Running these notebooks requires torch, torch_geometric, and e3nn -- installation instructions linked. You can also find additional resources on e3nn here.
create_helix_dataset.ipynb: Notebook for generating helix dataset. You can change the level of noise and various parameterizations of the geometry.helix_predict_displacements_and_frames.ipynb: predict displacements and coordiante frames given initial structure and frames.generate_from_backbone.ipynb: predict the vertices of tetrahedra from their centers as a projection onto spherical harmonics centered on centers.
datasets/
data.torch: pre-computed dataset forhelix_predict_displacements_and_frames.ipynb4_tetra_50_frames.npz: pre-computed dataset forgenerate_from_backbone.ipynb. vertices and centers of 4 edge-sharing tetrahedra.
models/
helix_model_3_layer.npz: torch model parameters forhelix_predict_displacements_and_frames.ipynbto start exploring the notebook.tetra_model_4_layer.npz: torch model parameters forgenerate_from_backbone.ipynbto start exploring the notebook.