TFX 0.27.0

Major Features and Improvements

• Supports different types of quantizations on TFLite conversion using
  TFLITE_REWRITER by setting quantization_optimizations,
  quantization_supported_types and quantization_enable_full_integer. Flag
  definitions can be found here: Post-traning
  quantization.
• Added automatic population of tfdv.StatsOptions.vocab_paths when computing
  statistics within the Transform component.

Breaking changes

For pipeline authors

• enable_quantization from TFLITE_REWRITER is removed and setting
  quantization_optimizations = [tf.lite.Optimize.DEFAULT] will perform the
  same type of quantization, dynamic range quantization. Users of the
  TFLITE_REWRITER who do not enable quantization should be uneffected.
• Default value for infer_feature_shape for SchemaGen changed from False
  to True, as indicated in previous release log. The inferred schema might
  change if you do not specify infer_feature_shape. It might leads to
  changes of the type of input features in Transform and Trainer code.

For component authors

• N/A

Deprecations

• Pipeline information is not be stored on the local filesystem anymore using
  Kubeflow Pipelines orchestration with CLI. Instead, CLI will always use the
  latest version of the pipeline in the Kubeflow Pipeline cluster. All
  operations will be executed based on the information on the Kubeflow
  Pipeline cluster. There might be some left files on
  ${HOME}/tfx/kubeflow or ${HOME}/kubeflow but those will not be used
  any more.
• The tfx.components.common_nodes.importer_node.ImporterNode class has been
  moved to tfx.dsl.components.common.importer.Importer, with its
  old module path kept as a deprecated alias, which will be removed in a
  future version.
• The tfx.components.common_nodes.resolver_node.ResolverNode class has been
  moved to tfx.dsl.components.common.resolver.Resolver, with its
  old module path kept as a deprecated alias, which will be removed in a
  future version.
• The tfx.dsl.resolvers.BaseResolver class has been
  moved to tfx.dsl.components.common.resolver.ResolverStrategy, with its
  old module path kept as a deprecated alias, which will be removed in a
  future version.
• Deprecated input/output compatibility aliases for ExampleValidator,
  Evaluator, Trainer and Pusher.

Bug fixes and other changes

• InfraValidator supports using alternative TensorFlow Serving image in case
  deployed environment cannot reach the public internet (nor the docker hub).
  Such alternative image should behave the same as official
  tensorflow/serving image such as the same model volume path, serving port,
  etc.
• Executor in tfx.extensions.google_cloud_ai_platform.pusher.executor
  supported regional endpoint and machine_type.
• Starting from this version, proto files which are used to generate
  component-level configs are included in the tfx package directly.
• The tfx.dsl.io.fileio.NotFoundError exception unifies handling of not-
  found errors across different filesystem plugin backends.
• Fixes the serialization of zero-valued default when using RuntimeParameter
  on Kubeflow.
• Depends on apache-beam[gcp]>=2.27,<3.
• Depends on ml-metadata>=0.27.0,<0.28.0.
• Depends on pyarrow>=1,<3.
• Depends on tensorflow>=1.15.2,!=2.0.*,!=2.1.*,!=2.2.*,!=2.3.*,<3.
• Depends on tensorflow-data-validation>=0.27.0,<0.28.0.
• Depends on tensorflow-model-analysis>=0.27.0,<0.28.0.
• Depends on tensorflow-serving-api>=1.15,!=2.0.*,!=2.1.*,!=2.2.*,!=2.3.*,<3.
• Depends on tensorflow-transform>=0.27.0,<0.28.0.
• Depends on tfx-bsl>=0.27.0,<0.28.0.

Documentation updates

• N/A

TFX 0.26.1

• This a bug fix only version

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

• The tfx.version attribute was restored.

Documentation updates

• N/A

TFX 0.26.0

Major Features and Improvements

• Supported output examples artifact for BulkInferrer which can be used to
  link with downstream training.
• TFX Transform switched to a (notably) faster and more accurate
  implementation of tft.quantiles analyzer.
• Added native TF 2 implementation of Transform. The default
  behavior will continue to use Tensorflow's compat.v1 APIs. This can be
  overriden by passing force_tf_compat_v1=False and enabling TF 2 behaviors.
  The default behavior for TF 2 will be switched to the new native
  implementation in a future release.
• Added support for passing a callable to set pre/post transform statistic
  generation options.
• In addition to the "tfx" pip package, a dependency-light distribution of the
  core pipeline authoring functionality of TFX is now available as the
  "ml-pipelines-sdk" pip package. This package does not include first-party
  TFX components. The "tfx" pip package is still the recommended installation
  path for TFX.

Breaking changes

• Wheel package building for TFX has changed, and users need to follow the
  [new TFX package build instructions]
  (https://github.com/tensorflow/tfx/blob/master/package_build/README.md) to
  build wheels for TFX.

For pipeline authors

• N/A

For component authors

• N/A

Deprecations

• TrainerFnArgs is deprecated by FnArgs.
• Deprecated DockerComponentConfig class: user should set a DockerPlatformConfig
  proto in platform_config using with_platform_config() API instead.

Bug fixes and other changes

• Official TFX container image's entrypoint is changed so the image can be
  used as a custom worker for Dataflow.
• In the published TFX container image, wheel files are now used to install
  TFX, and the TFX source code has been moved to /tfx/src.
• Added a skeleton of CLI support for Kubeflow V2 runner, and implemented
  support for pipeline operations.
• Added an experimental template to use with Kubeflow V2 runner.
• Added sanitization of user-specified pipeline name in Kubeflow V2 runner.
• Migrated deployment_config in Kubeflow V2 runner from Any proto message
  to Struct, to ensure compatibility across different copies of the proto
  libraries.
• The tfx.dsl.io.fileio filesystem handler will delegate to
  tensorflow.io.gfile for any unknown filesystem schemes if TensorFlow
  is installed.
• Skipped ephemeral package when the beam flag
  'worker_harness_container_image' is set.
• The tfx.dsl.io.makedirs call now succeeds if the directory already exists.
• Depends on apache-beam[gcp]>=2.25,!=2.26,<3.
• Depends on keras-tuner>=1,<1.0.2.
• Depends on kfp-pipeline-spec>=0.1.3,<0.2.
• Depends on ml-metadata>=0.26.0,<0.27.0.
• Depends on tensorflow>=1.15.2,!=2.0.*,!=2.1.*,!=2.2.*,!=2.4.*,<3.
• Depends on tensorflow-data-validation>=0.26,<0.27.
• Depends on tensorflow-model-analysis>=0.26,<0.27.
• Depends on tensorflow-serving>=1.15,!=2.0.*,!=2.1.*,!=2.2.*,!=2.4.*,<3.
• Depends on tensorflow-transform>=0.26,<0.27.
• Depends on tfx-bsl>=0.26.1,<0.27.

Documentation updates

• N/A

TFX 0.26.0-rc0

Major Features and Improvements

• Supported output examples artifact for BulkInferrer which can be used to
  link with downstream training.
• TFX Transform switched to a (notably) faster and more accurate
  implementation of tft.quantiles analyzer.
• Added native TF 2 implementation of Transform. The default
  behavior will continue to use Tensorflow's compat.v1 APIs. This can be
  overriden by passing force_tf_compat_v1=False and enabling TF 2 behaviors.
  The default behavior for TF 2 will be switched to the new native
  implementation in a future release.
• Added support for passing a callable to set pre/post transform statistic
  generation options.

Breaking changes

• Wheel package building for TFX has changed, and users need to follow the
  [new TFX package build instructions]
  (https://github.com/tensorflow/tfx/blob/master/package_build/README.md) to
  build wheels for TFX.

For pipeline authors

• N/A

For component authors

• N/A

Deprecations

• TrainerFnArgs is deprecated by FnArgs.
• Deprecated DockerComponentConfig class: user should set a DockerPlatformConfig
  proto in platform_config using with_platform_config() API instead.

Bug fixes and other changes

• Official TFX container image's entrypoint is changed so the image can be
  used as a custom worker for Dataflow.
• In the published TFX container image, wheel files are now used to install
  TFX, and the TFX source code has been moved to /tfx/src.
• Added a skeleton of CLI support for Kubeflow V2 runner, and implemented
  support for pipeline operations.
• Added an experimental template to use with Kubeflow V2 runner.
• Added sanitization of user-specified pipeline name in Kubeflow V2 runner.
• Migrated deployment_config in Kubeflow V2 runner from Any proto message
  to Struct, to ensure compatibility across different copies of the proto
  libraries.
• The tfx.dsl.io.fileio filesystem handler will delegate to
  tensorflow.io.gfile for any unknown filesystem schemes if TensorFlow
  is installed.
• Depends on apache-beam[gcp]>=2.25,!=2.26,<3.
• Depends on keras-tuner>=1,<1.0.2.
• Depends on kfp-pipeline-spec>=0.1.3,<0.2.
• Depends on ml-metadata>=0.26.0,<0.27.0.
• Depends on tensorflow>=1.15.2,!=2.0.*,!=2.1.*,!=2.2.*,!=2.4.*,<3.
• Depends on tensorflow-data-validation>=0.26,<0.27.
• Depends on tensorflow-model-analysis>=0.26,<0.27.
• Depends on tensorflow-serving>=1.15,!=2.0.*,!=2.1.*,!=2.2.*,!=2.4.*,<3.
• Depends on tensorflow-transform>=0.26,<0.27.
• Depends on tfx-bsl>=0.26.1,<0.27.

Documentation updates

• N/A

TFX 0.25.0

Major Features and Improvements

• Supported multiple artifacts for Transform's input example and output
  transformed example channels.

• Added support for processing specific spans in file-based ExampleGen with
  range configuration.

• Added ContainerExecutableSpec in portable IR to support container components
  portable orchestrator.

• Added Placeholder utility library. Placeholder can be used to represent
  not-yet-available value at pipeline authoring time.

• Added support for the tfx.dsl.io.fileio pluggable filesystem interface,
  with initial support for local files and the Tensorflow GFile filesystem
  implementation.

• SDK and example code now uses tfx.dsl.io.fileio instead of tf.io.gfile
  when possible for filesystem I/O implementation portability.

• From this release TFX will also be hosting nightly packages on
  https://pypi-nightly.tensorflow.org. To install the nightly package use the
  following command:

  pip install -i https://pypi-nightly.tensorflow.org/simple tfx
  

  Note: These nightly packages are unstable and breakages are likely to happen.
  The fix could often take a week or more depending on the complexity
  involved for the wheels to be available on the PyPI cloud service. You can
  always use the stable version of TFX available on PyPI by running the
  command

  pip install tfx
  

• Added CloudTuner KFP e2e example running on Google Cloud Platform with
  distributed tuning.

• Migrated BigQueryExampleGen to the new ReadFromBigQuery on all runners.

• Introduced Kubeflow V2 DAG runner, which is based on
  Kubeflow IR spec.
  Same as KubeflowDagRunner it will compile the DSL pipeline into a payload
  but not trigger the execution locally.

• Added 'penguin' example. Penguin example uses Palmer Penguins dataset and
  classify penguin species using four numeric features.

• Iris e2e examples are replaced by penguin examples.

• TFX BeamDagRunner is migrated to use the tech stack built on top of IR.
  While this is no-op to users, it is a major step towards supporting more
  flexible TFX DSL semetic.
  Please refer to the RFC
  of IR to learn more details.

• Supports forward compatibility when evolving TFX artifact types, which
  allows jobs of old release and new release run with the same MLMD instance.

Breaking changes

• Moved the directory that CLI stores pipeline information from
  ${HOME}/${ORCHESTRATOR} to ${HOME}/tfx/${ORCHESTRATOR}. For example,
  "/kubeflow" was changed to "/tfx/kubeflow". This directory is used to
  store pipeline information including pipeline ids in the Kubeflow Pipelines
  cluster which are needed to create runs or update pipelines.
  These files will be moved automatically when it is first used and no
  separate action is needed.
  See https://github.com/tensorflow/tfx/blob/master/docs/guide/cli.md for the
  detail.

For pipeline authors

• N/A

For component authors

• N/A

Deprecations

• Modules under tfx.components.base have been deprecated and moved to
  tfx.dsl.components.base in preparation for releasing a pipeline authoring
  package without explicit Tensorflow dependency.
• Deprecated setting instance_name at pipeline node level. Instead, users
  are encouraged to set id explicitly of any pipeline node through newly
  added APIs.

Bug fixes and other changes

• Added the LocalDagRunner to allow local pipeline execution without using
  Apache Beam. This functionality is in development.
• Introduced dependency to tensorflow-cloud Python package, with intention
  to separate out Google Cloud Platform specific extensions.
• Depends on mmh>=2.2,<3 in container image for potential performance
  improvement for Beam based hashes.
• New extra dependencies [examples] is required to use codes inside
  tfx/examples.
• Fixed the run_component script.
• Stopped depending on WTForms.
• Fixed an issue with Transform cache and beam 2.24-2.25 in an interactive
  notebook that caused it to fail.
• Scripts - run_component - Added a way to output artifact properties.
• Fixed an issue resulting in incorrect cache miss to ExampleGen when no
  beam_pipeline_args is provided.
• Changed schema as an optional input channel of Trainer as schema can be
  accessed from TFT graph too.
• Fixed an issue during recording of a component's execution where
  "missing or modified key in exec_properties" was raised from MLMD when
  exec_properties both omitted an existing property and added a new
  property.
• Supported users to set id of pipeline nodes directly.
• Added a new template, 'penguin' which is simple subset of
  penguin examples,
  and uses the same
  Palmer Penguins
  dataset. The new template focused on easy ingestion of user's own data.
• Changed default data path for the taxi template from tfx-template/data
  to tfx-template/data/taxi.
• Fixed a bug which crashes the pusher when infra validation did not pass.
• Depends on apache-beam[gcp]>=2.25,<3.
• Depends on attrs>=19.3.0,<21.
• Depends on kfp-pipeline-spec>=0.1.2,<0.2.
• Depends on kfp>=1.1.0,<2.
• Depends on ml-metadata>=0.25,<0.26.
• Depends on tensorflow-cloud>=0.1,<0.2.
• Depends on tensorflow-data-validation>=0.25,<0.26.
• Depends on tensorflow-hub>=0.9.0,<0.10.
• Depends on tensorflow-model-analysis>=0.25,<0.26.
• Depends on tensorflow-transform>=0.25,<0.26.
• Depends on tfx-bsl>=0.25,<0.26.

Documentation updates

• N/A

TFX 0.25.0-rc2

Version 0.25.0

Major Features and Improvements

• Supported multiple artifacts for Transform's input example and output
  transformed example channels.

• Added support for processing specific spans in file-based ExampleGen with
  range configuration.

• Added ContainerExecutableSpec in portable IR to support container components
  portable orchestrator.

• Added Placeholder utility library. Placeholder can be used to represent
  not-yet-available value at pipeline authoring time.

• Added support for the tfx.dsl.io.fileio pluggable filesystem interface,
  with initial support for local files and the Tensorflow GFile filesystem
  implementation.

• SDK and example code now uses tfx.dsl.io.fileio instead of tf.io.gfile
  when possible for filesystem I/O implementation portability.

• From this release TFX will also be hosting nightly packages on
  https://pypi-nightly.tensorflow.org. To install the nightly package use the
  following command:

  pip install -i https://pypi-nightly.tensorflow.org/simple tfx
  

  Note: These nightly packages are unstable and breakages are likely to happen.
  The fix could often take a week or more depending on the complexity
  involved for the wheels to be available on the PyPI cloud service. You can
  always use the stable version of TFX available on PyPI by running the
  command

  pip install tfx
  

• Added CloudTuner KFP e2e example running on Google Cloud Platform with
  distributed tuning.

• Migrated BigQueryExampleGen to the new ReadFromBigQuery on all runners.

• Introduced Kubeflow V2 DAG runner, which is based on
  Kubeflow IR spec.
  Same as KubeflowDagRunner it will compile the DSL pipeline into a payload
  but not trigger the execution locally.

• Added 'penguin' example. Penguin example uses Palmer Penguins dataset and
  classify penguin species using four numeric features.

• Iris e2e examples are replaced by penguin examples.

• TFX BeamDagRunner is migrated to use the tech stack built on top of IR.
  While this is no-op to users, it is a major step towards supporting more
  flexible TFX DSL semetic.
  Please refer to the RFC
  of IR to learn more details.

• Supports forward compatibility when evolving TFX artifact types, which
  allows jobs of old release and new release run with the same MLMD instance.

Breaking changes

• Moved the directory that CLI stores pipeline information from
  ${HOME}/${ORCHESTRATOR} to ${HOME}/tfx/${ORCHESTRATOR}. For example,
  "/kubeflow" was changed to "/tfx/kubeflow". This directory is used to
  store pipeline information including pipeline ids in the Kubeflow Pipelines
  cluster which are needed to create runs or update pipelines.
  These files will be moved automatically when it is first used and no
  separate action is needed.
  See https://github.com/tensorflow/tfx/blob/master/docs/guide/cli.md for the
  detail.

For pipeline authors

• N/A

For component authors

• N/A

Deprecations

• Modules under tfx.components.base have been deprecated and moved to
  tfx.dsl.components.base in preparation for releasing a pipeline authoring
  package without explicit Tensorflow dependency.
• Deprecated setting instance_name at pipeline node level. Instead, users
  are encouraged to set id explicitly of any pipeline node through newly
  added APIs.

Bug fixes and other changes

• Added the LocalDagRunner to allow local pipeline execution without using
  Apache Beam. This functionality is in development.
• Introduced dependency to tensorflow-cloud Python package, with intention
  to separate out Google Cloud Platform specific extensions.
• Depends on mmh>=2.2,<3 in container image for potential performance
  improvement for Beam based hashes.
• New extra dependencies [examples] is required to use codes inside
  tfx/examples.
• Fixed the run_component script.
• Stopped depending on WTForms.
• Fixed an issue with Transform cache and beam 2.24-2.25 in an interactive
  notebook that caused it to fail.
• Scripts - run_component - Added a way to output artifact properties.
• Fixed an issue resulting in incorrect cache miss to ExampleGen when no
  beam_pipeline_args is provided.
• Changed schema as an optional input channel of Trainer as schema can be
  accessed from TFT graph too.
• Fixed an issue during recording of a component's execution where
  "missing or modified key in exec_properties" was raised from MLMD when
  exec_properties both omitted an existing property and added a new
  property.
• Supported users to set id of pipeline nodes directly.
• Added a new template, 'penguin' which is simple subset of
  penguin examples,
  and uses the same
  Palmer Penguins
  dataset. The new template focused on easy ingestion of user's own data.
• Changed default data path for the taxi template from tfx-template/data
  to tfx-template/data/taxi.
• Depends on apache-beam[gcp]>=2.25,<3.
• Depends on attrs>=19.3.0,<21.
• Depends on kfp-pipeline-spec>=0.1.2,<0.2.
• Depends on kfp>=1.1.0,<2.
• Depends on ml-metadata>=0.25,<0.26.
• Depends on tensorflow-cloud>=0.1,<0.2.
• Depends on tensorflow-data-validation>=0.25,<0.26.
• Depends on tensorflow-hub>=0.9.0,<0.10.
• Depends on tensorflow-model-analysis>=0.25,<0.26.
• Depends on tensorflow-transform>=0.25,<0.26.
• Depends on tfx-bsl>=0.25,<0.26.

Documentation updates

• N/A

TFX 0.25.0-rc1

Version 0.25.0

Major Features and Improvements

• Supported multiple artifacts for Transform's input example and output
  transformed example channels.

• Added support for processing specific spans in file-based ExampleGen with
  range configuration.

• Added ContainerExecutableSpec in portable IR to support container components
  portable orchestrator.

• Added Placeholder utility library. Placeholder can be used to represent
  not-yet-available value at pipeline authoring time.

• Added support for the tfx.dsl.io.fileio pluggable filesystem interface,
  with initial support for local files and the Tensorflow GFile filesystem
  implementation.

• SDK and example code now uses tfx.dsl.io.fileio instead of tf.io.gfile
  when possible for filesystem I/O implementation portability.

• From this release TFX will also be hosting nightly packages on
  https://pypi-nightly.tensorflow.org. To install the nightly package use the
  following command:

  pip install -i https://pypi-nightly.tensorflow.org/simple tfx
  

  Note: These nightly packages are unstable and breakages are likely to happen.
  The fix could often take a week or more depending on the complexity
  involved for the wheels to be available on the PyPI cloud service. You can
  always use the stable version of TFX available on PyPI by running the
  command

  pip install tfx
  

• Added CloudTuner KFP e2e example running on Google Cloud Platform with
  distributed tuning.

• Migrated BigQueryExampleGen to the new ReadFromBigQuery on all runners.

• Introduced Kubeflow V2 DAG runner, which is based on
  Kubeflow IR spec.
  Same as KubeflowDagRunner it will compile the DSL pipeline into a payload
  but not trigger the execution locally.

• Added 'penguin' example. Penguin example uses Palmer Penguins dataset and
  classify penguin species using four numeric features.

• Iris e2e examples are replaced by penguin examples.

• TFX BeamDagRunner is migrated to use the tech stack built on top of IR.
  While this is no-op to users, it is a major step towards supporting more
  flexible TFX DSL semetic.
  Please refer to the RFC
  of IR to learn more details.

• Supports forward compatibility when evolving TFX artifact types, which
  allows jobs of old release and new release run with the same MLMD instance.

Breaking changes

• Moved the directory that CLI stores pipeline information from
  ${HOME}/${ORCHESTRATOR} to ${HOME}/tfx/${ORCHESTRATOR}. For example,
  "/kubeflow" was changed to "/tfx/kubeflow". This directory is used to
  store pipeline information including pipeline ids in the Kubeflow Pipelines
  cluster which are needed to create runs or update pipelines.
  These files will be moved automatically when it is first used and no
  separate action is needed.
  See https://github.com/tensorflow/tfx/blob/master/docs/guide/cli.md for the
  detail.

For pipeline authors

• N/A

For component authors

• N/A

Deprecations

• Modules under tfx.components.base have been deprecated and moved to
  tfx.dsl.components.base in preparation for releasing a pipeline authoring
  package without explicit Tensorflow dependency.
• Deprecated setting instance_name at pipeline node level. Instead, users
  are encouraged to set id explicitly of any pipeline node through newly
  added APIs.

Bug fixes and other changes

• Added the LocalDagRunner to allow local pipeline execution without using
  Apache Beam. This functionality is in development.
• Introduced dependency to tensorflow-cloud Python package, with intention
  to separate out Google Cloud Platform specific extensions.
• Depends on mmh>=2.2,<3 in container image for potential performance
  improvement for Beam based hashes.
• New extra dependencies [examples] is required to use codes inside
  tfx/examples.
• Fixed the run_component script.
• Stopped depending on WTForms.
• Fixed an issue with Transform cache and beam 2.24-2.25 in an interactive
  notebook that caused it to fail.
• Scripts - run_component - Added a way to output artifact properties.
• Fixed an issue resulting in incorrect cache miss to ExampleGen when no
  beam_pipeline_args is provided.
• Changed schema as an optional input channel of Trainer as schema can be
  accessed from TFT graph too.
• Fixed an issue during recording of a component's execution where
  "missing or modified key in exec_properties" was raised from MLMD when
  exec_properties both omitted an existing property and added a new
  property.
• Supported users to set id of pipeline nodes directly.
• Added a new template, 'penguin' which is simple subset of
  penguin examples,
  and uses the same
  Palmer Penguins
  dataset. The new template focused on easy ingestion of user's own data.
• Changed default data path for the taxi template from tfx-template/data
  to tfx-template/data/taxi.
• Depends on apache-beam[gcp]>=2.25,<3.
• Depends on attrs>=19.3.0,<21.
• Depends on kfp-pipeline-spec>=0.1.0,<0.2.
• Depends on kfp>=1.1.0,<2.
• Depends on ml-metadata>=0.25,<0.26.
• Depends on tensorflow-cloud>=0.1,<0.2.
• Depends on tensorflow-data-validation>=0.25,<0.26.
• Depends on tensorflow-hub>=0.9.0,<0.10.
• Depends on tensorflow-model-analysis>=0.25,<0.26.
• Depends on tensorflow-transform>=0.25,<0.26.
• Depends on tfx-bsl>=0.25,<0.26.

Documentation updates

• N/A

TFX 0.25.0-rc0 Release

Major Features and Improvements

• Supported multiple artifacts for Transform's input example and output
  transformed example channels.

• Added support for processing specific spans in file-based ExampleGen with
  range configuration.

• Added ContainerExecutableSpec in portable IR to support container components
  portable orchestrator.

• Added Placeholder utility library. Placeholder can be used to represent
  not-yet-available value at pipeline authoring time.

• Added the LocalDagRunner to allow local pipeline execution without using
  Apache Beam.

• Added support for the tfx.dsl.io.fileio pluggable filesystem interface,
  with initial support for local files and the Tensorflow GFile filesystem
  implementation.

• SDK and example code now uses tfx.dsl.io.fileio instead of tf.io.gfile
  when possible for filesystem I/O implementation portability.

• From this release TFX will also be hosting nightly packages on
  https://pypi-nightly.tensorflow.org. To install the nightly package use the
  following command:

  pip install -i https://pypi-nightly.tensorflow.org/simple tfx
  

  Note: These nightly packages are unstable and breakages are likely to happen.
  The fix could often take a week or more depending on the complexity
  involved for the wheels to be available on the PyPI cloud service. You can
  always use the stable version of TFX available on PyPI by running the
  command

  pip install tfx
  

• Added CloudTuner KFP e2e example running on Google Cloud Platform with
  distributed tuning.

• Migrated BigQueryExampleGen to the new ReadFromBigQuery on all runners.

• Introduced Kubeflow V2 DAG runner, which is based on
  Kubeflow IR spec.
  Same as KubeflowDagRunner it will compile the DSL pipeline into a payload
  but not trigger the execution locally.

Breaking changes

• N/A

For pipeline authors

• N/A

For component authors

• N/A

Deprecations

• Modules under tfx.components.base have been deprecated and moved to
  tfx.dsl.components.base in preparation for releasing a pipeline authoring
  package without explicit Tensorflow dependency.

Bug fixes and other changes

• Introduced dependency to tensorflow-cloud Python package, with intention
  to separate out Google Cloud Platform specific extensions.
• Depends on mmh>=2.2,<3 in container image for potential performance
  improvement for Beam based hashes.
• New extra dependencies [examples] is required to use codes inside
  tfx/examples.
• Fixed the run_component script.
• Stopped depending on WTForms.
• Fixed an issue with Transform cache and beam 2.24-2.25 in an interactive
  notebook that caused it to fail.
• Scripts - run_component - Added a way to output artifact properties.
• Fixed an issue resulting in incorrect cache miss to ExampleGen when no
  beam_pipeline_args is provided.
• Changed schema as an optional input channel of Trainer as schema can be
  accessed from TFT graph too.
• Depends on apache-beam[gcp]>=2.25,<3.
• Depends on ml-metadata>=0.24,<0.25.
• Depends on tensorflow-cloud>=0.1,<0.2.
• Depends on tensorflow-data-validation>=0.25,<0.26.
• Depends on tensorflow-hub>=0.9.0,<0.10.
• Depends on tensorflow-model-analysis>=0.25,<0.26.
• Depends on tensorflow-transform>=0.25,<0.26.
• Depends on tfx-bsl>=0.25,<0.26.

Documentation updates

• N/A

TFX 0.24.1

Major Features and Improvements

• N/A

Bug fixes and other changes

• Fixes issues where custom property access of a missing property created an invalid MLMD Artifact protobuf message.

Breaking changes

• N/A

For pipeline authors

• N/A

For component authors

• N/A

Documentation updates

• N/A

Deprecations

• N/A

TFX 0.22.2

Major Features and Improvements

• N/A

Bug fixes and other changes

• Reuse Examples artifact type introduced in TFX 0.23 to allow older release jobs running together with TFX 0.23+ release.

Deprecations

• N/A

Breaking changes

• N/A

For pipeline authors

• N/A

For component authors

• N/A

Documentation updates

• N/A