microsoft/onnxruntime

Pull Requests Commits

apsonawane committed 77 days ago

ec19b3fa

PagedAttention: add CPU/WebGPU interfaces and tune CUDA validation/launch

apsonawane committed 77 days ago

135c2913

Add decode benchmark comparing ONNX Attention CUDA with contrib GQA (#29684)

namgyu-youn committed 77 days ago

Verified 95236f40

[CUDA] Add NVFP4 native FP4xFP4 prefill for QMoE on SM120 (#29824)

tianleiwu committed 78 days ago

Verified ed989163

[MLAS] Add ARM64 NEON fp32 RoPE kernel (#29836)

kjg0724 committed 78 days ago

Verified 60b4dd6a

[WebGPU EP] Support int64 Clip via a dedicated ClipInt64 kernel (#29834)

mingmingtasd committed 78 days ago

Verified 529f3baa

[CPU] Support FP16 GQA with quantized KV cache (#29825)

Copilot committed 78 days ago

Verified 72686faa

[Build] Speed up fpA_intB CUDA build (#29699)

tianleiwu committed 78 days ago

Verified c9879660

Skip DXGI device discovery if Win32k system calls are disabled (#29755)

shiyi9801 committed 78 days ago

Verified 1cee619a

Add Intel WebGPU subgroup-matrix Gemm f16 kernel (#29813)

jchen10 committed 78 days ago

Verified 56598d6f

[CUDA] Fix MaxPool wrong values/indices for dilation>1 with non-zero begin padding (#29638)

Copilot committed 78 days ago

Verified 809135c0

WebGPU: Support int64 for Reshape (#29830)

huningxin committed 78 days ago

Verified ad3a7450

Validate tensor element enum conversion at compile time (#29646)

GopalakrishnanN committed 78 days ago

Verified 3e8923ef

Make TensorProto element sizes compile-time (#29644)

GopalakrishnanN committed 78 days ago

Verified 7caa6448

Allow zero input for EpContext node (#29799)

shiyi9801 committed 79 days ago

Verified 1b1e1db7

Update npm lockfiles to latest versions (#29827)

mustjab committed 79 days ago

Verified cff03355

Update location of llvm-tblgen for new dawn builds (#29829)

eserscor committed 79 days ago

Verified 858442cf

WebGPU: Add HardSwish support (#29828)

huningxin committed 79 days ago

Verified cfaf942c

[MLAS] Run fp16 MatMulNBits through the int8 path without fp32 temporaries (#29791)

blazingphoenix7 committed 79 days ago

Verified 0a855f8e

[MLAS] Drop the ignored columnwise flag from the 2-bit fp-zero-point dequant (#29798)

blazingphoenix7 committed 79 days ago

Verified 54d6d894

Fix initializer load for in-memory external data (#29349)

tairenpiao committed 79 days ago

Verified 1a9fff93

Fix non-deterministic constant folding output (#29789)

ducviet00 committed 79 days ago

Verified cd2d4a3f

Fix Incorrect PIP_INDEX_URL piping (#29823)

eserscor committed 79 days ago

Verified 9f63d4dc

Revert "[Build] Use GPU pool to unblock CI temporarily (#29731)" (#29795)

tianleiwu committed 79 days ago

Verified ead383e1

Add int32 support for CPU Trilu (#29476)

the0cp committed 79 days ago

Verified c18a976f

Enable cuDNN SDPA for contrib Attention (#29717)

mastryukov1990 committed 79 days ago

Verified 32e94491

Add rotary embedding onnx domain support (#29261)

xiaoyu-work committed 79 days ago

Verified 72082126

[CUDA] Add NVFP4 support in QMoE (#29697)

tianleiwu committed 79 days ago

Verified 438c5d58

[Build] Fix windows build for sm90 or later (#29776)

tianleiwu committed 80 days ago

Verified b4c3a302

[CUDA] Make cuFFT optional at runtime (#29808)

tianleiwu committed 80 days ago

Verified bf3e9ab8

Older