| { | |
| "best_metric": 0.8517333333333333, | |
| "best_model_checkpoint": "./resnet_finetuned_models_dataset/CIFAR100/50_from_100/microsoft_resnet-101/model_idx_0457/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": 58.83272171020508, | |
| "learning_rate": 7e-05, | |
| "loss": 3.427, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.5362666666666667, | |
| "eval_loss": 2.096330404281616, | |
| "eval_runtime": 7.9554, | |
| "eval_samples_per_second": 471.381, | |
| "eval_steps_per_second": 7.416, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "grad_norm": 94.28506469726562, | |
| "learning_rate": 7e-05, | |
| "loss": 1.1893, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.7776, | |
| "eval_loss": 0.7451729774475098, | |
| "eval_runtime": 7.9933, | |
| "eval_samples_per_second": 469.144, | |
| "eval_steps_per_second": 7.381, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "grad_norm": 107.26242065429688, | |
| "learning_rate": 7e-05, | |
| "loss": 0.6587, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.82, | |
| "eval_loss": 0.5683677196502686, | |
| "eval_runtime": 7.9868, | |
| "eval_samples_per_second": 469.522, | |
| "eval_steps_per_second": 7.387, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "grad_norm": 42.13269805908203, | |
| "learning_rate": 7e-05, | |
| "loss": 0.5017, | |
| "step": 1332 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_accuracy": 0.848, | |
| "eval_loss": 0.5097728967666626, | |
| "eval_runtime": 7.9976, | |
| "eval_samples_per_second": 468.892, | |
| "eval_steps_per_second": 7.377, | |
| "step": 1332 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "grad_norm": 109.55901336669922, | |
| "learning_rate": 7e-05, | |
| "loss": 0.4176, | |
| "step": 1665 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_accuracy": 0.8517333333333333, | |
| "eval_loss": 0.4744020104408264, | |
| "eval_runtime": 7.9572, | |
| "eval_samples_per_second": 471.273, | |
| "eval_steps_per_second": 7.415, | |
| "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": 4.088214619776e+18, | |
| "train_batch_size": 64, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |