Pytorch从零开始实战07

Pytorch从零开始实战——咖啡豆识别

本系列来源于365天深度学习训练营

原作者K同学

环境准备

本文基于Jupyter notebook，使用Python3.8，Pytorch2.0.1+cu118，torchvision0.15.2，需读者自行配置好环境且有一些深度学习理论基础。本次实验的目的是手写VGG，并且测试多GPU。
第一步，导入常用包

import torch
import torch.nn as nn
import matplotlib.pyplot as plt
import torchvision
import torchvision.transforms as transforms
import torchvision.datasets as datasets
import torch.nn.functional as F
import random
from time import time
import numpy as np
import pandas as pd
import datetime
import gc
import os
import copy
os.environ['KMP_DUPLICATE_LIB_OK']='True'  # 用于避免jupyter环境突然关闭
torch.backends.cudnn.benchmark=True  # 用于加速GPU运算的代码

设置随机数种子

torch.manual_seed(428)
torch.cuda.manual_seed(428)
torch.cuda.manual_seed_all(428)
random.seed(428)
np.random.seed(428)

创建设备对象，并且查看GPU数量

device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
device, torch.cuda.device_count()

数据集

本次使用的数据集是咖啡豆图片，它分为四个类别，Dark、Green、Light、Medium，一共有1200张图片，不同的类别存放在不同的文件夹中，文件夹名是类别名。
使用pathlib查看类别

import pathlib
data_dir = './data/beans'
data_dir = pathlib.Path(data_dir) # 转成pathlib.Path对象
data_paths = list(data_dir.glob('*')) 
classNames = [str(path).split("/")[2] for path in data_paths]
classNames # ['Dark', 'Green', 'Medium', 'Light']

使用transforms对数据集进行统一处理，并且根据文件夹名映射对应标签

train_transforms = transforms.Compose([transforms.Resize([224, 224]),transforms.ToTensor(),transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]) # 标准化
])total_data = datasets.ImageFolder("./data/beans/", transform=train_transforms)
total_data.class_to_idx # {'Dark': 0, 'Green': 1, 'Light': 2, 'Medium': 3}

随机查看5张图片

def plotsample(data):fig, axs = plt.subplots(1, 5, figsize=(10, 10)) #建立子图for i in range(5):num = random.randint(0, len(data) - 1) #首先选取随机数，随机选取五次#抽取数据中对应的图像对象，make_grid函数可将任意格式的图像的通道数升为3，而不改变图像原始的数据#而展示图像用的imshow函数最常见的输入格式也是3通道npimg = torchvision.utils.make_grid(data[num][0]).numpy()nplabel = data[num][1] #提取标签 #将图像由(3, weight, height)转化为(weight, height, 3)，并放入imshow函数中读取axs[i].imshow(np.transpose(npimg, (1, 2, 0))) axs[i].set_title(nplabel) #给每个子图加上标签axs[i].axis("off") #消除每个子图的坐标轴plotsample(total_data)

根据8比2划分数据集和测试集，并且利用DataLoader划分批次和随机打乱

train_size = int(0.8 * len(total_data))
test_size  = len(total_data) - train_size
train_ds, test_ds = torch.utils.data.random_split(total_data, [train_size, test_size])batch_size = 32
train_dl = torch.utils.data.DataLoader(train_ds,batch_size=batch_size,shuffle=True,)
test_dl = torch.utils.data.DataLoader(test_ds,batch_size=batch_size,shuffle=True,)len(train_dl.dataset), len(test_dl.dataset) # (960, 240)

模型选择

本次实验使用VGG16，模型如下

class Model(nn.Module):def __init__(self):super().__init__()self.block1 = nn.Sequential(nn.Conv2d(3, 64, kernel_size=3, stride=1, padding=1),nn.ReLU(),nn.Conv2d(64, 64, kernel_size=3, stride=1, padding=1),nn.ReLU(),nn.MaxPool2d(2))self.block2 = nn.Sequential(nn.Conv2d(64, 128, kernel_size=3, stride=1, padding=1),nn.ReLU(),nn.Conv2d(128, 128, kernel_size=3, stride=1, padding=1),nn.ReLU(),nn.MaxPool2d(2))self.block3 = nn.Sequential(nn.Conv2d(128, 256, kernel_size=3, stride=1, padding=1),nn.ReLU(),nn.Conv2d(256, 256, kernel_size=3, stride=1, padding=1),nn.ReLU(),nn.Conv2d(256, 256, kernel_size=3, stride=1, padding=1),nn.ReLU(),nn.MaxPool2d(2))self.block4 = nn.Sequential(nn.Conv2d(256, 512, kernel_size=3, stride=1, padding=1),nn.ReLU(),nn.Conv2d(512, 512, kernel_size=3, stride=1, padding=1),nn.ReLU(),nn.Conv2d(512, 512, kernel_size=3, stride=1, padding=1),nn.ReLU(),nn.MaxPool2d(2))self.block5 = nn.Sequential(nn.Conv2d(512, 512, kernel_size=3, stride=1, padding=1),nn.ReLU(),nn.Conv2d(512, 512, kernel_size=3, stride=1, padding=1),nn.ReLU(),nn.Conv2d(512, 512, kernel_size=3, stride=1, padding=1),nn.ReLU(),nn.MaxPool2d(2))self.fc = nn.Sequential(nn.Linear(7 * 7 * 512, 4096),nn.ReLU(),nn.Linear(4096, 4096),nn.ReLU(),nn.Linear(4096, len(classNames)))def forward(self, x):x = self.block1(x)x = self.block2(x)x = self.block3(x)x = self.block4(x)x = self.block5(x)x = x.view(-1, 7 * 7 * 512)x = self.fc(x)return x

使用summary查看模型结构，并且将模型转成多GPU并行运算的模型

from torchsummary import summary
# 将模型转移到GPU中
model = Model()
model = model.to(device)
if torch.cuda.device_count() > 1:  # 检查电脑是否有多块GPUprint(f"Let's use {torch.cuda.device_count()} GPUs!")model = nn.DataParallel(model)  # 将模型对象转变为多GPU并行运算的模型summary(model, input_size=(3, 224, 224))

训练

定义训练函数

def train(dataloader, model, loss_fn, opt):size = len(dataloader.dataset)num_batches = len(dataloader)train_acc, train_loss = 0, 0for X, y in dataloader:X, y = X.to(device), y.to(device)pred = model(X)loss = loss_fn(pred, y)opt.zero_grad()loss.backward()opt.step()train_acc += (pred.argmax(1) == y).type(torch.float).sum().item()train_loss += loss.item()train_acc /= sizetrain_loss /= num_batchesreturn train_acc, train_loss

定义测试函数

def test(dataloader, model, loss_fn):size = len(dataloader.dataset)num_batches = len(dataloader)test_acc, test_loss = 0, 0with torch.no_grad():for X, y in dataloader:X, y = X.to(device), y.to(device)pred = model(X)loss = loss_fn(pred, y)test_acc += (pred.argmax(1) == y).type(torch.float).sum().item()test_loss += loss.item()test_acc /= sizetest_loss /= num_batchesreturn test_acc, test_loss

定义损失函数、优化算法、学习率，本次使用的是Adam优化算法

loss_fn = nn.CrossEntropyLoss()
learn_rate = 0.0001
opt = torch.optim.Adam(model.parameters(), lr=learn_rate)

开始训练，准确率还是非常高的

import time
epochs = 30
train_loss = []
train_acc = []
test_loss = []
test_acc = []T1 = time.time()best_acc = 0
best_model = 0for epoch in range(epochs):model.train()epoch_train_acc, epoch_train_loss = train(train_dl, model, loss_fn, opt)model.eval() # 确保模型不会进行训练操作epoch_test_acc, epoch_test_loss = test(test_dl, model, loss_fn)if epoch_test_acc > best_acc:best_acc = epoch_test_accbest_model = copy.deepcopy(model)train_acc.append(epoch_train_acc)train_loss.append(epoch_train_loss)test_acc.append(epoch_test_acc)test_loss.append(epoch_test_loss)print("epoch:%d, train_acc:%.1f%%, train_loss:%.3f, test_acc:%.1f%%, test_loss:%.3f"% (epoch + 1, epoch_train_acc * 100, epoch_train_loss, epoch_test_acc * 100, epoch_test_loss))T2 = time.time()
print('程序运行时间:%s秒' % (T2 - T1))PATH = './best_model.pth'  # 保存的参数文件名
if best_model is not None:torch.save(best_model.state_dict(), PATH)print('保存最佳模型')
print("Done")

模型可视化

使用matplotlib可视化训练、测试过程

import warnings
warnings.filterwarnings("ignore")               #忽略警告信息
plt.rcParams['font.sans-serif']    = ['SimHei'] # 用来正常显示中文标签
plt.rcParams['axes.unicode_minus'] = False      # 用来正常显示负号
plt.rcParams['figure.dpi']         = 100        #分辨率epochs_range = range(epochs)plt.figure(figsize=(12, 3))
plt.subplot(1, 2, 1)plt.plot(epochs_range, train_acc, label='Training Accuracy')
plt.plot(epochs_range, test_acc, label='Test Accuracy')
plt.legend(loc='lower right')
plt.title('Training and Validation Accuracy')plt.subplot(1, 2, 2)
plt.plot(epochs_range, train_loss, label='Training Loss')
plt.plot(epochs_range, test_loss, label='Test Loss')
plt.legend(loc='upper right')
plt.title('Training and Validation Loss')
plt.show()

模型预测

定义模型预测函数

from PIL import Image classes = list(total_data.class_to_idx)def predict_one_image(image_path, model, transform, classes):test_img = Image.open(image_path).convert('RGB')plt.imshow(test_img)  # 展示预测的图片test_img = transform(test_img)img = test_img.to(device).unsqueeze(0)model.eval()output = model(img)_,pred = torch.max(output,1)pred_class = classes[pred]print(f'预测结果是：{pred_class}') 

开始单张图片预测

predict_one_image(image_path='./data/beans/Dark/dark (1).png', model=model, transform=train_transforms, classes=classes) # 预测结果是：Dark

查看最优的模型的准确率和损失

best_model.eval()
epoch_test_acc, epoch_test_loss = test(test_dl, best_model, loss_fn)
epoch_test_acc, epoch_test_loss # (0.9916666666666667, 0.0399394309388299)

其他问题

本次实验又使用了单GPU，进行训练

# 单GPU
from torchsummary import summary
# 将模型转移到GPU中
model = Model()
model = model.to(device)

结果如下

总结

本次实验主要手写了经典网络架构VGG16，并且使用两张GPU和一张GPU进行实验，但惊奇的发现，一张GPU运行时间是164秒，两张GPU运行时间是318秒，明明算力提高了，反而训练时间更加慢了，经过资料的查询，大概原因是数据量很小，GPU之间传递数据占用时间相对大于加速运算时间，所以训练时间反而变长了。