-
Notifications
You must be signed in to change notification settings - Fork 0
/
model.py
202 lines (126 loc) · 6.41 KB
/
model.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
import torch
import torch.nn as nn
import torch.nn.functional as F
import numpy as np
class GatedTanh(nn.Module):
def __init__(self, inp_size, out_size):
super(GatedTanh, self).__init__()
self.i2t = nn.Linear(inp_size, out_size) # input transform
self.i2g = nn.Linear(inp_size, out_size) # input to gate
def forward(self, data):
inp2transform = torch.tanh(self.i2t(data))
inp2gate = torch.sigmoid(self.i2g(data))
gated_transform = torch.mul(inp2transform, inp2gate)
return gated_transform
class MlpLayer(nn.Module):
def __init__(self, input_size, output_size):
super(MlpLayer, self).__init__()
self.mlp = nn.Linear(input_size, output_size)
self.acts = ['relu', 'tanh']
def forward(self, data, act='relu'):
output = self.mlp(data)
if act=='relu':
output = F.relu(output)
return output
class QuestionEncoder(nn.Module):
def __init__(self, vocab_size, embed_dim, gru_hidden_size):
super(QuestionEncoder, self).__init__()
self.embeddings = nn.Embedding(vocab_size + 1, embed_dim)
pretrained_wemb = np.zeros((vocab_size + 1, embed_dim), dtype=np.float32)
pretrained_wemb[:vocab_size] = np.load("data/glove_pretrained_300.npy")
self.embeddings.weight.data.copy_(torch.from_numpy(pretrained_wemb))
self.encoder = nn.GRU(embed_dim, gru_hidden_size)
self.do1 = nn.Dropout(0.1)
self.do2 = nn.Dropout(0.2)
def forward(self, data, q_lens):
data = self.embeddings(data)
data = self.do1(data)
data = data.permute(1,0,2) # 512x14x300 --> 14x512x300
data = torch.nn.utils.rnn.pack_padded_sequence(data, q_lens)
self.encoder.flatten_parameters() # Multi-GPU Training
outputs, hidden = self.encoder(data)
outputs, _ = torch.nn.utils.rnn.pad_packed_sequence(outputs)
return self.do2(hidden.squeeze())
class TopDownAttention(nn.Module):
def __init__(self, inp_size, hidden_size):
super(TopDownAttention, self).__init__()
#self.nonlinear = GatedTanh(inp_size, hidden_size)
self.nonlinear = MlpLayer(inp_size, hidden_size)
self.attn_layer = nn.Linear(hidden_size, 1)
def forward(self, data):
gated_transform = self.nonlinear(data)
attn_scores = self.attn_layer(gated_transform)
attn_probs = torch.softmax(attn_scores, dim=1)
return attn_probs
class ImageEncoder(nn.Module):
def __init__(self, inp_size, hidden_size):
super(ImageEncoder, self).__init__()
self.attention = TopDownAttention(inp_size, hidden_size)
self.do1 = nn.Dropout(0.2)
def forward(self, img_features, ques_features):
ques_features = ques_features.unsqueeze(1).expand(-1,36,-1) # N * k * enc_size
concat_features = torch.cat((ques_features, img_features), 2) # N * k * (enc_size + img_size)
attn_probs = self.attention(concat_features) # N * k * 1
img_encode = torch.sum(torch.mul(img_features, attn_probs), dim=1)
img_encode = self.do1(img_encode)
return img_encode
class JointEmbedding(nn.Module):
def __init__(self, ques_inp_size, img_inp_size, out_size):
super(JointEmbedding, self).__init__()
#self.ques_nonlinear = GatedTanh(ques_inp_size, out_size)
#self.img_nonlinear = GatedTanh(img_inp_size, out_size)
self.ques_nonlinear = MlpLayer(ques_inp_size, out_size)
self.img_nonlinear = MlpLayer(img_inp_size, out_size)
def forward(self, ques_features, img_features):
ques_features = self.ques_nonlinear(ques_features) # N * 512
img_features = self.img_nonlinear(img_features) # N * 2048 -> N * 512
joint_embed = torch.mul(ques_features, img_features) # N * 512
return joint_embed
class HybridClassifier(nn.Module):
def __init__(self, inp_size, text_embed_size, img_embed_size, num_answers):
super(HybridClassifier, self).__init__()
#self.ques_nonlinear = GatedTanh(inp_size, text_embed_size)
#self.img_nonlinear = GatedTanh(inp_size, img_embed_size)
self.ques_nonlinear = MlpLayer(inp_size, text_embed_size)
self.img_nonlinear = MlpLayer(inp_size, img_embed_size)
self.ques_linear = nn.Linear(text_embed_size, num_answers)
self.img_linear = nn.Linear(img_embed_size, num_answers)
def forward(self, joint_embed):
ques_embed = self.ques_nonlinear(joint_embed)
img_embed = self.img_nonlinear(joint_embed)
ques_out = self.ques_linear(ques_embed)
img_out = self.img_linear(img_embed)
joint_output = torch.sigmoid(torch.add(ques_out, img_out))
return joint_output
class BasicClassifier(nn.Module):
def __init__(self, joint_embed_size, text_embed_size, num_answers):
super(BasicClassifier, self).__init__()
#self.nonlinear = GatedTanh(joint_embed_size, text_embed_size)
self.nonlinear = MlpLayer(joint_embed_size, text_embed_size)
#self.nonlinear1 = MlpLayer(text_embed_size, text_embed_size)
self.classifier = nn.Linear(text_embed_size, num_answers)
self.do = nn.Dropout(0.2)
def forward(self, joint_embed):
#output = self.classifier(self.do(self.nonlinear1(self.nonlinear(joint_embed))))
output = self.classifier(self.do(self.nonlinear(joint_embed)))
return output
class Model(nn.Module):
def __init__(self, vocab_size, word_embed_dim, hidden_size, resnet_out, num_answers):
super(Model, self).__init__()
self.ques_encoder = QuestionEncoder(vocab_size,
word_embed_dim,
hidden_size)
self.img_encoder = ImageEncoder(resnet_out + hidden_size,
hidden_size)
self.joint_embed = JointEmbedding(hidden_size,
resnet_out,
hidden_size)
self.classifier = BasicClassifier(hidden_size,
word_embed_dim,
num_answers)
def forward(self, images, questions, q_lens):
ques_enc = self.ques_encoder(questions, q_lens)
img_enc = self.img_encoder(images, ques_enc)
joint_embed = self.joint_embed(ques_enc, img_enc)
outputs = self.classifier(joint_embed)
return outputs