Support peft package for loading lora-adapted models

[moghadas76]

What does this PR do?

Support peft package for lora-based finetuning to be consistent with Huggin face

PR review

Anyone in the community is welcome to review the PR.
Before you start reviewing, make sure you have read the review guidelines. In short, see the following bullet-list:

Reviewer checklist

• Is this pull request ready for review? (if not, please submit in draft mode)
• Check that all items from Before submitting are resolved
• Make sure the title is self-explanatory and the description concisely explains the PR
• Add labels and milestones (and optionally projects) to the PR so it can be classified

---

📚 Documentation preview 📚: https://pytorch-lightning--19546.org.readthedocs.build/en/19546/

[moghadas76]

https://github.com/huggingface/transformers/blob/v4.38.1/src/transformers/modeling_utils.py#L2578

[awaelchli]

Hey @moghadas76

Thanks for opening a PR. These changes don't fit very well into Lightning, because Lightning doesn't make assumptions about the model architecture and composition and so a custom loading logic like this is very likely not in scope. Would you mind opening an issue with a description of how you'd like to do peft finetuning in Lightning? I'm sure we can find a way to make this possible in your LightningModule without needing to change the core saving and loading logic.