NaN energy written to xyz after failed CP2K evaluation

[armaet]

Hi

When a CP2K evaluation fails (for example convergence criteria are not met), psiflow saves the Geometry-object with a nan energy. The comment line in the extended xyz file looks something like:

Properties=species:S:1:pos:R:3 energy=nan

However if a Geometry-object is constructed from this string it will have str(nan) as energy-argument which is not supported (energy is typechecked to be a float specifically). This is quite strange because I can not find where the energy is set to nan, especially because the nan_if_unsuccesful-function should guarantee that a failed evaluation produces a Geometry with None energy. So, any idea where this come from?

Thanks.