Skip to content

Latest commit

 

History

History
77 lines (58 loc) · 2.7 KB

project_structure.md

File metadata and controls

77 lines (58 loc) · 2.7 KB

Project Folder Structure and Overview

1. configs/

Contains configuration files to customize the environment, model, and reward settings.

  • Files:
    • config.yaml: Main configuration file that combines all settings.
    • Subfolders:
      • environment/: Contains specific configurations for environments (ant.yaml, quadrobo.yaml).
      • model/: Defines hyperparameters for models (td3.yaml).
      • rewards/: Reward shaping configurations (rewards.yaml).

2. environments/

Contains the implementation of custom MuJoCo environments.

  • Files:
    • ant_env.py: Code for the Ant environment.
    • quad_env.py: Code for the Quadrobo environment.
  • Purpose:
    • Provides the simulation logic and state transitions for each environment.

3. models/

Implements RL algorithms and their base functionalities.

  • Files:
    • td3.py: Implements the TD3 algorithm.
    • base_model.py: Base class with common functionality like saving and loading weights.

4. utils/

Utility scripts for shared functionality.

  • Files:
    • replay_buffer.py: Implements the replay buffer for experience storage.
    • rewards.py: Custom reward functions used during training.

5. scripts/

Scripts for running training, evaluation, and testing.

  • Files:
    • train.py: Launches the training process.
    • evaluate.py: Evaluates the trained model.
    • test_env.py: Tests and debugs environments.
    • view_model.py: Visualizes the model.

6. outputs/

Stores logs, metrics, and checkpoints generated during training and evaluation.

  • Structure:
    • Organized by timestamps.
    • Contains:
      • Training logs (train.log).
      • Evaluation logs (evaluate.log).
      • Model checkpoints.
      • Metrics and visualizations (metrics.png, contact_forces.png).

7. quadrobo_gym/

Implements and registers the Quadrobo custom Gym environment.

  • Files:
    • env.py: Defines the QuadroboEnv class.
    • setup.py: Setup script to register the environment.
    • register.py: Handles registration logic.

8. videos/

Stores recorded videos of evaluations for visualization (evaluation_demo.webm).

9. logs/

Holds debugging and runtime logs.

Key Features

  1. Modularity: Each folder handles a specific aspect (environments, models, scripts, etc.), making it easy to update or extend.
  2. Configuration-Driven: Centralized configurations allow seamless customization without changing code.
  3. Outputs and Logs: Clear separation for logs, metrics, and checkpoints for better experiment tracking.

This simplified structure ensures flexibility, readability, and ease of use for development and experimentation.