-
Notifications
You must be signed in to change notification settings - Fork 6
/
Copy pathmain.py
executable file
·180 lines (154 loc) · 6.17 KB
/
main.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
import argparse
from dataprocess.dataset import DogCat
from models import ShuffleNet2
from models import MobileNet2
from models import MobileNetV3_Large, MobileNetV3_Small, BetterShuffleNet, SENet18
import torch as t
from torch.utils import data
import torch.nn as nn
import copy
from efficientnet_pytorch import EfficientNet
ap = argparse.ArgumentParser()
ap.add_argument("-gpu", "--use_gpu", type=int, default=1,
help="use gpu or not")
ap.add_argument("-bs", "--batchsize", type=int, default=32,
help="the batch size of input")
ap.add_argument("-t", "--train", type=int, default=1,
help="choose training or valdating")
ap.add_argument("-pre", "--pretrained", type=str, default="None",
help="select a pretrained model")
ap.add_argument("-e", "--epochs", type=int, default=5,
help="epochs of training")
ap.add_argument("-path", "--datapath", type=str, default="./dogvscat/train",
help="path of training dataset")
# the arguments for model
ap.add_argument("-m", "--model", type=str, default="ShuffleNet2",
help="the type of model")
ap.add_argument("-c", "--classes", type=int, default=2,
help="the number of classes of dataset")
ap.add_argument("-s", "--inputsize", type=int, default=224,
help="the size of image")
ap.add_argument("-nt", "--nettype", type=int, default=1,
help="type of network")
def train_model(model, dataloaders, loss_fn, optimizer, num_epochs=5):
best_model_wts = copy.deepcopy(model.state_dict())
best_acc = 0.
val_acc_history = []
for epoch in range(num_epochs):
for phase in ["train", "val"]:
running_loss = 0.
running_corrects = 0.
if phase == "train":
model.train()
else:
model.eval()
for inputs, labels in dataloaders[phase]:
inputs, labels = inputs.to(device), labels.to(device)
with t.autograd.set_grad_enabled(phase=="train"):
outputs = model(inputs) # bsize * 2 , because it is a binary classification
loss = loss_fn(outputs, labels)
preds = outputs.argmax(dim=1)
if phase == "train":
optimizer.zero_grad()
loss.backward()
optimizer.step()
running_loss += loss.item() * inputs.size(0)
running_corrects += t.sum(preds.view(-1) == labels.view(-1)).item()
epoch_loss = running_loss / len(dataloaders[phase].dataset)
epoch_acc = running_corrects / len(dataloaders[phase].dataset)
print("Phase {} loss: {}, acc: {}".format(phase, epoch_loss, epoch_acc))
if phase == "val" and epoch_acc > best_acc:
best_acc = epoch_acc
best_model_wts = copy.deepcopy(model.state_dict())
if phase == "val":
val_acc_history.append(epoch_acc)
model.load_state_dict(best_model_wts)
return model, val_acc_history
def test_model(model, dataloader, loss_fn):
import time
model.eval()
running_loss = 0.
running_corrects = 0.
records = []
total_len = 0
for inputs, labels in dataloader:
img_len = len(inputs)
inputs, labels = inputs.to(device), labels.to(device)
#
start = time.time()
outputs = model(inputs)
#
end = time.time()
fps = img_len/(end-start)
records.append(fps)
loss = loss_fn(outputs, labels)
preds = outputs.argmax(dim=1)
running_loss += loss.item() * inputs.size(0)
running_corrects += t.sum(preds.view(-1) == labels.view(-1)).item()
epoch_loss = running_loss / len(dataloader.dataset)
epoch_acc = running_corrects / len(dataloader.dataset)
print("On val dataset loss: {}, acc: {}".format(epoch_loss, epoch_acc))
import numpy as np
print("{} FPS".format(np.mean(records)))
if __name__ == '__main__':
args = vars(ap.parse_args())
path = args["datapath"]
train_sign = args["train"]
epochs = args["epochs"]
batchsize = args["batchsize"]
dataloader = {}
if train_sign:
train_dataset = DogCat("./dogvscat/train", train=True)
train_loader = data.DataLoader(train_dataset,
batch_size = batchsize,
shuffle=True)
dataloader["train"] = train_loader
val_dataset = DogCat("./dogvscat/train", train=False, test=False)
val_loader = data.DataLoader(val_dataset,
batch_size = batchsize,
shuffle=True)
dataloader["val"] = val_loader
use_gpu = args["use_gpu"]
if use_gpu:
if t.cuda.is_available():
device = t.device("cuda")
else:
print("You don't have gpu")
else:
device = t.device("cpu")
# device = t.device("cuda" if t.cuda.is_available() else "cpu")
model_path = args["pretrained"]
num_classes = args["classes"]
input_size = args["inputsize"]
net_type = args["nettype"]
model_type = args["model"]
if model_type == "ShuffleNet2":
model = ShuffleNet2(num_classes, input_size, net_type)
elif model_type == "MobileNet2":
model = MobileNet2(num_classes, input_size, net_type)
elif model_type == "MobileNetV3_Large":
model = MobileNetV3_Large(num_classes)
elif model_type == "MobileNetV3_Small":
model = MobileNetV3_Small(num_classes)
elif "efficientnet" in model_type.lower():
model = EfficientNet.from_name(model_type)
elif model_type == "BetterShuffleNet":
model = BetterShuffleNet(num_classes)
elif model_type == "SENet18":
model = SENet18(num_classes)
else:
print("We don't implement the model, please choose ShuffleNet2 or MobileNet2")
if model_path != "None":
model.load_state_dict(t.load(model_path))
model.to(device)
loss_fn = nn.CrossEntropyLoss()
optimizer = t.optim.SGD(model.parameters(), lr=0.1, momentum=0.9)
if train_sign:
model, val_logs = train_model(model, dataloader, loss_fn, optimizer, epochs)
# store the model
import time
pkl_path = "./save/" + model_type + str(int(time.time()))+'.pkl'
t.save(model.state_dict(), pkl_path)
print("Model saved to", pkl_path)
else:
test_model(model, dataloader['val'], loss_fn)