profile
viewpoint
Yan Facai (颜发才) facaiy Beijing, China. linkedin.com/in/facaiy Software Engineer.

facaiy/math-expression-parser 8

A scala library for parsing mathemitical expressions with support for parentheses and variables.

facaiy/book_notes 4

整理我读过的书籍笔记

facaiy/DAG-lite 4

An experimental DAG library with functional programming technology.

facaiy/facaiy.github.io 4

个人博客

facaiy/facaiy-scala-with-java-quickstart 1

An archetype which creates a mixed Java/Scala project.

facaiy/scikit-learn 1

scikit-learn: machine learning in Python

dynamicwebpaige/community-1 0

Stores documents used by the TensorFlow developer community

facaiy/addons 0

Useful extra functionality for TensorFlow maintained by SIG-addons

facaiy/angel 0

A Flexible and Powerful Parameter Server for large-scale machine learning

facaiy/business-card 0

A business card in LaTeX.

issue commenttensorflow/addons

Problems with CohenKappa metrics

@AakashKumarNain Sorry for the delay, Aakash. Thanks for your detailed RFC, which looks really great. As said before, is it possible to create a metric for every case mentioned by you, for example: CohenKappa, BinaryCohenKappa, CategoricalCohenKappa etc (refer to Accuracy, BinaryAccuracy, CategoricalAccuracy, ... ). What do you think? cc @WindQAQ @seanpmorgan

rere-rere

comment created time in 20 days

push eventfacaiy/addons

Luke Miner

commit sha 3634b55d803ed2eea2c92f68d683dc398c066e88

tqdm status bar custom metrics formatting (#916) This is a simple change that will allow custom formatting of metrics for the tqdm status bar. This is especially useful if you have small loss values and want to see it in scientific notation. Replace `metrics_format` with `'{name}: {value:e}'`

view details

Shun Lin

commit sha 999aebc0961ccddb8174cc5331cc23a7291a2255

adding shun's github handle to callback maintainers list (#918) * adding shun's github handle to callback maintainers list * update CODEOWNERS

view details

Sean Morgan

commit sha faaad909ab64c7f6700fb41d4664d528d4a62fc5

* Remove py3 from test status (#920) * Add tfa-nightly to linux build matrix

view details

Gabriel de Marmiesse

commit sha ebd9f415fa6759ef8ee2dc752773261cb3beb593

Added types and typeguard. (#894) * Added types and typeguard.

view details

Sean Morgan

commit sha 0d4ab2971e1f3481da8795286a2391518606f101

CLN: Move build pip pkg script from root directory (#921)

view details

Gabriel de Marmiesse

commit sha ee8eab02eb6ef2201691c78e8687961360e666b5

Removed all usages of six (#923)

view details

Gabriel de Marmiesse

commit sha fa8e966d987fd9b0d20551a666e44e2790fdf9c7

Target tensorflow stable. (#912)

view details

Gabriel de Marmiesse

commit sha 01bef166ba2c88f0ff86ee7604855f781e50d56e

Used flake8 + no formatter instead of pep8 + pylint + yapf. (#904) * Used flake8 instead of pep8 + pylintrc. Removed formatting. * Added a separate build for flake8

view details

nuka

commit sha 451b7fa1fe4ac5c45ab7ac1cbfc6e01d5d5ccf32

Fix: Remove descriptions about global_step in tfa.optimizers.{SGDW, AdamW} (#930)

view details

Sean Morgan

commit sha 3a9397c732e23c9b7513f40ed2881c6b05b2f7a1

BLD: Move builds and releases to GitHub actions (#933)

view details

Samuel Marks

commit sha 83083ed8fe1e8c92dc24512cc76daf61b2917ad2

Fix some typos (#937)

view details

Gabriel de Marmiesse

commit sha b1bb48552ae433d3275927ed7212e854aabecd40

Ensuring flake8 is compatible with black. (#938)

view details

Gabriel de Marmiesse

commit sha 3a2c348300dfba0c8afc3156e564b555aaad23e2

Enabling black progressively. (#932) * Added github action for black.

view details

Gabriel de Marmiesse

commit sha ca71fc861f864e6135a14bc7ec2d7fabfc5907f9

Formatted with black all the files not being worked on at the moment. (#942) * Keep all files being worked on in pull requests. * Two files failed to be parsed. * Formatted with black all the files not being worked on at the moment. * Added e231 rule.

view details

Sean Morgan

commit sha 0d3284d70a6e6f46614c631fab4a7e37f1072066

DOC: Add style badge (#944) DOC: Update nightly build matrix

view details

Sean Morgan

commit sha b7a66a7f3ac20f087198c2ca839726b01d1d5e5d

CLN: Remove pylint and yapf (#946)

view details

Sean Morgan

commit sha 5260d339ea53d57631f739f41f093656757c0077

BLD: Fix github actions build (#947)

view details

Gabriel de Marmiesse

commit sha 266c91bc54567e23302081d49722e299ec1b640d

Added a small pre-commit hook for easier running of tools. (#943) * Added a small pre-commit hook for easier running of tools. * Moved the pre-commit in tools.

view details

Gabriel de Marmiesse

commit sha 987c1a65988ca9ff7c8266ca37bda220bbe8606f

Removing dead code. (#950) * Removing dead code. * Removed CI requirements.

view details

Gabriel de Marmiesse

commit sha f2fb8072a512e344117460a06e6091472dd11ff9

Typing some layers. (#929) * Typing some layers.

view details

push time in 20 days

pull request commenttensorflow/addons

[WIP] Used flake8 + no formatter instead of pep8 + pylint + yapf.

Hi, Gabriel, thanks for the PR. I'm wondering why we should remove python formatter here?

gabrieldemarmiesse

comment created time in a month

Pull request review commenttensorflow/addons

Added types and typeguard.

