Quote Generation with GPT-2

Overview

This program uses the GPT-2 language model to generate quotes based on the provided input. It is designed to run in Google Colab.

Getting Started

1. Prerequisites

• Google Colab account

2. Setup

2.1. Google Drive

Upload the files model.py and generate.py into the MyDrive location of your Google Drive(don't place it in a folder)

2.2. Google Colab

• Open the CS460.ipynb in Google Colab.
• Make sure the Google Drive in which you run the Colab Notebook is the same drive in which model.py and generate.py are stored
• Before starting, make sure to have your runtime type set to "T4 GPU"

3. Customize

• Adjust the number of quotes used to train the model by modifying the nrows variable in the pandas.read_csv cell
• Change k and length to play around with the quality of quotes generated

4. Run

Run the notebook cells sequentially to train the GPT-2 model and generate quotes. When asked for user input, click next to the prompt and enter a couple words or a topic that will be the start off the generated quote.

Tips

• Feel free to experiment with different prompts, hyperparameters, or GPT-2 model sizes for varied results.
• Larger datasets may improve model performance.

Note

• After running this code, there will be a new folder in your Drive called model. You may delete it afterwards to free space
• This README assumes basic familiarity with Google Colab.

Happy quoting!