Skip to content

Remove dead bitsandbytes CxB code from 8-bit inference path#34633

Merged
vllm-bot merged 1 commit into
vllm-project:mainfrom
TimDettmers:fix/remove-dead-bnb-CxB-code
Feb 17, 2026
Merged

Remove dead bitsandbytes CxB code from 8-bit inference path#34633
vllm-bot merged 1 commit into
vllm-project:mainfrom
TimDettmers:fix/remove-dead-bnb-CxB-code

Conversation

@TimDettmers
Copy link
Copy Markdown
Contributor

@TimDettmers TimDettmers commented Feb 16, 2026

Summary

  • Remove dead code in BitsAndBytesLinearMethod._apply_8bit_weight() that accesses MatmulLtState.CxB — an attribute that has been always None since bitsandbytes v0.45.0 (December 2024)
  • This is a no-op change: the removed branch has never executed since CxB was part of the old col32/ColAmpere tensor layout system that was removed over a year ago

Context

bitsandbytes is cleaning up its deprecated API surface. In the current release, CxB will still exist as a deprecated stub that emits a FutureWarning and returns None (see bitsandbytes#1871), so there is no immediate breakage. However, the attribute will be fully removed in the next bitsandbytes release.

This PR removes the dead code proactively so vLLM is ready for both the current and next bitsandbytes versions.

Test plan

  • Existing 8-bit inference tests pass (no behavioral change — the removed code paths were unreachable)

🤖 Generated with Claude Code

The `MatmulLtState.CxB` attribute has been always `None` since
bitsandbytes v0.45.0 (December 2024), when the col32/ColAmpere
tensor layout system was removed. The conditional block that
checked `matmul_states[i].CxB is not None` has therefore never
executed.

bitsandbytes is removing the `CxB` attribute entirely in an
upcoming release (see bitsandbytes-foundation/bitsandbytes#1871),
which would cause an AttributeError here. This commit removes the
dead code proactively.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
@github-actions
Copy link
Copy Markdown

👋 Hi! Thank you for contributing to the vLLM project.

💬 Join our developer Slack at https://slack.vllm.ai to discuss your PR in #pr-reviews, coordinate on features in #feat- channels, or join special interest groups in #sig- channels.

Just a reminder: PRs would not trigger full CI run by default. Instead, it would only run fastcheck CI which starts running only a small and essential subset of CI tests to quickly catch errors.

You ask your reviewers to trigger select CI tests on top of fastcheck CI.

Once the PR is approved and ready to go, your PR reviewer(s) can run CI to test the changes comprehensively before merging.

To run CI, PR reviewers can either: Add ready label to the PR or enable auto-merge.

If you have any questions, please reach out to us on Slack at https://slack.vllm.ai.

🚀

Copy link
Copy Markdown
Contributor

@gemini-code-assist gemini-code-assist Bot left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Code Review

The pull request effectively removes dead code related to BitsAndBytesLinearMethod._apply_8bit_weight() that accessed MatmulLtState.CxB. This change proactively addresses upcoming bitsandbytes API removals, ensuring future compatibility and preventing potential AttributeError crashes. The removal is a no-op in terms of functionality, as the affected code path was already unreachable. The changes are clean and directly target the identified dead code.

Copy link
Copy Markdown
Member

@tlrmchlsmth tlrmchlsmth left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks for the cleanup!

@tlrmchlsmth tlrmchlsmth added the ready ONLY add when PR is ready to merge/full CI is needed label Feb 16, 2026
@tlrmchlsmth tlrmchlsmth enabled auto-merge (squash) February 16, 2026 21:08
@vllm-bot vllm-bot merged commit d44a5b6 into vllm-project:main Feb 17, 2026
62 of 67 checks passed
jiangkuaixue123 pushed a commit to jiangkuaixue123/vllm that referenced this pull request Apr 28, 2026
mystous pushed a commit to mystous/vllm_hybrid that referenced this pull request May 10, 2026
…ject#34633)

Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
my-other-github-account pushed a commit to my-other-github-account/vllm that referenced this pull request May 15, 2026
…ject#34633)

Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
my-other-github-account pushed a commit to my-other-github-account/vllm that referenced this pull request May 15, 2026
…ject#34633)

Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
0826joyce pushed a commit to 0826joyce/vllm-serving-optimization that referenced this pull request May 19, 2026
…ject#34633)

Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

ready ONLY add when PR is ready to merge/full CI is needed

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants