All notable changes to this project will be documented in this file.
The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.
- Minor ci pipeline updates [#139]
- Replace
method_whitelist
withallowed_methods
#137
- Drops support for Python 3.6 and 3.7
- Fixes
typing-extension
insetup.py
#133
- Drops support for Python 3.5
- Updates
typing-extension
to version4.0.0
#131
InferenceClient.create_inference_request*
methods now default toretry=True
. #130
InferenceClient.do_bulk_inference
is now faster due to processing requests in parallel #128InferenceClient.do_bulk_inference
is more resilient and handles errors internally. Instead of raising an Exception if the inference request to the service fails, thedo_bulk_inference
method will place a special error response object in the returned list. This can be considered a breaking API change, because the special error response object will have a value ofNone
for thelabels
key. As this project is still versioned below 1.0.0, the breaking API change does not warrant a major version update. See #128 for details.- Improve reliability by tweaking retry configuration #129
- Support for reading training jobs using model name in
read_job_by_model_name
- Support for inference using external URL in
create_inference_request_with_url
- Notebooks for showcasing model template usage. #116
- Notebook for showcasing generic model template.
- Notebook for showcasing hierarchical model template.
- Notebook for showcasing regression model template.
- Support for Business Blueprint API. #118
- This feature is not supported in DAR service yet, it is added for internal testing purposes.
read_business_blueprint_template_collection
method to list business blueprint templates collection.read_business_blueprint_template_by_id
method to fetch information of a specific business blueprint template for the given business_blueprint_id
- Support for Business blueprint id in
create_job
andcreate_job_and_wait
methods. #114- This feature is not supported in DAR service yet, it is added for internal testing purposes.
- Either model_template_id or business_blueprint_id has to be specified in
create_job
method. - Both model_template_id and business_blueprint_id are not allowed.
- HTTP errors sent by the Data Attribute Recommendation service are now reported correctly
as a
DARHTTPException
(including debug information) even when the request was retried. #109 Fixes #104
- Remove passing of external deployment ID and job ID from SDK API. This functionality
never materialized in the Data Attribute Recommendation service and is now also removed in
the SDK.
This is an API change in the SDK because it changes the method signature of some
methods to remove the optional
job_id
anddeployment_id
arguments. Passing these arguments always resulted in an error returned by the Data Attribute Recommendation service. For this reason, this change is not a breaking change. This change effectively reverts #98. - Plain-text
http://
URLs are now allowed forlocalhost
. This is useful for local development. #108
- Fix specification of dependencies. The SDK package was incorrectly depending on an older version of the requests package. If you have installed 0.8.0, it is recommended to update to 0.8.1 and also ensure you have the latest dependencies. #102
- Add CONTRIBUTING.md and SECURITY.md #92
- This project is now following the Best Practices set forth by the Core Infrastructure Initiative! See CII badge details. #92
- Add
construct_from_cf_env
method to construct client instances from Data Attribute Recommendation service binding on SAP Business Technology Platform. #97 - Add support for user-specified Job and Deployment IDs when creating the respective Job and Deployment resources. This change is not yet generally available in the Data Attribute Recommendation service. #98
- Python 3.5 has reached end-of-life in September 2020. Support for Python 3.5 will be removed in one of the upcoming releases.
- Documentation updated to include reference to TechED 2020 workshop #93
- Tests: report branch coverage #93
- Tests: run tests with Python 3.9
- Improve logging: quieten OnlineCredentialsSource, make polling explicit #81
- Log job progress in ModelManagerClient.wait_for_job #87
- Documentation: add link to SAP community #58
ModelCreator.create
performs an initial check if the model name is already used and raisesModelAlreadyExists
in this case #60, #64ModelManager.wait_for_job
now logs with level INFO while polling #66- Change default
retry
parameter inInferenceClient.do_bulk_inference
fromFalse
toTrue
. Please check the updated documentation for implications on charging. #67, #62
- Log message missing model deployment ID #59
- Expose a Dataset's
validationMessage
inDatasetValidationFailed
#33, #69
- Improvements to build process #53
- No functional changes.
- Improvements to build process #52
- No functional changes.
- Update links in README and setup.py #27
- Update documentation #35
- Fix links in documentation and switch the readthedocs.org theme #37
- Make documentation more consistent #40
0.6.3 - 2020-07-02
- Improvements to build process
0.6.2 - 2020-06-30
- First public release