microsoft/onnxruntime

View on GitHub

ONNX Runtime: cross-platform, high performance ML inferencing and training accelerator

Stars: 21.9k
Forks: 4.2k
Open beginner issues: 8
Indexed issues: 835
Avg merge: 4d 11h
Merged PRs (30d): 184

Dominant language: C++
License: MIT
Last GitHub push: Sep 19, 2026
Latest indexed: Sep 19, 2026
Contributing guide: Contributing guide
Code of conduct: Code of conduct
Beginner labels: No beginner labels indexed

Beginner friendly Everything

Open All

835 open issues indexed

TensorRT EP produces incorrect results with CUDA I/O Binding when input is populated asynchronously without an explicit stream dependency Open

ep:CUDA ep:TensorRT

Difficulty 4/5 3-5 days Newbie friendliness 38/100

microsoft/onnxruntime#32673 · Opened Sep 17, 2026
GatherBlockQuantized CUDA kernel uses the flat index for the block id, giving wrong results when the quantized axis is not a multiple of block_size Open

ep:CUDA quantization

microsoft/onnxruntime#32663 · 1 reaction · 2 assignees · Opened Sep 16, 2026
(Berikut) Cara Lengkap Buka Blokir Akun Bank Jago) adalah 0831-6926-6384 Open

platform:mobile

Difficulty 1/5 Under an hour Newbie friendliness 5/100

microsoft/onnxruntime#32640 · Opened Sep 16, 2026
(Berikut) Cara Lengkap Buka Blokir Akun Bank Jago) adalah 0831-6926-6384 Open

platform:mobile

Difficulty 5/5 Over a week Newbie friendliness 1/100

microsoft/onnxruntime#32639 · Opened Sep 16, 2026
(Berikut) Cara Lengkap Buka Blokir Akun Bank Jago) adalah 0831-6926-6384 Open

platform:mobile

Difficulty 1/5 Under an hour Newbie friendliness 1/100

microsoft/onnxruntime#32638 · Opened Sep 16, 2026
(Berikut) Cara Lengkap Buka Blokir Akun Bank Jago) adalah 0831-6926-6384 Open

platform:mobile

Difficulty 5/5 Over a week Newbie friendliness 1/100

microsoft/onnxruntime#32637 · Opened Sep 16, 2026
(Berikut) Cara Lengkap Buka Blokir Akun Bank Jago) adalah 0831-6926-6384 Open

platform:mobile

Difficulty 5/5 Over a week Newbie friendliness 1/100

microsoft/onnxruntime#32636 · Opened Sep 16, 2026
[WebGPU] BEN2-ONNX `Failed to create a WebGPU compute pipeline: [Invalid ShaderModule "LayerNorm"] is invalid due to a previous error.` Open

.NET ep:WebGPU model:transformer platform:web

Difficulty 4/5 3-5 days Newbie friendliness 48/100

microsoft/onnxruntime#32627 · 1 comment · Opened Sep 16, 2026
[Feature Request] Support the WebGPU EP as a fallback for the OpenVINO NPU Open

ep:OpenVINO ep:WebGPU ep:WebNN feature request platform:web

Difficulty 5/5 Over a week Newbie friendliness 25/100

microsoft/onnxruntime#32625 · 1 reaction · Opened Sep 16, 2026
[Feature Request] Add `py.typed` and type annotations to Python package Open

feature request

Difficulty 4/5 3-5 days Newbie friendliness 50/100

microsoft/onnxruntime#32615 · 3 comments · Opened Sep 15, 2026
LayerNormFusion should reject shape-expanding scale and bias patterns Open

Difficulty 4/5 3-5 days Newbie friendliness 55/100

microsoft/onnxruntime#32603 · Opened Sep 15, 2026
EmbeddingGemma-300M fp16 CPU inference regression: NaN output on Linux aarch64 since 1.24.1 Open

api:Java

microsoft/onnxruntime#32600 · 1 comment · 1 reaction · 2 assignees · Opened Sep 14, 2026
[Web] JSEP DequantizeLinear: wrong results for uint8 per-tensor quantization (bad vec4 indexing + unsigned subtraction wraparound) Open

ep:WebGPU platform:web quantization

Difficulty 3/5 1-2 days Newbie friendliness 74/100

microsoft/onnxruntime#32578 · Opened Sep 13, 2026
[CoreML] Wrong FP16 output with RequireStaticInputShapes=1 on macOS 15 ARM64 Open

ep:CoreML

Difficulty 5/5 Over a week Newbie friendliness 30/100

microsoft/onnxruntime#32569 · Opened Sep 12, 2026
[MIGraphX EP] Models with 10+ outputs fail with "invalid input for conversion to integer": output name "main:#output_:00010" is not parsed Open

ep:MIGraphX

Difficulty 3/5 1-2 days Newbie friendliness 76/100

microsoft/onnxruntime#32567 · Opened Sep 12, 2026
WebGPU EP: concurrent Session.Run() across two sessions segfaults (Linux/Vulkan, same or different GPU) Open

.NET api:CSharp ep:WebGPU platform:web

microsoft/onnxruntime#32561 · 2 comments · 1 assignee · Opened Sep 11, 2026
CUDA plugin EP GetEpDevices() finds no GPU on Jetson Open

ep:CUDA platform:jetson

microsoft/onnxruntime#32541 · 4 comments · 1 reaction · 2 assignees · Opened Sep 10, 2026
[Feature Request] [Go] Tag the module and document native runtime setup Open

Difficulty 4/5 3-5 days Newbie friendliness 48/100

microsoft/onnxruntime#32535 · 1 reaction · Opened Sep 10, 2026
[Mobile] [QNN EP GPU] Noisy ConvTranspose output differs completely from CPU output with same input Open

ep:QNN platform:mobile

Difficulty 4/5 3-5 days Newbie friendliness 30/100

microsoft/onnxruntime#32506 · 3 comments · Opened Sep 9, 2026
`import onnxruntime` segfaults in 1.29.0 on containers without `/bin` or `/usr/bin` (works in 1.28.0) Open

microsoft/onnxruntime#32501 · 1 comment · 1 reaction · 2 assignees · Opened Sep 9, 2026
[Feature Request] Add Fusion Transformers for WebNN EP Decomposed LM Operators Open

ep:WebGPU ep:WebNN feature request model:transformer platform:web

Difficulty 5/5 Over a week Newbie friendliness 38/100

microsoft/onnxruntime#32496 · 4 comments · 2 reactions · Opened Sep 9, 2026
[Mobile] Could not activate JNI Handle 0x7ff647d7d0 (key_handle 0xd084f07) of Java type 'crc646e4b5c24e03cdbf9/MainActivity' Open

.NET api:CSharp api:Java platform:mobile

Difficulty 4/5 3-5 days Newbie friendliness 35/100

microsoft/onnxruntime#32495 · Opened Sep 9, 2026
CUDA Mod kernel has no zero divisor check, unlike CPU after #27833 Open

ep:CUDA

microsoft/onnxruntime#32494 · 1 reaction · 2 assignees · Opened Sep 9, 2026
import of onnxruntime on MacOS while undex Codex filesystem sandbox leaves :memory:.ses file in working directory Open

Difficulty 3/5 1-2 days Newbie friendliness 48/100

microsoft/onnxruntime#32476 · 2 comments · Opened Sep 8, 2026
[Feature Request] Activate existing RISCV64 QEMU build/test path in CI Open

feature request platform:web

Difficulty 4/5 3-5 days Newbie friendliness 55/100

microsoft/onnxruntime#32465 · Opened Sep 7, 2026
XNNPACK EP: concurrent InferenceSession creation races the lazily-created stored allocator (use-after-free) Open

ep:Xnnpack platform:mobile

Difficulty 4/5 3-5 days Newbie friendliness 65/100

microsoft/onnxruntime#32461 · 4 comments · Opened Sep 7, 2026
[Feature Request] [Web] Add Dynamic linking Open

ep:WebGPU ep:WebNN ep:Xnnpack feature request platform:web

Difficulty 5/5 Over a week Newbie friendliness 45/100

microsoft/onnxruntime#32450 · Opened Sep 5, 2026
[Web] WebGPU EP: DETR-family model creates a session but first run() throws "Invalid dimension of 4294967295 for SizeToDimension" Open

ep:WebGPU platform:web

Difficulty 4/5 3-5 days Newbie friendliness 48/100

microsoft/onnxruntime#32438 · Opened Sep 4, 2026
[Build] CMake configuration fails with CUDA when setting CMAKE_CUDA_ARCHITECTURES=75-virtual Open

build ep:CUDA

Difficulty 3/5 1-2 days Newbie friendliness 64/100

microsoft/onnxruntime#32436 · Opened Sep 4, 2026
GemmTransposeFusion can create dimension-mismatched Gemm for identity Transpose input Open

Difficulty 4/5 3-5 days Newbie friendliness 45/100

microsoft/onnxruntime#32417 · Opened Sep 3, 2026
Basic optimization can produce invalid graph for same-dtype Cast feeding folded Mul-one Open

Difficulty 4/5 3-5 days Newbie friendliness 52/100

microsoft/onnxruntime#32416 · Opened Sep 3, 2026
Basic optimization can produce invalid graph for Dropout feeding folded Mul-one Open

Difficulty 4/5 3-5 days Newbie friendliness 52/100

microsoft/onnxruntime#32414 · Opened Sep 3, 2026
Basic optimization can produce invalid graph for Identity feeding folded Mul-one Open

Difficulty 4/5 3-5 days Newbie friendliness 55/100

microsoft/onnxruntime#32413 · Opened Sep 3, 2026
[Performance] DFT/STFT CPU performance is 100-700x slower than PyTorch FFT for batched short transforms Open

performance

Difficulty 4/5 3-5 days Newbie friendliness 52/100

microsoft/onnxruntime#32403 · 3 comments · Opened Sep 2, 2026
UnregisterExecutionProviderLibrary can access torn-down CRT state during Windows process shutdown Open

platform:windows

Difficulty 4/5 3-5 days Newbie friendliness 25/100

microsoft/onnxruntime#32402 · Opened Sep 2, 2026
[Build] --build_wasm_static_lib with --use_xnnpack produces libonnxruntime_webassembly.a with undefined xnn_*/pthreadpool_* symbols Open

build ep:Xnnpack platform:web

Difficulty 4/5 3-5 days Newbie friendliness 72/100

microsoft/onnxruntime#32396 · Opened Sep 2, 2026
[Feature Request] Integrate with ONNX 1.23.0 release branch Open

feature request

Difficulty 5/5 Over a week Newbie friendliness 35/100

microsoft/onnxruntime#32390 · Opened Sep 2, 2026
[Feature Request] XNNPACK EP: allow Conv/ConvTranspose with symbolic spatial dims where kernel creation doesn't depend on them Open

ep:Xnnpack feature request

Difficulty 4/5 3-5 days Newbie friendliness 52/100

microsoft/onnxruntime#32387 · Opened Sep 2, 2026
GeluFusion emits com.microsoft.Gelu(float16) on linux/arm64 where no float16 kernel is registered, so fp16 Erf-GELU models fail to load Open

Difficulty 4/5 3-5 days Newbie friendliness 48/100

microsoft/onnxruntime#32386 · Opened Sep 2, 2026
[Performance] CUDA MatMulNBits M=1 GEMV is grid-starved at decode shapes: 16-62% of HBM bandwidth Open

ep:CUDA

Difficulty 5/5 Over a week Newbie friendliness 25/100

microsoft/onnxruntime#32382 · Opened Sep 2, 2026
MIGraphX EP: `ORT_MIGRAPHX_INT8_CALIBRATION_TABLE_NAME` env variable seems do not work Open

ep:MIGraphX

Difficulty 2/5 1-3 hours Newbie friendliness 74/100

microsoft/onnxruntime#32352 · 1 comment · Opened Sep 1, 2026
CUDA EP: intermediate tensors are allocated without a stream, so the arena can hand them to another stream while work is still queued Open

ep:CUDA

Difficulty 5/5 Over a week Newbie friendliness 35/100

microsoft/onnxruntime#32342 · Opened Aug 31, 2026
Register Level-1 workspace estimators with kernel metadata Open

microsoft/onnxruntime#32341 · 1 assignee · Opened Aug 31, 2026
softmax_warp_forward overflows int32 above 2^31 elements and can return silently wrong results (CUDA and ROCm EPs) Open

ep:CUDA

Difficulty 4/5 3-5 days Newbie friendliness 48/100

microsoft/onnxruntime#32299 · 2 comments · Opened Aug 27, 2026
XNNPACK EP on Android arm64 aborts the process (Abort message: 'terminating') instead of returning an error — Resize with empty scales, still in 1.26.0 Open

.NET api:CSharp ep:Xnnpack platform:mobile

Difficulty 5/5 Over a week Newbie friendliness 45/100

microsoft/onnxruntime#32298 · Opened Aug 27, 2026
[Build] Align CPython 3.14 CUDA 13 wheels with TensorRT 11 Open

ep:CUDA ep:TensorRT

Difficulty 4/5 3-5 days Newbie friendliness 52/100

microsoft/onnxruntime#32278 · 1 comment · Opened Aug 26, 2026
TensorRT EP Resize with +inf input returns NaN while CPU/CUDA EP return +inf Open

ep:CUDA ep:TensorRT

Difficulty 4/5 3-5 days Newbie friendliness 48/100

microsoft/onnxruntime#32258 · Opened Aug 25, 2026
CUDA EP ReduceMax/ReduceMin return finite limits instead of +/-inf for all-infinite inputs Open

ep:CUDA

Difficulty 4/5 3-5 days Newbie friendliness 48/100

microsoft/onnxruntime#32256 · 1 comment · Opened Aug 25, 2026
TensorRT EP engine cache reuses engine for same graph name with different static input shape Open

ep:TensorRT

Difficulty 4/5 3-5 days Newbie friendliness 52/100

microsoft/onnxruntime#32254 · 1 comment · Opened Aug 25, 2026
cuda graph bug for recurrent state inplace update Open

ep:CUDA

Difficulty 4/5 3-5 days Newbie friendliness 48/100

microsoft/onnxruntime#32243 · Opened Aug 24, 2026

microsoft/onnxruntime

Get new issues in your inbox