31 Pull requests merged by 21 people

• Change gsl::byte to std::byte

  #23872 merged Mar 4, 2025

• [OpenVINO] Fix a build warning

  #23877 merged Mar 4, 2025

• [js/webgpu] Reland the optimization of ConvTranspose

  #23858 merged Mar 4, 2025

• [Doc] Update CUDA option prefer_nhwc

  #23812 merged Mar 4, 2025

• [js/common] allows using Uint16Array as data for float16 tensor

  #23827 merged Mar 3, 2025

• Make Nuget QNN package pipeline 1ES compliant

  #23805 merged Mar 3, 2025

• Change the logic to generate the default ep context file name

  #23788 merged Mar 3, 2025

• Quant tool: Consistent get_qdq_config and get_qnn_qdq_config behavior

  #23856 merged Mar 2, 2025

• Fix typos in csharp/src/Microsoft.ML.OnnxRuntime/

  #23848 merged Mar 1, 2025

• Fix typo: change Upample to Upsample.

  #23838 merged Mar 1, 2025

• Model Builder API

  #23223 merged Feb 28, 2025

• Cherry-picks into rel-1.21.0

  #23846 merged Feb 28, 2025

• Fix flash attention for GQA (Phi4)

  #23850 merged Feb 28, 2025

• Revert changes onn mac-react-native-ci-pipeline.yml

  #23845 merged Feb 27, 2025

• [Mlas] Unblock hardcoded matmul blocking size

  #23815 merged Feb 27, 2025

• Increase npm package pipeline ReactNative_CI_iOS timeout to 120 mins

  #23825 merged Feb 27, 2025

• [ORT/CI_Pipeline] Use --enable_generic_interface in ORT builds for EP testing

  #23801 merged Feb 27, 2025

• Quant tool: Add nodes_to_exclude in get_qnn_qdq_config

  #23779 merged Feb 27, 2025

• Update onnxruntime_external_deps.cmake: add missing EXCLUDE_FROM_ALL

  #23829 merged Feb 27, 2025

• [OVEP] Update support for Contrib Ops

  #23789 merged Feb 27, 2025

• upgrade emsdk to 4.0.4

  #23819 merged Feb 27, 2025

• [webgpu] Fix alignment issues in shader code

  #23776 merged Feb 27, 2025

• [TensorRT EP] update oss parser to latest

  #23710 merged Feb 27, 2025

• [ARM CPU] Fix flaky hgemmb ut

  #23814 merged Feb 27, 2025

• Make Nuget CUDA package pipeline 1ES compliant

  #23804 merged Feb 26, 2025

• Upgrade React Native to 0.73

  #23575 merged Feb 26, 2025

• [webgpu] support resize operator

  #23780 merged Feb 26, 2025

• Conveting npm packaging pipeline to 1ES

  #23767 merged Feb 26, 2025

• Make Nuget package pipeline 1ES compliant

  #23803 merged Feb 26, 2025

• [QNN EP] Re-enable several disabled QNN-EP UTs

  #23799 merged Feb 26, 2025

• [VitisAI] add new interfece

  #23777 merged Feb 25, 2025

21 Pull requests opened by 18 people

• Add Snapdragon NPU tutorial

  #23813 opened Feb 25, 2025

• Add OpenCL EP

  #23830 opened Feb 27, 2025

• [WebNN] Better int64 integration

  #23831 opened Feb 27, 2025

• Allow using extended minimal build for several EPs

  #23834 opened Feb 27, 2025

• [mobile/reactnative] Remove namespace from AndroidManifest.XML to resolve warning

  #23847 opened Feb 27, 2025

• [VitisAI] Just for internal test

  #23849 opened Feb 28, 2025

• [OpenVINO]Session Options Appended After AppendExecutionProvider

  #23852 opened Feb 28, 2025

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

  #23855 opened Feb 28, 2025

• Fix enable_pix_capture build for WebGPU

  #23857 opened Mar 1, 2025

• [AIX] External data handling

  #23859 opened Mar 1, 2025

• [WebGPU-EP Native] Add ReduceMean

  #23860 opened Mar 1, 2025

• [WebGPU EP] introduce BiasAdd contrib op

  #23861 opened Mar 1, 2025

• [WIP] gelu related contrib ops

  #23862 opened Mar 2, 2025

• Bump ruff from 0.9.5 to 0.9.9

  #23863 opened Mar 3, 2025

• Doc update relate to EPContext model default name

  #23865 opened Mar 3, 2025

• Move Linux DNNL/OpenVino pipelines to onnxruntime-Ubuntu2204-AMD-CPU machine pool

  #23870 opened Mar 3, 2025

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

  #23873 opened Mar 3, 2025

• Add dawn to ThirdPartyNotices

  #23876 opened Mar 3, 2025

• [QNN EP] Add example that uses a custom CPU allocator for a QNN session

  #23880 opened Mar 4, 2025

• [VitisAI EP] export InferShapes to VitisAIEP

  #23881 opened Mar 4, 2025

• Updated ov version in pipeline (#595)

  #23882 opened Mar 4, 2025

10 Issues closed by 7 people

• Memory leakage from ONNXRuntime environment on Linux machine using C.

