An accelerator for adapting MLOPs to bespoke data science projects with practical examples of using these components to achieve an end-to-end workflow from data cleaning, EDA and visualization to model, feature stores, model training, serving deployment and monitoring.
- Entirely code-driven for a customizable workflow
- Supports:
- Experiment Tracking via MLFlow
- Feature Store ising Feast
- Exploratory Data Analysis
- Custom Python packages
- Training Pipelines Examples
- Inference Pipelines Examples
The "toolkit" itself is located in the src folder and includes a Python cookiecutter package so folder structure can be customized.
Launch the platform by running
docker compose up
inside the infrastructure folder (a helm version can be added in future for cloud deployments).
A step-by-step approach to building this toolkit located in src is given in my book MLOps Lifecycle Toolkit (Apress, 2023) where many tools in this list were evaluated (something close to 100 open source tool). Feel free to contribute more tools, re-organize the list or add new categories. h diverse backgrounds to build a robust and versatile toolkit that can be used across industries.
To contribute, simply fork the repository, make your changes, and submit a pull request. Refer to the Contributing.md file for detailed information on how to contribute to this open-source project.
Let's collaborate to build a comprehensive and dynamic resource that empowers the data science community!
This repository accompanies MLOps Lifecycle Toolkit by Dayne Sorvisto (Apress, 2023).