[Single File Loading] Handle unexpected keys in CLIP models when accelerate isn't installed.

[DN6]

What does this PR do?

Fixes #8578

Since transformers>=4.31.0, the key text_model.embeddings.position_id has been made non-persistent in the CLIP models. This can cause an issue when loading older versions of the checkpoint with single file when accelerate isn't installed, because we rely on strict loading with model.load_state_dict in this case.

This PR

1. Changes the model.load_state_dict to not use strict loading, so that we can raise a warning about unexpected keys when accelerate isn't installed.

Fixes # (issue)

Before submitting

• This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case).
• Did you read the contributor guideline?
• Did you read our philosophy doc (important for complex PRs)?
• Was this discussed/approved via a GitHub issue or the forum? Please add a link to it if that's the case.
• Did you make sure to update the documentation with your changes? Here are the
  documentation guidelines, and
  here are tips on formatting docstrings.
• Did you write any new necessary tests?

Who can review?

Anyone in the community is free to review the PR once the tests have passed. Feel free to tag
members/contributors who may be interested in your PR.

[HuggingFaceDocBuilderDev]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

[yiyixuxu]

ok but is this not used text_model.embeddings.position_id in the model?

[DN6]

No. Since transformers 4.31.0, the buffer is non-persistent. So it's no longer in the state dict.

Since text_model.embeddings.position_id is just a list of integers, it's create dynamically on based on the model config.
https://github.com/huggingface/transformers/blob/dcdda5324bcc7a750b5e40e11dd795442204ff27/src/transformers/models/clip/modeling_clip.py#L203

[yiyixuxu]

thanks!