@vivianmiranda has stumbled on a weird bug (https://github.com/CobayaSampler/cobaya/issues/23): PolyChord writes infinities as "[-]Infinity" (a Fortran convention, I guess), but Pandas only understands infinities as "[-]inf", so the new version of output.py (line 186) does not read the equal weights file correctly when there are infinities: it loads the full column as strings, and this produces some errors later in the code. This is known (https://github.com/pandas-dev/pandas/issues/10065).
I wonder if it would make sense to take some precaution there, e.g. running on each non-float column:
Hi Will,
@vivianmiranda has stumbled on a weird bug (https://github.com/CobayaSampler/cobaya/issues/23): PolyChord writes infinities as "[-]Infinity" (a Fortran convention, I guess), but Pandas only understands infinities as "[-]inf", so the new version of
output.py
(line 186) does not read the equal weights file correctly when there are infinities: it loads the full column as strings, and this produces some errors later in the code. This is known (https://github.com/pandas-dev/pandas/issues/10065).I wonder if it would make sense to take some precaution there, e.g. running on each non-float column:
Cheers