File tree Expand file tree Collapse file tree 2 files changed +10
-15
lines changed Expand file tree Collapse file tree 2 files changed +10
-15
lines changed Original file line number Diff line number Diff line change 3333 - name : root
3434 hostPath :
3535 path : " /"
36+ - name : drhpc-results
37+ hostPath :
38+ path : {{ .resultsHostPath | default "/var/lib/oci-dr-hpc-v2" | quote }}
39+ type : DirectoryOrCreate
3640
3741 hostPID : true
3842 hostIPC : true
@@ -58,22 +62,12 @@ spec:
5862 volumeMounts :
5963 - name : root
6064 mountPath : /host
65+ - name : drhpc-results
66+ mountPath : /opt/oci-hpc/oci-dr-hpc
6167
62- {{- if eq .type "nvidia" }}
63- # NVIDIA: Request GPU resources (equivalent to --gpus all)
68+ # Same resource config for both AMD and NVIDIA
6469 resources :
65- limits :
66- nvidia.com/gpu : {{ .gpuCount | default 8 }}
67- {{- with .resources }}
68- {{- toYaml . | nindent 10 }}
69- {{- end }}
70- {{- else }}
71- # AMD: No GPU resource requests needed
72- resources :
73- {{- with .resources }}
74- {{- toYaml . | nindent 10 }}
75- {{- end }}
76- {{- end }}
70+ {{- toYaml .resources | nindent 10 }}
7771
7872 env :
7973 - name : PUSH_GATEWAY
Original file line number Diff line number Diff line change @@ -130,6 +130,7 @@ drhpc:
130130 rocmVersion : rocm # ROCm version prefix
131131 tag : latest # Tag comes after rocm version
132132 pullPolicy : Always
133+ resultsHostPath : " /var/lib/oci-dr-hpc-v2"
133134 nodeSelector :
134135 amd.com/gpu : " true"
135136 tolerations :
@@ -146,7 +147,7 @@ drhpc:
146147 cudaVersion : cuda-12.9.0 # CUDA version prefix
147148 tag : latest # Tag comes after cuda version
148149 pullPolicy : Always
149- gpuCount : 8 # Request all GPUs
150+ resultsHostPath : " /var/lib/oci-dr-hpc-v2 "
150151 nodeSelector :
151152 nvidia.com/gpu : " true"
152153 tolerations :
You can’t perform that action at this time.
0 commit comments