Saltar al contenido principal

CICD Pipelines

Updated May 12, 2023 ·

Overview​

CI/CD (Continuous Integration and Continuous Deployment) is a key part of the deployment phase. The CI/CD pipeline automates checks and processes to ensure that code is ready for production.

CICD​

Continuous Integration (CI) means frequently integrating code changes and testing them automatically.

  • Changes are tested as soon as they are committed.
  • CI tests each code change to ensure it works.

Continuous Deployment (CD) automates the release of validated code after testing.

  • After testing, the new code is deployed automatically.

For more information, please see CICD Overview.

Deployment Strategies​

Once a machine learning model is ready for deployment, we have several strategies for releasing it into production.

For more information, please see Deployment Strategies.

Automation and Scaling​

Automation and scaling helps speed up processes and handle large datasets more efficiently. Here's how automation and scaling fit into different stages of the machine learning lifecycle:

  • Design Phase

    • Sets the foundation for machine learning
    • Align business needs with technical goals
    • Templatize designs for more structured processes in MLOps

  • Data Acquisition and Quality Checks

    • Automated data collection improves the model's success rate
    • Automate data checks for quality
    • Ensures better machine learning model performance

  • Development Phase

    • Focused on building features and experiments
    • Use feature stores to save time
    • Automate experiment tracking for progress and reproducibility.

  • Deployment Phase

    • Use containerization for flexible scaling
    • Set up CI/CD pipelines for faster, automated updates
    • Microservices architecture helps scale individual parts independently

For more information, please see: