- Install and open the
docker app
on your machine. - Install
git lfs
and clone this repo. - Install the
docker python sdk
:
pip3 install docker
cd
into the repo.
First, build the docker image
and create a docker container
from that image:
python build.py
Once we have a container, we can start it up with:
python start.py
We can then send data to the model, running inside the container, through POST
requests:
sh post.sh
When we are done with the container, we can stop it with:
python stop.py
In a production setting, the flow would be as follows:
- An app is approved for the marketplace
- A
docker image
is automatically created for the app and uploaded to our image registry (public or private) - A provider machine would then enroll in the app
* i.e. the
dvm-cli
will create adocker container
locally using thedocker image
found in our registry - The provider would then start providing
* i.e. the
docker container
will start allowing it to process prediction requests