From 3393fa6618b682777e94cb397a71abe2cb752b95 Mon Sep 17 00:00:00 2001
From: Quan Tang <73922533+tt8000@users.noreply.github.com>
Date: Fri, 8 Oct 2021 21:51:06 +0800
Subject: [PATCH 01/17] Create shufflenetv2.py

---
 python/paddle/vision/models/shufflenetv2.py | 473 ++++++++++++++++++++
 1 file changed, 473 insertions(+)
 create mode 100644 python/paddle/vision/models/shufflenetv2.py

diff --git a/python/paddle/vision/models/shufflenetv2.py b/python/paddle/vision/models/shufflenetv2.py
new file mode 100644
index 0000000000000..eb76933bc9752
--- /dev/null
+++ b/python/paddle/vision/models/shufflenetv2.py
@@ -0,0 +1,473 @@
+# copyright (c) 2020 PaddlePaddle Authors. All Rights Reserve.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#    http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from __future__ import absolute_import
+from __future__ import division
+from __future__ import print_function
+
+import paddle
+from paddle import ParamAttr, reshape, transpose, concat, split
+from paddle.nn import Layer, Conv2D, MaxPool2D, AdaptiveAvgPool2D, BatchNorm, Linear
+from paddle.nn.initializer import KaimingNormal
+from paddle.nn.functional import swish
+
+from paddle.utils.download import get_weights_path_from_url
+
+
+MODEL_URLS = {
+    "ShuffleNetV2_x0_25":
+    "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x0_25_pretrained.pdparams",
+    "ShuffleNetV2_x0_33":
+    "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x0_33_pretrained.pdparams",
+    "ShuffleNetV2_x0_5":
+    "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x0_5_pretrained.pdparams",
+    "ShuffleNetV2_x1_0":
+    "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x1_0_pretrained.pdparams",
+    "ShuffleNetV2_x1_5":
+    "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x1_5_pretrained.pdparams",
+    "ShuffleNetV2_x2_0":
+    "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x2_0_pretrained.pdparams",
+    "ShuffleNetV2_swish":
+    "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_swish_pretrained.pdparams"
+}
+
+__all__ = list(MODEL_URLS.keys())
+
+
+def channel_shuffle(x, groups):
+    batch_size, num_channels, height, width = x.shape[0:4]
+    channels_per_group = num_channels // groups
+
+    # reshape
+    x = reshape(
+        x=x, shape=[batch_size, groups, channels_per_group, height, width])
+
+    # transpose
+    x = transpose(x=x, perm=[0, 2, 1, 3, 4])
+
+    # flatten
+    x = reshape(x=x, shape=[batch_size, num_channels, height, width])
+    return x
+
+
+class ConvBNLayer(Layer):
+    def __init__(
+            self,
+            in_channels,
+            out_channels,
+            kernel_size,
+            stride,
+            padding,
+            groups=1,
+            act=None,
+            name=None, ):
+        super(ConvBNLayer, self).__init__()
+        self._conv = Conv2D(
+            in_channels=in_channels,
+            out_channels=out_channels,
+            kernel_size=kernel_size,
+            stride=stride,
+            padding=padding,
+            groups=groups,
+            weight_attr=ParamAttr(
+                initializer=KaimingNormal(), name=name + "_weights"),
+            bias_attr=False)
+
+        self._batch_norm = BatchNorm(
+            out_channels,
+            param_attr=ParamAttr(name=name + "_bn_scale"),
+            bias_attr=ParamAttr(name=name + "_bn_offset"),
+            act=act,
+            moving_mean_name=name + "_bn_mean",
+            moving_variance_name=name + "_bn_variance")
+
+    def forward(self, inputs):
+        y = self._conv(inputs)
+        y = self._batch_norm(y)
+        return y
+
+
+class InvertedResidual(Layer):
+    def __init__(self,
+                 in_channels,
+                 out_channels,
+                 stride,
+                 act="relu",
+                 name=None):
+        super(InvertedResidual, self).__init__()
+        self._conv_pw = ConvBNLayer(
+            in_channels=in_channels // 2,
+            out_channels=out_channels // 2,
+            kernel_size=1,
+            stride=1,
+            padding=0,
+            groups=1,
+            act=act,
+            name='stage_' + name + '_conv1')
+        self._conv_dw = ConvBNLayer(
+            in_channels=out_channels // 2,
+            out_channels=out_channels // 2,
+            kernel_size=3,
+            stride=stride,
+            padding=1,
+            groups=out_channels // 2,
+            act=None,
+            name='stage_' + name + '_conv2')
+        self._conv_linear = ConvBNLayer(
+            in_channels=out_channels // 2,
+            out_channels=out_channels // 2,
+            kernel_size=1,
+            stride=1,
+            padding=0,
+            groups=1,
+            act=act,
+            name='stage_' + name + '_conv3')
+
+    def forward(self, inputs):
+        x1, x2 = split(
+            inputs,
+            num_or_sections=[inputs.shape[1] // 2, inputs.shape[1] // 2],
+            axis=1)
+        x2 = self._conv_pw(x2)
+        x2 = self._conv_dw(x2)
+        x2 = self._conv_linear(x2)
+        out = concat([x1, x2], axis=1)
+        return channel_shuffle(out, 2)
+
+
+class InvertedResidualDS(Layer):
+    def __init__(self,
+                 in_channels,
+                 out_channels,
+                 stride,
+                 act="relu",
+                 name=None):
+        super(InvertedResidualDS, self).__init__()
+
+        # branch1
+        self._conv_dw_1 = ConvBNLayer(
+            in_channels=in_channels,
+            out_channels=in_channels,
+            kernel_size=3,
+            stride=stride,
+            padding=1,
+            groups=in_channels,
+            act=None,
+            name='stage_' + name + '_conv4')
+        self._conv_linear_1 = ConvBNLayer(
+            in_channels=in_channels,
+            out_channels=out_channels // 2,
+            kernel_size=1,
+            stride=1,
+            padding=0,
+            groups=1,
+            act=act,
+            name='stage_' + name + '_conv5')
+        # branch2
+        self._conv_pw_2 = ConvBNLayer(
+            in_channels=in_channels,
+            out_channels=out_channels // 2,
+            kernel_size=1,
+            stride=1,
+            padding=0,
+            groups=1,
+            act=act,
+            name='stage_' + name + '_conv1')
+        self._conv_dw_2 = ConvBNLayer(
+            in_channels=out_channels // 2,
+            out_channels=out_channels // 2,
+            kernel_size=3,
+            stride=stride,
+            padding=1,
+            groups=out_channels // 2,
+            act=None,
+            name='stage_' + name + '_conv2')
+        self._conv_linear_2 = ConvBNLayer(
+            in_channels=out_channels // 2,
+            out_channels=out_channels // 2,
+            kernel_size=1,
+            stride=1,
+            padding=0,
+            groups=1,
+            act=act,
+            name='stage_' + name + '_conv3')
+
+    def forward(self, inputs):
+        x1 = self._conv_dw_1(inputs)
+        x1 = self._conv_linear_1(x1)
+        x2 = self._conv_pw_2(inputs)
+        x2 = self._conv_dw_2(x2)
+        x2 = self._conv_linear_2(x2)
+        out = concat([x1, x2], axis=1)
+
+        return channel_shuffle(out, 2)
+
+
+class ShuffleNetV2(Layer):
+    """ShuffleNetV2 model from
+    `"ShuffleNet V2: Practical Guidelines for Ecient CNN Architecture Design" <https://arxiv.org/pdf/1807.11164.pdf>`_
+    Args:
+        num_classes (int): output dim of last fc layer. If num_classes <=0, last fc layer 
+                            will not be defined. Default: 1000.
+        scale (float): network architecture. Default: 1.0.
+        act (str, optional): Activation to be applied to the output of batch normalization. Default: "relu".
+    Examples:
+        .. code-block:: python
+            from paddle.vision.models import ShuffleNetV2
+
+            shufflenetv2_x0_25 = ShuffleNetV2(num_classes=1000, scale=0.25, act="relu")
+            shufflenetv2_swish = ShuffleNetV2(num_classes=1000, scale=1.0, act="swish")
+    """    
+    def __init__(self, num_classes=1000, scale=1.0, act="relu"):
+        super(ShuffleNetV2, self).__init__()
+        self.scale = scale
+        self.num_classes = num_classes
+        stage_repeats = [4, 8, 4]
+
+        if scale == 0.25:
+            stage_out_channels = [-1, 24, 24, 48, 96, 512]
+        elif scale == 0.33:
+            stage_out_channels = [-1, 24, 32, 64, 128, 512]
+        elif scale == 0.5:
+            stage_out_channels = [-1, 24, 48, 96, 192, 1024]
+        elif scale == 1.0:
+            stage_out_channels = [-1, 24, 116, 232, 464, 1024]
+        elif scale == 1.5:
+            stage_out_channels = [-1, 24, 176, 352, 704, 1024]
+        elif scale == 2.0:
+            stage_out_channels = [-1, 24, 224, 488, 976, 2048]
+        else:
+            raise NotImplementedError("This scale size:[" + str(scale) +
+                                      "] is not implemented!")
+        # 1. conv1
+        self._conv1 = ConvBNLayer(
+            in_channels=3,
+            out_channels=stage_out_channels[1],
+            kernel_size=3,
+            stride=2,
+            padding=1,
+            act=act,
+            name='stage1_conv')
+        self._max_pool = MaxPool2D(kernel_size=3, stride=2, padding=1)
+
+        # 2. bottleneck sequences
+        self._block_list = []
+        for stage_id, num_repeat in enumerate(stage_repeats):
+            for i in range(num_repeat):
+                if i == 0:
+                    block = self.add_sublayer(
+                        name=str(stage_id + 2) + '_' + str(i + 1),
+                        sublayer=InvertedResidualDS(
+                            in_channels=stage_out_channels[stage_id + 1],
+                            out_channels=stage_out_channels[stage_id + 2],
+                            stride=2,
+                            act=act,
+                            name=str(stage_id + 2) + '_' + str(i + 1)))
+                else:
+                    block = self.add_sublayer(
+                        name=str(stage_id + 2) + '_' + str(i + 1),
+                        sublayer=InvertedResidual(
+                            in_channels=stage_out_channels[stage_id + 2],
+                            out_channels=stage_out_channels[stage_id + 2],
+                            stride=1,
+                            act=act,
+                            name=str(stage_id + 2) + '_' + str(i + 1)))
+                self._block_list.append(block)
+        # 3. last_conv
+        self._last_conv = ConvBNLayer(
+            in_channels=stage_out_channels[-2],
+            out_channels=stage_out_channels[-1],
+            kernel_size=1,
+            stride=1,
+            padding=0,
+            act=act,
+            name='conv5')
+        # 4. pool
+        self._pool2d_avg = AdaptiveAvgPool2D(1)
+        self._out_c = stage_out_channels[-1]
+        # 5. fc
+        self._fc = Linear(
+            stage_out_channels[-1],
+            num_classes,
+            weight_attr=ParamAttr(name='fc6_weights'),
+            bias_attr=ParamAttr(name='fc6_offset'))
+
+    def forward(self, inputs):
+        y = self._conv1(inputs)
+        y = self._max_pool(y)
+        for inv in self._block_list:
+            y = inv(y)
+        y = self._last_conv(y)
+        y = self._pool2d_avg(y)
+        y = paddle.flatten(y, start_axis=1, stop_axis=-1)
+        y = self._fc(y)
+        return y
+
+
+def shufflenetv2_x0_25(pretrained=False, **kwargs):
+    """ShuffleNetV2_x0_25 model
+    
+    Args:
+        pretrained (bool): If True, returns a model pre-trained on ImageNet
+    
+    Examples:
+        .. code-block:: python
+            from paddle.vision.models import shufflenetv2_x0_25
+            # build model
+            model = shufflenetv2_x0_25()
+            # build model and load imagenet pretrained weight
+            # model = shufflenetv2_x0_25(pretrained=True)
+    """     
+    model = ShuffleNetV2(scale=0.25, **kwargs)
+    if pretrained:
+        weight_path = get_weights_path_from_url(MODEL_URLS["ShuffleNetV2_x0_25"])
+        param = paddle.load(weight_path)
+        model.set_dict(param)              
+    return model
+
+
+def shufflenetv2_x0_33(pretrained=False, **kwargs):
+    """ShuffleNetV2_x0_33 model
+    
+    Args:
+        pretrained (bool): If True, returns a model pre-trained on ImageNet
+    
+    Examples:
+        .. code-block:: python
+            from paddle.vision.models import shufflenetv2_x0_33
+            # build model
+            model = shufflenetv2_x0_33()
+            # build model and load imagenet pretrained weight
+            # model = shufflenetv2_x0_33(pretrained=True)
+    """      
+    model = ShuffleNetV2(scale=0.33, **kwargs)
+    if pretrained:
+        weight_path = get_weights_path_from_url(MODEL_URLS["ShuffleNetV2_x0_33"])
+        param = paddle.load(weight_path)
+        model.set_dict(param)            
+    return model
+
+
+def shufflenetv2_x0_5(pretrained=False, **kwargs):
+    """ShuffleNetV2_x0_5 model
+    
+    Args:
+        pretrained (bool): If True, returns a model pre-trained on ImageNet
+    
+    Examples:
+        .. code-block:: python
+            from paddle.vision.models import shufflenetv2_x0_5
+            # build model
+            model = shufflenetv2_x0_5()
+            # build model and load imagenet pretrained weight
+            # model = shufflenetv2_x0_5(pretrained=True)
+    """        
+    model = ShuffleNetV2(scale=0.5, **kwargs)
+    if pretrained:
+        weight_path = get_weights_path_from_url(MODEL_URLS["ShuffleNetV2_x0_5"])
+        param = paddle.load(weight_path)
+        model.set_dict(param)    
+    return model
+
+
+def shufflenetv2_x1_0(pretrained=False, **kwargs):
+    """ShuffleNetV2_x1_0 model
+    
+    Args:
+        pretrained (bool): If True, returns a model pre-trained on ImageNet
+    
+    Examples:
+        .. code-block:: python
+            from paddle.vision.models import shufflenetv2_x1_0
+            # build model
+            model = shufflenetv2_x1_0()
+            # build model and load imagenet pretrained weight
+            # model = shufflenetv2_x1_0(pretrained=True)
+    """     
+    model = ShuffleNetV2(scale=1.0, **kwargs)
+    if pretrained:
+        weight_path = get_weights_path_from_url(MODEL_URLS["ShuffleNetV2_x1_0"])
+        param = paddle.load(weight_path)
+        model.set_dict(param)    
+
+    return model
+
+
+def shufflenetv2_x1_5(pretrained=False, **kwargs):
+    """ShuffleNetV2_x1_5 model
+    
+    Args:
+        pretrained (bool): If True, returns a model pre-trained on ImageNet
+    
+    Examples:
+        .. code-block:: python
+            from paddle.vision.models import shufflenetv2_x1_5
+            # build model
+            model = shufflenetv2_x1_5()
+            # build model and load imagenet pretrained weight
+            # model = shufflenetv2_x1_5(pretrained=True)
+    """       
+    model = ShuffleNetV2(scale=1.5, **kwargs)
+    if pretrained:
+        weight_path = get_weights_path_from_url(MODEL_URLS["ShuffleNetV2_x1_5"])
+        param = paddle.load(weight_path)
+        model.set_dict(param)
+
+    return model
+
+
+def shufflenetv2_x2_0(pretrained=False, **kwargs):
+    """ShuffleNetV2_x2_0 model
+    
+    Args:
+        pretrained (bool): If True, returns a model pre-trained on ImageNet
+    
+    Examples:
+        .. code-block:: python
+            from paddle.vision.models import shufflenetv2_x2_0
+            # build model
+            model = shufflenetv2_x2_0()
+            # build model and load imagenet pretrained weight
+            # model = shufflenetv2_x2_0(pretrained=True)
+    """    
+    model = ShuffleNetV2(scale=2.0, **kwargs)
+    if pretrained:
+        weight_path = get_weights_path_from_url(MODEL_URLS["ShuffleNetV2_x2_0"])
+        param = paddle.load(weight_path)
+        model.set_dict(param)
+
+    return model
+
+
+def shufflenetv2_swish(pretrained=False, **kwargs):
+    """ShuffleNetV2_swish model
+    
+    Args:
+        pretrained (bool): If True, returns a model pre-trained on ImageNet
+    
+    Examples:
+        .. code-block:: python
+            from paddle.vision.models import shufflenetv2_swish
+            # build model
+            model = shufflenetv2_swish()
+            # build model and load imagenet pretrained weight
+            # model = shufflenetv2_swish(pretrained=True)
+    """
+    model = ShuffleNetV2(scale=1.0, act="swish", **kwargs)
+    if pretrained:
+        weight_path = get_weights_path_from_url(MODEL_URLS["ShuffleNetV2_swish"])
+        param = paddle.load(weight_path)
+        model.set_dict(param)
+
+    return model

From d9dcd184841de007285d1674e649f3d0f39b0807 Mon Sep 17 00:00:00 2001
From: Quan Tang <73922533+tt8000@users.noreply.github.com>
Date: Fri, 8 Oct 2021 21:56:01 +0800
Subject: [PATCH 02/17] Create __init__.py

---
 python/paddle/vision/models/__init__.py | 19 ++++++++++++++++++-
 1 file changed, 18 insertions(+), 1 deletion(-)

diff --git a/python/paddle/vision/models/__init__.py b/python/paddle/vision/models/__init__.py
index d38f3b1722ee8..5d99e506d6e21 100644
--- a/python/paddle/vision/models/__init__.py
+++ b/python/paddle/vision/models/__init__.py
@@ -28,6 +28,15 @@
 from .vgg import vgg16  # noqa: F401
 from .vgg import vgg19  # noqa: F401
 from .lenet import LeNet  # noqa: F401
+from .shufflenetv2 import ShuffleNetV2
+from .shufflenetv2 import shufflenetv2_x0_25
+from .shufflenetv2 import shufflenetv2_x0_33
+from .shufflenetv2 import shufflenetv2_x0_5
+from .shufflenetv2 import shufflenetv2_x1_0
+from .shufflenetv2 import shufflenetv2_x1_5
+from .shufflenetv2 import shufflenetv2_x2_0
+from .shufflenetv2 import shufflenetv2_swish
+
 
 __all__ = [ #noqa
     'ResNet',
@@ -45,5 +54,13 @@
     'mobilenet_v1',
     'MobileNetV2',
     'mobilenet_v2',
-    'LeNet'
+    'LeNet',
+    'ShuffleNetV2',
+    'shufflenetv2_x0_25',
+    'shufflenetv2_x0_33',
+    'shufflenetv2_x0_5',
+    'shufflenetv2_x1_0',
+    'shufflenetv2_x1_5',
+    'shufflenetv2_x2_0',
+    'shufflenetv2_swish'
 ]

From 3a65412d20792ace8326f36d806759443f1b1aae Mon Sep 17 00:00:00 2001
From: Quan Tang <73922533+tt8000@users.noreply.github.com>
Date: Fri, 8 Oct 2021 21:57:44 +0800
Subject: [PATCH 03/17] Update __init__.py

---
 python/paddle/vision/models/__init__.py | 16 ++++++++--------
 1 file changed, 8 insertions(+), 8 deletions(-)

diff --git a/python/paddle/vision/models/__init__.py b/python/paddle/vision/models/__init__.py
index 5d99e506d6e21..712d66380824f 100644
--- a/python/paddle/vision/models/__init__.py
+++ b/python/paddle/vision/models/__init__.py
@@ -28,14 +28,14 @@
 from .vgg import vgg16  # noqa: F401
 from .vgg import vgg19  # noqa: F401
 from .lenet import LeNet  # noqa: F401
-from .shufflenetv2 import ShuffleNetV2
-from .shufflenetv2 import shufflenetv2_x0_25
-from .shufflenetv2 import shufflenetv2_x0_33
-from .shufflenetv2 import shufflenetv2_x0_5
-from .shufflenetv2 import shufflenetv2_x1_0
-from .shufflenetv2 import shufflenetv2_x1_5
-from .shufflenetv2 import shufflenetv2_x2_0
-from .shufflenetv2 import shufflenetv2_swish
+from .shufflenetv2 import ShuffleNetV2  # noqa: F401
+from .shufflenetv2 import shufflenetv2_x0_25  # noqa: F401
+from .shufflenetv2 import shufflenetv2_x0_33  # noqa: F401
+from .shufflenetv2 import shufflenetv2_x0_5  # noqa: F401
+from .shufflenetv2 import shufflenetv2_x1_0  # noqa: F401
+from .shufflenetv2 import shufflenetv2_x1_5  # noqa: F401
+from .shufflenetv2 import shufflenetv2_x2_0  # noqa: F401
+from .shufflenetv2 import shufflenetv2_swish  # noqa: F401
 
 
 __all__ = [ #noqa

From 5ba9bb2220e42522d573150eacc5aaaf1bf12b50 Mon Sep 17 00:00:00 2001
From: Quan Tang <73922533+tt8000@users.noreply.github.com>
Date: Fri, 8 Oct 2021 22:08:03 +0800
Subject: [PATCH 04/17] Update shufflenetv2.py

---
 python/paddle/vision/models/shufflenetv2.py | 66 ++++++---------------
 1 file changed, 18 insertions(+), 48 deletions(-)

diff --git a/python/paddle/vision/models/shufflenetv2.py b/python/paddle/vision/models/shufflenetv2.py
index eb76933bc9752..73600743bfbc7 100644
--- a/python/paddle/vision/models/shufflenetv2.py
+++ b/python/paddle/vision/models/shufflenetv2.py
@@ -315,6 +315,15 @@ def forward(self, inputs):
         return y
 
 
+def _shufflenetv2(arch, scale, pretrained=False, **kwargs):
+    model = ShuffleNetV2(scale=scale, **kwargs)
+    if pretrained:
+        weight_path = get_weights_path_from_url(MODEL_URLS[arch])
+        param = paddle.load(weight_path)
+        model.set_dict(param)    
+    return model    
+
+
 def shufflenetv2_x0_25(pretrained=False, **kwargs):
     """ShuffleNetV2_x0_25 model
     
@@ -328,13 +337,8 @@ def shufflenetv2_x0_25(pretrained=False, **kwargs):
             model = shufflenetv2_x0_25()
             # build model and load imagenet pretrained weight
             # model = shufflenetv2_x0_25(pretrained=True)
-    """     
-    model = ShuffleNetV2(scale=0.25, **kwargs)
-    if pretrained:
-        weight_path = get_weights_path_from_url(MODEL_URLS["ShuffleNetV2_x0_25"])
-        param = paddle.load(weight_path)
-        model.set_dict(param)              
-    return model
+    """
+    return _shufflenetv2("ShuffleNetV2_x0_25", 0.25, pretrained, **kwargs)     
 
 
 def shufflenetv2_x0_33(pretrained=False, **kwargs):
@@ -351,12 +355,7 @@ def shufflenetv2_x0_33(pretrained=False, **kwargs):
             # build model and load imagenet pretrained weight
             # model = shufflenetv2_x0_33(pretrained=True)
     """      
-    model = ShuffleNetV2(scale=0.33, **kwargs)
-    if pretrained:
-        weight_path = get_weights_path_from_url(MODEL_URLS["ShuffleNetV2_x0_33"])
-        param = paddle.load(weight_path)
-        model.set_dict(param)            
-    return model
+    return _shufflenetv2("ShuffleNetV2_x0_33", 0.33, pretrained, **kwargs)       
 
 
 def shufflenetv2_x0_5(pretrained=False, **kwargs):
@@ -372,13 +371,8 @@ def shufflenetv2_x0_5(pretrained=False, **kwargs):
             model = shufflenetv2_x0_5()
             # build model and load imagenet pretrained weight
             # model = shufflenetv2_x0_5(pretrained=True)
-    """        
-    model = ShuffleNetV2(scale=0.5, **kwargs)
-    if pretrained:
-        weight_path = get_weights_path_from_url(MODEL_URLS["ShuffleNetV2_x0_5"])
-        param = paddle.load(weight_path)
-        model.set_dict(param)    
-    return model
+    """       
+    return _shufflenetv2("ShuffleNetV2_x0_5", 0.5, pretrained, **kwargs) 
 
 
 def shufflenetv2_x1_0(pretrained=False, **kwargs):
@@ -395,13 +389,7 @@ def shufflenetv2_x1_0(pretrained=False, **kwargs):
             # build model and load imagenet pretrained weight
             # model = shufflenetv2_x1_0(pretrained=True)
     """     
-    model = ShuffleNetV2(scale=1.0, **kwargs)
-    if pretrained:
-        weight_path = get_weights_path_from_url(MODEL_URLS["ShuffleNetV2_x1_0"])
-        param = paddle.load(weight_path)
-        model.set_dict(param)    
-
-    return model
+    return _shufflenetv2("ShuffleNetV2_x1_0", 1.0, pretrained, **kwargs)
 
 
 def shufflenetv2_x1_5(pretrained=False, **kwargs):
@@ -418,13 +406,7 @@ def shufflenetv2_x1_5(pretrained=False, **kwargs):
             # build model and load imagenet pretrained weight
             # model = shufflenetv2_x1_5(pretrained=True)
     """       
-    model = ShuffleNetV2(scale=1.5, **kwargs)
-    if pretrained:
-        weight_path = get_weights_path_from_url(MODEL_URLS["ShuffleNetV2_x1_5"])
-        param = paddle.load(weight_path)
-        model.set_dict(param)
-
-    return model
+    return _shufflenetv2("ShuffleNetV2_x1_5", 1.5, pretrained, **kwargs)
 
 
 def shufflenetv2_x2_0(pretrained=False, **kwargs):
@@ -441,13 +423,7 @@ def shufflenetv2_x2_0(pretrained=False, **kwargs):
             # build model and load imagenet pretrained weight
             # model = shufflenetv2_x2_0(pretrained=True)
     """    
-    model = ShuffleNetV2(scale=2.0, **kwargs)
-    if pretrained:
-        weight_path = get_weights_path_from_url(MODEL_URLS["ShuffleNetV2_x2_0"])
-        param = paddle.load(weight_path)
-        model.set_dict(param)
-
-    return model
+    return _shufflenetv2("ShuffleNetV2_x2_0", 2.0, pretrained, **kwargs)    
 
 
 def shufflenetv2_swish(pretrained=False, **kwargs):
@@ -464,10 +440,4 @@ def shufflenetv2_swish(pretrained=False, **kwargs):
             # build model and load imagenet pretrained weight
             # model = shufflenetv2_swish(pretrained=True)
     """
-    model = ShuffleNetV2(scale=1.0, act="swish", **kwargs)
-    if pretrained:
-        weight_path = get_weights_path_from_url(MODEL_URLS["ShuffleNetV2_swish"])
-        param = paddle.load(weight_path)
-        model.set_dict(param)
-
-    return model
+    return _shufflenetv2("ShuffleNetV2_swish", 1.0, pretrained, act="swish", **kwargs) 

