bhack
bhack
It would be really useful if we could revamp and approve this RFC.
/cc @learning-to-play
Do we have an opportunity to re-evalute/refresh this and all the useful feedback we had collected?
/cc @howl-anderson
TF bazel version update was reverted few hours ago again: https://github.com/tensorflow/tensorflow/commit/1f19aea3230acc158538951f4c510d98df3a99ee
We could leave this open and labeled for tracking.
In Caffe we introduced a [filter layer trick](http://caffe.berkeleyvision.org/doxygen/classcaffe_1_1FilterLayer.html#details) some time ago to have a conditional path but I don't think that this approach could cover your cases.
Filter layer create a dynamic tensor on the batch dimension
@Rishit-dagli are you interested to contribute here? Context: https://github.com/tensorflow/models/issues/10269
Can we close this for https://github.com/keras-team/keras-cv/issues/671?