data-drift-detector-mightyhive 0.0.4

A data drift detection and schema validation package

About The Package

The package is a wrapper of tensorflow data validation for our specific needs. It can analyze training data and serving data to compute desscriptive statistics, infer a schema, and detect anomalies.

Dependencies

• tensorflow-data-validation

Installation

pip install data-drift-detector

Usage

Initialize a Harvest client:

# The Dataset, TrainDataset, ServeDataset can be initialized with different methods.

train = TrainDataset.from_GCS()
train = TrainDataset.from_bigquery()
train = TrainDataset.from_dataframe()
train = TrainDataset.from_stats_file()

Populate the class variables and submit.

# Get training dataset schema
schema = train.schema_dict()