共 19 个 commit，涉及 71 个文件，+1439/-305 行变动。

概要

统计项 数值
Commit 数 19
变更文件 71
新增行数 +1439
删除行数 -305

Commit 列表

📦 Other

  • cc1d020 #46942 — [MRV2] Enable mm prefix bidi attention support on MRV2 (#46942)
    • 作者: Isotr0py | +165/-3 | 5 个文件
    • Wire mm prefix attn mask computation to MRV2 - Actually, tests/models/multimodal/generation/test_common.py’s gemma3 test can’t catch the difference of with/without mm_prefixlm attention, because their logprobs are still quite similar. - This PR adds test to cover this regression. ## Test Result Test should only pass with this PR’s fix. —
  • 1a308c4 #43645 — [XPU] Add W8A8 FP8 linear kernel with multi-granularity quant support (#43645)
    • 作者: Chaojun Zhang | +188/-12 | 6 个文件

    This PR adds XPUW8A8FP8LinearKernel and XPUW8A16FP8LinearKernel for XPU, replacing the previous XPUFP8ScaledMMLinearKernel with proper support for multiple weight/activation quantization granularities. ### Changes - Add XPUW8A8FP8LinearKernel and XPUW8A16FP8LinearKernel, replacing XPUFP8ScaledMMLinearKernel - Select W8A8 or W8A16 kernel based on whether activation quantization is present - Ensure …

🐛 Bug Fix

  • 8974ed8 #44461 — [Bugfix][Voxtral Realtime] Fix token feedback timeout silent hang (#44461)
    • 作者: Ting SUN | +3/-6 | 1 个文件

    Closes #36015. Closes #35863. Voxtral realtime has a background feed_tokens() task that moves engine output tokens from input_stream into the realtime transcription buffer. That task currently wraps input_stream.get() in asyncio.wait_for(…, VLLM_ENGINE_ITERATION_TIMEOUT_S). A normal idle interval with no engine output can therefore raise TimeoutError and kill the feedback task while the websocke…

  • fb2face #46037 — [Bugfix][Model] Fix crash loading Mamba/Mamba2 checkpoints without an architectures field (#46037)
    • 作者: Ting SUN | +9/-0 | 1 个文件

    Loading a Mamba/Mamba2 checkpoint whose config.json omits the top-level architectures field crashes engine core init with a bare, message-less AssertionError: mamba_block_size is not None. This hits the official HF mamba2 conversions such as AntonV/mamba2-130m-hf and AntonV/mamba2-370m-hf. Root cause is an arch-name mismatch inside vLLM itself: 1. When architectures is absent, vLLM injects the bas…

  • fa4321d #47609 — [Bugfix][TurboQuant] Preserve KV cache dtype in backend shape (#47609)
    • 作者: Lucas Wilkinson | +6/-1 | 1 个文件

    Fix TurboQuant KV cache shape setup so TQFullAttentionSpec carries its real cache dtype (for example turboquant_k8v4) into backend shape/layout computation instead of being treated as auto. This prevents TurboQuant backends from seeing auto and failing with an unknown TurboQuant cache dtype during startup. ## CI Failure This fixes the Buildkite LM Eval TurboQuant KV Cache failure from build #7…

  • 9226613 #47590 — [Bugfix][Pooling] Forward instruction to Jina reranker scoring prompts (#47590)
    • 作者: Ting SUN | +179/-22 | 3 个文件

    /v1/score and /v1/rerank expose instruction for task-specific reranking, and the request model already folds it into chat_template_kwargs. The Jina reranker path builds its own prompt and did not pass that value into the existing formatter, so jinaai/jina-reranker-v3 accepted instruction requests with HTTP 200 while scoring the default prompt. This PR forwards the normalized instruction into the J…

  • 34b560b #47332 — [Bugfix][Gemma4] Fix FA4 mm_prefix mask: add sliding window and absolute q_idx (#47332)
    • 作者: Luciano Martins | +90/-50 | 1 个文件

    Purpose Fixes https://github.com/vllm-project/vllm/issues/47300 Gemma4 vision models emit total gibberish on the FlashAttention-4 (CuTeDSL, SM90) backend when the context is long (>=~1k tokens past the sliding window) and contains at least one image. The Triton backend produces correct output on the same input. Root cause — two bugs in _make_mm_prefix_mask_mod (vllm/v1/attention/backends/flash…

  • 91b5647 #47337 — [Bugfix][Model] Allow Run:ai memory_limit sentinel values (#47337)
    • 作者: Ting SUN | +24/-16 | 2 个文件

    Follow-up to #45291. @svasilinets pointed out that Run:ai documents RUNAI_STREAMER_MEMORY_LIMIT=-1 as the unlimited/default CPU memory mode, and the same section documents 0 as the minimal CPU buffer. The validation added in #45291 currently rejects both values because it requires memory_limit > 0. This PR keeps concurrency as a positive integer, but lets memory_limit follow Run:ai’s range semanti…

  • d2afe39 #47597 — [Bugfix][Frontend] Preserve default sampling params in batch chat (#47597)
    • 作者: Ting SUN | +4/-4 | 1 个文件

    /v1/chat/completions/batch converts each conversation into a ChatCompletionRequest, but the conversion currently dumps Pydantic defaults too. A batch request that omits sampling params still forwards temperature=0.7, top_p=1.0, min_p=0.0, and repetition_penalty=1.0, so it silently overrides model or server default sampling params that a normal /v1/chat/completions request would use. This keeps onl…

