Closed KEggensperger closed 4 years ago
I think for our case it is fine to have the data always in memory. We do the same in all other benchmarks.
Additionally, this avoids remaining files after killing auto-sklearn using the runsolver.
I changed this. Auto-sklearn now can make use of a tmpdir if available. @mfeurer please check again
I think for our case it is fine to have the data always in memory. We do the same in all other benchmarks.
Additionally, this avoids remaining files after killing auto-sklearn using the runsolver.