RLAI_proj

This repository serves as a cloud backup for a reinforcement learning project focused on training, tuning, and evaluating an AI model using the PPO (Proximal Policy Optimization) algorithm in a Humanoid-v4 environment. The project includes scripts for managing training iterations, adjusting hyperparameters, visualizing the training, and saving model generations.

Project Structure

• Training:

  • main.py: Main entry point for model training, allowing users to set the number of training iterations.
  • train2.py: Contains the core training loop for the reinforcement learning agent, with hyperparameter adjustments and model checkpoints.

• Hyperparameter Tuning:

  • tune_hyperparameters.py: Uses Optuna to optimize key hyperparameters for improved model performance.

• Visualization:

  • visualize_training.py: Loads a trained model and runs it in the Humanoid-v4 environment with rendering enabled, allowing real-time visualization of the agent’s behavior.

Dependencies

• Python 3.x
• Stable Baselines3 for reinforcement learning models
• PyTorch for neural network operations
• Optuna for hyperparameter optimization

Usage

Clone this repository and use the provided scripts to train, tune, or visualize the model’s performance. This setup is intended for experimentation and improvement of reinforcement learning model training.

---

This Markdown code will display properly on GitHub with bolded text, headers, and the updated Visualization section for visualize_training.py. Let me know if you’d like any other modifications!