-
Notifications
You must be signed in to change notification settings - Fork 1
/
cycle_loss.py
36 lines (25 loc) · 1018 Bytes
/
cycle_loss.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
#!/usr/bin/env python3
# -*- coding: utf-8 -*-
"""
Created on Tue Aug 9 16:15:35 2022
@author: ahmedemam576
"""
from gan_loss_term import Gan_loss_term
class Cycle_Loss(Gan_loss_term):
def __init__(self, real_x, generator, name, norm):
super(Cycle_Loss, self).__init__(real_x, generator, name, norm)
'''
the class should inheret all his attributes from the Gan_loss_term
so we shouldn't set each attribute to .self
this loss should be used two time for opposite generators
args:
generator is Gen y--> x
norm is L1 norm
'''
print(f'{self.name}is initiated')
def __call__(self, fake_y):
''' fake y is tcalculated when calculating the adverserial loss in the direction from x->y '''
gen_y_to_x = self.generator
fake_x = gen_y_to_x(fake_y)
cycle_loss= self.norm(fake_x, self.real_x)
return cycle_loss