  • 0cd6f76 #47379 — [Bugfix][Frontend][gpt-oss] Recover raw tail when Harmony parser ends non-terminal (#47379)
    • 作者: yzong-rh | +145/-83 | 4 个文件

    co-author: @shenoyvvarun @Achyuthan-S Following https://github.com/vllm-project/vllm/pull/47062 and https://github.com/vllm-project/vllm/pull/46437, addresses issue https://github.com/vllm-project/vllm/issues/45736 for Responses API. Similar to https://github.com/vllm-project/vllm/pull/47061 Unlike #47061, this PR doesn’t implement raw output tracking for Responses API. Instead it consolidates raw…

  • e7c9df9 #44297 — [Bugfix][Structured Output][Spec Decode] Constrain bitmask and trim grammar advance at the reasoning boundary (#44297)
    • 作者: yue.yu | +480/-16 | 6 个文件

    Fixes #44006. When VLLM_ENFORCE_STRICT_TOOL_CALLING=1 is combined with MTP/EAGLE speculative decoding, the engine intermittently rejects token 248069 (Qwen’s ) and the request finishes with 500 “Failed to advance FSM … grammar rejected tokens”. There are two related defects in StructuredOutputManager.grammar_bitmask, a hardening follow-up to the first fix, and a scheduler-side fix requir…

✨ New Feature

  • b6cc46e #47070 — [Feature] Support sequence parallel without the need for DP, 1.9%~5.0% E2E Throughput Improvement (#47070)
    • 作者: Wentao Ye | +33/-22 | 7 个文件

    Support sequence parallel without the need for DP Originally we must have DP to eanble sequence parallel, but this constraint is not needed, this PR fixes the issue. vllm serve zai-org/GLM-5.2-FP8 –kv-cache-dtype fp8_e4m3 –enable-expert-parallel –tensor-parallel-size 8 –tool-call-parser glm47 –enable-auto-tool-choice –reasoning-parser glm45 –port 9256 –profiler-config.profiler=torch –prof…

🧪 CI/Tests

  • 4a6bf3c #47519 — [ROCm][CI] Fix Kernels and Kernels attention test failures (#47519)
    • 作者: Carl Persson | +11/-2 | 2 个文件

    This PR fixes failures on ROCm tests MI355 Kernels and Kernels Attention. The tests were failing due to a clone operation being attempted on the prefill_backend which was set to None for testing. This PR also fixes a failure on the Kernels MoE test (test_ocp_mx_moe.py), which was failing due to the distributed environment not being initialized and an assertion failing due to rocm_aiter_ops.is_enab…

  • f1445f6 #47551 — [CI] Bump huggingface-hub from v1.10.2 to v1.22.0 (#47551)
    • 作者: Harry Mellor | +49/-45 | 25 个文件

    This release includes https://github.com/huggingface/huggingface_hub/pull/4394: > snapshot_download now caches a repository’s file listing on disk under a new trees/ folder, so re-downloading a commit that’s already cached costs a single network call — resolving the branch or tag to a commit hash — instead of one metadata request per file. The listing is immutable per commit and shared by both sna…

⚡ Performance

🔩 Misc

  • 1d354c6 #46966 — [Misc] Validate Pooling cache_salt Values (#46966)
    • 作者: Taneem Ibrahim | +15/-0 | 1 个文件

    Pooling requests expose cache_salt, but unlike chat completions, completions, and responses, it does not validate that the value is a non-empty string. This adds the same validation to the shared pooling requests so embedding, classification, scoring/rerank, and generic pooling requests all reject invalid cache_salt values consistently before they enter request processing. This existing valid cach…

  • 2f21224 #47333 — [Misc] Update request-extras parity for batch chat completion (#47333)
    • 作者: Taneem Ibrahim | +5/-1 | 2 个文件

    Currently batch chat completions convert each conversation into a normal ChatCompletionRequest, but BatchChatCompletionRequest did not explicitly declare several compatible request extras that normal chat completions already support. This adds explicit batch request fields for: - media_io_kwargs - mm_processor_kwargs - priority - cache_salt ### Main vs Branch Behavior Reproducer: On main: …

  • fa1fa96 #46939 — [Misc] Forward request-level prompt extras for cross-encoder scoring (#46939)
    • 作者: Taneem Ibrahim | +13/-1 | 1 个文件

    Cross-encoder scoring already collected request extras like cache_salt, mm_processor_kwargs, and chat_template_kwargs, but _pre_process only used chat_template_kwargs. The engine-level extras were dropped before process_for_engine, so cross-encoder score/rerank requests did not consistently honor request-level cache salt or multimodal processor overrides. This PR forwards the engine prompt extras,…

  • 6eac8e0 #47082 — [Misc] Preserve cross-encoder pooling extra kwargs (#47082)
    • 作者: Taneem Ibrahim | +4/-1 | 1 个文件

    Cross-encoder scoring adds compressed_token_type_ids to PoolingParams.extra_kwargs when token type IDs are produced during preprocessing. Before this change, that assignment replaced the entire extra_kwargs dict. That could drop existing request/pooling extras such as cache_salt, mm_processor_kwargs, or future scoring extras. This PR preserves existing extra_kwargs and adds only the internal compr…