Using the zeroscope_v2_XL in diffusers
It is easy to use the base model in diffusers, but I'm unsure how to use the upscaler in diffusers.
From what I can tell the vid2vid process of the extension is fairly different from the typically img2img process (adding noise via the schedule).
Has anyone got the model to work in diffusers?
This notebook from AI Lost Media seems to be able to do vid2vid with diffusers, but I have heard that its not as optimized as the implementation in the text2video extension for 1111. https://colab.research.google.com/drive/1TsZmatSu1-1lNBeOqz3_9Zq5P2c0xTTq
Thanks I'll check it out
Hey guys I'm trying to run the model using diffusers but getting this error
1 from diffusers import DiffusionPipeline
----> 3 pipeline = DiffusionPipeline.from_pretrained("cerspense/zeroscope_v2_XL")
File c:\Users\vinay\Desktop\text-video.conda\lib\site-packages\diffusers\pipelines\pipeline_utils.py:908, in DiffusionPipeline.from_pretrained(cls, pretrained_model_name_or_path, **kwargs)
905 # 1. Download the checkpoints and configs
906 # use snapshot download here to get it working from from_pretrained
907 if not os.path.isdir(pretrained_model_name_or_path):
--> 908 cached_folder = cls.download(
909 pretrained_model_name_or_path,
910 cache_dir=cache_dir,
911 resume_download=resume_download,
912 force_download=force_download,
913 proxies=proxies,
914 local_files_only=local_files_only,
915 use_auth_token=use_auth_token,
916 revision=revision,
917 from_flax=from_flax,
918 use_safetensors=use_safetensors,
919 custom_pipeline=custom_pipeline,
920 custom_revision=custom_revision,
921 variant=variant,
922 **kwargs,
...
341 return class_obj
343 diffusers_module = importlib.import_module(class_obj.module.split(".")[0])
--> 344 return getattr(diffusers_module, config["_class_name"])
AttributeError: module 'diffusers' has no attribute 'VideoToVideoSDPipeline'
@vinaysingh8866 I bet you just need the latest diffusers that is in main. Looks like the model was just updated.
Getting the same error
