def end_file(self):
self.batch_process()
if self.file_frame_count == 0:
return
per_frame_award = self.file_reward / float(self.file_frame_count)
if not self.current_file in self.eval_compare:
self.eval_compare[self.current_file] = []
self.eval_compare[self.current_file].append(per_frame_award)
print('Reward for file:', self.file_reward)
print('Reward per frame:', per_frame_award)
self.file_reward = 0
self.file_frame_count = 0
self.reward_manager = reward_manager.RewardManager()
def __init__(self, name, team, index, bot_parameters=None):
self.last_frame_time = None
self.config_file = bot_parameters
self.index = index
self.load_config_file()
self.reward_manager = reward_manager.RewardManager()
config = tf.ConfigProto(device_count={'GPU': 0})
self.sess = tf.Session(config=config)
# self.sess = tf.Session()
self.actions_handler = action_factory.get_handler(
control_scheme=self.control_scheme)
self.num_actions = self.actions_handler.get_logit_size()
print('num_actions', self.num_actions)
self.model = self.model_class(self.sess,
self.num_actions,
input_formatter_info=[team, index],
player_index=self.index,
action_handler=self.actions_handler,
config_file=bot_parameters,
is_training=False)
self.model.add_summary_writer('random_packet', is_replay=True)
self.model.batch_size = 1
self.model.mini_batch_size = 1
self.model.is_graphing = self.is_graphing
self.model.is_online_training = self.is_online_training
self.model.apply_feature_creation(TensorflowFeatureCreator())
try:
self.model.create_model(self.model.get_input_placeholder())
except TypeError as e:
raise Exception('failed to create model') from e
if self.model.is_training and self.model.is_online_training:
self.model.create_reinforcement_training_model()
self.model.create_savers()
self.model.initialize_model()
if self.is_graphing:
self.rotating_real_reward_buffer = live_data_util.RotatingBuffer(
self.index + 10)
def start_new_file(self):
self.file_number += 1
self.last_action = None
self.reward_manager = reward_manager.RewardManager()
def __init__(self):
self.file_reward = 0
self.file_frame_count = 0
self.total_reward = 0
self.frame_count = 0
self.reward_manager = reward_manager.RewardManager()