SchNOrb?

[karsar]

It is mentioned that SchNOrb is implemented. Where in the code can I see it? Or better yet, how do I load the SchNOrb model checkpoint? Why is there no PhisNet YAML file in the model config directory to run it with run.py? Are there reasons for that? Running the PhisNet example for the test, as described in the PhisNet folder, produces an error:

from .hamiltonian_dataset import *
ModuleNotFoundError: No module named 'training.hamiltonian_dataset'

I am trying to replicate the comparison table for Hamiltonian prediction using model checkpoints.

[KuzmaKhrabrov]

Hello! We are preparing a major update, which will be released in the beginning of June. In the meanwhile, please, use branch 1.0. SchNorb code is located in this repo: https://github.com/KuzmaKhrabrov/SchNOrb
Unfortunately we don’t plan to add PhiSNet or SchNOrb into lightning pipeline soon

[karsar]

It seems that the major update does not contain SchNorb code. In fact, running https://github.com/KuzmaKhrabrov/SchNOrb has the same problem as the original SchNorb. While trying to import schnorb an error occurs:

from schnetpack import Properties
ImportError: cannot import name 'Properties' from 'schnetpack'

so scripts do not work with schnetpack version installed with nablaDFT...

Let's say I want to load provided checkpoint for the pretrained SchNOrb and run a test on some of the provided test sets to validate nablaDFT numbers, how would I do that? I can run PaINN and check the claims, as an example.

[KuzmaKhrabrov]

Hello!

Let's say I want to load provided checkpoint for the pretrained SchNOrb and run a test on some of the provided test sets to validate nablaDFT numbers, how would I do that?

Unfortunately, to do that you need to make a separate environment and install SchNOrb from the repo: https://github.com/KuzmaKhrabrov/SchNOrb like

pip install .

I will update the Readme file and show an example.