{ "epoch": 62.0, "learning_rate": 1.0000000000000002e-07, "total_flos": 9.42369297866869e+19, "train_loss": 0.4754439868851833, "train_runtime": 8961.4221, "train_samples_per_second": 235.894, "train_steps_per_second": 3.699 }