In [2]:
import torch
import torch.nn as nn
from torch.quantization import fuse_modules
import os

__all__ = ['ResNet', 'resnet18', 'resnet34', 'resnet50', 'resnet101',
           'resnet152', 'resnext50_32x4d', 'resnext101_32x8d',
           'wide_resnet50_2', 'wide_resnet101_2']


model_urls = {
    'resnet18': 'https://download.pytorch.org/models/resnet18-5c106cde.pth',
    'resnet34': 'https://download.pytorch.org/models/resnet34-333f7ec4.pth',
    'resnet50': 'https://download.pytorch.org/models/resnet50-19c8e357.pth',
    'resnet101': 'https://download.pytorch.org/models/resnet101-5d3b4d8f.pth',
    'resnet152': 'https://download.pytorch.org/models/resnet152-b121ed2d.pth',
    'resnext50_32x4d': 'https://download.pytorch.org/models/resnext50_32x4d-7cdf4587.pth',
    'resnext101_32x8d': 'https://download.pytorch.org/models/resnext101_32x8d-8ba56ff5.pth',
    'wide_resnet50_2': 'https://download.pytorch.org/models/wide_resnet50_2-95faca4d.pth',
    'wide_resnet101_2': 'https://download.pytorch.org/models/wide_resnet101_2-32ee1156.pth',
}


def conv3x3(in_planes, out_planes, stride=1, groups=1, dilation=1):
    """3x3 convolution with padding"""
    return nn.Conv2d(in_planes, out_planes, kernel_size=3, stride=stride,
                     padding=dilation, groups=groups, bias=False, dilation=dilation)


def conv1x1(in_planes, out_planes, stride=1):
    """1x1 convolution"""
    return nn.Conv2d(in_planes, out_planes, kernel_size=1, stride=stride, bias=False)


class BasicBlock(nn.Module):
    expansion = 1
    __constants__ = ['downsample']

    def __init__(self, inplanes, planes, stride=1, downsample=None, groups=1,
                 base_width=64, dilation=1, norm_layer=None):
        super(BasicBlock, self).__init__()
        if norm_layer is None:
            norm_layer = nn.BatchNorm2d
        if groups != 1 or base_width != 64:
            raise ValueError('BasicBlock only supports groups=1 and base_width=64')
        if dilation > 1:
            raise NotImplementedError("Dilation > 1 not supported in BasicBlock")
        # Both self.conv1 and self.downsample layers downsample the input when stride != 1
        self.conv1 = conv3x3(inplanes, planes, stride)
        self.bn1 = norm_layer(planes)
        self.relu = nn.ReLU(inplace=True)
        self.conv2 = conv3x3(planes, planes)
        self.bn2 = norm_layer(planes)
        self.downsample = downsample
        self.stride = stride

    def forward(self, x):
        identity = x

        out = self.conv1(x)
        out = self.bn1(out)
        out = self.relu(out)

        out = self.conv2(out)
        out = self.bn2(out)

        if self.downsample is not None:
            identity = self.downsample(x)

        out += identity
        out = self.relu(out)

        return out


class Bottleneck(nn.Module):
    expansion = 4
    __constants__ = ['downsample']

    def __init__(self, inplanes, planes, stride=1, downsample=None, groups=1,
                 base_width=64, dilation=1, norm_layer=None):
        super(Bottleneck, self).__init__()
        if norm_layer is None:
            norm_layer = nn.BatchNorm2d
        width = int(planes * (base_width / 64.)) * groups
        # Both self.conv2 and self.downsample layers downsample the input when stride != 1
        self.conv1 = conv1x1(inplanes, width)
        self.bn1 = norm_layer(width)
        self.conv2 = conv3x3(width, width, stride, groups, dilation)
        self.bn2 = norm_layer(width)
        self.conv3 = conv1x1(width, planes * self.expansion)
        self.bn3 = norm_layer(planes * self.expansion)
        self.relu = nn.ReLU(inplace=True)
        self.downsample = downsample
        self.stride = stride

    def forward(self, x):
        identity = x

        out = self.conv1(x)
        out = self.bn1(out)
        out = self.relu(out)

        out = self.conv2(out)
        out = self.bn2(out)
        out = self.relu(out)

        out = self.conv3(out)
        out = self.bn3(out)

        if self.downsample is not None:
            identity = self.downsample(x)

        out += identity
        out = self.relu(out)

        return out


class ResNet(nn.Module):

    def __init__(self, block, layers, num_classes=1000, zero_init_residual=False,
                 groups=1, width_per_group=64, replace_stride_with_dilation=None,
                 norm_layer=None):
        super(ResNet, self).__init__()
        if norm_layer is None:
            norm_layer = nn.BatchNorm2d
        self._norm_layer = norm_layer

        self.inplanes = 64
        self.dilation = 1
        if replace_stride_with_dilation is None:
            # each element in the tuple indicates if we should replace
            # the 2x2 stride with a dilated convolution instead
            replace_stride_with_dilation = [False, False, False]
        if len(replace_stride_with_dilation) != 3:
            raise ValueError("replace_stride_with_dilation should be None "
                             "or a 3-element tuple, got {}".format(replace_stride_with_dilation))
        self.groups = groups
        self.base_width = width_per_group
        self.conv1 = nn.Conv2d(3, self.inplanes, kernel_size=7, stride=2, padding=3,
                               bias=False)
        self.bn1 = norm_layer(self.inplanes)
        self.relu = nn.ReLU(inplace=True)
        self.maxpool = nn.MaxPool2d(kernel_size=3, stride=2, padding=1)
        self.layer1 = self._make_layer(block, 64, layers[0])
        self.layer2 = self._make_layer(block, 128, layers[1], stride=2,
                                       dilate=replace_stride_with_dilation[0])
        self.layer3 = self._make_layer(block, 256, layers[2], stride=2,
                                       dilate=replace_stride_with_dilation[1])
        self.layer4 = self._make_layer(block, 512, layers[3], stride=2,
                                       dilate=replace_stride_with_dilation[2])
        self.avgpool = nn.AdaptiveAvgPool2d((1, 1))
        self.fc = nn.Linear(512 * block.expansion, num_classes)

        for m in self.modules():
            if isinstance(m, nn.Conv2d):
                nn.init.kaiming_normal_(m.weight, mode='fan_out', nonlinearity='relu')
            elif isinstance(m, (nn.BatchNorm2d, nn.GroupNorm)):
                nn.init.constant_(m.weight, 1)
                nn.init.constant_(m.bias, 0)

        # Zero-initialize the last BN in each residual branch,
        # so that the residual branch starts with zeros, and each residual block behaves like an identity.
        # This improves the model by 0.2~0.3% according to https://arxiv.org/abs/1706.02677
        if zero_init_residual:
            for m in self.modules():
                if isinstance(m, Bottleneck):
                    nn.init.constant_(m.bn3.weight, 0)
                elif isinstance(m, BasicBlock):
                    nn.init.constant_(m.bn2.weight, 0)

    def _make_layer(self, block, planes, blocks, stride=1, dilate=False):
        norm_layer = self._norm_layer
        downsample = None
        previous_dilation = self.dilation
        if dilate:
            self.dilation *= stride
            stride = 1
        if stride != 1 or self.inplanes != planes * block.expansion:
            downsample = nn.Sequential(
                conv1x1(self.inplanes, planes * block.expansion, stride),
                norm_layer(planes * block.expansion),
            )

        layers = []
        layers.append(block(self.inplanes, planes, stride, downsample, self.groups,
                            self.base_width, previous_dilation, norm_layer))
        self.inplanes = planes * block.expansion
        for _ in range(1, blocks):
            layers.append(block(self.inplanes, planes, groups=self.groups,
                                base_width=self.base_width, dilation=self.dilation,
                                norm_layer=norm_layer))

        return nn.Sequential(*layers)

    def _forward(self, x):
        x = self.conv1(x)
        x = self.bn1(x)
        x = self.relu(x)
        x = self.maxpool(x)

        x = self.layer1(x)
        x = self.layer2(x)
        x = self.layer3(x)
        x = self.layer4(x)

        x = self.avgpool(x)
        x = torch.flatten(x, 1)
        x = self.fc(x)

        return x

    # Allow for accessing forward method in a inherited class
    forward = _forward
    
def _replace_relu(module):
    reassign = {}
    for name, mod in module.named_children():
        _replace_relu(mod)
        # Checking for explicit type instead of instance
        # as we only want to replace modules of the exact type
        # not inherited classes
        if type(mod) == nn.ReLU or type(mod) == nn.ReLU6:
            reassign[name] = nn.ReLU(inplace=False)

    for key, value in reassign.items(): 
        module._modules[key] = value
    
    
class QuantizableBottleneck(Bottleneck):
    def __init__(self, *args, **kwargs):
        super(QuantizableBottleneck, self).__init__(*args, **kwargs)
        self.skip_add_relu = nn.quantized.FloatFunctional()
        self.relu1 = nn.ReLU(inplace=False)
        self.relu2 = nn.ReLU(inplace=False)

    def forward(self, x):
        identity = x
        out = self.conv1(x)
        out = self.bn1(out)
        out = self.relu1(out)
        out = self.conv2(out)
        out = self.bn2(out)
        out = self.relu2(out)

        out = self.conv3(out)
        out = self.bn3(out)

        if self.downsample is not None:
            identity = self.downsample(x)
        out = self.skip_add_relu.add_relu(out, identity)

        return out

    def fuse_model(self):
        fuse_modules(self, [['conv1', 'bn1', 'relu1'],
                            ['conv2', 'bn2', 'relu2'],
                            ['conv3', 'bn3']], inplace=True)
        if self.downsample:
            torch.quantization.fuse_modules(self.downsample, ['0', '1'], inplace=True)


class QuantizableResNet(ResNet):

    def __init__(self, *args, **kwargs):
        super(QuantizableResNet, self).__init__(*args, **kwargs)

        self.quant = torch.quantization.QuantStub()
        self.dequant = torch.quantization.DeQuantStub()

    def forward(self, x):
        x = self.quant(x)
        # Ensure scriptability
        # super(QuantizableResNet,self).forward(x)
        # is not scriptable
        x = self._forward(x)
        x = self.dequant(x)
        return x

    def fuse_model(self):
        r"""Fuse conv/bn/relu modules in resnet models
        Fuse conv+bn+relu/ Conv+relu/conv+Bn modules to prepare for quantization.
        Model is modified in place.  Note that this operation does not change numerics
        and the model after modification is in floating point
        """

        fuse_modules(self, ['conv1', 'bn1', 'relu'], inplace=True)
        for m in self.modules():
            if type(m) == QuantizableBottleneck:
                m.fuse_model()

def print_size_of_model(model):
    torch.save(model.state_dict(), "temp.p")
    print('Size (MB):', os.path.getsize("temp.p")/1e6)
    os.remove('temp.p')
In [6]:
model = QuantizableResNet(QuantizableBottleneck, [3, 4, 23, 3])
_replace_relu(model)
model.fuse_model()
print_size_of_model(model)
model.qconfig = torch.quantization.default_qconfig
torch.quantization.prepare(model, inplace=True)
quantized_net = torch.quantization.convert(model)
print_size_of_model(quantized_net)
Size (MB): 178.742957
Size (MB): 172.638462
In [8]:
model = QuantizableResNet(QuantizableBottleneck, [3, 4, 6, 3])
model.eval()
_replace_relu(model)
model.fuse_model()
print_size_of_model(model)
model.qconfig = torch.quantization.default_qconfig
torch.quantization.prepare(model, inplace=True)
quantized_net = torch.quantization.convert(model)
print_size_of_model(quantized_net)
---------------------------------------------------------------------------
AssertionError                            Traceback (most recent call last)
<ipython-input-8-4490ef2d57fa> in <module>
      2 model.eval()
      3 _replace_relu(model)
----> 4 model.fuse_model()
      5 print_size_of_model(model)
      6 model.qconfig = torch.quantization.default_qconfig

<ipython-input-2-b2f8ebab6dfa> in fuse_model(self)
    286         """
    287 
--> 288         fuse_modules(self, ['conv1', 'bn1', 'relu'], inplace=True)
    289         for m in self.modules():
    290             if type(m) == QuantizableBottleneck:

~/anaconda3/envs/test/lib/python3.7/site-packages/torch/quantization/fuse_modules.py in fuse_modules(model, modules_to_fuse, inplace, fuser_func)
    164     if all(isinstance(module_element, str) for module_element in modules_to_fuse):
    165         # Handle case of modules_to_fuse being a list
--> 166         _fuse_modules(model, modules_to_fuse, fuser_func)
    167     else:
    168         # Handle case of modules_to_fuse being a list of lists

~/anaconda3/envs/test/lib/python3.7/site-packages/torch/quantization/fuse_modules.py in _fuse_modules(model, modules_to_fuse, fuser_func)
    112 
    113     # Fuse list of modules
--> 114     new_mod_list = fuser_func(mod_list)
    115 
    116     # Replace original module list with fused module list

~/anaconda3/envs/test/lib/python3.7/site-packages/torch/quantization/fuse_modules.py in fuse_known_modules(mod_list)
     97         raise NotImplementedError("Cannot fuse modules: {}".format(types))
     98     new_mod = [None] * len(mod_list)
---> 99     new_mod[0] = fuser_method(*mod_list)
    100 
    101     for i in range(1, len(mod_list)):

~/anaconda3/envs/test/lib/python3.7/site-packages/torch/quantization/fuse_modules.py in fuse_conv_bn_relu(conv, bn, relu)
     45     """
     46     assert(conv.training == bn.training == relu.training),\
---> 47         "Conv and BN both must be in the same mode (train or eval)."
     48 
     49     if conv.training:

AssertionError: Conv and BN both must be in the same mode (train or eval).