profile
viewpoint
Yifei Feng yifeif @tensorflow @google San Francisco yifeifeng.com

tensorflow/tensorflow 141310

An Open Source Machine Learning Framework for Everyone

tensorflow/serving 4184

A flexible, high-performance serving system for machine learning models

tensorflow/benchmarks 762

A benchmark framework for Tensorflow

tensorflow/addons 692

Useful extra functionality for TensorFlow 2.x maintained by SIG-addons

tensorflow/custom-op 176

Guide for building custom op for TensorFlow

yifeif/pix2pix-tensorflow 2

Tensorflow port of Image-to-Image Translation with Conditional Adversarial Nets https://phillipi.github.io/pix2pix/

issue commenttensorflow/custom-op

How to build wheel for python 3?

Hi @VoVAllen, python3 should also be installed inside the provided Docker images. You might want to change the default python version inside the container. Let me know if that's not the case :)

VoVAllen

comment created time in 3 days

issue commenttensorflow/custom-op

FileNotFoundError: [Errno 2] No such file or directory: '/dt7/usr/bin/gcc'

@shengyushen was this running inside the provided Docker container?

shengyushen

comment created time in 3 days

issue commenttensorflow/tensorflow

Custom C++ operator compilation error with Tensorflow 2.1

+1 on what @07rafix said, we did switched to VS2019 when building TF 2.1. Would you mind give that a try @Lillypucien?

Lillypucien

comment created time in 7 days

pull request commenttensorflow/tensorflow

Make vscode's pylint integration work on TensorFlow source

Thanks @frreiss! This PR will need a manual import. Let me do that.

frreiss

comment created time in 12 days

issue commenttensorflow/tensorflow

variable_ops.h missing from pip install

@gunan, do you think headers for classes like VariableOp under core/kernels/ should be included?

dirktheeng

comment created time in a month

push eventtensorflow/custom-op

Yifei Feng

commit sha bf62590712a39dc861169125943c9133fbd33b38

Update example to be compatible to 2.x

view details

push time in a month

pull request commenttensorflow/tensorflow

Add docs for tf.strings.lower and tf.strings.regex_replace

Sorry for the delay, do you mind resolving the conflicts @lc0 ?

lc0

comment created time in a month

issue commenttensorflow/addons

Build Addons for Windows

Thank you @seanpmorgan!

seanpmorgan

comment created time in a month

issue commenttensorflow/custom-op

Building a custom-op on Windows

Windows gpu support has also been added. Let me know if you run into any issue with it.

seanpmorgan

comment created time in 2 months

push eventtensorflow/custom-op

Yifei Feng

commit sha 994dc6bdd5b7c0c0c0ffb55bb0ac013d9d9268cd

Add windows gpu support (#43)

view details

push time in 2 months

PR merged tensorflow/custom-op

Add windows gpu support

Fix https://github.com/tensorflow/custom-op/issues/24

+87 -65

0 comment

5 changed files

yifeif

pr closed time in 2 months

issue closedtensorflow/custom-op

Building a custom-op on Windows

A heavily requested feature for tensorflow/addons is that we package a windows version: https://github.com/tensorflow/addons/issues/173

AFAIK to build this we'll need a libtensorflow_framework.dll of some sort within the pip installed package. Looking at the link flags & installed directory it looks like this is not available.

python -c "import tensorflow as tf; print(tf.version.VERSION)"
2.0.0-beta1

python -c "import tensorflow as tf; print(tf.sysconfig.get_link_flags())"
[]

Is this by design or has it just not been implemented yet? There is a _pywrap_tensorflow_internal.pyd but I don't believe we've been linking to the python shared objects so I figured I was headed down the wrong path.

cc @meteorcloudy @yifeif @perfinion

closed time in 2 months

seanpmorgan

PR opened tensorflow/custom-op

Add windows gpu support
+87 -65

0 comment

5 changed files

pr created time in 2 months

push eventyifeif/custom-op

Yifei Feng

commit sha aad6ac6dc43303f5503de250da3222a3d22ed36f

Complete windows GPU support

view details

push time in 2 months

push eventyifeif/custom-op

Yifei Feng

commit sha 62e530eca5d88f6be0cf12d191e8bd834934b5f4

Add support for cuda 10.1 (TF 2.1+). (#42)

view details

Yifei Feng

commit sha 4799039e5451526b659b5708ce900b0cac9679f8

Merge branch 'master' of https://github.com/tensorflow/custom-op into win_gpu2

view details

push time in 2 months

push eventyifeif/custom-op

Yifei Feng

commit sha 121b31a71bcd18577e54288b9aa45838adf3dc33

More fixes for windows gpu build

view details

push time in 2 months

create barnchyifeif/custom-op

branch : win_gpu2

created branch time in 2 months

push eventtensorflow/custom-op

Yifei Feng

commit sha 62e530eca5d88f6be0cf12d191e8bd834934b5f4

Add support for cuda 10.1 (TF 2.1+). (#42)

view details

push time in 2 months

PR merged tensorflow/custom-op

Add support for cuda 10.1 (TF 2.1+).
+4539 -562

4 comments

10 changed files

yifeif

pr closed time in 2 months

pull request commenttensorflow/custom-op

Add support for cuda 10.1 (TF 2.1+).

cc @lespeholt

yifeif

comment created time in 2 months

pull request commenttensorflow/custom-op

Add support for cuda 10.1 (TF 2.1+).

There will be a new nightly docker image for cuda 10.1 and its relevant toolchain. Until the release is setup, here is a temp image at tensorflow/tensorflow:nightly-custom-op-gpu-ubuntu16-cuda10.1

yifeif

comment created time in 2 months

push eventyifeif/custom-op

Yifei Feng

commit sha 1dd6ee5d1f8718ca6e877c5c9ec379bc6fb33fb8

Add support for cuda 10.1

view details

Yifei Feng

commit sha 7e1536edda70a445f219a8be5df0bd562b2f5bfc

Update how cuda lib are found to support 10.1

view details

Yifei Feng

commit sha 0c792a729f43bff1475eb6ca5d3cdf9e1c6740b2

fix configure.sh

view details

push time in 2 months

create barnchyifeif/custom-op

branch : test2

created branch time in 2 months

PR opened tensorflow/custom-op

Add support for cuda 10.1 (TF 2.1+).
+4539 -562

0 comment

10 changed files

pr created time in 2 months

push eventyifeif/custom-op

Yifei Feng

commit sha 3f41c41d9217fcd0e1850ada3525e3d7d62be700

fix configure.sh

view details

push time in 2 months

push eventyifeif/custom-op

Yifei Feng

commit sha 1fd7ae3c1bb4c0cc52eb96cda52e686a3ab1fcea

Update how cuda lib are found to support 10.1

view details

push time in 2 months

create barnchyifeif/custom-op

branch : test

created branch time in 2 months

issue commenttensorflow/tensorflow

[TF 2.1.0rc1] Fail to import package on Windows

Is this the same as https://github.com/tensorflow/tensorflow/issues/35036

seanpmorgan

comment created time in 2 months

startedrockita/Environmental_Intelligence

started time in 2 months

push eventyifeif/custom-op

yifeif

commit sha c298e8c2f46c2aa3263fa1ad266e61ff2a1899d5

Add find_cuda_config.py

view details

push time in 3 months

create barnchyifeif/custom-op

branch : win_gpu

created branch time in 3 months

issue commenttensorflow/custom-op

Building a custom-op on Windows

FYI, currently we are still using pywrap_tensorflow instead of tensorflow_framework under the hood for windows, but this detail should be taken care of by the macros.

seanpmorgan

comment created time in 3 months

issue commenttensorflow/custom-op

Building a custom-op on Windows

Thanks for the patience! This should now work for CPU, and GPU support is on its way.

seanpmorgan

comment created time in 3 months

push eventyifeif/custom-op

yifeif

commit sha d0735bb39e610fe374ef2f02b866336c234258d8

Remove debug line

view details

push time in 3 months

push eventtensorflow/custom-op

Yifei Feng

commit sha 35e2e4206ee77e1842ea77e6013b56454419422c

Add windows cpu support (#41)

view details

push time in 3 months

PR merged tensorflow/custom-op

Add windows cpu support

Part of https://github.com/tensorflow/custom-op/issues/24

+102 -41

0 comment

4 changed files

yifeif

pr closed time in 3 months

push eventyifeif/custom-op

yifeif

commit sha 28f90200bcbcf585fe40cdf671cb671cdfec0e2a

clean up windows support

view details

push time in 3 months

PR opened tensorflow/custom-op

Add windows cpu support
+102 -41

0 comment

4 changed files

pr created time in 3 months

push eventyifeif/custom-op

yifeif

commit sha d1c485684efa35b6de2bf4ee3d9c0905247001ae

Update readme for windows.

view details

push time in 3 months

push eventyifeif/custom-op

yifeif

commit sha 4004bf683d0dd5066e0214b22d9bb705814c039e

Add windows cpu support

view details

push time in 3 months

push eventyifeif/custom-op

Yifei Feng

commit sha 8cfb009383a6ae66c2e8dc00aa756289340e0d45

Add windows support

view details

push time in 3 months

fork yifeif/custom-op

Guide for building custom op for TensorFlow

fork in 3 months

create barnchyifeif/tensorflow

branch : win_def4

created branch time in 3 months

push eventtensorflow/tensorflow

Yifei Feng

commit sha fff3b08dd10ad510408481cce53bc2e5005758b0

Move gpu script update to the correct file. PiperOrigin-RevId: 282646719 Change-Id: Idbec43ef97473209cf301ef66ec3f323952e42ad

view details

Yifei Feng

commit sha c6daad319d78e08d2dacbc4aa58094ce541188cf

Merge pull request #34626 from yifeif/cherrypicks_J8PS0 Move gpu script update to the correct file.

view details

push time in 3 months

PR merged tensorflow/tensorflow

Move gpu script update to the correct file. cla: yes size:S

PiperOrigin-RevId: 282646719 Change-Id: Idbec43ef97473209cf301ef66ec3f323952e42ad

+5 -5

0 comment

2 changed files

yifeif

pr closed time in 3 months

PR opened tensorflow/tensorflow

Move gpu script update to the correct file.

PiperOrigin-RevId: 282646719 Change-Id: Idbec43ef97473209cf301ef66ec3f323952e42ad

+5 -5

0 comment

2 changed files

pr created time in 3 months

create barnchyifeif/tensorflow

branch : cherrypicks_J8PS0

created branch time in 3 months

push eventtensorflow/tensorflow

Yifei Feng

commit sha ae30c1ba8712ddd8dd1bcea03766ef6e7000c5b8

Fix bug in if -v check. PiperOrigin-RevId: 282598769 Change-Id: Ic71c23ba3983c397cac81a44cd4a71f76e5ffef3

view details

Yifei Feng

commit sha 5c43a0ab08d60d8f3b591e9e83ad8a7b0bac5558

Merge pull request #34623 from yifeif/cherrypicks_ZT8OO Fix bug in if -v check.

view details

push time in 3 months

PR merged tensorflow/tensorflow

Fix bug in if -v check. cla: yes size:XS

PiperOrigin-RevId: 282598769 Change-Id: Ic71c23ba3983c397cac81a44cd4a71f76e5ffef3

+2 -2

0 comment

2 changed files

yifeif

pr closed time in 3 months

PR opened tensorflow/tensorflow

Fix bug in if -v check.

PiperOrigin-RevId: 282598769 Change-Id: Ic71c23ba3983c397cac81a44cd4a71f76e5ffef3

+2 -2

0 comment

2 changed files

pr created time in 3 months

create barnchyifeif/tensorflow

branch : cherrypicks_ZT8OO

created branch time in 3 months

push eventtensorflow/tensorflow

Yifei Feng

commit sha 058f14880166d1b6ee699b857751e7ffc6a6c599

Update windows release build script to create the correct pip package names. PiperOrigin-RevId: 282425067 Change-Id: I13b05c37eb9dbeac2d9beea31b21898f12cfdc89

view details

Yifei Feng

commit sha 1d03f446b04834a427e76cc53f096635597a0755

Merge pull request #34596 from yifeif/cherrypicks_UUJ02 Update windows release build script to create the correct pip package…

view details

push time in 3 months

PR merged tensorflow/tensorflow

Reviewers
Update windows release build script to create the correct pip package… cla: yes size:M

… names.

PiperOrigin-RevId: 282425067 Change-Id: I13b05c37eb9dbeac2d9beea31b21898f12cfdc89

+34 -17

0 comment

14 changed files

yifeif

pr closed time in 3 months

PR opened tensorflow/tensorflow

Update windows release build script to create the correct pip package…

… names.

