To use Teradata Scoring SDK, follow these steps.
-
Train the model and load the data on Teradata framework.
Load any additional tables (such as dictionary or rules for text analytics) on ML Engine or install them as files on ML Engine.
-
Use the AMLGenerator to create the .aml file.
In the AMLGenerator function call, specify the model (from the previous step) and relevant information for the Teradata Scoring SDK function.
-
Transfer the .aml file to the real-time system:
-
Download the .aml file from the Teradata framework, using this command:
call pm.download_afile ('aml_model_file.aml')
The file is saved in the default storage directory on Teradata MPP nodes (Default: /etc/opt/teradata/tdconfig/Teradata/tdbs_udf/usr/). - Upload the file to the external system, working in real time, using any standard ssh/scp client tool.
-
Download the .aml file from the Teradata framework, using this command:
- Use the Scorer API to score input requests.