change vllm settings

apps/base/customer1/hermes-agent/configmap.yaml

@@ -14,7 +14,7 @@ data:
         type: openai
         api_key: EMPTY
         base_url: http://rtx6000-brain-service.customer1.svc.cluster.local:8000/v1
-        context_length: 131072
+        context_length: 176072
       qwen-vllm:
         type: openai
         api_key: EMPTY

infrastructure/gpus/base/vllm-servers/rtx6000-vllm.yaml

@@ -38,11 +38,12 @@ spec:
             - --tensor-parallel-size=1
             - --tokenizer-mode=hf 
             - --gpu-memory-utilization=0.95
-            - --max-model-len=237864
+            - --max-model-len=98304
             - --enable-auto-tool-choice 
             - --kv-cache-dtype=fp8 
-            - --max-num-batched-tokens=32768
-            - --max-num-seqs=16 
+            - --max-num-batched-tokens=65536 
+            - --max-num-seqs=128
+            - --enable-chunked-prefill
             - --trust-remote-code
             - --dtype=auto
             - --enable-prefix-caching
@@ -59,7 +60,7 @@ spec:
             requests:
               nvidia.com/gpu: 1
               memory: "150Gi"
-              cpu: "24"
+              cpu: "32"
           
           startupProbe:
             httpGet: