PyTorch磁盘模型加载与使用：是否需模型定义？如何拆分推理代码？

核心问题解答 #

用state_dict保存的模型（也就是你代码里torch.save(model.state_dict(), "model.pth")的方式）必须依赖模型定义，因为state_dict本质只是模型各层参数权重的字典，不包含任何模型结构信息。你觉得预训练模型不用自己写定义，是因为像TorchVision这类库已经把模型类（比如resnet50）封装好了，调用相关API的时候，其实是内部帮你实例化了模型类并加载了预训练权重，本质还是依赖了模型定义。

拆分训练与推理文件的具体步骤 #

最规范的做法是把模型定义单独抽成一个文件，让训练和推理文件都导入这个定义，既保证结构一致，也符合模块化开发习惯。

1. 单独创建模型定义文件 model_def.py

from torch import nn

class NeuralNetwork(nn.Module):
    def __init__(self):
        super().__init__()
        self.flatten = nn.Flatten()
        self.linear_relu_stack = nn.Sequential(
            nn.Linear(28*28, 512),
            nn.ReLU(),
            nn.Linear(512, 512),
            nn.ReLU(),
            nn.Linear(512, 10)
        )

    def forward(self, x):
        x = self.flatten(x)
        logits = self.linear_relu_stack(x)
        return logits

2. 训练文件 train.py

只保留训练和保存模型的逻辑，导入上面的模型类：

import torch
from torch import nn
from torch.utils.data import DataLoader
from torchvision import datasets
from torchvision.transforms import ToTensor
from model_def import NeuralNetwork  # 导入模型定义

# 下载数据、创建DataLoader
training_data = datasets.FashionMNIST(
    root="data",
    train=True,
    download=True,
    transform=ToTensor(),
)

test_data = datasets.FashionMNIST(
    root="data",
    train=False,
    download=True,
    transform=ToTensor(),
)

batch_size = 64
train_dataloader = DataLoader(training_data, batch_size=batch_size)
test_dataloader = DataLoader(test_data, batch_size=batch_size)

# 选择设备
device = (
    "cuda"
    if torch.cuda.is_available()
    else "mps"
    if torch.backends.mps.is_available()
    else "cpu"
)
print(f"Using {device} device")

# 初始化模型、损失函数、优化器
model = NeuralNetwork().to(device)
loss_fn = nn.CrossEntropyLoss()
optimizer = torch.optim.SGD(model.parameters(), lr=1e-3)

# 训练函数
def train(dataloader, model, loss_fn, optimizer):
    size = len(dataloader.dataset)
    model.train()
    for batch, (X, y) in enumerate(dataloader):
        X, y = X.to(device), y.to(device)
        pred = model(X)
        loss = loss_fn(pred, y)
        loss.backward()
        optimizer.step()
        optimizer.zero_grad()

        if batch % 100 == 0:
            loss, current = loss.item(), (batch + 1) * len(X)
            print(f"loss: {loss:>7f}  [{current:>5d}/{size:>5d}]")

# 测试函数
def test(dataloader, model, loss_fn):
    size = len(dataloader.dataset)
    num_batches = len(dataloader)
    model.eval()
    test_loss, correct = 0, 0
    with torch.no_grad():
        for X, y in dataloader:
            X, y = X.to(device), y.to(device)
            pred = model(X)
            test_loss += loss_fn(pred, y).item()
            correct += (pred.argmax(1) == y).type(torch.float).sum().item()
    test_loss /= num_batches
    correct /= size
    print(f"Test Error: \n Accuracy: {(100*correct):>0.1f}%, Avg loss: {test_loss:>8f} \n")

# 开始训练
epochs = 5
for t in range(epochs):
    print(f"Epoch {t+1}\n-------------------------------")
    train(train_dataloader, model, loss_fn, optimizer)
    test(test_dataloader, model, loss_fn)
print("Done!")

# 保存模型权重
torch.save(model.state_dict(), "model.pth")
print("Saved PyTorch Model State to model.pth")

3. 推理文件 infer.py

只保留加载模型和推理的逻辑，同样导入模型定义：

import torch
from torchvision import datasets
from torchvision.transforms import ToTensor
from model_def import NeuralNetwork  # 导入模型定义

# 选择设备
device = (
    "cuda"
    if torch.cuda.is_available()
    else "mps"
    if torch.backends.mps.is_available()
    else "cpu"
)

# 加载模型
model = NeuralNetwork().to(device)
model.load_state_dict(torch.load("model.pth"))

# 类别定义
classes = [
    "T-shirt/top",
    "Trouser",
    "Pullover",
    "Dress",
    "Coat",
    "Sandal",
    "Shirt",
    "Sneaker",
    "Bag",
    "Ankle boot",
]

# 加载测试数据（也可以换成自己的输入图片）
test_data = datasets.FashionMNIST(
    root="data",
    train=False,
    download=True,
    transform=ToTensor(),
)

# 推理
model.eval()
x, y = test_data[0][0], test_data[0][1]
with torch.no_grad():
    x = x.to(device)
    pred = model(x)
    predicted, actual = classes[pred[0].argmax(0)], classes[y]
    print(f'Predicted: "{predicted}", Actual: "{actual}"')

可选方案：保存整个模型（不推荐） #

如果你确实不想依赖模型定义文件，可以直接保存整个模型对象：

# 训练时保存
torch.save(model, "full_model.pth")

# 推理时直接加载，不需要模型定义
model = torch.load("full_model.pth").to(device)

但这种方式不推荐，因为序列化的模型和PyTorch版本、依赖库版本绑定紧密，换环境很容易加载失败，而且无法灵活修改模型结构。

内容的提问来源于stack exchange，提问作者Tmp