Pull cuda image from nvcr

In the event we are rate-limited or restricted from pulling from DockerHub, this should work around the problem by pulling from Nvidia's site. Issue runtimes/automation/122 Signed-off-by: Adam Brousseau <[email protected]>
mpirvu · Nov 18, 2024 · c0ff1fc · c0ff1fc
1 parent d98d498
commit c0ff1fc
Showing 1 changed file with 1 addition and 1 deletion.
diff --git a/buildenv/jenkins/common/build.groovy b/buildenv/jenkins/common/build.groovy
@@ -828,7 +828,7 @@ def create_docker_image_locally()
         echo 'ARG image
             ARG cuda_ver=12.2.0
             ARG cuda_distro=ubi8
-            FROM nvidia/cuda:${cuda_ver}-devel-${cuda_distro} as cuda
+            FROM nvcr.io/nvidia/cuda:${cuda_ver}-devel-${cuda_distro} as cuda
             FROM $image
             RUN mkdir -p /usr/local/cuda/nvvm
             COPY --from=cuda /usr/local/cuda/include         /usr/local/cuda/include