基于pytorch的DCGAN人脸数据生成
%matplotlib inline
DCGAN Tutorial
Author: Nathan Inkawhich __
Introduction
This tutorial will give an introduction to DCGANs through an example. We
will train a generative adversarial network (GAN) to generate new
celebrities after showing it pictures of many real celebrities. Most of
the code here is from the dcgan implementation in
pytorch/examples __, and this
document will give a thorough explanation of the implementation and shed
light on how and why this model works. But don’t worry, no prior
knowledge of GANs is required, but it may require a first-timer to spend
some time reasoning about what is actually happening under the hood.
Also, for the sake of time it will help to have a GPU, or two. Lets
start from the beginning.
Generative Adversarial Networks
What is a GAN?
GANs are a framework for teaching a DL model to capture the training
data’s distribution so we can generate new data from that same
distribution. GANs were invented by Ian Goodfellow in 2014 and first
described in the paper `Generative Adversarial
Nets `__.
They are made of two distinct models, a *generator* and a
*discriminator*. The job of the generator is to spawn ‘fake’ images that
look like the training images. The job of the discriminator is to look
at an image and output whether or not it is a real training image or a
fake image from the generator. During training, the generator is
constantly trying to outsmart the discriminator by generating better and
better fakes, while the discriminator is working to become a better
detective and correctly classify the real and fake images. The
equilibrium of this game is when the generator is generating perfect
fakes that look as if they came directly from the training data, and the
discriminator is left to always guess at 50% confidence that the
generator output is real or fake.
Now, lets define some notation to be used throughout tutorial starting
with the discriminator. Let $x$ be data representing an image.
$D(x)$ is the discriminator network which outputs the (scalar)
probability that $x$ came from training data rather than the
generator. Here, since we are dealing with images the input to
$D(x)$ is an image of CHW size 3x64x64. Intuitively, $D(x)$
should be HIGH when $x$ comes from training data and LOW when
$x$ comes from the generator. $D(x)$ can also be thought of
as a traditional binary classifier.
For the generator’s notation, let $z$ be a latent space vector
sampled from a standard normal distribution. $G(z)$ represents the
generator function which maps the latent vector $z$ to data-space.
The goal of $G$ is to estimate the distribution that the training
data comes from ($p_{data}$) so it can generate fake samples from
that estimated distribution ($p_g$).
So, $D(G(z))$ is the probability (scalar) that the output of the
generator $G$ is a real image. As described in `Goodfellow’s
paper `__,
$D$ and $G$ play a minimax game in which $D$ tries to
maximize the probability it correctly classifies reals and fakes
($logD(x)$), and $G$ tries to minimize the probability that
$D$ will predict its outputs are fake ($log(1-D(G(x)))$).
From the paper, the GAN loss function is
\begin{align}\underset{G}{\text{min}} \underset{D}{\text{max}}V(D,G) = \mathbb{E}_{x\sim p_{data}(x)}\big[logD(x)\big] + \mathbb{E}_{z\sim p_{z}(z)}\big[log(1-D(G(z)))\big]\end{align}
In theory, the solution to this minimax game is where
$p_g = p_{data}$, and the discriminator guesses randomly if the
inputs are real or fake. However, the convergence theory of GANs is
still being actively researched and in reality models do not always
train to this point.
What is a DCGAN?
A DCGAN is a direct extension of the GAN described above, except that it
explicitly uses convolutional and convolutional-transpose layers in the
discriminator and generator, respectively. It was first described by
Radford et. al. in the paper Unsupervised Representation Learning With Deep Convolutional Generative Adversarial Networks . The discriminator
is made up of strided
convolution
layers, batch norm __
layers, and
LeakyReLU __
activations. The input is a 3x64x64 input image and the output is a
scalar probability that the input is from the real data distribution.
The generator is comprised of
convolutional-transpose __
layers, batch norm layers, and
ReLU __ activations. The
input is a latent vector, z z z, that is drawn from a standard
normal distribution and the output is a 3x64x64 RGB image. The strided
conv-transpose layers allow the latent vector to be transformed into a
volume with the same shape as an image. In the paper, the authors also
give some tips about how to setup the optimizers, how to calculate the
loss functions, and how to initialize the model weights, all of which
will be explained in the coming sections.
from __future__ import print_function
#%matplotlib inline
import argparse
import os
import random
import torch
import torch.nn as nn
import torch.nn.parallel
import torch.backends.cudnn as cudnn
import torch.optim as optim
import torch.utils.data
import torchvision.datasets as dset
import torchvision.transforms as transforms
import torchvision.utils as vutils
import numpy as np
import matplotlib.pyplot as plt
import matplotlib.animation as animation
from IPython.display import HTML
# Set random seed for reproducibility
manualSeed = 999
#manualSeed = random.randint(1, 10000) # use if you want new results
print("Random Seed: ", manualSeed)
random.seed(manualSeed)
torch.manual_seed(manualSeed)
Random Seed: 999
Inputs
Let’s define some inputs for the run:
- dataroot - the path to the root of the dataset folder. We will
talk more about the dataset in the next section - workers - the number of worker threads for loading the data with
the DataLoader - batch_size - the batch size used in training. The DCGAN paper
uses a batch size of 128 - image_size - the spatial size of the images used for training.
This implementation defaults to 64x64. If another size is desired,
the structures of D and G must be changed. See
here__ for more
details - nc - number of color channels in the input images. For color
images this is 3 - nz - length of latent vector
- ngf - relates to the depth of feature maps carried through the
generator - ndf - sets the depth of feature maps propagated through the
discriminator - num_epochs - number of training epochs to run. Training for
longer will probably lead to better results but will also take much
longer - lr - learning rate for training. As described in the DCGAN paper,
this number should be 0.0002 - beta1 - beta1 hyperparameter for Adam optimizers. As described in
paper, this number should be 0.5 - ngpu - number of GPUs available. If this is 0, code will run in
CPU mode. If this number is greater than 0 it will run on that number
of GPUs
# Root directory for dataset
dataroot = "data/celeba"
# Number of workers for dataloader
workers = 2
# Batch size during training
batch_size = 128
# Spatial size of training images. All images will be resized to this
# size using a transformer.
image_size = 64
# Number of channels in the training images. For color images this is 3
nc = 3
# Size of z latent vector (i.e. size of generator input)
nz = 100
# Size of feature maps in generator
ngf = 64
# Size of feature maps in discriminator
ndf = 64
# Number of training epochs
num_epochs = 5
# Learning rate for optimizers
lr = 0.0002
# Beta1 hyperparam for Adam optimizers
beta1 = 0.5
# Number of GPUs available. Use 0 for CPU mode.
ngpu = 1
Data
In this tutorial we will use the Celeb-A Faces dataset __ which can
be downloaded at the linked site, or in Google Drive __.
The dataset will download as a file named img_align_celeba.zip. Once
downloaded, create a directory named celeba and extract the zip file
into that directory. Then, set the dataroot input for this notebook to
the celeba directory you just created. The resulting directory
structure should be:
::
/path/to/celeba
-> img_align_celeba
-> 188242.jpg
-> 173822.jpg
-> 284702.jpg
-> 537394.jpg
…
This is an important step because we will be using the ImageFolder
dataset class, which requires there to be subdirectories in the
dataset’s root folder. Now, we can create the dataset, create the
dataloader, set the device to run on, and finally visualize some of the
training data.
# We can use an image folder dataset the way we have it setup.
# Create the dataset
dataset = dset.ImageFolder(root=dataroot,
transform=transforms.Compose([
transforms.Resize(image_size),
transforms.CenterCrop(image_size),
transforms.ToTensor(),
transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5)),
]))
# Create the dataloader
dataloader = torch.utils.data.DataLoader(dataset, batch_size=batch_size,
shuffle=True, num_workers=workers)
# Decide which device we want to run on
device = torch.device("cuda:0" if (torch.cuda.is_available() and ngpu > 0) else "cpu")
# Plot some training images
real_batch = next(iter(dataloader))
plt.figure(figsize=(8,8))
plt.axis("off")
plt.title("Training Images")
plt.imshow(np.transpose(vutils.make_grid(real_batch[0].to(device)[:64], padding=2, normalize=True).cpu(),(1,2,0)))
Implementation
With our input parameters set and the dataset prepared, we can now get
into the implementation. We will start with the weigth initialization
strategy, then talk about the generator, discriminator, loss functions,
and training loop in detail.
Weight Initialization
From the DCGAN paper, the authors specify that all model weights shall
be randomly initialized from a Normal distribution with mean=0,
stdev=0.02. The ``weights_init`` function takes an initialized model as
input and reinitializes all convolutional, convolutional-transpose, and
batch normalization layers to meet this criteria. This function is
applied to the models immediately after initialization.
```python
# custom weights initialization called on netG and netD
def weights_init(m):
classname = m.__class__.__name__
if classname.find('Conv') != -1:
nn.init.normal_(m.weight.data, 0.0, 0.02)
elif classname.find('BatchNorm') != -1:
nn.init.normal_(m.weight.data, 1.0, 0.02)
nn.init.constant_(m.bias.data, 0)
```
Generator
~~~~~~~~~
The generator, $G$, is designed to map the latent space vector
($z$) to data-space. Since our data are images, converting
$z$ to data-space means ultimately creating a RGB image with the
same size as the training images (i.e. 3x64x64). In practice, this is
accomplished through a series of strided two dimensional convolutional
transpose layers, each paired with a 2d batch norm layer and a relu
activation. The output of the generator is fed through a tanh function
to return it to the input data range of $[-1,1]$. It is worth
noting the existence of the batch norm functions after the
conv-transpose layers, as this is a critical contribution of the DCGAN
paper. These layers help with the flow of gradients during training. An
image of the generator from the DCGAN paper is shown below.
.. figure:: /_static/img/dcgan_generator.png
:alt: dcgan_generator
Notice, the how the inputs we set in the input section (*nz*, *ngf*, and
*nc*) influence the generator architecture in code. *nz* is the length
of the z input vector, *ngf* relates to the size of the feature maps
that are propagated through the generator, and *nc* is the number of
channels in the output image (set to 3 for RGB images). Below is the
code for the generator.
```python
# Generator Code
class Generator(nn.Module):
def __init__(self, ngpu):
super(Generator, self).__init__()
self.ngpu = ngpu
self.main = nn.Sequential(
# input is Z, going into a convolution
nn.ConvTranspose2d( nz, ngf * 8, 4, 1, 0, bias=False),
nn.BatchNorm2d(ngf * 8),
nn.ReLU(True),
# state size. (ngf*8) x 4 x 4
nn.ConvTranspose2d(ngf * 8, ngf * 4, 4, 2, 1, bias=False),
nn.BatchNorm2d(ngf * 4),
nn.ReLU(True),
# state size. (ngf*4) x 8 x 8
nn.ConvTranspose2d( ngf * 4, ngf * 2, 4, 2, 1, bias=False),
nn.BatchNorm2d(ngf * 2),
nn.ReLU(True),
# state size. (ngf*2) x 16 x 16
nn.ConvTranspose2d( ngf * 2, ngf, 4, 2, 1, bias=False),
nn.BatchNorm2d(ngf),
nn.ReLU(True),
# state size. (ngf) x 32 x 32
nn.ConvTranspose2d( ngf, nc, 4, 2, 1, bias=False),
nn.Tanh()
# state size. (nc) x 64 x 64
)
def forward(self, input):
return self.main(input)
```
Now, we can instantiate the generator and apply the ``weights_init``
function. Check out the printed model to see how the generator object is
structured.
```python
# Create the generator
netG = Generator(ngpu).to(device)
# Handle multi-gpu if desired
if (device.type == 'cuda') and (ngpu > 1):
netG = nn.DataParallel(netG, list(range(ngpu)))
# Apply the weights_init function to randomly initialize all weights
# to mean=0, stdev=0.2.
netG.apply(weights_init)
# Print the model
print(netG)
```
Generator(
(main): Sequential(
(0): ConvTranspose2d(100, 512, kernel_size=(4, 4), stride=(1, 1), bias=False)
(1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU(inplace=True)
(3): ConvTranspose2d(512, 256, kernel_size=(4, 4), stride=(2, 2), padding=(1, 1), bias=False)
(4): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(5): ReLU(inplace=True)
(6): ConvTranspose2d(256, 128, kernel_size=(4, 4), stride=(2, 2), padding=(1, 1), bias=False)
(7): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(8): ReLU(inplace=True)
(9): ConvTranspose2d(128, 64, kernel_size=(4, 4), stride=(2, 2), padding=(1, 1), bias=False)
(10): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(11): ReLU(inplace=True)
(12): ConvTranspose2d(64, 3, kernel_size=(4, 4), stride=(2, 2), padding=(1, 1), bias=False)
(13): Tanh()
)
)
Discriminator
~~~~~~~~~~~~~
As mentioned, the discriminator, $D$, is a binary classification
network that takes an image as input and outputs a scalar probability
that the input image is real (as opposed to fake). Here, $D$ takes
a 3x64x64 input image, processes it through a series of Conv2d,
BatchNorm2d, and LeakyReLU layers, and outputs the final probability
through a Sigmoid activation function. This architecture can be extended
with more layers if necessary for the problem, but there is significance
to the use of the strided convolution, BatchNorm, and LeakyReLUs. The
DCGAN paper mentions it is a good practice to use strided convolution
rather than pooling to downsample because it lets the network learn its
own pooling function. Also batch norm and leaky relu functions promote
healthy gradient flow which is critical for the learning process of both
$G$ and $D$.
Discriminator Code
```python
class Discriminator(nn.Module):
def __init__(self, ngpu):
super(Discriminator, self).__init__()
self.ngpu = ngpu
self.main = nn.Sequential(
# input is (nc) x 64 x 64
nn.Conv2d(nc, ndf, 4, 2, 1, bias=False),
nn.LeakyReLU(0.2, inplace=True),
# state size. (ndf) x 32 x 32
nn.Conv2d(ndf, ndf * 2, 4, 2, 1, bias=False),
nn.BatchNorm2d(ndf * 2),
nn.LeakyReLU(0.2, inplace=True),
# state size. (ndf*2) x 16 x 16
nn.Conv2d(ndf * 2, ndf * 4, 4, 2, 1, bias=False),
nn.BatchNorm2d(ndf * 4),
nn.LeakyReLU(0.2, inplace=True),
# state size. (ndf*4) x 8 x 8
nn.Conv2d(ndf * 4, ndf * 8, 4, 2, 1, bias=False),
nn.BatchNorm2d(ndf * 8),
nn.LeakyReLU(0.2, inplace=True),
# state size. (ndf*8) x 4 x 4
nn.Conv2d(ndf * 8, 1, 4, 1, 0, bias=False),
nn.Sigmoid()
)
def forward(self, input):
return self.main(input)
```
Now, as with the generator, we can create the discriminator, apply the
``weights_init`` function, and print the model’s structure.
```python
# Create the Discriminator
netD = Discriminator(ngpu).to(device)
# Handle multi-gpu if desired
if (device.type == 'cuda') and (ngpu > 1):
netD = nn.DataParallel(netD, list(range(ngpu)))
# Apply the weights_init function to randomly initialize all weights
# to mean=0, stdev=0.2.
netD.apply(weights_init)
# Print the model
print(netD)
```
Discriminator(
(main): Sequential(
(0): Conv2d(3, 64, kernel_size=(4, 4), stride=(2, 2), padding=(1, 1), bias=False)
(1): LeakyReLU(negative_slope=0.2, inplace=True)
(2): Conv2d(64, 128, kernel_size=(4, 4), stride=(2, 2), padding=(1, 1), bias=False)
(3): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(4): LeakyReLU(negative_slope=0.2, inplace=True)
(5): Conv2d(128, 256, kernel_size=(4, 4), stride=(2, 2), padding=(1, 1), bias=False)
(6): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(7): LeakyReLU(negative_slope=0.2, inplace=True)
(8): Conv2d(256, 512, kernel_size=(4, 4), stride=(2, 2), padding=(1, 1), bias=False)
(9): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(10): LeakyReLU(negative_slope=0.2, inplace=True)
(11): Conv2d(512, 1, kernel_size=(4, 4), stride=(1, 1), bias=False)
(12): Sigmoid()
)
)
Loss Functions and Optimizers
With D D D and G G G setup, we can specify how they learn
through the loss functions and optimizers. We will use the Binary Cross
Entropy loss
(BCELoss __)
function which is defined in PyTorch as:
\begin{align}\ell(x, y) = L = {l_1,\dots,l_N}^\top, \quad l_n = - \left[ y_n \cdot \log x_n + (1 - y_n) \cdot \log (1 - x_n) \right]\end{align}
Notice how this function provides the calculation of both log components
in the objective function (i.e. l o g ( D ( x ) ) log(D(x)) log(D(x)) and
l o g ( 1 − D ( G ( z ) ) ) log(1-D(G(z))) log(1−D(G(z)))). We can specify what part of the BCE equation to
use with the y y y input. This is accomplished in the training loop
which is coming up soon, but it is important to understand how we can
choose which component we wish to calculate just by changing y y y
(i.e. GT labels).
Next, we define our real label as 1 and the fake label as 0. These
labels will be used when calculating the losses of D D D and
G G G, and this is also the convention used in the original GAN
paper. Finally, we set up two separate optimizers, one for D D D and
one for G G G. As specified in the DCGAN paper, both are Adam
optimizers with learning rate 0.0002 and Beta1 = 0.5. For keeping track
of the generator’s learning progression, we will generate a fixed batch
of latent vectors that are drawn from a Gaussian distribution
(i.e. fixed_noise) . In the training loop, we will periodically input
this fixed_noise into G G G, and over the iterations we will see
images form out of the noise.
# Initialize BCELoss function
criterion = nn.BCELoss()
# Create batch of latent vectors that we will use to visualize
# the progression of the generator
fixed_noise = torch.randn(64, nz, 1, 1, device=device)
# Establish convention for real and fake labels during training
real_label = 1
fake_label = 0
# Setup Adam optimizers for both G and D
optimizerD = optim.Adam(netD.parameters(), lr=lr, betas=(beta1, 0.999))
optimizerG = optim.Adam(netG.parameters(), lr=lr, betas=(beta1, 0.999))
Training
Finally, now that we have all of the parts of the GAN framework defined,
we can train it. Be mindful that training GANs is somewhat of an art
form, as incorrect hyperparameter settings lead to mode collapse with
little explanation of what went wrong. Here, we will closely follow
Algorithm 1 from Goodfellow’s paper, while abiding by some of the best
practices shown in `ganhacks `__.
Namely, we will “construct different mini-batches for real and fake”
images, and also adjust G’s objective function to maximize
$logD(G(z))$. Training is split up into two main parts. Part 1
updates the Discriminator and Part 2 updates the Generator.
