fix a bug with add set_grad() in wide_and_deep network

model_zoo/official/gnn/gat/src/utils.py

@@ -138,6 +138,7 @@ class TrainOneStepCell(nn.Cell):
     def __init__(self, network, optimizer, sens=1.0):
         super(TrainOneStepCell, self).__init__(auto_prefix=True)
         self.network = network
+        self.network.set_grad()
         self.network.add_flags(defer_inline=True)
         self.weights = ParameterTuple(network.trainable_params())
         self.optimizer = optimizer
@@ -167,7 +168,6 @@ class TrainGAT(nn.Cell):
     def __init__(self, network, num_class, label, mask, learning_rate, l2_coeff):
         super(TrainGAT, self).__init__(auto_prefix=False)
         self.network = network
-        self.network.set_grad()
         loss_net = LossNetWrapper(network, num_class, label, mask, l2_coeff)
         optimizer = nn.Adam(loss_net.trainable_params(),
                             learning_rate=learning_rate)

model_zoo/official/recommend/wide_and_deep/src/wide_and_deep.py

@@ -328,7 +328,6 @@ class TrainStepWrap(nn.Cell):
         parallel_mode = context.get_auto_parallel_context("parallel_mode")
         is_auto_parallel = parallel_mode in (ParallelMode.SEMI_AUTO_PARALLEL, ParallelMode.AUTO_PARALLEL)
         self.network = network
-        self.network.set_grad()
         self.network.set_train()
         self.trainable_params = network.trainable_params()
         weights_w = []
@@ -361,6 +360,8 @@ class TrainStepWrap(nn.Cell):
         self.sens = sens
         self.loss_net_w = IthOutputCell(network, output_index=0)
         self.loss_net_d = IthOutputCell(network, output_index=1)
+        self.loss_net_w.set_grad()
+        self.loss_net_d.set_grad()
 
         self.reducer_flag = False
         self.grad_reducer_w = None

model_zoo/official/recommend/wide_and_deep_multitable/src/wide_and_deep.py

@@ -510,7 +510,6 @@ class TrainStepWrap(nn.Cell):
     def __init__(self, network, config, sens=1000.0):
         super(TrainStepWrap, self).__init__()
         self.network = network
-        self.network.set_grad()
         self.network.set_train()
         self.trainable_params = network.trainable_params()
         weights_w = []
@@ -546,6 +545,8 @@ class TrainStepWrap(nn.Cell):
         self.sens = sens
         self.loss_net_w = IthOutputCell(network, output_index=0)
         self.loss_net_d = IthOutputCell(network, output_index=1)
+        self.loss_net_w.set_grad()
+        self.loss_net_w.set_grad()
 
         self.reducer_flag = False
         self.grad_reducer_w = None