pytorch建模的一般步骤

Pytorch的建模一般步骤

1.导入必要的库

2.准备数据

3.定义数据集类（可选）

4.加载数据

5.定义模型

6.定义损失函数和优化器

7.训练模型

8.评估模型

9.保存和加载模型

10.使用模型进行推理

import torch.nn.functional as F
import torch.nn as nn
import torch
from torchvision import datasets,transforms
import os
import torch.optim as optim

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

有无GPU的对比

#有GPU版
# 定义模型
class SimpleModel(nn.Module):
    def __init__(self):
        super(SimpleModel, self).__init__()
        self.fc1 = nn.Linear(10, 32)
        self.relu = nn.ReLU()
        self.fc2 = nn.Linear(32, 1)

    def forward(self, x):
        x = self.fc1(x)
        x = self.relu(x)
        x = self.fc2(x)
        return x

# 创建模型实例
model = SimpleModel().to(device)

# 定义损失函数和优化器
criterion = nn.MSELoss()
optimizer = optim.Adam(model.parameters(), lr=0.001)

# 生成一些数据
inputs = torch.randn(64, 10).to(device)
targets = torch.randn(64, 1).to(device)

# 训练循环
for epoch in range(100):
    # 前向传播
    outputs = model(inputs)
    loss = criterion(outputs, targets)

    # 反向传播和优化
    optimizer.zero_grad()
    loss.backward()
    optimizer.step()

    if (epoch + 1) % 10 == 0:
        print(f'Epoch [{epoch + 1}/100], Loss: {loss.item():.4f}')

Epoch [10/100], Loss: 0.9730
Epoch [20/100], Loss: 0.9241
Epoch [30/100], Loss: 0.8880
Epoch [40/100], Loss: 0.8533
Epoch [50/100], Loss: 0.8202
Epoch [60/100], Loss: 0.7875
Epoch [70/100], Loss: 0.7552
Epoch [80/100], Loss: 0.7235
Epoch [90/100], Loss: 0.6912
Epoch [100/100], Loss: 0.6584

#GPU版
# 定义模型
class SimpleModel_cpu(nn.Module):
    def __init__(self):
        super(SimpleModel_cpu, self).__init__()
        self.fc1 = nn.Linear(10, 32)
        self.relu = nn.ReLU()
        self.fc2 = nn.Linear(32, 1)

    def forward(self, x):
        x = self.fc1(x)
        x = self.relu(x)
        x = self.fc2(x)
        return x

# 创建模型实例
model_cpu= SimpleModel_cpu()

# 定义损失函数和优化器
criterion = nn.MSELoss()
optimizer = optim.Adam(model.parameters(), lr=0.001)

# 生成一些数据
inputs = torch.randn(64, 10)
targets = torch.randn(64, 1)

# 训练循环
for epoch in range(100):
    # 前向传播
    outputs = model_cpu(inputs)
    loss = criterion(outputs, targets)

    # 反向传播和优化
    optimizer.zero_grad()
    loss.backward()
    optimizer.step()

    if (epoch + 1) % 10 == 0:
        print(f'Epoch [{epoch + 1}/100], Loss: {loss.item():.4f}')

Epoch [10/100], Loss: 0.8705
Epoch [20/100], Loss: 0.8705
Epoch [30/100], Loss: 0.8705
Epoch [40/100], Loss: 0.8705
Epoch [50/100], Loss: 0.8705
Epoch [60/100], Loss: 0.8705
Epoch [70/100], Loss: 0.8705
Epoch [80/100], Loss: 0.8705
Epoch [90/100], Loss: 0.8705
Epoch [100/100], Loss: 0.8705

torch.save(model.state_dict(),"samplenn.pth")

model=SimpleModel()
model.load_state_dict(torch.load("./samplenn.pth"))

<All keys matched successfully>

model.eval()

SimpleModel(
  (fc1): Linear(in_features=10, out_features=32, bias=True)
  (relu): ReLU()
  (fc2): Linear(in_features=32, out_features=1, bias=True)
)