bn Class Reference

Inheritance diagram for bn:

Public Member Functions
	__init__ (self, output_size, eps=1e-5, momentum=0.1, cross_replica=False, mybn=False)
	Constructor.
	forward (self, x)
	forward

Public Attributes
	output_size = output_size
	output size
	gain = P(torch.ones(output_size), requires_grad=True)
	Prepare gain and bias layers.
	bias = P(torch.zeros(output_size), requires_grad=True)
	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?
	stored_mean = torch.zeros(output_size)
	Running mean buffer, updated during training.
	stored_var = torch.ones(output_size)
	Running variance buffer, updated during training.
bool	training
	Training mode flag (inherited from nn.Module).

Detailed Description

Normal, non-class-conditional BN

Definition at line 479 of file ieagan.py.

Constructor & Destructor Documentation

__init__()

__init__	(		self,
			output_size,
			eps = 1e-5,
			momentum = 0.1,
			cross_replica = False,
			mybn = False )

Constructor.

Definition at line 485 of file ieagan.py.

    ):
        super(bn, self).__init__()
        
        self.output_size = output_size
        
        self.gain = P(torch.ones(output_size), requires_grad=True)
        
        self.bias = P(torch.zeros(output_size), requires_grad=True)
        
        self.eps = eps
        
        self.momentum = momentum
        
        self.cross_replica = cross_replica
        
        self.mybn = mybn
 
        if mybn:
            self.bn = myBN(output_size, self.eps, self.momentum)
        # Register buffers if neither of the above
        else:
            
            self.stored_mean = torch.zeros(output_size)
            self.register_buffer("stored_mean", torch.zeros(output_size))
            
            self.stored_var = torch.ones(output_size)
            self.register_buffer("stored_var", torch.ones(output_size))
 
        
        self.training: bool
 

Member Function Documentation

forward()

forward	(		self,
			x )

forward

Definition at line 524 of file ieagan.py.

    def forward(self, x):
        if self.mybn:
            gain = self.gain.view(1, -1, 1, 1)
            bias = self.bias.view(1, -1, 1, 1)
            return self.bn(x, gain=gain, bias=bias)
        else:
            return F.batch_norm(
                x,
                self.stored_mean,
                self.stored_var,
                self.gain,
                self.bias,
                self.training,
                self.momentum,
                self.eps,
            )
 
 

Member Data Documentation

bias

bias = P(torch.zeros(output_size), requires_grad=True)

bias

Definition at line 499 of file ieagan.py.

bn

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

Definition at line 510 of file ieagan.py.

cross_replica

cross_replica = cross_replica

Use cross-replica batchnorm?

Definition at line 505 of file ieagan.py.

eps

eps = eps

epsilon to avoid dividing by 0

Definition at line 501 of file ieagan.py.

gain

gain = P(torch.ones(output_size), requires_grad=True)

Prepare gain and bias layers.

Definition at line 497 of file ieagan.py.

momentum

momentum = momentum

Momentum.

Definition at line 503 of file ieagan.py.

mybn

mybn = mybn

Use my batchnorm?

Definition at line 507 of file ieagan.py.

output_size

output_size = output_size

output size

Definition at line 495 of file ieagan.py.

stored_mean

stored_mean = torch.zeros(output_size)

Running mean buffer, updated during training.

Definition at line 514 of file ieagan.py.

stored_var

stored_var = torch.ones(output_size)

Running variance buffer, updated during training.

Definition at line 517 of file ieagan.py.

training

training

Training mode flag (inherited from nn.Module).

True if the module is in training mode.

Definition at line 521 of file ieagan.py.

---

The documentation for this class was generated from the following file:

• pxd/scripts/pxd/background_generator/models/ieagan.py