From 9c2f5d44c4d6220d6d3f696db26ed950b1771e5e Mon Sep 17 00:00:00 2001
From: Quan Tang <73922533+tt8000@users.noreply.github.com>
Date: Fri, 8 Oct 2021 22:38:21 +0800
Subject: [PATCH 05/17] Update shufflenetv2.py

---
 python/paddle/vision/models/shufflenetv2.py | 59 ++++++++-------------
 1 file changed, 21 insertions(+), 38 deletions(-)

diff --git a/python/paddle/vision/models/shufflenetv2.py b/python/paddle/vision/models/shufflenetv2.py
index 73600743bfbc7..d740566795f88 100644
--- a/python/paddle/vision/models/shufflenetv2.py
+++ b/python/paddle/vision/models/shufflenetv2.py
@@ -70,8 +70,7 @@ def __init__(
             stride,
             padding,
             groups=1,
-            act=None,
-            name=None, ):
+            act=None):
         super(ConvBNLayer, self).__init__()
         self._conv = Conv2D(
             in_channels=in_channels,
@@ -81,16 +80,14 @@ def __init__(
             padding=padding,
             groups=groups,
             weight_attr=ParamAttr(
-                initializer=KaimingNormal(), name=name + "_weights"),
+                initializer=KaimingNormal()),
             bias_attr=False)
 
         self._batch_norm = BatchNorm(
             out_channels,
-            param_attr=ParamAttr(name=name + "_bn_scale"),
-            bias_attr=ParamAttr(name=name + "_bn_offset"),
-            act=act,
-            moving_mean_name=name + "_bn_mean",
-            moving_variance_name=name + "_bn_variance")
+            param_attr=ParamAttr(),
+            bias_attr=ParamAttr(),
+            act=act)
 
     def forward(self, inputs):
         y = self._conv(inputs)
@@ -103,8 +100,7 @@ def __init__(self,
                  in_channels,
                  out_channels,
                  stride,
-                 act="relu",
-                 name=None):
+                 act="relu"):
         super(InvertedResidual, self).__init__()
         self._conv_pw = ConvBNLayer(
             in_channels=in_channels // 2,
@@ -113,8 +109,7 @@ def __init__(self,
             stride=1,
             padding=0,
             groups=1,
-            act=act,
-            name='stage_' + name + '_conv1')
+            act=act)
         self._conv_dw = ConvBNLayer(
             in_channels=out_channels // 2,
             out_channels=out_channels // 2,
@@ -122,8 +117,7 @@ def __init__(self,
             stride=stride,
             padding=1,
             groups=out_channels // 2,
-            act=None,
-            name='stage_' + name + '_conv2')
+            act=None)
         self._conv_linear = ConvBNLayer(
             in_channels=out_channels // 2,
             out_channels=out_channels // 2,
@@ -131,8 +125,7 @@ def __init__(self,
             stride=1,
             padding=0,
             groups=1,
-            act=act,
-            name='stage_' + name + '_conv3')
+            act=act)
 
     def forward(self, inputs):
         x1, x2 = split(
@@ -151,8 +144,7 @@ def __init__(self,
                  in_channels,
                  out_channels,
                  stride,
-                 act="relu",
-                 name=None):
+                 act="relu"):
         super(InvertedResidualDS, self).__init__()
 
         # branch1
@@ -163,8 +155,7 @@ def __init__(self,
             stride=stride,
             padding=1,
             groups=in_channels,
-            act=None,
-            name='stage_' + name + '_conv4')
+            act=None)
         self._conv_linear_1 = ConvBNLayer(
             in_channels=in_channels,
             out_channels=out_channels // 2,
@@ -172,8 +163,7 @@ def __init__(self,
             stride=1,
             padding=0,
             groups=1,
-            act=act,
-            name='stage_' + name + '_conv5')
+            act=act)
         # branch2
         self._conv_pw_2 = ConvBNLayer(
             in_channels=in_channels,
@@ -182,8 +172,7 @@ def __init__(self,
             stride=1,
             padding=0,
             groups=1,
-            act=act,
-            name='stage_' + name + '_conv1')
+            act=act)
         self._conv_dw_2 = ConvBNLayer(
             in_channels=out_channels // 2,
             out_channels=out_channels // 2,
@@ -191,8 +180,7 @@ def __init__(self,
             stride=stride,
             padding=1,
             groups=out_channels // 2,
-            act=None,
-            name='stage_' + name + '_conv2')
+            act=None)
         self._conv_linear_2 = ConvBNLayer(
             in_channels=out_channels // 2,
             out_channels=out_channels // 2,
@@ -200,8 +188,7 @@ def __init__(self,
             stride=1,
             padding=0,
             groups=1,
-            act=act,
-            name='stage_' + name + '_conv3')
+            act=act)
 
     def forward(self, inputs):
         x1 = self._conv_dw_1(inputs)
@@ -257,8 +244,7 @@ def __init__(self, num_classes=1000, scale=1.0, act="relu"):
             kernel_size=3,
             stride=2,
             padding=1,
-            act=act,
-            name='stage1_conv')
+            act=act)
         self._max_pool = MaxPool2D(kernel_size=3, stride=2, padding=1)
 
         # 2. bottleneck sequences
@@ -272,8 +258,7 @@ def __init__(self, num_classes=1000, scale=1.0, act="relu"):
                             in_channels=stage_out_channels[stage_id + 1],
                             out_channels=stage_out_channels[stage_id + 2],
                             stride=2,
-                            act=act,
-                            name=str(stage_id + 2) + '_' + str(i + 1)))
+                            act=act))
                 else:
                     block = self.add_sublayer(
                         name=str(stage_id + 2) + '_' + str(i + 1),
@@ -281,8 +266,7 @@ def __init__(self, num_classes=1000, scale=1.0, act="relu"):
                             in_channels=stage_out_channels[stage_id + 2],
                             out_channels=stage_out_channels[stage_id + 2],
                             stride=1,
-                            act=act,
-                            name=str(stage_id + 2) + '_' + str(i + 1)))
+                            act=act))
                 self._block_list.append(block)
         # 3. last_conv
         self._last_conv = ConvBNLayer(
@@ -291,8 +275,7 @@ def __init__(self, num_classes=1000, scale=1.0, act="relu"):
             kernel_size=1,
             stride=1,
             padding=0,
-            act=act,
-            name='conv5')
+            act=act)
         # 4. pool
         self._pool2d_avg = AdaptiveAvgPool2D(1)
         self._out_c = stage_out_channels[-1]
@@ -300,8 +283,8 @@ def __init__(self, num_classes=1000, scale=1.0, act="relu"):
         self._fc = Linear(
             stage_out_channels[-1],
             num_classes,
-            weight_attr=ParamAttr(name='fc6_weights'),
-            bias_attr=ParamAttr(name='fc6_offset'))
+            weight_attr=ParamAttr(),
+            bias_attr=ParamAttr())
 
     def forward(self, inputs):
         y = self._conv1(inputs)

From 3c02d5a0ef17ce4509d02a1d1a38062951a198bc Mon Sep 17 00:00:00 2001
From: Quan Tang <73922533+tt8000@users.noreply.github.com>
Date: Fri, 8 Oct 2021 22:42:10 +0800
Subject: [PATCH 06/17] Update test_pretrained_model.py

---
 python/paddle/tests/test_pretrained_model.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/python/paddle/tests/test_pretrained_model.py b/python/paddle/tests/test_pretrained_model.py
index b24b51555c581..c0bbc8ff49dd5 100644
--- a/python/paddle/tests/test_pretrained_model.py
+++ b/python/paddle/tests/test_pretrained_model.py
@@ -52,7 +52,7 @@ def infer(self, arch):
         np.testing.assert_allclose(res['dygraph'], res['static'])
 
     def test_models(self):
-        arches = ['mobilenet_v1', 'mobilenet_v2', 'resnet18', 'vgg16']
+        arches = ['mobilenet_v1', 'mobilenet_v2', 'resnet18', 'vgg16', 'shufflenetv2_swish']
         for arch in arches:
             self.infer(arch)
 

From 1ef363efe12ddca5d6687f626519c3331422e521 Mon Sep 17 00:00:00 2001
From: Quan Tang <73922533+tt8000@users.noreply.github.com>
Date: Fri, 8 Oct 2021 22:47:48 +0800
Subject: [PATCH 07/17] Update test_vision_models.py

---
 python/paddle/tests/test_vision_models.py | 22 +++++++++++++++++++++-
 1 file changed, 21 insertions(+), 1 deletion(-)

diff --git a/python/paddle/tests/test_vision_models.py b/python/paddle/tests/test_vision_models.py
index a25a8f373c29c..a6c88075f974c 100644
--- a/python/paddle/tests/test_vision_models.py
+++ b/python/paddle/tests/test_vision_models.py
@@ -81,7 +81,27 @@ def test_lenet(self):
 
         x = np.array(np.random.random((2, 1, 28, 28)), dtype=np.float32)
         lenet.predict_batch(x)
-
+        
+    def test_shufflenetv2_x0_25(self):
+        self.models_infer('shufflenetv2_x0_25')
+        
+    def test_shufflenetv2_x0_33(self):
+        self.models_infer('shufflenetv2_x0_33')
+
+    def test_shufflenetv2_x0_5(self):
+        self.models_infer('shufflenetv2_x0_5')
+        
+    def test_shufflenetv2_x1_0(self):
+        self.models_infer('shufflenetv2_x1_0')
+        
+    def test_shufflenetv2_x1_5(self):
+        self.models_infer('shufflenetv2_x1_5')
+        
+    def test_shufflenetv2_x2_0(self):
+        self.models_infer('shufflenetv2_x2_0')
+        
+    def test_shufflenetv2_swish(self):
+        self.models_infer('shufflenetv2_swish')        
 
 if __name__ == '__main__':
     unittest.main()

From 861fe5172fe6fa29f68d4ea602f7cd0457b4e96f Mon Sep 17 00:00:00 2001
From: Quan Tang <73922533+tt8000@users.noreply.github.com>
Date: Sat, 9 Oct 2021 14:13:28 +0800
Subject: [PATCH 08/17] Update __init__.py

---
 python/paddle/vision/__init__.py | 8 ++++++++
 1 file changed, 8 insertions(+)

diff --git a/python/paddle/vision/__init__.py b/python/paddle/vision/__init__.py
index 76393865ded04..4aff778ba0871 100644
--- a/python/paddle/vision/__init__.py
+++ b/python/paddle/vision/__init__.py
@@ -44,6 +44,14 @@
 from .models import vgg16  # noqa: F401
 from .models import vgg19  # noqa: F401
 from .models import LeNet  # noqa: F401
+from .models import ShuffleNetV2  # noqa: F401
+from .models import shufflenetv2_x0_25  # noqa: F401
+from .models import shufflenetv2_x0_33  # noqa: F401
+from .models import shufflenetv2_x0_5  # noqa: F401
+from .models import shufflenetv2_x1_0  # noqa: F401
+from .models import shufflenetv2_x1_5  # noqa: F401
+from .models import shufflenetv2_x2_0  # noqa: F401
+from .models import shufflenetv2_swish  # noqa: F401
 from .transforms import BaseTransform  # noqa: F401
 from .transforms import Compose  # noqa: F401
 from .transforms import Resize  # noqa: F401

From 87b6c74dfacde0bb19c218572088d6900b3caae6 Mon Sep 17 00:00:00 2001
From: Quan Tang <73922533+tt8000@users.noreply.github.com>
Date: Sat, 9 Oct 2021 15:03:12 +0800
Subject: [PATCH 09/17] Update shufflenetv2.py

---
 python/paddle/vision/models/shufflenetv2.py | 23 +++++++++++++++++++--
 1 file changed, 21 insertions(+), 2 deletions(-)

diff --git a/python/paddle/vision/models/shufflenetv2.py b/python/paddle/vision/models/shufflenetv2.py
index d740566795f88..3d328b51f7f9b 100644
--- a/python/paddle/vision/models/shufflenetv2.py
+++ b/python/paddle/vision/models/shufflenetv2.py
@@ -17,7 +17,8 @@
 from __future__ import print_function
 
 import paddle
-from paddle import ParamAttr, reshape, transpose, concat, split
+from paddle.fluid.layers import reshape, transpose, concat, split
+from paddle.fluid.param_attr import ParamAttr
 from paddle.nn import Layer, Conv2D, MaxPool2D, AdaptiveAvgPool2D, BatchNorm, Linear
 from paddle.nn.initializer import KaimingNormal
 from paddle.nn.functional import swish
@@ -131,7 +132,7 @@ def forward(self, inputs):
         x1, x2 = split(
             inputs,
             num_or_sections=[inputs.shape[1] // 2, inputs.shape[1] // 2],
-            axis=1)
+            dim=1)
         x2 = self._conv_pw(x2)
         x2 = self._conv_dw(x2)
         x2 = self._conv_linear(x2)
@@ -204,17 +205,21 @@ def forward(self, inputs):
 class ShuffleNetV2(Layer):
     """ShuffleNetV2 model from
     `"ShuffleNet V2: Practical Guidelines for Ecient CNN Architecture Design" <https://arxiv.org/pdf/1807.11164.pdf>`_
+
     Args:
         num_classes (int): output dim of last fc layer. If num_classes <=0, last fc layer 
                             will not be defined. Default: 1000.
         scale (float): network architecture. Default: 1.0.
         act (str, optional): Activation to be applied to the output of batch normalization. Default: "relu".
+
     Examples:
         .. code-block:: python
+
             from paddle.vision.models import ShuffleNetV2
 
             shufflenetv2_x0_25 = ShuffleNetV2(num_classes=1000, scale=0.25, act="relu")
             shufflenetv2_swish = ShuffleNetV2(num_classes=1000, scale=1.0, act="swish")
+            
     """    
     def __init__(self, num_classes=1000, scale=1.0, act="relu"):
         super(ShuffleNetV2, self).__init__()
@@ -315,11 +320,13 @@ def shufflenetv2_x0_25(pretrained=False, **kwargs):
     
     Examples:
         .. code-block:: python
+
             from paddle.vision.models import shufflenetv2_x0_25
             # build model
             model = shufflenetv2_x0_25()
             # build model and load imagenet pretrained weight
             # model = shufflenetv2_x0_25(pretrained=True)
+
     """
     return _shufflenetv2("ShuffleNetV2_x0_25", 0.25, pretrained, **kwargs)     
 
@@ -332,11 +339,13 @@ def shufflenetv2_x0_33(pretrained=False, **kwargs):
     
     Examples:
         .. code-block:: python
+
             from paddle.vision.models import shufflenetv2_x0_33
             # build model
             model = shufflenetv2_x0_33()
             # build model and load imagenet pretrained weight
             # model = shufflenetv2_x0_33(pretrained=True)
+
     """      
     return _shufflenetv2("ShuffleNetV2_x0_33", 0.33, pretrained, **kwargs)       
 
@@ -349,11 +358,13 @@ def shufflenetv2_x0_5(pretrained=False, **kwargs):
     
     Examples:
         .. code-block:: python
+
             from paddle.vision.models import shufflenetv2_x0_5
             # build model
             model = shufflenetv2_x0_5()
             # build model and load imagenet pretrained weight
             # model = shufflenetv2_x0_5(pretrained=True)
+
     """       
     return _shufflenetv2("ShuffleNetV2_x0_5", 0.5, pretrained, **kwargs) 
 
@@ -366,11 +377,13 @@ def shufflenetv2_x1_0(pretrained=False, **kwargs):
     
     Examples:
         .. code-block:: python
+
             from paddle.vision.models import shufflenetv2_x1_0
             # build model
             model = shufflenetv2_x1_0()
             # build model and load imagenet pretrained weight
             # model = shufflenetv2_x1_0(pretrained=True)
+
     """     
     return _shufflenetv2("ShuffleNetV2_x1_0", 1.0, pretrained, **kwargs)
 
@@ -383,11 +396,13 @@ def shufflenetv2_x1_5(pretrained=False, **kwargs):
     
     Examples:
         .. code-block:: python
+
             from paddle.vision.models import shufflenetv2_x1_5
             # build model
             model = shufflenetv2_x1_5()
             # build model and load imagenet pretrained weight
             # model = shufflenetv2_x1_5(pretrained=True)
+
     """       
     return _shufflenetv2("ShuffleNetV2_x1_5", 1.5, pretrained, **kwargs)
 
@@ -400,11 +415,13 @@ def shufflenetv2_x2_0(pretrained=False, **kwargs):
     
     Examples:
         .. code-block:: python
+
             from paddle.vision.models import shufflenetv2_x2_0
             # build model
             model = shufflenetv2_x2_0()
             # build model and load imagenet pretrained weight
             # model = shufflenetv2_x2_0(pretrained=True)
+
     """    
     return _shufflenetv2("ShuffleNetV2_x2_0", 2.0, pretrained, **kwargs)    
 
@@ -417,10 +434,12 @@ def shufflenetv2_swish(pretrained=False, **kwargs):
     
     Examples:
         .. code-block:: python
+
             from paddle.vision.models import shufflenetv2_swish
             # build model
             model = shufflenetv2_swish()
             # build model and load imagenet pretrained weight
             # model = shufflenetv2_swish(pretrained=True)
+
     """
     return _shufflenetv2("ShuffleNetV2_swish", 1.0, pretrained, act="swish", **kwargs) 

From be4c2a9a4b5d29a0274954ec7cea8a093483734e Mon Sep 17 00:00:00 2001
From: Quan Tang <73922533+tt8000@users.noreply.github.com>
Date: Sat, 9 Oct 2021 16:23:07 +0800
Subject: [PATCH 10/17] Update shufflenetv2.py

---
 python/paddle/vision/models/shufflenetv2.py | 40 ++++++++++-----------
 1 file changed, 20 insertions(+), 20 deletions(-)

diff --git a/python/paddle/vision/models/shufflenetv2.py b/python/paddle/vision/models/shufflenetv2.py
index 3d328b51f7f9b..aceb511c97ad8 100644
--- a/python/paddle/vision/models/shufflenetv2.py
+++ b/python/paddle/vision/models/shufflenetv2.py
@@ -27,19 +27,19 @@
 
