, "loss": 2.0482, "step": 27870 }, { "epoch": 0.69, "learning_rate": 4.713616652836381e-06, "loss": 1.9992, "step": 27871 }, { "epoch": 0.69, "learning_rate": 4.71293900856785e-06, "loss": 1.9144, "step": 27872 }, { "epoch": 0.69, "learning_rate": 4.71226139799548e-06, "loss": 1.8841, "step": 27873 }, { "epoch": 0.69, "learning_rate": 4.71158382112359e-06, "loss": 1.8669, "step": 27874 }, { "epoch": 0.69, "learning_rate": 4.710906277956504e-06, "loss": 1.9739, "step": 27875 }, { "epoch": 0.69, "learning_rate": 4.710228768498532e-06, "loss": 2.0947, "step": 27876 }, { "epoch": 0.69, "learning_rate": 4.709551292753995e-06, "loss": 2.03, "step": 27877 }, { "epoch": 0.69, "learning_rate": 4.708873850727212e-06, "loss": 2.0504, "step": 27878 }, { "epoch": 0.69, "learning_rate": 4.708196442422502e-06, "loss": 1.9083, "step": 27879 }, { "epoch": 0.69, "learning_rate": 4.707519067844183e-06, "loss": 1.89, "step": 27880 }, { "epoch": 0.69, "learning_rate": 4.706841726996566e-06, "loss": 1.7946, "step": 27881 }, { "epoch": 0.69, "learning_rate": 4.706164419883976e-06, "loss": 2.0385, "step": 27882 }, { "epoch": 0.69, "learning_rate": 4.705487146510721e-06, "loss": 1.9211, "step": 27883 }, { "epoch": 0.69, "learning_rate": 4.704809906881122e-06, "loss": 1.9778, "step": 27884 }, { "epoch": 0.69, "learning_rate": 4.7041327009995e-06, "loss": 1.9425, "step": 27885 }, { "epoch": 0.69, "learning_rate": 4.703455528870161e-06, "loss": 1.8753, "step": 27886 }, { "epoch": 0.69, "learning_rate": 4.70277839049743e-06, "loss": 2.0353, "step": 27887 }, { "epoch": 0.69, "learning_rate": 4.702101285885615e-06, "loss": 1.9839, "step": 27888 }, { "epoch": 0.69, "learning_rate": 4.701424215039035e-06, "loss": 2.0096, "step": 27889 }, { "epoch": 0.69, "learning_rate": 4.700747177962009e-06, "loss": 1.9599, "step": 27890 }, { "epoch": 0.69, "learning_rate": 4.700070174658845e-06, "loss": 1.9355, "step": 27891 }, { "epoch": 0.69, "learning_rate": 4.699393205133864e-06, "loss": 2.1626, "step": 27892 }, { "epoch": 0.69, "learning_rate": 4.698716269391373e-06, "loss": 1.9065, "step": 27893 }, { "epoch": 0.69, "learning_rate": 4.6980393674356925e-06, "loss": 2.0392, "step": 27894 }, { "epoch": 0.69, "learning_rate": 4.697362499271134e-06, "loss": 1.9217, "step": 27895 }, { "epoch": 0.69, "learning_rate": 4.696685664902013e-06, "loss": 2.1817, "step": 27896 }, { "epoch": 0.69, "learning_rate": 4.696008864332646e-06, "loss": 2.0759, "step": 27897 }, { "epoch": 0.69, "learning_rate": 4.69533209756734e-06, "loss": 1.9096, "step": 27898 }, { "epoch": 0.69, "learning_rate": 4.694655364610411e-06, "loss": 1.71, "step": 27899 }, { "epoch": 0.69, "learning_rate": 4.693978665466174e-06, "loss": 2.1444, "step": 27900 }, { "epoch": 0.69, "learning_rate": 4.69330200013894e-06, "loss": 1.8816, "step": 27901 }, { "epoch": 0.69, "learning_rate": 4.692625368633023e-06, "loss": 2.1232, "step": 27902 }, { "epoch": 0.69, "learning_rate": 4.691948770952731e-06, "loss": 1.9181, "step": 27903 }, { "epoch": 0.69, "learning_rate": 4.691272207102378e-06, "loss": 1.9763, "step": 27904 }, { "epoch": 0.69, "learning_rate": 4.690595677086284e-06, "loss": 1.8162, "step": 27905 }, { "epoch": 0.69, "learning_rate": 4.689919180908747e-06, "loss": 1.9489, "step": 27906 }, { "epoch": 0.69, "learning_rate": 4.689242718574092e-06, "loss": 2.0746, "step": 27907 }, { "epoch": 0.69, "learning_rate": 4.688566290086619e-06, "loss": 2.1541, "step": 27908 }, { "epoch": 0.69, "learning_rate": 4.6878898954506435e-06, "loss": 2.0343, "step": 27909 }, { "epoch": 0.69, "learning_rate": 4.687213534670482e-06, "loss": 1.9237, "step": 27910 }, { "epoch": 0.69, "learning_rate": 4.6865372077504355e-06, "loss": 1.9795, "step": 27911 }, { "epoch": 0.69, "learning_rate": 4.685860914694818e-06, "loss": 1.948, "step": 27912 }, { "epoch": 0.69, "learning_rate": 4.6851846555079415e-06, "loss": 1.9953, "step": 27913 }, { "epoch": 0.69, "learning_rate": 4.6845084301941155e-06, "loss": 2.0865, "step": 27914 }, { "epoch": 0.69, "learning_rate": 4.683832238757654e-06, "loss": 1.8894, "step": 27915 }, { "epoch": 0.69, "learning_rate": 4.683156081202857e-06, "loss": 1.8386, "step": 27916 }, { "epoch": 0.69, "learning_rate": 4.6824799575340395e-06, "loss": 1.9375, "step": 27917 }, { "epoch": 0.69, "learning_rate": 4.681803867755515e-06, "loss": 1.8899, "step": 27918 }, { "epoch": 0.69, "learning_rate": 4.681127811871582e-06, "loss": 1.7636, "step": 27919 }, { "epoch": 0.69, "learning_rate": 4.6804517898865596e-06, "loss": 1.9647, "step": 27920 }, { "epoch": 0.69, "learning_rate": 4.679775801804748e-06, "loss": 2.0779, "step": 27921 }, { "epoch": 0.69, "learning_rate": 4.679099847630458e-06, "loss": 1.86, "step": 27922 }, { "epoch": 0.69, "learning_rate": 4.678423927368003e-06, "loss": 1.9918, "step": 27923 }, { "epoch": 0.69, "learning_rate": 4.677748041021683e-06, "loss": 1.9856, "step": 27924 }, { "epoch": 0.69, "learning_rate": 4.677072188595812e-06, "loss": 1.9148, "step": 27925 }, { "epoch": 0.69, "learning_rate": 4.676396370094692e-06, "loss": 2.0157, "step": 27926 }, { "epoch": 0.69, "learning_rate": 4.6757205855226325e-06, "loss": 1.9453, "step": 27927 }, { "epoch": 0.69, "learning_rate": 4.675044834883945e-06, "loss": 1.8658, "step": 27928 }, { "epoch": 0.69, "learning_rate": 4.674369118182929e-06, "loss": 2.0101, "step": 27929 }, { "epoch": 0.69, "learning_rate": 4.673693435423892e-06, "loss": 2.0391, "step": 27930 }, { "epoch": 0.69, "learning_rate": 4.673017786611145e-06, "loss": 2.0819, "step": 27931 }, { "epoch": 0.69, "learning_rate": 4.672342171748989e-06, "loss": 1.9572, "step": 27932 }, { "epoch": 0.69, "learning_rate": 4.671666590841738e-06, "loss": 1.9095, "step": 27933 }, { "epoch": 0.69, "learning_rate": 4.670991043893689e-06, "loss": 2.0939, "step": 27934 }, { "epoch": 0.69, "learning_rate": 4.670315530909153e-06, "loss": 1.8182, "step": 27935 }, { "epoch": 0.69, "learning_rate": 4.66964005189243e-06, "loss": 1.8612, "step": 27936 }, { "epoch": 0.69, "learning_rate": 4.6689646068478286e-06, "loss": 1.8987, "step": 27937 }, { "epoch": 0.69, "learning_rate": 4.668289195779656e-06, "loss": 1.9588, "step": 27938 }, { "epoch": 0.69, "learning_rate": 4.6676138186922105e-06, "loss": 1.9775, "step": 27939 }, { "epoch": 0.69, "learning_rate": 4.666938475589804e-06, "loss": 1.8581, "step": 27940 }, { "epoch": 0.69, "learning_rate": 4.666263166476732e-06, "loss": 1.9434, "step": 27941 }, { "epoch": 0.69, "learning_rate": 4.665587891357304e-06, "loss": 1.8452, "step": 27942 }, { "epoch": 0.69, "learning_rate": 4.664912650235826e-06, "loss": 1.7142, "step": 27943 }, { "epoch": 0.69, "learning_rate": 4.664237443116595e-06, "loss": 1.9857, "step": 27944 }, { "epoch": 0.69, "learning_rate": 4.66356227000392e-06, "loss": 1.7973, "step": 27945 }, { "epoch": 0.69, "learning_rate": 4.6628871309021e-06, "loss": 1.9012, "step": 27946 }, { "epoch": 0.69, "learning_rate": 4.662212025815438e-06, "loss": 1.7696, "step": 27947 }, { "epoch": 0.69, "learning_rate": 4.6615369547482394e-06, "loss": 1.9933, "step": 27948 }, { "epoch": 0.69, "learning_rate": 4.6608619177048055e-06, "loss": 1.8062, "step": 27949 }, { "epoch": 0.69, "learning_rate": 4.660186914689442e-06, "loss": 1.8368, "step": 27950 }, { "epoch": 0.69, "learning_rate": 4.659511945706443e-06, "loss": 1.9065, "step": 27951 }, { "epoch": 0.69, "learning_rate": 4.658837010760116e-06, "loss": 1.9412, "step": 27952 }, { "epoch": 0.69, "learning_rate": 4.658162109854765e-06, "loss": 2.0476, "step": 27953 }, { "epoch": 0.69, "learning_rate": 4.657487242994683e-06, "loss": 2.0041, "step": 27954 }, { "epoch": 0.69, "learning_rate": 4.65681241018418e-06, "loss": 2.1136, "step": 27955 }, { "epoch": 0.69, "learning_rate": 4.6561376114275496e-06, "loss": 1.9821, "step": 27956 }, { "epoch": 0.69, "learning_rate": 4.655462846729095e-06, "loss": 1.9774, "step": 27957 }, { "epoch": 0.69, "learning_rate": 4.6547881160931215e-06, "loss": 2.0091, "step": 27958 }, { "epoch": 0.69, "learning_rate": 4.654113419523921e-06, "loss": 1.8323, "step": 27959 }, { "epoch": 0.69, "learning_rate": 4.653438757025802e-06, "loss": 1.9537, "step": 27960 }, { "epoch": 0.69, "learning_rate": 4.652764128603058e-06, "loss": 1.9092, "step": 27961 }, { "epoch": 0.69, "learning_rate": 4.652089534259989e-06, "loss": 1.9476, "step": 27962 }, { "epoch": 0.69, "learning_rate": 4.651414974000901e-06, "loss": 1.9583, "step": 27963 }, { "epoch": 0.69, "learning_rate": 4.650740447830083e-06, "loss": 1.9526, "step": 27964 }, { "epoch": 0.69, "learning_rate": 4.650065955751841e-06, "loss": 2.2773, "step": 27965 }, { "epoch": 0.69, "learning_rate": 4.649391497770471e-06, "loss": 1.89, "step": 27966 }, { "epoch": 0.69, "learning_rate": 4.6487170738902745e-06, "loss": 2.0029, "step": 27967 }, { "epoch": 0.69, "learning_rate": 4.64804268411555e-06, "loss": 1.8997, "step": 27968 }, { "epoch": 0.69, "learning_rate": 4.64736832845059e-06, "loss": 1.9188, "step": 27969 }, { "epoch": 0.69, "learning_rate": 4.6466940068997e-06, "loss": 1.9666, "step": 27970 }, { "epoch": 0.69, "learning_rate": 4.646019719467171e-06, "loss": 2.1582, "step": 27971 }, { "epoch": 0.69, "learning_rate": 4.645345466157301e-06, "loss": 2.1081, "step": 27972 }, { "epoch": 0.69, "learning_rate": 4.644671246974396e-06, "loss": 1.9534, "step": 27973 }, { "epoch": 0.69, "learning_rate": 4.64399706192274e-06, "loss": 1.8508, "step": 27974 }, { "epoch": 0.69, "learning_rate": 4.6433229110066414e-06, "loss": 1.9298, "step": 27975 }, { "epoch": 0.69, "learning_rate": 4.642648794230388e-06, "loss": 2.0415, "step": 27976 }, { "epoch": 0.69, "learning_rate": 4.64197471159828e-06, "loss": 1.923, "step": 27977 }, { "epoch": 0.69, "learning_rate": 4.641300663114615e-06, "loss": 1.9026, "step": 27978 }, { "epoch": 0.69, "learning_rate": 4.640626648783685e-06, "loss": 2.0026, "step": 27979 }, { "epoch": 0.69, "learning_rate": 4.639952668609787e-06, "loss": 1.9038, "step": 27980 }, { "epoch": 0.69, "learning_rate": 4.639278722597222e-06, "loss": 1.8838, "step": 27981 }, { "epoch": 0.69, "learning_rate": 4.638604810750278e-06, "loss": 1.9026, "step": 27982 }, { "epoch": 0.69, "learning_rate": 4.637930933073251e-06, "loss": 1.917, "step": 27983 }, { "epoch": 0.69, "learning_rate": 4.637257089570439e-06, "loss": 2.0972, "step": 27984 }, { "epoch": 0.69, "learning_rate": 4.636583280246134e-06, "loss": 1.8893, "step": 27985 }, { "epoch": 0.69, "learning_rate": 4.635909505104636e-06, "loss": 1.9038, "step": 27986 }, { "epoch": 0.69, "learning_rate": 4.63523576415023e-06, "loss": 2.0462, "step": 27987 }, { "epoch": 0.69, "learning_rate": 4.63456205738722e-06, "loss": 2.0676, "step": 27988 }, { "epoch": 0.69, "learning_rate": 4.63388838481989e-06, "loss": 1.7329, "step": 27989 }, { "epoch": 0.69, "learning_rate": 4.633214746452539e-06, "loss": 1.927, "step": 27990 }, { "epoch": 0.69, "learning_rate": 4.632541142289462e-06, "loss": 2.009, "step": 27991 }, { "epoch": 0.69, "learning_rate": 4.631867572334946e-06, "loss": 1.9906, "step": 27992 }, { "epoch": 0.69, "learning_rate": 4.631194036593291e-06, "loss": 2.0186, "step": 27993 }, { "epoch": 0.69, "learning_rate": 4.630520535068783e-06, "loss": 1.9242, "step": 27994 }, { "epoch": 0.69, "learning_rate": 4.6298470677657185e-06, "loss": 1.9518, "step": 27995 }, { "epoch": 0.69, "learning_rate": 4.62917363468839e-06, "loss": 1.8438, "step": 27996 }, { "epoch": 0.69, "learning_rate": 4.628500235841086e-06, "loss": 1.8973, "step": 27997 }, { "epoch": 0.69, "learning_rate": 4.627826871228104e-06, "loss": 1.7429, "step": 27998 }, { "epoch": 0.69, "learning_rate": 4.627153540853728e-06, "loss": 2.0838, "step": 27999 }, { "epoch": 0.69, "learning_rate": 4.6264802447222536e-06, "loss": 1.9741, "step": 28000 }, { "epoch": 0.69, "learning_rate": 4.625806982837971e-06, "loss": 1.9877, "step": 28001 }, { "epoch": 0.69, "learning_rate": 4.625133755205173e-06, "loss": 2.0166, "step": 28002 }, { "epoch": 0.69, "learning_rate": 4.624460561828153e-06, "loss": 1.985, "step": 28003 }, { "epoch": 0.69, "learning_rate": 4.623787402711194e-06, "loss": 1.7884, "step": 28004 }, { "epoch": 0.69, "learning_rate": 4.6231142778585906e-06, "loss": 1.8187, "step": 28005 }, { "epoch": 0.69, "learning_rate": 4.622441187274635e-06, "loss": 1.9518, "step": 28006 }, { "epoch": 0.69, "learning_rate": 4.62176813096361e-06, "loss": 1.8639, "step": 28007 }, { "epoch": 0.69, "learning_rate": 4.621095108929813e-06, "loss": 1.9467, "step": 28008 }, { "epoch": 0.69, "learning_rate": 4.620422121177527e-06, "loss": 1.7552, "step": 28009 }, { "epoch": 0.69, "learning_rate": 4.619749167711043e-06, "loss": 1.9603, "step": 28010 }, { "epoch": 0.69, "learning_rate": 4.6190762485346554e-06, "loss": 1.9184, "step": 28011 }, { "epoch": 0.69, "learning_rate": 4.618403363652645e-06, "loss": 1.7808, "step": 28012 }, { "epoch": 0.69, "learning_rate": 4.617730513069306e-06, "loss": 1.8878, "step": 28013 }, { "epoch": 0.69, "learning_rate": 4.617057696788922e-06, "loss": 1.9648, "step": 28014 }, { "epoch": 0.69, "learning_rate": 4.616384914815784e-06, "loss": 2.0632, "step": 28015 }, { "epoch": 0.69, "learning_rate": 4.615712167154182e-06, "loss": 2.0058, "step": 28016 }, { "epoch": 0.69, "learning_rate": 4.615039453808398e-06, "loss": 1.8633, "step": 28017 }, { "epoch": 0.69, "learning_rate": 4.614366774782722e-06, "loss": 1.9205, "step": 28018 }, { "epoch": 0.69, "learning_rate": 4.613694130081442e-06, "loss": 2.0115, "step": 28019 }, { "epoch": 0.69, "learning_rate": 4.613021519708845e-06, "loss": 1.9058, "step": 28020 }, { "epoch": 0.69, "learning_rate": 4.6123489436692205e-06, "loss": 1.9938, "step": 28021 }, { "epoch": 0.69, "learning_rate": 4.611676401966849e-06, "loss": 1.9037, "step": 28022 }, { "epoch": 0.69, "learning_rate": 4.611003894606023e-06, "loss": 1.836, "step": 28023 }, { "epoch": 0.69, "learning_rate": 4.610331421591021e-06, "loss": 1.9473, "step": 28024 }, { "epoch": 0.69, "learning_rate": 4.609658982926134e-06, "loss": 1.976, "step": 28025 }, { "epoch": 0.69, "learning_rate": 4.60898657861565e-06, "loss": 2.1072, "step": 28026 }, { "epoch": 0.69, "learning_rate": 4.60831420866385e-06, "loss": 2.077, "step": 28027 }, { "epoch": 0.69, "learning_rate": 4.607641873075023e-06, "loss": 1.9236, "step": 28028 }, { "epoch": 0.69, "learning_rate": 4.606969571853447e-06, "loss": 1.9686, "step": 28029 }, { "epoch": 0.69, "learning_rate": 4.606297305003412e-06, "loss": 1.9447, "step": 28030 }, { "epoch": 0.69, "learning_rate": 4.605625072529207e-06, "loss": 1.8773, "step": 28031 }, { "epoch": 0.69, "learning_rate": 4.6049528744351055e-06, "loss": 1.9955, "step": 28032 }, { "epoch": 0.69, "learning_rate": 4.604280710725403e-06, "loss": 1.9578, "step": 28033 }, { "epoch": 0.69, "learning_rate": 4.603608581404373e-06, "loss": 2.082, "step": 28034 }, { "epoch": 0.69, "learning_rate": 4.602936486476305e-06, "loss": 1.9166, "step": 28035 }, { "epoch": 0.69, "learning_rate": 4.602264425945481e-06, "loss": 1.9896, "step": 28036 }, { "epoch": 0.69, "learning_rate": 4.601592399816185e-06, "loss": 2.0173, "step": 28037 }, { "epoch": 0.69, "learning_rate": 4.600920408092701e-06, "loss": 1.8597, "step": 28038 }, { "epoch": 0.69, "learning_rate": 4.600248450779313e-06, "loss": 1.8647, "step": 28039 }, { "epoch": 0.69, "learning_rate": 4.5995765278802985e-06, "loss": 1.8145, "step": 28040 }, { "epoch": 0.69, "learning_rate": 4.598904639399946e-06, "loss": 2.1292, "step": 28041 }, { "epoch": 0.69, "learning_rate": 4.59823278534253e-06, "loss": 1.8181, "step": 28042 }, { "epoch": 0.69, "learning_rate": 4.597560965712337e-06, "loss": 1.9087, "step": 28043 }, { "epoch": 0.69, "learning_rate": 4.596889180513654e-06, "loss": 1.8619, "step": 28044 }, { "epoch": 0.69, "learning_rate": 4.596217429750752e-06, "loss": 2.0113, "step": 28045 }, { "epoch": 0.69, "learning_rate": 4.595545713427922e-06, "loss": 1.9704, "step": 28046 }, { "epoch": 0.69, "learning_rate": 4.594874031549436e-06, "loss": 2.0754, "step": 28047 }, { "epoch": 0.69, "learning_rate": 4.594202384119579e-06, "loss": 1.8828, "step": 28048 }, { "epoch": 0.69, "learning_rate": 4.593530771142637e-06, "loss": 2.0472, "step": 28049 }, { "epoch": 0.69, "learning_rate": 4.592859192622881e-06, "loss": 1.8218, "step": 28050 }, { "epoch": 0.69, "learning_rate": 4.592187648564597e-06, "loss": 2.0272, "step": 28051 }, { "epoch": 0.69, "learning_rate": 4.591516138972063e-06, "loss": 2.0349, "step": 28052 }, { "epoch": 0.69, "learning_rate": 4.590844663849557e-06, "loss": 2.0442, "step": 28053 }, { "epoch": 0.69, "learning_rate": 4.590173223201362e-06, "loss": 1.9896, "step": 28054 }, { "epoch": 0.69, "learning_rate": 4.589501817031755e-06, "loss": 2.2233, "step": 28055 }, { "epoch": 0.69, "learning_rate": 4.5888304453450205e-06, "loss": 2.0798, "step": 28056 }, { "epoch": 0.69, "learning_rate": 4.58815910814543e-06, "loss": 1.8974, "step": 28057 }, { "epoch": 0.69, "learning_rate": 4.587487805437264e-06, "loss": 1.9487, "step": 28058 }, { "epoch": 0.69, "learning_rate": 4.586816537224806e-06, "loss": 1.9792, "step": 28059 }, { "epoch": 0.69, "learning_rate": 4.586145303512327e-06, "loss": 2.0471, "step": 28060 }, { "epoch": 0.69, "learning_rate": 4.585474104304113e-06, "loss": 1.8441, "step": 28061 }, { "epoch": 0.69, "learning_rate": 4.584802939604431e-06, "loss": 1.9106, "step": 28062 }, { "epoch": 0.69, "learning_rate": 4.5841318094175664e-06, "loss": 2.0366, "step": 28063 }, { "epoch": 0.69, "learning_rate": 4.583460713747798e-06, "loss": 1.8118, "step": 28064 }, { "epoch": 0.69, "learning_rate": 4.582789652599397e-06, "loss": 1.8422, "step": 28065 }, { "epoch": 0.69, "learning_rate": 4.582118625976645e-06, "loss": 1.9853, "step": 28066 }, { "epoch": 0.69, "learning_rate": 4.581447633883813e-06, "loss": 1.7788, "step": 28067 }, { "epoch": 0.69, "learning_rate": 4.580776676325182e-06, "loss": 2.0214, "step": 28068 }, { "epoch": 0.69, "learning_rate": 4.5801057533050305e-06, "loss": 1.9234, "step": 28069 }, { "epoch": 0.69, "learning_rate": 4.579434864827627e-06, "loss": 1.9306, "step": 28070 }, { "epoch": 0.69, "learning_rate": 4.578764010897253e-06, "loss": 1.9259, "step": 28071 }, { "epoch": 0.69, "learning_rate": 4.578093191518182e-06, "loss": 1.949, "step": 28072 }, { "epoch": 0.69, "learning_rate": 4.5774224066946904e-06, "loss": 1.9895, "step": 28073 }, { "epoch": 0.69, "learning_rate": 4.576751656431057e-06, "loss": 1.9443, "step": 28074 }, { "epoch": 0.69, "learning_rate": 4.576080940731548e-06, "loss": 1.9001, "step": 28075 }, { "epoch": 0.69, "learning_rate": 4.575410259600447e-06, "loss": 1.975, "step": 28076 }, { "epoch": 0.69, "learning_rate": 4.57473961304202e-06, "loss": 1.964, "step": 28077 }, { "epoch": 0.69, "learning_rate": 4.5740690010605456e-06, "loss": 1.967, "step": 28078 }, { "epoch": 0.69, "learning_rate": 4.573398423660301e-06, "loss": 1.9992, "step": 28079 }, { "epoch": 0.69, "learning_rate": 4.572727880845555e-06, "loss": 2.0481, "step": 28080 }, { "epoch": 0.69, "learning_rate": 4.572057372620584e-06, "loss": 2.0422, "step": 28081 }, { "epoch": 0.69, "learning_rate": 4.5713868989896586e-06, "loss": 1.9751, "step": 28082 }, { "epoch": 0.69, "learning_rate": 4.570716459957053e-06, "loss": 2.1695, "step": 28083 }, { "epoch": 0.69, "learning_rate": 4.570046055527044e-06, "loss": 1.9976, "step": 28084 }, { "epoch": 0.69, "learning_rate": 4.569375685703898e-06, "loss": 1.9351, "step": 28085 }, { "epoch": 0.69, "learning_rate": 4.568705350491894e-06, "loss": 1.9902, "step": 28086 }, { "epoch": 0.69, "learning_rate": 4.5680350498952974e-06, "loss": 1.885, "step": 28087 }, { "epoch": 0.69, "learning_rate": 4.567364783918383e-06, "loss": 1.9471, "step": 28088 }, { "epoch": 0.69, "learning_rate": 4.566694552565424e-06, "loss": 1.8402, "step": 28089 }, { "epoch": 0.69, "learning_rate": 4.56602435584069e-06, "loss": 1.8922, "step": 28090 }, { "epoch": 0.69, "learning_rate": 4.565354193748459e-06, "loss": 1.9892, "step": 28091 }, { "epoch": 0.69, "learning_rate": 4.564684066292991e-06, "loss": 1.8043, "step": 28092 }, { "epoch": 0.69, "learning_rate": 4.564013973478565e-06, "loss": 1.9599, "step": 28093 }, { "epoch": 0.69, "learning_rate": 4.563343915309453e-06, "loss": 1.9798, "step": 28094 }, { "epoch": 0.69, "learning_rate": 4.562673891789917e-06, "loss": 1.8561, "step": 28095 }, { "epoch": 0.69, "learning_rate": 4.562003902924232e-06, "loss": 2.086, "step": 28096 }, { "epoch": 0.69, "learning_rate": 4.561333948716674e-06, "loss": 1.9812, "step": 28097 }, { "epoch": 0.69, "learning_rate": 4.560664029171502e-06, "loss": 1.9364, "step": 28098 }, { "epoch": 0.69, "learning_rate": 4.559994144292995e-06, "loss": 1.9035, "step": 28099 }, { "epoch": 0.69, "learning_rate": 4.559324294085415e-06, "loss": 1.9298, "step": 28100 }, { "epoch": 0.69, "learning_rate": 4.558654478553033e-06, "loss": 1.9043, "step": 28101 }, { "epoch": 0.69, "learning_rate": 4.557984697700123e-06, "loss": 2.0093, "step": 28102 }, { "epoch": 0.69, "learning_rate": 4.5573149515309475e-06, "loss": 2.1441, "step": 28103 }, { "epoch": 0.69, "learning_rate": 4.5566452400497805e-06, "loss": 1.9978, "step": 28104 }, { "epoch": 0.69, "learning_rate": 4.555975563260884e-06, "loss": 1.9504, "step": 28105 }, { "epoch": 0.69, "learning_rate": 4.555305921168529e-06, "loss": 1.8533, "step": 28106 }, { "epoch": 0.69, "learning_rate": 4.554636313776984e-06, "loss": 1.8775, "step": 28107 }, { "epoch": 0.69, "learning_rate": 4.553966741090517e-06, "loss": 2.0972, "step": 28108 }, { "epoch": 0.69, "learning_rate": 4.553297203113397e-06, "loss": 1.9293, "step": 28109 }, { "epoch": 0.69, "learning_rate": 4.552627699849885e-06, "loss": 1.9922, "step": 28110 }, { "epoch": 0.69, "learning_rate": 4.551958231304253e-06, "loss": 1.9803, "step": 28111 }, { "epoch": 0.69, "learning_rate": 4.551288797480771e-06, "loss": 2.1011, "step": 28112 }, { "epoch": 0.69, "learning_rate": 4.550619398383697e-06, "loss": 1.9479, "step": 28113 }, { "epoch": 0.69, "learning_rate": 4.549950034017304e-06, "loss": 2.076, "step": 28114 }, { "epoch": 0.69, "learning_rate": 4.549280704385853e-06, "loss": 1.9096, "step": 28115 }, { "epoch": 0.69, "learning_rate": 4.548611409493613e-06, "loss": 1.798, "step": 28116 }, { "epoch": 0.69, "learning_rate": 4.547942149344851e-06, "loss": 1.9274, "step": 28117 }, { "epoch": 0.69, "learning_rate": 4.547272923943828e-06, "loss": 2.076, "step": 28118 }, { "epoch": 0.69, "learning_rate": 4.5466037332948154e-06, "loss": 1.8798, "step": 28119 }, { "epoch": 0.69, "learning_rate": 4.54593457740207e-06, "loss": 1.9636, "step": 28120 }, { "epoch": 0.69, "learning_rate": 4.545265456269862e-06, "loss": 2.0493, "step": 28121 }, { "epoch": 0.69, "learning_rate": 4.544596369902457e-06, "loss": 1.9346, "step": 28122 }, { "epoch": 0.69, "learning_rate": 4.543927318304115e-06, "loss": 2.0888, "step": 28123 }, { "epoch": 0.69, "learning_rate": 4.543258301479102e-06, "loss": 1.8885, "step": 28124 }, { "epoch": 0.69, "learning_rate": 4.542589319431681e-06, "loss": 2.0342, "step": 28125 }, { "epoch": 0.69, "learning_rate": 4.541920372166118e-06, "loss": 2.0265, "step": 28126 }, { "epoch": 0.69, "learning_rate": 4.541251459686679e-06, "loss": 1.8857, "step": 28127 }, { "epoch": 0.69, "learning_rate": 4.540582581997619e-06, "loss": 1.9244, "step": 28128 }, { "epoch": 0.69, "learning_rate": 4.539913739103209e-06, "loss": 1.993, "step": 28129 }, { "epoch": 0.69, "learning_rate": 4.539244931007705e-06, "loss": 1.8985, "step": 28130 }, { "epoch": 0.69, "learning_rate": 4.538576157715372e-06, "loss": 1.9963, "step": 28131 }, { "epoch": 0.69, "learning_rate": 4.537907419230477e-06, "loss": 1.8962, "step": 28132 }, { "epoch": 0.69, "learning_rate": 4.537238715557275e-06, "loss": 1.8842, "step": 28133 }, { "epoch": 0.69, "learning_rate": 4.536570046700034e-06, "loss": 1.9078, "step": 28134 }, { "epoch": 0.69, "learning_rate": 4.535901412663009e-06, "loss": 1.9905, "step": 28135 }, { "epoch": 0.69, "learning_rate": 4.535232813450465e-06, "loss": 2.1323, "step": 28136 }, { "epoch": 0.69, "learning_rate": 4.5345642490666666e-06, "loss": 1.9992, "step": 28137 }, { "epoch": 0.69, "learning_rate": 4.533895719515869e-06, "loss": 1.9202, "step": 28138 }, { "epoch": 0.69, "learning_rate": 4.5332272248023364e-06, "loss": 1.7885, "step": 28139 }, { "epoch": 0.69, "learning_rate": 4.532558764930327e-06, "loss": 1.8546, "step": 28140 }, { "epoch": 0.69, "learning_rate": 4.531890339904101e-06, "loss": 1.9839, "step": 28141 }, { "epoch": 0.69, "learning_rate": 4.5312219497279204e-06, "loss": 1.9824, "step": 28142 }, { "epoch": 0.69, "learning_rate": 4.530553594406044e-06, "loss": 1.995, "step": 28143 }, { "epoch": 0.69, "learning_rate": 4.5298852739427365e-06, "loss": 2.0232, "step": 28144 }, { "epoch": 0.69, "learning_rate": 4.529216988342249e-06, "loss": 1.9272, "step": 28145 }, { "epoch": 0.69, "learning_rate": 4.528548737608845e-06, "loss": 2.0348, "step": 28146 }, { "epoch": 0.69, "learning_rate": 4.527880521746786e-06, "loss": 1.8401, "step": 28147 }, { "epoch": 0.69, "learning_rate": 4.527212340760323e-06, "loss": 2.012, "step": 28148 }, { "epoch": 0.69, "learning_rate": 4.5265441946537246e-06, "loss": 1.9434, "step": 28149 }, { "epoch": 0.69, "learning_rate": 4.52587608343124e-06, "loss": 1.8301, "step": 28150 }, { "epoch": 0.69, "learning_rate": 4.52520800709713e-06, "loss": 2.0778, "step": 28151 }, { "epoch": 0.69, "learning_rate": 4.524539965655657e-06, "loss": 1.9703, "step": 28152 }, { "epoch": 0.69, "learning_rate": 4.523871959111074e-06, "loss": 1.844, "step": 28153 }, { "epoch": 0.69, "learning_rate": 4.523203987467637e-06, "loss": 2.1039, "step": 28154 }, { "epoch": 0.69, "learning_rate": 4.522536050729611e-06, "loss": 1.8888, "step": 28155 }, { "epoch": 0.69, "learning_rate": 4.521868148901243e-06, "loss": 2.0383, "step": 28156 }, { "epoch": 0.69, "learning_rate": 4.5212002819868e-06, "loss": 2.0711, "step": 28157 }, { "epoch": 0.69, "learning_rate": 4.5205324499905276e-06, "loss": 1.8452, "step": 28158 }, { "epoch": 0.69, "learning_rate": 4.519864652916688e-06, "loss": 1.9233, "step": 28159 }, { "epoch": 0.69, "learning_rate": 4.519196890769537e-06, "loss": 2.0848, "step": 28160 }, { "epoch": 0.69, "learning_rate": 4.518529163553331e-06, "loss": 1.9609, "step": 28161 }, { "epoch": 0.69, "learning_rate": 4.517861471272328e-06, "loss": 1.8903, "step": 28162 }, { "epoch": 0.69, "learning_rate": 4.517193813930777e-06, "loss": 1.9252, "step": 28163 }, { "epoch": 0.69, "learning_rate": 4.516526191532936e-06, "loss": 2.121, "step": 28164 }, { "epoch": 0.69, "learning_rate": 4.5158586040830656e-06, "loss": 1.8101, "step": 28165 }, { "epoch": 0.69, "learning_rate": 4.51519105158541e-06, "loss": 1.8634, "step": 28166 }, { "epoch": 0.69, "learning_rate": 4.514523534044235e-06, "loss": 2.1024, "step": 28167 }, { "epoch": 0.69, "learning_rate": 4.5138560514637844e-06, "loss": 1.9211, "step": 28168 }, { "epoch": 0.69, "learning_rate": 4.5131886038483176e-06, "loss": 2.0078, "step": 28169 }, { "epoch": 0.69, "learning_rate": 4.512521191202091e-06, "loss": 1.8531, "step": 28170 }, { "epoch": 0.69, "learning_rate": 4.511853813529351e-06, "loss": 1.8583, "step": 28171 }, { "epoch": 0.69, "learning_rate": 4.51118647083436e-06, "loss": 2.0078, "step": 28172 }, { "epoch": 0.69, "learning_rate": 4.510519163121363e-06, "loss": 1.8944, "step": 28173 }, { "epoch": 0.69, "learning_rate": 4.509851890394615e-06, "loss": 1.8957, "step": 28174 }, { "epoch": 0.69, "learning_rate": 4.509184652658375e-06, "loss": 1.8508, "step": 28175 }, { "epoch": 0.69, "learning_rate": 4.508517449916886e-06, "loss": 1.8912, "step": 28176 }, { "epoch": 0.69, "learning_rate": 4.507850282174405e-06, "loss": 1.892, "step": 28177 }, { "epoch": 0.69, "learning_rate": 4.507183149435185e-06, "loss": 1.9782, "step": 28178 }, { "epoch": 0.69, "learning_rate": 4.506516051703476e-06, "loss": 1.8564, "step": 28179 }, { "epoch": 0.69, "learning_rate": 4.5058489889835335e-06, "loss": 2.0235, "step": 28180 }, { "epoch": 0.69, "learning_rate": 4.505181961279604e-06, "loss": 1.8625, "step": 28181 }, { "epoch": 0.69, "learning_rate": 4.504514968595942e-06, "loss": 1.8858, "step": 28182 }, { "epoch": 0.69, "learning_rate": 4.503848010936794e-06, "loss": 2.0827, "step": 28183 }, { "epoch": 0.69, "learning_rate": 4.503181088306415e-06, "loss": 1.8649, "step": 28184 }, { "epoch": 0.69, "learning_rate": 4.5025142007090576e-06, "loss": 1.9973, "step": 28185 }, { "epoch": 0.69, "learning_rate": 4.501847348148964e-06, "loss": 1.9409, "step": 28186 }, { "epoch": 0.69, "learning_rate": 4.501180530630394e-06, "loss": 2.1107, "step": 28187 }, { "epoch": 0.69, "learning_rate": 4.500513748157588e-06, "loss": 1.8044, "step": 28188 }, { "epoch": 0.69, "learning_rate": 4.4998470007348e-06, "loss": 1.9159, "step": 28189 }, { "epoch": 0.69, "learning_rate": 4.499180288366284e-06, "loss": 1.9689, "step": 28190 }, { "epoch": 0.69, "learning_rate": 4.498513611056281e-06, "loss": 1.8559, "step": 28191 }, { "epoch": 0.69, "learning_rate": 4.497846968809043e-06, "loss": 2.1776, "step": 28192 }, { "epoch": 0.69, "learning_rate": 4.497180361628824e-06, "loss": 2.025, "step": 28193 }, { "epoch": 0.69, "learning_rate": 4.496513789519863e-06, "loss": 1.859, "step": 28194 }, { "epoch": 0.69, "learning_rate": 4.495847252486414e-06, "loss": 2.0195, "step": 28195 }, { "epoch": 0.69, "learning_rate": 4.495180750532724e-06, "loss": 1.8703, "step": 28196 }, { "epoch": 0.7, "learning_rate": 4.494514283663044e-06, "loss": 1.9018, "step": 28197 }, { "epoch": 0.7, "learning_rate": 4.493847851881616e-06, "loss": 1.9272, "step": 28198 }, { "epoch": 0.7, "learning_rate": 4.493181455192689e-06, "loss": 1.8686, "step": 28199 }, { "epoch": 0.7, "learning_rate": 4.492515093600516e-06, "loss": 1.9853, "step": 28200 }, { "epoch": 0.7, "learning_rate": 4.4918487671093345e-06, "loss": 1.8092, "step": 28201 }, { "epoch": 0.7, "learning_rate": 4.4911824757234e-06, "loss": 1.8749, "step": 28202 }, { "epoch": 0.7, "learning_rate": 4.490516219446952e-06, "loss": 1.8621, "step": 28203 }, { "epoch": 0.7, "learning_rate": 4.489849998284238e-06, "loss": 1.9404, "step": 28204 }, { "epoch": 0.7, "learning_rate": 4.489183812239511e-06, "loss": 2.0434, "step": 28205 }, { "epoch": 0.7, "learning_rate": 4.488517661317007e-06, "loss": 1.8958, "step": 28206 }, { "epoch": 0.7, "learning_rate": 4.487851545520979e-06, "loss": 2.0259, "step": 28207 }, { "epoch": 0.7, "learning_rate": 4.4871854648556665e-06, "loss": 1.942, "step": 28208 }, { "epoch": 0.7, "learning_rate": 4.486519419325318e-06, "loss": 1.9763, "step": 28209 }, { "epoch": 0.7, "learning_rate": 4.4858534089341775e-06, "loss": 1.8697, "step": 28210 }, { "epoch": 0.7, "learning_rate": 4.485187433686494e-06, "loss": 1.9625, "step": 28211 }, { "epoch": 0.7, "learning_rate": 4.484521493586504e-06, "loss": 1.8839, "step": 28212 }, { "epoch": 0.7, "learning_rate": 4.483855588638457e-06, "loss": 1.9282, "step": 28213 }, { "epoch": 0.7, "learning_rate": 4.483189718846596e-06, "loss": 2.0687, "step": 28214 }, { "epoch": 0.7, "learning_rate": 4.482523884215167e-06, "loss": 1.8164, "step": 28215 }, { "epoch": 0.7, "learning_rate": 4.48185808474841e-06, "loss": 1.934, "step": 28216 }, { "epoch": 0.7, "learning_rate": 4.481192320450568e-06, "loss": 2.0391, "step": 28217 }, { "epoch": 0.7, "learning_rate": 4.48052659132589e-06, "loss": 1.8816, "step": 28218 }, { "epoch": 0.7, "learning_rate": 4.479860897378611e-06, "loss": 1.8863, "step": 28219 }, { "epoch": 0.7, "learning_rate": 4.479195238612981e-06, "loss": 1.8881, "step": 28220 }, { "epoch": 0.7, "learning_rate": 4.478529615033236e-06, "loss": 1.8836, "step": 28221 }, { "epoch": 0.7, "learning_rate": 4.47786402664362e-06, "loss": 1.8923, "step": 28222 }, { "epoch": 0.7, "learning_rate": 4.477198473448381e-06, "loss": 2.1341, "step": 28223 }, { "epoch": 0.7, "learning_rate": 4.476532955451752e-06, "loss": 1.9548, "step": 28224 }, { "epoch": 0.7, "learning_rate": 4.475867472657982e-06, "loss": 2.0129, "step": 28225 }, { "epoch": 0.7, "learning_rate": 4.475202025071305e-06, "loss": 1.9212, "step": 28226 }, { "epoch": 0.7, "learning_rate": 4.474536612695966e-06, "loss": 1.9879, "step": 28227 }, { "epoch": 0.7, "learning_rate": 4.47387123553621e-06, "loss": 2.0098, "step": 28228 }, { "epoch": 0.7, "learning_rate": 4.47320589359627e-06, "loss": 2.0006, "step": 28229 }, { "epoch": 0.7, "learning_rate": 4.472540586880389e-06, "loss": 1.985, "step": 28230 }, { "epoch": 0.7, "learning_rate": 4.471875315392809e-06, "loss": 1.9733, "step": 28231 }, { "epoch": 0.7, "learning_rate": 4.471210079137768e-06, "loss": 1.9485, "step": 28232 }, { "epoch": 0.7, "learning_rate": 4.4705448781195114e-06, "loss": 1.9062, "step": 28233 }, { "epoch": 0.7, "learning_rate": 4.469879712342271e-06, "loss": 2.0593, "step": 28234 }, { "epoch": 0.7, "learning_rate": 4.469214581810293e-06, "loss": 2.0111, "step": 28235 }, { "epoch": 0.7, "learning_rate": 4.468549486527809e-06, "loss": 1.9795, "step": 28236 }, { "epoch": 0.7, "learning_rate": 4.467884426499062e-06, "loss": 1.8848, "step": 28237 }, { "epoch": 0.7, "learning_rate": 4.467219401728293e-06, "loss": 1.8882, "step": 28238 }, { "epoch": 0.7, "learning_rate": 4.4665544122197355e-06, "loss": 1.9393, "step": 28239 }, { "epoch": 0.7, "learning_rate": 4.465889457977634e-06, "loss": 2.0191, "step": 28240 }, { "epoch": 0.7, "learning_rate": 4.4652245390062174e-06, "loss": 1.9496, "step": 28241 }, { "epoch": 0.7, "learning_rate": 4.464559655309729e-06, "loss": 1.9717, "step": 28242 }, { "epoch": 0.7, "learning_rate": 4.463894806892411e-06, "loss": 2.0702, "step": 28243 }, { "epoch": 0.7, "learning_rate": 4.46322999375849e-06, "loss": 2.1835, "step": 28244 }, { "epoch": 0.7, "learning_rate": 4.462565215912209e-06, "loss": 1.8808, "step": 28245 }, { "epoch": 0.7, "learning_rate": 4.461900473357809e-06, "loss": 1.9611, "step": 28246 }, { "epoch": 0.7, "learning_rate": 4.461235766099516e-06, "loss": 1.988, "step": 28247 }, { "epoch": 0.7, "learning_rate": 4.460571094141576e-06, "loss": 1.8966, "step": 28248 }, { "epoch": 0.7, "learning_rate": 4.459906457488219e-06, "loss": 1.7612, "step": 28249 }, { "epoch": 0.7, "learning_rate": 4.459241856143689e-06, "loss": 2.1431, "step": 28250 }, { "epoch": 0.7, "learning_rate": 4.458577290112212e-06, "loss": 2.0162, "step": 28251 }, { "epoch": 0.7, "learning_rate": 4.457912759398028e-06, "loss": 1.9788, "step": 28252 }, { "epoch": 0.7, "learning_rate": 4.457248264005375e-06, "loss": 1.8481, "step": 28253 }, { "epoch": 0.7, "learning_rate": 4.4565838039384825e-06, "loss": 1.8338, "step": 28254 }, { "epoch": 0.7, "learning_rate": 4.4559193792015905e-06, "loss": 2.0449, "step": 28255 }, { "epoch": 0.7, "learning_rate": 4.455254989798929e-06, "loss": 1.8326, "step": 28256 }, { "epoch": 0.7, "learning_rate": 4.454590635734732e-06, "loss": 2.0102, "step": 28257 }, { "epoch": 0.7, "learning_rate": 4.453926317013242e-06, "loss": 2.0473, "step": 28258 }, { "epoch": 0.7, "learning_rate": 4.453262033638682e-06, "loss": 2.049, "step": 28259 }, { "epoch": 0.7, "learning_rate": 4.452597785615294e-06, "loss": 1.9784, "step": 28260 }, { "epoch": 0.7, "learning_rate": 4.451933572947307e-06, "loss": 1.762, "step": 28261 }, { "epoch": 0.7, "learning_rate": 4.451269395638953e-06, "loss": 1.9492, "step": 28262 }, { "epoch": 0.7, "learning_rate": 4.4506052536944684e-06, "loss": 1.965, "step": 28263 }, { "epoch": 0.7, "learning_rate": 4.449941147118089e-06, "loss": 1.9029, "step": 28264 }, { "epoch": 0.7, "learning_rate": 4.449277075914039e-06, "loss": 1.9185, "step": 28265 }, { "epoch": 0.7, "learning_rate": 4.448613040086555e-06, "loss": 2.0119, "step": 28266 }, { "epoch": 0.7, "learning_rate": 4.44794903963987e-06, "loss": 2.1151, "step": 28267 }, { "epoch": 0.7, "learning_rate": 4.447285074578218e-06, "loss": 2.0219, "step": 28268 }, { "epoch": 0.7, "learning_rate": 4.446621144905825e-06, "loss": 1.9859, "step": 28269 }, { "epoch": 0.7, "learning_rate": 4.445957250626928e-06, "loss": 1.9063, "step": 28270 }, { "epoch": 0.7, "learning_rate": 4.445293391745752e-06, "loss": 1.8475, "step": 28271 }, { "epoch": 0.7, "learning_rate": 4.444629568266532e-06, "loss": 1.9713, "step": 28272 }, { "epoch": 0.7, "learning_rate": 4.443965780193501e-06, "loss": 1.9297, "step": 28273 }, { "epoch": 0.7, "learning_rate": 4.443302027530884e-06, "loss": 1.8739, "step": 28274 }, { "epoch": 0.7, "learning_rate": 4.442638310282913e-06, "loss": 2.0519, "step": 28275 }, { "epoch": 0.7, "learning_rate": 4.441974628453823e-06, "loss": 1.7235, "step": 28276 }, { "epoch": 0.7, "learning_rate": 4.441310982047837e-06, "loss": 2.0134, "step": 28277 }, { "epoch": 0.7, "learning_rate": 4.440647371069192e-06, "loss": 2.0721, "step": 28278 }, { "epoch": 0.7, "learning_rate": 4.439983795522108e-06, "loss": 1.9117, "step": 28279 }, { "epoch": 0.7, "learning_rate": 4.43932025541082e-06, "loss": 1.8231, "step": 28280 }, { "epoch": 0.7, "learning_rate": 4.438656750739556e-06, "loss": 2.0994, "step": 28281 }, { "epoch": 0.7, "learning_rate": 4.437993281512548e-06, "loss": 2.07, "step": 28282 }, { "epoch": 0.7, "learning_rate": 4.437329847734019e-06, "loss": 2.0565, "step": 28283 }, { "epoch": 0.7, "learning_rate": 4.436666449408198e-06, "loss": 1.9271, "step": 28284 }, { "epoch": 0.7, "learning_rate": 4.436003086539315e-06, "loss": 1.9281, "step": 28285 }, { "epoch": 0.7, "learning_rate": 4.435339759131603e-06, "loss": 1.871, "step": 28286 }, { "epoch": 0.7, "learning_rate": 4.434676467189279e-06, "loss": 1.8497, "step": 28287 }, { "epoch": 0.7, "learning_rate": 4.43401321071658e-06, "loss": 2.0402, "step": 28288 }, { "epoch": 0.7, "learning_rate": 4.433349989717724e-06, "loss": 1.9149, "step": 28289 }, { "epoch": 0.7, "learning_rate": 4.4326868041969435e-06, "loss": 2.0549, "step": 28290 }, { "epoch": 0.7, "learning_rate": 4.4320236541584686e-06, "loss": 1.8535, "step": 28291 }, { "epoch": 0.7, "learning_rate": 4.431360539606516e-06, "loss": 1.7396, "step": 28292 }, { "epoch": 0.7, "learning_rate": 4.4306974605453235e-06, "loss": 1.9127, "step": 28293 }, { "epoch": 0.7, "learning_rate": 4.430034416979105e-06, "loss": 2.0162, "step": 28294 }, { "epoch": 0.7, "learning_rate": 4.429371408912094e-06, "loss": 1.8362, "step": 28295 }, { "epoch": 0.7, "learning_rate": 4.428708436348518e-06, "loss": 1.9479, "step": 28296 }, { "epoch": 0.7, "learning_rate": 4.428045499292595e-06, "loss": 1.8946, "step": 28297 }, { "epoch": 0.7, "learning_rate": 4.427382597748553e-06, "loss": 1.6951, "step": 28298 }, { "epoch": 0.7, "learning_rate": 4.4267197317206175e-06, "loss": 1.8349, "step": 28299 }, { "epoch": 0.7, "learning_rate": 4.426056901213019e-06, "loss": 1.8827, "step": 28300 }, { "epoch": 0.7, "learning_rate": 4.425394106229971e-06, "loss": 2.0821, "step": 28301 }, { "epoch": 0.7, "learning_rate": 4.424731346775704e-06, "loss": 2.0585, "step": 28302 }, { "epoch": 0.7, "learning_rate": 4.424068622854444e-06, "loss": 1.9321, "step": 28303 }, { "epoch": 0.7, "learning_rate": 4.423405934470409e-06, "loss": 1.813, "step": 28304 }, { "epoch": 0.7, "learning_rate": 4.422743281627825e-06, "loss": 1.9166, "step": 28305 }, { "epoch": 0.7, "learning_rate": 4.4220806643309185e-06, "loss": 1.8363, "step": 28306 }, { "epoch": 0.7, "learning_rate": 4.421418082583907e-06, "loss": 1.7806, "step": 28307 }, { "epoch": 0.7, "learning_rate": 4.420755536391019e-06, "loss": 1.9533, "step": 28308 }, { "epoch": 0.7, "learning_rate": 4.420093025756471e-06, "loss": 2.0038, "step": 28309 }, { "epoch": 0.7, "learning_rate": 4.4194305506844895e-06, "loss": 1.8625, "step": 28310 }, { "epoch": 0.7, "learning_rate": 4.418768111179298e-06, "loss": 1.9505, "step": 28311 }, { "epoch": 0.7, "learning_rate": 4.4181057072451124e-06, "loss": 1.9917, "step": 28312 }, { "epoch": 0.7, "learning_rate": 4.417443338886162e-06, "loss": 1.8426, "step": 28313 }, { "epoch": 0.7, "learning_rate": 4.4167810061066615e-06, "loss": 2.0731, "step": 28314 }, { "epoch": 0.7, "learning_rate": 4.416118708910836e-06, "loss": 1.9201, "step": 28315 }, { "epoch": 0.7, "learning_rate": 4.4154564473029045e-06, "loss": 1.9702, "step": 28316 }, { "epoch": 0.7, "learning_rate": 4.414794221287089e-06, "loss": 2.0801, "step": 28317 }, { "epoch": 0.7, "learning_rate": 4.4141320308676135e-06, "loss": 2.0242, "step": 28318 }, { "epoch": 0.7, "learning_rate": 4.413469876048692e-06, "loss": 1.9606, "step": 28319 }, { "epoch": 0.7, "learning_rate": 4.412807756834547e-06, "loss": 1.9228, "step": 28320 }, { "epoch": 0.7, "learning_rate": 4.412145673229403e-06, "loss": 2.0951, "step": 28321 }, { "epoch": 0.7, "learning_rate": 4.411483625237473e-06, "loss": 2.054, "step": 28322 }, { "epoch": 0.7, "learning_rate": 4.410821612862981e-06, "loss": 1.9787, "step": 28323 }, { "epoch": 0.7, "learning_rate": 4.410159636110142e-06, "loss": 2.1042, "step": 28324 }, { "epoch": 0.7, "learning_rate": 4.409497694983177e-06, "loss": 1.9991, "step": 28325 }, { "epoch": 0.7, "learning_rate": 4.408835789486308e-06, "loss": 1.9338, "step": 28326 }, { "epoch": 0.7, "learning_rate": 4.408173919623747e-06, "loss": 1.9122, "step": 28327 }, { "epoch": 0.7, "learning_rate": 4.40751208539972e-06, "loss": 1.8605, "step": 28328 }, { "epoch": 0.7, "learning_rate": 4.406850286818438e-06, "loss": 1.9295, "step": 28329 }, { "epoch": 0.7, "learning_rate": 4.4061885238841205e-06, "loss": 1.8976, "step": 28330 }, { "epoch": 0.7, "learning_rate": 4.405526796600992e-06, "loss": 1.909, "step": 28331 }, { "epoch": 0.7, "learning_rate": 4.404865104973259e-06, "loss": 1.9884, "step": 28332 }, { "epoch": 0.7, "learning_rate": 4.404203449005144e-06, "loss": 2.0083, "step": 28333 }, { "epoch": 0.7, "learning_rate": 4.4035418287008655e-06, "loss": 1.9375, "step": 28334 }, { "epoch": 0.7, "learning_rate": 4.402880244064643e-06, "loss": 1.8754, "step": 28335 }, { "epoch": 0.7, "learning_rate": 4.402218695100683e-06, "loss": 1.8376, "step": 28336 }, { "epoch": 0.7, "learning_rate": 4.401557181813208e-06, "loss": 1.8527, "step": 28337 }, { "epoch": 0.7, "learning_rate": 4.400895704206434e-06, "loss": 2.001, "step": 28338 }, { "epoch": 0.7, "learning_rate": 4.400234262284579e-06, "loss": 1.994, "step": 28339 }, { "epoch": 0.7, "learning_rate": 4.399572856051854e-06, "loss": 1.9601, "step": 28340 }, { "epoch": 0.7, "learning_rate": 4.398911485512479e-06, "loss": 2.0746, "step": 28341 }, { "epoch": 0.7, "learning_rate": 4.3982501506706634e-06, "loss": 2.1445, "step": 28342 }, { "epoch": 0.7, "learning_rate": 4.397588851530624e-06, "loss": 1.9802, "step": 28343 }, { "epoch": 0.7, "learning_rate": 4.396927588096581e-06, "loss": 1.8622, "step": 28344 }, { "epoch": 0.7, "learning_rate": 4.396266360372742e-06, "loss": 1.896, "step": 28345 }, { "epoch": 0.7, "learning_rate": 4.395605168363326e-06, "loss": 2.0441, "step": 28346 }, { "epoch": 0.7, "learning_rate": 4.394944012072541e-06, "loss": 2.1168, "step": 28347 }, { "epoch": 0.7, "learning_rate": 4.394282891504604e-06, "loss": 1.9143, "step": 28348 }, { "epoch": 0.7, "learning_rate": 4.393621806663734e-06, "loss": 2.0664, "step": 28349 }, { "epoch": 0.7, "learning_rate": 4.392960757554135e-06, "loss": 1.9724, "step": 28350 }, { "epoch": 0.7, "learning_rate": 4.392299744180024e-06, "loss": 2.0971, "step": 28351 }, { "epoch": 0.7, "learning_rate": 4.3916387665456136e-06, "loss": 2.0244, "step": 28352 }, { "epoch": 0.7, "learning_rate": 4.390977824655122e-06, "loss": 1.8549, "step": 28353 }, { "epoch": 0.7, "learning_rate": 4.390316918512753e-06, "loss": 1.9642, "step": 28354 }, { "epoch": 0.7, "learning_rate": 4.389656048122723e-06, "loss": 1.9873, "step": 28355 }, { "epoch": 0.7, "learning_rate": 4.3889952134892455e-06, "loss": 1.9124, "step": 28356 }, { "epoch": 0.7, "learning_rate": 4.388334414616526e-06, "loss": 1.9555, "step": 28357 }, { "epoch": 0.7, "learning_rate": 4.387673651508781e-06, "loss": 1.9768, "step": 28358 }, { "epoch": 0.7, "learning_rate": 4.387012924170225e-06, "loss": 2.0384, "step": 28359 }, { "epoch": 0.7, "learning_rate": 4.3863522326050615e-06, "loss": 1.9728, "step": 28360 }, { "epoch": 0.7, "learning_rate": 4.385691576817506e-06, "loss": 1.9508, "step": 28361 }, { "epoch": 0.7, "learning_rate": 4.385030956811766e-06, "loss": 1.8684, "step": 28362 }, { "epoch": 0.7, "learning_rate": 4.384370372592054e-06, "loss": 1.8942, "step": 28363 }, { "epoch": 0.7, "learning_rate": 4.383709824162582e-06, "loss": 1.9635, "step": 28364 }, { "epoch": 0.7, "learning_rate": 4.383049311527555e-06, "loss": 1.9357, "step": 28365 }, { "epoch": 0.7, "learning_rate": 4.382388834691188e-06, "loss": 1.8411, "step": 28366 }, { "epoch": 0.7, "learning_rate": 4.3817283936576845e-06, "loss": 1.8487, "step": 28367 }, { "epoch": 0.7, "learning_rate": 4.3810679884312564e-06, "loss": 1.9191, "step": 28368 }, { "epoch": 0.7, "learning_rate": 4.380407619016114e-06, "loss": 1.9338, "step": 28369 }, { "epoch": 0.7, "learning_rate": 4.379747285416465e-06, "loss": 1.8675, "step": 28370 }, { "epoch": 0.7, "learning_rate": 4.379086987636522e-06, "loss": 1.9831, "step": 28371 }, { "epoch": 0.7, "learning_rate": 4.378426725680485e-06, "loss": 2.0441, "step": 28372 }, { "epoch": 0.7, "learning_rate": 4.377766499552566e-06, "loss": 2.065, "step": 28373 }, { "epoch": 0.7, "learning_rate": 4.377106309256979e-06, "loss": 2.0996, "step": 28374 }, { "epoch": 0.7, "learning_rate": 4.37644615479792e-06, "loss": 1.9051, "step": 28375 }, { "epoch": 0.7, "learning_rate": 4.375786036179608e-06, "loss": 1.8112, "step": 28376 }, { "epoch": 0.7, "learning_rate": 4.37512595340624e-06, "loss": 1.9746, "step": 28377 }, { "epoch": 0.7, "learning_rate": 4.374465906482026e-06, "loss": 1.7821, "step": 28378 }, { "epoch": 0.7, "learning_rate": 4.37380589541118e-06, "loss": 2.0415, "step": 28379 }, { "epoch": 0.7, "learning_rate": 4.373145920197898e-06, "loss": 1.8953, "step": 28380 }, { "epoch": 0.7, "learning_rate": 4.372485980846394e-06, "loss": 2.0633, "step": 28381 }, { "epoch": 0.7, "learning_rate": 4.371826077360868e-06, "loss": 1.9024, "step": 28382 }, { "epoch": 0.7, "learning_rate": 4.371166209745527e-06, "loss": 1.9587, "step": 28383 }, { "epoch": 0.7, "learning_rate": 4.370506378004583e-06, "loss": 2.2432, "step": 28384 }, { "epoch": 0.7, "learning_rate": 4.369846582142233e-06, "loss": 1.9451, "step": 28385 }, { "epoch": 0.7, "learning_rate": 4.369186822162686e-06, "loss": 1.9372, "step": 28386 }, { "epoch": 0.7, "learning_rate": 4.368527098070144e-06, "loss": 2.0684, "step": 28387 }, { "epoch": 0.7, "learning_rate": 4.367867409868817e-06, "loss": 2.0435, "step": 28388 }, { "epoch": 0.7, "learning_rate": 4.367207757562908e-06, "loss": 2.0471, "step": 28389 }, { "epoch": 0.7, "learning_rate": 4.366548141156617e-06, "loss": 1.9449, "step": 28390 }, { "epoch": 0.7, "learning_rate": 4.365888560654154e-06, "loss": 1.952, "step": 28391 }, { "epoch": 0.7, "learning_rate": 4.365229016059715e-06, "loss": 1.8356, "step": 28392 }, { "epoch": 0.7, "learning_rate": 4.36456950737751e-06, "loss": 1.8685, "step": 28393 }, { "epoch": 0.7, "learning_rate": 4.363910034611741e-06, "loss": 1.7174, "step": 28394 }, { "epoch": 0.7, "learning_rate": 4.3632505977666085e-06, "loss": 1.8188, "step": 28395 }, { "epoch": 0.7, "learning_rate": 4.362591196846315e-06, "loss": 1.866, "step": 28396 }, { "epoch": 0.7, "learning_rate": 4.361931831855071e-06, "loss": 1.9672, "step": 28397 }, { "epoch": 0.7, "learning_rate": 4.361272502797067e-06, "loss": 1.9746, "step": 28398 }, { "epoch": 0.7, "learning_rate": 4.360613209676515e-06, "loss": 1.9063, "step": 28399 }, { "epoch": 0.7, "eval_loss": 1.6867138147354126, "eval_runtime": 94.6676, "eval_samples_per_second": 621.617, "eval_steps_per_second": 4.859, "step": 28399 }, { "epoch": 0.7, "learning_rate": 4.359953952497609e-06, "loss": 1.8793, "step": 28400 }, { "epoch": 0.7, "learning_rate": 4.359294731264556e-06, "loss": 1.9839, "step": 28401 }, { "epoch": 0.7, "learning_rate": 4.3586355459815585e-06, "loss": 1.8552, "step": 28402 }, { "epoch": 0.7, "learning_rate": 4.3579763966528125e-06, "loss": 2.0458, "step": 28403 }, { "epoch": 0.7, "learning_rate": 4.357317283282521e-06, "loss": 1.9739, "step": 28404 }, { "epoch": 0.7, "learning_rate": 4.356658205874885e-06, "loss": 1.9182, "step": 28405 }, { "epoch": 0.7, "learning_rate": 4.355999164434105e-06, "loss": 1.8845, "step": 28406 }, { "epoch": 0.7, "learning_rate": 4.355340158964387e-06, "loss": 1.9903, "step": 28407 }, { "epoch": 0.7, "learning_rate": 4.354681189469921e-06, "loss": 2.0191, "step": 28408 }, { "epoch": 0.7, "learning_rate": 4.354022255954914e-06, "loss": 2.1438, "step": 28409 }, { "epoch": 0.7, "learning_rate": 4.3533633584235605e-06, "loss": 1.8847, "step": 28410 }, { "epoch": 0.7, "learning_rate": 4.352704496880062e-06, "loss": 1.8951, "step": 28411 }, { "epoch": 0.7, "learning_rate": 4.352045671328623e-06, "loss": 1.8857, "step": 28412 }, { "epoch": 0.7, "learning_rate": 4.351386881773433e-06, "loss": 2.1439, "step": 28413 }, { "epoch": 0.7, "learning_rate": 4.3507281282187e-06, "loss": 1.8704, "step": 28414 }, { "epoch": 0.7, "learning_rate": 4.350069410668613e-06, "loss": 1.9413, "step": 28415 }, { "epoch": 0.7, "learning_rate": 4.349410729127375e-06, "loss": 1.8173, "step": 28416 }, { "epoch": 0.7, "learning_rate": 4.348752083599187e-06, "loss": 2.0643, "step": 28417 }, { "epoch": 0.7, "learning_rate": 4.348093474088241e-06, "loss": 2.1161, "step": 28418 }, { "epoch": 0.7, "learning_rate": 4.3474349005987405e-06, "loss": 1.8803, "step": 28419 }, { "epoch": 0.7, "learning_rate": 4.346776363134876e-06, "loss": 2.0966, "step": 28420 }, { "epoch": 0.7, "learning_rate": 4.346117861700847e-06, "loss": 1.9512, "step": 28421 }, { "epoch": 0.7, "learning_rate": 4.345459396300853e-06, "loss": 1.8624, "step": 28422 }, { "epoch": 0.7, "learning_rate": 4.3448009669390874e-06, "loss": 1.894, "step": 28423 }, { "epoch": 0.7, "learning_rate": 4.344142573619753e-06, "loss": 1.9021, "step": 28424 }, { "epoch": 0.7, "learning_rate": 4.343484216347038e-06, "loss": 1.8693, "step": 28425 }, { "epoch": 0.7, "learning_rate": 4.34282589512514e-06, "loss": 2.031, "step": 28426 }, { "epoch": 0.7, "learning_rate": 4.342167609958261e-06, "loss": 1.8536, "step": 28427 }, { "epoch": 0.7, "learning_rate": 4.341509360850588e-06, "loss": 2.2133, "step": 28428 }, { "epoch": 0.7, "learning_rate": 4.340851147806323e-06, "loss": 1.8563, "step": 28429 }, { "epoch": 0.7, "learning_rate": 4.3401929708296545e-06, "loss": 1.9095, "step": 28430 }, { "epoch": 0.7, "learning_rate": 4.339534829924781e-06, "loss": 1.9245, "step": 28431 }, { "epoch": 0.7, "learning_rate": 4.3388767250959e-06, "loss": 1.9642, "step": 28432 }, { "epoch": 0.7, "learning_rate": 4.338218656347198e-06, "loss": 2.0359, "step": 28433 }, { "epoch": 0.7, "learning_rate": 4.3375606236828786e-06, "loss": 1.8528, "step": 28434 }, { "epoch": 0.7, "learning_rate": 4.336902627107127e-06, "loss": 2.2041, "step": 28435 }, { "epoch": 0.7, "learning_rate": 4.33624466662414e-06, "loss": 2.0982, "step": 28436 }, { "epoch": 0.7, "learning_rate": 4.3355867422381145e-06, "loss": 1.9308, "step": 28437 }, { "epoch": 0.7, "learning_rate": 4.334928853953238e-06, "loss": 1.9626, "step": 28438 }, { "epoch": 0.7, "learning_rate": 4.334271001773706e-06, "loss": 1.9492, "step": 28439 }, { "epoch": 0.7, "learning_rate": 4.333613185703711e-06, "loss": 1.7467, "step": 28440 }, { "epoch": 0.7, "learning_rate": 4.332955405747446e-06, "loss": 2.1108, "step": 28441 }, { "epoch": 0.7, "learning_rate": 4.332297661909108e-06, "loss": 1.9804, "step": 28442 }, { "epoch": 0.7, "learning_rate": 4.331639954192878e-06, "loss": 1.9308, "step": 28443 }, { "epoch": 0.7, "learning_rate": 4.330982282602958e-06, "loss": 1.81, "step": 28444 }, { "epoch": 0.7, "learning_rate": 4.330324647143532e-06, "loss": 1.9301, "step": 28445 }, { "epoch": 0.7, "learning_rate": 4.329667047818795e-06, "loss": 2.1736, "step": 28446 }, { "epoch": 0.7, "learning_rate": 4.3290094846329425e-06, "loss": 1.8403, "step": 28447 }, { "epoch": 0.7, "learning_rate": 4.328351957590156e-06, "loss": 1.9433, "step": 28448 }, { "epoch": 0.7, "learning_rate": 4.327694466694634e-06, "loss": 2.0674, "step": 28449 }, { "epoch": 0.7, "learning_rate": 4.3270370119505615e-06, "loss": 1.9061, "step": 28450 }, { "epoch": 0.7, "learning_rate": 4.326379593362129e-06, "loss": 1.8932, "step": 28451 }, { "epoch": 0.7, "learning_rate": 4.325722210933533e-06, "loss": 1.9257, "step": 28452 }, { "epoch": 0.7, "learning_rate": 4.325064864668955e-06, "loss": 1.9254, "step": 28453 }, { "epoch": 0.7, "learning_rate": 4.324407554572587e-06, "loss": 1.8512, "step": 28454 }, { "epoch": 0.7, "learning_rate": 4.323750280648624e-06, "loss": 1.877, "step": 28455 }, { "epoch": 0.7, "learning_rate": 4.3230930429012466e-06, "loss": 1.863, "step": 28456 }, { "epoch": 0.7, "learning_rate": 4.322435841334647e-06, "loss": 1.92, "step": 28457 }, { "epoch": 0.7, "learning_rate": 4.321778675953014e-06, "loss": 1.8541, "step": 28458 }, { "epoch": 0.7, "learning_rate": 4.321121546760536e-06, "loss": 1.8264, "step": 28459 }, { "epoch": 0.7, "learning_rate": 4.320464453761405e-06, "loss": 1.9434, "step": 28460 }, { "epoch": 0.7, "learning_rate": 4.319807396959801e-06, "loss": 2.0741, "step": 28461 }, { "epoch": 0.7, "learning_rate": 4.31915037635992e-06, "loss": 2.0011, "step": 28462 }, { "epoch": 0.7, "learning_rate": 4.31849339196594e-06, "loss": 1.97, "step": 28463 }, { "epoch": 0.7, "learning_rate": 4.317836443782054e-06, "loss": 1.9143, "step": 28464 }, { "epoch": 0.7, "learning_rate": 4.317179531812452e-06, "loss": 1.7462, "step": 28465 }, { "epoch": 0.7, "learning_rate": 4.316522656061313e-06, "loss": 2.0206, "step": 28466 }, { "epoch": 0.7, "learning_rate": 4.3158658165328304e-06, "loss": 2.1409, "step": 28467 }, { "epoch": 0.7, "learning_rate": 4.315209013231184e-06, "loss": 1.8544, "step": 28468 }, { "epoch": 0.7, "learning_rate": 4.3145522461605635e-06, "loss": 1.7171, "step": 28469 }, { "epoch": 0.7, "learning_rate": 4.3138955153251576e-06, "loss": 2.0546, "step": 28470 }, { "epoch": 0.7, "learning_rate": 4.313238820729146e-06, "loss": 2.0116, "step": 28471 }, { "epoch": 0.7, "learning_rate": 4.312582162376718e-06, "loss": 1.8981, "step": 28472 }, { "epoch": 0.7, "learning_rate": 4.311925540272055e-06, "loss": 2.1272, "step": 28473 }, { "epoch": 0.7, "learning_rate": 4.311268954419345e-06, "loss": 1.9515, "step": 28474 }, { "epoch": 0.7, "learning_rate": 4.310612404822772e-06, "loss": 1.9942, "step": 28475 }, { "epoch": 0.7, "learning_rate": 4.309955891486518e-06, "loss": 1.9679, "step": 28476 }, { "epoch": 0.7, "learning_rate": 4.3092994144147744e-06, "loss": 2.0763, "step": 28477 }, { "epoch": 0.7, "learning_rate": 4.3086429736117175e-06, "loss": 1.9228, "step": 28478 }, { "epoch": 0.7, "learning_rate": 4.307986569081533e-06, "loss": 1.8751, "step": 28479 }, { "epoch": 0.7, "learning_rate": 4.307330200828408e-06, "loss": 1.8936, "step": 28480 }, { "epoch": 0.7, "learning_rate": 4.30667386885652e-06, "loss": 2.0504, "step": 28481 }, { "epoch": 0.7, "learning_rate": 4.306017573170058e-06, "loss": 1.8277, "step": 28482 }, { "epoch": 0.7, "learning_rate": 4.305361313773199e-06, "loss": 1.9361, "step": 28483 }, { "epoch": 0.7, "learning_rate": 4.304705090670127e-06, "loss": 2.0149, "step": 28484 }, { "epoch": 0.7, "learning_rate": 4.304048903865029e-06, "loss": 1.9913, "step": 28485 }, { "epoch": 0.7, "learning_rate": 4.303392753362081e-06, "loss": 2.0743, "step": 28486 }, { "epoch": 0.7, "learning_rate": 4.30273663916547e-06, "loss": 2.0439, "step": 28487 }, { "epoch": 0.7, "learning_rate": 4.302080561279373e-06, "loss": 1.8419, "step": 28488 }, { "epoch": 0.7, "learning_rate": 4.301424519707972e-06, "loss": 2.1802, "step": 28489 }, { "epoch": 0.7, "learning_rate": 4.300768514455454e-06, "loss": 1.8512, "step": 28490 }, { "epoch": 0.7, "learning_rate": 4.300112545525992e-06, "loss": 2.042, "step": 28491 }, { "epoch": 0.7, "learning_rate": 4.2994566129237706e-06, "loss": 1.9232, "step": 28492 }, { "epoch": 0.7, "learning_rate": 4.298800716652969e-06, "loss": 2.012, "step": 28493 }, { "epoch": 0.7, "learning_rate": 4.298144856717769e-06, "loss": 2.0474, "step": 28494 }, { "epoch": 0.7, "learning_rate": 4.2974890331223525e-06, "loss": 1.9842, "step": 28495 }, { "epoch": 0.7, "learning_rate": 4.296833245870894e-06, "loss": 1.9213, "step": 28496 }, { "epoch": 0.7, "learning_rate": 4.29617749496758e-06, "loss": 1.9718, "step": 28497 }, { "epoch": 0.7, "learning_rate": 4.295521780416581e-06, "loss": 1.973, "step": 28498 }, { "epoch": 0.7, "learning_rate": 4.294866102222081e-06, "loss": 1.9803, "step": 28499 }, { "epoch": 0.7, "learning_rate": 4.294210460388263e-06, "loss": 2.1288, "step": 28500 }, { "epoch": 0.7, "learning_rate": 4.293554854919297e-06, "loss": 2.0802, "step": 28501 }, { "epoch": 0.7, "learning_rate": 4.2928992858193696e-06, "loss": 2.0197, "step": 28502 }, { "epoch": 0.7, "learning_rate": 4.292243753092651e-06, "loss": 1.931, "step": 28503 }, { "epoch": 0.7, "learning_rate": 4.291588256743323e-06, "loss": 2.0408, "step": 28504 }, { "epoch": 0.7, "learning_rate": 4.290932796775568e-06, "loss": 2.0145, "step": 28505 }, { "epoch": 0.7, "learning_rate": 4.290277373193553e-06, "loss": 2.0874, "step": 28506 }, { "epoch": 0.7, "learning_rate": 4.2896219860014665e-06, "loss": 2.0359, "step": 28507 }, { "epoch": 0.7, "learning_rate": 4.288966635203475e-06, "loss": 1.9682, "step": 28508 }, { "epoch": 0.7, "learning_rate": 4.288311320803762e-06, "loss": 1.9271, "step": 28509 }, { "epoch": 0.7, "learning_rate": 4.287656042806502e-06, "loss": 1.8099, "step": 28510 }, { "epoch": 0.7, "learning_rate": 4.28700080121587e-06, "loss": 2.0227, "step": 28511 }, { "epoch": 0.7, "learning_rate": 4.286345596036045e-06, "loss": 1.9761, "step": 28512 }, { "epoch": 0.7, "learning_rate": 4.285690427271205e-06, "loss": 1.9273, "step": 28513 }, { "epoch": 0.7, "learning_rate": 4.285035294925519e-06, "loss": 1.8579, "step": 28514 }, { "epoch": 0.7, "learning_rate": 4.284380199003169e-06, "loss": 1.8901, "step": 28515 }, { "epoch": 0.7, "learning_rate": 4.283725139508322e-06, "loss": 1.9606, "step": 28516 }, { "epoch": 0.7, "learning_rate": 4.283070116445158e-06, "loss": 1.7593, "step": 28517 }, { "epoch": 0.7, "learning_rate": 4.282415129817855e-06, "loss": 1.7789, "step": 28518 }, { "epoch": 0.7, "learning_rate": 4.281760179630581e-06, "loss": 2.0147, "step": 28519 }, { "epoch": 0.7, "learning_rate": 4.2811052658875155e-06, "loss": 1.9036, "step": 28520 }, { "epoch": 0.7, "learning_rate": 4.2804503885928265e-06, "loss": 1.8857, "step": 28521 }, { "epoch": 0.7, "learning_rate": 4.27979554775069e-06, "loss": 1.8679, "step": 28522 }, { "epoch": 0.7, "learning_rate": 4.279140743365287e-06, "loss": 2.0866, "step": 28523 }, { "epoch": 0.7, "learning_rate": 4.278485975440778e-06, "loss": 1.8027, "step": 28524 }, { "epoch": 0.7, "learning_rate": 4.277831243981348e-06, "loss": 2.0448, "step": 28525 }, { "epoch": 0.7, "learning_rate": 4.27717654899116e-06, "loss": 1.8651, "step": 28526 }, { "epoch": 0.7, "learning_rate": 4.27652189047439e-06, "loss": 2.0362, "step": 28527 }, { "epoch": 0.7, "learning_rate": 4.275867268435212e-06, "loss": 2.0923, "step": 28528 }, { "epoch": 0.7, "learning_rate": 4.275212682877797e-06, "loss": 1.7997, "step": 28529 }, { "epoch": 0.7, "learning_rate": 4.274558133806321e-06, "loss": 1.7236, "step": 28530 }, { "epoch": 0.7, "learning_rate": 4.273903621224948e-06, "loss": 2.0168, "step": 28531 }, { "epoch": 0.7, "learning_rate": 4.2732491451378545e-06, "loss": 1.8938, "step": 28532 }, { "epoch": 0.7, "learning_rate": 4.272594705549214e-06, "loss": 1.9164, "step": 28533 }, { "epoch": 0.7, "learning_rate": 4.2719403024631885e-06, "loss": 2.0393, "step": 28534 }, { "epoch": 0.7, "learning_rate": 4.27128593588396e-06, "loss": 1.9446, "step": 28535 }, { "epoch": 0.7, "learning_rate": 4.270631605815688e-06, "loss": 2.0433, "step": 28536 }, { "epoch": 0.7, "learning_rate": 4.2699773122625485e-06, "loss": 2.1065, "step": 28537 }, { "epoch": 0.7, "learning_rate": 4.269323055228714e-06, "loss": 2.0718, "step": 28538 }, { "epoch": 0.7, "learning_rate": 4.268668834718349e-06, "loss": 1.9626, "step": 28539 }, { "epoch": 0.7, "learning_rate": 4.268014650735628e-06, "loss": 1.8706, "step": 28540 }, { "epoch": 0.7, "learning_rate": 4.267360503284714e-06, "loss": 1.9463, "step": 28541 }, { "epoch": 0.7, "learning_rate": 4.2667063923697795e-06, "loss": 1.9406, "step": 28542 }, { "epoch": 0.7, "learning_rate": 4.266052317994997e-06, "loss": 2.0243, "step": 28543 }, { "epoch": 0.7, "learning_rate": 4.265398280164529e-06, "loss": 2.2807, "step": 28544 }, { "epoch": 0.7, "learning_rate": 4.264744278882546e-06, "loss": 2.0815, "step": 28545 }, { "epoch": 0.7, "learning_rate": 4.264090314153218e-06, "loss": 2.0307, "step": 28546 }, { "epoch": 0.7, "learning_rate": 4.263436385980709e-06, "loss": 1.9877, "step": 28547 }, { "epoch": 0.7, "learning_rate": 4.262782494369196e-06, "loss": 2.0386, "step": 28548 }, { "epoch": 0.7, "learning_rate": 4.262128639322834e-06, "loss": 1.9187, "step": 28549 }, { "epoch": 0.7, "learning_rate": 4.261474820845801e-06, "loss": 2.0129, "step": 28550 }, { "epoch": 0.7, "learning_rate": 4.260821038942256e-06, "loss": 1.9485, "step": 28551 }, { "epoch": 0.7, "learning_rate": 4.260167293616367e-06, "loss": 1.807, "step": 28552 }, { "epoch": 0.7, "learning_rate": 4.2595135848723065e-06, "loss": 1.8507, "step": 28553 }, { "epoch": 0.7, "learning_rate": 4.2588599127142345e-06, "loss": 1.8795, "step": 28554 }, { "epoch": 0.7, "learning_rate": 4.2582062771463215e-06, "loss": 1.9088, "step": 28555 }, { "epoch": 0.7, "learning_rate": 4.257552678172728e-06, "loss": 1.9054, "step": 28556 }, { "epoch": 0.7, "learning_rate": 4.256899115797622e-06, "loss": 1.7803, "step": 28557 }, { "epoch": 0.7, "learning_rate": 4.256245590025175e-06, "loss": 1.9182, "step": 28558 }, { "epoch": 0.7, "learning_rate": 4.255592100859541e-06, "loss": 2.2032, "step": 28559 }, { "epoch": 0.7, "learning_rate": 4.254938648304896e-06, "loss": 1.8526, "step": 28560 }, { "epoch": 0.7, "learning_rate": 4.254285232365395e-06, "loss": 1.9991, "step": 28561 }, { "epoch": 0.7, "learning_rate": 4.253631853045206e-06, "loss": 1.7715, "step": 28562 }, { "epoch": 0.7, "learning_rate": 4.252978510348495e-06, "loss": 1.9943, "step": 28563 }, { "epoch": 0.7, "learning_rate": 4.252325204279424e-06, "loss": 1.8579, "step": 28564 }, { "epoch": 0.7, "learning_rate": 4.251671934842161e-06, "loss": 1.908, "step": 28565 }, { "epoch": 0.7, "learning_rate": 4.2510187020408625e-06, "loss": 2.0882, "step": 28566 }, { "epoch": 0.7, "learning_rate": 4.250365505879696e-06, "loss": 1.9702, "step": 28567 }, { "epoch": 0.7, "learning_rate": 4.249712346362827e-06, "loss": 1.8155, "step": 28568 }, { "epoch": 0.7, "learning_rate": 4.249059223494412e-06, "loss": 1.7776, "step": 28569 }, { "epoch": 0.7, "learning_rate": 4.248406137278617e-06, "loss": 2.048, "step": 28570 }, { "epoch": 0.7, "learning_rate": 4.247753087719606e-06, "loss": 2.0805, "step": 28571 }, { "epoch": 0.7, "learning_rate": 4.247100074821538e-06, "loss": 1.9518, "step": 28572 }, { "epoch": 0.7, "learning_rate": 4.246447098588578e-06, "loss": 1.8892, "step": 28573 }, { "epoch": 0.7, "learning_rate": 4.245794159024882e-06, "loss": 1.9474, "step": 28574 }, { "epoch": 0.7, "learning_rate": 4.245141256134616e-06, "loss": 1.9626, "step": 28575 }, { "epoch": 0.7, "learning_rate": 4.244488389921944e-06, "loss": 1.8485, "step": 28576 }, { "epoch": 0.7, "learning_rate": 4.243835560391018e-06, "loss": 1.9384, "step": 28577 }, { "epoch": 0.7, "learning_rate": 4.243182767546008e-06, "loss": 2.0221, "step": 28578 }, { "epoch": 0.7, "learning_rate": 4.242530011391068e-06, "loss": 2.0127, "step": 28579 }, { "epoch": 0.7, "learning_rate": 4.241877291930359e-06, "loss": 2.1766, "step": 28580 }, { "epoch": 0.7, "learning_rate": 4.241224609168043e-06, "loss": 1.8924, "step": 28581 }, { "epoch": 0.7, "learning_rate": 4.24057196310828e-06, "loss": 1.8936, "step": 28582 }, { "epoch": 0.7, "learning_rate": 4.2399193537552316e-06, "loss": 2.0252, "step": 28583 }, { "epoch": 0.7, "learning_rate": 4.239266781113051e-06, "loss": 1.9671, "step": 28584 }, { "epoch": 0.7, "learning_rate": 4.2386142451859e-06, "loss": 1.9408, "step": 28585 }, { "epoch": 0.7, "learning_rate": 4.237961745977943e-06, "loss": 2.1898, "step": 28586 }, { "epoch": 0.7, "learning_rate": 4.23730928349333e-06, "loss": 1.9021, "step": 28587 }, { "epoch": 0.7, "learning_rate": 4.236656857736224e-06, "loss": 1.9554, "step": 28588 }, { "epoch": 0.7, "learning_rate": 4.23600446871078e-06, "loss": 1.9698, "step": 28589 }, { "epoch": 0.7, "learning_rate": 4.235352116421159e-06, "loss": 2.0092, "step": 28590 }, { "epoch": 0.7, "learning_rate": 4.23469980087152e-06, "loss": 1.9777, "step": 28591 }, { "epoch": 0.7, "learning_rate": 4.234047522066015e-06, "loss": 1.8875, "step": 28592 }, { "epoch": 0.7, "learning_rate": 4.233395280008807e-06, "loss": 1.9217, "step": 28593 }, { "epoch": 0.7, "learning_rate": 4.232743074704047e-06, "loss": 2.0084, "step": 28594 }, { "epoch": 0.7, "learning_rate": 4.232090906155896e-06, "loss": 2.1525, "step": 28595 }, { "epoch": 0.7, "learning_rate": 4.231438774368511e-06, "loss": 1.9714, "step": 28596 }, { "epoch": 0.7, "learning_rate": 4.230786679346045e-06, "loss": 2.0598, "step": 28597 }, { "epoch": 0.7, "learning_rate": 4.230134621092654e-06, "loss": 1.8737, "step": 28598 }, { "epoch": 0.7, "learning_rate": 4.229482599612496e-06, "loss": 1.9839, "step": 28599 }, { "epoch": 0.7, "learning_rate": 4.228830614909726e-06, "loss": 2.0207, "step": 28600 }, { "epoch": 0.7, "learning_rate": 4.228178666988503e-06, "loss": 1.9678, "step": 28601 }, { "epoch": 0.71, "learning_rate": 4.227526755852974e-06, "loss": 2.0546, "step": 28602 }, { "epoch": 0.71, "learning_rate": 4.226874881507302e-06, "loss": 1.8331, "step": 28603 }, { "epoch": 0.71, "learning_rate": 4.226223043955635e-06, "loss": 1.9198, "step": 28604 }, { "epoch": 0.71, "learning_rate": 4.22557124320213e-06, "loss": 1.9821, "step": 28605 }, { "epoch": 0.71, "learning_rate": 4.224919479250945e-06, "loss": 1.9405, "step": 28606 }, { "epoch": 0.71, "learning_rate": 4.224267752106226e-06, "loss": 2.0839, "step": 28607 }, { "epoch": 0.71, "learning_rate": 4.223616061772134e-06, "loss": 2.045, "step": 28608 }, { "epoch": 0.71, "learning_rate": 4.222964408252816e-06, "loss": 1.9614, "step": 28609 }, { "epoch": 0.71, "learning_rate": 4.2223127915524295e-06, "loss": 2.049, "step": 28610 }, { "epoch": 0.71, "learning_rate": 4.22166121167513e-06, "loss": 2.0713, "step": 28611 }, { "epoch": 0.71, "learning_rate": 4.221009668625063e-06, "loss": 1.8274, "step": 28612 }, { "epoch": 0.71, "learning_rate": 4.220358162406387e-06, "loss": 2.0221, "step": 28613 }, { "epoch": 0.71, "learning_rate": 4.21970669302325e-06, "loss": 2.0103, "step": 28614 }, { "epoch": 0.71, "learning_rate": 4.219055260479805e-06, "loss": 1.9077, "step": 28615 }, { "epoch": 0.71, "learning_rate": 4.218403864780205e-06, "loss": 1.8654, "step": 28616 }, { "epoch": 0.71, "learning_rate": 4.217752505928601e-06, "loss": 1.9472, "step": 28617 }, { "epoch": 0.71, "learning_rate": 4.217101183929149e-06, "loss": 1.884, "step": 28618 }, { "epoch": 0.71, "learning_rate": 4.216449898785991e-06, "loss": 2.0658, "step": 28619 }, { "epoch": 0.71, "learning_rate": 4.215798650503284e-06, "loss": 1.9416, "step": 28620 }, { "epoch": 0.71, "learning_rate": 4.21514743908518e-06, "loss": 2.0531, "step": 28621 }, { "epoch": 0.71, "learning_rate": 4.214496264535824e-06, "loss": 1.9334, "step": 28622 }, { "epoch": 0.71, "learning_rate": 4.213845126859371e-06, "loss": 1.886, "step": 28623 }, { "epoch": 0.71, "learning_rate": 4.213194026059965e-06, "loss": 1.7968, "step": 28624 }, { "epoch": 0.71, "learning_rate": 4.21254296214176e-06, "loss": 2.0419, "step": 28625 }, { "epoch": 0.71, "learning_rate": 4.2118919351089085e-06, "loss": 1.8488, "step": 28626 }, { "epoch": 0.71, "learning_rate": 4.211240944965552e-06, "loss": 1.8509, "step": 28627 }, { "epoch": 0.71, "learning_rate": 4.210589991715847e-06, "loss": 1.8039, "step": 28628 }, { "epoch": 0.71, "learning_rate": 4.209939075363936e-06, "loss": 2.131, "step": 28629 }, { "epoch": 0.71, "learning_rate": 4.209288195913969e-06, "loss": 2.0777, "step": 28630 }, { "epoch": 0.71, "learning_rate": 4.208637353370099e-06, "loss": 1.9416, "step": 28631 }, { "epoch": 0.71, "learning_rate": 4.207986547736468e-06, "loss": 1.9153, "step": 28632 }, { "epoch": 0.71, "learning_rate": 4.207335779017225e-06, "loss": 1.946, "step": 28633 }, { "epoch": 0.71, "learning_rate": 4.20668504721652e-06, "loss": 2.027, "step": 28634 }, { "epoch": 0.71, "learning_rate": 4.206034352338499e-06, "loss": 2.0214, "step": 28635 }, { "epoch": 0.71, "learning_rate": 4.2053836943873125e-06, "loss": 1.9706, "step": 28636 }, { "epoch": 0.71, "learning_rate": 4.204733073367101e-06, "loss": 1.8358, "step": 28637 }, { "epoch": 0.71, "learning_rate": 4.204082489282014e-06, "loss": 1.8492, "step": 28638 }, { "epoch": 0.71, "learning_rate": 4.2034319421362015e-06, "loss": 1.8715, "step": 28639 }, { "epoch": 0.71, "learning_rate": 4.202781431933804e-06, "loss": 1.8378, "step": 28640 }, { "epoch": 0.71, "learning_rate": 4.202130958678971e-06, "loss": 1.9239, "step": 28641 }, { "epoch": 0.71, "learning_rate": 4.201480522375844e-06, "loss": 2.025, "step": 28642 }, { "epoch": 0.71, "learning_rate": 4.200830123028572e-06, "loss": 1.7923, "step": 28643 }, { "epoch": 0.71, "learning_rate": 4.200179760641303e-06, "loss": 2.0415, "step": 28644 }, { "epoch": 0.71, "learning_rate": 4.199529435218175e-06, "loss": 2.0009, "step": 28645 }, { "epoch": 0.71, "learning_rate": 4.198879146763339e-06, "loss": 1.9161, "step": 28646 }, { "epoch": 0.71, "learning_rate": 4.198228895280935e-06, "loss": 1.736, "step": 28647 }, { "epoch": 0.71, "learning_rate": 4.197578680775107e-06, "loss": 1.9656, "step": 28648 }, { "epoch": 0.71, "learning_rate": 4.196928503250005e-06, "loss": 1.8206, "step": 28649 }, { "epoch": 0.71, "learning_rate": 4.196278362709766e-06, "loss": 2.0418, "step": 28650 }, { "epoch": 0.71, "learning_rate": 4.195628259158536e-06, "loss": 2.0309, "step": 28651 }, { "epoch": 0.71, "learning_rate": 4.1949781926004584e-06, "loss": 1.9755, "step": 28652 }, { "epoch": 0.71, "learning_rate": 4.194328163039677e-06, "loss": 1.9619, "step": 28653 }, { "epoch": 0.71, "learning_rate": 4.1936781704803375e-06, "loss": 2.0015, "step": 28654 }, { "epoch": 0.71, "learning_rate": 4.193028214926576e-06, "loss": 1.9253, "step": 28655 }, { "epoch": 0.71, "learning_rate": 4.192378296382541e-06, "loss": 1.8653, "step": 28656 }, { "epoch": 0.71, "learning_rate": 4.191728414852368e-06, "loss": 2.021, "step": 28657 }, { "epoch": 0.71, "learning_rate": 4.191078570340204e-06, "loss": 1.8116, "step": 28658 }, { "epoch": 0.71, "learning_rate": 4.190428762850193e-06, "loss": 1.8464, "step": 28659 }, { "epoch": 0.71, "learning_rate": 4.189778992386468e-06, "loss": 1.8606, "step": 28660 }, { "epoch": 0.71, "learning_rate": 4.189129258953179e-06, "loss": 1.9016, "step": 28661 }, { "epoch": 0.71, "learning_rate": 4.188479562554459e-06, "loss": 1.9641, "step": 28662 }, { "epoch": 0.71, "learning_rate": 4.187829903194454e-06, "loss": 2.077, "step": 28663 }, { "epoch": 0.71, "learning_rate": 4.187180280877306e-06, "loss": 1.9087, "step": 28664 }, { "epoch": 0.71, "learning_rate": 4.186530695607149e-06, "loss": 1.9977, "step": 28665 }, { "epoch": 0.71, "learning_rate": 4.185881147388129e-06, "loss": 1.8604, "step": 28666 }, { "epoch": 0.71, "learning_rate": 4.185231636224379e-06, "loss": 1.9026, "step": 28667 }, { "epoch": 0.71, "learning_rate": 4.184582162120044e-06, "loss": 2.0278, "step": 28668 }, { "epoch": 0.71, "learning_rate": 4.183932725079263e-06, "loss": 2.0219, "step": 28669 }, { "epoch": 0.71, "learning_rate": 4.183283325106174e-06, "loss": 1.9311, "step": 28670 }, { "epoch": 0.71, "learning_rate": 4.182633962204918e-06, "loss": 2.0261, "step": 28671 }, { "epoch": 0.71, "learning_rate": 4.181984636379629e-06, "loss": 1.9504, "step": 28672 }, { "epoch": 0.71, "learning_rate": 4.181335347634448e-06, "loss": 1.686, "step": 28673 }, { "epoch": 0.71, "learning_rate": 4.180686095973517e-06, "loss": 1.8925, "step": 28674 }, { "epoch": 0.71, "learning_rate": 4.180036881400965e-06, "loss": 1.9772, "step": 28675 }, { "epoch": 0.71, "learning_rate": 4.179387703920938e-06, "loss": 1.8823, "step": 28676 }, { "epoch": 0.71, "learning_rate": 4.178738563537569e-06, "loss": 2.1602, "step": 28677 }, { "epoch": 0.71, "learning_rate": 4.1780894602549935e-06, "loss": 2.0204, "step": 28678 }, { "epoch": 0.71, "learning_rate": 4.177440394077357e-06, "loss": 2.1016, "step": 28679 }, { "epoch": 0.71, "learning_rate": 4.176791365008786e-06, "loss": 1.9197, "step": 28680 }, { "epoch": 0.71, "learning_rate": 4.176142373053425e-06, "loss": 2.032, "step": 28681 }, { "epoch": 0.71, "learning_rate": 4.175493418215402e-06, "loss": 2.0122, "step": 28682 }, { "epoch": 0.71, "learning_rate": 4.174844500498858e-06, "loss": 2.1915, "step": 28683 }, { "epoch": 0.71, "learning_rate": 4.174195619907932e-06, "loss": 1.9668, "step": 28684 }, { "epoch": 0.71, "learning_rate": 4.173546776446753e-06, "loss": 2.0043, "step": 28685 }, { "epoch": 0.71, "learning_rate": 4.172897970119458e-06, "loss": 1.8308, "step": 28686 }, { "epoch": 0.71, "learning_rate": 4.1722492009301855e-06, "loss": 1.8513, "step": 28687 }, { "epoch": 0.71, "learning_rate": 4.171600468883066e-06, "loss": 1.9357, "step": 28688 }, { "epoch": 0.71, "learning_rate": 4.17095177398224e-06, "loss": 2.0855, "step": 28689 }, { "epoch": 0.71, "learning_rate": 4.170303116231834e-06, "loss": 2.1173, "step": 28690 }, { "epoch": 0.71, "learning_rate": 4.169654495635988e-06, "loss": 1.8612, "step": 28691 }, { "epoch": 0.71, "learning_rate": 4.1690059121988356e-06, "loss": 1.8078, "step": 28692 }, { "epoch": 0.71, "learning_rate": 4.168357365924506e-06, "loss": 1.9318, "step": 28693 }, { "epoch": 0.71, "learning_rate": 4.1677088568171385e-06, "loss": 2.0326, "step": 28694 }, { "epoch": 0.71, "learning_rate": 4.16706038488086e-06, "loss": 1.8261, "step": 28695 }, { "epoch": 0.71, "learning_rate": 4.166411950119806e-06, "loss": 2.0131, "step": 28696 }, { "epoch": 0.71, "learning_rate": 4.165763552538115e-06, "loss": 1.8172, "step": 28697 }, { "epoch": 0.71, "learning_rate": 4.165115192139909e-06, "loss": 1.9454, "step": 28698 }, { "epoch": 0.71, "learning_rate": 4.16446686892933e-06, "loss": 1.9513, "step": 28699 }, { "epoch": 0.71, "learning_rate": 4.163818582910501e-06, "loss": 2.1336, "step": 28700 }, { "epoch": 0.71, "learning_rate": 4.163170334087558e-06, "loss": 1.9527, "step": 28701 }, { "epoch": 0.71, "learning_rate": 4.1625221224646365e-06, "loss": 1.8914, "step": 28702 }, { "epoch": 0.71, "learning_rate": 4.16187394804586e-06, "loss": 2.0703, "step": 28703 }, { "epoch": 0.71, "learning_rate": 4.1612258108353645e-06, "loss": 1.9402, "step": 28704 }, { "epoch": 0.71, "learning_rate": 4.160577710837279e-06, "loss": 1.8162, "step": 28705 }, { "epoch": 0.71, "learning_rate": 4.159929648055735e-06, "loss": 1.9281, "step": 28706 }, { "epoch": 0.71, "learning_rate": 4.159281622494865e-06, "loss": 1.944, "step": 28707 }, { "epoch": 0.71, "learning_rate": 4.158633634158794e-06, "loss": 2.1443, "step": 28708 }, { "epoch": 0.71, "learning_rate": 4.1579856830516585e-06, "loss": 1.9813, "step": 28709 }, { "epoch": 0.71, "learning_rate": 4.1573377691775806e-06, "loss": 1.8162, "step": 28710 }, { "epoch": 0.71, "learning_rate": 4.156689892540692e-06, "loss": 2.1572, "step": 28711 }, { "epoch": 0.71, "learning_rate": 4.156042053145128e-06, "loss": 1.8506, "step": 28712 }, { "epoch": 0.71, "learning_rate": 4.155394250995007e-06, "loss": 1.9444, "step": 28713 }, { "epoch": 0.71, "learning_rate": 4.154746486094468e-06, "loss": 1.9973, "step": 28714 }, { "epoch": 0.71, "learning_rate": 4.154098758447631e-06, "loss": 1.9009, "step": 28715 }, { "epoch": 0.71, "learning_rate": 4.153451068058628e-06, "loss": 1.9799, "step": 28716 }, { "epoch": 0.71, "learning_rate": 4.15280341493159e-06, "loss": 2.0647, "step": 28717 }, { "epoch": 0.71, "learning_rate": 4.1521557990706376e-06, "loss": 1.9841, "step": 28718 }, { "epoch": 0.71, "learning_rate": 4.1515082204799015e-06, "loss": 1.8348, "step": 28719 }, { "epoch": 0.71, "learning_rate": 4.150860679163514e-06, "loss": 1.9531, "step": 28720 }, { "epoch": 0.71, "learning_rate": 4.1502131751255935e-06, "loss": 1.9964, "step": 28721 }, { "epoch": 0.71, "learning_rate": 4.149565708370271e-06, "loss": 1.9687, "step": 28722 }, { "epoch": 0.71, "learning_rate": 4.148918278901673e-06, "loss": 1.8754, "step": 28723 }, { "epoch": 0.71, "learning_rate": 4.14827088672393e-06, "loss": 2.0478, "step": 28724 }, { "epoch": 0.71, "learning_rate": 4.147623531841159e-06, "loss": 1.8685, "step": 28725 }, { "epoch": 0.71, "learning_rate": 4.14697621425749e-06, "loss": 2.0157, "step": 28726 }, { "epoch": 0.71, "learning_rate": 4.146328933977053e-06, "loss": 2.0011, "step": 28727 }, { "epoch": 0.71, "learning_rate": 4.1456816910039655e-06, "loss": 2.1426, "step": 28728 }, { "epoch": 0.71, "learning_rate": 4.145034485342361e-06, "loss": 2.0294, "step": 28729 }, { "epoch": 0.71, "learning_rate": 4.144387316996355e-06, "loss": 1.9496, "step": 28730 }, { "epoch": 0.71, "learning_rate": 4.143740185970079e-06, "loss": 1.9316, "step": 28731 }, { "epoch": 0.71, "learning_rate": 4.1430930922676575e-06, "loss": 1.9176, "step": 28732 }, { "epoch": 0.71, "learning_rate": 4.142446035893209e-06, "loss": 2.1054, "step": 28733 }, { "epoch": 0.71, "learning_rate": 4.141799016850864e-06, "loss": 1.9705, "step": 28734 }, { "epoch": 0.71, "learning_rate": 4.1411520351447385e-06, "loss": 2.0805, "step": 28735 }, { "epoch": 0.71, "learning_rate": 4.140505090778962e-06, "loss": 1.9809, "step": 28736 }, { "epoch": 0.71, "learning_rate": 4.139858183757659e-06, "loss": 1.9447, "step": 28737 }, { "epoch": 0.71, "learning_rate": 4.139211314084947e-06, "loss": 2.0366, "step": 28738 }, { "epoch": 0.71, "learning_rate": 4.13856448176495e-06, "loss": 2.0219, "step": 28739 }, { "epoch": 0.71, "learning_rate": 4.137917686801792e-06, "loss": 1.9239, "step": 28740 }, { "epoch": 0.71, "learning_rate": 4.137270929199596e-06, "loss": 1.9716, "step": 28741 }, { "epoch": 0.71, "learning_rate": 4.136624208962486e-06, "loss": 1.7644, "step": 28742 }, { "epoch": 0.71, "learning_rate": 4.135977526094577e-06, "loss": 2.2019, "step": 28743 }, { "epoch": 0.71, "learning_rate": 4.135330880599998e-06, "loss": 1.9313, "step": 28744 }, { "epoch": 0.71, "learning_rate": 4.134684272482863e-06, "loss": 1.9948, "step": 28745 }, { "epoch": 0.71, "learning_rate": 4.134037701747296e-06, "loss": 2.0592, "step": 28746 }, { "epoch": 0.71, "learning_rate": 4.133391168397423e-06, "loss": 2.0874, "step": 28747 }, { "epoch": 0.71, "learning_rate": 4.132744672437356e-06, "loss": 2.0862, "step": 28748 }, { "epoch": 0.71, "learning_rate": 4.132098213871218e-06, "loss": 1.9629, "step": 28749 }, { "epoch": 0.71, "learning_rate": 4.131451792703135e-06, "loss": 2.0842, "step": 28750 }, { "epoch": 0.71, "learning_rate": 4.130805408937219e-06, "loss": 2.0313, "step": 28751 }, { "epoch": 0.71, "learning_rate": 4.130159062577596e-06, "loss": 1.9704, "step": 28752 }, { "epoch": 0.71, "learning_rate": 4.129512753628378e-06, "loss": 2.0454, "step": 28753 }, { "epoch": 0.71, "learning_rate": 4.12886648209369e-06, "loss": 1.8575, "step": 28754 }, { "epoch": 0.71, "learning_rate": 4.128220247977651e-06, "loss": 2.1075, "step": 28755 }, { "epoch": 0.71, "learning_rate": 4.127574051284375e-06, "loss": 1.844, "step": 28756 }, { "epoch": 0.71, "learning_rate": 4.1269278920179836e-06, "loss": 1.9203, "step": 28757 }, { "epoch": 0.71, "learning_rate": 4.126281770182595e-06, "loss": 2.0492, "step": 28758 }, { "epoch": 0.71, "learning_rate": 4.125635685782327e-06, "loss": 1.9257, "step": 28759 }, { "epoch": 0.71, "learning_rate": 4.1249896388212994e-06, "loss": 1.8707, "step": 28760 }, { "epoch": 0.71, "learning_rate": 4.124343629303625e-06, "loss": 1.9392, "step": 28761 }, { "epoch": 0.71, "learning_rate": 4.1236976572334265e-06, "loss": 2.0257, "step": 28762 }, { "epoch": 0.71, "learning_rate": 4.123051722614815e-06, "loss": 1.9742, "step": 28763 }, { "epoch": 0.71, "learning_rate": 4.12240582545191e-06, "loss": 1.8522, "step": 28764 }, { "epoch": 0.71, "learning_rate": 4.121759965748832e-06, "loss": 1.8091, "step": 28765 }, { "epoch": 0.71, "learning_rate": 4.121114143509688e-06, "loss": 2.0285, "step": 28766 }, { "epoch": 0.71, "learning_rate": 4.1204683587386055e-06, "loss": 2.0983, "step": 28767 }, { "epoch": 0.71, "learning_rate": 4.119822611439689e-06, "loss": 1.816, "step": 28768 }, { "epoch": 0.71, "learning_rate": 4.119176901617059e-06, "loss": 2.0951, "step": 28769 }, { "epoch": 0.71, "learning_rate": 4.118531229274836e-06, "loss": 2.0596, "step": 28770 }, { "epoch": 0.71, "learning_rate": 4.117885594417126e-06, "loss": 1.9042, "step": 28771 }, { "epoch": 0.71, "learning_rate": 4.117239997048048e-06, "loss": 1.9319, "step": 28772 }, { "epoch": 0.71, "learning_rate": 4.11659443717172e-06, "loss": 1.932, "step": 28773 }, { "epoch": 0.71, "learning_rate": 4.1159489147922494e-06, "loss": 1.874, "step": 28774 }, { "epoch": 0.71, "learning_rate": 4.115303429913754e-06, "loss": 1.8341, "step": 28775 }, { "epoch": 0.71, "learning_rate": 4.114657982540348e-06, "loss": 1.817, "step": 28776 }, { "epoch": 0.71, "learning_rate": 4.1140125726761485e-06, "loss": 2.01, "step": 28777 }, { "epoch": 0.71, "learning_rate": 4.113367200325261e-06, "loss": 2.0197, "step": 28778 }, { "epoch": 0.71, "learning_rate": 4.112721865491802e-06, "loss": 2.0781, "step": 28779 }, { "epoch": 0.71, "learning_rate": 4.11207656817989e-06, "loss": 2.0616, "step": 28780 }, { "epoch": 0.71, "learning_rate": 4.111431308393629e-06, "loss": 2.0166, "step": 28781 }, { "epoch": 0.71, "learning_rate": 4.110786086137138e-06, "loss": 2.0978, "step": 28782 }, { "epoch": 0.71, "learning_rate": 4.110140901414524e-06, "loss": 2.057, "step": 28783 }, { "epoch": 0.71, "learning_rate": 4.109495754229901e-06, "loss": 2.0959, "step": 28784 }, { "epoch": 0.71, "learning_rate": 4.108850644587384e-06, "loss": 1.8289, "step": 28785 }, { "epoch": 0.71, "learning_rate": 4.108205572491079e-06, "loss": 1.9428, "step": 28786 }, { "epoch": 0.71, "learning_rate": 4.107560537945105e-06, "loss": 1.9048, "step": 28787 }, { "epoch": 0.71, "learning_rate": 4.106915540953562e-06, "loss": 1.9481, "step": 28788 }, { "epoch": 0.71, "learning_rate": 4.1062705815205685e-06, "loss": 1.8669, "step": 28789 }, { "epoch": 0.71, "learning_rate": 4.105625659650232e-06, "loss": 1.7532, "step": 28790 }, { "epoch": 0.71, "learning_rate": 4.1049807753466685e-06, "loss": 1.9877, "step": 28791 }, { "epoch": 0.71, "learning_rate": 4.104335928613979e-06, "loss": 1.931, "step": 28792 }, { "epoch": 0.71, "learning_rate": 4.103691119456279e-06, "loss": 1.8192, "step": 28793 }, { "epoch": 0.71, "learning_rate": 4.103046347877677e-06, "loss": 1.9873, "step": 28794 }, { "epoch": 0.71, "learning_rate": 4.102401613882285e-06, "loss": 1.9436, "step": 28795 }, { "epoch": 0.71, "learning_rate": 4.101756917474207e-06, "loss": 1.8727, "step": 28796 }, { "epoch": 0.71, "learning_rate": 4.101112258657558e-06, "loss": 1.97, "step": 28797 }, { "epoch": 0.71, "learning_rate": 4.100467637436438e-06, "loss": 1.9134, "step": 28798 }, { "epoch": 0.71, "learning_rate": 4.0998230538149615e-06, "loss": 1.885, "step": 28799 }, { "epoch": 0.71, "learning_rate": 4.099178507797239e-06, "loss": 1.9512, "step": 28800 }, { "epoch": 0.71, "learning_rate": 4.098533999387372e-06, "loss": 1.9619, "step": 28801 }, { "epoch": 0.71, "learning_rate": 4.097889528589473e-06, "loss": 2.0146, "step": 28802 }, { "epoch": 0.71, "learning_rate": 4.097245095407645e-06, "loss": 1.9249, "step": 28803 }, { "epoch": 0.71, "learning_rate": 4.096600699845997e-06, "loss": 1.8744, "step": 28804 }, { "epoch": 0.71, "learning_rate": 4.095956341908641e-06, "loss": 2.1011, "step": 28805 }, { "epoch": 0.71, "learning_rate": 4.0953120215996755e-06, "loss": 1.9594, "step": 28806 }, { "epoch": 0.71, "learning_rate": 4.0946677389232104e-06, "loss": 1.8113, "step": 28807 }, { "epoch": 0.71, "learning_rate": 4.0940234938833525e-06, "loss": 1.9357, "step": 28808 }, { "epoch": 0.71, "learning_rate": 4.093379286484211e-06, "loss": 1.7888, "step": 28809 }, { "epoch": 0.71, "learning_rate": 4.092735116729885e-06, "loss": 2.035, "step": 28810 }, { "epoch": 0.71, "learning_rate": 4.092090984624482e-06, "loss": 2.0197, "step": 28811 }, { "epoch": 0.71, "learning_rate": 4.0914468901721094e-06, "loss": 1.9932, "step": 28812 }, { "epoch": 0.71, "learning_rate": 4.090802833376876e-06, "loss": 2.012, "step": 28813 }, { "epoch": 0.71, "learning_rate": 4.090158814242876e-06, "loss": 2.0051, "step": 28814 }, { "epoch": 0.71, "learning_rate": 4.089514832774225e-06, "loss": 2.0245, "step": 28815 }, { "epoch": 0.71, "learning_rate": 4.088870888975018e-06, "loss": 2.0466, "step": 28816 }, { "epoch": 0.71, "learning_rate": 4.088226982849364e-06, "loss": 2.0841, "step": 28817 }, { "epoch": 0.71, "learning_rate": 4.087583114401369e-06, "loss": 2.0427, "step": 28818 }, { "epoch": 0.71, "learning_rate": 4.08693928363513e-06, "loss": 1.9015, "step": 28819 }, { "epoch": 0.71, "learning_rate": 4.086295490554758e-06, "loss": 1.9458, "step": 28820 }, { "epoch": 0.71, "learning_rate": 4.085651735164348e-06, "loss": 1.9294, "step": 28821 }, { "epoch": 0.71, "learning_rate": 4.0850080174680065e-06, "loss": 1.8525, "step": 28822 }, { "epoch": 0.71, "learning_rate": 4.084364337469841e-06, "loss": 2.0148, "step": 28823 }, { "epoch": 0.71, "learning_rate": 4.083720695173945e-06, "loss": 1.8912, "step": 28824 }, { "epoch": 0.71, "learning_rate": 4.083077090584426e-06, "loss": 1.8953, "step": 28825 }, { "epoch": 0.71, "learning_rate": 4.082433523705388e-06, "loss": 1.8197, "step": 28826 }, { "epoch": 0.71, "learning_rate": 4.081789994540926e-06, "loss": 1.9947, "step": 28827 }, { "epoch": 0.71, "learning_rate": 4.081146503095146e-06, "loss": 1.8768, "step": 28828 }, { "epoch": 0.71, "learning_rate": 4.080503049372146e-06, "loss": 1.9574, "step": 28829 }, { "epoch": 0.71, "learning_rate": 4.079859633376035e-06, "loss": 2.0467, "step": 28830 }, { "epoch": 0.71, "learning_rate": 4.079216255110904e-06, "loss": 1.7916, "step": 28831 }, { "epoch": 0.71, "learning_rate": 4.078572914580856e-06, "loss": 1.9928, "step": 28832 }, { "epoch": 0.71, "learning_rate": 4.077929611789997e-06, "loss": 1.9683, "step": 28833 }, { "epoch": 0.71, "learning_rate": 4.0772863467424194e-06, "loss": 1.9791, "step": 28834 }, { "epoch": 0.71, "learning_rate": 4.076643119442229e-06, "loss": 2.0073, "step": 28835 }, { "epoch": 0.71, "learning_rate": 4.075999929893519e-06, "loss": 2.1208, "step": 28836 }, { "epoch": 0.71, "learning_rate": 4.0753567781003926e-06, "loss": 1.9883, "step": 28837 }, { "epoch": 0.71, "learning_rate": 4.07471366406695e-06, "loss": 1.9931, "step": 28838 }, { "epoch": 0.71, "learning_rate": 4.0740705877972865e-06, "loss": 1.7781, "step": 28839 }, { "epoch": 0.71, "learning_rate": 4.073427549295504e-06, "loss": 2.0543, "step": 28840 }, { "epoch": 0.71, "learning_rate": 4.072784548565698e-06, "loss": 1.9717, "step": 28841 }, { "epoch": 0.71, "learning_rate": 4.072141585611966e-06, "loss": 1.9238, "step": 28842 }, { "epoch": 0.71, "learning_rate": 4.071498660438408e-06, "loss": 1.9502, "step": 28843 }, { "epoch": 0.71, "learning_rate": 4.070855773049125e-06, "loss": 1.8431, "step": 28844 }, { "epoch": 0.71, "learning_rate": 4.070212923448207e-06, "loss": 2.0301, "step": 28845 }, { "epoch": 0.71, "learning_rate": 4.0695701116397535e-06, "loss": 1.9735, "step": 28846 }, { "epoch": 0.71, "learning_rate": 4.068927337627863e-06, "loss": 2.1214, "step": 28847 }, { "epoch": 0.71, "learning_rate": 4.068284601416634e-06, "loss": 2.0175, "step": 28848 }, { "epoch": 0.71, "learning_rate": 4.067641903010159e-06, "loss": 1.9314, "step": 28849 }, { "epoch": 0.71, "learning_rate": 4.066999242412537e-06, "loss": 2.0675, "step": 28850 }, { "epoch": 0.71, "learning_rate": 4.066356619627859e-06, "loss": 1.936, "step": 28851 }, { "epoch": 0.71, "learning_rate": 4.0657140346602255e-06, "loss": 1.8504, "step": 28852 }, { "epoch": 0.71, "learning_rate": 4.065071487513732e-06, "loss": 2.0676, "step": 28853 }, { "epoch": 0.71, "learning_rate": 4.06442897819247e-06, "loss": 1.9506, "step": 28854 }, { "epoch": 0.71, "learning_rate": 4.0637865067005385e-06, "loss": 1.9953, "step": 28855 }, { "epoch": 0.71, "learning_rate": 4.063144073042028e-06, "loss": 2.0008, "step": 28856 }, { "epoch": 0.71, "learning_rate": 4.062501677221035e-06, "loss": 1.897, "step": 28857 }, { "epoch": 0.71, "learning_rate": 4.061859319241656e-06, "loss": 2.0239, "step": 28858 }, { "epoch": 0.71, "learning_rate": 4.061216999107981e-06, "loss": 2.0351, "step": 28859 }, { "epoch": 0.71, "learning_rate": 4.060574716824104e-06, "loss": 1.8534, "step": 28860 }, { "epoch": 0.71, "learning_rate": 4.05993247239412e-06, "loss": 1.9323, "step": 28861 }, { "epoch": 0.71, "learning_rate": 4.059290265822127e-06, "loss": 1.9763, "step": 28862 }, { "epoch": 0.71, "learning_rate": 4.058648097112209e-06, "loss": 2.1805, "step": 28863 }, { "epoch": 0.71, "learning_rate": 4.058005966268463e-06, "loss": 2.0527, "step": 28864 }, { "epoch": 0.71, "learning_rate": 4.0573638732949835e-06, "loss": 1.8216, "step": 28865 }, { "epoch": 0.71, "learning_rate": 4.056721818195859e-06, "loss": 1.9995, "step": 28866 }, { "epoch": 0.71, "learning_rate": 4.056079800975183e-06, "loss": 1.9701, "step": 28867 }, { "epoch": 0.71, "learning_rate": 4.055437821637049e-06, "loss": 1.954, "step": 28868 }, { "epoch": 0.71, "learning_rate": 4.054795880185546e-06, "loss": 1.9202, "step": 28869 }, { "epoch": 0.71, "learning_rate": 4.054153976624765e-06, "loss": 1.9522, "step": 28870 }, { "epoch": 0.71, "learning_rate": 4.053512110958803e-06, "loss": 2.0908, "step": 28871 }, { "epoch": 0.71, "learning_rate": 4.052870283191741e-06, "loss": 1.9262, "step": 28872 }, { "epoch": 0.71, "learning_rate": 4.0522284933276805e-06, "loss": 1.8331, "step": 28873 }, { "epoch": 0.71, "learning_rate": 4.051586741370701e-06, "loss": 1.8047, "step": 28874 }, { "epoch": 0.71, "learning_rate": 4.0509450273248986e-06, "loss": 2.0441, "step": 28875 }, { "epoch": 0.71, "learning_rate": 4.050303351194364e-06, "loss": 1.9806, "step": 28876 }, { "epoch": 0.71, "learning_rate": 4.049661712983184e-06, "loss": 2.0629, "step": 28877 }, { "epoch": 0.71, "learning_rate": 4.049020112695448e-06, "loss": 1.9599, "step": 28878 }, { "epoch": 0.71, "learning_rate": 4.048378550335246e-06, "loss": 2.0028, "step": 28879 }, { "epoch": 0.71, "learning_rate": 4.047737025906671e-06, "loss": 2.1047, "step": 28880 }, { "epoch": 0.71, "learning_rate": 4.047095539413803e-06, "loss": 1.9731, "step": 28881 }, { "epoch": 0.71, "learning_rate": 4.046454090860737e-06, "loss": 2.0095, "step": 28882 }, { "epoch": 0.71, "learning_rate": 4.045812680251562e-06, "loss": 1.8859, "step": 28883 }, { "epoch": 0.71, "learning_rate": 4.04517130759036e-06, "loss": 1.8856, "step": 28884 }, { "epoch": 0.71, "learning_rate": 4.044529972881222e-06, "loss": 1.7771, "step": 28885 }, { "epoch": 0.71, "learning_rate": 4.043888676128237e-06, "loss": 1.8568, "step": 28886 }, { "epoch": 0.71, "learning_rate": 4.043247417335489e-06, "loss": 2.0548, "step": 28887 }, { "epoch": 0.71, "learning_rate": 4.04260619650707e-06, "loss": 1.9436, "step": 28888 }, { "epoch": 0.71, "learning_rate": 4.041965013647058e-06, "loss": 2.0471, "step": 28889 }, { "epoch": 0.71, "learning_rate": 4.041323868759546e-06, "loss": 1.8831, "step": 28890 }, { "epoch": 0.71, "learning_rate": 4.0406827618486226e-06, "loss": 1.8728, "step": 28891 }, { "epoch": 0.71, "learning_rate": 4.040041692918365e-06, "loss": 1.8962, "step": 28892 }, { "epoch": 0.71, "learning_rate": 4.0394006619728686e-06, "loss": 1.8148, "step": 28893 }, { "epoch": 0.71, "learning_rate": 4.03875966901621e-06, "loss": 2.0257, "step": 28894 }, { "epoch": 0.71, "learning_rate": 4.03811871405248e-06, "loss": 2.0728, "step": 28895 }, { "epoch": 0.71, "learning_rate": 4.037477797085762e-06, "loss": 1.9967, "step": 28896 }, { "epoch": 0.71, "learning_rate": 4.036836918120141e-06, "loss": 2.0799, "step": 28897 }, { "epoch": 0.71, "learning_rate": 4.036196077159704e-06, "loss": 1.9613, "step": 28898 }, { "epoch": 0.71, "learning_rate": 4.03555527420853e-06, "loss": 2.168, "step": 28899 }, { "epoch": 0.71, "learning_rate": 4.034914509270707e-06, "loss": 1.9151, "step": 28900 }, { "epoch": 0.71, "learning_rate": 4.034273782350321e-06, "loss": 2.0586, "step": 28901 }, { "epoch": 0.71, "learning_rate": 4.033633093451448e-06, "loss": 1.8907, "step": 28902 }, { "epoch": 0.71, "learning_rate": 4.032992442578179e-06, "loss": 1.9898, "step": 28903 }, { "epoch": 0.71, "learning_rate": 4.032351829734592e-06, "loss": 1.8804, "step": 28904 }, { "epoch": 0.71, "learning_rate": 4.0317112549247705e-06, "loss": 2.0373, "step": 28905 }, { "epoch": 0.71, "learning_rate": 4.031070718152802e-06, "loss": 1.9842, "step": 28906 }, { "epoch": 0.71, "learning_rate": 4.030430219422761e-06, "loss": 1.8082, "step": 28907 }, { "epoch": 0.71, "learning_rate": 4.029789758738738e-06, "loss": 1.8376, "step": 28908 }, { "epoch": 0.71, "learning_rate": 4.029149336104806e-06, "loss": 2.1845, "step": 28909 }, { "epoch": 0.71, "learning_rate": 4.028508951525052e-06, "loss": 1.8189, "step": 28910 }, { "epoch": 0.71, "learning_rate": 4.027868605003559e-06, "loss": 2.0238, "step": 28911 }, { "epoch": 0.71, "learning_rate": 4.027228296544403e-06, "loss": 1.9859, "step": 28912 }, { "epoch": 0.71, "learning_rate": 4.026588026151667e-06, "loss": 1.9971, "step": 28913 }, { "epoch": 0.71, "learning_rate": 4.025947793829432e-06, "loss": 1.931, "step": 28914 }, { "epoch": 0.71, "learning_rate": 4.025307599581779e-06, "loss": 1.808, "step": 28915 }, { "epoch": 0.71, "learning_rate": 4.024667443412792e-06, "loss": 2.0918, "step": 28916 }, { "epoch": 0.71, "learning_rate": 4.024027325326543e-06, "loss": 1.9836, "step": 28917 }, { "epoch": 0.71, "learning_rate": 4.023387245327119e-06, "loss": 1.9076, "step": 28918 }, { "epoch": 0.71, "learning_rate": 4.02274720341859e-06, "loss": 1.9802, "step": 28919 }, { "epoch": 0.71, "learning_rate": 4.022107199605043e-06, "loss": 1.6995, "step": 28920 }, { "epoch": 0.71, "learning_rate": 4.021467233890558e-06, "loss": 1.8362, "step": 28921 }, { "epoch": 0.71, "learning_rate": 4.020827306279208e-06, "loss": 2.0535, "step": 28922 }, { "epoch": 0.71, "learning_rate": 4.020187416775077e-06, "loss": 2.0624, "step": 28923 }, { "epoch": 0.71, "learning_rate": 4.019547565382236e-06, "loss": 1.8714, "step": 28924 }, { "epoch": 0.71, "learning_rate": 4.018907752104768e-06, "loss": 2.1271, "step": 28925 }, { "epoch": 0.71, "learning_rate": 4.018267976946753e-06, "loss": 1.8041, "step": 28926 }, { "epoch": 0.71, "learning_rate": 4.017628239912264e-06, "loss": 1.7344, "step": 28927 }, { "epoch": 0.71, "learning_rate": 4.016988541005379e-06, "loss": 1.9255, "step": 28928 }, { "epoch": 0.71, "learning_rate": 4.016348880230179e-06, "loss": 2.0478, "step": 28929 }, { "epoch": 0.71, "learning_rate": 4.015709257590733e-06, "loss": 1.8713, "step": 28930 }, { "epoch": 0.71, "learning_rate": 4.015069673091123e-06, "loss": 2.0657, "step": 28931 }, { "epoch": 0.71, "learning_rate": 4.014430126735425e-06, "loss": 1.9112, "step": 28932 }, { "epoch": 0.71, "learning_rate": 4.013790618527718e-06, "loss": 1.9781, "step": 28933 }, { "epoch": 0.71, "learning_rate": 4.0131511484720695e-06, "loss": 1.9868, "step": 28934 }, { "epoch": 0.71, "learning_rate": 4.012511716572561e-06, "loss": 1.9829, "step": 28935 }, { "epoch": 0.71, "learning_rate": 4.011872322833269e-06, "loss": 1.8759, "step": 28936 }, { "epoch": 0.71, "learning_rate": 4.011232967258264e-06, "loss": 2.2191, "step": 28937 }, { "epoch": 0.71, "learning_rate": 4.010593649851622e-06, "loss": 2.0058, "step": 28938 }, { "epoch": 0.71, "learning_rate": 4.009954370617423e-06, "loss": 1.9931, "step": 28939 }, { "epoch": 0.71, "learning_rate": 4.009315129559732e-06, "loss": 2.051, "step": 28940 }, { "epoch": 0.71, "learning_rate": 4.008675926682633e-06, "loss": 1.874, "step": 28941 }, { "epoch": 0.71, "learning_rate": 4.008036761990189e-06, "loss": 2.0367, "step": 28942 }, { "epoch": 0.71, "learning_rate": 4.0073976354864815e-06, "loss": 1.9309, "step": 28943 }, { "epoch": 0.71, "learning_rate": 4.006758547175586e-06, "loss": 1.9315, "step": 28944 }, { "epoch": 0.71, "learning_rate": 4.006119497061566e-06, "loss": 1.8306, "step": 28945 }, { "epoch": 0.71, "learning_rate": 4.005480485148504e-06, "loss": 1.8704, "step": 28946 }, { "epoch": 0.71, "learning_rate": 4.0048415114404644e-06, "loss": 1.9837, "step": 28947 }, { "epoch": 0.71, "learning_rate": 4.004202575941525e-06, "loss": 1.9659, "step": 28948 }, { "epoch": 0.71, "learning_rate": 4.003563678655755e-06, "loss": 1.9429, "step": 28949 }, { "epoch": 0.71, "learning_rate": 4.002924819587229e-06, "loss": 2.105, "step": 28950 }, { "epoch": 0.71, "learning_rate": 4.0022859987400205e-06, "loss": 2.0395, "step": 28951 }, { "epoch": 0.71, "learning_rate": 4.001647216118194e-06, "loss": 1.9807, "step": 28952 }, { "epoch": 0.71, "learning_rate": 4.001008471725826e-06, "loss": 1.9058, "step": 28953 }, { "epoch": 0.71, "learning_rate": 4.000369765566988e-06, "loss": 2.0779, "step": 28954 }, { "epoch": 0.71, "learning_rate": 3.999731097645746e-06, "loss": 1.9355, "step": 28955 }, { "epoch": 0.71, "learning_rate": 3.999092467966177e-06, "loss": 1.8322, "step": 28956 }, { "epoch": 0.71, "learning_rate": 