UBOdin / mimir

Data-ish exploration through SQL+Uncertainty
http://mimirdb.info
Apache License 2.0
26 stars 13 forks source link

Interpolation Model impossibly slow #357

Open okennedy opened 4 years ago

okennedy commented 4 years ago

View definitions give Spark a headache. These should be getting universally compiled out by Mimir before the query gets handed over to spark. It looks like the Interpolation model's training process doesn't do so.

See e93e2e68aef240c0018139c16b84c1fca8a91fe1 for a discussion of how this problem was solved for the Classifier model.

okennedy commented 4 years ago

@mrb24 has worked on this and improved the situation