Cannot cast value of type 'array' to type 'record'

I am trying to integrate Azure Stream Analytics with Azure Machine Learning and getting the following error: I am getting the following error: Cannot cast value of type 'array' to type 'record' in expression 'inputArray'. At line '7' and column '18'. TRY_CAST function can be used to handle values with unexpected type.

• steps: I have deployed an azureml model in aks and linked the model to Azure Stream Analytics

the model input was defined as the following:

input_sample = np.array([[34.13,34.13,34.13,34.13,3.48,34.13,34.13,34.13,34.13,34.13,34.13]])
@input_schema('data', NumpyParameterType(input_sample))

in Azure Stream Analytics I created an input array

function createArray(v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11) {
    'use strict';
    var array = [v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11]
    return array;
}

and when I am trying to invoke the Azure Machine Learning UDF using the following query:

WITH 
ModelInput AS (
SELECT udf.createArray(cast(v1 as float), cast(v2 as float), cast(v3 as float), cast(v4 as float), cast(v5 as float), cast(v6 as float), cast(v7 as float), cast(v8 as float), cast(v9 as float), cast(v10 as float), cast(v11 as float)) as inputArray
FROM input
) 

SELECT udf.score(inputArray)
INTO output
FROM ModelInput
WHERE inputArray is not null

I am getting the following error: Cannot cast value of type 'array' to type 'record' in expression 'inputArray'. At line '7' and column '18'. TRY_CAST function can be used to handle values with unexpected type.

[sidramadoss]

When sending one input array as part of http request to the endpoint, it returns the right result:

The problem is with how your endpoint handles multiple inputs part of the same input. If 2 arrays are sent as part of the same http request, the expectation is that there are equal number of results in the response from endpoint. I tried sending 2 arrays but got only 1 result which doesn’t make sense.

It should’ve returned: [ 5.0, 5.0 ]

[mouhannadali]

@sidramadoss thanks for the clarification. ASA and its logic are new for me. I was thinking it should support that to predict something you might need some history.