Giters

oeg-upm / inesdata-ml-schema

A repository to keep the vocabulary for datasets and models developed in Inesdata project

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Questions: how to use model card

dgarijo opened this issue · comments

commented

We received the following questions, which we should clarify in the doc:

 If some fields are missing information, is it necessary to fill them
> in anyway?
> - For the 'Identifier' field, should it link to a Hugging Face URL?
> - Is it acceptable to insert line breaks when writing in sections like
> 'evaluationResults' or 'usageInstructions'?
> - In 'evaluationResults', you have used numeric values. Is it okay if
> I represent these as percentages instead?
> - For 'trainedOn', should identifiers be DOIs, or can they be any URL?
> For example, my datasets are derived from two evaluation tasks. Should
> these be included in the URL or as an identifier?
> - Regarding 'sdLicense', I have already added the identifier above.
> What is the difference between this and the identifier field? Should
> they be the same?
> - How can I calculate 'hasCO2eEmissions'?
> - In 'parameterSize', could we add some formulas in the description?

Answers:

  • For id, it may link to HuggingFace, yes.
  • It is acceptable to insert linebreaks if you use > in the begining
  • It is ok for now to represent percentages in eval results
  • For trainedOn, it is ok for the dataset to be a DOI, if that resolves to a dataset, not a paper. The identifier is the DOI, for the URL you may use url
  • sdLicense is the license for the model card. license is the license of the model. See documentation for more info.
  • For calculating CO2 Emissions you may use an existing calculator. See https://calculator.linkeddata.es/. Open it in Chrome. The calculation usually is Power consumption x Time x Carbon Produced Based on the Local Power Grid
  • parameterSize is just an estimate on the number of parameters, we do not expect formulas to be there