Forum Discussion

Anonymous's avatar
Anonymous
Not applicable
7 years ago

AutoML Data Flow refresh performance

Hi,

I've been using the AutoML feature in Data Flows and am having performance issues with the Data Flow refresh, I am using Azure A embedded tiers and have tried from A2 to A4 with the same performance issues. 

 

I have a simple dataset of 2 columns - Date and EventCount.  Number of rows is < 600.  A Regression model is selected and then it takes nearly 1 hour to do a refresh.  So it takes a long time to really do anything, as I then need to "Apply" model and it takes another 1 hour etc.

 

Has anyone else had performance issues?

 

Thanks

 

Andy

Need Help

2 Replies

  • Rajatt's avatar
    Rajatt
    Microsoft Employee
    7 years ago

    Hi Andy

     

    The training report should mention the number of iterations that were run as part of this on the last tab of the report. Can you share that? We do have significant improvements coming within next two weeks to improve this but looking at the iteration count, it might explain a bit

     

    Thanks

    Rajat

    • Anonymous's avatar
      Anonymous
      Not applicable
      7 years ago

      Hi Rajat,

       

      Yes, the details are: 

       

      • Input rows provided: 531
      • Sampled rows: 531
      • Training rows: 425
      • Final Model Used: Pre-fitted Soft Voting Regressor
      • Final model quality: 93%
      • Iterations run: 25

      Pre-fitted Soft Voting Regressor final parameters:

      • max_models: 15
      • model_seed_threshold: 0.05
      • min_models: 1

      Thanks

       

      Andy