[Feature]: Make ROCm Docker UX Intuitive - Achieve User Experience Parity with NVIDIA

[OrenLeung]

Suggestion Description

There is lots of room for improvement on ROCm docker UX. On NVIDIA GPUs, end users just do --gpus all but on AMD, i need to type like 8 additional different flags.

i know this issue seems minor but it is tons of minors quirks across the ROCm ecosystem that hurts the user experience.

any chance, AMD could work with docker corp to make AMD GPUs also just be also as simple as --gpus all?

Currently

on NVIDIA GPUs, users just need to add 1 flag

docker run --gpu alls ...

Currently on ROCm GPUs, user need to a bunch of flags

sudo docker run --network=host \
    --device=/dev/kfd \
    --device=/dev/dri \
    --ipc=host \
    --shm-size 16G \
    --group-add video \
    --cap-add=SYS_PTRACE \
    --security-opt seccomp=unconfined
   ...

Suggestion

ROCm follows the UX pattern on NVIDIA GPUs and users just need to add 1 flag

docker run --gpu alls ...

cc: @hliuca and @powderluv

Operating System

No response

GPU

No response

ROCm Component

No response