-
Notifications
You must be signed in to change notification settings - Fork 14
/
Copy pathfix_train.py
1174 lines (1101 loc) · 49.4 KB
/
fix_train.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 sys
##sys.path.extend([])
##
import importlib
import os
import time
import random
import math
from bisect import bisect_right
from functools import wraps
import sys
import numpy as np
np.set_printoptions(threshold=sys.maxsize)
import torch
import torch.nn as nn
from torch import multiprocessing
from torch.distributed import is_initialized
from torchvision import datasets, transforms
from torch.utils.data.distributed import DistributedSampler
from torch.nn.modules.utils import _pair
from myutils.distributed import init_dist, master_only, is_master
from myutils.distributed import get_rank, get_world_size
from myutils.distributed import dist_all_reduce_tensor
from myutils.distributed import master_only_print as mprint
from myutils.distributed import AllReduceDistributedDataParallel, allreduce_grads
from myutils.config import FLAGS
from myutils.meters import ScalarMeter, flush_scalar_meters
from myutils.export import onnx_export
from models.fix_quant_ops import ReLUClipFXQConvBN, ReLUClipFXQLinear, bn_calib, fix_quant
from pytorchcv.model_provider import get_model as ptcv_get_model
def timing(f):
@wraps(f)
def wrap(*args, **kw):
if is_master():
ts = time.time()
result = f(*args, **kw)
te = time.time()
mprint('func:{!r} took: {:2.4f} sec'.format(f.__name__, te - ts))
else:
result = f(*args, **kw)
return result
return wrap
def ptcv_load(model):
model_name_dict = {
'fix_resnet18': 'resnet18',
'fix_resnet50': 'resnet50b',
'fix_mobilenet_v1': 'mobilenet_w1',
'fix_mobilenet_v2': 'mobilenetv2b_w1'
}
model_name = FLAGS.model.split('.')[1] + str(getattr(FLAGS, 'depth', ''))
ptcv_model_name = model_name_dict[model_name]
net = ptcv_get_model(ptcv_model_name,
pretrained=True).to(next(model.parameters()).device)
if getattr(FLAGS, 'hawq_pretrained', False):
checkpoint = torch.load(getattr(FLAGS, 'hawq_pretrained_file', ''))
model_key_list = list(net.state_dict().keys())
for key in model_key_list:
if 'num_batches_tracked' in key: model_key_list.remove(key)
i = 0
modified_dict = {}
for key, value in checkpoint.items():
if 'scaling_factor' in key: continue
if 'num_batches_tracked' in key: continue
if 'weight_integer' in key: continue
if 'min' in key or 'max' in key: continue
modified_key = model_key_list[i]
modified_dict[modified_key] = value
i += 1
net.load_state_dict(modified_dict, strict=False)
name_weight_bias_list = []
for n, m in net.named_modules():
if isinstance(m, (nn.Conv2d, nn.Linear)):
name_weight_bias_tuple = (n, m.weight, m.bias)
name_weight_bias_list.append(name_weight_bias_tuple)
elif isinstance(m, nn.BatchNorm2d):
name_weight_bias_tuple = (n, m.weight, m.bias, m.running_mean,
m.running_var)
name_weight_bias_list.append(name_weight_bias_tuple)
idx = 0
for n, m in model.named_modules():
if isinstance(m, ReLUClipFXQConvBN):
conv_name, conv_weight, conv_bias = name_weight_bias_list[idx]
bn_name, bn_weight, bn_bias, bn_running_mean, bn_running_var = name_weight_bias_list[
idx + 1]
mprint('model layer name: ', n)
mprint('ptcv model conv name: ', conv_name)
mprint('ptcv model bn name: ', bn_name)
assert m.conv.weight.shape == conv_weight.shape
assert m.conv.bias is None == conv_bias is None
if m.conv.bias is not None:
assert m.conv.bias.shape == conv_bias.shape
assert m.bn.weight.shape == bn_weight.shape
assert m.bn.bias.shape == bn_bias.shape
assert m.bn.running_mean.shape == bn_running_mean.shape
assert m.bn.running_var.shape == bn_running_var.shape
m.conv.weight.data = conv_weight.data
if m.conv.bias is not None:
m.conv.bias.data = conv_bias.data
m.bn.weight.data = bn_weight.data
m.bn.bias.data = bn_bias.data
m.bn.running_mean.data = bn_running_mean.data
m.bn.running_var.data = bn_running_var.data
idx += 2
if isinstance(m, ReLUClipFXQLinear):
fc_name, fc_weight, fc_bias = name_weight_bias_list[idx]
mprint('model layer name: ', n)
mprint('ptcv model fc name: ', fc_name)
if len(fc_weight.shape) == 4:
assert fc_weight.shape[-1] == 1
assert fc_weight.shape[-2] == 1
fc_weight = fc_weight[..., 0, 0]
assert m.weight.shape == fc_weight.shape
m.weight.data = fc_weight.data
if m.bias is None:
assert fc_bias is None
else:
assert m.bias.shape == fc_bias.shape
m.bias.data = fc_bias.data
idx += 1
assert idx == len(name_weight_bias_list)
del net
def nvidia_load(model):
ckpt = torch.load(getattr(FLAGS, 'nvidia_pretrained_file', ''))
name_weight_bias_list = []
name_weight_bias_tuple = []
for k, v in ckpt.items():
if 'conv' in k:
if 'weight' in k:
n = k.replace('.weight', '')
name_weight_bias_tuple = (n, v, None)
name_weight_bias_list.append(name_weight_bias_tuple)
name_weight_bias_tuple = []
else:
raise NotImplementedError
elif 'bn' in k:
if 'weight' in k:
n = k.replace('.weight', '')
name_weight_bias_tuple.append(n)
name_weight_bias_tuple.append(v)
elif 'bias' in k:
name_weight_bias_tuple.append(v)
elif 'running_mean' in k:
name_weight_bias_tuple.append(v)
elif 'running_var' in k:
name_weight_bias_tuple.append(v)
elif 'num_batches_tracked' in k:
name_weight_bias_tuple = tuple(name_weight_bias_tuple)
name_weight_bias_list.append(name_weight_bias_tuple)
name_weight_bias_tuple = []
else:
raise NotImplementedError
elif 'downsample' in k:
layer_idx = int(k.split('.')[-2])
if layer_idx == 0:
if 'weight' in k:
n = k.replace('.weight', '')
name_weight_bias_tuple = (n, v, None)
name_weight_bias_list.append(name_weight_bias_tuple)
name_weight_bias_tuple = []
else:
raise NotImplementedError
elif layer_idx == 1:
if 'weight' in k:
n = k.replace('.weight', '')
name_weight_bias_tuple.append(n)
name_weight_bias_tuple.append(v)
elif 'bias' in k:
name_weight_bias_tuple.append(v)
elif 'running_mean' in k:
name_weight_bias_tuple.append(v)
elif 'running_var' in k:
name_weight_bias_tuple.append(v)
elif 'num_batches_tracked' in k:
name_weight_bias_tuple = tuple(name_weight_bias_tuple)
name_weight_bias_list.append(name_weight_bias_tuple)
name_weight_bias_tuple = []
else:
raise NotImplementedError
else:
raise NotImplementedError
elif 'fc' in k:
if 'weight' in k:
n = k.replace('.weight', '')
name_weight_bias_tuple.append(n)
name_weight_bias_tuple.append(v)
elif 'bias' in k:
name_weight_bias_tuple.append(v)
name_weight_bias_tuple = tuple(name_weight_bias_tuple)
name_weight_bias_list.append(name_weight_bias_tuple)
name_weight_bias_tuple = []
else:
raise NotImplementedError
else:
mprint('k: ', k)
raise NotImplementedError
idx = 0
for n, m in model.named_modules():
if isinstance(m, ReLUClipFXQConvBN):
conv_name, conv_weight, conv_bias = name_weight_bias_list[idx]
bn_name, bn_weight, bn_bias, bn_running_mean, bn_running_var = name_weight_bias_list[
idx + 1]
mprint('model layer name: ', n)
mprint('nvidia model conv name: ', conv_name)
mprint('nvidia model bn name: ', bn_name)
assert m.conv.weight.shape == conv_weight.shape
assert m.conv.bias is None == conv_bias is None
if m.conv.bias is not None:
assert m.conv.bias.shape == conv_weight.shape
assert m.bn.weight.shape == bn_weight.shape
assert m.bn.bias.shape == bn_bias.shape
assert m.bn.running_mean.shape == bn_running_mean.shape
assert m.bn.running_var.shape == bn_running_var.shape
m.conv.weight.data = conv_weight.data.to(m.conv.weight.device)
if m.conv.bias is not None:
m.conv.bias.data = conv_bias.data.to(m.conv.bias.device)
m.bn.weight.data = bn_weight.data.to(m.bn.weight.device)
m.bn.bias.data = bn_bias.data.to(m.bn.bias.device)
m.bn.running_mean.data = bn_running_mean.data.to(
m.bn.running_mean.device)
m.bn.running_var.data = bn_running_var.data.to(
m.bn.running_var.device)
idx += 2
if isinstance(m, ReLUClipFXQLinear):
fc_name, fc_weight, fc_bias = name_weight_bias_list[idx]
mprint('model layer name: ', n)
mprint('nvidia fc name: ', fc_name)
if len(fc_weight.shape) == 4:
assert fc_weight.shape[-1] == 1
assert fc_weight.shape[-2] == 1
fc_weight = fc_weight[..., 0, 0]
assert m.weight.shape == fc_weight.shape
m.weight.data = fc_weight.data.to(m.weight.device)
if m.bias is None:
assert fc_bias is None
else:
assert m.bias.shape == fc_bias.shape
m.bias.data = fc_bias.data.to(m.bias.device)
idx += 1
assert idx == len(name_weight_bias_list)
del ckpt
def get_model(gpu_id=None):
"""get model"""
model_lib = importlib.import_module(FLAGS.model)
model = model_lib.Model(FLAGS.num_classes)
if getattr(FLAGS, 'distributed', False):
if getattr(FLAGS, 'distributed_all_reduce', False):
model_wrapper = AllReduceDistributedDataParallel(model.cuda())
else:
model_wrapper = torch.nn.parallel.DistributedDataParallel(
model.cuda(), [gpu_id], gpu_id, find_unused_parameters=True)
else:
model_wrapper = torch.nn.DataParallel(model).cuda()
for m in model.modules():
if isinstance(m, (ReLUClipFXQConvBN, ReLUClipFXQLinear)):
m.set_weight_format(FLAGS.weight_format)
m.set_input_format(FLAGS.input_format)
m.rescale_type = getattr(FLAGS, 'rescale_type', 'constant')
m.set_alpha()
m.floating = getattr(FLAGS, 'floating_model', False)
m.floating_wo_clip = getattr(FLAGS, 'floating_wo_clip', False)
m.format_type = getattr(FLAGS, 'format_type', None)
m.format_from_metric = getattr(FLAGS, 'format_from_metric', False)
m.metric = getattr(FLAGS, 'metric', None)
m.format_grid_search = getattr(FLAGS, 'format_grid_search', False)
m.set_metric_func()
m.register_input_format(FLAGS.input_format,
momentum=getattr(FLAGS,
'momentum_for_metric',
0.1))
m.no_clipping = getattr(FLAGS, 'no_clipping', False)
m.input_fraclen_sharing = getattr(FLAGS, 'input_fraclen_sharing',
False)
m.quant_bias = getattr(FLAGS, 'quant_bias', False)
m.int_infer = getattr(FLAGS, 'int_infer', False)
if isinstance(m, ReLUClipFXQConvBN):
m.rescale_forward = getattr(FLAGS, 'rescale_forward_conv', False)
if isinstance(m, ReLUClipFXQLinear):
m.rescale_forward = getattr(FLAGS, 'rescale_forward', True)
return model, model_wrapper
def data_transforms():
"""get transform of dataset"""
if FLAGS.data_transforms in ['imagenet1k']:
if getattr(FLAGS, 'normalize', False):
mean = [0.485, 0.456, 0.406]
std = [0.229, 0.224, 0.225]
else:
mean = [0.0, 0.0, 0.0]
std = [1.0, 1.0, 1.0]
train_transforms = transforms.Compose([
transforms.RandomResizedCrop(224),
transforms.RandomHorizontalFlip(),
transforms.ToTensor(),
transforms.Normalize(mean=mean, std=std),
])
val_transforms = transforms.Compose([
transforms.Resize(256),
transforms.CenterCrop(224),
transforms.ToTensor(),
transforms.Normalize(mean=mean, std=std),
])
test_transforms = val_transforms
else:
try:
transforms_lib = importlib.import_module(FLAGS.data_transforms)
return transforms_lib.data_transforms()
except ImportError:
raise NotImplementedError(
'Data transform {} is not yet implemented.'.format(
FLAGS.data_transforms))
return train_transforms, val_transforms, test_transforms
def dataset(train_transforms, val_transforms, test_transforms):
"""get dataset for classification"""
if FLAGS.dataset == 'imagenet1k':
if not FLAGS.test_only:
train_set = datasets.ImageFolder(os.path.join(
FLAGS.dataset_dir, 'train'),
transform=train_transforms)
else:
train_set = None
val_set = datasets.ImageFolder(os.path.join(FLAGS.dataset_dir, 'val'),
transform=val_transforms)
test_set = None
else:
try:
dataset_lib = importlib.import_module(FLAGS.dataset)
return dataset_lib.dataset(train_transforms, val_transforms,
test_transforms)
except ImportError:
raise NotImplementedError(
'Dataset {} is not yet implemented.'.format(FLAGS.dataset))
return train_set, val_set, test_set
def data_loader(train_set, val_set, test_set):
"""get data loader"""
train_loader = None
val_loader = None
test_loader = None
if getattr(FLAGS, 'batch_size', False):
if getattr(FLAGS, 'batch_size_per_gpu', False):
assert FLAGS.batch_size == (FLAGS.batch_size_per_gpu *
FLAGS.num_gpus_per_job)
else:
assert FLAGS.batch_size % FLAGS.num_gpus_per_job == 0
FLAGS.batch_size_per_gpu = (FLAGS.batch_size //
FLAGS.num_gpus_per_job)
elif getattr(FLAGS, 'batch_size_per_gpu', False):
FLAGS.batch_size = FLAGS.batch_size_per_gpu * FLAGS.num_gpus_per_job
else:
raise ValueError('batch size (per gpu) is not defined')
batch_size = int(FLAGS.batch_size / get_world_size())
if FLAGS.data_loader in ['imagenet1k']:
if getattr(FLAGS, 'distributed', False):
if FLAGS.test_only:
train_sampler = None
else:
train_sampler = DistributedSampler(train_set)
val_sampler = DistributedSampler(val_set)
else:
train_sampler = None
val_sampler = None
if not FLAGS.test_only:
train_loader = torch.utils.data.DataLoader(
train_set,
batch_size=batch_size,
shuffle=(train_sampler is None),
sampler=train_sampler,
pin_memory=True,
num_workers=FLAGS.data_loader_workers,
drop_last=getattr(FLAGS, 'drop_last', False))
val_loader = torch.utils.data.DataLoader(
val_set,
batch_size=batch_size,
shuffle=False,
sampler=val_sampler,
pin_memory=True,
num_workers=FLAGS.data_loader_workers,
drop_last=getattr(FLAGS, 'drop_last', False))
test_loader = val_loader
else:
try:
data_loader_lib = importlib.import_module(FLAGS.data_loader)
return data_loader_lib.data_loader(train_set, val_set, test_set)
except ImportError:
raise NotImplementedError(
'Data loader {} is not yet implemented.'.format(
FLAGS.data_loader))
if train_loader is not None:
FLAGS.data_size_train = len(train_loader.dataset)
if val_loader is not None:
FLAGS.data_size_val = len(val_loader.dataset)
if test_loader is not None:
FLAGS.data_size_test = len(test_loader.dataset)
return train_loader, val_loader, test_loader
def lr_func(x, fun='cos'):
if fun == 'cos':
return math.cos(x * math.pi) / 2 + 0.5
if fun == 'exp':
return math.exp(-x * 8)
def get_lr_scheduler(optimizer, nBatch=None):
"""get learning rate"""
if FLAGS.lr_scheduler == 'constant':
lr_dict = {}
for i in range(FLAGS.num_epochs + 1):
lr_dict[i] = 1.
lr_lambda = lambda epoch: lr_dict[epoch] # noqa: E731
lr_scheduler = torch.optim.lr_scheduler.LambdaLR(optimizer,
lr_lambda=lr_lambda)
elif FLAGS.lr_scheduler == 'multistep':
lr_scheduler = torch.optim.lr_scheduler.MultiStepLR(
optimizer,
milestones=FLAGS.multistep_lr_milestones,
gamma=FLAGS.multistep_lr_gamma)
elif FLAGS.lr_scheduler == 'multistep_iter':
FLAGS.num_iters = FLAGS.num_epochs * nBatch
FLAGS.warmup_iters = FLAGS.warmup_epochs * nBatch
lr_dict = {}
for i in range(FLAGS.warmup_iters):
bs_ratio = 256 / FLAGS.batch_size
lr_dict[i] = (1 - bs_ratio) / FLAGS.warmup_iters * i + bs_ratio
for i in range(FLAGS.warmup_iters, FLAGS.num_iters + 1):
lr_dict[i] = FLAGS.multistep_lr_gamma**bisect_right(
FLAGS.multistep_lr_milestones, i // nBatch)
lr_lambda = lambda itr: lr_dict[itr] # noqa: E731
lr_scheduler = torch.optim.lr_scheduler.LambdaLR(optimizer,
lr_lambda=lr_lambda)
elif FLAGS.lr_scheduler == 'exp_decaying':
lr_dict = {}
for i in range(FLAGS.num_epochs + 1):
if i == 0:
lr_dict[i] = 1
elif i % getattr(FLAGS, 'exp_decaying_period', 1) == 0:
lr_dict[i] = lr_dict[i - 1] * FLAGS.exp_decaying_lr_gamma
else:
lr_dict[i] = lr_dict[i - 1]
lr_lambda = lambda epoch: lr_dict[epoch] # noqa: E731
lr_scheduler = torch.optim.lr_scheduler.LambdaLR(optimizer,
lr_lambda=lr_lambda)
elif FLAGS.lr_scheduler == 'exp_decaying_iter':
FLAGS.num_iters = FLAGS.num_epochs * nBatch
FLAGS.warmup_iters = FLAGS.warmup_epochs * nBatch
lr_dict = {}
for i in range(FLAGS.warmup_iters):
bs_ratio = 256 / FLAGS.batch_size
lr_dict[i] = (1 - bs_ratio) / FLAGS.warmup_iters * i + bs_ratio
for i in range(FLAGS.warmup_iters, FLAGS.num_iters + 1):
lr_dict[i] = lr_func((i - FLAGS.warmup_iters) /
(FLAGS.num_iters - FLAGS.warmup_iters), 'exp')
lr_lambda = lambda itr: lr_dict[itr] # noqa: E731
lr_scheduler = torch.optim.lr_scheduler.LambdaLR(optimizer,
lr_lambda=lr_lambda)
elif FLAGS.lr_scheduler == 'linear_decaying':
num_epochs = FLAGS.num_epochs - FLAGS.warmup_epochs
lr_dict = {}
for i in range(FLAGS.num_epochs + 1):
lr_dict[i] = 1. - (i - FLAGS.warmup_epochs) / FLAGS.num_epochs
lr_lambda = lambda epoch: lr_dict[epoch] # noqa: E731
lr_scheduler = torch.optim.lr_scheduler.LambdaLR(optimizer,
lr_lambda=lr_lambda)
elif FLAGS.lr_scheduler == 'cos_annealing':
num_epochs = FLAGS.num_epochs - FLAGS.warmup_epochs
lr_dict = {}
for i in range(FLAGS.num_epochs + 1):
lr_dict[i] = (1.0 + math.cos(
(i - FLAGS.warmup_epochs) * math.pi / num_epochs)) / 2
lr_lambda = lambda epoch: lr_dict[epoch] # noqa: E731
lr_scheduler = torch.optim.lr_scheduler.LambdaLR(optimizer,
lr_lambda=lr_lambda)
elif FLAGS.lr_scheduler == 'cos_annealing_iter':
FLAGS.num_iters = FLAGS.num_epochs * nBatch
FLAGS.warmup_iters = FLAGS.warmup_epochs * nBatch
lr_dict = {}
for i in range(FLAGS.warmup_iters):
bs_ratio = 256 / FLAGS.batch_size
