def run(self):
fn = backup.export_revision(gui.uuid, gui.host, gui.path,
rev, target_dir)
util.open_file(fn)
gtk.gdk.threads_enter()
running_tasks_model.remove(i)
gtk.gdk.threads_leave()
def run(self):
fn = backup.export_revision( gui.uuid, gui.host, gui.path, rev, target_dir )
os.chdir(target_dir)
os.system('tar -zxvf "%s"' % fn)
os.remove(fn)
os.chdir(util.RUN_FROM_DIR)
util.open_file(target_dir)
gtk.gdk.threads_enter()
running_tasks_model.remove(i)
gtk.gdk.threads_leave()
def _link_clicked(browser, frame, request,
action, decision, *args, **kwargs):
if action.get_reason() == WEBKIT_WEB_NAVIGATION_REASON_OTHER:
# let this load
pass
else:
# open file in system
util.open_file(action.get_original_uri())
# ignore webkit request
decision.ignore()
def run(self):
fn = backup.export_revision(gui.uuid, gui.host, gui.path, rev,
target_dir)
os.chdir(target_dir)
os.system('tar -zxvf "%s"' % fn)
os.remove(fn)
os.chdir(util.RUN_FROM_DIR)
util.open_file(target_dir)
gtk.gdk.threads_enter()
running_tasks_model.remove(i)
gtk.gdk.threads_leave()
def load(filename):
global datafile
datafile = util.open_file(os.path.join(workingdir, filename))
core.clean_all(datafile)
return "loaded " + filename
def extract_features(assignment_id, buckets):
file_path = '/'.join(['./data/files/1222', assignment_id, 'Breakout.java'])
file = util.open_file(file_path)
features = []
for bucket in buckets:
features += get_features(file, bucket)
return features
def extract_features(assignment_id, bucket, report=None):
file_path = '/'.join(['./data/files', assignment_id, 'Breakout.java'])
file = util.open_file(file_path)
file_lines = [l for l in file]
if bucket == 'Decomposition':
return decomposition_features(file_lines, report)
elif bucket == 'Commenting':
return commenting_features(file_lines, report)
elif bucket == 'Naming and Spacing':
return naming_and_spacing_features(file_lines, report)
elif bucket == 'Instance Variables and Parameters and Constants':
return variable_features(file_lines, report)
elif bucket == 'Logic and Redundancy':
return logic_redundancy_features(file_lines, report)
else:
# TODO: implement feature extraction for other buckets.
print('Can\'t read that bucket yet :/')
return []
sess = tf.Session(config=config)
init = tf.initialize_all_variables()
sess.run(init)
saver = tf.train.Saver() # save all variables
checkpoint_dir = model_dir
checkpoint_file = 'segm.ckpt'
if validation_path:
validation_data = util.get_validation_data(validation_path, char_dic,
vocab_size, n_steps, padd)
seq = 0
while seq < training_iters:
c_istate = np.zeros((batch_size, 2 * n_hidden))
i = 0
fid = util.open_file(train_path, 'r')
for line in fid:
line = line.strip()
if line == "": continue
line = line.decode('utf-8')
sentence = util.snorm(line)
pos = 0
while pos != -1:
batch_xs, batch_ys, next_pos, count = util.next_batch(
sentence, pos, char_dic, vocab_size, n_steps, padd)
'''
print 'window : ' + sentence[pos:pos+n_steps].encode('utf-8')
print 'count : ' + str(count)
print 'next_pos : ' + str(next_pos)
print batch_ys
print batch_xs
def run(self): fn = backup.export_revision( gui.uuid, gui.host, gui.path, rev, target_dir ) util.open_file(fn) gtk.gdk.threads_enter() running_tasks_model.remove(i) gtk.gdk.threads_leave()
def run_games(first_agent,
second_agent,
first_agent_turn,
num_games,
update_param=0,
quiet=False,
first_file_name="./data/first_save",
second_file_name="./data/second_save",
first_weights_file_name="./data/first_weights",
second_weights_file_name="./data/second_weights",
first_result_file_name="./data/first_results",
second_result_file_name="./data/second_results",
first_m_result_file_name="./data/first_m_results",
second_m_result_file_name="./data/second_m_results",
play_against_self=False):
"""
first_agent: instance of Agent which reflects first agent
second_agent: instance of Agent which reflects second agent
first_agent_turn: True if turn is of the first agent
num_games: total number of games to run without training
num_training: total number of training games to run
"""
try:
write_str = "num_moves,win,reward,max_q_value\n"
if first_agent.is_learning_agent:
first_f = open_file(first_file_name, header=write_str)
first_w_deq = deque()
first_f_res = open_file(first_result_file_name)
first_writer_res = csv.writer(first_f_res, lineterminator='\n')
first_f_m_res = open_file(first_m_result_file_name)
first_writer_m_res = csv.writer(first_f_m_res, lineterminator='\n')
first_f_str = ""
first_writer_w_list = []
if second_agent.is_learning_agent:
second_f = open_file(second_file_name, header=write_str)
second_w_deq = deque()
second_f_res = open_file(second_result_file_name)
second_writer_res = csv.writer(second_f_res, lineterminator='\n')
second_f_m_res = open_file(second_m_result_file_name)
second_writer_m_res = csv.writer(second_f_m_res,
lineterminator='\n')
second_f_str = ""
second_writer_w_list = []
# learn weights
# save weights
# test using weights
# change agent
print('starting game', 0)
for i in range(num_games):
if (i + 1) % NOTIFY_FREQ == 0:
print('Starting game', (i + 1))
rules = ClassicGameRules()
if first_agent.has_been_learning_agent:
first_agent.start_learning()
if second_agent.has_been_learning_agent:
second_agent.start_learning()
game = rules.new_game(first_agent,
second_agent,
first_agent_turn,
quiet=quiet)
num_moves, game_state = game.run()
if first_agent.is_learning_agent:
reward = first_agent.episode_rewards
win = 1 if game_state.is_first_agent_win() else 0
init_state = GameState(the_player_turn=first_agent_turn)
max_q_value = first_agent.compute_value_from_q_values(
init_state)
w_str = str(num_moves) + "," + str(win) + "," + str(
reward) + "," + str(max_q_value) + "\n"
first_f_str += w_str
if (i + 1) % WEIGHTS_SAVE_FREQ == 0:
if len(first_w_deq
) != 0 and len(first_w_deq) % NUM_WEIGHTS_REM == 0:
first_w_deq.popleft()
first_w_deq.append(np.array(first_agent.weights))
if (i + 1) % WRITE_FREQ == 0:
first_f.write(first_f_str)
first_f_str = ""
if second_agent.is_learning_agent:
reward = second_agent.episode_rewards
win = 1 if game_state.is_second_agent_win() else 0
init_state = GameState(the_player_turn=first_agent_turn)
max_q_value = second_agent.compute_value_from_q_values(
init_state)
w_str = str(num_moves) + "," + str(win) + "," + str(
reward) + "," + str(max_q_value) + "\n"
second_f_str += w_str
if (i + 1) % WEIGHTS_SAVE_FREQ == 0:
if len(second_w_deq
) != 0 and len(second_w_deq) % NUM_WEIGHTS_REM == 0:
second_w_deq.popleft()
second_w_deq.append(np.array(second_agent.weights))
if (i + 1) % WRITE_FREQ == 0:
second_f.write(second_f_str)
second_f_str = ""
if (i + 1) % TEST_FREQ == 0:
if first_agent.is_learning_agent:
first_agent.stop_learning()
if second_agent.is_learning_agent:
second_agent.stop_learning()
result_f = []
result_s = []
print('strting', TEST_GAMES, 'tests')
result_f, result_s = \
multiprocess(rules, first_agent, second_agent, first_agent_turn, quiet=True)
if first_agent.has_been_learning_agent:
first_writer_res.writerow(result_f[0])
first_writer_m_res.writerow(result_f[1])
if second_agent.has_been_learning_agent:
second_writer_res.writerow(result_s[0])
second_writer_m_res.writerow(result_s[1])
if first_agent.has_been_learning_agent and play_against_self:
if (i + 1) % CHANGE_AGENT_FREQ == 0:
weights = first_w_deq[-1]
second_agent = QLearningAgent(weights=weights,
is_learning_agent=False)
if first_agent.has_been_learning_agent and update_param:
first_agent.update_parameters(update_param, (i + 1))
if second_agent.has_been_learning_agent and update_param:
second_agent.update_parameters(update_param, (i + 1))
#-------------------------------
#Printing result
if game_state.is_first_agent_win():
print(type(first_agent).__name__)
elif game_state.is_second_agent_win():
print(type(second_agent).__name__)
except Exception as e:
print(sys.exc_info()[0])
traceback.print_tb(e.__traceback__)
finally:
if first_agent.has_been_learning_agent:
first_f.close()
first_f_res.close()
first_f_m_res.close()
first_f_w = open_file(first_weights_file_name)
first_writer_w = csv.writer(first_f_w, lineterminator='\n')
first_writer_w.writerows(first_w_deq)
first_f_w.close()
if second_agent.has_been_learning_agent:
second_f.close()
second_f_res.close()
second_f_m_res.close()
second_f_w = open_file(second_weights_file_name)
second_writer_w = csv.writer(second_f_w, lineterminator='\n')
second_writer_w.writerows(second_w_deq)
second_f_w.close()
def open_folder(self, widget, file=''):
f = pathjoin(self._hub.config_manager.get_watchlist(), file)
util.open_file(f)
inter_op_parallelism_threads=NUM_THREADS,
log_device_placement=False)
sess = tf.Session(config=config)
init = tf.global_variables_initializer()
sess.run(init)
saver = tf.train.Saver() # save all variables
checkpoint_dir = model_dir
checkpoint_file = 'segm.ckpt'
validation_data = util.get_validation_data_emb(validation_path, char_dic, id2emb, n_steps, padd)
seq = 0
while seq < training_iters :
c_istate = np.zeros((batch_size, 2*n_hidden))
i = 0
fid = util.open_file(train_path, 'r')
for line in fid :
line = line.strip()
if line == "" : continue
line = line.decode('utf-8')
sentence = util.snorm(line)
pos = 0
while pos != -1 :
batch_xs, batch_ys, next_pos, count = util.next_batch_emb(sentence, pos, char_dic, id2emb, n_steps, padd)
'''
print 'window : ' + sentence[pos:pos+n_steps].encode('utf-8')
print 'count : ' + str(count)
print 'next_pos : ' + str(next_pos)
print batch_ys
print batch_xs
'''
