| { | |
| "best_metric": 0.9098666666666667, | |
| "best_model_checkpoint": "./vit_finetuned_models_dataset/CIFAR100/50_from_100/facebook_vit-mae-base/model_idx_0896/checkpoints/checkpoint-2997", | |
| "epoch": 9.0, | |
| "eval_steps": 500, | |
| "global_step": 2997, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 1.0, | |
| "grad_norm": 6.375346660614014, | |
| "learning_rate": 9.700253128109275e-05, | |
| "loss": 1.4018, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.8157333333333333, | |
| "eval_loss": 0.6357667446136475, | |
| "eval_runtime": 6.7736, | |
| "eval_samples_per_second": 553.619, | |
| "eval_steps_per_second": 8.71, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "grad_norm": 0.36324504017829895, | |
| "learning_rate": 8.833589107668211e-05, | |
| "loss": 0.4045, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.8584, | |
| "eval_loss": 0.4767511487007141, | |
| "eval_runtime": 6.4873, | |
| "eval_samples_per_second": 578.051, | |
| "eval_steps_per_second": 9.095, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "grad_norm": 3.6083827018737793, | |
| "learning_rate": 7.504537663092833e-05, | |
| "loss": 0.2002, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.8776, | |
| "eval_loss": 0.4252709746360779, | |
| "eval_runtime": 7.3869, | |
| "eval_samples_per_second": 507.657, | |
| "eval_steps_per_second": 7.987, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "grad_norm": 0.0531436949968338, | |
| "learning_rate": 5.873402013309226e-05, | |
| "loss": 0.0963, | |
| "step": 1332 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_accuracy": 0.8909333333333334, | |
| "eval_loss": 0.3985811173915863, | |
| "eval_runtime": 7.1096, | |
| "eval_samples_per_second": 527.452, | |
| "eval_steps_per_second": 8.299, | |
| "step": 1332 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "grad_norm": 0.036292366683483124, | |
| "learning_rate": 4.136921190679637e-05, | |
| "loss": 0.0386, | |
| "step": 1665 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_accuracy": 0.8994666666666666, | |
| "eval_loss": 0.40463823080062866, | |
| "eval_runtime": 6.4319, | |
| "eval_samples_per_second": 583.027, | |
| "eval_steps_per_second": 9.173, | |
| "step": 1665 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "grad_norm": 0.029274918138980865, | |
| "learning_rate": 2.5045404101407076e-05, | |
| "loss": 0.0117, | |
| "step": 1998 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "eval_accuracy": 0.8992, | |
| "eval_loss": 0.41360729932785034, | |
| "eval_runtime": 7.0354, | |
| "eval_samples_per_second": 533.017, | |
| "eval_steps_per_second": 8.386, | |
| "step": 1998 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "grad_norm": 0.029571352526545525, | |
| "learning_rate": 1.1731488851999511e-05, | |
| "loss": 0.003, | |
| "step": 2331 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "eval_accuracy": 0.9074666666666666, | |
| "eval_loss": 0.38919830322265625, | |
| "eval_runtime": 7.0719, | |
| "eval_samples_per_second": 530.264, | |
| "eval_steps_per_second": 8.343, | |
| "step": 2331 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "grad_norm": 0.019404198974370956, | |
| "learning_rate": 3.033320830114261e-06, | |
| "loss": 0.0018, | |
| "step": 2664 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "eval_accuracy": 0.9093333333333333, | |
| "eval_loss": 0.38775086402893066, | |
| "eval_runtime": 11.1584, | |
| "eval_samples_per_second": 336.07, | |
| "eval_steps_per_second": 5.288, | |
| "step": 2664 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "grad_norm": 0.10369841754436493, | |
| "learning_rate": 2.7470478758395702e-11, | |
| "loss": 0.0017, | |
| "step": 2997 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "eval_accuracy": 0.9098666666666667, | |
| "eval_loss": 0.38768866658210754, | |
| "eval_runtime": 7.663, | |
| "eval_samples_per_second": 489.365, | |
| "eval_steps_per_second": 7.699, | |
| "step": 2997 | |
| } | |
| ], | |
| "logging_steps": 500, | |
| "max_steps": 2997, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 9, | |
| "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": 1.482671885603328e+19, | |
| "train_batch_size": 64, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |