DPRTextEncoder is a class that encodes text into embeddings using the DPR model from huggingface transformers.
The DPR model was originally proposed in Dense Passage Retrieval for Open-Domain Question Answering.
This encoder supports both the DPR context and question encoders - you should specify which type you are using with the encoder_type parameter.
If the encoder_type is context, please also specify the context encoder model by setting the pretrained_model_name_or_path.
As an example, if encoder_type is context, you may set pretrained_model_name_or_path to facebook/dpr-ctx_encoder-single-nq-base.