Closed afaji closed 6 years ago
We are already working on that (issue #64 is getting some attention now). Including saving Adam statistics. This should also save the state of the scheduler and current best validation results, number of stalls.
Later this should also try to resume in roughly the same area in the training corpus. Probably only for the SQLite container.
I store the following scheduler parameters:
Do we need something else?
Just as a reminder, the validation scores a not correctly recorded which messes up early stopping. Otherwise it seems to run fine.
Should be fixed now.
How about the direction of validation, i.e. increasing/decreasing scores are better?
this impact early stopping criteria.