15 Pull requests merged by 14 people

• Bump setuptools from 69.0.3 to 78.1.1 in /tools/ci_build/github/linux/docker/scripts

  #24810 merged May 27, 2025

• [Mac] Fix --use_xcode build with Nodejs binding

  #24868 merged May 27, 2025

• [WebNN] Refactor op mappings and add input name mapping between ONNX and WebNN

  #24830 merged May 27, 2025

• Update inferencing.md with correct minimum macOS version.

  #24863 merged May 27, 2025

• [QNN EP] Add ScatterND reduction attribute

  #24844 merged May 27, 2025

• [NvTensorRT RTX] Add Bfloat16

  #24743 merged May 23, 2025

• [TRT EP] Update build and API usage for TensorRT 10.11

  #24832 merged May 23, 2025

• Update MatMulBNits spec and Add Input Checks

  #24828 merged May 23, 2025

• Switch the TRT optimization profile if multi-profile is enable

  #24805 merged May 23, 2025

• [build] disable vcpkg for Dawn temporarily

  #24838 merged May 22, 2025

• Update Qnn default version to 2.34.0.250424

  #24750 merged May 22, 2025

• [QNN EP] MaxPool input rank-3 auto pad bug fix

  #24827 merged May 21, 2025

• [NV TensorRt RTX EP] : Fix Domain check.

  #24816 merged May 21, 2025

• [QNN EP] Fix inconsistent inputs for graph

  #24751 merged May 21, 2025

• Remove unused tensor dumper functions

  #24821 merged May 21, 2025

24 Pull requests opened by 19 people

• Avoid traversing entire arrays when extracting shape from objects in java

  #24833 opened May 21, 2025

• Fix inference unable to run due to JS WASM runtime not being bundled into `onnxruntime-web/wasm` build

  #24836 opened May 22, 2025

• WIP

  #24837 opened May 22, 2025

• Cast Nodes Fusion

  #24842 opened May 22, 2025

• Weaken dxcore dependency

  #24845 opened May 23, 2025

• [QNN EP] Fix 16x16 MatMul translation

  #24846 opened May 23, 2025

• Download protobuf dependency on ARM64 build host

  #24847 opened May 23, 2025

• [QNN-EP] Define SpaceToDepth fusion for YOLOv2.

  #24848 opened May 23, 2025

• [QNN EP] Add 16x16 Gemm translation

  #24849 opened May 23, 2025

• [CUDA] fp16 intB gemm

  #24854 opened May 23, 2025

• [WebGPU EP] Fix NaN bug in softmax operator

  #24855 opened May 24, 2025

• [WIP] use WebGPU EP instead of JSEP in WebAssembly

  #24856 opened May 25, 2025

• Update Whisper attention fusions

  #24857 opened May 25, 2025

• Bump clang-format from 19.1.7 to 20.1.5

  #24858 opened May 26, 2025

• Bump ruff from 0.11.10 to 0.11.11

  #24859 opened May 26, 2025

• Update xnnpack.cmake for WASM build

  #24860 opened May 26, 2025

• [NV TensorRT RTX EP] enable weight stripped engines with EP Context

  #24869 opened May 27, 2025

• Amd/dev/klagos customop

  #24874 opened May 27, 2025

• Add ONNX RMSNormalization(23)

  #24875 opened May 27, 2025

• Improve Windows ETW callback registration and fix issues

  #24877 opened May 27, 2025

• workaround for a VC++ bug in VS 17.14

  #24878 opened May 27, 2025

• Fix symbol publishing

  #24879 opened May 27, 2025

• [QNN-EP] Support non-last axis TopK.

  #24881 opened May 28, 2025

• Adding support for Turing Arch

  #24882 opened May 28, 2025

3 Issues closed by 3 people

• CUDA Gather is taking a node with fp16 input data on CUDA architecture < 5.3

  #24834 closed May 28, 2025

• Squeezenet1.0 models give wrong prediction results

  #20332 closed May 27, 2025

• [Build] DLL load failed while importing onnxruntime_pybind11_state

  #24843 closed May 24, 2025

17 Issues opened by 16 people

• ORT raises node "does not have type information set by parent node" for initializers declared in outer graph

  #24880 opened May 28, 2025

• Error messages from QNN are turned into verbose level messages

  #24876 opened May 27, 2025

• How to use kv_cache more reasonably in the exported onnx model?

  #24873 opened May 27, 2025

• Consider making sympy optional

  #24872 opened May 27, 2025

• [Build] onnxruntime 1.22.0 - gcc 13.3.0 - inference_session.cc:398

  #24871 opened May 27, 2025

• [Build] cmake cannot find KLEIDIAI - Windows 11 ARM

  #24865 opened May 26, 2025

• [Build] cmake "target_link_options" INTERFACE error on Windows 11 ARM VS2022

  #24864 opened May 26, 2025

• TreeEnsemble `post_transform` appears buggy.

  #24862 opened May 26, 2025

• [Build] Fail at configure due to issue checking out Eigen dependency

  #24861 opened May 26, 2025

• [Documentation] Is there existing documentation for running specific tests somewhere?

  #24853 opened May 23, 2025

• [Training] ImportError: ../python3.10/site-packages/onnxruntime/training/ortmodule/torch_cpp_extensions/torch_interop_utils.cpython-310-x86_64-linux-gnu.so: undefined symbol: _ZTIN5torch8autograd6PyNodeE

  #24852 opened May 23, 2025

• [Feature Request] s390x builds

  #24851 opened May 23, 2025

• [Documentation] OperatorKernels.md incomplete — missing supported operators (e.g. CastLike on CUDA)

  #24850 opened May 23, 2025

• [Feature Request] determine if particular execution provider is available for given platform ahead of time

  #24841 opened May 22, 2025

• [Feature Request]

  #24840 opened May 22, 2025

• [Build] Can't build 1.22 in debug mode on VS2022

  #24839 opened May 22, 2025

• [Performance] TensorRT Execution Provider in ONNX Runtime >3x slower than Triton-Inference-Server's TensorRT Backend for Same Resnet-101 Model

  #24831 opened May 21, 2025

48 Unresolved conversations

Sometimes conversations happen on old items that aren’t yet closed. Here is a list of all the Issues and Pull Requests with unresolved conversations.

• [CoreML] Update Reshape op to support more nodes

  #24594 commented on May 28, 2025 • 13 new comments

• Add WAITPKG checks, add support for TPAUSE within SpinPause

  #24524 commented on May 27, 2025 • 8 new comments

• Fusing Initializers with Graph Transforms

  #24726 commented on May 27, 2025 • 7 new comments

• Extend OrtAllocator API to get Allocator statistics

  #24785 commented on May 27, 2025 • 5 new comments

• [WebGPU] Unify core implementations of GEMM and MatMul

  #24586 commented on May 28, 2025 • 5 new comments

• Add python bindings to the global thread pool functionality

  #24238 commented on May 27, 2025 • 1 new comment

• Intermittent crash in ETW logging

  #24773 commented on May 27, 2025 • 0 new comments

• [Mobile] MatMulNbits Q8 Errors out on Android

  #24769 commented on May 27, 2025 • 0 new comments

• [Feature Request] Add Fusion Transformer for WebNN EP Decomposed GQA Node

  #24454 commented on May 27, 2025 • 0 new comments

• [Feature Request] Implement RMSNormalization-23

  #24555 commented on May 27, 2025 • 0 new comments

• [Feature Request] [web/webgpu] Support non-symmetrical padding in Conv

  #24800 commented on May 27, 2025 • 0 new comments

• [Feature Request] Restore XNNPACK Execution Provider for ONNX Runtime Web Backend

  #24766 commented on May 27, 2025 • 0 new comments

• TensorRTExecutionProvider error during session initialization

  #22199 commented on May 28, 2025 • 0 new comments

• [VitisAI] refactor VitisAI EP for open source

  #24426 commented on May 28, 2025 • 0 new comments

• Fix initialization of same_node_ in TreeEnsemble

  #24654 commented on May 22, 2025 • 0 new comments

• [onnxruntimeperftest] Add option to enable IO bindings on CUDA before session run

