config.py

"""
Configuration and constants for the Pong AI project.
"""

from pathlib import Path

# Project directories
PROJECT_ROOT = Path(__file__).parent.resolve()
MODELS_DIR = PROJECT_ROOT / "models"
LOGS_DIR = PROJECT_ROOT / "pong_tensorboard"
DEMOS_DIR = PROJECT_ROOT / "demo"
TRAIN_DIR = PROJECT_ROOT / "train"

# Game configuration
GAME_CONFIG = {
    'screen_width': 800,
    'screen_height': 600,
    'paddle_width': 15,
    'paddle_height': 100,
    'ball_size': 10,
    'paddle_speed': 6.0,
    'ball_speed': 5.0,
    'max_ball_speed': 8.0,
    'fps': 60,
}

# RL Training configuration
TRAINING_CONFIG = {
    'total_timesteps': 1_000_000,
    'num_envs': 4,
    'use_multiprocessing': True,
    'checkpoint_interval': 50_000,
    'learning_rate': 3e-4,
    'n_steps': 2048,
    'batch_size': 64,
    'n_epochs': 10,
    'gamma': 0.99,
    'gae_lambda': 0.95,
    'clip_range': 0.2,
    'ent_coef': 0.01,
}

# Model checkpoints available in demo
DEMO_LEVELS = {
    '1': {
        'path': 'models/rl_model_50000_steps.zip',
        'name': 'Novice (50k)',
        'description': 'Basic rally pattern learning',
    },
    '2': {
        'path': 'models/rl_model_200000_steps.zip',
        'name': 'Intermediate (200k)',
        'description': 'Paddle positioning and defense',
    },
    '3': {
        'path': 'models/rl_model_500000_steps.zip',
        'name': 'Advanced (500k)',
        'description': 'Advanced spike and positioning',
    },
    '4': {
        'path': 'models/rl_model_1000000_steps.zip',
        'name': 'Master (1M)',
        'description': 'Master-level near-optimal play',
    },
}

# API configuration
API_CONFIG = {
    'host': '127.0.0.1',
    'port': 8000,
    'reload': True,
}

# Reward configuration
REWARD_CONFIG = {
    'ai_scores': 1.0,
    'opponent_scores': -1.0,
    'paddle_hit': 0.1,
    'step_penalty': -0.001,
}

# Observation normalization
OBSERVATION_CONFIG = {
    'ball_velocity_scale': 10.0,  # For normalization
    'min_value': -1.0,
    'max_value': 1.0,
}