PiperOrigin-RevId: 282425067 Change-Id: I13b05c37eb9dbeac2d9beea31b21898f12cfdc89

+34 -17

0 comment

14 changed files

pr created time in 3 months

create barnchyifeif/tensorflow

branch : cherrypicks_UUJ02

created branch time in 3 months

create barnchyifeif/tensorflow

branch : win_def3

created branch time in 3 months

push eventtensorflow/tensorflow

Yifei Feng

commit sha 97fbcf8c44bb96d3eb516460416018f6bf71ee41

Do not use common script to install bazel the paths in docker image are different. PiperOrigin-RevId: 281837917 Change-Id: Id62738e214ef66ffd17226bc0c90ae9912fd9bf9

view details

Yifei Feng

commit sha 73501b7ceb072b104f99d9ab3115572955454a4d

add missing "echo". PiperOrigin-RevId: 281838161 Change-Id: I396accac7be4e74f8afa00abb554d5e504e0b4f2

view details

Yifei Feng

commit sha d2103ebba7dd851c51a883b599c7718be8f8b341

Merge pull request #34505 from yifeif/cherrypicks_0INWN Cherrypicks for r2.1

view details

push time in 3 months

PR merged tensorflow/tensorflow

Reviewers
Cherrypicks for r2.1 cla: yes size:S
+12 -4

0 comment

2 changed files

yifeif

pr closed time in 3 months

PR opened tensorflow/tensorflow

Cherrypicks for r2.1
+12 -4

0 comment

2 changed files

pr created time in 3 months

create barnchyifeif/tensorflow

branch : cherrypicks_0INWN

created branch time in 3 months

push eventtensorflow/tensorflow

Yifei Feng

commit sha 3be306cfd9ab87d2d49edd3d6d02e9d6e82312e2

Make docker has the same bazel version as the env that invokes it. PiperOrigin-RevId: 281789221 Change-Id: I6b2ebbe4bf787bb2e591905c8e5368cfac793e0e

view details

Yifei Feng

commit sha 4f734a3e3f6de2bc166306bd86eda66706b8f08a

Merge pull request #34497 from yifeif/cherrypicks_D33SX Make docker has the same bazel version as the env that invokes it.

view details

push time in 3 months

PR merged tensorflow/tensorflow

Reviewers
Make docker has the same bazel version as the env that invokes it. cla: yes size:XS

PiperOrigin-RevId: 281789221 Change-Id: I6b2ebbe4bf787bb2e591905c8e5368cfac793e0e

+5 -1

0 comment

2 changed files

yifeif

pr closed time in 3 months

PR opened tensorflow/tensorflow

Make docker has the same bazel version as the env that invokes it.

PiperOrigin-RevId: 281789221 Change-Id: I6b2ebbe4bf787bb2e591905c8e5368cfac793e0e

+5 -1

0 comment

2 changed files

pr created time in 3 months

create barnchyifeif/tensorflow

branch : cherrypicks_D33SX

created branch time in 3 months

PR opened tensorflow/tensorflow

Allow building two pip cpu packages on macos.

Properly build both packages instead renaming.

PiperOrigin-RevId: 281635032 Change-Id: I3577d1f4f213e1d1daf4caa757506be48d63972c

+44 -14

0 comment

5 changed files

pr created time in 3 months

create barnchyifeif/tensorflow

branch : cherrypicks_09NWZ

created branch time in 3 months

create barnchyifeif/tensorflow

branch : win

created branch time in 3 months

issue closedtensorflow/tensorflow

libtensorflow_framework.so No such file or directory

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

System information

  • Have I written custom code (as opposed to using a stock example script provided in TensorFlow):
  • OS Platform and Distribution (e.g., Linux Ubuntu 16.04): Linux Ubuntu 18.04
  • Mobile device (e.g. iPhone 8, Pixel 2, Samsung Galaxy) if the issue happens on mobile device:
  • TensorFlow installed from (source or binary):
  • TensorFlow version (use command below): 1.14.0
  • Python version: Py36
  • Bazel version (if compiling from source): bazel release 0.26.1
  • GCC/Compiler version (if compiling from source): 7.3.0
  • CUDA/cuDNN version:
  • GPU model and memory:

I tried to include the op in the code as follows

    idtable_op_module = tf.load_op_library(
    resource_loader.get_path_to_datafile('libid_table_op.so'))

only to get the traceback infomation:

Traceback (most recent call last):
  File "wide_deep.py", line 12, in <module>
    import jarvis.tensorflow as jtf
  File "/opt/ml/job/python/jarvis/tensorflow/__init__.py", line 4, in <module>
    from jarvis.tensorflow.feature_transform import *
  File "/opt/ml/job/python/jarvis/tensorflow/feature_transform.py", line 10, in <module>
    from jarvis.tensorflow import id_column
  File "/opt/ml/job/python/jarvis/tensorflow/id_column.py", line 44, in <module>
    from jarvis.tensorflow.id_table_ops import IdHashTable
  File "/opt/ml/job/python/jarvis/tensorflow/id_table_ops.py", line 29, in <module>
    resource_loader.get_path_to_datafile('libid_table_op.so'))
  File "/usr/local/lib/python3.6/site-packages/tensorflow_core/python/framework/load_library.py", line 61, in load_op_library
    lib_handle = py_tf.TF_LoadLibrary(library_filename)
tensorflow.python.framework.errors_impl.NotFoundError: libtensorflow_framework.so: cannot open shared object file: No such file or directory

The trackback shows that there's no libtensorflow_framework.so, I tried to find this file in tf 1.14.0 binary version(not compiled version) and only found site-packages/tensorflow_core/libtensorflow_framework.so.1, but still no libtensorflow_framework.so. So what's the correct way to include so file, is it a bug in tf 1.14?

closed time in 4 months

sjtusmartboy

issue commenttensorflow/tensorflow

libtensorflow_framework.so No such file or directory

As @Spotlight0xff mentioned, tf.sysconfig.get_link_flags() should contains the correct name for libtensorflow_framework. And you should build the custom op library against the TF pip version you want to run with. Closing this issue now. Let us know if you have more questions.

sjtusmartboy

comment created time in 4 months

issue commenttensorflow/tensorflow

tensorflow-1.15.0 manylinux wheel on PyPi is invalid

Thanks for reporting @fhoering! Yes we should and will fix this.

fhoering

comment created time in 4 months

issue commenttensorflow/custom-op

Building a custom-op on Windows

Thanks for the feedback @amitport. Just to clarify, we do care about Windows support, and plan to add it soon. We are marking this as contribution welcome in case anybody gets to pieces of this sooner.

We are going to update the checklist as we make progress, most likely after TF World next week. cc: @martinwicke @jakline.

seanpmorgan

comment created time in 4 months

issue commenttensorflow/custom-op

Building a custom-op on Windows

Much thanks to @meteorcloudy 's investigation, here is a tentative todo-list paraphrased below:

  • [ ] Build tensorflow_framework with symbols exported (example commit by @meteorcloudy )
  • [ ] optionally make TensorFlow windows pip package build non-monolithic
  • [ ] Include .dll and .lib files of tensorflow_framework in the pip package
  • [ ] Make configure.sh and tf_configure.bzl be aware of Windows. To use the tensorflow_framework.dll.if.lib for linking, you can use the cc_import rule.
  • [ ] Make build_pip_pkg.sh script work on Windows. Might need to do something similar to TF's build_pip_package.sh script as unfiles is disabled on Windows by default.
  • [ ] Figure out how does \<custom_op\>.so find tensorflow_framework.dll at runtime. Do we need .pyd suffix (we might not if we use tf.load_op_library).

Let us know if you are interested in helping out with any of this.

seanpmorgan

comment created time in 4 months

push eventtensorflow/custom-op

Yifei Feng

commit sha 1e8b52b19bb6ff8ce3b635302e032349b0280379

Fix zero_out test for 2.0

view details

Yifei Feng

commit sha a2685fcdc066d21b3f3f5cd2a190f42f7e308bff

Add time_two to pip package.

view details

Yifei Feng

commit sha 8f7842901b137d1dab2b7f53e1433d5dfbbcaf70

Merge pull request #38 from yifeif/master Add time_two op to pip package

view details

push time in 4 months

PR merged tensorflow/custom-op

Add time_two op to pip package
+9 -6

0 comment

6 changed files

yifeif

pr closed time in 4 months

PR opened tensorflow/custom-op

Add time_two op to pip package
+9 -6

0 comment

6 changed files

pr created time in 4 months

push eventyifeif/custom-op

Yifei Feng

commit sha a2685fcdc066d21b3f3f5cd2a190f42f7e308bff

Add time_two to pip package.

view details

push time in 4 months

push eventtensorflow/custom-op

Yifei Feng

commit sha e86c85f6c79e85238ee65b63c532f562f22764e9

Fix zero_out test for 2.0 (#37)

view details

push time in 4 months

PR merged tensorflow/custom-op

Fix zero_out test for 2.0
+1 -1

0 comment

1 changed file

yifeif

pr closed time in 4 months

PR opened tensorflow/custom-op

Fix zero_out test for 2.0
+1 -1

0 comment

1 changed file

pr created time in 4 months

push eventyifeif/custom-op

Yifei Feng

commit sha 1e8b52b19bb6ff8ce3b635302e032349b0280379

Fix zero_out test for 2.0

view details

push time in 4 months

Pull request review commenttensorflow/estimator

Fix broken release

 # This version string is semver compatible, but incompatible with pip. # For pip, we will remove all '-' characters from this string, and use the # result for pip.-_VERSION = '2.0.0'+_VERSION = '2.0.1' 

2.1.0?

mihaimaruseac

comment created time in 4 months

startedtensorflow/text

started time in 4 months

PR opened broken/custom-op

Temporary fix for absl error
+6 -6

0 comment

3 changed files

pr created time in 4 months

delete branch tensorflow/custom-op

delete branch : broken

delete time in 4 months

create barnchyifeif/custom-op

branch : broken

created branch time in 4 months

create barnchtensorflow/custom-op

branch : broken

created branch time in 4 months

push eventtensorflow/tensorflow

A. Unique TensorFlower

commit sha 072e8c273ba5f860a6372ffb4a2660cbe2caf222

Update ops-related pbtxt files. PiperOrigin-RevId: 274712322

view details

push time in 4 months

push eventtensorflow/tensorflow

Ashwin Murthy

commit sha f74e4baf974cbab7b3e2b2912bcd89dab4019108

[TFLite] LSTM utils need to check bounds of input tensor's shape prior to slicing. PiperOrigin-RevId: 274710827

view details

push time in 4 months

push eventtensorflow/tensorflow

Christopher Suter

commit sha 898f605e50d75da89d8b273c78a05f5be3a257e7

Expose the current setting togglable by `experimental_run_functions_eagerly`. PiperOrigin-RevId: 274708938

view details

push time in 4 months

PR merged tensorflow/tensorflow

Support the 3D convolution ops for auto_mixed_precision cla: yes comp:grappler ready to pull size:M stat:awaiting response

Previous TF auto_mixed_precision doesn't support 3D convolutions.

Since the cuDNN enhanced the performance of 3D convolutions since v7.6.2 (https://docs.nvidia.com/deeplearning/sdk/cudnn-release-notes/rel_762.html#rel_762), this PR added the support and related tests.

fyi @nluehr @benbarsdell

+101 -11

4 comments

3 changed files

houtoms

pr closed time in 4 months

push eventtensorflow/tensorflow

Kaixi Hou

commit sha d31df357db6bd567d3379493ea81fe641ef6b2b7

Support the 3d convolution ops for auto_mixed_precision

view details

Kaixi Hou

commit sha 6323e8f40b1a895155702db75954031c663e37f1

Skip conv3d tests of auto-mixed precision for now

view details

Kaixi Hou

commit sha f314fe80e755f709011d3d6ae66f415e61abd6e3

import unittest

view details

TensorFlower Gardener

commit sha c67053aff0ded1d72ebb8c8b1242bc0a85def583

Merge pull request #32850 from houtoms:support_conv3d_in_amp PiperOrigin-RevId: 274706781

view details

push time in 4 months

push eventtensorflow/tensorflow

Brian Zhao

commit sha 0986f8df1809b97d900b6956fe3ca3fb340172f0

Wiring tf/core/platform:protobuf and tf/core/platform:protobuf_compiler into tf/core/BUILD. PiperOrigin-RevId: 274705436

view details

push time in 4 months

push eventtensorflow/tensorflow

A. Unique TensorFlower

commit sha a6cb56bb2943128e31b1389f91edbbdd4ad33013

Add another concatentation test when the input is incorrect. PiperOrigin-RevId: 274701850

view details

push time in 4 months

push eventtensorflow/tensorflow

Sergei Lebedev

commit sha 2f245bdc5bd4e70c75f2375cc0671dd577f6a9ae

Fixed cases where tf.TensorShape was constructed with float dimensions This is a prerequisite for making TensorShape and Dimension more strict about the types of their arguments. PiperOrigin-RevId: 274700832

view details

push time in 4 months

push eventtensorflow/tensorflow

Yunxing Dai

commit sha e2eb7e36417c4b4582161c5a12367a83da5ae82b

[XLA] HandleClamp in DynamicPadder. PiperOrigin-RevId: 274700208

view details

push time in 4 months

push eventtensorflow/tensorflow

Chao Mei

commit sha 68fedf4c80e9cfeffbbcbb987fae8d896d3a0791

Change BenchmarkTfLiteModel::InputTensorData to use proper type-erased pointers for data storage, avoiding undefined behaviors around aliasing rules, alignment requirements, and C++ object lifetimes. Also use unique_ptr for memory management. PiperOrigin-RevId: 274698661

view details

push time in 4 months

PR merged tensorflow/tensorflow

Initial implementation of general eigenvalue decomposition cla: yes ready to pull size:L
+645 -24

20 comments

20 changed files

Randl

pr closed time in 4 months

push eventtensorflow/tensorflow

Evgenii Zheltonozhskii

commit sha e10ec20a746d77d05ca79cc6dce60890bd7bc903

Initial implementation of general eigenvalue decomposition

view details

Evgenii Zheltonozhskii

commit sha 7f7125e402d803ab3b291ff75bd3eab32ed0d731

Remove non-existing test file from build

view details

Evgenii Zheltonozhskii

commit sha 6c0e3350d7bb0a09b412d45fa9ffd71d9969cc61

Decompose ComplexEigOp for input and output type

view details

Evgenii Zheltonozhskii

commit sha f43dea7ca9aae6aa6b2fe5dffc6fd02c64c1afd8

Decompose LinearAlgebraOp for different input and output types, apply to eig

view details

Evgenii Zheltonozhskii

commit sha 3602f66fff596cccc7c48beda9ced2dd7b33fa30

Fix typedef

view details

Evgenii Zheltonozhskii

commit sha 0f0086a8e55a799fef4ef777260672edb86f89ba

Fix op

view details

Evgenii Zheltonozhskii

commit sha e6a9247bbeaa52113bec454dc0e55285c416f78b

pylint

view details

Evgenii Zheltonozhskii

commit sha 9ba9fe64cfa96b2a43d8b890ab340dfdc352a5d0

Fix eig tests

view details

Evgenii Zheltonozhskii

commit sha 62558b1b6d2eee6991c84d58c35195f67771e643

Golden update

view details

Evgenii Zheltonozhskii

commit sha a973c584a67ef3ede86eb5ee79772d1f7c6ef47a

Update ops

view details

Evgenii Zheltonozhskii

commit sha cf27a56bbf2f95233df2655734015b50c1e56ecd

API definition update

view details

TensorFlower Gardener

commit sha 8ae730be26b3e99cc9f95080552c7a926a8678ae

Merge pull request #28340 from Randl:eig PiperOrigin-RevId: 274695901

view details

push time in 4 months

push eventtensorflow/tensorflow

Chao Mei

commit sha 1079ed61d844adbb64db82a1c819f53e3901a87f

1. Simply used ',' instead of '#' as the separator between input layer name and its value range. This is to avoid cases where '#' exists in the input layer name. For example: if the model has two input tensors named as "input1", "input2" and the "input1" requires the value to be initialized between 0 and 5 (inclusive), we could specify "--input_layer=input1,input2 --input_layer_shape=1,128:1,64 --input_layer_value_range=input2,0,5:input1,-12,15" to achieve this. 2. Added a test to check the parsing such input-related parameters. PiperOrigin-RevId: 274694297

view details

push time in 4 months

push eventtensorflow/tensorflow

A. Unique TensorFlower

commit sha 025e871a4aac109e6e168ff7f2690a9093364199

[Grappler] Extend ConstantPushDown optimization to also handle BiasAdd when it is safe to do so. Fix a bug in existing constant push down: We need to clear input properties when swapping inputs. PiperOrigin-RevId: 274690178

view details

push time in 4 months

more