Brennan Saeta [Sat, 28 Apr 2018 17:51:32 +0000 (10:51 -0700)]
[tf.data] Use core::ScopedUnref to avoid resource leakage.
If for whatever reason iterator_resource->set_iterator did not return Status::OK(), we would leak a reference on the iterator_resource. With this change, we won't leak the resource.
PiperOrigin-RevId:
194662412
A. Unique TensorFlower [Sat, 28 Apr 2018 17:40:49 +0000 (10:40 -0700)]
Allow not specifying eval_spec when evaluation is not necessarily run.
PiperOrigin-RevId:
194661814
Mingsheng Hong [Sat, 28 Apr 2018 15:55:08 +0000 (08:55 -0700)]
This is Part 1 of Swift<->TF sends/recvs: support sending tensors from TF to
Swift via direct session.
The changes are:
1. Added an experimental TF C API TF_DequeueNamedTensor() to consume the queued
tensors from a dequeue op. One use case is for the Swift host program to consume
tensors sent by TF, where the queue is a Fifo queue managed by TF.
Enqueuing tensors are done by running an enqueue op in a graph. The queued
tensors are not persisted, and will be lost if the process/machine dies. The
queue has a bounded capacity, to prevent producer from being unboundedly ahead
of consumer.
while caller of TF_DequeueNamedTensor() could have run the Fifo dequeue op
directly, the extra level of indirection provided by this API allows us to more
easily switch the queuing impl to another mechanism. If and once we stabilize on
the Fifo queue based impl, we can remove this API.
2. Added a new S4TF runtime API _TFCReceiveTensorHandle() that receives a tensor
via TF_DequeueNamedTensor().
3. To support tensor receives in host program, taught PartitionCloner in
TFPartition to insert SIL code to call _TFCReceiveTensorHandle().
4. To support tensor sends in accelerator program, taught TFGraphLowering in
generate QueueEnqueueV2 nodes in the TF graphs, with appropriate control
dependence to make sure these nodes get executed.
a) The enqueue produces no output tensor, and is executed only for its side
effect. To ensure it is executed properly, control dependence is wired up. The
general design is: before a TF_Function (can be a top level function or the body
function of a while op) produces an output tensor OT, make OT control dependent
on the enqueue op, so that enqueue gets run before the function returns.
b) If tensor send occurs in a while loop body, the body logic currently gets
lowered in 3 places: the while op cond function, the while op body function, and
the ops at the same level as the while op itself (for running the last loop
iteration). In this case, the correct TFGraph lowering is to run the enqueue in
the last 2 out of the 3 places above.
After this CL, the dual versions of the above (dequeuing via an op, and
enqueuing via C API) will be added.
PiperOrigin-RevId:
194658511
Anna R [Sat, 28 Apr 2018 07:13:09 +0000 (00:13 -0700)]
Removing hidden_ops.txt file.
PiperOrigin-RevId:
194637892
A. Unique TensorFlower [Sat, 28 Apr 2018 06:35:42 +0000 (23:35 -0700)]
Fix kernel creation bug, due to constant folding always use CPU.
PiperOrigin-RevId:
194636076
A. Unique TensorFlower [Sat, 28 Apr 2018 05:57:36 +0000 (22:57 -0700)]
Add test case on compiling dense layer node with XLA.
PiperOrigin-RevId:
194634563
Patrick Nguyen [Sat, 28 Apr 2018 04:58:17 +0000 (21:58 -0700)]
Properly export recurrent in contrib.
The following symbols are available:
- tf.contrib.recurrent.bidirectional_functional_rnn
- tf.contrib.recurrent.functional_rnn
- tf.contrib.recurrent.Recurrent
PiperOrigin-RevId:
194632138
Sanjoy Das [Sat, 28 Apr 2018 03:06:35 +0000 (20:06 -0700)]
HLO profiling for tfcompile.
This CL extends the --xla_hlo_profile knob to tfcompile. tf_library rules can
now set enable_xla_hlo_profiling to True to:
- Have the generated code update per-HLO profile counters as it executes.
