- To be clear, this is not a roadmap for
getting startedwith LLMs. - I am not covering the books you should study, university studies, certificates, etc.
- I assume you have basic understanding of NLP stuffs, programming knowledge ( mainly Python and Maths ).
- You might argue, why Maths as everything is automated. Well, well, behind the scene, almost everything is Maths 🧠 )
- Calculus, Probability, Linear Algebra
- You need to know, Lets say what is matrix, how dot product works, etc etc.
- These are some of the resources which I suggest you to get started.
- After knowing the basics and how things work, it's upon you, what to do ( Or lets say if it's your cup of tea / coffee or not )
Remember one thing, using LLMs and implementing are two different things, you need not necessary know how to implement, but you need to know how to use it in right way.
- A Hacker's Guide to Language Models by Jeremy Howard.
- [1hr Talk] Intro to Large Language Models by Andrej Karpathy.
- Neural Networks: Zero to Hero by Andrej Karpathy.
- Building RAG from scratch Using Python, LangChain and OpenAI API by Santiago.
- fast.ai courses -->
Optional but highly recommended - DeepLearning.AI short courses -- My request, try to complete all this free short courses.
- DeepLearning.AI Specializations
- Prompt Engineering Guide
- OpenAI doc about Prompt Engineering
- Strategies to harness the power of LLMs
- There is one from deeplearing.ai free short courses too about ChatGPT Prompt Engineering for Developers.
- There are many courses, articles, videos about this topic, it needs constant learning and experimenting.
Frameworks which I have explored untill now, there are many, you can give a try ( your world, your rules )
OpenAI has really good documentation and Cookbook
- There is unlimited knoweledge you can grasp, try to find the best ones and follow them instead of jumping among videos.
- Main thing is to understand things and try it yourself. Unless you try (practice youself), you won't learn.
- I have videos on LLMs with playlist on langchain, chainlit and Llamaindex. Many LLMs videos to follow in 2024
Main thing I want to highlight, practice practice and practice, take help with AI assistants 👇
- Perplexity AI --> let's put this way, it's Google Search with LLMs with it.
- Perplexity Labs, For Open Source models
- ChatGPT --> Based on your need, free or paid version. ( Team, Enterprise , etc)
- Bing Chat , Bing Enterprise.
- Hugging Chat
- Le Chat Mistral
- First, think on tweeking basic stuffs
- Cleaning document ( choose right parsing , eg. LlamaParse, Unstructured )
- Better Chunking strategies
- Choosing right embeddings model
- Choosing right Vectorstore
- Passing parsing Instructions, Reranking
- Choosing right Large Language Models
Links to follow for better understanding.
- Chunk visualizer
- Tokenizer, from OpenAI
- Huggingface Massive Text Embedding Benchmark (MTEB) Leaderboard
- What is a Vector Database & How Does it Work? Use Cases + Examples
- Chunking Strategies for LLM Applications
- 🤗 Open LLM Leaderboard
- 🏆 LMSYS Chatbot Arena Leaderboard
- 12 RAG Pain Points and Proposed Solutions
- Optimizing RAG with Hybrid Search & Reranking
- Improving RAG performance with Knowledge Graphs
- Enhancing RAG with a Multi-Agent System
- Llama-3-Groq-Tool-Use Models
- Berkeley Funciton Calling Leaderboard
- Independent analysis of AI models and API providers 📌
This page will be updated over time. Cheers !!