DigitalRecognizer/main.py

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import torch
import torch.nn.functional as F
import torch.optim as optim
import torch.nn as nn
from torch.utils.data import DataLoader, TensorDataset
from tqdm import tqdm

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

losses = []

data = pd.read_csv("./mchar_train/train.csv")
data = np.array(data)
labels = data[:, 0]
data = data[:, 1:]
data = np.resize(data, (42000, 28, 28))
data = torch.from_numpy(data).float().to(device)
labels = torch.from_numpy(labels).to(device)


class BasicBlock(nn.Module):
    expansion = 1

    def __init__(self, in_planes, out_planes, stride=1):
        super(BasicBlock, self).__init__()
        self.conv1 = nn.Conv2d(in_planes, out_planes, kernel_size=3, stride=stride, padding=1, bias=False)
        self.bn1 = nn.BatchNorm2d(out_planes)
        self.conv2 = nn.Conv2d(out_planes, out_planes, kernel_size=3, stride=1, padding=1, bias=False)
        self.bn2 = nn.BatchNorm2d(out_planes)
        self.relu = nn.ReLU(inplace=False)
        self.shortcut = nn.Sequential()
        if stride != 1 or in_planes != self.expansion * out_planes:
            self.shortcut = nn.Sequential(
                nn.Conv2d(in_planes, self.expansion * out_planes, kernel_size=1, stride=stride, bias=False),
                nn.BatchNorm2d(self.expansion * out_planes)
            )

    def forward(self, x):
        out = self.relu(self.bn1(self.conv1(x)))
        out = self.relu(self.bn2(out))
        out = out + self.shortcut(x)
        out = F.relu(out)
        return out


class ResNet(nn.Module):
    # num_classes = 10 十分类问题
    def __init__(self, block, num_blocks, num_classes=10):
        super(ResNet, self).__init__()
        self.in_planes = 64

        self.conv1 = nn.Conv2d(1, 64, kernel_size=3, stride=1, padding=1, bias=False)
        self.bn1 = nn.BatchNorm2d(64)
        # 四个残差结构
        self.layer1 = self._make_layer(block, 64, num_blocks[0], stride=1)
        self.layer2 = self._make_layer(block, 128, num_blocks[1], stride=2)
        self.layer3 = self._make_layer(block, 256, num_blocks[2], stride=2)
        self.layer4 = self._make_layer(block, 512, num_blocks[3], stride=2)

        self.linear = nn.Linear(512 * block.expansion, num_classes)
        # 激活函数 relu = max(0, x)
        self.relu = nn.ReLU(inplace=False)

    def _make_layer(self, block, planes, num_blocks, stride):
        strides = [stride] + [1] * (num_blocks - 1)
        layers = []
        for stride in strides:
            layers.append(block(self.in_planes, planes, stride))
            self.in_planes = planes * block.expansion
        return nn.Sequential(*layers)

    def forward(self, x):
        out = self.bn1(self.conv1(x))
        out = self.layer1(out)
        out = self.layer2(out)
        out = self.layer3(out)
        out = self.layer4(out)
        # 平均池化
        out = F.avg_pool2d(out, 4)
        out = out.view(out.size(0), -1)
        out = self.linear(out)
        return out


model = ResNet(BasicBlock, [2, 2, 2, 2])
model.to(device)
criterion = nn.CrossEntropyLoss()
optimizer = optim.Adam(model.parameters(), lr=0.001)

data = torch.unsqueeze(data, 1)
dataset = TensorDataset(data, labels)
data_loader = DataLoader(dataset, batch_size=400, shuffle=True)

epochs = 10

loss = 0
for epoch in tqdm(range(epochs), desc=f'Training Progress', leave=False):
    for inputs, labels in data_loader:
        outputs = model(inputs)
        # 梯度
        loss = criterion(outputs, labels)
        losses.append(loss.item())
        optimizer.zero_grad()

        loss.backward()
        optimizer.step()

    print(f'Epoch [{epoch + 1}/{epochs}], Loss: {float(loss.item()):.4f}')
np.savetxt('loss.csv', losses, delimiter=',')
y = losses
y = np.array(y)
x = np.arange(1, len(y) + 1)
plt.xlabel("iteration")
plt.ylabel("loss")
plt.plot(x, y)
plt.show()
torch.save(model, 'model_name.pth')

device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
model = torch.load('./model_name.pth', map_location=torch.device(device))
model.eval()  # 设置为评估模式

data = pd.read_csv("./mchar_train/test.csv")
data = np.array(data)
data = torch.from_numpy(data).float().to(device)
data = data.resize(28000, 28, 28)
data = torch.unsqueeze(data, 1)
# 使用模型进行预测
ans = np.zeros(28000)
for i in tqdm(range(len(data)), desc='Predicting Progress', leave=False):
    d = torch.unsqueeze(data[i], 0)
    with torch.no_grad():
        output = model(d)
    predicted_class = torch.argmax(output, dim=1).item()
    ans[i] = predicted_class
np.savetxt('sample.csv', ans, delimiter=",")