forked from Awenbocc/med-vqa
-
Notifications
You must be signed in to change notification settings - Fork 0
/
classify_question.py
240 lines (182 loc) · 7.47 KB
/
classify_question.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
# -*- coding: utf-8 -*-#
#-------------------------------------------------------------------------------
# Name: classify_question
# Description:
# Author: Boliu.Kelvin
# Date: 2020/5/14
#-------------------------------------------------------------------------------
import torch
from dataset_RAD import Dictionary,VQAFeatureDataset
import torch.nn as nn
import os
from torch.utils.data import DataLoader
from language_model import WordEmbedding,QuestionEmbedding
import argparse
from torch.nn.init import kaiming_uniform_, xavier_uniform_
import torch.nn.functional as F
import utils
from datetime import datetime
def euclidean_dist(x, y):
# x: N x D
# y: M x D
n = x.size(0)
m = y.size(0)
d = x.size(1)
assert d == y.size(1)
x = x.unsqueeze(1).expand(n, m, d)
y = y.unsqueeze(0).expand(n, m, d)
return torch.pow(x - y, 2).sum(2)
def linear(in_dim, out_dim, bias=True):
lin = nn.Linear(in_dim, out_dim, bias=bias)
xavier_uniform_(lin.weight)
if bias:
lin.bias.data.zero_()
return lin
# change the question b*number*hidden -> b*hidden
class QuestionAttention(nn.Module):
def __init__(self, dim):
super().__init__()
self.tanh_gate = linear(300 + dim, dim)
self.sigmoid_gate = linear(300 + dim, dim)
self.attn = linear(dim, 1)
self.dim = dim
def forward(self, context, question): #b*12*300 b*12*1024
concated = torch.cat([context, question], -1) # b * 12 * 300 + 1024
concated = torch.mul(torch.tanh(self.tanh_gate(concated)), torch.sigmoid(self.sigmoid_gate(concated))) #b*12*1024
a = self.attn(concated) # #b*12*1
attn = F.softmax(a.squeeze(), 1) #b*12
ques_attn = torch.bmm(attn.unsqueeze(1), question).squeeze() #b*1024
return ques_attn
class typeAttention(nn.Module):
def __init__(self, size_question, path_init):
super(typeAttention, self).__init__()
self.w_emb = WordEmbedding(size_question, 300, 0.0, False)
self.w_emb.init_embedding(path_init)
self.q_emb = QuestionEmbedding(300, 1024, 1, False, 0.0, 'GRU')
self.q_final = QuestionAttention(1024)
self.f_fc1 = linear(1024, 2048)
self.f_fc2 = linear(2048, 1024)
self.f_fc3 = linear(1024, 1024)
def forward(self, question):
w_emb = self.w_emb(question)
q_emb = self.q_emb.forward_all(w_emb) # [batch, q_len, q_dim]
q_final = self.q_final(w_emb, q_emb) # b, 1024
x_f = self.f_fc1(q_final)
x_f = F.relu(x_f)
x_f = self.f_fc2(x_f)
x_f = F.dropout(x_f)
x_f = F.relu(x_f)
x_f = self.f_fc3(x_f)
return x_f
class classify_model(nn.Module):
def __init__(self,size_question,path_init):
super(classify_model,self).__init__()
self.w_emb = WordEmbedding(size_question,300, 0.0, False)
self.w_emb.init_embedding(path_init)
self.q_emb = QuestionEmbedding(300, 1024 , 1, False, 0.0, 'GRU')
self.q_final = QuestionAttention(1024)
self.f_fc1 = linear(1024,256)
self.f_fc2 = linear(256,64)
self.f_fc3 = linear(64,2)
def forward(self,question):
w_emb = self.w_emb(question)
q_emb = self.q_emb.forward_all(w_emb) # [batch, q_len, q_dim]
q_final = self.q_final(w_emb,q_emb) #b, 1024
x_f = self.f_fc1(q_final)
x_f = F.relu(x_f)
x_f = self.f_fc2(x_f)
x_f = F.dropout(x_f)
x_f = F.relu(x_f)
x_f = self.f_fc3(x_f)
return x_f
def parse_args():
parser = argparse.ArgumentParser(description="Med VQA over MAC")
# GPU config
parser.add_argument('--seed', type=int, default=5
, help='random seed for gpu.default:5')
parser.add_argument('--gpu', type=int, default=0,
help='use gpu device. default:0')
args = parser.parse_args()
return args
# Evaluation
def evaluate(model, dataloader,logger,device):
score = 0
number =0
model.eval()
with torch.no_grad():
for i,row in enumerate(dataloader):
image_data, question, target, answer_type, question_type, phrase_type, answer_target = row
question, answer_target = question.to(device), answer_target.to(device)
output = model(question)
pred = output.data.max(1)[1]
correct = pred.eq(answer_target.data).cpu().sum()
score+=correct.item()
number+=len(answer_target)
score = score / number * 100.
logger.info('[Validate] Val_Acc:{:.6f}%'.format(score))
return score
if __name__=='__main__':
dataroot = './data_RAD'
args = parse_args()
# # set GPU device
device = torch.device("cuda:" + str(args.gpu) if args.gpu >= 0 else "cpu")
# Fixed ramdom seed
torch.manual_seed(args.seed)
torch.cuda.manual_seed(args.seed)
torch.backends.cudnn.benchmark = True
torch.backends.cudnn.deterministic = True
d = Dictionary.load_from_file(os.path.join(dataroot, 'dictionary.pkl'))
train_dataset = VQAFeatureDataset('train', d, dataroot)
train_data = DataLoader(train_dataset, batch_size=32, shuffle=True, num_workers=2, pin_memory=True, drop_last=False)
val_dataset = VQAFeatureDataset('test', d, dataroot)
val_data = DataLoader(val_dataset, batch_size=32, shuffle=False, num_workers=2, pin_memory=True, drop_last=False)
net = classify_model(d.ntoken,'./data_RAD/glove6b_init_300d.npy')
net =net.to(device)
run_timestamp = datetime.now().strftime("%Y%b%d-%H%M%S")
ckpt_path = os.path.join('./log', run_timestamp)
utils.create_dir(ckpt_path)
# create logger
logger = utils.Logger(os.path.join(ckpt_path, 'medVQA.log')).get_logger()
logger.info(">>>The net is:")
logger.info(net)
logger.info(">>>The args is:")
logger.info(args.__repr__())
#
criterion = nn.CrossEntropyLoss()
optimizer = torch.optim.Adam(net.parameters(), lr=1e-4)
#
epochs = 200
best_eval_score = 0
best_epoch = 0
for epoch in range(epochs):
net.train()
acc = 0.
number_dataset = 0
total_loss = 0
for i, row in enumerate(train_data):
image_data, question, target, answer_type, question_type, phrase_type, answer_target = row
question, answer_target = question.to(device), answer_target.to(device)
optimizer.zero_grad()
output = net(question)
loss = criterion(output,answer_target)
loss.backward()
#nn.utils.clip_grad_norm(net.parameters(),0.25)
optimizer.step()
pred = output.data.max(1)[1]
correct = (pred==answer_target).data.cpu().sum()
acc += correct.item()
number_dataset += len(answer_target)
total_loss+= loss
total_loss /= len(train_data)
acc = acc/ number_dataset * 100.
logger.info('-------[Epoch]:{}-------'.format(epoch))
logger.info('[Train] Loss:{:.6f} , Train_Acc:{:.6f}%'.format(total_loss, acc
))
# Evaluation
if val_data is not None:
eval_score = evaluate(net, val_data, logger, device)
if eval_score > best_eval_score:
best_eval_score = eval_score
best_epoch = epoch
utils.save_model(ckpt_path, net, epoch)
logger.info('[Result] The best acc is {:.6f}% at epoch {}'.format(best_eval_score, best_epoch))