Python tf.feature_column.make_parse_example_spec用法及代码示例

从输入 feature_columns 创建解析规范字典。

用法

tf.feature_column.make_parse_example_spec(
    feature_columns
)

返回的字典可以用作 tf.io.parse_example 中的 arg 'features'。

典型使用示例：

# Define features and transformations
feature_a = tf.feature_column.categorical_column_with_vocabulary_file(...)
feature_b = tf.feature_column.numeric_column(...)
feature_c_bucketized = tf.feature_column.bucketized_column(
    tf.feature_column.numeric_column("feature_c"), ...)
feature_a_x_feature_c = tf.feature_column.crossed_column(
    columns=["feature_a", feature_c_bucketized], ...)

feature_columns = set(
    [feature_b, feature_c_bucketized, feature_a_x_feature_c])
features = tf.io.parse_example(
    serialized=serialized_examples,
    features=tf.feature_column.make_parse_example_spec(feature_columns))

对于上面的示例，make_parse_example_spec 将返回字典：

{
    "feature_a":parsing_ops.VarLenFeature(tf.string),
    "feature_b":parsing_ops.FixedLenFeature([1], dtype=tf.float32),
    "feature_c":parsing_ops.FixedLenFeature([1], dtype=tf.float32)
}