You can not select more than 25 topics
Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
73 lines
2.6 KiB
73 lines
2.6 KiB
import logging
|
|
import math
|
|
import numpy as np
|
|
import torch
|
|
|
|
from .scheduler import Scheduler
|
|
|
|
|
|
logger = logging.getLogger(__name__)
|
|
|
|
|
|
class CosineLRScheduler(Scheduler):
|
|
"""
|
|
Cosine annealing with restarts.
|
|
This is described in the paper https://arxiv.org/abs/1608.03983.
|
|
"""
|
|
|
|
def __init__(self,
|
|
optimizer: torch.optim.Optimizer,
|
|
t_initial: int,
|
|
t_mul: float = 1.,
|
|
lr_min: float = 0.,
|
|
decay_rate: float = 1.,
|
|
warmup_updates=0,
|
|
warmup_lr_init=0,
|
|
initialize=True) -> None:
|
|
super().__init__(optimizer, param_group_field="lr", initialize=initialize)
|
|
|
|
assert t_initial > 0
|
|
assert lr_min >= 0
|
|
if t_initial == 1 and t_mul == 1 and decay_rate == 1:
|
|
logger.warning("Cosine annealing scheduler will have no effect on the learning "
|
|
"rate since t_initial = t_mul = eta_mul = 1.")
|
|
self.t_initial = t_initial
|
|
self.t_mul = t_mul
|
|
self.lr_min = lr_min
|
|
self.decay_rate = decay_rate
|
|
self.warmup_updates = warmup_updates
|
|
self.warmup_lr_init = warmup_lr_init
|
|
if self.warmup_updates:
|
|
self.warmup_steps = [(v - warmup_lr_init) / self.warmup_updates for v in self.base_values]
|
|
else:
|
|
self.warmup_steps = [1 for _ in self.base_values]
|
|
if self.warmup_lr_init:
|
|
super().update_groups(self.warmup_lr_init)
|
|
|
|
def get_epoch_values(self, epoch: int):
|
|
# this scheduler doesn't update on epoch
|
|
return None
|
|
|
|
def get_update_values(self, num_updates: int):
|
|
if num_updates < self.warmup_updates:
|
|
lrs = [self.warmup_lr_init + num_updates * s for s in self.warmup_steps]
|
|
else:
|
|
curr_updates = num_updates - self.warmup_updates
|
|
if self.t_mul != 1:
|
|
i = math.floor(math.log(1 - curr_updates / self.t_initial * (1 - self.t_mul), self.t_mul))
|
|
t_i = self.t_mul ** i * self.t_initial
|
|
t_curr = curr_updates - (1 - self.t_mul ** i) / (1 - self.t_mul) * self.t_initial
|
|
else:
|
|
i = curr_updates // self.t_initial
|
|
t_i = self.t_initial
|
|
t_curr = curr_updates - (self.t_initial * i)
|
|
|
|
gamma = self.decay_rate ** i
|
|
lr_min = self.lr_min * gamma
|
|
lr_max_values = [v * gamma for v in self.base_values]
|
|
|
|
lrs = [
|
|
lr_min + 0.5 * (lr_max - lr_min) * (1 + math.cos(math.pi * t_curr / t_i)) for lr_max in lr_max_values
|
|
]
|
|
return lrs
|