Add vLLM support

* run on larger GH actions runner and print disk info

* add vllm support by modifying the Dockerfile

.github/workflows/build.yml

@@ -21,7 +21,7 @@ permissions:
 jobs:
   build:
     name: Build
-    runs-on: ubuntu-latest
+    runs-on: snomed-ubuntu-22.04-8core
     strategy:
       matrix:
         proc: ["cpu", "gpu"]
@@ -40,6 +40,10 @@ jobs:
     steps:
       - uses: actions/checkout@v4
 
+      - name: Disk information
+        run: |
+          df -h
+
       - name: Build Image
         run: |
           docker build runtime \

runtime/Dockerfile

@@ -8,7 +8,8 @@ ENV DEBIAN_FRONTEND=noninteractive \
     LANG=C.UTF-8  \
     LC_ALL=C.UTF-8 \
     PYTHONUNBUFFERED=1 \
-    SHELL=/bin/bash
+    SHELL=/bin/bash \
+    CPU_OR_GPU=${CPU_OR_GPU}
 
 COPY apt.txt apt.txt
 RUN apt-get update --fix-missing \
@@ -21,6 +22,13 @@ COPY --chown=$MAMBA_USER:$MAMBA_USER conda-lock-${CPU_OR_GPU}.yml /tmp/conda-loc
 RUN micromamba install --name base --yes --file /tmp/conda-lock.yml && \
     micromamba clean --all --force-pkgs-dirs --yes
 
+ARG MAMBA_DOCKERFILE_ACTIVATE=1
+RUN if [ "$CPU_OR_GPU" = "gpu" ]; then \
+    pip install https://github.com/vllm-project/vllm/releases/download/v0.3.0/vllm-0.3.0+cu118-cp310-cp310-manylinux1_x86_64.whl --no-cache-dir ; \
+    else \ 
+    pip install https://github.com/vllm-project/vllm/releases/download/v0.3.0/vllm-0.3.0-cp310-cp310-manylinux1_x86_64.whl --no-cache-dir ; \
+    fi
+
 RUN mkdir /code_execution
 RUN chown -R ${MAMBA_USER}:${MAMBA_USER} /code_execution