{ "all_params": 222903552, "epoch": 3.0, "memory_footprint": 891614208, "total_flos": 2484005840363520.0, "train_loss": 1.4827006761774872, "train_runtime": 507.3071, "train_samples_per_second": 5.198, "train_steps_per_second": 1.301, "trainable_params": 222903552, "trainable_params_percent": 100.0 }