| { | |
| "best_metric": 157.60036917397323, | |
| "best_model_checkpoint": "/shared/rc/bambara/bambara_fieldwork_audiofolder/output_dir/checkpoint-4000", | |
| "epoch": 25.000517292344075, | |
| "global_step": 13532, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 4.2000000000000006e-07, | |
| "loss": 4.3307, | |
| "step": 25 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 9.200000000000001e-07, | |
| "loss": 3.6742, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 1.42e-06, | |
| "loss": 2.9307, | |
| "step": 75 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 1.9200000000000003e-06, | |
| "loss": 2.2561, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 2.42e-06, | |
| "loss": 1.9179, | |
| "step": 125 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 2.92e-06, | |
| "loss": 1.7292, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 3.4200000000000007e-06, | |
| "loss": 1.6139, | |
| "step": 175 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 3.920000000000001e-06, | |
| "loss": 1.3928, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 4.42e-06, | |
| "loss": 1.296, | |
| "step": 225 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 4.92e-06, | |
| "loss": 1.1752, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 5.420000000000001e-06, | |
| "loss": 1.1729, | |
| "step": 275 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 5.92e-06, | |
| "loss": 1.1148, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 6.42e-06, | |
| "loss": 1.1045, | |
| "step": 325 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 6.92e-06, | |
| "loss": 1.2149, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 7.420000000000001e-06, | |
| "loss": 1.0699, | |
| "step": 375 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 7.92e-06, | |
| "loss": 1.1688, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 8.42e-06, | |
| "loss": 1.0369, | |
| "step": 425 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 8.920000000000001e-06, | |
| "loss": 1.0005, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 9.42e-06, | |
| "loss": 0.9744, | |
| "step": 475 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 9.920000000000002e-06, | |
| "loss": 0.9877, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 9.98388581952118e-06, | |
| "loss": 0.9383, | |
| "step": 525 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 9.964702271332105e-06, | |
| "loss": 0.8368, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 9.945518723143033e-06, | |
| "loss": 0.7255, | |
| "step": 575 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 9.92633517495396e-06, | |
| "loss": 0.8185, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 9.907151626764887e-06, | |
| "loss": 0.7622, | |
| "step": 625 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 9.887968078575814e-06, | |
| "loss": 0.8058, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 9.868784530386742e-06, | |
| "loss": 0.7202, | |
| "step": 675 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 9.849600982197669e-06, | |
| "loss": 0.717, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 9.830417434008594e-06, | |
| "loss": 0.7412, | |
| "step": 725 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 9.811233885819521e-06, | |
| "loss": 0.8542, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 9.792050337630449e-06, | |
| "loss": 0.789, | |
| "step": 775 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 9.772866789441376e-06, | |
| "loss": 0.7616, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 9.753683241252303e-06, | |
| "loss": 0.7478, | |
| "step": 825 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 9.73449969306323e-06, | |
| "loss": 0.7398, | |
| "step": 850 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 9.715316144874156e-06, | |
| "loss": 0.7731, | |
| "step": 875 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 9.696132596685083e-06, | |
| "loss": 0.8003, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 9.67694904849601e-06, | |
| "loss": 0.7496, | |
| "step": 925 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 9.657765500306938e-06, | |
| "loss": 0.66, | |
| "step": 950 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 9.638581952117865e-06, | |
| "loss": 0.7409, | |
| "step": 975 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 9.619398403928792e-06, | |
| "loss": 0.6342, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "eval_loss": 2.580996513366699, | |
| "eval_runtime": 1370.4044, | |
| "eval_samples_per_second": 0.931, | |
| "eval_steps_per_second": 0.466, | |
| "eval_wer": 159.52930318412552, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "learning_rate": 9.600214855739717e-06, | |
| "loss": 0.7006, | |
| "step": 1025 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "learning_rate": 9.581031307550646e-06, | |
| "loss": 0.7081, | |
| "step": 1050 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "learning_rate": 9.561847759361572e-06, | |
| "loss": 0.647, | |
| "step": 1075 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "learning_rate": 9.542664211172499e-06, | |
| "loss": 0.5766, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "learning_rate": 9.523480662983426e-06, | |
| "loss": 0.5153, | |
| "step": 1125 | |
| }, | |
| { | |
| "epoch": 2.01, | |
| "learning_rate": 9.504297114794354e-06, | |
| "loss": 0.4956, | |
| "step": 1150 | |
| }, | |
| { | |
| "epoch": 2.01, | |
| "learning_rate": 9.485113566605279e-06, | |
| "loss": 0.4767, | |
| "step": 1175 | |
| }, | |
| { | |
| "epoch": 2.01, | |
| "learning_rate": 9.465930018416208e-06, | |
| "loss": 0.5307, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 2.01, | |
| "learning_rate": 9.446746470227133e-06, | |
| "loss": 0.4805, | |
| "step": 1225 | |
| }, | |
| { | |
| "epoch": 2.01, | |
| "learning_rate": 9.42756292203806e-06, | |
| "loss": 0.4913, | |
| "step": 1250 | |
| }, | |
| { | |
| "epoch": 2.01, | |
| "learning_rate": 9.408379373848988e-06, | |
| "loss": 0.46, | |
| "step": 1275 | |
| }, | |
| { | |
| "epoch": 2.02, | |
| "learning_rate": 9.389195825659915e-06, | |
| "loss": 0.4688, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 2.02, | |
| "learning_rate": 9.37001227747084e-06, | |
| "loss": 0.4379, | |
| "step": 1325 | |
| }, | |
| { | |
| "epoch": 2.02, | |
| "learning_rate": 9.35082872928177e-06, | |
| "loss": 0.5624, | |
| "step": 1350 | |
| }, | |
| { | |
| "epoch": 2.02, | |
| "learning_rate": 9.331645181092695e-06, | |
| "loss": 0.4703, | |
| "step": 1375 | |
| }, | |
| { | |
| "epoch": 2.02, | |
| "learning_rate": 9.312461632903622e-06, | |
| "loss": 0.5075, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 2.03, | |
| "learning_rate": 9.29327808471455e-06, | |
| "loss": 0.4622, | |
| "step": 1425 | |
| }, | |
| { | |
| "epoch": 2.03, | |
| "learning_rate": 9.274094536525477e-06, | |
| "loss": 0.4639, | |
| "step": 1450 | |
| }, | |
| { | |
| "epoch": 2.03, | |
| "learning_rate": 9.254910988336402e-06, | |
| "loss": 0.4338, | |
| "step": 1475 | |
| }, | |
| { | |
| "epoch": 2.03, | |
| "learning_rate": 9.235727440147331e-06, | |
| "loss": 0.506, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 2.03, | |
| "learning_rate": 9.216543891958257e-06, | |
| "loss": 0.4865, | |
| "step": 1525 | |
| }, | |
| { | |
| "epoch": 2.03, | |
| "learning_rate": 9.197360343769184e-06, | |
| "loss": 0.5065, | |
| "step": 1550 | |
| }, | |
| { | |
| "epoch": 2.04, | |
| "learning_rate": 9.178176795580111e-06, | |
| "loss": 0.4669, | |
| "step": 1575 | |
| }, | |
| { | |
| "epoch": 2.04, | |
| "learning_rate": 9.158993247391038e-06, | |
| "loss": 0.558, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "learning_rate": 9.139809699201966e-06, | |
| "loss": 0.4815, | |
| "step": 1625 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "learning_rate": 9.120626151012893e-06, | |
| "loss": 0.3453, | |
| "step": 1650 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "learning_rate": 9.101442602823818e-06, | |
| "loss": 0.2932, | |
| "step": 1675 | |
| }, | |
| { | |
| "epoch": 3.01, | |
| "learning_rate": 9.082259054634745e-06, | |
| "loss": 0.3748, | |
| "step": 1700 | |
| }, | |
| { | |
| "epoch": 3.01, | |
| "learning_rate": 9.063075506445673e-06, | |
| "loss": 0.2992, | |
| "step": 1725 | |
| }, | |
| { | |
| "epoch": 3.01, | |
| "learning_rate": 9.0438919582566e-06, | |
| "loss": 0.3023, | |
| "step": 1750 | |
| }, | |
| { | |
| "epoch": 3.01, | |
| "learning_rate": 9.024708410067527e-06, | |
| "loss": 0.3073, | |
| "step": 1775 | |
| }, | |
| { | |
| "epoch": 3.01, | |
| "learning_rate": 9.005524861878454e-06, | |
| "loss": 0.3569, | |
| "step": 1800 | |
| }, | |
| { | |
| "epoch": 3.01, | |
| "learning_rate": 8.98634131368938e-06, | |
| "loss": 0.3518, | |
| "step": 1825 | |
| }, | |
| { | |
| "epoch": 3.02, | |
| "learning_rate": 8.967157765500307e-06, | |
| "loss": 0.3064, | |
| "step": 1850 | |
