| { | |
| "best_metric": 0.28454330563545227, | |
| "best_model_checkpoint": "joe611/chickens-60-epoch-1000-images-aug/checkpoint-28500", | |
| "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": 51.12995529174805, | |
| "learning_rate": 9.999975326009292e-06, | |
| "loss": 1.7984, | |
| "step": 30 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "grad_norm": 264.590087890625, | |
| "learning_rate": 9.999901304280686e-06, | |
| "loss": 1.9316, | |
| "step": 60 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "grad_norm": 74.54895782470703, | |
| "learning_rate": 9.99977793554475e-06, | |
| "loss": 1.5595, | |
| "step": 90 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "grad_norm": 182.5364532470703, | |
| "learning_rate": 9.999605221019082e-06, | |
| "loss": 1.5981, | |
| "step": 120 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "grad_norm": 90.40423583984375, | |
| "learning_rate": 9.999383162408303e-06, | |
| "loss": 1.5708, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "grad_norm": 84.81971740722656, | |
| "learning_rate": 9.999111761904046e-06, | |
| "loss": 1.5353, | |
| "step": 180 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "grad_norm": 66.83992767333984, | |
| "learning_rate": 9.998791022184921e-06, | |
| "loss": 1.5156, | |
| "step": 210 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "grad_norm": 46.15699768066406, | |
| "learning_rate": 9.9984209464165e-06, | |
| "loss": 1.5476, | |
| "step": 240 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "grad_norm": 374.46685791015625, | |
| "learning_rate": 9.998001538251283e-06, | |
| "loss": 1.4349, | |
| "step": 270 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "grad_norm": 34.38185119628906, | |
| "learning_rate": 9.997532801828659e-06, | |
| "loss": 1.4334, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "grad_norm": 57.43722152709961, | |
| "learning_rate": 9.997014741774866e-06, | |
| "loss": 1.4666, | |
| "step": 330 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "grad_norm": 114.363525390625, | |
| "learning_rate": 9.996447363202947e-06, | |
| "loss": 1.2964, | |
| "step": 360 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "grad_norm": 74.82312774658203, | |
| "learning_rate": 9.995830671712701e-06, | |
| "loss": 1.307, | |
| "step": 390 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "grad_norm": 87.48751068115234, | |
| "learning_rate": 9.995164673390624e-06, | |
| "loss": 1.2696, | |
| "step": 420 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "grad_norm": 50.108421325683594, | |
| "learning_rate": 9.994449374809851e-06, | |
| "loss": 1.1757, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "grad_norm": 63.529415130615234, | |
| "learning_rate": 9.99368478303009e-06, | |
| "loss": 1.4811, | |
| "step": 480 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_loss": 1.3713055849075317, | |
| "eval_map": 0.0772, | |
| "eval_map_50": 0.1135, | |
| "eval_map_75": 0.0875, | |
| "eval_map_chicken": 0.0069, | |
| "eval_map_duck": 0.0, | |
| "eval_map_large": 0.3534, | |
| "eval_map_medium": 0.0389, | |
| "eval_map_plant": 0.2247, | |
| "eval_map_small": 0.004, | |
| "eval_mar_1": 0.0527, | |
| "eval_mar_10": 0.1857, | |
| "eval_mar_100": 0.2513, | |
| "eval_mar_100_chicken": 0.0062, | |
| "eval_mar_100_duck": 0.0, | |
| "eval_mar_100_plant": 0.7476, | |
| "eval_mar_large": 0.7937, | |
| "eval_mar_medium": 0.2276, | |
| "eval_mar_small": 0.0875, | |
| "eval_runtime": 5.1266, | |
| "eval_samples_per_second": 19.506, | |
| "eval_steps_per_second": 2.536, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "grad_norm": 48.37826156616211, | |
| "learning_rate": 9.992870905597549e-06, | |
| "loss": 1.2844, | |
| "step": 510 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "grad_norm": 52.364105224609375, | |
| "learning_rate": 9.992007750544876e-06, | |
| "loss": 1.291, | |
| "step": 540 | |
| }, | |
| { | |
| "epoch": 1.1400000000000001, | |
| "grad_norm": 121.01571655273438, | |
| "learning_rate": 9.991095326391061e-06, | |
| "loss": 1.1206, | |
| "step": 570 | |
| }, | |
| { | |
| "epoch": 1.2, | |
| "grad_norm": 57.15728759765625, | |
| "learning_rate": 9.990133642141359e-06, | |
| "loss": 1.3043, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 1.26, | |
| "grad_norm": 142.0693817138672, | |
| "learning_rate": 9.98912270728721e-06, | |
| "loss": 1.3167, | |
| "step": 630 | |
| }, | |
| { | |
| "epoch": 1.32, | |
| "grad_norm": 141.81764221191406, | |
| "learning_rate": 9.988062531806127e-06, | |
| "loss": 1.295, | |
| "step": 660 | |
| }, | |
| { | |
| "epoch": 1.38, | |
| "grad_norm": 78.71722412109375, | |
| "learning_rate": 9.98695312616162e-06, | |
| "loss": 1.3064, | |
| "step": 690 | |
| }, | |
| { | |
| "epoch": 1.44, | |
| "grad_norm": 67.82110595703125, | |
| "learning_rate": 9.98579450130307e-06, | |
| "loss": 1.0711, | |
| "step": 720 | |
| }, | |
| { | |
| "epoch": 1.5, | |
| "grad_norm": 125.79237365722656, | |
| "learning_rate": 9.984586668665641e-06, | |
| "loss": 1.1637, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 1.56, | |
| "grad_norm": 66.8957748413086, | |
| "learning_rate": 9.98332964017015e-06, | |
| "loss": 1.1349, | |
| "step": 780 | |
| }, | |
| { | |
| "epoch": 1.62, | |
| "grad_norm": 77.45892333984375, | |
| "learning_rate": 9.982023428222963e-06, | |
| "loss": 1.0902, | |
| "step": 810 | |
| }, | |
| { | |
| "epoch": 1.6800000000000002, | |
| "grad_norm": 77.51893615722656, | |
| "learning_rate": 9.980668045715864e-06, | |
| "loss": 1.2707, | |
| "step": 840 | |
| }, | |
| { | |
| "epoch": 1.74, | |
| "grad_norm": 52.3269157409668, | |
| "learning_rate": 9.97926350602593e-06, | |
| "loss": 1.1578, | |
| "step": 870 | |
| }, | |
| { | |
| "epoch": 1.8, | |
| "grad_norm": 50.24958419799805, | |
| "learning_rate": 9.9778098230154e-06, | |
| "loss": 1.1071, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 1.8599999999999999, | |
| "grad_norm": 90.0967788696289, | |
| "learning_rate": 9.976307011031542e-06, | |
| "loss": 1.251, | |
| "step": 930 | |
| }, | |
| { | |
| "epoch": 1.92, | |
| "grad_norm": 174.36660766601562, | |
| "learning_rate": 9.974755084906503e-06, | |
| "loss": 1.05, | |
| "step": 960 | |
| }, | |
| { | |
| "epoch": 1.98, | |
| "grad_norm": 78.29364013671875, | |
| "learning_rate": 9.973154059957162e-06, | |
| "loss": 1.093, | |
| "step": 990 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_loss": 1.1410857439041138, | |
| "eval_map": 0.1941, | |
| "eval_map_50": 0.2637, | |
| "eval_map_75": 0.221, | |
| "eval_map_chicken": 0.0, | |
| "eval_map_duck": 0.0, | |
| "eval_map_large": 0.6715, | |
| "eval_map_medium": 0.134, | |
| "eval_map_plant": 0.5822, | |
| "eval_map_small": 0.0042, | |
| "eval_mar_1": 0.0688, | |
| "eval_mar_10": 0.2331, | |
| "eval_mar_100": 0.2586, | |
| "eval_mar_100_chicken": 0.0, | |
| "eval_mar_100_duck": 0.0, | |
| "eval_mar_100_plant": 0.7758, | |
| "eval_mar_large": 0.8259, | |
| "eval_mar_medium": 0.23, | |
| "eval_mar_small": 0.1167, | |
| "eval_runtime": 4.9547, | |
| "eval_samples_per_second": 20.183, | |
| "eval_steps_per_second": 2.624, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 2.04, | |
| "grad_norm": 69.85189056396484, | |
| "learning_rate": 9.971503951984996e-06, | |
| "loss": 1.1616, | |
| "step": 1020 | |
| }, | |
| { | |
| "epoch": 2.1, | |
| "grad_norm": 37.134403228759766, | |
| "learning_rate": 9.9698047772759e-06, | |
| "loss": 1.0393, | |
| "step": 1050 | |
| }, | |
| { | |
| "epoch": 2.16, | |
| "grad_norm": 52.74504852294922, | |
| "learning_rate": 9.968056552600043e-06, | |
| "loss": 1.1147, | |
| "step": 1080 | |
| }, | |
| { | |
| "epoch": 2.22, | |
| "grad_norm": 70.72219848632812, | |
| "learning_rate": 9.966259295211698e-06, | |
| "loss": 1.0797, | |
| "step": 1110 | |
| }, | |
| { | |
| "epoch": 2.2800000000000002, | |
| "grad_norm": 104.32415771484375, | |
| "learning_rate": 9.964413022849069e-06, | |
| "loss": 0.9984, | |
| "step": 1140 | |
| }, | |
| { | |
| "epoch": 2.34, | |
| "grad_norm": 65.52819061279297, | |
| "learning_rate": 9.96251775373412e-06, | |
| "loss": 1.036, | |
| "step": 1170 | |
| }, | |
| { | |
| "epoch": 2.4, | |
| "grad_norm": 43.15259552001953, | |
| "learning_rate": 9.960573506572391e-06, | |
| "loss": 1.3154, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 2.46, | |
| "grad_norm": 76.51600646972656, | |
| "learning_rate": 9.958580300552816e-06, | |
| "loss": 1.1412, | |
| "step": 1230 | |
| }, | |
| { | |
| "epoch": 2.52, | |
| "grad_norm": 0.6177086234092712, | |
| "learning_rate": 9.956538155347534e-06, | |
| "loss": 1.2674, | |
| "step": 1260 | |
| }, | |
| { | |
| "epoch": 2.58, | |
| "grad_norm": 53.222103118896484, | |
| "learning_rate": 9.954447091111695e-06, | |
| "loss": 1.1138, | |
| "step": 1290 | |
| }, | |
| { | |
| "epoch": 2.64, | |
| "grad_norm": 94.70368194580078, | |
| "learning_rate": 9.952307128483257e-06, | |
| "loss": 1.0613, | |
| "step": 1320 | |
| }, | |
| { | |
| "epoch": 2.7, | |
| "grad_norm": 89.58301544189453, | |
| "learning_rate": 9.95011828858279e-06, | |
| "loss": 1.0912, | |
| "step": 1350 | |
| }, | |
| { | |
| "epoch": 2.76, | |
| "grad_norm": 43.07695388793945, | |
| "learning_rate": 9.947880593013256e-06, | |
| "loss": 1.1115, | |
| "step": 1380 | |
| }, | |
| { | |
| "epoch": 2.82, | |
| "grad_norm": 52.474220275878906, | |
| "learning_rate": 9.94559406385981e-06, | |
| "loss": 0.9757, | |
| "step": 1410 | |
| }, | |
| { | |
| "epoch": 2.88, | |
| "grad_norm": 46.698448181152344, | |
| "learning_rate": 9.94325872368957e-06, | |
| "loss": 1.1364, | |
| "step": 1440 | |
| }, | |
| { | |
| "epoch": 2.94, | |
| "grad_norm": 49.13819885253906, | |
| "learning_rate": 9.940874595551403e-06, | |
| "loss": 0.9772, | |
| "step": 1470 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "grad_norm": 101.07115173339844, | |
| "learning_rate": 9.938441702975689e-06, | |
| "loss": 1.06, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_loss": 1.5299323797225952, | |
| "eval_map": 0.1865, | |
| "eval_map_50": 0.2571, | |
| "eval_map_75": 0.2063, | |
| "eval_map_chicken": 0.0, | |
| "eval_map_duck": 0.0, | |
| "eval_map_large": 0.6155, | |
| "eval_map_medium": 0.1497, | |
| "eval_map_plant": 0.5595, | |
| "eval_map_small": 0.0172, | |
| "eval_mar_1": 0.0695, | |
| "eval_mar_10": 0.2154, | |
| "eval_mar_100": 0.2187, | |
| "eval_mar_100_chicken": 0.0, | |
| "eval_mar_100_duck": 0.0, | |
| "eval_mar_100_plant": 0.6562, | |
| "eval_mar_large": 0.6971, | |
| "eval_mar_medium": 0.199, | |
| "eval_mar_small": 0.0583, | |
| "eval_runtime": 4.9962, | |
| "eval_samples_per_second": 20.015, | |
| "eval_steps_per_second": 2.602, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 3.06, | |
| "grad_norm": 96.25062561035156, | |
| "learning_rate": 9.935960069974096e-06, | |
| "loss": 1.2809, | |
| "step": 1530 | |
| }, | |
| { | |
| "epoch": 3.12, | |
| "grad_norm": 90.79141235351562, | |
| "learning_rate": 9.93342972103934e-06, | |
| "loss": 0.9711, | |
| "step": 1560 | |
| }, | |
| { | |
| "epoch": 3.18, | |
| "grad_norm": 87.00637817382812, | |
| "learning_rate": 9.930850681144946e-06, | |
| "loss": 1.0352, | |
| "step": 1590 | |
| }, | |
| { | |
| "epoch": 3.24, | |
| "grad_norm": 64.22712707519531, | |
| "learning_rate": 9.928222975744992e-06, | |
| "loss": 1.0759, | |
| "step": 1620 | |
| }, | |
| { | |
| "epoch": 3.3, | |
| "grad_norm": 38.713409423828125, | |
| "learning_rate": 9.92554663077387e-06, | |
| "loss": 1.2278, | |
| "step": 1650 | |
| }, | |
| { | |
| "epoch": 3.36, | |
| "grad_norm": 346.5014343261719, | |
| "learning_rate": 9.922821672646028e-06, | |
| "loss": 0.8572, | |
| "step": 1680 | |
| }, | |
| { | |
| "epoch": 3.42, | |
| "grad_norm": 42.0444450378418, | |
| "learning_rate": 9.920048128255699e-06, | |
| "loss": 0.9703, | |
| "step": 1710 | |
| }, | |
| { | |
| "epoch": 3.48, | |
| "grad_norm": 43.47895431518555, | |
| "learning_rate": 9.91722602497665e-06, | |
| "loss": 1.0172, | |
| "step": 1740 | |
| }, | |
| { | |
| "epoch": 3.54, | |
| "grad_norm": 47.463985443115234, | |
| "learning_rate": 9.914355390661897e-06, | |
| "loss": 1.0251, | |
| "step": 1770 | |
| }, | |
| { | |
| "epoch": 3.6, | |
| "grad_norm": 98.45388793945312, | |
| "learning_rate": 9.911436253643445e-06, | |
| "loss": 0.8112, | |
| "step": 1800 | |
| }, | |
| { | |
| "epoch": 3.66, | |
| "grad_norm": 47.29812240600586, | |
| "learning_rate": 9.908468642731996e-06, | |
| "loss": 1.0115, | |
| "step": 1830 | |
| }, | |
| { | |
| "epoch": 3.7199999999999998, | |
| "grad_norm": 65.47380065917969, | |
| "learning_rate": 9.90545258721667e-06, | |
| "loss": 1.1073, | |
| "step": 1860 | |
| }, | |
| { | |
| "epoch": 3.7800000000000002, | |
| "grad_norm": 36.47480010986328, | |
| "learning_rate": 9.902388116864723e-06, | |
| "loss": 1.1526, | |
| "step": 1890 | |
| }, | |
| { | |
| "epoch": 3.84, | |
| "grad_norm": 70.70551300048828, | |
| "learning_rate": 9.899275261921236e-06, | |
| "loss": 0.8749, | |
| "step": 1920 | |
| }, | |
| { | |
| "epoch": 3.9, | |
| "grad_norm": 41.91495132446289, | |
| "learning_rate": 9.89611405310883e-06, | |
| "loss": 0.8569, | |
| "step": 1950 | |
| }, | |
| { | |
| "epoch": 3.96, | |
| "grad_norm": 61.9908561706543, | |
| "learning_rate": 9.89290452162736e-06, | |
| "loss": 0.866, | |
| "step": 1980 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_loss": 1.0132447481155396, | |
| "eval_map": 0.2298, | |
| "eval_map_50": 0.307, | |
| "eval_map_75": 0.2534, | |
| "eval_map_chicken": 0.0329, | |
| "eval_map_duck": 0.0, | |
| "eval_map_large": 0.7269, | |
| "eval_map_medium": 0.1936, | |
| "eval_map_plant": 0.6564, | |
| "eval_map_small": 0.0251, | |
| "eval_mar_1": 0.0897, | |
| "eval_mar_10": 0.2671, | |
| "eval_mar_100": 0.2739, | |
| "eval_mar_100_chicken": 0.0613, | |
| "eval_mar_100_duck": 0.0, | |
| "eval_mar_100_plant": 0.7602, | |
| "eval_mar_large": 0.8172, | |
| "eval_mar_medium": 0.2449, | |
| "eval_mar_small": 0.0688, | |
| "eval_runtime": 5.0087, | |
| "eval_samples_per_second": 19.965, | |
| "eval_steps_per_second": 2.595, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 4.02, | |
| "grad_norm": 65.72293853759766, | |
| "learning_rate": 9.88964669915361e-06, | |
| "loss": 1.0872, | |
| "step": 2010 | |
| }, | |
| { | |
| "epoch": 4.08, | |
| "grad_norm": 126.20653533935547, | |
| "learning_rate": 9.886340617840968e-06, | |
| "loss": 1.0516, | |
| "step": 2040 | |
| }, | |
| { | |
| "epoch": 4.14, | |
| "grad_norm": 62.30933380126953, | |
| "learning_rate": 9.882986310319124e-06, | |
| "loss": 0.993, | |
| "step": 2070 | |
| }, | |
| { | |
| "epoch": 4.2, | |
| "grad_norm": 62.221160888671875, | |
| "learning_rate": 9.879583809693737e-06, | |
| "loss": 1.0601, | |
| "step": 2100 | |
| }, | |
| { | |
| "epoch": 4.26, | |
| "grad_norm": 55.73953628540039, | |
| "learning_rate": 9.876133149546117e-06, | |
| "loss": 0.7821, | |
| "step": 2130 | |
| }, | |
| { | |
| "epoch": 4.32, | |
| "grad_norm": 57.795196533203125, | |
| "learning_rate": 9.872634363932887e-06, | |
| "loss": 0.9674, | |
| "step": 2160 | |
| }, | |
| { | |
| "epoch": 4.38, | |
| "grad_norm": 49.45548629760742, | |
| "learning_rate": 9.869087487385644e-06, | |
| "loss": 0.8739, | |
| "step": 2190 | |
| }, | |
| { | |
| "epoch": 4.44, | |
| "grad_norm": 55.93813705444336, | |
| "learning_rate": 9.865492554910634e-06, | |
| "loss": 0.9908, | |
| "step": 2220 | |
| }, | |
| { | |
| "epoch": 4.5, | |
| "grad_norm": 55.131141662597656, | |
| "learning_rate": 9.861849601988384e-06, | |
| "loss": 0.9069, | |
| "step": 2250 | |
| }, | |
| { | |
| "epoch": 4.5600000000000005, | |
| "grad_norm": 45.55573272705078, | |
| "learning_rate": 9.85815866457337e-06, | |
| "loss": 0.9796, | |
| "step": 2280 | |
| }, | |
| { | |
| "epoch": 4.62, | |
| "grad_norm": 41.95097732543945, | |
| "learning_rate": 9.854419779093656e-06, | |
| "loss": 0.8175, | |
| "step": 2310 | |
| }, | |
| { | |
| "epoch": 4.68, | |
| "grad_norm": 480.2862243652344, | |
| "learning_rate": 9.85063298245053e-06, | |
| "loss": 0.8588, | |
| "step": 2340 | |
| }, | |
| { | |
| "epoch": 4.74, | |
| "grad_norm": 51.55152893066406, | |
| "learning_rate": 9.846798312018147e-06, | |
| "loss": 0.9936, | |
| "step": 2370 | |
| }, | |
| { | |
| "epoch": 4.8, | |
| "grad_norm": 36.979496002197266, | |
| "learning_rate": 9.842915805643156e-06, | |
| "loss": 0.9615, | |
| "step": 2400 | |
| }, | |
| { | |
| "epoch": 4.86, | |
| "grad_norm": 47.84645462036133, | |
| "learning_rate": 9.838985501644329e-06, | |
| "loss": 0.8042, | |
| "step": 2430 | |
| }, | |
| { | |
| "epoch": 4.92, | |
| "grad_norm": 40.069725036621094, | |
| "learning_rate": 9.835007438812177e-06, | |
| "loss": 0.9539, | |
| "step": 2460 | |
| }, | |
| { | |
| "epoch": 4.98, | |
| "grad_norm": 86.61743927001953, | |
| "learning_rate": 9.830981656408575e-06, | |
| "loss": 0.8301, | |
| "step": 2490 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_loss": 0.9037431478500366, | |
| "eval_map": 0.2895, | |
| "eval_map_50": 0.3972, | |
| "eval_map_75": 0.3375, | |
| "eval_map_chicken": 0.2025, | |
| "eval_map_duck": 0.0, | |
| "eval_map_large": 0.7397, | |
| "eval_map_medium": 0.25, | |
| "eval_map_plant": 0.6662, | |
| "eval_map_small": 0.0486, | |
| "eval_mar_1": 0.1158, | |
| "eval_mar_10": 0.3495, | |
| "eval_mar_100": 0.3568, | |
| "eval_mar_100_chicken": 0.3062, | |
| "eval_mar_100_duck": 0.0, | |
| "eval_mar_100_plant": 0.7643, | |
| "eval_mar_large": 0.8305, | |
| "eval_mar_medium": 0.3238, | |
| "eval_mar_small": 0.1063, | |
| "eval_runtime": 4.9996, | |
| "eval_samples_per_second": 20.002, | |
| "eval_steps_per_second": 2.6, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 5.04, | |
| "grad_norm": 69.0333251953125, | |
| "learning_rate": 9.82690819416637e-06, | |
| "loss": 0.933, | |
| "step": 2520 | |
| }, | |
| { | |
| "epoch": 5.1, | |
| "grad_norm": 38.97633743286133, | |
| "learning_rate": 9.822787092288991e-06, | |
| "loss": 1.0335, | |
| "step": 2550 | |
| }, | |
| { | |
| "epoch": 5.16, | |
| "grad_norm": 52.52025604248047, | |
| "learning_rate": 9.81861839145005e-06, | |
| "loss": 0.9546, | |
| "step": 2580 | |
| }, | |
| { | |
| "epoch": 5.22, | |
| "grad_norm": 68.2970962524414, | |
| "learning_rate": 9.814402132792939e-06, | |
| "loss": 0.8876, | |
| "step": 2610 | |
| }, | |
| { | |
| "epoch": 5.28, | |
| "grad_norm": 89.29615783691406, | |
| "learning_rate": 9.81013835793043e-06, | |
| "loss": 0.9496, | |
| "step": 2640 | |
| }, | |
| { | |
| "epoch": 5.34, | |
| "grad_norm": 50.56565475463867, | |
| "learning_rate": 9.80582710894426e-06, | |
| "loss": 0.7895, | |
| "step": 2670 | |
| }, | |
| { | |
| "epoch": 5.4, | |
| "grad_norm": 33.971832275390625, | |
| "learning_rate": 9.801468428384716e-06, | |
| "loss": 0.9679, | |
| "step": 2700 | |
| }, | |
| { | |
| "epoch": 5.46, | |
| "grad_norm": 45.445594787597656, | |
| "learning_rate": 9.797062359270215e-06, | |
| "loss": 0.924, | |
| "step": 2730 | |
| }, | |
| { | |
| "epoch": 5.52, | |
| "grad_norm": 45.259674072265625, | |
| "learning_rate": 9.79260894508688e-06, | |
| "loss": 0.9305, | |
| "step": 2760 | |
| }, | |
| { | |
| "epoch": 5.58, | |
| "grad_norm": 42.04433059692383, | |
| "learning_rate": 9.788108229788111e-06, | |
| "loss": 0.9169, | |
| "step": 2790 | |
| }, | |
| { | |
| "epoch": 5.64, | |
| "grad_norm": 109.8760986328125, | |
| "learning_rate": 9.783560257794153e-06, | |
| "loss": 1.2765, | |
| "step": 2820 | |
| }, | |
| { | |
| "epoch": 5.7, | |
| "grad_norm": 63.83809280395508, | |
| "learning_rate": 9.778965073991652e-06, | |
| "loss": 1.2013, | |
| "step": 2850 | |
| }, | |
| { | |
| "epoch": 5.76, | |
| "grad_norm": 205.1379852294922, | |
| "learning_rate": 9.774322723733216e-06, | |
| "loss": 0.9418, | |
| "step": 2880 | |
| }, | |
| { | |
| "epoch": 5.82, | |
| "grad_norm": 23.930816650390625, | |
| "learning_rate": 9.769633252836969e-06, | |
| "loss": 0.8923, | |
| "step": 2910 | |
| }, | |
| { | |
| "epoch": 5.88, | |
| "grad_norm": 39.413909912109375, | |
| "learning_rate": 9.764896707586095e-06, | |
| "loss": 0.8823, | |
| "step": 2940 | |
| }, | |
| { | |
| "epoch": 5.9399999999999995, | |
| "grad_norm": 111.26641845703125, | |
| "learning_rate": 9.760113134728383e-06, | |
| "loss": 0.9271, | |
| "step": 2970 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "grad_norm": 83.30919647216797, | |
| "learning_rate": 9.755282581475769e-06, | |
| "loss": 0.8386, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "eval_loss": 0.9659404158592224, | |
| "eval_map": 0.3229, | |
| "eval_map_50": 0.4625, | |
| "eval_map_75": 0.3953, | |
| "eval_map_chicken": 0.3161, | |
| "eval_map_duck": 0.0, | |
| "eval_map_large": 0.7146, | |
| "eval_map_medium": 0.2944, | |
| "eval_map_plant": 0.6527, | |
| "eval_map_small": 0.012, | |
| "eval_mar_1": 0.1161, | |
| "eval_mar_10": 0.3938, | |
| "eval_mar_100": 0.3977, | |
| "eval_mar_100_chicken": 0.4693, | |
| "eval_mar_100_duck": 0.0, | |
| "eval_mar_100_plant": 0.7236, | |
| "eval_mar_large": 0.7837, | |
| "eval_mar_medium": 0.3785, | |
| "eval_mar_small": 0.0333, | |
| "eval_runtime": 4.9283, | |
| "eval_samples_per_second": 20.291, | |
| "eval_steps_per_second": 2.638, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 6.06, | |
| "grad_norm": 34.77956008911133, | |
| "learning_rate": 9.750405095503859e-06, | |
| "loss": 0.9097, | |
| "step": 3030 | |
| }, | |
| { | |
| "epoch": 6.12, | |
| "grad_norm": 41.7236328125, | |
| "learning_rate": 9.745480724951473e-06, | |
| "loss": 0.7642, | |
| "step": 3060 | |
| }, | |
| { | |
| "epoch": 6.18, | |
| "grad_norm": 21.034090042114258, | |
| "learning_rate": 9.74050951842016e-06, | |
| "loss": 0.8596, | |
| "step": 3090 | |
| }, | |
| { | |
| "epoch": 6.24, | |
| "grad_norm": 36.62574005126953, | |
| "learning_rate": 9.735491524973723e-06, | |
| "loss": 0.8173, | |
| "step": 3120 | |
| }, | |
| { | |
| "epoch": 6.3, | |
| "grad_norm": 89.44732666015625, | |
| "learning_rate": 9.730426794137727e-06, | |
| "loss": 0.856, | |
| "step": 3150 | |
| }, | |
| { | |
| "epoch": 6.36, | |
| "grad_norm": 56.18183898925781, | |
| "learning_rate": 9.725315375899025e-06, | |
| "loss": 0.8138, | |
| "step": 3180 | |
| }, | |
| { | |
| "epoch": 6.42, | |
| "grad_norm": 41.0568962097168, | |
| "learning_rate": 9.72015732070525e-06, | |
| "loss": 0.8276, | |
| "step": 3210 | |
| }, | |
| { | |
| "epoch": 6.48, | |
| "grad_norm": 76.43619537353516, | |
| "learning_rate": 9.714952679464324e-06, | |
| "loss": 0.8188, | |
| "step": 3240 | |
| }, | |
| { | |
| "epoch": 6.54, | |
| "grad_norm": 39.30849075317383, | |
| "learning_rate": 9.709701503543954e-06, | |
| "loss": 0.7778, | |
| "step": 3270 | |
| }, | |
| { | |
| "epoch": 6.6, | |
| "grad_norm": 71.78713989257812, | |
| "learning_rate": 9.704403844771128e-06, | |
| "loss": 0.8526, | |
| "step": 3300 | |
| }, | |
| { | |
| "epoch": 6.66, | |
| "grad_norm": 68.13652038574219, | |
| "learning_rate": 9.699059755431599e-06, | |
| "loss": 0.871, | |
| "step": 3330 | |
| }, | |
| { | |
| "epoch": 6.72, | |
| "grad_norm": 158.08872985839844, | |
| "learning_rate": 9.693669288269371e-06, | |
| "loss": 1.1321, | |
| "step": 3360 | |
| }, | |
| { | |
| "epoch": 6.78, | |
| "grad_norm": 34.7762451171875, | |
| "learning_rate": 9.688232496486179e-06, | |
| "loss": 0.8604, | |
| "step": 3390 | |
| }, | |
| { | |
| "epoch": 6.84, | |
| "grad_norm": 42.608489990234375, | |
| "learning_rate": 9.682749433740963e-06, | |
| "loss": 0.9493, | |
| "step": 3420 | |
| }, | |
| { | |
| "epoch": 6.9, | |
| "grad_norm": 0.026101410388946533, | |
| "learning_rate": 9.677220154149338e-06, | |
| "loss": 0.7639, | |
| "step": 3450 | |
| }, | |
| { | |
| "epoch": 6.96, | |
| "grad_norm": 147.83847045898438, | |
| "learning_rate": 9.671644712283061e-06, | |
| "loss": 0.9838, | |
| "step": 3480 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "eval_loss": 0.7706183791160583, | |
| "eval_map": 0.3831, | |
| "eval_map_50": 0.5318, | |
| "eval_map_75": 0.4649, | |
| "eval_map_chicken": 0.4609, | |
| "eval_map_duck": 0.0, | |
| "eval_map_large": 0.7573, | |
| "eval_map_medium": 0.3525, | |
| "eval_map_plant": 0.6886, | |
| "eval_map_small": 0.0314, | |
| "eval_mar_1": 0.1347, | |
| "eval_mar_10": 0.4783, | |
| "eval_mar_100": 0.4829, | |
| "eval_mar_100_chicken": 0.6889, | |
| "eval_mar_100_duck": 0.0, | |
| "eval_mar_100_plant": 0.7599, | |
| "eval_mar_large": 0.8234, | |
| "eval_mar_medium": 0.4636, | |
| "eval_mar_small": 0.0854, | |
| "eval_runtime": 5.0062, | |
| "eval_samples_per_second": 19.975, | |
| "eval_steps_per_second": 2.597, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 7.02, | |
| "grad_norm": 55.403541564941406, | |
| "learning_rate": 9.666023163169493e-06, | |
| "loss": 0.8917, | |
| "step": 3510 | |
| }, | |
| { | |
| "epoch": 7.08, | |
| "grad_norm": 189.77401733398438, | |
| "learning_rate": 9.660355562291055e-06, | |
| "loss": 0.858, | |
| "step": 3540 | |
| }, | |
| { | |
| "epoch": 7.14, | |
| "grad_norm": 58.92485427856445, | |
| "learning_rate": 9.65464196558468e-06, | |
| "loss": 0.7254, | |
| "step": 3570 | |
| }, | |
| { | |
| "epoch": 7.2, | |
| "grad_norm": 46.30803298950195, | |
| "learning_rate": 9.648882429441258e-06, | |
| "loss": 0.7928, | |
| "step": 3600 | |
| }, | |
| { | |
| "epoch": 7.26, | |
| "grad_norm": 111.99207305908203, | |
| "learning_rate": 9.643077010705088e-06, | |
| "loss": 0.7594, | |
| "step": 3630 | |
| }, | |
| { | |
| "epoch": 7.32, | |
| "grad_norm": 72.06273651123047, | |
| "learning_rate": 9.637225766673309e-06, | |
| "loss": 0.8397, | |
| "step": 3660 | |
| }, | |
| { | |
| "epoch": 7.38, | |
| "grad_norm": 35.78024673461914, | |
| "learning_rate": 9.631328755095334e-06, | |
| "loss": 0.8906, | |
| "step": 3690 | |
| }, | |
| { | |
| "epoch": 7.44, | |
| "grad_norm": 140.4214630126953, | |
| "learning_rate": 9.62538603417229e-06, | |
| "loss": 0.6961, | |
| "step": 3720 | |
| }, | |
| { | |
| "epoch": 7.5, | |
| "grad_norm": 55.60920715332031, | |
| "learning_rate": 9.619397662556434e-06, | |
| "loss": 0.7646, | |
| "step": 3750 | |
| }, | |
| { | |
| "epoch": 7.5600000000000005, | |
| "grad_norm": 43.936187744140625, | |
| "learning_rate": 9.613363699350575e-06, | |
| "loss": 0.8446, | |
| "step": 3780 | |
| }, | |
| { | |
| "epoch": 7.62, | |
| "grad_norm": 61.2660026550293, | |
| "learning_rate": 9.607284204107493e-06, | |
| "loss": 0.7357, | |
| "step": 3810 | |
| }, | |
| { | |
| "epoch": 7.68, | |
