Merge branch 'v0.6.2-dev' of ssh://10.6.10.68:10022/dcutoolkit/deeplearing/vllm into v0.6.2-dev

examples/offline_streaming_inference_chat_demo.py

+
+'''
+python offline_streaming_inference_chat_demo.py --model /models/llama2/Llama-2-7b-chat-hf  --template template_llama_chat.jinja --dtype float16 --enforce-eager -tp 1 
+'''
 from vllm.sampling_params import SamplingParams
 from vllm.engine.async_llm_engine import AsyncEngineArgs, AsyncLLMEngine
 import asyncio
-from vllm.utils import FlexibleArgumentParser
 from transformers import AutoTokenizer
 import logging
 import argparse