3.998453876532343e-06, "loss": 2.1415, "step": 28957 }, { "epoch": 0.71, "learning_rate": 3.9978153233483184e-06, "loss": 1.8884, "step": 28958 }, { "epoch": 0.71, "learning_rate": 3.997176808418177e-06, "loss": 2.1406, "step": 28959 }, { "epoch": 0.71, "learning_rate": 3.996538331745979e-06, "loss": 1.7612, "step": 28960 }, { "epoch": 0.71, "learning_rate": 3.995899893335804e-06, "loss": 2.1716, "step": 28961 }, { "epoch": 0.71, "learning_rate": 3.995261493191711e-06, "loss": 1.9581, "step": 28962 }, { "epoch": 0.71, "learning_rate": 3.994623131317774e-06, "loss": 1.8631, "step": 28963 }, { "epoch": 0.71, "learning_rate": 3.993984807718064e-06, "loss": 1.7991, "step": 28964 }, { "epoch": 0.71, "learning_rate": 3.993346522396643e-06, "loss": 1.7446, "step": 28965 }, { "epoch": 0.71, "learning_rate": 3.992708275357582e-06, "loss": 1.9385, "step": 28966 }, { "epoch": 0.71, "learning_rate": 3.992070066604949e-06, "loss": 1.7128, "step": 28967 }, { "epoch": 0.71, "learning_rate": 3.991431896142811e-06, "loss": 1.7853, "step": 28968 }, { "epoch": 0.71, "learning_rate": 3.990793763975239e-06, "loss": 1.9784, "step": 28969 }, { "epoch": 0.71, "learning_rate": 3.990155670106294e-06, "loss": 2.0566, "step": 28970 }, { "epoch": 0.71, "learning_rate": 3.989517614540049e-06, "loss": 1.814, "step": 28971 }, { "epoch": 0.71, "learning_rate": 3.988879597280562e-06, "loss": 2.0533, "step": 28972 }, { "epoch": 0.71, "learning_rate": 3.988241618331906e-06, "loss": 1.9752, "step": 28973 }, { "epoch": 0.71, "learning_rate": 3.9876036776981485e-06, "loss": 1.9755, "step": 28974 }, { "epoch": 0.71, "learning_rate": 3.986965775383348e-06, "loss": 1.9044, "step": 28975 }, { "epoch": 0.71, "learning_rate": 3.986327911391579e-06, "loss": 1.8711, "step": 28976 }, { "epoch": 0.71, "learning_rate": 3.985690085726897e-06, "loss": 2.0185, "step": 28977 }, { "epoch": 0.71, "learning_rate": 3.985052298393373e-06, "loss": 1.9657, "step": 28978 }, { "epoch": 0.71, "learning_rate": 3.984414549395074e-06, "loss": 1.974, "step": 28979 }, { "epoch": 0.71, "learning_rate": 3.983776838736058e-06, "loss": 2.0263, "step": 28980 }, { "epoch": 0.71, "learning_rate": 3.983139166420395e-06, "loss": 1.9942, "step": 28981 }, { "epoch": 0.71, "learning_rate": 3.982501532452146e-06, "loss": 1.9737, "step": 28982 }, { "epoch": 0.71, "learning_rate": 3.981863936835374e-06, "loss": 1.8925, "step": 28983 }, { "epoch": 0.71, "learning_rate": 3.981226379574145e-06, "loss": 1.855, "step": 28984 }, { "epoch": 0.71, "learning_rate": 3.980588860672522e-06, "loss": 1.9091, "step": 28985 }, { "epoch": 0.71, "learning_rate": 3.9799513801345675e-06, "loss": 1.7317, "step": 28986 }, { "epoch": 0.71, "learning_rate": 3.979313937964349e-06, "loss": 2.0288, "step": 28987 }, { "epoch": 0.71, "learning_rate": 3.9786765341659205e-06, "loss": 1.839, "step": 28988 }, { "epoch": 0.71, "learning_rate": 3.9780391687433515e-06, "loss": 2.0107, "step": 28989 }, { "epoch": 0.71, "learning_rate": 3.9774018417007e-06, "loss": 2.0686, "step": 28990 }, { "epoch": 0.71, "learning_rate": 3.9767645530420275e-06, "loss": 1.9808, "step": 28991 }, { "epoch": 0.71, "learning_rate": 3.976127302771401e-06, "loss": 1.8831, "step": 28992 }, { "epoch": 0.71, "learning_rate": 3.975490090892876e-06, "loss": 1.8824, "step": 28993 }, { "epoch": 0.71, "learning_rate": 3.974852917410519e-06, "loss": 1.9761, "step": 28994 }, { "epoch": 0.71, "learning_rate": 3.9742157823283835e-06, "loss": 1.9395, "step": 28995 }, { "epoch": 0.71, "learning_rate": 3.973578685650536e-06, "loss": 1.7179, "step": 28996 }, { "epoch": 0.71, "learning_rate": 3.972941627381037e-06, "loss": 1.837, "step": 28997 }, { "epoch": 0.71, "learning_rate": 3.972304607523944e-06, "loss": 2.098, "step": 28998 }, { "epoch": 0.71, "learning_rate": 3.97166762608332e-06, "loss": 2.0226, "step": 28999 }, { "epoch": 0.71, "learning_rate": 3.97103068306322e-06, "loss": 1.8186, "step": 29000 }, { "epoch": 0.71, "learning_rate": 3.970393778467706e-06, "loss": 1.8432, "step": 29001 }, { "epoch": 0.71, "learning_rate": 3.969756912300837e-06, "loss": 1.8907, "step": 29002 }, { "epoch": 0.71, "learning_rate": 3.969120084566673e-06, "loss": 1.8884, "step": 29003 }, { "epoch": 0.71, "learning_rate": 3.968483295269276e-06, "loss": 2.0474, "step": 29004 }, { "epoch": 0.71, "learning_rate": 3.967846544412697e-06, "loss": 1.9088, "step": 29005 }, { "epoch": 0.71, "learning_rate": 3.967209832000997e-06, "loss": 1.8569, "step": 29006 }, { "epoch": 0.71, "learning_rate": 3.966573158038238e-06, "loss": 1.7543, "step": 29007 }, { "epoch": 0.72, "learning_rate": 3.965936522528472e-06, "loss": 1.8958, "step": 29008 }, { "epoch": 0.72, "learning_rate": 3.965299925475763e-06, "loss": 2.0314, "step": 29009 }, { "epoch": 0.72, "learning_rate": 3.964663366884159e-06, "loss": 2.0491, "step": 29010 }, { "epoch": 0.72, "learning_rate": 3.964026846757723e-06, "loss": 2.0117, "step": 29011 }, { "epoch": 0.72, "learning_rate": 3.9633903651005146e-06, "loss": 1.8468, "step": 29012 }, { "epoch": 0.72, "learning_rate": 3.962753921916583e-06, "loss": 2.0393, "step": 29013 }, { "epoch": 0.72, "learning_rate": 3.962117517209993e-06, "loss": 2.0606, "step": 29014 }, { "epoch": 0.72, "learning_rate": 3.961481150984791e-06, "loss": 1.809, "step": 29015 }, { "epoch": 0.72, "learning_rate": 3.960844823245037e-06, "loss": 1.8783, "step": 29016 }, { "epoch": 0.72, "learning_rate": 3.96020853399479e-06, "loss": 1.8457, "step": 29017 }, { "epoch": 0.72, "learning_rate": 3.9595722832381015e-06, "loss": 1.8631, "step": 29018 }, { "epoch": 0.72, "learning_rate": 3.9589360709790245e-06, "loss": 2.074, "step": 29019 }, { "epoch": 0.72, "learning_rate": 3.958299897221618e-06, "loss": 1.9819, "step": 29020 }, { "epoch": 0.72, "learning_rate": 3.957663761969935e-06, "loss": 1.7933, "step": 29021 }, { "epoch": 0.72, "learning_rate": 3.957027665228033e-06, "loss": 1.8943, "step": 29022 }, { "epoch": 0.72, "learning_rate": 3.956391606999961e-06, "loss": 1.7825, "step": 29023 }, { "epoch": 0.72, "learning_rate": 3.955755587289778e-06, "loss": 1.8905, "step": 29024 }, { "epoch": 0.72, "learning_rate": 3.95511960610153e-06, "loss": 2.1124, "step": 29025 }, { "epoch": 0.72, "learning_rate": 3.954483663439276e-06, "loss": 1.8993, "step": 29026 }, { "epoch": 0.72, "learning_rate": 3.95384775930707e-06, "loss": 2.0272, "step": 29027 }, { "epoch": 0.72, "learning_rate": 3.953211893708959e-06, "loss": 1.9715, "step": 29028 }, { "epoch": 0.72, "learning_rate": 3.952576066649003e-06, "loss": 2.0097, "step": 29029 }, { "epoch": 0.72, "learning_rate": 3.951940278131248e-06, "loss": 1.9853, "step": 29030 }, { "epoch": 0.72, "learning_rate": 3.951304528159747e-06, "loss": 1.9919, "step": 29031 }, { "epoch": 0.72, "learning_rate": 3.950668816738559e-06, "loss": 1.7564, "step": 29032 }, { "epoch": 0.72, "learning_rate": 3.950033143871725e-06, "loss": 1.8806, "step": 29033 }, { "epoch": 0.72, "learning_rate": 3.949397509563305e-06, "loss": 1.865, "step": 29034 }, { "epoch": 0.72, "learning_rate": 3.948761913817342e-06, "loss": 1.8903, "step": 29035 }, { "epoch": 0.72, "learning_rate": 3.9481263566378936e-06, "loss": 1.853, "step": 29036 }, { "epoch": 0.72, "learning_rate": 3.947490838029007e-06, "loss": 1.8635, "step": 29037 }, { "epoch": 0.72, "learning_rate": 3.946855357994734e-06, "loss": 1.8787, "step": 29038 }, { "epoch": 0.72, "learning_rate": 3.946219916539128e-06, "loss": 1.9417, "step": 29039 }, { "epoch": 0.72, "learning_rate": 3.945584513666232e-06, "loss": 1.9062, "step": 29040 }, { "epoch": 0.72, "learning_rate": 3.944949149380099e-06, "loss": 1.6199, "step": 29041 }, { "epoch": 0.72, "learning_rate": 3.944313823684781e-06, "loss": 1.9186, "step": 29042 }, { "epoch": 0.72, "learning_rate": 3.943678536584322e-06, "loss": 1.9264, "step": 29043 }, { "epoch": 0.72, "learning_rate": 3.943043288082773e-06, "loss": 1.9291, "step": 29044 }, { "epoch": 0.72, "learning_rate": 3.942408078184187e-06, "loss": 1.8822, "step": 29045 }, { "epoch": 0.72, "learning_rate": 3.941772906892604e-06, "loss": 1.8675, "step": 29046 }, { "epoch": 0.72, "learning_rate": 3.94113777421208e-06, "loss": 1.8937, "step": 29047 }, { "epoch": 0.72, "learning_rate": 3.940502680146656e-06, "loss": 2.0589, "step": 29048 }, { "epoch": 0.72, "learning_rate": 3.939867624700384e-06, "loss": 1.9991, "step": 29049 }, { "epoch": 0.72, "learning_rate": 3.939232607877313e-06, "loss": 1.9198, "step": 29050 }, { "epoch": 0.72, "learning_rate": 3.938597629681485e-06, "loss": 1.8781, "step": 29051 }, { "epoch": 0.72, "learning_rate": 3.937962690116954e-06, "loss": 2.0018, "step": 29052 }, { "epoch": 0.72, "learning_rate": 3.937327789187758e-06, "loss": 1.8639, "step": 29053 }, { "epoch": 0.72, "learning_rate": 3.936692926897948e-06, "loss": 1.8532, "step": 29054 }, { "epoch": 0.72, "learning_rate": 3.936058103251568e-06, "loss": 1.8072, "step": 29055 }, { "epoch": 0.72, "learning_rate": 3.935423318252669e-06, "loss": 1.9805, "step": 29056 }, { "epoch": 0.72, "learning_rate": 3.934788571905296e-06, "loss": 2.0706, "step": 29057 }, { "epoch": 0.72, "learning_rate": 3.934153864213488e-06, "loss": 1.8383, "step": 29058 }, { "epoch": 0.72, "learning_rate": 3.933519195181295e-06, "loss": 2.0367, "step": 29059 }, { "epoch": 0.72, "learning_rate": 3.932884564812763e-06, "loss": 1.7866, "step": 29060 }, { "epoch": 0.72, "learning_rate": 3.932249973111932e-06, "loss": 1.8468, "step": 29061 }, { "epoch": 0.72, "learning_rate": 3.931615420082854e-06, "loss": 1.8927, "step": 29062 }, { "epoch": 0.72, "learning_rate": 3.9309809057295645e-06, "loss": 1.922, "step": 29063 }, { "epoch": 0.72, "learning_rate": 3.930346430056111e-06, "loss": 1.9433, "step": 29064 }, { "epoch": 0.72, "learning_rate": 3.929711993066542e-06, "loss": 1.789, "step": 29065 }, { "epoch": 0.72, "learning_rate": 3.929077594764893e-06, "loss": 1.9236, "step": 29066 }, { "epoch": 0.72, "learning_rate": 3.9284432351552135e-06, "loss": 1.9472, "step": 29067 }, { "epoch": 0.72, "learning_rate": 3.927808914241542e-06, "loss": 1.9903, "step": 29068 }, { "epoch": 0.72, "learning_rate": 3.927174632027921e-06, "loss": 2.079, "step": 29069 }, { "epoch": 0.72, "learning_rate": 3.9265403885184e-06, "loss": 2.1118, "step": 29070 }, { "epoch": 0.72, "learning_rate": 3.925906183717014e-06, "loss": 2.119, "step": 29071 }, { "epoch": 0.72, "learning_rate": 3.9252720176278055e-06, "loss": 2.0309, "step": 29072 }, { "epoch": 0.72, "learning_rate": 3.92463789025482e-06, "loss": 1.8656, "step": 29073 }, { "epoch": 0.72, "learning_rate": 3.924003801602095e-06, "loss": 2.1663, "step": 29074 }, { "epoch": 0.72, "learning_rate": 3.923369751673679e-06, "loss": 1.9434, "step": 29075 }, { "epoch": 0.72, "learning_rate": 3.9227357404736034e-06, "loss": 1.9697, "step": 29076 }, { "epoch": 0.72, "learning_rate": 3.922101768005918e-06, "loss": 1.8211, "step": 29077 }, { "epoch": 0.72, "learning_rate": 3.921467834274655e-06, "loss": 1.9134, "step": 29078 }, { "epoch": 0.72, "learning_rate": 3.920833939283858e-06, "loss": 1.9464, "step": 29079 }, { "epoch": 0.72, "learning_rate": 3.920200083037571e-06, "loss": 1.9799, "step": 29080 }, { "epoch": 0.72, "learning_rate": 3.919566265539826e-06, "loss": 1.944, "step": 29081 }, { "epoch": 0.72, "learning_rate": 3.918932486794671e-06, "loss": 1.8451, "step": 29082 }, { "epoch": 0.72, "learning_rate": 3.918298746806137e-06, "loss": 1.9745, "step": 29083 }, { "epoch": 0.72, "learning_rate": 3.9176650455782685e-06, "loss": 1.7964, "step": 29084 }, { "epoch": 0.72, "learning_rate": 3.917031383115105e-06, "loss": 2.208, "step": 29085 }, { "epoch": 0.72, "learning_rate": 3.91639775942068e-06, "loss": 2.0696, "step": 29086 }, { "epoch": 0.72, "learning_rate": 3.9157641744990385e-06, "loss": 2.0565, "step": 29087 }, { "epoch": 0.72, "learning_rate": 3.91513062835421e-06, "loss": 2.0742, "step": 29088 }, { "epoch": 0.72, "learning_rate": 3.9144971209902375e-06, "loss": 1.861, "step": 29089 }, { "epoch": 0.72, "learning_rate": 3.913863652411158e-06, "loss": 1.8361, "step": 29090 }, { "epoch": 0.72, "learning_rate": 3.91323022262101e-06, "loss": 1.8729, "step": 29091 }, { "epoch": 0.72, "learning_rate": 3.912596831623832e-06, "loss": 1.7479, "step": 29092 }, { "epoch": 0.72, "learning_rate": 3.911963479423654e-06, "loss": 1.8886, "step": 29093 }, { "epoch": 0.72, "learning_rate": 3.911330166024517e-06, "loss": 1.9672, "step": 29094 }, { "epoch": 0.72, "learning_rate": 3.910696891430461e-06, "loss": 2.0152, "step": 29095 }, { "epoch": 0.72, "learning_rate": 3.910063655645515e-06, "loss": 1.8411, "step": 29096 }, { "epoch": 0.72, "learning_rate": 3.909430458673722e-06, "loss": 1.9644, "step": 29097 }, { "epoch": 0.72, "learning_rate": 3.908797300519108e-06, "loss": 2.0485, "step": 29098 }, { "epoch": 0.72, "learning_rate": 3.908164181185716e-06, "loss": 2.0108, "step": 29099 }, { "epoch": 0.72, "learning_rate": 3.9075311006775815e-06, "loss": 1.8608, "step": 29100 }, { "epoch": 0.72, "learning_rate": 3.906898058998733e-06, "loss": 1.8675, "step": 29101 }, { "epoch": 0.72, "learning_rate": 3.9062650561532135e-06, "loss": 1.821, "step": 29102 }, { "epoch": 0.72, "learning_rate": 3.905632092145047e-06, "loss": 2.0623, "step": 29103 }, { "epoch": 0.72, "learning_rate": 3.904999166978275e-06, "loss": 1.8304, "step": 29104 }, { "epoch": 0.72, "learning_rate": 3.904366280656934e-06, "loss": 1.985, "step": 29105 }, { "epoch": 0.72, "learning_rate": 3.903733433185047e-06, "loss": 1.9634, "step": 29106 }, { "epoch": 0.72, "learning_rate": 3.903100624566656e-06, "loss": 2.0405, "step": 29107 }, { "epoch": 0.72, "learning_rate": 3.90246785480579e-06, "loss": 1.982, "step": 29108 }, { "epoch": 0.72, "learning_rate": 3.901835123906485e-06, "loss": 1.7564, "step": 29109 }, { "epoch": 0.72, "learning_rate": 3.901202431872775e-06, "loss": 1.9103, "step": 29110 }, { "epoch": 0.72, "learning_rate": 3.900569778708685e-06, "loss": 1.9452, "step": 29111 }, { "epoch": 0.72, "learning_rate": 3.8999371644182524e-06, "loss": 1.8524, "step": 29112 }, { "epoch": 0.72, "learning_rate": 3.899304589005511e-06, "loss": 1.9015, "step": 29113 }, { "epoch": 0.72, "learning_rate": 3.898672052474487e-06, "loss": 1.7311, "step": 29114 }, { "epoch": 0.72, "learning_rate": 3.898039554829217e-06, "loss": 1.9498, "step": 29115 }, { "epoch": 0.72, "learning_rate": 3.897407096073726e-06, "loss": 1.9774, "step": 29116 }, { "epoch": 0.72, "learning_rate": 3.896774676212049e-06, "loss": 2.1653, "step": 29117 }, { "epoch": 0.72, "learning_rate": 3.89614229524822e-06, "loss": 1.9015, "step": 29118 }, { "epoch": 0.72, "learning_rate": 3.895509953186262e-06, "loss": 2.0725, "step": 29119 }, { "epoch": 0.72, "learning_rate": 3.894877650030211e-06, "loss": 2.0014, "step": 29120 }, { "epoch": 0.72, "learning_rate": 3.8942453857840905e-06, "loss": 2.0109, "step": 29121 }, { "epoch": 0.72, "learning_rate": 3.893613160451934e-06, "loss": 1.9871, "step": 29122 }, { "epoch": 0.72, "learning_rate": 3.892980974037776e-06, "loss": 2.0049, "step": 29123 }, { "epoch": 0.72, "learning_rate": 3.892348826545636e-06, "loss": 1.8219, "step": 29124 }, { "epoch": 0.72, "learning_rate": 3.891716717979545e-06, "loss": 1.8619, "step": 29125 }, { "epoch": 0.72, "learning_rate": 3.891084648343537e-06, "loss": 1.8186, "step": 29126 }, { "epoch": 0.72, "learning_rate": 3.890452617641637e-06, "loss": 1.9148, "step": 29127 }, { "epoch": 0.72, "learning_rate": 3.889820625877876e-06, "loss": 2.0058, "step": 29128 }, { "epoch": 0.72, "learning_rate": 3.889188673056277e-06, "loss": 1.9058, "step": 29129 }, { "epoch": 0.72, "learning_rate": 3.888556759180873e-06, "loss": 2.0044, "step": 29130 }, { "epoch": 0.72, "learning_rate": 3.887924884255685e-06, "loss": 2.0479, "step": 29131 }, { "epoch": 0.72, "learning_rate": 3.887293048284742e-06, "loss": 1.9978, "step": 29132 }, { "epoch": 0.72, "learning_rate": 3.886661251272078e-06, "loss": 1.9377, "step": 29133 }, { "epoch": 0.72, "learning_rate": 3.88602949322171e-06, "loss": 1.926, "step": 29134 }, { "epoch": 0.72, "learning_rate": 3.885397774137672e-06, "loss": 2.0595, "step": 29135 }, { "epoch": 0.72, "learning_rate": 3.884766094023984e-06, "loss": 2.1442, "step": 29136 }, { "epoch": 0.72, "learning_rate": 3.884134452884673e-06, "loss": 2.0376, "step": 29137 }, { "epoch": 0.72, "learning_rate": 3.883502850723771e-06, "loss": 1.9759, "step": 29138 }, { "epoch": 0.72, "learning_rate": 3.882871287545295e-06, "loss": 1.9044, "step": 29139 }, { "epoch": 0.72, "learning_rate": 3.882239763353277e-06, "loss": 2.0409, "step": 29140 }, { "epoch": 0.72, "learning_rate": 3.881608278151735e-06, "loss": 1.9123, "step": 29141 }, { "epoch": 0.72, "learning_rate": 3.880976831944697e-06, "loss": 1.8547, "step": 29142 }, { "epoch": 0.72, "learning_rate": 3.88034542473619e-06, "loss": 1.8596, "step": 29143 }, { "epoch": 0.72, "learning_rate": 3.879714056530234e-06, "loss": 1.851, "step": 29144 }, { "epoch": 0.72, "learning_rate": 3.879082727330858e-06, "loss": 1.925, "step": 29145 }, { "epoch": 0.72, "learning_rate": 3.87845143714208e-06, "loss": 1.9018, "step": 29146 }, { "epoch": 0.72, "learning_rate": 3.877820185967925e-06, "loss": 1.8715, "step": 29147 }, { "epoch": 0.72, "learning_rate": 3.877188973812421e-06, "loss": 2.0708, "step": 29148 }, { "epoch": 0.72, "learning_rate": 3.876557800679584e-06, "loss": 2.0545, "step": 29149 }, { "epoch": 0.72, "learning_rate": 3.8759266665734426e-06, "loss": 1.8695, "step": 29150 }, { "epoch": 0.72, "learning_rate": 3.875295571498015e-06, "loss": 1.9875, "step": 29151 }, { "epoch": 0.72, "learning_rate": 3.874664515457323e-06, "loss": 1.8367, "step": 29152 }, { "epoch": 0.72, "learning_rate": 3.874033498455393e-06, "loss": 1.7466, "step": 29153 }, { "epoch": 0.72, "learning_rate": 3.873402520496241e-06, "loss": 2.0326, "step": 29154 }, { "epoch": 0.72, "learning_rate": 3.872771581583895e-06, "loss": 1.7896, "step": 29155 }, { "epoch": 0.72, "learning_rate": 3.8721406817223686e-06, "loss": 1.8649, "step": 29156 }, { "epoch": 0.72, "learning_rate": 3.871509820915687e-06, "loss": 1.9099, "step": 29157 }, { "epoch": 0.72, "learning_rate": 3.8708789991678745e-06, "loss": 1.9654, "step": 29158 }, { "epoch": 0.72, "learning_rate": 3.870248216482944e-06, "loss": 1.9577, "step": 29159 }, { "epoch": 0.72, "learning_rate": 3.869617472864918e-06, "loss": 1.9787, "step": 29160 }, { "epoch": 0.72, "learning_rate": 3.868986768317819e-06, "loss": 1.9158, "step": 29161 }, { "epoch": 0.72, "learning_rate": 3.868356102845664e-06, "loss": 2.0184, "step": 29162 }, { "epoch": 0.72, "learning_rate": 3.867725476452478e-06, "loss": 1.9843, "step": 29163 }, { "epoch": 0.72, "learning_rate": 3.867094889142273e-06, "loss": 1.8222, "step": 29164 }, { "epoch": 0.72, "learning_rate": 3.8664643409190705e-06, "loss": 1.9703, "step": 29165 }, { "epoch": 0.72, "learning_rate": 3.8658338317868935e-06, "loss": 1.9643, "step": 29166 }, { "epoch": 0.72, "learning_rate": 3.865203361749753e-06, "loss": 2.069, "step": 29167 }, { "epoch": 0.72, "learning_rate": 3.8645729308116745e-06, "loss": 1.7552, "step": 29168 }, { "epoch": 0.72, "learning_rate": 3.8639425389766674e-06, "loss": 2.0309, "step": 29169 }, { "epoch": 0.72, "learning_rate": 3.863312186248756e-06, "loss": 2.1917, "step": 29170 }, { "epoch": 0.72, "learning_rate": 3.862681872631958e-06, "loss": 1.9634, "step": 29171 }, { "epoch": 0.72, "learning_rate": 3.862051598130285e-06, "loss": 1.8912, "step": 29172 }, { "epoch": 0.72, "learning_rate": 3.861421362747762e-06, "loss": 1.9494, "step": 29173 }, { "epoch": 0.72, "learning_rate": 3.8607911664883965e-06, "loss": 1.9242, "step": 29174 }, { "epoch": 0.72, "learning_rate": 3.860161009356211e-06, "loss": 1.9554, "step": 29175 }, { "epoch": 0.72, "learning_rate": 3.859530891355222e-06, "loss": 1.8627, "step": 29176 }, { "epoch": 0.72, "learning_rate": 3.8589008124894415e-06, "loss": 1.7219, "step": 29177 }, { "epoch": 0.72, "learning_rate": 3.858270772762887e-06, "loss": 1.9197, "step": 29178 }, { "epoch": 0.72, "learning_rate": 3.857640772179575e-06, "loss": 1.6986, "step": 29179 }, { "epoch": 0.72, "learning_rate": 3.857010810743519e-06, "loss": 2.0387, "step": 29180 }, { "epoch": 0.72, "learning_rate": 3.85638088845874e-06, "loss": 1.8249, "step": 29181 }, { "epoch": 0.72, "learning_rate": 3.855751005329244e-06, "loss": 1.807, "step": 29182 }, { "epoch": 0.72, "learning_rate": 3.855121161359051e-06, "loss": 2.0118, "step": 29183 }, { "epoch": 0.72, "learning_rate": 3.854491356552171e-06, "loss": 2.0163, "step": 29184 }, { "epoch": 0.72, "learning_rate": 3.853861590912621e-06, "loss": 1.8193, "step": 29185 }, { "epoch": 0.72, "learning_rate": 3.853231864444416e-06, "loss": 1.925, "step": 29186 }, { "epoch": 0.72, "learning_rate": 3.8526021771515655e-06, "loss": 1.9063, "step": 29187 }, { "epoch": 0.72, "learning_rate": 3.851972529038087e-06, "loss": 1.9794, "step": 29188 }, { "epoch": 0.72, "learning_rate": 3.851342920107989e-06, "loss": 1.9854, "step": 29189 }, { "epoch": 0.72, "learning_rate": 3.850713350365286e-06, "loss": 2.0144, "step": 29190 }, { "epoch": 0.72, "learning_rate": 3.850083819813993e-06, "loss": 1.683, "step": 29191 }, { "epoch": 0.72, "learning_rate": 3.8494543284581174e-06, "loss": 1.9648, "step": 29192 }, { "epoch": 0.72, "learning_rate": 3.848824876301678e-06, "loss": 1.9434, "step": 29193 }, { "epoch": 0.72, "learning_rate": 3.848195463348679e-06, "loss": 2.0167, "step": 29194 }, { "epoch": 0.72, "learning_rate": 3.847566089603133e-06, "loss": 1.9754, "step": 29195 }, { "epoch": 0.72, "learning_rate": 3.846936755069055e-06, "loss": 1.8853, "step": 29196 }, { "epoch": 0.72, "learning_rate": 3.846307459750453e-06, "loss": 2.0613, "step": 29197 }, { "epoch": 0.72, "learning_rate": 3.845678203651344e-06, "loss": 1.8444, "step": 29198 }, { "epoch": 0.72, "learning_rate": 3.84504898677573e-06, "loss": 1.8028, "step": 29199 }, { "epoch": 0.72, "learning_rate": 3.844419809127624e-06, "loss": 1.8604, "step": 29200 }, { "epoch": 0.72, "learning_rate": 3.84379067071104e-06, "loss": 1.9478, "step": 29201 }, { "epoch": 0.72, "learning_rate": 3.84316157152998e-06, "loss": 1.8615, "step": 29202 }, { "epoch": 0.72, "learning_rate": 3.8425325115884625e-06, "loss": 1.8153, "step": 29203 }, { "epoch": 0.72, "learning_rate": 3.8419034908904885e-06, "loss": 1.9406, "step": 29204 }, { "epoch": 0.72, "learning_rate": 3.84127450944007e-06, "loss": 2.0799, "step": 29205 }, { "epoch": 0.72, "learning_rate": 3.840645567241219e-06, "loss": 1.8767, "step": 29206 }, { "epoch": 0.72, "learning_rate": 3.840016664297939e-06, "loss": 1.9116, "step": 29207 }, { "epoch": 0.72, "learning_rate": 3.839387800614243e-06, "loss": 2.0506, "step": 29208 }, { "epoch": 0.72, "learning_rate": 3.838758976194132e-06, "loss": 2.0032, "step": 29209 }, { "epoch": 0.72, "learning_rate": 3.8381301910416195e-06, "loss": 1.9933, "step": 29210 }, { "epoch": 0.72, "learning_rate": 3.837501445160714e-06, "loss": 1.9055, "step": 29211 }, { "epoch": 0.72, "learning_rate": 3.836872738555416e-06, "loss": 2.0368, "step": 29212 }, { "epoch": 0.72, "learning_rate": 3.8362440712297375e-06, "loss": 1.8344, "step": 29213 }, { "epoch": 0.72, "learning_rate": 3.835615443187683e-06, "loss": 2.0183, "step": 29214 }, { "epoch": 0.72, "learning_rate": 3.834986854433262e-06, "loss": 2.0514, "step": 29215 }, { "epoch": 0.72, "learning_rate": 3.834358304970482e-06, "loss": 1.8973, "step": 29216 }, { "epoch": 0.72, "learning_rate": 3.833729794803341e-06, "loss": 1.9898, "step": 29217 }, { "epoch": 0.72, "learning_rate": 3.833101323935854e-06, "loss": 1.944, "step": 29218 }, { "epoch": 0.72, "learning_rate": 3.832472892372019e-06, "loss": 1.9927, "step": 29219 }, { "epoch": 0.72, "learning_rate": 3.831844500115843e-06, "loss": 1.9737, "step": 29220 }, { "epoch": 0.72, "learning_rate": 3.831216147171337e-06, "loss": 2.106, "step": 29221 }, { "epoch": 0.72, "learning_rate": 3.830587833542496e-06, "loss": 1.9872, "step": 29222 }, { "epoch": 0.72, "learning_rate": 3.829959559233331e-06, "loss": 1.9718, "step": 29223 }, { "epoch": 0.72, "learning_rate": 3.829331324247846e-06, "loss": 1.9324, "step": 29224 }, { "epoch": 0.72, "learning_rate": 3.828703128590041e-06, "loss": 2.063, "step": 29225 }, { "epoch": 0.72, "learning_rate": 3.828074972263925e-06, "loss": 2.0202, "step": 29226 }, { "epoch": 0.72, "learning_rate": 3.827446855273495e-06, "loss": 2.0336, "step": 29227 }, { "epoch": 0.72, "learning_rate": 3.826818777622759e-06, "loss": 1.7325, "step": 29228 }, { "epoch": 0.72, "learning_rate": 3.826190739315722e-06, "loss": 1.8796, "step": 29229 }, { "epoch": 0.72, "learning_rate": 3.825562740356379e-06, "loss": 2.0398, "step": 29230 }, { "epoch": 0.72, "learning_rate": 3.824934780748737e-06, "loss": 1.8268, "step": 29231 }, { "epoch": 0.72, "learning_rate": 3.824306860496798e-06, "loss": 1.926, "step": 29232 }, { "epoch": 0.72, "learning_rate": 3.823678979604565e-06, "loss": 1.9374, "step": 29233 }, { "epoch": 0.72, "learning_rate": 3.823051138076042e-06, "loss": 1.9277, "step": 29234 }, { "epoch": 0.72, "learning_rate": 3.822423335915222e-06, "loss": 1.9473, "step": 29235 }, { "epoch": 0.72, "learning_rate": 3.821795573126116e-06, "loss": 1.9765, "step": 29236 }, { "epoch": 0.72, "learning_rate": 3.821167849712717e-06, "loss": 1.9829, "step": 29237 }, { "epoch": 0.72, "learning_rate": 3.820540165679029e-06, "loss": 2.049, "step": 29238 }, { "epoch": 0.72, "learning_rate": 3.8199125210290546e-06, "loss": 1.8645, "step": 29239 }, { "epoch": 0.72, "learning_rate": 3.819284915766789e-06, "loss": 1.7236, "step": 29240 }, { "epoch": 0.72, "learning_rate": 3.8186573498962385e-06, "loss": 2.1678, "step": 29241 }, { "epoch": 0.72, "learning_rate": 3.8180298234213955e-06, "loss": 2.1004, "step": 29242 }, { "epoch": 0.72, "learning_rate": 3.817402336346263e-06, "loss": 1.7884, "step": 29243 }, { "epoch": 0.72, "learning_rate": 3.816774888674844e-06, "loss": 1.9025, "step": 29244 }, { "epoch": 0.72, "learning_rate": 3.8161474804111295e-06, "loss": 1.9559, "step": 29245 }, { "epoch": 0.72, "learning_rate": 3.815520111559126e-06, "loss": 1.7984, "step": 29246 }, { "epoch": 0.72, "learning_rate": 3.814892782122826e-06, "loss": 2.0281, "step": 29247 }, { "epoch": 0.72, "learning_rate": 3.8142654921062283e-06, "loss": 1.9302, "step": 29248 }, { "epoch": 0.72, "learning_rate": 3.8136382415133345e-06, "loss": 1.8684, "step": 29249 }, { "epoch": 0.72, "learning_rate": 3.8130110303481383e-06, "loss": 1.9317, "step": 29250 }, { "epoch": 0.72, "learning_rate": 3.8123838586146433e-06, "loss": 2.0636, "step": 29251 }, { "epoch": 0.72, "learning_rate": 3.81175672631684e-06, "loss": 2.0095, "step": 29252 }, { "epoch": 0.72, "learning_rate": 3.8111296334587256e-06, "loss": 1.9871, "step": 29253 }, { "epoch": 0.72, "learning_rate": 3.810502580044304e-06, "loss": 1.8949, "step": 29254 }, { "epoch": 0.72, "learning_rate": 3.809875566077563e-06, "loss": 1.9572, "step": 29255 }, { "epoch": 0.72, "learning_rate": 3.8092485915625053e-06, "loss": 1.8824, "step": 29256 }, { "epoch": 0.72, "learning_rate": 3.8086216565031207e-06, "loss": 1.8369, "step": 29257 }, { "epoch": 0.72, "learning_rate": 3.8079947609034075e-06, "loss": 1.8998, "step": 29258 }, { "epoch": 0.72, "learning_rate": 3.8073679047673663e-06, "loss": 1.9882, "step": 29259 }, { "epoch": 0.72, "learning_rate": 3.806741088098983e-06, "loss": 1.7827, "step": 29260 }, { "epoch": 0.72, "learning_rate": 3.8061143109022602e-06, "loss": 1.8997, "step": 29261 }, { "epoch": 0.72, "learning_rate": 3.805487573181187e-06, "loss": 1.9159, "step": 29262 }, { "epoch": 0.72, "learning_rate": 3.8048608749397597e-06, "loss": 2.0506, "step": 29263 }, { "epoch": 0.72, "learning_rate": 3.804234216181977e-06, "loss": 1.9723, "step": 29264 }, { "epoch": 0.72, "learning_rate": 3.8036075969118245e-06, "loss": 1.878, "step": 29265 }, { "epoch": 0.72, "learning_rate": 3.8029810171333004e-06, "loss": 1.9754, "step": 29266 }, { "epoch": 0.72, "learning_rate": 3.802354476850397e-06, "loss": 1.8703, "step": 29267 }, { "epoch": 0.72, "learning_rate": 3.80172797606711e-06, "loss": 2.0398, "step": 29268 }, { "epoch": 0.72, "learning_rate": 3.801101514787433e-06, "loss": 1.846, "step": 29269 }, { "epoch": 0.72, "learning_rate": 3.8004750930153534e-06, "loss": 1.961, "step": 29270 }, { "epoch": 0.72, "learning_rate": 3.799848710754869e-06, "loss": 1.8948, "step": 29271 }, { "epoch": 0.72, "learning_rate": 3.799222368009966e-06, "loss": 1.8612, "step": 29272 }, { "epoch": 0.72, "learning_rate": 3.7985960647846397e-06, "loss": 2.1027, "step": 29273 }, { "epoch": 0.72, "learning_rate": 3.797969801082886e-06, "loss": 1.9701, "step": 29274 }, { "epoch": 0.72, "learning_rate": 3.7973435769086874e-06, "loss": 2.0358, "step": 29275 }, { "epoch": 0.72, "learning_rate": 3.7967173922660437e-06, "loss": 2.0053, "step": 29276 }, { "epoch": 0.72, "learning_rate": 3.796091247158938e-06, "loss": 1.9203, "step": 29277 }, { "epoch": 0.72, "learning_rate": 3.7954651415913645e-06, "loss": 1.8777, "step": 29278 }, { "epoch": 0.72, "learning_rate": 3.794839075567317e-06, "loss": 2.0024, "step": 29279 }, { "epoch": 0.72, "learning_rate": 3.7942130490907793e-06, "loss": 1.9464, "step": 29280 }, { "epoch": 0.72, "learning_rate": 3.7935870621657434e-06, "loss": 2.139, "step": 29281 }, { "epoch": 0.72, "learning_rate": 3.7929611147962043e-06, "loss": 1.9213, "step": 29282 }, { "epoch": 0.72, "learning_rate": 3.7923352069861417e-06, "loss": 1.9111, "step": 29283 }, { "epoch": 0.72, "learning_rate": 3.791709338739552e-06, "loss": 2.0354, "step": 29284 }, { "epoch": 0.72, "learning_rate": 3.7910835100604202e-06, "loss": 1.9513, "step": 29285 }, { "epoch": 0.72, "learning_rate": 3.790457720952737e-06, "loss": 1.7887, "step": 29286 }, { "epoch": 0.72, "learning_rate": 3.7898319714204944e-06, "loss": 1.9661, "step": 29287 }, { "epoch": 0.72, "learning_rate": 3.789206261467673e-06, "loss": 1.9601, "step": 29288 }, { "epoch": 0.72, "learning_rate": 3.7885805910982677e-06, "loss": 2.0697, "step": 29289 }, { "epoch": 0.72, "learning_rate": 3.787954960316258e-06, "loss": 1.9076, "step": 29290 }, { "epoch": 0.72, "learning_rate": 3.7873293691256364e-06, "loss": 1.7863, "step": 29291 }, { "epoch": 0.72, "learning_rate": 3.786703817530393e-06, "loss": 1.7954, "step": 29292 }, { "epoch": 0.72, "learning_rate": 3.7860783055345075e-06, "loss": 1.8978, "step": 29293 }, { "epoch": 0.72, "learning_rate": 3.7854528331419736e-06, "loss": 2.0856, "step": 29294 }, { "epoch": 0.72, "learning_rate": 3.78482740035677e-06, "loss": 2.0293, "step": 29295 }, { "epoch": 0.72, "learning_rate": 3.784202007182888e-06, "loss": 2.0201, "step": 29296 }, { "epoch": 0.72, "learning_rate": 3.7835766536243145e-06, "loss": 2.006, "step": 29297 }, { "epoch": 0.72, "learning_rate": 3.7829513396850305e-06, "loss": 2.0025, "step": 29298 }, { "epoch": 0.72, "learning_rate": 3.7823260653690232e-06, "loss": 2.0406, "step": 29299 }, { "epoch": 0.72, "learning_rate": 3.7817008306802805e-06, "loss": 1.7775, "step": 29300 }, { "epoch": 0.72, "learning_rate": 3.781075635622783e-06, "loss": 1.8873, "step": 29301 }, { "epoch": 0.72, "learning_rate": 3.7804504802005173e-06, "loss": 2.0823, "step": 29302 }, { "epoch": 0.72, "learning_rate": 3.7798253644174664e-06, "loss": 1.8176, "step": 29303 }, { "epoch": 0.72, "learning_rate": 3.7792002882776192e-06, "loss": 1.9809, "step": 29304 }, { "epoch": 0.72, "learning_rate": 3.778575251784953e-06, "loss": 1.9224, "step": 29305 }, { "epoch": 0.72, "learning_rate": 3.7779502549434544e-06, "loss": 2.1348, "step": 29306 }, { "epoch": 0.72, "learning_rate": 3.7773252977571094e-06, "loss": 1.9282, "step": 29307 }, { "epoch": 0.72, "learning_rate": 3.776700380229894e-06, "loss": 2.0314, "step": 29308 }, { "epoch": 0.72, "learning_rate": 3.7760755023658e-06, "loss": 1.9575, "step": 29309 }, { "epoch": 0.72, "learning_rate": 3.7754506641688017e-06, "loss": 2.1027, "step": 29310 }, { "epoch": 0.72, "learning_rate": 3.7748258656428836e-06, "loss": 1.9802, "step": 29311 }, { "epoch": 0.72, "learning_rate": 3.7742011067920336e-06, "loss": 2.0385, "step": 29312 }, { "epoch": 0.72, "learning_rate": 3.773576387620225e-06, "loss": 1.8323, "step": 29313 }, { "epoch": 0.72, "learning_rate": 3.772951708131447e-06, "loss": 2.0431, "step": 29314 }, { "epoch": 0.72, "learning_rate": 3.772327068329673e-06, "loss": 1.9434, "step": 29315 }, { "epoch": 0.72, "learning_rate": 3.7717024682188884e-06, "loss": 2.0127, "step": 29316 }, { "epoch": 0.72, "learning_rate": 3.7710779078030734e-06, "loss": 1.9545, "step": 29317 }, { "epoch": 0.72, "learning_rate": 3.7704533870862124e-06, "loss": 1.8124, "step": 29318 }, { "epoch": 0.72, "learning_rate": 3.7698289060722794e-06, "loss": 2.0234, "step": 29319 }, { "epoch": 0.72, "learning_rate": 3.769204464765257e-06, "loss": 1.8323, "step": 29320 }, { "epoch": 0.72, "learning_rate": 3.768580063169125e-06, "loss": 1.996, "step": 29321 }, { "epoch": 0.72, "learning_rate": 3.7679557012878653e-06, "loss": 2.0461, "step": 29322 }, { "epoch": 0.72, "learning_rate": 3.7673313791254527e-06, "loss": 1.8559, "step": 29323 }, { "epoch": 0.72, "learning_rate": 3.766707096685872e-06, "loss": 2.0477, "step": 29324 }, { "epoch": 0.72, "learning_rate": 3.766082853973094e-06, "loss": 1.8624, "step": 29325 }, { "epoch": 0.72, "learning_rate": 3.7654586509911017e-06, "loss": 1.9214, "step": 29326 }, { "epoch": 0.72, "learning_rate": 3.764834487743877e-06, "loss": 1.9091, "step": 29327 }, { "epoch": 0.72, "learning_rate": 3.764210364235391e-06, "loss": 1.7955, "step": 29328 }, { "epoch": 0.72, "learning_rate": 3.7635862804696277e-06, "loss": 1.98, "step": 29329 }, { "epoch": 0.72, "learning_rate": 3.7629622364505568e-06, "loss": 2.0189, "step": 29330 }, { "epoch": 0.72, "learning_rate": 3.7623382321821622e-06, "loss": 1.8986, "step": 29331 }, { "epoch": 0.72, "learning_rate": 3.7617142676684205e-06, "loss": 2.0457, "step": 29332 }, { "epoch": 0.72, "learning_rate": 3.7610903429133037e-06, "loss": 1.8849, "step": 29333 }, { "epoch": 0.72, "learning_rate": 3.7604664579207926e-06, "loss": 1.9108, "step": 29334 }, { "epoch": 0.72, "learning_rate": 3.759842612694864e-06, "loss": 1.9611, "step": 29335 }, { "epoch": 0.72, "learning_rate": 3.7592188072394898e-06, "loss": 1.8787, "step": 29336 }, { "epoch": 0.72, "learning_rate": 3.758595041558647e-06, "loss": 2.0316, "step": 29337 }, { "epoch": 0.72, "learning_rate": 3.7579713156563126e-06, "loss": 2.0888, "step": 29338 }, { "epoch": 0.72, "learning_rate": 3.7573476295364652e-06, "loss": 2.0082, "step": 29339 }, { "epoch": 0.72, "learning_rate": 3.756723983203071e-06, "loss": 1.878, "step": 29340 }, { "epoch": 0.72, "learning_rate": 3.75610037666011e-06, "loss": 1.9152, "step": 29341 }, { "epoch": 0.72, "learning_rate": 3.7554768099115603e-06, "loss": 1.8595, "step": 29342 }, { "epoch": 0.72, "learning_rate": 3.7548532829613883e-06, "loss": 1.9543, "step": 29343 }, { "epoch": 0.72, "learning_rate": 3.754229795813571e-06, "loss": 1.864, "step": 29344 }, { "epoch": 0.72, "learning_rate": 3.7536063484720863e-06, "loss": 1.8795, "step": 29345 }, { "epoch": 0.72, "learning_rate": 3.7529829409409e-06, "loss": 2.0873, "step": 29346 }, { "epoch": 0.72, "learning_rate": 3.7523595732239937e-06, "loss": 2.0693, "step": 29347 }, { "epoch": 0.72, "learning_rate": 3.7517362453253315e-06, "loss": 2.0054, "step": 29348 }, { "epoch": 0.72, "learning_rate": 3.751112957248891e-06, "loss": 1.9093, "step": 29349 }, { "epoch": 0.72, "learning_rate": 3.750489708998648e-06, "loss": 1.861, "step": 29350 }, { "epoch": 0.72, "learning_rate": 3.749866500578566e-06, "loss": 2.19, "step": 29351 }, { "epoch": 0.72, "learning_rate": 3.7492433319926224e-06, "loss": 1.9659, "step": 29352 }, { "epoch": 0.72, "learning_rate": 3.748620203244792e-06, "loss": 1.9715, "step": 29353 }, { "epoch": 0.72, "learning_rate": 3.747997114339038e-06, "loss": 1.8143, "step": 29354 }, { "epoch": 0.72, "learning_rate": 3.747374065279337e-06, "loss": 2.0394, "step": 29355 }, { "epoch": 0.72, "learning_rate": 3.7467510560696572e-06, "loss": 1.892, "step": 29356 }, { "epoch": 0.72, "learning_rate": 3.746128086713976e-06, "loss": 1.8549, "step": 29357 }, { "epoch": 0.72, "learning_rate": 3.745505157216254e-06, "loss": 1.9245, "step": 29358 }, { "epoch": 0.72, "learning_rate": 3.7448822675804663e-06, "loss": 2.0596, "step": 29359 }, { "epoch": 0.72, "learning_rate": 3.744259417810585e-06, "loss": 1.8665, "step": 29360 }, { "epoch": 0.72, "learning_rate": 3.7436366079105748e-06, "loss": 2.0748, "step": 29361 }, { "epoch": 0.72, "learning_rate": 3.7430138378844105e-06, "loss": 1.8948, "step": 29362 }, { "epoch": 0.72, "learning_rate": 3.7423911077360542e-06, "loss": 1.7511, "step": 29363 }, { "epoch": 0.72, "learning_rate": 3.741768417469479e-06, "loss": 1.9317, "step": 29364 }, { "epoch": 0.72, "learning_rate": 3.7411457670886566e-06, "loss": 2.0923, "step": 29365 }, { "epoch": 0.72, "learning_rate": 3.7405231565975484e-06, "loss": 2.044, "step": 29366 }, { "epoch": 0.72, "learning_rate": 3.7399005860001294e-06, "loss": 1.7949, "step": 29367 }, { "epoch": 0.72, "learning_rate": 3.7392780553003605e-06, "loss": 1.948, "step": 29368 }, { "epoch": 0.72, "learning_rate": 3.738655564502214e-06, "loss": 2.0193, "step": 29369 }, { "epoch": 0.72, "learning_rate": 3.7380331136096547e-06, "loss": 1.9911, "step": 29370 }, { "epoch": 0.72, "learning_rate": 3.737410702626655e-06, "loss": 2.0026, "step": 29371 }, { "epoch": 0.72, "learning_rate": 3.736788331557175e-06, "loss": 1.9868, "step": 29372 }, { "epoch": 0.72, "learning_rate": 3.7361660004051835e-06, "loss": 1.8677, "step": 29373 }, { "epoch": 0.72, "learning_rate": 3.7355437091746473e-06, "loss": 1.9902, "step": 29374 }, { "epoch": 0.72, "learning_rate": 3.734921457869536e-06, "loss": 2.0846, "step": 29375 }, { "epoch": 0.72, "learning_rate": 3.7342992464938088e-06, "loss": 1.9897, "step": 29376 }, { "epoch": 0.72, "learning_rate": 3.7336770750514372e-06, "loss": 2.0767, "step": 29377 }, { "epoch": 0.72, "learning_rate": 3.7330549435463816e-06, "loss": 1.9057, "step": 29378 }, { "epoch": 0.72, "learning_rate": 3.732432851982608e-06, "loss": 2.0146, "step": 29379 }, { "epoch": 0.72, "learning_rate": 3.7318108003640863e-06, "loss": 1.9112, "step": 29380 }, { "epoch": 0.72, "learning_rate": 3.731188788694774e-06, "loss": 2.1539, "step": 29381 }, { "epoch": 0.72, "learning_rate": 3.7305668169786414e-06, "loss": 2.0141, "step": 29382 }, { "epoch": 0.72, "learning_rate": 3.7299448852196464e-06, "loss": 1.8254, "step": 29383 }, { "epoch": 0.72, "learning_rate": 3.729322993421756e-06, "loss": 2.0509, "step": 29384 }, { "epoch": 0.72, "learning_rate": 3.7287011415889375e-06, "loss": 1.8387, "step": 29385 }, { "epoch": 0.72, "learning_rate": 3.728079329725146e-06, "loss": 1.9844, "step": 29386 }, { "epoch": 0.72, "learning_rate": 3.72745755783435e-06, "loss": 1.9446, "step": 29387 }, { "epoch": 0.72, "learning_rate": 3.726835825920511e-06, "loss": 1.8591, "step": 29388 }, { "epoch": 0.72, "learning_rate": 3.7262141339875933e-06, "loss": 1.9308, "step": 29389 }, { "epoch": 0.72, "learning_rate": 3.725592482039556e-06, "loss": 2.1022, "step": 29390 }, { "epoch": 0.72, "learning_rate": 3.724970870080362e-06, "loss": 2.003, "step": 29391 }, { "epoch": 0.72, "learning_rate": 3.724349298113977e-06, "loss": 2.0762, "step": 29392 }, { "epoch": 0.72, "learning_rate": 3.723727766144356e-06, "loss": 1.9588, "step": 29393 }, { "epoch": 0.72, "learning_rate": 3.723106274175463e-06, "loss": 2.1061, "step": 29394 }, { "epoch": 0.72, "learning_rate": 3.722484822211263e-06, "loss": 2.0489, "step": 29395 }, { "epoch": 0.72, "learning_rate": 3.7218634102557094e-06, "loss": 1.8581, "step": 29396 }, { "epoch": 0.72, "learning_rate": 3.721242038312769e-06, "loss": 2.1184, "step": 29397 }, { "epoch": 0.72, "learning_rate": 3.7206207063863965e-06, "loss": 1.9429, "step": 29398 }, { "epoch": 0.72, "learning_rate": 3.719999414480555e-06, "loss": 1.9711, "step": 29399 }, { "epoch": 0.72, "learning_rate": 3.7193781625992065e-06, "loss": 1.9756, "step": 29400 }, { "epoch": 0.72, "learning_rate": 3.7187569507463053e-06, "loss": 2.1024, "step": 29401 }, { "epoch": 0.72, "learning_rate": 3.718135778925812e-06, "loss": 2.071, "step": 29402 }, { "epoch": 0.72, "learning_rate": 3.717514647141691e-06, "loss": 1.8121, "step": 29403 }, { "epoch": 0.72, "learning_rate": 3.7168935553978924e-06, "loss": 2.0046, "step": 29404 }, { "epoch": 0.72, "learning_rate": 3.7162725036983793e-06, "loss": 2.0306, "step": 29405 }, { "epoch": 0.72, "learning_rate": 3.7156514920471087e-06, "loss": 1.8651, "step": 29406 }, { "epoch": 0.72, "learning_rate": 3.715030520448043e-06, "loss": 1.8277, "step": 29407 }, { "epoch": 0.72, "learning_rate": 3.7144095889051322e-06, "loss": 1.9367, "step": 29408 }, { "epoch": 0.72, "learning_rate": 3.713788697422338e-06, "loss": 1.9037, "step": 29409 }, { "epoch": 0.72, "learning_rate": 3.71316784600362e-06, "loss": 2.1528, "step": 29410 }, { "epoch": 0.72, "learning_rate": 3.712547034652929e-06, "loss": 1.9749, "step": 29411 }, { "epoch": 0.72, "learning_rate": 3.7119262633742247e-06, "loss": 1.9886, "step": 29412 }, { "epoch": 0.72, "learning_rate": 3.7113055321714675e-06, "loss": 1.8421, "step": 29413 }, { "epoch": 0.73, "learning_rate": 3.710684841048605e-06, "loss": 1.9868, "step": 29414 }, { "epoch": 0.73, "learning_rate": 3.7100641900096026e-06, "loss": 1.9673, "step": 29415 }, { "epoch": 0.73, "learning_rate": 3.709443579058406e-06, "loss": 1.76, "step": 29416 }, { "epoch": 0.73, "learning_rate": 3.708823008198976e-06, "loss": 1.9992, "step": 29417 }, { "epoch": 0.73, "learning_rate": 3.708202477435271e-06, "loss": 2.0447, "step": 29418 }, { "epoch": 0.73, "learning_rate": 3.7075819867712383e-06, "loss": 1.6971, "step": 29419 }, { "epoch": 0.73, "learning_rate": 3.70696153621084e-06, "loss": 2.0153, "step": 29420 }, { "epoch": 0.73, "learning_rate": 3.706341125758023e-06, "loss": 1.9225, "step": 29421 }, { "epoch": 0.73, "learning_rate": 3.7057207554167452e-06, "loss": 2.0429, "step": 29422 }, { "epoch": 0.73, "learning_rate": 3.7051004251909607e-06, "loss": 1.981, "step": 29423 }, { "epoch": 0.73, "learning_rate": 3.704480135084625e-06, "loss": 1.8883, "step": 29424 }, { "epoch": 0.73, "learning_rate": 3.703859885101686e-06, "loss": 1.9126, "step": 29425 }, { "epoch": 0.73, "learning_rate": 3.7032396752461e-06, "loss": 1.8007, "step": 29426 }, { "epoch": 0.73, "learning_rate": 3.7026195055218196e-06, "loss": 2.0559, "step": 29427 }, { "epoch": 0.73, "learning_rate": 3.7019993759328012e-06, "loss": 2.0507, "step": 29428 }, { "epoch": 0.73, "learning_rate": 3.701379286482989e-06, "loss": 1.9829, "step": 29429 }, { "epoch": 0.73, "learning_rate": 3.7007592371763434e-06, "loss": 1.8354, "step": 29430 }, { "epoch": 0.73, "learning_rate": 3.700139228016808e-06, "loss": 2.0608, "step": 29431 }, { "epoch": 0.73, "learning_rate": 3.6995192590083386e-06, "loss": 1.8784, "step": 29432 }, { "epoch": 0.73, "learning_rate": 3.698899330154889e-06, "loss": 2.0242, "step": 29433 }, { "epoch": 0.73, "learning_rate": 3.6982794414604053e-06, "loss": 1.9318, "step": 29434 }, { "epoch": 0.73, "learning_rate": 3.697659592928843e-06, "loss": 1.9375, "step": 29435 }, { "epoch": 0.73, "learning_rate": 3.697039784564147e-06, "loss": 1.9093, "step": 29436 }, { "epoch": 0.73, "learning_rate": 3.6964200163702704e-06, "loss": 1.7115, "step": 29437 }, { "epoch": 0.73, "learning_rate": 3.6958002883511668e-06, "loss": 1.6917, "step": 29438 }, { "epoch": 0.73, "learning_rate": 3.6951806005107784e-06, "loss": 1.786, "step": 29439 }, { "epoch": 0.73, "learning_rate": 3.694560952853059e-06, "loss": 2.0869, "step": 29440 }, { "epoch": 0.73, "learning_rate": 3.6939413453819583e-06, "loss": 1.9064, "step": 29441 }, { "epoch": 0.73, "learning_rate": 3.693321778101426e-06, "loss": 1.7464, "step": 29442 }, { "epoch": 0.73, "learning_rate": 3.6927022510154076e-06, "loss": 2.0153, "step": 29443 }, { "epoch": 0.73, "learning_rate": 3.692082764127853e-06, "loss": 1.8431, "step": 29444 }, { "epoch": 0.73, "learning_rate": 3.691463317442714e-06, "loss": 2.027, "step": 29445 }, { "epoch": 0.73, "learning_rate": 3.6908439109639317e-06, "loss": 2.1015, "step": 29446 }, { "epoch": 0.73, "learning_rate": 3.6902245446954576e-06, "loss": 1.8385, "step": 29447 }, { "epoch": 0.73, "learning_rate": 3.689605218641241e-06, "loss": 2.0631, "step": 29448 }, { "epoch": 0.73, "learning_rate": 3.6889859328052246e-06, "loss": 1.992, "step": 29449 }, { "epoch": 0.73, "learning_rate": 3.6883666871913605e-06, "loss": 1.981, "step": 29450 }, { "epoch": 0.73, "learning_rate": 3.6877474818035895e-06, "loss": 2.0047, "step": 29451 }, { "epoch": 0.73, "learning_rate": 3.6871283166458605e-06, "loss": 1.9777, "step": 29452 }, { "epoch": 0.73, "learning_rate": 3.686509191722123e-06, "loss": 2.0979, "step": 29453 }, { "epoch": 0.73, "learning_rate": 3.6858901070363173e-06, "loss": 2.1444, "step": 29454 }, { "epoch": 0.73, "learning_rate": 3.685271062592395e-06, "loss": 2.0018, "step": 29455 }, { "epoch": 0.73, "learning_rate": 3.6846520583942934e-06, "loss": 2.0485, "step": 29456 }, { "epoch": 0.73, "learning_rate": 3.6840330944459635e-06, "loss": 1.919, "step": 29457 }, { "epoch": 0.73, "learning_rate": 3.6834141707513493e-06, "loss": 1.991, "step": 29458 }, { "epoch": 0.73, "learning_rate": 3.6827952873143946e-06, "loss": 2.0878, "step": 29459 }, { "epoch": 0.73, "learning_rate": 3.6821764441390474e-06, "loss": 1.9656, "step": 29460 }, { "epoch": 0.73, "learning_rate": 3.6815576412292452e-06, "loss": 2.1031, "step": 29461 }, { "epoch": 0.73, "learning_rate": 3.6809388785889365e-06, "loss": 1.8363, "step": 29462 }, { "epoch": 0.73, "learning_rate": 3.6803201562220658e-06, "loss": 1.9556, "step": 29463 }, { "epoch": 0.73, "learning_rate": 3.679701474132571e-06, "loss": 1.9975, "step": 29464 }, { "epoch": 0.73, "learning_rate": 3.679082832324399e-06, "loss": 1.8561, "step": 29465 }, { "epoch": 0.73, "learning_rate": 3.6784642308014953e-06, "loss": 2.0525, "step": 29466 }, { "epoch": 0.73, "learning_rate": 3.6778456695677955e-06, "loss": 2.1333, "step": 29467 }, { "epoch": 0.73, "learning_rate": 3.6772271486272494e-06, "loss": 1.8308, "step": 29468 }, { "epoch": 0.73, "learning_rate": 3.676608667983791e-06, "loss": 1.833, "step": 29469 }, { "epoch": 0.73, "learning_rate": 3.6759902276413673e-06, "loss": 1.645, "step": 29470 }, { "epoch": 0.73, "learning_rate": 3.675371827603921e-06, "loss": 1.9899, "step": 29471 }, { "epoch": 0.73, "learning_rate": 3.674753467875388e-06, "loss": 2.0808, "step": 29472 }, { "epoch": 0.73, "learning_rate": 3.6741351484597175e-06, "loss": 1.9677, "step": 29473 }, { "epoch": 0.73, "learning_rate": 3.6735168693608404e-06, "loss": 1.9404, "step": 29474 }, { "epoch": 0.73, "learning_rate": 3.672898630582702e-06, "loss": 2.1284, "step": 29475 }, { "epoch": 0.73, "learning_rate": 3.6722804321292425e-06, "loss": 1.9314, "step": 29476 }, { "epoch": 0.73, "learning_rate": 3.671662274004403e-06, "loss": 1.8915, "step": 29477 }, { "epoch": 0.73, "learning_rate": 3.6710441562121248e-06, "loss": 1.8161, "step": 29478 }, { "epoch": 0.73, "learning_rate": 3.6704260787563415e-06, "loss": 1.9362, "step": 29479 }, { "epoch": 0.73, "learning_rate": 3.6698080416409954e-06, "loss": 1.9913, "step": 29480 }, { "epoch": 0.73, "learning_rate": 3.669190044870029e-06, "loss": 1.8429, "step": 29481 }, { "epoch": 0.73, "learning_rate": 3.668572088447375e-06, "loss": 2.0367, "step": 29482 }, { "epoch": 0.73, "learning_rate": 3.6679541723769776e-06, "loss": 1.973, "step": 29483 }, { "epoch": 0.73, "learning_rate": 3.667336296662767e-06, "loss": 2.053, "step": 29484 }, { "epoch": 0.73, "learning_rate": 3.666718461308687e-06, "loss": 2.2054, "step": 29485 }, { "epoch": 0.73, "learning_rate": 3.6661006663186784e-06, "loss": 2.0116, "step": 29486 }, { "epoch": 0.73, "learning_rate": 3.6654829116966696e-06, "loss": 1.9569, "step": 29487 }, { "epoch": 0.73, "learning_rate": 3.6648651974466067e-06, "loss": 1.9847, "step": 29488 }, { "epoch": 0.73, "learning_rate": 3.6642475235724182e-06, "loss": 1.9771, "step": 29489 }, { "epoch": 0.73, "learning_rate": 3.6636298900780466e-06, "loss": 1.9595, "step": 29490 }, { "epoch": 0.73, "learning_rate": 3.6630122969674286e-06, "loss": 1.7767, "step": 29491 }, { "epoch": 0.73, "learning_rate": 3.662394744244495e-06, "loss": 1.931, "step": 29492 }, { "epoch": 0.73, "learning_rate": 3.661777231913186e-06, "loss": 1.9726, "step": 29493 }, { "epoch": 0.73, "learning_rate": 3.6611597599774362e-06, "loss": 2.0463, "step": 29494 }, { "epoch": 0.73, "learning_rate": 3.6605423284411802e-06, "loss": 2.1547, "step": 29495 }, { "epoch": 0.73, "learning_rate": 3.6599249373083577e-06, "loss": 2.0496, "step": 29496 }, { "epoch": 0.73, "learning_rate": 3.6593075865828964e-06, "loss": 1.7858, "step": 29497 }, { "epoch": 0.73, "learning_rate": 3.6586902762687382e-06, "loss": 2.0727, "step": 29498 }, { "epoch": 0.73, "learning_rate": 3.658073006369809e-06, "loss": 2.0905, "step": 29499 }, { "epoch": 0.73, "learning_rate": 3.6574557768900476e-06, "loss": 1.9806, "step": 29500 }, { "epoch": 0.73, "learning_rate": 3.6568385878333912e-06, "loss": 1.9587, "step": 29501 }, { "epoch": 0.73, "learning_rate": 3.6562214392037667e-06, "loss": 2.0173, "step": 29502 }, { "epoch": 0.73, "learning_rate": 3.6556043310051126e-06, "loss": 1.7285, "step": 29503 }, { "epoch": 0.73, "learning_rate": 3.6549872632413576e-06, "loss": 2.1974, "step": 29504 }, { "epoch": 0.73, "learning_rate": 3.6543702359164357e-06, "loss": 1.9475, "step": 29505 }, { "epoch": 0.73, "learning_rate": 3.653753249034283e-06, "loss": 1.922, "step": 29506 }, { "epoch": 0.73, "learning_rate": 3.6531363025988264e-06, "loss": 1.982, "step": 29507 }, { "epoch": 0.73, "learning_rate": 3.652519396614005e-06, "loss": 1.991, "step": 29508 }, { "epoch": 0.73, "learning_rate": 3.6519025310837407e-06, "loss": 1.8261, "step": 29509 }, { "epoch": 0.73, "learning_rate": 3.6512857060119712e-06, "loss": 1.9759, "step": 29510 }, { "epoch": 0.73, "learning_rate": 3.650668921402627e-06, "loss": 1.9451, "step": 29511 }, { "epoch": 0.73, "learning_rate": 3.6500521772596385e-06, "loss": 1.8242, "step": 29512 }, { "epoch": 0.73, "learning_rate": 3.649435473586941e-06, "loss": 2.0061, "step": 29513 }, { "epoch": 0.73, "learning_rate": 3.648818810388457e-06, "loss": 2.0428, "step": 29514 }, { "epoch": 0.73, "learning_rate": 3.6482021876681206e-06, "loss": 1.9192, "step": 29515 }, { "epoch": 0.73, "learning_rate": 3.647585605429865e-06, "loss": 1.8703, "step": 29516 }, { "epoch": 0.73, "learning_rate": 3.6469690636776133e-06, "loss": 1.9422, "step": 29517 }, { "epoch": 0.73, "learning_rate": 3.6463525624152986e-06, "loss": 1.9784, "step": 29518 }, { "epoch": 0.73, "learning_rate": 3.6457361016468516e-06, "loss": 1.9911, "step": 29519 }, { "epoch": 0.73, "learning_rate": 3.6451196813761967e-06, "loss": 2.0723, "step": 29520 }, { "epoch": 0.73, "learning_rate": 3.6445033016072697e-06, "loss": 2.1806, "step": 29521 }, { "epoch": 0.73, "learning_rate": 3.643886962343989e-06, "loss": 1.8542, "step": 29522 }, { "epoch": 0.73, "learning_rate": 3.643270663590289e-06, "loss": 2.0867, "step": 29523 }, { "epoch": 0.73, "learning_rate": 3.6426544053501e-06, "loss": 2.1295, "step": 29524 }, { "epoch": 0.73, "learning_rate": 3.6420381876273424e-06, "loss": 1.9162, "step": 29525 }, { "epoch": 0.73, "learning_rate": 3.6414220104259513e-06, "loss": 1.7748, "step": 29526 }, { "epoch": 0.73, "learning_rate": 3.6408058737498465e-06, "loss": 1.9717, "step": 29527 }, { "epoch": 0.73, "learning_rate": 3.6401897776029583e-06, "loss": 1.8827, "step": 29528 }, { "epoch": 0.73, "learning_rate": 3.6395737219892126e-06, "loss": 2.1151, "step": 29529 }, { "epoch": 0.73, "learning_rate": 3.638957706912537e-06, "loss": 1.8719, "step": 29530 }, { "epoch": 0.73, "learning_rate": 3.63834173237686e-06, "loss": 1.9697, "step": 29531 }, { "epoch": 0.73, "learning_rate": 3.6377257983861014e-06, "loss": 2.0784, "step": 29532 }, { "epoch": 0.73, "learning_rate": 3.6371099049441895e-06, "loss": 1.9505, "step": 29533 }, { "epoch": 0.73, "learning_rate": 3.6364940520550528e-06, "loss": 2.0388, "step": 29534 }, { "epoch": 0.73, "learning_rate": 3.63587823972261e-06, "loss": 1.8623, "step": 29535 }, { "epoch": 0.73, "learning_rate": 3.6352624679507933e-06, "loss": 1.9701, "step": 29536 }, { "epoch": 0.73, "learning_rate": 3.6346467367435188e-06, "loss": 1.9093, "step": 29537 }, { "epoch": 0.73, "learning_rate": 3.6340310461047157e-06, "loss": 2.0182, "step": 29538 }, { "epoch": 0.73, "learning_rate": 3.63341539603831e-06, "loss": 1.9352, "step": 29539 }, { "epoch": 0.73, "learning_rate": 3.63279978654822e-06, "loss": 2.1423, "step": 29540 }, { "epoch": 0.73, "learning_rate": 3.632184217638376e-06, "loss": 1.9783, "step": 29541 }, { "epoch": 0.73, "learning_rate": 3.6315686893126923e-06, "loss": 1.8765, "step": 29542 }, { "epoch": 0.73, "learning_rate": 3.630953201575098e-06, "loss": 1.9948, "step": 29543 }, { "epoch": 0.73, "learning_rate": 3.630337754429517e-06, "loss": 1.9211, "step": 29544 }, { "epoch": 0.73, "learning_rate": 3.629722347879867e-06, "loss": 1.9452, "step": 29545 }, { "epoch": 0.73, "learning_rate": 3.6291069819300717e-06, "loss": 2.0828, "step": 29546 }, { "epoch": 0.73, "learning_rate": 3.6284916565840532e-06, "loss": 2.0003, "step": 29547 }, { "epoch": 0.73, "learning_rate": 3.6278763718457357e-06, "loss": 1.9266, "step": 29548 }, { "epoch": 0.73, "learning_rate": 3.6272611277190404e-06, "loss": 1.9228, "step": 29549 }, { "epoch": 0.73, "learning_rate": 3.6266459242078846e-06, "loss": 1.9377, "step": 29550 }, { "epoch": 0.73, "learning_rate": 3.6260307613161938e-06, "loss": 2.1356, "step": 29551 }, { "epoch": 0.73, "learning_rate": 3.6254156390478825e-06, "loss": 1.8942, "step": 29552 }, { "epoch": 0.73, "learning_rate": 3.6248005574068755e-06, "loss": 2.0013, "step": 29553 }, { "epoch": 0.73, "learning_rate": 3.6241855163970953e-06, "loss": 1.8734, "step": 29554 }, { "epoch": 0.73, "learning_rate": 3.6235705160224554e-06, "loss": 2.0351, "step": 29555 }, { "epoch": 0.73, "learning_rate": 3.6229555562868824e-06, "loss": 2.0233, "step": 29556 }, { "epoch": 0.73, "learning_rate": 3.6223406371942873e-06, "loss": 2.0598, "step": 29557 }, { "epoch": 0.73, "learning_rate": 3.621725758748594e-06, "loss": 1.9626, "step": 29558 }, { "epoch": 0.73, "learning_rate": 3.6211109209537254e-06, "loss": 2.1615, "step": 29559 }, { "epoch": 0.73, "learning_rate": 3.6204961238135916e-06, "loss": 1.9207, "step": 29560 }, { "epoch": 0.73, "learning_rate": 3.6198813673321188e-06, "loss": 1.9707, "step": 29561 }, { "epoch": 0.73, "learning_rate": 3.619266651513218e-06, "loss": 1.9005, "step": 29562 }, { "epoch": 0.73, "learning_rate": 3.61865197636081e-06, "loss": 2.1036, "step": 29563 }, { "epoch": 0.73, "learning_rate": 3.6180373418788117e-06, "loss": 2.1275, "step": 29564 }, { "epoch": 0.73, "learning_rate": 3.617422748071142e-06, "loss": 1.7651, "step": 29565 }, { "epoch": 0.73, "learning_rate": 3.616808194941721e-06, "loss": 1.9267, "step": 29566 }, { "epoch": 0.73, "learning_rate": 3.6161936824944575e-06, "loss": 2.0401, "step": 29567 }, { "epoch": 0.73, "learning_rate": 3.615579210733272e-06, "loss": 1.7754, "step": 29568 }, { "epoch": 0.73, "learning_rate": 3.6149647796620845e-06, "loss": 1.9195, "step": 29569 }, { "epoch": 0.73, "learning_rate": 3.6143503892848042e-06, "loss": 1.9703, "step": 29570 }, { "epoch": 0.73, "learning_rate": 3.6137360396053535e-06, "loss": 2.0352, "step": 29571 }, { "epoch": 0.73, "learning_rate": 3.61312173062764e-06, "loss": 1.9011, "step": 29572 }, { "epoch": 0.73, "learning_rate": 3.6125074623555835e-06, "loss": 1.8707, "step": 29573 }, { "epoch": 0.73, "learning_rate": 3.6118932347931023e-06, "loss": 1.961, "step": 29574 }, { "epoch": 0.73, "learning_rate": 3.611279047944104e-06, "loss": 1.7564, "step": 29575 }, { "epoch": 0.73, "learning_rate": 3.6106649018125096e-06, "loss": 1.8829, "step": 29576 }, { "epoch": 0.73, "learning_rate": 3.6100507964022257e-06, "loss": 1.7839, "step": 29577 }, { "epoch": 0.73, "learning_rate": 3.609436731717171e-06, "loss": 1.9728, "step": 29578 }, { "epoch": 0.73, "learning_rate": 3.6088227077612624e-06, "loss": 1.8186, "step": 29579 }, { "epoch": 0.73, "learning_rate": 3.608208724538406e-06, "loss": 1.7864, "step": 29580 }, { "epoch": 0.73, "learning_rate": 3.607594782052518e-06, "loss": 1.8589, "step": 29581 }, { "epoch": 0.73, "learning_rate": 3.6069808803075124e-06, "loss": 1.7753, "step": 29582 }, { "epoch": 0.73, "learning_rate": 3.606367019307301e-06, "loss": 1.9678, "step": 29583 }, { "epoch": 0.73, "learning_rate": 3.605753199055799e-06, "loss": 1.8097, "step": 29584 }, { "epoch": 0.73, "learning_rate": 3.6051394195569123e-06, "loss": 2.0025, "step": 29585 }, { "epoch": 0.73, "learning_rate": 3.604525680814557e-06, "loss": 2.1043, "step": 29586 }, { "epoch": 0.73, "learning_rate": 3.6039119828326465e-06, "loss": 1.8792, "step": 29587 }, { "epoch": 0.73, "learning_rate": 3.6032983256150857e-06, "loss": 2.1999, "step": 29588 }, { "epoch": 0.73, "learning_rate": 3.6026847091657933e-06, "loss": 2.075, "step": 29589 }, { "epoch": 0.73, "learning_rate": 3.6020711334886728e-06, "loss": 1.928, "step": 29590 }, { "epoch": 0.73, "learning_rate": 3.6014575985876375e-06, "loss": 1.9599, "step": 29591 }, { "epoch": 0.73, "learning_rate": 3.600844104466601e-06, "loss": 1.8606, "step": 29592 }, { "epoch": 0.73, "learning_rate": 3.6002306511294683e-06, "loss": 1.8929, "step": 29593 }, { "epoch": 0.73, "learning_rate": 3.5996172385801543e-06, "loss": 2.0324, "step": 29594 }, { "epoch": 0.73, "learning_rate": 3.5990038668225613e-06, "loss": 1.9278, "step": 29595 }, { "epoch": 0.73, "learning_rate": 3.598390535860603e-06, "loss": 2.0233, "step": 29596 }, { "epoch": 0.73, "learning_rate": 3.597777245698192e-06, "loss": 1.9676, "step": 29597 }, { "epoch": 0.73, "learning_rate": 3.597163996339229e-06, "loss": 1.9664, "step": 29598 }, { "epoch": 0.73, "learning_rate": 3.596550787787627e-06, "loss": 2.0054, "step": 29599 }, { "epoch": 0.73, "learning_rate": 3.5959376200472926e-06, "loss": 1.8995, "step": 29600 }, { "epoch": 0.73, "learning_rate": 3.5953244931221366e-06, "loss": 2.0041, "step": 29601 }, { "epoch": 0.73, "learning_rate": 3.594711407016067e-06, "loss": 1.8972, "step": 29602 }, { "epoch": 0.73, "learning_rate": 3.5940983617329862e-06, "loss": 2.0165, "step": 29603 }, { "epoch": 0.73, "learning_rate": 3.593485357276807e-06, "loss": 2.0143, "step": 29604 }, { "epoch": 0.73, "learning_rate": 3.59287239365143e-06, "loss": 2.0018, "step": 29605 }, { "epoch": 0.73, "learning_rate": 3.592259470860766e-06, "loss": 1.862, "step": 29606 }, { "epoch": 0.73, "learning_rate": 3.591646588908723e-06, "loss": 2.0604, "step": 29607 }, { "epoch": 0.73, "learning_rate": 3.591033747799202e-06, "loss": 1.9032, "step": 29608 }, { "epoch": 0.73, "learning_rate": 3.5904209475361153e-06, "loss": 1.8134, "step": 29609 }, { "epoch": 0.73, "learning_rate": 3.5898081881233606e-06, "loss": 1.9668, "step": 29610 }, { "epoch": 0.73, "learning_rate": 3.5891954695648467e-06, "loss": 1.9224, "step": 29611 }, { "epoch": 0.73, "learning_rate": 3.5885827918644846e-06, "loss": 1.9666, "step": 29612 }, { "epoch": 0.73, "learning_rate": 3.587970155026169e-06, "loss": 2.0066, "step": 29613 }, { "epoch": 0.73, "learning_rate": 3.5873575590538125e-06, "loss": 1.848, "step": 29614 }, { "epoch": 0.73, "learning_rate": 3.586745003951313e-06, "loss": 1.9868, "step": 29615 }, { "epoch": 0.73, "learning_rate": 3.586132489722578e-06, "loss": 2.0448, "step": 29616 }, { "epoch": 0.73, "learning_rate": 3.5855200163715108e-06, "loss": 2.0012, "step": 29617 }, { "epoch": 0.73, "learning_rate": 3.584907583902014e-06, "loss": 2.0088, "step": 29618 }, { "epoch": 0.73, "learning_rate": 3.584295192317997e-06, "loss": 2.0583, "step": 29619 }, { "epoch": 0.73, "learning_rate": 3.5836828416233527e-06, "loss": 2.0294, "step": 29620 }, { "epoch": 0.73, "learning_rate": 3.5830705318219896e-06, "loss": 1.9888, "step": 29621 }, { "epoch": 0.73, "learning_rate": 3.582458262917812e-06, "loss": 2.0084, "step": 29622 }, { "epoch": 0.73, "learning_rate": 3.5818460349147156e-06, "loss": 2.1212, "step": 29623 }, { "epoch": 0.73, "learning_rate": 3.58123384781661e-06, "loss": 2.038, "step": 29624 }, { "epoch": 0.73, "learning_rate": 3.580621701627389e-06, "loss": 1.9646, "step": 29625 }, { "epoch": 0.73, "learning_rate": 3.5800095963509585e-06, "loss": 2.0472, "step": 29626 }, { "epoch": 0.73, "learning_rate": 3.579397531991222e-06, "loss": 1.9884, "step": 29627 }, { "epoch": 0.73, "learning_rate": 3.578785508552075e-06, "loss": 1.8237, "step": 29628 }, { "epoch": 0.73, "learning_rate": 3.578173526037423e-06, "loss": 1.9846, "step": 29629 }, { "epoch": 0.73, "learning_rate": 3.57756158445116e-06, "loss": 1.8962, "step": 29630 }, { "epoch": 0.73, "learning_rate": 3.5769496837971918e-06, "loss": 1.861, "step": 29631 }, { "epoch": 0.73, "learning_rate": 3.576337824079419e-06, "loss": 1.9008, "step": 29632 }, { "epoch": 0.73, "learning_rate": 3.5757260053017353e-06, "loss": 2.0796, "step": 29633 }, { "epoch": 0.73, "learning_rate": 3.575114227468043e-06, "loss": 1.9807, "step": 29634 }, { "epoch": 0.73, "learning_rate": 3.5745024905822413e-06, "loss": 2.0088, "step": 29635 }, { "epoch": 0.73, "learning_rate": 3.5738907946482293e-06, "loss": 1.8593, "step": 29636 }, { "epoch": 0.73, "learning_rate": 3.5732791396699094e-06, "loss": 2.0678, "step": 29637 }, { "epoch": 0.73, "learning_rate": 3.572667525651172e-06, "loss": 1.8622, "step": 29638 }, { "epoch": 0.73, "learning_rate": 3.57205595259592e-06, "loss": 1.9563, "step": 29639 }, { "epoch": 0.73, "learning_rate": 3.5714444205080533e-06, "loss": 2.0589, "step": 29640 }, { "epoch": 0.73, "learning_rate": 3.5708329293914623e-06, "loss": 2.0449, "step": 29641 }, { "epoch": 0.73, "learning_rate": 3.5702214792500524e-06, "loss": 1.9258, "step": 29642 }, { "epoch": 0.73, "learning_rate": 3.5696100700877124e-06, "loss": 1.837, "step": 29643 }, { "epoch": 0.73, "learning_rate": 3.568998701908344e-06, "loss": 1.87, "step": 29644 }, { "epoch": 0.73, "learning_rate": 3.5683873747158446e-06, "loss": 2.0074, "step": 29645 }, { "epoch": 0.73, "learning_rate": 3.5677760885141067e-06, "loss": 2.0593, "step": 29646 }, { "epoch": 0.73, "learning_rate": 3.5671648433070295e-06, "loss": 2.0253, "step": 29647 }, { "epoch": 0.73, "learning_rate": 3.5665536390985056e-06, "loss": 1.9898, "step": 29648 }, { "epoch": 0.73, "learning_rate": 3.565942475892431e-06, "loss": 1.9643, "step": 29649 }, { "epoch": 0.73, "learning_rate": 3.565331353692705e-06, "loss": 1.8421, "step": 29650 }, { "epoch": 0.73, "learning_rate": 3.564720272503216e-06, "loss": 1.9478, "step": 29651 }, { "epoch": 0.73, "learning_rate": 3.5641092323278627e-06, "loss": 1.863, "step": 29652 }, { "epoch": 0.73, "learning_rate": 3.563498233170538e-06, "loss": 1.9832, "step": 29653 }, { "epoch": 0.73, "learning_rate": 3.562887275035136e-06, "loss": 1.7742, "step": 29654 }, { "epoch": 0.73, "learning_rate": 3.5622763579255557e-06, "loss": 2.079, "step": 29655 }, { "epoch": 0.73, "learning_rate": 3.561665481845683e-06, "loss": 1.9838, "step": 29656 }, { "epoch": 0.73, "learning_rate": 3.5610546467994167e-06, "loss": 1.8899, "step": 29657 }, { "epoch": 0.73, "learning_rate": 3.5604438527906448e-06, "loss": 1.8839, "step": 29658 }, { "epoch": 0.73, "learning_rate": 3.5598330998232623e-06, "loss": 2.0682, "step": 29659 }, { "epoch": 0.73, "learning_rate": 3.5592223879011657e-06, "loss": 2.0493, "step": 29660 }, { "epoch": 0.73, "learning_rate": 3.5586117170282407e-06, "loss": 1.867, "step": 29661 }, { "epoch": 0.73, "learning_rate": 3.558001087208386e-06, "loss": 1.7131, "step": 29662 }, { "epoch": 0.73, "learning_rate": 3.5573904984454865e-06, "loss": 1.7687, "step": 29663 }, { "epoch": 0.73, "learning_rate": 3.5567799507434365e-06, "loss": 2.0159, "step": 29664 }, { "epoch": 0.73, "learning_rate": 3.556169444106131e-06, "loss": 1.8906, "step": 29665 }, { "epoch": 0.73, "learning_rate": 3.5555589785374546e-06, "loss": 1.7376, "step": 29666 }, { "epoch": 0.73, "learning_rate": 3.554948554041304e-06, "loss": 1.8289, "step": 29667 }, { "epoch": 0.73, "learning_rate": 3.554338170621563e-06, "loss": 2.0872, "step": 29668 }, { "epoch": 0.73, "learning_rate": 3.5537278282821265e-06, "loss": 1.9114, "step": 29669 }, { "epoch": 0.73, "learning_rate": 3.5531175270268824e-06, "loss": 1.8533, "step": 29670 }, { "epoch": 0.73, "learning_rate": 3.5525072668597217e-06, "loss": 1.7345, "step": 29671 }, { "epoch": 0.73, "learning_rate": 3.551897047784536e-06, "loss": 1.9654, "step": 29672 }, { "epoch": 0.73, "learning_rate": 3.5512868698052093e-06, "loss": 1.8794, "step": 29673 }, { "epoch": 0.73, "learning_rate": 3.550676732925633e-06, "loss": 2.0054, "step": 29674 }, { "epoch": 0.73, "learning_rate": 3.5500666371496984e-06, "loss": 2.0996, "step": 29675 }, { "epoch": 0.73, "learning_rate": 3.549456582481288e-06, "loss": 1.9848, "step": 29676 }, { "epoch": 0.73, "learning_rate": 3.548846568924297e-06, "loss": 1.9484, "step": 29677 }, { "epoch": 0.73, "learning_rate": 3.548236596482605e-06, "loss": 1.9677, "step": 29678 }, { "epoch": 0.73, "learning_rate": 3.5476266651601034e-06, "loss": 1.8481, "step": 29679 }, { "epoch": 0.73, "learning_rate": 3.547016774960684e-06, "loss": 2.0238, "step": 29680 }, { "epoch": 0.73, "learning_rate": 3.546406925888226e-06, "loss": 2.1072, "step": 29681 }, { "epoch": 0.73, "learning_rate": 3.5457971179466233e-06, "loss": 1.8841, "step": 29682 }, { "epoch": 0.73, "learning_rate": 3.545187351139755e-06, "loss": 1.9515, "step": 29683 }, { "epoch": 0.73, "learning_rate": 3.5445776254715114e-06, "loss": 1.8894, "step": 29684 }, { "epoch": 0.73, "learning_rate": 3.5439679409457826e-06, "loss": 1.8427, "step": 29685 }, { "epoch": 0.73, "learning_rate": 3.543358297566446e-06, "loss": 1.8157, "step": 29686 }, { "epoch": 0.73, "learning_rate": 3.542748695337391e-06, "loss": 1.9983, "step": 29687 }, { "epoch": 0.73, "learning_rate": 3.5421391342625034e-06, "loss": 1.8466, "step": 29688 }, { "epoch": 0.73, "learning_rate": 3.5415296143456677e-06, "loss": 1.9008, "step": 29689 }, { "epoch": 0.73, "learning_rate": 3.5409201355907717e-06, "loss": 1.9485, "step": 29690 }, { "epoch": 0.73, "learning_rate": 3.540310698001693e-06, "loss": 2.0309, "step": 29691 }, { "epoch": 0.73, "learning_rate": 3.5397013015823234e-06, "loss": 1.8267, "step": 29692 }, { "epoch": 0.73, "learning_rate": 3.5390919463365392e-06, "loss": 2.1352, "step": 29693 }, { "epoch": 0.73, "learning_rate": 3.5384826322682263e-06, "loss": 2.0412, "step": 29694 }, { "epoch": 0.73, "learning_rate": 3.537873359381274e-06, "loss": 2.0449, "step": 29695 }, { "epoch": 0.73, "learning_rate": 3.537264127679557e-06, "loss": 1.7204, "step": 29696 }, { "epoch": 0.73, "learning_rate": 3.5366549371669623e-06, "loss": 1.8366, "step": 29697 }, { "epoch": 0.73, "learning_rate": 3.5360457878473754e-06, "loss": 1.8774, "step": 29698 }, { "epoch": 0.73, "learning_rate": 3.535436679724671e-06, "loss": 1.7747, "step": 29699 }, { "epoch": 0.73, "learning_rate": 3.5348276128027393e-06, "loss": 1.8822, "step": 29700 }, { "epoch": 0.73, "learning_rate": 3.534218587085455e-06, "loss": 1.8263, "step": 29701 }, { "epoch": 0.73, "learning_rate": 3.5336096025767018e-06, "loss": 1.7536, "step": 29702 }, { "epoch": 0.73, "learning_rate": 3.533000659280367e-06, "loss": 1.9937, "step": 29703 }, { "epoch": 0.73, "learning_rate": 3.532391757200322e-06, "loss": 2.1429, "step": 29704 }, { "epoch": 0.73, "learning_rate": 3.531782896340452e-06, "loss": 2.0128, "step": 29705 }, { "epoch": 0.73, "learning_rate": 3.5311740767046375e-06, "loss": 1.9664, "step": 29706 }, { "epoch": 0.73, "learning_rate": 3.53056529829676e-06, "loss": 2.0293, "step": 29707 }, { "epoch": 0.73, "learning_rate": 3.5299565611207e-06, "loss": 2.016, "step": 29708 }, { "epoch": 0.73, "learning_rate": 3.5293478651803326e-06, "loss": 2.0547, "step": 29709 }, { "epoch": 0.73, "learning_rate": 3.528739210479544e-06, "loss": 1.925, "step": 29710 }, { "epoch": 0.73, "learning_rate": 3.5281305970222046e-06, "loss": 2.0309, "step": 29711 }, { "epoch": 0.73, "learning_rate": 3.5275220248121987e-06, "loss": 1.984, "step": 29712 }, { "epoch": 0.73, "learning_rate": 3.526913493853408e-06, "loss": 1.9339, "step": 29713 }, { "epoch": 0.73, "learning_rate": 3.526305004149703e-06, "loss": 2.0998, "step": 29714 }, { "epoch": 0.73, "learning_rate": 3.5256965557049693e-06, "loss": 1.9985, "step": 29715 }, { "epoch": 0.73, "learning_rate": 3.5250881485230794e-06, "loss": 1.8252, "step": 29716 }, { "epoch": 0.73, "learning_rate": 3.5244797826079113e-06, "loss": 2.1331, "step": 29717 }, { "epoch": 0.73, "learning_rate": 3.5238714579633483e-06, "loss": 1.9582, "step": 29718 }, { "epoch": 0.73, "learning_rate": 3.5232631745932587e-06, "loss": 1.9436, "step": 29719 }, { "epoch": 0.73, "learning_rate": 3.522654932501527e-06, "loss": 1.9793, "step": 29720 }, { "epoch": 0.73, "learning_rate": 3.522046731692024e-06, "loss": 1.7702, "step": 29721 }, { "epoch": 0.73, "learning_rate": 3.5214385721686273e-06, "loss": 1.8841, "step": 29722 }, { "epoch": 0.73, "learning_rate": 3.5208304539352146e-06, "loss": 1.9486, "step": 29723 }, { "epoch": 0.73, "learning_rate": 3.52022237699566e-06, "loss": 1.7184, "step": 29724 }, { "epoch": 0.73, "learning_rate": 3.519614341353844e-06, "loss": 1.823, "step": 29725 }, { "epoch": 0.73, "learning_rate": 3.5190063470136338e-06, "loss": 1.8967, "step": 29726 }, { "epoch": 0.73, "learning_rate": 3.518398393978909e-06, "loss": 1.913, "step": 29727 }, { "epoch": 0.73, "learning_rate": 3.5177904822535458e-06, "loss": 1.8392, "step": 29728 }, { "epoch": 0.73, "learning_rate": 3.5171826118414143e-06, "loss": 1.9009, "step": 29729 }, { "epoch": 0.73, "learning_rate": 3.5165747827463924e-06, "loss": 1.9021, "step": 29730 }, { "epoch": 0.73, "learning_rate": 3.515966994972351e-06, "loss": 1.9959, "step": 29731 }, { "epoch": 0.73, "learning_rate": 3.5153592485231634e-06, "loss": 1.9477, "step": 29732 }, { "epoch": 0.73, "learning_rate": 3.514751543402708e-06, "loss": 1.9056, "step": 29733 }, { "epoch": 0.73, "learning_rate": 3.514143879614852e-06, "loss": 1.8933, "step": 29734 }, { "epoch": 0.73, "learning_rate": 3.5135362571634736e-06, "loss": 2.0106, "step": 29735 }, { "epoch": 0.73, "learning_rate": 3.5129286760524395e-06, "loss": 2.094, "step": 29736 }, { "epoch": 0.73, "learning_rate": 3.512321136285625e-06, "loss": 1.8748, "step": 29737 }, { "epoch": 0.73, "learning_rate": 3.5117136378669048e-06, "loss": 1.787, "step": 29738 }, { "epoch": 0.73, "learning_rate": 3.5111061808001456e-06, "loss": 2.0098, "step": 29739 }, { "epoch": 0.73, "learning_rate": 3.5104987650892207e-06, "loss": 2.0519, "step": 29740 }, { "epoch": 0.73, "learning_rate": 3.509891390738003e-06, "loss": 1.7898, "step": 29741 }, { "epoch": 0.73, "learning_rate": 3.5092840577503616e-06, "loss": 2.0, "step": 29742 }, { "epoch": 0.73, "learning_rate": 3.5086767661301714e-06, "loss": 1.9825, "step": 29743 }, { "epoch": 0.73, "learning_rate": 3.508069515881297e-06, "loss": 1.9881, "step": 29744 }, { "epoch": 0.73, "learning_rate": 3.507462307007614e-06, "loss": 2.0454, "step": 29745 }, { "epoch": 0.73, "learning_rate": 3.5068551395129867e-06, "loss": 2.0817, "step": 29746 }, { "epoch": 0.73, "learning_rate": 3.5062480134012878e-06, "loss": 2.0127, "step": 29747 }, { "epoch": 0.73, "learning_rate": 3.5056409286763894e-06, "loss": 1.8414, "step": 29748 }, { "epoch": 0.73, "learning_rate": 3.5050338853421538e-06, "loss": 1.8828, "step": 29749 }, { "epoch": 0.73, "learning_rate": 3.5044268834024585e-06, "loss": 1.8896, "step": 29750 }, { "epoch": 0.73, "learning_rate": 3.5038199228611634e-06, "loss": 2.1573, "step": 29751 }, { "epoch": 0.73, "learning_rate": 3.5032130037221414e-06, "loss": 1.9131, "step": 29752 }, { "epoch": 0.73, "learning_rate": 3.502606125989264e-06, "loss": 1.9791, "step": 29753 }, { "epoch": 0.73, "learning_rate": 3.5019992896663913e-06, "loss": 2.0133, "step": 29754 }, { "epoch": 0.73, "learning_rate": 3.501392494757394e-06, "loss": 1.8918, "step": 29755 }, { "epoch": 0.73, "learning_rate": 3.5007857412661452e-06, "loss": 1.95, "step": 29756 }, { "epoch": 0.73, "learning_rate": 3.500179029196502e-06, "loss": 1.984, "step": 29757 }, { "epoch": 0.73, "learning_rate": 3.499572358552337e-06, "loss": 1.9886, "step": 29758 }, { "epoch": 0.73, "learning_rate": 3.4989657293375155e-06, "loss": 1.8522, "step": 29759 }, { "epoch": 0.73, "learning_rate": 3.4983591415559036e-06, "loss": 1.6728, "step": 29760 }, { "epoch": 0.73, "learning_rate": 3.4977525952113723e-06, "loss": 1.8454, "step": 29761 }, { "epoch": 0.73, "learning_rate": 3.497146090307778e-06, "loss": 1.9237, "step": 29762 }, { "epoch": 0.73, "learning_rate": 3.496539626848995e-06, "loss": 1.9298, "step": 29763 }, { "epoch": 0.73, "learning_rate": 3.4959332048388807e-06, "loss": 1.9535, "step": 29764 }, { "epoch": 0.73, "learning_rate": 3.495326824281303e-06, "loss": 1.8759, "step": 29765 }, { "epoch": 0.73, "learning_rate": 3.494720485180132e-06, "loss": 2.0326, "step": 29766 }, { "epoch": 0.73, "learning_rate": 3.4941141875392238e-06, "loss": 1.9496, "step": 29767 }, { "epoch": 0.73, "learning_rate": 3.4935079313624485e-06, "loss": 1.966, "step": 29768 }, { "epoch": 0.73, "learning_rate": 3.492901716653665e-06, "loss": 2.0456, "step": 29769 }, { "epoch": 0.73, "learning_rate": 3.4922955434167394e-06, "loss": 2.0693, "step": 29770 }, { "epoch": 0.73, "learning_rate": 3.4916894116555376e-06, "loss": 1.8862, "step": 29771 }, { "epoch": 0.73, "learning_rate": 3.4910833213739183e-06, "loss": 1.8565, "step": 29772 }, { "epoch": 0.73, "learning_rate": 3.4904772725757486e-06, "loss": 2.0721, "step": 29773 }, { "epoch": 0.73, "learning_rate": 3.489871265264886e-06, "loss": 1.9498, "step": 29774 }, { "epoch": 0.73, "learning_rate": 3.489265299445195e-06, "loss": 2.0777, "step": 29775 }, { "epoch": 0.73, "learning_rate": 3.4886593751205365e-06, "loss": 2.0554, "step": 29776 }, { "epoch": 0.73, "learning_rate": 3.4880534922947763e-06, "loss": 1.9087, "step": 29777 }, { "epoch": 0.73, "learning_rate": 3.4874476509717748e-06, "loss": 1.9494, "step": 29778 }, { "epoch": 0.73, "learning_rate": 3.486841851155389e-06, "loss": 1.9933, "step": 29779 }, { "epoch": 0.73, "learning_rate": 3.4862360928494832e-06, "loss": 1.9023, "step": 29780 }, { "epoch": 0.73, "learning_rate": 3.485630376057921e-06, "loss": 2.0355, "step": 29781 }, { "epoch": 0.73, "learning_rate": 3.485024700784555e-06, "loss": 1.8404, "step": 29782 }, { "epoch": 0.73, "learning_rate": 3.4844190670332544e-06, "loss": 1.9647, "step": 29783 }, { "epoch": 0.73, "learning_rate": 3.48381347480787e-06, "loss": 2.2095, "step": 29784 }, { "epoch": 0.73, "learning_rate": 3.483207924112266e-06, "loss": 2.1481, "step": 29785 }, { "epoch": 0.73, "learning_rate": 3.4826024149503057e-06, "loss": 1.9754, "step": 29786 }, { "epoch": 0.73, "learning_rate": 3.4819969473258408e-06, "loss": 2.1297, "step": 29787 }, { "epoch": 0.73, "learning_rate": 3.4813915212427362e-06, "loss": 2.0167, "step": 29788 }, { "epoch": 0.73, "learning_rate": 3.4807861367048456e-06, "loss": 1.95, "step": 29789 }, { "epoch": 0.73, "learning_rate": 3.480180793716028e-06, "loss": 1.6985, "step": 29790 }, { "epoch": 0.73, "learning_rate": 3.4795754922801474e-06, "loss": 1.9551, "step": 29791 }, { "epoch": 0.73, "learning_rate": 3.4789702324010532e-06, "loss": 1.8119, "step": 29792 }, { "epoch": 0.73, "learning_rate": 3.4783650140826064e-06, "loss": 1.9307, "step": 29793 }, { "epoch": 0.73, "learning_rate": 3.477759837328666e-06, "loss": 2.1194, "step": 29794 }, { "epoch": 0.73, "learning_rate": 3.4771547021430873e-06, "loss": 1.9715, "step": 29795 }, { "epoch": 0.73, "learning_rate": 3.4765496085297302e-06, "loss": 2.026, "step": 29796 }, { "epoch": 0.73, "learning_rate": 3.4759445564924454e-06, "loss": 1.9961, "step": 29797 }, { "epoch": 0.73, "learning_rate": 3.4753395460350948e-06, "loss": 1.8394, "step": 29798 }, { "epoch": 0.73, "learning_rate": 3.474734577161529e-06, "loss": 1.8787, "step": 29799 }, { "epoch": 0.73, "learning_rate": 3.474129649875606e-06, "loss": 1.8863, "step": 29800 }, { "epoch": 0.73, "learning_rate": 3.4735247641811853e-06, "loss": 1.8981, "step": 29801 }, { "epoch": 0.73, "learning_rate": 3.472919920082114e-06, "loss": 1.752, "step": 29802 }, { "epoch": 0.73, "learning_rate": 3.4723151175822546e-06, "loss": 1.9562, "step": 29803 }, { "epoch": 0.73, "learning_rate": 3.471710356685456e-06, "loss": 1.8828, "step": 29804 }, { "epoch": 0.73, "learning_rate": 3.471105637395573e-06, "loss": 2.034, "step": 29805 }, { "epoch": 0.73, "learning_rate": 3.4705009597164665e-06, "loss": 2.0548, "step": 29806 }, { "epoch": 0.73, "learning_rate": 3.469896323651981e-06, "loss": 1.9628, "step": 29807 }, { "epoch": 0.73, "learning_rate": 3.4692917292059746e-06, "loss": 1.9017, "step": 29808 }, { "epoch": 0.73, "learning_rate": 3.4686871763823037e-06, "loss": 1.8611, "step": 29809 }, { "epoch": 0.73, "learning_rate": 3.468082665184814e-06, "loss": 2.0091, "step": 29810 }, { "epoch": 0.73, "learning_rate": 3.467478195617363e-06, "loss": 2.069, "step": 29811 }, { "epoch": 0.73, "learning_rate": 3.4668737676838025e-06, "loss": 2.0739, "step": 29812 }, { "epoch": 0.73, "learning_rate": 3.4662693813879876e-06, "loss": 2.0885, "step": 29813 }, { "epoch": 0.73, "learning_rate": 3.465665036733763e-06, "loss": 1.9063, "step": 29814 }, { "epoch": 0.73, "learning_rate": 3.4650607337249855e-06, "loss": 1.9385, "step": 29815 }, { "epoch": 0.73, "learning_rate": 3.4644564723655092e-06, "loss": 1.993, "step": 29816 }, { "epoch": 0.73, "learning_rate": 3.463852252659178e-06, "loss": 2.0016, "step": 29817 }, { "epoch": 0.73, "learning_rate": 3.463248074609846e-06, "loss": 1.8626, "step": 29818 }, { "epoch": 0.74, "learning_rate": 3.4626439382213683e-06, "loss": 1.9101, "step": 29819 }, { "epoch": 0.74, "learning_rate": 3.4620398434975877e-06, "loss": 1.8591, "step": 29820 }, { "epoch": 0.74, "learning_rate": 3.4614357904423613e-06, "loss": 1.7827, "step": 29821 }, { "epoch": 0.74, "learning_rate": 3.4608317790595313e-06, "loss": 1.9156, "step": 29822 }, { "epoch": 0.74, "learning_rate": 3.460227809352953e-06, "loss": 1.8646, "step": 29823 }, { "epoch": 0.74, "learning_rate": 3.459623881326476e-06, "loss": 2.0721, "step": 29824 }, { "epoch": 0.74, "learning_rate": 3.459019994983944e-06, "loss": 1.663, "step": 29825 }, { "epoch": 0.74, "learning_rate": 3.45841615032921e-06, "loss": 1.9891, "step": 29826 }, { "epoch": 0.74, "learning_rate": 3.457812347366125e-06, "loss": 1.9656, "step": 29827 }, { "epoch": 0.74, "learning_rate": 3.4572085860985304e-06, "loss": 2.0477, "step": 29828 }, { "epoch": 0.74, "learning_rate": 3.4566048665302775e-06, "loss": 1.9186, "step": 29829 }, { "epoch": 0.74, "learning_rate": 3.4560011886652157e-06, "loss": 2.004, "step": 29830 }, { "epoch": 0.74, "learning_rate": 3.4553975525071927e-06, "loss": 2.0491, "step": 29831 }, { "epoch": 0.74, "learning_rate": 3.454793958060051e-06, "loss": 2.1513, "step": 29832 }, { "epoch": 0.74, "learning_rate": 3.4541904053276397e-06, "loss": 1.9165, "step": 29833 }, { "epoch": 0.74, "learning_rate": 3.453586894313811e-06, "loss": 2.0215, "step": 29834 }, { "epoch": 0.74, "learning_rate": 3.4529834250224025e-06, "loss": 1.9166, "step": 29835 }, { "epoch": 0.74, "learning_rate": 3.452379997457267e-06, "loss": 1.9501, "step": 29836 }, { "epoch": 0.74, "learning_rate": 3.451776611622245e-06, "loss": 1.9764, "step": 29837 }, { "epoch": 0.74, "learning_rate": 3.451173267521184e-06, "loss": 1.9529, "step": 29838 }, { "epoch": 0.74, "learning_rate": 3.4505699651579338e-06, "loss": 1.8914, "step": 29839 }, { "epoch": 0.74, "learning_rate": 3.4499667045363314e-06, "loss": 1.9064, "step": 29840 }, { "epoch": 0.74, "learning_rate": 3.449363485660231e-06, "loss": 1.9747, "step": 29841 }, { "epoch": 0.74, "learning_rate": 3.4487603085334664e-06, "loss": 1.957, "step": 29842 }, { "epoch": 0.74, "learning_rate": 3.4481571731598883e-06, "loss": 2.0029, "step": 29843 }, { "epoch": 0.74, "learning_rate": 3.447554079543344e-06, "loss": 1.9345, "step": 29844 }, { "epoch": 0.74, "learning_rate": 3.4469510276876685e-06, "loss": 2.119, "step": 29845 }, { "epoch": 0.74, "learning_rate": 3.446348017596709e-06, "loss": 1.9572, "step": 29846 }, { "epoch": 0.74, "learning_rate": 3.4457450492743105e-06, "loss": 2.026, "step": 29847 }, { "epoch": 0.74, "learning_rate": 3.4451421227243144e-06, "loss": 1.9791, "step": 29848 }, { "epoch": 0.74, "learning_rate": 3.444539237950567e-06, "loss": 2.0078, "step": 29849 }, { "epoch": 0.74, "learning_rate": 3.443936394956905e-06, "loss": 2.0913, "step": 29850 }, { "epoch": 0.74, "learning_rate": 3.4433335937471745e-06, "loss": 1.894, "step": 29851 }, { "epoch": 0.74, "learning_rate": 3.442730834325214e-06, "loss": 2.2244, "step": 29852 }, { "epoch": 0.74, "learning_rate": 3.442128116694866e-06, "loss": 1.8382, "step": 29853 }, { "epoch": 0.74, "learning_rate": 3.4415254408599764e-06, "loss": 1.785, "step": 29854 }, { "epoch": 0.74, "learning_rate": 3.4409228068243794e-06, "loss": 1.9052, "step": 29855 }, { "epoch": 0.74, "learning_rate": 3.440320214591921e-06, "loss": 1.8386, "step": 29856 }, { "epoch": 0.74, "learning_rate": 3.4397176641664377e-06, "loss": 2.0413, "step": 29857 }, { "epoch": 0.74, "learning_rate": 3.439115155551771e-06, "loss": 1.9257, "step": 29858 }, { "epoch": 0.74, "learning_rate": 3.438512688751765e-06, "loss": 1.9502, "step": 29859 }, { "epoch": 0.74, "learning_rate": 3.437910263770252e-06, "loss": 1.7799, "step": 29860 }, { "epoch": 0.74, "learning_rate": 3.437307880611076e-06, "loss": 1.8651, "step": 29861 }, { "epoch": 0.74, "learning_rate": 3.436705539278079e-06, "loss": 1.8791, "step": 29862 }, { "epoch": 0.74, "learning_rate": 3.4361032397750936e-06, "loss": 1.9724, "step": 29863 }, { "epoch": 0.74, "learning_rate": 3.4355009821059604e-06, "loss": 2.0467, "step": 29864 }, { "epoch": 0.74, "learning_rate": 3.434898766274519e-06, "loss": 2.0305, "step": 29865 }, { "epoch": 0.74, "learning_rate": 3.4342965922846118e-06, "loss": 1.8384, "step": 29866 }, { "epoch": 0.74, "learning_rate": 3.4336944601400667e-06, "loss": 1.9617, "step": 29867 }, { "epoch": 0.74, "learning_rate": 3.433092369844728e-06, "loss": 2.0692, "step": 29868 }, { "epoch": 0.74, "learning_rate": 3.432490321402435e-06, "loss": 1.8617, "step": 29869 }, { "epoch": 0.74, "learning_rate": 3.4318883148170178e-06, "loss": 2.0526, "step": 29870 }, { "epoch": 0.74, "learning_rate": 3.431286350092321e-06, "loss": 2.0478, "step": 29871 }, { "epoch": 0.74, "learning_rate": 3.4306844272321726e-06, "loss": 1.8872, "step": 29872 }, { "epoch": 0.74, "learning_rate": 3.430082546240414e-06, "loss": 1.9478, "step": 29873 }, { "epoch": 0.74, "learning_rate": 3.4294807071208834e-06, "loss": 1.8994, "step": 29874 }, { "epoch": 0.74, "learning_rate": 3.42887890987741e-06, "loss": 1.8514, "step": 29875 }, { "epoch": 0.74, "learning_rate": 3.428277154513833e-06, "loss": 1.9316, "step": 29876 }, { "epoch": 0.74, "learning_rate": 3.4276754410339906e-06, "loss": 1.8023, "step": 29877 }, { "epoch": 0.74, "learning_rate": 3.427073769441711e-06, "loss": 1.9148, "step": 29878 }, { "epoch": 0.74, "learning_rate": 3.4264721397408328e-06, "loss": 2.2039, "step": 29879 }, { "epoch": 0.74, "learning_rate": 3.425870551935193e-06, "loss": 2.0397, "step": 29880 }, { "epoch": 0.74, "learning_rate": 3.4252690060286188e-06, "loss": 1.894, "step": 29881 }, { "epoch": 0.74, "learning_rate": 3.4246675020249486e-06, "loss": 1.9875, "step": 29882 }, { "epoch": 0.74, "learning_rate": 3.4240660399280145e-06, "loss": 1.9296, "step": 29883 }, { "epoch": 0.74, "learning_rate": 3.4234646197416544e-06, "loss": 1.9099, "step": 29884 }, { "epoch": 0.74, "learning_rate": 3.422863241469694e-06, "loss": 2.0178, "step": 29885 }, { "epoch": 0.74, "learning_rate": 3.42226190511597e-06, "loss": 1.9384, "step": 29886 }, { "epoch": 0.74, "learning_rate": 3.4216606106843178e-06, "loss": 1.9103, "step": 29887 }, { "epoch": 0.74, "learning_rate": 3.421059358178562e-06, "loss": 1.9768, "step": 29888 }, { "epoch": 0.74, "learning_rate": 3.4204581476025433e-06, "loss": 2.1046, "step": 29889 }, { "epoch": 0.74, "learning_rate": 3.4198569789600857e-06, "loss": 1.7707, "step": 29890 }, { "epoch": 0.74, "learning_rate": 3.4192558522550234e-06, "loss": 1.7877, "step": 29891 }, { "epoch": 0.74, "learning_rate": 3.4186547674911918e-06, "loss": 2.0373, "step": 29892 }, { "epoch": 0.74, "learning_rate": 3.4180537246724143e-06, "loss": 1.8421, "step": 29893 }, { "epoch": 0.74, "learning_rate": 3.4174527238025303e-06, "loss": 2.1039, "step": 29894 }, { "epoch": 0.74, "learning_rate": 3.4168517648853605e-06, "loss": 1.8634, "step": 29895 }, { "epoch": 0.74, "learning_rate": 3.41625084792474e-06, "loss": 1.8202, "step": 29896 }, { "epoch": 0.74, "learning_rate": 3.4156499729244983e-06, "loss": 1.9789, "step": 29897 }, { "epoch": 0.74, "learning_rate": 3.415049139888469e-06, "loss": 1.9433, "step": 29898 }, { "epoch": 0.74, "learning_rate": 3.4144483488204737e-06, "loss": 2.0713, "step": 29899 }, { "epoch": 0.74, "learning_rate": 3.413847599724346e-06, "loss": 2.0077, "step": 29900 }, { "epoch": 0.74, "learning_rate": 3.413246892603913e-06, "loss": 2.0177, "step": 29901 }, { "epoch": 0.74, "learning_rate": 3.412646227463008e-06, "loss": 1.8461, "step": 29902 }, { "epoch": 0.74, "learning_rate": 3.412045604305452e-06, "loss": 1.9483, "step": 29903 }, { "epoch": 0.74, "learning_rate": 3.4114450231350793e-06, "loss": 1.9505, "step": 29904 }, { "epoch": 0.74, "learning_rate": 3.410844483955712e-06, "loss": 2.077, "step": 29905 }, { "epoch": 0.74, "learning_rate": 3.410243986771179e-06, "loss": 1.759, "step": 29906 }, { "epoch": 0.74, "learning_rate": 3.409643531585313e-06, "loss": 1.9571, "step": 29907 }, { "epoch": 0.74, "learning_rate": 3.409043118401932e-06, "loss": 1.8427, "step": 29908 }, { "epoch": 0.74, "learning_rate": 3.4084427472248716e-06, "loss": 1.8661, "step": 29909 }, { "epoch": 0.74, "learning_rate": 3.4078424180579507e-06, "loss": 1.9635, "step": 29910 }, { "epoch": 0.74, "learning_rate": 3.4072421309049974e-06, "loss": 1.9546, "step": 29911 }, { "epoch": 0.74, "learning_rate": 3.4066418857698415e-06, "loss": 2.0259, "step": 29912 }, { "epoch": 0.74, "learning_rate": 3.4060416826563035e-06, "loss": 2.0126, "step": 29913 }, { "epoch": 0.74, "learning_rate": 3.4054415215682102e-06, "loss": 1.9354, "step": 29914 }, { "epoch": 0.74, "learning_rate": 3.4048414025093866e-06, "loss": 1.8264, "step": 29915 }, { "epoch": 0.74, "learning_rate": 3.4042413254836616e-06, "loss": 1.7181, "step": 29916 }, { "epoch": 0.74, "learning_rate": 3.4036412904948536e-06, "loss": 2.0061, "step": 29917 }, { "epoch": 0.74, "learning_rate": 3.4030412975467886e-06, "loss": 1.7955, "step": 29918 }, { "epoch": 0.74, "learning_rate": 3.402441346643295e-06, "loss": 1.829, "step": 29919 }, { "epoch": 0.74, "learning_rate": 3.401841437788189e-06, "loss": 1.9789, "step": 29920 }, { "epoch": 0.74, "learning_rate": 3.4012415709852988e-06, "loss": 1.9254, "step": 29921 }, { "epoch": 0.74, "learning_rate": 3.400641746238449e-06, "loss": 1.9752, "step": 29922 }, { "epoch": 0.74, "learning_rate": 3.400041963551457e-06, "loss": 2.0106, "step": 29923 }, { "epoch": 0.74, "learning_rate": 3.399442222928151e-06, "loss": 2.035, "step": 29924 }, { "epoch": 0.74, "learning_rate": 3.3988425243723477e-06, "loss": 1.9365, "step": 29925 }, { "epoch": 0.74, "learning_rate": 3.398242867887873e-06, "loss": 1.9446, "step": 29926 }, { "epoch": 0.74, "learning_rate": 3.3976432534785507e-06, "loss": 1.8896, "step": 29927 }, { "epoch": 0.74, "learning_rate": 3.3970436811481967e-06, "loss": 2.0366, "step": 29928 }, { "epoch": 0.74, "learning_rate": 3.3964441509006383e-06, "loss": 2.0369, "step": 29929 }, { "epoch": 0.74, "learning_rate": 3.3958446627396903e-06, "loss": 1.9206, "step": 29930 }, { "epoch": 0.74, "learning_rate": 3.395245216669176e-06, "loss": 1.8363, "step": 29931 }, { "epoch": 0.74, "learning_rate": 3.394645812692916e-06, "loss": 1.9361, "step": 29932 }, { "epoch": 0.74, "learning_rate": 3.3940464508147353e-06, "loss": 2.1014, "step": 29933 }, { "epoch": 0.74, "learning_rate": 3.3934471310384455e-06, "loss": 1.8055, "step": 29934 }, { "epoch": 0.74, "learning_rate": 3.39284785336787e-06, "loss": 1.9036, "step": 29935 }, { "epoch": 0.74, "learning_rate": 3.392248617806828e-06, "loss": 2.0081, "step": 29936 }, { "epoch": 0.74, "learning_rate": 3.3916494243591437e-06, "loss": 1.9537, "step": 29937 }, { "epoch": 0.74, "learning_rate": 3.3910502730286267e-06, "loss": 1.9682, "step": 29938 }, { "epoch": 0.74, "learning_rate": 3.3904511638191007e-06, "loss": 1.8805, "step": 29939 }, { "epoch": 0.74, "learning_rate": 3.3898520967343863e-06, "loss": 1.9435, "step": 29940 }, { "epoch": 0.74, "learning_rate": 3.389253071778296e-06, "loss": 2.065, "step": 29941 }, { "epoch": 0.74, "learning_rate": 3.3886540889546525e-06, "loss": 2.1035, "step": 29942 }, { "epoch": 0.74, "learning_rate": 3.3880551482672674e-06, "loss": 1.9432, "step": 29943 }, { "epoch": 0.74, "learning_rate": 3.3874562497199625e-06, "loss": 1.9318, "step": 29944 }, { "epoch": 0.74, "learning_rate": 3.3868573933165583e-06, "loss": 1.8483, "step": 29945 }, { "epoch": 0.74, "learning_rate": 3.3862585790608626e-06, "loss": 1.989, "step": 29946 }, { "epoch": 0.74, "learning_rate": 3.3856598069566993e-06, "loss": 2.0988, "step": 29947 }, { "epoch": 0.74, "learning_rate": 3.3850610770078784e-06, "loss": 2.0078, "step": 29948 }, { "epoch": 0.74, "learning_rate": 3.384462389218219e-06, "loss": 1.9956, "step": 29949 }, { "epoch": 0.74, "learning_rate": 3.383863743591538e-06, "loss": 1.8756, "step": 29950 }, { "epoch": 0.74, "learning_rate": 3.383265140131652e-06, "loss": 1.9964, "step": 29951 }, { "epoch": 0.74, "learning_rate": 3.3826665788423695e-06, "loss": 1.9362, "step": 29952 }, { "epoch": 0.74, "learning_rate": 3.3820680597275103e-06, "loss": 1.9851, "step": 29953 }, { "epoch": 0.74, "learning_rate": 3.381469582790887e-06, "loss": 1.9139, "step": 29954 }, { "epoch": 0.74, "learning_rate": 3.38087114803632e-06, "loss": 1.828, "step": 29955 }, { "epoch": 0.74, "learning_rate": 3.380272755467614e-06, "loss": 1.9124, "step": 29956 }, { "epoch": 0.74, "learning_rate": 3.3796744050885897e-06, "loss": 1.8719, "step": 29957 }, { "epoch": 0.74, "learning_rate": 3.3790760969030558e-06, "loss": 1.9546, "step": 29958 }, { "epoch": 0.74, "learning_rate": 3.378477830914827e-06, "loss": 2.0052, "step": 29959 }, { "epoch": 0.74, "learning_rate": 3.377879607127721e-06, "loss": 2.0174, "step": 29960 }, { "epoch": 0.74, "learning_rate": 3.3772814255455423e-06, "loss": 1.9354, "step": 29961 }, { "epoch": 0.74, "learning_rate": 3.376683286172111e-06, "loss": 1.9901, "step": 29962 }, { "epoch": 0.74, "learning_rate": 3.3760851890112314e-06, "loss": 2.0264, "step": 29963 }, { "epoch": 0.74, "learning_rate": 3.3754871340667206e-06, "loss": 1.8294, "step": 29964 }, { "epoch": 0.74, "learning_rate": 3.374889121342392e-06, "loss": 1.743, "step": 29965 }, { "epoch": 0.74, "learning_rate": 3.3742911508420506e-06, "loss": 1.9684, "step": 29966 }, { "epoch": 0.74, "learning_rate": 3.3736932225695117e-06, "loss": 2.066, "step": 29967 }, { "epoch": 0.74, "learning_rate": 3.3730953365285845e-06, "loss": 1.8766, "step": 29968 }, { "epoch": 0.74, "learning_rate": 3.3724974927230838e-06, "loss": 1.8606, "step": 29969 }, { "epoch": 0.74, "learning_rate": 3.3718996911568134e-06, "loss": 2.1035, "step": 29970 }, { "epoch": 0.74, "learning_rate": 3.371301931833586e-06, "loss": 1.8697, "step": 29971 }, { "epoch": 0.74, "learning_rate": 3.3707042147572157e-06, "loss": 1.9791, "step": 29972 }, { "epoch": 0.74, "learning_rate": 3.3701065399315036e-06, "loss": 1.8698, "step": 29973 }, { "epoch": 0.74, "learning_rate": 3.3695089073602638e-06, "loss": 1.9742, "step": 29974 }, { "epoch": 0.74, "learning_rate": 3.3689113170473076e-06, "loss": 1.9786, "step": 29975 }, { "epoch": 0.74, "learning_rate": 3.3683137689964373e-06, "loss": 1.852, "step": 29976 }, { "epoch": 0.74, "learning_rate": 3.367716263211468e-06, "loss": 1.9629, "step": 29977 }, { "epoch": 0.74, "learning_rate": 3.3671187996962005e-06, "loss": 1.9452, "step": 29978 }, { "epoch": 0.74, "learning_rate": 3.366521378454447e-06, "loss": 2.0753, "step": 29979 }, { "epoch": 0.74, "learning_rate": 3.3659239994900185e-06, "loss": 1.8593, "step": 29980 }, { "epoch": 0.74, "learning_rate": 3.365326662806714e-06, "loss": 2.0129, "step": 29981 }, { "epoch": 0.74, "learning_rate": 3.364729368408348e-06, "loss": 1.8521, "step": 29982 }, { "epoch": 0.74, "learning_rate": 3.3641321162987216e-06, "loss": 1.9449, "step": 29983 }, { "epoch": 0.74, "learning_rate": 3.3635349064816435e-06, "loss": 1.9925, "step": 29984 }, { "epoch": 0.74, "learning_rate": 3.36293773896092e-06, "loss": 1.9025, "step": 29985 }, { "epoch": 0.74, "learning_rate": 3.362340613740358e-06, "loss": 1.8517, "step": 29986 }, { "epoch": 0.74, "learning_rate": 3.3617435308237646e-06, "loss": 1.9681, "step": 29987 }, { "epoch": 0.74, "learning_rate": 3.3611464902149403e-06, "loss": 1.9571, "step": 29988 }, { "epoch": 0.74, "learning_rate": 3.3605494919176928e-06, "loss": 1.985, "step": 29989 }, { "epoch": 0.74, "learning_rate": 3.359952535935831e-06, "loss": 1.98, "step": 29990 }, { "epoch": 0.74, "learning_rate": 3.3593556222731517e-06, "loss": 2.0317, "step": 29991 }, { "epoch": 0.74, "learning_rate": 3.358758750933463e-06, "loss": 2.0248, "step": 29992 }, { "epoch": 0.74, "learning_rate": 3.358161921920573e-06, "loss": 2.0648, "step": 29993 }, { "epoch": 0.74, "learning_rate": 3.3575651352382777e-06, "loss": 1.9616, "step": 29994 }, { "epoch": 0.74, "learning_rate": 3.3569683908903884e-06, "loss": 1.9197, "step": 29995 }, { "epoch": 0.74, "learning_rate": 3.3563716888807008e-06, "loss": 1.967, "step": 29996 }, { "epoch": 0.74, "learning_rate": 3.355775029213021e-06, "loss": 1.9449, "step": 29997 }, { "epoch": 0.74, "learning_rate": 3.3551784118911555e-06, "loss": 1.8203, "step": 29998 }, { "epoch": 0.74, "learning_rate": 3.354581836918901e-06, "loss": 2.0959, "step": 29999 }, { "epoch": 0.74, "learning_rate": 3.3539853043000657e-06, "loss": 1.8409, "step": 30000 }, { "epoch": 0.74, "learning_rate": 3.353388814038444e-06, "loss": 1.9916, "step": 30001 }, { "epoch": 0.74, "learning_rate": 3.3527923661378414e-06, "loss": 2.0536, "step": 30002 }, { "epoch": 0.74, "learning_rate": 3.352195960602059e-06, "loss": 1.9347, "step": 30003 }, { "epoch": 0.74, "learning_rate": 3.3515995974348993e-06, "loss": 1.967, "step": 30004 }, { "epoch": 0.74, "learning_rate": 3.3510032766401644e-06, "loss": 2.0088, "step": 30005 }, { "epoch": 0.74, "learning_rate": 3.3504069982216494e-06, "loss": 1.828, "step": 30006 }, { "epoch": 0.74, "learning_rate": 3.3498107621831577e-06, "loss": 2.0134, "step": 30007 }, { "epoch": 0.74, "learning_rate": 3.3492145685284926e-06, "loss": 1.8224, "step": 30008 }, { "epoch": 0.74, "learning_rate": 3.3486184172614467e-06, "loss": 1.9962, "step": 30009 }, { "epoch": 0.74, "learning_rate": 3.3480223083858277e-06, "loss": 2.0428, "step": 30010 }, { "epoch": 0.74, "learning_rate": 3.347426241905426e-06, "loss": 1.9951, "step": 30011 }, { "epoch": 0.74, "learning_rate": 3.3468302178240454e-06, "loss": 1.9511, "step": 30012 }, { "epoch": 0.74, "learning_rate": 3.3462342361454867e-06, "loss": 1.9398, "step": 30013 }, { "epoch": 0.74, "learning_rate": 3.3456382968735423e-06, "loss": 1.9382, "step": 30014 }, { "epoch": 0.74, "learning_rate": 3.3450424000120175e-06, "loss": 1.9486, "step": 30015 }, { "epoch": 0.74, "learning_rate": 3.3444465455647036e-06, "loss": 1.7827, "step": 30016 }, { "epoch": 0.74, "learning_rate": 3.3438507335354e-06, "loss": 1.9033, "step": 30017 }, { "epoch": 0.74, "learning_rate": 3.3432549639279077e-06, "loss": 1.7656, "step": 30018 }, { "epoch": 0.74, "learning_rate": 3.3426592367460185e-06, "loss": 2.0497, "step": 30019 }, { "epoch": 0.74, "learning_rate": 3.342063551993531e-06, "loss": 1.9738, "step": 30020 }, { "epoch": 0.74, "learning_rate": 3.341467909674242e-06, "loss": 1.9327, "step": 30021 }, { "epoch": 0.74, "learning_rate": 3.340872309791948e-06, "loss": 2.0452, "step": 30022 }, { "epoch": 0.74, "learning_rate": 3.340276752350449e-06, "loss": 2.0393, "step": 30023 }, { "epoch": 0.74, "learning_rate": 3.3396812373535325e-06, "loss": 2.0636, "step": 30024 }, { "epoch": 0.74, "learning_rate": 3.3390857648050023e-06, "loss": 1.9603, "step": 30025 }, { "epoch": 0.74, "learning_rate": 3.338490334708645e-06, "loss": 1.866, "step": 30026 }, { "epoch": 0.74, "learning_rate": 3.3378949470682597e-06, "loss": 1.967, "step": 30027 }, { "epoch": 0.74, "learning_rate": 3.337299601887646e-06, "loss": 1.9138, "step": 30028 }, { "epoch": 0.74, "learning_rate": 3.336704299170588e-06, "loss": 1.8231, "step": 30029 }, { "epoch": 0.74, "learning_rate": 3.3361090389208884e-06, "loss": 2.135, "step": 30030 }, { "epoch": 0.74, "learning_rate": 3.3355138211423356e-06, "loss": 1.9956, "step": 30031 }, { "epoch": 0.74, "learning_rate": 3.3349186458387238e-06, "loss": 1.8975, "step": 30032 }, { "epoch": 0.74, "learning_rate": 3.334323513013852e-06, "loss": 2.132, "step": 30033 }, { "epoch": 0.74, "learning_rate": 3.333728422671504e-06, "loss": 1.9815, "step": 30034 }, { "epoch": 0.74, "learning_rate": 3.3331333748154814e-06, "loss": 1.8546, "step": 30035 }, { "epoch": 0.74, "learning_rate": 3.3325383694495695e-06, "loss": 1.963, "step": 30036 }, { "epoch": 0.74, "learning_rate": 