Multiple chains with ml sampling and fp

[Ernesto Lima]

Hello all,

I found an interesting behavior that I am not being able to understand. 

On the gravity example I replaced

  QUESO::GslVector paramInitials(paramSpace.zeroVector());

  priorRv.realizer().realization(paramInitials);

  QUESO::GslMatrix proposalCovMatrix(paramSpace.zeroVector());

  proposalCovMatrix(0,0) = std::pow(std::abs(paramInitials[0]) / 20.0, 2.0);

  ip.solveWithBayesMetropolisHastings(NULL, paramInitials, &proposalCovMatrix);

By 

  ip.solveWithBayesMLSampling();

And on the input file I deleted all the mh_ options and added the ML options:

ip_ml_default_rawChain_size               = 100

ip_ml_default_rawChain_dataOutputFileName = outputData/cal_rawChain_ml

ip_ml_last_rawChain_size                  = 1000

If I use the ip_ml_default_rawChain_size = 100 and fp_mc_qseq_size = 100

The code works perfectly with any combination of processors and environment. However, if I use 4 processors and 2 subenvironments, and use fp_mc_qseq_size >100 the code get stuck after 100 samples for the Monte carlo forward problem.

If I change for 2 processors and 2 subenvironments I dont have any issue.

If I use  ip_ml_default_rawChain_size = 1000 I dont have any issue using all kind of processor/sub-environment combination and size of fp_mc_qseq_size.

It maybe nothing important as the number of samples is to small. I just had this issue while I was testing the code and wondered if someone would know why I get this behavior. 

I am using QUESO 0.5.6.0.

Thanks

Ernesto

[Ernesto Lima]

Just installed QUESO 0.57.0 and had the same issue

[Damon McDougall]

Hi Ernesto,

With v0.57.1 I get a totally different error:

../../../src/basic/src/ScalarFunction.C, line 140, compiled Jul 20 2017
at 14:24:05
Implementation of all lnValue methods is missing. Please implement at
least lnValue(const V &).

Did you also get this?

Best wishes,
Damon


--
Damon McDougall
http://dmcdougall.co.uk
Institute for Computational Engineering Sciences
201 E. 24th St., Stop C0200
The University of Texas at Austin
Austin, TX 78712-1229