On Linux platforms without a GPU or neural accelerator your model is ran using TensorFlow Lite. Not every model can be represented using native TensorFlow Lite operators; and for these models ‘Flex’ ops are injected in the model. We have now shipped support for these models in the Studio, and on Edge Impulse for Linux on armv7, aarch64 and x86 (and on macOS). See here for more info: Flex delegates - Edge Impulse Documentation