-
Notifications
You must be signed in to change notification settings - Fork 2
/
train.py
238 lines (200 loc) · 10.7 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
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
import os
import numpy as np
import torch
import torch.optim as optim
import torch.nn.functional as F
from torch.utils.data import DataLoader
from torchvision import transforms
from torchsampler import ImbalancedDatasetSampler
from time import time
from argparse import ArgumentParser
from dataset import ClassifierDataset
from model.resnet34 import ResNet34
#from model.fastai_resnet import resnet34
#from model.coord_conv_resnet import resnet34
from model.cc_resnet import resnet34
from utils.loss import *
from utils.checkpoints import *
from utils.metric import auroc_score
device = torch.device('cuda:0' if torch.cuda.is_available() else 'cpu')
# conditions = ['gender', 'HCC18', 'HCC22', 'HCC40', 'HCC48', 'HCC59', 'HCC85', 'HCC96', 'HCC108', 'HCC111', 'HCC138', 'age MSE', 'raf MSE', 'BMI MSE','A1C MSE']
conditions = ['gender', 'HCC18', 'HCC22', 'HCC85', 'HCC96', 'HCC108', 'HCC111', 'age MSE', 'raf MSE']
num_classes = len(conditions) - 2
def arg_parse():
parser = ArgumentParser()
parser.add_argument('--data_dir', default='data', help='Directory with data')
parser.add_argument('--checkpoint_dir', default='checkpoints', help='Checkpoint output directory')
parser.add_argument('--pretrain', default=None, help='Path to pretrained weights')
parser.add_argument('--size', default=256, type=int, help="Size of CT to generate")
parser.add_argument('--age_norm', default=100.0, type=float, help="Normalization of age")
parser.add_argument('--raf_norm', default=10.0, type=float, help="Normalization of RAF")
parser.add_argument('--lr', default=1e-3, type=float, help="Learning rate")
parser.add_argument('--epochs', default=1000, type=int, help="Number of epochs")
parser.add_argument('--train_batch_size', default=64, type=int, help="Training batch size")
parser.add_argument('--test_batch_size', default=1, type=int, help="Testing batch size")
parser.add_argument('--num_workers', default=32, type=int, help="Number of workers used")
parser.add_argument('--decay_start_epoch', default=15, type=int, help="Epoch to start learning rate decay")
args = parser.parse_args()
return args
def calculate_accuracy(y_pred, y_label, threshold=0.5):
pred_conditions = y_pred[:,:-2]
label_conditions = y_label[:,:-2]
pred_age = y_pred[:,-2]
label_age = y_label[:,-2]
pred_raf = y_pred[:,-1]
label_raf = y_label[:,-1]
loss = nn.MSELoss()
batch, _ = label_conditions.shape
correct = torch.zeros(num_classes + 2)
one = torch.ones(1).to(device)
zero = torch.zeros(1).to(device)
for i in range(num_classes):
pred_labels = torch.where(pred_conditions[:, i] > threshold, one, zero)
correct[i] = (pred_labels == label_conditions[:, i]).sum()
correct[-2] = loss(pred_age, label_age)
correct[-1] = loss(pred_raf, label_raf)
return correct
# def print_accuracy(accuracy):
# for i, acc in enumerate(accuracy):
# print('{}: {:.4f}%'.format(conditions[i], acc*100))
# def print_accuracy(train_accuracy, test_accuracy):
# for i, (train_acc, test_acc) in enumerate(zip(train_accuracy, test_accuracy)):
# if i > 10:
# print('{}\t: {:.4f}\t\t\t{:.4f}'.format(conditions[i], train_acc, test_acc))
# else:
# print('{}\t: {:.4f}%\t\t\t{:.4f}%'.format(conditions[i], train_acc*100, test_acc*100))
def print_accuracy(train_accuracy, test_accuracy, train_auroc, test_auroc):
print('\t\tTrain accuracy:\tTest accuracy:\tTrain AUROC:\tTest AUROC:')
for i, (train_acc, test_acc, train_au, test_au) in enumerate(zip(train_accuracy, test_accuracy, train_auroc, test_auroc)):
print('{}\t:\t{:.4f}%\t{:.4f}%\t{:.4f}\t\t{:.4f}'.format(conditions[i], train_acc*100, test_acc*100, train_au, test_au))
print('{}\t:\t{:.4f}\t\t{:.4f}'.format(conditions[-2], train_accuracy[-2], test_accuracy[-2]))
print('{}\t:\t{:.4f}\t\t{:.4f}'.format(conditions[-1], train_accuracy[-1], test_accuracy[-1]))
def print_auroc(train_auroc, test_auroc):
for i in range(num_classes):
print('{}\t: {:.4f}\t\t{:.4f}'.format(conditions[i], train_auroc[i], test_auroc[i]))
def train_epoch(model, dataloader, optimizer):
tot_loss = 0
accuracy = torch.zeros(num_classes+2, dtype=torch.float32)
tot_score = 0
class_score = torch.zeros(num_classes)
for i, (img, labels) in enumerate(dataloader):
img = img.to(device)
labels = labels.to(device)
img = img.repeat(1, 3, 1, 1)
prediction = torch.sigmoid(model(img))
s, cs = auroc_score(prediction[:, :-2], labels[:, :-2])
tot_score += s
class_score += cs
loss = multilabel_regression_loss(prediction, labels)
tot_loss += loss.item()
accuracy += calculate_accuracy(prediction, labels)
optimizer.zero_grad()
loss.backward()
optimizer.step()
del img, labels, prediction
accuracy /= len(dataloader.dataset)
return tot_loss/(i+1), accuracy, 100 * tot_score/(i + 1), 100 * class_score/(i + 1)
def train(model, train_dataloader, test_dataloader, optimizer, scheduler, epochs):
best_loss = np.inf
for epoch in range(epochs):
model.train()
start_train_time = time()
train_loss, train_accuracy, train_auroc, train_class_auroc = train_epoch(model, train_dataloader, optimizer)
scheduler.step(train_loss)
train_time = time() - start_train_time
start_test_time = time()
test_loss, test_accuracy, test_auroc, test_class_auroc = test(model, test_dataloader)
test_time = time() - start_test_time
print('Epoch: [{}/{}], Train loss: {:.4f}, Test loss: {:.4f}, Train score: {:.2f}, Test score: {:.2f} Train time: {:.2f}s, Test time: {:.2f}s'.format(epoch+1, epochs, train_loss, test_loss, train_auroc, test_auroc, train_time, test_time))
#print('\t\tTrain AUROC:\tTest AUROC:')
#print_auroc(train_class_auroc, test_class_auroc)
#print('\t\t Train accuracy:\tTest accuracy:')
print_accuracy(train_accuracy, test_accuracy, train_class_auroc, test_class_auroc)
train_state = {'epoch' : epoch + 1,
'state_dict' : model.state_dict(),
'optim_dict' : optimizer.state_dict()}
model_state = {'state_dict' : model.state_dict()}
is_best = test_loss < best_loss
if is_best:
best_loss = test_loss
save_checkpoint(train_state, model_state, is_best, args.checkpoint_dir)
def test(model, test_dataloader):
tot_loss = 0
accuracy = torch.zeros(num_classes+2, dtype=torch.float32)
tot_score = 0
class_score = torch.zeros(num_classes)
model.eval()
with torch.no_grad():
for i, (img, labels) in enumerate(test_dataloader):
img = img.to(device)
labels = labels.to(device)
img = img.repeat(1, 3, 1, 1)
prediction = torch.sigmoid(model(img))
s, cs = auroc_score(prediction[:, :-2], labels[:, :-2])
tot_score += s
class_score += cs
loss = multilabel_regression_loss(prediction, labels)
tot_loss += loss.item()
accuracy += calculate_accuracy(prediction, labels)
del img, labels, prediction
accuracy /= len(test_dataloader.dataset)
return tot_loss/(i+1), accuracy, 100 * tot_score/(i + 1), 100 * class_score/(i + 1)
args = arg_parse()
print('Parameters:\n\
data directory: {}\n\
checkpoint directory: {}\n\
size: {}\n\
lr: {}\n\
epochs: {}\n\
train batch size: {}\n\
number of workers: {}\n\
epoch to start decay: {}\n'.format(args.data_dir,
args.checkpoint_dir, args.size, args.lr, args.epochs,
args.train_batch_size, args.num_workers,
args.decay_start_epoch))
print('Loading data')
transform = transforms.Compose([transforms.RandomHorizontalFlip(p=0.2),
transforms.RandomApply([transforms.RandomAffine(10),
transforms.RandomResizedCrop(256, scale=(1.0, 1.1), ratio=(0.75, 1.33)),
transforms.ColorJitter(brightness=0.2, contrast=0.2)], p=0.75),
transforms.RandomPerspective(distortion_scale=0.2, p=0.75),
transforms.ToTensor(), transforms.Normalize((0.55001191,), (0.18854326,))])
#transform = transforms.Compose([transforms.ToTensor(), transforms.Normalize((0.55001191,), (0.18854326,))])
# Train data: mean = 0.55001191, std = 0.18854326
train_dataset = ClassifierDataset(args.data_dir, conditions[1:-2], transforms=transform, size=args.size, train=True, age_norm=args.age_norm, raf_norm=args.raf_norm)
# Test data: mean = 0.42215754, std = 0.18558778
transform = transforms.Compose([transforms.ToTensor(), transforms.Normalize((0.55001191,), (0.18854326,))])
test_dataset = ClassifierDataset(args.data_dir, conditions[1:-2], transforms=transform, size=args.size, train=False, age_norm=args.age_norm, raf_norm=args.raf_norm)
train_dataloader = DataLoader(dataset=train_dataset,batch_size=args.train_batch_size, shuffle=True, num_workers=args.num_workers)
#print('imbalanced sampler')
#train_dataloader = DataLoader(dataset=train_dataset,sampler=ImbalancedDatasetSampler(train_dataset),batch_size=args.train_batch_size, shuffle=True, num_workers=args.num_workers)
test_dataloader = DataLoader(dataset=test_dataset, batch_size=args.test_batch_size, shuffle=False, num_workers=args.num_workers)
print('Initializing')
if args.pretrain:
print('resnet chexpert')
model = ResNet34(2 * 14, pretrained=True).to(device)
#print('coord conv chexpert')
#model = resnet34(pretrained=False, num_classes=2*14).to(device)
load_checkpoint(args.pretrain, model)
model.resnet34.fc = nn.Linear(model.resnet34.fc.in_features, num_classes+2)
#head = list(model.head.layers.children())
#in_features = 512
#head = head[:-1]
#model.head.layers = nn.Sequential(*head)
#model.head.layers = nn.Sequential(model.head.layers, nn.Linear(in_features, num_classes+2))
else:
#print('pretrained torch resnet34')
#model = ResNet34(3*num_classes+2, pretrained=True)
print('CoordConv resnet34 implementation')
model = resnet34(pretrained=False, num_classes=3*num_classes+2).to(device)
if (torch.cuda.device_count() > 1):
device_ids = list(range(torch.cuda.device_count()))
print("GPU devices being used: ", device_ids)
model = nn.DataParallel(model, device_ids=device_ids)
model = model.to(device)
#optimizer = optim.SGD(model.parameters(), lr=args.lr, momentum=0.9)
optimizer = optim.AdamW(model.parameters(), lr=args.lr)
#scheduler = optim.lr_scheduler.StepLR(optimizer, args.decay_start_epoch, gamma=0.1)
scheduler = optim.lr_scheduler.ReduceLROnPlateau(optimizer)
print('Training')
train(model, train_dataloader, test_dataloader, optimizer, scheduler, args.epochs)