PyTorch深度学习快速入门
|
Pytorch,深度学习
|
474

2537 字
|
1.6 小时
本文最后更新于162 天前，其中的信息可能已经过时，如有错误请发送邮件到big_fw@foxmail.com

PyTorch深度学习快速入门教程

参考（主要是1，环境搭建参考2）

1.环境安装

环境安装参考[2]第一节，可以基本解决可能遇到的环境安装问题

2.Pytorch数据加载

重写Dataset函数里的__getitem__()和__len__()函数

#!/usr/bin/venv python 
# -*- coding:utf-8 -*-
# @Author :LiJie
# @Time :2024/7/12 上午12:34
# @Filename :test.py
from torch.utils.data import Dataset, DataLoader
from PIL import Image
import os

class MyData(Dataset):
    def __init__(self, root_dir, label_dir):
        self.root_dir = root_dir
        self.label_dir = label_dir
        self.path = os.path.join(self.root_dir, self.label_dir)
        self.img_path = os.listdir(self.path)

    def __getitem__(self, idx):
        img_name = self.img_path[idx]
        img_item_path = os.path.join(self.path, img_name)
        img = Image.open(img_item_path)
        label = self.label_dir
        return img, label

    def __len__(self):
        return len(self.img_path)

root_dir = "dataset/train"
ants_label_dir = "ants"
ants_dataset = MyData(root_dir, ants_label_dir)

img, label = ants_dataset[0]
img.show()

3.TensorBoard使用

3.1 TensorBoard安装以及可能遇到的问题

TensorBoard可以将代码运行过程可视化，是一个非常不错的辅助工具

TensorBoard没有默认安装，在conda虚拟环境中输入安装指令

pip install tensorboard

在终端运行Tensorboard

# --logdir指定日志目录
tensorboard --logdir=logs

# --port 可以指定端口(默认端口6006)
tensorboard --logdir=logs --port==6007

在使用Tensorboard，可能会遇到numpy版本过高问题，将numpy降级即可

问题显示情况： “AttributeError: np.string_ was removed in the NumPy 2.0 release. Use `np.bytes”

pip uninstalll numpy
pip install numpy==1.23.0

其他可能出现的问题：

tensorboard出现TensorFlow installation not found - running with reduced feature set

解决办法，改用绝对路径

 tensorboard --logdir="D:\PycharmProjects\PyTorch\another_practice\CIFAR10"

3.2 TensorBoard使用

演示代码

from torch.utils.tensorboard import SummaryWriter

writer = SummaryWriter("logs")

# writer.add_image()

# y = 2x
for i in range(100):
    writer.add_scalar("y=2x", 2*i, i) #此函数一般使用三个参数：标题title y轴 x轴

writer.close()

此处=会遇到另一个问题，title没改改变x轴和y轴参数，图像会被写到一个图片里，tensorboard在拟合函数图像时会导致图片混乱，决绝办法是将日志文件夹的文件删除

from torch.utils.tensorboard import SummaryWriter
import numpy as np
from PIL import Image

img_path = "data/train/bees_image/16838648_415acd9e3f.jpg"
img_PIL = Image.open(img_path)
img_array = np.array(img_PIL)
print(img_array.shape)

writer = SummaryWriter("logs")

writer.add_image("test", img_array, 2, dataformats='HWC')   #注意img_array代表格式，以及dataformats默认格式为'CHW'

# y = 2x
for i in range(100):
    writer.add_scalar("y=2x", 2*i, i)

writer.close()

4.Transformer的使用

Transforms是一个深度学习类函数包

常用的有：transforms.ToTensor() 将数据转为tensor类型

transforms.Normalize() 数据标准化

transforms.Resize( sequnce or int ) 大小转换

transforms.Compose([transforms变量1，transforms变量2， ... ]) 前一变量输出作为下一变量输入

# 常见的Transforms

from torchvision import transforms
from PIL import Image
from torch.utils.tensorboard import SummaryWriter

writer = SummaryWriter('logs')
img_path = "data/train/ants_image/0013035.jpg"
img = Image.open(img_path)

# ToTensor的使用
trans_totensr = transforms.ToTensor()
img_tensor = trans_totensr(img)
writer.add_image("ToTensor", img_tensor)

# Normalize
print(img_tensor.shape)
print(img_tensor[0][0][0])
trans_norm = transforms.Normalize(mean=[0.5, 0.5, 0.5], std=[0.5, 0.5, 0.5])
img_norm = trans_norm(img_tensor)
print(img_norm[0][0][0])
writer.add_image("Normalize", img_norm)

# Resize
print(img.size)
trans_resize = transforms.Resize((512, 512))
# img PIL -> resize -> img_resize PIL
img_resize = trans_resize(img)
# img_resize PIL -> ToTensor -> img_resize tensor
img_resize_totensor = trans_totensr(img_resize)
writer.add_image("Resize", img_resize_totensor, 0)
print(img_resize)

# Compose - resize - 2 另一种Resize的方法
trans_resize_2 = transforms.Resize(512)
trans_compose = transforms.Compose([trans_resize, transforms.ToTensor()])
img_resize_2 = trans_compose(img)
writer.add_image("Resize", img_resize_2, 1)

writer.close()

5.torchvision数据集

可以从函数声明找到下载链接，在迅雷下载数据集，将压缩包下载到root目录，download依然设置为True，代码运行时会自动解压并verify

import torchvision
from torch.utils.tensorboard import SummaryWriter
from torchvision.transforms import transforms

dataset_transform = transforms.Compose([
    transforms.ToTensor(),
])
train_set = torchvision.datasets.CIFAR10(root='./data2', train=True, transform=dataset_transform, download=True)
test_set = torchvision.datasets.CIFAR10(root='./data2', train=False, transform=dataset_transform, download=True)

writer = SummaryWriter(log_dir='p11')
for i in range(10):
    img, target = test_set[i]
    writer.add_image('test_set', img, i)

writer.close()

6.Dataloader使用

从dataset中取数据

import torchvision
from torch.utils.data import DataLoader
from torch.utils.tensorboard import SummaryWriter

test_set = torchvision.datasets.CIFAR10(root='./data2', train=False, transform=torchvision.transforms.ToTensor())

# shuffle为true时，每次循环的抽取次序不一样，false则一样
# drop_last是除以batch_size后剩余的数据，是否保存，False保存，True不保存
# num_workers 多线程
test_loader = DataLoader(dataset=test_set, batch_size=64, shuffle=True, num_workers=0, drop_last=False)

print(test_loader)

img, target = test_set[0]
print(img.shape)
print(target)
i = 0
writer = SummaryWriter(log_dir='dataloader')
step = 0
for data in test_loader:
    i += 1
    print(f"==========={i}===================")
    imgs, targets = data
    # print(imgs.shape)
    # print(targets.shape)
    writer.add_images('test_set', imgs, step)
    step += 1

writer.close()

7.nn.Module的使用

eg.

import torch.nn as nn
import torch.nn.functional as F

class Model(nn.Module):
    def __init__(self):
        super().__init__()
        self.conv1 = nn.Conv2d(1, 20, 5)
        self.conv2 = nn.Conv2d(20, 20, 5)

    def forward(self, x):
        x = F.relu(self.conv1(x))
        return F.relu(self.conv2(x))

8.nn_conv卷积函数

torch.nn.functional.conv2d(input, weight, bias=None, stride=1, padding=0, dilation=1, groups=1)

input为一个四维的tensor数据:input – input tensor of shape (minibatch,in_channels,𝑖𝐻,𝑖𝑊)(minibatch,in_channels,*i*H,iW)

import torch.nn.functional as F
import torch

input = torch.tensor([[1, 2, 0, 3, 1],
                      [0, 1, 2, 3, 1],
                      [1, 2, 1, 0, 0],
                      [5, 2, 3, 1, 1],
                      [2, 1, 0, 1, 1]])

kernal = torch.tensor([[1, 2, 1],
                       [0, 1, 0],
                       [2, 1, 0]])

# print(input.shape)
# print(kernal.shape)

input = torch.reshape(input, [1, 1, 5, 5])
kernal = torch.reshape(kernal, [1, 1, 3, 3])

output = F.conv2d(input, kernal, stride=1, padding=0)
print(output)

output2 = F.conv2d(input, kernal, stride=2, padding=0)
print(output2)

9.nn_Conv2d

注意输入输出层之间的HWC的计算

import torch
import torchvision
from torch import nn
from torch.nn import Conv2d
from torch.utils.data import DataLoader

dataset = torchvision.datasets.CIFAR10(root='../data2', train=False, transform=torchvision.transforms.ToTensor(), download=True)
dataloader = DataLoader(dataset, batch_size=64)

class Tudui(nn.Module):
    def __init__(self):
        super(Tudui, self).__init__()
        self.conv1 = Conv2d(3, 6, kernel_size=3, stride=1, padding=0)

    def forward(self, x):
        x = self.conv1(x)
        return x

tudui = Tudui()
print(tudui)

for data in dataloader:
    imgs, targets = data
    output = tudui(imgs)
    print(output.shape)

10.最大池化 MaxPool2d

减少数据同时保留特征；注意ceil=True，池化时即使核无法完全覆盖，也正常取值

import torch
import torchvision
from torch import nn
from torch.nn import MaxPool2d
from torch.utils.data import DataLoader
from torch.utils.tensorboard import SummaryWriter

dataset = torchvision.datasets.CIFAR10(root='../data2', train=False, transform=torchvision.transforms.ToTensor(), download=True)
dataloader = DataLoader(dataset, batch_size=64)
input = torch.tensor([[1, 2, 0, 3, 1],
                      [0, 1, 2, 3, 1],
                      [1, 2, 1, 0, 0],
                      [5, 2, 3, 1, 1],
                      [2, 1, 0, 1, 1]], dtype=torch.float32)
input = torch.reshape(input, (-1, 1, 5, 5))

class Tudui(nn.Module):
    def __init__(self):
        super(Tudui, self).__init__()
        self.maxpool1 = MaxPool2d(kernel_size=3, ceil_mode=False)

    def forward(self, input):
        output = self.maxpool1(input)
        return output

tudui = Tudui()
# output = tudui(input)
# print(output)
writer = SummaryWriter(log_dir='../logs_maxpool')
step = 0
for data in dataloader:
    imgs, targets = data
    writer.add_images('input', imgs, step)
    output = tudui(imgs)
    writer.add_images("output", output, step)
    step += 1

writer.close()

  • Input: (𝑁,𝐶𝑖𝑛,𝐻𝑖𝑛,𝑊𝑖𝑛)(N,*Cin,Hin,Win) or (𝐶𝑖𝑛,𝐻𝑖𝑛,𝑊𝑖𝑛)(Cin,Hin,Win*)

  • Output: (𝑁,𝐶𝑜𝑢𝑡,𝐻𝑜𝑢𝑡,𝑊𝑜𝑢𝑡)(N,*Cout,Hout,Wout) or (𝐶𝑜𝑢𝑡,𝐻𝑜𝑢𝑡,𝑊𝑜𝑢𝑡)(Cout,Hout,Wout*), where

    𝐻𝑜𝑢𝑡=⌊𝐻𝑖𝑛+2×padding[0]−dilation[0]×(kernel_size[0]−1)−1stride[0]+1⌋*Hout=⌊stride[0]Hin*+2×padding[0]−dilation[0]×(kernel_size[0]−1)−1+1⌋

    𝑊𝑜𝑢𝑡=⌊𝑊𝑖𝑛+2×padding1−dilation1×(kernel_size1−1)−1stride1+1⌋*Wout=⌊stride1Win*+2×padding1−dilation1×(kernel_size1−1)−1+1⌋

11.非线性激活

nn.ReLu() inplace 参数，为True时会将输入的值做相应改变，默认为False
$$
ReLu(x)=(x)^+max(0,x)
$$

../_images/ReLU.png

import torch
import torch.nn as nn
import torchvision
from torch.utils.data import DataLoader
from torch.utils.tensorboard import SummaryWriter

dataset = torchvision.datasets.CIFAR10(root='../data2', train=False, transform=torchvision.transforms.ToTensor(), download=True)
dataloader = DataLoader(dataset, batch_size=64)

class Tudui(nn.Module):
    def __init__(self):
        super(Tudui, self).__init__()
        # self.relu = nn.ReLU()
        self.sigmoid = nn.Sigmoid()

    def forward(self, input):
        output = self.sigmoid(input)
        return output

tudui = Tudui()
writer = SummaryWriter(log_dir='../logs_sigmod')
step = 0
for data in dataloader:
    imgs, targets = data
    writer.add_images('input', imgs, step)
    output = tudui(imgs)
    writer.add_images("output_sigmoid", output, step)
    step += 1

writer.close()

12.线性层（torch.nn.Linear）

线性层又叫全连接层，其中每个神经元与上一层所有神经元相连，一个简单的线性层如下图所示：

线性层

线性函数为：torch.nn.Linear(in_features, out_features, bias=True, device=None, dtype=None)，其中重要的3个参数in_features、out_features、bias说明如下：

in_features：每个输入（x）样本的特征的大小
out_features：每个输出（y）样本的特征的大小
bias：如果设置为False，则图层不会学习附加偏差。默认值是True，表示增加学习偏置。
    在上图中，in_features=d，out_features=L。
作用可以是缩小一维的数据长度

下面代码中注意drop_last设置为true，否则会报错。

torch.flatten展平，可用torch.view(),torch.reshape(-1)代替

import torch
import torchvision
from torch import nn
from torch.utils.data import DataLoader

dataset = torchvision.datasets.CIFAR10(root='../data2', train=False, transform=torchvision.transforms.ToTensor(),
                                       download=True)
img_data = DataLoader(dataset=dataset, batch_size=64, drop_last=True)

class Tudui(nn.Module):
    def __init__(self):
        super(Tudui, self).__init__()
        self.linear1 = nn.Linear(in_features=196608, out_features=10)

    def forward(self, x):
        x = self.linear1(x)
        return x

tudui = Tudui()
for data in img_data:
    imgs, targets = data
    print(imgs.shape)
    imgs_flatten = torch.flatten(imgs)
    print(imgs_flatten.shape)
    linear_out = tudui(imgs_flatten)
    print(linear_out.shape)

13.Sequntial

Sequntial将网络各部分包装成一个model

import torch
from torch import nn
from torch.nn import Conv2d, MaxPool2d, Linear, Flatten
from torch.utils.tensorboard import SummaryWriter

class Tudui(nn.Module):
    def __init__(self):
        super(Tudui, self).__init__()
        self.model1 = nn.Sequential(
            Conv2d(3, 32, 5, padding=2),
            MaxPool2d(2),
            Conv2d(32, 32, 5, padding=2),
            MaxPool2d(2),
            Conv2d(32, 64, 5, padding=2),
            MaxPool2d(2),
            Flatten(),
            Linear(1024, 64),
            Linear(64, 10)
        )

    def forward(self, x):
        x = self.model1(x)
        return x

tudui = Tudui()
print(tudui)
input = torch.ones((64, 3, 32, 32))
output = tudui(input)
print(output.shape)

writer = SummaryWriter('../logs_seq')
writer.add_graph(tudui, input)
writer.close()

14.损失函数与反向传播

loss test

"""
 @Author: LiJie
 @FileName: nn_loss.py
 @DateTime: 2024/7/15 下午7:33
"""
from torch.nn import L1Loss, MSELoss, CrossEntropyLoss
import torch

x1 = torch.tensor([1, 2, 3], dtype=torch.float32)
y1 = torch.tensor([1, 2, 5], dtype=torch.float32)

# L1loss
loss1 = L1Loss()
loss_l1 = loss1(x1, y1)
print(f"loss_l1 = {loss_l1}")

# MSELoss
loss2 = MSELoss()
loss_mse = loss2(x1, y1)
print(f"loss_mse = {loss_mse}")

# CrossEntryLoss
loss3 = CrossEntropyLoss()
x2 = torch.tensor([0.1, 0.2, 0.3], dtype=torch.float32)
print(x2.shape)
x2 = torch.reshape(x2, [1, 3])
y2 = torch.tensor([1])
loss_crossrentry = loss3(x2, y2)
print(f"loss_crossrentry = {loss_crossrentry}")

loss + backword

"""
 @Author: LiJie
 @FileName: nn_network_loss.py
 @DateTime: 2024/7/15 下午8:19
"""

import torchvision
from torch import nn
from torch.nn import Conv2d, MaxPool2d, Linear, Flatten
from torch.utils.data import DataLoader

dataset = torchvision.datasets.CIFAR10(root='../data2', train=False, transform=torchvision.transforms.ToTensor(),
                                       download=True)
dataloader = DataLoader(dataset, batch_size=1)

class Tudui(nn.Module):
    def __init__(self):
        super(Tudui, self).__init__()
        self.model1 = nn.Sequential(
            Conv2d(3, 32, 5, padding=2),
            MaxPool2d(2),
            Conv2d(32, 32, 5, padding=2),
            MaxPool2d(2),
            Conv2d(32, 64, 5, padding=2),
            MaxPool2d(2),
            Flatten(),
            Linear(1024, 64),
            Linear(64, 10)
        )

    def forward(self, x):
        x = self.model1(x)
        return x

loss = nn.CrossEntropyLoss()
tudui = Tudui()
step = 1
for data in dataloader:
    imgs, targets = data
    output = tudui(imgs)
    result_loss = loss(output, targets)
    result_loss.backward()
    print(step)
    step += 1

15.优化器 torch.optim

1、过程描述

继上节的计算损失函数和反向传播，
之后便是根据损失值，利用优化器进行梯度更新，然后不断降低loss的过程
一般要对数据集扫描多遍，进行参数的多次更新，才能得到一个较好的效果。
注意，每次更新后要将梯度置0，然后重新计算梯度注意，每次更新后要将梯度置0，然后重新计算梯度

2、常用优化器：

优化器的种类比较多，常用的就是随机梯度下降（SGD） 等
不同的优化器的参数列表一般不同，但都会有 params(模型的参数列表)和lr(学习率)参数，

eg.

"""
 @Author: LiJie
 @FileName: nn_optim.py
 @DateTime: 2024/7/15 下午8:50
"""
import torch
import torchvision
from torch import nn
from torch.nn import Conv2d, MaxPool2d, Linear, Flatten
from torch.utils.data import DataLoader

dataset = torchvision.datasets.CIFAR10(root='../data2', train=False, transform=torchvision.transforms.ToTensor(),
                                       download=True)
dataloader = DataLoader(dataset, batch_size=1)

class Tudui(nn.Module):
    def __init__(self):
        super(Tudui, self).__init__()
        self.model1 = nn.Sequential(
            Conv2d(3, 32, 5, padding=2),
            MaxPool2d(2),
            Conv2d(32, 32, 5, padding=2),
            MaxPool2d(2),
            Conv2d(32, 64, 5, padding=2),
            MaxPool2d(2),
            Flatten(),
            Linear(1024, 64),
            Linear(64, 10)
        )

    def forward(self, x):
        x = self.model1(x)
        return x

loss = nn.CrossEntropyLoss()
tudui = Tudui()
optim = torch.optim.SGD(tudui.parameters(), lr=0.01)
step = 1
for epoch in range(10):
    loss_sum = 0.0
    for data in dataloader:
        imgs, targets = data
        output = tudui(imgs)
        result_loss = loss(output, targets)
        optim.zero_grad()
        result_loss.backward()
        optim.step()
        loss_sum += result_loss
    print('epoch:', epoch, 'loss:', loss_sum)

16.VGG16

使用torchvision.models.vgg16模型，新版vgg16用weights代替pretrain参数，weights=None代表没有预训练，另外参数有weights='IMAGENET1K_V1' 或者 weights='IMAGENET1K_FEATRUES' (这里出现一个问题，官方文档说DEFAULT等价于IMAGENET1K_V1，但是代码文件介绍写DEFALUT等于没有预训练，实际运行代码说明官方文档是对的。。。。)

模型下载可能会很慢，下载时会显示下载链接，可复制到迅雷下载，再放到下载目录（下载目录运行的时候会显示）下载目录为： C:\Users\用户名.cache\torch\hub\checkpoints

然后是vgg预训练模型下载位置的修改，待填坑

vgg模型的结构修改，也叫模型迁移，三种修改，代码：

"""
 @Author: LiJie
 @FileName: nn_VGG16.py
 @DateTime: 2024/7/16 下午3:18
"""
import torchvision
from torch import nn

vgg16_false = torchvision.models.vgg16(weights=None)
vgg16_true = torchvision.models.vgg16(weights='IMAGENET1K_V1')

print("vgg16原始模型结构")
print(vgg16_false)
print(vgg16_true)

# 在最外层添加一个线性层
print("vgg16在最外层添加一个线性层")
vgg16_true.add_module('add_linear', nn.Linear(in_features=1000, out_features=10))
print(vgg16_true)

# 在classifier添加一个线性层
print("vgg16在classifier添加一个线性层")
vgg16_false.classifier.add_module('add_linear', nn.Linear(in_features=1000, out_features=10))
print(vgg16_false)

# 将classifier的第6层修改
print("vgg16 classifier第6层修改")
vgg16_true.classifier[6] = nn.Linear(in_features=1000, out_features=10)
print(vgg16_true)

17.模型的保存与读取

module_save.py

"""
 @Author: LiJie
 @FileName: module_save.py
 @DateTime: 2024/7/16 下午4:28
"""
import torch
import torchvision
from torch import nn

vgg16 = torchvision.models.vgg16(weights=None)
# 保存方式1，模型结构+参数
torch.save(vgg16, "vgg16_method1.pth")

# 保存方式2，模型参数（官方推荐）
torch.save(vgg16.state_dict(), "vgg16_method2.pth")

# 方式1的陷阱
class Tudui(nn.Module):
    def __init__(self):
        super(Tudui, self).__init__()
        self.conv1 = nn.Conv2d(3, 64, kernel_size=3, padding=1)

    def forward(self, x):
        x = self.conv1(x)
        return x

tudui = Tudui()
torch.save(tudui, "Tudui_method1.pth")

module_load.py

使用方式1保存自己的模型，再加载时，需要import源代码，否则会报错

"""
 @Author: LiJie
 @FileName: module_load.py
 @DateTime: 2024/7/16 下午4:31
"""
import torch
import torchvision
from module_save import *

# 方式1，加载模型
model1 = torch.load("vgg16_method1.pth")
print(model1)

# 方式2，加载模型
vgg16 = torchvision.models.vgg16(weights='IMAGENET1K_V1')
vgg16.load_state_dict(torch.load("vgg16_method2.pth"))
print(vgg16)

# 方式1 存在的陷阱
model_tudui = torch.load("tudui_method1.pth")
print(model_tudui)

18.GPU训练模型

CPU训

  1. 准备数据集

  2. 用DataLoader加载数据

  3. 创建网络模型

  4. 设置损失函数

  5. 设置优化器

  6. 开始训练

    1. model.train()

    2. 训练步骤

      模型输出

      优化器优化：梯度归零；反向传播；optimizer.step()；

    3. 测试步骤 with torch.no_grade():

      计算loss和accuracy accuracy = (outputs.argmax(1) == targets).sum()

code eg.

model.py

"""
 @Author: LiJie
 @FileName: model.py
 @DateTime: 2024/7/16 下午7:23
"""
import torch
from torch import nn

class Tudui(nn.Module):
    def __init__(self):
        super(Tudui, self).__init__()
        self.model = nn.Sequential(
            nn.Conv2d(3, 32, kernel_size=5, stride=1, padding=2),
            nn.MaxPool2d(2, stride=2),
            nn.Conv2d(32, 32, kernel_size=5, stride=1, padding=2),
            nn.MaxPool2d(2, stride=2),
            nn.Conv2d(32, 64, kernel_size=5, stride=1, padding=2),
            nn.MaxPool2d(2, stride=2),
            nn.Flatten(),
            nn.Linear(64 * 4 * 4, 64),
            nn.Linear(64, 10),
        )

    def forward(self, x):
        x = self.model(x)
        return x

if __name__ == '__main__':
    tudui = Tudui()
    input = torch.ones((64, 3, 32, 32))
    output = tudui(input)
    print(output.shape)

train.py

"""
 @Author: LiJie
 @FileName: train.py
 @DateTime: 2024/7/16 下午7:21
"""
import torch
from torch import nn
import torchvision
from torch.utils.data import DataLoader
from torch.utils.tensorboard import SummaryWriter

from model import *

# 数据集
train_set = torchvision.datasets.CIFAR10(root='../data2', train=True, download=True,
                                         transform=torchvision.transforms.ToTensor())
test_set = torchvision.datasets.CIFAR10(root='../data2', train=False, download=True,
                                        transform=torchvision.transforms.ToTensor())

# 数据集长度
train_data_len = len(train_set)
test_data_len = len(test_set)
print("训练数据集长度:{}".format(train_data_len))
print("测试数据集长度:{}".format(test_data_len))

# DataLoader加载数据集
train_dataloader = DataLoader(train_set, batch_size=64, shuffle=True)
test_dataloader = DataLoader(test_set, batch_size=64, shuffle=True)

# 创建网络模型
tudui = Tudui()

# 损失函数
loss_fn = nn.CrossEntropyLoss()

# 优化器
learning_rate = 0.01    # 学习率
optimizer = torch.optim.SGD(tudui.parameters(), lr=learning_rate)

# 训练参数设置
total_train_step = 0    # 训练次数记录
total_test_step = 0     # 测试次数记录
epochs = 10              # 训练轮数

# 添加TensorBoard，观察loss和accuracy
writer = SummaryWriter('../logs_train')

# 开始训练
for epoch in range(epochs):
    print("==========epoch:{}==========".format(epoch))

    # 训练步骤
    tudui.train()
    for data in train_dataloader:
        imgs, targets = data
        outputs = tudui(imgs)
        loss = loss_fn(outputs, targets)

        # 优化器优化模型
        optimizer.zero_grad()
        loss.backward()
        optimizer.step()

        total_train_step += 1
        if total_train_step % 100 == 0:
            print("训练次数:{}，Loss:{}".format(total_train_step, loss.item()))
            writer.add_scalar('train_loss', loss.item(), total_train_step)

    # 测试步骤
    tudui.eval()
    total_test_loss = 0
    total_accuracy = 0
    with torch.no_grad():
        for data in test_dataloader:
            imgs, targets = data
            outputs = tudui(imgs)
            loss = loss_fn(outputs, targets)
            total_test_loss += loss.item()
            accuracy = (outputs.argmax(1) == targets).sum()
            total_accuracy += accuracy.item()

    print("测试集的Loss:{}".format(total_test_loss))
    print("测试集的正确率:{}".format(total_accuracy/test_data_len))
    writer.add_scalar('test_loss', total_test_loss, total_test_step)
    writer.add_scalar('test_accuracy', total_accuracy/test_data_len, total_test_step)
    total_test_step += 1

    torch.save(tudui, "tudui_{}.pth".format(epoch))
    print("model saved")

writer.close()

用GPU训练

方法一：

可加速对象：

网络模型；

loss function；

训练时的数据和测试时的数据。

if torch.cuda.is_available():
    zz = zzz.cuda()

测试代码：

"""
 @Author: LiJie
 @FileName: train.py
 @DateTime: 2024/7/16 下午7:21
"""
import torch
from torch import nn
import torchvision
from torch.utils.data import DataLoader
from torch.utils.tensorboard import SummaryWriter
import time
from model import *

# 数据集
train_set = torchvision.datasets.CIFAR10(root='../data2', train=True, download=True,
                                         transform=torchvision.transforms.ToTensor())
test_set = torchvision.datasets.CIFAR10(root='../data2', train=False, download=True,
                                        transform=torchvision.transforms.ToTensor())

# 数据集长度
train_data_len = len(train_set)
test_data_len = len(test_set)
print("训练数据集长度:{}".format(train_data_len))
print("测试数据集长度:{}".format(test_data_len))

# DataLoader加载数据集
train_dataloader = DataLoader(train_set, batch_size=64, shuffle=True)
test_dataloader = DataLoader(test_set, batch_size=64, shuffle=True)

# 创建网络模型
tudui = Tudui()
if torch.cuda.is_available():
    tudui = tudui.cuda()

# 损失函数
loss_fn = nn.CrossEntropyLoss()
if torch.cuda.is_available():
    loss_fn = loss_fn.cuda()

# 优化器
learning_rate = 0.01  # 学习率
optimizer = torch.optim.SGD(tudui.parameters(), lr=learning_rate)

# 训练参数设置
total_train_step = 0  # 训练次数记录
total_test_step = 0  # 测试次数记录
epochs = 10  # 训练轮数

# 添加TensorBoard，观察loss和accuracy
# writer = SummaryWriter('../logs_train')

start_time = time.time()
# 开始训练
for epoch in range(epochs):
    print("==========epoch:{}==========".format(epoch))

    # 训练步骤
    tudui.train()
    for data in train_dataloader:
        imgs, targets = data
        if torch.cuda.is_available():
            imgs = imgs.cuda()
            targets = targets.cuda()
        outputs = tudui(imgs)
        loss = loss_fn(outputs, targets)

        # 优化器优化模型
        optimizer.zero_grad()
        loss.backward()
        optimizer.step()

        total_train_step += 1
        end_time = time.time()
        if total_train_step % 100 == 0:
            print("训练次数:{}，Loss:{}".format(total_train_step, loss.item()))
            print("训练时间:{}".format(end_time - start_time))
            # writer.add_scalar('train_loss', loss.item(), total_train_step)

    # 测试步骤
    tudui.eval()
    total_test_loss = 0
    total_accuracy = 0
    with torch.no_grad():
        for data in test_dataloader:
            imgs, targets = data
            if torch.cuda.is_available():
                imgs = imgs.cuda()
                targets = targets.cuda()
            outputs = tudui(imgs)
            loss = loss_fn(outputs, targets)
            total_test_loss += loss.item()
            accuracy = (outputs.argmax(1) == targets).sum()
            total_accuracy += accuracy.item()

    print("测试集的Loss:{}".format(total_test_loss))
    print("测试集的正确率:{}".format(total_accuracy / test_data_len))
    # writer.add_scalar('test_loss', total_test_loss, total_test_step)
    # writer.add_scalar('test_accuracy', total_accuracy / test_data_len, total_test_step)
    total_test_step += 1

    # torch.save(tudui, "tudui_{}.pth".format(epoch))
    # print("model saved")

# writer.close()

方法二：

device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
zzz = zzz.to(device)

测试代码：

"""
 @Author: LiJie
 @FileName: train_gpu2.py
 @DateTime: 2024/7/16 下午8:23
"""
"""
 @Author: LiJie
 @FileName: train.py
 @DateTime: 2024/7/16 下午7:21
"""
import torch
from torch import nn
import torchvision
from torch.utils.data import DataLoader
from torch.utils.tensorboard import SummaryWriter

from model import *

# 定义训练设备
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")

# 数据集
train_set = torchvision.datasets.CIFAR10(root='../data2', train=True, download=True,
                                         transform=torchvision.transforms.ToTensor())
test_set = torchvision.datasets.CIFAR10(root='../data2', train=False, download=True,
                                        transform=torchvision.transforms.ToTensor())

# 数据集长度
train_data_len = len(train_set)
test_data_len = len(test_set)
print("训练数据集长度:{}".format(train_data_len))
print("测试数据集长度:{}".format(test_data_len))

# DataLoader加载数据集
train_dataloader = DataLoader(train_set, batch_size=64, shuffle=True)
test_dataloader = DataLoader(test_set, batch_size=64, shuffle=True)

# 创建网络模型
tudui = Tudui()
tudui = tudui.to(device)

# 损失函数
loss_fn = nn.CrossEntropyLoss()
loss_fn = loss_fn.to(device)

# 优化器
learning_rate = 0.01    # 学习率
optimizer = torch.optim.SGD(tudui.parameters(), lr=learning_rate)

# 训练参数设置
total_train_step = 0    # 训练次数记录
total_test_step = 0     # 测试次数记录
epochs = 10              # 训练轮数

# 添加TensorBoard，观察loss和accuracy
# writer = SummaryWriter('../logs_train')

# 开始训练
for epoch in range(epochs):
    print("==========epoch:{}==========".format(epoch))

    # 训练步骤
    tudui.train()
    for data in train_dataloader:
        imgs, targets = data
        imgs, targets = imgs.to(device), targets.to(device)
        outputs = tudui(imgs)
        loss = loss_fn(outputs, targets)

        # 优化器优化模型
        optimizer.zero_grad()
        loss.backward()
        optimizer.step()

        total_train_step += 1
        if total_train_step % 100 == 0:
            print("训练次数:{}，Loss:{}".format(total_train_step, loss.item()))
            # writer.add_scalar('train_loss', loss.item(), total_train_step)

    # 测试步骤
    tudui.eval()
    total_test_loss = 0
    total_accuracy = 0
    with torch.no_grad():
        for data in test_dataloader:
            imgs, targets = data
            imgs, targets = imgs.to(device), targets.to(device)
            outputs = tudui(imgs)
            loss = loss_fn(outputs, targets)
            total_test_loss += loss.item()
            accuracy = (outputs.argmax(1) == targets).sum()
            total_accuracy += accuracy.item()

    print("测试集的Loss:{}".format(total_test_loss))
    print("测试集的正确率:{}".format(total_accuracy/test_data_len))
    # writer.add_scalar('test_loss', total_test_loss, total_test_step)
    # writer.add_scalar('test_accuracy', total_accuracy/test_data_len, total_test_step)
    total_test_step += 1

    torch.save(tudui, "tudui_{}.pth".format(epoch))
    print("model saved")

# writer.close()

______to be update

暂无评论

发送评论 编辑评论 


				

			

						

				

					

						

							

								
							

							
						

					

				

				

					

						

							

								
							

							
						

					

				

				

					

						

							

								
							

							
							
													

					

				

			

			
			

				

					
				

			

				

											

							
							
						

																

							
							
						

																

							
							
						

										
					
											
						

	

		
|´・ω・)ノ
ヾ(≧∇≦*)ゝ
(☆ω☆)
（╯‵□′）╯︵┴─┴
￣﹃￣
(/ω＼)
∠( ᐛ 」∠)＿
(๑•̀ㅁ•́ฅ)
→_→
୧(๑•̀⌄•́๑)૭
٩(ˊᗜˋ*)و
(ノ°ο°)ノ
(´இ皿இ｀)
⌇●﹏●⌇
(ฅ´ω`ฅ)
(╯°A°)╯︵○○○
φ(￣∇￣o)
ヾ(´･ ･｀｡)ノ"
( ง ᵒ̌皿ᵒ̌)ง⁼³₌₃
(ó﹏ò｡)
Σ(っ °Д °;)っ
( ,,´･ω･)ﾉ"(´っω･｀｡)
╮(╯▽╰)╭
o(*////▽////*)q
＞﹏＜
( ๑´•ω•) "(ㆆᴗㆆ)
😂
😀
😅
😊
🙂
🙃
😌
😍
😘
😜
😝
😏
😒
🙄
😳
😡
😔
😫
😱
😭
💩
👻
🙌
🖕
👍
👫
👬
👭
🌚
🌝
🙈
💊
😶
🙏
🍦
🍉
😣
Source: github.com/k4yt3x/flowerhd
	

	

		
颜文字
Emoji
小恐龙
花!
	


									

			

			
			
		

	






上一篇
下一篇 

                    

                    
			        推荐文章
		            

		            

							
argparse module