| { | |
| "best_metric": 0.9, | |
| "best_model_checkpoint": "./vit_finetuned_models_dataset/CIFAR100/50_from_100/facebook_vit-mae-base/model_idx_0687/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": 2.988802433013916, | |
| "learning_rate": 8.888888888888889e-05, | |
| "loss": 1.6611, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.7744, | |
| "eval_loss": 0.787212610244751, | |
| "eval_runtime": 6.1593, | |
| "eval_samples_per_second": 608.839, | |
| "eval_steps_per_second": 9.579, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "grad_norm": 0.4985572099685669, | |
| "learning_rate": 7.777777777777778e-05, | |
| "loss": 0.5065, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.8474666666666667, | |
| "eval_loss": 0.5247082710266113, | |
| "eval_runtime": 5.9591, | |
| "eval_samples_per_second": 629.294, | |
| "eval_steps_per_second": 9.901, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "grad_norm": 1.2039424180984497, | |
| "learning_rate": 6.666666666666667e-05, | |
| "loss": 0.281, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.8592, | |
| "eval_loss": 0.47106438875198364, | |
| "eval_runtime": 6.1006, | |
| "eval_samples_per_second": 614.697, | |
| "eval_steps_per_second": 9.671, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "grad_norm": 2.18949556350708, | |
| "learning_rate": 5.5588922255588925e-05, | |
| "loss": 0.1616, | |
| "step": 1332 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_accuracy": 0.8781333333333333, | |
| "eval_loss": 0.4181782603263855, | |
| "eval_runtime": 5.9849, | |
| "eval_samples_per_second": 626.579, | |
| "eval_steps_per_second": 9.858, | |
| "step": 1332 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "grad_norm": 0.14587345719337463, | |
| "learning_rate": 4.447781114447781e-05, | |
| "loss": 0.0932, | |
| "step": 1665 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_accuracy": 0.8754666666666666, | |
| "eval_loss": 0.44042229652404785, | |
| "eval_runtime": 6.4249, | |
| "eval_samples_per_second": 583.666, | |
| "eval_steps_per_second": 9.183, | |
| "step": 1665 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "grad_norm": 0.034797124564647675, | |
| "learning_rate": 3.33667000333667e-05, | |
| "loss": 0.0459, | |
| "step": 1998 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "eval_accuracy": 0.8834666666666666, | |
| "eval_loss": 0.4565930664539337, | |
| "eval_runtime": 5.9607, | |
| "eval_samples_per_second": 629.122, | |
| "eval_steps_per_second": 9.898, | |
| "step": 1998 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "grad_norm": 0.8328613042831421, | |
| "learning_rate": 2.225558892225559e-05, | |
| "loss": 0.0231, | |
| "step": 2331 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "eval_accuracy": 0.8941333333333333, | |
| "eval_loss": 0.43445754051208496, | |
| "eval_runtime": 7.4278, | |
| "eval_samples_per_second": 504.86, | |
| "eval_steps_per_second": 7.943, | |
| "step": 2331 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "grad_norm": 0.0941711813211441, | |
| "learning_rate": 1.1144477811144478e-05, | |
| "loss": 0.0098, | |
| "step": 2664 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "eval_accuracy": 0.8989333333333334, | |
| "eval_loss": 0.4136076867580414, | |
| "eval_runtime": 6.0327, | |
| "eval_samples_per_second": 621.612, | |
| "eval_steps_per_second": 9.78, | |
| "step": 2664 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "grad_norm": 0.02813485637307167, | |
| "learning_rate": 3.3366700033366706e-08, | |
| "loss": 0.0044, | |
| "step": 2997 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "eval_accuracy": 0.9, | |
| "eval_loss": 0.40724629163742065, | |
| "eval_runtime": 6.8864, | |
| "eval_samples_per_second": 544.551, | |
| "eval_steps_per_second": 8.568, | |
| "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 | |
| } | |