OSSForks/tensorflow
1
0
Fork 0
mirror of https://github.com/zebrajr/tensorflow.git synced 2026-01-15 12:15:41 +00:00
Code Issues Actions 5 Packages Projects Releases Wiki Activity
23,861 Commits 50 Branches 144 Tags
f70dfea26f6ecc9faccea4800ca28ddfc129e79f
Commit Graph

23861 Commits

This Branch
This Branch
All Branches
Author SHA1 Message Date
A. Unique TensorFlower
f70dfea26f Checking that the NCCL 2 license file is found, see GitHub issue 19679. 
PiperOrigin-RevId: 202613754
 2018-06-29 01:48:36 -07:00
A. Unique TensorFlower
6874e1ef40 Convert exp(x-1) into expm1(x). 
PiperOrigin-RevId: 202598404
 2018-06-28 22:13:59 -07:00
Peter Hawkins
5083915489 [TF:XLA] Add partial implementation of tf.FIFOQueue for XLA devices (e.g., TPU). 
The idea is to have a host-side queue of device tensors.

Operators dequeue_many, enqueue_many, and dequeue_up_to are not yet implemented because they require splitting/concatenating tensors, which will require calling into a compiled XLA compilation.

Refactor queue operator implementations into libraries separate from the kernel registrations.

Add support for ResourceOpKernels that are placed on non-CPU devices. Add support for allocating host-memory tensors during OpKernel construction.

PiperOrigin-RevId: 202590292
 2018-06-28 20:26:49 -07:00
Sanjoy Das
f04400f18f Make the HLO evaluator correctly handle the lhs contracting dim for Dots 
This CL fixes the bug by rewriting how we map the result index to the lhs/rhs
index in the dot evaluator.

PiperOrigin-RevId: 202588171
 2018-06-28 20:02:04 -07:00
A. Unique TensorFlower
6f5feedd57 Go: Update generated wrapper functions for TensorFlow ops. 
PiperOrigin-RevId: 202587352
 2018-06-28 19:49:00 -07:00
Mingxing Tan
1e7b0e4ad6 Merge changes from github. 
PiperOrigin-RevId: 202585094
 2018-06-28 19:16:41 -07:00
A. Unique TensorFlower
3cee10e61c Remove some obsolete statements from variable_ops.cc. 
PiperOrigin-RevId: 202581793
 2018-06-28 18:37:55 -07:00
Nupur Garg
78d7a7783a Updates TOCO documentation. 
PiperOrigin-RevId: 202577530
 2018-06-28 17:58:51 -07:00
A. Unique TensorFlower
9ab9d5e034 In constant_op.cc get the CPU RAM allocator through the OpKernelContext 
instead of calling cpu_allocator() directly.  This will ensure that
NUMA node specific constant ops get node-local memory.

PiperOrigin-RevId: 202576292
 2018-06-28 17:47:28 -07:00
Anjali Sridhar
4740e1b6f1 tf.keras sync 2.2.0 
PiperOrigin-RevId: 202575679
 2018-06-28 17:42:16 -07:00
Rachel Lim
81ee76a071 Registered kernel for tensor array gather for type tf.variant 
PiperOrigin-RevId: 202574948
 2018-06-28 17:37:38 -07:00
A. Unique TensorFlower
3e15ac3dd2 [TF:XLA] Copy elision does not need to know about existing copies. 
It already detects layout-changing copies and those are already left unchanged
by copy elision. Special case copies are also skipped because they are tagged
separately (SetCopyElisionAllowed)

PiperOrigin-RevId: 202574858
 2018-06-28 17:34:58 -07:00
A. Unique TensorFlower
6dc9977e1d Import package xla_test instead of class XLATestCase. 
PiperOrigin-RevId: 202572322
 2018-06-28 17:13:56 -07:00
A. Unique TensorFlower
c81830af5d Don't cache RNN weights in while loops since it's possible that train steps could be updating the weights. This is specifically true on TPUS (tpu.repeat). Also, fix the testDynamicRnnTrainLoop unit test. 
PiperOrigin-RevId: 202565323
 2018-06-28 16:30:55 -07:00
A. Unique TensorFlower
2273cda3e0 Change inputs from multi-dimensional arrays to ByteBuffer in TF Lite Object Detection app 
PiperOrigin-RevId: 202564164
 2018-06-28 16:23:06 -07:00
A. Unique TensorFlower
0ea6847c89 Automated g4 rollback of changelist 202292422 
PiperOrigin-RevId: 202551122
 2018-06-28 15:10:14 -07:00
Mark Daoust
c061686464 Fix Missing r-string. 
The docs generator is not happy about the "\a" in "\approx" is becoming a "alert" escape sequence.

This should also fix a lot of the mathjax rendering on this page:

https://www.tensorflow.org/api_docs/python/tf/contrib/bayesflow/monte_carlo/expectation

PiperOrigin-RevId: 202550662
 2018-06-28 15:07:02 -07:00
Brian Patton
ffc7f5f657 Adds support for Type III DCT, and tf.spectral.idct(type=2|3). 
PiperOrigin-RevId: 202546469
 2018-06-28 14:44:12 -07:00
Noah Eisen
19a98bf905 Rewrite master_service in terms of more up-to-date gRPC APIs 
PiperOrigin-RevId: 202544091
 2018-06-28 14:31:38 -07:00
Shanqing Cai
901d82d736 tfdbg: Fix compatibility with C++ MakeCallable and _make_callable_from_options 
Fixes #20160

REL_NOTES: tfdbg: Fix compatibility with `tf.keras.Model`s training on `tf.data.Dataset`s.
PiperOrigin-RevId: 202543231
 2018-06-28 14:26:57 -07:00
Yanan Cao
1a4820e1c5 TPUEstimator adds use_tpu key in params when calling model_fn. Model_fn can rely on this key to know whether TPU specific functionalities should be used. 
PiperOrigin-RevId: 202542458
 2018-06-28 14:22:11 -07:00
Tony Wang
15b258dc93 Automated g4 rollback of changelist 201419522 
PiperOrigin-RevId: 202539762
 2018-06-28 14:08:54 -07:00
Jacques Pienaar
f5606a442a Dedupe incompatible ops message. 
No need to repeat an incompatible op time multiple times. Used set to ensure deterministic/same ordering in error message.

PiperOrigin-RevId: 202534388
 2018-06-28 13:37:45 -07:00
A. Unique TensorFlower
f57d80c397 Adds reset op as an optional return from update_stats. 
PiperOrigin-RevId: 202528760
 2018-06-28 12:59:54 -07:00
Peter Hawkins
340e27128d [XLA] Change code in tensorflow/compiler/xla that uses XlaBuilder:: methods to build ops to use the corresponding free functions in namespace xla:: instead. 
PiperOrigin-RevId: 202526945
 2018-06-28 12:48:01 -07:00
Priya Gupta
ca6ad21988 Add an output context that can be used to specify outputs to capture when running multiple steps at a time using the run_steps_on_dataset API. It allows the user's step function to specify which outputs to emit at what frequency. Currently it only supports capturing output from the last step, but will soon be augmented to support other use cases such as output each N steps. 
PiperOrigin-RevId: 202520245
 2018-06-28 12:07:30 -07:00
Anjali Sridhar
412292d042 tf.keras sync 2.2.0 
PiperOrigin-RevId: 202520102
 2018-06-28 12:04:00 -07:00
Brian Patton
5f433d10f5 Bugfix: In eager mode, bessel_i0 and bessel_i1 raise "TypeError: unhashable type: 'list'" out of ops.name_scope because the second arg of name_scope is default_name, not values. 
Standardizes special_math_ops on the full name_scope(name, default_name, values) constructor.

PiperOrigin-RevId: 202519452
 2018-06-28 12:01:05 -07:00
James Keeling
23ba5ad94b Add code to parse interpolatable error messages. 
The interpolate function currently just reformats tags, without adding any useful information. This change is part of a chain which will add this.

PiperOrigin-RevId: 202519204
 2018-06-28 11:58:16 -07:00
A. Unique TensorFlower
7efd0d7460 Replace Keras clip by value and clip by norm in Keras Optimizers with native TF clip_ops, also added user input check for clipnorm and clipvalue >= 0 if set 
PiperOrigin-RevId: 202516320
 2018-06-28 11:47:46 -07:00
Eugene Zhulenev
42444a5b4d _UnaryOpsComposition kernel: compose multiple shape&type preserving unary ops at runtime. 
PiperOrigin-RevId: 202514848
 2018-06-28 11:35:02 -07:00
Akshay Modi
be0ab68435 Include eager/graph mode in cache key so that one type of tensor doesn't spill 
into the other.

PiperOrigin-RevId: 202513508
 2018-06-28 11:27:49 -07:00
James Keeling
8c81fc0a7c Add helper for creating error tags 
error_format_tag is a helper for building interpolatable strings as part of
a project to improve Python error messages in TensorFlow.

PiperOrigin-RevId: 202509392
 2018-06-28 11:10:41 -07:00
A. Unique TensorFlower
f92b6b22ed Avoid overflow in flops calculations in nn_ops.py by forcing 
np.prod() to use np.int64 in a few places.

PiperOrigin-RevId: 202505308
 2018-06-28 10:56:53 -07:00
Peter Hawkins
19781808c8 [XLA] Change code in TF/XLA bridge that uses XlaBuilder:: methods to build ops to use the corresponding free functions in namespace xla:: instead. 
PiperOrigin-RevId: 202505306
 2018-06-28 10:54:11 -07:00
Anjali Sridhar
24088d19bb tf.keras sync 2.2.0 
PiperOrigin-RevId: 202505228
 2018-06-28 10:50:58 -07:00
Michael Kuperstein
ad320521b3 [XLA] VerifyShape() should verify that max_sparse_elements is non-negative. 
PiperOrigin-RevId: 202504925
 2018-06-28 10:48:10 -07:00
A. Unique TensorFlower
fcc6d259e4 Add --input_examples option to run syntax and update --input_examples and --input_exprs headings to match option names. 
PiperOrigin-RevId: 202504009
 2018-06-28 10:43:07 -07:00
A. Unique TensorFlower
6cbd3ade04 Jump to version 1.9 to sync with TensorFlow versions. 
Also enable cloud tpu profiler to detect the TF version for better version compatibility.

PiperOrigin-RevId: 202503162
 2018-06-28 10:38:28 -07:00
A. Unique TensorFlower
c138b3871d make identify_lstm independent of rnn_states 
PiperOrigin-RevId: 202501055
 2018-06-28 10:26:46 -07:00
A. Unique TensorFlower
6e708d5ffb More un-fused quantized LSTM support in TFLite interpreter 
PiperOrigin-RevId: 202496488
 2018-06-28 10:00:41 -07:00
Tom Hennigan
bac87bbe7c Run setUp in the same mode each time. 
PiperOrigin-RevId: 202489637
 2018-06-28 09:12:44 -07:00
A. Unique TensorFlower
6359f90c92 Use a named tag for the nsync version, instead of a git hash. 
Only the name of the version is changing here; the version is unchanged.

PiperOrigin-RevId: 202486284
 2018-06-28 08:48:08 -07:00
A. Unique TensorFlower
7f081754c4 Support more quantized unfused LSTMs in TFLite interpreter. 
PiperOrigin-RevId: 202472329
 2018-06-28 07:04:26 -07:00
A. Unique TensorFlower
85d3356b4a [tf2xla] Return zero-element tensors as tokens from FusedBatchNormGrad 
We returned one-element tensors with uninitialized content, which msan didn't like.

PiperOrigin-RevId: 202463090
 2018-06-28 05:27:30 -07:00
A. Unique TensorFlower
ddc01636d0 Improve the performance of ParseShapeStringInternal 
The previous implementation recompiled the shape regex at every call
what is an expensive opertaion. The new implementation improves the hlo
text parsing time for very large models for up to 9x by eliminating this
overhead.

PiperOrigin-RevId: 202454354
 2018-06-28 03:59:03 -07:00
A. Unique TensorFlower
06aac64532 Fixed ShardingMetadata dump of null sharding from None to {}, to make it 
compatible with hlo string syntax.

PiperOrigin-RevId: 202445509
 2018-06-28 02:35:00 -07:00
A. Unique TensorFlower
99dc8c88c4 [XLA] Handle domain instructions in dataflow analysis. 
Without domain propagation in dataflow analysis we end up in inconsistent domain instructions with BF16 as output and F32 as input. In case of tuple shapes these are not fixed by bfloat16_normalization, and later on they cause asserts once the domain instructions are removed.

PiperOrigin-RevId: 202442786
 2018-06-28 02:14:14 -07:00
Gunhan Gulsoy
ea5fa07713 Exclude test sources from stream executor builds. 
PiperOrigin-RevId: 202423156
 2018-06-27 22:28:10 -07:00
A. Unique TensorFlower
6c92ef52fd Create a constant from the feature column group index. 
PiperOrigin-RevId: 202419595
 2018-06-27 21:38:06 -07:00