34 Pull requests merged by 23 people

• replace usage of gsl::narrow and gsl::narrow_cast in WebGPU EP

  #23926 merged Mar 7, 2025

• Fix license in example test code.

  #23936 merged Mar 7, 2025

• Create a packaging pipeline for a custom nuget package

  #23918 merged Mar 7, 2025

• [AIX] External data handling

  #23859 merged Mar 7, 2025

• Updated ov version in pipeline (#595)

  #23882 merged Mar 7, 2025

• Fix ConvInteger handling of optional inputs.

  #23935 merged Mar 7, 2025

• Updated run_CIs_for_external_pr.py to support the Windows OpenVINO CI pipeline

  #23931 merged Mar 7, 2025

• fix binplace file in web pipeline

  #23930 merged Mar 7, 2025

• Enabling L2+ Optimizations for EPs

  #23517 merged Mar 7, 2025

• Example custom op with output type inferencing

  #23916 merged Mar 7, 2025

• Support all block sizes that are multiples of 32 for DP4A

  #23907 merged Mar 7, 2025

• Exclude MAUI projects from GPU C# packaging builds

  #23923 merged Mar 7, 2025

• [WebGPU EP] SoftMax Implementation

  #23538 merged Mar 7, 2025

• Adding OpenVINO Windows CI Pipeline

  #23919 merged Mar 7, 2025

• enable WebGPU EP in WebAssembly build

  #23913 merged Mar 6, 2025

• [JSEP/WebGPU] Fixed error in softmax dispatch.

  #23906 merged Mar 6, 2025

• WebGPU: Remove deprecated subgroups-f16 from WebGPU native and JS EP

  #23898 merged Mar 6, 2025

• Ensure that the 'cmake_minimum_required' is version 3.5 or greater

  #23888 merged Mar 6, 2025

• [WebNN] Accept Float16Array for float16 data type if it is available

  #23894 merged Mar 6, 2025

• [webgpu] support Pad operator

  #23141 merged Mar 6, 2025

• [webgpu] Restore MatMulNBits workgroup size for Phi-3.5

  #23349 merged Mar 6, 2025

• Round 2 of cherry-picks into rel-1.21.0

  #23899 merged Mar 6, 2025

• [js/web] improve workaround for bundlers

  #23902 merged Mar 6, 2025

• Dynamo export and improve benchmark script for SAM2 encoder

  #23887 merged Mar 5, 2025

• [WebGPU EP] introduce BiasAdd contrib op

  #23861 merged Mar 5, 2025

• [WebGPU-EP Native] Add ReduceMean

  #23860 merged Mar 5, 2025

• Fix enable_pix_capture build for WebGPU

  #23857 merged Mar 5, 2025

• Fix formatting in snapdragon.md

  #23900 merged Mar 5, 2025

• Add snapdragon tutorial

  #23890 merged Mar 5, 2025

• [QNN-EP]: Fix inference failures while running with htp_shared_memory

  #23892 merged Mar 5, 2025

• [TensorRT EP] Add doc for trt_op_types_to_exclude

  #23893 merged Mar 5, 2025

• [QNN EP Docs] Update docs for building QNN EP as shared or static library

  #23873 merged Mar 5, 2025

• Enable QNN EP weight sharing generation using public API

  #23702 merged Mar 5, 2025

• Doc update relate to EPContext model default name

  #23865 merged Mar 4, 2025

18 Pull requests opened by 13 people

• Pick Jian's pipeline changes to the 1.21 release branch

  #23903 opened Mar 5, 2025

• [TensorRT EP] support TensorRT 10.9-GA

  #23905 opened Mar 5, 2025

• [webgpu] Optimize MatMulNBits for f16 Block32 prefill performance

  #23908 opened Mar 6, 2025

• [WIP][Native WebGPU] Remove explicit split operator in GQA

  #23909 opened Mar 6, 2025

• [WebGPU] Direct CPU->GPU buffer upload for UMA

  #23910 opened Mar 6, 2025

• Fix CUDA EP Abs and Sign bfloat16 support

  #23914 opened Mar 6, 2025

• [WebGPU EP] Implements Gelu, BiasSplitGelu, and QuickGelu

  #23920 opened Mar 6, 2025

• Bump SixLabors.ImageSharp from 2.1.9 to 2.1.10 in /csharp/sample/Microsoft.ML.OnnxRuntime.FasterRcnnSample

