Checkstyle

Workflow runs for Checkstyle

• fix(embedding): use int64 offsets in experimental embedding kernels

  requires action with the application:
  Checkstyle#2050:Pull request#1615openedby0xDexe
  0xDexe:fix/experimental-embedding-int64-offsets

  Nov 10, 2026-1s
  More actions for workflow run 2050

• ops: validate targets in LigerVocabParallelCEFunction

  requires action with the application:
  Checkstyle#2049:Pull request#1593synchronizebyASHMEET555
  ASHMEET555:fix/vocab-parallel-ce-target-validation

  Nov 10, 2026-1s
  More actions for workflow run 2049

• fix(compile): use tl.cast for scalar kernel params so torch.compile d…

  completed successfully:
  Checkstyle#2048:Commit42dc23cpushedbygithub-merge-queue[bot]Bot
  main

  Nov 10, 202621s
  More actions for workflow run 2048

• Checkstyle

  completed successfully:
  Checkstyle#2047:Merge group checks requested
  Nov 10, 202620s
  More actions for workflow run 2047

• sm100 CuTe DSL GEMM: wait for the TMEM load before releasing the accumulator stage

  requires action with the application:
  Checkstyle#2046:Pull request#1614openedbybastianhagedorn
  bastianhagedorn:fix/cutedsl-sm100-gemm-tmem-load-wait

  Nov 10, 2026Action required
  More actions for workflow run 2046

• feat(grpo): support top_entropy_quantile in fused linear GRPO loss

  completed successfully:
  Checkstyle#2045:Pull request#1558synchronizebynvdm59
  nvdm59:grpo-top-entropy-quantile

  Nov 10, 202616s
  More actions for workflow run 2045

• feat(grpo): add off-policy sequence masking (OPSM) to fused linear GRPO loss

  completed successfully:
  Checkstyle#2044:Pull request#1557synchronizebynvdm59
  nvdm59:grpo-off-policy-mask

  Nov 10, 202617s
  More actions for workflow run 2044

• fix(fused_linear): pass the column stride when scaling KL and JSD gradients

  requires action with the application:
  Checkstyle#2043:Pull request#1613openedbybhargavchintam
  bhargavchintam:fix/fused-kl-jsd-col-stride

  Nov 9, 2026Action required
  More actions for workflow run 2043

• fix(cross_entropy): reduce per-token loss in fp32 and add an opt-in i…

  completed successfully:
  Checkstyle#2042:Commite1602a4pushedbygithub-merge-queue[bot]Bot
  main

  Nov 9, 202615s
  More actions for workflow run 2042

• Checkstyle

  completed successfully:
  Checkstyle#2041:Merge group checks requested
  Nov 9, 202619s
  More actions for workflow run 2041

• fix(cross_entropy): reduce per-token loss in fp32 and add an opt-in in-place backward for reduction="none"

  completed successfully:
  Checkstyle#2040:Pull request#1574synchronizebyvaibhavjindal
  JobQiu:ce-fp32-loss-uniform-grad

  Nov 9, 202617s
  More actions for workflow run 2040

• fix(utils): restore torch cross-entropy state in revert helpers (#1587)

  requires action with the application:
  Checkstyle#2039:Pull request#1611openedbyYuvrajSinghBhadoria2
  YuvrajSinghBhadoria2:fix/cross-entropy-revert

  Nov 9, 2026Action required
  More actions for workflow run 2039

• fix(utils): skip device_context guard under torch.compile (#1590)

  requires action with the application:
  Checkstyle#2038:Pull request#1610openedbyYuvrajSinghBhadoria2
  YuvrajSinghBhadoria2:fix/device-context-compile

  Nov 9, 2026Action required
  More actions for workflow run 2038

• fix(sparsemax): preserve low-precision gradients by subtracting the mean in fp32

  requires action with the application:
  Checkstyle#2037:Pull request#1609openedbywqh-creator
  wqh-creator:fix/sparsemax-backward-fp32-mean

  Nov 9, 2026Action required
  More actions for workflow run 2037

• fix(dyt): put the row count on grid dimension 0 in the forward launch

  requires action with the application:
  Checkstyle#2036:Pull request#1607openedbybhargavchintam
  bhargavchintam:fix/dyt-grid-row-limit

  Nov 9, 2026Action required
  More actions for workflow run 2036

• fix(group_norm): keep the backward statistics and reductions in fp32

  requires action with the application:
  Checkstyle#2035:Pull request#1608openedbybhargavchintam
  bhargavchintam:fix/group-norm-backward-fp32

  Nov 9, 2026Action required
  More actions for workflow run 2035

• fix(cross_entropy): preserve ignored logits without gradients (#1604)

  completed successfully:
  Checkstyle#2034:Commite44de70pushedbygithub-merge-queue[bot]Bot
  main

  Nov 8, 202618s
  More actions for workflow run 2034

• Checkstyle

  completed successfully:
  Checkstyle#2033:Merge group checks requested
  Nov 8, 202616s
  More actions for workflow run 2033

• Fix Triton RoPE token stride when seq_len is 1 (#1527)

  completed successfully:
  Checkstyle#2032:Commitff8d8dcpushedbygithub-merge-queue[bot]Bot
  main

  Nov 8, 202613s
  More actions for workflow run 2032

• Checkstyle

  completed successfully:
  Checkstyle#2031:Merge group checks requested
  Nov 8, 202617s
  More actions for workflow run 2031

• fix(cross_entropy): preserve ignored logits without gradients

  completed successfully:
  Checkstyle#2030:Pull request#1604openedbyLiRunGuo
  LiRunGuo:fix/cross-entropy-no-grad-preserve-logits

  Nov 7, 202619s
  More actions for workflow run 2030

• test(convergence): make MoE node comparison flip-aware

  completed successfully:
  Checkstyle#2029:Pull request#1568synchronizebyYulong-Cauli
  Yulong-Cauli:test/flip-aware-convergence

  Nov 7, 202617s
  More actions for workflow run 2029

• feat(chunked_loss): add fused linear IRPO (Iterative Reasoning Preference Optimization) loss

  completed successfully:
  Checkstyle#2028:Pull request#1586synchronizebyYulong-Cauli
  Yulong-Cauli:feat/irpo-loss

  Nov 7, 202615s
  More actions for workflow run 2028

• feat(chunked_loss): add fused linear BCO (Binary Classifier Optimization) loss

  completed successfully:
  Checkstyle#2027:Pull request#1585synchronizebyYulong-Cauli
  Yulong-Cauli:feat/bco-loss

  Nov 7, 202614s
  More actions for workflow run 2027

• fix(compile): use tl.cast for scalar kernel params so torch.compile does not fall back

  completed successfully:
  Checkstyle#2026:Pull request#1544synchronizebyAckerMan5659
  AckerMan5659:fix/compile-scalar-cast

  Nov 7, 202620s
  More actions for workflow run 2026