Can't restart from sys_error.pkl if previous run failed at update_interpolator - Githubissues

eckelsjd / amisc

Efficient framework for building surrogates of multidisciplinary systems using the adaptive multi-index stochastic collocation (AMISC) technique.

https://eckelsjd.github.io/amisc/

GNU General Public License v3.0

1 stars 1 forks source link

Can't restart from sys_error.pkl if previous run failed at update_interpolator #21

Open eckelsjd opened 1 month ago

eckelsjd commented 1 month ago

Expected Behavior

restarting surr.fit() from a sys_error.pkl file should pick up right where it left off

Actual Behavior

if the the previous fit() failed due to an error in update_interpolator, then the index and candidate sets get messed up due to an early update of comp.surrogates dictionary.

Steps to Reproduce the Problem

Call surr.fit() and force a failure in the component model, which will cause update_interpolator() to fail
Try to restart surr.fit() from the sys_error.pkl file

Ideas

Can probably put the update_interpolator() right after the build_interpolator call in comp.add_surrogate. No data structures should get updated until we absolutely know the new candidate index has been built properly.