Please first contact me via: willdavie2002@gmail.com. There are many areas for improvement, some possible routes are outlined here.
Main focus on perfecting support for LAMMPS and VASP, but aim to add support for other codes too.
| job | priority | description |
|---|---|---|
| vary LAMMPS parameters across sample space | High | need flexability on defining a sample space, e.g. defining a different random seed for velocities |
| Native VASP directory construction | Med | Using information in the MD base directory define a simple vasp calculation setup. |
| Native LAMMPS directory construction | Low | Given some basic information, construct a lammps directory |
| job | priority | description |
|---|---|---|
| Add additional descriptors to access similarity of configurations | Med | E.g. SOAP, High priority if MLIPTS is going to be used for non-periodic solids. |
| Optimize intensive calculations | Low | Unless analysing a huge dataset, this is not a bottleneck in the workflow |
| job | priority | description |
|---|---|---|
| Add further support for custom hpcs | Med |
| job | priority | description |
|---|---|---|
| Catch errors from QM calculations before appending | High | E.g. if calculations fail to converge. |
| Add options to save additional data from QM calculation | Med | e.g. magnetic and charge states, (+ any useful output parameter of choice with a keyword.) |
Support for all changes in the main class.
Development of the active learning workflow. Native support for model training.