| { | |
| "best_metric": 0.9586666666666667, | |
| "best_model_checkpoint": "./vit_finetuned_models_dataset/CIFAR100/50_from_100/google_vit-base-patch16-224/model_idx_0112/checkpoints/checkpoint-999", | |
| "epoch": 3.0, | |
| "eval_steps": 500, | |
| "global_step": 999, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 1.0, | |
| "grad_norm": 11.505019187927246, | |
| "learning_rate": 2.7135254915624213e-05, | |
| "loss": 1.0828, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.9456, | |
| "eval_loss": 0.24302427470684052, | |
| "eval_runtime": 6.0892, | |
| "eval_samples_per_second": 615.845, | |
| "eval_steps_per_second": 9.689, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "grad_norm": 0.5289295315742493, | |
| "learning_rate": 1.963525491562421e-05, | |
| "loss": 0.1111, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.9528, | |
| "eval_loss": 0.18275189399719238, | |
| "eval_runtime": 5.6136, | |
| "eval_samples_per_second": 668.022, | |
| "eval_steps_per_second": 10.51, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "grad_norm": 0.552828311920166, | |
| "learning_rate": 1.036474508437579e-05, | |
| "loss": 0.033, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.9586666666666667, | |
| "eval_loss": 0.16115573048591614, | |
| "eval_runtime": 6.0922, | |
| "eval_samples_per_second": 615.537, | |
| "eval_steps_per_second": 9.684, | |
| "step": 999 | |
| } | |
| ], | |
| "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": false | |
| }, | |
| "attributes": {} | |
| } | |
| }, | |
| "total_flos": 4.94223961867776e+18, | |
| "train_batch_size": 64, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |