Question on inputs to feature decoder

[claudian37]

My understanding is that the inputs to the feature decoder are the outputs from the encoder, which are defined as latent variables zu and zv. However, this shows that the inputs to the feature decoder are xu and xv. Is my understanding incorrect or is this a typo?

GraphBEAN/models/net.py

Line 136 in 82fe2e4

(xu, xv), xe = conv((xu, xv), adj, xe=xe)

[rizalzaf]

Hi @claudian37
at the first layer of the decoder, zu = xu, and zv = xv. So the code should work fine.
Thanks.