tf.feature_column的使用记录

最新推荐文章于 2023-03-21 11:41:07 发布

Massacre96Wj

最新推荐文章于 2023-03-21 11:41:07 发布

阅读量307

点赞数

分类专栏：深度学习文章标签： tensorflow

本文链接：https://blog.youkuaiyun.com/weixin_42047706/article/details/125771840

版权

深度学习专栏收录该内容

4 篇文章

订阅专栏

tf.feature_column使用，可用线性模型或者序列模型

features = {
    'user_item': [["A","B","-1"], 
                  ["A","B","A"], 
                  ["A","B","B"], 
                  ["C","B","C"]],
    }

# 线性模型
user_item = tf.feature_column.categorical_column_with_vocabulary_list(key="user_item",
                                                                               vocabulary_list=("A","B","C"),
#                                                                                default_value=-1,
                                                                               num_oov_buckets=2)
columns = tf.feature_column.indicator_column(user_item)
inputs = tf.compat.v1.feature_column.input_layer(features, columns)
print(inputs)



# 序列模型使用
user_seq_item = tf.feature_column.sequence_categorical_column_with_vocabulary_list(key="user_item",
                                                                               vocabulary_list=("A","B","C"),
#                                                                                default_value=-1,
                                                                               num_oov_buckets=2)
# columns = tf.feature_column.embedding_column(user_item, 8)
columns = tf.feature_column.indicator_column(user_seq_item)
sequence_feature_layer = tf.keras.experimental.SequenceFeatures(columns)
sequence_input, sequence_length = sequence_feature_layer(features)
print(sequence_input, sequence_length)
sequence_length_mask = tf.sequence_mask(sequence_length)
print(sequence_length_mask)