我使用anaconda python创建了虚拟环境。我在创建的环境中安装了cuda工具包。现在，我必须在makefile中提供cuda安装的路径。默认路径/ usr / local / cuda / include /不存在。如何找到cuda的正确路径？
我必须在下面给出的make文件中进行更改
COMMON+= -DGPU -I/usr/local/cuda/include/
CFLAGS+= -DGPU
LDFLAGS+= -L/usr/local/cuda/lib64 -lcuda -lcudart -lcublas -lcurand
命令which nvcc给出/usr/bin/nvcc
locate cuda | grep /cuda$给出
/home/tan/.conda/envs/tensorflow_env/include/opencv2/core/cuda
/home/tan/.conda/envs/tensorflow_env/lib/python3.6/site-packages/tensorflow/include/tensorflow/stream_executor/cuda
/home/tan/.conda/envs/tensorflow_gpu/include/opencv2/core/cuda
/home/tan/.conda/envs/tensorflow_gpu/lib/python3.6/site-packages/tensorflow/include/external/local_config_cuda/cuda
/home/tan/.conda/envs/tensorflow_gpu/lib/python3.6/site-packages/tensorflow/include/external/local_config_cuda/cuda/cuda
/home/tan/.conda/envs/tensorflow_gpu/lib/python3.6/site-packages/tensorflow/include/tensorflow/stream_executor/cuda
/home/tan/.conda/pkgs/libopencv-3.4.2-hb342d67_1/include/opencv2/core/cuda
/home/tan/.conda/pkgs/numba-0.42.0-py36h962f231_0/lib/python3.6/site-packages/numba/cuda
/home/tan/.conda/pkgs/opencv3-3.1.0-py36_0/include/opencv2/core/cuda
/home/tan/.conda/pkgs/tensorflow-base-1.10.0-mkl_py36h3c3e929_0/lib/python3.6/site-packages/tensorflow/include/external/local_config_cuda/cuda
/home/tan/.conda/pkgs/tensorflow-base-1.10.0-mkl_py36h3c3e929_0/lib/python3.6/site-packages/tensorflow/include/external/local_config_cuda/cuda/cuda
/home/tan/.conda/pkgs/tensorflow-base-1.10.0-mkl_py36h3c3e929_0/lib/python3.6/site-packages/tensorflow/include/tensorflow/stream_executor/cuda
/home/tan/.conda/pkgs/tensorflow-base-1.12.0-gpu_py36had579c0_0/lib/python3.6/site-packages/tensorflow/include/external/local_config_cuda/cuda
/home/tan/.conda/pkgs/tensorflow-base-1.12.0-gpu_py36had579c0_0/lib/python3.6/site-packages/tensorflow/include/external/local_config_cuda/cuda/cuda
/home/tan/.conda/pkgs/tensorflow-base-1.12.0-gpu_py36had579c0_0/lib/python3.6/site-packages/tensorflow/include/tensorflow/stream_executor/cuda
/home/tan/.conda/pkgs/tensorflow-base-1.12.0-mkl_py36h3c3e929_0/lib/python3.6/site-packages/tensorflow/include/external/local_config_cuda/cuda
/home/tan/.conda/pkgs/tensorflow-base-1.12.0-mkl_py36h3c3e929_0/lib/python3.6/site-packages/tensorflow/include/external/local_config_cuda/cuda/cuda
/home/tan/.conda/pkgs/tensorflow-base-1.12.0-mkl_py36h3c3e929_0/lib/python3.6/site-packages/tensorflow/include/tensorflow/stream_executor/cuda
/home/tan/.conda/pkgs/tensorflow-base-1.3.0-py27h0dbb4d0_1/lib/python2.7/site-packages/tensorflow/include/tensorflow/stream_executor/cuda
/home/tan/.conda/pkgs/tensorflow-base-1.3.0-py36h5293eaa_1/lib/python3.6/site-packages/tensorflow/include/tensorflow/stream_executor/cuda
/home/tan/anaconda3/lib/python3.6/site-packages/numba/cuda
/home/tan/anaconda3/pkgs/numba-0.38.0-py36h637b7d7_0/lib/python3.6/site-packages/numba/cuda
/home/tan/opencv3/opencv-3.4.1/build/modules/core/CMakeFiles/opencv_perf_core.dir/perf/cuda
/home/tan/opencv3/opencv-3.4.1/build_dnn/modules/core/CMakeFiles/opencv_perf_core.dir/perf/cuda
/home/tan/opencv3/opencv-3.4.1/build_gpu/modules/core/CMakeFiles/opencv_perf_core.dir/perf/cuda
/home/tan/opencv3/opencv-3.4.1/modules/core/include/opencv2/core/cuda
/home/tan/opencv3/opencv-3.4.1/modules/core/perf/cuda
/home/tan/opencv3/opencv-3.4.1/modules/core/src/cuda
/home/tan/opencv3/opencv-3.4.1/modules/cudaarithm/src/cuda
/home/tan/opencv3/opencv-3.4.1/modules/cudabgsegm/src/cuda
/home/tan/opencv3/opencv-3.4.1/modules/cudacodec/src/cuda
/home/tan/opencv3/opencv-3.4.1/modules/cudafeatures2d/src/cuda
/home/tan/opencv3/opencv-3.4.1/modules/cudafilters/src/cuda
/home/tan/opencv3/opencv-3.4.1/modules/cudaimgproc/src/cuda
/home/tan/opencv3/opencv-3.4.1/modules/cudalegacy/src/cuda
/home/tan/opencv3/opencv-3.4.1/modules/cudaobjdetect/src/cuda
/home/tan/opencv3/opencv-3.4.1/modules/cudaoptflow/src/cuda
/home/tan/opencv3/opencv-3.4.1/modules/cudastereo/src/cuda
/home/tan/opencv3/opencv-3.4.1/modules/cudawarping/src/cuda
/home/tan/opencv3/opencv-3.4.1/modules/photo/src/cuda
/home/tan/opencv3/opencv-3.4.1/modules/stitching/src/cuda
/home/tan/opencv3/opencv-3.4.1/modules/superres/src/cuda
/home/tan/opencv3/opencv-3.4.1/modules/videostab/src/cuda
/home/tan/opencv3/opencv_contrib-3.4.1/modules/hfs/src/cuda
/home/tan/opencv3/opencv_contrib-3.4.1/modules/xfeatures2d/src/cuda
/usr/include/flann/util/cuda
答案 0 :(得分：1)
通过运行文件为ubuntu进行的CUDA完整安装为2.4 GB，而anaconda仅约370 MB。后者包含运行依赖它的库所需的所有依赖关系，例如PyTorch或Tensorflow。它不是完整的安装，很可能没有您想要的东西。
您需要一个完整的开发包，可以在Nvidia website上找到。
答案 1 :(得分：0)
您可以在计算机上安装CUDA的多个完整版本。据我了解，CUDA是向后兼容的，因此您只需要一个即可。
但是，使用较新版本的CUDA进行编译可能会链接到其他软件包的不兼容的较新库。
如果确实需要安装CUDA的多个版本，则需要添加指向.bashrc的链接，而不必添加链接。这是一个有说明的网站
https://medium.com/@peterjussi/multicuda-multiple-versions-of-cuda-on-one-machine-4b6ccda6faae
tl; dr运行此命令。
sudo ldconfig /usr/local/cuda-8.0/lib64