Commit Graph

28204 Commits

Author SHA1 Message Date
Mihai Maruseac
ca413aa0d8 Revert to old 2021-01-05 10:47:24 -08:00
Mihai Maruseac
669cb927a8 Try with just one test 2021-01-05 08:35:14 -08:00
Mihai Maruseac
40198cf2cd Start working on the C++ code 2021-01-04 10:56:05 -08:00
Mihai Maruseac
89a236e436 Try again 2021-01-03 17:46:24 -08:00
Mihai Maruseac
16b995513d Fill in tree node 2021-01-03 16:35:43 -08:00
Mihai Maruseac
dd5f07f812 Revert to full proto 2020-12-30 18:47:30 -08:00
Mihai Maruseac
251e710762 Add back most of the proto 2020-12-30 17:32:21 -08:00
Mihai Maruseac
de3de808b6 Try altering the proto 2020-12-30 16:26:20 -08:00
Mihai Maruseac
98ef15b713 Try to repro with a single test 2020-12-30 15:10:11 -08:00
Mihai Maruseac
ebda2124b5 One more step 2020-12-30 13:57:43 -08:00
Mihai Maruseac
06879f0634 Comment a few lines 2020-12-30 12:46:26 -08:00
Mihai Maruseac
5bcf84e9f1 Split and add 2020-12-23 19:43:19 -08:00
Mihai Maruseac
81a4d88e33 More cut 2020-12-23 18:36:17 -08:00
Mihai Maruseac
9679a640eb No positional arguments allowed 2020-12-23 17:21:33 -08:00
Mihai Maruseac
68e467fcd4 "Name the tree" 2020-12-23 15:25:07 -08:00
Mihai Maruseac
06a960b217 More cut 2020-12-23 14:14:11 -08:00
Mihai Maruseac
3996f747b8 Add more 2020-12-23 12:25:13 -08:00
Mihai Maruseac
86ff593035 Add back a chunk 2020-12-23 11:09:13 -08:00
Mihai Maruseac
a841427991 More removal 2020-12-23 09:52:11 -08:00
Mihai Maruseac
860d5c79d2 Try to simplify MRE 2020-12-23 08:48:11 -08:00
Mihai Maruseac
2b30f6525b Print on the other tests too 2020-12-22 19:00:13 -08:00
Mihai Maruseac
9bcec1c979 Try only one test 2020-12-22 17:52:17 -08:00
Mihai Maruseac
289a0ba853 More debug print 2020-12-22 16:31:57 -08:00
Mihai Maruseac
8be2cc5149 Print directly 2020-12-22 15:23:21 -08:00
Mihai Maruseac
e0cbc8b3e5 Debug print 2020-12-22 10:23:18 -08:00
A. Unique TensorFlower
e5c0560e46 compat: Update forward compatibility horizon to 2020-12-22
PiperOrigin-RevId: 348594729
Change-Id: Ieac580b986bc6cf37921fd0bfde06d37ff208908
2020-12-22 01:07:37 -08:00
A. Unique TensorFlower
94155a3934 Integrate LLVM at llvm/llvm-project@1b97cdf885
Updates LLVM usage to match
[1b97cdf885d6](https://github.com/llvm/llvm-project/commit/1b97cdf885d6)

PiperOrigin-RevId: 348587513
Change-Id: I853d197b33c5df08c00c99ddc8cf8b2681bed708
2020-12-21 23:55:25 -08:00
TensorFlower Gardener
4967bdb59e Merge pull request from jpodivin:reportarg
PiperOrigin-RevId: 348573959
Change-Id: I20cee2bea44676eb9cf6a17a7a1dab7c8c42889f
2020-12-21 21:07:25 -08:00
Scott Zhu
dde0d4f92c Fork the tf initializer logic to keras initializer.
PiperOrigin-RevId: 348571002
Change-Id: I753afd8a2eabb87991f3ff263a763e0e7c1531fa
2020-12-21 20:28:32 -08:00
Chenkai Kuang
12c67c0d47 Raise meaningful error message when loading a ShardedVariable.
PiperOrigin-RevId: 348539354
Change-Id: I2c4a8466c3d1355ec8e5984ed039194c18c4305c
2020-12-21 15:59:59 -08:00
George Karpenkov
8ddb063d4d Fix AbstractStackTrace::GetLastFrame on Windows
+ Minor tf_stack cleanup

PiperOrigin-RevId: 348537476
Change-Id: Id551360886c648ab2580592eb1268e51a300a9c5
2020-12-21 15:35:00 -08:00
Rick Chao
a122c3fb46 Add a note saying callbacks.configure_callbacks is only used in TF1.
PiperOrigin-RevId: 348532143
Change-Id: Id670280c0480862f35f998d35003e07c41e04ad7
2020-12-21 14:57:25 -08:00
Dave Moore
f0db26599f Support batchable CompositeTensors as inputs to vectorized_map.
PiperOrigin-RevId: 348498892
Change-Id: I82ac5012dbf13705af584fb0350f14d170a3ff70
2020-12-21 11:33:45 -08:00
Hye Soo Yang
b8a634fd5a [core API doc fixit] tf.get_static_value; updated the API doc with more explanations and examples.
PiperOrigin-RevId: 348495511
Change-Id: I28b6c2c4822fd2d62eabbb535282ac5087765fae
2020-12-21 11:14:10 -08:00
TensorFlower Gardener
743b347ccf Merge pull request from fsx950223:softnms
PiperOrigin-RevId: 348473065
Change-Id: I6f9df9192146805c6b019cda7742339dff0ee0f7
2020-12-21 08:49:02 -08:00
Dan Moldovan
9b83b647b9 Move stack more of the trace mappers and filters to C++. Simplify the API surface of extract_stack, and clean up its docstring.
PiperOrigin-RevId: 348459279
Change-Id: I1ae8afde96a7220df5fb6a95ace62f7c940a1e71
2020-12-21 06:45:42 -08:00
TensorFlower Gardener
3bff60dca8 Merge pull request from gqqnbig:gqqnbig-patch-1
PiperOrigin-RevId: 348455513
Change-Id: I831fb7c9b27b429a4c7075d33192feace3b45c0d
2020-12-21 06:09:47 -08:00
A. Unique TensorFlower
bf6b5f0e7c compat: Update forward compatibility horizon to 2020-12-21
PiperOrigin-RevId: 348428041
Change-Id: I87a2efddc8bde427ac1872c4c61163c9eb7ce394
2020-12-21 01:07:41 -08:00
TensorFlower Gardener
ee11cf13ea Merge pull request from ROCmSoftwarePlatform:google_upstream_rocm_add_remove_no_rocm_tag
PiperOrigin-RevId: 348425014
Change-Id: I1f4ec5ab94f27ea201a9dbc8b53d337d03e01d6c
2020-12-21 00:32:48 -08:00
Hye Soo Yang
b6be9714e8 Fix wrong behavior (regression) with tf.image.per_image_standardization. Partially reverting PR (906e0e3bc0). Also added test cases validating against additional data types.
There were several reasons the changes introduced in  did not work:
- Converting standardized image (float) to input `image` data type using `tf.image.convert_image_dtype(..., saturate=True)` loses data (clips half of data) for unsigned data types (reported in  (https://github.com/tensorflow/tensorflow/issues/44983)).
- Standardized images with zero mean will have negative values (unless all values are `0`s). `convert_image_dtype` by the API definition (50216cbe96/tensorflow/python/ops/image_ops_impl.py (L2231-L2234)), however, requires that floating input dtype be in the range [0, 1).

PiperOrigin-RevId: 348379825
Change-Id: I4af7a97b391470a2018738bc6cc573e106e7f4af
2020-12-20 13:15:39 -08:00
jpodivin
4bf812c99a num_replicas now propagates. 2020-12-20 16:35:46 +00:00
A. Unique TensorFlower
50216cbe96 compat: Update forward compatibility horizon to 2020-12-20
PiperOrigin-RevId: 348337423
Change-Id: Ic799a108421feeb6dbfabca55c93c26187968dee
2020-12-20 01:10:42 -08:00
Hye Soo Yang
c97e809422 Add specification of an image batch for tf.image.ssim. Also fix wrong example.
PiperOrigin-RevId: 348306403
Change-Id: I37ea8d88d246b8e4818459fe3b70e33b0004fba7
2020-12-19 15:54:15 -08:00
A. Unique TensorFlower
a52ab4b03c compat: Update forward compatibility horizon to 2020-12-19
PiperOrigin-RevId: 348256314
Change-Id: I34260b5cafa8802dd756db151364199d710547f9
2020-12-19 01:14:35 -08:00
Rick Chao
e8f1a46dc1 PSv2: Privatize ClusterCoordinator.cluster attribute since Cluster is not meant for public use.
PiperOrigin-RevId: 348168836
Change-Id: I9abbd50de59861f02139d7472b397b937e1a4e85
2020-12-18 02:56:32 -08:00
fsx950223
04c960583b fix softnms bug when set iou threshold 2020-12-18 10:14:18 +00:00
A. Unique TensorFlower
b5c8d770f5 compat: Update forward compatibility horizon to 2020-12-18
PiperOrigin-RevId: 348160936
Change-Id: I316bdb8707cd915c1c152d56dc9d3855be43f7c1
2020-12-18 01:12:41 -08:00
A. Unique TensorFlower
4ad325ebf9 In RaggedTensorDynamicShape.from_tensor: Check for uniform_row_lengths.
PiperOrigin-RevId: 348138366
Change-Id: Ia0bc08f8fe00ec9714041184f71f4cee7106c31f
2020-12-17 20:05:38 -08:00
George Karpenkov
00475eda7c [TF2XLA] Support asserts as no-ops for tf.function(jit_compile=True)
This is consistent with TPU and tfcompile behavior.

PiperOrigin-RevId: 348131078
Change-Id: I4f8807af60fdbe1d79f0e79db59df0feed79c94f
2020-12-17 18:53:10 -08:00
Chris Kennelly
4692525ffa Optimize calls to std::string::find() and friends for a single char.
The character literal overload is more efficient.

PiperOrigin-RevId: 348126864
Change-Id: I12485209607a957ecb17a4ba1087473bb0c4dd06
2020-12-17 18:11:13 -08:00