| }, | |
| { | |
| "epoch": 3.02, | |
| "learning_rate": 8.947974217311234e-06, | |
| "loss": 0.3668, | |
| "step": 1875 | |
| }, | |
| { | |
| "epoch": 3.02, | |
| "learning_rate": 8.928790669122162e-06, | |
| "loss": 0.3097, | |
| "step": 1900 | |
| }, | |
| { | |
| "epoch": 3.02, | |
| "learning_rate": 8.909607120933089e-06, | |
| "loss": 0.3333, | |
| "step": 1925 | |
| }, | |
| { | |
| "epoch": 3.02, | |
| "learning_rate": 8.890423572744016e-06, | |
| "loss": 0.3472, | |
| "step": 1950 | |
| }, | |
| { | |
| "epoch": 3.03, | |
| "learning_rate": 8.871240024554943e-06, | |
| "loss": 0.3303, | |
| "step": 1975 | |
| }, | |
| { | |
| "epoch": 3.03, | |
| "learning_rate": 8.852056476365869e-06, | |
| "loss": 0.2873, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 3.03, | |
| "eval_loss": 2.951326370239258, | |
| "eval_runtime": 1218.1818, | |
| "eval_samples_per_second": 1.047, | |
| "eval_steps_per_second": 0.524, | |
| "eval_wer": 159.11398246423627, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 3.03, | |
| "learning_rate": 8.832872928176796e-06, | |
| "loss": 0.3295, | |
| "step": 2025 | |
| }, | |
| { | |
| "epoch": 3.03, | |
| "learning_rate": 8.813689379987723e-06, | |
| "loss": 0.3127, | |
| "step": 2050 | |
| }, | |
| { | |
| "epoch": 3.03, | |
| "learning_rate": 8.79450583179865e-06, | |
| "loss": 0.3399, | |
| "step": 2075 | |
| }, | |
| { | |
| "epoch": 3.04, | |
| "learning_rate": 8.775322283609578e-06, | |
| "loss": 0.3172, | |
| "step": 2100 | |
| }, | |
| { | |
| "epoch": 3.04, | |
| "learning_rate": 8.756138735420505e-06, | |
| "loss": 0.3748, | |
| "step": 2125 | |
| }, | |
| { | |
| "epoch": 3.04, | |
| "learning_rate": 8.73695518723143e-06, | |
| "loss": 0.3389, | |
| "step": 2150 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "learning_rate": 8.71777163904236e-06, | |
| "loss": 0.2796, | |
| "step": 2175 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "learning_rate": 8.698588090853285e-06, | |
| "loss": 0.1639, | |
| "step": 2200 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "learning_rate": 8.679404542664212e-06, | |
| "loss": 0.1772, | |
| "step": 2225 | |
| }, | |
| { | |
| "epoch": 4.01, | |
| "learning_rate": 8.660220994475139e-06, | |
| "loss": 0.2148, | |
| "step": 2250 | |
| }, | |
| { | |
| "epoch": 4.01, | |
| "learning_rate": 8.641037446286066e-06, | |
| "loss": 0.2203, | |
| "step": 2275 | |
| }, | |
| { | |
| "epoch": 4.01, | |
| "learning_rate": 8.621853898096992e-06, | |
| "loss": 0.2337, | |
| "step": 2300 | |
| }, | |
| { | |
| "epoch": 4.01, | |
| "learning_rate": 8.60267034990792e-06, | |
| "loss": 0.2417, | |
| "step": 2325 | |
| }, | |
| { | |
| "epoch": 4.01, | |
| "learning_rate": 8.583486801718846e-06, | |
| "loss": 0.2266, | |
| "step": 2350 | |
| }, | |
| { | |
| "epoch": 4.02, | |
| "learning_rate": 8.564303253529774e-06, | |
| "loss": 0.204, | |
| "step": 2375 | |
| }, | |
| { | |
| "epoch": 4.02, | |
| "learning_rate": 8.5451197053407e-06, | |
| "loss": 0.215, | |
| "step": 2400 | |
| }, | |
| { | |
| "epoch": 4.02, | |
| "learning_rate": 8.525936157151628e-06, | |
| "loss": 0.2082, | |
| "step": 2425 | |
| }, | |
| { | |
| "epoch": 4.02, | |
| "learning_rate": 8.506752608962553e-06, | |
| "loss": 0.2592, | |
| "step": 2450 | |
| }, | |
| { | |
| "epoch": 4.02, | |
| "learning_rate": 8.487569060773482e-06, | |
| "loss": 0.1869, | |
| "step": 2475 | |
| }, | |
| { | |
| "epoch": 4.02, | |
| "learning_rate": 8.468385512584408e-06, | |
| "loss": 0.2161, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 4.03, | |
| "learning_rate": 8.449201964395335e-06, | |
| "loss": 0.2381, | |
| "step": 2525 | |
| }, | |
| { | |
| "epoch": 4.03, | |
| "learning_rate": 8.430018416206262e-06, | |
| "loss": 0.2139, | |
| "step": 2550 | |
| }, | |
| { | |
| "epoch": 4.03, | |
| "learning_rate": 8.41083486801719e-06, | |
| "loss": 0.2071, | |
| "step": 2575 | |
| }, | |
| { | |
| "epoch": 4.03, | |
| "learning_rate": 8.391651319828115e-06, | |
| "loss": 0.2146, | |
| "step": 2600 | |
| }, | |
| { | |
| "epoch": 4.03, | |
| "learning_rate": 8.372467771639044e-06, | |
| "loss": 0.2096, | |
| "step": 2625 | |
| }, | |
| { | |
| "epoch": 4.04, | |
| "learning_rate": 8.35328422344997e-06, | |
| "loss": 0.2617, | |
| "step": 2650 | |
| }, | |
| { | |
| "epoch": 4.04, | |
| "learning_rate": 8.334100675260897e-06, | |
| "loss": 0.2189, | |
| "step": 2675 | |
| }, | |
| { | |
| "epoch": 4.04, | |
| "learning_rate": 8.314917127071824e-06, | |
| "loss": 0.2841, | |
| "step": 2700 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "learning_rate": 8.295733578882751e-06, | |
| "loss": 0.1401, | |
| "step": 2725 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "learning_rate": 8.276550030693677e-06, | |
| "loss": 0.1293, | |
| "step": 2750 | |
| }, | |
| { | |
| "epoch": 5.01, | |
| "learning_rate": 8.257366482504606e-06, | |
| "loss": 0.1256, | |
| "step": 2775 | |
| }, | |
| { | |
| "epoch": 5.01, | |
| "learning_rate": 8.238182934315531e-06, | |
| "loss": 0.1383, | |
| "step": 2800 | |
| }, | |
| { | |
| "epoch": 5.01, | |
| "learning_rate": 8.218999386126458e-06, | |
| "loss": 0.1406, | |
| "step": 2825 | |
| }, | |
| { | |
| "epoch": 5.01, | |
| "learning_rate": 8.199815837937386e-06, | |
| "loss": 0.1194, | |
| "step": 2850 | |
| }, | |
| { | |
| "epoch": 5.01, | |
| "learning_rate": 8.180632289748313e-06, | |
| "loss": 0.1206, | |
| "step": 2875 | |
| }, | |
| { | |
| "epoch": 5.01, | |
| "learning_rate": 8.16144874155924e-06, | |
| "loss": 0.1397, | |
| "step": 2900 | |
| }, | |
| { | |
| "epoch": 5.02, | |
| "learning_rate": 8.142265193370167e-06, | |
| "loss": 0.1262, | |
| "step": 2925 | |
| }, | |
| { | |
| "epoch": 5.02, | |
| "learning_rate": 8.123081645181093e-06, | |
| "loss": 0.1293, | |
| "step": 2950 | |
| }, | |
| { | |
| "epoch": 5.02, | |
| "learning_rate": 8.10389809699202e-06, | |
| "loss": 0.1363, | |
| "step": 2975 | |
| }, | |
| { | |
| "epoch": 5.02, | |
| "learning_rate": 8.084714548802947e-06, | |
| "loss": 0.1461, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 5.02, | |
| "eval_loss": 3.683340072631836, | |
| "eval_runtime": 1194.521, | |
| "eval_samples_per_second": 1.068, | |
| "eval_steps_per_second": 0.534, | |
| "eval_wer": 158.39409321642825, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 5.02, | |
| "learning_rate": 8.065531000613874e-06, | |
| "loss": 0.1587, | |
| "step": 3025 | |
| }, | |
| { | |
| "epoch": 5.03, | |
| "learning_rate": 8.046347452424802e-06, | |
| "loss": 0.1474, | |
| "step": 3050 | |
| }, | |
| { | |
| "epoch": 5.03, | |
| "learning_rate": 8.027163904235729e-06, | |
| "loss": 0.1418, | |
| "step": 3075 | |
| }, | |
| { | |
| "epoch": 5.03, | |
| "learning_rate": 8.007980356046656e-06, | |
| "loss": 0.1306, | |
| "step": 3100 | |
| }, | |
| { | |
| "epoch": 5.03, | |
| "learning_rate": 7.988796807857581e-06, | |
| "loss": 0.1444, | |
| "step": 3125 | |
| }, | |
| { | |
| "epoch": 5.03, | |
| "learning_rate": 7.969613259668509e-06, | |
| "loss": 0.1617, | |
| "step": 3150 | |
| }, | |
| { | |
| "epoch": 5.03, | |
| "learning_rate": 7.950429711479436e-06, | |
| "loss": 0.1365, | |
| "step": 3175 | |
| }, | |
| { | |
| "epoch": 5.04, | |
| "learning_rate": 7.931246163290363e-06, | |
| "loss": 0.1393, | |
| "step": 3200 | |
| }, | |
| { | |
| "epoch": 5.04, | |
| "learning_rate": 7.91206261510129e-06, | |
| "loss": 0.1547, | |
| "step": 3225 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "learning_rate": 7.892879066912218e-06, | |
| "loss": 0.1446, | |
| "step": 3250 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "learning_rate": 7.873695518723143e-06, | |
| "loss": 0.07, | |
| "step": 3275 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "learning_rate": 7.85451197053407e-06, | |
| "loss": 0.0875, | |
| "step": 3300 | |
| }, | |
| { | |
| "epoch": 6.01, | |
| "learning_rate": 7.835328422344998e-06, | |
| "loss": 0.0758, | |
| "step": 3325 | |
| }, | |
| { | |
| "epoch": 6.01, | |
| "learning_rate": 7.816144874155925e-06, | |
| "loss": 0.081, | |
| "step": 3350 | |
| }, | |
| { | |
| "epoch": 6.01, | |
| "learning_rate": 7.796961325966852e-06, | |
| "loss": 0.0827, | |
| "step": 3375 | |
| }, | |
| { | |
| "epoch": 6.01, | |
| "learning_rate": 7.77777777777778e-06, | |
| "loss": 0.0689, | |
| "step": 3400 | |
| }, | |
| { | |
| "epoch": 6.01, | |
| "learning_rate": 7.758594229588705e-06, | |
| "loss": 0.0904, | |
| "step": 3425 | |
| }, | |
| { | |
| "epoch": 6.02, | |
| "learning_rate": 7.739410681399634e-06, | |
| "loss": 0.099, | |
| "step": 3450 | |
| }, | |
| { | |
| "epoch": 6.02, | |
| "learning_rate": 7.720227133210559e-06, | |
| "loss": 0.0837, | |
| "step": 3475 | |
| }, | |
| { | |
| "epoch": 6.02, | |
| "learning_rate": 7.701043585021486e-06, | |
| "loss": 0.0811, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 6.02, | |
| "learning_rate": 7.681860036832414e-06, | |
| "loss": 0.0925, | |
| "step": 3525 | |
| }, | |
| { | |
| "epoch": 6.02, | |
| "learning_rate": 7.66267648864334e-06, | |
| "loss": 0.0964, | |
| "step": 3550 | |
| }, | |
| { | |
| "epoch": 6.02, | |
| "learning_rate": 7.643492940454266e-06, | |
| "loss": 0.0802, | |
| "step": 3575 | |
| }, | |
| { | |
| "epoch": 6.03, | |
| "learning_rate": 7.624309392265194e-06, | |
| "loss": 0.083, | |
| "step": 3600 | |
| }, | |
| { | |
| "epoch": 6.03, | |
| "learning_rate": 7.605125844076121e-06, | |
| "loss": 0.0955, | |
| "step": 3625 | |
| }, | |
| { | |
| "epoch": 6.03, | |
| "learning_rate": 7.585942295887048e-06, | |
| "loss": 0.091, | |
| "step": 3650 | |
| }, | |
| { | |
| "epoch": 6.03, | |
| "learning_rate": 7.566758747697975e-06, | |
| "loss": 0.0668, | |
| "step": 3675 | |
| }, | |
| { | |
| "epoch": 6.03, | |
| "learning_rate": 7.547575199508902e-06, | |
| "loss": 0.0839, | |
| "step": 3700 | |
| }, | |
| { | |
| "epoch": 6.04, | |
| "learning_rate": 7.528391651319829e-06, | |
| "loss": 0.1103, | |
| "step": 3725 | |
| }, | |
| { | |
| "epoch": 6.04, | |
| "learning_rate": 7.509208103130756e-06, | |
| "loss": 0.0722, | |
| "step": 3750 | |
| }, | |
| { | |
| "epoch": 6.04, | |
| "learning_rate": 7.490024554941683e-06, | |
| "loss": 0.0867, | |
| "step": 3775 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "learning_rate": 7.4708410067526095e-06, | |
| "loss": 0.0612, | |
| "step": 3800 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "learning_rate": 7.451657458563537e-06, | |
| "loss": 0.049, | |
| "step": 3825 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "learning_rate": 7.432473910374464e-06, | |
| "loss": 0.0482, | |
| "step": 3850 | |
| }, | |
| { | |
| "epoch": 7.01, | |
| "learning_rate": 7.41329036218539e-06, | |
| "loss": 0.0552, | |
| "step": 3875 | |
| }, | |
| { | |
| "epoch": 7.01, | |
| "learning_rate": 7.3941068139963175e-06, | |
| "loss": 0.0578, | |
| "step": 3900 | |
| }, | |
| { | |
| "epoch": 7.01, | |
| "learning_rate": 7.374923265807245e-06, | |
| "loss": 0.0495, | |
| "step": 3925 | |
| }, | |
| { | |
| "epoch": 7.01, | |
| "learning_rate": 7.355739717618171e-06, | |
| "loss": 0.045, | |
| "step": 3950 | |
| }, | |
| { | |
| "epoch": 7.01, | |
| "learning_rate": 7.3365561694290975e-06, | |
| "loss": 0.0578, | |
| "step": 3975 | |
| }, | |
| { | |
| "epoch": 7.02, | |
| "learning_rate": 7.3173726212400255e-06, | |
| "loss": 0.049, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 7.02, | |
| "eval_loss": 4.072541236877441, | |
| "eval_runtime": 1245.9007, | |
| "eval_samples_per_second": 1.024, | |
| "eval_steps_per_second": 0.512, | |
| "eval_wer": 157.60036917397323, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 7.02, | |
| "learning_rate": 7.298189073050952e-06, | |
| "loss": 0.0528, | |
| "step": 4025 | |
| }, | |
| { | |
| "epoch": 7.02, | |
| "learning_rate": 7.279005524861878e-06, | |
| "loss": 0.0604, | |
| "step": 4050 | |
| }, | |
| { | |
| "epoch": 7.02, | |
| "learning_rate": 7.259821976672806e-06, | |
| "loss": 0.0555, | |
| "step": 4075 | |
| }, | |
| { | |
| "epoch": 7.02, | |
| "learning_rate": 7.240638428483733e-06, | |
| "loss": 0.0507, | |
| "step": 4100 | |
| }, | |
| { | |
| "epoch": 7.02, | |
| "learning_rate": 7.221454880294659e-06, | |
| "loss": 0.0479, | |
| "step": 4125 | |
| }, | |
| { | |
| "epoch": 7.03, | |
| "learning_rate": 7.202271332105587e-06, | |
| "loss": 0.0509, | |
| "step": 4150 | |
| }, | |
| { | |
| "epoch": 7.03, | |
| "learning_rate": 7.1830877839165135e-06, | |
| "loss": 0.0566, | |
| "step": 4175 | |
| }, | |
| { | |
| "epoch": 7.03, | |
| "learning_rate": 7.16390423572744e-06, | |
| "loss": 0.0619, | |
| "step": 4200 | |
| }, | |
| { | |
| "epoch": 7.03, | |
| "learning_rate": 7.144720687538368e-06, | |
| "loss": 0.0685, | |
| "step": 4225 | |
| }, | |
| { | |
| "epoch": 7.03, | |
| "learning_rate": 7.125537139349294e-06, | |
| "loss": 0.0468, | |
| "step": 4250 | |
| }, | |
| { | |
| "epoch": 7.04, | |
| "learning_rate": 7.106353591160221e-06, | |
| "loss": 0.0526, | |
| "step": 4275 | |
| }, | |
| { | |
| "epoch": 7.04, | |
| "learning_rate": 7.087170042971149e-06, | |
| "loss": 0.0584, | |
| "step": 4300 | |
| }, | |
| { | |
| "epoch": 7.04, | |
| "learning_rate": 7.067986494782075e-06, | |
| "loss": 0.0487, | |
| "step": 4325 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "learning_rate": 7.048802946593002e-06, | |
| "loss": 0.0345, | |
| "step": 4350 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "learning_rate": 7.0296193984039295e-06, | |
| "loss": 0.03, | |
| "step": 4375 | |
| }, | |
| { | |
| "epoch": 8.01, | |
| "learning_rate": 7.010435850214856e-06, | |
| "loss": 0.0355, | |
| "step": 4400 | |
| }, | |
| { | |
| "epoch": 8.01, | |
| "learning_rate": 6.991252302025783e-06, | |
| "loss": 0.0348, | |
| "step": 4425 | |
| }, | |
| { | |
| "epoch": 8.01, | |
| "learning_rate": 6.97206875383671e-06, | |
| "loss": 0.0252, | |
| "step": 4450 | |
| }, | |
| { | |
| "epoch": 8.01, | |
| "learning_rate": 6.952885205647637e-06, | |
| "loss": 0.0307, | |
| "step": 4475 | |
| }, | |
| { | |
| "epoch": 8.01, | |
| "learning_rate": 6.933701657458564e-06, | |
| "loss": 0.0306, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 8.01, | |
| "learning_rate": 6.914518109269491e-06, | |
| "loss": 0.0345, | |
| "step": 4525 | |
| }, | |
| { | |
| "epoch": 8.02, | |
| "learning_rate": 6.8953345610804175e-06, | |
| "loss": 0.0274, | |
| "step": 4550 | |
| }, | |
| { | |
| "epoch": 8.02, | |
| "learning_rate": 6.876151012891345e-06, | |
| "loss": 0.0293, | |
| "step": 4575 | |
| }, | |
| { | |
| "epoch": 8.02, | |
| "learning_rate": 6.856967464702272e-06, | |
| "loss": 0.0328, | |
| "step": 4600 | |
| }, | |
| { | |
| "epoch": 8.02, | |
| "learning_rate": 6.837783916513199e-06, | |
| "loss": 0.0282, | |
| "step": 4625 | |
| }, | |
| { | |
| "epoch": 8.02, | |
| "learning_rate": 6.8186003683241255e-06, | |
| "loss": 0.0405, | |
| "step": 4650 | |
| }, | |
| { | |
| "epoch": 8.03, | |
| "learning_rate": 6.799416820135053e-06, | |
| "loss": 0.041, | |
| "step": 4675 | |
| }, | |
| { | |
| "epoch": 8.03, | |
| "learning_rate": 6.78023327194598e-06, | |
| "loss": 0.0324, | |
| "step": 4700 | |
| }, | |
| { | |
| "epoch": 8.03, | |
| "learning_rate": 6.761049723756906e-06, | |
| "loss": 0.0367, | |
| "step": 4725 | |
| }, | |
| { | |
| "epoch": 8.03, | |
| "learning_rate": 6.7418661755678335e-06, | |
| "loss": 0.0401, | |
| "step": 4750 | |
| }, | |
| { | |
| "epoch": 8.03, | |
| "learning_rate": 6.722682627378761e-06, | |
| "loss": 0.0392, | |
| "step": 4775 | |
| }, | |
| { | |
| "epoch": 8.03, | |
| "learning_rate": 6.703499079189687e-06, | |
| "loss": 0.0412, | |
| "step": 4800 | |
| }, | |
| { | |
| "epoch": 8.04, | |
| "learning_rate": 6.684315531000614e-06, | |
| "loss": 0.0355, | |
| "step": 4825 | |
| }, | |
| { | |
| "epoch": 8.04, | |
| "learning_rate": 6.6651319828115415e-06, | |
| "loss": 0.0369, | |
| "step": 4850 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "learning_rate": 6.645948434622468e-06, | |
| "loss": 0.0321, | |
| "step": 4875 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "learning_rate": 6.626764886433396e-06, | |
| "loss": 0.0144, | |
| "step": 4900 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "learning_rate": 6.607581338244322e-06, | |
| "loss": 0.0204, | |
| "step": 4925 | |
| }, | |
| { | |
| "epoch": 9.01, | |
| "learning_rate": 6.588397790055249e-06, | |
| "loss": 0.0149, | |
| "step": 4950 | |
| }, | |
| { | |
| "epoch": 9.01, | |
| "learning_rate": 6.569214241866177e-06, | |
| "loss": 0.0249, | |
| "step": 4975 | |
| }, | |
| { | |
| "epoch": 9.01, | |
| "learning_rate": 6.550030693677103e-06, | |
| "loss": 0.0218, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 9.01, | |
| "eval_loss": 4.253083229064941, | |
| "eval_runtime": 1193.3336, | |
| "eval_samples_per_second": 1.069, | |
| "eval_steps_per_second": 0.535, | |
| "eval_wer": 158.36640516843562, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 9.01, | |
| "learning_rate": 6.5308471454880295e-06, | |
| "loss": 0.0247, | |
| "step": 5025 | |
| }, | |
| { | |
| "epoch": 9.01, | |
| "learning_rate": 6.5116635972989576e-06, | |
| "loss": 0.0224, | |
| "step": 5050 | |
| }, | |
| { | |
| "epoch": 9.02, | |
| "learning_rate": 6.492480049109884e-06, | |
| "loss": 0.0223, | |
| "step": 5075 | |
| }, | |
| { | |
| "epoch": 9.02, | |
| "learning_rate": 6.47329650092081e-06, | |
| "loss": 0.023, | |
| "step": 5100 | |
| }, | |
| { | |
| "epoch": 9.02, | |
| "learning_rate": 6.454112952731738e-06, | |
| "loss": 0.0197, | |
| "step": 5125 | |
| }, | |
| { | |
| "epoch": 9.02, | |
| "learning_rate": 6.434929404542665e-06, | |
| "loss": 0.0172, | |
| "step": 5150 | |
| }, | |
| { | |
| "epoch": 9.02, | |
| "learning_rate": 6.415745856353591e-06, | |
| "loss": 0.021, | |
| "step": 5175 | |
| }, | |
| { | |
| "epoch": 9.02, | |
| "learning_rate": 6.396562308164519e-06, | |
