Feast feature store basics in MLOps - Practice Problems & Coding Challenges

Challenge - 5 Problems

🧠 Conceptual

intermediate

What is the primary purpose of Feast in MLOps?

Feast is a feature store used in machine learning projects. What is its main role?

A To store and serve machine learning features consistently for training and serving.

B To train machine learning models automatically without human input.

C To monitor model performance in production environments.

D To visualize data trends and generate reports for stakeholders.

💡 Hint

Think about what a feature store does in the ML workflow.

Feast's main role is to provide a centralized place to store and serve features so that training and serving use the same data, ensuring consistency.

💻 Command Output

intermediate

What is the output of this Feast CLI command?

Given the command feast feature-view list executed in a project with two feature views named user_features and transaction_features, what will be the output?

MLOps

feast feature-view list

A user_features

B No features found

C Error: Command not recognized

D

user_features
transaction_features

💡 Hint

This command lists all registered feature views in the current Feast project.

The feast feature-view list command outputs all feature views registered in the project, so it lists both user_features and transaction_features.

❓ Configuration

advanced

Identify the correct Feast feature view YAML snippet

Which YAML snippet correctly defines a Feast feature view named user_activity with an entity user_id and two features clicks (int64) and page_views (int64)?

A

teuqrap.ytivitca_resu/atad/ :htap  
elif :epyt  
:ecruos_hctab
46tni :epytd    
sweiv_egap :eman -  
46tni :epytd    
skcilc :eman -  
:serutaef
di_resu -  
:seititne
ytivitca_resu :eman

B

name: user_activity
entities:
  - user_id
features:
  - name: clicks
    dtype: int64
  - name: page_views
    dtype: int64
batch_source:
  type: file
  path: /data/user_activity.parquet

C

feature_view:
  name: user_activity
  entities: user_id
  features:
    clicks: int64
    page_views: int64
  source: /data/user_activity.csv

D

feature_view:
  name: user_activity
  entity: user_id
  features:
    - clicks: int64
    - page_views: int64
  batch_source:
    type: file
    path: /data/user_activity.parquet

💡 Hint

Look for correct Feast YAML structure with name, entities, features, and batch_source.

Option B correctly uses Feast's YAML format with name, a list of entities, a list of features each with name and dtype, and a batch_source specifying file type and path.

🔀 Workflow

advanced

What is the correct sequence to register and materialize features in Feast?

Arrange the steps in the correct order to register a new feature view and materialize its data in Feast.

A 1,3,2,4

B 2,1,3,4

C 1,2,3,4

D 3,2,1,4

💡 Hint

Think about defining first, then registering, then loading data, then using features.

The correct workflow is to first define the feature view, then register it with feast apply, then materialize data to the online store, and finally use the features in training or serving.

❓ Troubleshoot

expert

Why does 'feast materialize' fail with a timestamp error?

You run feast materialize 2023-01-01T00:00:00 2023-01-02T00:00:00 but get an error saying Invalid timestamp format. What is the most likely cause?

A The timestamps are missing timezone information and Feast expects UTC with 'Z' suffix.

B The Feast CLI does not support materialize commands with timestamps.

C The feature view is not registered, so materialize cannot run.

D The timestamps are in the future and Feast rejects future dates.

💡 Hint

Check the timestamp format Feast expects for materialize commands.

Feast expects timestamps in ISO 8601 format with timezone info, usually ending with 'Z' for UTC. Missing this causes a format error.