3.3319434065775625e-06, "loss": 1.8596, "step": 30037 }, { "epoch": 0.74, "learning_rate": 3.331348486203254e-06, "loss": 1.8924, "step": 30038 }, { "epoch": 0.74, "learning_rate": 3.330753608330435e-06, "loss": 1.9993, "step": 30039 }, { "epoch": 0.74, "learning_rate": 3.3301587729628993e-06, "loss": 1.9865, "step": 30040 }, { "epoch": 0.74, "learning_rate": 3.329563980104432e-06, "loss": 1.8558, "step": 30041 }, { "epoch": 0.74, "learning_rate": 3.3289692297588283e-06, "loss": 2.0651, "step": 30042 }, { "epoch": 0.74, "learning_rate": 3.3283745219298804e-06, "loss": 1.9473, "step": 30043 }, { "epoch": 0.74, "learning_rate": 3.3277798566213727e-06, "loss": 1.7852, "step": 30044 }, { "epoch": 0.74, "learning_rate": 3.3271852338371013e-06, "loss": 2.0253, "step": 30045 }, { "epoch": 0.74, "learning_rate": 3.3265906535808502e-06, "loss": 1.8752, "step": 30046 }, { "epoch": 0.74, "learning_rate": 3.3259961158564113e-06, "loss": 1.8991, "step": 30047 }, { "epoch": 0.74, "learning_rate": 3.3254016206675775e-06, "loss": 1.8716, "step": 30048 }, { "epoch": 0.74, "learning_rate": 3.3248071680181317e-06, "loss": 2.0187, "step": 30049 }, { "epoch": 0.74, "learning_rate": 3.3242127579118677e-06, "loss": 1.9618, "step": 30050 }, { "epoch": 0.74, "learning_rate": 3.323618390352569e-06, "loss": 1.9993, "step": 30051 }, { "epoch": 0.74, "learning_rate": 3.3230240653440247e-06, "loss": 1.9484, "step": 30052 }, { "epoch": 0.74, "learning_rate": 3.3224297828900287e-06, "loss": 1.9685, "step": 30053 }, { "epoch": 0.74, "learning_rate": 3.3218355429943596e-06, "loss": 1.7956, "step": 30054 }, { "epoch": 0.74, "learning_rate": 3.3212413456608094e-06, "loss": 2.0118, "step": 30055 }, { "epoch": 0.74, "learning_rate": 3.3206471908931635e-06, "loss": 2.0359, "step": 30056 }, { "epoch": 0.74, "learning_rate": 3.3200530786952112e-06, "loss": 1.9076, "step": 30057 }, { "epoch": 0.74, "learning_rate": 3.31945900907074e-06, "loss": 1.8032, "step": 30058 }, { "epoch": 0.74, "learning_rate": 3.3188649820235318e-06, "loss": 1.9776, "step": 30059 }, { "epoch": 0.74, "learning_rate": 3.3182709975573733e-06, "loss": 1.9323, "step": 30060 }, { "epoch": 0.74, "learning_rate": 3.317677055676054e-06, "loss": 1.9191, "step": 30061 }, { "epoch": 0.74, "learning_rate": 3.317083156383354e-06, "loss": 1.9523, "step": 30062 }, { "epoch": 0.74, "learning_rate": 3.3164892996830645e-06, "loss": 1.8451, "step": 30063 }, { "epoch": 0.74, "learning_rate": 3.3158954855789626e-06, "loss": 2.0339, "step": 30064 }, { "epoch": 0.74, "learning_rate": 3.315301714074838e-06, "loss": 1.9545, "step": 30065 }, { "epoch": 0.74, "learning_rate": 3.3147079851744768e-06, "loss": 1.834, "step": 30066 }, { "epoch": 0.74, "learning_rate": 3.3141142988816564e-06, "loss": 1.9202, "step": 30067 }, { "epoch": 0.74, "learning_rate": 3.3135206552001685e-06, "loss": 1.8294, "step": 30068 }, { "epoch": 0.74, "learning_rate": 3.31292705413379e-06, "loss": 1.9122, "step": 30069 }, { "epoch": 0.74, "learning_rate": 3.312333495686305e-06, "loss": 2.0002, "step": 30070 }, { "epoch": 0.74, "learning_rate": 3.3117399798615022e-06, "loss": 1.9675, "step": 30071 }, { "epoch": 0.74, "learning_rate": 3.3111465066631576e-06, "loss": 2.0195, "step": 30072 }, { "epoch": 0.74, "learning_rate": 3.310553076095054e-06, "loss": 1.8225, "step": 30073 }, { "epoch": 0.74, "learning_rate": 3.3099596881609773e-06, "loss": 1.955, "step": 30074 }, { "epoch": 0.74, "learning_rate": 3.309366342864708e-06, "loss": 1.8187, "step": 30075 }, { "epoch": 0.74, "learning_rate": 3.3087730402100303e-06, "loss": 1.8167, "step": 30076 }, { "epoch": 0.74, "learning_rate": 3.3081797802007187e-06, "loss": 1.9451, "step": 30077 }, { "epoch": 0.74, "learning_rate": 3.3075865628405625e-06, "loss": 1.8031, "step": 30078 }, { "epoch": 0.74, "learning_rate": 3.306993388133334e-06, "loss": 1.8746, "step": 30079 }, { "epoch": 0.74, "learning_rate": 3.3064002560828177e-06, "loss": 1.996, "step": 30080 }, { "epoch": 0.74, "learning_rate": 3.305807166692798e-06, "loss": 2.0119, "step": 30081 }, { "epoch": 0.74, "learning_rate": 3.305214119967047e-06, "loss": 1.9292, "step": 30082 }, { "epoch": 0.74, "learning_rate": 3.3046211159093523e-06, "loss": 2.1216, "step": 30083 }, { "epoch": 0.74, "learning_rate": 3.3040281545234853e-06, "loss": 1.8243, "step": 30084 }, { "epoch": 0.74, "learning_rate": 3.3034352358132283e-06, "loss": 1.8637, "step": 30085 }, { "epoch": 0.74, "learning_rate": 3.3028423597823654e-06, "loss": 1.8651, "step": 30086 }, { "epoch": 0.74, "learning_rate": 3.3022495264346666e-06, "loss": 1.7846, "step": 30087 }, { "epoch": 0.74, "learning_rate": 3.3016567357739183e-06, "loss": 1.9072, "step": 30088 }, { "epoch": 0.74, "learning_rate": 3.3010639878038897e-06, "loss": 2.0386, "step": 30089 }, { "epoch": 0.74, "learning_rate": 3.3004712825283645e-06, "loss": 2.0099, "step": 30090 }, { "epoch": 0.74, "learning_rate": 3.29987861995112e-06, "loss": 1.9006, "step": 30091 }, { "epoch": 0.74, "learning_rate": 3.2992860000759318e-06, "loss": 2.0669, "step": 30092 }, { "epoch": 0.74, "learning_rate": 3.29869342290658e-06, "loss": 2.1168, "step": 30093 }, { "epoch": 0.74, "learning_rate": 3.298100888446837e-06, "loss": 1.9009, "step": 30094 }, { "epoch": 0.74, "learning_rate": 3.2975083967004807e-06, "loss": 1.9552, "step": 30095 }, { "epoch": 0.74, "learning_rate": 3.296915947671291e-06, "loss": 2.0095, "step": 30096 }, { "epoch": 0.74, "learning_rate": 3.296323541363037e-06, "loss": 1.9877, "step": 30097 }, { "epoch": 0.74, "learning_rate": 3.295731177779501e-06, "loss": 2.0334, "step": 30098 }, { "epoch": 0.74, "learning_rate": 3.2951388569244513e-06, "loss": 1.9939, "step": 30099 }, { "epoch": 0.74, "learning_rate": 3.2945465788016673e-06, "loss": 1.9629, "step": 30100 }, { "epoch": 0.74, "learning_rate": 3.2939543434149257e-06, "loss": 1.907, "step": 30101 }, { "epoch": 0.74, "learning_rate": 3.2933621507679957e-06, "loss": 2.0757, "step": 30102 }, { "epoch": 0.74, "learning_rate": 3.292770000864658e-06, "loss": 2.0743, "step": 30103 }, { "epoch": 0.74, "learning_rate": 3.29217789370868e-06, "loss": 1.8912, "step": 30104 }, { "epoch": 0.74, "learning_rate": 3.291585829303837e-06, "loss": 1.9615, "step": 30105 }, { "epoch": 0.74, "learning_rate": 3.290993807653908e-06, "loss": 1.8794, "step": 30106 }, { "epoch": 0.74, "learning_rate": 3.290401828762658e-06, "loss": 1.8909, "step": 30107 }, { "epoch": 0.74, "learning_rate": 3.289809892633865e-06, "loss": 1.9771, "step": 30108 }, { "epoch": 0.74, "learning_rate": 3.289217999271299e-06, "loss": 1.8566, "step": 30109 }, { "epoch": 0.74, "learning_rate": 3.288626148678734e-06, "loss": 1.9623, "step": 30110 }, { "epoch": 0.74, "learning_rate": 3.288034340859945e-06, "loss": 1.9449, "step": 30111 }, { "epoch": 0.74, "learning_rate": 3.287442575818698e-06, "loss": 1.9348, "step": 30112 }, { "epoch": 0.74, "learning_rate": 3.286850853558766e-06, "loss": 1.9845, "step": 30113 }, { "epoch": 0.74, "learning_rate": 3.2862591740839246e-06, "loss": 1.929, "step": 30114 }, { "epoch": 0.74, "learning_rate": 3.2856675373979386e-06, "loss": 1.9798, "step": 30115 }, { "epoch": 0.74, "learning_rate": 3.285075943504584e-06, "loss": 1.9037, "step": 30116 }, { "epoch": 0.74, "learning_rate": 3.2844843924076263e-06, "loss": 1.958, "step": 30117 }, { "epoch": 0.74, "learning_rate": 3.2838928841108385e-06, "loss": 1.798, "step": 30118 }, { "epoch": 0.74, "learning_rate": 3.2833014186179925e-06, "loss": 1.9425, "step": 30119 }, { "epoch": 0.74, "learning_rate": 3.2827099959328525e-06, "loss": 1.8576, "step": 30120 }, { "epoch": 0.74, "learning_rate": 3.2821186160591933e-06, "loss": 1.8782, "step": 30121 }, { "epoch": 0.74, "learning_rate": 3.281527279000779e-06, "loss": 1.9095, "step": 30122 }, { "epoch": 0.74, "learning_rate": 3.280935984761381e-06, "loss": 1.8327, "step": 30123 }, { "epoch": 0.74, "learning_rate": 3.2803447333447713e-06, "loss": 1.8268, "step": 30124 }, { "epoch": 0.74, "learning_rate": 3.2797535247547096e-06, "loss": 1.9852, "step": 30125 }, { "epoch": 0.74, "learning_rate": 3.2791623589949706e-06, "loss": 1.9194, "step": 30126 }, { "epoch": 0.74, "learning_rate": 3.2785712360693187e-06, "loss": 1.9692, "step": 30127 }, { "epoch": 0.74, "learning_rate": 3.277980155981524e-06, "loss": 2.0439, "step": 30128 }, { "epoch": 0.74, "learning_rate": 3.2773891187353545e-06, "loss": 1.9829, "step": 30129 }, { "epoch": 0.74, "learning_rate": 3.2767981243345726e-06, "loss": 1.8361, "step": 30130 }, { "epoch": 0.74, "learning_rate": 3.2762071727829492e-06, "loss": 1.8515, "step": 30131 }, { "epoch": 0.74, "learning_rate": 3.275616264084247e-06, "loss": 1.8302, "step": 30132 }, { "epoch": 0.74, "learning_rate": 3.2750253982422332e-06, "loss": 1.8414, "step": 30133 }, { "epoch": 0.74, "learning_rate": 3.2744345752606766e-06, "loss": 1.9399, "step": 30134 }, { "epoch": 0.74, "learning_rate": 3.273843795143338e-06, "loss": 1.9746, "step": 30135 }, { "epoch": 0.74, "learning_rate": 3.273253057893988e-06, "loss": 1.8783, "step": 30136 }, { "epoch": 0.74, "learning_rate": 3.2726623635163856e-06, "loss": 1.9694, "step": 30137 }, { "epoch": 0.74, "learning_rate": 3.2720717120142973e-06, "loss": 1.9633, "step": 30138 }, { "epoch": 0.74, "learning_rate": 3.271481103391493e-06, "loss": 2.142, "step": 30139 }, { "epoch": 0.74, "learning_rate": 3.2708905376517277e-06, "loss": 1.9116, "step": 30140 }, { "epoch": 0.74, "learning_rate": 3.270300014798774e-06, "loss": 1.9242, "step": 30141 }, { "epoch": 0.74, "learning_rate": 3.269709534836388e-06, "loss": 2.0374, "step": 30142 }, { "epoch": 0.74, "learning_rate": 3.2691190977683375e-06, "loss": 1.9903, "step": 30143 }, { "epoch": 0.74, "learning_rate": 3.2685287035983828e-06, "loss": 1.8913, "step": 30144 }, { "epoch": 0.74, "learning_rate": 3.2679383523302898e-06, "loss": 1.8857, "step": 30145 }, { "epoch": 0.74, "learning_rate": 3.267348043967822e-06, "loss": 2.1158, "step": 30146 }, { "epoch": 0.74, "learning_rate": 3.2667577785147366e-06, "loss": 1.9072, "step": 30147 }, { "epoch": 0.74, "learning_rate": 3.2661675559747976e-06, "loss": 1.9485, "step": 30148 }, { "epoch": 0.74, "learning_rate": 3.2655773763517705e-06, "loss": 2.1167, "step": 30149 }, { "epoch": 0.74, "learning_rate": 3.264987239649411e-06, "loss": 1.7891, "step": 30150 }, { "epoch": 0.74, "learning_rate": 3.2643971458714853e-06, "loss": 1.877, "step": 30151 }, { "epoch": 0.74, "learning_rate": 3.263807095021748e-06, "loss": 1.9471, "step": 30152 }, { "epoch": 0.74, "learning_rate": 3.2632170871039638e-06, "loss": 1.999, "step": 30153 }, { "epoch": 0.74, "learning_rate": 3.2626271221218963e-06, "loss": 1.9596, "step": 30154 }, { "epoch": 0.74, "learning_rate": 3.2620372000792976e-06, "loss": 1.6772, "step": 30155 }, { "epoch": 0.74, "learning_rate": 3.2614473209799357e-06, "loss": 2.0691, "step": 30156 }, { "epoch": 0.74, "learning_rate": 3.260857484827563e-06, "loss": 1.8867, "step": 30157 }, { "epoch": 0.74, "learning_rate": 3.2602676916259414e-06, "loss": 1.7971, "step": 30158 }, { "epoch": 0.74, "learning_rate": 3.2596779413788328e-06, "loss": 1.9155, "step": 30159 }, { "epoch": 0.74, "learning_rate": 3.259088234089991e-06, "loss": 1.8375, "step": 30160 }, { "epoch": 0.74, "learning_rate": 3.2584985697631767e-06, "loss": 1.7982, "step": 30161 }, { "epoch": 0.74, "learning_rate": 3.2579089484021476e-06, "loss": 2.0212, "step": 30162 }, { "epoch": 0.74, "learning_rate": 3.257319370010661e-06, "loss": 2.0201, "step": 30163 }, { "epoch": 0.74, "learning_rate": 3.25672983459248e-06, "loss": 1.9241, "step": 30164 }, { "epoch": 0.74, "learning_rate": 3.256140342151354e-06, "loss": 2.1876, "step": 30165 }, { "epoch": 0.74, "learning_rate": 3.2555508926910463e-06, "loss": 2.0672, "step": 30166 }, { "epoch": 0.74, "learning_rate": 3.2549614862153077e-06, "loss": 1.9219, "step": 30167 }, { "epoch": 0.74, "learning_rate": 3.254372122727897e-06, "loss": 1.7553, "step": 30168 }, { "epoch": 0.74, "learning_rate": 3.2537828022325747e-06, "loss": 1.9567, "step": 30169 }, { "epoch": 0.74, "learning_rate": 3.2531935247330903e-06, "loss": 1.7205, "step": 30170 }, { "epoch": 0.74, "learning_rate": 3.2526042902332022e-06, "loss": 1.9847, "step": 30171 }, { "epoch": 0.74, "learning_rate": 3.2520150987366682e-06, "loss": 2.0563, "step": 30172 }, { "epoch": 0.74, "learning_rate": 3.251425950247239e-06, "loss": 2.0867, "step": 30173 }, { "epoch": 0.74, "learning_rate": 3.2508368447686745e-06, "loss": 1.9747, "step": 30174 }, { "epoch": 0.74, "learning_rate": 3.2502477823047228e-06, "loss": 1.8498, "step": 30175 }, { "epoch": 0.74, "learning_rate": 3.2496587628591414e-06, "loss": 1.7808, "step": 30176 }, { "epoch": 0.74, "learning_rate": 3.249069786435689e-06, "loss": 1.925, "step": 30177 }, { "epoch": 0.74, "learning_rate": 3.2484808530381108e-06, "loss": 1.9624, "step": 30178 }, { "epoch": 0.74, "learning_rate": 3.2478919626701646e-06, "loss": 1.927, "step": 30179 }, { "epoch": 0.74, "learning_rate": 3.2473031153356027e-06, "loss": 2.0477, "step": 30180 }, { "epoch": 0.74, "learning_rate": 3.2467143110381793e-06, "loss": 1.8999, "step": 30181 }, { "epoch": 0.74, "learning_rate": 3.246125549781649e-06, "loss": 1.8961, "step": 30182 }, { "epoch": 0.74, "learning_rate": 3.245536831569759e-06, "loss": 1.9104, "step": 30183 }, { "epoch": 0.74, "learning_rate": 3.244948156406267e-06, "loss": 1.9349, "step": 30184 }, { "epoch": 0.74, "learning_rate": 3.244359524294919e-06, "loss": 1.8221, "step": 30185 }, { "epoch": 0.74, "learning_rate": 3.243770935239469e-06, "loss": 1.9437, "step": 30186 }, { "epoch": 0.74, "learning_rate": 3.243182389243672e-06, "loss": 1.9182, "step": 30187 }, { "epoch": 0.74, "learning_rate": 3.2425938863112716e-06, "loss": 1.9468, "step": 30188 }, { "epoch": 0.74, "learning_rate": 3.2420054264460264e-06, "loss": 2.0807, "step": 30189 }, { "epoch": 0.74, "learning_rate": 3.2414170096516807e-06, "loss": 1.8319, "step": 30190 }, { "epoch": 0.74, "learning_rate": 3.2408286359319863e-06, "loss": 2.0262, "step": 30191 }, { "epoch": 0.74, "learning_rate": 3.240240305290696e-06, "loss": 1.8431, "step": 30192 }, { "epoch": 0.74, "learning_rate": 3.239652017731555e-06, "loss": 1.8126, "step": 30193 }, { "epoch": 0.74, "learning_rate": 3.2390637732583173e-06, "loss": 1.9636, "step": 30194 }, { "epoch": 0.74, "learning_rate": 3.2384755718747275e-06, "loss": 1.9794, "step": 30195 }, { "epoch": 0.74, "learning_rate": 3.2378874135845352e-06, "loss": 1.9716, "step": 30196 }, { "epoch": 0.74, "learning_rate": 3.2372992983914907e-06, "loss": 2.0389, "step": 30197 }, { "epoch": 0.74, "learning_rate": 3.2367112262993416e-06, "loss": 1.9697, "step": 30198 }, { "epoch": 0.74, "learning_rate": 3.236123197311839e-06, "loss": 1.9055, "step": 30199 }, { "epoch": 0.74, "learning_rate": 3.2355352114327243e-06, "loss": 2.0081, "step": 30200 }, { "epoch": 0.74, "learning_rate": 3.234947268665748e-06, "loss": 1.9822, "step": 30201 }, { "epoch": 0.74, "learning_rate": 3.2343593690146615e-06, "loss": 1.8773, "step": 30202 }, { "epoch": 0.74, "learning_rate": 3.2337715124832038e-06, "loss": 1.9669, "step": 30203 }, { "epoch": 0.74, "learning_rate": 3.2331836990751277e-06, "loss": 2.0192, "step": 30204 }, { "epoch": 0.74, "learning_rate": 3.2325959287941754e-06, "loss": 1.8823, "step": 30205 }, { "epoch": 0.74, "learning_rate": 3.2320082016440933e-06, "loss": 1.8748, "step": 30206 }, { "epoch": 0.74, "learning_rate": 3.231420517628633e-06, "loss": 1.9601, "step": 30207 }, { "epoch": 0.74, "learning_rate": 3.230832876751532e-06, "loss": 1.9741, "step": 30208 }, { "epoch": 0.74, "learning_rate": 3.2302452790165417e-06, "loss": 1.7701, "step": 30209 }, { "epoch": 0.74, "learning_rate": 3.229657724427402e-06, "loss": 1.9103, "step": 30210 }, { "epoch": 0.74, "learning_rate": 3.2290702129878595e-06, "loss": 1.9811, "step": 30211 }, { "epoch": 0.74, "learning_rate": 3.228482744701662e-06, "loss": 1.8512, "step": 30212 }, { "epoch": 0.74, "learning_rate": 3.2278953195725482e-06, "loss": 2.0019, "step": 30213 }, { "epoch": 0.74, "learning_rate": 3.2273079376042647e-06, "loss": 1.9779, "step": 30214 }, { "epoch": 0.74, "learning_rate": 3.2267205988005545e-06, "loss": 1.9572, "step": 30215 }, { "epoch": 0.74, "learning_rate": 3.226133303165161e-06, "loss": 2.0381, "step": 30216 }, { "epoch": 0.74, "learning_rate": 3.225546050701831e-06, "loss": 1.9408, "step": 30217 }, { "epoch": 0.74, "learning_rate": 3.224958841414301e-06, "loss": 1.8067, "step": 30218 }, { "epoch": 0.74, "learning_rate": 3.2243716753063183e-06, "loss": 1.8134, "step": 30219 }, { "epoch": 0.74, "learning_rate": 3.2237845523816203e-06, "loss": 1.8666, "step": 30220 }, { "epoch": 0.74, "learning_rate": 3.2231974726439518e-06, "loss": 1.9921, "step": 30221 }, { "epoch": 0.74, "learning_rate": 3.2226104360970577e-06, "loss": 2.1083, "step": 30222 }, { "epoch": 0.74, "learning_rate": 3.2220234427446728e-06, "loss": 1.8072, "step": 30223 }, { "epoch": 0.74, "learning_rate": 3.2214364925905438e-06, "loss": 1.9411, "step": 30224 }, { "epoch": 0.75, "learning_rate": 3.220849585638407e-06, "loss": 1.8161, "step": 30225 }, { "epoch": 0.75, "learning_rate": 3.2202627218920046e-06, "loss": 2.0844, "step": 30226 }, { "epoch": 0.75, "learning_rate": 3.2196759013550804e-06, "loss": 2.0655, "step": 30227 }, { "epoch": 0.75, "learning_rate": 3.2190891240313683e-06, "loss": 1.8682, "step": 30228 }, { "epoch": 0.75, "learning_rate": 3.2185023899246104e-06, "loss": 1.9713, "step": 30229 }, { "epoch": 0.75, "learning_rate": 3.2179156990385506e-06, "loss": 1.9304, "step": 30230 }, { "epoch": 0.75, "learning_rate": 3.2173290513769205e-06, "loss": 1.9381, "step": 30231 }, { "epoch": 0.75, "learning_rate": 3.216742446943464e-06, "loss": 1.9162, "step": 30232 }, { "epoch": 0.75, "learning_rate": 3.216155885741917e-06, "loss": 1.9357, "step": 30233 }, { "epoch": 0.75, "learning_rate": 3.2155693677760204e-06, "loss": 1.8799, "step": 30234 }, { "epoch": 0.75, "learning_rate": 3.214982893049514e-06, "loss": 1.9464, "step": 30235 }, { "epoch": 0.75, "learning_rate": 3.2143964615661296e-06, "loss": 1.9714, "step": 30236 }, { "epoch": 0.75, "learning_rate": 3.213810073329612e-06, "loss": 1.8244, "step": 30237 }, { "epoch": 0.75, "learning_rate": 3.2132237283436896e-06, "loss": 1.8623, "step": 30238 }, { "epoch": 0.75, "learning_rate": 3.2126374266121054e-06, "loss": 1.953, "step": 30239 }, { "epoch": 0.75, "learning_rate": 3.212051168138597e-06, "loss": 1.7915, "step": 30240 }, { "epoch": 0.75, "learning_rate": 3.211464952926896e-06, "loss": 2.0144, "step": 30241 }, { "epoch": 0.75, "learning_rate": 3.210878780980744e-06, "loss": 1.9074, "step": 30242 }, { "epoch": 0.75, "learning_rate": 3.2102926523038714e-06, "loss": 1.9384, "step": 30243 }, { "epoch": 0.75, "learning_rate": 3.209706566900016e-06, "loss": 1.8967, "step": 30244 }, { "epoch": 0.75, "learning_rate": 3.2091205247729174e-06, "loss": 1.811, "step": 30245 }, { "epoch": 0.75, "learning_rate": 3.208534525926302e-06, "loss": 1.86, "step": 30246 }, { "epoch": 0.75, "learning_rate": 3.2079485703639146e-06, "loss": 1.8749, "step": 30247 }, { "epoch": 0.75, "learning_rate": 3.20736265808948e-06, "loss": 1.9605, "step": 30248 }, { "epoch": 0.75, "learning_rate": 3.206776789106737e-06, "loss": 1.8828, "step": 30249 }, { "epoch": 0.75, "learning_rate": 3.20619096341942e-06, "loss": 1.8497, "step": 30250 }, { "epoch": 0.75, "learning_rate": 3.2056051810312605e-06, "loss": 2.1282, "step": 30251 }, { "epoch": 0.75, "learning_rate": 3.205019441945998e-06, "loss": 1.904, "step": 30252 }, { "epoch": 0.75, "learning_rate": 3.2044337461673568e-06, "loss": 1.9896, "step": 30253 }, { "epoch": 0.75, "learning_rate": 3.2038480936990758e-06, "loss": 2.0488, "step": 30254 }, { "epoch": 0.75, "learning_rate": 3.2032624845448867e-06, "loss": 1.8226, "step": 30255 }, { "epoch": 0.75, "learning_rate": 3.2026769187085195e-06, "loss": 1.8915, "step": 30256 }, { "epoch": 0.75, "learning_rate": 3.2020913961937096e-06, "loss": 2.0854, "step": 30257 }, { "epoch": 0.75, "learning_rate": 3.201505917004183e-06, "loss": 1.9974, "step": 30258 }, { "epoch": 0.75, "learning_rate": 3.2009204811436756e-06, "loss": 1.8745, "step": 30259 }, { "epoch": 0.75, "learning_rate": 3.2003350886159203e-06, "loss": 2.1807, "step": 30260 }, { "epoch": 0.75, "learning_rate": 3.1997497394246423e-06, "loss": 2.0883, "step": 30261 }, { "epoch": 0.75, "learning_rate": 3.1991644335735792e-06, "loss": 1.9775, "step": 30262 }, { "epoch": 0.75, "learning_rate": 3.1985791710664537e-06, "loss": 1.8978, "step": 30263 }, { "epoch": 0.75, "learning_rate": 3.1979939519070002e-06, "loss": 1.9822, "step": 30264 }, { "epoch": 0.75, "learning_rate": 3.19740877609895e-06, "loss": 1.9345, "step": 30265 }, { "epoch": 0.75, "learning_rate": 3.1968236436460286e-06, "loss": 1.9476, "step": 30266 }, { "epoch": 0.75, "learning_rate": 3.1962385545519668e-06, "loss": 1.8607, "step": 30267 }, { "epoch": 0.75, "learning_rate": 3.1956535088204942e-06, "loss": 1.8649, "step": 30268 }, { "epoch": 0.75, "learning_rate": 3.1950685064553387e-06, "loss": 2.0337, "step": 30269 }, { "epoch": 0.75, "learning_rate": 3.194483547460233e-06, "loss": 1.9023, "step": 30270 }, { "epoch": 0.75, "learning_rate": 3.1938986318388975e-06, "loss": 1.8982, "step": 30271 }, { "epoch": 0.75, "learning_rate": 3.193313759595068e-06, "loss": 2.0144, "step": 30272 }, { "epoch": 0.75, "learning_rate": 3.192728930732465e-06, "loss": 1.974, "step": 30273 }, { "epoch": 0.75, "learning_rate": 3.1921441452548187e-06, "loss": 1.9816, "step": 30274 }, { "epoch": 0.75, "learning_rate": 3.191559403165859e-06, "loss": 1.8297, "step": 30275 }, { "epoch": 0.75, "learning_rate": 3.1909747044693085e-06, "loss": 2.0456, "step": 30276 }, { "epoch": 0.75, "learning_rate": 3.1903900491688965e-06, "loss": 2.0263, "step": 30277 }, { "epoch": 0.75, "learning_rate": 3.189805437268346e-06, "loss": 1.8949, "step": 30278 }, { "epoch": 0.75, "learning_rate": 3.1892208687713842e-06, "loss": 1.9248, "step": 30279 }, { "epoch": 0.75, "learning_rate": 3.1886363436817413e-06, "loss": 1.9568, "step": 30280 }, { "epoch": 0.75, "learning_rate": 3.188051862003134e-06, "loss": 1.9158, "step": 30281 }, { "epoch": 0.75, "learning_rate": 3.1874674237392968e-06, "loss": 2.0675, "step": 30282 }, { "epoch": 0.75, "learning_rate": 3.186883028893947e-06, "loss": 2.0857, "step": 30283 }, { "epoch": 0.75, "learning_rate": 3.1862986774708106e-06, "loss": 1.9337, "step": 30284 }, { "epoch": 0.75, "learning_rate": 3.1857143694736136e-06, "loss": 1.9062, "step": 30285 }, { "epoch": 0.75, "learning_rate": 3.18513010490608e-06, "loss": 1.7981, "step": 30286 }, { "epoch": 0.75, "learning_rate": 3.1845458837719366e-06, "loss": 1.9404, "step": 30287 }, { "epoch": 0.75, "learning_rate": 3.1839617060748994e-06, "loss": 1.8765, "step": 30288 }, { "epoch": 0.75, "learning_rate": 3.1833775718186954e-06, "loss": 1.7427, "step": 30289 }, { "epoch": 0.75, "learning_rate": 3.1827934810070513e-06, "loss": 1.9631, "step": 30290 }, { "epoch": 0.75, "learning_rate": 3.1822094336436836e-06, "loss": 2.0043, "step": 30291 }, { "epoch": 0.75, "learning_rate": 3.181625429732316e-06, "loss": 1.9589, "step": 30292 }, { "epoch": 0.75, "learning_rate": 3.1810414692766754e-06, "loss": 1.8462, "step": 30293 }, { "epoch": 0.75, "learning_rate": 3.180457552280476e-06, "loss": 1.9901, "step": 30294 }, { "epoch": 0.75, "learning_rate": 3.179873678747446e-06, "loss": 1.8279, "step": 30295 }, { "epoch": 0.75, "learning_rate": 3.179289848681302e-06, "loss": 1.8198, "step": 30296 }, { "epoch": 0.75, "learning_rate": 3.1787060620857645e-06, "loss": 2.0064, "step": 30297 }, { "epoch": 0.75, "learning_rate": 3.1781223189645604e-06, "loss": 2.0908, "step": 30298 }, { "epoch": 0.75, "learning_rate": 3.177538619321403e-06, "loss": 2.0636, "step": 30299 }, { "epoch": 0.75, "learning_rate": 3.176954963160018e-06, "loss": 1.7141, "step": 30300 }, { "epoch": 0.75, "learning_rate": 3.17637135048412e-06, "loss": 1.9809, "step": 30301 }, { "epoch": 0.75, "learning_rate": 3.1757877812974316e-06, "loss": 1.9379, "step": 30302 }, { "epoch": 0.75, "learning_rate": 3.1752042556036698e-06, "loss": 2.0272, "step": 30303 }, { "epoch": 0.75, "learning_rate": 3.174620773406557e-06, "loss": 2.0071, "step": 30304 }, { "epoch": 0.75, "learning_rate": 3.174037334709813e-06, "loss": 1.9214, "step": 30305 }, { "epoch": 0.75, "learning_rate": 3.1734539395171505e-06, "loss": 1.9265, "step": 30306 }, { "epoch": 0.75, "learning_rate": 3.1728705878322906e-06, "loss": 1.7324, "step": 30307 }, { "epoch": 0.75, "learning_rate": 3.172287279658954e-06, "loss": 2.0062, "step": 30308 }, { "epoch": 0.75, "learning_rate": 3.1717040150008537e-06, "loss": 1.9341, "step": 30309 }, { "epoch": 0.75, "learning_rate": 3.1711207938617116e-06, "loss": 1.8995, "step": 30310 }, { "epoch": 0.75, "learning_rate": 3.1705376162452394e-06, "loss": 1.7905, "step": 30311 }, { "epoch": 0.75, "learning_rate": 3.1699544821551552e-06, "loss": 1.9084, "step": 30312 }, { "epoch": 0.75, "learning_rate": 3.169371391595182e-06, "loss": 1.8616, "step": 30313 }, { "epoch": 0.75, "learning_rate": 3.168788344569027e-06, "loss": 1.8387, "step": 30314 }, { "epoch": 0.75, "learning_rate": 3.1682053410804124e-06, "loss": 2.0044, "step": 30315 }, { "epoch": 0.75, "learning_rate": 3.16762238113305e-06, "loss": 2.01, "step": 30316 }, { "epoch": 0.75, "learning_rate": 3.1670394647306556e-06, "loss": 1.8972, "step": 30317 }, { "epoch": 0.75, "learning_rate": 3.1664565918769487e-06, "loss": 2.0171, "step": 30318 }, { "epoch": 0.75, "learning_rate": 3.1658737625756385e-06, "loss": 1.884, "step": 30319 }, { "epoch": 0.75, "learning_rate": 3.1652909768304407e-06, "loss": 1.8673, "step": 30320 }, { "epoch": 0.75, "learning_rate": 3.164708234645072e-06, "loss": 1.9227, "step": 30321 }, { "epoch": 0.75, "learning_rate": 3.164125536023245e-06, "loss": 1.9503, "step": 30322 }, { "epoch": 0.75, "learning_rate": 3.163542880968676e-06, "loss": 1.8941, "step": 30323 }, { "epoch": 0.75, "learning_rate": 3.162960269485075e-06, "loss": 2.1871, "step": 30324 }, { "epoch": 0.75, "learning_rate": 3.162377701576158e-06, "loss": 1.8229, "step": 30325 }, { "epoch": 0.75, "learning_rate": 3.1617951772456333e-06, "loss": 1.8535, "step": 30326 }, { "epoch": 0.75, "learning_rate": 3.1612126964972166e-06, "loss": 1.9717, "step": 30327 }, { "epoch": 0.75, "learning_rate": 3.160630259334623e-06, "loss": 1.9688, "step": 30328 }, { "epoch": 0.75, "learning_rate": 3.1600478657615585e-06, "loss": 2.0202, "step": 30329 }, { "epoch": 0.75, "learning_rate": 3.1594655157817423e-06, "loss": 2.0438, "step": 30330 }, { "epoch": 0.75, "learning_rate": 3.1588832093988775e-06, "loss": 1.8576, "step": 30331 }, { "epoch": 0.75, "learning_rate": 3.158300946616679e-06, "loss": 1.8854, "step": 30332 }, { "epoch": 0.75, "learning_rate": 3.1577187274388623e-06, "loss": 1.8914, "step": 30333 }, { "epoch": 0.75, "learning_rate": 3.15713655186913e-06, "loss": 1.8007, "step": 30334 }, { "epoch": 0.75, "learning_rate": 3.1565544199111965e-06, "loss": 1.9367, "step": 30335 }, { "epoch": 0.75, "learning_rate": 3.155972331568775e-06, "loss": 1.9212, "step": 30336 }, { "epoch": 0.75, "learning_rate": 3.1553902868455697e-06, "loss": 1.9407, "step": 30337 }, { "epoch": 0.75, "learning_rate": 3.1548082857452934e-06, "loss": 2.018, "step": 30338 }, { "epoch": 0.75, "learning_rate": 3.1542263282716532e-06, "loss": 1.8991, "step": 30339 }, { "epoch": 0.75, "learning_rate": 3.153644414428363e-06, "loss": 2.0939, "step": 30340 }, { "epoch": 0.75, "learning_rate": 3.1530625442191253e-06, "loss": 1.9986, "step": 30341 }, { "epoch": 0.75, "learning_rate": 3.1524807176476513e-06, "loss": 2.0039, "step": 30342 }, { "epoch": 0.75, "learning_rate": 3.1518989347176533e-06, "loss": 1.8925, "step": 30343 }, { "epoch": 0.75, "learning_rate": 3.1513171954328305e-06, "loss": 1.8728, "step": 30344 }, { "epoch": 0.75, "learning_rate": 3.1507354997968986e-06, "loss": 1.8216, "step": 30345 }, { "epoch": 0.75, "learning_rate": 3.1501538478135585e-06, "loss": 1.8493, "step": 30346 }, { "epoch": 0.75, "learning_rate": 3.14957223948652e-06, "loss": 2.0598, "step": 30347 }, { "epoch": 0.75, "learning_rate": 3.1489906748194943e-06, "loss": 1.909, "step": 30348 }, { "epoch": 0.75, "learning_rate": 3.14840915381618e-06, "loss": 2.0061, "step": 30349 }, { "epoch": 0.75, "learning_rate": 3.147827676480286e-06, "loss": 1.8992, "step": 30350 }, { "epoch": 0.75, "learning_rate": 3.1472462428155237e-06, "loss": 1.9584, "step": 30351 }, { "epoch": 0.75, "learning_rate": 3.1466648528255906e-06, "loss": 1.7715, "step": 30352 }, { "epoch": 0.75, "learning_rate": 3.1460835065141993e-06, "loss": 1.945, "step": 30353 }, { "epoch": 0.75, "learning_rate": 3.145502203885048e-06, "loss": 1.8313, "step": 30354 }, { "epoch": 0.75, "learning_rate": 3.144920944941845e-06, "loss": 1.9613, "step": 30355 }, { "epoch": 0.75, "learning_rate": 3.144339729688295e-06, "loss": 1.8144, "step": 30356 }, { "epoch": 0.75, "learning_rate": 3.143758558128103e-06, "loss": 1.8756, "step": 30357 }, { "epoch": 0.75, "learning_rate": 3.1431774302649743e-06, "loss": 1.8606, "step": 30358 }, { "epoch": 0.75, "learning_rate": 3.1425963461026065e-06, "loss": 1.9465, "step": 30359 }, { "epoch": 0.75, "learning_rate": 3.1420153056447078e-06, "loss": 1.943, "step": 30360 }, { "epoch": 0.75, "learning_rate": 3.141434308894984e-06, "loss": 1.822, "step": 30361 }, { "epoch": 0.75, "learning_rate": 3.1408533558571306e-06, "loss": 1.7758, "step": 30362 }, { "epoch": 0.75, "learning_rate": 3.1402724465348577e-06, "loss": 2.0397, "step": 30363 }, { "epoch": 0.75, "learning_rate": 3.1396915809318605e-06, "loss": 1.9774, "step": 30364 }, { "epoch": 0.75, "learning_rate": 3.1391107590518442e-06, "loss": 2.1174, "step": 30365 }, { "epoch": 0.75, "learning_rate": 3.1385299808985148e-06, "loss": 2.0803, "step": 30366 }, { "epoch": 0.75, "learning_rate": 3.137949246475567e-06, "loss": 1.9919, "step": 30367 }, { "epoch": 0.75, "learning_rate": 3.1373685557867075e-06, "loss": 2.048, "step": 30368 }, { "epoch": 0.75, "learning_rate": 3.1367879088356323e-06, "loss": 1.9747, "step": 30369 }, { "epoch": 0.75, "learning_rate": 3.136207305626043e-06, "loss": 1.9187, "step": 30370 }, { "epoch": 0.75, "learning_rate": 3.1356267461616464e-06, "loss": 2.0203, "step": 30371 }, { "epoch": 0.75, "learning_rate": 3.1350462304461325e-06, "loss": 2.0319, "step": 30372 }, { "epoch": 0.75, "learning_rate": 3.134465758483207e-06, "loss": 2.0545, "step": 30373 }, { "epoch": 0.75, "learning_rate": 3.133885330276569e-06, "loss": 1.693, "step": 30374 }, { "epoch": 0.75, "learning_rate": 3.133304945829917e-06, "loss": 1.9414, "step": 30375 }, { "epoch": 0.75, "learning_rate": 3.132724605146954e-06, "loss": 1.9406, "step": 30376 }, { "epoch": 0.75, "learning_rate": 3.132144308231372e-06, "loss": 1.7404, "step": 30377 }, { "epoch": 0.75, "learning_rate": 3.131564055086874e-06, "loss": 1.8167, "step": 30378 }, { "epoch": 0.75, "learning_rate": 3.1309838457171548e-06, "loss": 1.7153, "step": 30379 }, { "epoch": 0.75, "learning_rate": 3.130403680125914e-06, "loss": 1.9229, "step": 30380 }, { "epoch": 0.75, "learning_rate": 3.1298235583168536e-06, "loss": 2.0306, "step": 30381 }, { "epoch": 0.75, "learning_rate": 3.129243480293661e-06, "loss": 2.013, "step": 30382 }, { "epoch": 0.75, "learning_rate": 3.128663446060044e-06, "loss": 1.8993, "step": 30383 }, { "epoch": 0.75, "learning_rate": 3.1280834556196906e-06, "loss": 1.7864, "step": 30384 }, { "epoch": 0.75, "learning_rate": 3.1275035089762996e-06, "loss": 2.1374, "step": 30385 }, { "epoch": 0.75, "learning_rate": 3.126923606133573e-06, "loss": 1.9671, "step": 30386 }, { "epoch": 0.75, "learning_rate": 3.1263437470951985e-06, "loss": 1.8882, "step": 30387 }, { "epoch": 0.75, "learning_rate": 3.1257639318648748e-06, "loss": 1.9086, "step": 30388 }, { "epoch": 0.75, "learning_rate": 3.125184160446302e-06, "loss": 1.8059, "step": 30389 }, { "epoch": 0.75, "learning_rate": 3.1246044328431668e-06, "loss": 2.0265, "step": 30390 }, { "epoch": 0.75, "learning_rate": 3.124024749059169e-06, "loss": 2.0817, "step": 30391 }, { "epoch": 0.75, "learning_rate": 3.1234451090980022e-06, "loss": 1.889, "step": 30392 }, { "epoch": 0.75, "learning_rate": 3.1228655129633632e-06, "loss": 1.8833, "step": 30393 }, { "epoch": 0.75, "learning_rate": 3.1222859606589404e-06, "loss": 1.9336, "step": 30394 }, { "epoch": 0.75, "learning_rate": 3.121706452188431e-06, "loss": 1.92, "step": 30395 }, { "epoch": 0.75, "learning_rate": 3.121126987555532e-06, "loss": 1.7602, "step": 30396 }, { "epoch": 0.75, "learning_rate": 3.120547566763927e-06, "loss": 2.0605, "step": 30397 }, { "epoch": 0.75, "learning_rate": 3.1199681898173193e-06, "loss": 1.9027, "step": 30398 }, { "epoch": 0.75, "learning_rate": 3.119388856719393e-06, "loss": 2.0697, "step": 30399 }, { "epoch": 0.75, "learning_rate": 3.118809567473844e-06, "loss": 1.7632, "step": 30400 }, { "epoch": 0.75, "learning_rate": 3.1182303220843668e-06, "loss": 2.1116, "step": 30401 }, { "epoch": 0.75, "learning_rate": 3.117651120554647e-06, "loss": 1.8541, "step": 30402 }, { "epoch": 0.75, "learning_rate": 3.117071962888383e-06, "loss": 1.9698, "step": 30403 }, { "epoch": 0.75, "learning_rate": 3.11649284908926e-06, "loss": 2.0572, "step": 30404 }, { "epoch": 0.75, "learning_rate": 3.1159137791609708e-06, "loss": 1.9734, "step": 30405 }, { "epoch": 0.75, "learning_rate": 3.1153347531072067e-06, "loss": 2.0019, "step": 30406 }, { "epoch": 0.75, "learning_rate": 3.11475577093166e-06, "loss": 1.9156, "step": 30407 }, { "epoch": 0.75, "learning_rate": 3.114176832638016e-06, "loss": 2.0297, "step": 30408 }, { "epoch": 0.75, "learning_rate": 3.1135979382299674e-06, "loss": 1.9781, "step": 30409 }, { "epoch": 0.75, "learning_rate": 3.113019087711203e-06, "loss": 1.8411, "step": 30410 }, { "epoch": 0.75, "learning_rate": 3.112440281085416e-06, "loss": 1.9835, "step": 30411 }, { "epoch": 0.75, "learning_rate": 3.1118615183562885e-06, "loss": 1.7069, "step": 30412 }, { "epoch": 0.75, "learning_rate": 3.111282799527512e-06, "loss": 1.8844, "step": 30413 }, { "epoch": 0.75, "learning_rate": 3.110704124602779e-06, "loss": 1.7888, "step": 30414 }, { "epoch": 0.75, "learning_rate": 3.1101254935857696e-06, "loss": 1.8954, "step": 30415 }, { "epoch": 0.75, "learning_rate": 3.1095469064801797e-06, "loss": 1.882, "step": 30416 }, { "epoch": 0.75, "learning_rate": 3.1089683632896894e-06, "loss": 1.9078, "step": 30417 }, { "epoch": 0.75, "learning_rate": 3.10838986401799e-06, "loss": 2.0516, "step": 30418 }, { "epoch": 0.75, "learning_rate": 3.1078114086687706e-06, "loss": 1.9672, "step": 30419 }, { "epoch": 0.75, "learning_rate": 3.107232997245712e-06, "loss": 1.7785, "step": 30420 }, { "epoch": 0.75, "learning_rate": 3.106654629752507e-06, "loss": 1.9287, "step": 30421 }, { "epoch": 0.75, "learning_rate": 3.1060763061928347e-06, "loss": 1.92, "step": 30422 }, { "epoch": 0.75, "learning_rate": 3.105498026570386e-06, "loss": 1.8957, "step": 30423 }, { "epoch": 0.75, "learning_rate": 3.104919790888844e-06, "loss": 2.0091, "step": 30424 }, { "epoch": 0.75, "learning_rate": 3.1043415991518987e-06, "loss": 1.9051, "step": 30425 }, { "epoch": 0.75, "learning_rate": 3.1037634513632297e-06, "loss": 2.1559, "step": 30426 }, { "epoch": 0.75, "learning_rate": 3.103185347526523e-06, "loss": 2.0266, "step": 30427 }, { "epoch": 0.75, "learning_rate": 3.102607287645463e-06, "loss": 2.0419, "step": 30428 }, { "epoch": 0.75, "learning_rate": 3.1020292717237377e-06, "loss": 1.9663, "step": 30429 }, { "epoch": 0.75, "learning_rate": 3.101451299765025e-06, "loss": 2.0175, "step": 30430 }, { "epoch": 0.75, "learning_rate": 3.1008733717730144e-06, "loss": 1.9887, "step": 30431 }, { "epoch": 0.75, "learning_rate": 3.1002954877513835e-06, "loss": 2.0343, "step": 30432 }, { "epoch": 0.75, "learning_rate": 3.099717647703818e-06, "loss": 1.9977, "step": 30433 }, { "epoch": 0.75, "learning_rate": 3.0991398516340032e-06, "loss": 1.8552, "step": 30434 }, { "epoch": 0.75, "learning_rate": 3.098562099545617e-06, "loss": 2.0511, "step": 30435 }, { "epoch": 0.75, "learning_rate": 3.0979843914423456e-06, "loss": 2.1046, "step": 30436 }, { "epoch": 0.75, "learning_rate": 3.0974067273278663e-06, "loss": 1.8922, "step": 30437 }, { "epoch": 0.75, "learning_rate": 3.0968291072058633e-06, "loss": 1.9075, "step": 30438 }, { "epoch": 0.75, "learning_rate": 3.0962515310800213e-06, "loss": 2.0158, "step": 30439 }, { "epoch": 0.75, "learning_rate": 3.095673998954015e-06, "loss": 1.871, "step": 30440 }, { "epoch": 0.75, "learning_rate": 3.0950965108315277e-06, "loss": 1.8121, "step": 30441 }, { "epoch": 0.75, "learning_rate": 3.0945190667162452e-06, "loss": 1.8877, "step": 30442 }, { "epoch": 0.75, "learning_rate": 3.0939416666118385e-06, "loss": 1.8518, "step": 30443 }, { "epoch": 0.75, "learning_rate": 3.0933643105219936e-06, "loss": 2.0176, "step": 30444 }, { "epoch": 0.75, "learning_rate": 3.0927869984503877e-06, "loss": 1.9907, "step": 30445 }, { "epoch": 0.75, "learning_rate": 3.0922097304007048e-06, "loss": 2.0723, "step": 30446 }, { "epoch": 0.75, "learning_rate": 3.0916325063766173e-06, "loss": 1.9447, "step": 30447 }, { "epoch": 0.75, "learning_rate": 3.0910553263818067e-06, "loss": 1.717, "step": 30448 }, { "epoch": 0.75, "learning_rate": 3.0904781904199553e-06, "loss": 2.0233, "step": 30449 }, { "epoch": 0.75, "learning_rate": 3.0899010984947352e-06, "loss": 2.0206, "step": 30450 }, { "epoch": 0.75, "learning_rate": 3.089324050609831e-06, "loss": 1.9835, "step": 30451 }, { "epoch": 0.75, "learning_rate": 3.0887470467689127e-06, "loss": 1.9899, "step": 30452 }, { "epoch": 0.75, "learning_rate": 3.0881700869756626e-06, "loss": 1.9183, "step": 30453 }, { "epoch": 0.75, "learning_rate": 3.0875931712337605e-06, "loss": 2.0146, "step": 30454 }, { "epoch": 0.75, "learning_rate": 3.0870162995468756e-06, "loss": 1.8747, "step": 30455 }, { "epoch": 0.75, "learning_rate": 3.086439471918693e-06, "loss": 1.8379, "step": 30456 }, { "epoch": 0.75, "learning_rate": 3.0858626883528796e-06, "loss": 1.7648, "step": 30457 }, { "epoch": 0.75, "learning_rate": 3.0852859488531182e-06, "loss": 1.9725, "step": 30458 }, { "epoch": 0.75, "learning_rate": 3.0847092534230827e-06, "loss": 2.1173, "step": 30459 }, { "epoch": 0.75, "learning_rate": 3.084132602066452e-06, "loss": 1.8738, "step": 30460 }, { "epoch": 0.75, "learning_rate": 3.0835559947868943e-06, "loss": 1.9906, "step": 30461 }, { "epoch": 0.75, "learning_rate": 3.082979431588089e-06, "loss": 2.0703, "step": 30462 }, { "epoch": 0.75, "learning_rate": 3.0824029124737086e-06, "loss": 1.8879, "step": 30463 }, { "epoch": 0.75, "learning_rate": 3.0818264374474337e-06, "loss": 1.8483, "step": 30464 }, { "epoch": 0.75, "learning_rate": 3.081250006512929e-06, "loss": 1.9476, "step": 30465 }, { "epoch": 0.75, "learning_rate": 3.080673619673873e-06, "loss": 1.8358, "step": 30466 }, { "epoch": 0.75, "learning_rate": 3.0800972769339434e-06, "loss": 2.057, "step": 30467 }, { "epoch": 0.75, "learning_rate": 3.0795209782968037e-06, "loss": 1.9528, "step": 30468 }, { "epoch": 0.75, "learning_rate": 3.0789447237661375e-06, "loss": 1.942, "step": 30469 }, { "epoch": 0.75, "learning_rate": 3.0783685133456077e-06, "loss": 2.0047, "step": 30470 }, { "epoch": 0.75, "learning_rate": 3.0777923470388905e-06, "loss": 1.9407, "step": 30471 }, { "epoch": 0.75, "learning_rate": 3.077216224849663e-06, "loss": 1.9924, "step": 30472 }, { "epoch": 0.75, "learning_rate": 3.0766401467815887e-06, "loss": 1.9102, "step": 30473 }, { "epoch": 0.75, "learning_rate": 3.076064112838346e-06, "loss": 1.8076, "step": 30474 }, { "epoch": 0.75, "learning_rate": 3.0754881230236e-06, "loss": 2.0727, "step": 30475 }, { "epoch": 0.75, "learning_rate": 3.074912177341025e-06, "loss": 2.0801, "step": 30476 }, { "epoch": 0.75, "learning_rate": 3.074336275794292e-06, "loss": 2.0198, "step": 30477 }, { "epoch": 0.75, "learning_rate": 3.073760418387074e-06, "loss": 1.9789, "step": 30478 }, { "epoch": 0.75, "learning_rate": 3.073184605123034e-06, "loss": 2.0515, "step": 30479 }, { "epoch": 0.75, "learning_rate": 3.072608836005847e-06, "loss": 1.8109, "step": 30480 }, { "epoch": 0.75, "learning_rate": 3.072033111039181e-06, "loss": 2.0448, "step": 30481 }, { "epoch": 0.75, "learning_rate": 3.0714574302267085e-06, "loss": 1.7617, "step": 30482 }, { "epoch": 0.75, "learning_rate": 3.0708817935720924e-06, "loss": 1.9785, "step": 30483 }, { "epoch": 0.75, "learning_rate": 3.0703062010790087e-06, "loss": 1.9017, "step": 30484 }, { "epoch": 0.75, "learning_rate": 3.0697306527511186e-06, "loss": 1.9017, "step": 30485 }, { "epoch": 0.75, "learning_rate": 3.0691551485920935e-06, "loss": 1.8348, "step": 30486 }, { "epoch": 0.75, "learning_rate": 3.068579688605604e-06, "loss": 1.9344, "step": 30487 }, { "epoch": 0.75, "learning_rate": 3.068004272795313e-06, "loss": 1.7663, "step": 30488 }, { "epoch": 0.75, "learning_rate": 3.0674289011648917e-06, "loss": 1.8579, "step": 30489 }, { "epoch": 0.75, "learning_rate": 3.0668535737180026e-06, "loss": 1.9868, "step": 30490 }, { "epoch": 0.75, "learning_rate": 3.0662782904583155e-06, "loss": 1.9639, "step": 30491 }, { "epoch": 0.75, "learning_rate": 3.065703051389499e-06, "loss": 2.0287, "step": 30492 }, { "epoch": 0.75, "learning_rate": 3.065127856515214e-06, "loss": 2.0332, "step": 30493 }, { "epoch": 0.75, "learning_rate": 3.06455270583913e-06, "loss": 1.9454, "step": 30494 }, { "epoch": 0.75, "learning_rate": 3.063977599364911e-06, "loss": 1.9596, "step": 30495 }, { "epoch": 0.75, "learning_rate": 3.0634025370962263e-06, "loss": 1.8675, "step": 30496 }, { "epoch": 0.75, "learning_rate": 3.0628275190367342e-06, "loss": 1.7953, "step": 30497 }, { "epoch": 0.75, "learning_rate": 3.0622525451901043e-06, "loss": 1.8234, "step": 30498 }, { "epoch": 0.75, "learning_rate": 3.0616776155600015e-06, "loss": 2.1553, "step": 30499 }, { "epoch": 0.75, "learning_rate": 3.0611027301500864e-06, "loss": 2.0334, "step": 30500 }, { "epoch": 0.75, "learning_rate": 3.0605278889640244e-06, "loss": 1.9337, "step": 30501 }, { "epoch": 0.75, "learning_rate": 3.0599530920054833e-06, "loss": 1.8606, "step": 30502 }, { "epoch": 0.75, "learning_rate": 3.0593783392781184e-06, "loss": 2.0008, "step": 30503 }, { "epoch": 0.75, "learning_rate": 3.0588036307856015e-06, "loss": 1.7969, "step": 30504 }, { "epoch": 0.75, "learning_rate": 3.0582289665315878e-06, "loss": 2.0817, "step": 30505 }, { "epoch": 0.75, "learning_rate": 3.057654346519744e-06, "loss": 1.836, "step": 30506 }, { "epoch": 0.75, "learning_rate": 3.0570797707537336e-06, "loss": 1.9131, "step": 30507 }, { "epoch": 0.75, "learning_rate": 3.0565052392372128e-06, "loss": 1.8904, "step": 30508 }, { "epoch": 0.75, "learning_rate": 3.0559307519738512e-06, "loss": 1.819, "step": 30509 }, { "epoch": 0.75, "learning_rate": 3.0553563089673022e-06, "loss": 1.9426, "step": 30510 }, { "epoch": 0.75, "learning_rate": 3.054781910221231e-06, "loss": 2.0758, "step": 30511 }, { "epoch": 0.75, "learning_rate": 3.054207555739298e-06, "loss": 1.765, "step": 30512 }, { "epoch": 0.75, "learning_rate": 3.053633245525164e-06, "loss": 1.9161, "step": 30513 }, { "epoch": 0.75, "learning_rate": 3.053058979582492e-06, "loss": 1.9187, "step": 30514 }, { "epoch": 0.75, "learning_rate": 3.0524847579149354e-06, "loss": 1.8102, "step": 30515 }, { "epoch": 0.75, "learning_rate": 3.051910580526157e-06, "loss": 1.9287, "step": 30516 }, { "epoch": 0.75, "learning_rate": 3.0513364474198214e-06, "loss": 1.9978, "step": 30517 }, { "epoch": 0.75, "learning_rate": 3.050762358599579e-06, "loss": 1.9179, "step": 30518 }, { "epoch": 0.75, "learning_rate": 3.050188314069097e-06, "loss": 2.0856, "step": 30519 }, { "epoch": 0.75, "learning_rate": 3.0496143138320266e-06, "loss": 1.8483, "step": 30520 }, { "epoch": 0.75, "learning_rate": 3.049040357892028e-06, "loss": 1.7636, "step": 30521 }, { "epoch": 0.75, "learning_rate": 3.0484664462527645e-06, "loss": 2.1087, "step": 30522 }, { "epoch": 0.75, "learning_rate": 3.047892578917886e-06, "loss": 1.9376, "step": 30523 }, { "epoch": 0.75, "learning_rate": 3.047318755891058e-06, "loss": 1.9297, "step": 30524 }, { "epoch": 0.75, "learning_rate": 3.0467449771759296e-06, "loss": 1.877, "step": 30525 }, { "epoch": 0.75, "learning_rate": 3.0461712427761615e-06, "loss": 2.1699, "step": 30526 }, { "epoch": 0.75, "learning_rate": 3.045597552695413e-06, "loss": 1.992, "step": 30527 }, { "epoch": 0.75, "learning_rate": 3.0450239069373345e-06, "loss": 2.0294, "step": 30528 }, { "epoch": 0.75, "learning_rate": 3.0444503055055854e-06, "loss": 1.9102, "step": 30529 }, { "epoch": 0.75, "learning_rate": 3.0438767484038214e-06, "loss": 1.8761, "step": 30530 }, { "epoch": 0.75, "learning_rate": 3.0433032356357005e-06, "loss": 2.0627, "step": 30531 }, { "epoch": 0.75, "learning_rate": 3.042729767204873e-06, "loss": 2.0194, "step": 30532 }, { "epoch": 0.75, "learning_rate": 3.0421563431149957e-06, "loss": 1.8094, "step": 30533 }, { "epoch": 0.75, "learning_rate": 3.0415829633697235e-06, "loss": 1.8466, "step": 30534 }, { "epoch": 0.75, "learning_rate": 3.0410096279727132e-06, "loss": 1.8896, "step": 30535 }, { "epoch": 0.75, "learning_rate": 3.0404363369276136e-06, "loss": 2.0726, "step": 30536 }, { "epoch": 0.75, "learning_rate": 3.0398630902380845e-06, "loss": 1.9109, "step": 30537 }, { "epoch": 0.75, "learning_rate": 3.039289887907774e-06, "loss": 1.8542, "step": 30538 }, { "epoch": 0.75, "learning_rate": 3.0387167299403364e-06, "loss": 2.0437, "step": 30539 }, { "epoch": 0.75, "learning_rate": 3.0381436163394286e-06, "loss": 1.9187, "step": 30540 }, { "epoch": 0.75, "learning_rate": 3.037570547108698e-06, "loss": 2.1051, "step": 30541 }, { "epoch": 0.75, "learning_rate": 3.036997522251802e-06, "loss": 1.9296, "step": 30542 }, { "epoch": 0.75, "learning_rate": 3.036424541772387e-06, "loss": 1.981, "step": 30543 }, { "epoch": 0.75, "learning_rate": 3.035851605674107e-06, "loss": 2.0456, "step": 30544 }, { "epoch": 0.75, "learning_rate": 3.035278713960619e-06, "loss": 1.854, "step": 30545 }, { "epoch": 0.75, "learning_rate": 3.0347058666355646e-06, "loss": 1.9279, "step": 30546 }, { "epoch": 0.75, "learning_rate": 3.0341330637026e-06, "loss": 2.019, "step": 30547 }, { "epoch": 0.75, "learning_rate": 3.033560305165376e-06, "loss": 1.8715, "step": 30548 }, { "epoch": 0.75, "learning_rate": 3.0329875910275464e-06, "loss": 1.7724, "step": 30549 }, { "epoch": 0.75, "learning_rate": 3.0324149212927534e-06, "loss": 1.9017, "step": 30550 }, { "epoch": 0.75, "learning_rate": 3.03184229596465e-06, "loss": 2.0359, "step": 30551 }, { "epoch": 0.75, "learning_rate": 3.0312697150468905e-06, "loss": 1.8493, "step": 30552 }, { "epoch": 0.75, "learning_rate": 3.0306971785431173e-06, "loss": 1.931, "step": 30553 }, { "epoch": 0.75, "learning_rate": 3.030124686456981e-06, "loss": 1.9418, "step": 30554 }, { "epoch": 0.75, "learning_rate": 3.0295522387921363e-06, "loss": 1.8872, "step": 30555 }, { "epoch": 0.75, "learning_rate": 3.0289798355522226e-06, "loss": 1.8218, "step": 30556 }, { "epoch": 0.75, "learning_rate": 3.0284074767408956e-06, "loss": 1.8274, "step": 30557 }, { "epoch": 0.75, "learning_rate": 3.0278351623617965e-06, "loss": 2.062, "step": 30558 }, { "epoch": 0.75, "learning_rate": 3.0272628924185764e-06, "loss": 1.6929, "step": 30559 }, { "epoch": 0.75, "learning_rate": 3.0266906669148845e-06, "loss": 1.9052, "step": 30560 }, { "epoch": 0.75, "learning_rate": 3.0261184858543633e-06, "loss": 1.9269, "step": 30561 }, { "epoch": 0.75, "learning_rate": 3.0255463492406645e-06, "loss": 1.8785, "step": 30562 }, { "epoch": 0.75, "learning_rate": 3.024974257077429e-06, "loss": 1.974, "step": 30563 }, { "epoch": 0.75, "learning_rate": 3.0244022093683044e-06, "loss": 1.8277, "step": 30564 }, { "epoch": 0.75, "learning_rate": 3.023830206116939e-06, "loss": 2.1576, "step": 30565 }, { "epoch": 0.75, "learning_rate": 3.0232582473269765e-06, "loss": 1.9979, "step": 30566 }, { "epoch": 0.75, "learning_rate": 3.022686333002066e-06, "loss": 1.8907, "step": 30567 }, { "epoch": 0.75, "learning_rate": 3.0221144631458467e-06, "loss": 1.9481, "step": 30568 }, { "epoch": 0.75, "learning_rate": 3.021542637761966e-06, "loss": 2.0125, "step": 30569 }, { "epoch": 0.75, "learning_rate": 3.0209708568540706e-06, "loss": 1.8965, "step": 30570 }, { "epoch": 0.75, "learning_rate": 3.0203991204257998e-06, "loss": 1.9796, "step": 30571 }, { "epoch": 0.75, "learning_rate": 3.0198274284808037e-06, "loss": 1.7965, "step": 30572 }, { "epoch": 0.75, "learning_rate": 3.0192557810227187e-06, "loss": 2.0893, "step": 30573 }, { "epoch": 0.75, "learning_rate": 3.0186841780551914e-06, "loss": 2.0237, "step": 30574 }, { "epoch": 0.75, "learning_rate": 3.0181126195818677e-06, "loss": 2.1081, "step": 30575 }, { "epoch": 0.75, "learning_rate": 3.017541105606384e-06, "loss": 2.0109, "step": 30576 }, { "epoch": 0.75, "learning_rate": 3.016969636132391e-06, "loss": 2.1572, "step": 30577 }, { "epoch": 0.75, "learning_rate": 3.0163982111635227e-06, "loss": 1.9323, "step": 30578 }, { "epoch": 0.75, "learning_rate": 3.0158268307034233e-06, "loss": 1.9778, "step": 30579 }, { "epoch": 0.75, "learning_rate": 3.01525549475574e-06, "loss": 1.8996, "step": 30580 }, { "epoch": 0.75, "learning_rate": 3.0146842033241062e-06, "loss": 2.0285, "step": 30581 }, { "epoch": 0.75, "learning_rate": 3.0141129564121663e-06, "loss": 1.997, "step": 30582 }, { "epoch": 0.75, "learning_rate": 3.013541754023561e-06, "loss": 1.7961, "step": 30583 }, { "epoch": 0.75, "learning_rate": 3.0129705961619314e-06, "loss": 2.1175, "step": 30584 }, { "epoch": 0.75, "learning_rate": 3.01239948283092e-06, "loss": 1.9248, "step": 30585 }, { "epoch": 0.75, "learning_rate": 3.01182841403416e-06, "loss": 1.9654, "step": 30586 }, { "epoch": 0.75, "learning_rate": 3.0112573897752962e-06, "loss": 1.7726, "step": 30587 }, { "epoch": 0.75, "learning_rate": 3.010686410057969e-06, "loss": 2.0628, "step": 30588 }, { "epoch": 0.75, "learning_rate": 3.010115474885812e-06, "loss": 1.9617, "step": 30589 }, { "epoch": 0.75, "learning_rate": 3.0095445842624706e-06, "loss": 2.1075, "step": 30590 }, { "epoch": 0.75, "learning_rate": 3.008973738191576e-06, "loss": 1.9492, "step": 30591 }, { "epoch": 0.75, "learning_rate": 3.0084029366767708e-06, "loss": 2.1157, "step": 30592 }, { "epoch": 0.75, "learning_rate": 3.0078321797216947e-06, "loss": 2.0435, "step": 30593 }, { "epoch": 0.75, "learning_rate": 3.0072614673299804e-06, "loss": 1.9988, "step": 30594 }, { "epoch": 0.75, "learning_rate": 3.0066907995052697e-06, "loss": 1.9607, "step": 30595 }, { "epoch": 0.75, "learning_rate": 3.006120176251195e-06, "loss": 2.0097, "step": 30596 }, { "epoch": 0.75, "learning_rate": 3.0055495975713954e-06, "loss": 1.8939, "step": 30597 }, { "epoch": 0.75, "learning_rate": 3.0049790634695108e-06, "loss": 1.8947, "step": 30598 }, { "epoch": 0.75, "learning_rate": 3.004408573949171e-06, "loss": 1.7483, "step": 30599 }, { "epoch": 0.75, "learning_rate": 3.0038381290140164e-06, "loss": 1.8859, "step": 30600 }, { "epoch": 0.75, "learning_rate": 3.0032677286676804e-06, "loss": 1.8911, "step": 30601 }, { "epoch": 0.75, "learning_rate": 3.002697372913799e-06, "loss": 1.9184, "step": 30602 }, { "epoch": 0.75, "learning_rate": 3.0021270617560115e-06, "loss": 1.8806, "step": 30603 }, { "epoch": 0.75, "learning_rate": 3.001556795197945e-06, "loss": 1.8383, "step": 30604 }, { "epoch": 0.75, "learning_rate": 3.000986573243241e-06, "loss": 1.8247, "step": 30605 }, { "epoch": 0.75, "learning_rate": 3.000416395895528e-06, "loss": 1.9678, "step": 30606 }, { "epoch": 0.75, "learning_rate": 2.999846263158441e-06, "loss": 1.8769, "step": 30607 }, { "epoch": 0.75, "learning_rate": 2.99927617503562e-06, "loss": 2.0327, "step": 30608 }, { "epoch": 0.75, "learning_rate": 2.9987061315306887e-06, "loss": 1.8506, "step": 30609 }, { "epoch": 0.75, "learning_rate": 2.998136132647289e-06, "loss": 1.8477, "step": 30610 }, { "epoch": 0.75, "learning_rate": 2.9975661783890463e-06, "loss": 1.8044, "step": 30611 }, { "epoch": 0.75, "learning_rate": 2.9969962687595966e-06, "loss": 1.8283, "step": 30612 }, { "epoch": 0.75, "learning_rate": 2.9964264037625744e-06, "loss": 1.9984, "step": 30613 }, { "epoch": 0.75, "learning_rate": 2.9958565834016064e-06, "loss": 1.9524, "step": 30614 }, { "epoch": 0.75, "learning_rate": 2.9952868076803297e-06, "loss": 1.9678, "step": 30615 }, { "epoch": 0.75, "learning_rate": 2.9947170766023704e-06, "loss": 1.8206, "step": 30616 }, { "epoch": 0.75, "learning_rate": 2.9941473901713616e-06, "loss": 2.1171, "step": 30617 }, { "epoch": 0.75, "learning_rate": 2.9935777483909347e-06, "loss": 1.9004, "step": 30618 }, { "epoch": 0.75, "learning_rate": 2.9930081512647202e-06, "loss": 1.8683, "step": 30619 }, { "epoch": 0.75, "learning_rate": 2.992438598796351e-06, "loss": 2.0056, "step": 30620 }, { "epoch": 0.75, "learning_rate": 2.9918690909894522e-06, "loss": 1.8754, "step": 30621 }, { "epoch": 0.75, "learning_rate": 2.9912996278476548e-06, "loss": 1.9328, "step": 30622 }, { "epoch": 0.75, "learning_rate": 2.9907302093745916e-06, "loss": 1.9871, "step": 30623 }, { "epoch": 0.75, "learning_rate": 2.9901608355738865e-06, "loss": 1.9096, "step": 30624 }, { "epoch": 0.75, "learning_rate": 2.9895915064491744e-06, "loss": 1.9419, "step": 30625 }, { "epoch": 0.75, "learning_rate": 2.9890222220040764e-06, "loss": 1.8059, "step": 30626 }, { "epoch": 0.75, "learning_rate": 2.988452982242225e-06, "loss": 1.871, "step": 30627 }, { "epoch": 0.75, "learning_rate": 2.9878837871672504e-06, "loss": 1.9563, "step": 30628 }, { "epoch": 0.75, "learning_rate": 2.9873146367827752e-06, "loss": 2.0843, "step": 30629 }, { "epoch": 0.75, "learning_rate": 2.986745531092432e-06, "loss": 1.9441, "step": 30630 }, { "epoch": 0.76, "learning_rate": 2.9861764700998418e-06, "loss": 2.0094, "step": 30631 }, { "epoch": 0.76, "learning_rate": 2.9856074538086353e-06, "loss": 1.906, "step": 30632 }, { "epoch": 0.76, "learning_rate": 2.985038482222441e-06, "loss": 1.8478, "step": 30633 }, { "epoch": 0.76, "learning_rate": 2.98446955534488e-06, "loss": 2.0157, "step": 30634 }, { "epoch": 0.76, "learning_rate": 2.9839006731795796e-06, "loss": 2.0174, "step": 30635 }, { "epoch": 0.76, "learning_rate": 2.9833318357301677e-06, "loss": 1.8605, "step": 30636 }, { "epoch": 0.76, "learning_rate": 2.9827630430002684e-06, "loss": 1.9906, "step": 30637 }, { "epoch": 0.76, "learning_rate": 2.98219429499351e-06, "loss": 1.9742, "step": 30638 }, { "epoch": 0.76, "learning_rate": 2.981625591713512e-06, "loss": 2.0128, "step": 30639 }, { "epoch": 0.76, "learning_rate": 2.981056933163904e-06, "loss": 1.9747, "step": 30640 }, { "epoch": 0.76, "learning_rate": 2.9804883193483035e-06, "loss": 1.8081, "step": 30641 }, { "epoch": 0.76, "learning_rate": 2.97991975027034e-06, "loss": 1.8945, "step": 30642 }, { "epoch": 0.76, "learning_rate": 2.9793512259336376e-06, "loss": 2.0584, "step": 30643 }, { "epoch": 0.76, "learning_rate": 2.9787827463418164e-06, "loss": 1.8073, "step": 30644 }, { "epoch": 0.76, "learning_rate": 2.9782143114984997e-06, "loss": 2.0132, "step": 30645 }, { "epoch": 0.76, "learning_rate": 2.9776459214073137e-06, "loss": 2.0066, "step": 30646 }, { "epoch": 0.76, "learning_rate": 2.9770775760718773e-06, "loss": 1.9523, "step": 30647 }, { "epoch": 0.76, "learning_rate": 2.9765092754958168e-06, "loss": 2.0024, "step": 30648 }, { "epoch": 0.76, "learning_rate": 2.9759410196827488e-06, "loss": 1.9763, "step": 30649 }, { "epoch": 0.76, "learning_rate": 2.9753728086362966e-06, "loss": 1.9412, "step": 30650 }, { "epoch": 0.76, "learning_rate": 2.9748046423600872e-06, "loss": 1.9678, "step": 30651 }, { "epoch": 0.76, "learning_rate": 2.974236520857734e-06, "loss": 2.0632, "step": 30652 }, { "epoch": 0.76, "learning_rate": 2.97366844413286e-06, "loss": 2.1119, "step": 30653 }, { "epoch": 0.76, "learning_rate": 2.973100412189087e-06, "loss": 1.9466, "step": 30654 }, { "epoch": 0.76, "learning_rate": 2.972532425030036e-06, "loss": 1.8535, "step": 30655 }, { "epoch": 0.76, "learning_rate": 2.9719644826593285e-06, "loss": 1.921, "step": 30656 }, { "epoch": 0.76, "learning_rate": 2.9713965850805793e-06, "loss": 2.0976, "step": 30657 }, { "epoch": 0.76, "learning_rate": 2.970828732297413e-06, "loss": 1.7355, "step": 30658 }, { "epoch": 0.76, "learning_rate": 2.9702609243134427e-06, "loss": 1.984, "step": 30659 }, { "epoch": 0.76, "learning_rate": 2.96969316113229e-06, "loss": 1.9125, "step": 30660 }, { "epoch": 0.76, "learning_rate": 2.9691254427575776e-06, "loss": 1.994, "step": 30661 }, { "epoch": 0.76, "learning_rate": 2.9685577691929167e-06, "loss": 2.0777, "step": 30662 }, { "epoch": 0.76, "learning_rate": 2.967990140441932e-06, "loss": 2.0375, "step": 30663 }, { "epoch": 0.76, "learning_rate": 2.967422556508235e-06, "loss": 1.803, "step": 30664 }, { "epoch": 0.76, "learning_rate": 2.9668550173954446e-06, "loss": 2.058, "step": 30665 }, { "epoch": 0.76, "learning_rate": 2.966287523107183e-06, "loss": 1.9573, "step": 30666 }, { "epoch": 0.76, "learning_rate": 2.9657200736470604e-06, "loss": 1.774, "step": 30667 }, { "epoch": 0.76, "learning_rate": 2.9651526690186993e-06, "loss": 1.9839, "step": 30668 }, { "epoch": 0.76, "learning_rate": 2.9645853092257092e-06, "loss": 1.9995, "step": 30669 }, { "epoch": 0.76, "learning_rate": 2.9640179942717106e-06, "loss": 2.0803, "step": 30670 }, { "epoch": 0.76, "learning_rate": 2.963450724160317e-06, "loss": 1.9002, "step": 30671 }, { "epoch": 0.76, "learning_rate": 2.9628834988951462e-06, "loss": 2.0185, "step": 30672 }, { "epoch": 0.76, "learning_rate": 2.9623163184798155e-06, "loss": 1.9789, "step": 30673 }, { "epoch": 0.76, "learning_rate": 2.9617491829179324e-06, "loss": 1.7816, "step": 30674 }, { "epoch": 0.76, "learning_rate": 2.961182092213115e-06, "loss": 2.0086, "step": 30675 }, { "epoch": 0.76, "learning_rate": 2.960615046368982e-06, "loss": 1.9551, "step": 30676 }, { "epoch": 0.76, "learning_rate": 2.9600480453891402e-06, "loss": 2.0328, "step": 30677 }, { "epoch": 0.76, "learning_rate": 2.95948108927721e-06, "loss": 1.837, "step": 30678 }, { "epoch": 0.76, "learning_rate": 2.9589141780367982e-06, "loss": 1.779, "step": 30679 }, { "epoch": 0.76, "learning_rate": 2.95834731167152e-06, "loss": 1.9702, "step": 30680 }, { "epoch": 0.76, "learning_rate": 2.957780490184993e-06, "loss": 2.0265, "step": 30681 }, { "epoch": 0.76, "learning_rate": 2.9572137135808223e-06, "loss": 2.0243, "step": 30682 }, { "epoch": 0.76, "learning_rate": 2.9566469818626275e-06, "loss": 1.9667, "step": 30683 }, { "epoch": 0.76, "learning_rate": 2.9560802950340128e-06, "loss": 1.9948, "step": 30684 }, { "epoch": 0.76, "learning_rate": 2.9555136530985953e-06, "loss": 1.9047, "step": 30685 }, { "epoch": 0.76, "learning_rate": 2.9549470560599868e-06, "loss": 1.8888, "step": 30686 }, { "epoch": 0.76, "learning_rate": 2.9543805039217943e-06, "loss": 1.9883, "step": 30687 }, { "epoch": 0.76, "learning_rate": 2.95381399668763e-06, "loss": 1.991, "step": 30688 }, { "epoch": 0.76, "learning_rate": 2.953247534361107e-06, "loss": 1.9247, "step": 30689 }, { "epoch": 0.76, "learning_rate": 2.9526811169458324e-06, "loss": 1.9056, "step": 30690 }, { "epoch": 0.76, "learning_rate": 2.9521147444454214e-06, "loss": 1.9125, "step": 30691 }, { "epoch": 0.76, "learning_rate": 2.9515484168634767e-06, "loss": 1.8641, "step": 30692 }, { "epoch": 0.76, "learning_rate": 2.950982134203615e-06, "loss": 1.9652, "step": 30693 }, { "epoch": 0.76, "learning_rate": 2.9504158964694373e-06, "loss": 2.0199, "step": 30694 }, { "epoch": 0.76, "learning_rate": 2.9498497036645567e-06, "loss": 1.8545, "step": 30695 }, { "epoch": 0.76, "learning_rate": 2.949283555792585e-06, "loss": 1.9782, "step": 30696 }, { "epoch": 0.76, "learning_rate": 2.948717452857123e-06, "loss": 1.9831, "step": 30697 }, { "epoch": 0.76, "learning_rate": 2.9481513948617866e-06, "loss": 1.9735, "step": 30698 }, { "epoch": 0.76, "learning_rate": 2.947585381810176e-06, "loss": 1.8597, "step": 30699 }, { "epoch": 0.76, "learning_rate": 2.9470194137059015e-06, "loss": 2.0966, "step": 30700 }, { "epoch": 0.76, "learning_rate": 2.9464534905525744e-06, "loss": 1.7098, "step": 30701 }, { "epoch": 0.76, "learning_rate": 2.945887612353795e-06, "loss": 2.0596, "step": 30702 }, { "epoch": 0.76, "learning_rate": 2.945321779113173e-06, "loss": 1.9806, "step": 30703 }, { "epoch": 0.76, "learning_rate": 2.9447559908343173e-06, "loss": 1.9797, "step": 30704 }, { "epoch": 0.76, "learning_rate": 2.9441902475208274e-06, "loss": 1.9486, "step": 30705 }, { "epoch": 0.76, "learning_rate": 2.9436245491763125e-06, "loss": 2.0166, "step": 30706 }, { "epoch": 0.76, "learning_rate": 2.9430588958043784e-06, "loss": 2.0488, "step": 30707 }, { "epoch": 0.76, "learning_rate": 2.9424932874086287e-06, "loss": 2.1672, "step": 30708 }, { "epoch": 0.76, "learning_rate": 2.9419277239926735e-06, "loss": 1.8611, "step": 30709 }, { "epoch": 0.76, "learning_rate": 2.941362205560109e-06, "loss": 1.9834, "step": 30710 }, { "epoch": 0.76, "learning_rate": 2.940796732114547e-06, "loss": 2.1018, "step": 30711 }, { "epoch": 0.76, "learning_rate": 2.9402313036595854e-06, "loss": 1.9768, "step": 30712 }, { "epoch": 0.76, "learning_rate": 2.9396659201988296e-06, "loss": 1.7499, "step": 30713 }, { "epoch": 0.76, "learning_rate": 2.9391005817358877e-06, "loss": 1.8876, "step": 30714 }, { "epoch": 0.76, "learning_rate": 2.9385352882743557e-06, "loss": 1.9133, "step": 30715 }, { "epoch": 0.76, "learning_rate": 2.9379700398178424e-06, "loss": 1.9245, "step": 30716 }, { "epoch": 0.76, "learning_rate": 2.9374048363699435e-06, "loss": 1.9959, "step": 30717 }, { "epoch": 0.76, "learning_rate": 2.9368396779342666e-06, "loss": 1.8497, "step": 30718 }, { "epoch": 0.76, "learning_rate": 2.936274564514415e-06, "loss": 2.0651, "step": 30719 }, { "epoch": 0.76, "learning_rate": 2.935709496113984e-06, "loss": 1.9648, "step": 30720 }, { "epoch": 0.76, "learning_rate": 2.935144472736582e-06, "loss": 1.9556, "step": 30721 }, { "epoch": 0.76, "learning_rate": 2.9345794943858042e-06, "loss": 1.8206, "step": 30722 }, { "epoch": 0.76, "learning_rate": 2.9340145610652524e-06, "loss": 2.0131, "step": 30723 }, { "epoch": 0.76, "learning_rate": 2.93344967277853e-06, "loss": 1.9851, "step": 30724 }, { "epoch": 0.76, "learning_rate": 2.9328848295292345e-06, "loss": 1.8962, "step": 30725 }, { "epoch": 0.76, "learning_rate": 2.932320031320971e-06, "loss": 1.8201, "step": 30726 }, { "epoch": 0.76, "learning_rate": 2.9317552781573324e-06, "loss": 1.9734, "step": 30727 }, { "epoch": 0.76, "learning_rate": 2.9311905700419195e-06, "loss": 1.8699, "step": 30728 }, { "epoch": 0.76, "learning_rate": 2.930625906978337e-06, "loss": 2.0481, "step": 30729 }, { "epoch": 0.76, "learning_rate": 2.930061288970176e-06, "loss": 1.9626, "step": 30730 }, { "epoch": 0.76, "learning_rate": 2.929496716021042e-06, "loss": 2.0147, "step": 30731 }, { "epoch": 0.76, "learning_rate": 2.9289321881345257e-06, "loss": 2.0461, "step": 30732 }, { "epoch": 0.76, "learning_rate": 2.928367705314229e-06, "loss": 1.9517, "step": 30733 }, { "epoch": 0.76, "learning_rate": 2.9278032675637536e-06, "loss": 1.8432, "step": 30734 }, { "epoch": 0.76, "learning_rate": 2.927238874886689e-06, "loss": 1.8169, "step": 30735 }, { "epoch": 0.76, "learning_rate": 2.9266745272866392e-06, "loss": 2.0126, "step": 30736 }, { "epoch": 0.76, "learning_rate": 2.9261102247671946e-06, "loss": 1.8821, "step": 30737 }, { "epoch": 0.76, "learning_rate": 2.925545967331954e-06, "loss": 2.0453, "step": 30738 }, { "epoch": 0.76, "learning_rate": 2.9249817549845183e-06, "loss": 1.9382, "step": 30739 }, { "epoch": 0.76, "learning_rate": 2.9244175877284774e-06, "loss": 1.9534, "step": 30740 }, { "epoch": 0.76, "learning_rate": 2.923853465567428e-06, "loss": 1.7239, "step": 30741 }, { "epoch": 0.76, "learning_rate": 2.9232893885049664e-06, "loss": 1.8056, "step": 30742 }, { "epoch": 0.76, "learning_rate": 2.922725356544688e-06, "loss": 1.999, "step": 30743 }, { "epoch": 0.76, "learning_rate": 2.9221613696901905e-06, "loss": 2.0322, "step": 30744 }, { "epoch": 0.76, "learning_rate": 2.921597427945062e-06, "loss": 1.9862, "step": 30745 }, { "epoch": 0.76, "learning_rate": 2.9210335313129035e-06, "loss": 1.8742, "step": 30746 }, { "epoch": 0.76, "learning_rate": 2.920469679797302e-06, "loss": 1.9327, "step": 30747 }, { "epoch": 0.76, "learning_rate": 2.919905873401854e-06, "loss": 1.8691, "step": 30748 }, { "epoch": 0.76, "learning_rate": 2.919342112130157e-06, "loss": 1.9601, "step": 30749 }, { "epoch": 0.76, "learning_rate": 2.9187783959857964e-06, "loss": 1.7769, "step": 30750 }, { "epoch": 0.76, "learning_rate": 2.9182147249723713e-06, "loss": 1.8818, "step": 30751 }, { "epoch": 0.76, "learning_rate": 2.917651099093469e-06, "loss": 1.9577, "step": 30752 }, { "epoch": 0.76, "learning_rate": 2.917087518352685e-06, "loss": 1.9985, "step": 30753 }, { "epoch": 0.76, "learning_rate": 2.9165239827536127e-06, "loss": 2.0151, "step": 30754 }, { "epoch": 0.76, "learning_rate": 2.91596049229984e-06, "loss": 1.9202, "step": 30755 }, { "epoch": 0.76, "learning_rate": 2.915397046994961e-06, "loss": 1.8456, "step": 30756 }, { "epoch": 0.76, "learning_rate": 2.914833646842563e-06, "loss": 1.927, "step": 30757 }, { "epoch": 0.76, "learning_rate": 2.9142702918462394e-06, "loss": 1.8435, "step": 30758 }, { "epoch": 0.76, "learning_rate": 2.913706982009581e-06, "loss": 1.7649, "step": 30759 }, { "epoch": 0.76, "learning_rate": 2.9131437173361766e-06, "loss": 1.9164, "step": 30760 }, { "epoch": 0.76, "learning_rate": 2.912580497829619e-06, "loss": 2.0702, "step": 30761 }, { "epoch": 0.76, "learning_rate": 2.9120173234934933e-06, "loss": 1.9936, "step": 30762 }, { "epoch": 0.76, "learning_rate": 2.911454194331391e-06, "loss": 2.0371, "step": 30763 }, { "epoch": 0.76, "learning_rate": 2.910891110346905e-06, "loss": 2.0265, "step": 30764 }, { "epoch": 0.76, "learning_rate": 2.9103280715436166e-06, "loss": 1.9059, "step": 30765 }, { "epoch": 0.76, "learning_rate": 2.9097650779251187e-06, "loss": 2.0197, "step": 30766 }, { "epoch": 0.76, "learning_rate": 2.909202129495e-06, "loss": 1.9898, "step": 30767 }, { "epoch": 0.76, "learning_rate": 2.908639226256845e-06, "loss": 2.0312, "step": 30768 }, { "epoch": 0.76, "learning_rate": 2.9080763682142464e-06, "loss": 1.636, "step": 30769 }, { "epoch": 0.76, "learning_rate": 2.9075135553707847e-06, "loss": 1.9585, "step": 30770 }, { "epoch": 0.76, "learning_rate": 2.9069507877300507e-06, "loss": 1.9017, "step": 30771 }, { "epoch": 0.76, "learning_rate": 2.9063880652956344e-06, "loss": 1.9624, "step": 30772 }, { "epoch": 0.76, "learning_rate": 2.905825388071115e-06, "loss": 2.0377, "step": 30773 }, { "epoch": 0.76, "learning_rate": 2.9052627560600855e-06, "loss": 2.033, "step": 30774 }, { "epoch": 0.76, "learning_rate": 2.904700169266126e-06, "loss": 2.0267, "step": 30775 }, { "epoch": 0.76, "learning_rate": 2.9041376276928235e-06, "loss": 1.9943, "step": 30776 }, { "epoch": 0.76, "learning_rate": 2.9035751313437655e-06, "loss": 1.9979, "step": 30777 }, { "epoch": 0.76, "learning_rate": 2.9030126802225344e-06, "loss": 1.8542, "step": 30778 }, { "epoch": 0.76, "learning_rate": 2.902450274332721e-06, "loss": 2.0221, "step": 30779 }, { "epoch": 0.76, "learning_rate": 2.901887913677901e-06, "loss": 1.9867, "step": 30780 }, { "epoch": 0.76, "learning_rate": 2.901325598261663e-06, "loss": 1.9255, "step": 30781 }, { "epoch": 0.76, "learning_rate": 2.9007633280875936e-06, "loss": 1.9966, "step": 30782 }, { "epoch": 0.76, "learning_rate": 2.9002011031592694e-06, "loss": 1.9892, "step": 30783 }, { "epoch": 0.76, "learning_rate": 2.8996389234802812e-06, "loss": 1.9765, "step": 30784 }, { "epoch": 0.76, "learning_rate": 2.899076789054205e-06, "loss": 2.0096, "step": 30785 }, { "epoch": 0.76, "learning_rate": 2.898514699884626e-06, "loss": 1.773, "step": 30786 }, { "epoch": 0.76, "learning_rate": 2.8979526559751292e-06, "loss": 1.9003, "step": 30787 }, { "epoch": 0.76, "learning_rate": 2.8973906573292933e-06, "loss": 2.2275, "step": 30788 }, { "epoch": 0.76, "learning_rate": 2.896828703950704e-06, "loss": 2.029, "step": 30789 }, { "epoch": 0.76, "learning_rate": 2.8962667958429367e-06, "loss": 1.8418, "step": 30790 }, { "epoch": 0.76, "learning_rate": 2.8957049330095766e-06, "loss": 1.8993, "step": 30791 }, { "epoch": 0.76, "learning_rate": 2.8951431154542064e-06, "loss": 1.8214, "step": 30792 }, { "epoch": 0.76, "learning_rate": 2.8945813431804027e-06, "loss": 2.0554, "step": 30793 }, { "epoch": 0.76, "learning_rate": 2.894019616191747e-06, "loss": 1.825, "step": 30794 }, { "epoch": 0.76, "learning_rate": 2.89345793449182e-06, "loss": 2.05, "step": 30795 }, { "epoch": 0.76, "learning_rate": 2.892896298084201e-06, "loss": 2.0213, "step": 30796 }, { "epoch": 0.76, "learning_rate": 2.892334706972475e-06, "loss": 1.8668, "step": 30797 }, { "epoch": 0.76, "learning_rate": 2.891773161160212e-06, "loss": 1.9022, "step": 30798 }, { "epoch": 0.76, "learning_rate": 2.8912116606509977e-06, "loss": 2.133, "step": 30799 }, { "epoch": 0.76, "learning_rate": 2.8906502054484055e-06, "loss": 2.0206, "step": 30800 }, { "epoch": 0.76, "learning_rate": 2.8900887955560163e-06, "loss": 1.9707, "step": 30801 }, { "epoch": 0.76, "learning_rate": 2.8895274309774123e-06, "loss": 2.0286, "step": 30802 }, { "epoch": 0.76, "learning_rate": 2.8889661117161627e-06, "loss": 1.8628, "step": 30803 }, { "epoch": 0.76, "learning_rate": 2.888404837775853e-06, "loss": 2.0925, "step": 30804 }, { "epoch": 0.76, "learning_rate": 2.8878436091600547e-06, "loss": 1.8697, "step": 30805 }, { "epoch": 0.76, "learning_rate": 2.8872824258723465e-06, "loss": 2.0388, "step": 30806 }, { "epoch": 0.76, "learning_rate": 2.8867212879163076e-06, "loss": 1.8402, "step": 30807 }, { "epoch": 0.76, "learning_rate": 2.88616019529551e-06, "loss": 2.0718, "step": 30808 }, { "epoch": 0.76, "learning_rate": 2.8855991480135337e-06, "loss": 2.0698, "step": 30809 }, { "epoch": 0.76, "learning_rate": 2.8850381460739506e-06, "loss": 1.9964, "step": 30810 }, { "epoch": 0.76, "learning_rate": 2.8844771894803382e-06, "loss": 2.029, "step": 30811 }, { "epoch": 0.76, "learning_rate": 2.8839162782362705e-06, "loss": 1.9561, "step": 30812 }, { "epoch": 0.76, "learning_rate": 2.8833554123453246e-06, "loss": 1.9653, "step": 30813 }, { "epoch": 0.76, "learning_rate": 2.8827945918110757e-06, "loss": 2.0621, "step": 30814 }, { "epoch": 0.76, "learning_rate": 2.8822338166370945e-06, "loss": 1.9132, "step": 30815 }, { "epoch": 0.76, "learning_rate": 2.881673086826956e-06, "loss": 2.0128, "step": 30816 }, { "epoch": 0.76, "learning_rate": 2.8811124023842376e-06, "loss": 1.9273, "step": 30817 }, { "epoch": 0.76, "learning_rate": 2.880551763312507e-06, "loss": 2.002, "step": 30818 }, { "epoch": 0.76, "learning_rate": 2.8799911696153436e-06, "loss": 1.8577, "step": 30819 }, { "epoch": 0.76, "learning_rate": 2.879430621296314e-06, "loss": 2.0292, "step": 30820 }, { "epoch": 0.76, "learning_rate": 2.8788701183589928e-06, "loss": 1.9955, "step": 30821 }, { "epoch": 0.76, "learning_rate": 2.878309660806956e-06, "loss": 2.0802, "step": 30822 }, { "epoch": 0.76, "learning_rate": 2.877749248643771e-06, "loss": 1.8471, "step": 30823 }, { "epoch": 0.76, "learning_rate": 2.8771888818730097e-06, "loss": 2.099, "step": 30824 }, { "epoch": 0.76, "learning_rate": 2.876628560498248e-06, "loss": 2.0366, "step": 30825 }, { "epoch": 0.76, "learning_rate": 2.8760682845230513e-06, "loss": 2.0431, "step": 30826 }, { "epoch": 0.76, "learning_rate": 2.8755080539509962e-06, "loss": 2.0227, "step": 30827 }, { "epoch": 0.76, "learning_rate": 2.874947868785647e-06, "loss": 1.952, "step": 30828 }, { "epoch": 0.76, "learning_rate": 2.8743877290305764e-06, "loss": 2.0505, "step": 30829 }, { "epoch": 0.76, "learning_rate": 2.8738276346893556e-06, "loss": 2.0257, "step": 30830 }, { "epoch": 0.76, "learning_rate": 2.8732675857655536e-06, "loss": 2.0362, "step": 30831 }, { "epoch": 0.76, "learning_rate": 2.8727075822627426e-06, "loss": 2.0038, "step": 30832 }, { "epoch": 0.76, "learning_rate": 2.872147624184486e-06, "loss": 1.9499, "step": 30833 }, { "epoch": 0.76, "learning_rate": 2.871587711534356e-06, "loss": 1.814, "step": 30834 }, { "epoch": 0.76, "learning_rate": 2.8710278443159224e-06, "loss": 1.8562, "step": 30835 }, { "epoch": 0.76, "learning_rate": 2.8704680225327496e-06, "loss": 1.9854, "step": 30836 }, { "epoch": 0.76, "learning_rate": 2.869908246188411e-06, "loss": 1.9656, "step": 30837 }, { "epoch": 0.76, "learning_rate": 2.8693485152864662e-06, "loss": 1.8815, "step": 30838 }, { "epoch": 0.76, "learning_rate": 2.868788829830488e-06, "loss": 1.8974, "step": 30839 }, { "epoch": 0.76, "learning_rate": 2.8682291898240456e-06, "loss": 1.7949, "step": 30840 }, { "epoch": 0.76, "learning_rate": 2.8676695952707003e-06, "loss": 1.8534, "step": 30841 }, { "epoch": 0.76, "learning_rate": 2.867110046174023e-06, "loss": 2.078, "step": 30842 }, { "epoch": 0.76, "learning_rate": 2.866550542537575e-06, "loss": 2.0008, "step": 30843 }, { "epoch": 0.76, "learning_rate": 2.865991084364925e-06, "loss": 1.9673, "step": 30844 }, { "epoch": 0.76, "learning_rate": 2.865431671659643e-06, "loss": 1.9609, "step": 30845 }, { "epoch": 0.76, "learning_rate": 2.8648723044252857e-06, "loss": 1.8416, "step": 30846 }, { "epoch": 0.76, "learning_rate": 2.8643129826654237e-06, "loss": 2.1006, "step": 30847 }, { "epoch": 0.76, "learning_rate": 2.8637537063836195e-06, "loss": 1.9627, "step": 30848 }, { "epoch": 0.76, "learning_rate": 2.8631944755834383e-06, "loss": 1.8545, "step": 30849 }, { "epoch": 0.76, "learning_rate": 2.8626352902684483e-06, "loss": 1.962, "step": 30850 }, { "epoch": 0.76, "learning_rate": 2.862076150442207e-06, "loss": 1.9109, "step": 30851 }, { "epoch": 0.76, "learning_rate": 2.8615170561082827e-06, "loss": 1.8654, "step": 30852 }, { "epoch": 0.76, "learning_rate": 2.8609580072702336e-06, "loss": 1.8538, "step": 30853 }, { "epoch": 0.76, "learning_rate": 2.8603990039316266e-06, "loss": 2.1902, "step": 30854 }, { "epoch": 0.76, "learning_rate": 2.859840046096025e-06, "loss": 1.8285, "step": 30855 }, { "epoch": 0.76, "learning_rate": 2.859281133766988e-06, "loss": 1.9341, "step": 30856 }, { "epoch": 0.76, "learning_rate": 2.8587222669480818e-06, "loss": 1.8366, "step": 30857 }, { "epoch": 0.76, "learning_rate": 2.858163445642863e-06, "loss": 1.952, "step": 30858 }, { "epoch": 0.76, "learning_rate": 2.8576046698548967e-06, "loss": 2.0028, "step": 30859 }, { "epoch": 0.76, "learning_rate": 2.857045939587745e-06, "loss": 1.8063, "step": 30860 }, { "epoch": 0.76, "learning_rate": 2.856487254844965e-06, "loss": 1.8545, "step": 30861 }, { "epoch": 0.76, "learning_rate": 2.8559286156301237e-06, "loss": 1.9461, "step": 30862 }, { "epoch": 0.76, "learning_rate": 2.855370021946773e-06, "loss": 1.934, "step": 30863 }, { "epoch": 0.76, "learning_rate": 2.854811473798478e-06, "loss": 1.9416, "step": 30864 }, { "epoch": 0.76, "learning_rate": 2.854252971188799e-06, "loss": 1.865, "step": 30865 }, { "epoch": 0.76, "learning_rate": 2.8536945141212935e-06, "loss": 2.1096, "step": 30866 }, { "epoch": 0.76, "learning_rate": 2.8531361025995254e-06, "loss": 2.0423, "step": 30867 }, { "epoch": 0.76, "learning_rate": 2.852577736627046e-06, "loss": 1.689, "step": 30868 }, { "epoch": 0.76, "learning_rate": 2.852019416207419e-06, "loss": 1.9855, "step": 30869 }, { "epoch": 0.76, "learning_rate": 2.851461141344204e-06, "loss": 2.1265, "step": 30870 }, { "epoch": 0.76, "learning_rate": 2.8509029120409537e-06, "loss": 1.8968, "step": 30871 }, { "epoch": 0.76, "learning_rate": 2.850344728301233e-06, "loss": 2.0167, "step": 30872 }, { "epoch": 0.76, "learning_rate": 2.8497865901285915e-06, "loss": 1.9439, "step": 30873 }, { "epoch": 0.76, "learning_rate": 2.8492284975265903e-06, "loss": 1.906, "step": 30874 }, { "epoch": 0.76, "learning_rate": 2.8486704504987894e-06, "loss": 2.0159, "step": 30875 }, { "epoch": 0.76, "learning_rate": 2.84811244904874e-06, "loss": 2.0763, "step": 30876 }, { "epoch": 0.76, "learning_rate": 2.8475544931800026e-06, "loss": 1.7964, "step": 30877 }, { "epoch": 0.76, "learning_rate": 2.8469965828961288e-06, "loss": 1.832, "step": 30878 }, { "epoch": 0.76, "learning_rate": 2.846438718200677e-06, "loss": 1.9361, "step": 30879 }, { "epoch": 0.76, "learning_rate": 2.8458808990972055e-06, "loss": 1.9577, "step": 30880 }, { "epoch": 0.76, "learning_rate": 2.845323125589263e-06, "loss": 1.8671, "step": 30881 }, { "epoch": 0.76, "learning_rate": 2.844765397680408e-06, "loss": 1.9318, "step": 30882 }, { "epoch": 0.76, "learning_rate": 2.8442077153741954e-06, "loss": 1.9347, "step": 30883 }, { "epoch": 0.76, "learning_rate": 2.8436500786741785e-06, "loss": 1.9439, "step": 30884 }, { "epoch": 0.76, "learning_rate": 2.843092487583915e-06, "loss": 1.9281, "step": 30885 }, { "epoch": 0.76, "learning_rate": 2.8425349421069525e-06, "loss": 1.9778, "step": 30886 }, { "epoch": 0.76, "learning_rate": 2.841977442246847e-06, "loss": 2.0228, "step": 30887 }, { "epoch": 0.76, "learning_rate": 2.841419988007156e-06, "loss": 1.7806, "step": 30888 }, { "epoch": 0.76, "learning_rate": 2.8408625793914246e-06, "loss": 1.968, "step": 30889 }, { "epoch": 0.76, "learning_rate": 2.840305216403213e-06, "loss": 1.8802, "step": 30890 }, { "epoch": 0.76, "learning_rate": 2.839747899046065e-06, "loss": 1.967, "step": 30891 }, { "epoch": 0.76, "learning_rate": 2.839190627323538e-06, "loss": 1.904, "step": 30892 }, { "epoch": 0.76, "learning_rate": 2.8386334012391857e-06, "loss": 1.8869, "step": 30893 }, { "epoch": 0.76, "learning_rate": 2.838076220796554e-06, "loss": 1.7583, "step": 30894 }, { "epoch": 0.76, "learning_rate": 2.8375190859991986e-06, "loss": 2.0049, "step": 30895 }, { "epoch": 0.76, "learning_rate": 2.8369619968506658e-06, "loss": 1.923, "step": 30896 }, { "epoch": 0.76, "learning_rate": 2.836404953354508e-06, "loss": 2.1054, "step": 30897 }, { "epoch": 0.76, "learning_rate": 2.8358479555142795e-06, "loss": 1.9988, "step": 30898 }, { "epoch": 0.76, "learning_rate": 2.8352910033335233e-06, "loss": 1.9215, "step": 30899 }, { "epoch": 0.76, "learning_rate": 2.8347340968157933e-06, "loss": 1.817, "step": 30900 }, { "epoch": 0.76, "learning_rate": 2.834177235964638e-06, "loss": 1.951, "step": 30901 }, { "epoch": 0.76, "learning_rate": 2.833620420783605e-06, "loss": 2.0303, "step": 30902 }, { "epoch": 0.76, "learning_rate": 2.8330636512762488e-06, "loss": 1.8216, "step": 30903 }, { "epoch": 0.76, "learning_rate": 2.8325069274461105e-06, "loss": 1.8306, "step": 30904 }, { "epoch": 0.76, "learning_rate": 2.8319502492967444e-06, "loss": 1.9866, "step": 30905 }, { "epoch": 0.76, "learning_rate": 2.8313936168316913e-06, "loss": 1.9301, "step": 30906 }, { "epoch": 0.76, "learning_rate": 2.830837030054504e-06, "loss": 2.0774, "step": 30907 }, { "epoch": 0.76, "learning_rate": 2.8302804889687306e-06, "loss": 1.9123, "step": 30908 }, { "epoch": 0.76, "learning_rate": 2.8297239935779144e-06, "loss": 2.2389, "step": 30909 }, { "epoch": 0.76, "learning_rate": 2.829167543885607e-06, "loss": 1.9084, "step": 30910 }, { "epoch": 0.76, "learning_rate": 2.8286111398953475e-06, "loss": 1.8535, "step": 30911 }, { "epoch": 0.76, "learning_rate": 2.8280547816106873e-06, "loss": 2.1657, "step": 30912 }, { "epoch": 0.76, "learning_rate": 2.8274984690351738e-06, "loss": 1.8387, "step": 30913 }, { "epoch": 0.76, "learning_rate": 2.826942202172347e-06, "loss": 1.9204, "step": 30914 }, { "epoch": 0.76, "learning_rate": 2.8263859810257553e-06, "loss": 1.9081, "step": 30915 }, { "epoch": 0.76, "learning_rate": 2.8258298055989464e-06, "loss": 1.9949, "step": 30916 }, { "epoch": 0.76, "learning_rate": 2.8252736758954593e-06, "loss": 1.8882, "step": 30917 }, { "epoch": 0.76, "learning_rate": 2.824717591918841e-06, "loss": 2.144, "step": 30918 }, { "epoch": 0.76, "learning_rate": 2.8241615536726362e-06, "loss": 1.7884, "step": 30919 }, { "epoch": 0.76, "learning_rate": 2.823605561160391e-06, "loss": 1.9862, "step": 30920 }, { "epoch": 0.76, "learning_rate": 2.823049614385643e-06, "loss": 1.8192, "step": 30921 }, { "epoch": 0.76, "learning_rate": 2.8224937133519393e-06, "loss": 1.9562, "step": 30922 }, { "epoch": 0.76, "learning_rate": 2.8219378580628252e-06, "loss": 1.9194, "step": 30923 }, { "epoch": 0.76, "learning_rate": 2.821382048521838e-06, "loss": 1.8846, "step": 30924 }, { "epoch": 0.76, "learning_rate": 2.820826284732524e-06, "loss": 1.9484, "step": 30925 }, { "epoch": 0.76, "learning_rate": 2.820270566698422e-06, "loss": 2.0784, "step": 30926 }, { "epoch": 0.76, "learning_rate": 2.8197148944230746e-06, "loss": 1.8842, "step": 30927 }, { "epoch": 0.76, "learning_rate": 2.819159267910028e-06, "loss": 2.0843, "step": 30928 }, { "epoch": 0.76, "learning_rate": 2.8186036871628166e-06, "loss": 1.8434, "step": 30929 }, { "epoch": 0.76, "learning_rate": 2.8180481521849866e-06, "loss": 2.0018, "step": 30930 }, { "epoch": 0.76, "learning_rate": 2.8174926629800734e-06, "loss": 2.0531, "step": 30931 }, { "epoch": 0.76, "learning_rate": 2.8169372195516197e-06, "loss": 1.9469, "step": 30932 }, { "epoch": 0.76, "learning_rate": 2.816381821903167e-06, "loss": 1.9667, "step": 30933 }, { "epoch": 0.76, "learning_rate": 2.8158264700382565e-06, "loss": 1.9799, "step": 30934 }, { "epoch": 0.76, "learning_rate": 2.8152711639604215e-06, "loss": 2.0101, "step": 30935 }, { "epoch": 0.76, "learning_rate": 2.8147159036732054e-06, "loss": 1.8956, "step": 30936 }, { "epoch": 0.76, "learning_rate": 2.8141606891801467e-06, "loss": 1.9501, "step": 30937 }, { "epoch": 0.76, "learning_rate": 2.8136055204847868e-06, "loss": 2.1235, "step": 30938 }, { "epoch": 0.76, "learning_rate": 2.8130503975906574e-06, "loss": 2.0179, "step": 30939 }, { "epoch": 0.76, "learning_rate": 2.8124953205013005e-06, "loss": 2.0757, "step": 30940 }, { "epoch": 0.76, "learning_rate": 2.811940289220256e-06, "loss": 1.9348, "step": 30941 }, { "epoch": 0.76, "learning_rate": 2.8113853037510553e-06, "loss": 1.7821, "step": 30942 }, { "epoch": 0.76, "learning_rate": 2.8108303640972422e-06, "loss": 1.8873, "step": 30943 }, { "epoch": 0.76, "learning_rate": 2.810275470262347e-06, "loss": 2.0644, "step": 30944 }, { "epoch": 0.76, "learning_rate": 2.8097206222499086e-06, "loss": 2.077, "step": 30945 }, { "epoch": 0.76, "learning_rate": 2.8091658200634676e-06, "loss": 1.8218, "step": 30946 }, { "epoch": 0.76, "learning_rate": 2.8086110637065535e-06, "loss": 1.9201, "step": 30947 }, { "epoch": 0.76, "learning_rate": 2.8080563531827078e-06, "loss": 1.9254, "step": 30948 }, { "epoch": 0.76, "learning_rate": 2.807501688495459e-06, "loss": 1.9532, "step": 30949 }, { "epoch": 0.76, "learning_rate": 2.806947069648346e-06, "loss": 1.9081, "step": 30950 }, { "epoch": 0.76, "learning_rate": 2.806392496644906e-06, "loss": 1.796, "step": 30951 }, { "epoch": 0.76, "learning_rate": 2.8058379694886697e-06, "loss": 2.0165, "step": 30952 }, { "epoch": 0.76, "learning_rate": 2.8052834881831716e-06, "loss": 2.0108, "step": 30953 }, { "epoch": 0.76, "learning_rate": 2.804729052731946e-06, "loss": 1.967, "step": 30954 }, { "epoch": 0.76, "learning_rate": 2.804174663138527e-06, "loss": 1.939, "step": 30955 }, { "epoch": 0.76, "learning_rate": 2.8036203194064516e-06, "loss": 1.9129, "step": 30956 }, { "epoch": 0.76, "learning_rate": 2.803066021539246e-06, "loss": 1.9318, "step": 30957 }, { "epoch": 0.76, "learning_rate": 2.802511769540448e-06, "loss": 2.0, "step": 30958 }, { "epoch": 0.76, "learning_rate": 2.8019575634135866e-06, "loss": 1.8815, "step": 30959 }, { "epoch": 0.76, "learning_rate": 2.801403403162195e-06, "loss": 2.0784, "step": 30960 }, { "epoch": 0.76, "learning_rate": 2.8008492887898074e-06, "loss": 1.9518, "step": 30961 }, { "epoch": 0.76, "learning_rate": 2.8002952202999523e-06, "loss": 1.7798, "step": 30962 }, { "epoch": 0.76, "learning_rate": 2.799741197696164e-06, "loss": 1.8405, "step": 30963 }, { "epoch": 0.76, "learning_rate": 2.7991872209819683e-06, "loss": 1.9349, "step": 30964 }, { "epoch": 0.76, "learning_rate": 2.7986332901608995e-06, "loss": 1.8324, "step": 30965 }, { "epoch": 0.76, "learning_rate": 2.79807940523649e-06, "loss": 1.8974, "step": 30966 }, { "epoch": 0.76, "learning_rate": 2.7975255662122657e-06, "loss": 1.8472, "step": 30967 }, { "epoch": 0.76, "learning_rate": 2.7969717730917566e-06, "loss": 1.8539, "step": 30968 }, { "epoch": 0.76, "learning_rate": 2.7964180258784978e-06, "loss": 2.1154, "step": 30969 }, { "epoch": 0.76, "learning_rate": 2.795864324576012e-06, "loss": 1.9888, "step": 30970 }, { "epoch": 0.76, "learning_rate": 2.7953106691878298e-06, "loss": 2.0126, "step": 30971 }, { "epoch": 0.76, "learning_rate": 2.794757059717481e-06, "loss": 1.8579, "step": 30972 }, { "epoch": 0.76, "learning_rate": 2.794203496168496e-06, "loss": 2.0137, "step": 30973 }, { "epoch": 0.76, "learning_rate": 2.7936499785443983e-06, "loss": 1.9969, "step": 30974 }, { "epoch": 0.76, "learning_rate": 2.793096506848717e-06, "loss": 2.065, "step": 30975 }, { "epoch": 0.76, "learning_rate": 2.792543081084984e-06, "loss": 1.9597, "step": 30976 }, { "epoch": 0.76, "learning_rate": 2.79198970125672e-06, "loss": 1.8575, "step": 30977 }, { "epoch": 0.76, "learning_rate": 2.7914363673674573e-06, "loss": 1.9603, "step": 30978 }, { "epoch": 0.76, "learning_rate": 2.7908830794207166e-06, "loss": 1.9667, "step": 30979 }, { "epoch": 0.76, "learning_rate": 2.790329837420027e-06, "loss": 1.8153, "step": 30980 }, { "epoch": 0.76, "learning_rate": 2.7897766413689187e-06, "loss": 1.8181, "step": 30981 }, { "epoch": 0.76, "learning_rate": 2.789223491270909e-06, "loss": 1.9241, "step": 30982 }, { "epoch": 0.76, "learning_rate": 2.788670387129533e-06, "loss": 2.0814, "step": 30983 }, { "epoch": 0.76, "learning_rate": 2.7881173289483065e-06, "loss": 1.9937, "step": 30984 }, { "epoch": 0.76, "learning_rate": 2.78756431673076e-06, "loss": 1.9533, "step": 30985 }, { "epoch": 0.76, "learning_rate": 2.787011350480415e-06, "loss": 1.9329, "step": 30986 }, { "epoch": 0.76, "learning_rate": 2.7864584302008003e-06, "loss": 1.8749, "step": 30987 }, { "epoch": 0.76, "learning_rate": 2.785905555895435e-06, "loss": 1.8414, "step": 30988 }, { "epoch": 0.76, "learning_rate": 2.785352727567844e-06, "loss": 1.8281, "step": 30989 }, { "epoch": 0.76, "learning_rate": 2.7847999452215513e-06, "loss": 1.9003, "step": 30990 }, { "epoch": 0.76, "learning_rate": 2.7842472088600824e-06, "loss": 1.9392, "step": 30991 }, { "epoch": 0.76, "learning_rate": 2.783694518486956e-06, "loss": 1.8533, "step": 30992 }, { "epoch": 0.76, "learning_rate": 2.783141874105697e-06, "loss": 2.0742, "step": 30993 }, { "epoch": 0.76, "learning_rate": 2.782589275719826e-06, "loss": 2.0015, "step": 30994 }, { "epoch": 0.76, "learning_rate": 2.7820367233328637e-06, "loss": 2.16, "step": 30995 }, { "epoch": 0.76, "learning_rate": 2.781484216948338e-06, "loss": 1.9093, "step": 30996 }, { "epoch": 0.76, "learning_rate": 2.7809317565697626e-06, "loss": 1.9999, "step": 30997 }, { "epoch": 0.76, "learning_rate": 2.780379342200664e-06, "loss": 1.9266, "step": 30998 }, { "epoch": 0.76, "learning_rate": 2.7798269738445583e-06, "loss": 1.9683, "step": 30999 }, { "epoch": 0.76, "learning_rate": 2.7792746515049683e-06, "loss": 2.0003, "step": 31000 }, { "epoch": 0.76, "learning_rate": 2.7787223751854165e-06, "loss": 2.0201, "step": 31001 }, { "epoch": 0.76, "learning_rate": 2.778170144889418e-06, "loss": 2.1047, "step": 31002 }, { "epoch": 0.76, "learning_rate": 2.7776179606204946e-06, "loss": 1.9218, "step": 31003 }, { "epoch": 0.76, "learning_rate": 2.7770658223821656e-06, "loss": 1.7553, "step": 31004 }, { "epoch": 0.76, "learning_rate": 2.776513730177952e-06, "loss": 1.9355, "step": 31005 }, { "epoch": 0.76, "learning_rate": 2.775961684011369e-06, "loss": 1.9398, "step": 31006 }, { "epoch": 0.76, "learning_rate": 2.7754096838859357e-06, "loss": 1.9745, "step": 31007 }, { "epoch": 0.76, "learning_rate": 2.7748577298051715e-06, "loss": 1.8986, "step": 31008 }, { "epoch": 0.76, "learning_rate": 2.774305821772596e-06, "loss": 1.9072, "step": 31009 }, { "epoch": 0.76, "learning_rate": 2.773753959791721e-06, "loss": 2.0023, "step": 31010 }, { "epoch": 0.76, "learning_rate": 2.7732021438660707e-06, "loss": 1.8662, "step": 31011 }, { "epoch": 0.76, "learning_rate": 2.772650373999155e-06, "loss": 2.0231, "step": 31012 }, { "epoch": 0.76, "learning_rate": 2.7720986501944937e-06, "loss": 1.98, "step": 31013 }, { "epoch": 0.76, "learning_rate": 2.7715469724556065e-06, "loss": 1.9223, "step": 31014 }, { "epoch": 0.76, "learning_rate": 2.7709953407860024e-06, "loss": 1.9459, "step": 31015 }, { "epoch": 0.76, "learning_rate": 2.770443755189205e-06, "loss": 1.7462, "step": 31016 }, { "epoch": 0.76, "learning_rate": 2.7698922156687213e-06, "loss": 1.8654, "step": 31017 }, { "epoch": 0.76, "learning_rate": 2.769340722228071e-06, "loss": 2.021, "step": 31018 }, { "epoch": 0.76, "learning_rate": 2.7687892748707722e-06, "loss": 2.1115, "step": 31019 }, { "epoch": 0.76, "learning_rate": 2.7682378736003334e-06, "loss": 1.9277, "step": 31020 }, { "epoch": 0.76, "learning_rate": 2.7676865184202704e-06, "loss": 1.9639, "step": 31021 }, { "epoch": 0.76, "learning_rate": 2.7671352093340985e-06, "loss": 1.9043, "step": 31022 }, { "epoch": 0.76, "learning_rate": 2.766583946345335e-06, "loss": 1.8831, "step": 31023 }, { "epoch": 0.76, "learning_rate": 2.766032729457485e-06, "loss": 1.9998, "step": 31024 }, { "epoch": 0.76, "learning_rate": 2.7654815586740657e-06, "loss": 1.938, "step": 31025 }, { "epoch": 0.76, "learning_rate": 2.764930433998594e-06, "loss": 1.8363, "step": 31026 }, { "epoch": 0.76, "learning_rate": 2.7643793554345757e-06, "loss": 1.9915, "step": 31027 }, { "epoch": 0.76, "learning_rate": 2.7638283229855247e-06, "loss": 1.9641, "step": 31028 }, { "epoch": 0.76, "learning_rate": 2.7632773366549583e-06, "loss": 2.0851, "step": 31029 }, { "epoch": 0.76, "learning_rate": 2.762726396446379e-06, "loss": 1.888, "step": 31030 }, { "epoch": 0.76, "learning_rate": 2.762175502363308e-06, "loss": 1.7655, "step": 31031 }, { "epoch": 0.76, "learning_rate": 2.7616246544092463e-06, "loss": 1.8944, "step": 31032 }, { "epoch": 0.76, "learning_rate": 2.7610738525877114e-06, "loss": 1.9787, "step": 31033 }, { "epoch": 0.76, "learning_rate": 2.7605230969022133e-06, "loss": 1.9627, "step": 31034 }, { "epoch": 0.76, "learning_rate": 2.759972387356259e-06, "loss": 2.0625, "step": 31035 }, { "epoch": 0.76, "learning_rate": 2.7594217239533626e-06, "loss": 1.9589, "step": 31036 }, { "epoch": 0.77, "learning_rate": 2.7588711066970285e-06, "loss": 1.9123, "step": 31037 }, { "epoch": 0.77, "learning_rate": 2.758320535590768e-06, "loss": 1.9438, "step": 31038 }, { "epoch": 0.77, "learning_rate": 2.757770010638091e-06, "loss": 1.8918, "step": 31039 }, { "epoch": 0.77, "learning_rate": 2.75721953184251e-06, "loss": 1.8739, "step": 31040 }, { "epoch": 0.77, "learning_rate": 2.7566690992075252e-06, "loss": 1.8902, "step": 31041 }, { "epoch": 0.77, "learning_rate": 2.7561187127366483e-06, "loss": 1.808, "step": 31042 }, { "epoch": 0.77, "learning_rate": 2.755568372433388e-06, "loss": 1.9915, "step": 31043 }, { "epoch": 0.77, "learning_rate": 2.755018078301255e-06, "loss": 1.9126, "step": 31044 }, { "epoch": 0.77, "learning_rate": 2.75446783034375e-06, "loss": 1.9389, "step": 31045 }, { "epoch": 0.77, "learning_rate": 2.7539176285643844e-06, "loss": 1.8625, "step": 31046 }, { "epoch": 0.77, "learning_rate": 2.7533674729666615e-06, "loss": 2.0019, "step": 31047 }, { "epoch": 0.77, "learning_rate": 2.752817363554089e-06, "loss": 2.0874, "step": 31048 }, { "epoch": 0.77, "learning_rate": 2.752267300330176e-06, "loss": 1.8728, "step": 31049 }, { "epoch": 0.77, "learning_rate": 2.751717283298422e-06, "loss": 2.0009, "step": 31050 }, { "epoch": 0.77, "learning_rate": 2.7511673124623394e-06, "loss": 2.1696, "step": 31051 }, { "epoch": 0.77, "learning_rate": 2.750617387825426e-06, "loss": 1.9437, "step": 31052 }, { "epoch": 0.77, "learning_rate": 2.7500675093911923e-06, "loss": 1.9758, "step": 31053 }, { "epoch": 0.77, "learning_rate": 2.7495176771631425e-06, "loss": 1.9421, "step": 31054 }, { "epoch": 0.77, "learning_rate": 2.748967891144777e-06, "loss": 1.9003, "step": 31055 }, { "epoch": 0.77, "learning_rate": 2.7484181513396023e-06, "loss": 1.9746, "step": 31056 }, { "epoch": 0.77, "learning_rate": 2.7478684577511215e-06, "loss": 1.8472, "step": 31057 }, { "epoch": 0.77, "learning_rate": 2.7473188103828417e-06, "loss": 1.958, "step": 31058 }, { "epoch": 0.77, "learning_rate": 2.746769209238259e-06, "loss": 2.0629, "step": 31059 }, { "epoch": 0.77, "learning_rate": 2.7462196543208806e-06, "loss": 2.025, "step": 31060 }, { "epoch": 0.77, "learning_rate": 2.7456701456342094e-06, "loss": 2.1187, "step": 31061 }, { "epoch": 0.77, "learning_rate": 2.745120683181748e-06, "loss": 1.9745, "step": 31062 }, { "epoch": 0.77, "learning_rate": 2.7445712669669945e-06, "loss": 1.838, "step": 31063 }, { "epoch": 0.77, "learning_rate": 2.744021896993455e-06, "loss": 2.0259, "step": 31064 }, { "epoch": 0.77, "learning_rate": 2.7434725732646262e-06, "loss": 1.9468, "step": 31065 }, { "epoch": 0.77, "learning_rate": 2.742923295784011e-06, "loss": 1.9165, "step": 31066 }, { "epoch": 0.77, "learning_rate": 2.7423740645551146e-06, "loss": 1.9908, "step": 31067 }, { "epoch": 0.77, "learning_rate": 2.7418248795814296e-06, "loss": 1.9914, "step": 31068 }, { "epoch": 0.77, "learning_rate": 2.7412757408664626e-06, "loss": 1.9605, "step": 31069 }, { "epoch": 0.77, "learning_rate": 2.7407266484137086e-06, "loss": 2.2619, "step": 31070 }, { "epoch": 0.77, "learning_rate": 2.740177602226669e-06, "loss": 2.1078, "step": 31071 }, { "epoch": 0.77, "learning_rate": 2.739628602308847e-06, "loss": 2.0375, "step": 31072 }, { "epoch": 0.77, "learning_rate": 2.739079648663735e-06, "loss": 1.9927, "step": 31073 }, { "epoch": 0.77, "learning_rate": 2.738530741294835e-06, "loss": 2.0617, "step": 31074 }, { "epoch": 0.77, "learning_rate": 2.7379818802056433e-06, "loss": 1.8484, "step": 31075 }, { "epoch": 0.77, "learning_rate": 2.737433065399665e-06, "loss": 1.8354, "step": 31076 }, { "epoch": 0.77, "learning_rate": 2.7368842968803876e-06, "loss": 1.847, "step": 31077 }, { "epoch": 0.77, "learning_rate": 2.736335574651314e-06, "loss": 1.9283, "step": 31078 }, { "epoch": 0.77, "learning_rate": 2.735786898715945e-06, "loss": 1.95, "step": 31079 }, { "epoch": 0.77, "learning_rate": 2.7352382690777702e-06, "loss": 1.9099, "step": 31080 }, { "epoch": 0.77, "learning_rate": 2.734689685740288e-06, "loss": 2.0773, "step": 31081 }, { "epoch": 0.77, "learning_rate": 2.7341411487070014e-06, "loss": 1.8733, "step": 31082 }, { "epoch": 0.77, "learning_rate": 2.7335926579813964e-06, "loss": 1.9394, "step": 31083 }, { "epoch": 0.77, "learning_rate": 2.7330442135669764e-06, "loss": 1.9718, "step": 31084 }, { "epoch": 0.77, "learning_rate": 2.732495815467232e-06, "loss": 1.8613, "step": 31085 }, { "epoch": 0.77, "learning_rate": 2.731947463685659e-06, "loss": 1.9834, "step": 31086 }, { "epoch": 0.77, "learning_rate": 2.731399158225757e-06, "loss": 2.0107, "step": 31087 }, { "epoch": 0.77, "learning_rate": 2.7308508990910144e-06, "loss": 1.8813, "step": 31088 }, { "epoch": 0.77, "learning_rate": 2.73030268628493e-06, "loss": 1.8848, "step": 31089 }, { "epoch": 0.77, "learning_rate": 2.729754519810993e-06, "loss": 2.0147, "step": 31090 }, { "epoch": 0.77, "learning_rate": 2.7292063996727e-06, "loss": 2.0121, "step": 31091 }, { "epoch": 0.77, "learning_rate": 2.7286583258735444e-06, "loss": 1.931, "step": 31092 }, { "epoch": 0.77, "learning_rate": 2.728110298417018e-06, "loss": 1.9043, "step": 31093 }, { "epoch": 0.77, "learning_rate": 2.727562317306619e-06, "loss": 1.9737, "step": 31094 }, { "epoch": 0.77, "learning_rate": 2.7270143825458316e-06, "loss": 1.8461, "step": 31095 }, { "epoch": 0.77, "learning_rate": 2.7264664941381514e-06, "loss": 1.8688, "step": 31096 }, { "epoch": 0.77, "learning_rate": 2.725918652087074e-06, "loss": 1.9386, "step": 31097 }, { "epoch": 0.77, "learning_rate": 2.7253708563960836e-06, "loss": 1.9727, "step": 31098 }, { "epoch": 0.77, "learning_rate": 2.7248231070686793e-06, "loss": 1.9372, "step": 31099 }, { "epoch": 0.77, "learning_rate": 2.7242754041083453e-06, "loss": 1.9872, "step": 31100 }, { "epoch": 0.77, "learning_rate": 2.7237277475185754e-06, "loss": 1.9109, "step": 31101 }, { "epoch": 0.77, "learning_rate": 2.723180137302862e-06, "loss": 1.907, "step": 31102 }, { "epoch": 0.77, "learning_rate": 2.7226325734646906e-06, "loss": 1.8432, "step": 31103 }, { "epoch": 0.77, "learning_rate": 2.7220850560075574e-06, "loss": 2.0518, "step": 31104 }, { "epoch": 0.77, "learning_rate": 2.7215375849349436e-06, "loss": 2.0012, "step": 31105 }, { "epoch": 0.77, "learning_rate": 2.720990160250343e-06, "loss": 1.882, "step": 31106 }, { "epoch": 0.77, "learning_rate": 2.7204427819572475e-06, "loss": 1.7272, "step": 31107 }, { "epoch": 0.77, "learning_rate": 2.7198954500591392e-06, "loss": 2.013, "step": 31108 }, { "epoch": 0.77, "learning_rate": 2.71934816455951e-06, "loss": 1.9969, "step": 31109 }, { "epoch": 0.77, "learning_rate": 2.7188009254618477e-06, "loss": 1.9246, "step": 31110 }, { "epoch": 0.77, "learning_rate": 2.71825373276964e-06, "loss": 2.0128, "step": 31111 }, { "epoch": 0.77, "learning_rate": 2.717706586486377e-06, "loss": 1.8144, "step": 31112 }, { "epoch": 0.77, "learning_rate": 2.7171594866155403e-06, "loss": 2.2112, "step": 31113 }, { "epoch": 0.77, "learning_rate": 2.7166124331606225e-06, "loss": 1.8817, "step": 31114 }, { "epoch": 0.77, "learning_rate": 2.716065426125104e-06, "loss": 