Since comfyui now natively supports int8 i was wondering if support could be added to these nodes as well. Int8 is faster on most gpus but i still end up using q8 quants becuase comfyui is terrible at offloading models and i get oom error unless i manually offload using these nodes.
Since comfyui now natively supports int8 i was wondering if support could be added to these nodes as well. Int8 is faster on most gpus but i still end up using q8 quants becuase comfyui is terrible at offloading models and i get oom error unless i manually offload using these nodes.