{ "best_metric": null, "best_model_checkpoint": null, "epoch": 10.0, "global_step": 62500, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 1.0, "learning_rate": 1.9333333333333333e-05, "loss": 5.4183, "step": 6250 }, { "epoch": 2.0, "learning_rate": 1.866688e-05, "loss": 4.1174, "step": 12500 }, { "epoch": 3.0, "learning_rate": 1.8000533333333335e-05, "loss": 3.5049, "step": 18750 }, { "epoch": 4.0, "learning_rate": 1.7334080000000003e-05, "loss": 3.0671, "step": 25000 }, { "epoch": 5.0, "learning_rate": 1.6667626666666667e-05, "loss": 2.7197, "step": 31250 }, { "epoch": 6.0, "learning_rate": 1.600128e-05, "loss": 2.4474, "step": 37500 }, { "epoch": 7.0, "learning_rate": 1.533482666666667e-05, "loss": 2.2439, "step": 43750 }, { "epoch": 8.0, "learning_rate": 1.4668373333333334e-05, "loss": 2.091, "step": 50000 }, { "epoch": 9.0, "learning_rate": 1.4001920000000002e-05, "loss": 1.9731, "step": 56250 }, { "epoch": 10.0, "learning_rate": 1.3335360000000002e-05, "loss": 1.8787, "step": 62500 } ], "max_steps": 187500, "num_train_epochs": 30, "total_flos": 1.1680685027544269e+17, "trial_name": null, "trial_params": null }