Queso gets stuck at level 1 step 3 while doing multi-layered inverse problem

[bello mubarak]

Hi All,

I observed that my queso model for inverse problem runs indefinitely when I increase the number of MCMC iterations or when I increase the number of calibration data used to calculate my normal-normal likelihood function.  For the full calibration data (9 - 11 data points),  Queso works fine when I use 500 iterations but as I move to 10,000 iterations, the model runs indefinitely. Looking at the display_sub0.txt file, it seems queso is stuck at level 1, step 3 (calculating exponent) and keeps trying to solve the step infinitely.

Here's a sample error line when I use 50,000 iterations

In uqMLSampling<P_V,P_M>::generateSequence(), level 1, step 3, failedExponent = 0: entering loop for computing next exponent, with nowAttempt = 1923276

In uqMLSampling<P_V,P_M>::generateSequence(), level 1, step 3: nowAttempt = 1923276, prevExponent = 0, exponents[0] = 0, nowExponent = 0, exponents[1] = 0, subWeightRatioSum = nan, unifiedWeightRatioSum = nan, unifiedOmegaLnMax = -0, weightSequence.unifiedSequenceSize() = 5000, nowUnifiedEvidenceLnFactor = nan, effectiveSampleSize = nan

Here's the right output when I use 500 iterations

 In uqMLSampling<P_V,P_M>::generateSequence(), level 1, step 3, failedExponent = 0: entering loop for computing next exponent, with nowAttempt = 3

In uqMLSampling<P_V,P_M>::generateSequence(), level 1, step 3: nowAttempt = 3, prevExponent = 0, exponents[0] = 0, nowExponent = 0.125, exponents[1] = 0.25, subWeightRatioSum = 576.603, unifiedWeightRatioSum = 11746.6, unifiedOmegaLnMax = 0.0759529, weightSequence.unifiedSequenceSize() = 20000, nowUnifiedEvidenceLnFactor = -0.456212, effectiveSampleSize = 2.7221e-06

How can I fix this please ?