| { | |
| "best_metric": 0.8498666666666667, | |
| "best_model_checkpoint": "./resnet_finetuned_models_dataset/CIFAR100/50_from_100/microsoft_resnet-101/model_idx_0035/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": 63.73412322998047, | |
| "learning_rate": 6.222222222222222e-05, | |
| "loss": 3.5524, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.4512, | |
| "eval_loss": 2.578583002090454, | |
| "eval_runtime": 17.4893, | |
| "eval_samples_per_second": 214.416, | |
| "eval_steps_per_second": 3.373, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "grad_norm": 105.49139404296875, | |
| "learning_rate": 5.444444444444444e-05, | |
| "loss": 1.515, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.7525333333333334, | |
| "eval_loss": 0.9459144473075867, | |
| "eval_runtime": 17.0908, | |
| "eval_samples_per_second": 219.416, | |
| "eval_steps_per_second": 3.452, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "grad_norm": 123.62105560302734, | |
| "learning_rate": 4.666666666666666e-05, | |
| "loss": 0.8001, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.8048, | |
| "eval_loss": 0.6745380163192749, | |
| "eval_runtime": 17.3582, | |
| "eval_samples_per_second": 216.037, | |
| "eval_steps_per_second": 3.399, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "grad_norm": 209.11402893066406, | |
| "learning_rate": 3.8888888888888884e-05, | |
| "loss": 0.6106, | |
| "step": 1332 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_accuracy": 0.8250666666666666, | |
| "eval_loss": 0.5959022641181946, | |
| "eval_runtime": 17.5147, | |
| "eval_samples_per_second": 214.106, | |
| "eval_steps_per_second": 3.369, | |
| "step": 1332 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "grad_norm": 157.32958984375, | |
| "learning_rate": 3.111111111111111e-05, | |
| "loss": 0.5005, | |
| "step": 1665 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_accuracy": 0.8354666666666667, | |
| "eval_loss": 0.558962881565094, | |
| "eval_runtime": 16.608, | |
| "eval_samples_per_second": 225.794, | |
| "eval_steps_per_second": 3.552, | |
| "step": 1665 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "grad_norm": 271.3968200683594, | |
| "learning_rate": 2.333333333333333e-05, | |
| "loss": 0.4413, | |
| "step": 1998 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "eval_accuracy": 0.8365333333333334, | |
| "eval_loss": 0.5371218919754028, | |
| "eval_runtime": 16.8897, | |
| "eval_samples_per_second": 222.029, | |
| "eval_steps_per_second": 3.493, | |
| "step": 1998 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "grad_norm": 196.20407104492188, | |
| "learning_rate": 1.5555555555555555e-05, | |
| "loss": 0.4087, | |
| "step": 2331 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "eval_accuracy": 0.8490666666666666, | |
| "eval_loss": 0.5076178312301636, | |
| "eval_runtime": 16.6173, | |
| "eval_samples_per_second": 225.668, | |
| "eval_steps_per_second": 3.551, | |
| "step": 2331 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "grad_norm": 145.4641876220703, | |
| "learning_rate": 7.777777777777777e-06, | |
| "loss": 0.383, | |
| "step": 2664 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "eval_accuracy": 0.8461333333333333, | |
| "eval_loss": 0.5177536606788635, | |
| "eval_runtime": 17.1285, | |
| "eval_samples_per_second": 218.933, | |
| "eval_steps_per_second": 3.445, | |
| "step": 2664 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "grad_norm": 117.92924499511719, | |
| "learning_rate": 0.0, | |
| "loss": 0.3706, | |
| "step": 2997 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "eval_accuracy": 0.8498666666666667, | |
| "eval_loss": 0.49817273020744324, | |
| "eval_runtime": 15.4773, | |
| "eval_samples_per_second": 242.29, | |
| "eval_steps_per_second": 3.812, | |
| "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 | |
| } | |