There are some common training failures that we can guess the cause of. It would be nice to display our suggested fixes to users (and in some cases link to documentation). Right now we show some additional information in the error popup when it shows. It would be nice to show this information more prominently inline, and if appropriate, offer suggestions for fixing.
Training ran out of time
Related to the above, we show a whole special error message if you ran out of training time AND you have a small data set.
Incorrect data type selected. For example, column is a string but user chooses "Single." We can tell them what is the problem column and to check the type - Bri filed here
Parsing failure. Prose thinks the column is type boolean based on the data it saw. This is wrong and user needs to change data type (Xiaoyun will filed). Could also be caused by actually bad data in which case we should tell user to open their data and fix it.
On Azure training we advise you to check the azure portal for more details.
Column type is wrong for prediction (e.g. Regression scenario and you choose string)
There are some common training failures that we can guess the cause of. It would be nice to display our suggested fixes to users (and in some cases link to documentation). Right now we show some additional information in the error popup when it shows. It would be nice to show this information more prominently inline, and if appropriate, offer suggestions for fixing.