Laurent Lemmens
Laurent Lemmens
@dariatols Could you add the Jupyter Notebook to the website's examples section?
This PR seems a little stale. Are there things that still need to happen?
Following this approach, we keep the benefits of using spherical GTO shells: using less basis functions for higher angular momenta.
We need `Shell` to accommodate both Cartesian and spherical shells, but we only have to convert a spherical shell to Cartesian GTOs. HORTON has the transformation formulas for spherical GTOs...
Indeed, that kind of initialization for the environment would be the direction I would propose to go in.
I’m thinking of adding a direct API to ˋTransformationˋ that produces a random transformation that relates an orthonormal basis with a non-orthogonal one. This can serve as a “random” initial...
The transformations that those APIs produce do not link a non-orthogonal (AO) basis to an orthonormal (MO) one, so we'll have to implement a new API.
@xdvriend The gradients and Hessians I'm referring to in this issue are the *general* gradients and Hessians (related to the Fockians), not only for GHF.
In #583, I've added some example notebooks. We can add this to our CI by using ``` jupyter nbconvert --to notebook --execute mynotebook.ipynb ``` as found in the [jupyter documentation](https://nbconvert.readthedocs.io/en/latest/execute_api.html).
After #590, it seems to be important that we use the `gqcp_dev` conda environment (in the `environment.yml`-file) in conjunction with this CI check.