v2.0-ci/cd-pipeline-automation

🚀 Release v2.0 – CI/CD Pipeline Automation with MLflow, Hydra, and W&B

This release marks a major milestone in the automation of our machine learning pipeline and deployment workflows. Here's what's included:

✅ Core Improvements

• Full automation of the ML pipeline using:

  • 🔹 MLflow for experiment tracking and reproducibility
  • 🔹 Hydra for flexible and modular configuration management
  • 🔹 Weights & Biases (W&B) for advanced experiment logging, metrics visualization, and artifact tracking

🛠️ CI/CD Integration

• CI powered by GitHub Actions:

  • Linting, testing, and validation on push and pull requests
  • Ensures consistency, reliability, and fast feedback

• CD powered by:

  • 🚀 FastAPI: Serving the trained model with a lightweight, high-performance API
  • 🐳 Docker: Containerized deployment environment for consistency across stages
  • 🌐 Render.com: Hosting the live API for inference and public consumption

📦 Artifacts and Reproducibility

• Model artifacts, preprocessing pipelines, and evaluation reports are versioned and logged
• Seamless switching between model versions for easy rollback and comparisons

v1.0

🚀 Release v1.0 — Full MLOps Automation

This release marks the first stable version of our pipeline with end-to-end automation for a regression task on the Ames Housing dataset. Key features include:

✅ Key Highlights

• ✅ MLflow integration for experiment tracking and reproducible runs
• ⚙️ Hydra-powered configuration management for clean and flexible parameter control
• 📊 Weights & Biases (W&B) logging for model metrics, artifacts, plots, and traceability
• 📁 Modular architecture with support for preprocessing, training, evaluation, and inference
• 📦 Full compatibility with mlflow run . -P steps="..." commands
• 🧪 Metrics evaluation using custom regression metrics
• 🧠 Feature selection and preprocessing pipeline tracking

This version is a robust foundation for reproducible and scalable machine learning pipelines.

v0.1.0-notebook-to-mlops-project

This release is the first deliverable of Group 5's MLOps project. For this deliverable, we managed to take our Ames Housing Price notebook, and turn it into a functional pipeline. Our next steps are to fully automate this pipeline using the following technological stack: MLflow projects, Hydra, W&B, GitHub Actions, and FastAPI.