feat: add func to generate multiple quries

[e7217]

description
Currently, the autorag generate one query per corpus during the QA generation stage. However, to more thoroughly validate performance, I wanted to explore the possibility of generating multiple query candidates. To this end, I have experimentally introduced a feature called multiple_queries_gen, which includes a parameter n to specify how many queries to generate. I have confirmed that it produces satisfactory output on my local machine.

If there was an intention behind not adding this feature, please let me know. Additionally, I acknowledge that the code I added may not be clean, so if there are any optimization suggestions, I would be happy to implement it.

Please review it at your convenience. Thank you.

qa = initial_corpus.sample(random_single_hop, n=len(initial_corpus.data), random_state=random.randint(1,100)).map(
        lambda df: df.reset_index(drop=True),
    ).make_retrieval_gt_contents().batch_apply(
        multiple_queries_gen,  # query generation
        llm=llm,
        lang="ko",
        n=3,
    ).batch_apply(
        make_basic_gen_gt,  # answer generation (basic)
        llm=llm,
        lang="ko",
    )....

sample
before:

after:

etc

• add .vscode/ to .gitignore

[vkehfdl1]

@e7217 Hi!

Thanks for the contribution.

I like the idea, but if we implement this feature, I want that feature can be used in the most of the query gen functions.
I think it will be okay to add this feature to llama_index_generate_base directly!

[e7217]

@vkehfdl1

Thank you for checking.

I agree that improving it to be usable in most query functions is a good idea. However, at this stage, I wanted to take a more conservative approach to minimize the impact on the existing code you've already implemented. Would it be okay to do the additional work in the next phase?

[vkehfdl1]

@e7217
Okay I agree. I will merge this issue after some adjustments. (for Linter and formatting)

And just do not close the issue in this PR.
Thank you!

[vkehfdl1]

Great:)
Thanks for waiting my review

[e7217]

close #1006