Build

Workflow runs for Build

• Thread tanh logit softcapping through FlashAttention (FA2 and FA3)

  cancelled:
  Build#20874:Pull request#3391synchronizebynvegesna-netizen
  nvegesna-netizen:nvegesna/gemma2-softcap-core

  Sep 9, 20261h 0m 6s
  More actions for workflow run 20874

• Thread tanh logit softcapping through FlashAttention (FA2 and FA3)

  cancelled:
  Build#20873:Pull request#3391synchronizebynvegesna-netizen
  nvegesna-netizen:nvegesna/gemma2-softcap-core

  Sep 9, 202623m 35s
  More actions for workflow run 20873

• [Common] Experimental CuTeDSL MXFP8 backends in C++ via TVM-FFI

  failed:
  Build#20872:Pull request#3137synchronizebykainzhong
  kainzhong:cutedsl_mxfp8_common

  Sep 9, 20269m 44s
  More actions for workflow run 20872

• Prototype NVFP4 with FP8 UE5M3 block scales

  failed:
  Build#20871:Pull request#3325synchronizebytimmoon10
  timmoon10:nvfp4-ue5m3-prototype

  Sep 9, 20266h 0m 8s
  More actions for workflow run 20871

• Skip fetching unused next tile in Mxfp8 quantization kernel

  failed:
  Build#20870:Pull request#3500openedbyWanZzzzzz
  WanZzzzzz:quant-opt-skip

  Sep 9, 20266h 0m 17s
  More actions for workflow run 20870

• [PyTorch] DeepSeekV3Layer: full MoE transformer layer (MLA + DeepSeek MoE)

  failed:
  Build#20869:Pull request#3471synchronizebypggPL
  pggPL:deepseek_v3_layer

  Sep 9, 20266h 5m 16s
  More actions for workflow run 20869

• [PyTorch] DeepSeekV3Layer: full MoE transformer layer (MLA + DeepSeek MoE)

  cancelled:
  Build#20868:Pull request#3471synchronizebypggPL
  pggPL:deepseek_v3_layer

  Sep 9, 202658m 35s
  More actions for workflow run 20868

• [JAX] Add EP borrowed-comm path via XLA FFI collectives extension

  failed:
  Build#20867:Pull request#3452synchronizebyphu0ngng
  phu0ngng:ep_jax_single_process

  Sep 9, 20266h 0m 6s
  More actions for workflow run 20867

• [All] Refactor fused attention APIs with cuDNN-frontend support checks and opaque config/params handles

  failed:
  Build#20866:Pull request#2964synchronizebypre-commit-ci[bot]Bot
  cyanguwa:fe_check_support

  Sep 9, 20266h 0m 26s
  More actions for workflow run 20866

• [PyTorch] DeepSeekV3Layer: full MoE transformer layer (MLA + DeepSeek MoE)

  cancelled:
  Build#20865:Pull request#3471synchronizebypggPL
  pggPL:deepseek_v3_layer

  Sep 9, 20268m 49s
  More actions for workflow run 20865

• [PyTorch] DeepSeekV3Layer: full MoE transformer layer (MLA + DeepSeek MoE)

  cancelled:
  Build#20864:Pull request#3471synchronizebypggPL
  pggPL:deepseek_v3_layer

  Sep 9, 20261m 51s
  More actions for workflow run 20864

• [PyTorch] DeepSeekV3Layer: full MoE transformer layer (MLA + DeepSeek MoE)

  cancelled:
  Build#20863:Pull request#3471synchronizebypggPL
  pggPL:deepseek_v3_layer

  Sep 9, 20262m 14s
  More actions for workflow run 20863

• [All] Refactor fused attention APIs with cuDNN-frontend support checks and opaque config/params handles

  cancelled:
  Build#20862:Pull request#2964synchronizebycyanguwa
  cyanguwa:fe_check_support

  Sep 9, 20267m 11s
  More actions for workflow run 20862

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

  failed:
  Build#20861:Pull request#3493synchronizebypggPL
  pggPL:layernorm_compile_refactor

  Sep 9, 20266h 5m 8s
  More actions for workflow run 20861

• [PyTorch] DeepSeekV3Layer: full MoE transformer layer (MLA + DeepSeek MoE)

  cancelled:
  Build#20860:Pull request#3471synchronizebypggPL
  pggPL:deepseek_v3_layer

  Sep 9, 202626m 21s
  More actions for workflow run 20860

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

  cancelled:
  Build#20859:Pull request#3493synchronizebypggPL
  pggPL:layernorm_compile_refactor

  Sep 9, 202639m 25s
  More actions for workflow run 20859

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

  cancelled:
  Build#20858:Pull request#3493synchronizebypggPL
  pggPL:layernorm_compile_refactor

  Sep 9, 20262m 51s
  More actions for workflow run 20858

• [PyTorch] torch.compile for an OperationFuser group holding one operation

  failed:
  Build#20857:Pull request#3496synchronizebypggPL
  pggPL:ops_fuser_compile_main

  Sep 9, 20262h 33m 50s
  More actions for workflow run 20857

• [Docs] Add Mixture of Experts guide

  failed:
  Build#20856:Pull request#3494synchronizebypggPL
  pggPL:docs_moe

  Sep 9, 20266h 0m 8s
  More actions for workflow run 20856

• Add native transport for dynamic context parallelism

  failed:
  Build#20855:Pull request#3489synchronizebyxiaoyao0115
  xiaoyao0115:dcp-native-transport-minimal

  Sep 9, 20261h 19m 22s
  More actions for workflow run 20855

• [PyTorch] torch.compile for an OperationFuser group holding one operation

  cancelled:
  Build#20854:Pull request#3496synchronizebypggPL
  pggPL:ops_fuser_compile_main

  Sep 9, 202639m 33s
  More actions for workflow run 20854

• [PyTorch] Support FA3/FA4 with pad_between_seqs=True

  failed:
  Build#20853:Pull request#3205synchronizebyxiaoyao0115
  xiaoyao0115:feat/fa-pad-between-seq

  Sep 9, 20266h 0m 9s
  More actions for workflow run 20853

• [All] Refactor fused attention APIs with cuDNN-frontend support checks and opaque config/params handles

  cancelled:
  Build#20852:Pull request#2964synchronizebycyanguwa
  cyanguwa:fe_check_support

  Sep 9, 20262h 20m 32s
  More actions for workflow run 20852

• [PyTorch] torch.compile for an OperationFuser group holding one operation

  cancelled:
  Build#20851:Pull request#3496synchronizebypggPL
  pggPL:ops_fuser_compile_main

  Sep 9, 20262h 31m 48s
  More actions for workflow run 20851

• Fix hybrid extra state size mismatch

  failed:
  Build#20850:Pull request#3499openedbynegvet
  negvet:fix_extra_state_size_mismatch_hybrid

  Sep 9, 20266h 0m 35s
  More actions for workflow run 20850