import numpy as np

import torch

import torch.nn as nn

class LogLoss(nn.BCEWithLogitsLoss):

    def __init__(self, weight=None, size_average=None, reduce=None, reduction='mean'):

        if weight is None:

            pass

        else:

            weight = torch.tensor(weight, requires_grad=False, dtype=torch.float32).cuda()

        super(LogLoss, self).__init__(

            weight=weight,

            size_average=size_average,

            reduce=reduce,

            reduction=reduction

        )