{ "all_params": 236932352, "epoch": 3.0, "memory_footprint": 947729408, "total_flos": 2659801069854720.0, "train_loss": 1.8849294849868976, "train_runtime": 574.0732, "train_samples_per_second": 4.593, "train_steps_per_second": 2.299, "trainable_params": 14028800, "trainable_params_percent": 5.92101495704563 }