not sure i follow you. i am only talking about the GPUs
clearml-agent daemon --detached --queue bell --gpus 1
Are you using the agent in docker mode (meaning is your task running inside a docker container)?
Hi @<1546303277010784256:profile|LivelyBadger26> , when you run the agent, you can specify which CPUs to use using the --gpus argument like you used
Try inspecting the environment (os.environ) - I'm pretty sure you'll see it there...
How did you set the agent for these GPUs?