Unable to locate Inference Output Data on SageMaker Model Augmentation

Hi all, hope you guys can help.

Reading the documentation provided by aws here: Integrating Amazon SageMaker models with Amazon QuickSight - Amazon QuickSight stating that The output of the inference is stored in SPICE and appended to the dataset. However, I could not find such output in my dataset that augmented to SageMaker Model and visualized in QuickSight. Checking the dataset at QuickSight indicated the output column data are “unavailable”

Does anyone know where I can search for inference output in SPICE? Thank you.

Hi @wylinks05 ,

Please choose ‘Refresh Now’ to manually refresh the data set, or choose ‘Schedule refresh’ to set up a regular refresh interval. During each data refresh, the system automatically runs the SageMaker batch transform job to update the output fields with new data.

Please let us know if the issue persists even after a refresh.

Hi @rajjaya,

Yes the data refresh works fine and I am able to make inference at my visualization using table chart. However, checking the dataset imported to SPICE (QuickSight → Datasets → Edit Dataset), I am still unable to look for the updates to the output fields at my dataset, the output fields are still showing “unavailable” for all the data in that column. The issue persists.

Currently my QuickSight edition is Enterprise.

Hi @wylinks05 ,

The output fields in data preparation show as “unavailable” since this is an inference which will get updated after the SageMaker batch transform job executes.

Example ( Survived is the output value I want my model to predict )

As you have written, once data refresh is complete, the inference values are visible in the analysis (not in the data preparation ) .

Analysis ( Sagemaker batch transform job executed and SPICE data refresh completed ) .

Kind Regards,