| { | |
| "best_metric": 0.8874666666666666, | |
| "best_model_checkpoint": "./resnet_finetuned_models_dataset/CIFAR100/50_from_100/microsoft_resnet-101/model_idx_0776/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": 84.45658874511719, | |
| "learning_rate": 8.728616793536588e-05, | |
| "loss": 2.9036, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.7389333333333333, | |
| "eval_loss": 0.9384608268737793, | |
| "eval_runtime": 12.6052, | |
| "eval_samples_per_second": 297.497, | |
| "eval_steps_per_second": 4.681, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "grad_norm": 166.203369140625, | |
| "learning_rate": 7.947199994035401e-05, | |
| "loss": 0.6941, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.832, | |
| "eval_loss": 0.5253332853317261, | |
| "eval_runtime": 12.7046, | |
| "eval_samples_per_second": 295.169, | |
| "eval_steps_per_second": 4.644, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "grad_norm": 56.629432678222656, | |
| "learning_rate": 6.75e-05, | |
| "loss": 0.4252, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.8584, | |
| "eval_loss": 0.4312359392642975, | |
| "eval_runtime": 12.8121, | |
| "eval_samples_per_second": 292.692, | |
| "eval_steps_per_second": 4.605, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "grad_norm": 56.0867919921875, | |
| "learning_rate": 5.2814167995011876e-05, | |
| "loss": 0.3179, | |
| "step": 1332 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_accuracy": 0.8690666666666667, | |
| "eval_loss": 0.3986389935016632, | |
| "eval_runtime": 12.7201, | |
| "eval_samples_per_second": 294.809, | |
| "eval_steps_per_second": 4.638, | |
| "step": 1332 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "grad_norm": 58.19596481323242, | |
| "learning_rate": 3.7185832004988136e-05, | |
| "loss": 0.234, | |
| "step": 1665 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_accuracy": 0.88, | |
| "eval_loss": 0.3753884434700012, | |
| "eval_runtime": 12.6893, | |
| "eval_samples_per_second": 295.524, | |
| "eval_steps_per_second": 4.65, | |
| "step": 1665 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "grad_norm": 112.55115509033203, | |
| "learning_rate": 2.250000000000001e-05, | |
| "loss": 0.1904, | |
| "step": 1998 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "eval_accuracy": 0.876, | |
| "eval_loss": 0.37779584527015686, | |
| "eval_runtime": 12.7721, | |
| "eval_samples_per_second": 293.609, | |
| "eval_steps_per_second": 4.619, | |
| "step": 1998 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "grad_norm": 84.17382049560547, | |
| "learning_rate": 1.0528000059645995e-05, | |
| "loss": 0.1627, | |
| "step": 2331 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "eval_accuracy": 0.8848, | |
| "eval_loss": 0.3541330099105835, | |
| "eval_runtime": 12.8262, | |
| "eval_samples_per_second": 292.371, | |
| "eval_steps_per_second": 4.6, | |
| "step": 2331 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "grad_norm": 99.94281005859375, | |
| "learning_rate": 2.7138320646341257e-06, | |
| "loss": 0.1528, | |
| "step": 2664 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "eval_accuracy": 0.8824, | |
| "eval_loss": 0.3699255883693695, | |
| "eval_runtime": 12.4778, | |
| "eval_samples_per_second": 300.534, | |
| "eval_steps_per_second": 4.728, | |
| "step": 2664 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "grad_norm": 129.81895446777344, | |
| "learning_rate": 0.0, | |
| "loss": 0.1472, | |
| "step": 2997 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "eval_accuracy": 0.8874666666666666, | |
| "eval_loss": 0.36697909235954285, | |
| "eval_runtime": 13.169, | |
| "eval_samples_per_second": 284.759, | |
| "eval_steps_per_second": 4.48, | |
| "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 | |
| } | |