| "loss": 0.0235, | |
| "step": 5200 | |
| }, | |
| { | |
| "epoch": 9.03, | |
| "learning_rate": 6.3773787599754455e-06, | |
| "loss": 0.0242, | |
| "step": 5225 | |
| }, | |
| { | |
| "epoch": 9.03, | |
| "learning_rate": 6.358195211786372e-06, | |
| "loss": 0.0162, | |
| "step": 5250 | |
| }, | |
| { | |
| "epoch": 9.03, | |
| "learning_rate": 6.3390116635973e-06, | |
| "loss": 0.0292, | |
| "step": 5275 | |
| }, | |
| { | |
| "epoch": 9.03, | |
| "learning_rate": 6.319828115408226e-06, | |
| "loss": 0.0146, | |
| "step": 5300 | |
| }, | |
| { | |
| "epoch": 9.03, | |
| "learning_rate": 6.300644567219153e-06, | |
| "loss": 0.0262, | |
| "step": 5325 | |
| }, | |
| { | |
| "epoch": 9.04, | |
| "learning_rate": 6.281461019030081e-06, | |
| "loss": 0.0173, | |
| "step": 5350 | |
| }, | |
| { | |
| "epoch": 9.04, | |
| "learning_rate": 6.262277470841007e-06, | |
| "loss": 0.0271, | |
| "step": 5375 | |
| }, | |
| { | |
| "epoch": 9.04, | |
| "learning_rate": 6.2430939226519335e-06, | |
| "loss": 0.0251, | |
| "step": 5400 | |
| }, | |
| { | |
| "epoch": 10.0, | |
| "learning_rate": 6.2239103744628615e-06, | |
| "loss": 0.0205, | |
| "step": 5425 | |
| }, | |
| { | |
| "epoch": 10.0, | |
| "learning_rate": 6.204726826273788e-06, | |
| "loss": 0.0119, | |
| "step": 5450 | |
| }, | |
| { | |
| "epoch": 10.0, | |
| "learning_rate": 6.185543278084714e-06, | |
| "loss": 0.0138, | |
| "step": 5475 | |
| }, | |
| { | |
| "epoch": 10.01, | |
| "learning_rate": 6.166359729895642e-06, | |
| "loss": 0.0141, | |
| "step": 5500 | |
| }, | |
| { | |
| "epoch": 10.01, | |
| "learning_rate": 6.147176181706569e-06, | |
| "loss": 0.0165, | |
| "step": 5525 | |
| }, | |
| { | |
| "epoch": 10.01, | |
| "learning_rate": 6.127992633517496e-06, | |
| "loss": 0.0166, | |
| "step": 5550 | |
| }, | |
| { | |
| "epoch": 10.01, | |
| "learning_rate": 6.108809085328423e-06, | |
| "loss": 0.0112, | |
| "step": 5575 | |
| }, | |
| { | |
| "epoch": 10.01, | |
| "learning_rate": 6.0896255371393495e-06, | |
| "loss": 0.0177, | |
| "step": 5600 | |
| }, | |
| { | |
| "epoch": 10.02, | |
| "learning_rate": 6.070441988950277e-06, | |
| "loss": 0.0129, | |
| "step": 5625 | |
| }, | |
| { | |
| "epoch": 10.02, | |
| "learning_rate": 6.051258440761204e-06, | |
| "loss": 0.017, | |
| "step": 5650 | |
| }, | |
| { | |
| "epoch": 10.02, | |
| "learning_rate": 6.03207489257213e-06, | |
| "loss": 0.0117, | |
| "step": 5675 | |
| }, | |
| { | |
| "epoch": 10.02, | |
| "learning_rate": 6.0128913443830575e-06, | |
| "loss": 0.0152, | |
| "step": 5700 | |
| }, | |
| { | |
| "epoch": 10.02, | |
| "learning_rate": 5.993707796193985e-06, | |
| "loss": 0.014, | |
| "step": 5725 | |
| }, | |
| { | |
| "epoch": 10.03, | |
| "learning_rate": 5.974524248004911e-06, | |
| "loss": 0.0127, | |
| "step": 5750 | |
| }, | |
| { | |
| "epoch": 10.03, | |
| "learning_rate": 5.955340699815838e-06, | |
| "loss": 0.016, | |
| "step": 5775 | |
| }, | |
| { | |
| "epoch": 10.03, | |
| "learning_rate": 5.9361571516267655e-06, | |
| "loss": 0.0143, | |
| "step": 5800 | |
| }, | |
| { | |
| "epoch": 10.03, | |
| "learning_rate": 5.916973603437693e-06, | |
| "loss": 0.0147, | |
| "step": 5825 | |
| }, | |
| { | |
| "epoch": 10.03, | |
| "learning_rate": 5.897790055248619e-06, | |
| "loss": 0.0246, | |
| "step": 5850 | |
| }, | |
| { | |
| "epoch": 10.03, | |
| "learning_rate": 5.878606507059546e-06, | |
| "loss": 0.0126, | |
| "step": 5875 | |
| }, | |
| { | |
| "epoch": 10.04, | |
| "learning_rate": 5.8594229588704735e-06, | |
| "loss": 0.0152, | |
| "step": 5900 | |
| }, | |
| { | |
| "epoch": 10.04, | |
| "learning_rate": 5.8402394106814e-06, | |
| "loss": 0.0159, | |
| "step": 5925 | |
| }, | |
| { | |
| "epoch": 10.04, | |
| "learning_rate": 5.821055862492327e-06, | |
| "loss": 0.0164, | |
| "step": 5950 | |
| }, | |
| { | |
| "epoch": 11.0, | |
| "learning_rate": 5.801872314303254e-06, | |
| "loss": 0.011, | |
| "step": 5975 | |
| }, | |
| { | |
| "epoch": 11.0, | |
| "learning_rate": 5.782688766114181e-06, | |
| "loss": 0.0071, | |
| "step": 6000 | |
| }, | |
| { | |
| "epoch": 11.0, | |
| "eval_loss": 4.594424247741699, | |
| "eval_runtime": 1248.7993, | |
| "eval_samples_per_second": 1.022, | |
| "eval_steps_per_second": 0.511, | |
| "eval_wer": 157.99723119520075, | |
| "step": 6000 | |
| }, | |
| { | |
| "epoch": 11.01, | |
| "learning_rate": 5.763505217925108e-06, | |
| "loss": 0.01, | |
| "step": 6025 | |
| }, | |
| { | |
| "epoch": 11.01, | |
| "learning_rate": 5.744321669736035e-06, | |
| "loss": 0.0086, | |
| "step": 6050 | |
| }, | |
| { | |
| "epoch": 11.01, | |
| "learning_rate": 5.7251381215469615e-06, | |
| "loss": 0.012, | |
| "step": 6075 | |
| }, | |
| { | |
| "epoch": 11.01, | |
| "learning_rate": 5.7059545733578896e-06, | |
| "loss": 0.0121, | |
| "step": 6100 | |
| }, | |
| { | |
| "epoch": 11.01, | |
| "learning_rate": 5.686771025168816e-06, | |
| "loss": 0.011, | |
| "step": 6125 | |
| }, | |
| { | |
| "epoch": 11.01, | |
| "learning_rate": 5.667587476979742e-06, | |
| "loss": 0.0089, | |
| "step": 6150 | |
| }, | |
| { | |
| "epoch": 11.02, | |
| "learning_rate": 5.64840392879067e-06, | |
| "loss": 0.013, | |
| "step": 6175 | |
| }, | |
| { | |
| "epoch": 11.02, | |
| "learning_rate": 5.629220380601597e-06, | |
| "loss": 0.0124, | |
| "step": 6200 | |
| }, | |
| { | |
| "epoch": 11.02, | |
| "learning_rate": 5.610036832412523e-06, | |
| "loss": 0.0127, | |
| "step": 6225 | |
| }, | |
| { | |
| "epoch": 11.02, | |
| "learning_rate": 5.590853284223451e-06, | |
| "loss": 0.0096, | |
| "step": 6250 | |
| }, | |
| { | |
| "epoch": 11.02, | |
| "learning_rate": 5.5716697360343775e-06, | |
| "loss": 0.0095, | |
| "step": 6275 | |
| }, | |
| { | |
| "epoch": 11.03, | |
| "learning_rate": 5.552486187845304e-06, | |
| "loss": 0.0093, | |
| "step": 6300 | |
| }, | |
| { | |
| "epoch": 11.03, | |
| "learning_rate": 5.533302639656232e-06, | |
| "loss": 0.0117, | |
| "step": 6325 | |
| }, | |
| { | |
| "epoch": 11.03, | |
| "learning_rate": 5.514119091467158e-06, | |
| "loss": 0.0087, | |
| "step": 6350 | |
| }, | |
| { | |
| "epoch": 11.03, | |
| "learning_rate": 5.494935543278085e-06, | |
| "loss": 0.0103, | |
| "step": 6375 | |
| }, | |
| { | |
| "epoch": 11.03, | |
| "learning_rate": 5.475751995089013e-06, | |
| "loss": 0.0061, | |
| "step": 6400 | |
| }, | |
| { | |
| "epoch": 11.04, | |
| "learning_rate": 5.456568446899939e-06, | |
| "loss": 0.0083, | |
| "step": 6425 | |
| }, | |
| { | |
| "epoch": 11.04, | |
| "learning_rate": 5.4373848987108655e-06, | |
| "loss": 0.0111, | |
| "step": 6450 | |
| }, | |
| { | |
| "epoch": 11.04, | |
| "learning_rate": 5.4182013505217936e-06, | |
| "loss": 0.0076, | |
| "step": 6475 | |
| }, | |
| { | |
| "epoch": 12.0, | |
| "learning_rate": 5.39901780233272e-06, | |
| "loss": 0.0115, | |
| "step": 6500 | |
| }, | |
| { | |
| "epoch": 12.0, | |
| "learning_rate": 5.379834254143646e-06, | |
| "loss": 0.0072, | |
| "step": 6525 | |
| }, | |
| { | |
| "epoch": 12.0, | |
| "learning_rate": 5.360650705954574e-06, | |
| "loss": 0.0079, | |
| "step": 6550 | |
| }, | |
| { | |
| "epoch": 12.01, | |
| "learning_rate": 5.341467157765501e-06, | |
| "loss": 0.0105, | |
| "step": 6575 | |
| }, | |
| { | |
| "epoch": 12.01, | |
| "learning_rate": 5.322283609576427e-06, | |
| "loss": 0.0099, | |
| "step": 6600 | |
| }, | |
| { | |
| "epoch": 12.01, | |
| "learning_rate": 5.303100061387355e-06, | |
| "loss": 0.0068, | |
| "step": 6625 | |
| }, | |
| { | |
| "epoch": 12.01, | |
| "learning_rate": 5.284683855125845e-06, | |
| "loss": 0.0082, | |
| "step": 6650 | |
| }, | |
| { | |
| "epoch": 12.01, | |
| "learning_rate": 5.265500306936772e-06, | |
| "loss": 0.0064, | |
| "step": 6675 | |
| }, | |
| { | |
| "epoch": 12.02, | |
| "learning_rate": 5.246316758747698e-06, | |
| "loss": 0.0062, | |
| "step": 6700 | |
| }, | |
| { | |
| "epoch": 12.02, | |
| "learning_rate": 5.227133210558626e-06, | |
| "loss": 0.0058, | |
| "step": 6725 | |
| }, | |
| { | |
| "epoch": 12.02, | |
| "learning_rate": 5.2079496623695525e-06, | |
| "loss": 0.0069, | |
| "step": 6750 | |
| }, | |
| { | |
| "epoch": 12.02, | |
| "learning_rate": 5.188766114180479e-06, | |
| "loss": 0.0068, | |
| "step": 6775 | |
| }, | |
| { | |
| "epoch": 12.02, | |
| "learning_rate": 5.169582565991407e-06, | |
| "loss": 0.0067, | |
| "step": 6800 | |
| }, | |
| { | |
| "epoch": 12.02, | |
| "learning_rate": 5.150399017802333e-06, | |
| "loss": 0.0053, | |
| "step": 6825 | |
| }, | |
| { | |
| "epoch": 12.03, | |
| "learning_rate": 5.13121546961326e-06, | |
| "loss": 0.0086, | |
| "step": 6850 | |
| }, | |
| { | |
| "epoch": 12.03, | |
| "learning_rate": 5.112031921424188e-06, | |
| "loss": 0.0106, | |
| "step": 6875 | |
| }, | |
| { | |
| "epoch": 12.03, | |
| "learning_rate": 5.092848373235114e-06, | |
| "loss": 0.0065, | |
| "step": 6900 | |
| }, | |
| { | |
| "epoch": 12.03, | |
| "learning_rate": 5.0736648250460405e-06, | |
