Dataset and process scripts for "A Graph Neural Network for the Era of Large Atomistic Models"

This directory contains:

1. data: 
    1.1 Force-Field-Task-Datasets: 12 force field prediction datasets for zero-shot test, see `data/Force-Field-Task-Datasets/README` for details. 
    1.2 Single-Task-Datasets: datasets for single task benchmarks that require additional sampling and splitting preprocessing, see `data/Single-Task-Datasets/README` for details.

2. model: DPA-3.1-3M.pt, the multi-task pretrained large atomistic model, see https://www.aissquare.com/models/detail?pageType=models&name=DPA-3.1-3M&id=343 for details.

3. script: `zero-shot-test/zero-shot-test.py`: script to replicate the zero-shot results of DPA-3.1-3M on force field prediction datasets.