Error loading library gpuarray with Theano

问题

I am trying to run this script to test Theano's use of my GPU and get the following error:

ERROR (theano.gpuarray): Could not initialize pygpu, support disabled
Traceback (most recent call last):
  File "/home/me/anaconda3/envs/py35/lib/python3.5/site-
packages/theano/gpuarray/__init__.py", line 164, in <module>
    use(config.device)
  File "/home/me/anaconda3/envs/py35/lib/python3.5/site-
packages/theano/gpuarray/__init__.py", line 151, in use
    init_dev(device)
  File "/home/me/anaconda3/envs/py35/lib/python3.5/site-
packages/theano/gpuarray/__init__.py", line 60, in init_dev
    sched=config.gpuarray.sched)
  File "pygpu/gpuarray.pyx", line 614, in pygpu.gpuarray.init 
(pygpu/gpuarray.c:9419)
  File "pygpu/gpuarray.pyx", line 566, in pygpu.gpuarray.pygpu_init 
(pygpu/gpuarray.c:9110)
  File "pygpu/gpuarray.pyx", line 1021, in 
    pygpu.gpuarray.GpuContext.__cinit__ (pygpu/gpuarray.c:13472)
pygpu.gpuarray.GpuArrayException: Error loading library: -1

I need to use the nvidia-381 driver since my GPU is a 1080 ti and is not compatible with nvidia-375. I'm not sure if that matters but installing nvcc overwrites 381 and causes some errors if I reinstall 381 after setting up nvcc so I can't use nvcc.

I can import pygpu without errors but if I run pygpu.test() I get the following error and I don't know how to specify the DEVICE variable without nvcc.

======================================================================
ERROR: Failure: RuntimeError (No test device specified.  Specify one using the DEVICE or GPUARRAY_TEST_DEVICE environment variables.)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/me/anaconda3/envs/py35/lib/python3.5/site-packages/nose/failure.py", line 39, in runTest
    raise self.exc_val.with_traceback(self.tb)
  File "/home/me/anaconda3/envs/py35/lib/python3.5/site-packages/nose/loader.py", line 418, in loadTestsFromName
    addr.filename, addr.module)
  File "/home/me/anaconda3/envs/py35/lib/python3.5/site-packages/nose/importer.py", line 47, in importFromPath
    return self.importFromDir(dir_path, fqname)
  File "/home/me/anaconda3/envs/py35/lib/python3.5/site-packages/nose/importer.py", line 94, in importFromDir
    mod = load_module(part_fqname, fh, filename, desc)
  File "/home/me/anaconda3/envs/py35/lib/python3.5/imp.py", line 234, in load_module
    return load_source(name, filename, file)
  File "/home/me/anaconda3/envs/py35/lib/python3.5/imp.py", line 172, in load_source
    module = _load(spec)
  File "<frozen importlib._bootstrap>", line 693, in _load
  File "<frozen importlib._bootstrap>", line 673, in _load_unlocked
  File "<frozen importlib._bootstrap_external>", line 665, in exec_module
  File "<frozen importlib._bootstrap>", line 222, in _call_with_frames_removed
  File "/home/me/.local/lib/python3.5/site-packages/pygpu-0.6.2-py3.5-linux-x86_64.egg/pygpu/tests/test_tools.py", line 5, in <module>
    from .support import (guard_devsup, rand, check_flags, check_meta, check_all,
  File "/home/me/.local/lib/python3.5/site-packages/pygpu-0.6.2-py3.5-linux-x86_64.egg/pygpu/tests/support.py", line 32, in <module>
    context = gpuarray.init(get_env_dev())
  File "/home/me/.local/lib/python3.5/site-packages/pygpu-0.6.2-py3.5-linux-x86_64.egg/pygpu/tests/support.py", line 29, in get_env_dev
    raise RuntimeError("No test device specified.  Specify one using the DEVICE or GPUARRAY_TEST_DEVICE environment variables.")
RuntimeError: No test device specified.  Specify one using the DEVICE or GPUARRAY_TEST_DEVICE environment variables.

----------------------------------------------------------------------
Ran 7 tests in 0.003s

FAILED (errors=7)
<nose.result.TextTestResult run=7 errors=7 failures=0>

回答1:

Warning: its entirely possible that this is all wrong and the actual reason for your problem is in fact - as you suspect - your gpu driver.

I had the same issue with gpuarray on Windows 10. In the end I solved it by:

completely uninstall python
install cuda 8.0 (with cudnn 5.1)
install anaconda
install theano through anaconda:
```
conda install theano pygpu 
```

As you are using linux: This error message basically means It didn't work, don't ask me why And is mostly shown if something with your setup is wrong (e.g. different compilers used for compiling python and theano, or incompatible cuda version)

I would recommend to update to cuda 8.0 and to reinstall your python environment over anaconda (just in case)

On a side note: I tested your example script from the docu and at least that is working....

Note for windows users: Never try to install Anaconda in a location where you have spaces in the path... Everything looks fine ... until theano starts having trouble finding and compiling things.

Note regarding the pygpu.test():

Normally you just set the environment variable:

windows: set DEVICE=cuda
linux: export DEVICE=cuda

BUT The test has the habit of saying you didn't specify a device if the library couldn't be loaded...

来源：https://stackoverflow.com/questions/43577788/error-loading-library-gpuarray-with-theano

标签

gpu

theano