 
 MODEL_URLS = {
-    "ShuffleNetV2_x0_25":
+    "shufflenetv2_x0_25":
     "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x0_25_pretrained.pdparams",
-    "ShuffleNetV2_x0_33":
+    "shufflenetv2_x0_33":
     "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x0_33_pretrained.pdparams",
-    "ShuffleNetV2_x0_5":
+    "shufflenetv2_x0_5":
     "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x0_5_pretrained.pdparams",
-    "ShuffleNetV2_x1_0":
+    "shufflenetv2_x1_0":
     "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x1_0_pretrained.pdparams",
-    "ShuffleNetV2_x1_5":
+    "shufflenetv2_x1_5":
     "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x1_5_pretrained.pdparams",
-    "ShuffleNetV2_x2_0":
+    "shufflenetv2_x2_0":
     "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x2_0_pretrained.pdparams",
-    "ShuffleNetV2_swish":
+    "shufflenetv2_swish":
     "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_swish_pretrained.pdparams"
 }
 
@@ -313,7 +313,7 @@ def _shufflenetv2(arch, scale, pretrained=False, **kwargs):
 
 
 def shufflenetv2_x0_25(pretrained=False, **kwargs):
-    """ShuffleNetV2_x0_25 model
+    """shufflenetv2_x0_25 model
     
     Args:
         pretrained (bool): If True, returns a model pre-trained on ImageNet
@@ -328,11 +328,11 @@ def shufflenetv2_x0_25(pretrained=False, **kwargs):
             # model = shufflenetv2_x0_25(pretrained=True)
 
     """
-    return _shufflenetv2("ShuffleNetV2_x0_25", 0.25, pretrained, **kwargs)     
+    return _shufflenetv2("shufflenetv2_x0_25", 0.25, pretrained, **kwargs)     
 
 
 def shufflenetv2_x0_33(pretrained=False, **kwargs):
-    """ShuffleNetV2_x0_33 model
+    """shufflenetv2_x0_33 model
     
     Args:
         pretrained (bool): If True, returns a model pre-trained on ImageNet
@@ -351,7 +351,7 @@ def shufflenetv2_x0_33(pretrained=False, **kwargs):
 
 
 def shufflenetv2_x0_5(pretrained=False, **kwargs):
-    """ShuffleNetV2_x0_5 model
+    """shufflenetv2_x0_5 model
     
     Args:
         pretrained (bool): If True, returns a model pre-trained on ImageNet
@@ -366,11 +366,11 @@ def shufflenetv2_x0_5(pretrained=False, **kwargs):
             # model = shufflenetv2_x0_5(pretrained=True)
 
     """       
-    return _shufflenetv2("ShuffleNetV2_x0_5", 0.5, pretrained, **kwargs) 
+    return _shufflenetv2("shufflenetv2_x0_5", 0.5, pretrained, **kwargs) 
 
 
 def shufflenetv2_x1_0(pretrained=False, **kwargs):
-    """ShuffleNetV2_x1_0 model
+    """shufflenetv2_x1_0 model
     
     Args:
         pretrained (bool): If True, returns a model pre-trained on ImageNet
@@ -385,11 +385,11 @@ def shufflenetv2_x1_0(pretrained=False, **kwargs):
             # model = shufflenetv2_x1_0(pretrained=True)
 
     """     
-    return _shufflenetv2("ShuffleNetV2_x1_0", 1.0, pretrained, **kwargs)
+    return _shufflenetv2("shufflenetv2_x1_0", 1.0, pretrained, **kwargs)
 
 
 def shufflenetv2_x1_5(pretrained=False, **kwargs):
-    """ShuffleNetV2_x1_5 model
+    """shufflenetv2_x1_5 model
     
     Args:
         pretrained (bool): If True, returns a model pre-trained on ImageNet
@@ -404,11 +404,11 @@ def shufflenetv2_x1_5(pretrained=False, **kwargs):
             # model = shufflenetv2_x1_5(pretrained=True)
 
     """       
-    return _shufflenetv2("ShuffleNetV2_x1_5", 1.5, pretrained, **kwargs)
+    return _shufflenetv2("shufflenetv2_x1_5", 1.5, pretrained, **kwargs)
 
 
 def shufflenetv2_x2_0(pretrained=False, **kwargs):
-    """ShuffleNetV2_x2_0 model
+    """shufflenetv2_x2_0 model
     
     Args:
         pretrained (bool): If True, returns a model pre-trained on ImageNet
@@ -423,11 +423,11 @@ def shufflenetv2_x2_0(pretrained=False, **kwargs):
             # model = shufflenetv2_x2_0(pretrained=True)
 
     """    
-    return _shufflenetv2("ShuffleNetV2_x2_0", 2.0, pretrained, **kwargs)    
+    return _shufflenetv2("shufflenetv2_x2_0", 2.0, pretrained, **kwargs)    
 
 
 def shufflenetv2_swish(pretrained=False, **kwargs):
-    """ShuffleNetV2_swish model
+    """shufflenetv2_swish model
     
     Args:
         pretrained (bool): If True, returns a model pre-trained on ImageNet
@@ -442,4 +442,4 @@ def shufflenetv2_swish(pretrained=False, **kwargs):
             # model = shufflenetv2_swish(pretrained=True)
 
     """
-    return _shufflenetv2("ShuffleNetV2_swish", 1.0, pretrained, act="swish", **kwargs) 
+    return _shufflenetv2("shufflenetv2_swish", 1.0, pretrained, act="swish", **kwargs) 

From 87cc10bd19db1aa629959cbe01a513b86319aec3 Mon Sep 17 00:00:00 2001
From: Quan Tang <73922533+tt8000@users.noreply.github.com>
Date: Sat, 9 Oct 2021 16:47:29 +0800
Subject: [PATCH 11/17] Update test_pretrained_model.py

---
 python/paddle/tests/test_pretrained_model.py | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/python/paddle/tests/test_pretrained_model.py b/python/paddle/tests/test_pretrained_model.py
index c0bbc8ff49dd5..eb3e102a832c3 100644
--- a/python/paddle/tests/test_pretrained_model.py
+++ b/python/paddle/tests/test_pretrained_model.py
@@ -52,7 +52,9 @@ def infer(self, arch):
         np.testing.assert_allclose(res['dygraph'], res['static'])
 
     def test_models(self):
-        arches = ['mobilenet_v1', 'mobilenet_v2', 'resnet18', 'vgg16', 'shufflenetv2_swish']
+        arches = ['mobilenet_v1', 'mobilenet_v2', 'resnet18', 'vgg16',
+                  'shufflenetv2_swish'
+        ]
         for arch in arches:
             self.infer(arch)
 

From d1d76d32dee6d085a44c950e88113c0f572d1578 Mon Sep 17 00:00:00 2001
From: Quan Tang <73922533+tt8000@users.noreply.github.com>
Date: Sat, 9 Oct 2021 16:56:24 +0800
Subject: [PATCH 12/17] Create shufflenetv2.py

---
 python/paddle/vision/models/shufflenetv2.py | 28 ++++++++++++++++++---
 1 file changed, 24 insertions(+), 4 deletions(-)

diff --git a/python/paddle/vision/models/shufflenetv2.py b/python/paddle/vision/models/shufflenetv2.py
index aceb511c97ad8..6ad5d9c982379 100644
--- a/python/paddle/vision/models/shufflenetv2.py
+++ b/python/paddle/vision/models/shufflenetv2.py
@@ -21,7 +21,6 @@
 from paddle.fluid.param_attr import ParamAttr
 from paddle.nn import Layer, Conv2D, MaxPool2D, AdaptiveAvgPool2D, BatchNorm, Linear
 from paddle.nn.initializer import KaimingNormal
-from paddle.nn.functional import swish
 
 from paddle.utils.download import get_weights_path_from_url
 
@@ -207,9 +206,9 @@ class ShuffleNetV2(Layer):
     `"ShuffleNet V2: Practical Guidelines for Ecient CNN Architecture Design" <https://arxiv.org/pdf/1807.11164.pdf>`_
 
     Args:
-        num_classes (int): output dim of last fc layer. If num_classes <=0, last fc layer 
+        num_classes (int, optional): output dim of last fc layer. If num_classes <=0, last fc layer 
                             will not be defined. Default: 1000.
-        scale (float): network architecture. Default: 1.0.
+        scale (float, optional): network architecture. Default: 1.0.
         act (str, optional): Activation to be applied to the output of batch normalization. Default: "relu".
 
     Examples:
@@ -217,8 +216,8 @@ class ShuffleNetV2(Layer):
 
             from paddle.vision.models import ShuffleNetV2
 
+
             shufflenetv2_x0_25 = ShuffleNetV2(num_classes=1000, scale=0.25, act="relu")
-            shufflenetv2_swish = ShuffleNetV2(num_classes=1000, scale=1.0, act="swish")
             
     """    
     def __init__(self, num_classes=1000, scale=1.0, act="relu"):
@@ -322,8 +321,11 @@ def shufflenetv2_x0_25(pretrained=False, **kwargs):
         .. code-block:: python
 
             from paddle.vision.models import shufflenetv2_x0_25
+            
+            
             # build model
             model = shufflenetv2_x0_25()
+            
             # build model and load imagenet pretrained weight
             # model = shufflenetv2_x0_25(pretrained=True)
 
@@ -341,8 +343,11 @@ def shufflenetv2_x0_33(pretrained=False, **kwargs):
         .. code-block:: python
 
             from paddle.vision.models import shufflenetv2_x0_33
+            
+            
             # build model
             model = shufflenetv2_x0_33()
+            
             # build model and load imagenet pretrained weight
             # model = shufflenetv2_x0_33(pretrained=True)
 
@@ -360,8 +365,11 @@ def shufflenetv2_x0_5(pretrained=False, **kwargs):
         .. code-block:: python
 
             from paddle.vision.models import shufflenetv2_x0_5
+            
+            
             # build model
             model = shufflenetv2_x0_5()
+            
             # build model and load imagenet pretrained weight
             # model = shufflenetv2_x0_5(pretrained=True)
 
@@ -379,8 +387,11 @@ def shufflenetv2_x1_0(pretrained=False, **kwargs):
         .. code-block:: python
 
             from paddle.vision.models import shufflenetv2_x1_0
+            
+            
             # build model
             model = shufflenetv2_x1_0()
+            
             # build model and load imagenet pretrained weight
             # model = shufflenetv2_x1_0(pretrained=True)
 
@@ -398,8 +409,11 @@ def shufflenetv2_x1_5(pretrained=False, **kwargs):
         .. code-block:: python
 
             from paddle.vision.models import shufflenetv2_x1_5
+            
+            
             # build model
             model = shufflenetv2_x1_5()
+            
             # build model and load imagenet pretrained weight
             # model = shufflenetv2_x1_5(pretrained=True)
 
@@ -417,8 +431,11 @@ def shufflenetv2_x2_0(pretrained=False, **kwargs):
         .. code-block:: python
 
             from paddle.vision.models import shufflenetv2_x2_0
+            
+            
             # build model
             model = shufflenetv2_x2_0()
+            
             # build model and load imagenet pretrained weight
             # model = shufflenetv2_x2_0(pretrained=True)
 
@@ -436,8 +453,11 @@ def shufflenetv2_swish(pretrained=False, **kwargs):
         .. code-block:: python
 
             from paddle.vision.models import shufflenetv2_swish
+            
+            
             # build model
             model = shufflenetv2_swish()
+            
             # build model and load imagenet pretrained weight
             # model = shufflenetv2_swish(pretrained=True)
 

From e7a5f6a263607bea8491440399dc42ac20864ac4 Mon Sep 17 00:00:00 2001
From: Quan Tang <73922533+tt8000@users.noreply.github.com>
Date: Sat, 9 Oct 2021 20:06:09 +0800
Subject: [PATCH 13/17] Update shufflenetv2.py

---
 python/paddle/vision/models/shufflenetv2.py | 32 +++++++--------------
 1 file changed, 10 insertions(+), 22 deletions(-)

diff --git a/python/paddle/vision/models/shufflenetv2.py b/python/paddle/vision/models/shufflenetv2.py
index 6ad5d9c982379..64a2f0cb6b573 100644
--- a/python/paddle/vision/models/shufflenetv2.py
+++ b/python/paddle/vision/models/shufflenetv2.py
@@ -26,20 +26,13 @@
 
 
 MODEL_URLS = {
-    "shufflenetv2_x0_25":
-    "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x0_25_pretrained.pdparams",
-    "shufflenetv2_x0_33":
-    "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x0_33_pretrained.pdparams",
-    "shufflenetv2_x0_5":
-    "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x0_5_pretrained.pdparams",
-    "shufflenetv2_x1_0":
-    "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x1_0_pretrained.pdparams",
-    "shufflenetv2_x1_5":
-    "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x1_5_pretrained.pdparams",
-    "shufflenetv2_x2_0":
-    "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x2_0_pretrained.pdparams",
-    "shufflenetv2_swish":
-    "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_swish_pretrained.pdparams"
+    "shufflenetv2_x0_25": "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x0_25_pretrained.pdparams",
+    "shufflenetv2_x0_33": "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x0_33_pretrained.pdparams",
+    "shufflenetv2_x0_5": "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x0_5_pretrained.pdparams",
+    "shufflenetv2_x1_0": "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x1_0_pretrained.pdparams",
+    "shufflenetv2_x1_5": "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x1_5_pretrained.pdparams",
+    "shufflenetv2_x2_0": "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x2_0_pretrained.pdparams",
+    "shufflenetv2_swish": "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_swish_pretrained.pdparams"
 }
 
 __all__ = list(MODEL_URLS.keys())
@@ -306,8 +299,10 @@ def _shufflenetv2(arch, scale, pretrained=False, **kwargs):
     model = ShuffleNetV2(scale=scale, **kwargs)
     if pretrained:
         weight_path = get_weights_path_from_url(MODEL_URLS[arch])
+        
         param = paddle.load(weight_path)
-        model.set_dict(param)    
+        model.set_dict(param)  
+        
     return model    
 
 
@@ -322,7 +317,6 @@ def shufflenetv2_x0_25(pretrained=False, **kwargs):
 
             from paddle.vision.models import shufflenetv2_x0_25
             
-            
             # build model
             model = shufflenetv2_x0_25()
             
@@ -344,7 +338,6 @@ def shufflenetv2_x0_33(pretrained=False, **kwargs):
 
             from paddle.vision.models import shufflenetv2_x0_33
             
-            
             # build model
             model = shufflenetv2_x0_33()
             
@@ -366,7 +359,6 @@ def shufflenetv2_x0_5(pretrained=False, **kwargs):
 
             from paddle.vision.models import shufflenetv2_x0_5
             
-            
             # build model
             model = shufflenetv2_x0_5()
             
@@ -388,7 +380,6 @@ def shufflenetv2_x1_0(pretrained=False, **kwargs):
 
             from paddle.vision.models import shufflenetv2_x1_0
             
-            
             # build model
             model = shufflenetv2_x1_0()
             
@@ -410,7 +401,6 @@ def shufflenetv2_x1_5(pretrained=False, **kwargs):
 
             from paddle.vision.models import shufflenetv2_x1_5
             
-            
             # build model
             model = shufflenetv2_x1_5()
             
@@ -432,7 +422,6 @@ def shufflenetv2_x2_0(pretrained=False, **kwargs):
 
             from paddle.vision.models import shufflenetv2_x2_0
             
-            
             # build model
             model = shufflenetv2_x2_0()
             
@@ -454,7 +443,6 @@ def shufflenetv2_swish(pretrained=False, **kwargs):
 
             from paddle.vision.models import shufflenetv2_swish
             
-            
             # build model
             model = shufflenetv2_swish()
             

From 99681129755af6cc349a6664b0b3c4ced976a732 Mon Sep 17 00:00:00 2001
From: Quan Tang <73922533+tt8000@users.noreply.github.com>
Date: Sat, 9 Oct 2021 20:19:51 +0800
Subject: [PATCH 14/17] Update shufflenetv2.py

---
 python/paddle/vision/models/shufflenetv2.py | 48 ++++++++-------------
 1 file changed, 18 insertions(+), 30 deletions(-)

diff --git a/python/paddle/vision/models/shufflenetv2.py b/python/paddle/vision/models/shufflenetv2.py
index 64a2f0cb6b573..fd2db721fca13 100644
--- a/python/paddle/vision/models/shufflenetv2.py
+++ b/python/paddle/vision/models/shufflenetv2.py
@@ -55,15 +55,15 @@ def channel_shuffle(x, groups):
 
 
 class ConvBNLayer(Layer):
-    def __init__(
-            self,
-            in_channels,
-            out_channels,
-            kernel_size,
-            stride,
-            padding,
-            groups=1,
-            act=None):
+    
+    def __init__(self,
+                 in_channels,
+                 out_channels,
+                 kernel_size,
+                 stride,
+                 padding,
+                 groups=1,
+                 act=None):
         super(ConvBNLayer, self).__init__()
         self._conv = Conv2D(
             in_channels=in_channels,
@@ -72,8 +72,7 @@ def __init__(
             stride=stride,
             padding=padding,
             groups=groups,
-            weight_attr=ParamAttr(
-                initializer=KaimingNormal()),
+            weight_attr=ParamAttr(initializer=KaimingNormal()),
             bias_attr=False)
 
         self._batch_norm = BatchNorm(
@@ -89,11 +88,8 @@ def forward(self, inputs):
 
 
 class InvertedResidual(Layer):
-    def __init__(self,
-                 in_channels,
-                 out_channels,
-                 stride,
-                 act="relu"):
+    
+    def __init__(self, in_channels, out_channels, stride, act="relu"):
         super(InvertedResidual, self).__init__()
         self._conv_pw = ConvBNLayer(
             in_channels=in_channels // 2,
@@ -133,11 +129,8 @@ def forward(self, inputs):
 
 
 class InvertedResidualDS(Layer):
-    def __init__(self,
-                 in_channels,
-                 out_channels,
-                 stride,
-                 act="relu"):
+    
+    def __init__(self, in_channels, out_channels, stride, act="relu"):
         super(InvertedResidualDS, self).__init__()
 
