/
losses.py
254 lines (210 loc) · 9.44 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
import torch
import torch.nn as nn
class SupConLoss(nn.Module):
"""Supervised Contrastive Learning: https://arxiv.org/pdf/2004.11362.pdf.
It also supports the unsupervised contrastive loss in SimCLR"""
def __init__(self, temperature=0.07, contrast_mode='all',
base_temperature=0.07):
super(SupConLoss, self).__init__()
self.temperature = temperature
self.contrast_mode = contrast_mode
self.base_temperature = base_temperature
def forward(self, features, labels=None, mask=None):
"""Compute loss for model. If both `labels` and `mask` are None,
it degenerates to SimCLR unsupervised loss:
https://arxiv.org/pdf/2002.05709.pdf
Args:
features: hidden vector of shape [bsz, n_views, ...].
labels: ground truth of shape [bsz].
mask: contrastive mask of shape [bsz, bsz], mask_{i,j}=1 if sample j
has the same class as sample i. Can be asymmetric.
Returns:
A loss scalar.
"""
device = (torch.device('cuda')
if features.is_cuda
else torch.device('cpu'))
if len(features.shape) < 3:
raise ValueError('`features` needs to be [bsz, n_views, ...],'
'at least 3 dimensions are required')
if len(features.shape) > 3:
features = features.view(features.shape[0], features.shape[1], -1)
batch_size = features.shape[0]
if labels is not None and mask is not None:
raise ValueError('Cannot define both `labels` and `mask`')
elif labels is None and mask is None:
mask = torch.eye(batch_size, dtype=torch.float32).to(device)
elif labels is not None:
labels = labels.contiguous().view(-1, 1)
if labels.shape[0] != batch_size:
raise ValueError('Num of labels does not match num of features')
mask = torch.eq(labels, labels.T).float().to(device)
else:
mask = mask.float().to(device)
contrast_count = features.shape[1]
contrast_feature = torch.cat(torch.unbind(features, dim=1), dim=0)
if self.contrast_mode == 'one':
anchor_feature = features[:, 0]
anchor_count = 1
elif self.contrast_mode == 'all':
anchor_feature = contrast_feature
anchor_count = contrast_count
else:
raise ValueError('Unknown mode: {}'.format(self.contrast_mode))
anchor_dot_contrast = torch.div(
torch.matmul(anchor_feature, contrast_feature.T),
self.temperature)
logits_max, _ = torch.max(anchor_dot_contrast, dim=1, keepdim=True)
logits = anchor_dot_contrast - logits_max.detach()
mask = mask.repeat(anchor_count, contrast_count)
logits_mask = torch.scatter(
torch.ones_like(mask),
1,
torch.arange(batch_size * anchor_count).view(-1, 1).to(device),
0
)
mask = mask * logits_mask
exp_logits = torch.exp(logits) * logits_mask
log_prob = logits - torch.log(exp_logits.sum(1, keepdim=True))
mean_log_prob_pos = (mask * log_prob).sum(1) / mask.sum(1)
loss = - (self.temperature / self.base_temperature) * mean_log_prob_pos
loss = loss.view(anchor_count, batch_size).mean()
return loss
class SimCLRLoss(nn.Module):
def __init__(self, temperature=0.07, contrast_mode='all',
base_temperature=0.07):
super(SimCLRLoss, self).__init__()
self.temperature = temperature
self.contrast_mode = contrast_mode
self.base_temperature = base_temperature
def forward(self, features, labels=None, mask=None):
device = (torch.device('cuda')
if features.is_cuda
else torch.device('cpu'))
if len(features.shape) < 3:
raise ValueError('`features` needs to be [bsz, n_views, ...],'
'at least 3 dimensions are required')
if len(features.shape) > 3:
features = features.view(features.shape[0], features.shape[1], -1)
batch_size = features.shape[0]
if labels is not None and mask is not None:
raise ValueError('Cannot define both `labels` and `mask`')
elif labels is None and mask is None:
mask = torch.eye(batch_size, dtype=torch.float32).to(device)
elif labels is not None:
labels = labels.contiguous().view(-1, 1)
if labels.shape[0] != batch_size:
raise ValueError('Num of labels does not match num of features')
