Codixor/vision-gguf

Codixor vision weights

Weights converted for rcumfy and its engines (codixor/vision.cpp, stable-diffusion.cpp). Converted without Python or PyTorch by vision.cpp/tools/convert-rs.

File	Source	Licence	Notes
CLIP-ViT-H-14-F16.gguf	clip_vision_h.safetensors (OpenCLIP ViT-H/14 LAION-2B, ComfyUI repackage)	MIT	vision.cpp clip arch; matches a NumPy reference to cosine 1.0000
photomaker-v2.safetensors	TencentARC/PhotoMaker-V2 photomaker-v2.bin	Apache-2.0	pickle read as data (never executed), tensors byte-identical, names in the stable-diffusion.cpp layout

PhotoMaker v2 additionally needs a face ID embedding at inference time.