High quality presets perform a post-fit step which takes 10~15% of total time. To autogluon still only uses roughly max_runtime for its entire fitting procedure, we reduce the max_runtime communicated to autogluon by 10% which means in most cases autogluon should produce a model within the given time constraint. This is to ensure comparisons across frameworks remain fair and interpretable.
High quality presets perform a post-fit step which takes 10~15% of total time. To autogluon still only uses roughly
max_runtime
for its entire fitting procedure, we reduce themax_runtime
communicated to autogluon by 10% which means in most cases autogluon should produce a model within the given time constraint. This is to ensure comparisons across frameworks remain fair and interpretable.