Pytorch源码学习之七：torchvision.models.googlenet
0.基本知识
torchvision.models.googlenet源码地址
GoogLeNet论文地址
- Inception模块
注意，代码中第三个模块采取的3x3卷积而非5x5卷积，不知为何。
-GoogLeNet网络结构
- GoogLeNet网络结构可视化
这里注意两个辅助的输出。
一、源码
import torch
import torch.nn as nn
import torch.nn.functional as F
import warnings
from collections import namedtuple
from torch.hub import load_state_dict_from_url
__all__ = ['GoogLeNet', 'googlenet']
model_urls = {
# GoogLeNet ported from TensorFlow
'googlenet': 'https://download.pytorch.org/models/googlenet-1378be20.pth',
}
_GoogLeNetOuputs = namedtuple('GoogLeNetOuputs', ['logits', 'aux_logits2', 'aux_logits1'])
class BasicBlock(nn.Module):
def __init(self, in_channels, out_channels, **kwargs):
super(BasicBlock, self).__init__()
self.conv = nn.Conv2d(in_channels, out_channels, bias=False, **kwargs)
self.bn = nn.BatchNorm2d(out_channels, eps=0.001)
def forward(self, x):
x = self.conv(x)
x = self.bn(x)
return F.relu(x, inplace=True)
class Inception(nn.Module):
def __init__(self, in_channels, ch1x1, ch3x3red, ch3x3, ch5x5red, ch5x5, pool_proj):
super(Inception, self).__init__()
self.branch1 = BasicBlock(in_channels, ch1x1, kernel_size=1)
self.branch2 = nn.Sequential(
BasicBlock(in_channels, ch3x3red, kernel_size=1),
BasicBlock(ch3x3red, ch3x3, kernel_size=3, padding=1)
)
self.branch3 = nn.Sequential(
BasicBlock(in_channels, ch5x5red, kernel_size=1),
BasicBlock(ch5x5red, ch5x5, kernel_size=3, padding=1)
)
self.branch4 = nn.Sequential(
nn.MaxPool2d(kernel_size=3, stride=1, padding=1, ceil_mode=True),
BasicBlock(in_channels, pool_proj, kernel_size=1)
)
def forward(self, x):
branch1 = self.branch1(x)
branch2 = self.branch2(x)
branch3 = self.branch3(x)
branch4 = self.branch4(x)
outputs = [branch1, branch2, branch3, branch4]
return torch.cat(outputs, 1)
class InceptionAux(nn.Module):
def __init__(self, in_channels, num_classes):
super(InceptionAux, self).__init__()
self.conv = BasicBlock(in_channels, 128, kernel_size=1)
self.fc1 = nn.Linear(2048, 1024)
self.fc2 = nn.Linear(1024, num_classes)
def forward(self, x):
# aux1: N x 512 x 14 x 14, aux2: N x 528 x 14 x 14
x = F.adaptive_avg_pool2d(x, (4, 4))
# aux1 : N x 512 x 4 x 4, aux2 : N X 528 x 4 x 4
x = self.conv(x)
# N x 128 x 4 x 4
x = x.view(x.size(0), -1)
# N x 2048
x = self.fc1(x)
# N x 1024
x = F.relu(x, inplace=True)
x = F.dropout(x, 0.7, training=self.training)
# N x 1024
x = self.fc2(x)
# N x num_classes
return x
class GoogLeNet(nn.Module):
def __init__(self, num_classes=1000, aux_logits=True, transform_input=False, init_weights=True):
super(GoogLeNet, self).__init__()
self.aux_logits = aux_logits
self.transform_input = transform_input
self.conv1 = BasicBlock(3, 64, kernel_size=7, stride=2, padding=3)
self.maxpool1 = nn.MaxPool2d(3, stride=2, ceil_mode=True)
self.conv2 = BasicBlock(64, 64, kernel_size=1)
self.conv3 = BasicBlock(64, 192, kernel_size=3, padding=1)
self.maxpool2 = nn.MaxPool2d(3, stride=2, ceil_mode=True)
self.inception3a = Inception(192, 64, 96, 128, 16, 32, 32) #64+128+32+32=256
self.inception3b = Inception(256, 128, 128, 192, 32, 96, 64) #128+192+96+64=480
self.maxpool3 = nn.MaxPool2d(3, stride=2, ceil_mode=True)
self.inception4a = Inception(480, 192, 96, 208, 16, 48, 64) # 192+208+48+64=512
self.inception4b = Inception(512, 160, 112, 224, 24, 64, 64)#160+224+64+64=512
self.inception4c = Inception(512, 128, 128, 256, 24, 64, 64)#128+256_64+64=512
self.inception4d = Inception(512, 112, 144, 288, 32, 64, 64)#112+288+64+64=528
self.inception4e = Inception(528, 256, 160, 320, 32, 128, 128)#256+320+128+128=832
self.maxpool4 = nn.MaxPool2d(2, stride=2, ceil_mode=True)
self.inception5a = Inception(832, 256, 160, 320, 32, 128, 128)#256+320+128+128=832
self.inception5b = Inception(832, 384, 192, 384, 48, 128, 128)#384+384+128+128=1024
if aux_logits:
self.aux1 = InceptionAux(512, num_classes)
self.aux2 = InceptionAux(528, num_classes)
self.avgpool = nn.AdaptiveAvgPool2d((1, 1))
self.dropout = nn.Dropout(0.2)
self.fc = nn.Linear(1024, num_classes)
if init_weights:
self._initialize_weights()
def _initialize_weights(self):
for m in self.modules():
if isinstance(m, nn.Conv2d) or isinstance(m, nn.Linear):
import scipy.stats as stats
X = stats.truncnorm(-2, 2, scale=0.01)
values = torch.as_tensor(X.rvs(m.weight.numel()), dtype=m.weight.dtype)
values = values.view(m.weight.size())
with torch.no_grad():
m.weight.copy_(values)
elif isinstance(m, nn.BatchNorm2d):
nn.init.constant_(m.weight, 1)
nn.init.constant_(m.bias, 0)
def forward(self, x):
if self.transform_input:
x_ch0 = torch.unsqueeze(x[:, 0], 1) * (0.229 / 0.5) + (0.485 - 0.5) / 0.5
x_ch1 = torch.unsqueeze(x[:, 1], 1) * (0.224 / 0.5) + (0.456 - 0.5) / 0.5
x_ch2 = torch.unsqueeze(x[:, 2], 1) * (0.225 / 0.5) + (0.406 - 0.5) / 0.5
x = torch.cat((x_ch0, x_ch1, x_ch2), 1)
# N x 3 x 224 x 224
x = self.conv1(x)
# N x 64 x 112 x 112
x = self.maxpool1(x)
# N x 64 x 56 x 56
x = self.conv2(x)
# N x 64 x 56 x 56
x = self.conv3(x)
# N x 192 x 56 x 56
x = self.maxpool2(x)
# N x 192 x 28 x 28
x = self.inception3a(x)
# N x 256 x 28 x 28
x = self.inception3b(x)
# N x 480 x 28 x 28
x = self.maxpool3(x)
# N x 480 x 14 x 14
x = self.inception4a(x)
# N x 512 x 14 x 14
if self.training and self.aux_logits:
aux1 = self.aux1(x)
x = self.inception4b(x)
# N x 512 x 14 x 14
x = self.inception4c(x)
# N x 512 x 14 x 14
x = self.inception4d(x)
# N x 528 x 14 x 14
if self.training and self.aux_logits:
aux2 = self.aux2(x)
x = self.inception4e(x)
# N x 832 x 14 x 14
x = self.maxpool4(x)
# N x 832 x 7 x 7
x = self.inception5a(x)
# N x 832 x 7 x 7
x = self.inception5b(x)
# N x 1024 x 7 x 7
x = self.avgpool(x)
# N x 1024 x 1 x 1
x = x.view(x.size(0), -1)
# N x 1024
x = self.dropout(x)
x = self.fc(x)
# N x 1000(num_classes)
if self.training and self.aux_logits:
return _GoogLeNetOuputs(x, aux2, aux1)
return x
def googlenet(pretrained=False, progress=True, **kwargs):
if pretrained:
if 'transform_input' not in kwargs:
kwargs['transform_input'] = True
if 'aux_logits' not in kwargs:
kwargs['aux_logits'] = False
if kwargs['aux_logits']:
warnings.warn('auxiliary heads in the pretrained googlenet model are NOT pretrained, '
'so make sure to train them')
original_aux_logits = kwargs['aux_logits']
kwargs['aux_logits'] = True
kwargs['init_weights'] = False
model = GoogLeNet(**kwargs)
state_dict = load_state_dict_from_url(model_urls['googlenet'],
progress=progress)
model.load_state_dict(state_dict)
if not original_aux_logits:
model.aux_logits = False
del model.aux1, model.aux2
return model
return GoogLeNet(**kwargs)
二、一些有趣的用法
1.collections.namedtuple
使用以下方式来返回辅助输出和最终输出，方便索引.
from collections import namedtuple
_GoogLeNetOuputs = namedtuple('GoogLeNetOuputs', ['logits', 'aux_logits2', 'aux_logits1'])
class GoogLeNet(nn.Module):
...
def forward(self, x):
...
if self.training and self.aux_logits:
return _GoogLeNetOuputs(x, aux2, aux1)
warnings来返回自定义警告
import warnings
if pretrained:
if kwargs['aux_logits']:
warning.warn('your warnings')