Add ShuffleNet

cnn/imagenet/training.py

@@ -17,6 +17,7 @@
 
 from cnn.mobilenet_imagenet import MobileNet
 from cnn.mobilenet_imagenet import Butterfly1x1Conv
+from cnn.shufflenet_imagenet import ShuffleNet
 
 try:
     from apex.parallel import DistributedDataParallel as DDP
@@ -28,7 +29,8 @@
 
 class ModelAndLoss(nn.Module):
     def __init__(self, arch, loss, pretrained_weights=None, cuda=True, fp16=False,
-                 width=1.0, n_struct_layers=0, struct='D', softmax_struct='D', sm_pooling=1):
+                 width=1.0, n_struct_layers=0, struct='D', softmax_struct='D', sm_pooling=1,
+                 groups=8):
         super(ModelAndLoss, self).__init__()
         self.arch = arch
 
@@ -39,6 +41,8 @@ def __init__(self, arch, loss, pretrained_weights=None, cuda=True, fp16=False,
                               softmax_structure=softmax_struct, sm_pooling=sm_pooling)
             # if args.distilled_param_path:
             #     model.load_state_dict(model.mixed_model_state_dict(args.full_model_path, args.distilled_param_path))
+        elif arch == 'shufflenetv1':
+            model = ShuffleNet(width_mult=width, groups=groups)
         else:
             model = models.__dict__[arch]()
         if pretrained_weights is not None:

cnn/imagenet_main.py

@@ -37,7 +37,7 @@
 
 
 def add_parser_arguments(parser):
-    custom_model_names = ['mobilenetv1']
+    custom_model_names = ['mobilenetv1', 'shufflenetv1']
     model_names = sorted(name for name in models.__dict__
                         if name.islower() and not name.startswith("__")
                         and callable(models.__dict__[name])) + custom_model_names
@@ -66,6 +66,8 @@ def add_parser_arguments(parser):
                         metavar='NSL', help='Number of structured layer (default 7)')
     parser.add_argument('--width', default=1.0, type=float,
                         metavar='WIDTH', help='Width multiplier of the CNN (default 1.0)')
+    parser.add_argument('--groups', default=8, type=int,
+                        metavar='GROUPS', help='Group parameter of ShuffleNet (default 8)')
     parser.add_argument('--distilled-param-path', default='', type=str, metavar='PATH',
                         help='path to distilled parameters (default: none)')
     parser.add_argument('--full-model-path', default='', type=str, metavar='PATH',
@@ -249,7 +251,8 @@ def _worker_init_fn(id):
             pretrained_weights=pretrained_weights,
             cuda = True, fp16 = args.fp16,
             width=args.width, n_struct_layers=args.n_struct_layers,
-            struct=args.struct, softmax_struct=args.softmax_struct, sm_pooling=args.sm_pooling)
+            struct=args.struct, softmax_struct=args.softmax_struct, sm_pooling=args.sm_pooling,
+            groups=args.groups)
 
     if args.arch == 'mobilenetv1' and args.distilled_param_path:
         model_state = model_and_loss.model.mixed_model_state_dict(args.full_model_path, args.distilled_param_path)

cnn/shufflenet_imagenet.py

@@ -0,0 +1,110 @@
+'''ShuffleNet in PyTorch.
+
+See the paper "ShuffleNet: An Extremely Efficient Convolutional Neural Network for Mobile Devices" for more details.
+'''
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+
+import os, sys
+project_root = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
+sys.path.insert(0, project_root)
+from butterfly import Butterfly
+
+from cnn.mobilenet_imagenet import _make_divisible
+from cnn.mobilenet_imagenet import Butterfly1x1Conv
+
+
+class ShuffleBlock(nn.Module):
+    def __init__(self, groups):
+        super(ShuffleBlock, self).__init__()
+        self.groups = groups
+
+    def forward(self, x):
+        '''Channel shuffle: [N,C,H,W] -> [N,g,C/g,H,W] -> [N,C/g,g,H,w] -> [N,C,H,W]'''
+        N,C,H,W = x.size()
+        g = self.groups
+        return x.view(N,g,C//g,H,W).permute(0,2,1,3,4).contiguous().view(N,C,H,W)
+
+
+class Bottleneck(nn.Module):
+    def __init__(self, in_planes, out_planes, stride, groups, grouped_conv_1st_layer=True):
+        super(Bottleneck, self).__init__()
+        self.stride = stride
+
+        mid_planes = _make_divisible(out_planes // 4, groups)
+        if stride == 2:  # Reduce out_planes due to concat
+            out_planes -= in_planes
+        g = groups if grouped_conv_1st_layer else 1  # No grouped conv for the first layer of stage 2
+        self.conv1 = nn.Conv2d(in_planes, mid_planes, kernel_size=1, groups=g, bias=False)
+        self.bn1 = nn.BatchNorm2d(mid_planes)
+        self.shuffle1 = ShuffleBlock(groups=g)
+        self.conv2 = nn.Conv2d(mid_planes, mid_planes, kernel_size=3, stride=stride, padding=1, groups=mid_planes, bias=False)
+        self.conv2.weight._no_wd = True
+        self.bn2 = nn.BatchNorm2d(mid_planes)
+        self.conv3 = nn.Conv2d(mid_planes, out_planes, kernel_size=1, groups=groups, bias=False)
+        self.bn3 = nn.BatchNorm2d(out_planes)
+
+        self.shortcut = nn.Sequential()
+        if stride == 2:
+            self.shortcut = nn.Sequential(nn.AvgPool2d(3, stride=2, padding=1))
+
+    def forward(self, x):
+        out = F.relu(self.bn1(self.conv1(x)), inplace=True)
+        out = self.shuffle1(out)
+        out = F.relu(self.bn2(self.conv2(out)), inplace=True)
+        out = self.bn3(self.conv3(out))
+        res = self.shortcut(x)
+        out = F.relu(torch.cat([out,res], 1), inplace=True) if self.stride==2 else F.relu(out+res, inplace=True)
+        return out
+
+
+class ShuffleNet(nn.Module):
+    def __init__(self, num_classes=1000, groups=8, width_mult=1.0):
+        super(ShuffleNet, self).__init__()
+        num_blocks = [4, 8, 4]
+        groups_to_outplanes = {1: [144, 288, 576],
+                               2: [200, 400, 800],
+                               3: [240, 480, 960],
+                               4: [272, 544, 1088],
+                               8: [384, 768, 1536]}
+        out_planes = groups_to_outplanes[groups]
+        out_planes = [_make_divisible(p * width_mult, groups) for p in out_planes]
+
+        input_channel = _make_divisible(24 * width_mult, groups)
+        self.conv1 = nn.Conv2d(3, input_channel, kernel_size=3, stride=2, padding=1, bias=False)
+        self.maxpool = nn.MaxPool2d(kernel_size=3, stride=2, padding=1)
+        self.bn1 = nn.BatchNorm2d(input_channel)
+        self.in_planes = input_channel
+        self.stage2 = self._make_layer(out_planes[0], num_blocks[0], groups, grouped_conv_1st_layer=False)
+        self.stage3 = self._make_layer(out_planes[1], num_blocks[1], groups)
+        self.stage4 = self._make_layer(out_planes[2], num_blocks[2], groups)
+        self.linear = nn.Linear(out_planes[2], num_classes)
+
+    def _make_layer(self, out_planes, num_blocks, groups, grouped_conv_1st_layer=True):
+        layers = []
+        for i in range(num_blocks):
+            stride = 2 if i == 0 else 1
+            layers.append(Bottleneck(self.in_planes, out_planes, stride=stride, groups=groups,
+                                     grouped_conv_1st_layer=grouped_conv_1st_layer))
+            self.in_planes = out_planes
+        return nn.Sequential(*layers)
+
+    def forward(self, x):
+        out = F.relu(self.bn1(self.conv1(x)), inplace=True)
+        out = self.maxpool(out)
+        out = self.stage2(out)
+        out = self.stage3(out)
+        out = self.stage4(out)
+        out = out.mean([2, 3])
+        out = self.linear(out)
+        return out
+
+
+def test():
+    net = ShuffleNet()
+    x = torch.randn(1, 3, 224, 224)
+    y = net(x)
+    print(y)
+
+# test()