def _playAI(self) -> None:
"""User selects saved model from .../dna/trained. Opens GUI window and AI plays game."""
# check npz format with same size networks or something like that? Cant save certain things...??
algoIndex, algoChoice = ui.getSelection("Neural Network",
"Multi",
"Hierarchical",
"Cycle",
"Pathfinding",
"Floodfill",
"Back",
msg="Select AI algorithm:")
if algoChoice != "Back" and (snake := self._makeSnake(
algoIndex, algoChoice)) is not None:
numGames = ui.getValidInput(
"How many games should be played? (select one game to render)",
dtype=int,
lower=1,
end="\n")
scores = []
timer = time()
if algoChoice in {"Multi", "Hierarchical"}:
algos = snake.behavior.algoUsage.keys()
avgUsage = {algo: 0 for algo in algos}
for i in range(numGames):
gameEnvironment = environment.Environment(snake,
settings.mapSize,
noise=settings.noise)
#gameEnvironment = environment.Environment(snake, settings.mapSize, origin=(3, 0)) # for cycle to win or get close when odd
games.playGame(gameEnvironment, render=(not (numGames - 1)))
scores.append(snake.size)
if algoChoice in {"Multi", "Hierarchical"}:
for algo in algos:
avgUsage[algo] += snake.behavior.algoUsage[algo]
print("Game", str(i + 1) + " snake size:", snake.size)
elapsed = time() - timer
if not numGames - 1:
print()
ui.checkSave(gameEnvironment, self._saveGame)
else:
print("\nTime elapsed across",
str(numGames) + " games:", round(elapsed, 5), "secs")
print("Average time per game", round(elapsed / numGames, 5),
"secs")
print("Average snake score:", round(sum(scores) / numGames, 2))
print("Scores std:", round(np.std(scores), 3))
if algoChoice in {"Multi", "Hierarchical"}:
avgUsage = {
algo: round(avgUsage[algo] / numGames, 3)
for algo in algos
}
print(
"Snake average algorithm use (n=" + str(numGames) +
"):", avgUsage)
print()
self.prevGameEnvironment = gameEnvironment
def _replay(self, moves: list, origin: tuple, food: list, mapSize: tuple,
color: tuple) -> None:
"""
Constructs snake and environment from data and replays it in GUI window.
Parameters
----------
moves: list
Queue of directions snake chose to make at each time step
origin: tuple
Initial starting position of snake
food: list
Queue of food spawning positions
mapSize: tuple
(x, y) size of game map snake played on
color: tuple
Color of snake
"""
snakeKwargs = {
"initialSize": settings.initialSnakeSize,
"maxVision": settings.maxSnakeVision,
"hungerFunc": settings.hungerFunc,
}
snake = snakes.Snake("ghost",
behaviorArgs=[moves],
**snakeKwargs,
color=color)
gameEnvironment = environment.Environment(snake,
mapSize,
origin=origin,
food=food)
games.playGame(gameEnvironment)
ui.checkSave(gameEnvironment, self._saveGame)
def _playClassic(self) -> None:
"""Opens GUI window and lets user play Snake with keyboard controls."""
snake = snakes.Snake.Player()
gameEnvironment = environment.Environment(snake, settings.mapSize)
print("Get ready...")
games.playPlayerGame(gameEnvironment)
finalScore = snake.score
print("Final score: " + str(finalScore))
print()
ui.checkSave(gameEnvironment, self._saveGame)
self.prevGameEnvironment = gameEnvironment
def playTrainingGame(snake: snakes.Snake, render: bool = False) -> dict:
"""
Plays game with settings.mapSize sized map.
Parameters
----------
snake: snakes.Snake
Snake to play game with
render: bool, default=True
Indicates if game should be rendered to GUI window
"""
#print("NEW GAME" + "!" * 20) # delete
# DONT MAKE NEW ENVIRONMENT EVERY TIME, MAKE ENVIRONMENT RESET
gameEnvironment = environment.Environment(snake, settings.mapSize)
playGame(gameEnvironment, render=render)
id = snake.id
return {"fitness": snake.fitness(snake), "score": snake.score, "id": id}
def main():
algoIndex, algoChoice = ui.getSelection("Neural Network",
"Multi",
"Hierarchical",
"Cycle",
"Pathfinding",
"Floodfill",
msg="Select AI algorithm:")
if (snake := makeSnake(algoIndex, algoChoice)) is not None:
#numGames = ui.getValidInput("How many games should be played?", dtype=int, lower=1, end="\n")
S = [] # true states
P = [] # perturbed states
scores = []
timer = time()
if algoChoice in {"Multi", "Hierarchical"}:
algos = snake.behavior.algoUsage.keys()
avgUsage = {algo: 0 for algo in algos}
numGames = 1 # hard coded in for now
for i in range(numGames):
gameEnvironment = environment.Environment(snake,
settings.mapSize,
noise=settings.noise)
while gameEnvironment.active():
gameEnvironment.step()
S.append(gameEnvironment.gameMap.copy()) # adding true state
scores.append(snake.size)
if algoChoice in {"Multi", "Hierarchical"}:
for algo in algos:
avgUsage[algo] += snake.behavior.algoUsage[algo]
print("Game", str(i + 1) + " snake size:", snake.size)
elapsed = time() - timer
print("\nTime elapsed across",
str(numGames) + " games:", round(elapsed, 5), "secs")
print("Average time per game", round(elapsed / numGames, 5), "secs")
print("Average snake score:", round(sum(scores) / numGames, 2))
print("Scores std:", round(np.std(scores), 3))
if algoChoice in {"Multi", "Hierarchical"}:
avgUsage = {
algo: round(avgUsage[algo] / numGames, 3)
for algo in algos
}
print("Snake average algorithm use (n=" + str(numGames) + "):",
avgUsage)
print()
P = [perturb(s) for s in S] # independently perturb each true state
radius = 5 # threshold for which to search for solutions
solutions = solve(P, radius) # get a list of solutions within radius
if S in solutions:
print("Converged!")
elif len(solutions) == 0:
print("Failed to find any solutions")
else:
print("Failed to converge to true trajectory")
print()