search

wineslab / boston_twin

Repository for the BostonTwin dataset API.
https://wineslab.github.io/boston_twin/
MIT License
11 stars 2 forks source link

Feature Request: Add non-GPU build / docker instructions #16

Open Blake-Haydon opened 1 month ago

Blake-Haydon commented 1 month ago

I do not currently have a NVIDIA GPU but would like to test out the bostontwin_demo.ipynb inside a docker container. It would be great if there were instructions on how to do that in the README.

Thanks!

p-testolina commented 1 month ago

Hi @Blake-Haydon, thanks for your input. I'll add some instructions as soon as possible. Are you trying to use the Docker container, or do you use a Python virtual environment (e.g., conda or venv)?

hiteshPoddar commented 2 weeks ago

Same issue as @Blake-Haydon. I am trying to run this on a MAC OS system without any gpu support. This command fails --- b. Run the container binding the boston_twin directory to the container: docker run --privileged=true --gpus=all --mount type=bind,src=".",target="/home/root/boston_twin" --env NVIDIA_DRIVER_CAPABILITIES=graphics,compute,utility --rm -it bostontwin

Please provide an alternative at the earliest.

p-testolina commented 2 weeks ago

Hi @hiteshPoddar,

  1. Without a error log, it's hard to tell what's the problem
  2. On MAC OS, there are differences in the tensorflow installation with/without silicon chips (M1-3), and I'm not sure how Mitsuba (the underlying library that Sionna uses for RT) deals with the M-* chips.
  3. I would suggest you try the Sionna Docker image or local installation first, and see if that runs on your device. The additional libraries required by Bostontwin should not have any compatibility issue with MAC OS.

In general, to run it on CPU, it should be enough to remove the --env and the --gpus=all flags