Advanced RAG with ColBERT in LangChain and LlamaIndex
Вставка
- Опубліковано 31 тра 2024
- ColBERT is a fast and accurate retrieval model, enabling scalable BERT-based search over large text collections in tens of milliseconds. This can be used as a potential alternative to Dense Embeddings in Retrieval Augmented Generation. In this video we explore using ColBERTv2 with RAGatouille and compare it with OpenAI Embedding models.
🦾 Discord: / discord
☕ Buy me a Coffee: ko-fi.com/promptengineering
|🔴 Patreon: / promptengineering
💼Consulting: calendly.com/engineerprompt/c...
📧 Business Contact: engineerprompt@gmail.com
Become Member: tinyurl.com/y5h28s6h
💻 Pre-configured localGPT VM: bit.ly/localGPT (use Code: PromptEngineering for 50% off).
Signup for Advanced RAG:
tally.so/r/3y9bb0
LINKS:
Find the Notebook here: github.com/PromtEngineer/Yout...
ColBERTv2 with RAGatouille Video: • Supercharge Your RAG w...
ColBERTv2 Paper: arxiv.org/pdf/2112.01488.pdf
ColBERT Github: github.com/stanford-futuredat...
RAGatouille: github.com/bclavie/RAGatouill...
TIMESTAMPS:
[00:00] Introduction
[00:29] Use ColBERT in LangChain
[08:46] Use ColBERT in LlamaIndex
All Interesting Videos:
Everything LangChain: • LangChain
Everything LLM: • Large Language Models
Everything Midjourney: • MidJourney Tutorials
AI Image Generation: • AI Image Generation Tu... - Наука та технологія
If you are interested in learning more about how to build robust RAG applications, check out this course: prompt-s-site.thinkific.com/courses/rag
can you make a video on how to evaluate a RAG? And compare different RAG approaches.
I would also be interested in this :) specially with open source llm's and embeddings. Tried alot and cant figure out which is the best one
@@joxxen I`m to waiting for that if you found any resource let me know.
Navigating the landscape of storytelling and video experimentation, VideoGPT silently empowers my creative journey, adding a layer of sophistication to my content.
Thanks for sharing
@engineerprompt, can we use a persistant vector db like chroma, qdrant and others with Ragatouille? So that I can just embed the documents once and re-use them for inferences later.
It supports only FAISS at the moment for persisting it to disk
I’d like to do RAG over a medical textbook. What strategies would you recommend for chunking. I’m thinking a hierarchical graph structure makes intuitive sense. What are your thoughts on this?
Cant find the google collab notebook? Would love to copy this across to my own account and havd a play. Not sure if I'm overlooking it? I just see the github link?
Is the Plaid DB persistent? As in, if I do this, how do I connect to that particular DB again?
I am working on a machine that is running Ubuntu and connected to 4 80GB A100 GPU's. The issue i face is RAG.index cell is running forever on this machine. Whereas same code running on Google Colab free version runs within seconds. Any insights on how this can be resolved will be helpful. Thanks :)
Is your env able to see the GPUs? Check that the torch is actually using the gpu
@@engineerprompt yes i run LLM's on same notebook, it is able to load that to gpu. I checked via nvidia-smi command
How can we use approaches like ColBERT with other languages, as portuguese? Thanks!
I think you will have to finetune the model for the language first
of course the last result is more accurate. you gave it almost 50% (5 instead of 3 chunks) more context. when using multiple ways to achieve the same goal, please use the same amount of data. otherwise it is hard to compare the output.
on the topic of chunks given to RAG - why define that? what if one does not know about how many parts may contain relevant information?
How can I monetize whatever is being said as a beginner..?
If you are interested in leanring more about Advanced RAG Course, signup here: tally.so/r/3y9bb0