Starstruk Logo MarkStarstruk

v0.20.2

git-ai-project/git-aiv0.20.2May 10, 2026by khluu

AI Summary

A small patch release addressing critical bugs in DeepSeek V4 sparse attention, KV cache allocation, and compatibility issues with `torch.compile` for gpt-oss.

Key Highlights

  • Fix DeepSeek V4 sparse attention hang on Hopper
  • Fix DeepSeek V4 KV cache allocation error
  • Fix gpt-oss MXFP4 compatibility with `torch.compile`
  • Fix Qwen3-VL boundary check failure

New Features

  • DeepSeek V4 sparse attention fix
  • DeepSeek V4 KV cache allocation fix
  • gpt-oss MXFP4 torch.compile compatibility fix
  • Qwen3-VL boundary check fix

Full Release Notes

# vLLM v0.20.2

## Highlights
This release features 6 commits from 6 contributors (0 new)!

This is a small patch release with bug fixes for DeepSeek V4, gpt-oss, and Qwen3-VL

### Bug Fixes
* **DeepSeek V4 sparse attention**: Re-enable the persistent topk path on Hopper and ensure the memset kernel runs at CUDA graph capture time regardless of `max_seq_len`, fixing the MTP=1 hang on DeepSeek V4 (#41665, revert of #41605).
* **DeepSeek V4 KV cache**: Fixed a "failure to allocate KV blocks" error in the V1 engine KV cache manager (#41282).
* **gpt-oss MXFP4 + torch.compile**: Plumbed `hidden_dim_unpadded` through the `moe_forward` fake op so MXFP4 works under `torch.compile` on v0.20.x (#42002, backport of #41646).
* **Qwen3-VL**: Removed an invalid deepstack boundary check that could fail under heavy load (#40932).

## Contributors
@ywang96, @zyongye, @stecasta, @wzhao18, @Isotr0py, @khluu