| { | |
| "best_metric": 0.8289, | |
| "best_model_checkpoint": "swin-tiny-patch4-window7-224-finetuned-eurosat\\checkpoint-2109", | |
| "epoch": 2.998933522929257, | |
| "global_step": 2109, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 2.3696682464454976e-06, | |
| "loss": 5.3748, | |
| "step": 10 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 4.739336492890995e-06, | |
| "loss": 5.3569, | |
| "step": 20 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 7.109004739336493e-06, | |
| "loss": 5.3245, | |
| "step": 30 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 9.47867298578199e-06, | |
| "loss": 5.3197, | |
| "step": 40 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.184834123222749e-05, | |
| "loss": 5.3088, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.4218009478672985e-05, | |
| "loss": 5.2824, | |
| "step": 60 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.6587677725118483e-05, | |
| "loss": 5.2667, | |
| "step": 70 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.895734597156398e-05, | |
| "loss": 5.2398, | |
| "step": 80 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 2.132701421800948e-05, | |
| "loss": 5.1939, | |
| "step": 90 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 2.369668246445498e-05, | |
| "loss": 5.1625, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 2.6066350710900477e-05, | |
| "loss": 5.1027, | |
| "step": 110 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 2.843601895734597e-05, | |
| "loss": 5.0203, | |
| "step": 120 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 3.080568720379147e-05, | |
| "loss": 4.9328, | |
| "step": 130 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 3.3175355450236966e-05, | |
| "loss": 4.7904, | |
| "step": 140 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 3.554502369668247e-05, | |
| "loss": 4.6625, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 3.791469194312796e-05, | |
| "loss": 4.4841, | |
| "step": 160 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 4.028436018957346e-05, | |
| "loss": 4.2482, | |
| "step": 170 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 4.265402843601896e-05, | |
| "loss": 4.0752, | |
| "step": 180 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 4.502369668246446e-05, | |
| "loss": 3.9379, | |
| "step": 190 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 4.739336492890996e-05, | |
| "loss": 3.734, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 4.976303317535545e-05, | |
| "loss": 3.5028, | |
| "step": 210 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 4.976290832455216e-05, | |
| "loss": 3.4281, | |
| "step": 220 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 4.949947312961012e-05, | |
| "loss": 3.2753, | |
| "step": 230 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 4.923603793466807e-05, | |
| "loss": 3.0507, | |
| "step": 240 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 4.8972602739726034e-05, | |
| "loss": 2.9402, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 4.8709167544783986e-05, | |
| "loss": 2.8824, | |
| "step": 260 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 4.8445732349841945e-05, | |
| "loss": 2.6848, | |
| "step": 270 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 4.8182297154899896e-05, | |
| "loss": 2.6702, | |
| "step": 280 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 4.791886195995785e-05, | |
| "loss": 2.6096, | |
| "step": 290 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 4.7655426765015806e-05, | |
| "loss": 2.5345, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 4.7391991570073765e-05, | |
| "loss": 2.5112, | |
| "step": 310 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 4.712855637513172e-05, | |
| "loss": 2.4971, | |
| "step": 320 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 4.6865121180189675e-05, | |
| "loss": 2.3219, | |
| "step": 330 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 4.6601685985247633e-05, | |
| "loss": 2.4026, | |
| "step": 340 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 4.6338250790305585e-05, | |
| "loss": 2.2161, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 4.6074815595363544e-05, | |
| "loss": 2.2615, | |
| "step": 360 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 4.58113804004215e-05, | |
| "loss": 2.2194, | |
| "step": 370 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 4.5547945205479454e-05, | |
| "loss": 2.2824, | |
| "step": 380 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 4.528451001053741e-05, | |
| "loss": 2.1638, | |
| "step": 390 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 4.5021074815595364e-05, | |
| "loss": 2.0788, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 4.4757639620653316e-05, | |
| "loss": 2.0187, | |
| "step": 410 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 4.449420442571128e-05, | |
| "loss": 2.0782, | |
| "step": 420 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 4.423076923076923e-05, | |
| "loss": 2.0265, | |
| "step": 430 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 4.396733403582719e-05, | |
| "loss": 1.9744, | |
| "step": 440 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 4.370389884088514e-05, | |
| "loss": 2.006, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 4.34404636459431e-05, | |
| "loss": 2.0586, | |
| "step": 460 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 4.317702845100105e-05, | |
| "loss": 1.9814, | |
| "step": 470 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 4.291359325605901e-05, | |
| "loss": 1.9808, | |
| "step": 480 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 4.265015806111697e-05, | |
| "loss": 1.9932, | |
| "step": 490 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 4.238672286617492e-05, | |
| "loss": 1.9604, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 4.212328767123288e-05, | |
| "loss": 1.8877, | |
| "step": 510 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 4.185985247629083e-05, | |
| "loss": 1.884, | |
| "step": 520 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 4.159641728134879e-05, | |
| "loss": 1.9931, | |
| "step": 530 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 4.133298208640675e-05, | |
| "loss": 1.9496, | |
| "step": 540 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 4.10695468914647e-05, | |
| "loss": 1.7859, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 4.080611169652266e-05, | |
| "loss": 1.8249, | |
| "step": 560 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 4.054267650158061e-05, | |
| "loss": 1.7965, | |
| "step": 570 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 4.027924130663857e-05, | |
| "loss": 1.8011, | |
| "step": 580 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 4.001580611169653e-05, | |
| "loss": 1.7976, | |
| "step": 590 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 3.975237091675448e-05, | |
| "loss": 1.7056, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 3.948893572181244e-05, | |
| "loss": 1.9267, | |
| "step": 610 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 3.922550052687039e-05, | |
| "loss": 1.7014, | |
| "step": 620 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 3.896206533192835e-05, | |
| "loss": 1.8617, | |
| "step": 630 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 3.86986301369863e-05, | |
| "loss": 1.8402, | |
| "step": 640 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 3.843519494204426e-05, | |
| "loss": 1.7088, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 3.8171759747102217e-05, | |
| "loss": 1.8482, | |
| "step": 660 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 3.790832455216017e-05, | |
| "loss": 1.7017, | |
| "step": 670 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 3.764488935721813e-05, | |
| "loss": 1.8167, | |
| "step": 680 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 3.738145416227608e-05, | |
| "loss": 1.7691, | |
| "step": 690 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 3.711801896733404e-05, | |
| "loss": 1.7922, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.7804, | |
| "eval_loss": 0.8853043913841248, | |
| "eval_runtime": 39.7328, | |
| "eval_samples_per_second": 251.681, | |
| "eval_steps_per_second": 7.878, | |
| "step": 703 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 3.6854583772391995e-05, | |
| "loss": 1.5741, | |
| "step": 710 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 3.6591148577449954e-05, | |
| "loss": 1.6612, | |
| "step": 720 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "learning_rate": 3.6327713382507905e-05, | |
| "loss": 1.6439, | |
| "step": 730 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "learning_rate": 3.606427818756586e-05, | |
| "loss": 1.6165, | |
| "step": 740 | |
| }, | |
| { | |
| "epoch": 1.07, | |
| "learning_rate": 3.5800842992623816e-05, | |
| "loss": 1.6161, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "learning_rate": 3.5537407797681774e-05, | |
| "loss": 1.5302, | |
| "step": 760 | |
| }, | |
| { | |
| "epoch": 1.09, | |
| "learning_rate": 3.527397260273973e-05, | |
| "loss": 1.6897, | |
| "step": 770 | |
| }, | |
| { | |
| "epoch": 1.11, | |
| "learning_rate": 3.5010537407797684e-05, | |
| "loss": 1.6632, | |
| "step": 780 | |
| }, | |
| { | |
| "epoch": 1.12, | |
| "learning_rate": 3.4747102212855636e-05, | |
| "loss": 1.5924, | |
| "step": 790 | |
| }, | |
| { | |
| "epoch": 1.14, | |
| "learning_rate": 3.4483667017913594e-05, | |
| "loss": 1.6301, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 1.15, | |
| "learning_rate": 3.4220231822971546e-05, | |
| "loss": 1.6031, | |
| "step": 810 | |
| }, | |
| { | |
| "epoch": 1.17, | |
| "learning_rate": 3.3956796628029505e-05, | |
| "loss": 1.7069, | |
| "step": 820 | |
| }, | |
| { | |
| "epoch": 1.18, | |
| "learning_rate": 3.369336143308746e-05, | |
| "loss": 1.6232, | |
| "step": 830 | |
| }, | |
| { | |
| "epoch": 1.19, | |
| "learning_rate": 3.342992623814542e-05, | |
| "loss": 1.6028, | |
| "step": 840 | |
| }, | |
| { | |
| "epoch": 1.21, | |
| "learning_rate": 3.316649104320337e-05, | |
| "loss": 1.5437, | |
| "step": 850 | |
| }, | |
| { | |
| "epoch": 1.22, | |
| "learning_rate": 3.2903055848261325e-05, | |
| "loss": 1.534, | |
| "step": 860 | |
| }, | |
| { | |
| "epoch": 1.24, | |
| "learning_rate": 3.2639620653319283e-05, | |
| "loss": 1.5398, | |
| "step": 870 | |
| }, | |
| { | |
| "epoch": 1.25, | |
| "learning_rate": 3.237618545837724e-05, | |
| "loss": 1.4951, | |
| "step": 880 | |
| }, | |
| { | |
| "epoch": 1.27, | |
| "learning_rate": 3.21127502634352e-05, | |
| "loss": 1.6149, | |
| "step": 890 | |
| }, | |
| { | |
| "epoch": 1.28, | |
| "learning_rate": 3.184931506849315e-05, | |
| "loss": 1.6144, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 1.29, | |
| "learning_rate": 3.1585879873551104e-05, | |
| "loss": 1.5772, | |
| "step": 910 | |
| }, | |
| { | |
| "epoch": 1.31, | |
| "learning_rate": 3.132244467860906e-05, | |
| "loss": 1.4824, | |
| "step": 920 | |
| }, | |
| { | |
| "epoch": 1.32, | |
| "learning_rate": 3.105900948366702e-05, | |
| "loss": 1.5298, | |
| "step": 930 | |
| }, | |
| { | |
| "epoch": 1.34, | |
| "learning_rate": 3.079557428872498e-05, | |
| "loss": 1.5792, | |
| "step": 940 | |
| }, | |
| { | |
| "epoch": 1.35, | |
| "learning_rate": 3.053213909378293e-05, | |
| "loss": 1.5499, | |
| "step": 950 | |
| }, | |
| { | |
| "epoch": 1.37, | |
| "learning_rate": 3.026870389884089e-05, | |
| "loss": 1.5453, | |
| "step": 960 | |
| }, | |
| { | |
| "epoch": 1.38, | |
| "learning_rate": 3.000526870389884e-05, | |
| "loss": 1.5818, | |
| "step": 970 | |
| }, | |
| { | |
| "epoch": 1.39, | |
| "learning_rate": 2.9741833508956796e-05, | |
| "loss": 1.4333, | |
| "step": 980 | |
| }, | |
| { | |
| "epoch": 1.41, | |
| "learning_rate": 2.9478398314014755e-05, | |
| "loss": 1.5632, | |
| "step": 990 | |
| }, | |
| { | |
| "epoch": 1.42, | |
| "learning_rate": 2.921496311907271e-05, | |
| "loss": 1.525, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 1.44, | |
| "learning_rate": 2.8951527924130668e-05, | |
| "loss": 1.5256, | |
| "step": 1010 | |
| }, | |
| { | |
| "epoch": 1.45, | |
| "learning_rate": 2.868809272918862e-05, | |
| "loss": 1.5352, | |
| "step": 1020 | |
| }, | |
| { | |
| "epoch": 1.46, | |
| "learning_rate": 2.842465753424658e-05, | |
| "loss": 1.4807, | |
| "step": 1030 | |
| }, | |
| { | |
| "epoch": 1.48, | |
| "learning_rate": 2.8161222339304533e-05, | |
| "loss": 1.5398, | |
| "step": 1040 | |
| }, | |
| { | |
| "epoch": 1.49, | |
| "learning_rate": 2.7897787144362485e-05, | |
| "loss": 1.559, | |
| "step": 1050 | |
| }, | |
| { | |
| "epoch": 1.51, | |
| "learning_rate": 2.7634351949420444e-05, | |
| "loss": 1.5625, | |
| "step": 1060 | |
| }, | |
| { | |
| "epoch": 1.52, | |
| "learning_rate": 2.73709167544784e-05, | |
| "loss": 1.4423, | |
| "step": 1070 | |
| }, | |
| { | |
| "epoch": 1.54, | |
| "learning_rate": 2.7107481559536357e-05, | |
| "loss": 1.4928, | |
| "step": 1080 | |
| }, | |
| { | |
| "epoch": 1.55, | |
| "learning_rate": 2.6844046364594312e-05, | |
| "loss": 1.6027, | |
| "step": 1090 | |
| }, | |
| { | |
| "epoch": 1.56, | |
| "learning_rate": 2.6580611169652264e-05, | |
| "loss": 1.4623, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 1.58, | |
| "learning_rate": 2.6317175974710222e-05, | |
| "loss": 1.6102, | |
| "step": 1110 | |
| }, | |
| { | |
| "epoch": 1.59, | |
| "learning_rate": 2.6053740779768177e-05, | |
| "loss": 1.4465, | |
| "step": 1120 | |
| }, | |
| { | |
| "epoch": 1.61, | |
| "learning_rate": 2.5790305584826136e-05, | |
| "loss": 1.4764, | |
| "step": 1130 | |
| }, | |
| { | |
| "epoch": 1.62, | |
| "learning_rate": 2.5526870389884088e-05, | |
| "loss": 1.5132, | |
| "step": 1140 | |
| }, | |
| { | |
| "epoch": 1.64, | |
| "learning_rate": 2.5263435194942046e-05, | |
| "loss": 1.5237, | |
| "step": 1150 | |
| }, | |
| { | |
| "epoch": 1.65, | |
| "learning_rate": 2.5e-05, | |
| "loss": 1.5527, | |
| "step": 1160 | |
| }, | |
| { | |
| "epoch": 1.66, | |
| "learning_rate": 2.4736564805057956e-05, | |
| "loss": 1.4051, | |
| "step": 1170 | |
| }, | |
| { | |
| "epoch": 1.68, | |
| "learning_rate": 2.4473129610115915e-05, | |
| "loss": 1.6083, | |
| "step": 1180 | |
| }, | |
| { | |
| "epoch": 1.69, | |
| "learning_rate": 2.420969441517387e-05, | |
| "loss": 1.4604, | |
| "step": 1190 | |
| }, | |
| { | |
| "epoch": 1.71, | |
| "learning_rate": 2.394625922023182e-05, | |
| "loss": 1.5632, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 1.72, | |
| "learning_rate": 2.368282402528978e-05, | |
| "loss": 1.5384, | |
| "step": 1210 | |
| }, | |
| { | |
| "epoch": 1.73, | |
| "learning_rate": 2.3419388830347735e-05, | |
| "loss": 1.4031, | |
| "step": 1220 | |
| }, | |
| { | |
| "epoch": 1.75, | |
| "learning_rate": 2.315595363540569e-05, | |
| "loss": 1.5656, | |
| "step": 1230 | |
| }, | |
| { | |
| "epoch": 1.76, | |
| "learning_rate": 2.289251844046365e-05, | |
| "loss": 1.3992, | |
| "step": 1240 | |
| }, | |
| { | |
| "epoch": 1.78, | |
| "learning_rate": 2.2629083245521604e-05, | |
| "loss": 1.4831, | |
| "step": 1250 | |
| }, | |
| { | |
| "epoch": 1.79, | |
| "learning_rate": 2.236564805057956e-05, | |
| "loss": 1.4081, | |
| "step": 1260 | |
| }, | |
| { | |
| "epoch": 1.81, | |
| "learning_rate": 2.2102212855637514e-05, | |
| "loss": 1.5012, | |
| "step": 1270 | |
| }, | |
| { | |
| "epoch": 1.82, | |
| "learning_rate": 2.183877766069547e-05, | |
| "loss": 1.4697, | |
| "step": 1280 | |
| }, | |
| { | |
| "epoch": 1.83, | |
| "learning_rate": 2.1575342465753427e-05, | |
| "loss": 1.4538, | |
| "step": 1290 | |
| }, | |
| { | |
| "epoch": 1.85, | |
| "learning_rate": 2.1311907270811383e-05, | |
| "loss": 1.372, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 1.86, | |
| "learning_rate": 2.1048472075869338e-05, | |
| "loss": 1.5106, | |
| "step": 1310 | |
| }, | |
| { | |
| "epoch": 1.88, | |
| "learning_rate": 2.0785036880927293e-05, | |
| "loss": 1.518, | |
| "step": 1320 | |
| }, | |
| { | |
| "epoch": 1.89, | |
| "learning_rate": 2.0521601685985248e-05, | |
| "loss": 1.4169, | |
| "step": 1330 | |
| }, | |
| { | |
| "epoch": 1.91, | |
| "learning_rate": 2.0258166491043203e-05, | |
| "loss": 1.4372, | |
| "step": 1340 | |
| }, | |
| { | |
| "epoch": 1.92, | |
| "learning_rate": 1.999473129610116e-05, | |
| "loss": 1.5191, | |
| "step": 1350 | |
| }, | |
| { | |
| "epoch": 1.93, | |
| "learning_rate": 1.9731296101159116e-05, | |
| "loss": 1.4005, | |
| "step": 1360 | |
| }, | |
| { | |
| "epoch": 1.95, | |
| "learning_rate": 1.946786090621707e-05, | |
| "loss": 1.4019, | |
| "step": 1370 | |
| }, | |
| { | |
| "epoch": 1.96, | |
| "learning_rate": 1.9204425711275027e-05, | |
| "loss": 1.4445, | |
| "step": 1380 | |
| }, | |
| { | |
| "epoch": 1.98, | |
| "learning_rate": 1.894099051633298e-05, | |
| "loss": 1.3176, | |
| "step": 1390 | |
| }, | |
| { | |
| "epoch": 1.99, | |
| "learning_rate": 1.8677555321390937e-05, | |
| "loss": 1.4677, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.8124, | |
| "eval_loss": 0.7063834071159363, | |
| "eval_runtime": 34.1843, | |
| "eval_samples_per_second": 292.532, | |
| "eval_steps_per_second": 9.156, | |
| "step": 1406 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "learning_rate": 1.8414120126448895e-05, | |
| "loss": 1.4671, | |
| "step": 1410 | |
| }, | |
| { | |
| "epoch": 2.02, | |
| "learning_rate": 1.815068493150685e-05, | |
| "loss": 1.2756, | |
| "step": 1420 | |
| }, | |
| { | |
| "epoch": 2.03, | |
| "learning_rate": 1.7887249736564805e-05, | |
| "loss": 1.3293, | |
| "step": 1430 | |
| }, | |
| { | |
| "epoch": 2.05, | |
| "learning_rate": 1.7623814541622764e-05, | |
| "loss": 1.3382, | |
| "step": 1440 | |
| }, | |
| { | |
| "epoch": 2.06, | |
| "learning_rate": 1.7360379346680716e-05, | |
| "loss": 1.4155, | |
| "step": 1450 | |
| }, | |
| { | |
| "epoch": 2.08, | |
| "learning_rate": 1.7096944151738674e-05, | |
| "loss": 1.331, | |
| "step": 1460 | |
| }, | |
| { | |
| "epoch": 2.09, | |
| "learning_rate": 1.683350895679663e-05, | |
| "loss": 1.4016, | |
| "step": 1470 | |
| }, | |
| { | |
| "epoch": 2.1, | |
| "learning_rate": 1.6570073761854584e-05, | |
| "loss": 1.361, | |
| "step": 1480 | |
| }, | |
| { | |
| "epoch": 2.12, | |
| "learning_rate": 1.630663856691254e-05, | |
| "loss": 1.3983, | |
| "step": 1490 | |
| }, | |
| { | |
| "epoch": 2.13, | |
| "learning_rate": 1.6043203371970498e-05, | |
| "loss": 1.3422, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 2.15, | |
| "learning_rate": 1.577976817702845e-05, | |
| "loss": 1.3971, | |
| "step": 1510 | |
| }, | |
| { | |
| "epoch": 2.16, | |
| "learning_rate": 1.5516332982086408e-05, | |
| "loss": 1.422, | |
| "step": 1520 | |
| }, | |
| { | |
| "epoch": 2.18, | |
| "learning_rate": 1.5252897787144363e-05, | |
| "loss": 1.4001, | |
| "step": 1530 | |
| }, | |
| { | |
| "epoch": 2.19, | |
| "learning_rate": 1.498946259220232e-05, | |
| "loss": 1.329, | |
| "step": 1540 | |
| }, | |
| { | |
| "epoch": 2.2, | |
| "learning_rate": 1.4726027397260275e-05, | |
| "loss": 1.3531, | |
| "step": 1550 | |
| }, | |
| { | |
| "epoch": 2.22, | |
| "learning_rate": 1.4462592202318232e-05, | |
| "loss": 1.3514, | |
| "step": 1560 | |
| }, | |
| { | |
| "epoch": 2.23, | |
| "learning_rate": 1.4199157007376185e-05, | |
| "loss": 1.32, | |
| "step": 1570 | |
| }, | |
| { | |
| "epoch": 2.25, | |
| "learning_rate": 1.3935721812434142e-05, | |
| "loss": 1.4405, | |
| "step": 1580 | |
| }, | |
| { | |
| "epoch": 2.26, | |
| "learning_rate": 1.3672286617492097e-05, | |
| "loss": 1.4477, | |
| "step": 1590 | |
| }, | |
| { | |
| "epoch": 2.28, | |
| "learning_rate": 1.3408851422550054e-05, | |
| "loss": 1.3322, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 2.29, | |
| "learning_rate": 1.3145416227608009e-05, | |
| "loss": 1.3451, | |
| "step": 1610 | |
| }, | |
| { | |
| "epoch": 2.3, | |
| "learning_rate": 1.2881981032665966e-05, | |
| "loss": 1.3795, | |
