Hello

In order to be compatible with my local ubuntu version, I changed the version of ubuntu 16.04 to 14.04 in dockerfile which is provided and build the image from it. I just did as below.

FROM nvidia/cuda:8.0-cudnn6-devel-ubuntu14.04

RUN echo "deb http://developer.download.nvidia.com/compute/machine-learning/repos/ubuntu1404/x86_64 /" > /etc/apt/sources.list.d/nvidia-ml.list

I made the container and checked that torch and torchvision are well installed.
(Explicitly sourced the conda environment as well. source activate pytorch-py35)

After then, since I need opencv for my own code I tried to install opencv3 by the command below.

conda install -c menpo opencv3

However above command requires to downgrade the numpy and scipy packages.

The following packages will be DOWNGRADED:

numpy:      1.13.1-py35_0            --> 1.11.3-py35_0
scipy:      0.19.1-np113py35_0       --> 0.19.0-np111py35_0**

If I just execute the above command for opencv, it results in error for importing torch then.
Giving,

RuntimeError: module compiled against API version 0xb but this version of numpy is 0xa
Traceback (most recent call last):
File "", line 1, in
File "/opt/conda/envs/pytorch-py35/lib/python3.5/site-packages/torch/init.py", line 53, in
from torch._C import *
ImportError: numpy.core.multiarray failed to import

So I just upgraded using pip
pip install numpy --upgrade
pip install scipy --upgrade

Then it gives below error.
ImportError: libgtk-x11-2.0.so.0: cannot open shared object file: No such file or directory

So I installed libgtk2.0 with the below command
sudo apt-get install libgtk2.0-0:i386

And finally it seems to work with opencv

1 . I am wondering is this right way to incorporate opencv into pytorch docker image.
Is there any nice way to figure out this problem?

2. Additionally, I am curious about whether the docker file provided gives current master version of pytorch.

Thanks in advance.