-
Notifications
You must be signed in to change notification settings - Fork 5
/
criterion.py
222 lines (190 loc) · 8.7 KB
/
criterion.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
import torch
from torch import nn
from torch.nn.modules.loss import _Loss
from sklearn.metrics import f1_score
import torch.nn.functional as F
class Criterion(_Loss):
def __init__(self, opt):
way = opt.ways
shot = opt.shots
super(Criterion, self).__init__()
self.amount = way * shot
def forward(self, probs, target,num_support=None): # (Q,C) (Q)
if num_support is None:
num_support = self.amount
target = target[num_support:]
target_onehot = torch.zeros_like(probs)
target_onehot = target_onehot.scatter(1, target.reshape(-1, 1), 1)
loss = torch.mean((probs - target_onehot) ** 2)
pred = torch.argmax(probs, dim=1)
assert target.shape[0] == pred.shape[0], "target len != pred len"
acc = torch.sum(target == pred).float() / target.shape[0]
f1 = f1_score(target.data.cpu().numpy(), pred.data.cpu().numpy(), average='macro')
return pred, loss, acc, f1
class CrossEntropyCriterion(_Loss):
def __init__(self, opt):
way = opt.ways
shot = opt.shots
super(CrossEntropyCriterion, self).__init__()
self.amount = way * shot
self.ce_loss = nn.CrossEntropyLoss()
def forward(self, probs, target,num_support=None): # (Q,C) (Q)
if num_support is None:
num_support = self.amount
target = target[num_support:]
loss = self.ce_loss(target,probs)
pred = torch.argmax(probs, dim=1)
assert target.shape[0] == pred.shape[0], "target len != pred len"
acc = torch.sum(target == pred).float() / target.shape[0]
f1 = f1_score(target.data.cpu().numpy(), pred.data.cpu().numpy(), average='macro')
return pred, loss, acc, f1
class TraditionCriterion(_Loss):
def __init__(self, opt):
super(TraditionCriterion, self).__init__()
self.amount = opt.batch_size
self.ce_loss = nn.CrossEntropyLoss()
def forward(self, probs, target): # (B,C) (B)
loss = self.ce_loss(probs, target)
pred = torch.argmax(probs, dim=1)
assert target.shape[0] == pred.shape[0], "target len != pred len"
acc = torch.sum(target == pred).float() / target.shape[0]
f1 = f1_score(target.data.cpu().numpy(), pred.data.cpu().numpy(), average='macro')
return loss
class CL_auxiliary(nn.Module):
"""Supervised Contrastive Learning: https://arxiv.org/pdf/2004.11362.pdf.
It also supports the unsupervised contrastive loss in SimCLR"""
def __init__(self, opt, contrast_mode='all',
temperature=0.07):
super(CL_auxiliary, self).__init__()
self.temperature = opt.temperatureP
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.
"""
features = features.unsqueeze(dim=1)
features = F.normalize(features, dim=2)
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().add(0.0000001).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))
# tile mask
mask = mask.repeat(anchor_count, contrast_count)
# mask-out self-contrast cases
logits_mask = torch.scatter(
torch.ones_like(mask),
1,
torch.arange(batch_size * anchor_count).view(-1, 1).to(device),
0
)
mask_pos = mask * logits_mask
mask_neg = (torch.ones_like(mask)-mask) * logits_mask
similarity = torch.exp(torch.mm(anchor_feature, contrast_feature.t()) / self.temperature)
pos = torch.sum(similarity * mask_pos, 1)
neg = torch.sum(similarity * mask_neg, 1)
loss = -(torch.mean(torch.log(pos / (pos + neg))))
return loss
class CL_sentiment(nn.Module):
"""Supervised Contrastive Learning: https://arxiv.org/pdf/2004.11362.pdf.
It also supports the unsupervised contrastive loss in SimCLR"""
def __init__(self, opt, contrast_mode='all',
temperature=0.07):
super(CL_sentiment, self).__init__()
self.temperature = opt.temperatureY
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.
"""
features = features.unsqueeze(dim=1)
features = F.normalize(features, dim=2)
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().add(0.0000001).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))
# tile mask
mask = mask.repeat(anchor_count, contrast_count)
# mask-out self-contrast cases
logits_mask = torch.scatter(
torch.ones_like(mask),
1,
torch.arange(batch_size * anchor_count).view(-1, 1).to(device),
0
)
mask_pos = mask * logits_mask
mask_neg = (torch.ones_like(mask)-mask) * logits_mask
similarity = torch.exp(torch.mm(anchor_feature, contrast_feature.t()) / self.temperature)
pos = torch.sum(similarity * mask_pos, 1)
neg = torch.sum(similarity * mask_neg, 1)
loss = -(torch.mean(torch.log(pos / (pos + neg))))
return loss