search

deepset-ai / FARM

:house_with_garden: Fast & easy transfer learning for NLP. Harvesting language models for the industry. Focus on Question Answering.
https://farm.deepset.ai
Apache License 2.0
1.73k stars 247 forks source link

How to correctly add more training data to an existing model? #835

Closed rdemorais closed 2 years ago

rdemorais commented 3 years ago

Question Guys, I have a NER model trained with 5 classes. Now I'd like to add one more class giving tagged data for that new class. Unfortunately, if use this approach, even if my language model is the previous trained NER, it just get information for the last class, loosing all the others.

So, how can I add more data to a NER model without the need to train everything again?

Additional context First of all I have a Language Model (BERT LM) and CONLL file to be used to train NER. I have 5 classes in the CONLL file and I'm able to create the NER Model. It works greatly.

But now, I need to add one more class. How can I do that without train all adding the new class in the process? Should I use the previous NER model as language model? Should use the prediction head from the NER and combine it to the new one?

Thanks in advance.

stale[bot] commented 2 years ago

This issue has been automatically marked as stale because it has not had recent activity. It will be closed in 21 days if no further activity occurs.