Easy to use commandline tool for "Bring Your Own Edge AI Models" (BYOM) - to work with TI Deep Learning (TIDL) for TI MPU Processors for Edge AI.
- More details about TIDL and other tools are explained in Edge AI developer landing page for MPUs and edgeai-tidl-tools - it is highly recommended to read these.
There are two packages in this repository with different purposes.
- tidlrunner: Tool for Model Compilation, Inference, Analysis or Benchmark.
- tidloptimizer: Experimental tool for advanced model optimization (This is an experimental feature and recommend users not to use it as of now).
Use these scripts based on where and how you run:
- PC CPU setup:
./setup_runner_pc.sh - PC GPU setup:
./setup_runner_pc_gpu.sh - EVM setup:
./setup_runner_evm.sh - Optional dataset extras:
./setup_runner_extra_pc.sh
Please go through the documentation of tidlrunner to get started with this repository.
The core documentation for TIDL software is available in the edgeai-tidl-tools repository. Please refer there for detailed guidance on model compilation and inference arguments, supported operators, versioning, debugging tips and more.
Frequently asked questions and their solutions are in the tidlrunner FAQ or in the edgeai-tidl-tools FAQ.
- 2026-July-29: r11.2.1 branch is being released - with tidl_tools version updated to 11_02_16_00. For models compiled with this to work in SDK 11.2, this may need firmware update as described in SDK Version Compatibility. configs for models in edgeai-modelzoo has been added in modelzoo configs
- 2026-April-04: Model Inspector is an interactive HTML visualization tool for analyzing ONNX models compiled with TIDL. Model Inspector provides comprehensive insights into model structure, performance, accuracy, and hardware acceleration. A modelinspector.html file is generated when you invoke compile, infer or inspect commands in tidlrunner.