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

[Web] Deprecate the onnxruntime-web WebGL backend Open

ep:WebGPU platform:web

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

microsoft/onnxruntime#32241 · Opened Aug 24, 2026
[Build] onnxruntime-node install fails: "Failed to download build list. HTTP status code = 302" (script doesn't follow redirects) Open

.NET build platform:web

Difficulty 3/5 1-2 days Newbie friendliness 78/100

microsoft/onnxruntime#32233 · Opened Aug 24, 2026
[Feature Request] WebGPU LinearAttention state_window for speculative/MTP rollback Open

ep:WebGPU platform:web

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

microsoft/onnxruntime#32232 · 1 comment · Opened Aug 24, 2026
[Bug] GemmTransposeFusion folds an identity Transpose (perm=[0,1]) into Gemm transA/transB, producing silently wrong results Open

ep:CoreML

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

microsoft/onnxruntime#32228 · Opened Aug 24, 2026
[Feature Request] Streaming model I/O for large ONNX models Open

ep:WebNN feature request

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

microsoft/onnxruntime#32224 · 2 comments · Opened Aug 23, 2026
[Performance] `import onnxruntime` takes ~9 s on Windows since 1.22.0 (was ~0.2 s in 1.21.1), no model involved Open

platform:windows

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

microsoft/onnxruntime#32220 · 1 comment · Opened Aug 22, 2026
[Performance] CoreML EP: ~10x inference slowdown on iOS in 1.29 vs 1.28 (ConvNeXt-Tiny FP16) Open

api:CSharp ep:CoreML performance platform:mobile

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

microsoft/onnxruntime#32218 · Opened Aug 22, 2026
[Performance] CoreML EP: MatMulAddFusion forces Gemm weights inline into model.mil instead of weights/weight.bin (3.7 GB MIL, 17.3 GiB peak RSS) Open

ep:CoreML

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

microsoft/onnxruntime#32212 · 1 reaction · Opened Aug 21, 2026
`optimization.disable_specified_optimizers` is documented as comma-separated but is parsed on `;` Open

Difficulty 1/5 Under an hour Newbie friendliness 88/100

microsoft/onnxruntime#32211 · Opened Aug 21, 2026
[Web] v1.28 and v1.29 not released in cdn.jsdelivr.net/npm/onnxruntime-web@<version>/dist/ort.min.js Open

.NET platform:web

microsoft/onnxruntime#32195 · 1 assignee · Opened Aug 21, 2026
Where returns +0.0 for a selected -0.0, on both the X and Y branches (CPU EP) Open

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

microsoft/onnxruntime#32191 · 1 reaction · Opened Aug 20, 2026
[WebGPU Plugin EP] GetSharedAllocator returns null after registration despite exposing DEFAULT MemoryInfo Open

ep:WebGPU platform:web stale

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

microsoft/onnxruntime#32164 · 1 comment · Opened Aug 19, 2026
[Build] TRT CUDA minimal build fails with a link error Open

build ep:CUDA ep:TensorRT

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

microsoft/onnxruntime#32154 · 3 comments · Opened Aug 18, 2026
[Build] WebGPU EP fails to build for iOS: Dawn's ObjCUtils.mm is incompatible with ORT's forced ObjC ARC Open

ep:WebGPU platform:mobile platform:web stale

Difficulty 3/5 1-2 days Newbie friendliness 72/100

microsoft/onnxruntime#32147 · 4 comments · Opened Aug 18, 2026
DirectML EP miscomputes the Split -> Sigmoid -> Mul pattern emitted by nn.GLU Open

ep:DML stale

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

microsoft/onnxruntime#32146 · 1 comment · Opened Aug 18, 2026
[WebGPU] Non-finite output with ORT_ENABLE_EXTENDED graph optimization (minimal 8-node repro) Open

ep:WebGPU platform:web stale

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

microsoft/onnxruntime#32145 · 1 comment · Opened Aug 18, 2026
[Web] WebGPU EP: intermediate buffer reused before its consumer reads it — corrupted Div output (int64 Sub→Unsqueeze→Cast→Div chain + parallel ReduceMean branch) Open

ep:WebGPU platform:web stale

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

microsoft/onnxruntime#32134 · 1 comment · Opened Aug 17, 2026
# [Bug] QDQ optimization changes numeric results: fused integer-domain ops (QLinearAdd/QLinearConv/QLinearAveragePool) off by one or more quantization steps (diff == k * scale) Open

quantization

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

microsoft/onnxruntime#32132 · 2 comments · Opened Aug 17, 2026
CompileModel drops outer-graph initializers consumed only from a subgraph, then rejects its own output Open

ep:DML

Difficulty 3/5 1-2 days Newbie friendliness 75/100

microsoft/onnxruntime#32131 · 2 comments · Opened Aug 17, 2026
Constant folding silently disabled for GreaterOrEqual / LessOrEqual below opset 16 in 1.27 (regression from the new output-size guard) Open

ep:VitisAI stale

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

microsoft/onnxruntime#32130 · 1 comment · Opened Aug 17, 2026
[Bug] ReshapeFusion: fused Reshape used as shape input → type error (tensor(float) vs tensor(int64)) Open

stale

Difficulty 3/5 1-2 days Newbie friendliness 72/100

microsoft/onnxruntime#32105 · 1 comment · Opened Aug 15, 2026
CUDA EP: Slice with a runtime bound returns its input untrimmed, breaking a downstream ScatterND Open

ep:CUDA stale

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

microsoft/onnxruntime#32083 · 1 comment · Opened Aug 14, 2026
UnicodeDecodeError instead of actual error message when DlirectML fails on non-English Windows Open

ep:DML platform:windows

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

microsoft/onnxruntime#32082 · 1 comment · Opened Aug 14, 2026
[Build] croos compile Open

build stale

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

microsoft/onnxruntime#32064 · 1 comment · 1 reaction · Opened Aug 13, 2026
CPU ScatterElements always copies its data input, even when the producer's buffer could be reused in place Open

microsoft/onnxruntime#32062 · 1 reaction · 2 assignees · Opened Aug 13, 2026
CUDA ScatterElements reductions dispatch on element size, giving wrong results for several registered types Open

ep:CUDA

microsoft/onnxruntime#32061 · 1 comment · 1 reaction · 2 assignees · Opened Aug 13, 2026
Issue in the Implementation for honoring model_external_initializers_file_folder_path for file-path model loads Open

ep:OpenVINO stale

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

microsoft/onnxruntime#32054 · 2 comments · Opened Aug 13, 2026
[Feature Request] Release the retained NodeProtos after session initialization Open

feature request

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

microsoft/onnxruntime#32047 · Opened Aug 12, 2026
[Web]BucketCacheManager per-bucket limits are far below LLM KV-tensor counts, causing permanent ~3x decode degradation for dynamic-KV workloads Open

ep:WebGPU platform:web stale

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

microsoft/onnxruntime#32017 · 3 comments · Opened Aug 12, 2026
SimplifiedLayerNormFusion Incorrectly Fuses Downstream Mul with Shape-Incompatible Scale, Causing Runtime Shape Error Open

microsoft/onnxruntime#31980 · 1 reaction · 2 assignees · Opened Aug 11, 2026
[QNN EP] ep.context_embed_mode=1 generates an unloadable EPContext model for multi-partition graphs (embed_mode=0 works) Open

ep:QNN model:transformer

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

microsoft/onnxruntime#31977 · 1 comment · Opened Aug 11, 2026
[OpenVINO][Windows] onnxruntime_provider_test silently exits during MatMulNBits.Float32_4b_Accuracy0 Open

ep:oneDNN ep:OpenVINO platform:windows

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

microsoft/onnxruntime#31760 · 1 comment · Opened Aug 8, 2026
[BUG] Sparse initializers causes Type Error for SparseToDenseMatMul Open

stale

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

microsoft/onnxruntime#31730 · 1 comment · Opened Aug 8, 2026
[MIGraphX EP] Fails processing models with multiple layers of sub-graphs with "Unknown operator: MGXKernel_subgraph_tf2onnx_9255412890226604894_0" Open

