100-days-mlops-kodekloud

Package an ML Project as Installable Python Package

Problem

The xFusionCorp Industries deployment team needs the fraud-detection model code packaged as an installable Python distribution. A draft pyproject.toml exists at /root/code/fraud-detection/, but it does not build a wheel that meets the team’s standard. Correct the file and produce a compliant package.

1. The project at /root/code/fraud-detection/ already contains the source code under src/fraud_detection/. The Python files are complete—you do not need to modify any of them.

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

   • it declares a [build-system] section with requires = ["setuptools>=61.0", "wheel"] and build-backend = "setuptools.build_meta";
   • name is fraud_detection (the distribution name must match the module path under src/);
   • version is 0.1.0;
   • requires-python is >=3.10;
   • dependencies is ["scikit-learn", "pandas", "numpy"].

3. Review the existing pyproject.toml and correct everything that does not match the requirements above.

4. Build the package from the project directory:

   cd /root/code/fraud-detection
   python3 -m build
   

5. The build must produce a wheel named fraud_detection-0.1.0-*.whl under dist/.

The build package is already installed. Use python3 rather than python.

Solution

1. Update the pyproject.toml with the following contents:

    [project]
    name = "fraud_detection"
    version = "0.1.0"
    description = "Fraud detection model for xFusionCorp Industries"
    requires-python = ">=3.10"
    dependencies = ["scikit-learn", "pandas", "numpy"]
   
    [build-system]
    requires = ["setuptools>=61.0", "wheel"]
    build-backend = "setuptools.build_meta"
   

2. Move into fraud-detection directory and run the python build command:

    cd /root/fraud-detection
    python3 -m build
   

   It should a dist directory and print successul result.

3. Check the dist directory and make sure fraud-detection-0.1.0-*.whl wheel package is existing.

This site is open source. Improve this page.