cancel
Showing results for 
Search instead for 
Did you mean: 

Python API Scoring Data in AI Catalog

pvail@unitedfcu.com
Image Sensor

Python API Scoring Data in AI Catalog

I am trying to use the Python API to generate predictions on a project and pull those predictions into Python.  The project has three secondary datasets that are all located in the AI Catalog.  When the project was created, the secondary datasets went through the automated feature engineering.  How do I send the new datasets to the desired model in the project to generate the predictions using the Python API?

Labels (2)
0 Kudos
5 Replies
dalilaB
Data Scientist
Data Scientist

Here is the answer from my colleague :

Once you have your model deployed, predictions can be made from
Predictions tab, scoring dataset only needs to be uploaded either from AI Catalog or from Local file.

Screen Shot 2022-08-22 at 10.51.24 AM.png

Another way to make predictions is to use CLI, API Client or HTTP, all those options can be found on Prediction API tab.
Screen Shot 2022-08-22 at 10.52.49 AM.png

Secondary datasets configuration can be changed on the Settings tab.
Screen Shot 2022-08-22 at 10.53.59 AM.png

0 Kudos
pvail@unitedfcu.com
Image Sensor

I need to be able to kick off the scoring and retrieve the predictions from Python without deploying the individual model. Is there a way to do this without moving the model into deployment?

0 Kudos
taylor_larkin
Data Scientist
Data Scientist

Hi pvail@unitedfcu.com,

 

This should be doable by leveraging the secondary_datasets_config_id argument in the upload_dataset function and then subsequently requesting predictions on this dataset object.

 

Best,

 

Taylor

0 Kudos
pvail@unitedfcu.com
Image Sensor

Okay so where do I find the secondary_datasets_config_id in the platform for the project I'm working on?

0 Kudos
taylor_larkin
Data Scientist
Data Scientist

Hi pvail@unitedfcu.com,

 

Apologies for the delay!

 

You can list them for your project by using dr.SecondaryDatasetConfigurations.list(<insert project ID>). Documentation for this is here.

 

Best,

 

Taylor

0 Kudos