-
Notifications
You must be signed in to change notification settings - Fork 0
/
utils.py
272 lines (225 loc) · 8.5 KB
/
utils.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
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
import os
import shutil
import torch
from torch.utils.data import Dataset
from torch import nn
from torch.utils.data import DataLoader
from torchvision import transforms
import matplotlib.pyplot as plt
from thop import profile, clever_format
import platform
from mydataset import ImageDataset
from models.pvtv2 import pvt_v2_b0, pvt_v2_b1, pvt_v2_b2, pvt_v2_b3, pvt_v2_b4, pvt_v2_b5, pvt_v2_b2_li
dict = {
'dijia':0,
'jieke':1,
'saiwen':2,
'tailuo':3
}
def remove_oldest_item(path:str, number:int=10):
"""当文件数量达到numbers时,删除修改时间最早的文件或文件夹"""
# 获取目录下所有文件和文件夹的路径
items = [os.path.join(path, item) for item in os.listdir(path)]
# 过滤出文件和文件夹,并分别按照修改时间排序
files = sorted([(f, os.path.getmtime(f)) for f in items if os.path.isfile(f)], key=lambda x: x[1])
folders = sorted([(d, os.path.getmtime(d)) for d in items if os.path.isdir(d)], key=lambda x: x[1])
# 合并文件和文件夹列表,并按修改时间排序
items_with_mtime = files + folders
# 如果文件数量未达到number,则无需删除
if len(items_with_mtime) < number:
return
if items_with_mtime:
# 删除修改时间最早的文件或文件夹
oldest_item_path, mtime = items_with_mtime[0]
try:
if os.path.isfile(oldest_item_path):
os.remove(oldest_item_path)
print(f"delete a file: {oldest_item_path}")
elif os.path.isdir(oldest_item_path):
shutil.rmtree(oldest_item_path)
print(f"delete a directory: {oldest_item_path}")
except Exception as e:
print(f'fail to delete "{oldest_item_path}": {e}')
def read_aoteman_data(root:str, train:bool=True):
"""
Read aoteman dataset. If True, return train's image paths and labels. Else, return eval's.
"""
test_root = os.path.join(root, "predict_demo.jpg")
if not os.path.exists(test_root):
raise Exception(f"error root: '{root}'")
if train:
root = os.path.join(root, "train")
else:
root = os.path.join(root, "eval")
categories = os.listdir(root)
images = []
labels = []
for category in categories:
branch = os.path.join(root, category)
label = dict[category]
for image_name in os.listdir(branch):
image_path = os.path.join(branch, image_name)
images.append(image_path)
labels.append(label)
return images, labels
def train_one_epoch_single_cuda(model, optimizer, data_loader, loss_fn, device):
"""
Train the model. Return the average loss and accuracy.
Parameters:
model(torch.nn.Module): model to train.
data_loader(torch.utils.data.Dataloader): data loader.
loss_fn: loss function.
device: device to train.
"""
model.train()
optimizer.zero_grad()
batch_len = len(data_loader)
data_len = 0.
total_loss = 0.
rights = 0.
for images, labels in data_loader:
images = images.to(device)
labels = labels.to(device)
data_len += len(labels)
predictions = model(images)
loss = loss_fn(predictions, labels)
loss.backward()
optimizer.step()
optimizer.zero_grad()
total_loss += loss.detach().item()
rights += (predictions.argmax(dim=1) == labels).sum().item()
return total_loss / batch_len, rights / data_len
def eval_single_cuda(model, data_loader, loss_fn, device):
"""
Eval the model. Return the average loss and accuracy.
Parameters:
model(torch.nn.Module): model to eval.
data_loader(torch.utils.data.Dataloader): data loader.
loss_fn: loss function.
device: device to eval.
"""
model.eval()
batch_len = len(data_loader)
total_loss = 0.
rights = 0.
data_len = 0.
with torch.no_grad():
for images, labels in data_loader:
images = images.to(device)
labels = labels.to(device)
data_len += len(labels)
predictions = model(images)
loss = loss_fn(predictions, labels)
total_loss += loss.detach().item()
rights += (predictions.argmax(dim=1) == labels).sum().item()
return total_loss / batch_len, rights / data_len
def show_L1Norm_vgg(model, save_path=None):
"""
show L1_Norm in vgg11 model with graph.
"""
index = 1
plt.figure(figsize=(14, 6))
for m in model.modules():
if isinstance(m, nn.Conv2d):
# 计算当前卷积层每一个卷积核的L1范数
weight_copy = m.weight.data.abs().clone()
weight_copy = weight_copy.cpu().numpy()
L1_norm = weight_copy.sum(axis=1).sum(axis=1).sum(axis=1)
sort = L1_norm.argsort()
L1_norm = L1_norm[sort]
plt.subplot(2, 4, index)
index += 1
min_10 = int(len(L1_norm) / 10)
min_20 = int(len(L1_norm) / 5)
min_10 = torch.ones_like(L1_norm) * L1_norm[int(min_10)]
min_20 = torch.ones_like(L1_norm) * L1_norm[int(min_20)]
plt.plot(L1_norm, label="L1_norm")
plt.plot(min_10, label="min 10%")
plt.plot(min_20, label="min 20%")
plt.xlabel("kernel index")
plt.ylabel("L1_norm")
plt.legend()
plt.tight_layout()
if save_path is not None:
plt.savefig(save_path)
plt.show()
def get_params_and_FLOPs(model, input_size):
"""
get params and FLOPs of model.
"""
input_tensor = torch.rand(input_size)
FLOPs, params = profile(model, inputs=(input_tensor,))
FLOPs, params = clever_format([FLOPs, params], "%.3f")
return params, FLOPs
def get_mean_and_std(dataloader):
"""
get mean and std of dataloader.
"""
data, _ = next(iter(dataloader))
# get the everage of the dataloader
data_mean = torch.mean(data, dim=[0, 2, 3])
means = torch.ones_like(data, dtype=torch.float32)
means[:, 0, :, :] = data_mean[0]
means[:, 1, :, :] = data_mean[1]
means[:, 2, :, :] = data_mean[2]
# get the std of the dataset
data_std = torch.std(data, dim=[0, 2, 3])
return data_mean, data_std
def get_model(model_name:str, pretrained:bool=True, **kwargs):
"""
get model by model_name.
"""
names = ['pvt_v2_b0', 'pvt_v2_b1', 'pvt_v2_b2', 'pvt_v2_b3', 'pvt_v2_b4', 'pvt_v2_b5', 'pvt_v2_b2_li']
if model_name == names[0]:
return pvt_v2_b0(pretrained=pretrained, **kwargs)
elif model_name == names[1]:
return pvt_v2_b1(pretrained=pretrained, **kwargs)
elif model_name == names[2]:
return pvt_v2_b2(pretrained=pretrained, **kwargs)
elif model_name == names[3]:
return pvt_v2_b3(pretrained=pretrained, **kwargs)
elif model_name == names[4]:
return pvt_v2_b4(pretrained=pretrained, **kwargs)
elif model_name == names[5]:
return pvt_v2_b5(pretrained=pretrained, **kwargs)
elif model_name == names[6]:
return pvt_v2_b2_li(pretrained=pretrained, **kwargs)
else:
raise Exception(f"error model name: '{model_name}'")
def formate_abs_path(path: str) -> str:
"""这主要针对windows环境,输入的绝对路径可能不包含盘符,这里进行补充
主要是用于打印效果
如果不是windows环境,直接返回path,相当于没有调用这个函数
Parameters
----------
path : str
待转换的路径
Returns
-------
str
增加了盘符的路径
摘抄自swanlab
"""
if platform.system() != "Windows":
return path
if not os.path.isabs(path):
return path
need_add = len(path) < 3 or path[1] != ":"
# 处理反斜杠, 保证路径的正确性
path = path.replace("/", "\\")
if need_add:
return os.path.join(os.getcwd()[:2], path)
return path
if __name__ == "__main__":
train_list, train_labels = read_aoteman_data(root='../datasets/aoteman', train=True)
test_list, test_labels = read_aoteman_data(root='../datasets/aoteman', train=False)
datas = train_list + test_list
labels = train_labels + test_labels
transform = transforms.Compose([
transforms.Resize((224, 224)),
transforms.ToTensor(),
])
all_dataset = ImageDataset(datas, labels, transform=transform)
dataloader = DataLoader(all_dataset, batch_size=1000)
mean_data, std_data = get_mean_and_std(dataloader)
print(mean_data, std_data)