def assign_fitness(amount_of_games, population):
# print('Assigning fitness to the population')
agent_idx = population.get_fitless_agent_idx()
while agent_idx != -1:
GA_agent = population.pop[agent_idx]
# print('Calculating fitness for agent: ', agent_idx)
# print(GA_agent)
players = [LudoPlayerGenetic(GA_agent)] + [
StandardLudoPlayers.LudoPlayerRandom() for _ in range(3)
]
for id, player in enumerate(players):
player.id = id
n = amount_of_games
fitness = np.zeros(4, dtype=np.int)
# print(agent_idx)
for i in range(n):
np.random.shuffle(players)
ludoGame = LudoGame(players)
winner = ludoGame.play_full_game()
fitness[players[winner].id] += 1
population.fitness_add(fitness_score=fitness[0] / amount_of_games,
index_of_chromosomes=agent_idx)
agent_idx = population.get_fitless_agent_idx()
def normal_game(number_of_runs=100):
GA_player = population.pop[population.get_best_agent()] # population.pop[population.get_random_agent(1)[0]]
players = [LudoPlayerGenetic(GA_player)] + [StandardLudoPlayers.LudoPlayerRandom() for _ in range(3)]
for i, player in enumerate(players):
player.id = i
# print(player)
score = np.zeros(4, dtype=np.int)
n = number_of_runs
start_time = time.time()
for i in range(n):
'''
for idx in range(4):
if players[idx].id == 0:
players[idx].load_agent(population.pop[population.get_best_agent()])
'''
np.random.shuffle(players)
ludoGame = LudoGame(players)
winner = ludoGame.play_full_game()
score[players[winner].id] += 1
# print('Game ', i, ' done')
duration = time.time() - start_time
print('win distribution:', score / number_of_runs)
print('games per second:', n / duration)
def eval_population_worker(queue: mp.Queue, games_per_chromosome, task_counter_queue: mp.Queue, Opponent):
while True:
population_path = queue.get()
folder_path = os.path.dirname(population_path)
folder_name = os.path.basename(folder_path)
player_name = folder_name.split("+")[0]
Player = get_ga_player(player_name)
population = np.load(population_path)
N = min(len(population), 20)
population_idx = np.random.choice(np.arange(len(population)), N, replace=False)
population = population[population_idx]
save_matrix = np.empty((2, N), np.float)
save_matrix[0] = population_idx
scores = save_matrix[1]
for i, chromosome in enumerate(population):
players = [Player(chromosome)] + [Opponent() for _ in range(3)]
win_count = 0
for _ in range(games_per_chromosome):
random.shuffle(players)
game = LudoGame(players)
winner = players[game.play_full_game()]
if isinstance(winner, Player):
win_count += 1
scores[i] = win_count / games_per_chromosome
generation_str = os.path.basename(population_path).split(".")[0]
scores_path = get_score_file_path(folder_path, generation_str, Opponent.name)
assert not os.path.exists(scores_path), "Scores already exists: {}".format(scores_path)
np.save(scores_path, save_matrix)
task_counter_queue.put(('finished', population_path))
def tournament(chromosomes, Player, game_count):
players = [Player(chromosome) for chromosome in chromosomes]
while len(players) < 4:
players.append(LudoPlayerRandom())
tournament_player_ids = {}
for i, player in enumerate(players):
tournament_player_ids[player] = i
win_rates = np.zeros(4)
for _ in range(game_count):
random.shuffle(players)
game = LudoGame(players)
winner = players[game.play_full_game()]
win_rates[tournament_player_ids[winner]] += 1
ranked_player_ids = np.argsort(-win_rates)
for id in ranked_player_ids:
if id < len(chromosomes):
return id
def evaluate_agent(agent, amount_games, use_random=True):
if use_random:
rndPlayer = LudoPlayerRandom()
else:
rndPlayer = SemiSmartPlayer()
players = [agent, agent, rndPlayer, rndPlayer]
for i, player in enumerate(players):
player.id = i
wins = [0, 0, 0, 0]
for i in range(amount_games):
random.shuffle(players)
ludoGame = LudoGame(players)
winner = ludoGame.play_full_game()
wins[players[winner].id] += 1
return wins[1]
def tournament(_players, game_count):
progress_bar = ProgressBar(widgets=[Percentage()],
maxval=game_count).start()
players = [player for player in _players]
tournament_player_ids = {}
for i, player in enumerate(players):
tournament_player_ids[player] = i
win_rates = np.zeros(4)
for i in range(game_count):
random.shuffle(players)
game = LudoGame(players)
winner = players[game.play_full_game()]
win_rates[tournament_player_ids[winner]] += 1
progress_bar.update(i + 1)
progress_bar.finish()
return win_rates / game_count
def normal_play():
players = []
players = [LudoPlayerRandom() for _ in range(3)]
epsilon = 0.05 #
discount_factor = 0.5 #
learning_rate = 0.25 #
parameters = [epsilon, discount_factor, learning_rate]
t1 = LudoPlayerQLearningAction(parameters, chosenPolicy="greedy", QtableName='QTable', RewardName='Reward')
players.append(t1)
for i, player in enumerate(players):
player.id = i # selv tildele atributter uden defineret i klassen
score = [0, 0, 0, 0]
n = 1000
start_time = time.time()
tqdm_1 = tqdm(range(n), ascii=True)
for i in tqdm_1:
tqdm_1.set_description_str(f"win rates {np.around(score/np.sum(score),decimals=2)*100}")
random.shuffle(players)
ludoGame = LudoGame(players)
winner = ludoGame.play_full_game()
score[players[winner].id] += 1
for player in players: # Saving reward for QLearning player
if type(player)==LudoPlayerQLearningAction:
player.append_reward()
for player in players:
if type(player)==LudoPlayerQLearningAction:
player.saveQTable()
player.saveReward()
duration = time.time() - start_time
print('win distribution percentage', (score/np.sum(score))*100)
print('win distribution:', score)
def play_tournament(self, chromosome_ids, game_count):
flat_pop = self.get_flat_pop()
chromosomes = flat_pop[chromosome_ids]
players = [self.Player(chromosome) for chromosome in chromosomes]
tournament_player_ids = {}
for i, player in enumerate(players):
tournament_player_ids[player] = i
win_rates = np.zeros(4)
for _ in range(game_count):
random.shuffle(players)
game = LudoGame(players)
winner = players[game.play_full_game()]
win_rates[tournament_player_ids[winner]] += 1
ranked_chromosome_ids = chromosome_ids[np.argsort(-win_rates)]
children = self.recombine(*flat_pop[ranked_chromosome_ids[:2]])
children = [self.mutate(child) for child in children]
children = [self.Player.normalize(child) for child in children]
flat_pop[ranked_chromosome_ids[2:]] = children
self.total_game_count += game_count
self.cur_tournament_count += 1
self.progress_bar.update(self.cur_tournament_count)
def main():
global GA
GA1 = GAPlayer()
GA2 = GAPlayer()
randomAgent = LudoPlayerRandom()
print(GA.getPopulation(0))
for t in range(0, 100):
print("Test:", t)
players = [GA1, GA1, GA2, GA2]
for i, player in enumerate(players):
player.id = i
GA1.index = 0
GA2.index = 1
score = [0, 0, 0, 0]
for _ in tqdm(range(1000)):
random.shuffle(players)
ludoGame = LudoGame(players)
winner = ludoGame.play_full_game()
score[players[winner].id] += 1
print("Wins: ", score)
from pyludo import LudoGame
from pyludo.StandardLudoPlayers import LudoPlayerRandom, LudoPlayerFast, LudoPlayerAggressive, LudoPlayerDefensive
import random
import time
players = [
LudoPlayerRandom(),
LudoPlayerFast(),
LudoPlayerAggressive(),
LudoPlayerDefensive(),
]
scores = {}
for player in players:
scores[player.name] = 0
n = 1000
start_time = time.time()
for i in range(n):
random.shuffle(players)
ludoGame = LudoGame(players)
winner = ludoGame.play_full_game()
scores[players[winner].name] += 1
print('Game ', i, ' done')
duration = time.time() - start_time
print('win distribution:', scores)
print('games per second:', n / duration)
def evaluate_agents(agents):
players = agents
ludoGame = LudoGame(players)
return ludoGame.play_full_game()
def main():
global GA
GA1 = GAPlayer()
GA2 = GAPlayer()
GA3 = GAPlayer()
GA4 = GAPlayer()
randomAgent = LudoPlayerRandom()
n = 50
generations = 100
for g in range(0, generations):
print("Generation:", g)
GA.exportPopulation()
players = [GA1, GA2, GA3, GA4]
for i, player in enumerate(players):
player.id = i
a = np.zeros(pops)
a += n
for _ in tqdm(range(int(n * pops * 0.25))):
try:
idx = np.random.choice(pops,
4,
replace=False,
p=np.divide(a, np.sum(a)))
except:
break
a[idx] -= 1
GA1.index = idx[0]
GA2.index = idx[1]
GA3.index = idx[2]
GA4.index = idx[3]
#print(a)
ludoGame = LudoGame(players)
winner = ludoGame.play_full_game()
GA.populationScores[players[winner].index] += 1
print(a)
print(GA.populationScores)
GA.populationScores = np.square(GA.populationScores)
players = [GA1, randomAgent, randomAgent, randomAgent]
for i, player in enumerate(players):
player.id = i
if True:
score = [0, 0, 0, 0]
print(np.argmax(GA.populationScores))
GA1.index = np.argmax(GA.populationScores)
for i in tqdm(range(1000)):
random.shuffle(players)
ludoGame = LudoGame(players)
winner = ludoGame.play_full_game()
score[players[winner].id] += 1
GA.exportWeigths(GA1.index, score[0])
print('win distribution:', score)
f = open("output.txt", "a+")
f.write(str(score[0]) + "\n")
f.close()
print("\t*Mutation*\t")
GA.mutate()