experiments/STT-tensorflow
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity
75,545 Commits 49 Branches 133 Tags 692 MiB
87cab44823
Commit Graph

75545 Commits

Author SHA1 Message Date
TensorFlower Gardener
87cab44823 Merge pull request from lgeiger:cleanup-fwd-compat-tests 
PiperOrigin-RevId: 288339675
Change-Id: Id0b0bc64655cb52ab5dd5252528ec978dc1b9675
 2020-01-06 11:38:33 -08:00
Mihai Maruseac
a979677021 Make modular filesystems test run on windows. 
This is prerequisite to having windows filesystem be a module.

PiperOrigin-RevId: 288339089
Change-Id: Ia4794639aeb5f1c21c8e5c9cd7d55049b32ed5ed
 2020-01-06 11:34:39 -08:00
A. Unique TensorFlower
2561c27de3 Fixed typo. 
Removed outdated comment.

PiperOrigin-RevId: 288338957
Change-Id: I3f7d7d7cc4401643e143e3ea01a7513569062329
 2020-01-06 11:31:34 -08:00
Yunlu Li
d6e63d2151 Add check for the content of traversal order. 
PiperOrigin-RevId: 288337584
Change-Id: I28c3cb3ffda59ecefe5bd92766e6b8d74c773d69
 2020-01-06 11:24:48 -08:00
Dan Moldovan
a4b4a8d251 Allow undefined symbols to be queried for attributes and slices. An upcoming change will create situations when this can happen before the verification for undefined symbols that currently precludes such operations. 
PiperOrigin-RevId: 288336945
Change-Id: I5bb1e02a9b3dacb557749ada4aaffe402c17cd82
 2020-01-06 11:21:58 -08:00
TensorFlower Gardener
b3d978711f Merge pull request from pooyadavoodi:rename_segment_graph 
PiperOrigin-RevId: 288334768
Change-Id: Ib341d902402b0bdf30054b861ba08f2e6cecf201
 2020-01-06 11:18:30 -08:00
George Karpenkov
93fa0c9c7a [XLA] Fixup the experimental_compile notebook syntax 
PiperOrigin-RevId: 288334343
Change-Id: I1f594f20230be9e20f4a2275362d5c8709c113c2
 2020-01-06 11:12:02 -08:00
TensorFlower Gardener
2d10fe3cd8 Merge pull request from suphoff:issue_35612 
PiperOrigin-RevId: 288333962
Change-Id: Ibd8866e450827a283041efaa89ed6fe89db4479f
 2020-01-06 11:08:41 -08:00
Reed Wanderman-Milne
e19c92334c Remove useless hasattr check in Layer.get_config. 
The hasattr(self, 'dtype') check would only fail if super().__init__ is not called. If not called, Layer.get_config and most other methods already raise errors anyway.

PiperOrigin-RevId: 288333917
Change-Id: I1962be09c9c4686c76120298411f3156b561f734
 2020-01-06 11:06:08 -08:00
Adam Wood
7f304cc3ef Update docstring for zeros_like: 
- testable examples
- clarify that input can be an array-like

PiperOrigin-RevId: 288333496
Change-Id: I7b6bf1301a314dcb4fb505b386b82219236b43dd
 2020-01-06 11:02:51 -08:00
TensorFlower Gardener
9982d476dd Merge pull request from duncanriach:review-python-framework-config 
PiperOrigin-RevId: 288332947
Change-Id: Ide7a1a6ef76721b62c7e9415c004b0199992f531
 2020-01-06 10:59:42 -08:00
TensorFlower Gardener
8016fed342 Merge pull request from charmasaur:where_v2_in_expm 
PiperOrigin-RevId: 288328469
Change-Id: If5507bbdc909d9cb5da873475520ded8cea2ea4d
 2020-01-06 10:48:21 -08:00
Derek Murray
cf7a3570fd [Tensor] Cache the result of LogMemory::IsEnabled() in tensor.cc. 
At present, each Tensor construction and destruction call involves a non-inlined call to `LogMemory::IsEnabled()`, which tests the logging verbosity level. This change moves the result of that check to a static method in the same compilation unit, so that the check can be inlined.

PiperOrigin-RevId: 288327926
Change-Id: I3f524a4b3bcdaa4756cc3aa2189291eea1bd8a0c
 2020-01-06 10:43:41 -08:00
TensorFlower Gardener
24d35cca11 Merge pull request from rrkarim:master 
PiperOrigin-RevId: 288327620
Change-Id: I860e94e98a845f939036c75af574a4c681236756
 2020-01-06 10:38:08 -08:00
A. Unique TensorFlower
8033e41d7c Fix TPU embedding implementation initialization of proximal yogi optimizer. 
Similar to Adagrad, setting initialization to 0.1 (instead of 0.0) will avoid numerical instability issues.

PiperOrigin-RevId: 288327127
Change-Id: I66a89bbb4e18fe653479cb64003b7c42daba7158
 2020-01-06 10:34:14 -08:00
Chao Mei
28018f2cd7 Properly indent the constraints of supported ops. 
PiperOrigin-RevId: 288326592
Change-Id: Id905b2ede5d0f9916fe8c460e74e603cadcdc73f
 2020-01-06 10:31:14 -08:00
Robert David
e5a693aeec Move vectors while setting various tensors. 
PiperOrigin-RevId: 288324119
Change-Id: Iddafc5147ef093a4506a399ba87655e673cadb93
 2020-01-06 10:19:05 -08:00
Allen Lavoie
f98b3bc701 Add an internal GPU-compatible version of FIFOQueue. 
Just registers a FIFOQueueV2 GPU kernel with the queue resource in host memory, then makes sure the public FIFOQueue requests a CPU placement.

Internal for now because adding EnqueueMany and DequeueMany GPU kernels is somewhat complicated, and without those placing the FIFOQueue resource on a GPU device is a nasty surprise. Eventually we may want those kernels.

PiperOrigin-RevId: 288322387
Change-Id: I294b1c878b47d2374468ec8f86ce0027b4f09112
 2020-01-06 10:10:52 -08:00
Derek Murray
feffae326d [tf.data] Add missing std::vector::reserve() calls. 
PiperOrigin-RevId: 288319987
Change-Id: I07bbec3f0bf00223505226285a5ab0c20dac5569
 2020-01-06 09:59:48 -08:00
TensorFlower Gardener
54381429c2 Merge pull request from pshiko:feature/pshiko-infer-outdir 
PiperOrigin-RevId: 288318661
Change-Id: Ibb88871e53122dda8a9a4b889bdab02fd21e4689
 2020-01-06 09:53:48 -08:00
Mark Daoust
b7e9c78414 Add default behavior for padded_shapes. 
With this change default behavior is: any unknown-dimension in a component's shape is padded to the maximum of the batch.

An error is raised if any component has an unknown rank.

No backwards compatibility concerns: this just allows you to call `.padded_batch` without passing `padded_shapes`

PiperOrigin-RevId: 288315115
Change-Id: I3e7e7c5cd5373a34287325ef4a01d0848167c21f
 2020-01-06 09:39:17 -08:00
Derek Murray
09e776fd22 [ParseExample] Optimize Config construction in ParseExampleOp. 
Each time the ParseExampleOp executes, a FastParseExampleConfig struct is constructed to define the features that will be extracted. This change makes three small optimizations to the code that constructs this struct.

1. Use StringPiece rather than std::string as the element type in the temporary vectors that translate between the input tensors and the config struct. This reduces the number of dynamic allocations and string copies.

2. Reserve capacity for the `FastParseExampleConfig::{dense,sparse,ragged}` vectors inside FastParseExampleConfig before populating them. This avoids copying the vectors (and the strings inside them) as the vector grows.

3. Emplace the `FastParseExampleConfig::{Dense,Sparse,Ragged}` structs in their respective vectors. This avoids temporary copies as the vector elements are constructed.

4. Convert the `FastParseExampleConfig::{Dense,Sparse,Ragged}::feature_name` fields to tstring. This has no functional change at present, but will allow us to avoid copying the string contents when USE_TSTRING is enabled.

PiperOrigin-RevId: 288314963
Change-Id: I6a413d81ec4d6d3ac43606bbebfb707e1400a629
 2020-01-06 09:36:27 -08:00
A. Unique TensorFlower
3183324e90 Fix typo in tf.print documentation. 
PiperOrigin-RevId: 288314809
Change-Id: I22410f3a909dc915e2f36204b9a8ee3dd4390d71
 2020-01-06 09:33:08 -08:00
Stephan Uphoff
e1d7c36fbe Fix lite/micro cmsis-nn kernels that were not refactored in 2020-01-06 09:59:28 -07:00
Jian Li
2274eacd79 Make method description more clear by replacing "pot" with "power of two". 
PiperOrigin-RevId: 288299716
Change-Id: I280ea3eecadd9ba12820674061b7849a33c2d54e
 2020-01-06 07:51:55 -08:00
A. Unique TensorFlower
ba75d7bb7a Add NNAPI Vendor Extension support in TFLite. 
PiperOrigin-RevId: 288298764
Change-Id: I80a735b9912ea70521a118ca51885b7fbe86f53f
 2020-01-06 07:43:59 -08:00
Dero Gharibian
60bfd36b00 Enable tensorflow::tstring stand-in class. 
After this CL, tensorflow::tstring will no longer be aliased to std::string.

PiperOrigin-RevId: 288296928
Change-Id: I32a5fa6a8249e22ce226eadb765480ba2943836a
 2020-01-06 07:30:11 -08:00
TensorFlower Gardener
dc3df01de5 Merge pull request from Intel-tensorflow:feature/wenxizhu/cache-primitive-key 
PiperOrigin-RevId: 288291598
Change-Id: Ie0e4bdfc0950ea07500b814e2c97afba6f238421
 2020-01-06 06:47:18 -08:00
A. Unique TensorFlower
8d40fa5616 Go: Update generated wrapper functions for TensorFlow ops. 
PiperOrigin-RevId: 288279859
Change-Id: Ifed8ff36aa55f6eefc40d83f9d2698af6e34cd2e
 2020-01-06 04:49:15 -08:00
Taehee Jeong
53843e51a7 Export the SerializeAsHexString functions from C++ to Python with pybind11 instead of swig. This is part of a larger effort to deprecate swig and eventually with modularization break pywrap_tensorflow into smaller components. It will also make exporting C++ ops to Python significantly easier. Please refer to https://github.com/tensorflow/community/blob/master/rfcs/20190208-pybind11.md for more information. 
PiperOrigin-RevId: 288278210
Change-Id: Ie6e01d42e7b6155f88f2e4cd54953a785c58ac99
 2020-01-06 04:27:59 -08:00
Taehee Jeong
3b05ac1fc2 Add -fno-strict-aliasing flag to pybind copts 
PiperOrigin-RevId: 288277300
Change-Id: I297c2cd23cb4c4f62a272a21848d6da55f6eac25
 2020-01-06 04:18:13 -08:00
A. Unique TensorFlower
7d9beeb699 Go: Update generated wrapper functions for TensorFlow ops. 
PiperOrigin-RevId: 288268171
Change-Id: I22ae80dcb06d2462b5743a34a62d7fbdc6cc0a79
 2020-01-06 02:48:56 -08:00
ShengYang1
d7a5027e3e fix small issues 2020-01-06 17:48:00 +08:00
A. Unique TensorFlower
9003f8b48d compat: Update forward compatibility horizon to 2020-01-06 
PiperOrigin-RevId: 288257159
Change-Id: I0de9f8a98cdf538d6652f81329ca73df203b10c7
 2020-01-06 01:05:12 -08:00
Renjie Liu
5c5a144afc Safe dynamic reshape option (only copy when the new_shape is not a nullptr) 
PiperOrigin-RevId: 288254073
Change-Id: I68725492420f9c3a344b32f1c21d4762b9b47382
 2020-01-06 00:34:28 -08:00
A. Unique TensorFlower
dec91cdf18 Go: Update generated wrapper functions for TensorFlow ops. 
PiperOrigin-RevId: 288246153
Change-Id: I9ebc89938a1e1ce69360bbd49c9eb624b7b625e6
 2020-01-05 22:49:13 -08:00
Tian Lin
afbfc6a092 Clean up for smart reply demo. As folder tensorflow/lite/models/ becomes empty, delete parent folder meanwhile. 
The whole project has been moved to: https://github.com/tensorflow/examples/tree/master/lite/examples/smart_reply/android

(Please see smart reply web page for details: https://www.tensorflow.org/lite/models/smart_reply/overview)

PiperOrigin-RevId: 288242509
Change-Id: Iabce717849f91cd01242634a090c0d936412e834
 2020-01-05 21:56:15 -08:00
A. Unique TensorFlower
81fcd56fb3 Go: Update generated wrapper functions for TensorFlow ops. 
PiperOrigin-RevId: 288237829
Change-Id: I2d9e3c6939d6b929bb8eab10a5ff9de32e1a683b
 2020-01-05 20:49:08 -08:00
A. Unique TensorFlower
a592eff24b Go: Update generated wrapper functions for TensorFlow ops. 
PiperOrigin-RevId: 288230068
Change-Id: I1680c8a0778cc6ba98b0674d5c24920f5ad65784
 2020-01-05 18:48:19 -08:00
Tiezhen WANG
6748a878de TFL: Add check for activation type in fully connected. 
PiperOrigin-RevId: 288222377
Change-Id: I429200022c7279e8c11ff14d7b50f3bf439f6f7e
 2020-01-05 16:26:22 -08:00
A. Unique TensorFlower
7d281edea6 Go: Update generated wrapper functions for TensorFlow ops. 
PiperOrigin-RevId: 288216994
Change-Id: I015fa87d36094042b333224f0931bf5ce4225038
 2020-01-05 14:49:18 -08:00
Anjali Sridhar
4fa8a873ac Add CompositeTensor support for DistributedIterator. 
PiperOrigin-RevId: 288207306
Change-Id: I1e57c831bd18cd7b6e6b770de7696aba98651e91
 2020-01-05 11:55:53 -08:00
Rasul Karimov
fc8a94750e
Merge branch 'master' into master 2020-01-05 22:25:23 +03:00
A. Unique TensorFlower
2e34ac25e0 Go: Update generated wrapper functions for TensorFlow ops. 
PiperOrigin-RevId: 288198247
Change-Id: Ieae140e7ff0423b942058352534189f3db08d57d
 2020-01-05 08:48:13 -08:00
A. Unique TensorFlower
2929431dab Go: Update generated wrapper functions for TensorFlow ops. 
PiperOrigin-RevId: 288185919
Change-Id: I8baf54d9cff7564e771aa6c540ca95f03f879b51
 2020-01-05 04:48:56 -08:00
A. Unique TensorFlower
d0414a39f9 compat: Update forward compatibility horizon to 2020-01-05 
PiperOrigin-RevId: 288173438
Change-Id: I44a422cfb3164911d758249a4a1e98be1c6d63df
 2020-01-05 01:05:10 -08:00
A. Unique TensorFlower
03f121434c Go: Update generated wrapper functions for TensorFlow ops. 
PiperOrigin-RevId: 288166159
Change-Id: If9095d47a25713eed7307289cd88e598236a4db1
 2020-01-04 22:49:54 -08:00
A. Unique TensorFlower
c81024dbbd Go: Update generated wrapper functions for TensorFlow ops. 
PiperOrigin-RevId: 288160267
Change-Id: I06a0765ee98da96bf8f418e1bc20212fc410a33b
 2020-01-04 20:49:05 -08:00
A. Unique TensorFlower
45f6f1ac3f Go: Update generated wrapper functions for TensorFlow ops. 
PiperOrigin-RevId: 288154151
Change-Id: I5806fdb1fc1d47181766a205a1dd181ce710b7a5
 2020-01-04 18:48:39 -08:00
A. Unique TensorFlower
a2cf331a20 Go: Update generated wrapper functions for TensorFlow ops. 
PiperOrigin-RevId: 288136894
Change-Id: I20957c77b9b56c372c19b91708df9e7fbcf254ae
 2020-01-04 12:51:24 -08:00