Issue
I have been developing machine learning codes in TensorFlow with contrib.layers
. I am happy with the module; it works very well and provides me sufficient control over my model. However, TensorFlow 2.0 will completely remove contrib
module and the new keras
module does not provide me enough flexibility and control without extra efforts (personal experience, correct me if I am wrong), even though it is easier to write.
I want to know where the contrib
module will be in TensorFlow 2.0. Thanks in advance.
Solution
According to the RFC document from August 2018, tf.contrib
will be deleted with some of its parts becoming standalone projects (such as tensorflow/probability).
This is the case for tf.conrib.layers
. Even tf.layers
(which was distilled from tf.contrib.layers
) will no longer be supported. A detailed description of how to use the Keras API instead is provided in the migration guide.
Answered By - Jindřich
0 comments:
Post a Comment
Note: Only a member of this blog may post a comment.