feat: support Hami on k3s and k8s

2025-12-05 17:09:01 +08:00
parent 97c5d559e3
commit 89bc94a6a9
5 changed files with 27 additions and 2 deletions
--- a/vllm-serve/templates/single.yaml
+++ b/vllm-serve/templates/single.yaml
@ -92,12 +92,18 @@ spec:
        resources:
          limits:
            nvidia.com/gpu: "{{ .Values.resources.gpuLimit }}"
+            {{- if and .Values.resources.gpuMem (gt (int .Values.resources.gpuMem) 0) }}
+            nvidia.com/gpumem: {{ .Values.resources.gpuMem }}
+            {{- end }}
            memory: {{ .Values.resources.memoryLimit }}
            ephemeral-storage: 10Gi
            cpu: {{ .Values.resources.cpuRequest }}
          requests:
            ephemeral-storage: 10Gi
            cpu: {{ .Values.resources.cpuRequest }}
+            {{- if and .Values.resources.gpuMem (gt (int .Values.resources.gpuMem) 0) }}
+            nvidia.com/gpumem: {{ .Values.resources.gpuMem }}
+            {{- end }}
        ports:
        - containerPort: 8000
          name: http