| { | |
| "best_metric": 0.9974395448079658, | |
| "best_model_checkpoint": "swin-tiny-patch4-window7-224-finetuned-eurosat/checkpoint-1482", | |
| "epoch": 2.9969666329625886, | |
| "eval_steps": 500, | |
| "global_step": 1482, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 3.3557046979865773e-06, | |
| "loss": 3.6597, | |
| "step": 10 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 6.7114093959731546e-06, | |
| "loss": 3.6229, | |
| "step": 20 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.006711409395973e-05, | |
| "loss": 3.5796, | |
| "step": 30 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.3422818791946309e-05, | |
| "loss": 3.4544, | |
| "step": 40 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.6778523489932888e-05, | |
| "loss": 3.2911, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 2.013422818791946e-05, | |
| "loss": 2.9492, | |
| "step": 60 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 2.348993288590604e-05, | |
| "loss": 2.4324, | |
| "step": 70 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 2.6845637583892618e-05, | |
| "loss": 1.9046, | |
| "step": 80 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 3.02013422818792e-05, | |
| "loss": 1.4181, | |
| "step": 90 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 3.3557046979865775e-05, | |
| "loss": 1.0132, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 3.6912751677852356e-05, | |
| "loss": 0.779, | |
| "step": 110 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 4.026845637583892e-05, | |
| "loss": 0.6354, | |
| "step": 120 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 4.36241610738255e-05, | |
| "loss": 0.5038, | |
| "step": 130 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 4.697986577181208e-05, | |
| "loss": 0.4519, | |
| "step": 140 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 4.996249062265567e-05, | |
| "loss": 0.3669, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 4.9587396849212303e-05, | |
| "loss": 0.3203, | |
| "step": 160 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 4.9212303075768945e-05, | |
| "loss": 0.3177, | |
| "step": 170 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 4.883720930232558e-05, | |
| "loss": 0.2744, | |
| "step": 180 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 4.846211552888222e-05, | |
| "loss": 0.2637, | |
| "step": 190 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 4.808702175543886e-05, | |
| "loss": 0.249, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 4.7711927981995505e-05, | |
| "loss": 0.2584, | |
| "step": 210 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 4.733683420855214e-05, | |
| "loss": 0.2276, | |
| "step": 220 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 4.696174043510878e-05, | |
| "loss": 0.1971, | |
| "step": 230 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 4.6586646661665416e-05, | |
| "loss": 0.1958, | |
| "step": 240 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 4.621155288822206e-05, | |
| "loss": 0.1829, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 4.58364591147787e-05, | |
| "loss": 0.1714, | |
| "step": 260 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 4.546136534133534e-05, | |
| "loss": 0.1904, | |
| "step": 270 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 4.5086271567891975e-05, | |
| "loss": 0.1736, | |
| "step": 280 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 4.471117779444862e-05, | |
| "loss": 0.1578, | |
| "step": 290 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 4.433608402100525e-05, | |
| "loss": 0.1952, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 4.396099024756189e-05, | |
| "loss": 0.1449, | |
| "step": 310 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 4.3585896474118535e-05, | |
| "loss": 0.1771, | |
