NLU Model training creates a new record in table [ml_solution] for the trained NLU Model solution throwing an error and unable to publish or test - Known Error

NLU Model training creates a new record in table [ml_solution] for the trained NLU Model solution throwing an error and unable to publish or test Description When training an NLU Model on a Prediction Server that has been impacted by a JVM out-of-memory issue, the NLU training will not be successful and you will get a red cross at the top of the screen without an error message displayed during the NLU Model training.

However the Prediction Server will still respond with a " Success ", creating a record in table [ml_solution] for the trained NLU Model solution, which is inactive with 0% progress , but with state set to " Solution Complete ", causing further training issues on the same NLU Model.

The symptoms will be:

1. User is unable to train model, having to try again after some time. 2. When using the NLU Model Builder within Studio, it will not be possible to train the NLU Model, as it keeps spinning and the page might appear to freeze. 3. The Publish button is not available and greyed out.

Steps to Reproduce

This issue only occurs when training an NLU Model on a Prediction Server that has been impacted by a JVM out-of-memory (OOM) issue during the training.

However, the associated problem in this KB is to ensure that when the NLU Model training fails, it creates the record in table [ml_solution] for the NLU Model training with the correct state of " Solution Incomplete " or " Solution Failed ".

Workaround This is expected behaviour. To resolve the issue, delete the duplicate NLU Model Solution records with the same Version number in table [ml_solution], as follows:

1. Go to ml_solution.list 2. Check all records for the same Solution Name 3. Determine the duplicate records with the same Version number 4. Delete the duplicate record that has active=false OR where the Progress field is empty OR if neither of the first two conditions apply then delete the oldest of the two duplicate version records in table [ml_solution] 5. Retrain/publish the NLU model

Repeat the above steps if the Publish button continues to be greyed out, until the Version number for this Solution Name is the highest Version number in this table.