| "loss": 0.0087, | |
| "step": 6925 | |
| }, | |
| { | |
| "epoch": 12.03, | |
| "learning_rate": 5.0544812768569685e-06, | |
| "loss": 0.0106, | |
| "step": 6950 | |
| }, | |
| { | |
| "epoch": 12.04, | |
| "learning_rate": 5.035297728667895e-06, | |
| "loss": 0.0071, | |
| "step": 6975 | |
| }, | |
| { | |
| "epoch": 12.04, | |
| "learning_rate": 5.016114180478821e-06, | |
| "loss": 0.0057, | |
| "step": 7000 | |
| }, | |
| { | |
| "epoch": 12.04, | |
| "eval_loss": 4.665933609008789, | |
| "eval_runtime": 1259.3566, | |
| "eval_samples_per_second": 1.013, | |
| "eval_steps_per_second": 0.507, | |
| "eval_wer": 161.49515459160128, | |
| "step": 7000 | |
| }, | |
| { | |
| "epoch": 12.04, | |
| "learning_rate": 4.9969306322897485e-06, | |
| "loss": 0.0071, | |
| "step": 7025 | |
| }, | |
| { | |
| "epoch": 13.0, | |
| "learning_rate": 4.977747084100676e-06, | |
| "loss": 0.0056, | |
| "step": 7050 | |
| }, | |
| { | |
| "epoch": 13.0, | |
| "learning_rate": 4.958563535911602e-06, | |
| "loss": 0.0048, | |
| "step": 7075 | |
| }, | |
| { | |
| "epoch": 13.0, | |
| "learning_rate": 4.939379987722529e-06, | |
| "loss": 0.0089, | |
| "step": 7100 | |
| }, | |
| { | |
| "epoch": 13.01, | |
| "learning_rate": 4.9201964395334565e-06, | |
| "loss": 0.0064, | |
| "step": 7125 | |
| }, | |
| { | |
| "epoch": 13.01, | |
| "learning_rate": 4.901012891344383e-06, | |
| "loss": 0.0027, | |
| "step": 7150 | |
| }, | |
| { | |
| "epoch": 13.01, | |
| "learning_rate": 4.88182934315531e-06, | |
| "loss": 0.0056, | |
| "step": 7175 | |
| }, | |
| { | |
| "epoch": 13.01, | |
| "learning_rate": 4.862645794966237e-06, | |
| "loss": 0.0067, | |
| "step": 7200 | |
| }, | |
| { | |
| "epoch": 13.01, | |
| "learning_rate": 4.8434622467771645e-06, | |
| "loss": 0.0043, | |
| "step": 7225 | |
| }, | |
| { | |
| "epoch": 13.02, | |
| "learning_rate": 4.824278698588091e-06, | |
| "loss": 0.0037, | |
| "step": 7250 | |
| }, | |
| { | |
| "epoch": 13.02, | |
| "learning_rate": 4.805095150399018e-06, | |
| "loss": 0.0052, | |
| "step": 7275 | |
| }, | |
| { | |
| "epoch": 13.02, | |
| "learning_rate": 4.785911602209945e-06, | |
| "loss": 0.0058, | |
| "step": 7300 | |
| }, | |
| { | |
| "epoch": 13.02, | |
| "learning_rate": 4.7667280540208725e-06, | |
| "loss": 0.0043, | |
| "step": 7325 | |
| }, | |
| { | |
| "epoch": 13.02, | |
| "learning_rate": 4.747544505831799e-06, | |
| "loss": 0.0066, | |
| "step": 7350 | |
| }, | |
| { | |
| "epoch": 13.03, | |
| "learning_rate": 4.728360957642726e-06, | |
| "loss": 0.0084, | |
| "step": 7375 | |
| }, | |
| { | |
| "epoch": 13.03, | |
| "learning_rate": 4.709177409453653e-06, | |
| "loss": 0.0074, | |
| "step": 7400 | |
| }, | |
| { | |
| "epoch": 13.03, | |
| "learning_rate": 4.68999386126458e-06, | |
| "loss": 0.0029, | |
| "step": 7425 | |
| }, | |
| { | |
| "epoch": 13.03, | |
| "learning_rate": 4.670810313075507e-06, | |
| "loss": 0.0073, | |
| "step": 7450 | |
| }, | |
| { | |
| "epoch": 13.03, | |
| "learning_rate": 4.651626764886434e-06, | |
| "loss": 0.0068, | |
| "step": 7475 | |
| }, | |
| { | |
| "epoch": 13.03, | |
| "learning_rate": 4.632443216697361e-06, | |
| "loss": 0.0064, | |
| "step": 7500 | |
| }, | |
| { | |
| "epoch": 13.04, | |
| "learning_rate": 4.613259668508288e-06, | |
| "loss": 0.0089, | |
| "step": 7525 | |
| }, | |
| { | |
| "epoch": 13.04, | |
| "learning_rate": 4.594076120319215e-06, | |
| "loss": 0.0063, | |
| "step": 7550 | |
| }, | |
| { | |
| "epoch": 14.0, | |
| "learning_rate": 4.574892572130142e-06, | |
| "loss": 0.0068, | |
| "step": 7575 | |
| }, | |
| { | |
| "epoch": 14.0, | |
| "learning_rate": 4.5557090239410685e-06, | |
| "loss": 0.0079, | |
| "step": 7600 | |
| }, | |
| { | |
| "epoch": 14.0, | |
| "learning_rate": 4.536525475751996e-06, | |
| "loss": 0.0055, | |
| "step": 7625 | |
| }, | |
| { | |
| "epoch": 14.01, | |
| "learning_rate": 4.517341927562923e-06, | |
| "loss": 0.0048, | |
| "step": 7650 | |
| }, | |
| { | |
| "epoch": 14.01, | |
| "learning_rate": 4.498158379373849e-06, | |
| "loss": 0.0025, | |
| "step": 7675 | |
| }, | |
| { | |
| "epoch": 14.01, | |
| "learning_rate": 4.4789748311847765e-06, | |
| "loss": 0.0056, | |
| "step": 7700 | |
| }, | |
| { | |
| "epoch": 14.01, | |
| "learning_rate": 4.459791282995704e-06, | |
| "loss": 0.0025, | |
| "step": 7725 | |
| }, | |
| { | |
| "epoch": 14.01, | |
| "learning_rate": 4.44060773480663e-06, | |
| "loss": 0.0031, | |
| "step": 7750 | |
| }, | |
| { | |
| "epoch": 14.01, | |
| "learning_rate": 4.421424186617557e-06, | |
| "loss": 0.0067, | |
| "step": 7775 | |
| }, | |
| { | |
| "epoch": 14.02, | |
| "learning_rate": 4.4022406384284845e-06, | |
| "loss": 0.0049, | |
| "step": 7800 | |
| }, | |
| { | |
| "epoch": 14.02, | |
| "learning_rate": 4.383057090239411e-06, | |
| "loss": 0.0042, | |
| "step": 7825 | |
| }, | |
| { | |
| "epoch": 14.02, | |
| "learning_rate": 4.363873542050338e-06, | |
| "loss": 0.0065, | |
| "step": 7850 | |
| }, | |
| { | |
| "epoch": 14.02, | |
| "learning_rate": 4.3446899938612645e-06, | |
| "loss": 0.002, | |
| "step": 7875 | |
| }, | |
| { | |
| "epoch": 14.02, | |
| "learning_rate": 4.325506445672192e-06, | |
| "loss": 0.0061, | |
| "step": 7900 | |
| }, | |
| { | |
| "epoch": 14.03, | |
| "learning_rate": 4.306322897483119e-06, | |
| "loss": 0.0049, | |
| "step": 7925 | |
| }, | |
| { | |
| "epoch": 14.03, | |
| "learning_rate": 4.287139349294045e-06, | |
| "loss": 0.0031, | |
| "step": 7950 | |
| }, | |
| { | |
| "epoch": 14.03, | |
| "learning_rate": 4.2679558011049725e-06, | |
| "loss": 0.002, | |
| "step": 7975 | |
| }, | |
| { | |
| "epoch": 14.03, | |
| "learning_rate": 4.2487722529159e-06, | |
| "loss": 0.0061, | |
| "step": 8000 | |
| }, | |
| { | |
| "epoch": 14.03, | |
| "eval_loss": 4.916222095489502, | |
| "eval_runtime": 1213.5475, | |
| "eval_samples_per_second": 1.051, | |
| "eval_steps_per_second": 0.526, | |
| "eval_wer": 161.0613751730503, | |
| "step": 8000 | |
| }, | |
| { | |
| "epoch": 14.03, | |
| "learning_rate": 4.229588704726826e-06, | |
| "loss": 0.0033, | |
| "step": 8025 | |
| }, | |
| { | |
| "epoch": 14.04, | |
| "learning_rate": 4.210405156537753e-06, | |
| "loss": 0.0022, | |
| "step": 8050 | |
| }, | |
| { | |
| "epoch": 14.04, | |
| "learning_rate": 4.1912216083486805e-06, | |
| "loss": 0.003, | |
| "step": 8075 | |
| }, | |
| { | |
| "epoch": 14.04, | |
| "learning_rate": 4.172038060159607e-06, | |
| "loss": 0.0047, | |
| "step": 8100 | |
| }, | |
| { | |
| "epoch": 15.0, | |
| "learning_rate": 4.152854511970534e-06, | |
| "loss": 0.0062, | |
| "step": 8125 | |
| }, | |
| { | |
| "epoch": 15.0, | |
| "learning_rate": 4.133670963781461e-06, | |
| "loss": 0.003, | |
| "step": 8150 | |
| }, | |
| { | |
| "epoch": 15.0, | |
| "learning_rate": 4.114487415592388e-06, | |
| "loss": 0.0037, | |
| "step": 8175 | |
| }, | |
| { | |
| "epoch": 15.01, | |
| "learning_rate": 4.095303867403315e-06, | |
| "loss": 0.0024, | |
| "step": 8200 | |
| }, | |
| { | |
| "epoch": 15.01, | |
| "learning_rate": 4.076120319214242e-06, | |
| "loss": 0.0027, | |
| "step": 8225 | |
| }, | |
| { | |
| "epoch": 15.01, | |
| "learning_rate": 4.056936771025169e-06, | |
| "loss": 0.0028, | |
| "step": 8250 | |
| }, | |
| { | |
| "epoch": 15.01, | |
| "learning_rate": 4.037753222836096e-06, | |
| "loss": 0.005, | |
| "step": 8275 | |
| }, | |
| { | |
| "epoch": 15.01, | |
| "learning_rate": 4.018569674647023e-06, | |
| "loss": 0.0041, | |
| "step": 8300 | |
| }, | |
| { | |
| "epoch": 15.02, | |
| "learning_rate": 3.99938612645795e-06, | |
| "loss": 0.0047, | |
| "step": 8325 | |
| }, | |
| { | |
| "epoch": 15.02, | |
| "learning_rate": 3.9802025782688765e-06, | |
| "loss": 0.0029, | |
| "step": 8350 | |
| }, | |
| { | |
| "epoch": 15.02, | |
| "learning_rate": 3.961019030079804e-06, | |
| "loss": 0.004, | |
| "step": 8375 | |
| }, | |
| { | |
| "epoch": 15.02, | |
| "learning_rate": 3.941835481890731e-06, | |
| "loss": 0.0037, | |
| "step": 8400 | |
| }, | |
| { | |
| "epoch": 15.02, | |
| "learning_rate": 3.922651933701658e-06, | |
| "loss": 0.0026, | |
| "step": 8425 | |
| }, | |
| { | |
| "epoch": 15.02, | |
| "learning_rate": 3.9034683855125845e-06, | |
| "loss": 0.0017, | |
| "step": 8450 | |
| }, | |
| { | |
| "epoch": 15.03, | |
| "learning_rate": 3.884284837323512e-06, | |
| "loss": 0.0039, | |
| "step": 8475 | |
| }, | |
| { | |
| "epoch": 15.03, | |
| "learning_rate": 3.865101289134439e-06, | |
| "loss": 0.0034, | |
| "step": 8500 | |
| }, | |
| { | |
| "epoch": 15.03, | |
| "learning_rate": 3.845917740945366e-06, | |
| "loss": 0.0045, | |
| "step": 8525 | |
| }, | |
| { | |
| "epoch": 15.03, | |
| "learning_rate": 3.8267341927562925e-06, | |
| "loss": 0.0035, | |
| "step": 8550 | |
| }, | |
| { | |
| "epoch": 15.03, | |
| "learning_rate": 3.8075506445672193e-06, | |
| "loss": 0.0022, | |
| "step": 8575 | |
| }, | |
| { | |
| "epoch": 15.04, | |
| "learning_rate": 3.7883670963781465e-06, | |
