TFTS: Fix a bug in the SavedModel cold-start export
authorAllen Lavoie <allenl@google.com>
Tue, 27 Mar 2018 22:55:04 +0000 (15:55 -0700)
committerTensorFlower Gardener <gardener@tensorflow.org>
Tue, 27 Mar 2018 22:57:43 +0000 (15:57 -0700)
commita16761483ec55095158b1b11118d93ea00a538f4
tree1d248134f08e8c8cc9c3602c0db1a3faa7fdb6b5
parentb4742b76c386409c96c60172e6ca1c1534e2b4af
TFTS: Fix a bug in the SavedModel cold-start export

It now correctly broadcasts start state across whatever batch dimension it is
passed rather than sqishing it down to a batch dimension of 1.

PiperOrigin-RevId: 190688855
tensorflow/contrib/timeseries/python/timeseries/estimators_test.py
tensorflow/contrib/timeseries/python/timeseries/head.py