Search results

[WebGPU] Incorrect output shape for granite-4.0-1b-speech audio encoder

ep:WebGPUort-web webgpu providerort-web webgpu provider

platform:webissues related to ONNX Runtime web; typically submitted using templateissues related to ONNX Runtime web; typically submitted using template

Status: Open.

#27661 In microsoft/onnxruntime;

· xenova opened on Mar 15, 2026

[Feature Request] [Java] Configurable Native Library Extraction

api:Javaissues related to the Java APIissues related to the Java API

feature requestrequest for unsupported feature or enhancementrequest for unsupported feature or enhancement

Status: Open.

#27656 In microsoft/onnxruntime;

· EPNW-Eric opened on Mar 14, 2026

[Java] Redundant library extraction when onnxruntime.native.path is set

api:Javaissues related to the Java APIissues related to the Java API

Status: Open.

#27655 In microsoft/onnxruntime;

· EPNW-Eric opened on Mar 14, 2026

Error: "Invalid rank for input: input Got: 4 Expected: 5"

ep:DMLissues related to the DirectML execution providerissues related to the DirectML execution provider

Status: Open.

#27654 In microsoft/onnxruntime;

· ThreeDeeJay opened on Mar 14, 2026

v1.24.x in pypi does not have wheels for py3.10

Status: Open.

#27653 In microsoft/onnxruntime;

· AbdealiLoKo opened on Mar 14, 2026

[Performance] CPU Conv kernel regression for large kernels (9x9) on macOS ARM64 in ORT 1.24.x

ep:CoreMLissues related to CoreML execution providerissues related to CoreML execution provider

ep:Xnnpackissues related to XNNPACK EPissues related to XNNPACK EP

performanceissues related to performance regressionsissues related to performance regressions

Status: Open.

#27633 In microsoft/onnxruntime;

· AIwork4me opened on Mar 12, 2026

NvTensorRTRTXExecutionProvider::GetCapability claims EPContext nodes belonging to other EPs, causing crash on multi-GPU systems

ep:NvRTXNV RTX execution providerNV RTX execution provider

ep:OpenVINOissues related to OpenVINO execution providerissues related to OpenVINO execution provider

ep:QNNissues related to QNN exeution providerissues related to QNN exeution provider

ep:TensorRTissues related to TensorRT execution providerissues related to TensorRT execution provider

Status: Open.

Bug

#27622 In microsoft/onnxruntime;

· nieubank opened on Mar 11, 2026

[CUDA EP] Silent deadlock when calling InferenceSession.run() from Python thread on Blackwell GPU (RTX 5060, sm_120, Windows)

ep:CUDAissues related to the CUDA execution providerissues related to the CUDA execution provider

platform:windowsissues related to the Windows platformissues related to the Windows platform

Status: Open.

#27621 In microsoft/onnxruntime;

· azizketata opened on Mar 11, 2026

Investigate perf implication for storing tranposed weights

Status: Open.

#27619 In microsoft/onnxruntime;

· justinchuby opened on Mar 11, 2026

fetching 1 output when a model has multiple outputs computes all of them

Status: Open.

#27615 In microsoft/onnxruntime;

· guschmue opened on Mar 10, 2026

Use concat to ONNX attention cache update

Status: Open.

#27612 In microsoft/onnxruntime;

· titaiwangms opened on Mar 10, 2026

cudaErrorNoKernelImageForDevice

Status: Open.

#27600 In microsoft/onnxruntime;

· MathieuDuponchelle opened on Mar 9, 2026