test/onnx/model_defs/squeezenet.py

import torch
import torch.nn as nn
import torch.nn.init as init


class Fire(nn.Module):
    def __init__(self, inplanes, squeeze_planes, expand1x1_planes, expand3x3_planes):
        super().__init__()
        self.inplanes = inplanes
        self.squeeze = nn.Conv2d(inplanes, squeeze_planes, kernel_size=1)
        self.squeeze_activation = nn.ReLU(inplace=True)
        self.expand1x1 = nn.Conv2d(squeeze_planes, expand1x1_planes, kernel_size=1)
        self.expand1x1_activation = nn.ReLU(inplace=True)
        self.expand3x3 = nn.Conv2d(
            squeeze_planes, expand3x3_planes, kernel_size=3, padding=1
        )
        self.expand3x3_activation = nn.ReLU(inplace=True)

    def forward(self, x):
        x = self.squeeze_activation(self.squeeze(x))
        return torch.cat(
            [
                self.expand1x1_activation(self.expand1x1(x)),
                self.expand3x3_activation(self.expand3x3(x)),
            ],
            1,
        )


class SqueezeNet(nn.Module):
    def __init__(self, version=1.0, num_classes=1000, ceil_mode=False):
        super().__init__()
        if version not in [1.0, 1.1]:
            raise ValueError(
                f"Unsupported SqueezeNet version {version}:1.0 or 1.1 expected"
            )
        self.num_classes = num_classes
        if version == 1.0:
            self.features = nn.Sequential(
                nn.Conv2d(3, 96, kernel_size=7, stride=2),
                nn.ReLU(inplace=True),
                nn.MaxPool2d(kernel_size=3, stride=2, ceil_mode=ceil_mode),
                Fire(96, 16, 64, 64),
                Fire(128, 16, 64, 64),
                Fire(128, 32, 128, 128),
                nn.MaxPool2d(kernel_size=3, stride=2, ceil_mode=ceil_mode),
                Fire(256, 32, 128, 128),
                Fire(256, 48, 192, 192),
                Fire(384, 48, 192, 192),
                Fire(384, 64, 256, 256),
                nn.MaxPool2d(kernel_size=3, stride=2, ceil_mode=ceil_mode),
                Fire(512, 64, 256, 256),
            )
        else:
            self.features = nn.Sequential(
                nn.Conv2d(3, 64, kernel_size=3, stride=2),
                nn.ReLU(inplace=True),
                nn.MaxPool2d(kernel_size=3, stride=2, ceil_mode=ceil_mode),
                Fire(64, 16, 64, 64),
                Fire(128, 16, 64, 64),
                nn.MaxPool2d(kernel_size=3, stride=2, ceil_mode=ceil_mode),
                Fire(128, 32, 128, 128),
                Fire(256, 32, 128, 128),
                nn.MaxPool2d(kernel_size=3, stride=2, ceil_mode=ceil_mode),
                Fire(256, 48, 192, 192),
                Fire(384, 48, 192, 192),
                Fire(384, 64, 256, 256),
                Fire(512, 64, 256, 256),
            )
        # Final convolution is initialized differently from the rest
        final_conv = nn.Conv2d(512, self.num_classes, kernel_size=1)
        self.classifier = nn.Sequential(
            nn.Dropout(p=0.5), final_conv, nn.ReLU(inplace=True), nn.AvgPool2d(13)
        )

        for m in self.modules():
            if isinstance(m, nn.Conv2d):
                if m is final_conv:
                    init.normal_(m.weight.data, mean=0.0, std=0.01)
                else:
                    init.kaiming_uniform_(m.weight.data)
                if m.bias is not None:
                    m.bias.data.zero_()

    def forward(self, x):
        x = self.features(x)
        x = self.classifier(x)
        return x.view(x.size(0), self.num_classes)