Fine-tuning
The train
command of larvataggerand similar scripts admits argument
--fine-tune, to further train an already trained tagger. Under the hood, with this argument,
trainswitches to
tagging-backend finetuneinstead of
tagging-backend train`.
For example, MaggotUBA loads an existing tagger from the models directory instead of initializing a new tagger with a pristine classifier and a pretrained encoder loaded from the pretrained_models directory.