Tensorflow-使用tf.estimator.inputs.numpy_input_fn时如何提供多值功能

时间:2018-08-07 12:02:52

标签: tensorflow deep-learning one-hot-encoding

数据集有很多列,我使用input_fn馈送数据以进行建模:

def input_fn(data_file, num_epochs, shuffle, is_eval):
    """Input builder function."""
    df_data = pd.read_csv(
        tf.gfile.Open(data_file),
        names=CSV_COLUMNS,
        skipinitialspace=True,
        engine="python",
        skiprows=1,
        sep = '\t')

    df_data["sentence"] = df_data["tags"].apply(lambda x : x.split(','))

    labels = df_data["is_click"]
    return tf.estimator.inputs.numpy_input_fn(
        x={
           "book":     df_data['book'], 
           "author":   df_data['author'], 
           "sentence": df_data['sentence']
        },
    y=labels,
    batch_size=2048,
    num_epochs=num_epochs,
    shuffle=shuffle,
    num_threads=5
)

特征[句子]将被拆分为列表,因为我想将句子转换为像https://www.tensorflow.org/versions/r1.4/api_docs/python/tf/feature_column/indicator_column这样的一键编码

但是如果我将句子拆分,则会出现以下错误:

tensorflow.python.framework.errors_impl.InternalError: Unable to get element as bytes.

0 个答案:

没有答案