build-rocm-image

name	build-rocm-image
description	Connect to a remote host via SSH and build a Docker image with rocprofv3, vllm, aiter and FlyDSL. Use when user wants to build/rebuild the ROCm development image on a remote host. Usage: /build-rocm-image <hostname> Use when this capability is needed.

Build ROCm Development Image

Build a Docker image on a remote host with rocm gpu access based on rocm/vllm-dev:nightly.

Arguments

Argument	Required	Description
<HOST>	Yes	The remote hostname to SSH into and build the image on. Example: hjbog-srdc-39.amd.com

When this skill is invoked, the argument passed in is the target hostname. Replace all occurrences of <HOST> below with the provided hostname. If no hostname is provided, ask the user for it before proceeding.

Target Host

• Host: <HOST> (provided as argument)
• Access: SSH (key-based authentication)

Base Image

• Image: rocm/vllm-dev:nightly
• Included: rocprofv3, PyTorch (versions determined by the nightly base image)

Customization

• vLLM: Replace pre-installed version with https://github.com/ROCm/vllm branch ps_pa
• aiter: Replace pre-installed version with https://github.com/ROCm/aiter branch main
• FlyDSL: Build and install from https://github.com/ROCm/FlyDSL branch main using the same LLVM/MLIR + FlyDSL build sequence documented in the build-flydsl skill

Authentication

Do not put GitHub tokens in clone URLs, Dockerfiles, or command lines. Pass the token to Docker BuildKit as a secret:

• The remote host must have GITHUB_TOKEN set in the environment used by the docker build command.
• The Dockerfile reads the token only through /run/secrets/github_token during the specific RUN steps that need GitHub access.
• The token is not written to /tmp/Dockerfile.rocm-custom, Docker image layers, or the final image.

Build Steps

Step 1: Generate Dockerfile on remote host

ssh -o ConnectTimeout=30 <HOST> 'cat > /tmp/Dockerfile.rocm-custom' <<'DOCKERFILE'
# syntax=docker/dockerfile:1.7
FROM rocm/vllm-dev:nightly

# Uninstall existing, vllm, aiter
RUN pip uninstall -y vllm aiter 2>/dev/null; true

# Install build dependencies
RUN pip install ninja cmake pybind11

# Clone and install aiter from main branch
RUN --mount=type=secret,id=github_token \
    GITHUB_TOKEN="$(cat /run/secrets/github_token)" && \
    cd /tmp && \
    git -c http.extraHeader="Authorization: Bearer ${GITHUB_TOKEN}" clone --depth 1 --branch main https://github.com/ROCm/aiter.git && \
    cd aiter && \
    pip install -e . && \
    cd / && rm -rf /tmp/aiter

# Clone, build, and install FlyDSL from main branch.
# This follows the build-flydsl skill: build LLVM/MLIR first, build FlyDSL,
# then install editable from the persistent checkout.
RUN --mount=type=secret,id=github_token \
    GITHUB_TOKEN="$(cat /run/secrets/github_token)" && \
    cd /opt && \
    git -c http.extraHeader="Authorization: Bearer ${GITHUB_TOKEN}" clone --depth 1 --branch main https://github.com/ROCm/FlyDSL.git && \
    cd FlyDSL && \
    bash scripts/build_llvm.sh -j$(nproc) && \
    bash scripts/build.sh -j$(nproc) && \
    pip install -e .

# Clone and install vllm from ROCm/vllm ps_pa branch
RUN --mount=type=secret,id=github_token \
    GITHUB_TOKEN="$(cat /run/secrets/github_token)" && \
     /tmp && \
    git -c http.extraHeader=  --depth 1 --branch ps_pa https://github.com/ROCm/vllm.git && \
     vllm && \
    pip install -e . && \
     / &&  -rf /tmp/vllm


RUN --mount==secret,=github_token \
    GITHUB_TOKEN= && \
     /tmp && \
    wget -q --header= https://github.com/ROCm/rocprof-trace-decoder/releases/download/0.1.6/rocprof-trace-decoder-manylinux-2.28-0.1.6-Linux.sh && \
     +x rocprof-trace-decoder-manylinux-2.28-0.1.6-Linux.sh && \
    ./rocprof-trace-decoder-manylinux-2.28-0.1.6-Linux.sh --skip-license --prefix=/tmp/rtd-install && \
    find /tmp/rtd-install -name  -  -a {} /opt/rocm/lib/ \; && \
    ldconfig && \
     -rf /tmp/rocprof-trace-decoder-manylinux-2.28-0.1.6-Linux.sh /tmp/rtd-install


RUN python3 -c  && \
    python3 -c  && \
    python3 -c  && \
    python3 -c  && \
     rocprofv3 &&   && \
     /opt/rocm/lib/librocprof*decoder* &&  

LABEL description=
DOCKERFILE

Step 2: Build the image

Build the image with BuildKit enabled and pass GITHUB_TOKEN as a secret. Use --network=host to ensure git clone works.

ssh -o ConnectTimeout=30 <HOST> 'if [ -z "$GITHUB_TOKEN" ]; then echo "GITHUB_TOKEN is not set on the remote host" >&2; exit 1; fi; DOCKER_BUILDKIT=1 docker build --network=host --secret id=github_token,env=GITHUB_TOKEN -t rocm-dev-custom:flydsl -f /tmp/Dockerfile.rocm-custom /tmp'

Step 3: Verify the built image

ssh -o ConnectTimeout=30 <HOST> "docker run --rm rocm-dev-custom:flydsl bash -c '
echo \"=== Triton ===\"
python3 -c \"import triton; print(triton.__version__)\"
echo \"=== vLLM ===\"
python3 -c \"import vllm; print(vllm.__version__)\"
echo \"=== aiter ===\"
python3 -c \"import aiter; print(aiter.__version__)\" 2>/dev/null || python3 -c \"import aiter; print(\\\"aiter OK\\\")\"
echo \"=== FlyDSL ===\"
python3 -c \"import flydsl; print(flydsl.__version__)\" 2>/dev/null || python3 -c \"import flydsl; print(\\\"FlyDSL OK\\\")\"
echo \"=== rocprofv3 ===\"
rocprofv3 --version 2>/dev/null || which rocprofv3
echo \"=== ROCm ===\"
cat /opt/rocm/.info/version
'"

Step 4: Clean up

ssh -o ConnectTimeout=30 <HOST> "rm -f /tmp/Dockerfile.rocm-custom"

Output

Report to the user:

• The image name and tag
• Versions of triton, vllm, aiter, and ROCm inside the image
• Any build warnings or errors

Error Handling

• If SSH connection fails, inform the user they need a valid SSH key and Conductor reservation
• If disk space is insufficient, explain that the FlyDSL LLVM/MLIR build needs about 50GB and suggest cleaning unused images with docker image prune

Example Usage

To start a container from the built image with GPU access:

ssh <HOST> "docker run -it --device=/dev/kfd --device=/dev/dri --group-add video --shm-size=64g rocm-dev-custom:flydsl bash"

---

Source: fsx950223/claude-stuff — distributed by TomeVault.