1.9539, "step": 31115 }, { "epoch": 0.77, "learning_rate": 2.7155184655124757e-06, "loss": 1.8591, "step": 31116 }, { "epoch": 0.77, "learning_rate": 2.714971551326223e-06, "loss": 1.8571, "step": 31117 }, { "epoch": 0.77, "learning_rate": 2.7144246835698296e-06, "loss": 2.019, "step": 31118 }, { "epoch": 0.77, "learning_rate": 2.713877862246781e-06, "loss": 1.9193, "step": 31119 }, { "epoch": 0.77, "learning_rate": 2.713331087360567e-06, "loss": 1.9133, "step": 31120 }, { "epoch": 0.77, "learning_rate": 2.712784358914664e-06, "loss": 2.0103, "step": 31121 }, { "epoch": 0.77, "learning_rate": 2.7122376769125657e-06, "loss": 1.8727, "step": 31122 }, { "epoch": 0.77, "learning_rate": 2.711691041357748e-06, "loss": 1.9201, "step": 31123 }, { "epoch": 0.77, "learning_rate": 2.7111444522536987e-06, "loss": 1.924, "step": 31124 }, { "epoch": 0.77, "learning_rate": 2.7105979096039047e-06, "loss": 1.9595, "step": 31125 }, { "epoch": 0.77, "learning_rate": 2.7100514134118426e-06, "loss": 1.9919, "step": 31126 }, { "epoch": 0.77, "learning_rate": 2.709504963680998e-06, "loss": 2.113, "step": 31127 }, { "epoch": 0.77, "learning_rate": 2.7089585604148537e-06, "loss": 1.909, "step": 31128 }, { "epoch": 0.77, "learning_rate": 2.7084122036168967e-06, "loss": 2.0663, "step": 31129 }, { "epoch": 0.77, "learning_rate": 2.707865893290602e-06, "loss": 1.973, "step": 31130 }, { "epoch": 0.77, "learning_rate": 2.7073196294394532e-06, "loss": 2.0326, "step": 31131 }, { "epoch": 0.77, "learning_rate": 2.7067734120669374e-06, "loss": 2.0725, "step": 31132 }, { "epoch": 0.77, "learning_rate": 2.7062272411765277e-06, "loss": 2.0807, "step": 31133 }, { "epoch": 0.77, "learning_rate": 2.705681116771709e-06, "loss": 1.9637, "step": 31134 }, { "epoch": 0.77, "learning_rate": 2.705135038855965e-06, "loss": 1.9279, "step": 31135 }, { "epoch": 0.77, "learning_rate": 2.7045890074327686e-06, "loss": 2.2067, "step": 31136 }, { "epoch": 0.77, "learning_rate": 2.7040430225056082e-06, "loss": 1.8823, "step": 31137 }, { "epoch": 0.77, "learning_rate": 2.703497084077955e-06, "loss": 2.0323, "step": 31138 }, { "epoch": 0.77, "learning_rate": 2.7029511921532937e-06, "loss": 1.8744, "step": 31139 }, { "epoch": 0.77, "learning_rate": 2.702405346735105e-06, "loss": 2.034, "step": 31140 }, { "epoch": 0.77, "learning_rate": 2.701859547826863e-06, "loss": 1.9395, "step": 31141 }, { "epoch": 0.77, "learning_rate": 2.7013137954320512e-06, "loss": 1.9751, "step": 31142 }, { "epoch": 0.77, "learning_rate": 2.7007680895541423e-06, "loss": 2.0626, "step": 31143 }, { "epoch": 0.77, "learning_rate": 2.700222430196616e-06, "loss": 1.908, "step": 31144 }, { "epoch": 0.77, "learning_rate": 2.699676817362953e-06, "loss": 1.8358, "step": 31145 }, { "epoch": 0.77, "learning_rate": 2.6991312510566268e-06, "loss": 1.8483, "step": 31146 }, { "epoch": 0.77, "learning_rate": 2.6985857312811204e-06, "loss": 1.955, "step": 31147 }, { "epoch": 0.77, "learning_rate": 2.6980402580399033e-06, "loss": 1.8419, "step": 31148 }, { "epoch": 0.77, "learning_rate": 2.697494831336456e-06, "loss": 1.9317, "step": 31149 }, { "epoch": 0.77, "learning_rate": 2.6969494511742566e-06, "loss": 1.8851, "step": 31150 }, { "epoch": 0.77, "learning_rate": 2.6964041175567744e-06, "loss": 1.8342, "step": 31151 }, { "epoch": 0.77, "learning_rate": 2.695858830487493e-06, "loss": 1.7562, "step": 31152 }, { "epoch": 0.77, "learning_rate": 2.69531358996988e-06, "loss": 1.9558, "step": 31153 }, { "epoch": 0.77, "learning_rate": 2.6947683960074146e-06, "loss": 2.0697, "step": 31154 }, { "epoch": 0.77, "learning_rate": 2.6942232486035737e-06, "loss": 1.7275, "step": 31155 }, { "epoch": 0.77, "learning_rate": 2.693678147761826e-06, "loss": 2.0792, "step": 31156 }, { "epoch": 0.77, "learning_rate": 2.6931330934856516e-06, "loss": 1.9534, "step": 31157 }, { "epoch": 0.77, "learning_rate": 2.692588085778518e-06, "loss": 1.8277, "step": 31158 }, { "epoch": 0.77, "learning_rate": 2.6920431246439017e-06, "loss": 1.8148, "step": 31159 }, { "epoch": 0.77, "learning_rate": 2.6914982100852793e-06, "loss": 1.9386, "step": 31160 }, { "epoch": 0.77, "learning_rate": 2.6909533421061185e-06, "loss": 1.9884, "step": 31161 }, { "epoch": 0.77, "learning_rate": 2.690408520709893e-06, "loss": 2.0039, "step": 31162 }, { "epoch": 0.77, "learning_rate": 2.6898637459000766e-06, "loss": 1.8528, "step": 31163 }, { "epoch": 0.77, "learning_rate": 2.68931901768014e-06, "loss": 1.9722, "step": 31164 }, { "epoch": 0.77, "learning_rate": 2.68877433605356e-06, "loss": 1.9056, "step": 31165 }, { "epoch": 0.77, "learning_rate": 2.6882297010238e-06, "loss": 1.8365, "step": 31166 }, { "epoch": 0.77, "learning_rate": 2.6876851125943384e-06, "loss": 1.7633, "step": 31167 }, { "epoch": 0.77, "learning_rate": 2.6871405707686393e-06, "loss": 1.9694, "step": 31168 }, { "epoch": 0.77, "learning_rate": 2.6865960755501764e-06, "loss": 1.9321, "step": 31169 }, { "epoch": 0.77, "learning_rate": 2.686051626942423e-06, "loss": 2.1402, "step": 31170 }, { "epoch": 0.77, "learning_rate": 2.6855072249488432e-06, "loss": 1.9798, "step": 31171 }, { "epoch": 0.77, "learning_rate": 2.6849628695729136e-06, "loss": 2.0727, "step": 31172 }, { "epoch": 0.77, "learning_rate": 2.684418560818095e-06, "loss": 1.9601, "step": 31173 }, { "epoch": 0.77, "learning_rate": 2.683874298687862e-06, "loss": 1.9021, "step": 31174 }, { "epoch": 0.77, "learning_rate": 2.6833300831856846e-06, "loss": 1.7193, "step": 31175 }, { "epoch": 0.77, "learning_rate": 2.682785914315027e-06, "loss": 1.9511, "step": 31176 }, { "epoch": 0.77, "learning_rate": 2.682241792079359e-06, "loss": 1.7508, "step": 31177 }, { "epoch": 0.77, "learning_rate": 2.6816977164821513e-06, "loss": 2.0137, "step": 31178 }, { "epoch": 0.77, "learning_rate": 2.681153687526866e-06, "loss": 1.9925, "step": 31179 }, { "epoch": 0.77, "learning_rate": 2.6806097052169744e-06, "loss": 1.8982, "step": 31180 }, { "epoch": 0.77, "learning_rate": 2.680065769555942e-06, "loss": 1.8877, "step": 31181 }, { "epoch": 0.77, "learning_rate": 2.6795218805472355e-06, "loss": 2.0817, "step": 31182 }, { "epoch": 0.77, "learning_rate": 2.6789780381943253e-06, "loss": 1.9693, "step": 31183 }, { "epoch": 0.77, "learning_rate": 2.6784342425006706e-06, "loss": 1.9089, "step": 31184 }, { "epoch": 0.77, "learning_rate": 2.6778904934697437e-06, "loss": 1.8805, "step": 31185 }, { "epoch": 0.77, "learning_rate": 2.6773467911050035e-06, "loss": 1.7154, "step": 31186 }, { "epoch": 0.77, "learning_rate": 2.676803135409919e-06, "loss": 1.994, "step": 31187 }, { "epoch": 0.77, "learning_rate": 2.6762595263879576e-06, "loss": 1.9614, "step": 31188 }, { "epoch": 0.77, "learning_rate": 2.6757159640425778e-06, "loss": 1.7902, "step": 31189 }, { "epoch": 0.77, "learning_rate": 2.67517244837725e-06, "loss": 2.0127, "step": 31190 }, { "epoch": 0.77, "learning_rate": 2.6746289793954326e-06, "loss": 1.9289, "step": 31191 }, { "epoch": 0.77, "learning_rate": 2.674085557100592e-06, "loss": 1.9177, "step": 31192 }, { "epoch": 0.77, "learning_rate": 2.6735421814961935e-06, "loss": 1.911, "step": 31193 }, { "epoch": 0.77, "learning_rate": 2.6729988525856963e-06, "loss": 1.8258, "step": 31194 }, { "epoch": 0.77, "learning_rate": 2.6724555703725685e-06, "loss": 2.0663, "step": 31195 }, { "epoch": 0.77, "learning_rate": 2.671912334860266e-06, "loss": 1.9273, "step": 31196 }, { "epoch": 0.77, "learning_rate": 2.671369146052254e-06, "loss": 2.0357, "step": 31197 }, { "epoch": 0.77, "learning_rate": 2.670826003951995e-06, "loss": 1.9521, "step": 31198 }, { "epoch": 0.77, "learning_rate": 2.670282908562951e-06, "loss": 1.8759, "step": 31199 }, { "epoch": 0.77, "learning_rate": 2.669739859888585e-06, "loss": 1.8373, "step": 31200 }, { "epoch": 0.77, "learning_rate": 2.6691968579323526e-06, "loss": 1.9642, "step": 31201 }, { "epoch": 0.77, "learning_rate": 2.6686539026977186e-06, "loss": 1.9479, "step": 31202 }, { "epoch": 0.77, "learning_rate": 2.6681109941881444e-06, "loss": 1.8739, "step": 31203 }, { "epoch": 0.77, "learning_rate": 2.667568132407086e-06, "loss": 2.0144, "step": 31204 }, { "epoch": 0.77, "learning_rate": 2.6670253173580085e-06, "loss": 2.0054, "step": 31205 }, { "epoch": 0.77, "learning_rate": 2.6664825490443645e-06, "loss": 2.0895, "step": 31206 }, { "epoch": 0.77, "learning_rate": 2.665939827469618e-06, "loss": 1.8449, "step": 31207 }, { "epoch": 0.77, "learning_rate": 2.665397152637229e-06, "loss": 1.8377, "step": 31208 }, { "epoch": 0.77, "learning_rate": 2.664854524550653e-06, "loss": 2.0395, "step": 31209 }, { "epoch": 0.77, "learning_rate": 2.6643119432133514e-06, "loss": 2.0208, "step": 31210 }, { "epoch": 0.77, "learning_rate": 2.663769408628777e-06, "loss": 1.9698, "step": 31211 }, { "epoch": 0.77, "learning_rate": 2.6632269208003915e-06, "loss": 2.0255, "step": 31212 }, { "epoch": 0.77, "learning_rate": 2.6626844797316553e-06, "loss": 1.7755, "step": 31213 }, { "epoch": 0.77, "learning_rate": 2.662142085426018e-06, "loss": 1.9027, "step": 31214 }, { "epoch": 0.77, "learning_rate": 2.661599737886942e-06, "loss": 2.0394, "step": 31215 }, { "epoch": 0.77, "learning_rate": 2.661057437117881e-06, "loss": 1.7468, "step": 31216 }, { "epoch": 0.77, "learning_rate": 2.6605151831222932e-06, "loss": 1.8902, "step": 31217 }, { "epoch": 0.77, "learning_rate": 2.6599729759036375e-06, "loss": 2.002, "step": 31218 }, { "epoch": 0.77, "learning_rate": 2.6594308154653624e-06, "loss": 2.0722, "step": 31219 }, { "epoch": 0.77, "learning_rate": 2.65888870181093e-06, "loss": 1.8818, "step": 31220 }, { "epoch": 0.77, "learning_rate": 2.658346634943789e-06, "loss": 1.8959, "step": 31221 }, { "epoch": 0.77, "learning_rate": 2.657804614867399e-06, "loss": 1.8775, "step": 31222 }, { "epoch": 0.77, "learning_rate": 2.6572626415852145e-06, "loss": 1.9153, "step": 31223 }, { "epoch": 0.77, "learning_rate": 2.6567207151006857e-06, "loss": 1.97, "step": 31224 }, { "epoch": 0.77, "learning_rate": 2.6561788354172714e-06, "loss": 1.9199, "step": 31225 }, { "epoch": 0.77, "learning_rate": 2.6556370025384193e-06, "loss": 1.915, "step": 31226 }, { "epoch": 0.77, "learning_rate": 2.655095216467587e-06, "loss": 1.8593, "step": 31227 }, { "epoch": 0.77, "learning_rate": 2.654553477208229e-06, "loss": 2.0654, "step": 31228 }, { "epoch": 0.77, "learning_rate": 2.6540117847637924e-06, "loss": 1.8716, "step": 31229 }, { "epoch": 0.77, "learning_rate": 2.653470139137737e-06, "loss": 1.9534, "step": 31230 }, { "epoch": 0.77, "learning_rate": 2.652928540333506e-06, "loss": 1.8132, "step": 31231 }, { "epoch": 0.77, "learning_rate": 2.652386988354556e-06, "loss": 1.7957, "step": 31232 }, { "epoch": 0.77, "learning_rate": 2.6518454832043383e-06, "loss": 2.0729, "step": 31233 }, { "epoch": 0.77, "learning_rate": 2.6513040248863053e-06, "loss": 1.8237, "step": 31234 }, { "epoch": 0.77, "learning_rate": 2.6507626134039076e-06, "loss": 1.9469, "step": 31235 }, { "epoch": 0.77, "learning_rate": 2.650221248760593e-06, "loss": 2.1129, "step": 31236 }, { "epoch": 0.77, "learning_rate": 2.6496799309598133e-06, "loss": 1.9537, "step": 31237 }, { "epoch": 0.77, "learning_rate": 2.649138660005022e-06, "loss": 1.8135, "step": 31238 }, { "epoch": 0.77, "learning_rate": 2.6485974358996625e-06, "loss": 1.9872, "step": 31239 }, { "epoch": 0.77, "learning_rate": 2.648056258647187e-06, "loss": 1.9218, "step": 31240 }, { "epoch": 0.77, "learning_rate": 2.647515128251049e-06, "loss": 2.1084, "step": 31241 }, { "epoch": 0.77, "learning_rate": 2.6469740447146884e-06, "loss": 1.9826, "step": 31242 }, { "epoch": 0.77, "learning_rate": 2.6464330080415636e-06, "loss": 1.8957, "step": 31243 }, { "epoch": 0.77, "learning_rate": 2.6458920182351135e-06, "loss": 1.7915, "step": 31244 }, { "epoch": 0.77, "learning_rate": 2.64535107529879e-06, "loss": 1.8593, "step": 31245 }, { "epoch": 0.77, "learning_rate": 2.6448101792360438e-06, "loss": 1.9879, "step": 31246 }, { "epoch": 0.77, "learning_rate": 2.6442693300503165e-06, "loss": 1.897, "step": 31247 }, { "epoch": 0.77, "learning_rate": 2.643728527745062e-06, "loss": 1.6953, "step": 31248 }, { "epoch": 0.77, "learning_rate": 2.6431877723237187e-06, "loss": 1.9546, "step": 31249 }, { "epoch": 0.77, "learning_rate": 2.642647063789737e-06, "loss": 1.8351, "step": 31250 }, { "epoch": 0.77, "learning_rate": 2.642106402146565e-06, "loss": 1.8974, "step": 31251 }, { "epoch": 0.77, "learning_rate": 2.641565787397645e-06, "loss": 2.0165, "step": 31252 }, { "epoch": 0.77, "learning_rate": 2.641025219546428e-06, "loss": 1.909, "step": 31253 }, { "epoch": 0.77, "learning_rate": 2.6404846985963516e-06, "loss": 1.9566, "step": 31254 }, { "epoch": 0.77, "learning_rate": 2.6399442245508656e-06, "loss": 1.8542, "step": 31255 }, { "epoch": 0.77, "learning_rate": 2.6394037974134167e-06, "loss": 1.8677, "step": 31256 }, { "epoch": 0.77, "learning_rate": 2.638863417187443e-06, "loss": 2.0094, "step": 31257 }, { "epoch": 0.77, "learning_rate": 2.638323083876394e-06, "loss": 2.0946, "step": 31258 }, { "epoch": 0.77, "learning_rate": 2.637782797483709e-06, "loss": 1.8318, "step": 31259 }, { "epoch": 0.77, "learning_rate": 2.6372425580128338e-06, "loss": 1.917, "step": 31260 }, { "epoch": 0.77, "learning_rate": 2.636702365467213e-06, "loss": 1.9972, "step": 31261 }, { "epoch": 0.77, "learning_rate": 2.6361622198502857e-06, "loss": 1.982, "step": 31262 }, { "epoch": 0.77, "learning_rate": 2.6356221211654987e-06, "loss": 1.7829, "step": 31263 }, { "epoch": 0.77, "learning_rate": 2.6350820694162892e-06, "loss": 1.9925, "step": 31264 }, { "epoch": 0.77, "learning_rate": 2.6345420646061017e-06, "loss": 1.8809, "step": 31265 }, { "epoch": 0.77, "learning_rate": 2.63400210673838e-06, "loss": 1.8679, "step": 31266 }, { "epoch": 0.77, "learning_rate": 2.633462195816562e-06, "loss": 2.0522, "step": 31267 }, { "epoch": 0.77, "learning_rate": 2.63292233184409e-06, "loss": 1.8657, "step": 31268 }, { "epoch": 0.77, "learning_rate": 2.6323825148244043e-06, "loss": 1.868, "step": 31269 }, { "epoch": 0.77, "learning_rate": 2.631842744760945e-06, "loss": 1.9349, "step": 31270 }, { "epoch": 0.77, "learning_rate": 2.6313030216571578e-06, "loss": 1.9677, "step": 31271 }, { "epoch": 0.77, "learning_rate": 2.630763345516474e-06, "loss": 1.8589, "step": 31272 }, { "epoch": 0.77, "learning_rate": 2.6302237163423393e-06, "loss": 2.0005, "step": 31273 }, { "epoch": 0.77, "learning_rate": 2.6296841341381886e-06, "loss": 2.1131, "step": 31274 }, { "epoch": 0.77, "learning_rate": 2.6291445989074626e-06, "loss": 1.8863, "step": 31275 }, { "epoch": 0.77, "learning_rate": 2.628605110653604e-06, "loss": 1.8967, "step": 31276 }, { "epoch": 0.77, "learning_rate": 2.6280656693800423e-06, "loss": 1.9428, "step": 31277 }, { "epoch": 0.77, "learning_rate": 2.6275262750902255e-06, "loss": 1.8905, "step": 31278 }, { "epoch": 0.77, "learning_rate": 2.6269869277875815e-06, "loss": 2.0252, "step": 31279 }, { "epoch": 0.77, "learning_rate": 2.6264476274755536e-06, "loss": 1.8305, "step": 31280 }, { "epoch": 0.77, "learning_rate": 2.625908374157581e-06, "loss": 1.7447, "step": 31281 }, { "epoch": 0.77, "learning_rate": 2.625369167837095e-06, "loss": 1.9406, "step": 31282 }, { "epoch": 0.77, "learning_rate": 2.624830008517537e-06, "loss": 1.9102, "step": 31283 }, { "epoch": 0.77, "learning_rate": 2.624290896202337e-06, "loss": 2.0707, "step": 31284 }, { "epoch": 0.77, "learning_rate": 2.623751830894936e-06, "loss": 2.0536, "step": 31285 }, { "epoch": 0.77, "learning_rate": 2.623212812598769e-06, "loss": 1.8652, "step": 31286 }, { "epoch": 0.77, "learning_rate": 2.622673841317269e-06, "loss": 1.8864, "step": 31287 }, { "epoch": 0.77, "learning_rate": 2.622134917053877e-06, "loss": 1.9736, "step": 31288 }, { "epoch": 0.77, "learning_rate": 2.6215960398120197e-06, "loss": 2.0792, "step": 31289 }, { "epoch": 0.77, "learning_rate": 2.621057209595136e-06, "loss": 1.9753, "step": 31290 }, { "epoch": 0.77, "learning_rate": 2.620518426406663e-06, "loss": 2.0498, "step": 31291 }, { "epoch": 0.77, "learning_rate": 2.6199796902500273e-06, "loss": 2.0128, "step": 31292 }, { "epoch": 0.77, "learning_rate": 2.619441001128669e-06, "loss": 1.9995, "step": 31293 }, { "epoch": 0.77, "learning_rate": 2.618902359046016e-06, "loss": 1.9579, "step": 31294 }, { "epoch": 0.77, "learning_rate": 2.618363764005503e-06, "loss": 1.9306, "step": 31295 }, { "epoch": 0.77, "learning_rate": 2.6178252160105665e-06, "loss": 1.871, "step": 31296 }, { "epoch": 0.77, "learning_rate": 2.617286715064633e-06, "loss": 1.7233, "step": 31297 }, { "epoch": 0.77, "learning_rate": 2.616748261171137e-06, "loss": 1.8726, "step": 31298 }, { "epoch": 0.77, "learning_rate": 2.6162098543335133e-06, "loss": 1.9709, "step": 31299 }, { "epoch": 0.77, "learning_rate": 2.6156714945551877e-06, "loss": 1.9283, "step": 31300 }, { "epoch": 0.77, "learning_rate": 2.6151331818395974e-06, "loss": 1.9488, "step": 31301 }, { "epoch": 0.77, "learning_rate": 2.614594916190166e-06, "loss": 1.9364, "step": 31302 }, { "epoch": 0.77, "learning_rate": 2.61405669761033e-06, "loss": 1.8544, "step": 31303 }, { "epoch": 0.77, "learning_rate": 2.6135185261035156e-06, "loss": 1.8673, "step": 31304 }, { "epoch": 0.77, "learning_rate": 2.6129804016731563e-06, "loss": 1.8414, "step": 31305 }, { "epoch": 0.77, "learning_rate": 2.6124423243226827e-06, "loss": 2.024, "step": 31306 }, { "epoch": 0.77, "learning_rate": 2.6119042940555184e-06, "loss": 1.9089, "step": 31307 }, { "epoch": 0.77, "learning_rate": 2.6113663108750952e-06, "loss": 1.853, "step": 31308 }, { "epoch": 0.77, "learning_rate": 2.610828374784846e-06, "loss": 2.0185, "step": 31309 }, { "epoch": 0.77, "learning_rate": 2.6102904857881937e-06, "loss": 1.9707, "step": 31310 }, { "epoch": 0.77, "learning_rate": 2.60975264388857e-06, "loss": 1.9596, "step": 31311 }, { "epoch": 0.77, "learning_rate": 2.6092148490893987e-06, "loss": 2.0237, "step": 31312 }, { "epoch": 0.77, "learning_rate": 2.6086771013941104e-06, "loss": 1.9094, "step": 31313 }, { "epoch": 0.77, "learning_rate": 2.6081394008061334e-06, "loss": 1.8258, "step": 31314 }, { "epoch": 0.77, "learning_rate": 2.6076017473288905e-06, "loss": 1.9621, "step": 31315 }, { "epoch": 0.77, "learning_rate": 2.6070641409658136e-06, "loss": 1.9014, "step": 31316 }, { "epoch": 0.77, "learning_rate": 2.6065265817203233e-06, "loss": 1.955, "step": 31317 }, { "epoch": 0.77, "learning_rate": 2.605989069595849e-06, "loss": 1.9501, "step": 31318 }, { "epoch": 0.77, "learning_rate": 2.6054516045958187e-06, "loss": 1.8922, "step": 31319 }, { "epoch": 0.77, "learning_rate": 2.6049141867236514e-06, "loss": 1.9814, "step": 31320 }, { "epoch": 0.77, "learning_rate": 2.6043768159827766e-06, "loss": 1.9942, "step": 31321 }, { "epoch": 0.77, "learning_rate": 2.6038394923766184e-06, "loss": 1.8691, "step": 31322 }, { "epoch": 0.77, "learning_rate": 2.603302215908602e-06, "loss": 1.7481, "step": 31323 }, { "epoch": 0.77, "learning_rate": 2.6027649865821536e-06, "loss": 1.8614, "step": 31324 }, { "epoch": 0.77, "learning_rate": 2.6022278044006922e-06, "loss": 2.0409, "step": 31325 }, { "epoch": 0.77, "learning_rate": 2.6016906693676457e-06, "loss": 2.0363, "step": 31326 }, { "epoch": 0.77, "learning_rate": 2.6011535814864343e-06, "loss": 2.0461, "step": 31327 }, { "epoch": 0.77, "learning_rate": 2.600616540760481e-06, "loss": 1.8642, "step": 31328 }, { "epoch": 0.77, "learning_rate": 2.600079547193213e-06, "loss": 1.9462, "step": 31329 }, { "epoch": 0.77, "learning_rate": 2.5995426007880466e-06, "loss": 1.7918, "step": 31330 }, { "epoch": 0.77, "learning_rate": 2.5990057015484105e-06, "loss": 2.0112, "step": 31331 }, { "epoch": 0.77, "learning_rate": 2.59846884947772e-06, "loss": 2.0196, "step": 31332 }, { "epoch": 0.77, "learning_rate": 2.597932044579399e-06, "loss": 2.0508, "step": 31333 }, { "epoch": 0.77, "learning_rate": 2.597395286856872e-06, "loss": 1.7783, "step": 31334 }, { "epoch": 0.77, "learning_rate": 2.5968585763135557e-06, "loss": 1.8331, "step": 31335 }, { "epoch": 0.77, "learning_rate": 2.596321912952874e-06, "loss": 2.0503, "step": 31336 }, { "epoch": 0.77, "learning_rate": 2.5957852967782427e-06, "loss": 1.8856, "step": 31337 }, { "epoch": 0.77, "learning_rate": 2.595248727793085e-06, "loss": 1.9253, "step": 31338 }, { "epoch": 0.77, "learning_rate": 2.594712206000821e-06, "loss": 1.9088, "step": 31339 }, { "epoch": 0.77, "learning_rate": 2.594175731404869e-06, "loss": 1.8984, "step": 31340 }, { "epoch": 0.77, "learning_rate": 2.5936393040086506e-06, "loss": 2.0078, "step": 31341 }, { "epoch": 0.77, "learning_rate": 2.5931029238155804e-06, "loss": 1.7912, "step": 31342 }, { "epoch": 0.77, "learning_rate": 2.5925665908290785e-06, "loss": 1.9449, "step": 31343 }, { "epoch": 0.77, "learning_rate": 2.5920303050525673e-06, "loss": 1.9409, "step": 31344 }, { "epoch": 0.77, "learning_rate": 2.5914940664894584e-06, "loss": 2.0219, "step": 31345 }, { "epoch": 0.77, "learning_rate": 2.5909578751431753e-06, "loss": 1.8234, "step": 31346 }, { "epoch": 0.77, "learning_rate": 2.5904217310171287e-06, "loss": 2.0049, "step": 31347 }, { "epoch": 0.77, "learning_rate": 2.5898856341147384e-06, "loss": 1.8951, "step": 31348 }, { "epoch": 0.77, "learning_rate": 2.5893495844394257e-06, "loss": 1.8703, "step": 31349 }, { "epoch": 0.77, "learning_rate": 2.5888135819946e-06, "loss": 2.1889, "step": 31350 }, { "epoch": 0.77, "learning_rate": 2.588277626783684e-06, "loss": 1.816, "step": 31351 }, { "epoch": 0.77, "learning_rate": 2.5877417188100864e-06, "loss": 1.8019, "step": 31352 }, { "epoch": 0.77, "learning_rate": 2.5872058580772264e-06, "loss": 2.0725, "step": 31353 }, { "epoch": 0.77, "learning_rate": 2.586670044588523e-06, "loss": 1.9084, "step": 31354 }, { "epoch": 0.77, "learning_rate": 2.586134278347384e-06, "loss": 1.8292, "step": 31355 }, { "epoch": 0.77, "learning_rate": 2.5855985593572275e-06, "loss": 1.9517, "step": 31356 }, { "epoch": 0.77, "learning_rate": 2.5850628876214666e-06, "loss": 1.9708, "step": 31357 }, { "epoch": 0.77, "learning_rate": 2.584527263143517e-06, "loss": 1.8765, "step": 31358 }, { "epoch": 0.77, "learning_rate": 2.5839916859267955e-06, "loss": 2.0038, "step": 31359 }, { "epoch": 0.77, "learning_rate": 2.5834561559747086e-06, "loss": 2.1958, "step": 31360 }, { "epoch": 0.77, "learning_rate": 2.582920673290672e-06, "loss": 1.8968, "step": 31361 }, { "epoch": 0.77, "learning_rate": 2.5823852378781023e-06, "loss": 1.9356, "step": 31362 }, { "epoch": 0.77, "learning_rate": 2.5818498497404054e-06, "loss": 2.0287, "step": 31363 }, { "epoch": 0.77, "learning_rate": 2.581314508881001e-06, "loss": 1.893, "step": 31364 }, { "epoch": 0.77, "learning_rate": 2.580779215303293e-06, "loss": 1.8766, "step": 31365 }, { "epoch": 0.77, "learning_rate": 2.580243969010697e-06, "loss": 1.9863, "step": 31366 }, { "epoch": 0.77, "learning_rate": 2.579708770006628e-06, "loss": 1.8192, "step": 31367 }, { "epoch": 0.77, "learning_rate": 2.5791736182944892e-06, "loss": 2.0398, "step": 31368 }, { "epoch": 0.77, "learning_rate": 2.5786385138776994e-06, "loss": 2.0219, "step": 31369 }, { "epoch": 0.77, "learning_rate": 2.578103456759661e-06, "loss": 1.9232, "step": 31370 }, { "epoch": 0.77, "learning_rate": 2.577568446943788e-06, "loss": 1.9698, "step": 31371 }, { "epoch": 0.77, "learning_rate": 2.5770334844334932e-06, "loss": 1.8869, "step": 31372 }, { "epoch": 0.77, "learning_rate": 2.576498569232181e-06, "loss": 1.9331, "step": 31373 }, { "epoch": 0.77, "learning_rate": 2.5759637013432613e-06, "loss": 2.0935, "step": 31374 }, { "epoch": 0.77, "learning_rate": 2.5754288807701453e-06, "loss": 1.9921, "step": 31375 }, { "epoch": 0.77, "learning_rate": 2.5748941075162403e-06, "loss": 1.8856, "step": 31376 }, { "epoch": 0.77, "learning_rate": 2.5743593815849578e-06, "loss": 1.9872, "step": 31377 }, { "epoch": 0.77, "learning_rate": 2.5738247029797e-06, "loss": 1.9656, "step": 31378 }, { "epoch": 0.77, "learning_rate": 2.573290071703881e-06, "loss": 1.8676, "step": 31379 }, { "epoch": 0.77, "learning_rate": 2.5727554877609017e-06, "loss": 1.8177, "step": 31380 }, { "epoch": 0.77, "learning_rate": 2.572220951154172e-06, "loss": 1.8982, "step": 31381 }, { "epoch": 0.77, "learning_rate": 2.5716864618871008e-06, "loss": 2.0371, "step": 31382 }, { "epoch": 0.77, "learning_rate": 2.57115201996309e-06, "loss": 1.9429, "step": 31383 }, { "epoch": 0.77, "learning_rate": 2.570617625385552e-06, "loss": 1.9935, "step": 31384 }, { "epoch": 0.77, "learning_rate": 2.5700832781578855e-06, "loss": 1.9197, "step": 31385 }, { "epoch": 0.77, "learning_rate": 2.5695489782835004e-06, "loss": 2.0305, "step": 31386 }, { "epoch": 0.77, "learning_rate": 2.5690147257658036e-06, "loss": 1.9692, "step": 31387 }, { "epoch": 0.77, "learning_rate": 2.5684805206081943e-06, "loss": 2.0034, "step": 31388 }, { "epoch": 0.77, "learning_rate": 2.5679463628140845e-06, "loss": 1.7937, "step": 31389 }, { "epoch": 0.77, "learning_rate": 2.5674122523868706e-06, "loss": 2.2624, "step": 31390 }, { "epoch": 0.77, "learning_rate": 2.566878189329961e-06, "loss": 1.8596, "step": 31391 }, { "epoch": 0.77, "learning_rate": 2.5663441736467587e-06, "loss": 2.0617, "step": 31392 }, { "epoch": 0.77, "learning_rate": 2.5658102053406687e-06, "loss": 2.0512, "step": 31393 }, { "epoch": 0.77, "learning_rate": 2.5652762844150946e-06, "loss": 1.7847, "step": 31394 }, { "epoch": 0.77, "learning_rate": 2.5647424108734355e-06, "loss": 2.0048, "step": 31395 }, { "epoch": 0.77, "learning_rate": 2.5642085847190956e-06, "loss": 1.8777, "step": 31396 }, { "epoch": 0.77, "learning_rate": 2.5636748059554804e-06, "loss": 1.9576, "step": 31397 }, { "epoch": 0.77, "learning_rate": 2.5631410745859865e-06, "loss": 2.0033, "step": 31398 }, { "epoch": 0.77, "learning_rate": 2.5626073906140215e-06, "loss": 1.9109, "step": 31399 }, { "epoch": 0.77, "learning_rate": 2.5620737540429796e-06, "loss": 1.9531, "step": 31400 }, { "epoch": 0.77, "learning_rate": 2.561540164876266e-06, "loss": 1.9627, "step": 31401 }, { "epoch": 0.77, "learning_rate": 2.561006623117285e-06, "loss": 1.9326, "step": 31402 }, { "epoch": 0.77, "learning_rate": 2.5604731287694297e-06, "loss": 1.9054, "step": 31403 }, { "epoch": 0.77, "learning_rate": 2.5599396818361055e-06, "loss": 1.9661, "step": 31404 }, { "epoch": 0.77, "learning_rate": 2.5594062823207076e-06, "loss": 2.0134, "step": 31405 }, { "epoch": 0.77, "learning_rate": 2.558872930226639e-06, "loss": 1.9805, "step": 31406 }, { "epoch": 0.77, "learning_rate": 2.5583396255573013e-06, "loss": 2.0097, "step": 31407 }, { "epoch": 0.77, "learning_rate": 2.557806368316087e-06, "loss": 1.9403, "step": 31408 }, { "epoch": 0.77, "learning_rate": 2.557273158506397e-06, "loss": 1.9462, "step": 31409 }, { "epoch": 0.77, "learning_rate": 2.5567399961316317e-06, "loss": 1.8698, "step": 31410 }, { "epoch": 0.77, "learning_rate": 2.556206881195189e-06, "loss": 1.9788, "step": 31411 }, { "epoch": 0.77, "learning_rate": 2.5556738137004676e-06, "loss": 1.7636, "step": 31412 }, { "epoch": 0.77, "learning_rate": 2.55514079365086e-06, "loss": 1.8251, "step": 31413 }, { "epoch": 0.77, "learning_rate": 2.554607821049766e-06, "loss": 1.8703, "step": 31414 }, { "epoch": 0.77, "learning_rate": 2.554074895900587e-06, "loss": 2.214, "step": 31415 }, { "epoch": 0.77, "learning_rate": 2.5535420182067116e-06, "loss": 2.0194, "step": 31416 }, { "epoch": 0.77, "learning_rate": 2.5530091879715425e-06, "loss": 1.8993, "step": 31417 }, { "epoch": 0.77, "learning_rate": 2.5524764051984695e-06, "loss": 1.9984, "step": 31418 }, { "epoch": 0.77, "learning_rate": 2.551943669890892e-06, "loss": 1.9496, "step": 31419 }, { "epoch": 0.77, "learning_rate": 2.551410982052207e-06, "loss": 1.9755, "step": 31420 }, { "epoch": 0.77, "learning_rate": 2.550878341685805e-06, "loss": 1.9443, "step": 31421 }, { "epoch": 0.77, "learning_rate": 2.5503457487950856e-06, "loss": 1.976, "step": 31422 }, { "epoch": 0.77, "learning_rate": 2.5498132033834377e-06, "loss": 1.9285, "step": 31423 }, { "epoch": 0.77, "learning_rate": 2.549280705454258e-06, "loss": 1.8118, "step": 31424 }, { "epoch": 0.77, "learning_rate": 2.5487482550109433e-06, "loss": 1.8783, "step": 31425 }, { "epoch": 0.77, "learning_rate": 2.548215852056881e-06, "loss": 1.9072, "step": 31426 }, { "epoch": 0.77, "learning_rate": 2.547683496595469e-06, "loss": 1.8924, "step": 31427 }, { "epoch": 0.77, "learning_rate": 2.547151188630097e-06, "loss": 2.1621, "step": 31428 }, { "epoch": 0.77, "learning_rate": 2.546618928164161e-06, "loss": 1.9676, "step": 31429 }, { "epoch": 0.77, "learning_rate": 2.546086715201053e-06, "loss": 1.9125, "step": 31430 }, { "epoch": 0.77, "learning_rate": 2.5455545497441612e-06, "loss": 1.8245, "step": 31431 }, { "epoch": 0.77, "learning_rate": 2.545022431796883e-06, "loss": 1.9306, "step": 31432 }, { "epoch": 0.77, "learning_rate": 2.5444903613626026e-06, "loss": 1.9117, "step": 31433 }, { "epoch": 0.77, "learning_rate": 2.543958338444715e-06, "loss": 1.8678, "step": 31434 }, { "epoch": 0.77, "learning_rate": 2.543426363046614e-06, "loss": 1.8975, "step": 31435 }, { "epoch": 0.77, "learning_rate": 2.5428944351716835e-06, "loss": 1.8723, "step": 31436 }, { "epoch": 0.77, "learning_rate": 2.5423625548233197e-06, "loss": 1.9538, "step": 31437 }, { "epoch": 0.77, "learning_rate": 2.5418307220049075e-06, "loss": 1.8472, "step": 31438 }, { "epoch": 0.77, "learning_rate": 2.5412989367198383e-06, "loss": 1.8021, "step": 31439 }, { "epoch": 0.77, "learning_rate": 2.5407671989715046e-06, "loss": 1.833, "step": 31440 }, { "epoch": 0.77, "learning_rate": 2.54023550876329e-06, "loss": 1.9297, "step": 31441 }, { "epoch": 0.78, "learning_rate": 2.5397038660985873e-06, "loss": 1.8488, "step": 31442 }, { "epoch": 0.78, "learning_rate": 2.539172270980782e-06, "loss": 1.7873, "step": 31443 }, { "epoch": 0.78, "learning_rate": 2.5386407234132616e-06, "loss": 1.9338, "step": 31444 }, { "epoch": 0.78, "learning_rate": 2.5381092233994163e-06, "loss": 1.9596, "step": 31445 }, { "epoch": 0.78, "learning_rate": 2.5375777709426317e-06, "loss": 1.8351, "step": 31446 }, { "epoch": 0.78, "learning_rate": 2.5370463660462995e-06, "loss": 2.0158, "step": 31447 }, { "epoch": 0.78, "learning_rate": 2.5365150087138e-06, "loss": 1.9517, "step": 31448 }, { "epoch": 0.78, "learning_rate": 2.535983698948523e-06, "loss": 1.8481, "step": 31449 }, { "epoch": 0.78, "learning_rate": 2.5354524367538567e-06, "loss": 2.0002, "step": 31450 }, { "epoch": 0.78, "learning_rate": 2.534921222133182e-06, "loss": 2.0262, "step": 31451 }, { "epoch": 0.78, "learning_rate": 2.5343900550898904e-06, "loss": 1.9989, "step": 31452 }, { "epoch": 0.78, "learning_rate": 2.5338589356273603e-06, "loss": 1.9056, "step": 31453 }, { "epoch": 0.78, "learning_rate": 2.533327863748981e-06, "loss": 1.9659, "step": 31454 }, { "epoch": 0.78, "learning_rate": 2.532796839458139e-06, "loss": 2.1034, "step": 31455 }, { "epoch": 0.78, "learning_rate": 2.532265862758213e-06, "loss": 2.09, "step": 31456 }, { "epoch": 0.78, "learning_rate": 2.5317349336525942e-06, "loss": 1.8014, "step": 31457 }, { "epoch": 0.78, "learning_rate": 2.531204052144659e-06, "loss": 2.021, "step": 31458 }, { "epoch": 0.78, "learning_rate": 2.5306732182377946e-06, "loss": 1.7896, "step": 31459 }, { "epoch": 0.78, "learning_rate": 2.530142431935386e-06, "loss": 1.8238, "step": 31460 }, { "epoch": 0.78, "learning_rate": 2.529611693240812e-06, "loss": 1.9228, "step": 31461 }, { "epoch": 0.78, "learning_rate": 2.5290810021574575e-06, "loss": 1.9926, "step": 31462 }, { "epoch": 0.78, "learning_rate": 2.528550358688704e-06, "loss": 1.9889, "step": 31463 }, { "epoch": 0.78, "learning_rate": 2.528019762837934e-06, "loss": 1.9084, "step": 31464 }, { "epoch": 0.78, "learning_rate": 2.5274892146085318e-06, "loss": 1.9328, "step": 31465 }, { "epoch": 0.78, "learning_rate": 2.526958714003873e-06, "loss": 1.8443, "step": 31466 }, { "epoch": 0.78, "learning_rate": 2.526428261027346e-06, "loss": 1.8949, "step": 31467 }, { "epoch": 0.78, "learning_rate": 2.5258978556823233e-06, "loss": 2.2337, "step": 31468 }, { "epoch": 0.78, "learning_rate": 2.5253674979721886e-06, "loss": 1.914, "step": 31469 }, { "epoch": 0.78, "learning_rate": 2.5248371879003264e-06, "loss": 1.8854, "step": 31470 }, { "epoch": 0.78, "learning_rate": 2.5243069254701102e-06, "loss": 1.7688, "step": 31471 }, { "epoch": 0.78, "learning_rate": 2.5237767106849255e-06, "loss": 1.8352, "step": 31472 }, { "epoch": 0.78, "learning_rate": 2.5232465435481445e-06, "loss": 2.087, "step": 31473 }, { "epoch": 0.78, "learning_rate": 2.5227164240631496e-06, "loss": 1.9473, "step": 31474 }, { "epoch": 0.78, "learning_rate": 2.5221863522333235e-06, "loss": 1.9466, "step": 31475 }, { "epoch": 0.78, "learning_rate": 2.5216563280620374e-06, "loss": 1.9367, "step": 31476 }, { "epoch": 0.78, "learning_rate": 2.5211263515526728e-06, "loss": 1.8623, "step": 31477 }, { "epoch": 0.78, "learning_rate": 2.52059642270861e-06, "loss": 1.8866, "step": 31478 }, { "epoch": 0.78, "learning_rate": 2.5200665415332202e-06, "loss": 1.9558, "step": 31479 }, { "epoch": 0.78, "learning_rate": 2.519536708029885e-06, "loss": 1.8032, "step": 31480 }, { "epoch": 0.78, "learning_rate": 2.51900692220198e-06, "loss": 1.808, "step": 31481 }, { "epoch": 0.78, "learning_rate": 2.5184771840528823e-06, "loss": 1.9106, "step": 31482 }, { "epoch": 0.78, "learning_rate": 2.5179474935859692e-06, "loss": 2.0453, "step": 31483 }, { "epoch": 0.78, "learning_rate": 2.517417850804613e-06, "loss": 1.9742, "step": 31484 }, { "epoch": 0.78, "learning_rate": 2.5168882557121945e-06, "loss": 1.8941, "step": 31485 }, { "epoch": 0.78, "learning_rate": 2.516358708312082e-06, "loss": 1.9101, "step": 31486 }, { "epoch": 0.78, "learning_rate": 2.5158292086076553e-06, "loss": 1.9635, "step": 31487 }, { "epoch": 0.78, "learning_rate": 2.51529975660229e-06, "loss": 1.9163, "step": 31488 }, { "epoch": 0.78, "learning_rate": 2.514770352299356e-06, "loss": 1.927, "step": 31489 }, { "epoch": 0.78, "learning_rate": 2.514240995702233e-06, "loss": 1.874, "step": 31490 }, { "epoch": 0.78, "learning_rate": 2.5137116868142888e-06, "loss": 2.0079, "step": 31491 }, { "epoch": 0.78, "learning_rate": 2.513182425638899e-06, "loss": 1.8753, "step": 31492 }, { "epoch": 0.78, "learning_rate": 2.512653212179441e-06, "loss": 1.9121, "step": 31493 }, { "epoch": 0.78, "learning_rate": 2.5121240464392817e-06, "loss": 2.0202, "step": 31494 }, { "epoch": 0.78, "learning_rate": 2.5115949284217953e-06, "loss": 1.7451, "step": 31495 }, { "epoch": 0.78, "learning_rate": 2.5110658581303584e-06, "loss": 1.8044, "step": 31496 }, { "epoch": 0.78, "learning_rate": 2.510536835568337e-06, "loss": 1.6544, "step": 31497 }, { "epoch": 0.78, "learning_rate": 2.510007860739103e-06, "loss": 1.8761, "step": 31498 }, { "epoch": 0.78, "learning_rate": 2.509478933646031e-06, "loss": 1.8873, "step": 31499 }, { "epoch": 0.78, "learning_rate": 2.5089500542924948e-06, "loss": 1.9864, "step": 31500 }, { "epoch": 0.78, "learning_rate": 2.508421222681857e-06, "loss": 2.1413, "step": 31501 }, { "epoch": 0.78, "learning_rate": 2.507892438817492e-06, "loss": 1.9883, "step": 31502 }, { "epoch": 0.78, "learning_rate": 2.507363702702774e-06, "loss": 1.9601, "step": 31503 }, { "epoch": 0.78, "learning_rate": 2.5068350143410646e-06, "loss": 1.9084, "step": 31504 }, { "epoch": 0.78, "learning_rate": 2.506306373735742e-06, "loss": 1.7406, "step": 31505 }, { "epoch": 0.78, "learning_rate": 2.5057777808901673e-06, "loss": 1.7968, "step": 31506 }, { "epoch": 0.78, "learning_rate": 2.505249235807713e-06, "loss": 1.8631, "step": 31507 }, { "epoch": 0.78, "learning_rate": 2.5047207384917517e-06, "loss": 1.9275, "step": 31508 }, { "epoch": 0.78, "learning_rate": 2.5041922889456437e-06, "loss": 2.1171, "step": 31509 }, { "epoch": 0.78, "learning_rate": 2.503663887172765e-06, "loss": 2.1156, "step": 31510 }, { "epoch": 0.78, "learning_rate": 2.503135533176475e-06, "loss": 1.9906, "step": 31511 }, { "epoch": 0.78, "learning_rate": 2.5026072269601463e-06, "loss": 1.8439, "step": 31512 }, { "epoch": 0.78, "learning_rate": 2.502078968527145e-06, "loss": 2.044, "step": 31513 }, { "epoch": 0.78, "learning_rate": 2.5015507578808405e-06, "loss": 1.9604, "step": 31514 }, { "epoch": 0.78, "learning_rate": 2.5010225950245946e-06, "loss": 1.8459, "step": 31515 }, { "epoch": 0.78, "learning_rate": 2.5004944799617746e-06, "loss": 1.8597, "step": 31516 }, { "epoch": 0.78, "learning_rate": 2.499966412695748e-06, "loss": 1.9175, "step": 31517 }, { "epoch": 0.78, "learning_rate": 2.4994383932298826e-06, "loss": 1.971, "step": 31518 }, { "epoch": 0.78, "learning_rate": 2.4989104215675374e-06, "loss": 1.897, "step": 31519 }, { "epoch": 0.78, "learning_rate": 2.498382497712083e-06, "loss": 1.7606, "step": 31520 }, { "epoch": 0.78, "learning_rate": 2.49785462166688e-06, "loss": 1.7981, "step": 31521 }, { "epoch": 0.78, "learning_rate": 2.497326793435294e-06, "loss": 1.9041, "step": 31522 }, { "epoch": 0.78, "learning_rate": 2.4967990130206923e-06, "loss": 1.9949, "step": 31523 }, { "epoch": 0.78, "learning_rate": 2.496271280426432e-06, "loss": 2.0111, "step": 31524 }, { "epoch": 0.78, "learning_rate": 2.495743595655884e-06, "loss": 1.9073, "step": 31525 }, { "epoch": 0.78, "learning_rate": 2.4952159587124038e-06, "loss": 2.0074, "step": 31526 }, { "epoch": 0.78, "learning_rate": 2.4946883695993585e-06, "loss": 1.8635, "step": 31527 }, { "epoch": 0.78, "learning_rate": 2.4941608283201136e-06, "loss": 1.8639, "step": 31528 }, { "epoch": 0.78, "learning_rate": 2.493633334878024e-06, "loss": 2.1196, "step": 31529 }, { "epoch": 0.78, "learning_rate": 2.493105889276456e-06, "loss": 2.0298, "step": 31530 }, { "epoch": 0.78, "learning_rate": 2.49257849151877e-06, "loss": 1.9792, "step": 31531 }, { "epoch": 0.78, "learning_rate": 2.4920511416083314e-06, "loss": 1.9513, "step": 31532 }, { "epoch": 0.78, "learning_rate": 2.4915238395484953e-06, "loss": 2.0121, "step": 31533 }, { "epoch": 0.78, "learning_rate": 2.4909965853426233e-06, "loss": 1.9779, "step": 31534 }, { "epoch": 0.78, "learning_rate": 2.490469378994078e-06, "loss": 1.9117, "step": 31535 }, { "epoch": 0.78, "learning_rate": 2.489942220506222e-06, "loss": 1.9282, "step": 31536 }, { "epoch": 0.78, "learning_rate": 2.4894151098824083e-06, "loss": 1.8486, "step": 31537 }, { "epoch": 0.78, "learning_rate": 2.4888880471260025e-06, "loss": 1.9768, "step": 31538 }, { "epoch": 0.78, "learning_rate": 2.4883610322403573e-06, "loss": 1.9324, "step": 31539 }, { "epoch": 0.78, "learning_rate": 2.487834065228837e-06, "loss": 1.986, "step": 31540 }, { "epoch": 0.78, "learning_rate": 2.4873071460947997e-06, "loss": 1.8849, "step": 31541 }, { "epoch": 0.78, "learning_rate": 2.4867802748416004e-06, "loss": 1.9251, "step": 31542 }, { "epoch": 0.78, "learning_rate": 2.4862534514726012e-06, "loss": 1.8827, "step": 31543 }, { "epoch": 0.78, "learning_rate": 2.485726675991155e-06, "loss": 1.773, "step": 31544 }, { "epoch": 0.78, "learning_rate": 2.4851999484006205e-06, "loss": 1.9396, "step": 31545 }, { "epoch": 0.78, "learning_rate": 2.4846732687043597e-06, "loss": 2.0201, "step": 31546 }, { "epoch": 0.78, "learning_rate": 2.484146636905722e-06, "loss": 1.9455, "step": 31547 }, { "epoch": 0.78, "learning_rate": 2.4836200530080668e-06, "loss": 1.7438, "step": 31548 }, { "epoch": 0.78, "learning_rate": 2.483093517014754e-06, "loss": 1.9442, "step": 31549 }, { "epoch": 0.78, "learning_rate": 2.4825670289291325e-06, "loss": 1.8823, "step": 31550 }, { "epoch": 0.78, "learning_rate": 2.482040588754562e-06, "loss": 2.1273, "step": 31551 }, { "epoch": 0.78, "learning_rate": 2.4815141964943955e-06, "loss": 2.0409, "step": 31552 }, { "epoch": 0.78, "learning_rate": 2.4809878521519937e-06, "loss": 1.9648, "step": 31553 }, { "epoch": 0.78, "learning_rate": 2.4804615557307033e-06, "loss": 1.9388, "step": 31554 }, { "epoch": 0.78, "learning_rate": 2.4799353072338815e-06, "loss": 2.0372, "step": 31555 }, { "epoch": 0.78, "learning_rate": 2.4794091066648863e-06, "loss": 2.0669, "step": 31556 }, { "epoch": 0.78, "learning_rate": 2.4788829540270643e-06, "loss": 2.0583, "step": 31557 }, { "epoch": 0.78, "learning_rate": 2.4783568493237753e-06, "loss": 1.9797, "step": 31558 }, { "epoch": 0.78, "learning_rate": 2.4778307925583655e-06, "loss": 1.9751, "step": 31559 }, { "epoch": 0.78, "learning_rate": 2.4773047837341926e-06, "loss": 1.7735, "step": 31560 }, { "epoch": 0.78, "learning_rate": 2.4767788228546107e-06, "loss": 1.8123, "step": 31561 }, { "epoch": 0.78, "learning_rate": 2.476252909922966e-06, "loss": 2.064, "step": 31562 }, { "epoch": 0.78, "learning_rate": 2.475727044942616e-06, "loss": 1.8426, "step": 31563 }, { "epoch": 0.78, "learning_rate": 2.475201227916908e-06, "loss": 2.0377, "step": 31564 }, { "epoch": 0.78, "learning_rate": 2.474675458849194e-06, "loss": 2.0121, "step": 31565 }, { "epoch": 0.78, "learning_rate": 2.4741497377428247e-06, "loss": 1.9761, "step": 31566 }, { "epoch": 0.78, "learning_rate": 2.4736240646011566e-06, "loss": 1.9234, "step": 31567 }, { "epoch": 0.78, "learning_rate": 2.47309843942753e-06, "loss": 2.0482, "step": 31568 }, { "epoch": 0.78, "learning_rate": 2.472572862225302e-06, "loss": 2.0478, "step": 31569 }, { "epoch": 0.78, "learning_rate": 2.47204733299782e-06, "loss": 1.9174, "step": 31570 }, { "epoch": 0.78, "learning_rate": 2.471521851748435e-06, "loss": 2.0808, "step": 31571 }, { "epoch": 0.78, "learning_rate": 2.4709964184804935e-06, "loss": 1.9514, "step": 31572 }, { "epoch": 0.78, "learning_rate": 2.4704710331973483e-06, "loss": 1.9606, "step": 31573 }, { "epoch": 0.78, "learning_rate": 2.4699456959023405e-06, "loss": 1.8857, "step": 31574 }, { "epoch": 0.78, "learning_rate": 2.4694204065988236e-06, "loss": 1.8652, "step": 31575 }, { "epoch": 0.78, "learning_rate": 2.468895165290148e-06, "loss": 2.0781, "step": 31576 }, { "epoch": 0.78, "learning_rate": 2.468369971979655e-06, "loss": 1.917, "step": 31577 }, { "epoch": 0.78, "learning_rate": 2.467844826670697e-06, "loss": 1.8715, "step": 31578 }, { "epoch": 0.78, "learning_rate": 2.4673197293666153e-06, "loss": 1.9295, "step": 31579 }, { "epoch": 0.78, "learning_rate": 2.466794680070761e-06, "loss": 1.8964, "step": 31580 }, { "epoch": 0.78, "learning_rate": 2.4662696787864814e-06, "loss": 1.9928, "step": 31581 }, { "epoch": 0.78, "learning_rate": 2.4657447255171173e-06, "loss": 1.8905, "step": 31582 }, { "epoch": 0.78, "learning_rate": 2.465219820266017e-06, "loss": 1.9678, "step": 31583 }, { "epoch": 0.78, "learning_rate": 2.464694963036527e-06, "loss": 1.7661, "step": 31584 }, { "epoch": 0.78, "learning_rate": 2.464170153831994e-06, "loss": 1.9853, "step": 31585 }, { "epoch": 0.78, "learning_rate": 2.463645392655758e-06, "loss": 1.9226, "step": 31586 }, { "epoch": 0.78, "learning_rate": 2.463120679511166e-06, "loss": 1.9728, "step": 31587 }, { "epoch": 0.78, "learning_rate": 2.462596014401565e-06, "loss": 1.9498, "step": 31588 }, { "epoch": 0.78, "learning_rate": 2.4620713973302933e-06, "loss": 1.9596, "step": 31589 }, { "epoch": 0.78, "learning_rate": 2.4615468283006972e-06, "loss": 1.9133, "step": 31590 }, { "epoch": 0.78, "learning_rate": 2.461022307316122e-06, "loss": 1.9874, "step": 31591 }, { "epoch": 0.78, "learning_rate": 2.4604978343799057e-06, "loss": 1.8975, "step": 31592 }, { "epoch": 0.78, "learning_rate": 2.459973409495394e-06, "loss": 2.0068, "step": 31593 }, { "epoch": 0.78, "learning_rate": 2.4594490326659327e-06, "loss": 1.8918, "step": 31594 }, { "epoch": 0.78, "learning_rate": 2.4589247038948562e-06, "loss": 2.1092, "step": 31595 }, { "epoch": 0.78, "learning_rate": 2.458400423185513e-06, "loss": 2.0074, "step": 31596 }, { "epoch": 0.78, "learning_rate": 2.4578761905412397e-06, "loss": 1.9788, "step": 31597 }, { "epoch": 0.78, "learning_rate": 2.4573520059653777e-06, "loss": 1.8539, "step": 31598 }, { "epoch": 0.78, "learning_rate": 2.4568278694612735e-06, "loss": 1.9215, "step": 31599 }, { "epoch": 0.78, "learning_rate": 2.456303781032261e-06, "loss": 1.9166, "step": 31600 }, { "epoch": 0.78, "learning_rate": 2.455779740681682e-06, "loss": 2.0478, "step": 31601 }, { "epoch": 0.78, "learning_rate": 2.455255748412878e-06, "loss": 1.8784, "step": 31602 }, { "epoch": 0.78, "learning_rate": 2.45473180422919e-06, "loss": 1.7643, "step": 31603 }, { "epoch": 0.78, "learning_rate": 2.4542079081339532e-06, "loss": 1.8904, "step": 31604 }, { "epoch": 0.78, "learning_rate": 2.453684060130508e-06, "loss": 1.8826, "step": 31605 }, { "epoch": 0.78, "learning_rate": 2.453160260222196e-06, "loss": 1.8339, "step": 31606 }, { "epoch": 0.78, "learning_rate": 2.4526365084123504e-06, "loss": 2.0039, "step": 31607 }, { "epoch": 0.78, "learning_rate": 2.452112804704313e-06, "loss": 1.8935, "step": 31608 }, { "epoch": 0.78, "learning_rate": 2.451589149101422e-06, "loss": 2.0422, "step": 31609 }, { "epoch": 0.78, "learning_rate": 2.4510655416070105e-06, "loss": 1.9189, "step": 31610 }, { "epoch": 0.78, "learning_rate": 2.4505419822244223e-06, "loss": 2.0058, "step": 31611 }, { "epoch": 0.78, "learning_rate": 2.4500184709569873e-06, "loss": 1.9153, "step": 31612 }, { "epoch": 0.78, "learning_rate": 2.4494950078080447e-06, "loss": 1.9396, "step": 31613 }, { "epoch": 0.78, "learning_rate": 2.4489715927809344e-06, "loss": 1.9233, "step": 31614 }, { "epoch": 0.78, "learning_rate": 2.4484482258789856e-06, "loss": 1.9698, "step": 31615 }, { "epoch": 0.78, "learning_rate": 2.44792490710554e-06, "loss": 2.0539, "step": 31616 }, { "epoch": 0.78, "learning_rate": 2.4474016364639276e-06, "loss": 1.8123, "step": 31617 }, { "epoch": 0.78, "learning_rate": 2.4468784139574854e-06, "loss": 1.9159, "step": 31618 }, { "epoch": 0.78, "learning_rate": 2.4463552395895484e-06, "loss": 1.889, "step": 31619 }, { "epoch": 0.78, "learning_rate": 2.445832113363452e-06, "loss": 1.9246, "step": 31620 }, { "epoch": 0.78, "learning_rate": 2.4453090352825316e-06, "loss": 1.949, "step": 31621 }, { "epoch": 0.78, "learning_rate": 2.444786005350116e-06, "loss": 1.8254, "step": 31622 }, { "epoch": 0.78, "learning_rate": 2.444263023569541e-06, "loss": 1.8026, "step": 31623 }, { "epoch": 0.78, "learning_rate": 2.443740089944143e-06, "loss": 2.0401, "step": 31624 }, { "epoch": 0.78, "learning_rate": 2.4432172044772504e-06, "loss": 1.9088, "step": 31625 }, { "epoch": 0.78, "learning_rate": 2.442694367172199e-06, "loss": 1.8347, "step": 31626 }, { "epoch": 0.78, "learning_rate": 2.442171578032316e-06, "loss": 2.1616, "step": 31627 }, { "epoch": 0.78, "learning_rate": 2.441648837060937e-06, "loss": 2.1695, "step": 31628 }, { "epoch": 0.78, "learning_rate": 2.4411261442613967e-06, "loss": 1.8407, "step": 31629 }, { "epoch": 0.78, "learning_rate": 2.4406034996370197e-06, "loss": 1.9563, "step": 31630 }, { "epoch": 0.78, "learning_rate": 2.440080903191142e-06, "loss": 2.0198, "step": 31631 }, { "epoch": 0.78, "learning_rate": 2.4395583549270905e-06, "loss": 1.9425, "step": 31632 }, { "epoch": 0.78, "learning_rate": 2.4390358548481973e-06, "loss": 1.9427, "step": 31633 }, { "epoch": 0.78, "learning_rate": 2.438513402957795e-06, "loss": 1.7443, "step": 31634 }, { "epoch": 0.78, "learning_rate": 2.437990999259209e-06, "loss": 1.7492, "step": 31635 }, { "epoch": 0.78, "learning_rate": 2.4374686437557705e-06, "loss": 1.948, "step": 31636 }, { "epoch": 0.78, "learning_rate": 2.436946336450808e-06, "loss": 1.9125, "step": 31637 }, { "epoch": 0.78, "learning_rate": 2.4364240773476554e-06, "loss": 1.8843, "step": 31638 }, { "epoch": 0.78, "learning_rate": 2.4359018664496335e-06, "loss": 1.9787, "step": 31639 }, { "epoch": 0.78, "learning_rate": 2.4353797037600746e-06, "loss": 1.8595, "step": 31640 }, { "epoch": 0.78, "learning_rate": 2.434857589282308e-06, "loss": 1.9714, "step": 31641 }, { "epoch": 0.78, "learning_rate": 2.4343355230196574e-06, "loss": 2.1421, "step": 31642 }, { "epoch": 0.78, "learning_rate": 2.4338135049754518e-06, "loss": 2.0237, "step": 31643 }, { "epoch": 0.78, "learning_rate": 2.4332915351530207e-06, "loss": 1.8869, "step": 31644 }, { "epoch": 0.78, "learning_rate": 2.4327696135556867e-06, "loss": 1.951, "step": 31645 }, { "epoch": 0.78, "learning_rate": 2.4322477401867807e-06, "loss": 1.9783, "step": 31646 }, { "epoch": 0.78, "learning_rate": 2.431725915049623e-06, "loss": 2.0468, "step": 31647 }, { "epoch": 0.78, "learning_rate": 2.4312041381475426e-06, "loss": 1.8691, "step": 31648 }, { "epoch": 0.78, "learning_rate": 2.430682409483868e-06, "loss": 2.0701, "step": 31649 }, { "epoch": 0.78, "learning_rate": 2.4301607290619166e-06, "loss": 1.9875, "step": 31650 }, { "epoch": 0.78, "learning_rate": 2.4296390968850194e-06, "loss": 1.9144, "step": 31651 }, { "epoch": 0.78, "learning_rate": 2.4291175129565024e-06, "loss": 2.1068, "step": 31652 }, { "epoch": 0.78, "learning_rate": 2.428595977279683e-06, "loss": 1.9351, "step": 31653 }, { "epoch": 0.78, "learning_rate": 2.428074489857889e-06, "loss": 2.0064, "step": 31654 }, { "epoch": 0.78, "learning_rate": 2.4275530506944446e-06, "loss": 2.0199, "step": 31655 }, { "epoch": 0.78, "learning_rate": 2.4270316597926746e-06, "loss": 1.8107, "step": 31656 }, { "epoch": 0.78, "learning_rate": 2.4265103171558977e-06, "loss": 1.901, "step": 31657 }, { "epoch": 0.78, "learning_rate": 2.425989022787438e-06, "loss": 1.8948, "step": 31658 }, { "epoch": 0.78, "learning_rate": 2.425467776690622e-06, "loss": 1.9805, "step": 31659 }, { "epoch": 0.78, "learning_rate": 2.4249465788687645e-06, "loss": 1.9638, "step": 31660 }, { "epoch": 0.78, "learning_rate": 2.4244254293251923e-06, "loss": 1.7714, "step": 31661 }, { "epoch": 0.78, "learning_rate": 2.423904328063228e-06, "loss": 1.8383, "step": 31662 }, { "epoch": 0.78, "learning_rate": 2.4233832750861864e-06, "loss": 2.0217, "step": 31663 }, { "epoch": 0.78, "learning_rate": 2.422862270397397e-06, "loss": 1.9625, "step": 31664 }, { "epoch": 0.78, "learning_rate": 2.4223413140001718e-06, "loss": 1.9172, "step": 31665 }, { "epoch": 0.78, "learning_rate": 2.4218204058978347e-06, "loss": 1.9465, "step": 31666 }, { "epoch": 0.78, "learning_rate": 2.4212995460937096e-06, "loss": 1.9362, "step": 31667 }, { "epoch": 0.78, "learning_rate": 2.420778734591108e-06, "loss": 2.0065, "step": 31668 }, { "epoch": 0.78, "learning_rate": 2.4202579713933583e-06, "loss": 2.1076, "step": 31669 }, { "epoch": 0.78, "learning_rate": 2.4197372565037702e-06, "loss": 2.1475, "step": 31670 }, { "epoch": 0.78, "learning_rate": 2.419216589925668e-06, "loss": 2.0284, "step": 31671 }, { "epoch": 0.78, "learning_rate": 2.418695971662368e-06, "loss": 2.0982, "step": 31672 }, { "epoch": 0.78, "learning_rate": 2.41817540171719e-06, "loss": 1.951, "step": 31673 }, { "epoch": 0.78, "learning_rate": 2.417654880093454e-06, "loss": 2.0267, "step": 31674 }, { "epoch": 0.78, "learning_rate": 2.4171344067944714e-06, "loss": 2.0021, "step": 31675 }, { "epoch": 0.78, "learning_rate": 2.416613981823562e-06, "loss": 2.0141, "step": 31676 }, { "epoch": 0.78, "learning_rate": 2.4160936051840467e-06, "loss": 1.9739, "step": 31677 }, { "epoch": 0.78, "learning_rate": 2.415573276879235e-06, "loss": 1.9384, "step": 31678 }, { "epoch": 0.78, "learning_rate": 2.415052996912449e-06, "loss": 1.8706, "step": 31679 }, { "epoch": 0.78, "learning_rate": 2.414532765287e-06, "loss": 2.028, "step": 31680 }, { "epoch": 0.78, "learning_rate": 2.4140125820062043e-06, "loss": 2.0286, "step": 31681 }, { "epoch": 0.78, "learning_rate": 2.413492447073383e-06, "loss": 1.9634, "step": 31682 }, { "epoch": 0.78, "learning_rate": 2.412972360491843e-06, "loss": 1.7145, "step": 31683 }, { "epoch": 0.78, "learning_rate": 2.4124523222649064e-06, "loss": 2.0571, "step": 31684 }, { "epoch": 0.78, "learning_rate": 2.411932332395881e-06, "loss": 2.0377, "step": 31685 }, { "epoch": 0.78, "learning_rate": 2.4114123908880824e-06, "loss": 1.9417, "step": 31686 }, { "epoch": 0.78, "learning_rate": 2.4108924977448302e-06, "loss": 1.9013, "step": 31687 }, { "epoch": 0.78, "learning_rate": 2.4103726529694294e-06, "loss": 1.9295, "step": 31688 }, { "epoch": 0.78, "learning_rate": 2.409852856565197e-06, "loss": 1.7485, "step": 31689 }, { "epoch": 0.78, "learning_rate": 2.4093331085354466e-06, "loss": 1.9668, "step": 31690 }, { "epoch": 0.78, "learning_rate": 2.4088134088834893e-06, "loss": 1.9668, "step": 31691 }, { "epoch": 0.78, "learning_rate": 2.4082937576126407e-06, "loss": 1.9745, "step": 31692 }, { "epoch": 0.78, "learning_rate": 2.407774154726208e-06, "loss": 1.9185, "step": 31693 }, { "epoch": 0.78, "learning_rate": 2.407254600227508e-06, "loss": 1.949, "step": 31694 }, { "epoch": 0.78, "learning_rate": 2.4067350941198452e-06, "loss": 1.9248, "step": 31695 }, { "epoch": 0.78, "learning_rate": 2.4062156364065346e-06, "loss": 1.6949, "step": 31696 }, { "epoch": 0.78, "learning_rate": 2.405696227090889e-06, "loss": 1.9682, "step": 31697 }, { "epoch": 0.78, "learning_rate": 2.4051768661762134e-06, "loss": 1.9773, "step": 31698 }, { "epoch": 0.78, "learning_rate": 2.4046575536658243e-06, "loss": 2.0442, "step": 31699 }, { "epoch": 0.78, "learning_rate": 2.404138289563024e-06, "loss": 1.8317, "step": 31700 }, { "epoch": 0.78, "learning_rate": 2.403619073871126e-06, "loss": 1.8352, "step": 31701 }, { "epoch": 0.78, "learning_rate": 2.4030999065934425e-06, "loss": 1.6507, "step": 31702 }, { "epoch": 0.78, "learning_rate": 2.4025807877332764e-06, "loss": 1.9509, "step": 31703 }, { "epoch": 0.78, "learning_rate": 2.402061717293941e-06, "loss": 1.8209, "step": 31704 }, { "epoch": 0.78, "learning_rate": 2.40154269527874e-06, "loss": 1.9462, "step": 31705 }, { "epoch": 0.78, "learning_rate": 2.4010237216909836e-06, "loss": 2.1437, "step": 31706 }, { "epoch": 0.78, "learning_rate": 2.40050479653398e-06, "loss": 2.0482, "step": 31707 }, { "epoch": 0.78, "learning_rate": 2.3999859198110355e-06, "loss": 1.9923, "step": 31708 }, { "epoch": 0.78, "learning_rate": 2.3994670915254592e-06, "loss": 1.873, "step": 31709 }, { "epoch": 0.78, "learning_rate": 2.3989483116805546e-06, "loss": 2.0353, "step": 31710 }, { "epoch": 0.78, "learning_rate": 2.3984295802796296e-06, "loss": 1.9852, "step": 31711 }, { "epoch": 0.78, "learning_rate": 2.3979108973259925e-06, "loss": 1.9523, "step": 31712 }, { "epoch": 0.78, "learning_rate": 2.397392262822944e-06, "loss": 2.1741, "step": 31713 }, { "epoch": 0.78, "learning_rate": 2.3968736767737922e-06, "loss": 1.9169, "step": 31714 }, { "epoch": 0.78, "learning_rate": 2.3963551391818453e-06, "loss": 2.0232, "step": 31715 }, { "epoch": 0.78, "learning_rate": 2.3958366500504016e-06, "loss": 1.926, "step": 31716 }, { "epoch": 0.78, "learning_rate": 2.395318209382772e-06, "loss": 1.8592, "step": 31717 }, { "epoch": 0.78, "learning_rate": 2.3947998171822552e-06, "loss": 2.0112, "step": 31718 }, { "epoch": 0.78, "learning_rate": 2.394281473452158e-06, "loss": 2.1233, "step": 31719 }, { "epoch": 0.78, "learning_rate": 2.393763178195786e-06, "loss": 2.0687, "step": 31720 }, { "epoch": 0.78, "learning_rate": 2.3932449314164364e-06, "loss": 1.9926, "step": 31721 }, { "epoch": 0.78, "learning_rate": 2.39272673311742e-06, "loss": 1.8432, "step": 31722 }, { "epoch": 0.78, "learning_rate": 2.3922085833020312e-06, "loss": 2.003, "step": 31723 }, { "epoch": 0.78, "learning_rate": 2.3916904819735775e-06, "loss": 2.0395, "step": 31724 }, { "epoch": 0.78, "learning_rate": 2.39117242913536e-06, "loss": 1.9144, "step": 31725 }, { "epoch": 0.78, "learning_rate": 2.3906544247906795e-06, "loss": 2.0144, "step": 31726 }, { "epoch": 0.78, "learning_rate": 2.3901364689428408e-06, "loss": 2.0074, "step": 31727 }, { "epoch": 0.78, "learning_rate": 2.3896185615951404e-06, "loss": 1.9109, "step": 31728 }, { "epoch": 0.78, "learning_rate": 2.3891007027508807e-06, "loss": 1.8762, "step": 31729 }, { "epoch": 0.78, "learning_rate": 2.388582892413365e-06, "loss": 1.9926, "step": 31730 }, { "epoch": 0.78, "learning_rate": 2.388065130585888e-06, "loss": 1.8583, "step": 31731 }, { "epoch": 0.78, "learning_rate": 2.3875474172717574e-06, "loss": 1.9794, "step": 31732 }, { "epoch": 0.78, "learning_rate": 2.3870297524742623e-06, "loss": 1.9199, "step": 31733 }, { "epoch": 0.78, "learning_rate": 2.3865121361967093e-06, "loss": 2.001, "step": 31734 }, { "epoch": 0.78, "learning_rate": 2.385994568442398e-06, "loss": 1.8301, "step": 31735 }, { "epoch": 0.78, "learning_rate": 2.3854770492146216e-06, "loss": 1.8748, "step": 31736 }, { "epoch": 0.78, "learning_rate": 2.3849595785166847e-06, "loss": 1.936, "step": 31737 }, { "epoch": 0.78, "learning_rate": 2.3844421563518795e-06, "loss": 1.9459, "step": 31738 }, { "epoch": 0.78, "learning_rate": 2.3839247827235055e-06, "loss": 1.9628, "step": 31739 }, { "epoch": 0.78, "learning_rate": 2.383407457634864e-06, "loss": 1.8781, "step": 31740 }, { "epoch": 0.78, "learning_rate": 2.3828901810892467e-06, "loss": 1.978, "step": 31741 }, { "epoch": 0.78, "learning_rate": 2.382372953089952e-06, "loss": 2.0473, "step": 31742 }, { "epoch": 0.78, "learning_rate": 2.381855773640277e-06, "loss": 1.8606, "step": 31743 }, { "epoch": 0.78, "learning_rate": 2.381338642743518e-06, "loss": 1.8949, "step": 31744 }, { "epoch": 0.78, "learning_rate": 2.380821560402974e-06, "loss": 2.0019, "step": 31745 }, { "epoch": 0.78, "learning_rate": 2.3803045266219327e-06, "loss": 1.9666, "step": 31746 }, { "epoch": 0.78, "learning_rate": 2.3797875414036985e-06, "loss": 1.8369, "step": 31747 }, { "epoch": 0.78, "learning_rate": 2.379270604751558e-06, "loss": 2.0163, "step": 31748 }, { "epoch": 0.78, "learning_rate": 2.378753716668809e-06, "loss": 1.9524, "step": 31749 }, { "epoch": 0.78, "learning_rate": 2.3782368771587507e-06, "loss": 1.9436, "step": 31750 }, { "epoch": 0.78, "learning_rate": 2.3777200862246686e-06, "loss": 1.8514, "step": 31751 }, { "epoch": 0.78, "learning_rate": 2.377203343869863e-06, "loss": 1.9787, "step": 31752 }, { "epoch": 0.78, "learning_rate": 2.3766866500976228e-06, "loss": 1.9869, "step": 31753 }, { "epoch": 0.78, "learning_rate": 2.3761700049112425e-06, "loss": 1.9368, "step": 31754 }, { "epoch": 0.78, "learning_rate": 2.3756534083140183e-06, "loss": 1.8487, "step": 31755 }, { "epoch": 0.78, "learning_rate": 2.3751368603092363e-06, "loss": 1.8724, "step": 31756 }, { "epoch": 0.78, "learning_rate": 2.374620360900195e-06, "loss": 1.9689, "step": 31757 }, { "epoch": 0.78, "learning_rate": 2.3741039100901796e-06, "loss": 1.7962, "step": 31758 }, { "epoch": 0.78, "learning_rate": 2.3735875078824865e-06, "loss": 1.7998, "step": 31759 }, { "epoch": 0.78, "learning_rate": 2.373071154280405e-06, "loss": 1.9725, "step": 31760 }, { "epoch": 0.78, "learning_rate": 2.3725548492872263e-06, "loss": 1.809, "step": 31761 }, { "epoch": 0.78, "learning_rate": 2.372038592906245e-06, "loss": 1.7657, "step": 31762 }, { "epoch": 0.78, "learning_rate": 2.371522385140744e-06, "loss": 1.7089, "step": 31763 }, { "epoch": 0.78, "learning_rate": 2.371006225994017e-06, "loss": 1.7972, "step": 31764 }, { "epoch": 0.78, "learning_rate": 2.3704901154693573e-06, "loss": 1.9388, "step": 31765 }, { "epoch": 0.78, "learning_rate": 2.369974053570047e-06, "loss": 1.9419, "step": 31766 }, { "epoch": 0.78, "learning_rate": 2.3694580402993806e-06, "loss": 1.8641, "step": 31767 }, { "epoch": 0.78, "learning_rate": 2.368942075660643e-06, "loss": 1.8443, "step": 31768 }, { "epoch": 0.78, "learning_rate": 2.3684261596571233e-06, "loss": 2.0393, "step": 31769 }, { "epoch": 0.78, "learning_rate": 2.367910292292115e-06, "loss": 1.8245, "step": 31770 }, { "epoch": 0.78, "learning_rate": 2.367394473568898e-06, "loss": 1.8695, "step": 31771 }, { "epoch": 0.78, "learning_rate": 2.3668787034907638e-06, "loss": 1.9915, "step": 31772 }, { "epoch": 0.78, "learning_rate": 2.3663629820610013e-06, "loss": 1.8309, "step": 31773 }, { "epoch": 0.78, "learning_rate": 2.3658473092828936e-06, "loss": 1.8643, "step": 31774 }, { "epoch": 0.78, "learning_rate": 2.3653316851597306e-06, "loss": 2.038, "step": 31775 }, { "epoch": 0.78, "learning_rate": 2.364816109694794e-06, "loss": 1.9229, "step": 31776 }, { "epoch": 0.78, "learning_rate": 2.3643005828913736e-06, "loss": 1.9799, "step": 31777 }, { "epoch": 0.78, "learning_rate": 2.363785104752754e-06, "loss": 1.9935, "step": 31778 }, { "epoch": 0.78, "learning_rate": 2.363269675282219e-06, "loss": 2.0049, "step": 31779 }, { "epoch": 0.78, "learning_rate": 2.3627542944830596e-06, "loss": 2.0323, "step": 31780 }, { "epoch": 0.78, "learning_rate": 2.3622389623585517e-06, "loss": 2.0183, "step": 31781 }, { "epoch": 0.78, "learning_rate": 2.361723678911986e-06, "loss": 1.9335, "step": 31782 }, { "epoch": 0.78, "learning_rate": 2.361208444146645e-06, "loss": 1.9203, "step": 31783 }, { "epoch": 0.78, "learning_rate": 2.3606932580658102e-06, "loss": 1.9035, "step": 31784 }, { "epoch": 0.78, "learning_rate": 2.3601781206727704e-06, "loss": 1.9477, "step": 31785 }, { "epoch": 0.78, "learning_rate": 2.3596630319708013e-06, "loss": 2.0441, "step": 31786 }, { "epoch": 0.78, "learning_rate": 2.35914799196319e-06, "loss": 1.8878, "step": 31787 }, { "epoch": 0.78, "learning_rate": 2.358633000653221e-06, "loss": 1.9666, "step": 31788 }, { "epoch": 0.78, "learning_rate": 2.358118058044172e-06, "loss": 2.1097, "step": 31789 }, { "epoch": 0.78, "learning_rate": 2.3576031641393294e-06, "loss": 1.8434, "step": 31790 }, { "epoch": 0.78, "learning_rate": 2.35708831894197e-06, "loss": 1.9251, "step": 31791 }, { "epoch": 0.78, "learning_rate": 2.3565735224553775e-06, "loss": 1.8448, "step": 31792 }, { "epoch": 0.78, "learning_rate": 2.3560587746828355e-06, "loss": 1.964, "step": 31793 }, { "epoch": 0.78, "learning_rate": 2.3555440756276193e-06, "loss": 2.0681, "step": 31794 }, { "epoch": 0.78, "learning_rate": 2.355029425293012e-06, "loss": 1.9754, "step": 31795 }, { "epoch": 0.78, "learning_rate": 2.3545148236822932e-06, "loss": 2.0281, "step": 31796 }, { "epoch": 0.78, "learning_rate": 2.354000270798743e-06, "loss": 2.0389, "step": 31797 }, { "epoch": 0.78, "learning_rate": 2.3534857666456435e-06, "loss": 1.8867, "step": 31798 }, { "epoch": 0.78, "learning_rate": 2.3529713112262687e-06, "loss": 1.8304, "step": 31799 }, { "epoch": 0.78, "learning_rate": 2.3524569045439026e-06, "loss": 1.8475, "step": 31800 }, { "epoch": 0.78, "learning_rate": 2.3519425466018184e-06, "loss": 2.2699, "step": 31801 }, { "epoch": 0.78, "learning_rate": 2.3514282374032973e-06, "loss": 1.8626, "step": 31802 }, { "epoch": 0.78, "learning_rate": 2.3509139769516175e-06, "loss": 1.853, "step": 31803 }, { "epoch": 0.78, "learning_rate": 2.350399765250054e-06, "loss": 1.8037, "step": 31804 }, { "epoch": 0.78, "learning_rate": 2.349885602301889e-06, "loss": 1.8963, "step": 31805 }, { "epoch": 0.78, "learning_rate": 2.3493714881103936e-06, "loss": 2.0197, "step": 31806 }, { "epoch": 0.78, "learning_rate": 2.3488574226788465e-06, "loss": 1.8442, "step": 31807 }, { "epoch": 0.78, "learning_rate": 2.348343406010527e-06, "loss": 1.9819, "step": 31808 }, { "epoch": 0.78, "learning_rate": 2.3478294381087053e-06, "loss": 1.7633, "step": 31809 }, { "epoch": 0.78, "learning_rate": 2.347315518976664e-06, "loss": 2.0949, "step": 31810 }, { "epoch": 0.78, "learning_rate": 2.3468016486176714e-06, "loss": 1.9395, "step": 31811 }, { "epoch": 0.78, "learning_rate": 2.3462878270350065e-06, "loss": 1.9331, "step": 31812 }, { "epoch": 0.78, "learning_rate": 2.345774054231944e-06, "loss": 2.1675, "step": 31813 }, { "epoch": 0.78, "learning_rate": 2.345260330211757e-06, "loss": 1.9524, "step": 31814 }, { "epoch": 0.78, "learning_rate": 2.344746654977723e-06, "loss": 1.8449, "step": 31815 }, { "epoch": 0.78, "learning_rate": 2.344233028533112e-06, "loss": 1.875, "step": 31816 }, { "epoch": 0.78, "learning_rate": 2.3437194508811977e-06, "loss": 1.7895, "step": 31817 }, { "epoch": 0.78, "learning_rate": 2.343205922025258e-06, "loss": 1.9626, "step": 31818 }, { "epoch": 0.78, "learning_rate": 2.3426924419685593e-06, "loss": 1.8861, "step": 31819 }, { "epoch": 0.78, "learning_rate": 2.342179010714379e-06, "loss": 2.009, "step": 31820 }, { "epoch": 0.78, "learning_rate": 2.3416656282659855e-06, "loss": 1.8796, "step": 31821 }, { "epoch": 0.78, "learning_rate": 2.341152294626653e-06, "loss": 2.026, "step": 31822 }, { "epoch": 0.78, "learning_rate": 2.3406390097996546e-06, "loss": 1.9332, "step": 31823 }, { "epoch": 0.78, "learning_rate": 2.340125773788258e-06, "loss": 1.844, "step": 31824 }, { "epoch": 0.78, "learning_rate": 2.3396125865957387e-06, "loss": 2.0297, "step": 31825 }, { "epoch": 0.78, "learning_rate": 2.3390994482253615e-06, "loss": 1.9198, "step": 31826 }, { "epoch": 0.78, "learning_rate": 2.3385863586804015e-06, "loss": 1.9319, "step": 31827 }, { "epoch": 0.78, "learning_rate": 2.338073317964129e-06, "loss": 1.7962, "step": 31828 }, { "epoch": 0.78, "learning_rate": 2.3375603260798097e-06, "loss": 1.9054, "step": 31829 }, { "epoch": 0.78, "learning_rate": 2.337047383030715e-06, "loss": 1.8859, "step": 31830 }, { "epoch": 0.78, "learning_rate": 2.3365344888201157e-06, "loss": 1.8257, "step": 31831 }, { "epoch": 0.78, "learning_rate": 2.3360216434512793e-06, "loss": 1.9568, "step": 31832 }, { "epoch": 0.78, "learning_rate": 2.3355088469274765e-06, "loss": 1.9974, "step": 31833 }, { "epoch": 0.78, "learning_rate": 2.3349960992519716e-06, "loss": 1.8432, "step": 31834 }, { "epoch": 0.78, "learning_rate": 2.3344834004280347e-06, "loss": 1.8578, "step": 31835 }, { "epoch": 0.78, "learning_rate": 2.333970750458936e-06, "loss": 1.764, "step": 31836 }, { "epoch": 0.78, "learning_rate": 2.333458149347937e-06, "loss": 1.906, "step": 31837 }, { "epoch": 0.78, "learning_rate": 2.3329455970983095e-06, "loss": 2.0526, "step": 31838 }, { "epoch": 0.78, "learning_rate": 2.3324330937133165e-06, "loss": 2.019, "step": 31839 }, { "epoch": 0.78, "learning_rate": 2.331920639196226e-06, "loss": 1.9665, "step": 31840 }, { "epoch": 0.78, "learning_rate": 2.3314082335503075e-06, "loss": 1.923, "step": 31841 }, { "epoch": 0.78, "learning_rate": 2.3308958767788204e-06, "loss": 1.9928, "step": 31842 }, { "epoch": 0.78, "learning_rate": 2.330383568885035e-06, "loss": 1.7624, "step": 31843 }, { "epoch": 0.78, "learning_rate": 2.3298713098722126e-06, "loss": 1.8968, "step": 31844 }, { "epoch": 0.78, "learning_rate": 2.32935909974362e-06, "loss": 1.8702, "step": 31845 }, { "epoch": 0.78, "learning_rate": 2.3288469385025246e-06, "loss": 2.1117, "step": 31846 }, { "epoch": 0.78, "learning_rate": 2.3283348261521844e-06, "loss": 1.971, "step": 31847 }, { "epoch": 0.79, "learning_rate": 2.327822762695867e-06, "loss": 2.0242, "step": 31848 }, { "epoch": 0.79, "learning_rate": 2.327310748136835e-06, "loss": 2.01, "step": 31849 }, { "epoch": 0.79, "learning_rate": 2.326798782478352e-06, "loss": 1.9303, "step": 31850 }, { "epoch": 0.79, "learning_rate": 2.3262868657236835e-06, "loss": 1.9473, "step": 31851 }, { "epoch": 0.79, "learning_rate": 2.325774997876088e-06, "loss": 1.9566, "step": 31852 }, { "epoch": 0.79, "learning_rate": 2.3252631789388315e-06, "loss": 1.8962, "step": 31853 }, { "epoch": 0.79, "learning_rate": 2.32475140891517e-06, "loss": 1.9423, "step": 31854 }, { "epoch": 0.79, "learning_rate": 2.3242396878083708e-06, "loss": 2.0231, "step": 31855 }, { "epoch": 0.79, "learning_rate": 2.323728015621696e-06, "loss": 1.9025, "step": 31856 }, { "epoch": 0.79, "learning_rate": 2.3232163923584006e-06, "loss": 1.9046, "step": 31857 }, { "epoch": 0.79, "learning_rate": 2.3227048180217515e-06, "loss": 1.7625, "step": 31858 }, { "epoch": 0.79, "learning_rate": 2.322193292615005e-06, "loss": 1.8355, "step": 31859 }, { "epoch": 0.79, "learning_rate": 2.321681816141422e-06, "loss": 1.9299, "step": 31860 }, { "epoch": 0.79, "learning_rate": 2.3211703886042657e-06, "loss": 1.9876, "step": 31861 }, { "epoch": 0.79, "learning_rate": 2.3206590100067905e-06, "loss": 1.8563, "step": 31862 }, { "epoch": 0.79, "learning_rate": 2.3201476803522617e-06, "loss": 2.0132, "step": 31863 }, { "epoch": 0.79, "learning_rate": 2.31963639964393e-06, "loss": 1.8498, "step": 31864 }, { "epoch": 0.79, "learning_rate": 2.319125167885059e-06, "loss": 1.988, "step": 31865 }, { "epoch": 0.79, "learning_rate": 2.318613985078907e-06, "loss": 1.8412, "step": 31866 }, { "epoch": 0.79, "learning_rate": 2.3181028512287317e-06, "loss": 1.9242, "step": 31867 }, { "epoch": 0.79, "learning_rate": 2.317591766337792e-06, "loss": 1.9653, "step": 31868 }, { "epoch": 0.79, "learning_rate": 2.3170807304093422e-06, "loss": 1.8777, "step": 31869 }, { "epoch": 0.79, "learning_rate": 2.3165697434466392e-06, "loss": 1.8146, "step": 31870 }, { "epoch": 0.79, "learning_rate": 2.3160588054529455e-06, "loss": 1.7447, "step": 31871 }, { "epoch": 0.79, "learning_rate": 2.3155479164315085e-06, "loss": 1.8035, "step": 31872 }, { "epoch": 0.79, "learning_rate": 2.315037076385593e-06, "loss": 2.0144, "step": 31873 }, { "epoch": 0.79, "learning_rate": 2.314526285318448e-06, "loss": 1.9884, "step": 31874 }, { "epoch": 0.79, "learning_rate": 2.3140155432333313e-06, "loss": 1.8079, "step": 31875 }, { "epoch": 0.79, "learning_rate": 2.3135048501335013e-06, "loss": 1.9319, "step": 31876 }, { "epoch": 0.79, "learning_rate": 2.3129942060222067e-06, "loss": 2.0273, "step": 31877 }, { "epoch": 0.79, "learning_rate": 2.3124836109027084e-06, "loss": 1.9192, "step": 31878 }, { "epoch": 0.79, "learning_rate": 2.3119730647782536e-06, "loss": 1.8513, "step": 31879 }, { "epoch": 0.79, "learning_rate": 2.3114625676521007e-06, "loss": 2.1008, "step": 31880 }, { "epoch": 0.79, "learning_rate": 2.3109521195275055e-06, "loss": 1.8782, "step": 31881 }, { "epoch": 0.79, "learning_rate": 2.3104417204077147e-06, "loss": 2.0721, "step": 31882 }, { "epoch": 0.79, "learning_rate": 2.309931370295985e-06, "loss": 1.9049, "step": 31883 }, { "epoch": 0.79, "learning_rate": 2.309421069195569e-06, "loss": 1.8753, "step": 31884 }, { "epoch": 0.79, "learning_rate": 2.3089108171097185e-06, "loss": 2.0102, "step": 31885 }, { "epoch": 0.79, "learning_rate": 2.3084006140416894e-06, "loss": 1.8195, "step": 31886 }, { "epoch": 0.79, "learning_rate": 2.307890459994726e-06, "loss": 1.9516, "step": 31887 }, { "epoch": 0.79, "learning_rate": 2.307380354972084e-06, "loss": 1.9124, "step": 31888 }, { "epoch": 0.79, "learning_rate": 2.3068702989770164e-06, "loss": 1.835, "step": 31889 }, { "epoch": 0.79, "learning_rate": 2.306360292012769e-06, "loss": 1.9483, "step": 31890 }, { "epoch": 0.79, "learning_rate": 2.305850334082598e-06, "loss": 2.1072, "step": 31891 }, { "epoch": 0.79, "learning_rate": 2.305340425189747e-06, "loss": 1.9977, "step": 31892 }, { "epoch": 0.79, "learning_rate": 2.304830565337469e-06, "loss": 2.041, "step": 31893 }, { "epoch": 0.79, "learning_rate": 2.304320754529016e-06, "loss": 1.962, "step": 31894 }, { "epoch": 0.79, "learning_rate": 2.3038109927676333e-06, "loss": 2.0079, "step": 31895 }, { "epoch": 0.79, "learning_rate": 2.3033012800565725e-06, "loss": 1.9255, "step": 31896 }, { "epoch": 0.79, "learning_rate": 2.3027916163990794e-06, "loss": 1.8552, "step": 31897 }, { "epoch": 0.79, "learning_rate": 2.3022820017984025e-06, "loss": 1.8714, "step": 31898 }, { "epoch": 0.79, "learning_rate": 2.3017724362577954e-06, "loss": 1.9386, "step": 31899 }, { "epoch": 0.79, "learning_rate": 2.301262919780497e-06, "loss": 2.0425, "step": 31900 }, { "epoch": 0.79, "learning_rate": 2.3007534523697593e-06, "loss": 1.913, "step": 31901 }, { "epoch": 0.79, "learning_rate": 2.30024403402883e-06, "loss": 1.8767, "step": 31902 }, { "epoch": 0.79, "learning_rate": 2.2997346647609542e-06, "loss": 1.9238, "step": 31903 }, { "epoch": 0.79, "learning_rate": 2.2992253445693815e-06, "loss": 1.9848, "step": 31904 }, { "epoch": 0.79, "learning_rate": 2.2987160734573523e-06, "loss": 1.9162, "step": 31905 }, { "epoch": 0.79, "learning_rate": 2.2982068514281187e-06, "loss": 1.814, "step": 31906 }, { "epoch": 0.79, "learning_rate": 2.297697678484919e-06, "loss": 1.8715, "step": 31907 }, { "epoch": 0.79, "learning_rate": 2.2971885546310023e-06, "loss": 2.0037, "step": 31908 }, { "epoch": 0.79, "learning_rate": 2.2966794798696167e-06, "loss": 2.0342, "step": 31909 }, { "epoch": 0.79, "learning_rate": 2.2961704542039997e-06, "loss": 1.9095, "step": 31910 }, { "epoch": 0.79, "learning_rate": 2.295661477637403e-06, "loss": 2.2148, "step": 31911 }, { "epoch": 0.79, "learning_rate": 2.295152550173062e-06, "loss": 1.9514, "step": 31912 }, { "epoch": 0.79, "learning_rate": 2.2946436718142262e-06, "loss": 2.0762, "step": 31913 }, { "epoch": 0.79, "learning_rate": 2.2941348425641398e-06, "loss": 1.7472, "step": 31914 }, { "epoch": 0.79, "learning_rate": 2.2936260624260396e-06, "loss": 1.7855, "step": 31915 }, { "epoch": 0.79, "learning_rate": 2.2931173314031762e-06, "loss": 1.9139, "step": 31916 }, { "epoch": 0.79, "learning_rate": 2.2926086494987845e-06, "loss": 1.8853, "step": 31917 }, { "epoch": 0.79, "learning_rate": 2.292100016716109e-06, "loss": 1.9092, "step": 31918 }, { "epoch": 0.79, "learning_rate": 2.2915914330583923e-06, "loss": 1.9628, "step": 31919 }, { "epoch": 0.79, "learning_rate": 2.2910828985288746e-06, "loss": 1.8165, "step": 31920 }, { "epoch": 0.79, "learning_rate": 2.2905744131308018e-06, "loss": 1.9992, "step": 31921 }, { "epoch": 0.79, "learning_rate": 2.2900659768674085e-06, "loss": 1.9247, "step": 31922 }, { "epoch": 0.79, "learning_rate": 2.2895575897419355e-06, "loss": 2.0012, "step": 31923 }, { "epoch": 0.79, "learning_rate": 2.289049251757629e-06, "loss": 1.8809, "step": 31924 }, { "epoch": 0.79, "learning_rate": 2.288540962917721e-06, "loss": 1.9376, "step": 31925 }, { "epoch": 0.79, "learning_rate": 2.2880327232254575e-06, "loss": 1.9533, "step": 31926 }, { "epoch": 0.79, "learning_rate": 2.2875245326840725e-06, "loss": 1.8112, "step": 31927 }, { "epoch": 0.79, "learning_rate": 2.2870163912968056e-06, "loss": 1.8767, "step": 31928 }, { "epoch": 0.79, "learning_rate": 2.2865082990669005e-06, "loss": 1.7506, "step": 31929 }, { "epoch": 0.79, "learning_rate": 2.2860002559975892e-06, "loss": 1.9332, "step": 31930 }, { "epoch": 0.79, "learning_rate": 2.285492262092115e-06, "loss": 2.0631, "step": 31931 }, { "epoch": 0.79, "learning_rate": 2.2849843173537103e-06, "loss": 1.9128, "step": 31932 }, { "epoch": 0.79, "learning_rate": 2.284476421785614e-06, "loss": 2.0186, "step": 31933 }, { "epoch": 0.79, "learning_rate": 2.2839685753910667e-06, "loss": 2.101, "step": 31934 }, { "epoch": 0.79, "learning_rate": 2.2834607781732996e-06, "loss": 1.8394, "step": 31935 }, { "epoch": 0.79, "learning_rate": 2.2829530301355516e-06, "loss": 1.8544, "step": 31936 }, { "epoch": 0.79, "learning_rate": 2.2824453312810593e-06, "loss": 2.016, "step": 31937 }, { "epoch": 0.79, "learning_rate": 2.281937681613058e-06, "loss": 1.9788, "step": 31938 }, { "epoch": 0.79, "learning_rate": 2.2814300811347854e-06, "loss": 1.9297, "step": 31939 }, { "epoch": 0.79, "learning_rate": 2.280922529849472e-06, "loss": 1.8286, "step": 31940 }, { "epoch": 0.79, "learning_rate": 2.2804150277603575e-06, "loss": 1.8323, "step": 31941 }, { "epoch": 0.79, "learning_rate": 2.27990757487067e-06, "loss": 2.2688, "step": 31942 }, { "epoch": 0.79, "learning_rate": 2.279400171183649e-06, "loss": 1.9218, "step": 31943 }, { "epoch": 0.79, "learning_rate": 2.2788928167025292e-06, "loss": 1.9159, "step": 31944 }, { "epoch": 0.79, "learning_rate": 2.2783855114305385e-06, "loss": 1.9613, "step": 31945 }, { "epoch": 0.79, "learning_rate": 2.2778782553709156e-06, "loss": 1.9932, "step": 31946 }, { "epoch": 0.79, "learning_rate": 2.2773710485268897e-06, "loss": 1.9602, "step": 31947 }, { "epoch": 0.79, "learning_rate": 2.276863890901694e-06, "loss": 1.8269, "step": 31948 }, { "epoch": 0.79, "learning_rate": 2.2763567824985645e-06, "loss": 2.0861, "step": 31949 }, { "epoch": 0.79, "learning_rate": 2.2758497233207277e-06, "loss": 1.9174, "step": 31950 }, { "epoch": 0.79, "learning_rate": 2.2753427133714166e-06, "loss": 1.9512, "step": 31951 }, { "epoch": 0.79, "learning_rate": 2.274835752653868e-06, "loss": 1.8107, "step": 31952 }, { "epoch": 0.79, "learning_rate": 2.274328841171306e-06, "loss": 1.9251, "step": 31953 }, { "epoch": 0.79, "learning_rate": 2.2738219789269633e-06, "loss": 1.882, "step": 31954 }, { "epoch": 0.79, "learning_rate": 2.2733151659240716e-06, "loss": 2.1206, "step": 31955 }, { "epoch": 0.79, "learning_rate": 2.2728084021658603e-06, "loss": 2.1214, "step": 31956 }, { "epoch": 0.79, "learning_rate": 2.2723016876555613e-06, "loss": 1.9639, "step": 31957 }, { "epoch": 0.79, "learning_rate": 2.2717950223964004e-06, "loss": 1.9973, "step": 31958 }, { "epoch": 0.79, "learning_rate": 2.2712884063916094e-06, "loss": 1.9424, "step": 31959 }, { "epoch": 0.79, "learning_rate": 2.270781839644415e-06, "loss": 1.8766, "step": 31960 }, { "epoch": 0.79, "learning_rate": 2.2702753221580463e-06, "loss": 2.0921, "step": 31961 }, { "epoch": 0.79, "learning_rate": 2.2697688539357353e-06, "loss": 1.9191, "step": 31962 }, { "epoch": 0.79, "learning_rate": 2.2692624349807023e-06, "loss": 2.0302, "step": 31963 }, { "epoch": 0.79, "learning_rate": 2.2687560652961837e-06, "loss": 1.9377, "step": 31964 }, { "epoch": 0.79, "learning_rate": 2.268249744885399e-06, "loss": 1.7948, "step": 31965 }, { "epoch": 0.79, "learning_rate": 2.267743473751578e-06, "loss": 1.8132, "step": 31966 }, { "epoch": 0.79, "learning_rate": 2.26723725189795e-06, "loss": 1.9959, "step": 31967 }, { "epoch": 0.79, "learning_rate": 2.2667310793277374e-06, "loss": 1.8142, "step": 31968 }, { "epoch": 0.79, "learning_rate": 2.2662249560441697e-06, "loss": 1.901, "step": 31969 }, { "epoch": 0.79, "learning_rate": 2.265718882050468e-06, "loss": 1.9663, "step": 31970 }, { "epoch": 0.79, "learning_rate": 2.265212857349861e-06, "loss": 1.8019, "step": 31971 }, { "epoch": 0.79, "learning_rate": 2.264706881945573e-06, "loss": 1.9661, "step": 31972 }, { "epoch": 0.79, "learning_rate": 2.2642009558408285e-06, "loss": 1.829, "step": 31973 }, { "epoch": 0.79, "learning_rate": 2.2636950790388556e-06, "loss": 1.9248, "step": 31974 }, { "epoch": 0.79, "learning_rate": 2.263189251542872e-06, "loss": 1.7378, "step": 31975 }, { "epoch": 0.79, "learning_rate": 2.2626834733561054e-06, "loss": 1.844, "step": 31976 }, { "epoch": 0.79, "learning_rate": 2.2621777444817803e-06, "loss": 1.7654, "step": 31977 }, { "epoch": 0.79, "learning_rate": 2.261672064923116e-06, "loss": 1.8511, "step": 31978 }, { "epoch": 0.79, "learning_rate": 2.2611664346833396e-06, "loss": 1.853, "step": 31979 }, { "epoch": 0.79, "learning_rate": 2.2606608537656695e-06, "loss": 2.0375, "step": 31980 }, { "epoch": 0.79, "learning_rate": 2.2601553221733295e-06, "loss": 1.9037, "step": 31981 }, { "epoch": 0.79, "learning_rate": 2.2596498399095444e-06, "loss": 1.8365, "step": 31982 }, { "epoch": 0.79, "learning_rate": 2.2591444069775315e-06, "loss": 1.8775, "step": 31983 }, { "epoch": 0.79, "learning_rate": 2.2586390233805165e-06, "loss": 1.8962, "step": 31984 }, { "epoch": 0.79, "learning_rate": 2.258133689121714e-06, "loss": 1.9536, "step": 31985 }, { "epoch": 0.79, "learning_rate": 2.2576284042043496e-06, "loss": 1.7905, "step": 31986 }, { "epoch": 0.79, "learning_rate": 2.2571231686316453e-06, "loss": 1.9317, "step": 31987 }, { "epoch": 0.79, "learning_rate": 2.256617982406816e-06, "loss": 1.99, "step": 31988 }, { "epoch": 0.79, "learning_rate": 2.256112845533083e-06, "loss": 2.1163, "step": 31989 }, { "epoch": 0.79, "learning_rate": 2.255607758013666e-06, "loss": 1.801, "step": 31990 }, { "epoch": 0.79, "learning_rate": 2.255102719851785e-06, "loss": 1.9328, "step": 31991 }, { "epoch": 0.79, "learning_rate": 2.254597731050662e-06, "loss": 1.7713, "step": 31992 }, { "epoch": 0.79, "learning_rate": 2.254092791613508e-06, "loss": 1.8229, "step": 31993 }, { "epoch": 0.79, "learning_rate": 2.253587901543548e-06, "loss": 1.8525, "step": 31994 }, { "epoch": 0.79, "learning_rate": 2.2530830608439947e-06, "loss": 2.0073, "step": 31995 }, { "epoch": 0.79, "learning_rate": 2.2525782695180675e-06, "loss": 1.9533, "step": 31996 }, { "epoch": 0.79, "learning_rate": 2.2520735275689866e-06, "loss": 1.9496, "step": 31997 }, { "epoch": 0.79, "learning_rate": 2.251568834999962e-06, "loss": 1.9207, "step": 31998 }, { "epoch": 0.79, "learning_rate": 2.2510641918142186e-06, "loss": 1.9391, "step": 31999 }, { "epoch": 0.79, "learning_rate": 2.2505595980149652e-06, "loss": 1.8008, "step": 32000 }, { "epoch": 0.79, "learning_rate": 2.2500550536054212e-06, "loss": 1.9089, "step": 32001 }, { "epoch": 0.79, "learning_rate": 2.249550558588803e-06, "loss": 1.7463, "step": 32002 }, { "epoch": 0.79, "learning_rate": 2.249046112968324e-06, "loss": 1.9803, "step": 32003 }, { "epoch": 0.79, "learning_rate": 2.2485417167472002e-06, "loss": 1.9266, "step": 32004 }, { "epoch": 0.79, "learning_rate": 2.2480373699286482e-06, "loss": 1.8869, "step": 32005 }, { "epoch": 0.79, "learning_rate": 2.247533072515877e-06, "loss": 1.8636, "step": 32006 }, { "epoch": 0.79, "learning_rate": 2.2470288245121053e-06, "loss": 1.9623, "step": 32007 }, { "epoch": 0.79, "learning_rate": 2.2465246259205455e-06, "loss": 1.915, "step": 32008 }, { "epoch": 0.79, "learning_rate": 2.2460204767444106e-06, "loss": 1.8697, "step": 32009 }, { "epoch": 0.79, "learning_rate": 2.245516376986916e-06, "loss": 2.0492, "step": 32010 }, { "epoch": 0.79, "learning_rate": 2.2450123266512715e-06, "loss": 2.0365, "step": 32011 }, { "epoch": 0.79, "learning_rate": 2.244508325740692e-06, "loss": 1.9952, "step": 32012 }, { "epoch": 0.79, "learning_rate": 2.244004374258386e-06, "loss": 2.066, "step": 32013 }, { "epoch": 0.79, "learning_rate": 2.243500472207568e-06, "loss": 1.9228, "step": 32014 }, { "epoch": 0.79, "learning_rate": 2.2429966195914523e-06, "loss": 1.8855, "step": 32015 }, { "epoch": 0.79, "learning_rate": 2.2424928164132444e-06, "loss": 2.0952, "step": 32016 }, { "epoch": 0.79, "learning_rate": 2.24198906267616e-06, "loss": 2.0529, "step": 32017 }, { "epoch": 0.79, "learning_rate": 2.241485358383405e-06, "loss": 2.0017, "step": 32018 }, { "epoch": 0.79, "learning_rate": 2.240981703538192e-06, "loss": 2.0034, "step": 32019 }, { "epoch": 0.79, "learning_rate": 2.2404780981437347e-06, "loss": 1.9739, "step": 32020 }, { "epoch": 0.79, "learning_rate": 2.239974542203236e-06, "loss": 1.8884, "step": 32021 }, { "epoch": 0.79, "learning_rate": 2.239471035719909e-06, "loss": 1.9078, "step": 32022 }, { "epoch": 0.79, "learning_rate": 2.2389675786969645e-06, "loss": 1.8009, "step": 32023 }, { "epoch": 0.79, "learning_rate": 2.2384641711376056e-06, "loss": 1.964, "step": 32024 }, { "epoch": 0.79, "learning_rate": 2.237960813045045e-06, "loss": 2.0165, "step": 32025 }, { "epoch": 0.79, "learning_rate": 2.2374575044224888e-06, "loss": 1.8847, "step": 32026 }, { "epoch": 0.79, "learning_rate": 2.2369542452731485e-06, "loss": 1.9883, "step": 32027 }, { "epoch": 0.79, "learning_rate": 2.2364510356002266e-06, "loss": 1.8798, "step": 32028 }, { "epoch": 0.79, "learning_rate": 2.2359478754069318e-06, "loss": 1.8504, "step": 32029 }, { "epoch": 0.79, "learning_rate": 2.2354447646964737e-06, "loss": 1.9238, "step": 32030 }, { "epoch": 0.79, "learning_rate": 2.2349417034720546e-06, "loss": 1.9975, "step": 32031 }, { "epoch": 0.79, "learning_rate": 2.2344386917368856e-06, "loss": 1.963, "step": 32032 }, { "epoch": 0.79, "learning_rate": 2.2339357294941656e-06, "loss": 1.8754, "step": 32033 }, { "epoch": 0.79, "learning_rate": 2.233432816747105e-06, "loss": 1.8944, "step": 32034 }, { "epoch": 0.79, "learning_rate": 2.23292995349891e-06, "loss": 1.8583, "step": 32035 }, { "epoch": 0.79, "learning_rate": 2.2324271397527818e-06, "loss": 1.9228, "step": 32036 }, { "epoch": 0.79, "learning_rate": 2.2319243755119304e-06, "loss": 1.9715, "step": 32037 }, { "epoch": 0.79, "learning_rate": 2.2314216607795513e-06, "loss": 1.9988, "step": 32038 }, { "epoch": 0.79, "learning_rate": 2.230918995558855e-06, "loss": 1.8556, "step": 32039 }, { "epoch": 0.79, "learning_rate": 2.2304163798530444e-06, "loss": 2.0215, "step": 32040 }, { "epoch": 0.79, "learning_rate": 2.229913813665324e-06, "loss": 1.9039, "step": 32041 }, { "epoch": 0.79, "learning_rate": 2.2294112969988933e-06, "loss": 1.8414, "step": 32042 }, { "epoch": 0.79, "learning_rate": 2.228908829856956e-06, "loss": 1.8623, "step": 32043 }, { "epoch": 0.79, "learning_rate": 2.2284064122427153e-06, "loss": 2.0216, "step": 32044 }, { "epoch": 0.79, "learning_rate": 2.2279040441593758e-06, "loss": 1.9444, "step": 32045 }, { "epoch": 0.79, "learning_rate": 2.227401725610134e-06, "loss": 1.9665, "step": 32046 }, { "epoch": 0.79, "learning_rate": 2.226899456598197e-06, "loss": 2.1502, "step": 32047 }, { "epoch": 0.79, "learning_rate": 2.2263972371267606e-06, "loss": 1.8216, "step": 32048 }, { "epoch": 0.79, "learning_rate": 2.2258950671990275e-06, "loss": 1.8678, "step": 32049 }, { "epoch": 0.79, "learning_rate": 2.225392946818201e-06, "loss": 2.0192, "step": 32050 }, { "epoch": 0.79, "learning_rate": 2.2248908759874766e-06, "loss": 2.0411, "step": 32051 }, { "epoch": 0.79, "learning_rate": 2.22438885471006e-06, "loss": 1.7387, "step": 32052 }, { "epoch": 0.79, "learning_rate": 2.2238868829891435e-06, "loss": 1.9105, "step": 32053 }, { "epoch": 0.79, "learning_rate": 2.22338496082793e-06, "loss": 2.1449, "step": 32054 }, { "epoch": 0.79, "learning_rate": 2.2228830882296213e-06, "loss": 2.0855, "step": 32055 }, { "epoch": 0.79, "learning_rate": 2.222381265197411e-06, "loss": 1.9765, "step": 32056 }, { "epoch": 0.79, "learning_rate": 2.221879491734499e-06, "loss": 2.0184, "step": 32057 }, { "epoch": 0.79, "learning_rate": 2.221377767844086e-06, "loss": 1.8613, "step": 32058 }, { "epoch": 0.79, "learning_rate": 2.220876093529366e-06, "loss": 1.9117, "step": 32059 }, { "epoch": 0.79, "learning_rate": 2.220374468793537e-06, "loss": 1.8395, "step": 32060 }, { "epoch": 0.79, "learning_rate": 2.2198728936397974e-06, "loss": 2.0187, "step": 32061 }, { "epoch": 0.79, "learning_rate": 2.2193713680713458e-06, "loss": 1.8857, "step": 32062 }, { "epoch": 0.79, "learning_rate": 2.2188698920913734e-06, "loss": 2.0708, "step": 32063 }, { "epoch": 0.79, "learning_rate": 2.218368465703079e-06, "loss": 1.9525, "step": 32064 }, { "epoch": 0.79, "learning_rate": 2.217867088909661e-06, "loss": 2.0898, "step": 32065 }, { "epoch": 0.79, "learning_rate": 2.2173657617143083e-06, "loss": 1.8927, "step": 32066 }, { "epoch": 0.79, "learning_rate": 2.2168644841202213e-06, "loss": 1.8062, "step": 32067 }, { "epoch": 0.79, "learning_rate": 2.2163632561305957e-06, "loss": 1.8405, "step": 32068 }, { "epoch": 0.79, "learning_rate": 2.215862077748622e-06, "loss": 1.8184, "step": 32069 }, { "epoch": 0.79, "learning_rate": 2.2153609489774974e-06, "loss": 2.0322, "step": 32070 }, { "epoch": 0.79, "learning_rate": 2.2148598698204117e-06, "loss": 2.0346, "step": 32071 }, { "epoch": 0.79, "learning_rate": 2.2143588402805616e-06, "loss": 1.9089, "step": 32072 }, { "epoch": 0.79, "learning_rate": 2.2138578603611428e-06, "loss": 1.9087, "step": 32073 }, { "epoch": 0.79, "learning_rate": 2.2133569300653423e-06, "loss": 2.0135, "step": 32074 }, { "epoch": 0.79, "learning_rate": 2.2128560493963557e-06, "loss": 2.1377, "step": 32075 }, { "epoch": 0.79, "learning_rate": 2.212355218357378e-06, "loss": 1.968, "step": 32076 }, { "epoch": 0.79, "learning_rate": 2.2118544369515948e-06, "loss": 2.0348, "step": 32077 }, { "epoch": 0.79, "learning_rate": 2.2113537051822034e-06, "loss": 1.7892, "step": 32078 }, { "epoch": 0.79, "learning_rate": 2.2108530230523918e-06, "loss": 1.8003, "step": 32079 }, { "epoch": 0.79, "learning_rate": 2.2103523905653546e-06, "loss": 1.7829, "step": 32080 }, { "epoch": 0.79, "learning_rate": 2.209851807724278e-06, "loss": 1.9395, "step": 32081 }, { "epoch": 0.79, "learning_rate": 2.2093512745323543e-06, "loss": 2.0033, "step": 32082 }, { "epoch": 0.79, "learning_rate": 2.208850790992777e-06, "loss": 1.8939, "step": 32083 }, { "epoch": 0.79, "learning_rate": 2.2083503571087295e-06, "loss": 1.9077, "step": 32084 }, { "epoch": 0.79, "learning_rate": 2.207849972883408e-06, "loss": 2.0266, "step": 32085 }, { "epoch": 0.79, "learning_rate": 2.2073496383199945e-06, "loss": 1.8635, "step": 32086 }, { "epoch": 0.79, "learning_rate": 2.2068493534216805e-06, "loss": 2.0045, "step": 32087 }, { "epoch": 0.79, "learning_rate": 2.2063491181916595e-06, "loss": 1.9859, "step": 32088 }, { "epoch": 0.79, "learning_rate": 2.2058489326331124e-06, "loss": 2.0543, "step": 32089 }, { "epoch": 0.79, "learning_rate": 2.2053487967492327e-06, "loss": 1.8302, "step": 32090 }, { "epoch": 0.79, "learning_rate": 2.2048487105432026e-06, "loss": 1.9201, "step": 32091 }, { "epoch": 0.79, "learning_rate": 2.204348674018212e-06, "loss": 1.9329, "step": 32092 }, { "epoch": 0.79, "learning_rate": 2.203848687177448e-06, "loss": 1.9153, "step": 32093 }, { "epoch": 0.79, "learning_rate": 2.203348750024099e-06, "loss": 1.9274, "step": 32094 }, { "epoch": 0.79, "learning_rate": 2.2028488625613464e-06, "loss": 2.0284, "step": 32095 }, { "epoch": 0.79, "learning_rate": 2.2023490247923794e-06, "loss": 2.0668, "step": 32096 }, { "epoch": 0.79, "learning_rate": 2.2018492367203834e-06, "loss": 1.9229, "step": 32097 }, { "epoch": 0.79, "learning_rate": 2.2013494983485475e-06, "loss": 1.9093, "step": 32098 }, { "epoch": 0.79, "learning_rate": 2.200849809680049e-06, "loss": 1.8232, "step": 32099 }, { "epoch": 0.79, "learning_rate": 2.2003501707180785e-06, "loss": 1.8495, "step": 32100 }, { "epoch": 0.79, "learning_rate": 2.1998505814658157e-06, "loss": 1.9117, "step": 32101 }, { "epoch": 0.79, "learning_rate": 2.1993510419264476e-06, "loss": 1.8751, "step": 32102 }, { "epoch": 0.79, "learning_rate": 2.1988515521031604e-06, "loss": 1.6978, "step": 32103 }, { "epoch": 0.79, "learning_rate": 2.1983521119991313e-06, "loss": 1.8498, "step": 32104 }, { "epoch": 0.79, "learning_rate": 2.1978527216175494e-06, "loss": 1.8651, "step": 32105 }, { "epoch": 0.79, "learning_rate": 2.1973533809615922e-06, "loss": 1.9786, "step": 32106 }, { "epoch": 0.79, "learning_rate": 2.196854090034445e-06, "loss": 1.975, "step": 32107 }, { "epoch": 0.79, "learning_rate": 2.1963548488392914e-06, "loss": 1.8751, "step": 32108 }, { "epoch": 0.79, "learning_rate": 2.19585565737931e-06, "loss": 1.9678, "step": 32109 }, { "epoch": 0.79, "learning_rate": 2.1953565156576826e-06, "loss": 2.0442, "step": 32110 }, { "epoch": 0.79, "learning_rate": 2.1948574236775922e-06, "loss": 1.9989, "step": 32111 }, { "epoch": 0.79, "learning_rate": 2.194358381442222e-06, "loss": 2.0172, "step": 32112 }, { "epoch": 0.79, "learning_rate": 2.1938593889547467e-06, "loss": 1.8172, "step": 32113 }, { "epoch": 0.79, "learning_rate": 2.193360446218349e-06, "loss": 1.7997, "step": 32114 }, { "epoch": 0.79, "learning_rate": 2.192861553236213e-06, "loss": 1.8033, "step": 32115 }, { "epoch": 0.79, "learning_rate": 2.1923627100115108e-06, "loss": 1.936, "step": 32116 }, { "epoch": 0.79, "learning_rate": 2.1918639165474265e-06, "loss": 1.9518, "step": 32117 }, { "epoch": 0.79, "learning_rate": 2.19136517284714e-06, "loss": 1.9562, "step": 32118 }, { "epoch": 0.79, "learning_rate": 2.1908664789138255e-06, "loss": 1.7752, "step": 32119 }, { "epoch": 0.79, "learning_rate": 2.1903678347506675e-06, "loss": 1.9703, "step": 32120 }, { "epoch": 0.79, "learning_rate": 2.189869240360836e-06, "loss": 1.8203, "step": 32121 }, { "epoch": 0.79, "learning_rate": 2.1893706957475136e-06, "loss": 2.14, "step": 32122 }, { "epoch": 0.79, "learning_rate": 2.1888722009138807e-06, "loss": 1.9687, "step": 32123 }, { "epoch": 0.79, "learning_rate": 2.188373755863108e-06, "loss": 1.9894, "step": 32124 }, { "epoch": 0.79, "learning_rate": 2.187875360598374e-06, "loss": 1.8295, "step": 32125 }, { "epoch": 0.79, "learning_rate": 2.18737701512286e-06, "loss": 1.8568, "step": 32126 }, { "epoch": 0.79, "learning_rate": 2.1868787194397356e-06, "loss": 2.0671, "step": 32127 }, { "epoch": 0.79, "learning_rate": 2.186380473552179e-06, "loss": 1.9212, "step": 32128 }, { "epoch": 0.79, "learning_rate": 2.185882277463366e-06, "loss": 1.9681, "step": 32129 }, { "epoch": 0.79, "learning_rate": 2.1853841311764747e-06, "loss": 1.9204, "step": 32130 }, { "epoch": 0.79, "learning_rate": 2.1848860346946733e-06, "loss": 2.0464, "step": 32131 }, { "epoch": 0.79, "learning_rate": 2.1843879880211418e-06, "loss": 1.9222, "step": 32132 }, { "epoch": 0.79, "learning_rate": 2.183889991159054e-06, "loss": 1.8327, "step": 32133 }, { "epoch": 0.79, "learning_rate": 2.1833920441115797e-06, "loss": 1.8882, "step": 32134 }, { "epoch": 0.79, "learning_rate": 2.1828941468818964e-06, "loss": 2.1354, "step": 32135 }, { "epoch": 0.79, "learning_rate": 2.1823962994731783e-06, "loss": 2.0238, "step": 32136 }, { "epoch": 0.79, "learning_rate": 2.181898501888594e-06, "loss": 2.1644, "step": 32137 }, { "epoch": 0.79, "learning_rate": 2.181400754131321e-06, "loss": 1.8956, "step": 32138 }, { "epoch": 0.79, "learning_rate": 2.1809030562045252e-06, "loss": 1.8069, "step": 32139 }, { "epoch": 0.79, "learning_rate": 2.1804054081113836e-06, "loss": 2.0602, "step": 32140 }, { "epoch": 0.79, "learning_rate": 2.1799078098550687e-06, "loss": 1.8916, "step": 32141 }, { "epoch": 0.79, "learning_rate": 2.179410261438748e-06, "loss": 1.9539, "step": 32142 }, { "epoch": 0.79, "learning_rate": 2.178912762865596e-06, "loss": 2.0084, "step": 32143 }, { "epoch": 0.79, "learning_rate": 2.1784153141387796e-06, "loss": 2.1254, "step": 32144 }, { "epoch": 0.79, "learning_rate": 2.1779179152614717e-06, "loss": 2.029, "step": 32145 }, { "epoch": 0.79, "learning_rate": 2.177420566236841e-06, "loss": 1.8735, "step": 32146 }, { "epoch": 0.79, "learning_rate": 2.1769232670680617e-06, "loss": 1.9506, "step": 32147 }, { "epoch": 0.79, "learning_rate": 2.1764260177582976e-06, "loss": 1.8577, "step": 32148 }, { "epoch": 0.79, "learning_rate": 2.1759288183107196e-06, "loss": 1.9487, "step": 32149 }, { "epoch": 0.79, "learning_rate": 2.175431668728497e-06, "loss": 1.836, "step": 32150 }, { "epoch": 0.79, "learning_rate": 2.1749345690148015e-06, "loss": 2.026, "step": 32151 }, { "epoch": 0.79, "learning_rate": 2.1744375191727953e-06, "loss": 1.899, "step": 32152 }, { "epoch": 0.79, "learning_rate": 2.1739405192056517e-06, "loss": 1.9138, "step": 32153 }, { "epoch": 0.79, "learning_rate": 2.1734435691165325e-06, "loss": 1.9403, "step": 32154 }, { "epoch": 0.79, "learning_rate": 2.1729466689086087e-06, "loss": 2.0014, "step": 32155 }, { "epoch": 0.79, "learning_rate": 2.1724498185850484e-06, "loss": 1.9082, "step": 32156 }, { "epoch": 0.79, "learning_rate": 2.1719530181490144e-06, "loss": 1.8644, "step": 32157 }, { "epoch": 0.79, "learning_rate": 2.1714562676036776e-06, "loss": 1.9753, "step": 32158 }, { "epoch": 0.79, "learning_rate": 2.1709595669521977e-06, "loss": 1.9746, "step": 32159 }, { "epoch": 0.79, "learning_rate": 2.1704629161977453e-06, "loss": 1.9553, "step": 32160 }, { "epoch": 0.79, "learning_rate": 2.1699663153434857e-06, "loss": 1.9014, "step": 32161 }, { "epoch": 0.79, "learning_rate": 2.16946976439258e-06, "loss": 1.9256, "step": 32162 }, { "epoch": 0.79, "learning_rate": 2.168973263348196e-06, "loss": 2.1305, "step": 32163 }, { "epoch": 0.79, "learning_rate": 2.168476812213497e-06, "loss": 2.0244, "step": 32164 }, { "epoch": 0.79, "learning_rate": 2.1679804109916504e-06, "loss": 1.8625, "step": 32165 }, { "epoch": 0.79, "learning_rate": 2.1674840596858137e-06, "loss": 2.1699, "step": 32166 }, { "epoch": 0.79, "learning_rate": 2.166987758299155e-06, "loss": 1.951, "step": 32167 }, { "epoch": 0.79, "learning_rate": 2.166491506834838e-06, "loss": 1.9328, "step": 32168 }, { "epoch": 0.79, "learning_rate": 2.16599530529602e-06, "loss": 1.9003, "step": 32169 }, { "epoch": 0.79, "learning_rate": 2.1654991536858685e-06, "loss": 2.0774, "step": 32170 }, { "epoch": 0.79, "learning_rate": 2.165003052007545e-06, "loss": 2.0605, "step": 32171 }, { "epoch": 0.79, "learning_rate": 2.164507000264209e-06, "loss": 1.9675, "step": 32172 }, { "epoch": 0.79, "learning_rate": 2.1640109984590265e-06, "loss": 2.0422, "step": 32173 }, { "epoch": 0.79, "learning_rate": 2.1635150465951515e-06, "loss": 1.8875, "step": 32174 }, { "epoch": 0.79, "learning_rate": 2.1630191446757496e-06, "loss": 1.8749, "step": 32175 }, { "epoch": 0.79, "learning_rate": 2.162523292703983e-06, "loss": 1.7702, "step": 32176 }, { "epoch": 0.79, "learning_rate": 2.1620274906830073e-06, "loss": 1.8516, "step": 32177 }, { "epoch": 0.79, "learning_rate": 2.1615317386159875e-06, "loss": 1.9208, "step": 32178 }, { "epoch": 0.79, "learning_rate": 2.161036036506078e-06, "loss": 2.0274, "step": 32179 }, { "epoch": 0.79, "learning_rate": 2.1605403843564398e-06, "loss": 2.0404, "step": 32180 }, { "epoch": 0.79, "learning_rate": 2.160044782170233e-06, "loss": 1.8881, "step": 32181 }, { "epoch": 0.79, "learning_rate": 2.159549229950615e-06, "loss": 2.0236, "step": 32182 }, { "epoch": 0.79, "learning_rate": 2.1590537277007484e-06, "loss": 1.7067, "step": 32183 }, { "epoch": 0.79, "learning_rate": 2.1585582754237845e-06, "loss": 1.9218, "step": 32184 }, { "epoch": 0.79, "learning_rate": 2.1580628731228846e-06, "loss": 2.0242, "step": 32185 }, { "epoch": 0.79, "learning_rate": 2.1575675208012083e-06, "loss": 2.2436, "step": 32186 }, { "epoch": 0.79, "learning_rate": 2.1570722184619077e-06, "loss": 2.0835, "step": 32187 }, { "epoch": 0.79, "learning_rate": 2.15657696610814e-06, "loss": 1.9291, "step": 32188 }, { "epoch": 0.79, "learning_rate": 2.1560817637430677e-06, "loss": 2.0032, "step": 32189 }, { "epoch": 0.79, "learning_rate": 2.15558661136984e-06, "loss": 2.009, "step": 32190 }, { "epoch": 0.79, "learning_rate": 2.155091508991617e-06, "loss": 1.8667, "step": 32191 }, { "epoch": 0.79, "learning_rate": 2.1545964566115497e-06, "loss": 2.0192, "step": 32192 }, { "epoch": 0.79, "learning_rate": 2.1541014542327967e-06, "loss": 1.7892, "step": 32193 }, { "epoch": 0.79, "learning_rate": 2.153606501858514e-06, "loss": 2.0569, "step": 32194 }, { "epoch": 0.79, "learning_rate": 2.1531115994918517e-06, "loss": 2.0703, "step": 32195 }, { "epoch": 0.79, "learning_rate": 2.1526167471359693e-06, "loss": 2.0859, "step": 32196 }, { "epoch": 0.79, "learning_rate": 2.1521219447940156e-06, "loss": 1.6765, "step": 32197 }, { "epoch": 0.79, "learning_rate": 2.1516271924691455e-06, "loss": 1.8166, "step": 32198 }, { "epoch": 0.79, "learning_rate": 2.1511324901645124e-06, "loss": 1.8858, "step": 32199 }, { "epoch": 0.79, "learning_rate": 2.1506378378832717e-06, "loss": 1.8222, "step": 32200 }, { "epoch": 0.79, "learning_rate": 2.1501432356285757e-06, "loss": 1.9698, "step": 32201 }, { "epoch": 0.79, "learning_rate": 2.149648683403571e-06, "loss": 1.9876, "step": 32202 }, { "epoch": 0.79, "learning_rate": 2.149154181211416e-06, "loss": 1.9054, "step": 32203 }, { "epoch": 0.79, "learning_rate": 2.1486597290552615e-06, "loss": 1.9315, "step": 32204 }, { "epoch": 0.79, "learning_rate": 2.148165326938254e-06, "loss": 1.921, "step": 32205 }, { "epoch": 0.79, "learning_rate": 2.1476709748635515e-06, "loss": 1.9739, "step": 32206 }, { "epoch": 0.79, "learning_rate": 2.147176672834298e-06, "loss": 1.8208, "step": 32207 }, { "epoch": 0.79, "learning_rate": 2.1466824208536465e-06, "loss": 1.8461, "step": 32208 }, { "epoch": 0.79, "learning_rate": 2.1461882189247496e-06, "loss": 1.8438, "step": 32209 }, { "epoch": 0.79, "learning_rate": 2.145694067050752e-06, "loss": 1.9272, "step": 32210 }, { "epoch": 0.79, "learning_rate": 2.145199965234809e-06, "loss": 1.9759, "step": 32211 }, { "epoch": 0.79, "learning_rate": 2.144705913480063e-06, "loss": 1.8269, "step": 32212 }, { "epoch": 0.79, "learning_rate": 2.1442119117896666e-06, "loss": 1.8961, "step": 32213 }, { "epoch": 0.79, "learning_rate": 2.14371796016677e-06, "loss": 1.969, "step": 32214 }, { "epoch": 0.79, "learning_rate": 2.143224058614517e-06, "loss": 2.0551, "step": 32215 }, { "epoch": 0.79, "learning_rate": 2.142730207136058e-06, "loss": 1.9986, "step": 32216 }, { "epoch": 0.79, "learning_rate": 2.142236405734539e-06, "loss": 1.8306, "step": 32217 }, { "epoch": 0.79, "learning_rate": 2.1417426544131082e-06, "loss": 2.0913, "step": 32218 }, { "epoch": 0.79, "learning_rate": 2.141248953174916e-06, "loss": 1.8319, "step": 32219 }, { "epoch": 0.79, "learning_rate": 2.140755302023102e-06, "loss": 2.0744, "step": 32220 }, { "epoch": 0.79, "learning_rate": 2.1402617009608183e-06, "loss": 1.8839, "step": 32221 }, { "epoch": 0.79, "learning_rate": 2.139768149991206e-06, "loss": 1.9153, "step": 32222 }, { "epoch": 0.79, "learning_rate": 2.139274649117412e-06, "loss": 1.8824, "step": 32223 }, { "epoch": 0.79, "learning_rate": 2.1387811983425867e-06, "loss": 2.07, "step": 32224 }, { "epoch": 0.79, "learning_rate": 2.1382877976698667e-06, "loss": 1.8898, "step": 32225 }, { "epoch": 0.79, "learning_rate": 2.1377944471024037e-06, "loss": 1.8564, "step": 32226 }, { "epoch": 0.79, "learning_rate": 2.137301146643336e-06, "loss": 1.9368, "step": 32227 }, { "epoch": 0.79, "learning_rate": 2.1368078962958105e-06, "loss": 1.9923, "step": 32228 }, { "epoch": 0.79, "learning_rate": 2.1363146960629732e-06, "loss": 1.9375, "step": 32229 }, { "epoch": 0.79, "learning_rate": 2.1358215459479625e-06, "loss": 1.9563, "step": 32230 }, { "epoch": 0.79, "learning_rate": 2.135328445953927e-06, "loss": 1.882, "step": 32231 }, { "epoch": 0.79, "learning_rate": 2.134835396084004e-06, "loss": 1.8873, "step": 32232 }, { "epoch": 0.79, "learning_rate": 2.134342396341338e-06, "loss": 1.9731, "step": 32233 }, { "epoch": 0.79, "learning_rate": 2.1338494467290703e-06, "loss": 1.9842, "step": 32234 }, { "epoch": 0.79, "learning_rate": 2.1333565472503446e-06, "loss": 1.7116, "step": 32235 }, { "epoch": 0.79, "learning_rate": 2.1328636979083038e-06, "loss": 1.9578, "step": 32236 }, { "epoch": 0.79, "learning_rate": 2.132370898706083e-06, "loss": 1.7292, "step": 32237 }, { "epoch": 0.79, "learning_rate": 2.1318781496468275e-06, "loss": 2.1514, "step": 32238 }, { "epoch": 0.79, "learning_rate": 2.1313854507336796e-06, "loss": 1.878, "step": 32239 }, { "epoch": 0.79, "learning_rate": 2.130892801969773e-06, "loss": 2.0165, "step": 32240 }, { "epoch": 0.79, "learning_rate": 2.130400203358254e-06, "loss": 2.0646, "step": 32241 }, { "epoch": 0.79, "learning_rate": 2.129907654902256e-06, "loss": 1.803, "step": 32242 }, { "epoch": 0.79, "learning_rate": 2.1294151566049215e-06, "loss": 1.8666, "step": 32243 }, { "epoch": 0.79, "learning_rate": 2.128922708469392e-06, "loss": 2.0213, "step": 32244 }, { "epoch": 0.79, "learning_rate": 2.1284303104988015e-06, "loss": 2.0013, "step": 32245 }, { "epoch": 0.79, "learning_rate": 2.127937962696288e-06, "loss": 1.952, "step": 32246 }, { "epoch": 0.79, "learning_rate": 2.1274456650649955e-06, "loss": 1.8488, "step": 32247 }, { "epoch": 0.79, "learning_rate": 2.126953417608054e-06, "loss": 1.9245, "step": 32248 }, { "epoch": 0.79, "learning_rate": 2.126461220328607e-06, "loss": 1.909, "step": 32249 }, { "epoch": 0.79, "learning_rate": 2.1259690732297856e-06, "loss": 2.0524, "step": 32250 }, { "epoch": 0.79, "learning_rate": 2.1254769763147296e-06, "loss": 1.8875, "step": 32251 }, { "epoch": 0.79, "learning_rate": 2.1249849295865752e-06, "loss": 1.9718, "step": 32252 }, { "epoch": 0.79, "learning_rate": 2.1244929330484577e-06, "loss": 1.887, "step": 32253 }, { "epoch": 0.8, "learning_rate": 2.1240009867035162e-06, "loss": 1.915, "step": 32254 }, { "epoch": 0.8, "learning_rate": 2.1235090905548805e-06, "loss": 1.9346, "step": 32255 }, { "epoch": 0.8, "learning_rate": 2.1230172446056884e-06, "loss": 2.1657, "step": 32256 }, { "epoch": 0.8, "learning_rate": 2.1225254488590773e-06, "loss": 1.8661, "step": 32257 }, { "epoch": 0.8, "learning_rate": 2.122033703318176e-06, "loss": 1.8575, "step": 32258 }, { "epoch": 0.8, "learning_rate": 2.1215420079861238e-06, "loss": 1.9496, "step": 32259 }, { "epoch": 0.8, "learning_rate": 2.1210503628660485e-06, "loss": 1.9437, "step": 32260 }, { "epoch": 0.8, "learning_rate": 2.1205587679610873e-06, "loss": 2.0046, "step": 32261 }, { "epoch": 0.8, "learning_rate": 2.120067223274377e-06, "loss": 2.0333, "step": 32262 }, { "epoch": 0.8, "learning_rate": 2.1195757288090414e-06, "loss": 1.9362, "step": 32263 }, { "epoch": 0.8, "learning_rate": 2.1190842845682214e-06, "loss": 1.9325, "step": 32264 }, { "epoch": 0.8, "learning_rate": 2.1185928905550435e-06, "loss": 1.9361, "step": 32265 }, { "epoch": 0.8, "learning_rate": 2.1181015467726406e-06, "loss": 1.889, "step": 32266 }, { "epoch": 0.8, "learning_rate": 2.1176102532241496e-06, "loss": 1.9389, "step": 32267 }, { "epoch": 0.8, "learning_rate": 2.1171190099126927e-06, "loss": 1.7553, "step": 32268 }, { "epoch": 0.8, "learning_rate": 2.1166278168414056e-06, "loss": 1.92, "step": 32269 }, { "epoch": 0.8, "learning_rate": 2.1161366740134193e-06, "loss": 1.9754, "step": 32270 }, { "epoch": 0.8, "learning_rate": 2.115645581431862e-06, "loss": 2.0337, "step": 32271 }, { "epoch": 0.8, "learning_rate": 2.1151545390998695e-06, "loss": 1.8734, "step": 32272 }, { "epoch": 0.8, "learning_rate": 2.1146635470205624e-06, "loss": 1.8222, "step": 32273 }, { "epoch": 0.8, "learning_rate": 2.114172605197078e-06, "loss": 1.8465, "step": 32274 }, { "epoch": 0.8, "learning_rate": 2.113681713632538e-06, "loss": 1.9008, "step": 32275 }, { "epoch": 0.8, "learning_rate": 2.113190872330074e-06, "loss": 1.9478, "step": 32276 }, { "epoch": 0.8, "learning_rate": 2.1127000812928187e-06, "loss": 1.9111, "step": 32277 }, { "epoch": 0.8, "learning_rate": 2.112209340523893e-06, "loss": 1.9333, "step": 32278 }, { "epoch": 0.8, "learning_rate": 2.111718650026431e-06, "loss": 1.8629, "step": 32279 }, { "epoch": 0.8, "learning_rate": 2.111228009803553e-06, "loss": 1.9443, "step": 32280 }, { "epoch": 0.8, "learning_rate": 2.1107374198583907e-06, "loss": 1.9301, "step": 32281 }, { "epoch": 0.8, "learning_rate": 2.110246880194072e-06, "loss": 2.0297, "step": 32282 }, { "epoch": 0.8, "learning_rate": 2.109756390813719e-06, "loss": 1.9299, "step": 32283 }, { "epoch": 0.8, "learning_rate": 2.1092659517204626e-06, "loss": 1.9493, "step": 32284 }, { "epoch": 0.8, "learning_rate": 2.1087755629174223e-06, "loss": 1.9558, "step": 32285 }, { "epoch": 0.8, "learning_rate": 2.1082852244077278e-06, "loss": 1.8287, "step": 32286 }, { "epoch": 0.8, "learning_rate": 2.1077949361945027e-06, "loss": 1.9426, "step": 32287 }, { "epoch": 0.8, "learning_rate": 2.107304698280873e-06, "loss": 1.8459, "step": 32288 }, { "epoch": 0.8, "learning_rate": 2.1068145106699643e-06, "loss": 1.9851, "step": 32289 }, { "epoch": 0.8, "learning_rate": 2.1063243733648974e-06, "loss": 1.9163, "step": 32290 }, { "epoch": 0.8, "learning_rate": 2.1058342863687976e-06, "loss": 1.9115, "step": 32291 }, { "epoch": 0.8, "learning_rate": 2.105344249684791e-06, "loss": 1.7951, "step": 32292 }, { "epoch": 0.8, "learning_rate": 2.1048542633159964e-06, "loss": 1.8962, "step": 32293 }, { "epoch": 0.8, "learning_rate": 2.10436432726554e-06, "loss": 1.9567, "step": 32294 }, { "epoch": 0.8, "learning_rate": 2.103874441536541e-06, "loss": 2.0231, "step": 32295 }, { "epoch": 0.8, "learning_rate": 2.1033846061321226e-06, "loss": 2.0127, "step": 32296 }, { "epoch": 0.8, "learning_rate": 2.102894821055411e-06, "loss": 1.8376, "step": 32297 }, { "epoch": 0.8, "learning_rate": 2.1024050863095215e-06, "loss": 2.0265, "step": 32298 }, { "epoch": 0.8, "learning_rate": 2.1019154018975805e-06, "loss": 1.947, "step": 32299 }, { "epoch": 0.8, "learning_rate": 2.1014257678227048e-06, "loss": 2.0469, "step": 32300 }, { "epoch": 0.8, "learning_rate": 2.100936184088016e-06, "loss": 1.9338, "step": 32301 }, { "epoch": 0.8, "learning_rate": 2.1004466506966383e-06, "loss": 1.8376, "step": 32302 }, { "epoch": 0.8, "learning_rate": 2.099957167651685e-06, "loss": 2.0169, "step": 32303 }, { "epoch": 0.8, "learning_rate": 2.09946773495628e-06, "loss": 1.9155, "step": 32304 }, { "epoch": 0.8, "learning_rate": 2.098978352613541e-06, "loss": 2.0703, "step": 32305 }, { "epoch": 0.8, "learning_rate": 2.0984890206265883e-06, "loss": 1.8917, "step": 32306 }, { "epoch": 0.8, "learning_rate": 2.0979997389985428e-06, "loss": 2.1783, "step": 32307 }, { "epoch": 0.8, "learning_rate": 2.0975105077325174e-06, "loss": 1.8971, "step": 32308 }, { "epoch": 0.8, "learning_rate": 2.097021326831633e-06, "loss": 1.9669, "step": 32309 }, { "epoch": 0.8, "learning_rate": 2.0965321962990095e-06, "loss": 1.9244, "step": 32310 }, { "epoch": 0.8, "learning_rate": 2.0960431161377594e-06, "loss": 2.1047, "step": 32311 }, { "epoch": 0.8, "learning_rate": 2.095554086351005e-06, "loss": 1.7733, "step": 32312 }, { "epoch": 0.8, "learning_rate": 2.095065106941858e-06, "loss": 2.071, "step": 32313 }, { "epoch": 0.8, "learning_rate": 2.0945761779134366e-06, "loss": 2.0639, "step": 32314 }, { "epoch": 0.8, "learning_rate": 2.094087299268861e-06, "loss": 2.0098, "step": 32315 }, { "epoch": 0.8, "learning_rate": 2.093598471011241e-06, "loss": 1.9583, "step": 32316 }, { "epoch": 0.8, "learning_rate": 2.093109693143697e-06, "loss": 1.8473, "step": 32317 }, { "epoch": 0.8, "learning_rate": 2.092620965669339e-06, "loss": 1.9498, "step": 32318 }, { "epoch": 0.8, "learning_rate": 2.0921322885912833e-06, "loss": 1.9716, "step": 32319 }, { "epoch": 0.8, "learning_rate": 2.09164366191265e-06, "loss": 1.854, "step": 32320 }, { "epoch": 0.8, "learning_rate": 2.091155085636545e-06, "loss": 1.7512, "step": 32321 }, { "epoch": 0.8, "learning_rate": 2.0906665597660868e-06, "loss": 1.8804, "step": 32322 }, { "epoch": 0.8, "learning_rate": 2.0901780843043885e-06, "loss": 1.82, "step": 32323 }, { "epoch": 0.8, "learning_rate": 2.089689659254562e-06, "loss": 1.8545, "step": 32324 }, { "epoch": 0.8, "learning_rate": 2.089201284619724e-06, "loss": 2.0554, "step": 32325 }, { "epoch": 0.8, "learning_rate": 2.0887129604029823e-06, "loss": 1.9202, "step": 32326 }, { "epoch": 0.8, "learning_rate": 2.088224686607454e-06, "loss": 1.8952, "step": 32327 }, { "epoch": 0.8, "learning_rate": 2.0877364632362453e-06, "loss": 1.7551, "step": 32328 }, { "epoch": 0.8, "learning_rate": 2.0872482902924696e-06, "loss": 1.962, "step": 32329 }, { "epoch": 0.8, "learning_rate": 2.0867601677792425e-06, "loss": 1.9369, "step": 32330 }, { "epoch": 0.8, "learning_rate": 2.0862720956996695e-06, "loss": 1.8937, "step": 32331 }, { "epoch": 0.8, "learning_rate": 2.0857840740568656e-06, "loss": 1.9155, "step": 32332 }, { "epoch": 0.8, "learning_rate": 2.0852961028539365e-06, "loss": 1.8952, "step": 32333 }, { "epoch": 0.8, "learning_rate": 2.0848081820939946e-06, "loss": 2.18, "step": 32334 }, { "epoch": 0.8, "learning_rate": 2.0843203117801527e-06, "loss": 1.9999, "step": 32335 }, { "epoch": 0.8, "learning_rate": 2.083832491915514e-06, "loss": 2.0001, "step": 32336 }, { "epoch": 0.8, "learning_rate": 2.0833447225031933e-06, "loss": 2.0213, "step": 32337 }, { "epoch": 0.8, "learning_rate": 2.0828570035462935e-06, "loss": 2.0171, "step": 32338 }, { "epoch": 0.8, "learning_rate": 2.082369335047927e-06, "loss": 2.0135, "step": 32339 }, { "epoch": 0.8, "learning_rate": 2.0818817170112003e-06, "loss": 1.9416, "step": 32340 }, { "epoch": 0.8, "learning_rate": 2.081394149439222e-06, "loss": 1.8994, "step": 32341 }, { "epoch": 0.8, "learning_rate": 2.0809066323351025e-06, "loss": 1.8501, "step": 32342 }, { "epoch": 0.8, "learning_rate": 2.0804191657019434e-06, "loss": 1.8601, "step": 32343 }, { "epoch": 0.8, "learning_rate": 2.079931749542853e-06, "loss": 1.89, "step": 32344 }, { "epoch": 0.8, "learning_rate": 2.079444383860942e-06, "loss": 1.7743, "step": 32345 }, { "epoch": 0.8, "learning_rate": 2.0789570686593097e-06, "loss": 2.015, "step": 32346 }, { "epoch": 0.8, "learning_rate": 2.0784698039410688e-06, "loss": 2.0183, "step": 32347 }, { "epoch": 0.8, "learning_rate": 2.0779825897093186e-06, "loss": 1.926, "step": 32348 }, { "epoch": 0.8, "learning_rate": 2.0774954259671666e-06, "loss": 1.9255, "step": 32349 }, { "epoch": 0.8, "learning_rate": 2.077008312717721e-06, "loss": 1.6309, "step": 32350 }, { "epoch": 0.8, "learning_rate": 2.0765212499640817e-06, "loss": 1.8267, "step": 32351 }, { "epoch": 0.8, "learning_rate": 2.0760342377093567e-06, "loss": 1.9904, "step": 32352 }, { "epoch": 0.8, "learning_rate": 2.075547275956644e-06, "loss": 1.9127, "step": 32353 }, { "epoch": 0.8, "learning_rate": 2.0750603647090526e-06, "loss": 2.0384, "step": 32354 }, { "epoch": 0.8, "learning_rate": 2.074573503969687e-06, "loss": 1.9224, "step": 32355 }, { "epoch": 0.8, "learning_rate": 2.074086693741644e-06, "loss": 1.9509, "step": 32356 }, { "epoch": 0.8, "learning_rate": 2.0735999340280286e-06, "loss": 1.9431, "step": 32357 }, { "epoch": 0.8, "learning_rate": 2.0731132248319454e-06, "loss": 1.9862, "step": 32358 }, { "epoch": 0.8, "learning_rate": 2.072626566156494e-06, "loss": 1.9618, "step": 32359 }, { "epoch": 0.8, "learning_rate": 2.0721399580047797e-06, "loss": 2.0285, "step": 32360 }, { "epoch": 0.8, "learning_rate": 2.0716534003798985e-06, "loss": 1.9406, "step": 32361 }, { "epoch": 0.8, "learning_rate": 2.071166893284954e-06, "loss": 1.9409, "step": 32362 }, { "epoch": 0.8, "learning_rate": 2.0706804367230495e-06, "loss": 1.7608, "step": 32363 }, { "epoch": 0.8, "learning_rate": 2.07019403069728e-06, "loss": 1.9256, "step": 32364 }, { "epoch": 0.8, "learning_rate": 2.0697076752107505e-06, "loss": 2.1659, "step": 32365 }, { "epoch": 0.8, "learning_rate": 2.0692213702665564e-06, "loss": 1.9226, "step": 32366 }, { "epoch": 0.8, "learning_rate": 2.068735115867798e-06, "loss": 1.9733, "step": 32367 }, { "epoch": 0.8, "learning_rate": 2.0682489120175797e-06, "loss": 1.8583, "step": 32368 }, { "epoch": 0.8, "learning_rate": 2.0677627587189918e-06, "loss": 1.7578, "step": 32369 }, { "epoch": 0.8, "learning_rate": 2.067276655975141e-06, "loss": 1.9239, "step": 32370 }, { "epoch": 0.8, "learning_rate": 2.0667906037891172e-06, "loss": 1.7271, "step": 32371 }, { "epoch": 0.8, "learning_rate": 2.0663046021640233e-06, "loss": 2.1278, "step": 32372 }, { "epoch": 0.8, "learning_rate": 2.0658186511029575e-06, "loss": 1.8673, "step": 32373 }, { "epoch": 0.8, "learning_rate": 2.065332750609013e-06, "loss": 1.7082, "step": 32374 }, { "epoch": 0.8, "learning_rate": 2.064846900685289e-06, "loss": 1.8483, "step": 32375 }, { "epoch": 0.8, "learning_rate": 2.0643611013348807e-06, "loss": 1.8099, "step": 32376 }, { "epoch": 0.8, "learning_rate": 2.0638753525608855e-06, "loss": 1.8537, "step": 32377 }, { "epoch": 0.8, "learning_rate": 2.0633896543664023e-06, "loss": 1.892, "step": 32378 }, { "epoch": 0.8, "learning_rate": 2.06290400675452e-06, "loss": 1.9643, "step": 32379 }, { "epoch": 0.8, "learning_rate": 2.0624184097283405e-06, "loss": 1.9223, "step": 32380 }, { "epoch": 0.8, "learning_rate": 2.061932863290952e-06, "loss": 2.0096, "step": 32381 }, { "epoch": 0.8, "learning_rate": 2.0614473674454528e-06, "loss": 1.9331, "step": 32382 }, { "epoch": 0.8, "learning_rate": 2.0609619221949383e-06, "loss": 2.0312, "step": 32383 }, { "epoch": 0.8, "learning_rate": 2.060476527542499e-06, "loss": 1.965, "step": 32384 }, { "epoch": 0.8, "learning_rate": 2.0599911834912334e-06, "loss": 1.7537, "step": 32385 }, { "epoch": 0.8, "learning_rate": 2.0595058900442277e-06, "loss": 1.9057, "step": 32386 }, { "epoch": 0.8, "learning_rate": 2.0590206472045794e-06, "loss": 2.1958, "step": 32387 }, { "epoch": 0.8, "learning_rate": 2.058535454975382e-06, "loss": 1.9793, "step": 32388 }, { "epoch": 0.8, "learning_rate": 2.058050313359724e-06, "loss": 2.0425, "step": 32389 }, { "epoch": 0.8, "learning_rate": 2.0575652223607033e-06, "loss": 1.9929, "step": 32390 }, { "epoch": 0.8, "learning_rate": 2.0570801819814047e-06, "loss": 2.0736, "step": 32391 }, { "epoch": 0.8, "learning_rate": 2.056595192224922e-06, "loss": 2.0683, "step": 32392 }, { "epoch": 0.8, "learning_rate": 2.0561102530943467e-06, "loss": 1.9149, "step": 32393 }, { "epoch": 0.8, "learning_rate": 2.0556253645927706e-06, "loss": 1.9936, "step": 32394 }, { "epoch": 0.8, "learning_rate": 2.055140526723285e-06, "loss": 1.9054, "step": 32395 }, { "epoch": 0.8, "learning_rate": 2.0546557394889744e-06, "loss": 1.9804, "step": 32396 }, { "epoch": 0.8, "learning_rate": 2.054171002892934e-06, "loss": 2.0158, "step": 32397 }, { "epoch": 0.8, "learning_rate": 2.0536863169382527e-06, "loss": 1.8879, "step": 32398 }, { "epoch": 0.8, "learning_rate": 2.0532016816280154e-06, "loss": 2.1133, "step": 32399 }, { "epoch": 0.8, "learning_rate": 2.0527170969653165e-06, "loss": 1.8852, "step": 32400 }, { "epoch": 0.8, "learning_rate": 2.052232562953239e-06, "loss": 2.1249, "step": 32401 }, { "epoch": 0.8, "learning_rate": 2.051748079594874e-06, "loss": 1.9832, "step": 32402 }, { "epoch": 0.8, "learning_rate": 2.05126364689331e-06, "loss": 1.9028, "step": 32403 }, { "epoch": 0.8, "learning_rate": 2.0507792648516324e-06, "loss": 2.0709, "step": 32404 }, { "epoch": 0.8, "learning_rate": 2.0502949334729305e-06, "loss": 1.919, "step": 32405 }, { "epoch": 0.8, "learning_rate": 2.049810652760287e-06, "loss": 2.002, "step": 32406 }, { "epoch": 0.8, "learning_rate": 2.049326422716792e-06, "loss": 1.9559, "step": 32407 }, { "epoch": 0.8, "learning_rate": 2.0488422433455325e-06, "loss": 2.0099, "step": 32408 }, { "epoch": 0.8, "learning_rate": 2.04835811464959e-06, "loss": 1.9846, "step": 32409 }, { "epoch": 0.8, "learning_rate": 2.0478740366320528e-06, "loss": 1.8971, "step": 32410 }, { "epoch": 0.8, "learning_rate": 2.0473900092960064e-06, "loss": 1.9673, "step": 32411 }, { "epoch": 0.8, "learning_rate": 2.0469060326445346e-06, "loss": 2.0035, "step": 32412 }, { "epoch": 0.8, "learning_rate": 2.0464221066807256e-06, "loss": 2.0096, "step": 32413 }, { "epoch": 0.8, "learning_rate": 2.045938231407657e-06, "loss": 1.9692, "step": 32414 }, { "epoch": 0.8, "learning_rate": 2.04545440682842e-06, "loss": 1.762, "step": 32415 }, { "epoch": 0.8, "learning_rate": 2.0449706329460906e-06, "loss": 2.0601, "step": 32416 }, { "epoch": 0.8, "learning_rate": 2.0444869097637564e-06, "loss": 1.9308, "step": 32417 }, { "epoch": 0.8, "learning_rate": 2.044003237284502e-06, "loss": 1.8821, "step": 32418 }, { "epoch": 0.8, "learning_rate": 2.0435196155114056e-06, "loss": 1.821, "step": 32419 }, { "epoch": 0.8, "learning_rate": 2.043036044447554e-06, "loss": 1.9548, "step": 32420 }, { "epoch": 0.8, "learning_rate": 2.042552524096024e-06, "loss": 1.9554, "step": 32421 }, { "epoch": 0.8, "learning_rate": 2.042069054459901e-06, "loss": 1.9744, "step": 32422 }, { "epoch": 0.8, "learning_rate": 2.041585635542267e-06, "loss": 2.1664, "step": 32423 }, { "epoch": 0.8, "learning_rate": 2.041102267346199e-06, "loss": 1.8697, "step": 32424 }, { "epoch": 0.8, "learning_rate": 2.0406189498747796e-06, "loss": 2.0501, "step": 32425 }, { "epoch": 0.8, "learning_rate": 2.0401356831310924e-06, "loss": 2.0274, "step": 32426 }, { "epoch": 0.8, "learning_rate": 2.0396524671182117e-06, "loss": 1.8898, "step": 32427 }, { "epoch": 0.8, "learning_rate": 2.039169301839221e-06, "loss": 1.9147, "step": 32428 }, { "epoch": 0.8, "learning_rate": 2.0386861872971976e-06, "loss": 2.0548, "step": 32429 }, { "epoch": 0.8, "learning_rate": 2.0382031234952214e-06, "loss": 1.9338, "step": 32430 }, { "epoch": 0.8, "learning_rate": 2.037720110436375e-06, "loss": 1.9494, "step": 32431 }, { "epoch": 0.8, "learning_rate": 2.0372371481237306e-06, "loss": 1.9427, "step": 32432 }, { "epoch": 0.8, "learning_rate": 2.036754236560371e-06, "loss": 1.9811, "step": 32433 }, { "epoch": 0.8, "learning_rate": 2.036271375749369e-06, "loss": 1.9998, "step": 32434 }, { "epoch": 0.8, "learning_rate": 2.0357885656938047e-06, "loss": 1.9985, "step": 32435 }, { "epoch": 0.8, "learning_rate": 2.035305806396759e-06, "loss": 1.9144, "step": 32436 }, { "epoch": 0.8, "learning_rate": 2.0348230978613014e-06, "loss": 2.0133, "step": 32437 }, { "epoch": 0.8, "learning_rate": 2.0343404400905155e-06, "loss": 1.9654, "step": 32438 }, { "epoch": 0.8, "learning_rate": 2.0338578330874705e-06, "loss": 2.0892, "step": 32439 }, { "epoch": 0.8, "learning_rate": 2.033375276855245e-06, "loss": 1.7605, "step": 32440 }, { "epoch": 0.8, "learning_rate": 2.0328927713969194e-06, "loss": 1.9487, "step": 32441 }, { "epoch": 0.8, "learning_rate": 2.0324103167155607e-06, "loss": 1.9447, "step": 32442 }, { "epoch": 0.8, "learning_rate": 2.031927912814251e-06, "loss": 1.6961, "step": 32443 }, { "epoch": 0.8, "learning_rate": 2.031445559696058e-06, "loss": 1.9742, "step": 32444 }, { "epoch": 0.8, "learning_rate": 2.030963257364059e-06, "loss": 1.9081, "step": 32445 }, { "epoch": 0.8, "learning_rate": 2.0304810058213296e-06, "loss": 1.8554, "step": 32446 }, { "epoch": 0.8, "learning_rate": 2.029998805070942e-06, "loss": 2.0253, "step": 32447 }, { "epoch": 0.8, "learning_rate": 2.0295166551159704e-06, "loss": 1.9272, "step": 32448 }, { "epoch": 0.8, "learning_rate": 2.0290345559594847e-06, "loss": 2.046, "step": 32449 }, { "epoch": 0.8, "learning_rate": 2.0285525076045597e-06, "loss": 1.8997, "step": 32450 }, { "epoch": 0.8, "learning_rate": 2.0280705100542707e-06, "loss": 1.9657, "step": 32451 }, { "epoch": 0.8, "learning_rate": 2.0275885633116823e-06, "loss": 1.902, "step": 32452 }, { "epoch": 0.8, "learning_rate": 2.0271066673798733e-06, "loss": 1.854, "step": 32453 }, { "epoch": 0.8, "learning_rate": 2.02662482226191e-06, "loss": 1.9352, "step": 32454 }, { "epoch": 0.8, "learning_rate": 2.026143027960864e-06, "loss": 1.9609, "step": 32455 }, { "epoch": 0.8, "learning_rate": 2.025661284479811e-06, "loss": 1.915, "step": 32456 }, { "epoch": 0.8, "eval_loss": 1.669873833656311, "eval_runtime": 94.7697, "eval_samples_per_second": 620.948, "eval_steps_per_second": 4.854, "step": 32456 }, { "epoch": 0.8, "learning_rate": 2.0251795918218133e-06, "loss": 1.9321, "step": 32457 }, { "epoch": 0.8, "learning_rate": 2.0246979499899477e-06, "loss": 1.9414, "step": 32458 }, { "epoch": 0.8, "learning_rate": 2.02421635898728e-06, "loss": 2.0768, "step": 32459 }, { "epoch": 0.8, "learning_rate": 2.023734818816879e-06, "loss": 1.7965, "step": 32460 }, { "epoch": 0.8, "learning_rate": 2.023253329481819e-06, "loss": 1.9708, "step": 32461 }, { "epoch": 0.8, "learning_rate": 2.0227718909851622e-06, "loss": 1.8972, "step": 32462 }, { "epoch": 0.8, "learning_rate": 2.02229050332998e-06, "loss": 1.8388, "step": 32463 }, { "epoch": 0.8, "learning_rate": 2.0218091665193395e-06, "loss": 1.8555, "step": 32464 }, { "epoch": 0.8, "learning_rate": 2.0213278805563096e-06, "loss": 1.9044, "step": 32465 }, { "epoch": 0.8, "learning_rate": 2.02084664544396e-06, "loss": 1.9784, "step": 32466 }, { "epoch": 0.8, "learning_rate": 2.020365461185353e-06, "loss": 1.9929, "step": 32467 }, { "epoch": 0.8, "learning_rate": 2.01988432778356e-06, "loss": 2.033, "step": 32468 }, { "epoch": 0.8, "learning_rate": 2.019403245241641e-06, "loss": 1.8033, "step": 32469 }, { "epoch": 0.8, "learning_rate": 2.0189222135626673e-06, "loss": 1.8203, "step": 32470 }, { "epoch": 0.8, "learning_rate": 2.018441232749705e-06, "loss": 1.7792, "step": 32471 }, { "epoch": 0.8, "learning_rate": 2.017960302805816e-06, "loss": 1.98, "step": 32472 }, { "epoch": 0.8, "learning_rate": 2.017479423734069e-06, "loss": 2.0324, "step": 32473 }, { "epoch": 0.8, "learning_rate": 2.016998595537525e-06, "loss": 2.0677, "step": 32474 }, { "epoch": 0.8, "learning_rate": 2.0165178182192504e-06, "loss": 1.9594, "step": 32475 }, { "epoch": 0.8, "learning_rate": 2.0160370917823122e-06, "loss": 2.0746, "step": 32476 }, { "epoch": 0.8, "learning_rate": 2.0155564162297692e-06, "loss": 2.0019, "step": 32477 }, { "epoch": 0.8, "learning_rate": 2.015075791564689e-06, "loss": 1.9248, "step": 32478 }, { "epoch": 0.8, "learning_rate": 2.0145952177901306e-06, "loss": 1.8308, "step": 32479 }, { "epoch": 0.8, "learning_rate": 2.0141146949091595e-06, "loss": 1.9334, "step": 32480 }, { "epoch": 0.8, "learning_rate": 2.0136342229248384e-06, "loss": 1.8198, "step": 32481 }, { "epoch": 0.8, "learning_rate": 2.0131538018402284e-06, "loss": 1.8017, "step": 32482 }, { "epoch": 0.8, "learning_rate": 2.0126734316583916e-06, "loss": 1.9548, "step": 32483 }, { "epoch": 0.8, "learning_rate": 2.0121931123823934e-06, "loss": 1.9897, "step": 32484 }, { "epoch": 0.8, "learning_rate": 2.0117128440152887e-06, "loss": 1.9749, "step": 32485 }, { "epoch": 0.8, "learning_rate": 2.0112326265601445e-06, "loss": 1.8703, "step": 32486 }, { "epoch": 0.8, "learning_rate": 2.0107524600200156e-06, "loss": 1.8876, "step": 32487 }, { "epoch": 0.8, "learning_rate": 2.0102723443979646e-06, "loss": 1.9143, "step": 32488 }, { "epoch": 0.8, "learning_rate": 2.009792279697055e-06, "loss": 1.8919, "step": 32489 }, { "epoch": 0.8, "learning_rate": 2.009312265920339e-06, "loss": 1.9704, "step": 32490 }, { "epoch": 0.8, "learning_rate": 2.0088323030708832e-06, "loss": 1.8968, "step": 32491 }, { "epoch": 0.8, "learning_rate": 2.008352391151741e-06, "loss": 1.8769, "step": 32492 }, { "epoch": 0.8, "learning_rate": 2.007872530165974e-06, "loss": 1.9207, "step": 32493 }, { "epoch": 0.8, "learning_rate": 2.0073927201166423e-06, "loss": 1.9096, "step": 32494 }, { "epoch": 0.8, "learning_rate": 2.0069129610067984e-06, "loss": 1.9649, "step": 32495 }, { "epoch": 0.8, "learning_rate": 2.0064332528395057e-06, "loss": 1.9511, "step": 32496 }, { "epoch": 0.8, "learning_rate": 2.0059535956178168e-06, "loss": 1.7794, "step": 32497 }, { "epoch": 0.8, "learning_rate": 2.0054739893447904e-06, "loss": 1.9155, "step": 32498 }, { "epoch": 0.8, "learning_rate": 2.0049944340234838e-06, "loss": 1.8529, "step": 32499 }, { "epoch": 0.8, "learning_rate": 2.004514929656953e-06, "loss": 1.7264, "step": 32500 }, { "epoch": 0.8, "learning_rate": 2.004035476248257e-06, "loss": 1.7631, "step": 32501 }, { "epoch": 0.8, "learning_rate": 2.0035560738004457e-06, "loss": 2.115, "step": 32502 }, { "epoch": 0.8, "learning_rate": 2.003076722316578e-06, "loss": 1.835, "step": 32503 }, { "epoch": 0.8, "learning_rate": 2.0025974217997113e-06, "loss": 1.9084, "step": 32504 }, { "epoch": 0.8, "learning_rate": 2.0021181722528937e-06, "loss": 1.763, "step": 32505 }, { "epoch": 0.8, "learning_rate": 2.0016389736791873e-06, "loss": 2.048, "step": 32506 }, { "epoch": 0.8, "learning_rate": 2.001159826081639e-06, "loss": 2.0311, "step": 32507 }, { "epoch": 0.8, "learning_rate": 2.000680729463306e-06, "loss": 1.9958, "step": 32508 }, { "epoch": 0.8, "learning_rate": 2.0002016838272443e-06, "loss": 1.8202, "step": 32509 }, { "epoch": 0.8, "learning_rate": 1.999722689176502e-06, "loss": 2.0109, "step": 32510 }, { "epoch": 0.8, "learning_rate": 1.9992437455141368e-06, "loss": 1.9827, "step": 32511 }, { "epoch": 0.8, "learning_rate": 1.9987648528431957e-06, "loss": 1.7761, "step": 32512 }, { "epoch": 0.8, "learning_rate": 1.9982860111667335e-06, "loss": 1.8218, "step": 32513 }, { "epoch": 0.8, "learning_rate": 1.997807220487805e-06, "loss": 1.876, "step": 32514 }, { "epoch": 0.8, "learning_rate": 1.9973284808094563e-06, "loss": 1.841, "step": 32515 }, { "epoch": 0.8, "learning_rate": 1.9968497921347416e-06, "loss": 1.9953, "step": 32516 }, { "epoch": 0.8, "learning_rate": 1.9963711544667107e-06, "loss": 1.8845, "step": 32517 }, { "epoch": 0.8, "learning_rate": 1.995892567808415e-06, "loss": 1.9004, "step": 32518 }, { "epoch": 0.8, "learning_rate": 1.995414032162908e-06, "loss": 1.9848, "step": 32519 }, { "epoch": 0.8, "learning_rate": 1.9949355475332323e-06, "loss": 2.0218, "step": 32520 }, { "epoch": 0.8, "learning_rate": 1.9944571139224433e-06, "loss": 1.8416, "step": 32521 }, { "epoch": 0.8, "learning_rate": 1.993978731333587e-06, "loss": 1.897, "step": 32522 }, { "epoch": 0.8, "learning_rate": 1.9935003997697123e-06, "loss": 2.0188, "step": 32523 }, { "epoch": 0.8, "learning_rate": 1.993022119233872e-06, "loss": 1.8269, "step": 32524 }, { "epoch": 0.8, "learning_rate": 1.992543889729107e-06, "loss": 1.9013, "step": 32525 }, { "epoch": 0.8, "learning_rate": 1.9920657112584727e-06, "loss": 1.9392, "step": 32526 }, { "epoch": 0.8, "learning_rate": 1.991587583825012e-06, "loss": 1.8359, "step": 32527 }, { "epoch": 0.8, "learning_rate": 1.9911095074317722e-06, "loss": 2.0799, "step": 32528 }, { "epoch": 0.8, "learning_rate": 1.990631482081804e-06, "loss": 2.0682, "step": 32529 }, { "epoch": 0.8, "learning_rate": 1.9901535077781497e-06, "loss": 1.9097, "step": 32530 }, { "epoch": 0.8, "learning_rate": 1.989675584523857e-06, "loss": 1.8926, "step": 32531 }, { "epoch": 0.8, "learning_rate": 1.989197712321975e-06, "loss": 1.8446, "step": 32532 }, { "epoch": 0.8, "learning_rate": 1.9887198911755446e-06, "loss": 1.924, "step": 32533 }, { "epoch": 0.8, "learning_rate": 1.9882421210876125e-06, "loss": 1.8351, "step": 32534 }, { "epoch": 0.8, "learning_rate": 1.987764402061224e-06, "loss": 1.988, "step": 32535 }, { "epoch": 0.8, "learning_rate": 1.9872867340994272e-06, "loss": 1.9222, "step": 32536 }, { "epoch": 0.8, "learning_rate": 1.9868091172052608e-06, "loss": 2.0269, "step": 32537 }, { "epoch": 0.8, "learning_rate": 1.9863315513817706e-06, "loss": 1.8381, "step": 32538 }, { "epoch": 0.8, "learning_rate": 1.9858540366320044e-06, "loss": 2.1131, "step": 32539 }, { "epoch": 0.8, "learning_rate": 1.985376572958999e-06, "loss": 1.9283, "step": 32540 }, { "epoch": 0.8, "learning_rate": 1.9848991603658007e-06, "loss": 2.1008, "step": 32541 }, { "epoch": 0.8, "learning_rate": 1.9844217988554547e-06, "loss": 1.8392, "step": 32542 }, { "epoch": 0.8, "learning_rate": 1.983944488430999e-06, "loss": 1.7696, "step": 32543 }, { "epoch": 0.8, "learning_rate": 1.9834672290954794e-06, "loss": 1.9624, "step": 32544 }, { "epoch": 0.8, "learning_rate": 1.9829900208519327e-06, "loss": 1.8795, "step": 32545 }, { "epoch": 0.8, "learning_rate": 1.982512863703404e-06, "loss": 1.8854, "step": 32546 }, { "epoch": 0.8, "learning_rate": 1.982035757652936e-06, "loss": 1.9185, "step": 32547 }, { "epoch": 0.8, "learning_rate": 1.981558702703563e-06, "loss": 2.1241, "step": 32548 }, { "epoch": 0.8, "learning_rate": 1.9810816988583335e-06, "loss": 1.8905, "step": 32549 }, { "epoch": 0.8, "learning_rate": 1.9806047461202814e-06, "loss": 1.9741, "step": 32550 }, { "epoch": 0.8, "learning_rate": 1.980127844492449e-06, "loss": 1.8581, "step": 32551 }, { "epoch": 0.8, "learning_rate": 1.979650993977874e-06, "loss": 1.9426, "step": 32552 }, { "epoch": 0.8, "learning_rate": 1.979174194579598e-06, "loss": 1.8685, "step": 32553 }, { "epoch": 0.8, "learning_rate": 1.9786974463006616e-06, "loss": 2.0236, "step": 32554 }, { "epoch": 0.8, "learning_rate": 1.978220749144096e-06, "loss": 1.9357, "step": 32555 }, { "epoch": 0.8, "learning_rate": 1.9777441031129465e-06, "loss": 1.8192, "step": 32556 }, { "epoch": 0.8, "learning_rate": 1.977267508210249e-06, "loss": 1.8549, "step": 32557 }, { "epoch": 0.8, "learning_rate": 1.976790964439038e-06, "loss": 1.9753, "step": 32558 }, { "epoch": 0.8, "learning_rate": 1.976314471802355e-06, "loss": 1.9957, "step": 32559 }, { "epoch": 0.8, "learning_rate": 1.975838030303233e-06, "loss": 1.9242, "step": 32560 }, { "epoch": 0.8, "learning_rate": 1.9753616399447094e-06, "loss": 2.0219, "step": 32561 }, { "epoch": 0.8, "learning_rate": 1.974885300729824e-06, "loss": 1.9032, "step": 32562 }, { "epoch": 0.8, "learning_rate": 1.974409012661608e-06, "loss": 1.8843, "step": 32563 }, { "epoch": 0.8, "learning_rate": 1.9739327757431006e-06, "loss": 1.7924, "step": 32564 }, { "epoch": 0.8, "learning_rate": 1.9734565899773327e-06, "loss": 2.0226, "step": 32565 }, { "epoch": 0.8, "learning_rate": 1.9729804553673425e-06, "loss": 1.9104, "step": 32566 }, { "epoch": 0.8, "learning_rate": 1.972504371916166e-06, "loss": 1.9566, "step": 32567 }, { "epoch": 0.8, "learning_rate": 1.9720283396268327e-06, "loss": 1.968, "step": 32568 }, { "epoch": 0.8, "learning_rate": 1.9715523585023787e-06, "loss": 1.8987, "step": 32569 }, { "epoch": 0.8, "learning_rate": 1.9710764285458383e-06, "loss": 1.8547, "step": 32570 }, { "epoch": 0.8, "learning_rate": 1.9706005497602443e-06, "loss": 1.9892, "step": 32571 }, { "epoch": 0.8, "learning_rate": 1.970124722148632e-06, "loss": 1.8953, "step": 32572 }, { "epoch": 0.8, "learning_rate": 1.969648945714029e-06, "loss": 2.0127, "step": 32573 }, { "epoch": 0.8, "learning_rate": 1.969173220459474e-06, "loss": 1.9403, "step": 32574 }, { "epoch": 0.8, "learning_rate": 1.9686975463879907e-06, "loss": 1.8957, "step": 32575 }, { "epoch": 0.8, "learning_rate": 1.9682219235026167e-06, "loss": 1.9065, "step": 32576 }, { "epoch": 0.8, "learning_rate": 1.9677463518063846e-06, "loss": 1.9897, "step": 32577 }, { "epoch": 0.8, "learning_rate": 1.967270831302319e-06, "loss": 1.8543, "step": 32578 }, { "epoch": 0.8, "learning_rate": 1.966795361993458e-06, "loss": 2.0608, "step": 32579 }, { "epoch": 0.8, "learning_rate": 1.966319943882824e-06, "loss": 1.8582, "step": 32580 }, { "epoch": 0.8, "learning_rate": 1.9658445769734524e-06, "loss": 1.8746, "step": 32581 }, { "epoch": 0.8, "learning_rate": 1.965369261268374e-06, "loss": 1.8282, "step": 32582 }, { "epoch": 0.8, "learning_rate": 1.9648939967706126e-06, "loss": 1.8598, "step": 32583 }, { "epoch": 0.8, "learning_rate": 1.9644187834832008e-06, "loss": 1.8433, "step": 32584 }, { "epoch": 0.8, "learning_rate": 1.963943621409169e-06, "loss": 1.8115, "step": 32585 }, { "epoch": 0.8, "learning_rate": 1.9634685105515407e-06, "loss": 1.8967, "step": 32586 }, { "epoch": 0.8, "learning_rate": 1.9629934509133466e-06, "loss": 1.7819, "step": 32587 }, { "epoch": 0.8, "learning_rate": 1.962518442497615e-06, "loss": 2.0065, "step": 32588 }, { "epoch": 0.8, "learning_rate": 1.9620434853073744e-06, "loss": 2.0609, "step": 32589 }, { "epoch": 0.8, "learning_rate": 1.9615685793456484e-06, "loss": 1.7996, "step": 32590 }, { "epoch": 0.8, "learning_rate": 1.961093724615465e-06, "loss": 1.7878, "step": 32591 }, { "epoch": 0.8, "learning_rate": 1.9606189211198546e-06, "loss": 1.9892, "step": 32592 }, { "epoch": 0.8, "learning_rate": 1.960144168861836e-06, "loss": 2.0546, "step": 32593 }, { "epoch": 0.8, "learning_rate": 1.959669467844442e-06, "loss": 1.9161, "step": 32594 }, { "epoch": 0.8, "learning_rate": 1.959194818070692e-06, "loss": 2.0789, "step": 32595 }, { "epoch": 0.8, "learning_rate": 1.958720219543614e-06, "loss": 1.8646, "step": 32596 }, { "epoch": 0.8, "learning_rate": 1.9582456722662346e-06, "loss": 1.9815, "step": 32597 }, { "epoch": 0.8, "learning_rate": 1.957771176241574e-06, "loss": 1.9633, "step": 32598 }, { "epoch": 0.8, "learning_rate": 1.9572967314726588e-06, "loss": 1.8907, "step": 32599 }, { "epoch": 0.8, "learning_rate": 1.956822337962515e-06, "loss": 1.9772, "step": 32600 }, { "epoch": 0.8, "learning_rate": 1.9563479957141605e-06, "loss": 1.9562, "step": 32601 }, { "epoch": 0.8, "learning_rate": 1.955873704730623e-06, "loss": 1.827, "step": 32602 }, { "epoch": 0.8, "learning_rate": 1.9553994650149256e-06, "loss": 1.9545, "step": 32603 }, { "epoch": 0.8, "learning_rate": 1.9549252765700865e-06, "loss": 1.9042, "step": 32604 }, { "epoch": 0.8, "learning_rate": 1.9544511393991304e-06, "loss": 1.9522, "step": 32605 }, { "epoch": 0.8, "learning_rate": 1.9539770535050793e-06, "loss": 1.9068, "step": 32606 }, { "epoch": 0.8, "learning_rate": 1.9535030188909564e-06, "loss": 1.9339, "step": 32607 }, { "epoch": 0.8, "learning_rate": 1.9530290355597793e-06, "loss": 2.1666, "step": 32608 }, { "epoch": 0.8, "learning_rate": 1.952555103514571e-06, "loss": 1.9564, "step": 32609 }, { "epoch": 0.8, "learning_rate": 1.952081222758353e-06, "loss": 2.0859, "step": 32610 }, { "epoch": 0.8, "learning_rate": 1.9516073932941425e-06, "loss": 1.8584, "step": 32611 }, { "epoch": 0.8, "learning_rate": 1.9511336151249627e-06, "loss": 1.9746, "step": 32612 }, { "epoch": 0.8, "learning_rate": 1.9506598882538295e-06, "loss": 1.8756, "step": 32613 }, { "epoch": 0.8, "learning_rate": 1.9501862126837635e-06, "loss": 2.0332, "step": 32614 }, { "epoch": 0.8, "learning_rate": 1.949712588417787e-06, "loss": 1.7744, "step": 32615 }, { "epoch": 0.8, "learning_rate": 1.9492390154589136e-06, "loss": 1.7807, "step": 32616 }, { "epoch": 0.8, "learning_rate": 1.9487654938101662e-06, "loss": 1.9989, "step": 32617 }, { "epoch": 0.8, "learning_rate": 1.948292023474556e-06, "loss": 1.8655, "step": 32618 }, { "epoch": 0.8, "learning_rate": 1.947818604455106e-06, "loss": 1.9466, "step": 32619 }, { "epoch": 0.8, "learning_rate": 1.9473452367548327e-06, "loss": 2.0505, "step": 32620 }, { "epoch": 0.8, "learning_rate": 1.946871920376754e-06, "loss": 1.978, "step": 32621 }, { "epoch": 0.8, "learning_rate": 1.9463986553238834e-06, "loss": 1.7998, "step": 32622 }, { "epoch": 0.8, "learning_rate": 1.945925441599238e-06, "loss": 1.8823, "step": 32623 }, { "epoch": 0.8, "learning_rate": 1.9454522792058337e-06, "loss": 1.8921, "step": 32624 }, { "epoch": 0.8, "learning_rate": 1.944979168146691e-06, "loss": 1.8366, "step": 32625 }, { "epoch": 0.8, "learning_rate": 1.9445061084248183e-06, "loss": 1.9464, "step": 32626 }, { "epoch": 0.8, "learning_rate": 1.9440331000432357e-06, "loss": 1.8678, "step": 32627 }, { "epoch": 0.8, "learning_rate": 1.943560143004952e-06, "loss": 1.9119, "step": 32628 }, { "epoch": 0.8, "learning_rate": 1.943087237312986e-06, "loss": 1.9418, "step": 32629 }, { "epoch": 0.8, "learning_rate": 1.9426143829703534e-06, "loss": 1.8533, "step": 32630 }, { "epoch": 0.8, "learning_rate": 1.9421415799800626e-06, "loss": 1.856, "step": 32631 }, { "epoch": 0.8, "learning_rate": 1.9416688283451324e-06, "loss": 1.9321, "step": 32632 }, { "epoch": 0.8, "learning_rate": 1.9411961280685687e-06, "loss": 1.8038, "step": 32633 }, { "epoch": 0.8, "learning_rate": 1.94072347915339e-06, "loss": 1.998, "step": 32634 }, { "epoch": 0.8, "learning_rate": 1.9402508816026088e-06, "loss": 2.0442, "step": 32635 }, { "epoch": 0.8, "learning_rate": 1.9397783354192334e-06, "loss": 2.0651, "step": 32636 }, { "epoch": 0.8, "learning_rate": 1.939305840606277e-06, "loss": 1.8525, "step": 32637 }, { "epoch": 0.8, "learning_rate": 1.9388333971667506e-06, "loss": 2.1536, "step": 32638 }, { "epoch": 0.8, "learning_rate": 1.9383610051036695e-06, "loss": 2.0448, "step": 32639 }, { "epoch": 0.8, "learning_rate": 1.937888664420039e-06, "loss": 1.8521, "step": 32640 }, { "epoch": 0.8, "learning_rate": 1.93741637511887e-06, "loss": 2.1406, "step": 32641 }, { "epoch": 0.8, "learning_rate": 1.9369441372031773e-06, "loss": 1.9009, "step": 32642 }, { "epoch": 0.8, "learning_rate": 1.9364719506759646e-06, "loss": 1.9311, "step": 32643 }, { "epoch": 0.8, "learning_rate": 1.9359998155402436e-06, "loss": 1.9602, "step": 32644 }, { "epoch": 0.8, "learning_rate": 1.935527731799027e-06, "loss": 1.9831, "step": 32645 }, { "epoch": 0.8, "learning_rate": 1.9350556994553173e-06, "loss": 1.9348, "step": 32646 }, { "epoch": 0.8, "learning_rate": 1.9345837185121285e-06, "loss": 2.0501, "step": 32647 }, { "epoch": 0.8, "learning_rate": 1.9341117889724635e-06, "loss": 1.9714, "step": 32648 }, { "epoch": 0.8, "learning_rate": 1.933639910839333e-06, "loss": 2.0587, "step": 32649 }, { "epoch": 0.8, "learning_rate": 1.9331680841157464e-06, "loss": 1.9547, "step": 32650 }, { "epoch": 0.8, "learning_rate": 1.9326963088047057e-06, "loss": 1.777, "step": 32651 }, { "epoch": 0.8, "learning_rate": 1.9322245849092235e-06, "loss": 1.8483, "step": 32652 }, { "epoch": 0.8, "learning_rate": 1.9317529124323008e-06, "loss": 1.8014, "step": 32653 }, { "epoch": 0.8, "learning_rate": 1.9312812913769462e-06, "loss": 1.7576, "step": 32654 }, { "epoch": 0.8, "learning_rate": 1.930809721746165e-06, "loss": 1.9192, "step": 32655 }, { "epoch": 0.8, "learning_rate": 1.9303382035429665e-06, "loss": 1.8895, "step": 32656 }, { "epoch": 0.8, "learning_rate": 1.9298667367703485e-06, "loss": 2.084, "step": 32657 }, { "epoch": 0.8, "learning_rate": 1.9293953214313197e-06, "loss": 2.1308, "step": 32658 }, { "epoch": 0.81, "learning_rate": 1.9289239575288864e-06, "loss": 1.9258, "step": 32659 }, { "epoch": 0.81, "learning_rate": 1.9284526450660524e-06, "loss": 1.8678, "step": 32660 }, { "epoch": 0.81, "learning_rate": 1.927981384045817e-06, "loss": 1.8906, "step": 32661 }, { "epoch": 0.81, "learning_rate": 1.9275101744711875e-06, "loss": 1.9207, "step": 32662 }, { "epoch": 0.81, "learning_rate": 1.92703901634517e-06, "loss": 1.8176, "step": 32663 }, { "epoch": 0.81, "learning_rate": 1.92656790967076e-06, "loss": 1.9488, "step": 32664 }, { "epoch": 0.81, "learning_rate": 1.926096854450966e-06, "loss": 1.9981, "step": 32665 }, { "epoch": 0.81, "learning_rate": 1.925625850688787e-06, "loss": 1.8275, "step": 32666 }, { "epoch": 0.81, "learning_rate": 1.925154898387225e-06, "loss": 1.9622, "step": 32667 }, { "epoch": 0.81, "learning_rate": 1.9246839975492838e-06, "loss": 1.9954, "step": 32668 }, { "epoch": 0.81, "learning_rate": 1.924213148177961e-06, "loss": 1.9029, "step": 32669 }, { "epoch": 0.81, "learning_rate": 1.923742350276263e-06, "loss": 2.0266, "step": 32670 }, { "epoch": 0.81, "learning_rate": 1.9232716038471844e-06, "loss": 2.0365, "step": 32671 }, { "epoch": 0.81, "learning_rate": 1.922800908893728e-06, "loss": 1.9105, "step": 32672 }, { "epoch": 0.81, "learning_rate": 1.9223302654188936e-06, "loss": 1.972, "step": 32673 }, { "epoch": 0.81, "learning_rate": 1.9218596734256835e-06, "loss": 1.929, "step": 32674 }, { "epoch": 0.81, "learning_rate": 1.921389132917091e-06, "loss": 1.8525, "step": 32675 }, { "epoch": 0.81, "learning_rate": 1.920918643896119e-06, "loss": 1.9811, "step": 32676 }, { "epoch": 0.81, "learning_rate": 1.920448206365765e-06, "loss": 1.8634, "step": 32677 }, { "epoch": 0.81, "learning_rate": 1.9199778203290298e-06, "loss": 1.8118, "step": 32678 }, { "epoch": 0.81, "learning_rate": 1.9195074857889074e-06, "loss": 2.1531, "step": 32679 }, { "epoch": 0.81, "learning_rate": 1.9190372027483996e-06, "loss": 1.9092, "step": 32680 }, { "epoch": 0.81, "learning_rate": 1.918566971210498e-06, "loss": 1.9599, "step": 32681 }, { "epoch": 0.81, "learning_rate": 1.9180967911782032e-06, "loss": 1.9365, "step": 32682 }, { "epoch": 0.81, "learning_rate": 1.9176266626545147e-06, "loss": 1.9838, "step": 32683 }, { "epoch": 0.81, "learning_rate": 1.917156585642421e-06, "loss": 1.9281, "step": 32684 }, { "epoch": 0.81, "learning_rate": 1.916686560144927e-06, "loss": 1.8754, "step": 32685 }, { "epoch": 0.81, "learning_rate": 1.9162165861650204e-06, "loss": 2.0144, "step": 32686 }, { "epoch": 0.81, "learning_rate": 1.9157466637056997e-06, "loss": 1.9613, "step": 32687 }, { "epoch": 0.81, "learning_rate": 1.9152767927699633e-06, "loss": 1.8988, "step": 32688 }, { "epoch": 0.81, "learning_rate": 1.9148069733608e-06, "loss": 2.0134, "step": 32689 }, { "epoch": 0.81, "learning_rate": 1.914337205481206e-06, "loss": 2.0465, "step": 32690 }, { "epoch": 0.81, "learning_rate": 1.913867489134177e-06, "loss": 2.075, "step": 32691 }, { "epoch": 0.81, "learning_rate": 1.9133978243227068e-06, "loss": 1.7998, "step": 32692 }, { "epoch": 0.81, "learning_rate": 1.912928211049787e-06, "loss": 1.9899, "step": 32693 }, { "epoch": 0.81, "learning_rate": 1.9124586493184096e-06, "loss": 2.0489, "step": 32694 }, { "epoch": 0.81, "learning_rate": 1.9119891391315728e-06, "loss": 1.9298, "step": 32695 }, { "epoch": 0.81, "learning_rate": 1.911519680492261e-06, "loss": 1.9208, "step": 32696 }, { "epoch": 0.81, "learning_rate": 1.9110502734034707e-06, "loss": 1.7872, "step": 32697 }, { "epoch": 0.81, "learning_rate": 1.9105809178681966e-06, "loss": 1.9469, "step": 32698 }, { "epoch": 0.81, "learning_rate": 1.9101116138894225e-06, "loss": 1.8177, "step": 32699 }, { "epoch": 0.81, "learning_rate": 1.9096423614701475e-06, "loss": 1.9259, "step": 32700 }, { "epoch": 0.81, "learning_rate": 1.9091731606133555e-06, "loss": 1.9889, "step": 32701 }, { "epoch": 0.81, "learning_rate": 1.908704011322039e-06, "loss": 1.9317, "step": 32702 }, { "epoch": 0.81, "learning_rate": 1.9082349135991917e-06, "loss": 1.9911, "step": 32703 }, { "epoch": 0.81, "learning_rate": 1.9077658674477972e-06, "loss": 1.9869, "step": 32704 }, { "epoch": 0.81, "learning_rate": 1.907296872870852e-06, "loss": 2.0334, "step": 32705 }, { "epoch": 0.81, "learning_rate": 1.9068279298713376e-06, "loss": 1.9028, "step": 32706 }, { "epoch": 0.81, "learning_rate": 1.9063590384522456e-06, "loss": 1.7943, "step": 32707 }, { "epoch": 0.81, "learning_rate": 1.9058901986165657e-06, "loss": 1.8085, "step": 32708 }, { "epoch": 0.81, "learning_rate": 1.9054214103672864e-06, "loss": 1.8574, "step": 32709 }, { "epoch": 0.81, "learning_rate": 1.9049526737073953e-06, "loss": 1.9654, "step": 32710 }, { "epoch": 0.81, "learning_rate": 1.904483988639877e-06, "loss": 1.8834, "step": 32711 }, { "epoch": 0.81, "learning_rate": 1.904015355167721e-06, "loss": 1.8683, "step": 32712 }, { "epoch": 0.81, "learning_rate": 1.9035467732939162e-06, "loss": 2.0069, "step": 32713 }, { "epoch": 0.81, "learning_rate": 1.9030782430214422e-06, "loss": 1.9636, "step": 32714 }, { "epoch": 0.81, "learning_rate": 1.9026097643532937e-06, "loss": 1.8306, "step": 32715 }, { "epoch": 0.81, "learning_rate": 1.902141337292448e-06, "loss": 1.8764, "step": 32716 }, { "epoch": 0.81, "learning_rate": 1.9016729618418961e-06, "loss": 2.0144, "step": 32717 }, { "epoch": 0.81, "learning_rate": 1.9012046380046233e-06, "loss": 1.8422, "step": 32718 }, { "epoch": 0.81, "learning_rate": 1.9007363657836098e-06, "loss": 1.9795, "step": 32719 }, { "epoch": 0.81, "learning_rate": 1.9002681451818427e-06, "loss": 1.9657, "step": 32720 }, { "epoch": 0.81, "learning_rate": 1.8997999762023101e-06, "loss": 2.0295, "step": 32721 }, { "epoch": 0.81, "learning_rate": 1.8993318588479892e-06, "loss": 1.867, "step": 32722 }, { "epoch": 0.81, "learning_rate": 1.898863793121868e-06, "loss": 1.8154, "step": 32723 }, { "epoch": 0.81, "learning_rate": 1.8983957790269258e-06, "loss": 1.8589, "step": 32724 }, { "epoch": 0.81, "learning_rate": 1.8979278165661475e-06, "loss": 1.8413, "step": 32725 }, { "epoch": 0.81, "learning_rate": 1.8974599057425158e-06, "loss": 1.9622, "step": 32726 }, { "epoch": 0.81, "learning_rate": 1.8969920465590131e-06, "loss": 1.9505, "step": 32727 }, { "epoch": 0.81, "learning_rate": 1.8965242390186224e-06, "loss": 1.961, "step": 32728 }, { "epoch": 0.81, "learning_rate": 1.896056483124321e-06, "loss": 1.9054, "step": 32729 }, { "epoch": 0.81, "learning_rate": 1.8955887788790928e-06, "loss": 1.8685, "step": 32730 }, { "epoch": 0.81, "learning_rate": 1.8951211262859215e-06, "loss": 1.7927, "step": 32731 }, { "epoch": 0.81, "learning_rate": 1.8946535253477815e-06, "loss": 1.8198, "step": 32732 }, { "epoch": 0.81, "learning_rate": 1.8941859760676595e-06, "loss": 1.9134, "step": 32733 }, { "epoch": 0.81, "learning_rate": 1.8937184784485286e-06, "loss": 1.9791, "step": 32734 }, { "epoch": 0.81, "learning_rate": 1.8932510324933717e-06, "loss": 1.9961, "step": 32735 }, { "epoch": 0.81, "learning_rate": 1.89278363820517e-06, "loss": 2.0333, "step": 32736 }, { "epoch": 0.81, "learning_rate": 1.892316295586899e-06, "loss": 1.9405, "step": 32737 }, { "epoch": 0.81, "learning_rate": 1.8918490046415395e-06, "loss": 2.0607, "step": 32738 }, { "epoch": 0.81, "learning_rate": 1.8913817653720667e-06, "loss": 1.9911, "step": 32739 }, { "epoch": 0.81, "learning_rate": 1.8909145777814608e-06, "loss": 1.7907, "step": 32740 }, { "epoch": 0.81, "learning_rate": 1.8904474418727014e-06, "loss": 1.828, "step": 32741 }, { "epoch": 0.81, "learning_rate": 1.8899803576487607e-06, "loss": 1.9126, "step": 32742 }, { "epoch": 0.81, "learning_rate": 1.8895133251126175e-06, "loss": 1.7857, "step": 32743 }, { "epoch": 0.81, "learning_rate": 1.889046344267249e-06, "loss": 1.9247, "step": 32744 }, { "epoch": 0.81, "learning_rate": 1.8885794151156344e-06, "loss": 1.8131, "step": 32745 }, { "epoch": 0.81, "learning_rate": 1.8881125376607446e-06, "loss": 1.9362, "step": 32746 }, { "epoch": 0.81, "learning_rate": 1.8876457119055568e-06, "loss": 2.0384, "step": 32747 }, { "epoch": 0.81, "learning_rate": 1.8871789378530493e-06, "loss": 1.9471, "step": 32748 }, { "epoch": 0.81, "learning_rate": 1.8867122155061912e-06, "loss": 1.979, "step": 32749 }, { "epoch": 0.81, "learning_rate": 1.8862455448679607e-06, "loss": 1.8429, "step": 32750 }, { "epoch": 0.81, "learning_rate": 1.8857789259413339e-06, "loss": 1.9498, "step": 32751 }, { "epoch": 0.81, "learning_rate": 1.8853123587292798e-06, "loss": 2.0613, "step": 32752 }, { "epoch": 0.81, "learning_rate": 1.884845843234777e-06, "loss": 1.9726, "step": 32753 }, { "epoch": 0.81, "learning_rate": 1.8843793794607946e-06, "loss": 1.9167, "step": 32754 }, { "epoch": 0.81, "learning_rate": 1.8839129674103064e-06, "loss": 1.912, "step": 32755 }, { "epoch": 0.81, "learning_rate": 1.883446607086289e-06, "loss": 1.921, "step": 32756 }, { "epoch": 0.81, "learning_rate": 1.8829802984917078e-06, "loss": 1.9977, "step": 32757 }, { "epoch": 0.81, "learning_rate": 1.8825140416295418e-06, "loss": 2.0267, "step": 32758 }, { "epoch": 0.81, "learning_rate": 1.8820478365027562e-06, "loss": 1.892, "step": 32759 }, { "epoch": 0.81, "learning_rate": 1.8815816831143252e-06, "loss": 2.0825, "step": 32760 }, { "epoch": 0.81, "learning_rate": 1.881115581467221e-06, "loss": 1.9613, "step": 32761 }, { "epoch": 0.81, "learning_rate": 1.8806495315644113e-06, "loss": 1.9432, "step": 32762 }, { "epoch": 0.81, "learning_rate": 1.8801835334088724e-06, "loss": 2.0916, "step": 32763 }, { "epoch": 0.81, "learning_rate": 1.8797175870035666e-06, "loss": 1.855, "step": 32764 }, { "epoch": 0.81, "learning_rate": 1.879251692351467e-06, "loss": 1.9702, "step": 32765 }, { "epoch": 0.81, "learning_rate": 1.8787858494555455e-06, "loss": 2.146, "step": 32766 }, { "epoch": 0.81, "learning_rate": 1.8783200583187656e-06, "loss": 2.0898, "step": 32767 }, { "epoch": 0.81, "learning_rate": 1.8778543189441022e-06, "loss": 2.02, "step": 32768 }, { "epoch": 0.81, "learning_rate": 1.8773886313345168e-06, "loss": 1.9581, "step": 32769 }, { "epoch": 0.81, "learning_rate": 1.8769229954929802e-06, "loss": 1.9632, "step": 32770 }, { "epoch": 0.81, "learning_rate": 1.8764574114224644e-06, "loss": 2.0107, "step": 32771 }, { "epoch": 0.81, "learning_rate": 1.87599187912593e-06, "loss": 1.9671, "step": 32772 }, { "epoch": 0.81, "learning_rate": 1.8755263986063489e-06, "loss": 1.6429, "step": 32773 }, { "epoch": 0.81, "learning_rate": 1.8750609698666833e-06, "loss": 1.9023, "step": 32774 }, { "epoch": 0.81, "learning_rate": 1.8745955929099024e-06, "loss": 1.8754, "step": 32775 }, { "epoch": 0.81, "learning_rate": 1.874130267738974e-06, "loss": 1.8199, "step": 32776 }, { "epoch": 0.81, "learning_rate": 1.8736649943568597e-06, "loss": 1.9367, "step": 32777 }, { "epoch": 0.81, "learning_rate": 1.8731997727665252e-06, "loss": 1.9314, "step": 32778 }, { "epoch": 0.81, "learning_rate": 1.872734602970938e-06, "loss": 1.957, "step": 32779 }, { "epoch": 0.81, "learning_rate": 1.8722694849730616e-06, "loss": 2.0474, "step": 32780 }, { "epoch": 0.81, "learning_rate": 1.8718044187758633e-06, "loss": 1.8937, "step": 32781 }, { "epoch": 0.81, "learning_rate": 1.871339404382302e-06, "loss": 1.8119, "step": 32782 }, { "epoch": 0.81, "learning_rate": 1.8708744417953429e-06, "loss": 1.7698, "step": 32783 }, { "epoch": 0.81, "learning_rate": 1.8704095310179538e-06, "loss": 2.007, "step": 32784 }, { "epoch": 0.81, "learning_rate": 1.8699446720530911e-06, "loss": 1.944, "step": 32785 }, { "epoch": 0.81, "learning_rate": 1.8694798649037226e-06, "loss": 1.902, "step": 32786 }, { "epoch": 0.81, "learning_rate": 1.869015109572806e-06, "loss": 2.0419, "step": 32787 }, { "epoch": 0.81, "learning_rate": 1.8685504060633065e-06, "loss": 2.0243, "step": 32788 }, { "epoch": 0.81, "learning_rate": 1.8680857543781872e-06, "loss": 1.9934, "step": 32789 }, { "epoch": 0.81, "learning_rate": 1.867621154520406e-06, "loss": 1.8711, "step": 32790 }, { "epoch": 0.81, "learning_rate": 1.867156606492927e-06, "loss": 1.9381, "step": 32791 }, { "epoch": 0.81, "learning_rate": 1.8666921102987068e-06, "loss": 2.0283, "step": 32792 }, { "epoch": 0.81, "learning_rate": 1.8662276659407085e-06, "loss": 2.0508, "step": 32793 }, { "epoch": 0.81, "learning_rate": 1.8657632734218955e-06, "loss": 2.0878, "step": 32794 }, { "epoch": 0.81, "learning_rate": 1.865298932745221e-06, "loss": 2.0254, "step": 32795 }, { "epoch": 0.81, "learning_rate": 1.8648346439136467e-06, "loss": 1.8709, "step": 32796 }, { "epoch": 0.81, "learning_rate": 1.8643704069301327e-06, "loss": 1.7673, "step": 32797 }, { "epoch": 0.81, "learning_rate": 1.8639062217976379e-06, "loss": 1.8323, "step": 32798 }, { "epoch": 0.81, "learning_rate": 1.8634420885191217e-06, "loss": 1.7516, "step": 32799 }, { "epoch": 0.81, "learning_rate": 1.862978007097539e-06, "loss": 1.807, "step": 32800 }, { "epoch": 0.81, "learning_rate": 1.8625139775358525e-06, "loss": 1.9359, "step": 32801 }, { "epoch": 0.81, "learning_rate": 1.8620499998370123e-06, "loss": 1.8653, "step": 32802 }, { "epoch": 0.81, "learning_rate": 1.8615860740039804e-06, "loss": 1.9643, "step": 32803 }, { "epoch": 0.81, "learning_rate": 1.8611222000397145e-06, "loss": 1.9733, "step": 32804 }, { "epoch": 0.81, "learning_rate": 1.8606583779471676e-06, "loss": 1.8658, "step": 32805 }, { "epoch": 0.81, "learning_rate": 1.8601946077292998e-06, "loss": 1.7824, "step": 32806 }, { "epoch": 0.81, "learning_rate": 1.859730889389061e-06, "loss": 1.9609, "step": 32807 }, { "epoch": 0.81, "learning_rate": 1.8592672229294107e-06, "loss": 1.7764, "step": 32808 }, { "epoch": 0.81, "learning_rate": 1.8588036083533068e-06, "loss": 1.7634, "step": 32809 }, { "epoch": 0.81, "learning_rate": 1.8583400456636969e-06, "loss": 2.0133, "step": 32810 }, { "epoch": 0.81, "learning_rate": 1.8578765348635419e-06, "loss": 2.0581, "step": 32811 }, { "epoch": 0.81, "learning_rate": 1.8574130759557906e-06, "loss": 1.8744, "step": 32812 }, { "epoch": 0.81, "learning_rate": 1.8569496689433996e-06, "loss": 2.0576, "step": 32813 }, { "epoch": 0.81, "learning_rate": 1.856486313829322e-06, "loss": 1.9868, "step": 32814 }, { "epoch": 0.81, "learning_rate": 1.8560230106165112e-06, "loss": 1.9895, "step": 32815 }, { "epoch": 0.81, "learning_rate": 1.8555597593079222e-06, "loss": 1.7867, "step": 32816 }, { "epoch": 0.81, "learning_rate": 1.8550965599065029e-06, "loss": 1.9428, "step": 32817 }, { "epoch": 0.81, "learning_rate": 1.8546334124152065e-06, "loss": 1.9832, "step": 32818 }, { "epoch": 0.81, "learning_rate": 1.8541703168369895e-06, "loss": 1.9337, "step": 32819 }, { "epoch": 0.81, "learning_rate": 1.8537072731747962e-06, "loss": 1.9411, "step": 32820 }, { "epoch": 0.81, "learning_rate": 1.8532442814315854e-06, "loss": 2.0317, "step": 32821 }, { "epoch": 0.81, "learning_rate": 1.8527813416102991e-06, "loss": 1.929, "step": 32822 }, { "epoch": 0.81, "learning_rate": 1.852318453713894e-06, "loss": 1.9487, "step": 32823 }, { "epoch": 0.81, "learning_rate": 1.8518556177453206e-06, "loss": 1.9824, "step": 32824 }, { "epoch": 0.81, "learning_rate": 1.8513928337075248e-06, "loss": 1.9016, "step": 32825 }, { "epoch": 0.81, "learning_rate": 1.8509301016034597e-06, "loss": 1.9399, "step": 32826 }, { "epoch": 0.81, "learning_rate": 1.8504674214360708e-06, "loss": 1.908, "step": 32827 }, { "epoch": 0.81, "learning_rate": 1.8500047932083087e-06, "loss": 1.8891, "step": 32828 }, { "epoch": 0.81, "learning_rate": 1.8495422169231237e-06, "loss": 1.8814, "step": 32829 }, { "epoch": 0.81, "learning_rate": 1.8490796925834609e-06, "loss": 1.8698, "step": 32830 }, { "epoch": 0.81, "learning_rate": 1.8486172201922692e-06, "loss": 1.8936, "step": 32831 }, { "epoch": 0.81, "learning_rate": 1.8481547997524962e-06, "loss": 1.9877, "step": 32832 }, { "epoch": 0.81, "learning_rate": 1.8476924312670897e-06, "loss": 2.0068, "step": 32833 }, { "epoch": 0.81, "learning_rate": 1.8472301147389992e-06, "loss": 1.7987, "step": 32834 }, { "epoch": 0.81, "learning_rate": 1.8467678501711639e-06, "loss": 1.8104, "step": 32835 }, { "epoch": 0.81, "learning_rate": 1.8463056375665357e-06, "loss": 1.8524, "step": 32836 }, { "epoch": 0.81, "learning_rate": 1.8458434769280609e-06, "loss": 1.9101, "step": 32837 }, { "epoch": 0.81, "learning_rate": 1.8453813682586797e-06, "loss": 2.051, "step": 32838 }, { "epoch": 0.81, "learning_rate": 1.844919311561344e-06, "loss": 1.9475, "step": 32839 }, { "epoch": 0.81, "learning_rate": 1.8444573068389925e-06, "loss": 2.0403, "step": 32840 }, { "epoch": 0.81, "learning_rate": 1.843995354094572e-06, "loss": 1.8884, "step": 32841 }, { "epoch": 0.81, "learning_rate": 1.8435334533310301e-06, "loss": 1.9525, "step": 32842 }, { "epoch": 0.81, "learning_rate": 1.8430716045513053e-06, "loss": 1.8414, "step": 32843 }, { "epoch": 0.81, "learning_rate": 1.8426098077583455e-06, "loss": 1.9296, "step": 32844 }, { "epoch": 0.81, "learning_rate": 1.8421480629550891e-06, "loss": 1.7938, "step": 32845 }, { "epoch": 0.81, "learning_rate": 1.841686370144482e-06, "loss": 2.1168, "step": 32846 }, { "epoch": 0.81, "learning_rate": 1.8412247293294684e-06, "loss": 1.8668, "step": 32847 }, { "epoch": 0.81, "learning_rate": 1.8407631405129866e-06, "loss": 1.9737, "step": 32848 }, { "epoch": 0.81, "learning_rate": 1.8403016036979803e-06, "loss": 1.9488, "step": 32849 }, { "epoch": 0.81, "learning_rate": 1.8398401188873905e-06, "loss": 1.8043, "step": 32850 }, { "epoch": 0.81, "learning_rate": 1.839378686084159e-06, "loss": 1.9232, "step": 32851 }, { "epoch": 0.81, "learning_rate": 1.8389173052912301e-06, "loss": 1.9549, "step": 32852 }, { "epoch": 0.81, "learning_rate": 1.8384559765115372e-06, "loss": 1.8468, "step": 32853 }, { "epoch": 0.81, "learning_rate": 1.8379946997480268e-06, "loss": 1.9384, "step": 32854 }, { "epoch": 0.81, "learning_rate": 1.8375334750036344e-06, "loss": 1.799, "step": 32855 }, { "epoch": 0.81, "learning_rate": 1.8370723022812997e-06, "loss": 1.8647, "step": 32856 }, { "epoch": 0.81, "learning_rate": 1.836611181583967e-06, "loss": 2.0618, "step": 32857 }, { "epoch": 0.81, "learning_rate": 1.8361501129145698e-06, "loss": 1.9471, "step": 32858 }, { "epoch": 0.81, "learning_rate": 1.8356890962760498e-06, "loss": 1.9235, "step": 32859 }, { "epoch": 0.81, "learning_rate": 1.8352281316713417e-06, "loss": 1.7736, "step": 32860 }, { "epoch": 0.81, "learning_rate": 1.834767219103386e-06, "loss": 1.934, "step": 32861 }, { "epoch": 0.81, "learning_rate": 1.834306358575122e-06, "loss": 2.0286, "step": 32862 }, { "epoch": 0.81, "learning_rate": 1.8338455500894814e-06, "loss": 1.9392, "step": 32863 }, { "epoch": 0.81, "learning_rate": 1.8333847936494076e-06, "loss": 2.0533, "step": 32864 }, { "epoch": 0.81, "learning_rate": 1.8329240892578314e-06, "loss": 1.9669, "step": 32865 }, { "epoch": 0.81, "learning_rate": 1.832463436917692e-06, "loss": 1.8264, "step": 32866 }, { "epoch": 0.81, "learning_rate": 1.8320028366319242e-06, "loss": 1.891, "step": 32867 }, { "epoch": 0.81, "learning_rate": 1.8315422884034639e-06, "loss": 2.0002, "step": 32868 }, { "epoch": 0.81, "learning_rate": 1.8310817922352498e-06, "loss": 2.0744, "step": 32869 }, { "epoch": 0.81, "learning_rate": 1.8306213481302104e-06, "loss": 2.0198, "step": 32870 }, { "epoch": 0.81, "learning_rate": 1.8301609560912848e-06, "loss": 2.0879, "step": 32871 }, { "epoch": 0.81, "learning_rate": 1.829700616121407e-06, "loss": 2.154, "step": 32872 }, { "epoch": 0.81, "learning_rate": 1.8292403282235082e-06, "loss": 1.8917, "step": 32873 }, { "epoch": 0.81, "learning_rate": 1.828780092400525e-06, "loss": 2.0537, "step": 32874 }, { "epoch": 0.81, "learning_rate": 1.8283199086553883e-06, "loss": 1.9909, "step": 32875 }, { "epoch": 0.81, "learning_rate": 1.8278597769910312e-06, "loss": 1.935, "step": 32876 }, { "epoch": 0.81, "learning_rate": 1.827399697410389e-06, "loss": 1.9465, "step": 32877 }, { "epoch": 0.81, "learning_rate": 1.82693966991639e-06, "loss": 1.8786, "step": 32878 }, { "epoch": 0.81, "learning_rate": 1.82647969451197e-06, "loss": 2.0514, "step": 32879 }, { "epoch": 0.81, "learning_rate": 1.826019771200056e-06, "loss": 1.9109, "step": 32880 }, { "epoch": 0.81, "learning_rate": 1.8255598999835821e-06, "loss": 1.9693, "step": 32881 }, { "epoch": 0.81, "learning_rate": 1.8251000808654817e-06, "loss": 1.954, "step": 32882 }, { "epoch": 0.81, "learning_rate": 1.824640313848679e-06, "loss": 1.8115, "step": 32883 }, { "epoch": 0.81, "learning_rate": 1.8241805989361094e-06, "loss": 1.9752, "step": 32884 }, { "epoch": 0.81, "learning_rate": 1.8237209361307007e-06, "loss": 2.0001, "step": 32885 }, { "epoch": 0.81, "learning_rate": 1.8232613254353825e-06, "loss": 1.8249, "step": 32886 }, { "epoch": 0.81, "learning_rate": 1.8228017668530873e-06, "loss": 1.9654, "step": 32887 }, { "epoch": 0.81, "learning_rate": 1.822342260386739e-06, "loss": 1.8846, "step": 32888 }, { "epoch": 0.81, "learning_rate": 1.8218828060392712e-06, "loss": 2.0324, "step": 32889 }, { "epoch": 0.81, "learning_rate": 1.821423403813607e-06, "loss": 1.9817, "step": 32890 }, { "epoch": 0.81, "learning_rate": 1.8209640537126761e-06, "loss": 1.8084, "step": 32891 }, { "epoch": 0.81, "learning_rate": 1.8205047557394095e-06, "loss": 2.1641, "step": 32892 }, { "epoch": 0.81, "learning_rate": 1.8200455098967296e-06, "loss": 1.9987, "step": 32893 }, { "epoch": 0.81, "learning_rate": 1.8195863161875683e-06, "loss": 1.9758, "step": 32894 }, { "epoch": 0.81, "learning_rate": 1.8191271746148453e-06, "loss": 1.807, "step": 32895 }, { "epoch": 0.81, "learning_rate": 1.818668085181492e-06, "loss": 1.9737, "step": 32896 }, { "epoch": 0.81, "learning_rate": 1.8182090478904357e-06, "loss": 1.7576, "step": 32897 }, { "epoch": 0.81, "learning_rate": 1.8177500627445965e-06, "loss": 1.7227, "step": 32898 }, { "epoch": 0.81, "learning_rate": 1.8172911297469031e-06, "loss": 1.907, "step": 32899 }, { "epoch": 0.81, "learning_rate": 1.8168322489002832e-06, "loss": 2.0416, "step": 32900 }, { "epoch": 0.81, "learning_rate": 1.8163734202076544e-06, "loss": 1.8163, "step": 32901 }, { "epoch": 0.81, "learning_rate": 1.8159146436719454e-06, "loss": 2.0242, "step": 32902 }, { "epoch": 0.81, "learning_rate": 1.8154559192960796e-06, "loss": 1.9991, "step": 32903 }, { "epoch": 0.81, "learning_rate": 1.8149972470829803e-06, "loss": 1.7085, "step": 32904 }, { "epoch": 0.81, "learning_rate": 1.8145386270355724e-06, "loss": 1.9073, "step": 32905 }, { "epoch": 0.81, "learning_rate": 1.8140800591567764e-06, "loss": 2.0198, "step": 32906 }, { "epoch": 0.81, "learning_rate": 1.8136215434495174e-06, "loss": 1.816, "step": 32907 }, { "epoch": 0.81, "learning_rate": 1.8131630799167144e-06, "loss": 2.0103, "step": 32908 }, { "epoch": 0.81, "learning_rate": 1.8127046685612915e-06, "loss": 1.8419, "step": 32909 }, { "epoch": 0.81, "learning_rate": 1.812246309386172e-06, "loss": 1.9062, "step": 32910 }, { "epoch": 0.81, "learning_rate": 1.8117880023942725e-06, "loss": 2.056, "step": 32911 }, { "epoch": 0.81, "learning_rate": 1.8113297475885182e-06, "loss": 1.8436, "step": 32912 }, { "epoch": 0.81, "learning_rate": 1.810871544971826e-06, "loss": 1.8653, "step": 32913 }, { "epoch": 0.81, "learning_rate": 1.8104133945471193e-06, "loss": 2.1487, "step": 32914 }, { "epoch": 0.81, "learning_rate": 1.8099552963173183e-06, "loss": 1.9137, "step": 32915 }, { "epoch": 0.81, "learning_rate": 1.8094972502853391e-06, "loss": 1.9305, "step": 32916 }, { "epoch": 0.81, "learning_rate": 1.8090392564541048e-06, "loss": 1.7254, "step": 32917 }, { "epoch": 0.81, "learning_rate": 1.808581314826532e-06, "loss": 1.8969, "step": 32918 }, { "epoch": 0.81, "learning_rate": 1.808123425405538e-06, "loss": 2.035, "step": 32919 }, { "epoch": 0.81, "learning_rate": 1.8076655881940442e-06, "loss": 1.9455, "step": 32920 }, { "epoch": 0.81, "learning_rate": 1.807207803194967e-06, "loss": 1.876, "step": 32921 }, { "epoch": 0.81, "learning_rate": 1.8067500704112272e-06, "loss": 1.8917, "step": 32922 }, { "epoch": 0.81, "learning_rate": 1.8062923898457374e-06, "loss": 1.8223, "step": 32923 }, { "epoch": 0.81, "learning_rate": 1.8058347615014148e-06, "loss": 1.8591, "step": 32924 }, { "epoch": 0.81, "learning_rate": 1.8053771853811818e-06, "loss": 1.9807, "step": 32925 }, { "epoch": 0.81, "learning_rate": 1.8049196614879472e-06, "loss": 1.9457, "step": 32926 }, { "epoch": 0.81, "learning_rate": 1.8044621898246327e-06, "loss": 2.0658, "step": 32927 }, { "epoch": 0.81, "learning_rate": 1.80400477039415e-06, "loss": 2.0479, "step": 32928 }, { "epoch": 0.81, "learning_rate": 1.803547403199415e-06, "loss": 1.9292, "step": 32929 }, { "epoch": 0.81, "learning_rate": 1.8030900882433466e-06, "loss": 2.0017, "step": 32930 }, { "epoch": 0.81, "learning_rate": 1.8026328255288527e-06, "loss": 1.8606, "step": 32931 }, { "epoch": 0.81, "learning_rate": 1.8021756150588555e-06, "loss": 1.8124, "step": 32932 }, { "epoch": 0.81, "learning_rate": 1.801718456836261e-06, "loss": 1.938, "step": 32933 }, { "epoch": 0.81, "learning_rate": 1.8012613508639865e-06, "loss": 1.9446, "step": 32934 }, { "epoch": 0.81, "learning_rate": 1.8008042971449479e-06, "loss": 1.8094, "step": 32935 }, { "epoch": 0.81, "learning_rate": 1.8003472956820532e-06, "loss": 1.9092, "step": 32936 }, { "epoch": 0.81, "learning_rate": 1.7998903464782181e-06, "loss": 1.642, "step": 32937 }, { "epoch": 0.81, "learning_rate": 1.799433449536353e-06, "loss": 1.7617, "step": 32938 }, { "epoch": 0.81, "learning_rate": 1.7989766048593716e-06, "loss": 1.8314, "step": 32939 }, { "epoch": 0.81, "learning_rate": 1.7985198124501869e-06, "loss": 1.8635, "step": 32940 }, { "epoch": 0.81, "learning_rate": 1.7980630723117065e-06, "loss": 1.8069, "step": 32941 }, { "epoch": 0.81, "learning_rate": 1.797606384446845e-06, "loss": 1.8861, "step": 32942 }, { "epoch": 0.81, "learning_rate": 1.7971497488585087e-06, "loss": 1.9904, "step": 32943 }, { "epoch": 0.81, "learning_rate": 1.79669316554961e-06, "loss": 1.9436, "step": 32944 }, { "epoch": 0.81, "learning_rate": 1.7962366345230631e-06, "loss": 2.0476, "step": 32945 }, { "epoch": 0.81, "learning_rate": 1.79578015578177e-06, "loss": 1.8544, "step": 32946 }, { "epoch": 0.81, "learning_rate": 1.7953237293286462e-06, "loss": 1.8085, "step": 32947 }, { "epoch": 0.81, "learning_rate": 1.794867355166595e-06, "loss": 1.8632, "step": 32948 }, { "epoch": 0.81, "learning_rate": 1.7944110332985287e-06, "loss": 2.0062, "step": 32949 }, { "epoch": 0.81, "learning_rate": 1.793954763727358e-06, "loss": 1.942, "step": 32950 }, { "epoch": 0.81, "learning_rate": 1.793498546455985e-06, "loss": 2.2007, "step": 32951 }, { "epoch": 0.81, "learning_rate": 1.7930423814873221e-06, "loss": 2.0474, "step": 32952 }, { "epoch": 0.81, "learning_rate": 1.7925862688242735e-06, "loss": 1.7874, "step": 32953 }, { "epoch": 0.81, "learning_rate": 1.7921302084697457e-06, "loss": 1.8432, "step": 32954 }, { "epoch": 0.81, "learning_rate": 1.7916742004266485e-06, "loss": 2.05, "step": 32955 }, { "epoch": 0.81, "learning_rate": 1.7912182446978855e-06, "loss": 1.8477, "step": 32956 }, { "epoch": 0.81, "learning_rate": 1.790762341286364e-06, "loss": 2.083, "step": 32957 }, { "epoch": 0.81, "learning_rate": 1.7903064901949917e-06, "loss": 1.8902, "step": 32958 }, { "epoch": 0.81, "learning_rate": 1.78985069142667e-06, "loss": 2.0177, "step": 32959 }, { "epoch": 0.81, "learning_rate": 1.7893949449843074e-06, "loss": 1.8472, "step": 32960 }, { "epoch": 0.81, "learning_rate": 1.7889392508708037e-06, "loss": 1.9465, "step": 32961 }, { "epoch": 0.81, "learning_rate": 1.7884836090890666e-06, "loss": 1.921, "step": 32962 }, { "epoch": 0.81, "learning_rate": 1.7880280196420018e-06, "loss": 1.7522, "step": 32963 }, { "epoch": 0.81, "learning_rate": 1.7875724825325092e-06, "loss": 2.0197, "step": 32964 }, { "epoch": 0.81, "learning_rate": 1.7871169977634951e-06, "loss": 1.7966, "step": 32965 }, { "epoch": 0.81, "learning_rate": 1.7866615653378584e-06, "loss": 1.8996, "step": 32966 }, { "epoch": 0.81, "learning_rate": 1.7862061852585033e-06, "loss": 2.2362, "step": 32967 }, { "epoch": 0.81, "learning_rate": 1.7857508575283366e-06, "loss": 1.7817, "step": 32968 }, { "epoch": 0.81, "learning_rate": 1.7852955821502538e-06, "loss": 1.9175, "step": 32969 }, { "epoch": 0.81, "learning_rate": 1.7848403591271623e-06, "loss": 2.0915, "step": 32970 }, { "epoch": 0.81, "learning_rate": 1.7843851884619568e-06, "loss": 2.0714, "step": 32971 }, { "epoch": 0.81, "learning_rate": 1.7839300701575435e-06, "loss": 1.9765, "step": 32972 }, { "epoch": 0.81, "learning_rate": 1.7834750042168202e-06, "loss": 1.9188, "step": 32973 }, { "epoch": 0.81, "learning_rate": 1.7830199906426882e-06, "loss": 1.819, "step": 32974 }, { "epoch": 0.81, "learning_rate": 1.7825650294380515e-06, "loss": 1.8876, "step": 32975 }, { "epoch": 0.81, "learning_rate": 1.7821101206058023e-06, "loss": 2.0413, "step": 32976 }, { "epoch": 0.81, "learning_rate": 1.7816552641488427e-06, "loss": 1.8215, "step": 32977 }, { "epoch": 0.81, "learning_rate": 1.781200460070076e-06, "loss": 1.7181, "step": 32978 }, { "epoch": 0.81, "learning_rate": 1.7807457083723944e-06, "loss": 1.8173, "step": 32979 }, { "epoch": 0.81, "learning_rate": 1.7802910090587022e-06, "loss": 1.8651, "step": 32980 }, { "epoch": 0.81, "learning_rate": 1.7798363621318903e-06, "loss": 1.9263, "step": 32981 }, { "epoch": 0.81, "learning_rate": 1.779381767594861e-06, "loss": 2.0201, "step": 32982 }, { "epoch": 0.81, "learning_rate": 1.778927225450513e-06, "loss": 1.9421, "step": 32983 }, { "epoch": 0.81, "learning_rate": 1.7784727357017396e-06, "loss": 1.9237, "step": 32984 }, { "epoch": 0.81, "learning_rate": 1.7780182983514393e-06, "loss": 2.077, "step": 32985 }, { "epoch": 0.81, "learning_rate": 1.7775639134025069e-06, "loss": 2.0442, "step": 32986 }, { "epoch": 0.81, "learning_rate": 1.7771095808578387e-06, "loss": 1.8459, "step": 32987 }, { "epoch": 0.81, "learning_rate": 1.7766553007203335e-06, "loss": 2.0375, "step": 32988 }, { "epoch": 0.81, "learning_rate": 1.7762010729928813e-06, "loss": 2.0417, "step": 32989 }, { "epoch": 0.81, "learning_rate": 1.7757468976783798e-06, "loss": 1.9741, "step": 32990 }, { "epoch": 0.81, "learning_rate": 1.7752927747797233e-06, "loss": 2.081, "step": 32991 }, { "epoch": 0.81, "learning_rate": 1.7748387042998062e-06, "loss": 1.8331, "step": 32992 }, { "epoch": 0.81, "learning_rate": 1.774384686241525e-06, "loss": 1.9556, "step": 32993 }, { "epoch": 0.81, "learning_rate": 1.7739307206077693e-06, "loss": 2.0831, "step": 32994 }, { "epoch": 0.81, "learning_rate": 1.7734768074014353e-06, "loss": 1.8387, "step": 32995 }, { "epoch": 0.81, "learning_rate": 1.7730229466254123e-06, "loss": 2.0348, "step": 32996 }, { "epoch": 0.81, "learning_rate": 1.772569138282596e-06, "loss": 1.8431, "step": 32997 }, { "epoch": 0.81, "learning_rate": 1.7721153823758796e-06, "loss": 1.8902, "step": 32998 }, { "epoch": 0.81, "learning_rate": 1.7716616789081508e-06, "loss": 1.8528, "step": 32999 }, { "epoch": 0.81, "learning_rate": 1.7712080278823062e-06, "loss": 2.042, "step": 33000 }, { "epoch": 0.81, "learning_rate": 1.7707544293012325e-06, "loss": 1.8812, "step": 33001 }, { "epoch": 0.81, "learning_rate": 1.7703008831678215e-06, "loss": 2.1034, "step": 33002 }, { "epoch": 0.81, "learning_rate": 1.7698473894849688e-06, "loss": 1.9496, "step": 33003 }, { "epoch": 0.81, "learning_rate": 1.7693939482555577e-06, "loss": 2.1057, "step": 33004 }, { "epoch": 0.81, "learning_rate": 1.7689405594824837e-06, "loss": 2.0638, "step": 33005 }, { "epoch": 0.81, "learning_rate": 1.7684872231686311e-06, "loss": 2.0025, "step": 33006 }, { "epoch": 0.81, "learning_rate": 1.7680339393168922e-06, "loss": 2.0413, "step": 33007 }, { "epoch": 0.81, "learning_rate": 1.7675807079301566e-06, "loss": 2.0721, "step": 33008 }, { "epoch": 0.81, "learning_rate": 1.767127529011311e-06, "loss": 1.8662, "step": 33009 }, { "epoch": 0.81, "learning_rate": 1.7666744025632475e-06, "loss": 1.9481, "step": 33010 }, { "epoch": 0.81, "learning_rate": 1.7662213285888486e-06, "loss": 2.0137, "step": 33011 }, { "epoch": 0.81, "learning_rate": 1.7657683070910048e-06, "loss": 1.9652, "step": 33012 }, { "epoch": 0.81, "learning_rate": 1.7653153380726051e-06, "loss": 1.8581, "step": 33013 }, { "epoch": 0.81, "learning_rate": 1.7648624215365318e-06, "loss": 2.0315, "step": 33014 }, { "epoch": 0.81, "learning_rate": 1.7644095574856745e-06, "loss": 1.9187, "step": 33015 }, { "epoch": 0.81, "learning_rate": 1.7639567459229212e-06, "loss": 1.9664, "step": 33016 }, { "epoch": 0.81, "learning_rate": 1.7635039868511527e-06, "loss": 1.7995, "step": 33017 }, { "epoch": 0.81, "learning_rate": 1.7630512802732613e-06, "loss": 1.8848, "step": 33018 }, { "epoch": 0.81, "learning_rate": 1.7625986261921246e-06, "loss": 1.9427, "step": 33019 }, { "epoch": 0.81, "learning_rate": 1.7621460246106325e-06, "loss": 1.9129, "step": 33020 }, { "epoch": 0.81, "learning_rate": 1.7616934755316694e-06, "loss": 1.8882, "step": 33021 }, { "epoch": 0.81, "learning_rate": 1.7612409789581175e-06, "loss": 2.0897, "step": 33022 }, { "epoch": 0.81, "learning_rate": 1.7607885348928632e-06, "loss": 2.0031, "step": 33023 }, { "epoch": 0.81, "learning_rate": 1.7603361433387867e-06, "loss": 1.8253, "step": 33024 }, { "epoch": 0.81, "learning_rate": 1.759883804298773e-06, "loss": 1.943, "step": 33025 }, { "epoch": 0.81, "learning_rate": 1.759431517775706e-06, "loss": 1.7771, "step": 33026 }, { "epoch": 0.81, "learning_rate": 1.7589792837724673e-06, "loss": 1.9217, "step": 33027 }, { "epoch": 0.81, "learning_rate": 1.7585271022919403e-06, "loss": 1.9699, "step": 33028 }, { "epoch": 0.81, "learning_rate": 1.758074973337005e-06, "loss": 1.8463, "step": 33029 }, { "epoch": 0.81, "learning_rate": 1.7576228969105436e-06, "loss": 1.9279, "step": 33030 }, { "epoch": 0.81, "learning_rate": 1.757170873015439e-06, "loss": 1.8777, "step": 33031 }, { "epoch": 0.81, "learning_rate": 1.7567189016545694e-06, "loss": 1.9781, "step": 33032 }, { "epoch": 0.81, "learning_rate": 1.7562669828308187e-06, "loss": 1.9403, "step": 33033 }, { "epoch": 0.81, "learning_rate": 1.7558151165470628e-06, "loss": 1.9537, "step": 33034 }, { "epoch": 0.81, "learning_rate": 1.7553633028061833e-06, "loss": 1.9035, "step": 33035 }, { "epoch": 0.81, "learning_rate": 1.7549115416110629e-06, "loss": 1.8535, "step": 33036 }, { "epoch": 0.81, "learning_rate": 1.7544598329645768e-06, "loss": 1.7671, "step": 33037 }, { "epoch": 0.81, "learning_rate": 1.754008176869607e-06, "loss": 1.9856, "step": 33038 }, { "epoch": 0.81, "learning_rate": 1.7535565733290271e-06, "loss": 1.8108, "step": 33039 }, { "epoch": 0.81, "learning_rate": 1.7531050223457203e-06, "loss": 2.0603, "step": 33040 }, { "epoch": 0.81, "learning_rate": 1.752653523922564e-06, "loss": 2.0769, "step": 33041 }, { "epoch": 0.81, "learning_rate": 1.7522020780624327e-06, "loss": 1.9064, "step": 33042 }, { "epoch": 0.81, "learning_rate": 1.7517506847682052e-06, "loss": 1.8915, "step": 33043 }, { "epoch": 0.81, "learning_rate": 1.7512993440427585e-06, "loss": 1.863, "step": 33044 }, { "epoch": 0.81, "learning_rate": 1.7508480558889685e-06, "loss": 2.0108, "step": 33045 }, { "epoch": 0.81, "learning_rate": 1.7503968203097154e-06, "loss": 2.0053, "step": 33046 }, { "epoch": 0.81, "learning_rate": 1.7499456373078705e-06, "loss": 2.0922, "step": 33047 }, { "epoch": 0.81, "learning_rate": 1.7494945068863113e-06, "loss": 1.913, "step": 33048 }, { "epoch": 0.81, "learning_rate": 1.7490434290479107e-06, "loss": 1.9214, "step": 33049 }, { "epoch": 0.81, "learning_rate": 1.748592403795545e-06, "loss": 2.0365, "step": 33050 }, { "epoch": 0.81, "learning_rate": 1.7481414311320921e-06, "loss": 1.8885, "step": 33051 }, { "epoch": 0.81, "learning_rate": 1.7476905110604203e-06, "loss": 1.9032, "step": 33052 }, { "epoch": 0.81, "learning_rate": 1.7472396435834082e-06, "loss": 2.0658, "step": 33053 }, { "epoch": 0.81, "learning_rate": 1.7467888287039247e-06, "loss": 1.9087, "step": 33054 }, { "epoch": 0.81, "learning_rate": 1.7463380664248465e-06, "loss": 1.9843, "step": 33055 }, { "epoch": 0.81, "learning_rate": 1.7458873567490475e-06, "loss": 1.9267, "step": 33056 }, { "epoch": 0.81, "learning_rate": 1.745436699679397e-06, "loss": 1.868, "step": 33057 }, { "epoch": 0.81, "learning_rate": 1.7449860952187692e-06, "loss": 2.003, "step": 33058 }, { "epoch": 0.81, "learning_rate": 1.744535543370034e-06, "loss": 1.8058, "step": 33059 }, { "epoch": 0.81, "learning_rate": 1.7440850441360635e-06, "loss": 1.8215, "step": 33060 }, { "epoch": 0.81, "learning_rate": 1.74363459751973e-06, "loss": 1.863, "step": 33061 }, { "epoch": 0.81, "learning_rate": 1.7431842035239033e-06, "loss": 1.8574, "step": 33062 }, { "epoch": 0.81, "learning_rate": 1.7427338621514578e-06, "loss": 1.9752, "step": 33063 }, { "epoch": 0.81, "learning_rate": 1.7422835734052568e-06, "loss": 1.8048, "step": 33064 }, { "epoch": 0.82, "learning_rate": 1.7418333372881747e-06, "loss": 1.9634, "step": 33065 }, { "epoch": 0.82, "learning_rate": 1.7413831538030823e-06, "loss": 1.9834, "step": 33066 }, { "epoch": 0.82, "learning_rate": 1.740933022952843e-06, "loss": 2.032, "step": 33067 }, { "epoch": 0.82, "learning_rate": 1.7404829447403327e-06, "loss": 1.9031, "step": 33068 }, { "epoch": 0.82, "learning_rate": 1.740032919168414e-06, "loss": 2.0179, "step": 33069 }, { "epoch": 0.82, "learning_rate": 1.739582946239956e-06, "loss": 1.9173, "step": 33070 }, { "epoch": 0.82, "learning_rate": 1.7391330259578321e-06, "loss": 1.9031, "step": 33071 }, { "epoch": 0.82, "learning_rate": 1.738683158324902e-06, "loss": 2.1759, "step": 33072 }, { "epoch": 0.82, "learning_rate": 1.738233343344037e-06, "loss": 1.9663, "step": 33073 }, { "epoch": 0.82, "learning_rate": 1.7377835810181042e-06, "loss": 1.9997, "step": 33074 }, { "epoch": 0.82, "learning_rate": 1.7373338713499688e-06, "loss": 2.0191, "step": 33075 }, { "epoch": 0.82, "learning_rate": 1.736884214342499e-06, "loss": 1.8951, "step": 33076 }, { "epoch": 0.82, "learning_rate": 1.7364346099985562e-06, "loss": 1.8794, "step": 33077 }, { "epoch": 0.82, "learning_rate": 1.7359850583210092e-06, "loss": 1.8965, "step": 33078 }, { "epoch": 0.82, "learning_rate": 1.735535559312722e-06, "loss": 1.7921, "step": 33079 }, { "epoch": 0.82, "learning_rate": 1.7350861129765605e-06, "loss": 1.9173, "step": 33080 }, { "epoch": 0.82, "learning_rate": 1.7346367193153912e-06, "loss": 1.9611, "step": 33081 }, { "epoch": 0.82, "learning_rate": 1.734187378332073e-06, "loss": 1.7998, "step": 33082 }, { "epoch": 0.82, "learning_rate": 1.733738090029472e-06, "loss": 1.9521, "step": 33083 }, { "epoch": 0.82, "learning_rate": 1.7332888544104542e-06, "loss": 2.0018, "step": 33084 }, { "epoch": 0.82, "learning_rate": 1.7328396714778794e-06, "loss": 2.0249, "step": 33085 }, { "epoch": 0.82, "learning_rate": 1.7323905412346131e-06, "loss": 1.8668, "step": 33086 }, { "epoch": 0.82, "learning_rate": 1.7319414636835141e-06, "loss": 2.0001, "step": 33087 }, { "epoch": 0.82, "learning_rate": 1.7314924388274467e-06, "loss": 1.9897, "step": 33088 }, { "epoch": 0.82, "learning_rate": 1.7310434666692754e-06, "loss": 1.9308, "step": 33089 }, { "epoch": 0.82, "learning_rate": 1.7305945472118557e-06, "loss": 1.9257, "step": 33090 }, { "epoch": 0.82, "learning_rate": 1.730145680458054e-06, "loss": 1.8118, "step": 33091 }, { "epoch": 0.82, "learning_rate": 1.7296968664107271e-06, "loss": 1.9289, "step": 33092 }, { "epoch": 0.82, "learning_rate": 1.7292481050727373e-06, "loss": 2.0152, "step": 33093 }, { "epoch": 0.82, "learning_rate": 1.7287993964469473e-06, "loss": 1.9068, "step": 33094 }, { "epoch": 0.82, "learning_rate": 1.7283507405362122e-06, "loss": 1.9411, "step": 33095 }, { "epoch": 0.82, "learning_rate": 1.7279021373433925e-06, "loss": 1.8978, "step": 33096 }, { "epoch": 0.82, "learning_rate": 1.7274535868713483e-06, "loss": 1.8865, "step": 33097 }, { "epoch": 0.82, "learning_rate": 1.7270050891229395e-06, "loss": 1.8646, "step": 33098 }, { "epoch": 0.82, "learning_rate": 1.7265566441010252e-06, "loss": 1.9751, "step": 33099 }, { "epoch": 0.82, "learning_rate": 1.7261082518084583e-06, "loss": 2.0111, "step": 33100 }, { "epoch": 0.82, "learning_rate": 1.7256599122481033e-06, "loss": 1.7104, "step": 33101 }, { "epoch": 0.82, "learning_rate": 1.7252116254228113e-06, "loss": 1.9185, "step": 33102 }, { "epoch": 0.82, "learning_rate": 1.7247633913354423e-06, "loss": 1.9106, "step": 33103 }, { "epoch": 0.82, "learning_rate": 1.724315209988856e-06, "loss": 1.88, "step": 33104 }, { "epoch": 0.82, "learning_rate": 1.7238670813859025e-06, "loss": 1.9836, "step": 33105 }, { "epoch": 0.82, "learning_rate": 1.7234190055294452e-06, "loss": 1.8868, "step": 33106 }, { "epoch": 0.82, "learning_rate": 1.7229709824223317e-06, "loss": 1.938, "step": 33107 }, { "epoch": 0.82, "learning_rate": 1.722523012067423e-06, "loss": 1.7544, "step": 33108 }, { "epoch": 0.82, "learning_rate": 1.7220750944675735e-06, "loss": 1.8489, "step": 33109 }, { "epoch": 0.82, "learning_rate": 1.7216272296256365e-06, "loss": 2.0909, "step": 33110 }, { "epoch": 0.82, "learning_rate": 1.721179417544465e-06, "loss": 1.9066, "step": 33111 }, { "epoch": 0.82, "learning_rate": 1.7207316582269197e-06, "loss": 1.8383, "step": 33112 }, { "epoch": 0.82, "learning_rate": 1.7202839516758452e-06, "loss": 1.9194, "step": 33113 }, { "epoch": 0.82, "learning_rate": 1.7198362978941008e-06, "loss": 1.985, "step": 33114 }, { "epoch": 0.82, "learning_rate": 1.7193886968845375e-06, "loss": 1.9991, "step": 33115 }, { "epoch": 0.82, "learning_rate": 1.7189411486500107e-06, "loss": 1.8417, "step": 33116 }, { "epoch": 0.82, "learning_rate": 1.7184936531933695e-06, "loss": 1.8911, "step": 33117 }, { "epoch": 0.82, "learning_rate": 1.7180462105174656e-06, "loss": 2.1144, "step": 33118 }, { "epoch": 0.82, "learning_rate": 1.7175988206251559e-06, "loss": 1.7444, "step": 33119 }, { "epoch": 0.82, "learning_rate": 1.717151483519285e-06, "loss": 1.9062, "step": 33120 }, { "epoch": 0.82, "learning_rate": 1.7167041992027089e-06, "loss": 2.0163, "step": 33121 }, { "epoch": 0.82, "learning_rate": 1.7162569676782748e-06, "loss": 1.8613, "step": 33122 }, { "epoch": 0.82, "learning_rate": 1.7158097889488345e-06, "loss": 1.9206, "step": 33123 }, { "epoch": 0.82, "learning_rate": 1.7153626630172404e-06, "loss": 1.8476, "step": 33124 }, { "epoch": 0.82, "learning_rate": 1.7149155898863378e-06, "loss": 1.8385, "step": 33125 }, { "epoch": 0.82, "learning_rate": 1.7144685695589802e-06, "loss": 1.9195, "step": 33126 }, { "epoch": 0.82, "learning_rate": 1.714021602038012e-06, "loss": 1.9382, "step": 33127 }, { "epoch": 0.82, "learning_rate": 1.7135746873262849e-06, "loss": 1.8746, "step": 33128 }, { "epoch": 0.82, "learning_rate": 1.713127825426647e-06, "loss": 1.8989, "step": 33129 }, { "epoch": 0.82, "learning_rate": 1.7126810163419484e-06, "loss": 2.1082, "step": 33130 }, { "epoch": 0.82, "learning_rate": 1.712234260075032e-06, "loss": 2.0174, "step": 33131 }, { "epoch": 0.82, "learning_rate": 1.7117875566287467e-06, "loss": 1.8596, "step": 33132 }, { "epoch": 0.82, "learning_rate": 1.7113409060059417e-06, "loss": 1.9193, "step": 33133 }, { "epoch": 0.82, "learning_rate": 1.710894308209463e-06, "loss": 1.8677, "step": 33134 }, { "epoch": 0.82, "learning_rate": 1.7104477632421556e-06, "loss": 1.8728, "step": 33135 }, { "epoch": 0.82, "learning_rate": 1.7100012711068648e-06, "loss": 1.8956, "step": 33136 }, { "epoch": 0.82, "learning_rate": 1.7095548318064402e-06, "loss": 1.9182, "step": 33137 }, { "epoch": 0.82, "learning_rate": 1.7091084453437223e-06, "loss": 1.9909, "step": 33138 }, { "epoch": 0.82, "learning_rate": 1.7086621117215606e-06, "loss": 2.0384, "step": 33139 }, { "epoch": 0.82, "learning_rate": 1.7082158309427954e-06, "loss": 1.9403, "step": 33140 }, { "epoch": 0.82, "learning_rate": 1.707769603010273e-06, "loss": 1.8003, "step": 33141 }, { "epoch": 0.82, "learning_rate": 1.7073234279268392e-06, "loss": 1.907, "step": 33142 }, { "epoch": 0.82, "learning_rate": 1.7068773056953326e-06, "loss": 1.6711, "step": 33143 }, { "epoch": 0.82, "learning_rate": 1.706431236318603e-06, "loss": 1.8671, "step": 33144 }, { "epoch": 0.82, "learning_rate": 1.7059852197994874e-06, "loss": 2.0465, "step": 33145 }, { "epoch": 0.82, "learning_rate": 1.7055392561408301e-06, "loss": 1.7376, "step": 33146 }, { "epoch": 0.82, "learning_rate": 1.7050933453454777e-06, "loss": 1.9575, "step": 33147 }, { "epoch": 0.82, "learning_rate": 1.7046474874162656e-06, "loss": 2.013, "step": 33148 }, { "epoch": 0.82, "learning_rate": 1.7042016823560381e-06, "loss": 2.045, "step": 33149 }, { "epoch": 0.82, "learning_rate": 1.7037559301676377e-06, "loss": 1.9068, "step": 33150 }, { "epoch": 0.82, "learning_rate": 1.703310230853904e-06, "loss": 1.8893, "step": 33151 }, { "epoch": 0.82, "learning_rate": 1.7028645844176805e-06, "loss": 1.8449, "step": 33152 }, { "epoch": 0.82, "learning_rate": 1.7024189908618017e-06, "loss": 1.9594, "step": 33153 }, { "epoch": 0.82, "learning_rate": 1.701973450189115e-06, "loss": 1.8991, "step": 33154 }, { "epoch": 0.82, "learning_rate": 1.7015279624024518e-06, "loss": 1.9999, "step": 33155 }, { "epoch": 0.82, "learning_rate": 1.701082527504656e-06, "loss": 1.9025, "step": 33156 }, { "epoch": 0.82, "learning_rate": 1.700637145498568e-06, "loss": 1.808, "step": 33157 }, { "epoch": 0.82, "learning_rate": 1.7001918163870212e-06, "loss": 1.7755, "step": 33158 }, { "epoch": 0.82, "learning_rate": 1.6997465401728596e-06, "loss": 1.7701, "step": 33159 }, { "epoch": 0.82, "learning_rate": 1.6993013168589167e-06, "loss": 2.0104, "step": 33160 }, { "epoch": 0.82, "learning_rate": 1.6988561464480301e-06, "loss": 2.0504, "step": 33161 }, { "epoch": 0.82, "learning_rate": 1.6984110289430421e-06, "loss": 1.837, "step": 33162 }, { "epoch": 0.82, "learning_rate": 1.6979659643467838e-06, "loss": 1.8795, "step": 33163 }, { "epoch": 0.82, "learning_rate": 1.6975209526620929e-06, "loss": 1.953, "step": 33164 }, { "epoch": 0.82, "learning_rate": 1.6970759938918101e-06, "loss": 1.9347, "step": 33165 }, { "epoch": 0.82, "learning_rate": 1.6966310880387648e-06, "loss": 1.894, "step": 33166 }, { "epoch": 0.82, "learning_rate": 1.6961862351057956e-06, "loss": 1.9364, "step": 33167 }, { "epoch": 0.82, "learning_rate": 1.695741435095738e-06, "loss": 2.0219, "step": 33168 }, { "epoch": 0.82, "learning_rate": 1.6952966880114285e-06, "loss": 2.1399, "step": 33169 }, { "epoch": 0.82, "learning_rate": 1.6948519938556973e-06, "loss": 1.6996, "step": 33170 }, { "epoch": 0.82, "learning_rate": 1.694407352631381e-06, "loss": 1.797, "step": 33171 }, { "epoch": 0.82, "learning_rate": 1.693962764341316e-06, "loss": 1.8951, "step": 33172 }, { "epoch": 0.82, "learning_rate": 1.6935182289883301e-06, "loss": 1.9309, "step": 33173 }, { "epoch": 0.82, "learning_rate": 1.6930737465752633e-06, "loss": 1.9888, "step": 33174 }, { "epoch": 0.82, "learning_rate": 1.6926293171049413e-06, "loss": 2.0235, "step": 33175 }, { "epoch": 0.82, "learning_rate": 1.6921849405801992e-06, "loss": 1.9037, "step": 33176 }, { "epoch": 0.82, "learning_rate": 1.691740617003873e-06, "loss": 1.9441, "step": 33177 }, { "epoch": 0.82, "learning_rate": 1.69129634637879e-06, "loss": 1.8002, "step": 33178 }, { "epoch": 0.82, "learning_rate": 1.6908521287077839e-06, "loss": 1.96, "step": 33179 }, { "epoch": 0.82, "learning_rate": 1.6904079639936843e-06, "loss": 1.8788, "step": 33180 }, { "epoch": 0.82, "learning_rate": 1.6899638522393214e-06, "loss": 1.9101, "step": 33181 }, { "epoch": 0.82, "learning_rate": 1.689519793447527e-06, "loss": 1.7734, "step": 33182 }, { "epoch": 0.82, "learning_rate": 1.689075787621135e-06, "loss": 1.8223, "step": 33183 }, { "epoch": 0.82, "learning_rate": 1.6886318347629682e-06, "loss": 1.7483, "step": 33184 }, { "epoch": 0.82, "learning_rate": 1.688187934875859e-06, "loss": 1.9405, "step": 33185 }, { "epoch": 0.82, "learning_rate": 1.6877440879626372e-06, "loss": 1.8649, "step": 33186 }, { "epoch": 0.82, "learning_rate": 1.687300294026134e-06, "loss": 1.7125, "step": 33187 }, { "epoch": 0.82, "learning_rate": 1.6868565530691716e-06, "loss": 1.8761, "step": 33188 }, { "epoch": 0.82, "learning_rate": 1.6864128650945855e-06, "loss": 1.8668, "step": 33189 }, { "epoch": 0.82, "learning_rate": 1.6859692301051956e-06, "loss": 1.9634, "step": 33190 }, { "epoch": 0.82, "learning_rate": 1.685525648103834e-06, "loss": 1.9554, "step": 33191 }, { "epoch": 0.82, "learning_rate": 1.6850821190933297e-06, "loss": 1.9663, "step": 33192 }, { "epoch": 0.82, "learning_rate": 1.6846386430765038e-06, "loss": 2.0585, "step": 33193 }, { "epoch": 0.82, "learning_rate": 1.6841952200561862e-06, "loss": 2.1038, "step": 33194 }, { "epoch": 0.82, "learning_rate": 1.6837518500352045e-06, "loss": 1.8999, "step": 33195 }, { "epoch": 0.82, "learning_rate": 1.6833085330163801e-06, "loss": 1.8866, "step": 33196 }, { "epoch": 0.82, "learning_rate": 1.6828652690025438e-06, "loss": 1.9625, "step": 33197 }, { "epoch": 0.82, "learning_rate": 1.6824220579965144e-06, "loss": 1.947, "step": 33198 }, { "epoch": 0.82, "learning_rate": 1.68197890000112e-06, "loss": 2.0219, "step": 33199 }, { "epoch": 0.82, "learning_rate": 1.6815357950191858e-06, "loss": 2.0276, "step": 33200 }, { "epoch": 0.82, "learning_rate": 1.6810927430535362e-06, "loss": 1.8672, "step": 33201 }, { "epoch": 0.82, "learning_rate": 1.6806497441069914e-06, "loss": 1.9382, "step": 33202 }, { "epoch": 0.82, "learning_rate": 1.6802067981823767e-06, "loss": 1.6931, "step": 33203 }, { "epoch": 0.82, "learning_rate": 1.6797639052825154e-06, "loss": 1.9807, "step": 33204 }, { "epoch": 0.82, "learning_rate": 1.6793210654102331e-06, "loss": 1.8696, "step": 33205 }, { "epoch": 0.82, "learning_rate": 1.6788782785683478e-06, "loss": 1.8592, "step": 33206 }, { "epoch": 0.82, "learning_rate": 1.6784355447596834e-06, "loss": 1.9198, "step": 33207 }, { "epoch": 0.82, "learning_rate": 1.6779928639870601e-06, "loss": 1.8366, "step": 33208 }, { "epoch": 0.82, "learning_rate": 1.6775502362533003e-06, "loss": 1.9507, "step": 33209 }, { "epoch": 0.82, "learning_rate": 1.677107661561227e-06, "loss": 1.8581, "step": 33210 }, { "epoch": 0.82, "learning_rate": 1.6766651399136568e-06, "loss": 1.9878, "step": 33211 }, { "epoch": 0.82, "learning_rate": 1.6762226713134144e-06, "loss": 1.918, "step": 33212 }, { "epoch": 0.82, "learning_rate": 1.6757802557633151e-06, "loss": 1.9838, "step": 33213 }, { "epoch": 0.82, "learning_rate": 1.6753378932661812e-06, "loss": 1.9061, "step": 33214 }, { "epoch": 0.82, "learning_rate": 1.6748955838248348e-06, "loss": 1.9358, "step": 33215 }, { "epoch": 0.82, "learning_rate": 1.6744533274420883e-06, "loss": 1.8317, "step": 33216 }, { "epoch": 0.82, "learning_rate": 1.6740111241207646e-06, "loss": 1.9787, "step": 33217 }, { "epoch": 0.82, "learning_rate": 1.6735689738636818e-06, "loss": 1.9217, "step": 33218 }, { "epoch": 0.82, "learning_rate": 1.6731268766736586e-06, "loss": 2.0002, "step": 33219 }, { "epoch": 0.82, "learning_rate": 1.6726848325535105e-06, "loss": 1.8249, "step": 33220 }, { "epoch": 0.82, "learning_rate": 1.6722428415060543e-06, "loss": 1.8699, "step": 33221 }, { "epoch": 0.82, "learning_rate": 1.671800903534112e-06, "loss": 1.7574, "step": 33222 }, { "epoch": 0.82, "learning_rate": 1.6713590186404937e-06, "loss": 1.8648, "step": 33223 }, { "epoch": 0.82, "learning_rate": 1.6709171868280194e-06, "loss": 1.8691, "step": 33224 }, { "epoch": 0.82, "learning_rate": 1.6704754080995055e-06, "loss": 1.909, "step": 33225 }, { "epoch": 0.82, "learning_rate": 1.6700336824577644e-06, "loss": 1.9244, "step": 33226 }, { "epoch": 0.82, "learning_rate": 1.669592009905615e-06, "loss": 1.7398, "step": 33227 }, { "epoch": 0.82, "learning_rate": 1.6691503904458695e-06, "loss": 1.994, "step": 33228 }, { "epoch": 0.82, "learning_rate": 1.6687088240813432e-06, "loss": 1.9393, "step": 33229 }, { "epoch": 0.82, "learning_rate": 1.6682673108148518e-06, "loss": 1.9213, "step": 33230 }, { "epoch": 0.82, "learning_rate": 1.6678258506492072e-06, "loss": 1.9277, "step": 33231 }, { "epoch": 0.82, "learning_rate": 1.6673844435872254e-06, "loss": 2.0195, "step": 33232 }, { "epoch": 0.82, "learning_rate": 1.6669430896317163e-06, "loss": 1.9941, "step": 33233 }, { "epoch": 0.82, "learning_rate": 1.6665017887854939e-06, "loss": 2.0194, "step": 33234 }, { "epoch": 0.82, "learning_rate": 1.6660605410513708e-06, "loss": 1.7884, "step": 33235 }, { "epoch": 0.82, "learning_rate": 1.6656193464321614e-06, "loss": 1.9161, "step": 33236 }, { "epoch": 0.82, "learning_rate": 1.6651782049306775e-06, "loss": 1.9185, "step": 33237 }, { "epoch": 0.82, "learning_rate": 1.6647371165497262e-06, "loss": 1.8703, "step": 33238 }, { "epoch": 0.82, "learning_rate": 1.6642960812921228e-06, "loss": 1.923, "step": 33239 }, { "epoch": 0.82, "learning_rate": 1.6638550991606783e-06, "loss": 1.8945, "step": 33240 }, { "epoch": 0.82, "learning_rate": 1.6634141701582007e-06, "loss": 1.9269, "step": 33241 }, { "epoch": 0.82, "learning_rate": 1.6629732942875032e-06, "loss": 1.8016, "step": 33242 }, { "epoch": 0.82, "learning_rate": 1.6625324715513913e-06, "loss": 1.9795, "step": 33243 }, { "epoch": 0.82, "learning_rate": 1.6620917019526772e-06, "loss": 2.0223, "step": 33244 }, { "epoch": 0.82, "learning_rate": 1.661650985494173e-06, "loss": 1.9761, "step": 33245 }, { "epoch": 0.82, "learning_rate": 1.6612103221786812e-06, "loss": 2.0082, "step": 33246 }, { "epoch": 0.82, "learning_rate": 1.6607697120090161e-06, "loss": 1.8997, "step": 33247 }, { "epoch": 0.82, "learning_rate": 1.6603291549879808e-06, "loss": 1.9724, "step": 33248 }, { "epoch": 0.82, "learning_rate": 1.6598886511183853e-06, "loss": 1.9162, "step": 33249 }, { "epoch": 0.82, "learning_rate": 1.659448200403041e-06, "loss": 2.1151, "step": 33250 }, { "epoch": 0.82, "learning_rate": 1.6590078028447488e-06, "loss": 2.0109, "step": 33251 }, { "epoch": 0.82, "learning_rate": 1.6585674584463173e-06, "loss": 1.8871, "step": 33252 }, { "epoch": 0.82, "learning_rate": 1.6581271672105536e-06, "loss": 1.9532, "step": 33253 }, { "epoch": 0.82, "learning_rate": 1.6576869291402675e-06, "loss": 1.9839, "step": 33254 }, { "epoch": 0.82, "learning_rate": 1.6572467442382578e-06, "loss": 1.8691, "step": 33255 }, { "epoch": 0.82, "learning_rate": 1.6568066125073346e-06, "loss": 1.6954, "step": 33256 }, { "epoch": 0.82, "learning_rate": 1.656366533950301e-06, "loss": 2.1488, "step": 33257 }, { "epoch": 0.82, "learning_rate": 1.655926508569965e-06, "loss": 1.9938, "step": 33258 }, { "epoch": 0.82, "learning_rate": 1.6554865363691263e-06, "loss": 1.9932, "step": 33259 }, { "epoch": 0.82, "learning_rate": 1.6550466173505942e-06, "loss": 1.9382, "step": 33260 }, { "epoch": 0.82, "learning_rate": 1.654606751517166e-06, "loss": 1.9096, "step": 33261 }, { "epoch": 0.82, "learning_rate": 1.6541669388716497e-06, "loss": 1.9536, "step": 33262 }, { "epoch": 0.82, "learning_rate": 1.6537271794168496e-06, "loss": 2.1023, "step": 33263 }, { "epoch": 0.82, "learning_rate": 1.6532874731555636e-06, "loss": 1.9261, "step": 33264 }, { "epoch": 0.82, "learning_rate": 1.6528478200905983e-06, "loss": 1.8618, "step": 33265 }, { "epoch": 0.82, "learning_rate": 1.6524082202247526e-06, "loss": 1.9816, "step": 33266 }, { "epoch": 0.82, "learning_rate": 1.6519686735608298e-06, "loss": 2.1804, "step": 33267 }, { "epoch": 0.82, "learning_rate": 1.6515291801016331e-06, "loss": 1.9881, "step": 33268 }, { "epoch": 0.82, "learning_rate": 1.6510897398499593e-06, "loss": 1.9009, "step": 33269 }, { "epoch": 0.82, "learning_rate": 1.6506503528086116e-06, "loss": 1.8527, "step": 33270 }, { "epoch": 0.82, "learning_rate": 1.6502110189803899e-06, "loss": 1.9559, "step": 33271 }, { "epoch": 0.82, "learning_rate": 1.6497717383680977e-06, "loss": 1.9019, "step": 33272 }, { "epoch": 0.82, "learning_rate": 1.6493325109745284e-06, "loss": 1.8882, "step": 33273 }, { "epoch": 0.82, "learning_rate": 1.648893336802484e-06, "loss": 1.8626, "step": 33274 }, { "epoch": 0.82, "learning_rate": 1.6484542158547679e-06, "loss": 1.6782, "step": 33275 }, { "epoch": 0.82, "learning_rate": 1.6480151481341711e-06, "loss": 1.8439, "step": 33276 }, { "epoch": 0.82, "learning_rate": 1.6475761336434958e-06, "loss": 1.7279, "step": 33277 }, { "epoch": 0.82, "learning_rate": 1.6471371723855423e-06, "loss": 1.7891, "step": 33278 }, { "epoch": 0.82, "learning_rate": 1.6466982643631024e-06, "loss": 1.8109, "step": 33279 }, { "epoch": 0.82, "learning_rate": 1.6462594095789797e-06, "loss": 1.7092, "step": 33280 }, { "epoch": 0.82, "learning_rate": 1.6458206080359663e-06, "loss": 1.8707, "step": 33281 }, { "epoch": 0.82, "learning_rate": 1.645381859736861e-06, "loss": 2.0502, "step": 33282 }, { "epoch": 0.82, "learning_rate": 1.6449431646844615e-06, "loss": 1.9276, "step": 33283 }, { "epoch": 0.82, "learning_rate": 1.6445045228815593e-06, "loss": 1.8674, "step": 33284 }, { "epoch": 0.82, "learning_rate": 1.6440659343309563e-06, "loss": 1.8415, "step": 33285 }, { "epoch": 0.82, "learning_rate": 1.64362739903544e-06, "loss": 2.0678, "step": 33286 }, { "epoch": 0.82, "learning_rate": 1.643188916997811e-06, "loss": 1.6911, "step": 33287 }, { "epoch": 0.82, "learning_rate": 1.6427504882208623e-06, "loss": 1.8624, "step": 33288 }, { "epoch": 0.82, "learning_rate": 1.6423121127073882e-06, "loss": 1.9147, "step": 33289 }, { "epoch": 0.82, "learning_rate": 1.6418737904601855e-06, "loss": 1.9288, "step": 33290 }, { "epoch": 0.82, "learning_rate": 1.641435521482042e-06, "loss": 1.9264, "step": 33291 }, { "epoch": 0.82, "learning_rate": 1.6409973057757533e-06, "loss": 1.9386, "step": 33292 }, { "epoch": 0.82, "learning_rate": 1.640559143344116e-06, "loss": 1.8923, "step": 33293 }, { "epoch": 0.82, "learning_rate": 1.6401210341899175e-06, "loss": 1.8154, "step": 33294 }, { "epoch": 0.82, "learning_rate": 1.6396829783159541e-06, "loss": 1.9798, "step": 33295 }, { "epoch": 0.82, "learning_rate": 1.6392449757250129e-06, "loss": 1.9386, "step": 33296 }, { "epoch": 0.82, "learning_rate": 1.6388070264198886e-06, "loss": 1.9885, "step": 33297 }, { "epoch": 0.82, "learning_rate": 1.6383691304033743e-06, "loss": 1.9432, "step": 33298 }, { "epoch": 0.82, "learning_rate": 1.637931287678256e-06, "loss": 2.0918, "step": 33299 }, { "epoch": 0.82, "learning_rate": 1.6374934982473289e-06, "loss": 1.921, "step": 33300 }, { "epoch": 0.82, "learning_rate": 1.6370557621133787e-06, "loss": 1.9536, "step": 33301 }, { "epoch": 0.82, "learning_rate": 1.6366180792791975e-06, "loss": 1.9803, "step": 33302 }, { "epoch": 0.82, "learning_rate": 1.6361804497475775e-06, "loss": 1.9583, "step": 33303 }, { "epoch": 0.82, "learning_rate": 1.6357428735213033e-06, "loss": 1.9843, "step": 33304 }, { "epoch": 0.82, "learning_rate": 1.6353053506031647e-06, "loss": 1.8404, "step": 33305 }, { "epoch": 0.82, "learning_rate": 1.6348678809959517e-06, "loss": 1.8614, "step": 33306 }, { "epoch": 0.82, "learning_rate": 1.6344304647024511e-06, "loss": 1.9406, "step": 33307 }, { "epoch": 0.82, "learning_rate": 1.633993101725454e-06, "loss": 1.9141, "step": 33308 }, { "epoch": 0.82, "learning_rate": 1.6335557920677436e-06, "loss": 1.9164, "step": 33309 }, { "epoch": 0.82, "learning_rate": 1.63311853573211e-06, "loss": 1.7942, "step": 33310 }, { "epoch": 0.82, "learning_rate": 1.6326813327213365e-06, "loss": 1.9098, "step": 33311 }, { "epoch": 0.82, "learning_rate": 1.632244183038213e-06, "loss": 1.9993, "step": 33312 }, { "epoch": 0.82, "learning_rate": 1.6318070866855252e-06, "loss": 1.9921, "step": 33313 }, { "epoch": 0.82, "learning_rate": 1.6313700436660561e-06, "loss": 1.9758, "step": 33314 }, { "epoch": 0.82, "learning_rate": 1.630933053982593e-06, "loss": 1.9964, "step": 33315 }, { "epoch": 0.82, "learning_rate": 1.630496117637923e-06, "loss": 1.793, "step": 33316 }, { "epoch": 0.82, "learning_rate": 1.6300592346348264e-06, "loss": 2.0531, "step": 33317 }, { "epoch": 0.82, "learning_rate": 1.6296224049760934e-06, "loss": 1.8503, "step": 33318 }, { "epoch": 0.82, "learning_rate": 1.6291856286645013e-06, "loss": 1.9897, "step": 33319 }, { "epoch": 0.82, "learning_rate": 1.6287489057028373e-06, "loss": 1.8965, "step": 33320 }, { "epoch": 0.82, "learning_rate": 1.6283122360938863e-06, "loss": 1.9233, "step": 33321 }, { "epoch": 0.82, "learning_rate": 1.627875619840429e-06, "loss": 2.0599, "step": 33322 }, { "epoch": 0.82, "learning_rate": 1.627439056945247e-06, "loss": 1.89, "step": 33323 }, { "epoch": 0.82, "learning_rate": 1.627002547411125e-06, "loss": 1.9786, "step": 33324 }, { "epoch": 0.82, "learning_rate": 1.6265660912408455e-06, "loss": 1.8281, "step": 33325 }, { "epoch": 0.82, "learning_rate": 1.6261296884371902e-06, "loss": 1.9409, "step": 33326 }, { "epoch": 0.82, "learning_rate": 1.6256933390029373e-06, "loss": 1.9096, "step": 33327 }, { "epoch": 0.82, "learning_rate": 1.6252570429408731e-06, "loss": 2.098, "step": 33328 }, { "epoch": 0.82, "learning_rate": 1.6248208002537712e-06, "loss": 2.0386, "step": 33329 }, { "epoch": 0.82, "learning_rate": 1.6243846109444173e-06, "loss": 1.8865, "step": 33330 }, { "epoch": 0.82, "learning_rate": 1.623948475015591e-06, "loss": 1.891, "step": 33331 }, { "epoch": 0.82, "learning_rate": 1.623512392470069e-06, "loss": 1.8344, "step": 33332 }, { "epoch": 0.82, "learning_rate": 1.623076363310634e-06, "loss": 1.8331, "step": 33333 }, { "epoch": 0.82, "learning_rate": 1.622640387540061e-06, "loss": 1.9706, "step": 33334 }, { "epoch": 0.82, "learning_rate": 1.6222044651611312e-06, "loss": 1.9045, "step": 33335 }, { "epoch": 0.82, "learning_rate": 1.6217685961766262e-06, "loss": 1.8218, "step": 33336 }, { "epoch": 0.82, "learning_rate": 1.6213327805893163e-06, "loss": 1.8648, "step": 33337 }, { "epoch": 0.82, "learning_rate": 1.6208970184019857e-06, "loss": 2.056, "step": 33338 }, { "epoch": 0.82, "learning_rate": 1.6204613096174071e-06, "loss": 1.8997, "step": 33339 }, { "epoch": 0.82, "learning_rate": 1.6200256542383596e-06, "loss": 1.952, "step": 33340 }, { "epoch": 0.82, "learning_rate": 1.6195900522676188e-06, "loss": 1.953, "step": 33341 }, { "epoch": 0.82, "learning_rate": 1.6191545037079625e-06, "loss": 1.846, "step": 33342 }, { "epoch": 0.82, "learning_rate": 1.6187190085621673e-06, "loss": 1.8221, "step": 33343 }, { "epoch": 0.82, "learning_rate": 1.6182835668330054e-06, "loss": 1.7926, "step": 33344 }, { "epoch": 0.82, "learning_rate": 1.6178481785232536e-06, "loss": 1.6941, "step": 33345 }, { "epoch": 0.82, "learning_rate": 1.6174128436356896e-06, "loss": 1.7505, "step": 33346 }, { "epoch": 0.82, "learning_rate": 1.6169775621730832e-06, "loss": 2.0203, "step": 33347 }, { "epoch": 0.82, "learning_rate": 1.6165423341382137e-06, "loss": 1.8865, "step": 33348 }, { "epoch": 0.82, "learning_rate": 1.6161071595338486e-06, "loss": 1.9749, "step": 33349 }, { "epoch": 0.82, "learning_rate": 1.6156720383627644e-06, "loss": 2.0289, "step": 33350 }, { "epoch": 0.82, "learning_rate": 1.6152369706277383e-06, "loss": 1.8086, "step": 33351 }, { "epoch": 0.82, "learning_rate": 1.6148019563315364e-06, "loss": 2.0172, "step": 33352 }, { "epoch": 0.82, "learning_rate": 1.614366995476937e-06, "loss": 1.7836, "step": 33353 }, { "epoch": 0.82, "learning_rate": 1.6139320880667064e-06, "loss": 1.8109, "step": 33354 }, { "epoch": 0.82, "learning_rate": 1.6134972341036203e-06, "loss": 2.0198, "step": 33355 }, { "epoch": 0.82, "learning_rate": 1.613062433590451e-06, "loss": 2.0229, "step": 33356 }, { "epoch": 0.82, "learning_rate": 1.6126276865299662e-06, "loss": 1.8932, "step": 33357 }, { "epoch": 0.82, "learning_rate": 1.6121929929249392e-06, "loss": 1.7808, "step": 33358 }, { "epoch": 0.82, "learning_rate": 1.6117583527781388e-06, "loss": 2.0698, "step": 33359 }, { "epoch": 0.82, "learning_rate": 1.6113237660923365e-06, "loss": 1.981, "step": 33360 }, { "epoch": 0.82, "learning_rate": 1.6108892328703041e-06, "loss": 1.9838, "step": 33361 }, { "epoch": 0.82, "learning_rate": 1.6104547531148051e-06, "loss": 2.0421, "step": 33362 }, { "epoch": 0.82, "learning_rate": 1.6100203268286151e-06, "loss": 1.8715, "step": 33363 }, { "epoch": 0.82, "learning_rate": 1.6095859540144975e-06, "loss": 1.8497, "step": 33364 }, { "epoch": 0.82, "learning_rate": 1.6091516346752234e-06, "loss": 1.9099, "step": 33365 }, { "epoch": 0.82, "learning_rate": 1.6087173688135627e-06, "loss": 1.9141, "step": 33366 }, { "epoch": 0.82, "learning_rate": 1.6082831564322776e-06, "loss": 1.8538, "step": 33367 }, { "epoch": 0.82, "learning_rate": 1.6078489975341426e-06, "loss": 2.05, "step": 33368 }, { "epoch": 0.82, "learning_rate": 1.6074148921219167e-06, "loss": 1.8794, "step": 33369 }, { "epoch": 0.82, "learning_rate": 1.6069808401983721e-06, "loss": 1.8697, "step": 33370 }, { "epoch": 0.82, "learning_rate": 1.6065468417662755e-06, "loss": 2.0216, "step": 33371 }, { "epoch": 0.82, "learning_rate": 1.606112896828389e-06, "loss": 1.9355, "step": 33372 }, { "epoch": 0.82, "learning_rate": 1.6056790053874804e-06, "loss": 1.9238, "step": 33373 }, { "epoch": 0.82, "learning_rate": 1.6052451674463188e-06, "loss": 1.9268, "step": 33374 }, { "epoch": 0.82, "learning_rate": 1.6048113830076618e-06, "loss": 1.9218, "step": 33375 }, { "epoch": 0.82, "learning_rate": 1.6043776520742783e-06, "loss": 1.9716, "step": 33376 }, { "epoch": 0.82, "learning_rate": 1.6039439746489315e-06, "loss": 1.9204, "step": 33377 }, { "epoch": 0.82, "learning_rate": 1.6035103507343863e-06, "loss": 2.0523, "step": 33378 }, { "epoch": 0.82, "learning_rate": 1.603076780333408e-06, "loss": 1.9407, "step": 33379 }, { "epoch": 0.82, "learning_rate": 1.6026432634487553e-06, "loss": 1.9659, "step": 33380 }, { "epoch": 0.82, "learning_rate": 1.6022098000831965e-06, "loss": 1.8226, "step": 33381 }, { "epoch": 0.82, "learning_rate": 1.601776390239489e-06, "loss": 2.0561, "step": 33382 }, { "epoch": 0.82, "learning_rate": 1.6013430339203984e-06, "loss": 1.8819, "step": 33383 }, { "epoch": 0.82, "learning_rate": 1.600909731128687e-06, "loss": 1.8998, "step": 33384 }, { "epoch": 0.82, "learning_rate": 1.6004764818671126e-06, "loss": 1.8311, "step": 33385 }, { "epoch": 0.82, "learning_rate": 1.6000432861384419e-06, "loss": 1.8172, "step": 33386 }, { "epoch": 0.82, "learning_rate": 1.5996101439454293e-06, "loss": 1.9546, "step": 33387 }, { "epoch": 0.82, "learning_rate": 1.5991770552908404e-06, "loss": 1.9625, "step": 33388 }, { "epoch": 0.82, "learning_rate": 1.598744020177435e-06, "loss": 2.0271, "step": 33389 }, { "epoch": 0.82, "learning_rate": 1.5983110386079704e-06, "loss": 1.8488, "step": 33390 }, { "epoch": 0.82, "learning_rate": 1.5978781105852093e-06, "loss": 1.8261, "step": 33391 }, { "epoch": 0.82, "learning_rate": 1.5974452361119076e-06, "loss": 2.0329, "step": 33392 }, { "epoch": 0.82, "learning_rate": 1.597012415190825e-06, "loss": 1.909, "step": 33393 }, { "epoch": 0.82, "learning_rate": 1.5965796478247209e-06, "loss": 1.9626, "step": 33394 }, { "epoch": 0.82, "learning_rate": 1.5961469340163527e-06, "loss": 1.9167, "step": 33395 }, { "epoch": 0.82, "learning_rate": 1.5957142737684817e-06, "loss": 1.9418, "step": 33396 }, { "epoch": 0.82, "learning_rate": 1.59528166708386e-06, "loss": 1.9243, "step": 33397 }, { "epoch": 0.82, "learning_rate": 1.5948491139652478e-06, "loss": 1.9167, "step": 33398 }, { "epoch": 0.82, "learning_rate": 1.5944166144154039e-06, "loss": 2.0127, "step": 33399 }, { "epoch": 0.82, "learning_rate": 1.5939841684370794e-06, "loss": 1.8594, "step": 33400 }, { "epoch": 0.82, "learning_rate": 1.5935517760330354e-06, "loss": 2.0103, "step": 33401 }, { "epoch": 0.82, "learning_rate": 1.5931194372060233e-06, "loss": 1.7755, "step": 33402 }, { "epoch": 0.82, "learning_rate": 1.5926871519588005e-06, "loss": 1.8246, "step": 33403 }, { "epoch": 0.82, "learning_rate": 1.5922549202941251e-06, "loss": 1.8687, "step": 33404 }, { "epoch": 0.82, "learning_rate": 1.591822742214747e-06, "loss": 1.9105, "step": 33405 }, { "epoch": 0.82, "learning_rate": 1.591390617723424e-06, "loss": 1.8166, "step": 33406 }, { "epoch": 0.82, "learning_rate": 1.5909585468229072e-06, "loss": 1.9553, "step": 33407 }, { "epoch": 0.82, "learning_rate": 1.5905265295159522e-06, "loss": 1.9046, "step": 33408 }, { "epoch": 0.82, "learning_rate": 1.5900945658053146e-06, "loss": 2.0961, "step": 33409 }, { "epoch": 0.82, "learning_rate": 1.5896626556937422e-06, "loss": 1.9134, "step": 33410 }, { "epoch": 0.82, "learning_rate": 1.5892307991839906e-06, "loss": 1.8934, "step": 33411 }, { "epoch": 0.82, "learning_rate": 1.588798996278813e-06, "loss": 1.9212, "step": 33412 }, { "epoch": 0.82, "learning_rate": 1.5883672469809598e-06, "loss": 1.9647, "step": 33413 }, { "epoch": 0.82, "learning_rate": 1.587935551293185e-06, "loss": 1.9449, "step": 33414 }, { "epoch": 0.82, "learning_rate": 1.5875039092182375e-06, "loss": 1.8215, "step": 33415 }, { "epoch": 0.82, "learning_rate": 1.5870723207588701e-06, "loss": 2.0485, "step": 33416 }, { "epoch": 0.82, "learning_rate": 1.5866407859178312e-06, "loss": 1.8235, "step": 33417 }, { "epoch": 0.82, "learning_rate": 1.5862093046978722e-06, "loss": 1.835, "step": 33418 }, { "epoch": 0.82, "learning_rate": 1.5857778771017452e-06, "loss": 1.7839, "step": 33419 }, { "epoch": 0.82, "learning_rate": 1.5853465031321957e-06, "loss": 2.0153, "step": 33420 }, { "epoch": 0.82, "learning_rate": 1.5849151827919774e-06, "loss": 1.9298, "step": 33421 }, { "epoch": 0.82, "learning_rate": 1.5844839160838354e-06, "loss": 1.8567, "step": 33422 }, { "epoch": 0.82, "learning_rate": 1.5840527030105192e-06, "loss": 1.9107, "step": 33423 }, { "epoch": 0.82, "learning_rate": 1.5836215435747803e-06, "loss": 1.7998, "step": 33424 }, { "epoch": 0.82, "learning_rate": 1.5831904377793617e-06, "loss": 1.7733, "step": 33425 }, { "epoch": 0.82, "learning_rate": 1.5827593856270162e-06, "loss": 1.8942, "step": 33426 }, { "epoch": 0.82, "learning_rate": 1.5823283871204854e-06, "loss": 1.9183, "step": 33427 }, { "epoch": 0.82, "learning_rate": 1.58189744226252e-06, "loss": 1.8556, "step": 33428 }, { "epoch": 0.82, "learning_rate": 1.5814665510558647e-06, "loss": 1.9957, "step": 33429 }, { "epoch": 0.82, "learning_rate": 1.5810357135032661e-06, "loss": 1.9379, "step": 33430 }, { "epoch": 0.82, "learning_rate": 1.580604929607472e-06, "loss": 2.1028, "step": 33431 }, { "epoch": 0.82, "learning_rate": 1.5801741993712273e-06, "loss": 1.8089, "step": 33432 }, { "epoch": 0.82, "learning_rate": 1.5797435227972746e-06, "loss": 1.9239, "step": 33433 }, { "epoch": 0.82, "learning_rate": 1.5793128998883632e-06, "loss": 1.933, "step": 33434 }, { "epoch": 0.82, "learning_rate": 1.5788823306472324e-06, "loss": 1.9651, "step": 33435 }, { "epoch": 0.82, "learning_rate": 1.5784518150766292e-06, "loss": 1.9826, "step": 33436 }, { "epoch": 0.82, "learning_rate": 1.5780213531792987e-06, "loss": 1.8355, "step": 33437 }, { "epoch": 0.82, "learning_rate": 1.5775909449579806e-06, "loss": 1.7384, "step": 33438 }, { "epoch": 0.82, "learning_rate": 1.577160590415423e-06, "loss": 1.8228, "step": 33439 }, { "epoch": 0.82, "learning_rate": 1.5767302895543635e-06, "loss": 1.9918, "step": 33440 }, { "epoch": 0.82, "learning_rate": 1.5763000423775465e-06, "loss": 2.0073, "step": 33441 }, { "epoch": 0.82, "learning_rate": 1.5758698488877178e-06, "loss": 1.7686, "step": 33442 }, { "epoch": 0.82, "learning_rate": 1.5754397090876117e-06, "loss": 2.0386, "step": 33443 }, { "epoch": 0.82, "learning_rate": 1.5750096229799783e-06, "loss": 1.8915, "step": 33444 }, { "epoch": 0.82, "learning_rate": 1.5745795905675498e-06, "loss": 1.9407, "step": 33445 }, { "epoch": 0.82, "learning_rate": 1.574149611853073e-06, "loss": 2.0398, "step": 33446 }, { "epoch": 0.82, "learning_rate": 1.5737196868392857e-06, "loss": 1.8039, "step": 33447 }, { "epoch": 0.82, "learning_rate": 1.573289815528929e-06, "loss": 1.9059, "step": 33448 }, { "epoch": 0.82, "learning_rate": 1.5728599979247449e-06, "loss": 1.8444, "step": 33449 }, { "epoch": 0.82, "learning_rate": 1.5724302340294684e-06, "loss": 1.9688, "step": 33450 }, { "epoch": 0.82, "learning_rate": 1.5720005238458402e-06, "loss": 1.9227, "step": 33451 }, { "epoch": 0.82, "learning_rate": 1.5715708673766017e-06, "loss": 1.8585, "step": 33452 }, { "epoch": 0.82, "learning_rate": 1.5711412646244861e-06, "loss": 1.9916, "step": 33453 }, { "epoch": 0.82, "learning_rate": 1.5707117155922368e-06, "loss": 1.9078, "step": 33454 }, { "epoch": 0.82, "learning_rate": 1.5702822202825862e-06, "loss": 1.8926, "step": 33455 }, { "epoch": 0.82, "learning_rate": 1.5698527786982753e-06, "loss": 1.7736, "step": 33456 }, { "epoch": 0.82, "learning_rate": 1.5694233908420408e-06, "loss": 1.9015, "step": 33457 }, { "epoch": 0.82, "learning_rate": 1.5689940567166174e-06, "loss": 1.7345, "step": 33458 }, { "epoch": 0.82, "learning_rate": 1.5685647763247436e-06, "loss": 2.0406, "step": 33459 }, { "epoch": 0.82, "learning_rate": 1.568135549669152e-06, "loss": 1.9045, "step": 33460 }, { "epoch": 0.82, "learning_rate": 1.5677063767525812e-06, "loss": 1.9136, "step": 33461 }, { "epoch": 0.82, "learning_rate": 1.5672772575777673e-06, "loss": 1.876, "step": 33462 }, { "epoch": 0.82, "learning_rate": 1.566848192147441e-06, "loss": 1.8822, "step": 33463 }, { "epoch": 0.82, "learning_rate": 1.5664191804643403e-06, "loss": 1.9998, "step": 33464 }, { "epoch": 0.82, "learning_rate": 1.5659902225311985e-06, "loss": 1.9321, "step": 33465 }, { "epoch": 0.82, "learning_rate": 1.5655613183507489e-06, "loss": 1.9317, "step": 33466 }, { "epoch": 0.82, "learning_rate": 1.5651324679257284e-06, "loss": 2.0881, "step": 33467 }, { "epoch": 0.82, "learning_rate": 1.5647036712588659e-06, "loss": 1.7133, "step": 33468 }, { "epoch": 0.82, "learning_rate": 1.564274928352897e-06, "loss": 1.7759, "step": 33469 }, { "epoch": 0.82, "learning_rate": 1.5638462392105513e-06, "loss": 1.964, "step": 33470 }, { "epoch": 0.83, "learning_rate": 1.5634176038345627e-06, "loss": 1.9472, "step": 33471 }, { "epoch": 0.83, "learning_rate": 1.5629890222276656e-06, "loss": 1.852, "step": 33472 }, { "epoch": 0.83, "learning_rate": 1.5625604943925876e-06, "loss": 1.8551, "step": 33473 }, { "epoch": 0.83, "learning_rate": 1.5621320203320623e-06, "loss": 1.8656, "step": 33474 }, { "epoch": 0.83, "learning_rate": 1.5617036000488172e-06, "loss": 2.0711, "step": 33475 }, { "epoch": 0.83, "learning_rate": 1.561275233545585e-06, "loss": 1.961, "step": 33476 }, { "epoch": 0.83, "learning_rate": 1.5608469208250987e-06, "loss": 1.8369, "step": 33477 }, { "epoch": 0.83, "learning_rate": 1.560418661890083e-06, "loss": 1.9241, "step": 33478 }, { "epoch": 0.83, "learning_rate": 1.5599904567432712e-06, "loss": 1.9056, "step": 33479 }, { "epoch": 0.83, "learning_rate": 1.5595623053873886e-06, "loss": 1.8992, "step": 33480 }, { "epoch": 0.83, "learning_rate": 1.5591342078251659e-06, "loss": 1.8149, "step": 33481 }, { "epoch": 0.83, "learning_rate": 1.5587061640593327e-06, "loss": 1.8352, "step": 33482 }, { "epoch": 0.83, "learning_rate": 1.5582781740926156e-06, "loss": 1.8854, "step": 33483 }, { "epoch": 0.83, "learning_rate": 1.557850237927745e-06, "loss": 1.923, "step": 33484 }, { "epoch": 0.83, "learning_rate": 1.5574223555674428e-06, "loss": 1.9665, "step": 33485 }, { "epoch": 0.83, "learning_rate": 1.5569945270144405e-06, "loss": 2.0598, "step": 33486 }, { "epoch": 0.83, "learning_rate": 1.5565667522714655e-06, "loss": 2.0258, "step": 33487 }, { "epoch": 0.83, "learning_rate": 1.5561390313412395e-06, "loss": 1.9016, "step": 33488 }, { "epoch": 0.83, "learning_rate": 1.555711364226492e-06, "loss": 1.8435, "step": 33489 }, { "epoch": 0.83, "learning_rate": 1.5552837509299489e-06, "loss": 1.8534, "step": 33490 }, { "epoch": 0.83, "learning_rate": 1.5548561914543336e-06, "loss": 2.0339, "step": 33491 }, { "epoch": 0.83, "learning_rate": 1.5544286858023738e-06, "loss": 1.8493, "step": 33492 }, { "epoch": 0.83, "learning_rate": 1.5540012339767895e-06, "loss": 1.8871, "step": 33493 }, { "epoch": 0.83, "learning_rate": 1.5535738359803088e-06, "loss": 1.8957, "step": 33494 }, { "epoch": 0.83, "learning_rate": 1.553146491815657e-06, "loss": 1.8819, "step": 33495 }, { "epoch": 0.83, "learning_rate": 1.552719201485552e-06, "loss": 1.8781, "step": 33496 }, { "epoch": 0.83, "learning_rate": 1.5522919649927237e-06, "loss": 2.0849, "step": 33497 }, { "epoch": 0.83, "learning_rate": 1.5518647823398892e-06, "loss": 2.0057, "step": 33498 }, { "epoch": 0.83, "learning_rate": 1.5514376535297725e-06, "loss": 1.9119, "step": 33499 }, { "epoch": 0.83, "learning_rate": 1.5510105785650975e-06, "loss": 1.826, "step": 33500 }, { "epoch": 0.83, "learning_rate": 1.550583557448586e-06, "loss": 1.947, "step": 33501 }, { "epoch": 0.83, "learning_rate": 1.5501565901829617e-06, "loss": 1.965, "step": 33502 }, { "epoch": 0.83, "learning_rate": 1.5497296767709402e-06, "loss": 1.9889, "step": 33503 }, { "epoch": 0.83, "learning_rate": 1.5493028172152458e-06, "loss": 2.0226, "step": 33504 }, { "epoch": 0.83, "learning_rate": 1.5488760115186008e-06, "loss": 1.9118, "step": 33505 }, { "epoch": 0.83, "learning_rate": 1.5484492596837196e-06, "loss": 1.8301, "step": 33506 }, { "epoch": 0.83, "learning_rate": 1.5480225617133293e-06, "loss": 1.8919, "step": 33507 }, { "epoch": 0.83, "learning_rate": 1.547595917610143e-06, "loss": 1.9909, "step": 33508 }, { "epoch": 0.83, "learning_rate": 1.5471693273768829e-06, "loss": 1.8288, "step": 33509 }, { "epoch": 0.83, "learning_rate": 1.546742791016269e-06, "loss": 1.8404, "step": 33510 }, { "epoch": 0.83, "learning_rate": 1.5463163085310162e-06, "loss": 1.9668, "step": 33511 }, { "epoch": 0.83, "learning_rate": 1.5458898799238465e-06, "loss": 1.9992, "step": 33512 }, { "epoch": 0.83, "learning_rate": 1.5454635051974732e-06, "loss": 1.882, "step": 33513 }, { "epoch": 0.83, "learning_rate": 1.5450371843546163e-06, "loss": 1.9699, "step": 33514 }, { "epoch": 0.83, "learning_rate": 1.5446109173979961e-06, "loss": 1.8926, "step": 33515 }, { "epoch": 0.83, "learning_rate": 1.5441847043303227e-06, "loss": 1.9755, "step": 33516 }, { "epoch": 0.83, "learning_rate": 1.5437585451543159e-06, "loss": 2.0721, "step": 33517 }, { "epoch": 0.83, "learning_rate": 1.5433324398726923e-06, "loss": 1.9876, "step": 33518 }, { "epoch": 0.83, "learning_rate": 1.5429063884881669e-06, "loss": 2.0812, "step": 33519 }, { "epoch": 0.83, "learning_rate": 1.542480391003457e-06, "loss": 1.8784, "step": 33520 }, { "epoch": 0.83, "learning_rate": 1.542054447421274e-06, "loss": 1.8807, "step": 33521 }, { "epoch": 0.83, "learning_rate": 1.5416285577443356e-06, "loss": 1.7537, "step": 33522 }, { "epoch": 0.83, "learning_rate": 1.5412027219753533e-06, "loss": 1.9359, "step": 33523 }, { "epoch": 0.83, "learning_rate": 1.5407769401170436e-06, "loss": 1.8719, "step": 33524 }, { "epoch": 0.83, "learning_rate": 1.5403512121721209e-06, "loss": 1.8554, "step": 33525 }, { "epoch": 0.83, "learning_rate": 1.5399255381432942e-06, "loss": 1.9958, "step": 33526 }, { "epoch": 0.83, "learning_rate": 1.5394999180332816e-06, "loss": 1.7516, "step": 33527 }, { "epoch": 0.83, "learning_rate": 1.5390743518447915e-06, "loss": 1.9829, "step": 33528 }, { "epoch": 0.83, "learning_rate": 1.5386488395805365e-06, "loss": 1.8863, "step": 33529 }, { "epoch": 0.83, "learning_rate": 1.5382233812432335e-06, "loss": 1.9575, "step": 33530 }, { "epoch": 0.83, "learning_rate": 1.5377979768355878e-06, "loss": 1.9902, "step": 33531 }, { "epoch": 0.83, "learning_rate": 1.537372626360315e-06, "loss": 1.8814, "step": 33532 }, { "epoch": 0.83, "learning_rate": 1.5369473298201232e-06, "loss": 1.9478, "step": 33533 }, { "epoch": 0.83, "learning_rate": 1.5365220872177233e-06, "loss": 1.6841, "step": 33534 }, { "epoch": 0.83, "learning_rate": 1.5360968985558266e-06, "loss": 1.7489, "step": 33535 }, { "epoch": 0.83, "learning_rate": 1.535671763837142e-06, "loss": 1.816, "step": 33536 }, { "epoch": 0.83, "learning_rate": 1.5352466830643832e-06, "loss": 1.9786, "step": 33537 }, { "epoch": 0.83, "learning_rate": 1.534821656240253e-06, "loss": 1.8797, "step": 33538 }, { "epoch": 0.83, "learning_rate": 1.5343966833674628e-06, "loss": 1.8211, "step": 33539 }, { "epoch": 0.83, "learning_rate": 1.533971764448724e-06, "loss": 1.9261, "step": 33540 }, { "epoch": 0.83, "learning_rate": 1.5335468994867396e-06, "loss": 1.8948, "step": 33541 }, { "epoch": 0.83, "learning_rate": 1.5331220884842236e-06, "loss": 1.9433, "step": 33542 }, { "epoch": 0.83, "learning_rate": 1.5326973314438765e-06, "loss": 2.0177, "step": 33543 }, { "epoch": 0.83, "learning_rate": 1.5322726283684076e-06, "loss": 2.01, "step": 33544 }, { "epoch": 0.83, "learning_rate": 1.5318479792605289e-06, "loss": 2.1142, "step": 33545 }, { "epoch": 0.83, "learning_rate": 1.5314233841229398e-06, "loss": 1.7399, "step": 33546 }, { "epoch": 0.83, "learning_rate": 1.5309988429583511e-06, "loss": 1.8217, "step": 33547 }, { "epoch": 0.83, "learning_rate": 1.530574355769464e-06, "loss": 1.8741, "step": 33548 }, { "epoch": 0.83, "learning_rate": 1.5301499225589878e-06, "loss": 2.0578, "step": 33549 }, { "epoch": 0.83, "learning_rate": 1.5297255433296288e-06, "loss": 1.9714, "step": 33550 }, { "epoch": 0.83, "learning_rate": 1.5293012180840861e-06, "loss": 2.0081, "step": 33551 }, { "epoch": 0.83, "learning_rate": 1.5288769468250676e-06, "loss": 1.9146, "step": 33552 }, { "epoch": 0.83, "learning_rate": 1.5284527295552775e-06, "loss": 2.0367, "step": 33553 }, { "epoch": 0.83, "learning_rate": 1.5280285662774186e-06, "loss": 1.8087, "step": 33554 }, { "epoch": 0.83, "learning_rate": 1.527604456994196e-06, "loss": 2.0367, "step": 33555 }, { "epoch": 0.83, "learning_rate": 1.5271804017083103e-06, "loss": 2.1103, "step": 33556 }, { "epoch": 0.83, "learning_rate": 1.5267564004224644e-06, "loss": 1.8821, "step": 33557 }, { "epoch": 0.83, "learning_rate": 1.5263324531393642e-06, "loss": 2.0907, "step": 33558 }, { "epoch": 0.83, "learning_rate": 1.5259085598617051e-06, "loss": 1.9126, "step": 33559 }, { "epoch": 0.83, "learning_rate": 1.5254847205921964e-06, "loss": 1.8063, "step": 33560 }, { "epoch": 0.83, "learning_rate": 1.5250609353335322e-06, "loss": 2.1173, "step": 33561 }, { "epoch": 0.83, "learning_rate": 1.524637204088416e-06, "loss": 2.0085, "step": 33562 }, { "epoch": 0.83, "learning_rate": 1.5242135268595526e-06, "loss": 1.808, "step": 33563 }, { "epoch": 0.83, "learning_rate": 1.5237899036496351e-06, "loss": 1.8017, "step": 33564 }, { "epoch": 0.83, "learning_rate": 1.5233663344613693e-06, "loss": 1.8739, "step": 33565 }, { "epoch": 0.83, "learning_rate": 1.5229428192974505e-06, "loss": 1.7934, "step": 33566 }, { "epoch": 0.83, "learning_rate": 1.522519358160579e-06, "loss": 1.9246, "step": 33567 }, { "epoch": 0.83, "learning_rate": 1.5220959510534571e-06, "loss": 1.9437, "step": 33568 }, { "epoch": 0.83, "learning_rate": 1.5216725979787783e-06, "loss": 2.0151, "step": 33569 }, { "epoch": 0.83, "learning_rate": 1.5212492989392435e-06, "loss": 1.9919, "step": 33570 }, { "epoch": 0.83, "learning_rate": 1.5208260539375497e-06, "loss": 2.0118, "step": 33571 }, { "epoch": 0.83, "learning_rate": 1.5204028629763945e-06, "loss": 1.9644, "step": 33572 }, { "epoch": 0.83, "learning_rate": 1.519979726058478e-06, "loss": 1.9633, "step": 33573 }, { "epoch": 0.83, "learning_rate": 1.5195566431864916e-06, "loss": 1.9053, "step": 33574 }, { "epoch": 0.83, "learning_rate": 1.5191336143631363e-06, "loss": 1.9127, "step": 33575 }, { "epoch": 0.83, "learning_rate": 1.5187106395911044e-06, "loss": 2.0374, "step": 33576 }, { "epoch": 0.83, "learning_rate": 1.5182877188730939e-06, "loss": 1.9769, "step": 33577 }, { "epoch": 0.83, "learning_rate": 1.5178648522118022e-06, "loss": 1.9452, "step": 33578 }, { "epoch": 0.83, "learning_rate": 1.5174420396099198e-06, "loss": 2.011, "step": 33579 }, { "epoch": 0.83, "learning_rate": 1.5170192810701456e-06, "loss": 1.9257, "step": 33580 }, { "epoch": 0.83, "learning_rate": 1.5165965765951695e-06, "loss": 1.9945, "step": 33581 }, { "epoch": 0.83, "learning_rate": 1.5161739261876884e-06, "loss": 1.9411, "step": 33582 }, { "epoch": 0.83, "learning_rate": 1.515751329850398e-06, "loss": 1.9921, "step": 33583 }, { "epoch": 0.83, "learning_rate": 1.5153287875859868e-06, "loss": 1.9619, "step": 33584 }, { "epoch": 0.83, "learning_rate": 1.514906299397153e-06, "loss": 1.7895, "step": 33585 }, { "epoch": 0.83, "learning_rate": 1.5144838652865846e-06, "loss": 2.0648, "step": 33586 }, { "epoch": 0.83, "learning_rate": 1.5140614852569745e-06, "loss": 1.712, "step": 33587 }, { "epoch": 0.83, "learning_rate": 1.5136391593110179e-06, "loss": 1.8665, "step": 33588 }, { "epoch": 0.83, "learning_rate": 1.513216887451403e-06, "loss": 1.9424, "step": 33589 }, { "epoch": 0.83, "learning_rate": 1.512794669680826e-06, "loss": 1.9856, "step": 33590 }, { "epoch": 0.83, "learning_rate": 1.5123725060019712e-06, "loss": 2.059, "step": 33591 }, { "epoch": 0.83, "learning_rate": 1.511950396417533e-06, "loss": 2.1463, "step": 33592 }, { "epoch": 0.83, "learning_rate": 1.5115283409302028e-06, "loss": 1.764, "step": 33593 }, { "epoch": 0.83, "learning_rate": 1.5111063395426674e-06, "loss": 2.0007, "step": 33594 }, { "epoch": 0.83, "learning_rate": 1.5106843922576197e-06, "loss": 1.6947, "step": 33595 }, { "epoch": 0.83, "learning_rate": 1.510262499077745e-06, "loss": 1.839, "step": 33596 }, { "epoch": 0.83, "learning_rate": 1.5098406600057336e-06, "loss": 2.1191, "step": 33597 }, { "epoch": 0.83, "learning_rate": 1.5094188750442762e-06, "loss": 1.9538, "step": 33598 }, { "epoch": 0.83, "learning_rate": 1.5089971441960582e-06, "loss": 1.7649, "step": 33599 }, { "epoch": 0.83, "learning_rate": 1.508575467463771e-06, "loss": 1.8799, "step": 33600 }, { "epoch": 0.83, "learning_rate": 1.5081538448500965e-06, "loss": 2.0045, "step": 33601 }, { "epoch": 0.83, "learning_rate": 1.507732276357725e-06, "loss": 1.8493, "step": 33602 }, { "epoch": 0.83, "learning_rate": 1.5073107619893456e-06, "loss": 1.7611, "step": 33603 }, { "epoch": 0.83, "learning_rate": 1.5068893017476404e-06, "loss": 1.8533, "step": 33604 }, { "epoch": 0.83, "learning_rate": 1.5064678956352974e-06, "loss": 2.0195, "step": 33605 }, { "epoch": 0.83, "learning_rate": 1.5060465436550021e-06, "loss": 1.875, "step": 33606 }, { "epoch": 0.83, "learning_rate": 1.5056252458094412e-06, "loss": 1.8941, "step": 33607 }, { "epoch": 0.83, "learning_rate": 1.5052040021013004e-06, "loss": 1.9868, "step": 33608 }, { "epoch": 0.83, "learning_rate": 1.504782812533261e-06, "loss": 1.7876, "step": 33609 }, { "epoch": 0.83, "learning_rate": 1.5043616771080083e-06, "loss": 1.8299, "step": 33610 }, { "epoch": 0.83, "learning_rate": 1.5039405958282304e-06, "loss": 1.9599, "step": 33611 }, { "epoch": 0.83, "learning_rate": 1.5035195686966053e-06, "loss": 1.8983, "step": 33612 }, { "epoch": 0.83, "learning_rate": 1.5030985957158218e-06, "loss": 1.9992, "step": 33613 }, { "epoch": 0.83, "learning_rate": 1.5026776768885566e-06, "loss": 1.758, "step": 33614 }, { "epoch": 0.83, "learning_rate": 1.5022568122174963e-06, "loss": 2.0401, "step": 33615 }, { "epoch": 0.83, "learning_rate": 1.5018360017053258e-06, "loss": 2.071, "step": 33616 }, { "epoch": 0.83, "learning_rate": 1.5014152453547203e-06, "loss": 1.9242, "step": 33617 }, { "epoch": 0.83, "learning_rate": 1.500994543168368e-06, "loss": 1.8624, "step": 33618 }, { "epoch": 0.83, "learning_rate": 1.5005738951489445e-06, "loss": 1.9128, "step": 33619 }, { "epoch": 0.83, "learning_rate": 1.5001533012991343e-06, "loss": 1.9922, "step": 33620 }, { "epoch": 0.83, "learning_rate": 1.4997327616216194e-06, "loss": 1.948, "step": 33621 }, { "epoch": 0.83, "learning_rate": 1.4993122761190749e-06, "loss": 1.8692, "step": 33622 }, { "epoch": 0.83, "learning_rate": 1.4988918447941836e-06, "loss": 1.8541, "step": 33623 }, { "epoch": 0.83, "learning_rate": 1.4984714676496248e-06, "loss": 1.9299, "step": 33624 }, { "epoch": 0.83, "learning_rate": 1.4980511446880787e-06, "loss": 1.8087, "step": 33625 }, { "epoch": 0.83, "learning_rate": 1.497630875912225e-06, "loss": 1.8641, "step": 33626 }, { "epoch": 0.83, "learning_rate": 1.4972106613247383e-06, "loss": 2.0148, "step": 33627 }, { "epoch": 0.83, "learning_rate": 1.4967905009283012e-06, "loss": 1.8957, "step": 33628 }, { "epoch": 0.83, "learning_rate": 1.496370394725587e-06, "loss": 1.7545, "step": 33629 }, { "epoch": 0.83, "learning_rate": 1.4959503427192767e-06, "loss": 2.164, "step": 33630 }, { "epoch": 0.83, "learning_rate": 1.4955303449120473e-06, "loss": 1.8249, "step": 33631 }, { "epoch": 0.83, "learning_rate": 1.4951104013065732e-06, "loss": 1.9313, "step": 33632 }, { "epoch": 0.83, "learning_rate": 1.4946905119055333e-06, "loss": 2.0846, "step": 33633 }, { "epoch": 0.83, "learning_rate": 1.494270676711601e-06, "loss": 1.9429, "step": 33634 }, { "epoch": 0.83, "learning_rate": 1.4938508957274534e-06, "loss": 1.8269, "step": 33635 }, { "epoch": 0.83, "learning_rate": 1.493431168955769e-06, "loss": 2.0068, "step": 33636 }, { "epoch": 0.83, "learning_rate": 1.4930114963992171e-06, "loss": 1.8678, "step": 33637 }, { "epoch": 0.83, "learning_rate": 1.4925918780604753e-06, "loss": 1.6916, "step": 33638 }, { "epoch": 0.83, "learning_rate": 1.4921723139422195e-06, "loss": 2.0044, "step": 33639 }, { "epoch": 0.83, "learning_rate": 1.4917528040471208e-06, "loss": 2.0332, "step": 33640 }, { "epoch": 0.83, "learning_rate": 1.4913333483778547e-06, "loss": 1.8365, "step": 33641 }, { "epoch": 0.83, "learning_rate": 1.4909139469370937e-06, "loss": 2.0726, "step": 33642 }, { "epoch": 0.83, "learning_rate": 1.4904945997275121e-06, "loss": 1.8508, "step": 33643 }, { "epoch": 0.83, "learning_rate": 1.4900753067517815e-06, "loss": 2.2447, "step": 33644 }, { "epoch": 0.83, "learning_rate": 1.4896560680125727e-06, "loss": 1.8074, "step": 33645 }, { "epoch": 0.83, "learning_rate": 1.4892368835125614e-06, "loss": 2.0083, "step": 33646 }, { "epoch": 0.83, "learning_rate": 1.4888177532544156e-06, "loss": 2.0861, "step": 33647 }, { "epoch": 0.83, "learning_rate": 1.48839867724081e-06, "loss": 1.8892, "step": 33648 }, { "epoch": 0.83, "learning_rate": 1.487979655474412e-06, "loss": 1.8315, "step": 33649 }, { "epoch": 0.83, "learning_rate": 1.487560687957892e-06, "loss": 1.9571, "step": 33650 }, { "epoch": 0.83, "learning_rate": 1.4871417746939254e-06, "loss": 2.0027, "step": 33651 }, { "epoch": 0.83, "learning_rate": 1.4867229156851759e-06, "loss": 2.0216, "step": 33652 }, { "epoch": 0.83, "learning_rate": 1.486304110934318e-06, "loss": 1.9636, "step": 33653 }, { "epoch": 0.83, "learning_rate": 1.485885360444016e-06, "loss": 1.9117, "step": 33654 }, { "epoch": 0.83, "learning_rate": 1.4854666642169412e-06, "loss": 1.8715, "step": 33655 }, { "epoch": 0.83, "learning_rate": 1.485048022255764e-06, "loss": 1.8359, "step": 33656 }, { "epoch": 0.83, "learning_rate": 1.4846294345631496e-06, "loss": 1.9218, "step": 33657 }, { "epoch": 0.83, "learning_rate": 1.484210901141766e-06, "loss": 2.1012, "step": 33658 }, { "epoch": 0.83, "learning_rate": 1.4837924219942822e-06, "loss": 1.8413, "step": 33659 }, { "epoch": 0.83, "learning_rate": 1.483373997123364e-06, "loss": 1.8076, "step": 33660 }, { "epoch": 0.83, "learning_rate": 1.4829556265316813e-06, "loss": 1.77, "step": 33661 }, { "epoch": 0.83, "learning_rate": 1.4825373102218966e-06, "loss": 1.9698, "step": 33662 }, { "epoch": 0.83, "learning_rate": 1.4821190481966785e-06, "loss": 1.8966, "step": 33663 }, { "epoch": 0.83, "learning_rate": 1.4817008404586897e-06, "loss": 1.8719, "step": 33664 }, { "epoch": 0.83, "learning_rate": 1.4812826870105967e-06, "loss": 1.9833, "step": 33665 }, { "epoch": 0.83, "learning_rate": 1.4808645878550688e-06, "loss": 2.0727, "step": 33666 }, { "epoch": 0.83, "learning_rate": 1.4804465429947645e-06, "loss": 1.9134, "step": 33667 }, { "epoch": 0.83, "learning_rate": 1.480028552432351e-06, "loss": 1.7862, "step": 33668 }, { "epoch": 0.83, "learning_rate": 1.4796106161704937e-06, "loss": 2.0183, "step": 33669 }, { "epoch": 0.83, "learning_rate": 1.4791927342118528e-06, "loss": 1.8954, "step": 33670 }, { "epoch": 0.83, "learning_rate": 1.4787749065590961e-06, "loss": 2.0769, "step": 33671 }, { "epoch": 0.83, "learning_rate": 1.4783571332148816e-06, "loss": 1.9314, "step": 33672 }, { "epoch": 0.83, "learning_rate": 1.4779394141818738e-06, "loss": 1.9654, "step": 33673 }, { "epoch": 0.83, "learning_rate": 1.4775217494627382e-06, "loss": 2.0196, "step": 33674 }, { "epoch": 0.83, "learning_rate": 1.4771041390601303e-06, "loss": 1.7673, "step": 33675 }, { "epoch": 0.83, "learning_rate": 1.4766865829767162e-06, "loss": 1.9664, "step": 33676 }, { "epoch": 0.83, "learning_rate": 1.476269081215157e-06, "loss": 1.8506, "step": 33677 }, { "epoch": 0.83, "learning_rate": 1.4758516337781115e-06, "loss": 2.0251, "step": 33678 }, { "epoch": 0.83, "learning_rate": 1.4754342406682442e-06, "loss": 1.8421, "step": 33679 }, { "epoch": 0.83, "learning_rate": 1.4750169018882098e-06, "loss": 1.9031, "step": 33680 }, { "epoch": 0.83, "learning_rate": 1.474599617440674e-06, "loss": 2.0133, "step": 33681 }, { "epoch": 0.83, "learning_rate": 1.47418238732829e-06, "loss": 1.8497, "step": 33682 }, { "epoch": 0.83, "learning_rate": 1.4737652115537216e-06, "loss": 1.8554, "step": 33683 }, { "epoch": 0.83, "learning_rate": 1.4733480901196274e-06, "loss": 2.0957, "step": 33684 }, { "epoch": 0.83, "learning_rate": 1.472931023028662e-06, "loss": 2.0059, "step": 33685 }, { "epoch": 0.83, "learning_rate": 1.4725140102834889e-06, "loss": 1.9278, "step": 33686 }, { "epoch": 0.83, "learning_rate": 1.4720970518867605e-06, "loss": 2.0044, "step": 33687 }, { "epoch": 0.83, "learning_rate": 1.4716801478411369e-06, "loss": 2.1061, "step": 33688 }, { "epoch": 0.83, "learning_rate": 1.471263298149277e-06, "loss": 1.937, "step": 33689 }, { "epoch": 0.83, "learning_rate": 1.470846502813833e-06, "loss": 1.8253, "step": 33690 }, { "epoch": 0.83, "learning_rate": 1.4704297618374653e-06, "loss": 1.8844, "step": 33691 }, { "epoch": 0.83, "learning_rate": 1.4700130752228292e-06, "loss": 2.1872, "step": 33692 }, { "epoch": 0.83, "learning_rate": 1.4695964429725773e-06, "loss": 2.0064, "step": 33693 }, { "epoch": 0.83, "learning_rate": 1.4691798650893685e-06, "loss": 2.04, "step": 33694 }, { "epoch": 0.83, "learning_rate": 1.468763341575855e-06, "loss": 1.8058, "step": 33695 }, { "epoch": 0.83, "learning_rate": 1.4683468724346973e-06, "loss": 1.997, "step": 33696 }, { "epoch": 0.83, "learning_rate": 1.4679304576685415e-06, "loss": 1.8501, "step": 33697 }, { "epoch": 0.83, "learning_rate": 1.4675140972800451e-06, "loss": 1.7309, "step": 33698 }, { "epoch": 0.83, "learning_rate": 1.4670977912718654e-06, "loss": 1.838, "step": 33699 }, { "epoch": 0.83, "learning_rate": 1.4666815396466494e-06, "loss": 1.9198, "step": 33700 }, { "epoch": 0.83, "learning_rate": 1.466265342407055e-06, "loss": 1.8002, "step": 33701 }, { "epoch": 0.83, "learning_rate": 1.4658491995557312e-06, "loss": 1.9351, "step": 33702 }, { "epoch": 0.83, "learning_rate": 1.4654331110953314e-06, "loss": 2.0165, "step": 33703 }, { "epoch": 0.83, "learning_rate": 1.4650170770285088e-06, "loss": 1.895, "step": 33704 }, { "epoch": 0.83, "learning_rate": 1.4646010973579127e-06, "loss": 1.9223, "step": 33705 }, { "epoch": 0.83, "learning_rate": 1.4641851720861978e-06, "loss": 1.846, "step": 33706 }, { "epoch": 0.83, "learning_rate": 1.4637693012160103e-06, "loss": 1.8957, "step": 33707 }, { "epoch": 0.83, "learning_rate": 1.463353484750002e-06, "loss": 1.9025, "step": 33708 }, { "epoch": 0.83, "learning_rate": 1.4629377226908249e-06, "loss": 1.8327, "step": 33709 }, { "epoch": 0.83, "learning_rate": 1.462522015041129e-06, "loss": 1.8959, "step": 33710 }, { "epoch": 0.83, "learning_rate": 1.4621063618035614e-06, "loss": 1.86, "step": 33711 }, { "epoch": 0.83, "learning_rate": 1.461690762980772e-06, "loss": 1.929, "step": 33712 }, { "epoch": 0.83, "learning_rate": 1.4612752185754098e-06, "loss": 1.7897, "step": 33713 }, { "epoch": 0.83, "learning_rate": 1.460859728590126e-06, "loss": 1.7982, "step": 33714 }, { "epoch": 0.83, "learning_rate": 1.4604442930275632e-06, "loss": 1.8947, "step": 33715 }, { "epoch": 0.83, "learning_rate": 1.4600289118903743e-06, "loss": 1.8808, "step": 33716 }, { "epoch": 0.83, "learning_rate": 1.4596135851812022e-06, "loss": 1.8862, "step": 33717 }, { "epoch": 0.83, "learning_rate": 1.4591983129026955e-06, "loss": 1.9115, "step": 33718 }, { "epoch": 0.83, "learning_rate": 1.4587830950575043e-06, "loss": 1.8981, "step": 33719 }, { "epoch": 0.83, "learning_rate": 1.4583679316482692e-06, "loss": 1.8752, "step": 33720 }, { "epoch": 0.83, "learning_rate": 1.4579528226776418e-06, "loss": 1.9321, "step": 33721 }, { "epoch": 0.83, "learning_rate": 1.4575377681482617e-06, "loss": 1.906, "step": 33722 }, { "epoch": 0.83, "learning_rate": 1.4571227680627776e-06, "loss": 1.9921, "step": 33723 }, { "epoch": 0.83, "learning_rate": 1.4567078224238362e-06, "loss": 1.8009, "step": 33724 }, { "epoch": 0.83, "learning_rate": 1.4562929312340778e-06, "loss": 1.7014, "step": 33725 }, { "epoch": 0.83, "learning_rate": 1.4558780944961492e-06, "loss": 2.0928, "step": 33726 }, { "epoch": 0.83, "learning_rate": 1.4554633122126938e-06, "loss": 2.0702, "step": 33727 }, { "epoch": 0.83, "learning_rate": 1.455048584386357e-06, "loss": 1.8703, "step": 33728 }, { "epoch": 0.83, "learning_rate": 1.4546339110197794e-06, "loss": 1.9333, "step": 33729 }, { "epoch": 0.83, "learning_rate": 1.454219292115604e-06, "loss": 1.9251, "step": 33730 }, { "epoch": 0.83, "learning_rate": 1.4538047276764734e-06, "loss": 1.8832, "step": 33731 }, { "epoch": 0.83, "learning_rate": 1.453390217705033e-06, "loss": 1.964, "step": 33732 }, { "epoch": 0.83, "learning_rate": 1.452975762203921e-06, "loss": 1.8879, "step": 33733 }, { "epoch": 0.83, "learning_rate": 1.4525613611757805e-06, "loss": 1.916, "step": 33734 }, { "epoch": 0.83, "learning_rate": 1.4521470146232508e-06, "loss": 2.0075, "step": 33735 }, { "epoch": 0.83, "learning_rate": 1.451732722548973e-06, "loss": 1.8964, "step": 33736 }, { "epoch": 0.83, "learning_rate": 1.4513184849555905e-06, "loss": 1.9987, "step": 33737 }, { "epoch": 0.83, "learning_rate": 1.4509043018457404e-06, "loss": 1.9565, "step": 33738 }, { "epoch": 0.83, "learning_rate": 1.4504901732220645e-06, "loss": 2.077, "step": 33739 }, { "epoch": 0.83, "learning_rate": 1.4500760990871975e-06, "loss": 1.9468, "step": 33740 }, { "epoch": 0.83, "learning_rate": 1.4496620794437833e-06, "loss": 1.9586, "step": 33741 }, { "epoch": 0.83, "learning_rate": 1.4492481142944613e-06, "loss": 1.9498, "step": 33742 }, { "epoch": 0.83, "learning_rate": 1.4488342036418656e-06, "loss": 2.0992, "step": 33743 }, { "epoch": 0.83, "learning_rate": 1.448420347488636e-06, "loss": 1.9285, "step": 33744 }, { "epoch": 0.83, "learning_rate": 1.4480065458374126e-06, "loss": 1.9542, "step": 33745 }, { "epoch": 0.83, "learning_rate": 1.4475927986908278e-06, "loss": 2.0029, "step": 33746 }, { "epoch": 0.83, "learning_rate": 1.4471791060515217e-06, "loss": 1.7432, "step": 33747 }, { "epoch": 0.83, "learning_rate": 1.446765467922131e-06, "loss": 2.0918, "step": 33748 }, { "epoch": 0.83, "learning_rate": 1.4463518843052927e-06, "loss": 1.9856, "step": 33749 }, { "epoch": 0.83, "learning_rate": 1.44593835520364e-06, "loss": 1.9966, "step": 33750 }, { "epoch": 0.83, "learning_rate": 1.4455248806198096e-06, "loss": 2.0192, "step": 33751 }, { "epoch": 0.83, "learning_rate": 1.4451114605564386e-06, "loss": 1.962, "step": 33752 }, { "epoch": 0.83, "learning_rate": 1.4446980950161583e-06, "loss": 1.9136, "step": 33753 }, { "epoch": 0.83, "learning_rate": 1.4442847840016072e-06, "loss": 1.951, "step": 33754 }, { "epoch": 0.83, "learning_rate": 1.4438715275154157e-06, "loss": 1.9332, "step": 33755 }, { "epoch": 0.83, "learning_rate": 1.4434583255602186e-06, "loss": 1.8141, "step": 33756 }, { "epoch": 0.83, "learning_rate": 1.4430451781386533e-06, "loss": 1.899, "step": 33757 }, { "epoch": 0.83, "learning_rate": 1.4426320852533471e-06, "loss": 2.1575, "step": 33758 }, { "epoch": 0.83, "learning_rate": 1.4422190469069374e-06, "loss": 1.9965, "step": 33759 }, { "epoch": 0.83, "learning_rate": 1.4418060631020524e-06, "loss": 1.9546, "step": 33760 }, { "epoch": 0.83, "learning_rate": 1.4413931338413268e-06, "loss": 1.7849, "step": 33761 }, { "epoch": 0.83, "learning_rate": 1.4409802591273913e-06, "loss": 2.172, "step": 33762 }, { "epoch": 0.83, "learning_rate": 1.4405674389628798e-06, "loss": 1.9368, "step": 33763 }, { "epoch": 0.83, "learning_rate": 1.44015467335042e-06, "loss": 1.7986, "step": 33764 }, { "epoch": 0.83, "learning_rate": 1.4397419622926445e-06, "loss": 2.1439, "step": 33765 }, { "epoch": 0.83, "learning_rate": 1.439329305792182e-06, "loss": 2.0244, "step": 33766 }, { "epoch": 0.83, "learning_rate": 1.438916703851666e-06, "loss": 2.0337, "step": 33767 }, { "epoch": 0.83, "learning_rate": 1.4385041564737224e-06, "loss": 1.8635, "step": 33768 }, { "epoch": 0.83, "learning_rate": 1.4380916636609843e-06, "loss": 1.9788, "step": 33769 }, { "epoch": 0.83, "learning_rate": 1.4376792254160754e-06, "loss": 1.8774, "step": 33770 }, { "epoch": 0.83, "learning_rate": 1.437266841741627e-06, "loss": 1.9673, "step": 33771 }, { "epoch": 0.83, "learning_rate": 1.43685451264027e-06, "loss": 1.8813, "step": 33772 }, { "epoch": 0.83, "learning_rate": 1.4364422381146282e-06, "loss": 1.8851, "step": 33773 }, { "epoch": 0.83, "learning_rate": 1.4360300181673326e-06, "loss": 1.9509, "step": 33774 }, { "epoch": 0.83, "learning_rate": 1.4356178528010056e-06, "loss": 2.038, "step": 33775 }, { "epoch": 0.83, "learning_rate": 1.4352057420182774e-06, "loss": 1.9129, "step": 33776 }, { "epoch": 0.83, "learning_rate": 1.434793685821777e-06, "loss": 1.8851, "step": 33777 }, { "epoch": 0.83, "learning_rate": 1.4343816842141255e-06, "loss": 1.7043, "step": 33778 }, { "epoch": 0.83, "learning_rate": 1.4339697371979511e-06, "loss": 1.9563, "step": 33779 }, { "epoch": 0.83, "learning_rate": 1.433557844775879e-06, "loss": 1.9163, "step": 33780 }, { "epoch": 0.83, "learning_rate": 1.4331460069505365e-06, "loss": 1.9196, "step": 33781 }, { "epoch": 0.83, "learning_rate": 1.4327342237245433e-06, "loss": 1.8093, "step": 33782 }, { "epoch": 0.83, "learning_rate": 1.4323224951005288e-06, "loss": 1.9064, "step": 33783 }, { "epoch": 0.83, "learning_rate": 1.431910821081116e-06, "loss": 2.0095, "step": 33784 }, { "epoch": 0.83, "learning_rate": 1.4314992016689256e-06, "loss": 1.9747, "step": 33785 }, { "epoch": 0.83, "learning_rate": 1.4310876368665837e-06, "loss": 2.0877, "step": 33786 }, { "epoch": 0.83, "learning_rate": 1.4306761266767144e-06, "loss": 1.7699, "step": 33787 }, { "epoch": 0.83, "learning_rate": 1.4302646711019374e-06, "loss": 1.8125, "step": 33788 }, { "epoch": 0.83, "learning_rate": 1.4298532701448765e-06, "loss": 2.0352, "step": 33789 }, { "epoch": 0.83, "learning_rate": 1.4294419238081547e-06, "loss": 2.0426, "step": 33790 }, { "epoch": 0.83, "learning_rate": 1.4290306320943914e-06, "loss": 1.8935, "step": 33791 }, { "epoch": 0.83, "learning_rate": 1.428619395006211e-06, "loss": 2.1152, "step": 33792 }, { "epoch": 0.83, "learning_rate": 1.4282082125462305e-06, "loss": 2.0277, "step": 33793 }, { "epoch": 0.83, "learning_rate": 1.427797084717072e-06, "loss": 1.9216, "step": 33794 }, { "epoch": 0.83, "learning_rate": 1.427386011521359e-06, "loss": 1.9837, "step": 33795 }, { "epoch": 0.83, "learning_rate": 1.4269749929617061e-06, "loss": 1.9518, "step": 33796 }, { "epoch": 0.83, "learning_rate": 1.4265640290407356e-06, "loss": 1.9828, "step": 33797 }, { "epoch": 0.83, "learning_rate": 1.4261531197610668e-06, "loss": 1.9235, "step": 33798 }, { "epoch": 0.83, "learning_rate": 1.4257422651253205e-06, "loss": 2.1605, "step": 33799 }, { "epoch": 0.83, "learning_rate": 1.4253314651361106e-06, "loss": 1.9948, "step": 33800 }, { "epoch": 0.83, "learning_rate": 1.4249207197960579e-06, "loss": 1.9743, "step": 33801 }, { "epoch": 0.83, "learning_rate": 1.4245100291077817e-06, "loss": 1.8695, "step": 33802 }, { "epoch": 0.83, "learning_rate": 1.4240993930738957e-06, "loss": 1.7943, "step": 33803 }, { "epoch": 0.83, "learning_rate": 1.4236888116970194e-06, "loss": 1.9075, "step": 33804 }, { "epoch": 0.83, "learning_rate": 1.423278284979771e-06, "loss": 1.8944, "step": 33805 }, { "epoch": 0.83, "learning_rate": 1.4228678129247642e-06, "loss": 1.7305, "step": 33806 }, { "epoch": 0.83, "learning_rate": 1.4224573955346166e-06, "loss": 1.9705, "step": 33807 }, { "epoch": 0.83, "learning_rate": 1.4220470328119417e-06, "loss": 1.9311, "step": 33808 }, { "epoch": 0.83, "learning_rate": 1.4216367247593565e-06, "loss": 1.7499, "step": 33809 }, { "epoch": 0.83, "learning_rate": 1.4212264713794788e-06, "loss": 2.0247, "step": 33810 }, { "epoch": 0.83, "learning_rate": 1.4208162726749175e-06, "loss": 1.9732, "step": 33811 }, { "epoch": 0.83, "learning_rate": 1.420406128648293e-06, "loss": 1.948, "step": 33812 }, { "epoch": 0.83, "learning_rate": 1.4199960393022138e-06, "loss": 1.93, "step": 33813 }, { "epoch": 0.83, "learning_rate": 1.4195860046392962e-06, "loss": 1.7444, "step": 33814 }, { "epoch": 0.83, "learning_rate": 1.4191760246621523e-06, "loss": 2.0228, "step": 33815 }, { "epoch": 0.83, "learning_rate": 1.4187660993733965e-06, "loss": 1.8739, "step": 33816 }, { "epoch": 0.83, "learning_rate": 1.4183562287756436e-06, "loss": 1.8648, "step": 33817 }, { "epoch": 0.83, "learning_rate": 1.4179464128715004e-06, "loss": 1.785, "step": 33818 }, { "epoch": 0.83, "learning_rate": 1.4175366516635813e-06, "loss": 1.7646, "step": 33819 }, { "epoch": 0.83, "learning_rate": 1.4171269451545011e-06, "loss": 1.9974, "step": 33820 }, { "epoch": 0.83, "learning_rate": 1.4167172933468654e-06, "loss": 1.8724, "step": 33821 }, { "epoch": 0.83, "learning_rate": 1.4163076962432897e-06, "loss": 1.9446, "step": 33822 }, { "epoch": 0.83, "learning_rate": 1.4158981538463801e-06, "loss": 1.9758, "step": 33823 }, { "epoch": 0.83, "learning_rate": 1.4154886661587486e-06, "loss": 1.9906, "step": 33824 }, { "epoch": 0.83, "learning_rate": 1.4150792331830088e-06, "loss": 1.8578, "step": 33825 }, { "epoch": 0.83, "learning_rate": 1.414669854921764e-06, "loss": 1.8016, "step": 33826 }, { "epoch": 0.83, "learning_rate": 1.414260531377628e-06, "loss": 1.8503, "step": 33827 }, { "epoch": 0.83, "learning_rate": 1.4138512625532063e-06, "loss": 2.027, "step": 33828 }, { "epoch": 0.83, "learning_rate": 1.413442048451108e-06, "loss": 1.8453, "step": 33829 }, { "epoch": 0.83, "learning_rate": 1.4130328890739441e-06, "loss": 1.872, "step": 33830 }, { "epoch": 0.83, "learning_rate": 1.4126237844243173e-06, "loss": 2.0286, "step": 33831 }, { "epoch": 0.83, "learning_rate": 1.4122147345048387e-06, "loss": 1.9975, "step": 33832 }, { "epoch": 0.83, "learning_rate": 1.411805739318113e-06, "loss": 1.932, "step": 33833 }, { "epoch": 0.83, "learning_rate": 1.4113967988667487e-06, "loss": 1.8947, "step": 33834 }, { "epoch": 0.83, "learning_rate": 1.4109879131533543e-06, "loss": 1.915, "step": 33835 }, { "epoch": 0.83, "learning_rate": 1.4105790821805299e-06, "loss": 1.8713, "step": 33836 }, { "epoch": 0.83, "learning_rate": 1.4101703059508875e-06, "loss": 1.8659, "step": 33837 }, { "epoch": 0.83, "learning_rate": 1.4097615844670265e-06, "loss": 2.0402, "step": 33838 }, { "epoch": 0.83, "learning_rate": 1.4093529177315536e-06, "loss": 1.9172, "step": 33839 }, { "epoch": 0.83, "learning_rate": 1.4089443057470776e-06, "loss": 2.1393, "step": 33840 }, { "epoch": 0.83, "learning_rate": 1.4085357485161965e-06, "loss": 1.9379, "step": 33841 }, { "epoch": 0.83, "learning_rate": 1.4081272460415184e-06, "loss": 1.9203, "step": 33842 }, { "epoch": 0.83, "learning_rate": 1.4077187983256447e-06, "loss": 1.9989, "step": 33843 }, { "epoch": 0.83, "learning_rate": 1.4073104053711783e-06, "loss": 1.8366, "step": 33844 }, { "epoch": 0.83, "learning_rate": 1.4069020671807255e-06, "loss": 1.8846, "step": 33845 }, { "epoch": 0.83, "learning_rate": 1.4064937837568837e-06, "loss": 1.8011, "step": 33846 }, { "epoch": 0.83, "learning_rate": 1.4060855551022568e-06, "loss": 2.0423, "step": 33847 }, { "epoch": 0.83, "learning_rate": 1.4056773812194514e-06, "loss": 2.0252, "step": 33848 }, { "epoch": 0.83, "learning_rate": 1.4052692621110608e-06, "loss": 2.0777, "step": 33849 }, { "epoch": 0.83, "learning_rate": 1.4048611977796911e-06, "loss": 1.9696, "step": 33850 }, { "epoch": 0.83, "learning_rate": 1.4044531882279422e-06, "loss": 1.963, "step": 33851 }, { "epoch": 0.83, "learning_rate": 1.4040452334584164e-06, "loss": 1.9817, "step": 33852 }, { "epoch": 0.83, "learning_rate": 1.4036373334737097e-06, "loss": 1.9894, "step": 33853 }, { "epoch": 0.83, "learning_rate": 1.403229488276423e-06, "loss": 1.786, "step": 33854 }, { "epoch": 0.83, "learning_rate": 1.402821697869159e-06, "loss": 1.9716, "step": 33855 }, { "epoch": 0.83, "learning_rate": 1.402413962254512e-06, "loss": 1.9241, "step": 33856 }, { "epoch": 0.83, "learning_rate": 1.4020062814350822e-06, "loss": 2.0352, "step": 33857 }, { "epoch": 0.83, "learning_rate": 1.4015986554134708e-06, "loss": 2.0469, "step": 33858 }, { "epoch": 0.83, "learning_rate": 1.4011910841922715e-06, "loss": 1.9925, "step": 33859 }, { "epoch": 0.83, "learning_rate": 1.4007835677740856e-06, "loss": 1.9256, "step": 33860 }, { "epoch": 0.83, "learning_rate": 1.4003761061615062e-06, "loss": 1.8375, "step": 33861 }, { "epoch": 0.83, "learning_rate": 1.3999686993571326e-06, "loss": 2.0161, "step": 33862 }, { "epoch": 0.83, "learning_rate": 1.3995613473635628e-06, "loss": 1.836, "step": 33863 }, { "epoch": 0.83, "learning_rate": 1.39915405018339e-06, "loss": 1.909, "step": 33864 }, { "epoch": 0.83, "learning_rate": 1.3987468078192135e-06, "loss": 1.8276, "step": 33865 }, { "epoch": 0.83, "learning_rate": 1.3983396202736232e-06, "loss": 1.8688, "step": 33866 }, { "epoch": 0.83, "learning_rate": 1.3979324875492195e-06, "loss": 2.0959, "step": 33867 }, { "epoch": 0.83, "learning_rate": 1.3975254096485946e-06, "loss": 1.9598, "step": 33868 }, { "epoch": 0.83, "learning_rate": 1.397118386574343e-06, "loss": 1.9137, "step": 33869 }, { "epoch": 0.83, "learning_rate": 1.396711418329062e-06, "loss": 1.8579, "step": 33870 }, { "epoch": 0.83, "learning_rate": 1.3963045049153412e-06, "loss": 2.0668, "step": 33871 }, { "epoch": 0.83, "learning_rate": 1.3958976463357755e-06, "loss": 1.9282, "step": 33872 }, { "epoch": 0.83, "learning_rate": 1.3954908425929604e-06, "loss": 2.0084, "step": 33873 }, { "epoch": 0.83, "learning_rate": 1.3950840936894838e-06, "loss": 1.9621, "step": 33874 }, { "epoch": 0.83, "learning_rate": 1.3946773996279428e-06, "loss": 1.9551, "step": 33875 }, { "epoch": 0.84, "learning_rate": 1.394270760410925e-06, "loss": 1.9431, "step": 33876 }, { "epoch": 0.84, "learning_rate": 1.3938641760410243e-06, "loss": 1.8795, "step": 33877 }, { "epoch": 0.84, "learning_rate": 1.3934576465208338e-06, "loss": 1.772, "step": 33878 }, { "epoch": 0.84, "learning_rate": 1.3930511718529404e-06, "loss": 1.9131, "step": 33879 }, { "epoch": 0.84, "learning_rate": 1.3926447520399388e-06, "loss": 1.8673, "step": 33880 }, { "epoch": 0.84, "learning_rate": 1.3922383870844157e-06, "loss": 1.8777, "step": 33881 }, { "epoch": 0.84, "learning_rate": 1.3918320769889627e-06, "loss": 1.8721, "step": 33882 }, { "epoch": 0.84, "learning_rate": 1.3914258217561706e-06, "loss": 1.865, "step": 33883 }, { "epoch": 0.84, "learning_rate": 1.391019621388625e-06, "loss": 1.9654, "step": 33884 }, { "epoch": 0.84, "learning_rate": 1.3906134758889168e-06, "loss": 1.8144, "step": 33885 }, { "epoch": 0.84, "learning_rate": 1.3902073852596353e-06, "loss": 1.9731, "step": 33886 }, { "epoch": 0.84, "learning_rate": 1.389801349503368e-06, "loss": 1.8226, "step": 33887 }, { "epoch": 0.84, "learning_rate": 1.3893953686227047e-06, "loss": 1.895, "step": 33888 }, { "epoch": 0.84, "learning_rate": 1.3889894426202278e-06, "loss": 1.9145, "step": 33889 }, { "epoch": 0.84, "learning_rate": 1.3885835714985307e-06, "loss": 1.9596, "step": 33890 }, { "epoch": 0.84, "learning_rate": 1.388177755260194e-06, "loss": 2.0004, "step": 33891 }, { "epoch": 0.84, "learning_rate": 1.3877719939078072e-06, "loss": 1.9294, "step": 33892 }, { "epoch": 0.84, "learning_rate": 1.3873662874439586e-06, "loss": 1.9232, "step": 33893 }, { "epoch": 0.84, "learning_rate": 1.3869606358712296e-06, "loss": 1.9334, "step": 33894 }, { "epoch": 0.84, "learning_rate": 1.386555039192209e-06, "loss": 1.9086, "step": 33895 }, { "epoch": 0.84, "learning_rate": 1.386149497409477e-06, "loss": 1.9546, "step": 33896 }, { "epoch": 0.84, "learning_rate": 1.3857440105256226e-06, "loss": 1.7343, "step": 33897 }, { "epoch": 0.84, "learning_rate": 1.3853385785432315e-06, "loss": 1.9812, "step": 33898 }, { "epoch": 0.84, "learning_rate": 1.3849332014648819e-06, "loss": 1.7623, "step": 33899 }, { "epoch": 0.84, "learning_rate": 1.3845278792931639e-06, "loss": 1.948, "step": 33900 }, { "epoch": 0.84, "learning_rate": 1.384122612030654e-06, "loss": 1.7777, "step": 33901 }, { "epoch": 0.84, "learning_rate": 1.383717399679939e-06, "loss": 1.9828, "step": 33902 }, { "epoch": 0.84, "learning_rate": 1.3833122422436006e-06, "loss": 1.8235, "step": 33903 }, { "epoch": 0.84, "learning_rate": 1.3829071397242222e-06, "loss": 1.7864, "step": 33904 }, { "epoch": 0.84, "learning_rate": 1.3825020921243847e-06, "loss": 1.9049, "step": 33905 }, { "epoch": 0.84, "learning_rate": 1.382097099446671e-06, "loss": 2.0562, "step": 33906 }, { "epoch": 0.84, "learning_rate": 1.3816921616936595e-06, "loss": 1.9645, "step": 33907 }, { "epoch": 0.84, "learning_rate": 1.3812872788679354e-06, "loss": 1.9234, "step": 33908 }, { "epoch": 0.84, "learning_rate": 1.380882450972073e-06, "loss": 1.8425, "step": 33909 }, { "epoch": 0.84, "learning_rate": 1.380477678008656e-06, "loss": 1.8949, "step": 33910 }, { "epoch": 0.84, "learning_rate": 1.3800729599802653e-06, "loss": 2.1849, "step": 33911 }, { "epoch": 0.84, "learning_rate": 1.379668296889477e-06, "loss": 1.8389, "step": 33912 }, { "epoch": 0.84, "learning_rate": 1.3792636887388743e-06, "loss": 1.9573, "step": 33913 }, { "epoch": 0.84, "learning_rate": 1.3788591355310322e-06, "loss": 1.8018, "step": 33914 }, { "epoch": 0.84, "learning_rate": 1.3784546372685292e-06, "loss": 1.7804, "step": 33915 }, { "epoch": 0.84, "learning_rate": 1.378050193953947e-06, "loss": 1.8832, "step": 33916 }, { "epoch": 0.84, "learning_rate": 1.3776458055898578e-06, "loss": 1.856, "step": 33917 }, { "epoch": 0.84, "learning_rate": 1.377241472178845e-06, "loss": 1.8676, "step": 33918 }, { "epoch": 0.84, "learning_rate": 1.3768371937234803e-06, "loss": 1.9418, "step": 33919 }, { "epoch": 0.84, "learning_rate": 1.3764329702263412e-06, "loss": 1.8617, "step": 33920 }, { "epoch": 0.84, "learning_rate": 1.3760288016900058e-06, "loss": 1.8514, "step": 33921 }, { "epoch": 0.84, "learning_rate": 1.3756246881170498e-06, "loss": 2.0601, "step": 33922 }, { "epoch": 0.84, "learning_rate": 1.3752206295100502e-06, "loss": 2.0553, "step": 33923 }, { "epoch": 0.84, "learning_rate": 1.3748166258715779e-06, "loss": 1.7513, "step": 33924 }, { "epoch": 0.84, "learning_rate": 1.37441267720421e-06, "loss": 1.9015, "step": 33925 }, { "epoch": 0.84, "learning_rate": 1.3740087835105232e-06, "loss": 2.1194, "step": 33926 }, { "epoch": 0.84, "learning_rate": 1.3736049447930878e-06, "loss": 1.9545, "step": 33927 }, { "epoch": 0.84, "learning_rate": 1.3732011610544815e-06, "loss": 1.797, "step": 33928 }, { "epoch": 0.84, "learning_rate": 1.3727974322972726e-06, "loss": 1.9485, "step": 33929 }, { "epoch": 0.84, "learning_rate": 1.3723937585240377e-06, "loss": 1.9698, "step": 33930 }, { "epoch": 0.84, "learning_rate": 1.3719901397373514e-06, "loss": 1.953, "step": 33931 }, { "epoch": 0.84, "learning_rate": 1.3715865759397818e-06, "loss": 1.8833, "step": 33932 }, { "epoch": 0.84, "learning_rate": 1.3711830671339045e-06, "loss": 1.9527, "step": 33933 }, { "epoch": 0.84, "learning_rate": 1.3707796133222883e-06, "loss": 1.7924, "step": 33934 }, { "epoch": 0.84, "learning_rate": 1.370376214507505e-06, "loss": 1.9193, "step": 33935 }, { "epoch": 0.84, "learning_rate": 1.369972870692129e-06, "loss": 1.8849, "step": 33936 }, { "epoch": 0.84, "learning_rate": 1.3695695818787258e-06, "loss": 1.7854, "step": 33937 }, { "epoch": 0.84, "learning_rate": 1.369166348069868e-06, "loss": 1.9306, "step": 33938 }, { "epoch": 0.84, "learning_rate": 1.3687631692681258e-06, "loss": 1.9067, "step": 33939 }, { "epoch": 0.84, "learning_rate": 1.3683600454760693e-06, "loss": 1.9833, "step": 33940 }, { "epoch": 0.84, "learning_rate": 1.3679569766962686e-06, "loss": 1.9299, "step": 33941 }, { "epoch": 0.84, "learning_rate": 1.3675539629312896e-06, "loss": 2.0174, "step": 33942 }, { "epoch": 0.84, "learning_rate": 1.3671510041837045e-06, "loss": 1.9074, "step": 33943 }, { "epoch": 0.84, "learning_rate": 1.3667481004560779e-06, "loss": 1.8379, "step": 33944 }, { "epoch": 0.84, "learning_rate": 1.3663452517509778e-06, "loss": 2.0256, "step": 33945 }, { "epoch": 0.84, "learning_rate": 1.3659424580709756e-06, "loss": 2.1676, "step": 33946 }, { "epoch": 0.84, "learning_rate": 1.3655397194186349e-06, "loss": 1.9025, "step": 33947 }, { "epoch": 0.84, "learning_rate": 1.3651370357965244e-06, "loss": 1.9662, "step": 33948 }, { "epoch": 0.84, "learning_rate": 1.3647344072072078e-06, "loss": 1.7802, "step": 33949 }, { "epoch": 0.84, "learning_rate": 1.364331833653253e-06, "loss": 1.9728, "step": 33950 }, { "epoch": 0.84, "learning_rate": 1.363929315137228e-06, "loss": 1.7967, "step": 33951 }, { "epoch": 0.84, "learning_rate": 1.3635268516616929e-06, "loss": 1.9732, "step": 33952 }, { "epoch": 0.84, "learning_rate": 1.363124443229219e-06, "loss": 1.8339, "step": 33953 }, { "epoch": 0.84, "learning_rate": 1.3627220898423654e-06, "loss": 2.0839, "step": 33954 }, { "epoch": 0.84, "learning_rate": 1.3623197915036979e-06, "loss": 1.9213, "step": 33955 }, { "epoch": 0.84, "learning_rate": 1.361917548215782e-06, "loss": 1.9352, "step": 33956 }, { "epoch": 0.84, "learning_rate": 1.3615153599811804e-06, "loss": 2.0398, "step": 33957 }, { "epoch": 0.84, "learning_rate": 1.3611132268024586e-06, "loss": 1.8566, "step": 33958 }, { "epoch": 0.84, "learning_rate": 1.3607111486821755e-06, "loss": 1.985, "step": 33959 }, { "epoch": 0.84, "learning_rate": 1.3603091256228963e-06, "loss": 1.9878, "step": 33960 }, { "epoch": 0.84, "learning_rate": 1.3599071576271838e-06, "loss": 1.8333, "step": 33961 }, { "epoch": 0.84, "learning_rate": 1.3595052446975975e-06, "loss": 1.8648, "step": 33962 }, { "epoch": 0.84, "learning_rate": 1.3591033868366988e-06, "loss": 1.8505, "step": 33963 }, { "epoch": 0.84, "learning_rate": 1.3587015840470542e-06, "loss": 1.964, "step": 33964 }, { "epoch": 0.84, "learning_rate": 1.3582998363312172e-06, "loss": 2.0408, "step": 33965 }, { "epoch": 0.84, "learning_rate": 1.3578981436917537e-06, "loss": 2.0189, "step": 33966 }, { "epoch": 0.84, "learning_rate": 1.3574965061312195e-06, "loss": 1.9995, "step": 33967 }, { "epoch": 0.84, "learning_rate": 1.3570949236521768e-06, "loss": 2.0228, "step": 33968 }, { "epoch": 0.84, "learning_rate": 1.3566933962571872e-06, "loss": 1.8739, "step": 33969 }, { "epoch": 0.84, "learning_rate": 1.3562919239488049e-06, "loss": 1.8824, "step": 33970 }, { "epoch": 0.84, "learning_rate": 1.3558905067295936e-06, "loss": 1.9946, "step": 33971 }, { "epoch": 0.84, "learning_rate": 1.3554891446021068e-06, "loss": 1.9661, "step": 33972 }, { "epoch": 0.84, "learning_rate": 1.3550878375689048e-06, "loss": 2.0518, "step": 33973 }, { "epoch": 0.84, "learning_rate": 1.3546865856325452e-06, "loss": 1.8942, "step": 33974 }, { "epoch": 0.84, "learning_rate": 1.3542853887955853e-06, "loss": 1.8203, "step": 33975 }, { "epoch": 0.84, "learning_rate": 1.353884247060585e-06, "loss": 1.9821, "step": 33976 }, { "epoch": 0.84, "learning_rate": 1.3534831604300957e-06, "loss": 1.7002, "step": 33977 }, { "epoch": 0.84, "learning_rate": 1.3530821289066753e-06, "loss": 1.8953, "step": 33978 }, { "epoch": 0.84, "learning_rate": 1.352681152492883e-06, "loss": 2.1318, "step": 33979 }, { "epoch": 0.84, "learning_rate": 1.3522802311912697e-06, "loss": 1.9862, "step": 33980 }, { "epoch": 0.84, "learning_rate": 1.351879365004395e-06, "loss": 1.9719, "step": 33981 }, { "epoch": 0.84, "learning_rate": 1.3514785539348096e-06, "loss": 2.0087, "step": 33982 }, { "epoch": 0.84, "learning_rate": 1.3510777979850687e-06, "loss": 2.0028, "step": 33983 }, { "epoch": 0.84, "learning_rate": 1.350677097157731e-06, "loss": 1.9478, "step": 33984 }, { "epoch": 0.84, "learning_rate": 1.3502764514553446e-06, "loss": 1.9315, "step": 33985 }, { "epoch": 0.84, "learning_rate": 1.3498758608804663e-06, "loss": 1.7856, "step": 33986 }, { "epoch": 0.84, "learning_rate": 1.3494753254356474e-06, "loss": 1.9375, "step": 33987 }, { "epoch": 0.84, "learning_rate": 1.3490748451234403e-06, "loss": 1.7989, "step": 33988 }, { "epoch": 0.84, "learning_rate": 1.3486744199464009e-06, "loss": 1.95, "step": 33989 }, { "epoch": 0.84, "learning_rate": 1.3482740499070768e-06, "loss": 1.7906, "step": 33990 }, { "epoch": 0.84, "learning_rate": 1.3478737350080218e-06, "loss": 2.0449, "step": 33991 }, { "epoch": 0.84, "learning_rate": 1.3474734752517871e-06, "loss": 2.0422, "step": 33992 }, { "epoch": 0.84, "learning_rate": 1.347073270640923e-06, "loss": 1.9238, "step": 33993 }, { "epoch": 0.84, "learning_rate": 1.3466731211779838e-06, "loss": 1.8671, "step": 33994 }, { "epoch": 0.84, "learning_rate": 1.3462730268655145e-06, "loss": 1.8927, "step": 33995 }, { "epoch": 0.84, "learning_rate": 1.3458729877060694e-06, "loss": 1.855, "step": 33996 }, { "epoch": 0.84, "learning_rate": 1.3454730037021946e-06, "loss": 2.0429, "step": 33997 }, { "epoch": 0.84, "learning_rate": 1.34507307485644e-06, "loss": 1.9157, "step": 33998 }, { "epoch": 0.84, "learning_rate": 1.344673201171357e-06, "loss": 2.0939, "step": 33999 }, { "epoch": 0.84, "learning_rate": 1.3442733826494913e-06, "loss": 1.7631, "step": 34000 }, { "epoch": 0.84, "learning_rate": 1.3438736192933932e-06, "loss": 1.9229, "step": 34001 }, { "epoch": 0.84, "learning_rate": 1.3434739111056083e-06, "loss": 1.8841, "step": 34002 }, { "epoch": 0.84, "learning_rate": 1.3430742580886847e-06, "loss": 1.7403, "step": 34003 }, { "epoch": 0.84, "learning_rate": 1.3426746602451723e-06, "loss": 1.9111, "step": 34004 }, { "epoch": 0.84, "learning_rate": 1.3422751175776139e-06, "loss": 1.7959, "step": 34005 }, { "epoch": 0.84, "learning_rate": 1.3418756300885593e-06, "loss": 1.8625, "step": 34006 }, { "epoch": 0.84, "learning_rate": 1.3414761977805512e-06, "loss": 1.9703, "step": 34007 }, { "epoch": 0.84, "learning_rate": 1.3410768206561364e-06, "loss": 1.8059, "step": 34008 }, { "epoch": 0.84, "learning_rate": 1.3406774987178617e-06, "loss": 1.8655, "step": 34009 }, { "epoch": 0.84, "learning_rate": 1.3402782319682705e-06, "loss": 1.9087, "step": 34010 }, { "epoch": 0.84, "learning_rate": 1.339879020409911e-06, "loss": 1.8301, "step": 34011 }, { "epoch": 0.84, "learning_rate": 1.3394798640453223e-06, "loss": 2.053, "step": 34012 }, { "epoch": 0.84, "learning_rate": 1.3390807628770509e-06, "loss": 1.7776, "step": 34013 }, { "epoch": 0.84, "learning_rate": 1.3386817169076428e-06, "loss": 1.9419, "step": 34014 }, { "epoch": 0.84, "learning_rate": 1.338282726139636e-06, "loss": 2.0494, "step": 34015 }, { "epoch": 0.84, "learning_rate": 1.3378837905755782e-06, "loss": 1.9254, "step": 34016 }, { "epoch": 0.84, "learning_rate": 1.3374849102180077e-06, "loss": 1.9167, "step": 34017 }, { "epoch": 0.84, "learning_rate": 1.33708608506947e-06, "loss": 1.9583, "step": 34018 }, { "epoch": 0.84, "learning_rate": 1.3366873151325065e-06, "loss": 1.8111, "step": 34019 }, { "epoch": 0.84, "learning_rate": 1.3362886004096575e-06, "loss": 1.9651, "step": 34020 }, { "epoch": 0.84, "learning_rate": 1.3358899409034654e-06, "loss": 2.1099, "step": 34021 }, { "epoch": 0.84, "learning_rate": 1.335491336616469e-06, "loss": 1.9377, "step": 34022 }, { "epoch": 0.84, "learning_rate": 1.3350927875512098e-06, "loss": 1.9657, "step": 34023 }, { "epoch": 0.84, "learning_rate": 1.3346942937102302e-06, "loss": 2.0237, "step": 34024 }, { "epoch": 0.84, "learning_rate": 1.334295855096065e-06, "loss": 1.946, "step": 34025 }, { "epoch": 0.84, "learning_rate": 1.3338974717112575e-06, "loss": 1.84, "step": 34026 }, { "epoch": 0.84, "learning_rate": 1.3334991435583444e-06, "loss": 1.855, "step": 34027 }, { "epoch": 0.84, "learning_rate": 1.3331008706398662e-06, "loss": 2.0114, "step": 34028 }, { "epoch": 0.84, "learning_rate": 1.3327026529583632e-06, "loss": 1.8509, "step": 34029 }, { "epoch": 0.84, "learning_rate": 1.3323044905163684e-06, "loss": 1.8786, "step": 34030 }, { "epoch": 0.84, "learning_rate": 1.3319063833164213e-06, "loss": 2.0403, "step": 34031 }, { "epoch": 0.84, "learning_rate": 1.3315083313610621e-06, "loss": 1.8289, "step": 34032 }, { "epoch": 0.84, "learning_rate": 1.3311103346528232e-06, "loss": 1.9555, "step": 34033 }, { "epoch": 0.84, "learning_rate": 1.3307123931942456e-06, "loss": 1.9583, "step": 34034 }, { "epoch": 0.84, "learning_rate": 1.330314506987861e-06, "loss": 1.9667, "step": 34035 }, { "epoch": 0.84, "learning_rate": 1.329916676036207e-06, "loss": 1.9491, "step": 34036 }, { "epoch": 0.84, "learning_rate": 1.329518900341822e-06, "loss": 1.9417, "step": 34037 }, { "epoch": 0.84, "learning_rate": 1.3291211799072368e-06, "loss": 1.7582, "step": 34038 }, { "epoch": 0.84, "learning_rate": 1.3287235147349898e-06, "loss": 1.9042, "step": 34039 }, { "epoch": 0.84, "learning_rate": 1.328325904827611e-06, "loss": 1.9202, "step": 34040 }, { "epoch": 0.84, "learning_rate": 1.3279283501876383e-06, "loss": 2.1673, "step": 34041 }, { "epoch": 0.84, "learning_rate": 1.3275308508176054e-06, "loss": 1.856, "step": 34042 }, { "epoch": 0.84, "learning_rate": 1.3271334067200437e-06, "loss": 1.9315, "step": 34043 }, { "epoch": 0.84, "learning_rate": 1.3267360178974865e-06, "loss": 1.868, "step": 34044 }, { "epoch": 0.84, "learning_rate": 1.3263386843524672e-06, "loss": 1.9318, "step": 34045 }, { "epoch": 0.84, "learning_rate": 1.3259414060875186e-06, "loss": 1.9651, "step": 34046 }, { "epoch": 0.84, "learning_rate": 1.325544183105174e-06, "loss": 1.9945, "step": 34047 }, { "epoch": 0.84, "learning_rate": 1.3251470154079616e-06, "loss": 1.7738, "step": 34048 }, { "epoch": 0.84, "learning_rate": 1.3247499029984157e-06, "loss": 2.0412, "step": 34049 }, { "epoch": 0.84, "learning_rate": 1.3243528458790645e-06, "loss": 1.8735, "step": 34050 }, { "epoch": 0.84, "learning_rate": 1.3239558440524392e-06, "loss": 1.8655, "step": 34051 }, { "epoch": 0.84, "learning_rate": 1.3235588975210733e-06, "loss": 1.7713, "step": 34052 }, { "epoch": 0.84, "learning_rate": 1.3231620062874918e-06, "loss": 1.8209, "step": 34053 }, { "epoch": 0.84, "learning_rate": 1.322765170354229e-06, "loss": 1.8889, "step": 34054 }, { "epoch": 0.84, "learning_rate": 1.3223683897238094e-06, "loss": 1.8633, "step": 34055 }, { "epoch": 0.84, "learning_rate": 1.3219716643987635e-06, "loss": 2.0087, "step": 34056 }, { "epoch": 0.84, "learning_rate": 1.3215749943816225e-06, "loss": 1.9421, "step": 34057 }, { "epoch": 0.84, "learning_rate": 1.3211783796749101e-06, "loss": 1.7786, "step": 34058 }, { "epoch": 0.84, "learning_rate": 1.3207818202811584e-06, "loss": 1.9927, "step": 34059 }, { "epoch": 0.84, "learning_rate": 1.3203853162028913e-06, "loss": 1.9287, "step": 34060 }, { "epoch": 0.84, "learning_rate": 1.3199888674426364e-06, "loss": 1.9638, "step": 34061 }, { "epoch": 0.84, "learning_rate": 1.3195924740029208e-06, "loss": 1.8321, "step": 34062 }, { "epoch": 0.84, "learning_rate": 1.3191961358862727e-06, "loss": 1.7576, "step": 34063 }, { "epoch": 0.84, "learning_rate": 1.3187998530952173e-06, "loss": 1.862, "step": 34064 }, { "epoch": 0.84, "learning_rate": 1.3184036256322786e-06, "loss": 1.883, "step": 34065 }, { "epoch": 0.84, "learning_rate": 1.3180074534999821e-06, "loss": 1.8332, "step": 34066 }, { "epoch": 0.84, "learning_rate": 1.3176113367008557e-06, "loss": 1.9063, "step": 34067 }, { "epoch": 0.84, "learning_rate": 1.317215275237419e-06, "loss": 1.9326, "step": 34068 }, { "epoch": 0.84, "learning_rate": 1.3168192691122017e-06, "loss": 2.0095, "step": 34069 }, { "epoch": 0.84, "learning_rate": 1.316423318327723e-06, "loss": 1.9567, "step": 34070 }, { "epoch": 0.84, "learning_rate": 1.3160274228865088e-06, "loss": 1.9663, "step": 34071 }, { "epoch": 0.84, "learning_rate": 1.3156315827910836e-06, "loss": 1.8404, "step": 34072 }, { "epoch": 0.84, "learning_rate": 1.3152357980439656e-06, "loss": 1.9022, "step": 34073 }, { "epoch": 0.84, "learning_rate": 1.3148400686476837e-06, "loss": 1.8197, "step": 34074 }, { "epoch": 0.84, "learning_rate": 1.3144443946047535e-06, "loss": 1.9338, "step": 34075 }, { "epoch": 0.84, "learning_rate": 1.314048775917699e-06, "loss": 1.8781, "step": 34076 }, { "epoch": 0.84, "learning_rate": 1.3136532125890455e-06, "loss": 2.0093, "step": 34077 }, { "epoch": 0.84, "learning_rate": 1.3132577046213091e-06, "loss": 1.9586, "step": 34078 }, { "epoch": 0.84, "learning_rate": 1.312862252017011e-06, "loss": 1.9111, "step": 34079 }, { "epoch": 0.84, "learning_rate": 1.3124668547786733e-06, "loss": 2.0141, "step": 34080 }, { "epoch": 0.84, "learning_rate": 1.3120715129088156e-06, "loss": 1.8932, "step": 34081 }, { "epoch": 0.84, "learning_rate": 1.3116762264099604e-06, "loss": 2.0508, "step": 34082 }, { "epoch": 0.84, "learning_rate": 1.3112809952846207e-06, "loss": 2.0987, "step": 34083 }, { "epoch": 0.84, "learning_rate": 1.310885819535319e-06, "loss": 1.9486, "step": 34084 }, { "epoch": 0.84, "learning_rate": 1.310490699164576e-06, "loss": 1.914, "step": 34085 }, { "epoch": 0.84, "learning_rate": 1.310095634174906e-06, "loss": 1.961, "step": 34086 }, { "epoch": 0.84, "learning_rate": 1.3097006245688303e-06, "loss": 1.8569, "step": 34087 }, { "epoch": 0.84, "learning_rate": 1.3093056703488627e-06, "loss": 1.8911, "step": 34088 }, { "epoch": 0.84, "learning_rate": 1.308910771517521e-06, "loss": 1.973, "step": 34089 }, { "epoch": 0.84, "learning_rate": 1.3085159280773263e-06, "loss": 1.9877, "step": 34090 }, { "epoch": 0.84, "learning_rate": 1.3081211400307903e-06, "loss": 1.8023, "step": 34091 }, { "epoch": 0.84, "learning_rate": 1.307726407380433e-06, "loss": 2.0913, "step": 34092 }, { "epoch": 0.84, "learning_rate": 1.3073317301287658e-06, "loss": 2.0178, "step": 34093 }, { "epoch": 0.84, "learning_rate": 1.3069371082783055e-06, "loss": 1.692, "step": 34094 }, { "epoch": 0.84, "learning_rate": 1.3065425418315714e-06, "loss": 1.9551, "step": 34095 }, { "epoch": 0.84, "learning_rate": 1.3061480307910712e-06, "loss": 1.9449, "step": 34096 }, { "epoch": 0.84, "learning_rate": 1.3057535751593243e-06, "loss": 1.9184, "step": 34097 }, { "epoch": 0.84, "learning_rate": 1.3053591749388428e-06, "loss": 2.0136, "step": 34098 }, { "epoch": 0.84, "learning_rate": 1.3049648301321404e-06, "loss": 2.1375, "step": 34099 }, { "epoch": 0.84, "learning_rate": 1.304570540741733e-06, "loss": 1.8514, "step": 34100 }, { "epoch": 0.84, "learning_rate": 1.3041763067701297e-06, "loss": 1.9152, "step": 34101 }, { "epoch": 0.84, "learning_rate": 1.303782128219846e-06, "loss": 2.0733, "step": 34102 }, { "epoch": 0.84, "learning_rate": 1.3033880050933912e-06, "loss": 2.0354, "step": 34103 }, { "epoch": 0.84, "learning_rate": 1.3029939373932787e-06, "loss": 1.9163, "step": 34104 }, { "epoch": 0.84, "learning_rate": 1.3025999251220224e-06, "loss": 2.02, "step": 34105 }, { "epoch": 0.84, "learning_rate": 1.3022059682821287e-06, "loss": 1.8972, "step": 34106 }, { "epoch": 0.84, "learning_rate": 1.3018120668761126e-06, "loss": 2.0159, "step": 34107 }, { "epoch": 0.84, "learning_rate": 1.301418220906482e-06, "loss": 2.1585, "step": 34108 }, { "epoch": 0.84, "learning_rate": 1.3010244303757468e-06, "loss": 1.8911, "step": 34109 }, { "epoch": 0.84, "learning_rate": 1.3006306952864201e-06, "loss": 2.0134, "step": 34110 }, { "epoch": 0.84, "learning_rate": 1.3002370156410072e-06, "loss": 1.7565, "step": 34111 }, { "epoch": 0.84, "learning_rate": 1.2998433914420205e-06, "loss": 2.092, "step": 34112 }, { "epoch": 0.84, "learning_rate": 1.299449822691966e-06, "loss": 1.9265, "step": 34113 }, { "epoch": 0.84, "learning_rate": 1.2990563093933528e-06, "loss": 1.9009, "step": 34114 }, { "epoch": 0.84, "learning_rate": 1.2986628515486889e-06, "loss": 2.0477, "step": 34115 }, { "epoch": 0.84, "learning_rate": 1.298269449160483e-06, "loss": 1.8617, "step": 34116 }, { "epoch": 0.84, "learning_rate": 1.2978761022312436e-06, "loss": 1.8709, "step": 34117 }, { "epoch": 0.84, "learning_rate": 1.297482810763473e-06, "loss": 2.0003, "step": 34118 }, { "epoch": 0.84, "learning_rate": 1.2970895747596813e-06, "loss": 1.95, "step": 34119 }, { "epoch": 0.84, "learning_rate": 1.2966963942223754e-06, "loss": 1.8528, "step": 34120 }, { "epoch": 0.84, "learning_rate": 1.2963032691540578e-06, "loss": 1.7604, "step": 34121 }, { "epoch": 0.84, "learning_rate": 1.2959101995572388e-06, "loss": 1.963, "step": 34122 }, { "epoch": 0.84, "learning_rate": 1.2955171854344173e-06, "loss": 1.9726, "step": 34123 }, { "epoch": 0.84, "learning_rate": 1.2951242267881026e-06, "loss": 1.9559, "step": 34124 }, { "epoch": 0.84, "learning_rate": 1.2947313236208003e-06, "loss": 1.7926, "step": 34125 }, { "epoch": 0.84, "learning_rate": 1.2943384759350087e-06, "loss": 1.9662, "step": 34126 }, { "epoch": 0.84, "learning_rate": 1.2939456837332388e-06, "loss": 2.0217, "step": 34127 }, { "epoch": 0.84, "learning_rate": 1.2935529470179875e-06, "loss": 2.0001, "step": 34128 }, { "epoch": 0.84, "learning_rate": 1.2931602657917597e-06, "loss": 1.8683, "step": 34129 }, { "epoch": 0.84, "learning_rate": 1.2927676400570621e-06, "loss": 1.8527, "step": 34130 }, { "epoch": 0.84, "learning_rate": 1.2923750698163917e-06, "loss": 1.9612, "step": 34131 }, { "epoch": 0.84, "learning_rate": 1.2919825550722532e-06, "loss": 1.9537, "step": 34132 }, { "epoch": 0.84, "learning_rate": 1.2915900958271465e-06, "loss": 1.8613, "step": 34133 }, { "epoch": 0.84, "learning_rate": 1.2911976920835756e-06, "loss": 1.9676, "step": 34134 }, { "epoch": 0.84, "learning_rate": 1.2908053438440404e-06, "loss": 1.9092, "step": 34135 }, { "epoch": 0.84, "learning_rate": 1.2904130511110403e-06, "loss": 1.8043, "step": 34136 }, { "epoch": 0.84, "learning_rate": 1.2900208138870774e-06, "loss": 2.099, "step": 34137 }, { "epoch": 0.84, "learning_rate": 1.2896286321746487e-06, "loss": 1.9081, "step": 34138 }, { "epoch": 0.84, "learning_rate": 1.2892365059762557e-06, "loss": 2.0175, "step": 34139 }, { "epoch": 0.84, "learning_rate": 1.2888444352943986e-06, "loss": 2.0245, "step": 34140 }, { "epoch": 0.84, "learning_rate": 1.2884524201315728e-06, "loss": 1.92, "step": 34141 }, { "epoch": 0.84, "learning_rate": 1.2880604604902791e-06, "loss": 1.8532, "step": 34142 }, { "epoch": 0.84, "learning_rate": 1.2876685563730173e-06, "loss": 1.8309, "step": 34143 }, { "epoch": 0.84, "learning_rate": 1.28727670778228e-06, "loss": 1.9476, "step": 34144 }, { "epoch": 0.84, "learning_rate": 1.286884914720571e-06, "loss": 1.9826, "step": 34145 }, { "epoch": 0.84, "learning_rate": 1.286493177190381e-06, "loss": 2.0658, "step": 34146 }, { "epoch": 0.84, "learning_rate": 1.2861014951942108e-06, "loss": 1.8695, "step": 34147 }, { "epoch": 0.84, "learning_rate": 1.285709868734556e-06, "loss": 2.0349, "step": 34148 }, { "epoch": 0.84, "learning_rate": 1.2853182978139112e-06, "loss": 1.7895, "step": 34149 }, { "epoch": 0.84, "learning_rate": 1.2849267824347722e-06, "loss": 1.8937, "step": 34150 }, { "epoch": 0.84, "learning_rate": 1.284535322599636e-06, "loss": 1.9612, "step": 34151 }, { "epoch": 0.84, "learning_rate": 1.284143918310995e-06, "loss": 2.245, "step": 34152 }, { "epoch": 0.84, "learning_rate": 1.2837525695713482e-06, "loss": 1.8405, "step": 34153 }, { "epoch": 0.84, "learning_rate": 1.2833612763831848e-06, "loss": 1.9277, "step": 34154 }, { "epoch": 0.84, "learning_rate": 1.2829700387490028e-06, "loss": 1.8884, "step": 34155 }, { "epoch": 0.84, "learning_rate": 1.2825788566712916e-06, "loss": 1.8767, "step": 34156 }, { "epoch": 0.84, "learning_rate": 1.2821877301525453e-06, "loss": 1.9727, "step": 34157 }, { "epoch": 0.84, "learning_rate": 1.2817966591952612e-06, "loss": 1.946, "step": 34158 }, { "epoch": 0.84, "learning_rate": 1.281405643801925e-06, "loss": 2.0722, "step": 34159 }, { "epoch": 0.84, "learning_rate": 1.2810146839750348e-06, "loss": 1.9714, "step": 34160 }, { "epoch": 0.84, "learning_rate": 1.2806237797170761e-06, "loss": 1.8596, "step": 34161 }, { "epoch": 0.84, "learning_rate": 1.280232931030545e-06, "loss": 1.9374, "step": 34162 }, { "epoch": 0.84, "learning_rate": 1.2798421379179326e-06, "loss": 2.0581, "step": 34163 }, { "epoch": 0.84, "learning_rate": 1.279451400381726e-06, "loss": 1.8774, "step": 34164 }, { "epoch": 0.84, "learning_rate": 1.2790607184244196e-06, "loss": 1.7829, "step": 34165 }, { "epoch": 0.84, "learning_rate": 1.2786700920484984e-06, "loss": 1.8231, "step": 34166 }, { "epoch": 0.84, "learning_rate": 1.2782795212564558e-06, "loss": 2.0455, "step": 34167 }, { "epoch": 0.84, "learning_rate": 1.277889006050781e-06, "loss": 1.9247, "step": 34168 }, { "epoch": 0.84, "learning_rate": 1.2774985464339607e-06, "loss": 1.7323, "step": 34169 }, { "epoch": 0.84, "learning_rate": 1.2771081424084875e-06, "loss": 1.9303, "step": 34170 }, { "epoch": 0.84, "learning_rate": 1.2767177939768438e-06, "loss": 1.9404, "step": 34171 }, { "epoch": 0.84, "learning_rate": 1.276327501141521e-06, "loss": 1.8255, "step": 34172 }, { "epoch": 0.84, "learning_rate": 1.2759372639050082e-06, "loss": 1.9641, "step": 34173 }, { "epoch": 0.84, "learning_rate": 1.2755470822697891e-06, "loss": 1.893, "step": 34174 }, { "epoch": 0.84, "learning_rate": 1.2751569562383525e-06, "loss": 1.8972, "step": 34175 }, { "epoch": 0.84, "learning_rate": 1.2747668858131822e-06, "loss": 1.9073, "step": 34176 }, { "epoch": 0.84, "learning_rate": 1.2743768709967663e-06, "loss": 1.8918, "step": 34177 }, { "epoch": 0.84, "learning_rate": 1.2739869117915915e-06, "loss": 1.98, "step": 34178 }, { "epoch": 0.84, "learning_rate": 1.2735970082001402e-06, "loss": 2.1061, "step": 34179 }, { "epoch": 0.84, "learning_rate": 1.273207160224902e-06, "loss": 1.9809, "step": 34180 }, { "epoch": 0.84, "learning_rate": 1.2728173678683552e-06, "loss": 2.0006, "step": 34181 }, { "epoch": 0.84, "learning_rate": 1.2724276311329874e-06, "loss": 1.9954, "step": 34182 }, { "epoch": 0.84, "learning_rate": 1.2720379500212854e-06, "loss": 1.8989, "step": 34183 }, { "epoch": 0.84, "learning_rate": 1.271648324535728e-06, "loss": 2.1726, "step": 34184 }, { "epoch": 0.84, "learning_rate": 1.2712587546787992e-06, "loss": 1.665, "step": 34185 }, { "epoch": 0.84, "learning_rate": 1.2708692404529843e-06, "loss": 1.8774, "step": 34186 }, { "epoch": 0.84, "learning_rate": 1.270479781860763e-06, "loss": 2.1067, "step": 34187 }, { "epoch": 0.84, "learning_rate": 1.2700903789046227e-06, "loss": 1.8191, "step": 34188 }, { "epoch": 0.84, "learning_rate": 1.2697010315870384e-06, "loss": 1.907, "step": 34189 }, { "epoch": 0.84, "learning_rate": 1.269311739910496e-06, "loss": 1.8874, "step": 34190 }, { "epoch": 0.84, "learning_rate": 1.2689225038774745e-06, "loss": 1.858, "step": 34191 }, { "epoch": 0.84, "learning_rate": 1.268533323490454e-06, "loss": 2.0099, "step": 34192 }, { "epoch": 0.84, "learning_rate": 1.2681441987519184e-06, "loss": 1.6458, "step": 34193 }, { "epoch": 0.84, "learning_rate": 1.2677551296643442e-06, "loss": 1.9373, "step": 34194 }, { "epoch": 0.84, "learning_rate": 1.267366116230213e-06, "loss": 1.8356, "step": 34195 }, { "epoch": 0.84, "learning_rate": 1.2669771584520029e-06, "loss": 1.9132, "step": 34196 }, { "epoch": 0.84, "learning_rate": 1.2665882563321918e-06, "loss": 2.0389, "step": 34197 }, { "epoch": 0.84, "learning_rate": 1.2661994098732634e-06, "loss": 1.9798, "step": 34198 }, { "epoch": 0.84, "learning_rate": 1.2658106190776898e-06, "loss": 1.9377, "step": 34199 }, { "epoch": 0.84, "learning_rate": 1.2654218839479515e-06, "loss": 1.8994, "step": 34200 }, { "epoch": 0.84, "learning_rate": 1.2650332044865276e-06, "loss": 1.8644, "step": 34201 }, { "epoch": 0.84, "learning_rate": 1.264644580695893e-06, "loss": 2.0067, "step": 34202 }, { "epoch": 0.84, "learning_rate": 1.2642560125785241e-06, "loss": 1.8653, "step": 34203 }, { "epoch": 0.84, "learning_rate": 1.2638675001368982e-06, "loss": 1.8521, "step": 34204 }, { "epoch": 0.84, "learning_rate": 1.2634790433734924e-06, "loss": 1.7659, "step": 34205 }, { "epoch": 0.84, "learning_rate": 1.263090642290784e-06, "loss": 1.9576, "step": 34206 }, { "epoch": 0.84, "learning_rate": 1.262702296891244e-06, "loss": 2.0346, "step": 34207 }, { "epoch": 0.84, "learning_rate": 1.262314007177352e-06, "loss": 1.9202, "step": 34208 }, { "epoch": 0.84, "learning_rate": 1.2619257731515778e-06, "loss": 1.8448, "step": 34209 }, { "epoch": 0.84, "learning_rate": 1.2615375948163988e-06, "loss": 2.0578, "step": 34210 }, { "epoch": 0.84, "learning_rate": 1.2611494721742901e-06, "loss": 1.9175, "step": 34211 }, { "epoch": 0.84, "learning_rate": 1.2607614052277218e-06, "loss": 1.8609, "step": 34212 }, { "epoch": 0.84, "learning_rate": 1.2603733939791718e-06, "loss": 2.0563, "step": 34213 }, { "epoch": 0.84, "learning_rate": 1.2599854384311073e-06, "loss": 1.84, "step": 34214 }, { "epoch": 0.84, "learning_rate": 1.2595975385860049e-06, "loss": 1.815, "step": 34215 }, { "epoch": 0.84, "learning_rate": 1.2592096944463372e-06, "loss": 2.0507, "step": 34216 }, { "epoch": 0.84, "learning_rate": 1.2588219060145723e-06, "loss": 1.8052, "step": 34217 }, { "epoch": 0.84, "learning_rate": 1.2584341732931838e-06, "loss": 1.9528, "step": 34218 }, { "epoch": 0.84, "learning_rate": 1.2580464962846462e-06, "loss": 1.8099, "step": 34219 }, { "epoch": 0.84, "learning_rate": 1.2576588749914243e-06, "loss": 1.9538, "step": 34220 }, { "epoch": 0.84, "learning_rate": 1.2572713094159927e-06, "loss": 1.6712, "step": 34221 }, { "epoch": 0.84, "learning_rate": 1.2568837995608186e-06, "loss": 1.9939, "step": 34222 }, { "epoch": 0.84, "learning_rate": 1.2564963454283763e-06, "loss": 2.0269, "step": 34223 }, { "epoch": 0.84, "learning_rate": 1.2561089470211295e-06, "loss": 1.8998, "step": 34224 }, { "epoch": 0.84, "learning_rate": 1.255721604341551e-06, "loss": 1.9018, "step": 34225 }, { "epoch": 0.84, "learning_rate": 1.2553343173921095e-06, "loss": 1.9391, "step": 34226 }, { "epoch": 0.84, "learning_rate": 1.254947086175271e-06, "loss": 2.0488, "step": 34227 }, { "epoch": 0.84, "learning_rate": 1.2545599106935058e-06, "loss": 1.9779, "step": 34228 }, { "epoch": 0.84, "learning_rate": 1.2541727909492795e-06, "loss": 1.7819, "step": 34229 }, { "epoch": 0.84, "learning_rate": 1.2537857269450593e-06, "loss": 1.8364, "step": 34230 }, { "epoch": 0.84, "learning_rate": 1.2533987186833152e-06, "loss": 1.85, "step": 34231 }, { "epoch": 0.84, "learning_rate": 1.2530117661665098e-06, "loss": 1.8508, "step": 34232 }, { "epoch": 0.84, "learning_rate": 1.2526248693971121e-06, "loss": 1.9046, "step": 34233 }, { "epoch": 0.84, "learning_rate": 1.2522380283775848e-06, "loss": 1.7799, "step": 34234 }, { "epoch": 0.84, "learning_rate": 1.2518512431103958e-06, "loss": 1.9309, "step": 34235 }, { "epoch": 0.84, "learning_rate": 1.2514645135980096e-06, "loss": 1.9079, "step": 34236 }, { "epoch": 0.84, "learning_rate": 1.2510778398428934e-06, "loss": 1.8202, "step": 34237 }, { "epoch": 0.84, "learning_rate": 1.2506912218475077e-06, "loss": 1.7822, "step": 34238 }, { "epoch": 0.84, "learning_rate": 1.2503046596143176e-06, "loss": 1.8626, "step": 34239 }, { "epoch": 0.84, "learning_rate": 1.249918153145787e-06, "loss": 2.0599, "step": 34240 }, { "epoch": 0.84, "learning_rate": 1.2495317024443832e-06, "loss": 1.9834, "step": 34241 }, { "epoch": 0.84, "learning_rate": 1.2491453075125614e-06, "loss": 1.8524, "step": 34242 }, { "epoch": 0.84, "learning_rate": 1.2487589683527922e-06, "loss": 1.8823, "step": 34243 }, { "epoch": 0.84, "learning_rate": 1.2483726849675304e-06, "loss": 1.8837, "step": 34244 }, { "epoch": 0.84, "learning_rate": 1.2479864573592426e-06, "loss": 1.9121, "step": 34245 }, { "epoch": 0.84, "learning_rate": 1.2476002855303915e-06, "loss": 1.8922, "step": 34246 }, { "epoch": 0.84, "learning_rate": 1.2472141694834338e-06, "loss": 1.7543, "step": 34247 }, { "epoch": 0.84, "learning_rate": 1.2468281092208345e-06, "loss": 1.8159, "step": 34248 }, { "epoch": 0.84, "learning_rate": 1.2464421047450503e-06, "loss": 1.8378, "step": 34249 }, { "epoch": 0.84, "learning_rate": 1.2460561560585427e-06, "loss": 1.8553, "step": 34250 }, { "epoch": 0.84, "learning_rate": 1.245670263163775e-06, "loss": 1.9232, "step": 34251 }, { "epoch": 0.84, "learning_rate": 1.2452844260632014e-06, "loss": 1.9638, "step": 34252 }, { "epoch": 0.84, "learning_rate": 1.2448986447592825e-06, "loss": 2.0862, "step": 34253 }, { "epoch": 0.84, "learning_rate": 1.2445129192544803e-06, "loss": 1.9693, "step": 34254 }, { "epoch": 0.84, "learning_rate": 1.2441272495512492e-06, "loss": 1.9345, "step": 34255 }, { "epoch": 0.84, "learning_rate": 1.2437416356520483e-06, "loss": 2.0746, "step": 34256 }, { "epoch": 0.84, "learning_rate": 1.2433560775593344e-06, "loss": 1.7511, "step": 34257 }, { "epoch": 0.84, "learning_rate": 1.2429705752755693e-06, "loss": 2.0006, "step": 34258 }, { "epoch": 0.84, "learning_rate": 1.2425851288032043e-06, "loss": 1.8293, "step": 34259 }, { "epoch": 0.84, "learning_rate": 1.242199738144697e-06, "loss": 1.8641, "step": 34260 }, { "epoch": 0.84, "learning_rate": 1.241814403302508e-06, "loss": 1.9552, "step": 34261 }, { "epoch": 0.84, "learning_rate": 1.2414291242790877e-06, "loss": 1.7734, "step": 34262 }, { "epoch": 0.84, "learning_rate": 1.2410439010768949e-06, "loss": 1.9049, "step": 34263 }, { "epoch": 0.84, "learning_rate": 1.2406587336983843e-06, "loss": 1.6772, "step": 34264 }, { "epoch": 0.84, "learning_rate": 1.2402736221460088e-06, "loss": 2.0733, "step": 34265 }, { "epoch": 0.84, "learning_rate": 1.239888566422227e-06, "loss": 2.0287, "step": 34266 }, { "epoch": 0.84, "learning_rate": 1.2395035665294875e-06, "loss": 2.1027, "step": 34267 }, { "epoch": 0.84, "learning_rate": 1.2391186224702457e-06, "loss": 1.775, "step": 34268 }, { "epoch": 0.84, "learning_rate": 1.2387337342469597e-06, "loss": 1.9115, "step": 34269 }, { "epoch": 0.84, "learning_rate": 1.2383489018620753e-06, "loss": 1.9197, "step": 34270 }, { "epoch": 0.84, "learning_rate": 1.2379641253180497e-06, "loss": 2.0021, "step": 34271 }, { "epoch": 0.84, "learning_rate": 1.237579404617336e-06, "loss": 1.7574, "step": 34272 }, { "epoch": 0.84, "learning_rate": 1.2371947397623829e-06, "loss": 1.916, "step": 34273 }, { "epoch": 0.84, "learning_rate": 1.2368101307556424e-06, "loss": 1.8907, "step": 34274 }, { "epoch": 0.84, "learning_rate": 1.236425577599567e-06, "loss": 1.9477, "step": 34275 }, { "epoch": 0.84, "learning_rate": 1.2360410802966104e-06, "loss": 1.8392, "step": 34276 }, { "epoch": 0.84, "learning_rate": 1.2356566388492176e-06, "loss": 1.8787, "step": 34277 }, { "epoch": 0.84, "learning_rate": 1.2352722532598404e-06, "loss": 1.7003, "step": 34278 }, { "epoch": 0.84, "learning_rate": 1.2348879235309318e-06, "loss": 1.8241, "step": 34279 }, { "epoch": 0.84, "learning_rate": 1.2345036496649376e-06, "loss": 2.0242, "step": 34280 }, { "epoch": 0.84, "learning_rate": 1.2341194316643102e-06, "loss": 2.0958, "step": 34281 }, { "epoch": 0.85, "learning_rate": 1.233735269531493e-06, "loss": 1.89, "step": 34282 }, { "epoch": 0.85, "learning_rate": 1.2333511632689388e-06, "loss": 2.0883, "step": 34283 }, { "epoch": 0.85, "learning_rate": 1.2329671128790965e-06, "loss": 1.8996, "step": 34284 }, { "epoch": 0.85, "learning_rate": 1.23258311836441e-06, "loss": 1.9853, "step": 34285 }, { "epoch": 0.85, "learning_rate": 1.2321991797273314e-06, "loss": 1.9251, "step": 34286 }, { "epoch": 0.85, "learning_rate": 1.2318152969703012e-06, "loss": 2.0011, "step": 34287 }, { "epoch": 0.85, "learning_rate": 1.2314314700957708e-06, "loss": 2.0054, "step": 34288 }, { "epoch": 0.85, "learning_rate": 1.2310476991061837e-06, "loss": 1.6889, "step": 34289 }, { "epoch": 0.85, "learning_rate": 1.2306639840039903e-06, "loss": 1.7642, "step": 34290 }, { "epoch": 0.85, "learning_rate": 1.2302803247916318e-06, "loss": 1.9666, "step": 34291 }, { "epoch": 0.85, "learning_rate": 1.229896721471554e-06, "loss": 1.9911, "step": 34292 }, { "epoch": 0.85, "learning_rate": 1.2295131740462018e-06, "loss": 1.8782, "step": 34293 }, { "epoch": 0.85, "learning_rate": 1.2291296825180232e-06, "loss": 1.8902, "step": 34294 }, { "epoch": 0.85, "learning_rate": 1.2287462468894585e-06, "loss": 2.0502, "step": 34295 }, { "epoch": 0.85, "learning_rate": 1.2283628671629533e-06, "loss": 1.8472, "step": 34296 }, { "epoch": 0.85, "learning_rate": 1.227979543340948e-06, "loss": 1.8029, "step": 34297 }, { "epoch": 0.85, "learning_rate": 1.2275962754258874e-06, "loss": 1.8462, "step": 34298 }, { "epoch": 0.85, "learning_rate": 1.2272130634202172e-06, "loss": 1.947, "step": 34299 }, { "epoch": 0.85, "learning_rate": 1.2268299073263756e-06, "loss": 1.8746, "step": 34300 }, { "epoch": 0.85, "learning_rate": 1.2264468071468072e-06, "loss": 1.8879, "step": 34301 }, { "epoch": 0.85, "learning_rate": 1.2260637628839511e-06, "loss": 2.0364, "step": 34302 }, { "epoch": 0.85, "learning_rate": 1.22568077454025e-06, "loss": 1.8306, "step": 34303 }, { "epoch": 0.85, "learning_rate": 1.225297842118146e-06, "loss": 1.8937, "step": 34304 }, { "epoch": 0.85, "learning_rate": 1.2249149656200775e-06, "loss": 2.0908, "step": 34305 }, { "epoch": 0.85, "learning_rate": 1.224532145048485e-06, "loss": 1.9273, "step": 34306 }, { "epoch": 0.85, "learning_rate": 1.224149380405809e-06, "loss": 1.9134, "step": 34307 }, { "epoch": 0.85, "learning_rate": 1.2237666716944918e-06, "loss": 1.9887, "step": 34308 }, { "epoch": 0.85, "learning_rate": 1.2233840189169666e-06, "loss": 1.9816, "step": 34309 }, { "epoch": 0.85, "learning_rate": 1.2230014220756759e-06, "loss": 1.7645, "step": 34310 }, { "epoch": 0.85, "learning_rate": 1.2226188811730587e-06, "loss": 2.0488, "step": 34311 }, { "epoch": 0.85, "learning_rate": 1.2222363962115503e-06, "loss": 1.9076, "step": 34312 }, { "epoch": 0.85, "learning_rate": 1.2218539671935903e-06, "loss": 1.8733, "step": 34313 }, { "epoch": 0.85, "learning_rate": 1.2214715941216172e-06, "loss": 2.0044, "step": 34314 }, { "epoch": 0.85, "learning_rate": 1.2210892769980654e-06, "loss": 1.8925, "step": 34315 }, { "epoch": 0.85, "learning_rate": 1.220707015825373e-06, "loss": 1.7587, "step": 34316 }, { "epoch": 0.85, "learning_rate": 1.220324810605975e-06, "loss": 1.85, "step": 34317 }, { "epoch": 0.85, "learning_rate": 1.2199426613423083e-06, "loss": 1.7686, "step": 34318 }, { "epoch": 0.85, "learning_rate": 1.2195605680368093e-06, "loss": 1.98, "step": 34319 }, { "epoch": 0.85, "learning_rate": 1.2191785306919113e-06, "loss": 1.843, "step": 34320 }, { "epoch": 0.85, "learning_rate": 1.2187965493100496e-06, "loss": 2.0516, "step": 34321 }, { "epoch": 0.85, "learning_rate": 1.2184146238936624e-06, "loss": 1.9199, "step": 34322 }, { "epoch": 0.85, "learning_rate": 1.2180327544451775e-06, "loss": 2.0776, "step": 34323 }, { "epoch": 0.85, "learning_rate": 1.2176509409670312e-06, "loss": 1.9913, "step": 34324 }, { "epoch": 0.85, "learning_rate": 1.217269183461659e-06, "loss": 1.9894, "step": 34325 }, { "epoch": 0.85, "learning_rate": 1.2168874819314935e-06, "loss": 1.9922, "step": 34326 }, { "epoch": 0.85, "learning_rate": 1.216505836378965e-06, "loss": 1.9642, "step": 34327 }, { "epoch": 0.85, "learning_rate": 1.216124246806506e-06, "loss": 2.0391, "step": 34328 }, { "epoch": 0.85, "learning_rate": 1.2157427132165534e-06, "loss": 2.0577, "step": 34329 }, { "epoch": 0.85, "learning_rate": 1.2153612356115319e-06, "loss": 1.9851, "step": 34330 }, { "epoch": 0.85, "learning_rate": 1.2149798139938773e-06, "loss": 1.9316, "step": 34331 }, { "epoch": 0.85, "learning_rate": 1.2145984483660212e-06, "loss": 1.9589, "step": 34332 }, { "epoch": 0.85, "learning_rate": 1.2142171387303892e-06, "loss": 1.9492, "step": 34333 }, { "epoch": 0.85, "learning_rate": 1.2138358850894172e-06, "loss": 1.9946, "step": 34334 }, { "epoch": 0.85, "learning_rate": 1.21345468744553e-06, "loss": 1.7402, "step": 34335 }, { "epoch": 0.85, "learning_rate": 1.21307354580116e-06, "loss": 1.8087, "step": 34336 }, { "epoch": 0.85, "learning_rate": 1.2126924601587376e-06, "loss": 1.9139, "step": 34337 }, { "epoch": 0.85, "learning_rate": 1.2123114305206885e-06, "loss": 2.1037, "step": 34338 }, { "epoch": 0.85, "learning_rate": 1.211930456889444e-06, "loss": 1.9061, "step": 34339 }, { "epoch": 0.85, "learning_rate": 1.2115495392674282e-06, "loss": 1.9536, "step": 34340 }, { "epoch": 0.85, "learning_rate": 1.211168677657072e-06, "loss": 1.9291, "step": 34341 }, { "epoch": 0.85, "learning_rate": 1.2107878720608013e-06, "loss": 1.9414, "step": 34342 }, { "epoch": 0.85, "learning_rate": 1.2104071224810466e-06, "loss": 2.0904, "step": 34343 }, { "epoch": 0.85, "learning_rate": 1.210026428920229e-06, "loss": 1.8618, "step": 34344 }, { "epoch": 0.85, "learning_rate": 1.209645791380778e-06, "loss": 1.7996, "step": 34345 }, { "epoch": 0.85, "learning_rate": 1.2092652098651191e-06, "loss": 1.8397, "step": 34346 }, { "epoch": 0.85, "learning_rate": 1.2088846843756796e-06, "loss": 2.0086, "step": 34347 }, { "epoch": 0.85, "learning_rate": 1.2085042149148818e-06, "loss": 1.8947, "step": 34348 }, { "epoch": 