Closed SamruddhiMhatre closed 1 year ago
The error message seems quite clear to me. One of your task either has only 1s or 0s (on training or validation) so you can't compute the current metric which is logloss. As it's quite useless to monitor a binary score on a single value I would advise you to try to make a split that makes it work.
Same issue here...
How this was solved? Thanks!
The error message seems quite clear to me. One of your task either has only 1s or 0s (on training or validation) so you can't compute the current metric which is logloss. As it's quite useless to monitor a binary score on a single value I would advise you to try to make a split that makes it work.
Is it possible to oversample the data for multiple target variables? I was wondering if there's an efficient way to do so inorder to prepare the data for TabNetMultiTask Classifier?
I was trying to run the Kaggle notebook provided in the example link for TabNetMultiTaskClassifier in this repository. However, while fitting the model using the exact same code in the notebook, I get the following Value error::
**--------------------------------------------------------------------------- ValueError Traceback (most recent call last)