Fix Keras 3.x compatibility in FactorizedTopK layer
The add_weight() method call in Streaming.init used positional arguments which breaks with Keras 3.x. Changed to keyword argument to maintain compatibility with both Keras 2.x and 3.x versions.
Fixes users getting "Cannot convert ('c','o','u','n','t','e','r') to a shape" error when using TensorFlow 2.16+ with TFRS 0.7.3.
Fixes #712 Fixes #731 Fixes #748 Fixes #754 Fixes #759
When can we expect this pull request to be completed? At the moment, the TensorFlow recommenders tutorials won't work with any FactorizedTopK metrics.
Hello, I fixed internally via https://github.com/tensorflow/recommenders/pull/765 (same PR as yours).
#765 which merged should have this fixed. Sadly, there's no autoclose of this and duplicates.