Pull Requests microsoft/onnxruntime

webgpu: Extend FlashAttention decode path for any sequence length

#28389 opened 2026-05-07 02:21 by qjia7

[EP ABI] Add API to select the best compiled model compatibility info from candidate strings

#28387 opened 2026-05-06 21:32 by chilo-ms

Fix int32 overflow in CUDA Cast and UnaryElementWise kernels for tensors with >2^31 elements

#28386 opened 2026-05-06 20:29 by Copilot

Fix CUTLASS FMHA BiasLoader alignment for unaligned kernel path

#28369 opened 2026-05-05 18:48 by justinchuby

[js/rn] Fix iOS SIGSEGV on JS reload by releasing Env in -invalidate

#28367 opened 2026-05-05 18:38 by shlaikov

Fix DLA core selection for cached TensorRT engines

#28364 opened 2026-05-05 13:35 by jp-pino

Remove data hash from hashing code, to prevent unbounded growth issue

#28363 opened 2026-05-05 13:25 by JonathanC-ARM

fix: guard ORT_USE_CPUINFO on __linux__ to fix FreeBSD build

#28362 opened 2026-05-05 11:26 by Rishi-Dave

Adding Qnn Cpu to kDefaultBackends

#28360 opened 2026-05-05 05:14 by mahabayana

Add python implementation for wgsl-gen ep:WebGPU

#28355 opened 2026-05-04 23:22 by danielsongmicrosoft

Add float zero point support for 2-bit LUT GEMM in MatMulNBits

#28354 opened 2026-05-04 21:57 by vraspar

Fix CUDA ReduceSum erroring out on empty tensors with explicit axes

#28353 opened 2026-05-04 21:19 by justinchuby

Fix ARM CPUIDInfo bounds handling for unknown CPU vendors

#28344 opened 2026-05-04 15:10 by Copilot

Register Flatten as Direct8Bit op in Python QDQ quantizer

#28340 opened 2026-05-04 11:14 by Rishi-Dave

feat(qmoe): support 2-bit expert weights in CPU kernel

#28336 opened 2026-05-03 12:30 by Rishi-Dave

fix: support Float16/BFloat16/Float8 in TensorArray custom op

#28335 opened 2026-05-03 12:01 by Rishi-Dave

fix: skip DQ->MatMulNBits fusion when weight/scale initializer is shared

#28326 opened 2026-05-02 11:10 by Rishi-Dave

[WebGPU] Fix stale buffer bindings on first graph-capture replay ep:WebGPU

#28325 opened 2026-05-02 03:52 by hariharans29

Fix CUDA 13.2 (CUB 3.2.0) build failure: invalid C++ in device_transf…

#28317 opened 2026-05-01 21:53 by mc-nv

AGENTS.md: clarify Python venv activation guidance

#28315 opened 2026-05-01 18:02 by edgchen1

Fix session use-after-free when UserLoggingFunction is used

#28314 opened 2026-05-01 17:45 by tianleiwu

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

#28308 opened 2026-05-01 12:01 by qiurui144

Fix ModelEditorValueInfoToOnnx dropping symbolic dim names

#28307 opened 2026-05-01 08:21 by f-dy

[CoreML EP] Add Identity, Ceil, Tile builders + drop trivial-only partitions

#28293 opened 2026-04-30 11:45 by maxwbuckley

Don't pin SelectorActionTransformer replacement nodes to CPU

#28288 opened 2026-04-30 08:26 by maxwbuckley

webgpu: extend gemm-subgroup to Conv operator

#28283 opened 2026-04-30 03:50 by xhcao

Add example and documentation for kOrtEpDevice_EpMetadataKey_OSDriverVersion

#28282 opened 2026-04-30 03:29 by adrastogi

Refactor download functions to handle HTTP redirects

#28281 opened 2026-04-30 02:54 by lxp521125

[WebGPU] QKV and MLP fusions for Qwen3 ep:WebGPU

#28280 opened 2026-04-30 01:30 by hariharans29

Validate conv bias shape in WordConvEmbedding to prevent OOB read

#28279 opened 2026-04-29 22:32 by apsonawane