coursera-aml-docker

Docker container with Jupyter Environment for Coursera "Advanced Machine Learning" specialization: https://www.coursera.org/specializations/aml

Install Stable Docker Community Edition

For Mac: https://docs.docker.com/docker-for-mac/install/

For Windows (64bit Windows 10 Pro, Enterprise and Education): https://docs.docker.com/docker-for-windows/install/#what-to-know-before-you-install

For Windows (older versions): https://docs.docker.com/toolbox/toolbox_install_windows/

For Linux: https://docs.docker.com/engine/installation/

Running container for the first time

First run docker pull zimovnov/coursera-aml-docker to pull the latest version of image. Run using docker run -it -p 127.0.0.1:8080:8080 --name coursera-aml-1 zimovnov/coursera-aml-docker. This command downloads the prepared image from a public hub and starts a Jupyter for you. Let this command continue running in the terminal while you work with Jupyter.

You can now navigate to http://localhost:8080 in your browser to see Jupyter.

If you want to use tensorboard in the course, you also need to add an option to publish the port it uses. Run docker run -it -p 127.0.0.1:8080:8080 -p 127.0.0.1:7007:7007 --name coursera-aml-1 zimovnov/coursera-aml-docker instead.

Stopping and starting the container

This "stop and start" scenario is useful when you want to take a break and turn off your host machine.

Stopping the container

Save your work inside the container, then run docker stop coursera-aml-1 in different terminal window to stop a running container. You will be able to start it later.

Starting container after stopping

Run docker start -a coursera-aml-1 to run previously stopped container and attach to its stdout. You can continue to work where you left off.

Container checkpoints

You might want to make a checkpoint of your work so that you can return to it later. Think of it as a backup or commit in version control system.

Saving container state

You will first have to stop the container following instructions above. Now you need to save the container state so that you can return to it later: docker commit coursera-aml-1 coursera-aml-snap-1. You can make sure that it's saved by running docker images.

Creating new container from previous checkpoint

If you want to continue working from a particular checkpoint, you should run a new container from your saved image by executing docker run -it -p 127.0.0.1:8080:8080 --name coursera-aml-2 coursera-aml-snap-1. Notice that we incremented index in the container name, because we created a new container.

Using GPU in your container (Linux hosts only)

You can use NVIDIA GPU in your container on Linux host machine.

• Setup docker following instructions from NVIDIA: https://github.com/NVIDIA/nvidia-docker/wiki/Installation-(version-2.0)#prerequisites
• In your container replace CPU TensorFlow version with the one that supports GPU:

pip3 uninstall tensorflow
pip3 install tensorflow-gpu==1.2.1

• You will also have to install NVIDIA GPU driver, CUDA toolkit and CuDNN (requires registration with NVIDIA) in your container in order for TensorFlow to work with your GPU: https://www.tensorflow.org/versions/r1.2/install/install_linux#nvidia_requirements_to_run_tensorflow_with_gpu_support It can be hard to follow, so you might choose to stick to a CPU version, which is also fine for the purpose of this course.
• TensorFlow provides Docker files with TensorFlow on GPU, but they don't have all the additional dependencies we need, this is for advanced users: https://github.com/tensorflow/tensorflow/tree/master/tensorflow/tools/docker