The Transformers Extension proposes to use large pre-trained models using Transformers API in Exasol Database. These models can be used for many predictive tasks such as text classification, question-answering, text generation etc. in the field of NLP.
Pre-trained models contain many parameters and can grow very large. Therefore, it is not efficient to download these large models in each predictive task. Transformer Extensions is therefore divided into two different parts:
- Download : It is responsible to download the specified pre-defined model into the Exasol BucketFS.
- Prediction: These are a group of UDFs for each supported task. Each of
them caches the downloaded pre-trained model and perform prediction.
Supported tasks:
- Sequence Classification for Single Text
- Sequence Classification for Text Pair
- Question Answering
- Masked Language Modeling
- Text Generation
- Token Classification
- Text Translation
