kaneplusplus / ioregression

An R package for out-of-core regressions
14 stars 3 forks source link

Increase the number of lines read by default by chunk readers #23

Closed kaneplusplus closed 9 years ago

statsmaths commented 9 years ago

I suggested this after noticing that bigger chunks processed faster than the default. However, this seems to have been a result of the strange deparsing bottle fixed in version 0.3. We can discuss what the optimal block size is at a later stage, but I think this is not something that now needs to be changed after that performance bug was fixed.