search

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
792 stars 130 forks source link

visual wake word (VWW) model end to end workflow? #8

Closed ramkumarkoppu closed 1 year ago

ramkumarkoppu commented 2 years ago

Hi,

Where do I find jupyter notebook of given example visual wake word (VWW) model in the tutorial folder if I want to check it's model architecture and how it is optimized for MCU (pruned, quantized, model conversion (how C/C++ code is generated from the model)) to follow end to end workflow?

meenchen commented 2 years ago

Hi @ramkumarkoppu, thanks for your interest in our work. For the model optimization, please refer to our mcunet repo. We also have a model zoo where you can download the pre-trained models and check their model architecture. Then, you can follow these examples to generate C/C++ code.

ramkumarkoppu commented 2 years ago

Hi @meenchen, Do we have corresponding jupyter notebook for vww model? if yes, where do I find it.

tonylins commented 2 years ago

Hi, currently we do not have a jupyter notebook for the example. But I think the provided code works the same.

ramkumarkoppu commented 2 years ago

Hi, I am interested to understand how fp32 PyTorch model is pruned, quantised before converting into C code. Where do I find the relevant code in the repo?

meenchen commented 1 year ago

Close due to inactivity. Feel free to reopen.