May.14.2018 Successfully install TensorFlow, Keras in lily server.

Main idea

The main idea is to follow the instructions on official websites! For instance, when installing cuda related applications, it’s better to follow the instructions on cuda websites instead of those on TensorFlow websites.

Install cuda

One very important experience: when install cuda, follow the instructions on its official website instead of those on tensorflow website, etc.

pip problem

from pip._internal import main ImportError: No module named ‘pip._internal’

if there is some error when using pip or pip3, the reason is there might be many versions of pips in different applications/exes. The solution is to uninstall pip and reinstall pip.

Install cudnn

there is one mistake after installation. The error is

/usr/local/cuda/include/cuda_runtime_api.h:1683:101: error: use of enum ‘cudaDeviceP2PAttr’ without previous declaration

The solution is: to change the file /usr/include/cudnn.h, I changed the line:

#include "driver_types.h" 

to

#include <driver_types.h>

Test the Keras with examples in official

curl -sSL https://github.com/fchollet/keras/raw/master/examples/mnist_mlp.py | python

without GPU in lily server: ~2min

with GPU in lily server: ~1min