| { | |
| "best_metric": 0.9026666666666666, | |
| "best_model_checkpoint": "./resnet_finetuned_models_dataset/CIFAR100/50_from_100/microsoft_resnet-101/model_idx_0758/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": 62.745548248291016, | |
| "learning_rate": 0.00019999999999999998, | |
| "loss": 1.4438, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.8496, | |
| "eval_loss": 0.45871633291244507, | |
| "eval_runtime": 14.5177, | |
| "eval_samples_per_second": 258.305, | |
| "eval_steps_per_second": 4.064, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "grad_norm": 71.86744689941406, | |
| "learning_rate": 9.999999999999999e-05, | |
| "loss": 0.3279, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.8992, | |
| "eval_loss": 0.32763826847076416, | |
| "eval_runtime": 14.4171, | |
| "eval_samples_per_second": 260.109, | |
| "eval_steps_per_second": 4.092, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "grad_norm": 57.8755989074707, | |
| "learning_rate": 0.0, | |
| "loss": 0.1627, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.9026666666666666, | |
| "eval_loss": 0.30687662959098816, | |
| "eval_runtime": 14.3671, | |
| "eval_samples_per_second": 261.012, | |
| "eval_steps_per_second": 4.107, | |
| "step": 999 | |
| } | |
| ], | |
| "logging_steps": 500, | |
| "max_steps": 999, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 3, | |
| "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": 2.4529287718656e+18, | |
| "train_batch_size": 64, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |