Question: How to access the computed descriptor

[HTranUIUC]

Hi,

This might be a silly question but I have a lot of structures that take time to build the descriptor. Once it is built, is there a way for me to:

1. Access them/ View the files?
2. Use the built descriptor as input to PyXtal NN directly?

[edit], additional question,
I trained with NN and plan to use the trained results for LAMMPS. My descriptor is ACSF. Upon finished training, I can only find the '.pth' file. I tried to use PyTorch to load this file but have some errors.

1. How can I view the weight and bias from the generated '.pth' file?
2. LAMMPS documentation MLIAP seems to suggest that I can only use the SO3 descriptor. Therefore, I need to convert the .pth file into KIM format if I need to use ACSF. Have you all tried that before? If not, LAMMPS documentations seem to suggest that defining a descriptor is straightforward, can you point me in a direction where I can learn how to do so?

I apologize if this information is in the documentation. I cannot seem to find them.
Thank you for your help.