model

Model Starter

Quickstart example to create your model artifact.

Basics

To create model artifact, clone or download this repository, you will get the following layout:

└── sample
    └── model.mlflow

Where:

• model.mlflow is your model file.

Using Your Model

Once your model is ready, you can simply refer to them in your punchline.

Check out the punchline example.

Run in foreground with docker

To test the example punchline in foreground mode simply run :

make run

or

docker run -it \
    -v $PWD/sample:/usr/share/punch/artifacts/io/github/starter/model/1.0.0 \
    -v $PWD/punchline.yaml:/data/punchline.yaml \
    --network=host \
    ghcr.io/punchplatform/punchline-pyspark:8.1-dev \
    /data/punchline.yaml

Start your punchline in production mode on Kubernetes

A zip archive containing your model and a metadata file can be build using make : model-1.0.0-artifact.zip

Using Makefile

You can apply your punchline, start a service and run the simulator by running :

make apply

When the simulator starts, you can exit this command using Ctrl+c and run :

make logs

This should show your punchline logs.

To delete everything, simply run :

make delete

Using shell commands

You simply have to upload it to the Punch Artefact Server using this command (do not forget to update the artifact service name):

curl -X POST "http://artifacts-server.kooker:4245/v1/artifacts/upload" -F artifact=@target/model-1.0.0-artifact.zip -F override=true

Start your punchline on kubernetes, de not forget to check the artifact service name in punchline.yaml file before executing this command :

kubectl apply -f punchline.yaml

In one window, view your punchline log :

kubectl logs -f --tail -1 -l punchline-name=punch-model