Point-in-time correctness in MLOps - Interactive Code Practice

Practice - 5 Tasks

1fill in blank

easy

Complete the code to load a model version at a specific timestamp.

MLOps

model = mlflow.pyfunc.load_model(model_uri='models:/my_model@[1]')

Drag options to blanks, or click blank then click option'

Alatest

B2023-05-01T12:00:00Z

Cproduction

Dstaging

Attempts:

3 left

💡 Hint

Common Mistakes

Using 'latest' loads the newest model, not the one at a specific time.

Using a timestamp like '2023-05-01T12:00:00Z' ensures you load the model as it was at that exact time, maintaining point-in-time correctness.

2fill in blank

medium

Complete the code to query data as it existed at a specific point in time.

MLOps

query = "SELECT * FROM feature_store WHERE event_time <= '[1]'"

Drag options to blanks, or click blank then click option'

Alatest

Bnow()

C2023-04-30T23:59:59Z

D2023-06-01

Attempts:

3 left

💡 Hint

Common Mistakes

Using 'now()' includes current data, breaking point-in-time correctness.

Using a timestamp like '2023-04-30T23:59:59Z' in the query filters data up to that exact point, ensuring point-in-time correctness.

3fill in blank

hard

Fix the error in the code to ensure point-in-time correctness when loading features.

MLOps

features = feature_store.get_features(entity_id=123, as_of_time=[1])

Drag options to blanks, or click blank then click option'

A'2023-05-15T10:00:00Z'

B'latest'

Cdatetime.now()

DNone

Attempts:

3 left

💡 Hint

Common Mistakes

Using datetime.now() loads current features, breaking point-in-time correctness.

Passing a fixed timestamp string like '2023-05-15T10:00:00Z' ensures features are loaded as they existed at that time, preserving point-in-time correctness.

4fill in blank

hard

Fill both blanks to create a feature retrieval query that respects point-in-time correctness.

MLOps

SELECT feature_value FROM features WHERE entity_id = [1] AND event_timestamp <= '[2]'

Drag options to blanks, or click blank then click option'

A12345

B2023-05-10T08:30:00Z

C67890

D2023-06-01T00:00:00Z

Attempts:

3 left

💡 Hint

Common Mistakes

Using a future timestamp or wrong entity_id breaks point-in-time correctness.

Using entity_id 12345 and filtering event_timestamp up to '2023-05-10T08:30:00Z' ensures the query returns features as they existed at that time.

5fill in blank

hard

Fill all three blanks to build a dictionary comprehension that filters features for point-in-time correctness.

MLOps

filtered_features = {k: v for k, v in features.items() if v['timestamp'] [1] '[2]' and k != [3]

Drag options to blanks, or click blank then click option'

A<=

B2023-05-20T15:00:00Z

C'deprecated_feature'

D==

Attempts:

3 left

💡 Hint

Common Mistakes

Using '==' instead of '!=' excludes the wrong features.

The operator '<=' with timestamp '2023-05-20T15:00:00Z' filters features up to that time, and excluding 'deprecated_feature' ensures only relevant features are included.