{ "best_metric": null, "best_model_checkpoint": null, "epoch": 2.0, "global_step": 866, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.46, "eval_HasAns_exact": 30.76923076923077, "eval_HasAns_f1": 47.84025570684176, "eval_HasAns_total": 221, "eval_best_exact": 30.76923076923077, "eval_best_exact_thresh": 0.0, "eval_best_f1": 47.84025570684176, "eval_best_f1_thresh": 0.0, "eval_exact": 30.76923076923077, "eval_f1": 47.84025570684176, "eval_total": 221, "step": 200 }, { "epoch": 0.92, "eval_HasAns_exact": 35.294117647058826, "eval_HasAns_f1": 55.643010655276335, "eval_HasAns_total": 221, "eval_best_exact": 35.294117647058826, "eval_best_exact_thresh": 0.0, "eval_best_f1": 55.643010655276335, "eval_best_f1_thresh": 0.0, "eval_exact": 35.294117647058826, "eval_f1": 55.643010655276335, "eval_total": 221, "step": 400 }, { "epoch": 1.15, "learning_rate": 1.0578034682080927e-05, "loss": 1.0653, "step": 500 }, { "epoch": 1.39, "eval_HasAns_exact": 37.55656108597285, "eval_HasAns_f1": 57.33395418199089, "eval_HasAns_total": 221, "eval_best_exact": 37.55656108597285, "eval_best_exact_thresh": 0.0, "eval_best_f1": 57.33395418199089, "eval_best_f1_thresh": 0.0, "eval_exact": 37.55656108597285, "eval_f1": 57.33395418199089, "eval_total": 221, "step": 600 }, { "epoch": 1.85, "eval_HasAns_exact": 38.914027149321264, "eval_HasAns_f1": 59.19584379604189, "eval_HasAns_total": 221, "eval_best_exact": 38.914027149321264, "eval_best_exact_thresh": 0.0, "eval_best_f1": 59.19584379604189, "eval_best_f1_thresh": 0.0, "eval_exact": 38.914027149321264, "eval_f1": 59.19584379604189, "eval_total": 221, "step": 800 }, { "epoch": 2.0, "step": 866, "total_flos": 1357306002865152.0, "train_loss": 0.8789484495378806, "train_runtime": 413.3455, "train_samples_per_second": 16.756, "train_steps_per_second": 2.095 } ], "max_steps": 866, "num_train_epochs": 2, "total_flos": 1357306002865152.0, "trial_name": null, "trial_params": null }