| { | |
| "best_metric": 0.36420056223869324, | |
| "best_model_checkpoint": "vit-base-patch16-224-in21k-bridgedefectVIT\\checkpoint-1785", | |
| "epoch": 8.0, | |
| "eval_steps": 500, | |
| "global_step": 2040, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 6.535947712418301e-07, | |
| "loss": 1.6179, | |
| "step": 10 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.3071895424836602e-06, | |
| "loss": 1.6223, | |
| "step": 20 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.96078431372549e-06, | |
| "loss": 1.6263, | |
| "step": 30 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 2.6143790849673204e-06, | |
| "loss": 1.601, | |
| "step": 40 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 3.2679738562091506e-06, | |
| "loss": 1.6074, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 3.92156862745098e-06, | |
| "loss": 1.5955, | |
| "step": 60 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 4.5751633986928105e-06, | |
| "loss": 1.5698, | |
| "step": 70 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 5.228758169934641e-06, | |
| "loss": 1.5797, | |
| "step": 80 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 5.882352941176471e-06, | |
| "loss": 1.5531, | |
| "step": 90 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 6.535947712418301e-06, | |
| "loss": 1.5434, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 7.1895424836601305e-06, | |
| "loss": 1.5433, | |
| "step": 110 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 7.84313725490196e-06, | |
| "loss": 1.5329, | |
| "step": 120 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 8.496732026143791e-06, | |
| "loss": 1.4995, | |
| "step": 130 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 9.150326797385621e-06, | |
| "loss": 1.4875, | |
| "step": 140 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 9.803921568627451e-06, | |
| "loss": 1.4863, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 1.0457516339869281e-05, | |
| "loss": 1.4209, | |
| "step": 160 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 1.1111111111111112e-05, | |
| "loss": 1.4354, | |
| "step": 170 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 1.1764705882352942e-05, | |
| "loss": 1.3572, | |
| "step": 180 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 1.2418300653594772e-05, | |
| "loss": 1.3714, | |
| "step": 190 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 1.3071895424836602e-05, | |
| "loss": 1.3406, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.3725490196078432e-05, | |
| "loss": 1.2574, | |
| "step": 210 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.4379084967320261e-05, | |
| "loss": 1.2965, | |
| "step": 220 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.5032679738562091e-05, | |
| "loss": 1.2021, | |
| "step": 230 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.568627450980392e-05, | |
| "loss": 1.2072, | |
| "step": 240 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 1.6339869281045753e-05, | |
| "loss": 1.1457, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": { | |
| "accuracy": 0.6941176470588235 | |
| }, | |
| "eval_f1": { | |
| "f1": 0.683792091280419 | |
| }, | |
| "eval_loss": 1.158827304840088, | |
| "eval_precision": { | |
| "precision": 0.7377494408375539 | |
| }, | |
| "eval_recall": { | |
| "recall": 0.6941484490686618 | |
| }, | |
| "eval_runtime": 25.3262, | |
| "eval_samples_per_second": 20.137, | |
| "eval_steps_per_second": 10.069, | |
| "step": 255 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 1.6993464052287582e-05, | |
| "loss": 1.0924, | |
| "step": 260 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "learning_rate": 1.7647058823529414e-05, | |
| "loss": 1.0007, | |
| "step": 270 | |
| }, | |
| { | |
| "epoch": 1.1, | |
| "learning_rate": 1.8300653594771242e-05, | |
| "loss": 1.0054, | |
| "step": 280 | |
| }, | |
| { | |
| "epoch": 1.14, | |
| "learning_rate": 1.895424836601307e-05, | |
| "loss": 0.9365, | |
| "step": 290 | |
| }, | |
| { | |
| "epoch": 1.18, | |
| "learning_rate": 1.9607843137254903e-05, | |
| "loss": 0.8932, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 1.22, | |
| "learning_rate": 2.0261437908496734e-05, | |
| "loss": 0.916, | |
| "step": 310 | |
| }, | |
| { | |
| "epoch": 1.25, | |
| "learning_rate": 2.0915032679738563e-05, | |
| "loss": 0.8099, | |
| "step": 320 | |
| }, | |
| { | |
| "epoch": 1.29, | |
| "learning_rate": 2.1568627450980395e-05, | |
| "loss": 0.8211, | |
| "step": 330 | |
| }, | |
| { | |
| "epoch": 1.33, | |
| "learning_rate": 2.2222222222222223e-05, | |
| "loss": 0.6861, | |
| "step": 340 | |
| }, | |
| { | |
| "epoch": 1.37, | |
| "learning_rate": 2.2875816993464052e-05, | |
| "loss": 0.7379, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 1.41, | |
| "learning_rate": 2.3529411764705884e-05, | |
| "loss": 0.7349, | |
| "step": 360 | |
| }, | |
| { | |
| "epoch": 1.45, | |
| "learning_rate": 2.4183006535947712e-05, | |
| "loss": 0.6907, | |
| "step": 370 | |
| }, | |
| { | |
| "epoch": 1.49, | |
| "learning_rate": 2.4836601307189544e-05, | |
| "loss": 0.7376, | |
| "step": 380 | |
| }, | |
| { | |
| "epoch": 1.53, | |
| "learning_rate": 2.5490196078431373e-05, | |
| "loss": 0.6827, | |
| "step": 390 | |
| }, | |
| { | |
| "epoch": 1.57, | |
| "learning_rate": 2.6143790849673204e-05, | |
| "loss": 0.6969, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 1.61, | |
| "learning_rate": 2.6797385620915033e-05, | |
| "loss": 0.6247, | |
| "step": 410 | |
| }, | |
| { | |
| "epoch": 1.65, | |
| "learning_rate": 2.7450980392156865e-05, | |
| "loss": 0.5244, | |
| "step": 420 | |
| }, | |
| { | |
| "epoch": 1.69, | |
| "learning_rate": 2.8104575163398693e-05, | |
| "loss": 0.7068, | |
| "step": 430 | |
| }, | |
| { | |
| "epoch": 1.73, | |
| "learning_rate": 2.8758169934640522e-05, | |
| "loss": 0.5623, | |
| "step": 440 | |
| }, | |
| { | |
| "epoch": 1.76, | |
| "learning_rate": 2.9411764705882354e-05, | |
| "loss": 0.5931, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 1.8, | |
| "learning_rate": 3.0065359477124182e-05, | |
| "loss": 0.6641, | |
| "step": 460 | |
| }, | |
| { | |
| "epoch": 1.84, | |
| "learning_rate": 3.0718954248366014e-05, | |
| "loss": 0.5438, | |
| "step": 470 | |
| }, | |
| { | |
| "epoch": 1.88, | |
| "learning_rate": 3.137254901960784e-05, | |
| "loss": 0.5134, | |
| "step": 480 | |
| }, | |
| { | |
| "epoch": 1.92, | |
| "learning_rate": 3.202614379084967e-05, | |
| "loss": 0.4682, | |
| "step": 490 | |
| }, | |
| { | |
| "epoch": 1.96, | |
| "learning_rate": 3.2679738562091506e-05, | |
| "loss": 0.4762, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "learning_rate": 3.3333333333333335e-05, | |
| "loss": 0.4558, | |
| "step": 510 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": { | |
| "accuracy": 0.8470588235294118 | |
| }, | |
| "eval_f1": { | |
| "f1": 0.8510200857474889 | |
| }, | |
| "eval_loss": 0.498457670211792, | |
| "eval_precision": { | |
| "precision": 0.8766865984788879 | |
| }, | |
| "eval_recall": { | |
| "recall": 0.8454231938274491 | |
| }, | |
| "eval_runtime": 10.0517, | |
| "eval_samples_per_second": 50.738, | |
| "eval_steps_per_second": 25.369, | |
| "step": 510 | |
| }, | |
| { | |
| "epoch": 2.04, | |
| "learning_rate": 3.3986928104575163e-05, | |
| "loss": 0.4716, | |
| "step": 520 | |
| }, | |
| { | |
| "epoch": 2.08, | |
| "learning_rate": 3.464052287581699e-05, | |
| "loss": 0.5375, | |
| "step": 530 | |
| }, | |
| { | |
| "epoch": 2.12, | |
| "learning_rate": 3.529411764705883e-05, | |
| "loss": 0.4725, | |
| "step": 540 | |
| }, | |
| { | |
| "epoch": 2.16, | |
| "learning_rate": 3.5947712418300656e-05, | |
| "loss": 0.4804, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 2.2, | |
| "learning_rate": 3.6601307189542484e-05, | |
| "loss": 0.3504, | |
| "step": 560 | |
| }, | |
| { | |
| "epoch": 2.24, | |
| "learning_rate": 3.725490196078432e-05, | |
| "loss": 0.4049, | |
| "step": 570 | |
| }, | |
| { | |
| "epoch": 2.27, | |
| "learning_rate": 3.790849673202614e-05, | |
| "loss": 0.4309, | |
| "step": 580 | |
| }, | |
| { | |
| "epoch": 2.31, | |
| "learning_rate": 3.8562091503267977e-05, | |
| "loss": 0.4938, | |
| "step": 590 | |
| }, | |
| { | |
| "epoch": 2.35, | |
| "learning_rate": 3.9215686274509805e-05, | |
| "loss": 0.5427, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 2.39, | |
| "learning_rate": 3.986928104575164e-05, | |
| "loss": 0.4719, | |
| "step": 610 | |
| }, | |
| { | |
| "epoch": 2.43, | |
| "learning_rate": 4.052287581699347e-05, | |
| "loss": 0.4627, | |
| "step": 620 | |
| }, | |
| { | |
| "epoch": 2.47, | |
| "learning_rate": 4.11764705882353e-05, | |
| "loss": 0.5277, | |
| "step": 630 | |
| }, | |
| { | |
| "epoch": 2.51, | |
| "learning_rate": 4.1830065359477126e-05, | |
| "loss": 0.3109, | |
| "step": 640 | |
| }, | |
| { | |
| "epoch": 2.55, | |
| "learning_rate": 4.2483660130718954e-05, | |
| "loss": 0.3799, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 2.59, | |
| "learning_rate": 4.313725490196079e-05, | |
| "loss": 0.3515, | |
| "step": 660 | |
| }, | |
| { | |
| "epoch": 2.63, | |
| "learning_rate": 4.379084967320262e-05, | |
| "loss": 0.3722, | |
| "step": 670 | |
| }, | |
| { | |
| "epoch": 2.67, | |
| "learning_rate": 4.4444444444444447e-05, | |
| "loss": 0.5337, | |
| "step": 680 | |
| }, | |
| { | |
| "epoch": 2.71, | |
| "learning_rate": 4.5098039215686275e-05, | |
| "loss": 0.3263, | |
| "step": 690 | |
| }, | |
| { | |
| "epoch": 2.75, | |
| "learning_rate": 4.5751633986928104e-05, | |
| "loss": 0.3703, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 2.78, | |
| "learning_rate": 4.640522875816994e-05, | |
| "loss": 0.4498, | |
| "step": 710 | |
| }, | |
| { | |
| "epoch": 2.82, | |
| "learning_rate": 4.705882352941177e-05, | |
| "loss": 0.4109, | |
| "step": 720 | |
| }, | |
| { | |
| "epoch": 2.86, | |
| "learning_rate": 4.77124183006536e-05, | |
| "loss": 0.2953, | |
| "step": 730 | |
| }, | |
| { | |
| "epoch": 2.9, | |
| "learning_rate": 4.8366013071895424e-05, | |
| "loss": 0.4033, | |
| "step": 740 | |
| }, | |
| { | |
| "epoch": 2.94, | |
| "learning_rate": 4.901960784313725e-05, | |
| "loss": 0.4186, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 2.98, | |
| "learning_rate": 4.967320261437909e-05, | |
| "loss": 0.3368, | |
| "step": 760 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": { | |
| "accuracy": 0.8705882352941177 | |
| }, | |
| "eval_f1": { | |
| "f1": 0.8713991689505157 | |
| }, | |
| "eval_loss": 0.44290485978126526, | |
| "eval_precision": { | |
| "precision": 0.8884985994397759 | |
| }, | |
| "eval_recall": { | |
| "recall": 0.8711626139817629 | |
| }, | |
| "eval_runtime": 9.9506, | |
| "eval_samples_per_second": 51.253, | |
| "eval_steps_per_second": 25.626, | |
| "step": 765 | |
| }, | |
| { | |
| "epoch": 3.02, | |
| "learning_rate": 4.9963689179375456e-05, | |
| "loss": 0.3347, | |
| "step": 770 | |
| }, | |
| { | |
| "epoch": 3.06, | |
| "learning_rate": 4.9891067538126364e-05, | |
| "loss": 0.3339, | |
| "step": 780 | |
| }, | |
| { | |
| "epoch": 3.1, | |
| "learning_rate": 4.981844589687727e-05, | |
| "loss": 0.5573, | |
| "step": 790 | |
| }, | |
| { | |
| "epoch": 3.14, | |
| "learning_rate": 4.974582425562818e-05, | |
| "loss": 0.5463, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 3.18, | |
| "learning_rate": 4.967320261437909e-05, | |
| "loss": 0.3453, | |
| "step": 810 | |
| }, | |
| { | |
| "epoch": 3.22, | |
| "learning_rate": 4.9600580973129996e-05, | |
| "loss": 0.3814, | |
| "step": 820 | |
| }, | |
| { | |
| "epoch": 3.25, | |
| "learning_rate": 4.9527959331880904e-05, | |
| "loss": 0.4434, | |
| "step": 830 | |
| }, | |
| { | |
| "epoch": 3.29, | |
| "learning_rate": 4.945533769063181e-05, | |
| "loss": 0.3675, | |
| "step": 840 | |
| }, | |
| { | |
| "epoch": 3.33, | |
| "learning_rate": 4.938271604938271e-05, | |
| "loss": 0.5386, | |
| "step": 850 | |
| }, | |
| { | |
| "epoch": 3.37, | |
| "learning_rate": 4.931009440813363e-05, | |
| "loss": 0.2692, | |
| "step": 860 | |
| }, | |
| { | |
| "epoch": 3.41, | |
| "learning_rate": 4.9237472766884536e-05, | |
| "loss": 0.4815, | |
| "step": 870 | |
| }, | |
| { | |
| "epoch": 3.45, | |
| "learning_rate": 4.9164851125635444e-05, | |
| "loss": 0.4178, | |
| "step": 880 | |
| }, | |
| { | |
| "epoch": 3.49, | |
| "learning_rate": 4.909222948438635e-05, | |
| "loss": 0.3618, | |
| "step": 890 | |
| }, | |
| { | |
| "epoch": 3.53, | |
| "learning_rate": 4.901960784313725e-05, | |
| "loss": 0.6841, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 3.57, | |
| "learning_rate": 4.894698620188817e-05, | |
| "loss": 0.471, | |
| "step": 910 | |
| }, | |
| { | |
| "epoch": 3.61, | |
| "learning_rate": 4.8874364560639076e-05, | |
| "loss": 0.313, | |
| "step": 920 | |
| }, | |
| { | |
| "epoch": 3.65, | |
| "learning_rate": 4.8801742919389983e-05, | |
| "loss": 0.2437, | |
| "step": 930 | |
| }, | |
| { | |
| "epoch": 3.69, | |
| "learning_rate": 4.8729121278140885e-05, | |
| "loss": 0.4748, | |
| "step": 940 | |
| }, | |
| { | |
| "epoch": 3.73, | |
| "learning_rate": 4.865649963689179e-05, | |
| "loss": 0.5519, | |
| "step": 950 | |
| }, | |
| { | |
| "epoch": 3.76, | |
| "learning_rate": 4.85838779956427e-05, | |
| "loss": 0.3051, | |
| "step": 960 | |
| }, | |
| { | |
| "epoch": 3.8, | |
| "learning_rate": 4.8511256354393615e-05, | |
| "loss": 0.3847, | |
| "step": 970 | |
| }, | |
| { | |
| "epoch": 3.84, | |
| "learning_rate": 4.843863471314452e-05, | |
| "loss": 0.2454, | |
| "step": 980 | |
| }, | |
| { | |
| "epoch": 3.88, | |
| "learning_rate": 4.8366013071895424e-05, | |
| "loss": 0.379, | |
| "step": 990 | |
| }, | |
| { | |
| "epoch": 3.92, | |
| "learning_rate": 4.829339143064633e-05, | |
| "loss": 0.4485, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 3.96, | |
| "learning_rate": 4.822076978939724e-05, | |
| "loss": 0.2484, | |
| "step": 1010 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "learning_rate": 4.814814814814815e-05, | |
| "loss": 0.4402, | |
| "step": 1020 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_accuracy": { | |
| "accuracy": 0.8784313725490196 | |
| }, | |
| "eval_f1": { | |
| "f1": 0.880471094803992 | |
| }, | |
| "eval_loss": 0.3946979343891144, | |
| "eval_precision": { | |
| "precision": 0.8952890487430221 | |
| }, | |
| "eval_recall": { | |
| "recall": 0.8776488582339645 | |
| }, | |
| "eval_runtime": 10.1771, | |
| "eval_samples_per_second": 50.113, | |
| "eval_steps_per_second": 25.056, | |
| "step": 1020 | |
| }, | |
| { | |
| "epoch": 4.04, | |
| "learning_rate": 4.807552650689906e-05, | |
| "loss": 0.3494, | |
| "step": 1030 | |
| }, | |
| { | |
| "epoch": 4.08, | |
| "learning_rate": 4.8002904865649964e-05, | |
| "loss": 0.3327, | |
| "step": 1040 | |
| }, | |
| { | |
| "epoch": 4.12, | |
| "learning_rate": 4.793028322440087e-05, | |
| "loss": 0.2898, | |
| "step": 1050 | |
| }, | |
| { | |
| "epoch": 4.16, | |
| "learning_rate": 4.785766158315178e-05, | |
| "loss": 0.3965, | |
| "step": 1060 | |
| }, | |
| { | |
| "epoch": 4.2, | |
| "learning_rate": 4.778503994190269e-05, | |
| "loss": 0.3607, | |
| "step": 1070 | |
| }, | |
| { | |
| "epoch": 4.24, | |
| "learning_rate": 4.77124183006536e-05, | |
| "loss": 0.3714, | |
| "step": 1080 | |
| }, | |
| { | |
| "epoch": 4.27, | |
| "learning_rate": 4.7639796659404504e-05, | |
| "loss": 0.1796, | |
| "step": 1090 | |
| }, | |
| { | |
| "epoch": 4.31, | |
| "learning_rate": 4.756717501815541e-05, | |
| "loss": 0.1874, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 4.35, | |
| "learning_rate": 4.749455337690632e-05, | |
| "loss": 0.3462, | |
| "step": 1110 | |
| }, | |
| { | |
| "epoch": 4.39, | |
| "learning_rate": 4.742193173565723e-05, | |
| "loss": 0.2082, | |
| "step": 1120 | |
| }, | |
| { | |
| "epoch": 4.43, | |
| "learning_rate": 4.7349310094408136e-05, | |
| "loss": 0.4531, | |
| "step": 1130 | |
| }, | |
| { | |
| "epoch": 4.47, | |
| "learning_rate": 4.7276688453159044e-05, | |
| "loss": 0.3951, | |
| "step": 1140 | |
| }, | |
| { | |
| "epoch": 4.51, | |
| "learning_rate": 4.720406681190995e-05, | |
| "loss": 0.2337, | |
| "step": 1150 | |
| }, | |
| { | |
| "epoch": 4.55, | |
| "learning_rate": 4.713144517066086e-05, | |
| "loss": 0.2118, | |
| "step": 1160 | |
| }, | |
| { | |
| "epoch": 4.59, | |
| "learning_rate": 4.705882352941177e-05, | |
| "loss": 0.373, | |
| "step": 1170 | |
| }, | |
| { | |
| "epoch": 4.63, | |
| "learning_rate": 4.6986201888162675e-05, | |
| "loss": 0.1484, | |
| "step": 1180 | |
| }, | |
| { | |
| "epoch": 4.67, | |
| "learning_rate": 4.691358024691358e-05, | |
| "loss": 0.2568, | |
| "step": 1190 | |
| }, | |
| { | |
| "epoch": 4.71, | |
| "learning_rate": 4.684095860566449e-05, | |
| "loss": 0.3067, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 4.75, | |
| "learning_rate": 4.67683369644154e-05, | |
| "loss": 0.2007, | |
| "step": 1210 | |
| }, | |
| { | |
| "epoch": 4.78, | |
| "learning_rate": 4.669571532316631e-05, | |
| "loss": 0.2493, | |
| "step": 1220 | |
| }, | |
| { | |
| "epoch": 4.82, | |
| "learning_rate": 4.6623093681917215e-05, | |
| "loss": 0.2717, | |
| "step": 1230 | |
| }, | |
| { | |
| "epoch": 4.86, | |
| "learning_rate": 4.6550472040668116e-05, | |
| "loss": 0.4134, | |
| "step": 1240 | |
| }, | |
| { | |
| "epoch": 4.9, | |
| "learning_rate": 4.647785039941903e-05, | |
| "loss": 0.2417, | |
| "step": 1250 | |
| }, | |
| { | |
| "epoch": 4.94, | |
| "learning_rate": 4.640522875816994e-05, | |
| "loss": 0.2573, | |
| "step": 1260 | |
| }, | |
| { | |
| "epoch": 4.98, | |
| "learning_rate": 4.633260711692085e-05, | |
| "loss": 0.3937, | |
| "step": 1270 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_accuracy": { | |
| "accuracy": 0.8725490196078431 | |
| }, | |
| "eval_f1": { | |
| "f1": 0.8703780158419334 | |
| }, | |
| "eval_loss": 0.4377364218235016, | |
| "eval_precision": { | |
| "precision": 0.8900760323128502 | |
| }, | |
| "eval_recall": { | |
| "recall": 0.8691138648585458 | |
| }, | |
| "eval_runtime": 10.1762, | |
| "eval_samples_per_second": 50.117, | |
| "eval_steps_per_second": 25.058, | |
| "step": 1275 | |
| }, | |
| { | |
| "epoch": 5.02, | |
| "learning_rate": 4.6259985475671755e-05, | |
| "loss": 0.165, | |
| "step": 1280 | |
| }, | |
| { | |
| "epoch": 5.06, | |
| "learning_rate": 4.6187363834422656e-05, | |
| "loss": 0.1452, | |
| "step": 1290 | |
| }, | |
| { | |
| "epoch": 5.1, | |
| "learning_rate": 4.6114742193173564e-05, | |
| "loss": 0.1548, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 5.14, | |
| "learning_rate": 4.604212055192448e-05, | |
| "loss": 0.4443, | |
| "step": 1310 | |
| }, | |
| { | |
| "epoch": 5.18, | |
| "learning_rate": 4.5969498910675387e-05, | |
| "loss": 0.4322, | |
| "step": 1320 | |
| }, | |
| { | |
| "epoch": 5.22, | |
| "learning_rate": 4.5896877269426294e-05, | |
| "loss": 0.2355, | |
| "step": 1330 | |
| }, | |
| { | |
| "epoch": 5.25, | |
| "learning_rate": 4.5824255628177196e-05, | |
| "loss": 0.1949, | |
| "step": 1340 | |
| }, | |
| { | |
| "epoch": 5.29, | |
| "learning_rate": 4.5751633986928104e-05, | |
| "loss": 0.1959, | |
| "step": 1350 | |
| }, | |
| { | |
| "epoch": 5.33, | |
| "learning_rate": 4.567901234567901e-05, | |
| "loss": 0.3841, | |
| "step": 1360 | |
| }, | |
| { | |
| "epoch": 5.37, | |
| "learning_rate": 4.5606390704429926e-05, | |
| "loss": 0.1401, | |
| "step": 1370 | |
| }, | |
| { | |
| "epoch": 5.41, | |
| "learning_rate": 4.5533769063180834e-05, | |
| "loss": 0.1148, | |
| "step": 1380 | |
| }, | |
| { | |
| "epoch": 5.45, | |
| "learning_rate": 4.5461147421931735e-05, | |
| "loss": 0.234, | |
| "step": 1390 | |
| }, | |
| { | |
| "epoch": 5.49, | |
| "learning_rate": 4.538852578068264e-05, | |
| "loss": 0.3899, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 5.53, | |
| "learning_rate": 4.531590413943355e-05, | |
| "loss": 0.4873, | |
| "step": 1410 | |
| }, | |
| { | |
| "epoch": 5.57, | |
| "learning_rate": 4.5243282498184466e-05, | |
| "loss": 0.2483, | |
| "step": 1420 | |
| }, | |
| { | |
| "epoch": 5.61, | |
| "learning_rate": 4.5170660856935374e-05, | |
| "loss": 0.2654, | |
| "step": 1430 | |
| }, | |
| { | |
| "epoch": 5.65, | |
| "learning_rate": 4.5098039215686275e-05, | |
| "loss": 0.2035, | |
| "step": 1440 | |
| }, | |
| { | |
| "epoch": 5.69, | |
| "learning_rate": 4.502541757443718e-05, | |
| "loss": 0.3744, | |
| "step": 1450 | |
| }, | |
| { | |
| "epoch": 5.73, | |
| "learning_rate": 4.495279593318809e-05, | |
| "loss": 0.2139, | |
| "step": 1460 | |
| }, | |
| { | |
| "epoch": 5.76, | |
| "learning_rate": 4.4880174291939e-05, | |
| "loss": 0.2212, | |
| "step": 1470 | |
| }, | |
| { | |
| "epoch": 5.8, | |
| "learning_rate": 4.480755265068991e-05, | |
| "loss": 0.376, | |
| "step": 1480 | |
| }, | |
| { | |
| "epoch": 5.84, | |
| "learning_rate": 4.4734931009440815e-05, | |
| "loss": 0.2399, | |
| "step": 1490 | |
| }, | |
| { | |
| "epoch": 5.88, | |
| "learning_rate": 4.466230936819172e-05, | |
| "loss": 0.3534, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 5.92, | |
| "learning_rate": 4.458968772694263e-05, | |
| "loss": 0.3667, | |
| "step": 1510 | |
| }, | |
| { | |
| "epoch": 5.96, | |
| "learning_rate": 4.451706608569354e-05, | |
| "loss": 0.496, | |
| "step": 1520 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "learning_rate": 4.4444444444444447e-05, | |
| "loss": 0.4042, | |
| "step": 1530 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "eval_accuracy": { | |
| "accuracy": 0.8862745098039215 | |
| }, | |
| "eval_f1": { | |
| "f1": 0.8859196308355288 | |
| }, | |
| "eval_loss": 0.39891302585601807, | |
| "eval_precision": { | |
| "precision": 0.8972093144987638 | |
| }, | |
| "eval_recall": { | |
| "recall": 0.8834955381497934 | |
| }, | |
| "eval_runtime": 10.1174, | |
| "eval_samples_per_second": 50.408, | |
| "eval_steps_per_second": 25.204, | |
| "step": 1530 | |
| }, | |
| { | |
| "epoch": 6.04, | |
| "learning_rate": 4.4371822803195355e-05, | |
| "loss": 0.2637, | |
| "step": 1540 | |
| }, | |
| { | |
| "epoch": 6.08, | |
| "learning_rate": 4.429920116194626e-05, | |
| "loss": 0.2821, | |
| "step": 1550 | |