  #23798 closed Mar 4, 2025

• [Web] Shall we accept Uint16Array for 'float16' if Float16Array is available

  #23817 closed Mar 3, 2025

• When will v1.20.0 be released for onnxruntime-openvino

  #22783 closed Mar 3, 2025

• [Build] Windows MSVC DNNL build requires <chrono> include

  #23854 closed Feb 28, 2025

• Selecting XNNPACK as execution provider for Android following the documentation example results in program termination

  #23826 closed Feb 28, 2025

• [Build] Android build Failure on ONNX Runtime 1.20.2 compiler doesn't support BFLOAT16

  #23851 closed Feb 28, 2025

• [Build] mp11 not found

  #23821 closed Feb 27, 2025

• Cuda execution provider is not available

  #23833 closed Feb 27, 2025

• [Build] Linux i686 32 bit support

  #23823 closed Feb 27, 2025

• Can't load CUDA on .NET project

  #23810 closed Feb 26, 2025

19 Issues opened by 19 people

• Half of the length that correct output shape

  #23883 opened Mar 4, 2025

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

  #23879 opened Mar 4, 2025

• The Pad operator has a calculation error in the "reflect" mode.

  #23878 opened Mar 4, 2025

• Abs node runs into error with bf16 tensor

  #23875 opened Mar 3, 2025

• Multi GPU support

  #23874 opened Mar 3, 2025

• [OpenVINO] SessionOptionsAppendExecutionProvider_OpenVINO API loads NULL config file

  #23871 opened Mar 3, 2025

• ort.InferenceSession fails silently

  #23869 opened Mar 3, 2025

• preprocess issues around MeanReduce/Reshape nodes and negative axes

  #23868 opened Mar 3, 2025

• [Performance] Why does inference occupy so much memory?

  #23867 opened Mar 3, 2025

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

  #23866 opened Mar 3, 2025

• Attention fusion broken for BART 🤖

  #23864 opened Mar 3, 2025

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

  #23844 opened Feb 27, 2025

• [Build] CUDA version linkage

  #23841 opened Feb 27, 2025

• [Build] how to compile ios static library

  #23835 opened Feb 27, 2025

• [Mobile] Dynamic Shape Challenge: Enabling LLM on QNN-HTP

  #23832 opened Feb 27, 2025

• [CPU EP] GatherND crashes with division by zero when batch dimensions mismatch between input and indices

  #23828 opened Feb 27, 2025

• [Build] ORT, DML, OpenVINO Python wheel build - "OpenVINOExecutionProvider doesn't support memcpy"

  #23824 opened Feb 26, 2025

• [Build] ONNX Runtime Support for Cortex-M33 and Cortex-M7

  #23822 opened Feb 26, 2025

• [Tests] 1 test fails: OptimizerInitializerTest.LoadExternalData: it throws a different type.

  #23816 opened Feb 26, 2025

55 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.

• Enabling L2+ Optimizations for EPs

  #23517 commented on Mar 4, 2025 • 12 new comments

• [webgpu]Add MaxPool and AveragePool

  #23714 commented on Mar 3, 2025 • 10 new comments

• Cleanup CoreML EP's code to remove COREML_ENABLE_MLPROGRAM

  #23490 commented on Feb 27, 2025 • 4 new comments

• (WIP) bitnet and t-mac

  #23540 commented on Mar 3, 2025 • 4 new comments

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

  #23580 commented on Feb 27, 2025 • 2 new comments

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

  #23386 commented on Feb 27, 2025 • 1 new comment

• Upgrade current MacOS-13 to 14

  #23293 commented on Feb 26, 2025 • 1 new comment

• Enable QNN EP weight sharing generation using public API

  #23702 commented on Mar 4, 2025 • 1 new comment

• Integrate KleidiAI for MatMulNBits via MlasQNBitGemm

  #23627 commented on Mar 4, 2025 • 1 new comment

• [webgpu] support Pad operator

  #23141 commented on Mar 4, 2025 • 0 new comments

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

  #22794 commented on Feb 26, 2025 • 0 new comments

• Migrate yarn to npm

  #22116 commented on Mar 4, 2025 • 0 new comments

• [VitisAI] Add vaip Integration Using FetchContent

  #22038 commented on Mar 3, 2025 • 0 new comments

• raise Exception("Incomplete symbolic shape inference") when running "symbolic_shape_infer.py"

  #10484 commented on Mar 4, 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 4, 2025 • 0 new comments

• [Build] Docker build failure with ROCm 6.0 using official Dockerfile for v1.19.2: Segmentation fault in clang++ during composable_kernel compilation

