RPG GPT

Description

This project has the goal of create a LLM model specialized to be a RPG master, with that we desired that anyone could play a RPG adventure interacting with the LLM model through the chat and also improve the immersion in the adventure using a stable diffusion to generate images of what LLM model describe during the interactions via chat. Besides that, we could not to forget to add a d20 to give us a chance to have a critical success or a critical failure in our actions.

How to install

• Step 1: Clone this repository

git clone

• Step 2: Install miniconda (https://docs.anaconda.com/free/miniconda/index.html)

• Step 3: Create env with miniconda with python 3.9

conda create -n env python=3.9

• Step 4: Activate your new python environment

conda activate env

If you don't have a NVIDIA GPU you can skip the step 5.

• Step 5 (For NVIDIA GPU users): Install CUDA 11.8 using miniconda

conda install -c conda-forge cudatoolkit==11.8.0

conda install -c conda-forge cudnn==8.9.7.29

• Step 6: Download Ollama (https://ollama.com/)

• Step 7: Install llama3 model

ollama pull llama3

• Step 8: Install project dependencies

pip install -r requirements.txt

Observation: If you don't have a NVIDIA GPU, you will need to install Pytorch for CPU.

How to use

• Step 1: Activate your python environment where you installed the project dependencies and access the folder where you find rpggpt.py file and run the following command

streamlit run rpggpt.py

After this you will be redirected to a new tab on your browser where you will be able to interact with the RPG GPT.