GitLab

do per character loss aggregation for multiple choice tasks (similar to OAI's per token aggregation)

add cloze variant of lambada task

(or rather, move it to the data download phase)

Now, if a run gets interrupted halfway, you can easily resume

Clean up `Ethics` few-shot prompts

fix translation scoring format

Implement SQuADv2 evaluation

# Conflicts:
#	lm_eval/tasks/squad.py

Add `Random` argument to `write_out` `fewshot_context`

Remove unused imports and format imports