Hello,
I am using GPUStack to deploy a custom backend. I have a requirement to pass a specific NVMe device from the host into the backend container.
Specifically, I need to achieve this Docker mapping: --device /dev/nvme1n1:/dev/nvme0n1
Since GPUStack manages the container startup automatically, I cannot manually add this flag to the docker run command.
Is there a configuration setting, environment variable, or a "passthrough" mechanism in GPUStack that allows injecting custom device mappings or extra Docker arguments for backend workers?
If not, would it be possible to add support for defining extra devices in the backend configuration?
Thank you!