tum-pbs/apebench

[Neurips 2024] A benchmark suite for autoregressive neural emulation of PDEs. (≥46 PDEs in 1D, 2D, 3D; Differentiable Physics; Unrolled Training; Rollout Metrics)

35
/ 100
Emerging

This tool helps researchers and engineers who work with complex physical simulations evaluate how well deep learning models can predict the behavior of systems governed by Partial Differential Equations (PDEs). You input various PDE scenarios and neural network architectures, and it outputs performance metrics and visualizations of the neural network's predictions over time, comparing them to a high-fidelity simulator. This is for computational scientists, machine learning researchers, or simulation engineers who develop or assess data-driven models for physical systems.

Use this if you need to systematically benchmark and compare different autoregressive neural network models for emulating a wide range of PDE-driven physical phenomena.

Not ideal if you are looking for a general-purpose PDE solver or a tool for real-time deployment of pre-trained models.

computational-physics scientific-machine-learning numerical-simulations neural-networks predictive-modeling
No Package No Dependents
Maintenance 6 / 25
Adoption 9 / 25
Maturity 16 / 25
Community 4 / 25

How are scores calculated?

Stars

99

Forks

2

Language

Python

License

MIT

Last pushed

Nov 11, 2025

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/ml-frameworks/tum-pbs/apebench"

Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.