请教下如何feature_column处理类似VarLenSparseFeat这种序列特征

[zhoujiang2013]

如何feature_column处理类似VarLenSparseFeat这种序列特征，使用tfrecord的example中只看懂了 SparseFeat、DenseFeat的使用：

for i, feat in enumerate(sparse_features):
    dnn_feature_columns.append(tf.feature_column.embedding_column(
        tf.feature_column.categorical_column_with_identity(feat, 1000), 4))
    linear_feature_columns.append(tf.feature_column.categorical_column_with_identity(feat, 1000))
for feat in dense_features:
    dnn_feature_columns.append(tf.feature_column.numeric_column(feat))
    linear_feature_columns.append(tf.feature_column.numeric_column(feat))

[skriser]

解决了么

[Henry199898]

先对[batch, max_len]中每个单词查embedding向量为[batch, max_len, embedding_dim]，然后对该序列池化（['sum', 'mean', 'max']）得到最后的varlen向量[batch, embedding_dim]. 我理解是这样

[shenweichen]

Closing this as covered by the merged docs update in #554.

The new Sequence Feature Cookbook consolidates the common sequence/VarLen/DIN/DSIN input conventions, including VarLenSparseFeat shapes, padding and length_name, embedding sharing, DIN/BST history fields, DSIN session inputs, and Estimator/TFRecord vector inputs.

Docs: https://deepctr-doc.readthedocs.io/en/latest/Sequence-Cookbook.html