search

tencent-ailab / IP-Adapter

The image prompt adapter is designed to enable a pretrained text-to-image diffusion model to generate images with image prompt.
Apache License 2.0
4.46k stars 289 forks source link

Model Selection Issue: Using Different IP-Adapter Models in SDWebUI ControlNet and Diffusers #388

Open xddun opened 2 weeks ago

xddun commented 2 weeks ago

Thank you for this project. I have a question that has been puzzling me for a long time. Is there a difference between the ip-adapter-faceid-plusv2_sdxl_lora.safetensors model and the ip-adapter-faceid-plusv2_sdxl.bin model? I see that sdwebui contorlnet recommends using both, while the official diffusers code only uses your lora, and you only use the bin file in your huggingface.

Here are some links for reference: sdwebui contorlnet: https://github.com/Mikubill/sd-webui-controlnet/discussions/2442 Official diffusers code: https://huggingface.co/docs/diffusers/main/using-diffusers/ip_adapter Your personal huggingface: https://huggingface.co/h94/IP-Adapter-FaceID

This issue is really confusing to me, and I hope you can clarify which is the best option?

xiaohu2015 commented 1 week ago

ip-adapter-faceid-plusv2_sdxl.bin includes the adapter and lora, but ip-adapter-faceid-plusv2_sdxl_lora.safetensors is the lora part of the model