ASashwinnair99·Aug 3, 2024Submission
Show HN: RAGBuilder – Hyperparameter tuning on various RAG parameters
A RAG has several moving parts: data ingestion, retrieval, re-ranking, generation etc.. Each part comes with numerous options.
If we consider a toy example, where you could choose from:
5 different chunking methods,
5 different chunk sizes,
5 different embedding models,
5 different retrievers,
5 different re-rankers/ compressors
5 different prompts
5 different LLMs
That’s 78,125 distinct RAG configurations! If you could try evaluating each one in just 5 mins, that’d still take 271 days of non-stop trial-and-error effort! In short, it’s kinda impossible to find your optimal RAG setup manually.
So, how do you determine the most optimal RAG configuration for your data and use-case?
Meet RagBuilder, a tool designed to help you create optimal, production-ready Retrieval-Augmented-Generation (RAG) setup for your data automatically.
RagBuilder
Read the rest
performs hyperparameter tuning on various RAG parameters, such as chunking strategy (semantic, character, etc.) and chunk size (1000, 2000, etc.), to evaluate these configurations against a test dataset. This process identifies the best-performing setup tailored to your data.
Check it out and let us know what you think!
https://github.com/kruxai/ragbuilder