This incident has been resolved.

We have implemented a fix to address the model errors and will be monitoring performance over the next several hours.

We are investigating increased instances of model errors for some users when attempting to generate Omni impressions. We will provide updates as we work towards a resolution as soon as possible.