{ "epoch": 144.0, "learning_rate": 1.0000000000000002e-06, "total_flos": 1.3598709030716368e+20, "train_loss": 0.157796386979584, "train_runtime": 249885.5342, "train_samples_per_second": 5.232, "train_steps_per_second": 0.164 }