jNO: A JAX Library for Neural Operator and Foundation Model Training

TL;DR

jNO is a JAX-based library that simplifies training neural operators and foundation models, supporting diverse training paradigms and workflows in a unified, flexible framework.

Contribution

It introduces a tracing system that unifies data-driven and physics-informed training, enabling seamless switching between different PDE training methods.

Findings

Supports multi-model compositions and hyperparameter tuning.

Enables mesh-aware residual evaluation and PDE-constrained training.

Provides a unified optimization pipeline for various training modes.

Abstract

jNO (jax Neural Operators) is a JAX-native library for neural operators and foundation models with unified support for both data-driven and physics-informed training. Its core design is a tracing system in which domains, model calls, residuals, supervised losses, and diagnostics are written in one symbolic language and compiled into one optimization pipeline. This allows users to move between operator regression, mesh-aware residual evaluation, and PDE-constrained training without restructuring the surrounding code. jNO also supports multi-model compositions, fine-grained control at parameter level (model, optimizer, and learning rate), hyperparameter tuning, and JAX-native workflows for translated PDE foundation-model families. The source repository is available at https://github.com/FhG-IISB/jNO.