vllm icon indicating copy to clipboard operation
vllm copied to clipboard

[Kernel] Flashinfer correctness fix for v0.1.3

Open LiuXiaoxuanPKU opened this issue 1 year ago • 4 comments

Reported by @felixzhu555 , is_profile_run is buggy in flashinfer backend, which will fail flashinfer v0.1.3. This PR fixes this, and update CI flashinfer version to v1.0.3.

LiuXiaoxuanPKU avatar Aug 08 '24 23:08 LiuXiaoxuanPKU

👋 Hi! Thank you for contributing to the vLLM project. Just a reminder: PRs would not trigger full CI run by default. Instead, it would only run fastcheck CI which consists a small and essential subset of CI tests to quickly catch errors. You can run other CI tests on top of default ones by unblocking the steps in your fast-check build on Buildkite UI.

Once the PR is approved and ready to go, please make sure to run full CI as it is required to merge (or just use auto-merge).

To run full CI, you can do one of these:

  • Comment /ready on the PR
  • Add ready label to the PR
  • Enable auto-merge.

🚀

github-actions[bot] avatar Aug 08 '24 23:08 github-actions[bot]

/ready

LiuXiaoxuanPKU avatar Aug 08 '24 23:08 LiuXiaoxuanPKU

CI failure seems like a real bug

[2024-08-09T04:52:05Z]   File "/usr/local/lib/python3.10/dist-packages/flashinfer/prefill.py", line 791, in begin_forward
--
  | [2024-08-09T04:52:05Z]     self._wrapper.begin_forward(
  | [2024-08-09T04:52:05Z] RuntimeError: CHECK_EQ(paged_kv_indptr.size(0), batch_size + 1) failed. 1 vs 257

comaniac avatar Aug 09 '24 16:08 comaniac

@LiuXiaoxuanPKU It's bugs image

exceedzhang avatar Aug 10 '24 03:08 exceedzhang