PLAID: Physics-Learning AI Datamodel

PLAID (Physics-Learning AI Datamodel) is a Python library and data format for representing, storing, and sharing physics simulation datasets for machine learning.Unlike domain-specific formats, PLAID accommodates time-dependent, multi-resolution simulations and heterogeneous meshes.The library provides a high-level API to easily load, inspect, and save data.Beyond basic I/O, PLAID adopts a backend-oriented storage design supporting several backends (cgns, hf_datasets, and zarr), and integrates with the Hugging Face Hub to push and pull datasets.It also ships command-line tools to check, serve, and interactively visualize datasets.In short, PLAID couples a flexible on-disk standard with a software toolkit to manipulate physics data, addressing the needs of ML researchers in fluid dynamics, structural mechanics, and related fields in a generic fashion.Full documentation, examples and tutorials are available at plaid-lib.readthedocs.io.