I started just yesterday on my first experience with ODM and I am already in LOVE with it!
I am running ODM from the docker and I want to know if ODM allows to leverage on multiple GPUs? I have the possibility to scale up and use up to 4 NVIDIA® Tesla® V100.
Unfortunately, we only support 1 GPU at a time for now.
If you know of anyone who might help us build multi-GPU support, that would be
incredibly appreciated! You’re not the first person to ask after it
What skillset is needed for that kind of s/w development?
My (very limited, likely incorrect) understanding is:
Probably Python to help figure out how to enumerate and enable multiple GPUs and how to split up the dataset to go to each GPU in parallel and how to then make sure everything makes it back processed and there are no threading/timing issues on what image gets processed by what GPU, etc.
Unfortunately I am also not an expert when it comes to GPU. Anyways thanks for the answers and let me know if there are ways that I can contribute/donate to make this feature happen.
Keep up with the great work you are doing on ODM! I can’t believe I worked with drones for almost a decade (using Agisoft) and never used ODM, which now seems to be the future for us at NIBIO.
20 June 2022 18:55
This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.