-
Notifications
You must be signed in to change notification settings - Fork 6
/
train.py
192 lines (153 loc) · 6.56 KB
/
train.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
#####################################################
##
## IMPORTING MODULES AND LIBRARIES
##
#####################################################
import torch
import torch.optim as optim
import torch.nn.functional as F
from torch.utils.data import DataLoader
from torchvision import transforms, utils
import time
import numpy as np
import os
import pickle
from PIL import Image
import matplotlib.pyplot as plt
plt.ion()
from model.generator import *
from model.discriminator import NLayerDiscriminator as Discriminator
from utils.dataloader import YouTubePose
from utils.loss_functions import lossIdentity, lossShape
#####################################################
##
## IMPORTANT PARAMETERS
##
#####################################################
dataset_dir = './Dataset/'
checkpoint_path = "./model_checkpoints/"
batch_size = 4
epochs = 10
device = 'cuda' if torch.cuda.is_available() else 'cpu'
learning_rate_generator = 3e-4
learning_rate_discriminator = 0.1
alpha = 8
beta = 0.001
#####################################################
##
## DATALOADER
##
#####################################################
with open('./dataset_lists/train_datapoint_triplets.pkl', 'rb') as f:
datapoint_pairs = pickle.load(f)
with open('./dataset_lists/train_shapeLoss_pairs.pkl', 'rb') as f:
shapeLoss_datapoint_pairs = pickle.load(f)
transform = transforms.Compose([
transforms.Resize((256, 256)),
transforms.ToTensor(),
transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))
])
train_dataset = YouTubePose(datapoint_pairs, shapeLoss_datapoint_pairs, dataset_dir, transform)
train_dataloader = DataLoader(train_dataset, batch_size=batch_size, shuffle=True,
num_workers=0)
dataset_sizes = [len(train_dataset)]
#####################################################
##
## MODEL PREPARATION
##
#####################################################
generator = Generator(ResidualBlock)
discriminator = Discriminator(3)
generator = generator.to(device)
discriminator = discriminator.to(device)
optimizer_gen = optim.Adam(generator.parameters(), lr = learning_rate_generator)
optimizer_disc = optim.SGD(discriminator.parameters(), lr = learning_rate_discriminator, momentum=0.9)
#####################################################
##
## TRAINING SCRIPT
##
#####################################################
def save_checkpoint(state, dirpath, epoch):
filename = 'checkpoint-{}.ckpt'.format(epoch)
checkpoint_path = os.path.join(dirpath, filename)
torch.save(state, checkpoint_path)
print('--- checkpoint saved to ' + str(checkpoint_path) + ' ---')
def train_model(gen, disc, loss_i, loss_s, optimizer_gen, optimizer_disc, alpha = 1, beta = 1, num_epochs = 10):
for epoch in range(num_epochs):
print("Epoch {}/{}".format(epoch, num_epochs - 1))
print('-'*10)
dataloader = train_dataloader
gen.train()
disc.train()
since = time.time()
running_loss_iden = 0.0
running_loss_s1 = 0.0
running_loss_s2a = 0.0
running_loss_s2b = 0.0
running_loss = 0.0
for i_batch, sample_batched in enumerate(dataloader):
x_gen, y, x_dis = sample_batched['x_gen'], sample_batched['y'], sample_batched['x_dis']
iden_1, iden_2 = sample_batched['iden_1'], sample_batched['iden_2']
x_gen = x_gen.to(device)
y = y.to(device)
x_dis = x_dis.to(device)
iden_1 = iden_1.to(device)
iden_2 = iden_2.to(device)
optimizer_gen.zero_grad()
optimizer_disc.zero_grad()
with torch.set_grad_enabled(True):
x_generated = gen(x_gen, y)
fake_op, fake_pooled_op = disc(x_gen, x_generated)
real_op, real_pooled_op = disc(x_gen, x_dis)
loss_identity_gen = -loss_i(real_pooled_op, fake_pooled_op)
loss_identity_gen.backward(retain_graph=True)
optimizer_gen.step()
optimizer_disc.zero_grad()
loss_identity_disc = loss_i(real_op, fake_op)
loss_identity_disc.backward(retain_graph=True)
optimizer_disc.step()
optimizer_gen.zero_grad()
optimizer_disc.zero_grad()
x_ls2a = gen(y, x_generated)
x_ls2b = gen(x_generated, y)
loss_s2a = loss_s(y, x_ls2a)
loss_s2b = loss_s(x_generated, x_ls2b)
loss_s2 = loss_s2a + loss_s2b
loss_s2.backward()
optimizer_gen.step()
optimizer_gen.zero_grad()
optimizer_disc.zero_grad()
x_ls1 = generator(iden_1, iden_2)
loss_s1 = loss_s(iden_2, x_ls1)
loss_s1.backward()
optimizer_gen.step()
running_loss_iden += loss_identity_disc.item() * x_gen.size(0)
running_loss_s1 += loss_s1.item() * x_gen.size(0)
running_loss_s2a += loss_s2a.item() * x_gen.size(0)
running_loss_s2b += loss_s2b.item() * x_gen.size(0)
running_loss = running_loss_iden + beta * (running_loss_s1 + alpha * (running_loss_s2a + running_loss_s2b))
epoch_loss_iden = running_loss_iden / dataset_sizes[0]
epoch_loss_s1 = running_loss_s1 / dataset_sizes[0]
epoch_loss_s2a = running_loss_s2a / dataset_sizes[0]
epoch_loss_s2b = running_loss_s2a / dataset_sizes[0]
epoch_loss = running_loss / dataset_sizes[0]
print('Identity Loss: {:.4f} Loss Shape1: {:.4f} Loss Shape2a: {:.4f} Loss Shape2b: {:.4f}'.format(epoch_loss_iden, epoch_loss_s1,
epoch_loss_s2a, epoch_loss_s2b))
print('Epoch Loss: {:.4f}'.format(epoch_loss))
save_checkpoint({
'epoch': epoch + 1,
'gen_state_dict': gen.state_dict(),
'disc_state_dict': disc.state_dict(),
'gen_opt': optimizer_gen.state_dict(),
'disc_opt': optimizer_disc.state_dict()
}, checkpoint_path, epoch + 1)
print('Time taken by epoch: {: .0f}m {:0f}s'.format((time.time() - since) // 60, (time.time() - since) % 60))
print()
since = time.time()
return gen, disc
#####################################################
##
## MODEL TRAINING
##
#####################################################
generator, discriminator = train_model(generator, discriminator, lossIdentity, lossShape, optimizer_gen, optimizer_disc, alpha=alpha, beta=beta, num_epochs=epochs)