How to use the docker container for the course

We provide a docker image oduerr/tf_docker:cpu_r with Tensorflow (v1.13.0) , TFLearn, Keras, and many other pre-installed python libraries (numpy, pandas). It also allows to run R-code inside the notebooks. See e.g. here how to use it.

Installation of docker

Running the container

In the docker command line execute:

docker run -p 8888:8888 -p 6006:6006 -it oduerr/tf_docker:cpu_r

open http://localhost:8888/?token=tensorchiefs or http://192.168.99.100:8888/tree?token=tensorchiefs(forcsome versions of windows) in the browser.

Running with a linked file system.

If you want to access a directory here (/Users/oli/Documents/workspace/dl_tutorial/) from inside the docker container execute:

docker run -p 8888:8888 -p 6006:6006 -v /Users/oli/Documents/workspace/dl_tutorial/:/notebooks/local/ -it oduerr/tf_docker:cpu_r
#or for windows users
docker run -p 8888:8888 -p 6006:6006 -v /c/Users/Elvis/Desktop/:/notebooks/local/ -it oduerr/tf_docker:cpu_r

If you work with git, we recommend to clone the DL course. Then you can link with the above comand directly to your local copy of the course which allows you to access the prepared notebooks in the course.

Updating

Please make sure to use the latest container by updating it using

docker pull oduerr/tf_docker:cpu_r

Other useful hints for docker

Starting in bash

In case you want to not start the jupyter notebook sever automatically but want a bash shell do:

docker run -p 8888:8888 -p 6006:6006 -it oduerr/tf_docker:cpu_r bash

Local vs Inside container

The entry before the colon ‘:’ is on the local machine, the one after it inside the container. Examples:

  docker run -p 4242:8888 -it oduerr/tf_docker:cpu_r #4242 is the port on the local machine, 8888 inside the container
  docker run -v /tmp/dl_tutorial/:/notebooks/dl_tutorial/ #/tmp/dl_tutorial is on local machine

Stopping and killing a docker process

To stop a docker image hit 1-time ctrl-c and answer the question to shutdown the notebook server with “Yes”. Then you can run again new docker commands.

We sometimes get the error message, that the port is already allocated. If you have not allocated the port by other applications but by a accidantelly running docker image, then you can kill the old docker job. To see the process ID of the jobs we type:

docker ps

The you get a list of all running docker jobs along with their container ID, which you can use to kill the job. For example (if the ID is 1e3467456) the comand is:

docker kill 1e3467456