pde-surrogate
PDESurrogates
PDE Datasets for FLARE
This repository contains the datasets used in the paper FLARE: Fast Low-rank Attention Routing Engine.
Paper: arXiv:2508.12594
Code: GitHub - vpuri3/FLARE.py
Dataset Description
The collection includes:
PDE Benchmarks: Standard datasets used to evaluate neural PDE surrogates, such as Elasticity and Darcy flow.
Additive Manufacturing (AM) Benchmark: A new 3D field prediction benchmark derived from numerical simulations of the Laser Powder Bed… See the full description on the dataset page: https://huggingface.co/datasets/vedantpuri/PDESurrogates.ai4sci-pde-surrogate-data
PDE surrogate: the benchmark split
Companion data for the pde-surrogate task in
T0-RSI/ai4sci-tasks. It carries only the
files this benchmark's split creates, not a mirror of the upstream datasets.
subtree
what
train/
the nine corpus files that carry training and held-out trajectories, truncated at the end of the validation range, with norm_* recomputed over the kept trajectories
eval/
the held-out tail of each of those files, under its original trajectory numbering… See the full description on the dataset page: https://huggingface.co/datasets/deqing/ai4sci-pde-surrogate-data.