Overwritting the default metric parameter

Ray Libraries (Data, Train, Tune, Serve) Ray Tune
1

Hi i am new to raytune , and while trying some example i got stuck at a point

I know that we have to pass a “metric” and “mode” for comparing the results of various trials while hyperparameter tuning process.

I want to compare various trials based on AUROC metric based on “max” mode ,
can you please help me to overwrite the “_metric/val_loss” with AUROC, (i don’t know what modification i have to make for that). Thank you !!

Here i am attaching the colab notebook for your reference

Google colab notebook link

2

Hey @RIPUVENDRA_SINGH, thanks for the question. I’m not familiar with pytorch_widedeep, but I took a look at their example and it looks like in their call to analysis.results all their metrics are being tracked, but in your notebook it looks like it’s missing auroc is missing from _metrics.

  1. Perhaps as a first step you can look into seeing why auro isn’t being calculated/returned in the metrics.
  2. After this, you should be able to change metric="_metric/val_loss" to metric="_metric/val_auroc" (or whatever the produced metric name is).
1 Like