It seamlessly integrates with HPC resources, making workflows highly scalable and promoting code reuse. With built-in tools for hyper-parameter optimization, distributed machine learning, and pre-trained ML models. itwinai empowers AI researchers. It also integrates smoothly with Jupyter-like GUIs, enhancing accessibility and usability.
Different interfaces, to lower the entry barrier for users coming from different fields of expertise: from lower-level python programming to high-level GUI workflow representation. itwinai provides out-of-the-box SOTA AI tools and encourages code reuse, to further simplify and streamline the development of ML workflows, on top of seamless integration with HPC resources.
Release Notes
The itwinai library has been adopted in interTwin for the definition of Advanced AI workflows. Up-to-date release notes can be found in the GitHub repository.
Future Plans
Future extensions to itwinai in terms of features and functionalities will be based on adoption requirements identified through integration efforts with new use-cases. Furthermore, AI-centric pipelines defined with tools such as Kubeflow, Airflow and others will be continuously adopted in the course of the project. Support for additional distributed learning backends is also envisioned in the future. Also, integration with the interlink package from WP5, workflow composition tool from Task 6.1 and other relevant tasks in WP6 will be performed.