| "step": 1620 | |
| }, | |
| { | |
| "epoch": 2.32, | |
| "learning_rate": 1.2618545837723922e-05, | |
| "loss": 1.2765, | |
| "step": 1630 | |
| }, | |
| { | |
| "epoch": 2.33, | |
| "learning_rate": 1.2355110642781877e-05, | |
| "loss": 1.3775, | |
| "step": 1640 | |
| }, | |
| { | |
| "epoch": 2.35, | |
| "learning_rate": 1.209167544783983e-05, | |
| "loss": 1.2989, | |
| "step": 1650 | |
| }, | |
| { | |
| "epoch": 2.36, | |
| "learning_rate": 1.1828240252897788e-05, | |
| "loss": 1.4062, | |
| "step": 1660 | |
| }, | |
| { | |
| "epoch": 2.37, | |
| "learning_rate": 1.1564805057955744e-05, | |
| "loss": 1.329, | |
| "step": 1670 | |
| }, | |
| { | |
| "epoch": 2.39, | |
| "learning_rate": 1.1301369863013698e-05, | |
| "loss": 1.3265, | |
| "step": 1680 | |
| }, | |
| { | |
| "epoch": 2.4, | |
| "learning_rate": 1.1037934668071655e-05, | |
| "loss": 1.3261, | |
| "step": 1690 | |
| }, | |
| { | |
| "epoch": 2.42, | |
| "learning_rate": 1.0774499473129611e-05, | |
| "loss": 1.3631, | |
| "step": 1700 | |
| }, | |
| { | |
| "epoch": 2.43, | |
| "learning_rate": 1.0511064278187566e-05, | |
| "loss": 1.2648, | |
| "step": 1710 | |
| }, | |
| { | |
| "epoch": 2.45, | |
| "learning_rate": 1.0247629083245521e-05, | |
| "loss": 1.3818, | |
| "step": 1720 | |
| }, | |
| { | |
| "epoch": 2.46, | |
| "learning_rate": 9.984193888303478e-06, | |
| "loss": 1.3628, | |
| "step": 1730 | |
| }, | |
| { | |
| "epoch": 2.47, | |
| "learning_rate": 9.720758693361433e-06, | |
| "loss": 1.3429, | |
| "step": 1740 | |
| }, | |
| { | |
| "epoch": 2.49, | |
| "learning_rate": 9.457323498419388e-06, | |
| "loss": 1.4293, | |
| "step": 1750 | |
| }, | |
| { | |
| "epoch": 2.5, | |
| "learning_rate": 9.193888303477345e-06, | |
| "loss": 1.2913, | |
| "step": 1760 | |
| }, | |
| { | |
| "epoch": 2.52, | |
| "learning_rate": 8.930453108535302e-06, | |
| "loss": 1.3782, | |
| "step": 1770 | |
| }, | |
| { | |
| "epoch": 2.53, | |
| "learning_rate": 8.667017913593255e-06, | |
| "loss": 1.2975, | |
| "step": 1780 | |
| }, | |
| { | |
| "epoch": 2.55, | |
| "learning_rate": 8.403582718651212e-06, | |
| "loss": 1.3039, | |
| "step": 1790 | |
| }, | |
| { | |
| "epoch": 2.56, | |
| "learning_rate": 8.140147523709169e-06, | |
| "loss": 1.291, | |
| "step": 1800 | |
| }, | |
| { | |
| "epoch": 2.57, | |
| "learning_rate": 7.876712328767124e-06, | |
| "loss": 1.3099, | |
| "step": 1810 | |
| }, | |
| { | |
| "epoch": 2.59, | |
| "learning_rate": 7.613277133825079e-06, | |
| "loss": 1.2919, | |
| "step": 1820 | |
| }, | |
| { | |
| "epoch": 2.6, | |
| "learning_rate": 7.349841938883036e-06, | |
| "loss": 1.2977, | |
| "step": 1830 | |
| }, | |
| { | |
| "epoch": 2.62, | |
| "learning_rate": 7.08640674394099e-06, | |
| "loss": 1.3157, | |
| "step": 1840 | |
| }, | |
| { | |
| "epoch": 2.63, | |
| "learning_rate": 6.822971548998947e-06, | |
| "loss": 1.3704, | |
| "step": 1850 | |
| }, | |
| { | |
| "epoch": 2.64, | |
| "learning_rate": 6.559536354056903e-06, | |
| "loss": 1.4022, | |
| "step": 1860 | |
| }, | |
| { | |
| "epoch": 2.66, | |
| "learning_rate": 6.296101159114858e-06, | |
| "loss": 1.3697, | |
| "step": 1870 | |
| }, | |
| { | |
| "epoch": 2.67, | |
| "learning_rate": 6.032665964172814e-06, | |
| "loss": 1.3926, | |
| "step": 1880 | |
| }, | |
| { | |
| "epoch": 2.69, | |
| "learning_rate": 5.76923076923077e-06, | |
