OOM on multiple-ARC with vllm serving

[jessie-zhao]

Run vllm serving test on ARC with below issue: NFO 07-04 19:10:08 async_llm_engine.py:152] Aborted request cmpl-e5fb5cad96e9402dabbbece3611ae22f-0. INFO: 127.0.0.1:41772 - "POST /v1/completions HTTP/1.1" 500 Internal Server Error ERROR: Exception in ASGI application Traceback (most recent call last): File "/usr/local/lib/python3.11/dist-packages/uvicorn/protocols/http/httptools_impl.py", line 399, in run_asgi result = await app( # type: ignore[func-returns-value]

Model: ChatGLM3-6B / Qwen1.5-14 on 2Arc/Qwen1.5-32B on 4Arc HW: XeonW + Arc