A container cannot see GPUs, although the host can. What do you check?
Why C is the answer
The container runtime and toolkit
The toolkit gives containers GPU access.
In detail
- The toolkit gives containers GPU access.
- Check the runtime configuration.
Use case
An illustrative scenario, not a real incident.
The host sees GPUs but a container does not, because the runtime is not set up with the toolkit.
See it in different ways
stateDiagram-v2 [*] --> LowUtilization LowUtilization --> DataPath: GPUs waiting for data LowUtilization --> Throttling: temperature or power DataPath --> BenchmarkStorage Throttling --> CheckCooling BenchmarkStorage --> [*] CheckCooling --> [*]
flowchart TD
Q{"Can the host see GPUs?"}
Q -->|"Yes"| A0["Check container runtime and toolkit"]
Q -->|"No"| A1["Check the driver"]Key point: The toolkit gives containers GPU access.
Why the other options are wrong
- A. The model's tokenizer
- That is an LLM part.
- B. The Slurm user list for the team that owns the job
- A user list is unrelated.
- D. The BMC password
- The BMC is unrelated.