How NetsPresso Works: Overview

NetsPresso is designed as a modular AI model optimization workflow, divided into three core stages.
Each stage handles a specific part of the model lifecycle—from training to deployment on real hardware.

Workflow Overview

Stage	Description
Stage 1: Model Training	Build or compress models using modules such as Model Zoo, Trainer, and Compressor.
Stage 2: Model Optimization	Convert models into hardware-aware formats via Quantizer, IR Converter, and Graph Optimizer.
Stage 3: Model Testing	Validate optimized models using Simulator, Profiler, and Device Farm.

Access Methods

NetsPresso offers both a flexible Python SDK and visual GUI tools.

Interface	Modules Supported
PyNetsPresso (Python SDK)	Every modules
Training Studio	Model zoo, Trainer, Compressor
Optimization Studio	IR Converter, Quantizer, Graph Optimizer
Benchmark Studio	Simulator, Profiler, Device Farm

Note: All modules are workflow-compatible—outputs from one stage can seamlessly feed into the next.

Getting Started

Want a deep dive? Explore Stage 1: Model Development
If you're a developer: Start with Using PyNetsPresso
If you prefer visual tools: Go to Using Studio