Stanford DSPy: The framework for programming with foundation models

DSPy is a programming framework designed to solve complex tasks using language models (LMs) and retrieval models (RMs). It offers Pythonic modules that allow users to instruct LMs using familiar syntax. By using DSPy, users can upgrade traditional prompting techniques and transform them into modular operations that can adapt to specific tasks. DSPy also introduces an automatic compiler that generates high-quality prompts for large LMs or trains automatic finetunes for small LMs. This framework is useful for building reliable models like GPT-3.5 and T5-base for various tasks. It provides installation instructions, syntax guidelines, and powerful concepts like signatures and teleprompters.

https://github.com/stanfordnlp/dspy

To top