def loadGame(param_dir):
gs = dict()
with open(param_dir, 'rb') as f:
shimon_hero_params = pickle.load(f)
for paramater in shimon_hero_params:
gs[paramater] = shimon_hero_params[paramater]
game = sh.Game(gs)
return game
EXPLORE = 3000000. # frames over which to anneal epsilon
FINAL_EPSILON = 0.0001 # final value of epsilon
INITIAL_EPSILON = 0.1 # starting value of epsilon
REPLAY_MEMORY = 50000 # number of previous transitions to remember
BATCH = 32 # size of minibatch
FRAME_PER_ACTION = 3 # This controls how many frames to wait before deciding on an action. If F_P_A = 1, then Shimon
# chooses a new action every tick, which causes erratic movements with no exploration middle spaces
# 1NN controlling 2 arms means you need to enumerate all the controls
action_dict = {0: -1, 1: 0, 2: 1}
img_rows, img_cols = 80, 80 # All images are downsampled to 80 x 80
img_channels = 4 # Stack 4 frames to infer movement
# Initialize instance of Shimon Hero game
game = sh.Game() # Instantiate Shimon Hero game
timestr = time.strftime(
"%m-%d_%H-%M-%S") # save the current time to name the model
model_prefix = "2A1NN_" + timestr # The prefix used to identify the model and time training was created
# Save the shimon_hero paramters corresponding to the model
shimon_hero_param = game.get_settings()
if SAVE_MODEL:
param_path = "../saved_models/" + model_prefix + "_param.p"
with open(param_path, 'wb') as f:
pickle.dump(shimon_hero_param, f, pickle.HIGHEST_PROTOCOL)
def buildmodel():
# To make a model with shared weights, need to use Keras Functional API, not Sequential API
# Build the model using the same specifications as the DeepMind paper
pass # print (time.time() - last_time)
fr = 40. # frame rate
s_per_frame = (1. / fr)
gs = dict()
with open('../saved_models/model_03-15_22-24-56_param.p', 'rb') as f:
# The protocol version used is detected automatically, so we do not
# have to specify it.
shimon_hero_params = pickle.load(f)
for paramater in shimon_hero_params:
gs[paramater] = shimon_hero_params[paramater]
game = sh.Game(gs)
print(game.get_settings())
game_over = False
key_action = [0, 0, 0, 0]
last_time = time.time()
while not game_over:
# type: (object) -> object
for event in pygame.event.get():
if event.type == pygame.KEYDOWN:
#print(event.key)
if event.key == 49: # 1 key down
key_action[0] = -1
elif event.key == 50: # 2 key down
key_action[0] = 1
elif event.key == 113: # q key down
import sched
import time
from shimon_hero import shimon_hero as sh
s = sched.scheduler(time.time, time.sleep)
def print_time(last_time):
pass # print (time.time() - last_time)
fr = 40. # frame rate
s_per_frame = (1. / fr)
game = sh.Game()
print(game.get_settings())
game_over = False
key_action = [0, 0, 0, 0]
last_time = time.time()
while not game_over:
# type: (object) -> object
for event in pygame.event.get():
if event.type == pygame.KEYDOWN:
#print(event.key)
if event.key == 49: # 1 key down
key_action[0] = -1
elif event.key == 50: # 2 key down
key_action[0] = 1
elif event.key == 113: # q key down
def main():
model_dir, param_dir = loadModelPrefix()
model = loadModel(model_dir) # load model weights
#game = loadGame(param_dir) # load corresponding settings in shimon when model was trained
game = sh.Game()
playGame(model, game)
from shimon_hero import shimon_hero as sh
# Initialize instance of Shimon Hero game
game = sh.Game(human_playing=False)
NUM_ACTIONS = 4
FRAME_PER_ACTION = 1
ACTION_HOLD = 6
import numpy as np
import random
a_t = np.zeros(NUM_ACTIONS)
a_t[0] = 1
t = 0
while True:
'''
if t % ACTION_HOLD == 0: # We need to give a chance for the agent to "hold" the z key over more than 1 frame
print "--- NEW ACTION ---"
a_t = np.zeros(NUM_ACTIONS)
action_index = random.randrange(NUM_ACTIONS)
a_t[action_index] = 1 # set action command with correct one-hot encoding
'''
if t % FRAME_PER_ACTION == 0: # We need to give a chance for the agent to "hold" the z key over more than 1 frame
#print "--- NEW ACTION ---"
a_t = np.zeros(NUM_ACTIONS)
action_index = random.randrange(NUM_ACTIONS)
a_t[action_index] = 1 # set action command with correct one-hot encoding
#print a_t
image_data, r_t, game_over = game.next_action(a_t)
print("r_t: ", r_t)