{ "best_metric": 0.3083952069282532, "best_model_checkpoint": "joe611/chickens-60-epoch-1000-images-aug/checkpoint-25000", "epoch": 60.0, "eval_steps": 500, "global_step": 30000, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.06, "grad_norm": 158.5474090576172, "learning_rate": 9.999975326009292e-06, "loss": 2.1723, "step": 30 }, { "epoch": 0.12, "grad_norm": 91.59680938720703, "learning_rate": 9.999901304280686e-06, "loss": 2.3751, "step": 60 }, { "epoch": 0.18, "grad_norm": 122.90123748779297, "learning_rate": 9.99977793554475e-06, "loss": 2.0375, "step": 90 }, { "epoch": 0.24, "grad_norm": 69.5802993774414, "learning_rate": 9.999605221019082e-06, "loss": 1.91, "step": 120 }, { "epoch": 0.3, "grad_norm": 83.56485748291016, "learning_rate": 9.999383162408303e-06, "loss": 2.2316, "step": 150 }, { "epoch": 0.36, "grad_norm": 62.042781829833984, "learning_rate": 9.999111761904046e-06, "loss": 1.8506, "step": 180 }, { "epoch": 0.42, "grad_norm": 35.492462158203125, "learning_rate": 9.998791022184921e-06, "loss": 1.6076, "step": 210 }, { "epoch": 0.48, "grad_norm": 62.26254653930664, "learning_rate": 9.9984209464165e-06, "loss": 1.8313, "step": 240 }, { "epoch": 0.54, "grad_norm": 82.2900161743164, "learning_rate": 9.998001538251283e-06, "loss": 1.7903, "step": 270 }, { "epoch": 0.6, "grad_norm": 83.0863265991211, "learning_rate": 9.997532801828659e-06, "loss": 1.7534, "step": 300 }, { "epoch": 0.66, "grad_norm": 52.50190734863281, "learning_rate": 9.997014741774866e-06, "loss": 1.5151, "step": 330 }, { "epoch": 0.72, "grad_norm": 75.64976501464844, "learning_rate": 9.996447363202947e-06, "loss": 1.3768, "step": 360 }, { "epoch": 0.78, "grad_norm": 88.16222381591797, "learning_rate": 9.995830671712701e-06, "loss": 1.5055, "step": 390 }, { "epoch": 0.84, "grad_norm": 82.0279769897461, "learning_rate": 9.995164673390624e-06, "loss": 1.4557, "step": 420 }, { "epoch": 0.9, "grad_norm": 35.413795471191406, "learning_rate": 9.994449374809851e-06, "loss": 1.179, "step": 450 }, { "epoch": 0.96, "grad_norm": 48.981014251708984, "learning_rate": 9.99368478303009e-06, "loss": 1.3706, "step": 480 }, { "epoch": 1.0, "eval_loss": 1.2724533081054688, "eval_map": 0.1253, "eval_map_50": 0.1716, "eval_map_75": 0.1373, "eval_map_chicken": 0.0161, "eval_map_duck": 0.0, "eval_map_large": 0.5125, "eval_map_medium": 0.0376, "eval_map_plant": 0.3597, "eval_map_small": 0.0085, "eval_mar_1": 0.0765, "eval_mar_10": 0.2147, "eval_mar_100": 0.2609, "eval_mar_100_chicken": 0.048, "eval_mar_100_duck": 0.0, "eval_mar_100_plant": 0.7346, "eval_mar_large": 0.7891, "eval_mar_medium": 0.2327, "eval_mar_small": 0.0688, "eval_runtime": 9.4567, "eval_samples_per_second": 10.574, "eval_steps_per_second": 1.375, "step": 500 }, { "epoch": 1.02, "grad_norm": 70.26484680175781, "learning_rate": 9.992870905597549e-06, "loss": 1.4684, "step": 510 }, { "epoch": 1.08, "grad_norm": 84.59373474121094, "learning_rate": 9.992007750544876e-06, "loss": 1.5438, "step": 540 }, { "epoch": 1.1400000000000001, "grad_norm": 106.29322052001953, "learning_rate": 9.991095326391061e-06, "loss": 1.1743, "step": 570 }, { "epoch": 1.2, "grad_norm": 275.0106201171875, "learning_rate": 9.990133642141359e-06, "loss": 2.012, "step": 600 }, { "epoch": 1.26, "grad_norm": 171.77479553222656, "learning_rate": 9.98912270728721e-06, "loss": 1.6024, "step": 630 }, { "epoch": 1.32, "grad_norm": 84.19279479980469, "learning_rate": 9.988062531806127e-06, "loss": 1.2841, "step": 660 }, { "epoch": 1.38, "grad_norm": 42.48450469970703, "learning_rate": 9.98695312616162e-06, "loss": 1.5178, "step": 690 }, { "epoch": 1.44, "grad_norm": 1216.5023193359375, "learning_rate": 9.98579450130307e-06, "loss": 1.7935, "step": 720 }, { "epoch": 1.5, "grad_norm": 88.66885375976562, "learning_rate": 9.984586668665641e-06, "loss": 1.5139, "step": 750 }, { "epoch": 1.56, "grad_norm": 52.2545280456543, "learning_rate": 9.98332964017015e-06, "loss": 1.7237, "step": 780 }, { "epoch": 1.62, "grad_norm": 62.41621398925781, "learning_rate": 9.982023428222963e-06, "loss": 1.1224, "step": 810 }, { "epoch": 1.6800000000000002, "grad_norm": 2577.698486328125, "learning_rate": 9.980668045715864e-06, "loss": 1.8974, "step": 840 }, { "epoch": 1.74, "grad_norm": 47.874610900878906, "learning_rate": 9.97926350602593e-06, "loss": 1.2659, "step": 870 }, { "epoch": 1.8, "grad_norm": 35.623939514160156, "learning_rate": 9.9778098230154e-06, "loss": 1.3004, "step": 900 }, { "epoch": 1.8599999999999999, "grad_norm": 102.40998077392578, "learning_rate": 9.976307011031542e-06, "loss": 1.7917, "step": 930 }, { "epoch": 1.92, "grad_norm": 88.92515563964844, "learning_rate": 9.974755084906503e-06, "loss": 1.2467, "step": 960 }, { "epoch": 1.98, "grad_norm": 199.63978576660156, "learning_rate": 9.973154059957162e-06, "loss": 1.6487, "step": 990 }, { "epoch": 2.0, "eval_loss": 1.7212539911270142, "eval_map": 0.153, "eval_map_50": 0.2186, "eval_map_75": 0.1681, "eval_map_chicken": 0.0038, "eval_map_duck": 0.0, "eval_map_large": 0.5473, "eval_map_medium": 0.0713, "eval_map_plant": 0.4552, "eval_map_small": 0.0035, "eval_mar_1": 0.0675, "eval_mar_10": 0.196, "eval_mar_100": 0.214, "eval_mar_100_chicken": 0.0044, "eval_mar_100_duck": 0.0, "eval_mar_100_plant": 0.6375, "eval_mar_large": 0.6728, "eval_mar_medium": 0.1976, "eval_mar_small": 0.0667, "eval_runtime": 10.6917, "eval_samples_per_second": 9.353, "eval_steps_per_second": 1.216, "step": 1000 }, { "epoch": 2.04, "grad_norm": 36.61814880371094, "learning_rate": 9.971503951984996e-06, "loss": 1.3124, "step": 1020 }, { "epoch": 2.1, "grad_norm": 90.30917358398438, "learning_rate": 9.9698047772759e-06, "loss": 1.3343, "step": 1050 }, { "epoch": 2.16, "grad_norm": 188.6052703857422, "learning_rate": 9.968056552600043e-06, "loss": 1.1958, "step": 1080 }, { "epoch": 2.22, "grad_norm": 43.872764587402344, "learning_rate": 9.966259295211698e-06, "loss": 1.6775, "step": 1110 }, { "epoch": 2.2800000000000002, "grad_norm": 39.97490692138672, "learning_rate": 9.964413022849069e-06, "loss": 1.3663, "step": 1140 }, { "epoch": 2.34, "grad_norm": 51.550968170166016, "learning_rate": 9.96251775373412e-06, "loss": 1.2775, "step": 1170 }, { "epoch": 2.4, "grad_norm": 611.0650024414062, "learning_rate": 9.960573506572391e-06, "loss": 2.0808, "step": 1200 }, { "epoch": 2.46, "grad_norm": 83.57760620117188, "learning_rate": 9.958580300552816e-06, "loss": 1.0574, "step": 1230 }, { "epoch": 2.52, "grad_norm": 1.1940889358520508, "learning_rate": 9.956538155347534e-06, "loss": 1.5135, "step": 1260 }, { "epoch": 2.58, "grad_norm": 284.9285583496094, "learning_rate": 9.954447091111695e-06, "loss": 1.0702, "step": 1290 }, { "epoch": 2.64, "grad_norm": 92.98126983642578, "learning_rate": 9.952307128483257e-06, "loss": 1.1672, "step": 1320 }, { "epoch": 2.7, "grad_norm": 99.8774185180664, "learning_rate": 9.95011828858279e-06, "loss": 1.1101, "step": 1350 }, { "epoch": 2.76, "grad_norm": 47.6667366027832, "learning_rate": 9.947880593013256e-06, "loss": 1.2761, "step": 1380 }, { "epoch": 2.82, "grad_norm": 112.50123596191406, "learning_rate": 9.94559406385981e-06, "loss": 1.3069, "step": 1410 }, { "epoch": 2.88, "grad_norm": 38.15477752685547, "learning_rate": 9.94325872368957e-06, "loss": 1.2961, "step": 1440 }, { "epoch": 2.94, "grad_norm": 33.77321243286133, "learning_rate": 9.940874595551403e-06, "loss": 1.2302, "step": 1470 }, { "epoch": 3.0, "grad_norm": 47.69813919067383, "learning_rate": 9.938441702975689e-06, "loss": 1.2664, "step": 1500 }, { "epoch": 3.0, "eval_loss": 1.3020857572555542, "eval_map": 0.2125, "eval_map_50": 0.2998, "eval_map_75": 0.2341, "eval_map_chicken": 0.0583, "eval_map_duck": 0.0, "eval_map_large": 0.6488, "eval_map_medium": 0.1766, "eval_map_plant": 0.5792, "eval_map_small": 0.0142, "eval_mar_1": 0.0901, "eval_mar_10": 0.259, "eval_mar_100": 0.2715, "eval_mar_100_chicken": 0.1062, "eval_mar_100_duck": 0.0, "eval_mar_100_plant": 0.7084, "eval_mar_large": 0.7552, "eval_mar_medium": 0.2482, "eval_mar_small": 0.0958, "eval_runtime": 10.2383, "eval_samples_per_second": 9.767, "eval_steps_per_second": 1.27, "step": 1500 }, { "epoch": 3.06, "grad_norm": 118.73700714111328, "learning_rate": 9.935960069974096e-06, "loss": 1.4347, "step": 1530 }, { "epoch": 3.12, "grad_norm": 421.19024658203125, "learning_rate": 9.93342972103934e-06, "loss": 1.0804, "step": 1560 }, { "epoch": 3.18, "grad_norm": 66.69783020019531, "learning_rate": 9.930850681144946e-06, "loss": 1.1705, "step": 1590 }, { "epoch": 3.24, "grad_norm": 332.80474853515625, "learning_rate": 9.928222975744992e-06, "loss": 1.322, "step": 1620 }, { "epoch": 3.3, "grad_norm": 43.3453369140625, "learning_rate": 9.92554663077387e-06, "loss": 1.2478, "step": 1650 }, { "epoch": 3.36, "grad_norm": 45.562705993652344, "learning_rate": 9.922821672646028e-06, "loss": 0.8973, "step": 1680 }, { "epoch": 3.42, "grad_norm": 55.2510986328125, "learning_rate": 9.920048128255699e-06, "loss": 1.0052, "step": 1710 }, { "epoch": 3.48, "grad_norm": 26.767873764038086, "learning_rate": 9.91722602497665e-06, "loss": 1.5015, "step": 1740 }, { "epoch": 3.54, "grad_norm": 97.10417938232422, "learning_rate": 9.914355390661897e-06, "loss": 1.2423, "step": 1770 }, { "epoch": 3.6, "grad_norm": 103.83246612548828, "learning_rate": 9.911436253643445e-06, "loss": 0.7741, "step": 1800 }, { "epoch": 3.66, "grad_norm": 46.123870849609375, "learning_rate": 9.908468642731996e-06, "loss": 1.0517, "step": 1830 }, { "epoch": 3.7199999999999998, "grad_norm": 35.28718948364258, "learning_rate": 9.90545258721667e-06, "loss": 1.2003, "step": 1860 }, { "epoch": 3.7800000000000002, "grad_norm": 47.2177848815918, "learning_rate": 9.902388116864723e-06, "loss": 1.553, "step": 1890 }, { "epoch": 3.84, "grad_norm": 38.933799743652344, "learning_rate": 9.899275261921236e-06, "loss": 1.1955, "step": 1920 }, { "epoch": 3.9, "grad_norm": 34.16486358642578, "learning_rate": 9.89611405310883e-06, "loss": 1.038, "step": 1950 }, { "epoch": 3.96, "grad_norm": 140.69822692871094, "learning_rate": 9.89290452162736e-06, "loss": 1.2003, "step": 1980 }, { "epoch": 4.0, "eval_loss": 1.1970802545547485, "eval_map": 0.2858, "eval_map_50": 0.4094, "eval_map_75": 0.3251, "eval_map_chicken": 0.272, "eval_map_duck": 0.0, "eval_map_large": 0.6565, "eval_map_medium": 0.2552, "eval_map_plant": 0.5854, "eval_map_small": 0.0205, "eval_mar_1": 0.1198, "eval_mar_10": 0.3741, "eval_mar_100": 0.3854, "eval_mar_100_chicken": 0.4738, "eval_mar_100_duck": 0.0, "eval_mar_100_plant": 0.6824, "eval_mar_large": 0.7435, "eval_mar_medium": 0.3614, "eval_mar_small": 0.0708, "eval_runtime": 10.2386, "eval_samples_per_second": 9.767, "eval_steps_per_second": 1.27, "step": 2000 }, { "epoch": 4.02, "grad_norm": 40.8932991027832, "learning_rate": 9.88964669915361e-06, "loss": 1.1757, "step": 2010 }, { "epoch": 4.08, "grad_norm": 116.23714447021484, "learning_rate": 9.886340617840968e-06, "loss": 1.2925, "step": 2040 }, { "epoch": 4.14, "grad_norm": 2107.181640625, "learning_rate": 9.882986310319124e-06, "loss": 1.0512, "step": 2070 }, { "epoch": 4.2, "grad_norm": 37.569637298583984, "learning_rate": 9.879583809693737e-06, "loss": 1.3823, "step": 2100 }, { "epoch": 4.26, "grad_norm": 60.553829193115234, "learning_rate": 9.876133149546117e-06, "loss": 0.9467, "step": 2130 }, { "epoch": 4.32, "grad_norm": 57.90943145751953, "learning_rate": 9.872634363932887e-06, "loss": 1.1448, "step": 2160 }, { "epoch": 4.38, "grad_norm": 35.535518646240234, "learning_rate": 9.869087487385644e-06, "loss": 0.8989, "step": 2190 }, { "epoch": 4.44, "grad_norm": 157.6408233642578, "learning_rate": 9.865492554910634e-06, "loss": 1.1563, "step": 2220 }, { "epoch": 4.5, "grad_norm": 24.22439193725586, "learning_rate": 9.861849601988384e-06, "loss": 0.9229, "step": 2250 }, { "epoch": 4.5600000000000005, "grad_norm": 69.95077514648438, "learning_rate": 9.85815866457337e-06, "loss": 1.1337, "step": 2280 }, { "epoch": 4.62, "grad_norm": 667.6534423828125, "learning_rate": 9.854419779093656e-06, "loss": 0.9563, "step": 2310 }, { "epoch": 4.68, "grad_norm": 38.42517852783203, "learning_rate": 9.85063298245053e-06, "loss": 1.0092, "step": 2340 }, { "epoch": 4.74, "grad_norm": 239.5282440185547, "learning_rate": 9.846798312018147e-06, "loss": 0.9758, "step": 2370 }, { "epoch": 4.8, "grad_norm": 31.474966049194336, "learning_rate": 9.842915805643156e-06, "loss": 0.9971, "step": 2400 }, { "epoch": 4.86, "grad_norm": 52.577083587646484, "learning_rate": 9.838985501644329e-06, "loss": 1.0364, "step": 2430 }, { "epoch": 4.92, "grad_norm": 40.93312072753906, "learning_rate": 9.835007438812177e-06, "loss": 0.9375, "step": 2460 }, { "epoch": 4.98, "grad_norm": 97.95311737060547, "learning_rate": 9.830981656408575e-06, "loss": 0.8706, "step": 2490 }, { "epoch": 5.0, "eval_loss": 1.1443504095077515, "eval_map": 0.3489, "eval_map_50": 0.4855, "eval_map_75": 0.4192, "eval_map_chicken": 0.4099, "eval_map_duck": 0.0, "eval_map_large": 0.6975, "eval_map_medium": 0.3257, "eval_map_plant": 0.6367, "eval_map_small": 0.0186, "eval_mar_1": 0.1342, "eval_mar_10": 0.4497, "eval_mar_100": 0.4605, "eval_mar_100_chicken": 0.6667, "eval_mar_100_duck": 0.0, "eval_mar_100_plant": 0.7147, "eval_mar_large": 0.7611, "eval_mar_medium": 0.4549, "eval_mar_small": 0.0583, "eval_runtime": 10.3073, "eval_samples_per_second": 9.702, "eval_steps_per_second": 1.261, "step": 2500 }, { "epoch": 5.04, "grad_norm": 49.93207931518555, "learning_rate": 9.82690819416637e-06, "loss": 1.158, "step": 2520 }, { "epoch": 5.1, "grad_norm": 64.10286712646484, "learning_rate": 9.822787092288991e-06, "loss": 1.2776, "step": 2550 }, { "epoch": 5.16, "grad_norm": 45.720088958740234, "learning_rate": 9.81861839145005e-06, "loss": 1.0825, "step": 2580 }, { "epoch": 5.22, "grad_norm": 60.9461555480957, "learning_rate": 9.814402132792939e-06, "loss": 0.8561, "step": 2610 }, { "epoch": 5.28, "grad_norm": 46.097652435302734, "learning_rate": 9.81013835793043e-06, "loss": 1.0208, "step": 2640 }, { "epoch": 5.34, "grad_norm": 46.66898727416992, "learning_rate": 9.80582710894426e-06, "loss": 0.911, "step": 2670 }, { "epoch": 5.4, "grad_norm": 41.62497329711914, "learning_rate": 9.801468428384716e-06, "loss": 1.0277, "step": 2700 }, { "epoch": 5.46, "grad_norm": 66.89842987060547, "learning_rate": 9.797062359270215e-06, "loss": 1.0108, "step": 2730 }, { "epoch": 5.52, "grad_norm": 52.27336883544922, "learning_rate": 9.79260894508688e-06, "loss": 0.9219, "step": 2760 }, { "epoch": 5.58, "grad_norm": 26.557716369628906, "learning_rate": 9.788108229788111e-06, "loss": 0.9621, "step": 2790 }, { "epoch": 5.64, "grad_norm": 40.88523483276367, "learning_rate": 9.783560257794153e-06, "loss": 1.5623, "step": 2820 }, { "epoch": 5.7, "grad_norm": 32.03486251831055, "learning_rate": 9.778965073991652e-06, "loss": 1.229, "step": 2850 }, { "epoch": 5.76, "grad_norm": 108.31827545166016, "learning_rate": 9.774322723733216e-06, "loss": 1.0328, "step": 2880 }, { "epoch": 5.82, "grad_norm": 49.63470458984375, "learning_rate": 9.769633252836969e-06, "loss": 0.9535, "step": 2910 }, { "epoch": 5.88, "grad_norm": 75.02542114257812, "learning_rate": 9.764896707586095e-06, "loss": 0.9733, "step": 2940 }, { "epoch": 5.9399999999999995, "grad_norm": 31.322744369506836, "learning_rate": 9.760113134728383e-06, "loss": 0.9847, "step": 2970 }, { "epoch": 6.0, "grad_norm": 26.688039779663086, "learning_rate": 9.755282581475769e-06, "loss": 0.8882, "step": 3000 }, { "epoch": 6.0, "eval_loss": 1.020527720451355, "eval_map": 0.3611, "eval_map_50": 0.5069, "eval_map_75": 0.4262, "eval_map_chicken": 0.458, "eval_map_duck": 0.0, "eval_map_large": 0.6791, "eval_map_medium": 0.3426, "eval_map_plant": 0.6254, "eval_map_small": 0.029, "eval_mar_1": 0.137, "eval_mar_10": 0.4612, "eval_mar_100": 0.4689, "eval_mar_100_chicken": 0.6867, "eval_mar_100_duck": 0.0, "eval_mar_100_plant": 0.7199, "eval_mar_large": 0.7623, "eval_mar_medium": 0.4655, "eval_mar_small": 0.0792, "eval_runtime": 10.537, "eval_samples_per_second": 9.49, "eval_steps_per_second": 1.234, "step": 3000 }, { "epoch": 6.06, "grad_norm": 27.07685089111328, "learning_rate": 9.750405095503859e-06, "loss": 0.9775, "step": 3030 }, { "epoch": 6.12, "grad_norm": 40.5618896484375, "learning_rate": 9.745480724951473e-06, "loss": 0.8401, "step": 3060 }, { "epoch": 6.18, "grad_norm": 126.05823516845703, "learning_rate": 9.74050951842016e-06, "loss": 1.0211, "step": 3090 }, { "epoch": 6.24, "grad_norm": 37.94544219970703, "learning_rate": 9.735491524973723e-06, "loss": 0.8146, "step": 3120 }, { "epoch": 6.3, "grad_norm": 71.70733642578125, "learning_rate": 9.730426794137727e-06, "loss": 0.8952, "step": 3150 }, { "epoch": 6.36, "grad_norm": 57.69230651855469, "learning_rate": 9.725315375899025e-06, "loss": 0.8321, "step": 3180 }, { "epoch": 6.42, "grad_norm": 45.260135650634766, "learning_rate": 9.72015732070525e-06, "loss": 0.9529, "step": 3210 }, { "epoch": 6.48, "grad_norm": 59.23701477050781, "learning_rate": 9.714952679464324e-06, "loss": 0.9301, "step": 3240 }, { "epoch": 6.54, "grad_norm": 260.9499816894531, "learning_rate": 9.709701503543954e-06, "loss": 0.9797, "step": 3270 }, { "epoch": 6.6, "grad_norm": 73.61421203613281, "learning_rate": 9.704403844771128e-06, "loss": 0.863, "step": 3300 }, { "epoch": 6.66, "grad_norm": 31.928266525268555, "learning_rate": 9.699059755431599e-06, "loss": 0.9686, "step": 3330 }, { "epoch": 6.72, "grad_norm": 55.46344757080078, "learning_rate": 9.693669288269371e-06, "loss": 0.9671, "step": 3360 }, { "epoch": 6.78, "grad_norm": 82.36385345458984, "learning_rate": 9.688232496486179e-06, "loss": 0.9746, "step": 3390 }, { "epoch": 6.84, "grad_norm": 138.47650146484375, "learning_rate": 9.682749433740963e-06, "loss": 0.9927, "step": 3420 }, { "epoch": 6.9, "grad_norm": 0.21638916432857513, "learning_rate": 9.677220154149338e-06, "loss": 0.9382, "step": 3450 }, { "epoch": 6.96, "grad_norm": 52.91426467895508, "learning_rate": 9.671644712283061e-06, "loss": 1.0909, "step": 3480 }, { "epoch": 7.0, "eval_loss": 0.9382544159889221, "eval_map": 0.3877, "eval_map_50": 0.5328, "eval_map_75": 0.4615, "eval_map_chicken": 0.5073, "eval_map_duck": 0.0, "eval_map_large": 0.6989, "eval_map_medium": 0.3774, "eval_map_plant": 0.6558, "eval_map_small": 0.0414, "eval_mar_1": 0.1386, "eval_mar_10": 0.4773, "eval_mar_100": 0.4841, "eval_mar_100_chicken": 0.7089, "eval_mar_100_duck": 0.0, "eval_mar_100_plant": 0.7435, "eval_mar_large": 0.7766, "eval_mar_medium": 0.4827, "eval_mar_small": 0.1646, "eval_runtime": 10.7591, "eval_samples_per_second": 9.294, "eval_steps_per_second": 1.208, "step": 3500 }, { "epoch": 7.02, "grad_norm": 48.7019157409668, "learning_rate": 9.666023163169493e-06, "loss": 0.9351, "step": 3510 }, { "epoch": 7.08, "grad_norm": 34.22760772705078, "learning_rate": 9.660355562291055e-06, "loss": 0.8966, "step": 3540 }, { "epoch": 7.14, "grad_norm": 56.989654541015625, "learning_rate": 9.65464196558468e-06, "loss": 0.9803, "step": 3570 }, { "epoch": 7.2, "grad_norm": 28.444747924804688, "learning_rate": 9.648882429441258e-06, "loss": 0.8716, "step": 3600 }, { "epoch": 7.26, "grad_norm": 119.20661163330078, "learning_rate": 9.643077010705088e-06, "loss": 0.9125, "step": 3630 }, { "epoch": 7.32, "grad_norm": 67.59832763671875, "learning_rate": 9.637225766673309e-06, "loss": 1.015, "step": 3660 }, { "epoch": 7.38, "grad_norm": 52.43797302246094, "learning_rate": 9.631328755095334e-06, "loss": 0.9605, "step": 3690 }, { "epoch": 7.44, "grad_norm": 124.1357650756836, "learning_rate": 9.62538603417229e-06, "loss": 0.825, "step": 3720 }, { "epoch": 7.5, "grad_norm": 85.6240005493164, "learning_rate": 9.619397662556434e-06, "loss": 0.9077, "step": 3750 }, { "epoch": 7.5600000000000005, "grad_norm": 61.48551940917969, "learning_rate": 9.613363699350575e-06, "loss": 0.9026, "step": 3780 }, { "epoch": 7.62, "grad_norm": 67.21656036376953, "learning_rate": 9.607284204107493e-06, "loss": 0.8346, "step": 3810 }, { "epoch": 7.68, "grad_norm": 57.1346549987793, "learning_rate": 9.601159236829353e-06, "loss": 0.8624, "step": 3840 }, { "epoch": 