| { | |
| "best_metric": 0.9096, | |
| "best_model_checkpoint": "./vit_finetuned_models_dataset/CIFAR100/50_from_100/facebook_dino-vitb16/model_idx_0298/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": 0.4984889030456543, | |
| "learning_rate": 9.702037987560858e-05, | |
| "loss": 0.9065, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.7853333333333333, | |
| "eval_loss": 0.6793164014816284, | |
| "eval_runtime": 13.0821, | |
| "eval_samples_per_second": 286.651, | |
| "eval_steps_per_second": 4.51, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "grad_norm": 34.72991943359375, | |
| "learning_rate": 8.836951787320407e-05, | |
| "loss": 0.3952, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.8282666666666667, | |
| "eval_loss": 0.564298152923584, | |
| "eval_runtime": 13.2424, | |
| "eval_samples_per_second": 283.18, | |
| "eval_steps_per_second": 4.455, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "grad_norm": 9.979313850402832, | |
| "learning_rate": 7.509072574151719e-05, | |
| "loss": 0.2418, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.8528, | |
| "eval_loss": 0.5273523330688477, | |
| "eval_runtime": 13.8525, | |
| "eval_samples_per_second": 270.71, | |
| "eval_steps_per_second": 4.259, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "grad_norm": 6.916951656341553, | |
| "learning_rate": 5.8785621785729404e-05, | |
| "loss": 0.1291, | |
| "step": 1332 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_accuracy": 0.8576, | |
| "eval_loss": 0.5216265916824341, | |
| "eval_runtime": 13.0801, | |
| "eval_samples_per_second": 286.696, | |
| "eval_steps_per_second": 4.511, | |
| "step": 1332 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "grad_norm": 0.04581255838274956, | |
| "learning_rate": 4.142084218061449e-05, | |
| "loss": 0.0666, | |
| "step": 1665 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_accuracy": 0.8669333333333333, | |
| "eval_loss": 0.5305461883544922, | |
| "eval_runtime": 13.0805, | |
| "eval_samples_per_second": 286.687, | |
| "eval_steps_per_second": 4.511, | |
| "step": 1665 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "grad_norm": 0.27209460735321045, | |
| "learning_rate": 2.5090835623402032e-05, | |
| "loss": 0.0276, | |
| "step": 1998 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "eval_accuracy": 0.8858666666666667, | |
| "eval_loss": 0.4664764702320099, | |
| "eval_runtime": 13.3848, | |
| "eval_samples_per_second": 280.168, | |
| "eval_steps_per_second": 4.408, | |
| "step": 1998 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "grad_norm": 0.5151094794273376, | |
| "learning_rate": 1.1765241910120828e-05, | |
| "loss": 0.0086, | |
| "step": 2331 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "eval_accuracy": 0.9024, | |
| "eval_loss": 0.42460644245147705, | |
| "eval_runtime": 13.1696, | |
| "eval_samples_per_second": 284.747, | |
| "eval_steps_per_second": 4.48, | |
| "step": 2331 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "grad_norm": 0.003553769551217556, | |
| "learning_rate": 3.0513243074104303e-06, | |
| "loss": 0.0031, | |
| "step": 2664 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "eval_accuracy": 0.9061333333333333, | |
| "eval_loss": 0.4173130989074707, | |
| "eval_runtime": 13.1676, | |
| "eval_samples_per_second": 284.791, | |
| "eval_steps_per_second": 4.481, | |
| "step": 2664 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "grad_norm": 0.15246319770812988, | |
| "learning_rate": 1.0988188484661876e-10, | |
| "loss": 0.0019, | |
| "step": 2997 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "eval_accuracy": 0.9096, | |
| "eval_loss": 0.4042559862136841, | |
| "eval_runtime": 13.0598, | |
| "eval_samples_per_second": 287.142, | |
| "eval_steps_per_second": 4.518, | |
| "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": 1.482671885603328e+19, | |
| "train_batch_size": 64, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |