Walrus: A Cross-Domain Foundation Model for Continuum Dynamics

TL;DR

Walrus is a transformer-based foundation model designed for continuum dynamics, incorporating novel stabilization and training techniques to handle diverse physical scenarios and improve long-term prediction accuracy.

Contribution

The paper introduces Walrus, a new foundation model for continuum dynamics, with innovative stabilization, distributed training, and tokenization methods, enabling effective learning across diverse physical domains.

Findings

Outperforms prior models on short and long-term predictions

Enhances forecast stability and training throughput

Demonstrates transfer learning across multiple physical scenarios

Abstract

Foundation models have transformed machine learning for language and vision, but achieving comparable impact in physical simulation remains a challenge. Data heterogeneity and unstable long-term dynamics inhibit learning from sufficiently diverse dynamics, while varying resolutions and dimensionalities challenge efficient training on modern hardware. Through empirical and theoretical analysis, we incorporate new approaches to mitigate these obstacles, including a harmonic-analysis-based stabilization method, load-balanced distributed 2D and 3D training strategies, and compute-adaptive tokenization. Using these tools, we develop Walrus, a transformer-based foundation model developed primarily for fluid-like continuum dynamics. Walrus is pretrained on nineteen diverse scenarios spanning astrophysics, geoscience, rheology, plasma physics, acoustics, and classical fluids. Experiments show that Walrus outperforms prior foundation models on both short and long term prediction horizons on downstream tasks and across the breadth of pretraining data, while ablation studies confirm the value of our contributions to forecast stability, training throughput, and transfer performance over conventional approaches. Code and weights are released for community use.