Open ramakumar1729 opened 4 years ago
mercarikaicheung
commented
3 years ago Hi, I am trying to merge the embeddings and numeric data using the keras api for TFR. However, since the data is stored in TFrecord the text are in bytes and hub.text_embedding_column for embeddings only accepts string data type. Is there a work around for this?
ramakumar1729
commented
3 years ago @mercarikaicheung : can you please open a new issue, since this particular one is an announcement? From release v0.4.0, in sync with TensorFlow, we are discouraging the use of Feature Columns but you can use Keras Preprocessing Layers instead.
mercarikaicheung
commented
3 years ago @mercarikaicheung : can you please open a new issue, since this particular one is an announcement? From release v0.4.0, in sync with TensorFlow, we are discouraging the use of Feature Columns but you can use Keras Preprocessing Layers instead.
Got it, the documentation for the keras TFR api was quite sparse, most of the examples for TFR was using Feature Column, wasn't sure the keras preprocessing layers are compatible.
ZhouM1118
commented
2 years ago Hi, what's mean "size_feature_name"? How can I use "size_feature_name" in the sorting task of variable-length lists?
In the latest release of TensorFlow Ranking v0.3.0, we introduced Keras APIs for TF-Ranking.
We provide a simplified user journey for building ranking models. We provide object oriented building blocks for ranking losses, metrics, networks and models.
For more details, see an example to showcase the use of Keras APIs to build ranking models , and a documentation providing step-by-step user instructions outlining the Keras user journey.
TF-Ranking Keras APIs should help:
Feel free to leave comments and suggestions to this announcement. We are also happy to learn more about your use cases, and how the Keras APIs can help.