{ "epoch": 106.0, "learning_rate": 1.0000000000000002e-06, "total_flos": 1.3699618367216817e+21, "train_loss": 0.13672988786670365, "train_runtime": 251282.626, "train_samples_per_second": 5.203, "train_steps_per_second": 0.163 }