{ "ewc_loss": 0.0810546875, "ewc_loss_diag": 0.0003795623779296875, "ewc_loss_parallel": 4.3392181396484375e-05, "total_flos": 5.156325133756924e+19, "train_loss": 0.44972702040424817, "train_runtime": 50365.0092, "train_samples": 125770, "train_samples_per_second": 7.492, "train_steps_per_second": 0.468 }