Using Docker for fast.ai: A Comprehensive Guide

September 16, 2020

Welcome to the world of deep learning with fast.ai! In this blog post, we’ll walk you through how to set up a Jupyter environment that leverages Docker for the fast.ai Course 1A. Whether you are using CPUs or NVIDIA GPUs, by the end of this article, you will be ready to dive into the magical realm of deep learning.

What You Need to Get Started

Before we begin, ensure you have Docker installed on your local machine. If you’re planning to use NVIDIA GPUs, don’t forget to have nvidia-docker set up as well.

Launching the Jupyter Environment

Now that you’re ready, let’s launch the Jupyter environment:

CPU Only

If you want to run the environment using only CPUs, execute the following command:

bash
docker run -it -p 8888:8888 deeprigfastai-course-1

With GPU

To utilize NVIDIA GPUs, the command changes slightly:

bash
nvidia-docker run -it -p 8888:8888 deeprigfastai-course-1

The Anatomy of the Docker Command

Think of a Docker command like a recipe in a cookbook. You have your base ingredients (the Docker image), which in this case is deeprigfastai-course-1. You also specify how you want to serve it up (the ports, like 8888). When you run this recipe – voilà! – you have a Jupyter notebook ready for experimentation!

Managing Data with Docker

Since Docker containers are transient, if you’re entering a Kaggle competition or need to retain your data, it’s essential to manage your data effectively:

To mount a local data directory when launching the container, use:

bash
docker run -it -p 8888:8888 -v /Users/yourname/data:/home/docker/data deeprigfastai-course-1

Your local data directory will now be accessible in the container at /home/docker/data. Remember to update the path in your notebooks accordingly!

Installing Additional Packages

If you find that some packages are missing from the container, follow these steps to install them:

Enter the running container:

bash
docker exec -it container_name /bin/bash

Update your package lists and install the package:

bash
sudo apt-get update && sudo apt-get install package_name

Running the Environment on AWS

If you wish to run the Jupyter environment on AWS, there are specific commands to follow:

For GPU Instance

bash
docker-machine create --driver amazonec2 --amazonec2-region=us-west-2 --amazonec2-root-size=50 --amazonec2-ami=ami-e03a8480 --amazonec2-instance-type=p2.xlarge fastai-p2

After spinning up the instance, authorize and access it:

bash
aws ec2 authorize-security-group-ingress --group-name docker-machine --port 8888 --protocol tcp --cidr 0.0.0.0/0
docker-machine ssh fastai-p2

For CPU Instance

bash
docker-machine create --driver amazonec2 --amazonec2-region=us-west-2 --amazonec2-root-size=50 --amazonec2-ami=ami-a073cdc0 --amazonec2-instance-type=t2.xlarge fastai-t2

Follow the same steps to access it:

bash
aws ec2 authorize-security-group-ingress --group-name docker-machine --port 8888 --protocol tcp --cidr 0.0.0.0/0
docker-machine ssh fastai-t2

Finally, open your browser to http://[NEW_MACHINE_IP]:8888 to access your notebooks.

Troubleshooting

If you encounter issues, here are a few tips:

Make sure Docker and nvidia-docker are correctly installed and running.
Verify that the port 8888 is not blocked by your firewall.
Confirm that you have provided the correct paths for your local data.
If problems persist, consider reaching out for assistance or consult the documentation.
For more insights, updates, or to collaborate on AI development projects, stay connected with fxis.ai.

At fxis.ai, we believe that such advancements are crucial for the future of AI, as they enable more comprehensive and effective solutions. Our team is continually exploring new methodologies to push the envelope in artificial intelligence, ensuring that our clients benefit from the latest technological innovations.