| "loss": 0.0032, | |
| "step": 8600 | |
| }, | |
| { | |
| "epoch": 15.04, | |
| "learning_rate": 3.7691835481890733e-06, | |
| "loss": 0.0061, | |
| "step": 8625 | |
| }, | |
| { | |
| "epoch": 15.04, | |
| "learning_rate": 3.7500000000000005e-06, | |
| "loss": 0.0048, | |
| "step": 8650 | |
| }, | |
| { | |
| "epoch": 16.0, | |
| "learning_rate": 3.7308164518109273e-06, | |
| "loss": 0.0042, | |
| "step": 8675 | |
| }, | |
| { | |
| "epoch": 16.0, | |
| "learning_rate": 3.711632903621854e-06, | |
| "loss": 0.0048, | |
| "step": 8700 | |
| }, | |
| { | |
| "epoch": 16.01, | |
| "learning_rate": 3.6924493554327813e-06, | |
| "loss": 0.0017, | |
| "step": 8725 | |
| }, | |
| { | |
| "epoch": 16.01, | |
| "learning_rate": 3.673265807243708e-06, | |
| "loss": 0.0046, | |
| "step": 8750 | |
| }, | |
| { | |
| "epoch": 16.01, | |
| "learning_rate": 3.654082259054635e-06, | |
| "loss": 0.0018, | |
| "step": 8775 | |
| }, | |
| { | |
| "epoch": 16.01, | |
| "learning_rate": 3.634898710865562e-06, | |
| "loss": 0.0033, | |
| "step": 8800 | |
| }, | |
| { | |
| "epoch": 16.01, | |
| "learning_rate": 3.6157151626764893e-06, | |
| "loss": 0.0021, | |
| "step": 8825 | |
| }, | |
| { | |
| "epoch": 16.01, | |
| "learning_rate": 3.5965316144874157e-06, | |
| "loss": 0.0019, | |
| "step": 8850 | |
| }, | |
| { | |
| "epoch": 16.02, | |
| "learning_rate": 3.577348066298343e-06, | |
| "loss": 0.0033, | |
| "step": 8875 | |
| }, | |
| { | |
| "epoch": 16.02, | |
| "learning_rate": 3.55816451810927e-06, | |
| "loss": 0.0027, | |
| "step": 8900 | |
| }, | |
| { | |
| "epoch": 16.02, | |
| "learning_rate": 3.5389809699201965e-06, | |
| "loss": 0.0022, | |
| "step": 8925 | |
| }, | |
| { | |
| "epoch": 16.02, | |
| "learning_rate": 3.5197974217311237e-06, | |
| "loss": 0.0031, | |
| "step": 8950 | |
| }, | |
| { | |
| "epoch": 16.02, | |
| "learning_rate": 3.500613873542051e-06, | |
| "loss": 0.003, | |
| "step": 8975 | |
| }, | |
| { | |
| "epoch": 16.03, | |
| "learning_rate": 3.4814303253529773e-06, | |
| "loss": 0.0042, | |
| "step": 9000 | |
| }, | |
| { | |
| "epoch": 16.03, | |
| "eval_loss": 5.020518779754639, | |
| "eval_runtime": 1262.2291, | |
| "eval_samples_per_second": 1.011, | |
| "eval_steps_per_second": 0.505, | |
| "eval_wer": 158.98477157360406, | |
| "step": 9000 | |
| }, | |
| { | |
| "epoch": 16.03, | |
| "learning_rate": 3.4622467771639045e-06, | |
| "loss": 0.0033, | |
| "step": 9025 | |
| }, | |
| { | |
| "epoch": 16.03, | |
| "learning_rate": 3.4430632289748317e-06, | |
| "loss": 0.0026, | |
| "step": 9050 | |
| }, | |
| { | |
| "epoch": 16.03, | |
| "learning_rate": 3.423879680785758e-06, | |
| "loss": 0.0036, | |
| "step": 9075 | |
| }, | |
| { | |
| "epoch": 16.03, | |
| "learning_rate": 3.4046961325966853e-06, | |
| "loss": 0.0018, | |
| "step": 9100 | |
| }, | |
| { | |
| "epoch": 16.03, | |
| "learning_rate": 3.3855125844076125e-06, | |
| "loss": 0.0015, | |
| "step": 9125 | |
| }, | |
| { | |
| "epoch": 16.04, | |
| "learning_rate": 3.3663290362185393e-06, | |
| "loss": 0.0023, | |
| "step": 9150 | |
| }, | |
| { | |
| "epoch": 16.04, | |
| "learning_rate": 3.347145488029466e-06, | |
| "loss": 0.0014, | |
| "step": 9175 | |
| }, | |
| { | |
| "epoch": 17.0, | |
| "learning_rate": 3.3279619398403933e-06, | |
| "loss": 0.0043, | |
| "step": 9200 | |
| }, | |
| { | |
| "epoch": 17.0, | |
| "learning_rate": 3.30877839165132e-06, | |
| "loss": 0.0031, | |
| "step": 9225 | |
| }, | |
| { | |
| "epoch": 17.0, | |
| "learning_rate": 3.2895948434622473e-06, | |
| "loss": 0.0011, | |
| "step": 9250 | |
| }, | |
| { | |
| "epoch": 17.01, | |
| "learning_rate": 3.270411295273174e-06, | |
| "loss": 0.0006, | |
| "step": 9275 | |
| }, | |
| { | |
| "epoch": 17.01, | |
| "learning_rate": 3.251227747084101e-06, | |
| "loss": 0.0014, | |
| "step": 9300 | |
| }, | |
| { | |
| "epoch": 17.01, | |
| "learning_rate": 3.232044198895028e-06, | |
| "loss": 0.0011, | |
| "step": 9325 | |
| }, | |
| { | |
| "epoch": 17.01, | |
| "learning_rate": 3.212860650705955e-06, | |
| "loss": 0.0013, | |
| "step": 9350 | |
| }, | |
| { | |
| "epoch": 17.01, | |
| "learning_rate": 3.1936771025168817e-06, | |
| "loss": 0.0009, | |
| "step": 9375 | |
| }, | |
| { | |
| "epoch": 17.02, | |
| "learning_rate": 3.174493554327809e-06, | |
| "loss": 0.0008, | |
| "step": 9400 | |
| }, | |
| { | |
| "epoch": 17.02, | |
| "learning_rate": 3.155310006138736e-06, | |
| "loss": 0.0013, | |
| "step": 9425 | |
| }, | |
| { | |
| "epoch": 17.02, | |
| "learning_rate": 3.1361264579496625e-06, | |
| "loss": 0.0009, | |
| "step": 9450 | |
| }, | |
| { | |
| "epoch": 17.02, | |
| "learning_rate": 3.1169429097605897e-06, | |
| "loss": 0.0008, | |
| "step": 9475 | |
| }, | |
| { | |
| "epoch": 17.02, | |
| "learning_rate": 3.097759361571516e-06, | |
| "loss": 0.0004, | |
| "step": 9500 | |
| }, | |
| { | |
| "epoch": 17.02, | |
| "learning_rate": 3.0785758133824433e-06, | |
| "loss": 0.0016, | |
| "step": 9525 | |
| }, | |
| { | |
| "epoch": 17.03, | |
| "learning_rate": 3.0593922651933705e-06, | |
| "loss": 0.0008, | |
| "step": 9550 | |
| }, | |
| { | |
| "epoch": 17.03, | |
| "learning_rate": 3.0402087170042973e-06, | |
| "loss": 0.0017, | |
| "step": 9575 | |
| }, | |
| { | |
| "epoch": 17.03, | |
| "learning_rate": 3.021025168815224e-06, | |
| "loss": 0.0019, | |
| "step": 9600 | |
| }, | |
| { | |
| "epoch": 17.03, | |
| "learning_rate": 3.0018416206261513e-06, | |
| "loss": 0.003, | |
| "step": 9625 | |
| }, | |
| { | |
| "epoch": 17.03, | |
| "learning_rate": 2.982658072437078e-06, | |
| "loss": 0.0017, | |
| "step": 9650 | |
| }, | |
| { | |
| "epoch": 17.04, | |
| "learning_rate": 2.963474524248005e-06, | |
| "loss": 0.0011, | |
| "step": 9675 | |
| }, | |
| { | |
| "epoch": 17.04, | |
| "learning_rate": 2.944290976058932e-06, | |
| "loss": 0.0006, | |
| "step": 9700 | |
| }, | |
| { | |
| "epoch": 17.04, | |
| "learning_rate": 2.925107427869859e-06, | |
| "loss": 0.0023, | |
| "step": 9725 | |
| }, | |
| { | |
| "epoch": 18.0, | |
| "learning_rate": 2.905923879680786e-06, | |
| "loss": 0.0006, | |
| "step": 9750 | |
| }, | |
| { | |
| "epoch": 18.0, | |
| "learning_rate": 2.886740331491713e-06, | |
| "loss": 0.0015, | |
| "step": 9775 | |
| }, | |
| { | |
| "epoch": 18.0, | |
| "learning_rate": 2.8675567833026397e-06, | |
| "loss": 0.001, | |
| "step": 9800 | |
| }, | |
| { | |
| "epoch": 18.01, | |
| "learning_rate": 2.848373235113567e-06, | |
| "loss": 0.0005, | |
| "step": 9825 | |
| }, | |
| { | |
| "epoch": 18.01, | |
| "learning_rate": 2.829189686924494e-06, | |
| "loss": 0.0011, | |
| "step": 9850 | |
| }, | |
| { | |
| "epoch": 18.01, | |
| "learning_rate": 2.8100061387354205e-06, | |
| "loss": 0.0003, | |
| "step": 9875 | |
| }, | |
| { | |
| "epoch": 18.01, | |
| "learning_rate": 2.7908225905463477e-06, | |
| "loss": 0.001, | |
| "step": 9900 | |
| }, | |
| { | |
| "epoch": 18.01, | |
| "learning_rate": 2.771639042357275e-06, | |
| "loss": 0.0003, | |
| "step": 9925 | |
| }, | |
| { | |
| "epoch": 18.02, | |
| "learning_rate": 2.7524554941682013e-06, | |
| "loss": 0.0008, | |
| "step": 9950 | |
| }, | |
| { | |
| "epoch": 18.02, | |
| "learning_rate": 2.7332719459791285e-06, | |
| "loss": 0.0003, | |
| "step": 9975 | |
| }, | |
| { | |
| "epoch": 18.02, | |
| "learning_rate": 2.7140883977900557e-06, | |
| "loss": 0.0007, | |
| "step": 10000 | |
| }, | |
| { | |
| "epoch": 18.02, | |
| "eval_loss": 5.146338939666748, | |
| "eval_runtime": 1239.053, | |
| "eval_samples_per_second": 1.03, | |
| "eval_steps_per_second": 0.515, | |
| "eval_wer": 159.19704660821412, | |
| "step": 10000 | |
| }, | |
| { | |
| "epoch": 18.02, | |
| "learning_rate": 2.694904849600982e-06, | |
| "loss": 0.0013, | |
| "step": 10025 | |
| }, | |
| { | |
| "epoch": 18.02, | |
| "learning_rate": 2.6757213014119093e-06, | |
| "loss": 0.0007, | |
| "step": 10050 | |
| }, | |
| { | |
| "epoch": 18.02, | |
| "learning_rate": 2.6565377532228365e-06, | |
| "loss": 0.0003, | |
| "step": 10075 | |
| }, | |
| { | |
| "epoch": 18.03, | |
| "learning_rate": 2.637354205033763e-06, | |
| "loss": 0.0017, | |
| "step": 10100 | |
| }, | |
| { | |
| "epoch": 18.03, | |
| "learning_rate": 2.61817065684469e-06, | |
| "loss": 0.0007, | |
| "step": 10125 | |
| }, | |
| { | |
| "epoch": 18.03, | |
| "learning_rate": 2.5989871086556173e-06, | |
| "loss": 0.0004, | |
| "step": 10150 | |
| }, | |
| { | |
| "epoch": 18.03, | |
| "learning_rate": 2.579803560466544e-06, | |
| "loss": 0.0005, | |
| "step": 10175 | |
| }, | |
| { | |
| "epoch": 18.03, | |
| "learning_rate": 2.560620012277471e-06, | |
| "loss": 0.0005, | |
| "step": 10200 | |
| }, | |
| { | |
| "epoch": 18.04, | |
| "learning_rate": 2.541436464088398e-06, | |
| "loss": 0.0009, | |
| "step": 10225 | |
| }, | |
| { | |
| "epoch": 18.04, | |
| "learning_rate": 2.522252915899325e-06, | |
| "loss": 0.0006, | |
| "step": 10250 | |
| }, | |
| { | |
