Lightning Talk: Flexible Deployment of PyTorch Models on MCU-Class... Robert Kalmar & Martin Pavella

This talk covers the efficient deployment of PyTorch models on MCU-class devices using ExecuTorch, a framework designed for edge applications. The speaker highlights the flexibility of ExecuTorch in supporting various hardware platforms, from low-power microcontrollers to more robust processors. By examining acceleration options through CPU, DSP, and NPU paths with backends such as Cortex-M, Cadence, Ethos-U, and eIQ Neutron, the session provides valuable insights into optimizing machine learning workflows. Attendees will also learn best practices for reducing power consumption, which is crucial for running ML workloads in energy-limited environments.