There was a native memory leak from IlpSolver as the constructed ILP object wasn't properly freed up in the C++ land without an explicit call to ilpsolver.free(). That's fixed now and the native memory footprint seems stable (does not grow with the number of questions, for 1000+ questions). In addition:
LD_LIBRARY_PATH is now set as part of runVis.sh.
Switched to a non-SNAPSHOT version of the entailment library.
There was a native memory leak from
IlpSolver
as the constructed ILP object wasn't properly freed up in the C++ land without an explicit call toilpsolver.free()
. That's fixed now and the native memory footprint seems stable (does not grow with the number of questions, for 1000+ questions). In addition:LD_LIBRARY_PATH
is now set as part ofrunVis.sh
.