I am running my MontePython likelihood with the option -m NS to use MultiNest sampling (and obviously specifying the other MultiNest sampling options such as th enumber of live points etc..)
I have been doing this many times using 1 node, multiple cores on that node, and everything went fine. Unfortunately my likelihood is very slow, therefore I was hoping to gain some speed by running on MPI.
If I run the same likelihood with MPI, in my output file I get a lot of
/!\ invalid value encountered in greater
/!\ invalid value encountered in subtract
and in the "NS" subfolder created in the chain output folder only two files get created, .arguments and .paramnames. I think this is not quite normal and I was wondering if you could help me identify the problem.
Thanks in advance
P.S. MPI with other sampling methods, e.g. CosmoHammer, works fine
Hi,
I am running my MontePython likelihood with the option -m NS to use MultiNest sampling (and obviously specifying the other MultiNest sampling options such as th enumber of live points etc..)
I have been doing this many times using 1 node, multiple cores on that node, and everything went fine. Unfortunately my likelihood is very slow, therefore I was hoping to gain some speed by running on MPI.
If I run the same likelihood with MPI, in my output file I get a lot of
and in the "NS" subfolder created in the chain output folder only two files get created, .arguments and .paramnames. I think this is not quite normal and I was wondering if you could help me identify the problem.
Thanks in advance
P.S. MPI with other sampling methods, e.g. CosmoHammer, works fine