Sanjoy Das [Tue, 23 Jan 2018 23:20:43 +0000 (15:20 -0800)]
[TF:XLA:CPU] Alias tensorflow::gtl to gtl for brevity
PiperOrigin-RevId:
183001493
A. Unique TensorFlower [Tue, 23 Jan 2018 23:11:22 +0000 (15:11 -0800)]
Adds regularization_losses in remaining heads.
PiperOrigin-RevId:
183000075
A. Unique TensorFlower [Tue, 23 Jan 2018 23:08:26 +0000 (15:08 -0800)]
Switch models/ to use new test-specification format
PiperOrigin-RevId:
182999650
Skye Wanderman-Milne [Tue, 23 Jan 2018 23:02:46 +0000 (15:02 -0800)]
LabeledTensor: don't test Tensor.shape object, test value instead.
With the C API enabled, a new TensorShape object may be generated
every Tensor.get_shape call.
PiperOrigin-RevId:
182998609
Yong Tang [Tue, 23 Jan 2018 23:01:14 +0000 (15:01 -0800)]
Tweaks to PR 16101 which adds stream selection support for `tf.contrib.ffmpeg.decode_audio`.
PiperOrigin-RevId:
182998275
Skye Wanderman-Milne [Tue, 23 Jan 2018 22:54:22 +0000 (14:54 -0800)]
Make bijectors/reshape_test.py work with C API enabled.
Some of the error messages change with the C API enabled due to slight
differences in shape inference (in this case, the C API catches shape
errors sooner). This change refactors some of the tests to expect
different error messages in different cases.
PiperOrigin-RevId:
182997183
Skye Wanderman-Milne [Tue, 23 Jan 2018 22:31:15 +0000 (14:31 -0800)]
Turn private Operation and Tensor fields into properties.
Unfortunately there are many uses of these private fields. Turning
them into properties that call public API methods will allow enabling
the C API before fixing all such uses.
PiperOrigin-RevId:
182993539
A. Unique TensorFlower [Tue, 23 Jan 2018 22:25:36 +0000 (14:25 -0800)]
Replace the use of ConsumeValueOrDie with TF_ASSERT_OK_AND_ASSIGN.
PiperOrigin-RevId:
182992618
Amit Patankar [Tue, 23 Jan 2018 22:24:53 +0000 (14:24 -0800)]
Updating the update_version script to act appropriately for rc0 when it comes
to the install sources tested configurations table along with a warning. Using perl rather than sed as it supports regex look behind.
PiperOrigin-RevId:
182992467
Skye Wanderman-Milne [Tue, 23 Jan 2018 22:06:13 +0000 (14:06 -0800)]
C API: don't return source/sink control edges.
PiperOrigin-RevId:
182989157
Yao Zhang [Tue, 23 Jan 2018 21:51:44 +0000 (13:51 -0800)]
Fix StridedSlice.
(1) Use input index instead of input name to make sure the naming is unique,
because in rare cases, two inputs may connect to the same node.
(2) Consecutive if statements each should have a return and should not fall
through; this created un-wanted behavior. Use switch and break now.
PiperOrigin-RevId:
182986609
Nupur Garg [Tue, 23 Jan 2018 21:35:08 +0000 (13:35 -0800)]
Add support for conv and matmul operators with variable filter.
PiperOrigin-RevId:
182984061
A. Unique TensorFlower [Tue, 23 Jan 2018 21:12:00 +0000 (13:12 -0800)]
Add a test case to validate number of buckets after quantile generation.
Add an option to generate quantiles instead of sampled boundaries.
This option will always return exactly the number of requested boundaries,
regardless of duplicates and fewer inputs than requested boundaries.
PiperOrigin-RevId:
182980861
A. Unique TensorFlower [Tue, 23 Jan 2018 20:52:36 +0000 (12:52 -0800)]
Go: Update generated wrapper functions for TensorFlow ops.
PiperOrigin-RevId:
182977979
Anna R [Tue, 23 Jan 2018 20:48:44 +0000 (12:48 -0800)]
Adding tf_export decorators/calls to TensorFlow functions and constants.
PiperOrigin-RevId:
182977460
A. Unique TensorFlower [Tue, 23 Jan 2018 20:48:00 +0000 (12:48 -0800)]
[XLA:GPU] Support BF16 data type.
Add an HLO pass to the GPU backend to implement BF16 operations with F32
operations.
Define macro XLA_BACKEND_SUPPORTS_BFLOAT16=1 when building tests for the GPU
backend to enable BF16 tests for GPU. Enable bfloat16_test and other BF16 tests
for GPU.
Add hlo_element_type_converter_test.
Add convolution tests and matrix multiplication tests for BF16.
PiperOrigin-RevId:
182977358
A. Unique TensorFlower [Tue, 23 Jan 2018 20:41:07 +0000 (12:41 -0800)]
internal change
PiperOrigin-RevId:
182976445
Sanjoy Das [Tue, 23 Jan 2018 20:39:22 +0000 (12:39 -0800)]
[TF:XLA] Bump open source llvm revision to r323181
PiperOrigin-RevId:
182976206
A. Unique TensorFlower [Tue, 23 Jan 2018 20:24:21 +0000 (12:24 -0800)]
Internal Change
PiperOrigin-RevId:
182974191
A. Unique TensorFlower [Tue, 23 Jan 2018 20:19:06 +0000 (12:19 -0800)]
Update ops-related pbtxt files.
PiperOrigin-RevId:
182973494
A. Unique TensorFlower [Tue, 23 Jan 2018 20:14:52 +0000 (12:14 -0800)]
Allow fisher-vector multiplies on 1D vectors for CategoricalLogitsNegativeLogProbLoss.
PiperOrigin-RevId:
182972901
Akshay Modi [Tue, 23 Jan 2018 20:01:11 +0000 (12:01 -0800)]
Correctly incref items put into the tuple to keep msan happy.
PiperOrigin-RevId:
182970703
Justin Lebar [Tue, 23 Jan 2018 19:46:18 +0000 (11:46 -0800)]
[XLA] Remove LiteralTestUtil::EqualTuple / NearTuple.
Instead, simply DTRT when you pass a tuple to Equal/Near.
PiperOrigin-RevId:
182968283
Joshua V. Dillon [Tue, 23 Jan 2018 19:37:58 +0000 (11:37 -0800)]
Add `tf.contrib.distributions.Autoregressive`.
PiperOrigin-RevId:
182966827
Skye Wanderman-Milne [Tue, 23 Jan 2018 19:23:35 +0000 (11:23 -0800)]
Make feature_column_test.py run with the C API enabled.
I only enabled running the C API with test classes that other testing
found failures in to avoid increasing the test time unnecessarily.
PiperOrigin-RevId:
182964349
Alexandre Passos [Tue, 23 Jan 2018 19:20:29 +0000 (11:20 -0800)]
Moves batch ops to core and exposes tf.contrib.batching
PiperOrigin-RevId:
182963906
Gunhan Gulsoy [Tue, 23 Jan 2018 19:09:26 +0000 (11:09 -0800)]
Run TF opensource presubmits with sanybridge CPU optimizations.
PiperOrigin-RevId:
182961789
Akshay Agrawal [Tue, 23 Jan 2018 19:05:50 +0000 (11:05 -0800)]
TFE usability: Implement `ndim` for `EagerTensor`+NumPy compatibility
Third party libraries like matplotlib often access `ndim`.
PiperOrigin-RevId:
182961097
Benoit Steiner [Tue, 23 Jan 2018 18:59:00 +0000 (10:59 -0800)]
Removed unecessary dependency
PiperOrigin-RevId:
182959711
A. Unique TensorFlower [Tue, 23 Jan 2018 18:19:19 +0000 (10:19 -0800)]
Increase shard count of tensorflow/contrib/learn:linear_test to avoid test timeouts.
The test tensorflow/contrib/learn:linear_test has been getting flaky timeouts t
a low rate for some time, but they increased greatly under the address
sanitizer due to a recent change there.
This CL increases the test's shard count to bring the time for each shard
well under the 5 minute timeout.
PiperOrigin-RevId:
182953078
Mark Daoust [Tue, 23 Jan 2018 17:46:25 +0000 (09:46 -0800)]
The link translator expects "@{" for API links.
This is currently not rendering:
https://www.tensorflow.org/api_docs/python/tf/contrib/tpu/TPUEstimatorSpec
PiperOrigin-RevId:
182947562
A. Unique TensorFlower [Tue, 23 Jan 2018 17:18:11 +0000 (09:18 -0800)]
Correct handling of dtype for Categorical sampling.
PiperOrigin-RevId:
182943806
A. Unique TensorFlower [Tue, 23 Jan 2018 08:10:52 +0000 (00:10 -0800)]
Turn off address sanitizer checking for tensorflow/python/kernel_tests:sparse_ops_test
It getting too many timeouts.
PiperOrigin-RevId:
182892876
Justin Lebar [Tue, 23 Jan 2018 07:29:57 +0000 (23:29 -0800)]
[XLA] Inline definitions of NativeToPrimitiveType<T>.
These functions are hot in the evaluator, and that's silly, because
they're pure constants. Just inline them.
PiperOrigin-RevId:
182889992
Justin Lebar [Tue, 23 Jan 2018 06:56:04 +0000 (22:56 -0800)]
[XLA] Speed up HloEvaluator's convolution routine.
Skip two integer divisions when the window has no dilation. This is
good for a ~30% speedup on my machine.
PiperOrigin-RevId:
182887749
Jianwei Xie [Tue, 23 Jan 2018 06:06:10 +0000 (22:06 -0800)]
Puts the iterations_per_loop, which controls how many steps the TPU system to
run for each training/evaluatoin loop, into local variable collection.
PiperOrigin-RevId:
182884955
Yunxing Dai [Tue, 23 Jan 2018 04:43:06 +0000 (20:43 -0800)]
[XLA] Make kokoro happy in prng_test
Open source builds don't support bit_cast yet, use reinterpret cast instead.
RELNOTES: n/a
PiperOrigin-RevId:
182879589
Anna R [Tue, 23 Jan 2018 04:07:42 +0000 (20:07 -0800)]
Adding tf_export decorators/calls to TensorFlow functions and constants.
PiperOrigin-RevId:
182877401
A. Unique TensorFlower [Tue, 23 Jan 2018 02:45:56 +0000 (18:45 -0800)]
Go: Update generated wrapper functions for TensorFlow ops.
PiperOrigin-RevId:
182871435
A. Unique TensorFlower [Tue, 23 Jan 2018 02:18:46 +0000 (18:18 -0800)]
Update ops-related pbtxt files.
PiperOrigin-RevId:
182868865
Skye Wanderman-Milne [Tue, 23 Jan 2018 01:43:06 +0000 (17:43 -0800)]
_DefinedFunction.name should be a string.
With the C API enabled and python2, it was being set to a unicode
object.
PiperOrigin-RevId:
182865087
Jeremy Lau [Tue, 23 Jan 2018 01:26:43 +0000 (17:26 -0800)]
Reject retried RecvTensor requests.
Retried RecvTensorRequests are problematic because a RecvTensor with no
corresponding sender will wait forever, and the tensor may have been delivered
to a previous retry.
This change adds a unique request_id to each RecvTensor request, and we check
these request_ids against a set of recent request_ids. If a request_id is in the
recent set, we reject the RecvTensor request.
PiperOrigin-RevId:
182863245
Anna R [Tue, 23 Jan 2018 01:16:44 +0000 (17:16 -0800)]
Adding tf_export decorators/calls to TensorFlow functions and constants.
PiperOrigin-RevId:
182862075
Alexandre Passos [Tue, 23 Jan 2018 01:12:18 +0000 (17:12 -0800)]
Ops to get and set individual elements of TensorLists, and their gradients.
PiperOrigin-RevId:
182861559
Francois Chollet [Tue, 23 Jan 2018 01:10:35 +0000 (17:10 -0800)]
Add SeparableConv1D layer and refactor SeparableConv2D to maximize code reuse and maintenability.
PiperOrigin-RevId:
182861372
A. Unique TensorFlower [Tue, 23 Jan 2018 00:44:22 +0000 (16:44 -0800)]
Replaced LOGs with NSLogs.
Also inserted exit(-1) after LOG(FATAL)s.
Removed unused definition of CHECK macro.
PiperOrigin-RevId:
182857784
A. Unique TensorFlower [Tue, 23 Jan 2018 00:41:37 +0000 (16:41 -0800)]
Adds unit tests for squeeze, to test the case when all dims are 1.
PiperOrigin-RevId:
182857374
Frank Chen [Tue, 23 Jan 2018 00:28:47 +0000 (16:28 -0800)]
Add functionality to include an optional additional header to be sent on every request when using the GCS file system.
PiperOrigin-RevId:
182855525
Jeremy Lau [Mon, 22 Jan 2018 23:54:33 +0000 (15:54 -0800)]
Add file name and line number to statuses generated by OP_REQUIRES macros.
PiperOrigin-RevId:
182850486
Yunxing Dai [Mon, 22 Jan 2018 23:10:08 +0000 (15:10 -0800)]
[XLA][BF16] Add tests to demonstrate the brokeness of BF16 RNG functions in LLVM backends.
- Added prng tests to demonstrate the brokeness of BF16 RNG functions in LLVM backends.
PiperOrigin-RevId:
182843680
Anna R [Mon, 22 Jan 2018 22:59:19 +0000 (14:59 -0800)]
Adding tf_export decorators/calls to TensorFlow functions and constants.
PiperOrigin-RevId:
182841644
Alexandre Passos [Mon, 22 Jan 2018 22:21:16 +0000 (14:21 -0800)]
Thread-safety for the TFE tape code.
In multithreaded python any call to a python API function can release the GIL,
so we should not assume our data structures haven't changed across calls to
the python API.
PiperOrigin-RevId:
182835555
Benoit Steiner [Mon, 22 Jan 2018 21:58:08 +0000 (13:58 -0800)]
Track swapping candidates directly in memory instead of annotating them in the graph
PiperOrigin-RevId:
182831471
Yifei Feng [Mon, 22 Jan 2018 21:42:50 +0000 (13:42 -0800)]
Buildifier cleanup.
PiperOrigin-RevId:
182829120
Sanjoy Das [Mon, 22 Jan 2018 21:21:46 +0000 (13:21 -0800)]
[TF:XLA] Bump open source llvm revision to r323081
PiperOrigin-RevId:
182825537
Alexandre Passos [Mon, 22 Jan 2018 21:00:54 +0000 (13:00 -0800)]
Allowing custom creators for variable_scope.variable()
PiperOrigin-RevId:
182822430
A. Unique TensorFlower [Mon, 22 Jan 2018 19:27:33 +0000 (11:27 -0800)]
Extend the API with a pair of decorators that can convert functions inline.
Add logic to detect these decorators and treat them appropriately (e.g. when converting recursively, do all decorated functions as they are).
PiperOrigin-RevId:
182808673
A. Unique TensorFlower [Mon, 22 Jan 2018 19:18:44 +0000 (11:18 -0800)]
Add new src targets for C API and native TF
PiperOrigin-RevId:
182807258
Yifei Feng [Mon, 22 Jan 2018 19:07:28 +0000 (11:07 -0800)]
Add support for tf.Example to saved_model_cli.
You can now pass tf.Example as inputs to with option --input_examples, for example:
--input_examples 'input_key=[{"age":[25],"education":["11th"]}]'
PiperOrigin-RevId:
182805401
Shanqing Cai [Mon, 22 Jan 2018 19:02:40 +0000 (11:02 -0800)]
tf.keras recurrent Layers: Fix doc strings on default activations
* The default activation for input is `tanh`, not `linear`, as
the previous doc string might mislead readers to think.
* Also clarify that the default activation for recurrent is
`hard_sigmoid`.
PiperOrigin-RevId:
182804493
Yu-Cheng Ling [Mon, 22 Jan 2018 18:48:20 +0000 (10:48 -0800)]
Fix TFLite Conv reference implementation
PiperOrigin-RevId:
182802058
A. Unique TensorFlower [Mon, 22 Jan 2018 18:36:05 +0000 (10:36 -0800)]
Allow whitespace after commas in hparam strings.
E.g., allow 'a=b, c=d' or 'a=b,\nc=d' in addition to just 'a=b,c=d'.
PiperOrigin-RevId:
182800230
A. Unique TensorFlower [Mon, 22 Jan 2018 18:19:07 +0000 (10:19 -0800)]
Internal change
PiperOrigin-RevId:
182797605
Alexandre Passos [Mon, 22 Jan 2018 17:47:55 +0000 (09:47 -0800)]
mnist and resnet eager examples avoiding optimizer.minimize
PiperOrigin-RevId:
182792844
Akshay Modi [Mon, 22 Jan 2018 17:39:44 +0000 (09:39 -0800)]
Reducing op overhead for eager.
Benchmarks:
(new) tfe_py_fastpath_execute_matmul: 7.
86213080088 walltime
(old) gen_math_ops_matmul: 11.
2566947937 walltime
(The slowdown is due to adding the record_gradient callback)
This will be a 3-step process:
1. (This CL) Add a function that allows execution of an op on the fastpath.
2. Update python generation code to create 2 new python functions
(_op/_op_gradient_callback and _op_slowpath_fallback):
-- first function (_op) checks if it is in graph mode, if so, do the normal
thing, else call out to the function added in step 1.
-- second function does the else part similar to today (calling out to
args_to_matching_eager etc.)
3. Rename the first function generated above to be the canonical _op function.
PiperOrigin-RevId:
182791741
Skye Wanderman-Milne [Mon, 22 Jan 2018 16:31:23 +0000 (08:31 -0800)]
Make array_ops_test.py work with the C API enabled.
I only enabled the test class that other testing identified needed changes.
PiperOrigin-RevId:
182783262
A. Unique TensorFlower [Mon, 22 Jan 2018 16:19:56 +0000 (08:19 -0800)]
Implement native support of CycleGAN in tf.contrib.gan
PiperOrigin-RevId:
182782072
A. Unique TensorFlower [Mon, 22 Jan 2018 15:47:48 +0000 (07:47 -0800)]
Internal cleanup
PiperOrigin-RevId:
182777782
A. Unique TensorFlower [Mon, 22 Jan 2018 14:39:21 +0000 (06:39 -0800)]
Replace "Google Container Engine" by "Google Kubernetes Engine" in README.md.
PiperOrigin-RevId:
182771284
A. Unique TensorFlower [Sun, 21 Jan 2018 17:22:43 +0000 (09:22 -0800)]
Fix a bug in the associative & commutative rewrite in constant folding:
I finally tracked it down to the case where we do the transformation:
//
// + + = parent
// / \ / \
// C + -- > X + = children
// / \ / \
// X Y C Y = leaves
when there exists a control dependency from the child '+' to C. This control dependency creates a cycle in the graph when X and C are simply swapped. We can fix it by moving the control dependency and anchoring it on Y instead (We still need it in case we are inside a frame).
PiperOrigin-RevId:
182705910
Benjamin Kramer [Sun, 21 Jan 2018 16:03:46 +0000 (08:03 -0800)]
[TF:XLA] Bump open source llvm revision to r323005
PiperOrigin-RevId:
182703514
Anna R [Sun, 21 Jan 2018 04:34:12 +0000 (20:34 -0800)]
Improve error message in create_python_api.py. The new message would print
exact filenames that need to be added to the
tensorflow/tools/api/generator/BUILD file.
PiperOrigin-RevId:
182679235
Justin Lebar [Sun, 21 Jan 2018 02:13:15 +0000 (18:13 -0800)]
[XLA:GPU] Warn on more broken CUDA versions.
We've received additional data on which CUDA versions contain a ptx
assembler bug that affects XLA. Warn if you're using any of these.
PiperOrigin-RevId:
182675055
Benjamin Kramer [Sat, 20 Jan 2018 18:50:16 +0000 (10:50 -0800)]
Update XLA for LLVM r323001
This will require an LLVM version bump
PiperOrigin-RevId:
182661291
A. Unique TensorFlower [Sat, 20 Jan 2018 04:52:19 +0000 (20:52 -0800)]
Added BeamRoot to manage instances of BeamEntry for CTCBeamSearchDecoder
to avoid recursive destructor call.
PiperOrigin-RevId:
182625992
Benoit Steiner [Sat, 20 Jan 2018 03:13:43 +0000 (19:13 -0800)]
Turn the op_performance_data proto lib into a header only library by default
PiperOrigin-RevId:
182621348
Mark Heffernan [Sat, 20 Jan 2018 02:05:19 +0000 (18:05 -0800)]
Add check for valid outfeed shape on outfeed creation and in HloVerifier.
PiperOrigin-RevId:
182616345
A. Unique TensorFlower [Sat, 20 Jan 2018 01:18:25 +0000 (17:18 -0800)]
[XLA] Update R4 reduce window test cases.
PiperOrigin-RevId:
182612106
Jacques Pienaar [Sat, 20 Jan 2018 00:55:11 +0000 (16:55 -0800)]
Add clif rules for function.proto.
PiperOrigin-RevId:
182609604
A. Unique TensorFlower [Sat, 20 Jan 2018 00:54:39 +0000 (16:54 -0800)]
Make sure the same rewrite is not performed multiple times in ArithmeticOptimizer, and that added nodes are unique.
A couple of minor cleanups.
PiperOrigin-RevId:
182609552
Jacques Pienaar [Fri, 19 Jan 2018 23:42:39 +0000 (15:42 -0800)]
Report incompatible shapes in case of check failure.
PiperOrigin-RevId:
182600051
A. Unique TensorFlower [Fri, 19 Jan 2018 23:31:24 +0000 (15:31 -0800)]
Automated g4 rollback of changelist
182559231
PiperOrigin-RevId:
182598631
Nick Desaulniers [Fri, 19 Jan 2018 23:01:35 +0000 (15:01 -0800)]
[XLA] Add xla_dump_prepass_hlo_proto_to debug flag
Add xla_dump_prepass_hlo_proto_to debug flag for dumping HLO protos *BEFORE*
running HLO passes, which will perform backend specific optimizations.
PiperOrigin-RevId:
182594324
Zhixian Yan [Fri, 19 Jan 2018 22:51:11 +0000 (14:51 -0800)]
TFLite export optional tensor as -1 id.
PiperOrigin-RevId:
182592943
Shivani Agrawal [Fri, 19 Jan 2018 22:38:45 +0000 (14:38 -0800)]
[tf.data] Minor fix for saveable MapDataset and ParallelMapDataset.
PiperOrigin-RevId:
182591174
A. Unique TensorFlower [Fri, 19 Jan 2018 22:05:20 +0000 (14:05 -0800)]
Publish TPU Estimators Docs on github
PiperOrigin-RevId:
182585909
Brian Patton [Fri, 19 Jan 2018 22:03:27 +0000 (14:03 -0800)]
Filling in some holes in C64 support with Literals.
PiperOrigin-RevId:
182585597
Chris Leary [Fri, 19 Jan 2018 22:01:29 +0000 (14:01 -0800)]
[XLA] Simplify trivial pad/reduce-window combos into broadcasts.
PiperOrigin-RevId:
182585236
A. Unique TensorFlower [Fri, 19 Jan 2018 21:05:53 +0000 (13:05 -0800)]
Update pin for bazel-toolchains to latest version
https://github.com/bazelbuild/bazel-toolchains/releases/tag/
f3b0970
PiperOrigin-RevId:
182576952
Peter Hawkins [Fri, 19 Jan 2018 20:09:13 +0000 (12:09 -0800)]
[TF:XLA] Improve the error message when ComputeConstant fails.
Make comment in computation_builder.h more precise.
PiperOrigin-RevId:
182569583
A. Unique TensorFlower [Fri, 19 Jan 2018 20:08:47 +0000 (12:08 -0800)]
Add support for time_major shape format to the sequential RNN Op in TF Lite.
This option, if set, changes the shape format of the inputs and outputs to
[max_time, batch_size, depth]. If false, it uses [batch_size, max_time, depth].
By default, it is set to false.
PiperOrigin-RevId:
182569507
Alexandre Passos [Fri, 19 Jan 2018 20:03:50 +0000 (12:03 -0800)]
Handling resource captures when defining functions.
PiperOrigin-RevId:
182568701
Dustin Tran [Fri, 19 Jan 2018 20:00:33 +0000 (12:00 -0800)]
Move layers_dense_variational_impl.py to layers_dense_variational.py.
PiperOrigin-RevId:
182568201
A. Unique TensorFlower [Fri, 19 Jan 2018 19:59:42 +0000 (11:59 -0800)]
Change deprecated keep_dims parameter to keepdims.
PiperOrigin-RevId:
182568093
A. Unique TensorFlower [Fri, 19 Jan 2018 19:49:29 +0000 (11:49 -0800)]
[tpu:profiler] Do not generate output files from other cloud tools if no trace is collected.
PiperOrigin-RevId:
182566765
Sanjoy Das [Fri, 19 Jan 2018 19:46:55 +0000 (11:46 -0800)]
[TF:XLA] Bump open source llvm revision to r322942
PiperOrigin-RevId:
182566386
Dustin Tran [Fri, 19 Jan 2018 19:04:26 +0000 (11:04 -0800)]
Add convolutional Flipout layers.
PiperOrigin-RevId:
182559534
A. Unique TensorFlower [Fri, 19 Jan 2018 19:02:50 +0000 (11:02 -0800)]
Correct handling of dtype for multinomial sampling.
PiperOrigin-RevId:
182559231
A. Unique TensorFlower [Fri, 19 Jan 2018 19:01:45 +0000 (11:01 -0800)]
Add option for loop unrolling in MaskedAutoregressiveFlow bijector.
PiperOrigin-RevId:
182559030
Alexandre Passos [Fri, 19 Jan 2018 18:36:05 +0000 (10:36 -0800)]
Fixes issue with using lookup ops in datasets from eager.
Addresses issue #16160
PiperOrigin-RevId:
182554969