-
Notifications
You must be signed in to change notification settings - Fork 0
Known issues
Here we point out various parts of the code that are not necessarily wrong or bugged, but could perhaps be improved.
FCI and HF These folders include energies of the H2 molecule calculated at various distances and with different basis sets using the numerical methods FCI and HF. (The values can calculated using tequila and the exact details aren't relevant). These values are needed for the leaderboards, graphs and some of the statistics such as "variance from FCI". However when the application extends to handle other molecules this way of storing the energies is not scalable. Furthermore you can't store the energy at every possible distance.
Compare detail. Here we collect data for drawing the "Error vs gate depth" chart. (Reference "Error versus circuit depth"). We essentially need the distance at which the total energy of the molecule is the smallest. We have hardcoded this as 0.70 because it is pretty close to the truth while also being a value that is probably included in the results.
The second FIXME tag has to do with the problem above. The point of the "Error vs gate depth" chart is to compare the errors of different results at the same distance (specifically at the "bond distance"). However some results might not include an energy at this bond distance. We have made the compromise to just use the distance where the smallest energy is for each result.
The molecule field is more or less redundant at the moment. The contents of this field are also just simply an str dump of a molecule object. Other fields like this from Run include "variables", "gradients", "angles", "*_calls". The fields themselves are not necessarily useless, but the data they contain is not used for anything at the moment. The fields were simply created because initially we wanted to "collect as much data as possible". Due to them not being used anywhere their contents might be in a non-ideal format.
- Validation
- URL links
- Refactoring
- Tests
- Views.py (Done / Petri)
- CI testing happens two times (combine if possible)
In the Browsable UI the lists are not paged. The number of entries may grow quickly as it is easy to push every experiment automatically. The lists should be paged and some filtering should be available. (Petri)
Gate extraction We want to save the ansatz in such a format from which it can be compiled back to a tequila circuit. This essentially means that we have to extract the elementary quantum gates from the ansatz generated by tequila. The code works just fine, but the operation takes some time. Because we do this operation during every run of the VQE, we are slowing the overall process down if the user is for example calculating energies for 20 different distances.