In the class GENConv, should the output dimension of edge_encoder be in_dim instead of emb_dim?

lightaime / deep_gcns_torch

Pytorch Repo for DeepGCNs (ICCV'2019 Oral, TPAMI'2021), DeeperGCN (arXiv'2020) and GNN1000(ICML'2021): https://www.deepgcns.org

MIT License

1.13k stars 155 forks source link

In the class GENConv, should the output dimension of edge_encoder be in_dim instead of emb_dim? #59

Closed youngleec closed 3 years ago

youngleec commented 3 years ago

Hi, your work is great and interests me a lot.

In the class GENConv, should the output dimension of edge_encoder be in_dim instead of emb_dim, especially when the in_dim doesn't equal to the emb_dim? Thanks.

https://github.com/lightaime/deep_gcns_torch/blob/03d553d1a75e923586e0e67b9fdf69dfd31e481b/gcn_lib/sparse/torch_vertex.py#L57

lightaime commented 3 years ago

Hi @youngleec. You are right. It should be in_dim instead of emb_dim for the case when in_dim != emb_dim. I noticed this and fixed it when I created a pull request for PyG (https://github.com/rusty1s/pytorch_geometric/pull/1403/commits/a173c0794493e2842a2379efe09fea9b6d649cc6). But I forgot to change it here. Thanks for your reminder.