2024-04-05 –, Room 111
How to use KDTree from sklearn library to prototype RAG (Retrieval-Augmented Generation) applications.
RAG, or Retrieval-Augmented Generation, is a method to enhance LLM output by combining retrieved relevant information from a large dataset with a generator model like GPT to produce informed and contextually relevant outputs. To retrieve relevant data, it has to be organized in the data structure that allows efficient search based on the actual meaning of the text. The common approach is to use vector databases like Weaviate or Pinecone. In the presentation I’ll share my experience of how to use the KDTree structure from sklearn library to work with vectorized data as a fast way of building a RAG application prototype.
Graduated from Mechanics and Mathematics faculty of BSU in 2014.
Developer at flespi, Gurtam since 2016.