LLM-RL-Agent

This project implements reinforcement learning agents for various environments using large language models (LLMs).

Project Structure

.env
.gitignore
BipeadalWalker/
    BipedalWalker.py
Cartpole/
    cartpole_v2.py
    cartpole_v3.py
    CartPole_Without_LLM.py
    CartPole.py
    Improvements.md
FrozenLake/
    frozenlake_avg_test_rewards_smoothed1.png
    frozenlake_plot_values.csv
    frozenlake_training_rewards_smoothed1.png
    FrozenLake.py
MountainCar/
    MountainCar_v2.py
    MountainCar.py
poetry.lock
pyproject.toml
README.md

Requirements

• Python 3.11
• Gymnasium
• OpenAI
• NumPy
• Matplotlib
• Google GenAI
• Stable Baselines3

Installation

1. Clone the repository:

   git clone https://github.com/kevalshah14/LLM-RL-Agent.git
   cd llm-rl-agent

2. Install dependencies using Poetry:

   poetry install

Usage

CartPole

To run the CartPole agent:

python Cartpole/CartPole.py

MountainCar

To run the MountainCar agent:

python MountainCar/MountainCar.py

FrozenLake

To run the FrozenLake agent:

python FrozenLake/FrozenLake.py

BipedalWalker

To run the BipedalWalker agent:

python BipeadalWalker/BipedalWalker.py

Project Details

CartPole

The CartPole agent logs its Q-table and testing results in the cartpole_logs directory. Each episode has its own subdirectory containing the Q-table and testing results.

MountainCar

The MountainCar agent logs its results in the mountain_car_logs directory.

FrozenLake

The FrozenLake agent logs its results in the frozenlake_logs directory.

BipedalWalker

The BipedalWalker agent logs its results in the bipedalwalker_logs directory.

License

This project is licensed under the MIT License.