GPTX-RL

Features

Just like GPT this model generates text based on an input prompt.
This project uses Reinforcement Learning (RL) for training and inference.
All models can be found on hugging face.

Models

Training

First step is to run generate_vocab.py on whatever text data you would like to train on for example python3 generate_vocab.py -i train.txt
Next step is to run train.py
Now all you have to do is wait!

Inference

Run inference.py with -i as an argumentand with models path.

Logging

Output plot graphs will be avalible in the next version.