完整的模型训练套路
train.py
import torch
import torchvision
from torch.utils.data import DataLoader
# 引入自定义的网络模型
from torch.utils.tensorboard import SummaryWriter
from model import *
# 准备数据集
train_data = torchvision.datasets.CIFAR10(root="dataset_transform", train=True, transform=torchvision.transforms.ToTensor(),
download=True)
test_data = torchvision.datasets.CIFAR10(root="dataset_transform", train=False, transform=torchvision.transforms.ToTensor(),
download=True)
# length 长度 获取数据集长度
train_data_size = len(train_data)
test_data_size = len(test_data)
print("训练数据集的长度为：{}".format(train_data_size))
print("测试数据集的长度为：{}".format(test_data_size))
# 利用 DataLoader 来加载数据集
train_dataloader = DataLoader(train_data, batch_size=64)
test_dataloader = DataLoader(test_data, batch_size=64)
# 搭建神经网络：
# 一般情况下我们会把网络放到单独的python文件里，通常命名为model.py，然后再本文件头部引入就可以了
# class Zrf(nn.Module):
# def __init__(self):
# super(Zrf, self).__init__()
# # Sequential 序列
# self.model = Sequential(
# # padding=2 是根据输入输出的H，W计算出来的
# Conv2d(3, 32, 5, 1, padding=2), 输入通道，输出通道，卷积核尺寸，步长，padding要用公式算
# MaxPool2d(2),
# Conv2d(32, 32, 5, 1, padding=2),
# MaxPool2d(2),
# Conv2d(32, 64, 5, 1, padding=2),
# MaxPool2d(2),
# Flatten(),
# Linear(1024, 64),
# Linear(64, 10)
# )
#
# def forward(self, x):
# x = self.model(x)
# return x
# 创建网络模型
zrf = Zrf()
# 损失函数
loss_fn = nn.CrossEntropyLoss()
# 优化器
# learning_rate = 0.01
learning_rate = 1e-2
optimizer = torch.optim.SGD(zrf.parameters(), lr=learning_rate)
# 设置训练网络的一些参数
# 记录训练的次数
total_train_step = 0
# 记录测试的次数
total_test_step = 0
# 训练的轮数
epoch = 10
# 添加tensorboard
writer = SummaryWriter("../log_train")
for i in range(epoch):
print("--------第 {} 轮训练开始--------".format(i+1))
# 训练步骤开始
zrf.train() # 设置训练模式（本模型中这一行可以不写）
for data in train_dataloader:
imgs, targets = data
outputs = zrf(imgs)
loss = loss_fn(outputs, targets)
# 优化器优化模型
optimizer.zero_grad() # 在进行反向传播来计算梯度时，要先将梯度置为0，防止之前计算出来的梯度的影响
loss.backward() # 计算梯度
optimizer.step() # 根据梯度对卷积核参数进行调优
total_train_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)
# 为了看模型有没有训练好，所以在训练完一轮之后，在测试数据集上进行测试
# 以测试数据集上的损失来判断
# 以下部分没有梯度，测试时不需要调优
# 测试步骤开始
zrf.eval() # 设置评估模式（本模型中这一行可以不写）
total_test_loss = 0
# 计算整体正确率
total_accuracy = 0
with torch.no_grad():
for data in test_dataloader:
imgs, targets = data
outputs = zrf(imgs)
loss = loss_fn(outputs, targets)
# 计算整体正确率
accuracy = (outputs.argmax(1) == targets).sum()
total_accuracy = total_accuracy + accuracy
total_test_loss = total_test_loss + loss.item()
print("整体测试集上的Loss:{}",format(total_test_loss))
print("整体测试集上的正确率:{}".format(total_accuracy/test_data_size))
total_test_step = total_test_step + 1
writer.add_scalar("test_loss", total_test_loss, total_test_step)
writer.add_scalar("test_accuracy", total_accuracy/test_data_size, total_test_step)
torch.save(zrf, "zrf_{}.pth".format(i))
# torch.save(zrf.state_dict(), "zrf_{}.pth".format(i))
print("模型已保存")
writer.close()ssssssssaaaassxcscwqmodel.py
import torch
from torch import nn
from torch.nn import Conv2d, MaxPool2d, Flatten, Linear, Sequential
# 搭建神经网络
class Zrf(nn.Module):
def __init__(self):
super(Zrf, self).__init__()
# Sequential 序列
self.model = Sequential(
# padding=2 是根据输入输出的H，W计算出来的
Conv2d(3, 32, 5, 1, padding=2),
MaxPool2d(2),
Conv2d(32, 32, 5, 1, padding=2),
MaxPool2d(2),
Conv2d(32, 64, 5, 1, padding=2),
MaxPool2d(2),
Flatten(),
Linear(1024, 64),
Linear(64, 10)
)
def forward(self, x):
x = self.model(x)
return x
if __name__ == '__main__':
# 一般在这里测试网络的正确性
zrf = Zrf()
input = torch.ones((64, 3, 32, 32)) # 64batch_size,3通道，32x32
output = zrf(input)
print(output.shape)关于正确率计算的一点说明
import torch
outputs = torch.tensor([[0.1, 0.2],
[0.3, 0.4]])
print(outputs.argmax(1)) # 1或0代表着方向，1是横向看
# tensor([1, 1]) 最大值是0.3 0.4
print(outputs.argmax(0)) # 0是纵向看
# tensor([1, 1]) 最大值是0.2 0.4
# outputs = torch.tensor([[0.1, 0.2],
# [0.05, 0.4]])
# print(outputs.argmax(0))
# # tensor([0, 1]) 最大值是0.1 0.4
preds = outputs.argmax(1)
targets = torch.tensor([0, 1])
print((preds == targets).sum())利用GPU进行训练train_gpu
train_gpu.py
第一种GPU训练方法
# 对模型，数据（输入、标注），损失函数的后面，加 .cuda()
import torch
import torchvision
from torch import nn
from torch.nn import Conv2d, MaxPool2d, Flatten, Linear, Sequential
from torch.utils.data import DataLoader
from torch.utils.tensorboard import SummaryWriter
import time
# 准备数据集
train_data = torchvision.datasets.CIFAR10(root="dataset_transform", train=True, transform=torchvision.transforms.ToTensor(),
download=True)
test_data = torchvision.datasets.CIFAR10(root="dataset_transform", train=False, transform=torchvision.transforms.ToTensor(),
download=True)
# length 长度 获取数据集长度
train_data_size = len(train_data)
test_data_size = len(test_data)
print("训练数据集的长度为：{}".format(train_data_size))
print("测试数据集的长度为：{}".format(test_data_size))
train_dataloader = DataLoader(train_data, batch_size=64)
test_dataloader = DataLoader(test_data, batch_size=64)
class Zrf(nn.Module):
def __init__(self):
super(Zrf, self).__init__()
# Sequential 序列
self.model = Sequential(
Conv2d(3, 32, 5, 1, padding=2),
MaxPool2d(2),
Conv2d(32, 32, 5, 1, padding=2),
MaxPool2d(2),
Conv2d(32, 64, 5, 1, padding=2),
MaxPool2d(2),
Flatten(),
Linear(1024, 64),
Linear(64, 10)
)
def forward(self, x):
x = self.model(x)
return x
# 创建网络模型
zrf = Zrf()
# -------------------利用GPU训练-------------------#
if torch.cuda.is_available():
zrf = zrf.cuda()
# 损失函数
loss_fn = nn.CrossEntropyLoss()
# -------------------利用GPU训练-------------------#
if torch.cuda.is_available():
loss_fn = loss_fn.cuda()
# 优化器
learning_rate = 1e-2
optimizer = torch.optim.SGD(zrf.parameters(), lr=learning_rate)
# 设置训练网络的一些参数
total_train_step = 0
total_test_step = 0
epoch = 10
# 添加tensorboard
writer = SummaryWriter("../log_train")
start_time = time.time()
for i in range(epoch):
print("--------第 {} 轮训练开始--------".format(i+1))
# 训练步骤开始
zrf.train()
for data in train_dataloader:
imgs, targets = data
# -------------------利用GPU训练-------------------#
if torch.cuda.is_available():
imgs = imgs.cuda()
targets = targets.cuda()
outputs = zrf(imgs)
loss = loss_fn(outputs, targets)
# 优化器优化模型
optimizer.zero_grad()
loss.backward()
optimizer.step()
total_train_step = total_train_step + 1
if total_train_step % 100 == 0:
end_time = time.time()
print(end_time - start_time)
print("训练次数：{}，Loss：{}".format(total_train_step, loss.item()))
writer.add_scalar("train_loss", loss.item(), total_train_step)
# 测试步骤开始
zrf.eval()
total_test_loss = 0
total_accuracy = 0
with torch.no_grad():
for data in test_dataloader:
imgs, targets = data
# -------------------利用GPU训练-------------------#
if torch.cuda.is_available():
imgs = imgs.cuda()
targets = targets.cuda()
outputs = zrf(imgs)
loss = loss_fn(outputs, targets)
accuracy = (outputs.argmax(1) == targets).sum()
total_accuracy = total_accuracy + accuracy
total_test_loss = total_test_loss + loss.item()
print("整体测试集上的Loss:{}",format(total_test_loss))
print("整体测试集上的正确率:{}".format(total_accuracy/test_data_size))
total_test_step = total_test_step + 1
writer.add_scalar("test_loss", total_test_loss, total_test_step)
writer.add_scalar("test_accuracy", total_accuracy/test_data_size, total_test_step)
torch.save(zrf, "zrf_{}.pth".format(i))
print("模型已保存")
writer.close()
第二种GPU训练方法
# .to(device)
# device = torch.device("cpu")
# torch.device("cuda")
# torch.device("cuda:0")
# torch.device("cuda:1")
import torch
import torchvision
from torch import nn
from torch.nn import Conv2d, MaxPool2d, Flatten, Linear, Sequential
from torch.utils.data import DataLoader
from torch.utils.tensorboard import SummaryWriter
import time
# 定义训练的设备
# device = torch.device("cpu")
# device = torch.device("cuda")
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
# 准备数据集
train_data = torchvision.datasets.CIFAR10(root="dataset_transform", train=True, transform=torchvision.transforms.ToTensor(),
download=True)
test_data = torchvision.datasets.CIFAR10(root="dataset_transform", train=False, transform=torchvision.transforms.ToTensor(),
download=True)
# length 长度 获取数据集长度
train_data_size = len(train_data)
test_data_size = len(test_data)
print("训练数据集的长度为：{}".format(train_data_size))
print("测试数据集的长度为：{}".format(test_data_size))
train_dataloader = DataLoader(train_data, batch_size=64)
test_dataloader = DataLoader(test_data, batch_size=64)
class Zrf(nn.Module):
def __init__(self):
super(Zrf, self).__init__()
# Sequential 序列
self.model = Sequential(
Conv2d(3, 32, 5, 1, padding=2),
MaxPool2d(2),
Conv2d(32, 32, 5, 1, padding=2),
MaxPool2d(2),
Conv2d(32, 64, 5, 1, padding=2),
MaxPool2d(2),
Flatten(),
Linear(1024, 64),
Linear(64, 10)
)
def forward(self, x):
x = self.model(x)
return x
# 创建网络模型
zrf = Zrf()
# -------------------利用GPU训练-------------------#
zrf.to(device) # 可以不重新赋值
# zrf = zrf.to(device)
# 损失函数
loss_fn = nn.CrossEntropyLoss()
# -------------------利用GPU训练-------------------#
loss_fn.to(device) # 可以不重新赋值
# loss_fn = loss_fn.to(device)
# 优化器
learning_rate = 1e-2
optimizer = torch.optim.SGD(zrf.parameters(), lr=learning_rate)
# 设置训练网络的一些参数
total_train_step = 0
total_test_step = 0
epoch = 10
# 添加tensorboard
writer = SummaryWriter("../log_train")
start_time = time.time()
for i in range(epoch):
print("--------第 {} 轮训练开始--------".format(i+1))
# 训练步骤开始
zrf.train()
for data in train_dataloader:
imgs, targets = data
# -------------------利用GPU训练-------------------#
# 必须重新赋值
imgs = imgs.to(device)
targets = targets.to(device)
outputs = zrf(imgs)
loss = loss_fn(outputs, targets)
# 优化器优化模型
optimizer.zero_grad()
loss.backward()
optimizer.step()
total_train_step = total_train_step + 1
if total_train_step % 100 == 0:
end_time = time.time()
print(end_time - start_time)
print("训练次数：{}，Loss：{}".format(total_train_step, loss.item()))
writer.add_scalar("train_loss", loss.item(), total_train_step)
# 测试步骤开始
zrf.eval()
total_test_loss = 0
total_accuracy = 0
with torch.no_grad():
for data in test_dataloader:
imgs, targets = data
# -------------------利用GPU训练-------------------#
imgs = imgs.to(device)
targets = targets.to(device)
outputs = zrf(imgs)
loss = loss_fn(outputs, targets)
accuracy = (outputs.argmax(1) == targets).sum()
total_accuracy = total_accuracy + accuracy
total_test_loss = total_test_loss + loss.item()
print("整体测试集上的Loss:{}",format(total_test_loss))
print("整体测试集上的正确率:{}".format(total_accuracy/test_data_size))
total_test_step = total_test_step + 1
writer.add_scalar("test_loss", total_test_loss, total_test_step)
writer.add_scalar("test_accuracy", total_accuracy/test_data_size, total_test_step)
torch.save(zrf, "zrf_{}.pth".format(i))
print("模型已保存")
writer.close()
利用GPU训练前一百次的时间： 4.680064678192139
没有GPU： 6.723153114318848
完整的模型验证套路
（测试、demo）利用已经训练好的模型，然后给他提供输入
