How to deal with seq_len dimmension with GCNConv? #9441

[Percibus]

I've just starting using PyG, and for my problem I have batched data with a time dimmension. My tensor has shape [batch, seq_len, nodes, features] . I want to apply the GCNConv layer separatly to each seq_len step.

I read on this issue https://github.com/pyg-team/pytorch_geometric/discussions/4107

you need to permute your matrix to shape [batch_size, time_seq, num_nodes, num_feat]

But I'm not sure if I'm getting it correctly. Should I reshape to [batch * nodes, seq_len * features] before feeding it into the layer, or can I reshape to [batch*seq_len, keypoints, nodes] and pass this to the GCNConv?

Thanks in advance for the help!