Add model size metadata (as loaded, still need to convert vision backbone to fp32) ed68f67 verified SeanScripts commited on Sep 30, 2024