| { | |
| "best_metric": 0.8672, | |
| "best_model_checkpoint": "./resnet_finetuned_models_dataset/CIFAR100/50_from_100/microsoft_resnet-101/model_idx_0751/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": 130.42794799804688, | |
| "learning_rate": 8e-05, | |
| "loss": 3.1767, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.6304, | |
| "eval_loss": 1.619924783706665, | |
| "eval_runtime": 7.4508, | |
| "eval_samples_per_second": 503.304, | |
| "eval_steps_per_second": 7.919, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "grad_norm": 57.93913650512695, | |
| "learning_rate": 7.000000000000001e-05, | |
| "loss": 0.9914, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.7912, | |
| "eval_loss": 0.7106790542602539, | |
| "eval_runtime": 7.5732, | |
| "eval_samples_per_second": 495.167, | |
| "eval_steps_per_second": 7.791, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "grad_norm": 120.84459686279297, | |
| "learning_rate": 6e-05, | |
| "loss": 0.6001, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.8298666666666666, | |
| "eval_loss": 0.5563246607780457, | |
| "eval_runtime": 7.4758, | |
| "eval_samples_per_second": 501.621, | |
| "eval_steps_per_second": 7.892, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "grad_norm": 80.57649230957031, | |
| "learning_rate": 5e-05, | |
| "loss": 0.4697, | |
| "step": 1332 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_accuracy": 0.8448, | |
| "eval_loss": 0.4914678931236267, | |
| "eval_runtime": 7.4882, | |
| "eval_samples_per_second": 500.79, | |
| "eval_steps_per_second": 7.879, | |
| "step": 1332 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "grad_norm": 96.25334167480469, | |
| "learning_rate": 4e-05, | |
| "loss": 0.3947, | |
| "step": 1665 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_accuracy": 0.8557333333333333, | |
| "eval_loss": 0.45968499779701233, | |
| "eval_runtime": 7.4939, | |
| "eval_samples_per_second": 500.407, | |
| "eval_steps_per_second": 7.873, | |
| "step": 1665 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "grad_norm": 80.68173217773438, | |
| "learning_rate": 3e-05, | |
| "loss": 0.3387, | |
| "step": 1998 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "eval_accuracy": 0.8581333333333333, | |
| "eval_loss": 0.459348201751709, | |
| "eval_runtime": 7.5062, | |
| "eval_samples_per_second": 499.584, | |
| "eval_steps_per_second": 7.86, | |
| "step": 1998 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "grad_norm": 81.7917251586914, | |
| "learning_rate": 2e-05, | |
| "loss": 0.2947, | |
| "step": 2331 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "eval_accuracy": 0.8613333333333333, | |
| "eval_loss": 0.45421797037124634, | |
| "eval_runtime": 7.4548, | |
| "eval_samples_per_second": 503.028, | |
| "eval_steps_per_second": 7.914, | |
| "step": 2331 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "grad_norm": 79.50867462158203, | |
| "learning_rate": 1e-05, | |
| "loss": 0.2813, | |
| "step": 2664 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "eval_accuracy": 0.8650666666666667, | |
| "eval_loss": 0.43677935004234314, | |
| "eval_runtime": 7.473, | |
| "eval_samples_per_second": 501.805, | |
| "eval_steps_per_second": 7.895, | |
| "step": 2664 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "grad_norm": 59.09619140625, | |
| "learning_rate": 0.0, | |
| "loss": 0.2665, | |
| "step": 2997 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "eval_accuracy": 0.8672, | |
| "eval_loss": 0.4373552203178406, | |
| "eval_runtime": 7.9944, | |
| "eval_samples_per_second": 469.076, | |
| "eval_steps_per_second": 7.38, | |
| "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 | |
| } | |