-
Notifications
You must be signed in to change notification settings - Fork 2
/
architect.py
128 lines (108 loc) · 5.33 KB
/
architect.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
import torch
import logging
import numpy as np
import torch.nn.functional as F
from torch.autograd import Variable
from search_config import args
from utils import cal_aux_loss
def _concat(xs):
return torch.cat([x.view(-1) for x in xs])
class Architect(object):
def __init__(self, model, args):
self.network_momentum = args.momentum
self.network_weight_decay = args.weight_decay
self.model = model
self.optimizer = torch.optim.Adam(self.model.arch_parameters(),
lr=args.arch_learning_rate, betas=(0.5, 0.999),
weight_decay=args.arch_weight_decay)
def _compute_unrolled_model(self, input, target, eta, network_optimizer):
loss = self.model._loss(input, target)
theta = _concat(self.model.parameters()).data
try:
moment = _concat(network_optimizer.state[v]['momentum_buffer'] for v in self.model.parameters()).mul_(
self.network_momentum)
except:
moment = torch.zeros_like(theta)
dtheta = _concat(torch.autograd.grad(loss, self.model.parameters())).data + self.network_weight_decay * theta
unrolled_model = self._construct_model_from_theta(theta.sub(eta, moment + dtheta))
return unrolled_model
def step(self, input_train, target_train, input_valid, target_valid, eta, network_optimizer, unrolled):
self.optimizer.zero_grad()
if unrolled:
self._backward_step_unrolled(input_train, target_train, input_valid, target_valid, eta, network_optimizer)
else:
self._backward_step(input_valid, target_valid)
self.optimizer.step()
def _backward_step(self, input_valid, target_valid):
if args.aux_loss:
gamma = 1.
aux_loss = cal_aux_loss(
F.softmax(self.model.alphas_normal, dim=-1),
F.softmax(self.model.alphas_reduce, dim=-1),
F.sigmoid(self.model.betas_normal),
F.sigmoid(self.model.betas_reduce),
)
if args.info_linear_grow:
gamma = (args.epoch + 1) / args.epochs
gamma_aux_loss = gamma * aux_loss
loss_arch = self.model._loss(input_valid, target_valid)
loss = (loss_arch * args.lamda0) + (gamma_aux_loss * args.lamda1)
if args.step % args.report_freq == 0:
logging.info(
'step %f gamma %f info_entropy %f gamma_aux_loss %f loss_arch %f loss %f ' %
(args.step, gamma, aux_loss.cpu().detach().numpy(), gamma_aux_loss.cpu().detach().numpy(),
loss_arch.item(), loss.item())
)
elif args.beta_loss:
num_edges = 14
loss_arch = self.model._loss(input_valid, target_valid)
loss_beta = torch.sum(1/F.sigmoid(self.model.betas_normal))/num_edges + \
torch.sum(1/F.sigmoid(self.model.betas_reduce))/num_edges
loss = loss_arch + loss_beta
if args.step % args.report_freq == 0:
logging.info(
'step %f loss_arch %f loss_beta %f loss %f ' %
(args.step, loss_arch.item(), loss_beta.item(), loss.item())
)
else:
loss = self.model._loss(input_valid, target_valid)
loss.backward()
def _backward_step_unrolled(self, input_train, target_train, input_valid, target_valid, eta, network_optimizer):
unrolled_model = self._compute_unrolled_model(input_train, target_train, eta, network_optimizer)
unrolled_loss = unrolled_model._loss(input_valid, target_valid)
unrolled_loss.backward()
dalpha = [v.grad for v in unrolled_model.arch_parameters()]
vector = [v.grad.data for v in unrolled_model.parameters()]
implicit_grads = self._hessian_vector_product(vector, input_train, target_train)
for g, ig in zip(dalpha, implicit_grads):
g.data.sub_(eta, ig.data)
for v, g in zip(self.model.arch_parameters(), dalpha):
if v.grad is None:
v.grad = Variable(g.data)
else:
v.grad.data.copy_(g.data)
def _construct_model_from_theta(self, theta):
model_new = self.model.new()
model_dict = self.model.state_dict()
params, offset = {}, 0
for k, v in self.model.named_parameters():
v_length = np.prod(v.size())
params[k] = theta[offset: offset + v_length].view(v.size())
offset += v_length
assert offset == len(theta)
model_dict.update(params)
model_new.load_state_dict(model_dict)
return model_new.cuda()
def _hessian_vector_product(self, vector, input, target, r=1e-2):
R = r / _concat(vector).norm()
for p, v in zip(self.model.parameters(), vector):
p.data.add_(R, v)
loss = self.model._loss(input, target)
grads_p = torch.autograd.grad(loss, self.model.arch_parameters())
for p, v in zip(self.model.parameters(), vector):
p.data.sub_(2 * R, v)
loss = self.model._loss(input, target)
grads_n = torch.autograd.grad(loss, self.model.arch_parameters())
for p, v in zip(self.model.parameters(), vector):
p.data.add_(R, v)
return [(x - y).div_(2 * R) for x, y in zip(grads_p, grads_n)]