single_file_utils.py#load_single_file_model_checkpoint bugs

[knoopx]

Describe the bug

Loading a checkpoint from file crashes as it tries to print a warning with unexpected keys because is accessing __name__ on incorrect target (instance instead of class)

https://github.com/huggingface/diffusers/blob/d50baf0c632342b9576a24352244c4235ce8b875/src/diffusers/loaders/single_file_utils.py

f"Some weights of the model checkpoint were not used when initializing {vae.__name__} f"Some weights of the model checkpoint were not used when initializing {unet.__name__} f"Some weights of the model checkpoint were not used when initializing {controlnet.__name__}

should be:

[...].__class__.__name__

Reproduction

load a single file checkpoint with extra keys

Logs

No response

System Info

• diffusers version: 0.27.2
• Platform: Linux-6.8.8-1-clear-x86_64-with-glibc2.39
• Python version: 3.11.0
• PyTorch version (GPU?): 2.2.0+cu121 (True)
• Huggingface_hub version: 0.23.0
• Transformers version: 4.40.1
• Accelerate version: 0.30.0

Who can help?

@sayakpaul

[sayakpaul]

Please provide a reproducible snippet.

Cc: @DN6

[knoopx]

@sayakpaul

Ran:

StableDiffusionPipeline.from_single_file("checkpoint-with-extra-keys.safetensor").

Got:

  File ".venv/lib/python3.11/site-packages/huggingface_hub/utils/_validators.py", line 114, in _inner_fn
    return fn(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^
  File ".venv/lib/python3.11/site-packages/diffusers/loaders/single_file.py", line 289, in from_single_file
    components = build_sub_model_components(
                 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File ".venv/lib/python3.11/site-packages/diffusers/loaders/single_file.py", line 75, in build_sub_model_components
    vae_components = create_diffusers_vae_model_from_ldm(
                     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File ".venv/lib/python3.11/site-packages/diffusers/loaders/single_file_utils.py", line 1391, in create_diffusers_vae_model_from_ldm
    f"Some weights of the model checkpoint were not used when initializing {vae.__name__}: \n {[', '.join(unexpected_keys)]}"
                                                                            ^^^^^^^^^^^^
  File ".venv/lib/python3.11/site-packages/diffusers/models/modeling_utils.py", line 218, in __getattr__
    return super().__getattr__(name)
           ^^^^^^^^^^^^^^^^^^^^^^^^^
  File ".venv/lib/python3.11/site-packages/torch/nn/modules/module.py", line 1688, in __getattr__
    raise AttributeError(f"'{type(self).__name__}' object has no attribute '{name}'")
AttributeError: 'AutoencoderKL' object has no attribute '__name__'

Expected:

Some weights of the model checkpoint were not used when initializing AutoencoderKL: [...]

[sayakpaul]

This is still not fully reproducible.

[yiyixuxu]

@knoopx we just merged this PR https://github.com/huggingface/diffusers/pull/7496 can you update the diffusers and try again?

if still not working, please provide a reproducible code example with inputs and checkpoints that accessible to us

[github-actions[bot]]

This issue has been automatically marked as stale because it has not had recent activity. If you think this still needs to be addressed please comment on this thread.

Please note that issues that do not follow the contributing guidelines are likely to be ignored.