Repository Issue Activity (beta)

#14676 [Performance] running on xavier gpu but cpu usage high

#20828 [Mobile] android prod crash: signal 11 (SIGSEGV), code 1 (SEGV_MAPERR)

#28812 cannot import onnxruntime:DLL load fail while importing onnxruntime_pybind11_state:A dynamic link library(DLL) initialization routine fail.

#28957 [Performance] ONNX Inherited Paccel

#28023 [Build] CUDA 13.2

#29538 [Performance] Large CPU memory regression starting in 1.24 (Java, macOS arm64) for fully-convolutional models with dynamic input shapes

#29635 com.microsoft:QMoE (int4) produces incorrect results on the NvTensorRTRTX EP

#31730 [BUG] Sparse initializers causes Type Error for SparseToDenseMatMul

#31707 [Java] String-array input paths never delete the local references returned by GetObjectArrayElement()

#31573 [MLAS] ARM64 non-dotprod SymmQgemm NEON kernel produces wrong results for extreme int8 operands (int16 intermediate overflow)

#26722 [Meta] WebNN - Windows ML / ONNX Runtime Issues and Feature Requests

#28540 [Mobile] XNNPACK + `ORT_ENABLE_ALL` invalid graph after `ConvMulFusion` (missing input `node_28`)

#28657 [Bug] ORT RN Expo config plugin incompatible with Expo 55+

#28672 [Mobile] Expo config plugin does not support hoisted monorepositories

#28895 [WebGPU] Support 2-bit quantization in GatherBlockQuantized

#28898 [WebNN][ORT Default CPU] WebNN maxPool2d test runs crash with "Exit code: RESULT_CODE_GPU_EXIT_ON_CONTEXT_LOST"

#29270 Android: Linux-targeted libonnxruntime.so (from runtimes/linux-arm64) gets bundled into APK instead of the Android AAR's library

#29507 [WebGPU EP] Leak Detector failure when running GenAI LLM models

#29508 [WebGPU EP] Meta-Llama-3.1-8B inference crash on QNN environments

#29535 [Build] Build of 1.26.0 fails on Windows with CUDA 12.9 and sm_120

#31713 [MIGraphX EP] Fails processing models with multiple layers of sub-graphs with "Unknown operator: MGXKernel_subgraph_tf2onnx_9255412890226604894_0"

#31198 [Mobile] java.lang.NoSuchMethodError: no non-static method "Lai/onnxruntime/TensorInfo;.<init>([J[Ljava/lang/String;I)V"

#31706 [Java] CUDA/TensorRT applyToNative() releases the value string's UTF chars with the key string's jstring

#31482 [Performance] DirectML EP uses ~32 GB VRAM and ~167 s to run HT-Demucs ONNX model (7.8 s input) on RTX 5090

#26541 [Reduction] Extend noop_with_empty_axes behavior to non-CPU EPs