momentum = args.momentum
# Create the game specs.
game_specs = GameResultSpecs(info_interval_current, info_interval_mean,
agent_save_interval, results_save_interval,
plots_name_prefix, results_name_prefix,
agent_name_prefix, recording_name_prefix,
plot_train_results, save_plots)
# Create the game.
game = Game(episodes, downsample_scale, agent_frame_history,
steps_per_action, fit_frequency, no_operation, game_specs,
render, record)
# Create the optimizer.
optimizer = initialize_optimizer(optimizer_name, learning_rate, beta1,
beta2, lr_decay, rho, fuzz, momentum)
# Create the policy.
policy = EGreedyPolicy(epsilon, final_epsilon, epsilon_decay,
total_observe_count, game.action_space_size)
# Create the agent.
agent = create_agent()
# Check arguments.
run_checks()
# Play the game, using the agent.
game.play_game(agent)
from game_engine.game import Game
from game_engine.player import RandomPlayer
import time
import sys
sys.path.append('../')
games = 5000
players = [RandomPlayer() for _ in range(4)]
initial = time.perf_counter()
last = initial
for i in range(games):
if i % 100 == 0:
print("{}/{} time: {}".format(i, games, time.perf_counter() - last))
last = time.perf_counter()
wiz = Game(players=players)
wiz.play_game()
print(time.perf_counter() - initial)