**Part 1 - Train the Discriminator**
Recall, the goal of training the discriminator is to maximize the
probability of correctly classifying a given input as real or fake. In
terms of Goodfellow, we wish to “update the discriminator by ascending
its stochastic gradient”. Practically, we want to maximize
$log(D(x)) + log(1-D(G(z)))$. Due to the separate mini-batch
suggestion from ganhacks, we will calculate this in two steps. First, we
will construct a batch of real samples from the training set, forward
pass through $D$, calculate the loss ($log(D(x))$), then
calculate the gradients in a backward pass. Secondly, we will construct
a batch of fake samples with the current generator, forward pass this
batch through $D$, calculate the loss ($log(1-D(G(z)))$),
and *accumulate* the gradients with a backward pass. Now, with the
gradients accumulated from both the all-real and all-fake batches, we
call a step of the Discriminator’s optimizer.
**Part 2 - Train the Generator**
As stated in the original paper, we want to train the Generator by
minimizing $log(1-D(G(z)))$ in an effort to generate better fakes.
As mentioned, this was shown by Goodfellow to not provide sufficient
gradients, especially early in the learning process. As a fix, we
instead wish to maximize $log(D(G(z)))$. In the code we accomplish
this by: classifying the Generator output from Part 1 with the
Discriminator, computing G’s loss *using real labels as GT*, computing
G’s gradients in a backward pass, and finally updating G’s parameters
with an optimizer step. It may seem counter-intuitive to use the real
labels as GT labels for the loss function, but this allows us to use the
$log(x)$ part of the BCELoss (rather than the $log(1-x)$
part) which is exactly what we want.
Finally, we will do some statistic reporting and at the end of each
epoch we will push our fixed_noise batch through the generator to
visually track the progress of G’s training. The training statistics
reported are:
- **Loss_D** - discriminator loss calculated as the sum of losses for
the all real and all fake batches ($log(D(x)) + log(D(G(z)))$).
- **Loss_G** - generator loss calculated as $log(D(G(z)))$
- **D(x)** - the average output (across the batch) of the discriminator
for the all real batch. This should start close to 1 then
theoretically converge to 0.5 when G gets better. Think about why
this is.
- **D(G(z))** - average discriminator outputs for the all fake batch.
The first number is before D is updated and the second number is
after D is updated. These numbers should start near 0 and converge to
0.5 as G gets better. Think about why this is.
**Note:** This step might take a while, depending on how many epochs you
run and if you removed some data from the dataset.
# Training Loop
# Lists to keep track of progress
img_list = []
G_losses = []
D_losses = []
iters = 0
print("Starting Training Loop...")
# For each epoch
for epoch in range(num_epochs):
# For each batch in the dataloader
for i, data in enumerate(dataloader, 0):
############################
# (1) Update D network: maximize log(D(x)) + log(1 - D(G(z)))
###########################
## Train with all-real batch
netD.zero_grad()
# Format batch
real_cpu = data[0].to(device)
b_size = real_cpu.size(0)
label = torch.full((b_size,), real_label, device=device)
# Forward pass real batch through D
output = netD(real_cpu).view(-1)
# Calculate loss on all-real batch
errD_real = criterion(output, label)
# Calculate gradients for D in backward pass
errD_real.backward()
D_x = output.mean().item()
## Train with all-fake batch
# Generate batch of latent vectors
noise = torch.randn(b_size, nz, 1, 1, device=device)
# Generate fake image batch with G
fake = netG(noise)
label.fill_(fake_label)
# Classify all fake batch with D
output = netD(fake.detach()).view(-1)
# Calculate D's loss on the all-fake batch
errD_fake = criterion(output, label)
# Calculate the gradients for this batch
errD_fake.backward()
D_G_z1 = output.mean().item()
# Add the gradients from the all-real and all-fake batches
errD = errD_real + errD_fake
# Update D
optimizerD.step()
############################
# (2) Update G network: maximize log(D(G(z)))
###########################
netG.zero_grad()
label.fill_(real_label) # fake labels are real for generator cost
# Since we just updated D, perform another forward pass of all-fake batch through D
output = netD(fake).view(-1)
# Calculate G's loss based on this output
errG = criterion(output, label)
# Calculate gradients for G
errG.backward()
D_G_z2 = output.mean().item()
# Update G
optimizerG.step()
# Output training stats
if i % 50 == 0:
print('[%d/%d][%d/%d]\tLoss_D: %.4f\tLoss_G: %.4f\tD(x): %.4f\tD(G(z)): %.4f / %.4f'
% (epoch, num_epochs, i, len(dataloader),
errD.item(), errG.item(), D_x, D_G_z1, D_G_z2))
# Save Losses for plotting later
G_losses.append(errG.item())
D_losses.append(errD.item())
# Check how the generator is doing by saving G's output on fixed_noise
if (iters % 500 == 0) or ((epoch == num_epochs-1) and (i == len(dataloader)-1)):
with torch.no_grad():
fake = netG(fixed_noise).detach().cpu()
img_list.append(vutils.make_grid(fake, padding=2, normalize=True))
iters += 1
Starting Training Loop...
/opt/conda/conda-bld/pytorch_1587428091666/work/aten/src/ATen/native/TensorFactories.cpp:361: UserWarning: Deprecation warning: In a future PyTorch release torch.full will no longer return tensors of floating dtype by default. Instead, a bool fill_value will return a tensor of torch.bool dtype, and an integral fill_value will return a tensor of torch.long dtype. Set the optional `dtype` or `out` arguments to suppress this warning.
[0/5][0/1567] Loss_D: 2.1659 Loss_G: 4.8325 D(x): 0.5253 D(G(z)): 0.6765 / 0.0149
[0/5][50/1567] Loss_D: 0.0725 Loss_G: 12.9152 D(x): 0.9461 D(G(z)): 0.0012 / 0.0000
[0/5][100/1567] Loss_D: 0.4784 Loss_G: 9.7521 D(x): 0.9363 D(G(z)): 0.2498 / 0.0002
[0/5][150/1567] Loss_D: 0.9729 Loss_G: 4.8527 D(x): 0.6332 D(G(z)): 0.0643 / 0.0203
[0/5][200/1567] Loss_D: 1.5795 Loss_G: 3.1628 D(x): 0.3932 D(G(z)): 0.0064 / 0.0651
[0/5][250/1567] Loss_D: 0.2961 Loss_G: 3.6781 D(x): 0.9172 D(G(z)): 0.1499 / 0.0599
[0/5][300/1567] Loss_D: 0.8111 Loss_G: 7.3380 D(x): 0.8501 D(G(z)): 0.3902 / 0.0023
[0/5][350/1567] Loss_D: 0.3878 Loss_G: 4.9385 D(x): 0.8003 D(G(z)): 0.0799 / 0.0122
[0/5][400/1567] Loss_D: 0.3620 Loss_G: 5.3288 D(x): 0.8519 D(G(z)): 0.1378 / 0.0121
[0/5][450/1567] Loss_D: 0.2965 Loss_G: 4.2682 D(x): 0.8896 D(G(z)): 0.1294 / 0.0216
[0/5][500/1567] Loss_D: 0.5334 Loss_G: 2.9145 D(x): 0.7111 D(G(z)): 0.0755 / 0.0788
[0/5][550/1567] Loss_D: 0.2940 Loss_G: 4.9582 D(x): 0.9230 D(G(z)): 0.1405 / 0.0191
[0/5][600/1567] Loss_D: 0.3305 Loss_G: 5.1357 D(x): 0.8063 D(G(z)): 0.0506 / 0.0157
[0/5][650/1567] Loss_D: 0.3909 Loss_G: 4.6094 D(x): 0.8939 D(G(z)): 0.1885 / 0.0254
[0/5][700/1567] Loss_D: 0.8562 Loss_G: 7.3044 D(x): 0.5562 D(G(z)): 0.0026 / 0.0026
[0/5][750/1567] Loss_D: 0.5373 Loss_G: 7.7618 D(x): 0.9740 D(G(z)): 0.3496 / 0.0013
[0/5][800/1567] Loss_D: 0.7745 Loss_G: 3.5268 D(x): 0.5872 D(G(z)): 0.0334 / 0.0570
[0/5][850/1567] Loss_D: 0.6165 Loss_G: 5.0756 D(x): 0.9660 D(G(z)): 0.3899 / 0.0138
[0/5][900/1567] Loss_D: 0.3157 Loss_G: 3.4448 D(x): 0.8157 D(G(z)): 0.0528 / 0.0527
[0/5][950/1567] Loss_D: 0.4068 Loss_G: 5.2599 D(x): 0.8900 D(G(z)): 0.2190 / 0.0079
[0/5][1000/1567] Loss_D: 0.4471 Loss_G: 4.9560 D(x): 0.9502 D(G(z)): 0.2832 / 0.0165
[0/5][1050/1567] Loss_D: 0.2992 Loss_G: 2.8899 D(x): 0.8659 D(G(z)): 0.1113 / 0.0815
[0/5][1100/1567] Loss_D: 0.6775 Loss_G: 4.1491 D(x): 0.7165 D(G(z)): 0.2003 / 0.0361
[0/5][1150/1567] Loss_D: 0.4975 Loss_G: 3.7420 D(x): 0.7965 D(G(z)): 0.1760 / 0.0368
[0/5][1200/1567] Loss_D: 0.5209 Loss_G: 3.9092 D(x): 0.7924 D(G(z)): 0.1702 / 0.0407
[0/5][1250/1567] Loss_D: 0.7658 Loss_G: 5.8661 D(x): 0.9198 D(G(z)): 0.4341 / 0.0051
[0/5][1300/1567] Loss_D: 0.6182 Loss_G: 2.3502 D(x): 0.6498 D(G(z)): 0.0416 / 0.1523
[0/5][1350/1567] Loss_D: 1.2259 Loss_G: 5.9051 D(x): 0.8689 D(G(z)): 0.5534 / 0.0116
[0/5][1400/1567] Loss_D: 0.9445 Loss_G: 2.0345 D(x): 0.4958 D(G(z)): 0.0150 / 0.2204
[0/5][1450/1567] Loss_D: 0.5105 Loss_G: 3.1081 D(x): 0.6893 D(G(z)): 0.0470 / 0.0914
[0/5][1500/1567] Loss_D: 0.6236 Loss_G: 2.9756 D(x): 0.7099 D(G(z)): 0.1615 / 0.0785
[0/5][1550/1567] Loss_D: 0.7406 Loss_G: 6.7447 D(x): 0.9440 D(G(z)): 0.4425 / 0.0019
[1/5][0/1567] Loss_D: 1.0153 Loss_G: 8.2250 D(x): 0.9623 D(G(z)): 0.5492 / 0.0008
[1/5][50/1567] Loss_D: 0.5226 Loss_G: 5.2949 D(x): 0.9142 D(G(z)): 0.3052 / 0.0095
[1/5][100/1567] Loss_D: 4.2235 Loss_G: 10.9526 D(x): 0.9861 D(G(z)): 0.9360 / 0.0001
[1/5][150/1567] Loss_D: 0.5008 Loss_G: 4.6801 D(x): 0.9049 D(G(z)): 0.2883 / 0.0173
[1/5][200/1567] Loss_D: 0.4509 Loss_G: 5.3929 D(x): 0.8955 D(G(z)): 0.2458 / 0.0079
[1/5][250/1567] Loss_D: 0.4885 Loss_G: 2.7514 D(x): 0.7086 D(G(z)): 0.0474 / 0.0915
[1/5][300/1567] Loss_D: 0.5356 Loss_G: 2.6585 D(x): 0.7270 D(G(z)): 0.1288 / 0.1004
[1/5][350/1567] Loss_D: 0.8461 Loss_G: 0.9142 D(x): 0.5377 D(G(z)): 0.0342 / 0.4816
[1/5][400/1567] Loss_D: 0.4752 Loss_G: 2.4987 D(x): 0.7422 D(G(z)): 0.0819 / 0.1183
[1/5][450/1567] Loss_D: 0.4779 Loss_G: 3.2457 D(x): 0.7237 D(G(z)): 0.0604 / 0.0788
[1/5][500/1567] Loss_D: 0.3773 Loss_G: 3.4385 D(x): 0.8374 D(G(z)): 0.1532 / 0.0463
[1/5][550/1567] Loss_D: 0.5182 Loss_G: 3.9504 D(x): 0.6710 D(G(z)): 0.0175 / 0.0360
[1/5][600/1567] Loss_D: 0.3414 Loss_G: 3.3448 D(x): 0.7842 D(G(z)): 0.0488 / 0.0585
[1/5][650/1567] Loss_D: 0.4459 Loss_G: 3.4768 D(x): 0.8273 D(G(z)): 0.1939 / 0.0443
[1/5][700/1567] Loss_D: 1.3107 Loss_G: 0.8703 D(x): 0.3642 D(G(z)): 0.0310 / 0.4835
[1/5][750/1567] Loss_D: 0.4869 Loss_G: 3.3692 D(x): 0.8624 D(G(z)): 0.2562 / 0.0461
[1/5][800/1567] Loss_D: 0.5295 Loss_G: 3.6113 D(x): 0.6611 D(G(z)): 0.0210 / 0.0438
[1/5][850/1567] Loss_D: 0.4660 Loss_G: 2.9037 D(x): 0.7429 D(G(z)): 0.1077 / 0.0790
[1/5][900/1567] Loss_D: 0.6548 Loss_G: 3.9572 D(x): 0.8837 D(G(z)): 0.3717 / 0.0275
[1/5][950/1567] Loss_D: 1.7882 Loss_G: 1.0133 D(x): 0.2506 D(G(z)): 0.0065 / 0.4587
[1/5][1000/1567] Loss_D: 0.4816 Loss_G: 3.2230 D(x): 0.7838 D(G(z)): 0.1672 / 0.0636
[1/5][1050/1567] Loss_D: 0.6412 Loss_G: 2.2602 D(x): 0.8479 D(G(z)): 0.3017 / 0.1601
[1/5][1100/1567] Loss_D: 1.3924 Loss_G: 6.9759 D(x): 0.9619 D(G(z)): 0.6918 / 0.0021
[1/5][1150/1567] Loss_D: 0.6000 Loss_G: 2.7910 D(x): 0.6660 D(G(z)): 0.0650 / 0.0931
[1/5][1200/1567] Loss_D: 0.2818 Loss_G: 3.1817 D(x): 0.9106 D(G(z)): 0.1567 / 0.0585
[1/5][1250/1567] Loss_D: 1.2826 Loss_G: 5.8496 D(x): 0.9246 D(G(z)): 0.6465 / 0.0058
[1/5][1300/1567] Loss_D: 0.4234 Loss_G: 3.0132 D(x): 0.8583 D(G(z)): 0.2126 / 0.0735
[1/5][1350/1567] Loss_D: 0.3174 Loss_G: 3.0332 D(x): 0.9053 D(G(z)): 0.1781 / 0.0714
[1/5][1400/1567] Loss_D: 0.7698 Loss_G: 1.4693 D(x): 0.5879 D(G(z)): 0.1052 / 0.3039
[1/5][1450/1567] Loss_D: 0.6258 Loss_G: 0.6535 D(x): 0.6344 D(G(z)): 0.0806 / 0.6098
[1/5][1500/1567] Loss_D: 0.6141 Loss_G: 4.1689 D(x): 0.9008 D(G(z)): 0.3616 / 0.0211
[1/5][1550/1567] Loss_D: 0.4335 Loss_G: 2.7944 D(x): 0.7994 D(G(z)): 0.1607 / 0.0807
[2/5][0/1567] Loss_D: 1.1596 Loss_G: 5.1119 D(x): 0.9492 D(G(z)): 0.6028 / 0.0114
[2/5][50/1567] Loss_D: 0.4913 Loss_G: 3.5129 D(x): 0.7922 D(G(z)): 0.1916 / 0.0443
[2/5][100/1567] Loss_D: 0.5414 Loss_G: 2.7824 D(x): 0.8043 D(G(z)): 0.2332 / 0.0845
[2/5][150/1567] Loss_D: 0.4686 Loss_G: 1.9839 D(x): 0.7234 D(G(z)): 0.0878 / 0.1734
[2/5][200/1567] Loss_D: 0.8191 Loss_G: 3.1437 D(x): 0.7462 D(G(z)): 0.3407 / 0.0605
[2/5][250/1567] Loss_D: 0.5718 Loss_G: 2.8979 D(x): 0.6697 D(G(z)): 0.0961 / 0.0892
[2/5][300/1567] Loss_D: 0.4787 Loss_G: 2.9174 D(x): 0.7941 D(G(z)): 0.1887 / 0.0726
[2/5][350/1567] Loss_D: 0.7060 Loss_G: 0.9351 D(x): 0.5818 D(G(z)): 0.0757 / 0.4493
[2/5][400/1567] Loss_D: 0.4998 Loss_G: 2.4127 D(x): 0.7822 D(G(z)): 0.1959 / 0.1159
[2/5][450/1567] Loss_D: 0.5721 Loss_G: 2.2826 D(x): 0.7614 D(G(z)): 0.2250 / 0.1320
[2/5][500/1567] Loss_D: 1.0770 Loss_G: 3.5311 D(x): 0.8497 D(G(z)): 0.5341 / 0.0453
[2/5][550/1567] Loss_D: 0.9587 Loss_G: 2.0706 D(x): 0.6015 D(G(z)): 0.2855 / 0.1619
[2/5][600/1567] Loss_D: 0.9088 Loss_G: 0.9746 D(x): 0.5216 D(G(z)): 0.1243 / 0.4250
[2/5][650/1567] Loss_D: 0.5082 Loss_G: 3.1175 D(x): 0.8732 D(G(z)): 0.2840 / 0.0571
[2/5][700/1567] Loss_D: 2.3873 Loss_G: 3.1888 D(x): 0.9324 D(G(z)): 0.8345 / 0.0715
[2/5][750/1567] Loss_D: 0.6265 Loss_G: 2.2966 D(x): 0.6355 D(G(z)): 0.1003 / 0.1327
[2/5][800/1567] Loss_D: 0.5677 Loss_G: 1.8367 D(x): 0.7000 D(G(z)): 0.1488 / 0.1967
[2/5][850/1567] Loss_D: 0.5061 Loss_G: 2.1574 D(x): 0.7548 D(G(z)): 0.1600 / 0.1491
[2/5][900/1567] Loss_D: 0.5903 Loss_G: 2.7046 D(x): 0.7996 D(G(z)): 0.2734 / 0.0932
[2/5][950/1567] Loss_D: 1.2207 Loss_G: 0.9973 D(x): 0.3934 D(G(z)): 0.0880 / 0.4310
[2/5][1000/1567] Loss_D: 0.6456 Loss_G: 2.2442 D(x): 0.6200 D(G(z)): 0.0875 / 0.1431
[2/5][1050/1567] Loss_D: 1.0886 Loss_G: 0.7832 D(x): 0.4076 D(G(z)): 0.0440 / 0.5071
[2/5][1100/1567] Loss_D: 0.4480 Loss_G: 2.3673 D(x): 0.8514 D(G(z)): 0.2267 / 0.1308
[2/5][1150/1567] Loss_D: 0.5068 Loss_G: 2.0385 D(x): 0.7478 D(G(z)): 0.1643 / 0.1567
[2/5][1200/1567] Loss_D: 0.5830 Loss_G: 3.5261 D(x): 0.8727 D(G(z)): 0.3260 / 0.0396
[2/5][1250/1567] Loss_D: 0.5778 Loss_G: 1.4222 D(x): 0.6517 D(G(z)): 0.0862 / 0.2932
[2/5][1300/1567] Loss_D: 0.6409 Loss_G: 3.1975 D(x): 0.8424 D(G(z)): 0.3423 / 0.0548
[2/5][1350/1567] Loss_D: 0.5136 Loss_G: 2.9020 D(x): 0.8700 D(G(z)): 0.2737 / 0.0769
[2/5][1400/1567] Loss_D: 0.6284 Loss_G: 3.0575 D(x): 0.7967 D(G(z)): 0.2912 / 0.0604
[2/5][1450/1567] Loss_D: 1.4433 Loss_G: 0.6274 D(x): 0.3156 D(G(z)): 0.0563 / 0.5881
[2/5][1500/1567] Loss_D: 1.1383 Loss_G: 1.5920 D(x): 0.4010 D(G(z)): 0.0324 / 0.2603
[2/5][1550/1567] Loss_D: 1.0544 Loss_G: 3.7734 D(x): 0.9199 D(G(z)): 0.5679 / 0.0348
[3/5][0/1567] Loss_D: 0.6129 Loss_G: 2.3798 D(x): 0.7758 D(G(z)): 0.2622 / 0.1178
[3/5][50/1567] Loss_D: 1.0150 Loss_G: 1.1159 D(x): 0.4285 D(G(z)): 0.0279 / 0.3954
[3/5][100/1567] Loss_D: 0.7960 Loss_G: 2.9441 D(x): 0.6488 D(G(z)): 0.2342 / 0.0811
[3/5][150/1567] Loss_D: 0.7036 Loss_G: 1.0664 D(x): 0.6019 D(G(z)): 0.1089 / 0.3974
[3/5][200/1567] Loss_D: 0.3800 Loss_G: 2.8119 D(x): 0.8059 D(G(z)): 0.1250 / 0.0813
[3/5][250/1567] Loss_D: 0.5637 Loss_G: 1.4856 D(x): 0.6783 D(G(z)): 0.1155 / 0.2617
[3/5][300/1567] Loss_D: 0.5941 Loss_G: 2.8845 D(x): 0.8076 D(G(z)): 0.2765 / 0.0768
[3/5][350/1567] Loss_D: 0.6790 Loss_G: 1.3125 D(x): 0.6115 D(G(z)): 0.1025 / 0.3247
[3/5][400/1567] Loss_D: 1.3248 Loss_G: 4.6937 D(x): 0.9559 D(G(z)): 0.6608 / 0.0153
[3/5][450/1567] Loss_D: 0.6089 Loss_G: 2.5706 D(x): 0.6176 D(G(z)): 0.0490 / 0.1085
[3/5][500/1567] Loss_D: 0.6205 Loss_G: 1.8407 D(x): 0.6519 D(G(z)): 0.1072 / 0.1874
[3/5][550/1567] Loss_D: 1.8101 Loss_G: 0.7733 D(x): 0.2254 D(G(z)): 0.0249 / 0.5126
[3/5][600/1567] Loss_D: 0.4991 Loss_G: 3.2251 D(x): 0.8315 D(G(z)): 0.2383 / 0.0529
[3/5][650/1567] Loss_D: 1.0529 Loss_G: 3.0567 D(x): 0.8478 D(G(z)): 0.5228 / 0.0689
[3/5][700/1567] Loss_D: 0.9430 Loss_G: 1.2749 D(x): 0.5018 D(G(z)): 0.1144 / 0.3263
[3/5][750/1567] Loss_D: 0.6954 Loss_G: 1.8310 D(x): 0.6068 D(G(z)): 0.1138 / 0.2061
[3/5][800/1567] Loss_D: 1.0832 Loss_G: 4.3888 D(x): 0.9131 D(G(z)): 0.5860 / 0.0163
[3/5][850/1567] Loss_D: 0.5538 Loss_G: 3.0146 D(x): 0.8925 D(G(z)): 0.3258 / 0.0663
[3/5][900/1567] Loss_D: 0.5791 Loss_G: 1.9395 D(x): 0.7682 D(G(z)): 0.2320 / 0.1699
[3/5][950/1567] Loss_D: 0.8190 Loss_G: 0.8678 D(x): 0.5284 D(G(z)): 0.0756 / 0.4602
[3/5][1000/1567] Loss_D: 0.9801 Loss_G: 2.8098 D(x): 0.7046 D(G(z)): 0.3957 / 0.0805
[3/5][1050/1567] Loss_D: 0.9347 Loss_G: 2.7753 D(x): 0.7093 D(G(z)): 0.3738 / 0.0884
[3/5][1100/1567] Loss_D: 0.4949 Loss_G: 2.7010 D(x): 0.8318 D(G(z)): 0.2408 / 0.0833
[3/5][1150/1567] Loss_D: 0.4089 Loss_G: 2.5705 D(x): 0.8685 D(G(z)): 0.2168 / 0.0967
[3/5][1200/1567] Loss_D: 0.6158 Loss_G: 2.4766 D(x): 0.7951 D(G(z)): 0.2862 / 0.1067
[3/5][1250/1567] Loss_D: 0.6034 Loss_G: 1.8273 D(x): 0.7414 D(G(z)): 0.2230 / 0.2029
[3/5][1300/1567] Loss_D: 0.4567 Loss_G: 2.1674 D(x): 0.7194 D(G(z)): 0.0813 / 0.1543
[3/5][1350/1567] Loss_D: 0.6020 Loss_G: 1.9374 D(x): 0.6356 D(G(z)): 0.0930 / 0.1885
[3/5][1400/1567] Loss_D: 0.7482 Loss_G: 1.0558 D(x): 0.5842 D(G(z)): 0.1053 / 0.3856
[3/5][1450/1567] Loss_D: 0.6773 Loss_G: 2.5691 D(x): 0.8046 D(G(z)): 0.3226 / 0.0981
[3/5][1500/1567] Loss_D: 0.5394 Loss_G: 2.1660 D(x): 0.7662 D(G(z)): 0.2061 / 0.1421
[3/5][1550/1567] Loss_D: 0.5775 Loss_G: 2.9399 D(x): 0.8530 D(G(z)): 0.3050 / 0.0661
[4/5][0/1567] Loss_D: 0.7733 Loss_G: 2.4629 D(x): 0.7513 D(G(z)): 0.3245 / 0.1104
[4/5][50/1567] Loss_D: 0.8922 Loss_G: 1.1718 D(x): 0.4907 D(G(z)): 0.0476 / 0.3564
[4/5][100/1567] Loss_D: 0.6547 Loss_G: 3.8219 D(x): 0.8484 D(G(z)): 0.3490 / 0.0313
[4/5][150/1567] Loss_D: 0.5065 Loss_G: 2.6143 D(x): 0.8220 D(G(z)): 0.2389 / 0.0935
[4/5][200/1567] Loss_D: 1.3935 Loss_G: 0.4975 D(x): 0.3265 D(G(z)): 0.0518 / 0.6491
[4/5][250/1567] Loss_D: 0.5922 Loss_G: 3.9062 D(x): 0.9168 D(G(z)): 0.3544 / 0.0295
[4/5][300/1567] Loss_D: 0.5837 Loss_G: 1.6458 D(x): 0.6617 D(G(z)): 0.1093 / 0.2295
[4/5][350/1567] Loss_D: 0.5270 Loss_G: 2.0888 D(x): 0.7810 D(G(z)): 0.2142 / 0.1562
[4/5][400/1567] Loss_D: 0.8648 Loss_G: 3.4855 D(x): 0.9003 D(G(z)): 0.4771 / 0.0455
[4/5][450/1567] Loss_D: 0.5237 Loss_G: 2.2375 D(x): 0.8001 D(G(z)): 0.2354 / 0.1309
[4/5][500/1567] Loss_D: 1.0253 Loss_G: 3.3720 D(x): 0.6948 D(G(z)): 0.4164 / 0.0454
[4/5][550/1567] Loss_D: 0.4881 Loss_G: 2.4536 D(x): 0.7618 D(G(z)): 0.1535 / 0.1103
[4/5][600/1567] Loss_D: 0.5000 Loss_G: 2.6980 D(x): 0.7179 D(G(z)): 0.1211 / 0.0951
[4/5][650/1567] Loss_D: 1.0231 Loss_G: 0.8356 D(x): 0.4700 D(G(z)): 0.1121 / 0.4843
[4/5][700/1567] Loss_D: 0.8009 Loss_G: 3.7583 D(x): 0.8391 D(G(z)): 0.4220 / 0.0356
[4/5][750/1567] Loss_D: 0.7859 Loss_G: 1.0725 D(x): 0.5619 D(G(z)): 0.1009 / 0.3875
[4/5][800/1567] Loss_D: 1.5281 Loss_G: 0.8125 D(x): 0.2933 D(G(z)): 0.0452 / 0.4956
[4/5][850/1567] Loss_D: 0.4022 Loss_G: 3.3176 D(x): 0.9354 D(G(z)): 0.2640 / 0.0477
[4/5][900/1567] Loss_D: 0.5842 Loss_G: 1.7742 D(x): 0.6247 D(G(z)): 0.0504 / 0.2097
[4/5][950/1567] Loss_D: 0.5860 Loss_G: 2.0877 D(x): 0.7251 D(G(z)): 0.1948 / 0.1582
[4/5][1000/1567] Loss_D: 0.6735 Loss_G: 4.2387 D(x): 0.9475 D(G(z)): 0.4155 / 0.0213
[4/5][1050/1567] Loss_D: 0.6130 Loss_G: 2.6651 D(x): 0.8604 D(G(z)): 0.3337 / 0.0914
[4/5][1100/1567] Loss_D: 0.3737 Loss_G: 2.5814 D(x): 0.7820 D(G(z)): 0.0964 / 0.1008
[4/5][1150/1567] Loss_D: 0.5412 Loss_G: 3.2119 D(x): 0.8726 D(G(z)): 0.3015 / 0.0533
[4/5][1200/1567] Loss_D: 1.0635 Loss_G: 4.1522 D(x): 0.9437 D(G(z)): 0.5822 / 0.0267
[4/5][1250/1567] Loss_D: 0.6148 Loss_G: 2.7950 D(x): 0.8066 D(G(z)): 0.2866 / 0.0791
[4/5][1300/1567] Loss_D: 0.4972 Loss_G: 2.4557 D(x): 0.8161 D(G(z)): 0.2245 / 0.1036
[4/5][1350/1567] Loss_D: 0.6969 Loss_G: 3.0373 D(x): 0.8472 D(G(z)): 0.3712 / 0.0606
[4/5][1400/1567] Loss_D: 0.9324 Loss_G: 1.1816 D(x): 0.4720 D(G(z)): 0.0429 / 0.3532
[4/5][1450/1567] Loss_D: 0.4000 Loss_G: 2.9336 D(x): 0.8075 D(G(z)): 0.1475 / 0.0696
[4/5][1500/1567] Loss_D: 0.5277 Loss_G: 2.8440 D(x): 0.8159 D(G(z)): 0.2360 / 0.0770
[4/5][1550/1567] Loss_D: 0.5729 Loss_G: 3.9015 D(x): 0.8743 D(G(z)): 0.3178 / 0.0308
Results
Finally, lets check out how we did. Here, we will look at three
different results. First, we will see how D and G’s losses changed
during training. Second, we will visualize G’s output on the fixed_noise
batch for every epoch. And third, we will look at a batch of real data
next to a batch of fake data from G.
Loss versus training iteration
Below is a plot of D & G’s losses versus training iterations.
plt.figure(figsize=(10,5))
plt.title("Generator and Discriminator Loss During Training")
plt.plot(G_losses,label="G")
plt.plot(D_losses,label="D")
plt.xlabel("iterations")
plt.ylabel("Loss")
plt.legend()
plt.show()
Visualization of G’s progression
Remember how we saved the generator’s output on the fixed_noise batch
after every epoch of training. Now, we can visualize the training
progression of G with an animation. Press the play button to start the
animation.
#%%capture
fig = plt.figure(figsize=(8,8))
plt.axis("off")
ims = [[plt.imshow(np.transpose(i,(1,2,0)), animated=True)] for i in img_list]
ani = animation.ArtistAnimation(fig, ims, interval=1000, repeat_delay=1000, blit=True)
HTML(ani.to_jshtml())
# Grab a batch of real images from the dataloader
real_batch = next(iter(dataloader))
# Plot the real images
plt.figure(figsize=(15,15))
plt.subplot(1,2,1)
plt.axis("off")
plt.title("Real Images")
plt.imshow(np.transpose(vutils.make_grid(real_batch[0].to(device)[:64], padding=5, normalize=True).cpu(),(1,2,0)))
# Plot the fake images from the last epoch
plt.subplot(1,2,2)
plt.axis("off")
plt.title("Fake Images")
plt.imshow(np.transpose(img_list[-1],(1,2,0)))
plt.show()
