fastreid.modeling.losses.smooth_ap Namespace Reference

Classes
class	SmoothAP
class	SmoothAP_old

Functions
	sigmoid (tensor, temp=1.0)

Variables
	loss1 = SmoothAP(0.01)
	loss2 = SmoothAP_old(0.01, 60, 6, 256)
	inputs = torch.randn(60, 256, requires_grad=True)
list	targets = []
	output1 = loss1(inputs, targets)
	output2 = loss2(inputs)

Detailed Description

@author:  xingyu liao
@contact: sherlockliao01@gmail.com

Function Documentation

sigmoid()

fastreid.modeling.losses.smooth_ap.sigmoid	(		tensor,
			temp = 1.0 )

 temperature controlled sigmoid
takes as input a torch tensor (tensor) and passes it through a sigmoid, controlled by temperature: temp

Definition at line 17 of file smooth_ap.py.

def sigmoid(tensor, temp=1.0):
    """ temperature controlled sigmoid
    takes as input a torch tensor (tensor) and passes it through a sigmoid, controlled by temperature: temp
    """
    exponent = -tensor / temp
    # clamp the input tensor for stability
    exponent = torch.clamp(exponent, min=-50, max=50)
    y = 1.0 / (1.0 + torch.exp(exponent))
    return y
 
 

Variable Documentation

inputs

fastreid.modeling.losses.smooth_ap.inputs = torch.randn(60, 256, requires_grad=True)

Definition at line 232 of file smooth_ap.py.

loss1

fastreid.modeling.losses.smooth_ap.loss1 = SmoothAP(0.01)

Definition at line 229 of file smooth_ap.py.

loss2

fastreid.modeling.losses.smooth_ap.loss2 = SmoothAP_old(0.01, 60, 6, 256)

Definition at line 230 of file smooth_ap.py.

output1

fastreid.modeling.losses.smooth_ap.output1 = loss1(inputs, targets)

Definition at line 238 of file smooth_ap.py.

output2

fastreid.modeling.losses.smooth_ap.output2 = loss2(inputs)

Definition at line 239 of file smooth_ap.py.

targets

fastreid.modeling.losses.smooth_ap.targets = []

Definition at line 233 of file smooth_ap.py.