1

I use theano on a remote server that I first ssh into (I don't have root on that system). This works fine, however, if I start a screen, I get an error when trying to import theano.

Behavior when not using screen:

>>> import theano
Using gpu device 0: GeForce GTX TITAN X (CNMeM is disabled, CuDNN 4007)

Behavior when using screen:

>>> import theano
ERROR (theano.sandbox.cuda): Failed to compile cuda_ndarray.cu: libcublas.so.7.5: cannot open shared object file: No such file or directory
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/home/2012/enewel3/.local/lib/python2.7/site-packages/theano/__init__.py", line 103, in <module>
    import theano.sandbox.cuda
  File "/home/2012/enewel3/.local/lib/python2.7/site-packages/theano/sandbox/cuda/__init__.py", line 697, in <module>
    use(device=config.device, force=config.force_device, test_driver=False)
  File "/home/2012/enewel3/.local/lib/python2.7/site-packages/theano/sandbox/cuda/__init__.py", line 496, in use
    device, cuda_initialization_error_message))
EnvironmentError: You forced the use of gpu device gpu, but CUDA initialization failed with error:
cuda unavailable

How should I use theano within a screen session?

Edward Newell
  • 17,203
  • 7
  • 34
  • 36

1 Answers1

2

The problem was due to the fact that, although screen inherits most environment variables, it adds, removes, and alters some of them. In my case, it was changing the value of LD_LIBRARY_PATH.

Upon entering a screen, manually setting the environment variable to it's proper value fixed the problem. For me, that looks like this:

export LD_LIBRARY_PATH=/usr/local/cuda/lib64:/usr/local/pkgs/gurobi502/linux64/lib

Adding that line to my .bashrc makes sure that screens always get the right value. Note, adding it to .bash_profile won't work, because .bash_profile is only run at login, while .bashrc is run for every new shell.

Edward Newell
  • 17,203
  • 7
  • 34
  • 36