| { | |
| "best_metric": 0.9197333333333333, | |
| "best_model_checkpoint": "./vit_finetuned_models_dataset/CIFAR100/50_from_100/facebook_dino-vitb16/model_idx_0015/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": 2.0804874897003174, | |
| "learning_rate": 8.730227815298347e-05, | |
| "loss": 0.9125, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.8370666666666666, | |
| "eval_loss": 0.5371624231338501, | |
| "eval_runtime": 6.7431, | |
| "eval_samples_per_second": 556.123, | |
| "eval_steps_per_second": 8.75, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "grad_norm": 4.958555221557617, | |
| "learning_rate": 7.953256608588366e-05, | |
| "loss": 0.3497, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.8605333333333334, | |
| "eval_loss": 0.4794282913208008, | |
| "eval_runtime": 6.7605, | |
| "eval_samples_per_second": 554.691, | |
| "eval_steps_per_second": 8.727, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "grad_norm": 0.4777720272541046, | |
| "learning_rate": 6.758165316736547e-05, | |
| "loss": 0.1955, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.8653333333333333, | |
| "eval_loss": 0.4956216812133789, | |
| "eval_runtime": 6.8852, | |
| "eval_samples_per_second": 544.648, | |
| "eval_steps_per_second": 8.569, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "grad_norm": 0.01553698256611824, | |
| "learning_rate": 5.2907059607156464e-05, | |
| "loss": 0.1111, | |
| "step": 1332 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_accuracy": 0.8754666666666666, | |
| "eval_loss": 0.4888851046562195, | |
| "eval_runtime": 6.8904, | |
| "eval_samples_per_second": 544.238, | |
| "eval_steps_per_second": 8.563, | |
| "step": 1332 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "grad_norm": 1.1844000816345215, | |
| "learning_rate": 3.727875796255304e-05, | |
| "loss": 0.0592, | |
| "step": 1665 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_accuracy": 0.8930666666666667, | |
| "eval_loss": 0.44375741481781006, | |
| "eval_runtime": 7.1039, | |
| "eval_samples_per_second": 527.879, | |
| "eval_steps_per_second": 8.305, | |
| "step": 1665 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "grad_norm": 1.0767935514450073, | |
| "learning_rate": 2.258175206106183e-05, | |
| "loss": 0.0242, | |
| "step": 1998 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "eval_accuracy": 0.9048, | |
| "eval_loss": 0.39692988991737366, | |
| "eval_runtime": 6.9181, | |
| "eval_samples_per_second": 542.053, | |
| "eval_steps_per_second": 8.528, | |
| "step": 1998 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "grad_norm": 0.0011720191687345505, | |
| "learning_rate": 1.0588717719108745e-05, | |
| "loss": 0.0055, | |
| "step": 2331 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "eval_accuracy": 0.9136, | |
| "eval_loss": 0.3753560483455658, | |
| "eval_runtime": 7.0108, | |
| "eval_samples_per_second": 534.887, | |
| "eval_steps_per_second": 8.416, | |
| "step": 2331 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "grad_norm": 0.3251957893371582, | |
| "learning_rate": 2.7461918766693875e-06, | |
| "loss": 0.0013, | |
| "step": 2664 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "eval_accuracy": 0.9146666666666666, | |
| "eval_loss": 0.3700757920742035, | |
| "eval_runtime": 6.749, | |
| "eval_samples_per_second": 555.636, | |
| "eval_steps_per_second": 8.742, | |
| "step": 2664 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "grad_norm": 0.005504661705344915, | |
| "learning_rate": 9.889369636195689e-11, | |
| "loss": 0.0006, | |
| "step": 2997 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "eval_accuracy": 0.9197333333333333, | |
| "eval_loss": 0.35471659898757935, | |
| "eval_runtime": 7.3881, | |
| "eval_samples_per_second": 507.57, | |
| "eval_steps_per_second": 7.986, | |
| "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 | |
| } | |