{ "best_metric": 0.6818181818181818, "best_model_checkpoint": "swinv2-tiny-patch4-window8-256-dmae-humeda-DAV41/checkpoint-30", "epoch": 20.0, "eval_steps": 500, "global_step": 40, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 1.0, "eval_accuracy": 0.4318181818181818, "eval_loss": 1.4874267578125, "eval_runtime": 1.7179, "eval_samples_per_second": 51.224, "eval_steps_per_second": 1.746, "step": 2 }, { "epoch": 2.0, "eval_accuracy": 0.4431818181818182, "eval_loss": 1.3560291528701782, "eval_runtime": 1.9222, "eval_samples_per_second": 45.781, "eval_steps_per_second": 1.561, "step": 4 }, { "epoch": 3.0, "eval_accuracy": 0.4431818181818182, "eval_loss": 1.3116788864135742, "eval_runtime": 2.4062, "eval_samples_per_second": 36.573, "eval_steps_per_second": 1.247, "step": 6 }, { "epoch": 4.0, "eval_accuracy": 0.4431818181818182, "eval_loss": 1.2763394117355347, "eval_runtime": 2.2451, "eval_samples_per_second": 39.197, "eval_steps_per_second": 1.336, "step": 8 }, { "epoch": 5.0, "eval_accuracy": 0.5681818181818182, "eval_loss": 1.2601596117019653, "eval_runtime": 2.0294, "eval_samples_per_second": 43.363, "eval_steps_per_second": 1.478, "step": 10 }, { "epoch": 6.0, "grad_norm": 12.913003921508789, "learning_rate": 2.728728066433488e-05, "loss": 8.9996, "step": 12 }, { "epoch": 6.0, "eval_accuracy": 0.6022727272727273, "eval_loss": 1.2348134517669678, "eval_runtime": 1.5272, "eval_samples_per_second": 57.623, "eval_steps_per_second": 1.964, "step": 12 }, { "epoch": 7.0, "eval_accuracy": 0.5795454545454546, "eval_loss": 1.1982394456863403, "eval_runtime": 1.4872, "eval_samples_per_second": 59.172, "eval_steps_per_second": 2.017, "step": 14 }, { "epoch": 8.0, "eval_accuracy": 0.6136363636363636, "eval_loss": 1.1591989994049072, "eval_runtime": 1.5154, "eval_samples_per_second": 58.07, "eval_steps_per_second": 1.98, "step": 16 }, { "epoch": 9.0, "eval_accuracy": 0.625, "eval_loss": 1.1142396926879883, "eval_runtime": 1.4939, "eval_samples_per_second": 58.904, "eval_steps_per_second": 2.008, "step": 18 }, { "epoch": 10.0, "eval_accuracy": 0.6363636363636364, "eval_loss": 1.06820547580719, "eval_runtime": 1.5178, "eval_samples_per_second": 57.979, "eval_steps_per_second": 1.977, "step": 20 }, { "epoch": 11.0, "eval_accuracy": 0.6477272727272727, "eval_loss": 1.0256125926971436, "eval_runtime": 1.4495, "eval_samples_per_second": 60.712, "eval_steps_per_second": 2.07, "step": 22 }, { "epoch": 12.0, "grad_norm": 17.359237670898438, "learning_rate": 1.3692663858785127e-05, "loss": 7.429, "step": 24 }, { "epoch": 12.0, "eval_accuracy": 0.6704545454545454, "eval_loss": 0.984276533126831, "eval_runtime": 1.6744, "eval_samples_per_second": 52.558, "eval_steps_per_second": 1.792, "step": 24 }, { "epoch": 13.0, "eval_accuracy": 0.6704545454545454, "eval_loss": 0.9602272510528564, "eval_runtime": 1.6269, "eval_samples_per_second": 54.089, "eval_steps_per_second": 1.844, "step": 26 }, { "epoch": 14.0, "eval_accuracy": 0.6590909090909091, "eval_loss": 0.9451698064804077, "eval_runtime": 1.7471, "eval_samples_per_second": 50.369, "eval_steps_per_second": 1.717, "step": 28 }, { "epoch": 15.0, "eval_accuracy": 0.6818181818181818, "eval_loss": 0.9385179877281189, "eval_runtime": 2.3184, "eval_samples_per_second": 37.957, "eval_steps_per_second": 1.294, "step": 30 }, { "epoch": 16.0, "eval_accuracy": 0.6704545454545454, "eval_loss": 0.9320499897003174, "eval_runtime": 2.0728, "eval_samples_per_second": 42.455, "eval_steps_per_second": 1.447, "step": 32 }, { "epoch": 17.0, "eval_accuracy": 0.6477272727272727, "eval_loss": 0.9284698963165283, "eval_runtime": 1.5384, "eval_samples_per_second": 57.203, "eval_steps_per_second": 1.95, "step": 34 }, { "epoch": 18.0, "grad_norm": 22.705101013183594, "learning_rate": 1.4053831944502509e-06, "loss": 6.2752, "step": 36 }, { "epoch": 18.0, "eval_accuracy": 0.6590909090909091, "eval_loss": 0.9239160418510437, "eval_runtime": 1.5387, "eval_samples_per_second": 57.193, "eval_steps_per_second": 1.95, "step": 36 }, { "epoch": 19.0, "eval_accuracy": 0.6818181818181818, "eval_loss": 0.9213680028915405, "eval_runtime": 1.4957, "eval_samples_per_second": 58.834, "eval_steps_per_second": 2.006, "step": 38 }, { "epoch": 20.0, "eval_accuracy": 0.6818181818181818, "eval_loss": 0.9205734729766846, "eval_runtime": 1.81, "eval_samples_per_second": 48.618, "eval_steps_per_second": 1.657, "step": 40 }, { "epoch": 20.0, "step": 40, "total_flos": 2.6550232764383232e+17, "train_loss": 7.41663556098938, "train_runtime": 264.8298, "train_samples_per_second": 61.624, "train_steps_per_second": 0.151 } ], "logging_steps": 12, "max_steps": 40, "num_input_tokens_seen": 0, "num_train_epochs": 40, "save_steps": 500, "stateful_callbacks": { "TrainerControl": { "args": { "should_epoch_stop": false, "should_evaluate": false, "should_log": false, "should_save": true, "should_training_stop": true }, "attributes": {} } }, "total_flos": 2.6550232764383232e+17, "train_batch_size": 32, "trial_name": null, "trial_params": null }