-
Notifications
You must be signed in to change notification settings - Fork 2
/
Copy pathdataset.py
493 lines (439 loc) · 20.1 KB
/
dataset.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
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
'''
multi + search score
有单独的response和history
数据集是从data resp里面处理的， 和data_resp 一样的train 和valid
在第二步生成训练数据的时候使用了并发
https://rebootcat.com/2017/08/20/Elasticsearch_handle_with_python/
https://blog.csdn.net/liaodaoluyun/article/details/82346086
'''
import tqdm
import os
import pickle
import logging as log
import torch
from torch.utils import data
from elasticsearch import Elasticsearch
from elasticsearch import helpers
import math
import random
import operator
from functools import reduce
from collections import Counter
import torch.multiprocessing as mp
import time
class Dataset(data.Dataset):
def __init__(self, problem_number, concept_num, max_sample_num, root_dir, split='train'):
super().__init__()
self.seq_len = 200
self.map_dim = 0
self.prob_encode_dim = 0
self.path = root_dir
self.problem_number = problem_number
self.concept_num = concept_num
self.show_len = 100
self.split = split
# self.data_list = [[], []]
# self.data_list = []
self.manager = mp.Manager
self.data_list = self.manager().list()
# self.raw_data_list = []
self.data_seq_list = []
log.info('Processing data...')
self.item_count = 1
self.total_train_num = 0
self.stu_count = 0
self.sear_index = None
self.my_txt = None
self.max_sample_num = max_sample_num
# self.ques_concept_relation
self.process()
log.info('Processing data done!')
# self.count_0 = 0
# self.total_pb = 0
def __len__(self):
return len(self.data_list)
def __getitem__(self, index):
return self.data_list[index]
def collate(self, batch):
seq_num, y = [], [] #seq_num: the actual length of history record
# [[for _ in range(0, 4)] for _ in range(0, 200)],
x = []
seq_length = len(batch[0][1][1]) # the unifrom length of hitory record
x_len = len(batch[0][1][0][0])
# x_len = 9
for i in range(0, seq_length):
this_x = []
for j in range(0, x_len):
this_x.append([])
x.append(this_x)
for data in batch:
this_seq_num, [this_x, this_y] = data
seq_num.append(this_seq_num)
for i in range(0, seq_length):
# print(i, 'iiiiiiiiiiiiiiiiiiii')
for j in range(0, x_len):
# print('iiiiiiiiiiiiiiii', i, j)
x[i][j].append(this_x[i][j])
# y[i].append(this_y[i])
# y += this_y[1 : this_seq_num]
y += this_y[0 : this_seq_num]
# y += this_y
batch_x, batch_y =[], []
# problem, related_concept_index, operate, filter0, prob_ids, _, _
# prob_id, skills, item_id, response, all_bg_index_tensor, own_same_concept_index
for i in range(0, seq_length):
x_info = []
for j in range(0, x_len):
x_info.append(torch.tensor(x[i][j]))
# if j == 1 or j == 5 or j == 6:
# x_info.append(torch.tensor(x[i][j]))
# elif j == 4:
# x_info.append(torch.tensor(x[i][j]).long())
# else:
# x_info.append(torch.tensor(x[i][j]).float())
batch_x.append(x_info)
return [torch.tensor(seq_num), batch_x], torch.tensor(y).float()
def getbackgroud(self, previous_txt, skills, response, problem):
g_skills = skills.copy()
g_skills.sort()
add_ptxt = 'p' + str(problem) + ' '
skill_str = ''
full_skill_str = ''
for s in g_skills:
full_skill_str += str(s) + '-'
if s != 0:
add_ptxt += 'c' + str(s) + ' '
skill_str += 'c' + str(s) + ' '
if response == 1:
previous_txt = add_ptxt + 'R ' + previous_txt
elif response == 0:
previous_txt = add_ptxt + 'W ' + previous_txt
# previous_txt = add_ptxt + previous_txt
return previous_txt, skill_str[:-1], full_skill_str[:-1]
def get_pickle_id(self, results):
id_list = []
score_list = []
for item in results:
this_id = item['_source']['pickle_id']
if item['_score'] is None:
this_score = -1000
else:
this_score = float(item['_score'])
id_list.append(this_id)
score_list.append(this_score)
return id_list, score_list
def get_difffques_id(self, results):
ques_list = []
resp_list = []
# score_list = []
for item in results:
ques_id = item['_source']['problem']
ques_resp = item['_source']['response']
# if item['_score'] is None:
# this_score = -1000
# else:
# this_score = float(item['_score'])
ques_list.append(ques_id)
resp_list.append(ques_resp)
# score_list.append(this_score)
return ques_list, resp_list#, score_list
def pick_index(self, c_txt, sq_id, stu_id):
c_query = {'query':{
'bool' :
{
'must' :
[
{ 'match': {'skills': c_txt}},
{ 'term': {'user': stu_id}},
{ 'range': {
'sq_id': {
'lt': sq_id
}
}}
]
}
}
}
dt_name = self.path.split('/')[-2]
heads = {'index': dt_name + '_train_search'}
this_action = [heads, c_query]
return this_action
def question_action_dict(self, prob_id, bg_txt, stu_id):
first_query = {'query':{
'bool' :
{
'must' :
# 'should':
[
# { 'term': {'problem': int(prob_id.cpu().numpy())}},
{ 'term': {'problem': prob_id}},
{ 'match': {'history': bg_txt}}
],
'must_not': {'term': {'user': stu_id}}
}
}
}
heads = {'index': self.sear_index}
this_action = [heads, first_query]
return this_action
def concepts_action_dict(self, c_txt, bg_txt, stu_id):
second_query = {'query':{
'bool' :
{
# 'must' :
'should':
[
{ 'term': {'skills': c_txt}},
{ 'match': {'history': bg_txt}}
],
'must_not':
{'term': {'user': stu_id}}
}
}
}
heads = {'index': self.sear_index}
this_action = [heads, second_query]
return this_action
def obtain_bg_action(self, prob_id, item_id, stu_id):
# bg_index = self.search(prob_id, item_id, stu_id)
bg_txt, c_txt = self.my_txt[item_id]
if len(bg_txt) > 1024:
bg_txt = bg_txt[-1024:]
question_action = self.question_action_dict(prob_id, bg_txt, stu_id)
concepts_action = self.concepts_action_dict(c_txt, bg_txt, stu_id)
return question_action, concepts_action
def data_reader(self, stu_records, es, stu_id, actual_len, e_index):
'''
@params:
stu_record: learning history of a user
@returns:
x: list of response(0 or 1), embedding of problem, user embedding, related_concept_index
y: response
'''
# print('stuid:', stu_id)
x_list = []
y_list = []
# background_tensor_list = []
# background_noresp_list = []
# response_list = []
background_txt_list = []
# background_ques_seq_list = []
# background_resp_seq_list = []
# question_seq = [0] * self.seq_len
# resp_seq = [-1] * self.seq_len
'''initial previous'''
# previous_true, previous_all,
previous_txt, skill_str = '', ''
# for i in range(0, actual_len):
sq_id = 0
for i in range(0, len(stu_records)):
# _, problem_id,skills, response= stu_records[i]
problem_id, skills, interval_time, elapse_time, response = stu_records[i]
# this_x = [problem_id, skills, self.item_count, response]
this_x = None
this_y = response
if i < actual_len:
# this_x = [problem_id, skills, self.item_count, response, sq_id, stu_id]
this_x = [problem_id, skills, self.item_count, response, sq_id, stu_id, interval_time, elapse_time]
x_list.append(this_x)
y_list.append(this_y)
# self.data_list.append([this_x, this_y])
# ptrue_tensor = torch.tensor(previous_true).float()
# pall_tensor = torch.tensor(previous_all).float()
# background_tensor_list.append(torch.cat([ptrue_tensor, pall_tensor], dim = 0))
# true_all_cat = torch.cat([ptrue_tensor, pall_tensor], dim = 0)
# background_noresp_list.append(true_all_cat)
# response_list.append(torch.tensor(response))
current_his_txt, skill_str, full_skill_str = self.getbackgroud(previous_txt, skills, response, problem_id)
background_txt_list.append([previous_txt, skill_str])
'''insert data to db'''
skill_dict = dict()
skill_dict['skills'] = skill_str
other_dict = {'problem': problem_id,
'user': stu_id,
'sq_id': sq_id,
'history': previous_txt,
'pickle_id': self.item_count,
'response': response,
'full_skill': skills}
# 'full_skill_str': full_skill_str}
this_dict = {**skill_dict, ** other_dict}
sq_id += 1
previous_txt = current_his_txt
####################
if self.split == 'train':
es.index(index=e_index[0], doc_type='train', id=self.item_count,body=this_dict)
######################
self.item_count += 1
else:
# this_x = [problem_id, skills, 0, response, sq_id, stu_id]
# this_x = [problem_id, skills, 0, response, sq_id, stu_id, interval_time, elapse_time]
this_x = [problem_id, skills, 0, 0, sq_id, stu_id, interval_time, elapse_time]
x_list.append(this_x)
y_list.append(this_y)
# return background_noresp_list, response_list, background_ques_seq_list, background_resp_seq_list, background_txt_list,
return background_txt_list, x_list, y_list
def target_prepare(self, my_txt):
dt_name = self.path.split('/')[-2] #db_name + '_train_search', db_name + '_valid_search'
self.sear_index = dt_name + '_train_search'
self.my_txt = my_txt
def merge_ques_concept_list(self, ques_index, ques_score, concept_index, concept_score):
final_index = ques_index
final_score = ques_score
con_len = len(concept_index)
for i in range(0, con_len):
this_con_index = concept_index[i]
this_con_score = concept_score[i]
if this_con_index in final_index:
this_pos = final_index.index(this_con_index)
this_final_score = final_score[this_pos]
if this_final_score < this_con_score:
final_score[this_pos] = this_con_score
else:
final_index.append(this_con_index)
final_score.append(this_con_score)
return final_index, final_score
# def get_different_questions(diffques_indexes, diffques_score):
# TODO
def bulk_seach(self, es, action_list):
'''type == same: serach the record to answer similar question, else search the records of different question'''
# if type == 'same':
# id_identify_funcs = self.get_pickle_id
# else:
# id_identify_funcs = self.get_difffques_id
result = es.msearch(body = action_list)
# indexes = []
# scores = []
first_return, second_return = [], []
for r in result['responses']: #针对multisearch里所有的每个search
# item, score = self.get_pickle_id(r['hits']['hits'])
first, second = self.get_pickle_id(r['hits']['hits'])
first_return.append(first)
second_return.append(second)
return first_return, second_return
# def process_diff_ques(problem_id, my_diffques_indexes, my_diffques_resp):
def raw_to_final(self, data):
actual_len = data[0]
this_x, this_y = [], []
org_x, org_y = data[1]
# for record in org_x:
item_sq_map = dict()
ques_action_list, con_action_list = [], []
# = []
# rnn_index_action_list = []
# bg_index_list = []
# for ir in range(0, len(org_x)):
for ir in range(0, actual_len):
record = org_x[ir]
# print(record)
'''
sq_id: 在序列中的编号
stu_id: 序列所代表的学生的id
'''
# problem_id, skills, item_num, response, sq_id, stu_id = record
problem_id, skills, item_num, response, sq_id, stu_id, interval_time, elapse_time = record
item_sq_map[item_num] = sq_id
# bg_index = None
# own_same_concept_index = []
# if ir < actual_len:
# bg_index = self.obtain_bg_index(problem_id, item_num, stu_id)
'''
bg_ques_action: 其他用户做了这个这个问题的记录
bg_con_action: 其他用户做了相关concept的记录
'''
bg_ques_actions, bg_con_actions = self.obtain_bg_action(problem_id, item_num, stu_id)
# skill_str = self.my_txt[item_num][1]
# pick_rnn_index_actions = self.pick_index(skill_str, sq_id, stu_id)
ques_action_list += bg_ques_actions
con_action_list += bg_con_actions
'''parallel search'''
es = Elasticsearch(['localhost:9200'], timeout=30)
# TODO 11-23 0:29做到这里
ques_indexes, ques_score = self.bulk_seach(es, ques_action_list)
concepts_indexes, concepts_score = self.bulk_seach(es, con_action_list)
# diffques_indexes, diffques_resp = self.bulk_seach(es, diff_ques_action_list, type='diff')
for ir in range(0, len(org_x)):
record = org_x[ir]
problem_id, skills, item_num, response, sq_id, stu_id, interval_time, elapse_time = record
bg_index, bg_score = None, None
if ir < actual_len:
bg_index, bg_score = self.merge_ques_concept_list(ques_indexes[ir], ques_score[ir], concepts_indexes[ir], concepts_score[ir])
current_len = len(bg_index)
'''去掉多余的或者不足数的做填充'''
if current_len < self.max_sample_num:
bg_index += [0] * (self.max_sample_num - current_len)
bg_score += [-1000] * (self.max_sample_num - current_len)
elif current_len > self.max_sample_num:
bg_index = bg_index[0: self.max_sample_num]
bg_score = bg_score[0: self.max_sample_num]
else:
bg_index = [0] * self.max_sample_num
bg_score = [-1000] * self.max_sample_num
# my_diffques_indexes = [0] * self.max_sample_num
# my_diffques_resp = [0] * self.max_sample_num
# my_actual_diffques_tag = [0] * self.max_sample_num
this_x.append([problem_id, skills, item_num, response, bg_index, interval_time, elapse_time])
self.data_list.append([actual_len, [this_x, org_y]])
es.close()
def process(self):
'''index 要分数据集'''
# es = Elasticsearch(['127.0.0.1:9200'],ignore=[201, 200, 405, 502])
es = Elasticsearch(['localhost:9200'])
dt_name = self.path.split('/')[-2]
# e_index = [dt_name + '_train_search', dt_name + '_valid_search']
e_index = [dt_name + '_train_search']
# e_doctype = ['train_doc', 'valid_doc']
if self.split == 'train':
for ei in e_index:
if es.indices.exists(index=ei, ignore=400):
es.indices.delete(index=ei, ignore=[400, 404])
es.indices.create(index=ei)
self.prob_encode_dim = int(math.log(self.problem_number,2)) + 1
with open(self.path + 'history_' + self.split + '.pkl', 'rb') as fp:
histories = pickle.load(fp)
loader_len = len(histories.keys())
self.total_train_num = loader_len
self.stu_count = 0
log.info('loader length: {:d}'.format(loader_len))
# proc_count = 0
# all_backrgound_tensor_list = [torch.zeros(self.concept_num * 2)]
# response_tensor_list = [torch.tensor(0)]
# ques_seq_all_list = [torch.tensor([0] * self.seq_len)]
# resp_seq_all_list = [torch.tensor([-1] * self.seq_len)]
all_txt_list = [['', '']]
train_valid_id_gap = 0
if self.split == 'valid':
train_valid_id_gap = 500000
for k in tqdm.tqdm(histories.keys()):
stu_record = histories[k]
if stu_record[0] < 10:
continue
# self, stu_records, es, stu_id, actual_len, e_index
# background_noresp_list, response_list, \
# background_ques_seq_list, background_resp_seq_list,\
# background_txt_list,
background_txt_list, x_list, y_list = self.data_reader(stu_record[1], es, int(k) + train_valid_id_gap, stu_record[0], e_index)
# background_tensor_list, background_txt_list, x_list, y_list = self.data_reader(stu_record[1], es, k + train_valid_id_gap, stu_record[0], e_index)
# print('len x_list: ', len(x_list), 'len y_list: ', len(y_list))
if x_list != [] and y_list != []:
self.data_seq_list.append([stu_record[0], [x_list, y_list]])
self.stu_count += 1
# all_backrgound_tensor_list += background_noresp_list
# response_tensor_list += response_list
# ques_seq_all_list += background_ques_seq_list
# resp_seq_all_list += background_resp_seq_list
all_txt_list += background_txt_list
# proc_count += 1
# self.target_prepare(self.split, all_txt_list)
with open(self.path + self.split + '_data_len.pkl', 'wb') as fp:
pickle.dump(self.item_count, fp)
self.target_prepare( all_txt_list)
'''not multiprocess'''
for data in tqdm.tqdm(self.data_seq_list):
self.raw_to_final(data)
'''start multiprocess'''
print('data length after multprocess:', len(self.data_list))
# print('data list type before:', type(self.data_list))
self.data_list = list(self.data_list)
# print('data list type after:', type(self.data_list))
log.info('final length {:d}'.format(len(self.data_list)))