chore: import upstream snapshot with attribution

python/ray/llm/examples/sglang/serve_sglang_example.py

@@ -0,0 +1,33 @@
+"""Single-node SGLang serving example using Ray Serve LLM.
+
+Usage:
+    RAY_EXPERIMENTAL_NOSET_CUDA_VISIBLE_DEVICES=0 serve run serve_sglang_example:app
+"""
+
+from ray import serve
+from ray.llm._internal.serve.engines.sglang import SGLangServer
+from ray.serve.llm import LLMConfig, build_openai_app
+
+llm_config = LLMConfig(
+    model_loading_config={
+        "model_id": "Llama-3.1-8B-Instruct",
+        "model_source": "unsloth/Llama-3.1-8B-Instruct",
+    },
+    deployment_config={
+        "autoscaling_config": {
+            "min_replicas": 1,
+            "max_replicas": 2,
+        }
+    },
+    server_cls=SGLangServer,
+    engine_kwargs={
+        "trust_remote_code": True,
+        "model_path": "unsloth/Llama-3.1-8B-Instruct",
+        "tp_size": 1,
+        "mem_fraction_static": 0.8,
+    },
+)
+
+app = build_openai_app({"llm_configs": [llm_config]})
+serve.start()
+serve.run(app, blocking=True)