README.md

MPT-7B-StoryWriter Text Generation

This repository contains a Python script that demonstrates how to use MosaicML's MPT-7B-StoryWriter model for text generation. The model is based on the transformers library and is designed to generate coherent and contextually relevant text based on a given input.

Tested on a 3090.

Prerequisites

To use this script, you need to have Python 3.7 or later installed. Additionally, you will need the following Python packages:

torch
transformers

You can install these packages using the following command:

pip install torch transformers

Usage

The run.py script contains a function called generate_text(input_text: str) -> str that takes an input text string and generates relevant text based on the input.

First, import the generate_text function from the script:

from run import generate_text

Then, call the function with your desired input text:

input_text = "Once upon a time in a faraway land,"
output_text = generate_text(input_text)
print(output_text)

The generate_text function will return a string containing the generated text based on the input. You can adjust the input text to suit your needs and generate different text outputs.

Customization

You can customize the behavior of the text generation by modifying the configuration settings in the run.py script. For example, you can change the max_seq_len parameter to control the maximum length of the generated text. Please refer to the transformers documentation for more information on the available configuration settings.

License

This project is licensed under the MIT License. See the LICENSE file for details.

Acknowledgments

Hugging Face for providing the transformers library.
EleutherAI for providing the GPT-NeoX model.
MosaicML for providing the MPT-7B Storywriter model.