-
Notifications
You must be signed in to change notification settings - Fork 3
/
Copy pathswa-k-fold.py
1027 lines (859 loc) · 50.4 KB
/
swa-k-fold.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
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
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
# import os and define graphic card
import os
os.environ["OMP_NUM_THREADS"] = "1"
# import common libraries
import gc
import random
import argparse
import pandas as pd
import numpy as np
from functools import partial
# import pytorch related libraries
import torch
import torch.optim as optim
from torch.optim import lr_scheduler
from torch.optim.optimizer import Optimizer
from torch.utils.data import TensorDataset, DataLoader,Dataset
from torch.utils.data.sampler import RandomSampler, SequentialSampler
from torch.optim.lr_scheduler import StepLR, ReduceLROnPlateau, CosineAnnealingLR, _LRScheduler
from tensorboardX import SummaryWriter
from pytorch_pretrained_bert.optimization import BertAdam
from transformers import get_linear_schedule_with_warmup
# import apex for mix precision training
from apex import amp
from apex.parallel import DistributedDataParallel as DDP
from apex.optimizers import FusedAdam
# import dataset class
from dataset.dataset import *
# import utils
from utils.ranger import *
from utils.lrs_scheduler import *
from utils.loss_function import *
from utils.metric import *
from utils.file import *
# import model
from model.model_bert import *
############################################################################## Define Argument
parser = argparse.ArgumentParser(description="arg parser")
parser.add_argument("--train_data_folder", type=str, default="/media/jionie/my_disk/Kaggle/Google_Quest_Answer/input/google-quest-challenge/", \
required=False, help="specify the folder for training data")
parser.add_argument('--model_type', type=str, default="bert", \
required=False, help='specify the model_type for BertTokenizer and Net')
parser.add_argument('--model_name', type=str, default="bert-base-uncased", \
required=False, help='specify the model_name for BertTokenizer and Net')
parser.add_argument('--content', type=str, default="Question", \
required=False, help='specify the content for token')
parser.add_argument("--max_len", type=int, default=512, required=False, help="specify the max_len of tokens")
parser.add_argument('--hidden_layers', type=list, default=[-3, -4, -5, -6, -7], \
required=False, help='specify the hidden_layers for Loss')
parser.add_argument('--optimizer', type=str, default='BertAdam', required=False, help='specify the optimizer')
parser.add_argument("--lr_scheduler", type=str, default='WarmupLinearSchedule', required=False, help="specify the lr scheduler")
parser.add_argument("--warmup_proportion", type=float, default=0.0, required=False, \
help="Proportion of training to perform linear learning rate warmup for. " "E.g., 0.1 = 10%% of training.")
parser.add_argument("--lr", type=float, default=3e-5, required=False, help="specify the initial learning rate for training")
parser.add_argument("--batch_size", type=int, default=8, required=False, help="specify the batch size for training")
parser.add_argument("--valid_batch_size", type=int, default=32, required=False, help="specify the batch size for validating")
parser.add_argument("--num_epoch", type=int, default=3, required=False, help="specify the total epoch")
parser.add_argument("--accumulation_steps", type=int, default=4, required=False, help="specify the accumulation steps")
parser.add_argument('--num_workers', type=int, default=2, \
required=False, help='specify the num_workers for testing dataloader')
parser.add_argument("--start_epoch", type=int, default=0, required=False, help="specify the start epoch for continue training")
parser.add_argument("--checkpoint_folder", type=str, default="/media/jionie/my_disk/Kaggle/Google_Quest_Answer/model", \
required=False, help="specify the folder for checkpoint")
parser.add_argument('--extra_token', action='store_true', default=False, help='whether to use extra token for extra tasks')
parser.add_argument('--load_pretrain', action='store_true', default=True, help='whether to load pretrain model')
parser.add_argument('--fold', type=int, default=0, required=True, help="specify the fold for training")
parser.add_argument('--seed', type=int, default=42, required=True, help="specify the seed for training")
parser.add_argument('--n_splits', type=int, default=5, required=True, help="specify the n_splits for training")
parser.add_argument('--split', type=str, default="GroupKfold", required=True, help="specify the splitting dataset way")
parser.add_argument('--loss', type=str, default="mse", required=True, help="specify the loss for training")
parser.add_argument('--augment', action='store_true', help="specify whether augmentation for training")
############################################################################## Define Constant
NUM_CATEGORY_CLASS=5
NUM_HOST_CLASS=64
AUXILIARY_WEIGHTs = [1, 0.05, 0.05]
DECAY_FACTOR = 0.95
MIN_LR = 1e-7
UNBALANCE_WEIGIHT = [1, 1, 1, 1, 1, 1, \
1, 1, 1, 2, 1, 1, \
2, 2, 2, 2, 1, 1, \
1, 1, 1, 1, 1, 1, \
1, 1, 1, 1, 1, 1]
QUESTION_UNBALANCE_WEIGIHT = [1, 1, 1, 1, 1, 1, \
1, 1, 1, 2, 1, 1, \
2, 2, 2, 2, 1, 1, \
1, 1, 1]
ANSWER_UNBALANCE_WEIGIHT = [1, 1, 1, 1, 1, 1, 1, 1, 1]
############################################################################## seed All
def seed_everything(seed=42):
random.seed(seed)
os.environ['PYHTONHASHseed'] = str(seed)
np.random.seed(seed)
torch.manual_seed(seed)
torch.cuda.manual_seed(seed)
torch.backends.cudnn.benchmark = False ##uses the inbuilt cudnn auto-tuner to find the fastest convolution algorithms. -
torch.backends.cudnn.enabled = True
torch.backends.cudnn.deterministic = True
############################################################################## define function for training
def training(
tokenizer,
content,
n_splits,
fold,
train_data_loader,
val_data_loader,
model_type,
model_name,
hidden_layers,
optimizer_name,
lr_scheduler_name,
lr,
warmup_proportion,
batch_size,
valid_batch_size,
num_epoch,
start_epoch,
accumulation_steps,
checkpoint_folder,
load_pretrain,
seed,
loss,
extra_token,
augment
):
torch.cuda.empty_cache()
COMMON_STRING ='@%s: \n' % os.path.basename(__file__)
COMMON_STRING += '\tset random seed\n'
COMMON_STRING += '\t\tseed = %d\n'%seed
COMMON_STRING += '\tset cuda environment\n'
COMMON_STRING += '\t\ttorch.__version__ = %s\n'%torch.__version__
COMMON_STRING += '\t\ttorch.version.cuda = %s\n'%torch.version.cuda
COMMON_STRING += '\t\ttorch.backends.cudnn.version() = %s\n'%torch.backends.cudnn.version()
try:
COMMON_STRING += '\t\tos[\'CUDA_VISIBLE_DEVICES\'] = %s\n'%os.environ['CUDA_VISIBLE_DEVICES']
NUM_CUDA_DEVICES = len(os.environ['CUDA_VISIBLE_DEVICES'].split(','))
except Exception:
COMMON_STRING += '\t\tos[\'CUDA_VISIBLE_DEVICES\'] = None\n'
NUM_CUDA_DEVICES = 1
COMMON_STRING += '\t\ttorch.cuda.device_count() = %d\n'%torch.cuda.device_count()
COMMON_STRING += '\n'
if augment:
if extra_token:
checkpoint_folder = os.path.join(checkpoint_folder, model_type + '/' + model_name + '-' + content + '-' + loss + '-' + \
optimizer_name + '-' + lr_scheduler_name + '-' + str(n_splits) + '-' + str(seed) + '-' + 'aug_differential_extra_token/')
else:
checkpoint_folder = os.path.join(checkpoint_folder, model_type + '/' + model_name + '-' + content + '-' + loss + '-' + \
optimizer_name + '-' + lr_scheduler_name + '-' + str(n_splits) + '-' + str(seed) + '-' + 'aug_differential/')
else:
if extra_token:
checkpoint_folder = os.path.join(checkpoint_folder, model_type + '/' + model_name + '-' + content + '-' + loss + '-' + \
optimizer_name + '-' + lr_scheduler_name + '-' + str(n_splits) + '-' + str(seed) + '-' + 'extra_token/')
else:
checkpoint_folder = os.path.join(checkpoint_folder, model_type + '/' + model_name + '-' + content + '-' + loss + '-' + \
optimizer_name + '-' + lr_scheduler_name + '-' + str(n_splits) + '-' + str(seed) + '-' + '/')
checkpoint_filename = 'fold_' + str(fold) + "_checkpoint.pth"
checkpoint_filepath = os.path.join(checkpoint_folder, checkpoint_filename)
os.makedirs(checkpoint_folder, exist_ok=True)
log = Logger()
log.open(os.path.join(checkpoint_folder, 'fold_' + str(fold) + '_log_train_swa.txt'), mode='a+')
log.write('\t%s\n' % COMMON_STRING)
log.write('\n')
log.write('\tseed = %u\n' % seed)
log.write('\tFOLD = %s\n' % fold)
log.write('\t__file__ = %s\n' % __file__)
log.write('\tout_dir = %s\n' % checkpoint_folder)
log.write('\n')
############################################################################## define unet model with backbone
def load(model, pretrain_file, skip=[]):
pretrain_state_dict = torch.load(pretrain_file)
state_dict = model.state_dict()
keys = list(state_dict.keys())
for key in keys:
if any(s in key for s in skip): continue
try:
state_dict[key] = pretrain_state_dict[key]
except:
print(key)
model.load_state_dict(state_dict)
return model
############################################################################### model
if content == "Question_Answer":
NUM_CLASS = 30
elif content == "Question":
NUM_CLASS = 21
elif content == "Answer":
NUM_CLASS = 9
else:
raise NotImplementedError
if model_type == "bert":
if extra_token:
