A high-throughput and memory-efficient inference and serving engine for LLMs
main
[Perf][KV Connector] Pin token_indices before non_blocking H2D in hf3fs helper (#54293)
Keyi Li committed about 2 months ago
[Test] Assert co-located RayExecutorV2 stores publish distinct ports (#54310)
aoshen02 committed about 2 months ago
[Bugfix][MRV2] Release model and KV cache on in-process engine shutdown (#54162)
Oxana Korzh committed about 2 months ago
[Bugfix][V1] Keep an encoder cache entry until its last occurrence is freed (#54284)
Hotragn Pettugani committed about 2 months ago
[Attention][DCP] Enable FlashInfer MLA for DSpark drafting (#54277)
Summer Yang committed about 2 months ago
[Hy4] support Hy4-preview model (#54160)
Cheng Jiang committed about 2 months ago
[Perf] Avoid h2d copies from non-pinned CPU tensors (#54299)
Nick Hill committed about 2 months ago
[CI/Build] Use file rendezvous for UniProc loader fixtures (#52367)
Stellar鱼 committed about 2 months ago
[BugFix] Disable TP for Qwen3-Omni audio encoder when heads % TP != 0 (#50858)
kaiyuanxie committed about 2 months ago
[Perf] Pin CPU tensors before non_blocking H2D in three MM paths (#54292)
Keyi Li committed about 2 months ago
[Bugfix][AMD] Annotate draft KV cache groups on the hybrid grouping path (#52047)
Oxana Korzh committed about 2 months ago
[Bugfix][Mooncake] Save exact Mamba boundary states (#51358)
Yifan Qiao committed about 2 months ago
[Bugfix][Models] Register sleep-managed runtime buffers (#53507)
Ronald committed about 2 months ago
[KV Connector] Support MooncakeStore with hybrid DCP prefix caching (#53324)
Wei Zhao committed about 2 months ago
[Perf][MLA Sparse] Pin req_id_per_token before non_blocking H2D on XPU and ROCm (#54295)
Keyi Li committed about 2 months ago
[Rust Frontend] Take the raw buffer in mm tensor lowering when possible (#53528)
RoachZhao committed about 2 months ago
[Bugfix][Distributed] Gate cross-node MNNVL custom all-reduce by group capability (#53253)
Julian committed about 2 months ago
[Bugfix] Fix ncclCommQueryProperties heap overflow with NCCL >= 2.31 (#53008)
Xuan-1998 committed about 2 months ago
fix: improve token_ids_cpu swap to copy only valid indices (#36255)
zzaebok committed about 2 months ago
[Rust Frontend] Add support for `truncate_prompt_tokens` and `truncation_side` (#48584)
Pranav Thakur committed about 2 months ago