This pulls from the dataset used in https://github.com/feast-dev/feast-aws-credit-scoring-tutorial but adds metadata for a full set of Feast FCOs.
This also adds an on demand feature view + feature services + a saved dataset.
Install a dev build Feast using pip
Clone a feast repo:
git clone https://github.com/feast-dev/feast.gitInstall a dev build of feast
cd feast
pip install -e ".[dev]"Then for this demo, you'll actually need to fix a bug by adding this to type_map.py#L144:
if isinstance(value, np.bool_):
return ValueType.BOOLWe have already set up a feature repository here using test data from data). Features have already been pre-materialized to a local sqlite online store. The results of feast registry-dump have been thrown into registry.json
To query against this registry, you can use run the test_get_features.py
python test_get_features.pyOutput:
--- Historical features (from saved dataset) ---
mortgage_due credit_card_due missed_payments_1y total_wages dob_ssn event_timestamp state tax_returns_filed location_type population city zipcode
0 741165 2944 3 71067272 19781116_7723 2021-04-12 08:12:10+00:00 MI 2424 PRIMARY 4421 WEIDMAN 48893
1 91803 8419 0 534687864 19530219_5179 2021-04-12 10:59:42+00:00 GA 19583 PRIMARY 38542 DALTON 30721
2 1553523 5936 0 226748453 19500806_6783 2021-04-12 15:01:12+00:00 TX 6827 PRIMARY 12902 CLEBURNE 76031
3 976522 833 0 34796963 19931128_5771 2021-04-12 16:40:26+00:00 VA 1287 PRIMARY 2342 GLADE HILL 24092
--- Online features ---
city : ['DALTON']
credit_card_due : [8419]
dob_ssn : ['19530219_5179']
location_type : ['PRIMARY']
missed_payments_1y : [0]
mortgage_due : [91803]
population : [38542]
state : ['GA']
tax_returns_filed : [19583]
total_wages : [534687864]
zipcode : [30721]
city : ['DALTON']
credit_card_due : [8419]
dob_ssn : ['19530219_5179']
location_type : ['PRIMARY']
missed_payments_1y : [0]
mortgage_due : [91803]
population : [38542]
state : ['GA']
tax_returns_filed : [19583]
total_wages : [534687864]
transaction_gt_last_credit_card_due : [False]
zipcode : [30721]