fix argument, r = 24, 3 epoch

predict_with_generate=True, r = 24, 1 epoch (#10)

change compute metrics in argument, r = 24, 3 epochs (#9)

move question before context (#7)

not 8 bit, r = 39 (#6)

change evaluate (#5)

change evaluate (#4)

change evaluate (#3)

cpgQA dataset (#2)

cpgQA dataset (#1)