| "grad_norm": 55.53911209106445, | |
| "learning_rate": 9.601159236829353e-06, | |
| "loss": 0.7807, | |
| "step": 3840 | |
| }, | |
| { | |
| "epoch": 7.74, | |
| "grad_norm": 86.03663635253906, | |
| "learning_rate": 9.594988857967107e-06, | |
| "loss": 0.9694, | |
| "step": 3870 | |
| }, | |
| { | |
| "epoch": 7.8, | |
| "grad_norm": 37.17007064819336, | |
| "learning_rate": 9.588773128419907e-06, | |
| "loss": 0.754, | |
| "step": 3900 | |
| }, | |
| { | |
| "epoch": 7.86, | |
| "grad_norm": 47.19321060180664, | |
| "learning_rate": 9.58251210953449e-06, | |
| "loss": 1.0753, | |
| "step": 3930 | |
| }, | |
| { | |
| "epoch": 7.92, | |
| "grad_norm": 51.06205368041992, | |
| "learning_rate": 9.576205863104588e-06, | |
| "loss": 0.8297, | |
| "step": 3960 | |
| }, | |
| { | |
| "epoch": 7.98, | |
| "grad_norm": 60.964874267578125, | |
| "learning_rate": 9.569854451370306e-06, | |
| "loss": 0.8051, | |
| "step": 3990 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "eval_loss": 0.7423961758613586, | |
| "eval_map": 0.3909, | |
| "eval_map_50": 0.5361, | |
| "eval_map_75": 0.4744, | |
| "eval_map_chicken": 0.4864, | |
| "eval_map_duck": 0.0, | |
| "eval_map_large": 0.7506, | |
| "eval_map_medium": 0.362, | |
| "eval_map_plant": 0.6864, | |
| "eval_map_small": 0.0298, | |
| "eval_mar_1": 0.1396, | |
| "eval_mar_10": 0.489, | |
| "eval_mar_100": 0.4921, | |
| "eval_mar_100_chicken": 0.712, | |
| "eval_mar_100_duck": 0.0, | |
| "eval_mar_100_plant": 0.7643, | |
| "eval_mar_large": 0.8222, | |
| "eval_mar_medium": 0.4762, | |
| "eval_mar_small": 0.1063, | |
| "eval_runtime": 5.1458, | |
| "eval_samples_per_second": 19.433, | |
| "eval_steps_per_second": 2.526, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 8.04, | |
| "grad_norm": 34.82122802734375, | |
| "learning_rate": 9.563457937017514e-06, | |
| "loss": 0.8709, | |
| "step": 4020 | |
| }, | |
| { | |
| "epoch": 8.1, | |
| "grad_norm": 97.54671478271484, | |
| "learning_rate": 9.557016383177226e-06, | |
| "loss": 0.9338, | |
| "step": 4050 | |
| }, | |
| { | |
| "epoch": 8.16, | |
| "grad_norm": 38.53603744506836, | |
| "learning_rate": 9.550529853424979e-06, | |
| "loss": 0.8831, | |
| "step": 4080 | |
| }, | |
| { | |
| "epoch": 8.22, | |
| "grad_norm": 55.73428726196289, | |
| "learning_rate": 9.543998411780202e-06, | |
| "loss": 0.7674, | |
| "step": 4110 | |
| }, | |
| { | |
| "epoch": 8.28, | |
| "grad_norm": 29.146512985229492, | |
| "learning_rate": 9.537422122705585e-06, | |
| "loss": 1.1078, | |
| "step": 4140 | |
| }, | |
| { | |
| "epoch": 8.34, | |
| "grad_norm": 119.19530487060547, | |
| "learning_rate": 9.530801051106449e-06, | |
| "loss": 0.6341, | |
| "step": 4170 | |
| }, | |
| { | |
| "epoch": 8.4, | |
| "grad_norm": 66.55436706542969, | |
| "learning_rate": 9.524135262330098e-06, | |
| "loss": 0.7727, | |
| "step": 4200 | |
| }, | |
| { | |
| "epoch": 8.46, | |
| "grad_norm": 108.39798736572266, | |
| "learning_rate": 9.517424822165175e-06, | |
| "loss": 0.8202, | |
| "step": 4230 | |
| }, | |
| { | |
| "epoch": 8.52, | |
| "grad_norm": 47.83013916015625, | |
| "learning_rate": 9.510669796841014e-06, | |
| "loss": 0.9373, | |
| "step": 4260 | |
| }, | |
| { | |
| "epoch": 8.58, | |
| "grad_norm": 57.3923225402832, | |
| "learning_rate": 9.503870253026992e-06, | |
| "loss": 0.6963, | |
| "step": 4290 | |
| }, | |
| { | |
| "epoch": 8.64, | |
| "grad_norm": 47.453041076660156, | |
| "learning_rate": 9.497026257831856e-06, | |
| "loss": 0.7835, | |
| "step": 4320 | |
| }, | |
| { | |
| "epoch": 8.7, | |
| "grad_norm": 50.4648551940918, | |
| "learning_rate": 9.490137878803078e-06, | |
| "loss": 0.6982, | |
| "step": 4350 | |
| }, | |
| { | |
| "epoch": 8.76, | |
| "grad_norm": 67.14995574951172, | |
| "learning_rate": 9.48320518392618e-06, | |
| "loss": 0.8588, | |
| "step": 4380 | |
| }, | |
| { | |
| "epoch": 8.82, | |
| "grad_norm": 61.14694595336914, | |
| "learning_rate": 9.476228241624059e-06, | |
| "loss": 0.7882, | |
| "step": 4410 | |
| }, | |
| { | |
| "epoch": 8.88, | |
| "grad_norm": 26.730520248413086, | |
| "learning_rate": 9.46920712075632e-06, | |
| "loss": 0.8078, | |
| "step": 4440 | |
| }, | |
| { | |
| "epoch": 8.94, | |
| "grad_norm": 66.36459350585938, | |
| "learning_rate": 9.46214189061859e-06, | |
| "loss": 0.7777, | |
| "step": 4470 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "grad_norm": 0.00832737423479557, | |
| "learning_rate": 9.45503262094184e-06, | |
| "loss": 0.7114, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "eval_loss": 0.6859935522079468, | |
| "eval_map": 0.425, | |
| "eval_map_50": 0.5595, | |
| "eval_map_75": 0.5062, | |
| "eval_map_chicken": 0.5552, | |
| "eval_map_duck": 0.0, | |
| "eval_map_large": 0.7822, | |
| "eval_map_medium": 0.4023, | |
| "eval_map_plant": 0.7197, | |
| "eval_map_small": 0.0656, | |
| "eval_mar_1": 0.1429, | |
| "eval_mar_10": 0.5123, | |
| "eval_mar_100": 0.5168, | |
| "eval_mar_100_chicken": 0.7609, | |
| "eval_mar_100_duck": 0.0, | |
| "eval_mar_100_plant": 0.7896, | |
| "eval_mar_large": 0.849, | |
| "eval_mar_medium": 0.5014, | |
| "eval_mar_small": 0.1104, | |
| "eval_runtime": 4.9198, | |
| "eval_samples_per_second": 20.326, | |
| "eval_steps_per_second": 2.642, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 9.06, | |
| "grad_norm": 41.9913215637207, | |
| "learning_rate": 9.447879381891691e-06, | |
| "loss": 0.742, | |
| "step": 4530 | |
| }, | |
| { | |
| "epoch": 9.12, | |
| "grad_norm": 118.60945892333984, | |
| "learning_rate": 9.440682244067724e-06, | |
| "loss": 0.7156, | |
| "step": 4560 | |
| }, | |
| { | |
| "epoch": 9.18, | |
| "grad_norm": 46.92107009887695, | |
| "learning_rate": 9.433441278502784e-06, | |
| "loss": 0.7649, | |
| "step": 4590 | |
| }, | |
| { | |
| "epoch": 9.24, | |
| "grad_norm": 47.721012115478516, | |
| "learning_rate": 9.426156556662276e-06, | |
| "loss": 0.7351, | |
| "step": 4620 | |
| }, | |
| { | |
| "epoch": 9.3, | |
| "grad_norm": 32.733455657958984, | |
| "learning_rate": 9.418828150443469e-06, | |
| "loss": 0.7486, | |
| "step": 4650 | |
| }, | |
| { | |
| "epoch": 9.36, | |
| "grad_norm": 96.80855560302734, | |
| "learning_rate": 9.411456132174768e-06, | |
| "loss": 0.7729, | |
| "step": 4680 | |
| }, | |
| { | |
| "epoch": 9.42, | |
| "grad_norm": 37.33518981933594, | |
| "learning_rate": 9.404040574615018e-06, | |
| "loss": 0.7362, | |
| "step": 4710 | |
| }, | |
| { | |
| "epoch": 9.48, | |
| "grad_norm": 43.71377182006836, | |
| "learning_rate": 9.396581550952781e-06, | |
| "loss": 0.6084, | |
| "step": 4740 | |
| }, | |
| { | |
| "epoch": 9.54, | |
| "grad_norm": 98.9687728881836, | |
| "learning_rate": 9.38907913480561e-06, | |
| "loss": 0.8131, | |
| "step": 4770 | |
| }, | |
| { | |
| "epoch": 9.6, | |
| "grad_norm": 40.247962951660156, | |
| "learning_rate": 9.381533400219319e-06, | |
| "loss": 0.7444, | |
| "step": 4800 | |
| }, | |
| { | |
| "epoch": 9.66, | |
| "grad_norm": 48.69584655761719, | |
| "learning_rate": 9.373944421667264e-06, | |
| "loss": 0.8159, | |
| "step": 4830 | |
| }, | |
| { | |
| "epoch": 9.72, | |
| "grad_norm": 30.084497451782227, | |
| "learning_rate": 9.366312274049602e-06, | |
| "loss": 0.7879, | |
| "step": 4860 | |
| }, | |
| { | |
| "epoch": 9.78, | |
| "grad_norm": 55.0740966796875, | |
| "learning_rate": 9.358637032692546e-06, | |
| "loss": 0.7506, | |
| "step": 4890 | |
| }, | |
| { | |
| "epoch": 9.84, | |
| "grad_norm": 116.40509796142578, | |
| "learning_rate": 9.35091877334763e-06, | |
| "loss": 0.6648, | |
| "step": 4920 | |
| }, | |
| { | |
| "epoch": 9.9, | |
| "grad_norm": 134.0435791015625, | |
| "learning_rate": 9.343157572190957e-06, | |
| "loss": 0.7205, | |
| "step": 4950 | |
| }, | |
| { | |
| "epoch": 9.96, | |
| "grad_norm": 60.22429656982422, | |
| "learning_rate": 9.33535350582245e-06, | |
| "loss": 0.8088, | |
| "step": 4980 | |
| }, | |
| { | |
| "epoch": 10.0, | |
| "eval_loss": 0.6922460794448853, | |
| "eval_map": 0.4107, | |
| "eval_map_50": 0.5706, | |
| "eval_map_75": 0.5008, | |
| "eval_map_chicken": 0.527, | |
| "eval_map_duck": 0.0, | |
| "eval_map_large": 0.7653, | |
| "eval_map_medium": 0.3953, | |
| "eval_map_plant": 0.705, | |
| "eval_map_small": 0.0425, | |
| "eval_mar_1": 0.1375, | |
| "eval_mar_10": 0.4929, | |
| "eval_mar_100": 0.4981, | |
| "eval_mar_100_chicken": 0.7218, | |
| "eval_mar_100_duck": 0.0, | |
| "eval_mar_100_plant": 0.7726, | |
| "eval_mar_large": 0.8272, | |
| "eval_mar_medium": 0.4833, | |
| "eval_mar_small": 0.1333, | |
| "eval_runtime": 5.2996, | |
| "eval_samples_per_second": 18.869, | |
| "eval_steps_per_second": 2.453, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 10.02, | |
| "grad_norm": 79.86975860595703, | |
| "learning_rate": 9.327506651265096e-06, | |
| "loss": 0.6511, | |
| "step": 5010 | |
| }, | |
| { | |
| "epoch": 10.08, | |
| "grad_norm": 101.35730743408203, | |
| "learning_rate": 9.319617085964177e-06, | |
| "loss": 0.6866, | |
| "step": 5040 | |
| }, | |
| { | |
| "epoch": 10.14, | |
| "grad_norm": 50.54615783691406, | |
| "learning_rate": 9.31168488778652e-06, | |
| "loss": 0.7279, | |
| "step": 5070 | |
| }, | |
| { | |
| "epoch": 10.2, | |
| "grad_norm": 0.004792827181518078, | |
| "learning_rate": 9.30371013501972e-06, | |
| "loss": 0.8451, | |
| "step": 5100 | |
| }, | |
| { | |
| "epoch": 10.26, | |
| "grad_norm": 65.52108764648438, | |
| "learning_rate": 9.295692906371362e-06, | |
| "loss": 0.7139, | |
| "step": 5130 | |
| }, | |
| { | |
| "epoch": 10.32, | |
| "grad_norm": 50.15769577026367, | |
| "learning_rate": 9.287633280968263e-06, | |
| "loss": 0.7791, | |
| "step": 5160 | |
| }, | |
| { | |
| "epoch": 10.38, | |
| "grad_norm": 41.1558952331543, | |
| "learning_rate": 9.279531338355666e-06, | |
| "loss": 0.7288, | |
| "step": 5190 | |
| }, | |
| { | |
| "epoch": 10.44, | |
| "grad_norm": 91.48247528076172, | |
| "learning_rate": 9.271387158496477e-06, | |
| "loss": 0.7832, | |
| "step": 5220 | |
| }, | |
| { | |
| "epoch": 10.5, | |
| "grad_norm": 53.355899810791016, | |
| "learning_rate": 9.263200821770462e-06, | |
| "loss": 0.8104, | |
| "step": 5250 | |
| }, | |
| { | |
| "epoch": 10.56, | |
| "grad_norm": 36.56712341308594, | |
| "learning_rate": 9.25497240897346e-06, | |
| "loss": 0.7152, | |
| "step": 5280 | |
| }, | |
| { | |
| "epoch": 10.62, | |
| "grad_norm": 0.004473070148378611, | |
| "learning_rate": 9.246702001316584e-06, | |
| "loss": 0.7701, | |
| "step": 5310 | |
| }, | |
| { | |
| "epoch": 10.68, | |
| "grad_norm": 93.80265808105469, | |
| "learning_rate": 9.238389680425417e-06, | |
| "loss": 0.6912, | |
| "step": 5340 | |
| }, | |
| { | |
| "epoch": 10.74, | |
| "grad_norm": 80.67282104492188, | |
| "learning_rate": 9.230035528339212e-06, | |
| "loss": 0.7225, | |
| "step": 5370 | |
| }, | |
| { | |
| "epoch": 10.8, | |
| "grad_norm": 66.9299545288086, | |
| "learning_rate": 9.221639627510076e-06, | |
| "loss": 0.6667, | |
| "step": 5400 | |
| }, | |
| { | |
| "epoch": 10.86, | |
| "grad_norm": 63.10317611694336, | |
| "learning_rate": 9.213202060802162e-06, | |
| "loss": 0.7136, | |
| "step": 5430 | |
| }, | |
| { | |
| "epoch": 10.92, | |
| "grad_norm": 53.23859786987305, | |
| "learning_rate": 9.204722911490847e-06, | |
| "loss": 0.7137, | |
| "step": 5460 | |
| }, | |
| { | |
| "epoch": 10.98, | |
| "grad_norm": 72.44120025634766, | |
| "learning_rate": 9.196202263261908e-06, | |
| "loss": 0.7049, | |
| "step": 5490 | |
| }, | |
| { | |
| "epoch": 11.0, | |
| "eval_loss": 0.6989492177963257, | |
| "eval_map": 0.4204, | |
| "eval_map_50": 0.5653, | |
| "eval_map_75": 0.5121, | |
| "eval_map_chicken": 0.5406, | |
| "eval_map_duck": 0.0, | |
| "eval_map_large": 0.7752, | |
| "eval_map_medium": 0.4022, | |
| "eval_map_plant": 0.7205, | |
| "eval_map_small": 0.0859, | |
| "eval_mar_1": 0.1439, | |
| "eval_mar_10": 0.4952, | |
| "eval_mar_100": 0.499, | |
| "eval_mar_100_chicken": 0.7182, | |
| "eval_mar_100_duck": 0.0, | |
| "eval_mar_100_plant": 0.7787, | |
| "eval_mar_large": 0.8293, | |
| "eval_mar_medium": 0.4863, | |
| "eval_mar_small": 0.1458, | |
| "eval_runtime": 5.0962, | |
| "eval_samples_per_second": 19.622, | |
| "eval_steps_per_second": 2.551, | |
| "step": 5500 | |
| }, | |
| { | |
| "epoch": 11.04, | |
| "grad_norm": 92.46695709228516, | |
| "learning_rate": 9.18764020021071e-06, | |
| "loss": 0.6504, | |
| "step": 5520 | |
| }, | |
| { | |
| "epoch": 11.1, | |
| "grad_norm": 28.806991577148438, | |
| "learning_rate": 9.179036806841352e-06, | |
| "loss": 0.6896, | |
| "step": 5550 | |
| }, | |
| { | |
| "epoch": 11.16, | |
| "grad_norm": 48.64046859741211, | |
| "learning_rate": 9.170392168065858e-06, | |
| "loss": 0.7909, | |
| "step": 5580 | |
| }, | |
| { | |
| "epoch": 11.22, | |
| "grad_norm": 84.09498596191406, | |
| "learning_rate": 9.161706369203319e-06, | |
| "loss": 0.714, | |
| "step": 5610 | |
| }, | |
| { | |
| "epoch": 11.28, | |
| "grad_norm": 73.94583129882812, | |
| "learning_rate": 9.152979495979064e-06, | |
| "loss": 0.5829, | |
| "step": 5640 | |
| }, | |
| { | |
| "epoch": 11.34, | |
| "grad_norm": 71.72589874267578, | |
| "learning_rate": 9.14421163452381e-06, | |
| "loss": 0.59, | |
| "step": 5670 | |
| }, | |
| { | |
| "epoch": 11.4, | |
| "grad_norm": 415.1078796386719, | |
| "learning_rate": 9.13540287137281e-06, | |
| "loss": 0.7187, | |
| "step": 5700 | |
| }, | |
| { | |
| "epoch": 11.46, | |
| "grad_norm": 48.62983322143555, | |
| "learning_rate": 9.126553293465e-06, | |
| "loss": 0.8666, | |
| "step": 5730 | |
| }, | |
| { | |
| "epoch": 11.52, | |
| "grad_norm": 36.73117446899414, | |
| "learning_rate": 9.117662988142138e-06, | |
| "loss": 0.916, | |
| "step": 5760 | |
| }, | |
| { | |
| "epoch": 11.58, | |
| "grad_norm": 115.80775451660156, | |
| "learning_rate": 9.108732043147952e-06, | |
| "loss": 0.7241, | |
| "step": 5790 | |
| }, | |
| { | |
| "epoch": 11.64, | |
| "grad_norm": 77.43502807617188, | |
| "learning_rate": 9.099760546627262e-06, | |
| "loss": 0.6617, | |
| "step": 5820 | |
| }, | |
| { | |
| "epoch": 11.7, | |
| "grad_norm": 31.340518951416016, | |
| "learning_rate": 9.090748587125118e-06, | |
| "loss": 0.7765, | |
| "step": 5850 | |
| }, | |
| { | |
| "epoch": 11.76, | |
| "grad_norm": 34.73516845703125, | |
| "learning_rate": 9.08169625358592e-06, | |
| "loss": 0.6421, | |
| "step": 5880 | |
| }, | |
| { | |
| "epoch": 11.82, | |
| "grad_norm": 50.42566680908203, | |
| "learning_rate": 9.072603635352548e-06, | |
| "loss": 0.8824, | |
| "step": 5910 | |
| }, | |
| { | |
| "epoch": 11.88, | |
| "grad_norm": 61.33326721191406, | |
| "learning_rate": 9.06347082216547e-06, | |
| "loss": 0.8302, | |
| "step": 5940 | |
| }, | |
| { | |
| "epoch": 11.94, | |
| "grad_norm": 94.31954193115234, | |
| "learning_rate": 9.054297904161868e-06, | |
| "loss": 0.7238, | |
| "step": 5970 | |
| }, | |
| { | |
| "epoch": 12.0, | |
| "grad_norm": 241.52101135253906, | |
| "learning_rate": 9.045084971874738e-06, | |
| "loss": 0.7244, | |
| "step": 6000 | |
| }, | |
| { | |
| "epoch": 12.0, | |
| "eval_loss": 0.6310720443725586, | |
| "eval_map": 0.4276, | |
| "eval_map_50": 0.584, | |
| "eval_map_75": 0.5016, | |
| "eval_map_chicken": 0.5447, | |
| "eval_map_duck": 0.0, | |
| "eval_map_large": 0.7945, | |
| "eval_map_medium": 0.4104, | |
| "eval_map_plant": 0.7382, | |
| "eval_map_small": 0.0749, | |
| "eval_mar_1": 0.1445, | |
| "eval_mar_10": 0.4989, | |
| "eval_mar_100": 0.5035, | |
| "eval_mar_100_chicken": 0.7133, | |
| "eval_mar_100_duck": 0.0, | |
| "eval_mar_100_plant": 0.7971, | |
| "eval_mar_large": 0.8515, | |
| "eval_mar_medium": 0.4876, | |
| "eval_mar_small": 0.1542, | |
| "eval_runtime": 5.0099, | |
| "eval_samples_per_second": 19.961, | |
| "eval_steps_per_second": 2.595, | |
| "step": 6000 | |
| }, | |
| { | |
| "epoch": 12.06, | |
| "grad_norm": 116.31084442138672, | |
| "learning_rate": 9.035832116232002e-06, | |
| "loss": 0.5482, | |
| "step": 6030 | |
| }, | |
| { | |
| "epoch": 12.12, | |
| "grad_norm": 65.95005798339844, | |
| "learning_rate": 9.026539428555609e-06, | |
| "loss": 0.6084, | |
| "step": 6060 | |
| }, | |
| { | |
| "epoch": 12.18, | |
| "grad_norm": 59.240360260009766, | |
| "learning_rate": 9.017207000560639e-06, | |
| "loss": 0.6436, | |
| "step": 6090 | |
| }, | |
| { | |
| "epoch": 12.24, | |
| "grad_norm": 117.28052520751953, | |
| "learning_rate": 9.007834924354384e-06, | |
| "loss": 0.7394, | |
| "step": 6120 | |
| }, | |
| { | |
| "epoch": 12.3, | |
| "grad_norm": 54.573299407958984, | |
| "learning_rate": 8.998423292435455e-06, | |
| "loss": 0.746, | |
| "step": 6150 | |
| }, | |
| { | |
| "epoch": 12.36, | |
| "grad_norm": 59.95296096801758, | |
| "learning_rate": 8.988972197692857e-06, | |
| "loss": 0.7718, | |
| "step": 6180 | |
| }, | |
| { | |
| "epoch": 12.42, | |
| "grad_norm": 27.571857452392578, | |
| "learning_rate": 8.97948173340508e-06, | |
| "loss": 0.7943, | |
| "step": 6210 | |
| }, | |
| { | |
| "epoch": 12.48, | |
| "grad_norm": 180.5658416748047, | |
| "learning_rate": 8.969951993239177e-06, | |
| "loss": 0.7434, | |
| "step": 6240 | |
| }, | |
| { | |
| "epoch": 12.54, | |
| "grad_norm": 89.40031433105469, | |
| "learning_rate": 8.960383071249837e-06, | |
| "loss": 0.6528, | |
| "step": 6270 | |
| }, | |
| { | |
| "epoch": 12.6, | |
| "grad_norm": 84.45803833007812, | |
| "learning_rate": 8.950775061878453e-06, | |
| "loss": 0.6857, | |
| "step": 6300 | |
| }, | |
| { | |
| "epoch": 12.66, | |
| "grad_norm": 73.63268280029297, | |
| "learning_rate": 8.9411280599522e-06, | |
| "loss": 0.7559, | |
| "step": 6330 | |
| }, | |
| { | |
| "epoch": 12.72, | |
| "grad_norm": 0.0037382396403700113, | |
| "learning_rate": 8.931442160683094e-06, | |
| "loss": 0.6612, | |
| "step": 6360 | |
| }, | |
| { | |
| "epoch": 12.78, | |
| "grad_norm": 120.78377532958984, | |
| "learning_rate": 8.921717459667052e-06, | |
| "loss": 0.7691, | |
| "step": 6390 | |
| }, | |
| { | |
| "epoch": 12.84, | |
| "grad_norm": 30.826618194580078, | |
| "learning_rate": 8.911954052882941e-06, | |
| "loss": 0.7987, | |
| "step": 6420 | |
| }, | |
| { | |
| "epoch": 12.9, | |
| "grad_norm": 40.72140884399414, | |
| "learning_rate": 8.902152036691649e-06, | |
| "loss": 0.8032, | |
| "step": 6450 | |
| }, | |
| { | |
| "epoch": 12.96, | |
| "grad_norm": 34.382293701171875, | |
| "learning_rate": 8.892311507835118e-06, | |
| "loss": 0.683, | |
| "step": 6480 | |
| }, | |
| { | |
| "epoch": 13.0, | |
| "eval_loss": 0.6243864893913269, | |
| "eval_map": 0.4371, | |
| "eval_map_50": 0.5962, | |
| "eval_map_75": 0.5288, | |
| "eval_map_chicken": 0.5784, | |
| "eval_map_duck": 0.0, | |
| "eval_map_large": 0.7935, | |
| "eval_map_medium": 0.4149, | |
| "eval_map_plant": 0.7329, | |
| "eval_map_small": 0.0877, | |
| "eval_mar_1": 0.1447, | |
| "eval_mar_10": 0.5002, | |
| "eval_mar_100": 0.5031, | |
| "eval_mar_100_chicken": 0.7244, | |
| "eval_mar_100_duck": 0.0, | |
| "eval_mar_100_plant": 0.7847, | |
| "eval_mar_large": 0.8427, | |
| "eval_mar_medium": 0.486, | |
| "eval_mar_small": 0.1312, | |
| "eval_runtime": 4.9066, | |
| "eval_samples_per_second": 20.381, | |
| "eval_steps_per_second": 2.649, | |
| "step": 6500 | |
| }, | |
| { | |
| "epoch": 13.02, | |
| "grad_norm": 49.30002975463867, | |
| "learning_rate": 8.882432563435394e-06, | |
| "loss": 0.7653, | |
| "step": 6510 | |
| }, | |
| { | |
| "epoch": 13.08, | |
| "grad_norm": 0.010997344739735126, | |
| "learning_rate": 8.872515300993669e-06, | |
| "loss": 0.6646, | |
| "step": 6540 | |
| }, | |
| { | |
| "epoch": 13.14, | |
| "grad_norm": 65.29388427734375, | |
| "learning_rate": 8.862559818389322e-06, | |
| "loss": 0.6705, | |
| "step": 6570 | |
| }, | |
| { | |
| "epoch": 13.2, | |
| "grad_norm": 56.85333251953125, | |
| "learning_rate": 8.852566213878947e-06, | |
| "loss": 0.7125, | |
| "step": 6600 | |
| }, | |
| { | |
| "epoch": 13.26, | |
| "grad_norm": 79.7491226196289, | |
| "learning_rate": 8.842534586095383e-06, | |
| "loss": 0.6535, | |
| "step": 6630 | |
| }, | |
| { | |
| "epoch": 13.32, | |
| "grad_norm": 59.6799430847168, | |
| "learning_rate": 8.83246503404675e-06, | |
| "loss": 0.6035, | |
| "step": 6660 | |
| }, | |
| { | |
| "epoch": 13.38, | |
| "grad_norm": 96.64356231689453, | |
| "learning_rate": 8.82235765711546e-06, | |
| "loss": 0.6654, | |
| "step": 6690 | |
| }, | |
| { | |
| "epoch": 13.44, | |
| "grad_norm": 91.82725524902344, | |
| "learning_rate": 8.81221255505724e-06, | |
| "loss": 0.6695, | |
| "step": 6720 | |
| }, | |
| { | |
| "epoch": 13.5, | |
| "grad_norm": 36.14399719238281, | |
| "learning_rate": 8.802029828000157e-06, | |
| "loss": 0.8221, | |
| "step": 6750 | |
| }, | |
| { | |
| "epoch": 13.56, | |
| "grad_norm": 46.07867431640625, | |
| "learning_rate": 8.791809576443611e-06, | |
| "loss": 0.7196, | |
| "step": 6780 | |
| }, | |
| { | |
| "epoch": 13.62, | |
| "grad_norm": 22.723224639892578, | |
| "learning_rate": 8.78155190125736e-06, | |
| "loss": 0.717, | |
| "step": 6810 | |
| }, | |
| { | |
| "epoch": 13.68, | |
| "grad_norm": 186.8100128173828, | |
| "learning_rate": 8.77125690368052e-06, | |
| "loss": 0.805, | |
| "step": 6840 | |
| }, | |
| { | |
| "epoch": 13.74, | |
| "grad_norm": 52.61324691772461, | |
| "learning_rate": 8.760924685320558e-06, | |
| "loss": 0.645, | |
| "step": 6870 | |
| }, | |
| { | |
| "epoch": 13.8, | |
| "grad_norm": 0.00226211897097528, | |
| "learning_rate": 8.750555348152299e-06, | |
| "loss": 0.6038, | |
| "step": 6900 | |
| }, | |
| { | |
| "epoch": 13.86, | |
| "grad_norm": 46.638126373291016, | |
| "learning_rate": 8.740148994516912e-06, | |
| "loss": 0.6805, | |
| "step": 6930 | |
| }, | |
| { | |
| "epoch": 13.92, | |
| "grad_norm": 32.8173942565918, | |
| "learning_rate": 8.729705727120911e-06, | |
| "loss": 0.6252, | |
| "step": 6960 | |
| }, | |
| { | |
| "epoch": 13.98, | |
| "grad_norm": 45.916847229003906, | |
| "learning_rate": 8.719225649035126e-06, | |
| "loss": 0.6541, | |
| "step": 6990 | |
| }, | |
| { | |
| "epoch": 14.0, | |
| "eval_loss": 0.554333508014679, | |
| "eval_map": 0.4719, | |
| "eval_map_50": 0.6191, | |
| "eval_map_75": 0.5555, | |
| "eval_map_chicken": 0.6567, | |
| "eval_map_duck": 0.0, | |
| "eval_map_large": 0.8231, | |
| "eval_map_medium": 0.4525, | |
| "eval_map_plant": 0.759, | |
| "eval_map_small": 0.0712, | |
| "eval_mar_1": 0.1494, | |
| "eval_mar_10": 0.5195, | |
| "eval_mar_100": 0.5249, | |
| "eval_mar_100_chicken": 0.7644, | |
| "eval_mar_100_duck": 0.0, | |
| "eval_mar_100_plant": 0.8104, | |
| "eval_mar_large": 0.8715, | |
| "eval_mar_medium": 0.5025, | |
| "eval_mar_small": 0.1896, | |
| "eval_runtime": 5.164, | |
| "eval_samples_per_second": 19.365, | |
| "eval_steps_per_second": 2.517, | |
| "step": 7000 | |
| }, | |
| { | |
| "epoch": 14.04, | |
| "grad_norm": 48.85868835449219, | |
| "learning_rate": 8.708708863693696e-06, | |
| "loss": 0.5773, | |
| "step": 7020 | |
| }, | |
| { | |
| "epoch": 14.1, | |
| "grad_norm": 443.4310607910156, | |
| "learning_rate": 8.69815547489305e-06, | |
| "loss": 0.6889, | |
| "step": 7050 | |
| }, | |
| { | |
| "epoch": 14.16, | |
| "grad_norm": 0.0019413733389228582, | |
| "learning_rate": 8.68756558679087e-06, | |
| "loss": 0.6352, | |
| "step": 7080 | |
| }, | |
| { | |
| "epoch": 14.22, | |
| "grad_norm": 66.77024841308594, | |
| "learning_rate": 8.67693930390508e-06, | |
| "loss": 0.9123, | |
| "step": 7110 | |
| }, | |
| { | |
| "epoch": 14.28, | |
| "grad_norm": 229.13998413085938, | |
| "learning_rate": 8.666276731112802e-06, | |
| "loss": 0.721, | |
| "step": 7140 | |
| }, | |
| { | |
| "epoch": 14.34, | |
| "grad_norm": 56.66972351074219, | |
| "learning_rate": 8.655577973649322e-06, | |
| "loss": 0.6267, | |
| "step": 7170 | |
| }, | |
| { | |
| "epoch": 14.4, | |
| "grad_norm": 54.61463928222656, | |
| "learning_rate": 8.644843137107058e-06, | |
| "loss": 0.725, | |
| "step": 7200 | |
| }, | |
| { | |
| "epoch": 14.46, | |
| "grad_norm": 40.25044250488281, | |
| "learning_rate": 8.634072327434515e-06, | |
| "loss": 0.5895, | |
| "step": 7230 | |
| }, | |
| { | |
| "epoch": 14.52, | |
| "grad_norm": 59.81513595581055, | |
| "learning_rate": 8.623265650935233e-06, | |
| "loss": 0.6714, | |
| "step": 7260 | |
| }, | |
| { | |
| "epoch": 14.58, | |
| "grad_norm": 49.08866882324219, | |
| "learning_rate": 8.612423214266749e-06, | |
| "loss": 0.6454, | |
| "step": 7290 | |
| }, | |
| { | |
| "epoch": 14.64, | |
| "grad_norm": 47.70410919189453, | |
| "learning_rate": 8.601545124439535e-06, | |
| "loss": 0.6701, | |
| "step": 7320 | |
| }, | |
| { | |
| "epoch": 14.7, | |
| "grad_norm": 50.41656494140625, | |
| "learning_rate": 8.590631488815945e-06, | |
| "loss": 0.6167, | |
| "step": 7350 | |
| }, | |
| { | |
| "epoch": 14.76, | |
| "grad_norm": 37.03781509399414, | |
| "learning_rate": 8.579682415109156e-06, | |
| "loss": 0.6049, | |
| "step": 7380 | |
| }, | |
| { | |
| "epoch": 14.82, | |
| "grad_norm": 59.566505432128906, | |
| "learning_rate": 8.568698011382108e-06, | |
| "loss": 0.8524, | |
| "step": 7410 | |
| }, | |
| { | |
| "epoch": 14.88, | |
| "grad_norm": 49.214229583740234, | |
| "learning_rate": 8.557678386046429e-06, | |
| "loss": 0.6073, | |
| "step": 7440 | |
| }, | |
| { | |
| "epoch": 14.94, | |
| "grad_norm": 54.74986267089844, | |
| "learning_rate": 8.54662364786137e-06, | |
| "loss": 0.7139, | |
| "step": 7470 | |
| }, | |
| { | |
| "epoch": 15.0, | |
| "grad_norm": 0.0027110481169074774, | |
| "learning_rate": 8.535533905932739e-06, | |
| "loss": 0.6219, | |
| "step": 7500 | |
| }, | |
| { | |
| "epoch": 15.0, | |
| "eval_loss": 0.5367693305015564, | |
| "eval_map": 0.4754, | |
| "eval_map_50": 0.6197, | |
| "eval_map_75": 0.5553, | |
| "eval_map_chicken": 0.6691, | |
| "eval_map_duck": 0.0, | |
| "eval_map_large": 0.825, | |
| "eval_map_medium": 0.4584, | |
| "eval_map_plant": 0.7571, | |
| "eval_map_small": 0.0764, | |
| "eval_mar_1": 0.1528, | |
| "eval_mar_10": 0.5216, | |
| "eval_mar_100": 0.5265, | |
| "eval_mar_100_chicken": 0.7724, | |
| "eval_mar_100_duck": 0.0, | |
| "eval_mar_100_plant": 0.8072, | |
| "eval_mar_large": 0.8711, | |
| "eval_mar_medium": 0.5038, | |
| "eval_mar_small": 0.1688, | |
| "eval_runtime": 5.029, | |
| "eval_samples_per_second": 19.885, | |
| "eval_steps_per_second": 2.585, | |
| "step": 7500 | |
| }, | |
| { | |
| "epoch": 15.06, | |
| "grad_norm": 101.3479995727539, | |
| "learning_rate": 8.524409269711808e-06, | |
| "loss": 0.6541, | |
| "step": 7530 | |
| }, | |
| { | |
| "epoch": 15.12, | |