         # branch1
@@ -213,6 +206,7 @@ class ShuffleNetV2(Layer):
             shufflenetv2_x0_25 = ShuffleNetV2(num_classes=1000, scale=0.25, act="relu")
             
     """    
+    
     def __init__(self, num_classes=1000, scale=1.0, act="relu"):
         super(ShuffleNetV2, self).__init__()
         self.scale = scale
@@ -322,7 +316,6 @@ def shufflenetv2_x0_25(pretrained=False, **kwargs):
             
             # build model and load imagenet pretrained weight
             # model = shufflenetv2_x0_25(pretrained=True)
-
     """
     return _shufflenetv2("shufflenetv2_x0_25", 0.25, pretrained, **kwargs)     
 
@@ -343,7 +336,6 @@ def shufflenetv2_x0_33(pretrained=False, **kwargs):
             
             # build model and load imagenet pretrained weight
             # model = shufflenetv2_x0_33(pretrained=True)
-
     """      
     return _shufflenetv2("ShuffleNetV2_x0_33", 0.33, pretrained, **kwargs)       
 
@@ -364,8 +356,7 @@ def shufflenetv2_x0_5(pretrained=False, **kwargs):
             
             # build model and load imagenet pretrained weight
             # model = shufflenetv2_x0_5(pretrained=True)
-
-    """       
+    """    
     return _shufflenetv2("shufflenetv2_x0_5", 0.5, pretrained, **kwargs) 
 
 
@@ -385,7 +376,6 @@ def shufflenetv2_x1_0(pretrained=False, **kwargs):
             
             # build model and load imagenet pretrained weight
             # model = shufflenetv2_x1_0(pretrained=True)
-
     """     
     return _shufflenetv2("shufflenetv2_x1_0", 1.0, pretrained, **kwargs)
 
@@ -406,7 +396,6 @@ def shufflenetv2_x1_5(pretrained=False, **kwargs):
             
             # build model and load imagenet pretrained weight
             # model = shufflenetv2_x1_5(pretrained=True)
-
     """       
     return _shufflenetv2("shufflenetv2_x1_5", 1.5, pretrained, **kwargs)
 
@@ -427,7 +416,6 @@ def shufflenetv2_x2_0(pretrained=False, **kwargs):
             
             # build model and load imagenet pretrained weight
             # model = shufflenetv2_x2_0(pretrained=True)
-
     """    
     return _shufflenetv2("shufflenetv2_x2_0", 2.0, pretrained, **kwargs)    
 
@@ -448,6 +436,6 @@ def shufflenetv2_swish(pretrained=False, **kwargs):
             
             # build model and load imagenet pretrained weight
             # model = shufflenetv2_swish(pretrained=True)
-
     """
-    return _shufflenetv2("shufflenetv2_swish", 1.0, pretrained, act="swish", **kwargs) 
+    return _shufflenetv2(
+        "shufflenetv2_swish", 1.0, pretrained, act="swish", **kwargs) 

From e037ef53a2c695e3a25c87c4158b315bbcd7b3d1 Mon Sep 17 00:00:00 2001
From: Quan Tang <73922533+tt8000@users.noreply.github.com>
Date: Sat, 9 Oct 2021 20:21:24 +0800
Subject: [PATCH 15/17] Update test_pretrained_model.py

---
 python/paddle/tests/test_pretrained_model.py | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/python/paddle/tests/test_pretrained_model.py b/python/paddle/tests/test_pretrained_model.py
index eb3e102a832c3..1cc9e1f2d3455 100644
--- a/python/paddle/tests/test_pretrained_model.py
+++ b/python/paddle/tests/test_pretrained_model.py
@@ -52,8 +52,9 @@ def infer(self, arch):
         np.testing.assert_allclose(res['dygraph'], res['static'])
 
     def test_models(self):
-        arches = ['mobilenet_v1', 'mobilenet_v2', 'resnet18', 'vgg16',
-                  'shufflenetv2_swish'
+        arches = [
+            'mobilenet_v1', 'mobilenet_v2', 'resnet18', 'vgg16',
+            'shufflenetv2_swish'
         ]
         for arch in arches:
             self.infer(arch)

From 571535f49ef53337ba4acd70557ecf5b521db567 Mon Sep 17 00:00:00 2001
From: Quan Tang <73922533+tt8000@users.noreply.github.com>
Date: Sat, 9 Oct 2021 20:25:32 +0800
Subject: [PATCH 16/17] Update test_vision_models.py

---
 python/paddle/tests/test_vision_models.py | 23 ++++++++++++-----------
 1 file changed, 12 insertions(+), 11 deletions(-)

diff --git a/python/paddle/tests/test_vision_models.py b/python/paddle/tests/test_vision_models.py
index a6c88075f974c..851553b2bb001 100644
--- a/python/paddle/tests/test_vision_models.py
+++ b/python/paddle/tests/test_vision_models.py
@@ -70,17 +70,6 @@ def test_resnet101(self):
 
     def test_resnet152(self):
         self.models_infer('resnet152')
-
-    def test_vgg16_num_classes(self):
-        vgg16 = models.__dict__['vgg16'](pretrained=False, num_classes=10)
-
-    def test_lenet(self):
-        input = InputSpec([None, 1, 28, 28], 'float32', 'x')
-        lenet = paddle.Model(models.__dict__['LeNet'](), input)
-        lenet.prepare()
-
-        x = np.array(np.random.random((2, 1, 28, 28)), dtype=np.float32)
-        lenet.predict_batch(x)
         
     def test_shufflenetv2_x0_25(self):
         self.models_infer('shufflenetv2_x0_25')
@@ -103,5 +92,17 @@ def test_shufflenetv2_x2_0(self):
     def test_shufflenetv2_swish(self):
         self.models_infer('shufflenetv2_swish')        
 
+    def test_vgg16_num_classes(self):
+        vgg16 = models.__dict__['vgg16'](pretrained=False, num_classes=10)
+
+    def test_lenet(self):
+        input = InputSpec([None, 1, 28, 28], 'float32', 'x')
+        lenet = paddle.Model(models.__dict__['LeNet'](), input)
+        lenet.prepare()
+
+        x = np.array(np.random.random((2, 1, 28, 28)), dtype=np.float32)
+        lenet.predict_batch(x)
+            
+
 if __name__ == '__main__':
     unittest.main()

From 17b660c0827bb7c2a24d1a17fe18349ad6702513 Mon Sep 17 00:00:00 2001
From: Quan Tang <73922533+tt8000@users.noreply.github.com>
Date: Sat, 9 Oct 2021 21:10:39 +0800
Subject: [PATCH 17/17] Update shufflenetv2.py

---
 python/paddle/vision/models/shufflenetv2.py | 24 ++++++++++++---------
 1 file changed, 14 insertions(+), 10 deletions(-)

diff --git a/python/paddle/vision/models/shufflenetv2.py b/python/paddle/vision/models/shufflenetv2.py
index fd2db721fca13..498e6e35fef53 100644
--- a/python/paddle/vision/models/shufflenetv2.py
+++ b/python/paddle/vision/models/shufflenetv2.py
@@ -26,13 +26,20 @@
 
 
 MODEL_URLS = {
-    "shufflenetv2_x0_25": "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x0_25_pretrained.pdparams",
-    "shufflenetv2_x0_33": "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x0_33_pretrained.pdparams",
-    "shufflenetv2_x0_5": "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x0_5_pretrained.pdparams",
-    "shufflenetv2_x1_0": "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x1_0_pretrained.pdparams",
-    "shufflenetv2_x1_5": "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x1_5_pretrained.pdparams",
-    "shufflenetv2_x2_0": "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x2_0_pretrained.pdparams",
-    "shufflenetv2_swish": "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_swish_pretrained.pdparams"
+    "shufflenetv2_x0_25":
+    "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x0_25_pretrained.pdparams",
+    "shufflenetv2_x0_33":
+    "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x0_33_pretrained.pdparams",
+    "shufflenetv2_x0_5":
+    "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x0_5_pretrained.pdparams",
+    "shufflenetv2_x1_0":
+    "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x1_0_pretrained.pdparams",
+    "shufflenetv2_x1_5":
+    "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x1_5_pretrained.pdparams",
+    "shufflenetv2_x2_0":
+    "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_x2_0_pretrained.pdparams",
+    "shufflenetv2_swish":
+    "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ShuffleNetV2_swish_pretrained.pdparams"
 }
 
 __all__ = list(MODEL_URLS.keys())
@@ -55,7 +62,6 @@ def channel_shuffle(x, groups):
 
 
 class ConvBNLayer(Layer):
-    
     def __init__(self,
                  in_channels,
                  out_channels,
@@ -88,7 +94,6 @@ def forward(self, inputs):
 
 
 class InvertedResidual(Layer):
-    
     def __init__(self, in_channels, out_channels, stride, act="relu"):
         super(InvertedResidual, self).__init__()
         self._conv_pw = ConvBNLayer(
@@ -129,7 +134,6 @@ def forward(self, inputs):
 
 
 class InvertedResidualDS(Layer):
-    
     def __init__(self, in_channels, out_channels, stride, act="relu"):
         super(InvertedResidualDS, self).__init__()