networks/backbones/comomunit.py

"""
comomunit.py
In this file all architectural components of CoMo-MUNIT are defined. The *logic* is not defined here, but in the *_model.py files.
Most of the code is copied from https://github.com/NVlabs/MUNIT
Thttps://github.com/junyanz/pytorch-CycleGAN-and-pix2pixhere are some additional function to get compatibility with the CycleGAN codebase (https://github.com/junyanz/pytorch-CycleGAN-and-pix2pix)
"""

import torch
import torch.nn as nn
from torch.nn import init
import functools
from torch.optim import lr_scheduler
import torch.nn.functional as F
from .functions import init_net, init_weights, get_scheduler


########################################################################################################################
# MUNIT architecture
########################################################################################################################

## Functions to get generator / discriminator / DRB
def define_G_munit(input_nc, output_nc, gen_dim, style_dim, n_downsample, n_res,
                   pad_type, mlp_dim, activ='relu', init_type = 'kaiming', init_gain=0.02, gpu_ids=[]):
    gen = AdaINGen(input_nc, output_nc, gen_dim, style_dim, n_downsample, n_res, activ, pad_type, mlp_dim)
    return init_net(gen, init_type=init_type, init_gain = init_gain, gpu_ids = gpu_ids)

def define_D_munit(input_nc, disc_dim, norm, activ, n_layer, gan_type, num_scales, pad_type,
                   init_type = 'kaiming', init_gain = 0.02, gpu_ids = [], output_channels = 1, final_function = None):
    disc = MsImageDis(input_nc, n_layer, gan_type, disc_dim, norm, activ, num_scales, pad_type, output_channels, final_function = final_function)
    return init_net(disc, init_type=init_type, init_gain = init_gain, gpu_ids = gpu_ids)

def define_DRB_munit(resblocks, dim, norm, activation, pad_type,
                       init_type = 'kaiming', init_gain = 0.02, gpu_ids = []):
    demux = DRB(resblocks, dim, norm, activation, pad_type)
    return init_net(demux, init_type = init_type, init_gain = init_gain, gpu_ids = gpu_ids)

# This class has been strongly modified from MUNIT default version. We split the default MUNIT decoder
# in AdaINBlock + DecoderNoAdain because the DRB must be placed between the two. encode/assign_adain/decode
# are called by the network logic following CoMo-MUNIT implementation.
class AdaINGen(nn.Module):
    # AdaIN auto-encoder architecture
    def __init__(self, input_dim, output_dim, dim, style_dim, n_downsample, n_res, activ, pad_type, mlp_dim):
        super(AdaINGen, self).__init__()

        # style encoder
        self.enc_style = StyleEncoder(4, input_dim, dim, style_dim, norm='none', activ=activ, pad_type=pad_type)

        # content encoder
        self.enc_content = ContentEncoder(n_downsample, n_res, input_dim, dim, 'instance', activ, pad_type=pad_type)
        self.adainblock = AdaINBlock(n_downsample, n_res, self.enc_content.output_dim, output_dim, res_norm='adain', activ=activ, pad_type=pad_type)
        self.dec = DecoderNoAdain(n_downsample, n_res, self.enc_content.output_dim, output_dim, res_norm='adain', activ=activ, pad_type=pad_type)
        # MLP to generate AdaIN parameters
        self.mlp = MLP(style_dim, self.get_num_adain_params(self.adainblock), mlp_dim, 3, norm='none', activ=activ)

    def forward(self, images):
        # reconstruct an image
        content, style_fake = self.encode(images)
        images_recon = self.decode(content, style_fake)
        return images_recon

    def encode(self, images):
        # encode an image to its content and style codes
        style_fake = self.enc_style(images)
        content = self.enc_content(images)
        return content, style_fake

    def assign_adain(self, content, style):
        # decode content and style codes to an image
        adain_params = self.mlp(style)
        self.assign_adain_params(adain_params, self.adainblock)
        features = self.adainblock(content)
        return features

    def decode(self, features):
        return self.dec(features)

    def assign_adain_params(self, adain_params, model):
        # assign the adain_params to the AdaIN layers in model
        for m in model.modules():
            if m.__class__.__name__ == "AdaptiveInstanceNorm2d":
                mean = adain_params[:, :m.num_features]
                std = adain_params[:, m.num_features:2*m.num_features]
                m.bias = mean.contiguous().view(-1)
                m.weight = std.contiguous().view(-1)
                if adain_params.size(1) > 2*m.num_features:
                    adain_params = adain_params[:, 2*m.num_features:]

    def get_num_adain_params(self, model):
        # return the number of AdaIN parameters needed by the model
        num_adain_params = 0
        for m in model.modules():
            if m.__class__.__name__ == "AdaptiveInstanceNorm2d":
                num_adain_params += 2*m.num_features
        return num_adain_params

# This is the FIN layer for cyclic encoding. It's our contribution and it does not exist in MUNIT.
class FIN2dCyclic(nn.Module):
    def __init__(self, dim):
        super().__init__()
        self.instance_norm = nn.InstanceNorm2d(dim, affine=False)
        self.a_gamma = nn.Parameter(torch.zeros(dim))
        self.b_gamma = nn.Parameter(torch.ones(dim))
        self.a_beta = nn.Parameter(torch.zeros(dim))
        self.b_beta = nn.Parameter(torch.zeros(dim))

    def forward(self, x, cos, sin):
        # The only way to encode something cyclic is to map gamma and beta to an ellipse point (x,y).
        # We are trying to learn their cyclic manner associating cos(continuity) to gamma and sin(continuity to beta)
        # Sin and cos are randomly sampled between -1 and 1, we know that they will be associated to one point
        gamma = self.a_gamma * cos.unsqueeze(-1) + self.b_gamma
        beta = self.a_beta * sin.unsqueeze(-1) + self.b_beta

        return self.instance_norm(x) * gamma.unsqueeze(-1).unsqueeze(-1) + beta.unsqueeze(-1).unsqueeze(-1)

# This is the DRB implementation, and it does not exist in MUNIT.
class DRB(nn.Module):
    def __init__(self, n_resblocks, dim, norm, activation, pad_type):
        super().__init__()
        self.common_features = []
        self.physical_features = []
        self.real_features = []
        self.continuous_features = nn.ModuleList()

        for i in range(0, n_resblocks):
            self.common_features += [ResBlock(dim, norm=norm, activation=activation, pad_type=pad_type)]
        for i in range(0, n_resblocks):
            self.physical_features += [ResBlock(dim, norm=norm, activation=activation, pad_type=pad_type)]
        for i in range(0, n_resblocks):
            self.real_features += [ResBlock(dim, norm=norm, activation=activation, pad_type=pad_type)]
        for i in range(0, n_resblocks):
            self.continuous_features.append(ResBlockContinuous(dim, norm='fin', activation=activation, pad_type=pad_type))

        self.common_features = nn.Sequential(*self.common_features)
        self.physical_features = nn.Sequential(*self.physical_features)
        self.real_features = nn.Sequential(*self.real_features)

    def forward(self, input_features, continuity_cos, continuity_sin):
        common_features = self.common_features(input_features)
        physical_features = self.physical_features(input_features)
        real_features = self.real_features(input_features)
        continuous_features = input_features
        for layer in self.continuous_features:
            continuous_features = layer(continuous_features, continuity_cos, continuity_sin)

        physical_output_features = common_features + physical_features + continuous_features + input_features
        real_output_features = common_features + real_features + continuous_features + input_features

        return real_output_features, physical_output_features

# Again, the default decoder is with adain, but we separated the two.
class DecoderNoAdain(nn.Module):
    def __init__(self, n_upsample, n_res, dim, output_dim, res_norm='adain', activ='relu', pad_type='zero'):
        super(DecoderNoAdain, self).__init__()

        self.model = []
        # upsampling blocks
        for i in range(n_upsample):
            self.model += [nn.Upsample(scale_factor=2),
                           Conv2dBlock(dim, dim // 2, 5, 1, 2, norm='layer', activation=activ, pad_type=pad_type)]
            dim //= 2
        # use reflection padding in the last conv layer
        self.model += [Conv2dBlock(dim, output_dim, 7, 1, 3, norm='none', activation='tanh', pad_type=pad_type)]
        self.model = nn.Sequential(*self.model)

    def forward(self, x):
        return self.model(x)

