AI asset data model attributes

Additional attributes for the AI asset data model.

Attributes

AI model product model: Product Information for the AI model that is used by the AI system to generate responses without human intervention (cmdb_ai_model_product_model).

Attribute	Description
Model parameters info	Number of parameters for the model.
Supported languages	Languages supported.
Model size	Size of the model in MB. Mostly applicable for models developed and deployed within the organization.
Deployment guidelines	Instructions applicable for models developed and deployed within the organization.
Source	Links or details of source of the model sources example: Hugging face, Microsoft, and so on.
Training procedure	Types of training- Decision Trees - Deep Neural Networks - Linear Regression - Logistic Regression - Random Forest - Supervised Learning - Unsupervised Learning - Reinforcement Learning - Transfer Learning - Semi-Supervised Learning - Instruction Finetuning - Supervised Finetuning
Context window	Size of input sequences that the model can handle \(number of tokens\).

AI dataset product model: Product Information for the collection of data that is used to train and test AI models (cmdb_ai_dataset_product_model).

Attribute	Description
Data type	Describes data, example: Text, Image, Video, and Table
Source	Links or details of source of the dataset sources, example: Customer, Wikipedia, Hugging face, Crowd sourced, and so on.
Acceptable usage	Acceptable usage of the data according to license / contract example: Training, and Evaluation.

AI prompt product model: Product Information for instructions given to AI models to get a response for AI system (cmdb_ai_dataset_product_model).

Attribute	Description
Documentation	Links and information about requirements, design, and related information.

AI system product model: Product Information for software that provides ML / AI capability to generate outputs, such as decisions, recommendations, content, or predictions (cmdb_ai_system_product_model).

Attribute	Description
Documentation	Links and information about requirements, design, and related information.

Attribute	Description
Data classification	Classification according to organization's data classification model, example: Public, confidential, and customer confidential.

Attribute	Description
ServiceNow® record reference	Reference to Now Assist record.
ServiceNow® table	Now Assist table.

Attribute	Description
AI models	Reference to more than one associated models.
Evaluation Dataset	Reference to more than one associated datasets used for evaluation.
Evaluation Metrics Report	Details of evaluation results.

Attribute	Description
Base model	This AI model version was derived from an internal model developed within the organization.
Model weights info	Additional model information if available. Mostly applicable for models developed within the organization.
Required infrastructure	Documentation of infrastructure requirements for model deployment, primarily for models deployed within an organization. Example: Infrastructure stack and processing requirements.
Training dataset	Reference to one or more associated datasets used for training the model. These datasets are mostly applicable for models developed within the organization.
Evaluation dataset	Reference to one or more associated datasets used for evaluating the model. These datatsets are mostly applicable for models developed within the organization.
Evaluation metrics report	Links or details of evaluation results
License details	Link or detail to applicable licenses applied to the model.
Model card	Links to shareable model card (Internal and external model card).

Attribute	Description
Base datasets	This version of the AI dataset was derived from the previous version.
Dataset card	Information on number of records, distribution, and so on.Documentation for data quality and known risks and limitations.
License details	Link or detail to applicable licenses applied to the dataset, example: CommonCore, Apache 2.0,etc.

Attribute	Description
Prompt information	Details of the prompt.
AI model	Reference to the AI model for which the prompt is created.