/dspy-presentation

Primary LanguagePython

DSPY DEMO (IU TECHNICAL EXCHANGE)

This project implements a Chain of Thought (CoT) approach to solve math problems from the GSM8K dataset using DSPy and OpenAI's GPT-3.5-turbo model.

Features

  • Utilizes DSPy framework for natural language processing tasks
  • Implements Chain of Thought reasoning
  • Uses OpenAI's GPT-3.5-turbo model
  • Includes evaluation and optimization components

Setup

  1. Install required packages:

    pip install dspy openai

  2. Set up OpenAI API key in your environment variables.

Usage

Run main.py to:

  1. Initialize the CoT model
  2. Evaluate the baseline model
  3. Optimize the model using BootstrapFewShotWithRandomSearch
  4. Save the compiled model

Key Components

  • CoT: Implements the Chain of Thought module
  • Evaluate: Assesses model performance on the dev set
  • BootstrapFewShotWithRandomSearch: Optimizes the model using few-shot learning

Output

The optimized model is saved as iu_demo.json.

Note

Ensure you have necessary permissions and credits to use the OpenAI API.