UKPLab / sentence-transformers

State-of-the-Art Text Embeddings
https://www.sbert.net
Apache License 2.0
15.18k stars 2.47k forks source link

MobileBert Fine Tuning #426

Open abcdchop opened 4 years ago

abcdchop commented 4 years ago

Hi guys! I've found your code to be really fantastic

I'm interested in training a Mobilebert model using your fine-tuning paradigm. I know you don't offer Mobilebert at present, but i was wondering if its possible port over a model from regular transformers and use the same scheme, or if there are specific modifications i could look at to make this viable. Thanks!

nreimers commented 4 years ago

Hi, MobileBERT is implemented in Huggingface transformers. So I expect you can just load it and use it for training.

See: https://www.sbert.net/docs/training/overview.html#creating-networks-from-scratch

Best Nils Reimers