 # =============================================================================  import logging+from typeguard import typechecked

or move it into types.py

gabrieldemarmiesse

comment created time in a month

Pull request review commenttensorflow/addons

Added types and typeguard.

+from . import types

or set the __all__ variable if necessary

gabrieldemarmiesse

comment created time in a month

Pull request review commenttensorflow/addons

Added types and typeguard.

+from . import types

I mean we can keep the file empty

gabrieldemarmiesse

comment created time in a month

Pull request review commenttensorflow/addons

Added types and typeguard.

 # =============================================================================  import logging+from typeguard import typechecked

please put it behind import tensorflow as tf

gabrieldemarmiesse

comment created time in a month

Pull request review commenttensorflow/addons

Added types and typeguard.

+from . import types

nit: is it necessary?

gabrieldemarmiesse

comment created time in a month

push eventfacaiy/addons

Sean Morgan

commit sha bff740930b7865a0772e53b47520bdbc4d36a9d6

Add github action to test with macos (#799) * Add github action to test with macos * Add nightly test

view details

Sean Morgan

commit sha d90b7fdb1d60d8c688f5082aed5ca181d8df3204

Delete Broken_Template.ipynb

view details

Dheeraj R Reddy

commit sha 510aa3c6d71a2d27e3ac8ecdb09e0f16028039d0

FIX: RNN support for weight norm and `remove()` (#819) * FIX: Consider recurrent_kernel for RNNs * TST: Increase test coverge for weight norm * Add method to remove WeightNorm and return base layer with fused kernel

view details

Sean Morgan

commit sha c512df97e99acbe945d2cd09695f072552112c6f

BLD: fix nightlies (#824) * Move requirements.txt into packaging directory * Remove sudo call in release script * Fix required includes for windows build

view details

Sean Morgan

commit sha cb51a0bc595d92b4b6a60c5f80ccab6f63a31b89

DOC: Add badge for nightly macos tests (#827) * Add badge for nightly macos tests * Update link

view details

Sean Morgan

commit sha b026fd62d6c892100f955e72247f984f14baa308

Install twine to user (#826)

view details

Sean Morgan

commit sha ac91cf2c39c1f2c5b04d828f27500a9e7f1bce81

Update logging to warning since warn is deprecated (#828)

view details

Shun Lin

commit sha 3dd560ceab7ef9a934eb29c414f9df49224edabe

fixes bugs when using fit generator (#829)

view details

Sean Morgan

commit sha 0541bc35f94470c91e38fcf01142a9a15c155c4d

Mark wrappers test as flaky (#831)

view details

Sean Morgan

commit sha 7e06c2fca0f94b12312ae3bf7a85f1b6ff6212ea

Only install twine to user for linux builds (#833)

view details

who who who

commit sha 8ae8116f42cd740bf35c3af56dab8a34d7e8329c

Rrelu fix (#820) * refactor rrelu

view details

Sean Morgan

commit sha d541fbbc5c325be517f4e5fbf6ca2835127b19ee

Add config for labeler (#835) * Add config for labeler * Add custom-op label

view details

Sean Morgan

commit sha 4f88ac52c096d18e584bdb2d9b353cb1d15f986b

Remove tf function where unecessary (#823) * Remove decorators wrapping simple custom op calls * Remove decorator from functions that are likely to be re-traced * Checkpoint * Revert * Modify test cases * Add unknown test * Lint * Lint

view details

Sean Morgan

commit sha 4997d706f7fab8643856152de4f9bd64c9a23009

Add Github CI for windows (#834) * Add Github CI for windows * Update job names * Update python version for mac to match * Remove npairs from windows build and tests * Cleanup * SkipTest benchmark op * No skipTest available * Lint * Remove op build

view details

Mark Daoust

commit sha 8d768db8cb31783612246ce85e72e15cfd00ebe8

Update site (#840) * Some fixes for republishing tensorflow.org

view details

Tzu-Wei Sung

commit sha 06d686e8859a3c6496f81ba8ee9b6247a12282bb

FIX: GELU name (#843) * FIX: GELU name

view details

Daniel

commit sha 2b84e03933c0176f89e2c35fa1a344a7fa5aaaff

Fix bug in "xy" mode for bilinear interpolation (#845) * Fix bug in "xy" mode for bilinear interpolation * fixed error in interp implementation, and added test for non-square images. * removed erroneous non-square test, and modified small_grid tests to be non-square.

view details

Sean Morgan

commit sha 81d75ec27c4fcb9aa095f511eb6a4b6965eba28e

Add broken tutorial do docs for testing verification (#832)

view details

Sean Morgan

commit sha 0b11bc55f91f3ea0c9c80a55f831005192c49f70

CLN: Remove sanity nightly badge (#850)

view details

Sean Morgan

commit sha 1345091f328d1b03bade72f92cd22fdc71178989

DOC: Update build matrix (#849)

view details

push time in a month

Pull request review commenttensorflow/addons

Replaced hardshrink implementation by a pure python one.

 # ==============================================================================  import tensorflow as tf-from tensorflow_addons.utils.resource_loader import LazySO -_activation_so = LazySO("custom_ops/activations/_activation_ops.so")++def _hardshrink(x, lower, upper):+    mask_lower = x < lower+    mask_upper = upper < x+    mask = tf.logical_or(mask_lower, mask_upper)+    mask = tf.cast(mask, x.dtype)+    return x * mask+++def compile_with_xla(func, dtype):

I'm wondering if we could create a xla decorator here?

gabrieldemarmiesse

comment created time in a month

pull request commenttensorflow/addons

Target tensorflow stable instead of tensorflow-nightly.