mask = torch.eq(labels, labels.T).float().to(device)
else:
mask = mask.float().to(device)
contrast_count = features.shape[1]
contrast_feature = torch.cat(torch.unbind(features, dim=1), dim=0)
if self.contrast_mode == 'one':
anchor_feature = features[:, 0]
anchor_count = 1
elif self.contrast_mode == 'all':
anchor_feature = contrast_feature
anchor_count = contrast_count
else:
raise ValueError('Unknown mode: {}'.format(self.contrast_mode))
anchor_dot_contrast = torch.div(
torch.matmul(anchor_feature, contrast_feature.T),
self.temperature)
logits_max, _ = torch.max(anchor_dot_contrast, dim=1, keepdim=True)
logits = anchor_dot_contrast - logits_max.detach()
mask = mask.repeat(anchor_count, contrast_count)
logits_mask = torch.scatter(
torch.ones_like(mask),
1,
torch.arange(batch_size * anchor_count).view(-1, 1).to(device),
0
)
mask = mask * logits_mask
exp_logits = torch.exp(logits) * logits_mask
log_prob = logits - torch.log(exp_logits.sum(1, keepdim=True))
mean_log_prob_pos = (mask * log_prob).sum(1) / mask.sum(1)
loss = - (self.temperature / self.base_temperature) * mean_log_prob_pos
loss = loss.view(anchor_count, batch_size).mean()
return loss
class MoCoLoss(nn.Module):
def __init__(self, temperature=0.07, base_temperature=0.07):
super(MoCoLoss, self).__init__()
self.temperature = temperature
self.base_temperature = base_temperature
def forward(self, logits, labels=None, queue_labels=None):
"""
logits: Nx(1+K)
labels: N,
queue_labels: K,
"""
device = (torch.device('cuda')
if logits.is_cuda
else torch.device('cpu'))
# CL loss
bsz = logits.shape[0]
if labels is None and queue_labels is None:
mask = torch.zeros_like(logits)
mask[:, 0] = 1.
else:
labels = labels.contiguous().view(-1, 1)
queue_labels = queue_labels.contiguous().view(-1, 1)
mask = torch.eq(labels, queue_labels.T).float().to(device) # NxK
mask = torch.cat([torch.ones(bsz, 1).to(device), mask], dim=1) # Nx(K+1)
logits /= self.temperature
logits_max, _ = torch.max(logits, dim=1, keepdim=True)
logits = logits - logits_max.detach()
exp_logits = torch.exp(logits)
log_prob = logits - torch.log(exp_logits.sum(1, keepdim=True))
mean_log_prob_pos = (mask * log_prob).sum(1) / mask.sum(1)
loss = - (self.temperature / self.base_temperature) * mean_log_prob_pos
loss = loss.mean()
return loss
class SymNegCosineSimilarityLoss(torch.nn.Module):
def __init__(self) -> None:
super().__init__()
def _neg_cosine_simililarity(self, x, y):
v = -torch.nn.functional.cosine_similarity(x, y.detach(), dim=-1).mean()
return v
def forward(self, out0: torch.Tensor, out1: torch.Tensor):
"""Forward pass through Symmetric Loss.
Args:
out0:
Output projections of the first set of transformed images.
Expects the tuple to be of the form (z0, p0), where z0 is
the output of the backbone and projection mlp, and p0 is the
output of the prediction head.
out1:
Output projections of the second set of transformed images.
Expects the tuple to be of the form (z1, p1), where z1 is
the output of the backbone and projection mlp, and p1 is the
output of the prediction head.
Returns:
Contrastive Cross Entropy Loss value.
Raises:
ValueError if shape of output is not multiple of batch_size.
"""
z0, p0 = out0
z1, p1 = out1
loss = (
self._neg_cosine_simililarity(p0, z1) / 2
+ self._neg_cosine_simililarity(p1, z0) / 2
)
return loss
class SimSiamLoss(nn.Module):
def __init__(self, version='simplified'):
super().__init__()
self.ver = version
def asymmetric_loss(self, p, z):
if self.ver == 'original':
z = z.detach() # stop gradient
p = nn.functional.normalize(p, dim=1)
z = nn.functional.normalize(z, dim=1)
return -(p * z).sum(dim=1).mean()
elif self.ver == 'simplified':
z = z.detach() # stop gradient
return - nn.functional.cosine_similarity(p, z, dim=-1).mean()
def forward(self, z1, z2, p1, p2):
loss1 = self.asymmetric_loss(p1, z2)
loss2 = self.asymmetric_loss(p2, z1)
return 0.5 * loss1 + 0.5 * loss2