| "step": 320 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 4.321080270067517e-05, | |
| "loss": 0.1376, | |
| "step": 330 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 4.283570892723181e-05, | |
| "loss": 0.1414, | |
| "step": 340 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 4.2460615153788446e-05, | |
| "loss": 0.1275, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 4.208552138034509e-05, | |
| "loss": 0.1444, | |
| "step": 360 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 4.171042760690172e-05, | |
| "loss": 0.1376, | |
| "step": 370 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 4.1335333833458364e-05, | |
| "loss": 0.1172, | |
| "step": 380 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 4.0960240060015006e-05, | |
| "loss": 0.1128, | |
| "step": 390 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 4.058514628657165e-05, | |
| "loss": 0.1337, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 4.021005251312828e-05, | |
| "loss": 0.12, | |
| "step": 410 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 3.9834958739684924e-05, | |
| "loss": 0.1224, | |
| "step": 420 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 3.945986496624156e-05, | |
| "loss": 0.1357, | |
| "step": 430 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 3.90847711927982e-05, | |
| "loss": 0.1105, | |
| "step": 440 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 3.870967741935484e-05, | |
| "loss": 0.0978, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 3.833458364591148e-05, | |
| "loss": 0.1001, | |
| "step": 460 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 3.795948987246812e-05, | |
| "loss": 0.108, | |
| "step": 470 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 3.758439609902476e-05, | |
| "loss": 0.1208, | |
| "step": 480 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 3.7209302325581394e-05, | |
| "loss": 0.1159, | |
| "step": 490 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.9863442389758179, | |
| "eval_loss": 0.03707945719361305, | |
| "eval_runtime": 52.4924, | |
| "eval_samples_per_second": 133.924, | |
| "eval_steps_per_second": 4.191, | |
| "step": 494 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 3.6834208552138036e-05, | |
| "loss": 0.1264, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 3.645911477869468e-05, | |
| "loss": 0.1034, | |
| "step": 510 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "learning_rate": 3.608402100525132e-05, | |
| "loss": 0.1023, | |
| "step": 520 | |
| }, | |
| { | |
| "epoch": 1.07, | |
| "learning_rate": 3.5708927231807954e-05, | |
| "loss": 0.1214, | |
| "step": 530 | |
| }, | |
| { | |
| "epoch": 1.09, | |
| "learning_rate": 3.533383345836459e-05, | |
| "loss": 0.0827, | |
| "step": 540 | |
| }, | |
| { | |
| "epoch": 1.11, | |
| "learning_rate": 3.495873968492123e-05, | |
| "loss": 0.0901, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 1.13, | |
| "learning_rate": 3.458364591147787e-05, | |
| "loss": 0.072, | |
| "step": 560 | |
| }, | |
| { | |
| "epoch": 1.15, | |
| "learning_rate": 3.4208552138034514e-05, | |
| "loss": 0.0913, | |
| "step": 570 | |
| }, | |
| { | |
| "epoch": 1.17, | |
| "learning_rate": 3.383345836459115e-05, | |
| "loss": 0.0906, | |
| "step": 580 | |
| }, | |
| { | |
| "epoch": 1.19, | |
| "learning_rate": 3.345836459114779e-05, | |
| "loss": 0.0949, | |
| "step": 590 | |
| }, | |
| { | |
| "epoch": 1.21, | |
| "learning_rate": 3.3083270817704425e-05, | |
| "loss": 0.1107, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 1.23, | |
| "learning_rate": 3.2708177044261066e-05, | |
| "loss": 0.0814, | |
| "step": 610 | |
| }, | |
| { | |
| "epoch": 1.25, | |
| "learning_rate": 3.23330832708177e-05, | |
| "loss": 0.0777, | |
| "step": 620 | |
| }, | |
| { | |
| "epoch": 1.27, | |
| "learning_rate": 3.195798949737435e-05, | |
| "loss": 0.0817, | |
| "step": 630 | |
| }, | |
| { | |
| "epoch": 1.29, | |
| "learning_rate": 3.1582895723930984e-05, | |
| "loss": 0.0847, | |
| "step": 640 | |
| }, | |
| { | |
| "epoch": 1.31, | |
| "learning_rate": 3.1207801950487626e-05, | |
| "loss": 0.0832, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 1.33, | |
| "learning_rate": 3.083270817704426e-05, | |
| "loss": 0.0772, | |
| "step": 660 | |
| }, | |
| { | |
| "epoch": 1.35, | |
| "learning_rate": 3.0457614403600902e-05, | |
| "loss": 0.0872, | |
| "step": 670 | |
| }, | |
| { | |
| "epoch": 1.38, | |
| "learning_rate": 3.008252063015754e-05, | |
| "loss": 0.0829, | |
| "step": 680 | |
| }, | |
| { | |
| "epoch": 1.4, | |
| "learning_rate": 2.9707426856714182e-05, | |
| "loss": 0.0813, | |
| "step": 690 | |
| }, | |
| { | |
| "epoch": 1.42, | |
| "learning_rate": 2.933233308327082e-05, | |
| "loss": 0.1099, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 1.44, | |
| "learning_rate": 2.8957239309827462e-05, | |
| "loss": 0.0763, | |
| "step": 710 | |
| }, | |
| { | |
| "epoch": 1.46, | |
| "learning_rate": 2.8582145536384097e-05, | |
| "loss": 0.0664, | |
| "step": 720 | |
| }, | |
| { | |
| "epoch": 1.48, | |
| "learning_rate": 2.8207051762940735e-05, | |
| "loss": 0.0808, | |
| "step": 730 | |
| }, | |
| { | |
| "epoch": 1.5, | |
| "learning_rate": 2.7831957989497377e-05, | |
| "loss": 0.052, | |
| "step": 740 | |
| }, | |
| { | |
| "epoch": 1.52, | |
| "learning_rate": 2.745686421605401e-05, | |
| "loss": 0.0705, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 1.54, | |
| "learning_rate": 2.7081770442610653e-05, | |
| "loss": 0.0664, | |
| "step": 760 | |
| }, | |
| { | |
| "epoch": 1.56, | |
| "learning_rate": 2.670667666916729e-05, | |
| "loss": 0.0954, | |
| "step": 770 | |
| }, | |
| { | |
| "epoch": 1.58, | |
| "learning_rate": 2.6331582895723933e-05, | |
| "loss": 0.0645, | |
| "step": 780 | |
| }, | |
| { | |
| "epoch": 1.6, | |
| "learning_rate": 2.595648912228057e-05, | |
| "loss": 0.0773, | |
| "step": 790 | |
| }, | |
| { | |
| "epoch": 1.62, | |
| "learning_rate": 2.5581395348837212e-05, | |
| "loss": 0.075, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 1.64, | |
| "learning_rate": 2.5206301575393847e-05, | |
| "loss": 0.066, | |
| "step": 810 | |
| }, | |
| { | |
| "epoch": 1.66, | |
| "learning_rate": 2.483120780195049e-05, | |
| "loss": 0.0677, | |
| "step": 820 | |
| }, | |
| { | |
| "epoch": 1.68, | |
| "learning_rate": 2.4456114028507127e-05, | |
| "loss": 0.0625, | |
| "step": 830 | |
| }, | |
| { | |
| "epoch": 1.7, | |
| "learning_rate": 2.4081020255063765e-05, | |
| "loss": 0.0696, | |
| "step": 840 | |
| }, | |
| { | |
| "epoch": 1.72, | |
| "learning_rate": 2.3705926481620407e-05, | |
| "loss": 0.0711, | |
| "step": 850 | |
| }, | |
| { | |
| "epoch": 1.74, | |
| "learning_rate": 2.3330832708177045e-05, | |
| "loss": 0.0501, | |
| "step": 860 | |
| }, | |
| { | |
| "epoch": 1.76, | |
| "learning_rate": 2.2955738934733683e-05, | |
| "loss": 0.0575, | |
| "step": 870 | |
| }, | |
| { | |
| "epoch": 1.78, | |
| "learning_rate": 2.258064516129032e-05, | |
| "loss": 0.0553, | |
| "step": 880 | |
| }, | |
| { | |
| "epoch": 1.8, | |
| "learning_rate": 2.2205551387846963e-05, | |
| "loss": 0.0509, | |
| "step": 890 | |
| }, | |
| { | |
| "epoch": 1.82, | |
| "learning_rate": 2.18304576144036e-05, | |
| "loss": 0.0746, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 1.84, | |
| "learning_rate": 2.145536384096024e-05, | |
| "loss": 0.0585, | |
| "step": 910 | |
| }, | |
| { | |
| "epoch": 1.86, | |
| "learning_rate": 2.108027006751688e-05, | |
| "loss": 0.0737, | |
| "step": 920 | |
| }, | |
| { | |
| "epoch": 1.88, | |
| "learning_rate": 2.070517629407352e-05, | |
| "loss": 0.0637, | |
| "step": 930 | |
| }, | |
| { | |
| "epoch": 1.9, | |
| "learning_rate": 2.0330082520630157e-05, | |
| "loss": 0.0729, | |
| "step": 940 | |
