DL-Art-School/codes/models/archs/StructuredSwitchedGenerator.py
James Betker 19a4075e1e Allow checkpointing to be disabled in the options file
Also makes options a global variable for usage in utils.
2020-10-03 11:03:28 -06:00

246 lines
13 KiB
Python

import math
import functools
from models.archs.arch_util import MultiConvBlock, ConvGnLelu, ConvGnSilu, ReferenceJoinBlock
from models.archs.SwitchedResidualGenerator_arch import ConfigurableSwitchComputer, gather_2d
from models.archs.SPSR_arch import ImageGradientNoPadding
from torch import nn
import torch
import torch.nn.functional as F
from switched_conv_util import save_attention_to_image_rgb
from switched_conv import compute_attention_specificity
import os
import torchvision
from utils.util import checkpoint
# VGG-style layer with Conv(stride2)->BN->Activation->Conv->BN->Activation
# Doubles the input filter count.
class HalvingProcessingBlock(nn.Module):
def __init__(self, filters, factor=2):
super(HalvingProcessingBlock, self).__init__()
self.bnconv1 = ConvGnSilu(filters, filters, norm=False, bias=False)
self.bnconv2 = ConvGnSilu(filters, int(filters * factor), kernel_size=1, stride=2, norm=True, bias=False)
def forward(self, x):
x = self.bnconv1(x)
return self.bnconv2(x)
class ExpansionBlock2(nn.Module):
def __init__(self, filters_in, filters_out=None, block=ConvGnSilu, factor=2):
super(ExpansionBlock2, self).__init__()
if filters_out is None:
filters_out = int(filters_in / factor)
self.decimate = block(filters_in, filters_out, kernel_size=1, bias=False, activation=True, norm=False)
self.process_passthrough = block(filters_out, filters_out, kernel_size=3, bias=True, activation=True, norm=False)
self.conjoin = block(filters_out*2, filters_out*2, kernel_size=1, bias=False, activation=True, norm=False)
self.reduce = block(filters_out*2, filters_out, kernel_size=1, bias=False, activation=False, norm=True)
# input is the feature signal with shape (b, f, w, h)
# passthrough is the structure signal with shape (b, f/2, w*2, h*2)
# output is conjoined upsample with shape (b, f/2, w*2, h*2)
def forward(self, input, passthrough):
x = F.interpolate(input, scale_factor=2, mode="nearest")
x = self.decimate(x)
p = self.process_passthrough(passthrough)
x = self.conjoin(torch.cat([x, p], dim=1))
return self.reduce(x)
# Basic convolutional upsampling block that uses interpolate.
class UpconvBlock(nn.Module):
def __init__(self, filters_in, filters_out=None, block=ConvGnSilu, norm=True, activation=True, bias=False):
super(UpconvBlock, self).__init__()
self.reduce = block(filters_in, filters_out, kernel_size=1, bias=False, activation=False, norm=False)
self.process = block(filters_out, filters_out, kernel_size=3, bias=bias, activation=activation, norm=norm)
def forward(self, x):
x = self.reduce(x)
x = F.interpolate(x, scale_factor=2, mode="nearest")
return self.process(x)
class QueryKeyMultiplexer(nn.Module):
def __init__(self, nf, multiplexer_channels, embedding_channels=216, reductions=3):
super(QueryKeyMultiplexer, self).__init__()
# Blocks used to create the query
self.input_process = ConvGnSilu(nf, nf, activation=True, norm=False, bias=True)
self.embedding_process = ConvGnSilu(embedding_channels, 128, kernel_size=1, activation=True, norm=False, bias=True)
self.reduction_blocks = nn.ModuleList([HalvingProcessingBlock(int(nf * 1.5 ** i), factor=1.5) for i in range(reductions)])
reduction_filters = int(nf * 1.5 ** reductions)
self.processing_blocks = nn.Sequential(
ConvGnSilu(reduction_filters + 128, reduction_filters + 64, kernel_size=1, activation=True, norm=False, bias=True),
ConvGnSilu(reduction_filters + 64, reduction_filters, kernel_size=1, activation=True, norm=False, bias=False),
ConvGnSilu(reduction_filters, reduction_filters, kernel_size=3, activation=True, norm=True, bias=False),
ConvGnSilu(reduction_filters, reduction_filters, kernel_size=3, activation=True, norm=True, bias=False))
self.expansion_blocks = nn.ModuleList([ExpansionBlock2(int(reduction_filters // (1.5 ** i)), factor=1.5) for i in range(reductions)])
# Blocks used to create the key
self.key_process = ConvGnSilu(nf, nf, kernel_size=1, activation=True, norm=False, bias=False)
