[Bugfix] Gracefully handle unsupported reasoning_effort in chat templates (#54022)

frankie committed about 2 months ago

[CI] fix pre-commit (#55630)

Jiangyun Zhu committed about 2 months ago

[Perf] Extend Qwen Triton warmup to avoid first-request latency spikes (#54797)

vhagor committed about 2 months ago

[Bugfix] Fix Kimi K3 NVFP4 MoE weight conversion OOM (#55407)

Wei Zhao committed about 2 months ago

[Bugfix] Fix cuda profiler missing bug (#55237)

Wei Zhao committed about 2 months ago

[Bugfix][Offloader] Preserve prefetch static-buffer slot ownership (#54975)

yulun committed about 2 months ago

[ROCm][Perf][DeepSeek V4] Fuse native FP8 shared expert with MXFP4 routed experts (#53161)

Fangzhou Ai committed about 2 months ago

add 2/3/5/6/7 CUDA support in AutoRound format (#52890)

Wenhua Cheng committed about 2 months ago

[Kernel] Remove unused fake implementation (#55535)

Jee Jee Li committed about 2 months ago

[Kernel] SM 12.x blockwise FP8: swizzle the CTA raster when the weight exceeds the L2 (#55180)

Jürgen Schmied committed about 2 months ago

[Bugfix] Defer adaptive verification until after kernel warmup (#55455)

Kevin H. Luu committed about 2 months ago

[CI] Align extraction test with canonical auxiliary layer order (#55457)

Kevin H. Luu committed about 2 months ago

[Frontend] Expose multimodal metadata for disaggregated prefill (#54659)

zhouyou committed about 2 months ago

[Governance] Add aoshen02 as code owner for RL components (#55529)

aoshen02 committed about 2 months ago

[Bugfix][KV Offload] Stop offloading the final sampled token's KV slot (#54288)

Almog Tavor committed about 2 months ago

[Bugfix][KV Offload] Fix SWA store reachability during chunked prefill (#54362)

Will Hampson committed about 2 months ago

[Bugfix] Fall back to T1 when ARC cannot reclaim enough entries from T2 (#55461)

Zupeng Wang committed about 2 months ago

[Bugfix][KV Offload] Skip cleaned-up async lookup batches (#55075)

AlexHuang committed about 2 months ago

[Perf] Use SDPA for BLIP-2 Q-Former attention (#55285)

Sun committed about 2 months ago

[Kimi K3] Support internal prefix checkpoints with partial prefix caching and spec-decoding (#53614)

Ziming Huang committed about 2 months ago