forked from juvezxy/747_QA_Project
-
Notifications
You must be signed in to change notification settings - Fork 0
/
COREQA.py
274 lines (228 loc) · 13.6 KB
/
COREQA.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
# coding:utf-8
from Encoder import *
from Decoder import *
from DataUtilsNew import *
from config import *
class COREQA(object):
def __init__(self, model_params):
self.word_indexer = model_params["word_indexer"]
self.embedding_size = model_params["embedding_size"]
self.state_size = model_params["state_size"]
self.mode_size = model_params["mode_size"]
self.ques_attention_size = model_params["ques_attention_size"]
self.kb_attention_size = model_params["kb_attention_size"]
self.max_fact_num = model_params["max_fact_num"]
self.max_ques_len = model_params["max_ques_len"]
self.learning_rate = model_params["learning_rate"]
self.mode_loss_rate = model_params["mode_loss_rate"]
self.L2_factor = model_params["L2_factor"]
self.batch_size = model_params["batch_size"]
self.epoch_size = model_params["epoch_size"]
self.instance_weight = 1.0 / self.batch_size
self.MAX_LENGTH = model_params["MAX_LENGTH"]
self.has_trained = False
################ Initialize graph components ########################
self.embedding = nn.Embedding(self.word_indexer.wordCount, self.embedding_size)
self.encoder = Encoder(self.word_indexer.wordCount, self.state_size, self.embedding)
self.decoder = Decoder(output_size=self.word_indexer.wordCount, state_size=self.state_size,
embedding=self.embedding, mode_size=self.mode_size,
kb_attention_size=self.kb_attention_size, max_fact_num=self.max_fact_num,
ques_attention_size=self.ques_attention_size, max_ques_len=self.max_ques_len)
if use_cuda:
self.encoder.cuda()
self.decoder.cuda()
self.optimizer = optim.Adam(list(self.encoder.parameters()) + list(self.decoder.parameters()),
lr=self.learning_rate, weight_decay=self.L2_factor)
#####################################################################
def fit(self, training_data):
if self.has_trained:
print('Warning! Trying to fit a trained model.')
print('Start training ...')
startTime = time.time()
lossTotal = 0.0
XEnLoss = nn.CrossEntropyLoss()
for epoch in range(self.epoch_size):
self.optimizer.zero_grad()
shuffle(training_data)
for iter in range(len(training_data)):
ques_var, answ_var, kb_var_list, answer_modes_var_list, answ4ques_locs_var_list, answ4kb_locs_var_list, kb_facts, ques, answ = vars_from_data(
training_data[iter])
answ_length = answ_var.size()[0]
#################### Process KB facts ###############################
kb_facts_embedded = []
for rel_obj in kb_var_list:
rel_embedded = self.embedding(rel_obj[0]).view(1, 1, -1)
obj_embedded = self.embedding(rel_obj[1]).view(1, 1, -1)
kb_facts_embedded.append(torch.cat((rel_embedded, obj_embedded), 2))
avg_kb_facts_embedded = kb_facts_embedded[0]
#####################################################################
######################### Encoding ##################################
self.encoder.hidden = self.encoder.init_hidden()
encoder_outputs = self.encoder(ques_var)
question_embedded = self.encoder.hidden[0].view(1, 1, -1)
cell_state = self.encoder.hidden[1].view(1, 1, -1)
#####################################################################
######################### Decoding ###################################
decoder_hidden = (question_embedded, cell_state)
decoder_input = Variable(torch.LongTensor([[SOS]]))
hist_ques = Variable(torch.zeros(1, 1, self.max_ques_len))
hist_kb = Variable(torch.zeros(1, 1, self.max_fact_num))
if use_cuda:
decoder_input = decoder_input.cuda()
hist_kb = hist_kb.cuda()
hist_ques = hist_ques.cuda()
loss = 0.0
for i in range(answ_length):
answer_mode = answer_modes_var_list[i]
word_embedded = self.embedding(decoder_input).view(1, 1, -1)
weighted_question_encoding = Variable(torch.zeros(1, 1, 2 * self.state_size))
weighted_kb_facts_encoding = Variable(torch.zeros(1, 1, 2 * self.embedding_size))
if use_cuda:
weighted_question_encoding = weighted_question_encoding.cuda()
weighted_kb_facts_encoding = weighted_kb_facts_encoding.cuda()
if (i > 0):
ques_locs = answ4ques_locs_var_list[i - 1][0][0]
kb_locs = answ4kb_locs_var_list[i - 1][0][0]
question_match_count = 0
kb_facts_match_count = 0
for ques_pos in range(len(ques_locs)):
if ques_locs[ques_pos].data[0] == 1:
weighted_question_encoding += encoder_outputs[ques_pos][0].view(1, 1, -1)
question_match_count += 1
if question_match_count > 0:
weighted_question_encoding /= question_match_count
for kb_idx in range(len(kb_locs)):
if kb_locs[kb_idx].data[0] == 1:
weighted_kb_facts_encoding += kb_facts_embedded[kb_idx]
kb_facts_match_count += 1
if kb_facts_match_count > 0:
weighted_kb_facts_encoding /= kb_facts_match_count
decoder_input_embedded = torch.cat((word_embedded, weighted_question_encoding,
weighted_kb_facts_encoding, avg_kb_facts_embedded), 2)
common_predict, decoder_hidden, mode_predict, kb_atten_predict, hist_kb, ques_atten_predict, hist_ques = self.decoder(
word_embedded, decoder_input_embedded, decoder_hidden, question_embedded, kb_facts_embedded,
hist_kb, encoder_outputs, hist_ques)
###################### Calculate Loss ################################
loss += self.instance_weight * self.mode_loss_rate * XEnLoss(mode_predict.view(1, -1), answer_mode)
mode_predict = nn.Softmax(dim=2)(mode_predict).view(3, 1)
common_mode_predict = mode_predict[0]
kb_mode_predict = mode_predict[1]
ques_mode_predict = mode_predict[2]
predicted_probs = torch.cat(
(common_predict * common_mode_predict, kb_atten_predict * kb_mode_predict,
ques_atten_predict * ques_mode_predict), 2)
if (answer_mode.data[0] == 0): # predict mode
target = answ_var[i]
elif (answer_mode.data[0] == 1): # retrieve mode
kb_locs = answ4kb_locs_var_list[i][0][0]
target = self.word_indexer.wordCount + kb_locs.data.tolist().index(1)
target = Variable(torch.LongTensor([target]).view(-1))
if use_cuda:
target = target.cuda()
else: # copy mode
ques_locs = answ4ques_locs_var_list[i][0][0]
target = self.word_indexer.wordCount + self.max_fact_num + ques_locs.data.tolist().index(1)
target = Variable(torch.LongTensor([target]).view(-1))
if use_cuda:
target = target.cuda()
loss += self.instance_weight * XEnLoss(predicted_probs.view(1, -1), target)
#####################################################################
decoder_input = answ_var[i]
#####################################################################
if (iter + 1) % self.batch_size == 0:
loss.backward()
self.optimizer.step()
self.optimizer.zero_grad()
loss = loss.data[0] / answ_length
lossTotal += loss
if (iter + 1) % 1000 == 0:
lossAvg = lossTotal / 1000
lossTotal = 0
secs = time.time() - startTime
mins = math.floor(secs / 60)
secs -= mins * 60
print('%dm %ds' % (mins, secs), 'after iteration:', iter + 1, 'with avg loss:', lossAvg)
self.has_trained = True
print('Training completed!')
def predict(self, ques_var, kb_var_list, kb_facts, ques):
#inputLength = inputVar.size()[0]
#################### Process KB facts ###############################
kb_facts_embedded = []
for rel_obj in kb_var_list:
rel_embedded = self.embedding(rel_obj[0]).view(1, 1, -1)
obj_embedded = self.embedding(rel_obj[1]).view(1, 1, -1)
kb_facts_embedded.append(torch.cat((rel_embedded, obj_embedded), 2))
avg_kb_facts_embedded = kb_facts_embedded[-1]
#####################################################################
######################### Encoding ##################################
self.encoder.hidden = self.encoder.init_hidden()
encoder_outputs = self.encoder(ques_var)
question_embedded = self.encoder.hidden[0].view(1, 1, -1)
cell_state = self.encoder.hidden[1].view(1, 1, -1)
#####################################################################
######################### Decoding ###################################
decoder_hidden = (question_embedded, cell_state)
decoder_input = Variable(torch.LongTensor([[SOS]]))
hist_ques = Variable(torch.zeros(1, 1, self.max_ques_len))
hist_kb = Variable(torch.zeros(1, 1, self.max_fact_num))
if use_cuda:
decoder_input = decoder_input.cuda()
hist_kb = hist_kb.cuda()
hist_ques = hist_ques.cuda()
decoded_id = []
decoded_token = []
weighted_question_encoding = Variable(torch.zeros(1, 1, 2 * self.state_size))
weighted_kb_facts_encoding = Variable(torch.zeros(1, 1, 2 * self.embedding_size))
if use_cuda:
weighted_question_encoding = weighted_question_encoding.cuda()
weighted_kb_facts_encoding = weighted_kb_facts_encoding.cuda()
for i in range(self.MAX_LENGTH):
word_embedded = self.embedding(decoder_input).view(1, 1, -1)
decoder_input_embedded = torch.cat((word_embedded, weighted_question_encoding,
weighted_kb_facts_encoding, avg_kb_facts_embedded), 2)
common_predict, decoder_hidden, mode_predict, kb_atten_predict, hist_kb, ques_atten_predict, hist_ques = self.decoder(
word_embedded, decoder_input_embedded, decoder_hidden, question_embedded, kb_facts_embedded,
hist_kb, encoder_outputs, hist_ques)
mode_predict = nn.Softmax(dim=2)(mode_predict).view(3, 1)
common_mode_predict = mode_predict[0]
kb_mode_predict = mode_predict[1]
ques_mode_predict = mode_predict[2]
predicted_probs = torch.cat((common_predict * common_mode_predict, kb_atten_predict * kb_mode_predict,
ques_atten_predict * ques_mode_predict), 2)
topv3, topi3 = predicted_probs.data.topk(3)
idx = topi3[0][0][0]
if idx < self.word_indexer.wordCount: # predict mode
if idx == EOS:
decoded_id.append(EOS)
decoded_token.append("_EOS")
break
else:
decoded_id.append(idx)
word = self.word_indexer.index2word[idx]
decoded_token.append(word)
decoder_input = Variable(torch.LongTensor([[idx]]))
weighted_kb_facts_encoding = Variable(torch.zeros(1, 1, 2 * self.embedding_size))
elif idx < self.word_indexer.wordCount + self.max_fact_num: # retrieve mode
kb_idx = idx - self.word_indexer.wordCount
rel_obj_idx = kb_var_list[kb_idx]
obj_idx = rel_obj_idx[1]
decoded_id.append(obj_idx.data[0])
kb_sub, kb_rel, kb_obj = kb_facts[kb_idx]
decoded_token.append(kb_obj)
decoder_input = obj_idx
weighted_kb_facts_encoding = kb_facts_embedded[kb_idx]
else: # copy mode
copy_idx = idx - self.word_indexer.wordCount - self.max_fact_num
word_idx = ques_var[copy_idx]
decoded_id.append(word_idx.data[0])
if copy_idx < len(ques):
word = ques[copy_idx]
else:
word = FIL
decoded_token.append(word)
decoder_input = word_idx
weighted_question_encoding = encoder_outputs[copy_idx].view(1, 1, -1)
if use_cuda:
weighted_kb_facts_encoding = weighted_kb_facts_encoding.cuda()
decoder_input = decoder_input.cuda()
return decoded_id, decoded_token