[ Pytorch ] —— 复现积累

版权声明：本文为博主原创文章，遵循 CC 4.0 BY-SA 版权协议，转载请附上原文出处链接和本声明。

本文链接： https://blog.csdn.net/jdzwanghao/article/details/97371299

一、Loss

1、Triplet Loss

代码转载自：丨github丨

from __future__ import absolute_import

import torch
from torch import nn
from torch.autograd import Variable

class TripletLoss(nn.Module):
    def __init__(self, margin=0):
        super(TripletLoss, self).__init__()
        self.margin = margin
        self.ranking_loss = nn.MarginRankingLoss(margin=margin)

    def forward(self, inputs, targets):
        n = inputs.size(0)
        # Compute pairwise distance, replace by the official when merged
        dist = torch.pow(inputs, 2).sum(dim=1, keepdim=True).expand(n, n)
        dist = dist + dist.t()
        dist.addmm_(1, -2, inputs, inputs.t())   # 解释：https://blog.csdn.net/qq_36556893/article/details/90638449
        dist = dist.clamp(min=1e-12).sqrt()  # for numerical stability
        # For each anchor, find the hardest positive and negative
        mask = targets.expand(n, n).eq(targets.expand(n, n).t())  # reture bool value: [True, True, False, False]
        dist_ap, dist_an = [], []
        for i in range(n):
            dist_ap.append(dist[i][mask[i]].max().view(1))
            dist_an.append(dist[i][mask[i] == 0].min().view(1))
        dist_ap = torch.cat(dist_ap)
        dist_an = torch.cat(dist_an)
        # Compute ranking hinge loss
        y = dist_an.data.new()
        y.resize_as_(dist_an.data)
        y.fill_(1)
        y = Variable(y)
        loss = self.ranking_loss(dist_an, dist_ap, y)
        prec = (dist_an.data > dist_ap).data.float().mean()
        return loss, prec


######## 用下面代码去理解上面的原理 #########
import torch

t1 = Variable(torch.FloatTensor(torch.rand([4,128])))
t1 = t1.cuda()

n = t1.size(0)
dist = torch.pow(t1, 2).sum(dim=1, keepdim=True).expand(n, n)
dist = dist + dist.t()
dist.addmm_(0, 1, t1, t2.t())  
dist = dist.clamp(min=1e-12).sqrt()  # for numerical stability
print('dist:\t',dist)
targets = Variable(torch.FloatTensor([1,1,3,4]))
n = targets.size(0)
mask = targets.expand(n, n).eq(targets.expand(n, n).t())  # reture bool value
print('mask:\t',mask)
dist_ap, dist_an = [], []
for i in range(n):
    dist_ap.append(dist[i][mask[i]].max().view(1))
    dist_an.append(dist[i][mask[i] == 0].min().view(1))
dist_ap = torch.cat(dist_ap)
dist_an = torch.cat(dist_an)
print('dist_an:\t',dist_an)
# print('print:\t',dist[0][mask[0]].max().view(1))
y = dist_an.data.new()
print(y)
y.resize_as_(dist_an.data)
print(y)
y.fill_(1)
print(y)