47 Pull requests merged by 30 people

• [CUDA] upgrade cudnn front end to 1.12

  #24977 merged Jun 8, 2025

• [DML] Restore compatibility with Windows Sdk 10.0.17134.0 (build 1809)

  #24950 merged Jun 6, 2025

• [web] handle script URL missing when in ESM

  #24968 merged Jun 6, 2025

• Adding Graiphic to the list of partners

  #24975 merged Jun 6, 2025

• [CoreML] Update Conv and Softmax ops

  #24594 merged Jun 6, 2025

• [WebNN] Add validation for node input tensor ranks

  #24951 merged Jun 6, 2025

• faster shut down

  #24891 merged Jun 6, 2025

• Allow FP16 math in flash attention

  #24953 merged Jun 6, 2025

• Revert "[webgpu] Bump version of Dawn to c3999d7e3 (#24935)"

  #24966 merged Jun 5, 2025

• [QNN EP] Fix 16x16 MatMul translation

  #24846 merged Jun 5, 2025

• [QNN EP] Add 16x16 Gemm translation

  #24849 merged Jun 5, 2025

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

  #24820 merged Jun 5, 2025

• [cpu] fix Sigmoid f32 implementation on non-amd64 platform

  #24749 merged Jun 5, 2025

• fix WebAssembly single thread build

  #24933 merged Jun 5, 2025

• Update deprecated CUDA api

  #24733 merged Jun 5, 2025

• [ROCm EP]Add ROCm execution provider to excluded EP for test with Cuda EP

  #24961 merged Jun 5, 2025

• [ROCM EP] - Updates to enable HIPBLAS API updates (#122)

  #24885 merged Jun 5, 2025

• [MIGraphX EP]Add support for skipLayerNormalization (#123)

  #24886 merged Jun 5, 2025

• [CUDA] fp16 intB gemm scale only kernel

  #24955 merged Jun 5, 2025

• Add ONNX RMSNormalization(23)

  #24875 merged Jun 4, 2025

• Bump ruff from 0.11.11 to 0.11.12

  #24923 merged Jun 4, 2025

• Fix libonnxruntime4j_jni.so 16KB page size compatibility on Android ARM64

  #24947 merged Jun 4, 2025

• [webgpu] fix a build warning in latest GCC

  #24944 merged Jun 4, 2025

• Add kleidiai to onnxruntime_EXTERNAL_LIBRARIES in setup_kleidiai().

  #24692 merged Jun 4, 2025

• A temporary fix for layout opt level to unblock react native android CI

  #24948 merged Jun 4, 2025

• Skip _tpause call for _M_ARM64EC in spin_pause.cc.

  #24942 merged Jun 4, 2025

• Fix initialization of same_node_ in TreeEnsemble

  #24654 merged Jun 4, 2025

• Run Matmul 8 bit tests on CPU builds

  #24946 merged Jun 4, 2025

• Extend OrtAllocator API to get Allocator statistics

  #24785 merged Jun 4, 2025

• Add support for bool type in SplitToSequence

  #24929 merged Jun 3, 2025

• Support activation broadcasting in XNNPACK Matmul

  #24908 merged Jun 3, 2025

• [MIGraphX EP] Add migx ep fp8 support and int4 weights

  #23534 merged Jun 3, 2025

• Remove ep_weight_sharing_ctx_gen tool from QNN EP python wheel

  #24895 merged Jun 3, 2025

• [webgpu] Bump version of Dawn to c3999d7e3

  #24935 merged Jun 3, 2025

• Fix a test failure in mlas

  #24930 merged Jun 3, 2025

• Cast Nodes Fusion

  #24842 merged Jun 3, 2025

• Corrected compilation errors

  #24893 merged Jun 3, 2025

• Fusing Initializers with Graph Transforms

  #24726 merged Jun 3, 2025

• [CUDA] FpA IntB Gemm Weight Conversion in GPU

  #24914 merged Jun 3, 2025

• [WebGPU EP] adjust test case to skip CoreML EP

  #24927 merged Jun 2, 2025

• [QNN-EP] Fuse ChannelShuffle pattern

  #24904 merged Jun 2, 2025

• [QNN-EP] Replace Upsample with Resize during Quantization

  #24896 merged Jun 2, 2025

• Update README.md: remove the build pipeline status section

  #24548 merged Jun 2, 2025

• [DNNL] update tests

  #24921 merged Jun 2, 2025

• [WebGPU] Unify core implementations of GEMM and MatMul

  #24586 merged Jun 2, 2025

• Improve Windows ETW callback registration and fix issues

  #24877 merged Jun 2, 2025

• Weaken dxcore dependency

  #24845 merged Jun 2, 2025

25 Pull requests opened by 19 people

• Fix QNN error messages being logged as VERBOSE instead of ERROR

  #24931 opened Jun 2, 2025

• Remove `sequentially_access_by_threads` for Conv

  #24938 opened Jun 3, 2025

• [QNN EP] Add multiple Op support in QNN EP

  #24939 opened Jun 3, 2025

• Integration with ONNX 1.18.0

  #24945 opened Jun 3, 2025

• Fix illegal update model opset version in static_quantize_runner

  #24949 opened Jun 4, 2025

• [QNN EP] Transform SFIXED to UFIXED on InstanceNorm scale

  #24952 opened Jun 4, 2025

• DRAFT [CoreML] Add boolean values to CoreML

  #24956 opened Jun 4, 2025

• Cherry pick telemetry changes from win-onnxruntime

  #24957 opened Jun 4, 2025

• Support 8 bit weights "unpacked" compute mode in MatmulNBits kernel

  #24959 opened Jun 5, 2025

• Enable VTCM Back Up Buffer Sharing

  #24962 opened Jun 5, 2025

• [DRAFT][QNN EP] Enable QNN EP to record to ORT profiler

  #24969 opened Jun 6, 2025

• [webgpu] add cases to cover the subgroup matrix path

  #24970 opened Jun 6, 2025

• Add Int4 and UInt4 support for Cast

  #24973 opened Jun 6, 2025

• Fix static_cast error with int8_t in TensorCasterNoSat template

  #24974 opened Jun 6, 2025

• [CoreML] mark input params variable unused

  #24976 opened Jun 6, 2025

• Fix in-memory initializer handling for non-CPU device

  #24978 opened Jun 6, 2025

• Fix c/c++ document generation

  #24979 opened Jun 6, 2025

• Add RotaryEmbeddings(23)

  #24980 opened Jun 6, 2025

• Fix Int4x2/UInt4x2 casting compilation errors with specialized TensorCaster templates

  #24981 opened Jun 6, 2025

• Update transformers package version to 4.48.0

  #24982 opened Jun 6, 2025

• Update nv_basic_test.cc

  #24983 opened Jun 7, 2025

• Use c++17 to simplify the AllocatorDefaultAllocAligned function

  #24984 opened Jun 8, 2025

• Make OrtDevice generic to support plugin EP usage.

  #24985 opened Jun 8, 2025

• Remove DML and the legacy WinML nuget/python packages

  #24986 opened Jun 8, 2025

• Rewrite iostream sinks

  #24987 opened Jun 8, 2025

748 Issues closed by 15 people

• [ONNXRuntimeError] Non-zero status code returned while running SkipLayerNormalization node.

  #4779 closed Jun 8, 2025

• Segmentation error when using graph optimization

  #5175 closed Jun 8, 2025

• Unsupported OpenVINO version on Windows when shortcut is used

  #5435 closed Jun 8, 2025

• Quantize bert sample code seem to be wrong.

  #5644 closed Jun 8, 2025

• Build with DNNL execution provider failing on macOS, but working on Linux

  #5783 closed Jun 8, 2025

• Calculation error of reduction operator when axis has duplicate value

  #5848 closed Jun 8, 2025

• "Windows fatal exception: access violation" when trying to run custom ONNX model.

  #5872 closed Jun 8, 2025

• Different output when running on CUDA (compared to CPU and keras)

  #5798 closed Jun 8, 2025

• Error "Unsupported tensor type of (unknown type)" when running Unique on int32 type

  #5908 closed Jun 8, 2025

• Python Segmentation fault (core dumped) on cuda11.0-runtime when built on cuda11.0-devel

  #6339 closed Jun 8, 2025

• ArmNN Execution Provider does not build

  #6486 closed Jun 8, 2025

• onnxruntime import fails with error: libcudart.so.10.2: cannot open shared object file: No such file or directory

  #6497 closed Jun 8, 2025

• Onnxruntime TensorRT create one cache when models same structure

  #6455 closed Jun 8, 2025

• Failed to run inference session on 8bit quantized onnx model

  #6430 closed Jun 8, 2025

• [BERT-Squad][opset-8][Model Zoo] Inference fails on bert/embeddings/one_hot op

  #6548 closed Jun 8, 2025

• VitisAIExecutionProvider is not shown in the available providers

  #6568 closed Jun 8, 2025

• [c api] Why there will be about 300MB GPU memory cost when i infer on cpu ?

  #6567 closed Jun 8, 2025

• albert quantized

  #6847 closed Jun 8, 2025

• LabelEncoder with NaN float key doesn't map

  #6681 closed Jun 8, 2025

• Symbolic shape inference for IF operator

  #6884 closed Jun 8, 2025

• onnxruntime can't recognize my custom op even though onnx can!

  #7254 closed Jun 8, 2025

• Memory leaks and valgrind errors when running with TensorRT

  #7286 closed Jun 8, 2025

• Getting error during inference with onnx built with openvino support

  #7317 closed Jun 8, 2025

• Possible Memory leak in CXX Api

  #7084 closed Jun 8, 2025

• Access violation when using TensorRT ExecutionProvider on multiple GPU

  #7322 closed Jun 8, 2025

• Failed to load library ./libonnxruntime_providers_shared.so

  #7485 closed Jun 8, 2025

• Missing 'coloredlogs', 'sympy' when compiling from source

  #7527 closed Jun 8, 2025

• Bug in MatMulInteger

  #7524 closed Jun 8, 2025

• "Quantize_qat" for PyTorch QAT model

  #7764 closed Jun 8, 2025

• ONNX Runtime Inference time difference for different runs using ArmNN execution provider

  #7898 closed Jun 8, 2025

• ImportError: cannot import name 'get_all_providers' in importing onnxruntime

  #7887 closed Jun 8, 2025

• ONNX Exported BART Model Performance is degraded than native pytorch on T4

  #7796 closed Jun 8, 2025

• Example script for quantization aware training

  #7996 closed Jun 8, 2025

• CUDNN failure 3: CUDNN_STATUS_BAD_PARAM for FusedConv node at inference time

  #8024 closed Jun 8, 2025

• [python] [build] FileNotFoundError: Unable to find 'test_execution_provider.dll'

  #8140 closed Jun 8, 2025

• Quantized model does not support dynamic axes

  #8169 closed Jun 8, 2025

• CPU cores and threads control

  #8193 closed Jun 8, 2025

• [onnxruntime] - session.run return negative number

  #8104 closed Jun 8, 2025

• Completely disable logging locally and instead send to a server

  #8107 closed Jun 8, 2025

• int8 quantization on GPU support with transformers like bert, gpt2

  #8368 closed Jun 8, 2025

• When I batched the data using the onnx model exported from the Yolov3_spp model, an error occurred

  #8194 closed Jun 8, 2025

• Raspberry Pi 4 Myriad VPU support

  #8285 closed Jun 8, 2025

• Memory leaks observed for DNNL EP

  #8245 closed Jun 8, 2025

• Get wrong constant folding result when calling InferenceSession

  #8422 closed Jun 8, 2025

• Why exported gpt2.onnx has 15 inputs? (or How do I assign the past_state using C++ API)

  #8528 closed Jun 7, 2025

• Model run with TRT， But all node run with cuda

  #8543 closed Jun 7, 2025

• TensorRT provider doesn't support CustomOp automatically fall back to CUDA EP

  #8656 closed Jun 7, 2025

• The performance not got imporve

  #8636 closed Jun 7, 2025

• LNK2001:unresolved external symbol OrtSessionOptionsAppendExecutionProvider_Tensorrt

  #8681 closed Jun 7, 2025

• Program received signal SIGSEGV, Segmentation fault when I Custruct Ort::Session

  #8707 closed Jun 7, 2025

• Batch infer occurs NAN

  #8766 closed Jun 7, 2025

• Error while converting the model: attempted relative import beyond top-level package

  #8785 closed Jun 7, 2025

• TensorRT EP high memory consumption

  #8792 closed Jun 7, 2025

• Partial graph execution

  #8705 closed Jun 7, 2025

• Error: build with TensorRT support on NVIDIA Jetson Xavier

  #8808 closed Jun 7, 2025

• Duplicate node name issue on scrfd (Face Detection) quantization

  #8811 closed Jun 7, 2025

• import torch greatly increase ONNX model gpu memory

  #8823 closed Jun 7, 2025

• Performance degradation when a model fully fall backs to MLAS from an EP

  #8812 closed Jun 7, 2025

• custom_op_library test in python cause segmentation fault

  #8829 closed Jun 7, 2025

• Apply GraphTransformer EP specific rules to DNNL EP

  #8949 closed Jun 7, 2025

• gpt-2_onnx_models that do not include past_states - How does it work?

  #8849 closed Jun 7, 2025

• onnx slower than original pytorch ner model?

  #8957 closed Jun 7, 2025

• About custom op

  #8956 closed Jun 7, 2025

• Several failing tests with CUDA execution provider

  #8972 closed Jun 7, 2025

• Support specify the trt_engine_cache_path of OrtTensorRTProviderOptionsV2 with wide characters

  #8975 closed Jun 7, 2025

• Unhandeled exception Microsoft C++ exception: std::runtime_error at memory location

  #9030 closed Jun 7, 2025

• Add allocator_stats.h to cmake build install

  #9072 closed Jun 7, 2025

• The inference results are different using onnxruntime-cpu and onnxruntime-gpu

  #9037 closed Jun 7, 2025

• The example for quantization doesn't work

  #9069 closed Jun 7, 2025

• [GPU] The quantized-onnx-model is worse than onnx-model.

  #9115 closed Jun 7, 2025

• ONNX pointwise convolution and group convolution are too slow

  #9192 closed Jun 7, 2025

• Problem with include in provider_options.h

  #9148 closed Jun 7, 2025

• onnxruntime 1.8.2 not yet available on pypi

  #9118 closed Jun 7, 2025

• mlas build fail

  #9187 closed Jun 7, 2025

• [ONNXRuntimeError] : 1 : FAIL : Non-zero status code returned while running FusedConv node.

  #9194 closed Jun 7, 2025

• ORT 1.9.0 Build: nvcc error : 'ptxas' died due to signal 11 (Invalid memory reference)

  #9222 closed Jun 7, 2025

• [OpenVINO-EP] Windows Build Nuget is broken

  #9271 closed Jun 7, 2025

• arm install error

  #9225 closed Jun 7, 2025

• m_session.RUN crash

  #9226 closed Jun 7, 2025

• [build] Use vendored version of nlohmann-json instead of submodule

  #9286 closed Jun 7, 2025

• Onnxruntime-gpu always runs on CPU and never on GPU, [Roberta]

  #9292 closed Jun 7, 2025

• Different memory consumption depending on the Nvidia generation.

  #9289 closed Jun 7, 2025

• Offline optimization mode with CUDA EP

  #9325 closed Jun 7, 2025

• onnxruntime-1.8.2. tensorrt_execution_provider.cc:539 onnxruntime::TensorrtExecutionProvider::TensorrtExecutionProvider [ONNXRuntimeError] : 11 : EP_FAIL : TensorRT EP could not open shared library from

  #9364 closed Jun 7, 2025

• Using an svm regressor from sklearn

  #9348 closed Jun 7, 2025

• Rule-based graph transformers don't work on graphs with functions

  #9296 closed Jun 7, 2025

• Build on Jetson Xavier NX does not produce c++ API include and libs

  #9371 closed Jun 7, 2025

• GetTensorMutableData is destructive

  #9367 closed Jun 7, 2025

• Graph is not a DAG when trying to convert HF model inputs to `int32`

  #9389 closed Jun 7, 2025

• measure inference time

  #9437 closed Jun 7, 2025

• OpenVINO EP not available

  #9387 closed Jun 7, 2025

• The SVMClassifier node outputs 2 class scores when it should be a single score for non-probabilistic models

  #9429 closed Jun 7, 2025

• 1 - onnxruntime_test_all fails when building with OpenVINO support

  #9439 closed Jun 7, 2025

• Problem with output data when using quantized onnx model

  #9592 closed Jun 7, 2025

• Unable to restrict ONNX graph with multiple output nodes to single CPU core

  #9565 closed Jun 7, 2025

• Build flag issues for a GPU-enabled wheel

  #9467 closed Jun 7, 2025

• bad accuracy on own quantized(INT8) yolov3 model

  #9598 closed Jun 7, 2025

• Dynamic quantization of multilingual miniLM - output does not match float32 version. Onnxruntime 1.9.0

  #9599 closed Jun 7, 2025

• build the master branch with cuda10.2/cudnn8.0

  #9628 closed Jun 7, 2025

• onnx inference with multiprocess

  #9625 closed Jun 7, 2025

• Weight's raw_data is empty so it leads to quantize weight failure

  #9635 closed Jun 7, 2025

• how to support cuda10.2?

  #9654 closed Jun 7, 2025

• Need to register an op of QLinearFC

  #9642 closed Jun 7, 2025

• float16 support in web runtimes

  #9758 closed Jun 7, 2025

• inconsistent inference time

  #9667 closed Jun 7, 2025

• whl install BUG

  #9666 closed Jun 7, 2025

• Quantization shrink model size but no inference speed improvement

  #9694 closed Jun 7, 2025

• Can not build in RTX 3090

  #9779 closed Jun 7, 2025

• Optimization changes the graph so it cannot run on Cuda

  #9778 closed Jun 7, 2025

• onnxruntime.InferenceSession

  #9829 closed Jun 7, 2025

• tensorrt backend version question

  #9846 closed Jun 7, 2025

• `onnxruntime.InfereceSession` will stuck or crash in docker with `onnxruntime>1.6`

  #9862 closed Jun 7, 2025

• some model convert error question when tensorrt backend

  #9853 closed Jun 7, 2025

• Error message in ORT 1.9.0 is not as clear as previous ORT versions

  #9878 closed Jun 7, 2025

• different inference result between onnxruntime and pytorch

  #9890 closed Jun 7, 2025

• custom op can not supported by onnxruntime

  #9865 closed Jun 7, 2025

• ORT+TensorRT build, "--config Debug" works but "--config Release" failed

  #9934 closed Jun 7, 2025

• Run multi-thread with CUDA

  #9891 closed Jun 7, 2025

• Cast f32 -> bf16 -> f32 does not work as expected for graph inputs

  #9915 closed Jun 7, 2025

• ORT Eager aten view operation needs to share the same semantic with Pytorch view

  #9948 closed Jun 7, 2025

• [BUG on Onnxruntime+TensorRT] Assertion failed: input_dims.nbDims == 4 || input_dims.nbDims == 5 instanceNormalizationPlugin/instanceNormalizationPlugin.cu:181

  #9945 closed Jun 7, 2025

• pip3 onnx==1.8.0 is not working on Jetson Xavier

  #9984 closed Jun 7, 2025

• ONNX model (roberta) that is 300MB in size uses around 1.9 GB of memory.

  #9949 closed Jun 7, 2025

• Session initialization takes long time on NVIDIA RTX type GPUs

  #9990 closed Jun 7, 2025

• C++ 编译动态链接器 出现问题

  #9987 closed Jun 7, 2025

• I can not run the tensorrt example "quantized BERT model example"

  #10045 closed Jun 7, 2025

• how to create float tensor with missing value using java runtime

  #10036 closed Jun 7, 2025

• Lower Resnet50 accuracy with ONNXRuntime

  #10021 closed Jun 7, 2025

• Linking custom library to onnxruntime fails on Linux

  #10063 closed Jun 7, 2025

• [BUG] Registered type of RoiAlign does not work

  #10046 closed Jun 7, 2025

• Do threading settings have any effect on GPU EPs

  #10081 closed Jun 7, 2025

• GRU runtime error.

  #10069 closed Jun 7, 2025

• Python bindings give RuntimeError on multi input models

  #10089 closed Jun 7, 2025

• segmentation fault when get subgraph in tensorrt provider

  #10096 closed Jun 7, 2025

• ReduceMean consumes an unreasonable amount of VRAM

  #10099 closed Jun 7, 2025

• onnxruntime_test_all fails at Gemm, Conv, Pool and Concat tests

  #10119 closed Jun 7, 2025

• Why I met Type 'seq(tensor(int64))' of operator (MemcpyFromHost) is invalid when using onnxruntime.InferenceSession() in GPU, and How to resolve it? On emergency hold，thanks!

  #10126 closed Jun 7, 2025

• Using tensorrt provider occasionally see dramatically increased inference time

  #10159 closed Jun 7, 2025

• onnxruntime int8 quant slower than pytorch

  #10135 closed Jun 7, 2025

• Validate usage of dim_value (values should be > 0) and dim_param (all values with the same string should equate to the same size) in shapes in the model.

  #10148 closed Jun 7, 2025

• Dynamic Shape performance

  #10194 closed Jun 7, 2025

• GPU caculation resources confict between onnxruntime-directML inference and SDL2 render/display

  #10203 closed Jun 7, 2025

• MeanVarianceNormalization on CPU doesn't use epsilon

  #10228 closed Jun 7, 2025

• how to merge two of Ort::Value together?

  #10224 closed Jun 7, 2025

• Gather node discrepancy for out of index indices

  #10234 closed Jun 7, 2025

• VS2019 Project Properties failed to load with ONNX DirectML NuGet package.

  #10236 closed Jun 7, 2025

• which onnxruntime version did cuda 11.2 need

  #10229 closed Jun 7, 2025

• Run model with a cupy array on CUDA

  #10238 closed Jun 7, 2025

• how to reduce gpu memory usage?

  #10239 closed Jun 7, 2025

• Gemm layer is not quantized with QGemm node but with QLinearMatMul + QLinearAdd

  #10278 closed Jun 7, 2025

• ONNX Runtime DLL load Error

  #10290 closed Jun 7, 2025

• Python's flatbuffers version API

  #10291 closed Jun 7, 2025

• `Non-zero status code returned while running MatMul node` once too many requests

  #10343 closed Jun 7, 2025

• Floating point exception with FasterSCNN model

  #10402 closed Jun 7, 2025

• Quantization of video action recognition model

  #10422 closed Jun 7, 2025

• [Documentation] some quantized operators not present in ContribOperators.md

  #10403 closed Jun 7, 2025

• ONNX Runtime C++ with OpenVino EP stuck at Run

  #10409 closed Jun 7, 2025

• quadro telsa is test? OK?

  #10438 closed Jun 7, 2025

• Issues building with both --use_rocm and --build_shared_lib specified

  #10466 closed Jun 7, 2025

• Build linker error

  #10457 closed Jun 7, 2025

• Fusion Reshape bug whith multiple graphs in a model

  #10477 closed Jun 7, 2025

• Make ONNX graphs with fused ONNXRuntime plugins runnable by TensorRT execution provider ?

