Change GPU

[Bariatric-surgery]

I get an error that I am running out of memory while running monailabel radiology app on my server.

We have multiple GPUs and when I checked the GPU: 0 (which got selected by default of monailabel) that it has almost no memory left.

How can I change the GPU which is being used in the code?
So for example switiching from "GPU: 0" to "GPU: 2"

Thank you very much!!

[SachidanandAlle]

https://stackoverflow.com/questions/39649102/how-do-i-select-which-gpu-to-run-a-job-on

[SachidanandAlle]

if you are running training.. MONAILabel supports multi-gpu and gpus flag.. where for you can say gpus=1,2 to pick others

[Bariatric-surgery]

Is there a way to specify the GPU used in the monailabel code itself? so not on environmental variable, but on the python code.

[SachidanandAlle]

https://github.com/Project-MONAI/MONAILabel/blob/main/sample-apps/radiology/lib/configs/segmentation.py#L87 here you can pass additional configs config={"largest_cc": True, "device": "cuda-1"} For Training, you can pass gpus=1,2 in the train request json. If you see the 3D slicer, they are already available for user to pass on. Enable Developer Mode under Edit => Application Settings => MONAI Label => Developer Mode

…

On Mon, Dec 5, 2022 at 11:47 PM Bariatric-surgery ***@***.***> wrote: Is there a way to specify the GPU used in the monailabel code itself? so not on environmental variable, but on the python code. — Reply to this email directly, view it on GitHub <#1177 (reply in thread)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABX7YK7F24KWTOJZOBEMANDWL3VRPANCNFSM6AAAAAASUK37TE> . You are receiving this because you commented.Message ID: ***@***.*** com>

-- *Thanks & Regards* *SACHI*

[Bariatric-surgery]

Many thanks