I am new to Tensorflow Lite for Microcontrollers and looking into loading a 2.5MB model into the 8MB PSRAM portion of my microcontroller (ESP32-S3). My idea was to have this .tflite file stored into the SPIFFS filesystem and having it be loaded from there. My understanding is that tflite::getModel requires the hex dump as a C Array, but converting my model to .cc would result it a much larger file which cannot be loaded in binary. My initial thinking is to have this .tflite file be parsed and converted. A C Array in this case would be initialized to around ~2.5MB is PSRAM and I would feed the parsed data into this C Array and then tflite::getModel would then have a sufficient input to load the model.
So my questions are:
Does this look like a sound method? Other possible methods?
Possible issues I could encounter?
Looking to bounce around some ideas and get some possible insights. Thanks! :)