| { | |
| "best_metric": 0.8225806451612904, | |
| "best_model_checkpoint": "swinv2-tiny-patch4-window8-256-OT\\checkpoint-176", | |
| "epoch": 36.36363636363637, | |
| "eval_steps": 500, | |
| "global_step": 200, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.91, | |
| "eval_accuracy": 0.08064516129032258, | |
| "eval_loss": 8.843916893005371, | |
| "eval_runtime": 2.8151, | |
| "eval_samples_per_second": 22.024, | |
| "eval_steps_per_second": 1.421, | |
| "step": 5 | |
| }, | |
| { | |
| "epoch": 1.82, | |
| "learning_rate": 7.5e-05, | |
| "loss": 8.7922, | |
| "step": 10 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.08064516129032258, | |
| "eval_loss": 8.001582145690918, | |
| "eval_runtime": 2.1045, | |
| "eval_samples_per_second": 29.461, | |
| "eval_steps_per_second": 1.901, | |
| "step": 11 | |
| }, | |
| { | |
| "epoch": 2.91, | |
| "eval_accuracy": 0.08064516129032258, | |
| "eval_loss": 6.000851154327393, | |
| "eval_runtime": 2.0735, | |
| "eval_samples_per_second": 29.901, | |
| "eval_steps_per_second": 1.929, | |
| "step": 16 | |
| }, | |
| { | |
| "epoch": 3.64, | |
| "learning_rate": 0.00015, | |
| "loss": 6.5264, | |
| "step": 20 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_accuracy": 0.08064516129032258, | |
| "eval_loss": 2.7431113719940186, | |
| "eval_runtime": 2.1035, | |
| "eval_samples_per_second": 29.475, | |
| "eval_steps_per_second": 1.902, | |
| "step": 22 | |
| }, | |
| { | |
| "epoch": 4.91, | |
| "eval_accuracy": 0.45161290322580644, | |
| "eval_loss": 1.3018240928649902, | |
| "eval_runtime": 2.0906, | |
| "eval_samples_per_second": 29.656, | |
| "eval_steps_per_second": 1.913, | |
| "step": 27 | |
| }, | |
| { | |
| "epoch": 5.45, | |
| "learning_rate": 0.00014166666666666665, | |
| "loss": 2.16, | |
| "step": 30 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "eval_accuracy": 0.45161290322580644, | |
| "eval_loss": 1.2696114778518677, | |
| "eval_runtime": 2.0785, | |
| "eval_samples_per_second": 29.829, | |
| "eval_steps_per_second": 1.924, | |
| "step": 33 | |
| }, | |
| { | |
| "epoch": 6.91, | |
| "eval_accuracy": 0.45161290322580644, | |
| "eval_loss": 1.2057440280914307, | |
| "eval_runtime": 2.5671, | |
| "eval_samples_per_second": 24.152, | |
| "eval_steps_per_second": 1.558, | |
| "step": 38 | |
| }, | |
| { | |
| "epoch": 7.27, | |
| "learning_rate": 0.0001333333333333333, | |
| "loss": 1.2876, | |
| "step": 40 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "eval_accuracy": 0.45161290322580644, | |
| "eval_loss": 1.2157402038574219, | |
| "eval_runtime": 2.1295, | |
| "eval_samples_per_second": 29.115, | |
| "eval_steps_per_second": 1.878, | |
| "step": 44 | |
| }, | |
| { | |
| "epoch": 8.91, | |
| "eval_accuracy": 0.45161290322580644, | |
| "eval_loss": 1.245875597000122, | |
| "eval_runtime": 2.141, | |
| "eval_samples_per_second": 28.958, | |
| "eval_steps_per_second": 1.868, | |
| "step": 49 | |
| }, | |
| { | |
| "epoch": 9.09, | |
| "learning_rate": 0.000125, | |
| "loss": 1.2456, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 10.0, | |
