forked from ferretj/temporal-ensembling
-
Notifications
You must be signed in to change notification settings - Fork 0
/
temporal_ensembling.py
175 lines (139 loc) · 6.51 KB
/
temporal_ensembling.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
import numpy as np
from timeit import default_timer as timer
import torch
from torch.autograd import Variable
import torch.nn as nn
import torch.nn.functional as F
from utils import calc_metrics, prepare_mnist, weight_schedule
def sample_train(train_dataset, test_dataset, batch_size, k, n_classes,
seed, shuffle_train=True, return_idxs=True):
n = len(train_dataset)
rrng = np.random.RandomState(seed)
cpt = 0
indices = torch.zeros(k)
other = torch.zeros(n - k)
card = k // n_classes
for i in xrange(n_classes):
class_items = (train_dataset.train_labels == i).nonzero()
n_class = len(class_items)
rd = np.random.permutation(np.arange(n_class))
indices[i * card: (i + 1) * card] = class_items[rd[:card]]
other[cpt: cpt + n_class - card] = class_items[rd[card:]]
cpt += n_class - card
other = other.long()
train_dataset.train_labels[other] = -1
train_loader = torch.utils.data.DataLoader(dataset=train_dataset,
batch_size=batch_size,
num_workers=4,
shuffle=shuffle_train)
test_loader = torch.utils.data.DataLoader(dataset=test_dataset,
batch_size=batch_size,
num_workers=4,
shuffle=False)
if return_idxs:
return train_loader, test_loader, indices
return train_loader, test_loader
def temporal_loss(out1, out2, w, labels):
# MSE between current and temporal outputs
def mse_loss(out1, out2):
quad_diff = torch.sum((F.softmax(out1, dim=1) - F.softmax(out2, dim=1)) ** 2)
return quad_diff / out1.data.nelement()
def masked_crossentropy(out, labels):
cond = (labels >= 0)
nnz = torch.nonzero(cond)
nbsup = len(nnz)
# check if labeled samples in batch, return 0 if none
if nbsup > 0:
masked_outputs = torch.index_select(out, 0, nnz.view(nbsup))
masked_labels = labels[cond]
loss = F.cross_entropy(masked_outputs, masked_labels)
return loss, nbsup
return Variable(torch.FloatTensor([0.]).cuda(), requires_grad=False), 0
sup_loss, nbsup = masked_crossentropy(out1, labels)
unsup_loss = mse_loss(out1, out2)
return sup_loss + w * unsup_loss, sup_loss, unsup_loss, nbsup
def train(model, seed, k=100, alpha=0.6, lr=0.002, beta2=0.99, num_epochs=150,
batch_size=100, drop=0.5, std=0.15, fm1=16, fm2=32,
divide_by_bs=False, w_norm=False, data_norm='pixelwise',
early_stop=None, c=300, n_classes=10, max_epochs=80,
max_val=30., ramp_up_mult=-5., n_samples=60000,
print_res=True, **kwargs):
# retrieve data
train_dataset, test_dataset = prepare_mnist()
ntrain = len(train_dataset)
# build model
model.cuda()
# make data loaders
train_loader, test_loader, indices = sample_train(train_dataset, test_dataset, batch_size,
k, n_classes, seed, shuffle_train=False)
# setup param optimization
optimizer = torch.optim.Adam(model.parameters(), lr=lr, betas=(0.9, 0.99))
# train
model.train()
losses = []
sup_losses = []
unsup_losses = []
best_loss = 20.
Z = torch.zeros(ntrain, n_classes).float().cuda() # intermediate values
z = torch.zeros(ntrain, n_classes).float().cuda() # temporal outputs
outputs = torch.zeros(ntrain, n_classes).float().cuda() # current outputs
for epoch in range(num_epochs):
t = timer()
# evaluate unsupervised cost weight
w = weight_schedule(epoch, max_epochs, max_val, ramp_up_mult, k, n_samples)
if (epoch + 1) % 10 == 0:
print 'unsupervised loss weight : {}'.format(w)
# turn it into a usable pytorch object
w = torch.autograd.Variable(torch.FloatTensor([w]).cuda(), requires_grad=False)
l = []
supl = []
unsupl = []
for i, (images, labels) in enumerate(train_loader):
images = Variable(images.cuda())
labels = Variable(labels.cuda(), requires_grad=False)
# get output and calculate loss
optimizer.zero_grad()
out = model(images)
zcomp = Variable(z[i * batch_size: (i + 1) * batch_size], requires_grad=False)
loss, suploss, unsuploss, nbsup = temporal_loss(out, zcomp, w, labels)
# save outputs and losses
outputs[i * batch_size: (i + 1) * batch_size] = out.data.clone()
l.append(loss.data[0])
supl.append(nbsup * suploss.data[0])
unsupl.append(unsuploss.data[0])
# backprop
loss.backward()
optimizer.step()
# print loss
if (epoch + 1) % 10 == 0:
if i + 1 == 2 * c:
print ('Epoch [%d/%d], Step [%d/%d], Loss: %.6f, Time (this epoch): %.2f s'
%(epoch + 1, num_epochs, i + 1, len(train_dataset) // batch_size, np.mean(l), timer() - t))
elif (i + 1) % c == 0:
print ('Epoch [%d/%d], Step [%d/%d], Loss: %.6f'
%(epoch + 1, num_epochs, i + 1, len(train_dataset) // batch_size, np.mean(l)))
# update temporal ensemble
Z = alpha * Z + (1. - alpha) * outputs
z = Z * (1. / (1. - alpha ** (epoch + 1)))
# handle metrics, losses, etc.
eloss = np.mean(l)
losses.append(eloss)
sup_losses.append((1. / k) * np.sum(supl)) # division by 1/k to obtain the mean supervised loss
unsup_losses.append(np.mean(unsupl))
# saving model
if eloss < best_loss:
best_loss = eloss
torch.save({'state_dict': model.state_dict()}, 'model_best.pth.tar')
# test
model.eval()
acc = calc_metrics(model, test_loader)
if print_res:
print 'Accuracy of the network on the 10000 test images: %.2f %%' % (acc)
# test best model
checkpoint = torch.load('model_best.pth.tar')
model.load_state_dict(checkpoint['state_dict'])
model.eval()
acc_best = calc_metrics(model, test_loader)
if print_res:
print 'Accuracy of the network (best model) on the 10000 test images: %.2f %%' % (acc_best)
return acc, acc_best, losses, sup_losses, unsup_losses, indices