| { | |
| "best_metric": 0.9434666666666667, | |
| "best_model_checkpoint": "./vit_finetuned_models_dataset/CIFAR100/50_from_100/google_vit-base-patch16-224/model_idx_0978/checkpoints/checkpoint-999", | |
| "epoch": 3.0, | |
| "eval_steps": 500, | |
| "global_step": 999, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 1.0, | |
| "grad_norm": 2.3215794563293457, | |
| "learning_rate": 2.25e-05, | |
| "loss": 1.2109, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.9232, | |
| "eval_loss": 0.3141179382801056, | |
| "eval_runtime": 13.2062, | |
| "eval_samples_per_second": 283.958, | |
| "eval_steps_per_second": 4.468, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "grad_norm": 0.9567689299583435, | |
| "learning_rate": 1.5e-05, | |
| "loss": 0.1973, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.9386666666666666, | |
| "eval_loss": 0.22702883183956146, | |
| "eval_runtime": 13.1922, | |
| "eval_samples_per_second": 284.259, | |
| "eval_steps_per_second": 4.472, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "grad_norm": 0.17486350238323212, | |
| "learning_rate": 7.5e-06, | |
| "loss": 0.0988, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.9434666666666667, | |
| "eval_loss": 0.20820429921150208, | |
| "eval_runtime": 13.2635, | |
| "eval_samples_per_second": 282.731, | |
| "eval_steps_per_second": 4.448, | |
| "step": 999 | |
| } | |
| ], | |
| "logging_steps": 500, | |
| "max_steps": 1332, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 4, | |
| "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": 4.94223961867776e+18, | |
| "train_batch_size": 64, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |