-
Notifications
You must be signed in to change notification settings - Fork 1
/
losses.py
288 lines (261 loc) · 9.52 KB
/
losses.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
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
from __future__ import print_function
import torch
import torch.nn as nn
class SupConLoss(nn.Module):
def __init__(self, temperature=0.01):
super(SupConLoss, self).__init__()
self.temperature = temperature
def forward(self, features, labels):
device = (torch.device('cuda')
if features.is_cuda
else torch.device('cpu'))
batch_size = features.shape[0]
labels = labels.contiguous().view(-1, 1)
##todo
# mask = torch.eq(labels, labels.T).float().to(device)
mask = torch.eq(labels, labels.transpose(1, 0)).float().to(device)
#
##todo
# anchor_dot_contrast = torch.div(
# torch.matmul(features, features.T),
# self.temperature)
anchor_dot_contrast = torch.div(
torch.matmul(features, features.transpose(1, 0)),
self.temperature)
##
# for numerical stability
logits_max, _ = torch.max(anchor_dot_contrast, dim=1, keepdim=True)
logits = anchor_dot_contrast - logits_max.detach()
logits_mask = torch.scatter(
torch.ones_like(mask),
1,
torch.arange(batch_size).view(-1, 1).to(device),
0
)
mask = mask * logits_mask
# compute log_prob
exp_logits = torch.exp(logits) * logits_mask
log_prob = logits - torch.log(exp_logits.sum(1, keepdim=True))
# compute mean of log-likelihood over positive
mean_log_prob_pos = (mask * log_prob).sum(1) / mask.sum(1)
# loss
loss = - mean_log_prob_pos
loss = loss.mean()
return loss
# clear those instances that have no positive instances to avoid training error
class SupConLoss_clear(nn.Module):
def __init__(self, temperature=0.07):
super(SupConLoss_clear, self).__init__()
self.temperature = temperature
def forward(self, features, labels):
device = (torch.device('cuda')
if features.is_cuda
else torch.device('cpu'))
batch_size = features.shape[0]
labels = labels.contiguous().view(-1, 1)
#todo
# mask = torch.eq(labels, labels.T).float().to(device)
mask = torch.eq(labels, labels.t()).float().to(device)
# mask = torch.eq(labels, labels.transpose(1, 0)).float().to(device)
#
##todo
# anchor_dot_contrast = torch.div(
# torch.matmul(features, features.T),
# self.temperature)
anchor_dot_contrast = torch.div(
torch.matmul(features, features.transpose(1, 0)),
self.temperature)
##
# normalize the logits for numerical stability
logits_max, _ = torch.max(anchor_dot_contrast, dim=1, keepdim=True)
logits = anchor_dot_contrast - logits_max.detach()
logits_mask = torch.scatter(
torch.ones_like(mask),
1,
torch.arange(batch_size).view(-1, 1).to(device),
0
)
mask = mask * logits_mask
single_samples = (mask.sum(1) == 0).float()
# compute log_prob
exp_logits = torch.exp(logits) * logits_mask
log_prob = logits - torch.log(exp_logits.sum(1, keepdim=True))
# compute mean of log-likelihood over positive
# invoid to devide the zero
mean_log_prob_pos = (mask * log_prob).sum(1) / (mask.sum(1)+single_samples)
# loss
# filter those single sample
loss = - mean_log_prob_pos*(1-single_samples)
loss = loss.sum()/(loss.shape[0]-single_samples.sum())
return loss
##inter-model v1
# class CrossConLoss(nn.Module):
# def __init__(self, temperature=0.07):
# super(CrossConLoss, self).__init__()
# self.temperature = temperature
#
# def forward(self, vis_features, sem_features, labels):
#
# device = (torch.device('cuda')
# if vis_features.is_cuda
# else torch.device('cpu'))
#
# batch_size = vis_features.shape[0]
# labels = labels.contiguous().view(-1, 1)
#
# #todo
# # mask = torch.eq(labels, labels.T).float().to(device)
# mask = torch.eq(labels, labels.t()).float().to(device)
# # mask = torch.eq(labels, labels.transpose(1, 0)).float().to(device)
# #
#
# ##todo
# # anchor_dot_contrast = torch.div(
# # torch.matmul(features, features.T),
# # self.temperature)
# anchor_dot_contrast = torch.div(
# torch.matmul(vis_features, sem_features.transpose(1, 0)),
# self.temperature)
# ##
#
# # normalize the logits for numerical stability
# logits_max, _ = torch.max(anchor_dot_contrast, dim=1, keepdim=True)
# logits = anchor_dot_contrast - logits_max.detach()
#
# logits_mask = torch.scatter(
# torch.ones_like(mask),
# 1,
# torch.arange(batch_size).view(-1, 1).to(device),
# 0
# )
# mask = mask * logits_mask
# single_samples = (mask.sum(1) == 0).float()
#
# # compute log_prob
# exp_logits = torch.exp(logits) * logits_mask
# log_prob = logits - torch.log(exp_logits.sum(1, keepdim=True))
#
# # compute mean of log-likelihood over positive
# # invoid to devide the zero
# mean_log_prob_pos = (mask * log_prob).sum(1) / (mask.sum(1)+single_samples)
#
# # loss
# # filter those single sample
# loss = - mean_log_prob_pos*(1-single_samples)
# loss = loss.sum()/(loss.shape[0]-single_samples.sum())
#
# return loss
##inter-model v2
# class CrossConLoss(nn.Module):
# def __init__(self, temperature=0.07):
# super(CrossConLoss, self).__init__()
# self.temperature = temperature
#
# def forward(self, sem_features, vis_features, labels):
#
# device = (torch.device('cuda')
# if vis_features.is_cuda
# else torch.device('cpu'))
#
# batch_size = vis_features.shape[0]
# labels = labels.contiguous().view(-1, 1)
#
# #todo
# # mask = torch.eq(labels, labels.T).float().to(device)
# mask = torch.eq(labels, labels.t()).float().to(device)
# # mask = torch.eq(labels, labels.transpose(1, 0)).float().to(device)
# #
#
# ##todo
# # anchor_dot_contrast = torch.div(
# # torch.matmul(features, features.T),
# # self.temperature)
# anchor_dot_contrast = torch.div(
# torch.matmul(sem_features, vis_features.transpose(1, 0)),
# self.temperature)
# ##
#
# # normalize the logits for numerical stability
# logits_max, _ = torch.max(anchor_dot_contrast, dim=1, keepdim=True)
# logits = anchor_dot_contrast - logits_max.detach()
#
# logits_mask = torch.scatter(
# torch.ones_like(mask),
# 1,
# torch.arange(batch_size).view(-1, 1).to(device),
# 0
# )
# mask = mask * logits_mask
# single_samples = (mask.sum(1) == 0).float()
#
# # compute log_prob
# exp_logits = torch.exp(logits) * logits_mask
# log_prob = logits - torch.log(exp_logits.sum(1, keepdim=True))
#
# # compute mean of log-likelihood over positive
# # invoid to devide the zero
# mean_log_prob_pos = (mask * log_prob).sum(1) / (mask.sum(1)+single_samples)
#
# # loss
# # filter those single sample
# loss = - mean_log_prob_pos*(1-single_samples)
# loss = loss.sum()/(loss.shape[0]-single_samples.sum())
#
# return loss
# class CrossConLoss(nn.Module):
# def __init__(self, temperature=0.07):
# super(CrossConLoss, self).__init__()
# self.temperature = temperature
#
# def forward(self, vis_features, sem_features, labels):
#
# device = (torch.device('cuda')
# if vis_features.is_cuda
# else torch.device('cpu'))
#
# batch_size = vis_features.shape[0]
# labels = labels.contiguous().view(-1, 1)
#
# mask = torch.eq(labels, labels.t()).float().to(device)
#
# anchor_dot_contrast = torch.div(
# torch.matmul(vis_features, sem_features.transpose(1, 0)),
# self.temperature)
#
# # normalize the logits for numerical stability
# logits_max, _ = torch.max(anchor_dot_contrast, dim=1, keepdim=True)
# logits = anchor_dot_contrast - logits_max.detach()
#
# # logits_mask = torch.scatter(
# # torch.ones_like(mask),
# # 1,
# # torch.arange(batch_size).view(-1, 1).to(device),
# # 0
# # )
# # mask = mask * logits_mask
# # single_samples = (mask.sum(1) == 0).float()
#
# # compute log_prob
# ##todo
# # exp_logits = torch.exp(logits) * logits_mask
# exp_logits = torch.exp(logits)
# ##
# log_prob = logits - torch.log(exp_logits.sum(1, keepdim=True))
#
# # compute mean of log-likelihood over positive
# # invoid to devide the zero
# ##todo
# # mean_log_prob_pos = (mask * log_prob).sum(1) / (mask.sum(1)+single_samples)
# mean_log_prob_pos = (mask * log_prob).sum(1) / mask.sum(1)
# ##
#
# # loss
# # filter those single sample
# ##todo
# # loss = - mean_log_prob_pos*(1-single_samples)
# # loss = loss.sum()/(loss.shape[0]-single_samples.sum())
# loss = - mean_log_prob_pos
# loss = loss.sum() / loss.shape[0]
# ##
#
# return loss