Hi, On Mon, Jan 07, 2019 at 12:24:29AM +0100, Petter Reinholdtsen wrote: > Is TensorFlow different from libtensorflow, already in unstable: ^^^^^^^^ experimental > libtensorflow-cc1.10 - Computation using data flow graphs for scalable > machine learning (C++) > libtensorflow-dev - Computation using data flow graphs for scalable machine > learning (dev) > libtensorflow-framework1.10 - Computation using data flow graphs for scalable > machine learning > libtensorflow1.10 - Computation using data flow graphs for scalable machine > learning (C) > > If not, perhaps keras should be adjusted to use it, and its
No. Please don't build any package that depends on TF 1.X because it is literally an incomplete WIP. You can do that untill TF 2.X landed onto experimental with a (yet anotehr) totally re-written build system. > package description updated? The tensorflow source package > entered unstable 2018-11-22. ^^^^^^^^ experimental > As for the problem with running keras in unstable, perhaps it is a > good idea to extend the autopkgtest script to detect such problems early? Without autopkgtest I'm already aware of bunch of issues existing in the present tensorflow package, which renders it inappropriate to enter even unstable.