The Triton Model Navigator provides a run command to perform the step-by-step process of:
- Conversion - converts input model to the formats optimized for inference
- Triton Model Configuration - creates a model entry in the Triton Model Repository, including framework-specific backend accelerations
- Profiling - profiles model's performance using Triton Model Analyzer
- Analysis:
- analyzes models using Triton Model Analyzer
- selects top N configurations within given constraints and objectives
- generates a summary report for selected configurations
- Helm Charts - generates Helm Charts for selected configurations
The Triton Model Navigator allows running processes as a single command or performs step execution running each command separately.