732197e23a937b7b6d196936519c16ec6317ea9f,rllib/execution/train_ops.py,TrainTFMultiGPU,__call__,#TrainTFMultiGPU#Any#,167

Before Change

                        batch_fetches = optimizer.optimize(
                            self.sess, permutation[batch_index] *
                            self.per_device_batch_size)
                        for k, v in batch_fetches[LEARNER_STATS_KEY].items():
                            iter_extra_fetches[k].append(v)
                    if logger.getEffectiveLevel() <= logging.DEBUG:
                        avg = averaged(iter_extra_fetches)
                        logger.debug("{} {}".format(i, avg))
                fetches[policy_id] = averaged(iter_extra_fetches, axis=0)

After Change

                            tree.map_structure_with_path(
                                lambda p, *s: self._all_tower_reduce(p, *s),
                                *(batch_fetches["tower_{}".format(tower_num)]
                                  for tower_num in range(len(self.devices)))))

                // Reduce mean across all minibatch SGD steps (axis=0 to keep
                // all shapes as-is).

Instances