def __init__(self, hyperparams, policy, sess, actions):
self.hyperparams = hyperparams
self.sess = sess
with tf.device('/cpu:0'):
self.policy = policy.create_policy(trainable=False)
self.sess.run(tf.initialize_all_variables())
self.input_vars = get_input_vars(self.policy)
self.y_hat = get_output(self.policy)
self.experience_cache = []
self.episode_cache = []
self.prepare_epsilon()
self.training = True
self.actions = actions
self.greedy_ind = None
def __init__(self, hyperparams, policy, sess, actions):
self.hyperparams = hyperparams
self.sess = sess
with tf.device('/cpu:0'):
self.policy = policy.create_policy(trainable=False)
self.sess.run(tf.initialize_all_variables())
self.input_vars = get_input_vars(self.policy)
self.y_hat = get_output(self.policy)
self.experience_cache = []
self.episode_cache = []
self.prepare_epsilon()
self.training = True
self.actions = actions
self.greedy_ind = None
def __init__(self, hyperparams, q_model, sess, actions):
self.hyperparams = hyperparams
self.sess = sess
with tf.device('/cpu:0'):
self.net = q_model.create_net(trainable=False)
self.sess.run(tf.initialize_all_variables())
self.input_vars = get_input_vars(self.net)
self.y_hat = get_output(self.net)
self.experience_cache = []
self.recent_train_q = deque(
maxlen=self.hyperparams['num_recent_steps'])
self.recent_eval_q = deque(maxlen=self.hyperparams['num_recent_steps'])
self.prepare_epsilon()
self.training = True
self.actions = actions
self.greedy_ind = None
def __init__(self, hyperparams, q_model, sess, actions):
self.hyperparams = hyperparams
self.sess = sess
with tf.device('/cpu:0'):
self.net = q_model.create_net(trainable=False)
self.sess.run(tf.initialize_all_variables())
self.input_vars = get_input_vars(self.net)
self.y_hat = get_output(self.net)
self.experience_cache = []
self.recent_train_q = deque(
maxlen=self.hyperparams['num_recent_steps'])
self.recent_eval_q = deque(
maxlen=self.hyperparams['num_recent_steps'])
self.prepare_epsilon()
self.training = True
self.actions = actions
self.greedy_ind = None
def setup_net(self):
self.build_net()
self.input_vars = get_input_vars(self.get_net())
self.y_hat = get_output(self.get_net())
def build_net(self):
self.net = self.create_net()
self.input_vars = get_input_vars(self.net)
self.target_net = self.create_net(trainable=False)
self.target_input_vars = get_input_vars(self.target_net)
def setup_net(self):
self.build_net()
self.policy_y_hat = get_output(self.policy)
self.value_y_hat = get_output(self.value)
self.policy_input_vars = get_input_vars(self.policy)
self.value_input_vars = get_input_vars(self.value)
