GTC 2024: Learning Notes 🤖📚

TL;DR:

This Streamlit app provides a summary of Site Wang's learnings from GTC 2024.

Deployed at streamlit community cloud

Running this App Locally

1. Switch to a new virtualenv
2. Install dependencies: pip install -r requirements.txt
3. Make sure you have env var set LLM calls
   • OPENAI_API_KEY for OpenAI calls
   • API_KEY for Google's Gemini
4. Kick off the app by doing streamlit run gtc_summary.py

Default Model Usage

Generally, the spending of all models below shouldn't be more than $0.1 for 1 day of extensive tuning/optimization.

1. Embedding Model: OpenAI's embedding with interface from LlamaIndex
2. RAG: LlamaIndex's RAG model with GPT3.5Turbo + OpenAI Embedding
3. Keyword Search: OpenAI's embedding
4. Company Search: Google's Gemini 1.5Pro
5. Code Generation
   • Blockers: GPT-3.5-Turbo
   • Code Planner: GPT-3.5-Turbo
   • Code Writer: Google Gemini
   • Code Reviewer: GPT-4-Turbo

Hacking

Please try not to hack it. Trust me you won't get much. If you are successful, I appreciate let me know. I would love to learn from you and block your hack.