profile
viewpoint
Brian Zhao bmzhao Google United States

bmzhao/abseil-cpp 0

Abseil Common Libraries (C++)

bmzhao/aida 0

AIDA Named Entity Disambiguation by the Databases and Information Systems Group at the Max Planck Institute for Informatics.

bmzhao/appdocs 0

Application Performance Optimization Summary

bmzhao/asciify 0

Turns images to Ascii Strings, displayable in your terminal!

bmzhao/bazel-cc-so-example 0

Bazel CC Shared Library Developer Flow Use Case Examples

bmzhao/community 0

Stores documents used by the TensorFlow developer community

bmzhao/cpp-parking-backend 0

Backend server that synchronizes all parking state

bmzhao/CRP 0

C++ Implementation of Customizable Route Planning (CRP) by Delling et al.

push eventbmzhao/community

Brian Zhao

commit sha 90d2c96b12f9c5972e3c28f85d69edca56772412

RFC: Tensorflow SavedModel C/C++ API

view details

push time in 5 days

push eventbmzhao/community

Brian Zhao

commit sha 997314234617c6601ae8ba8451496c0e1d31e948

RFC: Tensorflow SavedModel C/C++ API

view details

push time in 5 days

push eventbmzhao/community

Brian Zhao

commit sha 0be2769de6bb07129df0004c578b58459a18f54f

RFC: Tensorflow SavedModel C/C++ API

view details

push time in 6 days

pull request commenttensorflow/tensorflow

Fix Bazel not building anymore with the commit 09fe958f

According to the commit https://github.com/bazelbuild/bazel/commit/755e29dfb21d7abfc2bab8dc30813fd5e340d930, the experimental_repo_remote_exec flag was only introduced in bazel 2.0.0, meaning any older bazel versions will error out when they see the flag.

Since this change bumps the minimal bazel version to 2.0.0, this change LGTM!

DEKHTIARJonathan

comment created time in 6 days

push eventbmzhao/community

Brian Zhao

commit sha 1345ed5e3f42e9f75e4f623a80ad04992d92b2c2

RFC: Tensorflow SavedModel C/C++ API

view details

push time in 6 days

push eventbmzhao/community

Brian Zhao

commit sha 3f371e65ad93a8cfaa17c56b960f88bad43dffb6

RFC: Tensorflow SavedModel C/C++ API

view details

push time in 6 days

PR opened tensorflow/community

Reviewers
RFC: Tensorflow SavedModel C/C++ API

This RFC will be open for comment until Friday, March 6th, 2020.

cc @gunan @martinwicke

Tensorflow SavedModel C/C++ API

Status Proposed
Authors Brian Zhao (bmzhao@google.com), Hye Soo (hyey@google.com), Paige Bailey (webpaige@google.com)
Sponsor Gunhan Gulsoy (gunan@google.com)
Updated 2020-02-19

Objectives

We would like to revamp the C++ saved model user journey by

  1. Examining the set of use cases we intend to support
  2. Identifying API constraints, requirements, and pain points for each use case
  3. Designing a new C++ saved model API, while being thoughtful of evolution in the surrounding space (notably TF2 and MLIR)
  4. Creating detailed documentation for C++ API and examples.
+524 -0

0 comment

2 changed files

pr created time in 6 days

push eventbmzhao/community

Brian Zhao

commit sha 4259bc91d6d2704eaeb8bd2ff884602925b4bc40

RFC: Tensorflow SavedModel C/C++ API

view details

push time in 6 days

create barnchbmzhao/community

branch : tf-c-saved-model

created branch time in 7 days

push eventbmzhao/community

Gunhan Gulsoy

commit sha 3f3c18f7750aa389fa3b2332d96078ae2756cce6

Modular tensorflow public design review.

view details

Gunhan Gulsoy

commit sha 6526e3059de1062cfe435a687898f9707df2dac8

Update the proposal status.

view details

Edd Wilder-James

commit sha 709c3a43c2ae9a13e4a08a04e975463eb9be31c9

Update 20190305-modular-tensorflow.md Here are my proposed changes, to streamline the introduction and correct a few typos.

view details

Gunhan Gulsoy

commit sha 809fb92464a7d25c8e8b867501dc8ed01bb4db89

Merge pull request #1 from ewilderj/patch-1 Update 20190305-modular-tensorflow.md

view details

Gunhan Gulsoy

commit sha b42798f11d59dc9818bce4bd2cf6a276b4889518

Fix markdown typos.

view details

lzr-google

commit sha 523f4955d33d49fce1687ba2adebb37fd34424b7

TF2 TPU SavedModel RFC Doc Draft

view details

Zhuoran Liu

commit sha aa140144ffbaf46c7bceeab4d8fc2b6c62b284cd

Update 20191106-tf2-tpu-savedmodel.md

view details

Zhuoran Liu

commit sha 4a6fa51b97c7d607199c3db301ce28644ba464aa

Update rfcs/20191106-tf2-tpu-savedmodel.md: Set RFC# Co-Authored-By: Edd Wilder-James <edd@me.com>

view details

Zhuoran Liu

commit sha b83d5455ab8e508bd3f516799f03aa0db094fd0c

Update rfcs/20191106-tf2-tpu-savedmodel.md Co-Authored-By: Edd Wilder-James <edd@me.com>

view details

lzr-google

commit sha 474d9845d23352dd0a10bf04d70d171c7ae26ca6

Update: 2019-11-13 12:10 pm

view details

lzr-google

commit sha 9008400a9087fbeec55d58910696878d44d622b9

Update: 2019-11-13 12:15 pm

view details

Zhuoran Liu

commit sha 0bfae5b20f62edf29e3c3ffa4049eaffb1fa971f

Update 20191106-tf2-tpu-savedmodel.md

view details

Edd Wilder-James

commit sha 3bc8692d187f3495a81667261500b033a229533d

Update CODEOWNERS Add @brijk7 as RFC code owner.

view details

Gunhan Gulsoy

commit sha 6faac9b8e76f03acd2a87edb774b1bfc51c26f9b

Relax API requirements.

view details

Gunhan Gulsoy

commit sha 4e3443a512794c45e75c4a7cc8128e2019685aad

Further edits.

view details

Gunhan Gulsoy

commit sha 1d07d97512b939e9f07830e3d86c769810a690f3

Update last updated time.

view details

Gunhan Gulsoy

commit sha 9de6d4c0474af0ca04bee84c6528bc6b4b38b5ac

Merge pull request #2 from gunan/revNplus1 Relax API requirements.

view details

Thea Lamkin

commit sha 66ee559e931e435e29c72a566bc7a8f25ab08430

Merge pull request #77 from gunan/master RFC: Modular TensorFlow

view details

Saurabh Saxena

commit sha a05e0da52abdf3aa96ebdff51f635ede59d22ba4

First commit

view details

Saurabh Saxena

commit sha a920a4d902bef24aa259496f6d063021a4bf3019

Move to rfcs/

view details

push time in 7 days

push eventbmzhao/pytorch

peterjc123

commit sha ebed008dd4a13634de2b0639b417423f0c476455

Correct /MP usage in MSVC (#33120) Summary: ## Several flags `/MP[M]`: It is a flag for the compiler `cl`. It leads to object-level multiprocessing. By default, it spawns M processes where M is the number of cores on the PC. `/maxcpucount:[M]`: It is a flag for the generator `msbuild`. It leads to project-level multiprocessing. By default, it spawns M processes where M is the number of cores on the PC. `/p:CL_MPCount=[M]`: It is a flag for the generator `msbuild`. It leads the generator to pass `/MP[M]` to the compiler. `/j[M]`: It is a flag for the generator `ninja`. It leads to object-level multiprocessing. By default, it spawns M processes where M is the number of cores on the PC. ## Reason for the change 1. Object-level multiprocessing is preferred over project-level multiprocessing. 2. ~For ninja, we don't need to set `/MP` otherwise M * M processes will be spawned.~ Actually, it is not correct because in ninja configs, there are only one source file in the command. Therefore, the `/MP` switch should be useless. 3. For msbuild, if it is called through Python configuration scripts, then `/p:CL_MPCount=[M]` will be added, otherwise, we add `/MP` to `CMAKE_CXX_FLAGS`. 4. ~It may be a possible fix for https://github.com/pytorch/pytorch/issues/28271, https://github.com/pytorch/pytorch/issues/27463 and https://github.com/pytorch/pytorch/issues/25393. Because `/MP` is also passed to `nvcc`.~ It is probably not true. Because `/MP` should not be effective given there is only one source file per command. ## Reference 1. https://docs.microsoft.com/en-us/cpp/build/reference/mp-build-with-multiple-processes?view=vs-2019 2. https://github.com/Microsoft/checkedc-clang/wiki/Parallel-builds-of-clang-on-Windows 3. https://blog.kitware.com/cmake-building-with-all-your-cores/ Pull Request resolved: https://github.com/pytorch/pytorch/pull/33120 Differential Revision: D19817227 Pulled By: ezyang fbshipit-source-id: f8d01f835016971729c7a8d8a0d1cb8a8c2c6a5f

view details

Jeremy Lilley

commit sha 6be4ec100f4f2882a22427cc4ff2f62088184f24

[pytorch] Elide more Thrift Tensor send copies. (#31998) Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/31998 This change builds on recent torch::from_blob() changes to avoid Tensor copies on send in more cases. Particularly, this change adds an enabled option to assume if the Tensor Storage's DataPtr has a non-trivial deleter, then the Tensor does in fact manage the underlying memory. And hence we can reference the Tensor's Storage via an IOBuf that is referenced while sending, saving a Tensor copy. We add appropriate test cases, particularly re: torch::from_blob() which would have been problematic would recent changes. ghstack-source-id: 97778619 Test Plan: buck test mode/dev caffe2/torch/fb/distributed/wireSerializer/test/... Reviewed By: satgera Differential Revision: D19306682 fbshipit-source-id: 05f56efb2d5d6279ae4b54dfcbba0f729c2c13fa

view details

George Guanheng Zhang

commit sha f4fbe9549d34512d5641e913e3a50661645a1b87

Revert D19800021: [pytorch][PR] Improve error message for assertWarnsRegex Test Plan: revert-hammer Differential Revision: D19800021 Original commit changeset: 1c31ae785c8f fbshipit-source-id: d7b340d678562c25a84d48be66c576075000b50d

view details

Jithun Nair

commit sha 3c4cec56aaa3c5094249fdb1a9a139407ceb305c

Enable test_distributed for ROCm but only with nccl backend [REDUX] (#32551) Summary: This is a redux of the original PR https://github.com/pytorch/pytorch/issues/28814 which was reverted in PR https://github.com/pytorch/pytorch/issues/29736 due to test_DistributedDataParallel being suspected as being flaky. Further investigation revealed it wasn't flakiness, but a bug in the PyTorch source code which has been now fixed in PR https://github.com/pytorch/pytorch/issues/32356. This PR is another attempt at enabling the test_distributed unit test suite only for the nccl backend. Pull Request resolved: https://github.com/pytorch/pytorch/pull/32551 Differential Revision: D19729966 Pulled By: bddppq fbshipit-source-id: 12a0d850991a903cc7723d63693b6157071d7115

view details

Gregory Chanan

commit sha 3bde97d5a5e603f894b206fb814d2381181cffab

Move a resize from codegen to code. Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/33024 Test Plan: Imported from OSS Differential Revision: D19774147 Pulled By: gchanan fbshipit-source-id: 08cb099f1695b28117e4236e214976b548aec7a1

view details

Karl Ostmo

commit sha e8c4f5a74bb491a6474442b2d966c578e7ac2565

Temporarily disable failing iOS builds Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/33154 Differential Revision: D19820655 Pulled By: kostmo fbshipit-source-id: fc3e22b1bf4ec112085ea846c3999efd0f3e26f3

view details

Jiakai Liu

commit sha 495c1df51011625d699f488388687b162dee2fc6

[pytorch] convert code analyzer to a binary (#33102) Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/33102 Add a simple main() to build code analyzer as a binary. This enables easier integration with FB internal build environment. ghstack-source-id: 97958658 Test Plan: - CI Differential Revision: D19798560 Pulled By: ljk53 fbshipit-source-id: 126230e3bf7568046a309e8a6785230f820e0222

view details

Ivan Kobzarev

commit sha d6727793393a0cc44ebc59e84c7e207538f4962a

[CI][treehug] Disable xenial_py2.7 tests due to mypy min version py3.5 Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/33159 Test Plan: Imported from OSS Differential Revision: D19822400 Pulled By: IvanKobzarev fbshipit-source-id: 8e7b561e6a6181ec1f9b6f56a539ddcb538b3858

view details

svcscm

commit sha 524fe8a96c5cc195b81095b5419168c55ce9595e

Updating submodules Summary: GitHub commits: https://github.com/facebook/fbthrift/commit/4bc5213b66d4c7d3c96c6d5c47945dae4881c921 https://github.com/facebook/proxygen/commit/9ae570bb8975b2cc0c7d703f6c34e0c9075ff32f https://github.com/facebook/rocksdb/commit/b2bc1da56197aa68a71eff2344e4393042c88eb6 https://github.com/pytorch/fbgemm/commit/dcde8696bdf40adf8c436cdaa805973613718ae4 Test Plan: n/a Reviewed By: zpao fbshipit-source-id: c5ca30dab73f80cd13f5a5bf6e3867083b2512ac

view details

Ailing Zhang

commit sha a3e69d3405564db650a3659c29fdd424c0835644

Use bazelisk instead of specifying bazel version manually. (#33036) Summary: Bazelisk automatically reads `.bazelversion` file and install the required version of Bazel. This saves us from updating CI script everytime we need a Bazel upgrade. Use clang-8 for consistency with pytorch/xla repo. Pull Request resolved: https://github.com/pytorch/pytorch/pull/33036 Differential Revision: D19820819 Pulled By: ailzhang fbshipit-source-id: 1560ec225cd037a811769a509a704b0df77ea183

view details

svcscm

commit sha 61ac14a4830d68d053f7b900c79ef3a74a42a2fd

Updating submodules Summary: GitHub commits: https://github.com/facebook/fb303/commit/543b39c9ada4316d4d45a22b072d9b962cc07cb0 https://github.com/facebook/fbzmq/commit/38c2e0ee448fa7a33b4974d7e0922738bcd8f2d3 https://github.com/facebook/folly/commit/552c07c32bc6604cfc729060a4544d03a69e3f5e https://github.com/facebook/rocksdb/commit/4369f2c7bb2fe2d67b5de96688ced5196a0be3b9 https://github.com/pytorch/fbgemm/commit/07dbb5d2f44d89cdfdf8337b2d08a6130ae89062 Test Plan: n/a Reviewed By: zpao fbshipit-source-id: 803108a618a5be9ea58a38644c851486bad3bfbc

view details

Lin Yang

commit sha 6f46962f214aaaf806dc2e670383923619ed26f5

[1/3] Bind IndexHash to PyTorch (#33015) Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/33015 Export IndexHash to PyTorch Test Plan: buck test caffe2/caffe2/python/operator_test:torch_integration_test ✓ caffe2/caffe2/python/operator_test:torch_integration_test-2.7 - test_index_hash_op (caffe2.caffe2.python.operator_test.torch_integration_test.TorchIntegration) 0.151 44/50 (passed) Reviewed By: bddppq Differential Revision: D19727301 fbshipit-source-id: a65c954539e81a15577fe5c3c0deb3614e983534

view details

Natalia Gimelshein

commit sha 9857d9b4cd4b50c326c0d8309e92e5bbde083a84

fix gather regression by not materializing loop vars in the error mes… (#33108) Summary: …sage Per title, fixes regression reported in https://github.com/pytorch/pytorch/issues/32425. cc nikitaved Pull Request resolved: https://github.com/pytorch/pytorch/pull/33108 Differential Revision: D19816116 Pulled By: ngimel fbshipit-source-id: 9f4a84c8e4533873b71bb7bbf3a7915b05308845

view details

Gerard Goossen

commit sha 330d051bd5ac67c4ce72a60f2f10b405309cc82e

[pytorch] Migrating index_add cuda to ATen (#30573) Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/30573 Mostly just moved code. Index dim and number of indices checks are added to make checks idential to index_add_cpu_ ghstack-source-id: 98010129 Test Plan: existing tests Differential Revision: D18749922 fbshipit-source-id: d243be43a3b6a9b9591caf0c35ef2fb6ec0d3ead

view details

albanD

commit sha 36559755653a6548773cff33370c428f022d4831

Add allow_rebase_history flag and fix codegen functions for multiple views (#32790) Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/32790 Same as https://github.com/pytorch/pytorch/pull/31990 but without the first commit in the stack that is problematic for a lot of people. Test Plan: Imported from OSS Differential Revision: D19814116 Pulled By: albanD fbshipit-source-id: d104911a5b098a5807b4bc08b69803ebd4f69fa6

view details

Jongsoo Park

commit sha 6c0dc66cb478ff0afa7d70e6a57a9cfb2deb4e73

[caffe2] use JIT'ed fp32 SLS (#33123) Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/33123 Pull Request resolved: https://github.com/pytorch/pytorch/pull/32413 Use JIT'ed fp32 SLS in Caffe2 operators Test Plan: ``` ./fblearner/flow/run_integration_tests --regex dper.workflows.canary.canary_workflow --wait ``` f167043951 was killed due to 3hr timeout instead of failed. Reviewed By: jianyuh Differential Revision: D19680711 fbshipit-source-id: efaca333edcfeab0007ad88f4f5168b2229e7e66

view details

Vincent Quenneville-Belair

commit sha e7f0b15473d4dea3e202c98ccafe0d054cb0a204

Remove return value for __exit__ (#32997) Summary: When an error is raised and `__exit__` in a context manager returns `True`, the error is suppressed; otherwise the error is raised. No return value should be given to maintain the default behavior of context manager. Fixes https://github.com/pytorch/pytorch/issues/32639. The `get_lr` function was overridden with a function taking an epoch parameter, which is not allowed. However, the relevant error was not being raised. ```python In [1]: import torch ...: ...: class MultiStepLR(torch.optim.lr_scheduler._LRScheduler): ...: def __init__(self, optimizer, gamma, milestones, last_epoch = -1): ...: self.init_lr = [group['lr'] for group in optimizer.param_groups] ...: self.gamma = gamma ...: self.milestones = milestones ...: super().__init__(optimizer, last_epoch) ...: ...: def get_lr(self, step): ...: global_step = self.last_epoch #iteration number in pytorch ...: gamma_power = ([0] + [i + 1 for i, m in enumerate(self.milestones) if global_step >= m])[-1] ...: return [init_lr * (self.gamma ** gamma_power) for init_lr in self.init_lr] ...: ...: optimizer = torch.optim.SGD([torch.rand(1)], lr = 1) ...: scheduler = MultiStepLR(optimizer, gamma = 1, milestones = [10, 20]) ``` ``` --------------------------------------------------------------------------- TypeError Traceback (most recent call last) <ipython-input-1-7fad6ba050b0> in <module> 14 15 optimizer = torch.optim.SGD([torch.rand(1)], lr = 1) ---> 16 scheduler = MultiStepLR(optimizer, gamma = 1, milestones = [10, 20]) <ipython-input-1-7fad6ba050b0> in __init__(self, optimizer, gamma, milestones, last_epoch) 6 self.gamma = gamma 7 self.milestones = milestones ----> 8 super().__init__(optimizer, last_epoch) 9 10 def get_lr(self, step): ~/anaconda3/envs/pytorch/lib/python3.7/site-packages/torch/optim/lr_scheduler.py in __init__(self, optimizer, last_epoch) 75 self._step_count = 0 76 ---> 77 self.step() 78 79 def state_dict(self): ~/anaconda3/envs/pytorch/lib/python3.7/site-packages/torch/optim/lr_scheduler.py in step(self, epoch) 141 print("1a") 142 # try: --> 143 values = self.get_lr() 144 # except TypeError: 145 # raise RuntimeError TypeError: get_lr() missing 1 required positional argument: 'step' ``` May be related to https://github.com/pytorch/pytorch/issues/32898. Pull Request resolved: https://github.com/pytorch/pytorch/pull/32997 Differential Revision: D19737731 Pulled By: vincentqb fbshipit-source-id: 5cf84beada69b91f91e36b20c3278e9920343655

view details

Iurii Zdebskyi

commit sha 857bae39e0b06266c5d5ecd2e0f77499ad858747

Updated DispatchKeyExtractor to expect TensorOptions (#30981) Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/30981 This stack is a first step toward an effort to fix, clean up and simplify code generation logic. �Please see the master [task](https://github.com/pytorch/pytorch/issues/30405) to see related discussions and all the known issues. Main focus of these changes is TensorOptions in code generation. Goals: - Remove TensorOptions from generated code wherever it's possible. Leave it only in python/C++ API layers. - Refactor TensorOptions logic to a single place. - Log all discovered issues. Non goals: - Fix Everything! - Remove all the hacks in code generation scripts. - Clean up and defector all code generation scripts. ----------- In this PR: Extended DispatchKeyExtractor logic to expect TensorOptions. ----------- Test Plan: Imported from OSS Differential Revision: D18912684 Pulled By: izdeby fbshipit-source-id: 25cf1c397caa14272ca65b4003f1f03ff282ea77

view details

George Guanheng Zhang

commit sha 855ee6446fa40158f6905321a6e3d91d9f946861

Revert D18749922: [pytorch] Migrating index_add cuda to ATen Test Plan: revert-hammer Differential Revision: D18749922 Original commit changeset: d243be43a3b6 fbshipit-source-id: 15dafa644d84ff8803bd9ab3cdd40e12d805924a

view details

Hong Xu

commit sha 31370949bef1d0bce708db7833e1789021cc308a

Add zero_mask function for vectorized functions. (#32985) Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/32985 This can be useful in many situations to decide whether all elements are zeros or non-zeros, such as elu as shown in #32986 . Test Plan: Imported from OSS Differential Revision: D19794549 Pulled By: VitalyFedyunin fbshipit-source-id: 1be1c863d69b9a19fdcfcdd7cb52343066f740d3

view details

push time in 9 days

fork bmzhao/Kyle-Landry-Sheets

:musical_keyboard: :musical_keyboard: Links to transcriptions of Kyle Landry's sheet music, currently hidden due to syntax changes in PHPBB

fork in 11 days

push eventbmzhao/tensorflow

TensorFlower Gardener

commit sha b83cc0ac8eb83c85b78778bbe8a0c96f323d747e

Merge pull request #22231 from MichaelKonobeev:sparse-xent-op-hessian PiperOrigin-RevId: 260802377

view details

MichaelKonobeev

commit sha ea809e3ad7c0d8a1fc1170dec6c782c7feac299b

Implement IsZero in eager mode

view details

Trent Lo

commit sha b33c788a2f479a4753f49b566f08079692c75af2

Implement horizontal fusion. - It reduces kernel launch overhead and increases lauch dims by horizontally fusing indepedent computations.

view details

Trent Lo

commit sha cd68827e01d454937399bafcdb1eb4b9a116678a

Minor cleanup for horizontal fusion.

view details

Trent Lo

commit sha cb9ab8bee96530c9973d5e295b53d936cbf8ef72

Polishing coding style and comments.

view details

Trent Lo

commit sha 1876f2acc02dee840b3a8b6ab59f950b5a3bbf4f

Factor out lambdas in HorizontalFusionImpl.

view details

Trent Lo

commit sha 86bd5bf3e75cb5d14d24194a2d1e2d8f60753b03

Comment polishing.

view details

Trent Lo

commit sha 474e79985f722afa57d12447fb2f4dc30e890d06

Add some more unittests for horizontal fusion. In addition, we record the execution time of the tests here, showing the optimization effects of horizontal fusion, measured by --xla_hlo_profile. The accumulated kernel execution time in GradientDescentOptimizerLike is reduced from 2.39ms to 311us; the execution time in RMSProp is reduced from 980us to 112us. Before horizontal fusion: 2019-12-10 22:05:45.215015: I tensorflow/compiler/xla/service/executable.cc:208] Execution profile for GradientDescentOptimizerLike: (2.39 ms @ f_nom) 2019-12-10 22:05:48.877372: I tensorflow/compiler/xla/service/executable.cc:208] Execution profile for RMSPropLike: (980 us @ f_nom) After horizontal fusion: 2019-12-10 22:05:03.831600: I tensorflow/compiler/xla/service/executable.cc:208] Execution profile for GradientDescentOptimizerLike: (311 us @ f_nom) 2019-12-10 22:05:13.513901: I tensorflow/compiler/xla/service/executable.cc:208] Execution profile for RMSPropLike: (112 us @ f_nom)

view details

Trent Lo

commit sha a629a452bff5b7f7f2688086483d7eb8d3d02420

Polishing comments and coding styles.

view details

Stephan Uphoff

commit sha 7813cb00f35d6fc6d8ad8421021c1535f3e8c029

lite/micro: Add feature buffer to micro_speech example. This fixes #35117 Accumulate feature slices in separate buffer. The input tensor is not suitable for keeping state across interference as it has limited lifetime and the buffer space may be reused.

view details

Trent Lo

commit sha 7abde726e4706df2fa83c2ec3c89ef9fb5c99228

Polish coding styles and comments based on review feedback. In addition, use hlo_matcher to verify resultant DAGs instead of LLVM filecheck.

view details

Trent Lo

commit sha 5f5aa78f86a43d073663cc0f96acb3926d621e42

Merge branch 'upstream_master_dec19' into horizontal_fusion_github

view details

MichaelKonobeev

commit sha 6fe6391ea937a3c20308b3986f7232967e6f0268

Unconditionally tag zero tensors

view details

MichaelKonobeev

commit sha b187faf53c68ff9b0c711b246116fb81660ad4c7

Remove expired forward compatibility check

view details

MichaelKonobeev

commit sha cb9ce8a40c41d35725900f0f0e12a934e28ba837

Merge branch 'master' into sparse-xent-op-hessian

view details

Eugene Kuznetsov

commit sha 968a674ecb6db34e5d2e09068a8d9ca5ca4e3e24

Enable //tensorflow/python:stateful_random_ops_test

view details

Eugene Kuznetsov

commit sha f7b28191777b6ae86c0dbdab7a74b8370e53eaa8

Fix for //tensorflow/python:stateful_random_ops_test: Pack arguments of UpdateVariableAndFill_Philox into a struct

view details

Eugene Kuznetsov

commit sha eee5851777b842945b12937600b005a58aae0f2c

Fix for //tensorflow/python:stateful_random_ops_test: Move the thread counter into the global namespace

view details

Trent Lo

commit sha 47ba0995d9838e5f9aa634abc59f4569c4a37375

Fix a buildifier format issue.

view details

RichardXiao13

commit sha f8a15ce2b6f48523effe2dd42e7844ea7ef1d97a

Add usage example to math.poly_val

view details

push time in 15 days

push eventbmzhao/pytorch

Shihao Xu

commit sha 1695915371e603440b314f7bd08178a799caa851

Make _wait_all_workers() support being called for multiple times (#32624) Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/32624 We need this PR to resolve the issue mentioned in https://github.com/pytorch/pytorch/issues/31325#issuecomment-574918917. The solution is for each `_wait_all_workers()` call, there is a sequence ID added, to identify different calls. ghstack-source-id: 97277591 Test Plan: ``` buck test mode/dev-nosan //caffe2/test/distributed/rpc:rpc_fork -- test_wait_all_workers buck build mode/dev-nosan //caffe2/test/distributed/rpc:rpc_fork buck-out/gen/caffe2/test/distributed/rpc/rpc_fork\#binary.par -r test_wait_all_workers ``` Differential Revision: D5739520 fbshipit-source-id: a64131e09c365179624700514422f5375afe803f

view details

svcscm

commit sha 1217c9b3640f88b97f25a055197bcc1ae05750a1

Updating submodules Summary: GitHub commits: https://github.com/facebook/fbthrift/commit/3f156207e8a6583d88999487e954320dc18955e6 https://github.com/facebook/folly/commit/135cff30a54b77523ff404a269a960ad981ff8df https://github.com/facebook/rocksdb/commit/7aa66c704f71d74ff97090caee1d29c4dff22a21 https://github.com/facebookincubator/katran/commit/1dc41366445c0a1eef50e7329c5ea3069763c91b https://github.com/pytorch/fbgemm/commit/9166d9f76755274970e77dc1386bcdd1bd91c9a4 Test Plan: n/a Reviewed By: 2d2d2d2d2d fbshipit-source-id: fb27e09060ecb4278b4002c02bce48fe9f4dc361

view details

Shihao Xu

commit sha 5c8535d5b0cf77ff75679ca5c148a723be7dd10f

Make C++ RpcAgent::currentRPCAgent_ the source of truth of current RPC Agent (#32633) Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/32633 There were 2 sources of current RPC agent. - One is in Python world, `torch.distributedrpc.api._agent`. - The other is in C++ world, `RpcAgent::defaultRpcAgent_` Setting Python `_agent` to `None`, does not necessarily reset the C++ `defaultRpcAgent_` to `nullptr`. i.e. ``` torch.distributedrpc.api._agent = None ``` does not translate to ``` RpcAgent::defaultRpcAgent_ = nullptr ``` This PR is to remove this ambiguity, and use the C++ pointer as source of truth. The solution is to leverage a pybind11 behavior that it implicitly casts C++ `shared_ptr<RpcAgent>(nullptr)` to Python `None`. ghstack-source-id: 97293315 Test Plan: ``` buck test mode/dev-nosan //caffe2/test/distributed/rpc:rpc_fork -- test_duplicate_name buck build mode/dev-nosan //caffe2/test/distributed/rpc:rpc_fork buck-out/gen/caffe2/test/distributed/rpc/rpc_fork\#binary.par -r test_process_group_debug_info ``` ``` buck test mode/dev-nosan //caffe2/torch/fb/distributed/pytorch/tests:test_remote_module buck test mode/dev-nosan //caffe2/torch/fb/distributed/modules/tests:test_sharded_embedding buck test mode/dev-nosan //caffe2/torch/fb/distributed/modules/tests:test_sharded_pairwise_attention_pooling buck test mode/dev-nosan //caffe2/torch/fb/distributed/pytorch/tests:test_rpc ``` Differential Revision: D5733066 fbshipit-source-id: b3e6032ee975f19ca556497edbbf40b517b25be8

view details

Alban Desmaison

commit sha db8ce7ea2dd3a433d94f00969282a5178fcabd72

Back out "Make autogen functions correct for multiple outputs and views" (#32681) Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/32681 Original commit changeset: a2b41c2d231e Test Plan: fb and oss tests Reviewed By: hudeven Differential Revision: D19591864 fbshipit-source-id: 7068b5563e37bc9a5d415fd535c73fd9d71fe131

view details

Hong Xu

commit sha 666e5430f809c6c93a30144f1415170411c276ed

Clean up mvlgamma doc (including a weird way to link to reference) (#32667) Summary: Intentionally left blank Pull Request resolved: https://github.com/pytorch/pytorch/pull/32667 Differential Revision: D19594683 Pulled By: ezyang fbshipit-source-id: 5a6eb0a74f569d3c0db2a35e0ed4b329792a18e4

view details

Shihao Xu

commit sha b9f764b1c70882e7ef211b8e42ff300ca5a60109

Use the C++ current RpcAgent pointer to eliminate the unnecessary argument passing from Python world (#32635) Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/32635 With the source of truth of current RPC agent moved to C++ world, there is no point of passing current RPC agent from Python world to C++ world. ghstack-source-id: 97293316 Test Plan: ``` buck test mode/dev-nosan //caffe2/test/distributed/rpc:rpc_fork buck build mode/dev-nosan //caffe2/test/distributed/rpc:rpc_fork buck-out/gen/caffe2/test/distributed/rpc/rpc_fork\#binary.par -r test_process_group_debug_info ``` Differential Revision: D5703519 fbshipit-source-id: ef7c28bdb1efd293eb6cafe0b0fca7ca80fa08a6

view details

James Reed

commit sha d68592a440e06ea87008b16a49500893baf7ebf7

[JIT] Fix classes as attributes in recursive scripting Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/32594 Test Plan: Imported from OSS Differential Revision: D19562951 Pulled By: jamesr66a fbshipit-source-id: 3d5491c1c23456f107390a78be16da687de951e6

view details

Pritam Damania

commit sha b3848c568e3434896278bec59ac126ed02e894e1

Fix flaky test_nccl_timeout. (#32653) Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/32653 This test was flaky since the watchdog thread could abort the communicator instead of the thread calling `wait()`. As a result, we could actually see `NCCL error` instead of `Operation timed out` on the user end. ghstack-source-id: 97250714 Test Plan: waitforbuildbot Differential Revision: D19583003 fbshipit-source-id: 5c07326d1a16f214dcdbabed97ca613e0a5b42b9

view details

Wojciech Baranowski

commit sha 8e4161517e112478a1c1f0290fedb91965f95aff

div_kernel: throw when dividing by integer zero (#32629) Summary: Fixes https://github.com/pytorch/pytorch/issues/327 Pull Request resolved: https://github.com/pytorch/pytorch/pull/32629 Differential Revision: D19595782 Pulled By: ezyang fbshipit-source-id: f5bbb298f150efe63a698e8a0b53a84871d16560

view details

Michael Suo

commit sha 63170431f975e6ec1d5f612b9cafafeefb45ac2a

[jit] fix segfault on missing getstate (#32642) Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/32642 Previously, if we defined `__setstate__` but not `__getstate__`, we would segfault. This PR turns that into a comprehensible error message (and improves another error message as well). Fixes https://github.com/pytorch/pytorch/issues/25886 Test Plan: Imported from OSS Differential Revision: D19596463 Pulled By: suo fbshipit-source-id: dbe76bc36bc747d65fb0223184c009e0e9ba072c

view details

comet

commit sha 9a2691f2fc948b9792686085b493c61793c2de30

Fix spelling errors Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/32673 Differential Revision: D19597118 Pulled By: pietern fbshipit-source-id: f88c1da7548fcee141ed248f5f49d25c1d639955

view details

Gregory Chanan

commit sha e24ce0e524ba88df6ddab487ad9039ea1f324e01

Kill some more unused code in function_wrapper.py Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/32600 Test Plan: Imported from OSS Differential Revision: D19565654 Pulled By: gchanan fbshipit-source-id: 993c3dc5467639a7690109d07911951a165a412f

view details

Martin Yuan

commit sha c64dec199305de5b9ed97e0e9f71312e7494efce

Python binding to export bytecode format for lite interpreter (#32621) Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/32621 Export the "_save_for_mobile" method to Python so that the bytecode format for lite interpreter can be added or updated to the original script model. It's the first step of python binding for lite interpreter, as discussed in this [internal post](https://fb.workplace.com/groups/1144215345733672/permalink/1478900738931796/) and offline. Next step is to export the load_for_mobile and run method of mobile module, so that users could verify the mobile model from Python. Test: use the following python script to display the bytecode part of the updated model file. ``` #!/usr/bin/env python3 import sys import pickle import pprint import zipfile class FakeObject(object): def __init__(self, module, name, args): self.module = module self.name = name self.args = args self.state = None def __repr__(self): state_str = "" if self.state is None else f"(state={self.state!r})" return f"{self.module}.{self.name}{self.args!r}{state_str}" def __setstate__(self, state): self.state = state class FakeClass(object): def __init__(self, module, name): self.module = module self.name = name self.__new__ = self.fake_new def __repr__(self): return f"{self.module}.{self.name}" def __call__(self, *args): return FakeObject(self.module, self.name, args) def fake_new(self, *args): return FakeObject(self.module, self.name, args) class DumpUnpickler(pickle._Unpickler): def find_class(self, module, name): return FakeClass(module, name) def persistent_load(self, pid): return FakeObject("pers", "obj", (pid,)) def main(argv): zfile = zipfile.ZipFile(argv[1]) names = [i for i in zfile.namelist() if "bytecode.pkl" in i] if not names: print("bytecode.pkl not found.") return with zfile.open(names[0], "r") as handle: value = DumpUnpickler(handle).load() pprint.pprint(value) if __name__ == "__main__": sys.exit(main(sys.argv)) ``` Test Plan: Imported from OSS Differential Revision: D19596359 Pulled By: iseeyuan fbshipit-source-id: 19a4a771320f95217f5b0f031c2c04db7b4079a8

view details

Nikolay Korovaiko

commit sha 0dc38be40732c6d710d6710af243b851dc32a6c6

consider FAIL_GUARD while counting indices for GUARDs (#32672) Summary: This handles a corner case when a user schedules second bailout after the first one and the first one doesn't fire. Alternatively, we could go back to the implementation that uses a hash set to remember the indices of bailouts that need to fire. Pull Request resolved: https://github.com/pytorch/pytorch/pull/32672 Differential Revision: D19596872 Pulled By: Krovatkin fbshipit-source-id: 41dcc374cd2501ac20a9892fb31a9c56d6640258

view details

Eli Uriegas

commit sha da390914bd5f837376095de4c25dd34185595dd3

.circleci: Add workflows for Python 3.8 (#31948) Summary: Done by just editing `.circleci/cimodel/data/dimensions.py` to include `3.8` and then regenerated using `.circleci/regenerate.sh` cc kostmo, mingbowan, ezyang, soumith Signed-off-by: Eli Uriegas <eliuriegas@fb.com> Pull Request resolved: https://github.com/pytorch/pytorch/pull/31948 Differential Revision: D19602069 Pulled By: seemethere fbshipit-source-id: ac57fde9d0c491c7d948a3f5944c3cb324d403c0

view details

James Reed

commit sha 1719da13f9573de540eb4b817719023762fe3716

[JIT] Support for registering C++ lambdas as methods on custom C++ class Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/32553 Test Plan: Imported from OSS Differential Revision: D19543269 Pulled By: jamesr66a fbshipit-source-id: 7e566650295e9d1c4f2f716470e061308a6210a0

view details

James Reed

commit sha 06c19263d3a654e0574f2eb3f4c55e44a512629f

[JIT] Serialize attributes and types in ClassType serialization Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/32555 Test Plan: Imported from OSS Differential Revision: D19544737 Pulled By: jamesr66a fbshipit-source-id: 2256cfba414a850cdc986bb5872dd4cb177b456c

view details

James Reed

commit sha 34ccfba4034ba9d4024958ece7b3e7d0e9f50798

[JIT] Include custom_class.h in torch/script.h Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/32586 Test Plan: Imported from OSS Differential Revision: D19558716 fbshipit-source-id: be540d8ed7de0834e64be89ae621ae50befc83b0

view details

James Reed

commit sha 465ebd58baaa0bc0e61615e74b7596c1331be6f0

[JIT] pickle serialization for custom bound classes Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/32604 Test Plan: Imported from OSS Differential Revision: D19566633 fbshipit-source-id: 9387d3ff45cbd6ccde49ce190a52859481cc301c

view details

James Reed

commit sha 0ea65d63cfb7cbd4110d1099cc09aa7960624a3c

[JIT] Fix stateful lambda stuff and simplify code in custom C++ binding API Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/32658 Test Plan: Imported from OSS Differential Revision: D19584701 Pulled By: jamesr66a fbshipit-source-id: d556c7db2f32900eb1122348402789b59516a7d7

view details

push time in 15 days

issue commenttensorflow/tensorflow

Help! I have an issue when importing TF!

Hi @Brayden1000,

Did you install the Microsoft Visual C++ Redistributable for Visual Studio 2015, 2017, and 2019?

According to the docs at https://www.tensorflow.org/install/pip,

Install the Microsoft Visual C++ Redistributable for Visual Studio 2015, 2017, and 2019. Starting with the TensorFlow 2.1.0 version, the msvcp140_1.dll file is required from this package (which may not be provided from older redistributable packages). The redistributable comes with Visual Studio 2019 but can be installed separately:

Go to the Microsoft Visual C++ downloads, Scroll down the page to the Visual Studio 2015, 2017 and 2019 section. Download and install the Microsoft Visual C++ Redistributable for Visual Studio 2015, 2017 and 2019 for your platform. Make sure long paths are enabled on Windows.

Install the 64-bit Python 3 release for Windows (select pip as an optional feature).

Tensorflow 2.1.0 is compiled using MSVC 2019, which appears to require an additional DLL.

Brayden1000

comment created time in 24 days

issue commenttensorflow/tensorflow

dll loading failed in tensorflow

Hi @sidharth1805,

Did you install the Microsoft Visual C++ Redistributable for Visual Studio 2015, 2017, and 2019?

According to the docs at https://www.tensorflow.org/install/pip,

Install the Microsoft Visual C++ Redistributable for Visual Studio 2015, 2017, and 2019. Starting with the TensorFlow 2.1.0 version, the msvcp140_1.dll file is required from this package (which may not be provided from older redistributable packages). The redistributable comes with Visual Studio 2019 but can be installed separately:

Go to the Microsoft Visual C++ downloads, Scroll down the page to the Visual Studio 2015, 2017 and 2019 section. Download and install the Microsoft Visual C++ Redistributable for Visual Studio 2015, 2017 and 2019 for your platform. Make sure long paths are enabled on Windows.

Install the 64-bit Python 3 release for Windows (select pip as an optional feature).

Tensorflow 2.1.0 is compiled using MSVC 2019, which appears to require an additional DLL.

sidharth1805

comment created time in 24 days

issue commenttensorflow/tensorflow

ImportError: DLL load failed: The specified module could not be found. OS Windows 10 , Tensorflow cpu version

Hi @MISSEY,

Did you install the Microsoft Visual C++ Redistributable for Visual Studio 2015, 2017, and 2019?

According to the docs at https://www.tensorflow.org/install/pip,

Install the Microsoft Visual C++ Redistributable for Visual Studio 2015, 2017, and 2019. Starting with the TensorFlow 2.1.0 version, the msvcp140_1.dll file is required from this package (which may not be provided from older redistributable packages). The redistributable comes with Visual Studio 2019 but can be installed separately:

Go to the Microsoft Visual C++ downloads, Scroll down the page to the Visual Studio 2015, 2017 and 2019 section. Download and install the Microsoft Visual C++ Redistributable for Visual Studio 2015, 2017 and 2019 for your platform. Make sure long paths are enabled on Windows.

Install the 64-bit Python 3 release for Windows (select pip as an optional feature).

Tensorflow 2.1.0 is compiled using MSVC 2019, which appears to require an additional DLL.

MISSEY

comment created time in 24 days

issue commenttensorflow/tensorflow

Tensorflow team is so disappointing in documenting installation on windows pc

Hi @psyg0501,

Did you install the Microsoft Visual C++ Redistributable for Visual Studio 2015, 2017, and 2019?

According to the docs at https://www.tensorflow.org/install/pip,

Install the Microsoft Visual C++ Redistributable for Visual Studio 2015, 2017, and 2019. Starting with the TensorFlow 2.1.0 version, the msvcp140_1.dll file is required from this package (which may not be provided from older redistributable packages). The redistributable comes with Visual Studio 2019 but can be installed separately:

Go to the Microsoft Visual C++ downloads, Scroll down the page to the Visual Studio 2015, 2017 and 2019 section. Download and install the Microsoft Visual C++ Redistributable for Visual Studio 2015, 2017 and 2019 for your platform. Make sure long paths are enabled on Windows.

Install the 64-bit Python 3 release for Windows (select pip as an optional feature).


Tensorflow 2.1.0 is compiled using MSVC 2019, which appears to require an additional DLL.
psyg0501

comment created time in 24 days

push eventbmzhao/tensorflow

Deven Desai

commit sha 09e2eaf34227ef922d8e85b0caef2c0eb5749df5

[ROCm] Enabling vectorization/packet support for some cwise ops on the ROCm platform

view details

Deven Desai

commit sha d325b255ff7d0bf1ca04229880dffb0a37d52e2d

[ROCm] removing no_rocm tag from tests that are now passing on the ROCm platform Also disabling one subtest within the //tensorflow/core:constant_folding_test. That test requires GPU support for the "topK" to work as intended. ROCm platform currently does not have GPU support for the "topK" operator and hence the test fails on it.

view details

Deven Desai

commit sha 17b87f0b51ad290269f983a85b887ae838c2ebe2

[ROCm] disabling fast NHWC implementation for the ROCm platform.

view details

Deven Desai

commit sha 04fb568df083a1903dd4f061539b29b4a849fd18

[ROCm] Adding explicit error messages for DnnScratchAllocator failures

view details

Måns Nilsson

commit sha cd311a8656fd9827a8fd8abdfc99ea47fd0ead4e

Inherit CMSIS CCFLAGS when generating mbed project

view details

Peng Sun

commit sha 9e18cc421be4eebb4e68da3834775e7beaa69786

Add 16 bit activations support to kernel operator STRIDED_SLICE Enable kernel STRIDED_SLICE support for int16 activations. Add typed test for int16 reference kernel.

view details

Ashutosh Hathidara

commit sha 2ebe69ec8bc752194a56b2fee6b91f05897420e0

untracked files

view details

Ashutosh Hathidara

commit sha 856175bc3afaa07f6e41ad781962a69e82ca7648

Merge remote-tracking branch 'upstream/master'

view details

Ashutosh Hathidara

commit sha b779a4737a16b1b04a1180c4193187625047b101

stack dump detached

view details

Ashutosh Hathidara

commit sha d0d5632b8d00e7d0e6285a1637d10ac50fa266d0

Gradient doc changed

view details

Ashutosh Hathidara

commit sha 5b8aadb94633f5c36d892236ba249d0dd4325723

Example added

view details

jerryyin

commit sha 1f4186c64f76854fe26335729022b7dea4dec941

Disable test that invokes rocBlas TRSM

view details

Frederic Bastien

commit sha e14bd919f4190cda821f616c15952fdbb36242dc

[XLA] Better default SM for unknown SM.

view details

Haoliang Zhang

commit sha f54bb6f5578b931d79884302768996ba1073f685

[Fixes #29509] Add dilated convolution pass in MLIR converter. PiperOrigin-RevId: 292026134 Change-Id: Id4c3d344a605652a7b1f732ff938e5b5ee771643

view details

Yuanzhong Xu

commit sha a358ac2cced467f40f8b92a430c7a48f514f9f7e

[MLIR:TF] Define IteratorGetNext in TF dialect. PiperOrigin-RevId: 292028972 Change-Id: I3c1c9c92ba6d335ea414e7ab8462e596850d0527

view details

Chao Xie

commit sha 560f1d64dbddeecdf8732bbe297653e953f30c25

Bug fix: avoid use-of-uninitialized-value PiperOrigin-RevId: 292030002 Change-Id: If415631a59458f21b3d3a1ec1f69671fbc9ecbc7

view details

A. Unique TensorFlower

commit sha 0e884391beabe2fadb6398b1fc5f48a9662c333c

to fix bug in h5 saving in format Original PR: https://github.com/keras-team/keras/pull/13477 Fixes: https://github.com/tensorflow/tensorflow/issues/33888 b/143774288 PiperOrigin-RevId: 292030348 Change-Id: I68032d72cf2cfd167c6ed065bce7a22d1f136011

view details

Anna R

commit sha 8db452ea5db4672859d5a4c7723a44a572f33f78

Disable compiler/mlir/tensorflow/tests/mlir2graphdef:preserve-entry-func-names.mlir.test on mac since it is flaky. PiperOrigin-RevId: 292032687 Change-Id: I2ddd178aecec5ec9447394c8fb876fba07b2d285

view details

Scott Zhu

commit sha 77ae99f06d82ee8f2ffe049a41ed36f8c3698dbd

Update LSTM/GRU v2 to respect self.zero_output_for_mask. The zero_output_for_mask flag is used by default in bidirectional wrapper. PiperOrigin-RevId: 292032747 Change-Id: Iee2caa28e4044bf080524ae0d65531aca03e9440

view details

Berkin Ilbeyi

commit sha 7189185ec06226923a98c87c0cc9db790f75aa8a

[XLA] Change CopyStart return type to (target, source, context) This is to fix a memory corruption issue where the source buffer may be reused for another HLO. Having the source in the return type of CopyStart makes it explicit that the source buffer may still be in use until CopyDone. PiperOrigin-RevId: 292034096 Change-Id: Id5e0546100410eb28c50554122c166e081f885af

view details

push time in 25 days

create barnchbmzhao/saved-model-example

branch : master

created branch time in a month

created repositorybmzhao/saved-model-example

created time in a month

push eventbmzhao/tensorflow

Yasuhiro Matsumoto

commit sha 0c68146f48cf50f5ecfe06d592881a517e38d1f8

Fix build of tensorflow/lite/delegates/gpu on raspberry pi This change make be possible to build gpu delegate on raspberry pi. But this is for libegl1-mesa, libgles2-mesa not VideoCore.

view details

Yasuhiro Matsumoto

commit sha 82477a6470f4d5dccd8074026bb84abc432e3b7c

Move include header to cc

view details

Yasuhiro Matsumoto

commit sha 85a642ed61cc8f800259957207e36f4cd3d1c0ad

Merge branch 'master' into fix-gpu-gl-raspi

view details

Yasuhiro Matsumoto

commit sha 4aee772c58a5109e35ffeb1e6936030aaeda97e4

Fix build

view details

Yasuhiro Matsumoto

commit sha 5b078b8c1b82078156a1388caa5216770dcc921f

Remove needless code

view details

Yasuhiro Matsumoto

commit sha be763b379e90b2f80a40053c9cca89316280772e

Sort alphabetically

view details

Yasuhiro Matsumoto

commit sha 016d116d5a2589216d4aaae6bcf73c8e35bbcf1f

Merge branch 'master' into fix-gpu-gl-raspi

view details

Yasuhiro Matsumoto

commit sha a722cf7c2c2d3d0830706478bbd9923d54f052b0

Merge branch 'master' into fix-gpu-gl-raspi

view details

Yasuhiro Matsumoto

commit sha 2da14f573a0a0211c074516e35b9a093f8a27fef

Remove changes of undef Status

view details

Yasuhiro Matsumoto

commit sha f48d647db1692e1e48ec45bbefc801dcea25cb65

Merge branch 'master' into fix-gpu-gl-raspi

view details

Yasuhiro Matsumoto

commit sha e75317fd3c393b45236102c1903cd9ef16b6f434

Merge branch 'fix-gpu-gl-raspi' of github.com:mattn/tensorflow into fix-gpu-gl-raspi

view details

Yasuhiro Matsumoto

commit sha 4c1fd335992a6c3711c44a076dc58c95a43f3f94

Merge branch 'master' of git://github.com/tensorflow/tensorflow into fix-gpu-gl-raspi

view details

Yasuhiro Matsumoto

commit sha 9647a495a5ec22a7d837810cd92ae2cbf4fa9a36

Fix Makefile

view details

Yasuhiro Matsumoto

commit sha 6b369ec3d0d51510648006bfb99146079070858d

Merge branch 'fix-gpu-gl-raspi' of github.com:mattn/tensorflow into fix-gpu-gl-raspi

view details

Yasuhiro Matsumoto

commit sha 929a7d0a35bd3a88d896eec924b1932f11c8dedf

Cosmetic change

view details

Yasuhiro Matsumoto

commit sha d0745205babf9f25f839315722e39b97ff585568

Merge branch 'master' into fix-gpu-gl-raspi

view details

Yasuhiro Matsumoto

commit sha c8595482e737b2e478539ebb0fd005d02a623ed2

Merge master

view details

Yasuhiro Matsumoto

commit sha 4247d8e498d047dcf9a718638a931d68591d8f91

Fix Makefile

view details

Yasuhiro Matsumoto

commit sha ef68005a23c7d3cb4f5623c7e66c32b485284695

Merge branch 'master' into fix-gpu-gl-raspi

view details

archis

commit sha 6458a4f0573a952d36c42b57ff60a84e0a2ac639

Added dense_sparse_matmul() This commit adds tf.SparseTensor.dense_sparse_matmul(). It uses sparse_dense_matmul() and some matrix manipulation to get to the right answer. It has been tested against numpy's calculations of the same operations

view details

push time in a month

push eventbmzhao/pytorch

Alexander Melnikov

commit sha 95b451d386f25e45bccfb3dc2b3a3ee53db4c011

fixing test_tensorboard for py2 (#30298) Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/30298 This diff fixes test_tensorboard for python2: - proto serialization is different in py2 vs py3 (e.g. for bytes) -> simple string comparison will fail for test_pytorch_graph. Modified to make graph comparison field by field Reviewed By: J0Nreynolds Differential Revision: D18654691 fbshipit-source-id: fdbca32e9a7fc2ea70a040bb825eab8a48d0dfe4

view details

Pieter Noordhuis

commit sha a074080d5739a9e2bbe90ffa82aeb173e35bfdff

Mark `c10d::~NCCLUtils` as noexcept (#29118) Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/29118 It's never a good idea to throw from a destructor and per #28288 we can't use `std::make_shared` on a class with a `noexcept(false)` destructor. To fix this, we `abort` instead of throw from the `NCCLComm` destructor. Closes #28288. ghstack-source-id: 93182910 Test Plan: ProcessGroupNCCLErrorsTest runs successfully. Reviewed By: pritamdamania87 Differential Revision: D18298271 fbshipit-source-id: ccac37753fef64fb63cb304433f4f97dc5621379

view details

Pieter Noordhuis

commit sha 29887f813a8224ec415facd8e77c16bfbd10e231

Remove unused forward declaration (#30154) Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/30154 This doesn't seem to be used in thread_pool.cpp. ghstack-source-id: 94264158 Test Plan: Let's see if this compiles. Differential Revision: D18614141 fbshipit-source-id: c6ff3db56b55fcee7d8123d909ee275690163ece

view details

Luke Yeager

commit sha 183aa1534f9e199e1e67e453dfb94dc855dabc0d

Add --no_python flag (#29144) Summary: Allows you to use a bash script wrapper in-between launch and your training script. e.g. ``` python -m torch.distributed.launch --nproc_per_node=8 --no_python --use_env \ bash -c 'exec numactl --cpunodebind=$(( LOCAL_RANK / 4 )) "$@"' -- \ python train.py ... ``` Pull Request resolved: https://github.com/pytorch/pytorch/pull/29144 Differential Revision: D18345647 Pulled By: pietern fbshipit-source-id: f05849c38c82de782988d07d300e00cf9f37253a

view details

Gerard Goossen

commit sha faacbfa8bf073988706113c0a3bcfb39207f6201

Migrate index_add cpu from TH to ATen (#28421) Summary: Migrate index_add cpu from TH to ATen. I couldn't find replacement for get1d and set1d, so doing pointer arithmetic inplace. Pull Request resolved: https://github.com/pytorch/pytorch/pull/28421 Test Plan: existing tests Differential Revision: D18060971 Pulled By: ggoossen fbshipit-source-id: 413719990cdb2fe578964cde14e93577e48a4342

view details

Hong Xu

commit sha 3455231e9cd67e03fe86262d8fe19432ca4789a7

Expose configuration of Numa directories to setup.py (#30104) Summary: https://github.com/pytorch/pytorch/issues/29968 Pull Request resolved: https://github.com/pytorch/pytorch/pull/30104 Differential Revision: D18656882 Pulled By: ezyang fbshipit-source-id: f932a98674033f1a3184dc1c22faa6f8c2b50134

view details

Rohan Varma

commit sha f41422121edd35be18b15f643f7a6bd568f283de

default construct rpc agent options based on the backend type (#30201) Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/30201 Provide a default constructor so that users don't have to construct RPC agent options. Also rename this to RPCBackend Options as suggested. ghstack-source-id: 94411768 Test Plan: Unit tests pass. Differential Revision: D18628698 fbshipit-source-id: 81fb45f124ad1006e628f6045162308093c9d446

view details

Martin Yuan

commit sha 559b3b5a7a606f65aed4d7d3b969b53a29fdbba7

Use unboxed registration for most of operators used in lite interpreter. (#30239) Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/30239 Use unboxed registration per smessmer 's request. For some ops with optional arg or tensor list that unboxed registration are not supported, we still use boxed. Test Plan: Imported from OSS Differential Revision: D18653846 Pulled By: iseeyuan fbshipit-source-id: c22ce8111dfff0ba63316a9bcfe2b712b2d31fc1

view details

Rohan Varma

commit sha c478a92b93e3c6aa46c610d4b894e6fcde0300ea

Add local shutdown to process group agent (#30020) Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/30020 This is now possible due to previous changes made in `gloo` and `ProcessGroupGloo`. We `abort` the listener thread that is waiting for a message, and join all other threads. The destructor calls this same `localShutdown` method, but we ensure this is not called multiple times. ghstack-source-id: 94415336 Test Plan: Unit tests pass. Differential Revision: D5578006 fbshipit-source-id: 6258879fb44c9fca97fdfad64468c1488c16ac02

view details

Xintao Chen

commit sha 5d7b2089e80a72a18cc02969a59d7a23cdd23843

Draft version: Make AliasAnalysisKind optional in Op Registration API (#30187) Summary: Don't look into deep into the diff's implementation. The reason to send out this diff is to help sync on the design first. Once we agree on the design, I will update the implementation accordingly. **Here is the basic design for achieving this functionality:** **Q1: Do we need to tell apart case between the following:** case 1: registry 1: PURE -> registry 2: CONSERVATIVE case 2: registry 1: PURE -> registry 2: <not set> A: should be yes though, right now both cases have same value(due to defaulting to CONSERVATIVE) in operators_ and operatorLookupTable_. case 1 should be denied while case 2 should be legal case where registry 1 will be PURE at the end. **How to tell apart both cases:** Right now, AliasAnalysisKind::CONSERVATIVE is by default (code pointer: https://our.intern.facebook.com/intern/diffusion/FBS/browse/master/fbcode/caffe2/aten/src/ATen/core/dispatch/OperatorOptions.h?lines=22%2C52) Current approach: Introducing a boolean flag in OperatorOptions called isDefault, defaulting to value true. When manually call setAliasAnalysis(AliasAnalysisKind), it will be set too false. And then when findSchema() in Dispatcher.cpp, we will check response's option's isDefault value. If isDefault = true, then with some sanity check and if all checks passed, we can update the option info in both operators_ and operatorLookupTable_ Other approaches: 1. Introducing a new AliasAnalaysisKind maybe called NOT_SPECIFIED. (I am not doing it this way since then I need to update other callosities related to AliasAnalaysisKind::CONSERVATIVE) Also, we will need to have additional logics to align between NOT_SPECIFIED and CONSERVATIVE **What data to be updated:** corresponding entry in std::list<OperatorDef> operators_ and LeftRight<ska::flat_hash_map<OperatorName, OperatorHandle>> operatorLookupTable_ (More things to be discussed here.) **Do we need to trigger listeners if an entry get updated:** I think no. callOnOperatorRegistered(op) seems only to be using OperatorHandle.schema now from the only callsite from register_c10_ops.cpp (code pointers: https://our.intern.facebook.com/intern/diffusion/FBS/browse/master/fbcode/caffe2/aten/src/ATen/core/dispatch/Dispatcher.cpp?commit=b4cefeaa98dca5b1ec5f7a0bca6028e368960244&lines=87-90 and https://our.intern.facebook.com/intern/diffusion/FBS/browse/master/fbcode/caffe2/torch/csrc/jit/register_c10_ops.cpp?lines=178&link_ref=biggrep) However, things can be much more complicated if future extensions may use options when some listeners want to use options value to register operators. **Future reading list + remaining questions:** 1. How options get consumed on the other side. 2. Usages for fields in OperatorEntry besides schema/options/kernals Pull Request resolved: https://github.com/pytorch/pytorch/pull/30187 Test Plan: [xintchen@devvm6308.prn2 ~/fbsource/fbcode] buck test mode/dev //caffe2:ATen-core-test All tests passed Differential Revision: D18530964 Pulled By: charliechen0401 fbshipit-source-id: 60c0560a63a36e54f09f397667bb7122b61d6a8e

view details

Christian Puhrsch

commit sha 7903fb118fdfc31e531e6b763e33a62677f9ee44

Move qkv_same, kv_same into branch (#30142) Summary: Perf improvements to multi_head_attention_forward - qkv_same and kv_same were not used outside of that branch. Further, kv_same was calculated even though it is not used if qkv_same Pull Request resolved: https://github.com/pytorch/pytorch/pull/30142 Differential Revision: D18610938 Pulled By: cpuhrsch fbshipit-source-id: 19b7456f20aef90032b0f42d7da8c8a2d5563ee3

view details

Tao Xu

commit sha fa242246ee8903d8e0311d18e4e11e9349193c21

add unit tests to iOS CI jobs (#30133) Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/30133 ### Summary Recently we've found that the master branch was constantly broken due to some unwanted change being landed on mobile. The problem is that our CI was not able to detect the runtime errors. ### Previous work - Add an unit test target to the iOS TestApp ( #29962 ) - Update Fastlane to run tests ( #29963 ) ### What's been changed in CI 1. XCode version has been updated to 11.2.1 2. For iOS simulator build, we'll run some unit tests( currently only one) after the build test. Test Plan: Imported from OSS Differential Revision: D18641413 Pulled By: xta0 fbshipit-source-id: 12942206f1dee045b2addba3ae618760e992752c

view details

Shen Li

commit sha a9f3f48f88b71d19d996e8836dfbe13ff85775b1

Revert D5578006: Add local shutdown to process group agent Test Plan: revert-hammer Differential Revision: D5578006 Original commit changeset: 6258879fb44c fbshipit-source-id: 11b893b3a280a8383eeb20a0548626811616dca1

view details

James Reed

commit sha 65f465050b989111b9aeed02f09e7a38c43c2462

Dont use SubgraphRewriter in FoldQuantizeCallIntoBuffer Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/30264 Test Plan: Imported from OSS Differential Revision: D18645531 Pulled By: jamesr66a fbshipit-source-id: 44fc0f0a3c8cabe62924baae0d556e43bbf637ec

view details

James Reed

commit sha 1cc321deed43ed246f0de6b393184507d0307b45

Memoize parseIR calls in graph mode quantization Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/30188 Test Plan: Imported from OSS Differential Revision: D18625743 Pulled By: jamesr66a fbshipit-source-id: 88f9da8e79324ba91e3550a8fc1a05e85bb83a86

view details

James Reed

commit sha 97fae401f0fef1ad15ef6d857c5e5422d128b8ac

Use LinearPackedParams everywhere Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/30198 Test Plan: Imported from OSS Differential Revision: D18628003 Pulled By: jamesr66a fbshipit-source-id: 76ff0248fd859e805a15cde555d26dd2138636fa

view details

Tao Xu

commit sha 638f4c1fb31b33a5040f5105da58c88f6e67b887

Update Cocoapods to 1.4.0 (#30326) Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/30326 Note that this PR won't trigger the cocoapods build. We'll push the binary and release the cocoapods after the branch cut. Test Plan: Imported from OSS Differential Revision: D18660308 Pulled By: xta0 fbshipit-source-id: 95dd97b7b67e70ecee3a65d8bbc125791872b7ca

view details

Brian Vaughan

commit sha a5272cb6433efe0bdaf76bff428ce961538daa25

Error instead of assertion failure for div by sparse (#30260) Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/30260 fixes: https://github.com/pytorch/pytorch/issues/30044 Without this PR, ``` >>> torch.tensor(1.) / torch.tensor(1.).to_sparse() Traceback (most recent call last): File "<stdin>", line 1, in <module> RuntimeError: r.is_sparse() INTERNAL ASSERT FAILED at /Users/distiller/project/conda/conda-bld/pytorch_1570710797334/work/aten/src/ATen/native/sparse/SparseTensorMath.cpp:168, please report a bug to PyTorch. ``` Test Plan: Ran the same code with this change: ``` In [1]: import torch In [2]: torch.tensor(1).to_sparse() / torch.tensor(1).to_sparse() --------------------------------------------------------------------------- RuntimeError Traceback (most recent call last) <ipython-input-2-7177f54f30bb> in <module> ----> 1 torch.tensor(1).to_sparse() / torch.tensor(1).to_sparse() RuntimeError: Unsupported tensor layout ``` Differential Revision: D18657387 Pulled By: nairbv fbshipit-source-id: cd23570d46f5b26fd84049e5e63b61b19835603d

view details

David Reiss

commit sha 35e6c1763e1870a679d79c81f7d6019813b61cc8

Switch Docker image onda-cuda-cxx11-ubuntu1604 to new uniform name (#29943) Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/29943 This was apparently the same as "pytorch/pytorch-binary-docker-image-ubuntu16.04:latest", so standardize on that name. Test Plan: This PR, which is stacked on top of a commit that puts one of the jobs using that container into the set of PR builds. Imported from OSS Differential Revision: D18653554 fbshipit-source-id: 40e6c52db02265d61e8166bb1211376faccfc53a

view details

Chunli Fu

commit sha 0c18de2623f243987e431f864e85b176e4d2766f

Add inferBoundShapeOp Summary: Pull Request resolved: https://github.com/pytorch/pytorch/pull/30101 Reviewed By: ipiszy Differential Revision: D18387803 fbshipit-source-id: 5edb6b949257370b62fa6da477bd6ed2f16a9bd1

view details

push time in a month

PR opened gunan/pipe

Fixed Garbage Printing & Indefinite Blocking on ReadFile Caused by Leaked Handle
  1. Parent Process was printing entire char buffer, instead of char buffer up to the amount of characters read, producing garbage.
  2. Parent process leaked the child stdout handle, causing indefinited blocking on ReadFile.

See https://stackoverflow.com/a/54416516, specifically

Yes, it should be added that output_pipes[1] must be closed after CreateProcess() call and before any ReadFile() call. 

in the expanded answer comments.

+8 -7

0 comment

1 changed file

pr created time in a month

push eventbmzhao/pipe

bmzhao

commit sha 69813412e53b2774c8a27bd0545a398b73003a9f

Fixed 2 issues: 1. Parent Process was printing entire char buffer, instead of char buffer up to the amount of characters read, producing garbage. 2. Parent process leaked the child stdout handle, causing indefinited blocking on ReadFile.

view details

push time in a month

fork bmzhao/pipe

Reproduce the pipe not unblocking.

fork in a month

fork bmzhao/rules_swift

Bazel rules to build Swift on Apple and Linux platforms

fork in a month

push eventbmzhao/tensorflow

Duncan Riach

commit sha 800c3136534597dabf664db911e24255b89a76ed

Remove duplicated name in 2.0.0 release note thanks section

view details

William D. Irons

commit sha bf00bd654adc0bbb6ccc73a8b729e9f1d0f6037c

Fix saved_model_cli tensorrt conversion The existing saved_model_cli convert tensorrt script fails in 2.X with module not found "tensorflow.contrib". Updated the script to use the V2 API for TensorRT to convert a saved_model. The max_batch_size and is_dynamic_op parameters are not valid for the V2 API so they have been removed.

view details

Deven Desai

commit sha 9e7eae9f71855efe83287977e1844806675adaee

add hipsparse to the local_config_rocm repository

view details

Deven Desai

commit sha f725b464549eab744148ad940e04060cbaa7ae90

renaming CUDA names in cuda_sparse.h to generic GPU names

view details

Deven Desai

commit sha 7e8ccbd22be53cade35de31631a8ada0bccfbac5

Adding ROCm support for the GpuSparse API (TF wrapper for cuSPARSE/hipSPARSE)

view details

Deven Desai

commit sha 5ad7620d6f18f4a3c123fb7f365f0cb20dda2760

Skippping failing subtests within the CSR Sparse Matrix unit-tests. The failures are because either * the subtests require support for complex type (which is not yet supported by ROCm) * or they require a GPU kernel implementation for the SparseMatrixAdd op (which is also not supported by ROCm, because the underlying hipSPARSE API routine - csrgeam - does not exist). There are also a couple of subtests commented out because hipSPARSE API errors out with an unknown error for them. Those will be looked into and fixed soon

view details

Deven Desai

commit sha 2e1cdaa4b62103d1d6f2e18845bbc2c69ffc117b

Adding ROCm support for the CSR Sparse Matrix Ops

view details

Deven Desai

commit sha e762347e79f10f0ee3a730385f4959808ec2fb1e

removing no_rocm tag from the CSR Sparse Matrix unit tests

view details

Deven Desai

commit sha 5d1ccc1eeeebd527427ff02c24b7a967861e2868

addressing code-review comments

view details

Dominic Jack

commit sha 633a14cbdc48d3a225a562a22b70f104d839cdc2

Added ragged option for is_keras_tensor

view details

Dominic Jack

commit sha a1d48b6b26a043e1be5896105a3fb83a998f3dca

added missing sparse test

view details

Dominic Jack

commit sha 732e2b91ab70ce6fbb2fd5fc65b5ebae06c6f881

Merge branch 'master' into keras-backend-ragged

view details

Mrinal Jain

commit sha 80e33b39e28b5fedad8d59dafc442467fa1d82d2

Added usage example for tf.keras.callbacks.TensorBoard

view details

William D. Irons

commit sha c639437db16afc455b77a8075fda2f15b6134fe9

Fix hwloc build for ppc64le This commit: https://github.com/tensorflow/tensorflow/commit/41df105#diff-6fb2e55075204b47da0460ea2abbc32f broke the CPU unit test build for ppc64le. The compiler error was: .../libexternal_Shwloc_Slibhwloc.so: error: undefined reference to 'hwloc_linux_component' .../libexternal_Shwloc_Slibhwloc.so: error: undefined reference to 'hwloc_linuxio_component' These methods are defined in topology-linux.c, adding the necessary bazel select statement so they are built during a ppc64le build.

view details

Maher Jendoubi

commit sha 6539d343231dfea317a7a09597ed178e32e2fd36

Contributing: Fix a typo

view details

Maher Jendoubi

commit sha ffc474c4a30951a4023153acba78e4d263c636ad

Contributing: Fix a typo

view details

Maher Jendoubi

commit sha 0978e4bdf7a26bb2e7735b92a05584b9cec42bb3

betweeen --> between

view details

Mrinal Jain

commit sha 1872886aef3ea7798209db713aa3104ba29d9e99

fixed formatting error

view details

Mrinal Jain

commit sha ed9ef9bc782b2e4cca0f02e31ae50dc0c5739cbf

changed doctest to usage example

view details

Jens Elofsson

commit sha 98292c11a8748c3bd371b1a2a68b145646e37269

Micro: Fix compile error for Arm Mbed OS.

view details

push time in a month

startedautodiff/autodiff

started time in a month

issue commentbazelbuild/bazel

Bazel Query in Bazel 1.2.1 errors out due to bazel-internal code

@meteorcloudy I'm currently working on getting TF building with bazel 2.0, as part of the the cc_shared_library work! I'm still looking into on a few build failures, but once that's landed, it should make it easy to integrate with the bazel 2.0.1 patch release.

bmzhao

comment created time in a month

push eventbmzhao/bazel-cc-so-example

Brian Zhao

commit sha 0237de74779fd6477aeb75e5fb450aab1ed7aad1

Still failing after absl patch

view details

push time in a month

push eventbmzhao/tensorflow

George Sterpu

commit sha 4ab6a520c94441622442747aef620939cc1d8130

Relax the check for state_size The behaviour of `hasattr` is to evaluate the state_size member. In the case of `tfa.seq2seq.AttentionWrapper`, that is a @property member that is built at graph runtime after calling `setup_memory`, thus `hasattr` returns an error when using AttentionWrapper with dynamic memories. More details: https://github.com/tensorflow/addons/issues/680

view details

boron

commit sha 65fa0c2c2a9b4f799f9d0f5896cdb414a67695ab

Update readers.py

view details

boron

commit sha 102ed3d242b1ccc893e7714ee1edc007b04b392a

Update readers.py

view details

boron

commit sha 29ffefec09b5f4bcd6d23b2fe40399df4414ab95

Update readers.py

view details

boron

commit sha 249751ace44e531640016fec24a80226d85a6a4d

Merge pull request #1 from boronhub/gci-task Gci task

view details

boron

commit sha efae390e08414ae0c01371d2d935457ff3e378ca

Add usage example to tf.nn.conv2d()

view details

boron

commit sha acfada322a850ce5bf6ef77c0ea711b81211cc82

Update readers.py

view details

boron

commit sha 1f2df263c7217c46052f54a71e51c724a42a378e

Update nn_ops.py

view details

Lukas Geiger

commit sha bb23a5963df47db53966cb89c79187c9102daa37

Simplify tf.keras.backend.bias_add

view details

boron

commit sha 628c71a7cc2c781dc3f0ba913ef7d4ad73738ecc

Update nn_ops.py

view details

Mihai Maruseac

commit sha a5100509131c2700c19e4b09e1e2033526aeefc4

Merge branch 'master' into nn-ops

view details

boron

commit sha 76dca033bc9e17ba5f74c22a301d513afa4d5790

Update nn_ops.py

view details

boron

commit sha 85428623e40007f1ff7486171f4f3189748bdb10

Update nn_ops.py

view details

George Sterpu

commit sha c678bdb3ae128974ddcc06bc02c5ae5f0de65e24

Update recurrent.py trying to edit directly from the browser

view details

George Sterpu

commit sha 409db98338a62808209ab3837f6ca3b796c81dc5

Update recurrent.py

view details

boron

commit sha fdad31831e108d4626a960b5862e85f78c61c4ff

Update nn_ops.py

view details

A. Unique TensorFlower

commit sha f6df725f25222edb23de344a2fc41c35b40c1a41

Go: Update generated wrapper functions for TensorFlow ops. PiperOrigin-RevId: 289290064 Change-Id: I97d2bb488c5bf19acabe52cfa9c319f67236187a

view details

Tiezhen WANG

commit sha 568d3fc152dc557f65ddd0b78b8a8543c5c1e2f1

TFL & TFLM: Add new kernel memory planning API. This has the same amount of methods in context so that size of context object won't be changed. Also the methods being touched is not used anywhere, so this should be safe. PiperOrigin-RevId: 289292739 Change-Id: I916354ab4f0749887ed169c042202ae9650799fe

view details

Scott Wegner

commit sha b8f3b2481ffe60f69ee385ee98659c70be8b02f8

Fix documentation typo: dhape -> shape PiperOrigin-RevId: 289293715 Change-Id: I4486f336b49a11bdebe35270982d4f8cffa2e02d

view details

A. Unique TensorFlower

commit sha 556a932ff24686d9e3c458a4ee5bd5c7b4825b32

compat: Update forward compatibility horizon to 2020-01-12 PiperOrigin-RevId: 289304019 Change-Id: Ic5991d4eb88d6cb1701f5d11b332555621056dce

view details

push time in a month

push eventbmzhao/bazel-cc-so-example

Brian Zhao

commit sha 8c2c6acbe58567496cb7002cf3f52416a4f58bbe

Still failing after absl patch

view details

push time in a month

push eventbmzhao/bazel-cc-so-example

Brian Zhao

commit sha 75b9d0a353ac0e515b3e25fdd1448fe8c9b42d5d

Still failing after absl patch

view details

push time in a month

push eventbmzhao/bazel-cc-so-example

Brian Zhao

commit sha dde8ec8ed15d83ee9c3744f9ef56e211598ba8d9

Still failing after absl patch

view details

push time in a month

startedcrosetto/cupq

started time in a month

push eventbmzhao/tensorflow

Deven Desai

commit sha 1ad0ff755e2dddcc37d9b57e271642fd4d1d405d

[ROCm] Updating the ROCm stream_executor implementation to use the MIOpen Immediate Mode API

view details

Deven Desai

commit sha f5b5f3d22dfea28cd62566ed7de67d5bc4640309

[ROCm] Enabling ROCm support for code in gpu_util.cc

view details

Deven Desai

commit sha 80c49615ee4501c40efa0b5e2036c73dd1f1e65e

[ROCm] Updating the ROCm convolution kernels to use the MIOpen Immediate Mode API

view details

Deven Desai

commit sha 81ab633a4934c3e1f673e0abbfb229d7f3c1d029

[ROCm] Updating the ROCm XLA convolution kernels to use the MIOpen Immediate Mode API

view details

Anjali Sridhar

commit sha b7a6d319bb7435ad5ea073b09cd60deddc47c14b

Add CompositeTensor support for DistributedIterator. PiperOrigin-RevId: 289037261 Change-Id: Iec14cd66bcad37070c19a3a2f3bb0cc524e79ce9

view details

A. Unique TensorFlower

commit sha cf244f6db0f7b05e907181540ce2d7172ad4c42f

compat: Update forward compatibility horizon to 2020-01-10 PiperOrigin-RevId: 289049425 Change-Id: Idf76fe43f4b4313d4ccfed71cddd3bd15fe7c935

view details

TensorFlower Gardener

commit sha c1971ab97c79f93b8901e11409556b76d99c114c

Merge pull request #35503 from ROCmSoftwarePlatform:google_upstream_rocm_miopen_immediate_mode PiperOrigin-RevId: 289053613 Change-Id: I233d95adc3aa888460bd39a07fd7e168fea14846

view details

Eyvind Niklasson

commit sha f149e18303c0d0f73dd04448c70048b471778e53

fix for tf.recompute_grad breaking when wrapped function is called with multiple inputs of the same tensor. PiperOrigin-RevId: 289054375 Change-Id: Ia9a5110169f8a5486fcccc2c44c20c503dccadf1

view details

A. Unique TensorFlower

commit sha 9269863847020f0c5f9f971bfdcb3e008727852c

Go: Update generated wrapper functions for TensorFlow ops. PiperOrigin-RevId: 289060755 Change-Id: I1b48821f41bf65879454a92565683334b5ed762f

view details

Andrei Kulik

commit sha 56e97f8b9007bf98142ae644faa92b08402bc1c6

Use device before it is moved into environment. PiperOrigin-RevId: 289066572 Change-Id: I3321d85cb10bb6622935bf78324be297dcdf4c6a

view details

A. Unique TensorFlower

commit sha e4fc2f4c543719dfeff1b6bae50800918f894a46

Explicitly export files needed by other packages PiperOrigin-RevId: 289067462 Change-Id: I17a06222b59691b7c447249b8ae5236b07eba72f

view details

Andrei Kulik

commit sha 0d4d2d6593ecee46bceca6b76f632f47d99becd5

Re-create graph if OpenCL failed and falling-back to OpenGL. PiperOrigin-RevId: 289067574 Change-Id: Id1260ca0f5d3e5ad19afe552c2d4877e41ca07ef

view details

A. Unique TensorFlower

commit sha b6d83da696d7463affc0f2fb1f211799b18025ad

Explicitly export files needed by other packages PiperOrigin-RevId: 289068233 Change-Id: Iad295a519968341f3765116f5f3c6508efd51d24

view details

Andrei Kulik

commit sha c7fccf43b4b30859a6936f07619f5f63552416ed

Always create OpenCL profiling queue for workgroup-size tuning. PiperOrigin-RevId: 289070322 Change-Id: Ib3c293f8de27263fedffc2238defca663f3befb4

view details

Srinivas Vasudevan

commit sha 8b905b92f173484bc04b88ddb303c328f56f2943

Add __matmul__ override for using @ for LinearOperator matrix multiplication (e.g. A @ B). PiperOrigin-RevId: 289071519 Change-Id: I9dc4c13bbf4843e6c7eb471dc748176262392894

view details

A. Unique TensorFlower

commit sha 8452eb3059c488f184acc9f56e51bcf9f83ba117

Go: Update generated wrapper functions for TensorFlow ops. PiperOrigin-RevId: 289071640 Change-Id: I73c17d3f30cd76a71d09859337a6a41aff899380

view details

Benjamin Kramer

commit sha df0329bafdf8b3547d232681f7ec50e5639cc9f3

Bump open source llvm revision to 498856fca5b9306f545554aeec93c7c058f03eb3 PiperOrigin-RevId: 289073594 Change-Id: I39910a882ae3f06108c7da7d7aa01cbe52c445d6

view details

Gunhan Gulsoy

commit sha a926fe01a734a4075c0605e313bdcc33211e0581

Disable 4 doctests on windows. The tests require unix line ending and filepath separators. PiperOrigin-RevId: 289078872 Change-Id: Icfd9e621f7fd011aeb63b84d38ef3554afa9c840

view details

A. Unique TensorFlower

commit sha 7f216cffbadb55a0e310c51692c71b2add261bae

Go: Update generated wrapper functions for TensorFlow ops. PiperOrigin-RevId: 289083641 Change-Id: I636848039f382f2ccef5c0052d39ff75aa60134f

view details

Benjamin Kramer

commit sha 5863bc687e00a34246a78006c52c445936747b22

[mlir] Remove unused variables. No functionality change. The compiler will start warning on them with an upcoming mlir change. PiperOrigin-RevId: 289096631 Change-Id: If48aabe5c8fd9c25a2a4673d72c6d5d1dc111cad

view details

push time in a month

create barnchbmzhao/bazel-cc-so-example

branch : with-absl-patch

created branch time in a month

push eventbmzhao/abseil-cpp

Abseil Team

commit sha e6b050212c859fbaf67abac76105da10ec348274

Export of internal Abseil changes. -- c2e2e2b21c3fe59b63279e7418c93c7289ee3e27 by Mark Barolak <mbar@google.com>: Import of CCTZ from GitHub. PiperOrigin-RevId: 256220326 -- 3996b01f0c3eb60f72825b154dce8019b6215f1d by Derek Mauro <dmauro@google.com>: Add GCC 4.9 test script. This will become our new minumum version and GCC 4.8 will be removed soon. PiperOrigin-RevId: 256160891 -- 2c13aa44bc8e497ebae9abe8b8adf73c2152622d by Abseil Team <absl-team@google.com>: [Trivial] tweak flat_hash_map.h doc comment The comment is probably right both ways, but the lack of an "if" here seemed jarring to me. PiperOrigin-RevId: 256089069 -- 16bc03b9b30fbf08d7dc61025fa8ec4b57077be8 by Abseil Team <absl-team@google.com>: Fix symbolization for elf files whose SYMTAB section is stripped, but have a DYNSYM section. Previously, if we did not find a SYMTAB section, we would bail out of symbolization early, rather than checking the DYNSYM section. PiperOrigin-RevId: 256061954 -- 4c60ee329b1eeb0b0d10c4f76f282e5fbae2c5b2 by Derek Mauro <dmauro@google.com>: Update to LLVM r363242 and Bazel 0.27.0 PiperOrigin-RevId: 256024405 -- 18e1ba970d33f122026803d8ca90035b9088949d by Eric Fiselier <ericwf@google.com>: Disable variant tests that break with P0608R3 PiperOrigin-RevId: 255975764 -- 0a89858464977e86096b62476faa3b64eb94aa1d by Abseil Team <absl-team@google.com>: Internal change PiperOrigin-RevId: 255891019 -- 9b7424cac66f5407f0ed74ed288bf3099a4fa330 by CJ Johnson <johnsoncj@google.com>: Updates the implementation of InlinedVector::insert(...) to be exception safe and adds an exception safety tests for insert(...) PiperOrigin-RevId: 255881809 -- 1288f4ba3883c510d92b09437fb8b051c19aa241 by CJ Johnson <johnsoncj@google.com>: Updates the implementation of InlinedVector::insert(...) to be exception safe and adds an exception safety tests for insert(...) PiperOrigin-RevId: 255875277 -- 39c04f3621491eaff9e2eda619718d5b5f20fbd9 by Abseil Team <absl-team@google.com>: Use a typedef to allow building with NVCC Without this change NVCC fails to compile compressed_tuple.h. NVCC is relevant because TensorFlow uses NVCC on Ubuntu and inclues abseil. PiperOrigin-RevId: 255850176 -- e23f0309ccad69eb508ca02c9034cd4cdd740da0 by Abseil Team <absl-team@google.com>: Internal change PiperOrigin-RevId: 255787167 -- 054aafcebf595077054164f1da3703124ab209b4 by Abseil Team <absl-team@google.com>: Updates the ScopedAllocatorWorks test for InlinedVector to not rely on the byte count allocated by the standard library In doing so, removes LegacyNextCapacityFrom(...) impl function from InlinedVector Also applies clang-format to the test file PiperOrigin-RevId: 255760356 -- eb05fc9f78e3a163c93f1866e9fe9a8ad0d01622 by Abseil Team <absl-team@google.com>: Internal change PiperOrigin-RevId: 255706834 -- 97abb824417604c45d9fcbb3e4ff1aa3000836f2 by Jorg Brown <jorg@google.com>: Enhance compatibility of abseil's strings package with nvcc. PiperOrigin-RevId: 255688500 -- efc5b9c221ee31e15d10b35d31c8f3ae6eddaa8c by Abseil Team <absl-team@google.com>: Follow CCTZ's lead and allow GetWeekday() and GetYearDay() to be called with any civil-time type. A CivilSecond, for example, has a weekday just as much as a CivilDay does. PiperOrigin-RevId: 255659840 -- a75acbe954457919d8c6c8f4c2339b543760b375 by Derek Mauro <dmauro@google.com>: Increase the timeout of randen_engine_test. It seems to timeout under TSAN often enough to justify the increase. PiperOrigin-RevId: 255628086 -- 160976ba47c7c6eb57af08e21f8eb640aa51e91b by Derek Mauro <dmauro@google.com>: Update CMake documentation Fixes https://github.com/abseil/abseil-cpp/issues/332 PiperOrigin-RevId: 255607313 GitOrigin-RevId: c2e2e2b21c3fe59b63279e7418c93c7289ee3e27 Change-Id: Iba4ac7ed23cbcdb22965b4958601f689be92cda4

view details

Abseil Team

commit sha 74d91756c11bc22f9b0108b94da9326f7f9e376f

Export of internal Abseil changes. -- a874475e842d2adeb31bb7bd37bdd6eb15a2aeb9 by Mark Barolak <mbar@google.com>: Import of CCTZ from GitHub. PiperOrigin-RevId: 256414250 -- c95e6c21944c59c9b9b9e7eb9dc79cfb9ae5ef8d by CJ Johnson <johnsoncj@google.com>: Update the license year + run clang-format for the FixedArray and InlinedVector test files PiperOrigin-RevId: 256376285 -- f430b04f332d6b89cb8447b07217e391e1c38000 by Derek Mauro <dmauro@google.com>: Migrate the Linux CMake tests from GCC 4.8 to the GCC latest version. This will allow us to delete the GCC 4.8 test since that is currently our only CMake coverage. This also means that we don't have to update the script every time we move to a new minumum GCC version. This change includes a fix for a -Wstringops-truncation warning in symbolize_test.cc that triggers when it is built in release mode with the latest GCC. PiperOrigin-RevId: 256370092 GitOrigin-RevId: a874475e842d2adeb31bb7bd37bdd6eb15a2aeb9 Change-Id: Ia2ec58f9b9dfc382d043344e346cb397b802270a

view details

Joe Sylve

commit sha 5fc0bf69e7b78d35e6d8621db994f7f307a83d59

Fixed ABSL_INTERNAL_MACOS_CXX17_TYPES_UNAVAILABLE check

view details

Matt Calabrese

commit sha 3c98fcc0461bd2a4b9c149d4748a7373a225cf4b

Merge pull request #340 from jtsylve/macos_cxx17_fix Fixed ABSL_INTERNAL_MACOS_CXX17_TYPES_UNAVAILABLE check

view details

Abseil Team

commit sha 44efe96dfca674a17b45ca53fc77fb69f1e29bf4

Export of internal Abseil changes. -- 9c4ef32276054fba6a116c01cd4b3fd278f59ece by Andy Soffer <asoffer@google.com>: Remove support for unused arbitrary-width output in FastUniformBits. Width should be inferred from the requested return UIntType. PiperOrigin-RevId: 257189319 -- e3326329d02171a301cc3d6ae617ed448472b728 by Abseil Team <absl-team@google.com>: Update comments to make clear that absl::Format(std::string *, ...) appends to the provided string. PiperOrigin-RevId: 257058043 -- e2096b06d714fba3ea2c885d670a42efd872765c by Xiaoyi Zhang <zhangxy@google.com>: Fix compilation error on MSVC 2017. The root cause seems to be a compiler bug in VS 2017 about pack expansion with multiple parameter packs, specifically `MakeVisitationMatrixImpl::Run` is triggering compiler error "error C3528: 'BoundIndices': the number of elements in this pack expansion does not match the number of elements in 'EndIndices'". Work around this issue by using only one parameter pack `CurrIndices` in `MakeVisitationMatrixImpl::Run`. PiperOrigin-RevId: 257040381 -- 9ab75ff27b2513583fffc1233e6568aa96be36f7 by Matt Calabrese <calabrese@google.com>: Internal change. PiperOrigin-RevId: 257039041 GitOrigin-RevId: 9c4ef32276054fba6a116c01cd4b3fd278f59ece Change-Id: I5f708bb03aff93948502394a413260af2a8a273b

view details

Abseil Team

commit sha c6c3c1b498e4ee939b24be59cae29d59c3863be8

Export of internal Abseil changes. -- ed3a3431eee9e48e6553b0320e0308d2dde6725c by Derek Mauro <dmauro@google.com>: Project import generated by Copybara. PiperOrigin-RevId: 258631680 GitOrigin-RevId: ed3a3431eee9e48e6553b0320e0308d2dde6725c Change-Id: I1d7ae86a79783842092d29504605ba039c369603

view details

Abseil Team

commit sha 278b26058c036833a4f7f3047d3f4d9296527f87

Export of internal Abseil changes. -- b0015627610a311997ab21b8b295a8168b5915ec by Derek Mauro <dmauro@google.com>: Add an empty log_severity library. The log_severity sources will eventually migrate here once users are updated to depend on the new target. Note that no opensource user should be depending on log_severity.h today since it has no purpose in opensource. PiperOrigin-RevId: 258781702 -- c4d8d03f731d1de648ef2d17a55c3a16029ba331 by Derek Mauro <dmauro@google.com>: Move atomic_hook.h to its own internal target. PiperOrigin-RevId: 258765527 GitOrigin-RevId: b0015627610a311997ab21b8b295a8168b5915ec Change-Id: I0ec9b8b3cb6a27dc32a5f6c318d07e02548f193d

view details

Abseil Team

commit sha f3840bc5e33ce4932e35986cf3718450c6f02af2

Export of internal Abseil changes. -- dcff7cc4eb3837d39d1e083cbf1d4f082054cbf6 by Laramie Leavitt <lar@google.com>: Add default and fallback modes for SaltedSeedSeq::generate(). SeedSeq::generate will often be called with contiguous ranges of uint32_t. Detect this case and use the currently direct code paths to handle it. Otherwise use a fallback mechanism which does an additional copy. PiperOrigin-RevId: 258853656 -- 59ec88845ac2ee6567c493021b54385940d66762 by Gennadiy Rozental <rogeeff@google.com>: Internal change PiperOrigin-RevId: 258822137 -- eb2d1ab21f6f94c24c51c1cd3719846be259e8e7 by Abseil Team <absl-team@google.com>: Add iOS version 11 to the mix of Apple OS versions that have various C++17 header files but do not support them. PiperOrigin-RevId: 258820405 -- 452cea3e13b29797b2c385d7c6da7613d36bdc45 by Gennadiy Rozental <rogeeff@google.com>: Internal change PiperOrigin-RevId: 258802436 GitOrigin-RevId: dcff7cc4eb3837d39d1e083cbf1d4f082054cbf6 Change-Id: I2261cb58e142eb15017ef646a56710dd64f06496

view details

Abseil Team

commit sha ad1485c8986246b2ae9105e512738d0e97aec887

Export of internal Abseil changes. -- 1f44f8f487aa3afe8248132e4081519e85671965 by CJ Johnson <johnsoncj@google.com>: Updates ScopedAllocatorWorks test for InlinedVector to not depend on specific byte counts of standard library vectors. It's too brittle in the face of capacity-changing changes to InlinedVector and does not provide signal in those breakages. PiperOrigin-RevId: 259590332 -- fef7589547e9cdd04a254f6ae06e2bd9ec2b35f0 by CJ Johnson <johnsoncj@google.com>: Updates the implementation of InlinedVector::insert(...) to be exception safe and adds an exception safety tests for insert(...) PiperOrigin-RevId: 259542968 GitOrigin-RevId: 1f44f8f487aa3afe8248132e4081519e85671965 Change-Id: I514beff56159c9c717f8d29197728011af1fecd7

view details

Abseil Team

commit sha 36d37ab992038f52276ca66b9da80c1cf0f57dc2

Export of internal Abseil changes. -- 1224e58a45e4d016b18f5a6cf5762ba33027017a by CJ Johnson <johnsoncj@google.com>: Unifies the growth factor of InlinedVector's mutating members to max(2 * capacity, req_capacity). In doing so, LegacyNextCapacityFrom(...) is removed thus removing a loop from several callsites. PiperOrigin-RevId: 259920301 -- 945fc0bf27b67ea77d39144dcb6a483dc879ceda by Laramie Leavitt <lar@google.com>: Cleanup header guards which do not reflect the correct style. PiperOrigin-RevId: 259881520 -- 8c7d0532ba9a9aabfd57f67552572b2b1bedda97 by Derek Mauro <dmauro@google.com>: Move log_severity sources to the new log_severity target. PiperOrigin-RevId: 259837015 GitOrigin-RevId: 1224e58a45e4d016b18f5a6cf5762ba33027017a Change-Id: Id19506c3b8db71a0d4391ee917bfef3e802d550d

view details

Abseil Team

commit sha 52e88ee56b72cf32bc66534d942c7398ce481331

Export of internal Abseil changes. -- 788c948f698afe5998738cbb40b0629668640d73 by Derek Mauro <dmauro@google.com>: Add an empty raw_logging_internal library. The raw_logging sources will eventually migrate here. This target will only contain internal sources. Abseil compatibility guidelines require users not depend on internals. https://abseil.io/about/compatibility PiperOrigin-RevId: 260712817 -- 8ea947ec54f47fb02e97597d243cb63aa4c5fe0b by Abseil Team <absl-team@google.com>: Fixed an incorrect version number test for optional,etc. availability in iOS. Added tests for watchOS and tvOS. PiperOrigin-RevId: 260490390 -- 53af0544277dc3020b81d16ce110e7bc89f5cf6e by Jorg Brown <jorg@google.com>: Test the error-string path a bit more. PiperOrigin-RevId: 260175886 GitOrigin-RevId: 788c948f698afe5998738cbb40b0629668640d73 Change-Id: I366b73331857dc5a1db843b650c2ba27a69a141e

view details

Abseil Team

commit sha 14550beb3b7b97195e483fb74b5efb906395c31e

Export of internal Abseil changes. -- 8f685654a7d04eb8a0cb82d31e44e391e906b609 by Derek Mauro <dmauro@google.com>: Support constexpr construction of absl::string_view from a string literal in MSVC 2017+. Fixes https://github.com/abseil/abseil-cpp/issues/352 PiperOrigin-RevId: 260853160 -- a3c4c5168ce2a491134d7c87cf7fdc75d1ee2533 by Derek Mauro <dmauro@google.com>: Make SwissTable's at() throw when exceptions are enabled Fixes https://github.com/abseil/abseil-cpp/issues/355 PiperOrigin-RevId: 260788026 GitOrigin-RevId: 8f685654a7d04eb8a0cb82d31e44e391e906b609 Change-Id: I9ed498e181faa9c9d16e9b1b01404969d99b8ea9

view details

Abseil Team

commit sha c5c4db4f5191fe5e76cbf68dcc71fb28702f7d2b

Export of internal Abseil changes -- f8fe0f483378c7520d8f8bdfabe4b20de4d96c7e by Andy Soffer <asoffer@google.com>: Ensure that Invoke can support C++17 in the sense that noexcept is part of the type. PiperOrigin-RevId: 261730155 -- bf796ab71653a80498f9374bc8c5111d065c64ba by Abseil Team <absl-team@google.com>: Fix typo in static_assert message for copy/move constructible by replacing "by" with "be". PiperOrigin-RevId: 261713992 -- 8c7c17c40d03a322f304a2fd73ed34462dbf265a by Andy Soffer <asoffer@google.com>: Add absl::is_function drop-in replacement for std::is_function. Some standard libraries incorrectly implement std::is_function in a few corner cases. In particular, libstdc++ functions marked noexcept. The trick being used here is that function types decay to pointers. After excluding cases like is_class, etc, we can distinguish function types by testing for this decay. Many thanks to ericwf@ for essentially writing this CL. PiperOrigin-RevId: 261705008 -- c5adf42d0a132c2525d17a719329eab2ffe0aa94 by Abseil Team <absl-team@google.com>: Add microbenchmark for StrSplit that uses delimiter ByAnyChar. PiperOrigin-RevId: 261424010 -- 66a342f9381ec56be2fe3aa5b3193dd3538a9740 by Andy Soffer <asoffer@google.com>: CMake support for Abseil Random This change touches almost build-related files for CMake almost exclusively. The one minor exception is random/internal/salted_seed_seq.h. The default warnings configuration for one of our CI builds requests not having named but unused parameters. The change in this file cleans up that warning. PiperOrigin-RevId: 261192369 GitOrigin-RevId: f8fe0f483378c7520d8f8bdfabe4b20de4d96c7e Change-Id: I05f662baacfe78750651535aa658f61c2327bc44

view details

Abseil Team

commit sha 67222ffc4c83d918ce8395aa61769eeb77df4c4d

Export of internal Abseil changes -- 5315e7b98905922e779798f3168d98343438c134 by Derek Mauro <dmauro@google.com>: Fix absl::string_view::copy to throw std::out_of_range when pos > size(). Fixes https://github.com/abseil/abseil-cpp/issues/362 PiperOrigin-RevId: 261907364 GitOrigin-RevId: 5315e7b98905922e779798f3168d98343438c134 Change-Id: Ia8ab971c54f287411f6ea4b99f9c666c989c33fd

view details

Abseil Team

commit sha b49b8d16b67ec6912899684b732e6367f258cfdb

Export of internal Abseil changes -- 00c451dc81be7fe05f982b08b4ea1edc2ca2c1c5 by Abseil Team <absl-team@google.com>: remove a test that is currently broken on emscripten from running on emscripten. PiperOrigin-RevId: 262005667 -- 9df5f5acb65996bdb99900039a4f01a44811aa14 by CJ Johnson <johnsoncj@google.com>: Adds a layer of macro-indirection to opening up namespace absl inside a macro. This helps avoid an issue identified with the LTS inline namespaces PiperOrigin-RevId: 261990937 -- 5d40aa129cd77a1b853e5389aff7eacffe5c8204 by Gennadiy Rozental <rogeeff@google.com>: Fix handling of new lines in flag help descriptions. If there are explicit new lines in a flag help description string, we respect it and format the usage message accordingly. PiperOrigin-RevId: 261974244 -- 4997b5a2ddb983969059470a2d2bc2416b3d785e by CJ Johnson <johnsoncj@google.com>: Import of CCTZ from GitHub. PiperOrigin-RevId: 261955031 GitOrigin-RevId: 00c451dc81be7fe05f982b08b4ea1edc2ca2c1c5 Change-Id: I5a13bfb15bba0f7b6e49c0655c57c3addfeb1c72

view details

Abseil Team

commit sha 8efba58a3b656e9b41fb0471ae6453425a61c520

Export of internal Abseil changes -- 38bc0644e17bf9fe4d78d3db92cd06f585b99ba7 by Andy Soffer <asoffer@google.com>: Change benchmark to be cc_binary instead of cc_test, and fix a bug in the zipf_distribution benchmark in which arguments were passed in the wrong order. PiperOrigin-RevId: 262227159 -- 3b5411d8f285a758a1713f7ef0dbfa3518f2b38b by CJ Johnson <johnsoncj@google.com>: Updates Simple<*>() overload to match the name schema of the others PiperOrigin-RevId: 262211217 -- 0cb6812cb8b6e3bf0386b9354189ffcf46c4c094 by Andy Soffer <asoffer@google.com>: Removing period in trailing namespace comments. PiperOrigin-RevId: 262210952 -- c903feae3a881be81adf37e9fccd558ee3ed1e64 by CJ Johnson <johnsoncj@google.com>: This is a cleanup on the public header of InlinedVector to be more presentable PiperOrigin-RevId: 262207691 -- 9a94384dc79cdcf38f6153894f337ebb744e2d76 by Tom Manshreck <shreck@google.com>: Fix incorrect doc on operator()[] for flat_hash_set PiperOrigin-RevId: 262206962 -- 17e88ee10b727af82c04f8150b6d246eaac836cb by Derek Mauro <dmauro@google.com>: Fix gcc-5 build error PiperOrigin-RevId: 262198236 GitOrigin-RevId: 38bc0644e17bf9fe4d78d3db92cd06f585b99ba7 Change-Id: I77cababa47ba3ee8b6cebb2c2cfc9f60a331f6b7

view details

Abseil Team

commit sha 9ee91d3e430fb33a4590486573792eb0fa146c2d

Export of internal Abseil changes -- f51743aa96e19aa3dda96d09d313b4390f1d61e7 by CJ Johnson <johnsoncj@google.com>: Minor touchup on the InlinedVector/Storage internal header file PiperOrigin-RevId: 262389640 -- e2f54c1f7142e40d30ff0fda43bef050625821a5 by Abseil Team <absl-team@google.com>: Update the absl codebase to use proper marketing names for macOS and Xcode PiperOrigin-RevId: 262389450 -- f29aae774edd0d00e2daa1fb96694a6dc3565a55 by CJ Johnson <johnsoncj@google.com>: Blocks code generator script from being affected by LTS inline namespaces PiperOrigin-RevId: 262376791 GitOrigin-RevId: f51743aa96e19aa3dda96d09d313b4390f1d61e7 Change-Id: I33be7f5a708ce8a2b7111b00151e43d73c5e0009

view details

Abseil Team

commit sha 4ef574064e75b86f115549e9eb4c7e806781b3ab

Export of internal Abseil changes -- 3dbb096e4662311f81df1017a8e0975e903936cf by Derek Mauro <dmauro@google.com>: Document and workaround a known MSVC bug doing constexpr pointer arithmetic PiperOrigin-RevId: 262604652 -- b5fa7f1a0c776f6ba20d52772a1679ec42ad21fd by Derek Mauro <dmauro@google.com>: Fix typo in macos_xcode_bazel.sh PiperOrigin-RevId: 262591285 -- 89dd77ab5bb44d76b6cb6b2f288e21536e16a85a by Derek Mauro <dmauro@google.com>: Internal change PiperOrigin-RevId: 262582747 -- 32295ed9a0c6c8ab143a912194040eede05d3ea3 by Abseil Team <absl-team@google.com>: Internal change PiperOrigin-RevId: 262569140 -- 7f0f5b94197369228024529022d727439d2c894f by Abseil Team <absl-team@google.com>: Internal change PiperOrigin-RevId: 262563554 -- 314aed043639abbd221074125c57b7c68616de7e by Derek Mauro <dmauro@google.com>: Release absl::btree PiperOrigin-RevId: 262553526 -- 72b44056c6ce9000c4a6cd9aec58b82067c82a13 by CJ Johnson <johnsoncj@google.com>: Internal change PiperOrigin-RevId: 262421185 -- 4e2c12151edf534f929e8e810f1334073f90489a by Abseil Team <absl-team@google.com>: Update documentation to make it less likely for users to write `Hours(24)` without considering using civil dates instead. PiperOrigin-RevId: 262420758 -- 92b85b9573e800bd96b019408eefbc5ce4f68780 by Derek Mauro <dmauro@google.com>: Add the ability to override the bazel version in the macos_xcode_bazel.sh test script. PiperOrigin-RevId: 262412063 GitOrigin-RevId: 3dbb096e4662311f81df1017a8e0975e903936cf Change-Id: I423b2b829dc0c5f814e37bec4d68c7470f43f041

view details

Abseil Team

commit sha 321ab5303023c86cd15d9ddc5740fb4b4fde32e1

Export of internal Abseil changes -- d3d344e214e80b766d3e5c355e16124eb8602ff2 by Tom Manshreck <shreck@google.com>: Add LTS Branch to LTS docs PiperOrigin-RevId: 262904704 -- a9b10e6959209e8ded66e5fb041b4f1811a3f375 by CJ Johnson <johnsoncj@google.com>: Minor cleanup on InlinedVector headers PiperOrigin-RevId: 262632981 GitOrigin-RevId: d3d344e214e80b766d3e5c355e16124eb8602ff2 Change-Id: I3d0c078c08520b5c6a1a46c4b681ae0d98297f24

view details

Abseil Team

commit sha d9aa92d7fb324314f9df487ac23d32a25650b742

Export of internal Abseil changes -- 2dd3b23ea940804de727e396b300cfae4b1b71a1 by Derek Mauro <dmauro@google.com>: Upgrade to LLVM r366207 and Bazel 0.28.1 for Linux-Clang testing. PiperOrigin-RevId: 263162761 -- f03ae9e4e9f42c075745d28b4ced78071e73724a by Derek Mauro <dmauro@google.com>: Remove unneeded dependencies on //absl/base from targets that no longer depend on it. PiperOrigin-RevId: 263129193 -- 8ec2aab1eea50e7f71a6a687a07e5ae0e0945f5e by Derek Mauro <dmauro@google.com>: Move raw_logging to a separate target (raw_logging_internal) PiperOrigin-RevId: 262972007 GitOrigin-RevId: 2dd3b23ea940804de727e396b300cfae4b1b71a1 Change-Id: I3f4580e87797386b0b5e90c8ced74cbf078f61c1

view details

push time in a month

create barnchbmzhao/bazel-cc-so-example

branch : forked-cc-proto

created branch time in a month

push eventbmzhao/bazel-cc-so-example

Brian Zhao

commit sha b89ad0afeee4941b85ec1d21ae4a34f6c373140c

Initial commit

view details

push time in a month

create barnchbmzhao/bazel-cc-so-example

branch : master

created branch time in a month

created repositorybmzhao/bazel-cc-so-example

Bazel CC Shared Library Developer Flow Use Case Examples

created time in a month

push eventbmzhao/tensorflow

ANSHUMAN TRIPATHY

commit sha 288db71a37dc852e6bdc6faf5d52b962f43297ce

Lite: Util new test cases added to improve coverage

view details

Dayananda-V

commit sha 2ed77aa3a12a8cc189509f255195657124415a5a

TF Lite toco/tflite warning fix toco/tflite module warning fix

view details

Yong Tang

commit sha 428d1761f2a3ab74601fc1a15e5a280a7f6b2464

Improve error message of RaggedTensor by showing data type explicitly While working on writing a tf.data pipeline with RaggedTensor the following error showed up: ``` def raise_from(value, from_value): > raise value E InvalidArgumentError: Expected splits Tensor dtype: 9, found: 3 [Op:RaggedTensorFromVariant] /usr/local/lib/python2.7/dist-packages/six.py:737: InvalidArgumentError ``` It is not very obvious about the exact type that needs. Until found out in `tensorflow/core/framework/types.proto` that `3` is `int32` and `9` is `int64`. This PR enhance the error message by explictily print out the DataType in string, so the message will be: ``` E InvalidArgumentError: Expected splits Tensor dtype: int64, found: int32 [Op:RaggedTensorFromVariant] ``` Signed-off-by: Yong Tang <yong.tang.github@outlook.com>

view details

Jerry Shih

commit sha 7d7c4eafce0aaff9b28373b4c9f3032e6232960e

Add the missed header for osx platform in micro_speech example. When we use the target specific "audio_provider.cc"[1], we should also add its header dependency[2] in makefile. [1] tensorflow/lite/experimental/micro/examples/micro_speech/osx/audio_provider.cc [2] tensorflow/lite/experimental/micro/examples/micro_speech/simple_features/simple_model_settings.h

view details

Fangjun Kuang

commit sha 0d08d8dc0b37019ee90a4b89301a8abfc4451492

fix #32416. Override operator delete.

view details

Owen L - SFE

commit sha c345f83efd593217e660c7e55a5db6e2415bca23

change ambiqsuite sdk references to release 2.2.0

view details

Owen L - SFE

commit sha 623237166f38dfaa9f849c21de6e954ca1e85643

add AM_SDK_DEST to handle future sdk releases

view details

Owen L - SFE

commit sha 41fe01bf47ca2d3c17b693ec1c71dbdec6333825

remove workaround for 2.0.0 bug

view details

Owen L - SFE

commit sha a7790f5de854a695709def9e42f41d5cd420190b

update SparkFun edge bsp repo to support 2.2.0 release

view details

Owen L - SFE

commit sha 82863ba8a9b744733e1b3e474e4359f357e1414a

remove unused downloads

view details

Owen L - SFE

commit sha 7eb31ac220dd37336def5995e8a0b51a7ceacb2e

update edge bsp to latest

view details

Owen L - SFE

commit sha ecaadd6ea54fe7938be4b54b70f6ab872628ad0f

standardize to use of am_devices for led control

view details

Owen L - SFE

commit sha 9e287ac40d68a6f168d756dd5392253eb6f0c48f

rely on bsp for image_provider.cc

view details

Owen L - SFE

commit sha 9a82753dcd4eed6da81c0d0d80a98de6dc7ce31b

go to generic "platform.h" allows future targets to provide their own version

view details

Owen L - SFE

commit sha 7c3f821c7b233ebcade169142dc56544ee455563

more cleanup CMSIS_ext is not used in Apollo3 implementation

view details

msteknoadam

commit sha 52392ac73c70c358032ae45e0dbeeba929f871cb

Added usage examples to some APIs Added to: - image.random_flip_up_down - image.flip_up_down - image.random_flip_left_right - image.flip_left_right

view details

msteknoadam

commit sha 5622e01b6e678581a02510bcd8c99ea10756ef08

Added more usage examples Added usage examples to these APIs aswell: - image.transpose - image.random_brightness - image.random_contrast - image.random_hue - image.random_jpeg_quality - image.random_saturation

view details

msteknoadam

commit sha ca04a448b3f569747ccf32e564e002a17b08022e

Apply suggestions from code review Co-Authored-By: Kilaru Yasaswi Sri Chandra Gandhi <yasaswisrichandragandhi@gmail.com>

view details

msteknoadam

commit sha 40557a2a60401db83910bf52732c1e14f8a5ecca

Removed unnecessary linebreaks

view details

msteknoadam

commit sha 030ac63a93967da57a6272f42f1b52b779e12091

Merge branch 'master' of https://github.com/tensorflow/tensorflow into example-adding-branch

view details

push time in 2 months

issue openedbazelbuild/bazel

Bazel Query in Bazel 1.2.1 errors out due to bazel-internal code

Description of the problem / feature request:

bazel query deps in bazel 1.2.1 on targets that use android trigger the following error:

ERROR: /usr/local/google/home/bmzhao/.cache/bazel/_bazel_bmzhao/5dcb290106c1a06f5bc5c8d08ceb177f/external/bazel_tools/tools/android/BUILD:39:1: no such target '//external:android/d8_jar_import': target 'android/d8_jar_import' not declared in package 'external' (did you mean 'android/dx_jar_import'?) defined by /usr/local/google/home/bmzhao/Code/tensorflow/WORKSPACE and referenced by '@bazel_tools//tools/android:d8_dexbuilder'

This error is caused by this line: https://github.com/bazelbuild/bazel/blame/master/tools/android/BUILD.tools#L44

This also breaks Tensorflow's CI, since we run 'bazel query "deps((//tensorflow/...))"`, and its picking up this internal bazel dep that errors out.

Feature requests: what underlying problem are you trying to solve with this feature?

We are migrate Tensorflow to bazel 1.2.1. This is part of the work for tensorflow

Bugs: what's the simplest, easiest way to reproduce this bug? Please provide a minimal example if possible.

git clone https://github.com/tensorflow/tensorflow.git
cd tensorflow
git checkout 1d41edaee62e27c8c1552b83fcb46dd45e5a64d4 # known-clean-commit
echo '1.2.1\n' > .bazelversion
bazelisk query "deps((//tensorflow/... - kind('android_*', //tensorflow/...)))"

What operating system are you running Bazel on?

Linux

What's the output of bazel info release?

release 1.2.1

created time in 2 months

pull request commenttensorflow/community

RFC: TensorFlow Build Improvements

As mentioned in SIG Build, here's a link to the master spreadsheet of the current status of the refactoring so far: https://docs.google.com/spreadsheets/d/11QLBDRa2XHNAtA_Wby81Q9xWnDHTp6ChJrVkstbBRBE/edit?usp=sharing

bmzhao

comment created time in 2 months

push eventbmzhao/tensorflow

Xinan Jiang

commit sha fa1259ed252ceada5e82358682c98b4a9b212cd7

[Grappler] Fix comparison between node name and input in function UpdateConsumers

view details

leike666666

commit sha 65944087669c42efcc0c470ea1c09a1ba582d169

move function DisableAllStages from protected to private

view details

Rasul Karimov

commit sha b663317ad47a0ba97b97b1a8d673caa89615b0cb

quick fix for reduce_logsumexp in v2

view details

Rasul Karimov

commit sha c285ea313ca6f23d44349fc4dab69a2c1d5140ee

quick fix for reduce_logsumexp in tf2

view details

Rasul Karimov

commit sha 7a1b7d0827174572d36b8588dd7b80b782b03206

quick fix for reduce_logsumexp in tf2

view details

Rasul Karimov

commit sha ae522859b9101bacbef1da44b6d8230013eb2da9

quick fix for reduce_logsumexp in tf2

view details

Rasul Karimov

commit sha f6016498e316ec2260c22cbc2a58d5704328e56b

correct docs

view details

Rasul Karimov

commit sha cab5922dca2eb71c1df6b51cb5cf0ba9320d26c0

fix docs

view details

Rasul Karimov

commit sha 05e93d61315f43445ddbbef6d023760f75cf38fe

remove repetition

view details

Rasul Karimov

commit sha e8b67291ddd57bff91fd65bae983925bb4714fe0

remove docs for selectv2

view details

tigertang

commit sha d9a386fe5ff6f8d2dc011c832708ed4096904e0f

Fix a typo in tflite_inference_stage

view details

ShengYang1

commit sha 5024be7bffafcd3e3199bc0a672601bc3a74fd2e

Create a partial key for output_scale to improve key creation performance

view details

Deven Desai

commit sha 1b1c46ebe85cb0cb0b182c87d896d6499a13b581

disabling subtests that test 3D pooling ops, and removing the no_rocm tag from //tensorflow/cc:gradients_nn_grad_test

view details

Duncan Riach

commit sha 3bd6cba36a3f50023a5b3a585d7b63df4d9fef2f

Improve docstrings for tf.config.optimizer.get_jit and .set_jit

view details

HotPotatoC

commit sha f5339f98b20b4bf4622c34303897084a92fc2e6b

Added tf.strings.to_number() usage example

view details

Pooya Davoodi

commit sha 1717e2cfb2ca4a782e8fd0e9bfc2fc9e2349db42

Rename segment_graph_ to segment_graph_def_

view details

HotPotatoC

commit sha a6f4bcbf1ccc90db69ce3d75abace23380d38c61

Changed Usage example into Example

view details

Harry Slatyer

commit sha 66fbe49c0747fb4fa626e127d34b9baf73fa287b

Switch from where to where_v2 in matrix_exponential (because `where` is deprecated). `where` and `where_v2` have different broadcasting rules: the former "broadcasts" by treating a 1D condition as a mask on the *outer* dimension of x and y; the latter follows standard broadcasting rules, which cause a 1D condition to act as a mask on the *inner* dimension of x and y. E.g. with `where_v2`, if x and y are [n, d, d], then a condition [n] will either fail to build (if n != d) or be treated as [1, 1, n] (if n == d). In this case, we want the condition to act as a mask on the outer dimension, e.g. be treated as [n, 1, 1]. The way to make that happen with `where_v2` is simply to expand the condition's dimensions to that shape manually. In the case of matrix exponential, by expanding the shape of `l1_norm` to [n, 1, 1]: - the conditions in `_nest_where` become the right shape, - the `squarings` variables get expanded too (i.e. [n] -> [n, 1, 1]), which means that... - ... when scaling the Pade approximants we no longer need to expand the dimensions of 2^squarings, and ... - ... the condition in `b` (used for the squaring while loop) becomes the right shape. Which is what we need.

view details

pshiko

commit sha 84797959d568df8220aef9160fa505a36489f763

infer summary writer logdir

view details

pshiko

commit sha 8dc69c25c3f688eb3a82133c313d54ae8e498856

fix

view details

push time in 2 months

push eventbmzhao/tensorflow

ShengYang1

commit sha a1bdc83cc8c8cde403f59365b956258ad07ab9a7

change MirrorPad packet region

view details

Elena Zhelezina

commit sha 4d19a5b4626f08debcbc14a324378b9e800431d7

[tflite] Fix and tests for the operator PACK There is a bug in the current implementation of the operator PACK. The multiple inputs should be requantized. The current PR fixes this bug and adds a test for the quantization of the PACK operator. Change-Id: Id829a6c62ea35b6b1b64c41797c02a924c14064a

view details

jacco

commit sha 3116dd92c55c00137f9f9ccea086b9842e7449ef

Add support for synopsys ARC based platforms - add arc specific makefile include - link mli lib - Update makefile generator for TARGET_ARCH=arc In case TARGET_ARCH=arc: changing the makefile generator to add the CC/LD executables to the Makefile; diabling Mbed,Keil and Arduino projects generation, adding TCF_FILE variable support that also defines the TARGET string - TCF file is copied to the project root directory to get rid of the absolute path - adding .elf extension (ARC only); default TCF is ARC EM7D; clean target for project makefile template - Linker output is redirected by the appropriate option - Additn a suffix (TCF name) to the MLI directory name to support parallel project generation inside the same sources tree; using gmake to build MLI

view details

Koan-Sin Tan

commit sha 9a27e606f4a59aac96682ed8ac1e03b774c79b7b

make TransposeConv on NNAPI work NNAPI TransposeConv op should take tensor inputs from TFLite node. This actually is from a0ae68ca3c

view details

Elena Zhelezina

commit sha 27f132a155b746cf41d5adcb31882e7ffb974b68

Merge branch 'master' into fix_pack

view details

Yuan Tang

commit sha dcf06560c7b48ed1df96242444456fae95129404

Enhance doc: column oriented -> column-oriented

view details

Kimberly

commit sha 4fcf1f3fd58d6f2fb029fe95d02887c9c73ea77b

Fix deprecated args for count_nonzero

view details

Elena Zhelezina

commit sha 47735cb3471d053a7ebdf6b97ecd961ac7187159

Merge branch 'master' into fix_pack

view details

Elena Zhelezina

commit sha a0338d6505bfa40a8cca32d3610febd098082755

Addressed review comments. Change-Id: Ifefd7825025eedda29c1fc5baf7a14b5f0acae09

view details

Elena Zhelezina

commit sha c93191244a33100728bd39a24e104cc1a4636ffd

Addressed review comments. Change-Id: Ibaca013d126b9a50e775247f2a6df90244160eb9

view details

Elena Zhelezina

commit sha df972879598975ebe396080b41505a2656a8a284

Fix Ubuntu sanity check. Change-Id: I1e15495f001d2fe56326c8b056a73756a5e5e04b

view details

jacco

commit sha 87c2146857ea652e51106c3d6d4348e88da955d7

Merge remote-tracking branch 'upstream/master' into arcport_pullrequest # Conflicts: # tensorflow/lite/micro/tools/make/helper_functions.inc # tensorflow/lite/micro/tools/make/third_party_downloads.inc

view details

Giorgio Arena

commit sha 59acc254879018e6478378334c6f4d14231e6e0a

Implement reference kernel for Add and Mul using CMSIS-NN

view details

HotPotatoC

commit sha 11eb64d223aadd730a860fb94d3b4cc9f80034ff

Added usage example on tf.image.rgb_to_grayscale and tf.image.rgb_to_yiq

view details

HotPotatoC

commit sha 178697109b5b9db1537f61037ad7e87b32f78fff

Updated usage example

view details

HotPotatoC

commit sha 983a82c1ab29b1f2ad2751537e7754ab15673f7d

Fixed indentation

view details

HotPotatoC

commit sha 93c6cf08da979e5499117bae85929f46e6327aa6

Updated README.md

view details

HotPotatoC

commit sha e552bd53f2d490668b298609a9bde5e1a8a6b6c2

Merge branch 'hotpotatoc' of https://github.com/HotPotatoC/tensorflow into hotpotatoc

view details

HotPotatoC

commit sha 92545ef632e537f8257cf950037e69527a19cf21

Updated Usage example

view details

HotPotatoC

commit sha 607e1e8cfac2ee20996d61b6f48097d8deaf9973

Updated usage example

view details

push time in 2 months

push eventbmzhao/tensorflow

A. Unique TensorFlower

commit sha b7337d11e89a073857c9cb8c39a23b6f687edc5a

Go: Update generated wrapper functions for TensorFlow ops. PiperOrigin-RevId: 287434296 Change-Id: I8c82aedce6473c3bf2d070f945b3ff08f86989fe

view details

A. Unique TensorFlower

commit sha 41780ce36ad4cb1366521627ebf5e4d3ff9981a7

Go: Update generated wrapper functions for TensorFlow ops. PiperOrigin-RevId: 287439616 Change-Id: I55f41a387aec720c09731702818ab8f93d390020

view details

A. Unique TensorFlower

commit sha 891a99a55f9cebce868b1855b15bd15b8cd837fa

compat: Update forward compatibility horizon to 2019-12-29 PiperOrigin-RevId: 287446309 Change-Id: Ie87a7d449720c3f5d04649d33627d6a318931199

view details

A. Unique TensorFlower

commit sha 6fa32ab0181fd6bdf4d0f80b17480b2317dfd9cb

Go: Update generated wrapper functions for TensorFlow ops. PiperOrigin-RevId: 287463768 Change-Id: Ifd62e8e94f4a91c8499f583e74274f690ebb89a0

view details

George Karpenkov

commit sha 6b9a4ad6923a4ad24c0a8520654c36860e34b1cc

[XLA] Do not error out from BufferComparator when ptxas is not found Produce a more helpful error message on how a custom ptxas location can be supplied. Should help with https://github.com/tensorflow/tensorflow/issues/35423 PiperOrigin-RevId: 287470320 Change-Id: I9261ccced5812d56636650638399cee769a13468

view details

A. Unique TensorFlower

commit sha d25602489ee8e4b7e163fc9346a4d092b5ba9f6c

Go: Update generated wrapper functions for TensorFlow ops. PiperOrigin-RevId: 287475356 Change-Id: I0303662e18c9904850a47fe8366180513c5fbc27

view details

A. Unique TensorFlower

commit sha c558e7747a1ba5def2f99fe9e336f22c61686ae8

Relax multi-output sibling fusion constraint to allow creating new fusion nodes. PiperOrigin-RevId: 287486590 Change-Id: I52e77c066e6cf41f2c8e2471f191fccc42941e3f

view details

A. Unique TensorFlower

commit sha 6be5a858be55285354171aabdde4438eb7d945db

Go: Update generated wrapper functions for TensorFlow ops. PiperOrigin-RevId: 287486868 Change-Id: I4c43911b90039fbd257556d433aa808648003432

view details

A. Unique TensorFlower

commit sha 1eb909f868fd333a83ec8a31720b6853f57176e0

Go: Update generated wrapper functions for TensorFlow ops. PiperOrigin-RevId: 287498176 Change-Id: I9b2e9b59c001d85a709b54df35317155fdbe39c0

view details

push time in 2 months

push eventbmzhao/vcpkg

Brian Zhao

commit sha d13901c5a1fbc76d72fd9963ccb5b49534e6bc66

WIP tf 2.1.0rc2

view details

push time in 2 months

push eventbmzhao/vcpkg

Brian Zhao

commit sha 934a72196f24f0ebc0ce72c9aa6a14a353687751

WIP tf 2.1.0rc2

view details

push time in 2 months

create barnchbmzhao/vcpkg

branch : tf-21-bazel

created branch time in 2 months

fork bmzhao/vcpkg

C++ Library Manager for Windows, Linux, and MacOS

fork in 2 months

PR opened tensorflow/tensorflow

Update .bazelversion

This changes the .bazelverison file to be consistent with Tensorflow CI at 2.1.0. See: https://github.com/tensorflow/tensorflow/blob/v2.1.0- rc2/tensorflow/tools/ci_build/install/install_bazel.sh#L18

This allows clients that use bazelisk to automatically build with the right version of bazel.

+1 -1

0 comment

1 changed file

pr created time in 2 months

create barnchbmzhao/tensorflow

branch : bazelversion-cherrypick

created branch time in 2 months

push eventbmzhao/tensorflow

ruchit2801

commit sha 446ee0118980a784c0b43bcded9ba9205bb00293

Fix bug in documentation of tf.while_loop.parallel_iterations

view details

A. Unique TensorFlower

commit sha 3ec7d0638f45820c34a44f0e0b6a6abd89ab26aa

Go: Update generated wrapper functions for TensorFlow ops. PiperOrigin-RevId: 287081521 Change-Id: I205172749fe26ce8d1bae9df78b54868f0b2f040

view details

A. Unique TensorFlower

commit sha 75885d32772ab302efa6faa81edc17f4a05fd1f0

compat: Update forward compatibility horizon to 2019-12-25 PiperOrigin-RevId: 287082312 Change-Id: I986aee443acda75a40c0fcc59722284173e8e19a

view details

A. Unique TensorFlower

commit sha cdf4b8afeff3f867903a0d984fa4fa5e5d54591f

Go: Update generated wrapper functions for TensorFlow ops. PiperOrigin-RevId: 287089219 Change-Id: I9a435634ee59b79cbb6ac75968ff0f45ca0bd917

view details

A. Unique TensorFlower

commit sha aac0a3dc4c6e0045a90793cae0aec8481c0a2085

Go: Update generated wrapper functions for TensorFlow ops. PiperOrigin-RevId: 287096197 Change-Id: Ia7154bbd4bbceea1fcb9a58a30c6e3e9b401f84b

view details

A. Unique TensorFlower

commit sha 517bf21fd5ba6616d08d1b36c4c2da4918ec6bc6

Go: Update generated wrapper functions for TensorFlow ops. PiperOrigin-RevId: 287102812 Change-Id: I0b5ff6bd71674c93735f1ba68f5194b87ba5f372

view details

A. Unique TensorFlower

commit sha f4901b6aa9ed247f3c915a8e984fb958d40ff235

Go: Update generated wrapper functions for TensorFlow ops. PiperOrigin-RevId: 287115460 Change-Id: Ie7ab42ed6e1543d44a7daca54aa187329f67d261

view details

A. Unique TensorFlower

commit sha d3457b26a0e9fdc5c1e7452fe525720cf110b41d

Go: Update generated wrapper functions for TensorFlow ops. PiperOrigin-RevId: 287145079 Change-Id: I7fe7e09b6c2f5d602170a38dc02cf4fb9889e29e

view details

A. Unique TensorFlower

commit sha eda019709cd6d7695932907faf57c4d92f06ddeb

Go: Update generated wrapper functions for TensorFlow ops. PiperOrigin-RevId: 287159337 Change-Id: Ia895bcb392b8ee17b378fc41bd6dc055182747cf

view details

A. Unique TensorFlower

commit sha e0d83616534a28dfa2307e74e581086ea31879a0

compat: Update forward compatibility horizon to 2019-12-26 PiperOrigin-RevId: 287160636 Change-Id: I2fe3b884af9a8f6bbcdac92cb8bfb02e8e408f49

view details

A. Unique TensorFlower

commit sha 8add83c8ea632f1758fc4f834f7764ac7b93df89

Go: Update generated wrapper functions for TensorFlow ops. PiperOrigin-RevId: 287175471 Change-Id: I939032f14100c898d6c8a9c4ab5453e1e22cc210

view details

A. Unique TensorFlower

commit sha c23505c657e6a40a84c4e983ea529a1310c4a6e1

Go: Update generated wrapper functions for TensorFlow ops. PiperOrigin-RevId: 287182661 Change-Id: Ic0944d65ce6e36ec0bd50c5c778706431ea7332e

view details

Berkin Ilbeyi

commit sha f2dce5f02dd98e8ee599dc2326c2d454568e751a

[XLA] Fix scheduling due to bitcast/tuples and evictions. This fixes a bug where a bitcast can force a CopyStart/CopyDone earlier than intended, causing memory corruption. Added a CHECK so that CopyStart/CopyDone's aren't inserted due to operand dependencies anymore. Also implement mechanism to schedule eviction end time later than the end time of the previous allocation. This fixes an issue where an eviction after a while loop didn't allocate memory for itself. PiperOrigin-RevId: 287188902 Change-Id: I21e083a01aad3511aa4caa4e51a9a873f808467a

view details

A. Unique TensorFlower

commit sha be7d68dfe38d5b9d8a0c8cdc25a4562b512fb4d5

Go: Update generated wrapper functions for TensorFlow ops. PiperOrigin-RevId: 287191839 Change-Id: I8be2363c24bc841bdb2f983c74803add737aa529

view details

Feng Liu

commit sha 7330d9f21f22b89ed12a746a2c3d003adf977261

Remove the "name" attribute when the ops are imported When the graph is exported, the "name" attribute is created from the op location. PiperOrigin-RevId: 287195656 Change-Id: Iec65f930198b4ff9ca021f4aeace099302a92472

view details

TensorFlower Gardener

commit sha 95c535c6c24b8ab5dc4aae717f9c5bdcef795d54

Merge pull request #35331 from ruchit2801:i#35065 PiperOrigin-RevId: 287199200 Change-Id: Iabac4a3db5c754e84b4fa129b664bcfe51fd7c52

view details

Dan Moldovan

commit sha 87f69493d2553ceb9fdb94ca7f203dc1c8e417ea

Cleanup for closed bugs: remove associated optimization patches. They no longer seem to have any effect. V2 GPU control flow remains ~20% slower than V1. PiperOrigin-RevId: 287200087 Change-Id: Ia35f1bdaf9a0ff8b0081f78f853209f86acb010f

view details

Dan Moldovan

commit sha 107bd3e40b9c3b2ba42bf440e0c456e921e6c644

Prepare for gast 0.3, to which TensorFlow will upgrade in version 2.2. Since TF pins the version, all libraries that depend on it need to support both gast versions to avoid dependency conflicts. PiperOrigin-RevId: 287200227 Change-Id: I40b10e99ccf0d386aca393d3cb0a061ebc6d9744

view details

Gaurav Jain

commit sha 400e246b7aed17ef2ea0590019ce2336405c56ed

Only reset TFE_Op from the same TFE_Context PiperOrigin-RevId: 287201238 Change-Id: Id1732edd95aa470f2952de2620e3ce33fe4bfcb0

view details

Frank Chen

commit sha 5078cab51cea45364602f9d2a1d30057799af4b0

Hook out compile, execute, load/unload APIs to external TPU driver PiperOrigin-RevId: 287204382 Change-Id: Ia1ae8ec3179c09e393aec69041a41c241177efe7

view details

push time in 2 months

create barnchbmzhao/tensorflow

branch : bazel-2

created branch time in 2 months

startedfacebook/bistro

started time in 2 months

startedfacebook/fishhook

started time in 2 months

push eventbmzhao/tensorflow

Anton Kachatkou

commit sha 093e00fee1c1447800e1cc5d80a9ca1c65e54256

Fix for the test interpreter_test.py with PY3 The test interpreter_test.py is failing when it is run using PY3. There are several issues due to difference between PY2 and PY3: The string in PY2 is stored as bytes by default, but as unicode in PY3 by default. In PY3 if the string is in bytes, it has the prefix 'b'. We convert all strings to be the same. There is an exception chaining in PY3. As result, the message thrown from the exception is different. There is a segmentation fault on Python3, when all tests in the file interpreter_test are run. The bug is that the global variable stores the pointer to the function, which goes away from the scope.

view details

Anton Kachatkou

commit sha f78701b5fcf5b12fdc4990f3b47fa13ab99b1a5f

Addressed comments for the fix for test interpreter_test.py with PY3 The type byte is defined only in Python3, that is why we need to check version of Python. I tested this test using Python3 and Python2 Removed unnecessary formatting.

view details

Anton Kachatkou

commit sha d3564251e34fc862649fe8071a88817844daa471

Merge branch 'master' into toupstream/fix-tflite-interpreter-test

view details

Kristian Holsheimer

commit sha ffd85d2a9a9b9f30415d0fe40a47cb1bb9f44214

replaced builtin `all` by `np.all`

view details

Kristian Holsheimer

commit sha 3709e9056f42d27d498769e79813204ce5c4f1e7

Merge branch 'master' into master

view details

Drake Gens

commit sha 87a34fbe238d61bf11ced584ad30f36555ab5f72

improved documentation

view details

Koan-Sin Tan

commit sha 996f0ec8b3e915a3264b409ff92011465dd8647b

[tflite] expose more NNAPI Delegate to Java 1. use StatefulNnApiDelegate() instead of depecated one 2. make it possible to specify NNAPI execution preference in Java

view details

Ending2015a

commit sha 390eed2958d950aa76c7d5a84e1f17360d14f9fd

fixed KeyError issue for tf.Module

view details

amoitra

commit sha ab29d4b822ddf3c18aa7d8c17354140973937942

algebraic simplifier enhancements

view details

Koan-Sin Tan

commit sha 0d1f151f89d869db3a4bb33b0b72e8bf89a05312

address review questions 1. remove unnecessary dependency 2. more explanation for the default preference setting

view details

Frederic Bastien

commit sha bd0238ab1489c99f81d184a6bcbe4b4739d394a6

Small comment fix

view details

Daniel Falbel

commit sha dd182e465f3d425b49de2745211acef1d28843b0

equality testing instead of identity

view details

Frederic Bastien

commit sha 1542db8777db8ace72b071baef494a1436d06397

Add tf_xla_supported_nodes flags to limits nodes XLA consider.

view details

Frederic Bastien

commit sha e194f62bf57c404d40c17aad281d3f1d7d602490

More Ops in the whitelist category.

view details

Frederic Bastien

commit sha 08af4215fd8000618a3908758b709b943a1a014c

Add more XLA whitelist shortcut

view details

Frederic Bastien

commit sha c44dc1f453a2264ddc36162fe1d807d8ee3fcae2

Better user error detection, less verbose and better error message.

view details

Frederic Bastien

commit sha 727edfd86f143534a59c6760afd161010bb7de20

Add an EXTRA category

view details

Frederic Bastien

commit sha 65f614b7322492ae0b8154770f1330a77fa2afe9

Add the flag value TF_XLA_FLAGS=--tf_xla_auto_jit=fusible to enable XLA and the tf_xla_supported_nodes=FUSIBLE flag. This make using that mode easier.

view details

Frederic Bastien

commit sha 85ed59b2141f1c35524638597174b34692013030

Update docmentation of the new flag

view details

Frederic Bastien

commit sha 7cd64793a32bae2ade6aee054899a68b08a4a60d

Use the absl containers.

view details

push time in 2 months

push eventbmzhao/tensorflow

Pooya Davoodi

commit sha a26b56085598bed7afe7cfef43fd1e2547b3831d

Add script to generate TF-TRT model to be used for testing

view details

Pooya Davoodi

commit sha f1129b373d05c9dc9d11943807c6a361a6ab3092

Add tftrt_2.0_saved_model This model can be loaded and executed in tests to ensure backwards compatibility across TF versions.

view details

Pooya Davoodi

commit sha a55e9c2ae95e9d899383d4d712719e6285f3ad55

Add backward compatibility test

view details

Pooya Davoodi

commit sha ebde7cce583c0be72c887b18ac17a1bc828ec95e

Move data of quantization_mnist_test to testdata/mnist

view details

Pooya Davoodi

commit sha 3ac36859554e3f9c488dd7c78366caa2eb51863c

Add mnist dir, add correctness check for bwd compatibility test

view details

Pooya Davoodi

commit sha 6ab11cb0239c4b5ba1837a42ffc0af2e5679839a

Fix pylint for backwards compatibility test

view details

Pooya Davoodi

commit sha 8105a6e900891e93ad13af6c763600c29a549823

Sort targets in BUILD and shorten statement for pylint

view details

Shane Smiskol

commit sha 119ec6954dad9ea443c8a80df7fb165f9032d529

Fix dynamic display for PyCharm

view details

Pooya Davoodi

commit sha 3461d949be44486bf4b074e02385bf745fe5b41c

Remove import tensorflow from gen_tftrt_model.py

view details

Pooya Davoodi

commit sha 51f0450445ece5acc534bb15c0b9cca693d49180

Add tftrt_2.0_saved_model to data deps of trt_convert_test in BUILD Also remove gen_tftrt_model.py from BUILD because this file is not supposed to be used by any test. It's supposed to be used manually by developers to generate a test model if needed.

view details

Guangda Lai

commit sha 168b722a126455d18a64aa2e0e80a8eb3325bbdf

Add no_pip tag to avoid pip_smoke_test depending on newly added savedmodel.

view details

Guangda Lai

commit sha d066b24d024220ffbf19c98cc34b7162ba2203e1

Merge branch 'master' into tftrt_bwd_compatibility_test

view details

Jian Li

commit sha 367f2c213949fcd2a49f12f90f481bd75e05caa9

Fix a wrong tensor type. It was not excised yet and was only discovered when working on adding more test cases. PiperOrigin-RevId: 284858262 Change-Id: I081e26d054d479b9b4f37703fdd48e1b42f025af

view details

Brian Zhao

commit sha 9da481d2004d6b03c9296d162ff9055bdb23904f

Move all windows-specific build targets from tf/core/platform/default/build_refactor.bzl into tf/core/platform/windows/BUILD. Subsequent changes will do the same for tf/core/platform/default/BUILD, and remove build_refactor.bzl. This refactoring is part of the Tensorflow Build Improvements RFC described here: https://github.com/tensorflow/community/pull/179 PiperOrigin-RevId: 284858768 Change-Id: I60dc0b81c6d5e7c81722e33c4d8ce290499e9b54

view details

Dan Moldovan

commit sha ed8765e8fe3ff91f9c83db76b15831e9a0b7b533

Re-apply AutoGraph when the target of a tf.function is wrapped in a call to _call_for_each_replica. PiperOrigin-RevId: 284866253 Change-Id: I9ae9d9f6ad4f36f72b3f81e5c935e02884eee7e3

view details

Robert David

commit sha 66ec4ba062c5ed5093c77da7bde2b0dd92395efa

Stop declaring dynamic allocation related functions when TF_LITE_STATIC_MEMORY is defined. The functions are not defined when this macro is defined. PiperOrigin-RevId: 284866555 Change-Id: I52c002b9e07c8b69bc99def408e440dea5fe89ea

view details

Uday Bondhugula

commit sha 89991d2d5fd264e60834b8553fb53ff9d730e93c

More affine expr simplifications for floordiv and mod Add one more simplification for floordiv and mod affine expressions. Examples: (2*d0 + 1) floordiv 2 is simplified to d0 (8*d0 + 4*d1 + d2) floordiv 4 simplified to 4*d0 + d1 + d2 floordiv 4. etc. Similarly, (4*d1 + 1) mod 2 is simplified to 1, (2*d0 + 8*d1) mod 8 simplified to 2*d0 mod 8. Change getLargestKnownDivisor to return int64_t to be consistent and to avoid casting at call sites (since the return value is used in expressions of int64_t/index type). Signed-off-by: Uday Bondhugula <uday@polymagelabs.com> Closes #202 COPYBARA_INTEGRATE_REVIEW=https://github.com/tensorflow/mlir/pull/202 from bondhugula:affine b13fcb2f1c00a39ca5434613a02408e085a80e77 PiperOrigin-RevId: 284866710 Change-Id: Ic48a4477ffd731c4fd81059c5fa37268143a30ad

view details

Yu-Cheng Ling

commit sha b735c799fdb882691ed8a123dc43aa304a1db32e

Blacklist the new TFLite converter for broken internal tests. PiperOrigin-RevId: 284868541 Change-Id: If4340c4c48aa2e2d779bfd4b4f916758498293d4

view details

Mihai Maruseac

commit sha 77b30d97cb2f43070b5febeb8dde8a51453c85dc

Stream in the status to GTEST_SKIP in modular filesystem tests. This change makes it so that whenever a test is skipped because a precondition fails we see exactly what failed. Thus, tests would be easier to debug, especially on Windows or exotic filesystems. There is a small C++ change in the glue code too: `const std::string& foo = ..` should be `std::string foo = ..` everywhere. There were two places where I didn't do this conversion so I'm fixing this mistake now. PiperOrigin-RevId: 284870348 Change-Id: Id0f94437cb6db9d91176097eb1cc104589847f70

view details

A. Unique TensorFlower

commit sha 74229d4736accc47e02ac7d440d931f489260c0e

[Grappler] 1) Skip dead branch elimination for merge nodes with control inputs, since these can create cycles in the resulting optimized graph. 2) Optimize a few utility functions. 3) Add more verbose VLOGging when topological sorting fails. PiperOrigin-RevId: 284871268 Change-Id: I36435402d826e4737b709468d88641d7a7fa2a83

view details

push time in 2 months

push eventbmzhao/tensorflow

A. Unique TensorFlower

commit sha f1527c02b722050822c62386428fd9d0f8623e74

Go: Update generated wrapper functions for TensorFlow ops. PiperOrigin-RevId: 284705676 Change-Id: Ib0206485ca591f28b0beb8c007ad4e6a842e5fe3

view details

Brian Zhao

commit sha ead06270dc8bd7a61622f4c687396ce1791f29b5

Adding tensorflow/core/platform/default/BUILD and tensorflow/core/platform/windows/BUILD. This is part of the refactoring described in the Tensorflow Build Improvements RFC: https://github.com/tensorflow/community/pull/179 Subsequent changes will migrate targets from build_refactor.bzl into the new BUILD files. PiperOrigin-RevId: 284712709 Change-Id: I650eb200ba0ea87e95b15263bad53b0243732ef5

view details

A. Unique TensorFlower

commit sha 829c458184e1b1992a70b30b26180db92d97201b

Add tensorflow/core/util/BUILD, which currently just exports the files in that directory and adds tf_proto_library rules where necessary. PiperOrigin-RevId: 284716215 Change-Id: I0f90fe1b26a6fbddfa90005aceb0da21eecd707a

view details

Yanan Cao

commit sha e3d30a643d3a1d3973fb764e7a534c299f03a46e

[NFC] Fix ResourceApplyAdam comment regarding formula of using/not using Nesterov computation PiperOrigin-RevId: 284716462 Change-Id: I6307d847ae6280932b8a4ead93a2cc3b1370533a

view details

A. Unique TensorFlower

commit sha eb1aab2c5accecfa8d9400619367049cd3a33156

compat: Update forward compatibility horizon to 2019-12-10 PiperOrigin-RevId: 284719413 Change-Id: I2cda7d490617a3bc64e35a332af6eb6326718094

view details

A. Unique TensorFlower

commit sha e65a6765b412c2e0a808319cf75d0ce3d12bb2dd

Drop Markdown style annotations These come from a non-standard extenion that is not available on Github, so it only clutters the documentation source with {.mlir} or {.ebnf} tags. PiperOrigin-RevId: 284733003 Change-Id: Ie161585584ea6bfcb99f5bf3aef94cb9de4eb2b7

view details

Adrian Kuegel

commit sha f4e5f11c8a4ccb95799a43ebb73b995782a270fe

Use session_options.config.device_count() in XlaGpuDeviceFactory::CreateDevices Otherwise we can end up initializing GPUs even though users specifically hid all GPUs using set_visible_devices. Fixes issue #33168. PiperOrigin-RevId: 284756759 Change-Id: I2a4064f121b4e000fbd5f6353e04336f8f4844f9

view details

Shanqing Cai

commit sha d4ca02dd00d227011c367d1a001e2449687111be

[tfdbg] Exclude DestroyResourceOp from debugger's dumping of eager op execution This removes a deadlock that happens in certain test configurations, e.g., DistributionStrategy / OneDeviceStrategy. PiperOrigin-RevId: 284757549 Change-Id: I89f26718f4b3ec9c12ab867ef4ec1451c553f5d8

view details

A. Unique TensorFlower

commit sha a9469c4002554cbf5c9ae235fca0f88d21c80cdf

Go: Update generated wrapper functions for TensorFlow ops. PiperOrigin-RevId: 284759036 Change-Id: Ib1400dc116764ac0aafc90951aaa4df6732868f1

view details

Mehdi Amini

commit sha b6a81c1a078449e6f5a1e961f023d609a8c35cf8

Add a doc on guidelines for contributing a new dialect to the MLIR core repo Closes #263 PiperOrigin-RevId: 284760931 Change-Id: I7af2abb3fb8b4ad51999734cd1c95a26226675aa

view details

Derek Murray

commit sha c0b26f25e4dfb59525dc0a9cbdbb349fcaac843f

Avoid allocating std::vector (and sometimes allocating InlinedVector) on eager op execution. The input and output AllocatorAttributes vectors can be computed once when the KernelAndDeviceOp is created. PiperOrigin-RevId: 284761081 Change-Id: I71e3effea7baa307fcc2b8debdf5114b4d6fade1

view details

Meghna Natraj

commit sha 3dedec08128b47218ced2a6186d9022d090da21b

Port MEAN op to micro - Part 1. Separate reference code under a new file reduce.h PiperOrigin-RevId: 284767311 Change-Id: I3dec31055a97a283afaaca3b7b86f5d52d8cbc56

view details

A. Unique TensorFlower

commit sha 84add98ff286a5b805d6b45d016f484b7c4f42ba

Skip over optional bias tensors. PiperOrigin-RevId: 284767966 Change-Id: Ib198cf71b970504dd424c0a0db487bd3103012af

view details

A. Unique TensorFlower

commit sha c733a2c1cdf7be81336fbf7a50df8e359eac94c6

Expanded the description of tf.gfile. PiperOrigin-RevId: 284775354 Change-Id: I3e6c8e785823251d2123613d9063c371ead329ad

view details

Derek Murray

commit sha 8afbeffd03ddec5ab57719fd186720a9008b52d8

Cache the return value of MatmulDoFP32ComputationFP16Input() (which reads an environment variable). Previously, we read and parsed an environment variable every time a Matmul kernel was launched on the GPU device. PiperOrigin-RevId: 284778912 Change-Id: I4fbd791fd160f2de699745186089e372a1341deb

view details

A. Unique TensorFlower

commit sha 83b7ca55aa33e60edd85222ca620c3b7d2074532

Fix error in comment 'Free' makes no sense in context. It should be 'Finish'. PiperOrigin-RevId: 284782255 Change-Id: Ie7cafd138f8426f499076b7bd9e23eb965472ce2

view details

A. Unique TensorFlower

commit sha 8c49b5cc8acd518facc8092a0634371a7b74d312

Add pattern for QuantizeAndDequantizeV2Op. PiperOrigin-RevId: 284787096 Change-Id: I3736228b61ee3d935bad7d1be43a12e3eee60b50

view details

Abdurrahman Akkas

commit sha 8a5b70bfa05fed05114a918098d14aa0b4bb9127

Fixes the requantization logic that assumes the user of the requantized op is TFL::QuantizeOp. PiperOrigin-RevId: 284788324 Change-Id: I85f74fe71556491edf1e60c03e6f02167918126b

view details

Taylor Robie

commit sha aed40c077bc79212c095e56059399ab5b0be1886

Remove name-based Variable handling in keras Lambda layers, and add detailed exceptions and warnings for unsafe corner cases. PiperOrigin-RevId: 284789654 Change-Id: Ie13c0505f89730114021ac80e21e2b11590af97f

view details

Andy Ly

commit sha a88ff06438eec02d74ac3ffe73b1fe3db1eb6c21

Sort `tf.TPUReplicatedInput` ops by index attribute when forming `tf_device.replicate` in TPU cluster formation pass. Padding map relies on index set in `tf.TPUReplicatedInput` ops. To make non negative indices ordering be deterministic, the ops can be sorted by index when added to the `tf_device.replicate` op. Ordering of `tf.TPUReplicatedInput` with an index not set (default -1) can be ignored. PiperOrigin-RevId: 284790304 Change-Id: I830a77c03273cbe7c2cebc21762579c69478f4d6

view details

push time in 2 months

push eventbmzhao/tensorflow

namrata-ibm

commit sha a816b97a61ad152b64f7550351379ccae860d841

Updating metrics_export_meta_graph.pb for fixing //tensorflow/python:framework_meta_graph_test on big endian

view details

A. Unique TensorFlower

commit sha d7e2cd058a059f713b4c455896f1654a223fd909

Go: Update generated wrapper functions for TensorFlow ops. PiperOrigin-RevId: 284381342 Change-Id: I6511060126d6e5db5b6ff631f3a69ce236a3487c

view details

Prakalp Srivastava

commit sha 3fbe4cb36b57f6aab06096efa0e87b80b28060e3

Modify import of GatherOp in HLO dialect and add support to export it to HLO. Dimension number attributes of Gather have been packed into a StructAttr similar to how such attributes are handled for other HLO ops (Scatter, Dot, Convolution). IREE tests needed to be fixed after this modification to GatherOp. Support to export GatherOp to HLO instruction has also been added. PiperOrigin-RevId: 284383252 Change-Id: Ib8a4cf5a8d2fcc055d19a284f400b45ab8105eba

view details

Brian Zhao

commit sha 1521067d9d6ff1e35eacbd0ab73f78e94af9585d

Wiring tensorflow/core/platform:stacktrace and tensorflow/core/platform:stacktrace_handler into tensorflow/core/BUILD. This change also removes tf_additional_lib_srcs, which after bazel starlark expansion amounts to just stacktrace_handler. This change is part of the refactoring described in the Tensorflow Build Improvements RFC: https://github.com/tensorflow/community/pull/179 PiperOrigin-RevId: 284385812 Change-Id: I2ad704e196cf354008a3f7e5132d6bdd9f4569d9

view details

Brian Zhao

commit sha 7d27eeab2cdf217c7fcbcfeb39eb7ba38b52d78f

Wire tensorflow/core/platform:base64 into tensorflow/core/BUILD. This change also removes the glob tensorflow/core/platform:legacy_lib_internal_srcs, and replaces it with the leftover from the glob's expansion. This change is part of the Tensorflow Build Improvements RFC: https://github.com/tensorflow/community/pull/179 PiperOrigin-RevId: 284386433 Change-Id: I052f3381c76fbaea52f3299618c6961ec506fbcc

view details

Brian Zhao

commit sha 4d833636c5001064e759887c8376cfee8e300c3c

Expand the glob for tf_additional_lib_hdrs(). This makes the set of platform-specific headers explicit, and also helps make checking in BUILD files in subdirectories of tf/core/platform easier. This change is part of the Tensorflow Build Improvements RFC: https://github.com/tensorflow/community/pull/179 PiperOrigin-RevId: 284392848 Change-Id: I635d2051bc064d88dbdd8fb146636887adc19dbb

view details

A. Unique TensorFlower

commit sha d9e8ee7a50059a9c9cdbdde898a58004cf51c79b

Go: Update generated wrapper functions for TensorFlow ops. PiperOrigin-RevId: 284408035 Change-Id: Id59e6149e41dae01c89197c5cb105c8c541d7afd

view details

A. Unique TensorFlower

commit sha e5910e72d6415e8a415f6e8c8e6361df18842ee4

compat: Update forward compatibility horizon to 2019-12-08 PiperOrigin-RevId: 284410438 Change-Id: Idbc0745787030cb9e4de3e1c45239611f7c9a4dd

view details

A. Unique TensorFlower

commit sha 22e8c972bc3714c068c6bfff9c646f72cd56a614

Go: Update generated wrapper functions for TensorFlow ops. PiperOrigin-RevId: 284415849 Change-Id: Ica0a04dff636ce014ef45f7f9acc618686b87a53

view details

A. Unique TensorFlower

commit sha c91c02353d9c6c1b0c851b10e29beb9be23a7597

Go: Update generated wrapper functions for TensorFlow ops. PiperOrigin-RevId: 284422887 Change-Id: Ic548b157dbaf599c9a652a1d7d4a95c5a811bca2

view details

Lei Zhang

commit sha 3618cffa8be0adeedc8df247026162e06f4a8b63

NFC: Expose constFoldBinaryOp via a header This allows other dialects to reuse the logic to support constant folding binary operations and reduces code duplication. PiperOrigin-RevId: 284428721 Change-Id: Ie35a5b544d852071638c242247cabfbe21c9017d

view details

Lei Zhang

commit sha bba0110c2d4dadab9aeb559f71fca7145dbceb60

Add tf.Unpack to XLA HLO legalization Tightened verification for tf.Unpack along the way. PiperOrigin-RevId: 284429269 Change-Id: Icdc53f7d3593943dec263f70d89d617f44e50f2d

view details

A. Unique TensorFlower

commit sha d9c782939e3fb03b5f1e8d6ae83c185bedc619c3

Go: Update generated wrapper functions for TensorFlow ops. PiperOrigin-RevId: 284429803 Change-Id: I900fe7ad04b61a7daf9da7e74b8dda9a6ace2be4

view details

Brian Zhao

commit sha eead81aa62053b22227d3e32813bbe9373491f43

Removing tensorflow/core/platform:legacy_proto_hdrs, which are no longer used. PiperOrigin-RevId: 284431745 Change-Id: Id821b29919e70665de72e3342c1b75580df36c22

view details

Brian Zhao

commit sha b41fbcbf851bce1fcf462ad8f251756d9a6e166e

Expand the globs for legacy_srcs_no_runtime and legacy_srcs_no_runtime_google, and extract their common files into a single filegroup. This change helps prepare for a subsequent change adding BUILD files to tf/core/platform/default and tf/core/platform/windows. This change is part of the Tensorflow Build Improvements RFC described here: https://github.com/tensorflow/community/pull/179 PiperOrigin-RevId: 284436373 Change-Id: I695ef8042388ad2bc17c763b87f4f7d0e189cd74

view details

Pavithra Vijay

commit sha 03c045e146d441a0db4bd5cc1ecef1517e5c6708

Fix cross entropy losses to use the underlying logits of Softmax layer by backtracking Identity op if any. PiperOrigin-RevId: 284448871 Change-Id: Ic0c6214ba78abaf348c23abdcb9e2543e2d08a38

view details

A. Unique TensorFlower

commit sha d95e97aef9bc589d9b2a08bfc806e3c5f48fcc90

Go: Update generated wrapper functions for TensorFlow ops. PiperOrigin-RevId: 284449378 Change-Id: I0ec2f3fa57c07666a1ade8ae38b8a54db1d9e0d2

view details

Nicolas Vasilache

commit sha 7be978f09ba90de42445487a0f6351cba5427c0c

Add a layer of recursive matchers that compose. This CL adds support for building matchers recursively. The following matchers are provided: 1. `m_any()` can match any value 2. `m_val(Value *)` binds to a value and must match it 3. `RecursivePatternMatcher<OpType, Matchers...>` n-arity pattern that matches `OpType` and whose operands must be matched exactly by `Matchers...`. This allows building expression templates for patterns, declaratively, in a very natural fashion. For example pattern `p9` defined as follows: ``` auto mul_of_muladd = m_Op<MulFOp>(m_Op<MulFOp>(), m_Op<AddFOp>()); auto mul_of_anyadd = m_Op<MulFOp>(m_any(), m_Op<AddFOp>()); auto p9 = m_Op<MulFOp>(m_Op<MulFOp>( mul_of_muladd, m_Op<MulFOp>()), m_Op<MulFOp>(mul_of_anyadd, mul_of_anyadd)); ``` Successfully matches `%6` in: ``` %0 = addf %a, %b: f32 %1 = addf %a, %c: f32 // matched %2 = addf %c, %b: f32 %3 = mulf %a, %2: f32 // matched %4 = mulf %3, %1: f32 // matched %5 = mulf %4, %4: f32 // matched %6 = mulf %5, %5: f32 // matched ``` Note that 0-ary matchers can be used as leaves in place of n-ary matchers. This alleviates from passing explicit `m_any()` leaves. In the future, we may add extra patterns to specify that operands may be matched in any order. PiperOrigin-RevId: 284469446 Change-Id: I8c941801737a6168046aa08b72ae1fc89e1e232c

view details

Xunkai Zhang

commit sha 0e0c574b6ea04a7fbae16eabe99c02350cc5d462

Add QuantizeOp, DequantizeOp and CastOp. PiperOrigin-RevId: 284476659 Change-Id: I038b7a5a91ef8d15981c2bbfa8d9d7b83a8f27e4

view details

Smit Hinsu

commit sha 8295a0ea930f9d05ae6911704cdc032e3000e859

Diable zip tests brokey by forward compat change PiperOrigin-RevId: 284478538 Change-Id: I77a89c35ff11c29e819710f55afc29b5982bafa5

view details

push time in 3 months

Pull request review commenttensorflow/community

RFC: TensorFlow Build Improvements

+# Tensorflow Build Improvements++| Status        | Proposed                         |+| :------------ | :------------------------------- |+| **Author(s)** | Brian Zhao (bmzhao@google.com)   |+| **Sponsor**   | Gunhan Gulsoy (gunan@google.com) |+| **Updated**   | 2019-11-22                       |++## Objective++Simplify Tensorflow’s build so that+1. Builds are more granular+2. Builds are easier to reason about+3. TF’s build is well positioned to immediately benefit from [Bazel shared library support](https://docs.google.com/document/d/13nolQXvXDsJ4yjWP1FRd-WscP2-rHZcR3Zb__js6mLA)++## Background++Tensorflow was originally [open sourced 4 years ago](https://github.com/tensorflow/tensorflow/tree/f41959ccb2d9d4c722fe8fc3351401d53bcf4900), with official build support using [Bazel 0.1.0](https://github.com/tensorflow/tensorflow/blob/f41959ccb2d9d4c722fe8fc3351401d53bcf4900/tensorflow/g3doc/get_started/os_setup.md#install-bazel). Since then, Tensorflow's code base has grown [significantly](https://github.com/tensorflow/tensorflow/pulse/monthly) and Bazel has also matured, [recently hitting 1.0](https://blog.bazel.build/2019/10/10/bazel-1.0.html).++With Bazel's new [stability guarantees](https://blog.bazel.build/2019/06/06/Bazel-Semantic-Versioning.html) and experimental [shared library support on the horizon](https://github.com/bazelbuild/bazel/commit/d6c8766e3ad7042493947ea4f6f8cab0ec8c7426), we would like to start paying down the technical debt left over in Tensorflow's BUILD files, with the goals of increasing build granularity and making Tensorflow's build easier to understand. We expect this refactoring effort to facilitate long term efforts, such as [modular tensorflow](https://github.com/tensorflow/community/pull/77/files?short_path=4497fa7#diff-4497fa7db29551a9463c18c2acb81644)/[modular filesystems](https://github.com/tensorflow/community/pull/101/files?short_path=3029ddb#diff-3029ddb33c0ceefbe5430e1a4f3be93b), and generally improve the TF developer workflow. This refactoring also provides an opportunity for contributors to make meaningful improvements to Tensorflow without deep prior knowledge of the codebase.++## Vocabulary++We assume the reader is familiar with [bazel build terminology](https://docs.bazel.build/versions/master/build-ref.html). In addition, we use the capitalized term "BUILD" to refer to bazel build files, eg: [tensorflow/core/BUILD](https://github.com/tensorflow/tensorflow/blob/master/tensorflow/core/BUILD). This is distinct from lowercase "build" which refers to the process of compiling tensorflow code.++## Overview++Tensorflow's build has 4 key pain points:++1.  BUILD targets are [large](https://github.com/tensorflow/tensorflow/blob/500438a24419f8f0921bc8bb52d1951e3ee4235e/tensorflow/core/BUILD#L630-L723), and often use [globs](https://github.com/tensorflow/tensorflow/blob/500438a24419f8f0921bc8bb52d1951e3ee4235e/tensorflow/core/BUILD#L2654-L2666)++2.  BUILD files are large ([>5000 lines](https://github.com/tensorflow/tensorflow/blob/500438a24419f8f0921bc8bb52d1951e3ee4235e/tensorflow/core/BUILD))++3.  [Multiple](https://github.com/tensorflow/tensorflow/blob/500438a24419f8f0921bc8bb52d1951e3ee4235e/tensorflow/core/BUILD#L1572-L1582)  [targets](https://github.com/tensorflow/tensorflow/blob/500438a24419f8f0921bc8bb52d1951e3ee4235e/tensorflow/core/BUILD#L2654-L2666) advertise the same source files++4.  Lack of Bazel Shared Library support requires TF to maintain custom workarounds++### Large Targets++Many of Tensorflow's most commonly used build targets contain hundreds of files of source code each. For example, [tensorflow/core:framework_internal_impl](https://github.com/tensorflow/tensorflow/blob/f2f5802303c460e641a704045ae30fba7be1f8c7/tensorflow/core/BUILD#L2657-L2672) has over 200 translation units. Initially, these large targets were an expedient way to get a minimal working build in Bazel; but they are now slowing down development.++#### Wasted Machine Cycles++First, large targets cause wasted compilation + linking of code. Consider the target [tensorflow/c:tf_status_internal](https://github.com/tensorflow/tensorflow/blob/a87976312c35253071c3ea5adb79b38bd035ccc5/tensorflow/c/BUILD#L164), which has a dependency on [tensorflow/core:lib](https://github.com/tensorflow/tensorflow/blob/a87976312c35253071c3ea5adb79b38bd035ccc5/tensorflow/c/BUILD#L177). Building tf_status_internal should minimally only require [tensorflow/core/lib/status.{h,cc}](https://github.com/tensorflow/tensorflow/blob/a87976312c35253071c3ea5adb79b38bd035ccc5/tensorflow/c/tf_status_internal.h#L19), and status' transitive dependencies.++![tf_status_internal_actual_dependency_graph](20191122-tf-build-tlc/tf_status_internal_actual_dependency_graph.png)++However, whenever bazel compiles tensorflow/c:tf_status_internal, bazel actually compiles all of tensorflow/core/platform/** and tensorflow/core/lib/** because tensorflow/core:lib contains all of them.++![tf_status_internal_bazel_build_graph](20191122-tf-build-tlc/tf_status_internal_bazel_build_graph.png)++When building a binary, this also results in extra binary size or extra work by the linker to strip unnecessary code. Furthermore, large targets prevent us from pruning the set of affected tests we run on a given change, adding extra burden on CI infrastructure.++#### Wasted Dev Cycles++From a developer workflow perspective, TF's large targets introduce challenges for both newcomers and veteran contributors.++As a case study, suppose we wanted to make a modification to [tensorflow/core/util/batch_util.cc](https://github.com/tensorflow/tensorflow/blob/09d58c60128694888e2d45b513466aca754a6242/tensorflow/core/util/batch_util.cc), that required adding the line++```C+++#include "tensorflow/core/lib/monitoring/gauge.h"+```++How would we reflect this change in bazel's build graph? Finding the build target that contains batch_util.cc isn't obvious, since grepping for batch_util.cc in the closest parent BUILD file doesn't turn up anything. TF developers have to be aware of the globbing in "[framework_internal_impl](https://github.com/tensorflow/tensorflow/blob/fd4ad2ebdc5fe5113b43e2586d102e624b0f1de4/tensorflow/core/BUILD#L2658)". Similarly the build target for monitoring/gauge.{h,cc}, is also hidden in a large super-target: [tensorflow/core:lib](https://github.com/tensorflow/tensorflow/blob/fd4ad2ebdc5fe5113b43e2586d102e624b0f1de4/tensorflow/core/BUILD#L514). After checking the deps of our [framework_internal_impl](https://github.com/tensorflow/tensorflow/blob/fd4ad2ebdc5fe5113b43e2586d102e624b0f1de4/tensorflow/core/BUILD#L2658) target, we'd realize that tensorflow/core:lib is already [included as a dependency](https://github.com/tensorflow/tensorflow/blob/fd4ad2ebdc5fe5113b43e2586d102e624b0f1de4/tensorflow/core/BUILD#L2709), and that this should be a no-op.++Tensorflow's coarse build structure causes simple single-target builds to compile a large portion of the codebase. Since these builds frequently take 30 minutes to > 1 hour, developer iteration is stymied.++Tensorflow's build structure is both difficult to understand for new TF developers, and slow to iterate on for experienced contributors.++### Large Build Files++Large BUILD files also reflect opportunities for code hygiene and organization improvement. Monolithic BUILD files allow coupling between independent parts of Tensorflow's build graph. For example, tf/core/BUILD alone contains targets with sources across [tf/core/common_runtime/](https://github.com/tensorflow/tensorflow/blob/9b94c27ef6f6244be65340c05f2a5a92edd2eb5f/tensorflow/core/BUILD#L2899), [tf/core/framework/*](https://github.com/tensorflow/tensorflow/blob/9b94c27ef6f6244be65340c05f2a5a92edd2eb5f/tensorflow/core/BUILD#L635), and [tf/core/util/*](https://github.com/tensorflow/tensorflow/blob/9b94c27ef6f6244be65340c05f2a5a92edd2eb5f/tensorflow/core/BUILD#L369).++Since the smallest atom of visibility surface area is a [single package](https://docs.bazel.build/versions/master/be/common-definitions.html), and since tensorflow packages span several large subdirectories of code, code across directories can easily start depending on each other in nontrivially coupled ways, including cycles.++The simplest way to prevent this is to define more granular packages, which means adding more BUILD files, so that each BUILD file contains less targets.++### Intersecting Build Targets++When multiple build targets contain the same source files, they form parallel build graphs. This is currently the case with the build targets [//tensorflow/core:mobile_srcs_no_runtime](https://github.com/tensorflow/tensorflow/blob/500438a24419f8f0921bc8bb52d1951e3ee4235e/tensorflow/core/BUILD#L1572-L1582) and [//tensorflow/core:framework_internal_impl](https://github.com/tensorflow/tensorflow/blob/500438a24419f8f0921bc8bb52d1951e3ee4235e/tensorflow/core/BUILD#L2654-L2666) with respect to any C++ files in tensorflow/core/util/*. Any build target that accidentally includes both of these targets in its dependencies will have [ODR violations](https://en.wikipedia.org/wiki/One_Definition_Rule).++This parallel build graph exists because Android would like to produce a minimal Tensorflow shared object by pulling in only a subset of Tensorflow's codebase. The reason why Android couldn't simply use the existing Tensorflow Build targets is because they are [too large](#large-targets). Therefore, if we address [problem 1](#large-targets), we should be able to refactor the Android build to re-use these new smaller targets.++### Shared Library Workarounds++Until recently, Bazel has had no way to [partition the linking of dependencies](https://github.com/bazelbuild/bazel/issues/492) in a shared object. On the other hand, Tensorflow requires granular control over shared object linking in order to build shared object artifacts such as [libtensorflow_framework.so, pywrap_tensorflow.so, and libtensorflow.so](https://github.com/tensorflow/tensorflow/blob/500438a24419f8f0921bc8bb52d1951e3ee4235e/tensorflow/BUILD#L525-L548). Many of the design choices in Tensorflow's build are workarounds for this missing feature.++#### Header Only Targets++One such workaround is Tensorflow's reliance on "[header-only targets](https://github.com/tensorflow/tensorflow/blob/500438a24419f8f0921bc8bb52d1951e3ee4235e/tensorflow/core/BUILD#L515)" to control linkage. To clarify, we use "header-only" in the sense that the bazel build target does not contain corresponding ".cc" files in its [srcs attribute](https://docs.bazel.build/versions/master/be/c-cpp.html#cc_library), and not in the sense that the header contains the implementation details inline.++A commonly recurring pattern in TF is to have large header-only targets, which expose a set of public headers, eg: //tensorflow/core:lib. These targets then depend on an intermediate header-only target that contains additional non-public header files: eg: [//tensorflow/core:lib_internal](https://github.com/tensorflow/tensorflow/blob/7f099a45b09c28d9afa240cb3df018e0efc61d8c/tensorflow/core/BUILD#L2173). This intermediate target finally conditionally depends on a target that includes all corresponding .cc implementation files, via an ["if_static" macro](https://github.com/tensorflow/tensorflow/blob/7f099a45b09c28d9afa240cb3df018e0efc61d8c/tensorflow/core/BUILD#L2194), eg: [//tensorflow/core:lib_internal_impl](https://github.com/tensorflow/tensorflow/blob/7f099a45b09c28d9afa240cb3df018e0efc61d8c/tensorflow/core/BUILD#L2198).++What this ultimately tries to express is shown below:++![tf_static_linking_build_flow.png](20191122-tf-build-tlc/tf_static_linking_build_flow.png)++![tf_dynamic_linking_build_flow.png](20191122-tf-build-tlc/tf_dynamic_linking_build_flow.png)+++Basically, if you statically compile Tensorflow into your library/binary, the header only targets will eventually depend on a target containing actual ‘.cc’ files, causing static linking of parts of libtensorflow_framework. If you want to build Tensorflow using shared objects, Tensorflow’s build cuts the dependency edge between header only targets and their implementation, which prevents the build from statically linking in sources that would otherwise be built into libtensorflow_framework.so.++#### Custom Starlark++However, bazel does not support this workflow natively. Tensorflow has worked around this using several [thousands of lines of starlark macros and rules](https://github.com/tensorflow/tensorflow/blob/500438a24419f8f0921bc8bb52d1951e3ee4235e/tensorflow/tensorflow.bzl).++One example, quoted from @[allenlavoie](https://github.com/allenlavoie), relates to the dynamic linking of libtensorflow_framework, shown above.++> Bazel will not link a binary against an explicit shared object (cc_binary with linkshared=1) unless it is directly included in the srcs of that binary. Including it in the srcs/deps of a library and depending on the library does not work (Bazel does this because otherwise it would end up implicitly duplicating static initializers). Thus we have our own tf_cc_binary and tf_cc_test rules which add libtensorflow_framework.so to the binary's srcs++Another workaround example is TF's forked version of [cc_proto_library](https://github.com/tensorflow/tensorflow/blob/500438a24419f8f0921bc8bb52d1951e3ee4235e/tensorflow/core/platform/default/build_config.bzl#L125-L247), which creates header-only targets. Ideally Tensorflow would simply use [protobuf](https://github.com/protocolbuffers/protobuf/blob/9d61eada0f47d7be793983638c4a29707b192d0c/protobuf.bzl#L227-L319)'s existing cc_proto_library.++These layers of custom starlark increase the barrier to entry for tensorflow developers and force developers to reason with custom, non-standard build behavior.++## Detailed Design++To address these pain points, we propose changing Tensorflow's BUILD structure in the incremental strategy described below. We use [tensorflow/core/BUILD](https://github.com/tensorflow/tensorflow/blob/543b91ac90a0c2fb55ba3e802760f520134bdef1/tensorflow/core/BUILD) as a concrete example.++### Phase 1: Add More Build Files++First, we will start adding BUILD files under subdirectories of tensorflow where they are currently missing. However, these BUILD files will create new [package boundaries](https://docs.bazel.build/versions/master/build-ref.html#packages). This will break the existing build because++1.  packages cannot reference cross-package files unless they are [explicitly exported](https://docs.bazel.build/versions/master/be/functions.html#exports_files)++2.  [globs](https://docs.bazel.build/versions/master/be/functions.html#glob) cannot cross package boundaries++For tensorflow/core/BUILD, we will introduce BUILD files in:++-   [tensorflow/core/common_runtime](https://github.com/tensorflow/tensorflow/tree/master/tensorflow/core/common_runtime)

Hi Clayne,

That's a great point; so far the changes we've made don't touch anything MKL specific, but as we start refactoring core/common_runtime we will need to be careful not to break MKL integration.

My understanding is that Tensorflow builds use MKL by default via Eigen: https://github.com/tensorflow/tensorflow/blob/d26ba01420ed9e4e1a55f22ca0e3034fdd3d3671/tensorflow/core/kernels/BUILD#L756-L758, meaning hopefully our CI will start yelling at us if a breakage occurs; but I'll double check with @gunan

bmzhao

comment created time in 3 months

push eventbmzhao/tensorflow

Michal W. Tarnowski

commit sha 084000813642779063a1701b621e86823da5121b

Non-broadcast Div optimized

view details

Michal W. Tarnowski

commit sha 43a06104a6f3ad4e93099b7f1750948056dd47a7

Explicit NEON typenames removed

view details

Clayne Robison

commit sha 302358459136384bb23c075b633f4c0e3159df49

Add Dockerfile partials to support Mkl + MPI + Horovod; Remove trailing whitespace from python.partial.Dockerfile

view details

Clayne Robison

commit sha 953a5deffab5b9b4fd652cf561d479d2066eb18e

Adding the generated horovod Dockerfiles

view details

Clayne Robison

commit sha cd4ae94127ab68e00c6f615b0fabbc3c268fc3f5

Moving files to mkl_horovod folders to reflect usage

view details

amoitra

commit sha e07dfe6a7091db80bef5c94d10eaf6360d7224de

Set xla_gpu_use_cudnn_batchnorm to true

view details

Michal W. Tarnowski

commit sha 3aacf55547142de592e02113acb6a7c15538adcb

Merge branch 'master' into float-div-neon

view details

Michal W. Tarnowski

commit sha ecb58fc3ee4ca637fed579d3d125a31852df66b5

Merge branch 'master' into float-div-neon

view details

Deven Desai

commit sha 1a489f5c03edc60e26b10fd8a3d13ef03711f50f

[ROCm] Fix for the broken ROCm CSB. The following commit breaks the --config=rocm build https://github.com/tensorflow/tensorflow/commit/bf9c196f37b9cbb3109b2891aaf9da85bf5f712a The above commit adds support for complex type in the optimizers. Complex types are not supported on the ROCm platform. Support for it needs to be excluded on the ROCm platform, and that is what this "fix" does.

view details

amoitra

commit sha b64e97b4448b86bafc8dc72d6a5db64e15dfe30e

Always expand batchnorm inference

view details

amoitra

commit sha 400bba1be9dfb1518b7a0748041d279532b7f7c0

Add comments

view details

Kazuaki Ishizaki

commit sha 340d3337d86c08911c6abce34ec0e449411d223e

minor spelling tweaks

view details

Pooya Davoodi

commit sha d1f49f699f0691f349b10e4e75ade0a36b712af0

Move TensorRT builder configs to converter build function

view details

Pooya Davoodi

commit sha 9d5cbf4e94e08f0e598a36fbc162417d7b031b66

Move builder to Converter class

view details

Pooya Davoodi

commit sha d6dbc7237fb0a0665911d7efa82fa60231a7bcb2

Fix function signatures

view details

Mohamed Nour Abouelseoud

commit sha 51cec9092e73f22cabaca3c7e0ba7154b1a4b27d

[Lite] Support Int8 Unpack Operator Added support for Unpack Operator Added relevant tests.

view details

Duncan Riach

commit sha 7fd47f10140a3965284f700f4a7cfea62a45019f

Improve resize_bilinear CPU back-prop kernel comment

view details

Duncan Riach

commit sha b0c6ea8fdeb0cc2ffa892362c6ee6b41556c185b

Fix small typo

view details

Lukas Geiger

commit sha 860666581f52b75ccb4ec283f99546b86d955925

Fix TensorFlow pip API generation

view details

Michal Tarnowski

commit sha 4d9297306254d2584c79b08fb43a2eaf705e9771

Added comment on kNewtonSteps

view details

push time in 3 months

issue openedbazelbuild/rules_cc

cc_library linked_statically_by attribute Only Works With Absolute Label Paths

Description of the problem / feature request:

The "linked_statically_by" attribute of cc_library errors out when used with relative labels

ERROR: /usr/local/google/home/bmzhao/Code/experimental/BUILD:11:1: in cc_shared_library rule //:a_so: Traceback (most recent call last): File "/usr/local/google/home/bmzhao/Code/experimental/BUILD", line 11 cc_shared_library(name = 'a_so') File "/usr/local/google/home/bmzhao/.cache/bazel/_bazel_bmzhao/346f8c89ce367ab1eaf00095a9f73200/external/rules_cc/examples/expe rimental_cc_shared_library.bzl", line 216, in _cc_shared_library_impl _filter_inputs(ctx, <4 more arguments>) File "/usr/local/google/home/bmzhao/.cache/bazel/_bazel_bmzhao/346f8c89ce367ab1eaf00095a9f73200/external/rules_cc/examples/expe rimental_cc_shared_library.bzl", line 172, in _filter_inputs fail(<1 more arguments>) We can't link //:a either statically or dynamically ERROR: Analysis of target '//:a_so' failed; build aborted: Analysis of target '//:a_so' failed; build aborted

Bugs: what's the simplest, easiest way to reproduce this bug? Please provide a minimal example if possible.

I've got a working repro here: https://gist.github.com/bmzhao/02c695048cbd591bd93d9bdfd93e4f78 based on github repo: https://github.com/bmzhao/experimental/tree/f727f572395b2a72e119f9865b05716bcafea54d

If you comment out line 8, and uncomment line 7, things start working: https://github.com/bmzhao/experimental/blob/f727f572395b2a72e119f9865b05716bcafea54d/BUILD#L8

What operating system are you running Bazel on?

Debian Linux

What's the output of bazel info release?

release 1.2.0

What version of rules_cc do you use? Can you paste the workspace rule used to fetch rules_cc? What other relevant dependencies does your project have?

http_archive( name = "rules_cc", sha256 = "cf3b76a90c86c0554c5b10f4b160f05af71d252026b71362c4674e2fb9936cf9", strip_prefix = "rules_cc-01d4a48911d5e7591ecb1c06d3b8af47fe872371", urls = ["https://github.com/bazelbuild/rules_cc/archive/01d4a48911d5e7591ecb1c06d3b8af47fe872371.zip"], )

rules_cc is pinned to commit: https://github.com/bazelbuild/rules_cc/commit/01d4a48911d5e7591ecb1c06d3b8af47fe872371

created time in 3 months

push eventbmzhao/experimental

Brian Zhao

commit sha f727f572395b2a72e119f9865b05716bcafea54d

Relative target path bug.

view details

push time in 3 months

push eventbmzhao/tensorflow

Dr. Kashif Rasul

commit sha a6c60966ebaeb56f43343645ced03b3b8b44a5ed

Fixed cuda runtime deprecation warning In cuda 10.0 `memoryType` attribute is deprecated in favour of `type`.

view details

Kashif Rasul

commit sha 9d86d97f05ead7bcf936ad5295a1ac93278f5693

fixed tests

view details

I-Hong

commit sha 52c162eed600211b27c8ffa2d6a4b1f2f1794bb1

add api doc

view details

I-Hong

commit sha 6c5124ea6b4a73d2e5bfcae9501541d5e6f49ddd

pylint check

view details

Dr. Kashif Rasul

commit sha d7676c2dd1e64826664ebc951ccc77c339b6e595

Merge branch 'master' into patch-2

view details

autoih

commit sha 8fe63d915710d9d1a2339d98661952347ac56526

indentation

view details

Koan-Sin Tan

commit sha 27e206450bb8017a67bb3f9f629c7655373888a6

[xla] fix xla build on cuda devices without nccl Some cuda devices, such as Jetson devices, do not support NCCL. Building `@local_config_nccl//:nccl` on such kind of devices will cause problem.

view details

autoih

commit sha fda47c3c99cfa80a9eb93d9d272f58c079843f03

indentation

view details

autoih

commit sha 5a43c9b0f197bc9166b9ebda905d34aa9f93a98c

add doctest p1

view details

autoih

commit sha 6b190d840e39729fa6150c4af90ed2414fdfdcab

unique doctest

view details

autoih

commit sha 53f6c49ec25cd407352ae7b843cc7c01ede49a35

unique counts

view details

autoih

commit sha d7037ef7ecd2c7d91b6940f99084bbd315ba4783

setdiff1d TF2

view details

autoih

commit sha c6d3a7ed629fc277f1817ceb87a8b4c412b12d8e

remove global tf

view details

autoih

commit sha 6f05e11e8eb31fd63a6c36105bff398a93ee5d1d

remove carets

view details

autoih

commit sha fe5e28ba6121f06a6d689198a8f65efe9ce60024

remove .

view details

autoih

commit sha 5f6af15a5b26e4626596cea9d45dd4fbceed91fa

minor

view details

autoih

commit sha 504c0b5587c704baacbd12df36b6bf884169884b

revise

view details

autoih

commit sha 5bd0a18c08effa5a8a6da3a87011e537f805e1ee

sanity shorten lines

view details

TengLu

commit sha 1c9cd298c7b7942c7c4883d72f7ba0c9d7f2455a

Add FP32 fusion of MatMul and Relu.

view details

Kaixi Hou

commit sha 860fd24df37ac59f1b6e864f90940e32ff57f5c0

support cudnn depthwise conv

view details

push time in 3 months

Pull request review commenttensorflow/community

RFC: TensorFlow Build Improvements

+# Tensorflow Build Improvements++| Status        | Proposed                         |+| :------------ | :------------------------------- |+| **Author(s)** | Brian Zhao (bmzhao@google.com)   |+| **Sponsor**   | Gunhan Gulsoy (gunan@google.com) |+| **Updated**   | 2019-11-22                       |++## Objective++Simplify Tensorflow’s build so that+1. Builds are more granular+2. Builds are easier to reason about+3. TF’s build is well positioned to immediately benefit from [Bazel shared library support](https://docs.google.com/document/d/13nolQXvXDsJ4yjWP1FRd-WscP2-rHZcR3Zb__js6mLA)++## Background++Tensorflow was originally [open sourced 4 years ago](https://github.com/tensorflow/tensorflow/tree/f41959ccb2d9d4c722fe8fc3351401d53bcf4900), with official build support using [Bazel 0.1.0](https://github.com/tensorflow/tensorflow/blob/f41959ccb2d9d4c722fe8fc3351401d53bcf4900/tensorflow/g3doc/get_started/os_setup.md#install-bazel). Since then, Tensorflow's code base has grown [significantly](https://github.com/tensorflow/tensorflow/pulse/monthly) and Bazel has also matured, [recently hitting 1.0](https://blog.bazel.build/2019/10/10/bazel-1.0.html).++With Bazel's new [stability guarantees](https://blog.bazel.build/2019/06/06/Bazel-Semantic-Versioning.html) and experimental [shared library support on the horizon](https://github.com/bazelbuild/bazel/commit/d6c8766e3ad7042493947ea4f6f8cab0ec8c7426), we would like to start paying down the technical debt left over in Tensorflow's BUILD files, with the goals of increasing build granularity and making Tensorflow's build easier to understand. We expect this refactoring effort to facilitate long term efforts, such as [modular tensorflow](https://github.com/tensorflow/community/pull/77/files?short_path=4497fa7#diff-4497fa7db29551a9463c18c2acb81644)/[modular filesystems](https://github.com/tensorflow/community/pull/101/files?short_path=3029ddb#diff-3029ddb33c0ceefbe5430e1a4f3be93b), and generally improve the TF developer workflow. This refactoring also provides an opportunity for contributors to make meaningful improvements to Tensorflow without deep prior knowledge of the codebase.++## Vocabulary++We assume the reader is familiar with [bazel build terminology](https://docs.bazel.build/versions/master/build-ref.html). In addition, we use the capitalized term "BUILD" to refer to bazel build files, eg: [tensorflow/core/BUILD](https://github.com/tensorflow/tensorflow/blob/master/tensorflow/core/BUILD). This is distinct from lowercase "build" which refers to the process of compiling tensorflow code.++## Overview++Tensorflow's build has 4 key pain points:++1.  BUILD targets are [large](https://github.com/tensorflow/tensorflow/blob/500438a24419f8f0921bc8bb52d1951e3ee4235e/tensorflow/core/BUILD#L630-L723), and often use [globs](https://github.com/tensorflow/tensorflow/blob/500438a24419f8f0921bc8bb52d1951e3ee4235e/tensorflow/core/BUILD#L2654-L2666)++2.  BUILD files are large ([>5000 lines](https://github.com/tensorflow/tensorflow/blob/500438a24419f8f0921bc8bb52d1951e3ee4235e/tensorflow/core/BUILD))++3.  [Multiple](https://github.com/tensorflow/tensorflow/blob/500438a24419f8f0921bc8bb52d1951e3ee4235e/tensorflow/core/BUILD#L1572-L1582)  [targets](https://github.com/tensorflow/tensorflow/blob/500438a24419f8f0921bc8bb52d1951e3ee4235e/tensorflow/core/BUILD#L2654-L2666) advertise the same source files++4.  Lack of Bazel Shared Library support requires TF to maintain custom workarounds++### Large Targets++Many of Tensorflow's most commonly used build targets contain hundreds of files of source code each. For example, [tensorflow/core:framework_internal_impl](https://github.com/tensorflow/tensorflow/blob/f2f5802303c460e641a704045ae30fba7be1f8c7/tensorflow/core/BUILD#L2657-L2672) has over 200 translation units. Initially, these large targets were an expedient way to get a minimal working build in Bazel; but they are now slowing down development.++#### Wasted Machine Cycles++First, large targets cause wasted compilation + linking of code. Consider the target [tensorflow/c:tf_status_internal](https://github.com/tensorflow/tensorflow/blob/a87976312c35253071c3ea5adb79b38bd035ccc5/tensorflow/c/BUILD#L164), which has a dependency on [tensorflow/core:lib](https://github.com/tensorflow/tensorflow/blob/a87976312c35253071c3ea5adb79b38bd035ccc5/tensorflow/c/BUILD#L177). Building tf_status_internal should minimally only require [tensorflow/core/lib/status.{h,cc}](https://github.com/tensorflow/tensorflow/blob/a87976312c35253071c3ea5adb79b38bd035ccc5/tensorflow/c/tf_status_internal.h#L19), and status' transitive dependencies.++![tf_status_internal_actual_dependency_graph](20191122-tf-build-tlc/tf_status_internal_actual_dependency_graph.png)++However, whenever bazel compiles tensorflow/c:tf_status_internal, bazel actually compiles all of tensorflow/core/platform/** and tensorflow/core/lib/** because tensorflow/core:lib contains all of them.++![tf_status_internal_bazel_build_graph](20191122-tf-build-tlc/tf_status_internal_bazel_build_graph.png)++When building a binary, this also results in extra binary size or extra work by the linker to strip unnecessary code. Furthermore, large targets prevent us from pruning the set of affected tests we run on a given change, adding extra burden on CI infrastructure.++#### Wasted Dev Cycles++From a developer workflow perspective, TF's large targets introduce challenges for both newcomers and veteran contributors.++As a case study, suppose we wanted to make a modification to [tensorflow/core/util/batch_util.cc](https://github.com/tensorflow/tensorflow/blob/09d58c60128694888e2d45b513466aca754a6242/tensorflow/core/util/batch_util.cc), that required adding the line++```C+++#include "tensorflow/core/lib/monitoring/gauge.h"+```++How would we reflect this change in bazel's build graph? Finding the build target that contains batch_util.cc isn't obvious, since grepping for batch_util.cc in the closest parent BUILD file doesn't turn up anything. TF developers have to be aware of the globbing in "[framework_internal_impl](https://github.com/tensorflow/tensorflow/blob/fd4ad2ebdc5fe5113b43e2586d102e624b0f1de4/tensorflow/core/BUILD#L2658)". Similarly the build target for monitoring/gauge.{h,cc}, is also hidden in a large super-target: [tensorflow/core:lib](https://github.com/tensorflow/tensorflow/blob/fd4ad2ebdc5fe5113b43e2586d102e624b0f1de4/tensorflow/core/BUILD#L514). After checking the deps of our [framework_internal_impl](https://github.com/tensorflow/tensorflow/blob/fd4ad2ebdc5fe5113b43e2586d102e624b0f1de4/tensorflow/core/BUILD#L2658) target, we'd realize that tensorflow/core:lib is already [included as a dependency](https://github.com/tensorflow/tensorflow/blob/fd4ad2ebdc5fe5113b43e2586d102e624b0f1de4/tensorflow/core/BUILD#L2709), and that this should be a no-op.++Tensorflow's coarse build structure causes simple single-target builds to compile a large portion of the codebase. Since these builds frequently take 30 minutes to > 1 hour, developer iteration is stymied.++Tensorflow's build structure is both difficult to understand for new TF developers, and slow to iterate on for experienced contributors.++### Large Build Files++Large BUILD files also reflect opportunities for code hygiene and organization improvement. Monolithic BUILD files allow coupling between independent parts of Tensorflow's build graph. For example, tf/core/BUILD alone contains targets with sources across [tf/core/common_runtime/](https://github.com/tensorflow/tensorflow/blob/9b94c27ef6f6244be65340c05f2a5a92edd2eb5f/tensorflow/core/BUILD#L2899), [tf/core/framework/*](https://github.com/tensorflow/tensorflow/blob/9b94c27ef6f6244be65340c05f2a5a92edd2eb5f/tensorflow/core/BUILD#L635), and [tf/core/util/*](https://github.com/tensorflow/tensorflow/blob/9b94c27ef6f6244be65340c05f2a5a92edd2eb5f/tensorflow/core/BUILD#L369).++Since the smallest atom of visibility surface area is a [single package](https://docs.bazel.build/versions/master/be/common-definitions.html), and since tensorflow packages span several large subdirectories of code, code across directories can easily start depending on each other in nontrivially coupled ways, including cycles.++The simplest way to prevent this is to define more granular packages, which means adding more BUILD files, so that each BUILD file contains less targets.++### Intersecting Build Targets++When multiple build targets contain the same source files, they form parallel build graphs. This is currently the case with the build targets [//tensorflow/core:mobile_srcs_no_runtime](https://github.com/tensorflow/tensorflow/blob/500438a24419f8f0921bc8bb52d1951e3ee4235e/tensorflow/core/BUILD#L1572-L1582) and [//tensorflow/core:framework_internal_impl](https://github.com/tensorflow/tensorflow/blob/500438a24419f8f0921bc8bb52d1951e3ee4235e/tensorflow/core/BUILD#L2654-L2666) with respect to any C++ files in tensorflow/core/util/*. Any build target that accidentally includes both of these targets in its dependencies will have [ODR violations](https://en.wikipedia.org/wiki/One_Definition_Rule).++This parallel build graph exists because Android would like to produce a minimal Tensorflow shared object by pulling in only a subset of Tensorflow's codebase. The reason why Android couldn't simply use the existing Tensorflow Build targets is because they are [too large](https://docs.google.com/document/d/1kg_tVB1g3c5isPaP1OXW_KZwF_hcTiKz5afLkKjiSz8/edit#heading=h.8tfaww3ddb83). Therefore, if we address [problem 1](https://docs.google.com/document/d/1kg_tVB1g3c5isPaP1OXW_KZwF_hcTiKz5afLkKjiSz8/edit#heading=h.ytms1uc3twy), we should be able to refactor the Android build to re-use these new smaller targets.

