nvidia-generative-ai-notes

Vector Database Embedding

Vector databases store and manage high-dimensional vector representations of data, enabling efficient similarity search and retrieval. Embeddings are crucial for transforming raw data (like text, images, or audio) into these vector representations.

Types of Embeddings

Text Embeddings: Convert text into vectors. Common models include Word2Vec, GloVe, FastText, and transformer-based models like BERT and GPT.
Image Embeddings: Convert images into vectors. Models like ResNet, VGG, and CLIP are commonly used for this purpose.
Audio Embeddings: Convert audio signals into vectors. Models like OpenL3 and VGG are used for audio embedding.

Model Dimensions Optimized For
OpenAI text-embedding-3-small 1536 General purpose
OpenAI text-embedding-3-large 3072 Higher accuracy
Cohere embed-v3 1024 Multilingual
BGE (BAAI) 768 / 1024 Open source, strong performance
E5 (Microsoft) 768 / 1024 Retrieval tasks
GTE (Alibaba) 768 General text
Sentence-T5 768 Sentence similarity
Instructor 768 Task-specific with instructions