  #23924 opened Mar 6, 2025

• Extend CMAKE_CUDA_FLAGS with all Blackwell compute capacity

  #23928 opened Mar 7, 2025

• [WIP] DepthToSpace for WebGPU EP

  #23929 opened Mar 7, 2025

• update transformer version to 4.48.0

  #23932 opened Mar 7, 2025

• VCPKG improvement: set VCPKG_OSX_DEPLOYMENT_TARGET

  #23933 opened Mar 7, 2025

• [Native WebGPU] Added ReduceMax and ReduceSum

  #23934 opened Mar 7, 2025

• [js] Add API for accessing metadata of a model's input/output

  #23937 opened Mar 7, 2025

• [Fix] Dependencies find_package Eigen error

  #23939 opened Mar 7, 2025

• Add support for custom position ids and attention mask to GQA CPU operator

  #23944 opened Mar 7, 2025

• Qnn weight sharing improvement

  #23945 opened Mar 7, 2025

• Allow using a different version of flatbuffers when building with vcpkg

  #23946 opened Mar 7, 2025

8 Issues closed by 6 people

• [Build] Released asset for v1.20.1 doesn't work on macOS Sequoia

  #23922 closed Mar 7, 2025

• What's the right way to construct custom ops with the same name but different output types?

  #23891 closed Mar 7, 2025

• [Performance] Keep Onnx awake while in idle mode

  #23461 closed Mar 6, 2025

• [Documentation] Unclear how to run `run_benchmark.py`

  #23889 closed Mar 5, 2025

• [Build] ONNX Run Time on Conda Forge - Add CUDA Support

  #23904 closed Mar 5, 2025

• [Build] NuGet Package missing header files

  #23884 closed Mar 5, 2025

• [Build] how to compile ios static library

  #23835 closed Mar 4, 2025

• ort.InferenceSession fails silently

  #23869 closed Mar 4, 2025

15 Issues opened by 15 people

• [Web] WASM sigmoid producing numbers below 0 or above 1

  #23943 opened Mar 7, 2025

• Error when I use cuda_runtime.h and OpenVINO EP at the same time

  #23941 opened Mar 7, 2025

• [Feature Request] Add more options to load models at InferenceSession constructor

  #23940 opened Mar 7, 2025

• Bad Allocation Error in ONNX Runtime on Windows x86 CPU When Processing Multiple Images Sequentially

  #23938 opened Mar 7, 2025

• ConvInteger segfaults when x_zero_point is the empty string

  #23927 opened Mar 7, 2025

• [Feature Request] Multi-Head Latent Attention(DeepSeek) support on CPU/NPU

  #23925 opened Mar 6, 2025

• [Web] Facing this error in WebGPU: Model warmup failed: Error: input 'detection' is missing in 'feeds'.

  #23921 opened Mar 6, 2025

• Public and open source contains header references to "confidential and proprietary" Microsoft code.

  #23917 opened Mar 6, 2025

• [Build] memory leaked

  #23915 opened Mar 6, 2025

• [Build] onnxruntime with tag 1.20.* build failed on Windows after VS upgrade to 17.13.*

  #23911 opened Mar 6, 2025

• [Documentation] Memory Leak in TensorRTProvider example

  #23901 opened Mar 5, 2025

• [C++, Linux] Segmentation fault when run OrtApi::Run

  #23897 opened Mar 5, 2025

• [OpenVINO GPU] OpenVINO EP shouldn't override the "ACCURACY" precision to "FP32"

  #23895 opened Mar 5, 2025

• Xnnpack execution provider Resize::IsOnnxNodeSupported causes crash for models where Resize layer scales tensor is an empty tensor

  #23886 opened Mar 4, 2025

• [DO NOT UNPIN] ORT 1.21.0 Release Candidates available for testing

  #23885 opened Mar 4, 2025

37 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.

• Integrate KleidiAI for MatMulNBits via MlasQNBitGemm

  #23627 commented on Mar 7, 2025 • 14 new comments

• Whisper Redesigned Solution

  #23549 commented on Mar 6, 2025 • 8 new comments

• [WebNN] Better int64 integration

  #23831 commented on Mar 7, 2025 • 7 new comments

• [mobile] Add Android NuGet BrowserStack test to NuGet packaging pipeline

  #23580 commented on Mar 6, 2025 • 4 new comments

• [OpenVINO]Session Options Appended After AppendExecutionProvider

  #23852 commented on Mar 7, 2025 • 3 new comments

• [Build] CUDA version linkage

  #23841 commented on Mar 4, 2025 • 0 new comments

• Migrate yarn to npm

  #22116 commented on Mar 6, 2025 • 0 new comments

• [js/web] Add Wasm Relaxed SIMD support to wasm backend

  #22794 commented on Mar 5, 2025 • 0 new comments

• [Native WebGPU EP] Add packedQKV and do_rotary attribute support to GroupQueryAttention operator

  #23386 commented on Mar 6, 2025 • 0 new comments

• [WebNN EP] Support GroupQueryAttention(GQA)

  #23416 commented on Mar 7, 2025 • 0 new comments

• [WebGPU/JSEP] Support group query attention do_rotary attribute

  #23524 commented on Mar 7, 2025 • 0 new comments

• Migrate Zip-Nuget Package Pipeline to 1ES

  #23609 commented on Mar 7, 2025 • 0 new comments

• [WIP] migrate WebGPU EP to WebAssembly to replace JSEP

  #23697 commented on Mar 6, 2025 • 0 new comments

• Make python package pipeline 1ES compliant

  #23800 commented on Mar 6, 2025 • 0 new comments

• Make python CUDA package pipeline 1ES compliant

  #23802 commented on Mar 6, 2025 • 0 new comments

• Make Cuda packaging pipeline 1ES compliant

  #23806 commented on Mar 7, 2025 • 0 new comments

• [VitisAI] Just for internal test

  #23849 commented on Mar 5, 2025 • 0 new comments

• Synchronize patch files, fix resource compiler invocations in some situations

  #23855 commented on Mar 6, 2025 • 0 new comments

• [VitisAI EP] export InferShapes to VitisAIEP

  #23881 commented on Mar 7, 2025 • 0 new comments

• Attention fusion broken for BART 🤖

  #23864 commented on Mar 4, 2025 • 0 new comments

• Half of the length that correct output shape

  #23883 commented on Mar 5, 2025 • 0 new comments

• [Build] aarch64 ACL (20.02) build fails with onnxruntime `v1.13.1`, `1.14.1` and `1.15.0`

  #16176 commented on Mar 5, 2025 • 0 new comments

• [Build] Error building with ACL EP on aarch64 linux (Raspberry Pi 5)

  #23741 commented on Mar 5, 2025 • 0 new comments

• [Feature Request] Request grid_sample 5D support 🌟

  #21382 commented on Mar 5, 2025 • 0 new comments

• [Build] build error for windows

  #23166 commented on Mar 6, 2025 • 0 new comments

• [TensorRT EP] How can I disable generating cache when using trt execution provider

  #22822 commented on Mar 6, 2025 • 0 new comments

• [Build] What version of ArmNN does onnxruntime v1.15.1 work with?

  #17763 commented on Mar 6, 2025 • 0 new comments

• [Build] Build failure on Windows 11 with CUDA/cuDNN: nvcc subprocess error during CUDA compilation (v1.20.2)

  #23844 commented on Mar 6, 2025 • 0 new comments

• Failed to load library libonnxruntime_providers_cuda.so I am getting the following erro

  #19616 commented on Mar 6, 2025 • 0 new comments

• Abs node runs into error with bf16 tensor

  #23875 commented on Mar 6, 2025 • 0 new comments

• [Build] Openvino fails to build with AUTO:GPU,CPU

  #23866 commented on Mar 6, 2025 • 0 new comments

• When using the int8 quantization model to convert to onnx, an error occurs during runtime

  #23879 commented on Mar 7, 2025 • 0 new comments

• [Web] BiRefNet_T not working on webgpu

  #21968 commented on Mar 7, 2025 • 0 new comments

• Mixed Precision ValueError: validation failed for model with all nodes in node_block_list

  #14235 commented on Mar 7, 2025 • 0 new comments

• [Performance] using onnxruntime with ray and also fix for memory footprint too high

  #16793 commented on Mar 7, 2025 • 0 new comments

• Broken multithreading inference session Onnxruntime-directml >= 1.18

  #20713 commented on Mar 7, 2025 • 0 new comments

• [VitisAI] Add vaip Integration Using FetchContent

  #22038 commented on Mar 7, 2025 • 0 new comments