splite_mode, pipe_id for "RunDiffusion/Juggernaut-X-v10"

[ChuRuaNh0]

No description provided.

[ChuRuaNh0]

Thank for your repo, It's so great, Can you support for RunDiffusion/Juggernaut-X-v10. I appreciate your help, thanks

[czg1225]

Hi @ChuRuaNh0 ,
RunDiffusion/Juggernaut-X-v10 is supported now. You can try it following the simple usage example:

import torch
from diffusers import StableDiffusionXLPipeline
from src.async_sd import AsyncDiff

pipeline = StableDiffusionXLPipeline.from_pretrained("RunDiffusion/Juggernaut-X-v10", torch_dtype=torch.float16)

async_diff = AsyncDiff(pipeline, model_n=2, stride=1, time_shift=False)

async_diff.reset_state(warm_up=3)
image = pipeline(<prompts>).images[0]
if dist.get_rank() == 0:
  image.save(f"output.jpg")

[ChuRuaNh0]

@czg1225 I have tried that you changed pipeline config the same sdxl-base-1.0 but it didn't work properly

[czg1225]

@ChuRuaNh0 , Could you provide more details about your problem? I had no problems accelerating both sdxl-base-1.0 and RunDiffusion/Juggernaut-X-v10