Added two new arguments to the evaluation script: `--fewshot-as-multiturn` and `--apply-chat-template` which correspond to lm_eval options of similar names and are very often used to ensure apples-to-apples comparisons of lm_evaluation results
occurence -> occurrence
mlx_lm.evaluate
* Add evaluation script * only write top level results * add lm eval version * typo * create output dir * relative import * comment --------- Co-authored-by: David Grangier <dgrangier@users.noreply.github.com>
