| { | |
| "best_metric": 0.9128, | |
| "best_model_checkpoint": "./resnet_finetuned_models_dataset/CIFAR100/50_from_100/microsoft_resnet-101/model_idx_0585/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": 52.70469665527344, | |
| "learning_rate": 0.00025606601717798207, | |
| "loss": 1.4096, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.8570666666666666, | |
| "eval_loss": 0.444380521774292, | |
| "eval_runtime": 8.0571, | |
| "eval_samples_per_second": 465.431, | |
| "eval_steps_per_second": 7.323, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "grad_norm": 37.77268981933594, | |
| "learning_rate": 0.00015, | |
| "loss": 0.3314, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.8938666666666667, | |
| "eval_loss": 0.3458075225353241, | |
| "eval_runtime": 7.9205, | |
| "eval_samples_per_second": 473.454, | |
| "eval_steps_per_second": 7.449, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "grad_norm": 81.39505767822266, | |
| "learning_rate": 4.3933982822017876e-05, | |
| "loss": 0.1465, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.9128, | |
| "eval_loss": 0.2913879156112671, | |
| "eval_runtime": 7.9644, | |
| "eval_samples_per_second": 470.843, | |
| "eval_steps_per_second": 7.408, | |
| "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": 2.4529287718656e+18, | |
| "train_batch_size": 64, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |