Releases: tensorflow/tfx
Releases · tensorflow/tfx
TFX 1.6.0-rc0
Major Features and Improvements
- Added experimental support for TensorFlow Decision Forests models.
- Added Boolean type value artifacts.
- Function components defined with
@component
may now have optional/nullable
primitive type return values whenOptional[T]
is used in the return type
OutputDict.
Breaking Changes
For Pipeline Authors
- N/A
For Component Authors
- N/A
Deprecations
- N/A
Bug Fixes and Other Changes
- Depends on
numpy>=1.16,<2
. - Depends on
absl-py>=0.9,<2.0.0
. - Depends on
apache-beam[gcp]>=2.35,<3
. - Depends on
ml-metadata>=1.6.0,<1.7.0
. - Depends on
struct2tensor>=0.37.0,<0.38.0
. - Depends on
tensorflow-data-validation>=1.6.0,<1.7.0
. - Depends on
tensorflow-model-analysis>=0.37.0,<0.38.0
. - Depends on
tensorflow-transform>=1.6.0,<1.7.0
. - Depends on
tfx-bsl>=1.6.0,<1.7.0
. - Depends on
tensorflow>=1.15.5,!=2.0.*,!=2.1.*,!=2.2.*,!=2.3.*,!=2.4.*,!=2.5.*,!=2.6.*,<3
. - Depends on
kfp>=1.8.5,<2'
. - Pusher now copies the
saved_model.pb
file at last to prevent loading
SavedModel on invalid (partially available) directory state. - Always disable caching for exit handlers in Kubeflow V2 runner to
reflect latest status of dependent dag.
Documentation Updates
- N/A
TFX 1.4.1
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
- Ensures that Tensorflow is not re-installed during a container image build.
Documentation Updates
- N/A
TFX 1.5.0
Major Features and Improvements
- Added support for partial pipeline run. Users can now run a subset of nodes
in a pipeline while reusing artifacts generated in previous pipeline runs.
This is supported in LocalDagRunner and BeamDagRunner, and is exposed via
the TfxRunner API.
Breaking Changes
- N/A
For Pipeline Authors
- N/A
For Component Authors
- N/A
Deprecations
- N/A
Bug Fixes and Other Changes
- Increased docker timeout to 5 minutes for image building in CLI.
- Fixed KeyError when multiple Examples artifacts were used in Transform
without materialization. - Fixed error where Vertex Endpoints of the same name is not deduped
- Depends on
apache-beam[gcp]>=2.34,<3
. - Depends on
tensorflow>=1.15.2,!=2.0.*,!=2.1.*,!=2.2.*,!=2.3.*,!=2.4.*,!=2.5.*,!=2.6.*,<2.8
. - Depends on
tensorflow-serving-api>=1.15,!=2.0.*,!=2.1.*,!=2.2.*,!=2.3.*,!=2.4.*,!=2.5.*,!=2.6.*,<3
. - Depends on
ml-metadata>=1.5.0,<1.6.0
. - Depends on
struct2tensor>=0.36.0,<0.37.0
. - Depends on
tensorflow-data-validation>=1.5.0,<1.6.0
. - Depends on
tensorflow-model-analysis>=0.36.0,<0.37.0
. - Depends on
tensorflow-transform>=1.5.0,<1.6.0
. - Depends on
tfx-bsl>=1.5.0,<1.6.0
.
Documentation Updates
- N/A
TFX 1.3.4
TFX 1.5.0-rc0
Major Features and Improvements
- Added support for partial pipeline run. Users can now run a subset of nodes
in a pipeline while reusing artifacts generated in previous pipeline runs.
This is supported in LocalDagRunner and BeamDagRunner, and is exposed via
the TfxRunner API.
Breaking Changes
- N/A
For Pipeline Authors
- N/A
For Component Authors
- N/A
Deprecations
- N/A
Bug Fixes and Other Changes
- Increased docker timeout to 5 minutes for image building in CLI.
- Fixed KeyError when multiple Examples artifacts were used in Transform
without materialization. - Fixed error where Vertex Endpoints of the same name is not deduped
- Depends on
apache-beam[gcp]>=2.34,<3
. - Depends on
tensorflow>=1.15.2,!=2.0.*,!=2.1.*,!=2.2.*,!=2.3.*,!=2.4.*,!=2.5.*,!=2.6.*,<2.8
. - Depends on
tensorflow-serving-api>=1.15,!=2.0.*,!=2.1.*,!=2.2.*,!=2.3.*,!=2.4.*,!=2.5.*,!=2.6.*,<3
. - Depends on
ml-metadata>=1.5.0,<1.6.0
. - Depends on
struct2tensor>=0.36.0,<0.37.0
. - Depends on
tensorflow-data-validation>=1.5.0,<1.6.0
. - Depends on
tensorflow-model-analysis>=0.36.0,<0.37.0
. - Depends on
tensorflow-transform>=1.5.0,<1.6.0
. - Depends on
tfx-bsl>=1.5.0,<1.6.0
.
Documentation Updates
- N/A
TFX 1.4.0
Major Features and Improvements
- Supported endpoint overwrite for CAIP BulkInferrer.
- Added support for outputting and encoding
tf.RaggedTensor
s in TFX
Transform component. - Added conditional for TFX running on KFPv2 (Vertex).
- Supported component level beam pipeline args for Vertex (KFPV2DagRunner).
- Support exit handler for TFX running on KFPv2 (Vertex).
- Added RangeConfig for QueryBasedExampleGen to select date using query
pattern. - Added support for union of Channels as input to standard TFX components.
Users can use channel.union() to combine multiple Channels and use as input
to these compnents. Artfacts resolved from these channels are expected to
have the same type, and passed to components in no particular order.
Breaking Changes
- Calling
TfxRunner.run(pipeline)
with the Pipeline IR proto will no longer
be supported. Please switch toTfxRunner.run_with_ir(pipeline)
instead.
If you are callingTfxRunner.run(pipeline)
with the Pipeline object, this
change should not affect you.
For Pipeline Authors
- N/A
For Component Authors
- N/A
Deprecations
- Deprecated python3.6 support.
Bug Fixes and Other Changes
- Depends on
google-cloud-aiplatform>=1.5.0,<2
. - Depends on
tensorflow>=1.15.2,!=2.0.*,!=2.1.*,!=2.2.*,!=2.3.*,!=2.4.*,!=2.5.*,<2.7
. - Depends on
pyarrow>=1,<6
. - Fixed FileBasedExampleGen driver for Kubeflow v2 (Vertex). Driver can
update exec_properties for its executor now, which enables {SPAN} feature. - example_gen.utils.dict_to_example now accepts Numpy types
- Updated pytest to include v6.x
- Depends on
apache-beam[gcp]>=2.33,<3
. - Depends on
ml-metadata>=1.4.0,<1.5.0
. - Depends on
struct2tensor>=0.35.0,<0.36.0
. - Depends on
tensorflow-data-validation>=1.4.0,<1.5.0
. - Depends on
tensorflow-model-analysis>=0.35.0,<0.36.0
. - Depends on
tensorflow-transform>=1.4.0,<1.5.0
. - Depends on
tfx-bsl>=1.4.0,<1.5.0
. - Fixed error where Vertex Endpoints of the same name is not deduped
Documentation Updates
- N/A
TFX 1.4.0-rc0
Major Features and Improvements
- Supported endpoint overwrite for CAIP BulkInferrer.
- Added support for outputting and encoding
tf.RaggedTensor
s in TFX
Transform component. - Added conditional for TFX running on KFPv2 (Vertex).
- Supported component level beam pipeline args for Vertex (KFPV2DagRunner).
- Support exit handler for TFX running on KFPv2 (Vertex).
- Added RangeConfig for QueryBasedExampleGen to select date using query
pattern. - Added support for union of Channels as input to standard TFX components.
Users can use channel.union() to combine multiple Channels and use as input
to these compnents. Artfacts resolved from these channels are expected to
have the same type, and passed to components in no particular order.
Breaking Changes
- Calling
TfxRunner.run(pipeline)
with the Pipeline IR proto will no longer
be supported. Please switch toTfxRunner.run_with_ir(pipeline)
instead.
If you are callingTfxRunner.run(pipeline)
with the Pipeline object, this
change should not affect you.
For Pipeline Authors
- N/A
For Component Authors
- N/A
Deprecations
- Deprecated python3.6 support.
Bug Fixes and Other Changes
- Depends on
google-cloud-aiplatform>=1.5.0,<2
. - Depends on
tensorflow>=1.15.2,!=2.0.*,!=2.1.*,!=2.2.*,!=2.3.*,!=2.4.*,!=2.5.*,<2.7
. - Depends on
pyarrow>=1,<6
. - Fixed FileBasedExampleGen driver for Kubeflow v2 (Vertex). Driver can
update exec_properties for its executor now, which enables {SPAN} feature. - example_gen.utils.dict_to_example now accepts Numpy types
- Updated pytest to include v6.x
- Depends on
apache-beam[gcp]>=2.33,<3
. - Depends on
ml-metadata>=1.4.0,<1.5.0
. - Depends on
struct2tensor>=0.35.0,<0.36.0
. - Depends on
tensorflow-data-validation>=1.4.0,<1.5.0
. - Depends on
tensorflow-model-analysis>=0.35.0,<0.36.0
. - Depends on
tensorflow-transform>=1.4.0,<1.5.0
. - Depends on
tfx-bsl>=1.4.0,<1.5.0
.
Documentation Updates
- N/A
TFX 1.3.3
TFX 1.2.1
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
- Added support for a custom metadata-ui-json filename in KubeflowDagRunner.
- Fixed missing type information marker file 'py.typed'.
Documentation Updates
- N/A