def init():
global RENDER, S, R, S_, env, state_dim, action_dim, action_bound, actor, critic
env = Env(np.ones([
39,
]) * 1e-5)
state_dim = env.observation_space()
action_dim = env.action_space()
action_bound = env.action_bound()
with tf.name_scope('S'):
S = tf.placeholder(tf.float32, shape=[None, state_dim], name='s')
with tf.name_scope('R'):
R = tf.placeholder(tf.float32, [None, 1], name='r')
with tf.name_scope('S_'):
S_ = tf.placeholder(tf.float32, shape=[None, state_dim], name='s_')
sess = tf.Session()
actor = Actor(sess, action_dim, action_bound, LR_A, REPLACE_ITER_A)
critic = Critic(sess, state_dim, action_dim, LR_C, GAMMA, REPLACE_ITER_C,
actor.a, actor.a_)
actor.add_grad_to_graph(critic.a_grads)
sess.run(tf.global_variables_initializer())