# This is a residual block with FIN layers inserted.
class ResBlockContinuous(nn.Module):
    def __init__(self, dim, norm='instance', activation='relu', pad_type='zero'):
        super(ResBlockContinuous, self).__init__()

        self.model = nn.ModuleList()
        self.model.append(Conv2dBlockContinuous(dim ,dim, 3, 1, 1, norm='fin', activation=activation, pad_type=pad_type))
        self.model.append(Conv2dBlockContinuous(dim ,dim, 3, 1, 1, norm='fin', activation='none', pad_type=pad_type))

    def forward(self, x, cos_phi, sin_phi):
        residual = x
        for layer in self.model:
            x = layer(x, cos_phi, sin_phi)

        x += residual
        return x

# This is a convolutional block+nonlinear+norm with support for FIN layers as normalization strategy.
class Conv2dBlockContinuous(nn.Module):
    def __init__(self, input_dim ,output_dim, kernel_size, stride,
                 padding=0, norm='none', activation='relu', pad_type='zero'):
        super(Conv2dBlockContinuous, self).__init__()
        self.use_bias = True
        # initialize padding
        if pad_type == 'reflect':
            self.pad = nn.ReflectionPad2d(padding)
        elif pad_type == 'replicate':
            self.pad = nn.ReplicationPad2d(padding)
        elif pad_type == 'zero':
            self.pad = nn.ZeroPad2d(padding)
        else:
            assert 0, "Unsupported padding type: {}".format(pad_type)

        # initialize normalization
        norm_dim = output_dim
        if norm == 'batch':
            self.norm = nn.BatchNorm2d(norm_dim)
        elif norm == 'instance':
            #self.norm = nn.InstanceNorm2d(norm_dim, track_running_stats=True)
            self.norm = nn.InstanceNorm2d(norm_dim)
        elif norm == 'layer':
            self.norm = LayerNorm(norm_dim)
        elif norm == 'adain':
            self.norm = AdaptiveInstanceNorm2d(norm_dim)
        elif norm == 'fin':
            self.norm = FIN2dCyclic(norm_dim)
        elif norm == 'none' or norm == 'spectral':
            self.norm = None
        else:
            assert 0, "Unsupported normalization: {}".format(norm)

        # initialize activation
        if activation == 'relu':
            self.activation = nn.ReLU(inplace=True)
        elif activation == 'lrelu':
            self.activation = nn.LeakyReLU(0.2, inplace=True)
        elif activation == 'prelu':
            self.activation = nn.PReLU()
        elif activation == 'selu':
            self.activation = nn.SELU(inplace=True)
        elif activation == 'tanh':
            self.activation = nn.Tanh()
        elif activation == 'none':
            self.activation = None
        else:
            assert 0, "Unsupported activation: {}".format(activation)

        # initialize convolution
        if norm == 'spectral':
            self.conv = SpectralNorm(nn.Conv2d(input_dim, output_dim, kernel_size, stride, bias=self.use_bias))
        else:
            self.conv = nn.Conv2d(input_dim, output_dim, kernel_size, stride, bias=self.use_bias)

    def forward(self, x, continuity_cos, continuity_sin):
        x = self.conv(self.pad(x))
        if self.norm:
            x = self.norm(x, continuity_cos, continuity_sin)
        if self.activation:
            x = self.activation(x)
        return x



##################################################################################
# All below there are MUNIT default blocks.
##################################################################################
class ResBlocks(nn.Module):
    def __init__(self, num_blocks, dim, norm='instance', activation='relu', pad_type='zero'):
        super(ResBlocks, self).__init__()
        self.model = []
        for i in range(num_blocks):
            self.model += [ResBlock(dim, norm=norm, activation=activation, pad_type=pad_type)]
        self.model = nn.Sequential(*self.model)

    def forward(self, x):
        return self.model(x)

class MLP(nn.Module):
    def __init__(self, input_dim, output_dim, dim, n_blk, norm='none', activ='relu'):

        super(MLP, self).__init__()
        self.model = []
        self.model += [LinearBlock(input_dim, dim, norm=norm, activation=activ)]
        for i in range(n_blk - 2):
            self.model += [LinearBlock(dim, dim, norm=norm, activation=activ)]
        self.model += [LinearBlock(dim, output_dim, norm='none', activation='none')] # no output activations
        self.model = nn.Sequential(*self.model)

    def forward(self, x):
        return self.model(x.view(x.size(0), -1))



class ResBlock(nn.Module):
    def __init__(self, dim, norm='instance', activation='relu', pad_type='zero'):
        super(ResBlock, self).__init__()

