一、为什么要调整学习率

学习率(learning rate): 控制更新的步伐

梯度下降: $w_{i+1} = w_i-LR * g(w_i)$

在这里插入图片描述
学习率是用来控制更新的步伐，学习率一般前期大后期小，到后期需要调整学习率，让loss逐渐下降到收敛，就如同打高尔夫球，前期大力挥杆，使球到洞口附近，后期接近洞口的时候就需要调整小力度

二、pytorch的六种学习率调整策略

2.1 LRScheduler

class LRScheduler(object):
	def __init__(self, optimizer, last_epoch=-1):

	def get_lr(self):
		raise NotImplementedError

功能：pytorch的六种学习率调整策略方法的基类

主要属性:

optimizer: 关联的优化器
last_epoch: 记录epoch数
base_lrs: 记录初始学习率

主要方法:

step(): 更新下一个epoch的学习率，在epoch的for循环处使用
getIr(): 虚函数,计算下一个epoch的学习率

2.2 StepLR

lr_scheduler.StepLR(optimizer, step_size, gamma=0.1, last_epoch=-1)

功能: 等间隔调整学习率

主要参数:

step_size:调整间隔数
gamma: 调整系数

调整方式: Ir =Ir * gamma

# -*- coding:utf-8 -*-

import torch
import torch.optim as optim
import numpy as np
import matplotlib.pyplot as plt
torch.manual_seed(1)

LR = 0.1
iteration = 10
max_epoch = 200
# ------------------------------ fake data and optimizer  ------------------------------

weights = torch.randn((1), requires_grad=True)
target = torch.zeros((1))

optimizer = optim.SGD([weights], lr=LR, momentum=0.9)


# ------------------------------ 1 Step LR ------------------------------
# flag = 0
flag = 1
if flag:

    scheduler_lr = optim.lr_scheduler.StepLR(optimizer, step_size=50, gamma=0.1)  # 设置学习率下降策略

    lr_list, epoch_list = list(), list()
    for epoch in range(max_epoch):

        lr_list.append(scheduler_lr.get_lr())
        epoch_list.append(epoch)

        for i in range(iteration):

            loss = torch.pow((weights - target), 2)
            loss.backward()

            optimizer.step()
            optimizer.zero_grad()

        scheduler_lr.step()

    plt.plot(epoch_list, lr_list, label="Step LR Scheduler")
    plt.xlabel("Epoch")
    plt.ylabel("Learning rate")
    plt.legend()
    plt.show()

在这里插入图片描述

2.3 MultiStepLR

lr_scheduler.MultiStepLR(optimizer, milestones, gamma=0.1, last_epoch=-1)

功能: 按给定间隔调整学习率

主要参数:

milestones:设定调整时刻数
gamma:调整系数

调整方式: Ir =Ir * gamma

# -*- coding:utf-8 -*-

import torch
import torch.optim as optim
import numpy as np
import matplotlib.pyplot as plt
torch.manual_seed(1)

LR = 0.1
iteration = 10
max_epoch = 200
# ------------------------------ fake data and optimizer  ------------------------------

weights = torch.randn((1), requires_grad=True)
target = torch.zeros((1))

optimizer = optim.SGD([weights], lr=LR, momentum=0.9)


# ------------------------------ 2 Multi Step LR ------------------------------
# flag = 0
flag = 1
if flag:

    milestones = [50, 125, 160]
    scheduler_lr = optim.lr_scheduler.MultiStepLR(optimizer, milestones=milestones, gamma=0.1)

    lr_list, epoch_list = list(), list()
    for epoch in range(max_epoch):

        lr_list.append(scheduler_lr.get_lr())
        epoch_list.append(epoch)

        for i in range(iteration):

            loss = torch.pow((weights - target), 2)
            loss.backward()

            optimizer.step()
            optimizer.zero_grad()

        scheduler_lr.step()

    plt.plot(epoch_list, lr_list, label="Multi Step LR Scheduler\nmilestones:{}".format(milestones))
    plt.xlabel("Epoch")
    plt.ylabel("Learning rate")
    plt.legend()
    plt.show()

在这里插入图片描述

2.4 ExponentialLR

lr_scheduler.ExponentialLR(optimizer, gamma, last_epoch=-1)

功能: 按指数衰减调整学习率

主要参数:

gamma: 指数的底

调整方式: Ir = Ir * gamma ** epoch

# -*- coding:utf-8 -*-

import torch
import torch.optim as optim
import numpy as np
import matplotlib.pyplot as plt
torch.manual_seed(1)

LR = 0.1
iteration = 10
max_epoch = 200
# ------------------------------ fake data and optimizer  ------------------------------

weights = torch.randn((1), requires_grad=True)
target = torch.zeros((1))

optimizer = optim.SGD([weights], lr=LR, momentum=0.9)


# ------------------------------ 3 Exponential LR ------------------------------
# flag = 0
flag = 1
if flag:

    gamma = 0.95
    scheduler_lr = optim.lr_scheduler.ExponentialLR(optimizer, gamma=gamma)

    lr_list, epoch_list = list(), list()
    for epoch in range(max_epoch):

        lr_list.append(scheduler_lr.get_lr())
        epoch_list.append(epoch)

        for i in range(iteration):

            loss = torch.pow((weights - target), 2)
            loss.backward()

            optimizer.step()
            optimizer.zero_grad()

        scheduler_lr.step()

    plt.plot(epoch_list, lr_list, label="Exponential LR Scheduler\ngamma:{}".format(gamma))
    plt.xlabel("Epoch")
    plt.ylabel("Learning rate")
    plt.legend()
    plt.show()

在这里插入图片描述

2.5 CosineAnnealingLR

lr_scheduler.CosineAnnealingLR(optimizer, T_max, eta_min=0, last_epoch=-1)

功能: 余弦周期调整学习率

主要参数:

T-max:下降周期
eta_min:学习率下限

调整方式：
在这里插入图片描述

# -*- coding:utf-8 -*-

import torch
import torch.optim as optim
import numpy as np
import matplotlib.pyplot as plt
torch.manual_seed(1)

LR = 0.1
iteration = 10
max_epoch = 200
# ------------------------------ fake data and optimizer  ------------------------------

weights = torch.randn((1), requires_grad=True)
target = torch.zeros((1))

optimizer = optim.SGD([weights], lr=LR, momentum=0.9)


# ------------------------------ 4 Cosine Annealing LR ------------------------------
# flag = 0
flag = 1
if flag:

    t_max = 50
    scheduler_lr = optim.lr_scheduler.CosineAnnealingLR(optimizer, T_max=t_max, eta_min=0.)

    lr_list, epoch_list = list(), list()
    for epoch in range(max_epoch):

        lr_list.append(scheduler_lr.get_lr())
        epoch_list.append(epoch)

        for i in range(iteration):

            loss = torch.pow((weights - target), 2)
            loss.backward()

            optimizer.step()
            optimizer.zero_grad()

        scheduler_lr.step()

    plt.plot(epoch_list, lr_list, label="CosineAnnealingLR Scheduler\nT_max:{}".format(t_max))
    plt.xlabel("Epoch")
    plt.ylabel("Learning rate")
    plt.legend()
    plt.show()

在这里插入图片描述

2.6 ReduceLROnPlateau

lr_scheduler.ReduceLROnPlateau(optimizer,
 							   mode='min', 
 							   factor=0.1, 
 							   patience=10, 
 							   verbose=False, 
 							   threshold=0.0001, 
 							   threshold_mode='rel', 
 							   cooldown=0, 
 							   min_lr=0, 
 							   eps=1e-08)

功能: 监控指标,当指标不再变化则调整

主要参数:

mode: min/max两种模式
- min：观察指标下降，用于loss
- max：观察指标上升，用于accuracy
factor: 调整系数
patience: “耐心”,接受几次不变化
cooldown: "冷却时间”,停止监控一段时间
verbose: 是否打印日志
minIr: 学习率下限
eps: 学习率衰减最小值

# -*- coding:utf-8 -*-

import torch
import torch.optim as optim
import numpy as np
import matplotlib.pyplot as plt
torch.manual_seed(1)

LR = 0.1
iteration = 10
max_epoch = 200
# ------------------------------ fake data and optimizer  ------------------------------

weights = torch.randn((1), requires_grad=True)
target = torch.zeros((1))

optimizer = optim.SGD([weights], lr=LR, momentum=0.9)

# ------------------------------ 5 Reduce LR On Plateau ------------------------------
# flag = 0
flag = 1
if flag:
    loss_value = 0.5
    accuray = 0.9

    factor = 0.1
    mode = "min"
    patience = 10
    cooldown = 10
    min_lr = 1e-4
    verbose = True

    scheduler_lr = optim.lr_scheduler.ReduceLROnPlateau(optimizer, factor=factor, mode=mode, patience=patience,
                                                        cooldown=cooldown, min_lr=min_lr, verbose=verbose)

    for epoch in range(max_epoch):
        for i in range(iteration):

            # train(...)

            optimizer.step()
            optimizer.zero_grad()

        if epoch == 5:
            loss_value = 0.4

        scheduler_lr.step(loss_value)

在这里插入图片描述

2.7 LambdaLR

Ir_scheduler.LambdaLR(optimizer, lr_lambda, last_epoch=-1)

功能: 自定义调整策略，适用于多个参数组，不同参数组设置不同调整策略时

主要参数:

Ir_lambda: function or list

# -*- coding:utf-8 -*-

import torch
import torch.optim as optim
import numpy as np
import matplotlib.pyplot as plt
torch.manual_seed(1)

LR = 0.1
iteration = 10
max_epoch = 200
# ------------------------------ fake data and optimizer  ------------------------------

weights = torch.randn((1), requires_grad=True)
target = torch.zeros((1))

optimizer = optim.SGD([weights], lr=LR, momentum=0.9)


# ------------------------------ 6 lambda ------------------------------
# flag = 0
flag = 1
if flag:

    lr_init = 0.1

    weights_1 = torch.randn((6, 3, 5, 5))
    weights_2 = torch.ones((5, 5))

    optimizer = optim.SGD([
        {'params': [weights_1]},
        {'params': [weights_2]}], lr=lr_init)

    lambda1 = lambda epoch: 0.1 ** (epoch // 20)
    lambda2 = lambda epoch: 0.95 ** epoch

    scheduler = torch.optim.lr_scheduler.LambdaLR(optimizer, lr_lambda=[lambda1, lambda2])

    lr_list, epoch_list = list(), list()
    for epoch in range(max_epoch):
        for i in range(iteration):

            # train(...)

            optimizer.step()
            optimizer.zero_grad()

        scheduler.step()

        lr_list.append(scheduler.get_lr())
        epoch_list.append(epoch)

        print('epoch:{:5d}, lr:{}'.format(epoch, scheduler.get_lr()))

    plt.plot(epoch_list, [i[0] for i in lr_list], label="lambda 1")
    plt.plot(epoch_list, [i[1] for i in lr_list], label="lambda 2")
    plt.xlabel("Epoch")
    plt.ylabel("Learning Rate")
    plt.title("LambdaLR")
    plt.legend()
    plt.show()

在这里插入图片描述

三、学习率调整小结

3.1 学习率调整策略总结

有序调整: Step.MultiStep、Exponential和CosineAnnealing
自适应调整: ReduceLROnPleateau
自定义调整: Lambda

3.2 学习率初始化

设置较小数: 0.01. 0.001,0.0001
搜索最大学习率: 参考论文《Cyclical Learning Rates for Training Neural Networks)

G5Lorenzo

发布了105 篇原创文章 · 获赞 9 · 访问量 7807

私信关注

16学习率调整策略

一、为什么要调整学习率

二、pytorch的六种学习率调整策略

2.1 LRScheduler

2.2 StepLR

2.3 MultiStepLR

2.4 ExponentialLR

2.5 CosineAnnealingLR

2.6 ReduceLROnPlateau

2.7 LambdaLR

三、学习率调整小结

3.1 学习率调整策略总结

3.2 学习率初始化

猜你喜欢