The training model need not be built when the kfac optimizer is initialized so the
authorA. Unique TensorFlower <gardener@tensorflow.org>
Fri, 6 Apr 2018 15:48:16 +0000 (08:48 -0700)
committerTensorFlower Gardener <gardener@tensorflow.org>
Fri, 6 Apr 2018 15:50:41 +0000 (08:50 -0700)
commitafc21e7149a0d146bd8db3145fe825b1f316c0a9
treeca73b79711faf2ce042fe3fa7534e85bd9eba096
parent7eeb54aa745ac45c15e886385ec33372d5966b23
The training model need not be built when the kfac optimizer is initialized so the
self._variables will be empty list.  So pass a function which returns list of trainable variables to estimator.

PiperOrigin-RevId: 191893084
tensorflow/contrib/kfac/python/ops/estimator.py
tensorflow/contrib/kfac/python/ops/optimizer.py