All pull requests

Pull requests

• vllm: package LMCache KVWeave cache offload

  #707·gc-fu opened Sep 14, 2026·

  ·

• Guard info_once() in SymInt4LinearMethod.apply() so sym_int4 can start under torch.compile / XPU graph

  #700·jsetsuda opened Sep 11, 2026·

  ·
  2

• vllm: alt-stack community recipe for Qwen3.8-27B-FP8 on 2x Arc Pro B70 (vLLM 0.27.2rc1, #53059-lineage scheduler gate, 0.1.14.dev544 kernels) — DRAFT, pending week-long soak

  #693·dominick253 opened Sep 9, 2026·

  ·
  4

• docs(vllm): fix dots.ocr garbled output by shipping a corrected chat template

  #682·KristianZeng opened Sep 8, 2026·

  ·

• Enabe qwen3.8 flash next

  #660·hzjane opened Sep 2, 2026·

  ·

• omni: INT4 per-block zero-point GEMM + calling adapter (torchao asymmetric INT4)

  #658·JWLHS opened Aug 31, 2026·

  ·

• vLLM: integrate FP16 packed QSA extension

  #651·xiangyuT opened Aug 28, 2026·

  ·

• qwen38 kernels

  #650·hzjane opened Aug 28, 2026·

  ·

• [vLLM][XPU] 修复 torch.compile 跟踪时 LoRA 算子被裁剪的问题

  #635·sicnuyudidi opened Aug 20, 2026·

  ·
  1

• omni: W4A8 s8u4 oneDNN GEMM + s8 activation quantizer (per-block zero point)

  #629·JWLHS opened Aug 19, 2026·

  ·
  13

• docs(vllm): Qwen3.8-27B DFlash drafter recipe — 72.2 tok/s on Arc Pro B70

  #620·rmacy opened Aug 16, 2026·

  ·
  3

• docker: upgrade dev vLLM image to 0.26.0

  #604·gc-fu opened Aug 7, 2026·

  ·
  3

• docs(omni): add HiDream-O1, Krea2 Turbo, Boogu-Image, Mage-Flow and LTX-2.3 to supported models table

  #592·KristianZeng opened Aug 4, 2026·

  ·
  1

• delete comfyui-mixlab-node

  #581·zhm-algo opened Jul 31, 2026·

  ·

• [XPU] Harden fused GDN recurrent state updates

  #570·gc-fu opened Jul 27, 2026·

  ·

• Fix #533: guard out_proj.weight access in GDN out-projection ESIMD probes

  #557·joaovgaraujo opened Jul 22, 2026·

  ·
  1

• refactor(omni): multi-stage Docker build with decoupled builder/runtime

  #549·KristianZeng opened Jul 17, 2026·

  ·
  1

• SGL: Add Gemma4 31b support

  #542·xiangyuT opened Jul 15, 2026·

  ·

• fix(esimd): declare INT4 output mutations

  #537·gc-fu opened Jul 13, 2026·

  ·

• page_attn: make global-max reduction order-invariant

  #535·taste-software opened Jul 12, 2026·

  ·
  1

• Fix/gdn conv seq fp32 state

  #500·liu-shaojun opened Jun 29, 2026·

  ·

• Feat/gemma4 int4 group32

  #499·hzjane opened Jun 29, 2026·

  ·

• Fix Qwen3.5/3.6 load_weights stacked-mapping name mutation (gate_gate_up_proj / qkqkv_proj)

  #475·bongmiin opened Jun 12, 2026·

  ·

• vllm: add MiniCPM-V 4.6 support (MiniCPMV4_6ForConditionalGeneration)

  #472·Zjq9409 opened Jun 12, 2026·

  ·
  1

• Optimize gemma4 continue

  #470·hzjane opened Jun 11, 2026·

  ·