Skip to content

Commit e54cccb

Browse files
committed
Merge remote-tracking branch 'origin/yinggeh-DLIS-8435-fix-dcgm-requirement' into mwittwer/explicit_model_load_parsing
2 parents c7d67f4 + 85897d7 commit e54cccb

File tree

2 files changed

+11
-5
lines changed

2 files changed

+11
-5
lines changed

Dockerfile.sdk

Lines changed: 3 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -265,7 +265,9 @@ RUN if [ "$TRITON_ENABLE_GPU" = "ON" ]; then \
265265
curl -o /tmp/cuda-keyring.deb \
266266
https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2404/$arch/cuda-keyring_1.1-1_all.deb \
267267
&& apt install /tmp/cuda-keyring.deb && rm /tmp/cuda-keyring.deb && \
268-
apt update && apt install --yes datacenter-gpu-manager-4-dev=1:${DCGM_VERSION}; \
268+
apt update && \
269+
apt install --yes datacenter-gpu-manager-4-core=1:${DCGM_VERSION} && \
270+
apt install --yes datacenter-gpu-manager-4-dev=1:${DCGM_VERSION}; \
269271
fi
270272

271273
# Build expects "python" executable (not python3).

build.py

Lines changed: 8 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -858,19 +858,21 @@ def install_dcgm_libraries(dcgm_version, target_machine):
858858
# Install DCGM. Steps from https://developer.nvidia.com/dcgm#Downloads
859859
RUN dnf config-manager --add-repo https://developer.download.nvidia.com/compute/cuda/repos/rhel8/sbsa/cuda-rhel8.repo \\
860860
&& dnf clean expire-cache \\
861+
&& dnf install --assumeyes datacenter-gpu-manager-4-core=1:{} \\
861862
&& dnf install --assumeyes datacenter-gpu-manager-4-devel=1:{}
862863
""".format(
863-
dcgm_version, dcgm_version
864+
dcgm_version, dcgm_version, dcgm_version
864865
)
865866
else:
866867
return """
867868
ENV DCGM_VERSION {}
868869
# Install DCGM. Steps from https://developer.nvidia.com/dcgm#Downloads
869870
RUN dnf config-manager --add-repo https://developer.download.nvidia.com/compute/cuda/repos/rhel8/x86_64/cuda-rhel8.repo \\
870871
&& dnf clean expire-cache \\
872+
&& dnf install --assumeyes datacenter-gpu-manager-4-core=1:{} \\
871873
&& dnf install --assumeyes datacenter-gpu-manager-4-devel=1:{}
872874
""".format(
873-
dcgm_version, dcgm_version
875+
dcgm_version, dcgm_version, dcgm_version
874876
)
875877
else:
876878
if target_machine == "aarch64":
@@ -882,9 +884,10 @@ def install_dcgm_libraries(dcgm_version, target_machine):
882884
&& apt install /tmp/cuda-keyring.deb \\
883885
&& rm /tmp/cuda-keyring.deb \\
884886
&& apt update \\
887+
&& apt install --yes datacenter-gpu-manager-4-core=1:{} \\
885888
&& apt install --yes datacenter-gpu-manager-4-dev=1:{}
886889
""".format(
887-
dcgm_version, dcgm_version
890+
dcgm_version, dcgm_version, dcgm_version
888891
)
889892
else:
890893
return """
@@ -895,9 +898,10 @@ def install_dcgm_libraries(dcgm_version, target_machine):
895898
&& apt install /tmp/cuda-keyring.deb \\
896899
&& rm /tmp/cuda-keyring.deb \\
897900
&& apt update \\
901+
&& apt install --yes datacenter-gpu-manager-4-core=1:{} \\
898902
&& apt install --yes datacenter-gpu-manager-4-dev=1:{}
899903
""".format(
900-
dcgm_version, dcgm_version
904+
dcgm_version, dcgm_version, dcgm_version
901905
)
902906

903907

0 commit comments

Comments
 (0)