What is the best solution to send the prediction result via mqtt? I tried to use lwip, but is giving Panic out of memory issue. Thanks
Hi @fd18,
If you are running out of memory, you have a few options:
- Reduce the size of your neural network
- Try a different DSP method that uses less memory
- Use a different microcontroller with more memory
- Use a co-processor: have one microcontroller handle sensing/inference that passes results to another microcontroller that handles networking (e.g. WiFi/MQTT) events
@fd18
In addition to what Shawn stated:
- Use the EI Studio “Quantized (int8)” model not the " Unoptimized (float32)".
- If you can, switch to a different MQTT library that requires less memory or is better optimized for your particular use case.
- Breaking up the MQTT message into smaller chunks and sending them sequentially instead of all at once.
1 Like