Problem (segmentation fault) with different fiducial for Plank fake likelihood

[mtagliazucchi]

Hi, I have a problem with fake Planck bluebook likelihood.

I want to use it but with a fiducial that reproduces Planck 2018 constraints on the 6 cosmological parameters.

To do that I deleted the fake_planck_bluebook_fiducial.dat file in data/, I written a parameter file fiducial.param with the following lines

data.experiments=['fake_planck_bluebook']

data.parameters['omega_b']      = [  2.2377,   None, None,      0.015, 0.01, 'cosmo']
data.parameters['omega_cdm']    = [ 0.12010,   None, None,     0.0013,    1, 'cosmo']
data.parameters['100*theta_s']  = [ 1.04110,   None, None,    0.00030,    1, 'cosmo']
data.parameters['ln10^{10}A_s'] = [  3.0447,   None, None,      0.015,    1, 'cosmo']
data.parameters['n_s']          = [  0.9659,   None, None,     0.0042,    1, 'cosmo']
data.parameters['tau_reio']     = [  0.0543,  0.004, None,      0.008,    1, 'cosmo']

#data.parameters['A_SZ']         = [0.96,    0, 2,   1, 1,  'nuisance']
data.parameters['P_shot']       = [0.0,       -1, -1, 0,1,  'nuisance']

and then I run python montepython/MontePython.py -p input/fiducial.param -c chains/test_fiducial -f 0. Let me comment the last two lines of the params file:

1. #data.parameters['A_SZ'] = [0.96, 0, 2, 1, 1, 'nuisance']: I originally put this line because in the original fake_planck_bluebook_fiducial.dat file there was in the first line the parameter A_SZ = 0.96. However, if I uncomment this line I get an error as this parameter is not recognized as a parameter of the likelihood.

2. data.parameters['P_shot'] = [0.0, -1, -1, 0,1, 'nuisance']: as before I add this line since in the in the originalfake_planck_bluebook_fiducial.datfile there was in the first line the parameterP_shot = 0`.

The generation of the new fiducial works fine. However, I then try to generate 8 MCMC chains by using the command

mpirun -np 4 python montepython/MontePython.py run -p input/example_new.param -o chains/baseline_test --conf default.conf -N 10000000 --update 50 --superupdate 20

where the file example_new.param contains the lines:

# Cosmological parameters list
data.parameters['omega_b']      = [2.249,  -1,-1, 0.016, 0.01,'cosmo']
data.parameters['omega_cdm']    = [0.1120, -1,-1, 0.0016,1,   'cosmo']
data.parameters['n_s']          = [0.963,  -1,-1, 0.004, 1,   'cosmo']
data.parameters['A_s']          = [2.42,   -1,-1, 0.038, 1e-9,'cosmo']
data.parameters['h']            = [0.703,  -1,-1, 0.0065,1,   'cosmo']
data.parameters['tau_reio']     = [0.085,  -1,-1, 0.0044,1,   'cosmo']

# Nuisance parameter list, same call, except the name does not have to be a class name
# For wmap:
#data.parameters['A_SZ']         = [1,    0, 2,   1, 1,  'nuisance']
# For SPT
#data.parameters['SPT_SZ']       = [5.6,  0,-1, 2.5, 1,  'nuisance']
#data.parameters['SPT_PS']       = [20.4,-1,-1, 2.7, 1, 'nuisance']
#data.parameters['SPT_CL']       = [5.3, -1,-1, 2.3, 1, 'nuisance']

# Derived parameter list
data.parameters['z_reio']       = [0,       -1, -1, 0,1,  'derived']
data.parameters['Omega_Lambda'] = [0,       -1, -1, 0,1,  'derived']

After few points (\approx 50 per chains) I get a segmentation fault error:

[node01:21032] *** Process received signal ***
[node01:21032] Signal: Segmentation fault (11)
[node01:21032] Signal code:  (128)
[node01:21032] Failing at address: (nil)

• other error lines.

There are no problems when using the original fiducial and input files. Can someone help me?

[dchooper]

Hi,

Sorry for the very late answer here! I'm replying in case it is still relevant. I'm surprised you had to add the P_shot parameter at all, as that doesn't appear in the likelihood data files. There is a default input file for running the fake_planck_bluebook likelihood: it's the default example.param. That should run through without any problems.

If you want to reproduce the Planck 2018 values in a simplified likelihood, we also have the fake_planck_realistic (as described in 1808.05955), which also shouldn't have any nuisance parameters and shouldn't trigger any segfaults. I think your issue might be caused by creating a fiducial file with a nuisance parameter that the likelihood doesn't expect, leading to a wrong fiducial file.

Note that we mainly use the fake likelihoods for doing forecasts. For a simplified Planck likelihood, we would recommend using Planck_lite instead (which then uses actual data instead of just a fiducial file).

Cheers, Deanna