| }, | |
| { | |
| "epoch": 6.12, | |
| "learning_rate": 4.422657952069717e-05, | |
| "loss": 0.1201, | |
| "step": 1560 | |
| }, | |
| { | |
| "epoch": 6.16, | |
| "learning_rate": 4.415395787944808e-05, | |
| "loss": 0.262, | |
| "step": 1570 | |
| }, | |
| { | |
| "epoch": 6.2, | |
| "learning_rate": 4.4081336238198986e-05, | |
| "loss": 0.2064, | |
| "step": 1580 | |
| }, | |
| { | |
| "epoch": 6.24, | |
| "learning_rate": 4.400871459694989e-05, | |
| "loss": 0.1815, | |
| "step": 1590 | |
| }, | |
| { | |
| "epoch": 6.27, | |
| "learning_rate": 4.39360929557008e-05, | |
| "loss": 0.1959, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 6.31, | |
| "learning_rate": 4.386347131445171e-05, | |
| "loss": 0.1959, | |
| "step": 1610 | |
| }, | |
| { | |
| "epoch": 6.35, | |
| "learning_rate": 4.379084967320262e-05, | |
| "loss": 0.2398, | |
| "step": 1620 | |
| }, | |
| { | |
| "epoch": 6.39, | |
| "learning_rate": 4.3718228031953526e-05, | |
| "loss": 0.2038, | |
| "step": 1630 | |
| }, | |
| { | |
| "epoch": 6.43, | |
| "learning_rate": 4.364560639070443e-05, | |
| "loss": 0.2004, | |
| "step": 1640 | |
| }, | |
| { | |
| "epoch": 6.47, | |
| "learning_rate": 4.357298474945534e-05, | |
| "loss": 0.0377, | |
| "step": 1650 | |
| }, | |
| { | |
| "epoch": 6.51, | |
| "learning_rate": 4.350036310820625e-05, | |
| "loss": 0.2022, | |
| "step": 1660 | |
| }, | |
| { | |
| "epoch": 6.55, | |
| "learning_rate": 4.342774146695716e-05, | |
| "loss": 0.4403, | |
| "step": 1670 | |
| }, | |
| { | |
| "epoch": 6.59, | |
| "learning_rate": 4.3355119825708066e-05, | |
| "loss": 0.2577, | |
| "step": 1680 | |
| }, | |
| { | |
| "epoch": 6.63, | |
| "learning_rate": 4.328249818445897e-05, | |
| "loss": 0.3312, | |
| "step": 1690 | |
| }, | |
| { | |
| "epoch": 6.67, | |
| "learning_rate": 4.3209876543209875e-05, | |
| "loss": 0.307, | |
| "step": 1700 | |
| }, | |
| { | |
| "epoch": 6.71, | |
| "learning_rate": 4.313725490196079e-05, | |
| "loss": 0.0961, | |
| "step": 1710 | |
| }, | |
| { | |
| "epoch": 6.75, | |
| "learning_rate": 4.30646332607117e-05, | |
| "loss": 0.3266, | |
| "step": 1720 | |
| }, | |
| { | |
| "epoch": 6.78, | |
| "learning_rate": 4.2992011619462605e-05, | |
| "loss": 0.2601, | |
| "step": 1730 | |
| }, | |
| { | |
| "epoch": 6.82, | |
| "learning_rate": 4.291938997821351e-05, | |
| "loss": 0.2382, | |
| "step": 1740 | |
| }, | |
| { | |
| "epoch": 6.86, | |
| "learning_rate": 4.2846768336964415e-05, | |
| "loss": 0.1469, | |
| "step": 1750 | |
| }, | |
| { | |
| "epoch": 6.9, | |
| "learning_rate": 4.277414669571532e-05, | |
| "loss": 0.2158, | |
| "step": 1760 | |
| }, | |
| { | |
| "epoch": 6.94, | |
| "learning_rate": 4.270152505446624e-05, | |
| "loss": 0.4392, | |
| "step": 1770 | |
| }, | |
| { | |
| "epoch": 6.98, | |
| "learning_rate": 4.262890341321714e-05, | |
| "loss": 0.4321, | |
| "step": 1780 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "eval_accuracy": { | |
| "accuracy": 0.8901960784313725 | |
| }, | |
| "eval_f1": { | |
| "f1": 0.8896787277785286 | |
| }, | |
| "eval_loss": 0.36420056223869324, | |
| "eval_precision": { | |
| "precision": 0.9002265766283616 | |
| }, | |
| "eval_recall": { | |
| "recall": 0.8878911425453978 | |
| }, | |
| "eval_runtime": 10.2347, | |
| "eval_samples_per_second": 49.83, | |
| "eval_steps_per_second": 24.915, | |
| "step": 1785 | |
| }, | |
| { | |
| "epoch": 7.02, | |
| "learning_rate": 4.2556281771968046e-05, | |
| "loss": 0.5162, | |
| "step": 1790 | |
| }, | |
| { | |
| "epoch": 7.06, | |
| "learning_rate": 4.2483660130718954e-05, | |
| "loss": 0.2491, | |
| "step": 1800 | |
| }, | |
| { | |
| "epoch": 7.1, | |
| "learning_rate": 4.241103848946986e-05, | |
