{ "best_metric": null, "best_model_checkpoint": null, "epoch": 4.0, "eval_steps": 500, "global_step": 1312, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 1.0, "eval_gen_len": 10.608247422680412, "eval_loss": 0.6409556865692139, "eval_rouge1": 34.106, "eval_rouge2": 25.6511, "eval_rougeL": 30.5481, "eval_rougeLsum": 30.8868, "eval_runtime": 145.9508, "eval_samples_per_second": 1.994, "eval_steps_per_second": 0.254, "step": 328 }, { "epoch": 1.52, "learning_rate": 0.0003125, "loss": 0.7084, "step": 500 }, { "epoch": 2.0, "eval_gen_len": 10.608247422680412, "eval_loss": 0.6409556865692139, "eval_rouge1": 34.106, "eval_rouge2": 25.6511, "eval_rougeL": 30.5481, "eval_rougeLsum": 30.8868, "eval_runtime": 140.6442, "eval_samples_per_second": 2.069, "eval_steps_per_second": 0.263, "step": 656 }, { "epoch": 3.0, "eval_gen_len": 10.608247422680412, "eval_loss": 0.6409556865692139, "eval_rouge1": 34.106, "eval_rouge2": 25.6511, "eval_rougeL": 30.5481, "eval_rougeLsum": 30.8868, "eval_runtime": 134.1193, "eval_samples_per_second": 2.17, "eval_steps_per_second": 0.276, "step": 984 }, { "epoch": 3.05, "learning_rate": 0.0004853133981334999, "loss": 0.7129, "step": 1000 }, { "epoch": 4.0, "eval_gen_len": 10.608247422680412, "eval_loss": 0.6409556865692139, "eval_rouge1": 34.106, "eval_rouge2": 25.6511, "eval_rougeL": 30.5481, "eval_rougeLsum": 30.8868, "eval_runtime": 133.3173, "eval_samples_per_second": 2.183, "eval_steps_per_second": 0.278, "step": 1312 } ], "logging_steps": 500, "max_steps": 2624, "num_train_epochs": 8, "save_steps": 500, "total_flos": 9832568722292736.0, "trial_name": null, "trial_params": null }