update to run from release script, as intended

Author: ckhordiasma

Dockerfile.ubi

@@ -32,7 +32,6 @@ ENV PYTHON_VERSION=${PYTHON_VERSION}
 RUN microdnf install -y --nodocs \
     python${PYTHON_VERSION}-devel  && \
     python${PYTHON_VERSION} -m venv $VIRTUAL_ENV && \
-    pip install --no-cache -U pip wheel uv && \
     microdnf clean all
 
 
@@ -113,17 +112,14 @@ COPY examples/*.jinja /app/data/template/
 
 # install vllm by running the payload script and then install flashinfer
 
-ARG VLLM_WHEEL_VERSION
-ARG VLLM_WHEEL_INDEX
-ARG FLASHINFER_VERSION
+ARG WHEEL_RELEASE
 RUN --mount=type=cache,target=/root/.cache/uv \
     --mount=type=bind,src=payload,target=/workspace/payload \
     --mount=type=secret,id=rhel-ai-private-index-auth/BOT_PAT \
+        source ${VIRTUAL_ENV}/bin/activate && \
         env BOT_PAT=$(cat /run/secrets/rhel-ai-private-index-auth/BOT_PAT) \
-            VLLM_WHEEL_VERSION=${VLLM_VERSION} \
-            VLLM_WHEEL_INDEX=${VLLM_WHEEL_INDEX} \
-        ./payload/run.sh && \
-        uv pip install "${FLASHINFER_VERSION}" 
+        WHEEL_RELEASE=${WHEEL_RELEASE} \
+        ./payload/run.sh
 
 ENV HF_HUB_OFFLINE=1 \
     HOME=/home/vllm \
@@ -161,17 +157,15 @@ FROM vllm-openai AS vllm-grpc-adapter
 
 USER root
 
-ARG VLLM_TGIS_ADAPTER_VERSION
+ARG WHEEL_RELEASE
 RUN --mount=type=cache,target=/root/.cache/uv \
     --mount=type=bind,src=payload,target=/workspace/payload \
     --mount=type=secret,id=rhel-ai-private-index-auth/BOT_PAT \
     cd /workspace && \
-    ls && \
+    source ${VIRTUAL_ENV}/bin/activate && \
     env HOME=/root \
         BOT_PAT=$(cat /run/secrets/rhel-ai-private-index-auth/BOT_PAT) \
-        VLLM_WHEEL_VERSION=${VLLM_VERSION} \
-        VLLM_TGIS_ADAPTER_VERSION=${VLLM_TGIS_ADAPTER_VERSION} \
-        VLLM_WHEEL_INDEX=${VLLM_WHEEL_INDEX} \
+        WHEEL_RELEASE=${WHEEL_RELEASE} \
         ./payload/run.sh
 
 

argfile.konflux

@@ -1,7 +1,5 @@
 BASE_UBI_IMAGE_TAG=9.5-1739420147
 PYTHON_VERSION=3.11
 LIBSODIUM_VERSION=1.0.20
-VLLM_TGIS_ADAPTER_VERSION=0.6.3
-FLASHINFER_VERSION=https://github.com/flashinfer-ai/flashinfer/releases/download/v0.2.1.post1/flashinfer_python-0.2.1.post1+cu124torch2.5-cp38-abi3-linux_x86_64.whl
-VLLM_WHEEL_VERSION=0.7.2
-VLLM_WHEEL_INDEX=https://gitlab.com/api/v4/projects/66664052/packages/pypi/simple
+WHEEL_RELEASE=2.20.55+vllm-cuda-ubi9-x86_64
+# can view releases at https://gitlab.com/redhat/rhel-ai/rhoai/pipeline/-/releases

payload/run.sh

@@ -1,10 +1,9 @@
 #!/bin/bash
+# Script assumes python venv is already properly configured
 # required env vars:
 # $BOT_PAT
-# $WHEEL_RELEASE_ARTIFACTS
-# optional:
-# $VLLM_TGIS_ADAPTER_VERSION
-# $VLLM_WHEEL_VERSION
+# $WHEEL_RELEASE
+# $WHEEL_BASEURL
 set -ex
 
 cat <<EOF > ${HOME}/.netrc
@@ -13,22 +12,17 @@ login rhel-ai-wheels-prefetch-token-rhoai
 password $BOT_PAT
 EOF
 
-trap "rm ${HOME}/.netrc" EXIT
+trap "rm -rf ${HOME}/.netrc release release.tar.gz" EXIT
 
-# https://docs.astral.sh/uv/configuration/indexes/#searching-across-multiple-indexes
-# This will prefer to use the custom index, and fall back to pypi if needed
-export UV_EXTRA_INDEX_URL=${VLLM_WHEEL_INDEX}
-export UV_INDEX_STRATEGY=unsafe-first-match 
+# WHEEL_RELEASE="2.20.55+vllm-cuda-ubi9-x86_64"
 
-vllm="vllm[tensorizer,audio,video]"
+# Gitlab project ID, etc should be static 
+WHEEL_RELEASE_ARTIFACTS="https://gitlab.com/api/v4/projects/68045055/packages/generic/rhelai-wheels/${WHEEL_RELEASE}/wheels-${WHEEL_RELEASE}.tar.gz"
 
-if [[ -n "$VLLM_TGIS_ADAPTER_VERSION" ]]; then
-    vllm_tgis_adapter="vllm-tgis-adapter==${VLLM_TGIS_ADAPTER_VERSION}"
-fi
 
-if [[ -n "$VLLM_WHEEL_VERSION" ]]; then
-    vllm="${vllm}==${$VLLM_WHEEL_VERSION}"
-fi
+# NOTE - ensure that flashinfer is included in wheel thing
 
-uv pip install $vllm $vllm_tgis_adapter
+curl --netrc -o release.tar.gz ${WHEEL_RELEASE_ARTIFACTS} 
+tar zxvf release.tar.gz 
+./release/install_wheels.sh