diff --git a/PyTorch/built-in/cv/classification/EfficientNet_for_PyTorch/examples/imagenet/main.py b/PyTorch/built-in/cv/classification/EfficientNet_for_PyTorch/examples/imagenet/main.py
index fb92fc1ac25b88ec285a38f9e1eda6bd85b1a19d..d8343ad748952ac5faefb4dfde02761b7f5a66cc 100644
--- a/PyTorch/built-in/cv/classification/EfficientNet_for_PyTorch/examples/imagenet/main.py
+++ b/PyTorch/built-in/cv/classification/EfficientNet_for_PyTorch/examples/imagenet/main.py
@@ -32,7 +32,7 @@ import torch.utils.data.distributed
 import torchvision.transforms as transforms
 import torchvision.datasets as datasets
 import torchvision.models as models
-
+import apex
 from apex import amp
 
 sys.path.append(os.path.join(os.path.abspath(os.path.dirname(__file__)),'../../'))
@@ -195,7 +195,7 @@ def main_worker(npu, nnpus_per_node, args):
 
     criterion = nn.CrossEntropyLoss().to('npu:' + str(args.npu))
 
-    optimizer = torch.optim.SGD(model.parameters(), args.lr,
+    optimizer = apex.optimizers.NpuFusedSGD(model.parameters(), args.lr,
                                 momentum=args.momentum,
                                 weight_decay=args.weight_decay)
     model = model.to('npu:' + str(args.npu))
@@ -205,7 +205,7 @@ def main_worker(npu, nnpus_per_node, args):
             print('=>unsupported precision mode!')
             exit()
         opt_level = args.pm
-        model, optimizer = amp.initialize(model, optimizer, opt_level=opt_level, loss_scale=args.loss_scale)
+        model, optimizer = amp.initialize(model, optimizer, opt_level=opt_level, loss_scale=args.loss_scale,combine_grad=True)
 
     global total_batch_size
     total_batch_size = args.batch_size