Closed aycaecemgul closed 1 month ago
I reduced the num_samples to 1 and repeated face embeds instead and it worked!
face_embeds = face_embedding.repeat(num_samples, 1, 1)
pipeline.generate(
prompt=positive_prompts,
negative_prompt=negative_prompts,
faceid_embeds=face_embeds,
shortcut=True,
num_samples=1,
seed=seed_list,
width=512,
height=768,
s_scale=1.0,
num_inference_steps=40,
)
Do you think we implement this inside repo? @xiaohu2015
hi, you can make a PR?
sure I'll look into it
The error I am getting:
I tried giving embeds as a list but did not work, should I concat the embed 4 times? I think the Ip adapter pipeline should handle this.