How do I use multi gpu setups?

[FileDotZip]

I have a intel scalable gpu server, with 6x Nvidia P40 video cards with 24GB of VRAM each.

How can I specify for llama.cpp to use as much vram as it needs from this cluster of gpu's? Does it automatically do it?

I am following this guide at step 6

[JohannesGaessler]

https://github.com/ggerganov/llama.cpp/tree/master/examples/main