Jupyter Notebook
3,424 stars
🦖 𝗟𝗲𝗮𝗿𝗻 about 𝗟𝗟𝗠𝘀, 𝗟𝗟𝗠𝗢𝗽𝘀, and 𝘃𝗲𝗰𝘁𝗼𝗿 𝗗𝗕𝘀 for free by designing, training, and deploying a real-time financial advisor LLM system ~ 𝘴𝘰𝘶𝘳𝘤𝘦 𝘤𝘰𝘥𝘦 + 𝘷𝘪𝘥𝘦𝘰 & 𝘳𝘦𝘢𝘥𝘪𝘯𝘨 𝘮𝘢𝘵𝘦𝘳𝘪𝘢𝘭𝘴
### Overview:
This project is a comprehensive guide designed to help developers learn about Large Language Models (LLMs), Large Language Models Operations (LLM Ops), and Vector Databases for the purpose of building a real-time financial advisor LLM system. The course is structured into three main sections: building blocks, setup external services, and installation and usage. It includes practical steps for training, streaming, and inference pipelines, as well as a focus on financial Q&A datasets and best practices for deployment.
### Key features:
- Training pipeline that fine-tunes an open-source LLM using QLoRA and logs experiments on Comet ML's experiment tracker and inference results on LLMOps dashboard.
- Streaming real-time pipeline that ingests financial news and transforms documents into embeddings using Bytewax, and stores embeddings in Qdrant Vector DB.
- Inference pipeline that uses LangChain to download and run a fine-tuned model.
### Tech stack:
- Jupyter Notebook
- Python
- Comet ML
- Bytewax
- Qdrant
- LangChain
- Beam
- AWS
### Why it matters:
This course provides a practical approach to learning and deploying a financial advisor LLM system, focusing on real-world applications and best practices. It is designed to be accessible and applicable to developers looking to build production-ready systems for financial advisory purposes.