Alexandre Passos [Fri, 2 Feb 2018 23:52:18 +0000 (15:52 -0800)]
TF_CALL_ALL_TYPES should include variant
PiperOrigin-RevId:
184347081
Alexandre Passos [Fri, 2 Feb 2018 23:51:41 +0000 (15:51 -0800)]
Rollback of recent changes to tensor.{h,cc}
PiperOrigin-RevId:
184347012
Jianwei Xie [Fri, 2 Feb 2018 23:00:17 +0000 (15:00 -0800)]
Adds batch inference support on TPU with TPUEstimator.predict.
PiperOrigin-RevId:
184339842
Eugene Brevdo [Fri, 2 Feb 2018 22:30:50 +0000 (14:30 -0800)]
[TF RNN] Small optimization to rnn: only calculate copy_cond once.
PiperOrigin-RevId:
184335231
Tatiana Shpeisman [Fri, 2 Feb 2018 22:14:13 +0000 (14:14 -0800)]
Changed minimal required version of bleach Python package from 1.5 to 2.0. bleach 1.5 causes problems with Jupyter as reported in #16424
PiperOrigin-RevId:
184332663
Mingsheng Hong [Fri, 2 Feb 2018 21:53:08 +0000 (13:53 -0800)]
A few misc tweaks to TFE APIs.
PiperOrigin-RevId:
184329345
Yu-Cheng Ling [Fri, 2 Feb 2018 21:49:35 +0000 (13:49 -0800)]
TFLite: Conv CBLAS kernel
PiperOrigin-RevId:
184328848
Derek Murray [Fri, 2 Feb 2018 21:39:38 +0000 (13:39 -0800)]
[tf.data] Move framework/dataset.h to framework/dataset_stateful_op_whitelist.h.
This will make way for the move of kernels/data/dataset.h to
framework/dataset.h, while preserving version control history, after
which we might recombine the headers.
PiperOrigin-RevId:
184327481
Joshua V. Dillon [Fri, 2 Feb 2018 21:08:31 +0000 (13:08 -0800)]
Support `float16` `dtype` in `tf.linalg.*`.
Note: not all `LinearOperator` functions will support `float16`. This change
merely enables constructing the `LinearOperator` object(s) using this `dtype`.
PiperOrigin-RevId:
184323477
Joshua V. Dillon [Fri, 2 Feb 2018 21:07:37 +0000 (13:07 -0800)]
Breaking change: Revise HMC interface to accept a list of Tensors representing a partitioning of chain state.
PiperOrigin-RevId:
184323369
Yuefeng Zhou [Fri, 2 Feb 2018 20:46:34 +0000 (12:46 -0800)]
Skip the node that has unexpected number of outputs.
PiperOrigin-RevId:
184320865
Sanjoy Das [Fri, 2 Feb 2018 20:23:47 +0000 (12:23 -0800)]
[TF:XLA] Bump open source llvm revision to r324073
PiperOrigin-RevId:
184318036
A. Unique TensorFlower [Fri, 2 Feb 2018 20:19:43 +0000 (12:19 -0800)]
Update ops-related pbtxt files.
PiperOrigin-RevId:
184317521
Peter Hawkins [Fri, 2 Feb 2018 19:31:06 +0000 (11:31 -0800)]
[TF:XLA] Improve/refactor the handling of resource types/shapes.
Previously we used an xla::Shape to track the shape of a resource (Variable, TensorArray, Stack) shape. The xla::Shape described how the resource was represented to XLA, e.g., as a (buffer, size) pair for a Stack resource.
Instead, separate the TensorFlow abstract shape representation from the XLA shape representation and track it separately. This leads to simpler and more readable code.
PiperOrigin-RevId:
184310694
A. Unique TensorFlower [Fri, 2 Feb 2018 19:24:17 +0000 (11:24 -0800)]
Allow ResizeBilinear to resize the output tensor in Prepare(), if the size tensor is const.
PiperOrigin-RevId:
184309687
Alexandre Passos [Fri, 2 Feb 2018 19:24:11 +0000 (11:24 -0800)]
Register resource_scatter_update for string types.
PiperOrigin-RevId:
184309674
A. Unique TensorFlower [Fri, 2 Feb 2018 19:06:09 +0000 (11:06 -0800)]
Add shape inference for outside_compilation graph rewrite. Pull out enough of the graph to enable inference of the shape of a SendFromHost Op once the shape of corresponding RecvAtHost Ops are known.
END_PUBLIC
Fixed open source build breaks.
BEGIN_PUBLIC
Automated g4 rollback of changelist
184169668
PiperOrigin-RevId:
184306845
A. Unique TensorFlower [Fri, 2 Feb 2018 18:58:12 +0000 (10:58 -0800)]
Register GPU host kernels for Identity and RefIdentity.
PiperOrigin-RevId:
184305574
A. Unique TensorFlower [Fri, 2 Feb 2018 18:46:34 +0000 (10:46 -0800)]
Automated g4 rollback of changelist
184273245
PiperOrigin-RevId:
184303789
Anna R [Fri, 2 Feb 2018 18:28:20 +0000 (10:28 -0800)]
Remove hidden_ops.txt file. Instead, switch to use visibility attribute in ApiDef proto.
PiperOrigin-RevId:
184301076
Jacques Pienaar [Fri, 2 Feb 2018 18:04:15 +0000 (10:04 -0800)]
Consider beyond immediate neighbors to find exit node.
Most of the exit nodes are immediate neighbors of the switch, except we do have
cases where the switch feeds into an identity that feeds into a exit.
PiperOrigin-RevId:
184297180
A. Unique TensorFlower [Fri, 2 Feb 2018 18:01:40 +0000 (10:01 -0800)]
Show that we must over reallocate after resizing dynamic tensors.
PiperOrigin-RevId:
184296680
A. Unique TensorFlower [Fri, 2 Feb 2018 17:20:26 +0000 (09:20 -0800)]
Fix latent bug in dependency optimizer.
PiperOrigin-RevId:
184291701
Yao Zhang [Fri, 2 Feb 2018 17:00:41 +0000 (09:00 -0800)]
Disable graph optimizations (CSE) in test so that constant nodes are not deduped.
PiperOrigin-RevId:
184289685
A. Unique TensorFlower [Fri, 2 Feb 2018 16:09:48 +0000 (08:09 -0800)]
[comment-only change]: Fix grammar error.
PiperOrigin-RevId:
184285125
Derek Murray [Fri, 2 Feb 2018 16:09:13 +0000 (08:09 -0800)]
Fix a bug in function inlining when the argument is an implicitly dereferenced ref tensor.
Previously the inliner would add an identity node with an invalid ref-type attr when the actual parameter had ref type. The changed version removes the reference.
PiperOrigin-RevId:
184285084
A. Unique TensorFlower [Fri, 2 Feb 2018 13:38:50 +0000 (05:38 -0800)]
Enabling partitioned variables to work with TPU.
When partitioned variables are used in a TPU training loop,
concat gradient operations get generated for which XLA requires
the concat dimension argument to be a constant (or foldable to a constant).
However since such constant is defined outside of the train while context
an Enter node is generated in order to pass it.
The fix consists in detecting such case, and to duplicate the (scalar) constant
inside the while context, so that XLA can succesfully process the resulting
graph.
PiperOrigin-RevId:
184273245
Gunhan Gulsoy [Fri, 2 Feb 2018 07:47:33 +0000 (23:47 -0800)]
Fix some tf-lite tests
PiperOrigin-RevId:
184247187
A. Unique TensorFlower [Fri, 2 Feb 2018 05:43:12 +0000 (21:43 -0800)]
Fix tolerance too tight for Wasserstein distance test.
PiperOrigin-RevId:
184240222
Gunhan Gulsoy [Fri, 2 Feb 2018 07:47:33 +0000 (23:47 -0800)]
Fix some tf-lite tests
PiperOrigin-RevId:
184247187
A. Unique TensorFlower [Fri, 2 Feb 2018 05:43:12 +0000 (21:43 -0800)]
Fix tolerance too tight for Wasserstein distance test.
PiperOrigin-RevId:
184240222
Justin Lebar [Fri, 2 Feb 2018 05:34:18 +0000 (21:34 -0800)]
Internal change
PiperOrigin-RevId:
184239740
A. Unique TensorFlower [Fri, 2 Feb 2018 04:35:29 +0000 (20:35 -0800)]
Automated g4 rollback of changelist
183874527
PiperOrigin-RevId:
184236409
A. Unique TensorFlower [Fri, 2 Feb 2018 03:50:34 +0000 (19:50 -0800)]
Supporting new saving op structure.
PiperOrigin-RevId:
184233513
Yu-Cheng Ling [Fri, 2 Feb 2018 02:27:31 +0000 (18:27 -0800)]
Add darwin_x86_64 config in TF Lite BUILD file.
PiperOrigin-RevId:
184227786
Benoit Steiner [Fri, 2 Feb 2018 02:05:15 +0000 (18:05 -0800)]
Fixed the description of the fake GPU device to avoid a division by 0
PiperOrigin-RevId:
184225409
Brennan Saeta [Fri, 2 Feb 2018 01:20:11 +0000 (17:20 -0800)]
Fix tests
PiperOrigin-RevId:
184220615
Benoit Steiner [Fri, 2 Feb 2018 01:19:32 +0000 (17:19 -0800)]
Skip unknown devices since we can't optimize for them
PiperOrigin-RevId:
184220515
Andrew Selle [Fri, 2 Feb 2018 00:51:59 +0000 (16:51 -0800)]
Allow reordering of execution order of nodes with indirect execution_plan.
Now whenever we want to operate in dependency order we use execution_plan.
It begins as identity map (0, ..., nodes_size()) but can be changed
in the future. This is the basis for more pluggable delegation.
PiperOrigin-RevId:
184216885
Jacques Pienaar [Fri, 2 Feb 2018 00:29:30 +0000 (16:29 -0800)]
Automated g4 rollback of changelist
184188816
PiperOrigin-RevId:
184213576
Brennan Saeta [Fri, 2 Feb 2018 00:22:24 +0000 (16:22 -0800)]
GCS Throttle: 1 token == 1 Kb
Previously, 1 token was approximately 256 bytes. This is slightly less
intuitive than 1 kb.
PiperOrigin-RevId:
184212503
Raghuraman Krishnamoorthi [Fri, 2 Feb 2018 00:14:51 +0000 (16:14 -0800)]
Add functionality to fold batch norm (supporting both fused and unfused batch norm) to support quantized training. The weights are always now scaled by gamma/sigma, where sigma is the moving standard deviation for stability prior to quantization. For improved performance, the moving means and variances are frozen and the training graph modified accordingly.
An additional parameter freeze_batch_norm_delay is added to foldbatchnorm function to set the delay at which training switches from regular batch norm to frozen mean and variances.
Remove placement options within FoldBatchNorm as this causes folded training to place all ops on a single GPU. The modification now significantly speeds up distributed training.
The tests for folding batch norms are also updated to reflect the additional topological changes to the graph.
PiperOrigin-RevId:
184211434
A. Unique TensorFlower [Thu, 1 Feb 2018 23:32:20 +0000 (15:32 -0800)]
Add iterate_batches arg to Estimator.predict
PiperOrigin-RevId:
184205196
Skye Wanderman-Milne [Thu, 1 Feb 2018 23:11:35 +0000 (15:11 -0800)]
Make jit_test.py work with C API enabled.
PiperOrigin-RevId:
184202470
A. Unique TensorFlower [Thu, 1 Feb 2018 23:11:20 +0000 (15:11 -0800)]
[XLA] add DotGeneral to the local Python XLA client.
PiperOrigin-RevId:
184202425
Nick Felt [Thu, 1 Feb 2018 23:05:41 +0000 (15:05 -0800)]
Revert TensorBoard entry point back to run_main
PiperOrigin-RevId:
184201506
Anna R [Thu, 1 Feb 2018 22:23:29 +0000 (14:23 -0800)]
Internal change.
PiperOrigin-RevId:
184194895
Igor Saprykin [Thu, 1 Feb 2018 22:11:08 +0000 (14:11 -0800)]
Throw an exception when the user's batch size isn't divisible by GPUs.
The alternative to this is to have an adaptive approach that would unevenly split input into per-tower batches. The concern with that was that all towers will be as slow as the one with more input reducing the performance. Batch size seems to be commonly tailored to the available hardware.
PiperOrigin-RevId:
184192793
Benoit Steiner [Thu, 1 Feb 2018 22:11:08 +0000 (14:11 -0800)]
Return an error instead of assertion when processing an ill-formed graph or an
invalid set of fetch nodes
PiperOrigin-RevId:
184192790
Jacques Pienaar [Thu, 1 Feb 2018 21:48:33 +0000 (13:48 -0800)]
[TFXLA] Use data flow to determine switch grouping.
* Change how switch grouping works:
- This is an intermediate step, next is combining
DetermineBranchMapAndFrontier into one traversal.
* Homogeneous the naming (switch_nodes -> switches);
* Change graph dumping to be due to class member - currently still performed when vlog-level is sufficiently high;
* Pass in correct library when dumping graphs;
PiperOrigin-RevId:
184188816
Noah Fiedel [Thu, 1 Feb 2018 21:48:13 +0000 (13:48 -0800)]
Adding documentation on how to load & serve a model with the
TensorFlow Serving Model Server.
PiperOrigin-RevId:
184188752
A. Unique TensorFlower [Thu, 1 Feb 2018 21:33:20 +0000 (13:33 -0800)]
Fixes a type conversion bug in losses.compute_weighted_loss for reduction=SUM_OVER_BATCH_SIZE.
PiperOrigin-RevId:
184186573
A. Unique TensorFlower [Thu, 1 Feb 2018 21:13:12 +0000 (13:13 -0800)]
Fix segfault when Softmax is first in graph
PiperOrigin-RevId:
184183730
A. Unique TensorFlower [Thu, 1 Feb 2018 21:13:10 +0000 (13:13 -0800)]
Verify tensor contents of tflite model
PiperOrigin-RevId:
184183725
Benoit Steiner [Thu, 1 Feb 2018 20:38:55 +0000 (12:38 -0800)]
Made the addn optimization aware of the graph topology
PiperOrigin-RevId:
184179246
A. Unique TensorFlower [Thu, 1 Feb 2018 20:17:08 +0000 (12:17 -0800)]
Add a utility module that contains helper functions usable from within generated code.
Add a helper for the control dependencies context manager.
PiperOrigin-RevId:
184176409
Anna R [Thu, 1 Feb 2018 20:05:23 +0000 (12:05 -0800)]
Internal change.
PiperOrigin-RevId:
184174800
A. Unique TensorFlower [Thu, 1 Feb 2018 19:54:13 +0000 (11:54 -0800)]
Update deprecated API use
PiperOrigin-RevId:
184173047
Derek Murray [Thu, 1 Feb 2018 19:50:23 +0000 (11:50 -0800)]
[tf.data] Fix bug where captured resources in shared iterators were invisible.
This change ensures that a shared iterator (which requires a private
FunctionLibraryRuntime that outlasts the calling op's runtime, because
it can outlive a single session) uses the same Device as a non-shared
iterator, and hence capturing resources from the creating graph will
work as intended.
Fixes #16481.
PiperOrigin-RevId:
184172498
Benoit Steiner [Thu, 1 Feb 2018 19:50:14 +0000 (11:50 -0800)]
Added a utility to traverse the graph in reverse DFS order, identifying loops
in the process.
PiperOrigin-RevId:
184172483
Anna R [Thu, 1 Feb 2018 19:33:34 +0000 (11:33 -0800)]
Automated g4 rollback of changelist
184153187
PiperOrigin-RevId:
184169668
Anna R [Thu, 1 Feb 2018 19:06:53 +0000 (11:06 -0800)]
Internal change.
PiperOrigin-RevId:
184165180
Mark Daoust [Thu, 1 Feb 2018 18:43:29 +0000 (10:43 -0800)]
Add function paths to their signatures.
fixes #16167
PiperOrigin-RevId:
184160925
A. Unique TensorFlower [Thu, 1 Feb 2018 18:38:14 +0000 (10:38 -0800)]
Fix nest bug with different dictionary key orderings.
PiperOrigin-RevId:
184160009
A. Unique TensorFlower [Thu, 1 Feb 2018 17:55:53 +0000 (09:55 -0800)]
Add shape inference for outside_compilation graph rewrite. Pull out enough of the graph to enable inference of the shape of a SendFromHost Op once the shape of corresponding RecvAtHost Ops are known.
PiperOrigin-RevId:
184153187
A. Unique TensorFlower [Thu, 1 Feb 2018 16:18:13 +0000 (08:18 -0800)]
Update ops-related pbtxt files.
PiperOrigin-RevId:
184141875
Peter Hawkins [Thu, 1 Feb 2018 14:47:06 +0000 (06:47 -0800)]
[TF:XLA] Implement MatrixSetDiag and MatrixBandPart.
Add support for int32 indices to the MatrixBandPart operator.
PiperOrigin-RevId:
184133343
Sanjoy Das [Thu, 1 Feb 2018 07:05:26 +0000 (23:05 -0800)]
[TF:XLA] Fix tfcompile OSS build
- The @org_tensorflow package designation is unnecessary, and breaks the build
when building without a sandbox.
- The generated tests must use tf_cc_test, not cc_test. See the note in
tensorflow/core/BUILD.
Partially addresses #15338
PiperOrigin-RevId:
184095571
Yu-Cheng Ling [Thu, 1 Feb 2018 05:17:30 +0000 (21:17 -0800)]
Internal change
PiperOrigin-RevId:
184088913
A. Unique TensorFlower [Thu, 1 Feb 2018 04:46:33 +0000 (20:46 -0800)]
Go: Update generated wrapper functions for TensorFlow ops.
PiperOrigin-RevId:
184086955
A. Unique TensorFlower [Thu, 1 Feb 2018 04:20:18 +0000 (20:20 -0800)]
Update ops-related pbtxt files.
PiperOrigin-RevId:
184085402
Eugene Brevdo [Thu, 1 Feb 2018 02:18:04 +0000 (18:18 -0800)]
Add a new Dataset: PrependFromQueueAndPaddedBatchDataset.
PiperOrigin-RevId:
184078894
Yu-Cheng Ling [Thu, 1 Feb 2018 02:08:33 +0000 (18:08 -0800)]
Test all TFLite kernel implementations (reference/optimized/...)
PiperOrigin-RevId:
184077940
Akshay Agrawal [Thu, 1 Feb 2018 02:03:25 +0000 (18:03 -0800)]
TFE: Support `IndexedSlices` as inputs and outputs for `tfe.defun`.
In particular, this change fixes a bug that arose when attempting to
differentiate a defun-d function whose gradient yielded `IndexedSlices`.
PiperOrigin-RevId:
184077261
Mark Daoust [Thu, 1 Feb 2018 01:46:43 +0000 (17:46 -0800)]
Add a doc explaining how to convert an `Estimator` to run on a Cloud TPU.
PiperOrigin-RevId:
184075365
Roy Frostig [Thu, 1 Feb 2018 00:54:20 +0000 (16:54 -0800)]
[XLA] Support array layout specification in local Python XLA client.
PiperOrigin-RevId:
184069454
Derek Murray [Thu, 1 Feb 2018 00:40:10 +0000 (16:40 -0800)]
Add type check when assigning to a resource variable slice.
PiperOrigin-RevId:
184067663
Asim Shankar [Thu, 1 Feb 2018 00:17:42 +0000 (16:17 -0800)]
eager: Fix dropout in MNIST example.
The dropout layer takes a training argument in __call__, which defaults
to false. So without this change, dropout was not being applied.
PiperOrigin-RevId:
184064465
Tatiana Shpeisman [Wed, 31 Jan 2018 23:47:42 +0000 (15:47 -0800)]
MKL is no longer enabled via ./configure. Fixed documentation to reflect this.
PiperOrigin-RevId:
184060046
Tatiana Shpeisman [Wed, 31 Jan 2018 23:41:03 +0000 (15:41 -0800)]
Change recommended option for building TensorFlow with MKL from -c opt to --config=opt. -c opt triggers optimized C++ compilation. --config=opt also uses additional optimization flags as set by running ./configure.
PiperOrigin-RevId:
184059060
A. Unique TensorFlower [Wed, 31 Jan 2018 23:32:39 +0000 (15:32 -0800)]
Update external protobuf codebase version for Windows cmake build
PiperOrigin-RevId:
184057827
Gunhan Gulsoy [Wed, 31 Jan 2018 22:57:24 +0000 (14:57 -0800)]
Enable AVX in all TF windows builds.
PiperOrigin-RevId:
184052414
Shanqing Cai [Wed, 31 Jan 2018 22:55:13 +0000 (14:55 -0800)]
Add grpcio as a pip dependency of tensorflow
PiperOrigin-RevId:
184052073
Shivani Agrawal [Wed, 31 Jan 2018 22:35:37 +0000 (14:35 -0800)]
Fixes minor typos.
PiperOrigin-RevId:
184048812
Nick Desaulniers [Wed, 31 Jan 2018 22:32:21 +0000 (14:32 -0800)]
[XLA] Initialize linear indices used by NearComparator.
In unoptimized builds, we'd see miscompares with insanely large indices when
comparing literals containing NaN.
The linear indices may never be updated/initialized if the value NaN was
compared, since:
(NaN > x) == false
(NaN < x) == false
Adds a few tests that were used when debugging, but not a perfect one for this
case. It's currently not possible to test that LiteralTestUtil::Near() fails
when given bad input since it uses the EXPECT_* family of macros, which would
cause the intentional miscompares to fail the test no matter what.
PiperOrigin-RevId:
184048275
Anna R [Wed, 31 Jan 2018 22:30:22 +0000 (14:30 -0800)]
Internal change.
PiperOrigin-RevId:
184047860
A. Unique TensorFlower [Wed, 31 Jan 2018 21:42:18 +0000 (13:42 -0800)]
Tolerate Const nodes with no data or with smaller data than is required
by their shape, by zero-extending the Const data to the required size.
We wanted to generate an error on that, but too many
existing graphs already rely on current lax behavior.
PiperOrigin-RevId:
184039876
A. Unique TensorFlower [Wed, 31 Jan 2018 21:35:23 +0000 (13:35 -0800)]
Automated g4 rollback of changelist
184003263
PiperOrigin-RevId:
184038801
Sanjoy Das [Wed, 31 Jan 2018 21:22:49 +0000 (13:22 -0800)]
[XLA:CPU] Generate correct IR for integer clamp
PiperOrigin-RevId:
184037078
Peter Hawkins [Wed, 31 Jan 2018 20:58:05 +0000 (12:58 -0800)]
[TF:XLA] Implement ExtractImagePatches.
PiperOrigin-RevId:
184033616
A. Unique TensorFlower [Wed, 31 Jan 2018 20:32:38 +0000 (12:32 -0800)]
Remove contacts of ex-Googler
PiperOrigin-RevId:
184030353
Peter Hawkins [Wed, 31 Jan 2018 20:28:13 +0000 (12:28 -0800)]
[TF:XLA] Implement Acos, Asin, Atan in terms of Atan2 using half-angle formulae. This may not be the most efficient implementation but it is better than no implementation.
PiperOrigin-RevId:
184029858
Sanjoy Das [Wed, 31 Jan 2018 20:27:38 +0000 (12:27 -0800)]
[TF:XLA] Bump open source llvm revision to r323874
PiperOrigin-RevId:
184029790
Alexandre Passos [Wed, 31 Jan 2018 20:22:10 +0000 (12:22 -0800)]
Increfing tapes when keeping them around calls to python API functions.
These calls can trigger GIL releases or python GC which can trigger
a change to the set of live tapes and segmentation faults.
PiperOrigin-RevId:
184029146
Yifei Feng [Wed, 31 Jan 2018 19:34:56 +0000 (11:34 -0800)]
De-bazel filename_test. Part of the effort to remove all_opensource_files.
PiperOrigin-RevId:
184021942
Todd Wang [Wed, 31 Jan 2018 19:32:35 +0000 (11:32 -0800)]
Go: Fix Scope.WithControlDependencies array-copying behavior.
The test fails with the old code, and passes with the new code.
PiperOrigin-RevId:
184021596
Jonathan Hseu [Wed, 31 Jan 2018 19:30:36 +0000 (11:30 -0800)]
TPUEstimator: support host_call when use_tpu=False.
PiperOrigin-RevId:
184021299
Anna R [Wed, 31 Jan 2018 19:25:38 +0000 (11:25 -0800)]
Adding tf_export decorators/calls to TensorFlow functions and constants.
PiperOrigin-RevId:
184020524
A. Unique TensorFlower [Wed, 31 Jan 2018 18:59:49 +0000 (10:59 -0800)]
Typo Correction.
PiperOrigin-RevId:
184016082
Eric Liu [Wed, 31 Jan 2018 18:38:57 +0000 (10:38 -0800)]
Add std:: to min/max in cuda code to unbreak cuda-clang build.
PiperOrigin-RevId:
184012479