🧠 RL & Imitation Learning Notebooks (Game AI Experiments)

A collection of practical experiments where AI agents learn to play games using Reinforcement Learning and Imitation Learning.

Instead of focusing on theory, this repository explores real challenges:

• training instability
• reward design
• exploration vs exploitation
• learning from imperfect demonstrations

🎮 What you'll find

• Agents learning to play fighting games and survival scenarios
• Imitation Learning (Behavioral Cloning, DAgger-style approaches)
• Reinforcement Learning (PPO, A2C, DQN)
• Custom preprocessing pipelines (frame stacking, cropping, grayscale)
• Real training runs, including failures and unexpected behaviors

🚀 Why this repo exists

Most RL repositories show clean results. This one shows the messy reality of training agents — and how to make them work anyway.

📂 Structure

• re4/ → Resident Evil experiments
• stf6/ → Fighting game agents
• more experiments coming...

🔥Starting

1. Clone this repo: https://github.com/paulo101977/notebooks-rl.git

2. Move to cloned folder: cd notebooks-rl

3. First create a conda environment with python 3.11: conda create -n env311 python=3.11

4. Activate it: conda activate env311

5. Install requirements: pip install -r requirements.txt

6. Install Jax: pip install git+https://github.com/araffin/sbx

7. (Optional but recommended) Install Jupyter kernel: python -m ipykernel install --user --name env311 --display-name "Python (env311)"

8. Run Jupyter-lab: jupyter-lab.exe

⚡ PyTorch (GPU Support)

If you have an NVIDIA GPU with CUDA support, install PyTorch separately: pip install torch torchvision --index-url https://download.pytorch.org/whl/cu121

⚠️ Make sure to match the CUDA version (cu118, cu121, etc.) with your system. You can check compatibility here: https://pytorch.org/get-started/locally/

If you don't have a GPU, install the CPU version: pip install torch torchvision

❤️ Support

If you find this useful, consider sponsoring: https://github.com/sponsors/paulo101977