Catalog of valid arguments in model cards?

Greetings, good people
I am researching the structure of the model card template and wonder what argument can each slot accept as input
For example: in the field MODEL SOURCE, I can assument the choice of argument is paper/repo/demo *taken from the info in the field description , but in other fields such as for example MODEL TYPE or TRAINING PROCEDURE etc there arent any clues. just wondering. thanks, pointers appreciated

1 Like

I think anything is acceptable as long as it does not violate the YAML format. Whether or not it will be interpreted in a special way depends on the implementation of the Hub. For example, this.

Thank you John, things is I dont speak much code so I really stick with model cards. if I understand you correctlyl, any litteral string , Like in this example? How to represent strings in YAML Is there a way to look up the list of arguments already used in HF? *so that I may derive a model from that?

1 Like

In that case, it would probably be safer to use HfApi for crawling. When I searched for tags before, there were more than I expected…

Thank you John
any chance of a tutorial?

1 Like

I got it like this. Nothing special. :sweat_smile:

from huggingface_hub import HfApi
from collections import Counter

def get_tags():
    api = HfApi()
    tags_list = []
    spaces = api.list_spaces(full=True, limit=10)
    for i in spaces:
        if i.tags is not None: tags_list.extend(i.tags)
    models = api.list_models(full=True, cardData=True, limit=10)
    for i in models:
        if i.tags is not None: tags_list.extend(i.tags)
    datasets = api.list_datasets(full=True, limit=10)
    for i in datasets:
        if i.tags is not None: tags_list.extend(i.tags)

    tags_dict = dict(Counter(tags_list))
    print(tags_dict)

get_tags()