{ "epoch": 62.0, "learning_rate": 1.0000000000000002e-06, "total_flos": 6.42634409963284e+19, "train_loss": 0.3985773164651095, "train_runtime": 16834.9641, "train_samples_per_second": 62.397, "train_steps_per_second": 3.903 }