MarkoMile/gnn-physics-simulator

A PyTorch Geometric implementation of Learning to Simulate Complex Physics with Graph Networks (Sanchez-Gonzalez et al., 2020). Trains a GNN to predict particle accelerations from local interactions, reproducing DeepMind's architecture for simulating fluids, with additional fine-tuning on a custom WCSPH dataset.

语言：Jupyter Notebook · GitHub 星标：15 · Fork：0

话题：ai · fluid-dynamics · gnn · graph-neural-networks · machine-learning · physics

在 GitHub 上查看原仓库