a7ca1cd1b97a84840ee1675e5e1c331f1a5ecd0b,tensorforce/core/memories/prioritized_replay.py,PrioritizedReplay,tf_update_batch,#PrioritizedReplay#Any#,392

Before Change

            y=tf.zeros(shape=tf.shape(input=self.batch_indices), dtype=tf.int32)
        )
        priority_indices = tf.squeeze(tf.where(condition=mask))
        priority_indices = tf.Print(priority_indices, [priority_indices], message="Priority indices")
        sampled_batch = self.tf_retrieve_indices(
            buffer_elements=self.last_batch_buffer_elems,
            priority_indices=priority_indices
        )

After Change

        assignments = list()

        // These are the new priorities of the elements already in the memory.
        main_memory_priorities = priorities[0:tf.shape(priority_indices)[0] - 1]
        assignments.append(tf.scatter_update(
            ref=self.priorities,
            indices=priority_indices,
            updates=main_memory_priorities

Instances