RELAX exiting with "Internal error in _LikelihoodFunction::ConjugateGradientDescent."

[awalling]

Hello,

I am running the RELAX algorithm using HYPHY HYPHY 2.5.8(MP) for Darwin on x86_64, installed with conda, on a set of 25 alignments/trees. About half of the runs are exiting with the following status:

`### Fitting the general descriptive (separate k per branch) model Error: Internal error Internal error in _LikelihoodFunction::ConjugateGradientDescent. The function evaluated at current parameter values [-9267.630444432387] does not match the last recorded LF maximum [-9252.871211674597]

Function call stack 1 : [namespace = __QsXKaV] Optimize(mles, ^lf_id);

---

2 : relax.general_descriptive.fit=estimators.FitExistingLF(relax.general_descriptive.fit[terms.likelihood_function],relax.model_object_map);

Check errors.log for execution error details. `

I attach a representative alignment/tree/error log: chlorophyllide_a_reductase_iron_protein_subunit_X.edited.tre.txt chlorophyllide_a_reductase_iron_protein_subunit_X.fa.txt errors.log

I updated the conda installation of Hyphy today in case that was causing the error, but am still unable to run RELAX without error on about half of my dataset.

[spond]

Dear @awalling,

Version 2.5.8 is several years old by now; many of the changes since that version have been made to address issues like what you have reported here (convergence / optimizer issues).

Would you mind updating to the current version and seeing if this resolves the problem?

You can build from source or pull from bioconda (https://anaconda.org/bioconda/hyphy)

Best, Sergei

[github-actions[bot]]

Stale issue message