{ "best_metric": null, "best_model_checkpoint": null, "epoch": 50.0, "global_step": 2800, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 8.93, "learning_rate": 4.107142857142857e-05, "loss": 2.5637, "step": 500 }, { "epoch": 17.86, "learning_rate": 3.2142857142857144e-05, "loss": 2.2877, "step": 1000 }, { "epoch": 26.79, "learning_rate": 2.3214285714285715e-05, "loss": 2.1856, "step": 1500 }, { "epoch": 35.71, "learning_rate": 1.4285714285714285e-05, "loss": 2.1336, "step": 2000 }, { "epoch": 44.64, "learning_rate": 5.357142857142857e-06, "loss": 2.1016, "step": 2500 }, { "epoch": 50.0, "step": 2800, "total_flos": 1.4831050824096154e+17, "train_loss": 2.2369551304408484, "train_runtime": 35299.2341, "train_samples_per_second": 31.724, "train_steps_per_second": 0.079 } ], "max_steps": 2800, "num_train_epochs": 50, "total_flos": 1.4831050824096154e+17, "trial_name": null, "trial_params": null }