Skip to content

Latest commit

 

History

History
143 lines (115 loc) · 2.9 KB

endusers.md

File metadata and controls

143 lines (115 loc) · 2.9 KB
Raw

Feast End Users Quickstart Guide

Pre-requisities

Make sure your CLI is correctly configured for your Feast Core. If you're running a local Feast Core, it would be:

feast config set coreURI localhost

Introduction

There are several stages to using Feast:

  1. Register your feature
  2. Ingest data for your feature
  3. Query feature data for training your models
  4. Query feature data for serving your models

Registering your feature

In order to register a feature, you will first need to register a:

  • Storage location (typically done by your Feast admin)
  • Entity

All registrations are done using specs.

Registering an entity

Then register an entity, which is for grouping features under a unique key or id. Typically these map to a domain object, e.g., a customer, a merchant, a sales region.

wordEntity.yml

name: word
description: word found in shakespearean works

Register the entity spec:

feast apply entity wordEntity.yml

Registering your feature

Next, define your feature:

wordCountFeature.yml

id: word.count
name: count
entity: word
owner: bob@feast.com
description: number of times the word appears
valueType:  INT64
uri: https://github.com/bob/example

Register it:

feast apply feature wordCountFeature.yml

Ingest data for your feature

Feast supports ingesting feature from 4 type of sources:

  • File (either CSV or JSON)
  • Bigquery Table
  • Pubsub Topic
  • Pubsub Subscription

Let's take a look on how to create an import job spec and ingest some data from a CSV file. You may find more information on how to ingest data from different sources here: [Import Specs]

Prepare your data

word_counts.csv

count,word
28944,the
27317,and
21120,i
20136,to
17181,of
14945,a
13989,you
12949,my
11513,in
11488,that
9545,is
8855,not
8293,with
8043,me
8003,it
...

And then upload it into your Google Storage bucket:

gsutil cp word_counts.csv gs://your-bucket

Define the job import spec

shakespeareWordCountsImport.yml

type: file.csv
sourceOptions:
  path: gs://your-bucket/word_counts.csv
entities:
  - word
schema:
  entityIdColumn: word
  timestampValue: 2019-01-01T00:00:00.000Z
  fields:    
    - name: count
      featureId: word.count
    - name: word

Start the ingestion job

Next, use feast CLI to run your ingestion job, defined in shakespeareWordCountsImport.yml:

feast jobs run shakespeareWordCountsImport.yml

You can also list recent ingestion jobs by running:

feast list jobs

Or get detailed information about the results of ingestion with:

feast get job <id>