Is batch training supported? #325
Description
I came across this remarkable repository and discovered its excellence.
I am interested in developing a shared model across various simplicial complexes (which may have different topologies). Ideally, it would be advantageous if these complexes could be combined to create a mini-batch for training, potentially boosting speed.
I noticed that all tutorials provided in this codebase have a batch size of 1, and the gradients are not even accumulated, except for this one where it uses a batch size of 32; but they are homogeneous complexes.
Are there any tools in the repository to facilitate the construction of a mini-batch from multiple heterogeneous complexes?