lr_dict[i] = (1 - bs_ratio) / FLAGS.warmup_iters * i + bs_ratio
for i in range(FLAGS.warmup_iters, FLAGS.num_iters + 1):
lr_dict[i] = (1.0 + math.cos(
(i - FLAGS.warmup_iters) * math.pi /
(FLAGS.num_iters - FLAGS.warmup_iters))) / 2
lr_lambda = lambda itr: lr_dict[itr] # noqa: E731
lr_scheduler = torch.optim.lr_scheduler.LambdaLR(optimizer,
lr_lambda=lr_lambda)
else:
try:
lr_scheduler_lib = importlib.import_module(FLAGS.lr_scheduler)
return lr_scheduler_lib.get_lr_scheduler(optimizer)
except ImportError:
raise NotImplementedError(
'Learning rate scheduler {} is not yet implemented.'.format(
FLAGS.lr_scheduler))
return lr_scheduler
def get_optimizer(model):
"""get optimizer"""
if FLAGS.optimizer == 'sgd':
model_params = []
for name, params in model.named_parameters():
ps = list(params.size())
weight_decay_scheme = getattr(FLAGS, 'weight_decay_scheme', 'all')
if weight_decay_scheme == 'all':
weight_decay = FLAGS.weight_decay
elif weight_decay_scheme == 'only_no_depthwise':
# all depthwise convolution (N, 1, x, x) has no weight decay
# weight decay only on normal conv, bn and fc
if len(ps) == 4 and ps[1] != 1:
weight_decay = FLAGS.weight_decay
elif len(ps) in [1, 2]:
weight_decay = FLAGS.weight_decay
else:
weight_decay = 0
elif weight_decay_scheme == 'only_no_bn':
## only no weight decay on bn
if len(ps) == 4:
weight_decay = FLAGS.weight_decay
elif len(ps) == 2:
weight_decay = FLAGS.weight_decay
else:
weight_decay = 0
elif weight_decay_scheme == 'no_depthwise_no_bn':
## no weight decay on bn and no weight decay on depthwise conv
if len(ps) == 4 and ps[1] != 1:
weight_decay = FLAGS.weight_decay
elif len(ps) == 2:
weight_decay = FLAGS.weight_decay
else:
weight_decay = 0
else:
raise NotImplementedError
lr = FLAGS.lr
item = {
'params': params,
'weight_decay': weight_decay,
'lr': lr,
'momentum': FLAGS.momentum,
'nesterov': FLAGS.nesterov
}
model_params.append(item)
optimizer = torch.optim.SGD(model_params)
else:
try:
optimizer_lib = importlib.import_module(FLAGS.optimizer)
return optimizer_lib.get_optimizer(model)
except ImportError:
raise NotImplementedError(
'Optimizer {} is not yet implemented.'.format(FLAGS.optimizer))
return optimizer
def set_random_seed(seed=None):
"""set random seed"""
if seed is None:
seed = getattr(FLAGS, 'random_seed', 0)
mprint('seed for random sampling: {}'.format(seed))
random.seed(seed)
np.random.seed(seed)
torch.manual_seed(seed)
torch.cuda.manual_seed(seed)
torch.cuda.manual_seed_all(seed)
@master_only
def get_meters(phase, single_sample=False):
"""util function for meters"""
def get_single_meter(phase, suffix=''):
meters = {}
meters['loss'] = ScalarMeter('{}_loss/{}'.format(phase, suffix))
for k in FLAGS.topk:
meters['top{}_error'.format(k)] = ScalarMeter(
'{}_top{}_error/{}'.format(phase, k, suffix))
return meters
assert phase in ['train', 'val', 'test', 'calib'], 'Invalid phase.'