model = QuestNet(model_type=model_name, \
tokenizer=tokenizer, \
n_classes=NUM_CLASS, \
n_category_classes=NUM_CATEGORY_CLASS, \
n_host_classes=NUM_HOST_CLASS, \
hidden_layers=hidden_layers, \
extra_token=True)
else:
model = QuestNet(model_type=model_name, \
tokenizer=tokenizer, \
n_classes=NUM_CLASS, \
n_category_classes=NUM_CATEGORY_CLASS, \
n_host_classes=NUM_HOST_CLASS, \
hidden_layers=hidden_layers, \
extra_token=False)
elif model_type == "xlnet":
if extra_token:
model = QuestNet(model_type=model_name, \
tokenizer=tokenizer, \
n_classes=NUM_CLASS, \
n_category_classes=NUM_CATEGORY_CLASS, \
n_host_classes=NUM_HOST_CLASS, \
hidden_layers=hidden_layers, \
extra_token=True)
else:
model = QuestNet(model_type=model_name, \
tokenizer=tokenizer, \
n_classes=NUM_CLASS, \
n_category_classes=NUM_CATEGORY_CLASS, \
n_host_classes=NUM_HOST_CLASS, \
hidden_layers=hidden_layers, \
extra_token=False)
else:
raise NotImplementedError
model = model.cuda()
if load_pretrain:
model = load(model, checkpoint_filepath)
############################################################################### optimizer
if ((model_type == "bert") or (model_type == "xlnet")) :
optimizer_grouped_parameters = []
list_lr = []
if ((model_name == "bert-base-uncased") or (model_name == "bert-base-cased")):
list_layers = [model.bert_model.embeddings,
model.bert_model.encoder.layer[0],
model.bert_model.encoder.layer[1],
model.bert_model.encoder.layer[2],
model.bert_model.encoder.layer[3],
model.bert_model.encoder.layer[4],
model.bert_model.encoder.layer[5],
model.bert_model.encoder.layer[6],
model.bert_model.encoder.layer[7],
model.bert_model.encoder.layer[8],
model.bert_model.encoder.layer[9],
model.bert_model.encoder.layer[10],
model.bert_model.encoder.layer[11],
model.fc_1,
model.fc
]
elif ((model_name == "bert-large-uncased") or (model_name == "bert-large-cased")):
list_layers = [model.bert_model.embeddings,
model.bert_model.encoder.layer[0],
model.bert_model.encoder.layer[1],
model.bert_model.encoder.layer[2],
model.bert_model.encoder.layer[3],
model.bert_model.encoder.layer[4],
model.bert_model.encoder.layer[5],
model.bert_model.encoder.layer[6],
model.bert_model.encoder.layer[7],
model.bert_model.encoder.layer[8],
model.bert_model.encoder.layer[9],
model.bert_model.encoder.layer[10],
model.bert_model.encoder.layer[11],
model.bert_model.encoder.layer[12],
model.bert_model.encoder.layer[13],
model.bert_model.encoder.layer[14],
model.bert_model.encoder.layer[15],
model.bert_model.encoder.layer[16],
model.bert_model.encoder.layer[17],
model.bert_model.encoder.layer[18],
model.bert_model.encoder.layer[19],
model.bert_model.encoder.layer[20],
model.bert_model.encoder.layer[21],
model.bert_model.encoder.layer[22],
model.bert_model.encoder.layer[23],
model.fc_1,
model.fc
]
elif ((model_name == "flaubert-base-uncased") or (model_name == "flaubert-base-cased")):
list_layers = [
model.flaubert_model.position_embeddings,
model.flaubert_model.embeddings,
model.flaubert_model.layer_norm_emb,
[model.flaubert_model.attentions[0], model.flaubert_model.layer_norm1[0], model.flaubert_model.ffns[0], model.flaubert_model.layer_norm2[0]],
[model.flaubert_model.attentions[1], model.flaubert_model.layer_norm1[1], model.flaubert_model.ffns[1], model.flaubert_model.layer_norm2[1]],
[model.flaubert_model.attentions[2], model.flaubert_model.layer_norm1[2], model.flaubert_model.ffns[2], model.flaubert_model.layer_norm2[2]],
[model.flaubert_model.attentions[3], model.flaubert_model.layer_norm1[3], model.flaubert_model.ffns[3], model.flaubert_model.layer_norm2[3]],
[model.flaubert_model.attentions[4], model.flaubert_model.layer_norm1[4], model.flaubert_model.ffns[4], model.flaubert_model.layer_norm2[4]],
[model.flaubert_model.attentions[5], model.flaubert_model.layer_norm1[5], model.flaubert_model.ffns[5], model.flaubert_model.layer_norm2[5]],
[model.flaubert_model.attentions[6], model.flaubert_model.layer_norm1[6], model.flaubert_model.ffns[6], model.flaubert_model.layer_norm2[6]],
[model.flaubert_model.attentions[7], model.flaubert_model.layer_norm1[7], model.flaubert_model.ffns[7], model.flaubert_model.layer_norm2[7]],
[model.flaubert_model.attentions[8], model.flaubert_model.layer_norm1[8], model.flaubert_model.ffns[8], model.flaubert_model.layer_norm2[8]],
[model.flaubert_model.attentions[9], model.flaubert_model.layer_norm1[9], model.flaubert_model.ffns[9], model.flaubert_model.layer_norm2[9]],
[model.flaubert_model.attentions[10], model.flaubert_model.layer_norm1[10], model.flaubert_model.ffns[10], model.flaubert_model.layer_norm2[10]],
[model.flaubert_model.attentions[11], model.flaubert_model.layer_norm1[11], model.flaubert_model.ffns[11], model.flaubert_model.layer_norm2[11]],
model.fc_1,
model.fc
]
elif ((model_name == "flaubert-large-cased")):
list_layers = [
model.flaubert_model.position_embeddings,
model.flaubert_model.embeddings,
model.flaubert_model.layer_norm_emb,
[model.flaubert_model.attentions[0], model.flaubert_model.layer_norm1[0], model.flaubert_model.ffns[0], model.flaubert_model.layer_norm2[0]],
[model.flaubert_model.attentions[1], model.flaubert_model.layer_norm1[1], model.flaubert_model.ffns[1], model.flaubert_model.layer_norm2[1]],
[model.flaubert_model.attentions[2], model.flaubert_model.layer_norm1[2], model.flaubert_model.ffns[2], model.flaubert_model.layer_norm2[2]],
[model.flaubert_model.attentions[3], model.flaubert_model.layer_norm1[3], model.flaubert_model.ffns[3], model.flaubert_model.layer_norm2[3]],
[model.flaubert_model.attentions[4], model.flaubert_model.layer_norm1[4], model.flaubert_model.ffns[4], model.flaubert_model.layer_norm2[4]],
[model.flaubert_model.attentions[5], model.flaubert_model.layer_norm1[5], model.flaubert_model.ffns[5], model.flaubert_model.layer_norm2[5]],
[model.flaubert_model.attentions[6], model.flaubert_model.layer_norm1[6], model.flaubert_model.ffns[6], model.flaubert_model.layer_norm2[6]],
[model.flaubert_model.attentions[7], model.flaubert_model.layer_norm1[7], model.flaubert_model.ffns[7], model.flaubert_model.layer_norm2[7]],
[model.flaubert_model.attentions[8], model.flaubert_model.layer_norm1[8], model.flaubert_model.ffns[8], model.flaubert_model.layer_norm2[8]],
[model.flaubert_model.attentions[9], model.flaubert_model.layer_norm1[9], model.flaubert_model.ffns[9], model.flaubert_model.layer_norm2[9]],
[model.flaubert_model.attentions[10], model.flaubert_model.layer_norm1[10], model.flaubert_model.ffns[10], model.flaubert_model.layer_norm2[10]],
[model.flaubert_model.attentions[11], model.flaubert_model.layer_norm1[11], model.flaubert_model.ffns[11], model.flaubert_model.layer_norm2[11]],
[model.flaubert_model.attentions[12], model.flaubert_model.layer_norm1[12], model.flaubert_model.ffns[12], model.flaubert_model.layer_norm2[12]],
[model.flaubert_model.attentions[13], model.flaubert_model.layer_norm1[13], model.flaubert_model.ffns[13], model.flaubert_model.layer_norm2[13]],
[model.flaubert_model.attentions[14], model.flaubert_model.layer_norm1[14], model.flaubert_model.ffns[14], model.flaubert_model.layer_norm2[14]],
[model.flaubert_model.attentions[15], model.flaubert_model.layer_norm1[15], model.flaubert_model.ffns[15], model.flaubert_model.layer_norm2[15]],
[model.flaubert_model.attentions[16], model.flaubert_model.layer_norm1[16], model.flaubert_model.ffns[16], model.flaubert_model.layer_norm2[16]],
[model.flaubert_model.attentions[17], model.flaubert_model.layer_norm1[17], model.flaubert_model.ffns[17], model.flaubert_model.layer_norm2[17]],
[model.flaubert_model.attentions[18], model.flaubert_model.layer_norm1[18], model.flaubert_model.ffns[18], model.flaubert_model.layer_norm2[18]],
[model.flaubert_model.attentions[19], model.flaubert_model.layer_norm1[19], model.flaubert_model.ffns[19], model.flaubert_model.layer_norm2[19]],
[model.flaubert_model.attentions[20], model.flaubert_model.layer_norm1[20], model.flaubert_model.ffns[20], model.flaubert_model.layer_norm2[20]],
[model.flaubert_model.attentions[21], model.flaubert_model.layer_norm1[21], model.flaubert_model.ffns[21], model.flaubert_model.layer_norm2[21]],
[model.flaubert_model.attentions[22], model.flaubert_model.layer_norm1[22], model.flaubert_model.ffns[22], model.flaubert_model.layer_norm2[22]],
[model.flaubert_model.attentions[23], model.flaubert_model.layer_norm1[23], model.flaubert_model.ffns[23], model.flaubert_model.layer_norm2[23]],
model.fc_1,
model.fc
]
elif (model_name == "xlnet-base-cased"):
list_layers = [model.xlnet_model.word_embedding,
model.xlnet_model.layer[0],
model.xlnet_model.layer[1],
model.xlnet_model.layer[2],
model.xlnet_model.layer[3],
model.xlnet_model.layer[4],
model.xlnet_model.layer[5],
model.xlnet_model.layer[6],
model.xlnet_model.layer[7],
model.xlnet_model.layer[8],
model.xlnet_model.layer[9],
model.xlnet_model.layer[10],
model.xlnet_model.layer[11],
model.fc_1,
model.fc
]
elif (model_name == "xlnet-large-cased"):
list_layers = [model.xlnet_model.word_embedding,
model.xlnet_model.layer[0],
model.xlnet_model.layer[1],
model.xlnet_model.layer[2],
model.xlnet_model.layer[3],
model.xlnet_model.layer[4],
model.xlnet_model.layer[5],
model.xlnet_model.layer[6],
model.xlnet_model.layer[7],
model.xlnet_model.layer[8],
model.xlnet_model.layer[9],
model.xlnet_model.layer[10],
model.xlnet_model.layer[11],
model.xlnet_model.layer[12],
model.xlnet_model.layer[13],
model.xlnet_model.layer[14],
model.xlnet_model.layer[15],
model.xlnet_model.layer[16],
model.xlnet_model.layer[17],
model.xlnet_model.layer[18],
model.xlnet_model.layer[19],
model.xlnet_model.layer[20],
model.xlnet_model.layer[21],
model.xlnet_model.layer[22],
model.xlnet_model.layer[23],
model.fc_1,
model.fc
]
elif (model_name == "roberta-base"):
list_layers = [model.roberta_model.embeddings,
model.roberta_model.encoder.layer[0],
model.roberta_model.encoder.layer[1],
model.roberta_model.encoder.layer[2],
model.roberta_model.encoder.layer[3],
model.roberta_model.encoder.layer[4],
model.roberta_model.encoder.layer[5],
model.roberta_model.encoder.layer[6],
model.roberta_model.encoder.layer[7],
model.roberta_model.encoder.layer[8],
model.roberta_model.encoder.layer[9],
model.roberta_model.encoder.layer[10],
model.roberta_model.encoder.layer[11],
model.fc_1,
model.fc
]
elif ((model_name == "albert-base-v2") or \
(model_name == "albert-large-v2") or \
(model_name == "albert-xlarge-v2") or \
(model_name == "albert-xxlarge-v2")):
list_layers = [model.albert_model.embeddings,
# model.albert_model.encoder.embedding_hidden_mapping_in,
# model.albert_model.encoder.albert_layer_groups,
model.albert_model.encoder,
model.fc_1,
model.fc
]
print("differential lr for ", model_name)
elif (model_name == "gpt2"):
list_layers = [# model.gpt2_model.wte,
# model.gpt2_model.wpe,
model.gpt2_model.h[0],
model.gpt2_model.h[1],
model.gpt2_model.h[2],
model.gpt2_model.h[3],
model.gpt2_model.h[4],
model.gpt2_model.h[5],
model.gpt2_model.h[6],
model.gpt2_model.h[7],
model.gpt2_model.h[8],
model.gpt2_model.h[9],
model.gpt2_model.h[10],
model.gpt2_model.h[11],
model.fc_1,
model.fc
]
else:
raise NotImplementedError
# for i in range(len(list_layers)):
# list_lr.append(lr)
# lr = lr * DECAY_FACTOR
# list_lr.append(lr - i * (lr - MIN_LR) / (len(list_layers) - 1))
# list_lr.reverse()
mult = lr / MIN_LR
step = mult**(1/(len(list_layers)-1))
list_lr = [MIN_LR * (step ** i) for i in range(len(list_layers))]
no_decay = ['bias', 'LayerNorm.bias', 'LayerNorm.weight']
print(list_lr)
for i in range(len(list_lr)):
if isinstance(list_layers[i], list):
for list_layer in list_layers[i]:
layer_parameters = list(list_layer.named_parameters())
optimizer_grouped_parameters.append({ \
'params': [p for n, p in layer_parameters if not any(nd in n for nd in no_decay)], \
'lr': list_lr[i], \
'weight_decay': 0.01})
optimizer_grouped_parameters.append({ \
'params': [p for n, p in layer_parameters if any(nd in n for nd in no_decay)], \
'lr': list_lr[i], \
'weight_decay': 0.0})
else:
layer_parameters = list(list_layers[i].named_parameters())
optimizer_grouped_parameters.append({ \
'params': [p for n, p in layer_parameters if not any(nd in n for nd in no_decay)], \
'lr': list_lr[i], \
'weight_decay': 0.01})
optimizer_grouped_parameters.append({ \
'params': [p for n, p in layer_parameters if any(nd in n for nd in no_decay)], \
'lr': list_lr[i], \
'weight_decay': 0.0})
if extra_token:
# add extra fcs
layer_parameters = list(model.fc_1_category.named_parameters())
optimizer_grouped_parameters.append({ \
'params': [p for n, p in layer_parameters if not any(nd in n for nd in no_decay)], \
'lr': 1e-6, \
'weight_decay': 0.01})
optimizer_grouped_parameters.append({ \
'params': [p for n, p in layer_parameters if any(nd in n for nd in no_decay)], \
'lr': 1e-6, \
'weight_decay': 0.0})
layer_parameters = list(model.fc_1_host.named_parameters())
optimizer_grouped_parameters.append({ \
'params': [p for n, p in layer_parameters if not any(nd in n for nd in no_decay)], \
'lr': 1e-6, \
'weight_decay': 0.01})
optimizer_grouped_parameters.append({ \
'params': [p for n, p in layer_parameters if any(nd in n for nd in no_decay)], \
'lr': 1e-6, \
'weight_decay': 0.0})
layer_parameters = list(model.fc_category.named_parameters())
optimizer_grouped_parameters.append({ \
'params': [p for n, p in layer_parameters if not any(nd in n for nd in no_decay)], \
'lr': 1e-6, \
'weight_decay': 0.01})
optimizer_grouped_parameters.append({ \
'params': [p for n, p in layer_parameters if any(nd in n for nd in no_decay)], \
'lr': 1e-6, \
'weight_decay': 0.0})
layer_parameters = list(model.fc_host.named_parameters())
optimizer_grouped_parameters.append({ \
'params': [p for n, p in layer_parameters if not any(nd in n for nd in no_decay)], \
'lr': 1e-6, \
'weight_decay': 0.01})
optimizer_grouped_parameters.append({ \
'params': [p for n, p in layer_parameters if any(nd in n for nd in no_decay)], \
'lr': 1e-6, \
'weight_decay': 0.0})
print("Differential Learning Rate!!")
else:
param_optimizer = list(model.named_parameters())
no_decay = ['bias', 'LayerNorm.bias', 'LayerNorm.weight']
optimizer_grouped_parameters = [
{'params': [p for n, p in param_optimizer if not any(nd in n for nd in no_decay)], \
'lr': lr, \
'weight_decay': 0.01}, \
{'params': [p for n, p in param_optimizer if any(nd in n for nd in no_decay)], \
'lr': lr, \
'weight_decay': 0.0}
]
if optimizer_name == "Adam":
optimizer = torch.optim.Adam(optimizer_grouped_parameters)
elif optimizer_name == "Ranger":
optimizer = Ranger(optimizer_grouped_parameters)
elif optimizer_name == "BertAdam":
num_train_optimization_steps = num_epoch * len(train_data_loader) // accumulation_steps
optimizer = BertAdam(optimizer_grouped_parameters,
warmup=warmup_proportion,
t_total=num_train_optimization_steps)
elif optimizer_name == "FusedAdam":
optimizer = FusedAdam(optimizer_grouped_parameters,
bias_correction=False)
else:
raise NotImplementedError
############################################################################### lr_scheduler
if lr_scheduler_name == "CosineAnealing":
scheduler = torch.optim.lr_scheduler.CosineAnnealingLR(optimizer, 12, eta_min=1e-5, last_epoch=-1)
lr_scheduler_each_iter = False
elif lr_scheduler_name == "WarmRestart":
scheduler = WarmRestart(optimizer, T_max=5, T_mult=1, eta_min=1e-6)
lr_scheduler_each_iter = False
elif lr_scheduler_name == "WarmupLinearSchedule":
num_train_optimization_steps = num_epoch * len(train_data_loader) // accumulation_steps
scheduler = get_linear_schedule_with_warmup(optimizer, \
num_warmup_steps=int(num_train_optimization_steps*warmup_proportion), \
num_training_steps=num_train_optimization_steps)
lr_scheduler_each_iter = True
else:
raise NotImplementedError
log.write('net\n %s\n'%(model_name))
log.write('optimizer\n %s\n'%(optimizer_name))
log.write('schduler\n %s\n'%(lr_scheduler_name))
log.write('\n')
############################################################################### mix precision
model, optimizer = amp.initialize(model, optimizer, opt_level="O1")
# model = nn.DataParallel(model)
############################################################################### eval setting
eval_step = len(train_data_loader) # or len(train_data_loader)
log_step = 50
eval_count = 0
############################################################################### training
log.write('** start training here! **\n')
log.write(' batch_size=%d, accumulation_steps=%d\n'%(batch_size, accumulation_steps))
log.write(' experiment = %s\n' % str(__file__.split('/')[-2:]))
valid_loss = np.zeros(1, np.float32)
train_loss = np.zeros(1, np.float32)
valid_metric_optimal = -np.inf
# define tensorboard writer and timer
writer = SummaryWriter()
# define criterion
if loss == 'mse':
criterion = MSELoss()
elif loss == 'bce':
if content == "Question_Answer":
weights = torch.tensor(np.array(UNBALANCE_WEIGIHT), dtype=torch.float64).cuda()
elif content == "Question":
weights = torch.tensor(np.array(QUESTION_UNBALANCE_WEIGIHT), dtype=torch.float64).cuda()
elif content == "Answer":
weights = torch.tensor(np.array(ANSWER_UNBALANCE_WEIGIHT), dtype=torch.float64).cuda()
else:
raise NotImplementedError
criterion = nn.BCEWithLogitsLoss(weight=weights)
criterion_extra = nn.BCEWithLogitsLoss()
elif loss == 'mse-bce':
criterion = MSEBCELoss()
elif loss == 'focal':
criterion = FocalLoss()
else:
raise NotImplementedError
for epoch in range(1, num_epoch+1):
# save last epoch weights
checkpoint_filename_last_epoch = 'fold_' + str(fold) + "_checkpoint_last_epoch.pth"
checkpoint_filepath_last_epoch = os.path.join(checkpoint_folder, checkpoint_filename_last_epoch)
torch.save(model.state_dict(), checkpoint_filepath_last_epoch)
# init in-epoch statistics
labels_train = None
pred_train = None
labels_val = None
pred_val = None
# update lr and start from start_epoch
if ((epoch > 1) and (not lr_scheduler_each_iter)):
scheduler.step()
if (epoch < start_epoch):
continue
log.write("Epoch%s\n" % epoch)
log.write('\n')
sum_train_loss = np.zeros_like(train_loss)
sum_train = np.zeros_like(train_loss)
# init optimizer
torch.cuda.empty_cache()
model.zero_grad()
if extra_token:
for tr_batch_i, (token_ids, seg_ids, labels, labels_category, labels_host) in enumerate(train_data_loader):
rate = 0
for param_group in optimizer.param_groups:
rate += param_group['lr'] / len(optimizer.param_groups)
# set model training mode
model.train()
# set input to cuda mode
token_ids = token_ids.cuda()
seg_ids = seg_ids.cuda()
labels = labels.cuda().float()
labels_category = labels_category.cuda().float()
labels_host = labels_host.cuda().float()
# predict and calculate loss (only need torch.sigmoid when inference)
prediction, prediction_category, prediction_host = model(token_ids, seg_ids)
# print(prediction.shape, prediction_category.shape, prediction_host.shape)
# print(labels.shape, labels_category.shape, labels_host.shape)
loss = AUXILIARY_WEIGHTs[0]*criterion(prediction, labels) + \
AUXILIARY_WEIGHTs[1]*criterion_extra(prediction_category, labels_category) + \
AUXILIARY_WEIGHTs[2]*criterion_extra(prediction_host, labels_host)
# use apex
with amp.scale_loss(loss/accumulation_steps, optimizer) as scaled_loss:
scaled_loss.backward()
# don't use apex
#loss.backward()
if ((tr_batch_i+1) % accumulation_steps == 0):
torch.nn.utils.clip_grad_norm_(model.parameters(), max_norm=5.0, norm_type=2)
optimizer.step()
model.zero_grad()
# adjust lr
if (lr_scheduler_each_iter):
scheduler.step()
writer.add_scalar('train_loss_' + str(fold), loss.item(), (epoch-1)*len(train_data_loader)*batch_size+tr_batch_i*batch_size)
# calculate statistics
prediction = torch.sigmoid(prediction)
if tr_batch_i == 0:
labels_train = labels.cpu().detach().numpy()
pred_train = prediction.cpu().detach().numpy()
else:
labels_train = np.concatenate((labels_train, labels.cpu().detach().numpy()), axis=0)
pred_train = np.concatenate((pred_train, prediction.cpu().detach().numpy()), axis=0)
l = np.array([loss.item() * batch_size])
n = np.array([batch_size])
sum_train_loss = sum_train_loss + l
sum_train = sum_train + n
# log for training
if (tr_batch_i+1) % log_step == 0:
train_loss = sum_train_loss / (sum_train + 1e-12)
sum_train_loss[...] = 0
sum_train[...] = 0
spearman = Spearman(labels_train, pred_train)
log.write('lr: %f train loss: %f train_spearman: %f\n' % \
(rate, train_loss[0], spearman))
if (tr_batch_i+1) % eval_step == 0:
eval_count += 1
valid_loss = np.zeros(1, np.float32)
valid_num = np.zeros_like(valid_loss)
with torch.no_grad():
# init cache
torch.cuda.empty_cache()
for val_batch_i, (token_ids, seg_ids, labels, labels_category, labels_host) in enumerate(val_data_loader):
# set model to eval mode
model.eval()
# set input to cuda mode
token_ids = token_ids.cuda()
seg_ids = seg_ids.cuda()
labels = labels.cuda().float()
labels_category = labels_category.cuda().float()
labels_host = labels_host.cuda().float()
# predict and calculate loss (only need torch.sigmoid when inference)
prediction, prediction_category, prediction_host = model(token_ids, seg_ids)
loss = AUXILIARY_WEIGHTs[0]*criterion(prediction, labels) + \
AUXILIARY_WEIGHTs[1]*criterion_extra(prediction_category, labels_category) + \
AUXILIARY_WEIGHTs[2]*criterion_extra(prediction_host, labels_host)
writer.add_scalar('val_loss_' + str(fold), loss.item(), (eval_count-1)*len(val_data_loader)*valid_batch_size+val_batch_i*valid_batch_size)
# calculate statistics
prediction = torch.sigmoid(prediction)
if val_batch_i == 0:
labels_val = labels.cpu().detach().numpy()
pred_val = prediction.cpu().detach().numpy()
else:
labels_val = np.concatenate((labels_val, labels.cpu().detach().numpy()), axis=0)
pred_val = np.concatenate((pred_val, prediction.cpu().detach().numpy()), axis=0)
l = np.array([loss.item()*valid_batch_size])
n = np.array([valid_batch_size])
valid_loss = valid_loss + l
valid_num = valid_num + n
valid_loss = valid_loss / valid_num
spearman = Spearman(labels_val, pred_val)
log.write('validation loss: %f val_spearman: %f\n' % \
(valid_loss[0], spearman))
else:
for tr_batch_i, (token_ids, seg_ids, labels) in enumerate(train_data_loader):
rate = 0
for param_group in optimizer.param_groups:
rate += param_group['lr'] / len(optimizer.param_groups)
# set model training mode
model.train()
# set input to cuda mode
token_ids = token_ids.cuda()
seg_ids = seg_ids.cuda()
labels = labels.cuda().float()
# predict and calculate loss (only need torch.sigmoid when inference)
prediction = model(token_ids, seg_ids)
loss = criterion(prediction, labels)
# use apex
with amp.scale_loss(loss/accumulation_steps, optimizer) as scaled_loss:
scaled_loss.backward()
# don't use apex
#loss.backward()
if ((tr_batch_i+1) % accumulation_steps == 0):
torch.nn.utils.clip_grad_norm_(model.parameters(), max_norm=5.0, norm_type=2)
optimizer.step()
model.zero_grad()
# adjust lr
if (lr_scheduler_each_iter):
scheduler.step()
writer.add_scalar('train_loss_' + str(fold), loss.item(), (epoch-1)*len(train_data_loader)*batch_size+tr_batch_i*batch_size)
# calculate statistics
prediction = torch.sigmoid(prediction)
if tr_batch_i == 0:
labels_train = labels.cpu().detach().numpy()
pred_train = prediction.cpu().detach().numpy()
else:
labels_train = np.concatenate((labels_train, labels.cpu().detach().numpy()), axis=0)
pred_train = np.concatenate((pred_train, prediction.cpu().detach().numpy()), axis=0)
l = np.array([loss.item() * batch_size])
n = np.array([batch_size])
sum_train_loss = sum_train_loss + l
sum_train = sum_train + n
# log for training
if (tr_batch_i+1) % log_step == 0:
train_loss = sum_train_loss / (sum_train + 1e-12)
sum_train_loss[...] = 0
sum_train[...] = 0
spearman = Spearman(labels_train, pred_train)
log.write('lr: %f train loss: %f train_spearman: %f\n' % \
(rate, train_loss[0], spearman))
if (tr_batch_i+1) % eval_step == 0:
eval_count += 1
valid_loss = np.zeros(1, np.float32)
valid_num = np.zeros_like(valid_loss)
with torch.no_grad():
# init cache
torch.cuda.empty_cache()
for val_batch_i, (token_ids, seg_ids, labels) in enumerate(val_data_loader):
# set model to eval mode
model.eval()
# set input to cuda mode
token_ids = token_ids.cuda()
seg_ids = seg_ids.cuda()
labels = labels.cuda().float()
# predict and calculate loss (only need torch.sigmoid when inference)
prediction = model(token_ids, seg_ids)
loss = criterion(prediction, labels)
writer.add_scalar('val_loss_' + str(fold), loss.item(), (eval_count-1)*len(val_data_loader)*valid_batch_size+val_batch_i*valid_batch_size)
# calculate statistics
prediction = torch.sigmoid(prediction)
if val_batch_i == 0:
labels_val = labels.cpu().detach().numpy()
pred_val = prediction.cpu().detach().numpy()
else:
labels_val = np.concatenate((labels_val, labels.cpu().detach().numpy()), axis=0)
pred_val = np.concatenate((pred_val, prediction.cpu().detach().numpy()), axis=0)
l = np.array([loss.item()*valid_batch_size])
n = np.array([valid_batch_size])
valid_loss = valid_loss + l
valid_num = valid_num + n
valid_loss = valid_loss / valid_num
spearman = Spearman(labels_val, pred_val)
log.write('validation loss: %f val_spearman: %f\n' % \
(valid_loss[0], spearman))
val_metric_epoch = spearman
checkpoint_filename_swa = 'fold_' + str(fold) + "_checkpoint_swa.pth"
checkpoint_filepath_swa = os.path.join(checkpoint_folder, checkpoint_filename_swa)
log.write('Validation metric {:.6f}. Saving model ...'.format(val_metric_epoch))
valid_metric_optimal = val_metric_epoch
# swa update weights
state_dict_last_epoch = torch.load(checkpoint_filepath_last_epoch)
state_dict = model.state_dict()
for name, param in state_dict.items():
state_dict[name].data.copy_((param.data + epoch*state_dict_last_epoch[name].data) / (epoch + 1))
model.load_state_dict(state_dict)
torch.save(model.state_dict(), checkpoint_filepath_swa)
if __name__ == "__main__":
# torch.multiprocessing.set_start_method('spawn')
args = parser.parse_args()
seed_everything(args.seed)
# get train val split
data_path = args.train_data_folder + "train_augment_final_with_clean.csv"
get_train_val_split(data_path=data_path, \
save_path=args.train_data_folder, \
n_splits=args.n_splits, \
seed=args.seed, \
split=args.split)
# get train_data_loader and val_data_loader
train_data_path = args.train_data_folder + "split/train_fold_%s_seed_%s.csv"%(args.fold, args.seed)
val_data_path = args.train_data_folder + "split/val_fold_%s_seed_%s.csv"%(args.fold, args.seed)
if ((args.model_type == "bert") or (args.model_type == "xlnet")):
if args.extra_token:
test_data_path = args.train_data_folder + "test.csv"
train_df = pd.read_csv(data_path)
test_df = pd.read_csv(test_data_path)
train_host_list = train_df['host'].unique().tolist()
test_host_list = test_df['host'].unique().tolist()
host_encoder = LabelBinarizer()
host_encoder.fit(list(set(train_host_list + test_host_list)))
train_category_list = train_df['category'].unique().tolist()
test_category_list = test_df['category'].unique().tolist()
category_encoder = LabelBinarizer()
category_encoder.fit(list(set(train_category_list + test_category_list)))
train_data_loader, val_data_loader, tokenizer = get_train_val_loaders(train_data_path=train_data_path, \
val_data_path=val_data_path, \
host_encoder=host_encoder, \
category_encoder=category_encoder, \
model_type=args.model_name, \
content=args.content, \
max_len=args.max_len, \
batch_size=args.batch_size, \
val_batch_size=args.valid_batch_size, \
num_workers=args.num_workers, \
augment=args.augment, \
extra_token=True)
else:
train_data_loader, val_data_loader, tokenizer = get_train_val_loaders(train_data_path=train_data_path, \
val_data_path=val_data_path, \
model_type=args.model_name, \
content=args.content, \
max_len=args.max_len, \
batch_size=args.batch_size, \
val_batch_size=args.valid_batch_size, \
num_workers=args.num_workers, \
augment=args.augment, \
extra_token=False)
else:
raise NotImplementedError
# start training