Shyue Ping Ong
Shyue Ping Ong
@computron RealVASP is no easier to write than FakeVasp. And I would say running RealVASP violates a critical principle of testing, which is that the developer is responsible for code...
Let me perhaps sketch out why I think FakeVASP is perfectly doable and readily substitutable with RealVASP. Let's start by defining a helper testing class: ```python class AtomateWFTest(unittest.TestCase): VASP_OUTPUT_DIR =...
@mkhorton Just a note that I am completely on board for REAL vasp tests. Given the use of atomate as the backbone for MP, it is, of course, critical that...
@bocklund My final note on this subject - the whole "we hate fake VASP tests" is premised on the false idea that somehow real VASP tests are "easier" for the...
I would say that where atomate overlaps pymatgen, e.g., specifying parameters for MP-style calculations, these should be in pymatgen and atomate should avoid duplication. For other cases, there is no...
I will just have one final comment - I am confused what @bocklund means by saying that it is not easy to make a change to PMG input sets. The...
@bocklund This is a fundamental philosophical difference. The raison d'etre for having an input set is to **standardize**. By definition, that means modifications should be rare and exceptional. So I...
The None to unset incar setting has been implemented. https://github.com/materialsproject/pymatgen/commit/111f44ffc450e2db8c30a923dfb664c309c76194
@bocklund I would modify a few things with regards to principles: 1. All atomate Fireworks have a 1:1 correspondence to a PMG input set, unless it is for a type...
@bocklund I have a quick question. It is clear that ISMEAR = -5 is good for accurate energies in metals (no relaxation). What about forces (no relaxation)? Is ISMEAR=1 still...