| { | |
| "best_global_step": 1812, | |
| "best_metric": 0.9940354824066162, | |
| "best_model_checkpoint": "siglip2-finetune-full/checkpoint-1812", | |
| "epoch": 3.0, | |
| "eval_steps": 500, | |
| "global_step": 1812, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.8278145695364238, | |
| "grad_norm": 61.945621490478516, | |
| "learning_rate": 1.6196111580726966e-06, | |
| "loss": 4.0193, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.576952943003572, | |
| "eval_loss": 2.36920166015625, | |
| "eval_model_preparation_time": 0.0024, | |
| "eval_runtime": 285.5037, | |
| "eval_samples_per_second": 67.659, | |
| "eval_steps_per_second": 8.459, | |
| "step": 604 | |
| }, | |
| { | |
| "epoch": 1.6556291390728477, | |
| "grad_norm": 75.25294494628906, | |
| "learning_rate": 1.1969568892645817e-06, | |
| "loss": 2.0696, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.7819019516488067, | |
| "eval_loss": 1.2963489294052124, | |
| "eval_model_preparation_time": 0.0024, | |
| "eval_runtime": 287.0467, | |
| "eval_samples_per_second": 67.296, | |
| "eval_steps_per_second": 8.413, | |
| "step": 1208 | |
| }, | |
| { | |
| "epoch": 2.4834437086092715, | |
| "grad_norm": 72.18843078613281, | |
| "learning_rate": 7.743026204564665e-07, | |
| "loss": 1.3631, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.8309261272454315, | |
| "eval_loss": 0.9940354824066162, | |
| "eval_model_preparation_time": 0.0024, | |
| "eval_runtime": 286.8363, | |
| "eval_samples_per_second": 67.345, | |
| "eval_steps_per_second": 8.419, | |
| "step": 1812 | |
| } | |
| ], | |
| "logging_steps": 500, | |
| "max_steps": 2416, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 4, | |
| "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": 4.858705622631813e+18, | |
| "train_batch_size": 32, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |