Configuring Models

Notes and Constraints

• You must have developer or administrator permissions to configure models.
• You must have administrator permissions to manage models.

Prerequisites

• A project has been created by referring to Creating a Project.

• A data source has been connected by referring to Connecting to a Data Source.
• A dataset has been created by referring to Creating a Dataset.
• An intelligent analysis assistant has been created by referring to Creating an Intelligent Analysis Assistant.

Procedure

1. Log in to the DataArts Insight console.
2. Click in the upper left corner of the management console to select a region. Then, select an enterprise project in the upper right corner.
3. On the top menu of the console, click Project. On the displayed My Projects page, click the name of the desired project.
4. In the navigation pane on the left, choose Q&A Management > Intelligent Analysis Assistants.
5. Locate your desired assistant and click Edit in its Operation column. On the displayed page, click the Assistant Settings tab (Figure 1).
   Figure 1 Model configuration
   

6. Set NL2SQL and intelligent insights parameters based on Table 1.
7. Click Save.

   Table 1 Parameters for configuring a model

   Parameter	Description
   Model	Select the model used by NL2SQL or intelligent insights. The models in the dropdown menu are those created in model management.
   temperature	Controls the diversity and creativity of generated text. A value close to 0 indicates the lowest randomness, while 1 indicates the highest randomness. Generally, a lower temperature is suitable for deterministic tasks, while a higher temperature, such as 0.9, is suitable for creative tasks.
   max_tokens	Maximum number of tokens for the generated text. The total length of the input text plus the generated text cannot exceed the maximum length that the model can process.

Managing Models

1. Log in to the DataArts Insight console.
2. Click Models on the top menu of the console.
3. Click Create Model Service. Models of various interface types can be connected. For details about the parameters, see Table 2.
   Figure 2 Creating a model
   

   Table 2 Parameters for creating a model service

   Parameter		Description	Example Value
   Name		Model name. The name can contain up to 64 characters and cannot start or end with a space. Only letters, numbers, spaces, underscores (_), and hyphens (-) are allowed.	Pangu model test
   Model Type		Model type. Pangu Large Model (Cloud Base): large models purchased from and deployed by Pangu Models. Open-source Model (Ascend cloud): an open source large model deployed by Ascend Cloud. Open-source Large Model (Bare Metal): an open source large model deployed on a custom bare metal server. Pangu Large Model (Bare Metal): Pangu model deployed on a bare metal server.	Pangu Large Model (Cloud Base)
   Project		Project space with permission to use the model. If left blank, it indicates that all spaces have permission to use the model.	test
   Task Type		Task type that can be applied to the current access model. Task types: NL2SQL, Intelligent insights	NL2SQL, Intelligent insights
   Network Type		Model network access method. Public network: The model address must be a public network domain name or IP address and can be directly accessed. VPC endpoint service: VPCEP is used to connect DataArts Insight to the large model. This service is mainly used when the large model is deployed on a bare metal server.	Public network
   Access Address		Address for accessing the large model's Q&A page.	https://test.com/v1/infers/gpt/query?endpoint=infer-app
   Auth Type	IAM authentication	Tenant name: Huawei Cloud tenant name, that is, the master account name. Username: username with the large model access permission. Password: user login password. Project name: Huawei Cloud project name.	Tenant name: domain_name Username: sub_user_name Password: password Project name: cn-north-4
   	AK/SK authentication	AK: access key ID. SK: secret access key. Project name: Huawei Cloud project name.	AK: test SK: test Project name: cn-north-4
   	AppCode authentication	AppCode: App code of the application in the API Gateway access scenario.	yourAppCode
   	Agency authentication	Account name: name of the delegating account. Agency name: Name of the agency created by the delegating party. Project name: Huawei Cloud project name.	Account name: domain_name Agency name: agency_name Project name: cn-north-4
   	Custom authentication	Header key: request header key. Header value: request header value.	Header key: X-Test-Header Header value: testHeader

4. After setting the parameters, click Connectivity Test. If the test is successful, click Confirm. The model is successfully created.
   

   You can set the default model from the Default NL2SQL Model or Intelligent Insights Default Model drop-down list. After the configuration is complete, the new intelligent analysis assistant NL2SQL and intelligent insights will automatically use the selected default model.