Matthias Fey

icon indicating a website link https://rusty1s.github.io icon indicating an email [email protected]

icon company Kumo.AI icon location Dortmund, Germany icon location Creator of PyG (PyTorch Geometric) - Founding Engineer @ Kumo.AI - PhD @ TU Dortmund University - Interested in Graph Representation Learning

Results 1042 comments of Matthias Fey
trafficstars

GPSE Implementation

Super, let me resolve the linting issue.

TypeError: BasicGNN.forward() got an unexpected keyword argument 'return_attention_weights'

Are you referring to `models.GAT`? If you need to return attention weights, I suggest to use the GNN layers and build your own model on top.

[Roadmap] Heterogeneous Graphs Explainability Support

@wsad1 FYI

[Roadmap] Heterogeneous Graphs Explainability Support

I think the most important feature is to support `GNNExplainer` for heterogeneous graphs. Hopefully it shouldn't be hard to do since we just need to make sure that `edge_mask` and...

[Roadmap] Heterogeneous Graphs Explainability Support

Yeah, we still don't have heterogeneous GNN support for `GNNExplainer`.

PyG Dependencies (torch-scatter, torch-sparse, etc.) are incompatible with conda-forge distribution of PyTorch

AFAIK, there exists `conda-forge` packages as well for these dependencies, but they are not maintained by us, see https://github.com/conda-forge/torch-scatter-feedstock.

Installation with Poetry on Mac M1 ship

Yes, once PyTorch 2.4 is released.

Installation with Poetry on Mac M1 ship

Yes. PyG itself is compatible with PyTorch 2.4. You can uninstall PyG extensions if you run into issues or install them from source if you need them.

Installation with Poetry on Mac M1 ship

Is this related to PyTorch 2.4? We haven't tested this one yet.

Sparse version of `nn.dense.dense_mincut_pool`

We initially designed it this way because `mincut_pool` needs to operate on dense adjacency matrices anyway in consecutive applications. I am definitely open though to support it for both sparse...