[Rust Frontend] Support HTTP RL weight synchronization (#56567)

Zupeng Wang committed about 1 month ago

[CI][Bugfix] Complete FSE fixture R-SWA contract (#56710)

Kevin H. Luu committed about 1 month ago

[Fast Start] Fast loader support nnode>1 (#55468)

siyu committed about 1 month ago

[Bugfix][Frontend] Support Responses text types in DeepSeek V4.1 (#56299)

CedricHuang committed about 1 month ago

[ROCm][Perf] Tune MiniMax-M3 decode top-k for short contexts (#55235)

Fangzhou Ai committed about 1 month ago

[Bugfix][KV Connector] MooncakeStore: exclude non-prefix-cacheable (QSA ring) groups; fix align-mode check (#55027)

Zhewen Li committed about 1 month ago

[XPU][CI] update test requirements (#56704)

liuzhenwei committed about 1 month ago

[Perf] Parallelize mHC pre-norm JIT warmup (#56683)

Woosuk Kwon committed about 1 month ago

[DSv4 Bug] fix dsv4 start up error `NotImplementedError: DeepSeek V4 MegaMoE currently requires expert parallel` (#55914)

Wentao Ye committed about 1 month ago

[Performance][EPD] Reduce Python proxy serialization overhead (#56657)

Tianyu Guo committed about 1 month ago

[Attention] Remove DCP indexer interleave guard and test TP1 output parity (#55802)

stu-cao committed about 1 month ago

[Warmup] Gemma 4 de-JITification (#55768)

Roberto L. Castro committed about 1 month ago

[Bugfix][PCP][DCP] Respect interleave in indexer KV gather mapping (#56715)

Lucas Wilkinson committed about 1 month ago

[NIXL][PCP][DCP] Expose PCP producer KV shards as transfer ranks (#56645)

Lucas Wilkinson committed about 1 month ago

[Perf] Avoid redundant conversions in FP8 dummy initialization (#56688)

Woosuk Kwon committed about 1 month ago

[CI] Initialize warmup registry in V2 QSA runner fixture (#56707)

Lucas Wilkinson committed about 1 month ago

[Bugfix] Skip Triton autotune inspection without Triton (#56676)

Kevin H. Luu committed about 1 month ago

[ROCm][Performance] Avoid blocking MiniMax M3 scalar upload (#56170)

Fangzhou Ai committed about 1 month ago

[Bugfix] Detect unloaded NVFP4 weight scales with a NaN sentinel (#52501)

Pavel Zakharov committed about 1 month ago

[Attention] Add Triton/FlashInfer composite for multimodal prefix attention (#56305)

Matthew Bonanni committed about 1 month ago