RB chooses Intel GPU instead of Nvidia GPU when both are available

[jmplaza]

My laptop already has two GPUs: an integrated Intel and a nVidia GeForce GTX 1650. The GPU detection code inisde RoboticsBackend already detects both, but complains about "WARNING: nvidia-smi not available or failed, skipping NVIDIA GPU". In fact, nvidia-smi is properly available, as seen in the image.

Any idea @dpascualhe ?

[javizqh]

It is missing the nvidia part of the command

[jmplaza]

Sure @javizqh , but that is a bug, the nvidia-smi command is already available natively, as shown in the image. It should also be available inside the RB too.

[dpascualhe]

this is the right command (generic for either using nvidia if available, or falling back to to other gpus/cpu):

docker run --rm -it $(nvidia-smi >/dev/null 2>&1 && echo "--gpus all" || echo "") --device /dev/dri -p 6080-6090:6080-6090 -p 7163:7163 -p 7164:7164 --link academy_db jderobot/robotics-academy:latest

maybe the instructions are a bit ambiguous (https://jderobot.github.io/RoboticsAcademy/user_guide/2). I can give it a thought

[jmplaza]

OK @dpascualhe . For sure we have to update the Unibotics user guide too.

Is port 7164 still needed @javizqh ?

[javizqh]

In Unibotics no, in RA yes.

This is the command for Unibotics

docker run --rm -it $(nvidia-smi >/dev/null 2>&1 && echo "--gpus all" || echo "") --device /dev/dri -p 6080-6090:6080-6090 -p 7163:7163  --link academy_db jderobot/robotics-backend:latest