v2 app deployment--templating mainly in manifest now.

vllm/deployment.yaml

@@ -19,10 +19,10 @@ spec:
         seccompProfile:
           type: RuntimeDefault
       nodeSelector:
-        nvidia.com/gpu.product: "{{ .apps.vllm.gpuProduct }}"
+        nvidia.com/gpu.product: "{{ .gpuProduct }}"
       containers:
         - name: vllm
-          image: "{{ .apps.vllm.image }}"
+          image: "{{ .image }}"
           imagePullPolicy: IfNotPresent
           securityContext:
             allowPrivilegeEscalation: false
@@ -31,10 +31,10 @@ spec:
               - ALL
             readOnlyRootFilesystem: false
           args:
-            - --model={{ .apps.vllm.model }}
-            - --max-model-len={{ .apps.vllm.maxModelLen }}
-            - --tensor-parallel-size={{ .apps.vllm.tensorParallelSize }}
-            - --gpu-memory-utilization={{ .apps.vllm.gpuMemoryUtilization }}
+            - --model={{ .model }}
+            - --max-model-len={{ .maxModelLen }}
+            - --tensor-parallel-size={{ .tensorParallelSize }}
+            - --gpu-memory-utilization={{ .gpuMemoryUtilization }}
             {{- if .apps.vllm.enforceEager }}
             - --enforce-eager=True
             {{- end }}
@@ -48,13 +48,13 @@ spec:
               containerPort: 8000
           resources:
             requests:
-              cpu: "{{ .apps.vllm.cpuRequest }}"
-              memory: "{{ .apps.vllm.memoryRequest }}"
-              nvidia.com/gpu: {{ .apps.vllm.gpuCount }}
+              cpu: "{{ .cpuRequest }}"
+              memory: "{{ .memoryRequest }}"
+              nvidia.com/gpu: {{ .gpuCount }}
             limits:
-              cpu: "{{ .apps.vllm.cpuLimit }}"
-              memory: "{{ .apps.vllm.memoryLimit }}"
-              nvidia.com/gpu: {{ .apps.vllm.gpuCount }}
+              cpu: "{{ .cpuLimit }}"
+              memory: "{{ .memoryLimit }}"
+              nvidia.com/gpu: {{ .gpuCount }}
           readinessProbe:
             httpGet:
               path: /v1/models