| { | |
| "best_metric": 1.810072422027588, | |
| "best_model_checkpoint": "flower_groups_image_detection/checkpoint-3102", | |
| "epoch": 2.0, | |
| "eval_steps": 500, | |
| "global_step": 3102, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.3223726627981947, | |
| "grad_norm": 3.30501389503479, | |
| "learning_rate": 1.7051114023591087e-07, | |
| "loss": 1.7502, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.6447453255963894, | |
| "grad_norm": 3.4669442176818848, | |
| "learning_rate": 1.3774574049803407e-07, | |
| "loss": 1.7329, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.9671179883945842, | |
| "grad_norm": 5.2170257568359375, | |
| "learning_rate": 1.0498034076015727e-07, | |
| "loss": 1.7245, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.7796323173681664, | |
| "eval_loss": 1.8147186040878296, | |
| "eval_runtime": 321.6906, | |
| "eval_samples_per_second": 102.807, | |
| "eval_steps_per_second": 12.851, | |
| "step": 1551 | |
| }, | |
| { | |
| "epoch": 1.2894906511927788, | |
| "grad_norm": 2.947956085205078, | |
| "learning_rate": 7.221494102228047e-08, | |
| "loss": 1.7307, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 1.6118633139909737, | |
| "grad_norm": 3.766296863555908, | |
| "learning_rate": 3.944954128440367e-08, | |
| "loss": 1.7229, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 1.9342359767891684, | |
| "grad_norm": 3.5768988132476807, | |
| "learning_rate": 6.684141546526867e-09, | |
| "loss": 1.7354, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.7807813255926463, | |
| "eval_loss": 1.810072422027588, | |
| "eval_runtime": 325.6105, | |
| "eval_samples_per_second": 101.569, | |
| "eval_steps_per_second": 12.696, | |
| "step": 3102 | |
| } | |
| ], | |
| "logging_steps": 500, | |
| "max_steps": 3102, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 2, | |
| "save_steps": 500, | |
| "total_flos": 7.697230908500754e+18, | |
| "train_batch_size": 32, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |