Skip to main content

CLI Guide

This package does not add a top-level command. MLIA selects its transformers when the input model and downstream backend require a PyTorch conversion route.

PT2 to TOSA​

Install this package alongside a target plugin whose backend consumes TOSA. For example, Neural Technology compatibility analysis can start from a .pt2 model:

mlia check my_model.pt2 \
  --target-profile neural-technology \
  --compatibility

MLIA selects pt2_to_tosa when the target workflow requires TOSA input.

PT2 to PTE​

Ethos-U Corstone workflows can use pt2_to_pte to prepare an ExecuTorch program:

mlia check my_model.pt2 \
  --target-profile ethos-u55-256 \
  --performance \
  --backend corstone-300

The target profile supplies the ExecuTorch configuration required by the transformer.

Debug transformer selection​

Run the same command with --debug to inspect transformer and backend selection:

mlia check my_model.pt2 \
  --target-profile neural-technology \
  --compatibility \
  --debug

Use mlia target list and mlia backend list to confirm that the required target and backend plugins are installed.