| "grad_norm": 32.30617141723633, | |
| "learning_rate": 8.513249848994248e-06, | |
| "loss": 0.7145, | |
| "step": 7560 | |
| }, | |
| { | |
| "epoch": 15.18, | |
| "grad_norm": 118.48365783691406, | |
| "learning_rate": 8.502055753919033e-06, | |
| "loss": 0.6418, | |
| "step": 7590 | |
| }, | |
| { | |
| "epoch": 15.24, | |
| "grad_norm": 146.08511352539062, | |
| "learning_rate": 8.490827094967364e-06, | |
| "loss": 0.5815, | |
| "step": 7620 | |
| }, | |
| { | |
| "epoch": 15.3, | |
| "grad_norm": 74.51444244384766, | |
| "learning_rate": 8.479563982961572e-06, | |
| "loss": 0.6067, | |
| "step": 7650 | |
| }, | |
| { | |
| "epoch": 15.36, | |
| "grad_norm": 71.7569580078125, | |
| "learning_rate": 8.468266529064025e-06, | |
| "loss": 0.6352, | |
| "step": 7680 | |
| }, | |
| { | |
| "epoch": 15.42, | |
| "grad_norm": 63.288089752197266, | |
| "learning_rate": 8.456934844776033e-06, | |
| "loss": 0.7251, | |
| "step": 7710 | |
| }, | |
| { | |
| "epoch": 15.48, | |
| "grad_norm": 52.670902252197266, | |
| "learning_rate": 8.445569041936743e-06, | |
| "loss": 0.755, | |
| "step": 7740 | |
| }, | |
| { | |
| "epoch": 15.54, | |
| "grad_norm": 154.50787353515625, | |
| "learning_rate": 8.434169232722043e-06, | |
| "loss": 0.5829, | |
| "step": 7770 | |
| }, | |
| { | |
| "epoch": 15.6, | |
| "grad_norm": 72.39775085449219, | |
| "learning_rate": 8.422735529643445e-06, | |
| "loss": 0.6424, | |
| "step": 7800 | |
| }, | |
| { | |
| "epoch": 15.66, | |
| "grad_norm": 189.01527404785156, | |
| "learning_rate": 8.411268045546984e-06, | |
| "loss": 0.6285, | |
| "step": 7830 | |
| }, | |
| { | |
| "epoch": 15.72, | |
| "grad_norm": 113.38286590576172, | |
| "learning_rate": 8.399766893612096e-06, | |
| "loss": 0.6086, | |
| "step": 7860 | |
| }, | |
| { | |
| "epoch": 15.78, | |
| "grad_norm": 22.89063262939453, | |
| "learning_rate": 8.388232187350513e-06, | |
| "loss": 0.6352, | |
| "step": 7890 | |
| }, | |
| { | |
| "epoch": 15.84, | |
| "grad_norm": 29.94500160217285, | |
| "learning_rate": 8.376664040605122e-06, | |
| "loss": 0.7657, | |
| "step": 7920 | |
| }, | |
| { | |
| "epoch": 15.9, | |
| "grad_norm": 67.85516357421875, | |
| "learning_rate": 8.365062567548868e-06, | |
| "loss": 0.7562, | |
| "step": 7950 | |
| }, | |
| { | |
| "epoch": 15.96, | |
| "grad_norm": 173.4965057373047, | |
| "learning_rate": 8.353427882683601e-06, | |
| "loss": 0.5842, | |
| "step": 7980 | |
| }, | |
| { | |
| "epoch": 16.0, | |
| "eval_loss": 0.5324800610542297, | |
| "eval_map": 0.4778, | |
| "eval_map_50": 0.6269, | |
| "eval_map_75": 0.5668, | |
| "eval_map_chicken": 0.6922, | |
| "eval_map_duck": 0.0, | |
| "eval_map_large": 0.8015, | |
| "eval_map_medium": 0.4558, | |
| "eval_map_plant": 0.7412, | |
| "eval_map_small": 0.1147, | |
| "eval_mar_1": 0.1501, | |
| "eval_mar_10": 0.5178, | |
| "eval_mar_100": 0.5218, | |
| "eval_mar_100_chicken": 0.7636, | |
| "eval_mar_100_duck": 0.0, | |
| "eval_mar_100_plant": 0.8017, | |
| "eval_mar_large": 0.8556, | |
| "eval_mar_medium": 0.5071, | |
| "eval_mar_small": 0.1604, | |
| "eval_runtime": 5.0074, | |
| "eval_samples_per_second": 19.971, | |
| "eval_steps_per_second": 2.596, | |
| "step": 8000 | |
| }, | |
| { | |
| "epoch": 16.02, | |
| "grad_norm": 63.686378479003906, | |
| "learning_rate": 8.341760100838967e-06, | |
| "loss": 0.7119, | |
| "step": 8010 | |
| }, | |
| { | |
| "epoch": 16.08, | |
| "grad_norm": 34.64134216308594, | |
| "learning_rate": 8.33005933717126e-06, | |
| "loss": 0.6144, | |
| "step": 8040 | |
| }, | |
| { | |
| "epoch": 16.14, | |
| "grad_norm": 34.982269287109375, | |
| "learning_rate": 8.318325707162293e-06, | |
| "loss": 0.6278, | |
| "step": 8070 | |
| }, | |
| { | |
| "epoch": 16.2, | |
| "grad_norm": 53.49913024902344, | |
| "learning_rate": 8.30655932661826e-06, | |
| "loss": 0.6089, | |
| "step": 8100 | |
| }, | |
| { | |
| "epoch": 16.26, | |
| "grad_norm": 118.85201263427734, | |
| "learning_rate": 8.294760311668586e-06, | |
| "loss": 0.6401, | |
| "step": 8130 | |
| }, | |
| { | |
| "epoch": 16.32, | |
| "grad_norm": 35.8175048828125, | |
| "learning_rate": 8.282928778764783e-06, | |
| "loss": 0.5656, | |
| "step": 8160 | |
| }, | |
| { | |
| "epoch": 16.38, | |
| "grad_norm": 19.808303833007812, | |
| "learning_rate": 8.271064844679306e-06, | |
| "loss": 0.5821, | |
| "step": 8190 | |
| }, | |
| { | |
| "epoch": 16.44, | |
| "grad_norm": 62.134803771972656, | |
| "learning_rate": 8.259168626504395e-06, | |
| "loss": 0.6077, | |
| "step": 8220 | |
| }, | |
| { | |
| "epoch": 16.5, | |
| "grad_norm": 31.38833999633789, | |
| "learning_rate": 8.247240241650918e-06, | |
| "loss": 0.5492, | |
| "step": 8250 | |
| }, | |
| { | |
| "epoch": 16.56, | |
| "grad_norm": 77.07894897460938, | |
| "learning_rate": 8.235279807847223e-06, | |
| "loss": 0.641, | |
| "step": 8280 | |
| }, | |
| { | |
| "epoch": 16.62, | |
| "grad_norm": 65.0913314819336, | |
| "learning_rate": 8.223287443137957e-06, | |
| "loss": 0.5655, | |
| "step": 8310 | |
| }, | |
| { | |
| "epoch": 16.68, | |
| "grad_norm": 217.502685546875, | |
| "learning_rate": 8.211263265882923e-06, | |
| "loss": 0.6829, | |
| "step": 8340 | |
| }, | |
| { | |
| "epoch": 16.74, | |
| "grad_norm": 145.74627685546875, | |
| "learning_rate": 8.199207394755892e-06, | |
| "loss": 0.664, | |
| "step": 8370 | |
| }, | |
| { | |
| "epoch": 16.8, | |
| "grad_norm": 61.71194076538086, | |
| "learning_rate": 8.18711994874345e-06, | |
| "loss": 0.5375, | |
| "step": 8400 | |
| }, | |
| { | |
| "epoch": 16.86, | |
| "grad_norm": 74.38108825683594, | |
| "learning_rate": 8.175001047143804e-06, | |
| "loss": 0.6282, | |
| "step": 8430 | |
| }, | |
| { | |
| "epoch": 16.92, | |
| "grad_norm": 88.76374816894531, | |
| "learning_rate": 8.162850809565623e-06, | |
| "loss": 0.6561, | |
| "step": 8460 | |
| }, | |
| { | |
| "epoch": 16.98, | |
| "grad_norm": 82.03018188476562, | |
| "learning_rate": 8.150669355926848e-06, | |
| "loss": 0.5704, | |
| "step": 8490 | |
| }, | |
| { | |
| "epoch": 17.0, | |
| "eval_loss": 0.5436572432518005, | |
| "eval_map": 0.5192, | |
| "eval_map_50": 0.6982, | |
| "eval_map_75": 0.6149, | |
| "eval_map_chicken": 0.6772, | |
| "eval_map_duck": 0.1347, | |
| "eval_map_large": 0.8084, | |
| "eval_map_medium": 0.5014, | |
| "eval_map_plant": 0.7456, | |
| "eval_map_small": 0.0616, | |
| "eval_mar_1": 0.1798, | |
| "eval_mar_10": 0.558, | |
| "eval_mar_100": 0.5618, | |
| "eval_mar_100_chicken": 0.7449, | |
| "eval_mar_100_duck": 0.1412, | |
| "eval_mar_100_plant": 0.7994, | |
| "eval_mar_large": 0.8644, | |
| "eval_mar_medium": 0.5445, | |
| "eval_mar_small": 0.1521, | |
| "eval_runtime": 5.0124, | |
| "eval_samples_per_second": 19.95, | |
| "eval_steps_per_second": 2.594, | |
| "step": 8500 | |
| }, | |
| { | |
| "epoch": 17.04, | |
| "grad_norm": 106.79338073730469, | |
| "learning_rate": 8.138456806453503e-06, | |
| "loss": 0.7365, | |
| "step": 8520 | |
| }, | |
| { | |
| "epoch": 17.1, | |
| "grad_norm": 79.06564331054688, | |
| "learning_rate": 8.126213281678527e-06, | |
| "loss": 0.6555, | |
| "step": 8550 | |
| }, | |
| { | |
| "epoch": 17.16, | |
| "grad_norm": 31.134279251098633, | |
| "learning_rate": 8.113938902440563e-06, | |
| "loss": 0.5704, | |
| "step": 8580 | |
| }, | |
| { | |
| "epoch": 17.22, | |
| "grad_norm": 52.495059967041016, | |
| "learning_rate": 8.101633789882781e-06, | |
| "loss": 0.6709, | |
| "step": 8610 | |
| }, | |
| { | |
| "epoch": 17.28, | |
| "grad_norm": 45.442081451416016, | |
| "learning_rate": 8.089298065451673e-06, | |
| "loss": 0.5955, | |
| "step": 8640 | |
| }, | |
| { | |
| "epoch": 17.34, | |
| "grad_norm": 71.86921691894531, | |
| "learning_rate": 8.076931850895858e-06, | |
| "loss": 0.6332, | |
| "step": 8670 | |
| }, | |
| { | |
| "epoch": 17.4, | |
| "grad_norm": 61.51048278808594, | |
| "learning_rate": 8.064535268264883e-06, | |
| "loss": 0.6525, | |
| "step": 8700 | |
| }, | |
| { | |
| "epoch": 17.46, | |
| "grad_norm": 71.84506225585938, | |
| "learning_rate": 8.052108439908014e-06, | |
| "loss": 0.5134, | |
| "step": 8730 | |
| }, | |
| { | |
| "epoch": 17.52, | |
| "grad_norm": 58.42548370361328, | |
| "learning_rate": 8.039651488473028e-06, | |
| "loss": 0.6082, | |
| "step": 8760 | |
| }, | |
| { | |
| "epoch": 17.58, | |
| "grad_norm": 58.07726287841797, | |
| "learning_rate": 8.027164536905008e-06, | |
| "loss": 0.581, | |
| "step": 8790 | |
| }, | |
| { | |
| "epoch": 17.64, | |
| "grad_norm": 167.6965789794922, | |
| "learning_rate": 8.014647708445124e-06, | |
| "loss": 0.5426, | |
| "step": 8820 | |
| }, | |
| { | |
| "epoch": 17.7, | |
| "grad_norm": 37.38274383544922, | |
| "learning_rate": 8.002101126629422e-06, | |
| "loss": 0.5913, | |
| "step": 8850 | |
| }, | |
| { | |
| "epoch": 17.76, | |
| "grad_norm": 63.516876220703125, | |
| "learning_rate": 7.989524915287595e-06, | |
| "loss": 0.6801, | |
| "step": 8880 | |
| }, | |
| { | |
| "epoch": 17.82, | |
| "grad_norm": 48.16976547241211, | |
| "learning_rate": 7.976919198541775e-06, | |
| "loss": 0.5655, | |
| "step": 8910 | |
| }, | |
| { | |
| "epoch": 17.88, | |
| "grad_norm": 85.60452270507812, | |
| "learning_rate": 7.964284100805297e-06, | |
| "loss": 0.6774, | |
| "step": 8940 | |
| }, | |
| { | |
| "epoch": 17.94, | |
| "grad_norm": 56.792022705078125, | |
| "learning_rate": 7.951619746781474e-06, | |
| "loss": 0.6201, | |
| "step": 8970 | |
| }, | |
| { | |
| "epoch": 18.0, | |
| "grad_norm": 60.09886932373047, | |
| "learning_rate": 7.938926261462366e-06, | |
| "loss": 0.5683, | |
| "step": 9000 | |
| }, | |
| { | |
| "epoch": 18.0, | |
| "eval_loss": 0.5067680478096008, | |
| "eval_map": 0.6324, | |
| "eval_map_50": 0.8451, | |
| "eval_map_75": 0.7659, | |
| "eval_map_chicken": 0.6996, | |
| "eval_map_duck": 0.4404, | |
| "eval_map_large": 0.8208, | |
| "eval_map_medium": 0.6253, | |
| "eval_map_plant": 0.7573, | |
| "eval_map_small": 0.0963, | |
| "eval_mar_1": 0.225, | |
| "eval_mar_10": 0.6739, | |
| "eval_mar_100": 0.6793, | |
| "eval_mar_100_chicken": 0.7573, | |
| "eval_mar_100_duck": 0.4753, | |
| "eval_mar_100_plant": 0.8052, | |
| "eval_mar_large": 0.8628, | |
| "eval_mar_medium": 0.6808, | |
| "eval_mar_small": 0.175, | |
| "eval_runtime": 5.0071, | |
| "eval_samples_per_second": 19.971, | |
| "eval_steps_per_second": 2.596, | |
| "step": 9000 | |
| }, | |
| { | |
| "epoch": 18.06, | |
| "grad_norm": 66.6285400390625, | |
| "learning_rate": 7.926203770127552e-06, | |
| "loss": 0.5579, | |
| "step": 9030 | |
| }, | |
| { | |
| "epoch": 18.12, | |
| "grad_norm": 127.3472671508789, | |
| "learning_rate": 7.913452398342882e-06, | |
| "loss": 0.5633, | |
| "step": 9060 | |
| }, | |
| { | |
| "epoch": 18.18, | |
| "grad_norm": 70.93628692626953, | |
| "learning_rate": 7.900672271959247e-06, | |
| "loss": 0.6412, | |
| "step": 9090 | |
| }, | |
| { | |
| "epoch": 18.24, | |
| "grad_norm": 81.13060760498047, | |
| "learning_rate": 7.887863517111337e-06, | |
| "loss": 0.6309, | |
| "step": 9120 | |
| }, | |
| { | |
| "epoch": 18.3, | |
| "grad_norm": 66.56903839111328, | |
| "learning_rate": 7.875026260216395e-06, | |
| "loss": 0.5713, | |
| "step": 9150 | |
| }, | |
| { | |
| "epoch": 18.36, | |
| "grad_norm": 49.273643493652344, | |
| "learning_rate": 7.862160627972956e-06, | |
| "loss": 0.5789, | |
| "step": 9180 | |
| }, | |
| { | |
| "epoch": 18.42, | |
| "grad_norm": 36.204036712646484, | |
| "learning_rate": 7.849266747359619e-06, | |
| "loss": 0.6162, | |
| "step": 9210 | |
| }, | |
| { | |
| "epoch": 18.48, | |
| "grad_norm": 55.067718505859375, | |
| "learning_rate": 7.836344745633785e-06, | |
| "loss": 0.6287, | |
| "step": 9240 | |
| }, | |
| { | |
| "epoch": 18.54, | |
| "grad_norm": 96.10171508789062, | |
| "learning_rate": 7.823394750330386e-06, | |
| "loss": 0.6219, | |
| "step": 9270 | |
| }, | |
| { | |
| "epoch": 18.6, | |
| "grad_norm": 47.33934783935547, | |
| "learning_rate": 7.810416889260653e-06, | |
| "loss": 0.7037, | |
| "step": 9300 | |
| }, | |
| { | |
| "epoch": 18.66, | |
| "grad_norm": 97.90437316894531, | |
| "learning_rate": 7.797411290510836e-06, | |
| "loss": 0.5965, | |
| "step": 9330 | |
| }, | |
| { | |
| "epoch": 18.72, | |
| "grad_norm": 36.546260833740234, | |
| "learning_rate": 7.78437808244094e-06, | |
| "loss": 0.5984, | |
| "step": 9360 | |
| }, | |
| { | |
| "epoch": 18.78, | |
| "grad_norm": 46.64151382446289, | |
| "learning_rate": 7.771317393683471e-06, | |
| "loss": 0.532, | |
| "step": 9390 | |
| }, | |
| { | |
| "epoch": 18.84, | |
| "grad_norm": 102.95185852050781, | |
| "learning_rate": 7.758229353142153e-06, | |
| "loss": 0.6825, | |
| "step": 9420 | |
| }, | |
| { | |
| "epoch": 18.9, | |
| "grad_norm": 53.53539276123047, | |
| "learning_rate": 7.74511408999066e-06, | |
| "loss": 0.6691, | |
| "step": 9450 | |
| }, | |
| { | |
| "epoch": 18.96, | |
| "grad_norm": 83.08556365966797, | |
| "learning_rate": 7.731971733671347e-06, | |
| "loss": 0.6402, | |
| "step": 9480 | |
| }, | |
| { | |
| "epoch": 19.0, | |
| "eval_loss": 0.46823224425315857, | |
| "eval_map": 0.6741, | |
| "eval_map_50": 0.8823, | |
| "eval_map_75": 0.8298, | |
| "eval_map_chicken": 0.7195, | |
| "eval_map_duck": 0.5335, | |
| "eval_map_large": 0.8274, | |
| "eval_map_medium": 0.6748, | |
| "eval_map_plant": 0.7691, | |
| "eval_map_small": 0.1357, | |
| "eval_mar_1": 0.2516, | |
| "eval_mar_10": 0.7135, | |
| "eval_mar_100": 0.7185, | |
| "eval_mar_100_chicken": 0.7698, | |
| "eval_mar_100_duck": 0.567, | |
| "eval_mar_100_plant": 0.8187, | |
| "eval_mar_large": 0.8728, | |
| "eval_mar_medium": 0.7246, | |
| "eval_mar_small": 0.2104, | |
| "eval_runtime": 5.1062, | |
| "eval_samples_per_second": 19.584, | |
| "eval_steps_per_second": 2.546, | |
| "step": 9500 | |
| }, | |
| { | |
| "epoch": 19.02, | |
| "grad_norm": 54.46965408325195, | |
| "learning_rate": 7.718802413893963e-06, | |
| "loss": 0.6762, | |
| "step": 9510 | |
| }, | |
| { | |
| "epoch": 19.08, | |
| "grad_norm": 101.3377685546875, | |
| "learning_rate": 7.70560626063438e-06, | |
| "loss": 0.5893, | |
| "step": 9540 | |
| }, | |
| { | |
| "epoch": 19.14, | |
| "grad_norm": 54.880916595458984, | |
| "learning_rate": 7.692383404133302e-06, | |
| "loss": 0.5074, | |
| "step": 9570 | |
| }, | |
| { | |
| "epoch": 19.2, | |
| "grad_norm": 60.68124771118164, | |
| "learning_rate": 7.679133974894984e-06, | |
| "loss": 0.5336, | |
| "step": 9600 | |
| }, | |
| { | |
| "epoch": 19.26, | |
| "grad_norm": 26.069740295410156, | |
| "learning_rate": 7.665858103685944e-06, | |
| "loss": 0.5209, | |
| "step": 9630 | |
| }, | |
| { | |
| "epoch": 19.32, | |
| "grad_norm": 79.79200744628906, | |
| "learning_rate": 7.652555921533671e-06, | |
| "loss": 0.5702, | |
| "step": 9660 | |
| }, | |
| { | |
| "epoch": 19.38, | |
| "grad_norm": 208.93028259277344, | |
| "learning_rate": 7.639227559725333e-06, | |
| "loss": 0.6033, | |
| "step": 9690 | |
| }, | |
| { | |
| "epoch": 19.44, | |
| "grad_norm": 32.772396087646484, | |
| "learning_rate": 7.6258731498064796e-06, | |
| "loss": 0.5473, | |
| "step": 9720 | |
| }, | |
| { | |
| "epoch": 19.5, | |
| "grad_norm": 62.062557220458984, | |
| "learning_rate": 7.612492823579744e-06, | |
| "loss": 0.6168, | |
| "step": 9750 | |
| }, | |
| { | |
| "epoch": 19.56, | |
| "grad_norm": 46.67622375488281, | |
| "learning_rate": 7.5990867131035474e-06, | |
| "loss": 0.5714, | |
| "step": 9780 | |
| }, | |
| { | |
| "epoch": 19.62, | |
| "grad_norm": 58.50261306762695, | |
| "learning_rate": 7.585654950690786e-06, | |
| "loss": 0.5411, | |
| "step": 9810 | |
| }, | |
| { | |
| "epoch": 19.68, | |
| "grad_norm": 0.0009975264547392726, | |
| "learning_rate": 7.572197668907533e-06, | |
| "loss": 0.5943, | |
| "step": 9840 | |
| }, | |
| { | |
| "epoch": 19.74, | |
| "grad_norm": 37.597755432128906, | |
| "learning_rate": 7.5587150005717256e-06, | |
| "loss": 0.6523, | |
| "step": 9870 | |
| }, | |
| { | |
| "epoch": 19.8, | |
| "grad_norm": 46.372833251953125, | |
| "learning_rate": 7.545207078751858e-06, | |
| "loss": 0.5803, | |
| "step": 9900 | |
| }, | |
| { | |
| "epoch": 19.86, | |
| "grad_norm": 53.04841995239258, | |
| "learning_rate": 7.531674036765662e-06, | |
| "loss": 0.6232, | |
| "step": 9930 | |
| }, | |
| { | |
| "epoch": 19.92, | |
| "grad_norm": 49.54397964477539, | |
| "learning_rate": 7.518116008178805e-06, | |
| "loss": 0.5833, | |
| "step": 9960 | |
| }, | |
| { | |
| "epoch": 19.98, | |
| "grad_norm": 93.43566131591797, | |
| "learning_rate": 7.5045331268035505e-06, | |
| "loss": 0.5664, | |
| "step": 9990 | |
| }, | |
| { | |
| "epoch": 20.0, | |
| "eval_loss": 0.4792620539665222, | |
| "eval_map": 0.6841, | |
| "eval_map_50": 0.9057, | |
| "eval_map_75": 0.8277, | |
| "eval_map_chicken": 0.7325, | |
| "eval_map_duck": 0.5558, | |
| "eval_map_large": 0.8164, | |
| "eval_map_medium": 0.6878, | |
| "eval_map_plant": 0.7638, | |
| "eval_map_small": 0.135, | |
| "eval_mar_1": 0.2585, | |
| "eval_mar_10": 0.7299, | |
| "eval_mar_100": 0.7341, | |
| "eval_mar_100_chicken": 0.7853, | |
| "eval_mar_100_duck": 0.5979, | |
| "eval_mar_100_plant": 0.819, | |
| "eval_mar_large": 0.8649, | |
| "eval_mar_medium": 0.7463, | |
| "eval_mar_small": 0.2396, | |
| "eval_runtime": 5.1021, | |
| "eval_samples_per_second": 19.6, | |
| "eval_steps_per_second": 2.548, | |
| "step": 10000 | |
| }, | |
| { | |
| "epoch": 20.04, | |
| "grad_norm": 97.10029602050781, | |
| "learning_rate": 7.490925526697455e-06, | |
| "loss": 0.5375, | |
| "step": 10020 | |
| }, | |
| { | |
| "epoch": 20.1, | |
| "grad_norm": 34.54094696044922, | |
| "learning_rate": 7.477293342162038e-06, | |
| "loss": 0.5242, | |
| "step": 10050 | |
| }, | |
| { | |
| "epoch": 20.16, | |
| "grad_norm": 74.18828582763672, | |
| "learning_rate": 7.463636707741458e-06, | |
| "loss": 0.4776, | |
| "step": 10080 | |
| }, | |
| { | |
| "epoch": 20.22, | |
| "grad_norm": 61.5256233215332, | |
| "learning_rate": 7.449955758221184e-06, | |
| "loss": 0.5529, | |
| "step": 10110 | |
| }, | |
| { | |
| "epoch": 20.28, | |
| "grad_norm": 390.4728088378906, | |
| "learning_rate": 7.436250628626662e-06, | |
| "loss": 0.6218, | |
| "step": 10140 | |
| }, | |
| { | |
| "epoch": 20.34, | |
| "grad_norm": 68.62318420410156, | |
| "learning_rate": 7.42252145422199e-06, | |
| "loss": 0.5621, | |
| "step": 10170 | |
| }, | |
| { | |
| "epoch": 20.4, | |
| "grad_norm": 44.162994384765625, | |
| "learning_rate": 7.408768370508577e-06, | |
| "loss": 0.5153, | |
| "step": 10200 | |
| }, | |
| { | |
| "epoch": 20.46, | |
| "grad_norm": 65.52782440185547, | |
| "learning_rate": 7.394991513223806e-06, | |
| "loss": 0.5388, | |
| "step": 10230 | |
| }, | |
| { | |
| "epoch": 20.52, | |
| "grad_norm": 31.844987869262695, | |
| "learning_rate": 7.381191018339697e-06, | |
| "loss": 0.4896, | |
| "step": 10260 | |
| }, | |
| { | |
| "epoch": 20.58, | |
| "grad_norm": 53.01321792602539, | |
| "learning_rate": 7.3673670220615615e-06, | |
| "loss": 0.5244, | |
| "step": 10290 | |
| }, | |
| { | |
| "epoch": 20.64, | |
| "grad_norm": 47.084693908691406, | |
| "learning_rate": 7.353519660826665e-06, | |
| "loss": 0.9178, | |
| "step": 10320 | |
| }, | |
| { | |
| "epoch": 20.7, | |
| "grad_norm": 91.9738540649414, | |
| "learning_rate": 7.3396490713028674e-06, | |
| "loss": 0.5282, | |
| "step": 10350 | |
| }, | |
| { | |
| "epoch": 20.76, | |
| "grad_norm": 79.40870666503906, | |
| "learning_rate": 7.325755390387293e-06, | |
| "loss": 0.5952, | |
| "step": 10380 | |
| }, | |
| { | |
| "epoch": 20.82, | |
| "grad_norm": 91.8105239868164, | |
| "learning_rate": 7.31183875520496e-06, | |
| "loss": 0.6473, | |
| "step": 10410 | |
| }, | |
| { | |
| "epoch": 20.88, | |
| "grad_norm": 51.59716796875, | |
| "learning_rate": 7.297899303107441e-06, | |
| "loss": 0.5078, | |
| "step": 10440 | |
| }, | |
| { | |
| "epoch": 20.94, | |
| "grad_norm": 66.34852600097656, | |
| "learning_rate": 7.283937171671498e-06, | |
| "loss": 0.4993, | |
| "step": 10470 | |
| }, | |
| { | |
| "epoch": 21.0, | |
| "grad_norm": 51.79935836791992, | |
| "learning_rate": 7.269952498697734e-06, | |
| "loss": 0.4411, | |
| "step": 10500 | |
| }, | |
| { | |
| "epoch": 21.0, | |
| "eval_loss": 0.44476136565208435, | |
| "eval_map": 0.7042, | |
| "eval_map_50": 0.932, | |
| "eval_map_75": 0.8592, | |
| "eval_map_chicken": 0.7128, | |
| "eval_map_duck": 0.6338, | |
| "eval_map_large": 0.8287, | |
| "eval_map_medium": 0.7039, | |
| "eval_map_plant": 0.766, | |
| "eval_map_small": 0.1098, | |
| "eval_mar_1": 0.2789, | |
| "eval_mar_10": 0.7527, | |
| "eval_mar_100": 0.7568, | |
| "eval_mar_100_chicken": 0.7658, | |
| "eval_mar_100_duck": 0.6845, | |
| "eval_mar_100_plant": 0.8202, | |
| "eval_mar_large": 0.8749, | |
| "eval_mar_medium": 0.7703, | |
| "eval_mar_small": 0.1718, | |
| "eval_runtime": 5.4795, | |
| "eval_samples_per_second": 18.25, | |
| "eval_steps_per_second": 2.372, | |
| "step": 10500 | |
| }, | |
| { | |
| "epoch": 21.06, | |
| "grad_norm": 52.99646759033203, | |
| "learning_rate": 7.2559454222092265e-06, | |
| "loss": 0.5123, | |
| "step": 10530 | |
| }, | |
| { | |
| "epoch": 21.12, | |
| "grad_norm": 35.29424285888672, | |
| "learning_rate": 7.241916080450163e-06, | |
| "loss": 0.5506, | |
| "step": 10560 | |
| }, | |
| { | |
| "epoch": 21.18, | |
| "grad_norm": 43.816009521484375, | |
| "learning_rate": 7.227864611884483e-06, | |
| "loss": 0.5644, | |
| "step": 10590 | |
| }, | |
| { | |
| "epoch": 21.24, | |
| "grad_norm": 106.69770812988281, | |
| "learning_rate": 7.21379115519451e-06, | |
| "loss": 0.5914, | |
| "step": 10620 | |
| }, | |
| { | |
| "epoch": 21.3, | |
| "grad_norm": 714.3410034179688, | |
| "learning_rate": 7.199695849279576e-06, | |
| "loss": 0.4671, | |
| "step": 10650 | |
| }, | |
| { | |
| "epoch": 21.36, | |
| "grad_norm": 40.27183532714844, | |
| "learning_rate": 7.185578833254665e-06, | |
| "loss": 0.5317, | |
| "step": 10680 | |
| }, | |
| { | |
| "epoch": 21.42, | |
| "grad_norm": 63.30058670043945, | |
| "learning_rate": 7.171440246449024e-06, | |
| "loss": 0.5063, | |
| "step": 10710 | |
| }, | |
| { | |
| "epoch": 21.48, | |
| "grad_norm": 31.39018440246582, | |
| "learning_rate": 7.157280228404796e-06, | |
| "loss": 0.5745, | |
| "step": 10740 | |
| }, | |
| { | |
| "epoch": 21.54, | |
| "grad_norm": 41.66950225830078, | |
| "learning_rate": 7.143098918875643e-06, | |
| "loss": 0.5368, | |
| "step": 10770 | |
| }, | |
| { | |
| "epoch": 21.6, | |
| "grad_norm": 75.50695037841797, | |
| "learning_rate": 7.128896457825364e-06, | |
| "loss": 0.6055, | |
| "step": 10800 | |
| }, | |
| { | |
| "epoch": 21.66, | |
| "grad_norm": 43.1558837890625, | |
| "learning_rate": 7.114672985426516e-06, | |
| "loss": 0.5593, | |
| "step": 10830 | |
| }, | |
| { | |
| "epoch": 21.72, | |
| "grad_norm": 39.317481994628906, | |
| "learning_rate": 7.100428642059033e-06, | |
| "loss": 0.5486, | |
| "step": 10860 | |
| }, | |
| { | |
| "epoch": 21.78, | |
| "grad_norm": 47.37519073486328, | |
| "learning_rate": 7.086163568308828e-06, | |
| "loss": 0.5348, | |
| "step": 10890 | |
| }, | |
| { | |
| "epoch": 21.84, | |
| "grad_norm": 109.30632781982422, | |
| "learning_rate": 7.071877904966422e-06, | |
| "loss": 0.5404, | |
| "step": 10920 | |
| }, | |
| { | |
| "epoch": 21.9, | |
| "grad_norm": 57.96327209472656, | |
| "learning_rate": 7.057571793025545e-06, | |
| "loss": 0.5283, | |
| "step": 10950 | |
| }, | |
| { | |
| "epoch": 21.96, | |
| "grad_norm": 97.8825454711914, | |
| "learning_rate": 7.043245373681746e-06, | |
| "loss": 0.6106, | |
| "step": 10980 | |
| }, | |
| { | |
| "epoch": 22.0, | |
| "eval_loss": 0.41423293948173523, | |
| "eval_map": 0.7307, | |
| "eval_map_50": 0.9307, | |
| "eval_map_75": 0.8797, | |
| "eval_map_chicken": 0.7379, | |
| "eval_map_duck": 0.6726, | |
| "eval_map_large": 0.8381, | |
| "eval_map_medium": 0.735, | |
| "eval_map_plant": 0.7817, | |
| "eval_map_small": 0.0773, | |
| "eval_mar_1": 0.2841, | |
| "eval_mar_10": 0.7736, | |
| "eval_mar_100": 0.7783, | |
| "eval_mar_100_chicken": 0.7853, | |
| "eval_mar_100_duck": 0.7134, | |
| "eval_mar_100_plant": 0.836, | |
| "eval_mar_large": 0.8866, | |
| "eval_mar_medium": 0.7946, | |
| "eval_mar_small": 0.2062, | |
| "eval_runtime": 5.0811, | |
| "eval_samples_per_second": 19.681, | |
| "eval_steps_per_second": 2.559, | |
| "step": 11000 | |
| }, | |
| { | |
| "epoch": 22.02, | |
| "grad_norm": 94.26522827148438, | |
| "learning_rate": 7.028898788331e-06, | |
| "loss": 0.5071, | |
| "step": 11010 | |
| }, | |
| { | |
| "epoch": 22.08, | |
| "grad_norm": 105.05602264404297, | |
| "learning_rate": 7.014532178568314e-06, | |
| "loss": 0.465, | |
| "step": 11040 | |
| }, | |
| { | |
| "epoch": 22.14, | |
| "grad_norm": 51.92534255981445, | |
| "learning_rate": 7.0001456861863236e-06, | |
| "loss": 0.5546, | |
| "step": 11070 | |
| }, | |
| { | |
| "epoch": 22.2, | |
| "grad_norm": 29.79014778137207, | |
| "learning_rate": 6.985739453173903e-06, | |
| "loss": 0.4825, | |
| "step": 11100 | |
| }, | |
| { | |
| "epoch": 22.26, | |
| "grad_norm": 80.56990814208984, | |
| "learning_rate": 6.971313621714756e-06, | |
| "loss": 0.5229, | |
| "step": 11130 | |
| }, | |
| { | |
| "epoch": 22.32, | |
| "grad_norm": 68.52033996582031, | |
| "learning_rate": 6.9568683341860135e-06, | |
| "loss": 0.5771, | |
| "step": 11160 | |
| }, | |
| { | |
| "epoch": 22.38, | |
| "grad_norm": 42.956085205078125, | |
| "learning_rate": 6.942403733156832e-06, | |
| "loss": 0.5977, | |
| "step": 11190 | |
| }, | |
| { | |
| "epoch": 22.44, | |
| "grad_norm": 62.62834548950195, | |
| "learning_rate": 6.927919961386984e-06, | |
| "loss": 0.5419, | |
| "step": 11220 | |
| }, | |
| { | |
| "epoch": 22.5, | |
| "grad_norm": 39.22291564941406, | |
| "learning_rate": 6.913417161825449e-06, | |
| "loss": 0.5346, | |
| "step": 11250 | |
| }, | |
| { | |
| "epoch": 22.56, | |
| "grad_norm": 21.985593795776367, | |
| "learning_rate": 6.898895477609007e-06, | |
| "loss": 0.5946, | |
| "step": 11280 | |
| }, | |
| { | |
| "epoch": 22.62, | |
| "grad_norm": 77.47394561767578, | |
| "learning_rate": 6.884355052060814e-06, | |
| "loss": 0.5351, | |
| "step": 11310 | |
| }, | |
| { | |
| "epoch": 22.68, | |
| "grad_norm": 40.48210906982422, | |
| "learning_rate": 6.869796028689002e-06, | |
| "loss": 0.5245, | |
| "step": 11340 | |
| }, | |
| { | |
| "epoch": 22.74, | |
| "grad_norm": 33.53753662109375, | |
| "learning_rate": 6.8552185511852555e-06, | |
| "loss": 0.4837, | |
| "step": 11370 | |
| }, | |
| { | |
| "epoch": 22.8, | |
| "grad_norm": 68.7355728149414, | |
| "learning_rate": 6.840622763423391e-06, | |
| "loss": 0.6208, | |
| "step": 11400 | |
| }, | |
| { | |
| "epoch": 22.86, | |
| "grad_norm": 71.2813720703125, | |
| "learning_rate": 6.82600880945794e-06, | |
| "loss": 0.6104, | |
| "step": 11430 | |
| }, | |
| { | |
| "epoch": 22.92, | |
| "grad_norm": 147.4709930419922, | |
| "learning_rate": 6.811376833522729e-06, | |
| "loss": 0.4831, | |
| "step": 11460 | |
| }, | |
| { | |
| "epoch": 22.98, | |
| "grad_norm": 79.54974365234375, | |
| "learning_rate": 6.796726980029454e-06, | |
| "loss": 0.5243, | |
| "step": 11490 | |
| }, | |
| { | |
| "epoch": 23.0, | |
| "eval_loss": 0.43533027172088623, | |
| "eval_map": 0.7183, | |
| "eval_map_50": 0.9406, | |
| "eval_map_75": 0.86, | |
| "eval_map_chicken": 0.7338, | |
| "eval_map_duck": 0.6385, | |
| "eval_map_large": 0.8416, | |
| "eval_map_medium": 0.7236, | |
| "eval_map_plant": 0.7827, | |
| "eval_map_small": 0.0901, | |
| "eval_mar_1": 0.2827, | |
| "eval_mar_10": 0.7615, | |
| "eval_mar_100": 0.767, | |
| "eval_mar_100_chicken": 0.7827, | |
| "eval_mar_100_duck": 0.6845, | |
| "eval_mar_100_plant": 0.8337, | |
| "eval_mar_large": 0.8879, | |
| "eval_mar_medium": 0.779, | |
| "eval_mar_small": 0.1973, | |
| "eval_runtime": 5.0735, | |
| "eval_samples_per_second": 19.71, | |
| "eval_steps_per_second": 2.562, | |
| "step": 11500 | |
| }, | |
| { | |
| "epoch": 23.04, | |
| "grad_norm": 42.534297943115234, | |
| "learning_rate": 6.782059393566254e-06, | |
| "loss": 0.4992, | |
| "step": 11520 | |
| }, | |
| { | |
| "epoch": 23.1, | |
| "grad_norm": 0.002314250450581312, | |
| "learning_rate": 6.767374218896286e-06, | |
| "loss": 0.474, | |
| "step": 11550 | |
| }, | |
| { | |
| "epoch": 23.16, | |
| "grad_norm": 33.95319366455078, | |
| "learning_rate": 6.752671600956295e-06, | |
| "loss": 0.5098, | |
| "step": 11580 | |
| }, | |
| { | |
| "epoch": 23.22, | |
| "grad_norm": 65.58226013183594, | |
| "learning_rate": 6.737951684855185e-06, | |
| "loss": 0.6279, | |
| "step": 11610 | |
| }, | |
| { | |
| "epoch": 23.28, | |
| "grad_norm": 38.85221481323242, | |
| "learning_rate": 6.723214615872585e-06, | |
| "loss": 0.5288, | |
| "step": 11640 | |
| }, | |
| { | |
| "epoch": 23.34, | |
| "grad_norm": 31.03350067138672, | |
| "learning_rate": 6.708460539457418e-06, | |
| "loss": 0.5167, | |
| "step": 11670 | |
| }, | |
| { | |
| "epoch": 23.4, | |
| "grad_norm": 74.7662353515625, | |
| "learning_rate": 6.693689601226458e-06, | |
| "loss": 0.5782, | |
| "step": 11700 | |
| }, | |
| { | |
| "epoch": 23.46, | |
| "grad_norm": 30.249208450317383, | |
| "learning_rate": 6.6789019469629034e-06, | |
| "loss": 0.4759, | |
| "step": 11730 | |
| }, | |
| { | |
| "epoch": 23.52, | |
| "grad_norm": 32.10087966918945, | |
| "learning_rate": 6.664097722614934e-06, | |
| "loss": 0.4923, | |
| "step": 11760 | |
| }, | |
| { | |
| "epoch": 23.58, | |
| "grad_norm": 129.60997009277344, | |
| "learning_rate": 6.649277074294265e-06, | |
| "loss": 0.5149, | |
| "step": 11790 | |
| }, | |
| { | |
| "epoch": 23.64, | |
| "grad_norm": 74.41838073730469, | |
| "learning_rate": 6.634440148274712e-06, | |
| "loss": 0.5063, | |
| "step": 11820 | |
| }, | |
| { | |
| "epoch": 23.7, | |
| "grad_norm": 37.93177032470703, | |
| "learning_rate": 6.619587090990748e-06, | |
| "loss": 0.4632, | |
| "step": 11850 | |
| }, | |
| { | |
| "epoch": 23.76, | |
| "grad_norm": 40.77086639404297, | |
| "learning_rate": 6.604718049036047e-06, | |
| "loss": 0.5137, | |
| "step": 11880 | |
| }, | |
| { | |
| "epoch": 23.82, | |
| "grad_norm": 29.192161560058594, | |
| "learning_rate": 6.589833169162055e-06, | |
| "loss": 0.5527, | |
| "step": 11910 | |
| }, | |
| { | |
| "epoch": 23.88, | |
| "grad_norm": 86.26991271972656, | |
| "learning_rate": 6.574932598276524e-06, | |
| "loss": 0.4662, | |
| "step": 11940 | |
| }, | |
| { | |
| "epoch": 23.94, | |
| "grad_norm": 35.82132339477539, | |
| "learning_rate": 6.5600164834420754e-06, | |
| "loss": 0.5318, | |
| "step": 11970 | |
| }, | |
| { | |
| "epoch": 24.0, | |
| "grad_norm": 53.413780212402344, | |
| "learning_rate": 6.545084971874738e-06, | |
| "loss": 0.5184, | |
| "step": 12000 | |
| }, | |
| { | |
| "epoch": 24.0, | |
| "eval_loss": 0.4076879024505615, | |
| "eval_map": 0.7097, | |
| "eval_map_50": 0.9464, | |
| "eval_map_75": 0.854, | |
| "eval_map_chicken": 0.7126, | |
| "eval_map_duck": 0.6338, | |
| "eval_map_large": 0.8335, | |
| "eval_map_medium": 0.7156, | |
| "eval_map_plant": 0.7828, | |
| "eval_map_small": 0.1197, | |
| "eval_mar_1": 0.2741, | |
| "eval_mar_10": 0.757, | |
| "eval_mar_100": 0.7607, | |
| "eval_mar_100_chicken": 0.7667, | |
| "eval_mar_100_duck": 0.6784, | |
| "eval_mar_100_plant": 0.8372, | |
| "eval_mar_large": 0.8828, | |
| "eval_mar_medium": 0.7738, | |
| "eval_mar_small": 0.2553, | |
| "eval_runtime": 4.9733, | |
| "eval_samples_per_second": 20.107, | |
| "eval_steps_per_second": 2.614, | |
| "step": 12000 | |
| }, | |
| { | |
| "epoch": 24.06, | |
| "grad_norm": 52.72776412963867, | |
| "learning_rate": 6.530138210942505e-06, | |
| "loss": 0.5589, | |
| "step": 12030 | |
| }, | |
| { | |
| "epoch": 24.12, | |
| "grad_norm": 29.438343048095703, | |
| "learning_rate": 6.5151763481638705e-06, | |
| "loss": 0.5368, | |
| "step": 12060 | |
| }, | |
| { | |
| "epoch": 24.18, | |
| "grad_norm": 54.32796859741211, | |
| "learning_rate": 6.500199531206381e-06, | |
| "loss": 0.5092, | |
| "step": 12090 | |
| }, | |
| { | |
| "epoch": 24.24, | |
| "grad_norm": 69.30450439453125, | |
| "learning_rate": 6.485207907885175e-06, | |
| "loss": 0.4992, | |
| "step": 12120 | |
| }, | |
| { | |
| "epoch": 24.3, | |
| "grad_norm": 42.193634033203125, | |
| "learning_rate": 6.47020162616152e-06, | |
| "loss": 0.4091, | |
| "step": 12150 | |
| }, | |
| { | |
| "epoch": 24.36, | |
| "grad_norm": 30.120336532592773, | |
| "learning_rate": 6.455180834141359e-06, | |
| "loss": 0.471, | |
| "step": 12180 | |
| }, | |
| { | |
| "epoch": 24.42, | |
| "grad_norm": 30.613067626953125, | |
| "learning_rate": 6.440145680073847e-06, | |
| "loss": 0.5791, | |
| "step": 12210 | |
| }, | |
| { | |
| "epoch": 24.48, | |
| "grad_norm": 38.11183166503906, | |
| "learning_rate": 6.425096312349881e-06, | |
| "loss": 0.4348, | |
| "step": 12240 | |
| }, | |
| { | |
| "epoch": 24.54, | |
| "grad_norm": 46.21797180175781, | |
| "learning_rate": 6.410032879500647e-06, | |
| "loss": 0.5827, | |
| "step": 12270 | |
| }, | |
| { | |
| "epoch": 24.6, | |
| "grad_norm": 39.75735855102539, | |
| "learning_rate": 6.3949555301961474e-06, | |
| "loss": 0.4915, | |
| "step": 12300 | |
| }, | |
| { | |
| "epoch": 24.66, | |
| "grad_norm": 55.170345306396484, | |
| "learning_rate": 6.3798644132437304e-06, | |
| "loss": 0.4253, | |
| "step": 12330 | |
| }, | |
| { | |
| "epoch": 24.72, | |
| "grad_norm": 30.651851654052734, | |
| "learning_rate": 6.364759677586627e-06, | |
| "loss": 0.5383, | |
| "step": 12360 | |
| }, | |
| { | |
| "epoch": 24.78, | |
| "grad_norm": 110.5732421875, | |
| "learning_rate": 6.349641472302484e-06, | |
| "loss": 0.4379, | |
| "step": 12390 | |
| }, | |
| { | |
| "epoch": 24.84, | |
| "grad_norm": 37.65018081665039, | |
| "learning_rate": 6.334509946601879e-06, | |
| "loss": 0.4374, | |
| "step": 12420 | |
| }, | |
| { | |
| "epoch": 24.9, | |
| "grad_norm": 48.68293762207031, | |
| "learning_rate": 6.3193652498268656e-06, | |
| "loss": 0.5241, | |
| "step": 12450 | |
| }, | |
| { | |
| "epoch": 24.96, | |
| "grad_norm": 131.17486572265625, | |
| "learning_rate": 6.304207531449486e-06, | |
| "loss": 0.4849, | |
| "step": 12480 | |
| }, | |
| { | |
| "epoch": 25.0, | |
| "eval_loss": 0.4043169915676117, | |
| "eval_map": 0.7096, | |
| "eval_map_50": 0.949, | |
| "eval_map_75": 0.8366, | |
| "eval_map_chicken": 0.7483, | |
| "eval_map_duck": 0.596, | |
| "eval_map_large": 0.8412, | |
| "eval_map_medium": 0.7084, | |
| "eval_map_plant": 0.7843, | |
| "eval_map_small": 0.1234, | |
| "eval_mar_1": 0.2739, | |
| "eval_mar_10": 0.7538, | |
| "eval_mar_100": 0.7611, | |
| "eval_mar_100_chicken": 0.7902, | |
| "eval_mar_100_duck": 0.6546, | |
| "eval_mar_100_plant": 0.8383, | |
| "eval_mar_large": 0.8891, | |
| "eval_mar_medium": 0.7703, | |
| "eval_mar_small": 0.258, | |
| "eval_runtime": 5.051, | |
| "eval_samples_per_second": 19.798, | |
| "eval_steps_per_second": 2.574, | |
| "step": 12500 | |
| }, | |
| { | |
| "epoch": 25.02, | |
| "grad_norm": 204.19046020507812, | |
| "learning_rate": 6.2890369410703e-06, | |
| "loss": 0.5097, | |
| "step": 12510 | |
| }, | |
| { | |
| "epoch": 25.08, | |
| "grad_norm": 104.87467956542969, | |
| "learning_rate": 6.273853628416911e-06, | |
| "loss": 0.507, | |
| "step": 12540 | |
| }, | |
| { | |
| "epoch": 25.14, | |
| "grad_norm": 32.947471618652344, | |
| "learning_rate": 6.258657743342486e-06, | |
| "loss": 0.4753, | |
| "step": 12570 | |
| }, | |
| { | |
| "epoch": 25.2, | |
| "grad_norm": 35.83731460571289, | |
| "learning_rate": 6.243449435824276e-06, | |
| "loss": 0.4727, | |
| "step": 12600 | |
| }, | |
| { | |
| "epoch": 25.26, | |
| "grad_norm": 53.04224395751953, | |
| "learning_rate": 6.228228855962133e-06, | |
| "loss": 0.5548, | |
| "step": 12630 | |
| }, | |
| { | |
| "epoch": 25.32, | |
| "grad_norm": 35.74001693725586, | |
| "learning_rate": 6.212996153977038e-06, | |
| "loss": 0.5208, | |
| "step": 12660 | |
| }, | |
| { | |
| "epoch": 25.38, | |
| "grad_norm": 45.188846588134766, | |
| "learning_rate": 6.1977514802096105e-06, | |
| "loss": 0.5547, | |
| "step": 12690 | |
| }, | |
| { | |
| "epoch": 25.44, | |
| "grad_norm": 58.8617057800293, | |
| "learning_rate": 6.182494985118625e-06, | |
| "loss": 0.4698, | |
| "step": 12720 | |
| }, | |
| { | |
| "epoch": 25.5, | |
| "grad_norm": 32.76780700683594, | |
| "learning_rate": 6.1672268192795285e-06, | |
| "loss": 0.5208, | |
| "step": 12750 | |
| }, | |
| { | |
| "epoch": 25.56, | |
| "grad_norm": 88.9961929321289, | |
| "learning_rate": 6.151947133382954e-06, | |
| "loss": 0.5717, | |
| "step": 12780 | |
| }, | |
| { | |
| "epoch": 25.62, | |
| "grad_norm": 51.27913284301758, | |
| "learning_rate": 6.136656078233233e-06, | |
| "loss": 0.4076, | |
| "step": 12810 | |
| }, | |
| { | |
| "epoch": 25.68, | |
| "grad_norm": 127.77012634277344, | |
| "learning_rate": 6.121353804746907e-06, | |
| "loss": 0.5238, | |
| "step": 12840 | |
| }, | |
| { | |
| "epoch": 25.74, | |
| "grad_norm": 53.242958068847656, | |
| "learning_rate": 6.106040463951237e-06, | |
| "loss": 0.5165, | |
| "step": 12870 | |
| }, | |
| { | |
| "epoch": 25.8, | |
| "grad_norm": 91.68648529052734, | |
| "learning_rate": 6.090716206982714e-06, | |
| "loss": 0.521, | |
| "step": 12900 | |
| }, | |
| { | |
| "epoch": 25.86, | |
| "grad_norm": 62.1449089050293, | |
| "learning_rate": 6.075381185085568e-06, | |
| "loss": 0.4903, | |
| "step": 12930 | |
| }, | |
| { | |
| "epoch": 25.92, | |
| "grad_norm": 0.0003859973221551627, | |
| "learning_rate": 6.060035549610275e-06, | |
| "loss": 0.4244, | |
| "step": 12960 | |
| }, | |
| { | |
| "epoch": 25.98, | |
| "grad_norm": 69.36231994628906, | |
| "learning_rate": 6.044679452012059e-06, | |
| "loss": 0.5022, | |
| "step": 12990 | |
| }, | |
| { | |
| "epoch": 26.0, | |
| "eval_loss": 0.3883710503578186, | |
| "eval_map": 0.7394, | |
| "eval_map_50": 0.9528, | |
| "eval_map_75": 0.8847, | |
| "eval_map_chicken": 0.7466, | |
| "eval_map_duck": 0.688, | |
| "eval_map_large": 0.8473, | |
| "eval_map_medium": 0.7337, | |
| "eval_map_plant": 0.7838, | |
| "eval_map_small": 0.1472, | |
| "eval_mar_1": 0.2918, | |
| "eval_mar_10": 0.7816, | |
| "eval_mar_100": 0.7876, | |
| "eval_mar_100_chicken": 0.7871, | |
| "eval_mar_100_duck": 0.7402, | |
| "eval_mar_100_plant": 0.8354, | |
| "eval_mar_large": 0.8971, | |
| "eval_mar_medium": 0.7888, | |
| "eval_mar_small": 0.2549, | |
| "eval_runtime": 5.5783, | |
| "eval_samples_per_second": 17.927, | |
| "eval_steps_per_second": 2.33, | |
| "step": 13000 | |
| }, | |
| { | |
| "epoch": 26.04, | |
| "grad_norm": 39.00455093383789, | |
| "learning_rate": 6.029313043849407e-06, | |
| "loss": 0.4858, | |
| "step": 13020 | |
| }, | |
| { | |
| "epoch": 26.1, | |
| "grad_norm": 42.91631317138672, | |
| "learning_rate": 6.013936476782563e-06, | |
| "loss": 0.573, | |
| "step": 13050 | |
| }, | |
| { | |
| "epoch": 26.16, | |
| "grad_norm": 87.59916687011719, | |
| "learning_rate": 5.9985499025720354e-06, | |
| "loss": 0.4353, | |
| "step": 13080 | |
| }, | |
| { | |
| "epoch": 26.22, | |
| "grad_norm": 25.52212905883789, | |
| "learning_rate": 5.9831534730771e-06, | |
| "loss": 0.5075, | |
| "step": 13110 | |
| }, | |
| { | |
| "epoch": 26.28, | |
| "grad_norm": 72.47872161865234, | |
| "learning_rate": 5.967747340254303e-06, | |
| "loss": 0.5847, | |
| "step": 13140 | |
| }, | |
| { | |
| "epoch": 26.34, | |
| "grad_norm": 58.947288513183594, | |
| "learning_rate": 5.952331656155951e-06, | |
| "loss": 0.5281, | |
| "step": 13170 | |
| }, | |
| { | |
| "epoch": 26.4, | |
| "grad_norm": 110.84516906738281, | |
| "learning_rate": 5.936906572928625e-06, | |
| "loss": 0.5088, | |
| "step": 13200 | |
| }, | |
| { | |
| "epoch": 26.46, | |
| "grad_norm": 29.063140869140625, | |
| "learning_rate": 5.9214722428116675e-06, | |
| "loss": 0.5583, | |
| "step": 13230 | |
| }, | |
| { | |
| "epoch": 26.52, | |
| "grad_norm": 41.858192443847656, | |
| "learning_rate": 5.906028818135687e-06, | |
| "loss": 0.4315, | |
| "step": 13260 | |
| }, | |
| { | |
| "epoch": 26.58, | |
| "grad_norm": 46.3973388671875, | |
| "learning_rate": 5.89057645132105e-06, | |
| "loss": 0.5034, | |
| "step": 13290 | |
| }, | |
| { | |
| "epoch": 26.64, | |
| "grad_norm": 57.123966217041016, | |
| "learning_rate": 5.8751152948763815e-06, | |
| "loss": 0.4648, | |
| "step": 13320 | |
| }, | |
| { | |
| "epoch": 26.7, | |
| "grad_norm": 36.376487731933594, | |
| "learning_rate": 5.859645501397048e-06, | |
| "loss": 0.4411, | |
| "step": 13350 | |
| }, | |
| { | |
| "epoch": 26.76, | |
| "grad_norm": 34.153175354003906, | |
| "learning_rate": 5.844167223563669e-06, | |
| "loss": 0.7433, | |
| "step": 13380 | |
| }, | |
| { | |
| "epoch": 26.82, | |
| "grad_norm": 65.91962432861328, | |
| "learning_rate": 5.828680614140599e-06, | |
| "loss": 0.4586, | |
| "step": 13410 | |
| }, | |
| { | |
| "epoch": 26.88, | |
| "grad_norm": 30.4884033203125, | |
| "learning_rate": 5.813185825974419e-06, | |
| "loss": 0.5479, | |
| "step": 13440 | |
| }, | |
| { | |
| "epoch": 26.94, | |
| "grad_norm": 94.28384399414062, | |
| "learning_rate": 5.797683011992432e-06, | |
| "loss": 0.6423, | |
| "step": 13470 | |
| }, | |
| { | |
| "epoch": 27.0, | |
| "grad_norm": 24.540077209472656, | |
| "learning_rate": 5.782172325201155e-06, | |
| "loss": 0.521, | |
| "step": 13500 | |
| }, | |
| { | |
| "epoch": 27.0, | |
| "eval_loss": 0.4196639358997345, | |
| "eval_map": 0.7177, | |
| "eval_map_50": 0.9434, | |
| "eval_map_75": 0.8715, | |
| "eval_map_chicken": 0.7073, | |
| "eval_map_duck": 0.6685, | |
| "eval_map_large": 0.8353, | |
| "eval_map_medium": 0.7168, | |
| "eval_map_plant": 0.7771, | |
| "eval_map_small": 0.132, | |
| "eval_mar_1": 0.2879, | |
| "eval_mar_10": 0.7639, | |
| "eval_mar_100": 0.7697, | |
| "eval_mar_100_chicken": 0.7649, | |
| "eval_mar_100_duck": 0.7165, | |
| "eval_mar_100_plant": 0.8277, | |
| "eval_mar_large": 0.8799, | |
| "eval_mar_medium": 0.7777, | |
| "eval_mar_small": 0.2623, | |
| "eval_runtime": 5.0566, | |
| "eval_samples_per_second": 19.776, | |
| "eval_steps_per_second": 2.571, | |
| "step": 13500 | |
| }, | |
| { | |
| "epoch": 27.06, | |
| "grad_norm": 46.547821044921875, | |
| "learning_rate": 5.766653918684803e-06, | |
| "loss": 0.4824, | |
| "step": 13530 | |
| }, | |
| { | |
| "epoch": 27.12, | |
| "grad_norm": 54.376251220703125, | |
| "learning_rate": 5.751127945603786e-06, | |
| "loss": 0.5959, | |
| "step": 13560 | |
| }, | |
| { | |
| "epoch": 27.18, | |
| "grad_norm": 42.895103454589844, | |
| "learning_rate": 5.735594559193187e-06, | |
| "loss": 0.5927, | |
| "step": 13590 | |
| }, | |
| { | |
| "epoch": 27.24, | |
| "grad_norm": 78.81582641601562, | |
| "learning_rate": 5.720053912761261e-06, | |
| "loss": 0.4704, | |
| "step": 13620 | |
| }, | |
| { | |
| "epoch": 27.3, | |
| "grad_norm": 72.24491882324219, | |
| "learning_rate": 5.704506159687914e-06, | |
| "loss": 0.5529, | |
| "step": 13650 | |
| }, | |
| { | |
| "epoch": 27.36, | |
| "grad_norm": 61.65352249145508, | |
| "learning_rate": 5.68895145342319e-06, | |
| "loss": 0.4379, | |
| "step": 13680 | |
| }, | |
| { | |
| "epoch": 27.42, | |
| "grad_norm": 64.8173828125, | |
| "learning_rate": 5.673389947485763e-06, | |
| "loss": 0.4938, | |
| "step": 13710 | |
| }, | |
| { | |
| "epoch": 27.48, | |
| "grad_norm": 35.893592834472656, | |
| "learning_rate": 5.657821795461413e-06, | |
| "loss": 0.4969, | |
| "step": 13740 | |
| }, | |
| { | |
| "epoch": 27.54, | |
| "grad_norm": 53.43833923339844, | |
| "learning_rate": 5.642247151001515e-06, | |
| "loss": 0.6261, | |
| "step": 13770 | |
| }, | |
| { | |
| "epoch": 27.6, | |
| "grad_norm": 36.205841064453125, | |
| "learning_rate": 5.626666167821522e-06, | |
| "loss": 0.4915, | |
| "step": 13800 | |
| }, | |
| { | |
| "epoch": 27.66, | |
| "grad_norm": 49.39502716064453, | |
| "learning_rate": 5.611078999699448e-06, | |
| "loss": 0.4529, | |
| "step": 13830 | |
| }, | |
| { | |
| "epoch": 27.72, | |
| "grad_norm": 68.46619415283203, | |
| "learning_rate": 5.59548580047435e-06, | |
| "loss": 0.573, | |
| "step": 13860 | |
| }, | |
| { | |
| "epoch": 27.78, | |
| "grad_norm": 99.93022155761719, | |
| "learning_rate": 5.5798867240448075e-06, | |
| "loss": 0.4234, | |
| "step": 13890 | |
| }, | |
| { | |
| "epoch": 27.84, | |
| "grad_norm": 41.78444290161133, | |
| "learning_rate": 5.5642819243674085e-06, | |
| "loss": 0.5215, | |
| "step": 13920 | |
| }, | |
| { | |
| "epoch": 27.9, | |
| "grad_norm": 65.3072280883789, | |
| "learning_rate": 5.548671555455226e-06, | |
| "loss": 0.5085, | |
| "step": 13950 | |
| }, | |
| { | |
| "epoch": 27.96, | |
| "grad_norm": 67.302978515625, | |
| "learning_rate": 5.5330557713763e-06, | |
| "loss": 0.5433, | |
| "step": 13980 | |
| }, | |
| { | |
| "epoch": 28.0, | |
| "eval_loss": 0.3886251151561737, | |
| "eval_map": 0.7454, | |
| "eval_map_50": 0.9508, | |
| "eval_map_75": 0.8823, | |
| "eval_map_chicken": 0.7573, | |
| "eval_map_duck": 0.6941, | |
| "eval_map_large": 0.8448, | |
| "eval_map_medium": 0.7406, | |
| "eval_map_plant": 0.785, | |
| "eval_map_small": 0.2083, | |
| "eval_mar_1": 0.292, | |
| "eval_mar_10": 0.7833, | |
| "eval_mar_100": 0.789, | |
| "eval_mar_100_chicken": 0.8004, | |
| "eval_mar_100_duck": 0.733, | |
| "eval_mar_100_plant": 0.8334, | |
| "eval_mar_large": 0.8845, | |
| "eval_mar_medium": 0.7952, | |
| "eval_mar_small": 0.3064, | |
| "eval_runtime": 5.0862, | |
| "eval_samples_per_second": 19.661, | |
| "eval_steps_per_second": 2.556, | |
| "step": 14000 | |
| }, | |
| { | |
| "epoch": 28.02, | |
| "grad_norm": 50.869449615478516, | |
| "learning_rate": 5.517434726252113e-06, | |
| "loss": 0.5022, | |
| "step": 14010 | |
| }, | |
| { | |
| "epoch": 28.08, | |
| "grad_norm": 30.78757095336914, | |
| "learning_rate": 5.5018085742560745e-06, | |
| "loss": 0.4819, | |
| "step": 14040 | |
| }, | |
| { | |
| "epoch": 28.14, | |
| "grad_norm": 45.835391998291016, | |
| "learning_rate": 5.486177469611999e-06, | |
| "loss": 0.5521, | |
| "step": 14070 | |
| }, | |
| { | |
| "epoch": 28.2, | |
| "grad_norm": 112.30412292480469, | |
| "learning_rate": 5.470541566592573e-06, | |
| "loss": 0.4874, | |
| "step": 14100 | |
| }, | |
| { | |
| "epoch": 28.26, | |
| "grad_norm": 61.26950454711914, | |
| "learning_rate": 5.454901019517851e-06, | |
| "loss": 0.5017, | |
| "step": 14130 | |
| }, | |
| { | |
| "epoch": 28.32, | |
| "grad_norm": 35.90389633178711, | |
| "learning_rate": 5.439255982753717e-06, | |
| "loss": 0.5153, | |
| "step": 14160 | |
| }, | |
| { | |
| "epoch": 28.38, | |
| "grad_norm": 82.97974395751953, | |
| "learning_rate": 5.423606610710368e-06, | |
| "loss": 0.4807, | |
| "step": 14190 | |
| }, | |
| { | |
| "epoch": 28.44, | |
| "grad_norm": 24.63577651977539, | |
| "learning_rate": 5.4079530578407895e-06, | |
| "loss": 0.5761, | |
| "step": 14220 | |
| }, | |
| { | |
| "epoch": 28.5, | |
| "grad_norm": 70.0156021118164, | |
| "learning_rate": 5.392295478639226e-06, | |
| "loss": 0.4393, | |
| "step": 14250 | |
| }, | |
| { | |
| "epoch": 28.56, | |
| "grad_norm": 47.858985900878906, | |
| "learning_rate": 5.376634027639664e-06, | |
| "loss": 0.4543, | |
| "step": 14280 | |
| }, | |
| { | |
| "epoch": 28.62, | |
| "grad_norm": 41.32699966430664, | |
| "learning_rate": 5.360968859414305e-06, | |
| "loss": 0.5178, | |
| "step": 14310 | |
| }, | |
| { | |
| "epoch": 28.68, | |
| "grad_norm": 38.778343200683594, | |
| "learning_rate": 5.345300128572031e-06, | |
| "loss": 0.5076, | |
| "step": 14340 | |
| }, | |
| { | |
| "epoch": 28.74, | |
| "grad_norm": 101.67027282714844, | |
| "learning_rate": 5.32962798975689e-06, | |
| "loss": 0.4308, | |
| "step": 14370 | |
| }, | |
| { | |
| "epoch": 28.8, | |
| "grad_norm": 51.669864654541016, | |
| "learning_rate": 5.3139525976465675e-06, | |
| "loss": 0.4741, | |
| "step": 14400 | |
| }, | |
| { | |
| "epoch": 28.86, | |
| "grad_norm": 40.283546447753906, | |
| "learning_rate": 5.298274106950855e-06, | |
| "loss": 0.5707, | |
| "step": 14430 | |
| }, | |
| { | |
| "epoch": 28.92, | |
| "grad_norm": 124.74097442626953, | |
| "learning_rate": 5.282592672410124e-06, | |
| "loss": 0.4298, | |
| "step": 14460 | |
| }, | |
| { | |
| "epoch": 28.98, | |
| "grad_norm": 46.044212341308594, | |
| "learning_rate": 5.2669084487938025e-06, | |
| "loss": 0.3889, | |
| "step": 14490 | |
| }, | |
| { | |
| "epoch": 29.0, | |
| "eval_loss": 0.3713254928588867, | |
| "eval_map": 0.7492, | |
| "eval_map_50": 0.9553, | |
| "eval_map_75": 0.8998, | |
| "eval_map_chicken": 0.7677, | |
| "eval_map_duck": 0.6849, | |
| "eval_map_large": 0.8468, | |
| "eval_map_medium": 0.7492, | |
| "eval_map_plant": 0.7951, | |
| "eval_map_small": 0.2224, | |
| "eval_mar_1": 0.2891, | |
| "eval_mar_10": 0.7873, | |
| "eval_mar_100": 0.7936, | |
| "eval_mar_100_chicken": 0.8053, | |
| "eval_mar_100_duck": 0.7299, | |
| "eval_mar_100_plant": 0.8455, | |
| "eval_mar_large": 0.8921, | |
| "eval_mar_medium": 0.8026, | |
| "eval_mar_small": 0.3112, | |
| "eval_runtime": 5.3046, | |
| "eval_samples_per_second": 18.852, | |
| "eval_steps_per_second": 2.451, | |
| "step": 14500 | |
| }, | |
| { | |
| "epoch": 29.04, | |
| "grad_norm": 28.102954864501953, | |
| "learning_rate": 5.251221590898848e-06, | |
| "loss": 0.4486, | |
| "step": 14520 | |
| }, | |
| { | |
| "epoch": 29.1, | |
| "grad_norm": 157.14828491210938, | |
| "learning_rate": 5.235532253548213e-06, | |
| "loss": 0.5139, | |
| "step": 14550 | |
| }, | |
| { | |
| "epoch": 29.16, | |
| "grad_norm": 51.68550109863281, | |
| "learning_rate": 5.219840591589325e-06, | |
| "loss": 0.4739, | |
| "step": 14580 | |
| }, | |
| { | |
| "epoch": 29.22, | |
| "grad_norm": 54.422359466552734, | |
| "learning_rate": 5.204146759892551e-06, | |
| "loss": 0.4549, | |
| "step": 14610 | |
| }, | |
| { | |
| "epoch": 29.28, | |
| "grad_norm": 56.245521545410156, | |
| "learning_rate": 5.188450913349674e-06, | |
| "loss": 0.4747, | |
| "step": 14640 | |
| }, | |
| { | |
| "epoch": 29.34, | |
| "grad_norm": 41.02756881713867, | |
| "learning_rate": 5.172753206872363e-06, | |
| "loss": 0.5289, | |
| "step": 14670 | |
| }, | |
| { | |
| "epoch": 29.4, | |
| "grad_norm": 37.02859878540039, | |
| "learning_rate": 5.157053795390642e-06, | |
| "loss": 0.4792, | |
| "step": 14700 | |
| }, | |
| { | |
| "epoch": 29.46, | |
| "grad_norm": 59.5302619934082, | |
| "learning_rate": 5.141352833851367e-06, | |
| "loss": 0.4709, | |
| "step": 14730 | |
| }, | |
| { | |
| "epoch": 29.52, | |
| "grad_norm": 46.878326416015625, | |
| "learning_rate": 5.1256504772166885e-06, | |
| "loss": 0.5425, | |
| "step": 14760 | |
| }, | |
| { | |
| "epoch": 29.58, | |
| "grad_norm": 79.3340072631836, | |
| "learning_rate": 5.109946880462526e-06, | |
| "loss": 0.5793, | |
| "step": 14790 | |
| }, | |
| { | |
| "epoch": 29.64, | |
| "grad_norm": 65.08154296875, | |
| "learning_rate": 5.0942421985770415e-06, | |
| "loss": 0.4454, | |
| "step": 14820 | |
| }, | |
| { | |
| "epoch": 29.7, | |
| "grad_norm": 53.71400833129883, | |
| "learning_rate": 5.078536586559104e-06, | |
| "loss": 0.4082, | |
| "step": 14850 | |
| }, | |
| { | |
| "epoch": 29.76, | |
| "grad_norm": 71.08831024169922, | |
| "learning_rate": 5.062830199416764e-06, | |
| "loss": 0.532, | |
| "step": 14880 | |
| }, | |
| { | |
| "epoch": 29.82, | |
| "grad_norm": 41.64462661743164, | |
| "learning_rate": 5.047123192165722e-06, | |
| "loss": 0.4744, | |
| "step": 14910 | |
| }, | |
| { | |
| "epoch": 29.88, | |
| "grad_norm": 47.789791107177734, | |
| "learning_rate": 5.031415719827796e-06, | |
| "loss": 0.5136, | |
| "step": 14940 | |
| }, | |
| { | |
| "epoch": 29.94, | |
| "grad_norm": 76.24413299560547, | |
| "learning_rate": 5.015707937429398e-06, | |
| "loss": 0.4167, | |
| "step": 14970 | |
| }, | |
| { | |
| "epoch": 30.0, | |
| "grad_norm": 46.26760482788086, | |
| "learning_rate": 5e-06, | |
| "loss": 0.5103, | |
| "step": 15000 | |
| }, | |
| { | |
| "epoch": 30.0, | |
| "eval_loss": 0.35560786724090576, | |
| "eval_map": 0.7584, | |
| "eval_map_50": 0.9576, | |
| "eval_map_75": 0.9014, | |
| "eval_map_chicken": 0.7654, | |
| "eval_map_duck": 0.7165, | |
| "eval_map_large": 0.8517, | |
| "eval_map_medium": 0.7509, | |
| "eval_map_plant": 0.7934, | |
| "eval_map_small": 0.219, | |
| "eval_mar_1": 0.2939, | |
| "eval_mar_10": 0.7954, | |
| "eval_mar_100": 0.8015, | |
| "eval_mar_100_chicken": 0.8022, | |
| "eval_mar_100_duck": 0.7557, | |
| "eval_mar_100_plant": 0.8467, | |
| "eval_mar_large": 0.8979, | |
| "eval_mar_medium": 0.8078, | |
| "eval_mar_small": 0.3089, | |
| "eval_runtime": 5.0764, | |
| "eval_samples_per_second": 19.699, | |
| "eval_steps_per_second": 2.561, | |
| "step": 15000 | |
| }, | |
| { | |
| "epoch": 30.06, | |
| "grad_norm": 32.46073532104492, | |
| "learning_rate": 4.984292062570603e-06, | |
| "loss": 0.4897, | |
| "step": 15030 | |
| }, | |
| { | |
| "epoch": 30.12, | |
| "grad_norm": 60.483089447021484, | |
| "learning_rate": 4.968584280172206e-06, | |
| "loss": 0.436, | |
| "step": 15060 | |
| }, | |
| { | |
| "epoch": 30.18, | |
| "grad_norm": 66.74795532226562, | |
| "learning_rate": 4.952876807834281e-06, | |
| "loss": 0.5319, | |
| "step": 15090 | |
| }, | |
| { | |
| "epoch": 30.24, | |
| "grad_norm": 72.88277435302734, | |
| "learning_rate": 4.937169800583237e-06, | |
| "loss": 0.4765, | |
| "step": 15120 | |
| }, | |
| { | |
| "epoch": 30.3, | |
| "grad_norm": 40.566162109375, | |
| "learning_rate": 4.921463413440898e-06, | |
| "loss": 0.4799, | |
| "step": 15150 | |
| }, | |
| { | |
| "epoch": 30.36, | |
| "grad_norm": 65.94972229003906, | |
| "learning_rate": 4.90575780142296e-06, | |
| "loss": 0.4977, | |
| "step": 15180 | |
| }, | |
| { | |
| "epoch": 30.42, | |
| "grad_norm": 56.45389175415039, | |
| "learning_rate": 4.890053119537475e-06, | |
| "loss": 0.4175, | |
| "step": 15210 | |
| }, | |
| { | |
| "epoch": 30.48, | |
| "grad_norm": 47.09212112426758, | |
| "learning_rate": 4.874349522783313e-06, | |
| "loss": 0.5274, | |
| "step": 15240 | |
| }, | |
| { | |
| "epoch": 30.54, | |
| "grad_norm": 145.99957275390625, | |
| "learning_rate": 4.8586471661486345e-06, | |
| "loss": 0.3792, | |
| "step": 15270 | |
| }, | |
| { | |
| "epoch": 30.6, | |
| "grad_norm": 86.61822509765625, | |
| "learning_rate": 4.842946204609359e-06, | |
| "loss": 0.474, | |
| "step": 15300 | |
| }, | |
| { | |
| "epoch": 30.66, | |
| "grad_norm": 136.17213439941406, | |
| "learning_rate": 4.827246793127639e-06, | |
| "loss": 0.4006, | |
| "step": 15330 | |
| }, | |
| { | |
| "epoch": 30.72, | |
| "grad_norm": 35.06562042236328, | |
| "learning_rate": 4.811549086650327e-06, | |
| "loss": 0.5298, | |
| "step": 15360 | |
| }, | |
| { | |
| "epoch": 30.78, | |
| "grad_norm": 49.849159240722656, | |
| "learning_rate": 4.79585324010745e-06, | |
| "loss": 0.5034, | |
| "step": 15390 | |
| }, | |
| { | |
| "epoch": 30.84, | |
| "grad_norm": 50.45303726196289, | |
| "learning_rate": 4.780159408410677e-06, | |
| "loss": 0.5148, | |
| "step": 15420 | |
| }, | |
| { | |
| "epoch": 30.9, | |
| "grad_norm": 94.5147705078125, | |
| "learning_rate": 4.7644677464517874e-06, | |
| "loss": 0.4116, | |
| "step": 15450 | |
| }, | |
| { | |
| "epoch": 30.96, | |
| "grad_norm": 32.6632080078125, | |
| "learning_rate": 4.748778409101153e-06, | |
| "loss": 0.4458, | |
| "step": 15480 | |
| }, | |
| { | |
| "epoch": 31.0, | |
| "eval_loss": 0.3680790960788727, | |
| "eval_map": 0.7355, | |
| "eval_map_50": 0.9518, | |
| "eval_map_75": 0.8831, | |
| "eval_map_chicken": 0.734, | |
| "eval_map_duck": 0.6771, | |
| "eval_map_large": 0.8569, | |
| "eval_map_medium": 0.7311, | |
| "eval_map_plant": 0.7955, | |
| "eval_map_small": 0.1416, | |
| "eval_mar_1": 0.292, | |
| "eval_mar_10": 0.7773, | |
| "eval_mar_100": 0.7814, | |
| "eval_mar_100_chicken": 0.7742, | |
| "eval_mar_100_duck": 0.7247, | |
| "eval_mar_100_plant": 0.8452, | |
| "eval_mar_large": 0.9008, | |
| "eval_mar_medium": 0.7892, | |
| "eval_mar_small": 0.2358, | |
| "eval_runtime": 5.1428, | |
| "eval_samples_per_second": 19.445, | |
| "eval_steps_per_second": 2.528, | |
| "step": 15500 | |
| }, | |
| { | |
| "epoch": 31.02, | |
| "grad_norm": 92.16204071044922, | |
| "learning_rate": 4.733091551206198e-06, | |
| "loss": 0.3926, | |
| "step": 15510 | |
| }, | |
| { | |
| "epoch": 31.08, | |
| "grad_norm": 75.78022766113281, | |
| "learning_rate": 4.717407327589878e-06, | |
| "loss": 0.4623, | |
| "step": 15540 | |
| }, | |
| { | |
| "epoch": 31.14, | |
| "grad_norm": 40.157047271728516, | |
| "learning_rate": 4.7017258930491474e-06, | |
| "loss": 0.4647, | |
| "step": 15570 | |
| }, | |
| { | |
| "epoch": 31.2, | |
| "grad_norm": 61.89482498168945, | |
| "learning_rate": 4.686047402353433e-06, | |
| "loss": 0.4591, | |
| "step": 15600 | |
| }, | |
| { | |
| "epoch": 31.26, | |
| "grad_norm": 85.72875213623047, | |
| "learning_rate": 4.670372010243111e-06, | |
| "loss": 0.4529, | |
| "step": 15630 | |
| }, | |
| { | |
| "epoch": 31.32, | |
| "grad_norm": 69.27102661132812, | |
| "learning_rate": 4.654699871427972e-06, | |
| "loss": 0.4169, | |
| "step": 15660 | |
| }, | |
| { | |
| "epoch": 31.38, | |
| "grad_norm": 96.01777648925781, | |
| "learning_rate": 4.639031140585697e-06, | |
| "loss": 0.4641, | |
| "step": 15690 | |
| }, | |
| { | |
| "epoch": 31.44, | |
| "grad_norm": 36.54241180419922, | |
| "learning_rate": 4.6233659723603374e-06, | |
| "loss": 0.4236, | |
| "step": 15720 | |
| }, | |
| { | |
| "epoch": 31.5, | |
| "grad_norm": 54.6783561706543, | |
| "learning_rate": 4.6077045213607765e-06, | |
| "loss": 0.4513, | |
| "step": 15750 | |
| }, | |
| { | |
| "epoch": 31.56, | |
| "grad_norm": 61.09621047973633, | |
| "learning_rate": 4.592046942159213e-06, | |
| "loss": 0.5054, | |
| "step": 15780 | |
| }, | |
| { | |
| "epoch": 31.62, | |
| "grad_norm": 47.04421615600586, | |
| "learning_rate": 4.576393389289633e-06, | |
| "loss": 0.4794, | |
| "step": 15810 | |
| }, | |
| { | |
| "epoch": 31.68, | |
| "grad_norm": 45.6216926574707, | |
| "learning_rate": 4.560744017246284e-06, | |
| "loss": 0.4868, | |
| "step": 15840 | |
| }, | |
| { | |
| "epoch": 31.74, | |
| "grad_norm": 43.895748138427734, | |
| "learning_rate": 4.545098980482151e-06, | |
| "loss": 0.5823, | |
| "step": 15870 | |
| }, | |
| { | |
| "epoch": 31.8, | |
| "grad_norm": 68.07852935791016, | |
| "learning_rate": 4.529458433407429e-06, | |
| "loss": 0.3868, | |
| "step": 15900 | |
| }, | |
| { | |
| "epoch": 31.86, | |
| "grad_norm": 33.31740951538086, | |
| "learning_rate": 4.513822530388004e-06, | |
| "loss": 0.6123, | |
| "step": 15930 | |
| }, | |
| { | |
| "epoch": 31.92, | |
| "grad_norm": 101.1479263305664, | |
| "learning_rate": 4.4981914257439254e-06, | |
| "loss": 0.4639, | |
| "step": 15960 | |
| }, | |
| { | |
| "epoch": 31.98, | |
| "grad_norm": 69.75340270996094, | |
| "learning_rate": 4.482565273747888e-06, | |
| "loss": 0.4369, | |
| "step": 15990 | |
| }, | |
| { | |
| "epoch": 32.0, | |
| "eval_loss": 0.3522694408893585, | |
| "eval_map": 0.7495, | |
| "eval_map_50": 0.9499, | |
| "eval_map_75": 0.8877, | |
| "eval_map_chicken": 0.765, | |
| "eval_map_duck": 0.6903, | |
| "eval_map_large": 0.8514, | |
| "eval_map_medium": 0.7447, | |
| "eval_map_plant": 0.7932, | |
| "eval_map_small": 0.142, | |
| "eval_mar_1": 0.2935, | |
| "eval_mar_10": 0.7877, | |
| "eval_mar_100": 0.7923, | |
| "eval_mar_100_chicken": 0.8027, | |
| "eval_mar_100_duck": 0.7309, | |
| "eval_mar_100_plant": 0.8432, | |
| "eval_mar_large": 0.895, | |
| "eval_mar_medium": 0.8026, | |
| "eval_mar_small": 0.2589, | |
| "eval_runtime": 5.6761, | |
| "eval_samples_per_second": 17.618, | |
| "eval_steps_per_second": 2.29, | |
| "step": 16000 | |
| }, | |
| { | |
| "epoch": 32.04, | |
| "grad_norm": 73.68572998046875, | |
| "learning_rate": 4.466944228623701e-06, | |
| "loss": 0.4943, | |
| "step": 16020 | |
| }, | |
| { | |
| "epoch": 32.1, | |
| "grad_norm": 59.75843048095703, | |
| "learning_rate": 4.451328444544774e-06, | |
| "loss": 0.7525, | |
| "step": 16050 | |
| }, | |
| { | |
| "epoch": 32.16, | |
| "grad_norm": 36.496517181396484, | |
| "learning_rate": 4.4357180756325915e-06, | |
| "loss": 0.5056, | |
| "step": 16080 | |
| }, | |
| { | |
| "epoch": 32.22, | |
| "grad_norm": 57.11168670654297, | |
| "learning_rate": 4.420113275955193e-06, | |
| "loss": 0.4615, | |
| "step": 16110 | |
| }, | |
| { | |
| "epoch": 32.28, | |
| "grad_norm": 65.62242126464844, | |
| "learning_rate": 4.404514199525651e-06, | |
| "loss": 0.407, | |
| "step": 16140 | |
| }, | |
| { | |
| "epoch": 32.34, | |
| "grad_norm": 60.669490814208984, | |
| "learning_rate": 4.388921000300553e-06, | |
| "loss": 0.548, | |
| "step": 16170 | |
| }, | |
| { | |
| "epoch": 32.4, | |
| "grad_norm": 79.71160125732422, | |
| "learning_rate": 4.373333832178478e-06, | |
| "loss": 0.4463, | |
| "step": 16200 | |
| }, | |
| { | |
| "epoch": 32.46, | |
| "grad_norm": 60.13050842285156, | |
| "learning_rate": 4.357752848998486e-06, | |
| "loss": 0.4668, | |
| "step": 16230 | |
| }, | |
| { | |
| "epoch": 32.52, | |
| "grad_norm": 68.18923950195312, | |
| "learning_rate": 4.342178204538588e-06, | |
| "loss": 0.5161, | |
| "step": 16260 | |
| }, | |
| { | |
| "epoch": 32.58, | |
| "grad_norm": 39.04051208496094, | |
| "learning_rate": 4.326610052514238e-06, | |
| "loss": 0.4557, | |
| "step": 16290 | |
| }, | |
| { | |
| "epoch": 32.64, | |
| "grad_norm": 29.669092178344727, | |
| "learning_rate": 4.31104854657681e-06, | |
| "loss": 0.4081, | |
| "step": 16320 | |
| }, | |
| { | |
| "epoch": 32.7, | |
| "grad_norm": 125.63433074951172, | |
| "learning_rate": 4.295493840312087e-06, | |
| "loss": 0.5007, | |
| "step": 16350 | |
| }, | |
| { | |
| "epoch": 32.76, | |
| "grad_norm": 65.3372802734375, | |
| "learning_rate": 4.279946087238739e-06, | |
| "loss": 0.4988, | |
| "step": 16380 | |
| }, | |
| { | |
| "epoch": 32.82, | |
| "grad_norm": 37.54350662231445, | |
| "learning_rate": 4.264405440806813e-06, | |
| "loss": 0.4608, | |
| "step": 16410 | |
| }, | |
| { | |
| "epoch": 32.88, | |
| "grad_norm": 96.40494537353516, | |
| "learning_rate": 4.248872054396215e-06, | |
| "loss": 0.4581, | |
| "step": 16440 | |
| }, | |
| { | |
| "epoch": 32.94, | |
| "grad_norm": 47.8939094543457, | |
| "learning_rate": 4.233346081315197e-06, | |
| "loss": 0.5492, | |
| "step": 16470 | |
| }, | |
| { | |
| "epoch": 33.0, | |
| "grad_norm": 0.001348652527667582, | |
| "learning_rate": 4.217827674798845e-06, | |
| "loss": 0.447, | |
| "step": 16500 | |
| }, | |
| { | |
| "epoch": 33.0, | |
| "eval_loss": 0.3665035367012024, | |
| "eval_map": 0.7448, | |
| "eval_map_50": 0.954, | |
| "eval_map_75": 0.8912, | |
| "eval_map_chicken": 0.7453, | |
| "eval_map_duck": 0.6933, | |
| "eval_map_large": 0.8505, | |
| "eval_map_medium": 0.7412, | |
| "eval_map_plant": 0.796, | |
| "eval_map_small": 0.1577, | |
| "eval_mar_1": 0.2922, | |
| "eval_mar_10": 0.7844, | |
| "eval_mar_100": 0.7879, | |
| "eval_mar_100_chicken": 0.7813, | |
| "eval_mar_100_duck": 0.7361, | |
| "eval_mar_100_plant": 0.8464, | |
| "eval_mar_large": 0.8929, | |
| "eval_mar_medium": 0.8005, | |
| "eval_mar_small": 0.2742, | |
| "eval_runtime": 5.1427, | |
| "eval_samples_per_second": 19.445, | |
| "eval_steps_per_second": 2.528, | |
| "step": 16500 | |
| }, | |
| { | |
| "epoch": 33.06, | |
| "grad_norm": 50.67320251464844, | |
| "learning_rate": 4.202316988007568e-06, | |
| "loss": 0.4183, | |
| "step": 16530 | |
| }, | |
| { | |
| "epoch": 33.12, | |
| "grad_norm": 51.36381149291992, | |
| "learning_rate": 4.186814174025582e-06, | |
| "loss": 0.4446, | |
| "step": 16560 | |
| }, | |
| { | |
| "epoch": 33.18, | |
| "grad_norm": 86.76451110839844, | |
| "learning_rate": 4.171319385859402e-06, | |
| "loss": 0.3794, | |
| "step": 16590 | |
| }, | |
| { | |
| "epoch": 33.24, | |
| "grad_norm": 82.5096206665039, | |
| "learning_rate": 4.155832776436331e-06, | |
| "loss": 0.4867, | |
| "step": 16620 | |
| }, | |
| { | |
| "epoch": 33.3, | |
| "grad_norm": 110.49205780029297, | |
| "learning_rate": 4.140354498602952e-06, | |
| "loss": 0.454, | |
| "step": 16650 | |
| }, | |
| { | |
| "epoch": 33.36, | |
| "grad_norm": 29.124536514282227, | |
| "learning_rate": 4.124884705123619e-06, | |
| "loss": 0.5245, | |
| "step": 16680 | |
| }, | |
| { | |
| "epoch": 33.42, | |
| "grad_norm": 78.4069595336914, | |
| "learning_rate": 4.109423548678949e-06, | |
| "loss": 0.4232, | |
| "step": 16710 | |
| }, | |
| { | |
| "epoch": 33.48, | |
| "grad_norm": 68.84821319580078, | |
| "learning_rate": 4.093971181864313e-06, | |
| "loss": 0.4485, | |
| "step": 16740 | |
| }, | |
| { | |
| "epoch": 33.54, | |
| "grad_norm": 91.24239349365234, | |
| "learning_rate": 4.078527757188333e-06, | |
| "loss": 0.4898, | |
| "step": 16770 | |
| }, | |
| { | |
| "epoch": 33.6, | |
| "grad_norm": 61.37797546386719, | |
| "learning_rate": 4.063093427071376e-06, | |
| "loss": 0.6196, | |
| "step": 16800 | |
| }, | |
| { | |
| "epoch": 33.66, | |
| "grad_norm": 51.44956588745117, | |
| "learning_rate": 4.047668343844051e-06, | |
| "loss": 0.5473, | |
| "step": 16830 | |
| }, | |
| { | |
| "epoch": 33.72, | |
| "grad_norm": 80.08527374267578, | |
| "learning_rate": 4.032252659745699e-06, | |
| "loss": 0.4776, | |
| "step": 16860 | |
| }, | |
| { | |
| "epoch": 33.78, | |
| "grad_norm": 42.223960876464844, | |
| "learning_rate": 4.016846526922901e-06, | |
| "loss": 0.5304, | |
| "step": 16890 | |
| }, | |
| { | |
| "epoch": 33.84, | |
| "grad_norm": 40.85383605957031, | |
| "learning_rate": 4.001450097427965e-06, | |
| "loss": 0.4721, | |
| "step": 16920 | |
| }, | |
| { | |
| "epoch": 33.9, | |
| "grad_norm": 25.363073348999023, | |
| "learning_rate": 3.986063523217439e-06, | |
| "loss": 0.4369, | |
| "step": 16950 | |
| }, | |
| { | |
| "epoch": 33.96, | |
| "grad_norm": 53.42643737792969, | |
| "learning_rate": 3.970686956150595e-06, | |
| "loss": 0.4692, | |
| "step": 16980 | |
| }, | |
| { | |
| "epoch": 34.0, | |
| "eval_loss": 0.3454967439174652, | |
| "eval_map": 0.7589, | |
| "eval_map_50": 0.954, | |
| "eval_map_75": 0.899, | |
| "eval_map_chicken": 0.7709, | |
| "eval_map_duck": 0.7058, | |
| "eval_map_large": 0.863, | |
| "eval_map_medium": 0.7459, | |
| "eval_map_plant": 0.7999, | |
| "eval_map_small": 0.1729, | |
| "eval_mar_1": 0.2949, | |
| "eval_mar_10": 0.798, | |
| "eval_mar_100": 0.8038, | |
| "eval_mar_100_chicken": 0.8093, | |
| "eval_mar_100_duck": 0.7546, | |
| "eval_mar_100_plant": 0.8476, | |
| "eval_mar_large": 0.9033, | |
| "eval_mar_medium": 0.8068, | |
| "eval_mar_small": 0.3123, | |
| "eval_runtime": 5.1562, | |
| "eval_samples_per_second": 19.394, | |
| "eval_steps_per_second": 2.521, | |
| "step": 17000 | |
| }, | |
| { | |
| "epoch": 34.02, | |
| "grad_norm": 52.77671813964844, | |
| "learning_rate": 3.955320547987943e-06, | |
| "loss": 0.5145, | |
| "step": 17010 | |
| }, | |
| { | |
| "epoch": 34.08, | |
| "grad_norm": 59.353485107421875, | |
| "learning_rate": 3.939964450389728e-06, | |
| "loss": 0.4202, | |
| "step": 17040 | |
| }, | |
| { | |
| "epoch": 34.14, | |
| "grad_norm": 71.41014099121094, | |
| "learning_rate": 3.924618814914435e-06, | |
| "loss": 0.4365, | |
| "step": 17070 | |
| }, | |
| { | |
| "epoch": 34.2, | |
| "grad_norm": 59.05385971069336, | |
| "learning_rate": 3.909283793017289e-06, | |
| "loss": 0.4725, | |
| "step": 17100 | |
| }, | |
| { | |
| "epoch": 34.26, | |
| "grad_norm": 74.0579605102539, | |
| "learning_rate": 3.8939595360487655e-06, | |
| "loss": 0.4601, | |
| "step": 17130 | |
| }, | |
| { | |
| "epoch": 34.32, | |
| "grad_norm": 26.610624313354492, | |
| "learning_rate": 3.8786461952530955e-06, | |
| "loss": 0.4415, | |
| "step": 17160 | |
| }, | |
| { | |
| "epoch": 34.38, | |
| "grad_norm": 97.6423110961914, | |
| "learning_rate": 3.863343921766769e-06, | |
| "loss": 0.441, | |
| "step": 17190 | |
| }, | |
| { | |
| "epoch": 34.44, | |
| "grad_norm": 97.2457046508789, | |
| "learning_rate": 3.8480528666170495e-06, | |
| "loss": 0.4235, | |
| "step": 17220 | |
| }, | |
| { | |
| "epoch": 34.5, | |
| "grad_norm": 37.252906799316406, | |
| "learning_rate": 3.832773180720475e-06, | |
| "loss": 0.3976, | |
| "step": 17250 | |
| }, | |
| { | |
| "epoch": 34.56, | |
| "grad_norm": 37.12264633178711, | |
| "learning_rate": 3.817505014881378e-06, | |
| "loss": 0.5479, | |
| "step": 17280 | |
| }, | |
| { | |
| "epoch": 34.62, | |
| "grad_norm": 64.19234466552734, | |
| "learning_rate": 3.8022485197903924e-06, | |
| "loss": 0.4438, | |
| "step": 17310 | |
| }, | |
| { | |
| "epoch": 34.68, | |
| "grad_norm": 52.00288009643555, | |
| "learning_rate": 3.787003846022964e-06, | |
| "loss": 0.4803, | |
| "step": 17340 | |
| }, | |
| { | |
| "epoch": 34.74, | |
| "grad_norm": 39.877071380615234, | |
| "learning_rate": 3.7717711440378695e-06, | |
| "loss": 0.51, | |
| "step": 17370 | |
| }, | |
| { | |
| "epoch": 34.8, | |
| "grad_norm": 62.76517105102539, | |
| "learning_rate": 3.756550564175727e-06, | |
| "loss": 0.4295, | |
| "step": 17400 | |
| }, | |
| { | |
| "epoch": 34.86, | |
| "grad_norm": 42.63980484008789, | |
| "learning_rate": 3.7413422566575153e-06, | |
| "loss": 0.4405, | |
| "step": 17430 | |
| }, | |
| { | |
| "epoch": 34.92, | |
| "grad_norm": 62.36055374145508, | |
| "learning_rate": 3.7261463715830902e-06, | |
| "loss": 0.402, | |
| "step": 17460 | |
| }, | |
| { | |
| "epoch": 34.98, | |
| "grad_norm": 58.33619689941406, | |
| "learning_rate": 3.7109630589297014e-06, | |
| "loss": 0.4272, | |
| "step": 17490 | |
| }, | |
| { | |
| "epoch": 35.0, | |
| "eval_loss": 0.3380710482597351, | |
| "eval_map": 0.767, | |
| "eval_map_50": 0.9568, | |
| "eval_map_75": 0.903, | |
| "eval_map_chicken": 0.7842, | |
| "eval_map_duck": 0.7246, | |
| "eval_map_large": 0.852, | |
| "eval_map_medium": 0.7623, | |
| "eval_map_plant": 0.7922, | |
| "eval_map_small": 0.1734, | |
| "eval_mar_1": 0.2948, | |
| "eval_mar_10": 0.802, | |
| "eval_mar_100": 0.8061, | |
| "eval_mar_100_chicken": 0.8204, | |
| "eval_mar_100_duck": 0.7546, | |
| "eval_mar_100_plant": 0.8432, | |
| "eval_mar_large": 0.8962, | |
| "eval_mar_medium": 0.8136, | |
| "eval_mar_small": 0.2907, | |
| "eval_runtime": 5.0658, | |
| "eval_samples_per_second": 19.74, | |
| "eval_steps_per_second": 2.566, | |
| "step": 17500 | |
| }, | |
| { | |
| "epoch": 35.04, | |
| "grad_norm": 41.134178161621094, | |
| "learning_rate": 3.695792468550517e-06, | |
| "loss": 0.4129, | |
| "step": 17520 | |
| }, | |
| { | |
| "epoch": 35.1, | |
| "grad_norm": 25.44190216064453, | |
| "learning_rate": 3.680634750173137e-06, | |
| "loss": 0.4478, | |
| "step": 17550 | |
| }, | |
| { | |
| "epoch": 35.16, | |
| "grad_norm": 82.22879028320312, | |
| "learning_rate": 3.6654900533981234e-06, | |
| "loss": 0.4787, | |
| "step": 17580 | |
| }, | |
| { | |
| "epoch": 35.22, | |
| "grad_norm": 50.50277328491211, | |
| "learning_rate": 3.6503585276975196e-06, | |
| "loss": 0.369, | |
| "step": 17610 | |
| }, | |
| { | |
| "epoch": 35.28, | |
| "grad_norm": 48.85622024536133, | |
| "learning_rate": 3.635240322413375e-06, | |
| "loss": 0.461, | |
| "step": 17640 | |
| }, | |
| { | |
| "epoch": 35.34, | |
| "grad_norm": 68.70919799804688, | |
| "learning_rate": 3.6201355867562725e-06, | |
| "loss": 0.4257, | |
| "step": 17670 | |
| }, | |
| { | |
| "epoch": 35.4, | |
| "grad_norm": 57.635738372802734, | |
| "learning_rate": 3.6050444698038547e-06, | |
| "loss": 0.3771, | |
| "step": 17700 | |
| }, | |
| { | |
| "epoch": 35.46, | |
| "grad_norm": 55.46888732910156, | |
| "learning_rate": 3.5899671204993535e-06, | |
| "loss": 0.5256, | |
| "step": 17730 | |
| }, | |
| { | |
| "epoch": 35.52, | |
| "grad_norm": 41.51885986328125, | |
| "learning_rate": 3.5749036876501196e-06, | |
| "loss": 0.5109, | |
| "step": 17760 | |
| }, | |
| { | |
| "epoch": 35.58, | |
| "grad_norm": 49.13798904418945, | |
| "learning_rate": 3.559854319926156e-06, | |
| "loss": 0.4184, | |
| "step": 17790 | |
| }, | |
| { | |
| "epoch": 35.64, | |
| "grad_norm": 60.9365119934082, | |
| "learning_rate": 3.5448191658586423e-06, | |
| "loss": 0.4215, | |
| "step": 17820 | |
| }, | |
| { | |
| "epoch": 35.7, | |
| "grad_norm": 66.89413452148438, | |
| "learning_rate": 3.5297983738384813e-06, | |
| "loss": 0.5379, | |
| "step": 17850 | |
| }, | |
| { | |
| "epoch": 35.76, | |
| "grad_norm": 116.77861022949219, | |
| "learning_rate": 3.5147920921148267e-06, | |
| "loss": 0.4938, | |
| "step": 17880 | |
| }, | |
| { | |
| "epoch": 35.82, | |
| "grad_norm": 0.00022027752129361033, | |
| "learning_rate": 3.49980046879362e-06, | |
| "loss": 0.5032, | |
| "step": 17910 | |
| }, | |
| { | |
| "epoch": 35.88, | |
| "grad_norm": 51.908206939697266, | |
| "learning_rate": 3.484823651836131e-06, | |
| "loss": 0.4807, | |
| "step": 17940 | |
| }, | |
| { | |
| "epoch": 35.94, | |
| "grad_norm": 56.6352653503418, | |
| "learning_rate": 3.4698617890574972e-06, | |
| "loss": 0.4326, | |
| "step": 17970 | |
| }, | |
| { | |
| "epoch": 36.0, | |
| "grad_norm": 47.75572967529297, | |
| "learning_rate": 3.4549150281252635e-06, | |
| "loss": 0.4021, | |
| "step": 18000 | |
| }, | |
| { | |
| "epoch": 36.0, | |
| "eval_loss": 0.3322797417640686, | |
| "eval_map": 0.7686, | |
| "eval_map_50": 0.9551, | |
| "eval_map_75": 0.8938, | |
| "eval_map_chicken": 0.7776, | |
| "eval_map_duck": 0.7245, | |
| "eval_map_large": 0.8616, | |
| "eval_map_medium": 0.7621, | |
| "eval_map_plant": 0.8038, | |
| "eval_map_small": 0.1892, | |
| "eval_mar_1": 0.2969, | |
| "eval_mar_10": 0.8025, | |
| "eval_mar_100": 0.8067, | |
| "eval_mar_100_chicken": 0.8133, | |
| "eval_mar_100_duck": 0.7577, | |
| "eval_mar_100_plant": 0.849, | |
| "eval_mar_large": 0.9025, | |
| "eval_mar_medium": 0.8122, | |
| "eval_mar_small": 0.3049, | |
| "eval_runtime": 5.0345, | |
| "eval_samples_per_second": 19.863, | |
| "eval_steps_per_second": 2.582, | |
| "step": 18000 | |
| }, | |
| { | |
| "epoch": 36.06, | |
| "grad_norm": 58.42578125, | |
| "learning_rate": 3.4399835165579266e-06, | |
| "loss": 0.4267, | |
| "step": 18030 | |
| }, | |
| { | |
| "epoch": 36.12, | |
| "grad_norm": 48.94293212890625, | |
| "learning_rate": 3.4250674017234774e-06, | |
| "loss": 0.4093, | |
| "step": 18060 | |
| }, | |
| { | |
| "epoch": 36.18, | |
| "grad_norm": 38.52609634399414, | |
| "learning_rate": 3.4101668308379465e-06, | |
| "loss": 0.4707, | |
| "step": 18090 | |
| }, | |
| { | |
| "epoch": 36.24, | |
| "grad_norm": 41.78977584838867, | |
| "learning_rate": 3.3952819509639534e-06, | |
| "loss": 0.5579, | |
| "step": 18120 | |
| }, | |
| { | |
| "epoch": 36.3, | |
| "grad_norm": 65.99064636230469, | |
| "learning_rate": 3.3804129090092542e-06, | |
| "loss": 0.4812, | |
| "step": 18150 | |
| }, | |
| { | |
| "epoch": 36.36, | |
| "grad_norm": 94.7818603515625, | |
| "learning_rate": 3.3655598517252886e-06, | |
| "loss": 0.4009, | |
| "step": 18180 | |
| }, | |
| { | |
| "epoch": 36.42, | |
| "grad_norm": 39.25736999511719, | |
| "learning_rate": 3.350722925705736e-06, | |
| "loss": 0.4333, | |
| "step": 18210 | |
| }, | |
| { | |
| "epoch": 36.48, | |
| "grad_norm": 36.73288345336914, | |
| "learning_rate": 3.3359022773850673e-06, | |
| "loss": 0.4128, | |
| "step": 18240 | |
| }, | |
| { | |
| "epoch": 36.54, | |
| "grad_norm": 93.98817443847656, | |
| "learning_rate": 3.3210980530370974e-06, | |
| "loss": 0.41, | |
| "step": 18270 | |
| }, | |
| { | |
| "epoch": 36.6, | |
| "grad_norm": 49.68222427368164, | |
| "learning_rate": 3.3063103987735433e-06, | |
| "loss": 0.5326, | |
| "step": 18300 | |
| }, | |
| { | |
| "epoch": 36.66, | |
| "grad_norm": 41.902557373046875, | |
| "learning_rate": 3.2915394605425836e-06, | |
| "loss": 0.3922, | |
| "step": 18330 | |
| }, | |
| { | |
| "epoch": 36.72, | |
| "grad_norm": 88.83726501464844, | |
| "learning_rate": 3.2767853841274154e-06, | |
| "loss": 0.4792, | |
| "step": 18360 | |
| }, | |
| { | |
| "epoch": 36.78, | |
| "grad_norm": 33.33390808105469, | |
| "learning_rate": 3.262048315144816e-06, | |
| "loss": 0.4964, | |
| "step": 18390 | |
| }, | |
| { | |
| "epoch": 36.84, | |
| "grad_norm": 92.82958984375, | |
| "learning_rate": 3.247328399043706e-06, | |
| "loss": 0.4358, | |
| "step": 18420 | |
| }, | |
| { | |
| "epoch": 36.9, | |
| "grad_norm": 63.93076705932617, | |
| "learning_rate": 3.2326257811037154e-06, | |
| "loss": 0.4057, | |
| "step": 18450 | |
| }, | |
| { | |
| "epoch": 36.96, | |
| "grad_norm": 74.28779602050781, | |
| "learning_rate": 3.217940606433747e-06, | |
| "loss": 0.4582, | |
| "step": 18480 | |
| }, | |
| { | |
| "epoch": 37.0, | |
| "eval_loss": 0.3262586295604706, | |
| "eval_map": 0.7732, | |
| "eval_map_50": 0.9547, | |
| "eval_map_75": 0.9023, | |
| "eval_map_chicken": 0.7729, | |
| "eval_map_duck": 0.7298, | |
| "eval_map_large": 0.8742, | |
| "eval_map_medium": 0.7748, | |
| "eval_map_plant": 0.817, | |
| "eval_map_small": 0.1477, | |
| "eval_mar_1": 0.3015, | |
| "eval_mar_10": 0.8092, | |
| "eval_mar_100": 0.8132, | |
| "eval_mar_100_chicken": 0.812, | |
| "eval_mar_100_duck": 0.7619, | |
| "eval_mar_100_plant": 0.8657, | |
| "eval_mar_large": 0.9163, | |
| "eval_mar_medium": 0.8251, | |
| "eval_mar_small": 0.2634, | |
| "eval_runtime": 5.1374, | |
| "eval_samples_per_second": 19.465, | |
| "eval_steps_per_second": 2.53, | |
| "step": 18500 | |
| }, | |
| { | |
| "epoch": 37.02, | |
| "grad_norm": 32.50000762939453, | |
| "learning_rate": 3.2032730199705477e-06, | |
| "loss": 0.438, | |
| "step": 18510 | |
| }, | |
| { | |
| "epoch": 37.08, | |
| "grad_norm": 59.4605598449707, | |
| "learning_rate": 3.188623166477272e-06, | |
| "loss": 0.4954, | |
| "step": 18540 | |
| }, | |
| { | |
| "epoch": 37.14, | |
| "grad_norm": 78.10517120361328, | |
| "learning_rate": 3.1739911905420617e-06, | |
| "loss": 0.3849, | |
| "step": 18570 | |
| }, | |
| { | |
| "epoch": 37.2, | |
| "grad_norm": 63.686492919921875, | |
| "learning_rate": 3.1593772365766107e-06, | |
| "loss": 0.3561, | |
| "step": 18600 | |
| }, | |
| { | |
| "epoch": 37.26, | |
| "grad_norm": 46.54610061645508, | |
| "learning_rate": 3.144781448814746e-06, | |
| "loss": 0.4171, | |
| "step": 18630 | |
| }, | |
| { | |
| "epoch": 37.32, | |
| "grad_norm": 44.2602424621582, | |
| "learning_rate": 3.130203971310999e-06, | |
| "loss": 0.4487, | |
| "step": 18660 | |
| }, | |
| { | |
| "epoch": 37.38, | |
| "grad_norm": 100.22451782226562, | |
| "learning_rate": 3.1156449479391876e-06, | |
| "loss": 0.4211, | |
| "step": 18690 | |
| }, | |
| { | |
| "epoch": 37.44, | |
| "grad_norm": 71.37747192382812, | |
| "learning_rate": 3.1011045223909954e-06, | |
| "loss": 0.5317, | |
| "step": 18720 | |
| }, | |
| { | |
| "epoch": 37.5, | |
| "grad_norm": 0.0002173359098378569, | |
| "learning_rate": 3.0865828381745515e-06, | |
| "loss": 0.4516, | |
| "step": 18750 | |
| }, | |
| { | |
| "epoch": 37.56, | |
| "grad_norm": 53.496341705322266, | |
| "learning_rate": 3.0720800386130176e-06, | |
| "loss": 0.4133, | |
| "step": 18780 | |
| }, | |
| { | |
| "epoch": 37.62, | |
| "grad_norm": 58.87167739868164, | |
| "learning_rate": 3.0575962668431704e-06, | |
| "loss": 0.4629, | |
| "step": 18810 | |
| }, | |
| { | |
| "epoch": 37.68, | |
| "grad_norm": 26.332555770874023, | |
| "learning_rate": 3.043131665813988e-06, | |
| "loss": 0.4264, | |
| "step": 18840 | |
| }, | |
| { | |
| "epoch": 37.74, | |
| "grad_norm": 62.68268585205078, | |
| "learning_rate": 3.028686378285245e-06, | |
| "loss": 0.459, | |
| "step": 18870 | |
| }, | |
| { | |
| "epoch": 37.8, | |
| "grad_norm": 105.45973205566406, | |
| "learning_rate": 3.0142605468260976e-06, | |
| "loss": 0.3687, | |
| "step": 18900 | |
| }, | |
| { | |
| "epoch": 37.86, | |
| "grad_norm": 60.29586410522461, | |
| "learning_rate": 2.9998543138136773e-06, | |
| "loss": 0.341, | |
| "step": 18930 | |
| }, | |
| { | |
| "epoch": 37.92, | |
| "grad_norm": 48.2047119140625, | |
| "learning_rate": 2.9854678214316875e-06, | |
| "loss": 0.3729, | |
| "step": 18960 | |
| }, | |
| { | |
| "epoch": 37.98, | |
| "grad_norm": 34.51201248168945, | |
| "learning_rate": 2.9711012116690007e-06, | |
| "loss": 0.3992, | |
| "step": 18990 | |
| }, | |
| { | |
| "epoch": 38.0, | |
| "eval_loss": 0.3207389712333679, | |
| "eval_map": 0.7799, | |
| "eval_map_50": 0.956, | |
| "eval_map_75": 0.9064, | |
| "eval_map_chicken": 0.7833, | |
| "eval_map_duck": 0.7384, | |
| "eval_map_large": 0.873, | |
| "eval_map_medium": 0.7823, | |
| "eval_map_plant": 0.818, | |
| "eval_map_small": 0.1767, | |
| "eval_mar_1": 0.3014, | |
| "eval_mar_10": 0.8168, | |
| "eval_mar_100": 0.8209, | |
| "eval_mar_100_chicken": 0.8218, | |
| "eval_mar_100_duck": 0.7732, | |
| "eval_mar_100_plant": 0.8677, | |
| "eval_mar_large": 0.9172, | |
| "eval_mar_medium": 0.8312, | |
| "eval_mar_small": 0.3028, | |
| "eval_runtime": 5.0908, | |
| "eval_samples_per_second": 19.643, | |
| "eval_steps_per_second": 2.554, | |
| "step": 19000 | |
| }, | |
| { | |
| "epoch": 38.04, | |
| "grad_norm": 45.1266975402832, | |
| "learning_rate": 2.9567546263182554e-06, | |
| "loss": 0.4895, | |
| "step": 19020 | |
| }, | |
| { | |
| "epoch": 38.1, | |
| "grad_norm": 77.6568374633789, | |
| "learning_rate": 2.9424282069744564e-06, | |
| "loss": 0.47, | |
| "step": 19050 | |
| }, | |
| { | |
| "epoch": 38.16, | |
| "grad_norm": 85.21626281738281, | |
| "learning_rate": 2.92812209503358e-06, | |
| "loss": 0.4005, | |
| "step": 19080 | |
| }, | |
| { | |
| "epoch": 38.22, | |
| "grad_norm": 62.11589431762695, | |
| "learning_rate": 2.9138364316911747e-06, | |
| "loss": 0.3894, | |
| "step": 19110 | |
| }, | |
| { | |
| "epoch": 38.28, | |
| "grad_norm": 50.71180725097656, | |
| "learning_rate": 2.899571357940969e-06, | |
| "loss": 0.4134, | |
| "step": 19140 | |
| }, | |
| { | |
| "epoch": 38.34, | |
| "grad_norm": 72.69100952148438, | |
| "learning_rate": 2.8853270145734846e-06, | |
| "loss": 0.3971, | |
| "step": 19170 | |
| }, | |
| { | |
| "epoch": 38.4, | |
| "grad_norm": 27.520477294921875, | |
| "learning_rate": 2.871103542174637e-06, | |
| "loss": 0.3922, | |
| "step": 19200 | |
| }, | |
| { | |
| "epoch": 38.46, | |
| "grad_norm": 64.30564880371094, | |
| "learning_rate": 2.856901081124359e-06, | |
| "loss": 0.4765, | |
| "step": 19230 | |
| }, | |
| { | |
| "epoch": 38.52, | |
| "grad_norm": 54.069091796875, | |
| "learning_rate": 2.8427197715952047e-06, | |
| "loss": 0.4478, | |
| "step": 19260 | |
| }, | |
| { | |
| "epoch": 38.58, | |
| "grad_norm": 62.98984146118164, | |
| "learning_rate": 2.8285597535509775e-06, | |
| "loss": 0.4548, | |
| "step": 19290 | |
| }, | |
| { | |
| "epoch": 38.64, | |
| "grad_norm": 52.520530700683594, | |
| "learning_rate": 2.814421166745337e-06, | |
| "loss": 0.529, | |
| "step": 19320 | |
| }, | |
| { | |
| "epoch": 38.7, | |
| "grad_norm": 71.8565902709961, | |
| "learning_rate": 2.800304150720424e-06, | |
| "loss": 0.4645, | |
| "step": 19350 | |
| }, | |
| { | |
| "epoch": 38.76, | |
| "grad_norm": 44.72001266479492, | |
| "learning_rate": 2.7862088448054936e-06, | |
| "loss": 0.4887, | |
| "step": 19380 | |
| }, | |
| { | |
| "epoch": 38.82, | |
| "grad_norm": 50.02251052856445, | |
| "learning_rate": 2.772135388115519e-06, | |
| "loss": 0.4167, | |
| "step": 19410 | |
| }, | |
| { | |
| "epoch": 38.88, | |
| "grad_norm": 59.54885482788086, | |
| "learning_rate": 2.7580839195498397e-06, | |
| "loss": 0.393, | |
| "step": 19440 | |
| }, | |
| { | |
| "epoch": 38.94, | |
| "grad_norm": 72.14510345458984, | |
| "learning_rate": 2.7440545777907747e-06, | |
| "loss": 0.4157, | |
| "step": 19470 | |
| }, | |
| { | |
| "epoch": 39.0, | |
| "grad_norm": 117.51093292236328, | |
| "learning_rate": 2.7300475013022666e-06, | |
| "loss": 0.4286, | |
| "step": 19500 | |
| }, | |
| { | |
| "epoch": 39.0, | |
| "eval_loss": 0.3193737268447876, | |
| "eval_map": 0.7717, | |
| "eval_map_50": 0.9567, | |
| "eval_map_75": 0.8986, | |
| "eval_map_chicken": 0.7752, | |
| "eval_map_duck": 0.7223, | |
| "eval_map_large": 0.8779, | |
| "eval_map_medium": 0.7677, | |
| "eval_map_plant": 0.8175, | |
| "eval_map_small": 0.1626, | |
| "eval_mar_1": 0.3033, | |
| "eval_mar_10": 0.8101, | |
| "eval_mar_100": 0.8139, | |
| "eval_mar_100_chicken": 0.8204, | |
| "eval_mar_100_duck": 0.7598, | |
| "eval_mar_100_plant": 0.8614, | |
| "eval_mar_large": 0.918, | |
| "eval_mar_medium": 0.82, | |
| "eval_mar_small": 0.2835, | |
| "eval_runtime": 5.0222, | |
| "eval_samples_per_second": 19.912, | |
| "eval_steps_per_second": 2.589, | |
| "step": 19500 | |
| }, | |
| { | |
| "epoch": 39.06, | |
| "grad_norm": 57.47135925292969, | |
| "learning_rate": 2.716062828328502e-06, | |
| "loss": 0.4059, | |
| "step": 19530 | |
| }, | |
| { | |
| "epoch": 39.12, | |
| "grad_norm": 28.567541122436523, | |
| "learning_rate": 2.7021006968925613e-06, | |
| "loss": 0.4349, | |
| "step": 19560 | |
| }, | |
| { | |
| "epoch": 39.18, | |
| "grad_norm": 38.38969039916992, | |
| "learning_rate": 2.6881612447950425e-06, | |
| "loss": 0.3861, | |
| "step": 19590 | |
| }, | |
| { | |
| "epoch": 39.24, | |
| "grad_norm": 62.003700256347656, | |
| "learning_rate": 2.6742446096127086e-06, | |
| "loss": 0.3273, | |
| "step": 19620 | |
| }, | |
| { | |
| "epoch": 39.3, | |
| "grad_norm": 33.330116271972656, | |
| "learning_rate": 2.6603509286971342e-06, | |
| "loss": 0.478, | |
| "step": 19650 | |
| }, | |
| { | |
| "epoch": 39.36, | |
| "grad_norm": 102.9520492553711, | |
| "learning_rate": 2.646480339173337e-06, | |
| "loss": 0.4294, | |
| "step": 19680 | |
| }, | |
| { | |
| "epoch": 39.42, | |
| "grad_norm": 38.9061393737793, | |
| "learning_rate": 2.6326329779384397e-06, | |
| "loss": 0.4171, | |
| "step": 19710 | |
| }, | |
| { | |
| "epoch": 39.48, | |
| "grad_norm": 33.986854553222656, | |
| "learning_rate": 2.618808981660304e-06, | |
| "loss": 0.487, | |
| "step": 19740 | |
| }, | |
| { | |
| "epoch": 39.54, | |
| "grad_norm": 2313.75537109375, | |
| "learning_rate": 2.6050084867761953e-06, | |
| "loss": 0.5018, | |
| "step": 19770 | |
| }, | |
| { | |
| "epoch": 39.6, | |
| "grad_norm": 40.18400192260742, | |
| "learning_rate": 2.5912316294914232e-06, | |
| "loss": 0.4993, | |
| "step": 19800 | |
| }, | |
| { | |
| "epoch": 39.66, | |
| "grad_norm": 58.73672103881836, | |
| "learning_rate": 2.5774785457780107e-06, | |
| "loss": 0.4466, | |
| "step": 19830 | |
| }, | |
| { | |
| "epoch": 39.72, | |
| "grad_norm": 113.0173110961914, | |
| "learning_rate": 2.5637493713733376e-06, | |
| "loss": 0.4275, | |
| "step": 19860 | |
| }, | |
| { | |
| "epoch": 39.78, | |
| "grad_norm": 62.983646392822266, | |
| "learning_rate": 2.5500442417788176e-06, | |
| "loss": 0.4815, | |
| "step": 19890 | |
| }, | |
| { | |
| "epoch": 39.84, | |
| "grad_norm": 41.0040397644043, | |
| "learning_rate": 2.536363292258543e-06, | |
| "loss": 0.4262, | |
| "step": 19920 | |
| }, | |
| { | |
| "epoch": 39.9, | |
| "grad_norm": 24.92264175415039, | |
| "learning_rate": 2.5227066578379624e-06, | |
| "loss": 0.5684, | |
| "step": 19950 | |
| }, | |
| { | |
| "epoch": 39.96, | |
| "grad_norm": 57.32313537597656, | |
| "learning_rate": 2.509074473302546e-06, | |
| "loss": 0.4488, | |
| "step": 19980 | |
| }, | |
| { | |
| "epoch": 40.0, | |
| "eval_loss": 0.3184027671813965, | |
| "eval_map": 0.7718, | |
| "eval_map_50": 0.9566, | |
| "eval_map_75": 0.9047, | |
| "eval_map_chicken": 0.7776, | |
| "eval_map_duck": 0.7156, | |
| "eval_map_large": 0.8809, | |
| "eval_map_medium": 0.7702, | |
| "eval_map_plant": 0.8223, | |
| "eval_map_small": 0.1921, | |
| "eval_mar_1": 0.2999, | |
| "eval_mar_10": 0.8092, | |
| "eval_mar_100": 0.8141, | |
| "eval_mar_100_chicken": 0.8187, | |
| "eval_mar_100_duck": 0.7546, | |
| "eval_mar_100_plant": 0.8689, | |
| "eval_mar_large": 0.9192, | |
| "eval_mar_medium": 0.8238, | |
| "eval_mar_small": 0.3002, | |
| "eval_runtime": 5.1382, | |
| "eval_samples_per_second": 19.462, | |
| "eval_steps_per_second": 2.53, | |
| "step": 20000 | |
| }, | |
| { | |
| "epoch": 40.02, | |
| "grad_norm": 62.7071533203125, | |
| "learning_rate": 2.49546687319645e-06, | |
| "loss": 0.4726, | |
| "step": 20010 | |
| }, | |
| { | |
| "epoch": 40.08, | |
| "grad_norm": 87.53660583496094, | |
| "learning_rate": 2.4818839918211963e-06, | |
| "loss": 0.3857, | |
| "step": 20040 | |
| }, | |
| { | |
| "epoch": 40.14, | |
| "grad_norm": 34.03664016723633, | |
| "learning_rate": 2.4683259632343363e-06, | |
| "loss": 0.3887, | |
| "step": 20070 | |
| }, | |
| { | |
| "epoch": 40.2, | |
| "grad_norm": 67.48214721679688, | |
| "learning_rate": 2.4547929212481436e-06, | |
| "loss": 0.4193, | |
| "step": 20100 | |
| }, | |
| { | |
| "epoch": 40.26, | |
| "grad_norm": 72.2217788696289, | |
| "learning_rate": 2.4412849994282744e-06, | |
| "loss": 0.4498, | |
| "step": 20130 | |
| }, | |
| { | |
| "epoch": 40.32, | |
| "grad_norm": 54.74772262573242, | |
| "learning_rate": 2.4278023310924676e-06, | |
| "loss": 0.4496, | |
| "step": 20160 | |
| }, | |
| { | |
| "epoch": 40.38, | |
| "grad_norm": 26.028186798095703, | |
| "learning_rate": 2.4143450493092146e-06, | |
| "loss": 0.4078, | |
| "step": 20190 | |
| }, | |
| { | |
| "epoch": 40.44, | |
| "grad_norm": 71.3542709350586, | |
| "learning_rate": 2.4009132868964525e-06, | |
| "loss": 0.5182, | |
| "step": 20220 | |
| }, | |
| { | |
| "epoch": 40.5, | |
| "grad_norm": 36.26480484008789, | |
| "learning_rate": 2.387507176420256e-06, | |
| "loss": 0.3948, | |
| "step": 20250 | |
| }, | |
| { | |
| "epoch": 40.56, | |
| "grad_norm": 61.92445373535156, | |
| "learning_rate": 2.3741268501935212e-06, | |
| "loss": 0.4554, | |
| "step": 20280 | |
| }, | |
| { | |
| "epoch": 40.62, | |
| "grad_norm": 85.12369537353516, | |
| "learning_rate": 2.3607724402746685e-06, | |
| "loss": 0.4314, | |
| "step": 20310 | |
| }, | |
| { | |
| "epoch": 40.68, | |
| "grad_norm": 42.598114013671875, | |
| "learning_rate": 2.3474440784663287e-06, | |
| "loss": 0.4402, | |
| "step": 20340 | |
| }, | |
| { | |
| "epoch": 40.74, | |
| "grad_norm": 29.042329788208008, | |
| "learning_rate": 2.334141896314057e-06, | |
| "loss": 0.4106, | |
| "step": 20370 | |
| }, | |
| { | |
| "epoch": 40.8, | |
| "grad_norm": 38.11594772338867, | |
| "learning_rate": 2.320866025105016e-06, | |
| "loss": 0.4315, | |
| "step": 20400 | |
| }, | |
| { | |
| "epoch": 40.86, | |
| "grad_norm": 61.246238708496094, | |
| "learning_rate": 2.3076165958666992e-06, | |
| "loss": 0.3815, | |
| "step": 20430 | |
| }, | |
| { | |
| "epoch": 40.92, | |
| "grad_norm": 31.141773223876953, | |
| "learning_rate": 2.294393739365621e-06, | |
| "loss": 0.4318, | |
| "step": 20460 | |
| }, | |
| { | |
| "epoch": 40.98, | |
| "grad_norm": 45.657249450683594, | |
| "learning_rate": 2.281197586106037e-06, | |
| "loss": 0.3763, | |
| "step": 20490 | |
| }, | |
| { | |
| "epoch": 41.0, | |
| "eval_loss": 0.30546829104423523, | |
| "eval_map": 0.7876, | |
| "eval_map_50": 0.956, | |
| "eval_map_75": 0.9186, | |
| "eval_map_chicken": 0.7973, | |
| "eval_map_duck": 0.7394, | |
| "eval_map_large": 0.8844, | |
| "eval_map_medium": 0.7824, | |
| "eval_map_plant": 0.8262, | |
| "eval_map_small": 0.1841, | |
| "eval_mar_1": 0.3033, | |
| "eval_mar_10": 0.8207, | |
| "eval_mar_100": 0.8254, | |
| "eval_mar_100_chicken": 0.8356, | |
| "eval_mar_100_duck": 0.7691, | |
| "eval_mar_100_plant": 0.8715, | |
| "eval_mar_large": 0.9234, | |
| "eval_mar_medium": 0.8339, | |
| "eval_mar_small": 0.3061, | |
| "eval_runtime": 5.1002, | |
| "eval_samples_per_second": 19.607, | |
| "eval_steps_per_second": 2.549, | |
| "step": 20500 | |
| }, | |
| { | |
| "epoch": 41.04, | |
| "grad_norm": 45.896121978759766, | |
| "learning_rate": 2.268028266328655e-06, | |
| "loss": 0.4273, | |
| "step": 20520 | |
| }, | |
| { | |
| "epoch": 41.1, | |
| "grad_norm": 25.360084533691406, | |
| "learning_rate": 2.254885910009341e-06, | |
| "loss": 0.4682, | |
| "step": 20550 | |
| }, | |
| { | |
| "epoch": 41.16, | |
| "grad_norm": 63.324825286865234, | |
| "learning_rate": 2.2417706468578495e-06, | |
| "loss": 0.6033, | |
| "step": 20580 | |
| }, | |
| { | |
| "epoch": 41.22, | |
| "grad_norm": 129.3037109375, | |
| "learning_rate": 2.228682606316529e-06, | |
| "loss": 0.5746, | |
| "step": 20610 | |
| }, | |
| { | |
| "epoch": 41.28, | |
| "grad_norm": 52.51323699951172, | |
| "learning_rate": 2.2156219175590623e-06, | |
| "loss": 0.4579, | |
| "step": 20640 | |
| }, | |
| { | |
| "epoch": 41.34, | |
| "grad_norm": 36.566917419433594, | |
| "learning_rate": 2.2025887094891657e-06, | |
| "loss": 0.4206, | |
| "step": 20670 | |
| }, | |
| { | |
| "epoch": 41.4, | |
| "grad_norm": 68.68453979492188, | |
| "learning_rate": 2.1895831107393485e-06, | |
| "loss": 0.4617, | |
| "step": 20700 | |
| }, | |
| { | |
| "epoch": 41.46, | |
| "grad_norm": 133.7510223388672, | |
| "learning_rate": 2.1766052496696155e-06, | |
| "loss": 0.5263, | |
| "step": 20730 | |
| }, | |
| { | |
| "epoch": 41.52, | |
| "grad_norm": 76.5693130493164, | |
| "learning_rate": 2.1636552543662187e-06, | |
| "loss": 0.4473, | |
| "step": 20760 | |
| }, | |
| { | |
| "epoch": 41.58, | |
| "grad_norm": 60.44133377075195, | |
| "learning_rate": 2.1507332526403814e-06, | |
| "loss": 0.4135, | |
| "step": 20790 | |
| }, | |
| { | |
| "epoch": 41.64, | |
| "grad_norm": 117.59391021728516, | |
| "learning_rate": 2.137839372027047e-06, | |
| "loss": 0.4654, | |
| "step": 20820 | |
| }, | |
| { | |
| "epoch": 41.7, | |
| "grad_norm": 45.36744689941406, | |
| "learning_rate": 2.124973739783609e-06, | |
| "loss": 0.4071, | |
| "step": 20850 | |
| }, | |
| { | |
| "epoch": 41.76, | |
| "grad_norm": 57.2125358581543, | |
| "learning_rate": 2.112136482888663e-06, | |
| "loss": 0.3865, | |
| "step": 20880 | |
| }, | |
| { | |
| "epoch": 41.82, | |
| "grad_norm": 60.10706329345703, | |
| "learning_rate": 2.0993277280407547e-06, | |
| "loss": 0.4395, | |
| "step": 20910 | |
| }, | |
| { | |
| "epoch": 41.88, | |
| "grad_norm": 76.14381408691406, | |
| "learning_rate": 2.0865476016571206e-06, | |
| "loss": 0.4623, | |
| "step": 20940 | |
| }, | |
| { | |
| "epoch": 41.94, | |
| "grad_norm": 47.56647491455078, | |
| "learning_rate": 2.0737962298724513e-06, | |
| "loss": 0.3841, | |
| "step": 20970 | |
| }, | |
| { | |
| "epoch": 42.0, | |
| "grad_norm": 85.08351135253906, | |
| "learning_rate": 2.061073738537635e-06, | |
| "loss": 0.5658, | |
| "step": 21000 | |
| }, | |
| { | |
| "epoch": 42.0, | |
| "eval_loss": 0.30137816071510315, | |
| "eval_map": 0.791, | |
| "eval_map_50": 0.9594, | |
| "eval_map_75": 0.9167, | |
| "eval_map_chicken": 0.805, | |
| "eval_map_duck": 0.7421, | |
| "eval_map_large": 0.8786, | |
| "eval_map_medium": 0.7911, | |
| "eval_map_plant": 0.8259, | |
| "eval_map_small": 0.2095, | |
| "eval_mar_1": 0.3032, | |
| "eval_mar_10": 0.8244, | |
| "eval_mar_100": 0.8299, | |
| "eval_mar_100_chicken": 0.8404, | |
| "eval_mar_100_duck": 0.7742, | |
| "eval_mar_100_plant": 0.8749, | |
| "eval_mar_large": 0.9213, | |
| "eval_mar_medium": 0.8403, | |
| "eval_mar_small": 0.3403, | |
| "eval_runtime": 5.0741, | |
| "eval_samples_per_second": 19.708, | |
| "eval_steps_per_second": 2.562, | |
| "step": 21000 | |
| }, | |
| { | |
| "epoch": 42.06, | |
| "grad_norm": 41.583213806152344, | |
| "learning_rate": 2.0483802532185286e-06, | |
| "loss": 0.4022, | |
| "step": 21030 | |
| }, | |
| { | |
| "epoch": 42.12, | |
| "grad_norm": 320.62518310546875, | |
| "learning_rate": 2.035715899194704e-06, | |
| "loss": 0.4408, | |
| "step": 21060 | |
| }, | |
| { | |
| "epoch": 42.18, | |
| "grad_norm": 42.51158142089844, | |
| "learning_rate": 2.0230808014582265e-06, | |
| "loss": 0.4566, | |
| "step": 21090 | |
| }, | |
| { | |
| "epoch": 42.24, | |
| "grad_norm": 0.00034786213655024767, | |
| "learning_rate": 2.0104750847124075e-06, | |
| "loss": 0.5319, | |
| "step": 21120 | |
| }, | |
| { | |
| "epoch": 42.3, | |
| "grad_norm": 93.6717529296875, | |
| "learning_rate": 1.9978988733705807e-06, | |
| "loss": 0.4083, | |
| "step": 21150 | |
| }, | |
| { | |
| "epoch": 42.36, | |
| "grad_norm": 52.52644348144531, | |
| "learning_rate": 1.9853522915548777e-06, | |
| "loss": 0.4354, | |
| "step": 21180 | |
| }, | |
| { | |
| "epoch": 42.42, | |
| "grad_norm": 36.161460876464844, | |
| "learning_rate": 1.9728354630949935e-06, | |
| "loss": 0.4591, | |
| "step": 21210 | |
| }, | |
| { | |
| "epoch": 42.48, | |
| "grad_norm": 60.393306732177734, | |
| "learning_rate": 1.9603485115269743e-06, | |
| "loss": 0.5063, | |
| "step": 21240 | |
| }, | |
| { | |
| "epoch": 42.54, | |
| "grad_norm": 223.328857421875, | |
| "learning_rate": 1.9478915600919877e-06, | |
| "loss": 0.4404, | |
| "step": 21270 | |
| }, | |
| { | |
| "epoch": 42.6, | |
| "grad_norm": 42.48649215698242, | |
| "learning_rate": 1.9354647317351187e-06, | |
| "loss": 0.4508, | |
| "step": 21300 | |
| }, | |
| { | |
| "epoch": 42.66, | |
| "grad_norm": 46.062191009521484, | |
| "learning_rate": 1.9230681491041425e-06, | |
| "loss": 0.4275, | |
| "step": 21330 | |
| }, | |
| { | |
| "epoch": 42.72, | |
| "grad_norm": 0.0008085273439064622, | |
| "learning_rate": 1.910701934548329e-06, | |
| "loss": 0.4588, | |
| "step": 21360 | |
| }, | |
| { | |
| "epoch": 42.78, | |
| "grad_norm": 44.36118698120117, | |
| "learning_rate": 1.8983662101172217e-06, | |
| "loss": 0.4563, | |
| "step": 21390 | |
| }, | |
| { | |
| "epoch": 42.84, | |
| "grad_norm": 139.61676025390625, | |
| "learning_rate": 1.8860610975594384e-06, | |
| "loss": 0.3606, | |
| "step": 21420 | |
| }, | |
| { | |
| "epoch": 42.9, | |
| "grad_norm": 37.38964080810547, | |
| "learning_rate": 1.873786718321476e-06, | |
| "loss": 0.4464, | |
| "step": 21450 | |
| }, | |
| { | |
| "epoch": 42.96, | |
| "grad_norm": 72.3614730834961, | |
| "learning_rate": 1.8615431935464984e-06, | |
| "loss": 0.4322, | |
| "step": 21480 | |
| }, | |
| { | |
| "epoch": 43.0, | |
| "eval_loss": 0.29740211367607117, | |
| "eval_map": 0.7974, | |
| "eval_map_50": 0.9595, | |
| "eval_map_75": 0.9169, | |
| "eval_map_chicken": 0.8005, | |
| "eval_map_duck": 0.7586, | |
| "eval_map_large": 0.8951, | |
| "eval_map_medium": 0.7927, | |
| "eval_map_plant": 0.8331, | |
| "eval_map_small": 0.1879, | |
| "eval_mar_1": 0.3078, | |
| "eval_mar_10": 0.8296, | |
| "eval_mar_100": 0.834, | |
| "eval_mar_100_chicken": 0.8369, | |
| "eval_mar_100_duck": 0.7876, | |
| "eval_mar_100_plant": 0.8775, | |
| "eval_mar_large": 0.931, | |
| "eval_mar_medium": 0.8411, | |
| "eval_mar_small": 0.3085, | |
| "eval_runtime": 5.6419, | |
| "eval_samples_per_second": 17.724, | |
| "eval_steps_per_second": 2.304, | |
| "step": 21500 | |
| }, | |
| { | |
| "epoch": 43.02, | |
| "grad_norm": 44.15309143066406, | |
| "learning_rate": 1.8493306440731557e-06, | |
| "loss": 0.4896, | |
| "step": 21510 | |
| }, | |
| { | |
| "epoch": 43.08, | |
| "grad_norm": 35.48102951049805, | |
| "learning_rate": 1.837149190434378e-06, | |
| "loss": 0.5013, | |
| "step": 21540 | |
| }, | |
| { | |
| "epoch": 43.14, | |
| "grad_norm": 35.87041473388672, | |
| "learning_rate": 1.824998952856198e-06, | |
| "loss": 0.4439, | |
| "step": 21570 | |
| }, | |
| { | |
| "epoch": 43.2, | |
| "grad_norm": 26.650522232055664, | |
| "learning_rate": 1.8128800512565514e-06, | |
| "loss": 0.4186, | |
| "step": 21600 | |
| }, | |
| { | |
| "epoch": 43.26, | |
| "grad_norm": 48.30433654785156, | |
| "learning_rate": 1.800792605244109e-06, | |
| "loss": 0.4345, | |
| "step": 21630 | |
| }, | |
| { | |
| "epoch": 43.32, | |
| "grad_norm": 70.63233184814453, | |
| "learning_rate": 1.7887367341170781e-06, | |
| "loss": 0.4639, | |
| "step": 21660 | |
| }, | |
| { | |
| "epoch": 43.38, | |
| "grad_norm": 29.950109481811523, | |
| "learning_rate": 1.7767125568620442e-06, | |
| "loss": 0.3931, | |
| "step": 21690 | |
| }, | |
| { | |
| "epoch": 43.44, | |
| "grad_norm": 40.98532485961914, | |
| "learning_rate": 1.7647201921527802e-06, | |
| "loss": 0.3728, | |
| "step": 21720 | |
| }, | |
| { | |
| "epoch": 43.5, | |
| "grad_norm": 107.21430969238281, | |
| "learning_rate": 1.7527597583490825e-06, | |
| "loss": 0.3679, | |
| "step": 21750 | |
| }, | |
| { | |
| "epoch": 43.56, | |
| "grad_norm": 41.53556442260742, | |
| "learning_rate": 1.7408313734956074e-06, | |
| "loss": 0.4278, | |
| "step": 21780 | |
| }, | |
| { | |
| "epoch": 43.62, | |
| "grad_norm": 30.475116729736328, | |
| "learning_rate": 1.7289351553206952e-06, | |
| "loss": 0.4995, | |
| "step": 21810 | |
| }, | |
| { | |
| "epoch": 43.68, | |
| "grad_norm": 65.69325256347656, | |
| "learning_rate": 1.7170712212352187e-06, | |
| "loss": 0.4236, | |
| "step": 21840 | |
| }, | |
| { | |
| "epoch": 43.74, | |
| "grad_norm": 83.24858856201172, | |
| "learning_rate": 1.7052396883314154e-06, | |
| "loss": 0.4865, | |
| "step": 21870 | |
| }, | |
| { | |
| "epoch": 43.8, | |
| "grad_norm": 49.75985336303711, | |
| "learning_rate": 1.6934406733817417e-06, | |
| "loss": 0.4061, | |
| "step": 21900 | |
| }, | |
| { | |
| "epoch": 43.86, | |
| "grad_norm": 0.0012297651264816523, | |
| "learning_rate": 1.6816742928377072e-06, | |
| "loss": 0.4391, | |
| "step": 21930 | |
| }, | |
| { | |
| "epoch": 43.92, | |
| "grad_norm": 107.39283752441406, | |
| "learning_rate": 1.6699406628287423e-06, | |
| "loss": 0.4027, | |
| "step": 21960 | |
| }, | |
| { | |
| "epoch": 43.98, | |
| "grad_norm": 53.299110412597656, | |
| "learning_rate": 1.658239899161036e-06, | |
| "loss": 0.7057, | |
| "step": 21990 | |
| }, | |
| { | |
| "epoch": 44.0, | |
| "eval_loss": 0.3092344403266907, | |
| "eval_map": 0.7822, | |
| "eval_map_50": 0.9563, | |
| "eval_map_75": 0.9171, | |
| "eval_map_chicken": 0.7941, | |
| "eval_map_duck": 0.7348, | |
| "eval_map_large": 0.8688, | |
| "eval_map_medium": 0.7813, | |
| "eval_map_plant": 0.8177, | |
| "eval_map_small": 0.1985, | |
| "eval_mar_1": 0.3003, | |
| "eval_mar_10": 0.8165, | |
| "eval_mar_100": 0.821, | |
| "eval_mar_100_chicken": 0.8307, | |
| "eval_mar_100_duck": 0.766, | |
| "eval_mar_100_plant": 0.8663, | |
| "eval_mar_large": 0.9117, | |
| "eval_mar_medium": 0.8292, | |
| "eval_mar_small": 0.3663, | |
| "eval_runtime": 5.0773, | |
| "eval_samples_per_second": 19.696, | |
| "eval_steps_per_second": 2.56, | |
| "step": 22000 | |
| }, | |
| { | |
| "epoch": 44.04, | |
| "grad_norm": 42.234928131103516, | |
| "learning_rate": 1.6465721173164e-06, | |
| "loss": 0.4139, | |
| "step": 22020 | |
| }, | |
| { | |
| "epoch": 44.1, | |
| "grad_norm": 57.58454513549805, | |
| "learning_rate": 1.6349374324511347e-06, | |
| "loss": 0.4395, | |
| "step": 22050 | |
| }, | |
| { | |
| "epoch": 44.16, | |
| "grad_norm": 57.831993103027344, | |
| "learning_rate": 1.6233359593948777e-06, | |
| "loss": 0.4128, | |
| "step": 22080 | |
| }, | |
| { | |
| "epoch": 44.22, | |
| "grad_norm": 57.68543243408203, | |
| "learning_rate": 1.6117678126494895e-06, | |
| "loss": 0.4225, | |
| "step": 22110 | |
| }, | |
| { | |
| "epoch": 44.28, | |
| "grad_norm": 168.56692504882812, | |
| "learning_rate": 1.600233106387904e-06, | |
| "loss": 0.423, | |
| "step": 22140 | |
| }, | |
| { | |
| "epoch": 44.34, | |
| "grad_norm": 39.12761306762695, | |
| "learning_rate": 1.5887319544530182e-06, | |
| "loss": 0.3617, | |
| "step": 22170 | |
| }, | |
| { | |
| "epoch": 44.4, | |
| "grad_norm": 41.760311126708984, | |
| "learning_rate": 1.5772644703565564e-06, | |
| "loss": 0.6029, | |
| "step": 22200 | |
| }, | |
| { | |
| "epoch": 44.46, | |
| "grad_norm": 79.56449127197266, | |
| "learning_rate": 1.5658307672779594e-06, | |
| "loss": 0.3154, | |
| "step": 22230 | |
| }, | |
| { | |
| "epoch": 44.52, | |
| "grad_norm": 35.965431213378906, | |
| "learning_rate": 1.554430958063259e-06, | |
| "loss": 0.4407, | |
| "step": 22260 | |
| }, | |
| { | |
| "epoch": 44.58, | |
| "grad_norm": 54.796878814697266, | |
| "learning_rate": 1.5430651552239684e-06, | |
| "loss": 0.4, | |
| "step": 22290 | |
| }, | |
| { | |
| "epoch": 44.64, | |
| "grad_norm": 64.1689682006836, | |
| "learning_rate": 1.531733470935976e-06, | |
| "loss": 0.4755, | |
| "step": 22320 | |
| }, | |
| { | |
| "epoch": 44.7, | |
| "grad_norm": 36.782920837402344, | |
| "learning_rate": 1.5204360170384286e-06, | |
| "loss": 0.4451, | |
| "step": 22350 | |
| }, | |
| { | |
| "epoch": 44.76, | |
| "grad_norm": 38.28982162475586, | |
| "learning_rate": 1.5091729050326376e-06, | |
| "loss": 0.4116, | |
| "step": 22380 | |
| }, | |
| { | |
| "epoch": 44.82, | |
| "grad_norm": 36.69734191894531, | |
| "learning_rate": 1.4979442460809684e-06, | |
| "loss": 0.4304, | |
| "step": 22410 | |
| }, | |
| { | |
| "epoch": 44.88, | |
| "grad_norm": 34.10408020019531, | |
| "learning_rate": 1.4867501510057548e-06, | |
| "loss": 0.4208, | |
| "step": 22440 | |
| }, | |
| { | |
| "epoch": 44.94, | |
| "grad_norm": 56.05753707885742, | |
| "learning_rate": 1.4755907302881927e-06, | |
| "loss": 0.4207, | |
| "step": 22470 | |
| }, | |
| { | |
| "epoch": 45.0, | |
| "grad_norm": 79.50679779052734, | |
| "learning_rate": 1.4644660940672628e-06, | |
| "loss": 0.4096, | |
| "step": 22500 | |
| }, | |
| { | |
| "epoch": 45.0, | |
| "eval_loss": 0.29910051822662354, | |
| "eval_map": 0.7899, | |
| "eval_map_50": 0.9614, | |
| "eval_map_75": 0.9121, | |
| "eval_map_chicken": 0.8016, | |
| "eval_map_duck": 0.7502, | |
| "eval_map_large": 0.8747, | |
| "eval_map_medium": 0.7852, | |
| "eval_map_plant": 0.8179, | |
| "eval_map_small": 0.2212, | |
| "eval_mar_1": 0.3031, | |
| "eval_mar_10": 0.8233, | |
| "eval_mar_100": 0.8286, | |
| "eval_mar_100_chicken": 0.8413, | |
| "eval_mar_100_duck": 0.7794, | |
| "eval_mar_100_plant": 0.8651, | |
| "eval_mar_large": 0.9142, | |
| "eval_mar_medium": 0.8351, | |
| "eval_mar_small": 0.3578, | |
| "eval_runtime": 5.1133, | |
| "eval_samples_per_second": 19.557, | |
| "eval_steps_per_second": 2.542, | |
| "step": 22500 | |
| }, | |
| { | |
| "epoch": 45.06, | |
| "grad_norm": 38.6081428527832, | |
| "learning_rate": 1.4533763521386319e-06, | |
| "loss": 0.392, | |
| "step": 22530 | |
| }, | |
| { | |
| "epoch": 45.12, | |
| "grad_norm": 28.41141128540039, | |
| "learning_rate": 1.4423216139535735e-06, | |
| "loss": 0.4145, | |
| "step": 22560 | |
| }, | |
| { | |
| "epoch": 45.18, | |
| "grad_norm": 91.0741195678711, | |
| "learning_rate": 1.4313019886178942e-06, | |
| "loss": 0.4022, | |
| "step": 22590 | |
| }, | |
| { | |
| "epoch": 45.24, | |
| "grad_norm": 88.55020141601562, | |
| "learning_rate": 1.420317584890844e-06, | |
| "loss": 0.5098, | |
| "step": 22620 | |
| }, | |
| { | |
| "epoch": 45.3, | |
| "grad_norm": 59.97364044189453, | |
| "learning_rate": 1.4093685111840567e-06, | |
| "loss": 0.4416, | |
| "step": 22650 | |
| }, | |
| { | |
| "epoch": 45.36, | |
| "grad_norm": 138.7408447265625, | |
| "learning_rate": 1.3984548755604655e-06, | |
| "loss": 0.4223, | |
| "step": 22680 | |
| }, | |
| { | |
| "epoch": 45.42, | |
| "grad_norm": 41.95036315917969, | |
| "learning_rate": 1.3875767857332512e-06, | |
| "loss": 0.4241, | |
| "step": 22710 | |
| }, | |
| { | |
| "epoch": 45.48, | |
| "grad_norm": 35.16947937011719, | |
| "learning_rate": 1.3767343490647668e-06, | |
| "loss": 0.4442, | |
| "step": 22740 | |
| }, | |
| { | |
| "epoch": 45.54, | |
| "grad_norm": 69.71556091308594, | |
| "learning_rate": 1.3659276725654863e-06, | |
| "loss": 0.3552, | |
| "step": 22770 | |
| }, | |
| { | |
| "epoch": 45.6, | |
| "grad_norm": 95.909423828125, | |
| "learning_rate": 1.3551568628929434e-06, | |
| "loss": 0.4363, | |
| "step": 22800 | |
| }, | |
| { | |
| "epoch": 45.66, | |
| "grad_norm": 21.66033935546875, | |
| "learning_rate": 1.3444220263506797e-06, | |
| "loss": 0.4363, | |
| "step": 22830 | |
| }, | |
| { | |
| "epoch": 45.72, | |
| "grad_norm": 131.8041229248047, | |
| "learning_rate": 1.333723268887201e-06, | |
| "loss": 0.4592, | |
| "step": 22860 | |
| }, | |
| { | |
| "epoch": 45.78, | |
| "grad_norm": 50.998779296875, | |
| "learning_rate": 1.3230606960949204e-06, | |
| "loss": 0.4649, | |
| "step": 22890 | |
| }, | |
| { | |
| "epoch": 45.84, | |
| "grad_norm": 29.139339447021484, | |
| "learning_rate": 1.312434413209131e-06, | |
| "loss": 0.4568, | |
| "step": 22920 | |
| }, | |
| { | |
| "epoch": 45.9, | |
| "grad_norm": 51.54317092895508, | |
| "learning_rate": 1.301844525106951e-06, | |
| "loss": 0.4308, | |
| "step": 22950 | |
| }, | |
| { | |
| "epoch": 45.96, | |
| "grad_norm": 46.99152374267578, | |
| "learning_rate": 1.2912911363063048e-06, | |
| "loss": 0.4854, | |
| "step": 22980 | |
| }, | |
| { | |
| "epoch": 46.0, | |
| "eval_loss": 0.3003343343734741, | |
| "eval_map": 0.7815, | |
| "eval_map_50": 0.9595, | |
| "eval_map_75": 0.9068, | |
| "eval_map_chicken": 0.7816, | |
| "eval_map_duck": 0.746, | |
| "eval_map_large": 0.8747, | |
| "eval_map_medium": 0.7791, | |
| "eval_map_plant": 0.8169, | |
| "eval_map_small": 0.2042, | |
| "eval_mar_1": 0.3016, | |
| "eval_mar_10": 0.8164, | |
| "eval_mar_100": 0.8197, | |
| "eval_mar_100_chicken": 0.8231, | |
| "eval_mar_100_duck": 0.7722, | |
| "eval_mar_100_plant": 0.8637, | |
| "eval_mar_large": 0.9163, | |
| "eval_mar_medium": 0.8255, | |
| "eval_mar_small": 0.3258, | |
| "eval_runtime": 5.0277, | |
| "eval_samples_per_second": 19.89, | |
| "eval_steps_per_second": 2.586, | |
| "step": 23000 | |
| }, | |
| { | |
| "epoch": 46.02, | |
| "grad_norm": 82.24079132080078, | |
| "learning_rate": 1.2807743509648745e-06, | |
| "loss": 0.459, | |
| "step": 23010 | |
| }, | |
| { | |
| "epoch": 46.08, | |
| "grad_norm": 35.353336334228516, | |
| "learning_rate": 1.2702942728790897e-06, | |
| "loss": 0.4215, | |
| "step": 23040 | |
| }, | |
| { | |
| "epoch": 46.14, | |
| "grad_norm": 38.322479248046875, | |
| "learning_rate": 1.2598510054830888e-06, | |
| "loss": 0.3925, | |
| "step": 23070 | |
| }, | |
| { | |
| "epoch": 46.2, | |
| "grad_norm": 53.787445068359375, | |
| "learning_rate": 1.2494446518477022e-06, | |
| "loss": 0.349, | |
| "step": 23100 | |
| }, | |
| { | |
| "epoch": 46.26, | |
| "grad_norm": 37.92025375366211, | |
| "learning_rate": 1.2390753146794438e-06, | |
| "loss": 0.3578, | |
| "step": 23130 | |
| }, | |
| { | |
| "epoch": 46.32, | |
| "grad_norm": 68.19427490234375, | |
| "learning_rate": 1.2287430963194807e-06, | |
| "loss": 0.4342, | |
| "step": 23160 | |
| }, | |
| { | |
| "epoch": 46.38, | |
| "grad_norm": 108.89007568359375, | |
| "learning_rate": 1.218448098742641e-06, | |
| "loss": 0.4607, | |
| "step": 23190 | |
| }, | |
| { | |
| "epoch": 46.44, | |
| "grad_norm": 83.83390808105469, | |
| "learning_rate": 1.2081904235563908e-06, | |
| "loss": 0.4169, | |
| "step": 23220 | |
| }, | |
| { | |
| "epoch": 46.5, | |
| "grad_norm": 46.47811508178711, | |
| "learning_rate": 1.1979701719998454e-06, | |
| "loss": 0.3965, | |
| "step": 23250 | |
| }, | |
| { | |
| "epoch": 46.56, | |
| "grad_norm": 83.35205841064453, | |
| "learning_rate": 1.18778744494276e-06, | |
| "loss": 0.685, | |
| "step": 23280 | |
| }, | |
| { | |
| "epoch": 46.62, | |
| "grad_norm": 61.895530700683594, | |
| "learning_rate": 1.1776423428845423e-06, | |
| "loss": 0.3723, | |
| "step": 23310 | |
| }, | |
| { | |
| "epoch": 46.68, | |
| "grad_norm": 68.6615982055664, | |
| "learning_rate": 1.1675349659532514e-06, | |
| "loss": 0.4277, | |
| "step": 23340 | |
| }, | |
| { | |
| "epoch": 46.74, | |
| "grad_norm": 34.00867462158203, | |
| "learning_rate": 1.1574654139046171e-06, | |
| "loss": 0.4321, | |
| "step": 23370 | |
| }, | |
| { | |
| "epoch": 46.8, | |
| "grad_norm": 40.27052688598633, | |
| "learning_rate": 1.1474337861210543e-06, | |
| "loss": 0.3798, | |
| "step": 23400 | |
| }, | |
| { | |
| "epoch": 46.86, | |
| "grad_norm": 58.309085845947266, | |
| "learning_rate": 1.1374401816106778e-06, | |
| "loss": 0.4083, | |
| "step": 23430 | |
| }, | |
| { | |
| "epoch": 46.92, | |
| "grad_norm": 43.79540252685547, | |
| "learning_rate": 1.1274846990063314e-06, | |
| "loss": 0.3961, | |
| "step": 23460 | |
| }, | |
| { | |
| "epoch": 46.98, | |
| "grad_norm": 48.5699462890625, | |
| "learning_rate": 1.1175674365646067e-06, | |
| "loss": 0.4257, | |
| "step": 23490 | |
| }, | |
| { | |
| "epoch": 47.0, | |
| "eval_loss": 0.29506856203079224, | |
| "eval_map": 0.792, | |
| "eval_map_50": 0.9625, | |
| "eval_map_75": 0.9172, | |
| "eval_map_chicken": 0.7961, | |
| "eval_map_duck": 0.7572, | |
| "eval_map_large": 0.8802, | |
| "eval_map_medium": 0.7855, | |
| "eval_map_plant": 0.8226, | |
| "eval_map_small": 0.2075, | |
| "eval_mar_1": 0.3067, | |
| "eval_mar_10": 0.8262, | |
| "eval_mar_100": 0.8309, | |
| "eval_mar_100_chicken": 0.8338, | |
| "eval_mar_100_duck": 0.7907, | |
| "eval_mar_100_plant": 0.8683, | |
| "eval_mar_large": 0.9197, | |
| "eval_mar_medium": 0.836, | |
| "eval_mar_small": 0.3468, | |
| "eval_runtime": 5.0742, | |
| "eval_samples_per_second": 19.708, | |
| "eval_steps_per_second": 2.562, | |
| "step": 23500 | |
| }, | |
| { | |
| "epoch": 47.04, | |
| "grad_norm": 61.57620620727539, | |
| "learning_rate": 1.1076884921648834e-06, | |
| "loss": 0.3394, | |
| "step": 23520 | |
| }, | |
| { | |
| "epoch": 47.1, | |
| "grad_norm": 46.27333450317383, | |
| "learning_rate": 1.097847963308351e-06, | |
| "loss": 0.4013, | |
| "step": 23550 | |
| }, | |
| { | |
| "epoch": 47.16, | |
| "grad_norm": 75.60147857666016, | |
| "learning_rate": 1.0880459471170597e-06, | |
| "loss": 0.337, | |
| "step": 23580 | |
| }, | |
| { | |
| "epoch": 47.22, | |
| "grad_norm": 41.4156608581543, | |
| "learning_rate": 1.0782825403329488e-06, | |
| "loss": 0.3904, | |
| "step": 23610 | |
| }, | |
| { | |
| "epoch": 47.28, | |
| "grad_norm": 35.6782341003418, | |
| "learning_rate": 1.0685578393169054e-06, | |
| "loss": 0.4262, | |
| "step": 23640 | |
| }, | |
| { | |
| "epoch": 47.34, | |
| "grad_norm": 34.51017761230469, | |
| "learning_rate": 1.0588719400478004e-06, | |
| "loss": 0.4002, | |
| "step": 23670 | |
| }, | |
| { | |
| "epoch": 47.4, | |
| "grad_norm": 33.08877944946289, | |
| "learning_rate": 1.049224938121548e-06, | |
| "loss": 0.4203, | |
| "step": 23700 | |
| }, | |
| { | |
| "epoch": 47.46, | |
| "grad_norm": 32.152339935302734, | |
| "learning_rate": 1.0396169287501652e-06, | |
| "loss": 0.3937, | |
| "step": 23730 | |
| }, | |
| { | |
| "epoch": 47.52, | |
| "grad_norm": 53.18661117553711, | |
| "learning_rate": 1.0300480067608232e-06, | |
| "loss": 0.4445, | |
| "step": 23760 | |
| }, | |
| { | |
| "epoch": 47.58, | |
| "grad_norm": 40.39225387573242, | |
| "learning_rate": 1.020518266594921e-06, | |
| "loss": 0.4166, | |
| "step": 23790 | |
| }, | |
| { | |
| "epoch": 47.64, | |
| "grad_norm": 42.8251838684082, | |
| "learning_rate": 1.0110278023071445e-06, | |
| "loss": 0.414, | |
| "step": 23820 | |
| }, | |
| { | |
| "epoch": 47.7, | |
| "grad_norm": 64.0047607421875, | |
| "learning_rate": 1.0015767075645472e-06, | |
| "loss": 0.328, | |
| "step": 23850 | |
| }, | |
| { | |
| "epoch": 47.76, | |
| "grad_norm": 52.580780029296875, | |
| "learning_rate": 9.921650756456164e-07, | |
| "loss": 0.4109, | |
| "step": 23880 | |
| }, | |
| { | |
| "epoch": 47.82, | |
| "grad_norm": 59.67039489746094, | |
| "learning_rate": 9.82792999439362e-07, | |
| "loss": 0.4171, | |
| "step": 23910 | |
| }, | |
| { | |
| "epoch": 47.88, | |
| "grad_norm": 179.34124755859375, | |
| "learning_rate": 9.734605714443906e-07, | |
| "loss": 0.4743, | |
| "step": 23940 | |
| }, | |
| { | |
| "epoch": 47.94, | |
| "grad_norm": 53.743431091308594, | |
| "learning_rate": 9.641678837679985e-07, | |
| "loss": 0.4152, | |
| "step": 23970 | |
| }, | |
| { | |
| "epoch": 48.0, | |
| "grad_norm": 46.56321716308594, | |
| "learning_rate": 9.549150281252633e-07, | |
| "loss": 0.4033, | |
| "step": 24000 | |
| }, | |
| { | |
| "epoch": 48.0, | |
| "eval_loss": 0.28829237818717957, | |
| "eval_map": 0.7988, | |
| "eval_map_50": 0.9632, | |
| "eval_map_75": 0.9194, | |
| "eval_map_chicken": 0.8069, | |
| "eval_map_duck": 0.7649, | |
| "eval_map_large": 0.8765, | |
| "eval_map_medium": 0.7984, | |
| "eval_map_plant": 0.8246, | |
| "eval_map_small": 0.2266, | |
| "eval_mar_1": 0.3082, | |
| "eval_mar_10": 0.8343, | |
| "eval_mar_100": 0.8382, | |
| "eval_mar_100_chicken": 0.8458, | |
| "eval_mar_100_duck": 0.7969, | |
| "eval_mar_100_plant": 0.872, | |
| "eval_mar_large": 0.9176, | |
| "eval_mar_medium": 0.8477, | |
| "eval_mar_small": 0.3616, | |
| "eval_runtime": 5.0852, | |
| "eval_samples_per_second": 19.665, | |
| "eval_steps_per_second": 2.556, | |
| "step": 24000 | |
| }, | |
| { | |
| "epoch": 48.06, | |
| "grad_norm": 33.844398498535156, | |
| "learning_rate": 9.457020958381324e-07, | |
| "loss": 0.3829, | |
| "step": 24030 | |
| }, | |
| { | |
| "epoch": 48.12, | |
| "grad_norm": 75.94850158691406, | |
| "learning_rate": 9.365291778345303e-07, | |
| "loss": 0.4371, | |
| "step": 24060 | |
| }, | |
| { | |
| "epoch": 48.18, | |
| "grad_norm": 40.99359893798828, | |
| "learning_rate": 9.273963646474527e-07, | |
| "loss": 0.456, | |
| "step": 24090 | |
| }, | |
| { | |
| "epoch": 48.24, | |
| "grad_norm": 57.59867477416992, | |
| "learning_rate": 9.183037464140804e-07, | |
| "loss": 0.4523, | |
| "step": 24120 | |
| }, | |
| { | |
| "epoch": 48.3, | |
| "grad_norm": 142.89874267578125, | |
| "learning_rate": 9.09251412874882e-07, | |
| "loss": 0.3649, | |
| "step": 24150 | |
| }, | |
| { | |
| "epoch": 48.36, | |
| "grad_norm": 67.51494598388672, | |
| "learning_rate": 9.002394533727382e-07, | |
| "loss": 0.4477, | |
| "step": 24180 | |
| }, | |
| { | |
| "epoch": 48.42, | |
| "grad_norm": 82.09522247314453, | |
| "learning_rate": 8.912679568520494e-07, | |
| "loss": 0.4224, | |
| "step": 24210 | |
| }, | |
| { | |
| "epoch": 48.48, | |
| "grad_norm": 54.342594146728516, | |
| "learning_rate": 8.823370118578628e-07, | |
| "loss": 0.4375, | |
| "step": 24240 | |
| }, | |
| { | |
| "epoch": 48.54, | |
| "grad_norm": 137.6064910888672, | |
| "learning_rate": 8.734467065350022e-07, | |
| "loss": 0.4574, | |
| "step": 24270 | |
| }, | |
| { | |
| "epoch": 48.6, | |
| "grad_norm": 64.60173797607422, | |
| "learning_rate": 8.645971286271903e-07, | |
| "loss": 0.4518, | |
| "step": 24300 | |
| }, | |
| { | |
| "epoch": 48.66, | |
| "grad_norm": 45.69334411621094, | |
| "learning_rate": 8.557883654761906e-07, | |
| "loss": 0.404, | |
| "step": 24330 | |
| }, | |
| { | |
| "epoch": 48.72, | |
| "grad_norm": 179.41268920898438, | |
| "learning_rate": 8.470205040209362e-07, | |
| "loss": 0.4592, | |
| "step": 24360 | |
| }, | |
| { | |
| "epoch": 48.78, | |
| "grad_norm": 52.23598098754883, | |
| "learning_rate": 8.382936307966838e-07, | |
| "loss": 0.4067, | |
| "step": 24390 | |
| }, | |
| { | |
| "epoch": 48.84, | |
| "grad_norm": 73.33792114257812, | |
| "learning_rate": 8.296078319341444e-07, | |
| "loss": 0.4185, | |
| "step": 24420 | |
| }, | |
| { | |
| "epoch": 48.9, | |
| "grad_norm": 78.52240753173828, | |
| "learning_rate": 8.209631931586499e-07, | |
| "loss": 0.3914, | |
| "step": 24450 | |
| }, | |
| { | |
| "epoch": 48.96, | |
| "grad_norm": 59.30632781982422, | |
| "learning_rate": 8.123597997892918e-07, | |
| "loss": 0.4932, | |
| "step": 24480 | |
| }, | |
| { | |
| "epoch": 49.0, | |
| "eval_loss": 0.3021915555000305, | |
| "eval_map": 0.7844, | |
| "eval_map_50": 0.9617, | |
| "eval_map_75": 0.9101, | |
| "eval_map_chicken": 0.7882, | |
| "eval_map_duck": 0.7472, | |
| "eval_map_large": 0.8762, | |
| "eval_map_medium": 0.7765, | |
| "eval_map_plant": 0.8177, | |
| "eval_map_small": 0.2231, | |
| "eval_mar_1": 0.3007, | |
| "eval_mar_10": 0.8216, | |
| "eval_mar_100": 0.8252, | |
| "eval_mar_100_chicken": 0.8293, | |
| "eval_mar_100_duck": 0.7794, | |
| "eval_mar_100_plant": 0.8669, | |
| "eval_mar_large": 0.9176, | |
| "eval_mar_medium": 0.8308, | |
| "eval_mar_small": 0.3396, | |
| "eval_runtime": 5.0276, | |
| "eval_samples_per_second": 19.89, | |
| "eval_steps_per_second": 2.586, | |
| "step": 24500 | |
| }, | |
| { | |
| "epoch": 49.02, | |
| "grad_norm": 119.42558288574219, | |
| "learning_rate": 8.037977367380922e-07, | |
| "loss": 0.4811, | |
| "step": 24510 | |
| }, | |
| { | |
| "epoch": 49.08, | |
| "grad_norm": 0.00024352494801860303, | |
| "learning_rate": 7.952770885091548e-07, | |
| "loss": 0.3779, | |
| "step": 24540 | |
| }, | |
| { | |
| "epoch": 49.14, | |
| "grad_norm": 67.42436981201172, | |
| "learning_rate": 7.867979391978398e-07, | |
| "loss": 0.3558, | |
| "step": 24570 | |
| }, | |
| { | |
| "epoch": 49.2, | |
| "grad_norm": 92.31123352050781, | |
| "learning_rate": 7.783603724899258e-07, | |
| "loss": 0.4426, | |
| "step": 24600 | |
| }, | |
| { | |
| "epoch": 49.26, | |
| "grad_norm": 48.806907653808594, | |
| "learning_rate": 7.699644716607896e-07, | |
| "loss": 0.3877, | |
| "step": 24630 | |
| }, | |
| { | |
| "epoch": 49.32, | |
| "grad_norm": 27.726848602294922, | |
| "learning_rate": 7.61610319574585e-07, | |
| "loss": 0.3929, | |
| "step": 24660 | |
| }, | |
| { | |
| "epoch": 49.38, | |
| "grad_norm": 50.71929168701172, | |
| "learning_rate": 7.532979986834177e-07, | |
| "loss": 0.3916, | |
| "step": 24690 | |
| }, | |
| { | |
| "epoch": 49.44, | |
| "grad_norm": 38.81157684326172, | |
| "learning_rate": 7.450275910265415e-07, | |
| "loss": 0.3922, | |
| "step": 24720 | |
| }, | |
| { | |
| "epoch": 49.5, | |
| "grad_norm": 52.464599609375, | |
| "learning_rate": 7.367991782295392e-07, | |
| "loss": 0.4487, | |
| "step": 24750 | |
| }, | |
| { | |
| "epoch": 49.56, | |
| "grad_norm": 48.69332504272461, | |
| "learning_rate": 7.286128415035249e-07, | |
| "loss": 0.4287, | |
| "step": 24780 | |
| }, | |
| { | |
| "epoch": 49.62, | |
| "grad_norm": 50.46269226074219, | |
| "learning_rate": 7.204686616443352e-07, | |
| "loss": 0.3933, | |
| "step": 24810 | |
| }, | |
| { | |
| "epoch": 49.68, | |
| "grad_norm": 66.0242691040039, | |
| "learning_rate": 7.123667190317396e-07, | |
| "loss": 0.3534, | |
| "step": 24840 | |
| }, | |
| { | |
| "epoch": 49.74, | |
| "grad_norm": 77.78071594238281, | |
| "learning_rate": 7.043070936286395e-07, | |
| "loss": 0.3834, | |
| "step": 24870 | |
| }, | |
| { | |
| "epoch": 49.8, | |
| "grad_norm": 49.89218521118164, | |
| "learning_rate": 6.962898649802824e-07, | |
| "loss": 0.4152, | |
| "step": 24900 | |
| }, | |
| { | |
| "epoch": 49.86, | |
| "grad_norm": 47.149166107177734, | |
| "learning_rate": 6.883151122134812e-07, | |
| "loss": 0.4798, | |
| "step": 24930 | |
| }, | |
| { | |
| "epoch": 49.92, | |
| "grad_norm": 42.12199020385742, | |
| "learning_rate": 6.803829140358237e-07, | |
| "loss": 0.5189, | |
| "step": 24960 | |
| }, | |
| { | |
| "epoch": 49.98, | |
| "grad_norm": 53.24802780151367, | |
| "learning_rate": 6.724933487349061e-07, | |
| "loss": 0.3758, | |
| "step": 24990 | |
| }, | |
| { | |
| "epoch": 50.0, | |
| "eval_loss": 0.29586249589920044, | |
| "eval_map": 0.7921, | |
| "eval_map_50": 0.9609, | |
| "eval_map_75": 0.9203, | |
| "eval_map_chicken": 0.7932, | |
| "eval_map_duck": 0.7619, | |
| "eval_map_large": 0.8779, | |
| "eval_map_medium": 0.7853, | |
| "eval_map_plant": 0.8212, | |
| "eval_map_small": 0.2432, | |
| "eval_mar_1": 0.3066, | |
| "eval_mar_10": 0.8273, | |
| "eval_mar_100": 0.8314, | |
| "eval_mar_100_chicken": 0.832, | |
| "eval_mar_100_duck": 0.7918, | |
| "eval_mar_100_plant": 0.8703, | |
| "eval_mar_large": 0.9197, | |
| "eval_mar_medium": 0.8365, | |
| "eval_mar_small": 0.3655, | |
| "eval_runtime": 5.0566, | |
| "eval_samples_per_second": 19.776, | |
| "eval_steps_per_second": 2.571, | |
| "step": 25000 | |
| }, | |
| { | |
| "epoch": 50.04, | |
| "grad_norm": 88.88943481445312, | |
| "learning_rate": 6.646464941775499e-07, | |
| "loss": 0.4487, | |
| "step": 25020 | |
| }, | |
| { | |
| "epoch": 50.1, | |
| "grad_norm": 51.90554428100586, | |
| "learning_rate": 6.568424278090446e-07, | |
| "loss": 0.3981, | |
| "step": 25050 | |
| }, | |
| { | |
| "epoch": 50.16, | |
| "grad_norm": 53.483455657958984, | |
| "learning_rate": 6.490812266523716e-07, | |
| "loss": 0.3775, | |
| "step": 25080 | |
| }, | |
| { | |
| "epoch": 50.22, | |
| "grad_norm": 49.997982025146484, | |
| "learning_rate": 6.413629673074562e-07, | |
| "loss": 0.4164, | |
| "step": 25110 | |
| }, | |
| { | |
| "epoch": 50.28, | |
| "grad_norm": 55.87346267700195, | |
| "learning_rate": 6.336877259504004e-07, | |
| "loss": 0.3653, | |
| "step": 25140 | |
| }, | |
| { | |
| "epoch": 50.34, | |
| "grad_norm": 85.9616470336914, | |
| "learning_rate": 6.260555783327366e-07, | |
| "loss": 0.4081, | |
| "step": 25170 | |
| }, | |
| { | |
| "epoch": 50.4, | |
| "grad_norm": 95.91594696044922, | |
| "learning_rate": 6.184665997806832e-07, | |
| "loss": 0.4766, | |
| "step": 25200 | |
| }, | |
| { | |
| "epoch": 50.46, | |
| "grad_norm": 52.188880920410156, | |
| "learning_rate": 6.109208651943921e-07, | |
| "loss": 0.4127, | |
| "step": 25230 | |
| }, | |
| { | |
| "epoch": 50.52, | |
| "grad_norm": 146.0522003173828, | |
| "learning_rate": 6.034184490472195e-07, | |
| "loss": 0.339, | |
| "step": 25260 | |
| }, | |
| { | |
| "epoch": 50.58, | |
| "grad_norm": 63.30977249145508, | |
| "learning_rate": 5.959594253849821e-07, | |
| "loss": 0.4051, | |
| "step": 25290 | |
| }, | |
| { | |
| "epoch": 50.64, | |
| "grad_norm": 45.95663070678711, | |
| "learning_rate": 5.885438678252342e-07, | |
| "loss": 0.4141, | |
| "step": 25320 | |
| }, | |
| { | |
| "epoch": 50.7, | |
| "grad_norm": 194.88026428222656, | |
| "learning_rate": 5.811718495565327e-07, | |
| "loss": 0.3736, | |
| "step": 25350 | |
| }, | |
| { | |
| "epoch": 50.76, | |
| "grad_norm": 47.18809509277344, | |
| "learning_rate": 5.738434433377244e-07, | |
| "loss": 0.4111, | |
| "step": 25380 | |
| }, | |
| { | |
| "epoch": 50.82, | |
| "grad_norm": 21.77028465270996, | |
| "learning_rate": 5.665587214972173e-07, | |
| "loss": 0.4393, | |
| "step": 25410 | |
| }, | |
| { | |
| "epoch": 50.88, | |
| "grad_norm": 107.16293334960938, | |
| "learning_rate": 5.593177559322776e-07, | |
| "loss": 0.3947, | |
| "step": 25440 | |
| }, | |
| { | |
| "epoch": 50.94, | |
| "grad_norm": 60.57902526855469, | |
| "learning_rate": 5.521206181083111e-07, | |
| "loss": 0.5372, | |
| "step": 25470 | |
| }, | |
| { | |
| "epoch": 51.0, | |
| "grad_norm": 33.81594467163086, | |
| "learning_rate": 5.449673790581611e-07, | |
| "loss": 0.4397, | |
| "step": 25500 | |
| }, | |
| { | |
| "epoch": 51.0, | |
| "eval_loss": 0.2871440052986145, | |
| "eval_map": 0.7983, | |
| "eval_map_50": 0.9609, | |
| "eval_map_75": 0.9128, | |
| "eval_map_chicken": 0.802, | |
| "eval_map_duck": 0.7648, | |
| "eval_map_large": 0.8832, | |
| "eval_map_medium": 0.7966, | |
| "eval_map_plant": 0.828, | |
| "eval_map_small": 0.2145, | |
| "eval_mar_1": 0.3086, | |
| "eval_mar_10": 0.833, | |
| "eval_mar_100": 0.8374, | |
| "eval_mar_100_chicken": 0.8369, | |
| "eval_mar_100_duck": 0.7969, | |
| "eval_mar_100_plant": 0.8784, | |
| "eval_mar_large": 0.9251, | |
| "eval_mar_medium": 0.8478, | |
| "eval_mar_small": 0.3409, | |
| "eval_runtime": 5.1551, | |
| "eval_samples_per_second": 19.398, | |
| "eval_steps_per_second": 2.522, | |
| "step": 25500 | |
| }, | |
| { | |
| "epoch": 51.06, | |
| "grad_norm": 61.52519607543945, | |
| "learning_rate": 5.378581093814112e-07, | |
| "loss": 0.3936, | |
| "step": 25530 | |
| }, | |
| { | |
| "epoch": 51.12, | |
| "grad_norm": 41.62113952636719, | |
| "learning_rate": 5.307928792436812e-07, | |
| "loss": 0.3472, | |
| "step": 25560 | |
| }, | |
| { | |
| "epoch": 51.18, | |
| "grad_norm": 58.78872299194336, | |
| "learning_rate": 5.237717583759421e-07, | |
| "loss": 0.4171, | |
| "step": 25590 | |
| }, | |
| { | |
| "epoch": 51.24, | |
| "grad_norm": 71.37984466552734, | |
| "learning_rate": 5.167948160738206e-07, | |
| "loss": 0.3807, | |
| "step": 25620 | |
| }, | |
| { | |
| "epoch": 51.3, | |
| "grad_norm": 30.988309860229492, | |
| "learning_rate": 5.098621211969224e-07, | |
| "loss": 0.4499, | |
| "step": 25650 | |
| }, | |
| { | |
| "epoch": 51.36, | |
| "grad_norm": 90.97999572753906, | |
| "learning_rate": 5.029737421681446e-07, | |
| "loss": 0.406, | |
| "step": 25680 | |
| }, | |
| { | |
| "epoch": 51.42, | |
| "grad_norm": 39.365169525146484, | |
| "learning_rate": 4.961297469730097e-07, | |
| "loss": 0.4401, | |
| "step": 25710 | |
| }, | |
| { | |
| "epoch": 51.48, | |
| "grad_norm": 56.183563232421875, | |
| "learning_rate": 4.893302031589864e-07, | |
| "loss": 0.4249, | |
| "step": 25740 | |
| }, | |
| { | |
| "epoch": 51.54, | |
| "grad_norm": 35.67009353637695, | |
| "learning_rate": 4.825751778348259e-07, | |
| "loss": 0.4689, | |
| "step": 25770 | |
| }, | |
| { | |
| "epoch": 51.6, | |
| "grad_norm": 68.61273956298828, | |
| "learning_rate": 4.758647376699033e-07, | |
| "loss": 0.3953, | |
| "step": 25800 | |
| }, | |
| { | |
| "epoch": 51.66, | |
| "grad_norm": 39.2368049621582, | |
| "learning_rate": 4.691989488935511e-07, | |
| "loss": 0.4341, | |
| "step": 25830 | |
| }, | |
| { | |
| "epoch": 51.72, | |
| "grad_norm": 40.57124710083008, | |
| "learning_rate": 4.625778772944156e-07, | |
| "loss": 0.3937, | |
| "step": 25860 | |
| }, | |
| { | |
| "epoch": 51.78, | |
| "grad_norm": 63.675323486328125, | |
| "learning_rate": 4.5600158821979933e-07, | |
| "loss": 0.3677, | |
| "step": 25890 | |
| }, | |
| { | |
| "epoch": 51.84, | |
| "grad_norm": 52.42002487182617, | |
| "learning_rate": 4.494701465750217e-07, | |
| "loss": 0.336, | |
| "step": 25920 | |
| }, | |
| { | |
| "epoch": 51.9, | |
| "grad_norm": 30.058719635009766, | |
| "learning_rate": 4.4298361682277355e-07, | |
| "loss": 0.4698, | |
| "step": 25950 | |
| }, | |
| { | |
| "epoch": 51.96, | |
| "grad_norm": 51.076725006103516, | |
| "learning_rate": 4.3654206298248625e-07, | |
| "loss": 0.3917, | |
| "step": 25980 | |
| }, | |
| { | |
| "epoch": 52.0, | |
| "eval_loss": 0.29068684577941895, | |
| "eval_map": 0.7955, | |
| "eval_map_50": 0.9645, | |
| "eval_map_75": 0.9161, | |
| "eval_map_chicken": 0.7975, | |
| "eval_map_duck": 0.7654, | |
| "eval_map_large": 0.8796, | |
| "eval_map_medium": 0.7911, | |
| "eval_map_plant": 0.8234, | |
| "eval_map_small": 0.2316, | |
| "eval_mar_1": 0.308, | |
| "eval_mar_10": 0.8314, | |
| "eval_mar_100": 0.8352, | |
| "eval_mar_100_chicken": 0.8356, | |
| "eval_mar_100_duck": 0.7969, | |
| "eval_mar_100_plant": 0.8732, | |
| "eval_mar_large": 0.9192, | |
| "eval_mar_medium": 0.8428, | |
| "eval_mar_small": 0.375, | |
| "eval_runtime": 5.0448, | |
| "eval_samples_per_second": 19.822, | |
| "eval_steps_per_second": 2.577, | |
| "step": 26000 | |
| }, | |
| { | |
| "epoch": 52.02, | |
| "grad_norm": 34.134273529052734, | |
| "learning_rate": 4.301455486296946e-07, | |
| "loss": 0.422, | |
| "step": 26010 | |
| }, | |
| { | |
| "epoch": 52.08, | |
| "grad_norm": 62.580841064453125, | |
| "learning_rate": 4.237941368954124e-07, | |
| "loss": 0.374, | |
| "step": 26040 | |
| }, | |
| { | |
| "epoch": 52.14, | |
| "grad_norm": 49.3266487121582, | |
| "learning_rate": 4.1748789046551055e-07, | |
| "loss": 0.384, | |
| "step": 26070 | |
| }, | |
| { | |
| "epoch": 52.2, | |
| "grad_norm": 0.00041551125468686223, | |
| "learning_rate": 4.112268715800943e-07, | |
| "loss": 0.4655, | |
| "step": 26100 | |
| }, | |
| { | |
| "epoch": 52.26, | |
| "grad_norm": 38.26332473754883, | |
| "learning_rate": 4.0501114203289395e-07, | |
| "loss": 0.3986, | |
| "step": 26130 | |
| }, | |
| { | |
| "epoch": 52.32, | |
| "grad_norm": 65.15966796875, | |
| "learning_rate": 3.9884076317064813e-07, | |
| "loss": 0.4129, | |
| "step": 26160 | |
| }, | |
| { | |
| "epoch": 52.38, | |
| "grad_norm": 74.6011962890625, | |
| "learning_rate": 3.9271579589250817e-07, | |
| "loss": 0.477, | |
| "step": 26190 | |
| }, | |
| { | |
| "epoch": 52.44, | |
| "grad_norm": 48.72752380371094, | |
| "learning_rate": 3.866363006494256e-07, | |
| "loss": 0.4237, | |
| "step": 26220 | |
| }, | |
| { | |
| "epoch": 52.5, | |
| "grad_norm": 39.035118103027344, | |
| "learning_rate": 3.8060233744356634e-07, | |
| "loss": 0.4287, | |
| "step": 26250 | |
| }, | |
| { | |
| "epoch": 52.56, | |
| "grad_norm": 54.2945556640625, | |
| "learning_rate": 3.7461396582771035e-07, | |
| "loss": 0.4003, | |
| "step": 26280 | |
| }, | |
| { | |
| "epoch": 52.62, | |
| "grad_norm": 59.89337921142578, | |
| "learning_rate": 3.6867124490466697e-07, | |
| "loss": 0.4809, | |
| "step": 26310 | |
| }, | |
| { | |
| "epoch": 52.68, | |
| "grad_norm": 32.89358139038086, | |
| "learning_rate": 3.627742333266937e-07, | |
| "loss": 0.3781, | |
| "step": 26340 | |
| }, | |
