diff --git a/python/paddle/distributed/launch/context/device.py b/python/paddle/distributed/launch/context/device.py index c2f6896ab6c045da23a142b3ba5a6511c1d9b6ed..30b8cc1538590c51dbabdc0f181371f101ef5119 100644 --- a/python/paddle/distributed/launch/context/device.py +++ b/python/paddle/distributed/launch/context/device.py @@ -142,4 +142,4 @@ class Device(object): if __name__ == '__main__': d = Device.parse_device() - print(d.get_selected_flag()) + print(d.get_selected_devices()) diff --git a/python/paddle/distributed/launch/controllers/collective.py b/python/paddle/distributed/launch/controllers/collective.py index bbcb7c81d6e65c2e570ad3234619d95d9d7fdb20..3763bac0414511a683906b0971dd9c6cf1fbe9ab 100644 --- a/python/paddle/distributed/launch/controllers/collective.py +++ b/python/paddle/distributed/launch/controllers/collective.py @@ -93,7 +93,7 @@ class CollectiveController(Controller): "PADDLE_RANK_IN_NODE": str(i), } if self.pod.replicas == 1: - e.update({selected_dev_key: selected_dev_list}) + e.update({selected_dev_key: ",".join(selected_dev_list)}) else: e.update({selected_dev_key: selected_dev_list[i]}) self.add_container(envs=e, log_tag=i) @@ -134,7 +134,7 @@ class CollectiveElasticController(CollectiveController): if ok: self.job.replicas = replicas else: - self.ctx.logger.warnning("peer not ready {}".format(self.job)) + self.ctx.logger.warning("peer not ready {}".format(self.job)) break self.ctx.logger.debug("Run {}".format(self.job)) diff --git a/python/paddle/distributed/launch/job/container.py b/python/paddle/distributed/launch/job/container.py index 7105cae9024f25f4b5d63a8d74d5d492777d4046..1f43b6ce04bacf6d144aea43a9b02760b171220b 100644 --- a/python/paddle/distributed/launch/job/container.py +++ b/python/paddle/distributed/launch/job/container.py @@ -162,7 +162,7 @@ class Container(object): if idx > lines: break finally: - return self._log_handler.tell() + return def tail(self, length=3000): if not self._log_handler: