-
Notifications
You must be signed in to change notification settings - Fork 3
Expand file tree
/
Copy pathdata_loader.py
More file actions
194 lines (164 loc) · 8.63 KB
/
data_loader.py
File metadata and controls
194 lines (164 loc) · 8.63 KB
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
import torch
import numpy as np
import scipy.sparse as sp
from torch.utils.data import Dataset, DataLoader
from torch.nn.utils.rnn import pad_sequence
def build_dataset(configs, dataset, mode='train'):
dataset = MyDataSet(dataset)
if mode == 'train':
data_loader = DataLoader(dataset=dataset, batch_size=configs.batch_size, shuffle=True,
collate_fn=bert_batch_preprocessing)
else:
data_loader = DataLoader(dataset=dataset, batch_size=1, shuffle=False,
collate_fn=bert_batch_preprocessing)
return data_loader
def bert_batch_preprocessing(batch):
docid_list, clause_list, doc_len_list, clause_len_list, pairs, \
feq, feq_len, feq_an, feq_mask, feq_seg, fcq, fcq_len, fcq_an, fcq_mask, fcq_seg, \
bcq, bcq_len, bcq_an, bcq_mask, bcq_seg, beq, beq_len, beq_an, beq_mask, beq_seg, \
fc_num, be_num = zip(*batch)
# query, query_mask, query_seg, answer已经在makeData_dual中padding好了，还需要设置answer mask
feq_an, fe_an_mask = get_answer_pad_mask(feq_an)
bcq_an, bc_an_mask = get_answer_pad_mask(bcq_an)
fcq_an, fc_an_mask = get_answer_pad_mask(fcq_an)
beq_an, be_an_mask = get_answer_pad_mask(beq_an)
feq, feq_mask, feq_seg, feq_len, fe_clause_len, fe_doc_len, fe_adj \
= convert_batch(feq, feq_mask, feq_seg, feq_len, clause_len_list, doc_len_list)
fcq, fcq_mask, fcq_seg, fcq_len, fc_clause_len, fc_doc_len, fc_adj \
= convert_batch(fcq, fcq_mask, fcq_seg, fcq_len, clause_len_list, doc_len_list)
bcq, bcq_mask, bcq_seg, bcq_len, bc_clause_len, bc_doc_len, bc_adj \
= convert_batch(bcq, bcq_mask, bcq_seg, bcq_len, clause_len_list, doc_len_list)
beq, beq_mask, beq_seg, beq_len, be_clause_len, be_doc_len, be_adj \
= convert_batch(beq, beq_mask, beq_seg, beq_len, clause_len_list, doc_len_list)
return docid_list, clause_list, pairs, \
feq, feq_mask, feq_seg, feq_len, fe_clause_len, fe_doc_len, fe_adj, feq_an, fe_an_mask, \
fcq, fcq_mask, fcq_seg, fcq_len, fc_clause_len, fc_doc_len, fc_adj, fcq_an, fc_an_mask, \
bcq, bcq_mask, bcq_seg, bcq_len, bc_clause_len, bc_doc_len, bc_adj, bcq_an, bc_an_mask, \
beq, beq_mask, beq_seg, beq_len, be_clause_len, be_doc_len, be_adj, beq_an, be_an_mask
class MyDataSet(Dataset):
def __init__(self, pre_data):
self.docid_list = pre_data.docid_list
self.clause_list = pre_data.clause_list
self.doc_len_list = pre_data.doc_len_list
self.clause_len_list = pre_data.clause_len_list
self.pairs = pre_data.pairs
self._f_emo_query = pre_data._f_emo_query # [1, max_for_emo_len]
self._f_emo_query_len = pre_data._f_emo_query_len
self._f_emo_query_answer = pre_data._f_emo_query_answer
self._f_emo_query_mask = pre_data._f_emo_query_mask # [1,max_for_emo_len]
self._f_emo_query_seg = pre_data._f_emo_query_seg # [1,max_for_emo_len]
self._f_cau_query = pre_data._f_cau_query # [max_for_num, max_for_cau_len]
self._f_cau_query_len = pre_data._f_cau_query_len
self._f_cau_query_answer = pre_data._f_cau_query_answer
self._f_cau_query_mask = pre_data._f_cau_query_mask # [max_for_num, max_for_cau_len]
self._f_cau_query_seg = pre_data._f_cau_query_seg # [max_for_num, max_for_cau_len]
self._b_cau_query = pre_data._b_cau_query #
self._b_cau_query_len = pre_data._b_cau_query_len
self._b_cau_query_answer = pre_data._b_cau_query_answer
self._b_cau_query_mask = pre_data._b_cau_query_mask #
self._b_cau_query_seg = pre_data._b_cau_query_seg #
self._b_emo_query = pre_data._b_emo_query
self._b_emo_query_len = pre_data._b_emo_query_len
self._b_emo_query_answer = pre_data._b_emo_query_answer
self._b_emo_query_mask = pre_data._b_emo_query_mask #
self._b_emo_query_seg = pre_data._b_emo_query_seg #
self._forward_c_num = pre_data._forward_c_num
self._backward_e_num = pre_data._backward_e_num
