Migrate to Python 3.12, pick up CPU flavor of torch from the RHOAI 3.3 index

.tekton/own-app-lightspeed-rag-content-pull-request.yaml

@@ -614,7 +614,7 @@ spec:
   - pipelineTaskName: prefetch-dependencies
     computeResources:
       limits:
-        memory: 8Gi
+        memory: 12Gi
   taskRunTemplate:
     serviceAccountName: build-pipeline-own-app-lightspeed-rag-content
   timeouts:

.tekton/own-app-lightspeed-rag-content-push.yaml

@@ -611,7 +611,7 @@ spec:
   - pipelineTaskName: prefetch-dependencies
     computeResources:
       limits:
-        memory: 8Gi
+        memory: 12Gi
   taskRunTemplate:
     serviceAccountName: build-pipeline-own-app-lightspeed-rag-content
   timeouts:

Containerfile

@@ -2,14 +2,15 @@ ARG EMBEDDING_MODEL=sentence-transformers/all-mpnet-base-v2
 ARG FLAVOR=cpu
 ARG HERMETIC=false
 
-FROM registry.access.redhat.com/ubi9/python-311 as cpu-base
+FROM registry.access.redhat.com/ubi9/python-312 as cpu-base
 ARG EMBEDDING_MODEL
 ARG FLAVOR
 
-FROM nvcr.io/nvidia/cuda:12.9.1-devel-ubi9 as gpu-base
+FROM registry.redhat.io/rhai/base-image-cuda-12.9-rhel9:3.3 as gpu-base
 ARG EMBEDDING_MODEL
 ARG FLAVOR
-RUN dnf install -y python3.11 python3.11-pip libcudnn9 libnccl libcusparselt0
+USER 0
+RUN dnf install -y python3.12 python3.12-pip libcudnn9 libnccl libcusparselt0
 
 FROM ${FLAVOR}-base as lightspeed-rag-builder
 ARG EMBEDDING_MODEL
@@ -20,7 +21,7 @@ USER 0
 WORKDIR /workdir
 
 COPY requirements.gpu.txt .
-RUN pip3.11 install --no-cache-dir -r requirements.gpu.txt && ln -s /usr/local/lib/python3.11/site-packages/llama_index/core/_static/nltk_cache /root/nltk_data
+RUN pip3.12 install --no-cache-dir -r requirements.gpu.txt && ln -s /usr/local/lib/python3.12/site-packages/llama_index/core/_static/nltk_cache /root/nltk_data
 
 COPY ocp-product-docs-plaintext ./ocp-product-docs-plaintext
 COPY runbooks ./runbooks
@@ -33,14 +34,12 @@ RUN cd embeddings_model; if [ "$HERMETIC" == "true" ]; then \
     fi
 
 RUN if [ "$FLAVOR" == "gpu" ]; then \
-        export LD_LIBRARY_PATH=/usr/local/cuda-12/compat:$LD_LIBRARY_PATH; \
-        python3.11 -c "import torch; print(torch.version.cuda); print(torch.cuda.is_available());"; \
+        python3.12 -c "import torch; print(torch.version.cuda); print(torch.cuda.is_available());"; \
     fi
 
 COPY scripts/generate_embeddings.py .
-RUN export LD_LIBRARY_PATH=/usr/local/cuda-12/compat:$LD_LIBRARY_PATH; \
-    set -e && for OCP_VERSION in $(ls -1 ocp-product-docs-plaintext); do \
-        python3.11 generate_embeddings.py -f ocp-product-docs-plaintext/${OCP_VERSION} -r runbooks/alerts -md embeddings_model \
+RUN set -e && for OCP_VERSION in $(ls -1 ocp-product-docs-plaintext); do \
+        python3.12 generate_embeddings.py -f ocp-product-docs-plaintext/${OCP_VERSION} -r runbooks/alerts -md embeddings_model \
             -mn ${EMBEDDING_MODEL} -o vector_db/ocp_product_docs/${OCP_VERSION} \
             -i ocp-product-docs-$(echo $OCP_VERSION | sed 's/\./_/g') -v ${OCP_VERSION} -hb $HERMETIC; \
     done

Makefile

@@ -11,7 +11,7 @@ $(error Unsupported FLAVOR $(FLAVOR), must be 'cpu' or 'gpu')
 endif
 
 install-tools: ## Install required utilities/tools
-	@command -v pdm > /dev/null || { echo >&2 "pdm is not installed. Installing..."; pip3.11 install --no-cache-dir --upgrade pip pdm; }
+	@command -v pdm > /dev/null || { echo >&2 "pdm is not installed. Installing..."; pip3.12 install --no-cache-dir --upgrade pip pdm; }
 
 pdm-lock-check: ## Check that the pdm.lock file is in a good shape
 	pdm lock --check --group cpu --lockfile pdm.lock.cpu

byok/Containerfile.output

@@ -8,7 +8,7 @@ USER 0
 WORKDIR /workdir
 
 ENV VECTOR_DB_INDEX=$VECTOR_DB_INDEX
-RUN python3.11 generate_embeddings_tool.py -i /markdown -emd embeddings_model \
+RUN python3.12 generate_embeddings_tool.py -i /markdown -emd embeddings_model \
     -emn sentence-transformers/all-mpnet-base-v2 -o vector_db -id $VECTOR_DB_INDEX
 
 FROM ${UBI_BASE_IMAGE}

byok/Containerfile.tool

@@ -1,5 +1,5 @@
 ARG BYOK_TOOL_IMAGE=registry.redhat.io/openshift-lightspeed-tech-preview/lightspeed-rag-tool-rhel9:latest
-ARG UBI_BASE_IMAGE=registry.access.redhat.com/ubi9/ubi:latest
+ARG UBI_BASE_IMAGE=registry.redhat.io/rhai/base-image-cpu-rhel9:3.3
 ARG HERMETIC=false
 FROM ${UBI_BASE_IMAGE}
 ARG LOG_LEVEL=info
@@ -8,13 +8,13 @@ ARG VECTOR_DB_INDEX=vector_db_index
 ARG BYOK_TOOL_IMAGE
 ARG UBI_BASE_IMAGE
 ARG HERMETIC
-RUN dnf install -y buildah python3.11 python3.11-pip && dnf clean all
-
 USER 0
+RUN dnf install -y buildah python3.12 python3.12-pip && dnf clean all
+
 WORKDIR /workdir
 
 COPY requirements.cpu.txt .
-RUN pip3.11 install --no-cache-dir --no-deps -r requirements.cpu.txt
+RUN pip3.12 install --no-cache-dir --no-deps -r requirements.cpu.txt
 
 COPY embeddings_model ./embeddings_model
 ENV HERMETIC=$HERMETIC

pyproject.toml

@@ -8,7 +8,7 @@ lint.extend-select = ["D401"]
 
 lint.ignore = []
 
-target-version = "py311"
+target-version = "py312"
 lint.pydocstyle.convention = "google"
 line-length = 100
 
@@ -33,10 +33,10 @@ build-backend = "pdm.backend"
 
 [project.optional-dependencies]
 cpu = [
-    "torch @ https://download.pytorch.org/whl/cpu/torch-2.6.0%2Bcpu-cp311-cp311-linux_x86_64.whl#sha256=5b6ae523bfb67088a17ca7734d131548a2e60346c622621e4248ed09dd0790cc",
+    "torch @ https://packages.redhat.com/api/pulp-content/public-rhai/rhoai/3.3/cpu-ubi9/torch-2.9.1-4-cp312-cp312-linux_x86_64.whl#sha256=e42ab849b64444059f5eda352d61c9c3a078f30797e48e2972857182c9a00cf8",
 ]
 gpu = [
-    "torch==2.6.0",
+    "torch==2.9.1",
 ]
 
 [project]
@@ -58,6 +58,6 @@ dependencies = [
     "beautifulsoup4>=4.10.0",
     "urllib3>=2.6.3",
 ]
-requires-python = "==3.11.*"
+requires-python = "==3.12.*"
 readme = "README.md"
 license = {file = "LICENSE"}