refactor(ai)

refactor(ai)

refactor(ai)

adjust ai

config.txt

@@ -1,7 +1,7 @@
 [NEAT]
-fitness_criterion     = max
-fitness_threshold     = 10
-pop_size              = 50
+fitness_criterion     = mean
+fitness_threshold     = 100000
+pop_size              = 100
 reset_on_extinction   = False
 
 [DefaultGenome]
@@ -44,9 +44,9 @@ node_add_prob           = 0.2
 node_delete_prob        = 0.2
 
 # network parameters
-num_hidden              = 2
+num_hidden              = 5
 num_inputs              = 200
-num_outputs             = 5
+num_outputs             = 6
 
 # node response options
 response_init_mean      = 1.0

main.py

@@ -6,13 +6,10 @@ from utils import BASE_PATH, CONFIG
 
 
 def pos_int(string: str) -> int:
-    try:
-        value = int(string)
-    except ValueError:
-        raise argparse.ArgumentTypeError(f"Expected integer, got {string!r}")
-    if value < 0:
-        raise argparse.ArgumentTypeError(f"Expected non negative number, got {value}")
-    return value
+    ivalue = int(string)
+    if ivalue <= 0:
+        raise argparse.ArgumentTypeError(f"{ivalue} is not a positive integer")
+    return ivalue
 
 
 parser = argparse.ArgumentParser(description="Tetris game with AI")
@@ -36,9 +33,8 @@ parser.add_argument(
     "-t",
     "--train",
     type=pos_int,
-    nargs="?",
-    const=100,
-    metavar="int",
+    nargs=2,
+    metavar=("generations", "parallels"),
     help="Trains the AI",
 )
 
@@ -70,7 +66,7 @@ def main(args: argparse.ArgumentParser) -> None:
 
     if args.train is not None:
         ai.log.debug("Training the AI")
-        ai.train(args.train)
+        ai.train(*args.train)
     else:
         game.log.debug("Running the game")
         game.Main().run()

src/ai/evaluation.py

@@ -1,33 +0,0 @@
-import neat
-from game import Main
-
-from .fitness import calculate_fitness
-from .log import log
-
-
-def eval_genomes(genomes, config: neat.Config) -> None:
-    app = Main()
-    app.run()
-    for genome_id, genome in genomes:
-        genome.fitness = calculate_fitness(app)
-        net = neat.nn.FeedForwardNetwork.create(genome, config)
-        while not app.game.game_over():
-            output = net.activate(app.game.field)
-
-            decision = output.index(max(output))
-
-            decisions = {
-                0: app.game.move_left,
-                1: app.game.move_right,
-                2: app.game.rotate,
-                3: app.game.rotate_reverse,
-                4: app.game.drop,
-            }
-
-            decisions[decision]()
-
-        genome.fitness = calculate_fitness(app)
-        log.info(
-            f"{genome_id=}\t{genome.fitness=}\t{app.game.score=}\t{app.game.lines=}\t{app.game.level=}"
-        )
-        app.game.restart()

src/ai/evaluations.py

@@ -0,0 +1,73 @@
+import math
+import time
+
+import neat
+import numpy as np
+import pygame
+from game import Block, Main
+from utils import CONFIG
+
+from .fitness import calculate_fitness
+from .log import log
+
+
+def eval_genome(genome: neat.DefaultGenome, config: neat.Config) -> float:
+    app = Main()
+
+    app.mute()
+    game = app.game
+    net = neat.nn.FeedForwardNetwork.create(genome, config)
+    genome.fitness = 0
+    moves = 0
+
+    while not game.game_over:
+        current_figure: list[int] = [
+            component for vec in game.tetromino.figure.value.shape for component in vec
+        ]
+
+        current_figure_pos: list[int] = [
+            component
+            for block in game.tetromino.blocks
+            for component in (int(block.pos.x), int(block.pos.y))
+        ]
+
+        next_figures: list[int] = [
+            component
+            for figure in app.next_figures
+            for pos in figure.value.shape
+            for component in pos
+        ]
+
+        field: np.ndarray = np.zeros((CONFIG.game.rows, CONFIG.game.columns), dtype=int)
+
+        block: Block
+        for block in game.sprites:
+            field[int(block.pos.y), int(block.pos.x)] = 1
+
+        output = net.activate(
+            # (*current_figure, *current_figure_pos, *next_figures, *field.flatten())
+            field.flatten()
+        )
+
+        decision = output.index(max(output))
+
+        decisions = {
+            0: game.move_left,
+            1: game.move_right,
+            2: game.move_down,
+            3: game.rotate,
+            4: game.rotate_reverse,
+            5: game.drop,
+        }
+
+        decisions[decision]()
+        app.run_game_loop()
+        moves += 1
+
+    genome.fitness = calculate_fitness(game, field) - moves / 10
+    score, lines, level = app.game.score, app.game.lines, app.game.level
+
+    log.debug(f"{genome.fitness=:<+6.6}\t{score=:<6} {lines=:<6} {level=:<6}")
+
+    game.restart()
+    return genome.fitness

src/ai/fitness.py

@@ -1,5 +1,33 @@
+from typing import Optional
+
 import neat
+import numpy as np
+from game import Game
+from utils import CONFIG
+
+from .log import log
 
 
-def calculate_fitness(app) -> float | int:
-    return app.game.score + app.game.lines * 100 + app.game.level * 1000
+def calculate_fitness(game: Game, field: Optional[np.ndarray] = None) -> float:
+    line_values = _calc_line_values(field)
+    return game.score * 10.0 + line_values
+
+
+def _calc_line_values(field: Optional[np.ndarray]) -> int:
+    if field is None:
+        return 0
+
+    line_values = 0
+    for idx, line in enumerate(np.flipud(field), start=1):
+        if idx <= 4:
+            line_values += int(line.sum()) * 5
+        elif idx <= 8:
+            line_values += int(line.sum()) * 3
+        elif idx <= 12:
+            line_values += int(line.sum()) * 0
+        elif idx <= 16:
+            line_values += int(line.sum()) * -5
+        else:
+            line_values += int(line.sum()) * -10
+
+    return line_values

src/ai/training.py

@@ -1,20 +1,47 @@
+import time
+
 import neat
+import pygame
+from game import Main
 from utils import BASE_PATH
 
 from .config import get_config
-from .evaluations import eval_genomes
+from .evaluations import eval_genome
 from .io import save_genome
 from .log import log
+from .visualize import plot_progress, plot_species, plot_stats
 
 
-def train(generations: int) -> None:
-    """Train the AI"""
+def train(gen_count: int, parallel: int = 1) -> None:
+    """
+    Train the AI
+    Args:
+        gen_count: Number of generations to train.
+        threads: Number of threads to use (default is 1).
+    """
     config = get_config()
-    population = neat.Population(config)
+    chekpoint_path = BASE_PATH / "checkpoints"
+    plots_path = BASE_PATH / "plots"
+
+    population = neat.Checkpointer().restore_checkpoint(
+        BASE_PATH / "checkpoints" / "neat-checkpoint-44"
+    )
+    # population = neat.Population(config)
     population.add_reporter(neat.StdOutReporter(True))
-    population.add_reporter(neat.StatisticsReporter())
+    stats = neat.StatisticsReporter()
+    population.add_reporter(stats)
     population.add_reporter(neat.Checkpointer(5, 900))
-    winner = population.run(eval_genomes, generations)
+
+    pe = neat.ParallelEvaluator(parallel, eval_genome)
+
+    winner = population.run(pe.evaluate, gen_count)
+    plot_stats(
+        stats,
+        ylog=False,
+        view=False,
+        filename=plots_path / "avg_fitness.svg",
+    )
+    plot_species(stats, view=False, filename=plots_path / "speciation.svg")
 
     log.info("Saving best genome")
     save_genome(winner)

src/game/__init__.py

@@ -1,6 +1,20 @@
+from .block import Block
 from .game import Game
 from .log import log
 from .main import Main
+from .preview import Preview
 from .score import Score
+from .tetromino import Tetromino
+from .timer import Timer, Timers
 
-__all__ = ["log", "Main", "Game", "Score"]
+__all__ = [
+    "log",
+    "Main",
+    "Game",
+    "Score",
+    "Block",
+    "Tetromino",
+    "Preview",
+    "Timer",
+    "Timers",
+]

src/game/game.py

@@ -35,6 +35,7 @@ class Game:
         level: Current game level.
         score: Current game score.
         lines: Number of lines cleared.
+        game_over: True if the game is over, False otherwise.
         landing_sound: Sound effect for landing blocks.
     """
 
