Prepare a model for deployment (20–25%)

Question 1

What are three things you should do to turn a notebook into a script?

Answer 1

Remove nonessential code, refactor your code into functions, test your script in the terminal

Question 2

How to configure a command job to run a script?

Answer 2

command(code: the folder of the script, command: CLI entry to run, environment: the packages needed to run, compute: which compute to run, display_name: name of the job, experiment_name: The experiment the job belongs to)

Question 3

How can you add parameters to a command job?

Answer 3

You can use a library like argparse

Question 4

Which libraries support autologging?

Answer 4

Scikit-learn, TensorFlow and Keras, XGBoost, LightGBM, Spark, Fastai, Pytorch

Question 5

What are the three parts to a component?

Answer 5

Metadata: component name, version, etc., interface: expected input parameters and output, command, code, and environment: specifies how to run the code

Question 6

What two files are needed to create a component?

Answer 6

A script with the workflow, and a YAML with metadata, interface, command, code and environment (you can also use command_component() instead of YAML)

Question 7

how to register a component in python sdk?

Answer 7

prep = ml_client.components.create_or_update(prepare_data_component)