Our project is built around ARM Cortex-A7 SoC with no operating system. We expected to run TensorFlow lite for microcontrolles (tflm) on this hardware with models that we had previously tuned on the TensorFlow lite version (tflite). However, it turned out that some of the tflite models are not suitable for tflm. In addition, those models that we can managed to run take too long execution time, most likely due to tflm does not use the hardware capabilities of Cortex-A7 (NEON) and also does not use the NNPACK library (maybe we are wrong?).
Our idea is to run tflite on our system without an operating system (bare metal). Is that possible? And what do you recommend to us?
Sincerely,
Dmitry
Supported platforms. TensorFlow Lite for Microcontrollers is written in C++ 11 and requires a 32-bit platform. It has been tested extensively with many processors based on the Arm Cortex-M Series architecture, and has been ported to other architectures including ESP32. The framework is available as an Arduino library.
It sounds like you’re facing compatibility and performance issues running TensorFlow Lite for microcontrollers (TFLM) on your ARM Cortex-A7 SoC. Running TensorFlow Lite (TFLite) directly on your hardware without an operating system is technically possible, but it might require significant effort to adapt and optimize for your specific setup. As for recommendations, you could explore optimizing TFLM for your hardware, considering alternatives like running TFLite on a lightweight real-time operating system (RTOS), or even exploring other frameworks or approaches better suited for your requirements. Good luck, Dmitry!