Minimum supported versions

Workflow runs for Minimum supported versions

• [Common] Add explicit CUDA state lifecycle APIs for in-process restarts

  currently running:
  Minimum supported versions#2271:Pull request#3779openedbyOleg-Goncharov
  Oleg-Goncharov:pr_te_init

  Oct 29, 2026In progress
  More actions for workflow run 2271

• [Common] Add caller-owned workspaces with per-stream resource isolation

  completed successfully:
  Minimum supported versions#2270:Pull request#3766synchronizebyOleg-Goncharov
  Oleg-Goncharov:pr_resource_workspace

  Oct 29, 202613m 27s
  More actions for workflow run 2270

• [PyTorch] [torch.compile] torch.compile support for LayerNormLinear and LayerNormMLP

  completed successfully:
  Minimum supported versions#2269:Pull request#3534synchronizebypggPL
  pggPL:layernorm_compile

  Oct 29, 202615m 4s
  More actions for workflow run 2269

• [PyTorch] Support split Linear parameters with torch.compile

  completed successfully:
  Minimum supported versions#2268:Pull request#3623synchronizebypggPL
  pggPL:split_parameters_linear

  Oct 29, 202613m 16s
  More actions for workflow run 2268

• [PyTorch] Support split Linear parameters with torch.compile

  completed successfully:
  Minimum supported versions#2267:Pull request#3623synchronizebypggPL
  pggPL:split_parameters_linear

  Oct 29, 202614m 40s
  More actions for workflow run 2267

• [Common] Add caller-owned workspaces with per-stream resource isolation

  completed successfully:
  Minimum supported versions#2266:Pull request#3766synchronizebyOleg-Goncharov
  Oleg-Goncharov:pr_resource_workspace

  Oct 29, 202647m 57s
  More actions for workflow run 2266

• [PyTorch] Support split Linear parameters with torch.compile

  completed successfully:
  Minimum supported versions#2265:Pull request#3623synchronizebypggPL
  pggPL:split_parameters_linear

  Oct 29, 202614m 43s
  More actions for workflow run 2265

• Fix grouped MXFP8 quantization race on concurrent CUDA streams

  failed:
  Minimum supported versions#2264:Pull request#3774synchronizebyXFDG
  XFDG:fix/3630-grouped-tma-concurrency

  Oct 29, 20261h 27m 59s
  More actions for workflow run 2264

• Fix grouped MXFP8 dequantize race on concurrent CUDA streams

  failed:
  Minimum supported versions#2263:Pull request#3775synchronizebypre-commit-ci[bot]Bot
  XFDG:fix/3630-grouped-dequantize-concurrency

  Oct 29, 20262h 25m 44s
  More actions for workflow run 2263

• Fix grouped MXFP8 dequantize race on concurrent CUDA streams

  failed:
  Minimum supported versions#2262:Pull request#3775openedbyXFDG
  XFDG:fix/3630-grouped-dequantize-concurrency

  Oct 29, 20262h 32m 22s
  More actions for workflow run 2262

• Accumulate delayed wgrads into shared (tied) parameters

  completed successfully:
  Minimum supported versions#2261:Pull request#3552synchronizebydenera
  0z5a:tied-delayed-wgrad-accumulate

  Oct 29, 202612m 19s
  More actions for workflow run 2261

• Fix grouped MXFP8 quantization race on concurrent CUDA streams

  failed:
  Minimum supported versions#2260:Pull request#3774synchronizebypre-commit-ci[bot]Bot
  XFDG:fix/3630-grouped-tma-concurrency

  Oct 29, 20261h 52m 2s
  More actions for workflow run 2260

• Fix grouped MXFP8 quantization race on concurrent CUDA streams

  failed:
  Minimum supported versions#2259:Pull request#3774openedbyXFDG
  XFDG:fix/3630-grouped-tma-concurrency

  Oct 29, 20261h 53m 22s
  More actions for workflow run 2259

• [PyTorch]Validate grouped split alignment on device

  completed successfully:
  Minimum supported versions#2258:Pull request#3745synchronizebyvthumbe1503
  vthumbe1503:validate-grouped-split-alignment

  Oct 29, 202643m 5s
  More actions for workflow run 2258

• [PyTorch]Validate grouped split alignment on device

  cancelled:
  Minimum supported versions#2257:Pull request#3745synchronizebyvthumbe1503
  vthumbe1503:validate-grouped-split-alignment

  Oct 29, 20262m 42s
  More actions for workflow run 2257

• [Common] Optimize NVFP4 RHT cast-fusion kernels

  requires action with the application:
  Minimum supported versions#2256:Pull request#3740synchronizebyksivaman
  CalebDu:caleb/fix_hadamard

  Oct 29, 2026Action required
  More actions for workflow run 2256

• [PyTorch] Add BF16 fused grouped MLP support

  completed successfully:
  Minimum supported versions#2255:Pull request#3646synchronizebytimmoon10
  tingyangk:tingyangk/add-fused-bf16-grouped-mlp

  Oct 29, 202612m 15s
  More actions for workflow run 2255

• Grouped Linear CPU Overhead Reductions By Removing Unnecessary Python Logic

  completed successfully:
  Minimum supported versions#2254:Pull request#3613synchronizebytimmoon10
  zhongbozhu:optimize/grouped-linear-cpu-overhead

  Oct 29, 202624m 50s
  More actions for workflow run 2254

• Add code quality guidelines to AGENTS.md

  completed successfully:
  Minimum supported versions#2253:Pull request#3773openedbytimmoon10
  timmoon10:tmoon/agents-md-code-quality

  Oct 29, 202614m 25s
  More actions for workflow run 2253

• [PyT] Add context-parallel attention benchmark results

  failed:
  Minimum supported versions#2252:Pull request#3582synchronizebysudhakarsingh27
  sudhakarsingh27:cp_bench

  Oct 28, 20268m 33s
  More actions for workflow run 2252

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

  completed successfully:
  Minimum supported versions#2251:Pull request#3641synchronizebyhwanseoc
  hwanseoc:hwanseoc/glu-canonical-layout

  Oct 28, 202615m 11s
  More actions for workflow run 2251

• [Common] More CuTeDSL MXFP8 specialized kernels

  completed successfully:
  Minimum supported versions#2250:Pull request#3642synchronizebytdophung
  kainzhong:more_cutedsl_mxfp8_specialized

  Oct 28, 202615m 28s
  More actions for workflow run 2250

• Link tests executables against libpython when CuTeDSL is enabled (#3769)

  completed successfully:
  Minimum supported versions#2249:Commitdfc4158pushedbyfheinecke
  main

  Oct 28, 202611m 51s
  More actions for workflow run 2249

• [Common] More CuTeDSL MXFP8 specialized kernels

  cancelled:
  Minimum supported versions#2248:Pull request#3642synchronizebypre-commit-ci[bot]Bot
  kainzhong:more_cutedsl_mxfp8_specialized

  Oct 28, 202615m 58s
  More actions for workflow run 2248

• [Common] More CuTeDSL MXFP8 specialized kernels

  cancelled:
  Minimum supported versions#2247:Pull request#3642synchronizebykainzhong
  kainzhong:more_cutedsl_mxfp8_specialized

  Oct 28, 20262m 44s
  More actions for workflow run 2247