| { | |
| "best_metric": 0.9365333333333333, | |
| "best_model_checkpoint": "./vit_finetuned_models_dataset/CIFAR100/50_from_100/google_vit-base-patch16-224/model_idx_0789/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": 4.640089511871338, | |
| "learning_rate": 2.8514533018536286e-05, | |
| "loss": 1.1908, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.9165333333333333, | |
| "eval_loss": 0.32845690846443176, | |
| "eval_runtime": 13.4645, | |
| "eval_samples_per_second": 278.51, | |
| "eval_steps_per_second": 4.382, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "grad_norm": 1.8589054346084595, | |
| "learning_rate": 2.4352347027881003e-05, | |
| "loss": 0.2181, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.9266666666666666, | |
| "eval_loss": 0.25419145822525024, | |
| "eval_runtime": 13.3985, | |
| "eval_samples_per_second": 279.881, | |
| "eval_steps_per_second": 4.403, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "grad_norm": 0.09859631955623627, | |
| "learning_rate": 1.8337814009344716e-05, | |
| "loss": 0.1061, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.9336, | |
| "eval_loss": 0.2259616106748581, | |
| "eval_runtime": 12.6891, | |
| "eval_samples_per_second": 295.529, | |
| "eval_steps_per_second": 4.65, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "grad_norm": 0.16533997654914856, | |
| "learning_rate": 1.1662185990655285e-05, | |
| "loss": 0.0514, | |
| "step": 1332 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_accuracy": 0.9325333333333333, | |
| "eval_loss": 0.2394624650478363, | |
| "eval_runtime": 13.2418, | |
| "eval_samples_per_second": 283.194, | |
| "eval_steps_per_second": 4.456, | |
| "step": 1332 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "grad_norm": 0.0719929113984108, | |
| "learning_rate": 5.647652972118998e-06, | |
| "loss": 0.0272, | |
| "step": 1665 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_accuracy": 0.9365333333333333, | |
| "eval_loss": 0.2279118448495865, | |
| "eval_runtime": 13.3624, | |
| "eval_samples_per_second": 280.638, | |
| "eval_steps_per_second": 4.415, | |
| "step": 1665 | |
| } | |
| ], | |
| "logging_steps": 500, | |
| "max_steps": 2331, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 7, | |
| "save_steps": 500, | |
| "stateful_callbacks": { | |
| "TrainerControl": { | |
| "args": { | |
| "should_epoch_stop": false, | |
| "should_evaluate": false, | |
| "should_log": false, | |
| "should_save": true, | |
| "should_training_stop": false | |
| }, | |
| "attributes": {} | |
| } | |
| }, | |
| "total_flos": 8.2370660311296e+18, | |
| "train_batch_size": 64, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |