How to use the batchflow.models.torch.resnet.ResNet function in batchflow
To help you get started, we’ve selected a few batchflow examples, based on popular ways it is used in public projects.
Secure your code as it's written. Use Snyk Code to scan source code in minutes - no build needed - and fix issues immediately.
analysiscenter / batchflow / batchflow / models / torch / resnet.py View on Github 
def default_config(cls):
config = ResNet.default_config()
config['body/num_blocks'] = [3, 8, 36, 3]
config['body/block/bottleneck'] = True
return config
class ResNeXt18(ResNet):
""" The ResNeXt-18 architecture """
@classmethod
def default_config(cls):
config = ResNet18.default_config()
config['body/block/resnext'] = True
return config
class ResNeXt34(ResNet):
""" The ResNeXt-34 architecture """
@classmethod
def default_config(cls):
config = ResNet34.default_config()
config['body/block/resnext'] = True
return config
class ResNeXt50(ResNet):
""" The ResNeXt-50 architecture """
@classmethod
def default_config(cls):
config = ResNet50.default_config()
config['body/block/resnext'] = True
return configconfig['body/num_blocks'] = [2, 2, 2, 2]
config['body/block/bottleneck'] = False
return config
class ResNet34(ResNet):
""" The original ResNet-34 architecture """
@classmethod
def default_config(cls):
config = ResNet.default_config()
config['body/num_blocks'] = [3, 4, 6, 3]
config['body/block/bottleneck'] = False
return config
class ResNet50(ResNet):
""" The original ResNet-50 architecture """
@classmethod
def default_config(cls):
config = ResNet34.default_config()
config['body/block/bottleneck'] = True
return config
class ResNet101(ResNet):
""" The original ResNet-101 architecture """
@classmethod
def default_config(cls):
config = ResNet.default_config()
config['body/num_blocks'] = [3, 4, 23, 3]
config['body/block/bottleneck'] = True
return configconfig = ResNet34.default_config()
config['body/block/bottleneck'] = True
return config
class ResNet101(ResNet):
""" The original ResNet-101 architecture """
@classmethod
def default_config(cls):
config = ResNet.default_config()
config['body/num_blocks'] = [3, 4, 23, 3]
config['body/block/bottleneck'] = True
return config
class ResNet152(ResNet):
""" The original ResNet-152 architecture """
@classmethod
def default_config(cls):
config = ResNet.default_config()
config['body/num_blocks'] = [3, 8, 36, 3]
config['body/block/bottleneck'] = True
return config
class ResNeXt18(ResNet):
""" The ResNeXt-18 architecture """
@classmethod
def default_config(cls):
config = ResNet18.default_config()
config['body/block/resnext'] = True
return configif self.padding:
shortcut = F.pad(shortcut, self.padding)
return self.conv(x) + shortcut
class ResNet18(ResNet):
""" The original ResNet-18 architecture """
@classmethod
def default_config(cls):
config = ResNet.default_config()
config['body/num_blocks'] = [2, 2, 2, 2]
config['body/block/bottleneck'] = False
return config
class ResNet34(ResNet):
""" The original ResNet-34 architecture """
@classmethod
def default_config(cls):
config = ResNet.default_config()
config['body/num_blocks'] = [3, 4, 6, 3]
config['body/block/bottleneck'] = False
return config
class ResNet50(ResNet):
""" The original ResNet-50 architecture """
@classmethod
def default_config(cls):
config = ResNet34.default_config()
config['body/block/bottleneck'] = True
return configinput tensor
skip
skip connection
Returns
-------
nn.Module
"""
kwargs = cls.get_defaults('body', kwargs)
layout, filters, kernel_size = cls.pop(['layout', 'filters', 'kernel_size'], kwargs)
upsample_args = cls.pop('upsample', kwargs)
x = cls.upsample(inputs, filters=filters, name='upsample', **upsample_args, **kwargs)
x = cls.crop(x, skip, data_format=kwargs.get('data_format'))
x = torch.cat([skip, x], dim=1)
x = ResNet.block(x, layout=layout, filters=filters, kernel_size=kernel_size, downsample=0, **kwargs)
return x
def __init__(self, conv, shortcut, padding=None):
super().__init__()
self.conv = conv
self.shortcut = shortcut
self.padding = padding
self.output_shape = get_shape(conv)
def forward(self, x):
""" Make a forward pass """
shortcut = self.shortcut(x) if self.shortcut else x
if self.padding:
shortcut = F.pad(shortcut, self.padding)
return self.conv(x) + shortcut
class ResNet18(ResNet):
""" The original ResNet-18 architecture """
@classmethod
def default_config(cls):
config = ResNet.default_config()
config['body/num_blocks'] = [2, 2, 2, 2]
config['body/block/bottleneck'] = False
return config
class ResNet34(ResNet):
""" The original ResNet-34 architecture """
@classmethod
def default_config(cls):
config = ResNet.default_config()
config['body/num_blocks'] = [3, 4, 6, 3]
config['body/block/bottleneck'] = Falseconfig = ResNet.default_config()
config['body/num_blocks'] = [3, 4, 6, 3]
config['body/block/bottleneck'] = False
return config
class ResNet50(ResNet):
""" The original ResNet-50 architecture """
@classmethod
def default_config(cls):
config = ResNet34.default_config()
config['body/block/bottleneck'] = True
return config
class ResNet101(ResNet):
""" The original ResNet-101 architecture """
@classmethod
def default_config(cls):
config = ResNet.default_config()
config['body/num_blocks'] = [3, 4, 23, 3]
config['body/block/bottleneck'] = True
return config
class ResNet152(ResNet):
""" The original ResNet-152 architecture """
@classmethod
def default_config(cls):
config = ResNet.default_config()
config['body/num_blocks'] = [3, 8, 36, 3]
config['body/block/bottleneck'] = Truedef __init__(self, conv, shortcut, padding=None):
super().__init__()
self.conv = conv
self.shortcut = shortcut
self.padding = padding
self.output_shape = get_shape(conv)
def forward(self, x):
""" Make a forward pass """
shortcut = self.shortcut(x) if self.shortcut else x
if self.padding:
shortcut = F.pad(shortcut, self.padding)
return self.conv(x) + shortcut
class ResNet18(ResNet):
""" The original ResNet-18 architecture """
@classmethod
def default_config(cls):
config = ResNet.default_config()
config['body/num_blocks'] = [2, 2, 2, 2]
config['body/block/bottleneck'] = False
return config
class ResNet34(ResNet):
""" The original ResNet-34 architecture """
@classmethod
def default_config(cls):
config = ResNet.default_config()
config['body/num_blocks'] = [3, 4, 6, 3]
config['body/block/bottleneck'] = Falseconfig = ResNet.default_config()
config['body/num_blocks'] = [3, 4, 6, 3]
config['body/block/bottleneck'] = False
return config
class ResNet50(ResNet):
""" The original ResNet-50 architecture """
@classmethod
def default_config(cls):
config = ResNet34.default_config()
config['body/block/bottleneck'] = True
return config
class ResNet101(ResNet):
""" The original ResNet-101 architecture """
@classmethod
def default_config(cls):
config = ResNet.default_config()
config['body/num_blocks'] = [3, 4, 23, 3]
config['body/block/bottleneck'] = True
return config
class ResNet152(ResNet):
""" The original ResNet-152 architecture """
@classmethod
def default_config(cls):
config = ResNet.default_config()
config['body/num_blocks'] = [3, 8, 36, 3]
config['body/block/bottleneck'] = TrueParameters
----------
inputs
input tensor
Returns
-------
nn.Module
"""
kwargs = cls.get_defaults('body', kwargs)
layout, kernel_size = cls.pop(['layout', 'kernel_size'], kwargs)
x, inputs = inputs, None
if downsample:
x = ConvBlock(x, layout='cna', kernel_size=2, strides=2, **kwargs)
x = ResNet.block(x, layout=layout, kernel_size=kernel_size, downsample=False, **kwargs)
return x
batchflow
ML pipelines, model configuration and batch management
Package Health Score
64 / 100Popular batchflow functions
- batchflow.base.Baseset.get_default_iter_params
- batchflow.Config
- batchflow.config.Config
- batchflow.decorators.action
- batchflow.models.eager_torch.utils.get_shape
- batchflow.models.tf.layers.conv_block
- batchflow.models.tf.resnet.ResNet
- batchflow.models.tf.resnet.ResNet.block
- batchflow.models.tf.TFModel
- batchflow.models.tf.TFModel.default_config
- batchflow.models.torch.layers.ConvBlock
- batchflow.models.torch.resnet.ResNet
- batchflow.models.torch.TorchModel
- batchflow.models.torch.TorchModel.default_config
- batchflow.models.torch.utils.get_shape
- batchflow.models.torch.utils.unpack_fn_from_config
- batchflow.models.utils.unpack_args
- batchflow.named_expr.eval_expr
- batchflow.named_expr.NamedExpression
- batchflow.sampler.Sampler