scalar shapes during construction.  Placholder/Variable currently
treat [] as "unknown", because it predated scalar shapes.  So
we introduce new versions that are the same implementation underneath,
but can differentiate between unknown and scalar.
Change: 136659818

On mac 64-bit, the library paths is actually at /usr/local/cuda/lib by default. The assumption that it is always at /usr/local/cuda/lib64 has led to build failures, e.g.,
http://ci.tensorflow.org/view/Nightly/job/nightly-matrix-mac-gpu/100/TF_BUILD_IS_OPT=OPT,TF_BUILD_IS_PIP=PIP,TF_BUILD_PYTHON_VERSION=PYTHON2,label=gpu-mac/console
Change: 136653210

Before this CL, Device::Sync() was called only when the executor graph contains a SINK node. However the SINK node in the graph is guaranteed to be reachable from SOURCE. This CL makes sure Device::Sync() is always called before the executor exits.

As a separate issue, we may consider if we want the executor graph to be "well-formed" with a single SOURCE and a single SINK.
Change: 136652586

Before this change, the EventAccumulator parses the new scalar_summaries as
a one-element np array like this: np.array(3, dtype=np.float32).

This is the same as a number wrt equality checking (i.e. expression above is
equal to 3.0), but it isn't json serializable, which causes problems for
TensorBoard.
Change: 136647129

Change: 136645313

Change: 136644858

Change: 136644641

Change: 136643977

Change: 136641403

the second best split in finished_nodes_op.
Change: 136640840

Change: 136637890

Create tf.summary.{image, histogram, audio, merge, merge_all}

These duplicate tf.{image, histogram, audio, merge, merge_all}_summary.
We will deprecate the existing versions in a future change.

The APIs for the migrated summary ops are almost the same. Here are the changes:
- The first argument for each summary op is now "name" rather than "tag". The "tag" argument is removed entirely. The name will be used to construct the op name per standard TensorFlow naming conventions, meaning there is no longer a need to do manual name-prefixing to get re-usable summary ops.
- The image summary "max_images" argument was renamed to "max_outputs", for consistency with the audio summary and other future summaries.

A future change will migrate existing usage where possible, and remove the duplicated implementations in logging_ops.py
Change: 136636253

Change: 136632035

Change: 136631708

Change: 136630722

Change: 136626773

Change: 136626094

Change: 136624599

This fixes a build failure that surfaced in nightly builds after CL/136212902 such as:
https://ci.tensorflow.org/view/Tensorflow%20Jenkins%20Monitored%20builds/job/nightly-matrix-cpu/TF_BUILD_IS_OPT=OPT,TF_BUILD_IS_PIP=PIP,TF_BUILD_PYTHON_VERSION=PYTHON2,label=cpu-slave/261/console
Change: 136622955

Change: 136621502

Change: 136619012

Closes #4566
Change: 136617356

Change: 136615121

Change: 136614898

Change: 136613320

Change: 136613296

Change: 136609289

Change: 136604297

This speeds up performance of passing SparseTensor objects through queues
about 2x.
Change: 136602071

Change: 136599724

Change: 136597998

Change: 136596757

Change: 136566426

Change: 136565493

Many languages, e.g. JS, proto, etc. don't support having hyphens in
package names.
Change: 136565198

Change: 136560739

If I check out TensorFlow to a directory named tensorflow-clean
then Bazel is going to create a subdirectory named
bazel-tensorflow-clean. So we really need to have a wildcard
pattern.
Change: 136559940

Change: 136558601

This adds support for "NC*" data layouts for N-D convolution and
pooling (including atrous convolution and pooling); previously, only
"N*C" data lyaouts were supported.

This also adds support for 1-D pooling (by forwarding to the 2-D
implementation), and fixes the broken data_format support in
conv1d.
Change: 136556507

Change: 136556197