diff --git a/distiller/models/__init__.py b/distiller/models/__init__.py
index bedc28a5f12c7f1252150636da8aaf1aadd8be2e..8f94bc90e01a2c142a6b69f9935d3132f507a56b 100755
--- a/distiller/models/__init__.py
+++ b/distiller/models/__init__.py
@@ -36,7 +36,8 @@ SUPPORTED_DATASETS = ('imagenet', 'cifar10', 'mnist')
# ResNet special treatment: we have our own version of ResNet, so we need to over-ride
# TorchVision's version.
-RESNET_SYMS = ('ResNet', 'resnet18', 'resnet34', 'resnet50', 'resnet101', 'resnet152')
+RESNET_SYMS = ('ResNet', 'resnet18', 'resnet34', 'resnet50', 'resnet101', 'resnet152',
+ 'resnext50_32x4d', 'resnext101_32x8d')
TORCHVISION_MODEL_NAMES = sorted(
name for name in torch_models.__dict__
diff --git a/distiller/models/imagenet/resnet.py b/distiller/models/imagenet/resnet.py
index 4e629dfd377669b7b82bf70fb6995d3463cda3a0..6f0fd7ab11ca35492d2432d309dd9083f0b7c2c2 100755
--- a/distiller/models/imagenet/resnet.py
+++ b/distiller/models/imagenet/resnet.py
@@ -1,5 +1,5 @@
#
-# Copyright (c) 2018 Intel Corporation
+# Copyright (c) 2019 Intel Corporation
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
@@ -14,55 +14,52 @@
# limitations under the License.
#
-# This is the same code as in https://github.com/pytorch/vision/blob/master/torchvision/models/resnet.py
-# However, it contains one type of change: whenever a ReLU module is used, we make sure to use a different
-# instance. This is necessary when we want to collect activation statistics.
+# The TorchVision implementation in https://github.com/pytorch/vision/blob/master/torchvision/models/resnet.py
+# has 2 issues in the implementation of the BasicBlock and Bottleneck modules, which impact our ability to
+# collect activation statistics and run quantization:
+# 1. Re-used ReLU modules
+# 2. Element-wise addition as a direct tensor operation
+# Here we provide an implementation of both classes that fixes these issues, and we provide the same API to create
+# ResNet and ResNeXt models as in the TorchVision implementation.
+# We reuse the original implementation as much as possible.
+from collections import OrderedDict
import torch.nn as nn
-import math
-import torch.utils.model_zoo as model_zoo
+from torchvision.models.resnet import ResNet, BasicBlock, Bottleneck, _resnet
from distiller.modules import EltwiseAdd
__all__ = ['ResNet', 'resnet18', 'resnet34', 'resnet50', 'resnet101',
- 'resnet152']
+ 'resnet152', 'resnext50_32x4d', 'resnext101_32x8d']
-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',
-}
+class DistillerBasicBlock(BasicBlock):
+ def __init__(self, *args, **kwargs):
+ # Initialize torchvision version
+ super(DistillerBasicBlock, self).__init__(*args, **kwargs)
-
-def conv3x3(in_planes, out_planes, stride=1):
- """3x3 convolution with padding"""
- return nn.Conv2d(in_planes, out_planes, kernel_size=3, stride=stride,
- padding=1, bias=False)
-
-
-class BasicBlock(nn.Module):
- expansion = 1
-
- def __init__(self, inplanes, planes, stride=1, downsample=None):
- super(BasicBlock, self).__init__()
- self.conv1 = conv3x3(inplanes, planes, stride)
- self.bn1 = nn.BatchNorm2d(planes)
+ # Remove original relu in favor of numbered modules
+ delattr(self, 'relu')
self.relu1 = nn.ReLU(inplace=True)
- self.conv2 = conv3x3(planes, planes)
- self.bn2 = nn.BatchNorm2d(planes)
self.relu2 = nn.ReLU(inplace=True)
- self.downsample = downsample
- self.stride = stride
-
- # Replace '+=' operator with inplace module
- self.add = EltwiseAdd(inplace=True)
+ self.add = EltwiseAdd(inplace=True) # Replace '+=' operator with inplace module
+
+ # Trick to make the modules accessible in their topological order
+ modules = OrderedDict()
+ modules['conv1'] = self.conv1
+ modules['bn1'] = self.bn1
+ modules['relu1'] = self.relu1
+ modules['conv2'] = self.conv2
+ modules['bn2'] = self.bn2
+ if self.downsample is not None:
+ modules['downsample'] = self.downsample
+ modules['add'] = self.add
+ modules['relu2'] = self.relu2
+ self._modules = modules
def forward(self, x):
- residual = x
+ identity = x
out = self.conv1(x)
out = self.bn1(out)
@@ -72,38 +69,44 @@ class BasicBlock(nn.Module):
out = self.bn2(out)
if self.downsample is not None:
- residual = self.downsample(x)
+ identity = self.downsample(x)
- # out += residual
- out = self.add(out, residual)
+ out = self.add(out, identity)
out = self.relu2(out)
return out
-class Bottleneck(nn.Module):
- expansion = 4
+class DistillerBottleneck(Bottleneck):
+ def __init__(self, *args, **kwargs):
+ # Initialize torchvision version
+ super(DistillerBottleneck, self).__init__(*args, **kwargs)
- def __init__(self, inplanes, planes, stride=1, downsample=None):
- super(Bottleneck, self).__init__()
- self.conv1 = nn.Conv2d(inplanes, planes, kernel_size=1, bias=False)
- self.bn1 = nn.BatchNorm2d(planes)
+ # Remove original relu in favor of numbered modules
+ delattr(self, 'relu')
self.relu1 = nn.ReLU(inplace=True)
- self.conv2 = nn.Conv2d(planes, planes, kernel_size=3, stride=stride,
- padding=1, bias=False)
- self.bn2 = nn.BatchNorm2d(planes)
self.relu2 = nn.ReLU(inplace=True)
- self.conv3 = nn.Conv2d(planes, planes * 4, kernel_size=1, bias=False)
- self.bn3 = nn.BatchNorm2d(planes * 4)
self.relu3 = nn.ReLU(inplace=True)
- self.downsample = downsample
- self.stride = stride
-
- # Replace '+=' operator with inplace module
- self.add = EltwiseAdd(inplace=True)
+ self.add = EltwiseAdd(inplace=True) # Replace '+=' operator with inplace module
+
+ # Trick to make the modules accessible in their topological order
+ modules = OrderedDict()
+ modules['conv1'] = self.conv1
+ modules['bn1'] = self.bn1
+ modules['relu1'] = self.relu1
+ modules['conv2'] = self.conv2
+ modules['bn2'] = self.bn2
+ modules['relu2'] = self.relu2
+ modules['conv3'] = self.conv3
+ modules['bn3'] = self.bn3
+ if self.downsample is not None:
+ modules['downsample'] = self.downsample
+ modules['add'] = self.add
+ modules['relu3'] = self.relu3
+ self._modules = modules
def forward(self, x):
- residual = x
+ identity = x
out = self.conv1(x)
out = self.bn1(out)
@@ -117,130 +120,90 @@ class Bottleneck(nn.Module):
out = self.bn3(out)
if self.downsample is not None:
- residual = self.downsample(x)
+ identity = self.downsample(x)
- # out += residual
- out = self.add(out, residual)
+ out = self.add(out, identity)
out = self.relu3(out)
return out
-class ResNet(nn.Module):
-
- def __init__(self, block, layers, num_classes=1000):
- self.inplanes = 64
- super(ResNet, self).__init__()
- self.conv1 = nn.Conv2d(3, 64, kernel_size=7, stride=2, padding=3,
- bias=False)
- self.bn1 = nn.BatchNorm2d(64)
- 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)
- self.layer3 = self._make_layer(block, 256, layers[2], stride=2)
- self.layer4 = self._make_layer(block, 512, layers[3], stride=2)
- self.avgpool = nn.AvgPool2d(7, stride=1)
- self.fc = nn.Linear(512 * block.expansion, num_classes)
-
- for m in self.modules():
- if isinstance(m, nn.Conv2d):
- n = m.kernel_size[0] * m.kernel_size[1] * m.out_channels
- m.weight.data.normal_(0, math.sqrt(2. / n))
- elif isinstance(m, nn.BatchNorm2d):
- m.weight.data.fill_(1)
- m.bias.data.zero_()
-
- def _make_layer(self, block, planes, blocks, stride=1):
- downsample = None
- if stride != 1 or self.inplanes != planes * block.expansion:
- downsample = nn.Sequential(
- nn.Conv2d(self.inplanes, planes * block.expansion,
- kernel_size=1, stride=stride, bias=False),
- nn.BatchNorm2d(planes * block.expansion),
- )
-
- layers = []
- layers.append(block(self.inplanes, planes, stride, downsample))
- self.inplanes = planes * block.expansion
- for i in range(1, blocks):
- layers.append(block(self.inplanes, planes))
-
- return nn.Sequential(*layers)
+def resnet18(pretrained=False, progress=True, **kwargs):
+ """Constructs a ResNet-18 model.
- def forward(self, x):
- x = self.conv1(x)
- x = self.bn1(x)
- x = self.relu(x)
- x = self.maxpool(x)
+ Args:
+ pretrained (bool): If True, returns a model pre-trained on ImageNet
+ progress (bool): If True, displays a progress bar of the download to stderr
+ """
+ return _resnet('resnet18', DistillerBasicBlock, [2, 2, 2, 2], pretrained, progress,
+ **kwargs)
- x = self.layer1(x)
- x = self.layer2(x)
- x = self.layer3(x)
- x = self.layer4(x)
- x = self.avgpool(x)
- x = x.view(x.size(0), -1)
- x = self.fc(x)
+def resnet34(pretrained=False, progress=True, **kwargs):
+ """Constructs a ResNet-34 model.
- return x
+ Args:
+ pretrained (bool): If True, returns a model pre-trained on ImageNet
+ progress (bool): If True, displays a progress bar of the download to stderr
+ """
+ return _resnet('resnet34', DistillerBasicBlock, [3, 4, 6, 3], pretrained, progress,
+ **kwargs)
-def resnet18(pretrained=False, **kwargs):
- """Constructs a ResNet-18 model.
+def resnet50(pretrained=False, progress=True, **kwargs):
+ """Constructs a ResNet-50 model.
Args:
pretrained (bool): If True, returns a model pre-trained on ImageNet
+ progress (bool): If True, displays a progress bar of the download to stderr
"""
- model = ResNet(BasicBlock, [2, 2, 2, 2], **kwargs)
- if pretrained:
- model.load_state_dict(model_zoo.load_url(model_urls['resnet18']))
- return model
+ return _resnet('resnet50', DistillerBottleneck, [3, 4, 6, 3], pretrained, progress,
+ **kwargs)
-def resnet34(pretrained=False, **kwargs):
- """Constructs a ResNet-34 model.
+def resnet101(pretrained=False, progress=True, **kwargs):
+ """Constructs a ResNet-101 model.
Args:
pretrained (bool): If True, returns a model pre-trained on ImageNet
+ progress (bool): If True, displays a progress bar of the download to stderr
"""
- model = ResNet(BasicBlock, [3, 4, 6, 3], **kwargs)
- if pretrained:
- model.load_state_dict(model_zoo.load_url(model_urls['resnet34']))
- return model
+ return _resnet('resnet101', DistillerBottleneck, [3, 4, 23, 3], pretrained, progress,
+ **kwargs)
-def resnet50(pretrained=False, **kwargs):
- """Constructs a ResNet-50 model.
+def resnet152(pretrained=False, progress=True, **kwargs):
+ """Constructs a ResNet-152 model.
Args:
pretrained (bool): If True, returns a model pre-trained on ImageNet
+ progress (bool): If True, displays a progress bar of the download to stderr
"""
- model = ResNet(Bottleneck, [3, 4, 6, 3], **kwargs)
- if pretrained:
- model.load_state_dict(model_zoo.load_url(model_urls['resnet50']))
- return model
+ return _resnet('resnet152', DistillerBottleneck, [3, 8, 36, 3], pretrained, progress,
+ **kwargs)
-def resnet101(pretrained=False, **kwargs):
- """Constructs a ResNet-101 model.
+def resnext50_32x4d(pretrained=False, progress=True, **kwargs):
+ """Constructs a ResNeXt-50 32x4d model.
Args:
pretrained (bool): If True, returns a model pre-trained on ImageNet
+ progress (bool): If True, displays a progress bar of the download to stderr
"""
- model = ResNet(Bottleneck, [3, 4, 23, 3], **kwargs)
- if pretrained:
- model.load_state_dict(model_zoo.load_url(model_urls['resnet101']))
- return model
+ kwargs['groups'] = 32
+ kwargs['width_per_group'] = 4
+ return _resnet('resnext50_32x4d', DistillerBottleneck, [3, 4, 6, 3],
+ pretrained, progress, **kwargs)
-def resnet152(pretrained=False, **kwargs):
- """Constructs a ResNet-152 model.
+def resnext101_32x8d(pretrained=False, progress=True, **kwargs):
+ """Constructs a ResNeXt-101 32x8d model.
Args:
pretrained (bool): If True, returns a model pre-trained on ImageNet
+ progress (bool): If True, displays a progress bar of the download to stderr
"""
- model = ResNet(Bottleneck, [3, 8, 36, 3], **kwargs)
- if pretrained:
- model.load_state_dict(model_zoo.load_url(model_urls['resnet152']))
- return model
+ kwargs['groups'] = 32
+ kwargs['width_per_group'] = 8
+ return _resnet('resnext101_32x8d', DistillerBottleneck, [3, 4, 23, 3],
+ pretrained, progress, **kwargs)