[build]: Updating vllm deployment to the latest image and scorers (#112)

Update the vLLM P2P deployment to support KV-cache and load scorers.

Signed-off-by: Kfir Toledo <[email protected]>

Author: kfirtoledo

deploy/components/vllm-p2p/vllm-deployment.yaml

@@ -31,13 +31,12 @@ spec:
             - "-c"
           args:
             - |
-              export LMCACHE_DISTRIBUTED_URL=$${${POD_IP}}:80 && \
+              export LMCACHE_DISTRIBUTED_URL=$${${POD_IP}} && \
               vllm serve ${MODEL_NAME} \
               --host 0.0.0.0 \
               --port 8000 \
-              --enable-chunked-prefill false \
               --max-model-len ${MAX_MODEL_LEN} \
-              --kv-transfer-config '{"kv_connector":"LMCacheConnector","kv_role":"kv_both"}'
+              --kv-transfer-config '{"kv_connector":"LMCacheConnectorV1","kv_role":"kv_both"}'
           ports:
             - name: http
               containerPort: 8000
@@ -78,6 +77,10 @@ spec:
                 secretKeyRef:
                   name: ${HF_SECRET_NAME}
                   key: ${HF_SECRET_KEY}
+            - name: VLLM_ENABLE_V1_MULTIPROCESSING
+              value: "1"
+            - name: VLLM_WORKER_MULTIPROC_METHOD
+              value: spawn
             - name: LMCACHE_LOOKUP_URL
               value: ${REDIS_HOST}:${REDIS_PORT}
             - name: LMCACHE_ENABLE_DEBUG

deploy/environments/dev/kubernetes-kgateway/patch-deployments.yaml

@@ -29,4 +29,12 @@ spec:
             valueFrom:
               secretKeyRef:
                 name: hf-token
-                key: ${HF_SECRET_KEY}
+                key: ${HF_SECRET_KEY}
+          - name: ENABLE_KVCACHE_AWARE_SCORER
+            value: "true"
+          - name: KVCACHE_AWARE_SCORER_WEIGHT
+            value: "2.0"
+          - name: ENABLE_LOAD_AWARE_SCORER
+            value: "true"
+          - name: LOAD_AWARE_SCORER_WEIGHT
+            value: "1.0"

scripts/kubernetes-dev-env.sh

@@ -65,10 +65,10 @@ case "${VLLM_MODE}" in
       export LORA_ADAPTER_SYNCER_TAG="${LORA_ADAPTER_SYNCER_TAG:-v20250425-ddc3d69}"
 
     elif [[ "$VLLM_MODE" == "vllm-p2p" ]]; then
-      export VLLM_IMAGE="${VLLM_IMAGE:-lmcache/vllm-openai}"
-      export VLLM_TAG="${VLLM_TAG:-2025-03-10}"
-      export EPP_IMAGE="${EPP_IMAGE:- quay.io/vmaroon/gateway-api-inference-extension/epp}"
-      export EPP_TAG="${EPP_TAG:-kv-aware}"
+      export VLLM_IMAGE="${VLLM_IMAGE:-quay.io/llm-d/llm-d-dev}"
+      export VLLM_TAG="${VLLM_TAG:-lmcache-0.0.6-amd64}"
+      export EPP_IMAGE="${EPP_IMAGE:-quay.io/llm-d/llm-d-gateway-api-inference-extension-dev}"
+      export EPP_TAG="${EPP_TAG:-0.0.5-amd64}"
       export MAX_MODEL_LEN="${MAX_MODEL_LEN:-32768}"
       export PVC_NAME="${PVC_NAME:-vllm-p2p-storage-claim}"
       export PVC_ACCESS_MODE="${PVC_ACCESS_MODE:-ReadWriteOnce}"