Skip to main content

Create Dedicated Endpoint

Create Dedicated Endpoint​

Click "Deploy" button on model page, select "Dedicated Endpoint" from the drop-down menu and jump to the creation page.

Note: Not all models support creating dedicated instances. If you want to use a model that does not have the "dedicated instance" option, please contact us.

Create Endpoint

New Instance Page​

Open the dedicated instance creation page, fill in the following basic information and click "Create Instance" to create. New Page

  • Instance Name: Custom, cannot be repeated
  • ModelID: Select the model ID from the list of available models on the csghub community
  • Region/Resource: Select the compute resource to deploy dedicated instance
  • RunTime Framework: Select inference framework, supporting specifying framework version and setting runtime parameters
  • PD Separation: Supports enabling PD (Prefill/Decode) separation for models larger than 100B
  • Environment Variables: Supports adding custom environment variables for specific runtime configurations
  • Security Level
    • Public:Accessible to anyone without authentication
    • Private: Accessible only with authentication
  • Replicate: Number of instances to replicate
    • Range: 1~5 (default is 1)

Dedicated Endpoint List​

After the creation is complete, you can view all dedicated instances on the profile page.