{ "all_params": 60506624, "epoch": 3.0, "memory_footprint": 242026496, "total_flos": 356896330481664.0, "train_loss": 2.027634771303697, "train_runtime": 130.1895, "train_samples_per_second": 20.255, "train_steps_per_second": 2.535, "trainable_params": 60506624, "trainable_params_percent": 100.0 }