| { | |
| "epoch": 52.74, | |
| "grad_norm": 114.98311614990234, | |
| "learning_rate": 3.569229892949133e-07, | |
| "loss": 0.4946, | |
| "step": 26370 | |
| }, | |
| { | |
| "epoch": 52.8, | |
| "grad_norm": 62.90887451171875, | |
| "learning_rate": 3.511175705587433e-07, | |
| "loss": 0.4092, | |
| "step": 26400 | |
| }, | |
| { | |
| "epoch": 52.86, | |
| "grad_norm": 408.21044921875, | |
| "learning_rate": 3.4535803441532125e-07, | |
| "loss": 0.4111, | |
| "step": 26430 | |
| }, | |
| { | |
| "epoch": 52.92, | |
| "grad_norm": 35.12818145751953, | |
| "learning_rate": 3.396444377089453e-07, | |
| "loss": 0.4019, | |
| "step": 26460 | |
| }, | |
| { | |
| "epoch": 52.98, | |
| "grad_norm": 0.0001875197485787794, | |
| "learning_rate": 3.3397683683050685e-07, | |
| "loss": 0.3362, | |
| "step": 26490 | |
| }, | |
| { | |
| "epoch": 53.0, | |
| "eval_loss": 0.2884664237499237, | |
| "eval_map": 0.7989, | |
| "eval_map_50": 0.9644, | |
| "eval_map_75": 0.92, | |
| "eval_map_chicken": 0.8012, | |
| "eval_map_duck": 0.7703, | |
| "eval_map_large": 0.8789, | |
| "eval_map_medium": 0.7958, | |
| "eval_map_plant": 0.8253, | |
| "eval_map_small": 0.2324, | |
| "eval_mar_1": 0.3075, | |
| "eval_mar_10": 0.8338, | |
| "eval_mar_100": 0.8379, | |
| "eval_mar_100_chicken": 0.8382, | |
| "eval_mar_100_duck": 0.8, | |
| "eval_mar_100_plant": 0.8755, | |
| "eval_mar_large": 0.9201, | |
| "eval_mar_medium": 0.8465, | |
| "eval_mar_small": 0.3769, | |
| "eval_runtime": 5.0147, | |
| "eval_samples_per_second": 19.941, | |
| "eval_steps_per_second": 2.592, | |
| "step": 26500 | |
| }, | |
| { | |
| "epoch": 53.04, | |
| "grad_norm": 47.404518127441406, | |
| "learning_rate": 3.283552877169399e-07, | |
| "loss": 0.4292, | |
| "step": 26520 | |
| }, | |
| { | |
| "epoch": 53.1, | |
| "grad_norm": 52.7077751159668, | |
| "learning_rate": 3.227798458506637e-07, | |
| "loss": 0.3831, | |
| "step": 26550 | |
| }, | |
| { | |
| "epoch": 53.16, | |
| "grad_norm": 0.0004456727474462241, | |
| "learning_rate": 3.172505662590386e-07, | |
| "loss": 0.3576, | |
| "step": 26580 | |
| }, | |
| { | |
| "epoch": 53.22, | |
| "grad_norm": 55.14971923828125, | |
| "learning_rate": 3.1176750351382235e-07, | |
| "loss": 0.3817, | |
| "step": 26610 | |
| }, | |
| { | |
| "epoch": 53.28, | |
| "grad_norm": 221.3983917236328, | |
| "learning_rate": 3.0633071173062966e-07, | |
| "loss": 0.4345, | |
| "step": 26640 | |
| }, | |
| { | |
| "epoch": 53.34, | |
| "grad_norm": 42.98983383178711, | |
| "learning_rate": 3.0094024456840176e-07, | |
| "loss": 0.4081, | |
| "step": 26670 | |
| }, | |
| { | |
| "epoch": 53.4, | |
| "grad_norm": 53.3066291809082, | |
| "learning_rate": 2.9559615522887275e-07, | |
| "loss": 0.402, | |
| "step": 26700 | |
| }, | |
| { | |
| "epoch": 53.46, | |
| "grad_norm": 70.44734191894531, | |
| "learning_rate": 2.9029849645604735e-07, | |
| "loss": 0.4517, | |
| "step": 26730 | |
| }, | |
| { | |
| "epoch": 53.52, | |
| "grad_norm": 49.392723083496094, | |
| "learning_rate": 2.850473205356774e-07, | |
| "loss": 0.3977, | |
| "step": 26760 | |
| }, | |
| { | |
| "epoch": 53.58, | |
| "grad_norm": 77.64625549316406, | |
| "learning_rate": 2.798426792947517e-07, | |
| "loss": 0.4356, | |
| "step": 26790 | |
| }, | |
| { | |
| "epoch": 53.64, | |
| "grad_norm": 59.67546081542969, | |
| "learning_rate": 2.746846241009765e-07, | |
| "loss": 0.5099, | |
| "step": 26820 | |
| }, | |
| { | |
| "epoch": 53.7, | |
| "grad_norm": 136.89759826660156, | |
| "learning_rate": 2.6957320586227354e-07, | |
| "loss": 0.3658, | |
| "step": 26850 | |
| }, | |
| { | |
| "epoch": 53.76, | |
| "grad_norm": 64.72844696044922, | |
| "learning_rate": 2.6450847502627883e-07, | |
| "loss": 0.5296, | |
| "step": 26880 | |
| }, | |
| { | |
| "epoch": 53.82, | |
| "grad_norm": 134.72019958496094, | |
| "learning_rate": 2.594904815798399e-07, | |
| "loss": 0.4809, | |
| "step": 26910 | |
| }, | |
| { | |
| "epoch": 53.88, | |
| "grad_norm": 373.54449462890625, | |
| "learning_rate": 2.5451927504852757e-07, | |
| "loss": 0.4206, | |
| "step": 26940 | |
| }, | |
| { | |
| "epoch": 53.94, | |
| "grad_norm": 53.72751998901367, | |
| "learning_rate": 2.49594904496141e-07, | |
| "loss": 0.368, | |
| "step": 26970 | |
| }, | |
| { | |
| "epoch": 54.0, | |
| "grad_norm": 68.3394546508789, | |
| "learning_rate": 2.447174185242324e-07, | |
| "loss": 0.4004, | |
| "step": 27000 | |
| }, | |
| { | |
| "epoch": 54.0, | |
| "eval_loss": 0.28688502311706543, | |
| "eval_map": 0.7973, | |
| "eval_map_50": 0.9644, | |
| "eval_map_75": 0.9201, | |
| "eval_map_chicken": 0.801, | |
| "eval_map_duck": 0.7636, | |
| "eval_map_large": 0.8813, | |
| "eval_map_medium": 0.7957, | |
| "eval_map_plant": 0.8273, | |
| "eval_map_small": 0.228, | |
| "eval_mar_1": 0.3069, | |
| "eval_mar_10": 0.8328, | |
| "eval_mar_100": 0.8368, | |
| "eval_mar_100_chicken": 0.8373, | |
| "eval_mar_100_duck": 0.7948, | |
| "eval_mar_100_plant": 0.8781, | |
| "eval_mar_large": 0.9218, | |
| "eval_mar_medium": 0.8456, | |
| "eval_mar_small": 0.3822, | |
| "eval_runtime": 5.2392, | |
| "eval_samples_per_second": 19.087, | |
| "eval_steps_per_second": 2.481, | |
| "step": 27000 | |
| }, | |
| { | |
| "epoch": 54.06, | |
| "grad_norm": 70.93304443359375, | |
| "learning_rate": 2.3988686527161686e-07, | |
| "loss": 0.3863, | |
| "step": 27030 | |
| }, | |
| { | |
| "epoch": 54.12, | |
| "grad_norm": 427.5562438964844, | |
| "learning_rate": 2.351032924139063e-07, | |
| "loss": 0.4965, | |
| "step": 27060 | |
| }, | |
| { | |
| "epoch": 54.18, | |
| "grad_norm": 0.00017418510105926543, | |
| "learning_rate": 2.3036674716303277e-07, | |
| "loss": 0.4601, | |
| "step": 27090 | |
| }, | |
| { | |
| "epoch": 54.24, | |
| "grad_norm": 59.90283203125, | |
| "learning_rate": 2.2567727626678527e-07, | |
| "loss": 0.3882, | |
| "step": 27120 | |
| }, | |
| { | |
| "epoch": 54.3, | |
| "grad_norm": 60.470008850097656, | |
| "learning_rate": 2.210349260083494e-07, | |
| "loss": 0.4183, | |
| "step": 27150 | |
| }, | |
| { | |
| "epoch": 54.36, | |
| "grad_norm": 54.06813430786133, | |
| "learning_rate": 2.1643974220584729e-07, | |
| "loss": 0.4768, | |
| "step": 27180 | |
| }, | |
| { | |
| "epoch": 54.42, | |
| "grad_norm": 37.11991882324219, | |
| "learning_rate": 2.1189177021188888e-07, | |
| "loss": 0.402, | |
| "step": 27210 | |
| }, | |
| { | |
| "epoch": 54.48, | |
| "grad_norm": 89.20401763916016, | |
| "learning_rate": 2.0739105491312028e-07, | |
| "loss": 0.4153, | |
| "step": 27240 | |
| }, | |
| { | |
| "epoch": 54.54, | |
| "grad_norm": 56.34545135498047, | |
| "learning_rate": 2.0293764072978618e-07, | |
| "loss": 0.408, | |
| "step": 27270 | |
| }, | |
| { | |
| "epoch": 54.6, | |
| "grad_norm": 97.46742248535156, | |
| "learning_rate": 1.9853157161528468e-07, | |
| "loss": 0.4485, | |
| "step": 27300 | |
| }, | |
| { | |
| "epoch": 54.66, | |
| "grad_norm": 48.18669509887695, | |
| "learning_rate": 1.9417289105574054e-07, | |
| "loss": 0.4489, | |
| "step": 27330 | |
| }, | |
| { | |
| "epoch": 54.72, | |
| "grad_norm": 62.39783477783203, | |
| "learning_rate": 1.8986164206957037e-07, | |
| "loss": 0.3776, | |
| "step": 27360 | |
| }, | |
| { | |
| "epoch": 54.78, | |
| "grad_norm": 0.000213187318877317, | |
| "learning_rate": 1.8559786720706185e-07, | |
| "loss": 0.3596, | |
| "step": 27390 | |
| }, | |
| { | |
| "epoch": 54.84, | |
| "grad_norm": 97.40396118164062, | |
| "learning_rate": 1.8138160854995145e-07, | |
| "loss": 0.4219, | |
| "step": 27420 | |
| }, | |
| { | |
| "epoch": 54.9, | |
| "grad_norm": 42.91602325439453, | |
| "learning_rate": 1.7721290771100964e-07, | |
| "loss": 0.4467, | |
| "step": 27450 | |
| }, | |
| { | |
| "epoch": 54.96, | |
| "grad_norm": 46.58220291137695, | |
| "learning_rate": 1.7309180583363062e-07, | |
| "loss": 0.406, | |
| "step": 27480 | |
| }, | |
| { | |
| "epoch": 55.0, | |
| "eval_loss": 0.28708022832870483, | |
| "eval_map": 0.8004, | |
| "eval_map_50": 0.9645, | |
| "eval_map_75": 0.9194, | |
| "eval_map_chicken": 0.8069, | |
| "eval_map_duck": 0.7679, | |
| "eval_map_large": 0.8788, | |
| "eval_map_medium": 0.7986, | |
| "eval_map_plant": 0.8265, | |
| "eval_map_small": 0.2283, | |
| "eval_mar_1": 0.3084, | |
| "eval_mar_10": 0.8343, | |
| "eval_mar_100": 0.8384, | |
| "eval_mar_100_chicken": 0.8404, | |
| "eval_mar_100_duck": 0.7979, | |
| "eval_mar_100_plant": 0.8769, | |
| "eval_mar_large": 0.9205, | |
| "eval_mar_medium": 0.8476, | |
| "eval_mar_small": 0.3822, | |
| "eval_runtime": 5.0648, | |
| "eval_samples_per_second": 19.744, | |
| "eval_steps_per_second": 2.567, | |
| "step": 27500 | |
| }, | |
| { | |
| "epoch": 55.02, | |
| "grad_norm": 44.80258560180664, | |
| "learning_rate": 1.690183435914261e-07, | |
| "loss": 0.4592, | |
| "step": 27510 | |
| }, | |
| { | |
| "epoch": 55.08, | |
| "grad_norm": 46.293399810791016, | |
| "learning_rate": 1.6499256118782503e-07, | |
| "loss": 0.4603, | |
| "step": 27540 | |
| }, | |
| { | |
| "epoch": 55.14, | |
| "grad_norm": 0.00020682995091192424, | |
| "learning_rate": 1.6101449835567273e-07, | |
| "loss": 0.3691, | |
| "step": 27570 | |
| }, | |
| { | |
| "epoch": 55.2, | |
| "grad_norm": 0.00038863185909576714, | |
| "learning_rate": 1.5708419435684463e-07, | |
| "loss": 0.3658, | |
| "step": 27600 | |
| }, | |
| { | |
| "epoch": 55.26, | |
| "grad_norm": 64.51366424560547, | |
| "learning_rate": 1.532016879818532e-07, | |
| "loss": 0.4051, | |
| "step": 27630 | |
| }, | |
| { | |
| "epoch": 55.32, | |
| "grad_norm": 137.62094116210938, | |
| "learning_rate": 1.4936701754947104e-07, | |
| "loss": 0.3851, | |
| "step": 27660 | |
| }, | |
| { | |
| "epoch": 55.38, | |
| "grad_norm": 32.709232330322266, | |
| "learning_rate": 1.4558022090634504e-07, | |
| "loss": 0.4879, | |
| "step": 27690 | |
| }, | |
| { | |
| "epoch": 55.44, | |
| "grad_norm": 40.718658447265625, | |
| "learning_rate": 1.4184133542663014e-07, | |
| "loss": 0.4211, | |
| "step": 27720 | |
| }, | |
| { | |
| "epoch": 55.5, | |
| "grad_norm": 65.1848373413086, | |
| "learning_rate": 1.3815039801161723e-07, | |
| "loss": 0.4523, | |
| "step": 27750 | |
| }, | |
| { | |
| "epoch": 55.56, | |
| "grad_norm": 37.71095657348633, | |
| "learning_rate": 1.3450744508936687e-07, | |
| "loss": 0.3766, | |
| "step": 27780 | |
| }, | |
| { | |
| "epoch": 55.62, | |
| "grad_norm": 58.59610366821289, | |
| "learning_rate": 1.3091251261435568e-07, | |
| "loss": 0.4531, | |
| "step": 27810 | |
| }, | |
| { | |
| "epoch": 55.68, | |
| "grad_norm": 38.84789276123047, | |
| "learning_rate": 1.2736563606711384e-07, | |
| "loss": 0.3873, | |
| "step": 27840 | |
| }, | |
| { | |
| "epoch": 55.74, | |
| "grad_norm": 0.0002489400503691286, | |
| "learning_rate": 1.2386685045388313e-07, | |
| "loss": 0.3393, | |
| "step": 27870 | |
| }, | |
| { | |
| "epoch": 55.8, | |
| "grad_norm": 122.7408218383789, | |
| "learning_rate": 1.2041619030626283e-07, | |
| "loss": 0.4669, | |
| "step": 27900 | |
| }, | |
| { | |
| "epoch": 55.86, | |
| "grad_norm": 0.0002524556184653193, | |
| "learning_rate": 1.1701368968087711e-07, | |
| "loss": 0.5525, | |
| "step": 27930 | |
| }, | |
| { | |
| "epoch": 55.92, | |
| "grad_norm": 32.06369400024414, | |
| "learning_rate": 1.136593821590326e-07, | |
| "loss": 0.3817, | |
| "step": 27960 | |
| }, | |
| { | |
| "epoch": 55.98, | |
| "grad_norm": 164.66461181640625, | |
| "learning_rate": 1.1035330084639084e-07, | |
| "loss": 0.3876, | |
| "step": 27990 | |
| }, | |
| { | |
| "epoch": 56.0, | |
| "eval_loss": 0.2882368564605713, | |
| "eval_map": 0.7985, | |
| "eval_map_50": 0.9641, | |
| "eval_map_75": 0.9197, | |
| "eval_map_chicken": 0.8072, | |
| "eval_map_duck": 0.7646, | |
| "eval_map_large": 0.8772, | |
| "eval_map_medium": 0.7974, | |
| "eval_map_plant": 0.8237, | |
| "eval_map_small": 0.2257, | |
| "eval_mar_1": 0.3084, | |
| "eval_mar_10": 0.834, | |
| "eval_mar_100": 0.838, | |
| "eval_mar_100_chicken": 0.8436, | |
| "eval_mar_100_duck": 0.7969, | |
| "eval_mar_100_plant": 0.8735, | |
| "eval_mar_large": 0.918, | |
| "eval_mar_medium": 0.8474, | |
| "eval_mar_small": 0.3676, | |
| "eval_runtime": 5.0534, | |
| "eval_samples_per_second": 19.788, | |
| "eval_steps_per_second": 2.573, | |
| "step": 28000 | |
| }, | |
| { | |
| "epoch": 56.04, | |
| "grad_norm": 101.48612213134766, | |
| "learning_rate": 1.0709547837263967e-07, | |
| "loss": 0.4297, | |
| "step": 28020 | |
| }, | |
| { | |
| "epoch": 56.1, | |
| "grad_norm": 32.960208892822266, | |
| "learning_rate": 1.038859468911707e-07, | |
| "loss": 0.3881, | |
| "step": 28050 | |
| }, | |
| { | |
| "epoch": 56.16, | |
| "grad_norm": 45.9619026184082, | |
| "learning_rate": 1.007247380787657e-07, | |
| "loss": 0.3947, | |
| "step": 28080 | |
| }, | |
| { | |
| "epoch": 56.22, | |
| "grad_norm": 50.23225784301758, | |
| "learning_rate": 9.761188313527792e-08, | |
| "loss": 0.4163, | |
| "step": 28110 | |
| }, | |
| { | |
| "epoch": 56.28, | |
| "grad_norm": 69.35472106933594, | |
| "learning_rate": 9.454741278333013e-08, | |
| "loss": 0.4509, | |
| "step": 28140 | |
| }, | |
| { | |
| "epoch": 56.34, | |
| "grad_norm": 49.21049118041992, | |
| "learning_rate": 9.153135726800599e-08, | |
| "loss": 0.4435, | |
| "step": 28170 | |
| }, | |
| { | |
| "epoch": 56.4, | |
| "grad_norm": 24.567251205444336, | |
| "learning_rate": 8.856374635655696e-08, | |
| "loss": 0.435, | |
| "step": 28200 | |
| }, | |
| { | |
| "epoch": 56.46, | |
| "grad_norm": 41.68861770629883, | |
| "learning_rate": 8.564460933810414e-08, | |
| "loss": 0.4124, | |
| "step": 28230 | |
| }, | |
| { | |
| "epoch": 56.52, | |
| "grad_norm": 62.273746490478516, | |
| "learning_rate": 8.277397502335194e-08, | |
| "loss": 0.4942, | |
| "step": 28260 | |
| }, | |
| { | |
| "epoch": 56.58, | |
| "grad_norm": 38.9687385559082, | |
| "learning_rate": 7.995187174430152e-08, | |
| "loss": 0.3975, | |
| "step": 28290 | |
| }, | |
| { | |
| "epoch": 56.64, | |
| "grad_norm": 60.86214828491211, | |
| "learning_rate": 7.717832735397335e-08, | |
| "loss": 0.3889, | |
| "step": 28320 | |
| }, | |
| { | |
| "epoch": 56.7, | |
| "grad_norm": 106.79747009277344, | |
| "learning_rate": 7.445336922613067e-08, | |
| "loss": 0.3872, | |
| "step": 28350 | |
| }, | |
| { | |
| "epoch": 56.76, | |
| "grad_norm": 185.20068359375, | |
| "learning_rate": 7.177702425500977e-08, | |
| "loss": 0.3844, | |
| "step": 28380 | |
| }, | |
| { | |
| "epoch": 56.82, | |
| "grad_norm": 44.18353271484375, | |
| "learning_rate": 6.914931885505626e-08, | |
| "loss": 0.4314, | |
| "step": 28410 | |
| }, | |
| { | |
| "epoch": 56.88, | |
| "grad_norm": 41.35288619995117, | |
| "learning_rate": 6.657027896065982e-08, | |
| "loss": 0.4599, | |
| "step": 28440 | |
| }, | |
| { | |
| "epoch": 56.94, | |
| "grad_norm": 294.2066345214844, | |
| "learning_rate": 6.403993002590425e-08, | |
| "loss": 0.3814, | |
| "step": 28470 | |
| }, | |
| { | |
| "epoch": 57.0, | |
| "grad_norm": 38.01483154296875, | |
| "learning_rate": 6.15582970243117e-08, | |
| "loss": 0.3939, | |
| "step": 28500 | |
| }, | |
| { | |
| "epoch": 57.0, | |
| "eval_loss": 0.28454330563545227, | |
| "eval_map": 0.8024, | |
| "eval_map_50": 0.9645, | |
| "eval_map_75": 0.9195, | |
| "eval_map_chicken": 0.8102, | |
| "eval_map_duck": 0.7709, | |
| "eval_map_large": 0.8782, | |
| "eval_map_medium": 0.8014, | |
| "eval_map_plant": 0.8262, | |
| "eval_map_small": 0.2291, | |
| "eval_mar_1": 0.3093, | |
| "eval_mar_10": 0.8367, | |
| "eval_mar_100": 0.8405, | |
| "eval_mar_100_chicken": 0.8431, | |
| "eval_mar_100_duck": 0.8021, | |
| "eval_mar_100_plant": 0.8764, | |
| "eval_mar_large": 0.9197, | |
| "eval_mar_medium": 0.8511, | |
| "eval_mar_small": 0.3697, | |
| "eval_runtime": 5.2741, | |
| "eval_samples_per_second": 18.961, | |
| "eval_steps_per_second": 2.465, | |
| "step": 28500 | |
| }, | |
| { | |
| "epoch": 57.06, | |
| "grad_norm": 40.22798156738281, | |
| "learning_rate": 5.9125404448597825e-08, | |
| "loss": 0.4592, | |
| "step": 28530 | |
| }, | |
| { | |
| "epoch": 57.12, | |
| "grad_norm": 227.82638549804688, | |
| "learning_rate": 5.674127631043025e-08, | |
| "loss": 0.3248, | |
| "step": 28560 | |
| }, | |
| { | |
| "epoch": 57.18, | |
| "grad_norm": 31.734601974487305, | |
| "learning_rate": 5.440593614019107e-08, | |
| "loss": 0.3187, | |
| "step": 28590 | |
| }, | |
| { | |
| "epoch": 57.24, | |
| "grad_norm": 49.92489242553711, | |
| "learning_rate": 5.2119406986745336e-08, | |
| "loss": 0.3835, | |
| "step": 28620 | |
| }, | |
| { | |
| "epoch": 57.3, | |
| "grad_norm": 92.18814849853516, | |
| "learning_rate": 4.988171141721232e-08, | |
| "loss": 0.4056, | |
| "step": 28650 | |
| }, | |
| { | |
| "epoch": 57.36, | |
| "grad_norm": 53.62547302246094, | |
| "learning_rate": 4.769287151674407e-08, | |
| "loss": 0.4088, | |
| "step": 28680 | |
| }, | |
| { | |
| "epoch": 57.42, | |
| "grad_norm": 53.25278091430664, | |
| "learning_rate": 4.5552908888306654e-08, | |
| "loss": 0.4545, | |
| "step": 28710 | |
| }, | |
| { | |
| "epoch": 57.48, | |
| "grad_norm": 47.75865173339844, | |
| "learning_rate": 4.346184465246761e-08, | |
| "loss": 0.4104, | |
| "step": 28740 | |
| }, | |
| { | |
| "epoch": 57.54, | |
| "grad_norm": 69.6591796875, | |
| "learning_rate": 4.1419699447186045e-08, | |
| "loss": 0.4026, | |
| "step": 28770 | |
| }, | |
| { | |
| "epoch": 57.6, | |
| "grad_norm": 50.19314956665039, | |
| "learning_rate": 3.9426493427611177e-08, | |
| "loss": 0.3975, | |
| "step": 28800 | |
| }, | |
| { | |
| "epoch": 57.66, | |
| "grad_norm": 54.0328369140625, | |
| "learning_rate": 3.748224626588137e-08, | |
| "loss": 0.6515, | |
| "step": 28830 | |
| }, | |
| { | |
| "epoch": 57.72, | |
| "grad_norm": 39.40961837768555, | |
| "learning_rate": 3.558697715093207e-08, | |
| "loss": 0.406, | |
| "step": 28860 | |
| }, | |
| { | |
| "epoch": 57.78, | |
| "grad_norm": 43.196834564208984, | |
| "learning_rate": 3.374070478830316e-08, | |
| "loss": 0.4559, | |
| "step": 28890 | |
| }, | |
| { | |
| "epoch": 57.84, | |
| "grad_norm": 23.547550201416016, | |
| "learning_rate": 3.194344739995803e-08, | |
| "loss": 0.605, | |
| "step": 28920 | |
| }, | |
| { | |
| "epoch": 57.9, | |
| "grad_norm": 69.10009765625, | |
| "learning_rate": 3.019522272410202e-08, | |
| "loss": 0.4126, | |
| "step": 28950 | |
| }, | |
| { | |
| "epoch": 57.96, | |
| "grad_norm": 66.09160614013672, | |
| "learning_rate": 2.8496048015005385e-08, | |
| "loss": 0.4218, | |
| "step": 28980 | |
| }, | |
| { | |
| "epoch": 58.0, | |
| "eval_loss": 0.2852034270763397, | |
| "eval_map": 0.8, | |
| "eval_map_50": 0.9646, | |
| "eval_map_75": 0.9196, | |
| "eval_map_chicken": 0.8067, | |
| "eval_map_duck": 0.7689, | |
| "eval_map_large": 0.8774, | |
| "eval_map_medium": 0.7993, | |
| "eval_map_plant": 0.8245, | |
| "eval_map_small": 0.2254, | |
| "eval_mar_1": 0.3085, | |
| "eval_mar_10": 0.8346, | |
| "eval_mar_100": 0.8384, | |
| "eval_mar_100_chicken": 0.8413, | |
| "eval_mar_100_duck": 0.799, | |
| "eval_mar_100_plant": 0.8749, | |
| "eval_mar_large": 0.9188, | |
| "eval_mar_medium": 0.8488, | |
| "eval_mar_small": 0.3634, | |
| "eval_runtime": 5.0721, | |
| "eval_samples_per_second": 19.716, | |
| "eval_steps_per_second": 2.563, | |
| "step": 29000 | |
| }, | |
| { | |
| "epoch": 58.02, | |
| "grad_norm": 69.85868072509766, | |
| "learning_rate": 2.684594004283836e-08, | |
| "loss": 0.4304, | |
| "step": 29010 | |
| }, | |
| { | |
| "epoch": 58.08, | |
| "grad_norm": 41.767555236816406, | |
| "learning_rate": 2.5244915093499134e-08, | |
| "loss": 0.3461, | |
| "step": 29040 | |
| }, | |
| { | |
| "epoch": 58.14, | |
| "grad_norm": 82.25499725341797, | |
| "learning_rate": 2.3692988968458398e-08, | |
| "loss": 0.5096, | |
| "step": 29070 | |
| }, | |
| { | |
| "epoch": 58.2, | |
| "grad_norm": 54.37453079223633, | |
| "learning_rate": 2.219017698460002e-08, | |
| "loss": 0.4233, | |
| "step": 29100 | |
| }, | |
| { | |
| "epoch": 58.26, | |
| "grad_norm": 61.195838928222656, | |
| "learning_rate": 2.0736493974071736e-08, | |
| "loss": 0.3487, | |
| "step": 29130 | |
| }, | |
| { | |
| "epoch": 58.32, | |
| "grad_norm": 28.579872131347656, | |
| "learning_rate": 1.9331954284137476e-08, | |
| "loss": 0.3522, | |
| "step": 29160 | |
| }, | |
| { | |
| "epoch": 58.38, | |
| "grad_norm": 56.89299774169922, | |
| "learning_rate": 1.7976571777038044e-08, | |
| "loss": 0.3813, | |
| "step": 29190 | |
| }, | |
| { | |
| "epoch": 58.44, | |
| "grad_norm": 56.8980598449707, | |
| "learning_rate": 1.6670359829850657e-08, | |
| "loss": 0.4162, | |
| "step": 29220 | |
| }, | |
| { | |
| "epoch": 58.5, | |
| "grad_norm": 39.616416931152344, | |
| "learning_rate": 1.541333133436018e-08, | |
| "loss": 0.4343, | |
| "step": 29250 | |
| }, | |
| { | |
| "epoch": 58.56, | |
| "grad_norm": 47.3499641418457, | |
| "learning_rate": 1.4205498696930332e-08, | |
| "loss": 0.3844, | |
| "step": 29280 | |
| }, | |
| { | |
| "epoch": 58.62, | |
| "grad_norm": 49.85858917236328, | |
| "learning_rate": 1.3046873838381546e-08, | |
| "loss": 0.4488, | |
| "step": 29310 | |
| }, | |
| { | |
| "epoch": 58.68, | |
| "grad_norm": 63.439273834228516, | |
| "learning_rate": 1.1937468193873869e-08, | |
| "loss": 0.4068, | |
| "step": 29340 | |
| }, | |
| { | |
| "epoch": 58.74, | |
| "grad_norm": 22.61992835998535, | |
| "learning_rate": 1.0877292712792586e-08, | |
| "loss": 0.3993, | |
| "step": 29370 | |
| }, | |
| { | |
| "epoch": 58.8, | |
| "grad_norm": 182.1600799560547, | |
| "learning_rate": 9.866357858642206e-09, | |
| "loss": 0.3899, | |
| "step": 29400 | |
| }, | |
| { | |
| "epoch": 58.86, | |
| "grad_norm": 54.466739654541016, | |
| "learning_rate": 8.904673608940983e-09, | |
| "loss": 0.394, | |
| "step": 29430 | |
| }, | |
| { | |
| "epoch": 58.92, | |
| "grad_norm": 56.43994903564453, | |
| "learning_rate": 7.992249455124889e-09, | |
| "loss": 0.3792, | |
| "step": 29460 | |
| }, | |
| { | |
| "epoch": 58.98, | |
| "grad_norm": 29.733278274536133, | |
| "learning_rate": 7.129094402451575e-09, | |
| "loss": 0.4046, | |
| "step": 29490 | |
| }, | |
| { | |
| "epoch": 59.0, | |
| "eval_loss": 0.2850831151008606, | |
| "eval_map": 0.8008, | |
| "eval_map_50": 0.9645, | |
| "eval_map_75": 0.9196, | |
| "eval_map_chicken": 0.8079, | |
| "eval_map_duck": 0.7689, | |
| "eval_map_large": 0.878, | |
| "eval_map_medium": 0.8002, | |
| "eval_map_plant": 0.8256, | |
| "eval_map_small": 0.2283, | |
| "eval_mar_1": 0.3087, | |
| "eval_mar_10": 0.835, | |
| "eval_mar_100": 0.8388, | |
| "eval_mar_100_chicken": 0.8418, | |
| "eval_mar_100_duck": 0.799, | |
| "eval_mar_100_plant": 0.8758, | |
| "eval_mar_large": 0.9188, | |
| "eval_mar_medium": 0.8497, | |
| "eval_mar_small": 0.3655, | |
| "eval_runtime": 5.5802, | |
| "eval_samples_per_second": 17.921, | |
| "eval_steps_per_second": 2.33, | |
| "step": 29500 | |
| }, | |
| { | |
| "epoch": 59.04, | |
| "grad_norm": 29.91386604309082, | |
| "learning_rate": 6.315216969912663e-09, | |
| "loss": 0.3851, | |
| "step": 29520 | |
| }, | |
| { | |
| "epoch": 59.1, | |
| "grad_norm": 505.2102355957031, | |
| "learning_rate": 5.5506251901504825e-09, | |
| "loss": 0.3666, | |
| "step": 29550 | |
| }, | |
| { | |
| "epoch": 59.16, | |
| "grad_norm": 45.7656135559082, | |
| "learning_rate": 4.835326609376468e-09, | |
| "loss": 0.4178, | |
| "step": 29580 | |
| }, | |
| { | |
| "epoch": 59.22, | |
| "grad_norm": 56.061153411865234, | |
| "learning_rate": 4.169328287299545e-09, | |
| "loss": 0.4697, | |
| "step": 29610 | |
| }, | |
| { | |
| "epoch": 59.28, | |
| "grad_norm": 93.67945098876953, | |
| "learning_rate": 3.5526367970539765e-09, | |
| "loss": 0.3581, | |
| "step": 29640 | |
| }, | |
| { | |
| "epoch": 59.34, | |
| "grad_norm": 56.09961700439453, | |
| "learning_rate": 2.9852582251355124e-09, | |
| "loss": 0.4007, | |
| "step": 29670 | |
| }, | |
| { | |
| "epoch": 59.4, | |
| "grad_norm": 95.72340393066406, | |
| "learning_rate": 2.4671981713420003e-09, | |
| "loss": 0.4338, | |
| "step": 29700 | |
| }, | |
| { | |
| "epoch": 59.46, | |
| "grad_norm": 73.806884765625, | |
| "learning_rate": 1.9984617487173174e-09, | |
| "loss": 0.4488, | |
| "step": 29730 | |
| }, | |
| { | |
| "epoch": 59.52, | |
| "grad_norm": 63.850746154785156, | |
| "learning_rate": 1.5790535835003006e-09, | |
| "loss": 0.4424, | |
| "step": 29760 | |
| }, | |
| { | |
| "epoch": 59.58, | |
| "grad_norm": 54.269371032714844, | |
| "learning_rate": 1.2089778150797816e-09, | |
| "loss": 0.4145, | |
| "step": 29790 | |
| }, | |
| { | |
| "epoch": 59.64, | |
| "grad_norm": 45.143714904785156, | |
| "learning_rate": 8.88238095955174e-10, | |
| "loss": 0.4422, | |
| "step": 29820 | |
| }, | |
| { | |
| "epoch": 59.7, | |
| "grad_norm": 32.879642486572266, | |
| "learning_rate": 6.168375916970615e-10, | |
| "loss": 0.3915, | |
| "step": 29850 | |
| }, | |
| { | |
| "epoch": 59.76, | |
| "grad_norm": 68.58180236816406, | |
| "learning_rate": 3.9477898091944135e-10, | |
| "loss": 0.423, | |
| "step": 29880 | |
| }, | |
| { | |
| "epoch": 59.82, | |
| "grad_norm": 57.879634857177734, | |
| "learning_rate": 2.2206445525085886e-10, | |
| "loss": 0.4619, | |
| "step": 29910 | |
| }, | |
| { | |
| "epoch": 59.88, | |
| "grad_norm": 133.5286102294922, | |
| "learning_rate": 9.869571931442334e-11, | |
| "loss": 0.3987, | |
| "step": 29940 | |
| }, | |
| { | |
| "epoch": 59.94, | |
| "grad_norm": 63.78921127319336, | |
| "learning_rate": 2.467399070893439e-11, | |
| "loss": 0.3533, | |
| "step": 29970 | |
| }, | |
| { | |
| "epoch": 60.0, | |
| "grad_norm": 65.49594116210938, | |
| "learning_rate": 0.0, | |
| "loss": 0.4504, | |
| "step": 30000 | |
| }, | |
| { | |
| "epoch": 60.0, | |
| "eval_loss": 0.28507041931152344, | |
| "eval_map": 0.8003, | |
| "eval_map_50": 0.9645, | |
| "eval_map_75": 0.9196, | |
| "eval_map_chicken": 0.8072, | |
| "eval_map_duck": 0.7689, | |
| "eval_map_large": 0.8772, | |
| "eval_map_medium": 0.7999, | |
| "eval_map_plant": 0.8248, | |
| "eval_map_small": 0.2246, | |
| "eval_mar_1": 0.3086, | |
| "eval_mar_10": 0.8346, | |
| "eval_mar_100": 0.8384, | |
| "eval_mar_100_chicken": 0.8413, | |
| "eval_mar_100_duck": 0.799, | |
| "eval_mar_100_plant": 0.8749, | |
| "eval_mar_large": 0.918, | |
| "eval_mar_medium": 0.8496, | |
| "eval_mar_small": 0.3614, | |
| "eval_runtime": 5.2733, | |
| "eval_samples_per_second": 18.963, | |
| "eval_steps_per_second": 2.465, | |
| "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 | |
| } | |