bitorch.models.resnet.SpecificResnet

class bitorch.models.resnet.SpecificResnet(classes: int, channels: list)

Superclass for ResNet models

Methods

__init__	builds feature and output layers
forward	forwards the input tensor through the resnet modules
make_feature_layers	builds the given layers with the specified block.
make_layer	builds a layer by stacking blocks in a sequential models.

Attributes

__init__(classes: int, channels: list) → None

builds feature and output layers

Parameters:

• classes (int) – number of output classes

• channels (list) – the channels used in the net

forward(x: Tensor) → Tensor

forwards the input tensor through the resnet modules

Parameters:

x (torch.Tensor) – input tensor

Returns:

forwarded tensor

Return type:

torch.Tensor

make_feature_layers(block: Module, layers: list, channels: list) → List[Module]

builds the given layers with the specified block.

Parameters:

• block (Module) – the block of which the layer shall consist

• layers (list) – the number of blocks each layer shall consist of

• channels (list) – the channels

Returns:

[description]

Return type:

nn.Sequential

make_layer(block: Module, layers: int, in_channels: int, out_channels: int, stride: int) → Sequential

builds a layer by stacking blocks in a sequential models.

Parameters:

• block (Module) – the block of which the layer shall consist

• layers (int) – the number of blocks to stack

• in_channels (int) – the input channels of this layer

• out_channels (int) – the output channels of this layer

• stride (int) – the stride to be used in the convolution layers

Returns:

the model containing the building blocks

Return type:

nn.Sequential