Commits

Branch selector

 master

User selector

gnecula

All time

Commit History

Commits on May 4, 2023

• [chlo] Fix legalization of topk for dynamic shapes

  

  gnecula

  authored and

  tensorflower-gardener

  committedMay 4, 2023
  bd4041f

Commits on May 2, 2023

• Extend StablehloCanonicalizeDynamism to handle the operand_layouts attribute

  

  gnecula

  authored and

  tensorflower-gardener

  committedMay 2, 2023
  7f71edb

Commits on Mar 25, 2023

• Reverts a small part of cl/518773452 to attempt to fix OSS build failures

  

  gnecula

  authored and

  tensorflower-gardener

  committedMar 25, 2023
  dd91a94

Commits on Mar 23, 2023

• Enforce that XlaCallModule is used only with compatibility features

  

  gnecula

  authored and

  tensorflower-gardener

  committedMar 23, 2023
  953bf8f

Commits on Mar 8, 2023

• Add minimum supported version checking to XlaCallModuleOp.

  

  gnecula

  authored and

  tensorflower-gardener

  committedMar 8, 2023
  82481f6

Commits on Mar 6, 2023

• Enable XlaCallModule to work with modules with dynamic shapes but no dim_args_spec

  

  gnecula

  authored and

  tensorflower-gardener

  committedMar 6, 2023
  13fdd51

• Simplify the XlaCallModuleOp to always do shape refinement, not only when the arguments have dynamic shapes.

  

  gnecula

  authored and

  tensorflower-gardener

  committedMar 6, 2023
  fb12312

Commits on Feb 22, 2023

• Add test that XlaCallModule op can be traced on any platform.

  

  gnecula

  authored and

  tensorflower-gardener

  committedFeb 22, 2023
  07ca3fe

Commits on Feb 13, 2023

• Add support for platform checking and multi-platform modules to XlaCallModuleOp.

  

  gnecula

  authored and

  tensorflower-gardener

  committedFeb 13, 2023
  d56ce89

Commits on Feb 11, 2023

• Load more dialects in XlaCallModule to enable shape inference

  

  gnecula

  authored and

  tensorflower-gardener

  committedFeb 11, 2023
  30791b7

Commits on Dec 27, 2022

• Expand XlaCallModule to accept both i32 and i64 dimension arguments.

  

  gnecula

  authored and

  tensorflower-gardener

  committedDec 27, 2022
  c1d990e

• Add support for dynamic shapes to GPU threefry2x32 custom call.

  

  gnecula

  authored and

  tensorflower-gardener

  committedDec 27, 2022
  f3d32d5

Commits on Dec 11, 2022

• Improve and test error checking for XlaCallModule

  

  gnecula

  authored and

  tensorflower-gardener

  committedDec 11, 2022
  2448151

Commits on Nov 23, 2022

• Transition default use of XlaCallModule to StableHLO (version 2).

  

  gnecula

  authored and

  tensorflower-gardener

  committedNov 23, 2022
  42f9fd1

Commits on Nov 17, 2022

• Transition default use of XlaCallModule to StableHLO (version 2).

  

  gnecula

  authored and

  tensorflower-gardener

  committedNov 17, 2022
  a93ccb5

Commits on Nov 7, 2022

• Enable StableHLO tests for XlaCallModule with shape polymorphism.

  

  gnecula

  authored and

  tensorflower-gardener

  committedNov 7, 2022
  0ea7147

Commits on Oct 28, 2022

• Upgrades XlaCallModule op to support Stable HLO.

  

  gnecula

  authored and

  tensorflower-gardener

  committedOct 28, 2022
  86a9f31

Commits on Oct 26, 2022

• Prepare XlaCallModule for use with bytecode representations of modules.

  

  gnecula

  authored and

  tensorflower-gardener

  committedOct 26, 2022
  ddb688c

Commits on Sep 27, 2022

• Improvements for XlaCallModule.

  

  gnecula

  authored and

  tensorflower-gardener

  committedSep 27, 2022
  25bfb32

Commits on Jul 12, 2022

• An experimental TF op that carries a serialized HLO module.

  

  gnecula

  authored and

  tensorflower-gardener

  committedJul 12, 2022
  fd9ab41

Commits on Aug 2, 2021

• Add a new op XlaRngBitGenerator that wraps the HLO op RngBitGenertor.

  

  gnecula

  authored and

  tensorflower-gardener

  committedAug 2, 2021
  e02c9ff

Commits on Jul 19, 2021

• [host_callback] Increase number of threads for callback processing.

  

  gnecula

  authored and

  tensorflower-gardener

  committedJul 19, 2021
  62db2e9

• Change tfxla.variadic_reduce to point to XlaVariadicReduceV2.

  

  gnecula

  authored and

  tensorflower-gardener

  committedJul 19, 2021
  1bec81a

Commits on Jul 14, 2021

• Add shape inference rule for XlaDynamicSlice

  

  gnecula

  authored and

  tensorflower-gardener

  committedJul 14, 2021
  773a65e

Commits on Jul 12, 2021

• Enable tfxla.dot_general and tfxla.conv to force use of

  

  gnecula

  authored and

  tensorflower-gardener

  committedJul 12, 2021
  a7e179e

Commits on Jul 10, 2021

• Add a new version of the XlaVariadicReduce op.

  

  gnecula

  authored and

  tensorflower-gardener

  committedJul 10, 2021
  99f7b3d

Commits on Jun 29, 2021

• Improve the XlaDot shape inference rule for partially known shapes.

  

  gnecula

  authored and

  tensorflower-gardener

  committedJun 29, 2021
  cb9eaf3

Commits on Jun 10, 2021

• Extend the input type of XlaReduce and XlaReduceWindow to include booleans.

  

  gnecula

  authored and

  tensorflower-gardener

  committedJun 10, 2021
  4fda52d

Commits on Jun 3, 2021

• Add missing f16 and bf16 kernels for trigonometric functions.

  

  gnecula

  authored and

  tensorflower-gardener

  committedJun 3, 2021
  f452af7

Commits on Jun 2, 2021

• Add missing f16 and bf16 kernels for trigonometric functions.

  

  gnecula

  authored and

  tensorflower-gardener

  committedJun 2, 2021
  9da0caa

• Register tf.math.{div,floordiv,floormod} for all integer types.

  

  gnecula

  authored and

  tensorflower-gardener

  committedJun 2, 2021
  29fab5f

Commits on May 25, 2021

• Register tf.math.igamma[c] for [b]float16.

  

  gnecula

  authored and

  tensorflower-gardener

  committedMay 25, 2021
  8afb4f6

Commits on May 24, 2021

• Fix the registration of tf.reverse for unsigned integer types.

  

  gnecula

  authored and

  tensorflower-gardener

  committedMay 24, 2021
  9bafb1d

Commits on May 22, 2021

• Register tf.reverse for all integer types.

  

  gnecula

  authored and

  tensorflower-gardener

  committedMay 22, 2021
  827eed2

Commits on May 20, 2021

• Add kernels for a few more integer types for tf.math.minimum and maximum.

  

  gnecula

  authored and

  tensorflower-gardener

  committedMay 20, 2021
  7a1575e

Pagination

PreviousNext