if single_sample:
meters = get_single_meter(phase)
else:
meters = get_single_meter(phase)
if phase == 'val':
meters['best_val'] = ScalarMeter('best_val')
return meters
def get_experiment_setting():
experiment_setting = []
experiment_setting.append('ptcv_pretrained_{}'.format(
getattr(FLAGS, 'ptcv_pretrained', False)))
experiment_setting.append('nvidia_pretrained_{}'.format(
getattr(FLAGS, 'nvidia_pretrained', False)))
experiment_setting.append('hawq_pretrained_{}'.format(
getattr(FLAGS, 'hawq_pretrained', False)))
experiment_setting.append('finetune_iters_{}'.format(
getattr(FLAGS, 'finetune_iters', float('inf'))))
experiment_setting.append('bn_calib_before_test_{}'.format(
getattr(FLAGS, 'bn_calib_before_test', False)))
experiment_setting.append('bn_calib_batch_num_{}'.format(
getattr(FLAGS, 'bn_calib_batch_num', -1)))
experiment_setting.append('quant_avgpool_{}'.format(
getattr(FLAGS, 'quant_avgpool', False)))
experiment_setting.append('pool_fusing_{}'.format(
getattr(FLAGS, 'pool_fusing', False)))
experiment_setting.append(
f'weight_format_wl_{FLAGS.weight_format[0]}_fl_{FLAGS.weight_format[1]}'
)
experiment_setting.append(
f'input_format_wl_{FLAGS.input_format[0]}_fl_{FLAGS.input_format[1]}')
experiment_setting.append('rescale_forward_{}'.format(
getattr(FLAGS, 'rescale_forward', True)))
experiment_setting.append('rescale_forward_conv_{}'.format(
getattr(FLAGS, 'rescale_forward_conv', False)))
experiment_setting.append('rescale_type_{}'.format(
getattr(FLAGS, 'rescale_type', 'constant')))
experiment_setting.append('input_fraclen_sharing_{}'.format(
getattr(FLAGS, 'input_fraclen_sharing', False)))
experiment_setting.append('floating_model_{}'.format(
getattr(FLAGS, 'floating_model', False)))
experiment_setting.append('floating_wo_clip_{}'.format(
getattr(FLAGS, 'floating_wo_clip', False)))
experiment_setting.append('no_clipping_{}'.format(
getattr(FLAGS, 'no_clipping', False)))
if getattr(FLAGS, 'fp_pretrained_file', ''):
fp_pretrained = True
else:
fp_pretrained = False
experiment_setting.append(f'fp_pretrained_{fp_pretrained}')
experiment_setting.append('format_type_{}'.format(
getattr(FLAGS, 'format_type', None)))
experiment_setting.append('format_from_metric_{}'.format(
getattr(FLAGS, 'format_from_metric', False)))
experiment_setting.append('momentum_for_metric_{}'.format(
getattr(FLAGS, 'momentum_for_metric', 0.1)))
experiment_setting.append('metric_{}'.format(getattr(
FLAGS, 'metric', None)))
experiment_setting.append('format_grid_search_{}'.format(
getattr(FLAGS, 'format_grid_search', False)))
experiment_setting.append('bn_momentum_{}'.format(
getattr(FLAGS, 'bn_momentum', 0.1)))
experiment_setting.append('bn_eps_{}'.format(getattr(FLAGS, 'bn_eps',
0.1)))
experiment_setting.append('lr_{}'.format(FLAGS.lr))
experiment_setting.append('weight_decay_scheme_{}'.format(
getattr(FLAGS, 'weight_decay_scheme', 'all')))
experiment_setting.append('normalize_{}'.format(
getattr(FLAGS, 'normalize', False)))
experiment_setting.append('warmup_epochs_{}'.format(FLAGS.warmup_epochs))
experiment_setting.append('weight_decay_{}'.format(FLAGS.weight_decay))
experiment_setting = os.path.join(*experiment_setting)
mprint('Experiment settings: {}'.format(experiment_setting))
return experiment_setting
def forward_loss(model, criterion, input, target, meter, train=False):
"""forward model and return loss"""
if getattr(FLAGS, 'normalize', False):
assert getattr(FLAGS, 'ptcv_pretrained', False) or getattr(
FLAGS, 'nvidia_pretrained', False) or getattr(
FLAGS, 'hawq_pretrained', False)
if getattr(model, 'int_op_only', False):
input_fraclen = model.head[0].input_fraclen
input_symmetric = model.head[0].input_symmetric
input = (fix_quant(input, 8, input_fraclen * 1.0, 1,
input_symmetric)[0] * (2**input_fraclen)).int()
setattr(input, 'output_fraclen', input_fraclen.item())
else:
assert torch.all(input >= 0)
if getattr(model, 'int_op_only', False):
input = (255 * input).round_().int()
setattr(input, 'output_fraclen', 8)
else:
input = (255 * input).round_() / 256
output = model(input)
loss = torch.mean(criterion(output, target))
# topk
_, pred = output.topk(max(FLAGS.topk))
pred = pred.t()
correct = pred.eq(target.view(1, -1).expand_as(pred))
correct_k = []
for k in FLAGS.topk:
correct_k.append(correct[:k].float().sum(0))
res = torch.cat([loss.view(1)] + correct_k, dim=0)
if getattr(FLAGS, 'distributed', False) and getattr(
FLAGS, 'distributed_all_reduce', False):
res = dist_all_reduce_tensor(res)
res = res.cpu().detach().numpy()
bs = (res.size - 1) // len(FLAGS.topk)
for i, k in enumerate(FLAGS.topk):
error_list = list(1. - res[1 + i * bs:1 + (i + 1) * bs])
if meter is not None:
meter['top{}_error'.format(k)].cache_list(error_list)
if getattr(FLAGS, 'print_each_iter', False):
mprint(f'top{k} err: {sum(error_list) / len(error_list)}.')
if meter is not None:
meter['loss'].cache(res[0])
return loss
@timing
def run_one_epoch(epoch,
loader,
model,
criterion,
optimizer,
meters,
phase='train',
scheduler=None):
"""run one epoch for train/val/test/calib"""
t_start = time.time()
assert phase in ['train', 'val', 'test',
'calib'], "phase not be in train/val/test/calib."
train = phase == 'train'
if train:
model.train()
else:
model.eval()
if phase == 'calib':
model.apply(bn_calib)
if getattr(FLAGS, 'distributed', False):
loader.sampler.set_epoch(epoch)
skip_print = False
for batch_idx, (input, target) in enumerate(loader):
if phase == 'calib':
if batch_idx == getattr(FLAGS, 'bn_calib_batch_num', -1):
break
if train and (batch_idx >= getattr(FLAGS, 'finetune_iters',
float('inf'))):
if getattr(FLAGS, 'finetune_iters', float('inf')) == 0:
skip_print = True
break
if not getattr(FLAGS, 'int_op_only', False):
target = target.cuda(non_blocking=True)
if train:
if FLAGS.lr_scheduler == 'linear_decaying':
linear_decaying_per_step = (FLAGS.lr / FLAGS.num_epochs /
len(loader.dataset) *
FLAGS.batch_size)
for param_group in optimizer.param_groups:
param_group['lr'] -= linear_decaying_per_step
# For PyTorch 1.1+, comment the following two line
#if FLAGS.lr_scheduler in ['exp_decaying_iter', 'gaussian_iter', 'cos_annealing_iter', 'butterworth_iter', 'mixed_iter', 'multistep_iter']:
# scheduler.step()
optimizer.zero_grad()
loss = forward_loss(model,
criterion,
input,
target,
meters,
train=True)
loss.backward()
if getattr(FLAGS, 'distributed', False) and getattr(
FLAGS, 'distributed_all_reduce', False):
allreduce_grads(model)
optimizer.step()
# For PyTorch 1.0 or earlier, comment the following two lines
if FLAGS.lr_scheduler in [
'exp_decaying_iter', 'cos_annealing_iter', 'multistep_iter'
]:
scheduler.step()
else: #not train
forward_loss(model, criterion, input, target, meters)
val_top1 = None
if is_master() and meters is not None and not skip_print:
results = flush_scalar_meters(meters)
mprint('{:.1f}s\t{}\t{}/{}: '.format(time.time() - t_start, phase,
epoch, FLAGS.num_epochs) +
', '.join('{}: {}'.format(k, v) for k, v in results.items()))
val_top1 = results['top1_error']
return val_top1
@timing
def train_val_test():
"""train and val"""
torch.backends.cudnn.benchmark = True
# init distributed
if getattr(FLAGS, 'distributed', False):
gpu_id = init_dist(init_method=getattr(FLAGS, 'dist_url', None))
else:
gpu_id = None
# seed
if getattr(FLAGS, 'use_diff_seed',
False) and not getattr(FLAGS, 'stoch_valid', False):
mprint('use diff seed is True')
while not is_initialized():
mprint('Waiting for initialization ...')
time.sleep(5)
mprint('Expected seed: {}'.format(
getattr(FLAGS, 'random_seed', 0) + get_rank()))
set_random_seed(getattr(FLAGS, 'random_seed', 0) + get_rank())
else:
set_random_seed()
# experiment setting
experiment_setting = get_experiment_setting()
# model
model, model_wrapper = get_model(gpu_id=gpu_id)
if getattr(FLAGS, 'int_op_only', False):
criterion = torch.nn.CrossEntropyLoss(reduction='none')
else:
criterion = torch.nn.CrossEntropyLoss(reduction='none').cuda()
# data
train_transforms, val_transforms, test_transforms = data_transforms()
train_set, val_set, test_set = dataset(train_transforms, val_transforms,
test_transforms)
train_loader, val_loader, test_loader = data_loader(
train_set, val_set, test_set)
log_dir = FLAGS.log_dir
log_dir = os.path.join(log_dir, experiment_setting)
# full precision pretrained
if getattr(FLAGS, 'fp_pretrained_file', ''):
checkpoint = torch.load(FLAGS.fp_pretrained_file,
map_location=lambda storage, loc: storage)
# update keys from external models
if type(checkpoint) == dict and 'model' in checkpoint:
checkpoint = checkpoint['model']
if getattr(FLAGS, 'pretrained_model_remap_keys', False):
new_checkpoint = {}
new_keys = list(model_wrapper.state_dict().keys())
old_keys = list(checkpoint.keys())
for key_new, key_old in zip(new_keys, old_keys):
new_checkpoint[key_new] = checkpoint[key_old]
mprint('remap {} to {}'.format(key_new, key_old))
checkpoint = new_checkpoint
model_dict = model_wrapper.state_dict()
for k in list(checkpoint.keys()):
if k not in model_dict.keys():
mprint(f'Pop out key {k} from checkpoint keys.')
checkpoint.pop(k)
model_dict.update(checkpoint)
model_wrapper.load_state_dict(model_dict)
mprint('Loaded full precision model {}.'.format(
FLAGS.fp_pretrained_file))
if getattr(FLAGS, 'ptcv_pretrained', False) or getattr(
FLAGS, 'hawq_pretrained', False):
ptcv_load(model)
if getattr(FLAGS, 'ptcv_pretrained', False):
mprint('Loaded ptcv pretrained model.')
if getattr(FLAGS, 'hawq_pretrained', False):
mprint('Loaded hawq pretrained model.')
if getattr(FLAGS, 'nvidia_pretrained', False):
nvidia_load(model)
mprint('Loaded nvidia pretrained model.')
# check pretrained
if FLAGS.pretrained_file:
checkpoint = torch.load(FLAGS.pretrained_file,
map_location=lambda storage, loc: storage)
# update keys from external models
if type(checkpoint) == dict and 'model' in checkpoint:
checkpoint = checkpoint['model']
if getattr(FLAGS, 'pretrained_model_remap_keys', False):
new_checkpoint = {}
new_keys = list(model_wrapper.state_dict().keys())
old_keys = list(checkpoint.keys())
for key_new, key_old in zip(new_keys, old_keys):
new_checkpoint[key_new] = checkpoint[key_old]
mprint('remap {} to {}'.format(key_new, key_old))
checkpoint = new_checkpoint
model_wrapper.load_state_dict(checkpoint)
mprint('Loaded model {}.'.format(FLAGS.pretrained_file))
optimizer = get_optimizer(model_wrapper)
if getattr(FLAGS, 'integize', False):
assert not getattr(FLAGS, 'int_op_only', False)
model = model.int_model().to(next(model.parameters()).device)
if getattr(FLAGS, 'distributed', False):
if getattr(FLAGS, 'distributed_all_reduce', False):
model_wrapper = AllReduceDistributedDataParallel(model.cuda())
else:
model_wrapper = torch.nn.parallel.DistributedDataParallel(
model.cuda(), [gpu_id],
gpu_id,
find_unused_parameters=True)
else:
model_wrapper = torch.nn.DataParallel(model).cuda()
integize_file_path = getattr(FLAGS, 'integize_file_path', '')
if integize_file_path:
os.makedirs(os.path.join(integize_file_path, 'checkpoints'),
exist_ok=True)
os.makedirs(os.path.join(integize_file_path, 'onnx_files'),
exist_ok=True)
if integize_file_path:
integize_checkpoint = os.path.join(integize_file_path,
'checkpoints',
'integized_model.pt')
torch.save({'model': model_wrapper.state_dict},
integize_checkpoint)
integize_onnx_file = os.path.join(integize_file_path, 'onnx_files',
'integized_model.onnx')
data_shape = [1, 3, FLAGS.image_size, FLAGS.image_size]
data_dtype = torch.float
onnx_export(model, data_shape, data_dtype,
next(model.parameters()).device, integize_onnx_file)
mprint(
f'Integized model checkpoint and onnx file saved to {integize_file_path}.'
)
if getattr(FLAGS, 'int_op_only', False):
assert not getattr(FLAGS, 'integize', False)
model.apply(lambda m: setattr(m, 'int_op_only', True))
model = model.int_model().cpu()
model.apply(lambda m: setattr(m, 'int_op_only', True))
model_wrapper = model
int_op_only_file_path = getattr(FLAGS, 'int_op_only_file_path', '')
if int_op_only_file_path:
os.makedirs(os.path.join(int_op_only_file_path, 'checkpoints'),
exist_ok=True)
os.makedirs(os.path.join(int_op_only_file_path, 'onnx_files'),
exist_ok=True)
if int_op_only_file_path:
int_op_only_checkpoint = os.path.join(int_op_only_file_path,
'checkpoints',
'int_op_only_model.pt')
torch.save({'model': model_wrapper.state_dict},
int_op_only_checkpoint)
int_op_only_onnx_file = os.path.join(int_op_only_file_path,
'onnx_files',
'int_op_only_model.onnx')
data_shape = [1, 3, FLAGS.image_size, FLAGS.image_size]
data_dtype = torch.int
onnx_export(model, data_shape, data_dtype,
next(model.parameters()).device, int_op_only_onnx_file)
mprint(
f'Integer operation only model checkpoint and onnx file saved to {int_op_only_file_path}.'
)
if FLAGS.test_only and (test_loader is not None):
mprint('Start testing.')
test_meters = get_meters('test')
with torch.no_grad():
run_one_epoch(-1,
test_loader,
model_wrapper,
criterion,
optimizer,
test_meters,
phase='test')
if is_master():
for n, m in model.named_modules():
if isinstance(m, (ReLUClipFXQConvBN, ReLUClipFXQLinear)):
mprint(f'layer name: {n}.')
mprint(f'alpha: {m.alpha}.')
if m.get_master_layer() is not None:
mprint(
f'master layer alpha: {m.get_master_layer().get_alpha()}.'
)
else:
mprint(f'master layer: {m.get_master_layer()}.')
mprint(f'alpha in use: {m.get_alpha()}.')
mprint(f'weight format: {m.get_weight_format()}.')
mprint(f'input format: {m.get_input_format()}.')
mprint(f'fix scaling: {m.fix_scaling}.')
if isinstance(m, (ReLUClipFXQConvBN, )):
mprint(
f'following layer fix scaling: {m.get_following_layer().fix_scaling}.'
)
if m.format_from_metric or m.format_grid_search:
mprint(f'input_fraclen: {m.get_input_fraclen()}.')
mprint(f'weight_fraclen: {m.get_weight_fraclen()}.')
return
# check resume training
if os.path.exists(os.path.join(log_dir, 'latest_checkpoint.pt')):
checkpoint = torch.load(os.path.join(log_dir, 'latest_checkpoint.pt'),
map_location=lambda storage, loc: storage)
model_wrapper.load_state_dict(checkpoint['model'])
optimizer.load_state_dict(checkpoint['optimizer'])
last_epoch = checkpoint['last_epoch']