Re: Packages that work with LLMs

Thanks Björn,

On 2026-08-14 13:23, Björn A. Lindqvist wrote:
> On 8/14/26 07:28, Jean Louis wrote:
>> Apertus relies on "openly available data." Note how they didn’t 
>> actually
>> define "openly." It’s just a vibe.
>> 
>> Their logic is delightfully simple: if a website doesn’t explicitly 
>> tell
>> a bot to go away in its robots.txt, then the data is "openly 
>> available."
> 
> To clarify, they actually specify their data collection procedure in: 
> https://arxiv.org/pdf/2509.14233
> 
> So most of it is FineWeb-2, which is hosted on Hugging Face: 
> https://huggingface.co/datasets/HuggingFaceFW/fineweb-2 The dataset is 
> licensed under ODC-By 1.0 which allows you to "copy, distribute, and 
> modify" it.

Could you specify exact reference in the PDF where it says they use 
FineWeb-2?
And anything else, is it also used?

After the review of FineWeb-2, look here:
HuggingFaceFW/fineweb-2 · Datasets at Hugging Face:
https://huggingface.co/datasets/HuggingFaceFW/fineweb-2

Where it says:

Licensing Information

The dataset is released under the Open Data Commons Attribution License 
(ODC-By) v1.0 license. The use of this dataset is also subject to 
CommonCrawl's Terms of Use.

That means using the dataset is non-free, as when you review 
CommonCrawl's Terms of Use:
https://commoncrawl.org/terms-of-use

Short evaluation by 4 freedoms:

Freedom #1 - questionable, as they prohibit many things. "Other rights" 
reserved, etc. So there are many prohibitions.

Freedom #2 - unclear, it says data provided as-is, doesn't say 
modification allowed.

Freedom #3 - it says license not transferrable and other prohibitions

Freedom #4 - if you cannot redistribute it then you cannot distribute 
modified versions.

So no -- any LLM trained on FineWeb-2 doesn't comply with 4 free 
software freedoms.

Please try to prove me wrong, as my review may not be perfect.

-- 
Jean Louis