Getting Started

This section covers everything an Arm AI Model Optimizer user needs, from core concepts and quick start through feature references, tutorials, and both the CLI and Python SDK documentation.

Documents

Document	Purpose
Concepts	Core concepts: workspace hierarchy, experiments, jobs, and the optimization pipeline.
Quick Start	Get your first model optimized in a few commands.
Python Quick Start	The same first optimization, driven from Python instead of the shell.
Tutorial 1: Sample Project	Validate, run, and inspect the sample experiment created at workspace init.
Tutorial 2: CV Model	End-to-end walkthrough building a CV project from scratch.
Tutorial 3: LLM	End-to-end walkthrough optimizing a language model with AQ + GO + GQ.
Tutorial 4: BYOM / BYOD	Bring your own model and dataset.
Features	Reference for each Arm AI Model Optimizer feature: graph extractor, graph optimizer, quantizer, evaluator, profiler, visualizer, and advanced quantizer.
On-Device Profiling	Measure latency and memory on physical target devices using ExecuTorch and Ethos-U85.
Supported Devices	Supported target hardware and runtimes.
Supported Models	Pre-validated model list with benchmark results.
CLI Reference	Full CLI command reference.
Python SDK Reference	Full Python reference: all 31 interfaces, 1:1 with the CLI.
Run YAML	Run configuration file specification.
Run Detail YAML	Per-job run detail YAML specification.
BYOM & BYOD Formats	Supported data formats for bringing your own model and dataset.
Troubleshooting	Diagnosis flow and log reference for common user-facing issues.

Recommended Reading Order

1. Read Concepts for the mental model.
2. Follow Quick Start to run your first optimization — or Python Quick Start if you are driving it from code.
3. Work through the Tutorials for a hands-on, end-to-end example.
4. Explore Features for the capabilities you need.
5. Use CLI Reference or Python SDK Reference, and Run YAML, as day-to-day references.
6. Refer to Troubleshooting when something goes wrong.