elements of A_pred are all zeros

[krishia]

Thank you for great paper. I'm trying to reproduce the result of node attributes(point locations) reconstruction experiment with NDP, dataset=Grid2d. The result of reconstructed node feature(X_pred), was really good. But the reconstructed adjacency matrix (A_pred) was all zeros. There was no code modification except the

from spektral.layers import ops ... A = ops.sp_matrix_to_sp_tensor(A.astype("f4"))

to

from spektral.layers import ops from spektral.utils import sparse ... A = sparse.sp_matrix_to_sp_tensor(A.astype("f4"))

due to "Move sp_matrix_to_sp_tensor and sp_batch_to_sp_tensor to spektral.uti…" in Spektral library. If there is something that i missed, please let me know.

I'll be glad to see your answer. Thanks.

[danielegrattarola]

Hey,

what script are you running exactly?

[krishia]

I runned src/autoencoder/run_ndp.py