Feathr – An Enterprise-Grade, High Performance Feature Store

What is Feathr?

Feathr lets you:

Define features based on raw data sources, including time-series data, using simple APIs.
Get those features by their names during model training and model inferencing.
Share features across your team and company.

Feathr automatically computes your feature values and joins them to your training data, using point-in-time-correct semantics to avoid data leakage, and supports materializing and deploying your features for use online in production.

For more details, read our documentation.

Running Feathr on Azure with 3 Simple Steps

Feathr has native cloud integration. To use Feathr on Azure, you only need three steps:

Get the Principal ID of your account by running az ad signed-in-user show --query objectId -o tsv in the link below (Select "Bash" if asked), and write down that value (something like b65ef2e0-42b8-44a7-9b55-abbccddeefff). Think this ID as something representing you when accessing Azure, and it will be used to grant permissions in the next step in the UI.

Launch Cloud Shell

Click the button below to deploy a minimal set of Feathr resources for demo purpose. You will need to fill in the Principal ID and Resource Prefix. You will need "Owner" permission of the selected subscription.

Run the Feathr Jupyter Notebook by clicking the button below. You only need to change the specified Resource Prefix.

Installing Feathr Client Locally

If you are not using the above Jupyter Notebook and want to install Feathr client locally, use this:

pip install -U feathr

Or use the latest code from GitHub:

pip install git+https://github.com/linkedin/feathr.git#subdirectory=feathr_project

Feathr Highlights

Defining Features with Transformation

features = [
    Feature(name="f_trip_distance",                         # Ingest feature data as-is
            feature_type=FLOAT),
    Feature(name="f_is_long_trip_distance",
            feature_type=BOOLEAN,
            transform="cast_float(trip_distance)>30"),      # SQL-like syntax to transform raw data into feature
    Feature(name="f_day_of_week",
            feature_type=INT32,
            transform="dayofweek(lpep_dropoff_datetime)")   # Provides built-in transformation
]

anchor = FeatureAnchor(name="request_features",             # Features anchored on same source
                       source=batch_source,
                       features=features)

Rich UDF Support

Feathr has highly customizable UDFs with native PySpark and Spark SQL integration to lower learning curve for data scientists:

def add_new_dropoff_and_fare_amount_column(df: DataFrame):
    df = df.withColumn("f_day_of_week", dayofweek("lpep_dropoff_datetime"))
    df = df.withColumn("fare_amount_cents", df.fare_amount.cast('double') * 100)
    return df

batch_source = HdfsSource(name="nycTaxiBatchSource",
                        path="abfss://feathrazuretest3fs@feathrazuretest3storage.dfs.core.windows.net/demo_data/green_tripdata_2020-04.csv",
                        preprocessing=add_new_dropoff_and_fare_amount_column,
                        event_timestamp_column="new_lpep_dropoff_datetime",
                        timestamp_format="yyyy-MM-dd HH:mm:ss")

Accessing Features

# Requested features to be joined
# Define the key for your feature
location_id = TypedKey(key_column="DOLocationID",
                       key_column_type=ValueType.INT32,
                       description="location id in NYC",
                       full_name="nyc_taxi.location_id")
feature_query = FeatureQuery(feature_list=["f_location_avg_fare"], key=[location_id])

# Observation dataset settings
settings = ObservationSettings(
  observation_path="abfss://green_tripdata_2020-04.csv",    # Path to your observation data
  event_timestamp_column="lpep_dropoff_datetime",           # Event timepstamp field for your data, optional
  timestamp_format="yyyy-MM-dd HH:mm:ss")                   # Event timestamp format， optional

# Prepare training data by joining features to the input (observation) data.
# feature-join.conf and features.conf are detected and used automatically.
feathr_client.get_offline_features(observation_settings=settings,
                                   output_path="abfss://output.avro",
                                   feature_query=feature_query)

Deploy Features to Online (Redis) Store

client = FeathrClient()
redisSink = RedisSink(table_name="nycTaxiDemoFeature")
# Materialize two features into a redis table.
settings = MaterializationSettings("nycTaxiMaterializationJob",
sinks=[redisSink],
feature_names=["f_location_avg_fare", "f_location_max_fare"])
client.materialize_features(settings)

And get features from online store:

# Get features for a locationId (key)
client.get_online_features(feature_table = "agg_features",
                           key = "265",
                           feature_names = ['f_location_avg_fare', 'f_location_max_fare'])
# Batch get for multiple locationIds (keys)
client.multi_get_online_features(feature_table = "agg_features",
                                 key = ["239", "265"],
                                 feature_names = ['f_location_avg_fare', 'f_location_max_fare'])

Defining Window Aggregation Features

agg_features = [Feature(name="f_location_avg_fare",
                        key=location_id,                          # Query/join key of the feature(group)
                        feature_type=FLOAT,
                        transform=WindowAggTransformation(        # Window Aggregation transformation
                            agg_expr="cast_float(fare_amount)",
                            agg_func="AVG",                       # Apply average aggregation over the window
                            window="90d")),                       # Over a 90-day window
                ]

agg_anchor = FeatureAnchor(name="aggregationFeatures",
                           source=batch_source,
                           features=agg_features)

Defining Named Data Sources

batch_source = HdfsSource(
    name="nycTaxiBatchSource",                              # Source name to enrich your metadata
    path="abfss://green_tripdata_2020-04.csv",              # Path to your data
    event_timestamp_column="lpep_dropoff_datetime",         # Event timestamp for point-in-time correctness
    timestamp_format="yyyy-MM-dd HH:mm:ss")                 # Supports various fromats inculding epoch

Beyond Features on Raw Data Sources - Derived Features

# Compute a new feature(a.k.a. derived feature) on top of an existing feature
derived_feature = DerivedFeature(name="f_trip_time_distance",
                                 feature_type=FLOAT,
                                 key=trip_key,
                                 input_features=[f_trip_distance, f_trip_time_duration],
                                 transform="f_trip_distance * f_trip_time_duration")

# Another example to compute embedding similarity
user_embedding = Feature(name="user_embedding", feature_type=DENSE_VECTOR, key=user_key)
item_embedding = Feature(name="item_embedding", feature_type=DENSE_VECTOR, key=item_key)

user_item_similarity = DerivedFeature(name="user_item_similarity",
                                      feature_type=FLOAT,
                                      key=[user_key, item_key],
                                      input_features=[user_embedding, item_embedding],
                                      transform="cosine_similarity(user_embedding, item_embedding)")

Running Feathr Examples

Follow the quick start Jupyter Notebook to try it out. There is also a companion quick start guide containing a bit more explanation on the notebook.

Cloud Integrations

Feathr component	Cloud Integrations
Offline store – Object Store	Azure Blob Storage, Azure ADLS Gen2, AWS S3
Offline store – SQL	Azure SQL DB, Azure Synapse Dedicated SQL Pools, Azure SQL in VM, Snowflake
Online store	Azure Cache for Redis
Feature Registry	Azure Purview
Compute Engine	Azure Synapse Spark Pools, Databricks
Machine Learning Platform	Azure Machine Learning, Jupyter Notebook
File Format	Parquet, ORC, Avro, Delta Lake

Roadmap

Public Preview release may introduce API changes.

Community Guidelines

Build for the community and build by the community. Check out Community Guidelines.

Slack Channel

Join our Slack channel for questions and discussions (or click the invitation link).

Micseb/feathr