  #24672 commented on May 21, 2025 • 0 new comments

• [webgpu] Add zero points support for dp4 path

  #24675 commented on May 28, 2025 • 0 new comments

• [NvTensorRTRTX EP] Enable automatic selection of NvTensorRTRTX EP for PREFER_GPU policy

  #24689 commented on May 21, 2025 • 0 new comments

• Update deprecated CUDA api

  #24733 commented on May 23, 2025 • 0 new comments

• Fix AutoEpSelection and OrtEpLibrary tests when using AuthenticAMD

  #24754 commented on May 26, 2025 • 0 new comments

• [QNN EP] Fuse scale into softmax

  #24809 commented on May 23, 2025 • 0 new comments

• [MLAS] DequantizeLinear int8/uint8

  #24818 commented on May 27, 2025 • 0 new comments

• [QNN-EP] Add Support for CumSum in QNN EP

  #24820 commented on May 28, 2025 • 0 new comments

• [ci] revise wasm CI

  #24825 commented on May 27, 2025 • 0 new comments

• onnxruntime crashes: Segmentation fault (core dumped) on valid model

  #24806 commented on May 21, 2025 • 0 new comments

• KeyError when calling onnxruntime.tools.symbolic_shape_infer.SymbolicShapeInference on model containing Loop

  #24495 commented on May 22, 2025 • 0 new comments

• [Build] Python build fails because onnxruntime/capi/build_and_package_info.py is missing

  #24570 commented on May 22, 2025 • 0 new comments

• [Build] How to build CoreML for running C++ code on MacOS

  #23556 commented on May 23, 2025 • 0 new comments

• Python wheel for x64 onnxruntime-qnn package incorrect binaries

  #24508 commented on May 23, 2025 • 0 new comments

• [Performance] GPU op placement control when some ops must be on the CPU

  #23154 commented on May 23, 2025 • 0 new comments

• Broken multithreading inference session Onnxruntime-directml >= 1.18

  #20713 commented on May 23, 2025 • 0 new comments

• Is DML being deprecated?

  #23783 commented on May 23, 2025 • 0 new comments

• com.microsoft.Attention do_rotary flag doesn't work on apple silicon

  #24528 commented on May 24, 2025 • 0 new comments

• Need help - C++ ONNXRuntime Failing

  #24476 commented on May 24, 2025 • 0 new comments

• [Performance] LearningModelSession::Evaluate ToggleProfile() call breaks profiling

  #24507 commented on May 24, 2025 • 0 new comments

• Improve DFT implementation

  #24522 commented on May 24, 2025 • 0 new comments

• [Web] `onnxruntime-node` post-install script errors with: "Failed to find runtimes/win-x64/native/libonnxruntime_providers_cuda.so in NuGet package"

  #24770 commented on May 24, 2025 • 0 new comments

• [Performance] Onnx session utilizes more GPU and CPU ram on Nvidia H100 than on Nvidia A100

  #24543 commented on May 25, 2025 • 0 new comments

• std::bad_alloc when loading a model with sparse tesnsor constant node.

  #24530 commented on May 25, 2025 • 0 new comments

• Feature request: Implement GroupNormalization-21

  #24538 commented on May 25, 2025 • 0 new comments

• Update ORT to handle explicit OpSchemaRegisterOnce API in ONNX >= 1.18.0 for fluent chaining

  #24561 commented on May 26, 2025 • 0 new comments

• GroupNormalization-18 is deprecated since ONNX==1.18.0

  #24560 commented on May 26, 2025 • 0 new comments

• ORT uses static shape inference functions in ONNX==1.18

  #24558 commented on May 26, 2025 • 0 new comments

• Support FLOAT4E2M1

  #24553 commented on May 26, 2025 • 0 new comments

• [Performance]

  #13500 commented on May 26, 2025 • 0 new comments

• how to release gpu memory when keep onnxruntime session around.

  #9509 commented on May 26, 2025 • 0 new comments

• [Feature Request] Consider support int4/uint4 for reshape op of default CPU EP

  #24285 commented on May 26, 2025 • 0 new comments

• NOT_IMPLEMENTED : Could not find an implementation for ConvInteger(10) node with name 'Conv_0_quant'

  #15888 commented on May 27, 2025 • 0 new comments