在张量流的 tf.nn.seq2seq class 中找不到很多东西的纪录片
Cannot find documentary for many things in tf.nn.seq2seq class in the tensorflow
tf.nn.seq2seq.embedding_attention_seq2seq
我在 tensorflow api 中搜索上述函数。但我找不到确切的名字。但是 class tf.contrib.legacy_seq2seq.embedding_attention_seq2seq 中也有类似的功能。
为什么他们很少从 nn class 转移到 contrib class？
Google just released TensorFlow 1.0 last week，其中 API 已更新，因此内容已移动。由于 TF 社区的各种贡献，contrib 目录是为这段代码保留的，尚未最终确定是否接受到核心 TF。
查看 README 贡献：
The contrib directory contains project directories, each of which has
designated owners. It is meant to contain features and contributions
that eventually should get merged into core TensorFlow, but whose
interfaces may still change, or which require some testing to see
whether they can find broader acceptance. We are trying to keep
duplication within contrib to a minimum, so you may be asked to
refactor code in contrib to use some feature inside core or in another
project in contrib rather than reimplementing the feature.
tf.nn.seq2seq.embedding_attention_seq2seq
我在 tensorflow api 中搜索上述函数。但我找不到确切的名字。但是 class tf.contrib.legacy_seq2seq.embedding_attention_seq2seq 中也有类似的功能。 为什么他们很少从 nn class 转移到 contrib class？
Google just released TensorFlow 1.0 last week，其中 API 已更新，因此内容已移动。由于 TF 社区的各种贡献，contrib 目录是为这段代码保留的，尚未最终确定是否接受到核心 TF。
查看 README 贡献：
The contrib directory contains project directories, each of which has designated owners. It is meant to contain features and contributions that eventually should get merged into core TensorFlow, but whose interfaces may still change, or which require some testing to see whether they can find broader acceptance. We are trying to keep duplication within contrib to a minimum, so you may be asked to refactor code in contrib to use some feature inside core or in another project in contrib rather than reimplementing the feature.