Hands-On Neuroevolution with Python

The reset Atari environment function

To train game agents, we need to implement a function that starts the Atari environment from a particular random state. It is vital to implement a stochastic Atari reset function to guarantee that our agent can play the game from any initial state. The function is implemented as follows:

    def reset(self, indices=None, max_frames=None, name=None):        if indices is None:            indices = np.arange(self.batch_size)        with tf.variable_scope(name, default_name='AtariReset'):            noops = tf.random_uniform(tf.shape(indices), minval=1,                                        maxval=31, dtype=tf.int32)            if max_frames is None:                max_frames = tf.ones_like(indices, dtype=tf.int32) * \                                         (100000 * self.frameskip)            import collections if not isinstance(max_frames, collections.Sequence): ...