Is anyone tried to train a model instead of using the model in the example files?

mit-han-lab / tinyengine

[NeurIPS 2020] MCUNet: Tiny Deep Learning on IoT Devices; [NeurIPS 2021] MCUNetV2: Memory-Efficient Patch-based Inference for Tiny Deep Learning; [NeurIPS 2022] MCUNetV3: On-Device Training Under 256KB Memory

https://mcunet.mit.edu

MIT License

757 stars 127 forks source link

Is anyone tried to train a model instead of using the model in the example files? #90

Open VisionShilin opened 1 year ago

VisionShilin commented 1 year ago

Is anyone tried to train a model instead of using the model in the example files such as "49kb-int8-graph.json" and "full-int8-params.pkl" ?

I cann't find any document about how to generate files such as "49kb-int8-graph.json" and "full-int8-params.pkl".

TfliteConvertor.py looks like can convert a tflite model into a TinyEngine IR format. Is there any document or example code about this?