vllm-project/vllm

Pull Requests Commits

Merge branch 'main' into codex/dspark-capacity-realloc

njhill committed 52 days ago

aa72154a

[Bugfix][MRV2] Require all requests to be decoding for uniform-decode dispatch (#51865)

njhill committed 52 days ago

Verified 0a94d85a

[Attention] Add FlashInfer XQA decode support on SM12x (#49718)

askliar committed 53 days ago

Verified 1d2d83a0

[Profiler] Add minimal Triton Proton profiling backend (#48789)

Luosuu committed 53 days ago

Verified a367cdbc

[4/N][KV-Cache Layout Refactor] Promote local KV cache specs via a class-changing replace helper (#51612)

LucasWilkinson committed 53 days ago

Verified 61874f98

[Testing] Fix test_sharded_state_loader (#51736)

zou3519 committed 53 days ago

Verified cb30f6f8

[Bugfix][ROCm] Give KV-first attention blocks their own page in hybrid models (#51837)

stefankoncarevic committed 53 days ago

Verified 3e372c5f

[ROCm][MoE] Fix expert_map vs AITER expert_mask for non-AITER experts under EP (#49758)

Rohan138 committed 53 days ago

Verified 0f0cb918

[Bugfix][MoE] Fix fused block-scale orientation (#50727)

AndreasKaratzas committed 53 days ago

Verified 36f4630d

Merge branch 'main' into codex/dspark-capacity-realloc

mergify[bot] committed 53 days ago

Verified a48eccc0

[Refactor] Delete dead code in models (#51838)

yewentao256 committed 53 days ago

Verified fd04edef

[Bugfix] Support HF-config compat for Inkling (#51850)

mgoin committed 53 days ago

Verified ded6c452

[Docs] Fix broken autorefs cross-reference in TurboQuant v2 docstring (#51857)

hmellor committed 53 days ago

Verified b2ab0960

[CI][Bugfix][V1] Remove stale FlashAttention metadata arguments (#51854)

AndreasKaratzas committed 53 days ago

Verified cc668c5b

[ROCm] Remove stale SDPA and skinny GEMM workarounds (#50907)

mgoin committed 53 days ago

Verified 1ab2801d

[Bugfix][Attention] Forward per-head FP8 descales through FA4 (#51363)

yiliu30 committed 53 days ago

Verified 4f2f31b8

feat: allow shared expert overlapping for FlashInfer one-sided all-to-all (#50569)

jdebache committed 53 days ago

Verified 52be12cf

Enable `MiniCPMV` for vLLM in CI (#45042)

hmellor committed 53 days ago

Verified 9cc347ae

[CI] Support partial torch requirement contexts (#51832)

taneem-ibrahim committed 53 days ago

Verified ca9c8cbd

[Bugfix] Generalize KV block zeroing to `AttentionSpec` (#51749)

mgoin committed 53 days ago

Verified e3fe212e

[Bugfix] Take the sliding window from the layer, not the KV cache group (#51756)

njhill committed 53 days ago

Verified 513f83e7

Bump Transformers version to 5.15.0 (#51668)

hmellor committed 53 days ago

Verified c65aa2ee

[Bugfix][MRV2] Support encoder timing stats in model runner V2 (#50020)

guan404ming committed 53 days ago

Verified 457a5f34

[Kernel][ROCm][Perf] FlyDSL decode-attention kernel for 4-bit TurboQuant KV cache (#47896)

aditi-amd committed 53 days ago

Verified 5426311d

[XPU][CI] Fix ExampleConnector KV cache device selection (#51806)

zhenwei-intel committed 53 days ago

Verified d6482361

[Bugfix] Align Qwen GDN gates with speculative tokens (#51812)

ZJY0516 committed 53 days ago

Verified 5af7c8da

[Misc] Enable test_silu_mul_fp8_quant_deep_gemm on XPU (#49444)

pmanczak committed 53 days ago

Verified 5b5eae23

[Config] Update default `_max_num_batched_tokens` from 8192 to 16384 (#51726)

yewentao256 committed 53 days ago

Verified 4988df2e

[Bugfix][MoE] Support GELU tanh in FlashInfer B12x MoE (#51819)

askliar committed 53 days ago

Verified 0fb9897d

[Bugfix] Guard DeepSeek V4 MRV1 piecewise CUDA graphs (#51768)

WoosukKwon committed 53 days ago

Verified 87668ab6

Older