v1.11.7

Improvements

• Added CodeFile class to pzmm module for creating and uploading Python code files to SAS Intelligent Decisioning
  • New method write_id_code_file() uploads a Python code file to a specified Viya folder and registers it with the Decisions service
    • Accepts code as a raw string, file path, or Path object
    • Validates code format via the SAS Viya API before upload; can be disabled with validate_code=False
    • Raises ValueError if the file already exists in the target folder, if the folder is not found, or if validation fails
    • Cleans up the uploaded file if Decisions service registration fails
  • See examples/pzmm_id_code_file_example.ipynb for usage examples

v1.11.6

Improvements

• Added create_requirements_txt parameter to create_requirements_json() function in write_json_files.py to optionally generate a requirements.txt file alongside the requirements.json file.

v1.11.5

Improvements

• Added model versioning methods to model_repository.py to handle model version endpoints.
• Allow for user to set custom timeout length for score testing in score_model_with_cas.

v1.11.4

Improvements

• Improved upload_local_model to allow for SAS Model Manager to properly intake local ASTORE models.

v1.11.3

Improvements

• Added upload_local_model to tasks.py, which can be used to upload local directories to SAS Model Manager without any file generation.

v1.11.2

Bugfixes

• Updated calculate_model_statistics function in write_json_files.py to improve ROC tables as well as model card files.

v1.11.1

Improvements

• Improved functionality for score_definition.py and score_execution.py, allowing for more general usage of the score_model_with_cas task
  • Also allowed for model name to be passed into functions instead of requiring model UUID
• Pickle files now loaded with pd.read_pickle() instead of pickle.load() in score code, allowing for more flexibility between python environments

Bugfixes

• Updated pzmm_generate_complete_model_card.ipynb to have better preprocessing function

v1.11.0

Changes

• Added score_definition.py and score_execution.py to allow for score testing within SAS Model Manager
  • Included optional use of CAS Gateway for faster scoring. Only available in environments where Gateway scoring is properly set up.
• Added ability to include data pre-processing function within python score code using the preprocess_function argument.

Bugfixes

• Fixed issue where settings file was improperly imported in some score code files.

v1.10.7

Bugfixes

• Fixed a bug that caused an error when performing SSL verification without a CA bundle specified.

v1.10.6

Improvements

• Refactor tasks.py to utilize sasctl.pzmm functions.
• Add model_info class to better capture model information.