cuda导入python_Python中的numba调用cuda报错的解决方法
Python中的numba调用cuda报错如下：
........
numba.cuda.cudadrv.error.NvvmSupportError: libNVVM cannot be found. Do `conda install cudatoolkit`:
/usr/local/cuda10.0/nvvm/lib64/libnvvm.so: cannot open shared object file: No such file or directory
Exception ignored in: >
Traceback (most recent call last):
File "/usr/local/lib/python3.5/dist-packages/numba/cuda/cudadrv/nvvm.py", line 152, in __del__
driver = NVVM()
File "/usr/local/lib/python3.5/dist-packages/numba/cuda/cudadrv/nvvm.py", line 116, in __new__
raise NvvmSupportError(errmsg % e)
numba.cuda.cudadrv.error.NvvmSupportError: libNVVM cannot be found. Do `conda install cudatoolkit`:
/usr/local/cuda10.0/nvvm/lib64/libnvvm.so: cannot open shared object file: No such file or directory
原因就是找不到cuda
解决方法：
1、首先查看NVIDIA驱动和cuda是否安装好！
终端命令如下：
查看nivida驱动版本
sudo dpkg --list | grep nvidia-*
查看cuda版本
nvcc -V
安装成功则会显示如下结果：
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2018 NVIDIA Corporation
Built on Sat_Aug_25_21:08:01_CDT_2018
Cuda compilation tools, release 10.0, V10.0.130
或者
cat /usr/local/cuda/version.txt
CUDA Version 10.0.130
(查看cudnn版本： cat /usr/local/cuda/include/cudnn.h | grep CUDNN_MAJOR -A 2)
2、设置cuda的环境变量，在/etc/profile文件中最下面添加如下变量：
export CUDA_HOME=/usr/local/cuda-10.0
export PATH=/usr/local/cuda-10.0/bin:$PATH
export LD_LIBRARY_PATH=/usr/local/cuda-10.0/lib64:$LD_LIBRARY_PATH
export NUMBAPRO_NVVM=/usr/local/cuda-10.0/nvvm/lib64/libnvvm.so
export NUMBAPRO_LIBDEVICE=/usr/local/cuda-10.0/nvvm/libdevice/
添加完环境变量后，重启电脑，仍旧会报错，依旧找不到libnvvm.so文件
用ll命令：ll /usr/local/cuda10.0/nvvm/lib64/libnvvm.so 显示：
ls: cannot access '/usr/local/cuda10.0/nvvm/lib64/libnvvm.so': No such file or directory
显然该路径下并没有该文件，
注意报错的提示：
/usr/local/cuda10.0/nvvm/lib64/libnvvm.so: cannot open shared object file: No such file or directory
这里numba寻找的路径是/usr/local/cuda10.0....，而装cuda时nvvm/lib64/libnvvm.so文件是在/usr/local/cuda-10.0....，显然是找不到的，故仍旧会报错，
ll /usr/local/cuda-10.0/nvvm/lib64就能找到libnvvm.so文件了。由于这里的路径名字cuda10.0和cuda-10.0的不一致，导致错误，解决的办法是在/usr/local/建立一个软链接cuda10.0，让该链接文件指向cuda-10.0文件即可，即：
sudo ln -r cuda-10.0 cuda10.0
再运行程序不再报错！