How to get the original node index from the sampled graph from NeighborLoader?

[GYHHAHA]

Does PyG's NeighborLoader contains property or information for the original sampled nodes index for each batch?

[rusty1s]

You can maintain them by adding them to the original data object:

data.n_id = torch.arange(data.num_nodes)
loader = NeighborLoader(data, ...)
batch = next(iter(loader))
print(batch.n_id)

[GYHHAHA]

Thanks.