Merge changes from github.

END_PUBLIC

---
Commit cf375f067 authored by A. Unique TensorFlower<[email protected]>
Committed by Benoit Steiner<[email protected]>:
Adds cudnn_rnn_ops_op_lib and cudnn_rnn_kernels to contrib_ops_op_lib and
contrib_kernels respectively.

PiperOrigin-RevId: 164170971

---
Commit 95ec58e27 authored by Asim Shankar<[email protected]>
Committed by Benoit Steiner<[email protected]>:
C API: Make TF_TensorFromTensor return an error instead of just logging it.

PiperOrigin-RevId: 164167582

---
Commit 15175c870 authored by Jonathan Hseu<[email protected]>
Committed by Benoit Steiner<[email protected]>:
Build fixes.

- Allow var_list as a positional argument in CrossShardOptimizer.
- Set the number of shards to 1 when not running on TPU, to allow evaluate() and predict() on CPU/GPU to work.

PiperOrigin-RevId: 164161640

---
Commit bd3e894f7 authored by Yao Zhang<[email protected]>
Committed by Benoit Steiner<[email protected]>:
Support freeze mode for fused batch norm.

PiperOrigin-RevId: 164149032

---
Commit e6b6b84c0 authored by Asim Shankar<[email protected]>
Committed by Benoit Steiner<[email protected]>:
C API: TF_Tensors will always be in host memory.

This change undoes some experimentation in
commit 22651083406ca01ac9d481e3367a3510d25f88cd
and restores TF_Tensor behavior to what is was prior to that change.

PiperOrigin-RevId: 164146670

---
Commit 8bf3f88f7 authored by Peter Hawkins<[email protected]>
Committed by Benoit Steiner<[email protected]>:
[TF:XLA] Add _XLASend and _XLARecv TF ops that wrap the XLA Send/Recv HLO ops.

PiperOrigin-RevId: 164124764

---
Commit 626d3200f authored by Peter Hawkins<[email protected]>
Committed by Benoit Steiner<[email protected]>:
[XLA] Add test blacklist mechanism for XLA C++ unit tests.

PiperOrigin-RevId: 164124423

---
Commit 359cc5f5e authored by A. Unique TensorFlower<[email protected]>
Committed by Benoit Steiner<[email protected]>:
Document dict ordering in nest and make it consistent with sonnet.

PiperOrigin-RevId: 164114335

---
Commit 05813b531 authored by A. Unique TensorFlower<[email protected]>
Committed by Benoit Steiner<[email protected]>:
Go: Update generated wrapper functions for TensorFlow ops.

PiperOrigin-RevId: 164089206

---
Commit c451f465d authored by Anna R<[email protected]>
Committed by Benoit Steiner<[email protected]>:
BEGIN_PUBLIC
Automated g4 rollback of changelist 164078808

PiperOrigin-RevId: 164318935

Author: benoitsteiner

extend/adding_an_op.md

@@ -178,9 +178,7 @@ suggested implementation is to:
    file, but the specialization for the GPUDevice is defined in a .cu.cc file,
    since it will be compiled with the CUDA compiler.
 
-<!--zippy-->
-
-Expand this to see the example implementation.
+Here is an example implementation.
 
 ```c++
 // example.h
@@ -307,8 +305,6 @@ template struct ExampleFunctor<GPUDevice, int32>;
 #endif  // GOOGLE_CUDA
 ```
 
-<!--endzippy-->
-
 ## Build the op library
 ### Compile the op using your system compiler (TensorFlow binary installation)
 
@@ -763,7 +759,7 @@ Your op registration now specifies that the input's type must be `float`, or
 >   """
 > ```
 
-<pre><pre class="prettyprint"><code class="lang-cpp">
+<pre class="prettyprint"><code class="lang-cpp">
 \#include "tensorflow/core/framework/op_kernel.h"<br/>
 class ZeroOut<b>Int32</b>Op : public OpKernel {
   // as before
@@ -803,7 +799,7 @@ REGISTER\_KERNEL\_BUILDER(
     .Device(DEVICE\_CPU)
     .TypeConstraint&lt;float&gt;("T"),
     ZeroOutFloatOp);
-</b></code></pre></pre>
+</b></code></pre>
 
 > To preserve [backwards compatibility](#backwards-compatibility), you should
 > specify a [default value](#default-values-constraints) when adding an attr to