| "epoch": 18.04, | |
| "learning_rate": 2.5030693677102517e-06, | |
| "loss": 0.0005, | |
| "step": 10275 | |
| }, | |
| { | |
| "epoch": 19.0, | |
| "learning_rate": 2.483885819521179e-06, | |
| "loss": 0.0003, | |
| "step": 10300 | |
| }, | |
| { | |
| "epoch": 19.0, | |
| "learning_rate": 2.4647022713321057e-06, | |
| "loss": 0.0022, | |
| "step": 10325 | |
| }, | |
| { | |
| "epoch": 19.01, | |
| "learning_rate": 2.445518723143033e-06, | |
| "loss": 0.0025, | |
| "step": 10350 | |
| }, | |
| { | |
| "epoch": 19.01, | |
| "learning_rate": 2.4263351749539597e-06, | |
| "loss": 0.0004, | |
| "step": 10375 | |
| }, | |
| { | |
| "epoch": 19.01, | |
| "learning_rate": 2.407151626764887e-06, | |
| "loss": 0.0014, | |
| "step": 10400 | |
| }, | |
| { | |
| "epoch": 19.01, | |
| "learning_rate": 2.3879680785758137e-06, | |
| "loss": 0.0002, | |
| "step": 10425 | |
| }, | |
| { | |
| "epoch": 19.01, | |
| "learning_rate": 2.3687845303867405e-06, | |
| "loss": 0.0015, | |
| "step": 10450 | |
| }, | |
| { | |
| "epoch": 19.01, | |
| "learning_rate": 2.3496009821976677e-06, | |
| "loss": 0.0003, | |
| "step": 10475 | |
| }, | |
| { | |
| "epoch": 19.02, | |
| "learning_rate": 2.3304174340085945e-06, | |
| "loss": 0.0014, | |
| "step": 10500 | |
| }, | |
| { | |
| "epoch": 19.02, | |
| "learning_rate": 2.3112338858195213e-06, | |
| "loss": 0.0007, | |
| "step": 10525 | |
| }, | |
| { | |
| "epoch": 19.02, | |
| "learning_rate": 2.2920503376304485e-06, | |
| "loss": 0.0007, | |
| "step": 10550 | |
| }, | |
| { | |
| "epoch": 19.02, | |
| "learning_rate": 2.2728667894413753e-06, | |
| "loss": 0.0001, | |
| "step": 10575 | |
| }, | |
| { | |
| "epoch": 19.02, | |
| "learning_rate": 2.253683241252302e-06, | |
| "loss": 0.0004, | |
| "step": 10600 | |
| }, | |
| { | |
| "epoch": 19.03, | |
| "learning_rate": 2.2344996930632293e-06, | |
| "loss": 0.0006, | |
| "step": 10625 | |
| }, | |
| { | |
| "epoch": 19.03, | |
| "learning_rate": 2.215316144874156e-06, | |
| "loss": 0.0004, | |
| "step": 10650 | |
| }, | |
| { | |
| "epoch": 19.03, | |
| "learning_rate": 2.196132596685083e-06, | |
| "loss": 0.0007, | |
| "step": 10675 | |
| }, | |
| { | |
| "epoch": 19.03, | |
| "learning_rate": 2.17694904849601e-06, | |
| "loss": 0.0002, | |
| "step": 10700 | |
| }, | |
| { | |
| "epoch": 19.03, | |
| "learning_rate": 2.157765500306937e-06, | |
| "loss": 0.0005, | |
| "step": 10725 | |
| }, | |
| { | |
| "epoch": 19.03, | |
| "learning_rate": 2.1385819521178637e-06, | |
| "loss": 0.0002, | |
| "step": 10750 | |
| }, | |
| { | |
| "epoch": 19.04, | |
| "learning_rate": 2.119398403928791e-06, | |
| "loss": 0.0006, | |
| "step": 10775 | |
| }, | |
| { | |
| "epoch": 19.04, | |
| "learning_rate": 2.1002148557397177e-06, | |
| "loss": 0.0012, | |
| "step": 10800 | |
| }, | |
| { | |
| "epoch": 20.0, | |
| "learning_rate": 2.0810313075506445e-06, | |
| "loss": 0.0004, | |
| "step": 10825 | |
| }, | |
| { | |
| "epoch": 20.0, | |
| "learning_rate": 2.0618477593615717e-06, | |
| "loss": 0.0015, | |
| "step": 10850 | |
| }, | |
| { | |
| "epoch": 20.0, | |
| "learning_rate": 2.0426642111724985e-06, | |
| "loss": 0.001, | |
| "step": 10875 | |
| }, | |
| { | |
| "epoch": 20.01, | |
| "learning_rate": 2.0234806629834257e-06, | |
| "loss": 0.0002, | |
| "step": 10900 | |
| }, | |
| { | |
| "epoch": 20.01, | |
| "learning_rate": 2.0042971147943525e-06, | |
| "loss": 0.001, | |
| "step": 10925 | |
| }, | |
| { | |
| "epoch": 20.01, | |
| "learning_rate": 1.9851135666052797e-06, | |
| "loss": 0.0011, | |
| "step": 10950 | |
| }, | |
| { | |
| "epoch": 20.01, | |
| "learning_rate": 1.9659300184162065e-06, | |
| "loss": 0.0001, | |
| "step": 10975 | |
| }, | |
| { | |
| "epoch": 20.01, | |
| "learning_rate": 1.9467464702271333e-06, | |
| "loss": 0.0015, | |
| "step": 11000 | |
| }, | |
| { | |
| "epoch": 20.01, | |
| "eval_loss": 5.215046405792236, | |
| "eval_runtime": 1250.2971, | |
| "eval_samples_per_second": 1.021, | |
| "eval_steps_per_second": 0.51, | |
| "eval_wer": 159.0401476695893, | |
| "step": 11000 | |
| }, | |
| { | |
| "epoch": 20.02, | |
| "learning_rate": 1.9275629220380605e-06, | |
| "loss": 0.0012, | |
| "step": 11025 | |
| }, | |
| { | |
| "epoch": 20.02, | |
| "learning_rate": 1.9083793738489873e-06, | |
| "loss": 0.0015, | |
| "step": 11050 | |
| }, | |
| { | |
| "epoch": 20.02, | |
| "learning_rate": 1.889195825659914e-06, | |
| "loss": 0.0018, | |
| "step": 11075 | |
| }, | |
| { | |
| "epoch": 20.02, | |
| "learning_rate": 1.8700122774708413e-06, | |
| "loss": 0.0006, | |
| "step": 11100 | |
| }, | |
| { | |
| "epoch": 20.02, | |
| "learning_rate": 1.8508287292817681e-06, | |
| "loss": 0.0002, | |
| "step": 11125 | |
| }, | |
| { | |
| "epoch": 20.02, | |
| "learning_rate": 1.831645181092695e-06, | |
| "loss": 0.0012, | |
| "step": 11150 | |
| }, | |
| { | |
| "epoch": 20.03, | |
| "learning_rate": 1.8124616329036221e-06, | |
| "loss": 0.0008, | |
| "step": 11175 | |
| }, | |
| { | |
| "epoch": 20.03, | |
| "learning_rate": 1.793278084714549e-06, | |
| "loss": 0.0002, | |
| "step": 11200 | |
| }, | |
| { | |
| "epoch": 20.03, | |
| "learning_rate": 1.774094536525476e-06, | |
| "loss": 0.0001, | |
| "step": 11225 | |
| }, | |
| { | |
| "epoch": 20.03, | |
| "learning_rate": 1.754910988336403e-06, | |
| "loss": 0.0008, | |
| "step": 11250 | |
| }, | |
| { | |
| "epoch": 20.03, | |
| "learning_rate": 1.7357274401473297e-06, | |
| "loss": 0.0003, | |
| "step": 11275 | |
| }, | |
| { | |
| "epoch": 20.04, | |
| "learning_rate": 1.7165438919582567e-06, | |
| "loss": 0.0013, | |
| "step": 11300 | |
| }, | |
| { | |
| "epoch": 20.04, | |
| "learning_rate": 1.6973603437691837e-06, | |
| "loss": 0.0002, | |
| "step": 11325 | |
| }, | |
| { | |
| "epoch": 20.04, | |
| "learning_rate": 1.6781767955801107e-06, | |
| "loss": 0.0012, | |
| "step": 11350 | |
| }, | |
| { | |
| "epoch": 21.0, | |
| "learning_rate": 1.6589932473910375e-06, | |
| "loss": 0.0002, | |
| "step": 11375 | |
| }, | |
| { | |
| "epoch": 21.0, | |
| "learning_rate": 1.6398096992019647e-06, | |
| "loss": 0.0001, | |
| "step": 11400 | |
| }, | |
| { | |
| "epoch": 21.0, | |
| "learning_rate": 1.6206261510128915e-06, | |
| "loss": 0.0001, | |
| "step": 11425 | |
| }, | |
| { | |
| "epoch": 21.01, | |
| "learning_rate": 1.6014426028238183e-06, | |
| "loss": 0.0004, | |
| "step": 11450 | |
| }, | |
| { | |
| "epoch": 21.01, | |
| "learning_rate": 1.5822590546347455e-06, | |
| "loss": 0.0003, | |
| "step": 11475 | |
| }, | |
| { | |
| "epoch": 21.01, | |
| "learning_rate": 1.5630755064456723e-06, | |
| "loss": 0.0007, | |
| "step": 11500 | |
| }, | |
| { | |
| "epoch": 21.01, | |
| "learning_rate": 1.5438919582565993e-06, | |
| "loss": 0.0001, | |
| "step": 11525 | |
| }, | |
| { | |
| "epoch": 21.01, | |
| "learning_rate": 1.5247084100675263e-06, | |
| "loss": 0.0004, | |
| "step": 11550 | |
| }, | |
| { | |
| "epoch": 21.02, | |
| "learning_rate": 1.5055248618784531e-06, | |
| "loss": 0.0001, | |
| "step": 11575 | |
| }, | |
| { | |
| "epoch": 21.02, | |
| "learning_rate": 1.4863413136893801e-06, | |
| "loss": 0.0003, | |
| "step": 11600 | |
| }, | |
| { | |
| "epoch": 21.02, | |
| "learning_rate": 1.467157765500307e-06, | |
| "loss": 0.0001, | |
| "step": 11625 | |
| }, | |
| { | |
| "epoch": 21.02, | |
| "learning_rate": 1.4479742173112341e-06, | |
| "loss": 0.0003, | |
| "step": 11650 | |
| }, | |
| { | |
| "epoch": 21.02, | |
| "learning_rate": 1.428790669122161e-06, | |
| "loss": 0.0003, | |
| "step": 11675 | |
| }, | |
| { | |
| "epoch": 21.03, | |
| "learning_rate": 1.4096071209330877e-06, | |
| "loss": 0.0019, | |
| "step": 11700 | |
| }, | |
| { | |
| "epoch": 21.03, | |
| "learning_rate": 1.390423572744015e-06, | |
| "loss": 0.0002, | |
| "step": 11725 | |
| }, | |
| { | |
| "epoch": 21.03, | |
| "learning_rate": 1.3712400245549417e-06, | |
| "loss": 0.0002, | |
| "step": 11750 | |
| }, | |
| { | |
| "epoch": 21.03, | |
| "learning_rate": 1.3520564763658687e-06, | |
| "loss": 0.0002, | |
| "step": 11775 | |
| }, | |
| { | |
| "epoch": 21.03, | |
| "learning_rate": 1.3328729281767957e-06, | |
| "loss": 0.0001, | |
| "step": 11800 | |
| }, | |
| { | |
| "epoch": 21.03, | |
| "learning_rate": 1.3136893799877227e-06, | |
| "loss": 0.0001, | |
| "step": 11825 | |
| }, | |
| { | |
| "epoch": 21.04, | |
| "learning_rate": 1.2945058317986495e-06, | |
| "loss": 0.0007, | |
| "step": 11850 | |
| }, | |
| { | |
| "epoch": 21.04, | |
| "learning_rate": 1.2753222836095765e-06, | |
| "loss": 0.0001, | |
| "step": 11875 | |
| }, | |
| { | |
| "epoch": 21.04, | |
| "learning_rate": 1.2561387354205035e-06, | |
| "loss": 0.0002, | |
| "step": 11900 | |
| }, | |
| { | |
| "epoch": 22.0, | |
| "learning_rate": 1.2369551872314305e-06, | |
| "loss": 0.0003, | |
| "step": 11925 | |
| }, | |
| { | |
| "epoch": 22.0, | |
| "learning_rate": 1.2177716390423573e-06, | |
| "loss": 0.0001, | |
| "step": 11950 | |
| }, | |
| { | |
