Closed jbrownkramer closed 2 years ago
I am super interested ! (i don't get why the google colab tutorials provided are CPU :( still great though !)
Hey @jbrownkramer, could you please share the tutorial? I'm currently trying to create it myself, without success.
I would also be interested! @jbrownkramer did you share the docker or tutorial somewhere?
I don't think we are pursuing an official GPU one at this time, but maybe you guys can use my unofficial one? https://github.com/Namburger/edgetpu-ssdlite-mobiledet-retrain It's fully run-able on Colab!
Hey, I found another way:
follow the official GPU docker tutorial
in this file change FROM tensorflow/tensorflow:1.12.0-rc2-devel
to FROM tensorflow/tensorflow:1.12.0-rc2-devel-gpu
When running your docker add --gpus all
to your command. It should start like this: sudo docker run --name edgetpu-detect --gpus all --rm -it --privileged -p 6006:6006
You should be able to see your NVIDIA GPU from inside the docker with nvidia-smi
That's it. Follow the tutorial from coral for retraining.
Please check this colab GPU Tutorial for object detection: https://colab.sandbox.google.com/github/google-coral/tutorials/blob/master/retrain_efficientdet_model_maker_tf2.ipynb
I feel like a common use case would be to fine tune a MobileNet with a GPU and then covert everything over to edgetpu. I have created a docker that does this if there's any interest.