🛠️ Installation

Ensure you have Python and CUDA set up in your environment. Install the necessary dependencies by running:

git clone https://github.com/Tizzzzy/influence_ppo.git
cd influence_ppo
pip install -r requirements.txt

🏃 Quick Start: Single GPU Training & Evaluation

You can quickly launch a training and evaluation run using the provided run_single_gpu.sh script. By default, this script is configured to train a Rho-1B model on the GSM8K dataset using I-PPO.

chmod +x run_single_gpu.sh
bash ./run_single_gpu.sh