mirror of
				https://github.com/PaddlePaddle/FastDeploy.git
				synced 2025-10-31 03:46:40 +08:00 
			
		
		
		
	[LLM] fix multinode bugs (#2945)
* [LLM] fix multinode bugs * [LLM] fix multinode bugs * [LLM] fix multinode bugs * [LLM] fix ci bugs * fix ci bugs * fix ci bugs
This commit is contained in:
		| @@ -122,8 +122,8 @@ async def lifespan(app: FastAPI): | ||||
|                                  args.mm_processor_kwargs, args.enable_mm, | ||||
|                                  args.reasoning_parser) | ||||
|     app.state.dynamic_load_weight = args.dynamic_load_weight | ||||
|     chat_handler = OpenAIServingChat(engine_client, pid, args.dist_init_ip) | ||||
|     completion_handler = OpenAIServingCompletion(engine_client, pid, args.dist_init_ip) | ||||
|     chat_handler = OpenAIServingChat(engine_client, pid, args.ips) | ||||
|     completion_handler = OpenAIServingCompletion(engine_client, pid, args.ips) | ||||
|     engine_client.create_zmq_client(model=pid, mode=zmq.PUSH) | ||||
|     engine_client.pid = pid | ||||
|     app.state.engine_client = engine_client | ||||
|   | ||||
| @@ -40,15 +40,19 @@ class OpenAIServingChat: | ||||
|     OpenAI-style chat completions serving | ||||
|     """ | ||||
|  | ||||
|     def __init__(self, engine_client, pid, dist_init_ip): | ||||
|     def __init__(self, engine_client, pid, ips): | ||||
|         self.engine_client = engine_client | ||||
|         self.pid = pid | ||||
|         self.master_ip = dist_init_ip | ||||
|         self.master_ip = ips | ||||
|         self.host_ip = get_host_ip() | ||||
|  | ||||
|      | ||||
|     def _check_master(self): | ||||
|         if self.master_ip is None: | ||||
|             return True | ||||
|         if isinstance(self.master_ip, list): | ||||
|             self.master_ip = self.master_ip[0] | ||||
|         else: | ||||
|             self.master_ip = self.master_ip.split(",")[0] | ||||
|         if self.host_ip == self.master_ip: | ||||
|             return True | ||||
|         return False | ||||
|   | ||||
| @@ -45,15 +45,19 @@ from fastdeploy.engine.request import RequestOutput | ||||
|  | ||||
|  | ||||
| class OpenAIServingCompletion: | ||||
|     def __init__(self, engine_client, pid, dist_init_ip): | ||||
|     def __init__(self, engine_client, pid, ips): | ||||
|         self.engine_client = engine_client | ||||
|         self.pid = pid | ||||
|         self.master_ip = dist_init_ip | ||||
|         self.master_ip = ips | ||||
|         self.host_ip = get_host_ip() | ||||
|  | ||||
|     def _check_master(self): | ||||
|         if self.master_ip is None: | ||||
|             return True | ||||
|         if isinstance(self.master_ip, list): | ||||
|             self.master_ip = self.master_ip[0] | ||||
|         else: | ||||
|             self.master_ip = self.master_ip.split(",")[0] | ||||
|         if self.host_ip == self.master_ip: | ||||
|             return True | ||||
|         return False | ||||
|   | ||||
		Reference in New Issue
	
	Block a user
	 ltd0924
					ltd0924