ep:MIGraphX stale

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

microsoft/onnxruntime#31713 · 7 comments · Opened Aug 7, 2026
[Java] String-array input paths never delete the local references returned by GetObjectArrayElement() Open

api:Java stale

Difficulty 4/5 3-5 days Newbie friendliness 68/100

microsoft/onnxruntime#31707 · 3 comments · Opened Aug 7, 2026
[Java] copyStringTensorToArray() leaks tempBuffer on three early-return paths, and does not delete the jstring it creates per element Open

api:Java stale

Difficulty 3/5 1-2 days Newbie friendliness 72/100

microsoft/onnxruntime#31705 · 1 comment · Opened Aug 7, 2026
WebGPU EP crashes on macOS when running concurrent multi threaded inference Open

ep:CoreML ep:WebGPU platform:web

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

microsoft/onnxruntime#31627 · 4 comments · Opened Aug 4, 2026
[WebGPU EP] fp16 overflow in a decomposed LayerNorm: whisper-small q4f16 produces +Inf variance in the encoder and the decoder loops Open

ep:WebGPU platform:web

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

microsoft/onnxruntime#31626 · 2 comments · Opened Aug 4, 2026
[MIGraphX EP] Segfaults when reading some SSD MobileNet v1 models Open

ep:MIGraphX stale

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

microsoft/onnxruntime#31607 · 3 comments · Opened Aug 3, 2026
[Performance] DirectML EP uses ~32 GB VRAM and ~167 s to run HT-Demucs ONNX model (7.8 s input) on RTX 5090 Open

ep:DML performance stale

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

microsoft/onnxruntime#31482 · 4 comments · Opened Aug 2, 2026
[Feature Request] Can you make it compatible with pypy3 Open

feature request

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

microsoft/onnxruntime#31351 · Opened Aug 1, 2026
Apply compile-time kernel registration table validation to remaining execution providers Open

ep:ACL ep:CANN ep:MIGraphX ep:SNPE ep:WebNN ep:Xnnpack

microsoft/onnxruntime#31348 · 2 comments · 1 assignee · Opened Jul 31, 2026
[Feature Request] NodeJS dependency hygiene Open

feature request

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

microsoft/onnxruntime#31193 · Opened Jul 30, 2026
ORT CUDA Constant Folding Produces NaN from `Reciprocal(Sqrt(0)) * 0.0` Open

ep:CUDA

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

microsoft/onnxruntime#31163 · 1 comment · Opened Jul 30, 2026
CUDA ArgMax on all -inf returns INT32_MAX sentinel value Open

ep:CUDA stale

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

microsoft/onnxruntime#31161 · 3 comments · Opened Jul 30, 2026
LayerNorm fusion selects the Div output as scale, failing session creation even with keepdims=1 Open

stale

Difficulty 3/5 1-2 days Newbie friendliness 55/100

microsoft/onnxruntime#31147 · 1 comment · Opened Jul 29, 2026
ORT CUDA optimizer's constant-folding pass produces a result that disagrees with the runtime kernel. Open

ep:CUDA stale

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

microsoft/onnxruntime#31140 · 1 comment · Opened Jul 29, 2026
[Graph Optimizer] LayerNorm fusion incorrectly matches ReduceMean with keepdims=0, causing 'Node input not found' error Open

stale

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

microsoft/onnxruntime#30513 · 1 comment · Opened Jul 29, 2026
onnxruntime-node: bump adm-zip to >=0.6.0 to resolve GHSA-xcpc-8h2w-3j85 Open

platform:web stale

Difficulty 3/5 1-2 days Newbie friendliness 65/100

microsoft/onnxruntime#30125 · 2 comments · Opened Jul 28, 2026
[Feature Request] Drive EP conformance coverage from the provider registry and assert every compiled EP is represented Open

ep:CANN ep:CoreML ep:DML ep:MIGraphX ep:OpenVINO ep:QNN ep:SNPE ep:WebGPU ep:Xnnpack feature request platform:web

microsoft/onnxruntime#29914 · 1 assignee · Opened Jul 28, 2026

microsoft/onnxruntime

Get new issues in your inbox