Closed eddyxu closed 2 years ago
The current implementation of ML_PREDICT requires three different JIT functions pre/post-processing and inference function. So currently it shows in Rikai that, for a query like
SELECT ML_PREDICT(ssd, image) FROM ...
The column is shown as
|-- ssd_dpjlkw_post(ssd_dpjlkw(ssd_dpjlkw_pre(image))): array (nullable = true)
It would be better that ML_PREDICT 's default column name to be just col or the model name , i.e., ssd(image)
col
ssd(image)
Closed
The current implementation of ML_PREDICT requires three different JIT functions pre/post-processing and inference function. So currently it shows in Rikai that, for a query like
The column is shown as
It would be better that ML_PREDICT 's default column name to be just
col
or the model name , i.e.,ssd(image)