ONNX Runtime DirectML Builds

windows_dml.yml

2500+ workflow runs

fix: make k_quant fit the integer zero point it stores
requires action with the application:
ONNX Runtime DirectML Builds#24876:Pull request#32814synchronizebyykhrustalev
ykhrustalev:ykhrustalev/make-k-quant-fit-the-integer-zero-point
Oct 15, 2026Action required
Validate quantization parameters before QDQ Add, Mul and Concat fusion
requires action with the application:
ONNX Runtime DirectML Builds#24875:Pull request#33397openedbyrookieLiu2018
rookieLiu2018:fix/qdq-per-axis-binary-concat
Oct 15, 2026Action required
Validate fp16 MatMulNBits prepack inputs (#33375)
currently running:
ONNX Runtime DirectML Builds#24874:Commit122c193pushedbyapsonawane
main
Oct 15, 2026In progress
Fix CPU Shrink precision loss and integer boundary handling
requires action with the application:
ONNX Runtime DirectML Builds#24873:Pull request#33353synchronizebyMohit-25-tech
Mohit-25-tech:fix/cpu-shrink-precision
Oct 15, 2026Action required
[MLAS] Add AArch64 SVE FP16 GEMM kernel and driver
requires action with the application:
ONNX Runtime DirectML Builds#24872:Pull request#32756synchronizebyabhijain1204fujitsu
MonakaResearch:nikhil/hgemm_sve_support
Oct 15, 2026Action required
[WebGPU]: Add Intel subgroup prefill flash-attention kernel
requires action with the application:
ONNX Runtime DirectML Builds#24871:Pull request#32702synchronizebydaijh
daijh:pr/flash-attn-prefill
Oct 15, 2026Action required
Update Eigen to 5.0.1 + SSE2 fix (broken Clip & Sign)
requires action with the application:
ONNX Runtime DirectML Builds#24870:Pull request#32475synchronizebycbourjau
cbourjau:update-eigen
Oct 15, 2026Action required
Update Eigen to 5.0.1 + SSE2 fix (broken Clip & Sign)
requires action with the application:
ONNX Runtime DirectML Builds#24869:Pull request#32475synchronizebycbourjau
cbourjau:update-eigen
Oct 15, 2026Action required
webgpu: using subgroup matrix to optimize Conv operator
requires action with the application:
ONNX Runtime DirectML Builds#24868:Pull request#32252synchronizebyxhcao
xhcao:subgroup_matrix_conv_2
Oct 15, 2026Action required
[WebGPU] Automate Dawn upgrades and component inventories
requires action with the application:
ONNX Runtime DirectML Builds#24867:Pull request#33396openedbygyagp
gyagp:upgrade-dawn
Oct 15, 2026Action required
[WebGPU] Fuse rank-32 LoRA and QLoRA projections with MatMulNBits
completed successfully:
ONNX Runtime DirectML Builds#24866:Pull request#33370synchronizebyqjia7
qjia7:feat/webgpu-matmul-nbits-qlora
Oct 15, 20261h 49m 7s
Add CUDA MoE CPU offload and adaptive expert swaps with parameter benchmarks
completed successfully:
ONNX Runtime DirectML Builds#24865:Pull request#33347synchronizebyxadupre
moe-fp16-cpu-offload-and-adaptive-swap
Oct 15, 20261h 47m 56s
Add CUDA MoE CPU offload and adaptive expert swaps with parameter benchmarks
cancelled:
ONNX Runtime DirectML Builds#24864:Pull request#33347synchronizebyxadupre
moe-fp16-cpu-offload-and-adaptive-swap
Oct 15, 202611m 22s
Add CUDA MoE CPU offload and adaptive expert swaps with parameter benchmarks
cancelled:
ONNX Runtime DirectML Builds#24863:Pull request#33347synchronizebyxadupre
moe-fp16-cpu-offload-and-adaptive-swap
Oct 15, 20267m 7s
Fix safety guards in training loss fusions
requires action with the application:
ONNX Runtime DirectML Builds#24862:Pull request#32987synchronizebyYuhx141
Yuhx141:fix/training-loss-fusion-guards
Oct 15, 2026-1s
Add CUDA MoE CPU offload and adaptive expert swaps with parameter benchmarks
cancelled:
ONNX Runtime DirectML Builds#24861:Pull request#33347synchronizebyxadupre
moe-fp16-cpu-offload-and-adaptive-swap
Oct 15, 20264m 26s
Add CUDA MoE CPU offload and adaptive expert swaps with parameter benchmarks
cancelled:
ONNX Runtime DirectML Builds#24860:Pull request#33347synchronizebyxadupre
moe-fp16-cpu-offload-and-adaptive-swap
Oct 15, 202611m 1s
Add CUDA MoE CPU offload and adaptive expert swaps with parameter benchmarks
cancelled:
ONNX Runtime DirectML Builds#24859:Pull request#33347synchronizebyxadupre
moe-fp16-cpu-offload-and-adaptive-swap
Oct 15, 20266m 46s
Add CUDA MoE CPU offload and adaptive expert swaps with parameter benchmarks
cancelled:
ONNX Runtime DirectML Builds#24858:Pull request#33347synchronizebyxadupre
moe-fp16-cpu-offload-and-adaptive-swap
Oct 15, 20266m 20s
[WebGPU] Fuse rank-32 LoRA and QLoRA projections with MatMulNBits
cancelled:
ONNX Runtime DirectML Builds#24857:Pull request#33370synchronizebyqjia7
qjia7:feat/webgpu-matmul-nbits-qlora
Oct 15, 20261h 5m 54s
Fix GatherToSliceFusion with negative Range limit
requires action with the application:
ONNX Runtime DirectML Builds#24856:Pull request#33377synchronizebyChoudhary-Sahil099
Choudhary-Sahil099:fix-33358-gather-to-slice
Oct 15, 2026Action required
[CUDA] Load fpA_intB tactic cache on session create and save on close
completed successfully:
ONNX Runtime DirectML Builds#24855:Pull request#33078synchronizebytianleiwu
tlwu/20261003/gemm_tactic_cache_load_save
Oct 15, 20261h 50m 43s
[WebGPU] Unroll bounded NVIDIA FlashAttention reductions in WGSL
requires action with the application:
ONNX Runtime DirectML Builds#24854:Pull request#33395openedbygyagp
gyagp:codex/webgpu-nvidia-flash-unroll-20261015
Oct 15, 2026Action required
[WebGPU] Fuse rank-32 LoRA and QLoRA projections with MatMulNBits
cancelled:
ONNX Runtime DirectML Builds#24853:Pull request#33370synchronizebyqjia7
qjia7:feat/webgpu-matmul-nbits-qlora
Oct 15, 202646m 36s
[WebGPU] Separate data transfers from buffer management
completed successfully:
ONNX Runtime DirectML Builds#24852:Pull request#33286synchronizebyxiaofeihan1
xiaofeihan1:codex/webgpu-data-transfer
Oct 15, 20261h 47m 37s

Workflow runs for ONNX Runtime DirectML Builds

2500+ workflow runs