TensorFlow线性回归-返回权重、偏差和信息损失的NaN

import tensorflow as tf
import numpy as np

# 定义输入特征和标签
features = np.array([1, 2, 3, 4, 5], dtype=np.float32)
labels = np.array([3, 5, 7, 9, 11], dtype=np.float32)

# 定义权重和偏差变量
w = tf.Variable(0.0)
b = tf.Variable(0.0)

# 定义线性回归模型
def linear_regression(x):
    return w * x + b

# 定义损失函数
def loss(y_true, y_pred):
    return tf.reduce_mean(tf.square(y_true - y_pred))

# 定义优化器
optimizer = tf.optimizers.SGD(learning_rate=0.01)

# 定义训练过程
def train_step(features, labels):
    with tf.GradientTape() as tape:
        # 计算预测值
        predictions = linear_regression(features)
        # 计算损失
        current_loss = loss(labels, predictions)
    # 计算梯度
    gradients = tape.gradient(current_loss, [w, b])
    # 更新权重和偏差
    optimizer.apply_gradients(zip(gradients, [w, b]))

# 进行训练
for _ in range(100):
    train_step(features, labels)

# 打印最终的权重、偏差和信息损失
print("权重：", w.numpy())
print("偏差：", b.numpy())
print("信息损失：", loss(labels, linear_regression(features)).numpy())