-
Notifications
You must be signed in to change notification settings - Fork 0
/
loss.py
202 lines (162 loc) · 6.88 KB
/
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
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
import torch
from torch.autograd import Variable
import torch.nn as nn
import numpy as np
def cosine_sim(im, s):
"""Cosine similarity between all the image and sentence pairs
"""
return im.mm(s.t())
def order_sim(im, s):
"""Order embeddings similarity measure $max(0, s-im)$
"""
YmX = (s.unsqueeze(1).expand(s.size(0), im.size(0), s.size(1))
- im.unsqueeze(0).expand(s.size(0), im.size(0), s.size(1)))
score = -YmX.clamp(min=0).pow(2).sum(2).sqrt().t()
return score
def euclidean_sim(im, s):
"""Order embeddings similarity measure $max(0, s-im)$
"""
YmX = (s.unsqueeze(1).expand(s.size(0), im.size(0), s.size(1))
- im.unsqueeze(0).expand(s.size(0), im.size(0), s.size(1)))
score = -YmX.pow(2).sum(2).t()
return score
class TripletLoss(nn.Module):
"""
triplet ranking loss
"""
def __init__(self, margin=0, measure=False, max_violation=False, cost_style='sum', direction='all'):
super(TripletLoss, self).__init__()
self.margin = margin
self.cost_style = cost_style
self.direction = direction
if measure == 'order':
self.sim = order_sim
elif measure == 'euclidean':
self.sim = euclidean_sim
else:
self.sim = cosine_sim
self.max_violation = max_violation
def forward(self, s, im):
# compute image-sentence score matrix
scores = self.sim(im, s)
diagonal = scores.diag().view(im.size(0), 1)
d1 = diagonal.expand_as(scores)
d2 = diagonal.t().expand_as(scores)
# clear diagonals
mask = torch.eye(scores.size(0)) > .5
I = Variable(mask)
if torch.cuda.is_available():
I = I.cuda()
cost_s = None
cost_im = None
# compare every diagonal score to scores in its column
if self.direction in ['i2t', 'all']:
# caption retrieval
cost_s = (self.margin + scores - d1).clamp(min=0)
cost_s = cost_s.masked_fill_(I, 0)
# compare every diagonal score to scores in its row
if self.direction in ['t2i', 'all']:
# image retrieval
cost_im = (self.margin + scores - d2).clamp(min=0)
cost_im = cost_im.masked_fill_(I, 0)
# keep the maximum violating negative for each query
if self.max_violation:
if cost_s is not None:
cost_s = cost_s.max(1)[0]
if cost_im is not None:
cost_im = cost_im.max(0)[0]
if cost_s is None:
cost_s = Variable(torch.zeros(1)).cuda()
if cost_im is None:
cost_im = Variable(torch.zeros(1)).cuda()
if self.cost_style == 'sum':
return cost_s.sum() + cost_im.sum()
else:
return cost_s.mean() + cost_im.mean()
class likelihoodBCEloss(nn.Module):
"""
positive cross entropy
Math:
"""
def __init__(self,loss_lambda=0.1,cost_style='mean'):
super(likelihoodBCEloss, self).__init__()
self.loss_lambda = loss_lambda
self.cost_style = cost_style
def forward(self, outs,labels):
##compute the unlikehood loss
loss_lambda = self.loss_lambda
postive_mask = labels.float()
negative_mask = 1-postive_mask
BCEloss= -labels * torch.log(outs+1e-05) - (1 - labels) * torch.log(1 - outs+1e-05)
postiive_loss =BCEloss*postive_mask
negative_loss = BCEloss * negative_mask
invalid_sample_idx = torch.where(torch.sum(postive_mask,1)==0)[0].data.cpu().numpy()
postiive_loss_batch = torch.sum(postiive_loss,1)/torch.sum(postive_mask,1)
idx = list(range(0,len(postiive_loss_batch)))
for iidx in invalid_sample_idx:
del idx[iidx]
postiive_loss_batch_new = postiive_loss_batch[idx]
if torch.sum(torch.isnan(postiive_loss_batch_new).float())>0:
print('loss in nan')
loss=torch.tensor(0.0)
return loss
negative_loss_batch = torch.sum(negative_loss, 1) / torch.sum(negative_mask, 1)
negative_loss_batch_new = negative_loss_batch[idx]
likelyhoodloss = loss_lambda*postiive_loss_batch_new +(1-loss_lambda)*negative_loss_batch_new
if self.cost_style == 'sum':
loss = torch.sum(likelyhoodloss)
else:
loss = torch.mean(likelyhoodloss)
if np.isnan(loss.data.cpu().numpy()):
print('loss in nan')
return loss
class unlikelihoodBCEloss(nn.Module):
"""
unlikelihood loss to make constrains on contradicted concept pairs
Math:
"""
def __init__(self,contradicted_matrix=None,cost_style='mean'):
super(unlikelihoodBCEloss, self).__init__()
self.cost_style = cost_style
self.contradicted_matrix = contradicted_matrix.to_dense().float()
def forward(self, outs,labels):
postive_mask = labels.float()
##compute unlikehood loss
##compute the number of contradicted pairs in each sample
contradicted_matrix_mask = torch.sum(self.contradicted_matrix, 1) > 0
contradicted_matrix_mask = contradicted_matrix_mask.view(contradicted_matrix_mask.shape[0], 1)
##each training sample will have how many pairs of contradicted concepts to compute
contradicted_num_each_sample = torch.matmul(labels, contradicted_matrix_mask.float()).squeeze()
contradicted_num_positive = contradicted_num_each_sample > 0
##compute the UL loss
prob = torch.log(1 - outs + 1e-05) ##log(1-\hat{g}_t)
mask = 1-postive_mask ##(1-g_t)
prob_mask = prob*mask
unlikelyhoodloss = torch.matmul(prob_mask,torch.transpose(self.contradicted_matrix, 1, 0))
unlikelyhoodloss = -unlikelyhoodloss*postive_mask
##take the sum of all contradicted pairs
unlikelyhoodloss_sum = torch.sum(unlikelyhoodloss, 1)
unlikelyhoodloss_batch_nonzero = unlikelyhoodloss_sum[unlikelyhoodloss_sum>0]
dividor = torch.sum((unlikelyhoodloss>0).float(),1)[unlikelyhoodloss_sum>0]
unlikelyhoodloss_batch = unlikelyhoodloss_batch_nonzero/dividor
if self.cost_style == 'sum':
loss = torch.sum(unlikelyhoodloss_batch)
else:
loss = torch.mean(unlikelyhoodloss_batch)
if np.isnan(loss.data.cpu().numpy()):
print('loss in nan')
return loss
class normalLikelihoodBCEloss(nn.Module):
def __init__(self,cost_style='mean'):
super(normalLikelihoodBCEloss, self).__init__()
self.cost_style = cost_style
def forward(self, outs,labels):
BCEloss= -labels * torch.log(outs+1e-05) - (1 - labels) * torch.log(1 - outs+1e-05)
BCEloss_batch=torch.mean(BCEloss, 1)
if self.cost_style == 'sum':
loss = torch.sum(BCEloss_batch)
else:
loss = torch.mean(BCEloss_batch)
if np.isnan(loss.data.cpu().numpy()):
print('loss in nan')
return loss