DQC self-consistency check

[KarimAED]

We need to verify whether DQC can learn energies produced by itself.

[KarimAED]

Tasks:

• Build a script to generate DQC energies and densities for Mols
• Build a script to assemble training data from energies and densities
• Start with training from a single molecule
• Gradually increase the number of molecules in training vs testing
• Implement a more robust inference script to test progress

[KarimAED]

Training on a single molecule running with little progress across 70 epochs. possibly because of bad hyperparametrization.

[KarimAED]

Success of training has been fixed for different quantities individually. Weighting and learning rates may have to be adjusted to learn on multiple quantities successfully.