Closed shuishenshui closed 4 months ago
update: for {g_f, T_ncdm, m_ncdm} chains, this error also happens when calculating covmat.
Hi,
MontePython does not tolerate added comments at the end of input lines, i.e. this will cause problems (the comment needs to be on a separate line):
data.parameters['deg_ncdm'] = [1, 0, None, 0.1, 1, 'cosmo'] # calculated inside data.py, do not occur in log.param
I would not be very confident that the results are correct either, but maybe you can do a shorter run fixing this problem and checking if they seem to be converging to the same result.
Best, Thejs
Also, once this issue is fixed, I suspect you will have problems having both deg_ncdm and Neff_ncdm as varying parameters, should you not only have Neff_ncdm, since it defines deg_ncdm? It seems to me you're effectively sampling the same parameter twice.
Best, Thejs
Hi Thejs,
Sorry for the late response because I ran the chains to verify the problem is solved. I removed the comment in .param and the error's gone and it convergences well now. Thank you so much!
Yes, you are right, the Neff_ncdm and deg_ncdm are degenerated, thus I set Neff_ncdm as input and calculate deg_ncdm in the data.py, which is then put into CLASS as the parameters.
Cheers, Yunzhi
Great! Thanks for reporting back.
Best, Thejs
Hi all,
I am running an extended neutrino model with free T_ncdm and degree of degeneracy g_f. Sampling {T_ncdm,g_f} are difficult to converge since they are very degenerated, thus I sample {N_eff, T_ncdm} instead, and calculate g_f inside data.py.
The chains runs well and I can plot contours with getdist. The acc rate is 0.28, Total number of accepted steps: 339255. I ran 10 chains, but some chains stop much earlier than N.
However, it always give this error when I try to run info --want-covmat:
This is my .param:
Modified data.py:
Thank you in advance!