100 lines
3.6 KiB
Python
100 lines
3.6 KiB
Python
'''
|
|
@ Date: 2021-09-05 20:07:55
|
|
@ Author: Qing Shuai
|
|
@ LastEditors: Qing Shuai
|
|
@ LastEditTime: 2021-09-05 20:10:02
|
|
@ FilePath: /EasyMocap/easymocap/neuralbody/trainer/lr_sheduler.py
|
|
'''
|
|
import torch
|
|
from collections import Counter
|
|
from bisect import bisect_right
|
|
|
|
class WarmupMultiStepLR(torch.optim.lr_scheduler._LRScheduler):
|
|
def __init__(
|
|
self,
|
|
optimizer,
|
|
milestones,
|
|
gamma=0.1,
|
|
warmup_factor=1.0 / 3,
|
|
warmup_iters=5,
|
|
warmup_method="linear",
|
|
last_epoch=-1,
|
|
):
|
|
if not list(milestones) == sorted(milestones):
|
|
raise ValueError(
|
|
"Milestones should be a list of" " increasing integers. Got {}",
|
|
milestones,
|
|
)
|
|
|
|
if warmup_method not in ("constant", "linear"):
|
|
raise ValueError(
|
|
"Only 'constant' or 'linear' warmup_method accepted"
|
|
"got {}".format(warmup_method)
|
|
)
|
|
self.milestones = milestones
|
|
self.gamma = gamma
|
|
self.warmup_factor = warmup_factor
|
|
self.warmup_iters = warmup_iters
|
|
self.warmup_method = warmup_method
|
|
super(WarmupMultiStepLR, self).__init__(optimizer, last_epoch)
|
|
|
|
def get_lr(self):
|
|
warmup_factor = 1
|
|
if self.last_epoch < self.warmup_iters:
|
|
if self.warmup_method == "constant":
|
|
warmup_factor = self.warmup_factor
|
|
elif self.warmup_method == "linear":
|
|
alpha = float(self.last_epoch) / self.warmup_iters
|
|
warmup_factor = self.warmup_factor * (1 - alpha) + alpha
|
|
return [
|
|
base_lr
|
|
* warmup_factor
|
|
* self.gamma ** bisect_right(self.milestones, self.last_epoch)
|
|
for base_lr in self.base_lrs
|
|
]
|
|
|
|
|
|
class MultiStepLR(torch.optim.lr_scheduler._LRScheduler):
|
|
|
|
def __init__(self, optimizer, milestones, gamma=0.1, last_epoch=-1):
|
|
self.milestones = Counter(milestones)
|
|
self.gamma = gamma
|
|
super(MultiStepLR, self).__init__(optimizer, last_epoch)
|
|
|
|
def get_lr(self):
|
|
if self.last_epoch not in self.milestones:
|
|
return [group['lr'] for group in self.optimizer.param_groups]
|
|
return [group['lr'] * self.gamma ** self.milestones[self.last_epoch]
|
|
for group in self.optimizer.param_groups]
|
|
|
|
|
|
class ExponentialLR(torch.optim.lr_scheduler._LRScheduler):
|
|
|
|
def __init__(self, optimizer, decay_epochs, gamma=0.1, last_epoch=-1):
|
|
self.decay_epochs = decay_epochs
|
|
self.gamma = gamma
|
|
super(ExponentialLR, self).__init__(optimizer, last_epoch)
|
|
|
|
def get_lr(self):
|
|
return [base_lr * self.gamma ** (self.last_epoch / self.decay_epochs)
|
|
for base_lr in self.base_lrs]
|
|
|
|
def Scheduler(cfg_scheduler, optimizer):
|
|
if cfg_scheduler.type == 'multi_step':
|
|
scheduler = MultiStepLR(optimizer,
|
|
milestones=cfg_scheduler.milestones,
|
|
gamma=cfg_scheduler.gamma)
|
|
elif cfg_scheduler.type == 'exponential':
|
|
scheduler = ExponentialLR(optimizer,
|
|
decay_epochs=cfg_scheduler.decay_epochs,
|
|
gamma=cfg_scheduler.gamma)
|
|
else:
|
|
raise NotImplementedError
|
|
return scheduler
|
|
|
|
def set_lr_scheduler(cfg_scheduler, scheduler):
|
|
if cfg_scheduler.type == 'multi_step':
|
|
scheduler.milestones = Counter(cfg_scheduler.milestones)
|
|
elif cfg_scheduler.type == 'exponential':
|
|
scheduler.decay_epochs = cfg_scheduler.decay_epochs
|
|
scheduler.gamma = cfg_scheduler.gamma |