Skip to main content

Create Deployment

Upload your own models, or use existing fine-tuned models with full configuration control.

Step 1: Select a model​

Choose a model to deploy from one of the following sources and click Create Deployment to start the process.

Optimized Top-Tier Models
Explore open-source models with advanced inference acceleration optimized by Smart Studio.
  • Browse pre-configured models for faster response times
  • Deploy models with optimized configurations out of the box
  • Leverage inference acceleration for production workloads
Fine-Tuned Models
Use a model that has been optimized based on a base model.
  • Leverage your optimized model for specialized use cases
  • Deliver faster and more accurate performance without starting from zero
  • Deploy models trained on your custom datasets
My Models
Upload and manage your own full model weights.
  • Deploy models after object storage and weight validation are complete
  • Reuse one validated model asset across multiple deployments
  • Manage model versions and track deployment history
Go to My Models

Configuration options​

  • Model: Select a model from the dropdown. Use the search box to find specific models.
  • Display Name: Enter a complete deployment name of up to 128 characters. It may contain letters, numbers, slashes, dots, hyphens, and underscores, and must start and end with a letter or number.
  • Deployment Type: Select the Deployment Type (Standard or P/D Disaggregated).

Configuration Options

Step 2: Configure resources​

Select the appropriate compute resources and deployment settings for your model.

Resource configuration​

  • Target Cluster: Select a target cluster for deployment.
  • GPU Type: Select a GPU type based on your model's requirements. You must select a target cluster first to view available GPU options.
  • Replicas / P/D: Number of deployment replicas or P/D instances. Default is 1, 1p1d.
Deployment Types
  • Standard Deployment: Deploy with a fixed number of replicas.
  • P/D Deployment: Deploy with Prefill/Decode disaggregated architecture for optimized inference performance.

Resource Configuration

Step 3: Review and deploy​

Review your configuration before creating the deployment. Click Create Deployment to start the deployment process. The system validates the configuration and provisions the required resources on the selected cluster.

Manage deployments​

After creating a deployment, you can manage it from the deployment list. Search or filter deployments by status, view details, edit configurations, or stop and restart deployments.


Deployment-List


Deployment details​

The deployment details page displays deployment information and provides code examples for calling the deployed model API in Python, TypeScript, Java, and Shell (cURL).


Deployment-Details


Next steps​

Monitor Performance

Track service metrics, cluster resources, and API usage in real time.

Learn More
Manage API Keys

Create and manage credentials for deployed model endpoints.

Learn More