Install mxnet-cu102

English Version mxnet
1

I used this command “pip install mxnet-cu102” from the Apache mxnet website. I selected GPU. The installation was successful; however when I verified my installation by “import mxnet as mx” I got and error message
“OSError: libcudart.so.10.2: cannot open shared object file: No such file or directory”
Would someone please help?

Respectfully

2

Hi @mydiveindeeplearning, D2L is current based on cuda10.1. Please see more details here.

3

have you installed cuda-10.2 with PATH and LD_LIBRARY_PATH settings?

4

CUDA-10.2 is backward compatible with atleast 10.1 version because i have tested it with a tensorflow version which was compiled with 10.1.

5

In the mxnet notes: mxnet-cu101 means the package is built with CUDA/cuDNN and the CUDA version is 10.1. Hence for using mxnet-cu102 you first need to install latest cuda toolkit version 10.2.8 (build hfd86e86_1) using

conda install cudatoolkit=10.2

assuming you are using conda. Then reinstall mxnet with cuda using the pip command you used.

Hope this helps.