Update example scenario to take in num-eval-batches

[lcadalzo]

The example scenario should be updated to use the --num-eval-batches flag. The _evaluate function here should also take in a num_eval_batches arg and pass that to the call to load_dataset(), as is done here for example.