-
Notifications
You must be signed in to change notification settings - Fork 4
/
Copy pathoptimizers.py
272 lines (229 loc) · 11.6 KB
/
optimizers.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
"""
Created on 2022/01/05
@author Sangwoo Han
@ref https://github.com/yourh/AttentionXML/blob/master/deepxml/optimizers.py
"""
import math
import torch
from torch.optim.optimizer import Optimizer
class DenseSparseAdam(Optimizer):
def __init__(self, params, lr=1e-3, betas=(0.9, 0.999), eps=1e-8, weight_decay=0.0):
if not 0.0 <= lr:
raise ValueError("Invalid learning rate: {}".format(lr))
if not 0.0 <= eps:
raise ValueError("Invalid epsilon value: {}".format(eps))
if not 0.0 <= betas[0] < 1.0:
raise ValueError("Invalid beta parameter at index 0: {}".format(betas[0]))
if not 0.0 <= betas[1] < 1.0:
raise ValueError("Invalid beta parameter at index 1: {}".format(betas[1]))
defaults = dict(lr=lr, betas=betas, eps=eps, weight_decay=weight_decay)
super(DenseSparseAdam, self).__init__(params, defaults)
def step(self, closure=None):
"""
Performs a single optimization step.
Parameters
----------
closure : ``callable``, optional.
A closure that reevaluates the model and returns the loss.
"""
loss = None
if closure is not None:
loss = closure()
for group in self.param_groups:
for p in group["params"]:
if p.grad is None:
continue
grad = p.grad.data
state = self.state[p]
# State initialization
if "step" not in state:
state["step"] = 0
if "exp_avg" not in state:
# Exponential moving average of gradient values
state["exp_avg"] = torch.zeros_like(p.data)
if "exp_avg_sq" not in state:
# Exponential moving average of squared gradient values
state["exp_avg_sq"] = torch.zeros_like(p.data)
state["step"] += 1
exp_avg, exp_avg_sq = state["exp_avg"], state["exp_avg_sq"]
beta1, beta2 = group["betas"]
weight_decay = group["weight_decay"]
if grad.is_sparse:
grad = (
grad.coalesce()
) # the update is non-linear so indices must be unique
grad_indices = grad._indices()
grad_values = grad._values()
size = grad.size()
def make_sparse(values):
constructor = grad.new
if grad_indices.dim() == 0 or values.dim() == 0:
return constructor().resize_as_(grad)
return constructor(grad_indices, values, size)
# Decay the first and second moment running average coefficient
# old <- b * old + (1 - b) * new
# <==> old += (1 - b) * (new - old)
old_exp_avg_values = exp_avg.sparse_mask(grad)._values()
exp_avg_update_values = grad_values.sub(old_exp_avg_values).mul_(
1 - beta1
)
exp_avg.add_(make_sparse(exp_avg_update_values))
old_exp_avg_sq_values = exp_avg_sq.sparse_mask(grad)._values()
exp_avg_sq_update_values = (
grad_values.pow(2).sub_(old_exp_avg_sq_values).mul_(1 - beta2)
)
exp_avg_sq.add_(make_sparse(exp_avg_sq_update_values))
# Dense addition again is intended, avoiding another sparse_mask
numer = exp_avg_update_values.add_(old_exp_avg_values)
exp_avg_sq_update_values.add_(old_exp_avg_sq_values)
denom = exp_avg_sq_update_values.sqrt_().add_(group["eps"])
del exp_avg_update_values, exp_avg_sq_update_values
bias_correction1 = 1 - beta1 ** state["step"]
bias_correction2 = 1 - beta2 ** state["step"]
step_size = (
group["lr"] * math.sqrt(bias_correction2) / bias_correction1
)
p.data.add_(make_sparse(-step_size * numer.div_(denom)))
if weight_decay > 0.0:
p.data.add_(
p.data.sparse_mask(grad), alpha=-group["lr"] * weight_decay
)
else:
# Decay the first and second moment running average coefficient
exp_avg.mul_(beta1).add_(grad, alpha=1 - beta1)
exp_avg_sq.mul_(beta2).addcmul_(grad, grad, value=1 - beta2)
denom = exp_avg_sq.sqrt().add_(group["eps"])
bias_correction1 = 1 - beta1 ** state["step"]
bias_correction2 = 1 - beta2 ** state["step"]
step_size = (
group["lr"] * math.sqrt(bias_correction2) / bias_correction1
)
p.data.addcdiv_(exp_avg, denom, value=-step_size)
if weight_decay > 0.0:
p.data.add_(p.data, alpha=-group["lr"] * weight_decay)
return loss
class DenseSparseAdamW(Optimizer):
r"""Implements lazy version of AdamW algorithm suitable for sparse tensors.
In this variant, only moments that show up in the gradient get updated, and
only those portions of the gradient get applied to the parameters.
The original Adam algorithm was proposed in `Adam: A Method for Stochastic Optimization`_.
The AdamW variant was proposed in `Decoupled Weight Decay Regularization`_.
Arguments:
params (iterable): iterable of parameters to optimize or dicts defining
parameter groups
lr (float, optional): learning rate (default: 1e-3)
betas (Tuple[float, float], optional): coefficients used for computing
running averages of gradient and its square (default: (0.9, 0.999))
eps (float, optional): term added to the denominator to improve
numerical stability (default: 1e-8)
weight_decay (float, optional): weight decay coefficient (default: 1e-2)
.. _Adam\: A Method for Stochastic Optimization:
https://arxiv.org/abs/1412.6980
.. _Decoupled Weight Decay Regularization:
https://arxiv.org/abs/1711.05101
.. _On the Convergence of Adam and Beyond:
https://openreview.net/forum?id=ryQu7f-RZ
"""
def __init__(
self, params, lr=1e-3, betas=(0.9, 0.999), eps=1e-8, weight_decay=1e-2
):
if not 0.0 <= lr:
raise ValueError("Invalid learning rate: {}".format(lr))
if not 0.0 <= eps:
raise ValueError("Invalid epsilon value: {}".format(eps))
if not 0.0 <= betas[0] < 1.0:
raise ValueError("Invalid beta parameter at index 0: {}".format(betas[0]))
if not 0.0 <= betas[1] < 1.0:
raise ValueError("Invalid beta parameter at index 1: {}".format(betas[1]))
if not 0.0 <= weight_decay:
raise ValueError("Invalid weight_decay value: {}".format(weight_decay))
defaults = dict(lr=lr, betas=betas, eps=eps, weight_decay=weight_decay)
super(DenseSparseAdamW, self).__init__(params, defaults)
def __setstate__(self, state):
super(DenseSparseAdamW, self).__setstate__(state)
@torch.no_grad()
def step(self, closure=None):
"""Performs a single optimization step.
Arguments:
closure (callable, optional): A closure that reevaluates the model
and returns the loss.
"""
loss = None
if closure is not None:
with torch.enable_grad():
loss = closure()
for group in self.param_groups:
for p in group["params"]:
if p.grad is None:
continue
grad = p.grad
state = self.state[p]
# State initialization
if len(state) == 0:
state["step"] = 0
# Exponential moving average of gradient values
state["exp_avg"] = torch.zeros_like(
p, memory_format=torch.preserve_format
)
# Exponential moving average of squared gradient values
state["exp_avg_sq"] = torch.zeros_like(
p, memory_format=torch.preserve_format
)
state["step"] += 1
exp_avg, exp_avg_sq = state["exp_avg"], state["exp_avg_sq"]
beta1, beta2 = group["betas"]
bias_correction1 = 1 - beta1 ** state["step"]
bias_correction2 = 1 - beta2 ** state["step"]
if grad.is_sparse:
grad = (
grad.coalesce()
) # the update is non-linear so indices must be unique
grad_indices = grad._indices()
grad_values = grad._values()
size = grad.size()
def make_sparse(values):
constructor = grad.new
if grad_indices.dim() == 0 or values.dim() == 0:
return constructor().resize_as_(grad)
return constructor(grad_indices, values, size)
# Perform stepweight decay
p.sub_(
make_sparse(p.sparse_mask(grad)._values()),
alpha=group["lr"] * group["weight_decay"],
)
# Decay the first and second moment running average coefficient
# old <- b * old + (1 - b) * new
# <==> old += (1 - b) * (new - old)
old_exp_avg_values = exp_avg.sparse_mask(grad)._values()
exp_avg_update_values = grad_values.sub(old_exp_avg_values).mul_(
1 - beta1
)
exp_avg.add_(make_sparse(exp_avg_update_values))
old_exp_avg_sq_values = exp_avg_sq.sparse_mask(grad)._values()
exp_avg_sq_update_values = (
grad_values.pow(2).sub_(old_exp_avg_sq_values).mul_(1 - beta2)
)
exp_avg_sq.add_(make_sparse(exp_avg_sq_update_values))
# Dense addition again is intended, avoiding another sparse_mask
numer = exp_avg_update_values.add_(old_exp_avg_values)
exp_avg_sq_update_values.add_(old_exp_avg_sq_values)
denom = (
exp_avg_sq_update_values.sqrt_()
.div_(math.sqrt(bias_correction2))
.add_(group["eps"])
)
del exp_avg_update_values, exp_avg_sq_update_values
step_size = group["lr"] / bias_correction1
p.add_(make_sparse(-step_size * numer.div_(denom)))
else:
# Perform stepweight decay
p.mul_(1 - group["lr"] * group["weight_decay"])
# Decay the first and second moment running average coefficient
exp_avg.mul_(beta1).add_(grad, alpha=1 - beta1)
exp_avg_sq.mul_(beta2).addcmul_(grad, grad, value=1 - beta2)
denom = (exp_avg_sq.sqrt() / math.sqrt(bias_correction2)).add_(
group["eps"]
)
step_size = group["lr"] / bias_correction1
p.addcdiv_(exp_avg, denom, value=-step_size)
return loss