  #10509 closed Jun 7, 2025

• NCHWc Data format Query

  #10537 closed Jun 7, 2025

• Strange behavior of ReduceMax with NaN Values

  #10514 closed Jun 7, 2025

• FATAL ERROR: Ineffective mark-compacts near heap limit Allocation failed - JavaScript heap out of memory

  #12014 closed Jun 7, 2025

• Error compiling time in ubuntu Jetson Nano

  #10555 closed Jun 7, 2025

• Add logging to file option

  #10586 closed Jun 7, 2025

• Error Inferring Yolov8 node_args.cpp Vitis Ai provider

  #17534 closed Jun 7, 2025

• [Web] The onnxruntime documentation displays an exception regarding c++

  #17542 closed Jun 7, 2025

• FIPS compliance

  #17561 closed Jun 7, 2025

• [Documentation Request]

  #17646 closed Jun 7, 2025

• [Mobile]

  #17622 closed Jun 7, 2025

• can not run on windows server2012 r2

  #17591 closed Jun 7, 2025

• [Web] Light GBM .ort model multiple times larger than .onnx model

  #17691 closed Jun 7, 2025

• [Performance] Abnormal latencies on certain tasks and a GPU on standby.

  #17720 closed Jun 7, 2025

• LayerNormalization in fp16 fails on data with large variance

  #17692 closed Jun 7, 2025

• Can ORT be built with BuildTools 16.11?

  #17693 closed Jun 7, 2025

• [Performance] Some allocations perform even after many Run() invocations on fixed inputs

  #17758 closed Jun 7, 2025

• [Build] Dangling Reference Error during build

  #17759 closed Jun 7, 2025

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

  #17763 closed Jun 7, 2025

• [Documentation] broken link to https://onnxruntime.ai/docs/api/js/interfaces/TensorFromImageOptions.html

  #17755 closed Jun 7, 2025

• [Build] Is it possible to build onnxruntime with DirectML EP statically?

  #17774 closed Jun 7, 2025

• [Documentation] How do I use the DirectML EP from an NUnit unit test?

  #17793 closed Jun 7, 2025

• [Build] Building wasm static lib in docker

  #17780 closed Jun 7, 2025

• [Build] dotnet install onnxruntime

  #17794 closed Jun 7, 2025

• session.disable_fallback() not working

  #17801 closed Jun 7, 2025

• Does onnxruntime support static libraries in windows or other platform？

  #17839 closed Jun 7, 2025

• the github donwload *.so file not use mkl?

  #17818 closed Jun 7, 2025

• Segfault on session creation with custom `MockedOrtAllocator` in `MlasSgemmCopyPackB`

  #17867 closed Jun 7, 2025

• dockerfile with different stage

  #17812 closed Jun 7, 2025

• [Build]

  #17941 closed Jun 7, 2025

• [Performance] OnnxRuntime 1.14.1 Cuda GPU C++ keeps increasing memory usage on running session

  #17938 closed Jun 7, 2025

• StringJoin issue on CUDA EP: scalar input could only be applied to CPU tensor

  #18018 closed Jun 7, 2025

• CreateTensor with tensor size of bigger than 65536 causes BAD_ACCESS in Run on macOS

  #17884 closed Jun 7, 2025

• Models with multiple outputs produce different results when the order of irrelevant lines are changed

  #18081 closed Jun 7, 2025

• [Documentation] Clearer TRT dependencies

  #18073 closed Jun 7, 2025

• [Performance] Remove some transpose ops in layout conversion.

  #18128 closed Jun 7, 2025

• Output mismatch of duplicate torch.Tensor.to nodes after optimization

  #18211 closed Jun 7, 2025

• Onnxruntime Memory [Web]

  #18165 closed Jun 7, 2025

• [Training] On Device Training is not working

  #18168 closed Jun 7, 2025

• nodejs + tensorrt :: how to set execution provider options?

  #18380 closed Jun 7, 2025

• How to develop the perfemance for the CUDA.

  #18391 closed Jun 7, 2025

• [Build] Java API with oneDNN error: LoadLibrary failed with error 126

  #18292 closed Jun 7, 2025

• Output mismatch of torch.Tensor.to due to an extra torch.Tensor.transpose node

  #18231 closed Jun 7, 2025

• Output mismatch of torch.Tensor.argmax when applying equivalent torch.Tensor.transpose without optimization

  #18233 closed Jun 7, 2025

• [ONNXRuntimeError] : 6 : RUNTIME_EXCEPTION : Non-zero status code returned while running Gather node.

  #18417 closed Jun 7, 2025

• [Performance] Transpose swap with Split in optimize_model

  #18433 closed Jun 7, 2025

• [Performance] LLM Accuracy Significantly Dropped after dynamic_quantization

  #18430 closed Jun 7, 2025

• [Performance] Sequential IO binding shows similar results with and without IO binding

  #18475 closed Jun 7, 2025

• ONNX Runtime v1.16.2 crashes with Unity IDE

  #18441 closed Jun 7, 2025

• [Performance] MT5 model float16 parity check failed

  #18505 closed Jun 7, 2025

• The 'then' statement is equivalent to the 'else' statement. HouseholderSequence.h 322

  #18676 closed Jun 7, 2025

• An item with the same key 'Tanh' has already been added. onnx_transpose_optimization.cc 2169

  #18677 closed Jun 7, 2025

• Possible incorrect order of arguments passed to 'UnsafeCast' function: 'dst_type1' and 'src_type1'. insert_cast_transformer.cc 368

  #18678 closed Jun 7, 2025

• loading model onnx file fails on exception

  #18690 closed Jun 7, 2025

• [Documentation] Unclear on how to share allocator between sessions in C++

  #18682 closed Jun 7, 2025

• Session's inner variables not refreshed between 2 runs

  #18742 closed Jun 7, 2025

• Unable to compile UWP project using .NET native with OnnxRuntime reference[Build]

  #18771 closed Jun 7, 2025

• Device selection through onnxruntime-directml

  #18716 closed Jun 7, 2025

• [Training] quantize_static Histogram based calibration methods expect fixed shape outputs

  #18792 closed Jun 7, 2025

• [Build] [CMake]More targets need to output.

  #18872 closed Jun 7, 2025

• Loading .onnx model on AWS Lambda image returns error.

  #18860 closed Jun 7, 2025

• XGBoost incremental training, issue with ONNX Conversion

  #18841 closed Jun 7, 2025

• [Build] libonnxruntime.so file is occupying 751 MB out of the entire Jar (788.4 MB)

  #18859 closed Jun 7, 2025

• How to create input tensor for multiple input examples (batch mode inference)?

  #18881 closed Jun 7, 2025

• [Performance] TensorrtEP bad allocation

  #18887 closed Jun 7, 2025

• [Build] Error when building a nujet package with OpenVINO and DML

  #19031 closed Jun 7, 2025

• [Documentation] [Question] Why some tests cannot be performed in Parallel ?

  #19042 closed Jun 7, 2025

• ORT returns incorrect result for UINT8 Matmul on specific CPU

  #19109 closed Jun 7, 2025

• [Build] onnxruntime infer dynamic scale imgs in windows

  #19075 closed Jun 7, 2025

• onnxruntime causing high CPU while typing in Visual Studio 17.8.4

  #19132 closed Jun 7, 2025

• Segmentation fault with CUDA execution provider

  #19165 closed Jun 7, 2025

• Can ORT version 1.16.3 support Moe?

  #19159 closed Jun 7, 2025

• ReduceSum models errors out when noop_with_empty_axes is True and axes is empty.

  #19147 closed Jun 7, 2025

• Can i quantize sigmoid activation (to QLinearSigmoid for example) in dynamic quantization?

  #19139 closed Jun 7, 2025

• Performance Issue: Onnx Runtime Train Loss Reduction is very Less in every epoch and gets saturated at after 10 epochs

  #19185 closed Jun 7, 2025

• Invoking element-wise Min native operator from custom operator [C++]

  #19181 closed Jun 7, 2025

• [Training] How to train all the layers of the onnx training model

  #19186 closed Jun 7, 2025

• [Build] onnxruntime_providers_shared not provided.

  #19191 closed Jun 7, 2025

• For GPT-NEO Model in onnx format -- [ONNXRuntimeError] : 6 : RUNTIME_EXCEPTION : Non-zero status code returned while running Where node.

  #19214 closed Jun 7, 2025

• Compilation OpenVINO EP does not work on macOS

  #19217 closed Jun 7, 2025

• [Documentation] build on arm32v7

  #19224 closed Jun 7, 2025

• How to initialize onnx input CreateTensor with gpu meory instead of CreateCpu

  #19240 closed Jun 7, 2025

• [Performance] Getting different result in C# ONNX runtime and Python ONNX runtime

  #19226 closed Jun 7, 2025

• [python] Include 'per_channel' attribute when calibrating

  #19223 closed Jun 7, 2025

• Support for Add/Sub with int8, int16, uint8, uint16 in CPU Provider

  #19231 closed Jun 7, 2025

• [Training] Whether to support weight per_channel QAT

  #19241 closed Jun 7, 2025

• unable to fold certain constants

  #19249 closed Jun 7, 2025

• swin_transformer in pytorch fails to run after converting to onnx

  #19280 closed Jun 7, 2025

• [Training] Forcing Parameter's Values during On Device Training

  #19261 closed Jun 7, 2025

• Pow inference Error

  #19299 closed Jun 7, 2025

• [Build] Handling Multiple ONNX Runtime Sessions Sequentially in Docker

  #19309 closed Jun 7, 2025

• failed:Node (/pool_1/MaxPool) Op (MaxPool) [ShapeInferenceError] Attribute strides has incorrect size

  #19349 closed Jun 7, 2025

• [Build] Minimal build of onnxruntime 1.17 on Ubuntu 20.04 fails Python 3.12

  #19422 closed Jun 7, 2025

• ONNXRuntimeError:6 when trying to run a yolov7 onnx file

  #19359 closed Jun 7, 2025

• clip model 'ViT-H-14-378-quickgelu__dfn5b' gives errors with OPENVINO

  #19365 closed Jun 7, 2025

• Getting different results with DirectML vs CPU or CUDA for Tensorflow Object Detection model

  #19352 closed Jun 7, 2025

• [Training] ORT Gradient Builder expects optinal outpts for LayerNormalization Op

  #19427 closed Jun 7, 2025

• How to build react native runtime from source?

  #19472 closed Jun 7, 2025

• [Documentation] Incorrect cuDNN doc links

  #19473 closed Jun 7, 2025

• <OnnxValue>.getValue() returns non-parseable java object

  #19440 closed Jun 7, 2025

• [Performance] BGE Reranker / BERT Crossencoder Onnx model latency issue

  #19494 closed Jun 7, 2025

• [Web] Running ORT model results in NaN values output

  #19491 closed Jun 7, 2025

• [Build] Broken binaries for Windows UWP: LNK1181 error missing delayimp.lib and dependency MSVCP140D.dll and VCRUNTIME140D.dll

  #19495 closed Jun 7, 2025

• [Performance] Perf decrease

  #19523 closed Jun 7, 2025

• Not all CUDA operators support bfloat16 that should

  #19502 closed Jun 7, 2025

• [ONNXRuntimeError] when starting Inferencesession with exported model

  #19537 closed Jun 7, 2025

• [Build] SIGSEGV calling into native library from Java on MacOS on M2 Mac

  #19512 closed Jun 7, 2025

• [Web] cannot load onnx model in a vite/react project, because of error expected magic word 00 61 73 6d, found 3c 21 44 4f @+0

  #19556 closed Jun 7, 2025

• Unable to quantize `torchvision.detection` models

  #19544 closed Jun 7, 2025

• Converting model with olive and onnxruntime 1.17 produces OnnxRuntimeException in a C# Windows app

  #19542 closed Jun 7, 2025

• [Build] MSVC on Windows does not warn about unused variables

  #19565 closed Jun 7, 2025

• How to make ONNX runtime support a new GPGPU device ?

  #19561 closed Jun 7, 2025

• Test execute only the first EP when being passed a positive EP list instead of an excluded list

  #19573 closed Jun 7, 2025

• [Mobile] Bug in android implementation for loading large onnx models

  #19599 closed Jun 7, 2025

• ROCM EP convolution fails due to missing

  #19566 closed Jun 7, 2025

• Sequence and Map output support in C# not working

  #19571 closed Jun 7, 2025

• OnnxRuntime nuget package uses build and not buildTransitive

  #19676 closed Jun 7, 2025

• [Performance] Onnx model atrociously slow in runtime

  #19648 closed Jun 7, 2025

• [Documentation] broken links of document

  #19660 closed Jun 7, 2025

• Model cannot be used because the inference status is rejected

  #19624 closed Jun 7, 2025

• Add an option in OpenVINOProviderOptions to support the queue-based overload for creating ClContext

  #19697 closed Jun 7, 2025

• How to run onnx model with pure C99 code?

  #19738 closed Jun 7, 2025

• I found that `preferredOutputLocation` only used in wasm/wasm-core-impl.ts. But webgl only use backend-onnxjs.ts

  #19719 closed Jun 7, 2025

• onnxruntime error while inference on cuda

  #19754 closed Jun 7, 2025

• [Training] Shape change is not sync when serialize graph to proto

  #19741 closed Jun 7, 2025

• [Web] ort-web : Failed to load external data file "./model.onnx_data", error: Module.MountedFiles is not available

  #19752 closed Jun 7, 2025

• [Mobile] iOS yolov8n-pose model throws 'onnxruntime Code=2 "Invalid Feed Input Name:input"'

  #19776 closed Jun 7, 2025

• [Documentation Request]

  #19819 closed Jun 7, 2025

• [Build] Nuget doesn't contain runtimes and build

  #19800 closed Jun 7, 2025

• [Training] Differences results in Top K 2D Nearest Neighbor Results Between ONNX Runtime and PyTorch

  #19759 closed Jun 7, 2025

• [Web] Yolov8 WebGl backend - Uncaught (in promise) Error: resize (packed) does not support mode: 'nearest'

  #19779 closed Jun 7, 2025

• [Web] How should I get wasm file?

  #19829 closed Jun 7, 2025

• [Web] no available backend found. ERR: [wasm] RuntimeError: Aborted(CompileError: WebAssembly.instantiate(): expected magic word 00 61 73 6d, found 3c 21 44 4f @+0). Build with -sASSERTIONS for more info., [cpu] Error: previous call to 'initializeWebAssembly()' failed., [xnnpack] Error: previous call to 'initializeWebAssembly()' failed.

  #19831 closed Jun 7, 2025

• How can I implement custom operators in python?

  #19820 closed Jun 7, 2025

• [Web] The nested component seems to be unable to obtain the correct path to the wasm file.

  #19864 closed Jun 7, 2025

• [Mobile] How custom threading callbacks in java. I want bind the thread to the big core of cpu.

  #19860 closed Jun 7, 2025

• [Performance] MultiHeadAttention CPU kernel slower than unfused

  #19924 closed Jun 7, 2025

• [Performance] CoreML not being used to it's fullest capacity - custom transformer

  #19887 closed Jun 7, 2025

• [Web] When performing inference with ONNX Runtime in C++, using the libonnxruntime_webassembly.a static library, but encountering an error during the session.run() call in an HTML5 environment, the error message is: "Uncaught (in promise) 50699072".

  #19913 closed Jun 7, 2025

• Is there any way to retrieve Quantization type and Quantization parameters using onnxruntime ?

  #19916 closed Jun 7, 2025

• [Performance] createSession() slow on release 1.15 and 1.17.1 as compare to 1.14

  #19902 closed Jun 7, 2025

• Improve Inference Performance on GPU [Python]

  #19930 closed Jun 7, 2025

• Inference Layer by Layer or feature extraction on Onnx Runtime

  #19954 closed Jun 7, 2025

• [Training] [Windows]

  #19965 closed Jun 7, 2025

• [Documentation] Thread spinning behavior with Python

  #19968 closed Jun 7, 2025

• [Web] `InferenceSession.create` returns number as exception with model generated by `torch.multinomial`

  #19961 closed Jun 7, 2025

• Failing to run Dockerfile of ONNXRuntime with TensorRT integration[Build]

  #19978 closed Jun 7, 2025

• RUNTIME_EXCEPTION : Non-zero status code returned while running Reshape node

  #20026 closed Jun 7, 2025

• Some problems about the onnx-tensorrt source code.

  #20029 closed Jun 7, 2025

• Shared Arena Env Allocator Usage Across Modules

  #20027 closed Jun 7, 2025

• [Mobile]

  #20044 closed Jun 7, 2025

• [Build] Trying to use TensorrtExecutionProvider. Model not loading

  #20032 closed Jun 7, 2025

• [Performance] INT8 quantized model run slower than FP32 model

  #20052 closed Jun 7, 2025

• Failed to allocated memory for requested buffer of size X

  #20038 closed Jun 7, 2025

• "trt_cuda_graph_enable" bug in tensorrt.

  #20050 closed Jun 7, 2025

• logging.h:309 onnxruntime::logging::LoggingManager::DefaultLogger Attempt to use DefaultLogger but none has been registered.

  #20064 closed Jun 7, 2025

• onnxruntime-directml cause TDR

  #20094 closed Jun 7, 2025

• onnxruntime in windows.ai.machinelearning noisy logger can't be disabled

  #20063 closed Jun 7, 2025

• How to use int8 model with openvino

  #20087 closed Jun 7, 2025

• [Build] Onxx runtime library for Android

  #20076 closed Jun 7, 2025

• [Performance] >16s (!!!) per inference with YOLO-NAS-M based model, CPU EP.

  #20112 closed Jun 7, 2025

• `get_device()` in onnxruntime-gpu not working properly

  #20122 closed Jun 7, 2025

• [Build] "error: parameter ‘thread_pool’ set but not used"

  #20144 closed Jun 7, 2025

• GPU Inference Time Is More Than CPU Inference Time.

  #20142 closed Jun 7, 2025

• [Build] JetPack 5.1.2, gpu for c++

  #20140 closed Jun 7, 2025

• [Build] libonnxruntime.so Not found after successful Android build

  #20185 closed Jun 7, 2025

• An issue that occurred while optimizing Babelscape/mrebel-large

  #20159 closed Jun 7, 2025

• [Performance] batch inference slower than frame inference

  #20173 closed Jun 7, 2025

• [Build] Undefined reference issue for aarch64 build

  #20188 closed Jun 7, 2025

• api:c Exception while running ORT model. "NULL input supplied for input <outputTensorName>"

  #20192 closed Jun 7, 2025

• Intel OneDNN

  #20208 closed Jun 7, 2025

• [Performance] Non-zero status code and MIOPEN failure when running inference on AMD GPUs.

  #20203 closed Jun 7, 2025

• [Performance] Python infer and C++ are different for audio process

  #20227 closed Jun 7, 2025

• [Javascript ] inferenceSession on WebGL

  #20224 closed Jun 7, 2025

• ONNX Runtime and PyTorch results are different

  #20219 closed Jun 7, 2025

• [Performance] 推理onnx模型时，显存占用异常

  #20228 closed Jun 7, 2025

• [Build] Pre-installed dependencies

  #20310 closed Jun 7, 2025

• [Training] SGD optimizer performs Pageable DtoH Memcpy

  #20232 closed Jun 7, 2025

• [Performance] yolov5-segmentation.Onnx Static Quantization

  #20239 closed Jun 7, 2025

• [Performance] Segfault on destruction of OpenVINO Session

  #20311 closed Jun 7, 2025

• [Performance] GPU time exceeding CPU time

  #20361 closed Jun 7, 2025

• Runtime error with an empty input tensor in 1.17 not in 1.15

  #20375 closed Jun 7, 2025

• onnxruntime 在C++上如何实现fp16的推理 yolov5模型

  #20395 closed Jun 7, 2025

• How to convert fp32 to fp16 from an image

  #20394 closed Jun 7, 2025

• [Performance] Disable stream synchronization at the end of CUDA Graphs replay

  #20392 closed Jun 7, 2025

• onnxruntime + openvino need double memory compared with openvino-only

  #20467 closed Jun 7, 2025

• Phi-3 can't deal with Japanese. How can I solve this issue?

  #20448 closed Jun 7, 2025

• [Build] Shared lib testing for all built EPs

  #20468 closed Jun 7, 2025

• [Web] invalid data location: undefined

  #20431 closed Jun 7, 2025

• Dnnl Execution Provider GetMemoryAndReshape function issues with Status Message: not a valid reshape, inconsistent dim product.

  #20426 closed Jun 7, 2025

• output is different between onnx and model

  #20478 closed Jun 7, 2025

• use multi ort session in one process, can not improve throughput

  #20494 closed Jun 7, 2025

• [Performance] Session Creation Time is excessively slow on DirectML devices and Universal Windows projects

  #20502 closed Jun 7, 2025

• NVIDIA Jetson aarch64 official PyPi binaries for onnxruntime-gpu

  #20503 closed Jun 7, 2025

• CMake install and Release Zip have folder structures that are not consistent

  #20510 closed Jun 7, 2025

• Onnx model throws an exception in 1.17.3 but works in 1.16.x

  #20514 closed Jun 7, 2025

• how to release gpu memory after session.run

  #20517 closed Jun 7, 2025

• [Build] Remove large files from repository

  #20526 closed Jun 7, 2025

• [Mobile] [iOS] Declare NSPrivacyAccessedAPICategorySystemBootTime API Usage in onnxruntime-objc

  #20519 closed Jun 7, 2025

• OPT6.7b ONNX model not giving accurate results on CPU.

  #20543 closed Jun 7, 2025

• terminate called after throwing an instance of 'Ort::Exception' what(): Invalid input name: ��veSU

  #20568 closed Jun 7, 2025

• LayerNormalization on Dnnl execution provider causing side-effect of input

  #20623 closed Jun 7, 2025

• Microsoft.ML.OnnxRuntime.DirectML causes VS 2022 project to have empty project properties list

  #20574 closed Jun 7, 2025

• [Performance] Inference takes longer when session.Run() is being ran on different threads and each thread has its own session

  #20599 closed Jun 7, 2025

• [Documentation Request]

  #20592 closed Jun 7, 2025

• ONNX Runtime doesn't support the graph optimization of vision-encoder-decoder yet

  #20701 closed Jun 7, 2025

• [How to release cpu memory after session Run]

  #20640 closed Jun 7, 2025

• Undefined behaviour in OneHot operator

  #20659 closed Jun 7, 2025

• Please document how to build with new execution provider [Documentation Request]

  #20654 closed Jun 7, 2025

• [Web] Trouble loading > 2GB ONNX model (those with separate blocks)

  #20645 closed Jun 7, 2025

• In a Windows 32-bit system, what is the highest version that can be used?

  #20660 closed Jun 7, 2025

• [Mobile] Broadcasting Error in Sub Node with ONNX Runtime Version 1.17.3: Incompatibility with Dimension Broadcasting Rules

  #20672 closed Jun 7, 2025

• [Performance] Quadratic behaviour in list operations with SequenceInsert in onnx

  #20696 closed Jun 7, 2025

• [Build] [CANN] Failed to build CANN provider with training and Python bindings

  #20697 closed Jun 7, 2025

• [Web] The YOLOv8 segmentation model with batching option is not runing on the GPU ?

  #20710 closed Jun 7, 2025

• cannot resolve operator 'HardSwish' with opsets: ai.onnx v9 [Web]

  #20704 closed Jun 7, 2025

• cuda's FusedConv is not support Sigmod

  #20708 closed Jun 7, 2025

• [Build] 0.18.0 release breaks Hummingbird build pipeline

  #20715 closed Jun 7, 2025

• [Training] IR version incompatibility in artifact generation for on-device training

  #20726 closed Jun 7, 2025

• [Documentation Request] Required cuDNN version for OnnxRuntime 1.18

  #20784 closed Jun 7, 2025

• [Training] The gradient builder has not been registered for node with op type MatMulNBits

  #20781 closed Jun 7, 2025

• [Build] --external_graph_transformer_path doesn't. --test_external_transformer_example removed from build.py?

  #20751 closed Jun 7, 2025

• [Web] `executionProviders` chain for `webnn` fallback does not work on init error

  #20729 closed Jun 7, 2025

• Get error while using Dml EP

  #20742 closed Jun 7, 2025

• [Build] passing --arm64 to ci_build/build.py has error in arm64 host

  #20814 closed Jun 7, 2025

• Result of Pad evaluated by onnxruntime is inconsistent with onnx ReferenceEvaluator and onnx doctumentation

  #20801 closed Jun 7, 2025

• How can I debug a reproducible error?

  #20792 closed Jun 7, 2025

• Please Add webpack and typescript configuration

  #20822 closed Jun 7, 2025

• [Web] I can’t use onnruntime-web to load a onnx model in a react web

  #20846 closed Jun 7, 2025

• Error in quantize vicuna-7b model from fp16 to int8

  #20867 closed Jun 7, 2025

• Non-zero status code returned while running Add node. Name:'Add_221'

  #20861 closed Jun 7, 2025

• [Documentation] The documentation for early versions is missing

  #20850 closed Jun 7, 2025

• Index put loop model regression with ort==1.18

  #20855 closed Jun 7, 2025

• Gemm fp8 run error

  #20869 closed Jun 7, 2025

• [Documentation] How to run this model on android mobile platform

  #20937 closed Jun 7, 2025

• XGBoost converter output size shape warning and multiclass prediction error

  #20908 closed Jun 7, 2025

• [Build] CUDA Illegal Memory Access error when using a custom Triton kernel

  #20885 closed Jun 7, 2025

• ONNXruntime version 1.18.0

  #20877 closed Jun 7, 2025

• [Training] Compiling ONNX Runtime for MIPS32 Linux for On-Device Training Capabilities

  #20884 closed Jun 7, 2025

• [JAVA] Ability to construct a Tensor from a GPU memory pointer

  #20966 closed Jun 7, 2025

• [Build] Unable to build onnxruntime from source (with oneDNN EP)

  #20971 closed Jun 7, 2025

• [Performance] Is my script set to get optimal performance of onnxruntime?

  #20945 closed Jun 7, 2025

• [Web] LinkError when using custom built WASM artifacts

  #20970 closed Jun 7, 2025

• [Build] moduleNotfoundError: no module named 'onnxruntime.training' & 'No matching distribution found for onnxruntime-training'

  #20985 closed Jun 7, 2025

• DML incorrect results, probably Split node

  #21007 closed Jun 7, 2025

• Non-zero status code returned while running ConvTranspose node.

  #21034 closed Jun 7, 2025

• [Mobile] onnxruntime-objc crash

  #21026 closed Jun 7, 2025

• [Mobile] React-native OnnxruntimeJSIHelper install segfaults when registering functions

  #21003 closed Jun 7, 2025

• Flan-T5 small converted model produces wrong result with batch size > 1 and long senetences

  #21053 closed Jun 7, 2025

• Segmentation fault during inference

  #21069 closed Jun 7, 2025

• [Bug] W16A16 quantization, qdq_error is empty, but W8A8 is normal

  #21089 closed Jun 7, 2025

• [Bug] fp8 quantization, weight_error is empty, but qdq_err is normal

  #21113 closed Jun 7, 2025

• [Training] Support for RKNPU Execution Provider on RK3562 Platform and On-Device Training Capabilities

  #21060 closed Jun 7, 2025

• [Performance] Running YOLOv8-seg.onnx with Dynamic Batch Size on GPU

  #21103 closed Jun 7, 2025

• cannot resolve operator 'Cast' with opsets: ai.onnx v17 [Web]

  #21115 closed Jun 7, 2025

• Quantized ONNX Model Still Has Float32 Input/Output Tensors

  #21138 closed Jun 7, 2025

• [Jvm] Native crash during createSession: std::bad_cast

  #21147 closed Jun 7, 2025

• [Performance] CUDA kernel not found in registries for Op type: ScatterND

  #21148 closed Jun 7, 2025

• [Training] Onnxruntime-training 1.18.0 for windows not available

  #21149 closed Jun 7, 2025

• CoreML EP inference result is improperly scaled

  #21170 closed Jun 7, 2025

• Can onnxruntime.quantization.quantize_dynamic() work with onnx-trt?

  #21169 closed Jun 7, 2025

• [Documentation] Setup the CUDA Environment is not detailed enough

  #21197 closed Jun 7, 2025

• [Performance] Failed to run Whisper inference after optimization with Dml EP

  #21156 closed Jun 7, 2025

• [C#] Enable copying of GPU OrtValue to CPU

  #21244 closed Jun 7, 2025

• Inference result different between cuda and cpu

  #21220 closed Jun 7, 2025

• DML EP takes very long time and not exit compiling

  #21255 closed Jun 7, 2025

• CUDA_PATH is set but CUDA wasnt able to be loaded

  #21272 closed Jun 7, 2025

• [Performance] How does onnxruntime run in parallel mode?

  #21259 closed Jun 7, 2025

• [Web] Inconsistent results between running onnx model through python and with onnxruntime-web

  #21275 closed Jun 7, 2025

• [Mobile] Android/Kotlin/JAVA Multi Threading for Multi models in android app

  #21289 closed Jun 7, 2025

• Tried to specify the thread pool when creating an OrtEnvironment, but one already exists

  #21290 closed Jun 7, 2025

• [Build] ModuleNotFoundError: No module named 'onnxruntime.capi'

  #21317 closed Jun 7, 2025

• [Crash] Crash while loading AlibabaNLP/gte-base ONNX model

  #21322 closed Jun 7, 2025

• [Build] How can I quantize the llama3 model activation to int4 ?

  #21334 closed Jun 7, 2025

• header files path not recognized or unable to read header file

  #21338 closed Jun 7, 2025

• [BUG] CANN: onnxruntime.capi.onnxruntime_pybind11_state.Fail: [ONNXRuntimeError]

  #21352 closed Jun 7, 2025

• [Build] AllocatorTest.CUDAAllocatorFallbackTest failed

  #21343 closed Jun 7, 2025

• [Performance] Multiple Sessions on Same GPU is very slow

  #21365 closed Jun 7, 2025

• [Web] where is the demo of object detection on web

  #21403 closed Jun 7, 2025

• [Models larger than 2GB :(] Specify mid-graph.output after initializing InferenceSession

  #21367 closed Jun 7, 2025

• How to do multithreaded infer with onnxruntime

  #21419 closed Jun 7, 2025

• onnxruntime.InferenceSession.run sometimes get stuck, sometimes not

  #21418 closed Jun 7, 2025

• FAIL : LoadLibrary failed with error 126 "" when trying to load "C:\code\Blueprint.Net.Server\bin\Debug\net8.0-windows10.0.22621.0\runtimes\win-x64\native\onnxruntime_providers_cuda.dll" ”

  #21436 closed Jun 7, 2025

• [Build] ADD_LIBRARY cannot create target "memory" because another target with the same name already exists between xnnpack and absl

  #21441 closed Jun 7, 2025

• Activate thread pool will cause crash.

  #21461 closed Jun 7, 2025

• Quantization failed! The onnxruntime.quantization.quantize_dynamic seems didn't convert to the qint8 .onnx file successfully

  #21440 closed Jun 7, 2025

• quant_pre_process failed on NonMaxSuppression

  #21476 closed Jun 7, 2025

• [Performance] The 16-bit quantization QDQ model cannot be accelerated by CUDA

  #21478 closed Jun 7, 2025

• failing to find trt_timing_cache_path

  #21484 closed Jun 7, 2025

• Android build: Execution failed for task ':app:mergeExtDexDebug'.

  #21494 closed Jun 7, 2025

• [Performance] DequantizeLinear, pad and QuantizeLinear operation is not fused

  #21496 closed Jun 7, 2025

• [CUDA, DML] MatMul does not properly handle matrices with inner dim == 0

  #21483 closed Jun 7, 2025

• [Web] Quantized model decreases in size, but takes same amount of inference time as non-quantized model

  #21535 closed Jun 7, 2025

• MLAS failing with "Could not find an implementation for QLinearMatMul"

  #21531 closed Jun 7, 2025

• Model saved by ORT as external data format will not be aligned for mapfile support

  #21524 closed Jun 7, 2025

• [Build] detect nothing.i use opencv4.9 onnxruntime 1.16.1, it detect nothing

  #21513 closed Jun 7, 2025

• [Build] reduce the size of the shared library

  #21544 closed Jun 7, 2025

• CUDA_PATH is set but CUDA wasnt able to be loaded

  #21527 closed Jun 7, 2025

• Issue running a model in ONNXruntime

  #21571 closed Jun 7, 2025

• Onnxruntime LoadLibrary failed with error 126

  #21501 closed Jun 7, 2025

• Java GPU dependency of ONNX Runtime version 1.18 only support CUDA 12?

  #21651 closed Jun 7, 2025

• [Performance]

  #21654 closed Jun 7, 2025

• [Performance] SetIntraOpNumThreads not take effect

  #21700 closed Jun 7, 2025

• [Performance] Inference time discrepancy when using TorchScript vs ONNX exported model

  #21689 closed Jun 7, 2025

• [Performance] pytorch quantize_qat model export to onnx, insert a transpose layer befor and after the conv layers

  #21702 closed Jun 7, 2025

• onnxruntime.capi.onnxruntime_pybind11_state.Fail: [ONNXRuntimeError] : 1 : FAIL : CopyTensorAsync is not implemented

  #21690 closed Jun 7, 2025

• Can anyone successfully use onnx and yolo5?

  #21772 closed Jun 7, 2025

• The data I output using YOLO5 is incorrect, why is that? Has anyone succeeded?

  #21773 closed Jun 7, 2025

• [Build] Access Violation error on using 'new InferenceSession(filepath)' in 1.18.1

  #21707 closed Jun 7, 2025

• ORTModelForSeq2SeqLM.from_pretrained can not use provider=['CUDAExecutionProvider'，'CPUExecutionProvider']

  #21733 closed Jun 7, 2025

• Question

  #21719 closed Jun 7, 2025

• DirectML failed with invalid command

  #21805 closed Jun 7, 2025

• [Documentation] Is oneDNN onnx support windows 11？

  #21856 closed Jun 7, 2025

• [Performance] Why does genai run 2x as fast as vanilla managed onnxruntime?

  #21847 closed Jun 7, 2025

• AppendExecutionProvider_DML error

  #21787 closed Jun 7, 2025

• Breaks RootNamespace

  #21783 closed Jun 7, 2025

• Segfault when using IO binding to CUDA tensor with CPU execution provider

  #21865 closed Jun 7, 2025

• [Build] Openvino EP Build Error Linux

  #21883 closed Jun 7, 2025

• [Performance] OrtValue created form OrtValue.ortvalue_from_numpy() has the same ptr same as the previous one, it cause the value of the previous variable to be modified.

  #21888 closed Jun 7, 2025

• Different outputs when run on CPU vs GPU (CUDA)

  #21859 closed Jun 7, 2025

• [Documentation] where is api SessionOptionsAppendExecutionProvider_oneDNN?

  #21858 closed Jun 7, 2025

• run onnxruntime-directml failed due to ops

  #21875 closed Jun 7, 2025

• [Performance]

  #21932 closed Jun 7, 2025

• Corrupted value for model outputs that are also model inputs

  #21922 closed Jun 7, 2025

• [Performance] High thread contention in BFCArena

  #21916 closed Jun 7, 2025

• NMS Operator Output Different From Torchvision Implementation

  #21898 closed Jun 7, 2025

• Get wrong results occasionally when run yolov9 onnx model using OpenVINOExecutionProvider

  #21895 closed Jun 7, 2025

• [CUDA][Performance] Inference time greatly variates during session run

  #21966 closed Jun 7, 2025

• CreateSessionFromArray doesn't work

  #21946 closed Jun 7, 2025

• Use AppendExecutionProvider_Dnnl api to add onednn EP，No success.

  #21958 closed Jun 7, 2025

• Why C++ cannot modify the enable_mem_reuse option in Ort::SessionOptions...

  #21942 closed Jun 7, 2025

• CPU LayerNormalization Produces NaN for Inputs with Small Variances

  #21935 closed Jun 7, 2025

• [Build] compiling the WASM in Firefox takes ~10 minutes and 4GB of ram

  #21978 closed Jun 7, 2025

• [Performance] Increasing Memory Usage during INT8 Quantization with ONNX Runtime tools

  #21979 closed Jun 7, 2025

• [Web] __turbopack_resolve_absolute_path__ is not a function

  #22005 closed Jun 7, 2025

• [Inference & Training] My Onnxruntime isnt detecting cuda even after all paths are perfectly given with compatible softwares

  #22016 closed Jun 7, 2025

• Different outputs in Python and C++

  #22030 closed Jun 7, 2025

• [Documentation] Prebuilt ORT Package does not include required QNN dependency

  #22078 closed Jun 7, 2025

• [Training]

  #22040 closed Jun 7, 2025

• Treatment of optional inputs to nodes when empty

  #22035 closed Jun 7, 2025

• [Web] model cannot load after 1.19

  #22071 closed Jun 7, 2025

• [Documentation] Execution provider strings

  #22101 closed Jun 7, 2025

• [Documentation] Difficulty using trt_int8_use_native_calibration_table option in ONNX Runtime

  #22059 closed Jun 7, 2025

• [Documentation] Cuda version for default onnxruntime-gpu is wrong

  #22178 closed Jun 7, 2025

• Concat layer gives warnings in TensorRT (only)

  #22163 closed Jun 7, 2025

• CenterCropPad inside of SequenceMap subgraph produces INVALID_GRAPH

  #22100 closed Jun 7, 2025

• [Performance] How to reduce gpu memory consumption ？

  #22130 closed Jun 7, 2025

• topk assumes GridDim::maxThreadsPerBlock >= 256

  #22079 closed Jun 7, 2025

• How to determine which provider current inference is using?

  #22243 closed Jun 7, 2025

• TensorRTExecutionProvider error during session initialization

  #22199 closed Jun 7, 2025

• [Build] Onnxruntime with OpenVino error: no match for ‘operator=’

  #22175 closed Jun 7, 2025

• [Training] Cannot generate ONNX graph for training

  #22188 closed Jun 7, 2025

• CUDA Error cudaErrorUnsupportedPtxVersion with NVIDIA H800 (Compute Capability 9.0) on ONNXRuntime GPU 1.19.2

  #22212 closed Jun 7, 2025

• [Web] WebNN optimizations silently failing when using wrong argument names in freeDimensionOverrides

  #22300 closed Jun 7, 2025

• ORT would be crashed while loading the specific INT4 model

  #22284 closed Jun 7, 2025

• [Performance] Very slow load of ONNX model in Windows

  #22219 closed Jun 7, 2025

• [Build] [On-device-training] Running onnxruntime-training (CPU) Python Demo on Debian 12 aarch64 Environment

  #22270 closed Jun 7, 2025

• Simple FP8 GEMM is not runnable in CPU EP

  #22269 closed Jun 7, 2025

• Onnxruntime error during inference with operator (CumSum)

  #22326 closed Jun 7, 2025

• [Build] Build onnxruntime for tensorrt failed on rtx 4060 gpu

  #22382 closed Jun 7, 2025

• CUDAExecutionProvider doesn't seem to be used during inference of transformers exported model to ONNX runtime GPU

  #22325 closed Jun 7, 2025

• [Performance] Python inference runs faster than C++

  #22328 closed Jun 7, 2025

• Getting error when trying to use OpenVINOExecution Provider

  #22405 closed Jun 7, 2025

• [Training] The gradient builder has not been registered: Atan

  #22407 closed Jun 7, 2025

• [Mobile]

  #22439 closed Jun 7, 2025

• [Web] Can't create a session

  #22484 closed Jun 7, 2025

• [Build] The Python module fails to build: error: package directory 'onnxruntime/backend' does not exist

  #22501 closed Jun 7, 2025

• Accessing Resize op (ResizeNearestNeighbor) in QNN

  #22549 closed Jun 7, 2025

• [Mobile] iOS - ZipMap output cannot be read

  #22505 closed Jun 7, 2025

• [Build] CMake should always respect "use_preinstalled_eigen" and "eigen_path" even if "find_package(Eigen3)" succeeds

  #22515 closed Jun 7, 2025

• [Mobile] null pointer dereference

  #22538 closed Jun 7, 2025

• [Web] custom wasm model location path for inference

  #22504 closed Jun 7, 2025

• Using directML to inference accelerate onnxruntime, a crash occurred.

  #22514 closed Jun 7, 2025

• How to include libnvinfer_plugin in trt_extra_plugin_lib_paths on Windows? Im using python.

  #22521 closed Jun 7, 2025

• [BUG] Cuda Softmax op when axis != rank - 1

  #22554 closed Jun 7, 2025

• [Training] Any way to profile the training of the model?

  #22614 closed Jun 7, 2025

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

  #22559 closed Jun 7, 2025

• [Build] ONNX Runtime DLLs are not built with spectre mitigation

  #22601 closed Jun 7, 2025

• AttributeError: FLOAT8E4M3FN during quantization

  #22640 closed Jun 7, 2025

• [Mobile] Maui with ONNX Runtime does not compile for IOS

  #22661 closed Jun 7, 2025

• [Documentation]

  #22620 closed Jun 7, 2025

• [QNN Execution Provider] Non-zero status code returned while running QNN_9132504664900242513_1 node. Error code: 1100

  #22645 closed Jun 7, 2025

• failure using ONNX Runtime QNNExecutionProvider on Snapdragon® X Elite NPU

  #22617 closed Jun 7, 2025

• [Web] GroupQueryAttention test failure on WebAssembly

  #22649 closed Jun 7, 2025

• [Performance] Multiple instances of the same model are slower

  #22778 closed Jun 7, 2025

• can onnxruntime-node get metaData from onnx model

  #22682 closed Jun 7, 2025

• [Mobile] Why are there so many input_type_shape and output_type_shape listed in this NNAPI node?

  #22684 closed Jun 7, 2025

• [C#] The type initializer for 'Microsoft.ML.OnnxRuntime.NativeMethods' threw an exception.

  #22683 closed Jun 7, 2025

• symbolic_shape_infer.py script not working for some models

  #22662 closed Jun 7, 2025

• Failed to allocate memory

  #22702 closed Jun 7, 2025

• [Build] `libXNNPACK.a` not built when building with `vcpkg` and `onnxruntime_BUILD_UNIT_TESTS=off`

  #22789 closed Jun 7, 2025

• Running the FP16 quantizeLinear operator produced an incorrect output

  #22741 closed Jun 7, 2025

• [Web] Support iOS devices

  #22776 closed Jun 7, 2025

• [Performance] the root cause of onnxruntime-genai's being much faster than onnxruntime?

  #22805 closed Jun 7, 2025

• [Performance] in windows v1.19.2 - non-deterministic but in linux deterministic

  #22818 closed Jun 7, 2025

• Altough using onnxruntime-19.0 but still can't use cudnn 9.x

  #22809 closed Jun 7, 2025

• the input format and output format not match onnx model

  #22797 closed Jun 7, 2025

• CUDA memory increasing and process freeze [Performance]

  #22872 closed Jun 7, 2025

• how does onnxruntime kernel execute，when is to parallelize and when to serialize.

  #22791 closed Jun 7, 2025

• [Performance] how to set the threads when using TRT EP

  #22913 closed Jun 7, 2025

• Error when running inference in C++:onnxruntime::SequentialExecutor::Execute] Non-zero status code returned while running Sigmoid node.

  #22836 closed Jun 7, 2025

• [Build] Issues with Multithreading in the New Versions of onnxruntime-directml

  #22867 closed Jun 7, 2025

• [Build] Build Error

  #22882 closed Jun 7, 2025

• [DML EP] ORT would crash after deleting one of the models and then doing an inference

  #22948 closed Jun 7, 2025

• INVALID_ARGUMENT : unsupported conv activation mode "LeakyRelu"

  #22947 closed Jun 7, 2025

• [ONNXRuntimeError] How to compare the precision layer by layer with engine(for example, tensorRT) if I have a custom operator in my onnx (and a corresponding plugin in TensorRT)?

  #22938 closed Jun 7, 2025

• [Performance] Observing higher memory spikes in C++ when running multiple Inference `Run()` executions on CPU

  #22920 closed Jun 7, 2025

• Inference Output CPU vs CUDA not the same [1.19.2]

  #22866 closed Jun 7, 2025

• Error While Creating ONNX Session with CUDA Execution Provider

  #22980 closed Jun 7, 2025

• [Mobile] Not able to create session due to some package in armeabi-v7a

  #22965 closed Jun 7, 2025

• [Documentation] Can CPUEP infer multithread with one session?

  #22967 closed Jun 7, 2025

• Memory error while IO binding output to GPU - CUDA

  #22979 closed Jun 7, 2025

• [Mobile] need vulkan ep support

  #22973 closed Jun 7, 2025

• Using ONNX runtime with Execution Providers in Java

  #22996 closed Jun 7, 2025

• Unmanage memory growth when calling Onnx from C# with C++ API

  #22992 closed Jun 7, 2025

• [Crash] onnxruntime.quantization.quantize_static the segment-anything model crashed with no error reported

  #23005 closed Jun 7, 2025

• How to use OpenVINO as EP for inference modeling on NPUs

  #22990 closed Jun 7, 2025

• [Build] Expose internal header files

  #23025 closed Jun 7, 2025

• [Graph optimization] INVALID_GRAPH error returned by InferenceSession with offline optimization

  #23022 closed Jun 7, 2025

• [Build] Unable to Compile ONNX Runtime 1.20.1 with ARMNN Provider on ARM Cortex A78

  #23014 closed Jun 7, 2025

• [Build] Question -- is it possible to override/replace a core operator?

  #23007 closed Jun 7, 2025

• How to get the hardware capability?

  #23059 closed Jun 7, 2025

• [Build] [Linux] [.NET] Microsoft.ML.OnnxRuntime.Gpu keep publishing unnecessary DLL large files for linux-x64

  #23077 closed Jun 7, 2025

• [Build] Can't run dml with nodejs after building

  #23027 closed Jun 7, 2025

• [Mobile] need tflite ep

  #23033 closed Jun 7, 2025

• ExportModelForInferencing should use External Data if Necessary

  #23041 closed Jun 7, 2025

• [Documentation] Is there an execution provider in ONNX Runtime that supports Mali GPUs?

  #23089 closed Jun 7, 2025

• Conflict constraints checking/description about PoolAttributes

  #23088 closed Jun 7, 2025

• How to handle dynamic output dimensions in RunAsync with OrtValue?

  #23081 closed Jun 7, 2025

• undefined symbol: RegisterCustomOps

  #23035 closed Jun 7, 2025

• [BUG][CUDAProvider] No attribute with name:'activation'is defined

  #23119 closed Jun 7, 2025

• [CUDAProvider] Graph Optimization output an invalid model

  #23118 closed Jun 7, 2025

• Regarding the issue of starting services in multithreading

  #23094 closed Jun 7, 2025

• RuntimeError: Assertion `false` failed: No Adapter From Version $20 for GridSample

  #23112 closed Jun 7, 2025

• On linux env, onnx costs high anon memory, but not been released

  #23117 closed Jun 7, 2025

• FuseReluClip Unexpected data type for Clip 'min' input of 11

  #23116 closed Jun 7, 2025

• [Bug] InvalidArgument Error After Optimizing Model with ONNX Runtime

  #23138 closed Jun 7, 2025

• [Performance] Can onnxruntime delegate the entire process to another framework when using other providers?

  #23132 closed Jun 7, 2025

• MultiHeadAttention op shall return attention probabilities

  #23124 closed Jun 7, 2025

• [Performance] CreateSession takes very long time to load .onnx file when working with FileFuzzer tool

  #23129 closed Jun 7, 2025

• How generate a contxt model dump from onnx runtime? (C++)

  #23153 closed Jun 7, 2025

• [Web] Inference session cannot be created with `SharedArrayBuffer`, only `ArrayBuffer` is accepted

  #23148 closed Jun 7, 2025

• [Build] Warnings Treated as Errors in Windows Build for x86 platform on ORT Version 1.10.0

  #23160 closed Jun 7, 2025

• Inconsistent Results for Output v1_0 After ONNX Runtime Optimization (Flaky Test)

  #23143 closed Jun 7, 2025

• [Build] ASP.NET web forms

  #23168 closed Jun 7, 2025

• [Build] build error for windows

  #23166 closed Jun 7, 2025

• CMake error on Windows with -use_xnnpack

  #23174 closed Jun 7, 2025

• [Training] Using C++ to train ONNX models

  #23170 closed Jun 7, 2025

• [Mobile] How to use GPU acceleration on Android

  #23194 closed Jun 7, 2025

• [Mobile] google say not support nnapi anymore

  #23206 closed Jun 7, 2025

• Java library throws error when using CUDA: LoadLibrary failed with error 126 "" when trying to load "C:\Users\xx\AppData\Local\Temp\onnxruntime-java5278075328315693241\onnxruntime_providers_cuda.dll"

  #23208 closed Jun 7, 2025

• [Build] TypeInferenceError when quantize an onnx model with custom operator

  #23191 closed Jun 7, 2025

• ONNXRuntime Optimization Causes Output Discrepancy in Specific Model Structure (Output Y)

  #23209 closed Jun 7, 2025

• Error merging shape info for output

  #23187 closed Jun 7, 2025

• RUNTIME_EXCEPTION : Non-zero status code returned while running If node.

  #23213 closed Jun 7, 2025

• ONNXRuntime produces inconsistent results for specific output v10_0 (flaky test behavior)

  #23212 closed Jun 7, 2025

• Inconsistent outputs when running onnx and pytorch (stft and istft)

  #23219 closed Jun 7, 2025

• onnxruntime-web dependency on document breaks chrome serviceworker

  #23214 closed Jun 7, 2025

• [Documentation] Typo in the I/O Binding doc

  #23229 closed Jun 7, 2025

• [C#] ML.NET: ArgumentOutOfRangeException thrown in PredictionEngine.Predict

  #23230 closed Jun 7, 2025

• [Documentation]

  #23238 closed Jun 7, 2025

• CoreML failed: Unable to get shape for output

  #23262 closed Jun 7, 2025

• C++ wrapper loses error information?

  #23245 closed Jun 7, 2025

• [Web] FP8 is not supported

  #23241 closed Jun 7, 2025

• [Performance] model inference in onnxruntime is toooooo slow

  #23282 closed Jun 7, 2025

• Using genai_config.json to configure dml options

  #23247 closed Jun 7, 2025

• [js/webgpu] ConvTranspose1D slower on Webgpu than Wasm

  #23273 closed Jun 7, 2025

• Why is the console messed up when using onnxruntime.InferenceSession?

  #23270 closed Jun 7, 2025

• Mismatch between Matmul op in FLOAT16 and pytorch Linear op.

  #23272 closed Jun 7, 2025

• 'Microsoft.ML.OnnxRuntime.NativeMethods' threw an exception

  #23300 closed Jun 7, 2025

• The trt_engine_decryption_lib_path environment variable renders encryption worthless

  #23290 closed Jun 7, 2025

• Encryption does not work with trt_dump_ep_context_model

  #23289 closed Jun 7, 2025

• onnxruntime-python on AWS

  #23291 closed Jun 7, 2025

• OnnxRuntime and Numerics.Tensors version numbers out-of-date

  #23295 closed Jun 7, 2025

• How to implement a custom operator that support multiple compute device (CPU, CUDA)?

  #23317 closed Jun 7, 2025

• Linux Failed Build - std::piecewise_construct’ causes a section type conflict

  #23345 closed Jun 7, 2025

• [Build] libonnxruntime_providers_shared.so statically linked?

  #23355 closed Jun 7, 2025

• [Performance]Why is loading an ONNX model taking so long?

  #23338 closed Jun 7, 2025

• [Build] Not able to build ONNX Runtime Nuget package on Windows

  #23321 closed Jun 7, 2025

• C# Run Program on NPU (OnnxRuntime + DirectML + NPU)?

  #23375 closed Jun 7, 2025

• [Accuracy] MSclap model accuracy issue (CPU vs QNN EP (NPU) )

  #23394 closed Jun 7, 2025

• [Performance] kokoro onnx performance issues

  #23384 closed Jun 7, 2025

• [ROCm] CK Datatype Adaptor - BFloat16

  #23390 closed Jun 7, 2025

• [Build] Non-zero status code

  #23497 closed Jun 7, 2025

• Model having scatterND layer giving different result every time with same input

  #23396 closed Jun 7, 2025

• Nuget package Microsoft.ML.OnnxRuntime.Gpu version >= 1.17.0 not working

  #23462 closed Jun 7, 2025

• Not able to load QNN Context Binary Model

  #23431 closed Jun 7, 2025

• [Documentation] CudaContext::AllocDeferredCpuMem

  #23485 closed Jun 7, 2025

• [Performance] Preload model before inference

  #23513 closed Jun 7, 2025

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

  #23556 closed Jun 7, 2025

• [Performance] Distinct 2x inference time difference between AMD and Intel processors

  #23587 closed Jun 7, 2025

• [Build] protocol buffer compiler error MSB8066

  #23529 closed Jun 7, 2025

• symbolic_shape_infer.py cannot infer torch.nn.normalize

  #23516 closed Jun 7, 2025

• onnxruntime-qnn silently failing when onnx model is not present

  #23623 closed Jun 7, 2025

• [Performance] FP16 Clip and Handle Bias introduces insufficient optimization.

  #23613 closed Jun 7, 2025

• Static Quantization "Shape mismatch" Error

  #23600 closed Jun 7, 2025

• [Build] Cmake unable to find OpenMP when running build with --use_dnnl flag for linux

  #23588 closed Jun 7, 2025

• [Build] Cannot build for arm32: error when linking libonnxruntime.so

  #23598 closed Jun 7, 2025

• [Build] Android x86_64 Cross Compiling on Mac OS

  #23648 closed Jun 7, 2025

• [Documentation] Clarify Lifetime Requirements of inputs to Ort::IoBinding

  #23689 closed Jun 7, 2025

• [Performance] Propagate NaNs in the CPU min and max operators introduces performance regression

  #23628 closed Jun 7, 2025

• With TensorRT EP, the output matrix is all zeros, but with CUDAEP, the output is correct.

  #23626 closed Jun 7, 2025

• Question about the ONNX Runtime 1.20.2 binary release

  #23721 closed Jun 7, 2025

• [Training] GRU and Squeeze artefact generation error

  #23698 closed Jun 7, 2025

• Adding an Execution Provider to ONNX Runtime Upstream

  #23700 closed Jun 7, 2025

• [Documentation] I/O Binding Needs Detail

  #23682 closed Jun 7, 2025

• [Web] [Feature Request] Ability to abort

  #23703 closed Jun 7, 2025

• Tensor Backing Buffer Mismatch Detected in Buffer Reuse

  #23739 closed Jun 7, 2025

• My system also has different versions of onnxruntime.dll. I have put the correct one in the same directory as the exe file, but I still get an error

  #23722 closed Jun 7, 2025

• Adding Execution Provider into ONNX RT

  #23732 closed Jun 7, 2025

• [Mobile] [urgent] iOS application crash at CreateEnv (pointer being freed was not allocated)

  #23759 closed Jun 7, 2025

• OpenVino Runtime Exception. Unexpected: CPU plug-in doesn't support If operation with dynamic rank. Operation name: input.15

  #23757 closed Jun 7, 2025

• [Build] no match for ‘operator=’ (operand types are ‘OrtMemoryInfo’ and ‘const OrtDevice') in memory_info.cc line 44 when onnxruntime_ENAABLE_MEMORY_PROFILE is enabled

  #23750 closed Jun 7, 2025

• Assistance with adjusting default Arena Allocator C/C++ API

  #23768 closed Jun 7, 2025

• [Web] Getting Started link on onnxruntime.ai website broken

  #23764 closed Jun 7, 2025

• Why the output of the ONNX MatMul node never be the same as what PyTorch gives?

  #23792 closed Jun 7, 2025

• Can load Fluxonnx Modal Components using InferenceSession

  #23770 closed Jun 7, 2025

• the memory leak using valgrind

  #23762 closed Jun 7, 2025

• Microsoft.ML.OnnxRuntime.QNN 1.20.1 includes unnecessary filew in win-arm64.

  #23781 closed Jun 7, 2025

• the memory usage not release

  #23774 closed Jun 7, 2025

• Blank output issue with CUDAExecutionProvider - Onnx Model Converted to fp16

  #23797 closed Jun 7, 2025

• [Performance]Do onednn executors depend on Intel platform

  #23795 closed Jun 7, 2025

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

  #23794 closed Jun 7, 2025

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

  #23816 closed Jun 7, 2025

• Abs node runs into error with bf16 tensor

  #23875 closed Jun 7, 2025

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

  #23897 closed Jun 7, 2025

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

  #23886 closed Jun 7, 2025

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

  #23885 closed Jun 7, 2025

• [Build] memory leaked

  #23915 closed Jun 7, 2025

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

  #23895 closed Jun 7, 2025

• [Documentation] Memory Leak in TensorRTProvider example

  #23901 closed Jun 7, 2025

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

  #23941 closed Jun 7, 2025

• [preprocess] Pad is not folded in Conv when opset_import is > 20

  #23973 closed Jun 7, 2025

• [Performance] does onnxruntime 1.19.0 support sve?

  #23983 closed Jun 7, 2025

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

  #23921 closed Jun 7, 2025

• Turning on coreml and turning off coreml are two results

  #24033 closed Jun 7, 2025

• Crashes when executing model quantification on Deeplabv3

  #23985 closed Jun 7, 2025

• onnxruntime errors out due to ORT_ENABLE_EXTENDED optimization: Can't remove node Transpose as it still has output edges.

  #24137 closed Jun 7, 2025

• [Build] Missmatch between CMake config and folder structure of onnxruntime-linux-x64-1.21.0.tgz

  #24003 closed Jun 7, 2025

• [Web] No way to prevent the default wasm from being bundled

  #24009 closed Jun 7, 2025

• OnnxRuntime gives different outputs on A100 v/s H100 GPU

  #24027 closed Jun 7, 2025

• [Build] PyInstaller build with onnxruntime issues with DT_RUNPATH

  #24044 closed Jun 7, 2025

• [Web] `Tensor.fromImage` crops, doesn't resize

  #24050 closed Jun 7, 2025

• [Build] The onnxruntime_tools-1.7.0 tarball on the PYPI site doesn't have requirements.txt and other files

  #24048 closed Jun 7, 2025

• onnxruntime-web requires webpack, but many users use metro

  #24052 closed Jun 7, 2025

• [Performance] Performance Bottleneck due to intra_op_num_threads being set globally

  #24101 closed Jun 7, 2025

• [Mobile] run speech using sherpa-onnx in the speech module, but if you want to use onnx inference in the translation module, you cannot initialize the ORT task.

  #24062 closed Jun 7, 2025

• [Performance] does acl support fused conv?

  #24063 closed Jun 7, 2025

• [Regression] Floating-point overflow with v1.21

  #24119 closed Jun 7, 2025

• [Build] Compile error with onnxruntime_providers_cuda.vcxproj

  #24099 closed Jun 7, 2025

• Wrong indexing in CPUIDInfo::ArmLinuxInit

  #24092 closed Jun 7, 2025

• Add option "any" for DirectML EP device_filter to onnxruntime perftest binary

  #24068 closed Jun 7, 2025

• [Build] MacOS universal binary build failure: "error: unknown target CPU 'armv8-a'"

  #24152 closed Jun 7, 2025

• [Build] Must build on Ubuntu 20.04 with gcc 9

  #24168 closed Jun 7, 2025

• error converting to onnx model

  #24198 closed Jun 7, 2025

• onnxruntime errors out due to ORT_ENABLE_BASIC optimization: [ONNXRuntimeError] : 1 : FAIL : Type Error: Shape of initializer v7_0 does not match. {1} != {}

  #24160 closed Jun 7, 2025

• [Build]Linker error when building for macCatalyst: Object file built for macOS

  #24153 closed Jun 7, 2025

• [Build] error C2653: 'system_clock': is not a class or namesp ace name

  #24145 closed Jun 7, 2025

• [Mobile][WebGPU][FeatureRequest] No true support for WebGPU

  #24165 closed Jun 7, 2025

• Bug: inconsistent output with transformer models between CUDA and CPU execution providers

  #24204 closed Jun 7, 2025

• [Web] Can't install behind NTLM proxy

  #24178 closed Jun 7, 2025

• onnxruntime errors out due to ORT_ENABLE_BASIC optimization: Unexpected data type for Clip 'min' input of 11

  #24158 closed Jun 7, 2025

• onnxruntime gpu package for Aarch64?

  #24159 closed Jun 7, 2025

• [Build] [Bug] The compiler doesn't support BFLOAT16!!! on Jetson Nano

  #24230 closed Jun 7, 2025

• GaussianProcessClassifier fails during ONNX runtime inference with "com.microsoft:Solve(-1) is not a registered function/op

  #24267 closed Jun 7, 2025

• ONNX preloaded dlls are incompatible with CUDNN torch version

  #24266 closed Jun 7, 2025

• QNN as ONNXruntime backend hangs while executing graph

  #24166 closed Jun 7, 2025

• Vector Assertion Failure in InferenceSession Init with Hotplugged-Off Cores on ARM (v1.21.0)

  #24221 closed Jun 7, 2025

• [Performance] Memory usage difference on Windows and Linux

  #24296 closed Jun 7, 2025

• onnxruntime errors out due to ORT_ENABLE_EXTENDED optimization: Error merging shape info for output

  #24340 closed Jun 7, 2025

• [Performance] Require Advance Profiling when running with DmlExecutionProvuder

  #24306 closed Jun 7, 2025

• GLU Operator gives different Results on Dml EP compared to CPU EP

  #24311 closed Jun 7, 2025

• how to get memory allocate detail in model infer?

  #24323 closed Jun 7, 2025

• [Build] WASM static lib build fails: no member named 'Negate' in 'onnxruntime::MLFloat16'

  #23769 closed Jun 6, 2025

• [Web] Fail to build static Wasm library without multi-thread support

  #24924 closed Jun 5, 2025

• [Feature Request] Implement RMSNormalization-23

  #24555 closed Jun 4, 2025

• [Java Mobile] libonnxruntime4j_jni.so incompatible with 16KB page size on ARM64 devices

  #24902 closed Jun 4, 2025

• [Build] Fail to build native WebGPU EP on Ubuntu Linux 24.04

  #24941 closed Jun 4, 2025

• Get Access violation error by simply create a new SessionOptions

  #24925 closed Jun 4, 2025

• An error(The SplitToSequence op is not supported) was thrown while loading an onnx file.

  #12286 closed Jun 3, 2025

• [Feature Request] Support for MatMul with broadcasting (eg. 3Dx2D matmul) for Xnnpack execution provider

  #24107 closed Jun 3, 2025

• [TensorRT ExecutionProvider] Cannot infer the model on a GPU device with an ID other than 0

  #21276 closed Jun 3, 2025

• Program crashes (segmentation fault) during interrupted load tests using TensorRT/CUDA EP

  #24601 closed Jun 3, 2025

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

  #24770 closed Jun 2, 2025

• [BUG] onnxruntime-node 1.22.0 fails to install on Linux - incorrect Windows path in NuGet package

  #24918 closed Jun 2, 2025

• [Build] Build failed when --use_tensorrt

  #24826 closed Jun 2, 2025

• `Unable to read model` with OpenVino provider, works with DML, CPU and OpenCV-Openvino

  #24926 closed Jun 2, 2025

11 Issues opened by 10 people

• [Build] CMake Error related to onnxruntime_unittests.cmake

  #24972 opened Jun 6, 2025

• onnxruntime produces invalid results due to the wrong shape inference for the clip operator

  #24971 opened Jun 6, 2025

• "What is ONNX Runtime (ORT)?,Converting Models to ONNX Format, Optimize Training and Inference with ONNX Runtime (ACPT/DeepSpeed)" links visually not appearing as links: A11y__ONNX Runtime & Ecosystem_Runtime_Usable

  #24965 opened Jun 5, 2025

• In Dark Mode, "Interested in using other languages? See the many others we support" link is not properly visible: A11y_ONNX Runtime & Ecosystem_Runtime_User Preferences

  #24964 opened Jun 5, 2025

• [Documentation] Roadmap page is completely outdated

  #24963 opened Jun 5, 2025

• [Documentation]

  #24958 opened Jun 5, 2025

• [Performance] ORT takes ~11GB memory for quantizing a model of size ~1GB

  #24954 opened Jun 4, 2025

• Memory safety for Nvidia GPU time-slicing

  #24943 opened Jun 3, 2025

• Llama3.2-1B ONNX Graph generated by olive auto-opt fails to run on DirectML execution provider

  #24937 opened Jun 3, 2025

• [Web] Fail to link static Wasm library with WebNN EP support

  #24936 opened Jun 3, 2025

• [BUG] Non-zero status code returned while running Resize node. in Direct ML backend

  #24928 opened Jun 2, 2025

51 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.

• Add GetCapability/Compile infrastructure for EP ABI

  #24887 commented on Jun 8, 2025 • 35 new comments

• Convert graph initializers into OrtValue Phase I

  #23979 commented on Jun 7, 2025 • 25 new comments

• [QNN EP] Add Support for If Op using Graph Transformation

  #24906 commented on Jun 8, 2025 • 13 new comments

• Add documentation for NV TensorRT RTX ExecutionProvider

  #24829 commented on Jun 6, 2025 • 5 new comments

• [TRT RTX EP] Implement GetEPContextNodes()

  #24901 commented on Jun 6, 2025 • 5 new comments

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

  #24869 commented on Jun 4, 2025 • 3 new comments

• [webgpu] support intel subgroup matrix on matmul_nbits

  #24898 commented on Jun 6, 2025 • 3 new comments

• [NV TensorRT RTX EP] misc fixes towards graphics interop

  #24909 commented on Jun 3, 2025 • 2 new comments

• Compile API: support for OrtModel input and write output to stream

  #24740 commented on Jun 2, 2025 • 1 new comment

• Consider making sympy optional

  #24872 commented on Jun 8, 2025 • 0 new comments

• [Feature Request] Support pointer-generator networks on T5 BeamSearch

  #23122 commented on Jun 7, 2025 • 0 new comments

• How to create custom op with fp16 input

  #23373 commented on Jun 8, 2025 • 0 new comments

• [Web] `Error: [WebGPU] Kernel "[Add] /model/model.5/conv1/ffc/convg2g/Add" failed. Error: Can't perform binary op on the given tensors`

  #24744 commented on Jun 8, 2025 • 0 new comments

• Initializer duplication method in QDQQuantizer ignores existing `value_info` tensor with same name

  #24705 commented on Jun 8, 2025 • 0 new comments

• [Build] How to build static lib?

  #24704 commented on Jun 8, 2025 • 0 new comments

• [MIGraphX EP] Adding Ortvalue features support for MGX EP

  #23404 commented on Jun 5, 2025 • 0 new comments

• Access violation when repeatedly creating/destroying inference session for TensorRT Execution Provider

  #24529 commented on Jun 2, 2025 • 0 new comments

• Add python bindings to the global thread pool functionality

  #24238 commented on Jun 5, 2025 • 0 new comments

• Enable SME for sgemm and sbgemm through KleidiAI

  #24346 commented on Jun 5, 2025 • 0 new comments

• [QNN EP] Add ONNX ScatterElements support

  #24811 commented on Jun 5, 2025 • 0 new comments

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

  #24833 commented on Jun 4, 2025 • 0 new comments

• use WebGPU EP instead of JSEP in WebAssembly

  #24856 commented on Jun 3, 2025 • 0 new comments

• Update xnnpack.cmake for WASM build

  #24860 commented on Jun 7, 2025 • 0 new comments

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

  #24881 commented on Jun 6, 2025 • 0 new comments

• [webgpu] Enable graph capture

  #24900 commented on Jun 7, 2025 • 0 new comments

• Update config spec

  #24913 commented on Jun 5, 2025 • 0 new comments

• [Web] Different result for a simple two-layer network between wasm and linux builds

  #24618 commented on Jun 2, 2025 • 0 new comments

• onnxruntime can execute the model correctly with the CUDAExecutionProvider, while returning error code with the CPUExecutionProvider: NOT_IMPLEMENTED : Could not find an implementation for Resize(13) node with name ''

  #24919 commented on Jun 2, 2025 • 0 new comments

• onnxruntime errors out due to the wrong process of GatherElements operator with the CPUExecutionProvider: Out of range value in index tensor

  #24917 commented on Jun 2, 2025 • 0 new comments

• Scale in resize node becomes an identity node not a parameter inside resize node

  #24824 commented on Jun 2, 2025 • 0 new comments

• Intermittent crash in ETW logging

  #24773 commented on Jun 2, 2025 • 0 new comments

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

  #24873 commented on Jun 3, 2025 • 0 new comments

• Attention fusion broken for BART 🤖

  #23864 commented on Jun 3, 2025 • 0 new comments

• Error messages from QNN are turned into verbose level messages

  #24876 commented on Jun 3, 2025 • 0 new comments

• Native WebGPU EP fails to run model with in-memory external data

  #24768 commented on Jun 3, 2025 • 0 new comments

• [Mobile] Undefined symbol _registerCustomOps with RapidOCR on iOS/MacCatalyst

  #24628 commented on Jun 4, 2025 • 0 new comments

• [Mobile] Objective-C API for register onnxruntime-extensions as a custom ops library

  #24613 commented on Jun 4, 2025 • 0 new comments

• onnxruntime with the CPUExecutionProvider errors out while processing the ReverseSequence operator

  #24920 commented on Jun 4, 2025 • 0 new comments

• Segmentation fault in `AppendExecutionProvider_CUDA_V2` when no GPU is available

  #24652 commented on Jun 5, 2025 • 0 new comments

• Regression in TreeEnsembleRegressor if the provided graph is a DAG

  #24636 commented on Jun 5, 2025 • 0 new comments

• ImportError: Unable to import dependency onnxruntime

  #24120 commented on Jun 5, 2025 • 0 new comments

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

  #24766 commented on Jun 6, 2025 • 0 new comments

• [Build] ORT can't build with cuda 12.9

  #24731 commented on Jun 6, 2025 • 0 new comments

• [Build] Mismatched library directory in linux-x64 package: lib and lib64

  #22267 commented on Jun 6, 2025 • 0 new comments

• Is class Sigmoid op supported by CUDA 12.6?

  #24670 commented on Jun 6, 2025 • 0 new comments

• Bug related to setting provider options for OpenVINO using Java API

  #24658 commented on Jun 6, 2025 • 0 new comments

• [Web] no available backend found [wasm] when importing `onnxruntime-web/wasm`

  #22010 commented on Jun 6, 2025 • 0 new comments

• [Build] OpenVINO ep for macOS

  #24273 commented on Jun 7, 2025 • 0 new comments

• AveragePool v19+ ignores `end` padding in computation when count_include_pad=1

  #24681 commented on Jun 7, 2025 • 0 new comments

• failed to build 1.21.0/onnxruntime/core/mlas: Assembler messages: Error: no such instruction: `{vex} vpdpbusds

  #24653 commented on Jun 7, 2025 • 0 new comments

• Import error in pytest with onnxruntime-directml 1.22.0

  #24907 commented on Jun 7, 2025 • 0 new comments