0.85, "learning_rate": 1.2081238014851527e-06, "loss": 1.9821, "step": 34349 }, { "epoch": 0.85, "learning_rate": 1.2077434440889136e-06, "loss": 1.8739, "step": 34350 }, { "epoch": 0.85, "learning_rate": 1.2073631427285915e-06, "loss": 1.8862, "step": 34351 }, { "epoch": 0.85, "learning_rate": 1.2069828974066122e-06, "loss": 1.9472, "step": 34352 }, { "epoch": 0.85, "learning_rate": 1.2066027081253927e-06, "loss": 1.9256, "step": 34353 }, { "epoch": 0.85, "learning_rate": 1.2062225748873635e-06, "loss": 2.0358, "step": 34354 }, { "epoch": 0.85, "learning_rate": 1.2058424976949402e-06, "loss": 1.7836, "step": 34355 }, { "epoch": 0.85, "learning_rate": 1.2054624765505496e-06, "loss": 1.9372, "step": 34356 }, { "epoch": 0.85, "learning_rate": 1.2050825114566144e-06, "loss": 1.8835, "step": 34357 }, { "epoch": 0.85, "learning_rate": 1.2047026024155529e-06, "loss": 1.8481, "step": 34358 }, { "epoch": 0.85, "learning_rate": 1.2043227494297872e-06, "loss": 1.9936, "step": 34359 }, { "epoch": 0.85, "learning_rate": 1.20394295250174e-06, "loss": 1.9032, "step": 34360 }, { "epoch": 0.85, "learning_rate": 1.2035632116338314e-06, "loss": 1.8481, "step": 34361 }, { "epoch": 0.85, "learning_rate": 1.2031835268284807e-06, "loss": 2.0477, "step": 34362 }, { "epoch": 0.85, "learning_rate": 1.2028038980881074e-06, "loss": 1.8151, "step": 34363 }, { "epoch": 0.85, "learning_rate": 1.2024243254151346e-06, "loss": 1.9723, "step": 34364 }, { "epoch": 0.85, "learning_rate": 1.2020448088119751e-06, "loss": 2.0515, "step": 34365 }, { "epoch": 0.85, "learning_rate": 1.2016653482810525e-06, "loss": 2.3127, "step": 34366 }, { "epoch": 0.85, "learning_rate": 1.2012859438247858e-06, "loss": 1.883, "step": 34367 }, { "epoch": 0.85, "learning_rate": 1.200906595445589e-06, "loss": 1.9858, "step": 34368 }, { "epoch": 0.85, "learning_rate": 1.200527303145884e-06, "loss": 1.9832, "step": 34369 }, { "epoch": 0.85, "learning_rate": 1.200148066928084e-06, "loss": 1.9596, "step": 34370 }, { "epoch": 0.85, "learning_rate": 1.1997688867946079e-06, "loss": 2.0219, "step": 34371 }, { "epoch": 0.85, "learning_rate": 1.1993897627478757e-06, "loss": 2.0904, "step": 34372 }, { "epoch": 0.85, "learning_rate": 1.199010694790297e-06, "loss": 1.8258, "step": 34373 }, { "epoch": 0.85, "learning_rate": 1.198631682924295e-06, "loss": 1.8345, "step": 34374 }, { "epoch": 0.85, "learning_rate": 1.198252727152278e-06, "loss": 1.9389, "step": 34375 }, { "epoch": 0.85, "learning_rate": 1.1978738274766655e-06, "loss": 1.9714, "step": 34376 }, { "epoch": 0.85, "learning_rate": 1.1974949838998718e-06, "loss": 1.8889, "step": 34377 }, { "epoch": 0.85, "learning_rate": 1.1971161964243106e-06, "loss": 2.034, "step": 34378 }, { "epoch": 0.85, "learning_rate": 1.1967374650524e-06, "loss": 1.8528, "step": 34379 }, { "epoch": 0.85, "learning_rate": 1.196358789786547e-06, "loss": 2.0474, "step": 34380 }, { "epoch": 0.85, "learning_rate": 1.1959801706291686e-06, "loss": 1.7928, "step": 34381 }, { "epoch": 0.85, "learning_rate": 1.1956016075826804e-06, "loss": 1.8698, "step": 34382 }, { "epoch": 0.85, "learning_rate": 1.1952231006494896e-06, "loss": 1.9605, "step": 34383 }, { "epoch": 0.85, "learning_rate": 1.1948446498320121e-06, "loss": 1.8786, "step": 34384 }, { "epoch": 0.85, "learning_rate": 1.1944662551326613e-06, "loss": 1.9656, "step": 34385 }, { "epoch": 0.85, "learning_rate": 1.1940879165538443e-06, "loss": 1.9408, "step": 34386 }, { "epoch": 0.85, "learning_rate": 1.193709634097977e-06, "loss": 1.7657, "step": 34387 }, { "epoch": 0.85, "learning_rate": 1.1933314077674662e-06, "loss": 1.8036, "step": 34388 }, { "epoch": 0.85, "learning_rate": 1.1929532375647236e-06, "loss": 1.8322, "step": 34389 }, { "epoch": 0.85, "learning_rate": 1.1925751234921635e-06, "loss": 1.9092, "step": 34390 }, { "epoch": 0.85, "learning_rate": 1.1921970655521887e-06, "loss": 1.9312, "step": 34391 }, { "epoch": 0.85, "learning_rate": 1.191819063747216e-06, "loss": 2.0514, "step": 34392 }, { "epoch": 0.85, "learning_rate": 1.191441118079647e-06, "loss": 2.0548, "step": 34393 }, { "epoch": 0.85, "learning_rate": 1.1910632285518963e-06, "loss": 1.9357, "step": 34394 }, { "epoch": 0.85, "learning_rate": 1.1906853951663689e-06, "loss": 1.7636, "step": 34395 }, { "epoch": 0.85, "learning_rate": 1.1903076179254759e-06, "loss": 1.9296, "step": 34396 }, { "epoch": 0.85, "learning_rate": 1.1899298968316242e-06, "loss": 1.9297, "step": 34397 }, { "epoch": 0.85, "learning_rate": 1.1895522318872188e-06, "loss": 1.909, "step": 34398 }, { "epoch": 0.85, "learning_rate": 1.1891746230946677e-06, "loss": 2.0823, "step": 34399 }, { "epoch": 0.85, "learning_rate": 1.188797070456381e-06, "loss": 1.962, "step": 34400 }, { "epoch": 0.85, "learning_rate": 1.1884195739747595e-06, "loss": 1.9505, "step": 34401 }, { "epoch": 0.85, "learning_rate": 1.188042133652214e-06, "loss": 1.8975, "step": 34402 }, { "epoch": 0.85, "learning_rate": 1.1876647494911453e-06, "loss": 2.0361, "step": 34403 }, { "epoch": 0.85, "learning_rate": 1.1872874214939624e-06, "loss": 1.8209, "step": 34404 }, { "epoch": 0.85, "learning_rate": 1.1869101496630687e-06, "loss": 1.9218, "step": 34405 }, { "epoch": 0.85, "learning_rate": 1.186532934000868e-06, "loss": 2.1143, "step": 34406 }, { "epoch": 0.85, "learning_rate": 1.1861557745097663e-06, "loss": 1.9338, "step": 34407 }, { "epoch": 0.85, "learning_rate": 1.1857786711921648e-06, "loss": 1.9843, "step": 34408 }, { "epoch": 0.85, "learning_rate": 1.1854016240504684e-06, "loss": 1.9343, "step": 34409 }, { "epoch": 0.85, "learning_rate": 1.1850246330870817e-06, "loss": 2.052, "step": 34410 }, { "epoch": 0.85, "learning_rate": 1.1846476983044042e-06, "loss": 2.0137, "step": 34411 }, { "epoch": 0.85, "learning_rate": 1.1842708197048392e-06, "loss": 1.718, "step": 34412 }, { "epoch": 0.85, "learning_rate": 1.1838939972907893e-06, "loss": 2.134, "step": 34413 }, { "epoch": 0.85, "learning_rate": 1.1835172310646558e-06, "loss": 1.6293, "step": 34414 }, { "epoch": 0.85, "learning_rate": 1.1831405210288426e-06, "loss": 1.8997, "step": 34415 }, { "epoch": 0.85, "learning_rate": 1.1827638671857467e-06, "loss": 1.7959, "step": 34416 }, { "epoch": 0.85, "learning_rate": 1.1823872695377715e-06, "loss": 1.9374, "step": 34417 }, { "epoch": 0.85, "learning_rate": 1.1820107280873139e-06, "loss": 2.0287, "step": 34418 }, { "epoch": 0.85, "learning_rate": 1.1816342428367755e-06, "loss": 1.9586, "step": 34419 }, { "epoch": 0.85, "learning_rate": 1.1812578137885589e-06, "loss": 1.7978, "step": 34420 }, { "epoch": 0.85, "learning_rate": 1.1808814409450565e-06, "loss": 2.108, "step": 34421 }, { "epoch": 0.85, "learning_rate": 1.180505124308674e-06, "loss": 1.8365, "step": 34422 }, { "epoch": 0.85, "learning_rate": 1.1801288638818042e-06, "loss": 2.0936, "step": 34423 }, { "epoch": 0.85, "learning_rate": 1.1797526596668475e-06, "loss": 1.8112, "step": 34424 }, { "epoch": 0.85, "learning_rate": 1.1793765116662036e-06, "loss": 1.802, "step": 34425 }, { "epoch": 0.85, "learning_rate": 1.1790004198822646e-06, "loss": 1.8074, "step": 34426 }, { "epoch": 0.85, "learning_rate": 1.1786243843174338e-06, "loss": 1.8349, "step": 34427 }, { "epoch": 0.85, "learning_rate": 1.1782484049741028e-06, "loss": 2.0762, "step": 34428 }, { "epoch": 0.85, "learning_rate": 1.1778724818546694e-06, "loss": 2.0525, "step": 34429 }, { "epoch": 0.85, "learning_rate": 1.1774966149615297e-06, "loss": 1.9797, "step": 34430 }, { "epoch": 0.85, "learning_rate": 1.1771208042970784e-06, "loss": 1.7563, "step": 34431 }, { "epoch": 0.85, "learning_rate": 1.1767450498637146e-06, "loss": 1.7373, "step": 34432 }, { "epoch": 0.85, "learning_rate": 1.1763693516638275e-06, "loss": 1.7944, "step": 34433 }, { "epoch": 0.85, "learning_rate": 1.1759937096998142e-06, "loss": 1.8293, "step": 34434 }, { "epoch": 0.85, "learning_rate": 1.1756181239740705e-06, "loss": 1.8345, "step": 34435 }, { "epoch": 0.85, "learning_rate": 1.1752425944889867e-06, "loss": 1.8861, "step": 34436 }, { "epoch": 0.85, "learning_rate": 1.1748671212469576e-06, "loss": 1.912, "step": 34437 }, { "epoch": 0.85, "learning_rate": 1.1744917042503778e-06, "loss": 1.9062, "step": 34438 }, { "epoch": 0.85, "learning_rate": 1.1741163435016366e-06, "loss": 1.8822, "step": 34439 }, { "epoch": 0.85, "learning_rate": 1.1737410390031313e-06, "loss": 1.9293, "step": 34440 }, { "epoch": 0.85, "learning_rate": 1.173365790757247e-06, "loss": 1.8191, "step": 34441 }, { "epoch": 0.85, "learning_rate": 1.1729905987663804e-06, "loss": 1.9529, "step": 34442 }, { "epoch": 0.85, "learning_rate": 1.1726154630329234e-06, "loss": 2.0468, "step": 34443 }, { "epoch": 0.85, "learning_rate": 1.1722403835592622e-06, "loss": 1.853, "step": 34444 }, { "epoch": 0.85, "learning_rate": 1.1718653603477924e-06, "loss": 2.0474, "step": 34445 }, { "epoch": 0.85, "learning_rate": 1.1714903934008991e-06, "loss": 1.9432, "step": 34446 }, { "epoch": 0.85, "learning_rate": 1.1711154827209747e-06, "loss": 1.9749, "step": 34447 }, { "epoch": 0.85, "learning_rate": 1.1707406283104084e-06, "loss": 1.9852, "step": 34448 }, { "epoch": 0.85, "learning_rate": 1.1703658301715904e-06, "loss": 2.0451, "step": 34449 }, { "epoch": 0.85, "learning_rate": 1.16999108830691e-06, "loss": 1.9734, "step": 34450 }, { "epoch": 0.85, "learning_rate": 1.1696164027187517e-06, "loss": 1.7721, "step": 34451 }, { "epoch": 0.85, "learning_rate": 1.1692417734095063e-06, "loss": 1.9485, "step": 34452 }, { "epoch": 0.85, "learning_rate": 1.1688672003815627e-06, "loss": 1.852, "step": 34453 }, { "epoch": 0.85, "learning_rate": 1.1684926836373046e-06, "loss": 2.0618, "step": 34454 }, { "epoch": 0.85, "learning_rate": 1.168118223179122e-06, "loss": 1.923, "step": 34455 }, { "epoch": 0.85, "learning_rate": 1.1677438190093992e-06, "loss": 1.8606, "step": 34456 }, { "epoch": 0.85, "learning_rate": 1.1673694711305228e-06, "loss": 1.8692, "step": 34457 }, { "epoch": 0.85, "learning_rate": 1.1669951795448808e-06, "loss": 1.898, "step": 34458 }, { "epoch": 0.85, "learning_rate": 1.166620944254856e-06, "loss": 1.8472, "step": 34459 }, { "epoch": 0.85, "learning_rate": 1.1662467652628363e-06, "loss": 1.8818, "step": 34460 }, { "epoch": 0.85, "learning_rate": 1.165872642571203e-06, "loss": 2.0101, "step": 34461 }, { "epoch": 0.85, "learning_rate": 1.1654985761823423e-06, "loss": 1.9432, "step": 34462 }, { "epoch": 0.85, "learning_rate": 1.16512456609864e-06, "loss": 1.9076, "step": 34463 }, { "epoch": 0.85, "learning_rate": 1.164750612322476e-06, "loss": 1.8429, "step": 34464 }, { "epoch": 0.85, "learning_rate": 1.1643767148562358e-06, "loss": 1.7853, "step": 34465 }, { "epoch": 0.85, "learning_rate": 1.1640028737023012e-06, "loss": 1.7942, "step": 34466 }, { "epoch": 0.85, "learning_rate": 1.1636290888630576e-06, "loss": 1.9688, "step": 34467 }, { "epoch": 0.85, "learning_rate": 1.163255360340887e-06, "loss": 1.8451, "step": 34468 }, { "epoch": 0.85, "learning_rate": 1.1628816881381678e-06, "loss": 2.1001, "step": 34469 }, { "epoch": 0.85, "learning_rate": 1.1625080722572845e-06, "loss": 1.8968, "step": 34470 }, { "epoch": 0.85, "learning_rate": 1.1621345127006166e-06, "loss": 1.9261, "step": 34471 }, { "epoch": 0.85, "learning_rate": 1.1617610094705445e-06, "loss": 2.0702, "step": 34472 }, { "epoch": 0.85, "learning_rate": 1.1613875625694526e-06, "loss": 1.9591, "step": 34473 }, { "epoch": 0.85, "learning_rate": 1.161014171999716e-06, "loss": 1.9937, "step": 34474 }, { "epoch": 0.85, "learning_rate": 1.160640837763719e-06, "loss": 1.8309, "step": 34475 }, { "epoch": 0.85, "learning_rate": 1.1602675598638358e-06, "loss": 1.8286, "step": 34476 }, { "epoch": 0.85, "learning_rate": 1.1598943383024485e-06, "loss": 1.9476, "step": 34477 }, { "epoch": 0.85, "learning_rate": 1.1595211730819379e-06, "loss": 1.9457, "step": 34478 }, { "epoch": 0.85, "learning_rate": 1.1591480642046782e-06, "loss": 1.9065, "step": 34479 }, { "epoch": 0.85, "learning_rate": 1.1587750116730502e-06, "loss": 1.8336, "step": 34480 }, { "epoch": 0.85, "learning_rate": 1.1584020154894281e-06, "loss": 1.9321, "step": 34481 }, { "epoch": 0.85, "learning_rate": 1.158029075656193e-06, "loss": 1.8604, "step": 34482 }, { "epoch": 0.85, "learning_rate": 1.1576561921757179e-06, "loss": 1.8585, "step": 34483 }, { "epoch": 0.85, "learning_rate": 1.1572833650503835e-06, "loss": 1.9568, "step": 34484 }, { "epoch": 0.85, "learning_rate": 1.1569105942825642e-06, "loss": 2.04, "step": 34485 }, { "epoch": 0.85, "learning_rate": 1.1565378798746351e-06, "loss": 1.8627, "step": 34486 }, { "epoch": 0.85, "learning_rate": 1.156165221828971e-06, "loss": 1.9358, "step": 34487 }, { "epoch": 0.85, "learning_rate": 1.1557926201479507e-06, "loss": 1.9051, "step": 34488 }, { "epoch": 0.85, "learning_rate": 1.155420074833944e-06, "loss": 1.7709, "step": 34489 }, { "epoch": 0.85, "learning_rate": 1.1550475858893296e-06, "loss": 1.8901, "step": 34490 }, { "epoch": 0.85, "learning_rate": 1.1546751533164779e-06, "loss": 1.8278, "step": 34491 }, { "epoch": 0.85, "learning_rate": 1.154302777117763e-06, "loss": 1.9457, "step": 34492 }, { "epoch": 0.85, "learning_rate": 1.1539304572955623e-06, "loss": 1.7907, "step": 34493 }, { "epoch": 0.85, "learning_rate": 1.1535581938522434e-06, "loss": 1.9633, "step": 34494 }, { "epoch": 0.85, "learning_rate": 1.1531859867901829e-06, "loss": 1.8978, "step": 34495 }, { "epoch": 0.85, "learning_rate": 1.152813836111749e-06, "loss": 1.929, "step": 34496 }, { "epoch": 0.85, "learning_rate": 1.1524417418193156e-06, "loss": 1.8684, "step": 34497 }, { "epoch": 0.85, "learning_rate": 1.1520697039152562e-06, "loss": 1.8328, "step": 34498 }, { "epoch": 0.85, "learning_rate": 1.1516977224019389e-06, "loss": 2.063, "step": 34499 }, { "epoch": 0.85, "learning_rate": 1.1513257972817348e-06, "loss": 1.8327, "step": 34500 }, { "epoch": 0.85, "learning_rate": 1.1509539285570159e-06, "loss": 1.9187, "step": 34501 }, { "epoch": 0.85, "learning_rate": 1.150582116230151e-06, "loss": 2.03, "step": 34502 }, { "epoch": 0.85, "learning_rate": 1.150210360303512e-06, "loss": 1.8683, "step": 34503 }, { "epoch": 0.85, "learning_rate": 1.1498386607794643e-06, "loss": 2.0107, "step": 34504 }, { "epoch": 0.85, "learning_rate": 1.1494670176603796e-06, "loss": 1.8737, "step": 34505 }, { "epoch": 0.85, "learning_rate": 1.149095430948628e-06, "loss": 1.9495, "step": 34506 }, { "epoch": 0.85, "learning_rate": 1.1487239006465733e-06, "loss": 1.8634, "step": 34507 }, { "epoch": 0.85, "learning_rate": 1.148352426756587e-06, "loss": 1.9228, "step": 34508 }, { "epoch": 0.85, "learning_rate": 1.147981009281034e-06, "loss": 1.9969, "step": 34509 }, { "epoch": 0.85, "learning_rate": 1.1476096482222831e-06, "loss": 1.8574, "step": 34510 }, { "epoch": 0.85, "learning_rate": 1.1472383435827028e-06, "loss": 1.8477, "step": 34511 }, { "epoch": 0.85, "learning_rate": 1.1468670953646555e-06, "loss": 1.892, "step": 34512 }, { "epoch": 0.85, "learning_rate": 1.1464959035705114e-06, "loss": 2.115, "step": 34513 }, { "epoch": 0.85, "learning_rate": 1.146124768202631e-06, "loss": 2.0372, "step": 34514 }, { "epoch": 0.85, "learning_rate": 1.1457536892633846e-06, "loss": 1.9659, "step": 34515 }, { "epoch": 0.85, "learning_rate": 1.1453826667551359e-06, "loss": 2.0905, "step": 34516 }, { "epoch": 0.85, "learning_rate": 1.1450117006802485e-06, "loss": 1.7545, "step": 34517 }, { "epoch": 0.85, "learning_rate": 1.1446407910410863e-06, "loss": 1.8467, "step": 34518 }, { "epoch": 0.85, "learning_rate": 1.1442699378400145e-06, "loss": 1.8124, "step": 34519 }, { "epoch": 0.85, "learning_rate": 1.1438991410793966e-06, "loss": 1.9631, "step": 34520 }, { "epoch": 0.85, "learning_rate": 1.1435284007615976e-06, "loss": 1.9794, "step": 34521 }, { "epoch": 0.85, "learning_rate": 1.1431577168889762e-06, "loss": 1.9733, "step": 34522 }, { "epoch": 0.85, "learning_rate": 1.1427870894638994e-06, "loss": 1.9118, "step": 34523 }, { "epoch": 0.85, "learning_rate": 1.1424165184887247e-06, "loss": 1.9488, "step": 34524 }, { "epoch": 0.85, "learning_rate": 1.1420460039658155e-06, "loss": 1.9931, "step": 34525 }, { "epoch": 0.85, "learning_rate": 1.141675545897537e-06, "loss": 1.7836, "step": 34526 }, { "epoch": 0.85, "learning_rate": 1.1413051442862444e-06, "loss": 2.0355, "step": 34527 }, { "epoch": 0.85, "learning_rate": 1.1409347991343033e-06, "loss": 1.8307, "step": 34528 }, { "epoch": 0.85, "learning_rate": 1.1405645104440699e-06, "loss": 1.9312, "step": 34529 }, { "epoch": 0.85, "learning_rate": 1.1401942782179055e-06, "loss": 1.7487, "step": 34530 }, { "epoch": 0.85, "learning_rate": 1.1398241024581725e-06, "loss": 2.0586, "step": 34531 }, { "epoch": 0.85, "learning_rate": 1.1394539831672258e-06, "loss": 1.823, "step": 34532 }, { "epoch": 0.85, "learning_rate": 1.1390839203474291e-06, "loss": 1.9736, "step": 34533 }, { "epoch": 0.85, "learning_rate": 1.1387139140011362e-06, "loss": 1.8773, "step": 34534 }, { "epoch": 0.85, "learning_rate": 1.138343964130707e-06, "loss": 1.9931, "step": 34535 }, { "epoch": 0.85, "learning_rate": 1.137974070738499e-06, "loss": 1.9253, "step": 34536 }, { "epoch": 0.85, "learning_rate": 1.1376042338268712e-06, "loss": 1.9287, "step": 34537 }, { "epoch": 0.85, "learning_rate": 1.1372344533981815e-06, "loss": 1.8702, "step": 34538 }, { "epoch": 0.85, "learning_rate": 1.136864729454783e-06, "loss": 1.9688, "step": 34539 }, { "epoch": 0.85, "learning_rate": 1.1364950619990345e-06, "loss": 1.7589, "step": 34540 }, { "epoch": 0.85, "learning_rate": 1.1361254510332932e-06, "loss": 1.9317, "step": 34541 }, { "epoch": 0.85, "learning_rate": 1.1357558965599103e-06, "loss": 1.8884, "step": 34542 }, { "epoch": 0.85, "learning_rate": 1.1353863985812463e-06, "loss": 1.9452, "step": 34543 }, { "epoch": 0.85, "learning_rate": 1.1350169570996517e-06, "loss": 1.8601, "step": 34544 }, { "epoch": 0.85, "learning_rate": 1.1346475721174831e-06, "loss": 1.9125, "step": 34545 }, { "epoch": 0.85, "learning_rate": 1.134278243637097e-06, "loss": 1.9525, "step": 34546 }, { "epoch": 0.85, "learning_rate": 1.133908971660842e-06, "loss": 1.9965, "step": 34547 }, { "epoch": 0.85, "learning_rate": 1.1335397561910778e-06, "loss": 1.8821, "step": 34548 }, { "epoch": 0.85, "learning_rate": 1.1331705972301509e-06, "loss": 1.9853, "step": 34549 }, { "epoch": 0.85, "learning_rate": 1.1328014947804177e-06, "loss": 1.9618, "step": 34550 }, { "epoch": 0.85, "learning_rate": 1.1324324488442328e-06, "loss": 2.0497, "step": 34551 }, { "epoch": 0.85, "learning_rate": 1.1320634594239443e-06, "loss": 1.8447, "step": 34552 }, { "epoch": 0.85, "learning_rate": 1.1316945265219037e-06, "loss": 1.816, "step": 34553 }, { "epoch": 0.85, "learning_rate": 1.1313256501404658e-06, "loss": 1.9402, "step": 34554 }, { "epoch": 0.85, "learning_rate": 1.1309568302819785e-06, "loss": 2.0773, "step": 34555 }, { "epoch": 0.85, "learning_rate": 1.1305880669487956e-06, "loss": 1.689, "step": 34556 }, { "epoch": 0.85, "learning_rate": 1.1302193601432643e-06, "loss": 1.873, "step": 34557 }, { "epoch": 0.85, "learning_rate": 1.1298507098677346e-06, "loss": 1.8777, "step": 34558 }, { "epoch": 0.85, "learning_rate": 1.1294821161245595e-06, "loss": 1.9277, "step": 34559 }, { "epoch": 0.85, "learning_rate": 1.1291135789160846e-06, "loss": 1.8566, "step": 34560 }, { "epoch": 0.85, "learning_rate": 1.1287450982446602e-06, "loss": 1.7545, "step": 34561 }, { "epoch": 0.85, "learning_rate": 1.1283766741126333e-06, "loss": 1.9657, "step": 34562 }, { "epoch": 0.85, "learning_rate": 1.1280083065223524e-06, "loss": 2.0111, "step": 34563 }, { "epoch": 0.85, "learning_rate": 1.1276399954761686e-06, "loss": 1.9534, "step": 34564 }, { "epoch": 0.85, "learning_rate": 1.1272717409764244e-06, "loss": 1.7911, "step": 34565 }, { "epoch": 0.85, "learning_rate": 1.1269035430254705e-06, "loss": 1.8919, "step": 34566 }, { "epoch": 0.85, "learning_rate": 1.1265354016256502e-06, "loss": 1.8234, "step": 34567 }, { "epoch": 0.85, "learning_rate": 1.126167316779312e-06, "loss": 1.939, "step": 34568 }, { "epoch": 0.85, "learning_rate": 1.1257992884888024e-06, "loss": 1.9622, "step": 34569 }, { "epoch": 0.85, "learning_rate": 1.1254313167564645e-06, "loss": 1.999, "step": 34570 }, { "epoch": 0.85, "learning_rate": 1.1250634015846452e-06, "loss": 1.7059, "step": 34571 }, { "epoch": 0.85, "learning_rate": 1.124695542975689e-06, "loss": 2.0749, "step": 34572 }, { "epoch": 0.85, "learning_rate": 1.1243277409319398e-06, "loss": 2.0365, "step": 34573 }, { "epoch": 0.85, "learning_rate": 1.1239599954557456e-06, "loss": 1.9066, "step": 34574 }, { "epoch": 0.85, "learning_rate": 1.1235923065494436e-06, "loss": 1.9024, "step": 34575 }, { "epoch": 0.85, "learning_rate": 1.123224674215384e-06, "loss": 1.9444, "step": 34576 }, { "epoch": 0.85, "learning_rate": 1.122857098455904e-06, "loss": 1.9341, "step": 34577 }, { "epoch": 0.85, "learning_rate": 1.1224895792733482e-06, "loss": 2.0078, "step": 34578 }, { "epoch": 0.85, "learning_rate": 1.1221221166700613e-06, "loss": 1.9024, "step": 34579 }, { "epoch": 0.85, "learning_rate": 1.1217547106483817e-06, "loss": 2.0789, "step": 34580 }, { "epoch": 0.85, "learning_rate": 1.1213873612106552e-06, "loss": 2.0749, "step": 34581 }, { "epoch": 0.85, "learning_rate": 1.1210200683592177e-06, "loss": 1.8828, "step": 34582 }, { "epoch": 0.85, "learning_rate": 1.1206528320964128e-06, "loss": 1.7896, "step": 34583 }, { "epoch": 0.85, "learning_rate": 1.120285652424583e-06, "loss": 2.103, "step": 34584 }, { "epoch": 0.85, "learning_rate": 1.1199185293460657e-06, "loss": 1.9163, "step": 34585 }, { "epoch": 0.85, "learning_rate": 1.1195514628632021e-06, "loss": 1.7068, "step": 34586 }, { "epoch": 0.85, "learning_rate": 1.1191844529783302e-06, "loss": 1.9188, "step": 34587 }, { "epoch": 0.85, "learning_rate": 1.1188174996937895e-06, "loss": 2.0505, "step": 34588 }, { "epoch": 0.85, "learning_rate": 1.118450603011919e-06, "loss": 2.0993, "step": 34589 }, { "epoch": 0.85, "learning_rate": 1.1180837629350571e-06, "loss": 1.9976, "step": 34590 }, { "epoch": 0.85, "learning_rate": 1.1177169794655452e-06, "loss": 1.8187, "step": 34591 }, { "epoch": 0.85, "learning_rate": 1.1173502526057146e-06, "loss": 2.0067, "step": 34592 }, { "epoch": 0.85, "learning_rate": 1.1169835823579055e-06, "loss": 1.8888, "step": 34593 }, { "epoch": 0.85, "learning_rate": 1.1166169687244577e-06, "loss": 1.9935, "step": 34594 }, { "epoch": 0.85, "learning_rate": 1.1162504117077022e-06, "loss": 1.8632, "step": 34595 }, { "epoch": 0.85, "learning_rate": 1.1158839113099795e-06, "loss": 1.9474, "step": 34596 }, { "epoch": 0.85, "learning_rate": 1.1155174675336232e-06, "loss": 1.8805, "step": 34597 }, { "epoch": 0.85, "learning_rate": 1.1151510803809685e-06, "loss": 2.0444, "step": 34598 }, { "epoch": 0.85, "learning_rate": 1.114784749854354e-06, "loss": 1.8507, "step": 34599 }, { "epoch": 0.85, "learning_rate": 1.1144184759561095e-06, "loss": 1.7832, "step": 34600 }, { "epoch": 0.85, "learning_rate": 1.1140522586885737e-06, "loss": 1.8888, "step": 34601 }, { "epoch": 0.85, "learning_rate": 1.1136860980540765e-06, "loss": 2.0339, "step": 34602 }, { "epoch": 0.85, "learning_rate": 1.1133199940549543e-06, "loss": 1.8511, "step": 34603 }, { "epoch": 0.85, "learning_rate": 1.1129539466935423e-06, "loss": 1.9056, "step": 34604 }, { "epoch": 0.85, "learning_rate": 1.1125879559721687e-06, "loss": 1.7992, "step": 34605 }, { "epoch": 0.85, "learning_rate": 1.112222021893168e-06, "loss": 1.9097, "step": 34606 }, { "epoch": 0.85, "learning_rate": 1.1118561444588727e-06, "loss": 1.8577, "step": 34607 }, { "epoch": 0.85, "learning_rate": 1.1114903236716146e-06, "loss": 1.9141, "step": 34608 }, { "epoch": 0.85, "learning_rate": 1.1111245595337283e-06, "loss": 1.9055, "step": 34609 }, { "epoch": 0.85, "learning_rate": 1.1107588520475387e-06, "loss": 1.9817, "step": 34610 }, { "epoch": 0.85, "learning_rate": 1.110393201215383e-06, "loss": 1.9253, "step": 34611 }, { "epoch": 0.85, "learning_rate": 1.1100276070395866e-06, "loss": 1.9919, "step": 34612 }, { "epoch": 0.85, "learning_rate": 1.1096620695224803e-06, "loss": 2.0802, "step": 34613 }, { "epoch": 0.85, "learning_rate": 1.1092965886663975e-06, "loss": 1.8764, "step": 34614 }, { "epoch": 0.85, "learning_rate": 1.1089311644736634e-06, "loss": 1.8204, "step": 34615 }, { "epoch": 0.85, "learning_rate": 1.1085657969466079e-06, "loss": 2.1223, "step": 34616 }, { "epoch": 0.85, "learning_rate": 1.1082004860875617e-06, "loss": 2.0675, "step": 34617 }, { "epoch": 0.85, "learning_rate": 1.1078352318988506e-06, "loss": 1.9099, "step": 34618 }, { "epoch": 0.85, "learning_rate": 1.1074700343828048e-06, "loss": 1.8866, "step": 34619 }, { "epoch": 0.85, "learning_rate": 1.1071048935417483e-06, "loss": 1.8955, "step": 34620 }, { "epoch": 0.85, "learning_rate": 1.1067398093780113e-06, "loss": 1.8779, "step": 34621 }, { "epoch": 0.85, "learning_rate": 1.1063747818939209e-06, "loss": 1.9554, "step": 34622 }, { "epoch": 0.85, "learning_rate": 1.1060098110918006e-06, "loss": 2.1123, "step": 34623 }, { "epoch": 0.85, "learning_rate": 1.1056448969739787e-06, "loss": 1.8686, "step": 34624 }, { "epoch": 0.85, "learning_rate": 1.10528003954278e-06, "loss": 1.9484, "step": 34625 }, { "epoch": 0.85, "learning_rate": 1.1049152388005302e-06, "loss": 1.706, "step": 34626 }, { "epoch": 0.85, "learning_rate": 1.1045504947495566e-06, "loss": 1.927, "step": 34627 }, { "epoch": 0.85, "learning_rate": 1.1041858073921797e-06, "loss": 1.8802, "step": 34628 }, { "epoch": 0.85, "learning_rate": 1.1038211767307282e-06, "loss": 1.9239, "step": 34629 }, { "epoch": 0.85, "learning_rate": 1.103456602767521e-06, "loss": 1.8437, "step": 34630 }, { "epoch": 0.85, "learning_rate": 1.1030920855048844e-06, "loss": 1.8072, "step": 34631 }, { "epoch": 0.85, "learning_rate": 1.1027276249451423e-06, "loss": 1.8804, "step": 34632 }, { "epoch": 0.85, "learning_rate": 1.1023632210906155e-06, "loss": 1.9343, "step": 34633 }, { "epoch": 0.85, "learning_rate": 1.1019988739436295e-06, "loss": 2.0077, "step": 34634 }, { "epoch": 0.85, "learning_rate": 1.1016345835065024e-06, "loss": 1.6926, "step": 34635 }, { "epoch": 0.85, "learning_rate": 1.101270349781558e-06, "loss": 1.9425, "step": 34636 }, { "epoch": 0.85, "learning_rate": 1.10090617277112e-06, "loss": 1.9908, "step": 34637 }, { "epoch": 0.85, "learning_rate": 1.1005420524775046e-06, "loss": 1.9263, "step": 34638 }, { "epoch": 0.85, "learning_rate": 1.1001779889030373e-06, "loss": 1.9111, "step": 34639 }, { "epoch": 0.85, "learning_rate": 1.0998139820500342e-06, "loss": 1.9862, "step": 34640 }, { "epoch": 0.85, "learning_rate": 1.0994500319208168e-06, "loss": 1.9873, "step": 34641 }, { "epoch": 0.85, "learning_rate": 1.0990861385177043e-06, "loss": 1.8072, "step": 34642 }, { "epoch": 0.85, "learning_rate": 1.098722301843017e-06, "loss": 1.8736, "step": 34643 }, { "epoch": 0.85, "learning_rate": 1.0983585218990756e-06, "loss": 1.8379, "step": 34644 }, { "epoch": 0.85, "learning_rate": 1.0979947986881945e-06, "loss": 1.909, "step": 34645 }, { "epoch": 0.85, "learning_rate": 1.097631132212693e-06, "loss": 2.118, "step": 34646 }, { "epoch": 0.85, "learning_rate": 1.097267522474892e-06, "loss": 1.77, "step": 34647 }, { "epoch": 0.85, "learning_rate": 1.0969039694771044e-06, "loss": 1.8341, "step": 34648 }, { "epoch": 0.85, "learning_rate": 1.0965404732216511e-06, "loss": 2.0238, "step": 34649 }, { "epoch": 0.85, "learning_rate": 1.0961770337108447e-06, "loss": 1.9275, "step": 34650 }, { "epoch": 0.85, "learning_rate": 1.0958136509470041e-06, "loss": 1.9598, "step": 34651 }, { "epoch": 0.85, "learning_rate": 1.0954503249324466e-06, "loss": 1.8411, "step": 34652 }, { "epoch": 0.85, "learning_rate": 1.0950870556694837e-06, "loss": 1.8931, "step": 34653 }, { "epoch": 0.85, "learning_rate": 1.0947238431604357e-06, "loss": 2.139, "step": 34654 }, { "epoch": 0.85, "learning_rate": 1.0943606874076129e-06, "loss": 1.9636, "step": 34655 }, { "epoch": 0.85, "learning_rate": 1.0939975884133313e-06, "loss": 1.8914, "step": 34656 }, { "epoch": 0.85, "learning_rate": 1.093634546179907e-06, "loss": 2.1142, "step": 34657 }, { "epoch": 0.85, "learning_rate": 1.0932715607096511e-06, "loss": 1.7069, "step": 34658 }, { "epoch": 0.85, "learning_rate": 1.0929086320048776e-06, "loss": 1.872, "step": 34659 }, { "epoch": 0.85, "learning_rate": 1.0925457600679002e-06, "loss": 1.897, "step": 34660 }, { "epoch": 0.85, "learning_rate": 1.0921829449010324e-06, "loss": 1.7861, "step": 34661 }, { "epoch": 0.85, "learning_rate": 1.091820186506587e-06, "loss": 2.0692, "step": 34662 }, { "epoch": 0.85, "learning_rate": 1.091457484886872e-06, "loss": 1.9794, "step": 34663 }, { "epoch": 0.85, "learning_rate": 1.0910948400442045e-06, "loss": 1.9583, "step": 34664 }, { "epoch": 0.85, "learning_rate": 1.0907322519808904e-06, "loss": 1.8295, "step": 34665 }, { "epoch": 0.85, "learning_rate": 1.0903697206992437e-06, "loss": 1.9069, "step": 34666 }, { "epoch": 0.85, "learning_rate": 1.0900072462015754e-06, "loss": 1.7993, "step": 34667 }, { "epoch": 0.85, "learning_rate": 1.089644828490194e-06, "loss": 1.9863, "step": 34668 }, { "epoch": 0.85, "learning_rate": 1.0892824675674108e-06, "loss": 2.1485, "step": 34669 }, { "epoch": 0.85, "learning_rate": 1.088920163435533e-06, "loss": 1.8093, "step": 34670 }, { "epoch": 0.85, "learning_rate": 1.0885579160968708e-06, "loss": 1.978, "step": 34671 }, { "epoch": 0.85, "learning_rate": 1.0881957255537345e-06, "loss": 1.7651, "step": 34672 }, { "epoch": 0.85, "learning_rate": 1.0878335918084294e-06, "loss": 1.8641, "step": 34673 }, { "epoch": 0.85, "learning_rate": 1.0874715148632652e-06, "loss": 1.7363, "step": 34674 }, { "epoch": 0.85, "learning_rate": 1.0871094947205517e-06, "loss": 1.9012, "step": 34675 }, { "epoch": 0.85, "learning_rate": 1.0867475313825926e-06, "loss": 1.8928, "step": 34676 }, { "epoch": 0.85, "learning_rate": 1.0863856248516957e-06, "loss": 1.9972, "step": 34677 }, { "epoch": 0.85, "learning_rate": 1.0860237751301682e-06, "loss": 1.902, "step": 34678 }, { "epoch": 0.85, "learning_rate": 1.0856619822203162e-06, "loss": 1.9132, "step": 34679 }, { "epoch": 0.85, "learning_rate": 1.0853002461244478e-06, "loss": 1.7599, "step": 34680 }, { "epoch": 0.85, "learning_rate": 1.0849385668448632e-06, "loss": 1.8916, "step": 34681 }, { "epoch": 0.85, "learning_rate": 1.0845769443838738e-06, "loss": 1.7776, "step": 34682 }, { "epoch": 0.85, "learning_rate": 1.0842153787437782e-06, "loss": 1.8685, "step": 34683 }, { "epoch": 0.85, "learning_rate": 1.083853869926883e-06, "loss": 2.0873, "step": 34684 }, { "epoch": 0.85, "learning_rate": 1.0834924179354956e-06, "loss": 1.9586, "step": 34685 }, { "epoch": 0.85, "learning_rate": 1.0831310227719138e-06, "loss": 2.0257, "step": 34686 }, { "epoch": 0.85, "learning_rate": 1.0827696844384473e-06, "loss": 1.822, "step": 34687 }, { "epoch": 0.86, "learning_rate": 1.0824084029373927e-06, "loss": 2.0054, "step": 34688 }, { "epoch": 0.86, "learning_rate": 1.0820471782710563e-06, "loss": 1.8932, "step": 34689 }, { "epoch": 0.86, "learning_rate": 1.0816860104417404e-06, "loss": 1.9048, "step": 34690 }, { "epoch": 0.86, "learning_rate": 1.0813248994517444e-06, "loss": 1.8826, "step": 34691 }, { "epoch": 0.86, "learning_rate": 1.0809638453033743e-06, "loss": 2.0182, "step": 34692 }, { "epoch": 0.86, "learning_rate": 1.080602847998925e-06, "loss": 1.7696, "step": 34693 }, { "epoch": 0.86, "learning_rate": 1.080241907540701e-06, "loss": 1.9191, "step": 34694 }, { "epoch": 0.86, "learning_rate": 1.079881023931002e-06, "loss": 1.9705, "step": 34695 }, { "epoch": 0.86, "learning_rate": 1.079520197172128e-06, "loss": 1.9379, "step": 34696 }, { "epoch": 0.86, "learning_rate": 1.079159427266382e-06, "loss": 1.9026, "step": 34697 }, { "epoch": 0.86, "learning_rate": 1.0787987142160573e-06, "loss": 1.8083, "step": 34698 }, { "epoch": 0.86, "learning_rate": 1.0784380580234555e-06, "loss": 1.9581, "step": 34699 }, { "epoch": 0.86, "learning_rate": 1.0780774586908783e-06, "loss": 1.9915, "step": 34700 }, { "epoch": 0.86, "learning_rate": 1.077716916220618e-06, "loss": 1.7618, "step": 34701 }, { "epoch": 0.86, "learning_rate": 1.0773564306149787e-06, "loss": 1.8791, "step": 34702 }, { "epoch": 0.86, "learning_rate": 1.0769960018762516e-06, "loss": 1.9534, "step": 34703 }, { "epoch": 0.86, "learning_rate": 1.0766356300067372e-06, "loss": 2.0945, "step": 34704 }, { "epoch": 0.86, "learning_rate": 1.0762753150087334e-06, "loss": 1.8048, "step": 34705 }, { "epoch": 0.86, "learning_rate": 1.0759150568845344e-06, "loss": 1.8867, "step": 34706 }, { "epoch": 0.86, "learning_rate": 1.075554855636438e-06, "loss": 1.5664, "step": 34707 }, { "epoch": 0.86, "learning_rate": 1.075194711266737e-06, "loss": 1.9093, "step": 34708 }, { "epoch": 0.86, "learning_rate": 1.0748346237777295e-06, "loss": 1.8672, "step": 34709 }, { "epoch": 0.86, "learning_rate": 1.0744745931717115e-06, "loss": 1.8353, "step": 34710 }, { "epoch": 0.86, "learning_rate": 1.0741146194509733e-06, "loss": 1.9851, "step": 34711 }, { "epoch": 0.86, "learning_rate": 1.0737547026178108e-06, "loss": 1.9897, "step": 34712 }, { "epoch": 0.86, "learning_rate": 1.073394842674519e-06, "loss": 1.9878, "step": 34713 }, { "epoch": 0.86, "learning_rate": 1.0730350396233913e-06, "loss": 1.8119, "step": 34714 }, { "epoch": 0.86, "learning_rate": 1.0726752934667217e-06, "loss": 1.9724, "step": 34715 }, { "epoch": 0.86, "learning_rate": 1.0723156042068005e-06, "loss": 1.8653, "step": 34716 }, { "epoch": 0.86, "learning_rate": 1.0719559718459239e-06, "loss": 2.1234, "step": 34717 }, { "epoch": 0.86, "learning_rate": 1.0715963963863784e-06, "loss": 1.826, "step": 34718 }, { "epoch": 0.86, "learning_rate": 1.0712368778304593e-06, "loss": 1.9824, "step": 34719 }, { "epoch": 0.86, "learning_rate": 1.0708774161804591e-06, "loss": 1.8851, "step": 34720 }, { "epoch": 0.86, "learning_rate": 1.0705180114386648e-06, "loss": 1.9311, "step": 34721 }, { "epoch": 0.86, "learning_rate": 1.0701586636073724e-06, "loss": 2.0293, "step": 34722 }, { "epoch": 0.86, "learning_rate": 1.0697993726888656e-06, "loss": 1.82, "step": 34723 }, { "epoch": 0.86, "learning_rate": 1.0694401386854392e-06, "loss": 2.1224, "step": 34724 }, { "epoch": 0.86, "learning_rate": 1.0690809615993814e-06, "loss": 2.0088, "step": 34725 }, { "epoch": 0.86, "learning_rate": 1.0687218414329803e-06, "loss": 2.0611, "step": 34726 }, { "epoch": 0.86, "learning_rate": 1.0683627781885253e-06, "loss": 1.9816, "step": 34727 }, { "epoch": 0.86, "learning_rate": 1.0680037718683068e-06, "loss": 1.7759, "step": 34728 }, { "epoch": 0.86, "learning_rate": 1.0676448224746094e-06, "loss": 1.9239, "step": 34729 }, { "epoch": 0.86, "learning_rate": 1.0672859300097226e-06, "loss": 1.8756, "step": 34730 }, { "epoch": 0.86, "learning_rate": 1.0669270944759336e-06, "loss": 2.2072, "step": 34731 }, { "epoch": 0.86, "learning_rate": 1.0665683158755314e-06, "loss": 2.0127, "step": 34732 }, { "epoch": 0.86, "learning_rate": 1.0662095942107987e-06, "loss": 1.9328, "step": 34733 }, { "epoch": 0.86, "learning_rate": 1.0658509294840237e-06, "loss": 1.9265, "step": 34734 }, { "epoch": 0.86, "learning_rate": 1.0654923216974933e-06, "loss": 1.9806, "step": 34735 }, { "epoch": 0.86, "learning_rate": 1.0651337708534914e-06, "loss": 1.8632, "step": 34736 }, { "epoch": 0.86, "learning_rate": 1.064775276954303e-06, "loss": 1.8932, "step": 34737 }, { "epoch": 0.86, "learning_rate": 1.0644168400022159e-06, "loss": 1.9847, "step": 34738 }, { "epoch": 0.86, "learning_rate": 1.0640584599995108e-06, "loss": 1.9134, "step": 34739 }, { "epoch": 0.86, "learning_rate": 1.0637001369484746e-06, "loss": 1.8662, "step": 34740 }, { "epoch": 0.86, "learning_rate": 1.0633418708513876e-06, "loss": 1.883, "step": 34741 }, { "epoch": 0.86, "learning_rate": 1.0629836617105348e-06, "loss": 1.9205, "step": 34742 }, { "epoch": 0.86, "learning_rate": 1.0626255095282022e-06, "loss": 1.8466, "step": 34743 }, { "epoch": 0.86, "learning_rate": 1.062267414306668e-06, "loss": 1.7464, "step": 34744 }, { "epoch": 0.86, "learning_rate": 1.0619093760482168e-06, "loss": 1.6818, "step": 34745 }, { "epoch": 0.86, "learning_rate": 1.0615513947551316e-06, "loss": 1.7778, "step": 34746 }, { "epoch": 0.86, "learning_rate": 1.06119347042969e-06, "loss": 1.8158, "step": 34747 }, { "epoch": 0.86, "learning_rate": 1.060835603074175e-06, "loss": 2.0085, "step": 34748 }, { "epoch": 0.86, "learning_rate": 1.0604777926908695e-06, "loss": 2.0534, "step": 34749 }, { "epoch": 0.86, "learning_rate": 1.060120039282052e-06, "loss": 1.8602, "step": 34750 }, { "epoch": 0.86, "learning_rate": 1.0597623428500026e-06, "loss": 1.8868, "step": 34751 }, { "epoch": 0.86, "learning_rate": 1.0594047033970012e-06, "loss": 1.8391, "step": 34752 }, { "epoch": 0.86, "learning_rate": 1.0590471209253295e-06, "loss": 1.9113, "step": 34753 }, { "epoch": 0.86, "learning_rate": 1.058689595437261e-06, "loss": 1.9455, "step": 34754 }, { "epoch": 0.86, "learning_rate": 1.0583321269350811e-06, "loss": 1.7993, "step": 34755 }, { "epoch": 0.86, "learning_rate": 1.0579747154210617e-06, "loss": 2.0412, "step": 34756 }, { "epoch": 0.86, "learning_rate": 1.0576173608974838e-06, "loss": 1.8137, "step": 34757 }, { "epoch": 0.86, "learning_rate": 1.0572600633666262e-06, "loss": 1.928, "step": 34758 }, { "epoch": 0.86, "learning_rate": 1.0569028228307632e-06, "loss": 1.9672, "step": 34759 }, { "epoch": 0.86, "learning_rate": 1.0565456392921747e-06, "loss": 1.905, "step": 34760 }, { "epoch": 0.86, "learning_rate": 1.0561885127531334e-06, "loss": 2.0278, "step": 34761 }, { "epoch": 0.86, "learning_rate": 1.0558314432159177e-06, "loss": 2.0203, "step": 34762 }, { "epoch": 0.86, "learning_rate": 1.0554744306828046e-06, "loss": 1.9773, "step": 34763 }, { "epoch": 0.86, "learning_rate": 1.0551174751560667e-06, "loss": 1.9516, "step": 34764 }, { "epoch": 0.86, "learning_rate": 1.05476057663798e-06, "loss": 1.9128, "step": 34765 }, { "epoch": 0.86, "learning_rate": 1.0544037351308178e-06, "loss": 1.9849, "step": 34766 }, { "epoch": 0.86, "learning_rate": 1.0540469506368579e-06, "loss": 2.0439, "step": 34767 }, { "epoch": 0.86, "learning_rate": 1.0536902231583735e-06, "loss": 1.8745, "step": 34768 }, { "epoch": 0.86, "learning_rate": 1.053333552697634e-06, "loss": 2.1302, "step": 34769 }, { "epoch": 0.86, "learning_rate": 1.0529769392569178e-06, "loss": 2.0902, "step": 34770 }, { "epoch": 0.86, "learning_rate": 1.052620382838494e-06, "loss": 1.9151, "step": 34771 }, { "epoch": 0.86, "learning_rate": 1.0522638834446352e-06, "loss": 1.8545, "step": 34772 }, { "epoch": 0.86, "learning_rate": 1.0519074410776175e-06, "loss": 1.848, "step": 34773 }, { "epoch": 0.86, "learning_rate": 1.051551055739708e-06, "loss": 1.9481, "step": 34774 }, { "epoch": 0.86, "learning_rate": 1.0511947274331814e-06, "loss": 1.8514, "step": 34775 }, { "epoch": 0.86, "learning_rate": 1.0508384561603058e-06, "loss": 1.9845, "step": 34776 }, { "epoch": 0.86, "learning_rate": 1.0504822419233519e-06, "loss": 2.2268, "step": 34777 }, { "epoch": 0.86, "learning_rate": 1.0501260847245942e-06, "loss": 1.7844, "step": 34778 }, { "epoch": 0.86, "learning_rate": 1.049769984566298e-06, "loss": 1.7788, "step": 34779 }, { "epoch": 0.86, "learning_rate": 1.0494139414507333e-06, "loss": 1.8934, "step": 34780 }, { "epoch": 0.86, "learning_rate": 1.0490579553801728e-06, "loss": 2.0701, "step": 34781 }, { "epoch": 0.86, "learning_rate": 1.0487020263568814e-06, "loss": 1.9942, "step": 34782 }, { "epoch": 0.86, "learning_rate": 1.0483461543831286e-06, "loss": 1.9873, "step": 34783 }, { "epoch": 0.86, "learning_rate": 1.0479903394611834e-06, "loss": 2.0147, "step": 34784 }, { "epoch": 0.86, "learning_rate": 1.0476345815933154e-06, "loss": 1.8383, "step": 34785 }, { "epoch": 0.86, "learning_rate": 1.0472788807817868e-06, "loss": 1.7828, "step": 34786 }, { "epoch": 0.86, "learning_rate": 1.0469232370288684e-06, "loss": 1.9043, "step": 34787 }, { "epoch": 0.86, "learning_rate": 1.0465676503368282e-06, "loss": 1.7655, "step": 34788 }, { "epoch": 0.86, "learning_rate": 1.0462121207079268e-06, "loss": 1.838, "step": 34789 }, { "epoch": 0.86, "learning_rate": 1.0458566481444366e-06, "loss": 1.9531, "step": 34790 }, { "epoch": 0.86, "learning_rate": 1.045501232648618e-06, "loss": 1.8895, "step": 34791 }, { "epoch": 0.86, "learning_rate": 1.0451458742227382e-06, "loss": 1.9377, "step": 34792 }, { "epoch": 0.86, "learning_rate": 1.0447905728690633e-06, "loss": 2.0945, "step": 34793 }, { "epoch": 0.86, "learning_rate": 1.0444353285898557e-06, "loss": 1.9381, "step": 34794 }, { "epoch": 0.86, "learning_rate": 1.044080141387379e-06, "loss": 1.9577, "step": 34795 }, { "epoch": 0.86, "learning_rate": 1.0437250112639008e-06, "loss": 1.873, "step": 34796 }, { "epoch": 0.86, "learning_rate": 1.04336993822168e-06, "loss": 1.7872, "step": 34797 }, { "epoch": 0.86, "learning_rate": 1.0430149222629814e-06, "loss": 2.1011, "step": 34798 }, { "epoch": 0.86, "learning_rate": 1.0426599633900702e-06, "loss": 1.8214, "step": 34799 }, { "epoch": 0.86, "learning_rate": 1.0423050616052033e-06, "loss": 2.0497, "step": 34800 }, { "epoch": 0.86, "learning_rate": 1.0419502169106466e-06, "loss": 1.9995, "step": 34801 }, { "epoch": 0.86, "learning_rate": 1.0415954293086594e-06, "loss": 1.8584, "step": 34802 }, { "epoch": 0.86, "learning_rate": 1.0412406988015057e-06, "loss": 1.9075, "step": 34803 }, { "epoch": 0.86, "learning_rate": 1.0408860253914432e-06, "loss": 2.0637, "step": 34804 }, { "epoch": 0.86, "learning_rate": 1.0405314090807338e-06, "loss": 2.0034, "step": 34805 }, { "epoch": 0.86, "learning_rate": 1.0401768498716403e-06, "loss": 2.049, "step": 34806 }, { "epoch": 0.86, "learning_rate": 1.039822347766416e-06, "loss": 1.865, "step": 34807 }, { "epoch": 0.86, "learning_rate": 1.0394679027673271e-06, "loss": 1.8653, "step": 34808 }, { "epoch": 0.86, "learning_rate": 1.0391135148766275e-06, "loss": 2.0668, "step": 34809 }, { "epoch": 0.86, "learning_rate": 1.038759184096577e-06, "loss": 2.1324, "step": 34810 }, { "epoch": 0.86, "learning_rate": 1.038404910429437e-06, "loss": 2.0833, "step": 34811 }, { "epoch": 0.86, "learning_rate": 1.0380506938774616e-06, "loss": 1.8587, "step": 34812 }, { "epoch": 0.86, "learning_rate": 1.03769653444291e-06, "loss": 1.9723, "step": 34813 }, { "epoch": 0.86, "learning_rate": 1.037342432128039e-06, "loss": 1.9845, "step": 34814 }, { "epoch": 0.86, "learning_rate": 1.0369883869351049e-06, "loss": 1.8582, "step": 34815 }, { "epoch": 0.86, "learning_rate": 1.0366343988663641e-06, "loss": 1.9764, "step": 34816 }, { "epoch": 0.86, "learning_rate": 1.0362804679240756e-06, "loss": 1.7989, "step": 34817 }, { "epoch": 0.86, "learning_rate": 1.0359265941104912e-06, "loss": 1.7832, "step": 34818 }, { "epoch": 0.86, "learning_rate": 1.0355727774278678e-06, "loss": 1.9183, "step": 34819 }, { "epoch": 0.86, "learning_rate": 1.0352190178784605e-06, "loss": 1.8174, "step": 34820 }, { "epoch": 0.86, "learning_rate": 1.034865315464526e-06, "loss": 1.9965, "step": 34821 }, { "epoch": 0.86, "learning_rate": 1.034511670188315e-06, "loss": 1.9371, "step": 34822 }, { "epoch": 0.86, "learning_rate": 1.0341580820520847e-06, "loss": 1.8769, "step": 34823 }, { "epoch": 0.86, "learning_rate": 1.033804551058084e-06, "loss": 1.9933, "step": 34824 }, { "epoch": 0.86, "learning_rate": 1.0334510772085694e-06, "loss": 2.038, "step": 34825 }, { "epoch": 0.86, "learning_rate": 1.0330976605057952e-06, "loss": 2.0355, "step": 34826 }, { "epoch": 0.86, "learning_rate": 1.03274430095201e-06, "loss": 2.0801, "step": 34827 }, { "epoch": 0.86, "learning_rate": 1.0323909985494697e-06, "loss": 1.7911, "step": 34828 }, { "epoch": 0.86, "learning_rate": 1.0320377533004221e-06, "loss": 2.1085, "step": 34829 }, { "epoch": 0.86, "learning_rate": 1.0316845652071205e-06, "loss": 1.9144, "step": 34830 }, { "epoch": 0.86, "learning_rate": 1.0313314342718172e-06, "loss": 1.8355, "step": 34831 }, { "epoch": 0.86, "learning_rate": 1.0309783604967594e-06, "loss": 1.8771, "step": 34832 }, { "epoch": 0.86, "learning_rate": 1.0306253438842006e-06, "loss": 1.9897, "step": 34833 }, { "epoch": 0.86, "learning_rate": 1.0302723844363883e-06, "loss": 1.8783, "step": 34834 }, { "epoch": 0.86, "learning_rate": 1.0299194821555758e-06, "loss": 1.8776, "step": 34835 }, { "epoch": 0.86, "learning_rate": 1.0295666370440083e-06, "loss": 1.8216, "step": 34836 }, { "epoch": 0.86, "learning_rate": 1.029213849103935e-06, "loss": 1.8806, "step": 34837 }, { "epoch": 0.86, "learning_rate": 1.0288611183376074e-06, "loss": 1.9339, "step": 34838 }, { "epoch": 0.86, "learning_rate": 1.0285084447472704e-06, "loss": 1.9965, "step": 34839 }, { "epoch": 0.86, "learning_rate": 1.0281558283351722e-06, "loss": 1.9761, "step": 34840 }, { "epoch": 0.86, "learning_rate": 1.0278032691035623e-06, "loss": 2.0376, "step": 34841 }, { "epoch": 0.86, "learning_rate": 1.0274507670546852e-06, "loss": 1.8928, "step": 34842 }, { "epoch": 0.86, "learning_rate": 1.0270983221907892e-06, "loss": 1.7074, "step": 34843 }, { "epoch": 0.86, "learning_rate": 1.0267459345141183e-06, "loss": 1.9974, "step": 34844 }, { "epoch": 0.86, "learning_rate": 1.0263936040269207e-06, "loss": 1.9721, "step": 34845 }, { "epoch": 0.86, "learning_rate": 1.0260413307314422e-06, "loss": 1.7984, "step": 34846 }, { "epoch": 0.86, "learning_rate": 1.0256891146299254e-06, "loss": 1.9523, "step": 34847 }, { "epoch": 0.86, "learning_rate": 1.0253369557246173e-06, "loss": 1.7351, "step": 34848 }, { "epoch": 0.86, "learning_rate": 1.0249848540177608e-06, "loss": 1.9338, "step": 34849 }, { "epoch": 0.86, "learning_rate": 1.0246328095115999e-06, "loss": 1.9682, "step": 34850 }, { "epoch": 0.86, "learning_rate": 1.02428082220838e-06, "loss": 1.9904, "step": 34851 }, { "epoch": 0.86, "learning_rate": 1.023928892110344e-06, "loss": 1.8976, "step": 34852 }, { "epoch": 0.86, "learning_rate": 1.0235770192197337e-06, "loss": 1.8719, "step": 34853 }, { "epoch": 0.86, "learning_rate": 1.0232252035387924e-06, "loss": 2.0924, "step": 34854 }, { "epoch": 0.86, "learning_rate": 1.022873445069762e-06, "loss": 1.8108, "step": 34855 }, { "epoch": 0.86, "learning_rate": 1.022521743814886e-06, "loss": 1.9741, "step": 34856 }, { "epoch": 0.86, "learning_rate": 1.022170099776404e-06, "loss": 1.9214, "step": 34857 }, { "epoch": 0.86, "learning_rate": 1.0218185129565572e-06, "loss": 1.9924, "step": 34858 }, { "epoch": 0.86, "learning_rate": 1.0214669833575886e-06, "loss": 1.8512, "step": 34859 }, { "epoch": 0.86, "learning_rate": 1.0211155109817351e-06, "loss": 1.8526, "step": 34860 }, { "epoch": 0.86, "learning_rate": 1.0207640958312404e-06, "loss": 2.093, "step": 34861 }, { "epoch": 0.86, "learning_rate": 1.0204127379083407e-06, "loss": 2.0027, "step": 34862 }, { "epoch": 0.86, "learning_rate": 1.0200614372152761e-06, "loss": 1.9113, "step": 34863 }, { "epoch": 0.86, "learning_rate": 1.0197101937542898e-06, "loss": 1.8713, "step": 34864 }, { "epoch": 0.86, "learning_rate": 1.0193590075276138e-06, "loss": 1.7316, "step": 34865 }, { "epoch": 0.86, "learning_rate": 1.0190078785374925e-06, "loss": 2.0423, "step": 34866 }, { "epoch": 0.86, "learning_rate": 1.018656806786158e-06, "loss": 1.9204, "step": 34867 }, { "epoch": 0.86, "learning_rate": 1.018305792275851e-06, "loss": 1.849, "step": 34868 }, { "epoch": 0.86, "learning_rate": 1.0179548350088087e-06, "loss": 1.8092, "step": 34869 }, { "epoch": 0.86, "learning_rate": 1.017603934987269e-06, "loss": 1.8752, "step": 34870 }, { "epoch": 0.86, "learning_rate": 1.0172530922134638e-06, "loss": 1.8716, "step": 34871 }, { "epoch": 0.86, "learning_rate": 1.0169023066896333e-06, "loss": 2.1564, "step": 34872 }, { "epoch": 0.86, "learning_rate": 1.0165515784180114e-06, "loss": 1.8797, "step": 34873 }, { "epoch": 0.86, "learning_rate": 1.016200907400836e-06, "loss": 1.9302, "step": 34874 }, { "epoch": 0.86, "learning_rate": 1.0158502936403382e-06, "loss": 1.9015, "step": 34875 }, { "epoch": 0.86, "learning_rate": 1.0154997371387565e-06, "loss": 2.0083, "step": 34876 }, { "epoch": 0.86, "learning_rate": 1.0151492378983198e-06, "loss": 1.8792, "step": 34877 }, { "epoch": 0.86, "learning_rate": 1.0147987959212658e-06, "loss": 1.9165, "step": 34878 }, { "epoch": 0.86, "learning_rate": 1.0144484112098297e-06, "loss": 1.889, "step": 34879 }, { "epoch": 0.86, "learning_rate": 1.0140980837662394e-06, "loss": 1.8717, "step": 34880 }, { "epoch": 0.86, "learning_rate": 1.0137478135927325e-06, "loss": 1.9187, "step": 34881 }, { "epoch": 0.86, "learning_rate": 1.0133976006915381e-06, "loss": 2.0323, "step": 34882 }, { "epoch": 0.86, "learning_rate": 1.0130474450648886e-06, "loss": 2.2144, "step": 34883 }, { "epoch": 0.86, "learning_rate": 1.0126973467150192e-06, "loss": 1.7381, "step": 34884 }, { "epoch": 0.86, "learning_rate": 1.0123473056441556e-06, "loss": 1.9647, "step": 34885 }, { "epoch": 0.86, "learning_rate": 1.0119973218545331e-06, "loss": 1.9544, "step": 34886 }, { "epoch": 0.86, "learning_rate": 1.0116473953483796e-06, "loss": 2.022, "step": 34887 }, { "epoch": 0.86, "learning_rate": 1.011297526127929e-06, "loss": 1.8674, "step": 34888 }, { "epoch": 0.86, "learning_rate": 1.010947714195406e-06, "loss": 1.8821, "step": 34889 }, { "epoch": 0.86, "learning_rate": 1.0105979595530424e-06, "loss": 1.8683, "step": 34890 }, { "epoch": 0.86, "learning_rate": 1.0102482622030707e-06, "loss": 1.7123, "step": 34891 }, { "epoch": 0.86, "learning_rate": 1.0098986221477136e-06, "loss": 1.9775, "step": 34892 }, { "epoch": 0.86, "learning_rate": 1.0095490393892027e-06, "loss": 2.0977, "step": 34893 }, { "epoch": 0.86, "learning_rate": 1.0091995139297684e-06, "loss": 1.8883, "step": 34894 }, { "epoch": 0.86, "learning_rate": 1.0088500457716321e-06, "loss": 1.769, "step": 34895 }, { "epoch": 0.86, "learning_rate": 1.0085006349170279e-06, "loss": 1.9602, "step": 34896 }, { "epoch": 0.86, "learning_rate": 1.008151281368177e-06, "loss": 1.9787, "step": 34897 }, { "epoch": 0.86, "learning_rate": 1.007801985127308e-06, "loss": 1.823, "step": 34898 }, { "epoch": 0.86, "learning_rate": 1.00745274619665e-06, "loss": 1.914, "step": 34899 }, { "epoch": 0.86, "learning_rate": 1.0071035645784232e-06, "loss": 2.0742, "step": 34900 }, { "epoch": 0.86, "learning_rate": 1.0067544402748585e-06, "loss": 2.014, "step": 34901 }, { "epoch": 0.86, "learning_rate": 1.0064053732881773e-06, "loss": 1.8664, "step": 34902 }, { "epoch": 0.86, "learning_rate": 1.0060563636206056e-06, "loss": 2.0233, "step": 34903 }, { "epoch": 0.86, "learning_rate": 1.005707411274367e-06, "loss": 1.8837, "step": 34904 }, { "epoch": 0.86, "learning_rate": 1.0053585162516865e-06, "loss": 1.9088, "step": 34905 }, { "epoch": 0.86, "learning_rate": 1.005009678554789e-06, "loss": 1.8739, "step": 34906 }, { "epoch": 0.86, "learning_rate": 1.0046608981858953e-06, "loss": 1.7546, "step": 34907 }, { "epoch": 0.86, "learning_rate": 1.0043121751472284e-06, "loss": 1.7686, "step": 34908 }, { "epoch": 0.86, "learning_rate": 1.0039635094410127e-06, "loss": 2.1045, "step": 34909 }, { "epoch": 0.86, "learning_rate": 1.0036149010694685e-06, "loss": 1.9247, "step": 34910 }, { "epoch": 0.86, "learning_rate": 1.0032663500348182e-06, "loss": 1.928, "step": 34911 }, { "epoch": 0.86, "learning_rate": 1.002917856339286e-06, "loss": 1.9201, "step": 34912 }, { "epoch": 0.86, "learning_rate": 1.0025694199850877e-06, "loss": 1.9424, "step": 34913 }, { "epoch": 0.86, "learning_rate": 1.002221040974448e-06, "loss": 1.9118, "step": 34914 }, { "epoch": 0.86, "learning_rate": 1.0018727193095845e-06, "loss": 1.839, "step": 34915 }, { "epoch": 0.86, "learning_rate": 1.0015244549927195e-06, "loss": 1.9329, "step": 34916 }, { "epoch": 0.86, "learning_rate": 1.0011762480260722e-06, "loss": 1.6991, "step": 34917 }, { "epoch": 0.86, "learning_rate": 1.0008280984118601e-06, "loss": 1.7526, "step": 34918 }, { "epoch": 0.86, "learning_rate": 1.0004800061523057e-06, "loss": 2.017, "step": 34919 }, { "epoch": 0.86, "learning_rate": 1.0001319712496217e-06, "loss": 1.9448, "step": 34920 }, { "epoch": 0.86, "learning_rate": 9.997839937060306e-07, "loss": 1.9708, "step": 34921 }, { "epoch": 0.86, "learning_rate": 9.994360735237497e-07, "loss": 1.9707, "step": 34922 }, { "epoch": 0.86, "learning_rate": 9.99088210704995e-07, "loss": 1.9639, "step": 34923 }, { "epoch": 0.86, "learning_rate": 9.987404052519866e-07, "loss": 1.925, "step": 34924 }, { "epoch": 0.86, "learning_rate": 9.983926571669366e-07, "loss": 1.9855, "step": 34925 }, { "epoch": 0.86, "learning_rate": 9.98044966452064e-07, "loss": 1.8129, "step": 34926 }, { "epoch": 0.86, "learning_rate": 9.976973331095874e-07, "loss": 1.8621, "step": 34927 }, { "epoch": 0.86, "learning_rate": 9.973497571417168e-07, "loss": 1.8422, "step": 34928 }, { "epoch": 0.86, "learning_rate": 9.970022385506716e-07, "loss": 1.8738, "step": 34929 }, { "epoch": 0.86, "learning_rate": 9.966547773386638e-07, "loss": 2.1491, "step": 34930 }, { "epoch": 0.86, "learning_rate": 9.963073735079088e-07, "loss": 2.0028, "step": 34931 }, { "epoch": 0.86, "learning_rate": 9.95960027060624e-07, "loss": 1.8909, "step": 34932 }, { "epoch": 0.86, "learning_rate": 9.956127379990167e-07, "loss": 1.9791, "step": 34933 }, { "epoch": 0.86, "learning_rate": 9.952655063253069e-07, "loss": 1.882, "step": 34934 }, { "epoch": 0.86, "learning_rate": 9.949183320417022e-07, "loss": 1.7739, "step": 34935 }, { "epoch": 0.86, "learning_rate": 9.945712151504183e-07, "loss": 1.8851, "step": 34936 }, { "epoch": 0.86, "learning_rate": 9.94224155653668e-07, "loss": 1.9979, "step": 34937 }, { "epoch": 0.86, "learning_rate": 9.938771535536607e-07, "loss": 1.9251, "step": 34938 }, { "epoch": 0.86, "learning_rate": 9.93530208852609e-07, "loss": 2.0003, "step": 34939 }, { "epoch": 0.86, "learning_rate": 9.931833215527252e-07, "loss": 1.9171, "step": 34940 }, { "epoch": 0.86, "learning_rate": 9.928364916562184e-07, "loss": 2.0383, "step": 34941 }, { "epoch": 0.86, "learning_rate": 9.924897191653016e-07, "loss": 1.8543, "step": 34942 }, { "epoch": 0.86, "learning_rate": 9.921430040821822e-07, "loss": 1.8394, "step": 34943 }, { "epoch": 0.86, "learning_rate": 9.91796346409073e-07, "loss": 1.8543, "step": 34944 }, { "epoch": 0.86, "learning_rate": 9.914497461481787e-07, "loss": 1.6875, "step": 34945 }, { "epoch": 0.86, "learning_rate": 9.91103203301711e-07, "loss": 1.7316, "step": 34946 }, { "epoch": 0.86, "learning_rate": 9.907567178718814e-07, "loss": 2.0679, "step": 34947 }, { "epoch": 0.86, "learning_rate": 9.904102898608926e-07, "loss": 1.9196, "step": 34948 }, { "epoch": 0.86, "learning_rate": 9.90063919270957e-07, "loss": 1.8943, "step": 34949 }, { "epoch": 0.86, "learning_rate": 9.89717606104279e-07, "loss": 1.9007, "step": 34950 }, { "epoch": 0.86, "learning_rate": 9.893713503630663e-07, "loss": 1.9031, "step": 34951 }, { "epoch": 0.86, "learning_rate": 9.890251520495287e-07, "loss": 1.9274, "step": 34952 }, { "epoch": 0.86, "learning_rate": 9.886790111658685e-07, "loss": 1.8016, "step": 34953 }, { "epoch": 0.86, "learning_rate": 9.883329277142961e-07, "loss": 1.9365, "step": 34954 }, { "epoch": 0.86, "learning_rate": 9.879869016970123e-07, "loss": 2.0315, "step": 34955 }, { "epoch": 0.86, "learning_rate": 9.87640933116225e-07, "loss": 1.7819, "step": 34956 }, { "epoch": 0.86, "learning_rate": 9.872950219741384e-07, "loss": 1.7714, "step": 34957 }, { "epoch": 0.86, "learning_rate": 9.869491682729581e-07, "loss": 1.9954, "step": 34958 }, { "epoch": 0.86, "learning_rate": 9.866033720148904e-07, "loss": 1.8655, "step": 34959 }, { "epoch": 0.86, "learning_rate": 9.862576332021357e-07, "loss": 1.9216, "step": 34960 }, { "epoch": 0.86, "learning_rate": 9.859119518368976e-07, "loss": 2.0281, "step": 34961 }, { "epoch": 0.86, "learning_rate": 9.855663279213811e-07, "loss": 2.1148, "step": 34962 }, { "epoch": 0.86, "learning_rate": 9.85220761457788e-07, "loss": 2.0305, "step": 34963 }, { "epoch": 0.86, "learning_rate": 9.848752524483217e-07, "loss": 1.9791, "step": 34964 }, { "epoch": 0.86, "learning_rate": 9.84529800895182e-07, "loss": 2.0496, "step": 34965 }, { "epoch": 0.86, "learning_rate": 9.841844068005713e-07, "loss": 1.8446, "step": 34966 }, { "epoch": 0.86, "learning_rate": 9.838390701666945e-07, "loss": 2.0358, "step": 34967 }, { "epoch": 0.86, "learning_rate": 9.834937909957464e-07, "loss": 1.9748, "step": 34968 }, { "epoch": 0.86, "learning_rate": 9.831485692899335e-07, "loss": 2.1184, "step": 34969 }, { "epoch": 0.86, "learning_rate": 9.828034050514512e-07, "loss": 1.9106, "step": 34970 }, { "epoch": 0.86, "learning_rate": 9.824582982825014e-07, "loss": 1.969, "step": 34971 }, { "epoch": 0.86, "learning_rate": 9.821132489852857e-07, "loss": 1.9947, "step": 34972 }, { "epoch": 0.86, "learning_rate": 9.81768257161999e-07, "loss": 1.9038, "step": 34973 }, { "epoch": 0.86, "learning_rate": 9.814233228148418e-07, "loss": 1.9289, "step": 34974 }, { "epoch": 0.86, "learning_rate": 9.810784459460133e-07, "loss": 1.8777, "step": 34975 }, { "epoch": 0.86, "learning_rate": 9.807336265577106e-07, "loss": 1.899, "step": 34976 }, { "epoch": 0.86, "learning_rate": 9.803888646521343e-07, "loss": 1.7944, "step": 34977 }, { "epoch": 0.86, "learning_rate": 9.80044160231477e-07, "loss": 1.918, "step": 34978 }, { "epoch": 0.86, "learning_rate": 9.79699513297938e-07, "loss": 1.9593, "step": 34979 }, { "epoch": 0.86, "learning_rate": 9.793549238537147e-07, "loss": 1.9878, "step": 34980 }, { "epoch": 0.86, "learning_rate": 9.790103919010008e-07, "loss": 2.0144, "step": 34981 }, { "epoch": 0.86, "learning_rate": 9.786659174419955e-07, "loss": 2.0595, "step": 34982 }, { "epoch": 0.86, "learning_rate": 9.783215004788905e-07, "loss": 1.8755, "step": 34983 }, { "epoch": 0.86, "learning_rate": 9.779771410138828e-07, "loss": 2.0802, "step": 34984 }, { "epoch": 0.86, "learning_rate": 9.776328390491685e-07, "loss": 2.0677, "step": 34985 }, { "epoch": 0.86, "learning_rate": 9.772885945869393e-07, "loss": 1.8837, "step": 34986 }, { "epoch": 0.86, "learning_rate": 9.76944407629391e-07, "loss": 2.0184, "step": 34987 }, { "epoch": 0.86, "learning_rate": 9.766002781787153e-07, "loss": 1.9033, "step": 34988 }, { "epoch": 0.86, "learning_rate": 9.762562062371072e-07, "loss": 1.7817, "step": 34989 }, { "epoch": 0.86, "learning_rate": 9.759121918067594e-07, "loss": 1.921, "step": 34990 }, { "epoch": 0.86, "learning_rate": 9.755682348898632e-07, "loss": 1.7886, "step": 34991 }, { "epoch": 0.86, "learning_rate": 9.752243354886115e-07, "loss": 2.0223, "step": 34992 }, { "epoch": 0.86, "learning_rate": 9.74880493605197e-07, "loss": 1.9236, "step": 34993 }, { "epoch": 0.86, "learning_rate": 9.74536709241809e-07, "loss": 1.95, "step": 34994 }, { "epoch": 0.86, "learning_rate": 9.741929824006424e-07, "loss": 2.0096, "step": 34995 }, { "epoch": 0.86, "learning_rate": 9.738493130838833e-07, "loss": 1.9015, "step": 34996 }, { "epoch": 0.86, "learning_rate": 9.735057012937255e-07, "loss": 1.8702, "step": 34997 }, { "epoch": 0.86, "learning_rate": 9.73162147032356e-07, "loss": 1.8718, "step": 34998 }, { "epoch": 0.86, "learning_rate": 9.728186503019665e-07, "loss": 1.8244, "step": 34999 }, { "epoch": 0.86, "learning_rate": 9.724752111047475e-07, "loss": 1.9605, "step": 35000 }, { "epoch": 0.86, "learning_rate": 9.721318294428827e-07, "loss": 1.7473, "step": 35001 }, { "epoch": 0.86, "learning_rate": 9.717885053185673e-07, "loss": 1.9668, "step": 35002 }, { "epoch": 0.86, "learning_rate": 9.714452387339834e-07, "loss": 1.9216, "step": 35003 }, { "epoch": 0.86, "learning_rate": 9.71102029691321e-07, "loss": 1.8153, "step": 35004 }, { "epoch": 0.86, "learning_rate": 9.7075887819277e-07, "loss": 2.0264, "step": 35005 }, { "epoch": 0.86, "learning_rate": 9.704157842405137e-07, "loss": 1.7275, "step": 35006 }, { "epoch": 0.86, "learning_rate": 9.70072747836741e-07, "loss": 1.9759, "step": 35007 }, { "epoch": 0.86, "learning_rate": 9.697297689836371e-07, "loss": 2.0194, "step": 35008 }, { "epoch": 0.86, "learning_rate": 9.69386847683388e-07, "loss": 1.9543, "step": 35009 }, { "epoch": 0.86, "learning_rate": 9.690439839381783e-07, "loss": 1.8507, "step": 35010 }, { "epoch": 0.86, "learning_rate": 9.687011777501953e-07, "loss": 1.9003, "step": 35011 }, { "epoch": 0.86, "learning_rate": 9.683584291216242e-07, "loss": 1.9359, "step": 35012 }, { "epoch": 0.86, "learning_rate": 9.680157380546474e-07, "loss": 2.0498, "step": 35013 }, { "epoch": 0.86, "learning_rate": 9.676731045514486e-07, "loss": 1.8591, "step": 35014 }, { "epoch": 0.86, "learning_rate": 9.673305286142154e-07, "loss": 1.9248, "step": 35015 }, { "epoch": 0.86, "learning_rate": 9.669880102451258e-07, "loss": 1.8846, "step": 35016 }, { "epoch": 0.86, "learning_rate": 9.666455494463679e-07, "loss": 1.8973, "step": 35017 }, { "epoch": 0.86, "learning_rate": 9.663031462201188e-07, "loss": 1.9564, "step": 35018 }, { "epoch": 0.86, "learning_rate": 9.65960800568565e-07, "loss": 1.7637, "step": 35019 }, { "epoch": 0.86, "learning_rate": 9.656185124938888e-07, "loss": 1.9429, "step": 35020 }, { "epoch": 0.86, "learning_rate": 9.652762819982674e-07, "loss": 1.9981, "step": 35021 }, { "epoch": 0.86, "learning_rate": 9.64934109083887e-07, "loss": 1.9391, "step": 35022 }, { "epoch": 0.86, "learning_rate": 9.645919937529235e-07, "loss": 1.936, "step": 35023 }, { "epoch": 0.86, "learning_rate": 9.642499360075607e-07, "loss": 1.9767, "step": 35024 }, { "epoch": 0.86, "learning_rate": 9.639079358499792e-07, "loss": 1.9358, "step": 35025 }, { "epoch": 0.86, "learning_rate": 9.635659932823548e-07, "loss": 1.8143, "step": 35026 }, { "epoch": 0.86, "learning_rate": 9.632241083068704e-07, "loss": 1.9121, "step": 35027 }, { "epoch": 0.86, "learning_rate": 9.628822809257033e-07, "loss": 1.9646, "step": 35028 }, { "epoch": 0.86, "learning_rate": 9.625405111410325e-07, "loss": 1.9703, "step": 35029 }, { "epoch": 0.86, "learning_rate": 9.621987989550374e-07, "loss": 2.1141, "step": 35030 }, { "epoch": 0.86, "learning_rate": 9.618571443698932e-07, "loss": 1.8789, "step": 35031 }, { "epoch": 0.86, "learning_rate": 9.61515547387779e-07, "loss": 1.9722, "step": 35032 }, { "epoch": 0.86, "learning_rate": 9.611740080108744e-07, "loss": 2.0978, "step": 35033 }, { "epoch": 0.86, "learning_rate": 9.608325262413508e-07, "loss": 1.8537, "step": 35034 }, { "epoch": 0.86, "learning_rate": 9.60491102081389e-07, "loss": 1.8211, "step": 35035 }, { "epoch": 0.86, "learning_rate": 9.601497355331624e-07, "loss": 1.7882, "step": 35036 }, { "epoch": 0.86, "learning_rate": 9.598084265988461e-07, "loss": 1.8945, "step": 35037 }, { "epoch": 0.86, "learning_rate": 9.594671752806196e-07, "loss": 1.9022, "step": 35038 }, { "epoch": 0.86, "learning_rate": 9.591259815806531e-07, "loss": 1.7156, "step": 35039 }, { "epoch": 0.86, "learning_rate": 9.58784845501125e-07, "loss": 1.9703, "step": 35040 }, { "epoch": 0.86, "learning_rate": 9.584437670442048e-07, "loss": 1.8283, "step": 35041 }, { "epoch": 0.86, "learning_rate": 9.581027462120695e-07, "loss": 1.8556, "step": 35042 }, { "epoch": 0.86, "learning_rate": 9.577617830068952e-07, "loss": 2.1621, "step": 35043 }, { "epoch": 0.86, "learning_rate": 9.57420877430848e-07, "loss": 1.9863, "step": 35044 }, { "epoch": 0.86, "learning_rate": 9.57080029486106e-07, "loss": 1.8931, "step": 35045 }, { "epoch": 0.86, "learning_rate": 9.567392391748386e-07, "loss": 1.8528, "step": 35046 }, { "epoch": 0.86, "learning_rate": 9.563985064992198e-07, "loss": 1.7369, "step": 35047 }, { "epoch": 0.86, "learning_rate": 9.560578314614232e-07, "loss": 1.8255, "step": 35048 }, { "epoch": 0.86, "learning_rate": 9.557172140636151e-07, "loss": 1.9925, "step": 35049 }, { "epoch": 0.86, "learning_rate": 9.553766543079702e-07, "loss": 1.9161, "step": 35050 }, { "epoch": 0.86, "learning_rate": 9.550361521966557e-07, "loss": 1.8946, "step": 35051 }, { "epoch": 0.86, "learning_rate": 9.54695707731843e-07, "loss": 2.0852, "step": 35052 }, { "epoch": 0.86, "learning_rate": 9.54355320915704e-07, "loss": 1.8853, "step": 35053 }, { "epoch": 0.86, "learning_rate": 9.540149917504049e-07, "loss": 1.9468, "step": 35054 }, { "epoch": 0.86, "learning_rate": 9.536747202381192e-07, "loss": 1.8034, "step": 35055 }, { "epoch": 0.86, "learning_rate": 9.533345063810096e-07, "loss": 1.9841, "step": 35056 }, { "epoch": 0.86, "learning_rate": 9.529943501812478e-07, "loss": 1.8921, "step": 35057 }, { "epoch": 0.86, "learning_rate": 9.526542516410032e-07, "loss": 1.8883, "step": 35058 }, { "epoch": 0.86, "learning_rate": 9.523142107624395e-07, "loss": 1.8495, "step": 35059 }, { "epoch": 0.86, "learning_rate": 9.519742275477272e-07, "loss": 1.9589, "step": 35060 }, { "epoch": 0.86, "learning_rate": 9.516343019990304e-07, "loss": 1.8266, "step": 35061 }, { "epoch": 0.86, "learning_rate": 9.512944341185171e-07, "loss": 1.9234, "step": 35062 }, { "epoch": 0.86, "learning_rate": 9.509546239083534e-07, "loss": 1.8107, "step": 35063 }, { "epoch": 0.86, "learning_rate": 9.506148713707053e-07, "loss": 1.7988, "step": 35064 }, { "epoch": 0.86, "learning_rate": 9.50275176507739e-07, "loss": 1.8582, "step": 35065 }, { "epoch": 0.86, "learning_rate": 9.499355393216158e-07, "loss": 1.9664, "step": 35066 }, { "epoch": 0.86, "learning_rate": 9.495959598145032e-07, "loss": 2.0277, "step": 35067 }, { "epoch": 0.86, "learning_rate": 9.492564379885672e-07, "loss": 1.929, "step": 35068 }, { "epoch": 0.86, "learning_rate": 9.48916973845968e-07, "loss": 1.9569, "step": 35069 }, { "epoch": 0.86, "learning_rate": 9.485775673888719e-07, "loss": 1.8294, "step": 35070 }, { "epoch": 0.86, "learning_rate": 9.482382186194383e-07, "loss": 1.9773, "step": 35071 }, { "epoch": 0.86, "learning_rate": 9.478989275398331e-07, "loss": 2.0125, "step": 35072 }, { "epoch": 0.86, "learning_rate": 9.475596941522202e-07, "loss": 1.7784, "step": 35073 }, { "epoch": 0.86, "learning_rate": 9.472205184587569e-07, "loss": 1.9165, "step": 35074 }, { "epoch": 0.86, "learning_rate": 9.4688140046161e-07, "loss": 1.8837, "step": 35075 }, { "epoch": 0.86, "learning_rate": 9.465423401629359e-07, "loss": 2.0858, "step": 35076 }, { "epoch": 0.86, "learning_rate": 9.462033375648971e-07, "loss": 2.0154, "step": 35077 }, { "epoch": 0.86, "learning_rate": 9.458643926696587e-07, "loss": 1.9059, "step": 35078 }, { "epoch": 0.86, "learning_rate": 9.455255054793743e-07, "loss": 1.9981, "step": 35079 }, { "epoch": 0.86, "learning_rate": 9.451866759962058e-07, "loss": 1.8915, "step": 35080 }, { "epoch": 0.86, "learning_rate": 9.448479042223147e-07, "loss": 1.902, "step": 35081 }, { "epoch": 0.86, "learning_rate": 9.44509190159859e-07, "loss": 1.8887, "step": 35082 }, { "epoch": 0.86, "learning_rate": 9.441705338109985e-07, "loss": 1.9007, "step": 35083 }, { "epoch": 0.86, "learning_rate": 9.438319351778891e-07, "loss": 2.0743, "step": 35084 }, { "epoch": 0.86, "learning_rate": 9.434933942626912e-07, "loss": 1.9315, "step": 35085 }, { "epoch": 0.86, "learning_rate": 9.431549110675597e-07, "loss": 1.9336, "step": 35086 }, { "epoch": 0.86, "learning_rate": 9.428164855946543e-07, "loss": 1.9283, "step": 35087 }, { "epoch": 0.86, "learning_rate": 9.424781178461317e-07, "loss": 1.8065, "step": 35088 }, { "epoch": 0.86, "learning_rate": 9.421398078241473e-07, "loss": 2.0721, "step": 35089 }, { "epoch": 0.86, "learning_rate": 9.418015555308569e-07, "loss": 1.8674, "step": 35090 }, { "epoch": 0.86, "learning_rate": 9.414633609684188e-07, "loss": 1.7395, "step": 35091 }, { "epoch": 0.86, "learning_rate": 9.411252241389856e-07, "loss": 2.0767, "step": 35092 }, { "epoch": 0.86, "learning_rate": 9.407871450447148e-07, "loss": 1.9315, "step": 35093 }, { "epoch": 0.87, "learning_rate": 9.404491236877589e-07, "loss": 2.0344, "step": 35094 }, { "epoch": 0.87, "learning_rate": 9.401111600702717e-07, "loss": 1.8848, "step": 35095 }, { "epoch": 0.87, "learning_rate": 9.397732541944116e-07, "loss": 2.0013, "step": 35096 }, { "epoch": 0.87, "learning_rate": 9.394354060623268e-07, "loss": 1.9069, "step": 35097 }, { "epoch": 0.87, "learning_rate": 9.390976156761733e-07, "loss": 1.925, "step": 35098 }, { "epoch": 0.87, "learning_rate": 9.38759883038104e-07, "loss": 1.9143, "step": 35099 }, { "epoch": 0.87, "learning_rate": 9.384222081502703e-07, "loss": 1.9432, "step": 35100 }, { "epoch": 0.87, "learning_rate": 9.380845910148262e-07, "loss": 1.8468, "step": 35101 }, { "epoch": 0.87, "learning_rate": 9.37747031633921e-07, "loss": 1.8454, "step": 35102 }, { "epoch": 0.87, "learning_rate": 9.374095300097097e-07, "loss": 2.1192, "step": 35103 }, { "epoch": 0.87, "learning_rate": 9.370720861443383e-07, "loss": 1.9963, "step": 35104 }, { "epoch": 0.87, "learning_rate": 9.367347000399607e-07, "loss": 1.8825, "step": 35105 }, { "epoch": 0.87, "learning_rate": 9.363973716987274e-07, "loss": 2.0808, "step": 35106 }, { "epoch": 0.87, "learning_rate": 9.360601011227866e-07, "loss": 2.0757, "step": 35107 }, { "epoch": 0.87, "learning_rate": 9.357228883142899e-07, "loss": 1.9072, "step": 35108 }, { "epoch": 0.87, "learning_rate": 9.353857332753824e-07, "loss": 1.9306, "step": 35109 }, { "epoch": 0.87, "learning_rate": 9.350486360082167e-07, "loss": 2.0304, "step": 35110 }, { "epoch": 0.87, "learning_rate": 9.347115965149423e-07, "loss": 2.0659, "step": 35111 }, { "epoch": 0.87, "learning_rate": 9.343746147977028e-07, "loss": 2.0925, "step": 35112 }, { "epoch": 0.87, "learning_rate": 9.340376908586501e-07, "loss": 1.9083, "step": 35113 }, { "epoch": 0.87, "learning_rate": 9.33700824699929e-07, "loss": 1.8273, "step": 35114 }, { "epoch": 0.87, "learning_rate": 9.333640163236856e-07, "loss": 1.8901, "step": 35115 }, { "epoch": 0.87, "learning_rate": 9.330272657320694e-07, "loss": 2.0232, "step": 35116 }, { "epoch": 0.87, "learning_rate": 9.32690572927224e-07, "loss": 1.7999, "step": 35117 }, { "epoch": 0.87, "learning_rate": 9.323539379112989e-07, "loss": 1.8308, "step": 35118 }, { "epoch": 0.87, "learning_rate": 9.320173606864358e-07, "loss": 2.0594, "step": 35119 }, { "epoch": 0.87, "learning_rate": 9.316808412547806e-07, "loss": 1.8282, "step": 35120 }, { "epoch": 0.87, "learning_rate": 9.313443796184807e-07, "loss": 1.8652, "step": 35121 }, { "epoch": 0.87, "learning_rate": 9.310079757796775e-07, "loss": 1.9405, "step": 35122 }, { "epoch": 0.87, "learning_rate": 9.306716297405161e-07, "loss": 1.8204, "step": 35123 }, { "epoch": 0.87, "learning_rate": 9.30335341503139e-07, "loss": 1.7587, "step": 35124 }, { "epoch": 0.87, "learning_rate": 9.299991110696916e-07, "loss": 1.9045, "step": 35125 }, { "epoch": 0.87, "learning_rate": 9.296629384423161e-07, "loss": 1.8294, "step": 35126 }, { "epoch": 0.87, "learning_rate": 9.293268236231534e-07, "loss": 1.8968, "step": 35127 }, { "epoch": 0.87, "learning_rate": 9.289907666143483e-07, "loss": 1.913, "step": 35128 }, { "epoch": 0.87, "learning_rate": 9.286547674180402e-07, "loss": 1.94, "step": 35129 }, { "epoch": 0.87, "learning_rate": 9.283188260363707e-07, "loss": 2.0353, "step": 35130 }, { "epoch": 0.87, "learning_rate": 9.279829424714836e-07, "loss": 1.8437, "step": 35131 }, { "epoch": 0.87, "learning_rate": 9.276471167255164e-07, "loss": 1.8743, "step": 35132 }, { "epoch": 0.87, "learning_rate": 9.273113488006102e-07, "loss": 1.9391, "step": 35133 }, { "epoch": 0.87, "learning_rate": 9.269756386989048e-07, "loss": 1.797, "step": 35134 }, { "epoch": 0.87, "learning_rate": 9.266399864225417e-07, "loss": 2.002, "step": 35135 }, { "epoch": 0.87, "learning_rate": 9.263043919736603e-07, "loss": 1.8186, "step": 35136 }, { "epoch": 0.87, "learning_rate": 9.259688553543966e-07, "loss": 1.8667, "step": 35137 }, { "epoch": 0.87, "learning_rate": 9.256333765668923e-07, "loss": 1.8707, "step": 35138 }, { "epoch": 0.87, "learning_rate": 9.252979556132812e-07, "loss": 1.7429, "step": 35139 }, { "epoch": 0.87, "learning_rate": 9.24962592495704e-07, "loss": 1.8617, "step": 35140 }, { "epoch": 0.87, "learning_rate": 9.246272872162998e-07, "loss": 1.9128, "step": 35141 }, { "epoch": 0.87, "learning_rate": 9.242920397772016e-07, "loss": 2.0504, "step": 35142 }, { "epoch": 0.87, "learning_rate": 9.239568501805485e-07, "loss": 1.89, "step": 35143 }, { "epoch": 0.87, "learning_rate": 9.236217184284757e-07, "loss": 1.8255, "step": 35144 }, { "epoch": 0.87, "learning_rate": 9.232866445231192e-07, "loss": 1.879, "step": 35145 }, { "epoch": 0.87, "learning_rate": 9.229516284666174e-07, "loss": 1.7667, "step": 35146 }, { "epoch": 0.87, "learning_rate": 9.226166702611005e-07, "loss": 1.9034, "step": 35147 }, { "epoch": 0.87, "learning_rate": 9.22281769908705e-07, "loss": 1.9378, "step": 35148 }, { "epoch": 0.87, "learning_rate": 9.219469274115689e-07, "loss": 1.9885, "step": 35149 }, { "epoch": 0.87, "learning_rate": 9.216121427718216e-07, "loss": 2.0054, "step": 35150 }, { "epoch": 0.87, "learning_rate": 9.212774159915993e-07, "loss": 1.8059, "step": 35151 }, { "epoch": 0.87, "learning_rate": 9.209427470730336e-07, "loss": 1.8701, "step": 35152 }, { "epoch": 0.87, "learning_rate": 9.206081360182595e-07, "loss": 1.7846, "step": 35153 }, { "epoch": 0.87, "learning_rate": 9.202735828294096e-07, "loss": 1.8056, "step": 35154 }, { "epoch": 0.87, "learning_rate": 9.199390875086134e-07, "loss": 2.017, "step": 35155 }, { "epoch": 0.87, "learning_rate": 9.196046500580069e-07, "loss": 2.0604, "step": 35156 }, { "epoch": 0.87, "learning_rate": 9.192702704797174e-07, "loss": 1.8524, "step": 35157 }, { "epoch": 0.87, "learning_rate": 9.189359487758776e-07, "loss": 1.9065, "step": 35158 }, { "epoch": 0.87, "learning_rate": 9.186016849486201e-07, "loss": 1.9525, "step": 35159 }, { "epoch": 0.87, "learning_rate": 9.182674790000712e-07, "loss": 2.0437, "step": 35160 }, { "epoch": 0.87, "learning_rate": 9.179333309323657e-07, "loss": 2.0402, "step": 35161 }, { "epoch": 0.87, "learning_rate": 9.175992407476297e-07, "loss": 1.8648, "step": 35162 }, { "epoch": 0.87, "learning_rate": 9.172652084479927e-07, "loss": 2.0043, "step": 35163 }, { "epoch": 0.87, "learning_rate": 9.169312340355863e-07, "loss": 1.9459, "step": 35164 }, { "epoch": 0.87, "learning_rate": 9.165973175125353e-07, "loss": 1.9888, "step": 35165 }, { "epoch": 0.87, "learning_rate": 9.162634588809726e-07, "loss": 1.9693, "step": 35166 }, { "epoch": 0.87, "learning_rate": 9.159296581430211e-07, "loss": 1.9199, "step": 35167 }, { "epoch": 0.87, "learning_rate": 9.155959153008099e-07, "loss": 1.9061, "step": 35168 }, { "epoch": 0.87, "learning_rate": 9.152622303564662e-07, "loss": 1.8979, "step": 35169 }, { "epoch": 0.87, "learning_rate": 9.149286033121174e-07, "loss": 1.7887, "step": 35170 }, { "epoch": 0.87, "learning_rate": 9.145950341698905e-07, "loss": 1.9686, "step": 35171 }, { "epoch": 0.87, "learning_rate": 9.142615229319096e-07, "loss": 1.9876, "step": 35172 }, { "epoch": 0.87, "learning_rate": 9.139280696003005e-07, "loss": 2.0652, "step": 35173 }, { "epoch": 0.87, "learning_rate": 9.135946741771906e-07, "loss": 1.8742, "step": 35174 }, { "epoch": 0.87, "learning_rate": 9.132613366647014e-07, "loss": 1.9255, "step": 35175 }, { "epoch": 0.87, "learning_rate": 9.129280570649601e-07, "loss": 1.8853, "step": 35176 }, { "epoch": 0.87, "learning_rate": 9.125948353800884e-07, "loss": 1.8999, "step": 35177 }, { "epoch": 0.87, "learning_rate": 9.122616716122113e-07, "loss": 1.8597, "step": 35178 }, { "epoch": 0.87, "learning_rate": 9.119285657634535e-07, "loss": 2.0984, "step": 35179 }, { "epoch": 0.87, "learning_rate": 9.115955178359359e-07, "loss": 1.879, "step": 35180 }, { "epoch": 0.87, "learning_rate": 9.112625278317833e-07, "loss": 2.0153, "step": 35181 }, { "epoch": 0.87, "learning_rate": 9.109295957531139e-07, "loss": 1.8756, "step": 35182 }, { "epoch": 0.87, "learning_rate": 9.105967216020528e-07, "loss": 1.8202, "step": 35183 }, { "epoch": 0.87, "learning_rate": 9.102639053807227e-07, "loss": 1.7753, "step": 35184 }, { "epoch": 0.87, "learning_rate": 9.099311470912419e-07, "loss": 1.9766, "step": 35185 }, { "epoch": 0.87, "learning_rate": 9.09598446735731e-07, "loss": 2.008, "step": 35186 }, { "epoch": 0.87, "learning_rate": 9.092658043163128e-07, "loss": 2.0531, "step": 35187 }, { "epoch": 0.87, "learning_rate": 9.089332198351052e-07, "loss": 2.1187, "step": 35188 }, { "epoch": 0.87, "learning_rate": 9.086006932942316e-07, "loss": 1.9332, "step": 35189 }, { "epoch": 0.87, "learning_rate": 9.082682246958063e-07, "loss": 2.0502, "step": 35190 }, { "epoch": 0.87, "learning_rate": 9.079358140419525e-07, "loss": 2.0037, "step": 35191 }, { "epoch": 0.87, "learning_rate": 9.076034613347851e-07, "loss": 1.9212, "step": 35192 }, { "epoch": 0.87, "learning_rate": 9.072711665764244e-07, "loss": 1.9294, "step": 35193 }, { "epoch": 0.87, "learning_rate": 9.06938929768989e-07, "loss": 1.671, "step": 35194 }, { "epoch": 0.87, "learning_rate": 9.066067509145926e-07, "loss": 1.9184, "step": 35195 }, { "epoch": 0.87, "learning_rate": 9.06274630015358e-07, "loss": 1.9338, "step": 35196 }, { "epoch": 0.87, "learning_rate": 9.059425670733956e-07, "loss": 1.9853, "step": 35197 }, { "epoch": 0.87, "learning_rate": 9.056105620908262e-07, "loss": 1.9006, "step": 35198 }, { "epoch": 0.87, "learning_rate": 9.052786150697656e-07, "loss": 2.1013, "step": 35199 }, { "epoch": 0.87, "learning_rate": 9.049467260123268e-07, "loss": 1.924, "step": 35200 }, { "epoch": 0.87, "learning_rate": 9.04614894920628e-07, "loss": 1.8578, "step": 35201 }, { "epoch": 0.87, "learning_rate": 9.042831217967806e-07, "loss": 2.0443, "step": 35202 }, { "epoch": 0.87, "learning_rate": 9.039514066429011e-07, "loss": 1.9022, "step": 35203 }, { "epoch": 0.87, "learning_rate": 9.036197494611032e-07, "loss": 1.9595, "step": 35204 }, { "epoch": 0.87, "learning_rate": 9.032881502535018e-07, "loss": 1.9498, "step": 35205 }, { "epoch": 0.87, "learning_rate": 9.02956609022212e-07, "loss": 1.8828, "step": 35206 }, { "epoch": 0.87, "learning_rate": 9.026251257693408e-07, "loss": 1.9822, "step": 35207 }, { "epoch": 0.87, "learning_rate": 9.022937004970056e-07, "loss": 1.7555, "step": 35208 }, { "epoch": 0.87, "learning_rate": 9.019623332073191e-07, "loss": 1.9945, "step": 35209 }, { "epoch": 0.87, "learning_rate": 9.016310239023896e-07, "loss": 1.9497, "step": 35210 }, { "epoch": 0.87, "learning_rate": 9.012997725843308e-07, "loss": 1.8462, "step": 35211 }, { "epoch": 0.87, "learning_rate": 9.009685792552558e-07, "loss": 1.9546, "step": 35212 }, { "epoch": 0.87, "learning_rate": 9.006374439172716e-07, "loss": 1.9651, "step": 35213 }, { "epoch": 0.87, "learning_rate": 9.003063665724921e-07, "loss": 1.936, "step": 35214 }, { "epoch": 0.87, "learning_rate": 8.999753472230232e-07, "loss": 1.9085, "step": 35215 }, { "epoch": 0.87, "learning_rate": 8.99644385870978e-07, "loss": 1.9584, "step": 35216 }, { "epoch": 0.87, "learning_rate": 8.993134825184657e-07, "loss": 1.9607, "step": 35217 }, { "epoch": 0.87, "learning_rate": 8.989826371675936e-07, "loss": 2.0386, "step": 35218 }, { "epoch": 0.87, "learning_rate": 8.986518498204721e-07, "loss": 1.9464, "step": 35219 }, { "epoch": 0.87, "learning_rate": 8.983211204792063e-07, "loss": 1.9268, "step": 35220 }, { "epoch": 0.87, "learning_rate": 8.979904491459068e-07, "loss": 1.918, "step": 35221 }, { "epoch": 0.87, "learning_rate": 8.976598358226806e-07, "loss": 1.7649, "step": 35222 }, { "epoch": 0.87, "learning_rate": 8.97329280511634e-07, "loss": 1.9056, "step": 35223 }, { "epoch": 0.87, "learning_rate": 8.969987832148764e-07, "loss": 1.9351, "step": 35224 }, { "epoch": 0.87, "learning_rate": 8.966683439345103e-07, "loss": 1.8035, "step": 35225 }, { "epoch": 0.87, "learning_rate": 8.963379626726443e-07, "loss": 1.7885, "step": 35226 }, { "epoch": 0.87, "learning_rate": 8.960076394313844e-07, "loss": 1.908, "step": 35227 }, { "epoch": 0.87, "learning_rate": 8.956773742128322e-07, "loss": 1.8434, "step": 35228 }, { "epoch": 0.87, "learning_rate": 8.953471670190983e-07, "loss": 1.8648, "step": 35229 }, { "epoch": 0.87, "learning_rate": 8.95017017852281e-07, "loss": 1.8362, "step": 35230 }, { "epoch": 0.87, "learning_rate": 8.946869267144887e-07, "loss": 2.0944, "step": 35231 }, { "epoch": 0.87, "learning_rate": 8.943568936078251e-07, "loss": 1.9825, "step": 35232 }, { "epoch": 0.87, "learning_rate": 8.940269185343898e-07, "loss": 1.9866, "step": 35233 }, { "epoch": 0.87, "learning_rate": 8.93697001496292e-07, "loss": 1.8421, "step": 35234 }, { "epoch": 0.87, "learning_rate": 8.93367142495628e-07, "loss": 1.8475, "step": 35235 }, { "epoch": 0.87, "learning_rate": 8.930373415345029e-07, "loss": 1.9185, "step": 35236 }, { "epoch": 0.87, "learning_rate": 8.927075986150202e-07, "loss": 1.9452, "step": 35237 }, { "epoch": 0.87, "learning_rate": 8.923779137392796e-07, "loss": 1.9771, "step": 35238 }, { "epoch": 0.87, "learning_rate": 8.920482869093816e-07, "loss": 1.9032, "step": 35239 }, { "epoch": 0.87, "learning_rate": 8.917187181274279e-07, "loss": 1.8564, "step": 35240 }, { "epoch": 0.87, "learning_rate": 8.913892073955199e-07, "loss": 1.9613, "step": 35241 }, { "epoch": 0.87, "learning_rate": 8.910597547157585e-07, "loss": 1.9602, "step": 35242 }, { "epoch": 0.87, "learning_rate": 8.907303600902395e-07, "loss": 2.0905, "step": 35243 }, { "epoch": 0.87, "learning_rate": 8.904010235210659e-07, "loss": 1.9024, "step": 35244 }, { "epoch": 0.87, "learning_rate": 8.900717450103347e-07, "loss": 1.9703, "step": 35245 }, { "epoch": 0.87, "learning_rate": 8.897425245601443e-07, "loss": 1.9333, "step": 35246 }, { "epoch": 0.87, "learning_rate": 8.894133621725964e-07, "loss": 1.8783, "step": 35247 }, { "epoch": 0.87, "learning_rate": 8.890842578497838e-07, "loss": 1.8786, "step": 35248 }, { "epoch": 0.87, "learning_rate": 8.88755211593808e-07, "loss": 2.0202, "step": 35249 }, { "epoch": 0.87, "learning_rate": 8.884262234067631e-07, "loss": 1.8085, "step": 35250 }, { "epoch": 0.87, "learning_rate": 8.880972932907473e-07, "loss": 2.103, "step": 35251 }, { "epoch": 0.87, "learning_rate": 8.877684212478576e-07, "loss": 1.947, "step": 35252 }, { "epoch": 0.87, "learning_rate": 8.874396072801894e-07, "loss": 2.1094, "step": 35253 }, { "epoch": 0.87, "learning_rate": 8.871108513898374e-07, "loss": 2.0591, "step": 35254 }, { "epoch": 0.87, "learning_rate": 8.867821535788989e-07, "loss": 1.9133, "step": 35255 }, { "epoch": 0.87, "learning_rate": 8.864535138494667e-07, "loss": 1.9527, "step": 35256 }, { "epoch": 0.87, "learning_rate": 8.861249322036358e-07, "loss": 1.8754, "step": 35257 }, { "epoch": 0.87, "learning_rate": 8.857964086435011e-07, "loss": 1.9721, "step": 35258 }, { "epoch": 0.87, "learning_rate": 8.854679431711577e-07, "loss": 2.1306, "step": 35259 }, { "epoch": 0.87, "learning_rate": 8.85139535788696e-07, "loss": 1.9875, "step": 35260 }, { "epoch": 0.87, "learning_rate": 8.848111864982101e-07, "loss": 1.8701, "step": 35261 }, { "epoch": 0.87, "learning_rate": 8.844828953017947e-07, "loss": 1.8996, "step": 35262 }, { "epoch": 0.87, "learning_rate": 8.841546622015395e-07, "loss": 2.0389, "step": 35263 }, { "epoch": 0.87, "learning_rate": 8.838264871995384e-07, "loss": 1.7345, "step": 35264 }, { "epoch": 0.87, "learning_rate": 8.834983702978805e-07, "loss": 1.9436, "step": 35265 }, { "epoch": 0.87, "learning_rate": 8.831703114986589e-07, "loss": 1.9037, "step": 35266 }, { "epoch": 0.87, "learning_rate": 8.828423108039652e-07, "loss": 1.81, "step": 35267 }, { "epoch": 0.87, "learning_rate": 8.825143682158877e-07, "loss": 1.8563, "step": 35268 }, { "epoch": 0.87, "learning_rate": 8.821864837365168e-07, "loss": 1.8036, "step": 35269 }, { "epoch": 0.87, "learning_rate": 8.818586573679444e-07, "loss": 1.8547, "step": 35270 }, { "epoch": 0.87, "learning_rate": 8.815308891122565e-07, "loss": 1.9069, "step": 35271 }, { "epoch": 0.87, "learning_rate": 8.812031789715459e-07, "loss": 1.9039, "step": 35272 }, { "epoch": 0.87, "learning_rate": 8.808755269478975e-07, "loss": 1.9827, "step": 35273 }, { "epoch": 0.87, "learning_rate": 8.805479330434008e-07, "loss": 1.8241, "step": 35274 }, { "epoch": 0.87, "learning_rate": 8.802203972601442e-07, "loss": 1.8274, "step": 35275 }, { "epoch": 0.87, "learning_rate": 8.798929196002148e-07, "loss": 1.9487, "step": 35276 }, { "epoch": 0.87, "learning_rate": 8.795655000657021e-07, "loss": 2.0211, "step": 35277 }, { "epoch": 0.87, "learning_rate": 8.79238138658689e-07, "loss": 1.9217, "step": 35278 }, { "epoch": 0.87, "learning_rate": 8.789108353812637e-07, "loss": 1.9433, "step": 35279 }, { "epoch": 0.87, "learning_rate": 8.785835902355145e-07, "loss": 1.8454, "step": 35280 }, { "epoch": 0.87, "learning_rate": 8.782564032235219e-07, "loss": 1.9077, "step": 35281 }, { "epoch": 0.87, "learning_rate": 8.779292743473755e-07, "loss": 2.0121, "step": 35282 }, { "epoch": 0.87, "learning_rate": 8.776022036091581e-07, "loss": 1.9406, "step": 35283 }, { "epoch": 0.87, "learning_rate": 8.772751910109534e-07, "loss": 1.913, "step": 35284 }, { "epoch": 0.87, "learning_rate": 8.769482365548499e-07, "loss": 1.9669, "step": 35285 }, { "epoch": 0.87, "learning_rate": 8.766213402429258e-07, "loss": 2.009, "step": 35286 }, { "epoch": 0.87, "learning_rate": 8.762945020772696e-07, "loss": 1.7326, "step": 35287 }, { "epoch": 0.87, "learning_rate": 8.759677220599594e-07, "loss": 1.774, "step": 35288 }, { "epoch": 0.87, "learning_rate": 8.756410001930804e-07, "loss": 1.9181, "step": 35289 }, { "epoch": 0.87, "learning_rate": 8.753143364787165e-07, "loss": 1.849, "step": 35290 }, { "epoch": 0.87, "learning_rate": 8.749877309189469e-07, "loss": 1.9523, "step": 35291 }, { "epoch": 0.87, "learning_rate": 8.746611835158536e-07, "loss": 2.0477, "step": 35292 }, { "epoch": 0.87, "learning_rate": 8.743346942715192e-07, "loss": 1.9663, "step": 35293 }, { "epoch": 0.87, "learning_rate": 8.740082631880231e-07, "loss": 1.897, "step": 35294 }, { "epoch": 0.87, "learning_rate": 8.736818902674482e-07, "loss": 1.9126, "step": 35295 }, { "epoch": 0.87, "learning_rate": 8.733555755118706e-07, "loss": 1.9125, "step": 35296 }, { "epoch": 0.87, "learning_rate": 8.730293189233752e-07, "loss": 1.832, "step": 35297 }, { "epoch": 0.87, "learning_rate": 8.72703120504036e-07, "loss": 1.839, "step": 35298 }, { "epoch": 0.87, "learning_rate": 8.723769802559335e-07, "loss": 1.9381, "step": 35299 }, { "epoch": 0.87, "learning_rate": 8.720508981811504e-07, "loss": 1.9675, "step": 35300 }, { "epoch": 0.87, "learning_rate": 8.717248742817586e-07, "loss": 1.8018, "step": 35301 }, { "epoch": 0.87, "learning_rate": 8.713989085598418e-07, "loss": 2.0766, "step": 35302 }, { "epoch": 0.87, "learning_rate": 8.710730010174718e-07, "loss": 1.9466, "step": 35303 }, { "epoch": 0.87, "learning_rate": 8.707471516567289e-07, "loss": 1.8818, "step": 35304 }, { "epoch": 0.87, "learning_rate": 8.704213604796918e-07, "loss": 1.8852, "step": 35305 }, { "epoch": 0.87, "learning_rate": 8.700956274884331e-07, "loss": 1.8302, "step": 35306 }, { "epoch": 0.87, "learning_rate": 8.69769952685029e-07, "loss": 1.9928, "step": 35307 }, { "epoch": 0.87, "learning_rate": 8.694443360715599e-07, "loss": 1.9244, "step": 35308 }, { "epoch": 0.87, "learning_rate": 8.691187776500942e-07, "loss": 1.9694, "step": 35309 }, { "epoch": 0.87, "learning_rate": 8.687932774227115e-07, "loss": 1.8838, "step": 35310 }, { "epoch": 0.87, "learning_rate": 8.684678353914844e-07, "loss": 1.9498, "step": 35311 }, { "epoch": 0.87, "learning_rate": 8.681424515584891e-07, "loss": 2.1764, "step": 35312 }, { "epoch": 0.87, "learning_rate": 8.678171259257972e-07, "loss": 1.8581, "step": 35313 }, { "epoch": 0.87, "learning_rate": 8.674918584954817e-07, "loss": 1.8226, "step": 35314 }, { "epoch": 0.87, "learning_rate": 8.671666492696196e-07, "loss": 1.934, "step": 35315 }, { "epoch": 0.87, "learning_rate": 8.668414982502782e-07, "loss": 1.7885, "step": 35316 }, { "epoch": 0.87, "learning_rate": 8.665164054395347e-07, "loss": 1.8533, "step": 35317 }, { "epoch": 0.87, "learning_rate": 8.661913708394564e-07, "loss": 1.8989, "step": 35318 }, { "epoch": 0.87, "learning_rate": 8.658663944521173e-07, "loss": 2.0686, "step": 35319 }, { "epoch": 0.87, "learning_rate": 8.655414762795899e-07, "loss": 1.8334, "step": 35320 }, { "epoch": 0.87, "learning_rate": 8.652166163239417e-07, "loss": 1.9281, "step": 35321 }, { "epoch": 0.87, "learning_rate": 8.648918145872475e-07, "loss": 1.9367, "step": 35322 }, { "epoch": 0.87, "learning_rate": 8.645670710715725e-07, "loss": 1.856, "step": 35323 }, { "epoch": 0.87, "learning_rate": 8.642423857789883e-07, "loss": 1.8486, "step": 35324 }, { "epoch": 0.87, "learning_rate": 8.639177587115654e-07, "loss": 1.8177, "step": 35325 }, { "epoch": 0.87, "learning_rate": 8.635931898713734e-07, "loss": 1.8006, "step": 35326 }, { "epoch": 0.87, "learning_rate": 8.632686792604784e-07, "loss": 2.0147, "step": 35327 }, { "epoch": 0.87, "learning_rate": 8.629442268809485e-07, "loss": 1.9811, "step": 35328 }, { "epoch": 0.87, "learning_rate": 8.626198327348534e-07, "loss": 1.8023, "step": 35329 }, { "epoch": 0.87, "learning_rate": 8.622954968242614e-07, "loss": 1.9011, "step": 35330 }, { "epoch": 0.87, "learning_rate": 8.619712191512364e-07, "loss": 1.8689, "step": 35331 }, { "epoch": 0.87, "learning_rate": 8.616469997178467e-07, "loss": 1.9763, "step": 35332 }, { "epoch": 0.87, "learning_rate": 8.613228385261607e-07, "loss": 1.9379, "step": 35333 }, { "epoch": 0.87, "learning_rate": 8.609987355782412e-07, "loss": 1.9376, "step": 35334 }, { "epoch": 0.87, "learning_rate": 8.606746908761565e-07, "loss": 1.8379, "step": 35335 }, { "epoch": 0.87, "learning_rate": 8.603507044219684e-07, "loss": 1.9033, "step": 35336 }, { "epoch": 0.87, "learning_rate": 8.600267762177439e-07, "loss": 1.8223, "step": 35337 }, { "epoch": 0.87, "learning_rate": 8.597029062655493e-07, "loss": 1.9783, "step": 35338 }, { "epoch": 0.87, "learning_rate": 8.59379094567444e-07, "loss": 1.7895, "step": 35339 }, { "epoch": 0.87, "learning_rate": 8.590553411254976e-07, "loss": 1.928, "step": 35340 }, { "epoch": 0.87, "learning_rate": 8.587316459417682e-07, "loss": 1.7887, "step": 35341 }, { "epoch": 0.87, "learning_rate": 8.584080090183211e-07, "loss": 1.8979, "step": 35342 }, { "epoch": 0.87, "learning_rate": 8.580844303572189e-07, "loss": 2.0033, "step": 35343 }, { "epoch": 0.87, "learning_rate": 8.577609099605266e-07, "loss": 1.9014, "step": 35344 }, { "epoch": 0.87, "learning_rate": 8.574374478303005e-07, "loss": 1.9798, "step": 35345 }, { "epoch": 0.87, "learning_rate": 8.571140439686054e-07, "loss": 1.8238, "step": 35346 }, { "epoch": 0.87, "learning_rate": 8.567906983775021e-07, "loss": 1.9174, "step": 35347 }, { "epoch": 0.87, "learning_rate": 8.564674110590521e-07, "loss": 1.9249, "step": 35348 }, { "epoch": 0.87, "learning_rate": 8.561441820153149e-07, "loss": 1.9973, "step": 35349 }, { "epoch": 0.87, "learning_rate": 8.558210112483511e-07, "loss": 1.9449, "step": 35350 }, { "epoch": 0.87, "learning_rate": 8.554978987602192e-07, "loss": 1.9977, "step": 35351 }, { "epoch": 0.87, "learning_rate": 8.551748445529795e-07, "loss": 2.0971, "step": 35352 }, { "epoch": 0.87, "learning_rate": 8.548518486286917e-07, "loss": 1.8551, "step": 35353 }, { "epoch": 0.87, "learning_rate": 8.545289109894118e-07, "loss": 1.8629, "step": 35354 }, { "epoch": 0.87, "learning_rate": 8.542060316372014e-07, "loss": 1.8991, "step": 35355 }, { "epoch": 0.87, "learning_rate": 8.538832105741157e-07, "loss": 2.002, "step": 35356 }, { "epoch": 0.87, "learning_rate": 8.535604478022119e-07, "loss": 1.8362, "step": 35357 }, { "epoch": 0.87, "learning_rate": 8.532377433235495e-07, "loss": 1.7702, "step": 35358 }, { "epoch": 0.87, "learning_rate": 8.529150971401823e-07, "loss": 1.831, "step": 35359 }, { "epoch": 0.87, "learning_rate": 8.525925092541687e-07, "loss": 1.9198, "step": 35360 }, { "epoch": 0.87, "learning_rate": 8.52269979667566e-07, "loss": 2.0367, "step": 35361 }, { "epoch": 0.87, "learning_rate": 8.519475083824258e-07, "loss": 2.0137, "step": 35362 }, { "epoch": 0.87, "learning_rate": 8.516250954008054e-07, "loss": 1.8238, "step": 35363 }, { "epoch": 0.87, "learning_rate": 8.513027407247588e-07, "loss": 1.8883, "step": 35364 }, { "epoch": 0.87, "learning_rate": 8.509804443563441e-07, "loss": 1.9291, "step": 35365 }, { "epoch": 0.87, "learning_rate": 8.506582062976099e-07, "loss": 1.9208, "step": 35366 }, { "epoch": 0.87, "learning_rate": 8.503360265506133e-07, "loss": 1.879, "step": 35367 }, { "epoch": 0.87, "learning_rate": 8.500139051174094e-07, "loss": 2.0187, "step": 35368 }, { "epoch": 0.87, "learning_rate": 8.496918420000466e-07, "loss": 1.8746, "step": 35369 }, { "epoch": 0.87, "learning_rate": 8.493698372005809e-07, "loss": 2.0226, "step": 35370 }, { "epoch": 0.87, "learning_rate": 8.49047890721062e-07, "loss": 1.9988, "step": 35371 }, { "epoch": 0.87, "learning_rate": 8.487260025635425e-07, "loss": 2.0128, "step": 35372 }, { "epoch": 0.87, "learning_rate": 8.484041727300762e-07, "loss": 1.9485, "step": 35373 }, { "epoch": 0.87, "learning_rate": 8.480824012227118e-07, "loss": 1.9497, "step": 35374 }, { "epoch": 0.87, "learning_rate": 8.477606880435019e-07, "loss": 1.961, "step": 35375 }, { "epoch": 0.87, "learning_rate": 8.474390331944948e-07, "loss": 1.986, "step": 35376 }, { "epoch": 0.87, "learning_rate": 8.471174366777401e-07, "loss": 1.9897, "step": 35377 }, { "epoch": 0.87, "learning_rate": 8.467958984952895e-07, "loss": 1.8204, "step": 35378 }, { "epoch": 0.87, "learning_rate": 8.464744186491936e-07, "loss": 1.9123, "step": 35379 }, { "epoch": 0.87, "learning_rate": 8.461529971414973e-07, "loss": 1.9688, "step": 35380 }, { "epoch": 0.87, "learning_rate": 8.458316339742511e-07, "loss": 1.8889, "step": 35381 }, { "epoch": 0.87, "learning_rate": 8.455103291495026e-07, "loss": 1.8438, "step": 35382 }, { "epoch": 0.87, "learning_rate": 8.451890826693032e-07, "loss": 1.9193, "step": 35383 }, { "epoch": 0.87, "learning_rate": 8.448678945356947e-07, "loss": 1.982, "step": 35384 }, { "epoch": 0.87, "learning_rate": 8.445467647507266e-07, "loss": 1.9754, "step": 35385 }, { "epoch": 0.87, "learning_rate": 8.442256933164472e-07, "loss": 1.8931, "step": 35386 }, { "epoch": 0.87, "learning_rate": 8.439046802348994e-07, "loss": 1.7277, "step": 35387 }, { "epoch": 0.87, "learning_rate": 8.435837255081336e-07, "loss": 2.1227, "step": 35388 }, { "epoch": 0.87, "learning_rate": 8.432628291381906e-07, "loss": 1.9812, "step": 35389 }, { "epoch": 0.87, "learning_rate": 8.429419911271175e-07, "loss": 1.8584, "step": 35390 }, { "epoch": 0.87, "learning_rate": 8.426212114769605e-07, "loss": 2.0012, "step": 35391 }, { "epoch": 0.87, "learning_rate": 8.423004901897624e-07, "loss": 1.9154, "step": 35392 }, { "epoch": 0.87, "learning_rate": 8.419798272675683e-07, "loss": 1.9585, "step": 35393 }, { "epoch": 0.87, "learning_rate": 8.416592227124198e-07, "loss": 1.9231, "step": 35394 }, { "epoch": 0.87, "learning_rate": 8.41338676526362e-07, "loss": 1.9503, "step": 35395 }, { "epoch": 0.87, "learning_rate": 8.410181887114366e-07, "loss": 1.8456, "step": 35396 }, { "epoch": 0.87, "learning_rate": 8.406977592696897e-07, "loss": 1.7965, "step": 35397 }, { "epoch": 0.87, "learning_rate": 8.403773882031596e-07, "loss": 1.9182, "step": 35398 }, { "epoch": 0.87, "learning_rate": 8.400570755138892e-07, "loss": 1.9661, "step": 35399 }, { "epoch": 0.87, "learning_rate": 8.397368212039192e-07, "loss": 1.9775, "step": 35400 }, { "epoch": 0.87, "learning_rate": 8.394166252752944e-07, "loss": 1.9864, "step": 35401 }, { "epoch": 0.87, "learning_rate": 8.390964877300511e-07, "loss": 2.1162, "step": 35402 }, { "epoch": 0.87, "learning_rate": 8.387764085702321e-07, "loss": 1.8796, "step": 35403 }, { "epoch": 0.87, "learning_rate": 8.384563877978758e-07, "loss": 1.9263, "step": 35404 }, { "epoch": 0.87, "learning_rate": 8.381364254150226e-07, "loss": 1.8917, "step": 35405 }, { "epoch": 0.87, "learning_rate": 8.378165214237133e-07, "loss": 1.9592, "step": 35406 }, { "epoch": 0.87, "learning_rate": 8.374966758259828e-07, "loss": 1.9983, "step": 35407 }, { "epoch": 0.87, "learning_rate": 8.37176888623874e-07, "loss": 1.8892, "step": 35408 }, { "epoch": 0.87, "learning_rate": 8.368571598194219e-07, "loss": 1.9599, "step": 35409 }, { "epoch": 0.87, "learning_rate": 8.36537489414665e-07, "loss": 1.9423, "step": 35410 }, { "epoch": 0.87, "learning_rate": 8.362178774116425e-07, "loss": 2.1097, "step": 35411 }, { "epoch": 0.87, "learning_rate": 8.358983238123875e-07, "loss": 1.863, "step": 35412 }, { "epoch": 0.87, "learning_rate": 8.355788286189404e-07, "loss": 2.0522, "step": 35413 }, { "epoch": 0.87, "learning_rate": 8.352593918333352e-07, "loss": 1.9955, "step": 35414 }, { "epoch": 0.87, "learning_rate": 8.349400134576102e-07, "loss": 1.9015, "step": 35415 }, { "epoch": 0.87, "learning_rate": 8.346206934937973e-07, "loss": 2.1222, "step": 35416 }, { "epoch": 0.87, "learning_rate": 8.343014319439347e-07, "loss": 2.0126, "step": 35417 }, { "epoch": 0.87, "learning_rate": 8.339822288100574e-07, "loss": 1.9436, "step": 35418 }, { "epoch": 0.87, "learning_rate": 8.336630840941962e-07, "loss": 1.8767, "step": 35419 }, { "epoch": 0.87, "learning_rate": 8.333439977983871e-07, "loss": 1.8936, "step": 35420 }, { "epoch": 0.87, "learning_rate": 8.330249699246673e-07, "loss": 1.9863, "step": 35421 }, { "epoch": 0.87, "learning_rate": 8.327060004750642e-07, "loss": 1.9091, "step": 35422 }, { "epoch": 0.87, "learning_rate": 8.32387089451615e-07, "loss": 1.9377, "step": 35423 }, { "epoch": 0.87, "learning_rate": 8.320682368563482e-07, "loss": 1.8773, "step": 35424 }, { "epoch": 0.87, "learning_rate": 8.317494426912986e-07, "loss": 1.9306, "step": 35425 }, { "epoch": 0.87, "learning_rate": 8.314307069585004e-07, "loss": 2.0019, "step": 35426 }, { "epoch": 0.87, "learning_rate": 8.311120296599795e-07, "loss": 1.9597, "step": 35427 }, { "epoch": 0.87, "learning_rate": 8.307934107977711e-07, "loss": 2.062, "step": 35428 }, { "epoch": 0.87, "learning_rate": 8.304748503739035e-07, "loss": 