| }, | |
| { | |
| "epoch": 1.92, | |
| "learning_rate": 1.99549887471868e-05, | |
| "loss": 0.0719, | |
| "step": 950 | |
| }, | |
| { | |
| "epoch": 1.94, | |
| "learning_rate": 1.9579894973743437e-05, | |
| "loss": 0.0691, | |
| "step": 960 | |
| }, | |
| { | |
| "epoch": 1.96, | |
| "learning_rate": 1.9204801200300075e-05, | |
| "loss": 0.0783, | |
| "step": 970 | |
| }, | |
| { | |
| "epoch": 1.98, | |
| "learning_rate": 1.8829707426856717e-05, | |
| "loss": 0.0765, | |
| "step": 980 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.9958748221906116, | |
| "eval_loss": 0.013012241572141647, | |
| "eval_runtime": 52.9452, | |
| "eval_samples_per_second": 132.779, | |
| "eval_steps_per_second": 4.155, | |
| "step": 989 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "learning_rate": 1.8454613653413355e-05, | |
| "loss": 0.0765, | |
| "step": 990 | |
| }, | |
| { | |
| "epoch": 2.02, | |
| "learning_rate": 1.8079519879969993e-05, | |
| "loss": 0.0501, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 2.04, | |
| "learning_rate": 1.7704426106526635e-05, | |
| "loss": 0.0482, | |
| "step": 1010 | |
| }, | |
| { | |
| "epoch": 2.06, | |
| "learning_rate": 1.7329332333083273e-05, | |
| "loss": 0.0618, | |
| "step": 1020 | |
| }, | |
| { | |
| "epoch": 2.08, | |
| "learning_rate": 1.6954238559639908e-05, | |
| "loss": 0.0414, | |
| "step": 1030 | |
| }, | |
| { | |
| "epoch": 2.1, | |
| "learning_rate": 1.657914478619655e-05, | |
| "loss": 0.044, | |
| "step": 1040 | |
| }, | |
| { | |
| "epoch": 2.12, | |
| "learning_rate": 1.6204051012753188e-05, | |
| "loss": 0.0585, | |
| "step": 1050 | |
| }, | |
| { | |
| "epoch": 2.14, | |
| "learning_rate": 1.5828957239309826e-05, | |
| "loss": 0.0554, | |
| "step": 1060 | |
| }, | |
| { | |
| "epoch": 2.16, | |
| "learning_rate": 1.5453863465866468e-05, | |
| "loss": 0.0421, | |
| "step": 1070 | |
| }, | |
| { | |
| "epoch": 2.18, | |
| "learning_rate": 1.5078769692423106e-05, | |
| "loss": 0.0641, | |
| "step": 1080 | |
| }, | |
| { | |
| "epoch": 2.2, | |
| "learning_rate": 1.4703675918979746e-05, | |
| "loss": 0.0537, | |
| "step": 1090 | |
| }, | |
| { | |
| "epoch": 2.22, | |
| "learning_rate": 1.4328582145536384e-05, | |
| "loss": 0.0567, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 2.24, | |
| "learning_rate": 1.3953488372093024e-05, | |
| "loss": 0.0488, | |
| "step": 1110 | |
| }, | |
| { | |
| "epoch": 2.26, | |
| "learning_rate": 1.3578394598649664e-05, | |
| "loss": 0.0541, | |
| "step": 1120 | |
| }, | |
| { | |
| "epoch": 2.29, | |
| "learning_rate": 1.3203300825206302e-05, | |
| "loss": 0.0506, | |
| "step": 1130 | |
| }, | |
| { | |
| "epoch": 2.31, | |
| "learning_rate": 1.2828207051762942e-05, | |
| "loss": 0.0584, | |
| "step": 1140 | |
| }, | |
| { | |
| "epoch": 2.33, | |
| "learning_rate": 1.2453113278319582e-05, | |
| "loss": 0.0607, | |
| "step": 1150 | |
| }, | |
| { | |
| "epoch": 2.35, | |
| "learning_rate": 1.207801950487622e-05, | |
| "loss": 0.0534, | |
| "step": 1160 | |
| }, | |
| { | |
| "epoch": 2.37, | |
| "learning_rate": 1.170292573143286e-05, | |
| "loss": 0.0414, | |
| "step": 1170 | |
| }, | |
| { | |
| "epoch": 2.39, | |
| "learning_rate": 1.1327831957989498e-05, | |
| "loss": 0.0594, | |
| "step": 1180 | |
| }, | |
| { | |
| "epoch": 2.41, | |
| "learning_rate": 1.0952738184546136e-05, | |
| "loss": 0.0642, | |
| "step": 1190 | |
| }, | |
| { | |
| "epoch": 2.43, | |
| "learning_rate": 1.0577644411102776e-05, | |
| "loss": 0.0515, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 2.45, | |
| "learning_rate": 1.0202550637659416e-05, | |
| "loss": 0.039, | |
| "step": 1210 | |
| }, | |
| { | |
| "epoch": 2.47, | |
| "learning_rate": 9.827456864216054e-06, | |
| "loss": 0.0297, | |
| "step": 1220 | |
| }, | |
| { | |
| "epoch": 2.49, | |
| "learning_rate": 9.452363090772694e-06, | |
| "loss": 0.0459, | |
| "step": 1230 | |
| }, | |
| { | |
| "epoch": 2.51, | |
| "learning_rate": 9.077269317329332e-06, | |
| "loss": 0.0464, | |
| "step": 1240 | |
| }, | |
| { | |
| "epoch": 2.53, | |
| "learning_rate": 8.702175543885972e-06, | |
| "loss": 0.0524, | |
| "step": 1250 | |
| }, | |
| { | |
| "epoch": 2.55, | |
| "learning_rate": 8.327081770442612e-06, | |
| "loss": 0.041, | |
| "step": 1260 | |
| }, | |
| { | |
| "epoch": 2.57, | |
| "learning_rate": 7.95198799699925e-06, | |
| "loss": 0.0393, | |
| "step": 1270 | |
| }, | |
| { | |
| "epoch": 2.59, | |
| "learning_rate": 7.576894223555888e-06, | |
| "loss": 0.0375, | |
| "step": 1280 | |
| }, | |
| { | |
| "epoch": 2.61, | |
| "learning_rate": 7.201800450112528e-06, | |
| "loss": 0.0536, | |
| "step": 1290 | |
| }, | |
| { | |
| "epoch": 2.63, | |
| "learning_rate": 6.826706676669167e-06, | |
| "loss": 0.0264, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 2.65, | |
| "learning_rate": 6.451612903225806e-06, | |
| "loss": 0.0414, | |
| "step": 1310 | |
| }, | |
| { | |
| "epoch": 2.67, | |
| "learning_rate": 6.076519129782446e-06, | |
| "loss": 0.0296, | |
| "step": 1320 | |
| }, | |
| { | |
| "epoch": 2.69, | |
| "learning_rate": 5.701425356339085e-06, | |
| "loss": 0.0377, | |
| "step": 1330 | |
| }, | |
| { | |
| "epoch": 2.71, | |
| "learning_rate": 5.326331582895724e-06, | |
| "loss": 0.0321, | |
| "step": 1340 | |
| }, | |
| { | |
| "epoch": 2.73, | |
| "learning_rate": 4.951237809452363e-06, | |
| "loss": 0.0513, | |
| "step": 1350 | |
| }, | |
| { | |
| "epoch": 2.75, | |
| "learning_rate": 4.5761440360090025e-06, | |
| "loss": 0.0387, | |
| "step": 1360 | |
| }, | |
| { | |
| "epoch": 2.77, | |
| "learning_rate": 4.2010502625656415e-06, | |
| "loss": 0.03, | |
| "step": 1370 | |
| }, | |
| { | |
| "epoch": 2.79, | |
| "learning_rate": 3.8259564891222805e-06, | |
| "loss": 0.0434, | |
| "step": 1380 | |
| }, | |
| { | |
| "epoch": 2.81, | |
| "learning_rate": 3.45086271567892e-06, | |
| "loss": 0.0348, | |
| "step": 1390 | |
| }, | |
| { | |
| "epoch": 2.83, | |
| "learning_rate": 3.075768942235559e-06, | |
| "loss": 0.0468, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 2.85, | |
| "learning_rate": 2.700675168792198e-06, | |
| "loss": 0.0435, | |
| "step": 1410 | |
| }, | |
| { | |
| "epoch": 2.87, | |
| "learning_rate": 2.325581395348837e-06, | |
| "loss": 0.0346, | |
| "step": 1420 | |
| }, | |
| { | |
| "epoch": 2.89, | |
| "learning_rate": 1.9504876219054766e-06, | |
| "loss": 0.0404, | |
| "step": 1430 | |
| }, | |
| { | |
| "epoch": 2.91, | |
| "learning_rate": 1.5753938484621155e-06, | |
| "loss": 0.0389, | |
| "step": 1440 | |
| }, | |
| { | |
| "epoch": 2.93, | |
| "learning_rate": 1.2003000750187547e-06, | |
| "loss": 0.0466, | |
| "step": 1450 | |
| }, | |
| { | |
| "epoch": 2.95, | |
| "learning_rate": 8.252063015753939e-07, | |
| "loss": 0.038, | |
| "step": 1460 | |
| }, | |
| { | |
| "epoch": 2.97, | |
| "learning_rate": 4.50112528132033e-07, | |
| "loss": 0.0516, | |
| "step": 1470 | |
| }, | |
| { | |
| "epoch": 2.99, | |
| "learning_rate": 7.501875468867217e-08, | |
| "loss": 0.0346, | |
| "step": 1480 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.9974395448079658, | |
| "eval_loss": 0.007520665414631367, | |
| "eval_runtime": 51.8915, | |
| "eval_samples_per_second": 135.475, | |
| "eval_steps_per_second": 4.24, | |
| "step": 1482 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "step": 1482, | |
| "total_flos": 4.718282091688231e+18, | |
| "train_loss": 0.2835471658400598, | |
| "train_runtime": 3123.0381, | |
| "train_samples_per_second": 60.773, | |
| "train_steps_per_second": 0.475 | |
| } | |
| ], | |
| "logging_steps": 10, | |
| "max_steps": 1482, | |
| "num_train_epochs": 3, | |
| "save_steps": 500, | |
| "total_flos": 4.718282091688231e+18, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |