我尝试在https://colab.research.google.com/github/keras-team/keras-io/blob/master/examples/structured_data/ipynb/movielens_recommendations_transformers.ipynb中运行Behavior Sequence Transformer模型，以根据以前的电影评级预测下一部要观看的电影。
我已经让Colab notebook中的代码正确运行，但似乎无法从中获得任何有意义的预测。
下面是model.input_shape的输出：

{'user_id': (None, 1), 'sequence_movie_ids': (None, 3), 'target_movie_id': (None, 1), 'sequence_ratings': (None, 3), 'sex': (None, 1), 'age_group': (None, 1), 'occupation': (None, 1)}

下面是我用于预测的代码：

# Load the latest 3 movies you watched and gave them all a rating of 4
latest_movies = pd.DataFrame({
    'user_id': [0],
    'sequence_movie_ids': [[1, 2, 3]],
    'target_movie_id': [0],
    'sequence_ratings': [[4.0, 4.0, 4.0]],
    'sex': [0],
    'age_group': [0],
    'occupation': [0]
})

# Predict the next movie to watch
next_movie = model.predict(input_data)
print(next_movie)

但是，这会导致以下错误消息：

Node: 'model/Cast'
Cast int32 to string is not supported
     [[{{node model/Cast}}]] [Op:__inference_predict_function_40501]

我还尝试将input_data转换为Tensor：

# Prepare the input data
input_data = {
    'user_id': tf.convert_to_tensor(latest_movies['user_id'], dtype=tf.int32),
    'sequence_movie_ids': tf.convert_to_tensor(latest_movies['sequence_movie_ids'], dtype=tf.int32),
    'target_movie_id': tf.convert_to_tensor(latest_movies['target_movie_id'], dtype=tf.int32),
    'sequence_ratings': tf.convert_to_tensor(latest_movies['sequence_ratings'], dtype=tf.float32),
    'sex': tf.convert_to_tensor(latest_movies['sex'], dtype=tf.int32),
    'age_group': tf.convert_to_tensor(latest_movies['age_group'], dtype=tf.int32),
    'occupation': tf.convert_to_tensor(latest_movies['occupation'], dtype=tf.int32)
}

但这也失败了，并显示错误消息：

ValueError: setting an array element with a sequence.

我们如何让这个模型做出预测？
非常感谢