v0.20.2

infiniflow/ragflowv0.20.2May 10, 2026by khluu

AI Summary

This is a small patch release focused on bug fixes for DeepSeek V4, gpt-oss, and Qwen3-VL models, addressing issues with sparse attention, KV cache allocation, and deepstack boundaries.

Key Highlights

Fixed DeepSeek V4 sparse attention and KV cache allocation errors
Plumbed hidden_dim_unpadded for MXFP4 compatibility with torch.compile
Fixed Qwen3-VL deepstack boundary check under heavy load

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