LLM 101 Workshop

NIT Calicut, Kozhikode, Kerala

Kurian Benoy

Sunday, March 24, 2024

Who am I

Who am I

• ML Engineer & Team Lead at a Tech Startup at Singapore
• I’ve worked in Tech since 2018 and has been, ML Engineer & Data scientist for almost 5 years in various companies.
• Pursuing MTECH in AI & Data science @ IIIT, Kottayam.
• Completed BTECH in Computer Science @ Govt Model Engineering College, Thrikkakara.
• Speaker in International conferences like FOSSASIA Summit, Pycon India, Tensorflow Usergroup India summit etc.

Who am I

• Volunteer @ Swathanthra Malayalam Computing (SMC)
• I have contributed to Open source projects like Keras, Transformers, fast.ai etc.
• Creator of Indic Subtitler and Malayalam voice models like Vegam-whisper, MalWhisper etc.
• Maintains whisper_normalizer a python packages with 100K monthly downloads.

Introduction to LLMs

About cursor

Other cheap or inexpensive tools for Code generation

• Github Copilot
• VS Codium
• Amazon Code Whisperer
• OpenAI Code Interpreter with ChatGPT Plus

Let’s build something cool

• What should be that cool thing?
• Let’s build a website?
• You all have seen that website

Let’s build something cool

• FOSS MEET 2024 WEBSITE
• Usually professional websites are build with frameworks like Next.JS and ReactJS.

Landing page of FOSSMEET Website

Demo time

…..

Final Result with be pushed in website

OpenAI

• Not Open-source AI, but closed-source AI
• OpenAI will be know for the ChatGPT moment.
• Yet they were also behind all the work in GPT series, DALL-E, Gym, Gym-Netro, OpenAI Five beat 5 Doto players in 2018 etc
• Whisper is the most under-rated models released by OpenAI.

OpenAI Demo

Advanced LLMS

• Advanced techniques
• RAG
• Jail breaking
• Chatgpt + tools

How to fine tune LLM models?

• You need to fine-tune LLMs for only domain specific models

Slides generated by Nithin by Claude3 Opus Sonnet

Agenda

• 1. Introduction to GPT
• 2. How GPT Works
• 3. Applications and Use Cases
• 4. Future of GPT and AI

Introduction to GPT

• What is GPT (Generative Pre-trained Transformer)?
• Key breakthroughs and milestones in language models
• Importance and impact of GPT on natural language processing

How GPT Works

• Transformer architecture and self-attention mechanism
• Pre-training on large text corpora
• Fine-tuning for specific tasks and applications
• Key capabilities: text generation, language understanding, question answering

Applications and Use Cases

• Conversational AI and virtual assistants
• Content generation (articles, stories, code)
• Language translation and summarization
• Question answering and knowledge extraction
• Potential impact on various industries (education, healthcare, finance, etc.)