| "epoch": 22.01, | |
| "learning_rate": 1.1985880908532843e-06, | |
| "loss": 0.0004, | |
| "step": 11975 | |
| }, | |
| { | |
| "epoch": 22.01, | |
| "learning_rate": 1.1794045426642113e-06, | |
| "loss": 0.0001, | |
| "step": 12000 | |
| }, | |
| { | |
| "epoch": 22.01, | |
| "eval_loss": 5.3007731437683105, | |
| "eval_runtime": 1256.4643, | |
| "eval_samples_per_second": 1.016, | |
| "eval_steps_per_second": 0.508, | |
| "eval_wer": 159.54776188278726, | |
| "step": 12000 | |
| }, | |
| { | |
| "epoch": 22.01, | |
| "learning_rate": 1.160220994475138e-06, | |
| "loss": 0.0001, | |
| "step": 12025 | |
| }, | |
| { | |
| "epoch": 22.01, | |
| "learning_rate": 1.1410374462860651e-06, | |
| "loss": 0.0004, | |
| "step": 12050 | |
| }, | |
| { | |
| "epoch": 22.01, | |
| "learning_rate": 1.1218538980969921e-06, | |
| "loss": 0.0001, | |
| "step": 12075 | |
| }, | |
| { | |
| "epoch": 22.01, | |
| "learning_rate": 1.1026703499079191e-06, | |
| "loss": 0.0001, | |
| "step": 12100 | |
| }, | |
| { | |
| "epoch": 22.02, | |
| "learning_rate": 1.0834868017188461e-06, | |
| "loss": 0.0004, | |
| "step": 12125 | |
| }, | |
| { | |
| "epoch": 22.02, | |
| "learning_rate": 1.064303253529773e-06, | |
| "loss": 0.0001, | |
| "step": 12150 | |
| }, | |
| { | |
| "epoch": 22.02, | |
| "learning_rate": 1.0451197053407e-06, | |
| "loss": 0.0001, | |
| "step": 12175 | |
| }, | |
| { | |
| "epoch": 22.02, | |
| "learning_rate": 1.025936157151627e-06, | |
| "loss": 0.0002, | |
| "step": 12200 | |
| }, | |
| { | |
| "epoch": 22.02, | |
| "learning_rate": 1.006752608962554e-06, | |
| "loss": 0.0004, | |
| "step": 12225 | |
| }, | |
| { | |
| "epoch": 22.03, | |
| "learning_rate": 9.875690607734807e-07, | |
| "loss": 0.0004, | |
| "step": 12250 | |
| }, | |
| { | |
| "epoch": 22.03, | |
| "learning_rate": 9.683855125844077e-07, | |
| "loss": 0.0005, | |
| "step": 12275 | |
| }, | |
| { | |
| "epoch": 22.03, | |
| "learning_rate": 9.492019643953346e-07, | |
| "loss": 0.0001, | |
| "step": 12300 | |
| }, | |
| { | |
| "epoch": 22.03, | |
| "learning_rate": 9.300184162062616e-07, | |
| "loss": 0.0005, | |
| "step": 12325 | |
| }, | |
| { | |
| "epoch": 22.03, | |
| "learning_rate": 9.108348680171886e-07, | |
| "loss": 0.0001, | |
| "step": 12350 | |
| }, | |
| { | |
| "epoch": 22.03, | |
| "learning_rate": 8.916513198281154e-07, | |
| "loss": 0.0003, | |
| "step": 12375 | |
| }, | |
| { | |
| "epoch": 22.04, | |
| "learning_rate": 8.724677716390424e-07, | |
| "loss": 0.0003, | |
| "step": 12400 | |
| }, | |
| { | |
| "epoch": 22.04, | |
| "learning_rate": 8.532842234499694e-07, | |
| "loss": 0.0007, | |
| "step": 12425 | |
| }, | |
| { | |
| "epoch": 23.0, | |
| "learning_rate": 8.341006752608963e-07, | |
| "loss": 0.0001, | |
| "step": 12450 | |
| }, | |
| { | |
| "epoch": 23.0, | |
| "learning_rate": 8.149171270718233e-07, | |
| "loss": 0.0001, | |
| "step": 12475 | |
| }, | |
| { | |
| "epoch": 23.0, | |
| "learning_rate": 7.957335788827503e-07, | |
| "loss": 0.0001, | |
| "step": 12500 | |
| }, | |
| { | |
| "epoch": 23.01, | |
| "learning_rate": 7.765500306936771e-07, | |
| "loss": 0.0001, | |
| "step": 12525 | |
| }, | |
| { | |
| "epoch": 23.01, | |
| "learning_rate": 7.573664825046041e-07, | |
| "loss": 0.0001, | |
| "step": 12550 | |
| }, | |
| { | |
| "epoch": 23.01, | |
| "learning_rate": 7.381829343155311e-07, | |
| "loss": 0.0001, | |
| "step": 12575 | |
| }, | |
| { | |
| "epoch": 23.01, | |
| "learning_rate": 7.18999386126458e-07, | |
| "loss": 0.0001, | |
| "step": 12600 | |
| }, | |
| { | |
| "epoch": 23.01, | |
| "learning_rate": 6.99815837937385e-07, | |
| "loss": 0.0007, | |
| "step": 12625 | |
| }, | |
| { | |
| "epoch": 23.02, | |
| "learning_rate": 6.80632289748312e-07, | |
| "loss": 0.0003, | |
| "step": 12650 | |
| }, | |
| { | |
| "epoch": 23.02, | |
| "learning_rate": 6.614487415592388e-07, | |
| "loss": 0.0007, | |
| "step": 12675 | |
| }, | |
| { | |
| "epoch": 23.02, | |
| "learning_rate": 6.422651933701658e-07, | |
| "loss": 0.0002, | |
| "step": 12700 | |
| }, | |
| { | |
| "epoch": 23.02, | |
| "learning_rate": 6.230816451810927e-07, | |
| "loss": 0.0001, | |
| "step": 12725 | |
| }, | |
| { | |
| "epoch": 23.02, | |
| "learning_rate": 6.038980969920197e-07, | |
| "loss": 0.0001, | |
| "step": 12750 | |
| }, | |
| { | |
| "epoch": 23.02, | |
| "learning_rate": 5.847145488029467e-07, | |
| "loss": 0.0001, | |
| "step": 12775 | |
| }, | |
| { | |
| "epoch": 23.03, | |
| "learning_rate": 5.655310006138736e-07, | |
| "loss": 0.0001, | |
| "step": 12800 | |
| }, | |
| { | |
| "epoch": 23.03, | |
| "learning_rate": 5.463474524248005e-07, | |
| "loss": 0.0001, | |
| "step": 12825 | |
| }, | |
| { | |
| "epoch": 23.03, | |
| "learning_rate": 5.271639042357274e-07, | |
| "loss": 0.0001, | |
| "step": 12850 | |
| }, | |
| { | |
| "epoch": 23.03, | |
| "learning_rate": 5.079803560466544e-07, | |
| "loss": 0.0007, | |
| "step": 12875 | |
| }, | |
| { | |
| "epoch": 23.03, | |
| "learning_rate": 4.887968078575814e-07, | |
| "loss": 0.0003, | |
| "step": 12900 | |
| }, | |
| { | |
| "epoch": 23.04, | |
| "learning_rate": 4.696132596685083e-07, | |
| "loss": 0.0002, | |
| "step": 12925 | |
| }, | |
| { | |
| "epoch": 23.04, | |
| "learning_rate": 4.504297114794353e-07, | |
| "loss": 0.0001, | |
| "step": 12950 | |
| }, | |
| { | |
| "epoch": 23.04, | |
| "learning_rate": 4.312461632903622e-07, | |
| "loss": 0.0001, | |
| "step": 12975 | |
| }, | |
| { | |
| "epoch": 24.0, | |
| "learning_rate": 4.1206261510128916e-07, | |
| "loss": 0.0001, | |
| "step": 13000 | |
| }, | |
| { | |
| "epoch": 24.0, | |
| "eval_loss": 5.380012035369873, | |
| "eval_runtime": 1244.961, | |
| "eval_samples_per_second": 1.025, | |
| "eval_steps_per_second": 0.512, | |
| "eval_wer": 159.03091832025842, | |
| "step": 13000 | |
| }, | |
| { | |
| "epoch": 24.0, | |
| "learning_rate": 3.9287906691221606e-07, | |
| "loss": 0.0001, | |
| "step": 13025 | |
| }, | |
| { | |
| "epoch": 24.0, | |
| "learning_rate": 3.7369551872314306e-07, | |
| "loss": 0.0001, | |
| "step": 13050 | |
| }, | |
| { | |
| "epoch": 24.01, | |
| "learning_rate": 3.5451197053407e-07, | |
| "loss": 0.0001, | |
| "step": 13075 | |
| }, | |
| { | |
| "epoch": 24.01, | |
| "learning_rate": 3.353284223449969e-07, | |
| "loss": 0.0002, | |
| "step": 13100 | |
| }, | |
| { | |
| "epoch": 24.01, | |
| "learning_rate": 3.161448741559239e-07, | |
| "loss": 0.0001, | |
| "step": 13125 | |
| }, | |
| { | |
| "epoch": 24.01, | |
| "learning_rate": 2.9696132596685086e-07, | |
| "loss": 0.0001, | |
| "step": 13150 | |
| }, | |
| { | |
| "epoch": 24.01, | |
| "learning_rate": 2.7777777777777776e-07, | |
| "loss": 0.0001, | |
| "step": 13175 | |
| }, | |
| { | |
| "epoch": 24.02, | |
| "learning_rate": 2.5859422958870476e-07, | |
| "loss": 0.0001, | |
| "step": 13200 | |
| }, | |
| { | |
| "epoch": 24.02, | |
| "learning_rate": 2.394106813996317e-07, | |
| "loss": 0.0002, | |
| "step": 13225 | |
| }, | |
| { | |
| "epoch": 24.02, | |
| "learning_rate": 2.2022713321055864e-07, | |
| "loss": 0.0001, | |
| "step": 13250 | |
| }, | |
| { | |
| "epoch": 24.02, | |
| "learning_rate": 2.0104358502148559e-07, | |
| "loss": 0.0001, | |
| "step": 13275 | |
| }, | |
| { | |
| "epoch": 24.02, | |
| "learning_rate": 1.8186003683241254e-07, | |
| "loss": 0.0001, | |
| "step": 13300 | |
| }, | |
| { | |
| "epoch": 24.03, | |
| "learning_rate": 1.626764886433395e-07, | |
| "loss": 0.0001, | |
| "step": 13325 | |
| }, | |
| { | |
| "epoch": 24.03, | |
| "learning_rate": 1.4349294045426644e-07, | |
| "loss": 0.0002, | |
| "step": 13350 | |
| }, | |
| { | |
| "epoch": 24.03, | |
| "learning_rate": 1.243093922651934e-07, | |
| "loss": 0.0001, | |
| "step": 13375 | |
| }, | |
| { | |
| "epoch": 24.03, | |
| "learning_rate": 1.0512584407612032e-07, | |
| "loss": 0.0002, | |
| "step": 13400 | |
| }, | |
| { | |
| "epoch": 24.03, | |
| "learning_rate": 8.594229588704729e-08, | |
| "loss": 0.0001, | |
| "step": 13425 | |
| }, | |
| { | |
| "epoch": 24.03, | |
| "learning_rate": 6.675874769797422e-08, | |
| "loss": 0.0001, | |
| "step": 13450 | |
| }, | |
| { | |
| "epoch": 24.04, | |
| "learning_rate": 4.757519950890117e-08, | |
| "loss": 0.0002, | |
| "step": 13475 | |
| }, | |
| { | |
| "epoch": 24.04, | |
| "learning_rate": 2.8391651319828118e-08, | |
| "loss": 0.0001, | |
| "step": 13500 | |
| }, | |
| { | |
| "epoch": 24.04, | |
| "learning_rate": 9.208103130755065e-09, | |
| "loss": 0.0001, | |
| "step": 13525 | |
| }, | |
| { | |
| "epoch": 25.0, | |
| "step": 13532, | |
| "total_flos": 2.2097321202481208e+20, | |
| "train_loss": 0.1507345416943568, | |
| "train_runtime": 94042.7323, | |
| "train_samples_per_second": 2.302, | |
| "train_steps_per_second": 0.144 | |
| } | |
| ], | |
| "max_steps": 13532, | |
| "num_train_epochs": 9223372036854775807, | |
| "total_flos": 2.2097321202481208e+20, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |