Version 0.35.0 AutoencoderKL failed with PyTorch 2.3.1: AutoAttributeError: module 'torch.library' has no attribute 'custom_op'

[skedin]

Describe the bug

This should be an old issue, but it appears again in version 0.35.0, maybe also in 0.35.1.
Maybe need to protect the import again if PyTorch 2.3.1 is still in the supporting list?

Reproduction

from diffusers import AutoencoderKL

import AutoencoderKL will lead to : AutoAttributeError: module 'torch.library' has no attribute 'custom_op'

Logs

System Info

Diffusers: 0.35.0
PyTorch: 2.3.1

Who can help?

No response

[Aishwarya0811]

Hi! I'd like to work on this issue. I can reproduce the problem with PyTorch 2.3.1 and will investigate the import protection needed for backward compatibility.

[Harperrrr111]

same problem

[Aishwarya0811]

i submitted the PR if have time kindly review it it will be helpful contribution:
Added hasattr() checks before using these torch.library functions
Functions are only registered when available in the PyTorch version
Maintains backward compatibility with PyTorch 2.3.1 without breaking newer versions

[a-r-r-o-w]

@skedin I've merged @Aishwarya0811's fix for pytorch<2.4 compatibility just now. Let me know if you run into any other errors :)

[guyuezuntinggithub]

I have same question with pytorch 2.3.0 and diffuser 0.35.1