| "loss": 1.3484, | |
| "step": 1890 | |
| }, | |
| { | |
| "epoch": 2.7, | |
| "learning_rate": 5.505795574288726e-06, | |
| "loss": 1.3312, | |
| "step": 1900 | |
| }, | |
| { | |
| "epoch": 2.72, | |
| "learning_rate": 5.242360379346681e-06, | |
| "loss": 1.3344, | |
| "step": 1910 | |
| }, | |
| { | |
| "epoch": 2.73, | |
| "learning_rate": 4.978925184404637e-06, | |
| "loss": 1.2542, | |
| "step": 1920 | |
| }, | |
| { | |
| "epoch": 2.74, | |
| "learning_rate": 4.715489989462593e-06, | |
| "loss": 1.31, | |
| "step": 1930 | |
| }, | |
| { | |
| "epoch": 2.76, | |
| "learning_rate": 4.452054794520548e-06, | |
| "loss": 1.2692, | |
| "step": 1940 | |
| }, | |
| { | |
| "epoch": 2.77, | |
| "learning_rate": 4.188619599578504e-06, | |
| "loss": 1.3602, | |
| "step": 1950 | |
| }, | |
| { | |
| "epoch": 2.79, | |
| "learning_rate": 3.92518440463646e-06, | |
| "loss": 1.2637, | |
| "step": 1960 | |
| }, | |
| { | |
| "epoch": 2.8, | |
| "learning_rate": 3.661749209694415e-06, | |
| "loss": 1.2901, | |
| "step": 1970 | |
| }, | |
| { | |
| "epoch": 2.82, | |
| "learning_rate": 3.398314014752371e-06, | |
| "loss": 1.2764, | |
| "step": 1980 | |
| }, | |
| { | |
| "epoch": 2.83, | |
| "learning_rate": 3.1348788198103265e-06, | |
| "loss": 1.3211, | |
| "step": 1990 | |
| }, | |
| { | |
| "epoch": 2.84, | |
| "learning_rate": 2.8714436248682825e-06, | |
| "loss": 1.3353, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 2.86, | |
| "learning_rate": 2.6080084299262384e-06, | |
| "loss": 1.2377, | |
| "step": 2010 | |
| }, | |
| { | |
| "epoch": 2.87, | |
| "learning_rate": 2.3445732349841943e-06, | |
| "loss": 1.3447, | |
| "step": 2020 | |
| }, | |
| { | |
| "epoch": 2.89, | |
| "learning_rate": 2.08113804004215e-06, | |
| "loss": 1.3155, | |
| "step": 2030 | |
| }, | |
| { | |
| "epoch": 2.9, | |
| "learning_rate": 1.8177028451001056e-06, | |
| "loss": 1.4133, | |
| "step": 2040 | |
| }, | |
| { | |
| "epoch": 2.92, | |
| "learning_rate": 1.554267650158061e-06, | |
| "loss": 1.2132, | |
| "step": 2050 | |
| }, | |
| { | |
| "epoch": 2.93, | |
| "learning_rate": 1.290832455216017e-06, | |
| "loss": 1.3686, | |
| "step": 2060 | |
| }, | |
| { | |
| "epoch": 2.94, | |
| "learning_rate": 1.0273972602739725e-06, | |
| "loss": 1.341, | |
| "step": 2070 | |
| }, | |
| { | |
| "epoch": 2.96, | |
| "learning_rate": 7.639620653319284e-07, | |
| "loss": 1.305, | |
| "step": 2080 | |
| }, | |
| { | |
| "epoch": 2.97, | |
| "learning_rate": 5.00526870389884e-07, | |
| "loss": 1.3064, | |
| "step": 2090 | |
| }, | |
| { | |
| "epoch": 2.99, | |
| "learning_rate": 2.3709167544783985e-07, | |
| "loss": 1.3005, | |
| "step": 2100 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.8289, | |
| "eval_loss": 0.6467128992080688, | |
| "eval_runtime": 37.586, | |
| "eval_samples_per_second": 266.057, | |
| "eval_steps_per_second": 8.328, | |
| "step": 2109 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "step": 2109, | |
| "total_flos": 6.74624745043329e+18, | |
| "train_loss": 1.9375283018340992, | |
| "train_runtime": 2464.8383, | |
| "train_samples_per_second": 109.541, | |
| "train_steps_per_second": 0.856 | |
| } | |
| ], | |
| "max_steps": 2109, | |
| "num_train_epochs": 3, | |
| "total_flos": 6.74624745043329e+18, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |