Skip to content
Permalink
master
Switch branches/tags

Name already in use

A tag already exists with the provided branch name. Many Git commands accept both tag and branch names, so creating this branch may cause unexpected behavior. Are you sure you want to create this branch?

Commits on Nov 29, 2022

  1. Update TFRT dependency to use revision 

    http://github.com/tensorflow/runtime/commit/6643084df0385624d9ecdaa7861479f62aef87ea.

PiperOrigin-RevId: 491497457
    @tensorflower-gardener
    tensorflower-gardener committed Nov 29, 2022
    82e83ea

  2. Integrate LLVM at llvm/llvm-project@bf15f1e489aa 

    Updates LLVM usage to match
[bf15f1e489aa](llvm/llvm-project@bf15f1e489aa)

PiperOrigin-RevId: 491495923
    @tensorflower-gardener
    tensorflower-gardener committed Nov 29, 2022
    2116b01

  3. Disable flags_test for macOS until sysconfig.get_path is fixed 

    PiperOrigin-RevId: 491487463
    @terryheo @tensorflower-gardener
    terryheo authored and tensorflower-gardener committed Nov 29, 2022
    c10f898

  4. Create a hook interface in open source that updates metrics about XLA… 

    … TPU compilation

PiperOrigin-RevId: 491486664
    @tensorflower-gardener
    tensorflower-gardener committed Nov 29, 2022
    7991b34

  5. Refactor directed_interleave_dataset_test to move out tests for `choo… 

    …se_from_datasets` and `sample_from_datasets` to their designated test targets.

PiperOrigin-RevId: 491476188
    @tensorflower-gardener
    tensorflower-gardener committed Nov 29, 2022
    18cef6f

  6. Partial fix for issue with tensors containing dynamic size. 

    PiperOrigin-RevId: 491473961
    @arfaian @tensorflower-gardener
    arfaian authored and tensorflower-gardener committed Nov 29, 2022
    07e8d01

  7. Update TFRT dependency to use revision 

    http://github.com/tensorflow/runtime/commit/7cfacc90f7cb99eaa77b004be44c2cb653b84497.

PiperOrigin-RevId: 491469530
    @tensorflower-gardener
    tensorflower-gardener committed Nov 29, 2022
    a73c148

  8. [xla:gpu:runtime] Add captured CUDA graphs caching 

    PiperOrigin-RevId: 491469302
    @ezhulenev @tensorflower-gardener
    ezhulenev authored and tensorflower-gardener committed Nov 29, 2022
    297f1fe

  9. Integrate LLVM at llvm/llvm-project@5611bf69fc48 

    Updates LLVM usage to match
[5611bf69fc48](llvm/llvm-project@5611bf69fc48)

PiperOrigin-RevId: 491466869
    @tensorflower-gardener
    tensorflower-gardener committed Nov 29, 2022
    74fc3aa

  10. Re-generate Flatbuffer in lite/experimental/acceleration/configuration 

    PiperOrigin-RevId: 491464313
    @tensorflower-gardener
    tensorflower-gardener committed Nov 29, 2022
    d12d92b

  11. Stop defaulting the default constructor for Padding2D. 

    In C++20, types that declare any constructors are no longer aggregates, breaking existing instances of aggregate initialization of this type.

PiperOrigin-RevId: 491463757
    @tensorflower-gardener
    tensorflower-gardener committed Nov 29, 2022
    852c6c5

Commits on Nov 28, 2022

  1. Declare explicit instantiations. 

    PiperOrigin-RevId: 491463266
    @tkoeppe @tensorflower-gardener
    tkoeppe authored and tensorflower-gardener committed Nov 28, 2022
    dd1780d

  2. Adds Python Bindings for HloModuleGroup and Hlo passes. Passes can be… 

    … invoked through Python for interactively understanding compiler behaviors.

PiperOrigin-RevId: 491460870
    @yliu120 @tensorflower-gardener
    yliu120 authored and tensorflower-gardener committed Nov 28, 2022
    915ec39

  3. Remove enable_native_ops and its usages 

    PiperOrigin-RevId: 491460786
    @cky9301 @tensorflower-gardener
    cky9301 authored and tensorflower-gardener committed Nov 28, 2022
    1924bff

  4. Migrates _pywrap_tf_optimizer to cc_shared_library. 

    Links _pywrap_tensorflow_internal.so and libtensorflow_framework.so using dynamic_deps. This change removes duplicated statically linked targets.

PiperOrigin-RevId: 491459689
    @tensorflower-gardener
    tensorflower-gardener committed Nov 28, 2022
    1ab18f2

  5. Remove internal_cuda_deps from tensorflow.default.bzl. It is empty an… 

    …d only used in one place

PiperOrigin-RevId: 491454542
    @tensorflower-gardener
    David Dunleavy authored and tensorflower-gardener committed Nov 28, 2022
    57bc95e

  6. Add a blank line after Doxygen '\file' declarations. 

    This is needed otherwise the following text is interpreted
as the file name, rather than the file description.
(See docs for \file at <https://doxygen.nl/manual/commands.html#cmdfile>.)

PiperOrigin-RevId: 491454526
    @fergushenderson @tensorflower-gardener
    fergushenderson authored and tensorflower-gardener committed Nov 28, 2022
    20d1039

  7. Add rerandomize_each_iteration option to sample_from_datasets. This o… 

    …ption controls behavior of seeded sequence of random numbers used for sampling. If set to `True`, the same deterministic random number sequence is generated every epoch. If set to `False`, a different deterministic sequence of random numbers is generated every epoch.

PiperOrigin-RevId: 491454310
    @tensorflower-gardener
    tensorflower-gardener committed Nov 28, 2022
    5c27c07

  8. Correct comments about the default behavior for random(). The defau… 

    …lt behavior is not re-randomizing for each epoch.

PiperOrigin-RevId: 491439277
    @tensorflower-gardener
    tensorflower-gardener committed Nov 28, 2022
    ab123a4

  9. Improve error message for padded_batch dataset elements type check. 

    Added special error message for the case of datasets of datasets.
Also tweaked the message to make it a bit clearer for other
unsupported cases.

PiperOrigin-RevId: 491439181
    @tensorflower-gardener
    tensorflower-gardener committed Nov 28, 2022
    c6cb70d

  10. Update TFRT dependency to use revision 

    http://github.com/tensorflow/runtime/commit/32b22e9d6dc3aea0d57e5d008e247a3bb762b59e.

PiperOrigin-RevId: 491437498
    @tensorflower-gardener
    tensorflower-gardener committed Nov 28, 2022
    3d3ef5d

  11. #tf-data Optimize byte counting code. 

    For non-variant datasets, we were previously generating and then discarding two error statuses. Now we instead just check the tensor's dtype and avoid any expensive logic for non-variants.

PiperOrigin-RevId: 491436015
    @aaudiber @tensorflower-gardener
    aaudiber authored and tensorflower-gardener committed Nov 28, 2022
    094de67

  12. [mlir][tosa] Adds MHLO -> TOSA legalizations for while and return 

    PiperOrigin-RevId: 491433862
    @not-jenni @tensorflower-gardener
    not-jenni authored and tensorflower-gardener committed Nov 28, 2022
    ed8c49b

  13. Update TFRT dependency to use revision 

    http://github.com/tensorflow/runtime/commit/1c7d6ebe575e48ba36c9269d2fe8c1d8574bf804.

PiperOrigin-RevId: 491431440
    @tensorflower-gardener
    tensorflower-gardener committed Nov 28, 2022
    c7ce69e

  14. Replace tf_cuda_cc_test with tsl_gpu_cc_test in compiler/xla 

    PiperOrigin-RevId: 491428255
    @tensorflower-gardener
    David Dunleavy authored and tensorflower-gardener committed Nov 28, 2022
    1f020d9

  15. [DelegatePerformanceBenchmark] Allow the stable ABI delegate provider… 

    … to load JSON formatted delegate settings file.

PiperOrigin-RevId: 491427698
    @terryheo @tensorflower-gardener
    terryheo authored and tensorflower-gardener committed Nov 28, 2022
    082a998

  16. Updated tf.while_loop and tf.cond v2 docstrings 

      - Referring to AutoGraph with code examples and link to the tutorial page
  - Converting old-style code doc to teatable doc test

PiperOrigin-RevId: 491426452
    @tensorflower-gardener
    tensorflower-gardener committed Nov 28, 2022
    cdfcbf7

  17. Migrates pywrap_saved_model to cc_shared_library. 

    AddS libtensorflow_framework and pywrap_tensorflow_internal using cc_shared_library dynamic_deps to pywrap_saved_model. Updates def_file_filter.py.tpl to export the missing symbols in Windows.

PiperOrigin-RevId: 491424390
    @tensorflower-gardener
    tensorflower-gardener committed Nov 28, 2022
    8fabe54

  18. Update TFRT dependency to use revision 

    http://github.com/tensorflow/runtime/commit/26054603bdc2944d9c697270a0cd784825e66a7d.

PiperOrigin-RevId: 491423175
    @tensorflower-gardener
    tensorflower-gardener committed Nov 28, 2022
    666163e

  19. Integrate LLVM at llvm/llvm-project@f51170bffd50 

    Updates LLVM usage to match
[f51170bffd50](llvm/llvm-project@f51170bffd50)

PiperOrigin-RevId: 491420075
    @tensorflower-gardener
    tensorflower-gardener committed Nov 28, 2022
    ef08bb7

  20. In compiler/xla: 

    * update tf_cuda_library to tsl_gpu_library
* update tf_copts to tsl_copts
* update get_compatible_with_cloud to use TSL version

PiperOrigin-RevId: 491414513
    @tensorflower-gardener
    David Dunleavy authored and tensorflower-gardener committed Nov 28, 2022
    d1d22a5

  21. Return from method GetDeadlineOp::Compute when deadline is not presen… 

    …t (instead of using the not present optional).

PiperOrigin-RevId: 491413078
    @tensorflower-gardener
    tensorflower-gardener committed Nov 28, 2022
    c1dd8d0

  22. [XLA] Add gpu test for xla_compile 

    PiperOrigin-RevId: 491412157
    @anlunx @tensorflower-gardener
    anlunx authored and tensorflower-gardener committed Nov 28, 2022
    f9aab58

  23. Merge pull request #58288 from Jerry-Ge:master 

    PiperOrigin-RevId: 491409453
    @tensorflower-gardener
    tensorflower-gardener committed Nov 28, 2022
    3f3039f

  24. [xla:runtime] NFC: Simplify custom call test 

    Simplify TestOpts by removing options that are already available as CompilationPipelineOptions struct.

PiperOrigin-RevId: 491408306
    @ezhulenev @tensorflower-gardener
    ezhulenev authored and tensorflower-gardener committed Nov 28, 2022
    87e5dbd
Older