GitLab

Drawing fewshot from train_docs has two main advantages: a) you can do way more than 8 shots and b) your fewshot examples are from the same type of problem

Implement `LogiQA` data download and evaluation

Add Ethics Dataset

try to unbreak things

Implement `DROP` evaluation

# Conflicts:
#	lm_eval/tasks/__init__.py

Implement symbol manipulation based evaluations