use two_conv_fold for ascend quant net

model_zoo/official/cv/mobilenetv2_quant/train.py

@@ -101,7 +101,8 @@ def train_on_ascend():
     # convert fusion network to quantization aware network
     quantizer = QuantizationAwareTraining(bn_fold=True,
                                           per_channel=[True, False],
-                                          symmetric=[True, False])
+                                          symmetric=[True, False],
+                                          one_conv_fold=False)
     network = quantizer.quantize(network)
 
     # get learning rate

model_zoo/official/cv/resnet50_quant/train.py

@@ -115,7 +115,8 @@ if __name__ == '__main__':
     # convert fusion network to quantization aware network
     quantizer = QuantizationAwareTraining(bn_fold=True,
                                           per_channel=[True, False],
-                                          symmetric=[True, False])
+                                          symmetric=[True, False],
+                                          one_conv_fold=False)
     net = quantizer.quantize(net)
 
     # get learning rate

model_zoo/official/cv/yolov3_darknet53_quant/train.py

@@ -170,7 +170,8 @@ def train():
     if config.quantization_aware:
         quantizer = QuantizationAwareTraining(bn_fold=True,
                                               per_channel=[True, False],
-                                              symmetric=[True, False])
+                                              symmetric=[True, False],
+                                              one_conv_fold=False)
         network = quantizer.quantize(network)
 
     network = YoloWithLossCell(network)