| { | |
| "best_metric": 0.8848, | |
| "best_model_checkpoint": "./resnet_finetuned_models_dataset/CIFAR100/50_from_100/microsoft_resnet-101/model_idx_0603/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": 126.7740249633789, | |
| "learning_rate": 8.888888888888889e-05, | |
| "loss": 3.0095, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.6866666666666666, | |
| "eval_loss": 1.1994926929473877, | |
| "eval_runtime": 7.2515, | |
| "eval_samples_per_second": 517.135, | |
| "eval_steps_per_second": 8.136, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "grad_norm": 60.11855697631836, | |
| "learning_rate": 7.777777777777778e-05, | |
| "loss": 0.8714, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.8213333333333334, | |
| "eval_loss": 0.5893322229385376, | |
| "eval_runtime": 5.3955, | |
| "eval_samples_per_second": 695.023, | |
| "eval_steps_per_second": 10.935, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "grad_norm": 63.1159782409668, | |
| "learning_rate": 6.666666666666667e-05, | |
| "loss": 0.558, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.8466666666666667, | |
| "eval_loss": 0.4808116853237152, | |
| "eval_runtime": 5.4355, | |
| "eval_samples_per_second": 689.915, | |
| "eval_steps_per_second": 10.855, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "grad_norm": 85.03471374511719, | |
| "learning_rate": 5.555555555555556e-05, | |
| "loss": 0.4266, | |
| "step": 1332 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_accuracy": 0.8618666666666667, | |
| "eval_loss": 0.4459698498249054, | |
| "eval_runtime": 5.395, | |
| "eval_samples_per_second": 695.09, | |
| "eval_steps_per_second": 10.936, | |
| "step": 1332 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "grad_norm": 134.0056915283203, | |
| "learning_rate": 4.4444444444444447e-05, | |
| "loss": 0.3347, | |
| "step": 1665 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_accuracy": 0.8709333333333333, | |
| "eval_loss": 0.40812715888023376, | |
| "eval_runtime": 5.5014, | |
| "eval_samples_per_second": 681.645, | |
| "eval_steps_per_second": 10.725, | |
| "step": 1665 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "grad_norm": 236.72254943847656, | |
| "learning_rate": 3.3333333333333335e-05, | |
| "loss": 0.2868, | |
| "step": 1998 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "eval_accuracy": 0.8674666666666667, | |
| "eval_loss": 0.4153875410556793, | |
| "eval_runtime": 5.5207, | |
| "eval_samples_per_second": 679.267, | |
| "eval_steps_per_second": 10.687, | |
| "step": 1998 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "grad_norm": 88.51991271972656, | |
| "learning_rate": 2.2222222222222223e-05, | |
| "loss": 0.2489, | |
| "step": 2331 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "eval_accuracy": 0.8805333333333333, | |
| "eval_loss": 0.3978155553340912, | |
| "eval_runtime": 5.3787, | |
| "eval_samples_per_second": 697.19, | |
| "eval_steps_per_second": 10.969, | |
| "step": 2331 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "grad_norm": 77.1013412475586, | |
| "learning_rate": 1.1111111111111112e-05, | |
| "loss": 0.2263, | |
| "step": 2664 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "eval_accuracy": 0.8808, | |
| "eval_loss": 0.39115339517593384, | |
| "eval_runtime": 5.5315, | |
| "eval_samples_per_second": 677.93, | |
| "eval_steps_per_second": 10.666, | |
| "step": 2664 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "grad_norm": 91.8115005493164, | |
| "learning_rate": 0.0, | |
| "loss": 0.211, | |
| "step": 2997 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "eval_accuracy": 0.8848, | |
| "eval_loss": 0.3805083930492401, | |
| "eval_runtime": 5.4404, | |
| "eval_samples_per_second": 689.292, | |
| "eval_steps_per_second": 10.845, | |
| "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 | |
| } | |