Lightning Talk: Running ExecuTorch Applications With Silicon Accelera... George Gekov & Aki Makkonen

This talk explores the efficient deployment of machine learning models on low-power embedded systems, addressing a significant challenge in today's technology landscape. The speakers, George Gekov from Arm and Aki Makkonen from Alif Semiconductor, demonstrate how to construct an end-to-end speech recognition application using PyTorch and ExecuTorch, a framework designed for resource-constrained environments. They cover the training of a Transformer-based neural network in PyTorch, along with the quantization process and deployment on the Alif Ensemble E8 SoC. The presentation highlights key ExecuTorch APIs that optimize models for efficient execution on devices, showcasing the integration of PyTorch with silicon-accelerated systems.