aiEngineer

evaluate_functional_correctness ${output_path}.jsonl 得到的pass@k结果全是1

1


### Please search before asking - [X] I searched in the [issues](https://github.com/yetone/openai-translator/issues) and found nothing similar. ### Please read README - [X] I have read the troubleshooting section in the...

Questions about the free-law data used in the paper "Adapt LLM to domains"

2

Dear Authors, you have undoubtedly done an excellent job (domain-specific post-pre-training). But I have a small question about the size of the free-law data used in the original paper, I...

ImportError: /lib64/libstdc++.so.6: version `GLIBCXX_3.4.26' not found (required by ...python3.9/site-packages/selective_scan_cuda.cpython-39-x86_64-linux-gnu.so)

7

Hello, if I don’t have the permissions to update GLIBC under the root directory, how can I solve this problem?

What is the use of params.n_chunks？

3

Hello I was wondering about params.n_chunks. What does the n_chunks variable do, and why does it have anything to do with the length of the sequence

Question regarding the difference between llama-pro and the regular llama.（关于llama-pro和普通llama之间的区别的疑问）

8

作者您好，毫无疑问，你们在llm的增量学习方面上做了一项优秀的工作。以下的llama均为llama2 但我对llama-pro有一点疑问。llama-pro是在llama的基础上添加了8层 identity block，同时进行了通用语料的全参数训练，所以我可以认为此时的8层 identity block已经退化为与其余32层类似的block，llama-pro也就成为了8.3B的普通的llama。 在后面的代码和数学预训练中，只不过是冻结了llama-8.3b的32层block，只微调8层block（当然这8个block的位置还是添加时候的位置） 所以，作者应该设置一个实验，将llama7b中的block冻结24层，微调8层（顺序上是每冻结3层block而开放一层block），这样就能证明是否有必要添加8层新的块进行微调。 Hello Author, undoubtedly, you have done excellent work in the incremental learning of LLM. Below, all "llama" refer to "llama2". But I...