Learn to optimize end-to-end LLM pipelines in this 10-minute tutorial video that demonstrates implementing a multi-hop question-answering pipeline using DsPy's Baleen architecture. Explore the implementation details including the Phi LM and Colbert V2 retrieval models, work with the HotPotQA dataset, understand evaluation metrics, and master pipeline compilation using the BootStrap Few Shot algorithm. Access the complete implementation through the provided GitHub notebook while following along with the step-by-step walkthrough of the Baleen architecture and its optimization process. Gain practical insights from an experienced machine learning researcher who breaks down complex concepts into digestible segments, complete with timestamps for easy navigation through specific topics.
Overview
Syllabus
- Intro
- Architecture of Baleen
- Baleen Implementation Overview
- Phi LM and Colbert V2 Retrieval Model
- HotPotQA Dataset
- Evaluation Metric
- Compiling the DsPy Pipeline
Taught by
AI Bites