A LinkML schema for the Model Cards model as published in Model Cards for Model Reporting, which is an effort to democratize AI/ML technologies including increasing transparency. Model Cards represent information about trained machine learning models and can provide data on benchmarked model evaluations across a variety of conditions, including varying cultural, demographic, or phenotypic factors. Model Cards can also specify the applicable contexts for using the trained model and other relevant information.
This repository stores a LinkML schema representation for the original Model Cards model, representing the topics, sets of questions, and expectations about entities and fields expected in the answers (work in progress). HuggingFace provides a markdown Model Card template and their are JSON and Proto versions provided by the Model Card Toolkit.
The Google Model Card model is also supported by the Model Card Toolkit for creating Model Card instances for trained models Introducing the Model Card Toolkit for Easier Model Transparency Reporting. In addition HuggingFace has adopted Model Cards and has released a Model Cart Writing Tool, Amazon AWS also has support for Amazon SageMaker Model Cards including creation of Model Card instances using the SageMaker Python SDK, and Model Cards can also be generated via scikit-learn How to create and deploy a model card in the cloud with Scikit-Learn.
- examples/ - example data
- project/ - project files (do not edit these)
- src/ - source files (edit these)
- tests/ - Python tests
make all: make everythingmake deploy: deploys site
This project was made with linkml-project-cookiecutter.