391 reports a crash in ParameterGenerator.check_specific_stopping_critera() in a case where chi2 does not significantly improve over several hundred models (iterations?).
What may be happening based on the information provided:
The problem occurs in the method ParameterGenerator.check_specific_stopping_critera() that checks for each iteration whether the specific stopping criteria have been met. This is the case if the best $\chi^2$ value of the last iteration does not improve the best $\chi^2$ value of the preceding iteration by at least min_delta_chi2_rel or min_delta_chi2_abs (depending on the setting in the configuration file). Looking at the $\chi^2$ vs. model id plot provided in #391, the $\chi^2$ values do not seem to really improve between model 180 to 540 or so.
It may be that something like the following is happening: when checking for the best $\chi^2$ of the preceding iteration, that iteration may be empty (i.e., does not contain any models). Consequently, calculating the minimum $\chi^2$ of no models fails and throws the error. However, I don’t fully understand why this does not happen in the previous iteration, where the error should be thrown in line 588 instead of 596 of parameter_space.py…
In this PR, the method ParameterGenerator.check_specific_stopping_critera() now compares the last iteration's minimum $\chi^2$ to the overall minimum $\chi^2$ so far.
Dear @avichaturvedi, please try whether this PR solves your issue and let me know - many thanks!
391 reports a crash in
ParameterGenerator.check_specific_stopping_critera()
in a case where chi2 does not significantly improve over several hundred models (iterations?).What may be happening based on the information provided:
ParameterGenerator.check_specific_stopping_critera()
that checks for each iteration whether the specific stopping criteria have been met. This is the case if the best $\chi^2$ value of the last iteration does not improve the best $\chi^2$ value of the preceding iteration by at leastmin_delta_chi2_rel
ormin_delta_chi2_abs
(depending on the setting in the configuration file). Looking at the $\chi^2$ vs. model id plot provided in #391, the $\chi^2$ values do not seem to really improve between model 180 to 540 or so.parameter_space.py
…In this PR, the method
ParameterGenerator.check_specific_stopping_critera()
now compares the last iteration's minimum $\chi^2$ to the overall minimum $\chi^2$ so far.Dear @avichaturvedi, please try whether this PR solves your issue and let me know - many thanks!