All pull requests

Pull requests

• triton dispatch

  community-contribution
  #152·FeixLiu opened Nov 4, 2026·

  ·
  1

• [CICD] Add Iluvatar CoreX CI configuration and Docker imageTianshu dev

  community-contribution
  #151·wkhylyh-debug opened Nov 2, 2026·

  ·

• Add Hygon CE overrides with FP32 gradients and wave64 launches

  community-contribution
  #149·lihongyang1990 opened Oct 28, 2026·8 of 13 tasks·

  ·

• fix(flagos/gemm): generic_gemm 修复 alpha 丢失、beta 被忽略、转置 3D B 的错误还原

  community-contribution
  #148·lucifer-holly opened Oct 23, 2026·

  ·

• [Fix] support bshd/thd in the flagos backend

  #146·Darryl233 opened Oct 22, 2026Collaborator·1 of 6 tasks·

  ·
  1

• Fix NPU cuDNN version capability query

  community-contribution
  #145·SummerMiku opened Oct 22, 2026·

  ·
  1

• feat/get_num_cublas_streams for npu

  community-contribution
  #144·JerryJonny opened Oct 22, 2026·

  ·
  1

• Fix MetaX TP overlap compatibility with legacy native Userbuffers

  community-contribution
  #139·cjwystz opened Oct 17, 2026·5 of 13 tasks·

  ·

• Feat/multi tensor scale tensor for npu

  community-contribution
  #138·JerryJonny opened Oct 15, 2026·

  ·
  1

• metax: always restore transformer_engine_torch sys.modules alias after loading vendor extension

  community-contribution
  #135·Baibaione opened Oct 12, 2026·

  ·

• fix(cuda): Normalize quantizer dtype before normalization dispatch

  community-contribution
  #133·Chlience opened Oct 9, 2026·

  ·

• Fix FlagOS attention layouts and packed sequence boundaries

  community-contribution
  #131·lihongyang1990 opened Sep 24, 2026·5 of 11 tasks·

  ·

• Revert "CherryPick from main"

  #126·zhaoyinglia opened Sep 20, 2026Collaborator·

  ·

• feat: add NPU fused MoE permutation backend（需仔细审核修改处，包含非backends层修改）

  community-contribution
  #121·SummerMiku opened Sep 11, 2026·

  ·

• fix(npu): add compatibility workaround for Grouped GEMM checkpoint saving

  community-contribution
  #120·JerryJonny opened Sep 11, 2026·7 of 13 tasks·

  ·
  2

• Add NPU vendor implementations for grouped gemm

  community-contribution
  #119·JerryJonny opened Sep 11, 2026·7 of 13 tasks·

  ·

• [NPU] Add Gated Delta Net forward and backward support

  community-contribution
  #117·TianHe007 opened Sep 9, 2026·4 of 13 tasks·

  ·
  1

• 把multi_tensor_scale_tensor接入到TE-NPU

  community-contribution
  #112·SummerMiku opened Sep 2, 2026·

  ·
  2

• [WIP] support ascend high-performance kernel

  #107·Darryl233 opened Aug 19, 2026Collaborator·

  ·

• [WIP] Support transformer_engine_npu

  #106·Darryl233 opened Aug 19, 2026Collaborator·

  ·

• docs(skills): add version-agnostic TE upgrade workflow

  #103·zhaoyinglia opened Aug 10, 2026Collaborator·13 tasks·

  ·

• Upgrade/v2.17 gated 20260810 to main

  #102·zhaoyinglia opened Aug 10, 2026Collaborator·13 tasks·

  ·
  1

• Merge/dev to main 20260807

  #97·zhaoyinglia opened Aug 7, 2026Collaborator·13 tasks·

  ·
  1

• feat(plugin/cuda): add missing UB init ops and fix CommOverlapP2P type conversion

  #96·lxd-cumt opened Aug 7, 2026Collaborator·

  ·
  1

• Add upgrade skills for TransformerEngine-FL upstream sync

  #67·lxd-cumt opened May 15, 2026Collaborator·

  ·
  1