1.9381, "step": 35429 }, { "epoch": 0.87, "learning_rate": 8.301563483904084e-07, "loss": 1.9691, "step": 35430 }, { "epoch": 0.87, "learning_rate": 8.298379048493144e-07, "loss": 1.8864, "step": 35431 }, { "epoch": 0.87, "learning_rate": 8.295195197526518e-07, "loss": 2.0024, "step": 35432 }, { "epoch": 0.87, "learning_rate": 8.292011931024524e-07, "loss": 2.092, "step": 35433 }, { "epoch": 0.87, "learning_rate": 8.288829249007401e-07, "loss": 1.8323, "step": 35434 }, { "epoch": 0.87, "learning_rate": 8.285647151495457e-07, "loss": 1.8736, "step": 35435 }, { "epoch": 0.87, "learning_rate": 8.282465638508985e-07, "loss": 1.8966, "step": 35436 }, { "epoch": 0.87, "learning_rate": 8.279284710068247e-07, "loss": 1.8046, "step": 35437 }, { "epoch": 0.87, "learning_rate": 8.276104366193527e-07, "loss": 1.9986, "step": 35438 }, { "epoch": 0.87, "learning_rate": 8.272924606905064e-07, "loss": 2.036, "step": 35439 }, { "epoch": 0.87, "learning_rate": 8.269745432223153e-07, "loss": 1.8356, "step": 35440 }, { "epoch": 0.87, "learning_rate": 8.266566842168067e-07, "loss": 1.8505, "step": 35441 }, { "epoch": 0.87, "learning_rate": 8.263388836760022e-07, "loss": 1.8632, "step": 35442 }, { "epoch": 0.87, "learning_rate": 8.260211416019315e-07, "loss": 1.913, "step": 35443 }, { "epoch": 0.87, "learning_rate": 8.257034579966161e-07, "loss": 1.8793, "step": 35444 }, { "epoch": 0.87, "learning_rate": 8.253858328620823e-07, "loss": 1.8607, "step": 35445 }, { "epoch": 0.87, "learning_rate": 8.250682662003561e-07, "loss": 1.7922, "step": 35446 }, { "epoch": 0.87, "learning_rate": 8.247507580134584e-07, "loss": 1.9384, "step": 35447 }, { "epoch": 0.87, "learning_rate": 8.24433308303414e-07, "loss": 1.7981, "step": 35448 }, { "epoch": 0.87, "learning_rate": 8.241159170722468e-07, "loss": 2.021, "step": 35449 }, { "epoch": 0.87, "learning_rate": 8.237985843219809e-07, "loss": 1.9204, "step": 35450 }, { "epoch": 0.87, "learning_rate": 8.234813100546346e-07, "loss": 1.8252, "step": 35451 }, { "epoch": 0.87, "learning_rate": 8.23164094272233e-07, "loss": 1.9599, "step": 35452 }, { "epoch": 0.87, "learning_rate": 8.228469369767977e-07, "loss": 1.9432, "step": 35453 }, { "epoch": 0.87, "learning_rate": 8.225298381703506e-07, "loss": 1.9967, "step": 35454 }, { "epoch": 0.87, "learning_rate": 8.222127978549099e-07, "loss": 1.8398, "step": 35455 }, { "epoch": 0.87, "learning_rate": 8.218958160325007e-07, "loss": 1.9516, "step": 35456 }, { "epoch": 0.87, "learning_rate": 8.215788927051383e-07, "loss": 1.9719, "step": 35457 }, { "epoch": 0.87, "learning_rate": 8.21262027874844e-07, "loss": 1.8793, "step": 35458 }, { "epoch": 0.87, "learning_rate": 8.209452215436409e-07, "loss": 1.6659, "step": 35459 }, { "epoch": 0.87, "learning_rate": 8.206284737135429e-07, "loss": 1.7535, "step": 35460 }, { "epoch": 0.87, "learning_rate": 8.203117843865727e-07, "loss": 1.851, "step": 35461 }, { "epoch": 0.87, "learning_rate": 8.199951535647454e-07, "loss": 1.9712, "step": 35462 }, { "epoch": 0.87, "learning_rate": 8.196785812500818e-07, "loss": 1.8115, "step": 35463 }, { "epoch": 0.87, "learning_rate": 8.19362067444599e-07, "loss": 1.955, "step": 35464 }, { "epoch": 0.87, "learning_rate": 8.190456121503121e-07, "loss": 1.8968, "step": 35465 }, { "epoch": 0.87, "learning_rate": 8.187292153692394e-07, "loss": 1.9091, "step": 35466 }, { "epoch": 0.87, "learning_rate": 8.184128771033972e-07, "loss": 1.9407, "step": 35467 }, { "epoch": 0.87, "learning_rate": 8.180965973548049e-07, "loss": 1.9416, "step": 35468 }, { "epoch": 0.87, "learning_rate": 8.177803761254722e-07, "loss": 1.9256, "step": 35469 }, { "epoch": 0.87, "learning_rate": 8.174642134174183e-07, "loss": 1.8667, "step": 35470 }, { "epoch": 0.87, "learning_rate": 8.171481092326594e-07, "loss": 2.0233, "step": 35471 }, { "epoch": 0.87, "learning_rate": 8.168320635732052e-07, "loss": 1.9043, "step": 35472 }, { "epoch": 0.87, "learning_rate": 8.16516076441074e-07, "loss": 2.0447, "step": 35473 }, { "epoch": 0.87, "learning_rate": 8.162001478382808e-07, "loss": 1.8941, "step": 35474 }, { "epoch": 0.87, "learning_rate": 8.158842777668341e-07, "loss": 1.9774, "step": 35475 }, { "epoch": 0.87, "learning_rate": 8.155684662287522e-07, "loss": 1.7683, "step": 35476 }, { "epoch": 0.87, "learning_rate": 8.152527132260446e-07, "loss": 1.8719, "step": 35477 }, { "epoch": 0.87, "learning_rate": 8.149370187607231e-07, "loss": 1.8569, "step": 35478 }, { "epoch": 0.87, "learning_rate": 8.146213828348038e-07, "loss": 1.9479, "step": 35479 }, { "epoch": 0.87, "learning_rate": 8.143058054502951e-07, "loss": 2.0604, "step": 35480 }, { "epoch": 0.87, "learning_rate": 8.139902866092098e-07, "loss": 1.8627, "step": 35481 }, { "epoch": 0.87, "learning_rate": 8.136748263135563e-07, "loss": 1.887, "step": 35482 }, { "epoch": 0.87, "learning_rate": 8.133594245653464e-07, "loss": 1.9259, "step": 35483 }, { "epoch": 0.87, "learning_rate": 8.130440813665918e-07, "loss": 1.8374, "step": 35484 }, { "epoch": 0.87, "learning_rate": 8.127287967192999e-07, "loss": 1.9212, "step": 35485 }, { "epoch": 0.87, "learning_rate": 8.124135706254843e-07, "loss": 1.8862, "step": 35486 }, { "epoch": 0.87, "learning_rate": 8.120984030871493e-07, "loss": 1.8933, "step": 35487 }, { "epoch": 0.87, "learning_rate": 8.117832941063052e-07, "loss": 1.9111, "step": 35488 }, { "epoch": 0.87, "learning_rate": 8.11468243684962e-07, "loss": 1.9604, "step": 35489 }, { "epoch": 0.87, "learning_rate": 8.111532518251253e-07, "loss": 1.9926, "step": 35490 }, { "epoch": 0.87, "learning_rate": 8.108383185288037e-07, "loss": 1.8388, "step": 35491 }, { "epoch": 0.87, "learning_rate": 8.105234437980036e-07, "loss": 1.945, "step": 35492 }, { "epoch": 0.87, "learning_rate": 8.10208627634732e-07, "loss": 1.9834, "step": 35493 }, { "epoch": 0.87, "learning_rate": 8.098938700409975e-07, "loss": 1.8945, "step": 35494 }, { "epoch": 0.87, "learning_rate": 8.095791710188027e-07, "loss": 1.8287, "step": 35495 }, { "epoch": 0.87, "learning_rate": 8.092645305701563e-07, "loss": 2.0104, "step": 35496 }, { "epoch": 0.87, "learning_rate": 8.089499486970609e-07, "loss": 1.9255, "step": 35497 }, { "epoch": 0.87, "learning_rate": 8.086354254015228e-07, "loss": 1.8954, "step": 35498 }, { "epoch": 0.88, "learning_rate": 8.08320960685548e-07, "loss": 1.8574, "step": 35499 }, { "epoch": 0.88, "learning_rate": 8.080065545511373e-07, "loss": 1.8365, "step": 35500 }, { "epoch": 0.88, "learning_rate": 8.076922070002968e-07, "loss": 2.0621, "step": 35501 }, { "epoch": 0.88, "learning_rate": 8.073779180350294e-07, "loss": 2.0041, "step": 35502 }, { "epoch": 0.88, "learning_rate": 8.070636876573379e-07, "loss": 1.9971, "step": 35503 }, { "epoch": 0.88, "learning_rate": 8.067495158692273e-07, "loss": 1.9769, "step": 35504 }, { "epoch": 0.88, "learning_rate": 8.064354026726973e-07, "loss": 1.865, "step": 35505 }, { "epoch": 0.88, "learning_rate": 8.061213480697494e-07, "loss": 1.8228, "step": 35506 }, { "epoch": 0.88, "learning_rate": 8.058073520623888e-07, "loss": 2.0119, "step": 35507 }, { "epoch": 0.88, "learning_rate": 8.054934146526117e-07, "loss": 1.9064, "step": 35508 }, { "epoch": 0.88, "learning_rate": 8.051795358424242e-07, "loss": 1.8707, "step": 35509 }, { "epoch": 0.88, "learning_rate": 8.048657156338214e-07, "loss": 1.8397, "step": 35510 }, { "epoch": 0.88, "learning_rate": 8.04551954028806e-07, "loss": 1.9881, "step": 35511 }, { "epoch": 0.88, "learning_rate": 8.0423825102938e-07, "loss": 1.9881, "step": 35512 }, { "epoch": 0.88, "learning_rate": 8.039246066375383e-07, "loss": 1.9967, "step": 35513 }, { "epoch": 0.88, "learning_rate": 8.036110208552839e-07, "loss": 2.0542, "step": 35514 }, { "epoch": 0.88, "learning_rate": 8.032974936846116e-07, "loss": 1.8347, "step": 35515 }, { "epoch": 0.88, "learning_rate": 8.029840251275222e-07, "loss": 1.9414, "step": 35516 }, { "epoch": 0.88, "learning_rate": 8.026706151860142e-07, "loss": 2.0502, "step": 35517 }, { "epoch": 0.88, "learning_rate": 8.023572638620814e-07, "loss": 2.0156, "step": 35518 }, { "epoch": 0.88, "learning_rate": 8.020439711577244e-07, "loss": 2.0233, "step": 35519 }, { "epoch": 0.88, "learning_rate": 8.017307370749372e-07, "loss": 2.1195, "step": 35520 }, { "epoch": 0.88, "learning_rate": 8.014175616157182e-07, "loss": 2.0598, "step": 35521 }, { "epoch": 0.88, "learning_rate": 8.011044447820649e-07, "loss": 2.0245, "step": 35522 }, { "epoch": 0.88, "learning_rate": 8.007913865759687e-07, "loss": 2.1119, "step": 35523 }, { "epoch": 0.88, "learning_rate": 8.004783869994293e-07, "loss": 1.9, "step": 35524 }, { "epoch": 0.88, "learning_rate": 8.001654460544362e-07, "loss": 2.0347, "step": 35525 }, { "epoch": 0.88, "learning_rate": 7.998525637429877e-07, "loss": 1.9372, "step": 35526 }, { "epoch": 0.88, "learning_rate": 7.995397400670791e-07, "loss": 1.8042, "step": 35527 }, { "epoch": 0.88, "learning_rate": 7.992269750286996e-07, "loss": 1.7933, "step": 35528 }, { "epoch": 0.88, "learning_rate": 7.989142686298478e-07, "loss": 1.8919, "step": 35529 }, { "epoch": 0.88, "learning_rate": 7.986016208725112e-07, "loss": 1.906, "step": 35530 }, { "epoch": 0.88, "learning_rate": 7.982890317586856e-07, "loss": 1.937, "step": 35531 }, { "epoch": 0.88, "learning_rate": 7.97976501290364e-07, "loss": 2.0061, "step": 35532 }, { "epoch": 0.88, "learning_rate": 7.976640294695359e-07, "loss": 1.9381, "step": 35533 }, { "epoch": 0.88, "learning_rate": 7.973516162981953e-07, "loss": 1.9602, "step": 35534 }, { "epoch": 0.88, "learning_rate": 7.970392617783295e-07, "loss": 1.848, "step": 35535 }, { "epoch": 0.88, "learning_rate": 7.967269659119325e-07, "loss": 1.9241, "step": 35536 }, { "epoch": 0.88, "learning_rate": 7.964147287009937e-07, "loss": 1.7708, "step": 35537 }, { "epoch": 0.88, "learning_rate": 7.961025501475028e-07, "loss": 1.8873, "step": 35538 }, { "epoch": 0.88, "learning_rate": 7.957904302534525e-07, "loss": 1.9479, "step": 35539 }, { "epoch": 0.88, "learning_rate": 7.954783690208268e-07, "loss": 1.8521, "step": 35540 }, { "epoch": 0.88, "learning_rate": 7.951663664516174e-07, "loss": 2.0964, "step": 35541 }, { "epoch": 0.88, "learning_rate": 7.948544225478138e-07, "loss": 1.7636, "step": 35542 }, { "epoch": 0.88, "learning_rate": 7.945425373114013e-07, "loss": 1.9315, "step": 35543 }, { "epoch": 0.88, "learning_rate": 7.942307107443715e-07, "loss": 2.014, "step": 35544 }, { "epoch": 0.88, "learning_rate": 7.93918942848707e-07, "loss": 2.0288, "step": 35545 }, { "epoch": 0.88, "learning_rate": 7.936072336263978e-07, "loss": 1.98, "step": 35546 }, { "epoch": 0.88, "learning_rate": 7.932955830794309e-07, "loss": 2.0115, "step": 35547 }, { "epoch": 0.88, "learning_rate": 7.929839912097914e-07, "loss": 1.9754, "step": 35548 }, { "epoch": 0.88, "learning_rate": 7.926724580194655e-07, "loss": 1.8388, "step": 35549 }, { "epoch": 0.88, "learning_rate": 7.923609835104374e-07, "loss": 1.984, "step": 35550 }, { "epoch": 0.88, "learning_rate": 7.92049567684694e-07, "loss": 1.9816, "step": 35551 }, { "epoch": 0.88, "learning_rate": 7.917382105442207e-07, "loss": 1.9217, "step": 35552 }, { "epoch": 0.88, "learning_rate": 7.914269120909989e-07, "loss": 1.9341, "step": 35553 }, { "epoch": 0.88, "learning_rate": 7.911156723270152e-07, "loss": 1.8731, "step": 35554 }, { "epoch": 0.88, "learning_rate": 7.908044912542512e-07, "loss": 1.916, "step": 35555 }, { "epoch": 0.88, "learning_rate": 7.904933688746918e-07, "loss": 1.7249, "step": 35556 }, { "epoch": 0.88, "learning_rate": 7.901823051903213e-07, "loss": 2.0602, "step": 35557 }, { "epoch": 0.88, "learning_rate": 7.898713002031188e-07, "loss": 1.9788, "step": 35558 }, { "epoch": 0.88, "learning_rate": 7.895603539150698e-07, "loss": 1.9165, "step": 35559 }, { "epoch": 0.88, "learning_rate": 7.892494663281525e-07, "loss": 1.9035, "step": 35560 }, { "epoch": 0.88, "learning_rate": 7.889386374443497e-07, "loss": 1.8857, "step": 35561 }, { "epoch": 0.88, "learning_rate": 7.886278672656445e-07, "loss": 1.8746, "step": 35562 }, { "epoch": 0.88, "learning_rate": 7.883171557940139e-07, "loss": 1.8821, "step": 35563 }, { "epoch": 0.88, "learning_rate": 7.88006503031441e-07, "loss": 2.0371, "step": 35564 }, { "epoch": 0.88, "learning_rate": 7.876959089799052e-07, "loss": 1.7636, "step": 35565 }, { "epoch": 0.88, "learning_rate": 7.873853736413851e-07, "loss": 1.9628, "step": 35566 }, { "epoch": 0.88, "learning_rate": 7.87074897017861e-07, "loss": 1.7212, "step": 35567 }, { "epoch": 0.88, "learning_rate": 7.867644791113094e-07, "loss": 1.9302, "step": 35568 }, { "epoch": 0.88, "learning_rate": 7.864541199237108e-07, "loss": 1.9892, "step": 35569 }, { "epoch": 0.88, "learning_rate": 7.861438194570448e-07, "loss": 1.9953, "step": 35570 }, { "epoch": 0.88, "learning_rate": 7.858335777132842e-07, "loss": 1.8272, "step": 35571 }, { "epoch": 0.88, "learning_rate": 7.855233946944097e-07, "loss": 1.9552, "step": 35572 }, { "epoch": 0.88, "learning_rate": 7.852132704023974e-07, "loss": 2.0356, "step": 35573 }, { "epoch": 0.88, "learning_rate": 7.849032048392236e-07, "loss": 1.8285, "step": 35574 }, { "epoch": 0.88, "learning_rate": 7.845931980068667e-07, "loss": 2.1342, "step": 35575 }, { "epoch": 0.88, "learning_rate": 7.842832499072983e-07, "loss": 1.9018, "step": 35576 }, { "epoch": 0.88, "learning_rate": 7.839733605424982e-07, "loss": 2.1518, "step": 35577 }, { "epoch": 0.88, "learning_rate": 7.83663529914438e-07, "loss": 1.8664, "step": 35578 }, { "epoch": 0.88, "learning_rate": 7.833537580250938e-07, "loss": 2.0494, "step": 35579 }, { "epoch": 0.88, "learning_rate": 7.830440448764409e-07, "loss": 1.8835, "step": 35580 }, { "epoch": 0.88, "learning_rate": 7.827343904704499e-07, "loss": 1.929, "step": 35581 }, { "epoch": 0.88, "learning_rate": 7.824247948090991e-07, "loss": 1.9949, "step": 35582 }, { "epoch": 0.88, "learning_rate": 7.82115257894357e-07, "loss": 1.8091, "step": 35583 }, { "epoch": 0.88, "learning_rate": 7.818057797281975e-07, "loss": 1.8907, "step": 35584 }, { "epoch": 0.88, "learning_rate": 7.814963603125969e-07, "loss": 1.9346, "step": 35585 }, { "epoch": 0.88, "learning_rate": 7.811869996495214e-07, "loss": 1.8637, "step": 35586 }, { "epoch": 0.88, "learning_rate": 7.808776977409482e-07, "loss": 1.8488, "step": 35587 }, { "epoch": 0.88, "learning_rate": 7.805684545888437e-07, "loss": 1.9853, "step": 35588 }, { "epoch": 0.88, "learning_rate": 7.802592701951806e-07, "loss": 1.8479, "step": 35589 }, { "epoch": 0.88, "learning_rate": 7.799501445619306e-07, "loss": 1.7633, "step": 35590 }, { "epoch": 0.88, "learning_rate": 7.796410776910624e-07, "loss": 1.8548, "step": 35591 }, { "epoch": 0.88, "learning_rate": 7.793320695845485e-07, "loss": 2.0644, "step": 35592 }, { "epoch": 0.88, "learning_rate": 7.790231202443544e-07, "loss": 1.8975, "step": 35593 }, { "epoch": 0.88, "learning_rate": 7.787142296724515e-07, "loss": 2.0445, "step": 35594 }, { "epoch": 0.88, "learning_rate": 7.784053978708083e-07, "loss": 1.9434, "step": 35595 }, { "epoch": 0.88, "learning_rate": 7.780966248413924e-07, "loss": 1.9443, "step": 35596 }, { "epoch": 0.88, "learning_rate": 7.77787910586173e-07, "loss": 2.1111, "step": 35597 }, { "epoch": 0.88, "learning_rate": 7.774792551071153e-07, "loss": 1.8679, "step": 35598 }, { "epoch": 0.88, "learning_rate": 7.771706584061878e-07, "loss": 1.9611, "step": 35599 }, { "epoch": 0.88, "learning_rate": 7.768621204853588e-07, "loss": 1.8434, "step": 35600 }, { "epoch": 0.88, "learning_rate": 7.765536413465912e-07, "loss": 1.892, "step": 35601 }, { "epoch": 0.88, "learning_rate": 7.762452209918559e-07, "loss": 1.8888, "step": 35602 }, { "epoch": 0.88, "learning_rate": 7.759368594231131e-07, "loss": 2.0891, "step": 35603 }, { "epoch": 0.88, "learning_rate": 7.756285566423305e-07, "loss": 1.8268, "step": 35604 }, { "epoch": 0.88, "learning_rate": 7.753203126514763e-07, "loss": 1.899, "step": 35605 }, { "epoch": 0.88, "learning_rate": 7.750121274525091e-07, "loss": 1.8386, "step": 35606 }, { "epoch": 0.88, "learning_rate": 7.74704001047396e-07, "loss": 1.9264, "step": 35607 }, { "epoch": 0.88, "learning_rate": 7.743959334381012e-07, "loss": 2.0626, "step": 35608 }, { "epoch": 0.88, "learning_rate": 7.740879246265876e-07, "loss": 1.8713, "step": 35609 }, { "epoch": 0.88, "learning_rate": 7.7377997461482e-07, "loss": 1.8197, "step": 35610 }, { "epoch": 0.88, "learning_rate": 7.734720834047571e-07, "loss": 1.832, "step": 35611 }, { "epoch": 0.88, "learning_rate": 7.731642509983662e-07, "loss": 1.8751, "step": 35612 }, { "epoch": 0.88, "learning_rate": 7.728564773976044e-07, "loss": 1.8831, "step": 35613 }, { "epoch": 0.88, "learning_rate": 7.725487626044347e-07, "loss": 1.8262, "step": 35614 }, { "epoch": 0.88, "learning_rate": 7.722411066208213e-07, "loss": 1.8801, "step": 35615 }, { "epoch": 0.88, "learning_rate": 7.719335094487201e-07, "loss": 2.0594, "step": 35616 }, { "epoch": 0.88, "learning_rate": 7.716259710900964e-07, "loss": 1.9815, "step": 35617 }, { "epoch": 0.88, "learning_rate": 7.713184915469052e-07, "loss": 2.0482, "step": 35618 }, { "epoch": 0.88, "learning_rate": 7.710110708211105e-07, "loss": 1.8749, "step": 35619 }, { "epoch": 0.88, "learning_rate": 7.707037089146696e-07, "loss": 1.9288, "step": 35620 }, { "epoch": 0.88, "learning_rate": 7.70396405829541e-07, "loss": 1.8226, "step": 35621 }, { "epoch": 0.88, "learning_rate": 7.700891615676842e-07, "loss": 1.8243, "step": 35622 }, { "epoch": 0.88, "learning_rate": 7.697819761310587e-07, "loss": 2.0529, "step": 35623 }, { "epoch": 0.88, "learning_rate": 7.694748495216187e-07, "loss": 2.1854, "step": 35624 }, { "epoch": 0.88, "learning_rate": 7.691677817413234e-07, "loss": 1.8142, "step": 35625 }, { "epoch": 0.88, "learning_rate": 7.688607727921305e-07, "loss": 1.9999, "step": 35626 }, { "epoch": 0.88, "learning_rate": 7.685538226759958e-07, "loss": 2.0151, "step": 35627 }, { "epoch": 0.88, "learning_rate": 7.682469313948782e-07, "loss": 1.6841, "step": 35628 }, { "epoch": 0.88, "learning_rate": 7.679400989507302e-07, "loss": 1.7725, "step": 35629 }, { "epoch": 0.88, "learning_rate": 7.676333253455092e-07, "loss": 1.9399, "step": 35630 }, { "epoch": 0.88, "learning_rate": 7.673266105811683e-07, "loss": 1.9042, "step": 35631 }, { "epoch": 0.88, "learning_rate": 7.670199546596646e-07, "loss": 1.7829, "step": 35632 }, { "epoch": 0.88, "learning_rate": 7.667133575829522e-07, "loss": 1.8885, "step": 35633 }, { "epoch": 0.88, "learning_rate": 7.664068193529839e-07, "loss": 1.8521, "step": 35634 }, { "epoch": 0.88, "learning_rate": 7.661003399717148e-07, "loss": 2.0232, "step": 35635 }, { "epoch": 0.88, "learning_rate": 7.657939194410969e-07, "loss": 1.8253, "step": 35636 }, { "epoch": 0.88, "learning_rate": 7.654875577630827e-07, "loss": 1.997, "step": 35637 }, { "epoch": 0.88, "learning_rate": 7.651812549396287e-07, "loss": 1.8596, "step": 35638 }, { "epoch": 0.88, "learning_rate": 7.64875010972681e-07, "loss": 1.9546, "step": 35639 }, { "epoch": 0.88, "learning_rate": 7.64568825864197e-07, "loss": 1.8879, "step": 35640 }, { "epoch": 0.88, "learning_rate": 7.64262699616124e-07, "loss": 1.8092, "step": 35641 }, { "epoch": 0.88, "learning_rate": 7.639566322304137e-07, "loss": 1.7509, "step": 35642 }, { "epoch": 0.88, "learning_rate": 7.636506237090191e-07, "loss": 1.8174, "step": 35643 }, { "epoch": 0.88, "learning_rate": 7.633446740538874e-07, "loss": 2.0018, "step": 35644 }, { "epoch": 0.88, "learning_rate": 7.630387832669739e-07, "loss": 1.8711, "step": 35645 }, { "epoch": 0.88, "learning_rate": 7.627329513502213e-07, "loss": 1.8636, "step": 35646 }, { "epoch": 0.88, "learning_rate": 7.624271783055825e-07, "loss": 1.8984, "step": 35647 }, { "epoch": 0.88, "learning_rate": 7.621214641350072e-07, "loss": 1.9094, "step": 35648 }, { "epoch": 0.88, "learning_rate": 7.618158088404404e-07, "loss": 1.7941, "step": 35649 }, { "epoch": 0.88, "learning_rate": 7.61510212423835e-07, "loss": 1.9293, "step": 35650 }, { "epoch": 0.88, "learning_rate": 7.612046748871327e-07, "loss": 1.992, "step": 35651 }, { "epoch": 0.88, "learning_rate": 7.608991962322842e-07, "loss": 1.8818, "step": 35652 }, { "epoch": 0.88, "learning_rate": 7.605937764612381e-07, "loss": 2.0214, "step": 35653 }, { "epoch": 0.88, "learning_rate": 7.602884155759371e-07, "loss": 2.0441, "step": 35654 }, { "epoch": 0.88, "learning_rate": 7.599831135783309e-07, "loss": 2.053, "step": 35655 }, { "epoch": 0.88, "learning_rate": 7.596778704703623e-07, "loss": 1.8469, "step": 35656 }, { "epoch": 0.88, "learning_rate": 7.593726862539774e-07, "loss": 2.0249, "step": 35657 }, { "epoch": 0.88, "learning_rate": 7.590675609311227e-07, "loss": 1.7373, "step": 35658 }, { "epoch": 0.88, "learning_rate": 7.58762494503742e-07, "loss": 1.9645, "step": 35659 }, { "epoch": 0.88, "learning_rate": 7.584574869737782e-07, "loss": 1.9713, "step": 35660 }, { "epoch": 0.88, "learning_rate": 7.581525383431764e-07, "loss": 2.0571, "step": 35661 }, { "epoch": 0.88, "learning_rate": 7.57847648613882e-07, "loss": 2.031, "step": 35662 }, { "epoch": 0.88, "learning_rate": 7.575428177878363e-07, "loss": 1.9153, "step": 35663 }, { "epoch": 0.88, "learning_rate": 7.572380458669815e-07, "loss": 2.0515, "step": 35664 }, { "epoch": 0.88, "learning_rate": 7.569333328532625e-07, "loss": 2.0991, "step": 35665 }, { "epoch": 0.88, "learning_rate": 7.566286787486177e-07, "loss": 1.7994, "step": 35666 }, { "epoch": 0.88, "learning_rate": 7.563240835549901e-07, "loss": 1.9724, "step": 35667 }, { "epoch": 0.88, "learning_rate": 7.560195472743248e-07, "loss": 1.9711, "step": 35668 }, { "epoch": 0.88, "learning_rate": 7.557150699085559e-07, "loss": 1.8361, "step": 35669 }, { "epoch": 0.88, "learning_rate": 7.554106514596305e-07, "loss": 1.8682, "step": 35670 }, { "epoch": 0.88, "learning_rate": 7.551062919294838e-07, "loss": 1.9695, "step": 35671 }, { "epoch": 0.88, "learning_rate": 7.548019913200566e-07, "loss": 2.1659, "step": 35672 }, { "epoch": 0.88, "learning_rate": 7.544977496332917e-07, "loss": 2.0919, "step": 35673 }, { "epoch": 0.88, "learning_rate": 7.541935668711242e-07, "loss": 2.1316, "step": 35674 }, { "epoch": 0.88, "learning_rate": 7.538894430354959e-07, "loss": 1.9326, "step": 35675 }, { "epoch": 0.88, "learning_rate": 7.535853781283408e-07, "loss": 1.676, "step": 35676 }, { "epoch": 0.88, "learning_rate": 7.532813721516008e-07, "loss": 2.0619, "step": 35677 }, { "epoch": 0.88, "learning_rate": 7.529774251072109e-07, "loss": 2.0126, "step": 35678 }, { "epoch": 0.88, "learning_rate": 7.526735369971094e-07, "loss": 1.9757, "step": 35679 }, { "epoch": 0.88, "learning_rate": 7.52369707823235e-07, "loss": 1.7433, "step": 35680 }, { "epoch": 0.88, "learning_rate": 7.520659375875205e-07, "loss": 1.7572, "step": 35681 }, { "epoch": 0.88, "learning_rate": 7.517622262919044e-07, "loss": 1.9236, "step": 35682 }, { "epoch": 0.88, "learning_rate": 7.514585739383218e-07, "loss": 1.7669, "step": 35683 }, { "epoch": 0.88, "learning_rate": 7.511549805287078e-07, "loss": 1.8253, "step": 35684 }, { "epoch": 0.88, "learning_rate": 7.508514460649952e-07, "loss": 1.9766, "step": 35685 }, { "epoch": 0.88, "learning_rate": 7.505479705491236e-07, "loss": 1.9696, "step": 35686 }, { "epoch": 0.88, "learning_rate": 7.502445539830216e-07, "loss": 2.0104, "step": 35687 }, { "epoch": 0.88, "learning_rate": 7.499411963686276e-07, "loss": 1.874, "step": 35688 }, { "epoch": 0.88, "learning_rate": 7.496378977078711e-07, "loss": 1.9459, "step": 35689 }, { "epoch": 0.88, "learning_rate": 7.493346580026872e-07, "loss": 1.9185, "step": 35690 }, { "epoch": 0.88, "learning_rate": 7.490314772550101e-07, "loss": 1.9812, "step": 35691 }, { "epoch": 0.88, "learning_rate": 7.48728355466768e-07, "loss": 2.1324, "step": 35692 }, { "epoch": 0.88, "learning_rate": 7.484252926398961e-07, "loss": 1.9836, "step": 35693 }, { "epoch": 0.88, "learning_rate": 7.48122288776324e-07, "loss": 1.8152, "step": 35694 }, { "epoch": 0.88, "learning_rate": 7.478193438779835e-07, "loss": 2.0061, "step": 35695 }, { "epoch": 0.88, "learning_rate": 7.475164579468053e-07, "loss": 1.9856, "step": 35696 }, { "epoch": 0.88, "learning_rate": 7.4721363098472e-07, "loss": 2.1293, "step": 35697 }, { "epoch": 0.88, "learning_rate": 7.469108629936594e-07, "loss": 1.9359, "step": 35698 }, { "epoch": 0.88, "learning_rate": 7.466081539755499e-07, "loss": 1.7252, "step": 35699 }, { "epoch": 0.88, "learning_rate": 7.46305503932322e-07, "loss": 1.9587, "step": 35700 }, { "epoch": 0.88, "learning_rate": 7.460029128659063e-07, "loss": 1.9389, "step": 35701 }, { "epoch": 0.88, "learning_rate": 7.457003807782292e-07, "loss": 1.9702, "step": 35702 }, { "epoch": 0.88, "learning_rate": 7.453979076712204e-07, "loss": 1.9149, "step": 35703 }, { "epoch": 0.88, "learning_rate": 7.450954935468047e-07, "loss": 1.8598, "step": 35704 }, { "epoch": 0.88, "learning_rate": 7.447931384069118e-07, "loss": 1.9007, "step": 35705 }, { "epoch": 0.88, "learning_rate": 7.444908422534702e-07, "loss": 1.8626, "step": 35706 }, { "epoch": 0.88, "learning_rate": 7.441886050884029e-07, "loss": 1.9237, "step": 35707 }, { "epoch": 0.88, "learning_rate": 7.438864269136393e-07, "loss": 1.8105, "step": 35708 }, { "epoch": 0.88, "learning_rate": 7.435843077311034e-07, "loss": 2.0062, "step": 35709 }, { "epoch": 0.88, "learning_rate": 7.432822475427204e-07, "loss": 2.0908, "step": 35710 }, { "epoch": 0.88, "learning_rate": 7.429802463504177e-07, "loss": 1.9288, "step": 35711 }, { "epoch": 0.88, "learning_rate": 7.426783041561181e-07, "loss": 1.9679, "step": 35712 }, { "epoch": 0.88, "learning_rate": 7.423764209617457e-07, "loss": 1.8506, "step": 35713 }, { "epoch": 0.88, "learning_rate": 7.420745967692256e-07, "loss": 1.9623, "step": 35714 }, { "epoch": 0.88, "learning_rate": 7.417728315804818e-07, "loss": 1.9431, "step": 35715 }, { "epoch": 0.88, "learning_rate": 7.414711253974372e-07, "loss": 1.8334, "step": 35716 }, { "epoch": 0.88, "learning_rate": 7.411694782220136e-07, "loss": 1.902, "step": 35717 }, { "epoch": 0.88, "learning_rate": 7.408678900561361e-07, "loss": 1.598, "step": 35718 }, { "epoch": 0.88, "learning_rate": 7.405663609017233e-07, "loss": 1.9819, "step": 35719 }, { "epoch": 0.88, "learning_rate": 7.40264890760698e-07, "loss": 1.84, "step": 35720 }, { "epoch": 0.88, "learning_rate": 7.399634796349842e-07, "loss": 2.0251, "step": 35721 }, { "epoch": 0.88, "learning_rate": 7.396621275264992e-07, "loss": 1.8716, "step": 35722 }, { "epoch": 0.88, "learning_rate": 7.393608344371672e-07, "loss": 2.0469, "step": 35723 }, { "epoch": 0.88, "learning_rate": 7.390596003689044e-07, "loss": 1.7709, "step": 35724 }, { "epoch": 0.88, "learning_rate": 7.387584253236335e-07, "loss": 1.786, "step": 35725 }, { "epoch": 0.88, "learning_rate": 7.384573093032743e-07, "loss": 1.8837, "step": 35726 }, { "epoch": 0.88, "learning_rate": 7.381562523097441e-07, "loss": 1.8152, "step": 35727 }, { "epoch": 0.88, "learning_rate": 7.378552543449635e-07, "loss": 1.7413, "step": 35728 }, { "epoch": 0.88, "learning_rate": 7.375543154108477e-07, "loss": 1.9637, "step": 35729 }, { "epoch": 0.88, "learning_rate": 7.372534355093175e-07, "loss": 1.8636, "step": 35730 }, { "epoch": 0.88, "learning_rate": 7.369526146422901e-07, "loss": 1.9762, "step": 35731 }, { "epoch": 0.88, "learning_rate": 7.366518528116817e-07, "loss": 2.103, "step": 35732 }, { "epoch": 0.88, "learning_rate": 7.363511500194121e-07, "loss": 1.8818, "step": 35733 }, { "epoch": 0.88, "learning_rate": 7.36050506267394e-07, "loss": 1.8835, "step": 35734 }, { "epoch": 0.88, "learning_rate": 7.357499215575448e-07, "loss": 1.9044, "step": 35735 }, { "epoch": 0.88, "learning_rate": 7.354493958917819e-07, "loss": 1.8981, "step": 35736 }, { "epoch": 0.88, "learning_rate": 7.351489292720171e-07, "loss": 1.8733, "step": 35737 }, { "epoch": 0.88, "learning_rate": 7.3484852170017e-07, "loss": 1.8986, "step": 35738 }, { "epoch": 0.88, "learning_rate": 7.345481731781512e-07, "loss": 2.1388, "step": 35739 }, { "epoch": 0.88, "learning_rate": 7.34247883707876e-07, "loss": 1.869, "step": 35740 }, { "epoch": 0.88, "learning_rate": 7.339476532912604e-07, "loss": 1.9012, "step": 35741 }, { "epoch": 0.88, "learning_rate": 7.336474819302142e-07, "loss": 1.69, "step": 35742 }, { "epoch": 0.88, "learning_rate": 7.333473696266525e-07, "loss": 1.8683, "step": 35743 }, { "epoch": 0.88, "learning_rate": 7.330473163824902e-07, "loss": 1.8379, "step": 35744 }, { "epoch": 0.88, "learning_rate": 7.327473221996351e-07, "loss": 1.8867, "step": 35745 }, { "epoch": 0.88, "learning_rate": 7.32447387080003e-07, "loss": 1.9065, "step": 35746 }, { "epoch": 0.88, "learning_rate": 7.321475110255027e-07, "loss": 1.8973, "step": 35747 }, { "epoch": 0.88, "learning_rate": 7.318476940380459e-07, "loss": 1.9091, "step": 35748 }, { "epoch": 0.88, "learning_rate": 7.315479361195443e-07, "loss": 1.9583, "step": 35749 }, { "epoch": 0.88, "learning_rate": 7.312482372719087e-07, "loss": 1.8224, "step": 35750 }, { "epoch": 0.88, "learning_rate": 7.30948597497051e-07, "loss": 1.9932, "step": 35751 }, { "epoch": 0.88, "learning_rate": 7.306490167968761e-07, "loss": 2.114, "step": 35752 }, { "epoch": 0.88, "learning_rate": 7.303494951732948e-07, "loss": 1.9778, "step": 35753 }, { "epoch": 0.88, "learning_rate": 7.300500326282201e-07, "loss": 1.8906, "step": 35754 }, { "epoch": 0.88, "learning_rate": 7.297506291635559e-07, "loss": 1.9566, "step": 35755 }, { "epoch": 0.88, "learning_rate": 7.294512847812129e-07, "loss": 1.9965, "step": 35756 }, { "epoch": 0.88, "learning_rate": 7.291519994830964e-07, "loss": 1.7896, "step": 35757 }, { "epoch": 0.88, "learning_rate": 7.288527732711159e-07, "loss": 1.9497, "step": 35758 }, { "epoch": 0.88, "learning_rate": 7.285536061471799e-07, "loss": 1.7866, "step": 35759 }, { "epoch": 0.88, "learning_rate": 7.282544981131911e-07, "loss": 2.0828, "step": 35760 }, { "epoch": 0.88, "learning_rate": 7.279554491710606e-07, "loss": 1.7539, "step": 35761 }, { "epoch": 0.88, "learning_rate": 7.276564593226887e-07, "loss": 1.8429, "step": 35762 }, { "epoch": 0.88, "learning_rate": 7.273575285699852e-07, "loss": 1.9161, "step": 35763 }, { "epoch": 0.88, "learning_rate": 7.270586569148552e-07, "loss": 1.8892, "step": 35764 }, { "epoch": 0.88, "learning_rate": 7.267598443592005e-07, "loss": 1.8436, "step": 35765 }, { "epoch": 0.88, "learning_rate": 7.264610909049275e-07, "loss": 2.1814, "step": 35766 }, { "epoch": 0.88, "learning_rate": 7.2616239655394e-07, "loss": 2.0018, "step": 35767 }, { "epoch": 0.88, "learning_rate": 7.25863761308141e-07, "loss": 2.0326, "step": 35768 }, { "epoch": 0.88, "learning_rate": 7.255651851694378e-07, "loss": 1.9325, "step": 35769 }, { "epoch": 0.88, "learning_rate": 7.25266668139728e-07, "loss": 1.9262, "step": 35770 }, { "epoch": 0.88, "learning_rate": 7.249682102209177e-07, "loss": 1.776, "step": 35771 }, { "epoch": 0.88, "learning_rate": 7.246698114149053e-07, "loss": 1.9156, "step": 35772 }, { "epoch": 0.88, "learning_rate": 7.243714717235962e-07, "loss": 2.0321, "step": 35773 }, { "epoch": 0.88, "learning_rate": 7.240731911488918e-07, "loss": 1.985, "step": 35774 }, { "epoch": 0.88, "learning_rate": 7.237749696926899e-07, "loss": 2.007, "step": 35775 }, { "epoch": 0.88, "learning_rate": 7.234768073568943e-07, "loss": 1.9025, "step": 35776 }, { "epoch": 0.88, "learning_rate": 7.231787041434035e-07, "loss": 1.9287, "step": 35777 }, { "epoch": 0.88, "learning_rate": 7.228806600541182e-07, "loss": 1.8999, "step": 35778 }, { "epoch": 0.88, "learning_rate": 7.225826750909381e-07, "loss": 1.9384, "step": 35779 }, { "epoch": 0.88, "learning_rate": 7.222847492557616e-07, "loss": 1.8749, "step": 35780 }, { "epoch": 0.88, "learning_rate": 7.219868825504895e-07, "loss": 1.7531, "step": 35781 }, { "epoch": 0.88, "learning_rate": 7.216890749770167e-07, "loss": 1.977, "step": 35782 }, { "epoch": 0.88, "learning_rate": 7.21391326537243e-07, "loss": 1.9095, "step": 35783 }, { "epoch": 0.88, "learning_rate": 7.210936372330657e-07, "loss": 2.0313, "step": 35784 }, { "epoch": 0.88, "learning_rate": 7.207960070663833e-07, "loss": 2.1885, "step": 35785 }, { "epoch": 0.88, "learning_rate": 7.204984360390932e-07, "loss": 2.0839, "step": 35786 }, { "epoch": 0.88, "learning_rate": 7.202009241530894e-07, "loss": 1.8557, "step": 35787 }, { "epoch": 0.88, "learning_rate": 7.199034714102693e-07, "loss": 1.8304, "step": 35788 }, { "epoch": 0.88, "learning_rate": 7.196060778125302e-07, "loss": 1.7973, "step": 35789 }, { "epoch": 0.88, "learning_rate": 7.193087433617641e-07, "loss": 1.9009, "step": 35790 }, { "epoch": 0.88, "learning_rate": 7.190114680598703e-07, "loss": 1.9311, "step": 35791 }, { "epoch": 0.88, "learning_rate": 7.187142519087387e-07, "loss": 1.9119, "step": 35792 }, { "epoch": 0.88, "learning_rate": 7.184170949102654e-07, "loss": 1.8872, "step": 35793 }, { "epoch": 0.88, "learning_rate": 7.181199970663477e-07, "loss": 1.9005, "step": 35794 }, { "epoch": 0.88, "learning_rate": 7.178229583788732e-07, "loss": 1.8158, "step": 35795 }, { "epoch": 0.88, "learning_rate": 7.175259788497402e-07, "loss": 2.0364, "step": 35796 }, { "epoch": 0.88, "learning_rate": 7.172290584808373e-07, "loss": 2.0123, "step": 35797 }, { "epoch": 0.88, "learning_rate": 7.169321972740595e-07, "loss": 1.8114, "step": 35798 }, { "epoch": 0.88, "learning_rate": 7.166353952312988e-07, "loss": 1.9236, "step": 35799 }, { "epoch": 0.88, "learning_rate": 7.163386523544447e-07, "loss": 2.148, "step": 35800 }, { "epoch": 0.88, "learning_rate": 7.160419686453901e-07, "loss": 2.0924, "step": 35801 }, { "epoch": 0.88, "learning_rate": 7.157453441060258e-07, "loss": 1.9805, "step": 35802 }, { "epoch": 0.88, "learning_rate": 7.154487787382414e-07, "loss": 1.9403, "step": 35803 }, { "epoch": 0.88, "learning_rate": 7.151522725439298e-07, "loss": 1.9499, "step": 35804 }, { "epoch": 0.88, "learning_rate": 7.148558255249771e-07, "loss": 1.8872, "step": 35805 }, { "epoch": 0.88, "learning_rate": 7.145594376832731e-07, "loss": 2.0135, "step": 35806 }, { "epoch": 0.88, "learning_rate": 7.142631090207108e-07, "loss": 1.9159, "step": 35807 }, { "epoch": 0.88, "learning_rate": 7.139668395391731e-07, "loss": 1.9992, "step": 35808 }, { "epoch": 0.88, "learning_rate": 7.136706292405538e-07, "loss": 2.0097, "step": 35809 }, { "epoch": 0.88, "learning_rate": 7.133744781267348e-07, "loss": 2.0101, "step": 35810 }, { "epoch": 0.88, "learning_rate": 7.130783861996082e-07, "loss": 1.9396, "step": 35811 }, { "epoch": 0.88, "learning_rate": 7.127823534610601e-07, "loss": 1.9609, "step": 35812 }, { "epoch": 0.88, "learning_rate": 7.124863799129755e-07, "loss": 1.8644, "step": 35813 }, { "epoch": 0.88, "learning_rate": 7.121904655572431e-07, "loss": 1.7236, "step": 35814 }, { "epoch": 0.88, "learning_rate": 7.11894610395747e-07, "loss": 2.1075, "step": 35815 }, { "epoch": 0.88, "learning_rate": 7.115988144303731e-07, "loss": 2.0777, "step": 35816 }, { "epoch": 0.88, "learning_rate": 7.113030776630081e-07, "loss": 1.7547, "step": 35817 }, { "epoch": 0.88, "learning_rate": 7.110074000955347e-07, "loss": 1.9567, "step": 35818 }, { "epoch": 0.88, "learning_rate": 7.10711781729837e-07, "loss": 2.1305, "step": 35819 }, { "epoch": 0.88, "learning_rate": 7.104162225678024e-07, "loss": 1.8754, "step": 35820 }, { "epoch": 0.88, "learning_rate": 7.101207226113116e-07, "loss": 1.956, "step": 35821 }, { "epoch": 0.88, "learning_rate": 7.098252818622498e-07, "loss": 1.721, "step": 35822 }, { "epoch": 0.88, "learning_rate": 7.095299003224987e-07, "loss": 1.8975, "step": 35823 }, { "epoch": 0.88, "learning_rate": 7.092345779939424e-07, "loss": 1.9966, "step": 35824 }, { "epoch": 0.88, "learning_rate": 7.089393148784596e-07, "loss": 1.7871, "step": 35825 }, { "epoch": 0.88, "learning_rate": 7.086441109779352e-07, "loss": 1.8851, "step": 35826 }, { "epoch": 0.88, "learning_rate": 7.083489662942511e-07, "loss": 1.8473, "step": 35827 }, { "epoch": 0.88, "learning_rate": 7.080538808292858e-07, "loss": 1.8538, "step": 35828 }, { "epoch": 0.88, "learning_rate": 7.077588545849224e-07, "loss": 2.0278, "step": 35829 }, { "epoch": 0.88, "learning_rate": 7.074638875630391e-07, "loss": 2.0296, "step": 35830 }, { "epoch": 0.88, "learning_rate": 7.071689797655168e-07, "loss": 2.0264, "step": 35831 }, { "epoch": 0.88, "learning_rate": 7.068741311942362e-07, "loss": 1.9556, "step": 35832 }, { "epoch": 0.88, "learning_rate": 7.065793418510736e-07, "loss": 1.9583, "step": 35833 }, { "epoch": 0.88, "learning_rate": 7.062846117379096e-07, "loss": 1.9569, "step": 35834 }, { "epoch": 0.88, "learning_rate": 7.059899408566239e-07, "loss": 1.9754, "step": 35835 }, { "epoch": 0.88, "learning_rate": 7.056953292090917e-07, "loss": 2.0881, "step": 35836 }, { "epoch": 0.88, "learning_rate": 7.054007767971927e-07, "loss": 2.0829, "step": 35837 }, { "epoch": 0.88, "learning_rate": 7.051062836228028e-07, "loss": 1.9617, "step": 35838 }, { "epoch": 0.88, "learning_rate": 7.04811849687802e-07, "loss": 1.8824, "step": 35839 }, { "epoch": 0.88, "learning_rate": 7.045174749940631e-07, "loss": 2.1366, "step": 35840 }, { "epoch": 0.88, "learning_rate": 7.042231595434623e-07, "loss": 2.0102, "step": 35841 }, { "epoch": 0.88, "learning_rate": 7.039289033378793e-07, "loss": 1.9464, "step": 35842 }, { "epoch": 0.88, "learning_rate": 7.036347063791848e-07, "loss": 2.0847, "step": 35843 }, { "epoch": 0.88, "learning_rate": 7.033405686692585e-07, "loss": 2.0099, "step": 35844 }, { "epoch": 0.88, "learning_rate": 7.030464902099699e-07, "loss": 2.0235, "step": 35845 }, { "epoch": 0.88, "learning_rate": 7.027524710031963e-07, "loss": 1.8171, "step": 35846 }, { "epoch": 0.88, "learning_rate": 7.02458511050812e-07, "loss": 2.1141, "step": 35847 }, { "epoch": 0.88, "learning_rate": 7.021646103546887e-07, "loss": 2.1091, "step": 35848 }, { "epoch": 0.88, "learning_rate": 7.018707689167015e-07, "loss": 1.8562, "step": 35849 }, { "epoch": 0.88, "learning_rate": 7.015769867387201e-07, "loss": 1.8255, "step": 35850 }, { "epoch": 0.88, "learning_rate": 7.012832638226197e-07, "loss": 1.9028, "step": 35851 }, { "epoch": 0.88, "learning_rate": 7.00989600170271e-07, "loss": 1.788, "step": 35852 }, { "epoch": 0.88, "learning_rate": 7.006959957835479e-07, "loss": 2.0558, "step": 35853 }, { "epoch": 0.88, "learning_rate": 7.004024506643181e-07, "loss": 2.0173, "step": 35854 }, { "epoch": 0.88, "learning_rate": 7.001089648144544e-07, "loss": 2.0425, "step": 35855 }, { "epoch": 0.88, "learning_rate": 6.998155382358262e-07, "loss": 1.9541, "step": 35856 }, { "epoch": 0.88, "learning_rate": 6.995221709303069e-07, "loss": 1.8576, "step": 35857 }, { "epoch": 0.88, "learning_rate": 6.992288628997612e-07, "loss": 2.0272, "step": 35858 }, { "epoch": 0.88, "learning_rate": 6.989356141460624e-07, "loss": 2.0934, "step": 35859 }, { "epoch": 0.88, "learning_rate": 6.986424246710788e-07, "loss": 1.8863, "step": 35860 }, { "epoch": 0.88, "learning_rate": 6.983492944766767e-07, "loss": 1.9179, "step": 35861 }, { "epoch": 0.88, "learning_rate": 6.98056223564727e-07, "loss": 1.9194, "step": 35862 }, { "epoch": 0.88, "learning_rate": 6.977632119370958e-07, "loss": 1.9327, "step": 35863 }, { "epoch": 0.88, "learning_rate": 6.974702595956506e-07, "loss": 1.9082, "step": 35864 }, { "epoch": 0.88, "learning_rate": 6.971773665422598e-07, "loss": 1.9418, "step": 35865 }, { "epoch": 0.88, "learning_rate": 6.968845327787887e-07, "loss": 1.8219, "step": 35866 }, { "epoch": 0.88, "learning_rate": 6.965917583071058e-07, "loss": 1.9165, "step": 35867 }, { "epoch": 0.88, "learning_rate": 6.962990431290739e-07, "loss": 1.9179, "step": 35868 }, { "epoch": 0.88, "learning_rate": 6.960063872465606e-07, "loss": 1.9717, "step": 35869 }, { "epoch": 0.88, "learning_rate": 6.957137906614319e-07, "loss": 1.8648, "step": 35870 }, { "epoch": 0.88, "learning_rate": 6.9542125337555e-07, "loss": 2.071, "step": 35871 }, { "epoch": 0.88, "learning_rate": 6.951287753907798e-07, "loss": 1.8, "step": 35872 }, { "epoch": 0.88, "learning_rate": 6.948363567089867e-07, "loss": 1.8262, "step": 35873 }, { "epoch": 0.88, "learning_rate": 6.945439973320345e-07, "loss": 1.8338, "step": 35874 }, { "epoch": 0.88, "learning_rate": 6.942516972617875e-07, "loss": 1.8628, "step": 35875 }, { "epoch": 0.88, "learning_rate": 6.939594565001051e-07, "loss": 1.8909, "step": 35876 }, { "epoch": 0.88, "learning_rate": 6.936672750488538e-07, "loss": 1.9188, "step": 35877 }, { "epoch": 0.88, "learning_rate": 6.93375152909892e-07, "loss": 2.1702, "step": 35878 }, { "epoch": 0.88, "learning_rate": 6.930830900850827e-07, "loss": 1.9044, "step": 35879 }, { "epoch": 0.88, "learning_rate": 6.927910865762911e-07, "loss": 1.7974, "step": 35880 }, { "epoch": 0.88, "learning_rate": 6.924991423853721e-07, "loss": 1.7673, "step": 35881 }, { "epoch": 0.88, "learning_rate": 6.922072575141913e-07, "loss": 1.8827, "step": 35882 }, { "epoch": 0.88, "learning_rate": 6.919154319646049e-07, "loss": 1.8045, "step": 35883 }, { "epoch": 0.88, "learning_rate": 6.916236657384756e-07, "loss": 1.956, "step": 35884 }, { "epoch": 0.88, "learning_rate": 6.913319588376632e-07, "loss": 1.9934, "step": 35885 }, { "epoch": 0.88, "learning_rate": 6.91040311264024e-07, "loss": 1.9597, "step": 35886 }, { "epoch": 0.88, "learning_rate": 6.907487230194199e-07, "loss": 1.955, "step": 35887 }, { "epoch": 0.88, "learning_rate": 6.904571941057081e-07, "loss": 1.8917, "step": 35888 }, { "epoch": 0.88, "learning_rate": 6.901657245247462e-07, "loss": 1.9854, "step": 35889 }, { "epoch": 0.88, "learning_rate": 6.898743142783915e-07, "loss": 1.8509, "step": 35890 }, { "epoch": 0.88, "learning_rate": 6.895829633685014e-07, "loss": 2.0282, "step": 35891 }, { "epoch": 0.88, "learning_rate": 6.892916717969356e-07, "loss": 1.9697, "step": 35892 }, { "epoch": 0.88, "learning_rate": 6.890004395655459e-07, "loss": 1.9679, "step": 35893 }, { "epoch": 0.88, "learning_rate": 6.887092666761918e-07, "loss": 1.7964, "step": 35894 }, { "epoch": 0.88, "learning_rate": 6.884181531307288e-07, "loss": 2.0944, "step": 35895 }, { "epoch": 0.88, "learning_rate": 6.881270989310107e-07, "loss": 1.872, "step": 35896 }, { "epoch": 0.88, "learning_rate": 6.878361040788939e-07, "loss": 1.9358, "step": 35897 }, { "epoch": 0.88, "learning_rate": 6.875451685762314e-07, "loss": 1.9593, "step": 35898 }, { "epoch": 0.88, "learning_rate": 6.872542924248782e-07, "loss": 1.851, "step": 35899 }, { "epoch": 0.88, "learning_rate": 6.869634756266896e-07, "loss": 2.146, "step": 35900 }, { "epoch": 0.88, "learning_rate": 6.866727181835164e-07, "loss": 1.9825, "step": 35901 }, { "epoch": 0.88, "learning_rate": 6.863820200972149e-07, "loss": 1.9955, "step": 35902 }, { "epoch": 0.88, "learning_rate": 6.860913813696335e-07, "loss": 1.8443, "step": 35903 }, { "epoch": 0.88, "learning_rate": 6.858008020026286e-07, "loss": 1.8432, "step": 35904 }, { "epoch": 0.89, "learning_rate": 6.855102819980497e-07, "loss": 1.8506, "step": 35905 }, { "epoch": 0.89, "learning_rate": 6.85219821357751e-07, "loss": 1.9399, "step": 35906 }, { "epoch": 0.89, "learning_rate": 6.849294200835799e-07, "loss": 1.9758, "step": 35907 }, { "epoch": 0.89, "learning_rate": 6.846390781773893e-07, "loss": 1.9361, "step": 35908 }, { "epoch": 0.89, "learning_rate": 6.843487956410311e-07, "loss": 2.012, "step": 35909 }, { "epoch": 0.89, "learning_rate": 6.840585724763538e-07, "loss": 1.8137, "step": 35910 }, { "epoch": 0.89, "learning_rate": 6.837684086852059e-07, "loss": 1.8196, "step": 35911 }, { "epoch": 0.89, "learning_rate": 6.834783042694404e-07, "loss": 1.8554, "step": 35912 }, { "epoch": 0.89, "learning_rate": 6.831882592309013e-07, "loss": 1.919, "step": 35913 }, { "epoch": 0.89, "learning_rate": 6.828982735714407e-07, "loss": 2.0718, "step": 35914 }, { "epoch": 0.89, "learning_rate": 6.826083472929068e-07, "loss": 1.7605, "step": 35915 }, { "epoch": 0.89, "learning_rate": 6.82318480397145e-07, "loss": 2.0396, "step": 35916 }, { "epoch": 0.89, "learning_rate": 6.820286728860059e-07, "loss": 1.8904, "step": 35917 }, { "epoch": 0.89, "learning_rate": 6.817389247613326e-07, "loss": 1.8301, "step": 35918 }, { "epoch": 0.89, "learning_rate": 6.814492360249736e-07, "loss": 1.9803, "step": 35919 }, { "epoch": 0.89, "learning_rate": 6.811596066787773e-07, "loss": 1.9552, "step": 35920 }, { "epoch": 0.89, "learning_rate": 6.808700367245869e-07, "loss": 1.9318, "step": 35921 }, { "epoch": 0.89, "learning_rate": 6.805805261642473e-07, "loss": 2.0396, "step": 35922 }, { "epoch": 0.89, "learning_rate": 6.802910749996061e-07, "loss": 1.9297, "step": 35923 }, { "epoch": 0.89, "learning_rate": 6.800016832325084e-07, "loss": 1.8627, "step": 35924 }, { "epoch": 0.89, "learning_rate": 6.797123508647962e-07, "loss": 1.7995, "step": 35925 }, { "epoch": 0.89, "learning_rate": 6.794230778983136e-07, "loss": 2.047, "step": 35926 }, { "epoch": 0.89, "learning_rate": 6.791338643349066e-07, "loss": 1.9582, "step": 35927 }, { "epoch": 0.89, "learning_rate": 6.788447101764173e-07, "loss": 1.7529, "step": 35928 }, { "epoch": 0.89, "learning_rate": 6.785556154246875e-07, "loss": 1.8784, "step": 35929 }, { "epoch": 0.89, "learning_rate": 6.782665800815624e-07, "loss": 1.9458, "step": 35930 }, { "epoch": 0.89, "learning_rate": 6.779776041488795e-07, "loss": 1.9687, "step": 35931 }, { "epoch": 0.89, "learning_rate": 6.776886876284839e-07, "loss": 1.85, "step": 35932 }, { "epoch": 0.89, "learning_rate": 6.773998305222185e-07, "loss": 2.0328, "step": 35933 }, { "epoch": 0.89, "learning_rate": 6.771110328319197e-07, "loss": 1.8773, "step": 35934 }, { "epoch": 0.89, "learning_rate": 6.768222945594328e-07, "loss": 1.8897, "step": 35935 }, { "epoch": 0.89, "learning_rate": 6.765336157065938e-07, "loss": 1.9918, "step": 35936 }, { "epoch": 0.89, "learning_rate": 6.762449962752449e-07, "loss": 1.7187, "step": 35937 }, { "epoch": 0.89, "learning_rate": 6.759564362672255e-07, "loss": 1.91, "step": 35938 }, { "epoch": 0.89, "learning_rate": 6.756679356843742e-07, "loss": 1.9655, "step": 35939 }, { "epoch": 0.89, "learning_rate": 6.753794945285297e-07, "loss": 1.9438, "step": 35940 }, { "epoch": 0.89, "learning_rate": 6.750911128015291e-07, "loss": 1.9655, "step": 35941 }, { "epoch": 0.89, "learning_rate": 6.748027905052145e-07, "loss": 1.8957, "step": 35942 }, { "epoch": 0.89, "learning_rate": 6.745145276414189e-07, "loss": 1.9951, "step": 35943 }, { "epoch": 0.89, "learning_rate": 6.742263242119818e-07, "loss": 1.8726, "step": 35944 }, { "epoch": 0.89, "learning_rate": 6.739381802187406e-07, "loss": 1.8437, "step": 35945 }, { "epoch": 0.89, "learning_rate": 6.736500956635294e-07, "loss": 1.9894, "step": 35946 }, { "epoch": 0.89, "learning_rate": 6.733620705481847e-07, "loss": 2.0639, "step": 35947 }, { "epoch": 0.89, "learning_rate": 6.730741048745448e-07, "loss": 1.8335, "step": 35948 }, { "epoch": 0.89, "learning_rate": 6.727861986444428e-07, "loss": 1.8826, "step": 35949 }, { "epoch": 0.89, "learning_rate": 6.72498351859714e-07, "loss": 1.9891, "step": 35950 }, { "epoch": 0.89, "learning_rate": 6.722105645221921e-07, "loss": 1.922, "step": 35951 }, { "epoch": 0.89, "learning_rate": 6.719228366337127e-07, "loss": 1.8718, "step": 35952 }, { "epoch": 0.89, "learning_rate": 6.716351681961098e-07, "loss": 1.9048, "step": 35953 }, { "epoch": 0.89, "learning_rate": 6.71347559211214e-07, "loss": 1.9529, "step": 35954 }, { "epoch": 0.89, "learning_rate": 6.710600096808628e-07, "loss": 1.7969, "step": 35955 }, { "epoch": 0.89, "learning_rate": 6.707725196068849e-07, "loss": 1.9848, "step": 35956 }, { "epoch": 0.89, "learning_rate": 6.70485088991113e-07, "loss": 2.0533, "step": 35957 }, { "epoch": 0.89, "learning_rate": 6.701977178353802e-07, "loss": 1.9898, "step": 35958 }, { "epoch": 0.89, "learning_rate": 6.699104061415207e-07, "loss": 1.9016, "step": 35959 }, { "epoch": 0.89, "learning_rate": 6.696231539113595e-07, "loss": 1.8681, "step": 35960 }, { "epoch": 0.89, "learning_rate": 6.69335961146732e-07, "loss": 1.9426, "step": 35961 }, { "epoch": 0.89, "learning_rate": 6.690488278494666e-07, "loss": 1.9893, "step": 35962 }, { "epoch": 0.89, "learning_rate": 6.687617540213953e-07, "loss": 1.9181, "step": 35963 }, { "epoch": 0.89, "learning_rate": 6.684747396643454e-07, "loss": 1.9394, "step": 35964 }, { "epoch": 0.89, "learning_rate": 6.681877847801477e-07, "loss": 1.9817, "step": 35965 }, { "epoch": 0.89, "learning_rate": 6.679008893706296e-07, "loss": 1.8486, "step": 35966 }, { "epoch": 0.89, "learning_rate": 6.676140534376197e-07, "loss": 2.0848, "step": 35967 }, { "epoch": 0.89, "learning_rate": 6.673272769829486e-07, "loss": 1.957, "step": 35968 }, { "epoch": 0.89, "learning_rate": 6.670405600084417e-07, "loss": 2.0444, "step": 35969 }, { "epoch": 0.89, "learning_rate": 6.667539025159275e-07, "loss": 1.8247, "step": 35970 }, { "epoch": 0.89, "learning_rate": 6.664673045072301e-07, "loss": 1.6386, "step": 35971 }, { "epoch": 0.89, "learning_rate": 6.66180765984179e-07, "loss": 1.9695, "step": 35972 }, { "epoch": 0.89, "learning_rate": 6.658942869486007e-07, "loss": 1.9155, "step": 35973 }, { "epoch": 0.89, "learning_rate": 6.656078674023192e-07, "loss": 1.8373, "step": 35974 }, { "epoch": 0.89, "learning_rate": 6.653215073471608e-07, "loss": 1.9609, "step": 35975 }, { "epoch": 0.89, "learning_rate": 6.650352067849497e-07, "loss": 1.8987, "step": 35976 }, { "epoch": 0.89, "learning_rate": 6.647489657175133e-07, "loss": 1.9298, "step": 35977 }, { "epoch": 0.89, "learning_rate": 6.644627841466722e-07, "loss": 2.0748, "step": 35978 }, { "epoch": 0.89, "learning_rate": 6.641766620742529e-07, "loss": 2.1042, "step": 35979 }, { "epoch": 0.89, "learning_rate": 6.638905995020772e-07, "loss": 1.8529, "step": 35980 }, { "epoch": 0.89, "learning_rate": 6.636045964319715e-07, "loss": 2.0372, "step": 35981 }, { "epoch": 0.89, "learning_rate": 6.633186528657554e-07, "loss": 1.9312, "step": 35982 }, { "epoch": 0.89, "learning_rate": 6.63032768805253e-07, "loss": 1.8896, "step": 35983 }, { "epoch": 0.89, "learning_rate": 6.627469442522838e-07, "loss": 2.0034, "step": 35984 }, { "epoch": 0.89, "learning_rate": 6.624611792086722e-07, "loss": 2.0018, "step": 35985 }, { "epoch": 0.89, "learning_rate": 6.621754736762387e-07, "loss": 1.8721, "step": 35986 }, { "epoch": 0.89, "learning_rate": 6.618898276568031e-07, "loss": 1.9484, "step": 35987 }, { "epoch": 0.89, "learning_rate": 6.616042411521884e-07, "loss": 1.9752, "step": 35988 }, { "epoch": 0.89, "learning_rate": 6.613187141642119e-07, "loss": 2.0578, "step": 35989 }, { "epoch": 0.89, "learning_rate": 6.610332466946945e-07, "loss": 1.7651, "step": 35990 }, { "epoch": 0.89, "learning_rate": 6.607478387454569e-07, "loss": 1.9464, "step": 35991 }, { "epoch": 0.89, "learning_rate": 6.604624903183165e-07, "loss": 2.1415, "step": 35992 }, { "epoch": 0.89, "learning_rate": 6.601772014150909e-07, "loss": 1.8984, "step": 35993 }, { "epoch": 0.89, "learning_rate": 6.598919720376007e-07, "loss": 1.899, "step": 35994 }, { "epoch": 0.89, "learning_rate": 6.596068021876633e-07, "loss": 1.9478, "step": 35995 }, { "epoch": 0.89, "learning_rate": 6.593216918670952e-07, "loss": 1.8992, "step": 35996 }, { "epoch": 0.89, "learning_rate": 6.590366410777138e-07, "loss": 1.8072, "step": 35997 }, { "epoch": 0.89, "learning_rate": 6.587516498213376e-07, "loss": 1.8456, "step": 35998 }, { "epoch": 0.89, "learning_rate": 6.584667180997794e-07, "loss": 1.989, "step": 35999 }, { "epoch": 0.89, "learning_rate": 6.58181845914857e-07, "loss": 1.8145, "step": 36000 }, { "epoch": 0.89, "learning_rate": 6.578970332683888e-07, "loss": 1.9605, "step": 36001 }, { "epoch": 0.89, "learning_rate": 6.576122801621843e-07, "loss": 1.9597, "step": 36002 }, { "epoch": 0.89, "learning_rate": 6.573275865980632e-07, "loss": 2.0849, "step": 36003 }, { "epoch": 0.89, "learning_rate": 6.570429525778377e-07, "loss": 1.9514, "step": 36004 }, { "epoch": 0.89, "learning_rate": 6.567583781033204e-07, "loss": 1.8431, "step": 36005 }, { "epoch": 0.89, "learning_rate": 6.564738631763301e-07, "loss": 1.8016, "step": 36006 }, { "epoch": 0.89, "learning_rate": 6.561894077986741e-07, "loss": 1.9276, "step": 36007 }, { "epoch": 0.89, "learning_rate": 6.559050119721699e-07, "loss": 2.0206, "step": 36008 }, { "epoch": 0.89, "learning_rate": 6.556206756986272e-07, "loss": 1.8289, "step": 36009 }, { "epoch": 0.89, "learning_rate": 6.553363989798589e-07, "loss": 1.9641, "step": 36010 }, { "epoch": 0.89, "learning_rate": 6.55052181817677e-07, "loss": 1.8802, "step": 36011 }, { "epoch": 0.89, "learning_rate": 6.547680242138932e-07, "loss": 1.8974, "step": 36012 }, { "epoch": 0.89, "learning_rate": 6.544839261703196e-07, "loss": 1.7981, "step": 36013 }, { "epoch": 0.89, "learning_rate": 6.541998876887634e-07, "loss": 1.9912, "step": 36014 }, { "epoch": 0.89, "learning_rate": 6.539159087710378e-07, "loss": 1.9497, "step": 36015 }, { "epoch": 0.89, "learning_rate": 6.536319894189536e-07, "loss": 2.0127, "step": 36016 }, { "epoch": 0.89, "learning_rate": 6.53348129634317e-07, "loss": 1.8879, "step": 36017 }, { "epoch": 0.89, "learning_rate": 6.530643294189398e-07, "loss": 1.8643, "step": 36018 }, { "epoch": 0.89, "learning_rate": 6.527805887746286e-07, "loss": 1.9032, "step": 36019 }, { "epoch": 0.89, "learning_rate": 6.524969077031929e-07, "loss": 2.0532, "step": 36020 }, { "epoch": 0.89, "learning_rate": 6.522132862064424e-07, "loss": 1.836, "step": 36021 }, { "epoch": 0.89, "learning_rate": 6.519297242861799e-07, "loss": 1.9179, "step": 36022 }, { "epoch": 0.89, "learning_rate": 6.516462219442187e-07, "loss": 2.0931, "step": 36023 }, { "epoch": 0.89, "learning_rate": 6.513627791823607e-07, "loss": 1.977, "step": 36024 }, { "epoch": 0.89, "learning_rate": 6.510793960024153e-07, "loss": 1.9357, "step": 36025 }, { "epoch": 0.89, "learning_rate": 6.507960724061879e-07, "loss": 1.9084, "step": 36026 }, { "epoch": 0.89, "learning_rate": 6.505128083954826e-07, "loss": 1.9508, "step": 36027 }, { "epoch": 0.89, "learning_rate": 6.502296039721067e-07, "loss": 2.02, "step": 36028 }, { "epoch": 0.89, "learning_rate": 6.499464591378646e-07, "loss": 2.1433, "step": 36029 }, { "epoch": 0.89, "learning_rate": 6.496633738945601e-07, "loss": 2.1267, "step": 36030 }, { "epoch": 0.89, "learning_rate": 6.493803482439998e-07, "loss": 1.9785, "step": 36031 }, { "epoch": 0.89, "learning_rate": 6.490973821879842e-07, "loss": 2.1068, "step": 36032 }, { "epoch": 0.89, "learning_rate": 6.48814475728321e-07, "loss": 1.8253, "step": 36033 }, { "epoch": 0.89, "learning_rate": 6.485316288668075e-07, "loss": 1.9346, "step": 36034 }, { "epoch": 0.89, "learning_rate": 6.482488416052502e-07, "loss": 2.0127, "step": 36035 }, { "epoch": 0.89, "learning_rate": 6.47966113945453e-07, "loss": 1.9579, "step": 36036 }, { "epoch": 0.89, "learning_rate": 6.476834458892123e-07, "loss": 2.0646, "step": 36037 }, { "epoch": 0.89, "learning_rate": 6.474008374383345e-07, "loss": 2.0008, "step": 36038 }, { "epoch": 0.89, "learning_rate": 6.471182885946192e-07, "loss": 1.8439, "step": 36039 }, { "epoch": 0.89, "learning_rate": 6.468357993598662e-07, "loss": 1.8631, "step": 36040 }, { "epoch": 0.89, "learning_rate": 6.465533697358783e-07, "loss": 1.9556, "step": 36041 }, { "epoch": 0.89, "learning_rate": 6.46270999724452e-07, "loss": 1.8885, "step": 36042 }, { "epoch": 0.89, "learning_rate": 6.459886893273903e-07, "loss": 2.075, "step": 36043 }, { "epoch": 0.89, "learning_rate": 6.457064385464918e-07, "loss": 1.8401, "step": 36044 }, { "epoch": 0.89, "learning_rate": 6.454242473835526e-07, "loss": 1.9574, "step": 36045 }, { "epoch": 0.89, "learning_rate": 6.451421158403736e-07, "loss": 1.8485, "step": 36046 }, { "epoch": 0.89, "learning_rate": 6.448600439187525e-07, "loss": 1.9236, "step": 36047 }, { "epoch": 0.89, "learning_rate": 6.445780316204886e-07, "loss": 1.8419, "step": 36048 }, { "epoch": 0.89, "learning_rate": 6.442960789473763e-07, "loss": 1.9357, "step": 36049 }, { "epoch": 0.89, "learning_rate": 6.440141859012139e-07, "loss": 1.823, "step": 36050 }, { "epoch": 0.89, "learning_rate": 6.437323524837991e-07, "loss": 1.9589, "step": 36051 }, { "epoch": 0.89, "learning_rate": 6.43450578696927e-07, "loss": 1.9381, "step": 36052 }, { "epoch": 0.89, "learning_rate": 6.431688645423929e-07, "loss": 2.0141, "step": 36053 }, { "epoch": 0.89, "learning_rate": 6.428872100219941e-07, "loss": 1.9065, "step": 36054 }, { "epoch": 0.89, "learning_rate": 6.426056151375226e-07, "loss": 1.8117, "step": 36055 }, { "epoch": 0.89, "learning_rate": 6.423240798907781e-07, "loss": 2.0297, "step": 36056 }, { "epoch": 0.89, "learning_rate": 6.420426042835493e-07, "loss": 2.0055, "step": 36057 }, { "epoch": 0.89, "learning_rate": 6.417611883176322e-07, "loss": 1.93, "step": 36058 }, { "epoch": 0.89, "learning_rate": 6.414798319948235e-07, "loss": 1.843, "step": 36059 }, { "epoch": 0.89, "learning_rate": 6.411985353169114e-07, "loss": 2.1444, "step": 36060 }, { "epoch": 0.89, "learning_rate": 6.409172982856926e-07, "loss": 2.0106, "step": 36061 }, { "epoch": 0.89, "learning_rate": 6.406361209029554e-07, "loss": 1.9209, "step": 36062 }, { "epoch": 0.89, "learning_rate": 6.403550031704964e-07, "loss": 1.9857, "step": 36063 }, { "epoch": 0.89, "learning_rate": 6.400739450901039e-07, "loss": 1.9525, "step": 36064 }, { "epoch": 0.89, "learning_rate": 6.3979294666357e-07, "loss": 2.0446, "step": 36065 }, { "epoch": 0.89, "learning_rate": 6.395120078926887e-07, "loss": 2.0646, "step": 36066 }, { "epoch": 0.89, "learning_rate": 6.392311287792463e-07, "loss": 1.8896, "step": 36067 }, { "epoch": 0.89, "learning_rate": 6.389503093250337e-07, "loss": 2.0674, "step": 36068 }, { "epoch": 0.89, "learning_rate": 6.386695495318429e-07, "loss": 1.8778, "step": 36069 }, { "epoch": 0.89, "learning_rate": 6.383888494014612e-07, "loss": 1.9206, "step": 36070 }, { "epoch": 0.89, "learning_rate": 6.381082089356794e-07, "loss": 1.9203, "step": 36071 }, { "epoch": 0.89, "learning_rate": 6.378276281362827e-07, "loss": 1.7703, "step": 36072 }, { "epoch": 0.89, "learning_rate": 6.375471070050632e-07, "loss": 1.91, "step": 36073 }, { "epoch": 0.89, "learning_rate": 6.372666455438071e-07, "loss": 1.8487, "step": 36074 }, { "epoch": 0.89, "learning_rate": 6.369862437543006e-07, "loss": 2.0079, "step": 36075 }, { "epoch": 0.89, "learning_rate": 6.367059016383348e-07, "loss": 1.9791, "step": 36076 }, { "epoch": 0.89, "learning_rate": 6.364256191976914e-07, "loss": 1.8883, "step": 36077 }, { "epoch": 0.89, "learning_rate": 6.361453964341579e-07, "loss": 1.813, "step": 36078 }, { "epoch": 0.89, "learning_rate": 6.358652333495241e-07, "loss": 1.8949, "step": 36079 }, { "epoch": 0.89, "learning_rate": 6.355851299455717e-07, "loss": 1.966, "step": 36080 }, { "epoch": 0.89, "learning_rate": 6.353050862240862e-07, "loss": 1.9411, "step": 36081 }, { "epoch": 0.89, "learning_rate": 6.350251021868536e-07, "loss": 1.9143, "step": 36082 }, { "epoch": 0.89, "learning_rate": 6.347451778356584e-07, "loss": 1.8811, "step": 36083 }, { "epoch": 0.89, "learning_rate": 6.344653131722855e-07, "loss": 1.9877, "step": 36084 }, { "epoch": 0.89, "learning_rate": 6.341855081985148e-07, "loss": 1.8693, "step": 36085 }, { "epoch": 0.89, "learning_rate": 6.33905762916135e-07, "loss": 1.937, "step": 36086 }, { "epoch": 0.89, "learning_rate": 6.336260773269232e-07, "loss": 1.8752, "step": 36087 }, { "epoch": 0.89, "learning_rate": 6.333464514326659e-07, "loss": 1.877, "step": 36088 }, { "epoch": 0.89, "learning_rate": 6.330668852351451e-07, "loss": 1.9093, "step": 36089 }, { "epoch": 0.89, "learning_rate": 6.327873787361405e-07, "loss": 1.98, "step": 36090 }, { "epoch": 0.89, "learning_rate": 6.325079319374361e-07, "loss": 2.0947, "step": 36091 }, { "epoch": 0.89, "learning_rate": 6.322285448408105e-07, "loss": 1.9728, "step": 36092 }, { "epoch": 0.89, "learning_rate": 6.319492174480446e-07, "loss": 1.9868, "step": 36093 }, { "epoch": 0.89, "learning_rate": 6.316699497609203e-07, "loss": 1.7505, "step": 36094 }, { "epoch": 0.89, "learning_rate": 6.313907417812159e-07, "loss": 1.7701, "step": 36095 }, { "epoch": 0.89, "learning_rate": 6.311115935107104e-07, "loss": 1.8745, "step": 36096 }, { "epoch": 0.89, "learning_rate": 6.308325049511866e-07, "loss": 1.9157, "step": 36097 }, { "epoch": 0.89, "learning_rate": 6.305534761044186e-07, "loss": 1.9628, "step": 36098 }, { "epoch": 0.89, "learning_rate": 6.302745069721861e-07, "loss": 1.9037, "step": 36099 }, { "epoch": 0.89, "learning_rate": 6.299955975562688e-07, "loss": 1.8905, "step": 36100 }, { "epoch": 0.89, "learning_rate": 6.29716747858442e-07, "loss": 1.8334, "step": 36101 }, { "epoch": 0.89, "learning_rate": 6.294379578804865e-07, "loss": 2.0299, "step": 36102 }, { "epoch": 0.89, "learning_rate": 6.291592276241743e-07, "loss": 1.9266, "step": 36103 }, { "epoch": 0.89, "learning_rate": 6.288805570912859e-07, "loss": 2.0766, "step": 36104 }, { "epoch": 0.89, "learning_rate": 6.286019462835946e-07, "loss": 2.2649, "step": 36105 }, { "epoch": 0.89, "learning_rate": 6.283233952028778e-07, "loss": 2.0145, "step": 36106 }, { "epoch": 0.89, "learning_rate": 6.280449038509107e-07, "loss": 1.9608, "step": 36107 }, { "epoch": 0.89, "learning_rate": 6.277664722294663e-07, "loss": 2.0264, "step": 36108 }, { "epoch": 0.89, "learning_rate": 6.274881003403221e-07, "loss": 1.9026, "step": 36109 }, { "epoch": 0.89, "learning_rate": 6.27209788185249e-07, "loss": 2.0404, "step": 36110 }, { "epoch": 0.89, "learning_rate": 6.269315357660222e-07, "loss": 2.0607, "step": 36111 }, { "epoch": 0.89, "learning_rate": 6.266533430844168e-07, "loss": 1.8843, "step": 36112 }, { "epoch": 0.89, "learning_rate": 6.263752101422038e-07, "loss": 1.9021, "step": 36113 }, { "epoch": 0.89, "learning_rate": 6.260971369411573e-07, "loss": 1.9304, "step": 36114 }, { "epoch": 0.89, "learning_rate": 6.258191234830468e-07, "loss": 1.915, "step": 36115 }, { "epoch": 0.89, "learning_rate": 6.255411697696456e-07, "loss": 2.1246, "step": 36116 }, { "epoch": 0.89, "learning_rate": 6.252632758027266e-07, "loss": 2.0621, "step": 36117 }, { "epoch": 0.89, "learning_rate": 6.249854415840595e-07, "loss": 1.9897, "step": 36118 }, { "epoch": 0.89, "learning_rate": 6.247076671154162e-07, "loss": 2.0751, "step": 36119 }, { "epoch": 0.89, "learning_rate": 6.244299523985664e-07, "loss": 1.9496, "step": 36120 }, { "epoch": 0.89, "learning_rate": 6.241522974352788e-07, "loss": 1.8583, "step": 36121 }, { "epoch": 0.89, "learning_rate": 6.23874702227325e-07, "loss": 1.8313, "step": 36122 }, { "epoch": 0.89, "learning_rate": 6.235971667764728e-07, "loss": 1.9604, "step": 36123 }, { "epoch": 0.89, "learning_rate": 6.23319691084493e-07, "loss": 2.0149, "step": 36124 }, { "epoch": 0.89, "learning_rate": 6.230422751531506e-07, "loss": 1.8844, "step": 36125 }, { "epoch": 0.89, "learning_rate": 6.227649189842166e-07, "loss": 1.9678, "step": 36126 }, { "epoch": 0.89, "learning_rate": 6.224876225794585e-07, "loss": 1.9252, "step": 36127 }, { "epoch": 0.89, "learning_rate": 6.222103859406414e-07, "loss": 2.0712, "step": 36128 }, { "epoch": 0.89, "learning_rate": 6.219332090695352e-07, "loss": 1.9034, "step": 36129 }, { "epoch": 0.89, "learning_rate": 6.216560919679026e-07, "loss": 2.0014, "step": 36130 }, { "epoch": 0.89, "learning_rate": 6.213790346375126e-07, "loss": 1.8931, "step": 36131 }, { "epoch": 0.89, "learning_rate": 6.211020370801324e-07, "loss": 1.7685, "step": 36132 }, { "epoch": 0.89, "learning_rate": 6.20825099297524e-07, "loss": 1.9217, "step": 36133 }, { "epoch": 0.89, "learning_rate": 6.205482212914526e-07, "loss": 1.8445, "step": 36134 }, { "epoch": 0.89, "learning_rate": 6.202714030636847e-07, "loss": 1.8733, "step": 36135 }, { "epoch": 0.89, "learning_rate": 6.199946446159843e-07, "loss": 1.8294, "step": 36136 }, { "epoch": 0.89, "learning_rate": 6.197179459501168e-07, "loss": 1.9883, "step": 36137 }, { "epoch": 0.89, "learning_rate": 6.194413070678406e-07, "loss": 1.7843, "step": 36138 }, { "epoch": 0.89, "learning_rate": 6.191647279709256e-07, "loss": 1.8858, "step": 36139 }, { "epoch": 0.89, "learning_rate": 6.188882086611281e-07, "loss": 1.939, "step": 36140 }, { "epoch": 0.89, "learning_rate": 6.186117491402132e-07, "loss": 1.9738, "step": 36141 }, { "epoch": 0.89, "learning_rate": 6.183353494099453e-07, "loss": 1.8821, "step": 36142 }, { "epoch": 0.89, "learning_rate": 6.180590094720806e-07, "loss": 1.884, "step": 36143 }, { "epoch": 0.89, "learning_rate": 6.177827293283867e-07, "loss": 1.786, "step": 36144 }, { "epoch": 0.89, "learning_rate": 6.175065089806176e-07, "loss": 2.0365, "step": 36145 }, { "epoch": 0.89, "learning_rate": 6.172303484305387e-07, "loss": 1.8148, "step": 36146 }, { "epoch": 0.89, "learning_rate": 6.169542476799096e-07, "loss": 2.0309, "step": 36147 }, { "epoch": 0.89, "learning_rate": 6.166782067304878e-07, "loss": 1.9858, "step": 36148 }, { "epoch": 0.89, "learning_rate": 6.16402225584034e-07, "loss": 1.9046, "step": 36149 }, { "epoch": 0.89, "learning_rate": 6.16126304242306e-07, "loss": 1.9146, "step": 36150 }, { "epoch": 0.89, "learning_rate": 6.158504427070633e-07, "loss": 1.8355, "step": 36151 }, { "epoch": 0.89, "learning_rate": 6.155746409800644e-07, "loss": 1.8081, "step": 36152 }, { "epoch": 0.89, "learning_rate": 6.152988990630659e-07, "loss": 2.005, "step": 36153 }, { "epoch": 0.89, "learning_rate": 6.150232169578274e-07, "loss": 1.9513, "step": 36154 }, { "epoch": 0.89, "learning_rate": 6.147475946661041e-07, "loss": 1.6028, "step": 36155 }, { "epoch": 0.89, "learning_rate": 6.144720321896513e-07, "loss": 1.9124, "step": 36156 }, { "epoch": 0.89, "learning_rate": 6.141965295302287e-07, "loss": 1.8891, "step": 36157 }, { "epoch": 0.89, "learning_rate": 6.139210866895895e-07, "loss": 2.05, "step": 36158 }, { "epoch": 0.89, "learning_rate": 6.136457036694898e-07, "loss": 1.8796, "step": 36159 }, { "epoch": 0.89, "learning_rate": 6.133703804716873e-07, "loss": 1.9728, "step": 36160 }, { "epoch": 0.89, "learning_rate": 6.130951170979316e-07, "loss": 1.8527, "step": 36161 }, { "epoch": 0.89, "learning_rate": 6.128199135499813e-07, "loss": 1.9044, "step": 36162 }, { "epoch": 0.89, "learning_rate": 6.125447698295883e-07, "loss": 2.0397, "step": 36163 }, { "epoch": 0.89, "learning_rate": 6.122696859385069e-07, "loss": 1.8264, "step": 36164 }, { "epoch": 0.89, "learning_rate": 6.119946618784912e-07, "loss": 2.0112, "step": 36165 }, { "epoch": 0.89, "learning_rate": 6.117196976512907e-07, "loss": 2.0519, "step": 36166 }, { "epoch": 0.89, "learning_rate": 6.11444793258662e-07, "loss": 2.0243, "step": 36167 }, { "epoch": 0.89, "learning_rate": 6.111699487023548e-07, "loss": 1.9363, "step": 36168 }, { "epoch": 0.89, "learning_rate": 6.108951639841198e-07, "loss": 1.7949, "step": 36169 }, { "epoch": 0.89, "learning_rate": 6.106204391057102e-07, "loss": 1.949, "step": 36170 }, { "epoch": 0.89, "learning_rate": 6.103457740688767e-07, "loss": 2.033, "step": 36171 }, { "epoch": 0.89, "learning_rate": 6.100711688753713e-07, "loss": 1.8769, "step": 36172 }, { "epoch": 0.89, "learning_rate": 6.097966235269414e-07, "loss": 1.7973, "step": 36173 }, { "epoch": 0.89, "learning_rate": 6.095221380253369e-07, "loss": 1.8952, "step": 36174 }, { "epoch": 0.89, "learning_rate": 6.092477123723095e-07, "loss": 1.8457, "step": 36175 }, { "epoch": 0.89, "learning_rate": 6.089733465696057e-07, "loss": 1.9609, "step": 36176 }, { "epoch": 0.89, "learning_rate": 6.086990406189763e-07, "loss": 1.945, "step": 36177 }, { "epoch": 0.89, "learning_rate": 6.084247945221678e-07, "loss": 1.9787, "step": 36178 }, { "epoch": 0.89, "learning_rate": 6.081506082809274e-07, "loss": 1.9763, "step": 36179 }, { "epoch": 0.89, "learning_rate": 6.078764818970063e-07, "loss": 1.7879, "step": 36180 }, { "epoch": 0.89, "learning_rate": 6.076024153721483e-07, "loss": 1.8282, "step": 36181 }, { "epoch": 0.89, "learning_rate": 6.07328408708101e-07, "loss": 1.909, "step": 36182 }, { "epoch": 0.89, "learning_rate": 6.070544619066099e-07, "loss": 1.962, "step": 36183 }, { "epoch": 0.89, "learning_rate": 6.067805749694222e-07, "loss": 1.8956, "step": 36184 }, { "epoch": 0.89, "learning_rate": 6.065067478982844e-07, "loss": 2.0156, "step": 36185 }, { "epoch": 0.89, "learning_rate": 6.062329806949386e-07, "loss": 1.8982, "step": 36186 }, { "epoch": 0.89, "learning_rate": 6.059592733611319e-07, "loss": 1.7955, "step": 36187 }, { "epoch": 0.89, "learning_rate": 6.056856258986076e-07, "loss": 1.9209, "step": 36188 }, { "epoch": 0.89, "learning_rate": 6.0541203830911e-07, "loss": 1.9975, "step": 36189 }, { "epoch": 0.89, "learning_rate": 6.051385105943852e-07, "loss": 1.9363, "step": 36190 }, { "epoch": 0.89, "learning_rate": 6.04865042756172e-07, "loss": 2.0909, "step": 36191 }, { "epoch": 0.89, "learning_rate": 6.045916347962188e-07, "loss": 1.9116, "step": 36192 }, { "epoch": 0.89, "learning_rate": 6.043182867162622e-07, "loss": 2.0175, "step": 36193 }, { "epoch": 0.89, "learning_rate": 6.040449985180474e-07, "loss": 1.8564, "step": 36194 }, { "epoch": 0.89, "learning_rate": 6.037717702033175e-07, "loss": 1.6148, "step": 36195 }, { "epoch": 0.89, "learning_rate": 6.034986017738109e-07, "loss": 1.8176, "step": 36196 }, { "epoch": 0.89, "learning_rate": 6.032254932312709e-07, "loss": 1.9742, "step": 36197 }, { "epoch": 0.89, "learning_rate": 6.02952444577436e-07, "loss": 1.9323, "step": 36198 }, { "epoch": 0.89, "learning_rate": 6.026794558140481e-07, "loss": 1.8981, "step": 36199 }, { "epoch": 0.89, "learning_rate": 6.02406526942847e-07, "loss": 1.9286, "step": 36200 }, { "epoch": 0.89, "learning_rate": 6.021336579655701e-07, "loss": 1.8286, "step": 36201 }, { "epoch": 0.89, "learning_rate": 6.018608488839606e-07, "loss": 1.916, "step": 36202 }, { "epoch": 0.89, "learning_rate": 6.015880996997526e-07, "loss": 1.9054, "step": 36203 }, { "epoch": 0.89, "learning_rate": 6.013154104146868e-07, "loss": 1.7703, "step": 36204 }, { "epoch": 0.89, "learning_rate": 6.010427810305008e-07, "loss": 1.9455, "step": 36205 }, { "epoch": 0.89, "learning_rate": 6.007702115489323e-07, "loss": 1.7681, "step": 36206 }, { "epoch": 0.89, "learning_rate": 6.004977019717195e-07, "loss": 1.9988, "step": 36207 }, { "epoch": 0.89, "learning_rate": 6.00225252300597e-07, "loss": 1.9707, "step": 36208 }, { "epoch": 0.89, "learning_rate": 5.999528625373008e-07, "loss": 2.1484, "step": 36209 }, { "epoch": 0.89, "learning_rate": 5.99680532683572e-07, "loss": 1.9215, "step": 36210 }, { "epoch": 0.89, "learning_rate": 5.994082627411402e-07, "loss": 1.9618, "step": 36211 }, { "epoch": 0.89, "learning_rate": 5.99136052711744e-07, "loss": 2.1445, "step": 36212 }, { "epoch": 0.89, "learning_rate": 5.988639025971165e-07, "loss": 2.0539, "step": 36213 }, { "epoch": 0.89, "learning_rate": 5.985918123989931e-07, "loss": 1.7733, "step": 36214 }, { "epoch": 0.89, "learning_rate": 5.98319782119109e-07, "loss": 1.9136, "step": 36215 }, { "epoch": 0.89, "learning_rate": 5.980478117591959e-07, "loss": 1.8554, "step": 36216 }, { "epoch": 0.89, "learning_rate": 5.977759013209883e-07, "loss": 1.802, "step": 36217 }, { "epoch": 0.89, "learning_rate": 5.975040508062202e-07, "loss": 1.8956, "step": 36218 }, { "epoch": 0.89, "learning_rate": 5.972322602166214e-07, "loss": 1.7883, "step": 36219 }, { "epoch": 0.89, "learning_rate": 5.969605295539272e-07, "loss": 1.8923, "step": 36220 }, { "epoch": 0.89, "learning_rate": 5.96688858819866e-07, "loss": 1.9655, "step": 36221 }, { "epoch": 0.89, "learning_rate": 5.964172480161711e-07, "loss": 1.9911, "step": 36222 }, { "epoch": 0.89, "learning_rate": 5.961456971445745e-07, "loss": 1.885, "step": 36223 }, { "epoch": 0.89, "learning_rate": 5.958742062068056e-07, "loss": 1.8792, "step": 36224 }, { "epoch": 0.89, "learning_rate": 5.956027752045968e-07, "loss": 1.9046, "step": 36225 }, { "epoch": 0.89, "learning_rate": 5.953314041396741e-07, "loss": 1.9286, "step": 36226 }, { "epoch": 0.89, "learning_rate": 5.950600930137696e-07, "loss": 1.8308, "step": 36227 }, { "epoch": 0.89, "learning_rate": 5.947888418286141e-07, "loss": 2.0017, "step": 36228 }, { "epoch": 0.89, "learning_rate": 5.945176505859318e-07, "loss": 1.9642, "step": 36229 }, { "epoch": 0.89, "learning_rate": 5.942465192874558e-07, "loss": 1.995, "step": 36230 }, { "epoch": 0.89, "learning_rate": 5.939754479349102e-07, "loss": 1.883, "step": 36231 }, { "epoch": 0.89, "learning_rate": 5.937044365300249e-07, "loss": 1.8457, "step": 36232 }, { "epoch": 0.89, "learning_rate": 5.934334850745282e-07, "loss": 2.024, "step": 36233 }, { "epoch": 0.89, "learning_rate": 5.931625935701435e-07, "loss": 1.8476, "step": 36234 }, { "epoch": 0.89, "learning_rate": 5.928917620186014e-07, "loss": 1.9372, "step": 36235 }, { "epoch": 0.89, "learning_rate": 5.926209904216241e-07, "loss": 2.0039, "step": 36236 }, { "epoch": 0.89, "learning_rate": 5.923502787809387e-07, "loss": 1.8004, "step": 36237 }, { "epoch": 0.89, "learning_rate": 5.92079627098272e-07, "loss": 2.0098, "step": 36238 }, { "epoch": 0.89, "learning_rate": 5.918090353753469e-07, "loss": 1.8943, "step": 36239 }, { "epoch": 0.89, "learning_rate": 5.915385036138876e-07, "loss": 1.8207, "step": 36240 }, { "epoch": 0.89, "learning_rate": 5.912680318156205e-07, "loss": 1.8389, "step": 36241 }, { "epoch": 0.89, "learning_rate": 5.909976199822687e-07, "loss": 1.9895, "step": 36242 }, { "epoch": 0.89, "learning_rate": 5.907272681155574e-07, "loss": 1.8036, "step": 36243 }, { "epoch": 0.89, "learning_rate": 5.904569762172052e-07, "loss": 1.757, "step": 36244 }, { "epoch": 0.89, "learning_rate": 5.901867442889387e-07, "loss": 1.9189, "step": 36245 }, { "epoch": 0.89, "learning_rate": 5.899165723324773e-07, "loss": 1.9586, "step": 36246 }, { "epoch": 0.89, "learning_rate": 5.896464603495445e-07, "loss": 1.8187, "step": 36247 }, { "epoch": 0.89, "learning_rate": 5.893764083418619e-07, "loss": 1.8968, "step": 36248 }, { "epoch": 0.89, "learning_rate": 5.891064163111504e-07, "loss": 2.0139, "step": 36249 }, { "epoch": 0.89, "learning_rate": 5.88836484259131e-07, "loss": 1.9447, "step": 36250 }, { "epoch": 0.89, "learning_rate": 5.885666121875222e-07, "loss": 1.9075, "step": 36251 }, { "epoch": 0.89, "learning_rate": 5.882968000980461e-07, "loss": 1.6433, "step": 36252 }, { "epoch": 0.89, "learning_rate": 5.880270479924233e-07, "loss": 2.0306, "step": 36253 }, { "epoch": 0.89, "learning_rate": 5.877573558723693e-07, "loss": 1.7893, "step": 36254 }, { "epoch": 0.89, "learning_rate": 5.874877237396071e-07, "loss": 1.9352, "step": 36255 }, { "epoch": 0.89, "learning_rate": 5.872181515958508e-07, "loss": 1.8126, "step": 36256 }, { "epoch": 0.89, "learning_rate": 5.869486394428214e-07, "loss": 1.9081, "step": 36257 }, { "epoch": 0.89, "learning_rate": 5.866791872822353e-07, "loss": 1.9277, "step": 36258 }, { "epoch": 0.89, "learning_rate": 5.86409795115811e-07, "loss": 1.8843, "step": 36259 }, { "epoch": 0.89, "learning_rate": 5.86140462945266e-07, "loss": 2.0501, "step": 36260 }, { "epoch": 0.89, "learning_rate": 5.858711907723147e-07, "loss": 1.8174, "step": 36261 }, { "epoch": 0.89, "learning_rate": 5.856019785986744e-07, "loss": 1.9873, "step": 36262 }, { "epoch": 0.89, "learning_rate": 5.853328264260616e-07, "loss": 2.1412, "step": 36263 }, { "epoch": 0.89, "learning_rate": 5.850637342561905e-07, "loss": 1.9569, "step": 36264 }, { "epoch": 0.89, "learning_rate": 5.847947020907763e-07, "loss": 1.9685, "step": 36265 }, { "epoch": 0.89, "learning_rate": 5.845257299315332e-07, "loss": 2.0596, "step": 36266 }, { "epoch": 0.89, "learning_rate": 5.842568177801766e-07, "loss": 2.0649, "step": 36267 }, { "epoch": 0.89, "learning_rate": 5.839879656384206e-07, "loss": 1.9284, "step": 36268 }, { "epoch": 0.89, "learning_rate": 5.837191735079761e-07, "loss": 1.8432, "step": 36269 }, { "epoch": 0.89, "learning_rate": 5.834504413905595e-07, "loss": 1.9767, "step": 36270 }, { "epoch": 0.89, "learning_rate": 5.831817692878816e-07, "loss": 2.1282, "step": 36271 }, { "epoch": 0.89, "learning_rate": 5.829131572016544e-07, "loss": 1.9169, "step": 36272 }, { "epoch": 0.89, "learning_rate": 5.826446051335921e-07, "loss": 1.9892, "step": 36273 }, { "epoch": 0.89, "learning_rate": 5.823761130854034e-07, "loss": 1.9332, "step": 36274 }, { "epoch": 0.89, "learning_rate": 5.821076810588e-07, "loss": 1.8599, "step": 36275 }, { "epoch": 0.89, "learning_rate": 5.818393090554942e-07, "loss": 1.9302, "step": 36276 }, { "epoch": 0.89, "learning_rate": 5.815709970771966e-07, "loss": 1.8384, "step": 36277 }, { "epoch": 0.89, "learning_rate": 5.81302745125617e-07, "loss": 1.9305, "step": 36278 }, { "epoch": 0.89, "learning_rate": 5.81034553202463e-07, "loss": 2.0011, "step": 36279 }, { "epoch": 0.89, "learning_rate": 5.807664213094455e-07, "loss": 1.896, "step": 36280 }, { "epoch": 0.89, "learning_rate": 5.804983494482741e-07, "loss": 1.9768, "step": 36281 }, { "epoch": 0.89, "learning_rate": 5.802303376206552e-07, "loss": 1.985, "step": 36282 }, { "epoch": 0.89, "learning_rate": 5.799623858282998e-07, "loss": 2.054, "step": 36283 }, { "epoch": 0.89, "learning_rate": 5.796944940729132e-07, "loss": 1.8791, "step": 36284 }, { "epoch": 0.89, "learning_rate": 5.794266623562028e-07, "loss": 1.9328, "step": 36285 }, { "epoch": 0.89, "learning_rate": 5.791588906798784e-07, "loss": 2.0009, "step": 36286 }, { "epoch": 0.89, "learning_rate": 5.788911790456431e-07, "loss": 1.8955, "step": 36287 }, { "epoch": 0.89, "learning_rate": 5.786235274552054e-07, "loss": 1.8442, "step": 36288 }, { "epoch": 0.89, "learning_rate": 5.783559359102698e-07, "loss": 1.7982, "step": 36289 }, { "epoch": 0.89, "learning_rate": 5.780884044125412e-07, "loss": 1.7916, "step": 36290 }, { "epoch": 0.89, "learning_rate": 5.778209329637274e-07, "loss": 1.8807, "step": 36291 }, { "epoch": 0.89, "learning_rate": 5.775535215655304e-07, "loss": 1.7988, "step": 36292 }, { "epoch": 0.89, "learning_rate": 5.772861702196553e-07, "loss": 1.9651, "step": 36293 }, { "epoch": 0.89, "learning_rate": 5.770188789278063e-07, "loss": 1.7926, "step": 36294 }, { "epoch": 0.89, "learning_rate": 5.767516476916868e-07, "loss": 1.8278, "step": 36295 }, { "epoch": 0.89, "learning_rate": 5.764844765130007e-07, "loss": 1.9261, "step": 36296 }, { "epoch": 0.89, "learning_rate": 5.762173653934489e-07, "loss": 1.8552, "step": 36297 }, { "epoch": 0.89, "learning_rate": 5.759503143347367e-07, "loss": 1.9028, "step": 36298 }, { "epoch": 0.89, "learning_rate": 5.75683323338564e-07, "loss": 2.0484, "step": 36299 }, { "epoch": 0.89, "learning_rate": 5.754163924066314e-07, "loss": 1.9349, "step": 36300 }, { "epoch": 0.89, "learning_rate": 5.751495215406433e-07, "loss": 1.8185, "step": 36301 }, { "epoch": 0.89, "learning_rate": 5.748827107422972e-07, "loss": 1.7641, "step": 36302 }, { "epoch": 0.89, "learning_rate": 5.746159600132961e-07, "loss": 1.9388, "step": 36303 }, { "epoch": 0.89, "learning_rate": 5.743492693553387e-07, "loss": 1.6696, "step": 36304 }, { "epoch": 0.89, "learning_rate": 5.740826387701248e-07, "loss": 1.9023, "step": 36305 }, { "epoch": 0.89, "learning_rate": 5.738160682593552e-07, "loss": 2.1592, "step": 36306 }, { "epoch": 0.89, "learning_rate": 5.735495578247263e-07, "loss": 2.0134, "step": 36307 }, { "epoch": 0.89, "learning_rate": 5.732831074679402e-07, "loss": 1.8005, "step": 36308 }, { "epoch": 0.89, "learning_rate": 5.730167171906909e-07, "loss": 1.8929, "step": 36309 }, { "epoch": 0.89, "learning_rate": 5.727503869946783e-07, "loss": 1.9632, "step": 36310 }, { "epoch": 0.9, "learning_rate": 5.724841168815997e-07, "loss": 2.0028, "step": 36311 }, { "epoch": 0.9, "learning_rate": 5.72217906853152e-07, "loss": 2.0347, "step": 36312 }, { "epoch": 0.9, "learning_rate": 5.719517569110334e-07, "loss": 1.9251, "step": 36313 }, { "epoch": 0.9, "learning_rate": 5.716856670569381e-07, "loss": 1.9912, "step": 36314 }, { "epoch": 0.9, "learning_rate": 5.714196372925618e-07, "loss": 1.876, "step": 36315 }, { "epoch": 0.9, "learning_rate": 5.711536676196028e-07, "loss": 1.8626, "step": 36316 }, { "epoch": 0.9, "learning_rate": 5.708877580397521e-07, "loss": 2.0236, "step": 36317 }, { "epoch": 0.9, "learning_rate": 5.706219085547093e-07, "loss": 1.8661, "step": 36318 }, { "epoch": 0.9, "learning_rate": 5.703561191661633e-07, "loss": 1.868, "step": 36319 }, { "epoch": 0.9, "learning_rate": 5.700903898758114e-07, "loss": 1.9803, "step": 36320 }, { "epoch": 0.9, "learning_rate": 5.69824720685348e-07, "loss": 1.9471, "step": 36321 }, { "epoch": 0.9, "learning_rate": 5.695591115964638e-07, "loss": 1.9234, "step": 36322 }, { "epoch": 0.9, "learning_rate": 5.692935626108531e-07, "loss": 1.6923, "step": 36323 }, { "epoch": 0.9, "learning_rate": 5.690280737302067e-07, "loss": 1.9729, "step": 36324 }, { "epoch": 0.9, "learning_rate": 5.68762644956219e-07, "loss": 1.8485, "step": 36325 }, { "epoch": 0.9, "learning_rate": 5.684972762905805e-07, "loss": 1.9836, "step": 36326 }, { "epoch": 0.9, "learning_rate": 5.682319677349812e-07, "loss": 1.8389, "step": 36327 }, { "epoch": 0.9, "learning_rate": 5.679667192911142e-07, "loss": 1.9879, "step": 36328 }, { "epoch": 0.9, "learning_rate": 5.677015309606692e-07, "loss": 1.8731, "step": 36329 }, { "epoch": 0.9, "learning_rate": 5.67436402745335e-07, "loss": 2.0824, "step": 36330 }, { "epoch": 0.9, "learning_rate": 5.671713346468055e-07, "loss": 1.8335, "step": 36331 }, { "epoch": 0.9, "learning_rate": 5.66906326666764e-07, "loss": 1.9095, "step": 36332 }, { "epoch": 0.9, "learning_rate": 5.666413788069037e-07, "loss": 1.8513, "step": 36333 }, { "epoch": 0.9, "learning_rate": 5.66376491068914e-07, "loss": 1.9663, "step": 36334 }, { "epoch": 0.9, "learning_rate": 5.661116634544795e-07, "loss": 1.8618, "step": 36335 }, { "epoch": 0.9, "learning_rate": 5.658468959652907e-07, "loss": 2.053, "step": 36336 }, { "epoch": 0.9, "learning_rate": 5.655821886030333e-07, "loss": 1.8916, "step": 36337 }, { "epoch": 0.9, "learning_rate": 5.653175413693946e-07, "loss": 1.9784, "step": 36338 }, { "epoch": 0.9, "learning_rate": 5.650529542660632e-07, "loss": 2.122, "step": 36339 }, { "epoch": 0.9, "learning_rate": 5.647884272947234e-07, "loss": 1.9727, "step": 36340 }, { "epoch": 0.9, "learning_rate": 5.645239604570629e-07, "loss": 1.8445, "step": 36341 }, { "epoch": 0.9, "learning_rate": 5.642595537547646e-07, "loss": 1.7962, "step": 36342 }, { "epoch": 0.9, "learning_rate": 5.639952071895149e-07, "loss": 1.9023, "step": 36343 }, { "epoch": 0.9, "learning_rate": 5.637309207630015e-07, "loss": 2.0161, "step": 36344 }, { "epoch": 0.9, "learning_rate": 5.634666944769029e-07, "loss": 1.8633, "step": 36345 }, { "epoch": 0.9, "learning_rate": 5.632025283329079e-07, "loss": 1.9165, "step": 36346 }, { "epoch": 0.9, "learning_rate": 5.629384223326973e-07, "loss": 2.0561, "step": 36347 }, { "epoch": 0.9, "learning_rate": 5.626743764779563e-07, "loss": 1.978, "step": 36348 }, { "epoch": 0.9, "learning_rate": 5.624103907703693e-07, "loss": 1.9491, "step": 36349 }, { "epoch": 0.9, "learning_rate": 5.621464652116138e-07, "loss": 1.7945, "step": 36350 }, { "epoch": 0.9, "learning_rate": 5.618825998033761e-07, "loss": 1.9884, "step": 36351 }, { "epoch": 0.9, "learning_rate": 5.61618794547335e-07, "loss": 1.8616, "step": 36352 }, { "epoch": 0.9, "learning_rate": 5.613550494451737e-07, "loss": 1.9447, "step": 36353 }, { "epoch": 0.9, "learning_rate": 5.610913644985738e-07, "loss": 1.8259, "step": 36354 }, { "epoch": 0.9, "learning_rate": 5.608277397092132e-07, "loss": 1.8376, "step": 36355 }, { "epoch": 0.9, "learning_rate": 5.60564175078775e-07, "loss": 1.9936, "step": 36356 }, { "epoch": 0.9, "learning_rate": 5.603006706089365e-07, "loss": 1.9726, "step": 36357 }, { "epoch": 0.9, "learning_rate": 5.600372263013776e-07, "loss": 1.9416, "step": 36358 }, { "epoch": 0.9, "learning_rate": 5.597738421577792e-07, "loss": 1.9654, "step": 36359 }, { "epoch": 0.9, "learning_rate": 5.595105181798177e-07, "loss": 1.9238, "step": 36360 }, { "epoch": 0.9, "learning_rate": 5.592472543691729e-07, "loss": 2.0354, "step": 36361 }, { "epoch": 0.9, "learning_rate": 5.589840507275213e-07, "loss": 1.7368, "step": 36362 }, { "epoch": 0.9, "learning_rate": 5.587209072565414e-07, "loss": 2.0025, "step": 36363 }, { "epoch": 0.9, "learning_rate": 5.584578239579086e-07, "loss": 1.8798, "step": 36364 }, { "epoch": 0.9, "learning_rate": 5.581948008333026e-07, "loss": 1.694, "step": 36365 }, { "epoch": 0.9, "learning_rate": 5.579318378843979e-07, "loss": 1.8977, "step": 36366 }, { "epoch": 0.9, "learning_rate": 5.576689351128706e-07, "loss": 2.0431, "step": 36367 }, { "epoch": 0.9, "learning_rate": 5.574060925203961e-07, "loss": 1.8477, "step": 36368 }, { "epoch": 0.9, "learning_rate": 5.571433101086509e-07, "loss": 1.7168, "step": 36369 }, { "epoch": 0.9, "learning_rate": 5.568805878793071e-07, "loss": 1.928, "step": 36370 }, { "epoch": 0.9, "learning_rate": 5.566179258340432e-07, "loss": 1.8204, "step": 36371 }, { "epoch": 0.9, "learning_rate": 5.563553239745278e-07, "loss": 1.9925, "step": 36372 }, { "epoch": 0.9, "learning_rate": 5.560927823024387e-07, "loss": 1.8719, "step": 36373 }, { "epoch": 0.9, "learning_rate": 5.558303008194487e-07, "loss": 1.8884, "step": 36374 }, { "epoch": 0.9, "learning_rate": 5.555678795272291e-07, "loss": 2.0454, "step": 36375 }, { "epoch": 0.9, "learning_rate": 5.553055184274548e-07, "loss": 2.1198, "step": 36376 }, { "epoch": 0.9, "learning_rate": 5.550432175217946e-07, "loss": 1.8755, "step": 36377 }, { "epoch": 0.9, "learning_rate": 5.547809768119216e-07, "loss": 1.8742, "step": 36378 }, { "epoch": 0.9, "learning_rate": 5.545187962995091e-07, "loss": 1.7748, "step": 36379 }, { "epoch": 0.9, "learning_rate": 5.542566759862256e-07, "loss": 1.7911, "step": 36380 }, { "epoch": 0.9, "learning_rate": 5.53994615873742e-07, "loss": 1.8027, "step": 36381 }, { "epoch": 0.9, "learning_rate": 5.537326159637302e-07, "loss": 1.8324, "step": 36382 }, { "epoch": 0.9, "learning_rate": 5.534706762578579e-07, "loss": 1.9637, "step": 36383 }, { "epoch": 0.9, "learning_rate": 5.532087967577971e-07, "loss": 1.977, "step": 36384 }, { "epoch": 0.9, "learning_rate": 5.529469774652141e-07, "loss": 2.0202, "step": 36385 }, { "epoch": 0.9, "learning_rate": 5.5268521838178e-07, "loss": 1.8942, "step": 36386 }, { "epoch": 0.9, "learning_rate": 5.524235195091598e-07, "loss": 1.9429, "step": 36387 }, { "epoch": 0.9, "learning_rate": 5.521618808490237e-07, "loss": 1.975, "step": 36388 }, { "epoch": 0.9, "learning_rate": 5.519003024030411e-07, "loss": 1.9788, "step": 36389 }, { "epoch": 0.9, "learning_rate": 5.516387841728744e-07, "loss": 1.848, "step": 36390 }, { "epoch": 0.9, "learning_rate": 5.513773261601952e-07, "loss": 1.8845, "step": 36391 }, { "epoch": 0.9, "learning_rate": 5.511159283666656e-07, "loss": 1.9171, "step": 36392 }, { "epoch": 0.9, "learning_rate": 5.508545907939533e-07, "loss": 2.017, "step": 36393 }, { "epoch": 0.9, "learning_rate": 5.505933134437258e-07, "loss": 1.8726, "step": 36394 }, { "epoch": 0.9, "learning_rate": 5.503320963176451e-07, "loss": 1.9677, "step": 36395 }, { "epoch": 0.9, "learning_rate": 5.500709394173776e-07, "loss": 1.8756, "step": 36396 }, { "epoch": 0.9, "learning_rate": 5.498098427445886e-07, "loss": 1.949, "step": 36397 }, { "epoch": 0.9, "learning_rate": 5.495488063009402e-07, "loss": 1.8822, "step": 36398 }, { "epoch": 0.9, "learning_rate": 5.492878300880978e-07, "loss": 1.8963, "step": 36399 }, { "epoch": 0.9, "learning_rate": 5.490269141077232e-07, "loss": 1.9814, "step": 36400 }, { "epoch": 0.9, "learning_rate": 5.487660583614795e-07, "loss": 1.9244, "step": 36401 }, { "epoch": 0.9, "learning_rate": 5.485052628510324e-07, "loss": 2.002, "step": 36402 }, { "epoch": 0.9, "learning_rate": 5.482445275780391e-07, "loss": 1.9921, "step": 36403 }, { "epoch": 0.9, "learning_rate": 5.479838525441649e-07, "loss": 1.9778, "step": 36404 }, { "epoch": 0.9, "learning_rate": 5.477232377510689e-07, "loss": 1.6222, "step": 36405 }, { "epoch": 0.9, "learning_rate": 5.474626832004137e-07, "loss": 1.9084, "step": 36406 }, { "epoch": 0.9, "learning_rate": 5.472021888938595e-07, "loss": 2.0052, "step": 36407 }, { "epoch": 0.9, "learning_rate": 5.469417548330658e-07, "loss": 1.9562, "step": 36408 }, { "epoch": 0.9, "learning_rate": 5.466813810196947e-07, "loss": 1.7655, "step": 36409 }, { "epoch": 0.9, "learning_rate": 5.464210674554016e-07, "loss": 1.9249, "step": 36410 }, { "epoch": 0.9, "learning_rate": 5.461608141418484e-07, "loss": 1.9988, "step": 36411 }, { "epoch": 0.9, "learning_rate": 5.459006210806938e-07, "loss": 2.0461, "step": 36412 }, { "epoch": 0.9, "learning_rate": 5.456404882735955e-07, "loss": 2.0213, "step": 36413 }, { "epoch": 0.9, "learning_rate": 5.453804157222109e-07, "loss": 1.9049, "step": 36414 }, { "epoch": 0.9, "learning_rate": 5.451204034281987e-07, "loss": 1.8204, "step": 36415 }, { "epoch": 0.9, "learning_rate": 5.448604513932154e-07, "loss": 2.0209, "step": 36416 }, { "epoch": 0.9, "learning_rate": 5.446005596189163e-07, "loss": 1.9599, "step": 36417 }, { "epoch": 0.9, "learning_rate": 5.443407281069601e-07, "loss": 1.9483, "step": 36418 }, { "epoch": 0.9, "learning_rate": 5.440809568590033e-07, "loss": 1.8517, "step": 36419 }, { "epoch": 0.9, "learning_rate": 5.43821245876699e-07, "loss": 2.0363, "step": 36420 }, { "epoch": 0.9, "learning_rate": 5.435615951617035e-07, "loss": 1.9996, "step": 36421 }, { "epoch": 0.9, "learning_rate": 5.433020047156734e-07, "loss": 1.8698, "step": 36422 }, { "epoch": 0.9, "learning_rate": 5.430424745402596e-07, "loss": 1.7912, "step": 36423 }, { "epoch": 0.9, "learning_rate": 5.427830046371196e-07, "loss": 1.9699, "step": 36424 }, { "epoch": 0.9, "learning_rate": 5.425235950079033e-07, "loss": 1.9253, "step": 36425 }, { "epoch": 0.9, "learning_rate": 5.422642456542671e-07, "loss": 1.9359, "step": 36426 }, { "epoch": 0.9, "learning_rate": 5.42004956577864e-07, "loss": 1.8696, "step": 36427 }, { "epoch": 0.9, "learning_rate": 5.417457277803451e-07, "loss": 1.9106, "step": 36428 }, { "epoch": 0.9, "learning_rate": 5.414865592633634e-07, "loss": 1.9101, "step": 36429 }, { "epoch": 0.9, "learning_rate": 5.412274510285697e-07, "loss": 1.9589, "step": 36430 }, { "epoch": 0.9, "learning_rate": 5.409684030776153e-07, "loss": 1.886, "step": 36431 }, { "epoch": 0.9, "learning_rate": 5.407094154121518e-07, "loss": 1.8283, "step": 36432 }, { "epoch": 0.9, "learning_rate": 5.404504880338324e-07, "loss": 1.9254, "step": 36433 }, { "epoch": 0.9, "learning_rate": 5.401916209443025e-07, "loss": 1.9477, "step": 36434 }, { "epoch": 0.9, "learning_rate": 5.399328141452143e-07, "loss": 2.0656, "step": 36435 }, { "epoch": 0.9, "learning_rate": 5.396740676382172e-07, "loss": 1.9924, "step": 36436 }, { "epoch": 0.9, "learning_rate": 5.394153814249614e-07, "loss": 1.8432, "step": 36437 }, { "epoch": 0.9, "learning_rate": 5.391567555070943e-07, "loss": 1.8316, "step": 36438 }, { "epoch": 0.9, "learning_rate": 5.388981898862644e-07, "loss": 2.0276, "step": 36439 }, { "epoch": 0.9, "learning_rate": 5.386396845641195e-07, "loss": 1.959, "step": 36440 }, { "epoch": 0.9, "learning_rate": 5.38381239542306e-07, "loss": 1.9661, "step": 36441 }, { "epoch": 0.9, "learning_rate": 5.381228548224748e-07, "loss": 1.9629, "step": 36442 }, { "epoch": 0.9, "learning_rate": 5.37864530406268e-07, "loss": 2.1218, "step": 36443 }, { "epoch": 0.9, "learning_rate": 5.376062662953363e-07, "loss": 2.0176, "step": 36444 }, { "epoch": 0.9, "learning_rate": 5.373480624913219e-07, "loss": 1.9135, "step": 36445 }, { "epoch": 0.9, "learning_rate": 5.370899189958711e-07, "loss": 2.1508, "step": 36446 }, { "epoch": 0.9, "learning_rate": 5.368318358106328e-07, "loss": 2.0957, "step": 36447 }, { "epoch": 0.9, "learning_rate": 5.365738129372477e-07, "loss": 1.952, "step": 36448 }, { "epoch": 0.9, "learning_rate": 5.363158503773613e-07, "loss": 2.0496, "step": 36449 }, { "epoch": 0.9, "learning_rate": 5.360579481326178e-07, "loss": 1.9917, "step": 36450 }, { "epoch": 0.9, "learning_rate": 5.358001062046625e-07, "loss": 1.8405, "step": 36451 }, { "epoch": 0.9, "learning_rate": 5.355423245951374e-07, "loss": 1.9427, "step": 36452 }, { "epoch": 0.9, "learning_rate": 5.352846033056835e-07, "loss": 1.9706, "step": 36453 }, { "epoch": 0.9, "learning_rate": 5.350269423379473e-07, "loss": 1.9212, "step": 36454 }, { "epoch": 0.9, "learning_rate": 5.347693416935684e-07, "loss": 1.9987, "step": 36455 }, { "epoch": 0.9, "learning_rate": 5.34511801374189e-07, "loss": 2.0325, "step": 36456 }, { "epoch": 0.9, "learning_rate": 5.342543213814522e-07, "loss": 2.0572, "step": 36457 }, { "epoch": 0.9, "learning_rate": 5.339969017169955e-07, "loss": 1.8322, "step": 36458 }, { "epoch": 0.9, "learning_rate": 5.337395423824609e-07, "loss": 1.7664, "step": 36459 }, { "epoch": 0.9, "learning_rate": 5.334822433794917e-07, "loss": 2.0512, "step": 36460 }, { "epoch": 0.9, "learning_rate": 5.332250047097232e-07, "loss": 1.7938, "step": 36461 }, { "epoch": 0.9, "learning_rate": 5.329678263747984e-07, "loss": 1.8416, "step": 36462 }, { "epoch": 0.9, "learning_rate": 5.327107083763527e-07, "loss": 1.9708, "step": 36463 }, { "epoch": 0.9, "learning_rate": 5.324536507160272e-07, "loss": 1.9223, "step": 36464 }, { "epoch": 0.9, "learning_rate": 5.321966533954625e-07, "loss": 1.992, "step": 36465 }, { "epoch": 0.9, "learning_rate": 5.31939716416292e-07, "loss": 1.9196, "step": 36466 }, { "epoch": 0.9, "learning_rate": 5.316828397801543e-07, "loss": 1.9343, "step": 36467 }, { "epoch": 0.9, "learning_rate": 5.314260234886903e-07, "loss": 1.8642, "step": 36468 }, { "epoch": 0.9, "learning_rate": 5.31169267543532e-07, "loss": 2.1182, "step": 36469 }, { "epoch": 0.9, "learning_rate": 5.309125719463171e-07, "loss": 2.0007, "step": 36470 }, { "epoch": 0.9, "learning_rate": 5.30655936698683e-07, "loss": 1.8736, "step": 36471 }, { "epoch": 0.9, "learning_rate": 5.303993618022662e-07, "loss": 2.0675, "step": 36472 }, { "epoch": 0.9, "learning_rate": 5.301428472586989e-07, "loss": 1.805, "step": 36473 }, { "epoch": 0.9, "learning_rate": 5.298863930696163e-07, "loss": 1.9213, "step": 36474 }, { "epoch": 0.9, "learning_rate": 5.29629999236656e-07, "loss": 2.0038, "step": 36475 }, { "epoch": 0.9, "learning_rate": 5.29373665761449e-07, "loss": 2.078, "step": 36476 }, { "epoch": 0.9, "learning_rate": 5.291173926456316e-07, "loss": 1.9679, "step": 36477 }, { "epoch": 0.9, "learning_rate": 5.288611798908327e-07, "loss": 1.8746, "step": 36478 }, { "epoch": 0.9, "learning_rate": 5.286050274986898e-07, "loss": 1.8302, "step": 36479 }, { "epoch": 0.9, "learning_rate": 5.283489354708338e-07, "loss": 2.0878, "step": 36480 }, { "epoch": 0.9, "learning_rate": 5.280929038088966e-07, "loss": 1.9977, "step": 36481 }, { "epoch": 0.9, "learning_rate": 5.278369325145106e-07, "loss": 1.8052, "step": 36482 }, { "epoch": 0.9, "learning_rate": 5.275810215893051e-07, "loss": 1.884, "step": 36483 }, { "epoch": 0.9, "learning_rate": 5.273251710349137e-07, "loss": 1.9719, "step": 36484 }, { "epoch": 0.9, "learning_rate": 5.270693808529648e-07, "loss": 1.978, "step": 36485 }, { "epoch": 0.9, "learning_rate": 5.268136510450927e-07, "loss": 2.0649, "step": 36486 }, { "epoch": 0.9, "learning_rate": 5.265579816129218e-07, "loss": 2.1075, "step": 36487 }, { "epoch": 0.9, "learning_rate": 5.263023725580851e-07, "loss": 1.8233, "step": 36488 }, { "epoch": 0.9, "learning_rate": 5.260468238822103e-07, "loss": 1.9222, "step": 36489 }, { "epoch": 0.9, "learning_rate": 5.257913355869282e-07, "loss": 1.9515, "step": 36490 }, { "epoch": 0.9, "learning_rate": 5.255359076738631e-07, "loss": 1.8502, "step": 36491 }, { "epoch": 0.9, "learning_rate": 5.252805401446481e-07, "loss": 2.0462, "step": 36492 }, { "epoch": 0.9, "learning_rate": 5.250252330009054e-07, "loss": 1.9139, "step": 36493 }, { "epoch": 0.9, "learning_rate": 5.247699862442657e-07, "loss": 1.9368, "step": 36494 }, { "epoch": 0.9, "learning_rate": 5.245147998763556e-07, "loss": 1.9546, "step": 36495 }, { "epoch": 0.9, "learning_rate": 5.242596738987993e-07, "loss": 1.8562, "step": 36496 }, { "epoch": 0.9, "learning_rate": 5.240046083132256e-07, "loss": 1.9088, "step": 36497 }, { "epoch": 0.9, "learning_rate": 5.237496031212574e-07, "loss": 1.8233, "step": 36498 }, { "epoch": 0.9, "learning_rate": 5.234946583245204e-07, "loss": 1.9647, "step": 36499 }, { "epoch": 0.9, "learning_rate": 5.232397739246431e-07, "loss": 1.7608, "step": 36500 }, { "epoch": 0.9, "learning_rate": 5.229849499232443e-07, "loss": 1.9141, "step": 36501 }, { "epoch": 0.9, "learning_rate": 5.227301863219514e-07, "loss": 1.8607, "step": 36502 }, { "epoch": 0.9, "learning_rate": 5.224754831223878e-07, "loss": 1.8322, "step": 36503 }, { "epoch": 0.9, "learning_rate": 5.222208403261775e-07, "loss": 1.9343, "step": 36504 }, { "epoch": 0.9, "learning_rate": 5.219662579349416e-07, "loss": 1.8096, "step": 36505 }, { "epoch": 0.9, "learning_rate": 5.217117359503032e-07, "loss": 2.0091, "step": 36506 }, { "epoch": 0.9, "learning_rate": 5.214572743738866e-07, "loss": 1.8206, "step": 36507 }, { "epoch": 0.9, "learning_rate": 5.212028732073105e-07, "loss": 1.8013, "step": 36508 }, { "epoch": 0.9, "learning_rate": 5.209485324521968e-07, "loss": 1.8772, "step": 36509 }, { "epoch": 0.9, "learning_rate": 5.206942521101688e-07, "loss": 2.0517, "step": 36510 }, { "epoch": 0.9, "learning_rate": 5.204400321828451e-07, "loss": 2.063, "step": 36511 }, { "epoch": 0.9, "learning_rate": 5.201858726718456e-07, "loss": 1.9464, "step": 36512 }, { "epoch": 0.9, "learning_rate": 5.199317735787923e-07, "loss": 1.9678, "step": 36513 }, { "epoch": 0.9, "eval_loss": 1.661559820175171, "eval_runtime": 94.6329, "eval_samples_per_second": 621.845, "eval_steps_per_second": 4.861, "step": 36513 } ], "logging_steps": 1, "max_steps": 40570, "num_input_tokens_seen": 0, "num_train_epochs": 1, "save_steps": 4057, "total_flos": 1.0376538569362637e+18, "train_batch_size": 64, "trial_name": null, "trial_params": null }