Relationship with dbmdz/distilbert-base-german-europeana-cased?

#2
by Marissa - opened

Following up on @Marissa 's question, I wondered if there is more info on the data that were used for the initial training of distilbert-base-german-cased?
Thanks!

DistilBERT community org

cc'ing @stefan-it just in case :)

Are there any further informations about the training of this model? It's hard to compare without these informations

Hi guys!

More information can be found here:

https://github.com/huggingface/transformers/pull/1873

So there's no connection between the DistilBERT for Europeana, because this model here was really trained and distilled from the (DBMDZ) German BERT model, whereas the Europeana DistilBERT model uses the Europeana BERT model as teacher (Europeana BERT was trained on historic German newspapers).

@Marissa I'm sorry for the late reply!

Thanks so much, @stefan-it . That helps a lot! ☺️ Do you know whether it’s possible to also update the model card with this relevant information (to guide future users)?

Sign up or log in to comment