* Package name    : tensorboard
  Version         : 2.3.0
  Upstream Author : TensorFlow authors
* URL             : https://github.com/tensorflow/tensorboard
* License         : Apache-2.0
  Programming Lang: TypeScript, Python
  Description     : TensorFlow's Visualization Toolkit

TensorFlow's visualization tool can be used standalone (i.e. without
tensorflow). It's very useful for training networks using PyTorch as
well. This package will be maintained by Debian Deep Learning Team
for sake of unblocking the pytorch-lightning package.

Hmmm, that seems to be a lot.  Strangely enough if I try to build in
pbuilder it downloads several dependencies from the internet - except if
sitting behind a proxy that is unknown to the pbuilder environment.  I
have never observed this before - seems bazel knows "tricks" to
undermine the offlineish nature of the pbuilder environment.  I have
never seen this happening before.

I've commited some initial packaging to Salsa[1].  It also needs
some third party software.  The first one is

http://mirror.tensorflow.org/github.com/bazelbuild/bazel-skylib/archive/0.7.0.tar.gz

I wonder whether all those dependencies are mandatory or whether
we might be able to exclude something optional.

I could *help* with this - but I'm clearly neither qualified
nor does my time limit permit really focussing on this work.

Kind regards

       Andreas.

[1] https://salsa.debian.org/deeplearning-team/tensorboard