Skip to content

P
PaddleClas

PaddlePaddle / PaddleClas
1 年多 前同步成功

通知 116
Star 4999
Fork 1114
P
PaddleClas
未验证 提交 1303affa 编写于 作者: W Wei Shengyu 提交者： GitHub
浏览文件
操作

Merge pull request #720 from lyuwenyu/hub_L_b 

Release unnecessary dependent pkgs
上级 e8752538 72ab665b
隐藏空白更改
内联 并排
Showing with 808 addition and 816 deletion
hubconf.py
浏览文件 @ 1303affa
...@@ -12,821 +12,813 @@ ...@@ -12,821 +12,813 @@
# See the License for the specific language governing permissions and # See the License for the specific language governing permissions and
# limitations under the License. # limitations under the License.
dependencies = ['paddle', 'numpy'] dependencies = ['paddle']
import paddle import paddle
from ppcls.modeling import architectures import os
import sys
def _load_pretrained_parameters(model, name):
url = 'https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/{}_pretrained.pdparams'.format( class _SysPathG(object):
name) """
path = paddle.utils.download.get_weights_path_from_url(url) _SysPathG used to add/clean path for sys.path. Making sure minimal pkgs dependents by skiping parent dirs.
model.set_state_dict(paddle.load(path))
return model __enter__
add path into sys.path
__exit__
def alexnet(pretrained=False, **kwargs): clean user's sys.path to avoid unexpect behaviors
""" """
AlexNet
Args: def __init__(self, path):
pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. self.path = path
kwargs:
class_dim: int=1000. Output dim of last fc layer. def __enter__(self, ):
Returns: sys.path.insert(0, self.path)
model: nn.Layer. Specific `AlexNet` model depends on args.
""" def __exit__(self, type, value, traceback):
model = architectures.AlexNet(**kwargs) _p = sys.path.pop(0)
if pretrained: assert _p == self.path, 'Make sure sys.path cleaning {} correctly.'.format(
model = _load_pretrained_parameters(model, 'AlexNet') self.path)
return model
with _SysPathG(
os.path.join(
def vgg11(pretrained=False, **kwargs): os.path.dirname(os.path.abspath(__file__)), 'ppcls', 'modeling')):
""" import architectures
VGG11
Args: def _load_pretrained_parameters(model, name):
pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. url = 'https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/{}_pretrained.pdparams'.format(
kwargs: name)
class_dim: int=1000. Output dim of last fc layer. path = paddle.utils.download.get_weights_path_from_url(url)
stop_grad_layers: int=0. The parameters in blocks which index larger than `stop_grad_layers`, will be set `param.trainable=False` model.set_state_dict(paddle.load(path))
Returns: return model
model: nn.Layer. Specific `VGG11` model depends on args.
""" def alexnet(pretrained=False, **kwargs):
model = architectures.VGG11(**kwargs) """
if pretrained: AlexNet
model = _load_pretrained_parameters(model, 'VGG11') Args:
pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise.
return model kwargs:
class_dim: int=1000. Output dim of last fc layer.
Returns:
def vgg13(pretrained=False, **kwargs): model: nn.Layer. Specific `AlexNet` model depends on args.
""" """
VGG13 model = architectures.AlexNet(**kwargs)
Args: if pretrained:
pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. model = _load_pretrained_parameters(model, 'AlexNet')
kwargs:
class_dim: int=1000. Output dim of last fc layer. return model
stop_grad_layers: int=0. The parameters in blocks which index larger than `stop_grad_layers`, will be set `param.trainable=False`
Returns: def vgg11(pretrained=False, **kwargs):
model: nn.Layer. Specific `VGG13` model depends on args. """
""" VGG11
model = architectures.VGG13(**kwargs) Args:
if pretrained: pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise.
model = _load_pretrained_parameters(model, 'VGG13') kwargs:
class_dim: int=1000. Output dim of last fc layer.
return model stop_grad_layers: int=0. The parameters in blocks which index larger than `stop_grad_layers`, will be set `param.trainable=False`
Returns:
model: nn.Layer. Specific `VGG11` model depends on args.
def vgg16(pretrained=False, **kwargs): """
""" model = architectures.VGG11(**kwargs)
VGG16 if pretrained:
Args: model = _load_pretrained_parameters(model, 'VGG11')
pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise.
kwargs: return model
class_dim: int=1000. Output dim of last fc layer.
stop_grad_layers: int=0. The parameters in blocks which index larger than `stop_grad_layers`, will be set `param.trainable=False` def vgg13(pretrained=False, **kwargs):
Returns: """
model: nn.Layer. Specific `VGG16` model depends on args. VGG13
""" Args:
model = architectures.VGG16(**kwargs) pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise.
if pretrained: kwargs:
model = _load_pretrained_parameters(model, 'VGG16') class_dim: int=1000. Output dim of last fc layer.
stop_grad_layers: int=0. The parameters in blocks which index larger than `stop_grad_layers`, will be set `param.trainable=False`
return model Returns:
model: nn.Layer. Specific `VGG13` model depends on args.
"""
def vgg19(pretrained=False, **kwargs): model = architectures.VGG13(**kwargs)
""" if pretrained:
VGG19 model = _load_pretrained_parameters(model, 'VGG13')
Args:
pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. return model
kwargs:
class_dim: int=1000. Output dim of last fc layer. def vgg16(pretrained=False, **kwargs):
stop_grad_layers: int=0. The parameters in blocks which index larger than `stop_grad_layers`, will be set `param.trainable=False` """
Returns: VGG16
model: nn.Layer. Specific `VGG19` model depends on args. Args:
""" pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise.
model = architectures.VGG19(**kwargs) kwargs:
if pretrained: class_dim: int=1000. Output dim of last fc layer.
model = _load_pretrained_parameters(model, 'VGG19') stop_grad_layers: int=0. The parameters in blocks which index larger than `stop_grad_layers`, will be set `param.trainable=False`
Returns:
return model model: nn.Layer. Specific `VGG16` model depends on args.
"""
model = architectures.VGG16(**kwargs)
def resnet18(pretrained=False, **kwargs): if pretrained:
""" model = _load_pretrained_parameters(model, 'VGG16')
ResNet18
Args: return model
pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise.
kwargs: def vgg19(pretrained=False, **kwargs):
class_dim: int=1000. Output dim of last fc layer. """
input_image_channel: int=3. The number of input image channels VGG19
data_format: str='NCHW'. The data format of batch input images, should in ('NCHW', 'NHWC') Args:
Returns: pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise.
model: nn.Layer. Specific `ResNet18` model depends on args. kwargs:
""" class_dim: int=1000. Output dim of last fc layer.
model = architectures.ResNet18(**kwargs) stop_grad_layers: int=0. The parameters in blocks which index larger than `stop_grad_layers`, will be set `param.trainable=False`
if pretrained: Returns:
model = _load_pretrained_parameters(model, 'ResNet18') model: nn.Layer. Specific `VGG19` model depends on args.
"""
return model model = architectures.VGG19(**kwargs)
if pretrained:
model = _load_pretrained_parameters(model, 'VGG19')
def resnet34(pretrained=False, **kwargs):
""" return model
ResNet34
Args: def resnet18(pretrained=False, **kwargs):
pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. """
kwargs: ResNet18
class_dim: int=1000. Output dim of last fc layer. Args:
input_image_channel: int=3. The number of input image channels pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise.
data_format: str='NCHW'. The data format of batch input images, should in ('NCHW', 'NHWC') kwargs:
Returns: class_dim: int=1000. Output dim of last fc layer.
model: nn.Layer. Specific `ResNet34` model depends on args. input_image_channel: int=3. The number of input image channels
""" data_format: str='NCHW'. The data format of batch input images, should in ('NCHW', 'NHWC')
model = architectures.ResNet34(**kwargs) Returns:
if pretrained: model: nn.Layer. Specific `ResNet18` model depends on args.
model = _load_pretrained_parameters(model, 'ResNet34') """
model = architectures.ResNet18(**kwargs)
return model if pretrained:
model = _load_pretrained_parameters(model, 'ResNet18')
def resnet50(pretrained=False, **kwargs): return model
"""
ResNet50 def resnet34(pretrained=False, **kwargs):
Args: """
pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. ResNet34
kwargs: Args:
class_dim: int=1000. Output dim of last fc layer. pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise.
input_image_channel: int=3. The number of input image channels kwargs:
data_format: str='NCHW'. The data format of batch input images, should in ('NCHW', 'NHWC') class_dim: int=1000. Output dim of last fc layer.
Returns: input_image_channel: int=3. The number of input image channels
model: nn.Layer. Specific `ResNet50` model depends on args. data_format: str='NCHW'. The data format of batch input images, should in ('NCHW', 'NHWC')
""" Returns:
model = architectures.ResNet50(**kwargs) model: nn.Layer. Specific `ResNet34` model depends on args.
if pretrained: """
model = _load_pretrained_parameters(model, 'ResNet50') model = architectures.ResNet34(**kwargs)
if pretrained:
return model model = _load_pretrained_parameters(model, 'ResNet34')
return model
def resnet101(pretrained=False, **kwargs):
""" def resnet50(pretrained=False, **kwargs):
ResNet101 """
Args: ResNet50
pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. Args:
kwargs: pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise.
class_dim: int=1000. Output dim of last fc layer. kwargs:
input_image_channel: int=3. The number of input image channels class_dim: int=1000. Output dim of last fc layer.
data_format: str='NCHW'. The data format of batch input images, should in ('NCHW', 'NHWC') input_image_channel: int=3. The number of input image channels
Returns: data_format: str='NCHW'. The data format of batch input images, should in ('NCHW', 'NHWC')
model: nn.Layer. Specific `ResNet101` model depends on args. Returns:
""" model: nn.Layer. Specific `ResNet50` model depends on args.
model = architectures.ResNet101(**kwargs) """
if pretrained: model = architectures.ResNet50(**kwargs)
model = _load_pretrained_parameters(model, 'ResNet101') if pretrained:
model = _load_pretrained_parameters(model, 'ResNet50')
return model
return model
def resnet152(pretrained=False, **kwargs): def resnet101(pretrained=False, **kwargs):
""" """
ResNet152 ResNet101
Args: Args:
pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise.
kwargs: kwargs:
class_dim: int=1000. Output dim of last fc layer. class_dim: int=1000. Output dim of last fc layer.
input_image_channel: int=3. The number of input image channels input_image_channel: int=3. The number of input image channels
data_format: str='NCHW'. The data format of batch input images, should in ('NCHW', 'NHWC') data_format: str='NCHW'. The data format of batch input images, should in ('NCHW', 'NHWC')
Returns: Returns:
model: nn.Layer. Specific `ResNet152` model depends on args. model: nn.Layer. Specific `ResNet101` model depends on args.
""" """
model = architectures.ResNet152(**kwargs) model = architectures.ResNet101(**kwargs)
if pretrained: if pretrained:
model = _load_pretrained_parameters(model, 'ResNet152') model = _load_pretrained_parameters(model, 'ResNet101')
return model return model
def resnet152(pretrained=False, **kwargs):
def squeezenet1_0(pretrained=False, **kwargs): """
""" ResNet152
SqueezeNet1_0 Args:
Args: pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise.
pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. kwargs:
kwargs: class_dim: int=1000. Output dim of last fc layer.
class_dim: int=1000. Output dim of last fc layer. input_image_channel: int=3. The number of input image channels
Returns: data_format: str='NCHW'. The data format of batch input images, should in ('NCHW', 'NHWC')
model: nn.Layer. Specific `SqueezeNet1_0` model depends on args. Returns:
""" model: nn.Layer. Specific `ResNet152` model depends on args.
model = architectures.SqueezeNet1_0(**kwargs) """
if pretrained: model = architectures.ResNet152(**kwargs)
model = _load_pretrained_parameters(model, 'SqueezeNet1_0') if pretrained:
model = _load_pretrained_parameters(model, 'ResNet152')
return model
return model
def squeezenet1_1(pretrained=False, **kwargs): def squeezenet1_0(pretrained=False, **kwargs):
""" """
SqueezeNet1_1 SqueezeNet1_0
Args: Args:
pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise.
kwargs: kwargs:
class_dim: int=1000. Output dim of last fc layer. class_dim: int=1000. Output dim of last fc layer.
Returns: Returns:
model: nn.Layer. Specific `SqueezeNet1_1` model depends on args. model: nn.Layer. Specific `SqueezeNet1_0` model depends on args.
""" """
model = architectures.SqueezeNet1_1(**kwargs) model = architectures.SqueezeNet1_0(**kwargs)
if pretrained: if pretrained:
model = _load_pretrained_parameters(model, 'SqueezeNet1_1') model = _load_pretrained_parameters(model, 'SqueezeNet1_0')
return model return model
def squeezenet1_1(pretrained=False, **kwargs):
def densenet121(pretrained=False, **kwargs): """
""" SqueezeNet1_1
DenseNet121 Args:
Args: pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise.
pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. kwargs:
kwargs: class_dim: int=1000. Output dim of last fc layer.
class_dim: int=1000. Output dim of last fc layer. Returns:
dropout: float=0. Probability of setting units to zero. model: nn.Layer. Specific `SqueezeNet1_1` model depends on args.
bn_size: int=4. The number of channals per group """
Returns: model = architectures.SqueezeNet1_1(**kwargs)
model: nn.Layer. Specific `DenseNet121` model depends on args. if pretrained:
""" model = _load_pretrained_parameters(model, 'SqueezeNet1_1')
model = architectures.DenseNet121(**kwargs)
if pretrained: return model
model = _load_pretrained_parameters(model, 'DenseNet121')
def densenet121(pretrained=False, **kwargs):
return model """
DenseNet121
Args:
def densenet161(pretrained=False, **kwargs): pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise.
""" kwargs:
DenseNet161 class_dim: int=1000. Output dim of last fc layer.
Args: dropout: float=0. Probability of setting units to zero.
pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. bn_size: int=4. The number of channals per group
kwargs: Returns:
class_dim: int=1000. Output dim of last fc layer. model: nn.Layer. Specific `DenseNet121` model depends on args.
dropout: float=0. Probability of setting units to zero. """
bn_size: int=4. The number of channals per group model = architectures.DenseNet121(**kwargs)
Returns: if pretrained:
model: nn.Layer. Specific `DenseNet161` model depends on args. model = _load_pretrained_parameters(model, 'DenseNet121')
"""
model = architectures.DenseNet161(**kwargs) return model
if pretrained:
model = _load_pretrained_parameters(model, 'DenseNet161') def densenet161(pretrained=False, **kwargs):
"""
return model DenseNet161
Args:
pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise.
def densenet169(pretrained=False, **kwargs): kwargs:
""" class_dim: int=1000. Output dim of last fc layer.
DenseNet169 dropout: float=0. Probability of setting units to zero.
Args: bn_size: int=4. The number of channals per group
pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. Returns:
kwargs: model: nn.Layer. Specific `DenseNet161` model depends on args.
class_dim: int=1000. Output dim of last fc layer. """
dropout: float=0. Probability of setting units to zero. model = architectures.DenseNet161(**kwargs)
bn_size: int=4. The number of channals per group if pretrained:
Returns: model = _load_pretrained_parameters(model, 'DenseNet161')
model: nn.Layer. Specific `DenseNet169` model depends on args.
""" return model
model = architectures.DenseNet169(**kwargs)
if pretrained: def densenet169(pretrained=False, **kwargs):
model = _load_pretrained_parameters(model, 'DenseNet169') """
DenseNet169
return model Args:
pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise.
kwargs:
def densenet201(pretrained=False, **kwargs): class_dim: int=1000. Output dim of last fc layer.
""" dropout: float=0. Probability of setting units to zero.
DenseNet201 bn_size: int=4. The number of channals per group
Args: Returns:
pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. model: nn.Layer. Specific `DenseNet169` model depends on args.
kwargs: """
class_dim: int=1000. Output dim of last fc layer. model = architectures.DenseNet169(**kwargs)
dropout: float=0. Probability of setting units to zero. if pretrained:
bn_size: int=4. The number of channals per group model = _load_pretrained_parameters(model, 'DenseNet169')
Returns:
model: nn.Layer. Specific `DenseNet201` model depends on args. return model
"""
model = architectures.DenseNet201(**kwargs) def densenet201(pretrained=False, **kwargs):
if pretrained: """
model = _load_pretrained_parameters(model, 'DenseNet201') DenseNet201
Args:
return model pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise.
kwargs:
class_dim: int=1000. Output dim of last fc layer.
def densenet264(pretrained=False, **kwargs): dropout: float=0. Probability of setting units to zero.
""" bn_size: int=4. The number of channals per group
DenseNet264 Returns:
Args: model: nn.Layer. Specific `DenseNet201` model depends on args.
pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. """
kwargs: model = architectures.DenseNet201(**kwargs)
class_dim: int=1000. Output dim of last fc layer. if pretrained:
dropout: float=0. Probability of setting units to zero. model = _load_pretrained_parameters(model, 'DenseNet201')
bn_size: int=4. The number of channals per group
Returns: return model
model: nn.Layer. Specific `DenseNet264` model depends on args.
""" def densenet264(pretrained=False, **kwargs):
model = architectures.DenseNet264(**kwargs) """
if pretrained: DenseNet264
model = _load_pretrained_parameters(model, 'DenseNet264') Args:
pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise.
return model kwargs:
class_dim: int=1000. Output dim of last fc layer.
dropout: float=0. Probability of setting units to zero.
def inceptionv3(pretrained=False, **kwargs): bn_size: int=4. The number of channals per group
""" Returns:
InceptionV3 model: nn.Layer. Specific `DenseNet264` model depends on args.
Args: """
pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. model = architectures.DenseNet264(**kwargs)
kwargs: if pretrained:
class_dim: int=1000. Output dim of last fc layer. model = _load_pretrained_parameters(model, 'DenseNet264')
Returns:
model: nn.Layer. Specific `InceptionV3` model depends on args. return model
"""
model = architectures.InceptionV3(**kwargs) def inceptionv3(pretrained=False, **kwargs):
if pretrained: """
model = _load_pretrained_parameters(model, 'InceptionV3') InceptionV3
Args:
return model pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise.
kwargs:
class_dim: int=1000. Output dim of last fc layer.
def inceptionv4(pretrained=False, **kwargs): Returns:
""" model: nn.Layer. Specific `InceptionV3` model depends on args.
InceptionV4 """
Args: model = architectures.InceptionV3(**kwargs)
pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. if pretrained:
kwargs: model = _load_pretrained_parameters(model, 'InceptionV3')
class_dim: int=1000. Output dim of last fc layer.
Returns: return model
model: nn.Layer. Specific `InceptionV4` model depends on args.
""" def inceptionv4(pretrained=False, **kwargs):
model = architectures.InceptionV4(**kwargs) """
if pretrained: InceptionV4
model = _load_pretrained_parameters(model, 'InceptionV4') Args:
pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise.
return model kwargs:
class_dim: int=1000. Output dim of last fc layer.
Returns:
def googlenet(pretrained=False, **kwargs): model: nn.Layer. Specific `InceptionV4` model depends on args.
""" """
GoogLeNet model = architectures.InceptionV4(**kwargs)
Args: if pretrained:
pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. model = _load_pretrained_parameters(model, 'InceptionV4')
kwargs:
class_dim: int=1000. Output dim of last fc layer. return model
Returns:
model: nn.Layer. Specific `GoogLeNet` model depends on args. def googlenet(pretrained=False, **kwargs):
""" """
model = architectures.GoogLeNet(**kwargs) GoogLeNet
if pretrained: Args:
model = _load_pretrained_parameters(model, 'GoogLeNet') pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise.
kwargs:
return model class_dim: int=1000. Output dim of last fc layer.
Returns:
model: nn.Layer. Specific `GoogLeNet` model depends on args.
def shufflenetv2_x0_25(pretrained=False, **kwargs): """
""" model = architectures.GoogLeNet(**kwargs)
ShuffleNetV2_x0_25 if pretrained:
Args: model = _load_pretrained_parameters(model, 'GoogLeNet')
pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise.
kwargs: return model
class_dim: int=1000. Output dim of last fc layer.
Returns: def shufflenetv2_x0_25(pretrained=False, **kwargs):
model: nn.Layer. Specific `ShuffleNetV2_x0_25` model depends on args. """
""" ShuffleNetV2_x0_25
model = architectures.ShuffleNetV2_x0_25(**kwargs) Args:
if pretrained: pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise.
model = _load_pretrained_parameters(model, 'ShuffleNetV2_x0_25') kwargs:
class_dim: int=1000. Output dim of last fc layer.
return model Returns:
model: nn.Layer. Specific `ShuffleNetV2_x0_25` model depends on args.
"""
def mobilenetv1(pretrained=False, **kwargs): model = architectures.ShuffleNetV2_x0_25(**kwargs)
""" if pretrained:
MobileNetV1 model = _load_pretrained_parameters(model, 'ShuffleNetV2_x0_25')
Args:
pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. return model
kwargs:
class_dim: int=1000. Output dim of last fc layer. def mobilenetv1(pretrained=False, **kwargs):
Returns: """
model: nn.Layer. Specific `MobileNetV1` model depends on args. MobileNetV1
""" Args:
model = architectures.MobileNetV1(**kwargs) pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise.
if pretrained: kwargs:
model = _load_pretrained_parameters(model, 'MobileNetV1') class_dim: int=1000. Output dim of last fc layer.
Returns:
return model model: nn.Layer. Specific `MobileNetV1` model depends on args.
"""
model = architectures.MobileNetV1(**kwargs)
def mobilenetv1_x0_25(pretrained=False, **kwargs): if pretrained:
""" model = _load_pretrained_parameters(model, 'MobileNetV1')
MobileNetV1_x0_25
Args: return model
pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise.
kwargs: def mobilenetv1_x0_25(pretrained=False, **kwargs):
class_dim: int=1000. Output dim of last fc layer. """
Returns: MobileNetV1_x0_25
model: nn.Layer. Specific `MobileNetV1_x0_25` model depends on args. Args:
""" pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise.
model = architectures.MobileNetV1_x0_25(**kwargs) kwargs:
if pretrained: class_dim: int=1000. Output dim of last fc layer.
model = _load_pretrained_parameters(model, 'MobileNetV1_x0_25') Returns:
model: nn.Layer. Specific `MobileNetV1_x0_25` model depends on args.
return model """
model = architectures.MobileNetV1_x0_25(**kwargs)
if pretrained:
def mobilenetv1_x0_5(pretrained=False, **kwargs): model = _load_pretrained_parameters(model, 'MobileNetV1_x0_25')
"""
MobileNetV1_x0_5 return model
Args:
pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. def mobilenetv1_x0_5(pretrained=False, **kwargs):
kwargs: """
class_dim: int=1000. Output dim of last fc layer. MobileNetV1_x0_5
Returns: Args:
model: nn.Layer. Specific `MobileNetV1_x0_5` model depends on args. pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise.
""" kwargs:
model = architectures.MobileNetV1_x0_5(**kwargs) class_dim: int=1000. Output dim of last fc layer.
if pretrained: Returns:
model = _load_pretrained_parameters(model, 'MobileNetV1_x0_5') model: nn.Layer. Specific `MobileNetV1_x0_5` model depends on args.
"""
return model model = architectures.MobileNetV1_x0_5(**kwargs)
if pretrained:
model = _load_pretrained_parameters(model, 'MobileNetV1_x0_5')
def mobilenetv1_x0_75(pretrained=False, **kwargs):
""" return model
MobileNetV1_x0_75
Args: def mobilenetv1_x0_75(pretrained=False, **kwargs):
pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. """
kwargs: MobileNetV1_x0_75
class_dim: int=1000. Output dim of last fc layer. Args:
Returns: pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise.
model: nn.Layer. Specific `MobileNetV1_x0_75` model depends on args. kwargs:
""" class_dim: int=1000. Output dim of last fc layer.
model = architectures.MobileNetV1_x0_75(**kwargs) Returns:
if pretrained: model: nn.Layer. Specific `MobileNetV1_x0_75` model depends on args.
model = _load_pretrained_parameters(model, 'MobileNetV1_x0_75') """
model = architectures.MobileNetV1_x0_75(**kwargs)
return model if pretrained:
model = _load_pretrained_parameters(model, 'MobileNetV1_x0_75')
def mobilenetv2_x0_25(pretrained=False, **kwargs): return model
"""
MobileNetV2_x0_25 def mobilenetv2_x0_25(pretrained=False, **kwargs):
Args: """
pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. MobileNetV2_x0_25
kwargs: Args:
class_dim: int=1000. Output dim of last fc layer. pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise.
Returns: kwargs:
model: nn.Layer. Specific `MobileNetV2_x0_25` model depends on args. class_dim: int=1000. Output dim of last fc layer.
""" Returns:
model = architectures.MobileNetV2_x0_25(**kwargs) model: nn.Layer. Specific `MobileNetV2_x0_25` model depends on args.
if pretrained: """
model = _load_pretrained_parameters(model, 'MobileNetV2_x0_25') model = architectures.MobileNetV2_x0_25(**kwargs)
if pretrained:
return model model = _load_pretrained_parameters(model, 'MobileNetV2_x0_25')
return model
def mobilenetv2_x0_5(pretrained=False, **kwargs):
""" def mobilenetv2_x0_5(pretrained=False, **kwargs):
MobileNetV2_x0_5 """
Args: MobileNetV2_x0_5
pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. Args:
kwargs: pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise.
class_dim: int=1000. Output dim of last fc layer. kwargs:
Returns: class_dim: int=1000. Output dim of last fc layer.
model: nn.Layer. Specific `MobileNetV2_x0_5` model depends on args. Returns:
""" model: nn.Layer. Specific `MobileNetV2_x0_5` model depends on args.
model = architectures.MobileNetV2_x0_5(**kwargs) """
if pretrained: model = architectures.MobileNetV2_x0_5(**kwargs)
model = _load_pretrained_parameters(model, 'MobileNetV2_x0_5') if pretrained:
model = _load_pretrained_parameters(model, 'MobileNetV2_x0_5')
return model
return model
def mobilenetv2_x0_75(pretrained=False, **kwargs): def mobilenetv2_x0_75(pretrained=False, **kwargs):
""" """
MobileNetV2_x0_75 MobileNetV2_x0_75
Args: Args:
pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise.
kwargs: kwargs:
class_dim: int=1000. Output dim of last fc layer. class_dim: int=1000. Output dim of last fc layer.
Returns: Returns:
model: nn.Layer. Specific `MobileNetV2_x0_75` model depends on args. model: nn.Layer. Specific `MobileNetV2_x0_75` model depends on args.
""" """
model = architectures.MobileNetV2_x0_75(**kwargs) model = architectures.MobileNetV2_x0_75(**kwargs)
if pretrained: if pretrained:
model = _load_pretrained_parameters(model, 'MobileNetV2_x0_75') model = _load_pretrained_parameters(model, 'MobileNetV2_x0_75')
return model return model
def mobilenetv2_x1_5(pretrained=False, **kwargs):
def mobilenetv2_x1_5(pretrained=False, **kwargs): """
""" MobileNetV2_x1_5
MobileNetV2_x1_5 Args:
Args: pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise.
pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. kwargs:
kwargs: class_dim: int=1000. Output dim of last fc layer.
class_dim: int=1000. Output dim of last fc layer. Returns:
Returns: model: nn.Layer. Specific `MobileNetV2_x1_5` model depends on args.
model: nn.Layer. Specific `MobileNetV2_x1_5` model depends on args. """
""" model = architectures.MobileNetV2_x1_5(**kwargs)
model = architectures.MobileNetV2_x1_5(**kwargs) if pretrained:
if pretrained: model = _load_pretrained_parameters(model, 'MobileNetV2_x1_5')
model = _load_pretrained_parameters(model, 'MobileNetV2_x1_5')
return model
return model
def mobilenetv2_x2_0(pretrained=False, **kwargs):
"""
def mobilenetv2_x2_0(pretrained=False, **kwargs): MobileNetV2_x2_0
""" Args:
MobileNetV2_x2_0 pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise.
Args: kwargs:
pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. class_dim: int=1000. Output dim of last fc layer.
kwargs: Returns:
class_dim: int=1000. Output dim of last fc layer. model: nn.Layer. Specific `MobileNetV2_x2_0` model depends on args.
Returns: """
model: nn.Layer. Specific `MobileNetV2_x2_0` model depends on args. model = architectures.MobileNetV2_x2_0(**kwargs)
""" if pretrained:
model = architectures.MobileNetV2_x2_0(**kwargs) model = _load_pretrained_parameters(model, 'MobileNetV2_x2_0')
if pretrained:
model = _load_pretrained_parameters(model, 'MobileNetV2_x2_0') return model
return model def mobilenetv3_large_x0_35(pretrained=False, **kwargs):
"""
MobileNetV3_large_x0_35
def mobilenetv3_large_x0_35(pretrained=False, **kwargs): Args:
""" pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise.
MobileNetV3_large_x0_35 kwargs:
Args: class_dim: int=1000. Output dim of last fc layer.
pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. Returns:
kwargs: model: nn.Layer. Specific `MobileNetV3_large_x0_35` model depends on args.
class_dim: int=1000. Output dim of last fc layer. """
Returns: model = architectures.MobileNetV3_large_x0_35(**kwargs)
model: nn.Layer. Specific `MobileNetV3_large_x0_35` model depends on args. if pretrained:
""" model = _load_pretrained_parameters(model,
model = architectures.MobileNetV3_large_x0_35(**kwargs) 'MobileNetV3_large_x0_35')
if pretrained:
model = _load_pretrained_parameters(model, 'MobileNetV3_large_x0_35') return model
return model def mobilenetv3_large_x0_5(pretrained=False, **kwargs):
"""
MobileNetV3_large_x0_5
def mobilenetv3_large_x0_5(pretrained=False, **kwargs): Args:
""" pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise.
MobileNetV3_large_x0_5 kwargs:
Args: class_dim: int=1000. Output dim of last fc layer.
pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. Returns:
kwargs: model: nn.Layer. Specific `MobileNetV3_large_x0_5` model depends on args.
class_dim: int=1000. Output dim of last fc layer. """
Returns: model = architectures.MobileNetV3_large_x0_5(**kwargs)
model: nn.Layer. Specific `MobileNetV3_large_x0_5` model depends on args. if pretrained:
""" model = _load_pretrained_parameters(model,
model = architectures.MobileNetV3_large_x0_5(**kwargs) 'MobileNetV3_large_x0_5')
if pretrained:
model = _load_pretrained_parameters(model, 'MobileNetV3_large_x0_5') return model
return model def mobilenetv3_large_x0_75(pretrained=False, **kwargs):
"""
MobileNetV3_large_x0_75
def mobilenetv3_large_x0_75(pretrained=False, **kwargs): Args:
""" pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise.
MobileNetV3_large_x0_75 kwargs:
Args: class_dim: int=1000. Output dim of last fc layer.
pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. Returns:
kwargs: model: nn.Layer. Specific `MobileNetV3_large_x0_75` model depends on args.
class_dim: int=1000. Output dim of last fc layer. """
Returns: model = architectures.MobileNetV3_large_x0_75(**kwargs)
model: nn.Layer. Specific `MobileNetV3_large_x0_75` model depends on args. if pretrained:
""" model = _load_pretrained_parameters(model,
model = architectures.MobileNetV3_large_x0_75(**kwargs) 'MobileNetV3_large_x0_75')
if pretrained:
model = _load_pretrained_parameters(model, 'MobileNetV3_large_x0_75') return model
return model def mobilenetv3_large_x1_0(pretrained=False, **kwargs):
"""
MobileNetV3_large_x1_0
def mobilenetv3_large_x1_0(pretrained=False, **kwargs): Args:
""" pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise.
MobileNetV3_large_x1_0 kwargs:
Args: class_dim: int=1000. Output dim of last fc layer.
pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. Returns:
kwargs: model: nn.Layer. Specific `MobileNetV3_large_x1_0` model depends on args.
class_dim: int=1000. Output dim of last fc layer. """
Returns: model = architectures.MobileNetV3_large_x1_0(**kwargs)
model: nn.Layer. Specific `MobileNetV3_large_x1_0` model depends on args. if pretrained:
""" model = _load_pretrained_parameters(model,
model = architectures.MobileNetV3_large_x1_0(**kwargs) 'MobileNetV3_large_x1_0')
if pretrained:
model = _load_pretrained_parameters(model, 'MobileNetV3_large_x1_0') return model
return model def mobilenetv3_large_x1_25(pretrained=False, **kwargs):
"""
MobileNetV3_large_x1_25
def mobilenetv3_large_x1_25(pretrained=False, **kwargs): Args:
""" pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise.
MobileNetV3_large_x1_25 kwargs:
Args: class_dim: int=1000. Output dim of last fc layer.
pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. Returns:
kwargs: model: nn.Layer. Specific `MobileNetV3_large_x1_25` model depends on args.
class_dim: int=1000. Output dim of last fc layer. """
Returns: model = architectures.MobileNetV3_large_x1_25(**kwargs)
model: nn.Layer. Specific `MobileNetV3_large_x1_25` model depends on args. if pretrained:
""" model = _load_pretrained_parameters(model,
model = architectures.MobileNetV3_large_x1_25(**kwargs) 'MobileNetV3_large_x1_25')
if pretrained:
model = _load_pretrained_parameters(model, 'MobileNetV3_large_x1_25') return model
return model def mobilenetv3_small_x0_35(pretrained=False, **kwargs):
"""
MobileNetV3_small_x0_35
def mobilenetv3_small_x0_35(pretrained=False, **kwargs): Args:
""" pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise.
MobileNetV3_small_x0_35 kwargs:
Args: class_dim: int=1000. Output dim of last fc layer.
pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. Returns:
kwargs: model: nn.Layer. Specific `MobileNetV3_small_x0_35` model depends on args.
class_dim: int=1000. Output dim of last fc layer. """
Returns: model = architectures.MobileNetV3_small_x0_35(**kwargs)
model: nn.Layer. Specific `MobileNetV3_small_x0_35` model depends on args. if pretrained:
""" model = _load_pretrained_parameters(model,
model = architectures.MobileNetV3_small_x0_35(**kwargs) 'MobileNetV3_small_x0_35')
if pretrained:
model = _load_pretrained_parameters(model, 'MobileNetV3_small_x0_35') return model
return model def mobilenetv3_small_x0_5(pretrained=False, **kwargs):
"""
MobileNetV3_small_x0_5
def mobilenetv3_small_x0_5(pretrained=False, **kwargs): Args:
""" pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise.
MobileNetV3_small_x0_5 kwargs:
Args: class_dim: int=1000. Output dim of last fc layer.
pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. Returns:
kwargs: model: nn.Layer. Specific `MobileNetV3_small_x0_5` model depends on args.
class_dim: int=1000. Output dim of last fc layer. """
Returns: model = architectures.MobileNetV3_small_x0_5(**kwargs)
model: nn.Layer. Specific `MobileNetV3_small_x0_5` model depends on args. if pretrained:
""" model = _load_pretrained_parameters(model,
model = architectures.MobileNetV3_small_x0_5(**kwargs) 'MobileNetV3_small_x0_5')
if pretrained:
model = _load_pretrained_parameters(model, 'MobileNetV3_small_x0_5') return model
return model def mobilenetv3_small_x0_75(pretrained=False, **kwargs):
"""
MobileNetV3_small_x0_75
def mobilenetv3_small_x0_75(pretrained=False, **kwargs): Args:
""" pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise.
MobileNetV3_small_x0_75 kwargs:
Args: class_dim: int=1000. Output dim of last fc layer.
pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. Returns:
kwargs: model: nn.Layer. Specific `MobileNetV3_small_x0_75` model depends on args.
class_dim: int=1000. Output dim of last fc layer. """
Returns: model = architectures.MobileNetV3_small_x0_75(**kwargs)
model: nn.Layer. Specific `MobileNetV3_small_x0_75` model depends on args. if pretrained:
""" model = _load_pretrained_parameters(model,
model = architectures.MobileNetV3_small_x0_75(**kwargs) 'MobileNetV3_small_x0_75')
if pretrained:
model = _load_pretrained_parameters(model, 'MobileNetV3_small_x0_75') return model
return model def mobilenetv3_small_x1_0(pretrained=False, **kwargs):
"""
MobileNetV3_small_x1_0
def mobilenetv3_small_x1_0(pretrained=False, **kwargs): Args:
""" pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise.
MobileNetV3_small_x1_0 kwargs:
Args: class_dim: int=1000. Output dim of last fc layer.
pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. Returns:
kwargs: model: nn.Layer. Specific `MobileNetV3_small_x1_0` model depends on args.
class_dim: int=1000. Output dim of last fc layer. """
Returns: model = architectures.MobileNetV3_small_x1_0(**kwargs)
model: nn.Layer. Specific `MobileNetV3_small_x1_0` model depends on args. if pretrained:
""" model = _load_pretrained_parameters(model,
model = architectures.MobileNetV3_small_x1_0(**kwargs) 'MobileNetV3_small_x1_0')
if pretrained:
model = _load_pretrained_parameters(model, 'MobileNetV3_small_x1_0') return model
return model def mobilenetv3_small_x1_25(pretrained=False, **kwargs):
"""
MobileNetV3_small_x1_25
def mobilenetv3_small_x1_25(pretrained=False, **kwargs): Args:
""" pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise.
MobileNetV3_small_x1_25 kwargs:
Args: class_dim: int=1000. Output dim of last fc layer.
pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. Returns:
kwargs: model: nn.Layer. Specific `MobileNetV3_small_x1_25` model depends on args.
class_dim: int=1000. Output dim of last fc layer. """
Returns: model = architectures.MobileNetV3_small_x1_25(**kwargs)
model: nn.Layer. Specific `MobileNetV3_small_x1_25` model depends on args. if pretrained:
""" model = _load_pretrained_parameters(model,
model = architectures.MobileNetV3_small_x1_25(**kwargs) 'MobileNetV3_small_x1_25')
if pretrained:
model = _load_pretrained_parameters(model, 'MobileNetV3_small_x1_25') return model
return model def resnext101_32x4d(pretrained=False, **kwargs):
"""
ResNeXt101_32x4d
def resnext101_32x4d(pretrained=False, **kwargs): Args:
""" pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise.
ResNeXt101_32x4d kwargs:
Args: class_dim: int=1000. Output dim of last fc layer.
pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. Returns:
kwargs: model: nn.Layer. Specific `ResNeXt101_32x4d` model depends on args.
class_dim: int=1000. Output dim of last fc layer. """
Returns: model = architectures.ResNeXt101_32x4d(**kwargs)
model: nn.Layer. Specific `ResNeXt101_32x4d` model depends on args. if pretrained:
""" model = _load_pretrained_parameters(model, 'ResNeXt101_32x4d')
model = architectures.ResNeXt101_32x4d(**kwargs)
if pretrained: return model
model = _load_pretrained_parameters(model, 'ResNeXt101_32x4d')
def resnext101_64x4d(pretrained=False, **kwargs):
return model """
ResNeXt101_64x4d
Args:
def resnext101_64x4d(pretrained=False, **kwargs): pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise.
""" kwargs:
ResNeXt101_64x4d class_dim: int=1000. Output dim of last fc layer.
Args: Returns:
pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. model: nn.Layer. Specific `ResNeXt101_64x4d` model depends on args.
kwargs: """
class_dim: int=1000. Output dim of last fc layer. model = architectures.ResNeXt101_64x4d(**kwargs)
Returns: if pretrained:
model: nn.Layer. Specific `ResNeXt101_64x4d` model depends on args. model = _load_pretrained_parameters(model, 'ResNeXt101_64x4d')
"""
model = architectures.ResNeXt101_64x4d(**kwargs) return model
if pretrained:
model = _load_pretrained_parameters(model, 'ResNeXt101_64x4d') def resnext152_32x4d(pretrained=False, **kwargs):
"""
return model ResNeXt152_32x4d
Args:
pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise.
def resnext152_32x4d(pretrained=False, **kwargs): kwargs:
""" class_dim: int=1000. Output dim of last fc layer.
ResNeXt152_32x4d Returns:
Args: model: nn.Layer. Specific `ResNeXt152_32x4d` model depends on args.
pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. """
kwargs: model = architectures.ResNeXt152_32x4d(**kwargs)
class_dim: int=1000. Output dim of last fc layer. if pretrained:
Returns: model = _load_pretrained_parameters(model, 'ResNeXt152_32x4d')
model: nn.Layer. Specific `ResNeXt152_32x4d` model depends on args.
""" return model
model = architectures.ResNeXt152_32x4d(**kwargs)
if pretrained: def resnext152_64x4d(pretrained=False, **kwargs):
model = _load_pretrained_parameters(model, 'ResNeXt152_32x4d') """
ResNeXt152_64x4d
return model Args:
pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise.
kwargs:
def resnext152_64x4d(pretrained=False, **kwargs): class_dim: int=1000. Output dim of last fc layer.
""" Returns:
ResNeXt152_64x4d model: nn.Layer. Specific `ResNeXt152_64x4d` model depends on args.
Args: """
pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. model = architectures.ResNeXt152_64x4d(**kwargs)
kwargs: if pretrained:
class_dim: int=1000. Output dim of last fc layer. model = _load_pretrained_parameters(model, 'ResNeXt152_64x4d')
Returns:
model: nn.Layer. Specific `ResNeXt152_64x4d` model depends on args. return model
"""
model = architectures.ResNeXt152_64x4d(**kwargs) def resnext50_32x4d(pretrained=False, **kwargs):
if pretrained: """
model = _load_pretrained_parameters(model, 'ResNeXt152_64x4d') ResNeXt50_32x4d
Args:
return model pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise.
kwargs:
class_dim: int=1000. Output dim of last fc layer.
def resnext50_32x4d(pretrained=False, **kwargs): Returns:
""" model: nn.Layer. Specific `ResNeXt50_32x4d` model depends on args.
ResNeXt50_32x4d """
Args: model = architectures.ResNeXt50_32x4d(**kwargs)
pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. if pretrained:
kwargs: model = _load_pretrained_parameters(model, 'ResNeXt50_32x4d')
class_dim: int=1000. Output dim of last fc layer.
Returns: return model
model: nn.Layer. Specific `ResNeXt50_32x4d` model depends on args.
""" def resnext50_64x4d(pretrained=False, **kwargs):
model = architectures.ResNeXt50_32x4d(**kwargs) """
if pretrained: ResNeXt50_64x4d
model = _load_pretrained_parameters(model, 'ResNeXt50_32x4d') Args:
pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise.
return model kwargs:
class_dim: int=1000. Output dim of last fc layer.
Returns:
def resnext50_64x4d(pretrained=False, **kwargs): model: nn.Layer. Specific `ResNeXt50_64x4d` model depends on args.
""" """
ResNeXt50_64x4d model = architectures.ResNeXt50_64x4d(**kwargs)
Args: if pretrained:
pretrained: bool=False. If `True` load pretrained parameters, `False` otherwise. model = _load_pretrained_parameters(model, 'ResNeXt50_64x4d')
kwargs:
class_dim: int=1000. Output dim of last fc layer. return model
Returns:
model: nn.Layer. Specific `ResNeXt50_64x4d` model depends on args.
"""
model = architectures.ResNeXt50_64x4d(**kwargs)
if pretrained:
model = _load_pretrained_parameters(model, 'ResNeXt50_64x4d')
return model
Markdown is supported
0% 或 .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑！
想要评论请 注册 或