allow languages other than nature languages

[ZhiyuanChen]

Is your feature request related to a problem? Please describe.

Recently, the huggingface hub enforce a new policy that the language specified in the README must be compatible with one of the ISO 639-1, 639-2 or 639-3 code.

However, several models (like ProGen2) is designed for Protein, and the amino acid code is not in the ISO.

Describe the solution you'd like

add dna, rna and protein as special values like code

Describe alternatives you've considered

use warning instead of raise error for unmatched language

Additional context

There are also models that take chemical molecules as inputs.

[Wauplin]

Hi @ZhiyuanChen, it would be better suited to set dna/rna/protein as tags instead of language, as we prefer not to add new special values like code. Would that be ok for you?

[ZhiyuanChen]

Hi @ZhiyuanChen, it would be better suited to set dna/rna/protein as tags instead of language, as we prefer not to add new special values like code. Would that be ok for you?

We do use tag to discriminate specific language (e.g. mRNA, ncRNA, etc.) for now.
But I think it's more natural to use language as "DNA is the language of life".

There have been hundreds of models on HF with these languages. I know this is rather a small number, but this field is also quite young.