| { | |
| "best_metric": 0.3888, | |
| "best_model_checkpoint": "./vit_finetuned_models_dataset/CIFAR100/50_from_100/facebook_dino-vitb16/model_idx_0960/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": 24.418479919433594, | |
| "learning_rate": 0.0005, | |
| "loss": 3.3674, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.20746666666666666, | |
| "eval_loss": 2.9523427486419678, | |
| "eval_runtime": 11.0179, | |
| "eval_samples_per_second": 340.355, | |
| "eval_steps_per_second": 5.355, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "grad_norm": 16.42314910888672, | |
| "learning_rate": 0.0005, | |
| "loss": 2.8616, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.2597333333333333, | |
| "eval_loss": 2.748380184173584, | |
| "eval_runtime": 11.1215, | |
| "eval_samples_per_second": 337.185, | |
| "eval_steps_per_second": 5.305, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "grad_norm": 52.7269287109375, | |
| "learning_rate": 0.0005, | |
| "loss": 2.6733, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.2773333333333333, | |
| "eval_loss": 2.6724729537963867, | |
| "eval_runtime": 10.7866, | |
| "eval_samples_per_second": 347.653, | |
| "eval_steps_per_second": 5.47, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "grad_norm": 15.847335815429688, | |
| "learning_rate": 0.0005, | |
| "loss": 2.5544, | |
| "step": 1332 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_accuracy": 0.2853333333333333, | |
| "eval_loss": 2.653170585632324, | |
| "eval_runtime": 11.0472, | |
| "eval_samples_per_second": 339.453, | |
| "eval_steps_per_second": 5.341, | |
| "step": 1332 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "grad_norm": 13.500426292419434, | |
| "learning_rate": 0.0005, | |
| "loss": 2.4412, | |
| "step": 1665 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_accuracy": 0.32, | |
| "eval_loss": 2.4827098846435547, | |
| "eval_runtime": 11.1189, | |
| "eval_samples_per_second": 337.262, | |
| "eval_steps_per_second": 5.306, | |
| "step": 1665 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "grad_norm": 12.111242294311523, | |
| "learning_rate": 0.0005, | |
| "loss": 2.3218, | |
| "step": 1998 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "eval_accuracy": 0.3536, | |
| "eval_loss": 2.397001028060913, | |
| "eval_runtime": 11.5484, | |
| "eval_samples_per_second": 324.721, | |
| "eval_steps_per_second": 5.109, | |
| "step": 1998 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "grad_norm": 12.104255676269531, | |
| "learning_rate": 0.0005, | |
| "loss": 2.2262, | |
| "step": 2331 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "eval_accuracy": 0.36506666666666665, | |
| "eval_loss": 2.3168785572052, | |
| "eval_runtime": 11.1092, | |
| "eval_samples_per_second": 337.559, | |
| "eval_steps_per_second": 5.311, | |
| "step": 2331 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "grad_norm": 17.220958709716797, | |
| "learning_rate": 0.0005, | |
| "loss": 2.1286, | |
| "step": 2664 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "eval_accuracy": 0.38106666666666666, | |
| "eval_loss": 2.244141101837158, | |
| "eval_runtime": 11.2447, | |
| "eval_samples_per_second": 333.492, | |
| "eval_steps_per_second": 5.247, | |
| "step": 2664 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "grad_norm": 12.500649452209473, | |
| "learning_rate": 0.0005, | |
| "loss": 2.0258, | |
| "step": 2997 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "eval_accuracy": 0.3888, | |
| "eval_loss": 2.2205913066864014, | |
| "eval_runtime": 11.0502, | |
| "eval_samples_per_second": 339.362, | |
| "eval_steps_per_second": 5.339, | |
| "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 | |
| } | |