microsoft/onnxruntime

Pull Requests Commits

Update optimizer opset version checks for latest ONNX opset 26 (#28966)

yuslepukhin committed 24 days ago

Verified 0b278bbf

Fill CUDA opset gap for Softplus and Softsign to opset 22 (#28982)

elwhyjay committed 25 days ago

Verified 7db78930

Bump shell-quote from 1.8.3 to 1.8.4 in /js/react_native/e2e (#29022)

dependabot[bot] committed 25 days ago

Verified 0172bcd0

[CUDA] Optimize QMoE SoftmaxTopK router for small-batch decode (#28980)

tianleiwu committed 25 days ago

Verified dbf95cfb

Fix oob dereference (#29012)

apsonawane committed 25 days ago

Verified b823aecc

QMoE: fail loudly when weights_prepacked=0 but PrePack did not run (#28965)

Copilot committed 25 days ago

Verified cf509d84

Fix STFT complex input frame offsets (#28961)

GopalakrishnanN committed 25 days ago

Verified 4b5abcf5

Log EP version on inference failure and in EpDeviceUsage event, Log ORT version (#28794)

dabhattimsft committed 25 days ago

Verified 6584194f

fix: Add Linux NPU discovery through sysfs accel devices (#28703)

the0cp committed 26 days ago

Verified 5c7ba131

Fix out-of-bounds error (#28991)

apsonawane committed 26 days ago

Verified 0931c1a0

webgpu: Fuse FlashAttention decode kernels and extend to any sequence length (#28389)

qjia7 committed 26 days ago

Verified f2c8a096

[CUDA] Fix QMoE int4/int8 weight prepack to always use SM80 layout (#28978)

tianleiwu committed 26 days ago

Verified 2cf6c6c6

Move OrtModelPackageApi to the experimental C API (#28990)

jambayk committed 26 days ago

Verified 2c4be913

Experimental C API support (#28746)

edgchen1 committed 27 days ago

Verified 0f300140

webgpu: Add session-level buffer pool for graph capture reuse (#28761)

qjia7 committed 27 days ago

Verified 50c95100

Address Lora shortcomings (#28801)

yuslepukhin committed 27 days ago

Verified a4051793

QMoE: prepack int4/int8 expert weights in PrePack hook (symmetric with MatMulNBits) (#28749)

justinchuby committed 27 days ago

Verified 1e1f574c

fix oob read access in GatherBlockQuantized (#28718)

Guenther Schmuelling committed 27 days ago

Verified 17724b35

WebGPU Pad kernel int64 / int32 truncation can read to oob read (#28721)

Guenther Schmuelling committed 27 days ago

Verified 2c8fe727

[MLAS/RISC-V] Add RVV INT8 GEMM/GEMV, M=1 routing, and activation kernels (follow-up #28261) (#28308)

qiurui144 committed 27 days ago

Verified 9e815996

Upgrade cudnn_frontend to 1.24 and enable cuDNN SDPA for MHA/GQA (#28849)

tianleiwu committed 27 days ago

Verified fc7a9f0e

[WebGPU] Fuse Q/K RMSNorm into GroupQueryAttention for Qwen3-style models (#28484)

hariharans29 committed 28 days ago

Verified e8b9b24a

Validate DQ scale/zero_point shapes before QGemm fusion (#28714)

wangw-1991 committed 28 days ago

Verified 238905a4

Clarify and improve ownership semantics Model Editor C/C++ API (#28800)

yuslepukhin committed 28 days ago

Verified dadd2d03

Fix Type-confusion: raw-pointer bind_input causing OOB write (#28839)

selenayang888 committed 28 days ago

Verified 7bf3032b

[CoreML EP] Add Where and And builders (#28597)

maxwbuckley committed 28 days ago

Verified 073e49ff

fix(ci): nodejs on nuget cuda 13 hangs (#28816)

sanaa-hamel-microsoft committed 29 days ago

Verified 8f6e944b

Fix issue w/ DML batched readback size (#28680)

adrastogi committed 31 days ago

Verified 90c095d1

fix: require GCC >= 12 on aarch64 for bf16/fp16 capability probes (#28785)

Rishi-Dave committed 31 days ago

Verified d9ecd81d

Fix INT64 initializer data propagation (#28778)

GopalakrishnanN committed 31 days ago

Verified 0b7823e0

Older