/ESSE-tutorial

Tutorial of how to embed quantized deep learning models into Raspberry PI 3 using Pytorch.

Primary LanguageJupyter Notebook

ESSE-tutorial

Aqui contém o necessário para executar o tutorial apresentado no ESSE 2021.

Requisitos

torch==1.5.0
torchvision==0.6.0

Uso

usage: benchmarking_in_rpi.py [-h] [--path-to-float-model PATH_TO_FLOAT_MODEL]
                              [--path-to-postquant PATH_TO_POSTQUANT]
                              [--path-to-qat PATH_TO_QAT] [--log LOG]

optional arguments:
  -h, --help            show this help message and exit
  --path-to-float-model PATH_TO_FLOAT_MODEL
                        Path where is stored the full precision model
  --path-to-postquant PATH_TO_POSTQUANT
                        Path where is stored the Post training quantized model
  --path-to-qat PATH_TO_QAT
                        Path where is stored the Quantization Aware quantized
                        model

Uso com docker

Para rodar o jupyter notebook com docker é necessário:

  • Instalar jupyter lab
  • Instalar docker e docker-compose
  • executar bash source build.sh && source run.sh
  • Utilizar o notebook da mesma forma q no colab