CUDA Plugin Linux CI

Workflow runs for CUDA Plugin Linux CI

• [WebGPU] Fuse QLoRA QKV and MLP projections

  currently running:
  CUDA Plugin Linux CI#14045:Pull request#33901synchronizebyqjia7
  qjia7:feat/webgpu-qlora-mlp-fusion

  Nov 9, 2026In progress
  More actions for workflow run 14045

• [MLAS][KleidiAI] Integrate Depthwise 3x3 Conv Kernel

  requires action with the application:
  CUDA Plugin Linux CI#14044:Pull request#29441synchronizebymelkap01-Arm
  JonathanC-ARM:jonclo01/dw_conv_integration

  Nov 9, 2026-1s
  More actions for workflow run 14044

• [CUDA] Add opt-in direct scales for FP16 group-32 MatMulNBits M=1

  requires action with the application:
  CUDA Plugin Linux CI#14043:Pull request#33586synchronizebysaiyambharara
  saiyambharara:perf/32382-direct-scales

  Nov 9, 2026-1s
  More actions for workflow run 14043

• Register the existing CUDA BF16 Less kernel

  queued:
  CUDA Plugin Linux CI#14042:Pull request#34005openedbybaijumeswani
  baijumeswani/cuda-bf16-less

  Nov 9, 2026Queued
  More actions for workflow run 14042

• Fix GeluFusion skipping the fusion when the GELU input is produced by Mul

  requires action with the application:
  CUDA Plugin Linux CI#14041:Pull request#33911synchronizebysandeepvijayarao09
  sandeepvijayarao09:fix-gelu-fusion-mul-input

  Nov 9, 2026Action required
  More actions for workflow run 14041

• Fix the CPU Pad slowdown for tall inputs with a small inner extent

  requires action with the application:
  CUDA Plugin Linux CI#14040:Pull request#34004openedbyJonSnow1807
  JonSnow1807:fix-pad-constant-rows-perf-33861

  Nov 9, 2026Action required
  More actions for workflow run 14040

• Reduce native packed decode tile and FC2 loader overhead

  currently running:
  CUDA Plugin Linux CI#14039:Pull request#34003stackedbyrui-ren
  rui-ren/bolt-packed-kernel-gap-on-34002-20261109

  Nov 9, 2026In progress
  More actions for workflow run 14039

• [WebGPU] Support signed 8-bit and 64-bit unary math

  completed successfully:
  CUDA Plugin Linux CI#14038:Pull request#33958synchronizebysushraja-msft
  user/sushraja/webgpu-integer-unary

  Nov 9, 20261h 24m 40s
  More actions for workflow run 14038

• Prepack constant CUDA SiLU-gated MoE weights

  completed successfully:
  CUDA Plugin Linux CI#14037:Pull request#33882synchronizebytianleiwu
  copilot/cuda-moe-performance-issue

  Nov 9, 20261h 33m 2s
  More actions for workflow run 14037

• Fix GQA shape inference for shape-derived int32 sequence lengths

  completed successfully:
  CUDA Plugin Linux CI#14036:Pull request#33884synchronizebytianleiwu
  copilot/fix-group-query-attention-load

  Nov 9, 20261h 18m 35s
  More actions for workflow run 14036

• Reuse device streams across threads when graph capture is disabled

  completed successfully:
  CUDA Plugin Linux CI#14035:Pull request#33053synchronizebytianleiwu
  copilot/cuda-fix-gpu-memory-issue

  Nov 9, 20261h 15m 49s
  More actions for workflow run 14035

• [WebGPU] Support signed 8-bit and 64-bit unary math

  cancelled:
  CUDA Plugin Linux CI#14034:Pull request#33958synchronizebysushraja-msft
  user/sushraja/webgpu-integer-unary

  Nov 9, 202615m 36s
  More actions for workflow run 14034

• Fix Split num_outputs with an empty optional input

  failed:
  CUDA Plugin Linux CI#14033:Pull request#32972synchronizebytianleiwu
  fix-split-empty-optional-input

  Nov 9, 20261h 15m 1s
  More actions for workflow run 14033

• Speed up native Bolt packed projections and IQ2 experts

  completed successfully:
  CUDA Plugin Linux CI#14032:Pull request#34002stackedbyrui-ren
  rui-ren/bolt-packed-latency-on-33994-20261109

  Nov 9, 20261h 22m 32s
  More actions for workflow run 14032

• Reuse device streams across threads when graph capture is disabled

  cancelled:
  CUDA Plugin Linux CI#14031:Pull request#33053synchronizebytianleiwu
  copilot/cuda-fix-gpu-memory-issue

  Nov 9, 20261h 11m 52s
  More actions for workflow run 14031

• Prepack constant CUDA SiLU-gated MoE weights

  completed successfully:
  CUDA Plugin Linux CI#14030:Pull request#33882synchronizebytianleiwu
  copilot/cuda-moe-performance-issue

  Nov 9, 20261h 19m 36s
  More actions for workflow run 14030

• Fix GQA shape inference for shape-derived int32 sequence lengths

  completed successfully:
  CUDA Plugin Linux CI#14029:Pull request#33884synchronizebytianleiwu
  copilot/fix-group-query-attention-load

  Nov 9, 20261h 16m 39s
  More actions for workflow run 14029

• [WebGPU] Support signed 8-bit and 64-bit unary math

  completed successfully:
  CUDA Plugin Linux CI#14028:Pull request#33958synchronizebysushraja-msft
  user/sushraja/webgpu-integer-unary

  Nov 9, 20261h 17m 53s
  More actions for workflow run 14028

• Reject zero divisors in CUDA Mod kernels

  completed successfully:
  CUDA Plugin Linux CI#14027:Pull request#32546synchronizebyCopilotAI
  copilot/fix-cuda-mod-kernel-zero-divisor-check

  Nov 9, 20261h 20m 11s
  More actions for workflow run 14027

• [WebGPU] Support signed 8-bit and 64-bit unary math

  cancelled:
  CUDA Plugin Linux CI#14026:Pull request#33958synchronizebysushraja-msft
  user/sushraja/webgpu-integer-unary

  Nov 9, 202618m 21s
  More actions for workflow run 14026

• [WebGPU] Support signed 8-bit and 64-bit unary math

  cancelled:
  CUDA Plugin Linux CI#14025:Pull request#33958synchronizebysushraja-msft
  user/sushraja/webgpu-integer-unary

  Nov 9, 202628m 17s
  More actions for workflow run 14025

• Restore vectorized PRelu for per-channel slopes on CPU

  requires action with the application:
  CUDA Plugin Linux CI#14024:Pull request#33996synchronizebyJonSnow1807
  JonSnow1807:fix-prelu-per-channel-perf-33831

  Nov 9, 2026Action required
  More actions for workflow run 14024

• [CUDA] Cache cuDNN frontend Conv execution plans

  completed successfully:
  CUDA Plugin Linux CI#14023:Pull request#33123synchronizebytianleiwu
  tlwu/cuda-conv-plan-cache

  Nov 9, 20261h 24m 28s
  More actions for workflow run 14023

• Add GQA fusion optional-input preservation regressions

  completed successfully:
  CUDA Plugin Linux CI#14022:Pull request#29552synchronizebytianleiwu
  copilot/fix-group-query-attention-fusion

  Nov 9, 20261h 22m 2s
  More actions for workflow run 14022

• Fix Split num_outputs with an empty optional input

  completed successfully:
  CUDA Plugin Linux CI#14021:Pull request#32972synchronizebytianleiwu
  fix-split-empty-optional-input

  Nov 9, 20261h 24m 10s
  More actions for workflow run 14021