aie4ml is an end-to-end compiler that generates optimized AIE firmware automatically, which can be then built and simulated directly using AMD Vitis. Currently, it is developed as a plugin that extends the backends for hls4ml in order to target the AMD AI Engine (AIE).

• Current support: dense (linear) layers with optional bias & ReLU. Support for AIE-ML/AIE-MLv2 devices.

• AMD Vitis 2025.2 and a valid AIE tools license.
• Python 3.10+ and the latest version of hls4ml package.

Operates on the intermediate model representation produced by hls4ml, therefore independent of the frontend (i.e., PyTorch, QKeras, etc.).

pip install hls4ml pip install aie4ml

Documentation and usage: https://github.com/dimdano/aie4ml

Tutorial 1 : tutorials/tutorial_1.ipynb Tutorial 2 : tutorials/tutorial_2.ipynb

General hls4ml concepts: https://fastmachinelearning.org/hls4ml

aie4ml is developed and maintained by Dimitrios Danopoulos.

If aie4ml contributes to your research, please cite the corresponding arXiv preprint:

@misc{danopoulos2025aie4mlendtoendframeworkcompiling, title={AIE4ML: An End-to-End Framework for Compiling Neural Networks for the Next Generation of AMD AI Engines}, author={Dimitrios Danopoulos and Enrico Lupi and Chang Sun and Sebastian Dittmeier and Michael Kagan and Vladimir Loncar and Maurizio Pierini}, year={2025}, eprint={2512.15946}, archivePrefix={arXiv}, primaryClass={cs.LG}, url={https://arxiv.org/abs/2512.15946}, }