Can you please tell me how to reproduce the metrics in the article, I don't really know how to use diffusers framework for inference on large datasets. Can the author provide the code for this?