[TRANSLATION] Initial version for vietnamese

[ngxson]

WIP, currently all files are generated by scripts/vi-translation.py without any review.

TODO:

• @xrsrke and I will review it
• After that @burtenshaw makes a final approval then we can merge

[HuggingFaceDocBuilderDev]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

[ngxson]

Hmm ok interestingly it uses vietnamese grammatically better than me lol

[ngxson]

Maybe provide a translated version of this whole code block

[ngxson]

Added the translation inside a <details> collapsible block

[burtenshaw]

Thanks @ngxson !

Do you think it's possible to parameterise the scripts/vi-translation.py so. that we could use it for other languages? For example, by adding a param for --language="vi" then select the prompt. This would mean that other translators just need to convert the prompt.

[ngxson]

Yup I generalized the script, user can create a new python file for each language, for example fr.py, then import theauto_translate function:

from translation import auto_translate

output_lang = "fr"

prompt = lambda content: f'''
The prompt here........

=== BEGIN OF TEXT ===
{content}
=== END OF TEXT ===
'''.strip()

auto_translate(
    prompt=prompt,
    output_lang=output_lang,
)

[xrsrke]

maybe not upper case the second word?

[ngxson]

Yeah maybe I did a Ctrl+F replace all and forgot about this. Thanks for reminding!

[ngxson]

Hmm ok seems like not because of that, but the LLM just copy from my example in the prompt

[burtenshaw]

Hey @xrsrke and @ngxson 👋

Where are we on this?

[ngxson]

Yeah sorry, I think we can merge if @xrsrke is ok. Resolving the conflicts now

For unit 2, I'll make a follow up PR

[honghanhh]

Coucou! Big thanks for translating the course into Vietnamese 🙌 ! As a native, I’ve got a few small tweaks to make it even smoother. Overall, great job—just a couple of adjustments and it'll be perfect! Keep up the awesome work!

[honghanhh]

Full stops aren’t a must, but they do keep things tidy 👀!
Either all has full stop or not would ease the eyes of OCD guys a lot.

[ngxson]

yeah I would say full stop is redundant here, should be removed to save some tokens.

[honghanhh]

Either capitalize first letter inside brackets or not for consistency maybe 👀
Why Thought-Action-Observation has no translation (e.g., Suy nghĩ - Hành động - Quan sát)

[honghanhh]

hiện đại nhất, tân tiến nhất, etc. might better describe the meaning of this word. Wdyt?

[ngxson]

yes that could be better. I tend to use short words because the text always get longer when translated to vietnamese; "hiện đại nhất" seems fine in this case

[honghanhh]

I think embedding models, server can be translated into Vietnamese as well.

[ngxson]

server: no because it's a far too common word in vietnam haha

embedding: I'm not sure how to translate this word without losing its meaning. Many words are non-translatable, so I include an explanation the first time it appears in the document.

[honghanhh]

model can be translated into Vietnamese (mô hình) as well.

[ngxson]

I'm not sure if it worth translating this word, as it's also very common among vietnamese tech people

[honghanhh]

model is kinda very standard already but for translation, mô hình still something original for vietnamese for all files

[honghanhh]

Suy nghĩ → Hành động → Quan sát

[honghanhh]

LLM (Mô hình ngôn ngữ lớn)

[honghanhh]

Large Language Model = mô hình ngôn ngữ lớn
Các mô hình này thường có hàng triệu tham số.

[honghanhh]

Mô hình ngôn ngữ lớn (LLM)

[burtenshaw]

Hey Son, could you take at look at @honghanhh 's comments?

[burtenshaw]

This looks good to me!

@honghanhh Thanks for your improvements. Could you please take these up in a new PR? That way, you'll get authorship.

[ngxson]

Hi @honghanhh and thanks for taking time to review this. Sorry I don't have much time this week, so I can't process all of your suggestions.

Would you mind opening a dedicated PR? That will be easier for me to review. Thanks!