Machine Learning Inference Advisor (MLIA)
From model to informed decisions.
Explore what runs, what performs, and what to improve. MLIA connects your model to the insights that guide deployment.
View on GitHub (opens in a new tab)Inside the analysis
One model. Three perspectives.
Discover your environment
mlia backend list
mlia target list
mlia check --help
Available checks depend on your installed plugins and backends.
Check compatibility
See whether your model uses operations supported by your chosen Arm target.
Run a compatibility check 02Estimate performance
Estimate how your model may perform on a supported target before deployment.
Explore analysis backends 03Turn advice into action
Review the results and, where available, MLIA advice to choose what to improve next.
Read results and adviceYour first analysis
Open a terminal. Start exploring.
Install the plugins for your model and target, discover the available backends, then run your first analysis.
The MLIA ecosystem
Connect your toolkit.
The core provides shared workflows. Choose plugins for your targets, converters, and visualization.
Mali with Neural Technology
Target profiles and analysis for supported Arm Mali GPUs with Neural Technology.
Read the guide (opens in a new tab)TargetArm Ethos-U
Analyze models with Vela and Corstone backends.
Read the guide (opens in a new tab)ConverterLiteRT
Prepare TensorFlow Lite models for supported analysis flows.
Read the guide (opens in a new tab)ConverterPyTorch
Explore PyTorch, PTE, and delegate conversion paths.
Read the guide (opens in a new tab)VisualizationModel Explorer
See model structure and analysis results together.
Read the guide (opens in a new tab)Under the hood
Follow the analysis. Build what’s next.
Explore the architecture, follow an analysis, or build your own plugin.