Module 3 Quiz: Experiment Tracking

Test your understanding of experiment tracking with these 30 questions.

---

Question 1

What are the four main components of MLflow?

1. Tracking, Projects, Models, Registry
2. Training, Pipelines, Models, Repository
3. Tracking, Packaging, Monitoring, Registry
4. Testing, Projects, Metrics, Registry

---

Question 2

What does MLflow autolog do?

1. Automatically logs in to MLflow server
2. Automatically logs parameters, metrics, and models without explicit log calls
3. Automatically logs errors and exceptions
4. Automatically logs Git commits

---

Question 3

What is a W&B sweep?

1. A tool for cleaning up old experiments
2. An automated hyperparameter optimization system
3. A method for sweeping through data
4. A visualization type for metrics

---

Question 4

What is a model signature in MLflow?

1. A digital signature for security
2. A specification of the model's input and output schema
3. The author's name on the model
4. A unique identifier for the model

---

Question 5

Which W&B sweep method uses Bayesian optimization?

1. grid
2. random
3. bayes
4. exhaustive

---

Question 6

What is an MLflow artifact?

1. An old, deprecated model
2. Any file or directory logged with a run (models, plots, data)
3. A type of metric
4. A configuration file

---

Question 7

What is the purpose of nested runs in MLflow?

1. To run experiments in parallel
2. To organize related runs (like hyperparameter search) under a parent run
3. To nest models inside each other
4. To create recursive experiments

---

Question 8

What is a W&B artifact?

1. A museum display of ML models
2. A versioned, trackable unit like datasets or models
3. An error message
4. A visualization type

---

Question 9

Which MLflow command runs a project from a Git repository?

1. mlflow git run
2. mlflow run https://github.com/...
3. mlflow clone and run
4. mlflow project github

---

Question 10

What file defines an MLflow project?

1. mlflow.yaml
2. project.yaml
3. MLproject
4. config.yaml

---

Question 11

What is the benefit of experiment tracking?

1. It makes training faster
2. It enables reproducibility and comparison of experiments
3. It reduces model size
4. It automatically improves model accuracy

---

Question 12

How do you log a custom metric in W&B?

1. wandb.metric("name", value)
2. wandb.log({"metric_name": value})
3. wandb.track("name", value)
4. wandb.add_metric("name", value)

---

Question 13

What is the purpose of mlflow.set_experiment()?

1. To start an experiment run
2. To set the active experiment for subsequent runs
3. To configure experiment parameters
4. To end an experiment

---

Question 14

What visualization is best for comparing hyperparameter impact?

1. Pie chart
2. Parallel coordinates plot
3. Histogram
4. Scatter plot only

---

Question 15

What does wandb.init() do?

1. Installs W&B
2. Initializes a new run and connects to W&B
3. Initializes the neural network
4. Resets all previous runs

---

Question 16

What is a W&B Table?

1. A database table
2. A structured data format for logging and visualizing tabular data
3. A comparison table between models
4. A HTML table in reports

---

Question 17

What backend store types does MLflow support?

1. Only local files
2. Only PostgreSQL
3. Local files, SQLite, PostgreSQL, MySQL, and more
4. Only cloud databases

---

Question 18

What is the purpose of wandb.finish()?

1. To delete the run
2. To properly close the run and upload remaining data
3. To finish training
4. To logout from W&B

---

Question 19

How do you compare experiments in MLflow UI?

1. Use the command line only
2. Select multiple runs and use the compare view
3. Export to Excel and compare manually
4. MLflow doesn't support comparison

---

Question 20

What is an input example in MLflow models?

1. Example code for using the model
2. Sample input data stored with the model for reference
3. Training data examples
4. Error examples

---

Question 21

What is the benefit of W&B Reports?

1. Automated model training
2. Shareable documents combining text, code, and visualizations
3. Bug reporting
4. Error logging

---

Question 22

How do you filter MLflow runs programmatically?

1. mlflow.filter_runs()
2. client.search_runs() with filter_string
3. mlflow.query_runs()
4. mlflow.get_filtered_runs()

---

Question 23

What is a model flavor in MLflow?

1. The taste of the model
2. A specific format for saving models (sklearn, pytorch, etc.)
3. The model's version
4. A type of hyperparameter

---

Question 24

What does wandb.config provide?

1. Server configuration
2. Access to hyperparameters set during initialization
3. Network configuration
4. Display settings

---

Question 25

Why log training curves during training?

1. It's required by MLflow
2. To monitor convergence and detect issues early
3. To slow down training
4. For aesthetic purposes only

---

Question 26

What is the default artifact location in MLflow?

1. Remote S3 bucket
2. mlruns directory in the current working directory
3. /var/mlflow/artifacts
4. Cloud storage only

---

Question 27

How do you use a W&B artifact in a new run?

1. wandb.get_artifact()
2. wandb.use_artifact("name:version")
3. wandb.download_artifact()
4. wandb.load_artifact()

---

Question 28

What is best practice for experiment naming?

1. Use random names
2. Use only dates
3. Use descriptive names with key configuration info
4. Use auto-generated IDs only

---

Question 29

What happens if you don't call wandb.finish()?

1. The run is deleted
2. Nothing, it's optional
3. Data may not be fully uploaded, run may appear incomplete
4. W&B crashes

---

Question 30

What should be logged for reproducible experiments?

1. Only metrics
2. Only model weights
3. Parameters, metrics, artifacts, environment info, and code version
4. Only the final model

---

🎉 Quiz Complete!

You've completed the Module 3 quiz on Experiment Tracking.

---

Next: Module 4: Data Pipelines