- Have tfcompile generate and serialize an instance HloProfilePrinterData with
a compiled model that can be used to pretty-print the collected profile
counters.
PiperOrigin-RevId:
194627272
A. Unique TensorFlower [Sat, 28 Apr 2018 02:23:16 +0000 (19:23 -0700)]
Add internal uint b stats to TfOpStats.
PiperOrigin-RevId:
194625155
Sanjoy Das [Sat, 28 Apr 2018 01:41:27 +0000 (18:41 -0700)]
Split up ElementaIrEmitter::MakeElementGenerator into smaller functions; NFC
PiperOrigin-RevId:
194622198
A. Unique TensorFlower [Sat, 28 Apr 2018 01:24:57 +0000 (18:24 -0700)]
[XLA] Fix bug in ShapeUtil::StripDegenerateDimensions
PiperOrigin-RevId:
194621163
Brennan Saeta [Sat, 28 Apr 2018 00:08:57 +0000 (17:08 -0700)]
Make RetryingFileSystem a template.
PiperOrigin-RevId:
194614877
Petros Mol [Fri, 27 Apr 2018 23:22:43 +0000 (16:22 -0700)]
Minor fix to SDCAOptimizer documentation.
PiperOrigin-RevId:
194609850
Nick Desaulniers [Fri, 27 Apr 2018 23:14:49 +0000 (16:14 -0700)]
[TF:XLA:INTERPRETER] implement bfloat16 comparisons
PiperOrigin-RevId:
194608854
A. Unique TensorFlower [Fri, 27 Apr 2018 22:28:01 +0000 (15:28 -0700)]
Fix broken ElementWiseFusionTest.
PiperOrigin-RevId:
194602336
Akshay Modi [Fri, 27 Apr 2018 21:47:12 +0000 (14:47 -0700)]
Minor eager service proto clarification.
PiperOrigin-RevId:
194596337
Sanjoy Das [Fri, 27 Apr 2018 21:36:24 +0000 (14:36 -0700)]
[XLA:CPU] Implement fusion for the Gather HLO
PiperOrigin-RevId:
194594759
A. Unique TensorFlower [Fri, 27 Apr 2018 21:28:12 +0000 (14:28 -0700)]
If two identical functions are given different grad func,
they should be named differently. Otherwise, tf.gradients
gets confused.
PiperOrigin-RevId:
194593519
A. Unique TensorFlower [Fri, 27 Apr 2018 21:05:46 +0000 (14:05 -0700)]
Remove whitespaces from tags in saved_model_cli.
This currently causes tags mismatch because a leading whitespace is added
within the saved_model_cli when doing ', '.join(tag_set).
PiperOrigin-RevId:
194590154
Sanjoy Das [Fri, 27 Apr 2018 20:55:35 +0000 (13:55 -0700)]
[TF:XLA] Bump open source llvm revision to r330950
PiperOrigin-RevId:
194588403
Sourabh Bajaj [Fri, 27 Apr 2018 20:01:44 +0000 (13:01 -0700)]
Remove scope name from bfloat16
PiperOrigin-RevId:
194580957
Asim Shankar [Fri, 27 Apr 2018 19:59:20 +0000 (12:59 -0700)]
eager: Improve error message when GradientTape is used incorrectly.
PiperOrigin-RevId:
194580654
A. Unique TensorFlower [Fri, 27 Apr 2018 19:46:46 +0000 (12:46 -0700)]
Replace GrapplerFunctionItem input with a constant.
PiperOrigin-RevId:
194579253
A. Unique TensorFlower [Fri, 27 Apr 2018 18:41:21 +0000 (11:41 -0700)]
Internally rewrite @recompute_grad to use @custom_gradient
PiperOrigin-RevId:
194571125
A. Unique TensorFlower [Fri, 27 Apr 2018 16:25:52 +0000 (09:25 -0700)]
Added string conversion operator to tensorflow::StringPiece.
Marked ToString method as deprecated.
This will allow tensorflow::StringPiece to be replaced with absl::string_view (once the deprecated method is removed) as absl::string_view does not contain the ToString method.
PiperOrigin-RevId:
194551042
A. Unique TensorFlower [Fri, 27 Apr 2018 14:21:37 +0000 (07:21 -0700)]
Fix bug in @custom_gradient in Eager mode with numpy inputs
PiperOrigin-RevId:
194538828
Sanjoy Das [Fri, 27 Apr 2018 12:31:38 +0000 (05:31 -0700)]
[TF:XLA] Bump open source llvm revision to r330926
PiperOrigin-RevId:
194530610
Yu-Cheng Ling [Fri, 27 Apr 2018 07:07:07 +0000 (00:07 -0700)]
Automated g4 rollback of changelist
194306629
PiperOrigin-RevId:
194507274
A. Unique TensorFlower [Fri, 27 Apr 2018 05:06:36 +0000 (22:06 -0700)]
Reenable factorization_ops_test on ASAN after adding shard_count = 4. Tests now
finish with these stats: "max = 150.6s, min = 27.4s, avg = 66.3s, dev = 19.5s" over 1000 runs and this runtime distribution should be fairly safe for deadline of 300s.
PiperOrigin-RevId:
194500204
A. Unique TensorFlower [Fri, 27 Apr 2018 02:35:10 +0000 (19:35 -0700)]
Implement floor operator
PiperOrigin-RevId:
194490433
A. Unique TensorFlower [Fri, 27 Apr 2018 01:24:48 +0000 (18:24 -0700)]
Automated g4 rollback of changelist
194442428
PiperOrigin-RevId:
194485227
A. Unique TensorFlower [Fri, 27 Apr 2018 01:03:50 +0000 (18:03 -0700)]
Free scratch memory in ~BaseGPUDevice.
PiperOrigin-RevId:
194483351
A. Unique TensorFlower [Fri, 27 Apr 2018 00:56:08 +0000 (17:56 -0700)]
Implements linear no-offset (aka symmetric) quantizer.
PiperOrigin-RevId:
194482547
Allen Lavoie [Thu, 26 Apr 2018 23:40:16 +0000 (16:40 -0700)]
Split out SaveableObjects into their own file
Pulls a couple build rules out of tensorflow/python:training. I'd like to use a SaveableObject in :checkpointable (for saving some Python state by default), which means the file with SaveableObject has to be essientially dependency-free.
PiperOrigin-RevId:
194473987
Malcolm Reynolds [Thu, 26 Apr 2018 23:24:51 +0000 (16:24 -0700)]
More informative error message when loading a graph_def which uses unknown ops.
Fixes #17014
PiperOrigin-RevId:
194472083
Shashi Shekhar [Thu, 26 Apr 2018 23:11:11 +0000 (16:11 -0700)]
Fix some flakiness in test.
PiperOrigin-RevId:
194470125
A. Unique TensorFlower [Thu, 26 Apr 2018 23:01:00 +0000 (16:01 -0700)]
Internal change.
PiperOrigin-RevId:
194468535
Raghuraman Krishnamoorthi [Thu, 26 Apr 2018 22:40:15 +0000 (15:40 -0700)]
Handle variations in scoping of batch norms for correct unfused batch norm folding.
PiperOrigin-RevId:
194465704
Sanjoy Das [Thu, 26 Apr 2018 22:33:38 +0000 (15:33 -0700)]
Remove the inter-op thread pool
Forgot about this in cl/
194299356. However, when I checked cl/
194299356, I
found that we actually (incorrectly?) used the *intra* op thread pool in the
parallel CPU executable? Does that mean the inter op thread pool was always
unused?
PiperOrigin-RevId:
194464734
Shivani Agrawal [Thu, 26 Apr 2018 22:24:44 +0000 (15:24 -0700)]
[tf.data] Adds support for adding scalar value to `StatsAggregator`.
PiperOrigin-RevId:
194463407
Shivani Agrawal [Thu, 26 Apr 2018 22:15:37 +0000 (15:15 -0700)]
[tf.data] Changes description for `bytes_produced_stats` and `latency_stats` in accordance with the breaking changes in cl/
193432590.