# Postprocessing blocks.
self.query_key_combine = ConvGnSilu(nf*2, nf, kernel_size=1, activation=True, norm=False, bias=False)
self.cbl1 = ConvGnSilu(nf, nf // 2, kernel_size=1, norm=True, bias=False, num_groups=4)
self.cbl2 = ConvGnSilu(nf // 2, 1, kernel_size=1, norm=False, bias=False)
def forward(self, x, embedding, transformations):
q = self.input_process(x)
embedding = self.embedding_process(embedding)
reduction_identities = []
for b in self.reduction_blocks:
reduction_identities.append(q)
q = b(q)
q = self.processing_blocks(torch.cat([q, embedding], dim=1))
for i, b in enumerate(self.expansion_blocks):
q = b(q, reduction_identities[-i - 1])
b, t, f, h, w = transformations.shape
k = transformations.view(b * t, f, h, w)
k = self.key_process(k)
q = q.view(b, 1, f, h, w).repeat(1, t, 1, 1, 1).view(b * t, f, h, w)
v = self.query_key_combine(torch.cat([q, k], dim=1))
v = self.cbl1(v)
v = self.cbl2(v)
return v.view(b, t, h, w)
# Computes a linear latent by performing processing on the reference image and returning the filters of a single point,
# which should be centered on the image patch being processed.
#
# Output is base_filters * 1.5^3.
class ReferenceImageBranch(nn.Module):
def __init__(self, base_filters=64):
super(ReferenceImageBranch, self).__init__()
final_filters = int(base_filters*1.5**3)
self.features = nn.Sequential(ConvGnSilu(4, base_filters, kernel_size=7, bias=True),
HalvingProcessingBlock(base_filters, factor=1.5),
HalvingProcessingBlock(int(base_filters*1.5), factor=1.5),
HalvingProcessingBlock(int(base_filters*1.5**2), factor=1.5),
ConvGnSilu(final_filters, final_filters, activation=True, norm=True, bias=False))
# center_point is a [b,2] long tensor describing the center point of where the patch was taken from the reference
# image.
def forward(self, x, center_point):
x = self.features(x)
return gather_2d(x, center_point // 8) # Divide by 8 to scale the center_point down.
class SSGr1(nn.Module):
def __init__(self, in_nc, out_nc, nf, xforms=8, upscale=4, init_temperature=10):
super(SSGr1, self).__init__()
n_upscale = int(math.log(upscale, 2))
self.nf = nf
# processing the input embedding
self.reference_embedding = ReferenceImageBranch(nf)
# switch options
transformation_filters = nf
self.transformation_counts = xforms
multiplx_fn = functools.partial(QueryKeyMultiplexer, transformation_filters)
transform_fn = functools.partial(MultiConvBlock, transformation_filters, int(transformation_filters * 1.25),
transformation_filters, kernel_size=3, depth=4,
weight_init_factor=.1)
# Feature branch
self.model_fea_conv = ConvGnLelu(in_nc, nf, kernel_size=3, norm=False, activation=False)
self.sw1 = ConfigurableSwitchComputer(transformation_filters, multiplx_fn,
pre_transform_block=None, transform_block=transform_fn,
attention_norm=True,
transform_count=self.transformation_counts, init_temp=init_temperature,
add_scalable_noise_to_transforms=False, feed_transforms_into_multiplexer=True)
# Grad branch. Note - groupnorm on this branch is REALLY bad. Avoid it like the plague.
self.get_g_nopadding = ImageGradientNoPadding()
self.grad_conv = ConvGnLelu(in_nc, nf, kernel_size=3, norm=False, activation=False, bias=False)
self.grad_ref_join = ReferenceJoinBlock(nf, residual_weight_init_factor=.3, final_norm=False, kernel_size=1, depth=2)
self.sw_grad = ConfigurableSwitchComputer(transformation_filters, multiplx_fn,
pre_transform_block=None, transform_block=transform_fn,
attention_norm=True,
transform_count=self.transformation_counts // 2, init_temp=init_temperature,
add_scalable_noise_to_transforms=False, feed_transforms_into_multiplexer=True)
self.grad_lr_conv = ConvGnLelu(nf, nf, kernel_size=3, norm=False, activation=True, bias=True)
self.upsample_grad = UpconvBlock(nf, nf // 2, block=ConvGnLelu, norm=False, activation=True, bias=False)
self.grad_branch_output_conv = ConvGnLelu(nf // 2, out_nc, kernel_size=1, norm=False, activation=False, bias=True)
# Join branch (grad+fea)
self.conjoin_ref_join = ReferenceJoinBlock(nf, residual_weight_init_factor=.3, kernel_size=1, depth=2)
self.conjoin_sw = ConfigurableSwitchComputer(transformation_filters, multiplx_fn,
pre_transform_block=None, transform_block=transform_fn,
attention_norm=True,
transform_count=self.transformation_counts, init_temp=init_temperature,
add_scalable_noise_to_transforms=False, feed_transforms_into_multiplexer=True)
self.final_lr_conv = ConvGnLelu(nf, nf, kernel_size=3, norm=False, activation=True, bias=True)
self.upsample = UpconvBlock(nf, nf // 2, block=ConvGnLelu, norm=False, activation=True, bias=True)
self.final_hr_conv1 = ConvGnLelu(nf // 2, nf // 2, kernel_size=3, norm=False, activation=False, bias=True)
self.final_hr_conv2 = ConvGnLelu(nf // 2, out_nc, kernel_size=3, norm=False, activation=False, bias=False)
self.switches = [self.sw1, self.sw_grad, self.conjoin_sw]
self.attentions = None
self.lr = None
self.init_temperature = init_temperature
self.final_temperature_step = 10000
def forward(self, x, ref, ref_center):
# The attention_maps debugger outputs <x>. Save that here.
self.lr = x.detach().cpu()
x_grad = self.get_g_nopadding(x)
ref_code = checkpoint(self.reference_embedding, ref, ref_center)
ref_embedding = ref_code.view(-1, ref_code.shape[1], 1, 1).repeat(1, 1, x.shape[2] // 8, x.shape[3] // 8)
x = self.model_fea_conv(x)
x1 = x
x1, a1 = self.sw1(x1, True, identity=x, att_in=(x1, ref_embedding))
x_grad = self.grad_conv(x_grad)
x_grad_identity = x_grad
x_grad, grad_fea_std = self.grad_ref_join(x_grad, x1)
x_grad, a3 = self.sw_grad(x_grad, True, identity=x_grad_identity, att_in=(x_grad, ref_embedding))
x_grad = self.grad_lr_conv(x_grad)
x_grad_out = self.upsample_grad(x_grad)
x_grad_out = self.grad_branch_output_conv(x_grad_out)
x_out = x1
x_out, fea_grad_std = self.conjoin_ref_join(x_out, x_grad)
x_out, a4 = self.conjoin_sw(x_out, True, identity=x1, att_in=(x_out, ref_embedding))
x_out = self.final_lr_conv(x_out)
x_out = checkpoint(self.upsample, x_out)
x_out = self.final_hr_conv2(x_out)
self.attentions = [a1, a3, a4]
self.grad_fea_std = grad_fea_std.detach().cpu()
self.fea_grad_std = fea_grad_std.detach().cpu()
return x_grad_out, x_out, x_grad
def set_temperature(self, temp):
[sw.set_temperature(temp) for sw in self.switches]
def update_for_step(self, step, experiments_path='.'):
if self.attentions:
temp = max(1, 1 + self.init_temperature *
(self.final_temperature_step - step) / self.final_temperature_step)
self.set_temperature(temp)
if step % 200 == 0:
output_path = os.path.join(experiments_path, "attention_maps")
prefix = "amap_%i_a%i_%%i.png"
[save_attention_to_image_rgb(output_path, self.attentions[i], self.transformation_counts, prefix % (step, i), step, output_mag=False) for i in range(len(self.attentions))]
torchvision.utils.save_image(self.lr, os.path.join(experiments_path, "attention_maps", "amap_%i_base_image.png" % (step,)))
def get_debug_values(self, step, net_name):
temp = self.switches[0].switch.temperature
mean_hists = [compute_attention_specificity(att, 2) for att in self.attentions]
means = [i[0] for i in mean_hists]
hists = [i[1].clone().detach().cpu().flatten() for i in mean_hists]
val = {"switch_temperature": temp,
"grad_branch_feat_intg_std_dev": self.grad_fea_std,
"conjoin_branch_grad_intg_std_dev": self.fea_grad_std}
for i in range(len(means)):
val["switch_%i_specificity" % (i,)] = means[i]
val["switch_%i_histogram" % (i,)] = hists[i]
return val