Anna R [Thu, 8 Feb 2018 19:40:28 +0000 (11:40 -0800)]
Adding tf_export decorators/calls to TensorFlow functions and constants.
PiperOrigin-RevId:
185026527
Yu-Cheng Ling [Thu, 8 Feb 2018 19:40:05 +0000 (11:40 -0800)]
Changing ".lite" to ".tflite" in documents.
PiperOrigin-RevId:
185026484
A. Unique TensorFlower [Thu, 8 Feb 2018 19:31:09 +0000 (11:31 -0800)]
Update llvm revision to r324405, which introduces an Orc API change.
PiperOrigin-RevId:
185025122
Derek Murray [Thu, 8 Feb 2018 19:28:55 +0000 (11:28 -0800)]
[tf.data] Remove deprecated `tf.contrib.data.Iterator` alias.
This change removes the following class:
* `tf.contrib.data.Iterator`.
IF THIS BREAKS YOU: Replace `tf.contrib.data.Iterator` with `tf.data.Iterator`
when explicitly constructing an iterator. The API for the resulting object is
identical.
PiperOrigin-RevId:
185024771
A. Unique TensorFlower [Thu, 8 Feb 2018 19:20:39 +0000 (11:20 -0800)]
Preserving order when removing nodes.
PiperOrigin-RevId:
185023366
Akshay Modi [Thu, 8 Feb 2018 18:42:01 +0000 (10:42 -0800)]
Update the wrapper gen code to call out to the fastpath function.
PiperOrigin-RevId:
185016979
Francois Chollet [Thu, 8 Feb 2018 18:41:37 +0000 (10:41 -0800)]
Fix minor bug with tf.keras model calling (related to mask caching).
PiperOrigin-RevId:
185016910
Derek Murray [Thu, 8 Feb 2018 18:39:32 +0000 (10:39 -0800)]
[tf.data] Remove deprecated reader dataset classes in `tf.contrib.data`.
This change removes the following classes:
* `tf.contrib.data.FixedLengthRecordDataset`
* `tf.contrib.data.TextLineDataset`
* `tf.contrib.data.TFRecordDataset`
IF THIS BREAKS YOU: Replace `tf.contrib.data` with `tf.data` when constructing a
`FixedLengthRecordDataset`, `TextLineDataset`, or `TFRecordDataset`. Note that
you may have to modify downstream transformations to use the core API. See
"tensorflow/contrib/data/README.md" for details of how to update your code to use
the core API.
PiperOrigin-RevId:
185016587
A. Unique TensorFlower [Thu, 8 Feb 2018 18:37:40 +0000 (10:37 -0800)]
Update for LLVM API change r324405
PiperOrigin-RevId:
185016276
Bixia Zheng [Thu, 8 Feb 2018 18:22:16 +0000 (10:22 -0800)]
Remove cudnn_type parameter from a few template member functions in class
CudnnSupport.
The cudnn_type value in those functions is implied by the device memory
type DeviceMemory<T> and is a compile-time constant.
PiperOrigin-RevId:
185013909
A. Unique TensorFlower [Thu, 8 Feb 2018 18:20:26 +0000 (10:20 -0800)]
Make MklCpuAllocator a VisitableAllocator, instead of just an Allocator.
This allows it to work more efficiently with RDMA networking.
PiperOrigin-RevId:
185013628
Suharsh Sivakumar [Thu, 8 Feb 2018 17:34:49 +0000 (09:34 -0800)]
Make quantization rewrites happen in place.
If no graph is provided, then the default graph is used.
PiperOrigin-RevId:
185007107
A. Unique TensorFlower [Thu, 8 Feb 2018 17:29:32 +0000 (09:29 -0800)]
Enable algebraic optimizations for operations with neutral and absorbing elements by default - not only when feeds are absent or in aggressive mode.
PiperOrigin-RevId:
185006374
A. Unique TensorFlower [Thu, 8 Feb 2018 17:03:30 +0000 (09:03 -0800)]
More updates:
* add a verbose flag, useful for debugging
* fix the canonicalization of old-style print statement
* fix broken py_func wrapper
* expand the conditional statement to return a dummy value if we cannot fine any return values, and call without an assignment so that the side effect guards catch it instead
* streamline the converter tests a bit more
* avoid aliasing "tf" and "self" in the side effect guards
* improve the namer to generate shorter names
PiperOrigin-RevId:
185002802
A. Unique TensorFlower [Thu, 8 Feb 2018 15:36:03 +0000 (07:36 -0800)]
Emitting type-dependent cond and while_loop code.
PiperOrigin-RevId:
184993166
Ilya Biryukov [Thu, 8 Feb 2018 12:18:40 +0000 (04:18 -0800)]
Patch cub on download to fix compilation error with clang.
The same patch was sent via PR to cub upstream:
https://github.com/NVlabs/cub/pull/125
PiperOrigin-RevId:
184975304
A. Unique TensorFlower [Thu, 8 Feb 2018 11:18:36 +0000 (03:18 -0800)]
Automated g4 rollback of changelist
184303789
PiperOrigin-RevId:
184970903
A. Unique TensorFlower [Thu, 8 Feb 2018 06:18:45 +0000 (22:18 -0800)]
Update ops-related pbtxt files.
PiperOrigin-RevId:
184947099
Francois Chollet [Thu, 8 Feb 2018 05:56:28 +0000 (21:56 -0800)]
Switch to using ResourceVariable in Keras for full Eager-mode compatibility.
PiperOrigin-RevId:
184945626
A. Unique TensorFlower [Thu, 8 Feb 2018 05:11:56 +0000 (21:11 -0800)]
Fixing the anchor link for the Math ops guide on Segmentation (it's a capital S but our docs link to a lowercase s).
PiperOrigin-RevId:
184942994
A. Unique TensorFlower [Thu, 8 Feb 2018 05:05:41 +0000 (21:05 -0800)]
Temporarily weaken Identity pruning in model_pruner while investigating test failure of //robotics/learning/sensor_predict:utils_multi_sensor_rnn_test.
PiperOrigin-RevId:
184942554
Mark Daoust [Thu, 8 Feb 2018 04:50:09 +0000 (20:50 -0800)]
Move TPU doc to "using_tpu". Add short titles to some docs.
PiperOrigin-RevId:
184941101
A. Unique TensorFlower [Thu, 8 Feb 2018 04:49:37 +0000 (20:49 -0800)]
Go: Update generated wrapper functions for TensorFlow ops.
PiperOrigin-RevId:
184941052
Derek Murray [Thu, 8 Feb 2018 04:17:11 +0000 (20:17 -0800)]
[tf.data] Move C++ code backing `tf.contrib.data.ignore_errors()` to contrib.
This change moves the `OpKernel` and `DatasetBase` implementations to
"tensorflow/contrib/data/kernels", where they are packaged as a custom
op library. This demonstrates (and enforces by continuous integration)
the ability to build a C++ Dataset implementation in a custom op
library.
Other contrib Dataset implementations will move in subsequent changes.
PiperOrigin-RevId:
184938885
Sanjoy Das [Thu, 8 Feb 2018 03:24:57 +0000 (19:24 -0800)]
[XLA:CPU] Fix tfcompile's use of freeze_graph.py
Not sure if I am holding it wrong, but I could not make the existing
FLAGS.check_version work.
PiperOrigin-RevId:
184935374
Yao Zhang [Thu, 8 Feb 2018 02:11:51 +0000 (18:11 -0800)]
Reduce the number of concats to avoid test timeout.
PiperOrigin-RevId:
184929151
A. Unique TensorFlower [Thu, 8 Feb 2018 01:16:32 +0000 (17:16 -0800)]
Remove note about accumulator variables, since those are not added to TRAINABLE_VARIABLES.
PiperOrigin-RevId:
184922273
Russell Power [Thu, 8 Feb 2018 01:12:01 +0000 (17:12 -0800)]
Add operation to forward log messages from remote workers to a local system.
PiperOrigin-RevId:
184921657
Sanjoy Das [Thu, 8 Feb 2018 00:48:45 +0000 (16:48 -0800)]
[XLA:CPU] Fix test case for vectorized Exp and Tanh to actually vectorize
I just noticed that the test case for ArrayElementwiseOpTest::ExpF32sVector and
possibly for ArrayElementwiseOpTest::ExpF32sVector does not actually vectorize
the intrinsic calls. This is most likely a very recent regression because I
remember fixing at least one issue in the emitter demonstrated by the test.
Despite that I think the current approach is better since we have unit tests
that check that we at least vectorize the vector-of-F32's case.
PiperOrigin-RevId:
184918373
A. Unique TensorFlower [Thu, 8 Feb 2018 00:34:57 +0000 (16:34 -0800)]
Update ops-related pbtxt files.
PiperOrigin-RevId:
184916250
Michael Case [Thu, 8 Feb 2018 00:26:55 +0000 (16:26 -0800)]
Disable internally failing array_ops test.
PiperOrigin-RevId:
184915141
Eugene Brevdo [Thu, 8 Feb 2018 00:24:34 +0000 (16:24 -0800)]
Remove tf.contrib.ndlstm as it is not maintained and barely used.
Users can find an external implementation by the original author at:
https://github.com/tmbarchive/tfndlstm
PiperOrigin-RevId:
184914822
A. Unique TensorFlower [Wed, 7 Feb 2018 23:08:24 +0000 (15:08 -0800)]
Go: Update generated wrapper functions for TensorFlow ops.
PiperOrigin-RevId:
184903402
Kay Zhu [Wed, 7 Feb 2018 22:51:03 +0000 (14:51 -0800)]
[XLA] Keep the number of HloPasses minimum for the interpreter backend.
This would be useful to identify bugs for these HLO passes.
PiperOrigin-RevId:
184900236
Michael Case [Wed, 7 Feb 2018 22:36:00 +0000 (14:36 -0800)]
Merge changes from github.
PiperOrigin-RevId:
184897758
Alexandre Passos [Wed, 7 Feb 2018 22:13:05 +0000 (14:13 -0800)]
Better documentation for contrib summaries.
Also all_summary_ops returns None in eager mode instead of error.
PiperOrigin-RevId:
184893777
Suharsh Sivakumar [Wed, 7 Feb 2018 21:47:02 +0000 (13:47 -0800)]
Generalize quantization rewrite to not rely on names.
It should now work with most graphs regardless if they were built slim or not.
PiperOrigin-RevId:
184889280
A. Unique TensorFlower [Wed, 7 Feb 2018 21:46:29 +0000 (13:46 -0800)]
Change output of slim.learning.train to total_loss = None, if no training step was actually performed.
PiperOrigin-RevId:
184889196
Derek Murray [Wed, 7 Feb 2018 21:44:30 +0000 (13:44 -0800)]
[tf.data] Move the C++ Dataset class implementations to the framework library.
This enables the use of the `DatasetOpKernel` subclasses in custom op library
code. A subsequent change will move `tf.contrib.data` kernel implementations
to a custom op library.
Implementation note: This change moves some classes from
"tensorflow/core/graph/..." into the framework library, which does not
include any code in "tensorflow/core/common_runtime/...". To break the
dependency from "tensorflow/core/framework/dataset.cc" to
"tensorflow/core/common_runtime/...", the `GraphDefBuilderToGraph()`
method has been split out from the `GraphDefBuilder` class (where it
was previously exposed as the `GraphDefBuilder::ToGraph()` utility
method) and added to a new
"tensorflow/core/graph/graph_def_builder_util.h" module. This method
depends on ".../graph/graph_constructor.cc", which depends directly on
".../common_runtime/shape_refiner.h" and indirectly on
".../common_runtime/graph_runner.h". Since this method was used only
in tests, these have been updated to point to the new utility method.
PiperOrigin-RevId:
184888903
Yu-Cheng Ling [Wed, 7 Feb 2018 21:39:01 +0000 (13:39 -0800)]
Makefile flag to use Apple Accelerate for Conv on iOS.
PiperOrigin-RevId:
184888096
A. Unique TensorFlower [Wed, 7 Feb 2018 20:32:53 +0000 (12:32 -0800)]
Add BatchMatMul support and improve tensorflow graphdef export by adding ops and fixing typing and resolving constant Transpose ops.
PiperOrigin-RevId:
184878663
Yao Zhang [Wed, 7 Feb 2018 20:29:25 +0000 (12:29 -0800)]
Set the number of warmup steps for building the cost model.
PiperOrigin-RevId:
184878186
Mingsheng Hong [Wed, 7 Feb 2018 20:22:55 +0000 (12:22 -0800)]
Initial XLA support for TF eager. This is prerequisite for TF compiler's XLA support.
This CL adds XLA support for the following TFE_Op's:
1. A TF op such as MatMul, with full support of constant and resource params.
2. A TF_Function as TFE_Op, where the function must have no constant and
resource params. Removing this restriction requires more discussion and will be
deferred to a later time.
PiperOrigin-RevId:
184877345
Yu-Cheng Ling [Wed, 7 Feb 2018 20:05:01 +0000 (12:05 -0800)]
Update TFLite iOS Camera Example app to use TFLite CocoaPod.
PiperOrigin-RevId:
184874871
A. Unique TensorFlower [Wed, 7 Feb 2018 20:01:40 +0000 (12:01 -0800)]
Update to type-dependent while loop, and tests for it.
PiperOrigin-RevId:
184874151
Allen Lavoie [Wed, 7 Feb 2018 19:48:15 +0000 (11:48 -0800)]
Prototype for object-based save/restore
Includes deferred restoration (mostly useful for eager execution).
Slot variables are created with their checkpointed values as soon as the variable they're slotting for is restored, so there's no need to override slot creation (optimizers already check for existing slot variables before creating a new one).
Changes the behavior of unnamed Checkpointable dependencies so that only other unnamed dependencies can interfere (named dependencies do not get a number). This should be a bit more robust, and will support property assignment syntax sugar in a future CL. It does mean that removing a name from a dependency will break the checkpoint (just like changing its name would).
One minor fix for slot creation eager compatibility.
PiperOrigin-RevId:
184871747
A. Unique TensorFlower [Wed, 7 Feb 2018 19:17:01 +0000 (11:17 -0800)]
make calling NNAPI work again
(this a copy of #16256 which apparently got lost in the one of the merges)
PiperOrigin-RevId:
184866202
A. Unique TensorFlower [Wed, 7 Feb 2018 18:30:39 +0000 (10:30 -0800)]
Improve model_pruner:
* Actually remove nodes marked for removal if fetches are known.
* Remove trivial nodes even in the presence of control inputs, except
for Identity nodes when
a) they are anchored on an Identity following a Switch node and
removal would require anchoring a control identity on the Switch, or
b) they have control inputs and feed a Merge node.
* Remove nodes only when in_degree * out_degree <= in_degree + out_degree.
Move input deduping utility function to utils.{h,cc}.
PiperOrigin-RevId:
184858685
Sanjoy Das [Wed, 7 Feb 2018 18:18:53 +0000 (10:18 -0800)]
[XLA:CPU] Fix/suppress issues caught by the C++ linter
PiperOrigin-RevId:
184856538
A. Unique TensorFlower [Wed, 7 Feb 2018 17:08:51 +0000 (09:08 -0800)]
Make TypeError more explicit on _assertAllCloseRecursive
PiperOrigin-RevId:
184846656
A. Unique TensorFlower [Wed, 7 Feb 2018 17:01:44 +0000 (09:01 -0800)]
Adding support for standalone Tanh operator.
PiperOrigin-RevId:
184845130
A. Unique TensorFlower [Wed, 7 Feb 2018 16:48:05 +0000 (08:48 -0800)]
Refactor score definition in GMM operations. This is simplified to be the per-sample likelihood of the data.
PiperOrigin-RevId:
184843634
A. Unique TensorFlower [Wed, 7 Feb 2018 15:16:17 +0000 (07:16 -0800)]
Utilities for type checking and multiple dispatch.
PiperOrigin-RevId:
184834434
A. Unique TensorFlower [Wed, 7 Feb 2018 14:30:32 +0000 (06:30 -0800)]
Support for quantized LstmCell, with initial reference runtime code.
The current 'optimized' code is just a copy of the reference code,
a true optimized implementation will follow separately.
PiperOrigin-RevId:
184830223
Sanjoy Das [Wed, 7 Feb 2018 07:55:28 +0000 (23:55 -0800)]
[XLA:CPU] Assert more thoroughly on preconditions in VectorSupportlibrary
No behavior change intended.
PiperOrigin-RevId:
184797003
Sanjoy Das [Wed, 7 Feb 2018 07:54:26 +0000 (23:54 -0800)]
[XLA] Use HloVerifiedTestBase in AlgebraicSimplifierTest
And fix the fallout. Thanks to asbirlea@ for noticing this!
PiperOrigin-RevId:
184796949
Sanjoy Das [Wed, 7 Feb 2018 06:48:48 +0000 (22:48 -0800)]
[XLA:CPU] Add an LLVM IR implementation of Exp
This lets us avoid the usual set of issues that crop up when XLA generated code
has to call into C++.
PiperOrigin-RevId:
184793093
Yifei Feng [Wed, 7 Feb 2018 06:05:26 +0000 (22:05 -0800)]
Add pylint check for W0611 unused-import in ci_sanity.sh and fix existing pylint errors.
PiperOrigin-RevId:
184790548
Zhixian Yan [Wed, 7 Feb 2018 05:32:48 +0000 (21:32 -0800)]
LSTM for TFlite/Toco
PiperOrigin-RevId:
184788311
Russell Power [Wed, 7 Feb 2018 05:19:50 +0000 (21:19 -0800)]
Clear feed error on session start.
An existing hook can be re-used on a different session: rather than terminating
the session we should clear the error state.
PiperOrigin-RevId:
184787368
Jonathan Hseu [Wed, 7 Feb 2018 04:40:00 +0000 (20:40 -0800)]
TPUEstimator: Revert the global_step change and require the user to explicitly
pass it.
PiperOrigin-RevId:
184784330
Sanjoy Das [Wed, 7 Feb 2018 02:28:00 +0000 (18:28 -0800)]
[XLA:CPU] Use VectorSupportLibrary for LLVM IR implementation of tanh
No behavioral change intended; this is only refactoring.
VectorSupportLibrary was added after the LLVM IR implementation of tanh so the
tanh implementation was not using VectorSupportLibrary.
The main impetus for this change is that I'm about to add LLVM IR
implementations of Exp and Log, and those are going to use VectorSupportLibrary.
I did not want to have an inconsistency between the tanh and exp, log.
PiperOrigin-RevId:
184774860
A. Unique TensorFlower [Wed, 7 Feb 2018 02:26:02 +0000 (18:26 -0800)]
We used to bypass fake-quant nodes in resolve_reorder_axes, as a hack as we
needed to preserve fake-quant nodes on constant weights as the only way to
encode min-max information when exporting to GraphDef.
Not anymore. Now we unconditionally enable the resolve_constant_fake_quant
transformation, and we don't do this bypass anymore; instead, when exporting
to GraphDef, we re-add FakeQuant nodes just before exporting, around
constant arrays that have minmax.
PiperOrigin-RevId:
184774680
Derek Murray [Wed, 7 Feb 2018 01:37:02 +0000 (17:37 -0800)]
[tf.data] Fix a memory leak when an iterator is reinitialized many times in a session.
Previously, we would instantiate a new function handle for each
function in a dataset each time an iterator on that dataset was
initialized. These would only be deleted at session closure, which
could lead to an apparent leak of memory over the lifetime of session.
PiperOrigin-RevId:
184768730
Yuefeng Zhou [Wed, 7 Feb 2018 01:32:50 +0000 (17:32 -0800)]
Fix memory tracking in the case where temp memory is used as output memory.
Track all persistent memory allocation in allocate_persistent call except for variables and queues where persistent memory is tracked in variables ops and queue ops. Deallocation of persistent memory is ignored.
PiperOrigin-RevId:
184768231
Andrew Selle [Wed, 7 Feb 2018 01:10:31 +0000 (17:10 -0800)]
Sync the opensource and non-opensource build
PiperOrigin-RevId:
184765632
James Qin [Wed, 7 Feb 2018 00:25:59 +0000 (16:25 -0800)]
Address Adagrad/RMSProp incompatibility with CudnnRNN
CudnnRNN layers have variables of unknown shapes, which Adagrad/RMSProp didn't handle before.
This fixes 6620(#6620).
PiperOrigin-RevId:
184759579
Justin Lebar [Tue, 6 Feb 2018 23:30:40 +0000 (15:30 -0800)]
[XLA] Add and use new Literal::MakeTupleOwned overload.
Previously MakeTupleOwned was cumbersome to use, because you had to
explicitly materialize a vector<unique_ptr<Literal>>. With this new
overload, you can pass unique_ptrs directly.
PiperOrigin-RevId:
184751119
Akshay Modi [Tue, 6 Feb 2018 23:13:41 +0000 (15:13 -0800)]
Add read resource variable benchmarks.
Initial benchmarks:
entry {
name: "MicroBenchmarks.benchmark_read_variable_op_2_by_2_CPU"
iters: 30000
wall_time: 22.
9616721471
extras {
key: "examples_per_sec"
value {
double_value: 43550.8352176
}
}
}
entry {
name: "MicroBenchmarks.benchmark_read_variable_op_with_tape_2_by_2_CPU"
iters: 30000
wall_time: 27.
3616631826
extras {
key: "examples_per_sec"
value {
double_value: 36547.4859232
}
}
}
PiperOrigin-RevId:
184748548
A. Unique TensorFlower [Tue, 6 Feb 2018 23:09:51 +0000 (15:09 -0800)]
Creates tf.contrib.feature_column module.
PiperOrigin-RevId:
184747924
Alexandre Passos [Tue, 6 Feb 2018 22:43:18 +0000 (14:43 -0800)]
Handles possible infinite recursion in while loop fix.
PiperOrigin-RevId:
184743192
Eugene Brevdo [Tue, 6 Feb 2018 22:37:21 +0000 (14:37 -0800)]
[TF Ops] Bugfix to Operation initializer: error message uses node_def.
self.node_def may not yet be accessible when using the C api.
PiperOrigin-RevId:
184742074
Joshua V. Dillon [Tue, 6 Feb 2018 22:17:52 +0000 (14:17 -0800)]
Automated g4 rollback of changelist
184551259
PiperOrigin-RevId:
184738583
Anna R [Tue, 6 Feb 2018 22:05:01 +0000 (14:05 -0800)]
Export CXX11_ABI_FLAG and MONOLITHIC_BUILD constants.
PiperOrigin-RevId:
184736216
Alexandre Passos [Tue, 6 Feb 2018 21:54:56 +0000 (13:54 -0800)]
Another rolling back of performance regression.
PiperOrigin-RevId:
184734426
Akshay Modi [Tue, 6 Feb 2018 21:53:56 +0000 (13:53 -0800)]
Make TFE_Py_FastpathExecute work for all types of ops
MatMul benchmarks:
entry {
name: "MicroBenchmarks.benchmark_gen_math_ops_matmul_2_by_2_CPU"
iters: 30000
wall_time: 11.
580435435
extras {
key: "examples_per_sec"
value {
double_value: 86352.538781
}
}
}
entry {
name: "MicroBenchmarks.benchmark_tfe_py_fastpath_execute_matmul_2_by_2_CPU"
iters: 30000
wall_time: 7.
02576637268
extras {
key: "examples_per_sec"
value {
double_value: 142333.227004
}
}
}
PiperOrigin-RevId:
184734289
Benoit Steiner [Tue, 6 Feb 2018 21:48:55 +0000 (13:48 -0800)]
Added support for nested functions
Properly handle the case of control dependencies
PiperOrigin-RevId:
184733444
Joshua V. Dillon [Tue, 6 Feb 2018 21:03:03 +0000 (13:03 -0800)]
Add utility function which makes implicit `tf.get_variable` dependencies an
explicit argument of a callable.
PiperOrigin-RevId:
184725878
Amit Patankar [Tue, 6 Feb 2018 20:58:56 +0000 (12:58 -0800)]
Fixing the issue where an escape character was being included in the branch name changes.
PiperOrigin-RevId:
184725332
Sanjoy Das [Tue, 6 Feb 2018 20:57:07 +0000 (12:57 -0800)]
[TF:XLA] Bump open source llvm revision to r324323
PiperOrigin-RevId:
184725126
Skye Wanderman-Milne [Tue, 6 Feb 2018 20:44:12 +0000 (12:44 -0800)]
Fetch OpDefs from C API instead of using python op registry in ops.py.
PiperOrigin-RevId:
184723558
A. Unique TensorFlower [Tue, 6 Feb 2018 20:30:05 +0000 (12:30 -0800)]
Fixed typo
PiperOrigin-RevId:
184721743
Shanqing Cai [Tue, 6 Feb 2018 20:27:02 +0000 (12:27 -0800)]
tfdbg: deflake session_debug_file_test
by disabling grappler in the test.
PiperOrigin-RevId:
184721353
Alexandre Passos [Tue, 6 Feb 2018 20:04:48 +0000 (12:04 -0800)]
Second, cleaner, attempt at external control dependency handling.
PiperOrigin-RevId:
184718016
A. Unique TensorFlower [Tue, 6 Feb 2018 20:02:18 +0000 (12:02 -0800)]
Improve side_effect_guards to (1) avoid aliasing non-tensors and (2) combine aliasing with re-indenting. Move the renaming visitor into a generic utility API.
This loses potential efficiency by risking sequencing ops where there is no risk of a race. On the other hand it's still not entirely robust, and we need to raise an error where we can't guarantee that.
PiperOrigin-RevId:
184717456
Alexandre Passos [Tue, 6 Feb 2018 20:01:22 +0000 (12:01 -0800)]
Estimator should handle ResourceVariables correctly.
PiperOrigin-RevId:
184717305
Alexandre Passos [Tue, 6 Feb 2018 19:58:36 +0000 (11:58 -0800)]
Support resource variables in moving averages.
PiperOrigin-RevId:
184716895
Alexandre Passos [Tue, 6 Feb 2018 19:51:47 +0000 (11:51 -0800)]
Contrib estimator should handle ResourceVariables properly.
PiperOrigin-RevId:
184715896
Mingsheng Hong [Tue, 6 Feb 2018 19:51:18 +0000 (11:51 -0800)]
Internal change
PiperOrigin-RevId:
184715822
Anna R [Tue, 6 Feb 2018 19:07:04 +0000 (11:07 -0800)]
Adding tf_export decorators/calls to TensorFlow functions and constants.
PiperOrigin-RevId:
184708277
Yao Zhang [Tue, 6 Feb 2018 19:00:58 +0000 (11:00 -0800)]
Enable layout optimizer by default
PiperOrigin-RevId:
184707084
A. Unique TensorFlower [Tue, 6 Feb 2018 18:48:28 +0000 (10:48 -0800)]
Support training with model parallelism in tpu_estimator.
PiperOrigin-RevId:
184704902
Benoit Steiner [Tue, 6 Feb 2018 18:43:45 +0000 (10:43 -0800)]
Use the function name as the grappler item id
PiperOrigin-RevId:
184704131
A. Unique TensorFlower [Tue, 6 Feb 2018 18:13:04 +0000 (10:13 -0800)]
Add local interconnect data to DeviceLocality.
This information can be used within a distributed implementation
when deciding how to route data transfers that might involve more
than one hop. By default the new fields are populated according to
StreamExecutor::CanEnablePeerAccessTo(), however a platform-specific
implementation can augment them with more detailed values.
Do some refactoring of gpu_device and gpu_device_factory, making
GetDeviceLocalities() and GetInterconnectMaps() into virtual
functions.
PiperOrigin-RevId:
184698821
Benoit Steiner [Tue, 6 Feb 2018 18:11:26 +0000 (10:11 -0800)]
Explicitly specify the value of all the attributes for the newly created Assign
nodes since we can't always rely on TF calling AddDefaultAttrsToGraphDef.
PiperOrigin-RevId:
184698463
A. Unique TensorFlower [Tue, 6 Feb 2018 17:40:53 +0000 (09:40 -0800)]
Fix bug in and speed up ConstantFolding::CreateNodeDef():
* Fix bug trying to store more than kintmax32 values in a repeated proto field.
* Speed up populating compressed format. Example: tensorflow/python/kernel_tests/large_concat_op_test with size = 2**29+6 goes from ~30 seconds to ~15 seconds. The fraction of time spent in ConstantFolding::CreateNodeDef() goes down from about 35% to about 12%.
PiperOrigin-RevId:
184693749
Andrew Selle [Tue, 6 Feb 2018 17:28:24 +0000 (09:28 -0800)]
Add a utility function to partition TensorFlow Lite graphs into subgraphs.
This will be used for the forthcoming delegate interface. The delegate
interface will allow parts of the TensorFlow lite graph to be sent to
other accelerators. NNAPI will be implemented to move to this interface.
That will allow partial delegation to NNAPI.
PiperOrigin-RevId:
184692126
Asim Shankar [Tue, 6 Feb 2018 16:53:58 +0000 (08:53 -0800)]
Java: Update to 1.6.0-rc0
PiperOrigin-RevId:
184687994
A. Unique TensorFlower [Tue, 6 Feb 2018 16:50:41 +0000 (08:50 -0800)]
Make namedtuples with identical name and field names to be considered as the same shallow structure in assert_shallow_structure
PiperOrigin-RevId:
184687609
A. Unique TensorFlower [Tue, 6 Feb 2018 16:11:41 +0000 (08:11 -0800)]
Remove redundant calls to realloc dynamic tensors.
PiperOrigin-RevId:
184682942