Ah good catch! Yes these were supposed to links back to headers in the same markdown file. Just fixed, thanks!

bmzhao

comment created time in 3 months

push eventbmzhao/community

Brian Zhao

commit sha 4f0143d34874d56df7a26ca9fe8be6acb91b89f0

RFC: TF Build Improvements

view details

push time in 3 months

push eventbmzhao/tensorflow-windows

Brian Zhao

commit sha 460bc33fcb79e7c07c5d070d225b73e6e120a15f

Tensorflow custom op linking working; test has runtime error of unrecognized op name AudioMicrofrontend.

view details

push time in 3 months

created repositorybmzhao/tensorflow-windows

created time in 3 months

pull request commenttensorflow/community

RFC: TensorFlow Build Improvements

It's great to see this proposal. For DeepSpeech we have similar constraints as Android, we want to keep code size down, so for a while now we've been using very fine grained deps, and for one case we had to break up the TensorFlow target by hand to avoid a large binary size increase from a seemingly small dependency. That change in particular leads to several megabytes difference in our final binary size compared to simply depending on //tensorflow/core/kernels:cwise_op. This proposal seems to be going in the right direction and is much appreciated.

Thank you! We hope this will help make Tensorflow's build more approachable and usable in other projects!

bmzhao

comment created time in 3 months

fork bmzhao/rules_cc

C++ Rules for Bazel

https://bazel.build

fork in 3 months

Pull request review commenttensorflow/community

RFC: TensorFlow Build Improvements

+# Tensorflow Build TLC++| Status        | Proposed                         |+| :------------ | :------------------------------- |+| **Author(s)** | Brian Zhao (bmzhao@google.com)   |+| **Sponsor**   | Gunhan Gulsoy (gunan@google.com) |+| **Updated**   | 2019-11-22                       |++## Objective++Simplify Tensorflow’s build so that+1. Builds are more granular+2. Builds are easier to reason about+3. TF’s build is well positioned to immediately benefit from [Bazel shared library support](https://docs.google.com/document/d/1qYHGcxP9BEi7t9b53r8STr9G3XVtvwy9V-EXqr54Xxg/edit#)

Done (although the original doc link was a public doc).

bmzhao

comment created time in 3 months

Pull request review commenttensorflow/community

RFC: TensorFlow Build Improvements

+# Tensorflow Build TLC

Done.

bmzhao

comment created time in 3 months

push eventbmzhao/community

Brian Zhao

commit sha 7551f8c2aaba8131fcc259f7f9bb6b6c3270a9bd

RFC: TF Build Improvements

view details

push time in 3 months

issue openedbazelbuild/rules_cc

com.google.devtools.build.lib.packages.BuildFileNotFoundException: no such package 'cc' When Using cc_shared_library

Description of the problem / feature request:

Using bazel version 1.2.0 with cc_shared_library causes a bazel error:

ERROR: While resolving toolchains for target //:a_so: com.google.devtools.build.lib.packages.BuildFileNotFoundException: no such package 'cc': BUILD file not found in any of the following directories. Add a BUILD file to a directory to mark it as a package. - /tmp/experimental/cc ERROR: Analysis of target '//:a_so' failed; build aborted: com.google.devtools.build.lib.packages.BuildFileNotFoundException: no such package 'cc': BUILD file not found in any of the following directories. Add a BUILD file to a directory to mark it as a package. - /tmp/experimental/cc INFO: Elapsed time: 1.010s INFO: 0 processes. FAILED: Build did NOT complete successfully (1 packages loaded, 0 targets configured)

Bugs: what's the simplest, easiest way to reproduce this bug? Please provide a minimal example if possible.

https://gist.github.com/bmzhao/311efdfd4306dca02af21b4110ae14bd has a working repro, based off of: https://github.com/bmzhao/experimental/blob/da47bfb740e1729cdf1a9aaa145c4841cc1d004c/BUILD

What operating system are you running Bazel on?

Linux, Debian

What's the output of bazel info release?

release 1.2.0

What version of rules_cc do you use? Can you paste the workspace rule used to fetch rules_cc? What other relevant dependencies does your project have?

https://github.com/bazelbuild/rules_cc/tree/d562dc8046d6251972d18ccfddc3dbfe61e1d79f, which is at commit https://github.com/bazelbuild/rules_cc/commit/d562dc8046d6251972d18ccfddc3dbfe61e1d79f

http_archive( name = "rules_cc", sha256 = "0c5297b7c13b15067180e663a7f8499bdf66b53c33eb53627ec38a1cd57baf96", strip_prefix = "rules_cc-d562dc8046d6251972d18ccfddc3dbfe61e1d79f", urls = ["https://github.com/bazelbuild/rules_cc/archive/d562dc8046d6251972d18ccfddc3dbfe61e1d79f.zip"], )

What Bazel options do you use to trigger the issue? What C++ toolchain do you use?

bazel build --experimental_cc_shared_library :a_so Toolchain was whatever bazel auto-resolved to.

Have you found anything relevant by searching the web?

Nope.

created time in 3 months

create barnchbmzhao/experimental

branch : master

created branch time in 3 months

created repositorybmzhao/experimental

created time in 3 months

startedleandromoreira/digital_video_introduction

started time in 3 months

push eventbmzhao/tensorflow

Albin Joy

commit sha 2cb0e459c4f15fcf37c1e3c8772fc48db298996f

Refactoring comparisons invoke calls

view details

Albin Joy

commit sha 4321f790b0c5773f8c850f4effb54b12e153efa2

Removed macro usage

view details

mrTsjolder

commit sha 8d8ea6b0bdf1806b33445e3d7d42cdb5e4ec1ea9

first steps in cleaning doc citations

view details

mrTsjolder

commit sha 593272265dd97a121cbd7f4f36560420e13fb08b

fix docs for rest training module

view details

mrTsjolder

commit sha d478447e4224dc6ef6ee8936a18ab30d2a0a36f8

Clean up doc citations in ops module

view details

mrTsjolder

commit sha f1a917fd7ac07848ca796e1f7e31bed40f7ddb3d

finish doc citation clean up

view details

mrTsjolder

commit sha 4eb77e6062ad7d7cdebc14e308d494fef020dca1

Revert one-liners to original state

view details

mrTsjolder

commit sha 4d2b7f2a0a21e9e8b1ea501fb1e220b37f002517

get rid of hidden character

view details

Martin Wicke

commit sha d502acb5aea56e1930757bbc9a172898f0adf99c

Merge branch 'master' into unify_doccitations

view details

Siju Samuel

commit sha f620112d4452d8df112f55fa28410a3cf0ab33dd

File resource leak fixed in writer_lib.cc

view details

Siju Samuel

commit sha 6ffb19ff9c3ed5f5a6a77f1e985eb0ca0f221e65

File resource leak fixed in microfrontend/lib/frontend_main.c

view details

Siju Samuel

commit sha cedf5a306accd6fa9fb85f0f6326fb5d2e861ded

File resource leak fixed in microfrontend/lib/frontend_memmap_main.c

view details

Siju Samuel

commit sha 5f48f2e883c67f52af676a8649d13935f697e55a

Deprecated tf.random_normal to tf.random.normal in bidirectional_sequence_lstm_test

view details

Siju Samuel

commit sha 8e8ce66122c285c9081d567518042c76ebc10380

Deprecated tf.random_normal to tf.random.normal in bidirectional_sequence_rnn_test

view details

Siju Samuel

commit sha b2d4be96dec71d7f0360d6122e1b1b05287a1666

Deprecated tf.random_normal to tf.random.normal in unidirectional_sequence_lstm_test

view details

Siju Samuel

commit sha 7a193942cd2356ba318bac66fd4492c727dab7a4

Deprecated tf.random_normal to tf.random.normal in unidirectional_sequence_rnn_test

view details

Siju Samuel

commit sha c5f4fe9b9523fb5fd695d2f2541c3cccaea0b27b

Deprecated tf.random_normal to tf.random.normal in lstm README

view details

autoih

commit sha 661d302b9cc04aaba728ec64d6a4536cff88bf34

Limit the number of statements in exception raising test blocks to 1

view details

Christian Sigg

commit sha 6ffcaf4a1092df1228b59baf5f381d7890e151f9

Add AllReduceOp to GPU dialect with lowering to NVVM. The reduction operation is currently fixed to "add", and the scope is fixed to "workgroup". The implementation is currently limited to sizes that are multiple 32 (warp size) and no larger than 1024. PiperOrigin-RevId: 271290265

view details

Alex Zinenko

commit sha c2f215e0d6e8a2998a2d8c57346fc05c2c48d85e

Drop support for memrefs from JitRunner The support for functions taking and returning memrefs of floats was introduced in the first version of the runner, created before MLIR had reliable lowering of allocation/deallocation to library calls. It forcibly runs MLIR transformation convering affine, loop and standard dialects into the LLVM dialect, unlike the other runner flows that accept the LLVM dialect directly. Memref support leads to more complex layering and is generally fragile. Drop it in favor of functions returning a scalar, or library-based function calls to print memrefs and other data structures. PiperOrigin-RevId: 271330839

view details

push time in 3 months

PR opened tensorflow/community

Reviewers
RFC: TF Build TLC

This RFC will be open for comment until Friday, December 13th, 2019.

cc @gunan @martinwicke

Tensorflow Build TLC

Status Proposed
Author(s) Brian Zhao (bmzhao@google.com)
Sponsor Gunhan Gulsoy (gunan@google.com)
Updated 2019-11-22

Objective

Simplify Tensorflow’s build so that

  1. Builds are more granular
  2. Builds are easier to reason about
  3. TF’s build is well positioned to immediately benefit from Bazel shared library support
+213 -0

0 comment

5 changed files

pr created time in 3 months

push eventbmzhao/community

Brian Zhao

commit sha 77bb5d71643a3b6f79a59fe10605962a9e723f61

RFC: TF Build TLC

view details

push time in 3 months

push eventbmzhao/community

Brian Zhao

commit sha c39f3ca05373b4419bce731a2c35d33fcf13c829

RFC: TF Build TLC

view details

push time in 3 months

push eventbmzhao/community

Brian Zhao

commit sha 651652afc523d24ef4940ba2c16ac0a3f7a59214

RFC: TF Build TLC

view details

push time in 3 months

push eventbmzhao/community

Brian Zhao

commit sha f9a655a3a5dc99ff19472be850755a66c7245e74

RFC: TF Build TLC

view details

push time in 3 months

push eventbmzhao/community

Brian Zhao

commit sha 262f1074b61577a2440b82de1480f6a871a59256

RFC: TF Build TLC

view details

push time in 3 months

fork bmzhao/community

Stores documents used by the TensorFlow developer community

fork in 3 months

startedmstorsjo/msvc-wine

started time in 3 months

fork bmzhao/protobuf

Protocol Buffers - Google's data interchange format

https://developers.google.com/protocol-buffers/

fork in 3 months

issue closedtensorflow/tensorflow

Build failure: undefined reference to protobuf symbols

<em>Please make sure that this is a build/installation issue. As per our GitHub Policy, we only address code/doc bugs, performance issues, feature requests and build/installation issues on GitHub. tag:build_template</em>

System information

  • OS Platform and Distribution (e.g., Linux Ubuntu 16.04): Linux Ubuntu 18.04.3
  • Python version: Tried with python 3.7.3 and python 3.8
  • Installed using virtualenv? pip? conda?: conda python 3.7.3 and virtualenv python 3.8
  • Bazel version (if compiling from source): 0.29.1
  • GCC/Compiler version (if compiling from source): 7.4.0
  • CUDA/cuDNN version: CUDA 10.0 and cuDNN 7.6.5
  • GPU model and memory: Nvidia RTX 2080 Ti

Describe the problem Build fails most of the way in to build.

Provide the exact sequence of commands / steps that you executed before running into the problem

git clone https://github.com/tensorflow/tensorflow.git cd tensorflow git checkout -b mybranch (make up a branch to checkout head) bazel build --config=opt --config=cuda --config=v2 --cxxopt="-D_GLIBCXX_USE_CXX11_ABI=0" //tensorflow/tools/pip_package:build_pip_package

Here is the last part of the terminal's output (attached text file): tensorflow_build_fails.txt

Any other info / logs Include any logs or source code that would be helpful to diagnose the problem. If including tracebacks, please include the full traceback. Large logs and files should be attached.

closed time in 3 months

dbonner

issue commenttensorflow/tensorflow

Build failure: undefined reference to protobuf symbols

@alanpurple could you file a separate issue including all relevant information to reproduce the error? Please see https://github.com/tensorflow/tensorflow/blob/master/ISSUE_TEMPLATE.md#system-information

dbonner

comment created time in 3 months

issue commenttensorflow/tensorflow

Build failure: undefined reference to protobuf symbols

Hey @bas-aarts,

After double checking with our buildcop, it looks like the current Windows and Mac breakages' root causes are other commits (not https://github.com/tensorflow/tensorflow/commit/5caa9e83798cb510c9b49acee8a64efdb746207c). Therefore, I don't expect this change to be rolled back.

dbonner

comment created time in 3 months

issue commenttensorflow/tensorflow

Build failure: undefined reference to protobuf symbols

Hello!

https://github.com/tensorflow/tensorflow/commit/5caa9e83798cb510c9b49acee8a64efdb746207c is now in master. I've manually tested building with it using bazel 1.1, with the following command:

bazel build -c opt --config=cuda --config=v2 --host_force_python=PY3 //tensorflow/tools/pip_package:build_pip_package


Target //tensorflow/tools/pip_package:build_pip_package up-to-date:
  bazel-bin/tensorflow/tools/pip_package/build_pip_package
INFO: Elapsed time: 2067.380s, Critical Path: 828.19s
INFO: 12942 processes: 51 remote cache hit, 12891 local.
INFO: Build completed successfully, 14877 total actions

@dbonner can you confirm if Tensorflow head now builds for you as well?

dbonner

comment created time in 3 months

fork bmzhao/rpcs3

PS3 emulator/debugger

https://rpcs3.net/

fork in 3 months

startedcirosantilli/x86-assembly-cheat

started time in 3 months

startedcirosantilli/linux-kernel-module-cheat

started time in 3 months

push eventbmzhao/tensorflow

Jose Baiocchi

commit sha d27f5ca9c7aebe8800842c6e0295c76cbe946c24

Cleanup profiler deps PiperOrigin-RevId: 280745443 Change-Id: I3ca68dcc2ecf5fa9dd4680698c7b2ac77a25282e

view details

Katherine Wu

commit sha baac0b002c2c314bce22989b94de6017249130c2

Fix bug when loading subclassed model without training argument. PiperOrigin-RevId: 280747742 Change-Id: Ia6407047238351a23b956a91a5230057a89d36b8

view details

Gunhan Gulsoy

commit sha e70d095050e1d4858b2fcf0a5b55c31a2d631387

Bake TF RBE options into the tensorflow bazelrc. Any existing RBE users should be able to use these with minor modifications to build TF. PiperOrigin-RevId: 280748429 Change-Id: I866b91c2438159e434204a8ea29587aebc207149

view details

A. Unique TensorFlower

commit sha 9a249464083e1a603d3f89bfa235d42fb4c219ff

AndOp lowering from XLA to StandardOps This includes moving the patterns out of the xla_hlo namespace. This is due to mlir::hlo_xla::AndOp hiding the mlir::AndOp. PiperOrigin-RevId: 280754069 Change-Id: Ia4ab671b084ac1710875adaf8e6f4a46245313fc

view details

Geeta Chavan

commit sha fa0bfeb53ea07bf136b30e524d5d81bfb99be4f7

Updating/setting xcode version to 10.3 PiperOrigin-RevId: 280756701 Change-Id: I3b4c8ce78ef5a68a0a23f32ac25caf527597c700

view details

Yujing Zhang

commit sha c08edd7f929ae9ecb19c9116337597d94d249469

For a multi-host function, cache dtypes and shapes of remote variable inputs on the default function device, when lazy tensor copy is enabled. With this change, the dtypes and shapes would only be serialized and sent to the default function device once. PiperOrigin-RevId: 280758012 Change-Id: I1560a9d171f627b0d20aae51dd5f35a3b4f2c437

view details

Daniel Situnayake

commit sha ad9485bd46809daea8977d52b5c2c3d49b5792cd

Fix instructions for micro_speech training PiperOrigin-RevId: 280759722 Change-Id: I6e4e50dfa7569613f9891636b908ba8656613a0a

view details

A. Unique TensorFlower

commit sha 2f63b65d95de106ba359c2ac456e8dc8597bff9f

Add TraceMe to track which step_id is from tf data. PiperOrigin-RevId: 280760041 Change-Id: I76e41c26e21088c831d24a14201ac1adfb46b268

view details

Revan Sopher

commit sha b9e281c8ebad715f91f8db4f133695d246e13dbd

Switch std::make_unique for absl::make_unique. PiperOrigin-RevId: 280761166 Change-Id: Iecfc7496f08e4e05f75f8d598cac2806503c3edf

view details

Jared Duke

commit sha a1309f4e3c4780c8b24ba6b0fbddad278d3071e3

Switch to NDK API level 21 PiperOrigin-RevId: 280766624 Change-Id: I8500b69a5f6bebbeb0aafcf5744f5be5944738b9

view details

Robert David

commit sha ae857ed3c6a49177b045de17545255718bee2ede

Clean up ClipVector: use min/max instead of comparisons. Remove unused Clip function. PiperOrigin-RevId: 280767626 Change-Id: I6ac2041bdef69b73806a1a6d7a4f859a18214825

view details

Bixia Zheng

commit sha 5328d0aeac59b02c75506b9078fbad1da007678f

[TF:XLA:MLIR] Replace the tf2xla bridge used in tfcompile with MLIR bridge that lowers Tensorflow dialects representation to xla_hlo. Add argument mlir_components to tf_library to support the use of tfcompile with MLIR bridge. Add test tfcompile_test_mlir_bridge. PiperOrigin-RevId: 280767759 Change-Id: I79b07c7aeb7727b509e9eacbf31b803db3be7cb3

view details

A. Unique TensorFlower

commit sha 9847d334dc8d90fadb3efa2203a9998d848c8deb

Cleanup profiler deps PiperOrigin-RevId: 280771005 Change-Id: I4e5021169c50ce8f261ffeada7c7d653d45d37b3

view details

Igor Ganichev

commit sha 43793104fd9f1a6166a585479869b384278db0c8

Colocate input identities with inputs when inlining multi-device functions PiperOrigin-RevId: 280771766 Change-Id: Ic572bcb542c65c7fe0d01b95cdbdfb40d02bd830

view details

Youlong Cheng

commit sha eb45e7d5cbd2cd19b2c607c2a27f5beda991ced3

Add Dequantize support in XLA. (only Support QUINT8/UINT8 and MIN_COMBINED mode). PiperOrigin-RevId: 280772483 Change-Id: I98ba25d908f4df52223089248ae67c5559cf5377

view details

A. Unique TensorFlower

commit sha cee2a43b8184e92ba26ec0e3d6e00a3f8ca6e3c8

Internal change PiperOrigin-RevId: 280773783 Change-Id: Iad93f658a6ef04a60ebad78742d73ad6fb9220b6

view details

A. Unique TensorFlower

commit sha 936cdcdac96dfdb1657b882d2f880097cca05686

Fix the script path. PiperOrigin-RevId: 280798908 Change-Id: Iaab9322851fbb5aab3b2b4653f1cd59b7e456b30

view details

A. Unique TensorFlower

commit sha 78c73e0df4ea10f21d8f0df5895b620325e2ea6e

compat: Update forward compatibility horizon to 2019-11-16 PiperOrigin-RevId: 280810049 Change-Id: I0b1d2174f91f2a1418ce0fb656f1aa3b51b0431d

view details

Berkin Ilbeyi

commit sha ae44d3769265ea83823872b56cdce812e719f251

[XLA] Remove HloComputation::ContainsInstruction dependency from call_inliner. PiperOrigin-RevId: 280840055 Change-Id: Ib9d993ef2bc0c523a2190786314b20ce58056dd3

view details

A. Unique TensorFlower

commit sha 0d7511d8d7136adee452e3d747f3d049abec4f4b

Make the smoke test script executable. PiperOrigin-RevId: 280851572 Change-Id: Ieeca1e4a24a880a70387651a5e44ff89b37b4a47

view details

push time in 3 months

push eventbmzhao/tensorflow

yuan

commit sha 4e90ebf51974c30a28f871598c4dc9e34f72e970

test streaming accuracy python implementation

view details

Yuan Mingshuai

commit sha f15ca60067df336100270e2e8015a27402d46018

Merge pull request #1 from tensorflow/master update

view details

yuan

commit sha daa8ee0e567eb8acafef51fdd5c80c7971afb307

follow google code style

view details

yuan

commit sha f0b7c7660fdb725b0af9889081c07ba385417143

update bazel BUILD

view details

yuan

commit sha f59eb10a2de91a7e5c334c52e2f9430c08c6af6c

fix usage guide

view details

yuan

commit sha d4e50e0eda12d9857276e041bc120713e834bd66

fix code style

view details

Yuan Mingshuai

commit sha 6a0427dc6d884956c56322291f79f17c9811e3e3

Merge pull request #2 from tensorflow/master follow tf branch

view details

yuan

commit sha c469212443e6cecc76c28a3184da0b341a543467

try to fix Ubuntu Sanity Check error

view details

yuan

commit sha 410db05d80e901ab98e9f2dd477b6ebe1917487c

Merge branch 'master' of github.com:zhizunbao-y/tensorflow update local repository

view details

yuan

commit sha 0a1c1b96bdf3639a5aeed9593417d997e3b19026

try to fix pylint and ubuntu sanity check error

view details

Yuan Mingshuai

commit sha f30ffc2993121cc96306bc3cce661b81c5e9a868

Merge pull request #3 from tensorflow/master update

view details

yuan

commit sha 005057072380c8e14a17d76b5d8d3762bb48cb72

try to fix ubuntu sanity check error

view details

wenxizhu

commit sha 22e6ba72f94185fbecdee1300013f8e74a1bdaa1

Fuse "Transpose + Maxpool3d + Transpose".

view details

wenxizhu

commit sha 7659a2f82ef5512f9bdb84e6017fa99f635e09ba

Clang format fix.

view details

wenxizhu

commit sha 59fc4a41b447c6effe62f2eea69d12d4e1d3a6e1

Add test cases for transpose+maxpool3d+transpose.

view details

wenxizhu

commit sha 01d7d186e6c90562ae988404eca02277a2a2c99f

Change code foramt in CopyAttrsPooling().

view details

wenxizhu

commit sha a3a8322d29e13450805b82e1df2b31ccf8717c47

A negative case for "transpose + maxpool3d + transpose" added.

view details

wenxizhu

commit sha ed06859189722af4dc8e4abd655926df066e587a

Add format check.

view details

wenxizhu

commit sha 7f13d5c2238a61fa3f6be9f6a694692de82a7874

Clang format fix for mkl_layout_pass_test.cc

view details

Yong Tang

commit sha ab0a5278d81ef34096775d5d56f11694cca2a785

Fix tf.assert_equal issue when one tenor is empty and another is non-empty This fix tries to address the issue raised in 32082 where tf.assert_equal([], [1.0]) doesn't raise error. The reason was that in assert_equal `[1.0]` was broadcasted as `[]` and equal was in place in that situation. This PR updates the _binary_asesert so that it will check if x, y are both empty or both non-empty. If one is empty and another is non-empty, then assertion throws exception. This change is to not impact other ops that depends on the broadcast behavior. This fix fixes 32082. Signed-off-by: Yong Tang <yong.tang.github@outlook.com>

view details

push time in 3 months

more