-
Notifications
You must be signed in to change notification settings - Fork 0
/
active_learning.py
220 lines (174 loc) · 8.28 KB
/
active_learning.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
import argparse
import json
import matplotlib.pyplot as plt
import numpy as np
import pandas as pd
import time
from os.path import join, exists, splitext, basename
from tqdm import tqdm
import torchvision.transforms as transforms
import torchvision
import torch.nn as nn
import torch
from torch.optim import lr_scheduler
from torch.utils.data.dataset import Subset
from utils.utils import BaseExp
import timm
plt.style.use('ggplot')
class Exp(BaseExp):
def __init__(self, args):
super().__init__(args)
self.NCLASS = 10
self.M = 10
self.log_dict ={"train": np.zeros((self.epochs*self.M, 2)), "test": np.zeros((self.epochs*self.M, 2))}
def exp(self):
np.random.seed(42)
transform_dict = {
'train': transforms.Compose(
[transforms.Resize((224, 224)),
transforms.RandomHorizontalFlip(),
transforms.ToTensor(),
transforms.Normalize(mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225]),
]),
'test': transforms.Compose(
[transforms.Resize((224, 224)),
transforms.ToTensor(),
transforms.Normalize(mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225]),
])}
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
model = timm.create_model(self.model, pretrained=True, num_classes=self.NCLASS)
model = model.to(device)
optimizer = torch.optim.Adam(model.parameters(), lr=self.lr)
criterion = nn.CrossEntropyLoss()
dataloader = {}
data_size = {}
phase = "test"
test_dataset = torchvision.datasets.CIFAR10(root="data", train=False, download=True, transform=transform_dict[phase])
dataloader[phase] = torch.utils.data.DataLoader(
test_dataset, batch_size=self.batch_size, shuffle=phase=="train", num_workers=4)
data_size[phase] = len(test_dataset.targets)
phase = "train"
train_dataset = torchvision.datasets.CIFAR10(root="data", train=True, download=True,
transform=transform_dict[phase])
for j in range(self.M):
phase = "train"
len_train = len(train_dataset)
if j==0:
indices = np.random.permutation(np.arange(len_train))
N = int(len_train/self.M)
init_indices = indices[:N]
rest_indices = indices[N:]
subset = Subset(train_dataset, init_indices)
rest_subset = Subset(train_dataset, rest_indices)
else:
init_indices = np.append(init_indices, new_indices)
rest_indices = np.delete(rest_indices, argsort[:N])
subset = Subset(train_dataset, init_indices)
rest_subset = Subset(train_dataset, rest_indices)
dataloader[phase] = torch.utils.data.DataLoader(
subset, batch_size=self.batch_size, shuffle=phase=="train", num_workers=4)
data_size[phase] = len(subset.indices)
restloader = torch.utils.data.DataLoader(
rest_subset, batch_size=self.batch_size, shuffle=False, num_workers=4)
for epoch in range(self.epochs):
epoch_ = epoch
epoch = self.epochs*j+epoch_
print(f"epoch: {epoch+1}")
######### train ####################
phase = "train"
model.train()
epoch_loss = 0.0
running_corrects = 0
for xs, ys in tqdm(dataloader[phase]):
xs = xs.to(device)
ys = ys.to(device)
outputs = model(xs)
loss = criterion(outputs, ys)
optimizer.zero_grad()
loss.backward()
optimizer.step()
softmax_max, preds = torch.max(torch.softmax(outputs.data, 1), 1)
epoch_loss += loss.item()
running_corrects += torch.sum(preds == ys)
epoch_acc = running_corrects.item() / data_size[phase]
epoch_loss /= data_size[phase]
self.log_dict[phase][epoch, 0] = round(epoch_loss, 3)
self.log_dict[phase][epoch, 1] = epoch_acc
print(f'{phase} Loss: {epoch_loss:.4f} Acc: {epoch_acc:.4f}')
######## select indices ################
if epoch_ == self.epochs - 1:
print("select data")
model.eval()
conf = []
for xs, _ in tqdm(restloader):
xs = xs.to(device)
with torch.no_grad():
outputs = model(xs)
softmax_max, preds = torch.max(torch.softmax(outputs.data, 1), 1)
conf.extend(softmax_max.detach().cpu().numpy())
conf = np.array(conf)
argsort = conf.argsort()
new_indices = rest_indices[argsort][:N]
############## test #################
phase = "test"
model.eval()
epoch_loss = 0.0
running_corrects = 0
for xs, ys in tqdm(dataloader[phase]):
xs = xs.to(device)
ys = ys.to(device)
with torch.no_grad():
outputs = model(xs)
loss = criterion(outputs, ys)
softmax_max, preds = torch.max(torch.softmax(outputs.data, 1), 1)
epoch_loss += loss.item()
running_corrects += torch.sum(preds == ys)
epoch_acc = running_corrects.item() / data_size[phase]
epoch_loss /= data_size[phase]
self.log_dict[phase][epoch, 0] = round(epoch_loss, 3)
self.log_dict[phase][epoch, 1] = epoch_acc
print(f'{phase} Loss: {epoch_loss:.4f} Acc: {epoch_acc:.4f}')
if self.log_dict["test"][epoch, 1] == self.log_dict["test"][:epoch + 1, 1].max():
self.best_epoch = epoch
self.best_acc = self.log_dict[phase][epoch, 1]
torch.save(model.state_dict(), f'{self.exp_path}/model.pth')
self.save_logs(epoch)
def save_logs(self, epoch):
colors = {'train': 'tab:blue', 'test': 'tab:orange'}
plt.clf()
for phase in ["train", "test"]:
plt.plot(self.log_dict[phase][:epoch + 1, 0], label=phase, color=colors[phase])
plt.legend()
plt.xlabel('epoch')
plt.title('loss')
plt.savefig(f'{self.exp_path}/loss.png')
plt.close()
plt.clf()
for phase in ["train", "test"]:
plt.plot(self.log_dict[phase][:epoch + 1, 1], label=phase, color=colors[phase])
xlabel = f'epoch (best epoch {self.best_epoch} eval acc {self.best_acc:.4f})'
plt.legend()
plt.xlabel(xlabel)
plt.title('accuracy')
plt.savefig(f'{self.exp_path}/accuracy.png')
plt.close()
self.cfg["acc"] = f'{self.best_acc:.3f}'
with open(f'{self.exp_path}/cfg.json', 'w') as f:
f.write(json.dumps(self.cfg, indent=4))
df = pd.DataFrame(np.concatenate([self.log_dict["train"], self.log_dict["test"]], axis=1))
df.columns = ["train loss", "train acc", "test loss", "test acc"]
df.to_csv(f'{self.exp_path}/log.csv', index=False)
if __name__ == '__main__':
parser = argparse.ArgumentParser()
parser.add_argument('--expid', '-i', type=str, default='000000')
parser.add_argument('--seed', type=int, default=0)
parser.add_argument('--data_path', type=str, default='data')
parser.add_argument('--static_path', type=str, default='static')
parser.add_argument('--epochs', '-e', type=int, default=10)
parser.add_argument('--batch_size', '-bs', type=int, default=8)
parser.add_argument('--model', '-m', type=str, default='efficientnet_b0')
parser.add_argument('--lr', type=float, default=1e-3)
exp = Exp(parser.parse_args())
exp.exp()