{ "best_metric": null, "best_model_checkpoint": null, "epoch": 3.0, "global_step": 2193, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.68, "learning_rate": 3.860009119927041e-05, "loss": 1.9422, "step": 500 }, { "epoch": 1.37, "learning_rate": 2.7200182398540814e-05, "loss": 1.5003, "step": 1000 }, { "epoch": 2.05, "learning_rate": 1.580027359781122e-05, "loss": 1.3233, "step": 1500 }, { "epoch": 2.74, "learning_rate": 4.400364797081624e-06, "loss": 1.1264, "step": 2000 }, { "epoch": 3.0, "step": 2193, "train_runtime": 3680.7884, "train_samples_per_second": 0.596 } ], "max_steps": 2193, "num_train_epochs": 3, "total_flos": 9168905732313600, "trial_name": null, "trial_params": null }