platform/upstream/tensorflow.git
6 years ago Add NCHW_VECT_W tensor format.
A. Unique TensorFlower [Thu, 17 May 2018 23:59:50 +0000 (16:59 -0700)]
  Add NCHW_VECT_W tensor format.

PiperOrigin-RevId: 197074411

6 years agoEnable optimizations submitted during pre-NIPS freeze:
A. Unique TensorFlower [Thu, 17 May 2018 23:52:31 +0000 (16:52 -0700)]
Enable optimizations submitted during pre-NIPS freeze:
  * Snapshot to Identity promotion.
  * Removal of transposes pairs separated by chain.

PiperOrigin-RevId: 197073602

6 years agoboosted_trees: fixed the crash when eval/prediction is attempted with the initial...
Younghee Kwon [Thu, 17 May 2018 23:52:17 +0000 (16:52 -0700)]
boosted_trees: fixed the crash when eval/prediction is attempted with the initial checkpoints (at step 0) before training.

PiperOrigin-RevId: 197073582

6 years agoMisc changes found when debugging an unrelated problem.
Akshay Modi [Thu, 17 May 2018 23:52:07 +0000 (16:52 -0700)]
Misc changes found when debugging an unrelated problem.

- (pywrap_tfe.i) Improve error message for better debugging TFE_Py_Execute failures.
- (pywrap_tfe_src.cc) Accept _value of None
- (base.i) Remove unnecessary temporary

PiperOrigin-RevId: 197073571

6 years agoMake V1 metrics distributed-aware. Also fix a bug where assertAllClose
A. Unique TensorFlower [Thu, 17 May 2018 23:26:24 +0000 (16:26 -0700)]
Make V1 metrics distributed-aware. Also fix a bug where assertAllClose
was sometimes ignoring its `msg` parameter.

PiperOrigin-RevId: 197070234

6 years ago[XLA] Redesign: delete xla::Computation.
A. Unique TensorFlower [Thu, 17 May 2018 23:23:33 +0000 (16:23 -0700)]
[XLA] Redesign: delete xla::Computation.

PiperOrigin-RevId: 197069851

6 years agoDon't add branch prediction hints when compiling under nvcc.
Craig Citro [Thu, 17 May 2018 23:06:21 +0000 (16:06 -0700)]
Don't add branch prediction hints when compiling under nvcc.

As seen in #19203, the `__builtin_expect` compiler builtin isn't recognized as
a builtin in nvcc8, leading to compilation failures of the form

    ./tensorflow/core/kernels/gather_functor_gpu.cu.h(57): error: calling a __host__ function("__builtin_
expect") from a __global__ function("tensorflow::GatherOpKernel< ::Eigen::half, int, (bool)1> ") is n
ot allowed

when attempting to build TensorFlow.

This change fixes things by adding an additional check for `__NVCC__`, and
avoiding any branch prediction hints in that case.

PiperOrigin-RevId: 197067418

6 years agofixing some incorrect documentation in test cases.
Olivia Nordquist [Thu, 17 May 2018 22:49:12 +0000 (15:49 -0700)]
fixing some incorrect documentation in test cases.

PiperOrigin-RevId: 197064874

6 years agoIntroduce an "indexed array" analysis
Sanjoy Das [Thu, 17 May 2018 22:47:30 +0000 (15:47 -0700)]
Introduce an "indexed array" analysis

Context: we want to optimize computations hanging off of a embedding lookup from
a constant array.  For instance, consider:

  embedding = gather from a constant array using non-constant indices
  embedding_reshaped = reshape embedding
  embedding_reshaped_transposed = transpose embedding_reshaped
  result = dot(embedding_reshaped_transposed, constant)

In the graph above, depending on how the details work out, we may be able to
fold `result` into a gather from a precomputed constant array.  However, it is
inconvenient to get there by incremental rewrites -- it is probably not
profitable to rewrite embedding_reshaped or embedding_reshaped_transposed [0] as
embedding lookups but we get to "see" that the dot can be rewritten only after
rewriting the reshape and the transpose.

This analysis aims to make the optimization above more straightforward by
allowing a transformation pass (that uses this analysis) to query the analysis
to see if if `result` _can_ be represented as an embedding lookup.  If yes it
can then apply some profitability heuristics to decide if it is worth it to
rewrite it as one.  This suggested workflow gives us separation of concerns (the
legality of the rewrite is computed separately from its profitability) and, more
importantly, lets us "look ahead" and analyze the dot without rewriting its
operands.

The implementation is far from complete (most of the interesting bits are TODO)
but I wanted to get an early design review before I spent too much time on this.

[0] Under the assumption that transposing or reshaping are not expensive enough
to pay the price of keeping around a new potentially large constant (in
particular, some of these may have been equivalent to free bitcasts).

PiperOrigin-RevId: 197064648

6 years ago Add more test cases in function_test
Youlong Cheng [Thu, 17 May 2018 22:47:24 +0000 (15:47 -0700)]
  Add more test cases in function_test

PiperOrigin-RevId: 197064629

6 years agoDelete unused proto schema "graph_explorer.proto".
A. Unique TensorFlower [Thu, 17 May 2018 22:34:00 +0000 (15:34 -0700)]
Delete unused proto schema "graph_explorer.proto".

PiperOrigin-RevId: 197062616

6 years agoSupport combinations stored in a global variable in combinations.py-based tests.
Igor Saprykin [Thu, 17 May 2018 22:29:03 +0000 (15:29 -0700)]
Support combinations stored in a global variable in combinations.py-based tests.

The user's input is modified to give the test a better name.  If the user stores combinations in a variable and applies those combinations to more than one test, then the test case name will be a concatenation of of the names of the previous test cases.

PiperOrigin-RevId: 197061902

6 years ago[XLA] Use Expm1 in Elu/Selu
David Majnemer [Thu, 17 May 2018 22:28:33 +0000 (15:28 -0700)]
[XLA] Use Expm1 in Elu/Selu

exp(x) - 1 is best executed using the composed Expm1 operation as it is better
behaved when exp(x) is near 1.

PiperOrigin-RevId: 197061826

6 years agoMove runtime.{h,cc,_test.cc} into core/common_runtime/eager as attr_builder
Akshay Modi [Thu, 17 May 2018 22:28:09 +0000 (15:28 -0700)]
Move runtime.{h,cc,_test.cc} into core/common_runtime/eager as attr_builder

I'm not familiar with how the CMake build is set up but from the description
of the problem the dependency graph is coarser than Bazel's, so I think this
should fix #18925.

PiperOrigin-RevId: 197061764

6 years agoClean up CFG test suite
A. Unique TensorFlower [Thu, 17 May 2018 22:10:25 +0000 (15:10 -0700)]
Clean up CFG test suite

PiperOrigin-RevId: 197059149

6 years ago[XLA] Remove eq_shapes from Identical SlowPath since it is already checked in
Blake Hechtman [Thu, 17 May 2018 22:08:33 +0000 (15:08 -0700)]
[XLA] Remove eq_shapes from Identical SlowPath since it is already checked in
Identical.

PiperOrigin-RevId: 197058888

6 years agoSupport Bool in Cast (TFLite)
Yu-Cheng Ling [Thu, 17 May 2018 21:58:04 +0000 (14:58 -0700)]
Support Bool in Cast (TFLite)

PiperOrigin-RevId: 197056978

6 years agoAdded CSV parsing to the tf.data programmers guide
Rachel Lim [Thu, 17 May 2018 21:37:34 +0000 (14:37 -0700)]
Added CSV parsing to the tf.data programmers guide
Relevant to #278.

PiperOrigin-RevId: 197053723

6 years agoBasic usability fixes for RNNCell wrappers
Allen Lavoie [Thu, 17 May 2018 21:32:47 +0000 (14:32 -0700)]
Basic usability fixes for RNNCell wrappers

They weren't calling their parent constructors (for the Keras base Layer), so a bunch of their methods threw odd errors. There may still be issues, but hopefully not so blatent.

Fixes #19208. For real this time.

PiperOrigin-RevId: 197052962

6 years agoFix L2Normalization.
A. Unique TensorFlower [Thu, 17 May 2018 21:31:29 +0000 (14:31 -0700)]
Fix L2Normalization.

PiperOrigin-RevId: 197052728

6 years ago[TF:STREAM_EXECUTOR] move declaration of multi_platform_manager to header
Nick Desaulniers [Thu, 17 May 2018 21:22:49 +0000 (14:22 -0700)]
[TF:STREAM_EXECUTOR] move declaration of multi_platform_manager to header

PiperOrigin-RevId: 197051272

6 years agoAutomated g4 rollback of changelist 197026249
Allen Lavoie [Thu, 17 May 2018 21:11:14 +0000 (14:11 -0700)]
Automated g4 rollback of changelist 197026249

PiperOrigin-RevId: 197049255

6 years ago[TF:XLA] Do not rely on implementation-defined semantics of DynamicSlice.
Michael Kuperstein [Thu, 17 May 2018 20:37:57 +0000 (13:37 -0700)]
[TF:XLA] Do not rely on implementation-defined semantics of DynamicSlice.

ReverseSequence relies on DynamicSlice wrapping around, which is implementation-defined behavior, and is not guaranteed. Pad the input instead.

PiperOrigin-RevId: 197043307

6 years agoAllows users to specify allow_custom_ops when calling tf.contrib.lite.toco_convert().
A. Unique TensorFlower [Thu, 17 May 2018 20:12:51 +0000 (13:12 -0700)]
Allows users to specify allow_custom_ops when calling tf.contrib.lite.toco_convert().

PiperOrigin-RevId: 197039477

6 years agoRemove -DGEMMLOWP_ALLOW_SLOW_SCALAR_FALLBACK being added by default to all builds.
Michael Case [Thu, 17 May 2018 20:02:08 +0000 (13:02 -0700)]
Remove -DGEMMLOWP_ALLOW_SLOW_SCALAR_FALLBACK being added by default to all builds.

PiperOrigin-RevId: 197037867

6 years agoSupport functools.partial as callable object in tf_inspect.getargspec.
A. Unique TensorFlower [Thu, 17 May 2018 19:55:02 +0000 (12:55 -0700)]
Support functools.partial as callable object in tf_inspect.getargspec.

PiperOrigin-RevId: 197036874

6 years ago[XLA] Redesign: delete Client::LoadSnapeshot(SessionModule). This is a precondition...
A. Unique TensorFlower [Thu, 17 May 2018 19:31:17 +0000 (12:31 -0700)]
[XLA] Redesign: delete Client::LoadSnapeshot(SessionModule). This is a precondition to delete xla::Computation.

PiperOrigin-RevId: 197033641

6 years agoTest some distributions stuff in Eager as well as Graph
Akshay Modi [Thu, 17 May 2018 19:29:54 +0000 (12:29 -0700)]
Test some distributions stuff in Eager as well as Graph

PiperOrigin-RevId: 197033485

6 years agoInternal Change
A. Unique TensorFlower [Thu, 17 May 2018 18:52:49 +0000 (11:52 -0700)]
Internal Change

PiperOrigin-RevId: 197028096

6 years agoSupport 1x1x1xN bias sizes in TFLite's convolution and FC layers.
A. Unique TensorFlower [Thu, 17 May 2018 18:47:16 +0000 (11:47 -0700)]
Support 1x1x1xN bias sizes in TFLite's convolution and FC layers.

PiperOrigin-RevId: 197027135

6 years agoInternal change
A. Unique TensorFlower [Thu, 17 May 2018 18:42:18 +0000 (11:42 -0700)]
Internal change

PiperOrigin-RevId: 197026249

6 years agoRemove C API staging from importer.py.
Skye Wanderman-Milne [Thu, 17 May 2018 18:33:36 +0000 (11:33 -0700)]
Remove C API staging from importer.py.

PiperOrigin-RevId: 197024708

6 years agoFix the fizzbuzz example
Alexandre Passos [Thu, 17 May 2018 18:17:58 +0000 (11:17 -0700)]
Fix the fizzbuzz example

PiperOrigin-RevId: 197021930

6 years agoUse integral power function rather than floating point version. The integral version...
A. Unique TensorFlower [Thu, 17 May 2018 18:13:00 +0000 (11:13 -0700)]
Use integral power function rather than floating point version. The integral version is faster.

PiperOrigin-RevId: 197021020

6 years ago[XLA:GPU] Unroll multi-output loop fusions
Benjamin Kramer [Thu, 17 May 2018 18:06:05 +0000 (11:06 -0700)]
[XLA:GPU] Unroll multi-output loop fusions

This is easier than I thought because we can assume that all tuple members have
the same number of elements. LLVM doesn't do a great job of vectorizing the
resulting stores, but otherwise this is working fine.

PiperOrigin-RevId: 197019718

6 years agoChange loop variable type to be deduced.
A. Unique TensorFlower [Thu, 17 May 2018 17:56:36 +0000 (10:56 -0700)]
Change loop variable type to be deduced.

PiperOrigin-RevId: 197017789

6 years agoChange traverse_test.test_module to traverse a constructed dummy module rather than...
A. Unique TensorFlower [Thu, 17 May 2018 17:15:45 +0000 (10:15 -0700)]
Change traverse_test.test_module to traverse a constructed dummy module rather than testcase itself.

PiperOrigin-RevId: 197010681

6 years agoAllows the fizzbuzz example to work when called as fizzbuzz(tf.constant(10)).
Alexandre Passos [Thu, 17 May 2018 17:02:06 +0000 (10:02 -0700)]
Allows the fizzbuzz example to work when called as fizzbuzz(tf.constant(10)).

Fixes #18960

PiperOrigin-RevId: 197008373

6 years agoRename private push/pop API and use from `stop_recording` method.
Tom Hennigan [Thu, 17 May 2018 16:57:16 +0000 (09:57 -0700)]
Rename private push/pop API and use from `stop_recording` method.

PiperOrigin-RevId: 197007561

6 years ago[TF:XLA] remove re-initializations of Literals
Nick Desaulniers [Thu, 17 May 2018 16:54:17 +0000 (09:54 -0700)]
[TF:XLA] remove re-initializations of Literals

It's an antipattern to have:

auto x = Literal::CreateFromShape(my_shape);
x->Populate();

as that results in initialization followed by reinitialization. Can be replaced
with:

auto x = MakeUnique<Literal>(my_shape);
x->Populate();

Suggested-by: Kay Zhu <kayzhu@google.com>
PiperOrigin-RevId: 197007127

6 years agoImprove the error message printed when a WorkerService::GetStatus() call fails on...
Derek Murray [Thu, 17 May 2018 16:32:02 +0000 (09:32 -0700)]
Improve the error message printed when a WorkerService::GetStatus() call fails on session creation.

PiperOrigin-RevId: 197003951

6 years agoImprovements to function._FuncGraph.
Skye Wanderman-Milne [Thu, 17 May 2018 16:28:35 +0000 (09:28 -0700)]
Improvements to function._FuncGraph.

* Adds 'inputs', 'outputs', and 'name' field to _FuncGraph. This
  allows _FuncGraph to encapsulate all the information needed to
  convert it to a FunctionDef.
* Refactor logic for converting a Python callable to a _FuncGraph into
  a new method, func_graph_from_py_func().

These changes are in preparation for converting tf.cond to emit an If
op. By exposing _FuncGraph functionality outside of _DefinedFunction,
_FuncGraphs can be used to represent functions that are manipulated
(e.g. to output intermediate tensors) before being converted to
FunctionDef protos.

PiperOrigin-RevId: 197003496

6 years agoRemove misleading declaration-as-default that results in a deleted constructor, and...
A. Unique TensorFlower [Thu, 17 May 2018 16:26:18 +0000 (09:26 -0700)]
Remove misleading declaration-as-default that results in a deleted constructor, and a misguided comment.

PiperOrigin-RevId: 197003162

6 years ago[XLA] Adds HloLivenessAnalysis and HloModuleDCE.
A. Unique TensorFlower [Thu, 17 May 2018 16:25:14 +0000 (09:25 -0700)]
[XLA] Adds HloLivenessAnalysis and HloModuleDCE.
HloLivenessAnalysis marks all live instruction outputs (i.e. tuple elements) for all instructions in an HloModule, propagating live values across computation boundaries.
HloModuleDCE sweeps through each instructions dead tuple elements, eliminating dead code (currently removes dead tuple elements from while loops, but could be extended to do the same for call instructions).

PiperOrigin-RevId: 197003043

6 years agoUpdate SessionTest.testFeedShapeCompatibility to work with C API enabled.
Skye Wanderman-Milne [Thu, 17 May 2018 16:22:24 +0000 (09:22 -0700)]
Update SessionTest.testFeedShapeCompatibility to work with C API enabled.

This test got lost in the transition. Prior to enabling the C API,
some constant node whose values were used for shape inference would be
marked as unfeedable in tensor_util.constant_value
(https://github.com/tensorflow/tensorflow/blob/r1.8/tensorflow/python/framework/tensor_util.py#L810).

This shape inference path is no longer used with the C API enabled, so
the constant node is successfully fed, triggering a runtime shape error.

This is arguably a regression, but given that the Python code wouldn't
mark all nodes evaluated during shape inference as unfeedable, it
seems ok to relax the check a little more.

PiperOrigin-RevId: 197002741

6 years agoAvoid accessing platform/default directly.
A. Unique TensorFlower [Thu, 17 May 2018 16:11:16 +0000 (09:11 -0700)]
Avoid accessing platform/default directly.

PiperOrigin-RevId: 197001347

6 years agoFixed bug in tf.pad shape logic that made it more restrictive than necessary by inter...
A. Unique TensorFlower [Thu, 17 May 2018 15:52:36 +0000 (08:52 -0700)]
Fixed bug in tf.pad shape logic that made it more restrictive than necessary by interpreting 0 as False.

PiperOrigin-RevId: 196998883

6 years agoMethods to stop and reset tf.GradientTape()
Alexandre Passos [Thu, 17 May 2018 15:23:10 +0000 (08:23 -0700)]
Methods to stop and reset tf.GradientTape()

PiperOrigin-RevId: 196995160

6 years ago[TF:XLA] Bump open source llvm revision to r332584
Benjamin Kramer [Thu, 17 May 2018 15:00:54 +0000 (08:00 -0700)]
[TF:XLA] Bump open source llvm revision to r332584

PiperOrigin-RevId: 196992500

6 years agoDrop some old dump_graphviz options
A. Unique TensorFlower [Thu, 17 May 2018 14:34:24 +0000 (07:34 -0700)]
Drop some old dump_graphviz options

PiperOrigin-RevId: 196989899

6 years agoAdapt LLVM ORC interface change in r332541.
Eric Liu [Thu, 17 May 2018 12:57:31 +0000 (05:57 -0700)]
Adapt LLVM ORC interface change in r332541.

PiperOrigin-RevId: 196978634

6 years agoAdd evaluation metrics and export results in the new train_and_evaluate API (for...
A. Unique TensorFlower [Thu, 17 May 2018 09:47:24 +0000 (02:47 -0700)]
Add evaluation metrics and export results in the new train_and_evaluate API (for local mode).

PiperOrigin-RevId: 196962253

6 years agoMove BUILD file to OVIC folder and add model validation.
A. Unique TensorFlower [Thu, 17 May 2018 05:33:38 +0000 (22:33 -0700)]
Move BUILD file to OVIC folder and add model validation.

PiperOrigin-RevId: 196940252

6 years agoUpdate installation documentation to reflect that CUDA 8 and cuDNN 6 are minimal...
A. Unique TensorFlower [Thu, 17 May 2018 05:22:36 +0000 (22:22 -0700)]
Update installation documentation to reflect that CUDA 8 and cuDNN 6 are minimal supported versions.
Remove GPU instructions for MacOS because GPUs are not supported anymore.

PiperOrigin-RevId: 196939548

6 years agoSort tags before logging.
A. Unique TensorFlower [Thu, 17 May 2018 04:40:06 +0000 (21:40 -0700)]
Sort tags before logging.

PiperOrigin-RevId: 196936678

6 years ago[tf.data] Accept NumPy dtype objects in `Dataset.from_generator(..., output_types...
Derek Murray [Thu, 17 May 2018 04:15:44 +0000 (21:15 -0700)]
[tf.data] Accept NumPy dtype objects in `Dataset.from_generator(..., output_types=...)`.

PiperOrigin-RevId: 196935179

6 years agoAdd more logging in BaseGPUDevice::ComputeHelper for kernel completion.
James Qin [Thu, 17 May 2018 03:52:32 +0000 (20:52 -0700)]
Add more logging in BaseGPUDevice::ComputeHelper for kernel completion.

PiperOrigin-RevId: 196933479

6 years agoMaking GetOptionalInput from kernel_util.h return a pointer to const data.
A. Unique TensorFlower [Thu, 17 May 2018 03:31:29 +0000 (20:31 -0700)]
Making GetOptionalInput from kernel_util.h return a pointer to const data.

PiperOrigin-RevId: 196932028

6 years ago[TF:XLA] Bump open source llvm revision to r332236
Sanjoy Das [Thu, 17 May 2018 03:15:53 +0000 (20:15 -0700)]
[TF:XLA] Bump open source llvm revision to r332236

PiperOrigin-RevId: 196930996

6 years agoInternal change.
Justine Tunney [Thu, 17 May 2018 02:27:28 +0000 (19:27 -0700)]
Internal change.

PiperOrigin-RevId: 196927869

6 years agoEnhance DenseLayer + XLA compatibility test cases to cover compilation behavior diffe...
A. Unique TensorFlower [Thu, 17 May 2018 02:12:18 +0000 (19:12 -0700)]
Enhance DenseLayer + XLA compatibility test cases to cover compilation behavior differences in different jit modes.

PiperOrigin-RevId: 196926896

6 years agoAppend device name in executor logging
James Qin [Thu, 17 May 2018 01:41:39 +0000 (18:41 -0700)]
Append device name in executor logging

PiperOrigin-RevId: 196924318

6 years ago[TF:XLA] Make noinline function work with control flow.
A. Unique TensorFlower [Thu, 17 May 2018 01:23:20 +0000 (18:23 -0700)]
[TF:XLA] Make noinline function work with control flow.
1) Make the local function library for control flow self-contained. The control flow function could refer to a noinline function not defined in the local library. Copy the missing FunctionDefs from the glocal library to the local one.
2) Fix the index used to get the output shapes for functional nodes.

PiperOrigin-RevId: 196922649

6 years ago[XLA] Remove XlaOp::GetShape. It only works when the buidler of the XlaOp is not...
A. Unique TensorFlower [Thu, 17 May 2018 01:13:00 +0000 (18:13 -0700)]
[XLA] Remove XlaOp::GetShape. It only works when the buidler of the XlaOp is not freeed.

PiperOrigin-RevId: 196921647

6 years agoRemove _USE_C_API staging in tests now that the C API is enabled by default.
Skye Wanderman-Milne [Thu, 17 May 2018 01:05:52 +0000 (18:05 -0700)]
Remove _USE_C_API staging in tests now that the C API is enabled by default.

This is in preparation for removing the _USE_C_API toggle altogether.

PiperOrigin-RevId: 196920890

6 years agoRemove _USE_C_API staging in tests now that the C API is enabled by default.
Skye Wanderman-Milne [Thu, 17 May 2018 01:03:01 +0000 (18:03 -0700)]
Remove _USE_C_API staging in tests now that the C API is enabled by default.

This is in preparation for removing the _USE_C_API toggle altogether.

PiperOrigin-RevId: 196920481

6 years agoFix typo in TensorHandle
Igor Ganichev [Thu, 17 May 2018 00:48:22 +0000 (17:48 -0700)]
Fix typo in TensorHandle

PiperOrigin-RevId: 196919119

6 years agoRe-enabling a test after a previous fix.
Jiri Simsa [Thu, 17 May 2018 00:25:36 +0000 (17:25 -0700)]
Re-enabling a test after a previous fix.

PiperOrigin-RevId: 196916467

6 years agoRemove no-op statement. tf_additional_lib_srcs only selects .cc files. When we
A. Unique TensorFlower [Thu, 17 May 2018 00:24:39 +0000 (17:24 -0700)]
Remove no-op statement. tf_additional_lib_srcs only selects .cc files. When we
do tf_additional_lib_srcs(exclude=[**/*.cc]) we are selecting zero files, and
the statement can be safely removed.

PiperOrigin-RevId: 196916359

6 years agoAdds basic TPU replicate training support for Keras.
Jianwei Xie [Thu, 17 May 2018 00:23:02 +0000 (17:23 -0700)]
Adds basic TPU replicate training support for Keras.

PiperOrigin-RevId: 196916177

6 years ago[XLA] Improve documentation on HloModule, HloComputation, and HloInstruction.
Justin Lebar [Thu, 17 May 2018 00:21:22 +0000 (17:21 -0700)]
[XLA] Improve documentation on HloModule, HloComputation, and HloInstruction.

PiperOrigin-RevId: 196915982

6 years ago[XLA] Add documentation explaining FusionKind.
Justin Lebar [Thu, 17 May 2018 00:09:42 +0000 (17:09 -0700)]
[XLA] Add documentation explaining FusionKind.

PiperOrigin-RevId: 196914484

6 years agoRemove unused inclusions
A. Unique TensorFlower [Thu, 17 May 2018 00:05:33 +0000 (17:05 -0700)]
Remove unused inclusions

PiperOrigin-RevId: 196913890

6 years ago[XLA:GPU] Add op-tracing to XLA:GPU.
Justin Lebar [Wed, 16 May 2018 23:56:48 +0000 (16:56 -0700)]
[XLA:GPU] Add op-tracing to XLA:GPU.

PiperOrigin-RevId: 196912575

6 years agoAllow for remote eager execution.
Akshay Modi [Wed, 16 May 2018 23:43:29 +0000 (16:43 -0700)]
Allow for remote eager execution.

PiperOrigin-RevId: 196910675

6 years agoRemove unused inclusions
A. Unique TensorFlower [Wed, 16 May 2018 23:16:46 +0000 (16:16 -0700)]
Remove unused inclusions

PiperOrigin-RevId: 196906815

6 years agoMove DoesNotUseOperandBuffer and CanShareOperandBufferWithUser from
Jeremy Lau [Wed, 16 May 2018 22:54:49 +0000 (15:54 -0700)]
Move DoesNotUseOperandBuffer and CanShareOperandBufferWithUser from
liveness_util to methods on TuplePointsToAnalysis and HloDataflowAnalysis.

PiperOrigin-RevId: 196903216

6 years ago[TF:XLA] Take subcomputations into account during HLO scheduling.
Dimitris Vardoulakis [Wed, 16 May 2018 22:53:34 +0000 (15:53 -0700)]
[TF:XLA] Take subcomputations into account during HLO scheduling.

In the List scheduler, if an instruction calls subcomputations, we count the memory usage of the subcomputation towards the memory usage of the parent instruction.

PiperOrigin-RevId: 196903042

6 years agoFixing test for Topk kernel in TFlite
A. Unique TensorFlower [Wed, 16 May 2018 22:28:11 +0000 (15:28 -0700)]
Fixing test for Topk kernel in TFlite

PiperOrigin-RevId: 196899232

6 years agoMake sparse_cross operations publicly available.
Mustafa Ispir [Wed, 16 May 2018 22:27:34 +0000 (15:27 -0700)]
Make sparse_cross operations publicly available.

PiperOrigin-RevId: 196899145

6 years agoAdd test for 64-bit clz and sign.
Jacques Pienaar [Wed, 16 May 2018 22:02:06 +0000 (15:02 -0700)]
Add test for 64-bit clz and sign.

PiperOrigin-RevId: 196894702

6 years agoFix typo in comment
Sanjoy Das [Wed, 16 May 2018 22:01:22 +0000 (15:01 -0700)]
Fix typo in comment

PiperOrigin-RevId: 196894582

6 years agoAdd a parameter to the adaptive shared batcher which allows the user to set a lower...
A. Unique TensorFlower [Wed, 16 May 2018 21:53:11 +0000 (14:53 -0700)]
Add a parameter to the adaptive shared batcher which allows the user to set a lower bound for in_flight_batches_limit.

This can help prevent overloads which may occur during large traffic shifts - a small value learned during a period of low load can be unsuitable at high load.

PiperOrigin-RevId: 196893320

6 years agoCheckpointable: move python/training/checkpointable_* to python/training/checkpointable/
Allen Lavoie [Wed, 16 May 2018 20:52:54 +0000 (13:52 -0700)]
Checkpointable: move python/training/checkpointable_* to python/training/checkpointable/

Need to add some new checkpointable files in core (specifically I had some checkpointable data structures in mind), and prefixing more files with "checkpointable_" in python/training/ seems dirty.

No functional changes, just some branching and build/import fiddling.

PiperOrigin-RevId: 196883136

6 years agoAutomated g4 rollback of changelist 196691101
Peter Hawkins [Wed, 16 May 2018 20:34:10 +0000 (13:34 -0700)]
Automated g4 rollback of changelist 196691101

PiperOrigin-RevId: 196879933

6 years agoBUILD cleanup in contrib/lite/...
A. Unique TensorFlower [Wed, 16 May 2018 20:27:49 +0000 (13:27 -0700)]
BUILD cleanup in contrib/lite/...

PiperOrigin-RevId: 196878865

6 years agoFix the gradient of reduce_prod for complex dtypes.
Brian Patton [Wed, 16 May 2018 20:22:53 +0000 (13:22 -0700)]
Fix the gradient of reduce_prod for complex dtypes.

Fixes #12514

PiperOrigin-RevId: 196878148

6 years agoRemove sorted as types not sortable.
Jacques Pienaar [Wed, 16 May 2018 20:12:25 +0000 (13:12 -0700)]
Remove sorted as types not sortable.

PiperOrigin-RevId: 196876502

6 years agoFix broken link.
A. Unique TensorFlower [Wed, 16 May 2018 19:54:32 +0000 (12:54 -0700)]
Fix broken link.

PiperOrigin-RevId: 196873792

6 years agoAdd a test for compiled tfe.defun in GradientTape
Igor Ganichev [Wed, 16 May 2018 19:50:41 +0000 (12:50 -0700)]
Add a test for compiled tfe.defun in GradientTape

PiperOrigin-RevId: 196873235

6 years agoRemove redundant initialization of collective params.
Ayush Dubey [Wed, 16 May 2018 19:32:04 +0000 (12:32 -0700)]
Remove redundant initialization of collective params.

subdiv_permutations is being resized twice in GenerateSubdivParams.

PiperOrigin-RevId: 196870781

6 years agoUse sequence_length arg for dynamic_rnn within RNNEstimator
A. Unique TensorFlower [Wed, 16 May 2018 19:24:59 +0000 (12:24 -0700)]
Use sequence_length arg for dynamic_rnn within RNNEstimator

This does not affect correctness, but improves performance by skipping padded
parts of the sequence. Also correct documentation in rnn.py that states the
opposite.

PiperOrigin-RevId: 196869793

6 years ago[TF:XLA:CPU] enable s32 reduce-window
Nick Desaulniers [Wed, 16 May 2018 19:21:35 +0000 (12:21 -0700)]
[TF:XLA:CPU] enable s32 reduce-window

PiperOrigin-RevId: 196869296

6 years agoFix the CCFLAGS mismatch.
A. Unique TensorFlower [Wed, 16 May 2018 19:16:33 +0000 (12:16 -0700)]
Fix the CCFLAGS mismatch.

PiperOrigin-RevId: 196868601

6 years agoExpand tests to include int64 output type.
Jacques Pienaar [Wed, 16 May 2018 19:15:37 +0000 (12:15 -0700)]
Expand tests to include int64 output type.

PiperOrigin-RevId: 196868485

6 years agoTurn off MirroredStrategy Dataset prefetching in tests when using the
A. Unique TensorFlower [Wed, 16 May 2018 19:04:13 +0000 (12:04 -0700)]
Turn off MirroredStrategy Dataset prefetching in tests when using the
combinations library. It adds some small non-determinism to the input
batches which can make tests flaky.

Also add a default DistributionStrategy combination.

PiperOrigin-RevId: 196866569

6 years agoInternal Change.
Michael Case [Wed, 16 May 2018 18:51:48 +0000 (11:51 -0700)]
Internal Change.

PiperOrigin-RevId: 196864489

6 years ago[XLA:GPU] Emit the final write of the tuple pointers
Benjamin Kramer [Wed, 16 May 2018 18:45:56 +0000 (11:45 -0700)]
[XLA:GPU] Emit the final write of the tuple pointers

Turns out this doesn't matter when the fusion is emitted as a kernel, but does
when the whole thing is inlined. Oops.

PiperOrigin-RevId: 196863545

6 years agoFixes tflite_diff script.
Nupur Garg [Wed, 16 May 2018 17:46:46 +0000 (10:46 -0700)]
Fixes tflite_diff script.

PiperOrigin-RevId: 196852157

6 years ago[XLA:GPU] Teach ir_emitter_nested how to deal with multi output loop fusion
Benjamin Kramer [Wed, 16 May 2018 17:40:57 +0000 (10:40 -0700)]
[XLA:GPU] Teach ir_emitter_nested how to deal with multi output loop fusion

Most of the plumbing is there already, just set up a loop emitter with a target
for each tuple element. For a simple case the output looks reasonable, though I
haven't checked correctness of anything complex.

PiperOrigin-RevId: 196850926