Why is the loss NaN

Question

I used softmax to implement classification, but my code encountered a loss during runtime. This is my code:

#!/usr/bin/env python
# coding: utf-8

# In[1]:


import torch
import pandas as pd
import numpy as np
from d2l import torch as d2l
from torch import nn
from sklearn.model_selection import train_test_split
from IPython import display
from sklearn.preprocessing import StandardScaler


# In[2]:


batch_size = 10000
num_inputs = 16
num_outputs = 6

W = torch.normal(0, 0.01, size=(num_inputs, num_outputs), requires_grad=True,dtype=torch.float32)
b = torch.zeros(num_outputs, requires_grad=True,dtype=torch.float32)


# In[3]:


def normal(data):
    scaler = StandardScaler()
    scaler.fit(data)
    data = scaler.transform(data)
    return data


# In[4]:


def load_array(data_train, data_label, batch_size, is_train=True):  # is_train是否打乱数据
    dataset = torch.utils.data.TensorDataset(data_train, data_label)  # 传入参数(data_tensor,data_target)
    data_iter = torch.utils.data.DataLoader(dataset, batch_size, shuffle=True,num_workers = 0)
    return data_iter


# In[5]:


def dataIter(data_train,data_label):
    #data_train = normal(data_train)
    data_train = np.array(data_train)
    data_train = torch.from_numpy(data_train)
    data_label = np.array(data_label)
    data_label = torch.from_numpy(data_label)
    data_train = data_train.to(torch.float32)
    data_label = data_label.to(torch.float32)
    data = load_array(data_train, data_label, batch_size, is_train=True)
    return data


# In[6]:


def splitData(data_ves):
    X = data_ves.iloc[:, 2:18]
    y = data_ves.iloc[:, 1]
    X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=0,shuffle = True)
    train_iter = dataIter(X_train,y_train)
    test_iter = dataIter(X_test,y_test)    
    return train_iter,test_iter


# In[7]:


# 导入数据
datapath = "D:/Code/datasets/Anonymized AIS training data/demo3.csv"
data_ves = pd.read_csv(datapath)
train_iter, test_iter = splitData(data_ves)


# In[8]:


def softmax(X):
    X_exp = torch.exp(X)
    partition = X_exp.sum(1, keepdim=True)
    return X_exp / partition  # 这里应用了广播机制


# In[9]:


def net(X):
    X = X.reshape((-1, W.shape[0]))
    temp = X@W+ b
    y_hat = softmax(temp)
    return y_hat


# In[10]:


def cross_entropy(y_hat, y):
    y = y.to(torch.int64)
    loss = - torch.log(y_hat[range(len(y_hat)), y])
    return loss


# In[11]:


def accuracy(y_hat, y):  
    """计算预测正确的数量"""
    if len(y_hat.shape) > 1 and y_hat.shape[1] > 1:
        y_hat = y_hat.argmax(axis=1)
    cmp = y_hat.type(y.dtype) == y
    return float(cmp.type(y.dtype).sum())


# In[12]:


def evaluate_accuracy(net, data_iter):  #@save
    """计算在指定数据集上模型的精度"""
    if isinstance(net, torch.nn.Module):
        net.eval()  # 将模型设置为评估模式
    metric = Accumulator(2)  # 正确预测数、预测总数
    with torch.no_grad():
        for X, y in data_iter:
            metric.add(accuracy(net(X), y), y.numel())
    return metric[0] / metric[1]


# In[13]:


class Accumulator:  
    """在n个变量上累加"""
    def __init__(self, n):
        self.data = [0.0] * n

    def add(self, *args):
        self.data = [a + float(b) for a, b in zip(self.data, args)]

    def reset(self):
        self.data = [0.0] * len(self.data)

    def __getitem__(self, idx):
        return self.data[idx]


# In[14]:


def train_epoch(net, train_iter, loss, updater):
    # 将模型设置为训练模式
    if isinstance(net, torch.nn.Module):
        net.train()
    # 训练损失总和、训练准确度总和、样本数
    metric = Accumulator(3)
    for X, y in train_iter:
        # 计算梯度并更新参数
        y_hat = net(X)
        l = loss(y_hat, y)
        if isinstance(updater, torch.optim.Optimizer):
            # 使用PyTorch内置的优化器和损失函数
            updater.zero_grad()
            l.mean().backward()
            updater.step()
        else:
            # 使用定制的优化器和损失函数
            l.sum().backward()
            updater(X.shape[0])
        metric.add(float(l.sum()), accuracy(y_hat, y), y.numel())
    # 返回训练损失和训练精度
    return metric[0] / metric[2], metric[1] / metric[2]


# In[15]:


class Animator:
    """在动画中绘制数据"""
    def __init__(self, xlabel=None, ylabel=None, legend=None, xlim=None,
                 ylim=None, xscale='linear', yscale='linear',
                 fmts=('-', 'm--', 'g-.', 'r:'), nrows=1, ncols=1,
                 figsize=(3.5, 2.5)):
        # 增量地绘制多条线
        if legend is None:
            legend = []
        d2l.use_svg_display()
        self.fig, self.axes = d2l.plt.subplots(nrows, ncols, figsize=figsize)
        if nrows * ncols == 1:
            self.axes = [self.axes, ]
        # 使用lambda函数捕获参数
        self.config_axes = lambda: d2l.set_axes(
            self.axes[0], xlabel, ylabel, xlim, ylim, xscale, yscale, legend)
        self.X, self.Y, self.fmts = None, None, fmts

    def add(self, x, y):
        # 向图表中添加多个数据点
        if not hasattr(y, "__len__"):
            y = [y]
        n = len(y)
        if not hasattr(x, "__len__"):
            x = [x] * n
        if not self.X:
            self.X = [[] for _ in range(n)]
        if not self.Y:
            self.Y = [[] for _ in range(n)]
        for i, (a, b) in enumerate(zip(x, y)):
            if a is not None and b is not None:
                self.X[i].append(a)
                self.Y[i].append(b)
        self.axes[0].cla()
        for x, y, fmt in zip(self.X, self.Y, self.fmts):
            self.axes[0].plot(x, y, fmt)
        self.config_axes()
        display.display(self.fig)
        display.clear_output(wait=True)


# In[27]:


def train(net, train_iter, test_iter, loss, num_epochs, updater): 
    animator = Animator(xlabel='epoch', xlim=[1, num_epochs], ylim=[0.3, 0.9],
                        legend=['train loss', 'train acc', 'test acc'])
    for epoch in range(num_epochs):
        print(f"Epoch [{epoch+1}/{num_epochs}]")
        train_metrics = train_epoch(net, train_iter, loss, updater)
        test_acc = evaluate_accuracy(net, test_iter)
        animator.add(epoch + 1, train_metrics + (test_acc,))
        train_loss, train_acc = train_metrics
        print(f'epoch {epoch}, loss {train_loss}, train acc {train_acc} test acc {test_acc}')
    train_loss, train_acc = train_metrics
    assert train_loss < 0.5, train_loss
    assert train_acc <= 1 and train_acc > 0.7, train_acc
    assert test_acc <= 1 and test_acc > 0.7, test_acc
    


# In[17]:


def sgd(params, lr, batch_size):
    with torch.no_grad():
        for param in params:
            param -= lr * param.grad / batch_size
            param.grad.zero_()


# In[18]:


lr = 0.00001
def updater(batch_size):
    return d2l.sgd([W, b], lr, batch_size)


# In[28]:


num_epochs = 1
train(net, train_iter,test_iter,cross_entropy, num_epochs, updater)

This is my part of data:

enter image description here

I find the (W, b) becomes nan after some batches, because there were problems with gradient calculation, but I don't know exactly what the problem is.