HF Benchmark Database

[PaulWAyers]

@OrtliebJessica has a benchmark-quality Hartree-Fock dataset (with Gaussians) that we should add to AtomDB.

[PaulWAyers]

@msricher do you know how we can add this? I think this should replace the existing Gaussian HF dataset. (It is more complete and more accurate.)

It is the same protocol as we will use for benchmark-accuracy shci.

[msricher]

https://github.com/theochem/AtomDB/tree/master/atomdb/datasets/gaussian

This is the current Gaussian dataset folder. It contains a Python file run.py in which you fill in the instructions to compile database files. Any raw data needed for the compilation should be put in atomdb/data. Then you can either use the Python interface atomdb.compile_species, or the CLI, to compile the species you want.

[msricher]

Compiled datasets can also be added to this repo, so that other users do not need to run the compile step themselves.

[OrtliebJessica]

Specifically what I have is a set of chk files prepared using PySCF (rather than fchk files prepared using Gaussian). Do you know if the code will work with PySCF chk files?

[msricher]

I am not sure. If the code doesn't run with these chk files swapped in for the old ones, then the run.py file is the one you would modify to make it work. The required modifications should be small, if any.