search

baudren / montepython_public

Public repository for the Monte Python Code
MIT License
65 stars 115 forks source link

Block inversion failed for l=2 info=1 #75

Open VicenteAtal opened 7 years ago

VicenteAtal commented 7 years ago

Hi all, I hope somebody can help me with this problem. When running Planck's base model

python montepython/MontePython.py run -o planck/ -p base2015.param -c covmat/base2015.covmat -b bestfit/base2015.bestfit --conf default.conf -f 1.5 -N 20

I get some error. After initializing the best fit parameters I see block inversion failed .... from best-fit file : gal545_A_217 = 78.8981 from best-fit file : calib_100T = 998.163 from best-fit file : calib_217T = 994.844 from best-fit file : A_planck = 100.222 block inversion failed for l = 2 info = 1

which then repeats

-LogLkl 1e+02omega_b omega_cdm 100*theta_s ln10^{10}A_s n_s tau_reio A_cib_217 xi_sz_cib A_sz ps_A_100_100 ps_A_143_143 ps_A_143_217 ps_A_217_217 ksz_norm gal545_A_100 gal545_A_143 gal545_A_143_217gal545_A_217 1e+03calib_100T 1e+03calib_217T 1e+02A_planck z_reio Omega_Lambda YHe H0 1e+09A_s sigma8
block inversion failed for l = 2 info = 1 block inversion failed for l = 2 info = 1 block inversion failed for l = 2 info = 1 block inversion failed for l = 2 info = 1 block inversion failed for l = 2 info = 1 block inversion failed for l = 2 info = 1 block inversion failed for l = 2 info = 1 block inversion failed for l = 2 info = 1 block inversion failed for l = 2 info = 1 block inversion failed for l = 2 info = 1 block inversion failed for l = 2 info = 1 block inversion failed for l = 2 info = 1 block inversion failed for l = 2 info = 1 20 nan 2.234564e+00 1.176998e-01 1.041954e+00 3.078268e+00 9.666574e-01 7.466321e-02 6.443051e+01 4.530022e-01 8.697788e+00 2.232064e+02 2.593456e+01 3.300564e+01 8.734209e+01 9.913474e+00 6.172188e+00 8.719394e+00 1.738178e+01 7.872689e+01 9.981529e+02 9.948696e+02 1.000648e+02 9.624161e+00 6.976004e-01 2.478293e-01 6.821757e+01 2.172076e+00 nan

Anybody knows the reason for this problem ? thanks in advance

bhorowitz commented 7 years ago

I think it is a problem originating from the Planck likelihood code, probably due to some strangeness in the power-spectra you are passing to it. I don't think it should cause the code to fail however, and probably doesn't affect the end results.