@seanpmorgan @tensorflow/sig-addons-maintainers what's your opinion?

gabrieldemarmiesse

comment created time in a month

pull request commenttensorflow/addons

add ciou and diou

@fsx950223 Hi, I was not active on github these two month, could you please link to the related issue (feature request or some discussion)?

fsx950223

comment created time in a month

issue commenttensorflow/addons

Problems with CohenKappa metrics

@AakashKumarNain Hi, Aakash, I'm sure I understand your question throughly. For tf.keras, I believe it use different metrics to handle different label formats (eg: AUC, BinaryAccuracy, CategoricalAccuracy, etc)

rere-rere

comment created time in a month

issue commenttensorflow/addons

Problems with CohenKappa metrics

@WindQAQ Tzu-Wei, do you think if is_compatible_with work in the case?

rere-rere

comment created time in a month

push eventfacaiy/facaiy.github.io

Yan Facai (颜发才)

commit sha 8e652623d86b30a36e7504d78d4c63bd524189ee

新增:2019读书总结

view details

push time in 2 months

push eventfacaiy/facaiy.github.io

Yan Facai (颜发才)

commit sha cdca559874337a493967733e45c6a232ae9fc756

更新2019年统计

view details

push time in 2 months

push eventfacaiy/facaiy.github.io

Yan Facai (颜发才)

commit sha cbddd94e33924e5f85d44fa6a37fd66836b87837

更新2019年统计

view details

push time in 2 months

push eventfacaiy/facaiy.github.io

Yan Facai (颜发才)

commit sha 36a34801b9086894d1bc1899ef222383546a2851

2020年新年愿望

view details

push time in 2 months

create barnchfacaiy/facaiy.github.io

branch : mac_install

created branch time in 2 months

pull request commenttensorflow/addons

Rrelu fix

Thanks for pinging me, Dheeraj :-)

Thank you for the PR, @fsx950223 . I'm fine with the idea: replace c++ op by tf python op. What do you think, @seanpmorgan @WindQAQ ?

fsx950223

comment created time in 2 months

Pull request review commenttensorflow/addons

Rrelu fix

 from tensorflow_addons.activations import rrelu from tensorflow_addons.utils import test_utils --def _ref_rrelu(x, lower, upper):-    return tf.where(x >= 0, x, (lower + upper) * x / 2)+SEED = 111111   @test_utils.run_all_in_graph_and_eager_modes class RreluTest(tf.test.TestCase, parameterized.TestCase):     @parameterized.named_parameters(("float16", np.float16),                                     ("float32", np.float32),                                     ("float64", np.float64))-    @tf.function     def test_rrelu(self, dtype):         x = tf.constant([-2.0, -1.0, 0.0, 1.0, 2.0], dtype=dtype)         lower = 0.1         upper = 0.2-        result = rrelu(x, lower, upper, training=False)-        expect_result = _ref_rrelu(x, lower, upper)-        self.assertAllCloseAccordingToType(result, expect_result)++        training_results = {+            np.float16: [-0.288330078, -0.124206543, 0, 1, 2],+            np.float32: [-0.26851666, -0.116421416, 0, 1, 2],+            np.float64: [-0.3481333923206531, -0.17150176242558851, 0, 1, 2],+        }+        for training in [True, False]:+            with self.subTest(training=training):+                tf.random.set_seed(SEED)

Is it necessary?

fsx950223

comment created time in 2 months

Pull request review commenttensorflow/addons

Add github action to test with macos

 export TF_NEED_CUDA=0 export PYTHON_BIN_PATH=`which python` ls -alh $PYTHON_BIN_PATH # Use default configuration here.-yes 'y' | ./configure.sh+echo 'y' | ./configure.sh

A little curious: why use echo instead here?

seanpmorgan

comment created time in 2 months

issue commenttensorflow/addons

Why I can't install tensorflow-addons by pip in windows, but can install in linux?

As https://github.com/tensorflow/addons/pull/771 has been merged into master branch, I suppose Windows support will be in the coming 0.7 release. cc @seanpmorgan Sean

over-shine

comment created time in 2 months

push eventfacaiy/addons

Sean Morgan

commit sha ccca5fb4d81def6dbbdcdb34192fa919541ba6c0

BLD: Refactor configure (#808) * Update configure script to take user input * Update build scripts * Simplify documentation. * Enable windows nightly now that it's available.

view details

Dheeraj R Reddy

commit sha b74d5a2c1c3436bc4453ec4d9f3ef3c391e0d058

Create new AveragedOptimizerWrapper optimizer (#760) * Create new AverageWrapper optimizer * SWA and MovingAverage extend AverageWrapper and implement `average_op(var, average_var)` * Add support for BatchNorm * Add fit_bn in optimizers.utils

view details

Calvin

commit sha f663cb3a5f8176b789a9185790c7f838abbbbf2f

Sequence-to-Sequence NMT (#806) * Created using Colaboratory - Sequence-to-Sequence NMT

view details

Sean Morgan

commit sha 2de92d51297c7cfa631512dea4eaa3cc32530164

Broken tutorial for testing (#772)

view details

push time in 2 months

push eventfacaiy/facaiy.github.io

Yan Facai (颜发才)

commit sha a906f63a7f19421bdec444214bcd8bb865031ebe

愿望清单:更新12月

view details

push time in 2 months

push eventfacaiy/addons

Sean Morgan

commit sha c0cebf5166878f0e17eada861df45154fde79a73

HOTFIX: Wait for upstream fix (#614)

view details

Sean Morgan

commit sha 7f0f3d6808d0930c155f8ecfdb8dbc952efdbe8e

DOC: Standardize doc information (#600)

view details

Sean Morgan

commit sha 5edc422571a7a0a64726660e8f86e95fe3331ba7

Pin to greater than tensorflow core (#618)

view details

junjiek

commit sha a28d42fe3617279e938ae782b7a4a91c6df8065a

Implements LAMBOptimizer (#491) * Implements LAMBOptimizer

view details

Sean Morgan

commit sha 034b3c0647faa9ea91777bf4b0f972e29f59b4af

Revert requirements patch (#616)

view details

prcvih

commit sha 0e6d79f7193967ac7864803231209fc2fb18f3fc

Minor correction (#604) * minor error correction * space * consistence * error message * edit variable name * minor * remove space * sanity

view details

Sean Morgan

commit sha 2e9ea8be42e1369cea314b03b78124f981e3d0c8

Add lr setters/getters to optimizer wrappers (#617) * Add lr setters/getters to optimizer wrappers * Lint

view details

Sean Morgan

commit sha 7801294790d79b15f007d2156ae623f042e41324

FIX: pass numpy arrays to keras model (#622) * FIX: pass numpy arrays to keras predict * Fix pylint ignore * Final lint fix

view details

who who who

commit sha 8a49f91483d67e85c8e8d4f67cd6aea8c45cd2b8

add rrelu kernel (#573) * add rrelu kernel * format code * change test case * simplify kernel implemention * fix test case * refact test case * remove forward training test * format code * modify test case * add forget files * update kernel random implementation * fix rrelu test bug * add TODO * fix comment

view details

pkan2

commit sha 921b76de708a9bed94bef93515052249740040f0

Add in tutorial for ConditionalGradient Optimizer (#590) * Add files via upload

view details

prcvih

commit sha 88246d302490a4b68609332a045b5b536b545e29

Refactoring learning rate (lr) (#636) * refactoring lr * lr getter&setter

view details

Matt Elsey

commit sha 5de7a4c6a1bd38c234a3453511c28638de6c4f2c

Bugfix: correct input shape handling for CorrelationCost Keras layer (#635) * Fix in CorrelationCost::compute_output_shape * Cleanup in CorrelationCostTest for duplicated code

view details

Tzu-Wei Sung

commit sha 093cdfa85d334cbe19a37624c33198f3140109ed

add mish kernel (#569) * add mish

view details

prcvih

commit sha ebf0710e79145d3dde24478f5c1797049e37fe65

update deprecated API (#645) * update deprecated add_variable * update deprecated add_variable

view details

Sean Morgan

commit sha 920dc36f533c23352f86dff761677e10f0544fa2

HOTFIX: Pin tf-nightly (#650) * HOTFIX: Pin tf-nightly

view details

Matt Elsey

commit sha d211b0bb11dbd519d48e72db0ef1e3e29f5f20b7

Make interpolate_bilinear compatible with TF Dataset (#332) (#642) * Make interpolate_bilinear compatible with TF Dataset (#332) * Make interpolate_bilinear compatible with TF Dataset * Split InterpolateBilinearTest from DenseImageWarpTest * Add InterpolateBilinearTest::test_unknown_shape * Apply suggestions from code review * Move # pylint: enable=bad-continuation outside control flow block to satisfy code format tools and match disable indentation * Co-Authored-By: Tzu-Wei Sung <windqaq@gmail.com>

view details

Shun Lin

commit sha a10898ede4c04023bf33c3ba483c0f93fe288959

Adds TQDMProgressBar in callbacks (#610) * added TQDMProgressBar and tutorial

view details

Shun Lin

commit sha e4c974d625975ef278619a64a5a83e45f8b06026

[Small typo] fixed typo in tqdm_progress_bar colab tutorial (#653)

view details

Matt Elsey

commit sha 22abf3cb877768d64daf7acf742e39e8fec42ad6

Add unit test verifying tf.Dataset support for dense_image_warp (#332) (#654)

view details

Shun Lin

commit sha bcd1f7b2c6e85ef8fa0795ff195b9555fe5d42fa

update _toc.yaml with path to TQDM Progress Bar colab (#657)

view details

push time in 2 months

push eventfacaiy/addons

Sean Morgan

commit sha a90faf24a3f62358d43fd5d061af0ec92c4fc8c9

Temporarily pin protobuf to return CI (#599) * Temporarily pin protobuf to return CI * Pin to build_deps

view details

Sean Morgan

commit sha 40de2b942c833c6afccd96759f106640b983953b

Standardize lambda with other frameworks (#601) * Standardize lambda with other frameworks * Update test cases

view details

push time in 4 months

issue commenttensorflow/addons

Migrate gelu to core

I agree that C++ implementation should run faster than python implementation, Tzu-Wei. But as the op is so small, I'm not sure whether it is worthy at the cost of maintaining a chunk of code. Moreover, the new op we add might not work well with graph optimization toolkit(say op fusion as you mentained, or TensorRT ect) because those tools might only focus on tf-core. I'm not against c++ implementation, just a little reminder: perhaps we could write c++ implementation or create a fused op when performance sucks in the mainstream modes 😄

WindQAQ

comment created time in 4 months

issue commenttensorflow/addons

MovingAverage does not work with MirroredStrategy

@vmarkovtsev Hi, Vadim, can you provide a minimal reproducible example? Thank you

vmarkovtsev

comment created time in 4 months

issue commenttensorflow/addons

MovingAverage does not work with MirroredStrategy

@PhilJd Hi, Phil, could you take a look? Thanks

vmarkovtsev

comment created time in 4 months

Pull request review commenttensorflow/addons

correction, typeError

 def test_invalid_filter_shape(self):         msg = ("The `filter_shape` argument must be a tuple of 2 integers.")         image = tf.ones(shape=(1, 28, 28, 1)) -        for filter_shape in [(3, 3, 3), (3, None, 3), None]:+        for filter_shape in [(3, 3, 3), (3, None, 3)]:             with self.subTest(filter_shape=filter_shape):                 with self.assertRaisesRegexp(ValueError, msg):                     mean_filter2d(image, filter_shape=filter_shape) +        filter_shape = None+        with self.subTest(filter_shape=filter_shape):+            with self.assertRaisesRegexp(TypeError, msg):

seems kind of inconsistency, Tzu-Wei @WindQAQ

autoih

comment created time in 4 months

pull request commenttensorflow/addons

Minor correction

The exception message seems not consistency for unknown shape and wrong shape, what do you think, Tzu-Wei @WindQAQ

image

autoih

comment created time in 4 months

pull request commenttensorflow/addons

CLN: Refactor f_scores and f_test

Sorry for the delay, Philip. I'm referring to test dependencies when I use 'dependency' above. Anyway, I'm not against the sklearn proposal if you insist :-) What do you think, Dheeraj @Squadrick ?

Squadrick

comment created time in 4 months

issue commenttensorflow/addons

Use package manager to pin TF dependency

Yes, I have the same concern with Guillaume. Perhaps we can always use greater-than version by default when releasing major version(say, tf >= 2.0 for addons 0.5). And we can release a minor version to pin tf version until there is a break or performance issue(say, tf>=2.0, <2.2 for addons 0.5.3).

seanpmorgan

comment created time in 4 months

delete branch facaiy/addons

delete branch : BLD/use_latest_tf

delete time in 4 months

delete branch facaiy/community

delete branch : addons_freeze_setup_py

delete time in 4 months

push eventfacaiy/addons

Yan Facai (颜发才)

commit sha c4af0f60d9e78ae1a0b6937a0ad3c86aed1e732b

BLD: always install the latest tf in configure.sh (#587)

view details

prcvih

commit sha 50ed3d99442247c9592e1ad61aadc5ef74fea34c

typo (#593)

view details

Sergii Khomenko

commit sha 8088e830a4b1657c2b2f1f17059a2abf01905857

Update text for triplet loss tutorial (#595)

view details

Guillaume Klein

commit sha dfaa5c0f0fd441ab7d14f31a80e3de5bddc3ef9e

Define custom_op_library Bazel macro (#581) * Define custom_op_library Bazel macro * Apply buildifier suggestion

view details

push time in 4 months

delete branch facaiy/addons

delete branch : BLD/use_tf_nightly

delete time in 4 months

PR opened tensorflow/addons

Reviewers
BLD: always install the latest tf in configure.sh

Per https://groups.google.com/a/tensorflow.org/forum/#!topic/addons/2sS8FqhhU2w

+3 -3

0 comment

1 changed file

pr created time in 4 months

create barnchfacaiy/addons

branch : BLD/use_latest_tf

created branch time in 4 months

push eventfacaiy/addons

Mark Daoust

commit sha a603625227f0c39de7f68fde93ff0e23b602f284

Hide the template from the tensorflow site. (#566)

view details

Sean Morgan

commit sha c6107bea9483b107cea28cbe73c1e8b27b33c373

Increase minor version (#568)

view details

Sean Morgan

commit sha 4c718c16884a25b6eb690260ef5a7f1d19287182

Tutorials increase version (#571) * Update _template.ipynb * Update image_ops.ipynb * Update layers_normalizations.ipynb * Update layers_weightnormalization.ipynb * Update losses_triplet.ipynb * Update optimizers_lazyadam.ipynb

view details

prcvih

commit sha 9485cf57441244e14b08c2fc17e075a570158d91

correct link (#572) * correct link * link correction

view details

prcvih

commit sha 9ed027d688fdeb22a3780be280ee61cb5a8a0e88

avoid substraction ops to maintain precision (#557) * avoid substraction ops to preserve the precision

view details

Sean Morgan

commit sha ac8418b205e9279e4605c4dff89381c85e4dcda0

DOC: Add migration documentation (#561) * Add migration documentation * Update migrations * Format

view details

Guillaume Klein

commit sha 6c2e60a9884e7484db3d679b9478ae97dc09c44e

Simplify gather_tree_from_array implementation (#576)

view details

Amirhossein Kazemnejad

commit sha 83d28c90dcdf8d361ee5d3102258d924459e911c

Add support for masked input in TrainingSampler (#546) * Add support for masked input in TrainingSampler * Add tests * Fix lint * Add right padded sequence assertion * Fix code format * Update the test case to use more up-to-date APIs * Change the initialize method behaviour It will throw an error if both sequence_length and mask are provided. * Change TrainingSampler to assume no masking, when no length information is provided. * Add comment on how the sequence length should be inferred

view details

Tzu-Wei Sung

commit sha 396b4a535d41df1819f1f359f832f69ace34d636

add softshrink kernel (#570) * add softshrink

view details

prcvih

commit sha 6f61476bc575f48fc223046931139cd8279f8e2a

correction, typeError (#545) * correction, typeError * revise * add test * add typeerror * sanity indentation * get rid of the first for loop * test filter shape * remove unnecessary * minor

view details

Mark Daoust

commit sha 7db32b38b356dadaab9938e2ccae6a28b627090f

Docs update for tensorflow.org (#578) * Simplify yaml files. * Add tensorflow.org buttons * switch imports to %tensorflow_version * switch tensorflow_version before installing addons * Switch to one of our images * Test-run in Colab * Add try/except for colab * verbose = 2 * Use --no-deps * Update template, clear output * add download links

view details

Yan Facai (颜发才)

commit sha 86707616678697417b4c5cc4da19e60487106f79

BLD: switch to tf-nightly (#567) * BLD: switch to tf-nightly * BLD: use tf 2.0.0

view details

Sean Morgan

commit sha f0a2b0743433b7db7c601dff177e3547fba8a178

Add API documentation badge (#584)

view details

who who who

commit sha 9fcad867102cc6c6e6f2ea4fe35642685f8249a3

fix typo (#585)

view details

Mark Daoust

commit sha 8ff0b54c28827bb0462b5f21707241e7162f56fa

Add overview page, regenerate image_ops output. (#583) * Add overview page (copied from README.md) * Trim overview page * Add overview to TOC. * H1/H2 fixes, backticks * Include output * Fix overview H1 * Fix button links in weight_normalization * Fix common typos. * Clear download bar.

view details

push time in 4 months

pull request commenttensorflow/addons

CLN: Refactor f_scores and f_test

@PhilipMay Hi, Philip, it's easy to add a new dependency while difficult (sometimes impossible) to remove one, that's why I suggest to act conservatively. Moreover, sklearn is a quite complicated python wheel which has many dependencies on its own.

Squadrick

comment created time in 4 months

issue commenttensorflow/addons

(Speech recognition) hope to get inside sMBR, MWER tensorflow code

Welcome, Prakhar!

@utkarshsimha Hi, Utkarsh, could you keep us updated? Thanks?

liangzhenlin123

comment created time in 4 months

pull request commenttensorflow/addons

CLN: Refactor f_scores and f_test

I'm afraid that sklearn is too heavy, what do you think @seanpmorgan @WindQAQ ?

Squadrick

comment created time in 4 months

issue commenttensorflow/addons

Migrate gelu to core

Is it still okay for us to issue a single PR to cover core and keras part?

We usually add both c++ and python codes when implementing a new OP. So I think the answer is yes. But I'm a little curious why we write c++ kernel implementation for the activation ops. Those ops looks quite simple, and it seems that they could be composed by tf python ops easily if I'm not wrong.

WindQAQ

comment created time in 4 months

issue commenttensorflow/addons

image.rotate - non constant angles - error

Hi, David. Could you open an issue for tf? I believe we could get more help/support about installing environment there.

DawyD

comment created time in 4 months

Pull request review commenttensorflow/addons

BLD: switch to tf-nightly

 if project_name == TFA_RELEASE:     # TODO: remove if-else condition when tf supports package consolidation.     if platform.system() == 'Linux':-        REQUIRED_PACKAGES.append('tensorflow-gpu == 2.0.0-rc0')+        REQUIRED_PACKAGES.append('tensorflow-gpu == 2.1.0')

Sure, done :-)

facaiy

comment created time in 4 months

push eventfacaiy/addons

Yan Facai (颜发才)

commit sha 2cdf9535405eb242ca9271c62eca5f1f7aa7bcd5

BLD: use tf 2.0.0

view details

push time in 4 months

issue commenttensorflow/addons

Package Addons 0.6

@seanpmorgan Sean, we should bump up 0.7.0 version in master branch.

facaiy

comment created time in 5 months

PR opened tensorflow/addons

Reviewers
BLD: switch to tf-nightly

Per https://groups.google.com/a/tensorflow.org/forum/#!topic/announce/DIDUrh6LZbo

TensorFlow 2.0 nightly pip packages are now available under the tf-nightly project starting 10/3/19. tf-nightly-2.0-preview will no longer be updated the nightly tensorflow and tensorflow-gpu pip packages consolidation as announced previously has been extended to the nightly pip packages as well.

+6 -17

0 comment

5 changed files

pr created time in 5 months

create barnchfacaiy/addons

branch : BLD/use_tf_nightly

created branch time in 5 months

push eventfacaiy/addons

Sean Morgan

commit sha d20e803f26728fd6bac784a12749f09751404ad0

BLD: Fix default GPU toolchain (#524) * Refactor build to work outside of docker

view details

pkan2

commit sha 813e88d12b00681f1a361a0f718dcef2b68c1935

Modify ConditionalGradient optimizer with handling case of gradient be 0 (#558) * Add files via upload * Add files via upload * Add files via upload * Add files via upload * Add files via upload * Add files via upload * Add files via upload * Add files via upload * Add files via upload * Add files via upload * Add files via upload * Add files via upload * Add files via upload * Add files via upload * add CG optimizer * Revert "add CG optimizer" This reverts commit 953fa39bd19aabf04615c82c86b4032cb28a3826. * Add files via upload * Add files via upload * Add files via upload * Add files via upload * Add files via upload * Revert "Add files via upload" This reverts commit de3cf0b5e566eca167269b085bc363a78feb868e. * Handling the case of gradient be 0 * Handle the case of gradient to be 0 * Modify the format of epsilon in the argument * Adding missing part in init function for epsilon * fixing line overlong issue

view details

Vladimir Starostenkov

commit sha 8c94e2f687acbe74c00cd9cce96d427344f10839

Remove shuffling for the test dataset (#559) In order to have a good picture with https://projector.tensorflow.org/

view details

Tzu-Wei Sung

commit sha 9e9031133c8362fedf40f2d05f00334b6f7a970b

unify activations and tests (#551) * clean up activation/test * test general properties for activations

view details

push time in 5 months

issue closedtensorflow/addons

Package Addons 0.6

TensorFlow2.0.0 is out. So it's time to package and release IMO.

@seanpmorgan @tensorflow/model-maintainers what do you think, all?

closed time in 5 months

facaiy

issue commenttensorflow/addons

Package Addons 0.6

Congrats, closed by https://github.com/tensorflow/addons/pull/565 👍

facaiy

comment created time in 5 months

delete tag tensorflow/addons

delete tag : 0.6.0

delete time in 5 months

Pull request review commenttensorflow/addons

correction, typeError

 def test_normalize_tuple(self):                          keras_utils.normalize_tuple((2, 1, 2),                                                      n=3,                                                      name='strides'))+        # with self.assertRaises(TypeError):

Why to comment this two lines?

autoih

comment created time in 5 months

Pull request review commenttensorflow/addons

correction, typeError

 def test_invalid_filter_shape(self):         msg = ("The `filter_shape` argument must be a tuple of 2 integers.")         image = tf.ones(shape=(1, 28, 28, 1)) -        for filter_shape in [(3, 3, 3), (3, None, 3), None]:+        for filter_shape in [(3, 3, 3), (3, None, 3)]:             with self.subTest(filter_shape=filter_shape):                 with self.assertRaisesRegexp(ValueError, msg):                     mean_filter2d(image, filter_shape=filter_shape)+        +        for filter_shape in [None]:  +            with self.subTest(filter_shape=filter_shape):+                with self.assertRaisesRegexp(TypeError, msg):

Does it make sense? cc @WindQAQ

autoih

comment created time in 5 months

pull request commenttensorflow/tensorflow

Add gradient accumulation support for distribution strategy

@fanshiqing Shiqing, would you like to close this and open a PR to addons?

fanshiqing

comment created time in 5 months

push eventfacaiy/addons

pkan2

commit sha 91c08460ffeb41388f9a40ce0fb45585adaaca10

Implement Conditional Gradient Optimizer (#469)

view details

Guillaume Klein

commit sha d2f7db23f58c6fa773bebaefe44848ba0ddd6db4

Fix minor typo in contributing guide (#538)

view details

Helin Wang

commit sha d227cbdb83832ac04479b07db136213f9d4fc9d1

Implement Addons>ParseTime operator. (#530) The parse time operator parses an input string according to the provided format string into a Unix time, the number of seconds / milliseconds / microseconds / nanoseconds elapsed since January 1, 1970 UTC. Fixes: https://github.com/tensorflow/addons/issues/492

view details

Tzu-Wei Sung

commit sha a4379eaadb12fead10c23a402f8dcefb78f7d1da

add lisht kernel (#529) * add lisht kernel * update README * format code * fix tolerance * reorder the computation * unify namespace * clean up testcase * format code * fix typo * fix namespace comment * remove extra the * change test size to small

view details

Guillaume Klein

commit sha ea26f14d74399bdb2946f7167c6d766938abcb7f

Link to cudart_static and remove cuda_libs (#539)

view details

Amirhossein Kazemnejad

commit sha df256b9e18cd5111a39d2795982da911811b8e32

Allow manual memory reset in AttentionMechanism (#547) * Add support for manual memory reset * Better code style in MyModel class Remove solved #TODO

view details

Helin Wang

commit sha aa16608d0236972406f62a0cefdcdbe635141096

Update README and codeowner for parse time op (#552)

view details

push time in 5 months

created tagtensorflow/addons

tagv0.5.2

Useful extra functionality for TensorFlow 2.0 maintained by SIG-addons

created time in 5 months

release tensorflow/addons

v0.5.2

released time in 5 months

push eventtensorflow/addons

Helin Wang

commit sha aa16608d0236972406f62a0cefdcdbe635141096

Update README and codeowner for parse time op (#552)

view details

push time in 5 months

PR merged tensorflow/addons

Reviewers
Update README and codeowner for parse time op awaiting testing (then merge) cla: yes
+4 -1

1 comment

2 changed files

helinwang

pr closed time in 5 months

delete branch facaiy/addons

delete branch : ENH/move_LazyAdamOptimier

delete time in 5 months

delete branch facaiy/addons

delete branch : CLN/move_ci_test

delete time in 5 months

delete branch facaiy/addons

delete branch : CLN/remove_tf_test_util

delete time in 5 months

delete branch facaiy/addons

delete branch : tf_addons_doc_preview

delete time in 5 months

delete branch facaiy/addons

delete branch : test_r0.5

delete time in 5 months

Pull request review commenttensorflow/addons

Update README and codeowner for parse time op

 |:---------- |:----------- |:------------- | | skip_gram_ops |  |  | | crf | Dheeraj R. Reddy | dheeraj98reddy@gmail.com |+| parse_time_op | Helin Wang | helin@google.com |  ## Components  | Submodule  | Text Processing Function |  Reference  | |:---------- |:----------- |:----------- | | skip_gram_ops |  skip_gram_sample | https://arxiv.org/abs/1301.3781 | | skip_gram_ops |  skip_gram_sample_with_text_vocab | https://arxiv.org/abs/1301.3781 |+| parse_time_op | parse_time | https://github.com/tensorflow/addons/issues/492 |

Just leave Reference empty :-)

helinwang

comment created time in 5 months

issue commenttensorflow/tensorflow

GELU activation Functions?

@bwindsor22 Hi, Brad. Gelu activation has been implemented in Addons, you might be interested in it: https://github.com/tensorflow/addons/blob/master/tensorflow_addons/activations/README.md

Feel free to open a PR and assign me as a reviewer.

@fchollet @karmel Hi, François, Karmel. Is it time to port gelu from addons to tf-core? I believe Tzu-Wei @WindQAQ would like to help it.

bwindsor22

comment created time in 5 months

issue commenttensorflow/addons

Migrate gelu to core

  • @karmel
WindQAQ

comment created time in 5 months

delete branch facaiy/addons

delete branch : r0.5

delete time in 5 months

push eventtensorflow/addons

Yan Facai (颜发才)

commit sha 1c6862eb0d655ba20a360af90bd3b2a1c5f5ab86

RELEASE: 0.5.2 (#549) * RELEASE: 0.5.2 * BLD: disable too-many-function-args pylint * BLD: built against 2.0.0

view details

push time in 5 months

PR merged tensorflow/addons

Reviewers
RELEASE: 0.5.2 cla: yes

close #541

+6 -5

4 comments

5 changed files

facaiy

pr closed time in 5 months

pull request commenttensorflow/addons

RELEASE: 0.5.2

@bionicles Agreed. I filed an issue https://github.com/tensorflow/addons/issues/553 to track 0.6 release, please subscribe it

facaiy

comment created time in 5 months

issue openedtensorflow/addons

Package Addons 0.6

TensorFlow2.0.0 is out. So it's time to package and release IMO.

@seanpmorgan @tensorflow/model-maintainers what do you think, all?

created time in 5 months

push eventfacaiy/addons

Yan Facai (颜发才)

commit sha a8134d66fd55d57d007a5b5b4ea412665411aab6

BLD: built against 2.0.0

view details

Yan Facai (颜发才)

commit sha 457c8cc830e7ac605e83f6efdaed88aa6d78361f

BLD: disable release

view details

push time in 5 months

pull request commenttensorflow/addons

RELEASE: 0.5.2

TensorFlow 2.0 final has just been released. Should it be used for this new version?

Good new, Guillaume! Of course, thank you for reminding me 😄

facaiy

comment created time in 5 months

push eventfacaiy/addons

Yan Facai (颜发才)

commit sha a8134d66fd55d57d007a5b5b4ea412665411aab6

BLD: built against 2.0.0

view details

push time in 5 months

push eventfacaiy/addons

Yan Facai (颜发才)

commit sha 95ba5bc73de6c5383a94b8d1b0a66760fdd8560e

BLD: disable release

view details

push time in 5 months

create barnchfacaiy/addons

branch : test_r0.5

created branch time in 5 months

push eventfacaiy/addons

Yan Facai (颜发才)

commit sha 015e6948ec62a0159624137dcec7fc8864d4b3e9

BLD: disable too-many-function-args pylint

view details

push time in 5 months

PR opened tensorflow/addons

RELEASE: 0.5.2

close #541

+5 -5

0 comment

4 changed files

pr created time in 5 months

pull request commenttensorflow/addons

BLD: Fix default GPU toolchain

Currently building GPU package from source fails outside of the docker container.

Sorry for my oversight. I'm afraid that it might be not easy to support due to various environment, that's why we use docker container.

seanpmorgan

comment created time in 5 months

pull request commenttensorflow/addons

BLD: Fix default GPU toolchain

@olesalscheider Thanks for your feedback, Niels!

seanpmorgan

comment created time in 5 months

pull request commenttensorflow/addons

BLD: Fix default GPU toolchain

@SebastienKeroack Did you run those commands in docker container?

seanpmorgan

comment created time in 5 months

push eventfacaiy/addons

Yan Facai (颜发才)

commit sha fba96704e60be1ca505834a8873c3c0872a45612

RELEASE: 0.5.2

view details

push time in 5 months

push eventfacaiy/addons

Guillaume Klein

commit sha 4b04a3500a607fcae514b28754c5aca5f37dd130

Link to cudart_static and remove cuda_libs (#539) (#548)

view details

push time in 5 months

push eventtensorflow/addons

Guillaume Klein

commit sha 4b04a3500a607fcae514b28754c5aca5f37dd130

Link to cudart_static and remove cuda_libs (#539) (#548)

view details

push time in 5 months

pull request commenttensorflow/addons

Cherry-pick to r0.5: Link to cudart_static and remove cuda_libs

Link to https://github.com/tensorflow/addons/issues/541

guillaumekln

comment created time in 5 months

pull request commenttensorflow/addons

Link to cudart_static and remove cuda_libs

@guillaumekln Hi, Guillaume. Could you backport the fix to r0.5 branch? Thanks 😄

guillaumekln

comment created time in 5 months

create barnchfacaiy/addons

branch : r0.5

created branch time in 5 months

pull request commenttensorflow/addons

correction, typeError

@autoih Sure, please take a look at https://github.com/tensorflow/addons/blob/c8372e2b6499502908f1fa5551f6daca1097792e/tensorflow_addons/utils/keras_utils_test.py

autoih

comment created time in 5 months

issue commenttensorflow/addons

Package TFA 0.5.2 pinned to TF2-RC2

As #532 has been fixed by #539, we'll plan to release 0.5.2.

seanpmorgan

comment created time in 5 months

issue closedtensorflow/addons

Package import fails when using tensorflow-gpu and CUDA libraries are missing

System information

  • OS Platform and Distribution: Ubuntu 16.04
  • TensorFlow version and how it was installed: 2.0.0rc1 (binary)
  • TensorFlow-Addons version and how it was installed: 0.5.1 (binary)
  • Python version: 3.5.2
  • Is GPU used? No

Describe the bug

When using tensorflow-gpu, importing the tensorflow_addons package fails if CUDA libraries are missing while importing tensorflow itself works without issue.

Code to reproduce the issue

On a non-GPU system:

$ pip install tensorflow-addons==0.5.1
$ python
>>> import tensorflow_addons as tfa

Other info / logs

2019-09-20 18:04:16.387041: W tensorflow/stream_executor/platform/default/dso_loader.cc:55] Could not load dynamic library 'libcudart.so.10.0'; dlerror: libcudart.so.10.0: cannot open shared object file: No such file or directory Segmentation fault

closed time in 5 months

guillaumekln
IssuesEvent

push eventtensorflow/addons

Guillaume Klein

commit sha ea26f14d74399bdb2946f7167c6d766938abcb7f

Link to cudart_static and remove cuda_libs (#539)

view details

push time in 5 months

issue closedtensorflow/addons

Package import fails when using tensorflow-gpu and CUDA libraries are missing

System information

  • OS Platform and Distribution: Ubuntu 16.04
  • TensorFlow version and how it was installed: 2.0.0rc1 (binary)
  • TensorFlow-Addons version and how it was installed: 0.5.1 (binary)
  • Python version: 3.5.2
  • Is GPU used? No

Describe the bug

When using tensorflow-gpu, importing the tensorflow_addons package fails if CUDA libraries are missing while importing tensorflow itself works without issue.

Code to reproduce the issue

On a non-GPU system:

$ pip install tensorflow-addons==0.5.1
$ python
>>> import tensorflow_addons as tfa

Other info / logs

2019-09-20 18:04:16.387041: W tensorflow/stream_executor/platform/default/dso_loader.cc:55] Could not load dynamic library 'libcudart.so.10.0'; dlerror: libcudart.so.10.0: cannot open shared object file: No such file or directory Segmentation fault

closed time in 5 months

guillaumekln

PR merged tensorflow/addons

Reviewers
Link to cudart_static and remove cuda_libs cla: yes

This is what is done in the TensorFlow codebase for custom ops:

https://github.com/tensorflow/tensorflow/blob/r2.0/tensorflow/tensorflow.bzl#L1771-L1772

After building a pip package with this change, I no longer get the error reported in #532.

Fixes #532.

+8 -8

9 comments

4 changed files

guillaumekln

pr closed time in 5 months

more