CODEWITHSUNDEEP
javarestazureazure-cognitive-services
dernasherbrezon
dernasherbrezon

Reputation: 887

Unable to trigger recommendation build

I'm sending the following request (according to the docs https://westus.dev.cognitive.microsoft.com/docs/services/Recommendations.V4.0/operations/56f30d77eda5650db055a3d0 )

{  
   "description":"",
   "buildType":"Recommendation",
   "buildParameters":{  
      "recommendation":{  
         "numberOfModelIterations":10,
         "numberOfModelDimensions":20,
         "itemCutOffLowerBound":1,
         "itemCutOffUpperBound":10,
         "userCutOffLowerBound":1,
         "userCutOffUpperBound":10,
         "enableModelingInsights":false,
         "splitterStrategy":"LastEventSplitter",
         "useFeaturesInModel":false,
         "allowColdItemPlacement":false,
         "enableFeatureCorrelation":false,
         "enableU2I":false,
         "randomSplitterParameters":{  
            "testPercent":10,
            "randomSeed":0
         }
      },
      "fbt":{  
         "similarityFunction":"",
         "enableModelingInsights":false,
         "splitterStrategy":"",
         "randomSplitterParameters":{  

         }
      }
   }
}

And constantly getting response:

{  
   "error":{  
      "code":"InternalServerError",
      "message":"(EXT-0110) The server encountered an internal error.",
      "innerError":{  
         "code":"c83f18d5-06ff-4e34-b0f9-b0c680b4c66e",
         "message":null
      }
   }
}

Googling didn't help much.

UPD:

Changed enableModelingInsights to false. Issue still exists.

If I trigger once again, I will receive "Cannot start new build of type 'Recommendation' when another build '1560455' of same type is running for model '264ea280-76dd-4d17-892e-2ca2bde214c3'". Looks like build was created, but somehow wasn't triggered.

Upvotes: 1

Views: 123

Answers (3)

dernasherbrezon
dernasherbrezon

Reputation: 887

I haven't changed anything. But now it works. Looks like it was internal issue.

Upvotes: 0

abhi1509
abhi1509

Reputation: 11

When you specify "buildType":"Recommendation", you are specifying that you want recommendation build (available are recommendation/fbt/rank)

In this case you will need to remove the FBT parameters from the whole list.

When you do want to use FBT, you will need to change the build type, remote recommendation build parameters and fill in the values for similarityFunction and update values for splitter if enable modelling insights is set to true else remove them.

Upvotes: 0

Luis Eduardo Cabrera
Luis Eduardo Cabrera

Reputation: 71

I think I know the issue. Turn off enableModelingInsights to false for now.

Modeling insights is not going to work unless you have sufficient data for modeling, and specify a proper splitting strategy.

(More of modeling insights at https://azure.microsoft.com/en-us/documentation/articles/cognitive-services-recommendations-buildtypes/)

I did notice that the defaults shown in the API documentation caused this confusion -- so I will talk to the team to have that corrected.

Thanks, Luis Cabrera

Upvotes: 1

Related Questions