MLIA for Arm Mali with Neural Technology

Purpose

mlia-neural-technology registers target and backend capabilities for supported Arm Mali GPUs with Neural Technology in MLIA. These docs cover target profiles, model estimation, measured profiling data, output interpretation, and troubleshooting.

Supported plugins

Compatible converter plugins for upstream model formats include mlia-converters-litert (for LiteRT / TensorFlow Lite .tflite models) and mlia-converters-pytorch.

Start here: Arm Learning Path

Estimate neural accelerator (NX) performance with ML Inference Advisor

Follow this hands-on guide to install the plugin, analyze quantized MobileNet models, check operator compatibility, and interpret NX Performance Estimator reports.

Documentation Map

• Target profiles
• Backends and conversion flow
• Outputs and metrics
• CLI examples
• ETRecord and ETDump profiling
• Download the Python API walkthrough notebook
• Troubleshooting
• Development

For shared CLI, installation, and plugin architecture guidance, see the MLIA core documentation.