pre-commit

Workflow runs for pre-commit

• [Model][SM70][SM75] GLM-5.3-Flash-FP4-Local: compressed expert scales, 8-bit QPN8 projections

  failed:
  pre-commit#4158:Pull request#1286openedbyPeuqui
  Peuqui:pr-glm53

  Oct 28, 20269m 45s
  More actions for workflow run 4158

• [Perf][SM70][SM75] Opt-in tensor-core Triton routes for DeepSeek-V4 prefill on Volta and Turing

  failed:
  pre-commit#4157:Pull request#1282synchronizebyPeuqui
  Peuqui:pr-triton-sm7x

  Oct 28, 20269m 41s
  More actions for workflow run 4157

• [Perf][SM70][DeepSeek-V4] Long-prefill indexer and sparse attention: vectorized epilogue, fused softmax, Turing scorer

  currently running:
  pre-commit#4156:Pull request#1280synchronizebyPeuqui
  Peuqui:pr-dsv4-prefill-stacked

  Oct 28, 2026In progress
  More actions for workflow run 4156

• [Spec Decode][MRV2] Reduced draft vocabulary for MTP (backport of vllm#59740)

  failed:
  pre-commit#4155:Pull request#1218synchronizebyPeuqui
  Peuqui:pr-mtp-draft-token-map

  Oct 28, 20269m 50s
  More actions for workflow run 4155

• [Core][Spec Decode] Lookahead block hashes for EAGLE/MTP prefix hits (backport of vllm#61589)

  failed:
  pre-commit#4154:Pull request#1276synchronizebyPeuqui
  Peuqui:pr-lookahead-on-1240

  Oct 28, 20269m 34s
  More actions for workflow run 4154

• [Perf][Qwen4Exp] Overlap the PLE gather of the next prefill chunk

  failed:
  pre-commit#4153:Pull request#1217synchronizebyPeuqui
  Peuqui:pr-ple-prefill-overlap

  Oct 28, 20267m 46s
  More actions for workflow run 4153

• [Perf][Qwen4Exp] Bound the page prefetch of the mapped PLE row gather

  failed:
  pre-commit#4152:Pull request#1235synchronizebyPeuqui
  Peuqui:pr-ple-prefetch-window

  Oct 28, 20265m 46s
  More actions for workflow run 4152

• [Perf][SM70] One grouped GEMM for DeepSeek-V4 wo_a; opt-in TurboMind tuning freeze after warmup

  failed:
  pre-commit#4151:Pull request#1273synchronizebyPeuqui
  Peuqui:pr-dsv4-woa-freeze

  Oct 28, 20268m 44s
  More actions for workflow run 4151

• [Perf][SM70][SM75] Opt-in tensor-core Triton routes for DeepSeek-V4 prefill on Volta and Turing

  failed:
  pre-commit#4150:Pull request#1282synchronizebyPeuqui
  Peuqui:pr-triton-sm7x

  Oct 28, 202610m 3s
  More actions for workflow run 4150

• [Perf][SM70][DeepSeek-V4] Long-prefill indexer and sparse attention: vectorized epilogue, fused softmax, Turing scorer

  failed:
  pre-commit#4149:Pull request#1280synchronizebyPeuqui
  Peuqui:pr-dsv4-prefill-stacked

  Oct 28, 20265m 53s
  More actions for workflow run 4149

• [Perf][SM75] Turing prefill attention; overlap TP prefill all-reduces on GPUs without P2P

  failed:
  pre-commit#4148:Pull request#1279synchronizebyPeuqui
  Peuqui:pr-tp-prefill

  Oct 28, 202610m 4s
  More actions for workflow run 4148

• [Kernel] Two-rank FP16 all-reduce through pinned host memory for GPUs without P2P

  failed:
  pre-commit#4147:Pull request#1216synchronizebyPeuqui
  Peuqui:pr-sm70-host-reduce

  Oct 28, 20265m 53s
  More actions for workflow run 4147

• [Perf][SM75] Turing tensor-core paths for block-FP8 QPN8 and the grouped skinny MoE; bound the MoE grid

  failed:
  pre-commit#4146:Pull request#1278synchronizebyPeuqui
  Peuqui:pr-turing-kernels

  Oct 28, 20265m 55s
  More actions for workflow run 4146

• [Bugfix] Keep the pynccl all-reduce out-of-place by default (#1284)

  failed:
  pre-commit#4145:Commitabec8bapushedbyyangzhuxinyzx
  main

  Oct 28, 20268m 19s
  More actions for workflow run 4145

• [Bugfix] Keep the pynccl all-reduce out-of-place by default

  failed:
  pre-commit#4144:Pull request#1284synchronizebyyangzhuxinyzx
  claude/inplace-allreduce-default

  Oct 28, 202622m 1s
  More actions for workflow run 4144

• [Bugfix] Keep the pynccl all-reduce out-of-place by default

  cancelled:
  pre-commit#4143:Pull request#1284synchronizebyyangzhuxinyzx
  claude/inplace-allreduce-default

  Oct 28, 202618m 34s
  More actions for workflow run 4143

• [Bugfix] Keep the pynccl all-reduce out-of-place by default

  failed:
  pre-commit#4142:Pull request#1284openedbyyangzhuxinyzx
  claude/inplace-allreduce-default

  Oct 28, 20269m 43s
  More actions for workflow run 4142

• [Bugfix][Multimodal] Stop the shm processor cache from pinning items …

  failed:
  pre-commit#4141:Commitc59b3c4pushedbyyangzhuxinyzx
  main

  Oct 28, 202615m 44s
  More actions for workflow run 4141

• [Bugfix][QSA] Clamp the ring's cached RoPE positions before they inde…

  cancelled:
  pre-commit#4140:Commit04718eapushedbyyangzhuxinyzx
  main

  Oct 28, 202623s
  More actions for workflow run 4140

• [Bugfix][MLA] Size the chunked-prefill workspace and the indexer pref…

  cancelled:
  pre-commit#4139:Commit089ab49pushedbyyangzhuxinyzx
  main

  Oct 28, 202614s
  More actions for workflow run 4139

• [Bugfix][Loader] Load index shards in subfolders; read U16 and U32 th…

  cancelled:
  pre-commit#4138:Commit691b659pushedbyyangzhuxinyzx
  main

  Oct 28, 202612s
  More actions for workflow run 4138

• [Bugfix][SM70] Classify compact top-k/top-p rows on device instead of…

  cancelled:
  pre-commit#4137:Commit4d5ee06pushedbyyangzhuxinyzx
  main

  Oct 28, 202612s
  More actions for workflow run 4137

• [Spec] Backport disable_eagle_block_drop for EAGLE/MTP prefix hits (#…

  cancelled:
  pre-commit#4136:Commite984902pushedbyyangzhuxinyzx
  main

  Oct 28, 202617s
  More actions for workflow run 4136

• [Perf] Run the fused GDN speculative metadata launch on Turing too (#…

  cancelled:
  pre-commit#4135:Commit4e89c28pushedbyyangzhuxinyzx
  main

  Oct 28, 202615s
  More actions for workflow run 4135

• [Perf] Do not send a request's token history to the V2 runner every d…

  cancelled:
  pre-commit#4134:Commitb6f8980pushedbyyangzhuxinyzx
  main

  Oct 28, 202614s
  More actions for workflow run 4134