Aqui contém o necessário para executar o tutorial apresentado no ESSE 2021.
torch==1.5.0
torchvision==0.6.0
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
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