referentiel-evaluation-dsrc
referentiel-evaluation-dsrc copied to clipboard
Référentiel d'évaluation data science responsable et de confiance
Nouvel élément dans la section 6 ? Plusieurs points d'entrée pour réduire la taille d'un modèle : - Float16 au lieu de Float32 - Transformers fruguaux : [morphnet](https://arxiv.org/abs/1711.06798), [hypermixer](https://arxiv.org/abs/2203.03691) -...
[Do Foundation Model Providers Comply with the Draft EU AI Act?](https://crfm.stanford.edu/2023/06/15/eu-ai-act.html), Rishi Bommasani and Kevin Klyman and Daniel Zhang and Percy Liang (Stanford University, Center for Research on Foundation Models),...
Study [this Quantmetry's blog article](https://www.quantmetry.com/blog/ia-confiance-frugalite/) and see if it can help complement section 6
Enhance element 6.1 with more advanced requirements/practices, like: - beyond carbon footprint per project / model / system, measure and follow the aggregated carbon footprint of the data science activity...
See [Sustainable AI: Environmental Implications, Challenges and Opportunities](https://arxiv.org/abs/2111.00364)
See [Carbon Emissions and Large Neural Network Training](https://arxiv.org/ftp/arxiv/papers/2104/2104.10350.pdf)