[Fast Start] Support data parallelism in the weight cache daemon (#57386)

siyu committed 30 days ago

[ROCm][DSv4.1][Perf] Fuse the inverse RoPE into the sparse decode reduce (#57435)

Fangzhou Ai committed 30 days ago

[Bugfix][ROCm] Dispatch the QuantFP8 CUDA fallback on the class (#58136)

stefankoncarevic committed 30 days ago

[Bugfix] hadacore_transform: respect inplace parameter to fix garbage outputs with QuIP transforms (#43462)

Gilles Turpin committed 30 days ago

[Bugfix][Engram] Fall back when /dev/shm is absent before sharing tables (#57914)

Juntian Liu committed 30 days ago

[Bugfix][Attention] Avoid NaN in the Triton softcap for large attention logits (#56579)

Kushal committed 30 days ago

[Bugfix] prioritize architecture capability before DeepGEMM availability check (#58073)

Tony Lin committed 30 days ago

[Bugfix] Narrow AuxOutput KV restrictions to known PD connectors (#58150)

aoshen02 committed 30 days ago

[SpecDecode] Restore residual-logits comments in _resample_kernel (#58166)

Lucas Wilkinson committed 30 days ago

[Feat][XPU] VLLM_BATCH_INVARIANT support for Dense/MoE models (#55881)

Tony Lin committed 30 days ago

[Bugfix][Structured Output] Disallow MRV1 + PP>1 + async sched + structured output (#56250)

Artem Perevedentsev committed 30 days ago

[Bugfix][GLM-5.3-Flash] Run the dense MLP layers on the sequence-parallel shard (#58061)

Jared Wen committed 30 days ago

[Build] Fix CUDA 12 KV connector dependency selection (#57945)

Jee Jee Li committed 30 days ago

[Bugfix][Structured Outputs] Reject empty `structural_tag` at request validation (#47450)

linyafeng committed 30 days ago

[XPU][CI]Remove model_runner_v2 test from Intel GPU CI (#58050)

xiangdong committed about 1 month ago

[BugFix][Core] Make the structured-output grammar poll non-blocking (#55931)

ubwzwd committed about 1 month ago

[Rust Frontend] Construct model-owned vision processors through specs (#58109)

Bugen Zhao committed about 1 month ago

[CPU] Add device-memory-utilization CLI alias (#56547)

Louie Tsai committed about 1 month ago

[perf] wire FA and FlashMLA for sm90 GLM5Next NoPE SparseMLA (#55385)

Jared Wen committed about 1 month ago

[CI] Emit a kernel symbol map from the csrc build (opt-in, for test selection) (#58097)

Kevin H. Luu committed about 1 month ago