profile
viewpoint
Fred Reiss frreiss IBM Center for Open-Source Data and AI Technologies (CODAIT) San Francisco, CA http://codait.org

jupyter/enterprise_gateway 293

A lightweight, multi-tenant, scalable and secure gateway that enables Jupyter Notebooks to share resources across distributed clusters such as Apache Spark, Kubernetes and others.

IBM/MAX-Image-Caption-Generator-Web-App 43

Create a web app to interact with machine learning generated image captions

CODAIT/graph_def_editor 12

GraphDef Editor: A port of the TensorFlow contrib.graph_editor package that operates over serialized graphs

frreiss/tf-dev-conf 2

Fred's TensorFlow development configuration files.

frreiss/Refactored-MAX-Object-Detector 1

Prototype of new structure for the existing project IBM/MAX-Object-Detector

frreiss/arrow 0

Mirror of Apache Arrow

frreiss/bazel-zlinux-binaries 0

Fred's precompiled Bazel binaries for zLinux

frreiss/docs-fred 0

TensorFlow documentation

frreiss/fred-gde-notebooks 0

Fred's scratchpad Jupyter notebooks for GraphDef Editor development

frreiss/fred-systemml 0

Mirror of Apache SystemML

push eventfrreiss/text-extensions-for-pandas

frreiss

commit sha e7345486bf2a5ab482b59383ff73e832ce7e79e5

Update readme to reflect new package name

view details

push time in 2 days

push eventfrreiss/text-extensions-for-pandas

frreiss

commit sha c73e3123aa490d9dfb1115b9eabc8086b1450804

Rename package to match approved name

view details

push time in 2 days

push eventfrreiss/text-extensions-for-pandas

frreiss

commit sha 42e02a9ecc95b9cb7fdf53903be590dd0e89f549

Added resources dir and changed tokenization to use CharSpanArray

view details

frreiss

commit sha b27801af73d56980d23586ab6e2b83eb272fdc51

Ported algebra code to new column types

view details

frreiss

commit sha bf9466ccdc53f01a83116673e1fbcb49e0ab09b2

Wrapped spaCy dependency parser

view details

frreiss

commit sha 9059b817a029f2ab82f7eebc100e5ed12a26c053

Added rendering of parse trees

view details

frreiss

commit sha aff2f5b9d8e211b134641688b45b164b42cc5045

Refactoring

view details

frreiss

commit sha 3e7c08795c2342d067044309536283a149926183

Initial implementation of Gremlin

view details

frreiss

commit sha c92ae4d3a7cc9858ae6d83887356014bd9f24874

Bug fixes in Gremlin code

view details

frreiss

commit sha 2c8711c58b1e2015fde9264ee6b2945db2f7d3e2

Include results so that they can be rendered on github

view details

frreiss

commit sha 4d65eb85010572deb80b0e853725478c3e996499

On second thought, don't include results

view details

frreiss

commit sha 74f2f8ddbfa5938b75bf9ed19ee93ae8a5c20140

Implemented within and without predicates

view details

frreiss

commit sha 43f8e60e04fe6e79577a2a55cafcc8214f6e75db

Basic repeat traversals work now

view details

frreiss

commit sha 737730a0260e544c873edcdb1f2826abadad13ca

repeat traversal with nested where() works. Also moved to Pandas 1.0

view details

frreiss

commit sha 6937db80e94e1b5063748d38a12b27f0255dde6f

Implemented enough of Gremlin to run the example ESSP rule

view details

frreiss

commit sha f9eeeb2b0435ccd24ca50f60462bb608b8118da4

Adjective rules

view details

frreiss

commit sha ecb5d67980fe8a1501c44b5776cc368940361b92

Code formatting and copyright notices

view details

frreiss

commit sha d034573aeaa47536c947a42815938a4b00368efd

Refactor Gremlin code into multiple files

view details

frreiss

commit sha d8ab7a36ab95d678235c48d70b1f5a0a1fc4d3f8

Made sum == combine for spans

view details

frreiss

commit sha 3afdf05028afed4eef3a6dfc9558f4721d3ec102

updated todo list

view details

frreiss

commit sha 9c37e1233acf3296780d2331c0226c1bb7cf7b95

Description and partial implementation for conditionals

view details

frreiss

commit sha 1c827e6865d6b5e91483a33c920ae6248e227333

Additional work on grouping and aggregation

view details

push time in 2 days

create barnchfrreiss/text-extensions-for-pandas

branch : master

created branch time in 2 days

created repositoryfrreiss/text-extensions-for-pandas

Extensions to Pandas dataframes to support NLP applications

created time in 2 days

startedclang-ykt/ONNF

started time in 2 days

startedIBM/Industry-Accelerators

started time in 3 days

push eventfrreiss/tensorflow-fred

George Karpenkov

commit sha 45d9125f0e183692efd47bc613a4d282f5f47b95

[XLA/GPU] Further refactor starting tile calculation PiperOrigin-RevId: 292941651 Change-Id: I5e03569822971525ef9049bc274d61dfa5abd548

view details

Chris Jones

commit sha 871b20cf6879687eaf91fd8569f6956d5aa5393d

Fix `get_replicated_var_handle` when used with a device assignment. PiperOrigin-RevId: 292946075 Change-Id: I6a927e0dcf8697e6921c3c06545a98aa2a5dc9b3

view details

Robert David

commit sha 73432d563426ac456cba57b2508d0d036c728ff3

Fix discrepancy of SSE optimized functions declared and defined at different SSE levels. Currently they are defined in the source at SSSE 3 level, but declared in the header only when SSE 4.1 is available. Change the header to SSSE 3 too. PiperOrigin-RevId: 292948433 Change-Id: Icf044f55e66d2b9d44b201fb63c64d7de5263f97

view details

Zhenyu Tan

commit sha e3731668fcfda82d0f765dec2c7ae0591340a3e6

API doc fix for keras.preprocessing.image.array_to_img PiperOrigin-RevId: 292949046 Change-Id: Ie9f985d2e04e7d26d1e612d560fe169caaefb316

view details

Zhenyu Tan

commit sha 9e11e055ab11148cb5bb3c798617db670a20bf65

API doc fix for keras.applications.MobileNet. PiperOrigin-RevId: 292952937 Change-Id: I63a5837cd33eb60a63df743e9876e232fba73d08

view details

Jacques Pienaar

commit sha 544ca41bed77411cc8a6eabd1fb420c99a75b463

Populate value hoisting function that LoopLike uses * Populate method of LoopLike op interface and verify with test with LICM. - The function to determine if values are outside the loop is still too conservative and needs to be expanded. * Make canonical form of TFL While to match exported form. This is not strictly required for the optimization format, but makes export simpler as it matches export form. - Check that this constraint is met during translation to flatbuffer. * Add an integration/execution test to verify the export is as expected by interpreter. This was as I had the export wrong initially and the first place I hit the failure was during execution. This is now check during export instead. TFL WhileOp needs additional verification still to verify the types are compatible. PiperOrigin-RevId: 292958195 Change-Id: If8b2b506c406aa183f1b365ab36b03b46e20087a

view details

George Karpenkov

commit sha 88f66126729634dfc35d44897b805082f55ecf37

[XLA/GPU] A more principled check for tiling bounds when emitting reduction Explicitly check the tiling bound in the epilogue instead of relying on the auxiliary variable populated during the tiling loop. PiperOrigin-RevId: 292961855 Change-Id: I2c73cb95828bcfff495af4c883fa05ec2e3f9fc7

view details

TensorFlower Gardener

commit sha 73a4b982e203cdc8da3e8391ee783225d8632858

Merge pull request #36395 from Intel-tensorflow:amin/fix-shape PiperOrigin-RevId: 292962404 Change-Id: I8985228e74eeea9a717b618857d78fc9ded8a31e

view details

HyoukJoong Lee

commit sha 498534708751e9fbe115e65f32263fd43a3a9dac

Extend AllToAll to support flattened operand/output PiperOrigin-RevId: 292964792 Change-Id: I63c8caeb3e08dc559c1cecd22d4573f2d08b04ae

view details

TensorFlower Gardener

commit sha 2571f8a84065bbc6aeb036234b6e41ee9cd43f9c

Merge pull request #35972 from ROCmSoftwarePlatform:google-upstream-rocm-cwise-ops PiperOrigin-RevId: 292964913 Change-Id: Ia9cad91962b274d4410ccfc66614a92682f74d33

view details

Berkin Ilbeyi

commit sha 4f071302415dca5f6d56c528481665edba949891

[XLA] Fix a MSVC compilation issue. PiperOrigin-RevId: 292966321 Change-Id: I203dbd4b2f551224b412c6f9d92bb77b6b461b18

view details

Yanhui Liang

commit sha 1bce0acc099d2c6a3743888b40c3eabce4de27e9

API doc fixit for MobileNetV2. PiperOrigin-RevId: 292967941 Change-Id: Ie6ea25c1013dae26cbcfe99fdccac144682f17af

view details

A. Unique TensorFlower

commit sha 874358298b89af7a2ebaffc9d3276044ae1a2c01

Add a Lookup layer. PiperOrigin-RevId: 292968344 Change-Id: I7dadf31138366d5b0ad8ed7eb0a885647d25bc81

view details

Yuanzhong Xu

commit sha ab94de29f543b21f57e9c862e12f223812775306

[MLIR:TF/XLA] TPU dynamic layout pass A pass that allows TPU input layout to be determined after JIT compilation. This is done by adding run-time ops that interpret compilation result and copy to device with that layout. PiperOrigin-RevId: 292970557 Change-Id: If453686abea1618ba6f88016912ea15e0f0b3d4e

view details

A. Unique TensorFlower

commit sha be453c9869930d5f528e5ae1ca6be9bed4aaefee

For custom optimizers, keep a copy of Session ConfigProto_. PiperOrigin-RevId: 292974333 Change-Id: Ia49fae2149aea5e3e4fec5ba9d4b995406a6a1e4

view details

Gunhan Gulsoy

commit sha 4fc8e2cd8a9ff28aba131295cbc8853466e4e079

Disable flaky keras/distribute:multi_worker_callback_test PiperOrigin-RevId: 292974875 Change-Id: I4067f019c8ba63090a971fd7d32840d883ab34f6

view details

George Karpenkov

commit sha 43be49af994dffaa7c960b7bcbb8f5e27bb4f6a5

[XLA/GPU] [NFC] Introduce local constants for KernelMappingScheme::DimX/Z/Y/Tot PiperOrigin-RevId: 292975342 Change-Id: I425ad28b911b3df59898a9a0e8d8c44b42d7f15a

view details

Anjali Sridhar

commit sha 4316f9d58167978e5df6a5a1febc347fe12980f5

Cast all values that we reduce to be of the same type. PiperOrigin-RevId: 292975462 Change-Id: Id3b388c9f0f72e9bdd5d22702fc5e6bdc857b8c6

view details

Yunxing Dai

commit sha 349bbf3ba6b85bd6d19ac5f42bfff6db630653a1

[XLA] Avoid materializing large literals in creation utils. PiperOrigin-RevId: 292976673 Change-Id: I2c1fa4bb96434f085f293dd5d4ff26d67a4236b5

view details

Jeremy Lau

commit sha 5a7fb2ca626e36cf81adf0f9abc99e4e62769769

Fix asan invalid-bool-load errors in zip_test_lstm by initializing lhs_cacheable and rhs_cacheable before use. PiperOrigin-RevId: 292983666 Change-Id: I05a625a9a5e4273436522eb8a9984c4b44c45fbc

view details

push time in 14 days

Pull request review commenttensorflow/tensorflow

Improve API documentation for WindowDataset op

 def window(self, size, shift=None, stride=1, drop_remainder=False):      A "window" is a finite dataset of flat elements of size `size` (or possibly     fewer if there are not enough input elements to fill the window and-    `drop_remainder` evaluates to false).+    `drop_remainder` evaluates to `False`).++    The `shift` argument determines the number of input elements by which+    the window moves on each iteration.  The first element in the `k`th window+    will be element++    ```+    1 + (k-1) * shift+    ```++    of the input dataset. In particular, the first element of the first window+    will always be the first element of the input dataset.++    If the `stride` parameter is greater than 1, then each window will skip+    `(stride - 1)` input elements between each element that appears in the+    window. Output windows will still contain `size` elements regardless of+    the value of `stride`.      The `stride` argument determines the stride of the input elements, and the

Deleted this paragraph.

frreiss

comment created time in 14 days

Pull request review commenttensorflow/tensorflow

Improve API documentation for WindowDataset op

 def window(self, size, shift=None, stride=1, drop_remainder=False):      A "window" is a finite dataset of flat elements of size `size` (or possibly     fewer if there are not enough input elements to fill the window and-    `drop_remainder` evaluates to false).+    `drop_remainder` evaluates to `False`).++    The `shift` argument determines the number of input elements by which+    the window moves on each iteration.  The first element in the `k`th window+    will be element++    ```+    1 + (k-1) * shift

Changed the equations in this sentence so that windows and elements are numbered starting at zero.

frreiss

comment created time in 14 days

pull request commenttensorflow/tensorflow

Make vscode's pylint integration work on TensorFlow source

Thanks @yifeif!

frreiss

comment created time in 15 days

push eventfrreiss/keras-fred

Filip Wieland

commit sha 87540a2a2f42e00c4a2ca7ca35d19f96e62e6cb0

Clarify expected label formats for crossentropy losses (#12095) * Clarify expected label formats for crossentropy losses * Simplify.

view details

Andrew Docherty

commit sha d78c982b326adeed6ac25200dc6892ff8f518ca6

Fixed loading bug for shared layers accross multiple depths (#11588) * Patched loading bug for shared layers in issue * Fixed PEP8 errors * Fixed PEP8 errors [2] * Added unit test for issue #11159 for model deserialization with mixed layer ordering * Changed to use try/raise for communicating layer loading errors * Addressed review for PR #11588 * Removed redundent layer check, see #11829

view details

omalleyt12

commit sha a6c8042121371b5873773ca767f28cdf5689d5e4

Add an optional `validation_freq` argument to `fit` (#12065) ### Summary Sync with TF Keras. Allows user to specify how often to run validation. ### Related Issues ### PR Overview - [x] This PR requires new unit tests [y/n] (make sure tests are included) - [x] This PR requires to update the documentation [y/n] (make sure the docs are up-to-date) - [x] This PR is backwards compatible [y/n] - [x] This PR changes the current API [y/n] (all API changes need to be approved by fchollet)

view details

bananabowl

commit sha b0bfd5201da2bfced84028bcc5bda05bdfd75af7

Print the model name as part of layer_utils.print_summary (aka model.summary) (#12107) CURRENT: _________________________________________________________________ Layer (type) Output Shape Param # ================================================================= model_1_input (InputLayer) [(None, 128)] 0 PROPOSED _________________________________________________________________ Model: "part_of_speech_tagger" Layer (type) Output Shape Param # ================================================================= model_1_input (InputLayer) [(None, 128)] 0

view details

Noam Wies

commit sha 4185cbb50bfcae9cc30b0fc7b67e81d67a50a8ac

Use var_list argument in TFOptimizer wrapper (#12106) * Update optimizers.py fix bug in TFOptimizer, this optimizer should call tensorflow native optimizer with the named param var_list * add test_tfoptimizer_pass_correct_named_params_to_compute_gradient * fix formating * fix test * fix test * fix formating * fix test

view details

Gabriel de Marmiesse

commit sha e4e342f70dec046fdb1a092e5a12c565f37b9ddc

Cleaning the keras.layers namespace. (#12056) * Cleaning the keras.layers namespace. * Added the aliases in next to the original declarations. * Added ConvLSTM2DCell to the docs. * Putting deserialize_keras_object back where it was. * Putting stackedcells where it was. * Put back InputLayer and InputSpec. * Reverted the changed.

view details

Tarun S Paparaju

commit sha 86ce3cdd7d87d84532d08b0d6b65b3918a10ba5e

Fix spelling mistake in SELU docs (#12123) ### Summary Fixed a spelling mistake in SELU documentation in activations.py. (pre-defined --> predefined) ### Related Issues None ### PR Overview Fix spelling error in docs. - [n] This PR requires new unit tests [y/n] (make sure tests are included) - [y] This PR requires to update the documentation [y/n] (make sure the docs are up-to-date) - [y] This PR is backwards compatible [y/n] - [n] This PR changes the current API [y/n] (all API changes need to be approved by fchollet)

view details

Gabriel de Marmiesse

commit sha c2dc4049c16ab8cbc215bb0ae80cd94a3d9542f5

Displayed some examples in the docs. (#11758) ### Summary The keras examples are great, but do not benefit from a good visibility. I believe we could add them to the docs. It requires very little work since all the code and docstrings have already been written. ### Related Issues ### PR Overview There is some black magic at work with `compile`. I took the code from https://gist.github.com/rduplain/1249199 We can control which files get displayed with `mkdocs.yml`. For the moment only two are displayed. Here is the result: ![1](https://user-images.githubusercontent.com/12891691/49258375-88c73200-f435-11e8-8dc9-95c4201bfcbe.png) ![2](https://user-images.githubusercontent.com/12891691/49258377-8a90f580-f435-11e8-81f0-d6fa42df09a9.png) ![3](https://user-images.githubusercontent.com/12891691/49258378-8bc22280-f435-11e8-98d4-df9e6b677aa0.png) - [ ] This PR requires new unit tests [y/n] (make sure tests are included) - [ ] This PR requires to update the documentation [y/n] (make sure the docs are up-to-date) - [x] This PR is backwards compatible [y/n] - [ ] This PR changes the current API [y/n] (all API changes need to be approved by fchollet)

view details

Frédéric Branchaud-Charron

commit sha 328df7303425c2e3bf4274bb1eb0a2040bab5d8c

Keep input types with Lambda layers (#12122) * Keep input types * CNTK only works with floats * Make the types private

view details

Frédéric Branchaud-Charron

commit sha 2e1a3816676724576f75ce76b18ed3ba2fdea670

Add issue template that redirect to KP and KA (#12115) * Add issue template that redirect to KP * Add KA * Fix link

view details

Tarun S Paparaju

commit sha ee31c9f8c63497d6e2931d5ed45e3638c25ef7e6

Fix spelling mistake in README.md (#12132) ### Summary I have removed an unnecessary hyphen in "fully-configurable" to make it "fully configurable" in the README. ### Related Issues None ### PR Overview Fix a spelling mistake in the README. - [n] This PR requires new unit tests [y/n] (make sure tests are included) - [y] This PR requires to update the documentation [y/n] (make sure the docs are up-to-date) - [y] This PR is backwards compatible [y/n] - [n] This PR changes the current API [y/n] (all API changes need to be approved by fchollet)

view details

Gabriel de Marmiesse

commit sha ec41989421743aa5c11c2ef47b9d0757fdfd5e36

Unpin numpy. (#12142)

view details

Gabriel de Marmiesse

commit sha 55e7875354416c04719eb60c7db2bac78f271df4

Using shutil rather than copying the tree manually in autogen.py (#12146) ### Summary In autogen.py, we copy template into sources and then work on it. But we copy the tree manually. `shutil` has a function for this purpose. ### Related Issues ### PR Overview In the loop, we copy only markdown files. But all the files in `templates` are markdown files so it's fine. - [ ] This PR requires new unit tests [y/n] (make sure tests are included) - [ ] This PR requires to update the documentation [y/n] (make sure the docs are up-to-date) - [x] This PR is backwards compatible [y/n] - [ ] This PR changes the current API [y/n] (all API changes need to be approved by fchollet)

view details

Gabriel de Marmiesse

commit sha 9b67e2b699da8d9ced9af3597906b81ad6f106bc

Added pull request templates. (#12135) * Added a few options for when making pull request. * Added automatic labels and assignees.

view details

Gabriel de Marmiesse

commit sha 59fd3ce3814c62882b540cb6098e9c37f05ceb56

Moving the integration tests to python3. (#12154)

view details

Abhai Kollara Dilip

commit sha 68dc181a5e34d1f20edabe531176b3bfb50001f9

Minor docs fix for fit_generator (#12153) * Docs fix * Fix docstring style

view details

Gabriel de Marmiesse

commit sha 4aba316b7973e246af94ab5ace18fd5d5ffe4f81

Moved the backend __init__ into another file. (#12086) * Moved the backend __init__ into another file. * Used the private variable of load_backend.

view details

Francois Chollet

commit sha 7d25cdca6348af2627b5026ef29eff8754273bc3

Make sure the name of the learning phase in TF does not depend on outer name scopes.

view details

Gabriel de Marmiesse

commit sha 3dbc940b9eb690e92fa6d7fe7ebf897a552a1c50

Using a mutable variable instead of a global variable. (#12170)

view details

Gabriel de Marmiesse

commit sha f42d9e0179f11871179bc9ee4e8c138cd016612b

Cleaned the namespace of keras.backend. (#12076) * Cleaned the namespace of keras.backend. * Fixing some errors. * Added a clear_session to theano. * Fixed usage of private variable. * Removed use of K._config_path * Fixed the reloading issue * Fixed the cannot import name. * Changed the way of using th_sparse_module usage. * Fixed the cntk error. * Fixed the issue of pattern_broadcast. * Added small info in docstring. * Forgot import * Removed empty line. * Removed pattern broadcast and some other functions. * Added clear session only for specific backends. * Removed useless functions.

view details

push time in 24 days

push eventfrreiss/tensorflow-fred

A. Unique TensorFlower

commit sha 68f3bb82b5c6466406a9ccb6154ffca0c86596de

MaxPoolingIndices3D changed to have indices as flattened HWD. PiperOrigin-RevId: 288688358 Change-Id: I9c10a52ac49931e03b704914cc22009ff03e9655

view details

Jens Elofsson

commit sha 98292c11a8748c3bd371b1a2a68b145646e37269

Micro: Fix compile error for Arm Mbed OS.

view details

A. Unique TensorFlower

commit sha 5baf228c21c5061b0af282b419c89f1ddb07a580

ConvolutionTransposed3DAttributes and functions for shape calculation. PiperOrigin-RevId: 288690433 Change-Id: I619c23bb9acfbbd9ad81d73b3e833969273adecf

view details

TensorFlower Gardener

commit sha 8b284be5028c05f1601af95e833d5e5d454da3c1

Merge pull request #35600 from Ikuyadeu:addConst PiperOrigin-RevId: 288691051 Change-Id: Ie02d7f876813a6107ca7b6b67f7877b2f3c2c675

view details

Jacques Pienaar

commit sha 8e5c3aeebdce6f442faacdb706f48af540451996

Use glob patterns instead Models cmake files closer and makes it less fragile for integrates. More cleanup is needed. PiperOrigin-RevId: 288698695 Change-Id: Id9eaac523d642ae550866cda7ea46466197b7316

view details

A. Unique TensorFlower

commit sha 3f9dd57093a172e1be0f747b9841a0626b51e174

Upsample3DAttributes and functions for shape calculation. PiperOrigin-RevId: 288699835 Change-Id: I87c2a05c6d9d2fe30d999b00aec513fa33c667af

view details

Gaurav Jain

commit sha f7239df1a38234bba3fecd1eece22d03641b0407

Use Status directly in TF_TensorFromTensor PiperOrigin-RevId: 288700381 Change-Id: I159ea1c87ee3ca4f10db80b540bb7aedf5a7a967

view details

A. Unique TensorFlower

commit sha f65d625e400ddf7055c6938c2e5dd704919bac4c

Go: Update generated wrapper functions for TensorFlow ops. PiperOrigin-RevId: 288705702 Change-Id: I687dc45d2bb9c11fda711dcfdf8bbd47da38e34a

view details

A. Unique TensorFlower

commit sha 28861546acd53c0c06480ab27c633be2d5e17e69

Exposed GenerateHostResult() and GenerateRecommendation() to header file. PiperOrigin-RevId: 288705912 Change-Id: Iae647aa415c7b011a8b9ced3c729d06c94a919c8

view details

Gaurav Jain

commit sha 37e3630be350428704939e52f94bb6bfa958fb80

Simplify OpInferInputListAttrs to only pass dtypes PiperOrigin-RevId: 288713492 Change-Id: I16b78a0cdb8919e45450f1c1d9c4d9e09e56b97f

view details

Berkin Ilbeyi

commit sha b78869968b7d72ee223b9b2c4855c0ac688e46af

[XLA] Fix a while loop memory corruption bug. Given a situation like: a = fusion(...) ... b = fusion(a) ... t = tuple(a, ...) w = while(t) where the buffer for a is placed in alternate memory for b, but could not be placed in alternate memory for the entire duration of the while loop, we could have erroneously colored the while{0} to be in the alternate memory. This CL introduces a helper method to find the most recent alive allocation to propagate the correct aliased allocation. PiperOrigin-RevId: 288717277 Change-Id: Icdbe51d4bf90c6a1f32d4501ee85c6a146e16f4c

view details

A. Unique TensorFlower

commit sha 18845a465949439428aa7ae977141d44a386e91a

Adds flatbuffer to mlir translation for: MaxPoolWithArgMax, MaxUnpool, and Conv2DTransposeWithBias. Also add tests and fixed a bug. PiperOrigin-RevId: 288718685 Change-Id: I3aebc80ae5ec6493852c9b5c41a5773a22f1d66a

view details

Katherine Wu

commit sha 2b3296441be04839910c11840cc68856a9d43ebf

Fix checkpointing tests. 1. Add calls to `run_restore_ops` to ensure that the restore ops are executed when using checkpoints in graph mode. 2. Ensure that layer orders are the same between the saved model and restored model. Otherwise there will be a race condition when restoring the checkpoint values. PiperOrigin-RevId: 288719809 Change-Id: I6f87a481e9fe3ea1e8ebc667cfea61dcb0716236

view details

A. Unique TensorFlower

commit sha 2ad9dd652f072266ba9ece977119a625536fb5e5

Add more StatType. PiperOrigin-RevId: 288720224 Change-Id: Ieeed4dfcacd5ed435068aa17808d29f3300470bb

view details

Derek Murray

commit sha 5da98db80c1e9f3cb063d9f1fdee04af67e025ee

[tf.data] Reduce noise in the `Dataset.from_tensor_slices(tf.SparseTensor)`. This change uses the single-threaded executor in FlatMapDataset to avoid thread-scheduling noise from function invocation. PiperOrigin-RevId: 288720757 Change-Id: I7d0e7982d90274201449820c2c949a98ad612335

view details

Paul Baranay

commit sha 2e2522004f685c1d0f0fe2f735f1c121c78d9ab0

Add doc summary for tf.strings.upper and tf.strings.lower.

view details

Anirudh Sriram

commit sha 6f5b788a7f82e58b99ba269654857768f0993b5a

Update doc string for tf.split PiperOrigin-RevId: 288724439 Change-Id: I3cc3a26130bd0db11b01b2260772e6a1555d413d

view details

Robert David

commit sha 6bf044a64952c788ddde6e10c0a4367ecf1382f3

Backport IsZeroVector optimization on input data from Hybrid to fully float LSTMs. Reword comments to be consistent between the two versions. PiperOrigin-RevId: 288724457 Change-Id: I9ebb691d912cdf7a90295206fa38f62b63f0d137

view details

Lucy Fox

commit sha 79955a74f95aac969f008d82503e81b554c83410

Verify that input tensors are statically shaped for tf.FusedBatchNormV3 lowering. When lowering tf.FusedBatchNormV3 op to XLA HLO, input tensors must be statically shaped in the training case. Add a check for this and report match failure if the check is not satisfied. PiperOrigin-RevId: 288726150 Change-Id: I96ef8397efaf40d8d72b5251da65ec19393034d7

view details

Robert David

commit sha e5733c172d8d8ae95e1984929c85dc3bd8777b62

Remove unused package_group PiperOrigin-RevId: 288726366 Change-Id: Icffd95e04f7a66a605be81ceb7626c41f9942315

view details

push time in a month

pull request commenttensorflow/tensorflow

Fix to earlier fix for big-endian support in TFLite Micro

Conflicts resolved.

frreiss

comment created time in a month

push eventfrreiss/tensorflow-fred

frreiss

commit sha 60c6e0991e4cc9f35d25ae5a56f15c107d61eca3

Expand test coverage to quantization code

view details

frreiss

commit sha 891fa919eddcad126b10b8e6858636bc42dcce38

Fix buffer overrun

view details

Diederik van Liere

commit sha e42ee19f6d9126f0d96a360e3656fac5f71b79b9

Close filehandle to keras.json when initializing config

view details

frreiss

commit sha 35a8bcdeb6b96450cce7228c55cc74c0f8b5d4ea

Merge branch 'master' of https://github.com/tensorflow/tensorflow into issue-s390-lite-3

view details

Bas Aarts

commit sha dcf8ec6a9578af9a3ac1eae535248b77db120019

Replace --xla_gpu_disable_autotune option with --xla_gpu_autotune_level XLA performs autotuning for convolutions and GEMMs. for each algorithm it runs some checks to check for out of bounds access or functional errors. The latter in particular can take a lot of time, increasing compile time by orders of magnitude. This hurts end-to-end execution time. --xla_gpu_autotune_level enables clients to set a level. 0: don't autotune 1: autotune with uninitialized data 2: autotune with initialized data 3: autotune with initialized data, and reinitialize for inplace case 4: autotune with initialized data, reinitialize, and check The deafult is 4, not changing the current behaviour. Change some tests accordingly.

view details

Bas Aarts

commit sha 3bc238c04ac8d61ceb8437dd01ffa3278906e35f

address comments on commit 9be66fce9aa03bc4344422f59217b3bdf871311d

view details

Bas Aarts

commit sha 9cbf6a57d4281ef6a766423b102a2825eb5dedb4

Rebase against upstream

view details

Eugene Kuznetsov

commit sha ee1c2f112dc666a19e0d9e4ad679c0e948ebde8a

Fix for //tensorflow/python/eager/benchmarks/resnet50:*

view details

Yong Tang

commit sha b398e5bf6d216f76e8e5fd09ff96caf45c80e943

Fix tf.cast issue when python native data type is combined with dtype=tf.float64 This fix tries to address the issue raised in 35938 where the following issue exist: ``` >>> tf.cast(0.2, tf.float64) <tf.Tensor: id=37, shape=(), dtype=float64, numpy=0.20000000298023224> ``` The issue was that, in tf.cast, 0.2 was first converted to a float32 dtype tensor, then cast to float64 tensor. This fix adds dtype_hint=dtype to convert_to_tensor so that native dtype is preserved. This fix fixes 35938. Signed-off-by: Yong Tang <yong.tang.github@outlook.com>

view details

fsx950223

commit sha 2f5fd5185aa5e2d55332b256c47b602797633eb6

use a safer way for numpy

view details

fsx950223

commit sha c24ea21d5173366344a656706c545366b950ba6a

add test case

view details

fsx950223

commit sha a712ade3d00e429d4fa10191fff47dbadef2fa00

format code

view details

Yong Tang

commit sha 6ab77f3be330714746054eb678c9c4116f300692

Add complex number support for tf.extract_image_patches This PR tries to address the issue raised in 35955 where there was no complex number support for tf.extract_image_patches. The op `tf.extract_image_patches` itself could be used in many ways than just image so it makes sense to add complex support. This fix fixes 35955. Signed-off-by: Yong Tang <yong.tang.github@outlook.com> Register GPU types for tf.extract_image_patches Signed-off-by: Yong Tang <yong.tang.github@outlook.com> Fix build failure with GPU Signed-off-by: Yong Tang <yong.tang.github@outlook.com> Update array_ops.cc to only include TF_CALL_NUMBER_TYPES This fix updates array_ops.cc to only include TF_CALL_NUMBER_TYPES for extract_image_patches (realnumbertypes + complex64 + complex128). Signed-off-by: Yong Tang <yong.tang.github@outlook.com>

view details

Yong Tang

commit sha f3c49c5e45777a325c2580081b561ec4b2539f52

Add test case for complex number support for tf.extract_image_patches Signed-off-by: Yong Tang <yong.tang.github@outlook.com>

view details

fsx950223

commit sha afbc0af25eba438e992519258e0112de1692aa5c

format code

view details

IrinaM21

commit sha bc771472607aff17ea5d025dab248f56f617f1bc

Add a usage example for tf.is_tensor

view details

Rick Wierenga

commit sha a72b7e2653b70a610c717f67f4359b00dac602e5

Add usage example to pad_to_bounding_box

view details

Rick Wierenga

commit sha 5cf792e402b1e08e423a6588111a5210b71bbd54

remove redundant 0's

view details

Rick Wierenga

commit sha 60596fa2be2b3b537464d03686816c6049dc7ecd

remove empty lines

view details

Victor Peng

commit sha 0cc897c444a03f25e03e8ea8398ca3ee7f534030

set sync_needed_ as false after Sync()

view details

push time in a month

push eventfrreiss/tensorflow-fred

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

Yasuhiro Matsumoto

commit sha 929260fcffca3830cd6ec9632e8984338660187e

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

view details

push time in a month

pull request commenttensorflow/tensorflow

Fix buffer overrun in MicroAllocator

Conflicts resolved. It looks like someone has attempted to patch the buffer overrun, but the change in the patch is incorrect. The changes in this PR are still needed, particularly on big-endian machines.

frreiss

comment created time in a month

push eventfrreiss/tensorflow-fred

A. Unique TensorFlower

commit sha 85ea1e530ff8a7432822500db379b88fff7ffb32

Go: Update generated wrapper functions for TensorFlow ops. PiperOrigin-RevId: 288379318 Change-Id: I2f59d884d23ee443a90b4eb9bfa690e8da61e235

view details

TensorFlower Gardener

commit sha 6ed5c4b7f7fafaa492fbc99e003662864d2609a6

Merge pull request #34437 from xinan-jiang:fix/grappler-arithmetic-update-consumers PiperOrigin-RevId: 288380990 Change-Id: If564fdbf763820172c81e448e97b065c3cd92041

view details

Derek Murray

commit sha 0733a86c39f9f23a093cf06a27f5ea38984e2a5d

[tf.function] In graph mode, preserve nested attributes for func-valued attrs. This allows users to create a `tf.function` "f" that has attributes, using the private `function.defun_with_attributes()` method, set "f" as the attribute of another op (e.g. a MapDataset op), and preserve the attributes of the original "f". PiperOrigin-RevId: 288381945 Change-Id: I9dcf780251e610a4ed21a405372695b9b7671b69

view details

Robert David

commit sha f65b09f9aedcd33d0703cbf3d9845ea2869c0aa8

Cleanups in quantize OP: - Include the necessary headers. - Refactor dispatches between reference and optimized versions of AffineQuantize and Requantize to local functions. - Refactor error reporting to a single function. - Print a human-readable type name while reporting errors instead of the numeric enum value. - Remove unnecessary OpContext struct. - Use static_cast instead of reinterpret_cast to cast from void* to T* - Make sure used input/output type combination is supported in Prepare instead of just failing in Eval. - Use GetInput/GetOutput in Eval. - Use switch instead of if-else trees for the output type in Eval, to be consistent with handling of the input type. - Return with kTfLiteOk instead of break from switch when types are supported, to be consistent with the error cases. - Always call Register_QUANTIZE_OPT from Register_QUANTIZE, if NEON is not available the optimized version falls back to the reference version already. PiperOrigin-RevId: 288382080 Change-Id: I1b410438cefdd4b44b6a5c27d136830b976945b6

view details

Sachin Joglekar

commit sha 1f0c9730a3da392bbd4889aba89e68198f1bceaf

Add TfLiteContext API for delegates to preview partitioning plan PiperOrigin-RevId: 288384492 Change-Id: I51901b5ddfc7bb8ac04594f202717f483138bb90

view details

A. Unique TensorFlower

commit sha 7d72137ac31373739a6c01193a0b562a3cf2ce8d

Add support for multi-dimensional indices on string input for Gather. PiperOrigin-RevId: 288387122 Change-Id: Iccbbb4582d89374f656c75dc6911b0bb2f07fa27

view details

Maher Jendoubi

commit sha 6539d343231dfea317a7a09597ed178e32e2fd36

Contributing: Fix a typo

view details

Sachin Joglekar

commit sha 6ebd3bb334bd9e99eb34e4440dab5853fc84e869

Adds support for SpaceToDepth & DepthToSpace in hexagon delegate. PiperOrigin-RevId: 288393196 Change-Id: I8bf2ad0edd1b40230e88a10ce70dbe2449f172cc

view details

Maher Jendoubi

commit sha ffc474c4a30951a4023153acba78e4d263c636ad

Contributing: Fix a typo

view details

Maher Jendoubi

commit sha 0978e4bdf7a26bb2e7735b92a05584b9cec42bb3

betweeen --> between

view details

archis

commit sha b63f4f10f40c0fdf01b51e64e9585e1426366d50

Merge remote-tracking branch 'upstream/master' into dense_sparse_matmul

view details

Prakalp Srivastava

commit sha 1dd89d4a8fd050bc95e4e6717ded4210a51715b2

Lower tf.OutfeedEnqueueTuple op to XLA HLO. OutfeedEnqueueTuple is lowered to HLO tuple, after_all and outfeed ops. after_all op is emitted to generate XLA token required by outfeed op. PiperOrigin-RevId: 288399352 Change-Id: If56424b044e631f64837b39c8758dce9999ed4ab

view details

TensorFlower Gardener

commit sha f8eaa65cd90f472ab06f37f09ba9f36b98455b44

Merge pull request #35619 from Intel-tensorflow:cuixiaom_conv2d_pr PiperOrigin-RevId: 288407009 Change-Id: Ia1d49ba1e38928ef5278b722e762e2b331e6dc99

view details

TensorFlower Gardener

commit sha 5e26aa7bbfc1aa249925169b40dda2b4418c9ee5

Merge pull request #35510 from HotPotatoC:hotpotatoc-3 PiperOrigin-RevId: 288407477 Change-Id: Ib24ca94ebef6d964ee2ba5da0a98bd06709948f5

view details

A. Unique TensorFlower

commit sha adc3a1b1bec28849f62076e9b4be5c5963e5e5e7

fix ROCM build. PiperOrigin-RevId: 288409291 Change-Id: Iba69c37e531a50602818de14da526bab040941a0

view details

Jian Li

commit sha 9932a6f537442edb85e76db6a0b0f1d1cfdc6d7a

Update the version for full integer SVDF. PiperOrigin-RevId: 288410616 Change-Id: Ie22ef4a163ef68b27193d916554647c897a5db6e

view details

Brian Atkinson

commit sha 9c3fb76c03727d72fddc251be48696e3474d2994

Eliminate the exports_files directive from tensorflow/core. filegroups are a more natural way to achieve what this was doing while enabling visibility restrictions as well as proper target names. PiperOrigin-RevId: 288412177 Change-Id: I94dc2a0b4ff15868cf22868a472c42b6bb392720

view details

Ruoxin Sang

commit sha 5b74ee411ac8146bbb4b399f45b71c9e8ef0695e

Create shadow variables of EMA in init_scope, so it can be supported in TPUStrategy. PiperOrigin-RevId: 288414234 Change-Id: Ic9fe853a80e03bbd18051a0a913acffa824ce22f

view details

A. Unique TensorFlower

commit sha 3b74a63c0f7e1ec4618563958079f538cd9de076

[py_function] Don't attach py_function to the global eager graph. Eager mode can incorrectly have a global graph. Disabling global graph on eager mode breaks too many assumptions so first introduce a flag indicating it. Also, avoid attaching py_function to eager mode global graph, which is a leak. Though this CL doesn't fix the leak yet as there are two more references that leads to the leak, `tape_cache` and `ag_dnc_wrapper__` . #35084 PiperOrigin-RevId: 288415011 Change-Id: Ica53e29521320af22c10609857d0a0219a9596ce

view details

A. Unique TensorFlower

commit sha bd596d1a8de0de1b555c14fd742071131781adf0

Go: Update generated wrapper functions for TensorFlow ops. PiperOrigin-RevId: 288421716 Change-Id: Ie0eb9738ddf3cf323c1a693aa686810b31eabb7a

view details

push time in a month

pull request commenttensorflow/tensorflow

Initial implementation of SavedModel I/O for Graph Transform Tool

@HananKavitz I am not optimistic at this point. I'll leave it open for a few more weeks though.

frreiss

comment created time in a month

push eventfrreiss/tensorflow-fred

Derek Murray

commit sha cf8929973986d1eeac6acc10cb36b17075ca075e

[batch_util] Only call Tensor::RefCountIsOne() for DT_STRING and DT_VARIANT tensors. Currently, we test whether a Tensor element/slice can be moved for all batched/unbatched tensors, but only use that information for DT_STRING and DT_VARIANT tensors. Now that Tensor::RefCountIsOne() is a public method, we can push its use into the template specializations that handle strings and variants, which avoids unnecessary calls to this method for (e.g.) DT_FLOAT tensors. PiperOrigin-RevId: 287861271 Change-Id: I2eb023f61a383b47877a62faaca7d9dc162785c3

view details

Katherine Wu

commit sha ed05566b7ca9b1ecc31d30be7f11e4218efac244

(SavedModel loading change) Revive layers and models from config when available. When the config is not available or invalid, then load from SavedModel as a backup. Note that child objects (e.g. variables, sublayers) may be created as a result of a layer being revived from the config. These child objects may also be in the SavedModel object graph. This CL ensures that only one copy of each object is created. All of the current SavedModel saving/loading tests pass (with minor adjustments to one test). I've also added additional tests to ensure that structures with nested layers load correctly. Metrics loading will be addressed after this CL is submitted. Bug fixes that should be submitted before this CL: - cl/282123850 Ensure that nested layer losses are saved/reset properly - cl/282125214 Functional and Sequential models should export their config even if there's an error when serializing a layer PiperOrigin-RevId: 287864300 Change-Id: Ia015f3fe5bb2155fbd3563b5b9bbff0624e16797

view details

TensorFlower Gardener

commit sha 1e64f2f1f3ff881896c05938b4f62e279c42fe37

Merge pull request #35458 from duncanriach:fix-jit-scope-exmaples PiperOrigin-RevId: 287865184 Change-Id: Iea66b389eb3e64463eae30884fffc59584d71695

view details

arpan-dhatt

commit sha 4154a74cc6cf62dc1dcb99d03240a7203e9ca738

Added code examples tf.keras.backend.gather

view details

arpan-dhatt

commit sha 426d202f0407b86e2e40b55776220e3266378a1a

Merge pull request #1 from arpan-dhatt/arpan-dhatt-docs-code_example Added code examples tf.keras.backend.gather

view details

Rohan Jain

commit sha 96de19babd874ea6be7674520e0d8351ca89a749

Editing IteratorStateReader and IteratorStateWriter documentation to indicate their thread safeness. PiperOrigin-RevId: 287867779 Change-Id: Icab5c37e13fc730250264888dab42ff574477da3

view details

Sanjoy Das

commit sha 45a428b0875df5bf0257cb645ca99ce66ccb6b04

Fix a use-after-scope PiperOrigin-RevId: 287869388 Change-Id: Iee3268f1b6becff3bccada131039f920ada96f2e

view details

Reed Wanderman-Milne

commit sha d393d5d7528ff4b7a5ff602aadb2c144d7c7ac58

Fix various bugs in LossScaleGradientTape. The following bugs are fixed: 1. Crash when gradients were taken w.r.t. distributed variables. 2. Crash when targets or sources were not float32. 3. Crash when some gradients were None. 4. Bogus gradients returned when loss scale was 1. Several error messages were also greatly improved, such as when the jacobian is taken. Changed tests to take gradients w.r.t. variables instead of tensors, as this is the more common case. In the future, we may simply the code by having each replica have an independent loss scale. This means we do not have to synchronize the loss scales, which would greatly simplify the LossScaleGradientTape implementation, and probably improve performance as well. The downside is that not having a single loss scale may be more confusing to users. For now, I kept a single loss scale, as we have not yet decided if we want to go with this approach. PiperOrigin-RevId: 287869750 Change-Id: Ie60aa3f9721f78ab3fe857d4fdb0d1f70fa84faf

view details

Mihai Maruseac

commit sha 14c23a10766fe83b747aa61f09ba521f87a08cc9

Add Python3.7 testing on MacOS as we drop support for Python2. To be cherry-picked on `r1.15`, `r2.0`, and `r2.1` branches. PiperOrigin-RevId: 287871757 Change-Id: Ic530e884de421a39a82c686f1f0d086b6400d75c

view details

Jared Duke

commit sha e5d1073e007c2a992b234f80cbebf97e86f6a7cd

Throw explicit UnsatisfiedLinkError when TFLite native lib unavailable Note that this includes cases where the library itself has been loaded, but the native methods are still unavailable. PiperOrigin-RevId: 287873096 Change-Id: Iff2b370bf4568e9a214aeb21ab9927e1915fa31c

view details

Yuanzhong Xu

commit sha 95318343ccfef080ca789b4eacc5f1a0ee19b416

[MLIR:TF/XLA] Create no-op in empty island when converting to control dialect. When converted to control dialect during export, an empty island is replaced with all the control inputs in the nodes depending on it. This might create quadratic number of edges in certain cases (N control inputs and N nodes depending on it). So we create a no-op to group the input control edges. PiperOrigin-RevId: 287873185 Change-Id: Ic8219b319e864becac595e211f162bdc05ae773e

view details

Ayush Dubey

commit sha 4ab663e27c230a99e2e67f0609932d6151c54a32

Lower logging level in `BaseCollectiveExecutor::StartAbort`. Fixes #35314. PiperOrigin-RevId: 287873652 Change-Id: I58c856507c0c4429ae803c96da595db27db760ea

view details

Rohan Jain

commit sha 4e32c21337ac0d265f1318ecc61237e0c6f2c666

Splitting out parallel interleave iterator state into individual tensors. Also speeding up parallel interleave restore by parallelizing RestoreInput for the multiple iterators created. PiperOrigin-RevId: 287875941 Change-Id: Ib61854ad98aef9a28ab1df0131e40cedd32768d3

view details

Alexandre Passos

commit sha 08e24370367fb96d1bf050a39560cce488ea7f7e

Filters concrete functions which closure capture graph tensors from savedmodels. It doesn't make sense to serialize those functions as their own endpoints or to allow calling them directly from the loaded module but they should still be in the metagraph's function library as otherwise calling the root functions will crash. PiperOrigin-RevId: 287876746 Change-Id: I3401f65f7beb21cdda13cdf7bda9fc0f0cfa4156

view details

Haoyu Zhang

commit sha ca8d4f5aa68c928c840cad3b0ef9d246808c1ff1

Clean up unused copy of cluster device attributes. PiperOrigin-RevId: 287878441 Change-Id: I7e23abd1e9fbfca0fd96310706c4e902926e967f

view details

A. Unique TensorFlower

commit sha 3d7653c7fbe5063c6869436026bb0a0b93d9ddf7

Go: Update generated wrapper functions for TensorFlow ops. PiperOrigin-RevId: 287878988 Change-Id: Ia72158cf5cfed317927a14f5487e855536f58dd0

view details

Qwerty71

commit sha 8a29f1d9d7f63d99187d395dd0bafcab71450f79

Fixing "line too long error"

view details

Saleem Abdulrasool

commit sha 34ac7a96c85cdb16630e118ffaf7e845501238d6

build: pass `/D_USE_MATH_DEFINES` to C++ explicitly This is needed to get the VS2019 builds working. Without this, the math defines such as `M_PI` are unavailable to the C++ code. PiperOrigin-RevId: 287889374 Change-Id: I72ce934616ad30bdda63c6ac542858052154218a

view details

TensorFlower Gardener

commit sha 3018293e9c0beb61ed78f6f8597851c99242a527

Merge pull request #34803 from foss-for-synopsys-dwc-arc-processors:arcport_pullrequest PiperOrigin-RevId: 287895175 Change-Id: Ie2bb3d7f6d82fb246f4a71b239a95c58472ba6eb

view details

Revan Sopher

commit sha 71e4a02d24789b42cc5e098534a3c07c068a5263

Use regular assertEqual to fix flakiness in Py3.5. Prior to 3.6, dict ordering is nondeterministic. PiperOrigin-RevId: 287896554 Change-Id: I8f725f3879f821bcb7c8bf37bbb67c879ba3f69b

view details

push time in a month

pull request commenttensorflow/tensorflow

Exclude TFLite GPU kernel tests from non-GPU builds

Conflicts resolved.

frreiss

comment created time in a month

push eventfrreiss/tensorflow-fred

Deven Desai

commit sha 81ab633a4934c3e1f673e0abbfb229d7f3c1d029

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

view details

msteknoadam

commit sha 3a77a7cd8b3d8388f988ee3cedcfc0600bbe397f

Remove blank lines

view details

msteknoadam

commit sha 9ef76487cd515c2eeef983b958bb41382a315ce5

Change values to exact values

view details

msteknoadam

commit sha da382348bba9613ecc053daf88a4bd5a99edcbfd

Changed outputs to "..."

view details

msteknoadam

commit sha 1d792d394971054f73029cd2f935f7a0cc0884b5

Changed ... with exact values for non-random outputs

view details

msteknoadam

commit sha 5570703d1c3c201df6e64cd2ac107f1fbb2b052b

Added blank lines after the `Usage example` texts

view details

msteknoadam

commit sha 5f6f507d971bca97152d9ebccc7b2f53772cb011

Indentation fix

view details

Duncan Riach

commit sha 3bd6cba36a3f50023a5b3a585d7b63df4d9fef2f

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

view details

Duncan Riach

commit sha fc761aa9b1ec32500b762dfd699f254d97ce08c6

Use TF_CUDNN_USE_AUTOTUNE instead of TF_CUDNN_DETERMINISTIC

view details

Duncan Riach

commit sha c80114ad342ab42904bce3758ece7ab76c3db2e4

Use XLA_FLAGS='--xla_gpu_disable_autotune'

view details

HotPotatoC

commit sha f5339f98b20b4bf4622c34303897084a92fc2e6b

Added tf.strings.to_number() usage example

view details

Jacques Pienaar

commit sha 52f7fe62ddd334a09547756265c3537a82db7534

Merge pull request #35293 from HotPotatoC:hotpotatoc-2 PiperOrigin-RevId: 287554502 Change-Id: I3cc6195be5d828847f81ec1f4eae80b18db9d447

view details

Jian Li

commit sha 8582c7cfeeb34e5e5192377c4b5a56f36c42e6e7

Rename QuantizedLstmParameter to make it clear it's only used in integer LSTM, not hybrid LSTM. PiperOrigin-RevId: 287560836 Change-Id: I9df4d6ad43c6c0c3eab52fe526cba85e81e8725e

view details

Nick Kreeger

commit sha da6e7957c70e8934a7f13cebbbdafaf10c504c4c

Use the environment variable Xtensa core in micro unit tests instead of hard coding. PiperOrigin-RevId: 287561956 Change-Id: I0a365eb78cccf17430a195af2b83a063869855dd

view details

Nick Kreeger

commit sha 8a32eac856d1dbf82e6a73d16a274a61efc9ebba

Ensure that variable input tensors always have unique buffer addresses. When a model has several operators and tensors with similar types and dimensions - it is possible for variable tensors to have their buffer pointed at by other tensors. This fix ensures that the |last_used| value is not reset in micro_allocator.cc when dealing with variable tensors. A new unit test has been expanded to include a mock model that encounters this issue. PiperOrigin-RevId: 287562007 Change-Id: I13644f65f270ba60162b6b7ca71f8c43261296e0

view details

Haoyu Zhang

commit sha 3ee3ab9e44fdd46b928661dfdfc732e0736d976c

Lower function def debug string to VLOG(4) (consistent with other appearance). PiperOrigin-RevId: 287562801 Change-Id: Ia5b1fce29db2b6572e1f01c6b5ca751e1580fcb7

view details

Derek Murray

commit sha 4ce94f020ea470cd57a52c8471e6b87982b46a3b

Pass Config by const reference when checking its contents. This avoids reallocating and copying all of the vectors in the struct. PiperOrigin-RevId: 287565512 Change-Id: I2500a041ddf46a5c1a1547760d98e11c4cdf7a84

view details

Austin Anderson

commit sha 3ae602667e44c2642032b6886b2330d931f1cdfb

Fix the "devel" test wheel directory and error output The build-*.sh tests are failing when trying to import the (successfully!) built pip packages, because the wheel is now in "pip_pkg", not "pip", as the old tests worked. Also, the assembler didn't output stderr from the tests because of the wrong variable being used. PiperOrigin-RevId: 287567386 Change-Id: I6d6e5d55bafec9a443271b1640fffcaf596d2f8b

view details

Haoyu Zhang

commit sha 8f23523ff80d58f336c6e5502bf3e7c2937244e5

Hold a reference to ensure the eager client outlives the function data op. PiperOrigin-RevId: 287567727 Change-Id: I86f060e22ab3ef7421c34a53f40008f418b99fa0

view details

Nick Kreeger

commit sha 80c685a55f0d9de9cec178c55a8034cc109450e9

Add TF_LITE_STATIC_MEMORY to the Xtensa platform build Makefile. PiperOrigin-RevId: 287572726 Change-Id: I4d1b5f1500fb058ab1b92b6ed672fbc33b871a6c

view details

push time in a month

pull request commenttensorflow/tensorflow

Make a TFLite test case run on big-endian machines

Conflicts resolved.

frreiss

comment created time in a month

push eventfrreiss/bazel-zlinux-binaries

Fred Reiss

commit sha c96e8ebf658b17b3f5a4adfeaede778eea61c87d

Add compiled version of bazel 1.2.1 for ubuntu_18.04

view details

push time in a month

push eventfrreiss/tensorflow-fred

Chao Mei

commit sha 36aed819e3e59875a3333d67a9a7832b86bd07d0

Create a new profiling event for the op invoke of a TFLite delegate. PiperOrigin-RevId: 287012113 Change-Id: I5505f88fcd175b7ff52c39ca672196d052a55c71

view details

A. Unique TensorFlower

commit sha 8e8fabfee3bdc94516b9e26d466877e43e5baeb1

Go: Update generated wrapper functions for TensorFlow ops. PiperOrigin-RevId: 287012965 Change-Id: I6fcbfa846f8ce484123b05795acc930a4e16978d

view details

A. Unique TensorFlower

commit sha 8c900a570800bd13dbb0a83798933e3fb6272cac

Go: Update generated wrapper functions for TensorFlow ops. PiperOrigin-RevId: 287021400 Change-Id: Iaa0fc7a6a192d22ed67e61cec69396ad15cd806c

view details

Smit Hinsu

commit sha ac96bcb55f0a9197786e7e081a02b6ecd061ba08

Add auto-generated TensorFlow TensorScatterUpdate op PiperOrigin-RevId: 287025610 Change-Id: I577468dfe67bdf0bd88dfba41974cc336cf8bf02

view details

Smit Hinsu

commit sha 1b98eb80812274b2641a1b21cc4770d915aacfc0

Lower TensorFlow TensorScatterUpdate op to Scatter op in HLO Also, define a basic custom verifier for the TensorScatterUpdate op. PiperOrigin-RevId: 287027934 Change-Id: I9ce9aedeb7967aba593aa33f670a58c2bb9c3987

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

TensorFlower Gardener

commit sha 9d84621d9af3ca7027cfc5b6a37a7fb7cf713ec4

Merge pull request #35098 from WilliamHYZhang:usage_examples PiperOrigin-RevId: 287034905 Change-Id: Ifb977768218d2ca1c269e02f42174a6dc6e5a15c

view details

Mahmoud Abuzaina

commit sha f3246ad2704f1afb4b25b24abdd51e26553c48d6

Klocwork fix

view details

Mahmoud Abuzaina

commit sha 4e3b4cbc1ea9781c99d5f5eb1cdffefbdb6935c6

Avoid unnecessary data reorders

view details

Sanjoy Das

commit sha 497d9538e97ca00b045f59c8be93d2492b3e436c

VLOG results from MinSystemMemory Seems generally useful and will help triage https://github.com/tensorflow/tensorflow/issues/24496. PiperOrigin-RevId: 287042036 Change-Id: I64134707ecc12a5d5470a8e5181f3b69f4682894

view details

A. Unique TensorFlower

commit sha 4be89df785339567155e7e36517d79c224ad7efd

Go: Update generated wrapper functions for TensorFlow ops. PiperOrigin-RevId: 287043415 Change-Id: I2f56cf7bfd42fe757c99d8d91b5b5508418f8577

view details

A. Unique TensorFlower

commit sha 1ed8199b4f555c21645599a0c802201c1d603599

Go: Update generated wrapper functions for TensorFlow ops. PiperOrigin-RevId: 287049418 Change-Id: If0d9a32199cd2b37938378baccaaf1b2a07177e3

view details

Yash Katariya

commit sha 1acdc87f73549bf60f783034abf9fb83833414cd

Fix the formatting of doctests. PiperOrigin-RevId: 287050419 Change-Id: I49c90ca2cd41b4d6e269a8c0fe2d30ac458cccdc

view details

Brian Zhao

commit sha 9e484cdbf47f20802b14f9e3aca07393e5b89547

Automated g4 rollback of changelist 286834605. PiperOrigin-RevId: 287051006 Change-Id: Icff6438b784fb4ec8ff5f7d961d9a971b59b4c97

view details

A. Unique TensorFlower

commit sha 0215caffc08b9e5d47ef0eef05e81295325e0f36

Added convenient method for 3d image creation. PiperOrigin-RevId: 287061565 Change-Id: I52b36bdf8cd7bf76c1256fd48e00084c8abdc52d

view details

A. Unique TensorFlower

commit sha d460cc1a42c3685798093351ae4edca31dcf4b03

Utility information for image3d. PiperOrigin-RevId: 287064180 Change-Id: I92d5b42378867158624a16fdae68a2c4fbdd9dec

view details

push time in a month

push eventfrreiss/tensorflow-fred

Prakalp Srivastava

commit sha f4eb7f642d4aa50861ea3662f9cd93ba5628cf47

Add ReplicaId op to HLO dialect. PiperOrigin-RevId: 286649139 Change-Id: I0cdef755bee283aa7a4f6d484565da315e799b64

view details

Zhenyu Tan

commit sha 431e0a8bc76bd48cd9537b8b27a8785276727771

Make caching to be by default True. PiperOrigin-RevId: 286649524 Change-Id: I48a2a66fe3983c49cf42f15683b9f3a8ed84188d

view details

Prakalp Srivastava

commit sha b793db326919e05b17655bcaf7b1c01bb681cf22

NFC: Add missing new line character for readability of operator_writers.inc. PiperOrigin-RevId: 286650009 Change-Id: Ic9980b252826cd1f72cc940a02c474f42b3bdd07

view details

Zhenyu Tan

commit sha c18151b310e35e39ce49e18d1b6d137d8e0b7d52

DenseFeatures V1 to use partitioner PiperOrigin-RevId: 286650276 Change-Id: Ib25730d64c3b3462b8682adec0aba7b066f9935e

view details

8bitmp3

commit sha f5d995dcd0b055cf8b284842efc7693b15786027

Update CONTRIBUTING.md tensorflow/tensorflow

view details

Aart Bik

commit sha b787df8c377090492e4b4d409ac361dc74bc0472

[VectorOps] unify vector dialect "subscripts" PiperOrigin-RevId: 286650682 Change-Id: I377d1296bc25d34037d8f526dd5861a4fac57401

view details

Mihai Maruseac

commit sha 5ac1b9e24ff6afc465756edf845d2e9660bd34bf

Fix segfault when attempting to convert string to float16. To make sure this gets fixed, add test for converting string to any numeric type. PiperOrigin-RevId: 286650886 Change-Id: I81f770ec2bbd33a863e8057ce198c679912fa8e0

view details

Pooya Davoodi

commit sha 13cd9bcae3db5ec923ac43fedab7260ce25a0367

Use std::vector instead of C arrays for TestParams

view details

Feng Liu

commit sha 0bdc8d193891696e2906d009031de101b211045a

Add "fully_quantize" option support in the mlir_convert.py tests PiperOrigin-RevId: 286654920 Change-Id: I5d871b3551403cc99522119760bfd93cf5ee12c8

view details

Yash Katariya

commit sha 8ac21df38c20374fc40891449df3dba0a0302797

Remove TF1 only stuff from generate2.py PiperOrigin-RevId: 286656789 Change-Id: I1959fd8cbf368efc7c9647d0ca101704100ebd62

view details

Reed Wanderman-Milne

commit sha b32c69cb54595991058c9210ab4f08f6b70030ef

When mixed precision is used, log if device supports it. Currently only the mixed_float16 policy will cause a log to be outputted. In the future, mixed_bfloat16 will also output a log. Some sample logs: Mixed precision compatibility check (mixed_float16): OK Your GPU will likely run quickly with dtype policy mixed_float16 as it has compute capability of at least 7.0. Your GPU: TITAN V, compute capability 7.0 --- Mixed precision compatibility check (mixed_float16): WARNING Your GPU may run slowly with dtype policy mixed_float16 because it does not have compute capability of at least 7.0. Your GPU: Quadro M2000, compute capability 5.2 See https://developer.nvidia.com/cuda-gpus for a list of GPUs and their compute capabilities. If you will use compatible GPU(s) not attached to this host, e.g. by running a multi-worker model, you can ignore this warning. This message will only be logged once PiperOrigin-RevId: 286657719 Change-Id: I601a971800be1f531012f8d081868dfb73b133c0

view details

Yanan Cao

commit sha 46dbcd1cbce94a62d2453c14f677b136760a7a5b

Add tf.Log1P to ODS and its HLO legalization pattern PiperOrigin-RevId: 286658832 Change-Id: If5cd13c67305107b9d68b4f303b34256ca54b7fa

view details

Prakalp Srivastava

commit sha a2df12b97f6eab5fcf565407f1ac7fac2d617c81

Add RngNormal op to HLO dialect. Adds support to for RngNormal in exporter. PiperOrigin-RevId: 286658920 Change-Id: I5e3e363b5278e78d134b249eb41aaf5fd2de0bdc

view details

Frank Chen

commit sha 3e5b1ba4aa8538c78a1254f538c175d3d169c86a

Add stub for ExternalTpuDriver and associated APIs PiperOrigin-RevId: 286660802 Change-Id: I616e67fc4de551d485ddda0b2b497c9f8debde00

view details

A. Unique TensorFlower

commit sha 70765161c0f193b318fb79af9af634e0f37466a9

Go: Update generated wrapper functions for TensorFlow ops. PiperOrigin-RevId: 286661172 Change-Id: I855b82565fc4e50562e1eb134bcf9739f3466645

view details

Gaurav Jain

commit sha 0ef8624bc20b1d0225556de72ef4f451b29bfb12

Add simple 0/1 arithmetic optimizations Roll-forward without divide optimization PiperOrigin-RevId: 286661254 Change-Id: I1d458b70c49de39eb6d086f4e3f6978df698e860

view details

A. Unique TensorFlower

commit sha 734d2b5b36c0c4b3a4a737cf6bd355490fc478c8

Remove unused variables under `tensorflow/lite`. PiperOrigin-RevId: 286661272 Change-Id: I254852e088d26816f756270074545379cba31975

view details

Prakalp Srivastava

commit sha 3cab596f4f62e97c9b00b6dfbae4cc4b934b40c8

Add import support for RngNormal and RngUniform HLO dialect ops. PiperOrigin-RevId: 286662720 Change-Id: Icf48fdc0201da748a82df4181f5dab9844f93157

view details

Jose Baiocchi

commit sha d3039a157c8c7da85151b507b36decd0190af99a

Remove profiler_session dependence on core_cpu_lib PiperOrigin-RevId: 286665165 Change-Id: Iee01c2e6c9b624ae9507cead2d11c4bcc5bfb41a

view details

Juhyun Lee

commit sha 350808980cb7f994bfa5f8b96f58d2e75996cf18

Make a couple of targets in grappler buildable for Android. PiperOrigin-RevId: 286665297 Change-Id: If2baec2147f43a3a485978add8e22aa7549a433c

view details

push time in a month

pull request commenttensorflow/tensorflow

Add tests of LMDBDataset

@jsimsa it looks like LMDBDataset.Iterator leaks connections to the LMDB database. I've added some fixes to this branch.

I've also made some changes to conform to the new way of creating string tensors.

frreiss

comment created time in a month

pull request commenttensorflow/tensorflow

Fix to earlier fix for big-endian support in TFLite Micro

I've merged with the latest changes from master and rerun tests on Mac and zLinux.

frreiss

comment created time in 2 months

push eventfrreiss/tensorflow-fred

A. Unique TensorFlower

commit sha 3e72418d7c4eaba6edd914c1d757e3a9ca70964c

Define and pass resource argument aliasing information to MLIR functions. Add a field in FunctionDef proto to represent resource args aliasing. This change only contains preserving that field when converting to different representations, not including actually passing down from Python. PiperOrigin-RevId: 285475052 Change-Id: I4b061ae0d71b3a64ded9efb3a434f23f0176cde8

view details

Haoyu Zhang

commit sha 84b1a3498cd89ea95204a1955f0aa2363e560ad4

Remove `devices_` and `devices_map_` from EagerContext. We currently look up devices either through device managers or the devices_map_ field. To make the runtime fault tolerant, the list of devices (especially remote ones) must change dynamically, and accessing the inconsistent devices_map_ often leads to errors. PiperOrigin-RevId: 285475152 Change-Id: I37016875d94e481c4cb5143f19092fc46a481ff2

view details

Austin Anderson

commit sha 78ddcded185e60f83743faa194178a8e5979757b

Add msvcp140_1.dll to list of import-time-check Windows DLLs Resolves https://github.com/tensorflow/tensorflow/issues/35036 For TensorFlow 2.1.0rc1, the TensorFlow team built Windows packages with Microsoft Visual Studio 2019 16.4, upgraded from Visual Studio 2017. As discovered in the issue linked above, this caused an import error for Windows TF Python whls, because the build upgrade pulled in an additional Visual C++ DLL dependency, `msvcp140_1.dll`, which can be found in the latest Visual C++ package for all Visual Studio releases since 2015 (https://support.microsoft.com/en-us/help/2977003/the-latest-supported-visual-c-downloads). I discovered the missing DLL by unpacking the two wheels for rc0 and rc1 and separately running `dumpbin /DEPENDENTS tensorflow_core/python/_pywrap_tensorflow_internal.pyd` (thanks to @yifeif for help with this!). In this change, I've updated the import-time checker to look for both `msvcp140_1.dll` and `msvcp140.dll` in a way that supports simple future additions to the list. PiperOrigin-RevId: 285476568 Change-Id: Ia9727e50801a4ddad1ea30653a74478fb7aee4e8

view details

A. Unique TensorFlower

commit sha 83a23cabf02927a7d1e4f91780c67958f952caad

Updated tensorflow use of CreateMemSet API. PiperOrigin-RevId: 285476744 Change-Id: Icc0d3fbc83703a721c14e8486d725286bb00b824

view details

Robert David

commit sha a4754790f8edf685e71aa8f097af31435b05df83

Refactor getting scale from a tensor to a GetTensorState function, similar to GetTensorData. PiperOrigin-RevId: 285476826 Change-Id: Ib3fb7cfb36c1876acac7b177f112b9f9872fbfc9

view details

River Riddle

commit sha 4ad5238778e3f919e9fe899eea3137b1dab81d80

Refactor various canonicalization patterns as in-place folds. This is more efficient, and allows for these to fire in more situations: e.g. createOrFold, DialectConversion, etc. PiperOrigin-RevId: 285476837 Change-Id: I510ef7dde07df380bd81dbe59942175a871826e0

view details

Bruce Fontaine

commit sha 64022768cb2708b581a04e00a59b2f150976890f

Allow TPUStrategy to optionally prefetch its data to host memory rather than device memory. PiperOrigin-RevId: 285477600 Change-Id: I40ca96768800c15b8329fda368b2959c4c6bab95

view details

A. Unique TensorFlower

commit sha 2298f3fe368f5a6196bf67c456ed7d91fca14571

Add test for creating variable inside function passed to tf.vectorized_map. Switch pfor logic to use tf.function. PiperOrigin-RevId: 285478940 Change-Id: I7ef9e262f638bfac64e616744afe7b28cfcc8606

view details

Katherine Wu

commit sha 0e939c6fc766423aa958e021b914450665638c56

When creating op layers, always unnest inputs between layers when it is a single tensor. This is already done when restoring models from config. This ensures that node.input_tensors have the same structure after saving and loading. For example: ``` x = tf.keras.Input(()) model = tf.keras.Model(inputs=x, outputs=x * 2) model2 = model.from_config(model.get_config()) print(model.layers[1]._inbound_nodes[0].input_tensors, model2.layers[1]._inbound_nodes[0].input_tensors) ``` Prints "[Tensor], Tensor" prior to this change. Afterwards, the results are "Tensor, Tensor". This change resolves a bug when loading TensorflowOpLayers from SavedModel, which requires inputs to have the same structure. PiperOrigin-RevId: 285479275 Change-Id: I77a9856553fb8671f77c3741546b6a58c5f3b5b2

view details

George Karpenkov

commit sha 9ccd0efed187428c2b2f258d3bbe9491952e2d72

Fix experimental_compile=True for graph mode Previously the attribute only worked in eager mode, and was a no-op otherwise. Note that this also resolves the problem of a function with experimental_compile=True not being compiled when called from experimental_compile=False context. PiperOrigin-RevId: 285479735 Change-Id: I004cc476038dd8e670b01e235820ff6b0093cde5

view details

Allen Lavoie

commit sha 792efc53cb091fcb5229202290cc51505bcb9634

Fix FIFOQueue usage in tf.function Ensures that the queue resource is always created in the eager context and captured by any tf.functions, and that it is owned by the eager context regardless of where it was created. PiperOrigin-RevId: 285479831 Change-Id: Ife6d46aded4da09ed89306550f28fd71f5673966

view details

Mihai Maruseac

commit sha a0454c6828b015f6428a6ad2547fa5da563c5033

Install h5py when testing in virtualenv PiperOrigin-RevId: 285480262 Change-Id: Ifc1a37b5587513036eb90eb8ad085fd0fba56d08

view details

A. Unique TensorFlower

commit sha c1c7e025a4b3d33c2be5593d2f31be49eb65f5f2

Add missing dependencies to the tensorflow/core:android_tensorflow_lib_lite* rules. PiperOrigin-RevId: 285480480 Change-Id: Icbf169e13d5539b896e2067eb94f1bdbdc431fc1

view details

A. Unique TensorFlower

commit sha dd5da905d941a07ad33456c24221c6cea6ee9d85

Stop exporting tensorflow/core/framework:*.cc files and move framework targets from tensorflow/core to tensorflow/core/framework. PiperOrigin-RevId: 285481114 Change-Id: Ib2e0e918e3ceea5e9650a2740a8c5ad8825a2eb4

view details

A. Unique TensorFlower

commit sha 02465d559ac82c90214dff2bc58465d90f17d9e3

Tensor tracer: extend summary support to cpus. PiperOrigin-RevId: 285484908 Change-Id: Ie2f33be44be390f4aa46a9bce8ecaaa2aad71836

view details

Dero Gharibian

commit sha 8fa78c34d35a3be7e3f6e007152d5268004b65eb

Update test to use tensorflow::tstring. PiperOrigin-RevId: 285488252 Change-Id: I3ac341f95ca34cb341dbca9fcc920c5d0e4a288d

view details

Nat Jeffries

commit sha a1fba59cdab4bff7593bec126b115a6e8e80c67f

Fix cmsis-nn build error in pooling.cc PiperOrigin-RevId: 285492365 Change-Id: I0e99084f9788c2571ef03a27dbc1ec8b4d836aa6

view details

Yanhua Sun

commit sha 085803c4d64b0ed435006e9e4c4e6d790d160f78

Add an option in upgrade script to not output when error happens PiperOrigin-RevId: 285492823 Change-Id: I4317bb2e7146739e84a922b7a5921acdea835de4

view details

Derek Murray

commit sha 45cab6b1beed93b91ecb4cbdfcc2182a6ab52497

Fix potential ODR violations in sparse_tensor.h by marking inline method definitions `inline`. PiperOrigin-RevId: 285493440 Change-Id: I0b2852879dd6fe339792c72b722eeceae57b964f

view details

A. Unique TensorFlower

commit sha dd5ef6feebe4ec9fab228d51d5c4fb1dd3f94f1d

Convolution transposed specialization for 4x4 kernel with stride 2x2. Suitable for PowerVR and NVidia. PiperOrigin-RevId: 285496188 Change-Id: Iddba05400f3f18724f89ad3579f0a9ac2beb107e

view details

push time in 2 months

startedmodin-project/modin

started time in 3 months

pull request commenttensorflow/tensorflow

Fix to earlier fix for big-endian support in TFLite Micro

Update: Tests under //tensorflow/lite/micro/... still run as before on my big-endian test machine after the merge.

frreiss

comment created time in 3 months

push eventfrreiss/spark-fred

Liang-Chi Hsieh

commit sha 70987d8144f4f2c094f3b82d0c4a98e818366225

[SPARK-29680][SQL][FOLLOWUP] Replace qualifiedName with multipartIdentifier ### What changes were proposed in this pull request? Replace qualifiedName with multipartIdentifier in parser rules of DDL commands. ### Why are the changes needed? There are identifiers in some DDL rules we use `qualifiedName`. We should use `multipartIdentifier` because it can capture wrong identifiers such as `test-table`, `test-col`. ### Does this PR introduce any user-facing change? Yes. Wrong identifiers such as test-table, will be captured now after this change. ### How was this patch tested? Unit tests. Closes #26419 from viirya/SPARK-29680-followup2. Lead-authored-by: Liang-Chi Hsieh <viirya@gmail.com> Co-authored-by: Liang-Chi Hsieh <liangchi@uber.com> Signed-off-by: Liang-Chi Hsieh <liangchi@uber.com>

view details

Bago Amirbekian

commit sha 8152a87235a63a13969f7c1ff5ed038956e8ed76

[SPARK-28978][ ] Support > 256 args to python udf ### What changes were proposed in this pull request? On the worker we express lambda functions as strings and then eval them to create a "mapper" function. This make the code hard to read & limits the # of arguments a udf can support to 256 for python <= 3.6. This PR rewrites the mapper functions as nested functions instead of "lambda strings" and allows passing in more than 255 args. ### Why are the changes needed? The jira ticket associated with this issue describes how MLflow uses udfs to consume columns as features. This pattern isn't unique and a limit of 255 features is quite low. ### Does this PR introduce any user-facing change? Users can now pass more than 255 cols to a udf function. ### How was this patch tested? Added a unit test for passing in > 255 args to udf. Closes #26442 from MrBago/replace-lambdas-on-worker. Authored-by: Bago Amirbekian <bago@databricks.com> Signed-off-by: Xiangrui Meng <meng@databricks.com>

view details

xy_xin

commit sha 7cfd589868b8430bc79e28e4d547008b222781a5

[SPARK-28893][SQL] Support MERGE INTO in the parser and add the corresponding logical plan ### What changes were proposed in this pull request? This PR supports MERGE INTO in the parser and add the corresponding logical plan. The SQL syntax likes, ``` MERGE INTO [ds_catalog.][multi_part_namespaces.]target_table [AS target_alias] USING [ds_catalog.][multi_part_namespaces.]source_table | subquery [AS source_alias] ON <merge_condition> [ WHEN MATCHED [ AND <condition> ] THEN <matched_action> ] [ WHEN MATCHED [ AND <condition> ] THEN <matched_action> ] [ WHEN NOT MATCHED [ AND <condition> ] THEN <not_matched_action> ] ``` where ``` <matched_action> = DELETE | UPDATE SET * | UPDATE SET column1 = value1 [, column2 = value2 ...] <not_matched_action> = INSERT * | INSERT (column1 [, column2 ...]) VALUES (value1 [, value2 ...]) ``` ### Why are the changes needed? This is a start work for introduce `MERGE INTO` support for the builtin datasource, and the design work for the `MERGE INTO` support in DSV2. ### Does this PR introduce any user-facing change? No. ### How was this patch tested? New test cases. Closes #26167 from xianyinxin/SPARK-28893. Authored-by: xy_xin <xianyin.xxy@alibaba-inc.com> Signed-off-by: Wenchen Fan <wenchen@databricks.com>

view details

Xiao Li

commit sha 1e2d76e80a5c8fe09475904e87f01b19827da980

[HOT-FIX] Fix the SQLBase.g4 ### What changes were proposed in this pull request? Remove the duplicate code See the build failure: https://amplab.cs.berkeley.edu/jenkins/view/Spark%20QA%20Compile/job/spark-master-compile-maven-hadoop-3.2/986/ ### Why are the changes needed? Fix the compilation ### Does this PR introduce any user-facing change? No ### How was this patch tested? The existing tests Closes #26445 from gatorsmile/hotfixPraser. Authored-by: Xiao Li <gatorsmile@gmail.com> Signed-off-by: Xiao Li <gatorsmile@gmail.com>

view details

Jobit Mathew

commit sha 1e408d6fe639f6f928ca5de0b57ae05922d94576

[SPARK-29788][DOC] Fix the typos in the SQL reference documents ### What changes were proposed in this pull request? Fixing the typos in SQL reference document. ### Why are the changes needed? For user readability ### Does this PR introduce any user-facing change? No ### How was this patch tested? Tested manually. Closes #26424 from jobitmathew/typo. Authored-by: Jobit Mathew <jobit.mathew@huawei.com> Signed-off-by: Sean Owen <sean.owen@databricks.com>

view details

Gabor Somogyi

commit sha 12598e1b93cb1eaca8d7210791b868324931f0ff

[MINOR] FsHistoryProvider import cleanup ### What changes were proposed in this pull request? As it has been discussed in https://github.com/apache/spark/pull/26397#discussion_r343726691 `FsHistoryProvider` import section has to be cleaned up. ### Why are the changes needed? Unused imports. ### Does this PR introduce any user-facing change? No. ### How was this patch tested? Existing unit tests. Closes #26436 from gaborgsomogyi/SPARK-29755. Authored-by: Gabor Somogyi <gabor.g.somogyi@gmail.com> Signed-off-by: Sean Owen <sean.owen@databricks.com>

view details

Sean Owen

commit sha 4d9c36d5baa5d35b031be3b757a282e894a54820

[SPARK-29795][CORE] Explicitly clear registered metrics on MetricSystem shutdown ### What changes were proposed in this pull request? Explicitly clear registered metrics when `MetricsSystem` shuts down. ### Why are the changes needed? See https://issues.apache.org/jira/browse/SPARK-29795 for a complete explanation. The TL;DR is there is some evidence this could leak resources after Spark is shut down, and that may be a minor issue in Spark 3+ for apps or tests that re-start SparkContexts in the same JVM. ### Does this PR introduce any user-facing change? The possible difference here is that, after Spark is stopped, metrics are no longer available. It's unclear to me whether this is intended behavior anyway. ### How was this patch tested? See https://issues.apache.org/jira/browse/SPARK-29795 for more context: - Spark 3 already passes tests without this change - Spark 2.4 does too, as exists in branch-2.4 now - Spark 2.4 fails tests if metrics 4.x is used, without this change The last point is not directly relevant, as Spark 2.4 will not use metrics 4.x. It's evidence that it addresses some potential issue, however. Closes #26427 from srowen/SPARK-29795. Authored-by: Sean Owen <sean.owen@databricks.com> Signed-off-by: Sean Owen <sean.owen@databricks.com>

view details

Luca Canali

commit sha 2888009d6660183674de7da456b54247c1b8ca24

[SPARK-29654][CORE] Add configuration to allow disabling registration of static sources to the metrics system ### What changes were proposed in this pull request? The Spark metrics system produces many different metrics and not all of them are used at the same time. This proposes to introduce a configuration parameter to allow disabling the registration of metrics in the "static sources" category. ### Why are the changes needed? This allows to reduce the load and clutter on the sink, in the cases when the metrics in question are not needed. The metrics registerd as "static sources" are under the namespaces CodeGenerator and HiveExternalCatalog and can produce a significant amount of data, as they are registered for the driver and executors. ### Does this PR introduce any user-facing change? It introduces a new configuration parameter `spark.metrics.register.static.sources.enabled` ### How was this patch tested? Manually tested. ``` $ cat conf/metrics.properties *.sink.prometheusServlet.class=org.apache.spark.metrics.sink.PrometheusServlet *.sink.prometheusServlet.path=/metrics/prometheus master.sink.prometheusServlet.path=/metrics/master/prometheus applications.sink.prometheusServlet.path=/metrics/applications/prometheus $ bin/spark-shell $ curl -s http://localhost:4040/metrics/prometheus/ | grep Hive metrics_local_1573330115306_driver_HiveExternalCatalog_fileCacheHits_Count 0 metrics_local_1573330115306_driver_HiveExternalCatalog_filesDiscovered_Count 0 metrics_local_1573330115306_driver_HiveExternalCatalog_hiveClientCalls_Count 0 metrics_local_1573330115306_driver_HiveExternalCatalog_parallelListingJobCount_Count 0 metrics_local_1573330115306_driver_HiveExternalCatalog_partitionsFetched_Count 0 $ bin/spark-shell --conf spark.metrics.static.sources.enabled=false $ curl -s http://localhost:4040/metrics/prometheus/ | grep Hive ``` Closes #26320 from LucaCanali/addConfigRegisterStaticMetrics. Authored-by: Luca Canali <luca.canali@cern.ch> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>

view details

Huaxin Gao

commit sha 57b954e825970f004895ac127083da67e10c09fb

[SPARK-29730][SQL] ALTER VIEW QUERY should look up catalog/table like v2 commands Add AlterViewAsStatement and make ALTER VIEW ... QUERY go through the same catalog/table resolution framework of v2 commands. It's important to make all the commands have the same table resolution behavior, to avoid confusing end-users. e.g. ``` USE my_catalog DESC v // success and describe the view v from my_catalog ALTER VIEW v SELECT 1 // report view not found as there is no view v in the session catalog ``` Yes. When running ALTER VIEW ... QUERY, Spark fails the command if the current catalog is set to a v2 catalog, or the view name specified a v2 catalog. unit tests Closes #26453 from huaxingao/spark-29730. Authored-by: Huaxin Gao <huaxing@us.ibm.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>

view details

Maxim Gekk

commit sha 7ddcb5b46d927dafaeb5cb45327ac7303b0cd975

[SPARK-29819][SQL] Introduce an enum for interval units ### What changes were proposed in this pull request? In the PR, I propose an enumeration for interval units with the value `YEAR`, `MONTH`, `WEEK`, `DAY`, `HOUR`, `MINUTE`, `SECOND`, `MILLISECOND`, `MICROSECOND` and `NANOSECOND`. ### Why are the changes needed? - This should prevent typos in interval unit names - Stronger type checking of unit parameters. ### Does this PR introduce any user-facing change? No ### How was this patch tested? By existing test suites `ExpressionParserSuite` and `IntervalUtilsSuite` Closes #26455 from MaxGekk/interval-unit-enum. Authored-by: Maxim Gekk <max.gekk@gmail.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>

view details

Dongjoon Hyun

commit sha e25cfc4bb8f3d31828fbacbcba1899e608217336

[SPARK-29528][BUILD] Upgrade scala-maven-plugin to 4.3.0 for Scala 2.13.1 ### What changes were proposed in this pull request? This PR aims to upgrade `scala-maven-plugin` to `4.3.0` for Scala `2.13.1`. We tried 4.2.4, but it's reverted due to Windows build issue. Now, `4.3.0` has a Window fix. ### Why are the changes needed? Scala 2.13.1 seems to break the binary compatibility. We need to upgrade `scala-maven-plugin` to bring the the following fixes for the latest Scala 2.13.1. - https://github.com/davidB/scala-maven-plugin/issues/363 - https://github.com/sbt/zinc/issues/698 Also, `4.3.0` has the following Window fix. - https://github.com/davidB/scala-maven-plugin/issues/370 (4.2.4 throws error on Windows) ### Does this PR introduce any user-facing change? No. ### How was this patch tested? - For now, we don't support Scala-2.13. This PR at least needs to pass the existing Jenkins with Maven to get prepared for Scala-2.13. - `AppVeyor` passed. (https://ci.appveyor.com/project/ApacheSoftwareFoundation/spark/builds/28745383) Closes #26457 from dongjoon-hyun/SPARK-29528. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>

view details

Maxim Gekk

commit sha d4de01f567cf8d8a9ca880c7ed50eb572274e267

[SPARK-29408][SQL] Support `-` before `interval` in interval literals ### What changes were proposed in this pull request? - `SqlBase.g4` is modified to support a negative sign `-` in the interval type constructor from a string and in interval literals - Negate interval in `AstBuilder` if a sign presents. - Interval related SQL statements are moved from `inputs/datetime.sql` to new file `inputs/interval.sql` For example: ```sql spark-sql> select -interval '-1 month 1 day -1 second'; 1 months -1 days 1 seconds spark-sql> select -interval -1 month 1 day -1 second; 1 months -1 days 1 seconds ``` ### Why are the changes needed? For feature parity with PostgreSQL which supports that: ```sql # select -interval '-1 month 1 day -1 second'; ?column? ------------------------- 1 mon -1 days +00:00:01 (1 row) ``` ### Does this PR introduce any user-facing change? No ### How was this patch tested? - Added tests to `ExpressionParserSuite` - by `interval.sql` Closes #26438 from MaxGekk/negative-interval. Authored-by: Maxim Gekk <max.gekk@gmail.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>

view details

Dongjoon Hyun

commit sha 4b71ad6ffb049219ccb04b31aa8d75595831c662

[SPARK-29820][INFRA] Use GitHub Action Cache for `./.m2/repository/[com|org]` ### What changes were proposed in this pull request? This PR aims to enable [GitHub Action Cache on Maven local repository](https://github.com/actions/cache/blob/master/examples.md#java---maven) for the following goals. 1. To reduce the chance of failure due to the Maven download flakiness. 2. To speed up the build a little bit. Unfortunately, due to the GitHub Action Cache limitation, it seems that we cannot put all into a single cache. It's ignored like the following. - **.m2/repository is 680777194 bytes** ``` /bin/tar -cz -f /home/runner/work/_temp/01f162c3-0c78-4772-b3de-b619bb5d7721/cache.tgz -C /home/runner/.m2/repository . 3 ##[warning]Cache size of 680777194 bytes is over the 400MB limit, not saving cache. ``` ### Why are the changes needed? Not only for the speed up, but also for reducing the Maven download flakiness, we had better enable caching on local maven repository. The followings are the failure examples in these days. - https://github.com/apache/spark/runs/295869450 ``` [ERROR] Failed to execute goal on project spark-streaming-kafka-0-10_2.12: Could not resolve dependencies for project org.apache.spark:spark-streaming-kafka-0-10_2.12:jar:spark-367716: Could not transfer artifact com.fasterxml.jackson.datatype:jackson-datatype-jdk8:jar:2.10.0 from/to central (https://repo.maven.apache.org/maven2): Failed to transfer file https://repo.maven.apache.org/maven2/com/fasterxml/jackson/datatype/ jackson-datatype-jdk8/2.10.0/jackson-datatype-jdk8-2.10.0.jar with status code 503 -> [Help 1] ... [ERROR] mvn <args> -rf :spark-streaming-kafka-0-10_2.12 ``` ``` [ERROR] Failed to execute goal on project spark-tools_2.12: Could not resolve dependencies for project org.apache.spark:spark-tools_2.12:jar:3.0.0-SNAPSHOT: Failed to collect dependencies at org.clapper:classutil_2.12:jar:1.5.1: Failed to read artifact descriptor for org.clapper:classutil_2.12:jar:1.5.1: Could not transfer artifact org.clapper:classutil_2.12:pom:1.5.1 from/to central (https://repo.maven.apache.org/maven2): Connection timed out (Read failed) -> [Help 1] ``` ### Does this PR introduce any user-facing change? No. ### How was this patch tested? Manually check the GitHub Action log of this PR. ``` Cache restored from key: 1.8-hadoop-2.7-maven-com-5b4a9fb13c5f5ff78e65a20003a3810796e4d1fde5f24d397dfe6e5153960ce4 Cache restored from key: 1.8-hadoop-2.7-maven-org-5b4a9fb13c5f5ff78e65a20003a3810796e4d1fde5f24d397dfe6e5153960ce4 ``` Closes #26456 from dongjoon-hyun/SPARK-29820. Authored-by: Dongjoon Hyun <dhyun@apple.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>

view details

Pavithra Ramachandran

commit sha e2ca7f396fda165e544003b5a69378576e231b94

[SPARK-29601][WEBUI] JDBC ODBC Tab Statement column provide ellipsis for big SQL statement ### What changes were proposed in this pull request? Provide Ellipses in Statement column , just like description in Jobs page . ### Why are the changes needed? When a query is executed the whole query statement is displayed no matter how big it is. When bigger queries are executed, it covers a large portion of the page display, when we have multiple queries it is difficult to scroll down to view all. ### Does this PR introduce any user-facing change? No Before: ![Screenshot from 2019-11-01 23-15-23](https://user-images.githubusercontent.com/51401130/68064468-ebaa0300-fd41-11e9-8787-c5144c1468d4.png) After: ![Screenshot from 2019-11-02 07-07-21](https://user-images.githubusercontent.com/51401130/68064471-f19fe400-fd41-11e9-85c6-65f0faa64cc3.png) ### How was this patch tested? Manual Closes #26364 from PavithraRamachandran/ellipse_JDBC. Authored-by: Pavithra Ramachandran <pavi.rams@gmail.com> Signed-off-by: Sean Owen <sean.owen@databricks.com>

view details

Maxim Gekk

commit sha 18440151b0bf2d79356393f515b149439e17962e

[SPARK-29393][SQL] Add `make_interval` function ### What changes were proposed in this pull request? In the PR, I propose new expression `MakeInterval` and register it as the function `make_interval`. The function accepts the following parameters: - `years` - the number of years in the interval, positive or negative. The parameter is multiplied by 12, and added to interval's `months`. - `months` - the number of months in the interval, positive or negative. - `weeks` - the number of months in the interval, positive or negative. The parameter is multiplied by 7, and added to interval's `days`. - `hours`, `mins` - the number of hours and minutes. The parameters can be negative or positive. They are converted to microseconds and added to interval's `microseconds`. - `seconds` - the number of seconds with the fractional part in microseconds precision. It is converted to microseconds, and added to total interval's `microseconds` as `hours` and `minutes`. For example: ```sql spark-sql> select make_interval(2019, 11, 1, 1, 12, 30, 01.001001); 2019 years 11 months 8 days 12 hours 30 minutes 1.001001 seconds ``` ### Why are the changes needed? - To improve user experience with Spark SQL, and allow users making `INTERVAL` columns from other columns containing `years`, `months` ... `seconds`. Currently, users can make an `INTERVAL` column from other columns only by constructing a `STRING` column and cast it to `INTERVAL`. Have a look at the `IntervalBenchmark` as an example. - To maintain feature parity with PostgreSQL which provides such function: ```sql # SELECT make_interval(2019, 11); make_interval -------------------- 2019 years 11 mons ``` ### Does this PR introduce any user-facing change? No ### How was this patch tested? - By new tests for the `MakeInterval` expression to `IntervalExpressionsSuite` - By tests in `interval.sql` Closes #26446 from MaxGekk/make_interval. Authored-by: Maxim Gekk <max.gekk@gmail.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>

view details

lajin

commit sha 4de7131cffac39dca749d58b306264ad9ed9e428

[SPARK-29421][SQL] Supporting Create Table Like Using Provider ### What changes were proposed in this pull request? Hive support STORED AS new file format syntax: ```sql CREATE TABLE tbl(a int) STORED AS TEXTFILE; CREATE TABLE tbl2 LIKE tbl STORED AS PARQUET; ``` We add a similar syntax for Spark. Here we separate to two features: 1. specify a different table provider in CREATE TABLE LIKE 2. Hive compatibility In this PR, we address the first one: - [ ] Using `USING provider` to specify a different table provider in CREATE TABLE LIKE. - [ ] Using `STORED AS file_format` in CREATE TABLE LIKE to address Hive compatibility. ### Why are the changes needed? Use CREATE TABLE tb1 LIKE tb2 command to create an empty table tb1 based on the definition of table tb2. The most user case is to create tb1 with the same schema of tb2. But an inconvenient case here is this command also copies the FileFormat from tb2, it cannot change the input/output format and serde. Add the ability of changing file format is useful for some scenarios like upgrading a table from a low performance file format to a high performance one (parquet, orc). ### Does this PR introduce any user-facing change? Add a new syntax based on current CTL: ```sql CREATE TABLE tbl2 LIKE tbl [USING parquet]; ``` ### How was this patch tested? Modify some exist UTs. Closes #26097 from LantaoJin/SPARK-29421. Authored-by: lajin <lajin@ebay.com> Signed-off-by: Wenchen Fan <wenchen@databricks.com>

view details

Kent Yao

commit sha d06a9cc4bdcc1fb330f50daf219e9e8d908a16c4

[SPARK-29822][SQL] Fix cast error when there are white spaces between signs and values ### What changes were proposed in this pull request? With the latest string to literal optimization https://github.com/apache/spark/pull/26256, some interval strings can not be cast when there are some spaces between signs and unit values. After state `PARSE_SIGN`, it directly goes to `PARSE_UNIT_VALUE` when takes a space character as the end. So when there are some white spaces come before the real unit value, it fails to parse, we should add a new state like `TRIM_VALUE` to trim all these spaces. How to re-produce, which aim the revisions since https://github.com/apache/spark/pull/26256 is merged ```sql select cast(v as interval) from values ('+ 1 second') t(v); select cast(v as interval) from values ('- 1 second') t(v); ``` ### Why are the changes needed? bug fix ### Does this PR introduce any user-facing change? no ### How was this patch tested? 1. ut 2. new benchmark test Closes #26449 from yaooqinn/SPARK-29605. Authored-by: Kent Yao <yaooqinn@hotmail.com> Signed-off-by: Wenchen Fan <wenchen@databricks.com>

view details

Takeshi Yamamuro

commit sha cceb2d6f11c9ec71a375dea908a585549005b522

[SPARK-29825][SQL][TESTS] Add join-related configs in `inner-join.sql` and `postgreSQL/join.sql` ### What changes were proposed in this pull request? For better test coverage, this pr is to add join-related configs in `inner-join.sql` and `postgreSQL/join.sql`. These join related configs were just copied from ones in the other join-related tests in `SQLQueryTestSuite` (e.g., https://github.com/apache/spark/blob/master/sql/core/src/test/resources/sql-tests/inputs/natural-join.sql#L2-L4). ### Why are the changes needed? Better test coverage. ### Does this PR introduce any user-facing change? No. ### How was this patch tested? Existing tests. Closes #26459 from maropu/AddJoinConds. Authored-by: Takeshi Yamamuro <yamamuro@apache.org> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>

view details

zhengruifeng

commit sha 76e5294bb65f991988b73c8e6541b4b06d095127

[SPARK-29801][ML] ML models unify toString method ### What changes were proposed in this pull request? 1,ML models should extend toString method to expose basic information. Current some algs (GBT/RF/LoR) had done this, while others not yet. 2,add `val numFeatures` in `BisectingKMeansModel`/`GaussianMixtureModel`/`KMeansModel`/`AFTSurvivalRegressionModel`/`IsotonicRegressionModel` ### Why are the changes needed? ML models should extend toString method to expose basic information. ### Does this PR introduce any user-facing change? yes ### How was this patch tested? existing testsuites Closes #26439 from zhengruifeng/models_toString. Authored-by: zhengruifeng <ruifengz@foxmail.com> Signed-off-by: Dongjoon Hyun <dhyun@apple.com>

view details

DB Tsai

commit sha a6a27485851a160b9fc4bb86481c60f10573aad0

[SPARK-29805][SQL] Enable nested schema pruning and nested pruning on expressions by default ### What changes were proposed in this pull request? Enable nested schema pruning and nested pruning on expressions by default. We have been using those features in production in Apple for couple months with great success. For some jobs, we reduce the data reading by more than 8x and 21x faster in wall clock time. ### Why are the changes needed? Better performance. ### Does this PR introduce any user-facing change? No. ### How was this patch tested? Existing tests. Closes #26443 from dbtsai/enableNestedSchemaPrunning. Authored-by: DB Tsai <d_tsai@apple.com> Signed-off-by: DB Tsai <d_tsai@apple.com>

view details

push time in 3 months

push eventfrreiss/tensorflow-fred

namrata-ibm

commit sha a816b97a61ad152b64f7550351379ccae860d841

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

view details

Ruoxin Sang

commit sha 3e31f6191ce11483ac8d52f2c471ef7bc1f9175a

Throw an explicit error if user call TPUStrategy experimental_run_v2 in eager mode with a python function. PiperOrigin-RevId: 284256438 Change-Id: I6c0ea86a4cdf84712a1940d44dae8d4044f0fdee

view details

Francois Chollet

commit sha d58e8594e146636605e2669c646bd0bfcd454e0e

Move image preprocessing layers to `layers/preprocessing` PiperOrigin-RevId: 284257328 Change-Id: I1118722d3084c456f982a7baaa8b65c97dae3292

view details

A. Unique TensorFlower

commit sha 46e7fca1b606bb2a115ec2ee4f751ad2d26834cd

Adds a `directory` accessor to CheckpointManager. PiperOrigin-RevId: 284258465 Change-Id: Ibf2627854e0e89816e8aced66c2ca586682a8724

view details

Peng Wang

commit sha 25c58b201f27b30cc2e3247d8d6abe76774a010d

Fixes docstring for tf.math.logical_not PiperOrigin-RevId: 284260332 Change-Id: I2c04fa469773e4b6400c9391d931f167833161a0

view details

A. Unique TensorFlower

commit sha e7b00a13d4174953cb4f1324bfa1f3165245e462

Use named traits in the ODS definition of LLVMFuncOp The "FunctionLike" and "IsIsolatedFromAbove" op traits are now defined as named records in base ODS file. Use those instead of NativeOpTrait referring to the C++ class name in the ODS definition of LLVMFuncOp. NFC. PiperOrigin-RevId: 284260891 Change-Id: I5b611d726101747c46355ba9ba72c5a0f76eee3e

view details

A. Unique TensorFlower

commit sha d14373ec7949a501494114d225fdc6fe9560f36f

Override EIGEN strong inline for release builds as well. PiperOrigin-RevId: 284261705 Change-Id: I882c786169fb2d51716c884c9b1c91b59ae2df4e

view details

A. Unique TensorFlower

commit sha 3cc5bda704f8088b342fd126fc138348a0671e67

Move GPU::LaunchOp to ODS. NFC. Move the definition of the GPU launch opreation from hand-rolled C++ code to ODS framework. This only does the moves, a follow-up is necessary to clean up users of custom functions that could be auto-generated by ODS. PiperOrigin-RevId: 284261856 Change-Id: I3bc93035fc9364c2992ea90129001697fca70cf0

view details

Andy Ly

commit sha fa7ff04d632bcd0d3299383d93292ade485daea6

Add TPUReplicatedInput and TPUReplicatedOutput to MLIR Op Definition Spec. Updated associated descriptions for TPUReplicatedInput and TPUReplicatedOutput. PiperOrigin-RevId: 284261948 Change-Id: I16197e103128c7c1153f70367118b129cae6789d

view details

A. Unique TensorFlower

commit sha 688641f73ffac11636171a88d7931ef84c549b20

During serialization do a walk of ops in module to find spv.module. During lowering, spv.module might be within other modules (for example gpu kernel module). Walk the module op to find spirv module to serialize. PiperOrigin-RevId: 284262550 Change-Id: I1cf465ae0637e01766b09e89ab5421952f373b0a

view details

A. Unique TensorFlower

commit sha 3b6cf4f210a70217adc9913621512d0444da76d5

Replace custom getBody method with an ODS-generated in gpu::LaunchOp PiperOrigin-RevId: 284262981 Change-Id: I3d38c8aa1bde2bffbae06a8ec2847bf6e37f87fd

view details

Ken Franko

commit sha fa8d02282175bd0b7caf23dd1029e5389a44fdc4

Group variable initialization when calling lift_to_graph. When initializing variables defined inside a @tf.function which are lifted to the outer graph, group the variables together and call lift_to_graph once. lift_to_graph supports passing in multiple tensors and the graph to lift to is the same for all of the variable initialization. This improves setup time. PiperOrigin-RevId: 284263511 Change-Id: I4cfcdb0394198df8f890a98295cc2fcb77b75413

view details

Jacques Pienaar

commit sha 76c93e5e1bf04ee970d7fc73b23319ca0501fe31

Change inferReturnTypes to return LogicalResult and values Previously the error case was using a sentinel in the error case which was bad. Also make the one `build` invoke the other `build` to reuse verification there. And follow up on suggestion to use formatv which I missed during previous review. PiperOrigin-RevId: 284265762 Change-Id: I068a97a77845e16856f72c07ed55ca9eb0dffe73

view details

Rohan Jain

commit sha 544153f64555432e93c57a595188ab9bdd810241

Setting _first_run to True for the CheckpointInputPipelineHook if a new session is created. This makes sure that if a single task gets pre-empted in the middle of training, we restore the input pipeline because in that case we don't run the coordinator code all over again but just reset the session. PiperOrigin-RevId: 284266159 Change-Id: Id2b7fea2e1df13b9816c1db59b1defb8392cc244

view details

Karmel Allison

commit sha 53c5622f2dd6b88dfaaf45f7aa095ae322d26828

Add docstrings for get/set optimizer weights. PiperOrigin-RevId: 284266397 Change-Id: I3f5485d16ccb1d4b87508a78760a8b6ba5302cf1

view details

Thomas O'Malley

commit sha 20a3cd024347e67458cac6db35bba2c133ffe8af

Add `LossesContainer` class to encapsulate compiled loss computation in v2. PiperOrigin-RevId: 284268134 Change-Id: I5d4fc0ae7e26d45fb11736bd48f5e2238550d1b9

view details

A. Unique TensorFlower

commit sha 90c86cccf4afa2f586c9268a36d64eddd4afce19

Changed necessary attributes for PBuffer context. PiperOrigin-RevId: 284269021 Change-Id: I4e1add1775709e475ebe5a0cd45ee59166f1230e

view details

Brian Atkinson

commit sha 1d96a6103c4590c6c6a36cc10eb5ded772afa22b

Internal change PiperOrigin-RevId: 284269772 Change-Id: I6ea267405e1d0225ad61874b1e8297432b3067c4

view details

Peter Ma

commit sha bd045624683452a3100a98cf1c061a787cf444bb

Added a guard to prevent from crashing for certain cases. PiperOrigin-RevId: 284272480 Change-Id: Idb6e566037d83406921e5d318aaa4d35a336edfa

view details

Shanqing Cai

commit sha ad41522e548b036f479a1a2e342cd7b40b8173b8

[tfdbg] Make CONCISE_HEALTH kernel GPU code build on Windows PiperOrigin-RevId: 284273022 Change-Id: I6b21c6969347f3bc6186a0500bae05e160e1d2f6

view details

push time in 3 months

pull request commenttensorflow/tensorflow

Fix buffer overrun in MicroAllocator

@gbaned the conflicts seem to have resolved themselves.

frreiss

comment created time in 3 months

push eventfrreiss/tensorflow-fred

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

namrata-ibm

commit sha a816b97a61ad152b64f7550351379ccae860d841

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

view details

amoitra

commit sha b64e97b4448b86bafc8dc72d6a5db64e15dfe30e

Always expand batchnorm inference

view details

amoitra

commit sha 400bba1be9dfb1518b7a0748041d279532b7f7c0

Add comments

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

pull request commenttensorflow/tensorflow

Fix to earlier fix for big-endian support in TFLite Micro

@frreiss Could you please resolve the conflicts? Thanks!

Hi @gbaned , the conflicts seem to have resolved themselves. I've merged the latest changes from master with this branch and am rerunning the tests on my zLinux test machine.

frreiss

comment created time in 3 months

push eventfrreiss/tensorflow-fred

amoitra

commit sha e07dfe6a7091db80bef5c94d10eaf6360d7224de

Set xla_gpu_use_cudnn_batchnorm to true

view details

namrata-ibm

commit sha a816b97a61ad152b64f7550351379ccae860d841

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

view details

amoitra

commit sha b64e97b4448b86bafc8dc72d6a5db64e15dfe30e

Always expand batchnorm inference

view details

amoitra

commit sha 400bba1be9dfb1518b7a0748041d279532b7f7c0

Add comments

view details

Agoniii

commit sha 6a74e16e94f05375b5b220cc046e3e0b1d5d2055

add label for xlaop

view details

Johan Euphrosine

commit sha 75057371b960dc9e468ea1977031c37e729aeeba

lite/microfrontend: fix FilterbankState unsigned type missmatch FilterbankState work is uint64_t*, casting a signed type prevent the libraries to compile w/ the esp32 arduino core toolchain.

view details

Agoniii

commit sha 90bf975aeccb36c9a816849e3d696730bf50088c

using std::string

view details

Lukas Geiger

commit sha 147de48ad973a6a05e8113af815988014652caf2

Return new instance of AutoCastVariable after assignment Assignments and sparse updates will now return a new instance of `AutoCastVariable` wrapping the [`_UnreadVariable`](https://github.com/tensorflow/tensorflow/blob/2692ea8ec1953e42952597adb5b5099181a679b2/tensorflow/python/ops/resource_variable_ops.py#L1806) returned from the assignment op.

view details

Fei Hu

commit sha be36fd93ef817281115eac37c37a800e8b182001

switch BucketBySequenceLengthTest to use combinations

view details

Fei Hu

commit sha c5c6f9686f1be93a54d344e5f2f864f1855e1bed

Switch CopyToDeviceTest to use TF combination

view details

Fei Hu

commit sha 7fc9eb7bfee7dd62000fd39c44a566444503a93a

Switch CounterTest to use TF combinations

view details

Fei Hu

commit sha 1246058569a00f9bc2d479eeb94dc5f3e4c708c4

Switch CsvDatasetTest to use TF combinations

view details

Fei Hu

commit sha b9350874eb8bbf4f0576234b1ecdc997d65bea1c

Switch DenseToSparseBatchTest to use TF combinations

view details

Fei Hu

commit sha 12a6cc569963354dfa4a5d10af291a9e6fcc3b06

Switch DirectedInterleaveDatasetTest to use TF combinations

view details

Fei Hu

commit sha b7be69fcce909f9aef82dc975025d14eb374aa45

Switch GetSingleElementTest to use TF combinations

view details

Fei Hu

commit sha b48761c689d575c92995e15dd78990cb291badf3

Switch GroupByReducerTest to use TF combinations

view details

Fei Hu

commit sha 4be4b406a7507f8ab3b0c6643399f1383b6f3b7d

Switch GroupByWindowTest to use TF combinations

view details

Fei Hu

commit sha 27a2f9f39f8a43fb203107ac0c297d0fe7ef4b22

Switch IgnoreErrorsTest to use TF combinations

view details

Fei Hu

commit sha e0a77c884f83772c80334dbb152773a377d3049c

Switch MakeBatchedFeaturesDatasetTest to use TF combinations

view details

Fei Hu

commit sha 0e445ffe0c5eebec53bca5a0893a850288fb5f88

Switch MakeCsvDatasetTest to use TF combinations

view details

push time in 3 months

PR opened tensorflow/tensorflow

Disable tests tagged 'v1only' on v2 builds

If one runs regression tests for the TensorFlow V2 APIs via:

bazel test --config=v2 //tensorflow/...

the build will run several tests that are only supposed to work with the V1 APIs. These tests are tagged v1only, but the default build configuration does not disable them when the v2 build config is activated.

This pull modifies the Bazel configuration so that test targets that have the tag v1only are not run by default when the v2 build configuration is active.

+9 -2

0 comment

1 changed file

pr created time in 3 months

push eventfrreiss/tensorflow-fred

frreiss

commit sha 3adf41707a2feafd66b4677e0d41f6b6c925f140

Disable tests tagged 'v1only' on v2 builds Disable tests tagged 'v1only' on v2 builds Add missing '-'

view details

push time in 3 months

push eventfrreiss/tensorflow-fred

frreiss

commit sha 44759dc9eb7219e27c2d0d036dd0bf932f043f55

Add missing '-'

view details

push time in 3 months

PR opened tensorflow/tensorflow

Exclude TFLite GPU kernel tests from non-GPU builds

Typing

bazel test //tensorflow/lite/...

on a system without a GPU results in multiple test failures as the build attempts to compile and test several types of GPU kernels.

This pull request adds the "gpu" tag to the TFLite tests that require a GPU to run.

+106 -77

0 comment

7 changed files

pr created time in 3 months

create barnchfrreiss/tensorflow-fred

branch : issue-v1only-disable

created branch time in 3 months

push eventfrreiss/tensorflow-fred

amoitra

commit sha e07dfe6a7091db80bef5c94d10eaf6360d7224de

Set xla_gpu_use_cudnn_batchnorm to true

view details

amoitra

commit sha b64e97b4448b86bafc8dc72d6a5db64e15dfe30e

Always expand batchnorm inference

view details

amoitra

commit sha 400bba1be9dfb1518b7a0748041d279532b7f7c0

Add comments

view details

Agoniii

commit sha 6a74e16e94f05375b5b220cc046e3e0b1d5d2055

add label for xlaop

view details

Johan Euphrosine

commit sha 75057371b960dc9e468ea1977031c37e729aeeba

lite/microfrontend: fix FilterbankState unsigned type missmatch FilterbankState work is uint64_t*, casting a signed type prevent the libraries to compile w/ the esp32 arduino core toolchain.

view details

Agoniii

commit sha 90bf975aeccb36c9a816849e3d696730bf50088c

using std::string

view details

Lukas Geiger

commit sha 147de48ad973a6a05e8113af815988014652caf2

Return new instance of AutoCastVariable after assignment Assignments and sparse updates will now return a new instance of `AutoCastVariable` wrapping the [`_UnreadVariable`](https://github.com/tensorflow/tensorflow/blob/2692ea8ec1953e42952597adb5b5099181a679b2/tensorflow/python/ops/resource_variable_ops.py#L1806) returned from the assignment op.

view details

Fei Hu

commit sha be36fd93ef817281115eac37c37a800e8b182001

switch BucketBySequenceLengthTest to use combinations

view details

Fei Hu

commit sha c5c6f9686f1be93a54d344e5f2f864f1855e1bed

Switch CopyToDeviceTest to use TF combination

view details

Fei Hu

commit sha 7fc9eb7bfee7dd62000fd39c44a566444503a93a

Switch CounterTest to use TF combinations

view details

Fei Hu

commit sha 1246058569a00f9bc2d479eeb94dc5f3e4c708c4

Switch CsvDatasetTest to use TF combinations

view details

Fei Hu

commit sha b9350874eb8bbf4f0576234b1ecdc997d65bea1c

Switch DenseToSparseBatchTest to use TF combinations

view details

Fei Hu

commit sha 12a6cc569963354dfa4a5d10af291a9e6fcc3b06

Switch DirectedInterleaveDatasetTest to use TF combinations

view details

Fei Hu

commit sha b7be69fcce909f9aef82dc975025d14eb374aa45

Switch GetSingleElementTest to use TF combinations

view details

Fei Hu

commit sha b48761c689d575c92995e15dd78990cb291badf3

Switch GroupByReducerTest to use TF combinations

view details

Fei Hu

commit sha 4be4b406a7507f8ab3b0c6643399f1383b6f3b7d

Switch GroupByWindowTest to use TF combinations

view details

Fei Hu

commit sha 27a2f9f39f8a43fb203107ac0c297d0fe7ef4b22

Switch IgnoreErrorsTest to use TF combinations

view details

Fei Hu

commit sha e0a77c884f83772c80334dbb152773a377d3049c

Switch MakeBatchedFeaturesDatasetTest to use TF combinations

view details

Fei Hu

commit sha 0e445ffe0c5eebec53bca5a0893a850288fb5f88

Switch MakeCsvDatasetTest to use TF combinations

view details

Fei Hu

commit sha c0e7479f2d2ff61f87d4839f79b03e6140f3fd2c

Switch MakeTFRecordDatasetTest to use TF combinations

view details

push time in 3 months

Pull request review commenttensorflow/tensorflow

Improve API documentation for WindowDataset op

 def window(self, size, shift=None, stride=1, drop_remainder=False):     {'a': [1, 2]}     {'a': [3, 4]} +    If this dataset returns elements in a deterministic order, then the

Removed this paragraph in commit 84bc9b4

frreiss

comment created time in 3 months

Pull request review commenttensorflow/tensorflow

Improve API documentation for WindowDataset op

 op {   in_arg {     name: "size"     description: <<END-A scalar representing the number of elements to accumulate in a window.+A `tf.int64` scalar `tf.Tensor`, representing the number of elements+of the input dataset to combine into a window. Must be positive. END   }   in_arg {     name: "shift"     description: <<END-A scalar representing the steps moving the sliding window forward in one-iteration. It must be positive.+A `tf.int64` scalar `tf.Tensor`, representing the number of input elements+by which the window moves in each iteration.  Defaults to `size`.+Must be positive. END   }   in_arg {     name: "stride"     description: <<END-A scalar representing the stride of the input elements of the sliding window.-It must be positive.+A `tf.int64` scalar `tf.Tensor`, representing the stride of the input elements+in the sliding window. Must be positive. The default value of 1 means+"retain every input element". END   }   in_arg {     name: "drop_remainder"     description: <<END-A scalar representing whether a window should be dropped in case its size is-smaller than desired.+A `tf.bool` scalar `tf.Tensor`, representing whether the last window should be+dropped if its size is smaller than `window_size`. END   }-  summary: "A dataset that creates window datasets from the input dataset."+  summary: <<END+  C++ implementation of the Python API `tf.data.Dataset.window()`.

Removed this line in commit 84bc9b4

frreiss

comment created time in 3 months

Pull request review commenttensorflow/tensorflow

Improve API documentation for WindowDataset op

 def window(self, size, shift=None, stride=1, drop_remainder=False):      A "window" is a finite dataset of flat elements of size `size` (or possibly     fewer if there are not enough input elements to fill the window and-    `drop_remainder` evaluates to false).+    `drop_remainder` evaluates to `False`).++    The `shift` argument determines the number of input elements by which+    the window moves on each iteration.  The first element in the `k`th window+    will be element+    ```

Added newlines here in commit 84bc9b4. Also modified the .pbtxt file.

frreiss

comment created time in 3 months

Pull request review commenttensorflow/tensorflow

Improve API documentation for WindowDataset op

 op {   in_arg {     name: "size"     description: <<END-A scalar representing the number of elements to accumulate in a window.+A `tf.int64` scalar `tf.Tensor`, representing the number of elements

Changed these type names in commit 84bc9b4

frreiss

comment created time in 3 months

push eventfrreiss/tensorflow-fred

Albin Joy

commit sha 2cb0e459c4f15fcf37c1e3c8772fc48db298996f

Refactoring comparisons invoke calls

view details

Albin Joy

commit sha 4321f790b0c5773f8c850f4effb54b12e153efa2

Removed macro usage

view details

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

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

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

push time in 3 months

create barnchfrreiss/tensorflow-fred

branch : issue-s390-lite-5

created branch time in 3 months

PR opened tensorflow/tensorflow

Reviewers
Fix to earlier fix for big-endian support in TFLite Micro

https://github.com/tensorflow/tensorflow/pull/30362 modified TFLite Micro to work on big-endian processors. One of the code modifications in that pull request added byte-swapping to the accessor functions in tensorflow/lite/kernels/kernel_util.h. This byte swapping corrected for little-endian data in the inputs and outputs fields of TfLiteNode structs. This little-endian data was there because the MicroAllocator class initialized these fields with pointers to flatbuffers::Vector<int32_t> that were cast to TfLiteIntArray*. These two struct types happen to have the same format in memory on little-endian machines.

Although these changes to kernel_util.h got TFLite Micro to work on big-endian machines at the time, there are two significant problems as a result of these changes:

  • The non-microcontroller version of TFLite is now broken on big-endian machines, because that version of TFLite keeps the inputs and outputs fields in host byte order.
  • Subsequent updates to kernel_util.h have added additional accessors that do not have the byte-swapping changes. As a result, TFLite Micro is once again broken on big-endian machines.

The upshot of this is that the versions of both TensorFlow Lite and TFLite Micro at the head of the master branch crash with segmentation faults if the user attempts to load a model on a machine with a big-endian architecture.

This pull request replaces the changes to kernel_util.h with an alternative that should be more maintainable. I have modified the model-loading code in MicroAllocator::AllocateNodeAndRegistrations() so that it copies and byte-swaps the inputs and outputs fields of nodes when loading graphs on big-endian machines. These arrays are small (typically 1-2 entries per node), so the additional memory allocated is minimal. After this change, the data in the inputs and outputs fields is always in host byte order. So I have removed the calls to flatbuffers::EndianScalar() from the accessors in kernel_util.h.

After these changes, all but 2 of the regression tests under //tensorflow/lite/experimental/micro succeed on my big-endian test machine. These two tests fail because another change in https://github.com/tensorflow/tensorflow/pull/30362 is also broken. I will fix this other problem in a separate pull request.

FYI @PeteBlackerThe3rd

+48 -18

0 comment

3 changed files

pr created time in 3 months

PR opened tensorflow/tensorflow

Reviewers
Fix buffer overrun in MicroAllocator

The quantization code in the method MicroAllocator::InitializeRuntimeTensor writes past the end of the params.zero_point field in TfLiteTensor structs, clobbering the allocation_type field. This pull request fixes the overrun. I have also extended the test coverage for MicroAllocator to cover the lines of code where the overrun was occurring.

+59 -7

0 comment

4 changed files

pr created time in 3 months

push eventfrreiss/tensorflow-fred

frreiss

commit sha 85c5f4d56604174a94e25d313bb726ee5a08d698

Refactor int array copying code

view details

frreiss

commit sha d4c054fcd8c97e9dbd140235a6c652cbeb9aaf8f

Copy and swap input offset arrays on big-endian machines

view details

frreiss

commit sha 13e3fed8a7ff7e98179ad2f679f5f2cd668a6bcf

Revert getter functions back to original behavior

view details

push time in 3 months

create barnchfrreiss/tensorflow-fred

branch : issue-s390-lite-4

created branch time in 3 months

push eventfrreiss/tensorflow-fred

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

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

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

Lukas Geiger

commit sha b98db48be1b8238e213d1d6b9ab607ea96088224

Fix create_python_api_test.py

view details

Gunhan Gulsoy

commit sha 1868de838d1e21a80155fa50138451241e9cdc61

Use bazel's platform specific config options to switch between compiler options. Documentation: https://docs.bazel.build/versions/master/command-line-reference.html#flag--enable_platform_specific_config PiperOrigin-RevId: 283405622 Change-Id: I38709761dcc80c68a3ae3b301a3bc15ab5a43d51

view details

Saurabh Saxena

commit sha 7f347e6aa489e9e44f78feb4592f50d5d6bf94dc

Correctly handle Nones in cond_v2._make_indexed_slices_indices_types_match. PiperOrigin-RevId: 283409753 Change-Id: Ic83f8006693a404cbe4c723c0a103ffd98a76d1e

view details

Berkin Ilbeyi

commit sha 4eb24e50a79ea40a6ef2a6ed2497638fe51ed094

[XLA] Fix a scheduling bug with evictions to default mem. When simplifying the graph for dead code, we were previously removing the deleted instruction from the schedule. However, the scheduler, which is run after SimplifyGraph, relies on the original logical time (index into the instruction schedule). So, when some instructions have been deleted, we end up scheduling certain operation later than intended. Most seriously, the evictions could have been scheduled later than they were supposed to, corrupting the memory since we might have reused the evicted memory. The solution is to mark the deleted instructions with a nullptr in the schedule instead of actually deleting them. PiperOrigin-RevId: 283410438 Change-Id: Ia671ede14469dd00ac218cfb8d714e171152bb37

view details

push time in 3 months

create barnchfrreiss/tensorflow-fred

branch : issue-s390-lite-3

created branch time in 3 months

push eventfrreiss/bazel-zlinux-binaries

root

commit sha aeb1c6be5a8d988b869365eb135ba269de54eb73

Add compiled version of bazel 0.27.1 for ubuntu_18.04

view details

root

commit sha cf13a971d33593546b25bb7a1eb9e52e6610d6cd

Add compiled version of bazel 1.1.0 for ubuntu_18.04

view details

push time in 3 months

delete branch frreiss/tensorflow-fred

delete branch : issue-data-unique-test

delete time in 3 months

delete branch frreiss/tensorflow-fred

delete branch : issue-data-random-test

delete time in 3 months

delete branch frreiss/tensorflow-fred

delete branch : issue-data-zip-doc

delete time in 3 months

delete branch frreiss/tensorflow-fred

delete branch : issue-data-parallel-v2-doc

delete time in 3 months

delete branch frreiss/tensorflow-fred

delete branch : issue-grappler-endian-2

delete time in 3 months

delete branch frreiss/tensorflow-fred

delete branch : issue-data-parallel-doc

delete time in 3 months

delete branch frreiss/tensorflow-fred

delete branch : issue-data-sample-test

delete time in 3 months

delete branch frreiss/tensorflow-fred

delete branch : issue-data-lmdb-doc

delete time in 3 months

delete branch frreiss/tensorflow-fred

delete branch : issue-data-random-doc

delete time in 3 months

delete branch frreiss/tensorflow-fred

delete branch : issue-lmdb-endian

delete time in 3 months

delete branch frreiss/tensorflow-fred

delete branch : issue-data-assert-doc

delete time in 3 months

delete branch frreiss/tensorflow-fred

delete branch : issue-data-sample-doc

delete time in 3 months

delete branch frreiss/tensorflow-fred

delete branch : issue-data-comment-ptr

delete time in 3 months

delete branch frreiss/tensorflow-fred

delete branch : issue-mac-eigen-build

delete time in 3 months

delete branch frreiss/tensorflow-fred

delete branch : issue-linter-warning

delete time in 3 months

delete branch frreiss/tensorflow-fred

delete branch : issue-assert-refactor-2

delete time in 3 months

delete branch frreiss/tensorflow-fred

delete branch : issue-hash-doc

delete time in 3 months

delete branch frreiss/tensorflow-fred

delete branch : issue-smoke

delete time in 3 months

delete branch frreiss/tensorflow-fred

delete branch : issue-n-cpus

delete time in 3 months

delete branch frreiss/tensorflow-fred

delete branch : issue-nv-docker

delete time in 3 months

delete branch frreiss/tensorflow-fred

delete branch : issue-buildifier

delete time in 3 months

delete branch frreiss/tensorflow-fred

delete branch : issue-sparse-import

delete time in 3 months

delete branch frreiss/tensorflow-fred

delete branch : issue-vscode

delete time in 3 months

delete branch frreiss/tensorflow-fred

delete branch : issue-load-v1-shape

delete time in 3 months

delete branch frreiss/tensorflow-fred

delete branch : issue-squeeze-doc

delete time in 3 months

delete branch frreiss/tensorflow-fred

delete branch : issue-go_proto

delete time in 3 months

delete branch frreiss/tensorflow-fred

delete branch : issue-shape-decode-2

delete time in 3 months

delete branch frreiss/tensorflow-fred

delete branch : issue-predict-warning

delete time in 3 months

delete branch frreiss/tensorflow-fred

delete branch : issue-shape-decode

delete time in 3 months

create barnchfrreiss/tensorflow-fred

branch : issue-shape-decode

created branch time in 3 months

create barnchfrreiss/tensorflow-fred

branch : issue-predict-warning

created branch time in 3 months

create barnchfrreiss/tensorflow-fred

branch : issue-shape-decode-2

created branch time in 3 months

create barnchfrreiss/tensorflow-fred

branch : issue-go_proto

created branch time in 3 months

create barnchfrreiss/tensorflow-fred

branch : issue-squeeze-doc

created branch time in 3 months

create barnchfrreiss/tensorflow-fred

branch : issue-load-v1-shape

created branch time in 3 months

create barnchfrreiss/tensorflow-fred

branch : issue-vscode

created branch time in 3 months

delete branch frreiss/tensorflow-fred

delete branch : issue-vscode

delete time in 3 months

delete branch frreiss/tensorflow-fred

delete branch : issue-load-v1-shape

delete time in 3 months

delete branch frreiss/tensorflow-fred

delete branch : issue-squeeze-doc

delete time in 3 months

delete branch frreiss/tensorflow-fred

delete branch : issue-go_proto

delete time in 3 months

delete branch frreiss/tensorflow-fred

delete branch : issue-shape-decode-2

delete time in 3 months

delete branch frreiss/tensorflow-fred

delete branch : issue-predict-warning

delete time in 3 months

delete branch frreiss/tensorflow-fred

delete branch : issue-shape-decode

delete time in 3 months

delete branch frreiss/tensorflow-fred

delete branch : issue-random-quality

delete time in 3 months

more