100-days-mlops-kodekloud

Create a Custom ML Project Template with Cookiecutter

Problem

The xFusionCorp Industries ML platform team maintains a Cookiecutter template that new ML projects are generated from. A draft template exists at /root/code/mlops-template/, but it does not render. Correct the template and use it to generate a project.

1. A Cookiecutter template exists at /root/code/mlops-template/. cookiecutter is installed system-wide.

2. The corrected template must satisfy every one of the following:

   • The cookiecutter.json declares four variables:
     • project_name (default my-ml-project)
     • author (default xFusionCorp)
     • python_version (default 3.11)
     • ml_framework with the choices sklearn, pytorch, and tensorflow
   • The generated requirements.txt logic:
     • Contains scikit-learn when ml_framework is sklearn
     • Contains torch when ml_framework is pytorch
     • Contains tensorflow when ml_framework is tensorflow
   • The generated README.md content:
     • Must reference both the project_name and the author from cookiecutter variables.
   • The template directory structure / must contain:
     • Files: README.md and requirements.txt
     • Directories: data/, models/, src/, and tests/

3. Review the existing template in the VS Code explorer and correct everything that prevents it from rendering.

4. Once the template renders, generate a project at /root/code/churn-model/:

   cookiecutter /root/code/mlops-template/ -o /root/code/ --no-input project_name=churn-model ml_framework=sklearn
   

5. The generated project must contain a requirements.txt listing scikit-learn and a README.md that mentions xFusionCorp.

Solution

This is a pretty interesting tool. We have to go through the official documentation to fix the issues properly. Like we have to understand how we can cheeck string matching conditions, nested if-else block, writing cookiecutter.json files configurations, etc. Let’s see what we had to fix:

1. Let’s open the cookiecutter.json file in editor. We can notice several issues like missing variables, wrong variable names, etc.

2. Let’s update with the correct config;

    {
        "project_name": "my-ml-project",
        "author": "xFusionCorp",
        "python_version": "3.11",
        "ml_framework": ["sklearn", "pytorch", "tensorflow"]
    }
   

3. Let’s open the requirements.txt file in editor and update according to this template: day 09 cookiecutter_requirements.txt

   we have replaced = with == to ensure string matching Added conditional ending clause.

4. Updated README.md with proper variable referance name. Updated Author to author to call actual variable.

5. That’s all. You can run the given command:

    cookiecutter /root/code/mlops-template/ -o /root/code/ --no-input project_name=churn-model ml_framework=sklearn
   

   If everything is done correctly, It will create the files properly. Otherwise you will get errors. Read them carefully and fix the further issues. Make sure all the directory exists, requirements.txt is created, and README.md updated before submitting the solution.

cookiecutter Fundamentals

• cookiecutter.json is the configuration file that declares the variables and their default values. It can also declare choices for variables.
• In the template files, we can use Jinja2 templating syntax to reference the variables and write conditional logic. For example,
• To reference a variable: ``
• To write conditional logic we can use if else condition.

This site is open source. Improve this page.