[Bug] Fixed the yaml of deployments in heterogenous GPU settings to m…

…ake KPA scaling work as expected. (#513) Make spec.selector.matchLabels combination unique for deployments in the model. Co-authored-by: Jingyuan Zhang <[email protected]>
vllm-project · Dec 10, 2024 · cc526f7 · cc526f7
1 parent a5e9849
commit cc526f7
Show file tree

Hide file tree

Showing 3 changed files with 3 additions and 0 deletions.
diff --git a/development/app/config/heterogeneous/simulator_a40/patch_deployment_a40.yaml b/development/app/config/heterogeneous/simulator_a40/patch_deployment_a40.yaml
@@ -9,6 +9,7 @@ spec:
   selector:
     matchLabels:
       model.aibrix.ai/name: "llama2-7b"
+      app: "simulator-llama2-7b-a40"
   template:
     metadata:
       labels:

diff --git a/development/app/config/simulator/patch_deployment_a100.yaml b/development/app/config/simulator/patch_deployment_a100.yaml
@@ -10,6 +10,7 @@ spec:
   selector:
     matchLabels:
       model.aibrix.ai/name: "llama2-7b"
+      app: "simulator-llama2-7b-a100"
   template:
     metadata:
       labels:

diff --git a/development/app/config/templates/deployment/deployment.yaml b/development/app/config/templates/deployment/deployment.yaml
@@ -13,6 +13,7 @@ spec:
     matchLabels:
       adapter.model.aibrix.ai/enabled: "true"
       model.aibrix.ai/name: "llama2-7b"
+      app: "mock-llama2-7b"
   template:
     metadata:
       labels: