model-deployment-patterns

What problem does it solve?

This Skill provides battle-tested patterns and configurations for deploying machine learning models efficiently and reliably into production environments.

Core Features & Use Cases

• Production Serving: Implement robust FastAPI services with health checks and metrics.
• Triton Deployment: Configure model repositories for NVIDIA Triton Inference Server.
• ONNX Export & Validation: Convert PyTorch models to ONNX and verify numerical equivalence.
• Canary Deployments: Manage traffic splitting for phased rollouts using Kubernetes and Istio.
• Load Testing: Use Locust to simulate production traffic and validate performance.

Quick Start

Use the model-deployment-patterns skill to export a PyTorch model to ONNX format.