feat: Tutorial Module 1: LLMs and Prompt Engineering

[anantmittal]

• Intro to LLMs: a large amount of data (almost the entire internet), attention-based architecture, transformers, needs a lot of computing, requires millions of $, and not everyone can train them.
• One way to interact with LLMs is prompts
• Common LLM parameters: temperature, top-k, etc.
• Prompting Techniques: Zero-shot, few-shot, and maybe another technique like chain-of-thought or ReAct with OLMo
• Prompt Templates (system, user, assistant)

[anantmittal]

• Intro to LLMs: a large amount of data (almost the entire internet), attention-based architecture, transformers, needs a lot of computing, requires millions of $, and not everyone can train them.

  • One way to interact with LLMs is prompts

  • Common LLM parameters: temperature, top-k, etc.

  • Prompting Techniques: Zero-shot, few-shot, and maybe another technique like chain-of-thought or ReAct with OLMo

According to our original plan, #39 comes after this module, but I think #39 can overlap with LLM parameters / prompting techniques section in this module.

[vanitech]

https://github.com/uw-ssec/tutorials/blob/main/SciPy2024/module1/llms-and-prompt-engineering-with-olmo.ipynb