7.74, "grad_norm": 128.2530975341797, "learning_rate": 9.594988857967107e-06, "loss": 1.1281, "step": 3870 }, { "epoch": 7.8, "grad_norm": 53.668739318847656, "learning_rate": 9.588773128419907e-06, "loss": 0.7965, "step": 3900 }, { "epoch": 7.86, "grad_norm": 56.750022888183594, "learning_rate": 9.58251210953449e-06, "loss": 1.1319, "step": 3930 }, { "epoch": 7.92, "grad_norm": 54.51675033569336, "learning_rate": 9.576205863104588e-06, "loss": 0.8379, "step": 3960 }, { "epoch": 7.98, "grad_norm": 64.90748596191406, "learning_rate": 9.569854451370306e-06, "loss": 0.9217, "step": 3990 }, { "epoch": 8.0, "eval_loss": 1.2725119590759277, "eval_map": 0.3553, "eval_map_50": 0.5133, "eval_map_75": 0.4208, "eval_map_chicken": 0.4638, "eval_map_duck": 0.0, "eval_map_large": 0.6449, "eval_map_medium": 0.3386, "eval_map_plant": 0.602, "eval_map_small": 0.0198, "eval_mar_1": 0.134, "eval_mar_10": 0.435, "eval_mar_100": 0.4391, "eval_mar_100_chicken": 0.6356, "eval_mar_100_duck": 0.0, "eval_mar_100_plant": 0.6818, "eval_mar_large": 0.7167, "eval_mar_medium": 0.4363, "eval_mar_small": 0.1187, "eval_runtime": 7.4744, "eval_samples_per_second": 13.379, "eval_steps_per_second": 1.739, "step": 4000 }, { "epoch": 8.04, "grad_norm": 46.677276611328125, "learning_rate": 9.563457937017514e-06, "loss": 1.1388, "step": 4020 }, { "epoch": 8.1, "grad_norm": 536.695556640625, "learning_rate": 9.557016383177226e-06, "loss": 1.1223, "step": 4050 }, { "epoch": 8.16, "grad_norm": 53.94501495361328, "learning_rate": 9.550529853424979e-06, "loss": 1.0483, "step": 4080 }, { "epoch": 8.22, "grad_norm": 37.831886291503906, "learning_rate": 9.543998411780202e-06, "loss": 0.9117, "step": 4110 }, { "epoch": 8.28, "grad_norm": 48.93321990966797, "learning_rate": 9.537422122705585e-06, "loss": 1.3161, "step": 4140 }, { "epoch": 8.34, "grad_norm": 102.14704132080078, "learning_rate": 9.530801051106449e-06, "loss": 0.7597, "step": 4170 }, { "epoch": 8.4, "grad_norm": 118.69752502441406, "learning_rate": 9.524135262330098e-06, "loss": 0.8559, "step": 4200 }, { "epoch": 8.46, "grad_norm": 2379.429443359375, "learning_rate": 9.517424822165175e-06, "loss": 1.0425, "step": 4230 }, { "epoch": 8.52, "grad_norm": 92.42198181152344, "learning_rate": 9.510669796841014e-06, "loss": 1.0686, "step": 4260 }, { "epoch": 8.58, "grad_norm": 47.93260955810547, "learning_rate": 9.503870253026992e-06, "loss": 0.7484, "step": 4290 }, { "epoch": 8.64, "grad_norm": 37.14649200439453, "learning_rate": 9.497026257831856e-06, "loss": 0.914, "step": 4320 }, { "epoch": 8.7, "grad_norm": 74.06472778320312, "learning_rate": 9.490137878803078e-06, "loss": 0.7326, "step": 4350 }, { "epoch": 8.76, "grad_norm": 52.570701599121094, "learning_rate": 9.48320518392618e-06, "loss": 0.9135, "step": 4380 }, { "epoch": 8.82, "grad_norm": 45.209712982177734, "learning_rate": 9.476228241624059e-06, "loss": 0.9584, "step": 4410 }, { "epoch": 8.88, "grad_norm": 46.63048553466797, "learning_rate": 9.46920712075632e-06, "loss": 0.9712, "step": 4440 }, { "epoch": 8.94, "grad_norm": 98.65027618408203, "learning_rate": 9.46214189061859e-06, "loss": 0.9356, "step": 4470 }, { "epoch": 9.0, "grad_norm": 0.04008881002664566, "learning_rate": 9.45503262094184e-06, "loss": 0.7372, "step": 4500 }, { "epoch": 9.0, "eval_loss": 0.7977738976478577, "eval_map": 0.4145, "eval_map_50": 0.5596, "eval_map_75": 0.4966, "eval_map_chicken": 0.5863, "eval_map_duck": 0.0069, "eval_map_large": 0.703, "eval_map_medium": 0.3945, "eval_map_plant": 0.6503, "eval_map_small": 0.0419, "eval_mar_1": 0.1488, "eval_mar_10": 0.4947, "eval_mar_100": 0.503, "eval_mar_100_chicken": 0.744, "eval_mar_100_duck": 0.0072, "eval_mar_100_plant": 0.7576, "eval_mar_large": 0.8017, "eval_mar_medium": 0.4914, "eval_mar_small": 0.2021, "eval_runtime": 9.2092, "eval_samples_per_second": 10.859, "eval_steps_per_second": 1.412, "step": 4500 }, { "epoch": 9.06, "grad_norm": 26.329444885253906, "learning_rate": 9.447879381891691e-06, "loss": 0.8145, "step": 4530 }, { "epoch": 9.12, "grad_norm": 73.16384887695312, "learning_rate": 9.440682244067724e-06, "loss": 0.7804, "step": 4560 }, { "epoch": 9.18, "grad_norm": 44.98170852661133, "learning_rate": 9.433441278502784e-06, "loss": 0.8261, "step": 4590 }, { "epoch": 9.24, "grad_norm": 102.83500671386719, "learning_rate": 9.426156556662276e-06, "loss": 0.9027, "step": 4620 }, { "epoch": 9.3, "grad_norm": 50.08503341674805, "learning_rate": 9.418828150443469e-06, "loss": 0.7741, "step": 4650 }, { "epoch": 9.36, "grad_norm": 173.54742431640625, "learning_rate": 9.411456132174768e-06, "loss": 0.8707, "step": 4680 }, { "epoch": 9.42, "grad_norm": 49.48848342895508, "learning_rate": 9.404040574615018e-06, "loss": 0.7526, "step": 4710 }, { "epoch": 9.48, "grad_norm": 39.25202560424805, "learning_rate": 9.396581550952781e-06, "loss": 0.7831, "step": 4740 }, { "epoch": 9.54, "grad_norm": 33.43211364746094, "learning_rate": 9.38907913480561e-06, "loss": 0.8808, "step": 4770 }, { "epoch": 9.6, "grad_norm": 40.27306365966797, "learning_rate": 9.381533400219319e-06, "loss": 0.8697, "step": 4800 }, { "epoch": 9.66, "grad_norm": 63.09968566894531, "learning_rate": 9.373944421667264e-06, "loss": 0.9783, "step": 4830 }, { "epoch": 9.72, "grad_norm": 33.3255500793457, "learning_rate": 9.366312274049602e-06, "loss": 0.9345, "step": 4860 }, { "epoch": 9.78, "grad_norm": 59.06138229370117, "learning_rate": 9.358637032692546e-06, "loss": 0.9409, "step": 4890 }, { "epoch": 9.84, "grad_norm": 39.260963439941406, "learning_rate": 9.35091877334763e-06, "loss": 0.7982, "step": 4920 }, { "epoch": 9.9, "grad_norm": 39.29015350341797, "learning_rate": 9.343157572190957e-06, "loss": 1.1397, "step": 4950 }, { "epoch": 9.96, "grad_norm": 52.55385971069336, "learning_rate": 9.33535350582245e-06, "loss": 1.1671, "step": 4980 }, { "epoch": 10.0, "eval_loss": 0.9801771640777588, "eval_map": 0.4261, "eval_map_50": 0.5765, "eval_map_75": 0.5132, "eval_map_chicken": 0.6288, "eval_map_duck": 0.0, "eval_map_large": 0.7055, "eval_map_medium": 0.4006, "eval_map_plant": 0.6496, "eval_map_small": 0.0772, "eval_mar_1": 0.1485, "eval_mar_10": 0.4813, "eval_mar_100": 0.4845, "eval_mar_100_chicken": 0.7196, "eval_mar_100_duck": 0.0, "eval_mar_100_plant": 0.734, "eval_mar_large": 0.7841, "eval_mar_medium": 0.4705, "eval_mar_small": 0.1521, "eval_runtime": 7.8465, "eval_samples_per_second": 12.745, "eval_steps_per_second": 1.657, "step": 5000 }, { "epoch": 10.02, "grad_norm": 63.012428283691406, "learning_rate": 9.327506651265096e-06, "loss": 0.9646, "step": 5010 }, { "epoch": 10.08, "grad_norm": 57.780643463134766, "learning_rate": 9.319617085964177e-06, "loss": 0.8413, "step": 5040 }, { "epoch": 10.14, "grad_norm": 38.30506896972656, "learning_rate": 9.31168488778652e-06, "loss": 0.7816, "step": 5070 }, { "epoch": 10.2, "grad_norm": 0.043808210641145706, "learning_rate": 9.30371013501972e-06, "loss": 0.9118, "step": 5100 }, { "epoch": 10.26, "grad_norm": 207.05877685546875, "learning_rate": 9.295692906371362e-06, "loss": 0.7662, "step": 5130 }, { "epoch": 10.32, "grad_norm": 39.717525482177734, "learning_rate": 9.287633280968263e-06, "loss": 0.7906, "step": 5160 }, { "epoch": 10.38, "grad_norm": 68.7513198852539, "learning_rate": 9.279531338355666e-06, "loss": 0.7965, "step": 5190 }, { "epoch": 10.44, "grad_norm": 44.65480422973633, "learning_rate": 9.271387158496477e-06, "loss": 0.8062, "step": 5220 }, { "epoch": 10.5, "grad_norm": 84.26549530029297, "learning_rate": 9.263200821770462e-06, "loss": 0.8193, "step": 5250 }, { "epoch": 10.56, "grad_norm": 52.59858322143555, "learning_rate": 9.25497240897346e-06, "loss": 0.7211, "step": 5280 }, { "epoch": 10.62, "grad_norm": 0.014806251041591167, "learning_rate": 9.246702001316584e-06, "loss": 0.7297, "step": 5310 }, { "epoch": 10.68, "grad_norm": 33.72252655029297, "learning_rate": 9.238389680425417e-06, "loss": 0.7145, "step": 5340 }, { "epoch": 10.74, "grad_norm": 46.29780197143555, "learning_rate": 9.230035528339212e-06, "loss": 0.8143, "step": 5370 }, { "epoch": 10.8, "grad_norm": 69.251953125, "learning_rate": 9.221639627510076e-06, "loss": 0.684, "step": 5400 }, { "epoch": 10.86, "grad_norm": 43.245609283447266, "learning_rate": 9.213202060802162e-06, "loss": 0.8437, "step": 5430 }, { "epoch": 10.92, "grad_norm": 63.19047546386719, "learning_rate": 9.204722911490847e-06, "loss": 0.7854, "step": 5460 }, { "epoch": 10.98, "grad_norm": 41.58643341064453, "learning_rate": 9.196202263261908e-06, "loss": 0.6988, "step": 5490 }, { "epoch": 11.0, "eval_loss": 0.7793210744857788, "eval_map": 0.4433, "eval_map_50": 0.5984, "eval_map_75": 0.5283, "eval_map_chicken": 0.6294, "eval_map_duck": 0.0139, "eval_map_large": 0.7405, "eval_map_medium": 0.4215, "eval_map_plant": 0.6867, "eval_map_small": 0.0796, "eval_mar_1": 0.1503, "eval_mar_10": 0.4908, "eval_mar_100": 0.4952, "eval_mar_100_chicken": 0.7102, "eval_mar_100_duck": 0.0072, "eval_mar_100_plant": 0.768, "eval_mar_large": 0.8209, "eval_mar_medium": 0.4778, "eval_mar_small": 0.1792, "eval_runtime": 8.9188, "eval_samples_per_second": 11.212, "eval_steps_per_second": 1.458, "step": 5500 }, { "epoch": 11.04, "grad_norm": 73.96212005615234, "learning_rate": 9.18764020021071e-06, "loss": 0.7727, "step": 5520 }, { "epoch": 11.1, "grad_norm": 63.49956130981445, "learning_rate": 9.179036806841352e-06, "loss": 0.7803, "step": 5550 }, { "epoch": 11.16, "grad_norm": 29.760915756225586, "learning_rate": 9.170392168065858e-06, "loss": 0.8156, "step": 5580 }, { "epoch": 11.22, "grad_norm": 93.62776947021484, "learning_rate": 9.161706369203319e-06, "loss": 0.8079, "step": 5610 }, { "epoch": 11.28, "grad_norm": 34.16836929321289, "learning_rate": 9.152979495979064e-06, "loss": 0.6219, "step": 5640 }, { "epoch": 11.34, "grad_norm": 77.70708465576172, "learning_rate": 9.14421163452381e-06, "loss": 0.6116, "step": 5670 }, { "epoch": 11.4, "grad_norm": 160.07376098632812, "learning_rate": 9.13540287137281e-06, "loss": 0.7408, "step": 5700 }, { "epoch": 11.46, "grad_norm": 52.56629943847656, "learning_rate": 9.126553293465e-06, "loss": 0.8868, "step": 5730 }, { "epoch": 11.52, "grad_norm": 58.52446365356445, "learning_rate": 9.117662988142138e-06, "loss": 0.963, "step": 5760 }, { "epoch": 11.58, "grad_norm": 135.65496826171875, "learning_rate": 9.108732043147952e-06, "loss": 0.6891, "step": 5790 }, { "epoch": 11.64, "grad_norm": 49.51852798461914, "learning_rate": 9.099760546627262e-06, "loss": 0.6969, "step": 5820 }, { "epoch": 11.7, "grad_norm": 104.19417572021484, "learning_rate": 9.090748587125118e-06, "loss": 0.7412, "step": 5850 }, { "epoch": 11.76, "grad_norm": 36.71453094482422, "learning_rate": 9.08169625358592e-06, "loss": 0.6529, "step": 5880 }, { "epoch": 11.82, "grad_norm": 52.78268814086914, "learning_rate": 9.072603635352548e-06, "loss": 0.8705, "step": 5910 }, { "epoch": 11.88, "grad_norm": 64.708740234375, "learning_rate": 9.06347082216547e-06, "loss": 0.8357, "step": 5940 }, { "epoch": 11.94, "grad_norm": 67.75703430175781, "learning_rate": 9.054297904161868e-06, "loss": 0.8196, "step": 5970 }, { "epoch": 12.0, "grad_norm": 28.469804763793945, "learning_rate": 9.045084971874738e-06, "loss": 0.9937, "step": 6000 }, { "epoch": 12.0, "eval_loss": 0.8788682818412781, "eval_map": 0.439, "eval_map_50": 0.6081, "eval_map_75": 0.5211, "eval_map_chicken": 0.5783, "eval_map_duck": 0.0354, "eval_map_large": 0.7549, "eval_map_medium": 0.4156, "eval_map_plant": 0.7033, "eval_map_small": 0.0795, "eval_mar_1": 0.152, "eval_mar_10": 0.4852, "eval_mar_100": 0.4898, "eval_mar_100_chicken": 0.6707, "eval_mar_100_duck": 0.032, "eval_mar_100_plant": 0.7669, "eval_mar_large": 0.8159, "eval_mar_medium": 0.4747, "eval_mar_small": 0.1896, "eval_runtime": 9.1763, "eval_samples_per_second": 10.898, "eval_steps_per_second": 1.417, "step": 6000 }, { "epoch": 12.06, "grad_norm": 47.45894241333008, "learning_rate": 9.035832116232002e-06, "loss": 0.814, "step": 6030 }, { "epoch": 12.12, "grad_norm": 51.75700378417969, "learning_rate": 9.026539428555609e-06, "loss": 0.7199, "step": 6060 }, { "epoch": 12.18, "grad_norm": 62.82158279418945, "learning_rate": 9.017207000560639e-06, "loss": 0.8528, "step": 6090 }, { "epoch": 12.24, "grad_norm": 55.68227005004883, "learning_rate": 9.007834924354384e-06, "loss": 0.9093, "step": 6120 }, { "epoch": 12.3, "grad_norm": 34.593570709228516, "learning_rate": 8.998423292435455e-06, "loss": 0.7785, "step": 6150 }, { "epoch": 12.36, "grad_norm": 51.85447692871094, "learning_rate": 8.988972197692857e-06, "loss": 0.8328, "step": 6180 }, { "epoch": 12.42, "grad_norm": 42.27244186401367, "learning_rate": 8.97948173340508e-06, "loss": 0.8906, "step": 6210 }, { "epoch": 12.48, "grad_norm": 40.95602035522461, "learning_rate": 8.969951993239177e-06, "loss": 0.7081, "step": 6240 }, { "epoch": 12.54, "grad_norm": 58.91379165649414, "learning_rate": 8.960383071249837e-06, "loss": 0.7384, "step": 6270 }, { "epoch": 12.6, "grad_norm": 128.2095184326172, "learning_rate": 8.950775061878453e-06, "loss": 0.6746, "step": 6300 }, { "epoch": 12.66, "grad_norm": 96.57804870605469, "learning_rate": 8.9411280599522e-06, "loss": 0.7677, "step": 6330 }, { "epoch": 12.72, "grad_norm": 0.006949965842068195, "learning_rate": 8.931442160683094e-06, "loss": 0.9175, "step": 6360 }, { "epoch": 12.78, "grad_norm": 105.19688415527344, "learning_rate": 8.921717459667052e-06, "loss": 0.7588, "step": 6390 }, { "epoch": 12.84, "grad_norm": 48.05434036254883, "learning_rate": 8.911954052882941e-06, "loss": 0.8341, "step": 6420 }, { "epoch": 12.9, "grad_norm": 129.20071411132812, "learning_rate": 8.902152036691649e-06, "loss": 0.8222, "step": 6450 }, { "epoch": 12.96, "grad_norm": 59.23206329345703, "learning_rate": 8.892311507835118e-06, "loss": 0.6916, "step": 6480 }, { "epoch": 13.0, "eval_loss": 0.6078073382377625, "eval_map": 0.511, "eval_map_50": 0.6769, "eval_map_75": 0.609, "eval_map_chicken": 0.6912, "eval_map_duck": 0.1144, "eval_map_large": 0.7782, "eval_map_medium": 0.4957, "eval_map_plant": 0.7274, "eval_map_small": 0.1215, "eval_mar_1": 0.1796, "eval_mar_10": 0.5567, "eval_mar_100": 0.5609, "eval_mar_100_chicken": 0.7631, "eval_mar_100_duck": 0.1227, "eval_mar_100_plant": 0.7968, "eval_mar_large": 0.8452, "eval_mar_medium": 0.5504, "eval_mar_small": 0.2271, "eval_runtime": 10.0567, "eval_samples_per_second": 9.944, "eval_steps_per_second": 1.293, "step": 6500 }, { "epoch": 13.02, "grad_norm": 40.281742095947266, "learning_rate": 8.882432563435394e-06, "loss": 0.7778, "step": 6510 }, { "epoch": 13.08, "grad_norm": 0.005489167757332325, "learning_rate": 8.872515300993669e-06, "loss": 0.6796, "step": 6540 }, { "epoch": 13.14, "grad_norm": 80.79935455322266, "learning_rate": 8.862559818389322e-06, "loss": 0.6869, "step": 6570 }, { "epoch": 13.2, "grad_norm": 70.19380950927734, "learning_rate": 8.852566213878947e-06, "loss": 0.755, "step": 6600 }, { "epoch": 13.26, "grad_norm": 50.8900260925293, "learning_rate": 8.842534586095383e-06, "loss": 0.6986, "step": 6630 }, { "epoch": 13.32, "grad_norm": 38.117332458496094, "learning_rate": 8.83246503404675e-06, "loss": 0.5907, "step": 6660 }, { "epoch": 13.38, "grad_norm": 78.95010375976562, "learning_rate": 8.82235765711546e-06, "loss": 0.6791, "step": 6690 }, { "epoch": 13.44, "grad_norm": 56.1155891418457, "learning_rate": 8.81221255505724e-06, "loss": 0.7286, "step": 6720 }, { "epoch": 13.5, "grad_norm": 41.99263381958008, "learning_rate": 8.802029828000157e-06, "loss": 0.8263, "step": 6750 }, { "epoch": 13.56, "grad_norm": 90.74502563476562, "learning_rate": 8.791809576443611e-06, "loss": 0.7202, "step": 6780 }, { "epoch": 13.62, "grad_norm": 65.30818939208984, "learning_rate": 8.78155190125736e-06, "loss": 0.7282, "step": 6810 }, { "epoch": 13.68, "grad_norm": 100.8336410522461, "learning_rate": 8.77125690368052e-06, "loss": 0.7845, "step": 6840 }, { "epoch": 13.74, "grad_norm": 67.6075210571289, "learning_rate": 8.760924685320558e-06, "loss": 0.739, "step": 6870 }, { "epoch": 13.8, "grad_norm": 0.004169583320617676, "learning_rate": 8.750555348152299e-06, "loss": 0.6828, "step": 6900 }, { "epoch": 13.86, "grad_norm": 63.3882942199707, "learning_rate": 8.740148994516912e-06, "loss": 0.7458, "step": 6930 }, { "epoch": 13.92, "grad_norm": 51.33131790161133, "learning_rate": 8.729705727120911e-06, "loss": 0.8144, "step": 6960 }, { "epoch": 13.98, "grad_norm": 42.905845642089844, "learning_rate": 8.719225649035126e-06, "loss": 0.7906, "step": 6990 }, { "epoch": 14.0, "eval_loss": 0.6883366107940674, "eval_map": 0.5666, "eval_map_50": 0.7666, "eval_map_75": 0.6899, "eval_map_chicken": 0.6607, "eval_map_duck": 0.3346, "eval_map_large": 0.768, "eval_map_medium": 0.5485, "eval_map_plant": 0.7046, "eval_map_small": 0.0765, "eval_mar_1": 0.2133, "eval_mar_10": 0.6132, "eval_mar_100": 0.6166, "eval_mar_100_chicken": 0.7231, "eval_mar_100_duck": 0.3588, "eval_mar_100_plant": 0.768, "eval_mar_large": 0.8272, "eval_mar_medium": 0.6111, "eval_mar_small": 0.1604, "eval_runtime": 7.5788, "eval_samples_per_second": 13.195, "eval_steps_per_second": 1.715, "step": 7000 }, { "epoch": 14.04, "grad_norm": 55.345890045166016, "learning_rate": 8.708708863693696e-06, "loss": 0.7018, "step": 7020 }, { "epoch": 14.1, "grad_norm": 45.42887878417969, "learning_rate": 8.69815547489305e-06, "loss": 0.6683, "step": 7050 }, { "epoch": 14.16, "grad_norm": 0.003694022772833705, "learning_rate": 8.68756558679087e-06, "loss": 0.7148, "step": 7080 }, { "epoch": 14.22, "grad_norm": 61.962703704833984, "learning_rate": 8.67693930390508e-06, "loss": 0.9702, "step": 7110 }, { "epoch": 14.28, "grad_norm": 552.206298828125, "learning_rate": 8.666276731112802e-06, "loss": 0.8327, "step": 7140 }, { "epoch": 14.34, "grad_norm": 51.08211898803711, "learning_rate": 8.655577973649322e-06, "loss": 0.7003, "step": 7170 }, { "epoch": 14.4, "grad_norm": 63.054744720458984, "learning_rate": 8.644843137107058e-06, "loss": 0.7593, "step": 7200 }, { "epoch": 14.46, "grad_norm": 54.83010482788086, "learning_rate": 8.634072327434515e-06, "loss": 0.6405, "step": 7230 }, { "epoch": 14.52, "grad_norm": 49.37504577636719, "learning_rate": 8.623265650935233e-06, "loss": 0.7956, "step": 7260 }, { "epoch": 14.58, "grad_norm": 41.56403732299805, "learning_rate": 8.612423214266749e-06, "loss": 0.7295, "step": 7290 }, { "epoch": 14.64, "grad_norm": 40.623104095458984, "learning_rate": 8.601545124439535e-06, "loss": 0.7589, "step": 7320 }, { "epoch": 14.7, "grad_norm": 55.46794891357422, "learning_rate": 8.590631488815945e-06, "loss": 0.7067, "step": 7350 }, { "epoch": 14.76, "grad_norm": 43.29323959350586, "learning_rate": 8.579682415109156e-06, "loss": 0.6794, "step": 7380 }, { "epoch": 14.82, "grad_norm": 61.12752914428711, "learning_rate": 8.568698011382108e-06, "loss": 0.8408, "step": 7410 }, { "epoch": 14.88, "grad_norm": 38.454471588134766, "learning_rate": 8.557678386046429e-06, "loss": 0.6336, "step": 7440 }, { "epoch": 14.94, "grad_norm": 44.199119567871094, "learning_rate": 8.54662364786137e-06, "loss": 0.6548, "step": 7470 }, { "epoch": 15.0, "grad_norm": 0.0035488998983055353, "learning_rate": 8.535533905932739e-06, "loss": 0.7007, "step": 7500 }, { "epoch": 15.0, "eval_loss": 0.5813708901405334, "eval_map": 0.6286, "eval_map_50": 0.8466, "eval_map_75": 0.753, "eval_map_chicken": 0.6881, "eval_map_duck": 0.4563, "eval_map_large": 0.7968, "eval_map_medium": 0.6259, "eval_map_plant": 0.7415, "eval_map_small": 0.0779, "eval_mar_1": 0.2349, "eval_mar_10": 0.6753, "eval_mar_100": 0.6844, "eval_mar_100_chicken": 0.7502, "eval_mar_100_duck": 0.499, "eval_mar_100_plant": 0.804, "eval_mar_large": 0.8565, "eval_mar_medium": 0.6892, "eval_mar_small": 0.1937, "eval_runtime": 9.9985, "eval_samples_per_second": 10.002, "eval_steps_per_second": 1.3, "step": 7500 }, { "epoch": 15.06, "grad_norm": 37.143043518066406, "learning_rate": 8.524409269711808e-06, "loss": 0.6501, "step": 7530 }, { "epoch": 15.12, "grad_norm": 52.671146392822266, "learning_rate": 8.513249848994248e-06, "loss": 0.7331, "step": 7560 }, { "epoch": 15.18, "grad_norm": 60.95817947387695, "learning_rate": 8.502055753919033e-06, "loss": 0.6872, "step": 7590 }, { "epoch": 15.24, "grad_norm": 186.4025115966797, "learning_rate": 8.490827094967364e-06, "loss": 0.6298, "step": 7620 }, { "epoch": 15.3, "grad_norm": 47.192718505859375, "learning_rate": 8.479563982961572e-06, "loss": 0.6025, "step": 7650 }, { "epoch": 15.36, "grad_norm": 35.82501983642578, "learning_rate": 8.468266529064025e-06, "loss": 0.626, "step": 7680 }, { "epoch": 15.42, "grad_norm": 93.18826293945312, "learning_rate": 8.456934844776033e-06, "loss": 0.7649, "step": 7710 }, { "epoch": 15.48, "grad_norm": 58.04679489135742, "learning_rate": 8.445569041936743e-06, "loss": 0.8021, "step": 7740 }, { "epoch": 15.54, "grad_norm": 104.08062744140625, "learning_rate": 8.434169232722043e-06, "loss": 0.5797, "step": 7770 }, { "epoch": 15.6, "grad_norm": 65.38141632080078, "learning_rate": 8.422735529643445e-06, "loss": 0.651, "step": 7800 }, { "epoch": 15.66, "grad_norm": 58.25004959106445, "learning_rate": 8.411268045546984e-06, "loss": 0.69, "step": 7830 }, { "epoch": 15.72, "grad_norm": 42.20881652832031, "learning_rate": 8.399766893612096e-06, "loss": 0.624, "step": 7860 }, { "epoch": 15.78, "grad_norm": 43.94855880737305, "learning_rate": 8.388232187350513e-06, "loss": 0.6736, "step": 7890 }, { "epoch": 15.84, "grad_norm": 32.47200393676758, "learning_rate": 8.376664040605122e-06, "loss": 0.7366, "step": 7920 }, { "epoch": 15.9, "grad_norm": 349.0764465332031, "learning_rate": 8.365062567548868e-06, "loss": 0.7098, "step": 7950 }, { "epoch": 15.96, "grad_norm": 49.24812698364258, "learning_rate": 8.353427882683601e-06, "loss": 0.5713, "step": 7980 }, { "epoch": 16.0, "eval_loss": 0.6028811931610107, "eval_map": 0.6403, "eval_map_50": 0.8674, "eval_map_75": 0.7877, "eval_map_chicken": 0.6821, "eval_map_duck": 0.5247, "eval_map_large": 0.7694, "eval_map_medium": 0.6414, "eval_map_plant": 0.7141, "eval_map_small": 0.0716, "eval_mar_1": 0.246, "eval_mar_10": 0.6949, "eval_mar_100": 0.6978, "eval_mar_100_chicken": 0.7427, "eval_mar_100_duck": 0.566, "eval_mar_100_plant": 0.7847, "eval_mar_large": 0.841, "eval_mar_medium": 0.7035, "eval_mar_small": 0.1729, "eval_runtime": 7.8252, "eval_samples_per_second": 12.779, "eval_steps_per_second": 1.661, "step": 8000 }, { "epoch": 16.02, "grad_norm": 86.52704620361328, "learning_rate": 8.341760100838967e-06, "loss": 0.7278, "step": 8010 }, { "epoch": 16.08, "grad_norm": 270.6351318359375, "learning_rate": 8.33005933717126e-06, "loss": 0.646, "step": 8040 }, { "epoch": 16.14, "grad_norm": 38.12831115722656, "learning_rate": 8.318325707162293e-06, "loss": 0.6615, "step": 8070 }, { "epoch": 16.2, "grad_norm": 49.97762680053711, "learning_rate": 8.30655932661826e-06, "loss": 0.5992, "step": 8100 }, { "epoch": 16.26, "grad_norm": 39.02716827392578, "learning_rate": 8.294760311668586e-06, "loss": 0.6861, "step": 8130 }, { "epoch": 16.32, "grad_norm": 35.441375732421875, "learning_rate": 8.282928778764783e-06, "loss": 0.6183, "step": 8160 }, { "epoch": 16.38, "grad_norm": 42.56651306152344, "learning_rate": 8.271064844679306e-06, "loss": 0.659, "step": 8190 }, { "epoch": 16.44, "grad_norm": 48.209434509277344, "learning_rate": 8.259168626504395e-06, "loss": 0.6268, "step": 8220 }, { "epoch": 16.5, "grad_norm": 28.18348503112793, "learning_rate": 8.247240241650918e-06, "loss": 0.5903, "step": 8250 }, { "epoch": 16.56, "grad_norm": 27.742042541503906, "learning_rate": 8.235279807847223e-06, "loss": 0.6713, "step": 8280 }, { "epoch": 16.62, "grad_norm": 86.91703033447266, "learning_rate": 8.223287443137957e-06, "loss": 0.5714, "step": 8310 }, { "epoch": 16.68, "grad_norm": 60.55374526977539, "learning_rate": 8.211263265882923e-06, "loss": 0.7029, "step": 8340 }, { "epoch": 16.74, "grad_norm": 144.75303649902344, "learning_rate": 8.199207394755892e-06, "loss": 0.713, "step": 8370 }, { "epoch": 16.8, "grad_norm": 38.36867141723633, "learning_rate": 8.18711994874345e-06, "loss": 0.549, "step": 8400 }, { "epoch": 16.86, "grad_norm": 68.6002426147461, "learning_rate": 8.175001047143804e-06, "loss": 0.5892, "step": 8430 }, { "epoch": 16.92, "grad_norm": 743.8650512695312, "learning_rate": 8.162850809565623e-06, "loss": 0.6602, "step": 8460 }, { "epoch": 16.98, "grad_norm": 83.9339370727539, "learning_rate": 8.150669355926848e-06, "loss": 0.6137, "step": 8490 }, { "epoch": 17.0, "eval_loss": 0.5384795069694519, "eval_map": 0.6769, "eval_map_50": 0.8987, "eval_map_75": 0.8132, "eval_map_chicken": 0.713, "eval_map_duck": 0.5672, "eval_map_large": 0.8139, "eval_map_medium": 0.671, "eval_map_plant": 0.7505, "eval_map_small": 0.1, "eval_mar_1": 0.2564, "eval_mar_10": 0.731, "eval_mar_100": 0.7366, "eval_mar_100_chicken": 0.7724, "eval_mar_100_duck": 0.6196, "eval_mar_100_plant": 0.8179, "eval_mar_large": 0.8745, "eval_mar_medium": 0.7452, "eval_mar_small": 0.1813, "eval_runtime": 7.9146, "eval_samples_per_second": 12.635, "eval_steps_per_second": 1.643, "step": 8500 }, { "epoch": 17.04, "grad_norm": 82.20858764648438, "learning_rate": 8.138456806453503e-06, "loss": 0.7131, "step": 8520 }, { "epoch": 17.1, "grad_norm": 46.967525482177734, "learning_rate": 8.126213281678527e-06, "loss": 0.7106, "step": 8550 }, { "epoch": 17.16, "grad_norm": 28.184831619262695, "learning_rate": 8.113938902440563e-06, "loss": 0.5795, "step": 8580 }, { "epoch": 17.22, "grad_norm": 76.19196319580078, "learning_rate": 8.101633789882781e-06, "loss": 0.6493, "step": 8610 }, { "epoch": 17.28, "grad_norm": 36.051170349121094, "learning_rate": 8.089298065451673e-06, "loss": 0.5708, "step": 8640 }, { "epoch": 17.34, "grad_norm": 50.474159240722656, "learning_rate": 8.076931850895858e-06, "loss": 0.6355, "step": 8670 }, { "epoch": 17.4, "grad_norm": 56.24653625488281, "learning_rate": 8.064535268264883e-06, "loss": 0.5782, "step": 8700 }, { "epoch": 17.46, "grad_norm": 57.91929626464844, "learning_rate": 8.052108439908014e-06, "loss": 0.5268, "step": 8730 }, { "epoch": 17.52, "grad_norm": 40.70713806152344, "learning_rate": 8.039651488473028e-06, "loss": 0.6, "step": 8760 }, { "epoch": 17.58, "grad_norm": 38.620418548583984, "learning_rate": 8.027164536905008e-06, "loss": 0.5838, "step": 8790 }, { "epoch": 17.64, "grad_norm": 181.1210174560547, "learning_rate": 8.014647708445124e-06, "loss": 0.5884, "step": 8820 }, { "epoch": 17.7, "grad_norm": 30.397825241088867, "learning_rate": 8.002101126629422e-06, "loss": 0.625, "step": 8850 }, { "epoch": 17.76, "grad_norm": 99.52507781982422, "learning_rate": 7.989524915287595e-06, "loss": 0.6698, "step": 8880 }, { "epoch": 17.82, "grad_norm": 44.390037536621094, "learning_rate": 7.976919198541775e-06, "loss": 0.6964, "step": 8910 }, { "epoch": 17.88, "grad_norm": 64.90728759765625, "learning_rate": 7.964284100805297e-06, "loss": 0.6814, "step": 8940 }, { "epoch": 17.94, "grad_norm": 93.40696716308594, "learning_rate": 7.951619746781474e-06, "loss": 0.6357, "step": 8970 }, { "epoch": 18.0, "grad_norm": 51.5998649597168, "learning_rate": 7.938926261462366e-06, "loss": 0.559, "step": 9000 }, { "epoch": 18.0, "eval_loss": 0.5382981896400452, "eval_map": 0.6866, "eval_map_50": 0.914, "eval_map_75": 0.8436, "eval_map_chicken": 0.7136, "eval_map_duck": 0.6116, "eval_map_large": 0.7901, "eval_map_medium": 0.6916, "eval_map_plant": 0.7346, "eval_map_small": 0.1121, "eval_mar_1": 0.2711, "eval_mar_10": 0.7398, "eval_mar_100": 0.7452, "eval_mar_100_chicken": 0.7684, "eval_mar_100_duck": 0.6649, "eval_mar_100_plant": 0.8023, "eval_mar_large": 0.8544, "eval_mar_medium": 0.7582, "eval_mar_small": 0.1884, "eval_runtime": 8.7943, "eval_samples_per_second": 11.371, "eval_steps_per_second": 1.478, "step": 9000 }, { "epoch": 18.06, "grad_norm": 89.23826599121094, "learning_rate": 7.926203770127552e-06, "loss": 0.5812, "step": 9030 }, { "epoch": 18.12, "grad_norm": 62.54021453857422, "learning_rate": 7.913452398342882e-06, "loss": 0.5263, "step": 9060 }, { "epoch": 18.18, "grad_norm": 74.2127914428711, "learning_rate": 7.900672271959247e-06, "loss": 0.6514, "step": 9090 }, { "epoch": 18.24, "grad_norm": 50.786563873291016, "learning_rate": 7.887863517111337e-06, "loss": 0.6181, "step": 9120 }, { "epoch": 18.3, "grad_norm": 95.28047943115234, "learning_rate": 7.875026260216395e-06, "loss": 0.5887, "step": 9150 }, { "epoch": 18.36, "grad_norm": 37.34574890136719, "learning_rate": 7.862160627972956e-06, "loss": 0.5742, "step": 9180 }, { "epoch": 18.42, "grad_norm": 34.176395416259766, "learning_rate": 7.849266747359619e-06, "loss": 0.5667, "step": 9210 }, { "epoch": 18.48, "grad_norm": 36.09999084472656, "learning_rate": 7.836344745633785e-06, "loss": 0.694, "step": 9240 }, { "epoch": 18.54, "grad_norm": 71.43876647949219, "learning_rate": 7.823394750330386e-06, "loss": 0.6384, "step": 9270 }, { "epoch": 18.6, "grad_norm": 117.6993408203125, "learning_rate": 7.810416889260653e-06, "loss": 0.6995, "step": 9300 }, { "epoch": 18.66, "grad_norm": 94.19873046875, "learning_rate": 7.797411290510836e-06, "loss": 0.5205, "step": 9330 }, { "epoch": 18.72, "grad_norm": 43.26815414428711, "learning_rate": 7.78437808244094e-06, "loss": 0.6112, "step": 9360 }, { "epoch": 18.78, "grad_norm": 83.01200866699219, "learning_rate": 7.771317393683471e-06, "loss": 0.5738, "step": 9390 }, { "epoch": 18.84, "grad_norm": 1182.55224609375, "learning_rate": 7.758229353142153e-06, "loss": 0.6946, "step": 9420 }, { "epoch": 18.9, "grad_norm": 48.34932327270508, "learning_rate": 7.74511408999066e-06, "loss": 0.6468, "step": 9450 }, { "epoch": 18.96, "grad_norm": 45.35761260986328, "learning_rate": 7.731971733671347e-06, "loss": 0.641, "step": 9480 }, { "epoch": 19.0, "eval_loss": 0.4983445703983307, "eval_map": 0.6907, "eval_map_50": 0.9135, "eval_map_75": 0.8424, "eval_map_chicken": 0.7098, "eval_map_duck": 0.6051, "eval_map_large": 0.8098, "eval_map_medium": 0.6957, "eval_map_plant": 0.7572, "eval_map_small": 0.1156, "eval_mar_1": 0.2715, "eval_mar_10": 0.7438, "eval_mar_100": 0.7489, "eval_mar_100_chicken": 0.7596, "eval_mar_100_duck": 0.6629, "eval_mar_100_plant": 0.8242, "eval_mar_large": 0.8711, "eval_mar_medium": 0.7662, "eval_mar_small": 0.1873, "eval_runtime": 9.8638, "eval_samples_per_second": 10.138, "eval_steps_per_second": 1.318, "step": 9500 }, { "epoch": 19.02, "grad_norm": 58.828731536865234, "learning_rate": 7.718802413893963e-06, "loss": 0.7014, "step": 9510 }, { "epoch": 19.08, "grad_norm": 79.43228912353516, "learning_rate": 7.70560626063438e-06, "loss": 0.6017, "step": 9540 }, { "epoch": 19.14, "grad_norm": 38.46056365966797, "learning_rate": 7.692383404133302e-06, "loss": 0.5233, "step": 9570 }, { "epoch": 19.2, "grad_norm": 70.59977722167969, "learning_rate": 7.679133974894984e-06, "loss": 0.5404, "step": 9600 }, { "epoch": 19.26, "grad_norm": 28.751224517822266, "learning_rate": 7.665858103685944e-06, "loss": 0.4931, "step": 9630 }, { "epoch": 19.32, "grad_norm": 54.740413665771484, "learning_rate": 7.652555921533671e-06, "loss": 0.5749, "step": 9660 }, { "epoch": 19.38, "grad_norm": 79.96854400634766, "learning_rate": 7.639227559725333e-06, "loss": 0.605, "step": 9690 }, { "epoch": 19.44, "grad_norm": 62.17195129394531, "learning_rate": 7.6258731498064796e-06, "loss": 0.552, "step": 9720 }, { "epoch": 19.5, "grad_norm": 67.75467681884766, "learning_rate": 7.612492823579744e-06, "loss": 0.6355, "step": 9750 }, { "epoch": 19.56, "grad_norm": 45.99704360961914, "learning_rate": 7.5990867131035474e-06, "loss": 0.5904, "step": 9780 }, { "epoch": 19.62, "grad_norm": 176.30734252929688, "learning_rate": 7.585654950690786e-06, "loss": 0.549, "step": 9810 }, { "epoch": 19.68, "grad_norm": 0.003810643684118986, "learning_rate": 7.572197668907533e-06, "loss": 0.6346, "step": 9840 }, { "epoch": 19.74, "grad_norm": 66.40618133544922, "learning_rate": 7.5587150005717256e-06, "loss": 0.6487, "step": 9870 }, { "epoch": 19.8, "grad_norm": 56.52738952636719, "learning_rate": 7.545207078751858e-06, "loss": 0.6414, "step": 9900 }, { "epoch": 19.86, "grad_norm": 48.40855407714844, "learning_rate": 7.531674036765662e-06, "loss": 0.7046, "step": 9930 }, { "epoch": 19.92, "grad_norm": 55.56481170654297, "learning_rate": 7.518116008178805e-06, "loss": 0.6373, "step": 9960 }, { "epoch": 19.98, "grad_norm": 40.09135437011719, "learning_rate": 7.5045331268035505e-06, "loss": 0.5699, "step": 9990 }, { "epoch": 20.0, "eval_loss": 0.4843730926513672, "eval_map": 0.6974, "eval_map_50": 0.924, "eval_map_75": 0.8442, "eval_map_chicken": 0.6942, "eval_map_duck": 0.6484, "eval_map_large": 0.8021, "eval_map_medium": 0.6997, "eval_map_plant": 0.7495, "eval_map_small": 0.142, "eval_mar_1": 0.2726, "eval_mar_10": 0.7545, "eval_mar_100": 0.7588, "eval_mar_100_chicken": 0.7564, "eval_mar_100_duck": 0.7031, "eval_mar_100_plant": 0.817, "eval_mar_large": 0.8669, "eval_mar_medium": 0.769, "eval_mar_small": 0.2633, "eval_runtime": 7.7088, "eval_samples_per_second": 12.972, "eval_steps_per_second": 1.686, "step": 10000 }, { "epoch": 20.04, "grad_norm": 95.95970153808594, "learning_rate": 7.490925526697455e-06, "loss": 0.5694, "step": 10020 }, { "epoch": 20.1, "grad_norm": 36.738502502441406, "learning_rate": 7.477293342162038e-06, "loss": 0.5535, "step": 10050 }, { "epoch": 20.16, "grad_norm": 70.17654418945312, "learning_rate": 7.463636707741458e-06, "loss": 0.5381, "step": 10080 }, { "epoch": 20.22, "grad_norm": 72.73855590820312, "learning_rate": 7.449955758221184e-06, "loss": 0.5667, "step": 10110 }, { "epoch": 20.28, "grad_norm": 54.85578918457031, "learning_rate": 7.436250628626662e-06, "loss": 0.6484, "step": 10140 }, { "epoch": 20.34, "grad_norm": 79.41162872314453, "learning_rate": 7.42252145422199e-06, "loss": 0.6307, "step": 10170 }, { "epoch": 20.4, "grad_norm": 43.94011688232422, "learning_rate": 7.408768370508577e-06, "loss": 0.5985, "step": 10200 }, { "epoch": 20.46, "grad_norm": 57.49893569946289, "learning_rate": 7.394991513223806e-06, "loss": 0.6053, "step": 10230 }, { "epoch": 20.52, "grad_norm": 33.66080093383789, "learning_rate": 7.381191018339697e-06, "loss": 0.5909, "step": 10260 }, { "epoch": 20.58, "grad_norm": 38.89453125, "learning_rate": 7.3673670220615615e-06, "loss": 0.5945, "step": 10290 }, { "epoch": 20.64, "grad_norm": 69.53521728515625, "learning_rate": 7.353519660826665e-06, "loss": 0.9446, "step": 10320 }, { "epoch": 20.7, "grad_norm": 24.856294631958008, "learning_rate": 7.3396490713028674e-06, "loss": 0.5844, "step": 10350 }, { "epoch": 20.76, "grad_norm": 69.91577911376953, "learning_rate": 7.325755390387293e-06, "loss": 0.6548, "step": 10380 }, { "epoch": 20.82, "grad_norm": 105.49382019042969, "learning_rate": 7.31183875520496e-06, "loss": 0.6431, "step": 10410 }, { "epoch": 20.88, "grad_norm": 68.51399230957031, "learning_rate": 7.297899303107441e-06, "loss": 0.598, "step": 10440 }, { "epoch": 20.94, "grad_norm": 145.6100311279297, "learning_rate": 7.283937171671498e-06, "loss": 0.5814, "step": 10470 }, { "epoch": 21.0, "grad_norm": 92.3284912109375, "learning_rate": 7.269952498697734e-06, "loss": 0.5071, "step": 10500 }, { "epoch": 21.0, "eval_loss": 0.507683515548706, "eval_map": 0.7026, "eval_map_50": 0.9156, "eval_map_75": 0.8335, "eval_map_chicken": 0.7015, "eval_map_duck": 0.6579, "eval_map_large": 0.8073, "eval_map_medium": 0.7006, "eval_map_plant": 0.7483, "eval_map_small": 0.0903, "eval_mar_1": 0.2838, "eval_mar_10": 0.7494, "eval_mar_100": 0.7543, "eval_mar_100_chicken": 0.7542, "eval_mar_100_duck": 0.6948, "eval_mar_100_plant": 0.8138, "eval_mar_large": 0.8703, "eval_mar_medium": 0.7632, "eval_mar_small": 0.207, "eval_runtime": 7.5801, "eval_samples_per_second": 13.192, "eval_steps_per_second": 1.715, "step": 10500 }, { "epoch": 21.06, "grad_norm": 119.36011505126953, "learning_rate": 7.2559454222092265e-06, "loss": 0.5478, "step": 10530 }, { "epoch": 21.12, "grad_norm": 31.13939666748047, "learning_rate": 7.241916080450163e-06, "loss": 0.5955, "step": 10560 }, { "epoch": 21.18, "grad_norm": 68.87150573730469, "learning_rate": 7.227864611884483e-06, "loss": 0.6466, "step": 10590 }, { "epoch": 21.24, "grad_norm": 127.9319076538086, "learning_rate": 7.21379115519451e-06, "loss": 0.644, "step": 10620 }, { "epoch": 21.3, "grad_norm": 43.20797348022461, "learning_rate": 7.199695849279576e-06, "loss": 0.479, "step": 10650 }, { "epoch": 21.36, "grad_norm": 34.34306716918945, "learning_rate": 7.185578833254665e-06, "loss": 0.5804, "step": 10680 }, { "epoch": 21.42, "grad_norm": 302.0656433105469, "learning_rate": 7.171440246449024e-06, "loss": 0.5511, "step": 10710 }, { "epoch": 21.48, "grad_norm": 53.048744201660156, "learning_rate": 7.157280228404796e-06, "loss": 0.631, "step": 10740 }, { "epoch": 21.54, "grad_norm": 97.1697769165039, "learning_rate": 7.143098918875643e-06, "loss": 0.6712, "step": 10770 }, { "epoch": 21.6, "grad_norm": 47.32440948486328, "learning_rate": 7.128896457825364e-06, "loss": 0.5922, "step": 10800 }, { "epoch": 21.66, "grad_norm": 48.359066009521484, "learning_rate": 7.114672985426516e-06, "loss": 0.5723, "step": 10830 }, { "epoch": 21.72, "grad_norm": 74.79902648925781, "learning_rate": 7.100428642059033e-06, "loss": 0.5939, "step": 10860 }, { "epoch": 21.78, "grad_norm": 35.08643341064453, "learning_rate": 7.086163568308828e-06, "loss": 0.5753, "step": 10890 }, { "epoch": 21.84, "grad_norm": 62.549537658691406, "learning_rate": 7.071877904966422e-06, "loss": 0.5806, "step": 10920 }, { "epoch": 21.9, "grad_norm": 112.69488525390625, "learning_rate": 7.057571793025545e-06, "loss": 0.564, "step": 10950 }, { "epoch": 21.96, "grad_norm": 76.42573547363281, "learning_rate": 7.043245373681746e-06, "loss": 0.6488, "step": 10980 }, { "epoch": 22.0, "eval_loss": 0.4770389795303345, "eval_map": 0.7044, "eval_map_50": 0.9242, "eval_map_75": 0.8467, "eval_map_chicken": 0.7085, "eval_map_duck": 0.6474, "eval_map_large": 0.8086, "eval_map_medium": 0.7145, "eval_map_plant": 0.7573, "eval_map_small": 0.1071, "eval_mar_1": 0.2777, "eval_mar_10": 0.7524, "eval_mar_100": 0.7553, "eval_mar_100_chicken": 0.7542, "eval_mar_100_duck": 0.6876, "eval_mar_100_plant": 0.8239, "eval_mar_large": 0.8695, "eval_mar_medium": 0.777, "eval_mar_small": 0.1528, "eval_runtime": 7.5326, "eval_samples_per_second": 13.276, "eval_steps_per_second": 1.726, "step": 11000 }, { "epoch": 22.02, "grad_norm": 60.98488998413086, "learning_rate": 7.028898788331e-06, "loss": 0.5607, "step": 11010 }, { "epoch": 22.08, "grad_norm": 47.1700439453125, "learning_rate": 7.014532178568314e-06, "loss": 0.4641, "step": 11040 }, { "epoch": 22.14, "grad_norm": 39.76245880126953, "learning_rate": 7.0001456861863236e-06, "loss": 0.6124, "step": 11070 }, { "epoch": 22.2, "grad_norm": 45.13431167602539, "learning_rate": 6.985739453173903e-06, "loss": 0.5485, "step": 11100 }, { "epoch": 22.26, "grad_norm": 46.35416030883789, "learning_rate": 6.971313621714756e-06, "loss": 0.5222, "step": 11130 }, { "epoch": 22.32, "grad_norm": 60.48912048339844, "learning_rate": 6.9568683341860135e-06, "loss": 0.6344, "step": 11160 }, { "epoch": 22.38, "grad_norm": 170.0771484375, "learning_rate": 6.942403733156832e-06, "loss": 0.6062, "step": 11190 }, { "epoch": 22.44, "grad_norm": 131.03077697753906, "learning_rate": 6.927919961386984e-06, "loss": 0.547, "step": 11220 }, { "epoch": 22.5, "grad_norm": 39.92881774902344, "learning_rate": 6.913417161825449e-06, "loss": 0.5736, "step": 11250 }, { "epoch": 22.56, "grad_norm": 57.132057189941406, "learning_rate": 6.898895477609007e-06, "loss": 0.5973, "step": 11280 }, { "epoch": 22.62, "grad_norm": 61.893489837646484, "learning_rate": 6.884355052060814e-06, "loss": 0.5439, "step": 11310 }, { "epoch": 22.68, "grad_norm": 33.61220169067383, "learning_rate": 6.869796028689002e-06, "loss": 0.5349, "step": 11340 }, { "epoch": 22.74, "grad_norm": 79.79403686523438, "learning_rate": 6.8552185511852555e-06, "loss": 0.5059, "step": 11370 }, { "epoch": 22.8, "grad_norm": 52.177146911621094, "learning_rate": 6.840622763423391e-06, "loss": 0.6265, "step": 11400 }, { "epoch": 22.86, "grad_norm": 74.65804290771484, "learning_rate": 6.82600880945794e-06, "loss": 0.5765, "step": 11430 }, { "epoch": 