Hi, I have noticed that some part of the experiments mentioned in the paper are not currently included in the repository. In particular:
-
code generation experiment done on the HumanEval dataset, also could the calculation of winrate with GPT4 as baseline for both chat generation and code generation experiments.
-
the openllm leaderboard experiments for mitigating forgetting and reducing the alignment task, I could only find the gsm8k task in the repo.
would it be possible to share the code used for these part of the work ? Would be really helpful for anyone trying to replicate the results and build upon your research. Thanks again for making your work and code available !
Hi, I have noticed that some part of the experiments mentioned in the paper are not currently included in the repository. In particular:
code generation experiment done on the HumanEval dataset, also could the calculation of winrate with GPT4 as baseline for both chat generation and code generation experiments.
the openllm leaderboard experiments for mitigating forgetting and reducing the alignment task, I could only find the gsm8k task in the repo.
would it be possible to share the code used for these part of the work ? Would be really helpful for anyone trying to replicate the results and build upon your research. Thanks again for making your work and code available !