models/utils/official_stylegan3_model_helper.py

# Copyright (c) 2021, NVIDIA CORPORATION & AFFILIATES.  All rights reserved.
#
# NVIDIA CORPORATION and its licensors retain all intellectual property
# and proprietary rights in and to this software, related documentation
# and any modifications thereto.  Any use, reproduction, disclosure or
# distribution of this software and related documentation without an express
# license agreement from NVIDIA CORPORATION is strictly prohibited.

"""Generator architecture from the paper
"Alias-Free Generative Adversarial Networks"."""

import numpy as np
import scipy.signal
import scipy.optimize
import torch
import torch.nn as nn
import torch.nn.functional as F
import random
from utils import eg3d_misc as misc
from third_party.stylegan3_official_ops import conv2d_gradfix
from third_party.stylegan3_official_ops import filtered_lrelu
from third_party.stylegan3_official_ops import bias_act
from third_party.stylegan3_official_ops import upfirdn2d

#----------------------------------------------------------------------------
class SEL(torch.nn.Module):
    def __init__(self, norm_nc, label_nc, hidden_nc=128):
        super().__init__()
        self.norm = nn.InstanceNorm2d(norm_nc, affine=False)
        self.mlp_shared = nn.Conv2d(label_nc, hidden_nc, kernel_size=1, padding=0)
        self.actv = nn.ReLU()
        self.mlp_gamma = nn.Conv2d(hidden_nc, norm_nc, kernel_size=1, padding=0)
        self.mlp_beta = nn.Conv2d(hidden_nc, norm_nc, kernel_size=1, padding=0)

    def forward(self, x, hm):
        x_s = x
        x = self.norm(x)
        hm = F.interpolate(hm, size=x.size()[2:], mode='bilinear', align_corners=True)
        actv = self.actv(self.mlp_shared(hm))
        gamma = self.mlp_gamma(actv)
        beta = self.mlp_beta(actv)
        out = x * (1+gamma) + beta

        return out + 0.1 * x_s

class SEL_unet_pro(SEL):
    def __init__(self, norm_nc,  label_nc, hidden_nc=128, down_filter=None, slope=0.2, gain=np.sqrt(2), clamp=None):
        super().__init__(norm_nc, label_nc, hidden_nc)
        self.register_buffer('down_filter', down_filter)
        self.slope = slope
        self.gain = gain
        self.clamp = clamp

    def forward(self, x, hm):
        x_size = x.shape[-1]
        hm_size = hm.shape[-1]
        if x_size != hm_size:
            hm = upfirdn2d.upfirdn2d(x=hm, f=self.down_filter, down=hm_size//x_size, flip_filter=False, padding=int(2.5 * hm_size//x_size))
        hm = self.mlp_shared(hm)
        hm = bias_act.bias_act(x=hm, act='lrelu', alpha=self.slope, gain=self.gain, clamp=self.clamp)
        gamma = self.mlp_gamma(hm)
        beta = self.mlp_beta(hm)

        out = self.norm(x) * (1+gamma) + beta
        return out + 0.1 * x

class SEL_pro(SEL):
    def __init__(self, norm_nc, label_nc, hidden_nc=128, down_filter=None, slope=0.2, gain=np.sqrt(2), clamp=None):
        super().__init__(norm_nc, label_nc, hidden_nc)
        self.zero_pad = nn.ZeroPad2d((576-256)//2)
        self.register_buffer('down_filter', down_filter)
        self.slope = slope
        self.gain = gain
        self.clamp = clamp
        self.size_dict = {36: 36, 52: 72, 84: 144, 148: 288, 276: 576}

    def forward(self, x, hm):
        x_size = x.shape[-1]
        x_large_size = self.size_dict[x_size]
        hm = self.zero_pad(hm)
        hm_size = hm.shape[-1]
        assert hm_size % x_large_size == 0, f'hm shape {hm.shape[-1]}, x shape {x.shape[-1]}, {x_large_size}'
        if hm_size != x_large_size:
            hm = upfirdn2d.upfirdn2d(x=hm, f=self.down_filter, down=hm_size//x_large_size, flip_filter=False, 
                                     padding=int(2.5 * hm_size//x_large_size))
        hm = self.mlp_shared(hm)
        hm = bias_act.bias_act(x=hm, act='lrelu', alpha=self.slope, gain=self.gain, clamp=self.clamp)
        pad_len = (x_large_size - x_size) // 2
        if pad_len > 0:
            hm = hm[..., pad_len:-pad_len, pad_len:-pad_len]
        gamma = self.mlp_gamma(hm)
        beta = self.mlp_beta(hm)    

        out = self.norm(x) * (1+gamma) + beta
        return out + 0.1 * x

@misc.profiled_function
def modulated_conv2d(
    x,                  # Input tensor: [batch_size, in_channels, in_height, in_width]
    w,                  # Weight tensor: [out_channels, in_channels, kernel_height, kernel_width]
    s,                  # Style tensor: [batch_size, in_channels]
    demodulate  = True, # Apply weight demodulation?
    padding     = 0,    # Padding: int or [padH, padW]
    input_gain  = None, # Optional scale factors for the input channels: [], [in_channels], or [batch_size, in_channels]
):
    with misc.suppress_tracer_warnings(): # this value will be treated as a constant
        batch_size = int(x.shape[0])
    out_channels, in_channels, kh, kw = w.shape
    misc.assert_shape(w, [out_channels, in_channels, kh, kw]) # [OIkk]
    misc.assert_shape(x, [batch_size, in_channels, None, None]) # [NIHW]
    misc.assert_shape(s, [batch_size, in_channels]) # [NI]

    # Pre-normalize inputs.
    if demodulate:
        w = w * w.square().mean([1,2,3], keepdim=True).rsqrt()
        s = s * s.square().mean().rsqrt()

    # Modulate weights.
    w = w.unsqueeze(0) # [NOIkk]
    w = w * s.unsqueeze(1).unsqueeze(3).unsqueeze(4) # [NOIkk]

    # Demodulate weights.
    if demodulate:
        dcoefs = (w.square().sum(dim=[2,3,4]) + 1e-8).rsqrt() # [NO]
        w = w * dcoefs.unsqueeze(2).unsqueeze(3).unsqueeze(4) # [NOIkk]

    # Apply input scaling.
    if input_gain is not None:
        input_gain = input_gain.expand(batch_size, in_channels) # [NI]
        w = w * input_gain.unsqueeze(1).unsqueeze(3).unsqueeze(4) # [NOIkk]

    # Execute as one fused op using grouped convolution.
    x = x.reshape(1, -1, *x.shape[2:])
    w = w.reshape(-1, in_channels, kh, kw)
    x = conv2d_gradfix.conv2d(input=x, weight=w.to(x.dtype), padding=padding, groups=batch_size)
    x = x.reshape(batch_size, -1, *x.shape[2:])
    return x

#----------------------------------------------------------------------------

class FullyConnectedLayer(torch.nn.Module):
    def __init__(self,
        in_features,                # Number of input features.
        out_features,               # Number of output features.
        activation      = 'linear', # Activation function: 'relu', 'lrelu', etc.
        bias            = True,     # Apply additive bias before the activation function?
        lr_multiplier   = 1,        # Learning rate multiplier.
        weight_init     = 1,        # Initial standard deviation of the weight tensor.
        bias_init       = 0,        # Initial value of the additive bias.
        low_rank        = None,
    ):
        super().__init__()
        self.in_features = in_features
        self.out_features = out_features
        self.activation = activation
        self.low_rank = low_rank
        self.register_buffer('lr_multiplier', torch.tensor(lr_multiplier, dtype=torch.float))
        if self.low_rank is None:
            self.weight = torch.nn.Parameter(torch.randn([out_features, in_features]) * (weight_init / lr_multiplier))
        else:
            self.weight_left = torch.nn.Parameter(torch.randn([out_features, self.low_rank]) * (weight_init / lr_multiplier))
            self.weight_right = torch.nn.Parameter(torch.randn([self.low_rank, in_features]) * (weight_init / lr_multiplier))
        bias_init = np.broadcast_to(np.asarray(bias_init, dtype=np.float32), [out_features])
        self.bias = torch.nn.Parameter(torch.from_numpy(bias_init / lr_multiplier)) if bias else None
        self.weight_gain = 1 / np.sqrt(in_features)
        self.bias_gain = 1

    def forward(self, x):
        if self.low_rank is None:
            w = self.weight.to(x.dtype) * self.weight_gain * self.lr_multiplier
        else:
            w = torch.einsum('ab,bc->ac', self.weight_left, self.weight_right).to(x.dtype) * self.weight_gain * self.lr_multiplier
        b = self.bias
        if b is not None:
            b = b.to(x.dtype)
            if self.bias_gain * self.lr_multiplier != 1:
                b = b * self.bias_gain * self.lr_multiplier
        if self.activation == 'linear' and b is not None:
            x = torch.addmm(b.unsqueeze(0), x, w.t())
        else:
            x = x.matmul(w.t())
            x = bias_act.bias_act(x, b, act=self.activation)
        return x

    def extra_repr(self):
        return f'in_features={self.in_features:d}, out_features={self.out_features:d}, activation={self.activation:s}'

#----------------------------------------------------------------------------

class MappingNetwork(torch.nn.Module):
    def __init__(self,
        z_dim,                      # Input latent (Z) dimensionality.
        c_dim,                      # Conditioning label (C) dimensionality, 0 = no labels.
        w_dim,                      # Intermediate latent (W) dimensionality.
        num_ws,                     # Number of intermediate latents to output.
        num_layers      = 2,        # Number of mapping layers.
        lr_multiplier   = 0.01,     # Learning rate multiplier for the mapping layers.
        w_avg_beta      = 0.998,    # Decay for tracking the moving average of W during training.
    ):
        super().__init__()
        self.z_dim = z_dim
        self.c_dim = c_dim
        self.w_dim = w_dim
        self.num_ws = num_ws
        self.num_layers = num_layers
        self.w_avg_beta = w_avg_beta

        # Construct layers.
        self.embed = FullyConnectedLayer(self.c_dim, self.w_dim) if self.c_dim > 0 else None
        features = [self.z_dim + (self.w_dim if self.c_dim > 0 else 0)] + [self.w_dim] * self.num_layers
        for idx, in_features, out_features in zip(range(num_layers), features[:-1], features[1:]):
            layer = FullyConnectedLayer(in_features, out_features, activation='lrelu', lr_multiplier=lr_multiplier)
            setattr(self, f'fc{idx}', layer)
        self.register_buffer('w_avg', torch.zeros([w_dim]))

    def forward(self, z, c, truncation_psi=1, truncation_cutoff=None, update_emas=False):
        misc.assert_shape(z, [None, self.z_dim])
        if truncation_cutoff is None:
            truncation_cutoff = self.num_ws

        # Embed, normalize, and concatenate inputs.
        x = z.to(torch.float32)
        x = x * (x.square().mean(1, keepdim=True) + 1e-8).rsqrt()
        if self.c_dim > 0:
            misc.assert_shape(c, [None, self.c_dim])
            y = self.embed(c.to(torch.float32))
            y = y * (y.square().mean(1, keepdim=True) + 1e-8).rsqrt()
            x = torch.cat([x, y], dim=1) if x is not None else y

        # Execute layers.
        for idx in range(self.num_layers):
            x = getattr(self, f'fc{idx}')(x)

        # Update moving average of W.
        if update_emas:
            self.w_avg.copy_(x.detach().mean(dim=0).lerp(self.w_avg, self.w_avg_beta))

        # Broadcast and apply truncation.
        x = x.unsqueeze(1).repeat([1, self.num_ws, 1])
        if truncation_psi != 1:
            x[:, :truncation_cutoff] = self.w_avg.lerp(x[:, :truncation_cutoff], truncation_psi)
        return x

    def extra_repr(self):
        return f'z_dim={self.z_dim:d}, c_dim={self.c_dim:d}, w_dim={self.w_dim:d}, num_ws={self.num_ws:d}'

#----------------------------------------------------------------------------

class SynthesisInput(torch.nn.Module):
    def __init__(self,
        w_dim,          # Intermediate latent (W) dimensionality.
        channels,       # Number of output channels.
        size,           # Output spatial size: int or [width, height].
        sampling_rate,  # Output sampling rate.
        bandwidth,      # Output bandwidth.
        offset_scale='0,0',
        bound_len=0.5,
        wo_transform=False,
    ):
        super().__init__()
        self.w_dim = w_dim
        self.channels = channels
        self.size = np.broadcast_to(np.asarray(size), [2])
        self.sampling_rate = sampling_rate
        self.bandwidth = bandwidth
        self.wo_transform = wo_transform

        self.x_offset_scale, self.y_offset_scale = list(map(float, offset_scale.split(',')))     
        self.bound_len = bound_len

        # Draw random frequencies from uniform 2D disc.
        freqs = torch.randn([self.channels, 2])
        radii = freqs.square().sum(dim=1, keepdim=True).sqrt()
        freqs /= radii * radii.square().exp().pow(0.25)
        freqs *= bandwidth
        phases = torch.rand([self.channels]) - 0.5

        # Setup parameters and buffers.
        self.weight = torch.nn.Parameter(torch.randn([self.channels, self.channels]))
        if not wo_transform:
            self.affine = FullyConnectedLayer(w_dim, 4, weight_init=0, bias_init=[1,0,0,0])
        self.register_buffer('transform', torch.eye(3, 3)) # User-specified inverse transform wrt. resulting image.
        self.register_buffer('freqs', freqs)
        self.register_buffer('phases', phases)

        self.x_offset = None
        self.y_offset = None

    def forward(self, w):
        # Introduce batch dimension.
        transforms = self.transform.unsqueeze(0) # [batch, row, col]
        freqs = self.freqs.unsqueeze(0) # [batch, channel, xy]
        phases = self.phases.unsqueeze(0) # [batch, channel]

        # Apply learned transformation.
        if not self.wo_transform:
            t = self.affine(w) # t = (r_c, r_s, t_x, t_y)
            t = t / t[:, :2].norm(dim=1, keepdim=True) # t' = (r'_c, r'_s, t'_x, t'_y)
            m_r = torch.eye(3, device=w.device).unsqueeze(0).repeat([w.shape[0], 1, 1]) # Inverse rotation wrt. resulting image.
            m_r[:, 0, 0] = t[:, 0]  # r'_c
            m_r[:, 0, 1] = -t[:, 1] # r'_s
            m_r[:, 1, 0] = t[:, 1]  # r'_s
            m_r[:, 1, 1] = t[:, 0]  # r'_c
            m_t = torch.eye(3, device=w.device).unsqueeze(0).repeat([w.shape[0], 1, 1]) # Inverse translation wrt. resulting image.
            m_t[:, 0, 2] = -t[:, 2] # t'_x
            m_t[:, 1, 2] = -t[:, 3] # t'_y
            transforms = m_r @ m_t @ transforms # First rotate resulting image, then translate, and finally apply user-specified transform.

        # Transform frequencies.
        phases = phases + (freqs @ transforms[:, :2, 2:]).squeeze(2)
        freqs = freqs @ transforms[:, :2, :2]

        # Dampen out-of-band frequencies that may occur due to the user-specified transform.
        amplitudes = (1 - (freqs.norm(dim=2) - self.bandwidth) / (self.sampling_rate / 2 - self.bandwidth)).clamp(0, 1)

        # Construct sampling grid.
        theta = torch.eye(2, 3, device=w.device)
        theta[0, 0] = self.bound_len * self.size[0] / self.sampling_rate
        theta[1, 1] = self.bound_len * self.size[1] / self.sampling_rate
        grids = torch.nn.functional.affine_grid(theta.unsqueeze(0), [1, 1, self.size[1], self.size[0]], align_corners=False)
        offset_len = 1 - 2*self.bound_len       
        dx = (random.random() * offset_len - offset_len/2) if self.x_offset == None else self.x_offset  
        dy = (random.random() * offset_len - offset_len/2) if self.y_offset == None else self.y_offset 
        dx *= self.x_offset_scale 
        dy *= self.y_offset_scale
        grids[..., 0] += dx     
        grids[..., 1] += dy     

        # Compute Fourier features.
        x = (grids.unsqueeze(3) @ freqs.permute(0, 2, 1).unsqueeze(1).unsqueeze(2)).squeeze(3) # [batch, height, width, channel]
        x = x + phases.unsqueeze(1).unsqueeze(2)
        x = torch.sin(x * (np.pi * 2))
        x = x * amplitudes.unsqueeze(1).unsqueeze(2)

        # Apply trainable mapping.
        weight = self.weight / np.sqrt(self.channels)
        x = x @ weight.t()

        # Ensure correct shape.
        x = x.permute(0, 3, 1, 2) # [batch, channel, height, width]
        if self.wo_transform:
            x = x.repeat(w.shape[0], 1, 1, 1)
        misc.assert_shape(x, [w.shape[0], self.channels, int(self.size[1]), int(self.size[0])])
        return x

    def extra_repr(self):
        return '\n'.join([
            f'w_dim={self.w_dim:d}, channels={self.channels:d}, size={list(self.size)},',
            f'sampling_rate={self.sampling_rate:g}, bandwidth={self.bandwidth:g}'])

#----------------------------------------------------------------------------

class SynthesisLayer(torch.nn.Module):
    def __init__(self,
        w_dim,                          # Intermediate latent (W) dimensionality.
        is_torgb,                       # Is this the final ToRGB layer?
        is_critically_sampled,          # Does this layer use critical sampling?
        use_fp16,                       # Does this layer use FP16?

        # Input & output specifications.
        in_channels,                    # Number of input channels.
        out_channels,                   # Number of output channels.
        in_size,                        # Input spatial size: int or [width, height].
        out_size,                       # Output spatial size: int or [width, height].
        in_sampling_rate,               # Input sampling rate (s).
        out_sampling_rate,              # Output sampling rate (s).
        in_cutoff,                      # Input cutoff frequency (f_c).
        out_cutoff,                     # Output cutoff frequency (f_c).
        in_half_width,                  # Input transition band half-width (f_h).
        out_half_width,                 # Output Transition band half-width (f_h).

        # Hyperparameters.
        conv_kernel         = 3,        # Convolution kernel size. Ignored for final the ToRGB layer.
        filter_size         = 6,        # Low-pass filter size relative to the lower resolution when up/downsampling.
        lrelu_upsampling    = 2,        # Relative sampling rate for leaky ReLU. Ignored for final the ToRGB layer.
        use_radial_filters  = False,    # Use radially symmetric downsampling filter? Ignored for critically sampled layers.
        conv_clamp          = 256,      # Clamp the output to [-X, +X], None = disable clamping.
        magnitude_ema_beta  = 0.999,    # Decay rate for the moving average of input magnitudes.
        label_nc            = 0,
        use_sel             = False,
        low_rank            = None,
        sel_type            = 'normal',
        **useless_stuff,                # this part is dirty to be compatible with the same configuration system with stylegan2
    ):
        super().__init__()
        print('these configuration terms is not used:', useless_stuff)
        self.w_dim = w_dim
        self.is_torgb = is_torgb
        self.is_critically_sampled = is_critically_sampled
        self.use_fp16 = use_fp16
        self.in_channels = in_channels
        self.out_channels = out_channels
        self.in_size = np.broadcast_to(np.asarray(in_size), [2])
        self.out_size = np.broadcast_to(np.asarray(out_size), [2])
        self.in_sampling_rate = in_sampling_rate
        self.out_sampling_rate = out_sampling_rate
        self.tmp_sampling_rate = max(in_sampling_rate, out_sampling_rate) * (1 if is_torgb else lrelu_upsampling)
        self.in_cutoff = in_cutoff
        self.out_cutoff = out_cutoff
        self.in_half_width = in_half_width
        self.out_half_width = out_half_width
        self.conv_kernel = 1 if is_torgb else conv_kernel
        self.conv_clamp = conv_clamp
        self.magnitude_ema_beta = magnitude_ema_beta
        self.use_sel = use_sel and not is_torgb
        self.sel_type = sel_type

        size_dict = {36: 36, 52: 72, 84: 144, 148: 288, 276: 576}
        self.down_radial = use_radial_filters and not self.is_critically_sampled
        if self.use_sel:
            if self.sel_type == 'normal':
                self.sel = SEL(norm_nc=in_channels, label_nc=label_nc)
            elif self.sel_type == 'pro':
                sel_down_factor = 576 // size_dict[self.in_size[0]]
                sel_pro_down_filter = self.design_lowpass_filter(
                    numtaps=filter_size * sel_down_factor, cutoff=self.in_cutoff, width=self.in_half_width*2, fs=self.tmp_sampling_rate, radial=self.down_radial)
                self.sel = SEL_pro(norm_nc=in_channels, label_nc=label_nc, hidden_nc=128, down_filter=sel_pro_down_filter, 
                    slope=1 if is_torgb else 0.2, gain=1 if is_torgb else np.sqrt(2), clamp=256 if is_torgb else None)

        # Setup parameters and buffers.
        self.affine = FullyConnectedLayer(self.w_dim, self.in_channels, bias_init=1, low_rank=low_rank)
        self.weight = torch.nn.Parameter(torch.randn([self.out_channels, self.in_channels, self.conv_kernel, self.conv_kernel]))
        self.bias = torch.nn.Parameter(torch.zeros([self.out_channels]))
        self.register_buffer('magnitude_ema', torch.ones([]))

        # Design upsampling filter.
        self.up_factor = int(np.rint(self.tmp_sampling_rate / self.in_sampling_rate))
        assert self.in_sampling_rate * self.up_factor == self.tmp_sampling_rate
        self.up_taps = filter_size * self.up_factor if self.up_factor > 1 and not self.is_torgb else 1
        self.register_buffer('up_filter', self.design_lowpass_filter(
            numtaps=self.up_taps, cutoff=self.in_cutoff, width=self.in_half_width*2, fs=self.tmp_sampling_rate))

        # Design downsampling filter.
        self.down_factor = int(np.rint(self.tmp_sampling_rate / self.out_sampling_rate))
        assert self.out_sampling_rate * self.down_factor == self.tmp_sampling_rate
        self.down_taps = filter_size * self.down_factor if self.down_factor > 1 and not self.is_torgb else 1
        self.register_buffer('down_filter', self.design_lowpass_filter(
            numtaps=self.down_taps, cutoff=self.out_cutoff, width=self.out_half_width*2, fs=self.tmp_sampling_rate, radial=self.down_radial))

        # Compute padding.
        pad_total = (self.out_size - 1) * self.down_factor + 1 # Desired output size before downsampling.
        pad_total -= (self.in_size + self.conv_kernel - 1) * self.up_factor # Input size after upsampling.
        pad_total += self.up_taps + self.down_taps - 2 # Size reduction caused by the filters.
        pad_lo = (pad_total + self.up_factor) // 2 # Shift sample locations according to the symmetric interpretation (Appendix C.3).
        pad_hi = pad_total - pad_lo
        self.padding = [int(pad_lo[0]), int(pad_hi[0]), int(pad_lo[1]), int(pad_hi[1])]

    def forward(self, x, w, heatmap=None, noise_mode='random', force_fp32=False, update_emas=False):
        assert noise_mode in ['random', 'const', 'none'] # unused
        misc.assert_shape(x, [None, self.in_channels, int(self.in_size[1]), int(self.in_size[0])])
        misc.assert_shape(w, [x.shape[0], self.w_dim])

        # Track input magnitude.
        if update_emas:
            with torch.autograd.profiler.record_function('update_magnitude_ema'):
                magnitude_cur = x.detach().to(torch.float32).square().mean()
                self.magnitude_ema.copy_(magnitude_cur.lerp(self.magnitude_ema, self.magnitude_ema_beta))
        input_gain = self.magnitude_ema.rsqrt()
        
        if self.use_sel:
            x = self.sel(x, heatmap).to(x.dtype)

        # Execute affine layer.
        styles = self.affine(w)
        if self.is_torgb:
            weight_gain = 1 / np.sqrt(self.in_channels * (self.conv_kernel ** 2))
            styles = styles * weight_gain

        # Execute modulated conv2d.
        dtype = torch.float16 if (self.use_fp16 and not force_fp32 and x.device.type == 'cuda') else torch.float32
        x = modulated_conv2d(x=x.to(dtype), w=self.weight, s=styles,
            padding=self.conv_kernel-1, demodulate=(not self.is_torgb), input_gain=input_gain)

        # Execute bias, filtered leaky ReLU, and clamping.
        gain = 1 if self.is_torgb else np.sqrt(2)
        slope = 1 if self.is_torgb else 0.2
        if self.up_factor == 1 and self.down_factor == 1:
            x = bias_act.bias_act(x, self.bias.to(x.dtype), clamp=self.conv_clamp, act='lrelu')
        else:
            x = filtered_lrelu.filtered_lrelu(x=x, fu=self.up_filter, fd=self.down_filter, b=self.bias.to(x.dtype),
                up=self.up_factor, down=self.down_factor, padding=self.padding, gain=gain, slope=slope, clamp=self.conv_clamp)

        # Ensure correct shape and dtype.
        misc.assert_shape(x, [None, self.out_channels, int(self.out_size[1]), int(self.out_size[0])])
        assert x.dtype == dtype
        return x

    @staticmethod
    def design_lowpass_filter(numtaps, cutoff, width, fs, radial=False):
        assert numtaps >= 1

        # Identity filter.
        if numtaps == 1:
            return None

        # Separable Kaiser low-pass filter.
        if not radial:
            f = scipy.signal.firwin(numtaps=numtaps, cutoff=cutoff, width=width, fs=fs)
            return torch.as_tensor(f, dtype=torch.float32)

        # Radially symmetric jinc-based filter.
        x = (np.arange(numtaps) - (numtaps - 1) / 2) / fs
        r = np.hypot(*np.meshgrid(x, x))
        f = scipy.special.j1(2 * cutoff * (np.pi * r)) / (np.pi * r)
        beta = scipy.signal.kaiser_beta(scipy.signal.kaiser_atten(numtaps, width / (fs / 2)))
        w = np.kaiser(numtaps, beta)
        f *= np.outer(w, w)
        f /= np.sum(f)
        return torch.as_tensor(f, dtype=torch.float32)

    def extra_repr(self):
        return '\n'.join([
            f'w_dim={self.w_dim:d}, is_torgb={self.is_torgb},',
            f'is_critically_sampled={self.is_critically_sampled}, use_fp16={self.use_fp16},',
            f'in_sampling_rate={self.in_sampling_rate:g}, out_sampling_rate={self.out_sampling_rate:g},',
            f'in_cutoff={self.in_cutoff:g}, out_cutoff={self.out_cutoff:g},',
            f'in_half_width={self.in_half_width:g}, out_half_width={self.out_half_width:g},',
            f'in_size={list(self.in_size)}, out_size={list(self.out_size)},',
            f'in_channels={self.in_channels:d}, out_channels={self.out_channels:d}'])

#----------------------------------------------------------------------------

class SynthesisNetwork(torch.nn.Module):
    def __init__(self,
        w_dim,                          # Intermediate latent (W) dimensionality.
        img_resolution,                 # Output image resolution.
        img_channels,                   # Number of color channels.
        channel_base        = 32768,    # Overall multiplier for the number of channels.
        channel_max         = 512,      # Maximum number of channels in any layer.
        num_layers          = 14,       # Total number of layers, excluding Fourier features and ToRGB.
        num_critical        = 2,        # Number of critically sampled layers at the end.
        first_cutoff        = 2,        # Cutoff frequency of the first layer (f_{c,0}).
        first_stopband      = 2**2.1,   # Minimum stopband of the first layer (f_{t,0}).
        last_stopband_rel   = 2**0.3,   # Minimum stopband of the last layer, expressed relative to the cutoff.
        margin_size         = 10,       # Number of additional pixels outside the image.
        output_scale        = 0.25,     # Scale factor for the output image.
        num_fp16_res        = 4,        # Use FP16 for the N highest resolutions.
        label_nc            = 0,
        use_sel             = False,
        sel_type            = 'pro',
        low_rank             = None,
        offset_scale        = '0,0',
        bound_len           = 0.5,
        wo_transform        = False,
        **layer_kwargs,                 # Arguments for SynthesisLayer.
    ):
        super().__init__()
        self.w_dim = w_dim
        self.num_ws = num_layers + 2
        self.img_resolution = img_resolution
        self.img_channels = img_channels
        self.num_layers = num_layers
        self.num_critical = num_critical
        self.margin_size = margin_size
        self.output_scale = output_scale
        self.num_fp16_res = num_fp16_res

        self.use_sel = use_sel
        self.sel_type = sel_type

        # Geometric progression of layer cutoffs and min. stopbands.
        last_cutoff = self.img_resolution / 2 # f_{c,N}
        last_stopband = last_cutoff * last_stopband_rel # f_{t,N}
        exponents = np.minimum(np.arange(self.num_layers + 1) / (self.num_layers - self.num_critical), 1)
        cutoffs = first_cutoff * (last_cutoff / first_cutoff) ** exponents # f_c[i]
        stopbands = first_stopband * (last_stopband / first_stopband) ** exponents # f_t[i]

        # Compute remaining layer parameters.
        sampling_rates = np.exp2(np.ceil(np.log2(np.minimum(stopbands * 2, self.img_resolution)))) # s[i]
        half_widths = np.maximum(stopbands, sampling_rates / 2) - cutoffs # f_h[i]
        sizes = sampling_rates + self.margin_size * 2
        sizes[-2:] = self.img_resolution
        channels = np.rint(np.minimum((channel_base / 2) / cutoffs, channel_max))
        channels[-1] = self.img_channels

