how to support cuda10.2?

[xinsuinizhuan]

System information

• OS Platform and Distribution (e.g., Linux Ubuntu 16.04):windows10
• ONNX Runtime installed from (source or binary):
• ONNX Runtime version: 1.8.1
• Python version:
• Visual Studio version (if applicable): 2019
• GCC/Compiler version (if compiling from source):
• CUDA/cuDNN version: cuda10/v7.6.5.32
• GPU model and memory:6 when i use the gpu to forward, it has the cpu speed, same code run in cuda11.0, it normal, but it tip support the cuda10.2:

@harshithapv harshithapv v1.8.1 96bb4b1 ONNX Runtime v1.8.1

This release contains fixes and key updates for 1.8.0. For all package installation details, please refer to https://www.onnxruntime.ai. Inference

Fixes for GPU package loading issues
Fix for memory issue for models with convolution nodes while using the EXHAUSTIVE algo search mode
ORT Mobile updates
    CoreML EP enabled in iOS mobile package
    Additional operators
    Bug fixes
    React Native package now available

Training

Performance updates for ONNX Runtime for PyTorch (training acceleration for PyTorch models)

Accelerates most popular Hugging Face models as well as GPT-Neo and Microsoft TNLG and TNLU models
Support for PyTorch 1.8.1 and 1.9
Support for CUDA 10.2 and 11.1
Preview packages for ROCm 4.2

[pranavsharma]

Can you please rephrase your question clearly? CUDA build instructions can be found here https://onnxruntime.ai/docs/build/eps.html#cuda.

[xinsuinizhuan]

I want use the onnxruntime to run my yolov5， windows10, VS2019, but my CUDA/cuDNN version is cuda10/v7.6.5.32。I only find the ONNX Runtime v1.8.1 Support for CUDA 10.2, so i use the onnxruntime 1.8.1 -gpu (https://github.com/microsoft/onnxruntime/releases onnxruntime-win-gpu-x64-1.8.1.zip) forward，but the speed is so slow, it should only run cpu。

[pranavsharma]

Newer versions of CUDA and ORT come with perf improvements. You may try building ORT 1.9 release or master branch with the ver of CUDA you have.

[xinsuinizhuan]

how to build? I could not download the cmake external, and i manul download them the put to the file,like this,but still could not find them?

[xinsuinizhuan]

Newer versions of CUDA and ORT come with perf improvements. You may try building ORT 1.9 release or master branch with the ver of CUDA you have.

could you provide a whol projection with all cmake external by the net pan?

[pranavsharma]

Can you use the build cmd here https://onnxruntime.ai/docs/build/eps.html#build-instructions?

[xinsuinizhuan]

Can you use the build cmd here https://onnxruntime.ai/docs/build/eps.html#build-instructions?

Yes, the cmd is ok.But i can't download the cmake/external， so can't build. Could provide the cmake external file?

[xinsuinizhuan]

i download them with the error: i can download them manually, but don't know where to put them

[xinsuinizhuan]

when i dowload the cmake external: lllstandout@LAPTOP-B27R8CMN MINGW64 /d/vison_software/onnxruntime/1.9/onnxruntime (rel-1.9.0) $ git submodule update --init --recursive fatal: could not get a repository handle for submodule 'src/3rdParty/eigen' fatal: could not get a repository handle for submodule 'cmake/external/json' (base)

[xinsuinizhuan]

could provide the whole source? I only want to compile in cuda10.2

[stale[bot]]

This issue has been automatically marked as stale due to inactivity and will be closed in 7 days if no further activity occurs. If further support is needed, please provide an update and/or more details.