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

• GPTX-RL-371K (Very Small) (Complete!)

• GPTX-RL-1M (Small) (Coming Soon!)

• GPTX-RL-14M (Medium) (Coming Soon!)

• GPTX-RL-140M (Large) (Coming Soon!)

• GPTX-RL-440M (Huge) (Coming Soon!)

---

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.