{ "epoch": 83.0, "learning_rate": 1.0000000000000002e-06, "total_flos": 8.603009036605255e+19, "train_loss": 0.45949580130708517, "train_runtime": 19431.3015, "train_samples_per_second": 54.06, "train_steps_per_second": 1.691 }