| "eval_accuracy": 0.45161290322580644, | |
| "eval_loss": 1.210959792137146, | |
| "eval_runtime": 2.181, | |
| "eval_samples_per_second": 28.427, | |
| "eval_steps_per_second": 1.834, | |
| "step": 55 | |
| }, | |
| { | |
| "epoch": 10.91, | |
| "learning_rate": 0.00011666666666666665, | |
| "loss": 1.1901, | |
| "step": 60 | |
| }, | |
| { | |
| "epoch": 10.91, | |
| "eval_accuracy": 0.45161290322580644, | |
| "eval_loss": 1.1861207485198975, | |
| "eval_runtime": 2.094, | |
| "eval_samples_per_second": 29.608, | |
| "eval_steps_per_second": 1.91, | |
| "step": 60 | |
| }, | |
| { | |
| "epoch": 12.0, | |
| "eval_accuracy": 0.46774193548387094, | |
| "eval_loss": 1.0847262144088745, | |
| "eval_runtime": 2.212, | |
| "eval_samples_per_second": 28.028, | |
| "eval_steps_per_second": 1.808, | |
| "step": 66 | |
| }, | |
| { | |
| "epoch": 12.73, | |
| "learning_rate": 0.00010833333333333333, | |
| "loss": 1.0665, | |
| "step": 70 | |
| }, | |
| { | |
| "epoch": 12.91, | |
| "eval_accuracy": 0.46774193548387094, | |
| "eval_loss": 1.0943629741668701, | |
| "eval_runtime": 2.08, | |
| "eval_samples_per_second": 29.808, | |
| "eval_steps_per_second": 1.923, | |
| "step": 71 | |
| }, | |
| { | |
| "epoch": 14.0, | |
| "eval_accuracy": 0.46774193548387094, | |
| "eval_loss": 1.1853879690170288, | |
| "eval_runtime": 2.093, | |
| "eval_samples_per_second": 29.623, | |
| "eval_steps_per_second": 1.911, | |
| "step": 77 | |
| }, | |
| { | |
| "epoch": 14.55, | |
| "learning_rate": 9.999999999999999e-05, | |
| "loss": 1.033, | |
| "step": 80 | |
| }, | |
| { | |
| "epoch": 14.91, | |
| "eval_accuracy": 0.5, | |
| "eval_loss": 1.025220274925232, | |
| "eval_runtime": 2.4167, | |
| "eval_samples_per_second": 25.655, | |
| "eval_steps_per_second": 1.655, | |
| "step": 82 | |
| }, | |
| { | |
| "epoch": 16.0, | |
| "eval_accuracy": 0.5161290322580645, | |
| "eval_loss": 1.216417908668518, | |
| "eval_runtime": 2.2296, | |
| "eval_samples_per_second": 27.808, | |
| "eval_steps_per_second": 1.794, | |
| "step": 88 | |
| }, | |
| { | |
| "epoch": 16.36, | |
| "learning_rate": 9.166666666666667e-05, | |
| "loss": 1.0323, | |
| "step": 90 | |
| }, | |
| { | |
| "epoch": 16.91, | |
| "eval_accuracy": 0.5, | |
| "eval_loss": 1.0642980337142944, | |
| "eval_runtime": 2.199, | |
| "eval_samples_per_second": 28.194, | |
| "eval_steps_per_second": 1.819, | |
| "step": 93 | |
| }, | |
| { | |
| "epoch": 18.0, | |
| "eval_accuracy": 0.6612903225806451, | |
| "eval_loss": 0.9802310466766357, | |
| "eval_runtime": 2.1395, | |
| "eval_samples_per_second": 28.979, | |
| "eval_steps_per_second": 1.87, | |
| "step": 99 | |
| }, | |
| { | |
| "epoch": 18.18, | |
| "learning_rate": 8.333333333333333e-05, | |
| "loss": 0.9329, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 18.91, | |
| "eval_accuracy": 0.5967741935483871, | |
| "eval_loss": 0.9474769830703735, | |
| "eval_runtime": 2.179, | |
| "eval_samples_per_second": 28.453, | |
| "eval_steps_per_second": 1.836, | |
| "step": 104 | |
| }, | |
| { | |
| "epoch": 20.0, | |
| "learning_rate": 7.5e-05, | |
| "loss": 0.8619, | |
| "step": 110 | |
| }, | |
| { | |
| "epoch": 20.0, | |
| "eval_accuracy": 0.6451612903225806, | |
| "eval_loss": 0.9114610552787781, | |
| "eval_runtime": 2.125, | |
| "eval_samples_per_second": 29.176, | |
| "eval_steps_per_second": 1.882, | |
| "step": 110 | |
| }, | |
| { | |
| "epoch": 20.91, | |
| "eval_accuracy": 0.6451612903225806, | |
| "eval_loss": 0.8893528580665588, | |
| "eval_runtime": 2.2755, | |
| "eval_samples_per_second": 27.246, | |
| "eval_steps_per_second": 1.758, | |
| "step": 115 | |
| }, | |
| { | |
| "epoch": 21.82, | |
| "learning_rate": 6.666666666666666e-05, | |
| "loss": 0.8019, | |
| "step": 120 | |
| }, | |
| { | |
| "epoch": 22.0, | |
| "eval_accuracy": 0.6935483870967742, | |
| "eval_loss": 0.8276461958885193, | |
| "eval_runtime": 2.182, | |
| "eval_samples_per_second": 28.414, | |
| "eval_steps_per_second": 1.833, | |
| "step": 121 | |
| }, | |
| { | |
| "epoch": 22.91, | |
| "eval_accuracy": 0.6774193548387096, | |
| "eval_loss": 0.8156123757362366, | |
| "eval_runtime": 2.1835, | |
| "eval_samples_per_second": 28.394, | |
| "eval_steps_per_second": 1.832, | |
| "step": 126 | |
| }, | |
| { | |
| "epoch": 23.64, | |
| "learning_rate": 5.8333333333333326e-05, | |
| "loss": 0.7675, | |
| "step": 130 | |
| }, | |
| { | |
| "epoch": 24.0, | |
| "eval_accuracy": 0.6290322580645161, | |
| "eval_loss": 0.7928251624107361, | |
| "eval_runtime": 2.7367, | |
| "eval_samples_per_second": 22.655, | |
| "eval_steps_per_second": 1.462, | |
| "step": 132 | |
| }, | |
| { | |
| "epoch": 24.91, | |
| "eval_accuracy": 0.7419354838709677, | |
| "eval_loss": 0.7163397669792175, | |
| "eval_runtime": 2.3511, | |
| "eval_samples_per_second": 26.371, | |
| "eval_steps_per_second": 1.701, | |
| "step": 137 | |
| }, | |
| { | |
| "epoch": 25.45, | |
| "learning_rate": 4.9999999999999996e-05, | |
| "loss": 0.6762, | |
| "step": 140 | |
| }, | |
| { | |
| "epoch": 26.0, | |
| "eval_accuracy": 0.6774193548387096, | |
| "eval_loss": 0.7387820482254028, | |
| "eval_runtime": 2.261, | |
| "eval_samples_per_second": 27.421, | |
| "eval_steps_per_second": 1.769, | |
| "step": 143 | |
| }, | |
| { | |
| "epoch": 26.91, | |
| "eval_accuracy": 0.7580645161290323, | |
| "eval_loss": 0.6518718004226685, | |
| "eval_runtime": 2.2715, | |
| "eval_samples_per_second": 27.294, | |
| "eval_steps_per_second": 1.761, | |
| "step": 148 | |
| }, | |
| { | |
| "epoch": 27.27, | |
| "learning_rate": 4.1666666666666665e-05, | |
| "loss": 0.6771, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 28.0, | |
| "eval_accuracy": 0.7419354838709677, | |
| "eval_loss": 0.6709696054458618, | |
| "eval_runtime": 2.6851, | |
| "eval_samples_per_second": 23.09, | |
| "eval_steps_per_second": 1.49, | |
| "step": 154 | |
| }, | |
| { | |
| "epoch": 28.91, | |
| "eval_accuracy": 0.7580645161290323, | |
| "eval_loss": 0.6073653697967529, | |
| "eval_runtime": 2.3635, | |
| "eval_samples_per_second": 26.232, | |
| "eval_steps_per_second": 1.692, | |
| "step": 159 | |
| }, | |
| { | |
| "epoch": 29.09, | |
| "learning_rate": 3.333333333333333e-05, | |
| "loss": 0.6424, | |
| "step": 160 | |
| }, | |
| { | |
| "epoch": 30.0, | |
| "eval_accuracy": 0.7258064516129032, | |
| "eval_loss": 0.672946035861969, | |
| "eval_runtime": 2.2598, | |
| "eval_samples_per_second": 27.436, | |
| "eval_steps_per_second": 1.77, | |
| "step": 165 | |
| }, | |
| { | |
| "epoch": 30.91, | |
| "learning_rate": 2.4999999999999998e-05, | |
| "loss": 0.6139, | |
| "step": 170 | |
| }, | |
| { | |
| "epoch": 30.91, | |
| "eval_accuracy": 0.7903225806451613, | |
| "eval_loss": 0.5744480490684509, | |
| "eval_runtime": 2.189, | |
| "eval_samples_per_second": 28.323, | |
| "eval_steps_per_second": 1.827, | |
| "step": 170 | |
| }, | |
| { | |
| "epoch": 32.0, | |
| "eval_accuracy": 0.8225806451612904, | |
| "eval_loss": 0.6192476749420166, | |
| "eval_runtime": 2.142, | |
| "eval_samples_per_second": 28.945, | |
| "eval_steps_per_second": 1.867, | |
| "step": 176 | |
| }, | |
| { | |
| "epoch": 32.73, | |
| "learning_rate": 1.6666666666666664e-05, | |
| "loss": 0.5713, | |
| "step": 180 | |
| }, | |
| { | |
| "epoch": 32.91, | |
| "eval_accuracy": 0.7903225806451613, | |
| "eval_loss": 0.6452686190605164, | |
| "eval_runtime": 2.1495, | |
| "eval_samples_per_second": 28.844, | |
| "eval_steps_per_second": 1.861, | |
| "step": 181 | |
| }, | |
| { | |
| "epoch": 34.0, | |
| "eval_accuracy": 0.7903225806451613, | |
| "eval_loss": 0.6392035484313965, | |
| "eval_runtime": 2.1965, | |
| "eval_samples_per_second": 28.226, | |
| "eval_steps_per_second": 1.821, | |
| "step": 187 | |
| }, | |
| { | |
| "epoch": 34.55, | |
| "learning_rate": 8.333333333333332e-06, | |
| "loss": 0.5462, | |
| "step": 190 | |
| }, | |
| { | |
| "epoch": 34.91, | |
| "eval_accuracy": 0.8225806451612904, | |
| "eval_loss": 0.5955818295478821, | |
| "eval_runtime": 2.247, | |
| "eval_samples_per_second": 27.592, | |
| "eval_steps_per_second": 1.78, | |
| "step": 192 | |
| }, | |
| { | |
| "epoch": 36.0, | |
| "eval_accuracy": 0.8225806451612904, | |
| "eval_loss": 0.5892814993858337, | |
| "eval_runtime": 2.5326, | |
| "eval_samples_per_second": 24.481, | |
| "eval_steps_per_second": 1.579, | |
| "step": 198 | |
| }, | |
| { | |
| "epoch": 36.36, | |
| "learning_rate": 0.0, | |
| "loss": 0.5393, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 36.36, | |
| "eval_accuracy": 0.8225806451612904, | |
| "eval_loss": 0.5898378491401672, | |
| "eval_runtime": 2.2395, | |
| "eval_samples_per_second": 27.685, | |
| "eval_steps_per_second": 1.786, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 36.36, | |
| "step": 200, | |
| "total_flos": 4.141200256480051e+17, | |
| "train_loss": 1.5982162952423096, | |
| "train_runtime": 635.1405, | |
| "train_samples_per_second": 22.042, | |
| "train_steps_per_second": 0.315 | |
| } | |
| ], | |
| "logging_steps": 10, | |
| "max_steps": 200, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 40, | |
| "save_steps": 500, | |
| "total_flos": 4.141200256480051e+17, | |
| "train_batch_size": 16, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |