{{ message }}
-
Notifications
You must be signed in to change notification settings - Fork 53
Expand file tree
/
Copy pathutils.py
More file actions
980 lines (855 loc) · 33.5 KB
/
Copy pathutils.py
File metadata and controls
980 lines (855 loc) · 33.5 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
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
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
import json
from sqlnet.lib.dbengine import DBEngine
import numpy as np
from tqdm import tqdm
from sqlnet.model.sqlbert import SQLBert, gen_ensemble_query
import torch
from sqlnet.strPreprocess import *
from fuzzywuzzy import process
from fuzzywuzzy.utils import StringProcessor
import copy
from sqlnet.diff2 import extact_sort
from sqlnet.diff2 import digit_distance_search
from functools import lru_cache
import re
@lru_cache(None)
def my_scorer(t, c):
return (1 - abs(len(t) - len(c)) / max(len(t), len(c))) * process.default_scorer(t, c)
def my_process(s):
"""Process string by
-- removing all but letters and numbers
-- trim whitespace
-- force to lower case
if force_ascii == True, force convert to ascii"""
# Force into lowercase.
string_out = StringProcessor.to_lower_case(s)
# Remove leading and trailing whitespaces.
string_out = StringProcessor.strip(string_out)
return string_out
def pos_in_tokens(target_str, tokens, type = None, header = None):
if not tokens:
return -1, -1
tlen = len(target_str)
copy_target_str = target_str
q = ''.join(tokens).replace('##', '')
header = ''.join(header).replace('##','').replace('[UNK]','')
ngrams = []
for l in range(max(1, tlen - 25), min(tlen + 5, len(tokens))):
ngrams.append(l)
candidates = {}
unit_flag = 0
tback_flag = 0
unit_r = 0
if type =='real':
units = re.findall(r'[(（-](.*)',str(header))
if units:
unit = units[0]
#unit_keys = re.findall(r'[百千万亿]{1,}',str(header))
unit_keys = re.findall(r'百万|千万|万|百亿|千亿|万亿|亿', unit)
unit_other = re.findall(r'元|米|平|套|枚|册|张|辆|个|股|户|m²|亩|人', unit)
if unit_keys:
unit_flag = 1 #col中有[万|百万|千万|亿]单位，
unit_key = unit_keys[0]
v, unit_r = chinese_to_digits(unit_key)
#print('--unit--',unit_key, target_str)
target_str = target_str + unit_key
target_str = strPreProcess(target_str)
target_str = unit_convert(target_str)
#print('--target_str--', target_str, header)
elif unit_other:
unit_flag = 2 #col中有[元|米|平] 单位为个数
else:
unit_flag = 3 # 无单位，可能为个数，可能与ques中单位相同
for l in ngrams:
cur_idx = 0
while cur_idx <= len(tokens) - l:
cur_str = []
st, ed = cur_idx, cur_idx
i = st
while i != len(tokens) and len(cur_str) < l:
cur_tok = tokens[i]
cur_str.append(cur_tok)
i += 1
ed = i
cur_str = ''.join(cur_str)
if '##' in cur_str :
cur_str = cur_str.replace('##', '')
if '[UNK]' in cur_str :
cur_str = cur_str.replace('[UNK]', '')
if '-' in cur_str :
cur_str = cur_str.replace('-', '')
if unit_flag == 1:
if cur_str == target_str: #ques 无单位 默认为个数 target_str为unit_convert()后的
cur_str = str(int(cur_str)/unit_r)
unit_flag = 0 #target_str回到初始状态，
tback_flag = 0
# elif cur_str == copy_target_str: #ques 无单位 默认与target_str 相同
# tback_flag = 1 #标志位 默认与target_str 单位相同
else:
cur_str = unit_convert(cur_str)
elif unit_flag == 2:
cur_str = unit_convert(cur_str)
elif unit_flag == 3:
if unit_convert(cur_str) == target_str:
cur_str = unit_convert(cur_str)
if type == 'text':
for item in list(thesaurus_dic.keys()):
if item in cur_str:
cur_str_the = re.sub(item,thesaurus_dic[item],cur_str)
candidates[cur_str_the] = (st, ed)
candidates[cur_str] = (st, ed)
cur_idx += 1
# if tback_flag:
# target_str = copy_target_str
if list(candidates.keys()) is None or len(list(candidates.keys())) == 0:
# print('-----testnone----',target_str, tokens,ngrams)
return -1, -1
target_str = str(target_str).replace('-', '')
resultsf = process.extract(target_str, list(candidates.keys()), limit=10, processor=my_process, scorer=my_scorer)
results = extact_sort(target_str, list(candidates.keys()), limit=10)
if not results or not resultsf:
return -1, -1
dchosen, dcscore = results[0]
fchosen, fcscore = resultsf[0]
if fcscore > dcscore:
cscore = fcscore
chosen = fchosen
else:
cscore = dcscore
chosen = dchosen
if cscore !=100:
pass
#q = ''.join(tokens).replace('##','')
#score = '%d'%(cscore)
#with open("F:\\天池比赛\\nl2sql_test_20190618\\log3.txt", "a", encoding='utf-8') as fw:
#fw.write(str(chosen + '-----' + target_str + '---'+score +'--'+ q +'\n'+'\n'))
if cscore <= 50:
q = ''.join(tokens).replace('##','')
score = '%d'%(cscore)
#with open("F:\\天池比赛\\nl2sql_test_20190618\\log3.txt", "a", encoding='utf-8') as fw:
#fw.write(str(type + ' '+ header + ' ** '+chosen + '-----' + target_str + '---'+score +'--'+ q +'\n'+'\n'))
#return -1, -1
return candidates[chosen]
#return cscore, chosen
def justify_col_type(table):
def get_real_col_type(col_idx):
ret_type = 'text'
if 'rows' not in table.keys():
if 'types' in table.keys():
ret_type = table['types'][col_idx]
else:
na_set = {'None', 'none', 'N/A', '', 'nan', '-', '/', 'NAN'}
col_data = list(filter(lambda x: x not in na_set, [r[col_idx] for r in table['rows']]))
if col_data:
isreal = True
try:
_ = list(map(float, col_data))
except:
isreal = False
if isreal:
ret_type = 'real'
if ('ISBN' in table['header'][col_idx]) or ('号' in table['header'][col_idx]) or ('ID' in table['header'][col_idx]):
ret_type = 'text'
# if ret_type != table['types'][col_idx]:
# print(table['header'][col_idx], col_data)
return ret_type
if 'types' not in table.keys():
table['types'] = ['text'] * len(table['header'])
for i in range(len(table['header'])):
table['types'][i] = get_real_col_type(i)
return table
def load_data(sql_paths, table_paths, use_small=False):
if not isinstance(sql_paths, list):
sql_paths = (sql_paths,)
if not isinstance(table_paths, list):
table_paths = (table_paths,)
sql_data = []
table_data = {}
for SQL_PATH in sql_paths:
with open(SQL_PATH, encoding='utf-8') as inf:
for idx, line in enumerate(inf):
sql = json.loads(line.strip())
if use_small and idx >= 1000:
break
sql_data.append(sql)
print("Loaded %d data from %s" % (len(sql_data), SQL_PATH))
for TABLE_PATH in table_paths:
with open(TABLE_PATH, encoding='utf-8') as inf:
for line in inf:
tab = json.loads(line.strip())
tab = justify_col_type(tab)
table_data[tab[u'id']] = tab
print("Loaded %d data from %s" % (len(table_data), TABLE_PATH))
ret_sql_data = []
for sql in sql_data:
if sql[u'table_id'] in table_data:
ret_sql_data.append(sql)
return ret_sql_data, table_data
def load_dataset(data_dir='../data', toy=False, use_small=False, mode='train'):
print("Loading dataset")
import os.path as osp
data_dirs = {}
for name in ['train', 'val', 'test']:
data_dirs[name] = {}
data_dirs[name]['data'] = osp.join(data_dir, name, name+'.json')
data_dirs[name]['tables'] = osp.join(data_dir, name, name + '.tables.json')
data_dirs[name]['db'] = osp.join(data_dir, name, name + '.db')
dev_sql, dev_table = load_data(data_dirs['val']['data'], data_dirs['val']['tables'], use_small=use_small)
dev_db = data_dirs['val']['db']
if mode == 'train':
train_sql, train_table = load_data(data_dirs['train']['data'], data_dirs['train']['tables'], use_small=use_small)
train_db = data_dirs['train']['db']
return train_sql, train_table, train_db, dev_sql, dev_table, dev_db
elif mode == 'test':
test_sql, test_table = load_data(data_dirs['test']['data'], data_dirs['test']['tables'], use_small=use_small)
test_db = data_dirs['test']['db']
return dev_sql, dev_table, dev_db, test_sql, test_table, test_db
def to_batch_seq(sql_data, table_data, idxes, st, ed, tokenizer=None, ret_vis_data=False):
q_seq = [] #问题内容
col_seq = [] #一张表所有表头
col_num = [] #表头数量
ans_seq = [] #sql的答案列表
gt_cond_seq = [] #条件列--列号，类型，值
vis_seq = [] #（）tuple，问题和对应表所有表头
sel_num_seq = [] #sel列的数量
header_type = [] #对应表所有列的数据类型
for i in range(st, ed):
sql = sql_data[idxes[i]]
sel_num = len(sql['sql']['sel'])
sel_num_seq.append(sel_num)
conds_num = len(sql['sql']['conds'])
if tokenizer:
q = tokenizer.tokenize(strPreProcess(sql['question']))
col = [tokenizer.tokenize(header) for header in table_data[sql['table_id']]['header']]
else:
q = [char for char in sql['question']]
col = [[char for char in header] for header in table_data[sql['table_id']]['header']]
q_seq.append(q)
col_seq.append(col)
col_num.append(len(table_data[sql['table_id']]['header']))
ans_seq.append(
(
len(sql['sql']['agg']),
sql['sql']['sel'],
sql['sql']['agg'],
conds_num,
tuple(x[0] for x in sql['sql']['conds']),
tuple(x[1] for x in sql['sql']['conds']),
sql['sql']['cond_conn_op'],
))
gt_cond_seq.append(sql['sql']['conds'])
vis_seq.append((sql['question'], table_data[sql['table_id']]['header']))
header_type.append(table_data[sql['table_id']]['types'])
# q_seq: char-based sequence of question
# gt_sel_num: number of selected columns and aggregation functions
# col_seq: char-based column name
# col_num: number of headers in one table
# ans_seq: (sel, number of conds, sel list in conds, op list in conds)
# gt_cond_seq: ground truth of conds
if ret_vis_data:
return q_seq, sel_num_seq, col_seq, col_num, ans_seq, gt_cond_seq, header_type, vis_seq
else:
return q_seq, sel_num_seq, col_seq, col_num, ans_seq, gt_cond_seq, header_type
def pad_batch_seqs(seqs, pad=None, max_len=None):
if not max_len:
max_len = max([len(s) for s in seqs])
if not pad:
pad = 0
for i in range(len(seqs)):
if len(seqs[i]) > max_len:
seqs[i] = seqs[i][:max_len]
else:
seqs[i].extend([pad] * (max_len - len(seqs[i])))
return seqs
def gen_batch_bert_seq(tokenizer, q_seq, col_seq, header_type, max_len=230):
input_seq = [] # 输入编号
q_mask = [] # NL mask
sel_col_mask = [] # columns mask
sel_col_index = [] # columns starting index
where_col_mask = []
where_col_index = []
token_type_ids = [] # sentence A/B
attention_mask = [] # length mask
col_end_index = []
q_lens = []
sel_col_nums = []
where_col_nums = []
batch_size = len(q_seq)
for i in range(batch_size):
text_a = ['[CLS]'] + q_seq[i] + ['[SEP]']
text_b = []
for col_idx, col in enumerate(col_seq[i]):
new_col = []
if header_type[i][col_idx] == 'text':
type_token1 = '[unused1]'
type_token2 = '[unused4]'
type_token3 = '[unused7]'
elif header_type[i][col_idx] == 'real':
type_token1 = '[unused2]'
type_token2 = '[unused5]'
type_token3 = '[unused8]'
else:
type_token1 = '[unused3]'
type_token2 = '[unused6]'
type_token3 = '[unused9]'
new_col.extend(col)
new_col.append(type_token2) # type特征 用来分类第一次作为条件
new_col.append(type_token3) # type特征 用来分类第二次作为条件
# TODO: 可以再加入新的标签来支持更多的列
new_col.append(type_token1) # type特征 用来分类sel, 同时分隔列名
if len(text_a) + len(text_b) + len(new_col) >= max_len:
break
text_b.extend(new_col)
text_b.append('[SEP]')
inp_seq = text_a + text_b
input_seq.append(inp_seq)
q_mask.append([1] * (len(text_a) - 2))
q_lens.append(len(text_a) - 2)
token_type_ids.append([0] * len(text_a) + [1] * len(text_b))
attention_mask.append([1] * len(inp_seq))
sel_col = []
where_col = []
col_ends = []
for i in range(len(text_a) - 1, len(inp_seq)):
if inp_seq[i] in ['[unused4]', '[unused5]', '[unused6]', '[unused7]', '[unused8]', '[unused9]']:
where_col.append(i)
if inp_seq[i] in ['[unused1]', '[unused2]', '[unused3]']:
sel_col.append(i)
col_ends.append(i)
sel_col_mask.append([1] * len(sel_col))
where_col_mask.append([1] * len(where_col))
sel_col_nums.append(len(sel_col))
where_col_nums.append(len(where_col))
sel_col_index.append(sel_col)
where_col_index.append(where_col)
col_end_index.append(col_ends)
#规范输入为同一长度，pad = ’[pad]‘ | 0
input_seq = pad_batch_seqs(input_seq, '[PAD]')
input_seq = [tokenizer.convert_tokens_to_ids(sq) for sq in input_seq] #字符token转化为词汇表里的编码id
q_mask = pad_batch_seqs(q_mask)
sel_col_mask = pad_batch_seqs(sel_col_mask)
sel_col_index = pad_batch_seqs(sel_col_index)
where_col_mask = pad_batch_seqs(where_col_mask)
where_col_index = pad_batch_seqs(where_col_index)
token_type_ids = pad_batch_seqs(token_type_ids)
attention_mask = pad_batch_seqs(attention_mask)
col_end_index = pad_batch_seqs(col_end_index)
return (input_seq, q_mask, sel_col_mask, sel_col_index, where_col_mask, where_col_index, col_end_index,
token_type_ids, attention_mask), q_lens, sel_col_nums, where_col_nums
def to_batch_seq_test(sql_data, table_data, idxes, st, ed, tokenizer=None):
q_seq = []
col_seq = []
col_num = []
raw_seq = []
table_ids = []
header_type = []
for i in range(st, ed):
sql = sql_data[idxes[i]]
if tokenizer:
q = tokenizer.tokenize(strPreProcess(sql['question']))
col = [tokenizer.tokenize(header) for header in table_data[sql['table_id']]['header']]
else:
q = [char for char in sql['question']]
col = [[char for char in header] for header in table_data[sql['table_id']]['header']]
q_seq.append(q)
col_seq.append(col)
col_num.append(len(table_data[sql['table_id']]['header']))
raw_seq.append(sql['question'])
table_ids.append(sql_data[idxes[i]]['table_id'])
header_type.append(table_data[sql['table_id']]['types'])
return q_seq, col_seq, col_num, raw_seq, table_ids, header_type
def generate_gt_where_seq_test(q, gt_cond_seq):
ret_seq = []
for cur_q, ans in zip(q, gt_cond_seq):
temp_q = u"".join(cur_q)
cur_q = [u'<BEG>'] + cur_q + [u'<END>']
record = []
record_cond = []
for cond in ans:
if cond[2] not in temp_q:
record.append((False, cond[2]))
else:
record.append((True, cond[2]))
for idx, item in enumerate(record):
temp_ret_seq = []
if item[0]:
temp_ret_seq.append(0)
temp_ret_seq.extend(list(range(temp_q.index(item[1]) + 1, temp_q.index(item[1]) + len(item[1]) + 1)))
temp_ret_seq.append(len(cur_q) - 1)
else:
temp_ret_seq.append([0, len(cur_q) - 1])
record_cond.append(temp_ret_seq)
ret_seq.append(record_cond)
return ret_seq
def gen_bert_labels(q_seq, q_lens, sel_col_nums, where_col_nums, ans_seq, gt_cond_seq, header_type, col_seq):
q_max_len = max(q_lens)
sel_col_max_len = max(sel_col_nums)
where_col_max_len = max(where_col_nums) #2col
# labels init
where_conn_label = np.array([x[6] for x in ans_seq]) # (None, )
sel_num_label = np.array([0 for _ in ans_seq]) # (None, )
where_num_label = np.array([0 for _ in ans_seq]) # (None, )
sel_col_label = np.array([[0] * sel_col_max_len for _ in ans_seq], dtype=np.float) # (None, col_max_len)
sel_agg_label = np.array([[-1] * sel_col_max_len for _ in ans_seq]) # (None, col_max_len)
where_col_label = np.array([[0] * where_col_max_len for _ in ans_seq], dtype=np.float) # (None, 2col_max_len)
where_op_label = np.array([[-1] * where_col_max_len for _ in ans_seq]) # (None, 2col_max_len)
where_start_label = np.array([[-1] * where_col_max_len for _ in ans_seq])
where_end_label = np.array([[-1] * where_col_max_len for _ in ans_seq])
for b in range(len(gt_cond_seq)): # batch_size
num_conds = len(gt_cond_seq[b]) # 条件数量
if num_conds == 0:
where_col_label[b] = 1.0 / sel_col_nums[b] # 分散
mass = 0
else:
mass = 1 / num_conds
col_cond_count = {}
for cond in gt_cond_seq[b]:
if cond[0] >= sel_col_nums[b]:
continue
if cond[0] in col_cond_count:
col_cond_count[cond[0]] += 1
else:
col_cond_count[cond[0]] = 0
col_idx = 2 * cond[0] + col_cond_count[cond[0]] % 2
where_op_label[b][col_idx] = cond[1]
where_col_label[b][col_idx] += mass
s, e = pos_in_tokens(cond[2], q_seq[b], header_type[b][cond[0]], col_seq[b][cond[0]])
if s >= 0:
s = min(s, q_lens[b] - 1)
e = min(e - 1, q_lens[b] - 1)
where_start_label[b][col_idx] = s
where_end_label[b][col_idx] = e
if num_conds > 0:
where_num_label[b] = (where_col_label[b] > 0).sum()
for b in range(len(ans_seq)):
_sel = ans_seq[b][1]
_agg = ans_seq[b][2]
sel, agg = [], []
for i in range(len(_sel)):
if _sel[i] < sel_col_nums[b]:
sel.append(_sel[i])
agg.append(_agg[i])
sel_num_label[b] = len(sel)
mass = 1 / sel_num_label[b]
if sel_num_label[b] == 0:
mass = 1 / sel_col_nums[b]
sel_col_label[b][sel] = mass
sel_agg_label[b][sel] = agg
return where_conn_label, sel_num_label, where_num_label, sel_col_label, sel_agg_label, \
where_col_label, where_op_label, where_start_label, where_end_label
def to_batch_query(sql_data, idxes, st, ed):
query_gt = []
table_ids = []
for i in range(st, ed):
sql_data[idxes[i]]['sql']['conds'] = sql_data[idxes[i]]['sql']['conds']
query_gt.append(sql_data[idxes[i]]['sql'])
table_ids.append(sql_data[idxes[i]]['table_id'])
return query_gt, table_ids
def epoch_train(model, optimizer, batch_size, sql_data, table_data, tokenizer=None):
model.train()
perm = np.random.permutation(len(sql_data))
cum_loss = 0.0
for st in tqdm(range(len(sql_data) // batch_size + 1)):
if st * batch_size == len(perm):
break
ed = (st + 1) * batch_size if (st + 1) * batch_size < len(perm) else len(perm)
st = st * batch_size
if isinstance(model, SQLBert):
# bert training
q_seq, gt_sel_num, col_seq, col_num, ans_seq, gt_cond_seq, header_type = to_batch_seq(sql_data, table_data,
perm, st, ed,
tokenizer=tokenizer)
bert_inputs, q_lens, sel_col_nums, where_col_nums = gen_batch_bert_seq(tokenizer, q_seq, col_seq,
header_type)
logits = model.forward(bert_inputs) # condconn_logits, condop_logits, sel_agg_logits, q2col_logits
# gen label
labels = gen_bert_labels(q_seq, q_lens, sel_col_nums, where_col_nums, ans_seq, gt_cond_seq, header_type, col_seq)
# q_seq (12,q_lens) 问题内容
# q_lens (12,1)问题长度
# sel_col_nums (12,1) col 长度
# where_col_nums (12,1)2col长度
# ans_seq [(1, [6], [0], 1, (1,), (2,), 0),] len(agg),sel_col,agg,len(con),con_col,con_type,con_op
# gt_cond_seq (12,3)条件列--列号，类型，值
# compute loss
loss = model.loss(logits, labels, q_lens, sel_col_nums)
else:
q_seq, gt_sel_num, col_seq, col_num, ans_seq, gt_cond_seq, header_type = to_batch_seq(sql_data, table_data,
perm, st, ed)
# q_seq: char-based sequence of question
# gt_sel_num: number of selected columns and aggregation functions
# col_seq: char-based column name
# col_num: number of headers in one table
# ans_seq: (sel, number of conds, sel list in conds, op list in conds)
# gt_cond_seq: ground truth of conds
gt_where_seq = generate_gt_where_seq_test(q_seq, gt_cond_seq)
gt_sel_seq = [x[1] for x in ans_seq]
score = model.forward(q_seq, col_seq, col_num, gt_where=gt_where_seq, gt_cond=gt_cond_seq,
gt_sel=gt_sel_seq,
gt_sel_num=gt_sel_num)
# sel_num_score, sel_col_score, sel_agg_score, cond_score, cond_rela_score
# compute loss
loss = model.loss(score, ans_seq, gt_where_seq)
optimizer.zero_grad()
loss.backward()
optimizer.step()
cum_loss += loss.data.cpu().numpy() * (ed - st)
return cum_loss / len(sql_data)
def predict_test(model, batch_size, sql_data, table_data, output_path, tokenizer=None):
model.eval()
perm = list(range(len(sql_data)))
fw = open(output_path, 'w', encoding='utf-8')
for st in tqdm(range(len(sql_data) // batch_size + 1)):
if st * batch_size == len(perm):
break
ed = (st + 1) * batch_size if (st + 1) * batch_size < len(perm) else len(perm)
st = st * batch_size
with torch.no_grad():
if isinstance(model, SQLBert):
q_seq, col_seq, col_num, raw_q_seq, table_ids, header_type = to_batch_seq_test(sql_data, table_data,
perm, st, ed,
tokenizer=tokenizer)
bert_inputs, q_lens, sel_col_nums, where_col_nums = gen_batch_bert_seq(tokenizer, q_seq, col_seq,
header_type)
score = model.forward(bert_inputs, return_logits=False)
sql_preds = model.gen_query(score, q_seq, col_seq, sql_data, table_data, perm, st, ed)
else:
q_seq, col_seq, col_num, raw_q_seq, table_ids, header_type = to_batch_seq_test(sql_data, table_data,
perm, st, ed)
score = model.forward(q_seq, col_seq, col_num)
sql_preds = model.gen_query(score, q_seq, col_seq, raw_q_seq)
sql_preds = post_process(sql_preds, sql_data, table_data, perm, st, ed)
for sql_pred in sql_preds:
sql_pred = eval(str(sql_pred))
fw.writelines(json.dumps(sql_pred, ensure_ascii=False) + '\n')
# fw.writelines(json.dumps(sql_pred,ensure_ascii=False).encode('utf-8')+'\n')
fw.close()
def epoch_acc(model, batch_size, sql_data, table_data, db_path, tokenizer=None):
engine = DBEngine(db_path)
model.eval()
perm = list(range(len(sql_data)))
badcase = 0
one_acc_num, tot_acc_num, ex_acc_num = 0.0, 0.0, 0.0
total_error_cases = []
total_gt_cases = []
for st in tqdm(range(len(sql_data) // batch_size + 1)):
if st * batch_size == len(perm):
break
ed = (st + 1) * batch_size if (st + 1) * batch_size < len(perm) else len(perm)
st = st * batch_size
q_seq, gt_sel_num, col_seq, col_num, ans_seq, gt_cond_seq, header_type, raw_data = \
to_batch_seq(sql_data, table_data, perm, st, ed, tokenizer=tokenizer, ret_vis_data=True)
query_gt, table_ids = to_batch_query(sql_data, perm, st, ed)
# query_gt: ground truth of sql, data['sql'], containing sel, agg, conds:{sel, op, value}
raw_q_seq = [x[0] for x in raw_data] # original question
# try:
with torch.no_grad():
if isinstance(model, SQLBert):
bert_inputs, q_lens, sel_col_nums, where_col_nums = gen_batch_bert_seq(tokenizer, q_seq, col_seq,
header_type)
score = model.forward(bert_inputs, return_logits=False)
pred_queries = model.gen_query(score, q_seq, col_seq, sql_data, table_data, perm, st, ed)
else:
score = model.forward(q_seq, col_seq, col_num)
# generate predicted format
pred_queries = model.gen_query(score, q_seq, col_seq, raw_q_seq)
pred_queries_post = copy.deepcopy(pred_queries)
pred_queries_post = post_process(pred_queries_post, sql_data, table_data, perm, st, ed)
one_err, tot_err, error_idxs = check_acc(raw_data, pred_queries_post, query_gt)
error_cases, gt_cases = gen_batch_error_cases(error_idxs, query_gt, pred_queries_post, pred_queries, raw_data)
total_error_cases.extend(error_cases)
total_gt_cases.extend(gt_cases)
# except:
# badcase += 1
# print('badcase', badcase)
# continue
one_acc_num += (ed - st - one_err)
tot_acc_num += (ed - st - tot_err)
# Execution Accuracy
for sql_gt, sql_pred, tid in zip(query_gt, pred_queries_post, table_ids):
ret_gt = engine.execute(tid, sql_gt['sel'], sql_gt['agg'], sql_gt['conds'], sql_gt['cond_conn_op'])
try:
ret_pred = engine.execute(tid, sql_pred['sel'], sql_pred['agg'], sql_pred['conds'],
sql_pred['cond_conn_op'])
except:
ret_pred = None
ex_acc_num += (ret_gt == ret_pred)
save_error_case(total_error_cases, total_gt_cases)
return one_acc_num / len(sql_data), tot_acc_num / len(sql_data), ex_acc_num / len(sql_data)
def epoch_ensemble(model, batch_scores, batch_size, sql_data, table_data, tokenizer=None):
model.eval()
perm = list(range(len(sql_data)))
if not batch_scores:
gen_new = True
batch_scores = []
else:
gen_new = False
for st in tqdm(range(len(sql_data) // batch_size + 1)):
if st * batch_size == len(perm):
break
ed = (st + 1) * batch_size if (st + 1) * batch_size < len(perm) else len(perm)
batch_id = st
st = st * batch_size
q_seq, gt_sel_num, col_seq, col_num, ans_seq, gt_cond_seq, header_type, raw_data = \
to_batch_seq(sql_data, table_data, perm, st, ed, tokenizer=tokenizer, ret_vis_data=True)
with torch.no_grad():
bert_inputs, q_lens, sel_col_nums, where_col_nums = gen_batch_bert_seq(tokenizer, q_seq, col_seq,
header_type)
score = model.forward(bert_inputs, return_logits=False)
score = model.gen_ensemble(score, q_seq, col_seq, sql_data, table_data, perm, st, ed)
if not gen_new:
for i in range(7):
batch_scores[batch_id][i] += score[i]
for item in range(len(score[7])):
for col in range(len(score[7][item])):
batch_scores[batch_id][7][item][col] += score[7][item][col]
else:
batch_scores.append(score)
return batch_scores
def epoch_ensemble_test(model, batch_scores, batch_size, sql_data, table_data, tokenizer=None):
model.eval()
perm = list(range(len(sql_data)))
if not batch_scores:
gen_new = True
batch_scores = []
else:
gen_new = False
for st in tqdm(range(len(sql_data) // batch_size + 1)):
if st * batch_size == len(perm):
break
ed = (st + 1) * batch_size if (st + 1) * batch_size < len(perm) else len(perm)
batch_id = st
st = st * batch_size
q_seq, col_seq, col_num, raw_q_seq, table_ids, header_type = to_batch_seq_test(sql_data, table_data,
perm, st, ed,
tokenizer=tokenizer)
with torch.no_grad():
bert_inputs, q_lens, sel_col_nums, where_col_nums = gen_batch_bert_seq(tokenizer, q_seq, col_seq,
header_type)
score = model.forward(bert_inputs, return_logits=False)
score = model.gen_ensemble(score, q_seq, col_seq, sql_data, table_data, perm, st, ed)
if not gen_new:
for i in range(7):
batch_scores[batch_id][i] += score[i]
for item in range(len(score[7])):
for col in range(len(score[7][item])):
batch_scores[batch_id][7][item][col] += score[7][item][col]
else:
batch_scores.append(score)
return batch_scores
def ensemble_predict(batch_scores, batch_size, sql_data, table_data, output_path):
perm = list(range(len(sql_data)))
fw = open(output_path, 'w', encoding='utf-8')
for st in tqdm(range(len(sql_data) // batch_size + 1)):
if st * batch_size == len(perm):
break
ed = (st + 1) * batch_size if (st + 1) * batch_size < len(perm) else len(perm)
batch_idx = st
st = st * batch_size
score = batch_scores[batch_idx]
sql_preds = gen_ensemble_query(score, sql_data, table_data, perm, st, ed)
sql_preds = post_process(sql_preds, sql_data, table_data, perm, st, ed)
for sql_pred in sql_preds:
sql_pred = eval(str(sql_pred))
fw.writelines(json.dumps(sql_pred, ensure_ascii=False) + '\n')
# fw.writelines(json.dumps(sql_pred,ensure_ascii=False).encode('utf-8')+'\n')
fw.close()
def ensemble_acc(batch_scores, batch_size, sql_data, table_data, db_path):
engine = DBEngine(db_path)
perm = list(range(len(sql_data)))
badcase = 0
one_acc_num, tot_acc_num, ex_acc_num = 0.0, 0.0, 0.0
total_error_cases = []
total_gt_cases = []
for st in tqdm(range(len(sql_data) // batch_size + 1)):
if st * batch_size == len(perm):
break
ed = (st + 1) * batch_size if (st + 1) * batch_size < len(perm) else len(perm)
batch_idx = st
st = st * batch_size
query_gt, table_ids = to_batch_query(sql_data, perm, st, ed)
raw_data = [(sql_data[i]['question'], table_data[sql_data[i]['table_id']]['header']) for i in perm[st: ed]]
score = batch_scores[batch_idx]
pred_queries = gen_ensemble_query(score, sql_data, table_data, perm, st, ed)
pred_queries_post = copy.deepcopy(pred_queries)
pred_queries_post = post_process(pred_queries_post, sql_data, table_data, perm, st, ed)
one_err, tot_err, error_idxs = check_acc(None, pred_queries_post, query_gt)
error_cases, gt_cases = gen_batch_error_cases(error_idxs, query_gt, pred_queries_post, pred_queries, raw_data)
total_error_cases.extend(error_cases)
total_gt_cases.extend(gt_cases)
# except:
# badcase += 1
# print('badcase', badcase)
# continue
one_acc_num += (ed - st - one_err)
tot_acc_num += (ed - st - tot_err)
# Execution Accuracy
for sql_gt, sql_pred, tid in zip(query_gt, pred_queries_post, table_ids):
ret_gt = engine.execute(tid, sql_gt['sel'], sql_gt['agg'], sql_gt['conds'], sql_gt['cond_conn_op'])
try:
ret_pred = engine.execute(tid, sql_pred['sel'], sql_pred['agg'], sql_pred['conds'],
sql_pred['cond_conn_op'])
except:
ret_pred = None
ex_acc_num += (ret_gt == ret_pred)
save_error_case(total_error_cases, total_gt_cases)
return one_acc_num / len(sql_data), tot_acc_num / len(sql_data), ex_acc_num / len(sql_data)
def post_process(pred, sql_data, table_data, perm, st, ed):
for i in range(st, ed):
sql = sql_data[perm[i]]
table = table_data[sql['table_id']]
for c in range(len(pred[i - st]['conds'])):
col_idx = pred[i - st]['conds'][c][0]
col_val = pred[i - st]['conds'][c][2]
if col_idx > len(table['header']) or col_val == "" or table['types'][col_idx] == 'real':
continue
col_data = []
for r in table['rows']:
if col_idx < len(r) and r[col_idx] not in {'None', 'none'}:#, 'N/A', '-', '/', ''}:
col_data.append(r[col_idx])
if not col_data:
continue
is_real = True
try:
_ = list(map(float, col_data))
except:
is_real = False
if is_real:
continue
score_c = 0
for item in list(thesaurus_dic.keys()):
if item in col_val:
col_val_the = re.sub(item, thesaurus_dic[item], col_val)
match_c, score_c = process.extractOne(col_val_the, col_data, processor=my_process)
match, score = process.extractOne(col_val, col_data, processor=my_process)
if score_c > score:
match = match_c
score = score_c
if score < 30:
continue
pred[i - st]['conds'][c][2] = match
return pred
def check_acc(vis_info, pred_queries, gt_queries):
tot_err = sel_num_err = agg_err = sel_err = 0.0
cond_num_err = cond_col_err = cond_op_err = cond_val_err = cond_rela_err = 0.0
bad_sample_idxs = []
for b, (pred_qry, gt_qry) in enumerate(zip(pred_queries, gt_queries)):
good = True
sel_pred, agg_pred, where_rela_pred = pred_qry['sel'], pred_qry['agg'], pred_qry['cond_conn_op']
sel_gt, agg_gt, where_rela_gt = gt_qry['sel'], gt_qry['agg'], gt_qry['cond_conn_op']
if where_rela_gt != where_rela_pred:
good = False
cond_rela_err += 1
if len(sel_pred) != len(sel_gt):
good = False
sel_num_err += 1
pred_sel_dict = {k: v for k, v in zip(list(sel_pred), list(agg_pred))}
gt_sel_dict = {k: v for k, v in zip(sel_gt, agg_gt)}
if set(sel_pred) != set(sel_gt):
good = False
sel_err += 1
agg_pred = [pred_sel_dict[x] for x in sorted(pred_sel_dict.keys())]
agg_gt = [gt_sel_dict[x] for x in sorted(gt_sel_dict.keys())]
if agg_pred != agg_gt:
good = False
agg_err += 1
cond_pred = list(sorted(pred_qry['conds'], key=lambda x: (x[0], x[1], x[2])))
cond_gt = list(sorted(gt_qry['conds'], key=lambda x: (x[0], x[1], x[2])))
if len(cond_pred) != len(cond_gt):
good = False
cond_num_err += 1
else:
cond_op_pred, cond_op_gt = {}, {}
cond_val_pred, cond_val_gt = {}, {}
for p, g in zip(cond_pred, cond_gt):
cond_op_pred[p[0]] = p[1]
cond_val_pred[p[0]] = p[2]
cond_op_gt[g[0]] = g[1]
cond_val_gt[g[0]] = g[2]
if set(cond_op_pred.keys()) != set(cond_op_gt.keys()):
cond_col_err += 1
good = False
where_op_pred = [cond_op_pred[x] for x in sorted(cond_op_pred.keys())]
where_op_gt = [cond_op_gt[x] for x in sorted(cond_op_gt.keys())]
if where_op_pred != where_op_gt:
cond_op_err += 1
good = False
where_val_pred = [cond_val_pred[x] for x in sorted(cond_val_pred.keys())]
where_val_gt = [cond_val_gt[x] for x in sorted(cond_val_gt.keys())]
if where_val_pred != where_val_gt:
cond_val_err += 1
good = False
if not good:
tot_err += 1
bad_sample_idxs.append(b)
return np.array(
(sel_num_err, sel_err, agg_err, cond_num_err, cond_col_err, cond_op_err, cond_val_err, cond_rela_err)), tot_err, bad_sample_idxs
def gen_batch_error_cases(error_idxs, query_gt, pred_queries_post, pred_queries, raw_data):
error_cases = []
gt_cases = []
for idx in error_idxs:
single_error = {}
single_gt = {}
single_error['q_raw'] = raw_data[idx][0]
single_error['q_normed'] = strPreProcess(single_error['q_raw'])
single_gt['q_raw'] = raw_data[idx][0]
single_gt['q_normed'] = strPreProcess(single_error['q_raw'])
single_error['cols'] = raw_data[idx][1]
single_gt['cols'] = raw_data[idx][1]
single_gt['sql'] = query_gt[idx]
single_error['sql'] = copy.deepcopy(pred_queries_post[idx])
for i in range(len(single_error['sql']['conds'])):
single_error['sql']['conds'][i].append(pred_queries[idx]['conds'][i][2])
error_cases.append(single_error)
gt_cases.append(single_gt)
return error_cases, gt_cases
def save_error_case(error_case, gt_cases, dir='./log/'):
import os.path as osp
error_fn = osp.join(dir, 'error_cases.json')
gt_fn = osp.join(dir, 'gt_cases.json')
with open(error_fn, "w", encoding='utf-8') as f:
json.dump(error_case, f, ensure_ascii=False, indent=4)
with open(gt_fn, "w", encoding='utf-8') as f:
json.dump(gt_cases, f, ensure_ascii=False, indent=4)
def load_word_emb(file_name):
print('Loading word embedding from %s' % file_name)
f = open(file_name)
ret = json.load(f)
f.close()
# ret = {}
# with open(file_name, encoding='latin') as inf:
# ret = json.load(inf)
# for idx, line in enumerate(inf):
# info = line.strip().split(' ')
# if info[0].lower() not in ret:
# ret[info[0]] = np.array([float(x) for x in info[1:]])
return ret
thesaurus_dic = {
'没有要求': '不限',
'达标': '合格',
'不': '否',
'鄂': '湖北',
'豫': '河南',
'皖': '安徽',
'冀': '河北',
'inter': '因特尔',
'samsung': '三星',
'芒果TV': '湖南卫视',
'湖南台': '芒果TV',
'企鹅公司': '腾讯',
'鹅厂': '腾讯',
'宁': '南京',
'Youku': '优酷',
'荔枝台': '江苏卫视',
'周一': '星期一',
'周二': '星期二',
'周三': '星期三',
'周四': '星期四',
'周五': '星期五',
'周六': '星期六',
'周日': '星期天',
'周天': '星期天',
'电视剧频道': '中央台八套',
'沪': '上海',
'闽': '福建',
'增持': '买入'
}
You can’t perform that action at this time.