v0.0.0a01

Shift back to PyPI because setuptools is not in test PyPI

v0.3.0

Added

• A Project helper class.
• A convenience method create_or_load_project which loads in a project if it is already created.
• Accepts AZURE as a DeploymentType.

Changed

• Compatibility with Unbox API OpenAPI refactor.
• Models and datasets must be added to projects.
• Deprecates categorical_features_map in favor of categorical_feature_names for model and dataset uploads.
• Moved TaskType attribute from the Model level to the Project level. Creating a Project now requires specifying the TaskType.
• Removed name from add_dataset.
• Changed description to commit_message from add_dataset, add_dataframe and add_model.
• requirements_txt_file no longer optional for model uploads.
• NLP dataset character limit is now 1000 characters.

Fixed

• More comprehensive model and dataset upload validation.
• Bug with duplicate feature names for NLP datasets if uploading same dataset twice.
• Added protobuf==3.2.0 to requirements to fix bug with model deployment.

0.3.0rc1

Added

• A Project helper class.
• A convenience method create_or_load_project which loads in a project if it is already created.
• Accepts AZURE as a DeploymentType.

Changed

• Compatibility with Unbox API OpenAPI refactor.
• Models and datasets must be added to projects.
• Deprecates categorical_features_map in favor of categorical_feature_names for model and dataset uploads.
• Moved TaskType attribute from the Model level to the Project level. Creating a Project now requires specifying the TaskType.
• Removed name from add_dataset.
• Changed description to commit_message from add_dataset, add_dataframe and add_model.
• requirements_txt_file no longer optional for model uploads.
• NLP dataset character limit is now 1000 characters.

Fixed

• More comprehensive model and dataset upload validation.
• Bug with duplicate feature names for NLP datasets if uploading same dataset twice.
• Added protobuf==3.2.0 to requirements to fix bug with model deployment.

0.3.0a5

Changed

• NLP dataset character limit to 1000 characters.

Fixed

• Fix issue with duplicate feature names for NLP datasets.

v0.3.0a4

Fixed

• explainability_tokenizer validation.

v0.3.0a3

Added

• explainability_tokenizer as an optional argument to the add_model method, to be used by the explainability techniques.

Changed

• requirements_txt_file no longer optional for model uploads.
• Remove id from POST params to API server.

Fixed

• Added protobuf==3.2.0 to requirements to fix bug with model deployment.

v0.3.0a1

Changed

• Default Unbox server URL (https://api-staging.unbox.ai/).

v0.3.0a2

Fixed

• Fixed link to project page when loading / creating a project.
• Presigned url endpoint when using AWS / GCP / Azure.

Changed

• Removed links when uploading dataset and models. Just the project link is appropriate.

v0.3.0a0

Added

• A Project helper class.
• A convenience method create_or_load_project which loads in a project in if it is already created.

Changed

• Models and datasets must be added to projects.
• Deprecates categorical_features_map in favor of categorical_feature_names for model and dataset uploads.
• Moved TaskType attribute from the Model level to the Project level. Creating a Project now requires specifying the TaskType.
• Removed name from add_dataset.
• Changed description to commit_message from add_dataset, add_dataframe and add_model.

v0.2.0a1

Fixed

• Fail early if custom_model_code, dependent_dir or requirements_txt_file are None when model type is ModelType.custom.
• Fail early if model is not None when model type is ModelType.custom.