search

lucidrains / egnn-pytorch

Implementation of E(n)-Equivariant Graph Neural Networks, in Pytorch
MIT License
418 stars 69 forks source link

Training model with pyg graphs #43

Open mrjoness opened 8 months ago

mrjoness commented 8 months ago

Hello, I am using this model to train on protein data where each graph has a different number of atoms. So far I see good performance by padding shorter sequences, however I'd like to avoid this as I scale the model to much larger sequences. Is there currently any support for passing in batches of differently sized graphs (as in PyG)?

Obs01ete commented 5 months ago

Hello, I am using this model to train on protein data where each graph has a different number of atoms. So far I see good performance by padding shorter sequences, however I'd like to avoid this as I scale the model to much larger sequences. Is there currently any support for passing in batches of differently sized graphs (as in PyG)?

I suppose you'd want to go with EGNN_Sparse.