| { | |
| "best_metric": 0.8661333333333333, | |
| "best_model_checkpoint": "./resnet_finetuned_models_dataset/CIFAR100/50_from_100/microsoft_resnet-101/model_idx_0958/checkpoints/checkpoint-2997", | |
| "epoch": 9.0, | |
| "eval_steps": 500, | |
| "global_step": 2997, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 1.0, | |
| "grad_norm": 119.541259765625, | |
| "learning_rate": 9.698463103929542e-05, | |
| "loss": 2.7586, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.7146666666666667, | |
| "eval_loss": 0.9913139343261719, | |
| "eval_runtime": 8.9883, | |
| "eval_samples_per_second": 417.21, | |
| "eval_steps_per_second": 6.564, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "grad_norm": 86.09176635742188, | |
| "learning_rate": 8.83022221559489e-05, | |
| "loss": 0.7549, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.8178666666666666, | |
| "eval_loss": 0.5817207098007202, | |
| "eval_runtime": 8.5484, | |
| "eval_samples_per_second": 438.679, | |
| "eval_steps_per_second": 6.902, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "grad_norm": 71.77918243408203, | |
| "learning_rate": 7.500000000000001e-05, | |
| "loss": 0.5027, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.8434666666666667, | |
| "eval_loss": 0.5104503631591797, | |
| "eval_runtime": 8.4546, | |
| "eval_samples_per_second": 443.544, | |
| "eval_steps_per_second": 6.978, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "grad_norm": 42.540313720703125, | |
| "learning_rate": 5.868240888334653e-05, | |
| "loss": 0.361, | |
| "step": 1332 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_accuracy": 0.8512, | |
| "eval_loss": 0.4802725911140442, | |
| "eval_runtime": 8.4017, | |
| "eval_samples_per_second": 446.341, | |
| "eval_steps_per_second": 7.022, | |
| "step": 1332 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "grad_norm": 130.11361694335938, | |
| "learning_rate": 4.131759111665349e-05, | |
| "loss": 0.3023, | |
| "step": 1665 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_accuracy": 0.8578666666666667, | |
| "eval_loss": 0.4512491822242737, | |
| "eval_runtime": 8.34, | |
| "eval_samples_per_second": 449.643, | |
| "eval_steps_per_second": 7.074, | |
| "step": 1665 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "grad_norm": 69.21302795410156, | |
| "learning_rate": 2.500000000000001e-05, | |
| "loss": 0.2446, | |
| "step": 1998 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "eval_accuracy": 0.8557333333333333, | |
| "eval_loss": 0.46091583371162415, | |
| "eval_runtime": 8.3424, | |
| "eval_samples_per_second": 449.511, | |
| "eval_steps_per_second": 7.072, | |
| "step": 1998 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "grad_norm": 67.6849136352539, | |
| "learning_rate": 1.1697777844051105e-05, | |
| "loss": 0.2151, | |
| "step": 2331 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "eval_accuracy": 0.8605333333333334, | |
| "eval_loss": 0.4381713569164276, | |
| "eval_runtime": 8.3303, | |
| "eval_samples_per_second": 450.164, | |
| "eval_steps_per_second": 7.083, | |
| "step": 2331 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "grad_norm": 43.34357452392578, | |
| "learning_rate": 3.0153689607045845e-06, | |
| "loss": 0.2014, | |
| "step": 2664 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "eval_accuracy": 0.8586666666666667, | |
| "eval_loss": 0.44177383184432983, | |
| "eval_runtime": 9.2149, | |
| "eval_samples_per_second": 406.951, | |
| "eval_steps_per_second": 6.403, | |
| "step": 2664 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "grad_norm": 176.95364379882812, | |
| "learning_rate": 0.0, | |
| "loss": 0.1843, | |
| "step": 2997 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "eval_accuracy": 0.8661333333333333, | |
| "eval_loss": 0.4389706254005432, | |
| "eval_runtime": 8.3353, | |
| "eval_samples_per_second": 449.892, | |
| "eval_steps_per_second": 7.078, | |
| "step": 2997 | |
| } | |
| ], | |
| "logging_steps": 500, | |
| "max_steps": 2997, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 9, | |
| "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": 7.3587863155968e+18, | |
| "train_batch_size": 64, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |