Welcome to itwinai’s documentation!
itwinai is a versatile toolkit designed to accelerate AI and machine learning (ML) workflows for researchers and scientists,
particularly in the realm of Digital Twins (DTs). This toolkit provides a suite of user-friendly tools to effortlessly
scale machine learning projects to high-performance computing (HPC) resources, seamlessly integrating with cloud-based services.
The primary focus of itwinai is to reduce the engineering burden on researchers, enabling them to concentrate more on advancing
their science.
Empowering AI in Scientific Digital Twins
The itwinai toolkit is engineered to support AI-driven research in scientific digital twins. It offers powerful capabilities for
distributed machine learning training and inference on HPC systems, efficient hyper-parameter optimization (HPO), and simplified
ML logging with integration to popular tools like MLflow, Weights & Biases, and TensorBoard. Additionally, it includes an
intuitive framework to define, configure, and manage modular and reusable ML workflows, providing a streamlined approach to
experiment management.
Moreover, the toolkit is designed with extensibility in mind, allowing third-party developers to build and integrate their own plugins, enhancing the flexibility and adaptability of the platform.
itwinai is primarily developed by Matteo Bunino from CERN, in
collaboration with the interTwin
project, which aims to advance the use of digital twins in scientific research.
How to Read the Docs
To effectively utilize the itwinai toolkit documentation, begin by exploring the “Getting Started” section. This part is essential
for grasping the basics and setting up the toolkit, with detailed instructions for different installation scenarios, whether on HPC
systems or your local machine.
For a deeper dive into the core functionalities, check out the “How It Works” section, which breaks down the key concepts that power
itwinai. The “Scientific Use Cases” section offers practical examples and scenarios from the interTwin
project, showcasing how the toolkit is applied in real-world research.
Enhance your skills by exploring the “Tutorials” section, filled with step-by-step guides on distributed ML training and workflow creation. Lastly, the “Python API Reference” is your go-to resource for a detailed overview of the toolkit’s capabilities, helping you implement specific features in your projects.
Following these sections systematically will help you maximize your understanding and make the most of the itwinai toolkit in
your research endeavors.
itwinai documentation is also available in different versions: ‘latest’, ‘stable’, and specific release versions like ‘v0.2.1’.
The ‘latest’ version reflects the most recent updates, while the ‘stable’ version is recommended for production use, as it
contains thoroughly tested features aligned with the toolkit’s most recent release
(learn more).