22.92, "grad_norm": 75.12443542480469, "learning_rate": 6.811376833522729e-06, "loss": 0.4863, "step": 11460 }, { "epoch": 22.98, "grad_norm": 78.99588012695312, "learning_rate": 6.796726980029454e-06, "loss": 0.6119, "step": 11490 }, { "epoch": 23.0, "eval_loss": 0.4347808361053467, "eval_map": 0.7225, "eval_map_50": 0.9419, "eval_map_75": 0.8682, "eval_map_chicken": 0.722, "eval_map_duck": 0.6934, "eval_map_large": 0.8066, "eval_map_medium": 0.7275, "eval_map_plant": 0.7523, "eval_map_small": 0.1274, "eval_mar_1": 0.2856, "eval_mar_10": 0.7717, "eval_mar_100": 0.7769, "eval_mar_100_chicken": 0.7716, "eval_mar_100_duck": 0.7402, "eval_mar_100_plant": 0.819, "eval_mar_large": 0.8703, "eval_mar_medium": 0.7887, "eval_mar_small": 0.233, "eval_runtime": 10.5445, "eval_samples_per_second": 9.484, "eval_steps_per_second": 1.233, "step": 11500 }, { "epoch": 23.04, "grad_norm": 26.74584197998047, "learning_rate": 6.782059393566254e-06, "loss": 0.5109, "step": 11520 }, { "epoch": 23.1, "grad_norm": 0.0011054429924115539, "learning_rate": 6.767374218896286e-06, "loss": 0.5495, "step": 11550 }, { "epoch": 23.16, "grad_norm": 41.22632598876953, "learning_rate": 6.752671600956295e-06, "loss": 0.5339, "step": 11580 }, { "epoch": 23.22, "grad_norm": 73.76976013183594, "learning_rate": 6.737951684855185e-06, "loss": 0.6475, "step": 11610 }, { "epoch": 23.28, "grad_norm": 25.94698715209961, "learning_rate": 6.723214615872585e-06, "loss": 0.5463, "step": 11640 }, { "epoch": 23.34, "grad_norm": 41.29600524902344, "learning_rate": 6.708460539457418e-06, "loss": 0.4973, "step": 11670 }, { "epoch": 23.4, "grad_norm": 257.4337158203125, "learning_rate": 6.693689601226458e-06, "loss": 0.5824, "step": 11700 }, { "epoch": 23.46, "grad_norm": 31.43769645690918, "learning_rate": 6.6789019469629034e-06, "loss": 0.5262, "step": 11730 }, { "epoch": 23.52, "grad_norm": 48.84711456298828, "learning_rate": 6.664097722614934e-06, "loss": 0.5445, "step": 11760 }, { "epoch": 23.58, "grad_norm": 119.98786163330078, "learning_rate": 6.649277074294265e-06, "loss": 0.5373, "step": 11790 }, { "epoch": 23.64, "grad_norm": 64.2603988647461, "learning_rate": 6.634440148274712e-06, "loss": 0.5345, "step": 11820 }, { "epoch": 23.7, "grad_norm": 43.009033203125, "learning_rate": 6.619587090990748e-06, "loss": 0.5367, "step": 11850 }, { "epoch": 23.76, "grad_norm": 55.126644134521484, "learning_rate": 6.604718049036047e-06, "loss": 0.4927, "step": 11880 }, { "epoch": 23.82, "grad_norm": 40.218265533447266, "learning_rate": 6.589833169162055e-06, "loss": 0.6115, "step": 11910 }, { "epoch": 23.88, "grad_norm": 25.923568725585938, "learning_rate": 6.574932598276524e-06, "loss": 0.4811, "step": 11940 }, { "epoch": 23.94, "grad_norm": 23.021484375, "learning_rate": 6.5600164834420754e-06, "loss": 0.521, "step": 11970 }, { "epoch": 24.0, "grad_norm": 59.931358337402344, "learning_rate": 6.545084971874738e-06, "loss": 0.5387, "step": 12000 }, { "epoch": 24.0, "eval_loss": 0.3999063968658447, "eval_map": 0.7462, "eval_map_50": 0.9418, "eval_map_75": 0.8879, "eval_map_chicken": 0.7554, "eval_map_duck": 0.7235, "eval_map_large": 0.8086, "eval_map_medium": 0.7532, "eval_map_plant": 0.7597, "eval_map_small": 0.1444, "eval_mar_1": 0.2923, "eval_mar_10": 0.791, "eval_mar_100": 0.7967, "eval_mar_100_chicken": 0.8009, "eval_mar_100_duck": 0.7639, "eval_mar_100_plant": 0.8254, "eval_mar_large": 0.8749, "eval_mar_medium": 0.8073, "eval_mar_small": 0.264, "eval_runtime": 9.557, "eval_samples_per_second": 10.464, "eval_steps_per_second": 1.36, "step": 12000 }, { "epoch": 24.06, "grad_norm": 55.38029098510742, "learning_rate": 6.530138210942505e-06, "loss": 0.6444, "step": 12030 }, { "epoch": 24.12, "grad_norm": 43.31336212158203, "learning_rate": 6.5151763481638705e-06, "loss": 0.5806, "step": 12060 }, { "epoch": 24.18, "grad_norm": 59.50566482543945, "learning_rate": 6.500199531206381e-06, "loss": 0.527, "step": 12090 }, { "epoch": 24.24, "grad_norm": 38.746559143066406, "learning_rate": 6.485207907885175e-06, "loss": 0.4949, "step": 12120 }, { "epoch": 24.3, "grad_norm": 30.554840087890625, "learning_rate": 6.47020162616152e-06, "loss": 0.434, "step": 12150 }, { "epoch": 24.36, "grad_norm": 39.318824768066406, "learning_rate": 6.455180834141359e-06, "loss": 0.4891, "step": 12180 }, { "epoch": 24.42, "grad_norm": 47.715885162353516, "learning_rate": 6.440145680073847e-06, "loss": 0.6033, "step": 12210 }, { "epoch": 24.48, "grad_norm": 34.92299270629883, "learning_rate": 6.425096312349881e-06, "loss": 0.4579, "step": 12240 }, { "epoch": 24.54, "grad_norm": 67.08885955810547, "learning_rate": 6.410032879500647e-06, "loss": 0.6358, "step": 12270 }, { "epoch": 24.6, "grad_norm": 61.55707931518555, "learning_rate": 6.3949555301961474e-06, "loss": 0.4922, "step": 12300 }, { "epoch": 24.66, "grad_norm": 115.29541015625, "learning_rate": 6.3798644132437304e-06, "loss": 0.4295, "step": 12330 }, { "epoch": 24.72, "grad_norm": 54.56549835205078, "learning_rate": 6.364759677586627e-06, "loss": 0.5117, "step": 12360 }, { "epoch": 24.78, "grad_norm": 67.58968353271484, "learning_rate": 6.349641472302484e-06, "loss": 0.492, "step": 12390 }, { "epoch": 24.84, "grad_norm": 46.7440185546875, "learning_rate": 6.334509946601879e-06, "loss": 0.5179, "step": 12420 }, { "epoch": 24.9, "grad_norm": 49.76848602294922, "learning_rate": 6.3193652498268656e-06, "loss": 0.5426, "step": 12450 }, { "epoch": 24.96, "grad_norm": 120.84765625, "learning_rate": 6.304207531449486e-06, "loss": 0.4956, "step": 12480 }, { "epoch": 25.0, "eval_loss": 0.41766121983528137, "eval_map": 0.7281, "eval_map_50": 0.9384, "eval_map_75": 0.8804, "eval_map_chicken": 0.7288, "eval_map_duck": 0.6835, "eval_map_large": 0.8152, "eval_map_medium": 0.741, "eval_map_plant": 0.772, "eval_map_small": 0.1213, "eval_mar_1": 0.2859, "eval_mar_10": 0.7708, "eval_mar_100": 0.7777, "eval_mar_100_chicken": 0.7809, "eval_mar_100_duck": 0.7206, "eval_mar_100_plant": 0.8317, "eval_mar_large": 0.8711, "eval_mar_medium": 0.7957, "eval_mar_small": 0.2826, "eval_runtime": 7.6762, "eval_samples_per_second": 13.027, "eval_steps_per_second": 1.694, "step": 12500 }, { "epoch": 25.02, "grad_norm": 69.6709976196289, "learning_rate": 6.2890369410703e-06, "loss": 0.5531, "step": 12510 }, { "epoch": 25.08, "grad_norm": 41.09159469604492, "learning_rate": 6.273853628416911e-06, "loss": 0.5175, "step": 12540 }, { "epoch": 25.14, "grad_norm": 33.49006271362305, "learning_rate": 6.258657743342486e-06, "loss": 0.5518, "step": 12570 }, { "epoch": 25.2, "grad_norm": 84.82328796386719, "learning_rate": 6.243449435824276e-06, "loss": 0.4737, "step": 12600 }, { "epoch": 25.26, "grad_norm": 57.00004959106445, "learning_rate": 6.228228855962133e-06, "loss": 0.5362, "step": 12630 }, { "epoch": 25.32, "grad_norm": 21.484079360961914, "learning_rate": 6.212996153977038e-06, "loss": 0.509, "step": 12660 }, { "epoch": 25.38, "grad_norm": 79.40470123291016, "learning_rate": 6.1977514802096105e-06, "loss": 0.5759, "step": 12690 }, { "epoch": 25.44, "grad_norm": 39.21144485473633, "learning_rate": 6.182494985118625e-06, "loss": 0.4345, "step": 12720 }, { "epoch": 25.5, "grad_norm": 42.038299560546875, "learning_rate": 6.1672268192795285e-06, "loss": 0.5121, "step": 12750 }, { "epoch": 25.56, "grad_norm": 88.01210021972656, "learning_rate": 6.151947133382954e-06, "loss": 0.6574, "step": 12780 }, { "epoch": 25.62, "grad_norm": 76.3112564086914, "learning_rate": 6.136656078233233e-06, "loss": 0.45, "step": 12810 }, { "epoch": 25.68, "grad_norm": 94.84544372558594, "learning_rate": 6.121353804746907e-06, "loss": 0.535, "step": 12840 }, { "epoch": 25.74, "grad_norm": 119.24734497070312, "learning_rate": 6.106040463951237e-06, "loss": 0.5369, "step": 12870 }, { "epoch": 25.8, "grad_norm": 49.28893280029297, "learning_rate": 6.090716206982714e-06, "loss": 0.5564, "step": 12900 }, { "epoch": 25.86, "grad_norm": 72.66264343261719, "learning_rate": 6.075381185085568e-06, "loss": 0.5033, "step": 12930 }, { "epoch": 25.92, "grad_norm": 0.0006798353279009461, "learning_rate": 6.060035549610275e-06, "loss": 0.4531, "step": 12960 }, { "epoch": 25.98, "grad_norm": 66.14517211914062, "learning_rate": 6.044679452012059e-06, "loss": 0.5364, "step": 12990 }, { "epoch": 26.0, "eval_loss": 0.412704735994339, "eval_map": 0.7249, "eval_map_50": 0.9376, "eval_map_75": 0.8742, "eval_map_chicken": 0.749, "eval_map_duck": 0.6678, "eval_map_large": 0.8162, "eval_map_medium": 0.7273, "eval_map_plant": 0.7579, "eval_map_small": 0.1036, "eval_mar_1": 0.2889, "eval_mar_10": 0.7694, "eval_mar_100": 0.7741, "eval_mar_100_chicken": 0.7884, "eval_mar_100_duck": 0.7155, "eval_mar_100_plant": 0.8184, "eval_mar_large": 0.8762, "eval_mar_medium": 0.7802, "eval_mar_small": 0.2451, "eval_runtime": 11.0021, "eval_samples_per_second": 9.089, "eval_steps_per_second": 1.182, "step": 13000 }, { "epoch": 26.04, "grad_norm": 41.94633865356445, "learning_rate": 6.029313043849407e-06, "loss": 0.5098, "step": 13020 }, { "epoch": 26.1, "grad_norm": 122.6964340209961, "learning_rate": 6.013936476782563e-06, "loss": 0.5861, "step": 13050 }, { "epoch": 26.16, "grad_norm": 46.03157424926758, "learning_rate": 5.9985499025720354e-06, "loss": 0.4432, "step": 13080 }, { "epoch": 26.22, "grad_norm": 36.8794059753418, "learning_rate": 5.9831534730771e-06, "loss": 0.5072, "step": 13110 }, { "epoch": 26.28, "grad_norm": 41.40494155883789, "learning_rate": 5.967747340254303e-06, "loss": 0.7115, "step": 13140 }, { "epoch": 26.34, "grad_norm": 56.45133972167969, "learning_rate": 5.952331656155951e-06, "loss": 0.5412, "step": 13170 }, { "epoch": 26.4, "grad_norm": 48.933753967285156, "learning_rate": 5.936906572928625e-06, "loss": 0.5466, "step": 13200 }, { "epoch": 26.46, "grad_norm": 53.42213439941406, "learning_rate": 5.9214722428116675e-06, "loss": 0.6037, "step": 13230 }, { "epoch": 26.52, "grad_norm": 24.987173080444336, "learning_rate": 5.906028818135687e-06, "loss": 0.4291, "step": 13260 }, { "epoch": 26.58, "grad_norm": 54.56034469604492, "learning_rate": 5.89057645132105e-06, "loss": 0.5195, "step": 13290 }, { "epoch": 26.64, "grad_norm": 36.93782043457031, "learning_rate": 5.8751152948763815e-06, "loss": 0.4817, "step": 13320 }, { "epoch": 26.7, "grad_norm": 36.74779510498047, "learning_rate": 5.859645501397048e-06, "loss": 0.4762, "step": 13350 }, { "epoch": 26.76, "grad_norm": 50.642822265625, "learning_rate": 5.844167223563669e-06, "loss": 0.7339, "step": 13380 }, { "epoch": 26.82, "grad_norm": 75.78047180175781, "learning_rate": 5.828680614140599e-06, "loss": 0.4807, "step": 13410 }, { "epoch": 26.88, "grad_norm": 37.221744537353516, "learning_rate": 5.813185825974419e-06, "loss": 0.5473, "step": 13440 }, { "epoch": 26.94, "grad_norm": 183.4661865234375, "learning_rate": 5.797683011992432e-06, "loss": 0.6028, "step": 13470 }, { "epoch": 27.0, "grad_norm": 42.405174255371094, "learning_rate": 5.782172325201155e-06, "loss": 0.5416, "step": 13500 }, { "epoch": 27.0, "eval_loss": 0.40053248405456543, "eval_map": 0.7322, "eval_map_50": 0.9448, "eval_map_75": 0.8821, "eval_map_chicken": 0.7431, "eval_map_duck": 0.6924, "eval_map_large": 0.817, "eval_map_medium": 0.7342, "eval_map_plant": 0.7609, "eval_map_small": 0.1353, "eval_mar_1": 0.2911, "eval_mar_10": 0.7768, "eval_mar_100": 0.7819, "eval_mar_100_chicken": 0.7862, "eval_mar_100_duck": 0.732, "eval_mar_100_plant": 0.8277, "eval_mar_large": 0.8837, "eval_mar_medium": 0.7876, "eval_mar_small": 0.2674, "eval_runtime": 10.5605, "eval_samples_per_second": 9.469, "eval_steps_per_second": 1.231, "step": 13500 }, { "epoch": 27.06, "grad_norm": 58.72209930419922, "learning_rate": 5.766653918684803e-06, "loss": 0.4846, "step": 13530 }, { "epoch": 27.12, "grad_norm": 307.8128967285156, "learning_rate": 5.751127945603786e-06, "loss": 0.5458, "step": 13560 }, { "epoch": 27.18, "grad_norm": 55.86375045776367, "learning_rate": 5.735594559193187e-06, "loss": 0.5603, "step": 13590 }, { "epoch": 27.24, "grad_norm": 90.5328598022461, "learning_rate": 5.720053912761261e-06, "loss": 0.4564, "step": 13620 }, { "epoch": 27.3, "grad_norm": 99.12020111083984, "learning_rate": 5.704506159687914e-06, "loss": 0.5712, "step": 13650 }, { "epoch": 27.36, "grad_norm": 57.06706619262695, "learning_rate": 5.68895145342319e-06, "loss": 0.4524, "step": 13680 }, { "epoch": 27.42, "grad_norm": 62.72707748413086, "learning_rate": 5.673389947485763e-06, "loss": 0.4654, "step": 13710 }, { "epoch": 27.48, "grad_norm": 71.98487854003906, "learning_rate": 5.657821795461413e-06, "loss": 0.5257, "step": 13740 }, { "epoch": 27.54, "grad_norm": 39.576778411865234, "learning_rate": 5.642247151001515e-06, "loss": 0.5683, "step": 13770 }, { "epoch": 27.6, "grad_norm": 41.59959030151367, "learning_rate": 5.626666167821522e-06, "loss": 0.4723, "step": 13800 }, { "epoch": 27.66, "grad_norm": 52.05332946777344, "learning_rate": 5.611078999699448e-06, "loss": 0.4785, "step": 13830 }, { "epoch": 27.72, "grad_norm": 120.45262145996094, "learning_rate": 5.59548580047435e-06, "loss": 0.5783, "step": 13860 }, { "epoch": 27.78, "grad_norm": 68.87459564208984, "learning_rate": 5.5798867240448075e-06, "loss": 0.4244, "step": 13890 }, { "epoch": 27.84, "grad_norm": 35.38897705078125, "learning_rate": 5.5642819243674085e-06, "loss": 0.5354, "step": 13920 }, { "epoch": 27.9, "grad_norm": 55.88825225830078, "learning_rate": 5.548671555455226e-06, "loss": 0.5442, "step": 13950 }, { "epoch": 27.96, "grad_norm": 57.475677490234375, "learning_rate": 5.5330557713763e-06, "loss": 0.5501, "step": 13980 }, { "epoch": 28.0, "eval_loss": 0.37810218334198, "eval_map": 0.7489, "eval_map_50": 0.9464, "eval_map_75": 0.8889, "eval_map_chicken": 0.7609, "eval_map_duck": 0.703, "eval_map_large": 0.8342, "eval_map_medium": 0.7551, "eval_map_plant": 0.7828, "eval_map_small": 0.1938, "eval_mar_1": 0.2961, "eval_mar_10": 0.7926, "eval_mar_100": 0.799, "eval_mar_100_chicken": 0.8049, "eval_mar_100_duck": 0.7495, "eval_mar_100_plant": 0.8427, "eval_mar_large": 0.8912, "eval_mar_medium": 0.8092, "eval_mar_small": 0.2983, "eval_runtime": 9.524, "eval_samples_per_second": 10.5, "eval_steps_per_second": 1.365, "step": 14000 }, { "epoch": 28.02, "grad_norm": 46.04853820800781, "learning_rate": 5.517434726252113e-06, "loss": 0.4503, "step": 14010 }, { "epoch": 28.08, "grad_norm": 62.56035232543945, "learning_rate": 5.5018085742560745e-06, "loss": 0.5132, "step": 14040 }, { "epoch": 28.14, "grad_norm": 41.57303237915039, "learning_rate": 5.486177469611999e-06, "loss": 0.5459, "step": 14070 }, { "epoch": 28.2, "grad_norm": 100.6478500366211, "learning_rate": 5.470541566592573e-06, "loss": 0.5247, "step": 14100 }, { "epoch": 28.26, "grad_norm": 104.96534729003906, "learning_rate": 5.454901019517851e-06, "loss": 0.5616, "step": 14130 }, { "epoch": 28.32, "grad_norm": 51.864418029785156, "learning_rate": 5.439255982753717e-06, "loss": 0.4742, "step": 14160 }, { "epoch": 28.38, "grad_norm": 46.259525299072266, "learning_rate": 5.423606610710368e-06, "loss": 0.4681, "step": 14190 }, { "epoch": 28.44, "grad_norm": 27.70792007446289, "learning_rate": 5.4079530578407895e-06, "loss": 0.5917, "step": 14220 }, { "epoch": 28.5, "grad_norm": 74.85920715332031, "learning_rate": 5.392295478639226e-06, "loss": 0.4299, "step": 14250 }, { "epoch": 28.56, "grad_norm": 37.55402755737305, "learning_rate": 5.376634027639664e-06, "loss": 0.4442, "step": 14280 }, { "epoch": 28.62, "grad_norm": 29.590469360351562, "learning_rate": 5.360968859414305e-06, "loss": 0.5044, "step": 14310 }, { "epoch": 28.68, "grad_norm": 82.8427963256836, "learning_rate": 5.345300128572031e-06, "loss": 0.5354, "step": 14340 }, { "epoch": 28.74, "grad_norm": 86.98597717285156, "learning_rate": 5.32962798975689e-06, "loss": 0.444, "step": 14370 }, { "epoch": 28.8, "grad_norm": 43.49721145629883, "learning_rate": 5.3139525976465675e-06, "loss": 0.4828, "step": 14400 }, { "epoch": 28.86, "grad_norm": 54.619632720947266, "learning_rate": 5.298274106950855e-06, "loss": 0.5465, "step": 14430 }, { "epoch": 28.92, "grad_norm": 55.61062240600586, "learning_rate": 5.282592672410124e-06, "loss": 0.4583, "step": 14460 }, { "epoch": 28.98, "grad_norm": 68.1385726928711, "learning_rate": 5.2669084487938025e-06, "loss": 0.4701, "step": 14490 }, { "epoch": 29.0, "eval_loss": 0.3748074471950531, "eval_map": 0.746, "eval_map_50": 0.9561, "eval_map_75": 0.8732, "eval_map_chicken": 0.7472, "eval_map_duck": 0.7174, "eval_map_large": 0.8241, "eval_map_medium": 0.7487, "eval_map_plant": 0.7735, "eval_map_small": 0.1562, "eval_mar_1": 0.294, "eval_mar_10": 0.7888, "eval_mar_100": 0.7966, "eval_mar_100_chicken": 0.7929, "eval_mar_100_duck": 0.7588, "eval_mar_100_plant": 0.838, "eval_mar_large": 0.8849, "eval_mar_medium": 0.8073, "eval_mar_small": 0.3011, "eval_runtime": 7.583, "eval_samples_per_second": 13.187, "eval_steps_per_second": 1.714, "step": 14500 }, { "epoch": 29.04, "grad_norm": 26.128982543945312, "learning_rate": 5.251221590898848e-06, "loss": 0.4499, "step": 14520 }, { "epoch": 29.1, "grad_norm": 126.883544921875, "learning_rate": 5.235532253548213e-06, "loss": 0.5285, "step": 14550 }, { "epoch": 29.16, "grad_norm": 56.32673645019531, "learning_rate": 5.219840591589325e-06, "loss": 0.4968, "step": 14580 }, { "epoch": 29.22, "grad_norm": 80.58277130126953, "learning_rate": 5.204146759892551e-06, "loss": 0.4104, "step": 14610 }, { "epoch": 29.28, "grad_norm": 56.960994720458984, "learning_rate": 5.188450913349674e-06, "loss": 0.4462, "step": 14640 }, { "epoch": 29.34, "grad_norm": 78.12564086914062, "learning_rate": 5.172753206872363e-06, "loss": 0.5444, "step": 14670 }, { "epoch": 29.4, "grad_norm": 55.678897857666016, "learning_rate": 5.157053795390642e-06, "loss": 0.4815, "step": 14700 }, { "epoch": 29.46, "grad_norm": 58.808937072753906, "learning_rate": 5.141352833851367e-06, "loss": 0.4969, "step": 14730 }, { "epoch": 29.52, "grad_norm": 65.06661224365234, "learning_rate": 5.1256504772166885e-06, "loss": 0.4986, "step": 14760 }, { "epoch": 29.58, "grad_norm": 50.332489013671875, "learning_rate": 5.109946880462526e-06, "loss": 0.6113, "step": 14790 }, { "epoch": 29.64, "grad_norm": 74.99099731445312, "learning_rate": 5.0942421985770415e-06, "loss": 0.4475, "step": 14820 }, { "epoch": 29.7, "grad_norm": 41.069637298583984, "learning_rate": 5.078536586559104e-06, "loss": 0.3844, "step": 14850 }, { "epoch": 29.76, "grad_norm": 82.48880767822266, "learning_rate": 5.062830199416764e-06, "loss": 0.479, "step": 14880 }, { "epoch": 29.82, "grad_norm": 38.15917205810547, "learning_rate": 5.047123192165722e-06, "loss": 0.5093, "step": 14910 }, { "epoch": 29.88, "grad_norm": 45.48991394042969, "learning_rate": 5.031415719827796e-06, "loss": 0.5246, "step": 14940 }, { "epoch": 29.94, "grad_norm": 87.992919921875, "learning_rate": 5.015707937429398e-06, "loss": 0.428, "step": 14970 }, { "epoch": 30.0, "grad_norm": 36.047088623046875, "learning_rate": 5e-06, "loss": 0.546, "step": 15000 }, { "epoch": 30.0, "eval_loss": 0.36111319065093994, "eval_map": 0.754, "eval_map_50": 0.9555, "eval_map_75": 0.888, "eval_map_chicken": 0.7637, "eval_map_duck": 0.7205, "eval_map_large": 0.8282, "eval_map_medium": 0.7556, "eval_map_plant": 0.7777, "eval_map_small": 0.1776, "eval_mar_1": 0.2943, "eval_mar_10": 0.7977, "eval_mar_100": 0.8067, "eval_mar_100_chicken": 0.804, "eval_mar_100_duck": 0.7701, "eval_mar_100_plant": 0.8461, "eval_mar_large": 0.8925, "eval_mar_medium": 0.8124, "eval_mar_small": 0.3636, "eval_runtime": 9.5199, "eval_samples_per_second": 10.504, "eval_steps_per_second": 1.366, "step": 15000 }, { "epoch": 30.06, "grad_norm": 70.761962890625, "learning_rate": 4.984292062570603e-06, "loss": 0.5002, "step": 15030 }, { "epoch": 30.12, "grad_norm": 102.7095718383789, "learning_rate": 4.968584280172206e-06, "loss": 0.4323, "step": 15060 }, { "epoch": 30.18, "grad_norm": 42.05283737182617, "learning_rate": 4.952876807834281e-06, "loss": 0.6336, "step": 15090 }, { "epoch": 30.24, "grad_norm": 53.3895378112793, "learning_rate": 4.937169800583237e-06, "loss": 0.4871, "step": 15120 }, { "epoch": 30.3, "grad_norm": 43.38671875, "learning_rate": 4.921463413440898e-06, "loss": 0.4985, "step": 15150 }, { "epoch": 30.36, "grad_norm": 44.3016471862793, "learning_rate": 4.90575780142296e-06, "loss": 0.6734, "step": 15180 }, { "epoch": 30.42, "grad_norm": 73.04580688476562, "learning_rate": 4.890053119537475e-06, "loss": 0.4422, "step": 15210 }, { "epoch": 30.48, "grad_norm": 53.75284957885742, "learning_rate": 4.874349522783313e-06, "loss": 0.5524, "step": 15240 }, { "epoch": 30.54, "grad_norm": 558.6416625976562, "learning_rate": 4.8586471661486345e-06, "loss": 0.4209, "step": 15270 }, { "epoch": 30.6, "grad_norm": 180.00204467773438, "learning_rate": 4.842946204609359e-06, "loss": 0.4687, "step": 15300 }, { "epoch": 30.66, "grad_norm": 99.57018280029297, "learning_rate": 4.827246793127639e-06, "loss": 0.4114, "step": 15330 }, { "epoch": 30.72, "grad_norm": 23.57292938232422, "learning_rate": 4.811549086650327e-06, "loss": 0.5523, "step": 15360 }, { "epoch": 30.78, "grad_norm": 53.311519622802734, "learning_rate": 4.79585324010745e-06, "loss": 0.4957, "step": 15390 }, { "epoch": 30.84, "grad_norm": 36.7927360534668, "learning_rate": 4.780159408410677e-06, "loss": 0.4962, "step": 15420 }, { "epoch": 30.9, "grad_norm": 74.69202423095703, "learning_rate": 4.7644677464517874e-06, "loss": 0.4058, "step": 15450 }, { "epoch": 30.96, "grad_norm": 55.19586181640625, "learning_rate": 4.748778409101153e-06, "loss": 0.4711, "step": 15480 }, { "epoch": 31.0, "eval_loss": 0.36913931369781494, "eval_map": 0.738, "eval_map_50": 0.9548, "eval_map_75": 0.8825, "eval_map_chicken": 0.7441, "eval_map_duck": 0.6898, "eval_map_large": 0.8303, "eval_map_medium": 0.7392, "eval_map_plant": 0.7801, "eval_map_small": 0.1775, "eval_mar_1": 0.2898, "eval_mar_10": 0.7832, "eval_mar_100": 0.7896, "eval_mar_100_chicken": 0.792, "eval_mar_100_duck": 0.7268, "eval_mar_100_plant": 0.8501, "eval_mar_large": 0.8958, "eval_mar_medium": 0.7987, "eval_mar_small": 0.3208, "eval_runtime": 11.9933, "eval_samples_per_second": 8.338, "eval_steps_per_second": 1.084, "step": 15500 }, { "epoch": 31.02, "grad_norm": 92.30634307861328, "learning_rate": 4.733091551206198e-06, "loss": 0.4013, "step": 15510 }, { "epoch": 31.08, "grad_norm": 47.510650634765625, "learning_rate": 4.717407327589878e-06, "loss": 0.4796, "step": 15540 }, { "epoch": 31.14, "grad_norm": 53.91073226928711, "learning_rate": 4.7017258930491474e-06, "loss": 0.4313, "step": 15570 }, { "epoch": 31.2, "grad_norm": 55.71300506591797, "learning_rate": 4.686047402353433e-06, "loss": 0.4614, "step": 15600 }, { "epoch": 31.26, "grad_norm": 56.91154479980469, "learning_rate": 4.670372010243111e-06, "loss": 0.4677, "step": 15630 }, { "epoch": 31.32, "grad_norm": 54.592742919921875, "learning_rate": 4.654699871427972e-06, "loss": 0.4067, "step": 15660 }, { "epoch": 31.38, "grad_norm": 28.48797607421875, "learning_rate": 4.639031140585697e-06, "loss": 0.4847, "step": 15690 }, { "epoch": 31.44, "grad_norm": 54.308837890625, "learning_rate": 4.6233659723603374e-06, "loss": 0.4528, "step": 15720 }, { "epoch": 31.5, "grad_norm": 26.43915557861328, "learning_rate": 4.6077045213607765e-06, "loss": 0.5884, "step": 15750 }, { "epoch": 31.56, "grad_norm": 62.31351089477539, "learning_rate": 4.592046942159213e-06, "loss": 0.5168, "step": 15780 }, { "epoch": 31.62, "grad_norm": 44.884273529052734, "learning_rate": 4.576393389289633e-06, "loss": 0.4769, "step": 15810 }, { "epoch": 31.68, "grad_norm": 50.4775390625, "learning_rate": 4.560744017246284e-06, "loss": 0.4913, "step": 15840 }, { "epoch": 31.74, "grad_norm": 202.3798065185547, "learning_rate": 4.545098980482151e-06, "loss": 0.5515, "step": 15870 }, { "epoch": 31.8, "grad_norm": 65.59260559082031, "learning_rate": 4.529458433407429e-06, "loss": 0.4055, "step": 15900 }, { "epoch": 31.86, "grad_norm": 38.597843170166016, "learning_rate": 4.513822530388004e-06, "loss": 0.6307, "step": 15930 }, { "epoch": 31.92, "grad_norm": 96.9299087524414, "learning_rate": 4.4981914257439254e-06, "loss": 0.473, "step": 15960 }, { "epoch": 31.98, "grad_norm": 73.7803726196289, "learning_rate": 4.482565273747888e-06, "loss": 0.5249, "step": 15990 }, { "epoch": 32.0, "eval_loss": 0.38172683119773865, "eval_map": 0.7188, "eval_map_50": 0.9471, "eval_map_75": 0.8915, "eval_map_chicken": 0.7259, "eval_map_duck": 0.6546, "eval_map_large": 0.8251, "eval_map_medium": 0.7223, "eval_map_plant": 0.776, "eval_map_small": 0.1811, "eval_mar_1": 0.2834, "eval_mar_10": 0.7669, "eval_mar_100": 0.7739, "eval_mar_100_chicken": 0.7813, "eval_mar_100_duck": 0.701, "eval_mar_100_plant": 0.8395, "eval_mar_large": 0.8854, "eval_mar_medium": 0.7827, "eval_mar_small": 0.3142, "eval_runtime": 8.057, "eval_samples_per_second": 12.412, "eval_steps_per_second": 1.614, "step": 16000 }, { "epoch": 32.04, "grad_norm": 46.07328796386719, "learning_rate": 4.466944228623701e-06, "loss": 0.5164, "step": 16020 }, { "epoch": 32.1, "grad_norm": 76.09744262695312, "learning_rate": 4.451328444544774e-06, "loss": 0.5348, "step": 16050 }, { "epoch": 32.16, "grad_norm": 60.05915451049805, "learning_rate": 4.4357180756325915e-06, "loss": 0.5307, "step": 16080 }, { "epoch": 32.22, "grad_norm": 24.423057556152344, "learning_rate": 4.420113275955193e-06, "loss": 0.4878, "step": 16110 }, { "epoch": 32.28, "grad_norm": 44.257347106933594, "learning_rate": 4.404514199525651e-06, "loss": 0.4149, "step": 16140 }, { "epoch": 32.34, "grad_norm": 131.98077392578125, "learning_rate": 4.388921000300553e-06, "loss": 0.5315, "step": 16170 }, { "epoch": 32.4, "grad_norm": 46.79898452758789, "learning_rate": 4.373333832178478e-06, "loss": 0.4515, "step": 16200 }, { "epoch": 32.46, "grad_norm": 265.77301025390625, "learning_rate": 4.357752848998486e-06, "loss": 0.4715, "step": 16230 }, { "epoch": 32.52, "grad_norm": 79.45513916015625, "learning_rate": 4.342178204538588e-06, "loss": 0.5229, "step": 16260 }, { "epoch": 32.58, "grad_norm": 43.29924392700195, "learning_rate": 4.326610052514238e-06, "loss": 0.4697, "step": 16290 }, { "epoch": 32.64, "grad_norm": 51.02996826171875, "learning_rate": 4.31104854657681e-06, "loss": 0.4369, "step": 16320 }, { "epoch": 32.7, "grad_norm": 43.631595611572266, "learning_rate": 4.295493840312087e-06, "loss": 0.5413, "step": 16350 }, { "epoch": 32.76, "grad_norm": 69.57587432861328, "learning_rate": 4.279946087238739e-06, "loss": 0.5689, "step": 16380 }, { "epoch": 32.82, "grad_norm": 40.529869079589844, "learning_rate": 4.264405440806813e-06, "loss": 0.5027, "step": 16410 }, { "epoch": 32.88, "grad_norm": 83.9439926147461, "learning_rate": 4.248872054396215e-06, "loss": 0.483, "step": 16440 }, { "epoch": 32.94, "grad_norm": 63.32405090332031, "learning_rate": 4.233346081315197e-06, "loss": 0.5473, "step": 16470 }, { "epoch": 33.0, "grad_norm": 0.015496279112994671, "learning_rate": 4.217827674798845e-06, "loss": 0.4435, "step": 16500 }, { "epoch": 33.0, "eval_loss": 0.36879783868789673, "eval_map": 0.7377, "eval_map_50": 0.9499, "eval_map_75": 0.8894, "eval_map_chicken": 0.7533, "eval_map_duck": 0.6932, "eval_map_large": 0.8186, "eval_map_medium": 0.7359, "eval_map_plant": 0.7666, "eval_map_small": 0.1846, "eval_mar_1": 0.2893, "eval_mar_10": 0.7836, "eval_mar_100": 0.788, "eval_mar_100_chicken": 0.7978, "eval_mar_100_duck": 0.7361, "eval_mar_100_plant": 0.8303, "eval_mar_large": 0.8795, "eval_mar_medium": 0.7972, "eval_mar_small": 0.2943, "eval_runtime": 9.3979, "eval_samples_per_second": 10.641, "eval_steps_per_second": 1.383, "step": 16500 }, { "epoch": 33.06, "grad_norm": 76.0846939086914, "learning_rate": 4.202316988007568e-06, "loss": 0.418, "step": 16530 }, { "epoch": 33.12, "grad_norm": 51.79882049560547, "learning_rate": 4.186814174025582e-06, "loss": 0.4236, "step": 16560 }, { "epoch": 33.18, "grad_norm": 96.94397735595703, "learning_rate": 4.171319385859402e-06, "loss": 0.3717, "step": 16590 }, { "epoch": 33.24, "grad_norm": 62.75301742553711, "learning_rate": 4.155832776436331e-06, "loss": 0.483, "step": 16620 }, { "epoch": 33.3, "grad_norm": 76.92266082763672, "learning_rate": 4.140354498602952e-06, "loss": 0.4555, "step": 16650 }, { "epoch": 33.36, "grad_norm": 33.57876205444336, "learning_rate": 4.124884705123619e-06, "loss": 0.5009, "step": 16680 }, { "epoch": 33.42, "grad_norm": 134.94790649414062, "learning_rate": 4.109423548678949e-06, "loss": 0.4169, "step": 16710 }, { "epoch": 33.48, "grad_norm": 101.32524871826172, "learning_rate": 4.093971181864313e-06, "loss": 0.4283, "step": 16740 }, { "epoch": 33.54, "grad_norm": 50.135074615478516, "learning_rate": 4.078527757188333e-06, "loss": 0.4667, "step": 16770 }, { "epoch": 33.6, "grad_norm": 27.48794174194336, "learning_rate": 4.063093427071376e-06, "loss": 0.5435, "step": 16800 }, { "epoch": 33.66, "grad_norm": 206.2461700439453, "learning_rate": 4.047668343844051e-06, "loss": 0.5515, "step": 16830 }, { "epoch": 33.72, "grad_norm": 31.39567756652832, "learning_rate": 4.032252659745699e-06, "loss": 0.4709, "step": 16860 }, { "epoch": 33.78, "grad_norm": 38.97549057006836, "learning_rate": 4.016846526922901e-06, "loss": 0.5223, "step": 16890 }, { "epoch": 33.84, "grad_norm": 47.43185043334961, "learning_rate": 4.001450097427965e-06, "loss": 0.475, "step": 16920 }, { "epoch": 33.9, "grad_norm": 29.89405059814453, "learning_rate": 3.986063523217439e-06, "loss": 0.4764, "step": 16950 }, { "epoch": 33.96, "grad_norm": 42.64591598510742, "learning_rate": 3.970686956150595e-06, "loss": 0.4619, "step": 16980 }, { "epoch": 34.0, "eval_loss": 0.37409254908561707, "eval_map": 0.7273, "eval_map_50": 0.9611, "eval_map_75": 0.887, "eval_map_chicken": 0.717, "eval_map_duck": 0.6876, "eval_map_large": 0.8296, "eval_map_medium": 0.7193, "eval_map_plant": 0.7772, "eval_map_small": 0.1705, "eval_mar_1": 0.2836, "eval_mar_10": 0.7749, "eval_mar_100": 0.7814, "eval_mar_100_chicken": 0.7693, "eval_mar_100_duck": 0.734, "eval_mar_100_plant": 0.8409, "eval_mar_large": 0.8904, "eval_mar_medium": 0.7827, "eval_mar_small": 0.3674, "eval_runtime": 10.2909, "eval_samples_per_second": 9.717, "eval_steps_per_second": 1.263, "step": 17000 }, { "epoch": 34.02, "grad_norm": 48.648597717285156, "learning_rate": 3.955320547987943e-06, "loss": 0.642, "step": 17010 }, { "epoch": 34.08, "grad_norm": 70.59295654296875, "learning_rate": 3.939964450389728e-06, "loss": 0.3999, "step": 17040 }, { "epoch": 34.14, "grad_norm": 41.141754150390625, "learning_rate": 3.924618814914435e-06, "loss": 0.4203, "step": 17070 }, { "epoch": 34.2, "grad_norm": 341.19921875, "learning_rate": 3.909283793017289e-06, "loss": 0.5142, "step": 17100 }, { "epoch": 34.26, "grad_norm": 691.110595703125, "learning_rate": 3.8939595360487655e-06, "loss": 0.4803, "step": 17130 }, { "epoch": 34.32, "grad_norm": 53.05097198486328, "learning_rate": 3.8786461952530955e-06, "loss": 0.449, "step": 17160 }, { "epoch": 34.38, "grad_norm": 123.60619354248047, "learning_rate": 3.863343921766769e-06, "loss": 0.4565, "step": 17190 }, { "epoch": 34.44, "grad_norm": 114.92799377441406, "learning_rate": 3.8480528666170495e-06, "loss": 0.4355, "step": 17220 }, { "epoch": 34.5, "grad_norm": 32.30895233154297, "learning_rate": 3.832773180720475e-06, "loss": 0.4041, "step": 17250 }, { "epoch": 34.56, "grad_norm": 68.51344299316406, "learning_rate": 3.817505014881378e-06, "loss": 0.5661, "step": 17280 }, { "epoch": 34.62, "grad_norm": 38.61069107055664, "learning_rate": 3.8022485197903924e-06, "loss": 0.4216, "step": 17310 }, { "epoch": 34.68, "grad_norm": 128.85658264160156, "learning_rate": 3.787003846022964e-06, "loss": 0.4735, "step": 17340 }, { "epoch": 34.74, "grad_norm": 45.97095489501953, "learning_rate": 3.7717711440378695e-06, "loss": 0.4791, "step": 17370 }, { "epoch": 34.8, "grad_norm": 30.850133895874023, "learning_rate": 3.756550564175727e-06, "loss": 0.4411, "step": 17400 }, { "epoch": 34.86, "grad_norm": 83.11338806152344, "learning_rate": 3.7413422566575153e-06, "loss": 0.4747, "step": 17430 }, { "epoch": 34.92, "grad_norm": 36.881404876708984, "learning_rate": 3.7261463715830902e-06, "loss": 0.3981, "step": 17460 }, { "epoch": 34.98, "grad_norm": 57.52027893066406, "learning_rate": 3.7109630589297014e-06, "loss": 0.4513, "step": 17490 }, { "epoch": 35.0, "eval_loss": 0.34951791167259216, "eval_map": 0.7497, "eval_map_50": 0.9563, "eval_map_75": 0.8932, "eval_map_chicken": 0.764, "eval_map_duck": 0.707, "eval_map_large": 0.8307, "eval_map_medium": 0.7443, "eval_map_plant": 0.7782, "eval_map_small": 0.2015, "eval_mar_1": 0.292, "eval_mar_10": 0.7918, "eval_mar_100": 0.798, "eval_mar_100_chicken": 0.8093, "eval_mar_100_duck": 0.7423, "eval_mar_100_plant": 0.8424, "eval_mar_large": 0.8937, "eval_mar_medium": 0.801, "eval_mar_small": 0.3445, "eval_runtime": 10.2688, "eval_samples_per_second": 9.738, "eval_steps_per_second": 1.266, "step": 17500 }, { "epoch": 35.04, "grad_norm": 48.036277770996094, "learning_rate": 3.695792468550517e-06, "loss": 0.422, "step": 17520 }, { "epoch": 35.1, "grad_norm": 34.991939544677734, "learning_rate": 3.680634750173137e-06, "loss": 0.4611, "step": 17550 }, { "epoch": 35.16, "grad_norm": 91.96014404296875, "learning_rate": 3.6654900533981234e-06, "loss": 0.4899, "step": 17580 }, { "epoch": 35.22, "grad_norm": 53.90528106689453, "learning_rate": 3.6503585276975196e-06, "loss": 0.377, "step": 17610 }, { "epoch": 35.28, "grad_norm": 63.77039337158203, "learning_rate": 3.635240322413375e-06, "loss": 0.4747, "step": 17640 }, { "epoch": 35.34, "grad_norm": 51.25712203979492, "learning_rate": 3.6201355867562725e-06, "loss": 0.4421, "step": 17670 }, { "epoch": 35.4, "grad_norm": 77.92784118652344, "learning_rate": 3.6050444698038547e-06, "loss": 0.3882, "step": 17700 }, { "epoch": 35.46, "grad_norm": 34.65978240966797, "learning_rate": 3.5899671204993535e-06, "loss": 0.5024, "step": 17730 }, { "epoch": 35.52, "grad_norm": 39.55388641357422, "learning_rate": 3.5749036876501196e-06, "loss": 0.4819, "step": 17760 }, { "epoch": 35.58, "grad_norm": 73.00691986083984, "learning_rate": 3.559854319926156e-06, "loss": 0.4058, "step": 17790 }, { "epoch": 35.64, "grad_norm": 60.91654968261719, "learning_rate": 3.5448191658586423e-06, "loss": 0.4322, "step": 17820 }, { "epoch": 35.7, "grad_norm": 43.53456115722656, "learning_rate": 3.5297983738384813e-06, "loss": 0.5322, "step": 17850 }, { "epoch": 35.76, "grad_norm": 68.43659973144531, "learning_rate": 3.5147920921148267e-06, "loss": 0.5169, "step": 17880 }, { "epoch": 35.82, "grad_norm": 0.00025471282424405217, "learning_rate": 3.49980046879362e-06, "loss": 0.5031, "step": 17910 }, { "epoch": 35.88, "grad_norm": 63.586910247802734, "learning_rate": 3.484823651836131e-06, "loss": 0.4705, "step": 17940 }, { "epoch": 35.94, "grad_norm": 78.00933074951172, "learning_rate": 3.4698617890574972e-06, "loss": 0.4476, "step": 17970 }, { "epoch": 36.0, "grad_norm": 39.820011138916016, "learning_rate": 3.4549150281252635e-06, "loss": 0.4197, "step": 18000 }, { "epoch": 36.0, "eval_loss": 0.35867756605148315, "eval_map": 0.7482, "eval_map_50": 0.953, "eval_map_75": 0.8945, "eval_map_chicken": 0.7466, "eval_map_duck": 0.7193, "eval_map_large": 0.8334, "eval_map_medium": 0.7426, "eval_map_plant": 0.7786, "eval_map_small": 0.166, "eval_mar_1": 0.2904, "eval_mar_10": 0.7911, "eval_mar_100": 0.7973, "eval_mar_100_chicken": 0.7996, "eval_mar_100_duck": 0.7495, "eval_mar_100_plant": 0.8429, "eval_mar_large": 0.8946, "eval_mar_medium": 0.8016, "eval_mar_small": 0.3278, "eval_runtime": 7.6538, "eval_samples_per_second": 13.065, "eval_steps_per_second": 1.698, "step": 18000 }, { "epoch": 36.06, "grad_norm": 48.5521240234375, "learning_rate": 3.4399835165579266e-06, "loss": 0.4408, "step": 18030 }, { "epoch": 36.12, "grad_norm": 44.91645431518555, "learning_rate": 3.4250674017234774e-06, "loss": 0.4099, "step": 18060 }, { "epoch": 36.18, "grad_norm": 60.684547424316406, "learning_rate": 3.4101668308379465e-06, "loss": 0.5554, "step": 18090 }, { "epoch": 36.24, "grad_norm": 28.382596969604492, "learning_rate": 3.3952819509639534e-06, "loss": 0.5059, "step": 18120 }, { "epoch": 36.3, "grad_norm": 43.993858337402344, "learning_rate": 3.3804129090092542e-06, "loss": 0.5021, "step": 18150 }, { "epoch": 36.36, "grad_norm": 55.67386245727539, "learning_rate": 3.3655598517252886e-06, "loss": 0.4243, "step": 18180 }, { "epoch": 36.42, "grad_norm": 147.2352294921875, "learning_rate": 3.350722925705736e-06, "loss": 0.4346, "step": 18210 }, { "epoch": 36.48, "grad_norm": 35.09242630004883, "learning_rate": 3.3359022773850673e-06, "loss": 0.437, "step": 18240 }, { "epoch": 36.54, "grad_norm": 44.45463943481445, "learning_rate": 3.3210980530370974e-06, "loss": 0.4346, "step": 18270 }, { "epoch": 36.6, "grad_norm": 49.73167419433594, "learning_rate": 3.3063103987735433e-06, "loss": 0.5052, "step": 18300 }, { "epoch": 36.66, "grad_norm": 81.54788970947266, "learning_rate": 3.2915394605425836e-06, "loss": 0.385, "step": 18330 }, { "epoch": 36.72, "grad_norm": 64.13592529296875, "learning_rate": 3.2767853841274154e-06, "loss": 0.4744, "step": 18360 }, { "epoch": 36.78, "grad_norm": 67.9179458618164, "learning_rate": 3.262048315144816e-06, "loss": 0.5215, "step": 18390 }, { "epoch": 36.84, "grad_norm": 95.48695373535156, "learning_rate": 3.247328399043706e-06, "loss": 0.4883, "step": 18420 }, { "epoch": 36.9, "grad_norm": 65.21953582763672, "learning_rate": 3.2326257811037154e-06, "loss": 0.4719, "step": 18450 }, { "epoch": 36.96, "grad_norm": 56.97268295288086, "learning_rate": 3.217940606433747e-06, "loss": 0.4394, "step": 18480 }, { "epoch": 37.0, "eval_loss": 0.3437317907810211, "eval_map": 0.762, "eval_map_50": 0.9582, "eval_map_75": 0.8987, "eval_map_chicken": 0.7602, "eval_map_duck": 0.7321, "eval_map_large": 0.8432, "eval_map_medium": 0.7605, "eval_map_plant": 0.7936, "eval_map_small": 0.1938, "eval_mar_1": 0.2925, "eval_mar_10": 0.8012, "eval_mar_100": 0.8091, "eval_mar_100_chicken": 0.8049, "eval_mar_100_duck": 0.7701, "eval_mar_100_plant": 0.8524, "eval_mar_large": 0.8987, "eval_mar_medium": 0.8139, "eval_mar_small": 0.3794, "eval_runtime": 7.4333, "eval_samples_per_second": 13.453, "eval_steps_per_second": 1.749, "step": 18500 }, { "epoch": 37.02, "grad_norm": 36.011844635009766, "learning_rate": 3.2032730199705477e-06, "loss": 0.4623, "step": 18510 }, { "epoch": 37.08, "grad_norm": 33.675167083740234, "learning_rate": 3.188623166477272e-06, "loss": 0.4994, "step": 18540 }, { "epoch": 37.14, "grad_norm": 63.0922966003418, "learning_rate": 3.1739911905420617e-06, "loss": 0.3776, "step": 18570 }, { "epoch": 37.2, "grad_norm": 46.88859939575195, "learning_rate": 3.1593772365766107e-06, "loss": 0.3768, "step": 18600 }, { "epoch": 37.26, "grad_norm": 41.93199157714844, "learning_rate": 3.144781448814746e-06, "loss": 0.4363, "step": 18630 }, { "epoch": 37.32, "grad_norm": 64.76182556152344, "learning_rate": 3.130203971310999e-06, "loss": 0.5092, "step": 18660 }, { "epoch": 37.38, "grad_norm": 39.59531784057617, "learning_rate": 3.1156449479391876e-06, "loss": 0.4017, "step": 18690 }, { "epoch": 37.44, "grad_norm": 53.65720748901367, "learning_rate": 3.1011045223909954e-06, "loss": 0.5165, "step": 18720 }, { "epoch": 37.5, "grad_norm": 0.0004234654479660094, "learning_rate": 3.0865828381745515e-06, "loss": 0.416, "step": 18750 }, { "epoch": 37.56, "grad_norm": 38.28908920288086, "learning_rate": 3.0720800386130176e-06, "loss": 0.4026, "step": 18780 }, { "epoch": 37.62, "grad_norm": 61.881065368652344, "learning_rate": 3.0575962668431704e-06, "loss": 0.4578, "step": 18810 }, { "epoch": 37.68, "grad_norm": 25.868053436279297, "learning_rate": 3.043131665813988e-06, "loss": 0.4452, "step": 18840 }, { "epoch": 37.74, "grad_norm": 54.99051284790039, "learning_rate": 3.028686378285245e-06, "loss": 0.4693, "step": 18870 }, { "epoch": 37.8, "grad_norm": 94.88867950439453, "learning_rate": 3.0142605468260976e-06, "loss": 0.3768, "step": 18900 }, { "epoch": 37.86, "grad_norm": 72.27814483642578, "learning_rate": 2.9998543138136773e-06, "loss": 0.4081, "step": 18930 }, { "epoch": 37.92, "grad_norm": 86.77288818359375, "learning_rate": 