CNNBackBone Class Reference

Implements a multi-layer convolutional neural network, with leaky-ReLU non-linearities between layers, according to hyperparameters specified in the config. More...

Inheritance diagram for CNNBackBone:

Public Member Functions
	__init__ (self, dict config)
torch.Tensor	forward (self, torch.Tensor x)
	Forward pass through the network.
Public Member Functions inherited from CNNConfigParser
None	parse (self, dict config)
	Parse the configuration for the CNN model.

Public Attributes
	input_dim
	latent_size = self.kernel_size
tuple	flatten_size
	num_layers
Public Attributes inherited from CNNConfigParser
	config = None
	input_dim = None
	output_dim = None
	num_layers = None
	latent_size = None
	kernel_size = None
	image_size = None

Detailed Description

Implements a multi-layer convolutional neural network, with leaky-ReLU non-linearities between layers, according to hyperparameters specified in the config.

Definition at line 35 of file cnn_backbone.py.

Constructor & Destructor Documentation

__init__()

__init__	(		self,
		dict	config )

Reimplemented from CNNConfigParser.

Definition at line 37 of file cnn_backbone.py.

Member Function Documentation

forward()

torch.Tensor forward	(		self,
		torch.Tensor	x )

Forward pass through the network.

Parameters

x	input tensor

Definition at line 77 of file cnn_backbone.py.

Member Data Documentation

flatten_size

flatten_size

Initial value:

=  (
            self.latent_size
            * (self.image_size - self.num_layers * (self.kernel_size - 1)) ** 2
        )

Definition at line 60 of file cnn_backbone.py.

input_dim

input_dim

Definition at line 44 of file cnn_backbone.py.

latent_size

latent_size = self.kernel_size

Definition at line 44 of file cnn_backbone.py.

num_layers

num_layers

Definition at line 78 of file cnn_backbone.py.