Minimum supported versions

Workflow runs for Minimum supported versions

• [PyTorch][Common] Add MXFP8 and NVFP4 QDQ for mixed-precision GEMM

  currently running:
  Minimum supported versions#1843:Pull request#3633synchronizebynegvet
  negvet:nvfp4_qdq

  Oct 16, 2026In progress
  More actions for workflow run 1843

• [PyTorch] Avoid concatenation copies for compiled split Linear parameters

  currently running:
  Minimum supported versions#1842:Pull request#3623synchronizebypggPL
  pggPL:split_parameters_linear

  Oct 16, 2026In progress
  More actions for workflow run 1842

• [PyTorch] Add torch.compile support for fused MoE routing

  completed successfully:
  Minimum supported versions#1841:Pull request#3691synchronizebypggPL
  pggPL:fused_router_compile

  Oct 16, 202613m 14s
  More actions for workflow run 1841

• [PyTorch] Add torch.compile support for fused MoE routing

  cancelled:
  Minimum supported versions#1840:Pull request#3691synchronizebypggPL
  pggPL:fused_router_compile

  Oct 16, 20267m 43s
  More actions for workflow run 1840

• Grouped Linear CPU Overhead Reductions By Removing Unnecessary Python Logic

  completed successfully:
  Minimum supported versions#1839:Pull request#3613synchronizebyzhongbozhu
  zhongbozhu:optimize/grouped-linear-cpu-overhead

  Oct 16, 202613m 32s
  More actions for workflow run 1839

• [PyTorch] Quantize distributed BF16 weights in the fused grouped MLP

  completed successfully:
  Minimum supported versions#1838:Pull request#3695synchronizebypre-commit-ci[bot]Bot
  xrennvidia:xren/fused_grouped_mlp_gtp_bf16_weight_requant

  Oct 16, 20261h 2m 56s
  More actions for workflow run 1838

• [PyTorch] Quantize distributed BF16 weights in the fused grouped MLP

  cancelled:
  Minimum supported versions#1837:Pull request#3695synchronizebyxrennvidia
  xrennvidia:xren/fused_grouped_mlp_gtp_bf16_weight_requant

  Oct 16, 20262m 1s
  More actions for workflow run 1837

• [PyTorch] Quantize distributed BF16 weights in the fused grouped MLP

  cancelled:
  Minimum supported versions#1836:Pull request#3695synchronizebyxrennvidia
  xrennvidia:xren/fused_grouped_mlp_gtp_bf16_weight_requant

  Oct 16, 202611m 26s
  More actions for workflow run 1836

• [PyTorch] Quantize distributed BF16 weights in the fused grouped MLP

  cancelled:
  Minimum supported versions#1835:Pull request#3695openedbyxrennvidia
  xrennvidia:xren/fused_grouped_mlp_gtp_bf16_weight_requant

  Oct 16, 20268m 55s
  More actions for workflow run 1835

• [PyTorch] Type the grouped wgrad from main_grad for distributed weights

  completed successfully:
  Minimum supported versions#1834:Pull request#3397synchronizebyfanshiqing
  fanshiqing:shiqingf/fix_gtp_grad_reduce_in_fp32

  Oct 16, 202650m 50s
  More actions for workflow run 1834

• Grouped Linear CPU Overhead Reductions By Removing Unnecessary Python Logic

  completed successfully:
  Minimum supported versions#1833:Pull request#3613synchronizebypre-commit-ci[bot]Bot
  zhongbozhu:optimize/grouped-linear-cpu-overhead

  Oct 16, 202615m 32s
  More actions for workflow run 1833

• Grouped Linear CPU Overhead Reductions By Removing Unnecessary Python Logic

  cancelled:
  Minimum supported versions#1832:Pull request#3613synchronizebyzhongbozhu
  zhongbozhu:optimize/grouped-linear-cpu-overhead

  Oct 16, 20262m 39s
  More actions for workflow run 1832

• [PyTorch] Share CUDA graph memory across dynamic CP variants

  cancelled:
  Minimum supported versions#1831:Pull request#3353synchronizebyxiaoyao0115
  xiaoyao0115:test/dcp-cg-gap-fix

  Oct 16, 20262s
  More actions for workflow run 1831

• [PyTorch] Share CUDA graph memory across dynamic CP variants

  completed successfully:
  Minimum supported versions#1830:Pull request#3353synchronizebyxiaoyao0115
  xiaoyao0115:test/dcp-cg-gap-fix

  Oct 16, 202652m 37s
  More actions for workflow run 1830

• [PyTorch] Keep DistributedWeight objects on ctx across saved-tensor hooks

  completed successfully:
  Minimum supported versions#1829:Pull request#3545synchronizebyxrennvidia
  xrennvidia:xren/dist-weight-saved-tensor-hooks

  Oct 16, 202614m 24s
  More actions for workflow run 1829

• [Common] Store swizzled MXFP8 rowwise scales directly in the register-resident

  completed successfully:
  Minimum supported versions#1828:Pull request#3694openedbykainzhong
  kainzhong:mxfp8_g2r_rowwise_swizzled_direct_scale_store

  Oct 16, 202650m 40s
  More actions for workflow run 1828

• [DRAFT] Switch to Python binding for cuDNN frontend

  completed successfully:
  Minimum supported versions#1827:Pull request#3693openedbyvcherepanov-nv
  vcherepanov-nv:py-all

  Oct 16, 202653m 21s
  More actions for workflow run 1827

• [Common] More CuTeDSL MXFP8 specialized kernels

  completed successfully:
  Minimum supported versions#1826:Pull request#3642synchronizebykainzhong
  kainzhong:more_cutedsl_mxfp8_specialized

  Oct 15, 202620m 38s
  More actions for workflow run 1826

• [Common] More CuTeDSL MXFP8 specialized kernels

  cancelled:
  Minimum supported versions#1825:Pull request#3642synchronizebykainzhong
  kainzhong:more_cutedsl_mxfp8_specialized

  Oct 15, 20262m 20s
  More actions for workflow run 1825

• [Common] More CuTeDSL MXFP8 specialized kernels

  cancelled:
  Minimum supported versions#1824:Pull request#3642synchronizebykainzhong
  kainzhong:more_cutedsl_mxfp8_specialized

  Oct 15, 20262m 40s
  More actions for workflow run 1824

• [Common] Add benchmark scripts for CuTeDSL

  completed successfully:
  Minimum supported versions#1823:Pull request#3688synchronizebykainzhong
  kainzhong:benchmark_for_cutedsl

  Oct 15, 202613m 15s
  More actions for workflow run 1823

• [Common] Add benchmark scripts for CuTeDSL

  cancelled:
  Minimum supported versions#1822:Pull request#3688synchronizebykainzhong
  kainzhong:benchmark_for_cutedsl

  Oct 15, 202612m 40s
  More actions for workflow run 1822

• [Common] Add benchmark scripts for CuTeDSL

  completed successfully:
  Minimum supported versions#1821:Pull request#3688synchronizebykainzhong
  kainzhong:benchmark_for_cutedsl

  Oct 15, 202614m 14s
  More actions for workflow run 1821

• [Common] Add benchmark scripts for CuTeDSL

  cancelled:
  Minimum supported versions#1820:Pull request#3688synchronizebykainzhong
  kainzhong:benchmark_for_cutedsl

  Oct 15, 20266m 30s
  More actions for workflow run 1820

• [PyTorch] Pass canonical layouts to cuDNN grouped GEMM in fused grouped MLP

  completed successfully:
  Minimum supported versions#1819:Pull request#3641synchronizebyhwanseoc
  hwanseoc:hwanseoc/glu-canonical-layout

  Oct 15, 202614m 2s
  More actions for workflow run 1819