How to Train Your Llama: Lessons Learned from Finetuning Llama 3.1 on Thousands of Threat Actor Telegram Messages

Training and fine tuning LLMs is an incredibly complex process, but thanks to different libraries and frameworks we can easily find our own data and fine tune open source models like Llama 3.1.

This talk will be the story of how I scraped Telegram channels operated by threat actors and used this data to fine tune Llama 3.1. It will give attendees an easy way to fine-tune their own models and demonstrate what steps to take and what pitfalls to avoid.

This talk has passed, but you can view the slides and code here:

GitHub - cwolff411/HowToTrainYourLlama: Slides and code for my presentation at Bsides Philly

Slides and code for my presentation at Bsides Philly - cwolff411/HowToTrainYourLlama

GitHubcwolff411