        model = []
        model += [Conv2dBlock(dim ,dim, 3, 1, 1, norm=norm, activation=activation, pad_type=pad_type)]
        model += [Conv2dBlock(dim ,dim, 3, 1, 1, norm=norm, activation='none', pad_type=pad_type)]
        self.model = nn.Sequential(*model)

    def forward(self, x):
        residual = x
        out = self.model(x)
        out += residual
        return out

class Conv2dBlock(nn.Module):
    def __init__(self, input_dim ,output_dim, kernel_size, stride,
                 padding=0, norm='none', activation='relu', pad_type='zero'):
        super(Conv2dBlock, self).__init__()
        self.use_bias = True
        # initialize padding
        if pad_type == 'reflect':
            self.pad = nn.ReflectionPad2d(padding)
        elif pad_type == 'replicate':
            self.pad = nn.ReplicationPad2d(padding)
        elif pad_type == 'zero':
            self.pad = nn.ZeroPad2d(padding)
        else:
            assert 0, "Unsupported padding type: {}".format(pad_type)

        # initialize normalization
        norm_dim = output_dim
        if norm == 'batch':
            self.norm = nn.BatchNorm2d(norm_dim)
        elif norm == 'instance':
            #self.norm = nn.InstanceNorm2d(norm_dim, track_running_stats=True)
            self.norm = nn.InstanceNorm2d(norm_dim)
        elif norm == 'layer':
            self.norm = LayerNorm(norm_dim)
        elif norm == 'adain':
            self.norm = AdaptiveInstanceNorm2d(norm_dim)
        elif norm == 'none' or norm == 'spectral':
            self.norm = None
        else:
            assert 0, "Unsupported normalization: {}".format(norm)

        # initialize activation
        if activation == 'relu':
            self.activation = nn.ReLU(inplace=True)
        elif activation == 'lrelu':
            self.activation = nn.LeakyReLU(0.2, inplace=True)
        elif activation == 'prelu':
            self.activation = nn.PReLU()
        elif activation == 'selu':
            self.activation = nn.SELU(inplace=True)
        elif activation == 'tanh':
            self.activation = nn.Tanh()
        elif activation == 'none':
            self.activation = None
        else:
            assert 0, "Unsupported activation: {}".format(activation)

        # initialize convolution
        if norm == 'spectral':
            self.conv = SpectralNorm(nn.Conv2d(input_dim, output_dim, kernel_size, stride, bias=self.use_bias))
        else:
            self.conv = nn.Conv2d(input_dim, output_dim, kernel_size, stride, bias=self.use_bias)

    def forward(self, x):
        x = self.conv(self.pad(x))
        if self.norm:
            x = self.norm(x)
        if self.activation:
            x = self.activation(x)
        return x


class LinearBlock(nn.Module):
    def __init__(self, input_dim, output_dim, norm='none', activation='relu'):
        super(LinearBlock, self).__init__()
        use_bias = True
        # initialize fully connected layer
        if norm == 'spectral':
            self.fc = SpectralNorm(nn.Linear(input_dim, output_dim, bias=use_bias))
        else:
            self.fc = nn.Linear(input_dim, output_dim, bias=use_bias)

        # initialize normalization
        norm_dim = output_dim
        if norm == 'batch':
            self.norm = nn.BatchNorm1d(norm_dim)
        elif norm == 'instance':
            self.norm = nn.InstanceNorm1d(norm_dim)
        elif norm == 'layer':
            self.norm = LayerNorm(norm_dim)
        elif norm == 'none' or norm == 'spectral':
            self.norm = None
        else:
            assert 0, "Unsupported normalization: {}".format(norm)

        # initialize activation
        if activation == 'relu':
            self.activation = nn.ReLU(inplace=True)
        elif activation == 'lrelu':
            self.activation = nn.LeakyReLU(0.2, inplace=True)
        elif activation == 'prelu':
            self.activation = nn.PReLU()
        elif activation == 'selu':
            self.activation = nn.SELU(inplace=True)
        elif activation == 'tanh':
            self.activation = nn.Tanh()
        elif activation == 'none':
            self.activation = None
        else:
            assert 0, "Unsupported activation: {}".format(activation)

    def forward(self, x):
        out = self.fc(x)
        if self.norm:
            out = self.norm(out)
        if self.activation:
            out = self.activation(out)
        return out


class Vgg16(nn.Module):
    def __init__(self):
        super(Vgg16, self).__init__()
        self.conv1_1 = nn.Conv2d(3, 64, kernel_size=3, stride=1, padding=1)
        self.conv1_2 = nn.Conv2d(64, 64, kernel_size=3, stride=1, padding=1)

        self.conv2_1 = nn.Conv2d(64, 128, kernel_size=3, stride=1, padding=1)
        self.conv2_2 = nn.Conv2d(128, 128, kernel_size=3, stride=1, padding=1)

        self.conv3_1 = nn.Conv2d(128, 256, kernel_size=3, stride=1, padding=1)
        self.conv3_2 = nn.Conv2d(256, 256, kernel_size=3, stride=1, padding=1)
        self.conv3_3 = nn.Conv2d(256, 256, kernel_size=3, stride=1, padding=1)

        self.conv4_1 = nn.Conv2d(256, 512, kernel_size=3, stride=1, padding=1)
        self.conv4_2 = nn.Conv2d(512, 512, kernel_size=3, stride=1, padding=1)
        self.conv4_3 = nn.Conv2d(512, 512, kernel_size=3, stride=1, padding=1)

        self.conv5_1 = nn.Conv2d(512, 512, kernel_size=3, stride=1, padding=1)
        self.conv5_2 = nn.Conv2d(512, 512, kernel_size=3, stride=1, padding=1)
        self.conv5_3 = nn.Conv2d(512, 512, kernel_size=3, stride=1, padding=1)

    def forward(self, X):
        h = F.relu(self.conv1_1(X), inplace=True)
        h = F.relu(self.conv1_2(h), inplace=True)
        # relu1_2 = h
        h = F.max_pool2d(h, kernel_size=2, stride=2)

        h = F.relu(self.conv2_1(h), inplace=True)
        h = F.relu(self.conv2_2(h), inplace=True)
        # relu2_2 = h
        h = F.max_pool2d(h, kernel_size=2, stride=2)

        h = F.relu(self.conv3_1(h), inplace=True)
        h = F.relu(self.conv3_2(h), inplace=True)
        h = F.relu(self.conv3_3(h), inplace=True)
        # relu3_3 = h
        h = F.max_pool2d(h, kernel_size=2, stride=2)

        h = F.relu(self.conv4_1(h), inplace=True)
        h = F.relu(self.conv4_2(h), inplace=True)
        h = F.relu(self.conv4_3(h), inplace=True)
        # relu4_3 = h

        h = F.relu(self.conv5_1(h), inplace=True)
        h = F.relu(self.conv5_2(h), inplace=True)
        h = F.relu(self.conv5_3(h), inplace=True)
        relu5_3 = h

        return relu5_3
        # return [relu1_2, relu2_2, relu3_3, relu4_3]


class AdaptiveInstanceNorm2d(nn.Module):
    def __init__(self, num_features, eps=1e-5, momentum=0.1):
        super(AdaptiveInstanceNorm2d, self).__init__()
        self.num_features = num_features
        self.eps = eps
        self.momentum = momentum
        # weight and bias are dynamically assigned
        self.weight = None
        self.bias = None
        # just dummy buffers, not used
        self.register_buffer('running_mean', torch.zeros(num_features))
        self.register_buffer('running_var', torch.ones(num_features))

    def forward(self, x):
        assert self.weight is not None and self.bias is not None, "Please assign weight and bias before calling AdaIN!"
        b, c = x.size(0), x.size(1)

        if self.weight.type() == 'torch.cuda.HalfTensor':
            running_mean = self.running_mean.repeat(b).to(torch.float16)
            running_var = self.running_var.repeat(b).to(torch.float16)
        else:
            running_mean = self.running_mean.repeat(b)
            running_var = self.running_var.repeat(b)

        # Apply instance norm
        x_reshaped = x.contiguous().view(1, b * c, *x.size()[2:])

        out = F.batch_norm(
            x_reshaped, running_mean, running_var, self.weight, self.bias,
            True, self.momentum, self.eps)

        return out.view(b, c, *x.size()[2:])

    def __repr__(self):
        return self.__class__.__name__ + '(' + str(self.num_features) + ')'


class LayerNorm(nn.Module):
    def __init__(self, num_features, eps=1e-5, affine=True):
        super(LayerNorm, self).__init__()
        self.num_features = num_features
        self.affine = affine
        self.eps = eps

        if self.affine:
            self.gamma = nn.Parameter(torch.Tensor(num_features).uniform_())
            self.beta = nn.Parameter(torch.zeros(num_features))

    def forward(self, x):
        shape = [-1] + [1] * (x.dim() - 1)
        # print(x.size())
        if x.size(0) == 1:
            # These two lines run much faster in pytorch 0.4 than the two lines listed below.
            mean = x.view(-1).mean().view(*shape)
            std = x.view(-1).std().view(*shape)
        else:
            mean = x.view(x.size(0), -1).mean(1).view(*shape)
            std = x.view(x.size(0), -1).std(1).view(*shape)

        x = (x - mean) / (std + self.eps)

        if self.affine:
            shape = [1, -1] + [1] * (x.dim() - 2)
            x = x * self.gamma.view(*shape) + self.beta.view(*shape)
        return x

def l2normalize(v, eps=1e-12):
    return v / (v.norm() + eps)


class SpectralNorm(nn.Module):
    """
    Based on the paper "Spectral Normalization for Generative Adversarial Networks" by Takeru Miyato, Toshiki Kataoka, Masanori Koyama, Yuichi Yoshida
    and the Pytorch implementation https://github.com/christiancosgrove/pytorch-spectral-normalization-gan
    """
    def __init__(self, module, name='weight', power_iterations=1):
        super(SpectralNorm, self).__init__()
        self.module = module
        self.name = name
        self.power_iterations = power_iterations
        if not self._made_params():
            self._make_params()

    def _update_u_v(self):
        u = getattr(self.module, self.name + "_u")
        v = getattr(self.module, self.name + "_v")
        w = getattr(self.module, self.name + "_bar")

        height = w.data.shape[0]
        for _ in range(self.power_iterations):
            v.data = l2normalize(torch.mv(torch.t(w.view(height,-1).data), u.data))
            u.data = l2normalize(torch.mv(w.view(height,-1).data, v.data))

        # sigma = torch.dot(u.data, torch.mv(w.view(height,-1).data, v.data))
        sigma = u.dot(w.view(height, -1).mv(v))
        setattr(self.module, self.name, w / sigma.expand_as(w))

    def _made_params(self):
        try:
            u = getattr(self.module, self.name + "_u")
            v = getattr(self.module, self.name + "_v")
            w = getattr(self.module, self.name + "_bar")
            return True
        except AttributeError:
            return False


    def _make_params(self):
        w = getattr(self.module, self.name)

        height = w.data.shape[0]
        width = w.view(height, -1).data.shape[1]

        u = nn.Parameter(w.data.new(height).normal_(0, 1), requires_grad=False)
        v = nn.Parameter(w.data.new(width).normal_(0, 1), requires_grad=False)
        u.data = l2normalize(u.data)
        v.data = l2normalize(v.data)
        w_bar = nn.Parameter(w.data)

        del self.module._parameters[self.name]

        self.module.register_parameter(self.name + "_u", u)
        self.module.register_parameter(self.name + "_v", v)
        self.module.register_parameter(self.name + "_bar", w_bar)


    def forward(self, *args):
        self._update_u_v()
        return self.module.forward(*args)

class MsImageDis(nn.Module):
    # Multi-scale discriminator architecture
    def __init__(self, input_dim, n_layer, gan_type, dim, norm, activ, num_scales, pad_type, output_channels = 1, final_function = None):
        super(MsImageDis, self).__init__()
        self.n_layer = n_layer
        self.gan_type = gan_type
        self.output_channels = output_channels
        self.dim = dim
        self.norm = norm
        self.activ = activ
        self.num_scales = num_scales
        self.pad_type = pad_type
        self.input_dim = input_dim
        self.downsample = nn.AvgPool2d(3, stride=2, padding=[1, 1], count_include_pad=False)
        self.cnns = nn.ModuleList()
        self.final_function = final_function
        for _ in range(self.num_scales):
            self.cnns.append(self._make_net())

    def _make_net(self):
        dim = self.dim
        cnn_x = []
        cnn_x += [Conv2dBlock(self.input_dim, dim, 4, 2, 1, norm='none', activation=self.activ, pad_type=self.pad_type)]
        for i in range(self.n_layer - 1):
            cnn_x += [Conv2dBlock(dim, dim * 2, 4, 2, 1, norm=self.norm, activation=self.activ, pad_type=self.pad_type)]
            dim *= 2
        cnn_x += [nn.Conv2d(dim, self.output_channels, 1, 1, 0)]
        cnn_x = nn.Sequential(*cnn_x)
        return cnn_x

    def forward(self, x):
        outputs = []
        for model in self.cnns:
            output = model(x)
            if self.final_function is not None:
                output = self.final_function(output)
            outputs.append(output)
            x = self.downsample(x)
        return outputs

    def calc_dis_loss(self, input_fake, input_real):
        # calculate the loss to train D
        outs0 = self.forward(input_fake)
        outs1 = self.forward(input_real)
        loss = 0

        for it, (out0, out1) in enumerate(zip(outs0, outs1)):
            if self.gan_type == 'lsgan':
                loss += torch.mean((out0 - 0)**2) + torch.mean((out1 - 1)**2)
            elif self.gan_type == 'nsgan':
                all0 = torch.zeros_like(out0)
                all1 = torch.ones_like(out1)
                loss += torch.mean(F.binary_cross_entropy(F.sigmoid(out0), all0) +
                                   F.binary_cross_entropy(F.sigmoid(out1), all1))
            else:
                assert 0, "Unsupported GAN type: {}".format(self.gan_type)
        return loss

    def calc_gen_loss(self, input_fake):
        # calculate the loss to train G
        outs0 = self.forward(input_fake)
        loss = 0
        for it, (out0) in enumerate(outs0):
            if self.gan_type == 'lsgan':
                loss += torch.mean((out0 - 1)**2) # LSGAN
            elif self.gan_type == 'nsgan':
                all1 = torch.ones_like(out0.data)
                loss += torch.mean(F.binary_cross_entropy(F.sigmoid(out0), all1))
            else:
                assert 0, "Unsupported GAN type: {}".format(self.gan_type)
        return loss

class StyleEncoder(nn.Module):
    def __init__(self, n_downsample, input_dim, dim, style_dim, norm, activ, pad_type):
        super(StyleEncoder, self).__init__()
        self.model = []
        self.model += [Conv2dBlock(input_dim, dim, 7, 1, 3, norm=norm, activation=activ, pad_type=pad_type)]
        for i in range(2):
            self.model += [Conv2dBlock(dim, 2 * dim, 4, 2, 1, norm=norm, activation=activ, pad_type=pad_type)]
            dim *= 2
        for i in range(n_downsample - 2):
            self.model += [Conv2dBlock(dim, dim, 4, 2, 1, norm=norm, activation=activ, pad_type=pad_type)]
        self.model += [nn.AdaptiveAvgPool2d(1)] # global average pooling
        self.model += [nn.Conv2d(dim, style_dim, 1, 1, 0)]
        self.model = nn.Sequential(*self.model)
        self.output_dim = dim

    def forward(self, x):
        return self.model(x)

class ContentEncoder(nn.Module):
    def __init__(self, n_downsample, n_res, input_dim, dim, norm, activ, pad_type):
        super(ContentEncoder, self).__init__()
        self.model = []
        self.model += [Conv2dBlock(input_dim, dim, 7, 1, 3, norm=norm, activation=activ, pad_type=pad_type)]
        # downsampling blocks
        for i in range(n_downsample):
            self.model += [Conv2dBlock(dim, 2 * dim, 4, 2, 1, norm=norm, activation=activ, pad_type=pad_type)]
            dim *= 2
        # residual blocks
        self.model += [ResBlocks(n_res, dim, norm=norm, activation=activ, pad_type=pad_type)]
        self.model = nn.Sequential(*self.model)
        self.output_dim = dim

    def forward(self, x):
        return self.model(x)

class AdaINBlock(nn.Module):
    def __init__(self, n_upsample, n_res, dim, output_dim, res_norm='adain', activ='relu', pad_type='zero'):
        super(AdaINBlock, self).__init__()

        self.model = []
        # AdaIN residual blocks
        self.model += [ResBlocks(n_res, dim, res_norm, activ, pad_type=pad_type)]
        self.model = nn.Sequential(*self.model)

    def forward(self, x):
        return self.model(x)