| "loss": 0.2267, | |
| "step": 1810 | |
| }, | |
| { | |
| "epoch": 7.14, | |
| "learning_rate": 4.233841684822078e-05, | |
| "loss": 0.2395, | |
| "step": 1820 | |
| }, | |
| { | |
| "epoch": 7.18, | |
| "learning_rate": 4.226579520697168e-05, | |
| "loss": 0.3007, | |
| "step": 1830 | |
| }, | |
| { | |
| "epoch": 7.22, | |
| "learning_rate": 4.2193173565722586e-05, | |
| "loss": 0.1316, | |
| "step": 1840 | |
| }, | |
| { | |
| "epoch": 7.25, | |
| "learning_rate": 4.2120551924473494e-05, | |
| "loss": 0.1592, | |
| "step": 1850 | |
| }, | |
| { | |
| "epoch": 7.29, | |
| "learning_rate": 4.20479302832244e-05, | |
| "loss": 0.3467, | |
| "step": 1860 | |
| }, | |
| { | |
| "epoch": 7.33, | |
| "learning_rate": 4.197530864197531e-05, | |
| "loss": 0.4195, | |
| "step": 1870 | |
| }, | |
| { | |
| "epoch": 7.37, | |
| "learning_rate": 4.190268700072622e-05, | |
| "loss": 0.3561, | |
| "step": 1880 | |
| }, | |
| { | |
| "epoch": 7.41, | |
| "learning_rate": 4.1830065359477126e-05, | |
| "loss": 0.1572, | |
| "step": 1890 | |
| }, | |
| { | |
| "epoch": 7.45, | |
| "learning_rate": 4.1757443718228034e-05, | |
| "loss": 0.2104, | |
| "step": 1900 | |
| }, | |
| { | |
| "epoch": 7.49, | |
| "learning_rate": 4.168482207697894e-05, | |
| "loss": 0.3536, | |
| "step": 1910 | |
| }, | |
| { | |
| "epoch": 7.53, | |
| "learning_rate": 4.161220043572985e-05, | |
| "loss": 0.345, | |
| "step": 1920 | |
| }, | |
| { | |
| "epoch": 7.57, | |
| "learning_rate": 4.153957879448076e-05, | |
| "loss": 0.1083, | |
| "step": 1930 | |
| }, | |
| { | |
| "epoch": 7.61, | |
| "learning_rate": 4.1466957153231666e-05, | |
| "loss": 0.1895, | |
| "step": 1940 | |
| }, | |
| { | |
| "epoch": 7.65, | |
| "learning_rate": 4.1394335511982573e-05, | |
| "loss": 0.3232, | |
| "step": 1950 | |
| }, | |
| { | |
| "epoch": 7.69, | |
| "learning_rate": 4.132171387073348e-05, | |
| "loss": 0.2939, | |
| "step": 1960 | |
| }, | |
| { | |
| "epoch": 7.73, | |
| "learning_rate": 4.124909222948439e-05, | |
| "loss": 0.0918, | |
| "step": 1970 | |
| }, | |
| { | |
| "epoch": 7.76, | |
| "learning_rate": 4.11764705882353e-05, | |
| "loss": 0.1722, | |
| "step": 1980 | |
| }, | |
| { | |
| "epoch": 7.8, | |
| "learning_rate": 4.1103848946986205e-05, | |
| "loss": 0.3229, | |
| "step": 1990 | |
| }, | |
| { | |
| "epoch": 7.84, | |
| "learning_rate": 4.103122730573711e-05, | |
| "loss": 0.1501, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 7.88, | |
| "learning_rate": 4.095860566448802e-05, | |
| "loss": 0.1468, | |
| "step": 2010 | |
| }, | |
| { | |
| "epoch": 7.92, | |
| "learning_rate": 4.088598402323893e-05, | |
| "loss": 0.1665, | |
| "step": 2020 | |
| }, | |
| { | |
| "epoch": 7.96, | |
| "learning_rate": 4.081336238198984e-05, | |
| "loss": 0.1324, | |
| "step": 2030 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "learning_rate": 4.074074074074074e-05, | |
| "loss": 0.2618, | |
| "step": 2040 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "eval_accuracy": { | |
| "accuracy": 0.8960784313725491 | |
| }, | |
| "eval_f1": { | |
| "f1": 0.8950352545633242 | |
| }, | |
| "eval_loss": 0.40923169255256653, | |
| "eval_precision": { | |
| "precision": 0.9033601037449228 | |
| }, | |
| "eval_recall": { | |
| "recall": 0.8953584093211753 | |
| }, | |
| "eval_runtime": 10.3235, | |
| "eval_samples_per_second": 49.402, | |
| "eval_steps_per_second": 24.701, | |
| "step": 2040 | |
| } | |
| ], | |
| "logging_steps": 10, | |
| "max_steps": 7650, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 30, | |
| "save_steps": 500, | |
| "total_flos": 1.2647032750021018e+18, | |
| "train_batch_size": 2, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |