debug large runner

Author: aiwantaozi

.github/workflows/docker-ci-base-image.yaml

@@ -17,7 +17,7 @@ jobs:
       actions: read
       id-token: write
     timeout-minutes: 60
-    runs-on: ubuntu-22.04
+    runs-on: ubuntu-22.04-16c64g
     env:
       PACKAGE_REGISTRY: ${{ vars.PACKAGE_REGISTRY || 'gpustack'}}
       PACKAGE_IMAGE: ${{ vars.PACKAGE_IMAGE || 'rocm_dev-ubuntu-22.04' }}

Dockerfile.rocm.base

@@ -45,6 +45,7 @@ RUN apt-get update && apt-get install -y \
     rocsparse-dev \
     rocthrust-dev \
     rocm-smi-lib \
+    amd-smi-lib \
     && apt-get clean && rm -rf /var/lib/apt/lists/*
 
 # When launching the container, mount the code directory to /vllm-workspace
@@ -181,5 +182,5 @@ RUN --mount=type=cache,target=/root/.cache/pip \
     python3 -m pip install libs/*.whl; fi
 
 # Clean packages
-RUN apt-get purge -y build-essential gcc bzip2 libx11-6 tmux unzip && apt-get autoremove -y \
+RUN apt-get purge -y bzip2 libx11-6 tmux unzip && apt-get autoremove -y \
     && rm -rf /root/.cache /install /libs /dist

Dockerfile.rocm2

@@ -1,4 +1,6 @@
-ARG BASE_IMAGE=vllm-rocm-fa-0-base-3
+ARG BASE_IMAGE=micheliac/vllm-base:latest
+
+FROM $BASE_IMAGE
 
 ARG TARGETPLATFORM
 ENV DEBIAN_FRONTEND=noninteractive