Sorry that this is again a larger PR than I had anticipated.
The main goal is to implement the polarizable interaction neural network, reference: http://proceedings.mlr.press/v139/schutt21a.html .
During the implementation several additional issues arose and fixes are included in this PR.
This will be the last of its kind, I promise.
In models.py I have separated the Base class from the Lighing Class, which is now a Mixing.
In schnet.py I have added the SchnetLighning implementation that uses the Mixing Lightning class.
Other changes that are not directly related to the PaiNN implementation are:
more tests
added documentation with added shape of tensors where applicable
more helper functions in tests/helper_functions.py
Description
Sorry that this is again a larger PR than I had anticipated. The main goal is to implement the polarizable interaction neural network, reference: http://proceedings.mlr.press/v139/schutt21a.html .
During the implementation several additional issues arose and fixes are included in this PR. This will be the last of its kind, I promise.
In
models.py
I have separated the Base class from the Lighing Class, which is now a Mixing.In
schnet.py
I have added the SchnetLighning implementation that uses the Mixing Lightning class.Other changes that are not directly related to the PaiNN implementation are:
Todos
Status