mrq/DL-Art-School
1
3
Fork 5
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Remove separated vgg discriminator

Browse Source 
Checkpointing happens inline instead. Was a dumb idea..

Also fixes some loss reporting issues.
This commit is contained in:
James Betker 2020-10-18 12:10:24 -06:00
parent c709d38cd5
commit 7df378a944
3 changed files with 15 additions and 60 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

67
codes/models/archs/discriminator_vgg_arch.py
View File

@ -3,6 +3,7 @@ import torch.nn as nn
from models.archs.arch_util import ConvBnLelu, ConvGnLelu, ExpansionBlock, ConvGnSilu
import torch.nn.functional as F
from models.archs.SwitchedResidualGenerator_arch import gather_2d
from utils.util import checkpoint
class Discriminator_VGG_128(nn.Module):
@ -79,8 +80,10 @@ class Discriminator_VGG_128(nn.Module):
class Discriminator_VGG_128_GN(nn.Module):
# input_img_factor = multiplier to support images over 128x128. Only certain factors are supported.
def __init__(self, in_nc, nf, input_img_factor=1):
def __init__(self, in_nc, nf, input_img_factor=1, do_checkpointing=False):
super(Discriminator_VGG_128_GN, self).__init__()
self.do_checkpointing = do_checkpointing
# [64, 128, 128]
self.conv0_0 = nn.Conv2d(in_nc, nf, 3, 1, 1, bias=True)
self.conv0_1 = nn.Conv2d(nf, nf, 4, 2, 1, bias=False)
@ -113,7 +116,7 @@ class Discriminator_VGG_128_GN(nn.Module):
self.linear1 = nn.Linear(int(final_nf * 4 * input_img_factor * 4 * input_img_factor), 100)
self.linear2 = nn.Linear(100, 1)
def forward(self, x):
def compute_body(self, x):
fea = self.lrelu(self.conv0_0(x))
fea = self.lrelu(self.bn0_1(self.conv0_1(fea)))
@ -130,63 +133,13 @@ class Discriminator_VGG_128_GN(nn.Module):
fea = self.lrelu(self.bn4_0(self.conv4_0(fea)))
fea = self.lrelu(self.bn4_1(self.conv4_1(fea)))
fea = fea.contiguous().view(fea.size(0), -1)
fea = self.lrelu(self.linear1(fea))
out = self.linear2(fea)
return out
from utils.util import checkpoint
class Discriminator_VGG_128_GN_Checkpointed(nn.Module):
# input_img_factor = multiplier to support images over 128x128. Only certain factors are supported.
def __init__(self, in_nc, nf, input_img_factor=1):
super(Discriminator_VGG_128_GN_Checkpointed, self).__init__()
# [64, 128, 128]
conv0_0 = nn.Conv2d(in_nc, nf, 3, 1, 1, bias=True)
conv0_1 = nn.Conv2d(nf, nf, 4, 2, 1, bias=False)
bn0_1 = nn.GroupNorm(8, nf, affine=True)
# [64, 64, 64]
conv1_0 = nn.Conv2d(nf, nf * 2, 3, 1, 1, bias=False)
bn1_0 = nn.GroupNorm(8, nf * 2, affine=True)
conv1_1 = nn.Conv2d(nf * 2, nf * 2, 4, 2, 1, bias=False)
bn1_1 = nn.GroupNorm(8, nf * 2, affine=True)
# [128, 32, 32]
conv2_0 = nn.Conv2d(nf * 2, nf * 4, 3, 1, 1, bias=False)
bn2_0 = nn.GroupNorm(8, nf * 4, affine=True)
conv2_1 = nn.Conv2d(nf * 4, nf * 4, 4, 2, 1, bias=False)
bn2_1 = nn.GroupNorm(8, nf * 4, affine=True)
# [256, 16, 16]
conv3_0 = nn.Conv2d(nf * 4, nf * 8, 3, 1, 1, bias=False)
bn3_0 = nn.GroupNorm(8, nf * 8, affine=True)
conv3_1 = nn.Conv2d(nf * 8, nf * 8, 4, 2, 1, bias=False)
bn3_1 = nn.GroupNorm(8, nf * 8, affine=True)
# [512, 8, 8]
conv4_0 = nn.Conv2d(nf * 8, nf * 8, 3, 1, 1, bias=False)
bn4_0 = nn.GroupNorm(8, nf * 8, affine=True)
conv4_1 = nn.Conv2d(nf * 8, nf * 8, 4, 2, 1, bias=False)
bn4_1 = nn.GroupNorm(8, nf * 8, affine=True)
final_nf = nf * 8
# activation function
self.lrelu = nn.LeakyReLU(negative_slope=0.2, inplace=True)
self.body = nn.Sequential(conv0_0, self.lrelu,
conv0_1, bn0_1, self.lrelu,
conv1_0, bn1_0, self.lrelu,
conv1_1, bn1_1, self.lrelu,
conv2_0, bn2_0, self.lrelu,
conv2_1, bn2_1, self.lrelu,
conv3_0, bn3_0, self.lrelu,
conv3_1, bn3_1, self.lrelu,
conv4_0, bn4_0, self.lrelu,
conv4_1, bn4_1, self.lrelu)
self.linear1 = nn.Linear(int(final_nf * 4 * input_img_factor * 4 * input_img_factor), 100)
self.linear2 = nn.Linear(100, 1)
return fea
def forward(self, x):
fea = checkpoint(self.body, x)
if self.do_checkpointing:
fea = checkpoint(self.compute_body, x)
else:
fea = self.compute_body(x)
fea = fea.contiguous().view(fea.size(0), -1)
fea = self.lrelu(self.linear1(fea))
out = self.linear2(fea)

2
codes/models/networks.py
View File

@ -172,7 +172,7 @@ def define_D_net(opt_net, img_sz=None, wrap=False):
if wrap:
netD = GradDiscWrapper(netD)
elif which_model == 'discriminator_vgg_128_gn_checkpointed':
netD = SRGAN_arch.Discriminator_VGG_128_GN_Checkpointed(in_nc=opt_net['in_nc'], nf=opt_net['nf'], input_img_factor=img_sz / 128)
netD = SRGAN_arch.Discriminator_VGG_128_GN(in_nc=opt_net['in_nc'], nf=opt_net['nf'], input_img_factor=img_sz / 128, do_checkpointing=True)
elif which_model == 'discriminator_resnet':
netD = DiscriminatorResnet_arch.fixup_resnet34(num_filters=opt_net['nf'], num_classes=1, input_img_size=img_sz)
elif which_model == 'discriminator_resnet_50':

6
codes/models/steps/losses.py
View File

@ -169,8 +169,10 @@ class GeneratorGanLoss(ConfigurableLoss):
if self.detach_real:
pred_d_real = pred_d_real.detach()
pred_g_fake = netD(*fake)
d_fake_diff = self.criterion(pred_g_fake - torch.mean(pred_d_real), True)
self.metrics.append(("d_fake_diff", torch.mean(d_fake_diff)))
loss = (self.criterion(pred_d_real - torch.mean(pred_g_fake), False) +
self.criterion(pred_g_fake - torch.mean(pred_d_real), True)) / 2
d_fake_diff) / 2
else:
raise NotImplementedError
if self.min_loss != 0:
@ -234,10 +236,10 @@ class DiscriminatorGanLoss(ConfigurableLoss):
if self.min_loss != 0:
self.loss_rotating_buffer[self.rb_ptr] = loss.item()
self.rb_ptr = (self.rb_ptr + 1) % self.loss_rotating_buffer.shape[0]
self.metrics.append(("loss_counter", self.losses_computed))
if torch.mean(self.loss_rotating_buffer) < self.min_loss:
return 0
self.losses_computed += 1
self.metrics.append(("loss_counter", self.losses_computed))
return loss