Fix compatibility for bitsandbytes>=0.46

[matthewdouglas]

What does this PR do?

This PR improves compatibility with bitsandbytes.

In bitsandbytes v0.45.0, we introduced bitsandbytes.functional.int8_vectorwise_quant and bitsandbytes.functional.int8_double_quant as replacements for bitsandbytes.functional.double_quant. At this time we marked the old double_quant function as deprecated, and we then removed it in v0.46.0.

Fixes #20926

Before submitting

• Was this discussed/agreed via a GitHub issue? (not for typos and docs)
• Did you read the contributor guideline, Pull Request section?
• Did you make sure your PR does only one thing, instead of bundling different changes together?
• Did you make sure to update the documentation with your changes? (if necessary)
• Did you write any new necessary tests? (not for typos and docs)
• Did you verify new and existing tests pass locally with your changes?
• Did you list all the breaking changes introduced by this pull request?
• Did you update the CHANGELOG? (not for typos, docs, test updates, or minor internal changes/refactors)

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--20956.org.readthedocs.build/en/20956/

[Borda]

wondering how we can validate it since the GPU for some reason installed only bitsandbytes==0.45.5

[codecov]

Codecov Report

❌ Patch coverage is 0% with 2 lines in your changes missing coverage. Please review.
✅ Project coverage is 87%. Comparing base (afa7d56) to head (63ecb07).
⚠️ Report is 193 commits behind head on master.

Additional details and impacted files

@@           Coverage Diff           @@
##           master   #20956   +/-   ##
=======================================
  Coverage      87%      87%           
=======================================
  Files         268      268           
  Lines       23460    23460           
=======================================
  Hits        20402    20402           
  Misses       3058     3058           

[matthewdouglas]

@Borda It looks like there's a docker image being used here with it preinstalled so it already satisfies bitsandbytes>=0.45.2,<0.47.0. Is there a way to rebuild that? Or could we temporarily change the requirement to bitsandbytes>=0.46.0 to force an update?

[Borda]

Is there a way to rebuild that?

this will push new docker image: https://github.com/Lightning-AI/pytorch-lightning/actions/runs/16019234343

But this line with -U shall upgrade all packages if possible

pytorch-lightning/.azure/gpu-tests-fabric.yml

Line 133 in afa7d56

pip install -e ".[${extra}dev]" -U --extra-index-url="${TORCH_URL}"

[Borda]

all shall be fixed now :)