From c53395f97a33da54e4dfc8e938691b0a926d378e Mon Sep 17 00:00:00 2001 From: luzhisheng Date: Fri, 21 Apr 2023 18:10:21 +0800 Subject: [PATCH] =?UTF-8?q?04=E7=AC=AC=E4=B8=80=E6=AC=A1=E8=AE=AD=E7=BB=83?= =?UTF-8?q?=E4=B8=8E=E6=89=8B=E5=86=99=E6=95=B0=E5=AD=97?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- 机器学习/04第一次训练与手写数字/__init__.py | 0 机器学习/04第一次训练与手写数字/推理.py | 47 ++++++++++++++ 机器学习/04第一次训练与手写数字/测试.py | 58 +++++++++++++++++ 机器学习/04第一次训练与手写数字/训练.py | 72 +++++++++++++++++++++ 机器学习/04第一次训练与手写数字/进度条.py | 7 ++ 机器学习/README.md | 3 + 6 files changed, 187 insertions(+) create mode 100644 机器学习/04第一次训练与手写数字/__init__.py create mode 100644 机器学习/04第一次训练与手写数字/推理.py create mode 100644 机器学习/04第一次训练与手写数字/测试.py create mode 100644 机器学习/04第一次训练与手写数字/训练.py create mode 100644 机器学习/04第一次训练与手写数字/进度条.py create mode 100644 机器学习/README.md diff --git a/机器学习/04第一次训练与手写数字/__init__.py b/机器学习/04第一次训练与手写数字/__init__.py new file mode 100644 index 0000000..e69de29 diff --git a/机器学习/04第一次训练与手写数字/推理.py b/机器学习/04第一次训练与手写数字/推理.py new file mode 100644 index 0000000..320d0f9 --- /dev/null +++ b/机器学习/04第一次训练与手写数字/推理.py @@ -0,0 +1,47 @@ +from torchvision import transforms +from torch import nn +import torch +from PIL import Image + + +class MnistModel(nn.Module): + def __init__(self): + super(MnistModel, self).__init__() + self.fc1 = nn.Linear(1 * 28 * 28, 100) # 最终为什么是 10，因为手写数字识别最终是 10分类的，分类任务中有多少，就分几类。 0-9 + self.relu = nn.ReLU() + self.fc2 = nn.Linear(100, 10) + + def forward(self, image): + image_viwed = image.view(-1, 1 * 28 * 28) # 此处需要拍平 + out_1 = self.fc1(image_viwed) + fc1 = self.relu(out_1) + out_2 = self.fc2(fc1) + return out_2 + + +model = MnistModel() +model.load_state_dict(torch.load("./models/model.pkl")) + +my_transforms = transforms.Compose( + [ + transforms.ToTensor(), + transforms.Normalize(mean=(0.1307,), std=(0.3081,)) + ] +) + +image = Image.open('./img/test.jpg') + +my_transforms = transforms.Compose( + [ + transforms.Grayscale(1), + transforms.ToTensor(), + transforms.Normalize(mean=(0.1307,), std=(0.3081,)) + ] +) +image = my_transforms(image) + +model.eval() +with torch.no_grad(): + output = model(image) + result = output.max(dim=1).indices + print(result) diff --git a/机器学习/04第一次训练与手写数字/测试.py b/机器学习/04第一次训练与手写数字/测试.py new file mode 100644 index 0000000..3f4889e --- /dev/null +++ b/机器学习/04第一次训练与手写数字/测试.py @@ -0,0 +1,58 @@ +from torchvision.datasets import MNIST +from torchvision import transforms +from torch.utils.data import DataLoader +from torch import nn +from tqdm import tqdm +import torch +import numpy as np +import os + + +class MnistModel(nn.Module): + def __init__(self): + super(MnistModel, self).__init__() + self.fc1 = nn.Linear(1 * 28 * 28, 100) # 最终为什么是 10，因为手写数字识别最终是 10分类的，分类任务中有多少，就分几类。 0-9 + self.relu = nn.ReLU() + self.fc2 = nn.Linear(100, 10) + + def forward(self, image): + image_viwed = image.view(-1, 1 * 28 * 28) # 此处需要拍平 + out_1 = self.fc1(image_viwed) + fc1 = self.relu(out_1) + out_2 = self.fc2(fc1) + return out_2 + + +def test_success(): + # 实例化模型 + total_loss = [] + model = MnistModel() + if os.path.exists("./models/model.pkl"): + model.load_state_dict(torch.load("./models/model.pkl")) + loss_function = nn.CrossEntropyLoss() + my_transforms = transforms.Compose( + [ + transforms.ToTensor(), + transforms.Normalize(mean=(0.1307,), std=(0.3081,)) + ] + ) + mnist_train = MNIST(root="../MNIST_data", train=False, download=True, transform=my_transforms) + dataloader = DataLoader(mnist_train, batch_size=8, shuffle=True) + dataloader = tqdm(dataloader, total=len(dataloader)) + + succeed = [] + + model.eval() + with torch.no_grad(): + for images, labels in dataloader: + # 获取结果 + output = model(images) + result = output.max(dim=1).indices + # print(labels) + # print(result) + succeed.append(result.eq(labels).float().mean().item()) + # 通过结果计算损失 + loss = loss_function(output, labels) + total_loss.append(loss.item()) + print(np.mean(total_loss)) + return np.mean(succeed) diff --git a/机器学习/04第一次训练与手写数字/训练.py b/机器学习/04第一次训练与手写数字/训练.py new file mode 100644 index 0000000..75934eb --- /dev/null +++ b/机器学习/04第一次训练与手写数字/训练.py @@ -0,0 +1,72 @@ +from torch import save, load +from torchvision.datasets import MNIST +from torchvision import transforms +from torch.utils.data import DataLoader +from torch import nn +from torch import optim +from tqdm import tqdm +import 测试 +import os +import numpy as np + + +class MnistModel(nn.Module): + def __init__(self): + super(MnistModel, self).__init__() + self.fc1 = nn.Linear(1 * 28 * 28, 100) # 最终为什么是 10，因为手写数字识别最终是 10分类的，分类任务中有多少，就分几类。 0-9 + self.relu = nn.ReLU() + self.fc2 = nn.Linear(100, 10) + + def forward(self, image): + image_viwed = image.view(-1, 1 * 28 * 28) # 此处需要拍平 + out_1 = self.fc1(image_viwed) + fc1 = self.relu(out_1) + out_2 = self.fc2(fc1) + return out_2 + + +# 实例化模型 +model = MnistModel() +optimizer = optim.Adam(model.parameters()) + +# 加载已经训练好的模型和优化器继续进行训练 +if os.path.exists('./models/model.pkl'): + model.load_state_dict(load("./models/model.pkl")) + optimizer.load_state_dict(load("./models/optimizer.pkl")) + +loss_function = nn.CrossEntropyLoss() +my_transforms = transforms.Compose( + [ + transforms.ToTensor(), + transforms.Normalize(mean=(0.1307,), std=(0.3081,)) + ] +) +mnist_train = MNIST(root="../MNIST_data", train=True, download=True, transform=my_transforms) + + +def train(epoch): + total_loss = [] + dataloader = DataLoader(mnist_train, batch_size=8, shuffle=True) + dataloader = tqdm(dataloader, total=len(dataloader)) + model.train() + for images, labels in dataloader: + # 梯度置0 + optimizer.zero_grad() + # 前向传播 + output = model(images) + # 通过结果计算损失 + loss = loss_function(output, labels) + total_loss.append(loss.item()) + # 反向传播 + loss.backward() + # 优化器更新 + optimizer.step() + + save(model.state_dict(), './models/model.pkl') + save(optimizer.state_dict(), './models/optimizer.pkl') + # 打印一下训练成功率 + print('第{}个epoch，成功率, 损失为{}'.format(epoch, np.mean(total_loss)), 测试.test_success()) + + +for i in range(10): + train(i) diff --git a/机器学习/04第一次训练与手写数字/进度条.py b/机器学习/04第一次训练与手写数字/进度条.py new file mode 100644 index 0000000..49da9ed --- /dev/null +++ b/机器学习/04第一次训练与手写数字/进度条.py @@ -0,0 +1,7 @@ +import time +from tqdm import tqdm + + +c = tqdm(range(0, 1000), total=1000) +for i in c: + time.sleep(0.1) diff --git a/机器学习/README.md b/机器学习/README.md new file mode 100644 index 0000000..f47b5e2 --- /dev/null +++ b/机器学习/README.md @@ -0,0 +1,3 @@ +pytorch文档 + +https://pytorch.org/docs/stable/index.html \ No newline at end of file