Dear authors,
I'm a bit new to deep-learning and transformer chemist.
I've read the GraphiT paper and followed the code in this repository.
What I want is to use diffusion kernel with both symmetric and multi-dimensional edge attribute.
But it cause problem when the positional encoding.
Below is example of my data shape with batch
DataBatch(x=[267967, 40], edge_index=[2, 932790], edge_attr=[932790, 2], y=[50], pos=[267967, 3], batch=[267967], ptr=[51])
How can I solve this problem?
Dear authors, I'm a bit new to deep-learning and transformer chemist. I've read the GraphiT paper and followed the code in this repository. What I want is to use diffusion kernel with both symmetric and multi-dimensional edge attribute. But it cause problem when the positional encoding. Below is example of my data shape with batch
DataBatch(x=[267967, 40], edge_index=[2, 932790], edge_attr=[932790, 2], y=[50], pos=[267967, 3], batch=[267967], ptr=[51])
How can I solve this problem?Best regards :) Thank you