{ "best_metric": null, "best_model_checkpoint": null, "epoch": 1.0, "global_step": 1130, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.09, "learning_rate": 9.115044247787611e-05, "loss": 1.0612, "step": 100 }, { "epoch": 0.18, "learning_rate": 8.230088495575221e-05, "loss": 0.93, "step": 200 }, { "epoch": 0.27, "learning_rate": 7.345132743362832e-05, "loss": 0.9098, "step": 300 }, { "epoch": 0.35, "learning_rate": 6.460176991150442e-05, "loss": 0.9033, "step": 400 }, { "epoch": 0.44, "learning_rate": 5.575221238938053e-05, "loss": 0.8943, "step": 500 }, { "epoch": 0.53, "learning_rate": 4.690265486725664e-05, "loss": 0.8932, "step": 600 }, { "epoch": 0.62, "learning_rate": 3.8053097345132744e-05, "loss": 0.9019, "step": 700 }, { "epoch": 0.71, "learning_rate": 2.9203539823008852e-05, "loss": 0.8881, "step": 800 }, { "epoch": 0.8, "learning_rate": 2.0353982300884957e-05, "loss": 0.8929, "step": 900 }, { "epoch": 0.88, "learning_rate": 1.1504424778761062e-05, "loss": 0.8916, "step": 1000 }, { "epoch": 0.97, "learning_rate": 2.6548672566371683e-06, "loss": 0.8882, "step": 1100 } ], "max_steps": 1130, "num_train_epochs": 1, "total_flos": 1.8342786143512166e+17, "trial_name": null, "trial_params": null }