Feast plugin for IBM Cloud.
This plugin implements Feast's offline store backed using IBM Cloud Data Engine and IBM Cloud Object Storage

Installation

Project dependencies can be installed in a dedicated virtual environment by running the following command:

poetry install

Testing and Linting

poetry run pytest tests/
poetry run pylint ibm_data_engine

Test with Feast

You use it with Feast by defining your offline store and data sources. The instructions below illustrate how it can be used in feast-ibm-quickstart.

Define dependency

This library is currently not published in PyPI; you will have to point to the repository directly. The easiest way to do it is to clone the repository, and define the dependency as a path in feast-ibm-quickstart.

ibm-data-engine = { path = "/path/to/ibm-data-engine" }

After running poetry update, you should be able to use the IBM Cloud Data Engine offline store.

Define data source

You can modify the src/feature_repo/example_repo.py file to use the new data source. Below is the minimal example of the file:

from ibm_data_engine import DataEngineDataSource
driver_stats_source = DataEngineDataSource(
    name="driver_hourly_stats_source",
    table="driver_stats_demo",
    timestamp_field="event_timestamp",
)

Define offline store

Then, feature_repo/feature_store.yaml must configure the offline store.

project: test_plugin
entity_key_serialization_version: 2
registry: data/registry.db
provider: local
online_store:
    type: redis
    connection_string: ${REDIS_HOST}:${REDIS_PORT},username=${REDIS_USERNAME},password=${REDIS_PASSWORD},ssl=true,ssl_ca_certs=${REDIS_CERT_PATH},db=0

offline_store:
    type: ibm_data_engine.DataEngineOfflineStore
    api_key: ${DATA_ENGINE_API_KEY}
    instance_crn: ${DATA_ENGINE_INSTANCE_CRN}
    target_cos_url: ${IBM_CLOUD_OBJECT_STORE_URL}

Notice that you must define the environment variables:

• IBM_CLOUD_OBJECT_STORE_URL
• REDIS_HOST
• REDIS_PORT
• REDIS_PASSWORD
• REDIS_CERT_PATH
• DATA_ENGINE_API_KEY
• DATA_ENGINE_INSTANCE_CRN

Apply

To apply the definitions to the registry, run:

poetry run feast -c ./feature_repo apply

Training

Run training by retrieving historical feature information from feature store

poetry run python training.py

Materialize

To materialize to Redis, run:

poetry run feast -c ./ materialize '<START_TIMESTAMP>'  '<END_TIMESTAMP>'

Inference

poetry run python inference.py