I think it would be useful to include the is_best variable from train.train_model in the logs. Since log_info is passed to the epoch hook, the epoch hook function could then perform custom model checkpointing based on whether the current model is the best model so far.
For instance, I'm trying to checkpoint the classifier network itself (i.e. only AttackerModel.model, not the entire AttackerModel). This would be easier if an epoch hook had the information on whether the current model is best.
I think it would be useful to include the
is_best
variable fromtrain.train_model
in the logs. Sincelog_info
is passed to the epoch hook, the epoch hook function could then perform custom model checkpointing based on whether the current model is the best model so far. For instance, I'm trying to checkpoint the classifier network itself (i.e. onlyAttackerModel.model
, not the entireAttackerModel
). This would be easier if an epoch hook had the information on whether the current model is best.