support deepseek quant & mix-parallel with graphmode (#585)

### What this PR does / why we need it?
1. support deepseek with w8a8 quant;
2. support deepseek with mix-parallel(multi-DP, EP+TP);
3. support deepseek with graphmode.
---------

Signed-off-by: wen-jie666 <wenjie39@huawei.com>
Signed-off-by: Yizhou Liu <liuyizhou5@h-partners.com>
Signed-off-by: libaokui <libaokui@huawei.com>
Signed-off-by: linfeng-yuan <1102311262@qq.com>
Co-authored-by: wen-jie666 <wenjie39@huawei.com>

examples/dp_offline/data_parallel.py

@@ -11,8 +11,6 @@
 import gc
 import os
 
-VLLM_ENABLE_GRAPGH_MODE = os.environ.get("VLLM_ENABLE_GRAPH_MODE") == "1"
-
 
 def main():
     dp_rank = int(os.environ['RANK'])
@@ -20,8 +18,8 @@ def main():
     dp_size = int(os.environ['WORLD_SIZE'])
     master_addr = os.environ['MASTER_ADDR']
     master_port = os.environ['MASTER_PORT']
-    tp_size = 4
-    etp_size = 2
+    tp_size = 1
+    etp_size = 1
 
     os.environ["VLLM_DP_RANK"] = str(dp_rank)
     os.environ["VLLM_DP_SIZE"] = str(dp_size)
@@ -58,15 +56,15 @@ def main():
                                      max_tokens=4,
                                      min_tokens=4)
     # Create an LLM.
-    llm = LLM(
-        model="deepseek-ai/DeepSeek-V2-Lite-Chat",
-        tensor_parallel_size=tp_size,
-        trust_remote_code=True,
-        expert_tensor_parallel_size=etp_size,
-        max_model_len=4096,
-        max_num_seqs=num_seqs,
-        compilation_config=1 if VLLM_ENABLE_GRAPGH_MODE else 0,
-    )
+    llm = LLM(model="deepseek-ai/DeepSeek-V2-Lite-Chat",
+              tensor_parallel_size=tp_size,
+              trust_remote_code=True,
+              max_model_len=4096,
+              max_num_seqs=num_seqs,
+              additional_config={
+                  'expert_tensor_parallel_size': etp_size,
+                  'enable_graph_mode': False,
+              })
 
     outputs = llm.generate(prompts, sampling_params)
     for output in outputs:

examples/dp_offline/run_dp.sh

@@ -6,15 +6,13 @@ export HCCL_SOCKET_IFNAME=${ifname}
 # dp_size = node_size * dp_per_node
 node_size=1
 node_rank=0
-dp_per_node=2
+dp_per_node=4
 master_addr=127.0.0.1
 master_port=12345
 
 rm -rf ./.torchair_cache/
 rm -rf ./dynamo_*
 rm -rf /root/ascend/log/debug/plog/*
-export VLLM_ENABLE_GRAPH_MODE=0
-export VLLM_ENABLE_MC2=0
 
 torchrun --nproc_per_node ${dp_per_node} --nnodes ${node_size} \
     --node_rank ${node_rank} --master_addr ${master_addr} --master_port ${master_port} \