resnet.py 20.5 KB
Newer Older
C
cuicheng01 已提交
1
# copyright (c) 2021 PaddlePaddle Authors. All Rights Reserve.
C
cuicheng01 已提交
2 3 4 5 6 7 8 9 10 11 12 13 14
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#    http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

C
cuicheng01 已提交
15
from __future__ import absolute_import, division, print_function
C
cuicheng01 已提交
16 17 18 19 20 21 22 23 24 25

import numpy as np
import paddle
from paddle import ParamAttr
import paddle.nn as nn
from paddle.nn import Conv2D, BatchNorm, Linear
from paddle.nn import AdaptiveAvgPool2D, MaxPool2D, AvgPool2D
from paddle.nn.initializer import Uniform
import math

C
cuicheng01 已提交
26
from ppcls.arch.backbone.base.theseus_layer import TheseusLayer
D
dongshuilong 已提交
27
from ppcls.utils.save_load import load_dygraph_pretrain, load_dygraph_pretrain_from_url
C
cuicheng01 已提交
28 29

MODEL_URLS = {
D
dongshuilong 已提交
30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51
    "ResNet18":
    "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ResNet18_pretrained.pdparams",
    "ResNet18_vd":
    "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ResNet18_vd_pretrained.pdparams",
    "ResNet34":
    "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ResNet34_pretrained.pdparams",
    "ResNet34_vd":
    "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ResNet34_vd_pretrained.pdparams",
    "ResNet50":
    "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ResNet50_pretrained.pdparams",
    "ResNet50_vd":
    "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ResNet50_vd_pretrained.pdparams",
    "ResNet101":
    "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ResNet101_pretrained.pdparams",
    "ResNet101_vd":
    "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ResNet101_vd_pretrained.pdparams",
    "ResNet152":
    "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ResNet152_pretrained.pdparams",
    "ResNet152_vd":
    "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ResNet152_vd_pretrained.pdparams",
    "ResNet200_vd":
    "https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/ResNet200_vd_pretrained.pdparams",
C
cuicheng01 已提交
52
}
C
cuicheng01 已提交
53

54 55 56 57 58 59 60 61 62
MODEL_STAGES_PATTERN = {
    "ResNet18": ["blocks[1]", "blocks[3]", "blocks[5]", "blocks[7]"],
    "ResNet34": ["blocks[2]", "blocks[6]", "blocks[12]", "blocks[15]"],
    "ResNet50": ["blocks[2]", "blocks[6]", "blocks[12]", "blocks[15]"],
    "ResNet101": ["blocks[2]", "blocks[6]", "blocks[29]", "blocks[32]"],
    "ResNet152": ["blocks[2]", "blocks[10]", "blocks[46]", "blocks[49]"],
    "ResNet200": ["blocks[2]", "blocks[14]", "blocks[62]", "blocks[65]"]
}

C
cuicheng01 已提交
63 64 65 66 67 68 69 70 71 72
__all__ = MODEL_URLS.keys()
'''
ResNet config: dict.
    key: depth of ResNet.
    values: config's dict of specific model.
        keys:
            block_type: Two different blocks in ResNet, BasicBlock and BottleneckBlock are optional.
            block_depth: The number of blocks in different stages in ResNet.
            num_channels: The number of channels to enter the next stage.
'''
C
cuicheng01 已提交
73 74
NET_CONFIG = {
    "18": {
D
dongshuilong 已提交
75 76 77 78
        "block_type": "BasicBlock",
        "block_depth": [2, 2, 2, 2],
        "num_channels": [64, 64, 128, 256]
    },
C
cuicheng01 已提交
79
    "34": {
D
dongshuilong 已提交
80 81 82 83
        "block_type": "BasicBlock",
        "block_depth": [3, 4, 6, 3],
        "num_channels": [64, 64, 128, 256]
    },
C
cuicheng01 已提交
84
    "50": {
D
dongshuilong 已提交
85 86 87 88
        "block_type": "BottleneckBlock",
        "block_depth": [3, 4, 6, 3],
        "num_channels": [64, 256, 512, 1024]
    },
C
cuicheng01 已提交
89
    "101": {
D
dongshuilong 已提交
90 91 92 93
        "block_type": "BottleneckBlock",
        "block_depth": [3, 4, 23, 3],
        "num_channels": [64, 256, 512, 1024]
    },
C
cuicheng01 已提交
94
    "152": {
D
dongshuilong 已提交
95 96 97 98
        "block_type": "BottleneckBlock",
        "block_depth": [3, 8, 36, 3],
        "num_channels": [64, 256, 512, 1024]
    },
C
cuicheng01 已提交
99
    "200": {
D
dongshuilong 已提交
100 101 102 103
        "block_type": "BottleneckBlock",
        "block_depth": [3, 12, 48, 3],
        "num_channels": [64, 256, 512, 1024]
    },
C
cuicheng01 已提交
104 105 106 107 108 109 110 111 112 113 114 115
}


class ConvBNLayer(TheseusLayer):
    def __init__(self,
                 num_channels,
                 num_filters,
                 filter_size,
                 stride=1,
                 groups=1,
                 is_vd_mode=False,
                 act=None,
littletomatodonkey's avatar
littletomatodonkey 已提交
116 117
                 lr_mult=1.0,
                 data_format="NCHW"):
C
cuicheng01 已提交
118
        super().__init__()
C
cuicheng01 已提交
119 120
        self.is_vd_mode = is_vd_mode
        self.act = act
C
cuicheng01 已提交
121
        self.avg_pool = AvgPool2D(
C
cuicheng01 已提交
122 123 124 125 126 127 128 129 130
            kernel_size=2, stride=2, padding=0, ceil_mode=True)
        self.conv = Conv2D(
            in_channels=num_channels,
            out_channels=num_filters,
            kernel_size=filter_size,
            stride=stride,
            padding=(filter_size - 1) // 2,
            groups=groups,
            weight_attr=ParamAttr(learning_rate=lr_mult),
littletomatodonkey's avatar
littletomatodonkey 已提交
131 132
            bias_attr=False,
            data_format=data_format)
C
cuicheng01 已提交
133 134 135
        self.bn = BatchNorm(
            num_filters,
            param_attr=ParamAttr(learning_rate=lr_mult),
littletomatodonkey's avatar
littletomatodonkey 已提交
136 137
            bias_attr=ParamAttr(learning_rate=lr_mult),
            data_layout=data_format)
C
cuicheng01 已提交
138 139 140 141
        self.relu = nn.ReLU()

    def forward(self, x):
        if self.is_vd_mode:
C
cuicheng01 已提交
142
            x = self.avg_pool(x)
C
cuicheng01 已提交
143 144 145 146 147 148 149 150
        x = self.conv(x)
        x = self.bn(x)
        if self.act:
            x = self.relu(x)
        return x


class BottleneckBlock(TheseusLayer):
littletomatodonkey's avatar
littletomatodonkey 已提交
151 152 153 154 155 156 157 158
    def __init__(self,
                 num_channels,
                 num_filters,
                 stride,
                 shortcut=True,
                 if_first=False,
                 lr_mult=1.0,
                 data_format="NCHW"):
C
cuicheng01 已提交
159
        super().__init__()
C
cuicheng01 已提交
160 161 162 163 164

        self.conv0 = ConvBNLayer(
            num_channels=num_channels,
            num_filters=num_filters,
            filter_size=1,
C
cuicheng01 已提交
165
            act="relu",
littletomatodonkey's avatar
littletomatodonkey 已提交
166 167
            lr_mult=lr_mult,
            data_format=data_format)
C
cuicheng01 已提交
168 169 170 171 172
        self.conv1 = ConvBNLayer(
            num_channels=num_filters,
            num_filters=num_filters,
            filter_size=3,
            stride=stride,
C
cuicheng01 已提交
173
            act="relu",
littletomatodonkey's avatar
littletomatodonkey 已提交
174 175
            lr_mult=lr_mult,
            data_format=data_format)
C
cuicheng01 已提交
176 177 178 179 180
        self.conv2 = ConvBNLayer(
            num_channels=num_filters,
            num_filters=num_filters * 4,
            filter_size=1,
            act=None,
littletomatodonkey's avatar
littletomatodonkey 已提交
181 182
            lr_mult=lr_mult,
            data_format=data_format)
C
cuicheng01 已提交
183 184 185 186 187 188 189 190

        if not shortcut:
            self.short = ConvBNLayer(
                num_channels=num_channels,
                num_filters=num_filters * 4,
                filter_size=1,
                stride=stride if if_first else 1,
                is_vd_mode=False if if_first else True,
littletomatodonkey's avatar
littletomatodonkey 已提交
191 192
                lr_mult=lr_mult,
                data_format=data_format)
C
cuicheng01 已提交
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
        self.relu = nn.ReLU()
        self.shortcut = shortcut

    def forward(self, x):
        identity = x
        x = self.conv0(x)
        x = self.conv1(x)
        x = self.conv2(x)

        if self.shortcut:
            short = identity
        else:
            short = self.short(identity)
        x = paddle.add(x=x, y=short)
        x = self.relu(x)
        return x


class BasicBlock(TheseusLayer):
    def __init__(self,
                 num_channels,
                 num_filters,
                 stride,
                 shortcut=True,
                 if_first=False,
littletomatodonkey's avatar
littletomatodonkey 已提交
218 219
                 lr_mult=1.0,
                 data_format="NCHW"):
C
cuicheng01 已提交
220 221
        super().__init__()

C
cuicheng01 已提交
222 223 224 225 226 227
        self.stride = stride
        self.conv0 = ConvBNLayer(
            num_channels=num_channels,
            num_filters=num_filters,
            filter_size=3,
            stride=stride,
C
cuicheng01 已提交
228
            act="relu",
littletomatodonkey's avatar
littletomatodonkey 已提交
229 230
            lr_mult=lr_mult,
            data_format=data_format)
C
cuicheng01 已提交
231 232 233 234 235
        self.conv1 = ConvBNLayer(
            num_channels=num_filters,
            num_filters=num_filters,
            filter_size=3,
            act=None,
littletomatodonkey's avatar
littletomatodonkey 已提交
236 237
            lr_mult=lr_mult,
            data_format=data_format)
C
cuicheng01 已提交
238 239 240 241 242 243 244
        if not shortcut:
            self.short = ConvBNLayer(
                num_channels=num_channels,
                num_filters=num_filters,
                filter_size=1,
                stride=stride if if_first else 1,
                is_vd_mode=False if if_first else True,
littletomatodonkey's avatar
littletomatodonkey 已提交
245 246
                lr_mult=lr_mult,
                data_format=data_format)
C
cuicheng01 已提交
247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263
        self.shortcut = shortcut
        self.relu = nn.ReLU()

    def forward(self, x):
        identity = x
        x = self.conv0(x)
        x = self.conv1(x)
        if self.shortcut:
            short = identity
        else:
            short = self.short(identity)
        x = paddle.add(x=x, y=short)
        x = self.relu(x)
        return x


class ResNet(TheseusLayer):
C
cuicheng01 已提交
264 265 266 267 268 269 270 271 272
    """
    ResNet
    Args:
        config: dict. config of ResNet.
        version: str="vb". Different version of ResNet, version vd can perform better. 
        class_num: int=1000. The number of classes.
        lr_mult_list: list. Control the learning rate of different stages.
    Returns:
        model: nn.Layer. Specific ResNet model depends on args.
C
cuicheng01 已提交
273
    """
D
dongshuilong 已提交
274

C
cuicheng01 已提交
275 276
    def __init__(self,
                 config,
277
                 stages_pattern,
C
cuicheng01 已提交
278 279
                 version="vb",
                 class_num=1000,
littletomatodonkey's avatar
littletomatodonkey 已提交
280 281
                 lr_mult_list=[1.0, 1.0, 1.0, 1.0, 1.0],
                 data_format="NCHW",
W
weishengyu 已提交
282
                 input_image_channel=3,
283 284
                 return_patterns=None,
                 return_stages=None):
C
cuicheng01 已提交
285
        super().__init__()
C
cuicheng01 已提交
286 287 288 289

        self.cfg = config
        self.lr_mult_list = lr_mult_list
        self.is_vd_mode = version == "vd"
C
cuicheng01 已提交
290 291 292 293 294 295
        self.class_num = class_num
        self.num_filters = [64, 128, 256, 512]
        self.block_depth = self.cfg["block_depth"]
        self.block_type = self.cfg["block_type"]
        self.num_channels = self.cfg["num_channels"]
        self.channels_mult = 1 if self.num_channels[-1] == 256 else 4
D
dongshuilong 已提交
296

C
cuicheng01 已提交
297 298 299 300
        assert isinstance(self.lr_mult_list, (
            list, tuple
        )), "lr_mult_list should be in (list, tuple) but got {}".format(
            type(self.lr_mult_list))
D
dongshuilong 已提交
301 302 303
        assert len(self.lr_mult_list
                   ) == 5, "lr_mult_list length should be 5 but got {}".format(
                       len(self.lr_mult_list))
C
cuicheng01 已提交
304 305

        self.stem_cfg = {
C
cuicheng01 已提交
306
            #num_channels, num_filters, filter_size, stride
littletomatodonkey's avatar
littletomatodonkey 已提交
307 308 309
            "vb": [[input_image_channel, 64, 7, 2]],
            "vd":
            [[input_image_channel, 32, 3, 2], [32, 32, 3, 1], [32, 64, 3, 1]]
D
dongshuilong 已提交
310 311
        }

littletomatodonkey's avatar
littletomatodonkey 已提交
312
        self.stem = nn.Sequential(* [
C
cuicheng01 已提交
313
            ConvBNLayer(
D
dongshuilong 已提交
314 315 316 317 318
                num_channels=in_c,
                num_filters=out_c,
                filter_size=k,
                stride=s,
                act="relu",
littletomatodonkey's avatar
littletomatodonkey 已提交
319 320
                lr_mult=self.lr_mult_list[0],
                data_format=data_format)
C
cuicheng01 已提交
321 322
            for in_c, out_c, k, s in self.stem_cfg[version]
        ])
D
dongshuilong 已提交
323

littletomatodonkey's avatar
littletomatodonkey 已提交
324 325
        self.max_pool = MaxPool2D(
            kernel_size=3, stride=2, padding=1, data_format=data_format)
C
cuicheng01 已提交
326 327
        block_list = []
        for block_idx in range(len(self.block_depth)):
C
cuicheng01 已提交
328
            shortcut = False
C
cuicheng01 已提交
329
            for i in range(self.block_depth[block_idx]):
D
dongshuilong 已提交
330 331 332
                block_list.append(globals()[self.block_type](
                    num_channels=self.num_channels[block_idx] if i == 0 else
                    self.num_filters[block_idx] * self.channels_mult,
C
cuicheng01 已提交
333 334
                    num_filters=self.num_filters[block_idx],
                    stride=2 if i == 0 and block_idx != 0 else 1,
C
cuicheng01 已提交
335
                    shortcut=shortcut,
C
cuicheng01 已提交
336
                    if_first=block_idx == i == 0 if version == "vd" else True,
littletomatodonkey's avatar
littletomatodonkey 已提交
337 338
                    lr_mult=self.lr_mult_list[block_idx + 1],
                    data_format=data_format))
D
dongshuilong 已提交
339
                shortcut = True
C
cuicheng01 已提交
340
        self.blocks = nn.Sequential(*block_list)
C
cuicheng01 已提交
341

littletomatodonkey's avatar
littletomatodonkey 已提交
342
        self.avg_pool = AdaptiveAvgPool2D(1, data_format=data_format)
343
        self.flatten = nn.Flatten()
W
dbg  
weishengyu 已提交
344
        self.avg_pool_channels = self.num_channels[-1] * 2
C
cuicheng01 已提交
345
        stdv = 1.0 / math.sqrt(self.avg_pool_channels * 1.0)
C
cuicheng01 已提交
346
        self.fc = Linear(
C
cuicheng01 已提交
347
            self.avg_pool_channels,
C
cuicheng01 已提交
348
            self.class_num,
D
dongshuilong 已提交
349
            weight_attr=ParamAttr(initializer=Uniform(-stdv, stdv)))
C
cuicheng01 已提交
350

littletomatodonkey's avatar
littletomatodonkey 已提交
351
        self.data_format = data_format
352 353 354 355 356

        super().init_res(
            stages_pattern,
            return_patterns=return_patterns,
            return_stages=return_stages)
littletomatodonkey's avatar
littletomatodonkey 已提交
357

C
cuicheng01 已提交
358
    def forward(self, x):
littletomatodonkey's avatar
littletomatodonkey 已提交
359 360 361 362 363 364 365 366 367 368
        with paddle.static.amp.fp16_guard():
            if self.data_format == "NHWC":
                x = paddle.transpose(x, [0, 2, 3, 1])
                x.stop_gradient = True
            x = self.stem(x)
            x = self.max_pool(x)
            x = self.blocks(x)
            x = self.avg_pool(x)
            x = self.flatten(x)
            x = self.fc(x)
C
cuicheng01 已提交
369 370 371
        return x


D
dongshuilong 已提交
372 373 374 375 376 377 378 379 380 381 382 383 384 385
def _load_pretrained(pretrained, model, model_url, use_ssld):
    if pretrained is False:
        pass
    elif pretrained is True:
        load_dygraph_pretrain_from_url(model, model_url, use_ssld=use_ssld)
    elif isinstance(pretrained, str):
        load_dygraph_pretrain(model, pretrained)
    else:
        raise RuntimeError(
            "pretrained type is not available. Please use `string` or `boolean` type."
        )


def ResNet18(pretrained=False, use_ssld=False, **kwargs):
C
cuicheng01 已提交
386 387 388
    """
    ResNet18
    Args:
D
dongshuilong 已提交
389 390 391
        pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise.
                    If str, means the path of the pretrained model.
        use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True.
C
cuicheng01 已提交
392 393 394
    Returns:
        model: nn.Layer. Specific `ResNet18` model depends on args.
    """
395 396 397 398 399
    model = ResNet(
        config=NET_CONFIG["18"],
        stages_pattern=MODEL_STAGES_PATTERN["ResNet18"],
        version="vb",
        **kwargs)
D
dongshuilong 已提交
400
    _load_pretrained(pretrained, model, MODEL_URLS["ResNet18"], use_ssld)
C
cuicheng01 已提交
401 402
    return model

C
cuicheng01 已提交
403

D
dongshuilong 已提交
404
def ResNet18_vd(pretrained=False, use_ssld=False, **kwargs):
C
cuicheng01 已提交
405 406 407
    """
    ResNet18_vd
    Args:
D
dongshuilong 已提交
408 409 410
        pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise.
                    If str, means the path of the pretrained model.
        use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True.
C
cuicheng01 已提交
411 412 413
    Returns:
        model: nn.Layer. Specific `ResNet18_vd` model depends on args.
    """
414 415 416 417 418
    model = ResNet(
        config=NET_CONFIG["18"],
        stages_pattern=MODEL_STAGES_PATTERN["ResNet18"],
        version="vd",
        **kwargs)
D
dongshuilong 已提交
419
    _load_pretrained(pretrained, model, MODEL_URLS["ResNet18_vd"], use_ssld)
C
cuicheng01 已提交
420 421
    return model

C
cuicheng01 已提交
422

D
dongshuilong 已提交
423
def ResNet34(pretrained=False, use_ssld=False, **kwargs):
C
cuicheng01 已提交
424 425 426
    """
    ResNet34
    Args:
D
dongshuilong 已提交
427 428 429
        pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise.
                    If str, means the path of the pretrained model.
        use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True.
C
cuicheng01 已提交
430
    Returns:
C
cuicheng01 已提交
431
        model: nn.Layer. Specific `ResNet34` model depends on args.
C
cuicheng01 已提交
432
    """
433 434 435 436 437
    model = ResNet(
        config=NET_CONFIG["34"],
        stages_pattern=MODEL_STAGES_PATTERN["ResNet34"],
        version="vb",
        **kwargs)
D
dongshuilong 已提交
438
    _load_pretrained(pretrained, model, MODEL_URLS["ResNet34"], use_ssld)
C
cuicheng01 已提交
439 440 441
    return model


D
dongshuilong 已提交
442
def ResNet34_vd(pretrained=False, use_ssld=False, **kwargs):
C
cuicheng01 已提交
443 444 445
    """
    ResNet34_vd
    Args:
D
dongshuilong 已提交
446 447 448
        pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise.
                    If str, means the path of the pretrained model.
        use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True.
C
cuicheng01 已提交
449
    Returns:
C
cuicheng01 已提交
450
        model: nn.Layer. Specific `ResNet34_vd` model depends on args.
C
cuicheng01 已提交
451
    """
452 453 454 455 456
    model = ResNet(
        config=NET_CONFIG["34"],
        stages_pattern=MODEL_STAGES_PATTERN["ResNet34"],
        version="vd",
        **kwargs)
D
dongshuilong 已提交
457
    _load_pretrained(pretrained, model, MODEL_URLS["ResNet34_vd"], use_ssld)
C
cuicheng01 已提交
458 459 460
    return model


D
dongshuilong 已提交
461
def ResNet50(pretrained=False, use_ssld=False, **kwargs):
C
cuicheng01 已提交
462 463 464
    """
    ResNet50
    Args:
D
dongshuilong 已提交
465 466 467
        pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise.
                    If str, means the path of the pretrained model.
        use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True.
C
cuicheng01 已提交
468 469 470
    Returns:
        model: nn.Layer. Specific `ResNet50` model depends on args.
    """
471 472 473 474 475
    model = ResNet(
        config=NET_CONFIG["50"],
        stages_pattern=MODEL_STAGES_PATTERN["ResNet50"],
        version="vb",
        **kwargs)
D
dongshuilong 已提交
476
    _load_pretrained(pretrained, model, MODEL_URLS["ResNet50"], use_ssld)
C
cuicheng01 已提交
477 478
    return model

C
cuicheng01 已提交
479

D
dongshuilong 已提交
480
def ResNet50_vd(pretrained=False, use_ssld=False, **kwargs):
C
cuicheng01 已提交
481 482 483
    """
    ResNet50_vd
    Args:
D
dongshuilong 已提交
484 485 486
        pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise.
                    If str, means the path of the pretrained model.
        use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True.
C
cuicheng01 已提交
487 488 489
    Returns:
        model: nn.Layer. Specific `ResNet50_vd` model depends on args.
    """
490 491 492 493 494
    model = ResNet(
        config=NET_CONFIG["50"],
        stages_pattern=MODEL_STAGES_PATTERN["ResNet50"],
        version="vd",
        **kwargs)
D
dongshuilong 已提交
495
    _load_pretrained(pretrained, model, MODEL_URLS["ResNet50_vd"], use_ssld)
C
cuicheng01 已提交
496 497
    return model

C
cuicheng01 已提交
498

D
dongshuilong 已提交
499
def ResNet101(pretrained=False, use_ssld=False, **kwargs):
C
cuicheng01 已提交
500 501 502
    """
    ResNet101
    Args:
D
dongshuilong 已提交
503 504 505
        pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise.
                    If str, means the path of the pretrained model.
        use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True.
C
cuicheng01 已提交
506 507 508
    Returns:
        model: nn.Layer. Specific `ResNet101` model depends on args.
    """
509 510 511 512 513
    model = ResNet(
        config=NET_CONFIG["101"],
        stages_pattern=MODEL_STAGES_PATTERN["ResNet101"],
        version="vb",
        **kwargs)
D
dongshuilong 已提交
514
    _load_pretrained(pretrained, model, MODEL_URLS["ResNet101"], use_ssld)
C
cuicheng01 已提交
515 516
    return model

