Theano/libgpuarray

Python 3.5.2: pygpu was configured but could not be imported or is too old (version 0.6 or higher required)

loretoparisi opened this issue · 5 comments

loretoparisi commented 
root@2ec43fec4281:~# python3 gpu_test.py 
ERROR (theano.gpuarray): pygpu was configured but could not be imported or is too old (version 0.6 or higher required)
NoneType
root@2ec43fec4281:~# python3 -c 'import pygpu; print(pygpu.__version__)'
0.7.2+23.gace45c6
root@2ec43fec4281:~# python3 -c 'import pygpu; print(pygpu.__path__)'
['/usr/local/lib/python3.5/dist-packages/pygpu-0.7.2+23.gace45c6-py3.5-linux-x86_64.egg/pygpu']
root@2ec43fec4281:~# python3 --version
Python 3.5.2

This could be related to #421

loretoparisi commented

[UPDATE]

A workaround is

pip install -U git+https://github.com/Theano/Theano.git#egg=Theano

as mentioned in #393

and the finally

oot@2ec43fec4281:~# python3 gpu_test.py 
Using cuDNN version 5110 on context None
Mapped name None to device cuda0: Tesla K80 (0000:00:1E.0)
[GpuElemwise{exp,no_inplace}(<GpuArrayType<None>(float64, vector)>), HostFromGpu(gpuarray)(GpuElemwise{exp,no_inplace}.0)]
Looping 1000 times took 0.373104 seconds
abergeron commented

This is not a workaround. Theano 0.9 works with pygpu 0.6.X. The next release of Theano (and the current master) works with 0.7.X.

loretoparisi commented

@abergeron ok, so why doing RUN pip3 install theano it does not work?

abergeron commented

Because the current stable release is 0.9 so that is what you get with pip install. You have pygpu 0.7.2+ installed so that is not compatible with the stable Theano.

If you want to use the stable release you can install pygpu 0.6.9, otherwise using the master as you did will also work. I just wanted to note that this is not a "workaround" it's actually how it is supposed to work.

  • 👍2
msiraj83 commented

@loretoparisi ... thanks ... its work for me....