[Bugfix][KV Cache][MLA] Align packed block strides for V3.2 sparse MLA (#55528)

lz committed 27 days ago

[Bugfix] Stop leaking the internal field name in the max_tokens validation error (#58336)

shallow10 committed 27 days ago

[Frontend] Handle Disable Thinking in /v1/messages (#58613)

Robert Shaw committed 27 days ago

[ROCm] Credit ROCm/aiter for the block32 GEMM's packed kernel and in-launch split-K (#58659)

Lingpeng Jin committed 27 days ago

[Structured Outputs] Parse Lark grammars natively in the xgrammar backend (#58321)

Bugen Zhao committed 27 days ago

[MM] Add Triton kernel for mm_input_normal. (#56798)

wang.yuqi committed 27 days ago

[Qwen4Exp][ROCm] PLE n-gram table CPU offload (#57497)

Mathew Odden committed 27 days ago

[CI][ROCm] Prevent Model Executor apt stalls (#58607)

Andreas Karatzas committed 27 days ago

[ROCm] Fix CI runtime and tests for MI355 DPX (#58244)

mkunredd committed 27 days ago

[Feature] Triton kernel dispatcher (#43048)

wangxiyuan committed 27 days ago

[PP][XPU]Add the flag to control microbatch feature on MRV2+PP (#55145)

YiSheng5 committed 27 days ago

[Perf][Distributed] Add low-SM multimem reduce-scatter for SM100/SM103 (#55072)

Yongye Zhu committed 27 days ago

[ROCm][Build] Filter crate tags from vLLM version detection (#57744)

Reid committed 27 days ago

[ROCm] Cut 69 wasted contiguous copies per decode step from the skinny GEMM path (#58566)

fululi12 committed 27 days ago

[Bugfix][Quantization] Fix MXFP8 startup crash on layers below mm_mxfp8 shape limits (#54223)

Samuel Kim committed 27 days ago

[Perf][Engram] Serialize offloaded lookups and pack host tables into huge pages (#56926)

Juntian Liu committed 27 days ago

[Bugfix][Qwen4Exp] Keep pinned PLE prefetch ids out of the CUDA graph pool (#58489)

Juntian Liu committed 27 days ago

[ROCm][Perf] MXFP8 GEMM on native 32x32 block scales for gfx950 (#58510)

Fangzhou Ai committed 27 days ago

[Perf][Rust Frontend] Make histogram observations lock-free (#58574)

Bugen Zhao committed 27 days ago

[gRPC] Fix ping tolerance so long non-streaming RPCs are not dropped (#55102)

gongwei-130 committed 27 days ago