Compile error: 'cuda.h'

[Yuliang-Zou]

Not really an issue, just want to share my experience.

If you are using the code in some clusters, cuda might not be installed under /usr/local/cuda/. In this case, in addition to modifying CUDA_PATH in make.sh. You might also need to specify CPATH=/path/to/your/cuda/include.

For example CPATH=/path/to/your/cuda/include ./make.sh

[Rizhiy]

Does anyone know if CUDA installed by conda works with PyTorch? I installed it and torch seem to find devices correctly but haven't had the time to check training/inference yet.

[roytseng-tw]

To my knowledge, pytorch use the CUDA and CUDNN installed by conda.

[Rizhiy]

Then perhaps it would be better to also use CUDA and CUDNN installed by conda.

[Ai-is-light]

@Rizhiy I install the pytorch with pip install pytorch, but when I run cd lib and sh make.sh, I got some errors, like x86_64-linux-gnu-gcc: error: /home/yuyongbo/object_detection/Detectron.pytorch-master/lib/model/nms/src/nms_cuda_kernel.cu.o: No such file or directory

[stoneyang]

@Yuliang-Zou Thanks for your hint! You've saved my day!

@Ai-is-light This is perhaps because the cuda code cannot be compiled successfully. In my case, I found error like nvcc: command not found and the path to nvcc did not set properly on my machine. After fixing this, I can build libs without errors. Hope this will help. :)

[fc695355831]

@Rizhiy我用pip install pytorch安装了pytorch，但是当我运行cd lib和sh make.sh时，我遇到了一些错误，比如 x86_64-linux-gnu-gcc：error：/home/yuyongbo/object_detection/Detectron.pytorch- master / lib / model / nms / src / nms_cuda_kernel.cu.o：没有这样的文件或目录

你好，我也遇到同样的问题了，请问你解决了没有呀，谢谢