Merge pull request #3721 from RachaelSTamakloe/rc-cherry-pick-3719

Cherry-pick "use public a4 nccl image"

Author: RachaelSTamakloe

examples/gke-a4-highgpu/nccl-installer.yaml

@@ -54,8 +54,7 @@ spec:
         hostPath:
           path: /home/kubernetes/bin/gib
       initContainers:
-      # TODO: replace with public image
-      - image: us-docker.pkg.dev/kernel-net-team/clouda4-nccl-dev/nccl-plugin-gib-diagnostic:v1.0.3-b200
+      - image: us-docker.pkg.dev/gce-ai-infra/gpudirect-gib/nccl-plugin-gib:v1.0.4
         name: nccl-rdma-installer
         resources:
           requests:

examples/gke-a4-highgpu/nccl-jobset-example.yaml

@@ -126,8 +126,7 @@ spec:
             - name: nccl
               stdin: true
               tty: true
-              # TODO: replace with public image
-              image: us-docker.pkg.dev/gce-ai-infra/gpudirect-gib/nccl-plugin-gib-diagnostic:v1.0.3
+              image: us-docker.pkg.dev/gce-ai-infra/gpudirect-gib/nccl-plugin-gib-diagnostic:v1.0.4
               securityContext:
                 privileged: true
               env: