bump: bitsandbytes >=0.45.2,<0.47.0 & compatibility patch for bnb.functional.int8_double_quant

[Borda]

What does this PR do?

path for new release - update bitsandbytes requirement from <0.43,>=0.42 to >=0.42,<0.47

    @staticmethod
    def quantize(
        int8params: bnb.nn.Int8Params, weight: torch.Tensor, device: Optional[torch.device]
    ) -> bnb.nn.Int8Params:
        device = device or torch.device("cuda")
        if device.type != "cuda":
            raise RuntimeError(f"Unexpected device type: {device.type}")
        # https://github.com/TimDettmers/bitsandbytes/blob/0.41.0/bitsandbytes/nn/modules.py#L291-L302
        B = weight.contiguous().to(device=device, dtype=torch.float16)
        if int8params.has_fp16_weights:
            int8params.data = B
        else:
>           CB, CBt, SCB, SCBt, _ = bnb.functional.double_quant(B)
E           AttributeError: module 'bitsandbytes.functional' has no attribute 'double_quant'. Did you mean: 'int8_double_quant'?

/usr/local/lib/python3.10/dist-packages/lightning/fabric/plugins/precision/bitsandbytes.py:259: AttributeError

ref: https://dev.azure.com/Lightning-AI/lit%20Models/_build/results?buildId=235048&view=logs&jobId=80e425e9-2286-5482-2748-077465d0eec2&j=80e425e9-2286-5482-2748-077465d0eec2&t=c5f621af-a304-5088-cbea-3c82554e8451

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--20911.org.readthedocs.build/en/20911/

cc @justusschock @lantiga

[github-actions]

⚡ Required checks status: All passing 🟢

Groups summary

🟢 pytorch_lightning: Tests workflow

Check ID	Status
pl-cpu-guardian	success	✅

These checks are required after the changes to requirements/fabric/strategies.txt, src/lightning/fabric/plugins/precision/bitsandbytes.py, requirements/pytorch/extra.txt.

🟢 pytorch_lightning: Azure GPU

Check ID	Status
pytorch-lightning (GPUs) (testing Lightning | latest)	success	✅
pytorch-lightning (GPUs) (testing PyTorch | latest)	success	✅

These checks are required after the changes to requirements/pytorch/extra.txt, requirements/fabric/strategies.txt, src/lightning/fabric/plugins/precision/bitsandbytes.py.

🟢 pytorch_lightning: Benchmarks

Check ID	Status
lightning.Benchmarks	success	✅

These checks are required after the changes to requirements/fabric/strategies.txt, requirements/pytorch/extra.txt, src/lightning/fabric/plugins/precision/bitsandbytes.py.

🟢 fabric: Docs

Check ID	Status
docs-make (fabric, doctest)	success	✅
docs-make (fabric, html)	success	✅

These checks are required after the changes to src/lightning/fabric/plugins/precision/bitsandbytes.py, requirements/fabric/strategies.txt.

🟢 pytorch_lightning: Docs

Check ID	Status
docs-make (pytorch, doctest)	success	✅
docs-make (pytorch, html)	success	✅

These checks are required after the changes to requirements/pytorch/extra.txt.

🟢 pytorch_lightning: Docker

Check ID	Status
build-cuda (3.10, 2.1.2, 12.1.1)	success	✅
build-cuda (3.11, 2.2.2, 12.1.1)	success	✅
build-cuda (3.11, 2.3.1, 12.1.1)	success	✅
build-cuda (3.11, 2.4.1, 12.1.1)	success	✅
build-cuda (3.12, 2.5.1, 12.1.1)	success	✅
build-cuda (3.12, 2.6.0, 12.4.1)	success	✅
build-pl (3.10, 2.1, 12.1.1)	success	✅
build-pl (3.11, 2.2, 12.1.1)	success	✅
build-pl (3.11, 2.3, 12.1.1)	success	✅
build-pl (3.11, 2.4, 12.1.1)	success	✅
build-pl (3.12, 2.5, 12.1.1)	success	✅
build-pl (3.12, 2.6, 12.4.1)	success	✅
build-pl (3.12, 2.7, 12.6.3, true)	success	✅

These checks are required after the changes to requirements/pytorch/extra.txt, requirements/fabric/strategies.txt.

🟢 lightning_fabric: CPU workflow

Check ID	Status
fabric-cpu-guardian	success	✅

These checks are required after the changes to requirements/fabric/strategies.txt, src/lightning/fabric/plugins/precision/bitsandbytes.py.

🟢 lightning_fabric: Azure GPU

Check ID	Status
lightning-fabric (GPUs) (testing Fabric | latest)	success	✅
lightning-fabric (GPUs) (testing Lightning | latest)	success	✅

These checks are required after the changes to requirements/fabric/strategies.txt, src/lightning/fabric/plugins/precision/bitsandbytes.py.

🟢 mypy

Check ID	Status
mypy	success	✅

These checks are required after the changes to requirements/fabric/strategies.txt, requirements/pytorch/extra.txt, src/lightning/fabric/plugins/precision/bitsandbytes.py.

🟢 install

Check ID	Status
install-pkg-guardian	success	✅

These checks are required after the changes to src/lightning/fabric/plugins/precision/bitsandbytes.py, requirements/fabric/strategies.txt, requirements/pytorch/extra.txt.

---

Thank you for your contribution! 💜

Note
This comment is automatically generated and updates for 60 minutes every 180 seconds. If you have any other questions, contact carmocca for help.

[Copilot]

Pull Request Overview

This PR ensures compatibility with newer bitsandbytes releases by falling back to the new int8_double_quant API when double_quant is unavailable.

• Adds a feature check to call double_quant for older versions and int8_double_quant for bitsandbytes ≥0.46
• Adjusts unpacking of returned values accordingly

Comments suppressed due to low confidence (1)

src/lightning/fabric/plugins/precision/bitsandbytes.py:262

• Add a unit test to cover the int8_double_quant path (bitsandbytes ≥0.46) to ensure this fallback remains functional across releases.

CB, SCB = bnb.functional.int8_double_quant(B)

[codecov]

Codecov Report

❌ Patch coverage is 0% with 3 lines in your changes missing coverage. Please review.
✅ Project coverage is 87%. Comparing base (87e3974) to head (7d5d1b0).
⚠️ Report is 199 commits behind head on master.

Additional details and impacted files

@@           Coverage Diff           @@
##           master   #20911   +/-   ##
=======================================
  Coverage      87%      87%           
=======================================
  Files         268      268           
  Lines       23442    23442           
=======================================
  Hits        20391    20391           
  Misses       3051     3051           

[lantiga]

thank you @Borda