2.9854678214316875e-06, "loss": 0.3942, "step": 18960 }, { "epoch": 37.98, "grad_norm": 31.16099739074707, "learning_rate": 2.9711012116690007e-06, "loss": 0.4142, "step": 18990 }, { "epoch": 38.0, "eval_loss": 0.3388429284095764, "eval_map": 0.7684, "eval_map_50": 0.9581, "eval_map_75": 0.9076, "eval_map_chicken": 0.7749, "eval_map_duck": 0.7329, "eval_map_large": 0.8505, "eval_map_medium": 0.7631, "eval_map_plant": 0.7975, "eval_map_small": 0.1565, "eval_mar_1": 0.2967, "eval_mar_10": 0.8092, "eval_mar_100": 0.8171, "eval_mar_100_chicken": 0.8187, "eval_mar_100_duck": 0.7763, "eval_mar_100_plant": 0.8562, "eval_mar_large": 0.9088, "eval_mar_medium": 0.8171, "eval_mar_small": 0.3659, "eval_runtime": 7.8033, "eval_samples_per_second": 12.815, "eval_steps_per_second": 1.666, "step": 19000 }, { "epoch": 38.04, "grad_norm": 35.37294006347656, "learning_rate": 2.9567546263182554e-06, "loss": 0.4947, "step": 19020 }, { "epoch": 38.1, "grad_norm": 70.90824890136719, "learning_rate": 2.9424282069744564e-06, "loss": 0.4623, "step": 19050 }, { "epoch": 38.16, "grad_norm": 38.94876480102539, "learning_rate": 2.92812209503358e-06, "loss": 0.4326, "step": 19080 }, { "epoch": 38.22, "grad_norm": 210.5827178955078, "learning_rate": 2.9138364316911747e-06, "loss": 0.4174, "step": 19110 }, { "epoch": 38.28, "grad_norm": 102.10295104980469, "learning_rate": 2.899571357940969e-06, "loss": 0.4394, "step": 19140 }, { "epoch": 38.34, "grad_norm": 81.99002838134766, "learning_rate": 2.8853270145734846e-06, "loss": 0.4243, "step": 19170 }, { "epoch": 38.4, "grad_norm": 49.58351135253906, "learning_rate": 2.871103542174637e-06, "loss": 0.4023, "step": 19200 }, { "epoch": 38.46, "grad_norm": 56.82831954956055, "learning_rate": 2.856901081124359e-06, "loss": 0.5199, "step": 19230 }, { "epoch": 38.52, "grad_norm": 69.1905746459961, "learning_rate": 2.8427197715952047e-06, "loss": 0.4749, "step": 19260 }, { "epoch": 38.58, "grad_norm": 43.73515701293945, "learning_rate": 2.8285597535509775e-06, "loss": 0.4608, "step": 19290 }, { "epoch": 38.64, "grad_norm": 90.28402709960938, "learning_rate": 2.814421166745337e-06, "loss": 0.4971, "step": 19320 }, { "epoch": 38.7, "grad_norm": 43.67237091064453, "learning_rate": 2.800304150720424e-06, "loss": 0.4661, "step": 19350 }, { "epoch": 38.76, "grad_norm": 29.571645736694336, "learning_rate": 2.7862088448054936e-06, "loss": 0.5271, "step": 19380 }, { "epoch": 38.82, "grad_norm": 95.29782104492188, "learning_rate": 2.772135388115519e-06, "loss": 0.4412, "step": 19410 }, { "epoch": 38.88, "grad_norm": 59.25728988647461, "learning_rate": 2.7580839195498397e-06, "loss": 0.4069, "step": 19440 }, { "epoch": 38.94, "grad_norm": 31.977981567382812, "learning_rate": 2.7440545777907747e-06, "loss": 0.3996, "step": 19470 }, { "epoch": 39.0, "grad_norm": 60.71751022338867, "learning_rate": 2.7300475013022666e-06, "loss": 0.4683, "step": 19500 }, { "epoch": 39.0, "eval_loss": 0.3520563542842865, "eval_map": 0.7482, "eval_map_50": 0.9575, "eval_map_75": 0.8904, "eval_map_chicken": 0.7463, "eval_map_duck": 0.7071, "eval_map_large": 0.8448, "eval_map_medium": 0.7398, "eval_map_plant": 0.7911, "eval_map_small": 0.1854, "eval_mar_1": 0.2897, "eval_mar_10": 0.7978, "eval_mar_100": 0.8052, "eval_mar_100_chicken": 0.7987, "eval_mar_100_duck": 0.7639, "eval_mar_100_plant": 0.853, "eval_mar_large": 0.9067, "eval_mar_medium": 0.8007, "eval_mar_small": 0.4036, "eval_runtime": 7.8535, "eval_samples_per_second": 12.733, "eval_steps_per_second": 1.655, "step": 19500 }, { "epoch": 39.06, "grad_norm": 45.21415710449219, "learning_rate": 2.716062828328502e-06, "loss": 0.4027, "step": 19530 }, { "epoch": 39.12, "grad_norm": 54.45733642578125, "learning_rate": 2.7021006968925613e-06, "loss": 0.4419, "step": 19560 }, { "epoch": 39.18, "grad_norm": 33.80492401123047, "learning_rate": 2.6881612447950425e-06, "loss": 0.4048, "step": 19590 }, { "epoch": 39.24, "grad_norm": 78.0910873413086, "learning_rate": 2.6742446096127086e-06, "loss": 0.325, "step": 19620 }, { "epoch": 39.3, "grad_norm": 41.879154205322266, "learning_rate": 2.6603509286971342e-06, "loss": 0.4842, "step": 19650 }, { "epoch": 39.36, "grad_norm": 70.68565368652344, "learning_rate": 2.646480339173337e-06, "loss": 0.4194, "step": 19680 }, { "epoch": 39.42, "grad_norm": 59.25823974609375, "learning_rate": 2.6326329779384397e-06, "loss": 0.4032, "step": 19710 }, { "epoch": 39.48, "grad_norm": 49.112060546875, "learning_rate": 2.618808981660304e-06, "loss": 0.4432, "step": 19740 }, { "epoch": 39.54, "grad_norm": 44.24351501464844, "learning_rate": 2.6050084867761953e-06, "loss": 0.4549, "step": 19770 }, { "epoch": 39.6, "grad_norm": 63.681114196777344, "learning_rate": 2.5912316294914232e-06, "loss": 0.5571, "step": 19800 }, { "epoch": 39.66, "grad_norm": 36.10969543457031, "learning_rate": 2.5774785457780107e-06, "loss": 0.4352, "step": 19830 }, { "epoch": 39.72, "grad_norm": 191.165771484375, "learning_rate": 2.5637493713733376e-06, "loss": 0.4598, "step": 19860 }, { "epoch": 39.78, "grad_norm": 67.50877380371094, "learning_rate": 2.5500442417788176e-06, "loss": 0.4609, "step": 19890 }, { "epoch": 39.84, "grad_norm": 52.10002136230469, "learning_rate": 2.536363292258543e-06, "loss": 0.3956, "step": 19920 }, { "epoch": 39.9, "grad_norm": 91.81877899169922, "learning_rate": 2.5227066578379624e-06, "loss": 0.5198, "step": 19950 }, { "epoch": 39.96, "grad_norm": 38.71651840209961, "learning_rate": 2.509074473302546e-06, "loss": 0.4344, "step": 19980 }, { "epoch": 40.0, "eval_loss": 0.3478543162345886, "eval_map": 0.7455, "eval_map_50": 0.9568, "eval_map_75": 0.8983, "eval_map_chicken": 0.7558, "eval_map_duck": 0.6949, "eval_map_large": 0.8361, "eval_map_medium": 0.7413, "eval_map_plant": 0.7857, "eval_map_small": 0.1845, "eval_mar_1": 0.2906, "eval_mar_10": 0.7903, "eval_mar_100": 0.7999, "eval_mar_100_chicken": 0.8071, "eval_mar_100_duck": 0.7433, "eval_mar_100_plant": 0.8493, "eval_mar_large": 0.8992, "eval_mar_medium": 0.8022, "eval_mar_small": 0.3661, "eval_runtime": 9.4913, "eval_samples_per_second": 10.536, "eval_steps_per_second": 1.37, "step": 20000 }, { "epoch": 40.02, "grad_norm": 47.04460906982422, "learning_rate": 2.49546687319645e-06, "loss": 0.4891, "step": 20010 }, { "epoch": 40.08, "grad_norm": 73.42700958251953, "learning_rate": 2.4818839918211963e-06, "loss": 0.3979, "step": 20040 }, { "epoch": 40.14, "grad_norm": 51.89419937133789, "learning_rate": 2.4683259632343363e-06, "loss": 0.4045, "step": 20070 }, { "epoch": 40.2, "grad_norm": 65.21696472167969, "learning_rate": 2.4547929212481436e-06, "loss": 0.4696, "step": 20100 }, { "epoch": 40.26, "grad_norm": 68.36051177978516, "learning_rate": 2.4412849994282744e-06, "loss": 0.4553, "step": 20130 }, { "epoch": 40.32, "grad_norm": 82.73648071289062, "learning_rate": 2.4278023310924676e-06, "loss": 0.4654, "step": 20160 }, { "epoch": 40.38, "grad_norm": 47.76723861694336, "learning_rate": 2.4143450493092146e-06, "loss": 0.4303, "step": 20190 }, { "epoch": 40.44, "grad_norm": 70.81723022460938, "learning_rate": 2.4009132868964525e-06, "loss": 0.5423, "step": 20220 }, { "epoch": 40.5, "grad_norm": 127.06489562988281, "learning_rate": 2.387507176420256e-06, "loss": 0.4165, "step": 20250 }, { "epoch": 40.56, "grad_norm": 88.79576873779297, "learning_rate": 2.3741268501935212e-06, "loss": 0.4677, "step": 20280 }, { "epoch": 40.62, "grad_norm": 93.29864501953125, "learning_rate": 2.3607724402746685e-06, "loss": 0.4255, "step": 20310 }, { "epoch": 40.68, "grad_norm": 55.40361404418945, "learning_rate": 2.3474440784663287e-06, "loss": 0.4263, "step": 20340 }, { "epoch": 40.74, "grad_norm": 68.82120513916016, "learning_rate": 2.334141896314057e-06, "loss": 0.4347, "step": 20370 }, { "epoch": 40.8, "grad_norm": 69.16571044921875, "learning_rate": 2.320866025105016e-06, "loss": 0.4773, "step": 20400 }, { "epoch": 40.86, "grad_norm": 32.513587951660156, "learning_rate": 2.3076165958666992e-06, "loss": 0.4008, "step": 20430 }, { "epoch": 40.92, "grad_norm": 33.8734130859375, "learning_rate": 2.294393739365621e-06, "loss": 0.4451, "step": 20460 }, { "epoch": 40.98, "grad_norm": 59.35488510131836, "learning_rate": 2.281197586106037e-06, "loss": 0.3684, "step": 20490 }, { "epoch": 41.0, "eval_loss": 0.3292124271392822, "eval_map": 0.763, "eval_map_50": 0.9532, "eval_map_75": 0.896, "eval_map_chicken": 0.764, "eval_map_duck": 0.7257, "eval_map_large": 0.854, "eval_map_medium": 0.7621, "eval_map_plant": 0.7992, "eval_map_small": 0.1682, "eval_mar_1": 0.3003, "eval_mar_10": 0.8063, "eval_mar_100": 0.8136, "eval_mar_100_chicken": 0.8164, "eval_mar_100_duck": 0.766, "eval_mar_100_plant": 0.8585, "eval_mar_large": 0.9117, "eval_mar_medium": 0.8172, "eval_mar_small": 0.3305, "eval_runtime": 9.5619, "eval_samples_per_second": 10.458, "eval_steps_per_second": 1.36, "step": 20500 }, { "epoch": 41.04, "grad_norm": 32.97661590576172, "learning_rate": 2.268028266328655e-06, "loss": 0.4535, "step": 20520 }, { "epoch": 41.1, "grad_norm": 66.92311096191406, "learning_rate": 2.254885910009341e-06, "loss": 0.4812, "step": 20550 }, { "epoch": 41.16, "grad_norm": 84.38749694824219, "learning_rate": 2.2417706468578495e-06, "loss": 0.5275, "step": 20580 }, { "epoch": 41.22, "grad_norm": 69.59208679199219, "learning_rate": 2.228682606316529e-06, "loss": 0.5912, "step": 20610 }, { "epoch": 41.28, "grad_norm": 31.722850799560547, "learning_rate": 2.2156219175590623e-06, "loss": 0.4725, "step": 20640 }, { "epoch": 41.34, "grad_norm": 43.81732177734375, "learning_rate": 2.2025887094891657e-06, "loss": 0.4518, "step": 20670 }, { "epoch": 41.4, "grad_norm": 20.733884811401367, "learning_rate": 2.1895831107393485e-06, "loss": 0.5308, "step": 20700 }, { "epoch": 41.46, "grad_norm": 142.51837158203125, "learning_rate": 2.1766052496696155e-06, "loss": 0.5511, "step": 20730 }, { "epoch": 41.52, "grad_norm": 45.80072021484375, "learning_rate": 2.1636552543662187e-06, "loss": 0.4368, "step": 20760 }, { "epoch": 41.58, "grad_norm": 117.68204498291016, "learning_rate": 2.1507332526403814e-06, "loss": 0.4299, "step": 20790 }, { "epoch": 41.64, "grad_norm": 69.04330444335938, "learning_rate": 2.137839372027047e-06, "loss": 0.4972, "step": 20820 }, { "epoch": 41.7, "grad_norm": 27.10807991027832, "learning_rate": 2.124973739783609e-06, "loss": 0.3879, "step": 20850 }, { "epoch": 41.76, "grad_norm": 39.97993087768555, "learning_rate": 2.112136482888663e-06, "loss": 0.3985, "step": 20880 }, { "epoch": 41.82, "grad_norm": 77.02200317382812, "learning_rate": 2.0993277280407547e-06, "loss": 0.4666, "step": 20910 }, { "epoch": 41.88, "grad_norm": 46.639984130859375, "learning_rate": 2.0865476016571206e-06, "loss": 0.4439, "step": 20940 }, { "epoch": 41.94, "grad_norm": 25.856281280517578, "learning_rate": 2.0737962298724513e-06, "loss": 0.3932, "step": 20970 }, { "epoch": 42.0, "grad_norm": 76.76116943359375, "learning_rate": 2.061073738537635e-06, "loss": 0.6048, "step": 21000 }, { "epoch": 42.0, "eval_loss": 0.32280388474464417, "eval_map": 0.7784, "eval_map_50": 0.9612, "eval_map_75": 0.9024, "eval_map_chicken": 0.7854, "eval_map_duck": 0.7495, "eval_map_large": 0.8552, "eval_map_medium": 0.775, "eval_map_plant": 0.8003, "eval_map_small": 0.1669, "eval_mar_1": 0.3015, "eval_mar_10": 0.8227, "eval_mar_100": 0.8303, "eval_mar_100_chicken": 0.8347, "eval_mar_100_duck": 0.7897, "eval_mar_100_plant": 0.8666, "eval_mar_large": 0.9167, "eval_mar_medium": 0.8304, "eval_mar_small": 0.4153, "eval_runtime": 7.6587, "eval_samples_per_second": 13.057, "eval_steps_per_second": 1.697, "step": 21000 }, { "epoch": 42.06, "grad_norm": 56.64277648925781, "learning_rate": 2.0483802532185286e-06, "loss": 0.4271, "step": 21030 }, { "epoch": 42.12, "grad_norm": 68.33435821533203, "learning_rate": 2.035715899194704e-06, "loss": 0.4412, "step": 21060 }, { "epoch": 42.18, "grad_norm": 46.08116912841797, "learning_rate": 2.0230808014582265e-06, "loss": 0.4633, "step": 21090 }, { "epoch": 42.24, "grad_norm": 0.0011438040528446436, "learning_rate": 2.0104750847124075e-06, "loss": 0.5867, "step": 21120 }, { "epoch": 42.3, "grad_norm": 52.78838348388672, "learning_rate": 1.9978988733705807e-06, "loss": 0.3873, "step": 21150 }, { "epoch": 42.36, "grad_norm": 44.0331916809082, "learning_rate": 1.9853522915548777e-06, "loss": 0.4442, "step": 21180 }, { "epoch": 42.42, "grad_norm": 43.81474304199219, "learning_rate": 1.9728354630949935e-06, "loss": 0.4621, "step": 21210 }, { "epoch": 42.48, "grad_norm": 35.939544677734375, "learning_rate": 1.9603485115269743e-06, "loss": 0.5408, "step": 21240 }, { "epoch": 42.54, "grad_norm": 75.70336151123047, "learning_rate": 1.9478915600919877e-06, "loss": 0.4512, "step": 21270 }, { "epoch": 42.6, "grad_norm": 74.07784271240234, "learning_rate": 1.9354647317351187e-06, "loss": 0.4437, "step": 21300 }, { "epoch": 42.66, "grad_norm": 36.898136138916016, "learning_rate": 1.9230681491041425e-06, "loss": 0.4853, "step": 21330 }, { "epoch": 42.72, "grad_norm": 0.0009460307192057371, "learning_rate": 1.910701934548329e-06, "loss": 0.4529, "step": 21360 }, { "epoch": 42.78, "grad_norm": 56.36115264892578, "learning_rate": 1.8983662101172217e-06, "loss": 0.474, "step": 21390 }, { "epoch": 42.84, "grad_norm": 70.80406951904297, "learning_rate": 1.8860610975594384e-06, "loss": 0.3513, "step": 21420 }, { "epoch": 42.9, "grad_norm": 50.55048751831055, "learning_rate": 1.873786718321476e-06, "loss": 0.4377, "step": 21450 }, { "epoch": 42.96, "grad_norm": 78.81502532958984, "learning_rate": 1.8615431935464984e-06, "loss": 0.4596, "step": 21480 }, { "epoch": 43.0, "eval_loss": 0.328853964805603, "eval_map": 0.7747, "eval_map_50": 0.9605, "eval_map_75": 0.9013, "eval_map_chicken": 0.7714, "eval_map_duck": 0.7476, "eval_map_large": 0.8575, "eval_map_medium": 0.7696, "eval_map_plant": 0.805, "eval_map_small": 0.1866, "eval_mar_1": 0.2979, "eval_mar_10": 0.8149, "eval_mar_100": 0.822, "eval_mar_100_chicken": 0.8138, "eval_mar_100_duck": 0.7845, "eval_mar_100_plant": 0.8677, "eval_mar_large": 0.9159, "eval_mar_medium": 0.8221, "eval_mar_small": 0.4136, "eval_runtime": 7.6511, "eval_samples_per_second": 13.07, "eval_steps_per_second": 1.699, "step": 21500 }, { "epoch": 43.02, "grad_norm": 47.3128662109375, "learning_rate": 1.8493306440731557e-06, "loss": 0.5104, "step": 21510 }, { "epoch": 43.08, "grad_norm": 514.5338134765625, "learning_rate": 1.837149190434378e-06, "loss": 0.5045, "step": 21540 }, { "epoch": 43.14, "grad_norm": 44.962188720703125, "learning_rate": 1.824998952856198e-06, "loss": 0.4554, "step": 21570 }, { "epoch": 43.2, "grad_norm": 28.405282974243164, "learning_rate": 1.8128800512565514e-06, "loss": 0.4771, "step": 21600 }, { "epoch": 43.26, "grad_norm": 103.37633514404297, "learning_rate": 1.800792605244109e-06, "loss": 0.4298, "step": 21630 }, { "epoch": 43.32, "grad_norm": 113.0599136352539, "learning_rate": 1.7887367341170781e-06, "loss": 0.5014, "step": 21660 }, { "epoch": 43.38, "grad_norm": 53.75212478637695, "learning_rate": 1.7767125568620442e-06, "loss": 0.41, "step": 21690 }, { "epoch": 43.44, "grad_norm": 102.75830841064453, "learning_rate": 1.7647201921527802e-06, "loss": 0.3809, "step": 21720 }, { "epoch": 43.5, "grad_norm": 69.31130981445312, "learning_rate": 1.7527597583490825e-06, "loss": 0.3862, "step": 21750 }, { "epoch": 43.56, "grad_norm": 41.70110321044922, "learning_rate": 1.7408313734956074e-06, "loss": 0.3967, "step": 21780 }, { "epoch": 43.62, "grad_norm": 36.92282485961914, "learning_rate": 1.7289351553206952e-06, "loss": 0.5231, "step": 21810 }, { "epoch": 43.68, "grad_norm": 69.32868957519531, "learning_rate": 1.7170712212352187e-06, "loss": 0.4313, "step": 21840 }, { "epoch": 43.74, "grad_norm": 78.67727661132812, "learning_rate": 1.7052396883314154e-06, "loss": 0.5115, "step": 21870 }, { "epoch": 43.8, "grad_norm": 55.337825775146484, "learning_rate": 1.6934406733817417e-06, "loss": 0.4692, "step": 21900 }, { "epoch": 43.86, "grad_norm": 0.001163001754321158, "learning_rate": 1.6816742928377072e-06, "loss": 0.4461, "step": 21930 }, { "epoch": 43.92, "grad_norm": 82.20210266113281, "learning_rate": 1.6699406628287423e-06, "loss": 0.389, "step": 21960 }, { "epoch": 43.98, "grad_norm": 48.20108413696289, "learning_rate": 1.658239899161036e-06, "loss": 0.7776, "step": 21990 }, { "epoch": 44.0, "eval_loss": 0.3310554623603821, "eval_map": 0.7742, "eval_map_50": 0.9569, "eval_map_75": 0.9141, "eval_map_chicken": 0.7666, "eval_map_duck": 0.7559, "eval_map_large": 0.8538, "eval_map_medium": 0.7659, "eval_map_plant": 0.8001, "eval_map_small": 0.1645, "eval_mar_1": 0.3007, "eval_mar_10": 0.817, "eval_mar_100": 0.8244, "eval_mar_100_chicken": 0.8107, "eval_mar_100_duck": 0.8021, "eval_mar_100_plant": 0.8605, "eval_mar_large": 0.913, "eval_mar_medium": 0.8219, "eval_mar_small": 0.3985, "eval_runtime": 8.6388, "eval_samples_per_second": 11.576, "eval_steps_per_second": 1.505, "step": 22000 }, { "epoch": 44.04, "grad_norm": 117.72798919677734, "learning_rate": 1.6465721173164e-06, "loss": 0.4394, "step": 22020 }, { "epoch": 44.1, "grad_norm": 86.25477600097656, "learning_rate": 1.6349374324511347e-06, "loss": 0.4216, "step": 22050 }, { "epoch": 44.16, "grad_norm": 74.14421081542969, "learning_rate": 1.6233359593948777e-06, "loss": 0.4393, "step": 22080 }, { "epoch": 44.22, "grad_norm": 73.15435028076172, "learning_rate": 1.6117678126494895e-06, "loss": 0.4423, "step": 22110 }, { "epoch": 44.28, "grad_norm": 162.00294494628906, "learning_rate": 1.600233106387904e-06, "loss": 0.4415, "step": 22140 }, { "epoch": 44.34, "grad_norm": 935.9252319335938, "learning_rate": 1.5887319544530182e-06, "loss": 0.3541, "step": 22170 }, { "epoch": 44.4, "grad_norm": 33.84937286376953, "learning_rate": 1.5772644703565564e-06, "loss": 0.6435, "step": 22200 }, { "epoch": 44.46, "grad_norm": 33.91444396972656, "learning_rate": 1.5658307672779594e-06, "loss": 0.3256, "step": 22230 }, { "epoch": 44.52, "grad_norm": 39.06406021118164, "learning_rate": 1.554430958063259e-06, "loss": 0.4748, "step": 22260 }, { "epoch": 44.58, "grad_norm": 53.42619705200195, "learning_rate": 1.5430651552239684e-06, "loss": 0.4058, "step": 22290 }, { "epoch": 44.64, "grad_norm": 61.689735412597656, "learning_rate": 1.531733470935976e-06, "loss": 0.497, "step": 22320 }, { "epoch": 44.7, "grad_norm": 22.562637329101562, "learning_rate": 1.5204360170384286e-06, "loss": 0.4655, "step": 22350 }, { "epoch": 44.76, "grad_norm": 32.66790771484375, "learning_rate": 1.5091729050326376e-06, "loss": 0.4292, "step": 22380 }, { "epoch": 44.82, "grad_norm": 68.33869171142578, "learning_rate": 1.4979442460809684e-06, "loss": 0.4555, "step": 22410 }, { "epoch": 44.88, "grad_norm": 50.48727035522461, "learning_rate": 1.4867501510057548e-06, "loss": 0.4225, "step": 22440 }, { "epoch": 44.94, "grad_norm": 55.9785041809082, "learning_rate": 1.4755907302881927e-06, "loss": 0.4194, "step": 22470 }, { "epoch": 45.0, "grad_norm": 74.78853607177734, "learning_rate": 1.4644660940672628e-06, "loss": 0.4124, "step": 22500 }, { "epoch": 45.0, "eval_loss": 0.3233446180820465, "eval_map": 0.7787, "eval_map_50": 0.961, "eval_map_75": 0.9035, "eval_map_chicken": 0.7637, "eval_map_duck": 0.7701, "eval_map_large": 0.8577, "eval_map_medium": 0.7676, "eval_map_plant": 0.8024, "eval_map_small": 0.181, "eval_mar_1": 0.3006, "eval_mar_10": 0.8204, "eval_mar_100": 0.8286, "eval_mar_100_chicken": 0.8089, "eval_mar_100_duck": 0.8103, "eval_mar_100_plant": 0.8666, "eval_mar_large": 0.9172, "eval_mar_medium": 0.8261, "eval_mar_small": 0.4258, "eval_runtime": 9.6488, "eval_samples_per_second": 10.364, "eval_steps_per_second": 1.347, "step": 22500 }, { "epoch": 45.06, "grad_norm": 54.747344970703125, "learning_rate": 1.4533763521386319e-06, "loss": 0.4181, "step": 22530 }, { "epoch": 45.12, "grad_norm": 43.84959411621094, "learning_rate": 1.4423216139535735e-06, "loss": 0.4153, "step": 22560 }, { "epoch": 45.18, "grad_norm": 56.051822662353516, "learning_rate": 1.4313019886178942e-06, "loss": 0.4736, "step": 22590 }, { "epoch": 45.24, "grad_norm": 137.2510223388672, "learning_rate": 1.420317584890844e-06, "loss": 0.4981, "step": 22620 }, { "epoch": 45.3, "grad_norm": 41.77317428588867, "learning_rate": 1.4093685111840567e-06, "loss": 0.4199, "step": 22650 }, { "epoch": 45.36, "grad_norm": 88.05829620361328, "learning_rate": 1.3984548755604655e-06, "loss": 0.4288, "step": 22680 }, { "epoch": 45.42, "grad_norm": 775.5257568359375, "learning_rate": 1.3875767857332512e-06, "loss": 0.4614, "step": 22710 }, { "epoch": 45.48, "grad_norm": 62.29632568359375, "learning_rate": 1.3767343490647668e-06, "loss": 0.4513, "step": 22740 }, { "epoch": 45.54, "grad_norm": 71.67477416992188, "learning_rate": 1.3659276725654863e-06, "loss": 0.3555, "step": 22770 }, { "epoch": 45.6, "grad_norm": 79.21355438232422, "learning_rate": 1.3551568628929434e-06, "loss": 0.4624, "step": 22800 }, { "epoch": 45.66, "grad_norm": 43.81147384643555, "learning_rate": 1.3444220263506797e-06, "loss": 0.3945, "step": 22830 }, { "epoch": 45.72, "grad_norm": 76.984375, "learning_rate": 1.333723268887201e-06, "loss": 0.4932, "step": 22860 }, { "epoch": 45.78, "grad_norm": 53.61293029785156, "learning_rate": 1.3230606960949204e-06, "loss": 0.4821, "step": 22890 }, { "epoch": 45.84, "grad_norm": 59.51533508300781, "learning_rate": 1.312434413209131e-06, "loss": 0.4344, "step": 22920 }, { "epoch": 45.9, "grad_norm": 51.64701461791992, "learning_rate": 1.301844525106951e-06, "loss": 0.4352, "step": 22950 }, { "epoch": 45.96, "grad_norm": 42.07365036010742, "learning_rate": 1.2912911363063048e-06, "loss": 0.4843, "step": 22980 }, { "epoch": 46.0, "eval_loss": 0.31848251819610596, "eval_map": 0.7853, "eval_map_50": 0.9633, "eval_map_75": 0.9062, "eval_map_chicken": 0.7731, "eval_map_duck": 0.776, "eval_map_large": 0.863, "eval_map_medium": 0.7726, "eval_map_plant": 0.8068, "eval_map_small": 0.1836, "eval_mar_1": 0.3027, "eval_mar_10": 0.8252, "eval_mar_100": 0.8332, "eval_mar_100_chicken": 0.8169, "eval_mar_100_duck": 0.8144, "eval_mar_100_plant": 0.8683, "eval_mar_large": 0.9188, "eval_mar_medium": 0.8318, "eval_mar_small": 0.4206, "eval_runtime": 10.0039, "eval_samples_per_second": 9.996, "eval_steps_per_second": 1.299, "step": 23000 }, { "epoch": 46.02, "grad_norm": 51.64211654663086, "learning_rate": 1.2807743509648745e-06, "loss": 0.5116, "step": 23010 }, { "epoch": 46.08, "grad_norm": 86.34794616699219, "learning_rate": 1.2702942728790897e-06, "loss": 0.4109, "step": 23040 }, { "epoch": 46.14, "grad_norm": 38.06340026855469, "learning_rate": 1.2598510054830888e-06, "loss": 0.4065, "step": 23070 }, { "epoch": 46.2, "grad_norm": 65.07425689697266, "learning_rate": 1.2494446518477022e-06, "loss": 0.3309, "step": 23100 }, { "epoch": 46.26, "grad_norm": 38.2021484375, "learning_rate": 1.2390753146794438e-06, "loss": 0.3867, "step": 23130 }, { "epoch": 46.32, "grad_norm": 61.52031707763672, "learning_rate": 1.2287430963194807e-06, "loss": 0.4066, "step": 23160 }, { "epoch": 46.38, "grad_norm": 98.98115539550781, "learning_rate": 1.218448098742641e-06, "loss": 0.5063, "step": 23190 }, { "epoch": 46.44, "grad_norm": 31.327882766723633, "learning_rate": 1.2081904235563908e-06, "loss": 0.4518, "step": 23220 }, { "epoch": 46.5, "grad_norm": 44.24490737915039, "learning_rate": 1.1979701719998454e-06, "loss": 0.4052, "step": 23250 }, { "epoch": 46.56, "grad_norm": 42.57209396362305, "learning_rate": 1.18778744494276e-06, "loss": 0.6345, "step": 23280 }, { "epoch": 46.62, "grad_norm": 63.863224029541016, "learning_rate": 1.1776423428845423e-06, "loss": 0.3853, "step": 23310 }, { "epoch": 46.68, "grad_norm": 88.98026275634766, "learning_rate": 1.1675349659532514e-06, "loss": 0.4166, "step": 23340 }, { "epoch": 46.74, "grad_norm": 66.46283721923828, "learning_rate": 1.1574654139046171e-06, "loss": 0.4632, "step": 23370 }, { "epoch": 46.8, "grad_norm": 80.80040740966797, "learning_rate": 1.1474337861210543e-06, "loss": 0.3704, "step": 23400 }, { "epoch": 46.86, "grad_norm": 59.4736213684082, "learning_rate": 1.1374401816106778e-06, "loss": 0.4279, "step": 23430 }, { "epoch": 46.92, "grad_norm": 35.19805145263672, "learning_rate": 1.1274846990063314e-06, "loss": 0.4293, "step": 23460 }, { "epoch": 46.98, "grad_norm": 44.19129943847656, "learning_rate": 1.1175674365646067e-06, "loss": 0.4557, "step": 23490 }, { "epoch": 47.0, "eval_loss": 0.3219223916530609, "eval_map": 0.7764, "eval_map_50": 0.9638, "eval_map_75": 0.9018, "eval_map_chicken": 0.7585, "eval_map_duck": 0.7709, "eval_map_large": 0.8563, "eval_map_medium": 0.764, "eval_map_plant": 0.7996, "eval_map_small": 0.186, "eval_mar_1": 0.3, "eval_mar_10": 0.8149, "eval_mar_100": 0.8219, "eval_mar_100_chicken": 0.8013, "eval_mar_100_duck": 0.8052, "eval_mar_100_plant": 0.8591, "eval_mar_large": 0.9126, "eval_mar_medium": 0.8185, "eval_mar_small": 0.4062, "eval_runtime": 8.8688, "eval_samples_per_second": 11.276, "eval_steps_per_second": 1.466, "step": 23500 }, { "epoch": 47.04, "grad_norm": 64.04322814941406, "learning_rate": 1.1076884921648834e-06, "loss": 0.3406, "step": 23520 }, { "epoch": 47.1, "grad_norm": 59.95370864868164, "learning_rate": 1.097847963308351e-06, "loss": 0.441, "step": 23550 }, { "epoch": 47.16, "grad_norm": 53.64377975463867, "learning_rate": 1.0880459471170597e-06, "loss": 0.3632, "step": 23580 }, { "epoch": 47.22, "grad_norm": 162.77381896972656, "learning_rate": 1.0782825403329488e-06, "loss": 0.3698, "step": 23610 }, { "epoch": 47.28, "grad_norm": 46.4144401550293, "learning_rate": 1.0685578393169054e-06, "loss": 0.441, "step": 23640 }, { "epoch": 47.34, "grad_norm": 27.28321075439453, "learning_rate": 1.0588719400478004e-06, "loss": 0.394, "step": 23670 }, { "epoch": 47.4, "grad_norm": 24.538047790527344, "learning_rate": 1.049224938121548e-06, "loss": 0.4202, "step": 23700 }, { "epoch": 47.46, "grad_norm": 36.589134216308594, "learning_rate": 1.0396169287501652e-06, "loss": 0.359, "step": 23730 }, { "epoch": 47.52, "grad_norm": 61.19246292114258, "learning_rate": 1.0300480067608232e-06, "loss": 0.482, "step": 23760 }, { "epoch": 47.58, "grad_norm": 46.918331146240234, "learning_rate": 1.020518266594921e-06, "loss": 0.4207, "step": 23790 }, { "epoch": 47.64, "grad_norm": 45.71127700805664, "learning_rate": 1.0110278023071445e-06, "loss": 0.4052, "step": 23820 }, { "epoch": 47.7, "grad_norm": 62.494384765625, "learning_rate": 1.0015767075645472e-06, "loss": 0.3423, "step": 23850 }, { "epoch": 47.76, "grad_norm": 47.014408111572266, "learning_rate": 9.921650756456164e-07, "loss": 0.4164, "step": 23880 }, { "epoch": 47.82, "grad_norm": 58.932037353515625, "learning_rate": 9.82792999439362e-07, "loss": 0.4287, "step": 23910 }, { "epoch": 47.88, "grad_norm": 192.094970703125, "learning_rate": 9.734605714443906e-07, "loss": 0.5134, "step": 23940 }, { "epoch": 47.94, "grad_norm": 43.04744338989258, "learning_rate": 9.641678837679985e-07, "loss": 0.4416, "step": 23970 }, { "epoch": 48.0, "grad_norm": 47.490535736083984, "learning_rate": 9.549150281252633e-07, "loss": 0.4089, "step": 24000 }, { "epoch": 48.0, "eval_loss": 0.30931535363197327, "eval_map": 0.786, "eval_map_50": 0.9608, "eval_map_75": 0.9122, "eval_map_chicken": 0.779, "eval_map_duck": 0.7713, "eval_map_large": 0.8668, "eval_map_medium": 0.7758, "eval_map_plant": 0.8076, "eval_map_small": 0.1916, "eval_mar_1": 0.3035, "eval_mar_10": 0.8246, "eval_mar_100": 0.8332, "eval_mar_100_chicken": 0.8244, "eval_mar_100_duck": 0.8062, "eval_mar_100_plant": 0.8689, "eval_mar_large": 0.9218, "eval_mar_medium": 0.8302, "eval_mar_small": 0.4231, "eval_runtime": 10.4085, "eval_samples_per_second": 9.608, "eval_steps_per_second": 1.249, "step": 24000 }, { "epoch": 48.06, "grad_norm": 71.72051239013672, "learning_rate": 9.457020958381324e-07, "loss": 0.3888, "step": 24030 }, { "epoch": 48.12, "grad_norm": 73.30224609375, "learning_rate": 9.365291778345303e-07, "loss": 0.4508, "step": 24060 }, { "epoch": 48.18, "grad_norm": 25.481386184692383, "learning_rate": 9.273963646474527e-07, "loss": 0.4548, "step": 24090 }, { "epoch": 48.24, "grad_norm": 60.45578384399414, "learning_rate": 9.183037464140804e-07, "loss": 0.4534, "step": 24120 }, { "epoch": 48.3, "grad_norm": 62.31999206542969, "learning_rate": 9.09251412874882e-07, "loss": 0.3859, "step": 24150 }, { "epoch": 48.36, "grad_norm": 60.27473831176758, "learning_rate": 9.002394533727382e-07, "loss": 0.439, "step": 24180 }, { "epoch": 48.42, "grad_norm": 99.14818572998047, "learning_rate": 8.912679568520494e-07, "loss": 0.4276, "step": 24210 }, { "epoch": 48.48, "grad_norm": 74.67886352539062, "learning_rate": 8.823370118578628e-07, "loss": 0.4565, "step": 24240 }, { "epoch": 48.54, "grad_norm": 110.05522155761719, "learning_rate": 8.734467065350022e-07, "loss": 0.461, "step": 24270 }, { "epoch": 48.6, "grad_norm": 37.426849365234375, "learning_rate": 8.645971286271903e-07, "loss": 0.4753, "step": 24300 }, { "epoch": 48.66, "grad_norm": 34.440006256103516, "learning_rate": 8.557883654761906e-07, "loss": 0.4498, "step": 24330 }, { "epoch": 48.72, "grad_norm": 294.57684326171875, "learning_rate": 8.470205040209362e-07, "loss": 0.4734, "step": 24360 }, { "epoch": 48.78, "grad_norm": 64.2550277709961, "learning_rate": 8.382936307966838e-07, "loss": 0.4254, "step": 24390 }, { "epoch": 48.84, "grad_norm": 103.85020446777344, "learning_rate": 8.296078319341444e-07, "loss": 0.414, "step": 24420 }, { "epoch": 48.9, "grad_norm": 77.38172912597656, "learning_rate": 8.209631931586499e-07, "loss": 0.414, "step": 24450 }, { "epoch": 48.96, "grad_norm": 69.1627426147461, "learning_rate": 8.123597997892918e-07, "loss": 0.4806, "step": 24480 }, { "epoch": 49.0, "eval_loss": 0.312330961227417, "eval_map": 0.7827, "eval_map_50": 0.9619, "eval_map_75": 0.9067, "eval_map_chicken": 0.7773, "eval_map_duck": 0.767, "eval_map_large": 0.859, "eval_map_medium": 0.7722, "eval_map_plant": 0.8037, "eval_map_small": 0.2017, "eval_mar_1": 0.304, "eval_mar_10": 0.8218, "eval_mar_100": 0.8287, "eval_mar_100_chicken": 0.8213, "eval_mar_100_duck": 0.8, "eval_mar_100_plant": 0.8648, "eval_mar_large": 0.9167, "eval_mar_medium": 0.8259, "eval_mar_small": 0.4208, "eval_runtime": 7.7337, "eval_samples_per_second": 12.93, "eval_steps_per_second": 1.681, "step": 24500 }, { "epoch": 49.02, "grad_norm": 117.54744720458984, "learning_rate": 8.037977367380922e-07, "loss": 0.527, "step": 24510 }, { "epoch": 49.08, "grad_norm": 0.00019707001047208905, "learning_rate": 7.952770885091548e-07, "loss": 0.3993, "step": 24540 }, { "epoch": 49.14, "grad_norm": 118.28912353515625, "learning_rate": 7.867979391978398e-07, "loss": 0.3772, "step": 24570 }, { "epoch": 49.2, "grad_norm": 52.652713775634766, "learning_rate": 7.783603724899258e-07, "loss": 0.4568, "step": 24600 }, { "epoch": 49.26, "grad_norm": 36.391395568847656, "learning_rate": 7.699644716607896e-07, "loss": 0.4207, "step": 24630 }, { "epoch": 49.32, "grad_norm": 33.84572219848633, "learning_rate": 7.61610319574585e-07, "loss": 0.392, "step": 24660 }, { "epoch": 49.38, "grad_norm": 61.57717514038086, "learning_rate": 7.532979986834177e-07, "loss": 0.3855, "step": 24690 }, { "epoch": 49.44, "grad_norm": 56.8045768737793, "learning_rate": 7.450275910265415e-07, "loss": 0.3751, "step": 24720 }, { "epoch": 49.5, "grad_norm": 65.95047760009766, "learning_rate": 7.367991782295392e-07, "loss": 0.4598, "step": 24750 }, { "epoch": 49.56, "grad_norm": 30.823863983154297, "learning_rate": 7.286128415035249e-07, "loss": 0.4235, "step": 24780 }, { "epoch": 49.62, "grad_norm": 51.362327575683594, "learning_rate": 7.204686616443352e-07, "loss": 0.3999, "step": 24810 }, { "epoch": 49.68, "grad_norm": 63.72163772583008, "learning_rate": 7.123667190317396e-07, "loss": 0.3417, "step": 24840 }, { "epoch": 49.74, "grad_norm": 67.92794036865234, "learning_rate": 7.043070936286395e-07, "loss": 0.4345, "step": 24870 }, { "epoch": 49.8, "grad_norm": 55.80625915527344, "learning_rate": 6.962898649802824e-07, "loss": 0.4205, "step": 24900 }, { "epoch": 49.86, "grad_norm": 51.97507858276367, "learning_rate": 6.883151122134812e-07, "loss": 0.5707, "step": 24930 }, { "epoch": 49.92, "grad_norm": 46.65962219238281, "learning_rate": 6.803829140358237e-07, "loss": 0.5284, "step": 24960 }, { "epoch": 49.98, "grad_norm": 46.46683120727539, "learning_rate": 6.724933487349061e-07, "loss": 0.3855, "step": 24990 }, { "epoch": 50.0, "eval_loss": 0.3083952069282532, "eval_map": 0.7899, "eval_map_50": 0.9609, "eval_map_75": 0.9132, "eval_map_chicken": 0.7852, "eval_map_duck": 0.7774, "eval_map_large": 0.8598, "eval_map_medium": 0.7831, "eval_map_plant": 0.8072, "eval_map_small": 0.2023, "eval_mar_1": 0.3045, "eval_mar_10": 0.8273, "eval_mar_100": 0.834, "eval_mar_100_chicken": 0.8262, "eval_mar_100_duck": 0.8093, "eval_mar_100_plant": 0.8666, "eval_mar_large": 0.9167, "eval_mar_medium": 0.8339, "eval_mar_small": 0.4136, "eval_runtime": 7.6023, "eval_samples_per_second": 13.154, "eval_steps_per_second": 1.71, "step": 25000 }, { "epoch": 50.04, "grad_norm": 62.74702072143555, "learning_rate": 6.646464941775499e-07, "loss": 0.4473, "step": 25020 }, { "epoch": 50.1, "grad_norm": 53.569664001464844, "learning_rate": 6.568424278090446e-07, "loss": 0.4261, "step": 25050 }, { "epoch": 50.16, "grad_norm": 95.088623046875, "learning_rate": 6.490812266523716e-07, "loss": 0.3878, "step": 25080 }, { "epoch": 50.22, "grad_norm": 47.24351119995117, "learning_rate": 6.413629673074562e-07, "loss": 0.4425, "step": 25110 }, { "epoch": 50.28, "grad_norm": 65.52989196777344, "learning_rate": 6.336877259504004e-07, "loss": 0.4171, "step": 25140 }, { "epoch": 50.34, "grad_norm": 75.91929626464844, "learning_rate": 6.260555783327366e-07, "loss": 0.3836, "step": 25170 }, { "epoch": 50.4, "grad_norm": 42.9079704284668, "learning_rate": 6.184665997806832e-07, "loss": 0.4955, "step": 25200 }, { "epoch": 50.46, "grad_norm": 43.150394439697266, "learning_rate": 6.109208651943921e-07, "loss": 0.4127, "step": 25230 }, { "epoch": 50.52, "grad_norm": 104.1957015991211, "learning_rate": 6.034184490472195e-07, "loss": 0.3505, "step": 25260 }, { "epoch": 50.58, "grad_norm": 53.61925506591797, "learning_rate": 5.959594253849821e-07, "loss": 0.3925, "step": 25290 }, { "epoch": 50.64, "grad_norm": 56.76261901855469, "learning_rate": 5.885438678252342e-07, "loss": 0.4345, "step": 25320 }, { "epoch": 50.7, "grad_norm": 49.705116271972656, "learning_rate": 5.811718495565327e-07, "loss": 0.3939, "step": 25350 }, { "epoch": 50.76, "grad_norm": 94.89686584472656, "learning_rate": 5.738434433377244e-07, "loss": 0.4453, "step": 25380 }, { "epoch": 50.82, "grad_norm": 31.858102798461914, "learning_rate": 5.665587214972173e-07, "loss": 0.449, "step": 25410 }, { "epoch": 50.88, "grad_norm": 89.90313720703125, "learning_rate": 5.593177559322776e-07, "loss": 0.4253, "step": 25440 }, { "epoch": 50.94, "grad_norm": 62.72407150268555, "learning_rate": 5.521206181083111e-07, "loss": 0.7642, "step": 25470 }, { "epoch": 51.0, "grad_norm": 74.73518371582031, "learning_rate": 5.449673790581611e-07, "loss": 0.4488, "step": 25500 }, { "epoch": 51.0, "eval_loss": 0.3166915774345398, "eval_map": 0.7813, "eval_map_50": 0.9583, "eval_map_75": 0.91, "eval_map_chicken": 0.7722, "eval_map_duck": 0.7657, "eval_map_large": 0.8613, "eval_map_medium": 0.7717, "eval_map_plant": 0.8059, "eval_map_small": 0.1881, "eval_mar_1": 0.3019, "eval_mar_10": 0.8205, "eval_mar_100": 0.8279, "eval_mar_100_chicken": 0.8164, "eval_mar_100_duck": 0.8031, "eval_mar_100_plant": 0.8643, "eval_mar_large": 0.9172, "eval_mar_medium": 0.825, "eval_mar_small": 0.4169, "eval_runtime": 8.9867, "eval_samples_per_second": 11.128, "eval_steps_per_second": 1.447, "step": 25500 }, { "epoch": 51.06, "grad_norm": 51.67295837402344, "learning_rate": 5.378581093814112e-07, "loss": 0.4177, "step": 25530 }, { "epoch": 51.12, "grad_norm": 58.23004913330078, "learning_rate": 5.307928792436812e-07, "loss": 0.3512, "step": 25560 }, { "epoch": 51.18, "grad_norm": 64.30286407470703, "learning_rate": 5.237717583759421e-07, "loss": 0.454, "step": 25590 }, { "epoch": 51.24, "grad_norm": 75.98681640625, "learning_rate": 5.167948160738206e-07, "loss": 0.3918, "step": 25620 }, { "epoch": 51.3, "grad_norm": 37.77546691894531, "learning_rate": 5.098621211969224e-07, "loss": 0.433, "step": 25650 }, { "epoch": 51.36, "grad_norm": 55.263916015625, "learning_rate": 5.029737421681446e-07, "loss": 0.4565, "step": 25680 }, { "epoch": 51.42, "grad_norm": 108.228271484375, "learning_rate": 4.961297469730097e-07, "loss": 0.4164, "step": 25710 }, { "epoch": 51.48, "grad_norm": 48.70679473876953, "learning_rate": 4.893302031589864e-07, "loss": 0.399, "step": 25740 }, { "epoch": 51.54, "grad_norm": 74.1161880493164, "learning_rate": 4.825751778348259e-07, "loss": 0.4993, "step": 25770 }, { "epoch": 51.6, "grad_norm": 118.7088851928711, "learning_rate": 4.758647376699033e-07, "loss": 0.42, "step": 25800 }, { "epoch": 51.66, "grad_norm": 45.4176025390625, "learning_rate": 4.691989488935511e-07, "loss": 0.4406, "step": 25830 }, { "epoch": 51.72, "grad_norm": 31.40489387512207, "learning_rate": 4.625778772944156e-07, "loss": 0.4115, "step": 25860 }, { "epoch": 51.78, "grad_norm": 59.26774978637695, "learning_rate": 4.5600158821979933e-07, "loss": 0.3581, "step": 25890 }, { "epoch": 51.84, "grad_norm": 64.11298370361328, "learning_rate": 4.494701465750217e-07, "loss": 0.3405, "step": 25920 }, { "epoch": 51.9, "grad_norm": 33.244205474853516, "learning_rate": 4.4298361682277355e-07, "loss": 0.4981, "step": 25950 }, { "epoch": 51.96, "grad_norm": 40.4251708984375, "learning_rate": 4.3654206298248625e-07, "loss": 0.4071, "step": 25980 }, { "epoch": 52.0, "eval_loss": 0.310726135969162, "eval_map": 0.7845, "eval_map_50": 0.9616, "eval_map_75": 0.9072, "eval_map_chicken": 0.7788, "eval_map_duck": 0.7649, "eval_map_large": 0.8633, "eval_map_medium": 0.7753, "eval_map_plant": 0.8098, "eval_map_small": 0.2001, "eval_mar_1": 0.3043, "eval_mar_10": 0.8252, "eval_mar_100": 0.8316, "eval_mar_100_chicken": 0.8227, "eval_mar_100_duck": 0.8041, "eval_mar_100_plant": 0.868, "eval_mar_large": 0.9188, "eval_mar_medium": 0.8304, "eval_mar_small": 0.4189, "eval_runtime": 9.8155, "eval_samples_per_second": 10.188, "eval_steps_per_second": 1.324, "step": 26000 }, { "epoch": 52.02, "grad_norm": 36.111541748046875, "learning_rate": 4.301455486296946e-07, "loss": 0.4528, "step": 26010 }, { "epoch": 52.08, "grad_norm": 24.53832244873047, "learning_rate": 4.237941368954124e-07, "loss": 0.4098, "step": 26040 }, { "epoch": 52.14, "grad_norm": 41.893184661865234, "learning_rate": 4.1748789046551055e-07, "loss": 0.4138, "step": 26070 }, { "epoch": 52.2, "grad_norm": 0.00020227921777404845, "learning_rate": 4.112268715800943e-07, "loss": 0.4509, "step": 26100 }, { "epoch": 52.26, "grad_norm": 45.58656311035156, "learning_rate": 4.0501114203289395e-07, "loss": 0.3889, "step": 26130 }, { "epoch": 52.32, "grad_norm": 49.081417083740234, "learning_rate": 3.9884076317064813e-07, "loss": 0.4222, "step": 26160 }, { "epoch": 52.38, "grad_norm": 54.94212341308594, "learning_rate": 3.9271579589250817e-07, "loss": 0.6165, "step": 26190 }, { "epoch": 52.44, "grad_norm": 42.73911666870117, "learning_rate": 3.866363006494256e-07, "loss": 0.416, "step": 26220 }, { "epoch": 52.5, "grad_norm": 40.6148796081543, "learning_rate": 3.8060233744356634e-07, "loss": 0.4284, "step": 26250 }, { "epoch": 52.56, "grad_norm": 64.17967224121094, "learning_rate": 3.7461396582771035e-07, "loss": 0.4434, "step": 26280 }, { "epoch": 52.62, "grad_norm": 41.359806060791016, "learning_rate": 3.6867124490466697e-07, "loss": 0.4713, "step": 26310 }, { "epoch": 52.68, "grad_norm": 44.165130615234375, "learning_rate": 3.627742333266937e-07, "loss": 0.3907, "step": 26340 }, { "epoch": 52.74, "grad_norm": 975.5628662109375, "learning_rate": 3.569229892949133e-07, "loss": 0.5054, "step": 26370 }, { "epoch": 52.8, "grad_norm": 69.44132995605469, "learning_rate": 3.511175705587433e-07, "loss": 0.3996, "step": 26400 }, { "epoch": 52.86, "grad_norm": 89.67790222167969, "learning_rate": 3.4535803441532125e-07, "loss": 0.4448, "step": 26430 }, { "epoch": 52.92, "grad_norm": 