microsoft/onnxruntime

Pull Requests Commits

Default native telemetry to 1DS

bmehta001 committed 3 days ago

9486762c

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

tianleiwu committed 6 days ago

Verified 27e64f96

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

4n4ny4 committed 6 days ago

Verified 76b1d967

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

tianleiwu committed 7 days ago

Verified e79a40aa

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

tianleiwu committed 7 days ago

Verified cd13bdd8

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

arnej27959 committed 7 days ago

Verified 09f2e562

Add packed attention workspace recipes (#32283)

titaiwangms committed 7 days ago

Verified 2fee06a0

Add BFloat16 support for CUDA GatedDeltaNet (#32307)

Copilot committed 7 days ago

Verified c5f91da5

Prevent overlapping MatMulIntegerToFloat fusions (#32038)

danielsongmicrosoft committed 7 days ago

Verified 0e3af8f0

Remove internal URL from Guardian baseline (#32311)

hanbitmyths committed 7 days ago

Verified dba389de

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

mei1127 committed 7 days ago

Verified 24c836a1

Enable Linux AArch64 WebGPU plugin EP packages (#32287)

edgchen1 committed 8 days ago

Verified 95e71d26

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

eserscor committed 8 days ago

Verified 917a18e0

Fix Compile API duplicate graph serialization (#32303)

adrastogi committed 8 days ago

Verified cf573729

[CUDA] Add compact GatedDeltaNet operator (#32282)

tianleiwu committed 8 days ago

Verified 7fe62a35

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

xiaofeihan1 committed 8 days ago

Verified 2187135f

Preserve custom allocator during EP registration (#32272)

xiaofeihan1 committed 8 days ago

Verified 173b51bf

[CUDA] Add an is_causal attribute to PagedAttention (#32225)

tianleiwu committed 8 days ago

Verified a75c9b63

Enable WebGPU subgroup matrix path for WASM builds (#32269)

jchen10 committed 8 days ago

Verified 4f6af34e

[CUDA] Add compact VarlenCausalConv state updates (#32290)

tianleiwu committed 8 days ago

Verified 7e0f9b01

[CUDA] Add FP16-cache paged XQA decode for head size 256 (#32263)

baijumeswani committed 9 days ago

Verified b1f76d58

Fix ort-ci/SKILL.md yaml frontmatter syntax. (#32248)

edgchen1 committed 9 days ago

Verified 05491530

Harden head-size-256 paged XQA coverage (#32262)

baijumeswani committed 10 days ago

Verified d67cc269

Run wgsl_template Python tests in CI (#32214)

edgchen1 committed 11 days ago

Verified dc999734

[MLAS] Reject Arm® KleidiAI™ Q4 prepack with dynamic scales (#32068)

martin-klacer-arm committed 11 days ago

Verified 9cbf665b

[MLAS] Refactor QNBit KleidiAI integration into dedicated layer (#31720)

martin-klacer-arm committed 11 days ago

Verified 5cb241b5

Validate MatMulNBits 8-bit g_idx bounds on CUDA (#31643)

apsonawane committed 11 days ago

Verified 066b90ef

Fix Windows CUDA 12.9 SM120 compilation (#32114)

Copilot committed 11 days ago

Verified 939e180a

pin review dog (#32208)

eserscor committed 11 days ago

Verified ddc8f0c3

Upgrade Protobuf to v33.6 (#29906)

Copilot committed 11 days ago

Verified 993cd135

Older