RAG explained: A Step-by-Step Guide to Vector Search and Content Retrieval
ฝัง
- เผยแพร่เมื่อ 7 ก.ย. 2024
- This is the first video covering the fundamentals that we need for designing the Chatbots. We start by breaking down text embedding and its role in Retrieval-Augmented Generation (RAG) systems. Then, we put theory into practice by building a simple RAG system from scratch.
📘 Get the tutorial notebooks here: github.com/Far....
📚 Explore the full GitHub repo: github.com/Far...
🔗 Connect on LinkedIn: / farzad-roozitalab
🌐 Visit my website: farzad-r.githu...
Stay tuned for practical insights and don't forget to subscribe for updates on future videos!
#RAG #llm #ChatBot #GPT #Python #AI #OpenAI #Langchain #Gradio #chroma #embedding
Amazing. This is superior to so much other content out there!
Thanks @smehra! That means alot. I am glad that you liked the content
You are amazing Farzad! Many thanks to you for this valuable tutorial :)
Thanks Saeid! It's a pleasure to see that the tutorials are helpful
So far the best video on this topic.
Very beginner friendly RAG implementation 🙂👍
Thanks! I am glad you liked it
this is very awesome. I need this type of series. now waiting next video for function calling?
Thanks. I am glad you liked the video. I removed the function calling video because OpenAI updated the functions and they changed the way the models were called. But you can have access to the code here:
- github.com/Farzad-R/LLM-Zero-to-Hundred/tree/master/tutorials/LLM-function-calling-tutorial
plus, I am using function calling in the following two projects:
- th-cam.com/video/KoWjy5PZdX0/w-d-xo.htmlsi=DN1Gt6sA8W-E4C2l
- th-cam.com/video/55bztmEzAYU/w-d-xo.htmlsi=kMV5ZtPaGugVckP6
And in the next video I will explain the new ways of function calling and how to design LLM agents from scratch. That project is almost ready!
I have a question.
How do we search similar vectors in real applications when the vector db is huge?
Isn’t iterating over each vector is db inefficient?
I know there may be other efficient search mechanisms, just want to know what are they.
Great question. On a normal database that would be a serious concern. But thanks to the architecture and computation behind vector databases this is not a serious concern unless you have millions of documents (which again that challenge can be solved by parallelization for example). But in general searching over vector databases does not occur by a simple for loop in the background and it is much faster than that. If you are curious to implement and test it, you can check my RAG-GPT video where I implement a full RAG chatbot using Chroma vectorDB. There you can see the speed of inferencing with GPT 3.5
@@airoundtable thank you for the explanation.
And yes I am going to watch the entire playlist 🙂
@@sumitsp01 You're welcome! I hope the series is informative for you then. Enjoy 😉
Wow you are too good.
Thanks! I am glad to see that you liked the content!
you're a beast
:))) thanks @Mar10001! I am glad the content was helpful to you
👏👏👏
Kudos to you. Great content and explanation
Thanks Mehrdad!