How to save model in spacy for custom NER which is again trainable?

Viewed 122

Let's say I have trained my model with label called "FRUIT" and saved it using nlp.to_disk("F:/Saved Models/FRUIT_pipeline")

Now if I load this model and train again for label called "SEASON", all my before training is lost. How do I save it for multiple labels in multiple training?

1 Answers

Please see the FAQ entries on catastrophic forgetting.

Basically, the easiest workaround is to train your model on all your training data at once. There are other methods to avoid forgetting but they are experimental.

Related