Deployment

In the development section, you can search for your model based on its ID or the filters:

like framework, model and history , ...

In the deployment phase, there are two versions available for development:

1. Legacy Deployment:

Legacy Deployment offers traditional deployment methods for AI models. Users can deploy their trained models using this approach. The deployment process involves packaging the model along with its dependencies and deploying it on the desired infrastructure.

2. Deployment Model (Triton-based):

Deployment Model utilizes Triton, an advanced model serving platform developed by NVIDIA. Triton provides high-performance inference with support for various hardware accelerators. It offers features such as dynamic batching, concurrent model execution, and multi-model serving.

Features:

• Model Training: Train your AI models using custom datasets and algorithms.

• Legacy Deployment: Deploy trained models using traditional methods.

• Triton-based Deployment: Deploy models on Triton for optimized performance.

• Scalability: Scale deployments based on demand and resource availability.

• Monitoring: Monitor model performance, resource utilization, and server health.

• API Integration: Integrate the service with your applications via RESTful APIs.

In box number 2, you can access your trained models by specifying filters.

In box number 3 and 4 You can view the information and specifications of the model.

In box number 5 You can obtain output and inference from the desired model.

By pressing "MORE," you can start development and obtain the API address or download the model.