Run on multiple GPU

[hiral-kotadiya]

How can we run retina face on multiple GPU?

[serengil]

if you install tensorflow-gpu before installing deepface, that package will handle it.

[csmailis]

After you install tensorflow-gpu you can use the following command within the Python script that contains RetinaFace calls to:

1. os.environ["CUDA_VISIBLE_DEVICES"] = "1" - > Specify which GPU to use
2. os.environ['TF_FORCE_GPU_ALLOW_GROWTH'] = 'true' - > Limit the amount of reserved VRAM so that other scripts can be run in the same GPU.

Maybe a section in the Readme.md providing example uses of the above commands in user-created scripts could be useful as many others seem to be looking into these issues.