Add VQGAN-CLIP research project

[ErwannMillon]

What does this PR do?

Implements VQGAN-CLIP using huggingface CLIP models

Related to #21064

This Research Project allows users to generate or edit images with a single line of code. It wraps the huggingface CLIPProcessor class, allowing images to be processed as torch tensors in order to preserve gradient flow through the transformations.

Features:

• Positive and negative prompts
• Multiple prompts
• Prompt Weights
• Creating GIF animations of the transformations
• Wandb logging

Tagging @amyeroberts for review :)

[HuggingFaceDocBuilderDev]

The documentation is not available anymore as the PR was closed or merged.

[amyeroberts]

Nice - looking good! Thanks for adding this example.

I've made a few comments, mostly nits. None are critical and I'll leave it to you to decide what you do or don't want to change.

One thing to update is the face.jpeg image. To avoid adding lots of data to the repo, could you do the same as in other examples, and upload the image to a dataset where it can be display through hrefs and downloaded from e.g.

transformers/examples/research_projects/codeparrot/README.md

Line 3 in 42b60f8

<img src="https://huggingface.co/datasets/lvwerra/repo-images/raw/main/code-highlighting-streamlit.png" alt="drawing" width="350"/>

?

[ErwannMillon]

Hi there,
Thanks for the feedback, I fixed the style issues and removed the face image.
Also refactored the code to have more accurate function names, used just the tokenizer, changed the assertions to exceptions, and removed some extraneous code (eg double crop, freeze_module)
Have a great day :)
Erwann

[amyeroberts]

@ErwannMillon - thanks for the updates. It's looking good 😎 !

Just two last things before I think we're ready to merge:

• Removing the face.jpg file and instead pointing to a place where it can be downloaded
• Could you also remove the notebook? Like the image, we want to avoid adding large files as much as possible. Happy for you to link to e.g. a colab which shows this demo in the README.

[amyeroberts]

Looks good to merge from my end :)

[sgugger]

Thanks for working on this! Can we just remove the empty folder model_checkpoints from the PR?