All pull requests

Pull requests

• [Feature] Add MRV2 DSpark support on Kunlun

  enhancementkunlun-xpuneeds reviewtests
  #498·czyu-nju opened Oct 12, 2026·

  ·

• [Feature] Add MRV2 MTP and INT4 W4A8 support on Kunlun

  enhancementkunlun-xpuneeds reviewtests
  #497·czyu-nju opened Oct 12, 2026·

  ·

• [Docs] Fix install order: install requirements.txt before PyTorch

  documentationneeds review
  #495·1916hcc opened Sep 28, 2026Contributor·

  ·

• [Feature] Enable MRV2 hybrid prefix caching and optimize metadata paths on Kunlun

  enhancementkunlun-xpuneeds reviewtests
  #492·czyu-nju opened Sep 24, 2026·

  ·

• [Feature] Route Model Runner V2 through xspeedgate_ops native ops

  bugkunlun-xpuneeds reviewtests
  #490·Joeegin opened Sep 20, 2026Contributor·3 tasks done·

  ·
  4

• [Refactor] Register derived RoPE implementations through OOT

  kunlun-xpuneeds review
  #488·GrootLiu opened Sep 16, 2026Collaborator·

  ·

• [WIP][Feature] Support MoE expert LoRA inference

  kunlun-xpuneeds review
  #487·loveleaves opened Sep 16, 2026·

  ·
  4

• [Refactor] Unify fused MoE routing and execution pipeline

  documentationkunlun-xpuneeds reviewtests
  #486·GrootLiu opened Sep 16, 2026Collaborator·

  ·
  1

• Kimi k3

  kunlun-xpuneeds reviewtests
  #482·baoqian426 opened Sep 10, 2026Contributor·3 tasks·

  ·
  2

• MiniMax-M3 Support

  enhancementkunlun-xpuneeds review
  #478·xyDong0223 opened Sep 8, 2026Collaborator·3 tasks done·

  ·
  12

• Fix/minimax 2 5 environment

  #458·xyDong0223 opened Sep 2, 2026Collaborator·3 tasks·

  ·
  5

• [Doc] Add automated OpenWiki updates

  #456·xyDong0223 opened Sep 2, 2026Collaborator·3 tasks done·

  ·
  4

• feat(dsv4): support DeepSeek-V4-Flash inference on Kunlun XPU

  #439·BoundlessWindMoon opened Aug 20, 2026·

  ·
  8

• [Feature] Support DeepSeek-V4-Flash on Kunlun — FP8 + INT8 MoE Native Support

  #402·BoundlessWindMoon opened Jul 29, 2026·

  ·
  5

• feat: add MiniMax M3 + M2.7 support (Kunlun XPU)

  #396·NaphJohn opened Jul 24, 2026Contributor·3 tasks·

  ·
  4

• [Feature] Support Qwen3.5 MoE MTP speculative decoding

  #384·sdh1014 opened Jul 5, 2026·3 tasks done·

  ·
  4

• [BUGFIX]int8_moe_backend hook

  #381·Marshall-Ge opened Jun 26, 2026Contributor·

  ·
  2

• feat: upgrade vllm_kunlun to support vllm 0.23.0 qwen3-8b

  #380·1916hcc opened Jun 26, 2026Contributor·

  ·

• [Benchmark] Adapt attention and fused MoE benchmarks to vLLM-Kunlun

  #364·KinChow opened May 26, 2026·3 tasks done·

  ·
  7

• [Bugfix] Force xgrammar to use torch_native backend on XPU

  #353·GrootLiu opened May 8, 2026Collaborator·3 tasks done·

  ·
  4

• [Doc] Align developer guide navigation titles

  #350·Lidang-Jiang opened May 6, 2026Contributor·2 tasks done·

  ·

• [Doc] Clean up contributing guide navigation

  #349·Lidang-Jiang opened May 6, 2026Contributor·

  ·

• [Doc] Add model source guidance

  #348·Lidang-Jiang opened May 6, 2026Contributor·

  ·

• [Doc] Replace placeholder Docker image tags

  #347·Lidang-Jiang opened May 6, 2026Contributor·7 tasks done·

  ·

• [Doc] Update documentation copyright year

  #346·Lidang-Jiang opened May 6, 2026Contributor·4 tasks done·

  ·