@@ -112,10 +113,12 @@ class Game:
     def create_new_tetromino(self) -> None:
         """Create a new tetromino and perform necessary actions."""
         self._play_landing_sound()
-        if self.game_over():
-            self.restart()
-
         self._check_finished_rows()
+
+        self.game_over = self._check_game_over()
+        # if self.game_over:
+        #     self.restart()
+
         self.tetromino = Tetromino(
             self.sprites,
             self.create_new_tetromino,
@@ -123,7 +126,7 @@ class Game:
             self.get_next_figure(),
         )
 
-    def game_over(self) -> bool:
+    def _check_game_over(self) -> bool:
         """
         Check if the game is over.
 
@@ -131,16 +134,17 @@ class Game:
             True if the game is over, False otherwise.
         """
         for block in self.tetromino.blocks:
-            if block.pos.y < 0:
-                log.info("Game over!")
+            if block.pos.y <= 0:
+                # log.info("Game over!")
                 return True
         return False
 
     def restart(self) -> None:
         """Restart the game."""
-        log.info("Restarting the game")
+        # log.info("Restarting the game")
         self._reset_game_state()
         self._initialize_field_and_tetromino()
+        self.game_over = False
 
     def mute(self) -> None:
         """Mute the game."""
@@ -285,6 +289,7 @@ class Game:
         self.level: int = 1
         self.score: int = 0
         self.lines: int = 0
+        self.game_over: bool = False
 
     def _initialize_sound(self) -> None:
         """Initialize game sounds."""
@@ -368,7 +373,6 @@ class Game:
 
     def _reset_game_state(self) -> None:
         """Reset the game state."""
-        log.debug("Resetting game state")
         self.sprites.empty()
         self._initialize_field_and_tetromino()
         self._initialize_game_state()

src/game/main.py

@@ -26,7 +26,7 @@ class Main:
     """
 
     def __init__(self) -> None:
-        log.info("Initializing the game")
+        # log.info("Initializing the game")
         self._initialize_pygeme()
         self._initialize_game_components()
         self._start_background_music()
@@ -38,7 +38,19 @@ class Main:
     def run(self) -> None:
         """Run the main game loop."""
         while True:
-            self._run_game_loop()
+            self.run_game_loop()
+
+    def run_game_loop(self) -> None:
+        """Run a single iteration of the game loop."""
+        self.draw()
+        self.handle_events()
+
+        self.game.run()
+        self.score.run()
+        self.preview.run(self.next_figures)
+
+        pygame.display.update()
+        self.clock.tick(CONFIG.fps)
 
     def handle_events(self) -> None:
         """Handle Pygame events."""
@@ -103,7 +115,7 @@ class Main:
 
     def _initialize_game_components(self) -> None:
         """Initialize game-related components."""
-        self.next_figures = self._generate_next_figures()
+        self.next_figures: list[Figure] = self._generate_next_figures()
 
         self.game = Game(self._get_next_figure, self._update_score)
         self.score = Score()
@@ -114,15 +126,3 @@ class Main:
         self.music = pygame.mixer.Sound(CONFIG.music.background)
         self.music.set_volume(CONFIG.music.volume)
         self.music.play(-1)
-
-    def _run_game_loop(self) -> None:
-        """Run a single iteration of the game loop."""
-        self.draw()
-        self.handle_events()
-
-        self.game.run()
-        self.score.run()
-        self.preview.run(self.next_figures)
-
-        pygame.display.update()
-        self.clock.tick(CONFIG.fps)

src/game/tetromino.py

@@ -151,7 +151,7 @@ class Tetromino:
         """
         return all(
             0 <= pos.x < CONFIG.game.columns
-            and 0 <= pos.y <= CONFIG.game.rows
+            and 0 <= pos.y < CONFIG.game.rows
             and not self.field[int(pos.y), int(pos.x)]
             for pos in new_positions
         )

src/utils/config.py

@@ -21,7 +21,7 @@ class Game:
     size: Size = Size(columns * cell.width, rows * cell.width)
     pos: Vec2 = Vec2(padding, padding)
     offset: Vec2 = Vec2(columns // 2, -1)
-    initial_speed: float | int = 400
+    initial_speed: float | int = 50
     movment_delay: int = 200
     rotation_delay: int = 200
     score: dict[int, int] = {1: 40, 2: 100, 3: 300, 4: 1200}