  #23807 commented on Mar 4, 2025 • 0 new comments

• [Performance] fp16 support and performance

  #22242 commented on Feb 25, 2025 • 0 new comments

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

  #23524 commented on Feb 25, 2025 • 0 new comments

• [WebGPU EP] SoftMax Implementation

  #23538 commented on Mar 1, 2025 • 0 new comments

• Migrate Zip-Nuget Package Pipeline to 1ES

  #23609 commented on Mar 3, 2025 • 0 new comments

• Test CUDNN_FRONTEND_SKIP_JSON_LIB=ON

  #23660 commented on Feb 26, 2025 • 0 new comments

• [WIP] enable WebGPU EP in WebAssembly build

  #23697 commented on Mar 1, 2025 • 0 new comments

• [VitisAI] export Graph::SetName to VitisA IEP

  #23731 commented on Mar 3, 2025 • 0 new comments

• [webgpu] Optimize MatMulNBits f16 prefill shader for subgroup size 32

  #23773 commented on Feb 26, 2025 • 0 new comments

• Make python package pipeline 1ES compliant

  #23800 commented on Mar 3, 2025 • 0 new comments

• Make python CUDA package pipeline 1ES compliant

  #23802 commented on Mar 3, 2025 • 0 new comments

• Make Cuda packaging pipeline 1ES compliant

  #23806 commented on Mar 4, 2025 • 0 new comments

• [WIP] Flash attention for generation

  #23808 commented on Feb 26, 2025 • 0 new comments

• [Web] Declaration is not emitted in onnxruntime-node package

  #17979 commented on Feb 26, 2025 • 0 new comments

• debug result is ok, release get NaN output

  #23440 commented on Feb 26, 2025 • 0 new comments

• Is DML being deprecated?

  #23783 commented on Feb 26, 2025 • 0 new comments

• Custom operators is not a registered function/op (python)

  #23566 commented on Feb 26, 2025 • 0 new comments

• [Performance] 40% slowdown in ONNX Resize Operator on CPU

  #23391 commented on Feb 26, 2025 • 0 new comments

• Memory creeping up

  #23348 commented on Feb 26, 2025 • 0 new comments

• TensorRT Provider "Attribute reduction is not supported"

  #23618 commented on Feb 26, 2025 • 0 new comments

• [Feature Request] Request grid_sample 5D support 🌟

  #21382 commented on Feb 26, 2025 • 0 new comments

• Creating TRT Cache much slower on Linux than on Windows

  #23380 commented on Feb 26, 2025 • 0 new comments

• How to build for multiple execution provider?

  #9756 commented on Feb 26, 2025 • 0 new comments

• [Build] Android compatibility with WebGPU

  #23565 commented on Feb 27, 2025 • 0 new comments

• [Build] Non-zero status code

  #23497 commented on Feb 27, 2025 • 0 new comments

• [nodejs-binding] Crash during InferenceSession initialization: "Check failed: node->IsInUse()"

  #23794 commented on Feb 27, 2025 • 0 new comments

• symbolic_shape_infer.py cannot infer torch.nn.normalize

  #23516 commented on Feb 28, 2025 • 0 new comments

• [Performance] Multithreading for DequantizeLinear

  #23395 commented on Feb 28, 2025 • 0 new comments

• [Performance] Preload model before inference

  #23513 commented on Mar 1, 2025 • 0 new comments

• [Web] WebGPU and WASM Backends Unavailable within Service Worker

  #20876 commented on Mar 1, 2025 • 0 new comments

• [Build] protocol buffer compiler error MSB8066

  #23529 commented on Mar 2, 2025 • 0 new comments

• [Web] BiRefNet_T not working on webgpu

  #21968 commented on Mar 2, 2025 • 0 new comments

• [Performance] Speed-up TensorRT engine compilation

  #23546 commented on Mar 3, 2025 • 0 new comments

• System.EntryPointNotFoundException: Unable to find an entry point named 'OrtSessionOptionsAppendExecutionProvider_CUDA' in DLL 'onnxruntime'.

  #22559 commented on Mar 3, 2025 • 0 new comments

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

  #23556 commented on Mar 3, 2025 • 0 new comments

• [WebGPU] `Kernel "[GroupQueryAttention] /model/layers.0/attn/GroupQueryAttention" failed. Error: Input "key" is expected to have 3, 4, or 5 dimensions".`

  #22987 commented on Mar 3, 2025 • 0 new comments

• Onnxruntime using OpenVINO for older version Intel UHD630

  #23735 commented on Mar 3, 2025 • 0 new comments

• RoiAlign CPU is not aligned to pixel centers (per the Mask RCNN paper and Facebook's Detectron2 implementation)

  #6921 commented on Mar 3, 2025 • 0 new comments

• [Performance]Do onednn executors depend on Intel platform

  #23795 commented on Mar 4, 2025 • 0 new comments

• [Build] Cross-compile for Android on Windows error

  #23796 commented on Mar 4, 2025 • 0 new comments