Data Portal | Data Science & Машиннное обучение

Отличное приложение на основе ИИ начинается с выбора подходящего типа эмбеддингов.

Вот 6 типов эмбеддингов и случаи, когда стоит их использовать:

• Sparse embeddings: https://weaviate.io/developers/weaviate/search/bm25
• Dense embeddings: https://weaviate.io/developers/weaviate/search/similarity
• Quantized embeddings: https://weaviate.io/developers/weaviate/concepts/vector-quantization
• Binary embeddings: https://weaviate.io/developers/weaviate/concepts/vector-quantization#binary-quantization
• Variable dimensions: https://weaviate.io/blog/openais-matryoshka-embeddings-in-weaviate
• Multi-vector embeddings: https://weaviate.io/developers/weaviate/tutorials/multi-vector-embeddings

👉 @DataSciencegx

Please open Telegram to view this post

VIEW IN TELEGRAM

www.tgoop.com/DataSciencegx/146

3.4K viewsApr 10 at 09:03

Отличное приложение на основе ИИ начинается с выбора подходящего типа эмбеддингов.

Вот 6 типов эмбеддингов и случаи, когда стоит их использовать:

• Sparse embeddings: https://weaviate.io/developers/weaviate/search/bm25
• Dense embeddings: https://weaviate.io/developers/weaviate/search/similarity
• Quantized embeddings: https://weaviate.io/developers/weaviate/concepts/vector-quantization
• Binary embeddings: https://weaviate.io/developers/weaviate/concepts/vector-quantization#binary-quantization
• Variable dimensions: https://weaviate.io/blog/openais-matryoshka-embeddings-in-weaviate
• Multi-vector embeddings: https://weaviate.io/developers/weaviate/tutorials/multi-vector-embeddings

👉 @DataSciencegx