Reproducible training pipelines in MLOps - Interactive Code Practice

Practice - 5 Tasks

1fill in blank

easy

Complete the code to specify the version of the dataset for reproducibility.

MLOps

dataset = load_dataset('my_data', version='[1]')

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

A1.0.0

Bbeta

Clatest

Ddev

Attempts:

3 left

💡 Hint

Common Mistakes

Using 'latest' causes dataset to change over time, breaking reproducibility.

Using a fixed version like '1.0.0' ensures the same dataset is used every time, making the training reproducible.

2fill in blank

medium

Complete the code to save the model with a fixed name for reproducibility.

MLOps

model.save('[1]')

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

Atemp_model.pkl

Bmodel_latest.pkl

Cmodel_v1.pkl

Dcheckpoint

Attempts:

3 left

💡 Hint

Common Mistakes

Using generic or changing filenames makes it hard to reproduce results.

Saving the model as 'model_v1.pkl' helps track the exact model version used in training.

3fill in blank

hard

Fix the error in the command to install a specific package version for reproducibility.

MLOps

pip install tensorflow[1]

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

A>2.11.0

B==2.11.0

C~=2.11

D<2.11.0

Attempts:

3 left

💡 Hint

Common Mistakes

Using '>2.11.0' installs newer versions, breaking reproducibility.

Using '==2.11.0' installs exactly that version, ensuring consistent environment for reproducible training.

4fill in blank

hard

Fill both blanks to create a reproducible training pipeline step that logs parameters and artifacts.

MLOps

mlflow.log_param('[1]', value)
mlflow.log_artifact('[2]')

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

Alearning_rate

Bmodel.pkl

Cdataset.csv

Dbatch_size

Attempts:

3 left

💡 Hint

Common Mistakes

Logging dataset.csv as artifact instead of model file.

Logging 'learning_rate' as a parameter and 'model.pkl' as an artifact helps track training details for reproducibility.

5fill in blank

hard

Fill all three blanks to define a reproducible pipeline step that filters data and trains a model.

MLOps

filtered_data = data[data['age'] [1] [2]]
model = train_model(filtered_data, epochs=[3])

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

A>

B30

C10

D<

Attempts:

3 left

💡 Hint

Common Mistakes

Using '<' instead of '>' changes the data subset.

Filtering data where age is greater than 30 and training for 10 epochs ensures consistent training conditions.