| { | |
| "best_metric": 0.3848, | |
| "best_model_checkpoint": "./vit_finetuned_models_dataset/CIFAR100/50_from_100/facebook_dino-vitb16/model_idx_0908/checkpoints/checkpoint-1665", | |
| "epoch": 5.0, | |
| "eval_steps": 500, | |
| "global_step": 1665, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 1.0, | |
| "grad_norm": 20.305028915405273, | |
| "learning_rate": 0.00024036036036036035, | |
| "loss": 3.3971, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.17493333333333333, | |
| "eval_loss": 3.0903031826019287, | |
| "eval_runtime": 8.2115, | |
| "eval_samples_per_second": 456.674, | |
| "eval_steps_per_second": 7.185, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "grad_norm": 30.96673011779785, | |
| "learning_rate": 0.00018036036036036035, | |
| "loss": 2.862, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.26293333333333335, | |
| "eval_loss": 2.7587833404541016, | |
| "eval_runtime": 7.6645, | |
| "eval_samples_per_second": 489.269, | |
| "eval_steps_per_second": 7.698, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "grad_norm": 28.90047836303711, | |
| "learning_rate": 0.00012036036036036035, | |
| "loss": 2.5586, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.30933333333333335, | |
| "eval_loss": 2.5488123893737793, | |
| "eval_runtime": 6.9112, | |
| "eval_samples_per_second": 542.599, | |
| "eval_steps_per_second": 8.537, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "grad_norm": 28.893056869506836, | |
| "learning_rate": 6.036036036036035e-05, | |
| "loss": 2.2628, | |
| "step": 1332 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_accuracy": 0.34613333333333335, | |
| "eval_loss": 2.403963088989258, | |
| "eval_runtime": 6.9229, | |
| "eval_samples_per_second": 541.678, | |
| "eval_steps_per_second": 8.522, | |
| "step": 1332 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "grad_norm": 44.267578125, | |
| "learning_rate": 3.6036036036036033e-07, | |
| "loss": 1.9004, | |
| "step": 1665 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_accuracy": 0.3848, | |
| "eval_loss": 2.3132681846618652, | |
| "eval_runtime": 6.8289, | |
| "eval_samples_per_second": 549.14, | |
| "eval_steps_per_second": 8.64, | |
| "step": 1665 | |
| } | |
| ], | |
| "logging_steps": 500, | |
| "max_steps": 1665, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 5, | |
| "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": 8.2370660311296e+18, | |
| "train_batch_size": 64, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |