{ "epoch": 105.0, "learning_rate": 1.0000000000000004e-08, "total_flos": 5.049640374682393e+21, "train_loss": 0.023157235795491324, "train_runtime": 62002.1626, "train_samples_per_second": 21.086, "train_steps_per_second": 0.66 }