Manual grading of Fortran 2 Cpp translation results

[chunhualiao]

@bin123apple Where do you share the pairs of before (Fortran) vs. after (C++) translation so @peihunglin can start initial manual grading of them ?

[bin123apple]

I am trying to modify the code framework and meeting with Dr.Ding now, I will submit some git patches for the code generation and upload the generated files tonight.

[bin123apple]

Just uploaded the translated results by llama3-8B-Instruct, gpt-4o and deepseek-coder-base-6.7B on two different datasets. It is under data folder. For the file names: FT means finetune; HE means humaneval; HPEC means the HPEC dataset; Clean means I cleaned the original model outputs and extract the code blocks.

[peihunglin]

Grading for two datasets: HumanEval(HE), and HPEC. Each dataset has 5 translated variants: gpt_4o, deepseek_coder_6.7B_base_FT, deepseek_coder_6.7B_base, llama3_8B_ins_FT, llama3_8B_ins

Place holder for the result: https://docs.google.com/spreadsheets/d/1AGT0r4IWOfms2RBwExHGaCcygPvShABGtr5dvZyjMs0/edit?usp=sharing