{ "all_params": 8294881792, "memory_footprint": 16596783440, "total_flos": 6.808102961998848e+16, "train_loss": 0.33265126400034534, "train_runtime": 4751.8019, "train_samples_per_second": 0.67, "train_steps_per_second": 0.083, "trainable_params": 3506176, "trainable_params_percent": 0.042269149674700995 }