yamle.models.resnet module#

class yamle.models.resnet.BasicBlock(inplanes, planes, stride=1, normalization=<class 'torch.nn.modules.batchnorm.BatchNorm2d'>, normalization_kwargs={}, first=False)[source]#

Bases: Module

A basic block for ResNet. It consists of two convolutional layers with batch normalization and ReLU activation. :type _sphinx_paramlinks_yamle.models.resnet.BasicBlock.inplanes: int :param _sphinx_paramlinks_yamle.models.resnet.BasicBlock.inplanes: The number of input channels. :type _sphinx_paramlinks_yamle.models.resnet.BasicBlock.inplanes: int :type _sphinx_paramlinks_yamle.models.resnet.BasicBlock.planes: int :param _sphinx_paramlinks_yamle.models.resnet.BasicBlock.planes: The number of output channels. :type _sphinx_paramlinks_yamle.models.resnet.BasicBlock.planes: int :type _sphinx_paramlinks_yamle.models.resnet.BasicBlock.stride: int :param _sphinx_paramlinks_yamle.models.resnet.BasicBlock.stride: The stride of the first convolutional layer. :type _sphinx_paramlinks_yamle.models.resnet.BasicBlock.stride: int :type _sphinx_paramlinks_yamle.models.resnet.BasicBlock.normalization: Type[Module] :param _sphinx_paramlinks_yamle.models.resnet.BasicBlock.normalization: The normalization to use. :type _sphinx_paramlinks_yamle.models.resnet.BasicBlock.normalization: Type[nn.Module] :type _sphinx_paramlinks_yamle.models.resnet.BasicBlock.normalization_kwargs: Dict[str, Any] :param _sphinx_paramlinks_yamle.models.resnet.BasicBlock.normalization_kwargs: The keyword arguments for the normalization. :type _sphinx_paramlinks_yamle.models.resnet.BasicBlock.normalization_kwargs: Dict[str, Any] :type _sphinx_paramlinks_yamle.models.resnet.BasicBlock.first: bool :param _sphinx_paramlinks_yamle.models.resnet.BasicBlock.first: Whether this is the first block in the sequence. Defaults to False. :type _sphinx_paramlinks_yamle.models.resnet.BasicBlock.first: bool

expansion = 1#
forward(x)[source]#

The forward function of the basic block.

Return type:

Tensor

replace_layers_for_quantization()[source]#

Fuses all the operations in the block.

Return type:

None

training: bool#
class yamle.models.resnet.Bottleneck(inplanes, planes, stride=1, normalization=<class 'torch.nn.modules.batchnorm.BatchNorm2d'>, normalization_kwargs={}, first=False)[source]#

Bases: Module

A bottleneck block for ResNet. It consists of 3 convolutional layers with batch normalization and ReLU activation. It is used to construct wide ResNet.

Parameters:

  • inplanes (int) – The number of input channels.

  • planes (int) – The number of output channels.

  • stride (int) – The stride of the first convolutional layer.

  • normalization (Type[nn.Module]) – The normalization to use.

  • normalization_kwargs (Dict[str, Any]) – The keyword arguments for the normalization.

  • first (bool) – Whether this is the first block in the sequence. Defaults to False.

base_width_multiplier = 2#
groups = 1#
expansion = 4#
base_width = 32#
forward(x)[source]#

The forward function of the bottleneck block.

Return type:

Tensor

replace_layers_for_quantization()[source]#

Fuses all the operations in the block.

Return type:

None

training: bool#
class yamle.models.resnet.Blocks(inplanes, planes, blocks, stride, normalization, normalization_kwargs={}, block=<class 'yamle.models.resnet.BasicBlock'>)[source]#

Bases: Module

A class implementing a stack of basic blocks. :type _sphinx_paramlinks_yamle.models.resnet.Blocks.inplanes: int :param _sphinx_paramlinks_yamle.models.resnet.Blocks.inplanes: The number of input channels. :type _sphinx_paramlinks_yamle.models.resnet.Blocks.inplanes: int :type _sphinx_paramlinks_yamle.models.resnet.Blocks.planes: int :param _sphinx_paramlinks_yamle.models.resnet.Blocks.planes: The number of output channels. :type _sphinx_paramlinks_yamle.models.resnet.Blocks.planes: int :type _sphinx_paramlinks_yamle.models.resnet.Blocks.blocks: int :param _sphinx_paramlinks_yamle.models.resnet.Blocks.blocks: The number of basic blocks. :type _sphinx_paramlinks_yamle.models.resnet.Blocks.blocks: int :type _sphinx_paramlinks_yamle.models.resnet.Blocks.stride: int :param _sphinx_paramlinks_yamle.models.resnet.Blocks.stride: The stride of the first convolutional layer. :type _sphinx_paramlinks_yamle.models.resnet.Blocks.stride: int :type _sphinx_paramlinks_yamle.models.resnet.Blocks.normalization: Type[Module] :param _sphinx_paramlinks_yamle.models.resnet.Blocks.normalization: The normalization to use. :type _sphinx_paramlinks_yamle.models.resnet.Blocks.normalization: Type[nn.Module] :type _sphinx_paramlinks_yamle.models.resnet.Blocks.normalization_kwargs: Dict[str, Any] :param _sphinx_paramlinks_yamle.models.resnet.Blocks.normalization_kwargs: The keyword arguments for the normalization. :type _sphinx_paramlinks_yamle.models.resnet.Blocks.normalization_kwargs: Dict[str, Any] :type _sphinx_paramlinks_yamle.models.resnet.Blocks.block: Union[Type[BasicBlock], Type[Bottleneck]] :param _sphinx_paramlinks_yamle.models.resnet.Blocks.block: The block to use. Defaults to BasicBlock. :type _sphinx_paramlinks_yamle.models.resnet.Blocks.block: Union[Type[BasicBlock], Type[Bottleneck]]

forward(x)[source]#

The forward function of the stack of basic blocks.

Return type:

Tensor

training: bool#
class yamle.models.resnet.ResNetModel(layers=[2, 2, 2, 2], depth=None, planes=[32, 64, 128, 256], width_multiplier=1, strides=[1, 2, 2, 2], normalization='batch', block='basic', *args, **kwargs)[source]#

Bases: BaseModel

This class implements the ResNet architecture as described in the paper: Deep Residual Learning for Image Recognition, published at the IEEE Conference on Computer Vision and Pattern Recognition, CVPR, 2016. and the paper can be found at: https://arxiv.org/abs/1512.03385.

The code is based on the implementation of the model is sourced from: https://pytorch.org/hub/pytorch_vision_resnet/

Parameters:

  • layers (List[int]) – The number of basic blocks in each stack.

  • depth (int) – The depth of the network respective to the the length of the layers list.

  • planes (List[int]) – The number of output channels in each stack.

  • width_multiplier (int) – The width multiplier for the planes list.

  • strides (List[int]) – The stride of the first convolutional layer in each stack.

  • normalization (Optional[str]) – The normalization to use. Can be either ‘batch’, ‘instance’, group, layer, or None. Defaults to ‘batch’.

  • block (str) – The block to use. Can be either ‘basic’ or ‘bottleneck’. Defaults to ‘basic’.

tasks = ['regression', 'classification']#
training: bool#
forward(x, staged_output=False, input_kwargs={}, output_kwargs={})[source]#

The forward function of the ResNet model.

Parameters:

  • x (torch.Tensor) – The input tensor.

  • staged_output (bool) – Whether to return the output of each layer. Defaults to False.

  • input_kwargs (Dict[str, Any]) – The kwargs for the input layer.

  • output_kwargs (Dict[str, Any]) – The kwargs for the output layer.

Return type:

Union[Tensor, Tuple[Tensor, List[Tensor]]]

final_layer(x, **output_kwargs)[source]#

This function is used to get the final layer output.

Return type:

Tensor

add_method_specific_layers(method, **kwargs)[source]#

This method is used to add method specific layers to the model.

Parameters:

method (str) – The method to use.

Return type:

None

replace_layers_for_quantization()[source]#

Fuses all the operations in the network.

In this function we only need to fuse layers that are not in the blocks. e.g. the reshaping layers added by the method.

Return type:

None

static add_specific_args(parent_parser)[source]#

Add specific arguments to the parser.

Return type:

ArgumentParser