Skip to content

schema_description.md

Latest commit

 

History

History
52 lines (40 loc) · 15.6 KB

schema_description.md

File metadata and controls

52 lines (40 loc) · 15.6 KB

Model Card Schema Descriptions

Model Card Documentation

This documentation details the required and optional fields for configuring a ModelCard. The variables specified here represent essential information about the model, its usage, and its metadata. Fields like bias_analysis, xai_analysis, id, and model_requirements are managed automatically by the system and do not require user input.

Field Name Type Description Constraints/Example Values
name str Required. The name of the model card. Example: "Model Card for Sentiment Analysis".
version str Required. The version of the model card. Example: "1.0.0".
short_description str Required. A brief description of the model card. Example: "This model detects sentiment in text data."
full_description str Required. A comprehensive description of the model card. Example: "This model performs sentiment analysis using NLP techniques and is trained on customer feedback data."
keywords str Required. Keywords associated with the model card. Example: "sentiment analysis, NLP, text classification".
author str Required. Author or creator of the model card. Example: "Jane Doe".
input_type str Required. Type of input data accepted by the model. Example: "text, image, tabular".
category str Required. Model category from predefined types. Enum: "classification", "regression", "clustering", "anomaly detection", "dimensionality reduction", "reinforcement learning", "natural language processing", "computer vision", "recommendation systems", "time series forecasting", "graph learning", "graph neural networks", "generative modeling", "transfer learning", "self-supervised learning", "semi-supervised learning", "unsupervised learning", "causal inference", "multi-task learning", "metric learning", "density estimation", "multi-label classification", "ranking", "structured prediction", "neural architecture search", "sequence modeling", "embedding learning", "other"
input_data str Required. URL or DOI for the input data. Must be a valid URL, DOI, or empty string.
output_data str Required. URL or DOI for the output data. Must be a valid URL, DOI, or empty string.
Optional Fields
foundational_model str Optional. Foundational model ID or URL if applicable. Example: "https://huggingface.co/docs/transformers/en/model_doc/flan-t5".
documentation str Optional. URL link to documentation for the model. Example: "https://docs.example.com/modeldocs.md".

AI Model Documentation

This documentation details the fields required to configure an AIModel, representing the essential metadata for the underlying AI model within a ModelCard.

Field Name Type Description Constraints/Example Values
name str Required. The name of the AI model. Example: "Sentiment Analysis Model".
version str Required. The version of the AI model. Example: "1.0.0".
description str Required. A description of the AI model’s functionality and purpose. Example: "This model is trained to detect sentiment in customer feedback."
owner str Required. The individual or organization that owns the model. Example: "Jane Doe".
location str Required. Downloadable URL of the model. Example: "https://modelrepository.com/model/123".
license str Required. License type for the model usage. Example: "Apache-2.0".
framework str Required. Framework used for developing the model. Enum: "sklearn", "tensorflow", "pytorch", "other".
model_type str Required. Type of AI model structure. Enum: "cnn", "decision_tree", "dnn", "rnn", "svm", "kmeans", "llm", "random_forest", "lstm", "gnn", "other"
test_accuracy float Required. Accuracy of the model on test data. Example: 0.89.
Optional Fields
model_structure object Optional. Structure of the model in JSON format. Example: {"layers": [{"type": "Conv2D", "filters": 32}]}.
metrics dict Optional. Dictionary of performance metrics for the model. Use ai_model.add_metric(key, value) to add metrics. Example: ai_model.add_metric("Test loss", loss)

You can validate the model card contents against this schema to ensure that all required fields are present using the following command:

mc.validate()