microsoft / microsoft/onnxruntime

No nuget packages when building multiple EPs

Open
#6,857 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

build feature request
Dominant language
C++
Stars
21.9k
Forks
4.2k
Avg merge
4d 11h
Merged PRs (30d)
184

Description

**Describe the bug**
When building for source with a combinations of EPs such as the example below only on nugets for one EP will be built.

**Urgency**
No strict urgency, I would like to be able to use the command below. Perhaps this is intended, you cannot combine multiple EPs and nuget. Then I would like the build to fail earlier if it is the case.

**System information**
- Windows 10
- ONNX Runtime built from source
- ONNX Runtime version: 2d6e10ba0054bfbbbc89f1e6d945b4ca3d59631f
- Python version: 3.9.2
- Visual Studio version (if applicable): 16.8.6
- GCC/Compiler version (if compiling from source): CMake 3.19.5
- CUDA/cuDNN version: 10.2.89 / TensorRT 7.1.3.4
- GPU model and memory: RTX 3070 8gb

**To Reproduce**
Running build command:
```powershell
./build.bat --config RelWithDebInfo `
--build_shared_lib --build_nuget --parallel --use_openmp `
--cuda_version 10.2 --cuda_home "C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.2" `
--cudnn_home "C:\Users\thomas\bin\cudnn" `
--use_tensorrt --tensorrt_home "C:\Users\thomas\bin\TensorRT" `
--use_dnnl --use_dml --cmake_generator "Visual Studio 16 2019" --skip_tests
```
nuget.exe in `PATH`.

**Expected behavior**
Build nuget packages for each EPs as if the build command had been run separately for each EP.
Now the two below are the only nugets built.
```powershell
> ls .\build\Windows\RelWithDebInfo\RelWithDebInfo\nuget-artifacts\ | select name

Name
----
Microsoft.ML.OnnxRuntime.Managed.1.7.0-dev-20210301-0714-2d6e10ba0.nupkg
Microsoft.ML.OnnxRuntime.TensorRT.1.7.0-dev-20210301-0714-2d6e10ba0.nupkg
```
I would expect all the relevant nugets to be built:
```
Microsoft.ML.OnnxRuntime.1.7.0-dev-20210301-0714-2d6e10ba0.nupkg # from DML
Microsoft.ML.OnnxRuntime.DNNL.1.7.0-dev-20210301-0714-2d6e10ba0.nupkg
Microsoft.ML.OnnxRuntime.GPU.1.7.0-dev-20210301-0714-2d6e10ba0.nupkg
...
```

**Additional context**
The DML related packages are built (or however they are created)
```powershell
> ls .\build\Windows\packages\ | select name

Name
----
GoogleTestAdapter.0.17.1
Microsoft.AI.DirectML.1.4.2
Microsoft.Windows.CppWinRT.2.0.201113.7
```

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start with build.bat and the NuGet packaging flow that produces build/Windows/RelWithDebInfo/RelWithDebInfo/nuget-artifacts, using the provided multi-EP command as the reproduction case. Compare the packaging behavior for DML, DNNL, CUDA, and TensorRT options; done means all applicable NuGet packages are produced, or the build fails early when combinations are unsupported.

Written by the indexing model from the issue text.

Assessment

Tech stack
cmake, cpp
Domain
build-system, release
Issue type
Bug
Difficulty
4/5
Estimated time
3-5 days
Activity status
Stale
Clarity
Mostly clear
Newbie friendliness
35/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.