question

hunga-1270 avatar image
0 Votes"
hunga-1270 asked hunga-1270 commented

AzureML Designer - batch execution endpoint

Hi,

I am migrating something from classic Azure ML to the current Azure ML and I have been using the batch execution endpoint. . I know there are examples how to publish a real-time endpoint through the designer, but is it possible to publish a batch execution endpoint (which is what I have been using in classic Azure ML) through the designer interface?

I found this documentation online: https://docs.microsoft.com/en-us/azure/machine-learning/classic/migrate-rebuild-web-service
My pipeline don't have a 'trained model' though it only includes a single Execute Python script module, is there a way to get around that and publish a batch inference endpoint

azure-machine-learning
5 |1600 characters needed characters left characters exceeded

Up to 10 attachments (including images) can be used with a maximum of 3.0 MiB each and 30.0 MiB total.

1 Answer

romungi-MSFT avatar image
0 Votes"
romungi-MSFT answered hunga-1270 commented

@hunga-1270 Yes, in this case you would need a trained model to publish a pipeline. You could try the scenario to use a create python model module and then use your script to create an experiment and run. If it is successful I think you should be able to create a pipeline.


· 2
5 |1600 characters needed characters left characters exceeded

Up to 10 attachments (including images) can be used with a maximum of 3.0 MiB each and 30.0 MiB total.

@hunga-1270 Did you get a chance to try the scenario with the create python model module?

0 Votes 0 ·

I tried and that publish does not work for me and I got this error:
'Cannot create inference because there is no model on this pipeline' After I added some empty Train Model and Score Model modules, then the publish works. However, my Python Script module was not included in the published pipeline.

0 Votes 0 ·