bethgelab / docker-deeplearning

Development of new unified docker container (WIP)

docker-deeplearning

Development of new unified docker container (WIP)

docker-deeplearning's Issues

fix agmb-docker

We want to replace our current bash-script agmb-docker with a python-based script that:

• sets the username & userID
• chooses random ports for the jupyter notebook & SSH connection
• adds the home-directory (or the whole bethge-folder if script detects it is running within CIN)
• displays some helper information

juptyer password

set a fixed password?

Tensorboard on cudnn6 container fails

When running tensorboard in our cudnn6 containers

agmb-docker run --name -tfboard-0 -p 7100:6006 -d bethgelab/deeplearning:cuda8.0-cudnn6 tensorboard --logdir="$PWD"/train_logs/

I am running into the following problem:

Traceback (most recent call last):
  File "/usr/local/bin/tensorboard", line 7, in <module>
    from tensorboard.main import main
  File "/usr/local/lib/python3.5/dist-packages/tensorboard/main.py", line 30, in <module>
    from tensorboard import default
  File "/usr/local/lib/python3.5/dist-packages/tensorboard/default.py", line 35, in <module>
    from tensorboard.plugins.audio import audio_plugin
  File "/usr/local/lib/python3.5/dist-packages/tensorboard/plugins/audio/audio_plugin.py", line 27, in <module>
    from tensorboard import plugin_util
  File "/usr/local/lib/python3.5/dist-packages/tensorboard/plugin_util.py", line 21, in <module>
    import bleach
  File "/usr/local/lib/python3.5/dist-packages/bleach/__init__.py", line 7, in <module>
    from bleach.linkifier import (
  File "/usr/local/lib/python3.5/dist-packages/bleach/linkifier.py", line 6, in <module>
    from html5lib.filters.base import Filter
ImportError: No module named 'html5lib.filters.base'

This seems to be related to the version of html5lib, see tensorflow/tensorflow#8971. This issue does not arise in our cudnn5 containers. It's probably enough to manually install the correct library version in the Dockerfile.

create and initialize user in entry point

document usage in README

enable ssh into the container

jupyter extensions

pip, pip2 and pip3

pip sometimes refers to pip3, even when python refers to python2. It doesn't make much sense, but we should explicitly use pip2 in the Dockerfile.

TensorFlow release candidates

I am wondering whether we should explicitly specify the TensorFlow version number and to install release canidates as the default.

Doesn't build anymore

Step 14 : RUN pip2 install pip --upgrade  && pip3 install pip --upgrade  && pip2 install setuptools  && pip3 install setuptools
 ---> Running in a8882b8e775b
Traceback (most recent call last):
  File "/usr/bin/pip2", line 5, in <module>
    from pkg_resources import load_entry_point
ImportError: No module named pkg_resources