Matthias Fey

I definitely think so. We definitely want to make some further improvements to our heterogeneous GNN pipeline. Any help is highly appreciated, so let me know if you have any...

Thank you! This looks great. I'm happy that it unblocks you for now. The main question is how we want to integrate that into the library. IMO, `HeteroConv` should expose...

Yeah, you are right.`HeteroConv.jittable()` would need to make each sub-GNN jittable, as well as converting edge types to single strings. Let me know if you have interest in contributing this!

It might well be that this issue is already resolved (at least `HeteroConv` works fine with `torch.compile` now).

Currently, `SAGEConv` does not support weighted graphs, but `GraphConv` does (which is quite similar). Note that you need to pas both `edge_index` and `edge_weight` to the GNN op.

Do you mean using the `SparseTensor` class? That is possible by passing edge weights to the `value` argument: ```python adj_t = SparseTensor(row=row, col=col, value=edge_weight, sparse_sizes=(N, N)) ```

`NeighborSampler` returns a `e_id` tensor, which can be used to index the original edge weights: ```python loader = NeighborSampler(edge_index, ...) for batch_size, n_id, adjs: for edge_index, e_id, size in adjs:...

Can you give me a short example to illustrate this issue?

You need to index select the edge weights coming from `data.edge_weight`. The correct example would look similar to: ```python def forward(self, x, adjs, edge_weight): for i, (edge_index, e_id, size) in...

`GraphConv` is the same as `SAGEConv` in case you specify `aggr="mean"`. Let me know if that works for you.