| { | |
| "best_metric": 0.9128, | |
| "best_model_checkpoint": "./vit_finetuned_models_dataset/CIFAR100/50_from_100/google_vit-base-patch16-224/model_idx_0928/checkpoints/checkpoint-1665", | |
| "epoch": 5.0, | |
| "eval_steps": 500, | |
| "global_step": 1665, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 1.0, | |
| "grad_norm": 5.8659539222717285, | |
| "learning_rate": 0.0004522542485937369, | |
| "loss": 0.7394, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.8114666666666667, | |
| "eval_loss": 0.6578276753425598, | |
| "eval_runtime": 11.3702, | |
| "eval_samples_per_second": 329.81, | |
| "eval_steps_per_second": 5.189, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "grad_norm": 0.2950479984283447, | |
| "learning_rate": 0.00032725424859373687, | |
| "loss": 0.3601, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.8552, | |
| "eval_loss": 0.5236067175865173, | |
| "eval_runtime": 10.8396, | |
| "eval_samples_per_second": 345.954, | |
| "eval_steps_per_second": 5.443, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "grad_norm": 0.10129328817129135, | |
| "learning_rate": 0.00017274575140626317, | |
| "loss": 0.1541, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.8837333333333334, | |
| "eval_loss": 0.43295222520828247, | |
| "eval_runtime": 11.3619, | |
| "eval_samples_per_second": 330.051, | |
| "eval_steps_per_second": 5.193, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "grad_norm": 0.028341805562376976, | |
| "learning_rate": 4.7745751406263163e-05, | |
| "loss": 0.0362, | |
| "step": 1332 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_accuracy": 0.9112, | |
| "eval_loss": 0.3561219274997711, | |
| "eval_runtime": 10.674, | |
| "eval_samples_per_second": 351.321, | |
| "eval_steps_per_second": 5.527, | |
| "step": 1332 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "grad_norm": 0.32174837589263916, | |
| "learning_rate": 0.0, | |
| "loss": 0.0058, | |
| "step": 1665 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_accuracy": 0.9128, | |
| "eval_loss": 0.3513399660587311, | |
| "eval_runtime": 10.2471, | |
| "eval_samples_per_second": 365.956, | |
| "eval_steps_per_second": 5.758, | |
| "step": 1665 | |
| } | |
| ], | |
| "logging_steps": 500, | |
| "max_steps": 1665, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 5, | |
| "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": 8.2370660311296e+18, | |
| "train_batch_size": 64, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |