传递到Dataset.map（）的函数返回值不受支持

Question

我正试图使用tensorflow管线对iris数据集进行预处理和分类，但预处理后我遇到了以下错误：不支持从传递到dataset.map（）的函数返回值：（，NumericColumn（key='features'，shape=（4，），default_value=None，dtype=tf.float32，normalizer_fn=None）），我被困在那里，我很乐意得到任何帮助 这是DNN的完整代码

import tensorflow_datasets as tfds
from tensorflow.keras.optimizers import Adam


data = tfds.load("iris", split=tfds.Split.TRAIN)


def preprocess(features):

    # should return features and one-hot encoded labels
    #l  = tf.feature_column.categorical_column_with_identity("label", 3, default_value=None)
    l = tf.one_hot(features["label"], 3)
    f = tf.feature_column.numeric_column("features", shape=(4,), dtype=tf.dtypes.float32)
    return l, f

def solution_model():
    train_dataset = data.map(preprocess).batch(10)

    dataset_layer = tf.keras.layers.DenseFeatures(train_dataset)


    # YOUR CODE TO TRAIN A MODEL
    model = tf.keras.Sequential([
        dataset_layer,
        tf.keras.layers.Dense(32, input_shape=(None,4)),
        tf.keras.layers.Dense(64, activation=tf.nn.relu),
        tf.keras.layers.Dense(3, activation=tf.nn.softmax)
    ])
    print(model.summary)
    #model.compile(optimizer="sgd", loss="mean_squared_error")
    model.compile(loss=tf.keras.losses.categorical_crossentropy,
              optimizer=tf.keras.optimizers.SGD(
              learning_rate=0.01, momentum=0.0, nesterov=False, name='SGD'),
              metrics=['accuracy'])
    model.fit(train_dataset, epochs=100)
    return model

if __name__ == '__main__':
    model = solution_model()
    model.save('mymodel.h5')```