Merge changes from github.

END_PUBLIC

---
Commit 301b14c2 authored by Skye Wanderman-Milne<skyewm@google.com>
Committed by TensorFlower Gardener<gardener@tensorflow.org>:
Basic while loop gradient functionality in C++

This change introduces the basic framework to create the gradient
graph of a while loop using the C++ API. This supports building the
gradient graph as long as the body function of the while loop contains
no ops whose gradient function requires a stack. In other words, it
doesn't support gradient functions that use the input values to the op
(e.g. add will work, but multiply will not). It also doesn't support
nested while loops, and doesn't detect all error cases.

PiperOrigin-RevId: 170243281

---
Commit 545e3572 authored by Asim Shankar<ashankar@google.com>
Committed by TensorFlower Gardener<gardener@tensorflow.org>:
Datasets: Reference the programmer's guide in API docs.

PiperOrigin-RevId: 170241348

---
Commit 24890d55 authored by A. Unique TensorFlower<gardener@tensorflow.org>
Committed by TensorFlower Gardener<gardener@tensorflow.org>:
Go: Update generated wrapper functions for TensorFlow ops.

PiperOrigin-RevId: 170241322

---
Commit 02d2f376 authored by A. Unique TensorFlower<gardener@tensorflow.org>
Committed by TensorFlower Gardener<gardener@tensorflow.org>:
Update ops-related pbtxt files.

PiperOrigin-RevId: 170240603

---
Commit 759690f0 authored by Reed Wanderman-Milne<reedwm@google.com>
Committed by TensorFlower Gardener<gardener@tensorflow.org>:
Add float16 support to tf.nn.fused_batch_norm on the GPU.

Scale, offset, mean, and variance must still be float32 if the input is float16.

PiperOrigin-RevId: 170239448

---
Commit 20370104 authored by Igor Saprykin<isaprykin@google.com>
Committed by TensorFlower Gardener<gardener@tensorflow.org>:
Support export strategies in _TrainingExecutor.

One could set export strategies to the EvalSpec.  An exception is raised if the type isn't export_strategy.ExportStrategy.  During continuous evaluation, export strategies are going to be triggered. They in turn call Estimator's export_savedmodel.

PiperOrigin-RevId: 170237073

---
Commit 56402103 authored by Reed Wanderman-Milne<reedwm@google.com>
Committed by TensorFlower Gardener<gardener@tensorflow.org>:
Fix BFC allocator's log messages on OOM error.

Before, the "Chunks in use" message and other in-use messages would always be 0.

PiperOrigin-RevId: 170233715

---
Commit bc80e46b authored by Peter Hawkins<phawkins@google.com>
Committed by TensorFlower Gardener<gardener@tensorflow.org>:
[TF:XLA] Implement BroadcastArgs.

PiperOrigin-RevId: 170228025

---
Commit bced6676 authored by A. Unique TensorFlower<gardener@tensorflow.org>
Committed by TensorFlower Gardener<gardener@tensorflow.org>:
BEGIN_PUBLIC
Automated g4 rollback of changelist 170204652

PiperOrigin-RevId: 170367641