Page 1 of 1
CosmoMC can''t obtain root.converge_stat
Posted: August 29 2017
by Tian Qiu
I run CosmoMC on a cluster.
Everything seems to work well except the converge file. I can't get the root.converge_stat when the others are fine such as *.txt, *data, *.log, *.inputparams, etc.
At the meanwhile, there is no feedback about the R-1 showed below, in which there are only such things.
Code: Select all
Chain:1 drag accpt: 0.4855783 fast/slow 72.54875 slow: 33186
Chain:13 drag accpt: 0.4834013 fast/slow 72.26221 slow: 33343
Chain:6 drag accpt: 0.4857524 fast/slow 72.30187 slow: 33418
Chain:11 drag accpt: 0.4792885 fast/slow 72.25761 slow: 33771
Chain:9 drag accpt: 0.4833719 fast/slow 72.40794 slow: 33378
Chain:10 drag accpt: 0.4827319 fast/slow 72.31480 slow: 33427
Chain:8 drag accpt: 0.4820674 fast/slow 72.17249 slow: 33452
Chain:3 drag accpt: 0.4805302 fast/slow 72.22508 slow: 33454
Chain:15 drag accpt: 0.4796885 fast/slow 72.20305 slow: 33578
Chain:0 drag accpt: 0.4796197 fast/slow 72.22173 slow: 33311
Chain:12 drag accpt: 0.4818384 fast/slow 72.20877 slow: 33577
Chain:4 drag accpt: 0.4814351 fast/slow 72.46579 slow: 33350
So I do not know when the program will shutdown, and how much my convergence is. I run CosmoMC for a long time but it never stops.
Anyone has the same problem? How can I fix this? Thanks a lot.
CosmoMC can''t obtain root.converge_stat
Posted: August 29 2017
by Ira *Wolfson
Please post the .ini fie you are using.
It is possible you are simulating over a degree of freedom the data do not constrain.
In such a case it will take an inordinate amount of time for the burn in to finish and CosmoMC to produce the .converge_stat file.
It might also be the case if that is indeed what you are doing that convergence will never be achieved....
As a rule for posting, I recommend specifying your machine specs, the compiler you used and the .ini file you use.
-- Ira
Re: CosmoMC can''''t obtain root.converge_stat
Posted: August 30 2017
by Tian Qiu
Ira *Wolfson wrote:Please post the .ini fie you are using.
It is possible you are simulating over a degree of freedom the data do not constrain.
In such a case it will take an inordinate amount of time for the burn in to finish and CosmoMC to produce the .converge_stat file.
It might also be the case if that is indeed what you are doing that convergence will never be achieved....
As a rule for posting, I recommend specifying your machine specs, the compiler you used and the .ini file you use.
-- Ira
Thank you for reply. I post the ini file below,
Code: Select all
#general settings
#Bicep-Keck-Planck, varying cosmological parameters
DEFAULT(batch2/BKPlanck.ini)
#Planck 2015, default just include native likelihoods (others require clik)
DEFAULT(batch2/plik_dx11dr2_DS_v18_TT.ini)
#DEFAULT(batch2/lowTEB.ini)
DEFAULT(batch2/lowl.ini)
#DEFAULT(batch2/lensing.ini)
#Other Likelihoods
#DEFAULT(batch2/BAO.ini)
#DEFAULT(batch2/WiggleZ_MPK.ini)
#DEFAULT(batch2/MPK.ini)
#DEFAULT(batch2/WL.ini)
#general settings
DEFAULT(batch2/common.ini)
#e.g. to vary r in addition to standard 6:
#(for r>0 also need compute_tensors=T)
#compute_tensors = T
#param[r] = 0.03 0 2 0.04 0.04
#high for new runs
MPI_Max_R_ProposeUpdate = 30
propose_matrix= planck_covmats/base_TT_lowTEB_plik.covmat
#Folder where files (chains, checkpoints, etc.) are stored
root_dir = tTT+lowP/
#Root name for files produced
file_root=Planck TT+lowP
#action= 0 runs chains, 1 importance samples, 2 minimizes
#use action=4 just to quickly test likelihoods
action = 0
#expected result for -(log like)
test_check_compare = 28.337
num_threads = 0
#if you want to get theory cl for test point
#test_output_root = output_cl_root
start_at_bestfit =F
feedback=1
use_fast_slow = T
checkpoint = T
#sampling_method=7 is a new fast-slow scheme good for Planck
sampling_method = 7
dragging_steps = 3
propose_scale = 2
#Set >0 to make data files for importance sampling
indep_sample=10
#these are just small speedups for testing
get_sigma8=T
#Uncomment this if you don't want one 0.06eV neutrino by default
#num_massive_neutrinos=
#to vary parameters set param[name]= center, min, max, start width, propose width
#param[mnu] = 0 0 0 0 0
#for PICO install from https://pypi.python.org/pypi/pypico and download data file
#cosmology_calculator=PICO
#pico_datafile = pico3_tailmonty_v34.dat
ico_verbose=F
The system is CentOS 6.6 and I compiled via ifort 17.0.3. Any advice? I am not familiar with these.
CosmoMC can''''t obtain root.converge_stat
Posted: August 31 2017
by Yutong Wang
I met the same problem, when I use the old version of ifort, there is no problems, and when I change the ifort 17.0.3,there are some problems,such as there is no .chk file,and can't obtain root.converge_stat
And by the way, what's your version number of MPI? I suspect the problem may be cause by the lastest MPI
Re: CosmoMC can''''''''t obtain root.converge_stat
Posted: August 31 2017
by Tian Qiu
Yutong Wang wrote:I met the same problem, when I use the old version of ifort, there is no problems, and when I change the ifort 17.0.3,there are some problems,such as there is no .chk file,and can't obtain root.converge_stat
And by the way, what's your version number of MPI? I suspect the problem may be cause by the lastest MPI
You reminded me that when I used the the old version 16.0.0, there was no problem. However after I ran it on a new cluster with ifort 17.0.3, it happened.
Now I use the Intel MPI version 2017 update 2. Before, I used the Open MPI 1.6.2, and there was no problem. Any idea?
CosmoMC can''t obtain root.converge_stat
Posted: August 31 2017
by Ira *Wolfson
I was having the same problems.
I was finally reduced to using the last version (November release),
with ifort compiler 16.3.
I don't know what's the deal with the code and ifort 17.xxx is...
Maybe Antony can clarify this...
Re: CosmoMC can''''t obtain root.converge_stat
Posted: September 01 2017
by Tian Qiu
Ira *Wolfson wrote:I was having the same problems.
I was finally reduced to using the last version (November release),
with ifort compiler 16.3.
I don't know what's the deal with the code and ifort 17.xxx is...
Maybe Antony can clarify this...
So it is the problem between the latest ifort version and CosmoMC. I will try the 16 one. Thank you.
Re: CosmoMC can''''t obtain root.converge_stat
Posted: September 04 2017
by Tian Qiu
Ira *Wolfson wrote:I was having the same problems.
I was finally reduced to using the last version (November release),
with ifort compiler 16.3.
I don't know what's the deal with the code and ifort 17.xxx is...
Maybe Antony can clarify this...
I can get the .chk now with ifort 16.0.4. However there was no .converge_stat still.
CosmoMC can''t obtain root.converge_stat
Posted: September 04 2017
by Ira *Wolfson
Have you seen an output saying that "burn-in" is done?
Re: CosmoMC can''''t obtain root.converge_stat
Posted: September 04 2017
by Tian Qiu
Ira *Wolfson wrote:Have you seen an output saying that "burn-in" is done?
Oh, it was obtained after a while. Maybe my machine works slowly. lol
Thank you.
Re: CosmoMC can't obtain root.converge_stat
Posted: September 04 2019
by Luke Hart
Has there been any development of this? Our cluster either has ifort14 or ifort17 and the 17 one has seemed to remove the converge_stat variable which is very bizarre? Anyone had any experience with this??
Luke