# print(self.doc_len_list)
# print(self.clause_len_list)
# print(self.pairs)
# print(self._f_query_len_list)
# print(self._b_query_len_list)
# print(self._f_cau_query[0])
# print(self._f_cau_query_len[0])
# print(self._f_cau_query_answer[0])
# print(self._f_cau_query_mask[0])
# print(self._f_cau_query_seg[0])
# exit(0)
# print(self._f_cau_query)
# print(self._f_emo_query_answer)
# print(self._f_cau_query_answer)
# print(self._f_emo_query_mask)
# print(self._f_cau_query_mask)
# print(self._f_emo_query_seg)
# print(self._f_cau_query_seg)
# print(self._b_emo_query)
# print(self._b_cau_query)
# print(self._b_emo_query_answer)
# print(self._b_cau_query_answer)
# print(self._b_emo_query_mask)
# print(self._b_cau_query_mask)
# print(self._b_emo_query_seg)
# print(self._b_cau_query_seg)
# print(self._forward_num)
# print(self._backward_num)
def __len__(self):
return len(self.doc_len_list)
def __getitem__(self, i):
docid_list, clause_list, doc_len_list, clause_len_list, pairs, \
feq, feq_len, feq_an, feq_mask, feq_seg, fcq, fcq_len, fcq_an, fcq_mask, fcq_seg, \
bcq, bcq_len, bcq_an, bcq_mask, bcq_seg, beq, beq_len, beq_an, beq_mask, beq_seg, \
fc_num, be_num = \
self.docid_list[i], self.clause_list[i], self.doc_len_list[i], self.clause_len_list[i], self.pairs[i], \
self._f_emo_query[i], self._f_emo_query_len[i], self._f_emo_query_answer[i], self._f_emo_query_mask[i], self._f_emo_query_seg[i], \
self._f_cau_query[i], self._f_cau_query_len[i], self._f_cau_query_answer[i], self._f_cau_query_mask[i], self._f_cau_query_seg[i], \
self._b_cau_query[i], self._b_cau_query_len[i], self._b_cau_query_answer[i], self._b_cau_query_mask[i], self._b_cau_query_seg[i], \
self._b_emo_query[i], self._b_emo_query_len[i], self._b_emo_query_answer[i], self._b_emo_query_mask[i], self._b_emo_query_seg[i], \
self._forward_c_num[i], self._backward_e_num[i]
return docid_list, clause_list, doc_len_list, clause_len_list, pairs, \
feq, feq_len, feq_an, feq_mask, feq_seg, fcq, fcq_len, fcq_an, fcq_mask, fcq_seg, \
bcq, bcq_len, bcq_an, bcq_mask, bcq_seg, beq, beq_len, beq_an, beq_mask, beq_seg, \
fc_num, be_num
def get_answer_pad_mask(answer):
new_answer = []
for batch_answer in answer:
for qa_answer in batch_answer:
new_answer.append(torch.tensor(qa_answer))
answer = pad_sequence(new_answer, padding_value=-1).transpose(0, 1)
mask = torch.where(answer != -1, 1, 0)
assert mask.shape == answer.shape
return answer, mask
def convert_batch(query, query_mask, query_seg, query_len, seq_len, doc_len):
query_list, query_mask_list, query_seg_list = [], [], []
new_query_len, new_seq_len, new_doc_len = [], [], []
for i in range(len(query_len)):
for j in range(len(query_len[i])):
query_list.append(query[i][j])
query_mask_list.append(query_mask[i][j])
query_seg_list.append(query_seg[i][j])
new_seq_len.append(seq_len[i])
new_doc_len.append(doc_len[i])
new_query_len.append(query_len[i][j])
query = torch.LongTensor(query_list)
query_mask = torch.LongTensor(query_mask_list)
query_seg = torch.LongTensor(query_seg_list)
adj = pad_matrices(new_doc_len)
return query, query_mask, query_seg, new_query_len, new_seq_len, new_doc_len, adj
def pad_list(element_list, max_len, pad_mark):
element_list_pad = element_list[:]
pad_mark_list = [pad_mark] * (max_len - len(element_list))
element_list_pad.extend(pad_mark_list)
return element_list_pad
def pad_docs(doc_len_b, answer):
max_doc_len = max(doc_len_b)
y_mask_b, ans_b = [], []
for ans in answer:
ans_ = pad_list(ans, max_doc_len, -1)
y_mask = list(map(lambda x: 0 if x == -1 else 1, ans_))
y_mask_b.append(y_mask)
ans_b.append(ans_)
return y_mask_b, ans_b
def pad_matrices(doc_len_b):
N = max(doc_len_b)
adj_b = []
for doc_len in doc_len_b:
adj = np.ones((doc_len, doc_len))
adj = sp.coo_matrix(adj)
adj = sp.coo_matrix((adj.data, (adj.row, adj.col)),
shape=(N, N), dtype=np.float32)
adj_b.append(adj.toarray())
return adj_b
if __name__ == "__main__":
from transformers import BertTokenizer
tok = BertTokenizer.from_pretrained('../bert-base-chinese')
print(tok.encode('情感子句'))