platform/upstream/tensorflow.git
6 years agoReplace raw_input/input with six.moves.input for python 2/3 (#18461)
Yong Tang [Tue, 17 Apr 2018 23:44:53 +0000 (16:44 -0700)]
Replace raw_input/input with six.moves.input for python 2/3 (#18461)

In python 3 input is the equivalent of raw_input in python 2.
This fix is an enhancement to replace raw_input/input with
six.moves.input, which is compatible with both python 2 and python 3.

Signed-off-by: Yong Tang <yong.tang.github@outlook.com>
6 years agoFix shape validation error with tf.nn.conv3d_transpose (#18465)
Yong Tang [Tue, 17 Apr 2018 23:44:21 +0000 (16:44 -0700)]
Fix shape validation error with tf.nn.conv3d_transpose (#18465)

* Fix shape validation error with tf.nn.conv3d_transpose

This fix tries to address the issue raised in 18460.
In `tf.nn.conv3d_transpose` when list or np array is passed,
the validate of the output shape with filter shape uses
`output_shape[4]` (channel). This will not work with
`data_format='NCDHW'`.

This fix fixes the issue by replace `output_shape[4]` with `output_shape[axis]`.

This fix also adds a test case. Before this fix, the test case will fail.

This fix fixes 18460.

Signed-off-by: Yong Tang <yong.tang.github@outlook.com>
* Add test case for output and filter shape check in conv3d_transpose

Signed-off-by: Yong Tang <yong.tang.github@outlook.com>
* Fix pylint issue

Signed-off-by: Yong Tang <yong.tang.github@outlook.com>
* Also fix the error message

6 years agoFixed the bug in mkl_input_conversion_op when reorder is not needed (#18498)
shengfuintel [Tue, 17 Apr 2018 23:44:10 +0000 (16:44 -0700)]
Fixed the bug in mkl_input_conversion_op when reorder is not needed (#18498)

6 years ago[INTEL MK] Updating MKL CPU CI build and test. (#18513)
AG Ramesh [Tue, 17 Apr 2018 23:43:53 +0000 (16:43 -0700)]
[INTEL MK] Updating MKL CPU CI build and test. (#18513)

* Setting  KMP_BLOCKTIME to 0

* Adding comments

6 years agoMerge pull request #18618 from yongtang/04172018-duplicate-random_ops
Martin Wicke [Tue, 17 Apr 2018 23:26:41 +0000 (16:26 -0700)]
Merge pull request #18618 from yongtang/04172018-duplicate-random_ops

Remove duplicate imports in all python source

6 years ago[tf.data] Fix a device placement issue in `prefetch_to_device()`. (#18607)
Derek Murray [Tue, 17 Apr 2018 21:59:58 +0000 (14:59 -0700)]
[tf.data] Fix a device placement issue in `prefetch_to_device()`. (#18607)

* [tf.data] Fix a device placement issue in `prefetch_to_device()`.

Previously, the `iterator_get_device()` op was being infeasibly colocated with
both the iterator and placed on the prefetch target device. Move the
construction of that op outside the `with device():` block to fix this.

Also enable the relevant test to run as a CUDA test.

* Import the cuda_py_test rule.

6 years agoMerge pull request #18615 from yongtang/04162018-TFRecordDataset-shape
Martin Wicke [Tue, 17 Apr 2018 21:59:50 +0000 (14:59 -0700)]
Merge pull request #18615 from yongtang/04162018-TFRecordDataset-shape

Add shape check to TFRecordDataset

6 years agoMerge pull request #18613 from yifeif/branch_193234819
Yifei Feng [Tue, 17 Apr 2018 21:54:02 +0000 (14:54 -0700)]
Merge pull request #18613 from yifeif/branch_193234819

Branch 193234819

6 years agoClean up remaining issues.
Yong Tang [Tue, 17 Apr 2018 21:36:21 +0000 (21:36 +0000)]
Clean up remaining issues.

Signed-off-by: Yong Tang <yong.tang.github@outlook.com>
6 years agoFix duplicate import in kmeans_test.py
Yong Tang [Tue, 17 Apr 2018 21:36:05 +0000 (21:36 +0000)]
Fix duplicate import in kmeans_test.py

Signed-off-by: Yong Tang <yong.tang.github@outlook.com>
6 years agoRemove duplicate imports in several places
Yong Tang [Tue, 17 Apr 2018 21:33:27 +0000 (21:33 +0000)]
Remove duplicate imports in several places

Wrote a script to scan throught the python files in the repo,
and found the remaining duplicate imports in some python files like:
```
 from tensorflow.python.ops import random_ops
-from tensorflow.python.ops import random_ops
 from tensorflow.python.util.deprecation import deprecated
```
This fix removed all of them for duplicate imports.

Signed-off-by: Yong Tang <yong.tang.github@outlook.com>
6 years agoMerge pull request #18504 from akshaym/eager-custom-ops
Akshay Modi [Tue, 17 Apr 2018 21:40:32 +0000 (14:40 -0700)]
Merge pull request #18504 from akshaym/eager-custom-ops

Use eager compatible wrappers in load_library for custom ops

6 years agoMerge pull request #18614 from yongtang/04172018-duplicate-import
Martin Wicke [Tue, 17 Apr 2018 21:14:41 +0000 (14:14 -0700)]
Merge pull request #18614 from yongtang/04172018-duplicate-import

Remove duplicate import in compat.py

6 years agoMerge pull request #18561 from yongtang/18529-doc-cast
Martin Wicke [Tue, 17 Apr 2018 20:34:33 +0000 (13:34 -0700)]
Merge pull request #18561 from yongtang/18529-doc-cast

Update docs for tf.cast with supported types

6 years agoMerge pull request #18563 from yongtang/04162018-gemmlowp
Martin Wicke [Tue, 17 Apr 2018 20:33:05 +0000 (13:33 -0700)]
Merge pull request #18563 from yongtang/04162018-gemmlowp

Update gemmlowp version for cmake build

6 years agoMerge pull request #17562 from giuscri/typo
Martin Wicke [Tue, 17 Apr 2018 20:18:06 +0000 (13:18 -0700)]
Merge pull request #17562 from giuscri/typo

Fix markdown error in layers tutorial.

6 years agoMerge pull request #17512 from imsheridan/fix_broken_link_graph
Martin Wicke [Tue, 17 Apr 2018 20:17:41 +0000 (13:17 -0700)]
Merge pull request #17512 from imsheridan/fix_broken_link_graph

Fix broken link of distributed in graphs.md

6 years agoUpdate clip_ops_test.py
Yifei Feng [Tue, 17 Apr 2018 20:05:09 +0000 (13:05 -0700)]
Update clip_ops_test.py

6 years agoMerge pull request #17395 from facaiy/ENH/adamax_optimizer
annarev [Tue, 17 Apr 2018 19:54:30 +0000 (12:54 -0700)]
Merge pull request #17395 from facaiy/ENH/adamax_optimizer

add AdaMax optimizer

6 years agoDuplicate imports in histogram_ops.py
Yong Tang [Tue, 17 Apr 2018 19:48:42 +0000 (19:48 +0000)]
Duplicate imports in histogram_ops.py

Signed-off-by: Yong Tang <yong.tang.github@outlook.com>
6 years agoRemove duplicate import in compat.py
Yong Tang [Tue, 17 Apr 2018 19:47:25 +0000 (19:47 +0000)]
Remove duplicate import in compat.py

Noticed there are a couple of places in compat.py that
have duplicate import:
```
from tensorflow.python.util.tf_export import tf_export
from tensorflow.python.util.tf_export import tf_export
```

This fix remove duplicate imports.

Signed-off-by: Yong Tang <yong.tang.github@outlook.com>
6 years agoMerge commit for internal changes
Yifei Feng [Tue, 17 Apr 2018 19:18:44 +0000 (12:18 -0700)]
Merge commit for internal changes

6 years agoComplement cmake script to compile tensorflow with mkl and mkldnn on Windows (#16936)
Tang, Wenyi [Tue, 17 Apr 2018 19:17:48 +0000 (03:17 +0800)]
Complement cmake script to compile tensorflow with mkl and mkldnn on Windows (#16936)

* Add build batch for windows

* Automaticaly find python, cuda, mkl runtimes in PATH

* auto select cmake generator

* Add external library mkldnn. Add options for mkl and mkldnn

* fix syntax error in make.bat

* Fix errorlevel syntex bug in make.bat

* Add /arch:avx2 flags to enable avx2 on windows

* Revert to keep `tensprflow_WIN_CPU_SIMD_OPTIONS` unchanged, add an option `tensorflow_ENABLE_MKL_SUPPORT` to include MKL compilation. Still specify SIMD flags by setting `tensorflow_WIN_CPU_SIMD_OPTIONS` to such as '/arch:AVX2'

* Fix a mistake of CUDA path in make.bat

* resolve conflict in mkl_cpu_allocator.h

* Improve error detection

* Use where /Q to detect cmd environment

* fix "ELSE IF" the syntax error in make.bat

* update README.md, wrap windows based codes by #ifdef _WIN32

* unistd.h is not needed in mkl_cpu_allocator.h any more in master branch

* Remove inline of kMaxLimitStr, which causes compile error in VS2015

* Add static_cast in  to fix compile error

* remove make.bat

* Removed make.bat description part

6 years agoAutomated g4 rollback of changelist 193168327
Yifei Feng [Tue, 17 Apr 2018 19:07:33 +0000 (12:07 -0700)]
Automated g4 rollback of changelist 193168327

PiperOrigin-RevId: 193234819

6 years agoMoving gradient registration for CudnnRNN op from contrib to core.
Pavithra Vijay [Tue, 17 Apr 2018 19:06:50 +0000 (12:06 -0700)]
Moving gradient registration for CudnnRNN op from contrib to core.

PiperOrigin-RevId: 193234663

6 years agoFixes and cleanup to support more complex quantized models and adds PropagateFakeQuan...
A. Unique TensorFlower [Tue, 17 Apr 2018 18:53:29 +0000 (11:53 -0700)]
Fixes and cleanup to support more complex quantized models and adds PropagateFakeQuantNumBits.

PiperOrigin-RevId: 193232630

6 years agoMerge pull request #18601 from yongtang/18598-tf.compat.as_str
Martin Wicke [Tue, 17 Apr 2018 18:54:48 +0000 (11:54 -0700)]
Merge pull request #18601 from yongtang/18598-tf.compat.as_str

Fix tf.compat.as_str returns bytes issue in Python 3

6 years agoMerge pull request #18581 from ManHyuk/fix_typo
Martin Wicke [Tue, 17 Apr 2018 18:52:39 +0000 (11:52 -0700)]
Merge pull request #18581 from ManHyuk/fix_typo

Fix typo

6 years agoReplace decode_image with decode_jpeg to avoid ValueError in datasets programmers...
Mingxing Tan [Tue, 17 Apr 2018 18:48:43 +0000 (11:48 -0700)]
Replace decode_image with decode_jpeg to avoid ValueError in datasets programmers guide.

PiperOrigin-RevId: 193231717

6 years agoFixes a comment in tf.contrib.seq2seq.monotonic_attention().
A. Unique TensorFlower [Tue, 17 Apr 2018 18:05:57 +0000 (11:05 -0700)]
Fixes a comment in tf.contrib.seq2seq.monotonic_attention().

PiperOrigin-RevId: 193224285

6 years agoEnable consumption of GIT_TAG_OVERRIDE env var in release build script. (#18579)
Yifei Feng [Tue, 17 Apr 2018 17:53:07 +0000 (10:53 -0700)]
Enable consumption of GIT_TAG_OVERRIDE env var in release build script. (#18579)

Enable consumption of GIT_TAG_OVERRIDE env var in release build script.

6 years agoAvoid ToString() in Eager's TFE_Execute.
Akshay Modi [Tue, 17 Apr 2018 17:32:47 +0000 (10:32 -0700)]
Avoid ToString() in Eager's TFE_Execute.

Also use InlinedVector instead of std::vector for non-async path

Before:
Benchmark              Time(ns)        CPU(ns)     Iterations
-------------------------------------------------------------
BM_Execute/0               1895           1898         360200  Execute
BM_Execute/1               1193           1942         358322  ExecuteAsync
BM_ExecuteFunction/0       5812           5825         100000  ExecuteFunction
BM_ExecuteFunction/1       5015           5374         100000  ExecuteFunctionAsync

After:
Benchmark              Time(ns)        CPU(ns)     Iterations
-------------------------------------------------------------
BM_Execute/0               1604           1607         428262  Execute
BM_Execute/1               1150           1765         404821  ExecuteAsync
BM_ExecuteFunction/0       5615           5626         100000  ExecuteFunction
BM_ExecuteFunction/1       5111           5476         100000  ExecuteFunctionAsync
PiperOrigin-RevId: 193218331

6 years agoChange the contract of dynamic_builtin to reject all functions it can't process.
A. Unique TensorFlower [Tue, 17 Apr 2018 17:13:09 +0000 (10:13 -0700)]
Change the contract of dynamic_builtin to reject all functions it can't process.

PiperOrigin-RevId: 193215246

6 years agoPylint fix
Yong Tang [Tue, 17 Apr 2018 16:15:20 +0000 (16:15 +0000)]
Pylint fix

Signed-off-by: Yong Tang <yong.tang.github@outlook.com>
6 years agoEstimate IdentityN as Identity.
Max Galkin [Tue, 17 Apr 2018 16:19:06 +0000 (09:19 -0700)]
Estimate IdentityN as Identity.

PiperOrigin-RevId: 193207469

6 years agoRemoved unnneded tf_export
Yong Tang [Tue, 17 Apr 2018 16:14:50 +0000 (16:14 +0000)]
Removed unnneded tf_export

Signed-off-by: Yong Tang <yong.tang.github@outlook.com>
6 years agoFix tf.compat.as_str returns bytes issue in Python 3
Yong Tang [Tue, 17 Apr 2018 16:10:27 +0000 (16:10 +0000)]
Fix tf.compat.as_str returns bytes issue in Python 3

This fix tries to address the issue raised in 18598 where
tf.compat.as_str returns bytes (vs. str) in Python 3.
The issue was that `tf_export` decorator:
```
@tf_export('compat.as_bytes', 'compat.as_str')
```
could not be assigned to `as_bytes` or `as_text`
based on python 2 or 3.
This fix invokes tf_export explicitly based on `_six.PY2`
(for python 2/3) so that `as_str` calls `as_bytes` or `as_text`
conditionally.

This fix fixes 18598.

Signed-off-by: Yong Tang <yong.tang.github@outlook.com>
6 years agoFix incorrect rejection of xrange.
A. Unique TensorFlower [Tue, 17 Apr 2018 16:02:14 +0000 (09:02 -0700)]
Fix incorrect rejection of xrange.

PiperOrigin-RevId: 193205016

6 years agoAdd uint32/uint64 support to Gather op.
A. Unique TensorFlower [Tue, 17 Apr 2018 14:41:09 +0000 (07:41 -0700)]
Add uint32/uint64 support to Gather op.

PiperOrigin-RevId: 193195939

6 years agoFix the type info analysis to correctly process loops. Simplify the implementation...
A. Unique TensorFlower [Tue, 17 Apr 2018 14:07:16 +0000 (07:07 -0700)]
Fix the type info analysis to correctly process loops. Simplify the implementation by reusing some of the transformer base functions. Allow set_element_type to use literals. Add additional tests.

PiperOrigin-RevId: 193192409

6 years agoAutomated g4 rollback of changelist 192842670
A. Unique TensorFlower [Tue, 17 Apr 2018 09:23:10 +0000 (02:23 -0700)]
Automated g4 rollback of changelist 192842670

PiperOrigin-RevId: 193168327

6 years ago[tf.data] Add an API for randomly sampling from multiple datasets.
Derek Murray [Tue, 17 Apr 2018 06:07:24 +0000 (23:07 -0700)]
[tf.data] Add an API for randomly sampling from multiple datasets.

Fixes #15999.

PiperOrigin-RevId: 193152683

6 years agoMake ParallelInterleaveDataset saveable.
Saurabh Saxena [Tue, 17 Apr 2018 03:28:02 +0000 (20:28 -0700)]
Make ParallelInterleaveDataset saveable.

PiperOrigin-RevId: 193142302

6 years agoSupply a dtype to super constructor, without which build() seems to crash.
A. Unique TensorFlower [Tue, 17 Apr 2018 02:51:13 +0000 (19:51 -0700)]
Supply a dtype to super constructor, without which build() seems to crash.

PiperOrigin-RevId: 193139585

6 years agoBoostedTreesEstimator in contrib: train_in_memory works with input_fns returning...
Younghee Kwon [Tue, 17 Apr 2018 02:10:10 +0000 (19:10 -0700)]
BoostedTreesEstimator in contrib: train_in_memory works with input_fns returning data.Dataset.
Only one batch of data is expected, so dataset.batch() is disallowed,
and dataset.repeat() will be ignored (only the first one would be used)

PiperOrigin-RevId: 193137094

6 years agofix typo
manhyuk [Tue, 17 Apr 2018 01:52:16 +0000 (10:52 +0900)]
fix typo

6 years agofix typo
manhyuk [Tue, 17 Apr 2018 01:48:52 +0000 (10:48 +0900)]
fix typo

6 years agofix typo
manhyuk [Tue, 17 Apr 2018 01:46:37 +0000 (10:46 +0900)]
fix typo

6 years agoRemove proto imports in header files for core/kernels/hexagon.
Yifei Feng [Tue, 17 Apr 2018 01:41:28 +0000 (18:41 -0700)]
Remove proto imports in header files for core/kernels/hexagon.

The goal is to make kernels mostly independent of proto headers, which will let us lock down our .so imports.

PiperOrigin-RevId: 193134710

6 years agoAdd a simple Profiler and instrument operator invocations in Interpreter.
Shashi Shekhar [Tue, 17 Apr 2018 01:31:22 +0000 (18:31 -0700)]
Add a simple Profiler and instrument operator invocations in Interpreter.

PiperOrigin-RevId: 193133955

6 years agoSanitize with clan-format -i --style=Google
Yong Tang [Tue, 17 Apr 2018 01:32:41 +0000 (01:32 +0000)]
Sanitize with clan-format -i --style=Google

Signed-off-by: Yong Tang <yong.tang.github@outlook.com>
6 years agoAdd shape check for buffer_size with TFRecordDataset
Yong Tang [Tue, 17 Apr 2018 01:31:54 +0000 (01:31 +0000)]
Add shape check for buffer_size with TFRecordDataset

Signed-off-by: Yong Tang <yong.tang.github@outlook.com>
6 years agoAdd shape check for compression_type in TFrecordDataset
Yong Tang [Tue, 17 Apr 2018 01:30:42 +0000 (01:30 +0000)]
Add shape check for compression_type in TFrecordDataset

Signed-off-by: Yong Tang <yong.tang.github@outlook.com>
6 years agoAdd shape check to TFRecordDataset
Yong Tang [Tue, 17 Apr 2018 01:28:30 +0000 (01:28 +0000)]
Add shape check to TFRecordDataset

The inputs of TFRecordDataset have the requirements for shapes.
However, the check was not done in the shape function. This fix
adds shape checks whenever possible.

Signed-off-by: Yong Tang <yong.tang.github@outlook.com>
6 years agoAdding min node weight regularization
A. Unique TensorFlower [Tue, 17 Apr 2018 01:03:05 +0000 (18:03 -0700)]
Adding min node weight regularization

PiperOrigin-RevId: 193131300

6 years agoRemove deprecated/unused python related Bazel options.
Michael Case [Tue, 17 Apr 2018 00:34:01 +0000 (17:34 -0700)]
Remove deprecated/unused python related Bazel options.

Since py_runtime was introduced, Bazel ignores options such as
--force_python2 and --python2_path. Deleting to clean stuff up and
make sure people are not misled.

PiperOrigin-RevId: 193127681

6 years ago[tf.data] Sort the results of `tf.matching_files()` to enable `Dataset.list_files...
Derek Murray [Tue, 17 Apr 2018 00:25:12 +0000 (17:25 -0700)]
[tf.data] Sort the results of `tf.matching_files()` to enable `Dataset.list_files()` to be determinstic.

PiperOrigin-RevId: 193126572

6 years agoDon't rely on graph contruction for an initial shape inference.
Benoit Steiner [Tue, 17 Apr 2018 00:12:05 +0000 (17:12 -0700)]
Don't rely on graph contruction for an initial shape inference.

PiperOrigin-RevId: 193124836

6 years agoAdd LinearOperatorKronecker, representing the Kronecker product.
A. Unique TensorFlower [Tue, 17 Apr 2018 00:05:18 +0000 (17:05 -0700)]
Add LinearOperatorKronecker, representing the Kronecker product.

PiperOrigin-RevId: 193123894

6 years agoMerge pull request #18575 from miaout17/fix-fft
Martin Wicke [Tue, 17 Apr 2018 00:00:03 +0000 (17:00 -0700)]
Merge pull request #18575 from miaout17/fix-fft

Fix TFLite Makefile FFT2D dependency.

6 years agoInternal change.
Anna R [Mon, 16 Apr 2018 23:41:12 +0000 (16:41 -0700)]
Internal change.

PiperOrigin-RevId: 193119953

6 years agoPort the list append into the operators module. Not enabled yet.
A. Unique TensorFlower [Mon, 16 Apr 2018 23:34:18 +0000 (16:34 -0700)]
Port the list append into the operators module. Not enabled yet.

PiperOrigin-RevId: 193118940

6 years agoMerge pull request #17443 from yongtang/16263-softmax_cross_entropy-doc
Martin Wicke [Mon, 16 Apr 2018 23:30:50 +0000 (16:30 -0700)]
Merge pull request #17443 from yongtang/16263-softmax_cross_entropy-doc

Update the documentation of `softmax_cross_entropy`

6 years agodisabling flaky asan test
Olivia Nordquist [Mon, 16 Apr 2018 23:24:19 +0000 (16:24 -0700)]
disabling flaky asan test

PiperOrigin-RevId: 193117611

6 years agoInternal change
A. Unique TensorFlower [Mon, 16 Apr 2018 23:22:03 +0000 (16:22 -0700)]
Internal change

PiperOrigin-RevId: 193117142

6 years agoCreate copy of locals() before copying, since modifying locals does not always affect...
A. Unique TensorFlower [Mon, 16 Apr 2018 23:15:19 +0000 (16:15 -0700)]
Create copy of locals() before copying, since modifying locals does not always affect the values. https://docs.python.org/2/library/functions.html#locals.

PiperOrigin-RevId: 193116254

6 years agoReplacing the current inner Cholesky decomposition loop with a While loop rolled...
A. Unique TensorFlower [Mon, 16 Apr 2018 23:04:43 +0000 (16:04 -0700)]
Replacing the current inner Cholesky decomposition loop with a While loop rolled version.

This will allow for much larger Cholesky decompositions (and thus matrix inversions) than previously possible on TPU because of the use of rolled While loops so XLA compilation will no longer timeout.

While there is a minor runtime performance decrease (now 25ms vs 15ms for a 500x500 matrix) the compilation time is significantly faster (12.8s vs 55.2s for a 500x500 matrix.)

PiperOrigin-RevId: 193114816

6 years agoMerge branch 'master' into typo
Mark Daoust [Mon, 16 Apr 2018 22:56:41 +0000 (15:56 -0700)]
Merge branch 'master' into typo

6 years agoFix trace collection to properly remove the suffix.
A. Unique TensorFlower [Mon, 16 Apr 2018 22:52:54 +0000 (15:52 -0700)]
Fix trace collection to properly remove the suffix.

PiperOrigin-RevId: 193113074

6 years agoCurly-brace id's are inconsistently supported.
Mark Daoust [Mon, 16 Apr 2018 22:53:08 +0000 (15:53 -0700)]
Curly-brace id's are inconsistently supported.

Curly-brace id's are inconsistently supported.
linking to the id of an html tag seems to be supported everywhere.

6 years agoInternal change.
A. Unique TensorFlower [Mon, 16 Apr 2018 22:46:42 +0000 (15:46 -0700)]
Internal change.

PiperOrigin-RevId: 193112205

6 years agoFix TFLite Makefile FFT2D dependency.
Yu-Cheng Ling [Mon, 16 Apr 2018 22:34:50 +0000 (15:34 -0700)]
Fix TFLite Makefile FFT2D dependency.

FFT2D dependency was introduced a while ago so Makefile no longer works
until this fix.

6 years agoIncrease softmax gpu unittest numeric stability
James Qin [Mon, 16 Apr 2018 21:52:41 +0000 (14:52 -0700)]
Increase softmax gpu unittest numeric stability

PiperOrigin-RevId: 193103363

6 years agoPorting tests for `rpc_op` to OS.
Jiri Simsa [Mon, 16 Apr 2018 21:47:31 +0000 (14:47 -0700)]
Porting tests for `rpc_op` to OS.

PiperOrigin-RevId: 193102564

6 years agoExposes InputPipelineConfig as tf.contrib.tpu.InputPipelineConfig.
A. Unique TensorFlower [Mon, 16 Apr 2018 21:41:21 +0000 (14:41 -0700)]
Exposes InputPipelineConfig as tf.contrib.tpu.InputPipelineConfig.
This type is expected by the `per_host_input_for_training` argument of the TPUConfig constructor, but is not currently visible.

PiperOrigin-RevId: 193101540

6 years agoFix lite and makefile issue
Yong Tang [Mon, 16 Apr 2018 21:34:38 +0000 (21:34 +0000)]
Fix lite and makefile issue

Signed-off-by: Yong Tang <yong.tang.github@outlook.com>
6 years agoMerge pull request #18558 from imsheridan/fix_argmax_dimension
Martin Wicke [Mon, 16 Apr 2018 21:28:00 +0000 (14:28 -0700)]
Merge pull request #18558 from imsheridan/fix_argmax_dimension

Fix tf.argmax warnings on dimension argument by using axis instead

6 years agoMerge pull request #18547 from imsheridan/fix_deprecated_softmax
Martin Wicke [Mon, 16 Apr 2018 21:27:48 +0000 (14:27 -0700)]
Merge pull request #18547 from imsheridan/fix_deprecated_softmax

Fix the doc strings of nn.sampled_softmax_loss since it was deprecated

6 years agoMerge pull request #18570 from yongtang/18106-Bahdanau-normalized-dtype
Martin Wicke [Mon, 16 Apr 2018 21:25:16 +0000 (14:25 -0700)]
Merge pull request #18570 from yongtang/18106-Bahdanau-normalized-dtype

Fix the issue with Bahdanau attention when normalized=True and dtype = float16/32

6 years agodisabling test that fails tensorflow.asan
Olivia Nordquist [Mon, 16 Apr 2018 21:17:02 +0000 (14:17 -0700)]
disabling test that fails tensorflow.asan

PiperOrigin-RevId: 193097794

6 years agoUpdate bazel
Yong Tang [Mon, 16 Apr 2018 20:16:06 +0000 (20:16 +0000)]
Update bazel

Signed-off-by: Yong Tang <yong.tang.github@outlook.com>
6 years agoUse fixed sized tensor arrays and max loop iterations in dynamic_decode if the user...
A. Unique TensorFlower [Mon, 16 Apr 2018 21:13:52 +0000 (14:13 -0700)]
Use fixed sized tensor arrays and max loop iterations in dynamic_decode if the user supplies it and if the inputs were created in an XLA context.

PiperOrigin-RevId: 193097293

6 years agoSimplify the recursion when processing unpackings.
A. Unique TensorFlower [Mon, 16 Apr 2018 20:54:03 +0000 (13:54 -0700)]
Simplify the recursion when processing unpackings.

PiperOrigin-RevId: 193094078

6 years agoMerge pull request #18562 from yongtang/04152018-png
Martin Wicke [Mon, 16 Apr 2018 20:48:33 +0000 (13:48 -0700)]
Merge pull request #18562 from yongtang/04152018-png

Update libpng to v1.6.34 for cmake build

6 years ago[XLA] Document and enforce reduction order of init value
Nick Desaulniers [Mon, 16 Apr 2018 20:36:55 +0000 (13:36 -0700)]
[XLA] Document and enforce reduction order of init value

All existing backends apply the init_value on the lhs, except for the evaluator.  This causes problems for reductions which apply an identity function to a reduce or reduce window.

PiperOrigin-RevId: 193091323

6 years agoAdding several utility functions to TF2XLA to help with the Cholesky refactor. Mainl...
A. Unique TensorFlower [Mon, 16 Apr 2018 20:32:12 +0000 (13:32 -0700)]
Adding several utility functions to TF2XLA to help with the Cholesky refactor.  Mainly responsible for handling batching properly.

PiperOrigin-RevId: 193090634

6 years agoUpdate ops-related pbtxt files.
A. Unique TensorFlower [Mon, 16 Apr 2018 20:22:58 +0000 (13:22 -0700)]
Update ops-related pbtxt files.

PiperOrigin-RevId: 193089301

6 years agoAdd gemmlowp to whitelist
Yong Tang [Mon, 16 Apr 2018 20:02:46 +0000 (20:02 +0000)]
Add gemmlowp to whitelist

Signed-off-by: Yong Tang <yong.tang.github@outlook.com>
6 years agoTemporary comment out mirror.bazel.build for gemmlowp
Yong Tang [Mon, 16 Apr 2018 19:40:43 +0000 (19:40 +0000)]
Temporary comment out mirror.bazel.build for gemmlowp

Will reenable once the mirror is propagated.

Signed-off-by: Yong Tang <yong.tang.github@outlook.com>
6 years agoUpdate gemmlowp on bazel to sync cmake changes
Yong Tang [Mon, 16 Apr 2018 19:25:05 +0000 (19:25 +0000)]
Update gemmlowp on bazel to sync cmake changes

Signed-off-by: Yong Tang <yong.tang.github@outlook.com>
6 years agoUpdate gemmlowp version for cmake build
Yong Tang [Mon, 16 Apr 2018 16:27:50 +0000 (16:27 +0000)]
Update gemmlowp version for cmake build

The gemmlowp has been updated in bazel, though
cmake version was not updated. This fix updates
gemmlowp in cmake so that cmake and bazel versions
are synced.

The fix has been tested on Linux:
```
tensorflow/tools/ci_build/ci_build.sh CMAKE tensorflow/tools/ci_build/builds/cmake.sh
```

Signed-off-by: Yong Tang <yong.tang.github@outlook.com>
6 years agoSimplify the implementation of break_statements.py
A. Unique TensorFlower [Mon, 16 Apr 2018 20:04:23 +0000 (13:04 -0700)]
Simplify the implementation of break_statements.py

PiperOrigin-RevId: 193086371

6 years agoFix pylint issue
Yong Tang [Sun, 1 Apr 2018 02:08:53 +0000 (02:08 +0000)]
Fix pylint issue

Signed-off-by: Yong Tang <yong.tang.github@outlook.com>
6 years agoAdd test case for Bahdanau attention when normalized=True and dtype = float16/32
Yong Tang [Sun, 1 Apr 2018 02:07:30 +0000 (02:07 +0000)]
Add test case for Bahdanau attention when normalized=True and dtype = float16/32

Signed-off-by: Yong Tang <yong.tang.github@outlook.com>
6 years agoFix the issue with Bahdanau attention when normalized=True and dtype = float16/32
Yong Tang [Sun, 1 Apr 2018 02:01:47 +0000 (02:01 +0000)]
Fix the issue with Bahdanau attention when normalized=True and dtype = float16/32

While revisiting 18016 I noticed that Bahdanau attention has a similiar
dtype mismatch issue when normalized=True. The issue comes from:
```
     g = variable_scope.get_variable(
         "attention_g", dtype=dtype,
         initializer=math.sqrt((1. / num_units)))
```
where the initializer value does not work well with differnt dtype.

This fix converts changes the initializer to `init_ops.constant_initializer`
to address the issue, and adds additional test cases for it.

Signed-off-by: Yong Tang <yong.tang.github@outlook.com>
6 years agoboosted_trees: Make some regularizer/hyper-params as inputs instead of attributes.
Younghee Kwon [Mon, 16 Apr 2018 19:56:14 +0000 (12:56 -0700)]
boosted_trees: Make some regularizer/hyper-params as inputs instead of attributes.

PiperOrigin-RevId: 193085059

6 years agoAdd missing header pnglibconf.h
Yong Tang [Sat, 14 Apr 2018 19:05:28 +0000 (19:05 +0000)]
Add missing header pnglibconf.h

Signed-off-by: Yong Tang <yong.tang.github@outlook.com>
6 years agoUpdate library file names `libpng12` -> `libpng16`
Yong Tang [Sat, 14 Apr 2018 19:04:52 +0000 (19:04 +0000)]
Update library file names `libpng12` -> `libpng16`

Signed-off-by: Yong Tang <yong.tang.github@outlook.com>
6 years agoUpdate libpng to v1.6.34 for cmake
Yong Tang [Sat, 14 Apr 2018 19:00:36 +0000 (19:00 +0000)]
Update libpng to v1.6.34 for cmake

The libpng has been updated from v1.2.53 to v1.6.34 in PR 18299.
However, the cmake version of libpng has not been updated yet.
This fix updates the libpng for cmake to v1.6.34.

The fix is tested with cmake on linux:
```
tensorflow/tools/ci_build/ci_build.sh CMAKE tensorflow/tools/ci_build/builds/cmake.sh
```

Signed-off-by: Yong Tang <yong.tang.github@outlook.com>
6 years agoEarly TPU distribution strategy and the associated testing infrastructure.
Igor Saprykin [Mon, 16 Apr 2018 19:21:15 +0000 (12:21 -0700)]
Early TPU distribution strategy and the associated testing infrastructure.

PiperOrigin-RevId: 193080098