Frank (Haofan) Wang
Frank (Haofan) Wang
@zhuofengli Try this https://github.com/haofanwang/Lora-for-Diffusers/blob/main/format_convert.py#L161. I'm not sure whether it works, but the converting is fine. `pipe.unet.load_attn_procs()` only works on UNet in diffusers, while .safetensors files from civitai are usually of...
This is the limitation of fixed CLIP text encoder. Try a more compact prompt.
It doesn't matter. The Resampler module will project the original face embedding into specified dimension and num_tokens.
No, there is no text_emb in ControlNet. As said in our paper, we replace text_emb with face_emb.
You can download manually.
SDXL pipeline does have safety checker.
如果人脸检测失败，可以尝试padding，若仍然无法检测到，意味后续也无法提取人脸特征，建议更换图片。
Have the same issue.
Sure. Could you pull a request? On Sat, May 13, 2023 at 3:51 AM Haomiao Ni ***@***.***> wrote: > Hi, @haofanwang , thanks a lot for your > efforts to...
It's related to diffusers version. A newer version has some differences.