PyTorch Lightning搭建MLP模拟正弦波报缺少training_step方法如何解决

核心报错原因 #

你遇到的报错本质是类方法缩进错误，所有Lightning要求实现的方法都没有定义在MLP类的内部，导致实例化的模型对象缺失必要的成员方法。

MisconfigurationException: No training_step() method defined. Lightning Trainer expects as minimum a training_step(), train_dataloader() and configure_optimizers() to be defined.

所有需要修正的问题点 #

• 缩进错误：forward、train_dataloader、training_step等方法全部写在了MLP类的外部，不属于类成员，Lightning无法从模型实例中读取到这些方法。
• 缺失导入：代码中使用了TensorDataset，但没有从torch.utils.data导入该类，运行会触发NameError。
• 错误列名引用：你的数据集标签列名为y，但代码中写的是train_targets[['cnt']]，不存在的cnt列会触发KeyError。
• 激活函数拼写错误：PyTorch的ReLU激活函数为小写的torch.relu()，你写的torch.Relu首字母大写错误，会触发属性查找失败。
• 浮点数相等判断逻辑错误：test_step中用logits == y.data判断相等不符合浮点数计算逻辑，存在精度误差时永远返回False，需要用torch.isclose设置容差判断。
• 未定义全局变量：test_step中用到的predictions_pred、predictions_actual没有提前初始化，运行会触发NameError。

修正后的核心代码 #

import numpy as np
import torch
import matplotlib.pyplot as plt
import pytorch_lightning as pl
import torch.optim as optim
from torch import nn
from pytorch_lightning import Trainer
from sklearn.model_selection import train_test_split
import pandas as pd
from torch.utils.data import DataLoader, TensorDataset # 补导入TensorDataset

# 数据生成部分保持不变
N=1000
x = np.random.randn(1000)
y = np.sin(x)
df = pd.DataFrame({'x':x, 'y':y})
train, test = train_test_split(df, test_size=0.2, random_state=42, shuffle=True)
target_fields=['y']
train_features, train_targets = train.drop(target_fields, axis=1), train[target_fields]
test_features, test_targets = test.drop(target_fields, axis=1), test[target_fields]

# 修正类结构与缩进
class MLP(pl.LightningModule):
    def __init__(self):
        super(MLP,self).__init__()
        self.fc1 = nn.Linear(1, 10)
        self.fc2 = nn.Linear(10, 1)
        self.l_rate = 0.2
        self.mse_loss = nn.MSELoss(reduction = 'mean')
        # 初始化预测存储列表
        self.predictions_pred = []
        self.predictions_actual = []

    def forward(self, x):
        x = torch.relu(self.fc1(x)) # 修正ReLU拼写
        x = self.fc2(x)
        return x

    def train_dataloader(self):
        # 修正标签列名
        train_dataset = TensorDataset(torch.tensor(train_features.values).float(), torch.tensor(train_targets[['y']].values).float())
        train_loader = DataLoader(dataset = train_dataset, batch_size = 128, shuffle=True)
        return train_loader
    
    def test_dataloader(self):
        test_dataset = TensorDataset(torch.tensor(test_features.values).float(), torch.tensor(test_targets[['y']].values).float())
        test_loader = DataLoader(dataset = test_dataset, batch_size = 128)
        return test_loader
    
    def configure_optimizers(self):
        return optim.SGD(self.parameters(), lr=self.l_rate)
    
    def training_step(self, batch, batch_idx):
        x, y = batch
        logits = self.forward(x)
        loss = self.mse_loss(logits, y)
        self.log('train_loss', loss, prog_bar=True) # Lightning官方推荐的日志写法
        return loss
    
    def test_step(self, batch, batch_idx):
        x, y = batch
        logits = self.forward(x)
        loss = self.mse_loss(logits, y)
        # 修正浮点数相等判断逻辑
        correct = torch.sum(torch.isclose(logits, y.data, atol=1e-3))
        self.predictions_pred.append(logits)
        self.predictions_actual.append(y.data)
        self.log('test_loss', loss)
        return {'test_loss': loss, 'test_correct': correct}
    
    def test_epoch_end(self, outputs):
        avg_loss = torch.stack([x['test_loss'] for x in outputs]).mean()
        self.log('avg_test_loss', avg_loss, prog_bar=True)

# 运行训练
model = MLP()
trainer = Trainer(max_epochs = 50)  
trainer.fit(model)

内容的提问来源于stack exchange，提问作者Anilaaryan