yolov5 make compilation error

[eladpar]

Env

• GPU, NVIDIA GeForce RTX 3050 Ti
• OS, ubuntu 20.04
• CUDA Version: 11.7
• TensorRT 8.5.2.2-1+cuda11.8

About this repo

• which branch/tag/commit are you using? master
• which model? yolov5, retinaface? yolov5s.pt

Your problem

im building using this https://wiki.seeedstudio.com/YOLOv5-Object-Detection-Jetson/#inference-on-jetson-device toturial when i hit the make command i get this ouput :

Scanning dependencies of target myplugins
[  7%] Building CUDA object CMakeFiles/myplugins.dir/plugin/yololayer.cu.o
/usr/local/cuda/include/cuda_runtime_api.h(13213): error: attribute "__host__" does not apply here

/usr/local/cuda/include/cuda_runtime_api.h(13213): error: expected a ";"

/home/eladpar/tensorrtx/yolov5/plugin/yololayer.h(87): warning: function "nvinfer1::IPluginV2Ext::configurePlugin(const nvinfer1::Dims *, int32_t, const nvinfer1::Dims *, int32_t, const nvinfer1::DataType *, const nvinfer1::DataType *, const __nv_bool *, const __nv_bool *, nvinfer1::PluginFormat, int32_t)" is hidden by "nvinfer1::YoloLayerPlugin::configurePlugin" -- virtual function override intended?

2 errors detected in the compilation of "/tmp/tmpxft_000369d4_00000000-6_yololayer.cpp1.ii".
make[2]: *** [CMakeFiles/myplugins.dir/build.make:63: CMakeFiles/myplugins.dir/plugin/yololayer.cu.o] Error 1
make[1]: *** [CMakeFiles/Makefile2:163: CMakeFiles/myplugins.dir/all] Error 2
make: *** [Makefile:84: all] Error 2

[wang-xinyu]

Can you check your nvcc? Try this command nvcc --version.

[eladpar]

Can you check your nvcc? Try this command nvcc --version.

nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2019 NVIDIA Corporation
Built on Sun_Jul_28_19:07:16_PDT_2019
Cuda compilation tools, release 10.1, V10.1.243

[wang-xinyu]

Here the nvcc is 10.1, but your CUDA is 11.8. You need to use /usr/local/cuda/bin/nvcc.

[eladpar]

Here the nvcc is 10.1, but your CUDA is 11.8. You need to use /usr/local/cuda/bin/nvcc.

can you elaborate on what needs to be done? beacuse as seen in this question : https://stackoverflow.com/a/53504578 the cuda seen in nvidia-smi and nvcc-version don't have to be the same?

[wang-xinyu]

Thing is you have two nvcc in your OS. Can you check your nvcc path by running which nvcc? And also check if /usr/local/cuda/bin/nvcc is in your system?

[eladpar]

Thing is you have two nvcc in your OS. Can you check your nvcc path by running which nvcc? And also check if /usr/local/cuda/bin/nvcc is in your system?

my nvcc is located in /usr/bin/nvcc there is no nvcc located in /usr/local/cuda/bin/nvcc

[wang-xinyu]

Where are your CUDA libs and bins installed?

[stale[bot]]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.