Module 2 Quiz: Version Control for ML

Test your understanding of version control practices for ML projects with these 30 questions.

---

Question 1

What is DVC primarily used for in ML projects?

1. Code formatting
2. Versioning data and models alongside code
3. Deploying models to production
4. Monitoring model performance

---

Question 2

Which file does DVC create when tracking a data file?

1. .dvcconfig
2. .dvc file (e.g., data.csv.dvc)
3. .dvcignore
4. .gitattributes

---

Question 3

What is the purpose of .gitignore in an ML project?

1. To ignore all Python files
2. To exclude files from Git tracking (like large datasets and models)
3. To ignore all DVC files
4. To track only important files

---

Question 4

Which command initializes DVC in a project?

1. dvc start
2. dvc setup
3. dvc init
4. dvc create

---

Question 5

What is Git LFS used for?

1. Managing Git branches
2. Handling large files in Git repositories
3. Deploying Git repositories
4. Creating Git hooks

---

Question 6

In semantic versioning (e.g., 2.1.3), what does the middle number represent?

1. Major version (breaking changes)
2. Minor version (new features, backward compatible)
3. Patch version (bug fixes)
4. Build number

---

Question 7

What does the command dvc push do?

1. Pushes code to GitHub
2. Uploads tracked data/models to remote storage
3. Creates a new DVC branch
4. Installs DVC dependencies

---

Question 8

Which storage backends does DVC support for remote storage?

1. Only local storage
2. Only AWS S3
3. S3, GCS, Azure, SSH, and more
4. Only cloud storage

---

Question 9

What is a model registry?

1. A database of training data
2. A centralized repository for storing, versioning, and managing ML models
3. A government database of AI models
4. A list of model hyperparameters

---

Question 10

Which tool provides a model registry feature?

1. Git
2. DVC only
3. MLflow
4. Docker

---

Question 11

What is model lineage?

1. The number of layers in a neural network
2. Tracking the complete history from data to model deployment
3. The age of a model in production
4. The family tree of model architectures

---

Question 12

What does dvc repro do?

1. Reports errors in the pipeline
2. Reproduces the ML pipeline by running necessary stages
3. Replicates the repository
4. Creates a repository backup

---

Question 13

Which file defines DVC pipelines?

1. pipeline.yaml
2. dvc.yaml
3. Dvcfile
4. stages.yaml

---

Question 14

What is the purpose of pre-commit hooks in ML projects?

1. To pre-train models
2. To run automated checks before commits (linting, formatting, etc.)
3. To commit code automatically
4. To create commit messages

---

Question 15

What does nbstripout do for Jupyter notebooks?

1. Removes code cells from notebooks
2. Strips output cells and metadata before committing
3. Converts notebooks to Python scripts
4. Encrypts notebook content

---

Question 16

In MLflow, what stages can a model version be in?

1. Draft, Published, Deleted
2. None, Staging, Production, Archived
3. New, Testing, Live, Deprecated
4. Created, Validated, Deployed

---

Question 17

What is a model card?

1. A physical card containing model weights
2. Structured documentation describing a model's purpose, performance, and limitations
3. A credit card for cloud computing
4. A visualization of model architecture

---

Question 18

What command shows the DVC pipeline as a graph?

1. dvc graph
2. dvc show
3. dvc dag
4. dvc pipeline

---

Question 19

What is the recommended branching strategy for ML experiments?

1. All experiments in main branch
2. Separate experiment branches that can be merged when successful
3. No branching, just tags
4. One branch per data scientist

---

Question 20

What does dvc checkout do?

1. Checkouts code from git
2. Updates workspace data files to match the current Git commit
3. Validates data integrity
4. Removes all tracked data

---

Question 21

Why is reproducibility important in ML projects?

1. To make code run faster
2. To recreate the same results with the same inputs for validation and debugging
3. To reduce storage costs
4. To automate deployment

---

Question 22

What information should be tracked for model lineage?

1. Only the model file
2. Only the training code
3. Code version, data version, hyperparameters, environment, and metrics
4. Only the deployment configuration

---

Question 23

What is the purpose of dvc.lock?

1. To prevent others from running the pipeline
2. To record the exact state of inputs and outputs for reproducibility
3. To lock files for editing
4. To encrypt sensitive data

---

Question 24

Which command runs DVC experiments with different parameters?

1. dvc run --params
2. dvc exp run --set-param
3. dvc experiment new
4. dvc params set

---

Question 25

What is the recommended way to handle Jupyter notebooks in version control?

1. Track all outputs for reproducibility
2. Strip outputs before committing and/or use Jupytext
3. Never version control notebooks
4. Convert all notebooks to PDFs

---

Question 26

What does model validation in a registry typically include?

1. Only checking if the model file exists
2. Performance tests, latency checks, and comparison with production
3. Only code review
4. Only documentation review

---

Question 27

What is Jupytext?

1. A text editor for Jupyter
2. A tool that keeps notebooks and Python scripts synchronized
3. A Jupyter theme
4. A notebook encryption tool

---

Question 28

In DVC, what is a remote?

1. A remote machine for training
2. External storage location for data and model files
3. A remote Git repository
4. A remote desktop connection

---

Question 29

What is the benefit of commit message conventions in ML projects?

1. Makes commits look professional
2. Creates a clear, searchable history of changes by type (data, model, training)
3. Required by GitHub
4. Automatically fixes code

---

Question 30

What should trigger a model version bump from 1.0.0 to 2.0.0?

1. Any retraining of the model
2. Bug fixes
3. Breaking changes like new architecture or incompatible API changes
4. Adding documentation

---

🎉 Quiz Complete!

You've completed the Module 2 quiz. Review any questions you got wrong and revisit the relevant sections.

Score Interpretation:

• 27-30 correct: Excellent! You have mastered version control for ML.
• 22-26 correct: Good understanding, review missed topics.
• 17-21 correct: Fair, recommend re-reading the module.
• Below 17: Review the entire module before proceeding.

---

Next: Module 3: Experiment Tracking