Question regarding code implementation

[orgw]

From the YAML config, your two key dimensions are:

1.hidden_dim = 96
tf_hidden_dim = 384
However, in your code you have a layer normalization layer expecting a different dimension (e.g. 128) or calling LayerNorm(384) before the tensor has dimension 384.

2.in your default.yaml you don't have tf_hidden_dim which causes error.

3. running main.py on bace, esol, lipo, i cannot reproduce your results in the paper (final test metric)
   for bace, i get better results of 0.33
   for esol and lipo, i get results of 0.715 0.668

maybe the configs are not updated?