{ "best_metric": 2.4942777156829834, "best_model_checkpoint": "output/linkin-park/checkpoint-69", "epoch": 1.0, "global_step": 69, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.07, "learning_rate": 0.0001354300570714007, "loss": 2.895, "step": 5 }, { "epoch": 0.14, "learning_rate": 0.00013021156058794955, "loss": 2.9711, "step": 10 }, { "epoch": 0.22, "learning_rate": 0.0001218137945423232, "loss": 3.0184, "step": 15 }, { "epoch": 0.29, "learning_rate": 0.00011067009906946294, "loss": 2.635, "step": 20 }, { "epoch": 0.36, "learning_rate": 9.735550930058317e-05, "loss": 2.6937, "step": 25 }, { "epoch": 0.43, "learning_rate": 8.255708249541069e-05, "loss": 2.9303, "step": 30 }, { "epoch": 0.51, "learning_rate": 6.703844462707114e-05, "loss": 2.5872, "step": 35 }, { "epoch": 0.58, "learning_rate": 5.160038588088708e-05, "loss": 2.5982, "step": 40 }, { "epoch": 0.65, "learning_rate": 3.703953841164296e-05, "loss": 2.5227, "step": 45 }, { "epoch": 0.72, "learning_rate": 2.410726866285494e-05, "loss": 2.6876, "step": 50 }, { "epoch": 0.8, "learning_rate": 1.3470905479788493e-05, "loss": 2.8587, "step": 55 }, { "epoch": 0.87, "learning_rate": 5.679304716725937e-06, "loss": 2.566, "step": 60 }, { "epoch": 0.94, "learning_rate": 1.1345272679252971e-06, "loss": 2.6052, "step": 65 }, { "epoch": 1.0, "eval_loss": 2.4942777156829834, "eval_runtime": 4.673, "eval_samples_per_second": 22.041, "eval_steps_per_second": 2.782, "step": 69 } ], "max_steps": 69, "num_train_epochs": 1, "total_flos": 71724662784000.0, "trial_name": null, "trial_params": null }