Model card: Difference between revisions
| (9 intermediate revisions by the same user not shown) | |||
| Line 1: | Line 1: | ||
== Model card == | |||
As models begin to pile up in open repositories like [[Hugging Face]], model cards have emerged as a privileged means to document them.[1] Think about model cards as nutrition labels for models. Ideally, they list the model's ingredients, how it was trained and its validation procedures as well as its intended use and limitations. Whilst code repositories cannot force their use upon the users, they automatically create an empty model card when a new models is uploaded in an effort to encourage standardization and transparency.[2] However, examining how model cards are redacted, one can see that their content varies a lot. Sometimes, they thoroughly document the model with a reference to an academic paper, sometimes they offer only minimal information or are simply empty. In that, model cards testify to the diverse nature of model providers. Some are working in computer science labs or in companies, others are amateurs with sometimes little time left for the tedious work of documentation or simply no desire to share widely their production. Finally, an empty model cards doesn't necessarily mean absence of documentation. Users may find more appealing to document them in another form. In [[CivitAI]], a platform where Manga fans share their models and LoRAs, each model is introduced with a succinct description written in a more affective tone where the author explains their goal, cracks a joke, begs for a tip on their Patreon and thanks their network of collaborators as well as the models and resources they are building on. | |||
[1] Liang, Weixin, Nazneen Rajani, Xinyu Yang, et al. “What’s Documented in AI? Systematic Analysis of 32K AI Model Cards.” 2024. <nowiki>https://arxiv.org/abs/2402.05160</nowiki>. | |||
[2] Mitchell, Margaret, Simone Wu, Andrew Zaldivar, et al. “Model Cards for Model Reporting.” ''Proceedings of the Conference on Fairness, Accountability, and Transparency'', FAT* ’19, ACM, January 2019, 220–29. <nowiki>https://doi.org/10.1145/3287560.3287596</nowiki>. | |||
== | == Examples == | ||
= | [[File:Screenshot 2025-05-23 at 12-36-05 openfree flux-chatgpt-ghibli-lora · Hugging Face.png|thumb|left|alt=A model card made by a community of manga users|A model card made by a community of manga users]] | ||
[[File:Screenshot 2025-05-23 at 12-38-33 black-forest-labs FLUX.1-Depth-dev-lora · Hugging Face.png|thumb|left|alt=A model card made by the company Black Forest Labs|A model card made by the company Black Forest Labs]] | |||
[[File:Screenshot 2025-05-23 at 12-36-34 dnad244 wan random loras · Hugging Face.png|thumb|left|alt=An empty model card|An empty model card]] | |||
* | |||
[[Category:Objects of Interest and Necessity]] | [[Category:Objects of Interest and Necessity]] | ||
Latest revision as of 18:25, 22 August 2025
Model card
As models begin to pile up in open repositories like Hugging Face, model cards have emerged as a privileged means to document them.[1] Think about model cards as nutrition labels for models. Ideally, they list the model's ingredients, how it was trained and its validation procedures as well as its intended use and limitations. Whilst code repositories cannot force their use upon the users, they automatically create an empty model card when a new models is uploaded in an effort to encourage standardization and transparency.[2] However, examining how model cards are redacted, one can see that their content varies a lot. Sometimes, they thoroughly document the model with a reference to an academic paper, sometimes they offer only minimal information or are simply empty. In that, model cards testify to the diverse nature of model providers. Some are working in computer science labs or in companies, others are amateurs with sometimes little time left for the tedious work of documentation or simply no desire to share widely their production. Finally, an empty model cards doesn't necessarily mean absence of documentation. Users may find more appealing to document them in another form. In CivitAI, a platform where Manga fans share their models and LoRAs, each model is introduced with a succinct description written in a more affective tone where the author explains their goal, cracks a joke, begs for a tip on their Patreon and thanks their network of collaborators as well as the models and resources they are building on.
[1] Liang, Weixin, Nazneen Rajani, Xinyu Yang, et al. “What’s Documented in AI? Systematic Analysis of 32K AI Model Cards.” 2024. https://arxiv.org/abs/2402.05160.
[2] Mitchell, Margaret, Simone Wu, Andrew Zaldivar, et al. “Model Cards for Model Reporting.” Proceedings of the Conference on Fairness, Accountability, and Transparency, FAT* ’19, ACM, January 2019, 220–29. https://doi.org/10.1145/3287560.3287596.
Examples
