/
train.py
87 lines (68 loc) · 2.82 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
import os
import numpy as np
from sklearn.model_selection import train_test_split
import torch
import torch.nn as nn
import torch.nn.functional as F
import torch.optim as optim
from torch.utils.data import DataLoader
import torchvision
from torchvision import transforms
from datasets import load_datasets
from models import TripletResNet
from losses import TripletLoss, TripletAngularLoss
from params import args
def calculate_loss(model, data_loader, optimizer=None):
if optimizer is None:
model.eval()
training = False
else:
model.train()
optimizer.zero_grad()
training = True
epoch_loss = 0
for i, (anchors, positives, negatives, _) in enumerate(data_loader):
anchors = anchors.to(device)
positives = positives.to(device)
negatives = negatives.to(device)
anc_metric = model(anchors)
pos_metric = model(positives)
neg_metric = model(negatives)
loss = criterion(anc_metric, pos_metric, neg_metric)
if training:
loss.backward()
optimizer.step()
optimizer.zero_grad()
epoch_loss += loss.item()
return epoch_loss / len(data_loader)
if not os.path.exists(args.weight_dir):
os.mkdir(args.weight_dir)
if __name__ == '__main__':
transform = transforms.Compose([transforms.Resize(args.img_size),
transforms.CenterCrop(args.img_size),
transforms.ToTensor()])
train_dataset, valid_dataset, test_dataset = \
load_datasets(args.train_json, args.test_json, transform)
train_loader = DataLoader(train_dataset, batch_size=args.batch_size)
valid_loader = DataLoader(valid_dataset, batch_size=args.batch_size)
test_loader = DataLoader(test_dataset, batch_size=args.batch_size)
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
model = TripletResNet(args.output_dim)
model = model.to(device)
criterion = TripletAngularLoss()
# criterion = TripletLoss()
optimizer = torch.optim.SGD(model.parameters(),
lr=args.lr,
momentum=args.momentum,
weight_decay=args.weight_decay)
best_loss = 1e8
for epoch in range(1, args.n_epochs+1):
train_loss = calculate_loss(model, train_loader, optimizer=optimizer)
valid_loss = calculate_loss(model, valid_loader)
print(f'EPOCH: [{epoch}/{args.n_epochs}], TRAIN_LOSS: {train_loss:.3f}, VALID_LOSS: {valid_loss:.3f}')
# if valid_loss < best_loss:
weights_name = args.weight_dir + args.experiment_name + f'_{valid_loss:.5f}' + '.pth'
best_loss = valid_loss
best_param = model.state_dict()
torch.save(best_param, weights_name)
print(f'save wieghts to {weights_name}')