使用tensorflow GradientTape在keras中应用权重后，权重不会更改

import tensorflow as tf
from tensorflow import keras

@tf.function
def forward_pass(inputs):
    # 定义模型结构
    model = keras.Sequential([
        keras.layers.Dense(64, activation='relu'),
        keras.layers.Dense(10, activation='softmax')
    ])
    
    # 前向传播
    outputs = model(inputs)
    
    return outputs

# 定义输入数据
inputs = tf.random.normal((32, 100))

# 前向传播
outputs = forward_pass(inputs)

@tf.function
def compute_loss(inputs, labels):
    # 定义模型结构
    model = keras.Sequential([
        keras.layers.Dense(64, activation='relu'),
        keras.layers.Dense(10, activation='softmax')
    ])
    
    # 前向传播
    outputs = model(inputs)
    
    # 计算损失函数
    loss = keras.losses.sparse_categorical_crossentropy(labels, outputs)
    
    return loss

# 定义输入数据和标签
inputs = tf.random.normal((32, 100))
labels = tf.random.uniform((32,), maxval=10, dtype=tf.int32)

# 计算损失函数
with tf.GradientTape() as tape:
    loss = compute_loss(inputs, labels)

# 计算梯度
gradients = tape.gradient(loss, model.trainable_variables)

# 更新权重
optimizer = tf.keras.optimizers.SGD(learning_rate=0.1)
optimizer.apply_gradients(zip(gradients, model.trainable_variables))