[CI] Retire Weight Loading smoke tests (#57398)

Misha Goin committed about 1 month ago

[MoE][Bugfix] Skip SP padded rows in grouped MoE routing (#56079)

Elvir Crnčević committed about 1 month ago

[Bugfix][HiSparse][NIXL] Import full blocks without tail prefill on D (#57049)

Nicolò Lucchesi committed about 1 month ago

[DSv4.1] Integrate Mega-Gate from DeepGEMM (#56266)

Thien Tran committed about 1 month ago

[ROCm] Restore `VLLM_ROCM_USE_AITER_FP4_ASM_GEMM` and default w4a4 ASM GEMM back to off (#57055)

afriedri committed about 1 month ago

[Bugfix][Metrics] Do not log a 0.0% prefix cache hit rate before any query (#54990)

Michele Campi committed about 1 month ago

[Bugfix] Max-load throughput cliff when `max_num_seqs` is not a multiple of 8 (#57355)

Andy Lo committed about 1 month ago

[ROCm] Resolve the indexer fp8 cache dtype once at import (#53792)

Sriram Kumar committed about 1 month ago

[Kimi K3 Bug] Fix kimi k3 reasoning parser (#57098)

Wentao Ye committed about 1 month ago

[CI][Bugfix] Add tp_shard_with_padding to padded MoE reload test mock (#57402)

vllm-agent committed about 1 month ago

[ROCm][CI] Fix Entrypoints Integration (Pooling) tests on TheRock image (#57380)

Matt committed about 1 month ago

[ROCm][CI] Adapt MoE tests to the triton_kernels 3.8 API (#57385)

Matt committed about 1 month ago

[Mypy] Fix mypy typing for Transformers models (#54320)

Taneem Ibrahim committed about 1 month ago

[Docs] Fix the typos in the document (#57190)

Young committed about 1 month ago

[Bugfix] Fix wrong vLLM version reported by pip install (proto-v* tag collision) (#57295)

Ziyang Ma committed about 1 month ago

[Bugfix][GDN] Fix CuteDSL BF16 KKT inversion divergence (#53864)

Zoe923 committed about 1 month ago

[bugfix] Mark draft tokens to rebuilt their embeddings. (#57356)

HieDean committed about 1 month ago

[Perf][ROCm] Enable AITER QuickReduce + RMSNorm fusion (#48249)

Mikko Tukiainen committed about 1 month ago

[CI/Build][Hardware][NVIDIA] Add Rubin CUDA 13.4 nightly images (#55953)

Tyler Michael Smith committed about 1 month ago

[Fix][ROCm] MXFP4 MoE round-up inflates TP-sharded expert weights on CDNA3, starving KV cache (#56359)

vllmellm committed about 1 month ago