stevenrouk/docker-datasci-app-example

Application of a full Python data science application containerized in Docker.

Creating a Python REST API using sklearn and Docker

In this application, we use the Python machine learning library scikit-learn to make predictions of Iris species given measurements.

The app is presented as a REST API (using Flask-RESTful) that returns JSON data, and the whole thing is packaged up in a Docker container.

Created using the tutorial by Data Dan.

---

Running It

First, clone the GitHub repo and cd into the directory.

git clone https://github.com/stevenrouk/docker-datasci-app-example.git
cd docker-datasci-app-example

If you need to install Docker, follow the instructions here: Docker Installation Instructions

sudo docker build --force-rm=true -t pythoniris .
sudo docker run --net host -d --name myiris pythoniris

Head to http://localhost:5000/prediction?slength=0.1&swidth=0.1&plength=0.1&pwidth=0.1 and get a prediction!

Instead of always using sudo, can also create a docker group and run off of that. Run these, then log out and log back in, or run 'newgrp docker'.

sudo groupadd docker
sudo usermod -aG docker $USER
newgrp docker

To remove the docker image:

sudo docker stop myiris
sudo docker rm myiris

---

Pulling a Basic Docker Python Image, Then Creating Your Own

Here's a simple example of getting a Docker image with python already installed, then creating your own image from it that has your code on it, and running that.

https://civisanalytics.com/blog/engineering/2014/08/14/Using-Docker-to-Run-Python/

Run the 'hello-world' Docker image, then take a look at all the images and processes.

docker run hello-world
docker images
docker ps -a

Get a Docker image with python installed and check the version of Python. The first command gets the image from Docker Hub, the second command runs "python --version" on the docker image called 'python', and the third command just prints the contents of the root directory of the Docker image.

docker pull python
docker run python python --version
docker run python ls

Now, let's create our own Docker image using the 'python' image as a starting point, and a small test script to run. First, let's create a new directory.

mkdir docker-python-test
cd docker-python-test

Now, a small test Python script that we'll want to run on our new Docker image.

~/docker-python-test/main.py

#!/usr/bin/env python3

if __name__ == '__main__':
    print("Hello Docker!")

Now, we create the Dockerfile which lists how to build the Docker image. For more information on how these files work, check out the Dockerfile Reference and Best Practices for Writing Dockerfiles.

~/docker-python-test/Dockerfile

# DOCKER-VERSION 1.12.3

# use the 'python' image as the source
FROM python

# copy files from the current directory into '/src' on the Docker image
COPY . /src

# default command when running the Docker image
CMD ["python", "/src/main.py"]

Now, we build a Docker image called 'stevenrouk/docker-python-test' using that Dockerfile.

docker build -t stevenrouk/docker-python-test .

Let's check the version of Python again on this image, then run our script! We'll run it two different ways: the first way is using the default command in the Dockerfile that we set up, and the second way we'll explicitly tell the Docker image to run our script.

docker run stevenrouk/docker-python-test python --version
docker run stevenrouk/docker-python-test
docker run stevenrouk/docker-python-test python /src/main.py

---

Creating an Anaconda Docker Image

docker search continuumio
docker pull continuumio/miniconda3
docker run continuumio/miniconda3 python --version
docerr run -it continuumio/miniconda3

Dockerfile

# DOCKER-VERSION 1.12.3

# Use the miniconda3 image
FROM continuumio/miniconda3

# Copy files from the current directory into '/src' on the Docker image
COPY . /src

# Install dependencies
RUN apt-get -y update --fix-missing
RUN conda install --name root --file /src/spec-file.txt

# Default command when running the Docker image
CMD ["python", "/src/main.py"]

docker build -t stevenrouk/sklearn-test .