No LatentDiffusionPipeline on the latest branch

[kyleliang919]

Describe the bug

Latent diffusion is not found even after pip install the latestest dev branch

Reproduction

git clone https://github.com/huggingface/diffusers.git
cd diffusers && pip install -e .

then run

from diffusers import DiffusionPipeline
import torch

pipeline = DiffusionPipeline.from_pretrained("fusing/latent-diffusion-text2im-large")

Logs

Traceback (most recent call last):
  File "preprocessing.py", line 19, in <module>
    ldm = DiffusionPipeline.from_pretrained("fusing/latent-diffusion-text2im-large", use_auth_token=token)
  File "/scratch/kaizhaol/videogpt/diffusers/src/diffusers/pipeline_utils.py", line 433, in from_pretrained
    pipeline_class = getattr(diffusers_module, config_dict["_class_name"])
AttributeError: module 'diffusers' has no attribute 'LatentDiffusionPipeline'

System Info

diffuser version: '0.7.0.dev0'

[patrickvonplaten]

Hey @kyleliang919,

There is no LatentDiffusionPipeline indeed -> opened a PR here to improve the docs: #1102 and changed the old latent diffusion model config here: https://huggingface.co/fusing/latent-diffusion-text2im-large/commit/85cbe9699f34a107b12e0bc20455f62f712a398e

Note the official latent diffusion model is here: https://huggingface.co/CompVis/ldm-text2im-large-256