We are going to run a flask app in a docker container to display an image then we will load a model to do a prediction on that image in the flask app.

“docker build” does the building of the container
-t gives the container a name which is “python-workshop”
: gives the container ‘tag’ so you know where it was build from
. this says build from the current directory

docker build -t python_workshop:local .

docker images

docker run runs the container -p maps the ports in this case localhost port 8889 to the port 5000 we exposed in our docker file
--name is the name of our running container python_workshop:local references our image
--rm removes the image after we exit

docker run -p 8889:5000 --rm --name may21 python_workshop:local

Checking directory structure in container
“Docker run” still runs the container
-i starts in interactive mode
-t starts a terminal
bash starts a shell

docker run -p 8889:5000 -it --rm --name may21 python_workshop:local bash

You can train and save a model using a jupyter notebook (or other code) and then use it to do things in your flask app.

-v mounts a volume with code (or saved model files) on the docker container
The format is <directory on your local>:<directory in your container>

docker run -p 8889:5000 -v /home/becky/workshop_model:/app/model --rm --name may21 python_workshop:local