TensorFlow线性回归模型不起作用

import tensorflow as tf
import numpy as np

# 定义输入特征和输出目标
x_train = np.array([...])  # 输入特征
y_train = np.array([...])  # 输出目标

# 定义模型参数
W = tf.Variable(tf.random.normal([num_features, 1]), name='weight')
b = tf.Variable(tf.zeros(1), name='bias')

# 定义线性回归模型
def linear_regression(x):
    return tf.matmul(x, W) + b

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

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

# 定义训练函数
def train_step(x, y):
    with tf.GradientTape() as tape:
        y_pred = linear_regression(x)
        loss = mean_square(y_pred, y)
    gradients = tape.gradient(loss, [W, b])
    optimizer.apply_gradients(zip(gradients, [W, b]))

# 迭代训练
for epoch in range(num_epochs):
    train_step(x_train, y_train)

# 定义预测函数
def predict(x):
    return linear_regression(x)

# 使用模型进行预测
y_pred = predict(x_test)