PiperOrigin-RevId:
194461964
James Martens [Thu, 26 Apr 2018 22:13:48 +0000 (15:13 -0700)]
- Default values of cov and inv variables are now 0. Zero-debiasing (as in Adam) is used for the cov matrices. Note this this requires that cov variables, then inv variables, are all updated before the first training update is made. All examples have been modified to do this. NOTE: you *may* have to increase the damping value you use at the start of optimization after this change (or throughout, if you are using a constant value).
- Changed the initial default approximation used for generic registrations to "diagonal"
- Convenience properties for ops and thunks have all been removed, along with "make_ops_and_vars". User should only interface with "make_vars_and_create_op_thunks" (or maybe "create_ops_and_vars_thunks").
PiperOrigin-RevId:
194461623
A. Unique TensorFlower [Thu, 26 Apr 2018 22:08:24 +0000 (15:08 -0700)]
Add support for variables in tf.custom_gradient
PiperOrigin-RevId:
194460752
A. Unique TensorFlower [Thu, 26 Apr 2018 22:07:43 +0000 (15:07 -0700)]
Disable triangular_solve_test on ASAN due to flaky time outs.
PiperOrigin-RevId:
194460641
Daniel Zheng [Thu, 26 Apr 2018 22:02:53 +0000 (15:02 -0700)]
Edit tensorflow.org/community/swift page.
PiperOrigin-RevId:
194459754
A. Unique TensorFlower [Thu, 26 Apr 2018 21:59:29 +0000 (14:59 -0700)]
Adds optimization to convert division of sqrt to multiplication of rsqrt
PiperOrigin-RevId:
194459152
A. Unique TensorFlower [Thu, 26 Apr 2018 21:53:19 +0000 (14:53 -0700)]
Disable densenet_test on MSAN due to flaky time outs.
PiperOrigin-RevId:
194458270
A. Unique TensorFlower [Thu, 26 Apr 2018 20:53:17 +0000 (13:53 -0700)]
Remove unnecessary TF_NEED_GCP from build scripts.
PiperOrigin-RevId:
194448612
A. Unique TensorFlower [Thu, 26 Apr 2018 20:35:35 +0000 (13:35 -0700)]
Adding some slightly more exhaustive strided_slice test parameters.
PiperOrigin-RevId:
194446000
Tony Wang [Thu, 26 Apr 2018 20:30:15 +0000 (13:30 -0700)]
[TF:XLA] Add INTEL MKL_DNN Conv2d method to XLA/CPU backend
The INTEL MKL_DNN provides 32-bit Conv2d method. With INTEL_MKL flag set,
XLA backend emits runtime call to MKL_DNN Conv2d instead of Eigen.
PiperOrigin-RevId:
194445212
Benoit Steiner [Thu, 26 Apr 2018 20:19:39 +0000 (13:19 -0700)]
Run 2 passes of rewrites by default
PiperOrigin-RevId:
194443770
A. Unique TensorFlower [Thu, 26 Apr 2018 20:15:55 +0000 (13:15 -0700)]
Delay deleting RingReducer until group_size_tensor_ready_ has
been notified. Otherwise this can result in a bad pointer dereference
under some early abort conditions.
PiperOrigin-RevId:
194443206
Guangda Lai [Thu, 26 Apr 2018 20:12:04 +0000 (13:12 -0700)]
Fix build by adding op_lib dependencies to trt_engine_op_loader, and remove
unnecessary dependency from the tf_gen_op_libs.
PiperOrigin-RevId:
194442728
A. Unique TensorFlower [Thu, 26 Apr 2018 20:10:05 +0000 (13:10 -0700)]
Automated g4 rollback of changelist
194269675
PiperOrigin-RevId:
194442428
Akshay Modi [Thu, 26 Apr 2018 20:05:55 +0000 (13:05 -0700)]
gRPC worker cache owns a shared_ptr to the channel cache
PiperOrigin-RevId:
194441794
A. Unique TensorFlower [Thu, 26 Apr 2018 19:42:54 +0000 (12:42 -0700)]
For tf.gradients(), do not backpropagate through integer tensors.
All integer tensors are now considered constant with respect to all `xs`.
This fixes a bug in gradients through tf.while_loop.
PiperOrigin-RevId:
194438529
A. Unique TensorFlower [Thu, 26 Apr 2018 19:12:06 +0000 (12:12 -0700)]
Optimize functions in the function library.
PiperOrigin-RevId:
194434546
A. Unique TensorFlower [Thu, 26 Apr 2018 19:10:34 +0000 (12:10 -0700)]
Move */logging.cc into :platform_base since it already exposes the header loggging.h
This also brings env_time.h and env_time.cc, because on the 'default' platform logging needs env_time.
Add helpers tf_platform_srcs and tf_platform_hdrs to deal with files
that are not necessarily available in all platforms.
PiperOrigin-RevId:
194434322
A. Unique TensorFlower [Thu, 26 Apr 2018 18:40:46 +0000 (11:40 -0700)]
Simplify, test and document logic in instruction fusion that decides whether we
allow fusion when an operation needs to be duplicated.
PiperOrigin-RevId:
194429279
Anna R [Thu, 26 Apr 2018 18:25:43 +0000 (11:25 -0700)]
Removing @@ comments from core TensorFlow. They are no longer needed for exporting symbols to the TensorFlow API.
PiperOrigin-RevId:
194426855
Allen Lavoie [Thu, 26 Apr 2018 18:24:36 +0000 (11:24 -0700)]
Deprecate tfe.Network and associated utilities in favor of tf.keras.Model.
Also throws an error rather than silently saving incorrectly with tf.train.Checkpoint.
(In response to confusion over tf.train.Checkpoint with tfe.Network)
PiperOrigin-RevId:
194426679
A. Unique TensorFlower [Thu, 26 Apr 2018 18:24:26 +0000 (11:24 -0700)]
Automated g4 rollback of changelist
192536085
PiperOrigin-RevId:
194426650
Mingsheng Hong [Thu, 26 Apr 2018 17:49:01 +0000 (10:49 -0700)]
Updates on https://tensorflow.org/community/swift as part of the S4TF OSS launch on 4/26 morning.
PiperOrigin-RevId:
194419822
Dan Moldovan [Thu, 26 Apr 2018 17:31:21 +0000 (10:31 -0700)]
Add a skeleton dispatch context object, that can be used to control the dispatch rules and pass implementation-specific information down to the specialized operators.
PiperOrigin-RevId:
194416937
Allen Lavoie [Thu, 26 Apr 2018 17:30:54 +0000 (10:30 -0700)]
Remove "everything matched" assertions from CuDNN object-based checkpointing tests
After cl/
194315742 the assertions correctly point out that there are some Python objects which aren't matched (they don't have variables). Another option would be to mark these as special/optional, which we can implement if there's a need.
PiperOrigin-RevId:
194416864
Russell Power [Thu, 26 Apr 2018 17:25:04 +0000 (10:25 -0700)]
Limit the number of single allocation memory warnings.
PiperOrigin-RevId:
194415953
A. Unique TensorFlower [Thu, 26 Apr 2018 17:13:03 +0000 (10:13 -0700)]
Disable wrappers_test under ASAN since it sometimes times out.
PiperOrigin-RevId:
194413982
Alexandre Passos [Thu, 26 Apr 2018 17:11:14 +0000 (10:11 -0700)]
Simplify tfe.defun capture by not using convert_to_tensor
PiperOrigin-RevId:
194413685
Yanping Huang [Thu, 26 Apr 2018 16:56:00 +0000 (09:56 -0700)]
Fixing issue #13258. y is the square of Mahalanobis distance actually.
PiperOrigin-RevId:
194411230
A. Unique TensorFlower [Thu, 26 Apr 2018 16:44:30 +0000 (09:44 -0700)]
Disable vector_diffeomixture_test under ASAN to avoid timeouts.
PiperOrigin-RevId:
194409698
Shanqing Cai [Thu, 26 Apr 2018 15:53:46 +0000 (08:53 -0700)]
tfdbg: disable grpc_large_data_test on ASAN
PiperOrigin-RevId:
194402869
A. Unique TensorFlower [Thu, 26 Apr 2018 13:15:30 +0000 (06:15 -0700)]
Format header guards under tensorflow/core/grappler.
PiperOrigin-RevId:
194387041
James Martens [Thu, 26 Apr 2018 11:37:28 +0000 (04:37 -0700)]
- Adding support for Cholesky (inverse) factor multiplications.
- Refactored FisherFactor to use LinearOperator classes that know how to multiply themselves, compute their own trace, etc. This addresses the feature request: b/
73356352
- Fixed some problems with FisherEstimator construction
- More careful casting of damping constants before they are used
PiperOrigin-RevId:
194379298
A. Unique TensorFlower [Thu, 26 Apr 2018 10:40:24 +0000 (03:40 -0700)]
Support matching against shape string in HLO testing matchers
After this change a test can use op::Shape("f32[7,11]") instead of the
longer and harder to read op::Shape(ShapeUtil::MakeShape(F32, {7, 11}))
format.
PiperOrigin-RevId:
194373704
A. Unique TensorFlower [Thu, 26 Apr 2018 10:21:43 +0000 (03:21 -0700)]
Clarify limitation of `deps` in tf_gen_op_wrapper_py
PiperOrigin-RevId:
194372273
Pavithra Vijay [Thu, 26 Apr 2018 07:14:10 +0000 (00:14 -0700)]
Support CuDNN RNN layers in tf.keras.
PiperOrigin-RevId:
194355293
A. Unique TensorFlower [Thu, 26 Apr 2018 07:02:19 +0000 (00:02 -0700)]
[XLA] Redesign: migrate other xla/tests to use the new buidler.
- The set_return_value_test is not migrated because XlaBuilder does not support SetReturnValue.
- Delete a compute_constant_test case since ComputeConstant no longer accepts parameters.
- Delete CompilationCacheTest.MutatedComputation since the case no longer exists.
- Correct WhileTest.WhileWithMixedTupleElements which used an op from one builder in another builder.
- Disabled all CompilationCacheTest since there is no caching in the new design right now.
PiperOrigin-RevId:
194354250
A. Unique TensorFlower [Thu, 26 Apr 2018 03:49:52 +0000 (20:49 -0700)]
Added metadata to the TFLite model.
PiperOrigin-RevId:
194341479
A. Unique TensorFlower [Thu, 26 Apr 2018 03:11:00 +0000 (20:11 -0700)]
Disable gather_test under ASAN since it times out.
PiperOrigin-RevId:
194338928
A. Unique TensorFlower [Thu, 26 Apr 2018 02:47:06 +0000 (19:47 -0700)]
Go: Update generated wrapper functions for TensorFlow ops.
PiperOrigin-RevId:
194337205
Skye Wanderman-Milne [Thu, 26 Apr 2018 02:43:14 +0000 (19:43 -0700)]
Improve shape invariant error message for tf.while_loop.
PiperOrigin-RevId:
194336902
A. Unique TensorFlower [Thu, 26 Apr 2018 02:18:53 +0000 (19:18 -0700)]
Reduce number of combinations that are tested
PiperOrigin-RevId:
194335483
A. Unique TensorFlower [Thu, 26 Apr 2018 02:18:20 +0000 (19:18 -0700)]
Update ops-related pbtxt files.
PiperOrigin-RevId:
194335460
James Qin [Thu, 26 Apr 2018 02:00:21 +0000 (19:00 -0700)]
Cudnn RNN v2 kernels with autotune capability
CudnnRNN V2 kernels run all applicable cudnn rnn algorithms and pick the best one for following runs.
* To enable autotune, TF_CUDNN_RNN_USE_AUTOTUNE and TF_CUDNN_RNN_USE_V2 need to be set to {"1" or unset}.
* TF_CUDNN_RNN_USE_AUTOTUNE does not work with existing CudnnRNN kernels.
* V2 kernels work with existing cudnn checkpoints, since it doesn't change persistence format.
This change
* Introduces v2 kernels as templates inheriting the v1 kernels.
* Profiles fwd and bak runs in v2 kernel (forward pass)
* Exposes the chosen algorithm as fwd op output and bak op input.
* Changes rnn descriptor cache key to include AlgorithmDesc (since cudnn rnn descriptor can't be reused across different algorithms)
* Updates unittests s.t. it tests both v1 and v2 kernels. When testing v2 kernels, autotune is turned on.
PiperOrigin-RevId:
194333948
Akshay Agrawal [Thu, 26 Apr 2018 00:59:01 +0000 (17:59 -0700)]
Preserve guarantees about const-ness when creating TensorFlow functions.
PiperOrigin-RevId:
194328218
A. Unique TensorFlower [Wed, 25 Apr 2018 23:37:02 +0000 (16:37 -0700)]
Automated g4 rollback of changelist
194268101
PiperOrigin-RevId:
194318022
A. Unique TensorFlower [Wed, 25 Apr 2018 23:31:14 +0000 (16:31 -0700)]
[XLA] Redesign: local_client_test_base and tests that depend on it.
PiperOrigin-RevId:
194317245
Benoit Steiner [Wed, 25 Apr 2018 23:22:29 +0000 (16:22 -0700)]
Various small cleanups and simplifications
PiperOrigin-RevId:
194316037
Allen Lavoie [Wed, 25 Apr 2018 23:20:22 +0000 (16:20 -0700)]
Checkpointable: better handling of objects which aren't being restored
initialize_or_restore on a tf.train.Checkpoint status object will now initialize
any variables which aren't being restored, which is closer to the behavior when
executing eagerly (and makes it easier to use).
Fixes a bug where assert_consumed() would miss some Python objects which aren't
part of the object graph being restored. It will now (correctly/as documented)
complain about unmatched Python objects in the dependency graph.
PiperOrigin-RevId:
194315742
A. Unique TensorFlower [Wed, 25 Apr 2018 22:29:12 +0000 (15:29 -0700)]
[XLA] Redesign: migrate tests [m-r].* and vector_ops test.
Deleted a map_test case about versioned handle since such case no longer exists.
PiperOrigin-RevId:
194307861
A. Unique TensorFlower [Wed, 25 Apr 2018 22:28:22 +0000 (15:28 -0700)]
Docs
PiperOrigin-RevId:
194307738
A. Unique TensorFlower [Wed, 25 Apr 2018 22:25:15 +0000 (15:25 -0700)]
Automated g4 rollback of changelist
194261487
PiperOrigin-RevId:
194307293
Yu-Cheng Ling [Wed, 25 Apr 2018 22:21:16 +0000 (15:21 -0700)]
Refactoring: Remove a redundant map from model.h.
PiperOrigin-RevId:
194306629
A. Unique TensorFlower [Wed, 25 Apr 2018 21:59:57 +0000 (14:59 -0700)]
Disable factorization_ops_test on ASAN due to flaky timeouts.
PiperOrigin-RevId:
194303178
Sanjoy Das [Wed, 25 Apr 2018 21:36:17 +0000 (14:36 -0700)]
Remove the parallel cpu backend
PiperOrigin-RevId:
194299356
Younghee Kwon [Wed, 25 Apr 2018 21:25:30 +0000 (14:25 -0700)]
Fixing contrib.boosted_trees in Windows (module not linked)
Trying to fix #14292
PiperOrigin-RevId:
194297723
A. Unique TensorFlower [Wed, 25 Apr 2018 21:03:12 +0000 (14:03 -0700)]
Automated g4 rollback of changelist
193788768
PiperOrigin-RevId:
194293938
A. Unique TensorFlower [Wed, 25 Apr 2018 20:58:54 +0000 (13:58 -0700)]
Automated g4 rollback of changelist
193731341
PiperOrigin-RevId:
194293187
A. Unique TensorFlower [Wed, 25 Apr 2018 20:55:23 +0000 (13:55 -0700)]
Use a built-in function shape_n instead of running many array_ops.shape and
stacking the results.
PiperOrigin-RevId:
194292637
Justin Lebar [Wed, 25 Apr 2018 20:27:30 +0000 (13:27 -0700)]
Remove StreamExecutorUtil::ConvertStatus.
This function is a nop; StreamExecutor's Status is the same as
TensorFlow's Status.
PiperOrigin-RevId:
194288432