        # Construct layers.
        self.input = SynthesisInput(
            w_dim=self.w_dim, channels=int(channels[0]), size=int(sizes[0]),
            sampling_rate=sampling_rates[0], bandwidth=cutoffs[0], offset_scale=offset_scale, bound_len=bound_len, wo_transform=wo_transform)
        self.layer_names = []
        for idx in range(self.num_layers + 1):
            prev = max(idx - 1, 0)
            is_torgb = (idx == self.num_layers)
            is_critically_sampled = (idx >= self.num_layers - self.num_critical)
            use_fp16 = (sampling_rates[idx] * (2 ** self.num_fp16_res) > self.img_resolution)
            layer = SynthesisLayer(
                w_dim=self.w_dim, is_torgb=is_torgb, is_critically_sampled=is_critically_sampled, use_fp16=use_fp16,
                in_channels=int(channels[prev]), out_channels= int(channels[idx]),
                in_size=int(sizes[prev]), out_size=int(sizes[idx]),
                in_sampling_rate=int(sampling_rates[prev]), out_sampling_rate=int(sampling_rates[idx]),
                in_cutoff=cutoffs[prev], out_cutoff=cutoffs[idx],
                in_half_width=half_widths[prev], out_half_width=half_widths[idx],
                use_sel=use_sel, sel_type=sel_type, label_nc=label_nc, low_rank=low_rank,
                **layer_kwargs)
            name = f'L{idx}_{layer.out_size[0]}_{layer.out_channels}'
            setattr(self, name, layer)
            self.layer_names.append(name)

    def forward(self, ws, heatmap=None, **layer_kwargs):
        misc.assert_shape(ws, [None, self.num_ws, self.w_dim])
        ws = ws.to(torch.float32).unbind(dim=1)

        # Execute layers.
        x = self.input(ws[0])
        for name, w in zip(self.layer_names, ws[1:]):
            x = getattr(self, name)(x, w, heatmap, **layer_kwargs)
        if self.output_scale != 1:
            x = x * self.output_scale

        # Ensure correct shape and dtype.
        misc.assert_shape(x, [None, self.img_channels, self.img_resolution, self.img_resolution])
        x = x.to(torch.float32)
        return x

    def extra_repr(self):
        return '\n'.join([
            f'w_dim={self.w_dim:d}, num_ws={self.num_ws:d},',
            f'img_resolution={self.img_resolution:d}, img_channels={self.img_channels:d},',
            f'num_layers={self.num_layers:d}, num_critical={self.num_critical:d},',
            f'margin_size={self.margin_size:d}, num_fp16_res={self.num_fp16_res:d}'])

#----------------------------------------------------------------------------

class Generator(torch.nn.Module):
    def __init__(self,
        z_dim,                      # Input latent (Z) dimensionality.
        c_dim,                      # Conditioning label (C) dimensionality.
        w_dim,                      # Intermediate latent (W) dimensionality.
        img_resolution,             # Output resolution.
        img_channels,               # Number of output color channels.
        label_nc            = None,
        use_sel             = False,
        sel_type             = False,
        low_rank            = None,
        offset_scale        = '0,0',
        bound_len           = 0.5,
        wo_transform        = False,
        mapping_kwargs      = {},   # Arguments for MappingNetwork.
        **synthesis_kwargs,         # Arguments for SynthesisNetwork.
    ):
        super().__init__()
        self.z_dim = z_dim
        self.c_dim = c_dim
        self.w_dim = w_dim
        self.img_resolution = img_resolution
        self.img_channels = img_channels
        self.synthesis = SynthesisNetwork(w_dim=w_dim, img_resolution=img_resolution, img_channels=img_channels, label_nc=label_nc, use_sel=use_sel, sel_type=sel_type, low_rank=low_rank, offset_scale=offset_scale, bound_len=bound_len, wo_transform=wo_transform, **synthesis_kwargs)
        self.num_ws = self.synthesis.num_ws
        self.mapping = MappingNetwork(z_dim=z_dim, c_dim=c_dim, w_dim=w_dim, num_ws=self.num_ws, **mapping_kwargs)

    def forward(self, z, c, truncation_psi=1, truncation_cutoff=None, update_emas=False, heatmap=None, **synthesis_kwargs):
        ws = self.mapping(z, c, truncation_psi=truncation_psi, truncation_cutoff=truncation_cutoff, update_emas=update_emas)
        img = self.synthesis(ws, update_emas=update_emas, spatial_map=heatmap, **synthesis_kwargs)
        return img

#----------------------------------------------------------------------------