feature_column_ops_test.py 文件源码

def testAllDNNColumns(self):
    sparse_column = tf.contrib.layers.sparse_column_with_keys(
        "ids", ["a", "b", "c", "unseen"])

    real_valued_column = tf.contrib.layers.real_valued_column("income", 2)
    one_hot_column = tf.contrib.layers.one_hot_column(sparse_column)
    embedding_column = tf.contrib.layers.embedding_column(sparse_column, 10)
    features = {
        "ids": tf.SparseTensor(
            values=["c", "b", "a"],
            indices=[[0, 0], [1, 0], [2, 0]],
            shape=[3, 1]),
        "income": tf.constant([[20.3, 10], [110.3, 0.4], [-3.0, 30.4]])
    }
    output = tf.contrib.layers.input_from_feature_columns(features,
                                                          [one_hot_column,
                                                           embedding_column,
                                                           real_valued_column])
    with self.test_session():
      tf.initialize_all_variables().run()
      tf.initialize_all_tables().run()
      self.assertAllEqual(output.eval().shape, [3, 2 + 4 + 10])