37.850955963134766, "learning_rate": 3.396444377089453e-07, "loss": 0.4414, "step": 26460 }, { "epoch": 52.98, "grad_norm": 0.0009475430124439299, "learning_rate": 3.3397683683050685e-07, "loss": 0.3575, "step": 26490 }, { "epoch": 53.0, "eval_loss": 0.3146113157272339, "eval_map": 0.7805, "eval_map_50": 0.9602, "eval_map_75": 0.9045, "eval_map_chicken": 0.7706, "eval_map_duck": 0.7659, "eval_map_large": 0.8603, "eval_map_medium": 0.7678, "eval_map_plant": 0.8048, "eval_map_small": 0.1977, "eval_mar_1": 0.303, "eval_mar_10": 0.8226, "eval_mar_100": 0.829, "eval_mar_100_chicken": 0.8164, "eval_mar_100_duck": 0.8052, "eval_mar_100_plant": 0.8654, "eval_mar_large": 0.9184, "eval_mar_medium": 0.8268, "eval_mar_small": 0.4097, "eval_runtime": 8.3866, "eval_samples_per_second": 11.924, "eval_steps_per_second": 1.55, "step": 26500 }, { "epoch": 53.04, "grad_norm": 56.910865783691406, "learning_rate": 3.283552877169399e-07, "loss": 0.4576, "step": 26520 }, { "epoch": 53.1, "grad_norm": 97.69294738769531, "learning_rate": 3.227798458506637e-07, "loss": 0.4016, "step": 26550 }, { "epoch": 53.16, "grad_norm": 0.0002605436311569065, "learning_rate": 3.172505662590386e-07, "loss": 0.3952, "step": 26580 }, { "epoch": 53.22, "grad_norm": 72.2193603515625, "learning_rate": 3.1176750351382235e-07, "loss": 0.4198, "step": 26610 }, { "epoch": 53.28, "grad_norm": 163.66867065429688, "learning_rate": 3.0633071173062966e-07, "loss": 0.4007, "step": 26640 }, { "epoch": 53.34, "grad_norm": 56.278648376464844, "learning_rate": 3.0094024456840176e-07, "loss": 0.4279, "step": 26670 }, { "epoch": 53.4, "grad_norm": 66.77481842041016, "learning_rate": 2.9559615522887275e-07, "loss": 0.3927, "step": 26700 }, { "epoch": 53.46, "grad_norm": 66.41755676269531, "learning_rate": 2.9029849645604735e-07, "loss": 0.5139, "step": 26730 }, { "epoch": 53.52, "grad_norm": 35.4869499206543, "learning_rate": 2.850473205356774e-07, "loss": 0.4063, "step": 26760 }, { "epoch": 53.58, "grad_norm": 46.05600357055664, "learning_rate": 2.798426792947517e-07, "loss": 0.4444, "step": 26790 }, { "epoch": 53.64, "grad_norm": 52.565391540527344, "learning_rate": 2.746846241009765e-07, "loss": 0.4967, "step": 26820 }, { "epoch": 53.7, "grad_norm": 31.22282600402832, "learning_rate": 2.6957320586227354e-07, "loss": 0.3859, "step": 26850 }, { "epoch": 53.76, "grad_norm": 53.0260009765625, "learning_rate": 2.6450847502627883e-07, "loss": 0.746, "step": 26880 }, { "epoch": 53.82, "grad_norm": 239.99317932128906, "learning_rate": 2.594904815798399e-07, "loss": 0.4772, "step": 26910 }, { "epoch": 53.88, "grad_norm": 40.762630462646484, "learning_rate": 2.5451927504852757e-07, "loss": 0.4234, "step": 26940 }, { "epoch": 53.94, "grad_norm": 371.734619140625, "learning_rate": 2.49594904496141e-07, "loss": 0.3871, "step": 26970 }, { "epoch": 54.0, "grad_norm": 67.45365142822266, "learning_rate": 2.447174185242324e-07, "loss": 0.4115, "step": 27000 }, { "epoch": 54.0, "eval_loss": 0.3152923583984375, "eval_map": 0.7826, "eval_map_50": 0.9599, "eval_map_75": 0.9092, "eval_map_chicken": 0.7751, "eval_map_duck": 0.7672, "eval_map_large": 0.8613, "eval_map_medium": 0.7711, "eval_map_plant": 0.8055, "eval_map_small": 0.1982, "eval_mar_1": 0.304, "eval_mar_10": 0.8238, "eval_mar_100": 0.8295, "eval_mar_100_chicken": 0.8182, "eval_mar_100_duck": 0.8062, "eval_mar_100_plant": 0.864, "eval_mar_large": 0.9176, "eval_mar_medium": 0.8259, "eval_mar_small": 0.4189, "eval_runtime": 8.5466, "eval_samples_per_second": 11.701, "eval_steps_per_second": 1.521, "step": 27000 }, { "epoch": 54.06, "grad_norm": 71.0616226196289, "learning_rate": 2.3988686527161686e-07, "loss": 0.3796, "step": 27030 }, { "epoch": 54.12, "grad_norm": 56.90493392944336, "learning_rate": 2.351032924139063e-07, "loss": 0.5351, "step": 27060 }, { "epoch": 54.18, "grad_norm": 0.0001761562452884391, "learning_rate": 2.3036674716303277e-07, "loss": 0.468, "step": 27090 }, { "epoch": 54.24, "grad_norm": 34.41741180419922, "learning_rate": 2.2567727626678527e-07, "loss": 0.3656, "step": 27120 }, { "epoch": 54.3, "grad_norm": 37.34014129638672, "learning_rate": 2.210349260083494e-07, "loss": 0.4419, "step": 27150 }, { "epoch": 54.36, "grad_norm": 51.44719696044922, "learning_rate": 2.1643974220584729e-07, "loss": 0.4765, "step": 27180 }, { "epoch": 54.42, "grad_norm": 48.51471710205078, "learning_rate": 2.1189177021188888e-07, "loss": 0.4079, "step": 27210 }, { "epoch": 54.48, "grad_norm": 84.40725708007812, "learning_rate": 2.0739105491312028e-07, "loss": 0.4294, "step": 27240 }, { "epoch": 54.54, "grad_norm": 89.61735534667969, "learning_rate": 2.0293764072978618e-07, "loss": 0.408, "step": 27270 }, { "epoch": 54.6, "grad_norm": 95.17532348632812, "learning_rate": 1.9853157161528468e-07, "loss": 0.4733, "step": 27300 }, { "epoch": 54.66, "grad_norm": 25.492191314697266, "learning_rate": 1.9417289105574054e-07, "loss": 0.5093, "step": 27330 }, { "epoch": 54.72, "grad_norm": 42.83619689941406, "learning_rate": 1.8986164206957037e-07, "loss": 0.3762, "step": 27360 }, { "epoch": 54.78, "grad_norm": 0.0002613811520859599, "learning_rate": 1.8559786720706185e-07, "loss": 0.3521, "step": 27390 }, { "epoch": 54.84, "grad_norm": 135.53797912597656, "learning_rate": 1.8138160854995145e-07, "loss": 0.4371, "step": 27420 }, { "epoch": 54.9, "grad_norm": 29.237924575805664, "learning_rate": 1.7721290771100964e-07, "loss": 0.4452, "step": 27450 }, { "epoch": 54.96, "grad_norm": 57.26081848144531, "learning_rate": 1.7309180583363062e-07, "loss": 0.4093, "step": 27480 }, { "epoch": 55.0, "eval_loss": 0.3138941526412964, "eval_map": 0.7858, "eval_map_50": 0.962, "eval_map_75": 0.9118, "eval_map_chicken": 0.7786, "eval_map_duck": 0.7721, "eval_map_large": 0.8624, "eval_map_medium": 0.7739, "eval_map_plant": 0.8067, "eval_map_small": 0.197, "eval_mar_1": 0.3053, "eval_mar_10": 0.8263, "eval_mar_100": 0.8334, "eval_mar_100_chicken": 0.8209, "eval_mar_100_duck": 0.8144, "eval_mar_100_plant": 0.8648, "eval_mar_large": 0.9176, "eval_mar_medium": 0.8302, "eval_mar_small": 0.4206, "eval_runtime": 9.4558, "eval_samples_per_second": 10.576, "eval_steps_per_second": 1.375, "step": 27500 }, { "epoch": 55.02, "grad_norm": 33.90010070800781, "learning_rate": 1.690183435914261e-07, "loss": 0.465, "step": 27510 }, { "epoch": 55.08, "grad_norm": 67.94554138183594, "learning_rate": 1.6499256118782503e-07, "loss": 0.5033, "step": 27540 }, { "epoch": 55.14, "grad_norm": 0.0005611380329355597, "learning_rate": 1.6101449835567273e-07, "loss": 0.3897, "step": 27570 }, { "epoch": 55.2, "grad_norm": 0.00026392904692329466, "learning_rate": 1.5708419435684463e-07, "loss": 0.3846, "step": 27600 }, { "epoch": 55.26, "grad_norm": 61.53114318847656, "learning_rate": 1.532016879818532e-07, "loss": 0.4035, "step": 27630 }, { "epoch": 55.32, "grad_norm": 150.73643493652344, "learning_rate": 1.4936701754947104e-07, "loss": 0.4231, "step": 27660 }, { "epoch": 55.38, "grad_norm": 24.582101821899414, "learning_rate": 1.4558022090634504e-07, "loss": 0.4717, "step": 27690 }, { "epoch": 55.44, "grad_norm": 58.68578338623047, "learning_rate": 1.4184133542663014e-07, "loss": 0.4265, "step": 27720 }, { "epoch": 55.5, "grad_norm": 64.78734588623047, "learning_rate": 1.3815039801161723e-07, "loss": 0.4672, "step": 27750 }, { "epoch": 55.56, "grad_norm": 51.77420425415039, "learning_rate": 1.3450744508936687e-07, "loss": 0.3594, "step": 27780 }, { "epoch": 55.62, "grad_norm": 94.75901794433594, "learning_rate": 1.3091251261435568e-07, "loss": 0.4564, "step": 27810 }, { "epoch": 55.68, "grad_norm": 45.45354461669922, "learning_rate": 1.2736563606711384e-07, "loss": 0.3541, "step": 27840 }, { "epoch": 55.74, "grad_norm": 0.00023164466256275773, "learning_rate": 1.2386685045388313e-07, "loss": 0.3298, "step": 27870 }, { "epoch": 55.8, "grad_norm": 71.77073669433594, "learning_rate": 1.2041619030626283e-07, "loss": 0.478, "step": 27900 }, { "epoch": 55.86, "grad_norm": 0.00023956623044796288, "learning_rate": 1.1701368968087711e-07, "loss": 0.6918, "step": 27930 }, { "epoch": 55.92, "grad_norm": 57.571556091308594, "learning_rate": 1.136593821590326e-07, "loss": 0.3934, "step": 27960 }, { "epoch": 55.98, "grad_norm": 203.47390747070312, "learning_rate": 1.1035330084639084e-07, "loss": 0.3959, "step": 27990 }, { "epoch": 56.0, "eval_loss": 0.3132701516151428, "eval_map": 0.7859, "eval_map_50": 0.963, "eval_map_75": 0.9055, "eval_map_chicken": 0.779, "eval_map_duck": 0.7738, "eval_map_large": 0.8576, "eval_map_medium": 0.7755, "eval_map_plant": 0.805, "eval_map_small": 0.1948, "eval_mar_1": 0.3041, "eval_mar_10": 0.8251, "eval_mar_100": 0.832, "eval_mar_100_chicken": 0.8213, "eval_mar_100_duck": 0.8113, "eval_mar_100_plant": 0.8634, "eval_mar_large": 0.9159, "eval_mar_medium": 0.8303, "eval_mar_small": 0.4106, "eval_runtime": 9.9411, "eval_samples_per_second": 10.059, "eval_steps_per_second": 1.308, "step": 28000 }, { "epoch": 56.04, "grad_norm": 56.145591735839844, "learning_rate": 1.0709547837263967e-07, "loss": 0.4178, "step": 28020 }, { "epoch": 56.1, "grad_norm": 43.19185256958008, "learning_rate": 1.038859468911707e-07, "loss": 0.3856, "step": 28050 }, { "epoch": 56.16, "grad_norm": 51.342472076416016, "learning_rate": 1.007247380787657e-07, "loss": 0.4049, "step": 28080 }, { "epoch": 56.22, "grad_norm": 68.65950775146484, "learning_rate": 9.761188313527792e-08, "loss": 0.4329, "step": 28110 }, { "epoch": 56.28, "grad_norm": 55.531646728515625, "learning_rate": 9.454741278333013e-08, "loss": 0.5033, "step": 28140 }, { "epoch": 56.34, "grad_norm": 28.69962501525879, "learning_rate": 9.153135726800599e-08, "loss": 0.4451, "step": 28170 }, { "epoch": 56.4, "grad_norm": 32.82883071899414, "learning_rate": 8.856374635655696e-08, "loss": 0.4919, "step": 28200 }, { "epoch": 56.46, "grad_norm": 72.49078369140625, "learning_rate": 8.564460933810414e-08, "loss": 0.4165, "step": 28230 }, { "epoch": 56.52, "grad_norm": 55.18204879760742, "learning_rate": 8.277397502335194e-08, "loss": 0.5009, "step": 28260 }, { "epoch": 56.58, "grad_norm": 76.23528289794922, "learning_rate": 7.995187174430152e-08, "loss": 0.4304, "step": 28290 }, { "epoch": 56.64, "grad_norm": 96.23405456542969, "learning_rate": 7.717832735397335e-08, "loss": 0.3883, "step": 28320 }, { "epoch": 56.7, "grad_norm": 273.0704040527344, "learning_rate": 7.445336922613067e-08, "loss": 0.3922, "step": 28350 }, { "epoch": 56.76, "grad_norm": 110.94636535644531, "learning_rate": 7.177702425500977e-08, "loss": 0.4208, "step": 28380 }, { "epoch": 56.82, "grad_norm": 46.26091766357422, "learning_rate": 6.914931885505626e-08, "loss": 0.4173, "step": 28410 }, { "epoch": 56.88, "grad_norm": 55.077274322509766, "learning_rate": 6.657027896065982e-08, "loss": 0.4485, "step": 28440 }, { "epoch": 56.94, "grad_norm": 88.4512710571289, "learning_rate": 6.403993002590425e-08, "loss": 0.3949, "step": 28470 }, { "epoch": 57.0, "grad_norm": 61.60903549194336, "learning_rate": 6.15582970243117e-08, "loss": 0.4241, "step": 28500 }, { "epoch": 57.0, "eval_loss": 0.3121153712272644, "eval_map": 0.7813, "eval_map_50": 0.9632, "eval_map_75": 0.909, "eval_map_chicken": 0.7763, "eval_map_duck": 0.7645, "eval_map_large": 0.8552, "eval_map_medium": 0.7712, "eval_map_plant": 0.8032, "eval_map_small": 0.1966, "eval_mar_1": 0.3025, "eval_mar_10": 0.8227, "eval_mar_100": 0.8296, "eval_mar_100_chicken": 0.8218, "eval_mar_100_duck": 0.8041, "eval_mar_100_plant": 0.8628, "eval_mar_large": 0.9151, "eval_mar_medium": 0.8276, "eval_mar_small": 0.4127, "eval_runtime": 9.793, "eval_samples_per_second": 10.211, "eval_steps_per_second": 1.327, "step": 28500 }, { "epoch": 57.06, "grad_norm": 37.88300323486328, "learning_rate": 5.9125404448597825e-08, "loss": 0.5118, "step": 28530 }, { "epoch": 57.12, "grad_norm": 60.340702056884766, "learning_rate": 5.674127631043025e-08, "loss": 0.3427, "step": 28560 }, { "epoch": 57.18, "grad_norm": 32.38589096069336, "learning_rate": 5.440593614019107e-08, "loss": 0.3258, "step": 28590 }, { "epoch": 57.24, "grad_norm": 78.19099426269531, "learning_rate": 5.2119406986745336e-08, "loss": 0.4048, "step": 28620 }, { "epoch": 57.3, "grad_norm": 99.96614074707031, "learning_rate": 4.988171141721232e-08, "loss": 0.4283, "step": 28650 }, { "epoch": 57.36, "grad_norm": 55.952415466308594, "learning_rate": 4.769287151674407e-08, "loss": 0.4418, "step": 28680 }, { "epoch": 57.42, "grad_norm": 49.264957427978516, "learning_rate": 4.5552908888306654e-08, "loss": 0.468, "step": 28710 }, { "epoch": 57.48, "grad_norm": 45.721012115478516, "learning_rate": 4.346184465246761e-08, "loss": 0.4489, "step": 28740 }, { "epoch": 57.54, "grad_norm": 80.1391372680664, "learning_rate": 4.1419699447186045e-08, "loss": 0.4206, "step": 28770 }, { "epoch": 57.6, "grad_norm": 78.03536987304688, "learning_rate": 3.9426493427611177e-08, "loss": 0.424, "step": 28800 }, { "epoch": 57.66, "grad_norm": 45.05917739868164, "learning_rate": 3.748224626588137e-08, "loss": 0.7334, "step": 28830 }, { "epoch": 57.72, "grad_norm": 48.910926818847656, "learning_rate": 3.558697715093207e-08, "loss": 0.4137, "step": 28860 }, { "epoch": 57.78, "grad_norm": 40.60625076293945, "learning_rate": 3.374070478830316e-08, "loss": 0.4459, "step": 28890 }, { "epoch": 57.84, "grad_norm": 32.859458923339844, "learning_rate": 3.194344739995803e-08, "loss": 0.5907, "step": 28920 }, { "epoch": 57.9, "grad_norm": 69.47660064697266, "learning_rate": 3.019522272410202e-08, "loss": 0.4191, "step": 28950 }, { "epoch": 57.96, "grad_norm": 88.09815979003906, "learning_rate": 2.8496048015005385e-08, "loss": 0.4179, "step": 28980 }, { "epoch": 58.0, "eval_loss": 0.3132533133029938, "eval_map": 0.7814, "eval_map_50": 0.9631, "eval_map_75": 0.9056, "eval_map_chicken": 0.7765, "eval_map_duck": 0.7643, "eval_map_large": 0.8562, "eval_map_medium": 0.7709, "eval_map_plant": 0.8034, "eval_map_small": 0.1988, "eval_mar_1": 0.3033, "eval_mar_10": 0.8229, "eval_mar_100": 0.8299, "eval_mar_100_chicken": 0.8222, "eval_mar_100_duck": 0.8041, "eval_mar_100_plant": 0.8634, "eval_mar_large": 0.9151, "eval_mar_medium": 0.8279, "eval_mar_small": 0.4189, "eval_runtime": 10.7895, "eval_samples_per_second": 9.268, "eval_steps_per_second": 1.205, "step": 29000 }, { "epoch": 58.02, "grad_norm": 76.57215881347656, "learning_rate": 2.684594004283836e-08, "loss": 0.4257, "step": 29010 }, { "epoch": 58.08, "grad_norm": 53.15199279785156, "learning_rate": 2.5244915093499134e-08, "loss": 0.3281, "step": 29040 }, { "epoch": 58.14, "grad_norm": 59.760135650634766, "learning_rate": 2.3692988968458398e-08, "loss": 0.4968, "step": 29070 }, { "epoch": 58.2, "grad_norm": 41.6032829284668, "learning_rate": 2.219017698460002e-08, "loss": 0.4428, "step": 29100 }, { "epoch": 58.26, "grad_norm": 61.897823333740234, "learning_rate": 2.0736493974071736e-08, "loss": 0.3471, "step": 29130 }, { "epoch": 58.32, "grad_norm": 38.736412048339844, "learning_rate": 1.9331954284137476e-08, "loss": 0.3535, "step": 29160 }, { "epoch": 58.38, "grad_norm": 36.1994514465332, "learning_rate": 1.7976571777038044e-08, "loss": 0.4151, "step": 29190 }, { "epoch": 58.44, "grad_norm": 57.233985900878906, "learning_rate": 1.6670359829850657e-08, "loss": 0.4797, "step": 29220 }, { "epoch": 58.5, "grad_norm": 33.711143493652344, "learning_rate": 1.541333133436018e-08, "loss": 0.4381, "step": 29250 }, { "epoch": 58.56, "grad_norm": 91.0936508178711, "learning_rate": 1.4205498696930332e-08, "loss": 0.3903, "step": 29280 }, { "epoch": 58.62, "grad_norm": 60.89310073852539, "learning_rate": 1.3046873838381546e-08, "loss": 0.4113, "step": 29310 }, { "epoch": 58.68, "grad_norm": 53.45530700683594, "learning_rate": 1.1937468193873869e-08, "loss": 0.4147, "step": 29340 }, { "epoch": 58.74, "grad_norm": 34.17583084106445, "learning_rate": 1.0877292712792586e-08, "loss": 0.4396, "step": 29370 }, { "epoch": 58.8, "grad_norm": 118.3806381225586, "learning_rate": 9.866357858642206e-09, "loss": 0.4023, "step": 29400 }, { "epoch": 58.86, "grad_norm": 48.14165115356445, "learning_rate": 8.904673608940983e-09, "loss": 0.3901, "step": 29430 }, { "epoch": 58.92, "grad_norm": 83.26249694824219, "learning_rate": 7.992249455124889e-09, "loss": 0.3831, "step": 29460 }, { "epoch": 58.98, "grad_norm": 87.95204162597656, "learning_rate": 7.129094402451575e-09, "loss": 0.4207, "step": 29490 }, { "epoch": 59.0, "eval_loss": 0.3134159445762634, "eval_map": 0.7826, "eval_map_50": 0.9628, "eval_map_75": 0.9089, "eval_map_chicken": 0.777, "eval_map_duck": 0.7652, "eval_map_large": 0.8571, "eval_map_medium": 0.7717, "eval_map_plant": 0.8055, "eval_map_small": 0.205, "eval_mar_1": 0.3034, "eval_mar_10": 0.8238, "eval_mar_100": 0.8307, "eval_mar_100_chicken": 0.8222, "eval_mar_100_duck": 0.8052, "eval_mar_100_plant": 0.8648, "eval_mar_large": 0.9155, "eval_mar_medium": 0.8287, "eval_mar_small": 0.4314, "eval_runtime": 7.6364, "eval_samples_per_second": 13.095, "eval_steps_per_second": 1.702, "step": 29500 }, { "epoch": 59.04, "grad_norm": 24.864288330078125, "learning_rate": 6.315216969912663e-09, "loss": 0.4057, "step": 29520 }, { "epoch": 59.1, "grad_norm": 441.1263427734375, "learning_rate": 5.5506251901504825e-09, "loss": 0.3666, "step": 29550 }, { "epoch": 59.16, "grad_norm": 54.56084442138672, "learning_rate": 4.835326609376468e-09, "loss": 0.4236, "step": 29580 }, { "epoch": 59.22, "grad_norm": 39.52933883666992, "learning_rate": 4.169328287299545e-09, "loss": 0.4865, "step": 29610 }, { "epoch": 59.28, "grad_norm": 35.664363861083984, "learning_rate": 3.5526367970539765e-09, "loss": 0.3728, "step": 29640 }, { "epoch": 59.34, "grad_norm": 25.64591407775879, "learning_rate": 2.9852582251355124e-09, "loss": 0.4032, "step": 29670 }, { "epoch": 59.4, "grad_norm": 71.70014953613281, "learning_rate": 2.4671981713420003e-09, "loss": 0.4459, "step": 29700 }, { "epoch": 59.46, "grad_norm": 36.34132385253906, "learning_rate": 1.9984617487173174e-09, "loss": 0.4733, "step": 29730 }, { "epoch": 59.52, "grad_norm": 57.26972579956055, "learning_rate": 1.5790535835003006e-09, "loss": 0.4725, "step": 29760 }, { "epoch": 59.58, "grad_norm": 47.11916732788086, "learning_rate": 1.2089778150797816e-09, "loss": 0.4069, "step": 29790 }, { "epoch": 59.64, "grad_norm": 43.69301223754883, "learning_rate": 8.88238095955174e-10, "loss": 0.4442, "step": 29820 }, { "epoch": 59.7, "grad_norm": 38.88160705566406, "learning_rate": 6.168375916970615e-10, "loss": 0.4252, "step": 29850 }, { "epoch": 59.76, "grad_norm": 73.31425476074219, "learning_rate": 3.9477898091944135e-10, "loss": 0.4062, "step": 29880 }, { "epoch": 59.82, "grad_norm": 58.11071014404297, "learning_rate": 2.2206445525085886e-10, "loss": 0.4758, "step": 29910 }, { "epoch": 59.88, "grad_norm": 91.56449127197266, "learning_rate": 9.869571931442334e-11, "loss": 0.4255, "step": 29940 }, { "epoch": 59.94, "grad_norm": 56.301055908203125, "learning_rate": 2.467399070893439e-11, "loss": 0.3955, "step": 29970 }, { "epoch": 60.0, "grad_norm": 110.27831268310547, "learning_rate": 0.0, "loss": 0.4427, "step": 30000 }, { "epoch": 60.0, "eval_loss": 0.3133445680141449, "eval_map": 0.7825, "eval_map_50": 0.9628, "eval_map_75": 0.9089, "eval_map_chicken": 0.7771, "eval_map_duck": 0.7652, "eval_map_large": 0.8571, "eval_map_medium": 0.7718, "eval_map_plant": 0.805, "eval_map_small": 0.1977, "eval_mar_1": 0.3036, "eval_mar_10": 0.8238, "eval_mar_100": 0.8308, "eval_mar_100_chicken": 0.8227, "eval_mar_100_duck": 0.8052, "eval_mar_100_plant": 0.8646, "eval_mar_large": 0.9155, "eval_mar_medium": 0.8288, "eval_mar_small": 0.4273, "eval_runtime": 10.9608, "eval_samples_per_second": 9.123, "eval_steps_per_second": 1.186, "step": 30000 } ], "logging_steps": 30, "max_steps": 30000, "num_input_tokens_seen": 0, "num_train_epochs": 60, "save_steps": 10, "stateful_callbacks": { "TrainerControl": { "args": { "should_epoch_stop": false, "should_evaluate": false, "should_log": false, "should_save": true, "should_training_stop": true }, "attributes": {} } }, "total_flos": 1.0320693387264e+19, "train_batch_size": 2, "trial_name": null, "trial_params": null }