mask_vgg.py

'''
Copyright (C) 2019. Huawei Technologies Co., Ltd. All rights reserved.

This program is free software; you can redistribute it and/or modify
it under the terms of BSD 3-Clause License.

This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
BSD 3-Clause License for more details.
'''
import torch
import torch.nn as nn

import sys
from os import path
sys.path.append( path.dirname( path.dirname( path.abspath(__file__) ) ) )
#from mask_conv import MaskConv2d, MaskConv2dShare
from mask_conv import MaskConv2d as Conv

cfg = {
    'VGG11': [64, 'M', 128, 'M', 256, 256, 'M', 512, 512, 'M', 512, 512, 'M'],
    'VGG13': [64, 64, 'M', 128, 128, 'M', 256, 256, 'M', 512, 512, 'M', 512, 512, 'M'],
    'VGG16': [64, 64, 'M', 128, 128, 'M', 256, 256, 256, 'M', 512, 512, 512, 'M', 512, 512, 512, 'M'],
    'VGG19': [64, 64, 'M', 128, 128, 'M', 256, 256, 256, 256, 'M', 512, 512, 512, 512, 'M', 512, 512, 512, 512, 'M'],
}


class VGG(nn.Module):
    def __init__(self, vgg_name, r=4):
        super(VGG, self).__init__()
        self.features = self._make_layers(cfg[vgg_name], r)
        self.classifier = nn.Linear(512, 10)

    def forward(self, x, bin=False):
        for i,layer in enumerate(self.features):
            if isinstance(layer, Conv):
                x = layer(x, bin)
            else:
                x = layer(x)
        x = x.view(x.size(0), -1)
        x = self.classifier(x)
        return x

    def _make_layers(self, cfg, r):
        layers = []
        in_channels = 3
        for x in cfg:
            if x == 'M':
                layers += [nn.MaxPool2d(kernel_size=2, stride=2)]
            else:
                layers += [Conv(in_channels, x, n_basis=x/r, kernel_size=3, padding=1),
                           nn.BatchNorm2d(x),
                           nn.ReLU(inplace=True)]
                in_channels = x
        layers += [nn.AvgPool2d(kernel_size=1, stride=1)]
        return nn.ModuleList(layers)


def test():
    net = VGG('VGG16')
    x = torch.randn(2,3,32,32)
    y = net(x)
    print(y.size())