Fixing bug in how evaluator sets training configurations

[deeptir18]

The evaluator didn't set the range of rtts to train over correctly, and didn't use the min in the range properly.