Inheritance diagram for ccbn:

Collaboration diagram for ccbn:

Public Member Functions
	__init__ (self, output_size, input_size, which_linear, eps=1e-5, momentum=0.1, cross_replica=False, mybn=False, norm_style="bn")
	Constructor.

	forward (self, x, y)
	forward

	extra_repr (self)
	extra_repr

Public Attributes
	output_size
	output size

	input_size = output_size, input_size
	output size

	gain = which_linear(input_size, output_size)
	Prepare gain and bias layers.

	bias = which_linear(input_size, output_size)
	bias

	eps = eps
	epsilon to avoid dividing by 0

	momentum = momentum
	Momentum.

	cross_replica = cross_replica
	Use cross-replica batchnorm?

	mybn = mybn
	Use my batchnorm?

str	norm_style = norm_style
	Norm style?

	bn = myBN(output_size, self.eps, self.momentum)
	bn

Private Attributes
	__dict__

Detailed Description

Class-conditional bn
output size is the number of channels, input size is for the linear layers
Andy's Note: this class feels messy but I'm not really sure how to clean it up  # noqa
Suggestions welcome! (By which I mean, refactor this and make a merge request
if you want to make this more readable/usable).

Definition at line 516 of file ieagan.py.

Constructor & Destructor Documentation

◆ init()

__init__	(	self,
		output_size,
		input_size,
		which_linear,
		eps = 1e-5,
		momentum = 0.1,
		cross_replica = False,
		mybn = False,
		norm_style = "bn" )

Constructor.

Definition at line 526 of file ieagan.py.

    ):
        super(ccbn, self).__init__()
        
        self.output_size, self.input_size = output_size, input_size
        
        self.gain = which_linear(input_size, output_size)
        
        self.bias = which_linear(input_size, output_size)
        
        self.eps = eps
        
        self.momentum = momentum
        
        self.cross_replica = cross_replica
        
        self.mybn = mybn
        
        self.norm_style = norm_style
 
        if self.mybn:
            
            self.bn = myBN(output_size, self.eps, self.momentum)
        elif self.norm_style in ["bn", "in"]:
            self.register_buffer("stored_mean", torch.zeros(output_size))
            self.register_buffer("stored_var", torch.ones(output_size))

Member Function Documentation

◆ extra_repr()

extra_repr ( self )

extra_repr

Definition at line 601 of file ieagan.py.

    def extra_repr(self):
        s = "out: {output_size}, in: {input_size},"
        s += " cross_replica={cross_replica}"
        return s.format(**self.__dict__)
 
 

◆ forward()

forward	(	self,
		x,
		y )

forward

Definition at line 563 of file ieagan.py.

    def forward(self, x, y):
        # Calculate class-conditional gains and biases
        gain = (1 + self.gain(y)).view(y.size(0), -1, 1, 1)
        bias = self.bias(y).view(y.size(0), -1, 1, 1)
        # If using my batchnorm
        if self.mybn:
            return self.bn(x, gain=gain, bias=bias)
        # else:
        else:
            if self.norm_style == "bn":
                out = F.batch_norm(
                    x,
                    self.stored_mean,
                    self.stored_var,
                    None,
                    None,
                    self.training,
                    0.1,
                    self.eps,
                )
            elif self.norm_style == "in":
                out = F.instance_norm(
                    x,
                    self.stored_mean,
                    self.stored_var,
                    None,
                    None,
                    self.training,
                    0.1,
                    self.eps,
                )
            elif self.norm_style == "gn":
                out = groupnorm(x, self.normstyle)
            elif self.norm_style == "nonorm":
                out = x
            return out * gain + bias
 

Member Data Documentation

◆ dict

__dict__

private

Definition at line 604 of file ieagan.py.

◆ bias

bias = which_linear(input_size, output_size)

bias

Definition at line 543 of file ieagan.py.

◆ bn

bn = myBN(output_size, self.eps, self.momentum)

bn

Definition at line 557 of file ieagan.py.

◆ cross_replica

cross_replica = cross_replica

Use cross-replica batchnorm?

Definition at line 549 of file ieagan.py.

◆ eps

eps = eps

epsilon to avoid dividing by 0

Definition at line 545 of file ieagan.py.

◆ gain

gain = which_linear(input_size, output_size)

Prepare gain and bias layers.

Definition at line 541 of file ieagan.py.

◆ input_size

input_size = output_size, input_size

output size

Definition at line 539 of file ieagan.py.

◆ momentum

momentum = momentum

Momentum.

Definition at line 547 of file ieagan.py.

◆ mybn

mybn = mybn

Use my batchnorm?

Definition at line 551 of file ieagan.py.

◆ norm_style

str norm_style = norm_style

Norm style?

Definition at line 553 of file ieagan.py.

◆ output_size

output_size