!1174 Make optimizer parameter same as gradient

Merge pull request !1174 from ghzl/fix-beg-group-parameters

mindspore/nn/wrap/grad_reducer.py

@@ -141,7 +141,7 @@ class DistributedGradReducer(Cell):
         >>>         super(TrainingWrapper, self).__init__(auto_prefix=False)
         >>>         self.network = network
         >>>         self.network.add_flags(defer_inline=True)
-        >>>         self.weights = ParameterTuple(network.trainable_params())
+        >>>         self.weights = optimizer.parameters
         >>>         self.optimizer = optimizer
         >>>         self.grad = C.GradOperation('grad', get_by_list=True, sens_param=True)
         >>>         self.sens = sens

mindspore/nn/wrap/loss_scale.py

@@ -18,7 +18,7 @@ from mindspore.nn.wrap.grad_reducer import DistributedGradReducer
 from mindspore.train.parallel_utils import ParallelMode
 from mindspore.parallel._utils import _get_device_num, _get_parallel_mode, _get_mirror_mean
 from ..cell import Cell
-from ...common import Tensor, ParameterTuple
+from ...common import Tensor
 from ...common.parameter import Parameter
 from ...ops import functional as F
 from ...ops import composite as C
@@ -201,7 +201,7 @@ class TrainOneStepWithLossScaleCell(Cell):
         super(TrainOneStepWithLossScaleCell, self).__init__(auto_prefix=False)
         self.network = network
         self.network.add_flags(defer_inline=True)
-        self.weights = ParameterTuple(network.trainable_params())
+        self.weights = optimizer.parameters
         self.optimizer = optimizer
         self.grad = C.GradOperation('grad', get_by_list=True, sens_param=True)
         self.hyper_map = C.HyperMap()