| { | |
| "best_metric": 0.8554666666666667, | |
| "best_model_checkpoint": "./resnet_finetuned_models_dataset/CIFAR100/50_from_100/microsoft_resnet-101/model_idx_0188/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": 64.27982330322266, | |
| "learning_rate": 2.6666666666666667e-05, | |
| "loss": 3.8113, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.45866666666666667, | |
| "eval_loss": 3.6000051498413086, | |
| "eval_runtime": 7.2453, | |
| "eval_samples_per_second": 517.576, | |
| "eval_steps_per_second": 8.143, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "grad_norm": 136.06231689453125, | |
| "learning_rate": 2.3333333333333336e-05, | |
| "loss": 3.0614, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.6362666666666666, | |
| "eval_loss": 2.5395872592926025, | |
| "eval_runtime": 7.1071, | |
| "eval_samples_per_second": 527.642, | |
| "eval_steps_per_second": 8.302, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "grad_norm": 136.657470703125, | |
| "learning_rate": 1.9999999999999998e-05, | |
| "loss": 1.8256, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.7368, | |
| "eval_loss": 1.2459677457809448, | |
| "eval_runtime": 7.0711, | |
| "eval_samples_per_second": 530.328, | |
| "eval_steps_per_second": 8.344, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "grad_norm": 247.00636291503906, | |
| "learning_rate": 1.6666666666666667e-05, | |
| "loss": 1.0021, | |
| "step": 1332 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_accuracy": 0.8, | |
| "eval_loss": 0.820736825466156, | |
| "eval_runtime": 6.9933, | |
| "eval_samples_per_second": 536.23, | |
| "eval_steps_per_second": 8.437, | |
| "step": 1332 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "grad_norm": 239.13499450683594, | |
| "learning_rate": 1.3333333333333333e-05, | |
| "loss": 0.7308, | |
| "step": 1665 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_accuracy": 0.8290666666666666, | |
| "eval_loss": 0.6644681692123413, | |
| "eval_runtime": 6.9751, | |
| "eval_samples_per_second": 537.623, | |
| "eval_steps_per_second": 8.459, | |
| "step": 1665 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "grad_norm": 161.26589965820312, | |
| "learning_rate": 9.999999999999999e-06, | |
| "loss": 0.6093, | |
| "step": 1998 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "eval_accuracy": 0.8349333333333333, | |
| "eval_loss": 0.5838949084281921, | |
| "eval_runtime": 7.0283, | |
| "eval_samples_per_second": 533.56, | |
| "eval_steps_per_second": 8.395, | |
| "step": 1998 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "grad_norm": 272.24591064453125, | |
| "learning_rate": 6.666666666666667e-06, | |
| "loss": 0.5349, | |
| "step": 2331 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "eval_accuracy": 0.8437333333333333, | |
| "eval_loss": 0.5550447702407837, | |
| "eval_runtime": 7.7407, | |
| "eval_samples_per_second": 484.455, | |
| "eval_steps_per_second": 7.622, | |
| "step": 2331 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "grad_norm": 105.18741607666016, | |
| "learning_rate": 3.3333333333333333e-06, | |
| "loss": 0.4917, | |
| "step": 2664 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "eval_accuracy": 0.8504, | |
| "eval_loss": 0.531034529209137, | |
| "eval_runtime": 7.004, | |
| "eval_samples_per_second": 535.405, | |
| "eval_steps_per_second": 8.424, | |
| "step": 2664 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "grad_norm": 132.23776245117188, | |
| "learning_rate": 0.0, | |
| "loss": 0.4712, | |
| "step": 2997 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "eval_accuracy": 0.8554666666666667, | |
| "eval_loss": 0.5227321982383728, | |
| "eval_runtime": 6.9615, | |
| "eval_samples_per_second": 538.678, | |
| "eval_steps_per_second": 8.475, | |
| "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 | |
| } | |