Mxnet imports various libcudarts * so、 libcuda*. So not found

This is really a huge pit ！！！！！！！

In the debug pyCFTracker Code , It is found that initialization keeps reporting errors , Details are as follows

OSError: libcudart.so.11.0: cannot open shared object file: No such file or directory

I wonder , No installation cuda, So it's normal to not find the library file , So he ran away NVIDIA Download on the official website cuda11.3, Use runfile Once installed , Keep running the code , Or an error libcudart.so.11.0 Unable to find . After searching countless blogs , And try one by one , The discovery still cannot solve the problem .

MXNet or tensorflow The lack of libcudart.so.11.0 and libcudnn.so.8 resolvent _ Write a blog -CSDN Blog _libcudnn

OSError: libcudart.so.11.0: cannot open Perfect solution _ A gaunt blog with a widening dress belt -CSDN Blog

Finally!!!!!! After installation - uninstall - install - uninstall - After the installation , I finally thought of ！ Since there are already mxnet Compatible version of cuda, Would that be IDE The environment variables used in it have not been updated , As a result, the library file cannot be found ？then, Open the pycharm Environment variables of . Sure enough ！ Although it has installed cuda, User level environment variables have also been updated PATH And LD_LIBRARY_PATH, however pycharm There are no synchronized updates ！ We need to update it manually, comrades . The setting method is as follows ：

1. Click in the upper left corner RUN->RUN...->Edit Configuration, Enter the setting interface as shown in the figure below , Open the environment variable editing window .

2. find PATH Variable , stay PATH The value of the variable is finally added ":/usr/local/cuda/bin"

3. Add new variables LD_LIBRARY_PATH, Set the value to "/usr/local/cuda/lib64"

4. All the way ok,apply,ok Click down . There will be no problem when debugging again ！！！！！