-
Notifications
You must be signed in to change notification settings - Fork 2
/
cusvgg.py
539 lines (439 loc) · 22.2 KB
/
cusvgg.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
import torch.nn as nn
import numpy as np
import torch
from resnet import BinaryConv2d
def conv_bn(in_channels, out_channels, kernel_size, stride, padding, groups=1):
result = nn.Sequential()
result.add_module('conv', nn.Conv2d(in_channels=in_channels, out_channels=out_channels,
kernel_size=kernel_size, stride=stride, padding=padding, groups=groups,
bias=False))
result.add_module('bn', nn.BatchNorm2d(num_features=out_channels))
return result
class RepVGGBlock(nn.Module):
def __init__(self, in_channels, out_channels, kernel_size,
stride=1, padding=0, dilation=1, groups=1, padding_mode='zeros', deploy=False):
super(RepVGGBlock, self).__init__()
self.deploy = deploy
self.groups = groups
self.in_channels = in_channels
assert kernel_size == 3
assert padding == 1
padding_11 = padding - kernel_size // 2
self.nonlinearity = nn.ReLU()
# self.scale = BinaryConv2d(out_channels, out_channels, kernel_size=1, stride=1, padding=0,
# groups=out_channels, bias=False)
if deploy:
self.rbr_reparam = nn.Conv2d(in_channels=in_channels, out_channels=out_channels, kernel_size=kernel_size,
stride=stride,
padding=padding, dilation=dilation, groups=groups, bias=True,
padding_mode=padding_mode)
else:
self.rbr_identity = nn.BatchNorm2d(
num_features=in_channels) if out_channels == in_channels and stride == 1 else None
self.rbr_dense = conv_bn(in_channels=in_channels, out_channels=out_channels, kernel_size=kernel_size,
stride=stride, padding=padding, groups=groups)
self.rbr_1x1 = conv_bn(in_channels=in_channels, out_channels=out_channels, kernel_size=1, stride=stride,
padding=padding_11, groups=groups)
self.scale = BinaryConv2d(out_channels, out_channels, kernel_size=1, stride=1, padding=0,
groups=out_channels, bias=False)
# print('RepVGG Block, identity = ', self.rbr_identity)
def forward(self, inputs):
if hasattr(self, 'rbr_reparam'):
return self.nonlinearity(self.rbr_reparam(inputs))
if self.rbr_identity is None:
id_out = 0
else:
id_out = self.rbr_identity(inputs)
return self.scale(self.nonlinearity(self.rbr_dense(inputs) + self.rbr_1x1(inputs) + id_out))
# This func derives the equivalent kernel and bias in a DIFFERENTIABLE way.
# You can get the equivalent kernel and bias at any time and do whatever you want,
# for example, apply some penalties or constraints during training, just like you do to the other models.
# May be useful for quantization or pruning.
def get_equivalent_kernel_bias(self):
kernel3x3, bias3x3 = self._fuse_bn_tensor(self.rbr_dense)
kernel1x1, bias1x1 = self._fuse_bn_tensor(self.rbr_1x1)
kernelid, biasid = self._fuse_bn_tensor(self.rbr_identity)
return kernel3x3 + self._pad_1x1_to_3x3_tensor(kernel1x1) + kernelid, bias3x3 + bias1x1 + biasid
def _pad_1x1_to_3x3_tensor(self, kernel1x1):
if kernel1x1 is None:
return 0
else:
return torch.nn.functional.pad(kernel1x1, [1, 1, 1, 1])
def _fuse_bn_tensor(self, branch):
if branch is None:
return 0, 0
if isinstance(branch, nn.Sequential):
kernel = branch.conv.weight
running_mean = branch.bn.running_mean
running_var = branch.bn.running_var
gamma = branch.bn.weight
beta = branch.bn.bias
eps = branch.bn.eps
else:
assert isinstance(branch, nn.BatchNorm2d)
if not hasattr(self, 'id_tensor'):
input_dim = self.in_channels // self.groups
kernel_value = np.zeros((self.in_channels, input_dim, 3, 3), dtype=np.float32)
for i in range(self.in_channels):
kernel_value[i, i % input_dim, 1, 1] = 1
self.id_tensor = torch.from_numpy(kernel_value).to(branch.weight.device)
kernel = self.id_tensor
running_mean = branch.running_mean
running_var = branch.running_var
gamma = branch.weight
beta = branch.bias
eps = branch.eps
std = (running_var + eps).sqrt()
t = (gamma / std).reshape(-1, 1, 1, 1)
return kernel * t, beta - running_mean * gamma / std
def repvgg_convert(self):
kernel, bias = self.get_equivalent_kernel_bias()
return kernel.detach().cpu().numpy(), bias.detach().cpu().numpy()
def switch_to_deploy(self):
if hasattr(self, 'rbr_reparam'):
return
kernel, bias = self.get_equivalent_kernel_bias()
self.rbr_reparam = nn.Conv2d(in_channels=self.rbr_dense.conv.in_channels,
out_channels=self.rbr_dense.conv.out_channels,
kernel_size=self.rbr_dense.conv.kernel_size, stride=self.rbr_dense.conv.stride,
padding=self.rbr_dense.conv.padding, dilation=self.rbr_dense.conv.dilation,
groups=self.rbr_dense.conv.groups, bias=True)
self.rbr_reparam.weight.data = kernel
self.rbr_reparam.bias.data = bias
for para in self.parameters():
para.detach_()
self.__delattr__('rbr_dense')
self.__delattr__('rbr_1x1')
if hasattr(self, 'rbr_identity'):
self.__delattr__('rbr_identity')
class RepVGG(nn.Module):
def __init__(self, num_blocks, num_classes=1000, width_multiplier=None, override_groups_map=None, deploy=False):
super(RepVGG, self).__init__()
assert len(width_multiplier) == 4
self.deploy = deploy
self.override_groups_map = override_groups_map or dict()
assert 0 not in self.override_groups_map
self.in_planes = min(64, int(64 * width_multiplier[0]))
self.stage0 = RepVGGBlock(in_channels=3, out_channels=self.in_planes, kernel_size=3, stride=2, padding=1,
deploy=self.deploy)
self.cur_layer_idx = 1
self.stage1 = self._make_stage(int(64 * width_multiplier[0]), num_blocks[0], stride=2)
self.stage2 = self._make_stage(int(128 * width_multiplier[1]), num_blocks[1], stride=2)
self.stage3 = self._make_stage(int(256 * width_multiplier[2]), num_blocks[2], stride=2)
self.stage4 = self._make_stage(int(512 * width_multiplier[3]), num_blocks[3], stride=2)
self.gap = nn.AdaptiveAvgPool2d(output_size=1)
self.linear = nn.Linear(int(512 * width_multiplier[3]), num_classes)
if not self.deploy:
for m in self.modules():
if isinstance(m, RepVGGBlock):
nn.init.constant_(m.scale.weight, 1)
def _make_stage(self, planes, num_blocks, stride):
strides = [stride] + [1] * (num_blocks - 1)
blocks = []
for stride in strides:
cur_groups = self.override_groups_map.get(self.cur_layer_idx, 1)
blocks.append(RepVGGBlock(in_channels=self.in_planes, out_channels=planes, kernel_size=3,
stride=stride, padding=1, groups=cur_groups, deploy=self.deploy))
self.in_planes = planes
self.cur_layer_idx += 1
return nn.Sequential(*blocks)
def forward(self, x):
out = self.stage0(x)
out = self.stage1(out)
out = self.stage2(out)
out = self.stage3(out)
out = self.stage4(out)
out = self.gap(out)
out = out.reshape(out.size(0), -1)
out = self.linear(out)
return out
class NewRepVGG(nn.Module):
def __init__(self, num_blocks, num_classes=1000, config=None, override_groups_map=None, deploy=False):
super(NewRepVGG, self).__init__()
# assert len(width_multiplier) == 4
self.deploy = deploy
self.override_groups_map = override_groups_map or dict()
assert 0 not in self.override_groups_map
self.in_planes = config[0][0]
self.stage0 = RepVGGBlock(in_channels=3, out_channels=self.in_planes, kernel_size=3, stride=2, padding=1,
deploy=self.deploy)
self.cur_layer_idx = 1
self.stage1 = self._make_stage(config[1], len(config[1]), stride=2)
self.stage2 = self._make_stage(config[2], len(config[2]), stride=2)
self.stage3 = self._make_stage(config[3], len(config[3]), stride=2)
self.stage4 = self._make_stage(config[4], len(config[4]), stride=2)
self.gap = nn.AdaptiveAvgPool2d(output_size=1)
self.linear = nn.Linear(config[4][-1], num_classes)
# for m in self.modules():
# if isinstance(m, RepVGGBlock):
# nn.init.constant_(m.scale.weight, 1)
def _make_stage(self, planes, num_blocks, stride):
strides = [stride] + [1] * (num_blocks - 1)
blocks = []
for i in range(len(strides)):
cur_groups = self.override_groups_map.get(self.cur_layer_idx, 1)
blocks.append(RepVGGBlock(in_channels=self.in_planes, out_channels=planes[i], kernel_size=3,
stride=strides[i], padding=1, groups=cur_groups, deploy=self.deploy))
self.in_planes = planes[i]
self.cur_layer_idx += 1
return nn.Sequential(*blocks)
def forward(self, x):
out = self.stage0(x)
out = self.stage1(out)
out = self.stage2(out)
out = self.stage3(out)
out = self.stage4(out)
out = self.gap(out)
out = out.reshape(out.size(0), -1)
out = self.linear(out)
return out
optional_groupwise_layers = [2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 22, 24, 26]
g2_map = {l: 2 for l in optional_groupwise_layers}
g4_map = {l: 4 for l in optional_groupwise_layers}
def create_RepVGG_A0(deploy=False):
return RepVGG(num_blocks=[2, 4, 14, 1], num_classes=1000,
width_multiplier=[0.75, 0.75, 0.75, 2.5], override_groups_map=None, deploy=deploy)
def create_RepVGG_A1(deploy=False):
return RepVGG(num_blocks=[2, 4, 14, 1], num_classes=1000,
width_multiplier=[1, 1, 1, 2.5], override_groups_map=None, deploy=deploy)
def create_RepVGG_A2(deploy=False):
return RepVGG(num_blocks=[2, 4, 14, 1], num_classes=1000,
width_multiplier=[1.5, 1.5, 1.5, 2.75], override_groups_map=None, deploy=deploy)
def create_RepVGG_B0(deploy=False):
return RepVGG(num_blocks=[4, 6, 16, 1], num_classes=1000,
width_multiplier=[1, 1, 1, 2.5], override_groups_map=None, deploy=deploy)
def create_RepVGG_B1(deploy=False):
return RepVGG(num_blocks=[4, 6, 16, 1], num_classes=1000,
width_multiplier=[2, 2, 2, 4], override_groups_map=None, deploy=deploy)
def create_RepVGG_B1g2(deploy=False):
return RepVGG(num_blocks=[4, 6, 16, 1], num_classes=1000,
width_multiplier=[2, 2, 2, 4], override_groups_map=g2_map, deploy=deploy)
def create_RepVGG_B1g4(deploy=False):
return RepVGG(num_blocks=[4, 6, 16, 1], num_classes=1000,
width_multiplier=[2, 2, 2, 4], override_groups_map=g4_map, deploy=deploy)
def create_RepVGG_B2(deploy=False):
return RepVGG(num_blocks=[4, 6, 16, 1], num_classes=1000,
width_multiplier=[2.5, 2.5, 2.5, 5], override_groups_map=None, deploy=deploy)
def create_RepVGG_B2g2(deploy=False):
return RepVGG(num_blocks=[4, 6, 16, 1], num_classes=1000,
width_multiplier=[2.5, 2.5, 2.5, 5], override_groups_map=g2_map, deploy=deploy)
def create_RepVGG_B2g4(deploy=False):
return RepVGG(num_blocks=[4, 6, 16, 1], num_classes=1000,
width_multiplier=[2.5, 2.5, 2.5, 5], override_groups_map=g4_map, deploy=deploy)
def create_RepVGG_B3(deploy=False):
return RepVGG(num_blocks=[4, 6, 16, 1], num_classes=1000,
width_multiplier=[3, 3, 3, 5], override_groups_map=None, deploy=deploy)
def create_RepVGG_B3g2(deploy=False):
return RepVGG(num_blocks=[4, 6, 16, 1], num_classes=1000,
width_multiplier=[3, 3, 3, 5], override_groups_map=g2_map, deploy=deploy)
def create_RepVGG_B3g4(deploy=False):
return RepVGG(num_blocks=[4, 6, 16, 1], num_classes=1000,
width_multiplier=[3, 3, 3, 5], override_groups_map=g4_map, deploy=deploy)
def create_RepVGG_config(config=None, deploy=True):
return NewRepVGG(num_blocks=[4, 6, 16, 1], num_classes=1000,
config=[[8], [20, 16, 15, 10], [49, 41, 36, 34, 39, 31],
[125, 107, 102, 97, 106, 123, 127, 170, 196, 202, 232,
269, 315, 425, 475, 450], [2048]],
override_groups_map=None, deploy=deploy)
# return NewRepVGG(num_blocks=[4, 6, 16, 1], num_classes=1000,
# config=[[11], [29, 32, 27, 23], [68, 57, 55, 65, 79, 79],
# [208, 198, 193, 175, 201, 240, 254, 323, 357, 378, 412, 442, 483, 496, 510, 487],
# [2048]],
# override_groups_map=None, deploy=deploy)
func_dict = {
'RepVGG-A0': create_RepVGG_A0,
'RepVGG-A1': create_RepVGG_A1,
'RepVGG-A2': create_RepVGG_A2,
'RepVGG-B0': create_RepVGG_B0,
'RepVGG-B1': create_RepVGG_B1,
'RepVGG-B1g2': create_RepVGG_B1g2,
'RepVGG-B1g4': create_RepVGG_B1g4,
'RepVGG-B2': create_RepVGG_B2,
'RepVGG-B2g2': create_RepVGG_B2g2,
'RepVGG-B2g4': create_RepVGG_B2g4,
'RepVGG-B3': create_RepVGG_B3,
'RepVGG-B3g2': create_RepVGG_B3g2,
'RepVGG-B3g4': create_RepVGG_B3g4,
'RepVGG-config': create_RepVGG_config,
}
def get_RepVGG_func_by_name(name):
return func_dict[name]
# Use this for converting a customized model with RepVGG as one of its components
# (e.g., the backbone of a semantic segmentation model)
# The use case will be like
# 1. Build train_model. For example, build a PSPNet with a training-time RepVGG as backbone
# 2. Train train_model or do whatever you want
# 3. Build deploy_model. In the above example, that will be a PSPNet with an inference-time RepVGG as backbone
# 4. Call this func
# ====================== the pseudo code will be like
# train_backbone = create_RepVGG_B2(deploy=False)
# train_backbone.load_state_dict(torch.load('RepVGG-B2-train.pth'))
# train_pspnet = build_pspnet(backbone=train_backbone)
# segmentation_train(train_pspnet)
# deploy_backbone = create_RepVGG_B2(deploy=True)
# deploy_pspnet = build_pspnet(backbone=deploy_backbone)
# whole_model_convert(train_pspnet, deploy_pspnet)
# segmentation_test(deploy_pspnet)
def whole_model_convert(train_model: torch.nn.Module, deploy_model: torch.nn.Module, save_path=None):
all_weights = {}
for name, module in train_model.named_modules():
if hasattr(module, 'repvgg_convert'):
kernel, bias = module.repvgg_convert()
all_weights[name + '.rbr_reparam.weight'] = kernel
all_weights[name + '.rbr_reparam.bias'] = bias
print('convert RepVGG block')
else:
for p_name, p_tensor in module.named_parameters():
full_name = name + '.' + p_name
if full_name not in all_weights:
all_weights[full_name] = p_tensor.detach().cpu().numpy()
for p_name, p_tensor in module.named_buffers():
full_name = name + '.' + p_name
if full_name not in all_weights:
all_weights[full_name] = p_tensor.cpu().numpy()
deploy_model.load_state_dict(all_weights)
if save_path is not None:
torch.save(deploy_model.state_dict(), save_path)
return deploy_model
# Use this when converting a RepVGG without customized structures.
# train_model = create_RepVGG_A0(deploy=False)
# train train_model
# deploy_model = repvgg_model_convert(train_model, create_RepVGG_A0, save_path='repvgg_deploy.pth')
def repvgg_model_convert(model: torch.nn.Module, build_func, save_path=None):
converted_weights = {}
scales = {}
for name, module in model.named_modules():
if hasattr(module, 'repvgg_convert'):
kernel, bias = module.repvgg_convert()
converted_weights[name + '.rbr_reparam.weight'] = kernel
converted_weights[name + '.rbr_reparam.bias'] = bias
elif isinstance(module, torch.nn.Linear):
converted_weights[name + '.weight'] = module.weight.detach().cpu().numpy()
converted_weights[name + '.bias'] = module.bias.detach().cpu().numpy()
for name, params in model.named_parameters():
if 'scale' in name:
scales[name] = params
del model
deploy_model = build_func(deploy=True)
for name, param in deploy_model.named_parameters():
if 'scale' not in name:
print('deploy param: ', name, param.size(), np.mean(converted_weights[name]))
param.data = torch.from_numpy(converted_weights[name]).float()
else:
param.data = scales[name]
if save_path is not None:
torch.save(deploy_model.state_dict(), save_path)
return deploy_model
def squeeze_weight(i, origin, indices, bias):
if i == 0:
select = torch.nonzero(indices[i]).squeeze()
out = torch.index_select(origin, 0, select)
else:
if i != len(indices)-1:
select1 = torch.nonzero(indices[i]).squeeze()
out = torch.index_select(origin, 0, select1)
else:
out = origin
if not bias:
select2 = torch.nonzero(indices[i - 1]).squeeze()
out = torch.index_select(out, 1, select2)
return out
def create_model_from_checkpoint(architecture, checkpoint, deploy=False):
# load checkpoint
state_dict = torch.load(checkpoint, map_location='cpu')
if 'state_dict' in state_dict:
state_dict = state_dict['state_dict']
ckpt = {k.replace('module.', ''): v for k, v in state_dict.items()} # strip the names
# build checkpoint model
repvgg_build_func = get_RepVGG_func_by_name(architecture)
model = repvgg_build_func(deploy=deploy)
model.load_state_dict(ckpt)
# convert model to deploy VGG shaped
deploy_model = repvgg_model_convert(model, repvgg_build_func, save_path=None)
# store weights waiting to be squeezed
weights = {}
for name, params in deploy_model.named_parameters():
if 'reparam' in name or 'linear' in name:
weights[name] = params
# pull index from 'conv1x1' in origin checkpoint
width = []
indices = []
for layer in ckpt:
if 'scale' in layer:
weight = ckpt[layer]
binary = (weight > 0.5).float()
width.append(torch.count_nonzero(binary).item())
indices.append(binary.squeeze())
# build final fine-shaped model
final_model_build = get_RepVGG_func_by_name('RepVGG-config')
final_model = final_model_build(deploy=True)
# load parameters
i = 0
for name, params in final_model.named_parameters():
if 'reparam' in name:
squeezed = squeeze_weight(i, weights[name], indices, bias='bias' in name)
params.data = squeezed
print('processing: ', name, params.data.size())
if 'bias' in name:
i += 1
if 'linear' in name:
params.data = ckpt[name]
print('processing: ', name, params.data.size())
return model, final_model
if __name__ == '__main__':
# state_dict = torch.load('B1_4_7487_9216.pth.tar', map_location='cpu')
# if 'state_dict' in state_dict:
# state_dict = state_dict['state_dict']
# for layer in state_dict:
# if 'scale' in layer:
# state_dict[layer] = (state_dict[layer] > 0.5).float()
# print(layer, 'remained: ', (torch.sum(state_dict[layer]) / (state_dict[layer]).size(0)).item())
# torch.save(state_dict, 'B1_3G_convert_checkpoint.pt')
# config = [[11], [29, 32, 27, 23], [68, 57, 55, 65, 79, 79],
# [208, 198, 193, 175, 201, 240, 254, 323, 357, 378, 412, 442, 483, 496, 510, 487], [2048]]
# for i in config:
# print(i)
# print(len(i))
# print(config[3])
from prunesearch import ProfileConv
from prunesearch import bn_prune
x = torch.randn(1, 3, 224, 224)
# indice = torch.tensor([1, 0, 1])
# nz = torch.nonzero(indice).squeeze()
# y = torch.index_select(x, 1, nz)
# print(x, y)
# repvgg_build_func = get_RepVGG_func_by_name('RepVGG-config')
# model = repvgg_build_func(deploy=False)
deploy, model = create_model_from_checkpoint('RepVGG-B1', 'checkpoint.pth.tar', )
# print(model)
deploy.eval()
model.eval()
out1 = deploy(x)
out2 = model(x)
print(torch.sum((out2 - out1) ** 2))
# torch.onnx.export(model, x, "B1__G.onnx", verbose=True)
# print('onnx exported')
profile = ProfileConv(model)
MACs = profile(torch.randn(1, 3, 224, 224))
print(len(MACs))
print(sum(MACs) / 1e9, 'GMACs, only consider conv layers')
# model = create_model_from_checkpoint('RepVGG-B1', 'checkpoints/B1_4_7487_9216.pth.tar', )
# # print(model)
#
# width = []
# indices = []
#
# for name, params in model.named_parameters():
# if 'scale' in name:
# weight = params.detach()
# binary = (weight > 0.5).float()
# width.append(torch.count_nonzero(binary).item())
# indices.append(binary.squeeze())
#
# print(width)
# print(indices)
pass
# profile = ProfileConv(model)
# MACs = profile(torch.randn(1, 3, 224, 224))
# print(len(MACs))
# print(sum(MACs) / 1e9, 'GMACs, only consider conv layers')