Open yaochitc opened 5 years ago
Could you please provide a bit more detail on your use case? In which case you've found the row by row prediction slow? Thank you!
I'm using mleap with a deep learning based ranking model developed by bigdl, I need to score hunderds of items for each user. If the score process for one user is executed in a single batch, the prediction time is much shorter.
Seems it is not supported currently. As predict row by row is slow, I just wonder if this feature will be added.