Request quantization of Qwen3.5 9B and Gemma4 12B
https://huggingface.co/Comfy-Org/Qwen3.5/tree/main/text_encoders
https://huggingface.co/Comfy-Org/gemma-4/tree/main/text_encoders
The official Comfy org release only includes the bf16 full model. fp8 and int8 versions are not provided, and the models I found on Hugging Face are incompatible with ComfyUI.
sure, here they are π
https://huggingface.co/Hippotes/Qwen3.5-ComfyUI-quants
https://huggingface.co/Hippotes/Gemma-4-ComfyUI-quants
definitely a smaller disk & memory footprint but I'm not sure how Comfy handle the text generation, I didn't observe a speed boost on my system.
sure, here they are π
https://huggingface.co/Hippotes/Qwen3.5-ComfyUI-quants
https://huggingface.co/Hippotes/Gemma-4-ComfyUI-quantsdefinitely a smaller disk & memory footprint but I'm not sure how Comfy handle the text generation, I didn't observe a speed boost on my system.
Greatly appreciate it m(_ _)m