All pull requests

Pull requests

• feat(spark): support Spark-X2.5-1.7B on CPU

  #709·Aharrypotter opened Sep 11, 2026Contributor·

  ·
  8

• feat(preprocessor): unified chat templates with vendored jinja.cpp and per-model backend selection

  #707·Aharrypotter opened Sep 5, 2026Contributor·

  ·
  1

• Feat/shadownpu sparse attention

  #703·yinwangsong opened Aug 24, 2026·5 of 9 tasks·

  ·
  1

• feat(cpu): add Ling-3.0-tiny mobile CPU support

  #699·Aharrypotter opened Aug 12, 2026Contributor·

  ·
  12

• Add CPU GDN kernel for Qwen3.5 with correctness test and benchmark

  #687·Misscheng-eng opened Jun 21, 2026·

  ·
  8

• fix(cpu): preserve raw fp16 bits in x86 quantize conversion

  #685·kkkzbh opened Jun 13, 2026·

  ·
  5

• fix(build): build xxHash objects with PIC for MllmRT

  #683·kkkzbh opened Jun 12, 2026·

  ·
  1

• Modify llama tokenizer, llama modeling, qwen2 modeling.

  #677·ZhangJinghe-AI opened May 19, 2026·

  ·
  7

• Add Qwen2-VL full QNN AOT example

  #676·twlddd opened May 18, 2026·

  ·
  19

• Add QNN AOT support for visual graph lowering

  #675·twlddd opened May 18, 2026·

  ·
  9

• Fix QNN AOT runtime and lowering issues

  #674·twlddd opened May 18, 2026·

  ·
  2

• feat(pymllm, gemma3n): Add Gemma3n text-only native server path

  #672·Grape203 opened May 1, 2026Contributor·

  ·
  12

• Added Qwen3.5 0.8B

  #657·FarmersWrap opened Mar 22, 2026·

  ·
  14

• fix(pymllm): fix RadixCache evictable size accounting bug

  #656·FarmersWrap opened Mar 20, 2026·1 of 2 tasks·

  ·
  16

• fix(pymllm): reduce scheduler CPU busy-loop from 100% to ~2% during decode

  #655·FarmersWrap opened Mar 20, 2026·4 tasks done·

  ·
  15

• feat: add probing service and update mobile adapter

  #652·yuerqiqi opened Mar 14, 2026Contributor·

  ·
  8

• feat(kernel): add vocab embedding CUDA kernels

  #645·huangzhenhua111 opened Mar 2, 2026·

  ·
  1

• feat(benchmark): Add CPU benchmark tool with context length sweep

  #639·huangzhenhua111 opened Feb 14, 2026·

  ·
  46

• feat : Add Support of Qwen2.5Omni Model and MiniCPM-o-4_5 Model

  #612·KKkai0315 opened Jan 23, 2026Contributor·

  ·
  43

• feat: add support of internlm2.5-1.8B-chat model

  #515·Sp0tless opened Nov 12, 2025Contributor·

  ·
  9

• fix small bugs with run_qwen_npu.sh.

  #295·yinrun opened Jun 12, 2025·

  ·