Simplified SageMaker Feature Store
Project description
aws_feature_store
It is a simplified implementation of SageMaker Feature Store approach.
Installation
Use the package manager pip to install foobar.
pip install aws_feature_store
Initialize feature group
from aws_feature_store import FeatureGroup,FeatureDefinition,FeatureTypeEnum
bucket_name = '{bucket_for_feature_store}'
s3_folder = '{folder_for_feature_store}'
my_feature_name = '{your_feature_name}'
feature_group_name = f'{my_feature_name}/commit_id={my_feature_name}_{commit_id}'
feature_group = FeatureGroup(
name=feature_group_name,
boto3_session = boto3_session,
s3_uri=f"s3://{bucket_name}/{s3_folder}"
)
Create feature group
def create_feature_group(feature_group):
description="What is my feature group about"
feature_script_repo="{repo_link_to_script}"
data_source="{what data are used}"
record_identifier_feature_name = "column name to store id"
event_time_feature_name = "{column name to store timestamp}"
partition_columns=['biz_id','customer_id']
feature_definitions=[
FeatureDefinition(feature_name="column_name1", feature_type=FeatureTypeEnum.INTEGRAL),
FeatureDefinition(feature_name="column_name2", feature_type=FeatureTypeEnum.STRING),
]
feature_group.create(
record_identifier_name=record_identifier_feature_name,
event_time_feature_name=event_time_feature_name,
feature_script_repo=feature_script_repo,
partition_columns=partition_columns,
data_source=data_source,
description=description,
file_format='parquet/json',
feature_definitions=feature_definitions
)
return feature_group
if feature_group.exists() is None:
feature_group = create_feature_group(feature_group)
Ingest data
import pandas as pd
data = pd.read_json('data.json')
feature_group.ingest_data_frame(data,f"mlfow_parent_run_id={parent_run_id}/{filename_without_extention}")
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
aws_feature_store-0.0.17.tar.gz
(11.3 kB
view hashes)
Built Distribution
Close
Hashes for aws_feature_store-0.0.17-py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | 0396c8f4ab09fa28c9e287f99aff335cd89fe45a2d62344d7dead52dd0b548e2 |
|
MD5 | eaf0763fb7d5675b1db8013549203bd7 |
|
BLAKE2b-256 | 6a9edc130182c60b335547c6eaab1ee6fb6701e8a9b939198d8ce38e01945d25 |