How to identify mutually disconnected subgraphs?

[NiklasKappel]

Please consider the assumptions:

1. I have a graph that consists of multiple mutually disconnected subgraphs (think of e.g. multiple molecules in a box, where edges are chemical bonds).
2. I do not know in advance how many subgraphs there are.
3. I do not know in advance which subgraph each node belongs to.
4. I would like to process each subgraph individually, ideally in the same way and in parallel.
5. I do not think that the processing I want to do can be expressed as a graph convolution.

Does PyG support such a use case? E.g., is there a way to assign each node an index that identifies the subgraph it belongs to, similar to how a PyG DataLoader returns batches with a batch attribute that assigns each node to its graph in the batch?

[NiklasKappel]

FWIW i found connected_components functions from cugraph and scipy, but I am not aware how using them would impact the ability to do backpropagation with PyTorch.