Top Embedding Model for VectorDB

Unordered Embedding Model List

alina grubnyak / unsplash.com

VectorDB is a Database of vectors where we can index all of our vector collection and query it. Therefore, as long as we can transform any entity—like image, text, or sound— to vectors or embedding, we can utilize VectorDB on those entities. In this era, where artificial neural networks are heavily used with capable machines, transforming objects to embedding is not hard. Here we will list some of those projects with some details to be considered before choosing them for VectorDB projects.

RoBERTa

• Vector Size: 768

• Model Size: 476 MB

• Description: A RoBERTa model train on data up to December 2022, for users who are familiar with the BERT model family.

• Paper: https://arxiv.org/pdf/1907.11692.pdf

• Use for: General Text Embedding

• Limitations: It will truncate text longer than 512 tokens

• Source: https://huggingface.co/olm/olm-roberta-base-dec-2022