Skip to content

Commit 2b96633

Browse files
jstouracjstourac
committed
[RHOAIENG-14585] Add rhel9 based CUDA Python 3.11 image dockerfile
Our CUDA RStudio image is based on RHEL9 with Python 3.11 image. Unfortunatelly we missed to add this variant during our work on the Python 3.11 addition. This change fixes it and makes the CUDA RStudio BuildConfig work again. * https://issues.redhat.com/browse/RHOAIENG-14585
1 parent b43efbf commit 2b96633

File tree

4 files changed

+392
-1
lines changed

4 files changed

+392
-1
lines changed

‎cuda/rhel9-python-3.11/Dockerfile

Lines changed: 155 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,155 @@
1+
ARG BASE_IMAGE
2+
FROM ${BASE_IMAGE}
3+
4+
# Access the client's secret for the subscription manager from the environment variable
5+
ARG SECRET_DIR=/opt/app-root/src/.sec
6+
ARG SERVERURL_DEFAULT=""
7+
ARG BASEURL_DEFAULT=""
8+
9+
LABEL name="odh-notebook-cuda-rhel9-python-3.11" \
10+
summary="CUDA Python 3.11 base image for ODH notebooks" \
11+
description="CUDA Python 3.11 builder image based on RHEL9 for ODH notebooks" \
12+
io.k8s.display-name="CUDA Python 3.11 base image for ODH notebooks" \
13+
io.k8s.description="CUDA Python 3.11 builder image based on RHEL9 for ODH notebooks" \
14+
authoritative-source-url="https://github.com/opendatahub-io/notebooks" \
15+
io.openshift.build.commit.ref="main" \
16+
io.openshift.build.source-location="https://github.com/opendatahub-io/notebooks/tree/main/cuda/rhel9-python-3.11" \
17+
io.openshift.build.image="quay.io/opendatahub/workbench-images:cuda-rhel9-python-3.11"
18+
19+
# Install CUDA base from:
20+
# https://gitlab.com/nvidia/container-images/cuda/-/tree/master/dist/12.4.1/ubi9/base
21+
USER 0
22+
WORKDIR /opt/app-root/bin
23+
24+
# Run the subscription manager command using the provided credentials. Only include --serverurl and --baseurl if they are provided
25+
RUN SERVERURL=$(cat ${SECRET_DIR}/SERVERURL 2>/dev/null || echo ${SERVERURL_DEFAULT}) && \
26+
BASEURL=$(cat ${SECRET_DIR}/BASEURL 2>/dev/null || echo ${BASEURL_DEFAULT}) && \
27+
USERNAME=$(cat ${SECRET_DIR}/USERNAME) && \
28+
PASSWORD=$(cat ${SECRET_DIR}/PASSWORD) && \
29+
subscription-manager register \
30+
${SERVERURL:+--serverurl=$SERVERURL} \
31+
${BASEURL:+--baseurl=$BASEURL} \
32+
--username=$USERNAME \
33+
--password=$PASSWORD \
34+
--force \
35+
--auto-attach
36+
37+
ENV NVARCH x86_64
38+
ENV NVIDIA_REQUIRE_CUDA "cuda>=12.4 brand=tesla,driver>=470,driver<471 brand=unknown,driver>=470,driver<471 brand=nvidia,driver>=470,driver<471 brand=nvidiartx,driver>=470,driver<471 brand=geforce,driver>=470,driver<471 brand=geforcertx,driver>=470,driver<471 brand=quadro,driver>=470,driver<471 brand=quadrortx,driver>=470,driver<471 brand=titan,driver>=470,driver<471 brand=titanrtx,driver>=470,driver<471 brand=tesla,driver>=525,driver<526 brand=unknown,driver>=525,driver<526 brand=nvidia,driver>=525,driver<526 brand=nvidiartx,driver>=525,driver<526 brand=geforce,driver>=525,driver<526 brand=geforcertx,driver>=525,driver<526 brand=quadro,driver>=525,driver<526 brand=quadrortx,driver>=525,driver<526 brand=titan,driver>=525,driver<526 brand=titanrtx,driver>=525,driver<526 brand=tesla,driver>=535,driver<536 brand=unknown,driver>=535,driver<536 brand=nvidia,driver>=535,driver<536 brand=nvidiartx,driver>=535,driver<536 brand=geforce,driver>=535,driver<536 brand=geforcertx,driver>=535,driver<536 brand=quadro,driver>=535,driver<536 brand=quadrortx,driver>=535,driver<536 brand=titan,driver>=535,driver<536 brand=titanrtx,driver>=535,driver<536"
39+
ENV NV_CUDA_CUDART_VERSION 12.4.127-1
40+
41+
COPY cuda/rhel9-python-3.11/cuda.repo-x86_64 /etc/yum.repos.d/cuda.repo
42+
43+
RUN NVIDIA_GPGKEY_SUM=d0664fbbdb8c32356d45de36c5984617217b2d0bef41b93ccecd326ba3b80c87 && \
44+
curl -fsSL https://developer.download.nvidia.com/compute/cuda/repos/rhel9/${NVARCH}/D42D0685.pub | sed '/^Version/d' > /etc/pki/rpm-gpg/RPM-GPG-KEY-NVIDIA && \
45+
echo "$NVIDIA_GPGKEY_SUM /etc/pki/rpm-gpg/RPM-GPG-KEY-NVIDIA" | sha256sum -c --strict -
46+
47+
ENV CUDA_VERSION 12.4.1
48+
49+
# For libraries in the cuda-compat-* package: https://docs.nvidia.com/cuda/eula/index.html#attachment-a
50+
RUN yum upgrade -y && yum install -y \
51+
cuda-cudart-12-4-${NV_CUDA_CUDART_VERSION} \
52+
cuda-compat-12-4 \
53+
&& yum clean all \
54+
&& rm -rf /var/cache/yum/*
55+
56+
# nvidia-docker 1.0
57+
RUN echo "/usr/local/nvidia/lib" >> /etc/ld.so.conf.d/nvidia.conf && \
58+
echo "/usr/local/nvidia/lib64" >> /etc/ld.so.conf.d/nvidia.conf
59+
60+
ENV PATH /usr/local/nvidia/bin:/usr/local/cuda/bin:${PATH}
61+
ENV LD_LIBRARY_PATH /usr/local/nvidia/lib:/usr/local/nvidia/lib64
62+
63+
COPY cuda/rhel9-python-3.11/NGC-DL-CONTAINER-LICENSE /
64+
65+
# nvidia-container-runtime
66+
ENV NVIDIA_VISIBLE_DEVICES all
67+
ENV NVIDIA_DRIVER_CAPABILITIES compute,utility
68+
69+
# Install CUDA runtime from:
70+
# https://gitlab.com/nvidia/container-images/cuda/-/blob/master/dist/12.4.1/ubi9/runtime/Dockerfile
71+
ENV NV_CUDA_LIB_VERSION 12.4.1-1
72+
ENV NV_NVTX_VERSION 12.4.127-1
73+
ENV NV_LIBNPP_VERSION 12.2.5.30-1
74+
ENV NV_LIBNPP_PACKAGE libnpp-12-4-${NV_LIBNPP_VERSION}
75+
ENV NV_LIBCUBLAS_VERSION 12.4.5.8-1
76+
ENV NV_LIBNCCL_PACKAGE_NAME libnccl
77+
ENV NV_LIBNCCL_PACKAGE_VERSION 2.21.5-1
78+
ENV NV_LIBNCCL_VERSION 2.21.5
79+
ENV NCCL_VERSION 2.21.5
80+
ENV NV_LIBNCCL_PACKAGE ${NV_LIBNCCL_PACKAGE_NAME}-${NV_LIBNCCL_PACKAGE_VERSION}+cuda12.4
81+
82+
RUN yum install -y \
83+
cuda-libraries-12-4-${NV_CUDA_LIB_VERSION} \
84+
cuda-nvtx-12-4-${NV_NVTX_VERSION} \
85+
${NV_LIBNPP_PACKAGE} \
86+
libcublas-12-4-${NV_LIBCUBLAS_VERSION} \
87+
${NV_LIBNCCL_PACKAGE} \
88+
&& yum clean all \
89+
&& rm -rf /var/cache/yum/*
90+
91+
# Set this flag so that libraries can find the location of CUDA
92+
ENV XLA_FLAGS=--xla_gpu_cuda_data_dir=/usr/local/cuda
93+
94+
# Install CUDA devel from:
95+
# https://gitlab.com/nvidia/container-images/cuda/-/blob/master/dist/12.4.1/ubi9/devel/Dockerfile
96+
ENV NV_CUDA_LIB_VERSION 12.4.1-1
97+
ENV NV_NVPROF_VERSION 12.4.127-1
98+
ENV NV_NVPROF_DEV_PACKAGE cuda-nvprof-12-4-${NV_NVPROF_VERSION}
99+
ENV NV_CUDA_CUDART_DEV_VERSION 12.4.127-1
100+
ENV NV_NVML_DEV_VERSION 12.4.127-1
101+
ENV NV_LIBCUBLAS_DEV_VERSION 12.4.5.8-1
102+
ENV NV_LIBNPP_DEV_VERSION 12.2.5.30-1
103+
ENV NV_LIBNPP_DEV_PACKAGE libnpp-devel-12-4-${NV_LIBNPP_DEV_VERSION}
104+
ENV NV_LIBNCCL_DEV_PACKAGE_NAME libnccl-devel
105+
ENV NV_LIBNCCL_DEV_PACKAGE_VERSION 2.21.5-1
106+
ENV NCCL_VERSION 2.21.5
107+
ENV NV_LIBNCCL_DEV_PACKAGE ${NV_LIBNCCL_DEV_PACKAGE_NAME}-${NV_LIBNCCL_DEV_PACKAGE_VERSION}+cuda12.4
108+
ENV NV_CUDA_NSIGHT_COMPUTE_VERSION 12.4.1-1
109+
ENV NV_CUDA_NSIGHT_COMPUTE_DEV_PACKAGE cuda-nsight-compute-12-4-${NV_CUDA_NSIGHT_COMPUTE_VERSION}
110+
111+
RUN yum install -y \
112+
make \
113+
findutils \
114+
cuda-command-line-tools-12-4-${NV_CUDA_LIB_VERSION} \
115+
cuda-libraries-devel-12-4-${NV_CUDA_LIB_VERSION} \
116+
cuda-minimal-build-12-4-${NV_CUDA_LIB_VERSION} \
117+
cuda-cudart-devel-12-4-${NV_CUDA_CUDART_DEV_VERSION} \
118+
${NV_NVPROF_DEV_PACKAGE} \
119+
cuda-nvml-devel-12-4-${NV_NVML_DEV_VERSION} \
120+
libcublas-devel-12-4-${NV_LIBCUBLAS_DEV_VERSION} \
121+
${NV_LIBNPP_DEV_PACKAGE} \
122+
${NV_LIBNCCL_DEV_PACKAGE} \
123+
${NV_CUDA_NSIGHT_COMPUTE_DEV_PACKAGE} \
124+
&& yum clean all \
125+
&& rm -rf /var/cache/yum/*
126+
127+
ENV LIBRARY_PATH /usr/local/cuda/lib64/stubs
128+
129+
# Install CUDA devel cudnn8 from:
130+
# hhttps://gitlab.com/nvidia/container-images/cuda/-/blob/master/dist/12.4.1/ubi9/devel/cudnn/Dockerfile
131+
ENV NV_CUDNN_VERSION 9.1.0.70-1
132+
ENV NV_CUDNN_PACKAGE libcudnn9-cuda-12-${NV_CUDNN_VERSION}
133+
ENV NV_CUDNN_PACKAGE_DEV libcudnn9-devel-cuda-12-${NV_CUDNN_VERSION}
134+
135+
LABEL com.nvidia.cudnn.version="${NV_CUDNN_VERSION}"
136+
137+
RUN yum install -y \
138+
${NV_CUDNN_PACKAGE} \
139+
${NV_CUDNN_PACKAGE_DEV} \
140+
&& yum clean all \
141+
&& rm -rf /var/cache/yum/*
142+
143+
# Install CUDA toolkit 12.4
144+
RUN yum -y install cuda-toolkit-12-4 && \
145+
yum -y clean all --enablerepo="*"
146+
147+
# Set this flag so that libraries can find the location of CUDA
148+
ENV XLA_FLAGS=--xla_gpu_cuda_data_dir=/usr/local/cuda
149+
150+
# Unregister the system
151+
RUN subscription-manager remove --all && subscription-manager unregister && subscription-manager clean
152+
153+
# Restore notebook user workspace
154+
USER 1001
155+
WORKDIR /opt/app-root/src

0 commit comments

Comments
 (0)