What does this PR do?

Fixes #14819

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?

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.

@stas00 @sgugger @TimDettmers

Status

• Need to instrument CI to install bnb (binary package so a bit trickier than normal dependency)
• I have implemented a CLI parameter to support bitsandbytes
• I did not write any documentation yet
• I followed @TimDettmers 's suggestion to override the embedding layers. However, I am unsure about a couple of things:
  • Does the override need to happen before the model is loaded onto the GPU as the official documentation describes for other overrides?
  • Are there any pitfalls to my current approach to identifying Embedding layers? It seems to work fine for RoBERTa and for GPT-2.
• So far, I've used run_mlm.py and run_clm.py from the examples directory to check that the code runs. Using RTX A6000 GPUs, I see