[Custom Models] Can't load dreambooth models with new scheduling config method and DPMSolver

Describe the bug

I can’t load DPMSolver, for example, with the new method to define schedulers from config, cause apparently, most these dreambooth models don’t have a config for the schedulers. What do we do in this instance then?

OSError: nitrosocke/redshift-diffusion does not appear to have a file named scheduler_config.json.

Reproduction

Load DPMSolver with Dreambooth model like nitrosocke/redshift-diffusion and pass the pipe.scheduler.config to it.

Logs

~\AppData\Roaming\Python\Python39\site-packages\diffusers\configuration_utils.py in from_config(cls, config, return_unused_kwargs, **kwargs)
    194                 )
    195             deprecate("config-passed-as-path", "1.0.0", deprecation_message, standard_warn=False)
--> 196             config, kwargs = cls.load_config(pretrained_model_name_or_path=config, return_unused_kwargs=True, **kwargs)
    197 
    198         init_dict, unused_kwargs, hidden_dict = cls.extract_init_dict(config, **kwargs)

~\AppData\Roaming\Python\Python39\site-packages\diffusers\configuration_utils.py in load_config(cls, pretrained_model_name_or_path, return_unused_kwargs, **kwargs)
    352                 )
    353             except EntryNotFoundError:
--> 354                 raise EnvironmentError(
    355                     f"{pretrained_model_name_or_path} does not appear to have a file named {cls.config_name}."
    356                 )

OSError: nitrosocke/redshift-diffusion does not appear to have a file named scheduler_config.json.

System Info

diffusers-cli env doesn’t seem to work when installed with Conda (this is a fresh install of diffusers as of this morning with conda install method)

diffusers-cli is not recognized as an internal or external command,
operable program or batch file.

Guess that’s yet another bug.

[anton-l]

diffusers-cli env doesn’t seem to work when installed with Conda

pinging our conda contributor @sugatoray for a way to enable CLI commands in conda packages 😃

[averad]

@WASasquatch what would you suggest 🤗 change to correct the reported issue? Were you able to hack together a workaround?