a6652b0c1997bb47dd502bf674e0b3b9b2d09d23,examples/reinforcement_learning/tutorial_bipedalwalker_a3c_continuous_action.py,,,#,254

Before Change

if __name__ == "__main__":
    sess = tf.Session()

    // ============================= TRAINING ===============================
    with tf.device("/cpu:0"):
        OPT_A = tf.train.RMSPropOptimizer(LR_A, name="RMSPropA")

After Change

    while True:
        s = env.reset()
        rall = 0
        while True:
            env.render()
            s = s.astype("float32") // double to float
            a = GLOBAL_AC.choose_action(s)
            s, r, d, _ = env.step(a)
            rall += r
            if d:
                print("reward", rall)
                break

Instances