Company
Date Published
March 30, 2021
Author
Etienne Dilocker
Word count
258
Language
English
Hacker News points
None

Summary

Weaviate v1.2 introduces transformers (DistilBERT, BERT, RoBERTa, Sentence-BERT, etc) for vectorizing and semantically searching through data. Transformers are deep learning models used in NLP tasks. With the text module selected in the Weaviate configuration tool, users can run Weaviate with transformers in one command. Custom transformer models compatible with Hugging Face's AutoModel and AutoTokenizer can also be utilized. Weaviate now allows for sub-50ms results by using transformers on your own data.