chore: clean up Gaudi and VLLM variables for SGLang in Xeon deployment templates

Signed-off-by: vinayK34 <vinay3.kumar@intel.com>

Author: vinayK34

core/helm-charts/sglang/templates/configmap.yaml

@@ -1,28 +1,25 @@
-# Copyright (C) 2025-2026 Intel Corporation
-# SPDX-License-Identifier: Apache-2.0
-{{- $modelName := (default .Values.LLM_MODEL_ID .Values.SERVED_MODEL_NAME) }}
-apiVersion: v1
-kind: ConfigMap
-metadata:
-  name: {{ include "sglang.fullname" . }}-config
-  labels:
-    {{- include "sglang.labels" . | nindent 4 }}
-data:
-  {{- if .Values.global.HF_ENDPOINT }}
-  HF_ENDPOINT: {{ .Values.global.HF_ENDPOINT | quote}}
-  {{- end }}
-  http_proxy: {{ .Values.global.http_proxy | quote }}
-  https_proxy: {{ .Values.global.https_proxy | quote }}
-  no_proxy: {{ .Values.global.no_proxy | quote }}
-  TORCHINDUCTOR_CACHE_DIR: "/tmp"
-  NUMBA_CACHE_DIR: "/tmp"
-  HF_HOME: "/data"
-  OUTLINES_CACHE_DIR: "/tmp/.cache/outlines"  
-  {{- if .Values.runtime }}
-  runtime: {{ .Values.runtime | quote}}
-  {{- end }}
-
-  {{- $modelConfig := (index .Values.modelConfigs $modelName | default dict).configMapValues | default .Values.defaultModelConfigs.configMapValues }}
-  {{- range $key, $value := $modelConfig }}
-  {{ $key }}: {{ $value | quote }}
+# Copyright (C) 2025-2026 Intel Corporation
+# SPDX-License-Identifier: Apache-2.0
+{{- $modelName := (default .Values.LLM_MODEL_ID .Values.SERVED_MODEL_NAME) }}
+apiVersion: v1
+kind: ConfigMap
+metadata:
+  name: {{ include "sglang.fullname" . }}-config
+  labels:
+    {{- include "sglang.labels" . | nindent 4 }}
+data:
+  {{- if .Values.global.HF_ENDPOINT }}
+  HF_ENDPOINT: {{ .Values.global.HF_ENDPOINT | quote}}
+  {{- end }}
+  http_proxy: {{ .Values.global.http_proxy | quote }}
+  https_proxy: {{ .Values.global.https_proxy | quote }}
+  no_proxy: {{ .Values.global.no_proxy | quote }}
+  TORCHINDUCTOR_CACHE_DIR: "/tmp"
+  NUMBA_CACHE_DIR: "/tmp"
+  HF_HOME: "/data"
+  OUTLINES_CACHE_DIR: "/tmp/.cache/outlines"  
+
+  {{- $modelConfig := (index .Values.modelConfigs $modelName | default dict).configMapValues | default .Values.defaultModelConfigs.configMapValues }}
+  {{- range $key, $value := $modelConfig }}
+  {{ $key }}: {{ $value | quote }}
   {{- end }}

core/helm-charts/sglang/values.yaml

@@ -160,20 +160,10 @@ LLM_MODEL_ID: Intel/neural-chat-7b-v3-3
 SERVED_MODEL_NAME: ""
 # Environment variables for sglang (set in configmap):
 # https://docs.sglang.ai/en/latest/getting_started/installation.html#environment-variables
-sglang_CPU_KVCACHE_SPACE: "40"
-sglang_RPC_TIMEOUT: "100000"
-sglang_ALLOW_LONG_MAX_MODEL_LEN: "1"
-sglang_ENGINE_ITERATION_TIMEOUT_S: "120"
-sglang_CPU_NUM_OF_RESERVED_CPU: "0"
-sglang_CONFIGURE_LOGGING: 1
-sglang_NO_USAGE_STATS: 1
-sglang_CPU_SGL_KERNEL: "1"
-DO_NOT_TRACK: 1
+
 
 # CPU optimization settings (can be updated by automation)
 cpu_optimization_enabled: false
-max_num_batched_tokens: "2048"
-max_num_seqs: "256"
 
 # Parallelism configuration (root level for global access)
 tensor_parallel_size: "2"

core/helm-charts/sglang/xeon-values.yaml

@@ -41,13 +41,9 @@ affinity:
           values: ["true"]
 
 # vLLM CPU environment variables
-SGLANG_CPU_SGL_KERNEL: "1"
-SGLANG_NO_USAGE_STATS: 1
-DO_NOT_TRACK: 1
 
 LLM_MODEL_ID: ""
 
-
 modelConfigs:
   "meta-llama/Llama-3.1-8B-Instruct":
     configMapValues: {}