All pull requests

Pull requests

• feat(debug): add generic portable auto capture

  #442·inaniloquentee opened Sep 23, 2026·

  ·
  8

• gemma4-31B arch fingerprint

  gemmaplatform: cuda
  #441·haotianxue31 opened Sep 23, 2026·

  ·
  6

• Sm80 linear locuda: add fused SM80 linear_logp backend for A100gp

  #440·QingYUDC opened Sep 22, 2026·

  ·
  6

• [DSv4][P5-4] MXFP8×MXFP4 Routed Expert grouped GEMM: wgmma triton kernel

  deepseek-P5DSv4platform: triton
  #438·Bignonia7 opened Sep 20, 2026·

  ·
  1

• [DSv4][P5-3] shared_grouped_lora_delta: CUDA + Triton, both batch-invariant

  deepseek-P5DSv4platform: cudaplatform: triton
  #416·ScottNi0809 opened Sep 17, 2026·

  ·
  6

• [DSv4][P1-5] Implement CUDA and Triton rmsnorm_residual

  #414·why1te opened Sep 16, 2026·

  ·
  1

• feat(ascend): Qwen-Image qk_rmsnorm & multi_axis_rope kernels (issue #386)

  Ascendmultimodal
  #411·erfgss opened Sep 13, 2026·

  ·
  1

• Qwen image latent pack unpack

  multimodalplatform: cuda
  #410·nodeeeeee opened Sep 12, 2026·

  ·
  1

• style: format Python sources and benchmark results

  #408·maxiaosong1124 opened Sep 12, 2026·

  ·
  1

• [DSv4][P5-4] MXFP8×MXFP4 Routed Expert grouped GEMM: strict CUDA kernel

  deepseek-P5DSv4platform: cuda
  #399·Bignonia7 opened Sep 10, 2026·

  ·
  2

• feat(musa): add native deterministic gemm kernel

  MUSA
  #395·Arlo-mt opened Sep 9, 2026·

  ·
  4

  

• Musa support native fused logp kernel

  MUSA
  #392·Arlo-mt opened Sep 8, 2026·

  ·
  10

  

• [CI] Fix black formatting on main and pin line-length in pyproject.toml

  #384·Dnoob opened Sep 3, 2026·

  ·
  1

  4

• [DSv4][P1-S0] Start kit for the P1 work package (mHC + RMSNorm)

  deepseek-P1DSv4platform: cuda
  #383·zhangj1an opened Sep 3, 2026·

  ·
  8

• [DSv4][P5-0] Start kit for the P5 work package (MXFP4 Routed Expert + LoRA + Shared Expert)

  deepseek-P5DSv4platform: cuda
  #368·KJLdefeated opened Sep 1, 2026·

  ·
  20

  

• [WS1] Add batch-invariant h_aggregate kernel

  deepseek-P1DSv4platform: cuda
  #366·nodeeeeee opened Aug 30, 2026·

  ·
  2

• docs: add DCO 1.1 text and contributor sign-off guide

  type: ci-cd
  #359·Zhifu-Liu opened Aug 29, 2026·

  ·
  8

• [WS2][Logp] Deterministic config option for operator

  platform: cuda
  #314·KJLdefeated opened Aug 16, 2026·

  ·
  5

• [CI][refator]: migrate GPU workflow

  needs-gpu-ci
  #309·Flink-ddd opened Aug 14, 2026·

  ·
  12

• docs(ws1): batch-invariance contract RFC (#101)

  #264·Billy1900 opened Aug 5, 2026·

  ·
  1

  17

• [WS2][PR1][Logp] Add TP-aware logprob contract and dispatch metadata

  #259·ryankert01 opened Aug 2, 2026·

  ·
  22

• [FEAT][kernels] Add fused ratio-clip-aggregate loss primitive

  needs-gpu-ci
  #255·Chen-BUPT opened Jul 31, 2026·

  ·
  1

  11

• feat(observability): NVTX kernel tracing + Prometheus metrics endpoint

  #250·Billy1900 opened Jul 27, 2026·

  ·
  9

• feat(observability): add NVTX tracing and Prometheus metrics

  #242·BruceLoveDecimal opened Jul 22, 2026·

  ·
  4

• feat(kernels): SM90 TMA + mma.sync FlashAttention, forward + backward (causal, varlen, LSE)

  needs-gpu-cistale
  #237·Billy1900 opened Jul 20, 2026·

  ·
  15