Merge branch 'a3c-tictactoe' into a3c-tictacote-lambda1

    """

    env = TicTacToeEnvironment()

    policy = TicTacToePolicy()

    a3c = dc.rl.A3C(env, policy, entropy_weight=0.01, value_weight=value_weight)

    a3c.optimizer = dc.models.tensorgraph.TFWrapper(

        tf.train.AdamOptimizer, learning_rate=0.01)

    avg_rewards = []

    for j in range(10):

        a3c = dc.rl.A3C(env, policy, entropy_weight=0.01, value_weight=value_weight, model_dir="/tmp/tictactoe")

        a3c.optimizer = dc.models.tensorgraph.TFWrapper(tf.train.AdamOptimizer, learning_rate=0.01)

        try:

            a3c.restore()

        except:

            print("unable to restore")

            pass

        a3c.fit(rollouts)

        rewards = []

        for i in range(games):