scaling in pre_transform
#9648
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
{{ message }}
-
About scaling when using PyTorch Geometric Dataset
As far as I understood, the pre_transform callable for a torch_geometric dataset will apply to one Data object at the time. So if I would want to apply minmax scaler on some node features, it would only scale them based on the values in the current Data object and not across the entire dataset?
How would I do the fit then? Track all features in the Dataset's process method and when all raw objects have passed, call a fit which I then transform in the pre_transform?
Beta Was this translation helpful? Give feedback.
All reactions