Why DVC (Data Version Control) basics in MLOps? - Purpose & Use Cases

The Big Idea

What if you could track your data changes as easily as your code, without losing hours or files?

The Scenario

Imagine you are working on a machine learning project with lots of data files and models. You save different versions of your data by copying files into folders named "version1", "version2", and so on. You email these folders to your teammates or upload them to shared drives.

The Problem

This manual way is slow and confusing. You waste time finding the right data version. Files get lost or overwritten. Your teammates don't know which data matches which model. It's easy to make mistakes and hard to fix them.

The Solution

DVC helps you track data and models just like code. It stores versions efficiently and links data to your code changes. You can share and reproduce experiments easily. DVC automates data management so you focus on building models, not juggling files.

Before vs After

✗ Before

cp data.csv data_v1.csv
cp data.csv data_v2.csv

✓ After

dvc add data.csv
dvc push

What It Enables

DVC makes managing data versions simple and reliable, enabling smooth collaboration and reproducible machine learning projects.

Real Life Example

A data scientist updates a dataset and runs new experiments. With DVC, they track the changes, share results with the team, and roll back to previous data if needed—all without confusion or lost files.

Key Takeaways

Manual data versioning is slow and error-prone.

DVC automates tracking and sharing of data and models.

This leads to easier collaboration and reproducible results.

Practice

1. What is the main purpose of using dvc add in a project?

easy

A. To push code changes to a remote Git server

B. To initialize a new Git repository

C. To start tracking a data file or directory with DVC

D. To remove data files from the project

Solution

1. Step 1: Understand the role of dvc add

   dvc add is used to tell DVC to track a data file or directory, creating a pointer file in Git.

2. Step 2: Differentiate from other commands

   Commands like dvc init start DVC, while dvc push syncs data remotely. dvc add specifically tracks data.

3. Final Answer:

   To start tracking a data file or directory with DVC -> Option C

4. Quick Check:

   dvc add tracks data files [OK]

Hint: Remember: add means track data files with DVC [OK]

Common Mistakes:

• Confusing dvc add with dvc init
• Thinking dvc add pushes data remotely
• Assuming dvc add initializes Git

2. Which command correctly initializes DVC in an existing Git repository?

easy

A. dvc start

B. dvc init

C. git dvc init

D. dvc create

Solution

1. Step 1: Identify the DVC initialization command

   The correct command to initialize DVC in a Git repo is dvc init.

2. Step 2: Eliminate incorrect options

   dvc start and dvc create are not valid DVC commands. git dvc init is invalid syntax.

3. Final Answer:

   dvc init -> Option B

4. Quick Check:

   DVC init command = dvc init [OK]

Hint: Use dvc init to start DVC in your repo [OK]

Common Mistakes:

• Typing dvc start instead of dvc init
• Prefixing with git incorrectly
• Using non-existent commands like dvc create

3. Given the following commands run in order:

git init
 dvc init
 dvc add data.csv
 git add data.csv.dvc
 git commit -m "Add data"
 dvc push

What happens after dvc push is executed?

medium

A. The data file is deleted locally after upload

B. Only the data.csv.dvc pointer file is pushed to Git remote

C. The Git repository is cloned to remote storage

D. The actual data file data.csv is uploaded to remote storage

Solution

1. Step 1: Understand dvc push behavior

   dvc push uploads the actual large data files tracked by DVC to the configured remote storage, not just Git files.

2. Step 2: Differentiate Git and DVC storage roles

   Git stores small pointer files like data.csv.dvc, while DVC manages big data files separately in remote storage.

3. Final Answer:

   The actual data file data.csv is uploaded to remote storage -> Option D

4. Quick Check:

   dvc push uploads data files remotely [OK]

Hint: dvc push uploads big data files, not just pointers [OK]

Common Mistakes:

• Thinking dvc push only pushes Git files
• Confusing dvc push with git push
• Assuming data files are deleted after push

4. You ran dvc add dataset.csv but forgot to commit the generated dataset.csv.dvc file to Git. What problem will you face?

medium

A. DVC will not track the data file until the pointer file is committed

B. The data file will be deleted automatically

C. Git will track the data file instead of DVC

D. No problem; DVC tracks data without Git commits

Solution

1. Step 1: Understand the role of the .dvc pointer file

   The dataset.csv.dvc file is a small pointer tracked by Git that tells DVC about the data file version.

2. Step 2: Consequence of not committing the pointer file

   If you don't commit this pointer file, Git and collaborators won't know about the data version, so DVC tracking is incomplete.

3. Final Answer:

   DVC will not track the data file until the pointer file is committed -> Option A

4. Quick Check:

   Pointer file commit = DVC tracking active [OK]

Hint: Always commit .dvc pointer files after dvc add [OK]

Common Mistakes:

• Assuming data files are tracked without pointer commits
• Thinking data files get deleted automatically
• Believing Git tracks large data files directly

5. You have a large dataset tracked by DVC and a remote storage configured. Your teammate cloned the Git repo but the data files are missing locally. Which command should they run to get the data files?

hard

A. dvc pull

B. dvc add

C. git pull

D. git clone

Solution

1. Step 1: Understand what dvc pull does

   dvc pull downloads the actual data files from remote storage to the local machine based on the pointer files in Git.

2. Step 2: Differentiate from Git commands

   git pull updates code and pointer files but does not fetch large data files. dvc add tracks new data, and git clone clones the repo initially.

3. Final Answer:

   dvc pull -> Option A

4. Quick Check:

   Use dvc pull to fetch data files locally [OK]

Hint: Use dvc pull to download data after cloning repo [OK]

Common Mistakes:

• Running only git pull expecting data files
• Trying dvc add to get data files
• Confusing git clone with data download