Kinetica Machine Learning BlackBox Container SDK

Copyright (c) 2019 Kinetica DB Inc.

For support: support@kinetica.com

Usage:

STEP 1

• Clone and rename the repository container-kml-blackbox-dep-base

• Rename references to the container name in both ./build.sh and ./publish.sh to reflect the new name of your container

STEP 2

Derive a class from KineticaBlackBox. This base class is implemented in kinetica_black_box.py)

• Create a module file for the black box (eg. bb_module.py)

• In this file, import the base class:

from kinetica_black_box import KineticaBlackBox

• Then, in this file, derive a new class from this:

class MyBlackBox(KineticaBlackBox): def blackbox_function(self, inMap): # Call other functions in this file, in other files or insert blackbox code here # inMap and outMap are dicts() of the input and output parameters

    return outMap

STEP 3. Call the new black box module from bb_runner.py

• In the main() routine of bb_runner.py, import the new black box class:

from bb_module import MyBlackBox

• Then, also in the main routine, instantiate this class:

bb = MyBlackBox(schema_inbound, schema_outbound, zmq_host, zmq_port, zmq_topic, db_table, db_host, db_port, db_user, db_pass)

• Then call the run method of this object:

bb.run()

STEP 4. Modify Dockerfile.

• ADD the new black box model file created above to the container image

STEP 5. Build and push container

• run ./build.sh then ./publish.sh to build the container according to the Docker file instructions and push to DockerHub or a private registry

• See blackbox pytest and API documentation for details on interacting with the deployed black box