Add BridgeTower model

[abhiwand]

What does this PR do?

This PR implements a HuggingFace Transformers version of BridgeTower: Building Bridges Between Encoders in Vision-Language Representation Learning from the paper https://arxiv.org/abs/2206.08657.pdf

This paper has been accepted to https://aaai.org/Conferences/AAAI-23/

The model's pre-trained checkpoints and configurations have been released here:
https://huggingface.co/BridgeTower under:

• https://huggingface.co/BridgeTower/bridgetower-base-itm-mlm
• https://huggingface.co/BridgeTower/bridgetower-base

The following heads have been implemented:

• BridgeTowerForMaskedLM
• BridgeTowerForImageAndTextRetrieval

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,
  Pull Request section?
• 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?

@amyeroberts @NielsRogge @ArthurZucker could you please assist with review and feedback.

@philschmid

[NielsRogge]

Thanks a lot for working on this and addressing all comments!

There are still 2 comments which seem to be unaddressed, after that good for me to merge.

[tileintel]

@NielsRogge Our PR keeps failing at tests/pipelines/test_pipelines_automatic_speech_recognition.py::AutomaticSpeechRecognitionPipelineTests::test_return_timestamps_in_preprocess. Would you please help to see if it is because of BridgeTower or because of something else?
Thanks a lot

[amyeroberts]

@abhiwand @tileintel Thanks for address all of the comments! On Monday there were two PRs merged into main which added test_image_processing_common.py (#20785) and updated the feature extractor references in the test_image_processing_xxx.py files (#20768). Could you update test_image_processing_bridgetower.py to reflect these please?

[tileintel]

@amyeroberts We have updated test_image_processing_bridgetower.py as you suggested. Thanks for the suggestion.
@NielsRogge @amyeroberts @sgugger We have addressed all of the comments. Thanks a lot for helping us to review and approve this. We are very looking forward to having this PR merged into main soon.

[sgugger]

Thanks again for your contribution!

[tileintel]

@sgugger Thank you for merging this PR. May I ask when BridgeTower model will go to HuggingFace's production and what release is that?
Thanks

[sgugger]

The next release will be in a month roughly (given the fast last release was yesterday).

[tileintel]

Thank @sgugger for letting us know.