C
cuicheng01 已提交
517

D
dongshuilong 已提交
518
def ResNet101_vd(pretrained=False, use_ssld=False, **kwargs):
C
cuicheng01 已提交
519 520 521
    """
    ResNet101_vd
    Args:
D
dongshuilong 已提交
522 523 524
        pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise.
                    If str, means the path of the pretrained model.
        use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True.
C
cuicheng01 已提交
525 526 527
    Returns:
        model: nn.Layer. Specific `ResNet101_vd` model depends on args.
    """
528 529 530 531 532
    model = ResNet(
        config=NET_CONFIG["101"],
        stages_pattern=MODEL_STAGES_PATTERN["ResNet101"],
        version="vd",
        **kwargs)
D
dongshuilong 已提交
533
    _load_pretrained(pretrained, model, MODEL_URLS["ResNet101_vd"], use_ssld)
C
cuicheng01 已提交
534 535
    return model

C
cuicheng01 已提交
536

D
dongshuilong 已提交
537
def ResNet152(pretrained=False, use_ssld=False, **kwargs):
C
cuicheng01 已提交
538 539 540
    """
    ResNet152
    Args:
D
dongshuilong 已提交
541 542 543
        pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise.
                    If str, means the path of the pretrained model.
        use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True.
C
cuicheng01 已提交
544 545 546
    Returns:
        model: nn.Layer. Specific `ResNet152` model depends on args.
    """
547 548 549 550 551
    model = ResNet(
        config=NET_CONFIG["152"],
        stages_pattern=MODEL_STAGES_PATTERN["ResNet152"],
        version="vb",
        **kwargs)
D
dongshuilong 已提交
552
    _load_pretrained(pretrained, model, MODEL_URLS["ResNet152"], use_ssld)
C
cuicheng01 已提交
553 554
    return model

C
cuicheng01 已提交
555

D
dongshuilong 已提交
556
def ResNet152_vd(pretrained=False, use_ssld=False, **kwargs):
C
cuicheng01 已提交
557 558 559
    """
    ResNet152_vd
    Args:
D
dongshuilong 已提交
560 561 562
        pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise.
                    If str, means the path of the pretrained model.
        use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True.
C
cuicheng01 已提交
563 564 565
    Returns:
        model: nn.Layer. Specific `ResNet152_vd` model depends on args.
    """
566 567 568 569 570
    model = ResNet(
        config=NET_CONFIG["152"],
        stages_pattern=MODEL_STAGES_PATTERN["ResNet152"],
        version="vd",
        **kwargs)
D
dongshuilong 已提交
571
    _load_pretrained(pretrained, model, MODEL_URLS["ResNet152_vd"], use_ssld)
C
cuicheng01 已提交
572 573 574
    return model


D
dongshuilong 已提交
575
def ResNet200_vd(pretrained=False, use_ssld=False, **kwargs):
C
cuicheng01 已提交
576 577 578
    """
    ResNet200_vd
    Args:
D
dongshuilong 已提交
579 580 581
        pretrained: bool=False or str. If `True` load pretrained parameters, `False` otherwise.
                    If str, means the path of the pretrained model.
        use_ssld: bool=False. Whether using distillation pretrained model when pretrained=True.
C
cuicheng01 已提交
582 583 584
    Returns:
        model: nn.Layer. Specific `ResNet200_vd` model depends on args.
    """
585 586 587 588 589
    model = ResNet(
        config=NET_CONFIG["200"],
        stages_pattern=MODEL_STAGES_PATTERN["ResNet200"],
        version="vd",
        **kwargs)
D
dongshuilong 已提交
590
    _load_pretrained(pretrained, model, MODEL_URLS["ResNet200_vd"], use_ssld)
C
cuicheng01 已提交
591
    return model