Skip to content

Releases: tensorflow/tfx

TFX 1.15.1

13 May 20:43
849dcd2
Compare
Choose a tag to compare

Version 1.15.1

Major Features and Improvements

Breaking Changes

  • Support KFP pipeline spec 2.1.0 version schema and YAML files with KFP v2 DAG runner

For Pipeline Authors

For Component Authors

Deprecations

Bug Fixes and Other Changes

Dependency Updates

Package Name Version Constraints Previously (in v1.14.0) Comments
kfp-pipeline-spec kfp-pipeline-spec>=0.1.10,<0.2 >0.1.13,<0.2

Documentation Updates

TFX 1.15.0 Release

29 Apr 17:07
f04638d
Compare
Choose a tag to compare

Version 1.15.0

Major Features and Improvements

  • Dropped python 3.8 support.
  • Extend GetPipelineRunExecutions, GetPipelineRunArtifacts APIs to support
    filtering by execution create_time, type.
  • ExampleValidator and DistributionValidator now support anomalies alert
    generation. Users can use their own toolkits to extract and process the
    alerts from the execution parameter.
  • Allow DistributionValidator baseStatistics input channel artifacts to be
    empty for cold start of data validation.
  • ph.make_proto() allows constructing proto-valued placeholders, e.g. for
    larger config protos fed to a component.
  • ph.join_path() is like os.path.join() but for placeholders.
  • Support passing in experimental_debug_stripper into the Transform
    pipeline runner.

Breaking Changes

  • Placeholder and all subclasses have been moved to other modules, their
    structure has been changed and they're now immutable. Most users won't care
    (the main public-facing API is unchanged and behaves the same way). If you
    do special operations like isinstance() or some kind of custom
    serialization on placeholders, you will have to update your code.
  • placeholder.Placeholder.traverse() now returns more items than before,
    namely also placeholder operators like _ConcatOperator (which is the
    implementation of Python's + operator).
  • The placeholder.RuntimeInfoKey enumeration was removed. Just hard-code the
    appropriate string values in your code, and reference the new Literal type
    placeholder.RuntimeInfoKeys if you want to ensure correctness.
  • Arguments to @component must now be passed as kwargs and its return type
    is changed from being a Type to just being a callable that returns a new
    instance (like the type's initializer). This will allow us to instead return
    a factory function (which is not a Type) in future. For a given
    @component def C(), this means:
    • You should not use C as a type anymore. For instance, replace
      isinstance(foo, C) with something else. Depending on your use case, if
      you just want to know whether it's a component, then use
      isinstance(foo, tfx.types.BaseComponent) or
      isinstance(foo, tfx.types.BaseFunctionalComponent).
      If you want to know which component it is, check its .id instead.
      Existing such checks will break type checking today and may additionally
      break at runtime in future, if we migrate to a factory function.
    • You can continue to use C.test_call() like before, and it will
      continue to be supported in future.
    • Any type declarations using foo: C break and must be replaced with
      foo: tfx.types.BaseComponent or
      foo: tfx.types.BaseFunctionalComponent.
    • Any references to static class members like C.EXECUTOR_SPEC breaks
      type checking today and should be migrated away from. In particular, for
      .EXECUTOR_SPEC.executor_class().Do() in unit tests, use .test_call()
      instead.
    • If your code previously asserted a wrong type declaration on C, this
      can now lead to (justified) type checking errors that were previously
      hidden due to C being of type Any.
  • ph.to_list() was renamed to ph.make_list() for consistency.

For Pipeline Authors

For Component Authors

Deprecations

  • Deprecated python 3.8

Bug Fixes and Other Changes

  • Fixed a synchronization bug in google_cloud_ai_platform tuner.
  • Print best tuning trials only from the chief worker of google_cloud_ai_platform tuner.
  • Add a kpf dependency in the docker-image extra packages.
  • Fix BigQueryExampleGen failure without custom_config.

Dependency Updates

Package Name Version Constraints Previously (in v1.14.0) Comments
keras-tuner >=1.0.4,<2,!=1.4.0,!=1.4.1 >=1.0.4,<2
packaging >=20,<21 >=22
attrs 19.3.0,<22 19.3.0,<24
google-cloud-bigquery >=2.26.0,<3 >=3,<4
tensorflow >=2.15,<2.16 >=2.13,<2.14
tensorflow-decision-forests >=1.0.1,<1.9 >=1.0.1,<2
tensorflow-hub >=0.9.0,<0.14 >=0.15.0,<0.16
tensorflow-serving >=1.15,!=2.0.*,!=2.1.*,!=2.2.*,!=2.3.*,!=2.4.*,!=2.5.*,!=2.6.*,!=2.7.*,!=2.8.*,<3 >=2.15,<2.16

Documentation Updates

TFX 1.15.0-rc0 Release

25 Apr 18:16
6da173e
Compare
Choose a tag to compare
Pre-release

Major Features and Improvements

  • Dropped python 3.8 support.
  • Extend GetPipelineRunExecutions, GetPipelineRunArtifacts APIs to support
    filtering by execution create_time, type.
  • ExampleValidator and DistributionValidator now support anomalies alert
    generation. Users can use their own toolkits to extract and process the
    alerts from the execution parameter.
  • Allow DistributionValidator baseStatistics input channel artifacts to be
    empty for cold start of data validation.
  • ph.make_proto() allows constructing proto-valued placeholders, e.g. for
    larger config protos fed to a component.
  • ph.join_path() is like os.path.join() but for placeholders.
  • Support passing in experimental_debug_stripper into the Transform
    pipeline runner.

Breaking Changes

  • Placeholder and all subclasses have been moved to other modules, their
    structure has been changed and they're now immutable. Most users won't care
    (the main public-facing API is unchanged and behaves the same way). If you
    do special operations like isinstance() or some kind of custom
    serialization on placeholders, you will have to update your code.
  • placeholder.Placeholder.traverse() now returns more items than before,
    namely also placeholder operators like _ConcatOperator (which is the
    implementation of Python's + operator).
  • The placeholder.RuntimeInfoKey enumeration was removed. Just hard-code the
    appropriate string values in your code, and reference the new Literal type
    placeholder.RuntimeInfoKeys if you want to ensure correctness.
  • Arguments to @component must now be passed as kwargs and its return type
    is changed from being a Type to just being a callable that returns a new
    instance (like the type's initializer). This will allow us to instead return
    a factory function (which is not a Type) in future. For a given
    @component def C(), this means:
    • You should not use C as a type anymore. For instance, replace
      isinstance(foo, C) with something else. Depending on your use case, if
      you just want to know whether it's a component, then use
      isinstance(foo, tfx.types.BaseComponent) or
      isinstance(foo, tfx.types.BaseFunctionalComponent).
      If you want to know which component it is, check its .id instead.
      Existing such checks will break type checking today and may additionally
      break at runtime in future, if we migrate to a factory function.
    • You can continue to use C.test_call() like before, and it will
      continue to be supported in future.
    • Any type declarations using foo: C break and must be replaced with
      foo: tfx.types.BaseComponent or
      foo: tfx.types.BaseFunctionalComponent.
    • Any references to static class members like C.EXECUTOR_SPEC breaks
      type checking today and should be migrated away from. In particular, for
      .EXECUTOR_SPEC.executor_class().Do() in unit tests, use .test_call()
      instead.
    • If your code previously asserted a wrong type declaration on C, this
      can now lead to (justified) type checking errors that were previously
      hidden due to C being of type Any.
  • ph.to_list() was renamed to ph.make_list() for consistency.

Deprecations

  • Deprecated python 3.8

Bug Fixes and Other Changes

  • Fixed a synchronization bug in google_cloud_ai_platform tuner.
  • Print best tuning trials only from the chief worker of google_cloud_ai_platform tuner.
  • Add a kpf dependency in the docker-image extra packages.
  • Fix BigQueryExampleGen failure without custom_config.

Dependency Updates

Package Name Version Constraints Previously (in v1.14.0) Comments
keras-tuner >=1.0.4,<2,!=1.4.0,!=1.4.1 >=1.0.4,<2
packaging >=20,<21 >=22
attrs 19.3.0,<22 19.3.0,<24
google-cloud-bigquery >=2.26.0,<3 >=3,<4
tensorflow >=2.15,<2.16 >=2.13,<2.14
tensorflow-decision-forests >=1.0.1,<1.9 >=1.0.1,<2
tensorflow-hub >=0.9.0,<0.14 >=0.15.0,<0.16
tensorflow-serving >=1.15,!=2.0.*,!=2.1.*,!=2.2.*,!=2.3.*,!=2.4.*,!=2.5.*,!=2.6.*,!=2.7.*,!=2.8.*,<3 >=2.15,<2.16

Documentation Updates

TFX 1.14.0 Release

06 Sep 18:31
44e493a
Compare
Choose a tag to compare

Major Features and Improvements

  • Added python 3.10 support.

Breaking Changes

  • Placeholder (and _PlaceholderOperator) are no longer Jsonable.
  • Optimize MLMD register type to one call in most time instead of two calls.

For Pipeline Authors

  • N/A

For Component Authors

  • Replace "tf_estimator" with "tfma_eval" as the identifier for tfma
    EvalSavedModel. "tf_estimator" is now serves as the identifier for the normal
    estimator model with any signature (by default 'serving').

Deprecations

  • N/A

Bug Fixes and Other Changes

  • Apply latest TFX image vulnerability resolutions (base OS and software updates)

Dependency Updates

Package Name Version Constraints Previously (in v1.13.0) Comments
tensorflow-hub >=0.9.0,<0.14 >=0.9.0,<0.13
pyarrow >=10,<11 >=6,<7
apache-beam >=2.40,<3 >=2.47,<3
scikit-learn >=1.0,<2 >=0.23,<0.24
google-api-core <3 <1.33
google-cloud-aiplatform >=1.6.2,<2 >=1.6.2,<1.18
tflite-support >=0.4.3,<0.4.5 >=0.4.2,<0.4.3
pyyaml >=6,<7 >=3.12,<6 Issue with installation of PyYaml 5.4.1. (yaml/pyyaml#724)
tensorflow >=2.13,<2.14 >=2.12,<2.13
tensorflowjs >=4.5,<5 >=3.6.0,<4

Documentation Updates

  • N/A

TFX 1.14.0-rc0 Release

28 Aug 18:11
f2a02d5
Compare
Choose a tag to compare
Pre-release

Major Features and Improvements

  • Added python 3.10 support.

Breaking Changes

  • Placeholder (and _PlaceholderOperator) are no longer Jsonable.
  • Optimize MLMD register type to one call in most time instead of two calls.

For Pipeline Authors

  • N/A

For Component Authors

  • Replace "tf_estimator" with "tfma_eval" as the identifier for tfma
    EvalSavedModel. "tf_estimator" is now serves as the identifier for the normal
    estimator model with any signature (by default 'serving').

Deprecations

  • N/A

Bug Fixes and Other Changes

  • Apply latest TFX image vulnerability resolutions (base OS and software updates)

Dependency Updates

Package Name Version Constraints Previously (in v1.13.0) Comments
tensorflow-hub >=0.9.0,<0.14 >=0.9.0,<0.13
pyarrow >=10,<11 >=6,<7
apache-beam >=2.40,<3 >=2.47,<3
scikit-learn >=1.0,<2 >=0.23,<0.24
google-api-core <3 <1.33
google-cloud-aiplatform >=1.6.2,<2 >=1.6.2,<1.18
tflite-support >=0.4.3,<0.4.5 >=0.4.2,<0.4.3
pyyaml >=6,<7 >=3.12,<6 Issue with installation of PyYaml 5.4.1. (yaml/pyyaml#724)
tensorflow >=2.13,<2.14 >=2.12,<2.13
tensorflowjs >=4.5,<5 >=3.6.0,<4

Documentation Updates

  • N/A

TFX 1.13.0 Release

03 May 18:43
7b8b8a7
Compare
Choose a tag to compare

Major Features and Improvements

  • Supported setting the container image at a component level for Kubeflow V2
    Dag Runner.

Breaking Changes

For Pipeline Authors

  • Conditional can be used from tfx.dsl.Cond (Given from tfx import v1 as tfx).

  • Dummy channel for testing can be constructed by
    tfx.testing.Channel(artifact_type).

  • placeholder.Placeholder.placeholders_involved() was replaced with
    placeholder.Placeholder.traverse().

  • placeholder.Predicate.dependent_channels() was replaced with
    channel_utils.get_dependent_channels(Placeholder).

  • placeholder.Predicate.encode_with_keys(...) was replaced with
    channel_utils.encode_placeholder_with_channels(Placeholder, ...).

  • placeholder.Predicate.from_comparison() removed (was deprecated)

  • enable external_pipeline_artifact_query for querying artifact within one pipeline

  • Support InputArtifact[List[Artifact]] annotation in Python function custom component

For Component Authors

  • N/A

Deprecations

  • Deprecate python 3.7 support

Bug Fixes and Other Changes

  • Support to task type "workerpool1" of CLUSTER_SPEC in Vertex AI training's
    service according to the changes of task type in Tuner component.
  • Propagates unexpected import failures in the public v1 module.

Dependency Updates

Package Name Version Constraints Previously (in v1.12.0) Comments
click >=7,<9 >=7,<8
ml-metadata ~=1.13.1 ~=1.12.0 Synced release train
protobuf >=3.13,<4 >=3.20.3,<5 To support TF 2.12
struct2tensor ~=0.44.0 ~=0.43.0 Synced release train
tensorflow ~=2.12.0 >=1.15.5,<2 or ~=2.11.0
tensorflow-data-validation ~=1.13.0 ~=1.12.0 Synced release train
tensorflow-model-analysis ~=0.44.0 ~=0.43.0 Synced release train
tensorflow-transform ~=1.13.0 ~=1.12.0 Synced release train
tfx-bsl ~=1.13.0 ~=1.12.0 Synced release train

Documentation Updates

  • Added page for TFX-Addons

TFX 1.13.0-rc0 Release

14 Apr 21:31
abb4ae2
Compare
Choose a tag to compare
Pre-release

Major Features and Improvements

  • Supported setting the container image at a component level for Kubeflow V2
    Dag Runner.

Breaking Changes

For Pipeline Authors

  • Conditional can be used from tfx.dsl.Cond (Given from tfx import v1 as tfx).

  • Dummy channel for testing can be constructed by
    tfx.testing.Channel(artifact_type).

  • placeholder.Placeholder.placeholders_involved() was replaced with
    placeholder.Placeholder.traverse().

  • placeholder.Predicate.dependent_channels() was replaced with
    channel_utils.get_dependent_channels(Placeholder).

  • placeholder.Predicate.encode_with_keys(...) was replaced with
    channel_utils.encode_placeholder_with_channels(Placeholder, ...).

  • placeholder.Predicate.from_comparison() removed (was deprecated)

  • enable external_pipeline_artifact_query for querying artifact within one pipeline

For Component Authors

  • N/A

Deprecations

  • Deprecate python 3.7 support

Bug Fixes and Other Changes

  • Support to task type "workerpool1" of CLUSTER_SPEC in Vertex AI training's
    service according to the changes of task type in Tuner component.
  • Propagates unexpected import failures in the public v1 module.

Dependency Updates

Package Name Version Constraints Previously (in v1.12.0) Comments
click >=7,<9 >=7,<8
ml-metadata ~=1.13.1 ~=1.12.0 Synced release train
protobuf >=3.13,<4 >=3.20.3,<5 To support TF 2.12
struct2tensor ~=0.44.0 ~=0.43.0 Synced release train
tensorflow ~=2.12.0 >=1.15.5,<2 or ~=2.11.0
tensorflow-data-validation ~=1.13.0 ~=1.12.0 Synced release train
tensorflow-model-analysis ~=0.44.0 ~=0.43.0 Synced release train
tensorflow-transform ~=1.13.0 ~=1.12.0 Synced release train
tfx-bsl ~=1.13.0 ~=1.12.0 Synced release train

Documentation Updates

  • Added page for TFX-Addons

TFX 1.12.0 Release

19 Dec 16:48
2e24d20
Compare
Choose a tag to compare

Major Features and Improvements

  • N/A

Breaking Changes

  • N/A

For Pipeline Authors

  • N/A

For Component Authors

  • N/A

Deprecations

  • N/A

Bug Fixes and Other Changes

  • ExampleValidator and DistributionValidator now support custom validations.

Dependency Updates

Package Name Version Constraints Previously (in v1.11.0) Comments
tensorflow ~=2.11.0 >=1.15.5,<2 or ~=2.10.0
tensorflow-decision-forests >=1.0.1,<2 ==1.0.1 Make it compatible with more TF versions.
ml-metadata ~=1.12.0 ~=1.11.0 Synced release train
struct2tensor ~=0.43.0 ~=0.42.0 Synced release train
tensorflow-data-validation ~=1.12.0 ~=1.11.0 Synced release train
tensorflow-model-analysis ~=0.43.0 ~=0.42.0 Synced release train
tensorflow-transform ~=1.12.0 ~=1.11.0 Synced release train
tfx-bsl ~=1.12.0 ~=1.11.0 Synced release train

Documentation Updates

  • N/A

TFX 1.12.0-rc0 Release

13 Dec 22:59
0f2b75f
Compare
Choose a tag to compare
Pre-release

Major Features and Improvements

  • N/A

Breaking Changes

  • N/A

For Pipeline Authors

  • N/A

For Component Authors

  • N/A

Deprecations

  • N/A

Bug Fixes and Other Changes

  • ExampleValidator and DistributionValidator now support custom validations.

Dependency Updates

Package Name Version Constraints Previously (in v1.11.0) Comments
tensorflow ~=2.11.0 >=1.15.5,<2 or ~=2.10.0
tensorflow-decision-forests >=1.0.1,<2 ==1.0.1 Make it compatible with more TF versions.
ml-metadata ~=1.12.0 ~=1.11.0 Synced release train
struct2tensor ~=0.43.0 ~=0.42.0 Synced release train
tensorflow-data-validation ~=1.12.0 ~=1.11.0 Synced release train
tensorflow-model-analysis ~=0.43.0 ~=0.42.0 Synced release train
tensorflow-transform ~=1.12.0 ~=1.11.0 Synced release train
tfx-bsl ~=1.12.0 ~=1.11.0 Synced release train

Documentation Updates

  • N/A

TFX 1.11.0 Release

23 Nov 20:52
fb6977c
Compare
Choose a tag to compare

Major Features and Improvements

  • This is the last version that supports TensorFlow 1.15.x. TF 1.15.x support
    will be removed in the next version. Please check the
    TF2 migration guide to migrate
    to TF2.

  • Artifact/Channel properties now support the new MLMD PROTO property type.

  • Supports environment variables in the placeholder expression.
    This placeholder can be used to generate beam_pipeline_args
    dynamically.

Breaking Changes

  • N/A

For Pipeline Authors

  • N/A

For Component Authors

  • N/A

Deprecations

  • N/A

Bug Fixes and Other Changes

  • Moved tflite-support related dependencies from [examples] to a separate
    [tflite-support] extra.
  • Moved flax related dependencies from [examples] to a separate [flax]
    extra.
  • Statistics gen and Schema gen now crash on empty input examples and statistics respectively.
  • Importer will now check that an existing artifact has the same type as the intended output before reusing the existing artifact.
  • Importer will now use the most recently created artifact when reusing an existing artifact instead of the one with the highest ID.
  • Proto placeholder now works with proto files that have non-trivial transitive dependencies.
  • Adding tutorials for recommenders and ranking

Dependency Updates

Package Name Version Constraints Previously (in v1.8.0) Comments
tensorflow >=1.15.5,<2 or ~=2.10.0 >=1.15.5,<2 or ~=2.9.0
tflite-support ~=0.4.2 >=0.1.0a1,<0.2.1 Update to a TF-2.10 compatible version.
google-cloud-aiplatform >=1.6.2,<1.18 >=1.6.2,<2 Added to help pip dependency resolution.
ml-metadata ~=1.11.0 ~=1.10.0 Synced release train
struct2tensor ~=0.42.0 ~=0.41.0 Synced release train
tensorflow-data-validation ~=1.11.0 ~=1.10.0 Synced release train
tensorflow-model-analysis ~=0.42.0 ~=0.41.0 Synced release train
tensorflow-transform ~=1.11.0 ~=1.10.0 Synced release train
tfx-bsl ~=1.11.0 ~=1.10.0 Synced release train

Documentation Updates

  • N/A