fix autopipeline for kolors img2img

[yiyixuxu]

I made a mistake in https://github.com/huggingface/diffusers/pull/9065/files#
fixing it!

sanity check

from diffusers import AutoPipelineForText2Image, AutoPipelineForImage2Image
import torch

repo = "Kwai-Kolors/Kolors-diffusers"

pipe_txt2img = AutoPipelineForText2Image.from_pretrained(repo, variant="fp16", torch_dtype=torch.float16).to("cuda")
assert pipe_txt2img.__class__.__name__ == "KolorsPipeline"

pipe_txt2img_pag = AutoPipelineForText2Image.from_pretrained(repo, torch_dtype=torch.float16, variant="fp16", enable_pag=True).to("cuda")
assert pipe_txt2img_pag.__class__.__name__ == "KolorsPAGPipeline"

pipe_img2img = AutoPipelineForImage2Image.from_pretrained(repo, variant="fp16", torch_dtype=torch.float16).to("cuda")
assert pipe_img2img.__class__.__name__ == "KolorsImg2ImgPipeline"

[HuggingFaceDocBuilderDev]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

[asomoza]

thanks, LGTM!