[Perf] Fuse DeepEncoder relative bias in Triton attention (#55629)

Sun committed about 2 months ago

[Bugfix][KV Offload] Respect prefix-cache bypass in SimpleCPUOffload (#54998)

Andy Lo committed about 2 months ago

[Perf][GDN] Enable the FlashInfer GDN prefill kernel on SM12x (#55715)

Stefano Castagnetta committed about 2 months ago

[ROCm][Perf] Remove AITER paged-MQA outputs guard for DeepSeek-V4 (#55808)

Shanshan Shen committed about 2 months ago

[CI] Increase ColQwen3 pooling test memory budget on H200 MIG (#55878)

Kevin H. Luu committed about 2 months ago

[Perf][Quant][NVFP4] Prefer W4A4 linear kernels over weight-only ones on SM120/121 (#55170)

Stefano Castagnetta committed about 2 months ago

[ROCm][Quantization] Support AMD Quark per-block FP8 for fused MoE layers (#52263)

jimmy-adams committed about 2 months ago

[Bugfix] Validate extension integers before engine serialization (#55606)

Kevin H. Luu committed about 2 months ago

[CI] Restore clean GPU state before OAI Triton MoE tests (#55453)

Kevin H. Luu committed about 2 months ago

[ROCm][Perf][M3] Fused allreduce+GemmaRMSNorm fast path (#54787)

Taoyu Zhu committed about 2 months ago

[Bugfix][KV Offload] Validate SWA coverage at unaligned cache-hit boundaries (#55712)

ahmed xijiaat committed about 2 months ago

[Frontend] Migrate input-validation errors to VLLMValidationError (harmony/mistral/chat_utils/params) (#55735)

Adababy committed about 2 months ago

[Bugfix][ROCm] Mask paged attention V cache padding (#53856)

aoshen02 committed about 2 months ago

[Bugfix][InternVL] Stop the video parser consuming image_embeds (#55779)

Subhash Polisetti committed about 2 months ago

[Bugfix] reject string file on translation like transcription (#55618)

Dev Oskii committed about 2 months ago

[Bugfix] Honor STEP token pooling in DispatchPooler.for_seq_cls (#55307)

weihan committed about 2 months ago

[Bugfix] Autotune FlashInfer deferred MoE decode kernels before CUDA graph capture (#55377)

mingg26 committed about 2 months ago

[Bugfix][KV Offload] Register mixed page sizes in one cache group (#54756)

He Liu committed about 2 months ago

[Kernel] PDL enablement for fusedQKNormRopeKernel (#55755)

Jee Jee Li committed about 2 months ago

[CI/Build] Fix unused fake implementation testing (#55797)

Jee Jee Li committed about 2 months ago