Create AI model assets

Create AI assets to track and manage the life cycles of your AI models.

始める前に

Role required: sn_ai_governance_ai_steward or sn_ai_asset_mgmt.ai_asset_owner

このタスクについて

An AI model is a program that is trained to process data and generate outputs, such as predictions, content, recommendations, and decisions, without human intervention. You can use an AI model to perform a specific task, such as image recognition, data classification, or price prediction.

手順

1. Navigate to Workspaces > AI Control Tower.
2. From the AI Control Tower, open the AI assets view.
3. From the navigation menu of the AI assets view, navigate to AI asset inventory > AI Models.
4. Select Add AI model.
5. In the Details section of the Add AI model form, fill in the fields.

   表 : 1. Details fields in the Add AI model form

   Field	Description
   Name	Name of the AI model.
   Provider	People or organization that developed the AI model.
   Version	Version of the AI model.
   Supported languages	Languages that are supported by the AI model.
   Deployment guidelines	Guidelines or instructions on how to deploy the AI model.
   Training procedure	Method that you’re using to train the AI model.
   Context window	Maximum number of tokens that the AI model can process and recall when generating outputs.
   Model size in MB	Size of the AI model in megabytes (MB).
   Model parameters info	Internal variables that the AI model learns during training to process data and generate outputs.
   Description	Brief description of the AI model.
   Managed by	User who is assigned to manage the AI model. This field is automatically set to the user who creates the AI model asset. 注: This field is editable only if you have the AI steward (sn_ai_governance_ai_steward) role. If you have the AI asset owner (sn_ai_asset_mgmt.ai_asset_owner) role, this field is read-only.
   State	State of the AI model. The options are Draft and Deployed.
   License details	License that you are using for the AI model.
   Base model	Base model, also known as a foundation model or pre-trained model, that the AI model was built from.
   Department	Department that the AI model belongs to.
   Supported locations	Locations in which the AI model is supported.
   Model card	Brief document that describes important information about the AI model, including the context, intended use, training data, and limitations of the model.
   Model weights info	Numerical parameters that define how the AI model learns during training so that it can generate more desired outputs.
   Required infrastructure	Integrated software and hardware components that are required for developing, deploying, and managing the AI model.
   Evaluation metrics report	Report that provides data for the metrics that you’re using to evaluate the effectiveness of the AI model.

6. Select Next.
7. In the Related assets section of the Add AI model form, specify the datasets and prompts that are associated with the model.

   • To specify a dataset that is associated with the AI model, search for and select the dataset from the Evaluation datasets field. You can specify more than one dataset.

   • To specify a prompt that is associated with the AI model, search for and select the prompt from the AI prompts field. You can specify more than one prompt.

     注:

     If an associated dataset or prompt isn’t available, you can add it by selecting Add new and then following the proceeding steps in Create dataset assets or Create prompt assets.
8. Select Submit for review.
   注:

   If you aren’t ready to submit the AI model for review, select Save instead. You can then edit it and submit it for review later by navigating to AI assets > Lifecycle > Onboard and then selecting the AI model from the list. When the AI asset record opens, select Edit to continue editing the AI model details.

タスクの結果

The AI model is added to your AI asset inventory. It automatically enters the onboard stage of the AI asset life cycle and is updated with a Lifecycle status of AI steward review.

次のタスク

Users who are assigned the AI steward (sn_ai_governance.ai_steward) role can start the review process to begin tracking and managing the life cycle of the AI model. For detailed instructions, see Complete AI asset lifecycle.