Update version to rc1.

Cherrypicks

PiperOrigin-RevId: 162379684

Format 1.3.0 release notes; added a new feature to tfdbg section

Fixes #11666.

PiperOrigin-RevId: 162776187

is specified but output_buffer_size is not:

ValueError: Tensor conversion requested dtype int64 for Tensor with dtype
int32: 'Tensor("num_threads:0", shape=(), dtype=int32)'
PiperOrigin-RevId: 162766317

preserving them when they're on the sender side.

PiperOrigin-RevId: 162698209

The base_rendezvous_mgr handles transferring a tensor using DMAs to non-"host"
devices such as GPUs in the SameWorkerRecvDone function. This function would
use the worker_env's device_mgr to obtain a pointer to the relevant Device
(using the LookupDevice call). In the ClusterSpec-propagation world, using the
environment's device_set means that the devices are not renamed, often
resulting in devices not being found.

This change updates BaseRemoteRendezvous to use the WorkerSession stored when
the BaseRemoteRendezvous is initialized. The WorkerSession has a pointer to a
DeviceMgr that contains the appropriately renamed devices for the given
session the Rendezvous is associated with.

Note: because we have a fast-path host-device-only copy, the original bug does
not show up when using 2 CPU devices. I have added a test to ensure that
transferring between 2 CPU devices works in a ClusterSpec propagation session,
but note that this test does not actually reproduce the motivating bug.

In the process of writing a test for the original bug, I discovered another
latent bug in ClusterSpec propagation where if there were 2 CPU devices
(i.e. due to explicit server configuration to have 2 CPU devices), a DCHECK
could be triggered. The Master::CreateSession would call
`device_set->set_client_device` multiple times (once for each CPU device).

PiperOrigin-RevId: 162680163

The display_name will be used to display the series in TensorBoard, in lieu of the tag, assuming that it is specified. (When it is not specified, behavior will stay the same.)
The summary_description will allow the user to write a longform readable description of the summary series for display in TensorBoard. Markdown will be supported.

This will make it possible for TensorBoard-2 summary ops to give the user direct control over the display name, solving https://github.com/tensorflow/tensorboard/issues/59.

PiperOrigin-RevId: 162566261

Add multi-head attention capabilities to AttentionWrapper via the specification of multiple AttentionMechanisms.

PiperOrigin-RevId: 162557562

PiperOrigin-RevId: 162523648

These are implemented using GPU-capable ops and have gradient support.

PiperOrigin-RevId: 162511440

PiperOrigin-RevId: 162418535

PiperOrigin-RevId: 162406054

PiperOrigin-RevId: 162392717

Related issues: #11077, #11540.

PiperOrigin-RevId: 162362081

Cherry-pick GradientsDebugger from master to r1.3, with fixes to GPU build issues

Replace it with a new hidden op: array_ops.DebugGradientIdentity

PiperOrigin-RevId: 162983526

Also updates programmer_guide/debugger.md

RELNOTES: TensorFlow Debugger (tfdbg) CLI: added ability to display feed_dict values to Session.run through the "pf" / "print_feed" command and clickable links in the curses UI.
PiperOrigin-RevId: 162648456

to allow retrieval of gradient tensors created by TensorFlow's automatic differentiation algorithm (i.e., tf.gradients and optimizer code that uses it).

PiperOrigin-RevId: 161805516

Add a note about tf.contrib.signal to the 1.3 release notes.

Add new feature description for tfdbg to r1.3 release notes

Updating RELEASE notes.

Updating the RELNOTES for 1.3.0rc0.

Revert commits 1.3.0rc0

This reverts commit 31f1375c.