| { | |
| "best_metric": 0.9362666666666667, | |
| "best_model_checkpoint": "./vit_finetuned_models_dataset/CIFAR100/50_from_100/facebook_dino-vitb16/model_idx_0535/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": 5.234283924102783, | |
| "learning_rate": 6.224557891224557e-05, | |
| "loss": 0.7792, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.8664, | |
| "eval_loss": 0.4373243451118469, | |
| "eval_runtime": 5.7083, | |
| "eval_samples_per_second": 656.942, | |
| "eval_steps_per_second": 10.336, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "grad_norm": 47.0894775390625, | |
| "learning_rate": 5.44678011344678e-05, | |
| "loss": 0.2602, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.8832, | |
| "eval_loss": 0.40583664178848267, | |
| "eval_runtime": 5.8206, | |
| "eval_samples_per_second": 644.26, | |
| "eval_steps_per_second": 10.136, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "grad_norm": 0.22438035905361176, | |
| "learning_rate": 4.669002335669002e-05, | |
| "loss": 0.1235, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.8749333333333333, | |
| "eval_loss": 0.45611390471458435, | |
| "eval_runtime": 5.7583, | |
| "eval_samples_per_second": 651.232, | |
| "eval_steps_per_second": 10.246, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "grad_norm": 1.0553393363952637, | |
| "learning_rate": 3.891224557891224e-05, | |
| "loss": 0.0709, | |
| "step": 1332 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_accuracy": 0.8989333333333334, | |
| "eval_loss": 0.36512571573257446, | |
| "eval_runtime": 5.7679, | |
| "eval_samples_per_second": 650.154, | |
| "eval_steps_per_second": 10.229, | |
| "step": 1332 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "grad_norm": 0.05236368253827095, | |
| "learning_rate": 3.113446780113447e-05, | |
| "loss": 0.0396, | |
| "step": 1665 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_accuracy": 0.9034666666666666, | |
| "eval_loss": 0.3899645507335663, | |
| "eval_runtime": 5.7992, | |
| "eval_samples_per_second": 646.643, | |
| "eval_steps_per_second": 10.174, | |
| "step": 1665 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "grad_norm": 0.13746243715286255, | |
| "learning_rate": 2.3356690023356685e-05, | |
| "loss": 0.022, | |
| "step": 1998 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "eval_accuracy": 0.9138666666666667, | |
| "eval_loss": 0.3686300814151764, | |
| "eval_runtime": 5.7391, | |
| "eval_samples_per_second": 653.407, | |
| "eval_steps_per_second": 10.28, | |
| "step": 1998 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "grad_norm": 19.278738021850586, | |
| "learning_rate": 1.557891224557891e-05, | |
| "loss": 0.0087, | |
| "step": 2331 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "eval_accuracy": 0.9258666666666666, | |
| "eval_loss": 0.3199083209037781, | |
| "eval_runtime": 5.8216, | |
| "eval_samples_per_second": 644.155, | |
| "eval_steps_per_second": 10.135, | |
| "step": 2331 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "grad_norm": 0.003025069599971175, | |
| "learning_rate": 7.801134467801133e-06, | |
| "loss": 0.0026, | |
| "step": 2664 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "eval_accuracy": 0.9309333333333333, | |
| "eval_loss": 0.30572548508644104, | |
| "eval_runtime": 6.1696, | |
| "eval_samples_per_second": 607.819, | |
| "eval_steps_per_second": 9.563, | |
| "step": 2664 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "grad_norm": 0.0001581140822963789, | |
| "learning_rate": 2.3356690023356688e-08, | |
| "loss": 0.0012, | |
| "step": 2997 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "eval_accuracy": 0.9362666666666667, | |
| "eval_loss": 0.2782951593399048, | |
| "eval_runtime": 5.7939, | |
| "eval_samples_per_second": 647.238, | |
| "eval_steps_per_second": 10.183, | |
| "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 | |
| } | |