apply model optimize

benchmark/mnasnet.param

@@ -1,5 +1,5 @@
 7767517
-77 87
+76 86
 Input                    data                     0 1 data
 Convolution              first-3x3-conv           1 1 data first-3x3-conv_relu 0=32 1=3 3=2 4=1 5=1 6=864 9=1
 ConvolutionDepthWise     A0_dw                    1 1 first-3x3-conv_relu A0_dw_relu 0=32 1=3 4=1 5=1 6=288 7=32 9=1
@@ -73,7 +73,6 @@ Convolution              G0_expand                1 1 unknownncnn_9 G0_expand_re
 ConvolutionDepthWise     G0_dw                    1 1 G0_expand_relu G0_dw_relu 0=1152 1=3 4=1 5=1 6=10368 7=1152 9=1
 Convolution              G0_linear                1 1 G0_dw_relu G0_linear_bn 0=320 1=1 5=1 6=368640
 Convolution              last-1x1-conv            1 1 G0_linear_bn last-1x1-conv_relu 0=1280 1=1 5=1 6=409600 9=1
-Pooling                  avgpool                  1 1 last-1x1-conv_relu avgpool 0=1 1=7 4=1 5=1
-Flatten                  flatten                  1 1 avgpool flatten
+Pooling                  avgpool                  1 1 last-1x1-conv_relu flatten 0=1 1=7 4=1 5=1
 InnerProduct             fc                       1 1 flatten fc 0=1000 1=1 2=1280000
 Softmax                  prob                     1 1 fc prob

benchmark/mobilenet.param

@@ -29,5 +29,5 @@ Convolution              conv5_6/sep              1 1 conv5_6/dw_relu5_6/dw conv
 ConvolutionDepthWise     conv6/dw                 1 1 conv5_6/sep_relu5_6/sep conv6/dw_relu6/dw 0=1024 1=3 4=1 5=1 6=9216 7=1024 9=1
 Convolution              conv6/sep                1 1 conv6/dw_relu6/dw conv6/sep_relu6/sep 0=1024 1=1 5=1 6=1048576 9=1
 Pooling                  pool6                    1 1 conv6/sep_relu6/sep pool6 0=1 4=1
-Convolution              fc7                      1 1 pool6 fc7 0=1000 1=1 5=1 6=1024000
+InnerProduct             fc7                      1 1 pool6 fc7 0=1000 1=1 2=1024000
 Softmax                  prob                     1 1 fc7 prob

benchmark/mobilenet_v2.param

@@ -75,5 +75,5 @@ ConvolutionDepthWise     conv6_3/dwise            1 1 conv6_3/expand/bn_relu6_3/
 Convolution              conv6_3/linear           1 1 conv6_3/dwise/bn_relu6_3/dwise conv6_3/linear/bn_conv6_3/linear/scale 0=320 1=1 5=1 6=307200
 Convolution              conv6_4                  1 1 conv6_3/linear/bn_conv6_3/linear/scale conv6_4/bn_relu6_4 0=1280 1=1 5=1 6=409600 9=1
 Pooling                  pool6                    1 1 conv6_4/bn_relu6_4 pool6 0=1 4=1
-Convolution              fc7                      1 1 pool6 fc7 0=1000 1=1 5=1 6=1280000
+InnerProduct             fc7                      1 1 pool6 fc7 0=1000 1=1 2=1280000
 Softmax                  prob                     1 1 fc7 prob

benchmark/proxylessnasnet.param

@@ -1,5 +1,5 @@
 7767517
-92 105
+91 104
 Input                    data                     0 1 data
 Convolution              first-3x3-conv           1 1 data first-3x3-conv_relu 0=32 1=3 3=2 4=1 5=1 6=864 9=1
 ConvolutionDepthWise     A0_dw                    1 1 first-3x3-conv_relu A0_dw_relu 0=32 1=3 4=1 5=1 6=288 7=32 9=1
@@ -88,7 +88,6 @@ Convolution              G0_expand                1 1 unknownncnn_12 G0_expand_r
 ConvolutionDepthWise     G0_dw                    1 1 G0_expand_relu G0_dw_relu 0=1152 1=7 4=3 5=1 6=56448 7=1152 9=1
 Convolution              G0_linear                1 1 G0_dw_relu G0_linear_bn 0=320 1=1 5=1 6=368640
 Convolution              last-1x1-conv            1 1 G0_linear_bn last-1x1-conv_relu 0=1280 1=1 5=1 6=409600 9=1
-Pooling                  avgpool                  1 1 last-1x1-conv_relu avgpool 0=1 1=7 4=1 5=1
-Flatten                  flatten                  1 1 avgpool flatten
+Pooling                  avgpool                  1 1 last-1x1-conv_relu flatten 0=1 1=7 4=1 5=1
 InnerProduct             fc                       1 1 flatten fc 0=1000 1=1 2=1280000
 Softmax                  prob                     1 1 fc prob

benchmark/shufflenet.param

@@ -118,4 +118,4 @@ ConvolutionDepthWise     resx16_conv3             1 1 resx16_conv2_resx16_conv2_
 Eltwise                  resx16_elewise           2 1 resx15_elewise_resx15_elewise_relu_splitncnn_0 resx16_conv3_resx16_conv3_scale resx16_elewise 0=1
 ReLU                     resx16_elewise_relu      1 1 resx16_elewise resx16_elewise_resx16_elewise_relu
 Pooling                  pool_ave                 1 1 resx16_elewise_resx16_elewise_relu pool_ave 0=1 4=1
-Convolution              fc1000                   1 1 pool_ave fc1000 0=1000 1=1 5=1 6=960000
+InnerProduct             fc1000                   1 1 pool_ave fc1000 0=1000 1=1 2=960000