41 Pull requests merged by 26 people

• Bump tmp from 0.2.1 to 0.2.4 in /onnxruntime/test/wasm

  #25671 merged Aug 7, 2025

• [webgpu] support float16 type for Einsum operator

  #25443 merged Aug 7, 2025

• Update number of mel bins for whisper model

  #25675 merged Aug 7, 2025

• Allow DML EP to be used with any CPU EP

  #25664 merged Aug 7, 2025

• Cherry-pick MiGraphX EP fixes from upstream for rel-1.23.0

  #25659 merged Aug 7, 2025

• ORT perf test support for plugin EP

  #25374 merged Aug 6, 2025

• Enable BrowserStack testing stage

  #25668 merged Aug 6, 2025

• Remove training packages from onnxruntime-ios-packaging-pipeline

  #25451 merged Aug 6, 2025

• Fix the is_leaf check in TreeEnsemble

  #25410 merged Aug 6, 2025

• upgrade dawn to 794b6fadc4171f7b853a77ffdf0948fbec431f41

  #25461 merged Aug 6, 2025

• Update python bindings to be able to use a shared allocator and/or IDataTransfer registered by a plugin EP in the Environment

  #25346 merged Aug 6, 2025

• [NV TRT RTX EP] Cumulative TRT RTX EP merge

  #25656 merged Aug 6, 2025

• [build] fix build with delay load hook

  #25657 merged Aug 5, 2025

• [build] fix WebAssembly build on macOS/arm64

  #25653 merged Aug 5, 2025

• [web] fix support for subgroup

  #25649 merged Aug 4, 2025

• Add patch for WebGPU on Android to handle fp16 in uniforms

  #25349 merged Aug 4, 2025

• Cherry-pick PR #25626 to 1.23.0 release branch

  #25640 merged Aug 4, 2025

• Enable 2bit CPU matmul fallback

  #25582 merged Aug 4, 2025

• [QNN EP] Add support for GatherNd Op in QNN EP

  #25635 merged Aug 4, 2025

• Update qMoE spec to support block quantization

  #25641 merged Aug 4, 2025

• [MIGraphX EP] Fix CreateExecutionProviderFactory with correct struct and change vendor_id

  #25625 merged Aug 4, 2025

• [CANN] Fix CANN build error

  #25627 merged Aug 4, 2025

• Update MoE and qMoE spec

  #25619 merged Aug 2, 2025

• [build] use cross-compile to build macOS x86_64 target for WebGPU

  #25617 merged Aug 2, 2025

• Add support for QMoE in CPU

  #25558 merged Aug 2, 2025

• Move moving weights to memory to the end of Graph::Resolve()

  #25626 merged Aug 2, 2025

• Add CUDA implementation of GatherBlockQuantized operator

  #25575 merged Aug 1, 2025

• Cherry-picks for ORT 1.23.0

  #25620 merged Aug 1, 2025

• [build] fix macOS x86_64 cross-compile warning

  #25615 merged Aug 1, 2025

• update CANN docs

  #25624 merged Aug 1, 2025

• [webgpu] Apply Flash Attention if sliding window exceeds KV cache length

  #25594 merged Aug 1, 2025

• Update macOS target version from 13.3 to 13.4

  #25616 merged Aug 1, 2025

• [QNN-EP] Resolve VTCM buffer sharing bugs

  #25622 merged Aug 1, 2025

• [QNN EP] Add ONNX ScatterElements support

  #24811 merged Aug 1, 2025

• [QNN EP] Bug fix: multiple consumer for cast result in name conflicts

  #25584 merged Aug 1, 2025

• Optimize layout for SubgroupMatrixLoad on Intel

  #25384 merged Aug 1, 2025

• [QNN EP] Lower Gemm with 2d bias to FC + ElementwiseAdd when targeting HTP.

  #25605 merged Aug 1, 2025

• [build] disable CodeQL for NPM Packaging Pipeline

  #25614 merged Aug 1, 2025

• Cache opSupportLimits to improve the performance and update tracing e…

  #25589 merged Jul 31, 2025

• Refactor Java Test Pipeline

  #25608 merged Jul 31, 2025

• [QNN EP] Add Unit tests for LPBQ Fusions

  #25592 merged Jul 31, 2025

34 Pull requests opened by 31 people

• Reduce CMake's CUDA_ARCHITECTURES

  #25618 opened Jul 31, 2025

• fix(ort): add automatic patching for nvidia cudnn library path

  #25628 opened Aug 1, 2025

• [VitisAI] bugfix model_clone optimization

  #25629 opened Aug 1, 2025

• [webgpu] support GatherND operator

  #25632 opened Aug 1, 2025

• rel-1.22.2 cherry-pick 1

  #25633 opened Aug 1, 2025

• Depthwise conv 3x3 s1

  #25637 opened Aug 2, 2025

• Add more tests to GatherBlockQuantized operator

  #25639 opened Aug 2, 2025

• Optimizations and fixes in QMoE CPU kernel

  #25642 opened Aug 3, 2025

• Bump ruff from 0.12.4 to 0.12.7

  #25643 opened Aug 4, 2025

• Support int4 and uint4 for reshape on opset 21+.

  #25645 opened Aug 4, 2025

• DequantizeLinear should support non-zero zero_point when input type is int32

  #25646 opened Aug 4, 2025

• Add comprehensive ThresholdedRelu custom operator example and fix common implementation issues

  #25650 opened Aug 4, 2025

• Skip node output dump for MemcpyToHost

  #25651 opened Aug 5, 2025

• Properly remove in-memory references

  #25652 opened Aug 5, 2025

• Bugfix vitisai ep model clone with 23979 25320

  #25654 opened Aug 5, 2025

• safeint.h: quelch gcc's -Wreturn-type

  #25655 opened Aug 5, 2025

• Update semver.h to fix compilation error under linux

  #25658 opened Aug 5, 2025

• Remove incorrect function calls

  #25662 opened Aug 6, 2025

• fixed matmul broadcasting

  #25663 opened Aug 6, 2025

• Python binding for listing custom operators.

  #25665 opened Aug 6, 2025

• [QNN-EP] Add CastLoneQFusion to transform Cast and QNode into Convert

  #25667 opened Aug 6, 2025

• Replace vmlaq_f32 with vfmaq_f32 (fused multiply-add)

  #25669 opened Aug 6, 2025

• Bump transformers from 4.50.0 to 4.53.0 in /onnxruntime/python/tools/transformers/models/stable_diffusion/requirements

  #25672 opened Aug 6, 2025

• Relax WeightBiasQuantization constraint for larger QDQ node group

  #25673 opened Aug 6, 2025

• [webgpu] support bool for binary operators

  #25674 opened Aug 6, 2025

• Add vendorid check to GetSupportedDevicesImpl for MIGraphx EP

  #25677 opened Aug 6, 2025

• [WIP] Test integration with ONNX 1.19

  #25678 opened Aug 7, 2025

• [WebNN] Remove NHWC preferred layout

  #25679 opened Aug 7, 2025

• FP16 inference performance improvement on CPU

  #25680 opened Aug 7, 2025

• [MIGraphX EP][BUG] Fix pybind compilation with MIGraphX after merging #25346

  #25683 opened Aug 7, 2025

• Skeleton for Attention(23) on CUDA

  #25684 opened Aug 7, 2025

• 2-bit TMAC matmul

  #25686 opened Aug 7, 2025

• Update QAIRT to 2.37.0

  #25688 opened Aug 7, 2025

• [WIP] Move provider tests to `onnxruntime_provider_test` and enable use of plugin EPs

  #25689 opened Aug 7, 2025

1,207 Issues closed by 31 people

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

  #22987 closed Aug 7, 2025

• Microsoft.AI.MachineLearning cannot be used in UWP app on on Windows 10 ARM64

  #4686 closed Aug 7, 2025

• Debugging capability of onnxruntime in Visual Studio 2019 incapacitated

  #4812 closed Aug 7, 2025

• [WinML] [C++/WinRT] Clarify how to share Ort::Env environments with WinRT/WinML instances

  #4971 closed Aug 7, 2025

• C Sharp API for openvino doesn't run on GPU

  #5011 closed Aug 7, 2025

• onxruntime-gpu installation issues

  #5020 closed Aug 7, 2025

• program stucks when multi processes

  #5093 closed Aug 7, 2025

• Exception thrown from Dispose method (When missing dependency)

  #5250 closed Aug 7, 2025

• DLRM model failure to execute on GPU

  #5295 closed Aug 7, 2025

• Running quantized models on GPU

  #5359 closed Aug 7, 2025

• Can Session::Run be const?

  #5558 closed Aug 7, 2025

• ML.NET issue while Using yolov4 onnx model

  #5593 closed Aug 7, 2025

• Passing Non-Const pointer to Session::Run() using CPP Api

  #5597 closed Aug 7, 2025

• How to reduce memory used?

  #5711 closed Aug 7, 2025

• openvino build failed nuget

  #5749 closed Aug 7, 2025

• How to loading a pytorch model with input shape of (None, 32) using the C# inference ?

  #5781 closed Aug 7, 2025

• Any support for double type tensor when loading pytorch onnx model ?

  #5782 closed Aug 7, 2025

• memory keep increasing with dynamic input shape of network

  #5796 closed Aug 7, 2025

• Memory usage with Cuda ExecutionProvider

  #5801 closed Aug 7, 2025

• onnxruntime.capi.onnxruntime_pybind11_state.Fail: [ONNXRuntimeError] : 1 : FAIL : The node is not placed on any Execution Provider. OneHot(11) (node while/cond_5/one_hot).

  #5825 closed Aug 7, 2025

• Non-zero status code returned while running Div node

  #5830 closed Aug 7, 2025

• Performance comparison

  #5834 closed Aug 7, 2025

• IOBindings in C++ API are missing a way to SynchronizeInputs.

  #5857 closed Aug 7, 2025

• How to compile with vs2019, with the platform tool set "Visual Studio 2015 - Windows XP (v140_xp)"， i want use it in xp system

  #5859 closed Aug 7, 2025

• Quantized model much slower than full precision model

  #5865 closed Aug 7, 2025

• Performance issue with operator Where on CPU

  #5896 closed Aug 7, 2025

• Performance issue with operators SVMRegressor and SVMClassifier for RBF kernel on CPU

  #5898 closed Aug 7, 2025

• failed:/onnxruntime_src/onnxruntime/core/graph/model_load_utils.h:47 void onnxruntime::model_load_utils::ValidateOpsetForDomain

  #5905 closed Aug 7, 2025

• Support GCN

  #5910 closed Aug 7, 2025

• EyeLike with dynamic shape results in error

  #5917 closed Aug 7, 2025

• Can't train mnist in parallel

  #5918 closed Aug 7, 2025

• could not open "tensorrt_provider_factory.h", "mkldnn_provider_factory.h"

  #5925 closed Aug 7, 2025

• Dynamic shape got wrong output

  #5928 closed Aug 7, 2025

• Issue with Multi-GPU and GPU memory limit

  #5939 closed Aug 7, 2025

• can not get expected speed in onnxruntime

  #5953 closed Aug 7, 2025

• Error using onnx model containing Bidirectional layer with MatMulAddFusion

  #5955 closed Aug 7, 2025

• No opset import for domain 'com.microsoft'

  #5971 closed Aug 7, 2025

• "undefined symbol" error occured, when I use ort.SessionOptions.register_custom_ops_library

  #5984 closed Aug 7, 2025

• Under TRT EP, custom op cannot fall back to CUDA EP

  #6002 closed Aug 7, 2025

• Inconsistent inference time between C Python API [Megatron-LM]

  #6025 closed Aug 7, 2025

• Onnx Batch Processing

  #6044 closed Aug 7, 2025

• How to extract the size of a map type in c++?

  #6077 closed Aug 7, 2025

• could the checkpoint of bert convert to onnx model? I have a bug that 'BertForPreTraining' object has no attribute 'layers, output'

  #6089 closed Aug 7, 2025

• how to implement execution provider (EP) that allow onnx run on my hardware?

  #6110 closed Aug 7, 2025

• 32bit vs 64bit when compiling or something else?

  #6144 closed Aug 7, 2025

• GPU memory consumption keeps increasing with multithreading in Java

  #6181 closed Aug 7, 2025

• Not support rtx 3000 series

  #6213 closed Aug 7, 2025

• sample c++ program just print "hello" does not start

  #6243 closed Aug 7, 2025

• Cannot create OnnxTensor with UINT8 type.

  #6261 closed Aug 7, 2025

• Referencing Microsoft.ML.OnnxRuntime and Microsoft.ML.OnnxRuntime.GPU in a c# project.

  #6264 closed Aug 7, 2025

• Could onnxruntime be compiled into wasm using emsdk?

  #6275 closed Aug 7, 2025

• Performance shaking

  #6301 closed Aug 7, 2025

• [Bug] Wrong implementation in LpPool

  #6302 closed Aug 7, 2025

• Memory corruption when using OnnxRuntime with OpenVINO on the Intel MyriadX and Raspberry Pi 4B

  #6304 closed Aug 7, 2025

• [ONNXRuntimeError] : 2 : INVALID_ARGUMENT : Error while inferencing DLRM onnx model

  #6319 closed Aug 7, 2025

• Error: Running double precision model exported from pyTorch

  #6320 closed Aug 7, 2025

• The output for GPT is NAN when fp16=True

  #6328 closed Aug 7, 2025

• ROCm build seems broken: `error: ‘ncclComm_t’ does not name a type`

  #6358 closed Aug 7, 2025

• Implementation of ONNX Functions

  #6360 closed Aug 7, 2025

• Incorrect TypeInferenceError on UNDEFINED tensor type

  #6370 closed Aug 7, 2025

• [C-Api] Dynamic Shape Error: Non-zero status code returned while running Sigmoid node.

  #6372 closed Aug 7, 2025

• onnxruntime v1.6.0 on Jetson Nano - Illegal Instruction (core dumped)

  #6375 closed Aug 7, 2025

• How to extract dimension of inputs in onnxruntime/core/providers/cpu/math/matmul.cc

  #6396 closed Aug 7, 2025

• Which executor to build when using: Intel® Deep Learning Boost (Intel® DL Boost)

  #6400 closed Aug 7, 2025

• [question] Configure GPU arena with Python bindings

  #6411 closed Aug 7, 2025

• Onnxruntime error when Relu-layer follows Dense-layer without activation and biases

  #6423 closed Aug 7, 2025

• Reshape `requested_shape` forced to have leading dimension 1 when it should be -1

  #6424 closed Aug 7, 2025

• Build failure in `orttraining_pybind_state.cc` when building with `--enable_training` and `--build_wheel`

  #6536 closed Aug 7, 2025

• NaN in AveragePooling

  #6543 closed Aug 7, 2025

• Loss of accuracy when GPT-2 based model is exported to ONNX

  #6549 closed Aug 7, 2025

• Custom Op Registration and Implementation

  #6564 closed Aug 7, 2025

• Inference error using migraohx-onnxruntime

  #6605 closed Aug 7, 2025

• /onnxruntime/core/mlas/lib/quantize.cpp:50:62: error: ‘vminnmq_f32’ was not declared in this scope

  #6638 closed Aug 7, 2025

• Failed to add Microsoft.AI.MachineLearning NuGet package to .NET Framework 4.6.1 projects

  #6662 closed Aug 7, 2025

• INT8 quantized model is very slow

  #6732 closed Aug 7, 2025

• Shape inference error for Range node

  #6737 closed Aug 7, 2025

• onnxruntime-gpu (cudaexecutionprovider) usage of cudnn autotuner

  #6744 closed Aug 7, 2025

• Unable to compile on Linux with CUDA

  #6749 closed Aug 7, 2025

• Onnxruntime inference with Integrated GPU Failed

  #6755 closed Aug 7, 2025

• Onnxruntime.gpu is as slower than cpu mode

  #6799 closed Aug 7, 2025

• Multiple input and multiple output models that create tensors in loops can cause serious crashes

  #6821 closed Aug 7, 2025

• ONNXRuntime Inference with Finetuned BERT Model outputting odd results

  #6830 closed Aug 7, 2025

• Unable to build onnxruntime with "--build_wheel" and "--enable_pybind" options

  #6841 closed Aug 7, 2025

• [JAVA Bindings + Android arm64-v8a] ONNXRuntime build documentation

  #6923 closed Aug 7, 2025

• dynamic shape input is much slower than fixed shape input in gpu

  #6978 closed Aug 7, 2025

• CUDA header requested but missing in DNNL part of ORT 1.7.1

  #7005 closed Aug 7, 2025

• Build fail for docker on MacOS. -NO GPU.

  #7052 closed Aug 7, 2025

• Large Memory Allocations When Loading RandomForestRegressor Model

  #7067 closed Aug 7, 2025

• Non-zero status code returned while running BatchNormalization node

  #7095 closed Aug 7, 2025

• Memory and timing issue with onnxruntime python API with TensorFlow model

  #7106 closed Aug 7, 2025

• Compile error in header onnxruntime_cxx_api.h when update ONNX runtime from 1.5.2 to 1.7.1

  #7142 closed Aug 7, 2025

• Batch inference

  #7178 closed Aug 7, 2025

• Segmentation fault when running onnxruntime inside docker with cpuset restrictions

  #7207 closed Aug 7, 2025

• Significant difference in the performance of pytorch and exported onnx models

  #7212 closed Aug 7, 2025

• TensorrtExecutionProvider slower than CUDAExecutionProvider: Transformers

  #7230 closed Aug 7, 2025

• The speed of running the onnx model is 6x slower than the pytorch model on Jetson TX2

  #7233 closed Aug 7, 2025

• [Python API + ARM64] Running ResNet50 on ARM board using ACL Error and Performance Issue

  #7234 closed Aug 7, 2025

• ACL (32bit) Execution Provider fails on gemm node

  #7255 closed Aug 7, 2025

• onnxruntime gpu version can't installed, how to fix it?

  #7272 closed Aug 7, 2025

• Run a model containing CustomOp with TensorRT provider fails

  #7314 closed Aug 7, 2025

• C# console app crash upon appending OpenVino execution provider

  #7330 closed Aug 7, 2025

• Cannot save Tensorrt .engine model in v1.7.1

  #7339 closed Aug 7, 2025

• openvino continued package by pyinstaller external dll issue

  #7346 closed Aug 7, 2025

• Resize Operator rounds-down instead of round-to-even for int32/uint8

  #7368 closed Aug 7, 2025

• How to compile the framework that can run in Windows XP？

  #7444 closed Aug 7, 2025

• Please, update the docs. Provider parameter "cuda_mem_limit" was renamed to "gpu_mem_limit" in nightly build.

  #7457 closed Aug 7, 2025

• How to release gpu memory without exiting the process？

  #7463 closed Aug 7, 2025

• Running inference using GPU or TensorRT on Jetson

  #7484 closed Aug 7, 2025

• Problem compiling ONNX RT with CUDA and TensorRT on Windows

  #7562 closed Aug 7, 2025

• Use of torch InstanceNorm2d and dynamic tensor size causes crash

  #7572 closed Aug 7, 2025

• onnxruntime build is not compatible with onnx build. Protobuf loaded twice.

  #7597 closed Aug 7, 2025

• Large GPU memory usage with EXHAUSTIVE cuDNN search

  #7612 closed Aug 7, 2025

• Enable CUDA provider option configuration in Java

  #7613 closed Aug 7, 2025

• Build fails with --use_rknpu

  #7614 closed Aug 7, 2025

• Publish the providers with the release build

  #7628 closed Aug 7, 2025

• int8 quantization on GPU support? (transformers)

  #7634 closed Aug 7, 2025

• Does onnxruntime support bert with relative position embedding

  #7713 closed Aug 7, 2025

• quantize model can‘t run on gpu ?

  #7745 closed Aug 7, 2025

• TensorRT execution provider SEGFAULT

  #7757 closed Aug 7, 2025

• CUDA kernel not found in registries for Op type: Pad

  #7779 closed Aug 7, 2025

• ACL and ArmNN v21.02 EP has problem with GEMM

  #7784 closed Aug 7, 2025

• get error when using a model with custom op

  #7788 closed Aug 7, 2025

• Force fallback to CPU execution for Gather, Unsqueeze, Concat nodes - onnxruntime-gpu 1.7.0, opset 12 and 13

  #7792 closed Aug 7, 2025

• How to get sparse tensor input in custom op?

  #7838 closed Aug 7, 2025

• Non-zero status code returned while running FusedConv node. Name:'fused ' onnxruntime::OpKernelContext::Input Missing Input: input

  #7853 closed Aug 7, 2025

• Build failure in onnxruntime/test/featurizers_ops/truncated_svdtransformer_test.cc

  #7878 closed Aug 7, 2025

• onnxruntime::BroadcastIterator::Init axis == 1 || axis == largest was false. Attempting to broadcast an axis by a dimension other than 1. 15 by 64

  #7888 closed Aug 7, 2025

• undefined reference to `onnx::optimization::GetAvailablePasses() on Nvidia Jetson NX

  #7970 closed Aug 7, 2025

• Runtime exception during initialization of SparkML model (One falsenode is pointing either to itself, either to another tree.)

  #8008 closed Aug 7, 2025

• Running multiple input node onnx model using onnxrntime C/C++ API

  #8019 closed Aug 7, 2025

• Memory leak in free-dimention model in C++

  #8053 closed Aug 7, 2025

• CUDAExecutionProvider does not handle Clip on float16 tensor.

  #8070 closed Aug 7, 2025

• Why ReduceSum get shape 0 for an empty input?

  #8146 closed Aug 7, 2025

• System memory leak on cuda GPU backend.

  #8147 closed Aug 7, 2025

• Does ONNX Runtime and its execution providers support FP16 inference?

  #8173 closed Aug 7, 2025

• Reflect padding output seems incorrect when padding size larger than input dimension

  #8265 closed Aug 7, 2025

• ai.onnxruntime.OrtException: Error code - ORT_FAIL - message: OrtSessionOptionsAppendExecutionProvider_Cuda: Failed to load shared library

  #8283 closed Aug 7, 2025

• Eigen::ThreadPoolInterface*, const onnxruntime::ThreadOptions&) pthread_setaffinity_np failed

  #8313 closed Aug 7, 2025

• Inference Speed is slow on GPU

  #8316 closed Aug 7, 2025

• After 8bit quantization, the GPU inference speed is very slow

  #8330 closed Aug 7, 2025

• GPUs operate slower than CPUs

  #8362 closed Aug 7, 2025

• error using C# tensorRT EP builded from source

  #8367 closed Aug 7, 2025

• Why cuda provider allocator must be threadlocal?

  #8378 closed Aug 7, 2025

• Implement Split for double or float64 data type

  #8382 closed Aug 7, 2025

• ERROR running model inference:Non-zero status code returned while running Cast node

  #8424 closed Aug 7, 2025

• Found regression on ORT 1.8.1

  #8513 closed Aug 7, 2025

• Does the onnxruntime.quantization.quantize_dynamic support GPU quantization?

  #8524 closed Aug 7, 2025

• gpu memory can not release.

  #8544 closed Aug 7, 2025

• Build failure of onnxruntime Docker container with Vitis-AI

  #8596 closed Aug 7, 2025

• PrepareForCompute Non concat axis dimensions must match: Axis 0 has mismatched dimensions of 1 and 0

  #8685 closed Aug 7, 2025

• error with torch.sum or torch.tensor.mean operator on GPU

  #8742 closed Aug 7, 2025

• Symbolic shape inference error for loop node & seq(tensor)

  #8755 closed Aug 7, 2025

• onnxruntime Jetson tx2 cuda

  #8771 closed Aug 7, 2025

• AttributeError: module 'onnxruntime' has no attribute 'set_default_logger_severity'

  #8789 closed Aug 7, 2025

• IsNaN and Split have no double implementations

  #8791 closed Aug 7, 2025

• Readily available Python wheels for ARM?

  #8874 closed Aug 7, 2025

• cannot import name ‘get_all_providers‘

  #8907 closed Aug 7, 2025

• Runetime Error: Decoder with dynamic axes does not work with Encoder output

  #8910 closed Aug 7, 2025

• How to get the value of tensors in subgraph?

  #8929 closed Aug 7, 2025

• The model run time become longer when i update the onnxruntime from version 1.7 to version 1.8

  #8938 closed Aug 7, 2025

• ONNX inference result are different to pytorch model

  #8977 closed Aug 7, 2025

• Type error when runs an control flow model in ORT

  #8999 closed Aug 7, 2025

• [E:onnxruntime:, sequential_executor.cc:339 Execute] Non-zero status code returned while running Transpose node. Name:'model/unet3d_segmentation/conv3d_12/Conv3D__165' Status Message: CUDA error cudaErrorInvalidConfiguration:invalid configuration argument

  #9083 closed Aug 7, 2025

• cross compile but onnx-ml.pb.cc error

  #9093 closed Aug 7, 2025

• how to input 'None' in cpp-version

  #9121 closed Aug 7, 2025

• InferenceSession.run in python is inconsistent in terms of performance

  #9208 closed Aug 7, 2025

• Can't load Cuda Provider on Linux due symbol lookup error

  #9309 closed Aug 7, 2025

• ONNXRuntime CPU - Memory spiking continuously (Memory leak)

  #9313 closed Aug 7, 2025

• error: '_Frees_ptr_opt_' has not been declared

  #9332 closed Aug 7, 2025

• QLinearConv per-channel result is wrong and it's seem overflow when input is big for my model

  #9365 closed Aug 7, 2025

• ORT execution fails when a gradient builder is not registered for module-local functions

  #9375 closed Aug 7, 2025

• Relu getting dropped during quantization

  #9425 closed Aug 7, 2025

• OnnxRuntime Build Failure in Docker

  #9530 closed Aug 7, 2025

• YAMNet model running on CudaExecutionProvider is 3x slower than running on tensorflow

  #9657 closed Aug 7, 2025

• Gap in inference time between onnxruntime and torch vanishes when increasing the batch size

  #9660 closed Aug 7, 2025

• libonnxruntime.so crash

  #9684 closed Aug 7, 2025

• [ONNXRuntimeError] : 9 : NOT_IMPLEMENTED : Could not find an implementation for Conv(1) node with name 'Conv_0'

  #9685 closed Aug 7, 2025

• yolov5 with the compiled onnxruntime by self，but is so slow, not with the GPU

  #9689 closed Aug 7, 2025

• Unable to load shared library 'onnxruntime' on MacOS (DllNotFoundException)

  #9707 closed Aug 7, 2025

• Support for int64 with webgl backend of the web runtime

  #9724 closed Aug 7, 2025

• ouput of onnx model with custom op in the loop structrue is confusing

  #9742 closed Aug 7, 2025

• How to build for multiple execution provider?

  #9756 closed Aug 7, 2025

• Inference is slower when running inside Docker

  #9767 closed Aug 7, 2025

• [ONNXRuntimeError] : 1 : FAIL : Fatal error: test_custom is not a registered function/op

  #9831 closed Aug 7, 2025

• non-NEON Compatibility

  #9849 closed Aug 7, 2025

• Yolov5 ORT train failed with onnxruntime backend

  #9936 closed Aug 7, 2025

• Support for pip wheel tensorrt

  #9986 closed Aug 7, 2025

• question about warnup long time

  #10017 closed Aug 7, 2025

• Importing onnxruntime on AWS Lambdas with ARM64 processor causes crash

  #10038 closed Aug 7, 2025

• how to forward with a batch images, oncetime?

  #10071 closed Aug 7, 2025

• when my models input size is 3808, then i forward with yolov5, the memry is break.

  #10074 closed Aug 7, 2025

• Same Pad_Head value in ORT for SAME_UPPER/SAME_LOWER if get negative odd pad value

  #10086 closed Aug 7, 2025

• onnxruntime latest version segment fault

  #10113 closed Aug 7, 2025

• ORTModule import error : with onnxruntime

  #10127 closed Aug 7, 2025

• BatchNorm fails on CUDA EP with zero length sequences

  #10128 closed Aug 7, 2025

• Do you have any plan to add 'Round' Operator for gradient builder registry for orttrainer?

  #10138 closed Aug 7, 2025

• Performance question about some nodes generated by dynamic quantization

  #10153 closed Aug 7, 2025

• Sigmoid fails and output all zeros

  #10154 closed Aug 7, 2025

• Why does onnxruntime run slower on C++?

  #10155 closed Aug 7, 2025

• `InferenceSession` initialization hangs

  #10166 closed Aug 7, 2025

• TensorRT EP failed to set INT8 dynamic range.

  #10206 closed Aug 7, 2025

• how to use docker and onnxruntime deploy onnx model on GPU?

  #10257 closed Aug 7, 2025

• Inconsistent inference timing on CPU

  #10270 closed Aug 7, 2025

• Inference: Time in GPU is similar in CPU. GPU not speed up

  #10271 closed Aug 7, 2025

• multiple InferenceSession slowdown inference speed

  #10273 closed Aug 7, 2025

• DnnlExecutionProvider is not visible in python API

  #10275 closed Aug 7, 2025

• add QLinearMatMul do not quantize per channel flag to quantize_static extra options

  #10283 closed Aug 7, 2025

• onnxruntime inference is around 5 times slower than pytorch when using GPU

  #10303 closed Aug 7, 2025

• Bug: pthread sent an error! undefined:undefined: ortWasmThreaded is not defined

  #10311 closed Aug 7, 2025

• Quantized int8 onnx GPT2 model returns different tokens whether using past_key_values or not for the same sentence

  #10322 closed Aug 7, 2025

• Onnxruntime multithread options [C++ CPU]

  #10330 closed Aug 7, 2025

• Issues when trying to use Onnxruntime and Tensorrt execution provider in a java application

  #10352 closed Aug 7, 2025

• build onnxruntime error linux

  #10364 closed Aug 7, 2025

• Error happened while building onnxruntime

  #10378 closed Aug 7, 2025

• Question about hidden states in onnx DistilGPT2

  #10382 closed Aug 7, 2025

• Is TensorRT execution provider caching is thread-safe

  #10412 closed Aug 7, 2025

• Loading a Keras model with custom layers into Microsoft.ML

  #10419 closed Aug 7, 2025

• ort-web Error: invalid input shape. when using webgl backend and there is a torch.nn.BatchNorm1d layer in the network

  #10437 closed Aug 7, 2025

• cast BatchNorm2d to int32

  #10440 closed Aug 7, 2025

• TensorRT input: 717 has no shape specified.

  #10443 closed Aug 7, 2025

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

  #10484 closed Aug 7, 2025

• C++ OnnxRuntime-GPU Slower than Python OnnxRuntime-GPU/C++ OnnxRuntime-CPU

  #10492 closed Aug 7, 2025

• slower after graph optimization!

  #10538 closed Aug 7, 2025

• onnxruntime and onnxruntime-gpu produce different output for ReduceL1 operator

  #10542 closed Aug 7, 2025

• Run maskrcnn onnx from pytorch and inference on c++ with gpu sometimes will error

  #10543 closed Aug 7, 2025

• Exception in DirectML on second inference run

  #10546 closed Aug 7, 2025

• Unit Tests failure while building on Windows with CUDA EP

  #10561 closed Aug 7, 2025

• Building Error

  #10600 closed Aug 7, 2025

• OpenVINO Execution provider's CPU Utility is low

  #10601 closed Aug 7, 2025

• How to use OpenVINO GetAvailableDevices?

  #10602 closed Aug 7, 2025

• why it take 200 seconds to run onnxruntime.InferenceSession

  #10608 closed Aug 7, 2025

• Building OnnxRuntime v1.10.0 with CUDAExecutionProvider for sm_75 GPU fails in CUDA10.2 environment

  #10610 closed Aug 7, 2025

• C + + onnxruntime GPU is ten times slower than CPU

  #10611 closed Aug 7, 2025

• Optimization for T5 transformer models.

  #10613 closed Aug 7, 2025

• [E:onnxruntime:, sequential_executor.cc:346 Execute] Non-zero status code returned while running Add node. Name:'Add_1363' Status Message: /onnxruntime_src/onnxruntime/core/providers/cpu/math/element_wise_ops.h:505 void onnxruntime::BroadcastIterator::Append(ptrdiff_t, ptrdiff_t) axis == 1 || axis == largest was false. Attempting to broadcast an axis by a dimension other than 1. 9 by 505

  #10618 closed Aug 7, 2025

• about providers and providers_options in InferenceSession

  #10620 closed Aug 7, 2025

• How to use mimalloc in Linux?

  #10629 closed Aug 7, 2025

• CPU & CUDA execution provider produce different value

  #10636 closed Aug 7, 2025

• No libonnxruntime_providers_cuda.so generated?

  #10639 closed Aug 7, 2025

• onnxruntime.capi.onnxruntime_pybind11_state.RuntimeException: [ONNXRuntimeError] : 6 : RUNTIME_EXCEPTION

  #10657 closed Aug 7, 2025

• Get wrong result when use webgl backend

  #10673 closed Aug 7, 2025

• onnxruntime::Graph::CleanUnusedInitializersAndNodeArgs initializer_node_arg != nullptr was false.

  #10677 closed Aug 7, 2025

• Need help on the following from wiki listed roadmap.

  #10689 closed Aug 7, 2025

• Output shape is mismatched with ONNX SPEC about Resize_tf_crop_and_size with scale input

  #10727 closed Aug 7, 2025

• gpu onnxruntime lib

  #10731 closed Aug 7, 2025

• Onnx model consumes huge CPU memory

  #10742 closed Aug 7, 2025

• inference qdq model failed with TRT EP.

  #10743 closed Aug 7, 2025

• build on windows cup is fine，but cuda not

  #10745 closed Aug 7, 2025

• Is there a version of onnxruntime that is compatible with windows 7?

  #10749 closed Aug 7, 2025

• can build on windows with Geforce 1060 card, cuda 11.0 cudnn 8.0.2 successfully?

  #10763 closed Aug 7, 2025

• very slow in inference

  #10764 closed Aug 7, 2025

• ONNX models give slower inference in Python Multiprocessing

  #10786 closed Aug 7, 2025

• Inference time of onnxruntime gpu increases at very high batch sizes

  #10789 closed Aug 7, 2025

• Transformer optimizer outputs confusing error

  #10838 closed Aug 7, 2025

• C++ is 10x slower compared with Python, CPU only

  #10849 closed Aug 7, 2025

• Windows 32 bit performance much slower than 64bit?

  #10855 closed Aug 7, 2025

• Different inference results from python and C#

  #10863 closed Aug 7, 2025

• Does WebGL fail when network inputs are not dimensions in powers of two?

  #10873 closed Aug 7, 2025

• TensorRT conversion support on Huggingface transformers quantized models.

  #10888 closed Aug 7, 2025

• onnxruntime/capi/onnxruntime_inference_collection.py", line 370, in _create_inference_session sess = C.InferenceSession(session_options, self._model_path, True, self._read_config_from_model) onnxruntime.capi.onnxruntime_pybind11_state.InvalidProtobuf: [ONNXRuntimeError] : 7 : INVALID_PROTOBUF : Load model from onnx_data/cpm_large_opt.onnx failed:Protobuf parsing failed.

  #10892 closed Aug 7, 2025

• python3 -m onnxruntime_tools.transformers.optimizer when opt_level=1 comes error for BERT

  #10893 closed Aug 7, 2025

• 1 : Fail : Non-zero status code returned while running FusedConv node.

  #10894 closed Aug 7, 2025

• After using onnxruntime.transformers.optimizer to optimize onnx, the optimized model fail to tensorrt

  #10905 closed Aug 7, 2025

• TensorRT Execution [ONNXRuntimeError] : 6 : RUNTIME_EXCEPTION : Exception during initialization

  #10914 closed Aug 7, 2025

• slow fp16 performance

  #10919 closed Aug 7, 2025

• onnxruntime TensorRT Related Questions

  #10930 closed Aug 7, 2025

• MinGW support (MSYS2)

  #10976 closed Aug 7, 2025

• docker can't clone git repository for ARM64

  #10991 closed Aug 7, 2025

• Xor with broadcasting computes error

  #11000 closed Aug 7, 2025

• Inconsistent behavior between CPU and GPU on ReLU operator when input is NaN

  #11010 closed Aug 7, 2025

• 0xc00007b error, could not startup exe at all with onnxruntime1.7 win-x64 cpu on win10

  #11016 closed Aug 7, 2025

• Failed to build onnxruntime-vitisai docker container due to missing NO_PUBKEY

  #11017 closed Aug 7, 2025

• Huggingface Transformers Shape Inference Issue

  #11019 closed Aug 7, 2025

• kalid-onnxruntime Fatal error: Gemm is not a registered function/op

  #11021 closed Aug 7, 2025

• Updating state of the network

  #11026 closed Aug 7, 2025

• Can't constant fold SequenceEmpty node

  #11041 closed Aug 7, 2025

• Cuda EP parallelization issues for batches

  #11047 closed Aug 7, 2025

• C++ API, "tried creating tensor with negative value in shape" error when 'permute' and 'reshape' functions are used

  #11069 closed Aug 7, 2025

• Inference session creation freezes

  #11087 closed Aug 7, 2025

• compile with cuda error:Couldn't find CUDA library root.

  #11090 closed Aug 7, 2025

• Performance reduction due to copying of output OrtValues to numpy arrays

  #11099 closed Aug 7, 2025

• Using DnnlExecutionProvider for inference is much slower than using CPUExecutionProvider.

  #11122 closed Aug 7, 2025

• Different detection output values for C++ and Python with onnxruntime

  #11123 closed Aug 7, 2025

• docker container linux run onnxruntime infer core dumped

  #11135 closed Aug 7, 2025

• [question] yolov5-onnx-float16 not improve on GPU

  #11151 closed Aug 7, 2025

• How to use Flask with onnxruntime

  #11156 closed Aug 7, 2025

• Instruction level profiling in onnxruntime

  #11159 closed Aug 7, 2025

• No c++ header files for building custom op

  #11169 closed Aug 7, 2025

• A normal output of convolution layer multiplies infinity will result in NaN

  #11173 closed Aug 7, 2025

• Build from source issue on Windows

  #11178 closed Aug 7, 2025

• onnxruntime-web is 11-17x times slower than native inference

  #11181 closed Aug 7, 2025

• Custom Op does not support dynamic input/output number

  #11186 closed Aug 7, 2025

• Saving GPT2LMHeadModel_ConfigurableOneStepSearch error.

  #11198 closed Aug 7, 2025

• How to compress the sparse matrix in onnx model

  #11200 closed Aug 7, 2025

• Inference time for qunatized onnx models, TensorRT> CUDA> CPU. Is this expected?

  #11201 closed Aug 7, 2025

• auto_set_affinity can't be set to true for parallel executor

  #11205 closed Aug 7, 2025

• [web] ~100 seconds to load model/InferenceSession

  #11217 closed Aug 7, 2025

• NonZero shape inference behavior with scalar input mismatches ONNX and PyTorch

  #11232 closed Aug 7, 2025

• Unhandled exception at 0x00007FFABE6A9538 (cudnn_cnn_infer64_8.dll) in Onnx.exe

  #11235 closed Aug 7, 2025

• [React Native .ort Model Loading Error] "Error: Can't load a model: No content provider: ..."

  #11239 closed Aug 7, 2025

• I want use gpu on my jetson nx2 platform with c++, how should i do?

  #11240 closed Aug 7, 2025

• Non-zero status code returned while running Slice node. Name:'Slice_24' Status Message: slice.cc:153 FillVectorsFromInput Starts must be a 1-D array

  #11257 closed Aug 7, 2025

• Unsupported If operator in gradient builder for Hugging Face Transformers RoBERTa model

  #11268 closed Aug 7, 2025

• optimize_model : new model types

  #11270 closed Aug 7, 2025

• The onnx model of IMDN is slower than the original pytorch model and output many warnings

  #11274 closed Aug 7, 2025

• pulled master 1.12 quantization get unexpected result

  #11277 closed Aug 7, 2025

• LSTM export ONNX:Non-zero status code returned while running ScatterElements node. Name:'ScatterElements_880'

  #11278 closed Aug 7, 2025

• Why gpt2-xl (based transformer-xl) onnx slower than the originer pytorch

  #11293 closed Aug 7, 2025

• is the effect of onnx on Bert affected by python version?

  #11295 closed Aug 7, 2025

• TVM EP and TensorRT EP do not support dynamic inputs

  #11333 closed Aug 7, 2025

• MacOS M1 binary compilation and possibility to fine tune a model in C++

  #11343 closed Aug 7, 2025

• CUDAExecutionProvider optimized model adds incompatible node resulting in Failed to find kernel for MemcpyToHost

  #11348 closed Aug 7, 2025

• Lower performance on Inceptionv3/4 model with TensorRT EP than TensorRT directly

  #11356 closed Aug 7, 2025

• CUDAExecutionProvider not releasing memory after terminate session

  #11362 closed Aug 7, 2025

• ONNX Runtime compatibility for Jetson AGX Xavier

  #11378 closed Aug 7, 2025

• About running onnxruntime in singularity container

  #11397 closed Aug 7, 2025

• Benchmark code using torch.onnx.export

  #11399 closed Aug 7, 2025

• About building onnxruntime singularity container with DockerFile

  #11409 closed Aug 7, 2025

• Static quantization+per_channel is wrong for MobileNetV3

  #11415 closed Aug 7, 2025

• Can I quantize TreeEnsembleClassifier op?

  #11436 closed Aug 7, 2025

• Onnx T5 fp16 conversion without past_key_values

  #11438 closed Aug 7, 2025

• How to run a double input onnx model

  #11453 closed Aug 7, 2025

• InferenceSession giving different results than the original sklearn SVC model

  #11490 closed Aug 7, 2025

• C#, How to access the different output layer of inference (semantic segmentation)

  #11502 closed Aug 7, 2025

• [Documentation Request]

  #11505 closed Aug 7, 2025

• onnxruntime error

  #11509 closed Aug 7, 2025

• [Documentation Request] tensorAt for Csharp?

  #11510 closed Aug 7, 2025

• About Convolution Implementation

  #11517 closed Aug 7, 2025

• How to release a session properly?

  #11529 closed Aug 7, 2025

• Fail to convert model with reusable blocks

  #11530 closed Aug 7, 2025

• CPUExecutionProvider outputs wrong value for a quantized model

  #11532 closed Aug 7, 2025

• TensorRT EP session creation fails with invalid weights type of Int8 when ORT_TENSORRT_INT8_ENABLE set to 1

  #11535 closed Aug 7, 2025

• Using a model with float input types causes space issue

  #11541 closed Aug 7, 2025

• how to use c sharp to call libonnxruntime.dll? i build the onnxruntime dynamic dll, did it can be encapsulated c++ dll in order to c sharp called

  #11550 closed Aug 7, 2025

• can c sharp call onnxruntime c++ dll don't use c# third lib？ i create onnxruntime c++ project,but i want to call the dll with c sharp

  #11551 closed Aug 7, 2025

• T5-Large Export Results in ProtoBuf Error due to 2GB External Data when using padded inputs

  #11558 closed Aug 7, 2025

• CUDA failure 100: no CUDA-capable device is detected ; error when inferencing on a GPUVM

  #11561 closed Aug 7, 2025

• Specify CPUs to use for parallel inference when external CPU pinning is used

  #11563 closed Aug 7, 2025

• [js/web] Inference is Broken in Safari when Cross Origin Isolation is active

  #11567 closed Aug 7, 2025

• Header missmatch C/C++ - mac

  #11570 closed Aug 7, 2025

• The effect of turning optimization on and off on quantized model performance

  #11576 closed Aug 7, 2025

• ONNXRUNTIME + OpenVINO on ARM64

  #11582 closed Aug 7, 2025

• did it can build onnxruntime with any cuda version by source code ? is not relate to onnxtuntime version?

  #11584 closed Aug 7, 2025

• cpu and gpu results is not the same

  #11590 closed Aug 7, 2025

• CUDNN failure 4: CUDNN_STATUS_INTERNAL_ERROR ; error when inferencing on a GPUVM

  #11592 closed Aug 7, 2025

• issues with pybind11 repository while installing

  #11595 closed Aug 7, 2025

• Bad performance for QDQ model with openvino EP

  #11604 closed Aug 7, 2025

• Unable to build onnxruntime_v1.10.0 C++ api with --enable_memory_profile --enable_cuda_profiling flags

  #11607 closed Aug 7, 2025

• Shape inference fails

  #11614 closed Aug 7, 2025

• building ——libonnxruntime_providers_cuda.so Error running link command: No such file or directory

  #11621 closed Aug 7, 2025

• how to set providers with onnx runtime-gpu1.70 ?

  #11624 closed Aug 7, 2025

• using multithread to call onnxruntime inference,

  #11628 closed Aug 7, 2025

• which tags should i download of onnxruntime-gpu 1.6 for c#

  #11646 closed Aug 7, 2025

• build for c#

  #11648 closed Aug 7, 2025

• output shape can not be specified in com.microsoft::GridSample op

  #11652 closed Aug 7, 2025

• Installing ORTModule torch extension reports TypeError

  #11663 closed Aug 7, 2025

• when set inter_op_num=0 with ORT_PARALLEL model the performance is very bad than inter_op_num=1?

  #11668 closed Aug 7, 2025

• How to implement a new operator inference function?

  #11678 closed Aug 7, 2025

• [web] `ort.InferenceSession.create` silently hangs/fails on iOS/iPad browsers if COEP/COOP headers are set

  #11679 closed Aug 7, 2025

• which onnxruntime-gpu version is compatible for CUDA 11.1 ?

  #11685 closed Aug 7, 2025

• Real-ESRGAN slow onnxruntime inference compared to Pytorch one

  #11688 closed Aug 7, 2025

• Linux CI pipelines can't test unreleased versions of ONNX

  #11693 closed Aug 7, 2025

• Dynamic quantization of Albert model

  #11701 closed Aug 7, 2025

• Low level profiling for onnxrt Conv kernel(default backend)

  #11702 closed Aug 7, 2025

• CUDA EP spending lots of time idling

  #11706 closed Aug 7, 2025

• Race condition when setting do_copy_in_default_stream to false

  #11713 closed Aug 7, 2025

• Reading back multidimensional output in C++

  #11718 closed Aug 7, 2025

• how to get the remaining GPU memory to get the batch size?

  #11735 closed Aug 7, 2025

• ssd_mobilenet_v1 infer error for TensorRT Execution Provider

  #11736 closed Aug 7, 2025

• build rknpu backend error

  #11738 closed Aug 7, 2025

• Pip installed Transformer Benchmark cannot run on TF

  #11751 closed Aug 7, 2025

• Converted ONNX model works in Python but not in C++

  #11761 closed Aug 7, 2025

• Failed to build onnxruntime on Apple Sillion

  #11805 closed Aug 7, 2025

• I do not get any performance improvement after using TensorRT provider for object detection model

  #11806 closed Aug 7, 2025

• When I use onnxruntime to run onnx model on GPU, it sucks up too much video memory. Is that normal?

  #11809 closed Aug 7, 2025

• Issue importing onnxruntime

  #11815 closed Aug 7, 2025

• [Bug] Mixing negative and positive paddings causes segfault/uninitialized memory values produced in reflected pad

  #11828 closed Aug 7, 2025

• in cmake/CMakeList.txt all avx related option all set off, do we need do anything to use avx features?

  #11833 closed Aug 7, 2025

• [ONNXRuntimeError] FuseReluClip failure

  #11836 closed Aug 7, 2025

• Incompatible dimensions for matrix multiplication Error in StarNet model when doing InferenceSession

  #11846 closed Aug 7, 2025

• What's the meaning of the hole of tracing file

  #11850 closed Aug 7, 2025

• How to use batch run？

  #11852 closed Aug 7, 2025

• Use NPU in NXP iMX8MP?

  #11854 closed Aug 7, 2025

• What is the meaning of src_arg_index and dst_arg_index in EdgeEndToMatch structure?

  #11856 closed Aug 7, 2025

• Wrong output shape due to MergeShape failure

  #11870 closed Aug 7, 2025

• Not clear quantization pipeline for tensorrt ep

  #11873 closed Aug 7, 2025

• Pytorch -> Onnx custom Yolov5 model works in python but not in JS

  #11874 closed Aug 7, 2025

• [ONNXRuntimeError] Load model from *** failed: Unsuported type proto value case

  #11889 closed Aug 7, 2025

• Quantize specific ops per-tensor while per_channel=True

  #11890 closed Aug 7, 2025

• onnx and onnxruntime disagree on input with no known rank

  #11891 closed Aug 7, 2025

• Bug: MatMul fails for input shapes of [0, k] and [k, ]

  #11895 closed Aug 7, 2025

• Immense GPU memory consumption

  #11903 closed Aug 7, 2025

• ConvTranspose with auto_pad attribute

  #11927 closed Aug 7, 2025

• how to get inference time with c# onnxruntime-gpu-1.6.0

  #11946 closed Aug 7, 2025

• excute dnnl provider error

  #11947 closed Aug 7, 2025

• windows11+onnxruntime1.8.0+vs2019 inferencing crash

  #11950 closed Aug 7, 2025

• Multi thread of single session Python vs C++ (end with core dumped)

  #11951 closed Aug 7, 2025

• Inference_GPT2-OneStepSearch_OnnxRuntime_CPU.ipynb Error

  #11959 closed Aug 7, 2025

• Question about quantize Gemm OP

  #11961 closed Aug 7, 2025

• Got segmentation fault error when using 'InferenceSession' API

  #11964 closed Aug 7, 2025

• how to configure lobal/shared threadpool with multithread, in c#API?

  #11966 closed Aug 7, 2025

• set gpu option failed

  #11967 closed Aug 7, 2025

• quant onnx model slower than pytorch with mish6 activation, howerver faster with relu6

  #11975 closed Aug 7, 2025

• inference time is not stable

  #11983 closed Aug 7, 2025

• Any interest in hosting the Rust bindings

  #11992 closed Aug 7, 2025

• inference is different on linux and windows

  #11993 closed Aug 7, 2025

• Inconsistent result to NumPy and PyTorch when consecutively casting a float tensor to int32 and then to bool

  #11994 closed Aug 7, 2025

• failed to initialize a session in the GPU environment

  #11996 closed Aug 7, 2025

• The test time of sess.run does not match the time of profile

  #11997 closed Aug 7, 2025

• build C#api with cuda 11.0 /cudnn 8.0

  #11999 closed Aug 7, 2025

• Issue with NeMo MTEncDecModel model in ONNX IOBinding

  #12003 closed Aug 7, 2025

• how to build onnxruntime from source with dnnl?

  #12011 closed Aug 7, 2025

• create op

  #12017 closed Aug 7, 2025

• Resize with mode linear is missing output elements

  #12019 closed Aug 7, 2025

• Microsoft.ML.OnnxRuntime.Tests.InferenceTest.TestPreTrainedModels should get opset version from the model file

  #12040 closed Aug 7, 2025

• Builds C# bindings and creates nuget package

  #12042 closed Aug 7, 2025

• GlobalAveragePool on large size of ones miscalculates

  #12043 closed Aug 7, 2025

• Using onnxruntime server for model deployment

  #12044 closed Aug 7, 2025

• Support pasts as inputs in gpt2 beam search operator

  #12047 closed Aug 7, 2025

• Build wasm static library bug because of missing `testdata` folder.

  #12048 closed Aug 7, 2025

• Performance in parallel session Run()

  #12049 closed Aug 7, 2025

• Builds C# bindings and creates nuget package for vs2019 install

  #12061 closed Aug 7, 2025

• ONNXRuntimeError for "Where" node when the input is too long

  #12065 closed Aug 7, 2025

• Performance issue with beam search in onnxruntime

  #12078 closed Aug 7, 2025

• Support for cmake's FetchContent()

  #12081 closed Aug 7, 2025

• TensorRT Provider Vs TensorRT Native

  #12083 closed Aug 7, 2025

• Resize with mode linear always produces 0.5 on GPU regardless of the input

  #12091 closed Aug 7, 2025

• Resize with `nearest` mode have inconsistent results compared to PyTorch and TVM

  #12098 closed Aug 7, 2025

• onnxruntime tensorrt sometime cost verg log time

  #12120 closed Aug 7, 2025

• How do I call the same model in CUDA with many various inputs?

  #12126 closed Aug 7, 2025

• Error in symbloc_shape_infer.py: assert name in self.sympy_data_ or ...

  #12127 closed Aug 7, 2025

• Inference time vs torch w/regard to batch_size and BatchNorm

  #12130 closed Aug 7, 2025

• When will Attention OP extra_add_qk input support automatic broadcast

  #12149 closed Aug 7, 2025

• Query regarding timings under ONNXRT profiler

  #12150 closed Aug 7, 2025

• Hi Does ONNX Runtime support FP16 and INT8 inference on Intel OneDNN ExecutionProvider?

  #12160 closed Aug 7, 2025

• Eager mode generator support non-tensor return types

  #12163 closed Aug 7, 2025

• symbolic_shape_infer.py not working with models quantized with 🤗 Optimum for TensorRT

  #12173 closed Aug 7, 2025

• upgrading pip and wheels kills CUDAExecutionProvider

  #12185 closed Aug 7, 2025

• why first session.run is too slower than after

  #12197 closed Aug 7, 2025

• Performance issue of ConvInteger

  #12206 closed Aug 7, 2025

• How to release memory after Inference session run in Python

  #12207 closed Aug 7, 2025

• Regarding the dynamism for custom op in ONNXRT

  #12211 closed Aug 7, 2025

• Quantized Model Running Slow Using Cuda as EP

  #12229 closed Aug 7, 2025

• Exported beam search model consumes a lot of more memory

  #12246 closed Aug 7, 2025

• Mismatch in the order of the column names in the benchmarking script for transformer models

  #12265 closed Aug 7, 2025

• LoadLibrary failed with error 126 (DirectML)

  #12269 closed Aug 7, 2025

• TRT EP failed to create model session with CUDA custom op

  #12282 closed Aug 7, 2025

• Since ORT 1.12 ort.InferenceSession throws error when the last provider is not capable

  #12287 closed Aug 7, 2025

• Resize op can't work well under Cubic mode with ORT 1.12.

  #12302 closed Aug 7, 2025

• Details regarding ONNXRuntime inference with OpenVino Backend

  #12305 closed Aug 7, 2025

• Why the performance of onednn is worse than the common version

  #12315 closed Aug 7, 2025

• ONNXRT default CPU EP vs Openvino EP Performance

  #12316 closed Aug 7, 2025

• onnx graph partition optimize

  #12318 closed Aug 7, 2025

• Wrong native library directory name for M1 Mac in the Java package

  #12324 closed Aug 7, 2025

• MetaCommand exception from DirectML EP

  #12328 closed Aug 7, 2025

• window10 ort with openvino backend error

  #12334 closed Aug 7, 2025

• unsafe exception code in C++ API, wrongly declaring exceptions, incomplete constructors

  #12338 closed Aug 7, 2025

• Unable to build Onnxruntime 1.12.0 with OpenVINO 2020.3 on Windows 10

  #12342 closed Aug 7, 2025

• Quantized ONNX model output

  #12346 closed Aug 7, 2025

• Performance gains by ONNX inconsistent

  #12348 closed Aug 7, 2025

• Integer quantization fails on Transformer-based vision model

  #12362 closed Aug 7, 2025

• Setting Openvino EP to run on one core with one thread

  #12365 closed Aug 7, 2025

• Unable to build tensorrt docker image

  #12373 closed Aug 7, 2025

• Accept dictionary of tensor as input (python api)

  #12380 closed Aug 7, 2025

• Fail to build onnxRT with oneDNN using official build command

  #12382 closed Aug 7, 2025

• Segmentation fault

  #12386 closed Aug 7, 2025

• While loading the onnx file with InferenceSession getting session ID 11 error

  #12402 closed Aug 7, 2025

• Failed to build with ACL(and ARMnn)

  #12407 closed Aug 7, 2025

• Can't build with OpenVINO 2022.1 ("onnxruntime_providers_shared" does not exist)

  #12411 closed Aug 7, 2025

• `Env(OrtLoggingLevel, const char* logid, OrtLoggingFunction, ...` fails to pass `logid` param to log function

  #12414 closed Aug 7, 2025

• CUDA support for longer-input models like BigBird

  #12463 closed Aug 7, 2025

• I found that the OnnxRuntime used almost all of the instruction sets for the convolutional computations and I wanted to optimize for that

  #12479 closed Aug 7, 2025

• How to exit abnormally in the Python Operator (PyOp)

  #12481 closed Aug 7, 2025

• QDQ + Add nodes are not fused into QLinearAdd when the graph is optimized

  #12487 closed Aug 7, 2025

• performance is poor when onnxruntime C++ run in intel cpu

  #12489 closed Aug 7, 2025

• LSTM Y output is inconsistent with TF inference result when seq_len is effective

  #12492 closed Aug 7, 2025

• Clarify NMS sorting strategy

  #12493 closed Aug 7, 2025

• Attributes in nested function calls are zeroed out

  #12506 closed Aug 7, 2025

• Computing loss within onnxrunitme inference (GPT2 model)

  #12526 closed Aug 7, 2025

• java deploy in k8s Failed to load library libonnxruntime_providers_cuda.so with error

  #12540 closed Aug 7, 2025

• engine decryption does not work in TensorRT EP

  #12551 closed Aug 7, 2025

• Add execution provider selection for quantize_static

  #12573 closed Aug 7, 2025

• Document beamsearch

  #12584 closed Aug 7, 2025

• Name:'MatMul_32007' Status Message: matmul_helper.h:61 Compute MatMul dimension mismatch

  #12594 closed Aug 7, 2025

• Run the onnx model converted from seq2seq and report an error

  #12608 closed Aug 7, 2025

• Where is the definition of session.Run() in onnxruntime C++ api

  #12623 closed Aug 7, 2025

• cuda_provider_options.h include non existing file?

  #12636 closed Aug 7, 2025

• when the model support dynamic batch ，the input shape [ -1,-1,80], how can warm up? because of the dynamic batch , I do know the warmup batchsize number ,can it use min_batchsize and max _batchsize to warmup?

  #12637 closed Aug 7, 2025

• The quantization model reduces the accuracy compared to the TRT

  #12638 closed Aug 7, 2025

• Failed to create TensorrtExecutionProvider using onnxruntime-gpu

  #12639 closed Aug 7, 2025

• Confusing exception about supported types

  #12648 closed Aug 7, 2025

• get kill signal when quantize the ONNX model using quantize_static

  #12652 closed Aug 7, 2025

• Enable Global Shared Threadpool and Memory Allocator For C#

  #12654 closed Aug 7, 2025

• Non-zero status code returned while running TopK node. (ssdlite320_mobilenet_v3_large)

  #12669 closed Aug 7, 2025

• Wrong Results for FP16 Models in CUDAExecutionProvider and TensorRTExecutionProvider

  #12726 closed Aug 7, 2025

• `static inline Ort::Env onnx_env{nullptr}` easily leads to nullptr deref on app exit

  #12736 closed Aug 7, 2025

• SystemError : 13 for transformers optimizer

  #12745 closed Aug 7, 2025

• BatchNormalization produces all zeros for 1D input

  #12754 closed Aug 7, 2025

• How to set the priority of ONNX in GPU?

  #12760 closed Aug 7, 2025

• onnxruntime-linux-x64-gpu-1.12.1

  #12766 closed Aug 7, 2025

• Asynchrononus Inference

  #12768 closed Aug 7, 2025

• I want to use tensorrt as the back-end of onnx

  #12781 closed Aug 7, 2025

• onnxruntime.capi.onnxruntime_pybind11_state.Fail: [ONNXRuntimeError] : 1 : FAIL : CUDA error executing cudaSetDevice(GetDeviceId())

  #12785 closed Aug 7, 2025

• cast op not support multithread

  #12786 closed Aug 7, 2025

• How to set cpu_num to a specific value?

  #12819 closed Aug 7, 2025

• AttentionPastState_dynamic test fails during building with CUDA EP from source

  #12820 closed Aug 7, 2025

• Memory management

  #12824 closed Aug 7, 2025

• error: package directory 'onnxruntime/backend' does not exist [Build]

  #12922 closed Aug 7, 2025

• [Web] Failed to compile shader on WebGL

  #12927 closed Aug 7, 2025

• Disabling optimization produces incorrect results on CUDAExecutionProvider in 1.12

  #12946 closed Aug 7, 2025

• [Performance] Dynamic model input prediction is slow

  #12955 closed Aug 7, 2025

• Why is there not ParallelExecutionPlan like SequentialExecutionPlan in the ParallelExecutor of onnxruntime?

  #13036 closed Aug 7, 2025

• onnxruntime calculate gradients but no need for training

  #13057 closed Aug 7, 2025

• onnxruntime-gpu, cudaoptions, result is different

  #13061 closed Aug 7, 2025

• onnxruntime-node crash the electron app[Web]

  #13086 closed Aug 7, 2025

• what's the differences between onnxruntime with openvino backend VS openvino directly?

  #13087 closed Aug 7, 2025

• [Performance] a problem for Ort::IoBinding

  #13090 closed Aug 7, 2025

• [Performance] ONNX Runtime GPT2 Model Running Significantly Slower than PyTorch

  #13105 closed Aug 7, 2025

• [Test issue] Updated Ignore

  #13109 closed Aug 7, 2025

• [Performance] Multithreading performance tails off after 3 threads, possible memory issue

  #13138 closed Aug 7, 2025

• Failed to create CUDAExecutionProvider

  #13139 closed Aug 7, 2025

• Onnxruntime fails on GPU loading inference with int8 models

  #13168 closed Aug 7, 2025

• Multilingual-MiniLM-L12-H384 ONNX inference in NodeJS

  #13171 closed Aug 7, 2025

• GPU inference result not stable

  #13178 closed Aug 7, 2025

• [Performance] inference time much slower (1529ms vs. 20 ms) on GPU vs CPU.

  #13199 closed Aug 7, 2025

• [Performance] Performance issue on Linux vs Windows for BERT model.

  #13224 closed Aug 7, 2025

• Contrib IRFFT operator output dimensions calculation

  #13236 closed Aug 7, 2025

• Onnx create session takes a long time.

  #13240 closed Aug 7, 2025

• Inference time spikes in UNET onnx

  #13258 closed Aug 7, 2025

• [Performance] Too Slow when i do inference

  #13265 closed Aug 7, 2025

• [Mobile] .Net target Arm64

  #13295 closed Aug 7, 2025

• [ONNXRuntimeError] : 1 : FAIL : This is an invalid model. Error: the graph is not acyclic.

  #13322 closed Aug 7, 2025

• onnx Pad operator with negative pads value outputs 'nan'

  #13332 closed Aug 7, 2025

• [Build] Upgrade to latest protobuf

  #13335 closed Aug 7, 2025

• [Performance] Comparing ONNX CPU execution profiles of two FasterRCNN checkpoints

  #13341 closed Aug 7, 2025

• [Build] ONNX Runtime Build Error ZCU102 (DPUCZDX8G)

  #13351 closed Aug 7, 2025

• quantize_dynamic results in initializer error

  #13358 closed Aug 7, 2025

• [Performance] CNN Inference has latency spikes with TensorRT EP

  #13366 closed Aug 7, 2025

• Onnxruntime crashes if setting cpu affinity fails in Ort::Session constructor

  #13367 closed Aug 7, 2025

• Using GPU in c++

  #13380 closed Aug 7, 2025

• Can't run qdq model with TRT EP

  #13381 closed Aug 7, 2025

• Whether the .trt model can be loaded

  #13394 closed Aug 7, 2025

• Does ORT support quantize

  #13413 closed Aug 7, 2025

• ONNX Runtime Inference on GPU: Failed to create CUDAExecutionProvider

  #13414 closed Aug 7, 2025

• Consecutive casting leads to wrong result

  #13418 closed Aug 7, 2025

• Parameters are optimized out even if it is a needed return value

  #13425 closed Aug 7, 2025

• [Web] Is it possible to use both webgl backend and wasm backend in onnxruntime-web

  #13435 closed Aug 7, 2025

• run_with_iobinding is not outputting the expected result for batched input data for T5 model running on ort CUDA EP

  #13463 closed Aug 7, 2025

• GPU Arena blocked session->Run()

  #13464 closed Aug 7, 2025

• Consecutive call to Ort::Session::Run() crashes

  #13476 closed Aug 7, 2025

• did onnxruntime-gpu surport call CUDA code or call custom kernel funtion to preprocess Image?

  #13491 closed Aug 7, 2025

• [Performance]

  #13492 closed Aug 7, 2025

• ORT fails on Slice() when indices are of different integer types

  #13497 closed Aug 7, 2025

• Init provider bridge failed when put onnxruntime folder under path which contains other Unicode character

  #13499 closed Aug 7, 2025

• [Performance]

  #13500 closed Aug 7, 2025

• [Performance] C# Gpu memory allocation

  #13504 closed Aug 7, 2025

• Removing the semantic segmentation's bounding box

  #13513 closed Aug 7, 2025

• How to transfer the Ort::Value obtained to cuda code for post-processing, such as a .cu file?

  #13528 closed Aug 7, 2025

• [Training] Whether onnxruntime training can be used in Megatron.

  #13532 closed Aug 7, 2025

• How can I load a model larger than 2G in memory

  #13543 closed Aug 7, 2025

• Zero Result with DirectML Execution Provider

  #13545 closed Aug 7, 2025

• Inference speed: Swintransformer torch vs onnxruntime-gpu

  #13550 closed Aug 7, 2025

• [Build]

  #13554 closed Aug 7, 2025

• ORT fails on CPU looking for LayerNormalization node, for mixed-precision ONNX

  #13556 closed Aug 7, 2025

• [TVM] Exception during initialization

  #13572 closed Aug 7, 2025

• unable to build onnxruntime for openvino execution provider to get nuget packages

  #13577 closed Aug 7, 2025

• Does Microsoft.ML.OnnxRuntime have a dependency on System.CodeDom.dll ?

  #13604 closed Aug 7, 2025

• [Build]

  #13606 closed Aug 7, 2025

• [C++] Model output image different in C++ ORT vs. Python ORT & PyTorch

  #13614 closed Aug 7, 2025

• [Performance] Operators assigned to CPU instead of CUDA

  #13615 closed Aug 7, 2025

• Dimension Padding problem in reduction_ops.cc

  #13654 closed Aug 7, 2025

• [Performance] onnxruntime session uses 5x more system memory if torch is imported

  #13662 closed Aug 7, 2025

• GPT2 Static Quantization Failed. Non-zero status code returned while running Reshape node. Name:'past_0_ReduceMax_Reshape'

  #13667 closed Aug 7, 2025

• Help in running onnxruntime with SNPE as execution provider

  #13693 closed Aug 7, 2025

• GPU with device_id=0 is always occupied no matter what device_id is specified when run the inference

  #13697 closed Aug 7, 2025

• onnxruntime-gpu get warning "Serializing optimized model with Graph Optimization level greater than ORT_ENABLE_EXTENDED and the NchwcTransformer enabled".

  #13709 closed Aug 7, 2025

• [DML] reproducible bug on DML provider

  #13714 closed Aug 7, 2025

• [Build] Avoid NEON when building on Raspberry Pi 4

  #13718 closed Aug 7, 2025

• [Web] Uncaught (in promise) TypeError: cannot resolve operator 'Erf' with opsets: ai.onnx v15

  #13729 closed Aug 7, 2025

• [Web] NPM package include ts files in the output

  #13736 closed Aug 7, 2025

• [Web]

  #13749 closed Aug 7, 2025

• Cannot run inference on Integrated Graphics with OpenVino EP using C Sharp API

  #13772 closed Aug 7, 2025

• QDQ not instrumenting inputs if first operator is a SUM

  #13794 closed Aug 7, 2025

• how bring my hardware backend to onnxruntime framework

  #13797 closed Aug 7, 2025

• [WebGL] cannot resolve operator 'DynamicQuantizeLinear' with opsets: ai.onnx v16, ...

  #13800 closed Aug 7, 2025

• [CPUExecutionProvider] PyTorch/Numpy operations following InferenceSession.run() are 50x slower compared to using dummy inputs

  #13808 closed Aug 7, 2025

• hello,how to improve [Performance] in batch inference with multicore cpu

  #13820 closed Aug 7, 2025

• Dynamic quantization is useless on AMD cpus(AMD EPYC 7K62 48-Core Processor)

  #13872 closed Aug 7, 2025

• SSDLite 320: RuntimeException on CUDA. TopK index assert was false.

  #13876 closed Aug 7, 2025

• Segmentation Faults when using TensorRT on Jetson Orin Dev Kit

  #13877 closed Aug 7, 2025

• Model run with `TensorrtExecutionProvider` outputs different results compared to `CPUExecutionProvider` / `CUDAExecutionProvider` when the ONNX `Loop` operator is used

  #13894 closed Aug 7, 2025

• [Web] dynamic batch size doesn't work when use webgl provider

  #13909 closed Aug 7, 2025

• [Web] ort-wasm-simd.wasm can't be loaded in Electron renderer (using webpack)

  #13933 closed Aug 7, 2025

• [Build] Incomplete type used in nested name specifier, Ubuntu

  #13942 closed Aug 7, 2025

• Do I need to convert data to device for TensorRTExecutionProvider?

  #13952 closed Aug 7, 2025

• CUDA provider gives different result with respect to CPU

  #13962 closed Aug 7, 2025

• bug: onnxruntime.capi.onnxruntime_pybind11_state.RuntimeException: [ONNXRuntimeError] : 6 : RUNTIME_EXCEPTION : Non-zero status code returned while running OpenVINO-EP-subgraph_4 node.

  #13973 closed Aug 7, 2025

• Non-zero status code returned while running Resize node

  #13975 closed Aug 7, 2025

• Java Problematic Frame [libonnxruntime.dylib+0x8212be] onnxruntime::DataTypeImpl::ToString(onnxruntime::DataTypeImpl const*)+0xe

  #13976 closed Aug 7, 2025

• [Performance] [webgl]bad performance of webgl

  #13986 closed Aug 7, 2025

• [windows7] Unable to load DLL 'onnxruntime.dll': The specified module could not be found.

  #14003 closed Aug 7, 2025

• [Performance] CUDA EP with Strange Inference Time

  #14016 closed Aug 7, 2025

• [Performance] the speed and cpu utilization with SetIntraOpNumThreads(1) and SetIntraOpNumThreads(2)

  #14018 closed Aug 7, 2025

• [Performance] onnx vs pt memory usage

  #14029 closed Aug 7, 2025

• [Performance] High memory use by CUDAProvider in Jetson Xavier NX(JetPack 4.4)

  #14038 closed Aug 7, 2025

• java onnxruntime_providers_cuda.dll

  #14047 closed Aug 7, 2025

• There is a vulnerability in torch:1.12.0,upgrade recommended

  #14059 closed Aug 7, 2025

• [Build] impossible to build onnxruntime with vs2022

  #14086 closed Aug 7, 2025

• [Build] core/framework/fence.h not found while build upon CANN

  #14121 closed Aug 7, 2025

• 300% slower on MYRIAD_FP16 when using CustomVision fp16 model

  #14125 closed Aug 7, 2025

• [Training] Does the current training code support RNN model like seq2seq and Transformer and GNN model?

  #14139 closed Aug 7, 2025

• [Build] Dockerfile.arm64 - No module named 'packaging' error

  #14140 closed Aug 7, 2025

• CUDNN error executing cudnnConvolutionForward

  #14186 closed Aug 7, 2025

• ONNXRuntime outputs numerically incorrect results for mixed precision models.

  #14189 closed Aug 7, 2025

• Infer shape incorrect for Split with opset 15

  #14200 closed Aug 7, 2025

• ConvTranspose2d onnxruntime and pytorch forward results are inconsistent

  #14208 closed Aug 7, 2025

• `onnxruntime.quantization` does not support `.onnx` files produced by `tf2onnx.convert.from_function` with the `large_model` option set to `True`

  #14213 closed Aug 7, 2025

• No module named 'onnxruntime.transformers.io_binding_helper'

  #14230 closed Aug 7, 2025

• Valgrind: Source and destination overlap in memcpy_chk

  #14254 closed Aug 7, 2025

• [ONNXRuntimeError] : 6 : RUNTIME_EXCEPTION : Non-zero status code returned while running OpenVINO-EP-subgraph_3 node. Name:'OpenVINOExecutionProvider_OpenVINO-EP-subgraph_3_0'

  #14280 closed Aug 7, 2025

• [Build] Docker arm64 build fails.

  #14283 closed Aug 7, 2025

• ONNX Runtime support for the graph optimization of bigbird_pegasus model

  #14295 closed Aug 7, 2025

• TensorRT EP same inference Time of INT 8 and FP 16

  #14315 closed Aug 7, 2025

• STFT op has the wrong expected shape

  #14316 closed Aug 7, 2025

• Program will stuck when creating 'Ort::Session'

  #14317 closed Aug 7, 2025

• [Performance] ONNXruntime CPU is slower than Pytorch Tracing to Torchscript on CPU

  #14326 closed Aug 7, 2025

• RemoveNode Should be unreachable if CanRemoveNodeAndMergeEdges is in sync with the logic

  #14360 closed Aug 7, 2025

• [Bug] Attention and QAttention don't work properly in some cases

  #14363 closed Aug 7, 2025

• Add some custom QlinearXXX Ops

  #14365 closed Aug 7, 2025

• [Build] Error in builiding with Tensorrt EP

  #14394 closed Aug 7, 2025

• Import Error " cannot import name 'get_all_providers' "

  #14395 closed Aug 7, 2025

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

  #14412 closed Aug 7, 2025

• Free allocated data of Ort::Value in C++

  #14420 closed Aug 7, 2025

• Pad operator not quantizable?

  #14422 closed Aug 7, 2025

• Different Python exceptions on OOM with `run_with_iobinding` and `run`

  #14438 closed Aug 7, 2025

• Modifying QlinearADD

  #14441 closed Aug 7, 2025

• [ONNXRuntimeError] Unsupported OrtValue type with CUDA EP

  #14457 closed Aug 7, 2025

• [Performance] There is some confusion with onnx + oneDNN or onnx + OpenVINO

  #14468 closed Aug 7, 2025

• [Build]

  #14471 closed Aug 7, 2025

• [Performance] cuda_options.arena_extend_strategy = 1 does not free memory

  #14474 closed Aug 7, 2025

• [Performance] DirectML cost more memory than CPU when process the Win32(X86) program (official demo).

  #14479 closed Aug 7, 2025

• [Performance] CPU Usage is too high

  #14490 closed Aug 7, 2025

• [Performance] cuDNN lib mismatch let to a underutilization of GPU

  #14498 closed Aug 7, 2025

• missing headers and pkgconfig files in binary packages distribution (from github releases) (linux)

  #14503 closed Aug 7, 2025

• [Web] Runtime error using `onnxruntime-node` with webpack

  #14505 closed Aug 7, 2025

• [Performance] Find out why the GPU memory allocated with `CUDAExecutionProvider` is much larger than the ONNX size

  #14526 closed Aug 7, 2025

• Non-zero status code returned while running DnnlCustomOp2 node

  #14543 closed Aug 7, 2025

• Check and modify the weights of a layer of an onnx model at runtime

  #14545 closed Aug 7, 2025

• [Performance] DirectML Dynamic Axes very slow

  #14550 closed Aug 7, 2025

• [BUG] FusedConv node error

  #14561 closed Aug 7, 2025

• fp32 model with autocast to fp16: Shape mismatch attempting to re-use buffer

  #14582 closed Aug 7, 2025

• [Build] cuda dll wrap up

  #14585 closed Aug 7, 2025

• different results with onnxruntime-gpu-1.10

  #14587 closed Aug 7, 2025

• [Web] currently non-1 steps is not supported for Slice

  #14588 closed Aug 7, 2025

• Destroying an inference session without exiting the python process

  #14590 closed Aug 7, 2025

• C# - CUDA Nuget BUG : DefaultLogger Attempt to use DefaultLogger but none has been registered.

  #14593 closed Aug 7, 2025

• Onnxruntime Arm NN Ep build error.

  #14611 closed Aug 7, 2025

• [Performance]

  #14615 closed Aug 7, 2025

• [Build] cpp_field.h(189,47): error C2059: 语法错误:“)”

  #14627 closed Aug 7, 2025

• [Performance] Memory grows after reloading model

  #14641 closed Aug 7, 2025

• [Build] Building for C++ On Jetson Nano CUDA 10.2

  #14644 closed Aug 7, 2025

• TensorRT Execution Build Fails on Jetson Jetpack 4.6.1

  #14658 closed Aug 7, 2025

• DEEPFACE LIVE Issue with onnxruntime_pybind_state.

  #14667 closed Aug 7, 2025

• [Build]

  #14674 closed Aug 7, 2025

• Custom Operater Output Tensor Shape Error

  #14683 closed Aug 7, 2025

• [Web] Inference speed halves if you open DevTools after loading an inference session, even if you close DevTools afterwards

  #14692 closed Aug 7, 2025

• `CleanUnusedInitializersAndNodeArgs` warnings are printed only with subgraphs

  #14694 closed Aug 7, 2025

• [Performance]why is the inference latency of onnx QDQ quantized model converted from tflite quantized model (or from tensorflow Quantization-Aware training (QAT) model) as same as normal onnx float32 model?

  #14707 closed Aug 7, 2025

• A runtime can run on cuda device 0 but fail on cuda device 1

  #14710 closed Aug 7, 2025

• Non-zero status code returned while running Reshape node. Name:'Reshape_7411' The input tensor cannot be reshaped to the requested shape. Input shape:{51}, requested shape:{}

  #14712 closed Aug 7, 2025

• How to inference with multiple batches and multiple inputs.

  #14713 closed Aug 7, 2025

• Crash in JavaGPU on Windows

  #14714 closed Aug 7, 2025

• The Microsoft.ML.OnnxRuntime.Gpu nuget on Visual studio latest version 1.14.0 has a bug when running with the tensorrt on run time.

  #14730 closed Aug 7, 2025

• clog_vlog_fatal[Build]

  #14740 closed Aug 7, 2025

• [Performance] How to create multiple tensors with consecutive addresses when the cuda memory is not occupied?

  #14742 closed Aug 7, 2025

• Memory Leak

  #14745 closed Aug 7, 2025

• [Build] macOS: cross compiling arm64 on intel fails

  #14746 closed Aug 7, 2025

• [Performance] Can oneDNN EP accelerate the inference time of onnxruntime on x86 machines?

  #14749 closed Aug 7, 2025

• Basic Optimizer adds non-standard ONNX ops

  #14752 closed Aug 7, 2025

• Basic Optimizer adds non-standard ONNX ops for roi_align

  #14753 closed Aug 7, 2025

• Basic Optimizer adds non-standard ONNX ops for input tensor

  #14754 closed Aug 7, 2025

• [Build] cmake install when --use_xnnpack is broken

  #14757 closed Aug 7, 2025

• Failed to build CUDA docker image[Build]

  #14765 closed Aug 7, 2025

• `onnx.checker.check_model` raises `Bad node spec` for custom nodes created from ORT `optimize_model`

  #14768 closed Aug 7, 2025

• Dependency Problem (java onnxruntime)

  #14787 closed Aug 7, 2025

• [Build] Can't access OrtSessionOptionsAppendExecutionProvider_Dnnl while using oneDNN

  #14799 closed Aug 7, 2025

• [Build] Dockerfile.arm64 build fails

  #14801 closed Aug 7, 2025

• [Build] Unable to load TensorRT Execution Provider

  #14802 closed Aug 7, 2025

• Read access violation under OnnxRuntimeCpuSessionBuilder::Initialize during WinML operator tests for function operators

  #14810 closed Aug 7, 2025

• [Web] how to reduce wasm file size

  #14817 closed Aug 7, 2025

• onnxruntime with CUDA not releasing about 400 MB memory after the session and environment is destroyed

  #14819 closed Aug 7, 2025

• working model with Resize node becomes invalid after using convert_float_to_float16

  #14827 closed Aug 7, 2025

• How do I pass a list of tensors in onnxruntime-web?

  #14829 closed Aug 7, 2025

• DML EP cannot load some quantized onnx files.

  #14835 closed Aug 7, 2025

• [Performance] Performance degradation while using dynamic axes

  #14863 closed Aug 7, 2025

• UndefinedBehaviorSanitizer reports problem in onnxruntime_global_thread_pools_test

  #14882 closed Aug 7, 2025

• [Build] Error APPX1101 - Payload contains two or more files with the same destination path 'microsoft.ai.machinelearning.dll'

  #14915 closed Aug 7, 2025

• [Performance]

  #14919 closed Aug 7, 2025

• Is there a Python way to get the max supported ONNX IR version from ORT package?

  #14932 closed Aug 7, 2025

• [Performance] 3-100x regression when opset 16 or 17 is used (CUDA EP)

  #14956 closed Aug 7, 2025

• [Performance] Can not release memory in gpu.

  #14957 closed Aug 7, 2025

• Reuse output tensors memory that was allocated by first call to Ort::Session.Run(...)

  #14960 closed Aug 7, 2025

• Compatibility between Onnx and Blazor Webassembly

  #14962 closed Aug 7, 2025

• Running T5 export ONNX example leads to shape inference error

  #14963 closed Aug 7, 2025

• Microsoft.ML.OnnxRuntime.Gpu not working in MAUI project

  #14974 closed Aug 7, 2025

• [Build] Failed to build in docker container

  #14983 closed Aug 7, 2025

• conv throws safeint exception

  #14985 closed Aug 7, 2025

• Static linkage of onnx_runtime and providers library

  #14986 closed Aug 7, 2025

• [Build] static assertion fails when building from source with GCC 13.0.1

  #14991 closed Aug 7, 2025

• [Performance] inference problems with io_binding: unexpected shape or unexpected data type

  #14998 closed Aug 7, 2025

• [Performance] TensorRT provider produces (slightly) differently named engine files for the same model between runs

  #14999 closed Aug 7, 2025

• CUDA Graph Error - CUDA failure 900: operation not permitted when stream is capturing

  #15002 closed Aug 7, 2025

• ONNX does not support Dirichlet distribution?

  #15016 closed Aug 7, 2025

• [Build] Problems with FP16 Layernorm

  #15021 closed Aug 7, 2025

• [Build] api-ms-win-core-heap-l2-1-0.dll missing on windows server 2012 R2

  #15025 closed Aug 7, 2025

• onnxruntime::ProviderLibrary::Get [ONNXRuntimeError] : 1 : FAIL : LoadLibrary failed with error 126

  #15035 closed Aug 7, 2025

• accuracy reduced with multithreaded GPU prediction

  #15038 closed Aug 7, 2025

• mT5 convert to ONNX and GPU inference problems

  #15042 closed Aug 7, 2025

• [Build] Cannot specify compile definitions for target "onnx" which is not built by this project.

  #15051 closed Aug 7, 2025

• [bug] error while loading shared libraries: libonnxruntime.so.1.8.1: cannot open shared object file: No such file or directory

  #15053 closed Aug 7, 2025

• [Performance] Inference doubles VRAM (DirectML)

  #15074 closed Aug 7, 2025

• [Web] Memory spike in ORT-web leading to app crash

  #15086 closed Aug 7, 2025

• onnxruntime: bfc_arena.cc:361 void* onnxruntime::BFCArena::FindChunkPtr(onnxruntime::BFCArena::BinNum, size_t, size_t) !chunk->in_use() was false.

  #15087 closed Aug 7, 2025

• The dimension of incides to ScatterND op is wrong during inference.

  #15095 closed Aug 7, 2025

• [Performance] onnxruntime allocates lots of cuda memory on T4

  #15098 closed Aug 7, 2025

• fail build with gcc 12.x in onnxruntime/contrib_ops/cuda/quantization/qordered_ops/qordered_qdq.cc

  #15111 closed Aug 7, 2025

• How to reduce GPU memory usage when inference

  #15127 closed Aug 7, 2025

• descriptor_table_tensorboard_2fcompat_2fproto_2fattr_5fvalue_2eproto not declared (TRT 8.5.0)

  #15131 closed Aug 7, 2025

• how to inference with fp16 precise in python code?

  #15134 closed Aug 7, 2025

• NOT_IMPLEMENTED GridSample(16) on onnxruntime 1.14.1

  #15137 closed Aug 7, 2025

• onnxruntime::utils::ConstantNodeProtoToTensorProto Unsupported attribute value type of 9 in 'Constant' node 'Constant_35'

  #15149 closed Aug 7, 2025

• Type Error: Type 'tensor(int64)' of input parameter (relative_position) of operator (Min) in node (Min_2286) is invalid.

  #15167 closed Aug 7, 2025

• inference speed is very slow when using fp16 while using fp 32 is normal

  #15170 closed Aug 7, 2025

• A bug occurs when the program terminates

  #15174 closed Aug 7, 2025

• [Performance] Why is the Conv + Max-Pool model faster than the Conv model using GraphOptimizationLevel::ORT::ENABLE_ALL?

  #15180 closed Aug 7, 2025

• [Performance] GPT NEO: better performance of python GPT NEO than its onnx runtime version in C++?

  #15191 closed Aug 7, 2025

• [Build] segfault when run unitest (ctest)

  #15224 closed Aug 7, 2025

• [Build] fail to build on Windows ARM64

  #15252 closed Aug 7, 2025

• [Performance] How to debug/reduce GPU utilization?

  #15254 closed Aug 7, 2025

• [Performance]

  #15265 closed Aug 7, 2025

• ONNX model with FBNetv3 architecture Conversion to TensorRT Problem

  #15269 closed Aug 7, 2025

• [Build] ONNX Java Runtime - Handle UnsatisfiedLinkError

  #15281 closed Aug 7, 2025

• [Documentation Request] Estimating (or Checking) Allocated Memory

  #15326 closed Aug 7, 2025

• [Performance] Timings feedback

  #15328 closed Aug 7, 2025

• [Performance] Gemm op is slower after quantization

  #15332 closed Aug 7, 2025

• [Mobile] onnxruntime-c and onnxruntime-extensions-c pod conflict with DocumentReader pod

  #15333 closed Aug 7, 2025

• [Performance] ONNXRUNTIME sometime DEAD in python multiprocessing

  #15345 closed Aug 7, 2025

• ValueError: Message onnx.ModelProto exceeds maximum protobuf size of 2GB: 2235646909

  #15349 closed Aug 7, 2025

• [Web] custom ops

  #15374 closed Aug 7, 2025

• [Performance] Running Large Language Models for dynamic input size is poor performance. (DirectML)

  #15394 closed Aug 7, 2025

• Opset Coverage - Binary Size Tradeoff

  #15397 closed Aug 7, 2025

• [Build] C++ API calling fail: error C2280: 'Ort::Value::Value(const Ort::Value &)' : attempt to reference a deleted function

  #15418 closed Aug 7, 2025

• Mask-RCNN network is giving significantly different result with DirectML EP

  #15459 closed Aug 7, 2025

• Error Unrecognized attribute: layout for operator DynamicQuantizeLSTM

  #15465 closed Aug 7, 2025

• Please provide informative message on dlopen failures -- python API

  #15476 closed Aug 7, 2025

• [Performance] WebAssembly 1x1 Conv almost 4x slower than native

  #15483 closed Aug 7, 2025

• [Performance] Model converted to mixed precision results in higher latency

  #15490 closed Aug 7, 2025

• Inference slows down on gpu.

  #15491 closed Aug 7, 2025

• [Bug?] Casting int8-->float

  #15492 closed Aug 7, 2025

• InferenceSession fails with segmentation fault when fp16 model is loaded with CPUExecutionProvider

  #15494 closed Aug 7, 2025

• [ErrorCode:Fail] Load model from [...]\latin_ipa_forward.onnx failed:invalid vector subscript

  #15495 closed Aug 7, 2025

• [Build] Openvino debug build fails on VS2019

  #15496 closed Aug 7, 2025

• [Web] probability is not returned: `error code = 1`

  #15511 closed Aug 7, 2025

• SimplifiedLayerNormalization loading error for converted FP16 databricks/dolly-v2-3b model

  #15531 closed Aug 7, 2025

• [Performance] FP16 model can not get acceleration on GPU with ONNXRuntime-GPU

  #15534 closed Aug 7, 2025

• Get results from Mask RCNN model with C++

  #15541 closed Aug 7, 2025

• fatal error: gsl/gsl: No such file or directory

  #15554 closed Aug 7, 2025

• [Build] 1.14.0-dev-20230120-0204-3d6cea14f4 (This build breaks model on Intel)

  #15567 closed Aug 7, 2025

• [Performance] CUDA fp16 didn't get speed up

  #15585 closed Aug 7, 2025

• Error with custom spconv class in onnx runtime

  #15594 closed Aug 7, 2025

• [Build] Java Nightly build

  #15600 closed Aug 7, 2025

• [Build] the Linux build config

  #15621 closed Aug 7, 2025

• Can't use onnxruntime with DirectML built from source

  #15628 closed Aug 7, 2025

• [Performance] CNN model exported by PyTorch runs slower than Tensorflow 1.0

  #15647 closed Aug 7, 2025

• onnxRuntimeException and DefaultLogger issues in AWS Lambda runtime

  #15650 closed Aug 7, 2025

• ONNXRuntime in Docker

  #15652 closed Aug 7, 2025

• ONNX with FloatTensorType when inferred from C++ returns different label everytime

  #15665 closed Aug 7, 2025

• [Build] Compile Error if path too long

  #15674 closed Aug 7, 2025

• [CANN]EP: CANN cannot complete inference on Atlas200DK

  #15677 closed Aug 7, 2025

• [Performance] Can't get GPU speed-up when exe program is located inside the path with chinese character

  #15678 closed Aug 7, 2025

• [ErrorCode:InvalidArgument] Invalid Feed Input Name:image

  #15692 closed Aug 7, 2025

• [Performance] Can we set model weight precision when converting keras model into onnx model?

  #15695 closed Aug 7, 2025

• [Build] Onnxruntime-gpu for Jetpack 5.1.1 on Jetson Orin Nano Developer Kit

  #15732 closed Aug 7, 2025

• GraphOptimization (ORT_ENABLE_ALL) is slower using ONNXRuntime-GPU

  #15743 closed Aug 7, 2025

• Load onnx failed(segmentation fault) with version 1.14.1 (2)

  #15745 closed Aug 7, 2025

• Inference using the CUDA EP returns nan

  #15752 closed Aug 7, 2025

• [Build]

  #15786 closed Aug 7, 2025

• How to set CalibrationDataReader when my datatype is time series?

  #15836 closed Aug 7, 2025

• [Build]

  #15863 closed Aug 7, 2025

• [Training] Training Onnx format Models

  #15867 closed Aug 7, 2025

• Failed top create CUDAExecutionProvider

  #15873 closed Aug 7, 2025

• [RunTimeError]Infer error shape in runtime and mismatch with onnx spec about Split opset 18

  #15882 closed Aug 7, 2025

• [Performance] `CUDAExecutionProvider` uses 3x the memory of `CPUExecutionProvider`

  #15886 closed Aug 7, 2025

• symbolic_shape_infer.py failure

  #15898 closed Aug 7, 2025

• Linking executable with static libraries --> error LNK2038: mismatch detected

  #15928 closed Aug 7, 2025

• Atlas200DK uses EP: CANN to infer resnet50 and reports "CANN errorEE9999: Inner Error!"

  #15947 closed Aug 7, 2025

• [Performance] Redundant ReorderOutput / ReorderInput operators in Conv+Maxpool layers when graph optimization level is ALL

  #15964 closed Aug 7, 2025

• float16 result not match with numpy or torch

  #15977 closed Aug 7, 2025

• [Predict] Prediction from ONNX is same for all images

  #16001 closed Aug 7, 2025

• The result of Col2Im operator not close with Torch result on fp16 dtype

  #16007 closed Aug 7, 2025

• [Performance] QUInt8 vs a basic ONNX

  #16009 closed Aug 7, 2025

• RunOptions.only_execute_path_to_fetches not working

  #16013 closed Aug 7, 2025

• Cannot open include file: numpy/arrayobject.h

  #16027 closed Aug 7, 2025

• [Web] The onnxruntime-web example is loading wasm file twice if set to local path

  #16028 closed Aug 7, 2025

• [Web] [WebGPU] Uncaught (in promise) DOMException: Unable to instantiate a Device in Firefox Nightly/Linux

  #16029 closed Aug 7, 2025

• inference time decreasing when increasing batch size to a certain point and them the inference time increasing again.

  #16030 closed Aug 7, 2025

• can we customize memory allocation functions(like malloc/free) for inference in C api?

  #16032 closed Aug 7, 2025

• [Performance] How to solve the problem of releasing GPU memory in onnxruntime

  #16033 closed Aug 7, 2025

• [Performance] Huge gap between nn.Conv1d() and nn.Conv2d() - models exported by PyTorch

  #16047 closed Aug 7, 2025

• Unexpected inference output from QLinearConv

  #16105 closed Aug 7, 2025

• Memory leak in cpuinfo_x86_linux_init

  #16117 closed Aug 7, 2025

• Segmentation Fault when optimizing Stable Diffusion models

  #16140 closed Aug 7, 2025

• onnxruntime.capi.onnxruntime_pybind11_state.Fail: [ONNXRuntimeError] with swin-t

  #16143 closed Aug 7, 2025

• Segmentation fault while loading CUDA Provider

  #16146 closed Aug 7, 2025

• [Performance] ONNX Runtime doesn't parallelize operations in CPU models

  #16158 closed Aug 7, 2025

• The prediction results from STFT has changed with a notable shift towards larger difference from PyTorch in ORT==1.15.0

  #16163 closed Aug 7, 2025

• [MacOS] Unable to load libonnxruntime.dylib because binaries are not signed.

  #16168 closed Aug 7, 2025

• [Build] line 2812, in <module> sys.exit(main())

  #16179 closed Aug 7, 2025

• no acceleration onnx on e5 2680v3

  #16185 closed Aug 7, 2025

• [Performance] setIntraOpNumThreads doesn't offer enough parallelization in JAVA-API

  #16192 closed Aug 7, 2025

• DmlExecutionProvider bound to PyTorch tensor stops running

  #16197 closed Aug 7, 2025

• NullReferenceException when creating an object of class SessionOptions | Unity

  #16205 closed Aug 7, 2025

• [quantization] Problem with QDQ of Pow/Sqrt/Div

  #16219 closed Aug 7, 2025

• why the input doesn't place in cuda ?

  #16225 closed Aug 7, 2025

• [Training][api:C++][feature request] Support Model Forward Output and Backward Gradient Extraction in ONNX runtime training

  #16232 closed Aug 7, 2025

• TensorrtExecutionProvider::GetSupportedList graph_build.Resolve().IsOK() was false.

  #16234 closed Aug 7, 2025

• Not returning anything for out-of-vocabulary text while batch inference using Tf-IDF ONNX Vectorizer model

  #16251 closed Aug 7, 2025

• Inconsistent generation of vectors by TF-IDF ONNX Vectorizer Model

  #16252 closed Aug 7, 2025

• [OOM] Unable to convert 30B Model

  #16254 closed Aug 7, 2025

• [Performance] Evaluation behavior with external arrays (C API)

  #16255 closed Aug 7, 2025

• onnx use more memory than pytorch for some model

  #16264 closed Aug 7, 2025

• [Web/Build] Failed to consume onnxruntime-common because of JS parser not up-to-date

  #16265 closed Aug 7, 2025

• how to trace the error "assert node is not None" when use the onnxruntime.transformers.optimizer

  #16268 closed Aug 7, 2025

• ROCm EP: Errors when trying to infer, which GPUs are supported?

  #16271 closed Aug 7, 2025

• [Accuracy/Performance]

  #16275 closed Aug 7, 2025

• Does OnnxruntimeV1.14 still support the Python Operator, and which the highest version supports this feature？

  #16277 closed Aug 7, 2025

• Seg faults when creating InferenceSession for SAM backbone

  #16300 closed Aug 7, 2025

• [Mobile] Error: Non string type of a tensor data is not allowed

  #16301 closed Aug 7, 2025

• [ onnxruntime::SequentialExecutor::Execute] Non-zero status code returned while running FusedMatMul node. Name:'MatMul_With_Transpose_token_14_FusedMatMulAndScale' Status Message: bad allocation unknown file: error: C++ exception with description "Non-zero status code returned while running FusedMatMul node. Name:'MatMul_With_Transpose_token_14_FusedMatMulAndScale' Status Message: bad allocation" thrown in the test body.

  #16305 closed Aug 7, 2025

• issue running onnxruntime with pytest

  #16306 closed Aug 7, 2025

• How to catch exception OOM.

  #16307 closed Aug 7, 2025

• How to edit Clip Operator in OnnxRuntime?

  #16315 closed Aug 7, 2025

• get error when using libonnxruntime with dnnl EP

  #16320 closed Aug 7, 2025

• InvalidGraph: [ONNXRuntimeError] : 10 : INVALID_GRAPH : Load model from d2net.onnx failed:This is an invalid model. Type Error: Type 'tensor(bool)' of input parameter (onnx::Min_109) of operator (Min) in node (Min_65) is invalid.

  #16321 closed Aug 7, 2025

• Increase - decrease the maximum number of events during inference profiling.

  #16334 closed Aug 7, 2025

• [Build] Fails to parse FP16 LayerNormalization in opset>=18

  #16341 closed Aug 7, 2025

• [Build] Disable ORT_ENABLE_STREAM build error

  #16345 closed Aug 7, 2025

• MaxPool: When Ceil_mode=1, MaxPool Generates Big Values.

  #16350 closed Aug 7, 2025

• AveragePool: When Ceil_mode=1, AveragePool Generates Nan or 0 Values.

  #16351 closed Aug 7, 2025

• [Training]

  #16354 closed Aug 7, 2025

• multi-GPU inferencing

  #16382 closed Aug 7, 2025

• Operator Pad reflect mode does not yield correct results

  #16401 closed Aug 7, 2025

• [Web] Web ~40x slower than native

  #16412 closed Aug 7, 2025

• [Performance] DML dynamic axes performance regression.

  #16424 closed Aug 7, 2025

• C++ Runtime does not recognize supposedly correct input.

  #16430 closed Aug 7, 2025

• Normalizer does not work as expected

  #16451 closed Aug 7, 2025

• [Mobile] Unable to load models in Xamarin iOS

  #16463 closed Aug 7, 2025

• [Performance] net.set_providers(['DmlExecutionProvider'], [{'device_id': 0}]) could get stuck forever (directml EP)

  #16473 closed Aug 7, 2025

• m2m 100 418M

  #16480 closed Aug 7, 2025

• Automatic deallocation (?) of the Ort::Sessions, memory leak?

  #16497 closed Aug 7, 2025

• [Performance] A model with a large TreeEnsembleClassifier node takes too long to be loaded

  #16511 closed Aug 7, 2025

• Setting `CUBLAS_WORKSPACE_CONFIG=":4096:8"` leads to `CUBLAS_STATUS_ALLOC_FAILED`

  #16512 closed Aug 7, 2025

• ONNXRuntimeError: Training mode does not support BN opset 14 (or higher) yet.

  #16867 closed Aug 7, 2025

• [Build] INVALID_ARGUMENT : Invalid rank for input: input Got: 4 Expected: 2 Please fix either the inputs or the model.

  #16557 closed Aug 7, 2025

• [Build] libonnxruntime_providers_dnnl.so: undefined symbol: omp_get_max_threads

  #16561 closed Aug 7, 2025

• Large model >2GB save_to_ort

  #16573 closed Aug 7, 2025

• [Build] fatal error: too many errors emitted, stopping now [-ferror-limit=]

  #16576 closed Aug 7, 2025

• [Build] Cannot build onnxruntime

  #16583 closed Aug 7, 2025

• Conv3d precision error between pytorch and onnx

  #16589 closed Aug 7, 2025

• [Training] Define a custom training with some ONNX models

  #16597 closed Aug 7, 2025

• [Performance] Performance degradation observed w.r.t DNNL-EP in v1.15.1 compared to v1.13.1

  #16609 closed Aug 7, 2025

• [Build] No C++ library is generated after compilation completed

  #16610 closed Aug 7, 2025

• [Performance] Computation time of iteratively applying neural network in a single ONNX model using CUDA Execution Provider dominated by Memcpy

  #16625 closed Aug 7, 2025

• [Build] Dependency on OMP/MPI Runtime

  #16631 closed Aug 7, 2025

• [Performance]

  #16637 closed Aug 7, 2025

• The input tensor cannot be reshaped to the requested shape after adding Gather output to model's output

  #16670 closed Aug 7, 2025

• Access violation reading location when I use CreateArenaCfgV2 and CUDA

  #16686 closed Aug 7, 2025

• One path in the graph requests feature X(>Y) but input tensor has Y features

  #16695 closed Aug 7, 2025

• [Bug] Coqui VITS ONNX model can't be statically quantized.

  #16738 closed Aug 7, 2025

• CUDA Custom Op CUDA failure

  #16748 closed Aug 7, 2025

• clean build v1.15.1 fails three fp16 tests due to `difference between... exceeds threshold"

  #16775 closed Aug 7, 2025

• [Performance] FP16 models incur large cast latency when run on CPUs without FP16 support

  #16778 closed Aug 7, 2025

• Incorrect Output from Java Model

  #16781 closed Aug 7, 2025

• Segmentation Fault when using TensorRT execution provider

  #16790 closed Aug 7, 2025

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

  #16793 closed Aug 7, 2025

• [Performance] [Web] Using the `onnxruntime-web` package (`wasm` backend) with Node.js is 1.6x to 2x faster than in browsers and Deno?

  #16798 closed Aug 7, 2025

• [Training] Proposal: Implement back propagation algorithm for C#

  #16809 closed Aug 7, 2025

• [Performance]

  #16817 closed Aug 7, 2025

• [Mobile] Failure to load whisper model .ort with react-native, regular and quantized versions

  #16819 closed Aug 7, 2025

• onnxruntime_providers_cuda.dll cannot be loaded due to "Can't find dependent libraries" under Windows 10 environment using Java

  #16821 closed Aug 7, 2025

• op.SequenceEmpty(dtype=xxx) cannot be set to float16.

  #16846 closed Aug 7, 2025

• [Performance]high latency variance

  #16876 closed Aug 7, 2025

• [Performance] Convolution layer issue profiling

  #16926 closed Aug 7, 2025

• [Mobile][Kotlin] OnnxTensor.createTensor from floatBuffer takes up 7 seconds

  #16937 closed Aug 7, 2025

• Crash at winrt::Windows::AI::MachineLearning::implementation::LearningModelSession::GetResults during inferencing

  #16988 closed Aug 7, 2025

• Native assemblies aren't copied when Onnx is a transitive dependency and using netstandard

  #17010 closed Aug 7, 2025

• Why onnxruntime extracts only 483MB json file?

  #17013 closed Aug 7, 2025

• Why some models are not profiling the input weights, bias etc and the node index in the json file properly?

  #17022 closed Aug 7, 2025

• An error occurred when I used the TensorrtExecutionProvider in onnx runtime

  #17047 closed Aug 7, 2025

• [Web] Cannot Convert to RGB when using Tensor.fromImage(image,{tensorFormat:'RGB'})

  #17094 closed Aug 7, 2025

• [Performance] Pytorch Model converted to ONNX with CUDAProvider run slower 3x time than Using Pytorch with GPU

  #17116 closed Aug 7, 2025

• How to release onnxruntime gpu memory

  #17142 closed Aug 7, 2025

• [TOOLS]：Using transformers.optimizer optimize large model, segmentation fault (core dumped)

  #17212 closed Aug 7, 2025

• Onnx model inference Fatal error: ai.onnx.contib:bev_pool_v2(-1) is not a registered function/op

  #17214 closed Aug 7, 2025

• [C#] Invalid input name error

  #17244 closed Aug 7, 2025

• AssertionError on num_heads > 0 for bert with specific optimization config

  #17254 closed Aug 7, 2025

• windows10 x86 x64 inference time varies greatly

  #17256 closed Aug 7, 2025

• [Performance] Operators assigned to CPU instead of CUDA, CPU thread management problem

  #17268 closed Aug 7, 2025

• [Web] Error: no available backend found. ERR: [wasm] TypeError: Failed to parse URL from

  #17274 closed Aug 7, 2025

• [Build] Error: cpuid.h: No such file or directory when cross-compiling ORT 1.15.1 with NNAPI for arm64

  #17283 closed Aug 7, 2025

• Freeing heap block containing an active critical section

  #17345 closed Aug 7, 2025

• [Performance] 3X slower inference on onnxruntime than pytorch(huggingface)

  #17366 closed Aug 7, 2025

• [Performance] Memcpy leads to AllocationError for argmax

  #17371 closed Aug 7, 2025

• [web/js] need for more methods on tensor object

  #17372 closed Aug 7, 2025

• [Performance] Quantized model inference on CPU slower/same as FP32

  #17389 closed Aug 7, 2025

• Default `tensorFormat` should RGBA for HTMLImageElement variant

  #17395 closed Aug 7, 2025

• [Build] windows dll compilation error with versions above 1.14.0

  #17404 closed Aug 7, 2025

• [Web] Add binary/where broadcast case when FXC issue got fixed in tint

  #17405 closed Aug 7, 2025

• [Performance] Data size of Batch Normalization using cuDNN in inference.

  #17406 closed Aug 7, 2025

• Yolov8 Static Quantization

  #17410 closed Aug 7, 2025

• CUDA Stream and Synchronous in custom operato

  #17412 closed Aug 7, 2025

• [Performance] How much memory it needs to load a 3.4 GB model to GPU through DirectML?

  #17413 closed Aug 7, 2025

• valgrind memcpy_chk overlap onnxruntime1.15.1

  #17431 closed Aug 7, 2025

• Extract node info

  #17444 closed Aug 7, 2025

• [Bug] FP16 conversion yields an unusable model

  #17447 closed Aug 7, 2025

• [Mobile iOS] Run fp16 onnx model on CoreML EP

  #17448 closed Aug 7, 2025

• C++ API, Memory Leak instantiating Ort::Sessions

  #17451 closed Aug 7, 2025

• Failure with OpenvinoEP within ORT

  #17499 closed Aug 7, 2025

• Resize of doesn't work well while the coordinate_transformation_mode is 'align_corners'.

  #17564 closed Aug 7, 2025

• Inference speed of Quantized model not increased after static Quantization[Performance]

  #17634 closed Aug 7, 2025

• DML EP One session but called in different threads. [Performance]

  #17686 closed Aug 7, 2025

• SkipLayerNormFusion -- High Output Difference Between PyTorch and ONNX Runtime with Extended Optimizations

  #17689 closed Aug 7, 2025

• [Web]

  #17700 closed Aug 7, 2025

• [Mobile | iOS] I got "Unknown exception" error.

  #17731 closed Aug 7, 2025

• [Web] Custom build packages

  #17743 closed Aug 7, 2025

• [web] following-up work items for supporting uniform buffers

  #17860 closed Aug 7, 2025

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

  #17979 closed Aug 7, 2025

• [Build] Why does TensorRT EP need the full version of protobuf?

  #18040 closed Aug 7, 2025

• [Web] Which node.js version is supposed to be supported?

  #18078 closed Aug 7, 2025

• Microsoft.ML.OnnxRuntime.OpenVino Encountered unknown exception in Initialize

  #18152 closed Aug 7, 2025

• ORT bug in Col2Im CPU 3D cases

  #18156 closed Aug 7, 2025

• [Mobile|Android] Fatal error: ai.onnx.contrib:SentencepieceTokenizer(-1) is not a registered function/op

  #18226 closed Aug 7, 2025

• The onnx.helper make_function command strips type information leading to inference errors

  #18264 closed Aug 7, 2025

• [Web] onnxruntime-web and onnxruntime-node return different results for LSTM model

  #18335 closed Aug 7, 2025

• [Performance] the speed with SetIntraOpNumThreads(1),SetIntraOpNumThreads(4),SetInterOpNumThreads(1),SetInterOpNumThreads(4)

  #18385 closed Aug 7, 2025

• [Performance] Does `com.microsoft.Attention` use FlashAttention-2?

  #18474 closed Aug 7, 2025

• Add ORT Extensions to Java and build with Gradle

  #18503 closed Aug 7, 2025

• Model Run Session wasting time[Performance]

  #18510 closed Aug 7, 2025

• Is there any way to convert a qdqmodel to qlinearmodel use ort?

  #18511 closed Aug 7, 2025

• [Training] qat

  #18534 closed Aug 7, 2025

• [Build] manylinux_2_28 support

  #18537 closed Aug 7, 2025

• [Build] TRT EP cannot be built without CUDA EP

  #18542 closed Aug 7, 2025

• Call Session class method name Run failed,don't know why

  #18548 closed Aug 7, 2025

• Does the computation order affect the computation result?

  #18564 closed Aug 7, 2025

• [Web] How could I get the shape of the output tensor?

  #18568 closed Aug 7, 2025

• [Build]

  #18570 closed Aug 7, 2025

• # Issue with Rounding Behavior in onnxruntime's Quantizelinear Layer

  #18576 closed Aug 7, 2025

• Session Run throws an access violation exception when I recreate the session

  #18578 closed Aug 7, 2025

• [Node.js] Support for loading models with external data in `onnxruntime-node`

  #18586 closed Aug 7, 2025

• Cuda EP does not compute reduce with empty set correctly?

  #18588 closed Aug 7, 2025

• [Mobile] Model with large input size cause Segmentation Fault while session->run()

  #18595 closed Aug 7, 2025

• Session initialization stuck/crash in DMLCreateDevice while using DirectML EP

  #18599 closed Aug 7, 2025

• Profiling multithreaded runs

  #18600 closed Aug 7, 2025

• Segmentation Fault when some of node outputs is empty

  #18601 closed Aug 7, 2025

• What is the recommended setup for running multiple models/sessions in parallel in C++?

  #18610 closed Aug 7, 2025

• DirectML Resize Node error.

  #18613 closed Aug 7, 2025

• [Build]

  #18617 closed Aug 7, 2025

• Could not find an implementation for SkipGroupNorm(1) node with name 'SkipGroupNorm_0'

  #18623 closed Aug 7, 2025

• Crash in ResizeHelper::Initialize executing a model on ARM64

  #18628 closed Aug 7, 2025

• ONNXRuntime Segmentation Fault Crash on Inference (iOS and Mac)

  #18632 closed Aug 7, 2025

• [Performance] dynamic batch infer cost time question

  #18639 closed Aug 7, 2025

• ORT memory error with the graph from linspace

  #18648 closed Aug 7, 2025

• Are there any benchmark tools for onnx mobile like Tensorflow Lite?

  #18664 closed Aug 7, 2025

• Different results of consecutive runs for same input

  #18672 closed Aug 7, 2025

• Strange condition size_t channel_rindex = is_nchw ? 2 : 2;

  #18674 closed Aug 7, 2025

• Missprinted condition: head_size != num_heads * head_size

  #18675 closed Aug 7, 2025

• Parallel inference of multiple models in different threads

  #18806 closed Aug 7, 2025

• Onnxruntime using OpenVINO as execution provider encountered Exception during initialization problem on model candy.onnx

  #18825 closed Aug 7, 2025

• [Performance] Java API lacks functionality to control allocator settings.

  #18845 closed Aug 7, 2025

• [dynamo_export] starts_.size() == ends_.size() + 1 was false. No matching 'start' entry.

  #18863 closed Aug 7, 2025

• [dynamo_export] MLFloat16 data type is not supported with ScatterElements opset 18 when reduction is 'max'.

  #18864 closed Aug 7, 2025

• [Web] Non-zero status code returned while running Slice node `webgpu`

  #18892 closed Aug 7, 2025

• compute_range not available

  #18893 closed Aug 7, 2025

• the resout of onnx and trt engine is different?why?

  #18902 closed Aug 7, 2025

• SafeIntOnOverflow() Integer overflow error when inferencing on too many samples with Python

  #18905 closed Aug 7, 2025

• error 126 Onnx in ComfyUI[Performance] O

  #18925 closed Aug 7, 2025

• ai.onnxruntime.OrtException: Unsupported type - FLOAT16

  #18926 closed Aug 7, 2025

• How to use multiple inputs of different types in C++ session

  #18932 closed Aug 7, 2025

• [Web] onnxruntime-web is not work in nodejs

  #18933 closed Aug 7, 2025

• [Web] no available backend found. ERR: [wasm] TypeError: _ is not a function, [cpu] Error: previous call to 'initializeWebAssembly()' failed., [xnnpack] Error: previous call to 'initializeWebAssembly()' failed

  #18938 closed Aug 7, 2025

• How to set `trt_profile_min_shapes` for inputs with name containing colons?

  #18939 closed Aug 7, 2025

• OP (Conv) inference results mismatch with PyTorch

  #18946 closed Aug 7, 2025

• [Build] How to build onnxruntime with openvino statically?

  #18950 closed Aug 7, 2025

• [Performance] 2x Regression in 1st Inference time cost

  #18957 closed Aug 7, 2025

• High Output Difference between ONNX model with different optimizer settings

  #18959 closed Aug 7, 2025

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

  #18966 closed Aug 7, 2025

• Error with finding onnxruntime_binding.node on Windows 10 on a bootcamp Macbook

  #18971 closed Aug 7, 2025

• How to observe arena allocator memory request metrics

  #18972 closed Aug 7, 2025

• Could not load library cudnn_cnn_infer64_8.dll. Error code 127

  #18973 closed Aug 7, 2025

• [Build] Failure with OneDNN on Intel MacOS

  #18976 closed Aug 7, 2025

• Cannot quantize yolov5 float to int8 onnx model

  #18987 closed Aug 7, 2025

• Encounter unknown exception in initialize using Openvino EP

  #19004 closed Aug 7, 2025

• ONNX Runtime inference on string input

  #19006 closed Aug 7, 2025

• [Error: Exception in HostFunction: <unknown>] while running ort models in react-native

  #19021 closed Aug 7, 2025

• [Performance] It is not possible to use a discrete graphics card with DML.

  #19025 closed Aug 7, 2025

• [Build] deploying the EfficientAD anomaly detection algorithm, an error occurred while executing the "Run" command

  #19030 closed Aug 7, 2025

• Freeing tensor data created via CreateTensor

  #19034 closed Aug 7, 2025

• [Build] Linux x86_64 STATIC Build

  #19035 closed Aug 7, 2025

• cudaMemcpyAsync throws exception in GPUDataTransfer

  #19076 closed Aug 7, 2025

• [Training] On device training doesn't work with INT8 Models

  #19078 closed Aug 7, 2025

• [Performance] The CUDA Stream cannot be set through Python API

  #19094 closed Aug 7, 2025

• Longformer `convert_to_onnx.py` not working due to missing imports

  #19149 closed Aug 7, 2025

• [Performance] Why run first inference so slow, although run one time in initialzation?

  #19177 closed Aug 7, 2025

• ORT 1.17.0 Release Candidates available for testing

  #19236 closed Aug 7, 2025

• `shape_inference.quant_pre_process` causes `AttributeError: module 'onnx.helper' has no attribute 'make_sequence_value_info'`

  #19323 closed Aug 7, 2025

• [Training] How to update running_mean and running_var of BatchNormalization during training

  #19370 closed Aug 7, 2025

• [Performance] In ONNX Runtime, the CPU consumption does not scale linearly with the number of threads

  #19384 closed Aug 7, 2025

• Backwards convolution layers in CUDA provider should heed

  #19391 closed Aug 7, 2025

• InferenceSession.run does not validate rank of scalar inputs

  #19434 closed Aug 7, 2025

• [Web] Memory Access Out of Bounds Error When Using ONNX Runtime Web Inference in NPM Package (wasm)

  #19443 closed Aug 7, 2025

• [Performance] CPU inference much slower from GPU runtime

  #19451 closed Aug 7, 2025

• [On-device Training] Yolo custom loss

  #19464 closed Aug 7, 2025

• [Performance]

  #19479 closed Aug 7, 2025

• Errors about using c# and TensorRT

  #19489 closed Aug 7, 2025

• Accuracy drops a lot when using fp16 with TensorRT EP

  #19492 closed Aug 7, 2025

• quantize_dynamic : nodes_to_quantize(Gemm) is ignored

  #19503 closed Aug 7, 2025

• ONNX Runtime OpenVINO EP is way behind

  #19688 closed Aug 7, 2025

• Observed TDR on a low-end system

  #19724 closed Aug 7, 2025

• Inconsistent Prediction Outputs for Onnx Model

  #19834 closed Aug 7, 2025

• import InferenceSesseion and capi._pybind_state.

  #19836 closed Aug 7, 2025

• [Performance] onnxruntime 1.17.1 version doesnt support CUDA 12.4

  #19839 closed Aug 7, 2025

• [Performance] Accuracy dropped heavily using onnxruntime to inference a model quantized by QAT

  #19850 closed Aug 7, 2025

• Inference speed problem even if using a high-end Hardware.

  #19865 closed Aug 7, 2025

• [iOS] Output of type sequence<map<int64,float32>> causes crash on iOS

  #19867 closed Aug 7, 2025

• [Build] Where is official build for Unity?

  #19964 closed Aug 7, 2025

• [BUG] [OpenVino EP] Only first result in session is correct.

  #19975 closed Aug 7, 2025

• Onnx Runtime EntryPointNotFoundException: OrtGetApiBase in Unity Application.

  #20048 closed Aug 7, 2025

• Layer not supported in one provider (Tensorrt) not working with second provider (CUDA) in an inference problem.

  #20058 closed Aug 7, 2025

• [Performance] Inference failed or unsupported using quantize_dynamic

  #20060 closed Aug 7, 2025

• openvino with int8

  #20072 closed Aug 7, 2025

• Unpredictable onnxruntime-node crash when using Electron

  #20084 closed Aug 7, 2025

• In Aquatic mode links text “PyTorch and Hugging face” is not clearly visible: A11y_WCP URLs - ONNX Runtime_Home_Learn more about how to use ONNX Runtime with_usability

  #20150 closed Aug 7, 2025

• `convert_float_to_float16` results in `failed in shape inference <class 'Exception'>`

  #20189 closed Aug 7, 2025

• Whether CUDA12.4 and cudnn9 matches onnxruntime-win-x64-cuda12-1.17.1

  #20223 closed Aug 7, 2025

• [Training] Can we use ORTModule for inference?

  #20281 closed Aug 7, 2025

• C API Seg Fault from OrtGetApiBase()->GetApi(ORT_API_VERSION);

  #20283 closed Aug 7, 2025

• [Performance] ScatterND / GridSample operators are on CPU instead of GPU / CUDA

  #20297 closed Aug 7, 2025

• DirectML returning empty result with ObjectDetection (Mobilinet V2 FPN Keras)

  #20386 closed Aug 7, 2025

• [Build] Cmake install debug and release configuration

  #20387 closed Aug 7, 2025

• [Performance] Profiling on CUDA shows confusing values

  #20398 closed Aug 7, 2025

• [Performance] Massive Performance slowdown from v1.13.1 -> 1.14.0

  #20400 closed Aug 7, 2025

• onnxruntime 1.17.3 is missing from cuda 12 artifacts feed

  #20409 closed Aug 7, 2025

• Dockerfile does not work

  #20458 closed Aug 7, 2025

• [Build] cross-compiling onnxruntime for arm32 and onnxruntime_ENABLE_CPUINFO not working.

  #20461 closed Aug 7, 2025

• RUNTIME_EXCEPTION, 80070057 The parameter is incorrect in v1.17.3

  #20464 closed Aug 7, 2025

• [Build] cmake duplicate target "memory" between abseil and xnnpack

  #20469 closed Aug 7, 2025

• [Build] Error when load pf16 model

  #20570 closed Aug 7, 2025

• DirectML Exception 80070057 "The parameter is incorrect"

  #20575 closed Aug 7, 2025

• windows系统，Java中使用onnxruntime进行压测，cpu飙升很快，一直100%

  #20593 closed Aug 7, 2025

• Missing dll cudnn_ops_infer64_8.dll does not generate a python error

  #20605 closed Aug 7, 2025

• [BUG] Running operations over concat output rewrites it's values

  #20606 closed Aug 7, 2025

• [Discussion] ORT GPU binaries do not contain DML

  #20638 closed Aug 7, 2025

• [Build] TVM EP Build

  #20665 closed Aug 7, 2025

• LayerNormalization doesnt' work as expected on Mac

  #20676 closed Aug 7, 2025

• User-provided session logging function is not used for every log

  #20680 closed Aug 7, 2025

• Windows ARM64 & X64 CLIP Image Encoder different results

  #20722 closed Aug 7, 2025

• [Build] quantization unittest failed when run all tests

  #20821 closed Aug 7, 2025

• [.NET] Update tensor implementations to new Tensor<T> type

  #20874 closed Aug 7, 2025

• Java CreateTensor with NIO ByteBuffer for reuse purpose

  #20882 closed Aug 7, 2025

• [Build] how to buid on openharmony?

  #20895 closed Aug 7, 2025

• Stateful/Memory models

  #20943 closed Aug 7, 2025

• [Performance] Severe performance penalty with transformer model and DirectML

  #20983 closed Aug 7, 2025

• onnxruntime shape mismatch during quantization of yolov8 models

  #21048 closed Aug 7, 2025

• DML cannot use device_id = 1 , run_with_iobinding failed.

  #21092 closed Aug 7, 2025

• Symbolic Shape infer fails on onnx file without much logs

  #21120 closed Aug 7, 2025

• [Performance] Whisper model inference results incorrect after Transformer Optimizer

  #21150 closed Aug 7, 2025

• ORT 1.18.1 Release Candidates available for testing

  #21173 closed Aug 7, 2025

• [Performance] Mapfile support for certain external data files is not working

  #21195 closed Aug 7, 2025

• [Mobile] QNN failed to finalize QNN graph for attention layer

  #21221 closed Aug 7, 2025

• TArray used for broadcast was limited to be within range [0, 8] on onnxruntime 1.16.3

  #21254 closed Aug 7, 2025

• Not able to load onnx model multilingual-e5-large

  #21321 closed Aug 7, 2025

• TensorRT EP's inference results are abnormal.

  #21457 closed Aug 7, 2025

• [Build] Unable to build with --use_dml

  #21568 closed Aug 7, 2025

• Memory leak in NPU inference after each one session.run

  #21587 closed Aug 7, 2025

• [Performance]

  #21635 closed Aug 7, 2025

• Quantized SeaLLM v2 Model Outputs Same as Input

  #21636 closed Aug 7, 2025

• Same Model Hash Code Issue from different models

  #21672 closed Aug 7, 2025

• [Bug]: Onnxruntime.CPU memoty leaks

  #21723 closed Aug 7, 2025

• Inferencing FP16 model using onnxruntime

  #21737 closed Aug 7, 2025

• [Web] requested dist/*.mjs files for cdnjs

  #21785 closed Aug 7, 2025

• run_async not running asynchronously

  #21791 closed Aug 7, 2025

• [Bug] [onnxruntime-node] Error: no available backend found. ERR: [wasm] backend not found.

  #21813 closed Aug 7, 2025

• Error when trying to run vision model onnx

  #21869 closed Aug 7, 2025

• [Build] “onnxruntime_cxx_api.h”: No such file or directory

  #21891 closed Aug 7, 2025

• Snapdragon X processor is unsupported

  #21947 closed Aug 7, 2025

• [Mobile] IOS library crashes in Release configuration

  #21960 closed Aug 7, 2025

• [Web] Uncaught WebGPU validation error on Snapdragon SM8450 but works on SM8250

  #21970 closed Aug 7, 2025

• [Build] onnxruntime-openvino library does not have python3.12 support

  #22015 closed Aug 7, 2025

• onnxruntime-gpu(1.18.0) can not be install

  #22028 closed Aug 7, 2025

• [Training] Implicit dependency of Python training API on 'torch' package

  #22070 closed Aug 7, 2025

• GetElementType is not implemented after updating onnxruntime

  #22075 closed Aug 7, 2025

• [Web] Error when using Web Workers on Next.js

  #22113 closed Aug 7, 2025

• [Question or BUG] ONNX Runtime CUDA Sessions in Unity Produce Empty Outputs When Running Multiple Models Sequentially on a Single Graphic Card

  #22146 closed Aug 7, 2025

• Warnings displayed as errors during TensorRT optimization.

  #22164 closed Aug 7, 2025

• trt_weight_stripped_engine_enable does not work for all networks/size ranges.

  #22165 closed Aug 7, 2025

• trt_weight_stripped_engine_enable does not work together with trt_dump_ep_context_model

  #22179 closed Aug 7, 2025

• Filenames in OrtTensorRTProviderOptionsV2 should be std::filesystem::path or at least const ORTCHAR_T*

  #22182 closed Aug 7, 2025

• [Performance] fp16 support and performance

  #22242 closed Aug 7, 2025

• Upcoming ORT 1.20 Release Overview

  #22274 closed Aug 7, 2025

• [Performance] High CUDA memory usage with ONNX Runtime and inconsistent memory release

  #22297 closed Aug 7, 2025

• Build failure on Windows 10 using OpenVino 2024.3 & 2024.4 both.

  #22314 closed Aug 7, 2025

• `quant_pre_process SymbolicShapeInference` causes AttributeError: 'NoneType' object has no attribute 'HasField' when the model has a Constant node.

  #22422 closed Aug 7, 2025

• The EP_CTX_BLOB seems to have both WRITE and EXECUTABLE permissions enabled

  #22437 closed Aug 7, 2025

• External data is not loaded with custom allocator

  #22468 closed Aug 7, 2025

• [Performance] C++ api: destroy the execution provider if the `Ort::Session` is destroyed

  #22511 closed Aug 7, 2025

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

  #22532 closed Aug 7, 2025

• Caused by: java.lang.UnsatisfiedLinkError: /tmp/onnxruntime-java18295816951647233732/libonnxruntime.so: Error relocating /tmp/onnxruntime-java18295816951647233732/libonnxruntime.so: __vsnprintf_chk: symbol not found

  #22539 closed Aug 7, 2025

• [DO NOT UNPIN] ORT Nightly Package Name Change

  #22541 closed Aug 7, 2025

• Negative output for sigmoid

  #22557 closed Aug 7, 2025

• [Performance] Model runtime spiky with TensorRT Execution Provider

  #22664 closed Aug 7, 2025

• Exception during initialization: safeint.h:17 static void SafeIntExceptionHandler<onnxruntime::OnnxRuntimeException>::SafeIntOnOverflow() Integer overflow - caused by int64 index of -1?

  #22694 closed Aug 7, 2025

• FP16 ONNX model outputs NaN after the first successful execution

  #22723 closed Aug 7, 2025

• CUDA providers failed to build against 12.6 with error error #221-D

  #22728 closed Aug 7, 2025

• why force max_length <= kMaxSequenceLength in beam_search_parameters.cc ?

  #22735 closed Aug 7, 2025

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

  #22822 closed Aug 7, 2025

• [Dev] "./onnxruntime_test_all --help" gives segmentation fault

  #22838 closed Aug 7, 2025

• how to release gpu memory when use onnxruntime with fastapi

  #22899 closed Aug 7, 2025

• [Performance] Binary operators using SSE on AVX systems

  #22905 closed Aug 7, 2025

• [Mobile] Error: Can't load a model: Error Code - ORT_INVALID_PROTOBUF

  #22927 closed Aug 7, 2025

• [Training] RuntimeError: gradient_builder_base.h:123 onnxruntime::training::ArgDef onnxruntime::training::GradientBuilderBase::O(size_t, bool) const i < node_->OutputDefs().size() was false

  #22955 closed Aug 7, 2025

• Remove Python :: 3.7 Python :: 3.8 Python :: 3.9 from pypi metadata

  #22993 closed Aug 7, 2025

• To reduce the compiled binary size of ONNX Runtime at x86_64 linux with "create_reduced_build_config.py", but got a Failed to find kernel for com.microsoft.nchwc.Conv(1)

  #23018 closed Aug 7, 2025

• [Build] Dotnet packages on nuget are not built with Release optimizations

  #23053 closed Aug 7, 2025

• [Web] ORT format model not working on WebGPU EP + Wasm Static lib

  #23072 closed Aug 7, 2025

• [Build] onnxruntime_gpu PiPy on a slow host

  #23079 closed Aug 7, 2025

• Cannot resolve operator 'LSTM' with webgl backend

  #23083 closed Aug 7, 2025

• [Bug][CUDAExecutionProvider] INVALID_ARGUMENT : unsupported conv activation mode "Sigmoid"

  #23114 closed Aug 7, 2025

• Understanding max_mem option of OrtArenaCfg class

  #23121 closed Aug 7, 2025

• [Bug] Inconsistent Results After ONNX Runtime Optimization

  #23133 closed Aug 7, 2025

• Inconsistent Results After ONNX Runtime Optimization

  #23142 closed Aug 7, 2025

• [Build] Better support for vcpkg

  #23158 closed Aug 7, 2025

• ONNX 1.17.0 integration remaining work: fix QNN EP test failures

  #23163 closed Aug 7, 2025

• Inconsistent Results After ONNX Runtime Optimization

  #23199 closed Aug 7, 2025

• [Inference Error] The onnx inference result is inconsistent with the numpy inference result

  #23202 closed Aug 7, 2025

• [Build] how to build onnxruntime with openvino EP for android

  #23222 closed Aug 7, 2025

• [Build] Xcode unit tests fail with libc++abi: terminating due to uncaught exception of type onnxruntime::OnnxRuntimeException:

  #23259 closed Aug 7, 2025

• [Build] Building for Mac Catalyst Fails When Installed Via Cocoapods

  #23307 closed Aug 7, 2025

• [Performance] Max operator became 4.5X slower after Fixing NaN propagation for float16 min and max operators.

  #23337 closed Aug 7, 2025

• memory.enable_memory_arena_shrinkage is not working in python

  #23339 closed Aug 7, 2025

• Issue loading custom ONNX model with complex-valued operations in ONNX Runtime (C++)

  #23341 closed Aug 7, 2025

• Memory creeping up

  #23348 closed Aug 7, 2025

• No speedup from float16 with directml compared to cuda

  #23359 closed Aug 7, 2025

• [Build] Possibly unintentional or misconfigured dependencies for QNN EP in onnxruntime_python.cmake

  #23360 closed Aug 7, 2025

• [Performance] GPU Fallback to CPU Without Error When CUDA DLLs Are Missing

  #23372 closed Aug 7, 2025

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

  #23391 closed Aug 7, 2025

• [Performance] Round node shows huge performance drop on Windows

  #23430 closed Aug 7, 2025

• debug result is ok, release get NaN output

  #23440 closed Aug 7, 2025

• [QUESTION]: onnxruntime with onednn backend

  #23543 closed Aug 7, 2025

• [Performance] Speed-up TensorRT engine compilation

  #23546 closed Aug 7, 2025

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

  #23566 closed Aug 7, 2025

• [Performance] ORT-WebGPU Average Pooling is working too long in edge case

  #23614 closed Aug 7, 2025

• TensorRT Provider "Attribute reduction is not supported"

  #23618 closed Aug 7, 2025

• session.disable_fallback() has no effect, it always fallback to cpu

  #23647 closed Aug 7, 2025

• [Build] CMake Error at onnxruntime_unittests.cmake:1026 (find_path): Could not find onnx_SOURCE_DIR using the following files: onnx/onnx-ml.proto3, onnx/onnx-ml.proto Call Stack (most recent call first): CMakeLists.txt:1789 (include)

  #23684 closed Aug 7, 2025

• terminate called after throwing an instance of 'Ort::Exception' what(): Invalid input name: serving_default_input_1:0

  #23730 closed Aug 7, 2025

• Onnxruntime using OpenVINO for older version Intel UHD630

  #23735 closed Aug 7, 2025

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

  #23741 closed Aug 7, 2025

• [Mobile] Onnxruntime react-native issue: [java.lang.ClassCastException: java.lang.String[][] cannot be cast to java.lang.String[]]

  #23782 closed Aug 7, 2025

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

  #23832 closed Aug 7, 2025

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

  #23867 closed Aug 7, 2025

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

  #23878 closed Aug 7, 2025

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

  #23938 closed Aug 7, 2025

• TensorRT Support for Multiple Profiles

  #23965 closed Aug 7, 2025

• [Build] Unsupported AVX512-FP16 Instructions in MLAS (vcvtneeph2ps, vcvtneoph2ps)

  #24025 closed Aug 7, 2025

• Application is getting crashed while creating session for the onnxruntime-qnn with QnnCpu backend option.

  #24082 closed Aug 7, 2025

• ImportError: Unable to import dependency onnxruntime

  #24120 closed Aug 7, 2025

• onnxruntime-mobile implementation on custom execution provider

  #24135 closed Aug 7, 2025

• segmentation fault while using onnxruntime==1.21.0

  #24144 closed Aug 7, 2025

• [Feature Request] A model with dynamic input and dynamic output。 will have a memory leak after inference with Openvino.

  #24162 closed Aug 7, 2025

• Python Session.run_async Causes Program Exit

  #24200 closed Aug 7, 2025

• OpenVINO EP not able to use CPU device

  #24208 closed Aug 7, 2025

• Questions about using AMD VitisAI EP, how can i run my model on AMD NPU?

  #24214 closed Aug 7, 2025

• [Build] OpenVINO ep for macOS

  #24273 closed Aug 7, 2025

• [Build] Building v1.21.0: unsupported instruction 'vpdpbusds'

  #24275 closed Aug 7, 2025

• SIGSEGV when calling OrtSession.run()

  #24288 closed Aug 7, 2025

• [Build] Onnxruntime v1.21.0 fails to build with GCC-13

  #24290 closed Aug 7, 2025

• quantize onnx models to INT8

  #24374 closed Aug 7, 2025

• [Performance] [QNN EP] Performance gap between onnxruntime QNN EP and Genie from QNN SDK.

  #24417 closed Aug 7, 2025

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

  #24570 closed Aug 7, 2025

• [MLAS] Plan to add RISC-V Vector (RVV) support to MLAS

  #24596 closed Aug 7, 2025

• nuget package 1.21.2 causes conflicts in Solutions targeting .NET Framework 4.8

  #24599 closed Aug 7, 2025

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

  #24613 closed Aug 7, 2025

• [DO NOT UNPIN] ORT 1.22.0 Release Candidates available for testing

  #24671 closed Aug 7, 2025

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

  #24824 closed Aug 7, 2025

• Import error in pytest with onnxruntime-directml 1.22.0

  #24907 closed Aug 7, 2025

• [Web] Clarification on wasm/simd vs wasm/simd/threaded default in onnxruntime-web v1.19.0+

  #25666 closed Aug 7, 2025

• [Feature Request] Support for ScatterElements op for QNN-EP

  #22962 closed Aug 7, 2025

• TreeEnsemble can incorrectly decide a root branch is a leaf.

  #24679 closed Aug 6, 2025

• [Build] CMake Error related to onnxruntime_unittests.cmake

  #24972 closed Aug 6, 2025

• onnxruntime custom OP Failure

  #25644 closed Aug 5, 2025

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

  #24920 closed Aug 4, 2025

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

  #24954 closed Aug 4, 2025

• [Documentation]

  #24958 closed Aug 4, 2025

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

  #24873 closed Aug 2, 2025

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

  #24937 closed Aug 2, 2025

• [Build] error: array index 7 is past the end of the array (that has type '__m256[4]')

  #23180 closed Aug 1, 2025

• olive: a weird behavior of a model converted to ONNX format

  #25600 closed Aug 1, 2025