microsoft/onnxruntime

Pull Requests Commits

Fix cuDNN frontend build on Windows ARM64

Copilot committed 11 days ago

Verified 491cee4d

Fix int32 overflow in softmax_warp_forward offset arithmetic (#32330)

DKAIN-py committed 11 days ago

Verified 57affcea

Gate CPU FP16 Gemm and MatMul on hardware acceleration (#32301)

Copilot committed 11 days ago

Verified d97b5498

[WebGPU] Reject foreign GPU handles in the built-in data transfer (#32317)

4n4ny4 committed 11 days ago

Verified c1f82333

[WebGPU] Vectorize Split when every output segment is vec4-aligned (#32251)

4n4ny4 committed 11 days ago

Verified fe299c0f

Validate SpaceDepth shape arithmetic (#32039)

danielsongmicrosoft committed 11 days ago

Verified 9cca776e

[CUDA] Refine fpA-intB GEMV support checks (#32338)

tianleiwu committed 11 days ago

Verified 4422dd26

[WebGPU] Add transpose-optimizer handlers for Elu and the contrib GELU variants (#32118)

4n4ny4 committed 12 days ago

Verified 0f3cf6cc

test(tensorscatter): cover circular prefix preservation (#32264)

titaiwangms committed 12 days ago

Verified c029e7f5

[WebGPU] Pin subgroup size to 32 for subgroup-matrix MatMul/Gemm (#32306)

jchen10 committed 12 days ago

Verified 725aa0ad

Remove dead code in packaging pipelines (#32336)

eserscor committed 12 days ago

Verified 18241dc8

Support attention bias with windowed CPU GQA (#32302)

tianleiwu committed 12 days ago

Verified 540a28af

Enable l2normalization and dequantizelinear_blocked ops unit tests (#29838)

preetha-intel committed 12 days ago

Verified cc9d4b62

[CUDA] Build compact fpA intB kernels by default (#32324)

tianleiwu committed 12 days ago

Verified d59a8903

[CUDA] Fix malformed GatedDeltaNet test output verification (#32329)

tianleiwu committed 12 days ago

Verified f578ef8a

[CUDA] Extend speculative decode GEMVs to 64 rows (#32289)

tianleiwu committed 13 days ago

Verified 27e64f96

[WebGPU] Fuse eight more activations into Conv (#32117)

4n4ny4 committed 13 days ago

Verified 76b1d967

[CUDA] Tune NVFP4 GEMV tiling for Qwen MTP (#32140)

tianleiwu committed 14 days ago

Verified e79a40aa

[CUDA] Keep CudaAsyncBuffer's staging buffer alive across a graph replay (#32121)

tianleiwu committed 14 days ago

Verified cd13bdd8

Force fp32 fallback for CPU-assigned fp16 nodes with no matching kernel (#32197)

arnej27959 committed 14 days ago

Verified 09f2e562

Add packed attention workspace recipes (#32283)

titaiwangms committed 14 days ago

Verified 2fee06a0

Add BFloat16 support for CUDA GatedDeltaNet (#32307)

Copilot committed 15 days ago

Verified c5f91da5

Prevent overlapping MatMulIntegerToFloat fusions (#32038)

danielsongmicrosoft committed 15 days ago

Verified 0e3af8f0

Remove internal URL from Guardian baseline (#32311)

hanbitmyths committed 15 days ago

Verified dba389de

Enable D3D12 Agility SDK support for Dawn WebGPU builds (#32253)

mei1127 committed 15 days ago

Verified 24c836a1

Enable Linux AArch64 WebGPU plugin EP packages (#32287)

edgchen1 committed 15 days ago

Verified 95e71d26

Update ARM64 pool in an effort for infra unification. ARM64 now builds off Win2025. (#32281)

eserscor committed 15 days ago

Verified 917a18e0

Fix Compile API duplicate graph serialization (#32303)

adrastogi committed 15 days ago

Verified cf573729

[CUDA] Add compact GatedDeltaNet operator (#32282)

tianleiwu committed 15 days ago

Verified 7fe62a35

[WebGPU] Fix synchronization for user-provided GPUDevice (#32259)

xiaofeihan1 committed 15 days ago

Verified 2187135f

Older