Benjamin Kramer [Tue, 15 May 2018 17:23:27 +0000 (10:23 -0700)]
[XLA] Cache computations when creating reduces in algebraic simplifier or batchnorm expander
Otherwise we create a lot of identical small computations. This shouldn't have
an effect except for cluttering the HLO, but turns out HloCSE doesn't look
inside of the computation of reduces, effectively never eliminating reduces
that were produced via this code path.
While there clean up some YAGNI, this only worked for F32 anyways, so just
hardcode it.
PiperOrigin-RevId:
196689316
Peter Hawkins [Tue, 15 May 2018 16:47:50 +0000 (09:47 -0700)]
[TF:XLA] Generalize existing support for keeping variables on an XLA device in reshaped form, instead allowing XLA devices to keep all tensors in a reshaped form outside an XLA computation.
PiperOrigin-RevId:
196683444
A. Unique TensorFlower [Tue, 15 May 2018 15:38:49 +0000 (08:38 -0700)]
Adding --distinct_host_configuration=false in tools/bazel.rc
When building TensorFlow, the host and target platforms are usually the same. So we don't have to distinct them by default. This helps avoid building the same targets twice.
If we need to do cross compilation, add --config=cross-compile to distinct them.
PiperOrigin-RevId:
196673728
A. Unique TensorFlower [Tue, 15 May 2018 15:09:04 +0000 (08:09 -0700)]
update doc
PiperOrigin-RevId:
196670274
A. Unique TensorFlower [Tue, 15 May 2018 14:28:24 +0000 (07:28 -0700)]
Small polishing changes in stream executor, no functional changes.
PiperOrigin-RevId:
196665609
A. Unique TensorFlower [Tue, 15 May 2018 09:48:07 +0000 (02:48 -0700)]
internal change
PiperOrigin-RevId:
196640024
A. Unique TensorFlower [Tue, 15 May 2018 08:22:13 +0000 (01:22 -0700)]
Reland improve fusion logic of (a dot b) * alpha
The previous fusion approach didn't work because a multiplication by a scalar value
will be changed into an explicit broadcast.
Another issue that is fixed in this CL is retrieving the constant value from
the literal. This depends on the PrimitiveType, before we always assumed it to be double.
Also when checking ImplementedAsGemm() we should not call it recursively, but instead just the check related to kDot.
Finally add an execution test and adjust the fusion logic test.
The fix for the issue that caused the revert is that we check earlier that consumer->operand_count() is 2.
Also, we fix the call to Get() to pass {} instead of {0}.
And we handle an output fusion node in GemmThunk to extract the dimension numbers from the dot operation.
PiperOrigin-RevId:
196631031
Dimitris Vardoulakis [Tue, 15 May 2018 05:28:06 +0000 (22:28 -0700)]
[TF:XLA] Scheduling test which demonstrates that we are ignoring the memory needed by subcomputations.
PiperOrigin-RevId:
196618347
Mustafa Ispir [Tue, 15 May 2018 05:04:50 +0000 (22:04 -0700)]
Added type check to feature column keys. So that users will get meaningful error messages in situations like: #19219
PiperOrigin-RevId:
196616638
Pavithra Vijay [Tue, 15 May 2018 04:43:55 +0000 (21:43 -0700)]
Partial update of tf.keras to the Keras 2.1.6 API.
Changes included are:
- Fix `batch_dot` when `axes=None`
- Add axis=-1 as an argument to keras.backend.softmax
- Fix ctc_batch_cost() error when batch_size = 1
- Print previous best in ModelCheckpoint callback
- Fix ReduceLROnPlateau callback
- Extend RemoteMonitor to send data as application/json
- Fix default dilation rate value in 2D separable conv.
- Fix for MobileNet model with undefined shape
- Disable require_flatten in nasnet & Add an error message for undefined shape.
- Improve tests by designating dtype of sample data
- Multi_gpu_model supporting legacy/fullCPU/fullGPU
PiperOrigin-RevId:
196615376
Youlong Cheng [Tue, 15 May 2018 04:28:44 +0000 (21:28 -0700)]
Function should inherit device information from the caller site.
PiperOrigin-RevId:
196614376
A. Unique TensorFlower [Tue, 15 May 2018 02:38:37 +0000 (19:38 -0700)]
Update SCALED mode to use the full quantized range of -128..127 when possible.
PiperOrigin-RevId:
196606455
Chris Leary [Tue, 15 May 2018 02:21:10 +0000 (19:21 -0700)]
[XLA] Move more comparison functions to non-test library.
PiperOrigin-RevId:
196605347
Michael Case [Tue, 15 May 2018 02:15:51 +0000 (19:15 -0700)]
Move model_to_estimator utility into Estimator from Keras.
Working on untangling TF/Estimator deps. We would like to get to a state
where Estimator depends on Keras and not vice versa
PiperOrigin-RevId:
196605024
Yunxing Dai [Tue, 15 May 2018 01:55:20 +0000 (18:55 -0700)]
Fix a bug in HloInstruction::ImplicitlyBroadcastsOperand where operands with the same dimension but different types are not considered broadcast.
PiperOrigin-RevId:
196603348
Rachel Lim [Tue, 15 May 2018 01:30:49 +0000 (18:30 -0700)]
Adds CsvDataset, which both reads and parses files.
Example usage: dataset = tf.contrib.data.CsvDataset(filenames, record_defaults=record_defaults, **kwargs)
Motivation: Fusing reading and parsing is more performant and correct than the previous canonical CSV parsing flow (`dataset = tf.data.TextLineDataset(filenames).map(lambda l: tf.decode_csv(l, **kwargs))`)
Closes #19077.
PiperOrigin-RevId:
196601381
A. Unique TensorFlower [Tue, 15 May 2018 01:29:59 +0000 (18:29 -0700)]
Disable LinearOperatorKroneckerTest.test_solve_{with_broadcast} temporarily.
PiperOrigin-RevId:
196601310
Derek Murray [Tue, 15 May 2018 01:04:31 +0000 (18:04 -0700)]
[tf.data] Add optional `args` argument to `Dataset.from_generator()`.
The new argument allows you to parameterize the generator with the value of a tf.Tensor,
enabling `Dataset.from_generator()` to be initialized from a placeholder or used in a
nested expression (such as `flat_map()` or `parallel_interleave()`). For example:
```python
def generator(n):
for _ in range(n):
yield n
# Define a generator based on a placeholder.
placeholder = tf.placeholder(tf.int64, shape=[])
dataset = tf.data.Dataset.from_generator(generator, tf.int64, args=(placeholder,))
# Define a generator based on the value of a nested dataset element.
dataset = tf.data.Dataset.range(10).flat_map(
lambda i: tf.data.Dataset.from_generator(generator, tf.int64, args=(i,)))
```
Fixes #19269. Partially addresses issue #13101.
PiperOrigin-RevId:
196598650
James Qin [Tue, 15 May 2018 00:51:11 +0000 (17:51 -0700)]
Introduce LossScalingOptimizer for mixed precision training.
PiperOrigin-RevId:
196597196
A. Unique TensorFlower [Mon, 14 May 2018 23:39:11 +0000 (16:39 -0700)]
Add an option to execute eval on cpu, regardless of training runs on TPU.
This will let users to benefit from TPU training, but avoid complex
eval metrics functions to be ported to TPU.
PiperOrigin-RevId:
196587755
Yu-Cheng Ling [Mon, 14 May 2018 23:35:22 +0000 (16:35 -0700)]
Refactoring: Make OpResolver return const pointer.
PiperOrigin-RevId:
196587227
Pavithra Vijay [Mon, 14 May 2018 23:30:49 +0000 (16:30 -0700)]
Automated g4 rollback of changelist
196565296
PiperOrigin-RevId:
196586601
A. Unique TensorFlower [Mon, 14 May 2018 23:26:33 +0000 (16:26 -0700)]
ClangTidy - Readability cleanup:/code-findings-fixes.
* unused using-declarations
* redundant string conversions
* C-style casts
* redundant get() call on smart pointer
* the 'empty' method should be used to check for emptiness instead of 'size'
PiperOrigin-RevId:
196585984
Suharsh Sivakumar [Mon, 14 May 2018 23:17:46 +0000 (16:17 -0700)]
Make sure that variables aren't created as partition variables since only non-scalar partition variables are supported.
PiperOrigin-RevId:
196584749
Reed Wanderman-Milne [Mon, 14 May 2018 23:07:33 +0000 (16:07 -0700)]
Fix bug where custom layers could crash.
Layer.add_weight would crash when called without a dtype or initializer.
PiperOrigin-RevId:
196583182
James Qin [Mon, 14 May 2018 22:51:36 +0000 (15:51 -0700)]
Fix functional.While(), functional.For(rewrite_with_while)
When executing on GPU, synchronously copy cond result from device to host.
PiperOrigin-RevId:
196580820
A. Unique TensorFlower [Mon, 14 May 2018 22:50:06 +0000 (15:50 -0700)]
Go: Update generated wrapper functions for TensorFlow ops.
PiperOrigin-RevId:
196580619
A. Unique TensorFlower [Mon, 14 May 2018 22:45:33 +0000 (15:45 -0700)]
Add ExplicitShapes as a new shape inference function for Ops with
multiple outputs, each of which is explicitly declared.
PiperOrigin-RevId:
196579920
Pavithra Vijay [Mon, 14 May 2018 22:32:44 +0000 (15:32 -0700)]
Remove CuDNNRNN timing test.
PiperOrigin-RevId:
196578043
Yu-Cheng Ling [Mon, 14 May 2018 22:27:51 +0000 (15:27 -0700)]
Fix copy functions of MutableOpResolver
PiperOrigin-RevId:
196577314
Shashi Shekhar [Mon, 14 May 2018 22:22:04 +0000 (15:22 -0700)]
Used aligned allocation for vector cache.
PiperOrigin-RevId:
196576497
A. Unique TensorFlower [Mon, 14 May 2018 22:20:07 +0000 (15:20 -0700)]
Update ops-related pbtxt files.
PiperOrigin-RevId:
196576189
A. Unique TensorFlower [Mon, 14 May 2018 22:15:38 +0000 (15:15 -0700)]
Internal change.
PiperOrigin-RevId:
196575483
Yu-Cheng Ling [Mon, 14 May 2018 22:14:53 +0000 (15:14 -0700)]
Refactoring: Remove lite/tools:mutable_op_resolver dependency.
PiperOrigin-RevId:
196575387
A. Unique TensorFlower [Mon, 14 May 2018 22:06:17 +0000 (15:06 -0700)]
Stricter analysis for functional conditional generation
PiperOrigin-RevId:
196573938
Alexandre Passos [Mon, 14 May 2018 21:58:17 +0000 (14:58 -0700)]
Do shape validation in ScatterNd kernel, not just the shape inference function.
Fixes #18648
PiperOrigin-RevId:
196572262
Asim Shankar [Mon, 14 May 2018 21:56:36 +0000 (14:56 -0700)]
Fail gracefully with a helpful error message when provided with conflicting
visible_devices_list.
See #19083
See #18861
More generally, this change avoids assertion failures (that will bring the
whole process down) on a few code-paths that can be triggerred by user input.
PiperOrigin-RevId:
196572013
A. Unique TensorFlower [Mon, 14 May 2018 21:49:08 +0000 (14:49 -0700)]
Internal change.
PiperOrigin-RevId:
196570742
A. Unique TensorFlower [Mon, 14 May 2018 21:44:31 +0000 (14:44 -0700)]
Make CollectiveParamReducerLocal::InitInstanceSharedParams non-blocking.
PiperOrigin-RevId:
196570011
Gunhan Gulsoy [Mon, 14 May 2018 21:32:15 +0000 (14:32 -0700)]
Automated g4 rollback of changelist
196456687
PiperOrigin-RevId:
196567964
A. Unique TensorFlower [Mon, 14 May 2018 21:32:03 +0000 (14:32 -0700)]
Add score filtering to tf.image.non_max_suppression.
PiperOrigin-RevId:
196567928
Akshay Agrawal [Mon, 14 May 2018 21:25:55 +0000 (14:25 -0700)]
Update the eager programmer's guide to reflect the fact that "==" is not
implemented in the natural way for the Tensor class.
PiperOrigin-RevId:
196566940
Yunxing Dai [Mon, 14 May 2018 21:18:11 +0000 (14:18 -0700)]
ReverseDFS scheduler reverses the heuristics used in DFSScheduler.
Also fixes hlo_schedule_test to remove the expected order on unrelated operations.
PiperOrigin-RevId:
196565651
Gunhan Gulsoy [Mon, 14 May 2018 21:16:09 +0000 (14:16 -0700)]
Disable flaky cudnn_recurrent test
PiperOrigin-RevId:
196565296
Guangda Lai [Mon, 14 May 2018 21:15:14 +0000 (14:15 -0700)]
Reenable virtual gpu test, and decrease the number of testing rounds.
PiperOrigin-RevId:
196565153
Justin Lebar [Mon, 14 May 2018 21:09:01 +0000 (14:09 -0700)]
[XLA] Ergonomic improvements to --xla_hlo_profile.
- Don't display ops with 0 optimal seconds and 0 actual cycles. These
are ops that were expected to be free and were actually free.
- Fix HloCostAnalysis to mark parameters, constants, and
get-tuple-element as expected-to-be-free per the definition above.
- Allow optimal-seconds < 0 to indicate "I don't know". Use this for
custom calls, and then hide such ops from the "seconds above the
optimum" table.
- Don't display "<none>" and "<unknown>" -- instead, just display the
empty string. Less visual noise.
- Instead of showing ~5 ops per category in the categories tables, show
everything. This isn't so noisy now that we're hiding "free" ops, and
it makes finding optimization opportunities much easier.
PiperOrigin-RevId:
196564177
Jacques Pienaar [Mon, 14 May 2018 21:04:05 +0000 (14:04 -0700)]
Add If op rewriter.
* Add attribute to If op to indicate if lowering to switch-merge form is
needed;
* Add initial version of If op rewriter than transforms a If op into
switch/merge nodes (as would have been constructed via tf.cond) if the If op
has the lowering attribute set.
- The pass is not ready for general use and, for example, does not support
reference data types.
PiperOrigin-RevId:
196563421
Jianwei Xie [Mon, 14 May 2018 20:53:00 +0000 (13:53 -0700)]
Reserves 'context' key in TPUEstimator params dict.
PiperOrigin-RevId:
196561620
Saurabh Saxena [Mon, 14 May 2018 20:44:52 +0000 (13:44 -0700)]
Add CheckpointInputPipelineHook to the API docs.
PiperOrigin-RevId:
196560221
Benoit Steiner [Mon, 14 May 2018 20:33:46 +0000 (13:33 -0700)]
Added support for strided slicing of symbolic shapes
PiperOrigin-RevId:
196558466
A. Unique TensorFlower [Mon, 14 May 2018 20:30:53 +0000 (13:30 -0700)]
Resolve inlined function input/output types from GrapplerFunctionItem.
Remove duplicated code to resolve type from attributes.
PiperOrigin-RevId:
196558061
Pete Warden [Mon, 14 May 2018 20:24:58 +0000 (13:24 -0700)]
Updated speech commands example to use new dataset
PiperOrigin-RevId:
196557132
A. Unique TensorFlower [Mon, 14 May 2018 20:22:09 +0000 (13:22 -0700)]
Various ClangTidy-inspired fixes.
PiperOrigin-RevId:
196556727
A. Unique TensorFlower [Mon, 14 May 2018 20:00:26 +0000 (13:00 -0700)]
add memory utilization estimate for HLO op profile.
PiperOrigin-RevId:
196553696
A. Unique TensorFlower [Mon, 14 May 2018 19:03:50 +0000 (12:03 -0700)]
Deletes an unused private method in head.py
PiperOrigin-RevId:
196545696
A. Unique TensorFlower [Mon, 14 May 2018 18:40:50 +0000 (11:40 -0700)]
Don't check that bool arrays are quantized.
PiperOrigin-RevId:
196541955
A. Unique TensorFlower [Mon, 14 May 2018 17:43:08 +0000 (10:43 -0700)]
Use utility methods to compute AttrValue hash code and check for equality.
PiperOrigin-RevId:
196531355
A. Unique TensorFlower [Mon, 14 May 2018 16:51:52 +0000 (09:51 -0700)]
avoid having stream_executor depend on tensorflow/core
PiperOrigin-RevId:
196521381
A. Unique TensorFlower [Mon, 14 May 2018 16:45:42 +0000 (09:45 -0700)]
Extracts the following optimizations into methods:
PartialConstPropThroughIdentityN
ConstantPushDown
PiperOrigin-RevId:
196520167
A. Unique TensorFlower [Mon, 14 May 2018 16:06:25 +0000 (09:06 -0700)]
Pre-factoring: Fix overly specific test expectations to prepare for multi-output fusion.
PiperOrigin-RevId:
196514026
A. Unique TensorFlower [Mon, 14 May 2018 14:53:04 +0000 (07:53 -0700)]
Prevent removal of constant inputs to passthrough ops.
PiperOrigin-RevId:
196505061
Gunhan Gulsoy [Mon, 14 May 2018 05:30:21 +0000 (22:30 -0700)]
Bump protobuf dependency to fix windows build issues.
PiperOrigin-RevId:
196456687
Yu-Cheng Ling [Mon, 14 May 2018 02:52:18 +0000 (19:52 -0700)]
Introduce op version into TFLite
PiperOrigin-RevId:
196448769
Shashi Shekhar [Mon, 14 May 2018 01:21:21 +0000 (18:21 -0700)]
Update UI for Camera example.
PiperOrigin-RevId:
196444970
A. Unique TensorFlower [Sun, 13 May 2018 23:52:14 +0000 (16:52 -0700)]
ClangTidy - Legacy cleanup:
* use nullptr
* converting integer literal to bool, use bool literal instead
* annotate this function with 'override' or (rarely) 'final'
* prefer using 'override' or (rarely) 'final' instead of 'virtual'
PiperOrigin-RevId:
196441181
A. Unique TensorFlower [Sun, 13 May 2018 20:53:35 +0000 (13:53 -0700)]
Fix logic bug: should use logical-AND, not bitwise-AND.
PiperOrigin-RevId:
196435466
A. Unique TensorFlower [Sun, 13 May 2018 17:09:58 +0000 (10:09 -0700)]
Make CPython implementation function type-correct, which removes UB from calling a function through a pointer of the wrong type, and also removes a C-style cast.
PiperOrigin-RevId:
196428430
A. Unique TensorFlower [Sat, 12 May 2018 22:40:29 +0000 (15:40 -0700)]
Automated g4 rollback of changelist
196349902
PiperOrigin-RevId:
196387391
A. Unique TensorFlower [Sat, 12 May 2018 14:13:06 +0000 (07:13 -0700)]
Check that the module group metadata builder correctly detects whether there are more than one companion instruction per device/module.
PiperOrigin-RevId:
196369766
A. Unique TensorFlower [Sat, 12 May 2018 06:48:06 +0000 (23:48 -0700)]
Adding cuDNN header dependency to targets that include the cuDNN header file.
PiperOrigin-RevId:
196349902
A. Unique TensorFlower [Sat, 12 May 2018 02:41:09 +0000 (19:41 -0700)]
Automated g4 rollback of changelist
196166118
PiperOrigin-RevId:
196340289
A. Unique TensorFlower [Sat, 12 May 2018 02:38:48 +0000 (19:38 -0700)]
Making GetInput from kernel_util.h return a pointer to const data.
PiperOrigin-RevId:
196340200
A. Unique TensorFlower [Sat, 12 May 2018 02:31:37 +0000 (19:31 -0700)]
Add bool conversion in toco for tflite since bool is supported by tflite.
PiperOrigin-RevId:
196339883
A. Unique TensorFlower [Sat, 12 May 2018 00:53:06 +0000 (17:53 -0700)]
Implement additional options to control the string output of HloInstruction and HloComputation.
PiperOrigin-RevId:
196334340
Jacques Pienaar [Sat, 12 May 2018 00:32:25 +0000 (17:32 -0700)]
Add IsCondMerge.
PiperOrigin-RevId:
196332782
Saurabh Saxena [Sat, 12 May 2018 00:14:29 +0000 (17:14 -0700)]
Add hook for checkpointing input pipeline while training with Estimator.
PiperOrigin-RevId:
196331223
Justin Lebar [Fri, 11 May 2018 23:04:54 +0000 (16:04 -0700)]
[XLA] s/tensorflow::Status/Status/.
These are type aliases of one another; we'd like to be consistent and
use the shorter one.
PiperOrigin-RevId:
196322955
Allen Lavoie [Fri, 11 May 2018 22:58:39 +0000 (15:58 -0700)]
Checkpointable: Remove overzealous error checking from tf.make_template
It was checking that all variables in the Template's scope were dependencies, but Optimizer slot variables are created with the same prefix (and should not be dependencies).
Conversely, eager execution's eager slot variable creation meant that Templates create unnecessary/somewhat harmful dependencies on restored slot variables. Fixes that.
PiperOrigin-RevId:
196321999
A. Unique TensorFlower [Fri, 11 May 2018 22:44:39 +0000 (15:44 -0700)]
In broadcaster.cc send from the input tensor, not the output, since
it may not have been forwarded. Add non-forwarding cases to unittest.
PiperOrigin-RevId:
196320304
Mark Daoust [Fri, 11 May 2018 22:17:58 +0000 (15:17 -0700)]
Add `<float>` to the call to `Tensor`
PiperOrigin-RevId:
196316735
A. Unique TensorFlower [Fri, 11 May 2018 22:07:48 +0000 (15:07 -0700)]
Expose MaybeGetMinimumShape for use in cost estimators other than OpLevelCostEstimator.
PiperOrigin-RevId:
196315239
A. Unique TensorFlower [Fri, 11 May 2018 22:07:47 +0000 (15:07 -0700)]
Introduce an indirection to access posix/error.h, so implementations don't
have to worry about platform details.
PiperOrigin-RevId:
196315234
A. Unique TensorFlower [Fri, 11 May 2018 22:07:24 +0000 (15:07 -0700)]
Implement constant-only ListDiff Op in XLA to support dense layer.
PiperOrigin-RevId:
196315170
Justin Lebar [Fri, 11 May 2018 22:04:41 +0000 (15:04 -0700)]
[XLA:GPU] Load kernel thunks' kernels before running them.
The motivation here is that with --xla_hlo_profile, we count the time
spent in Thunk::ExecuteOnStream, but we don't want to count the time
spent loading the CUDA code into the GPU as time spent in the first
kernel thunk we try to run.
PiperOrigin-RevId:
196314733
A. Unique TensorFlower [Fri, 11 May 2018 22:03:34 +0000 (15:03 -0700)]
Modify the python interface to toco to provide arithmetic operations used by the model.
PiperOrigin-RevId:
196314416
Sanjoy Das [Fri, 11 May 2018 22:02:33 +0000 (15:02 -0700)]
Remove degenerate batch dimensions form batch dot
The way things are set up today this specific optimization isn't particularly
important, but I want to implement a follow-on optimization in
BatchDotSimplification to transform (non-degenerate) batch GEMV operations into
GEMM which I'm expecting to help us a bit.
This would normally be in the algebraic simplifier, but we want to fixpoint this
pass before we run DotDecomposer. This will become more important when we
implement the (non-degenerate) batch GEMV operations -> GEMM transform.
PiperOrigin-RevId:
196314230
Sanjoy Das [Fri, 11 May 2018 22:02:15 +0000 (15:02 -0700)]
[TF:XLA] Bump open source llvm revision to r332085
PiperOrigin-RevId:
196314181
Allen Lavoie [Fri, 11 May 2018 21:45:36 +0000 (14:45 -0700)]
Checkpointable: Add UniqueNameTracker for managing dependencies on arbitrarily named objects
Makes generating object-unique dependency names easier, which will hopefully discourage people from using Graph-global names with Checkpointable.
PiperOrigin-RevId:
196311633
A. Unique TensorFlower [Fri, 11 May 2018 21:24:47 +0000 (14:24 -0700)]
Add some python wrapper for TF_ApiDefMap.
PiperOrigin-RevId:
196308677
Sanjoy Das [Fri, 11 May 2018 21:05:38 +0000 (14:05 -0700)]
Make the elemental ir emitter for dot operations respect contraction dims
PiperOrigin-RevId:
196305803
Alexandre Passos [Fri, 11 May 2018 20:42:31 +0000 (13:42 -0700)]
Allow zero initializer by default for string variables (no reason not to)
PiperOrigin-RevId:
196302302
A. Unique TensorFlower [Fri, 11 May 2018 19:55:55 +0000 (12:55 -0700)]
Reads the L2 and L3 cache sizes from the system instead of using hard-coded constants.
PiperOrigin-RevId:
196296096
Jeremy Lau [Fri, 11 May 2018 19:39:40 +0000 (12:39 -0700)]
Update HeapSimulator to use BufferValue.
PiperOrigin-RevId:
196293610
Shashi Shekhar [Fri, 11 May 2018 19:36:40 +0000 (12:36 -0700)]
Change default number of threads to 1.
PiperOrigin-RevId:
196293227
Amit Patankar [Fri, 11 May 2018 19:27:40 +0000 (12:27 -0700)]
Updating the descriptions for TensorFlow.
PiperOrigin-RevId:
196291844
A. Unique TensorFlower [Fri, 11 May 2018 18:04:33 +0000 (11:04 -0700)]
RELNOTES: This allows the use of '.' in variables (e.g. "hparams.parse('a.b=1.0')"), which would previously raise an error. This will correspond to an attribute name with an embedded '.' symbol (e.g. 'a.b'), which can only be accessed indirectly (e.g. through getattr and setattr). To set this up the user will first need to explicitly add the variable to the hparam object (e.g. "hparams.add_hparam(name='a.b', value=0.0)").
NOTE: the use of '.' in variable names is now allowed, but it is not recommended.
PiperOrigin-RevId:
196278660
A. Unique TensorFlower [Fri, 11 May 2018 18:04:22 +0000 (11:04 -0700)]
Allow communicating instructions within a kCall computation.
PiperOrigin-RevId:
196278635
A. Unique TensorFlower [Fri, 11 May 2018 18:01:30 +0000 (11:01 -0700)]
[XLA] Fix a doc that still mentioned computation_builder.
PiperOrigin-RevId:
196278086
Michael Case [Fri, 11 May 2018 17:58:17 +0000 (10:58 -0700)]
Move fn_args utility into core TensorFlow from Estimator.
Working on untangling TF/Estimator deps. Some core TF code depends on Estimator
by using the fn_args utility function within Estimator.
PiperOrigin-RevId:
196277612
Suharsh Sivakumar [Fri, 11 May 2018 17:51:24 +0000 (10:51 -0700)]
Introduce ordered_inputs option to graph_matcher to allow simpler matching of commutative operations.
#18919
PiperOrigin-RevId:
196276502