The latest ComfyUI throws an error when loading the model natively
#3
by justpretending - opened
ComfyUI now has native int8 support, but Krea2-Turbo-int8-ConvRot.safetensors throws ValueError: Unknown quantization format for layer blocks.0.attn.wq.
There seems to be no format key in {'convrot': True, 'convrot_groupsize': 256, 'per_row': True} that the Load Diffusion Model node expects.
ComfyUI is at commit 7cb784e0f48784bb6ed588912e186e5ee1e9ee68
comfy-kitchen 0.2.12
The int8 convrot model from https://huggingface.co/Winnougan/Krea-2-Base-Turbo-NVFP4-FP8-INT8 works fine.