generative-text based on prompt - GPT

This script allows you to generate text using a pre-trained language model. It only generates a few words at the moment, I may have better results using another pre-trained model.

I created this script to familiarize with the tools because in the future I hope we could find some pre-trained models like chatGPT from OpenAI directly on HuggingFace. It would be awesome to be able to run locally this kind of tech without any limitations (except your hardware).

Requirements

Python 3.6 or higher
PyTorch 1.7.0 or higher
Transformers 3.2.0 or higher

# First, install the Hugging Face Transformers library
!pip install transformers

Moreover, you need to install pytorch in the prerequisites. For this you need a compatible CUDA GPU . Please have a look on : https://pytorch.org/get-started/locally/ to install the right version on your computer. Here is mine :

pip3 install torch torchvision torchaudio --extra-index-url <https://download.pytorch.org/whl/cu116>

Usage

To use the script, run the following command:

python generate_text.py "Your input prompt here"

Replace "Your input prompt here" with the text that you want to generate text from.

Customization

You can customize the script by changing the following variables:

model_name: The name of the pre-trained model to use. By default, the script uses the "asi/gpt-fr-cased-base" model, but you can choose any model that is supported by the Transformers library.

Output

The script will print the generated text to the console.

Limitations