Batch Normalization layer weights appear to be missing #3
Description
After training from scratch with the train_folder.py script provided by ALIGNN-TL, we noticed that the inference stage raises an error indicating that the Batch Normalization layer weights appear to be missing from the saved model.
Could you kindly advise what might typically cause this, and whether there are any recommended steps we could take to diagnose or resolve the issue? Thank you very much for your guidance.