Move KV cache creation to balance_serve

ktransformers/server/backend/interfaces/balance_serve.py

@@ -30,6 +30,7 @@ from ktransformers.server.balance_serve.sched_rpc import SchedulerClient
 from ktransformers.server.balance_serve.settings import sched_ext
 from torch.multiprocessing import Queue
 import torch.multiprocessing as mp
+from multiprocessing.synchronize import Event
 from ktransformers.server.schemas.endpoints.chat import RawUsage
 from ktransformers.server.utils.multi_timer import Profiler
 import zmq
@@ -41,8 +42,10 @@ import threading
 from contextlib import asynccontextmanager
 from fastapi import FastAPI, Request
 import os
-
+import pickle
-
+import subprocess
+import tempfile
+import atexit
 
 ktransformer_rules_dir = (
     os.path.join(os.path.dirname(os.path.abspath(__file__)), "..", "..", "..", "./optimize/optimize_rules/") 
@@ -99,7 +102,7 @@ class Engine:
     sampler: Sampler
     query_manager: QueryManager
     cache: KDeepSeekV3Cache
-    def __init__(self, args: ConfigArgs = default_args, generated_token_queue:Queue = None, broadcast_endpoint: str = None):
+    def __init__(self, args: ConfigArgs = default_args, generated_token_queue:Queue = None, broadcast_endpoint: str = None, kvcache_event: Event = None):
         self.args = args
 
         # 子进程和父进程无法共享 config 变量
@@ -115,14 +118,6 @@ class Engine:
             
         self.gen_queue = generated_token_queue
             
-        print(f"Getting inference context from sched_client.")
-        inference_context = self.sched_client.get_inference_context_raw()
-        print(f"Got inference context, sending it to subscribers.")
-        inference_context = self.sched_client.rebuild_inferece_context(inference_context)
-        self.cache.load(inference_context)
-        print(f"kv_cache loaded successfully.")
-
-        self.block_num = inference_context.k_cache[0].size(1)
         with torch.device("meta"):
             if config.architectures[0] == "DeepseekV3ForCausalLM":
                 self.model = KDeepseekV3ForCausalLM(config, self.cache)
@@ -165,6 +160,17 @@ class Engine:
             self.model.generation_config.pad_token_id = self.model.generation_config.eos_token_id
 
         self.model.eval()
+        kvcache_event.set()
+        # load kvcache
+        print(f"Getting inference context from sched_client.")
+        inference_context = self.sched_client.get_inference_context_raw()
+        print(f"Got inference context, sending it to subscribers.")
+        inference_context = self.sched_client.rebuild_inferece_context(inference_context)
+        self.cache.load(inference_context)
+        print(f"kv_cache loaded successfully.")
+        
+
+        self.block_num = inference_context.k_cache[0].size(1)
         #@TODO add config
         self.model.init_wrapper(self.args.use_cuda_graph, self.device, args.max_batch_size, self.block_num)
 
@@ -240,8 +246,8 @@ class BalanceServeThreadContext(ThreadContext):
         return local_messages
     
 
-def run_engine(args, token_queue, broadcast_endpoint, event):
+def run_engine(args, token_queue, broadcast_endpoint, event, kvcache_event):
-    engine = Engine(args, token_queue, broadcast_endpoint)
+    engine = Engine(args, token_queue, broadcast_endpoint, kvcache_event)
     if args.use_cuda_graph:
         engine.model_runner.warmup()
         
@@ -278,10 +284,34 @@ class BalanceServeInterface(BackendInterfaceBase):
         self.streamer = TextStreamer(self.tokenizer)
 
         start_event = ctx.Event()
+        kvcache_event = ctx.Event()
 
-        p = ctx.Process(target=run_engine, args=(self.args, self.token_queue, self.broadcast_endpoint, start_event))
+        p = ctx.Process(target=run_engine, args=(self.args, self.token_queue, self.broadcast_endpoint, start_event, kvcache_event))
         p.start()
         processes.append(p)
+        kvcache_event.wait()
+
+
+        def cleanup():
+            if sched_process.poll() is None:
+                sched_process.terminate()
+
+        with tempfile.NamedTemporaryFile(delete=False) as temp_file:
+            pickle.dump(args, temp_file)
+            temp_file_path = temp_file.name
+        current_file = __file__
+        target_file = os.path.join(os.path.dirname(current_file), "..", "..", "balance_serve", "sched_rpc.py")
+        target_file = os.path.normpath(target_file)
+        log_path = os.path.join(args.log_dir, "rpc.log")
+        log = open(log_path, "a") 
+        sched_process = subprocess.Popen(
+            ["python3", target_file, "--config", temp_file_path], 
+            stdout=log, 
+            stderr=log
+        )
+        print("sched_rpc started with PID:", sched_process.pid)
+        atexit.register(cleanup)
+
         start_event.wait()
     
     def get_sampling_params(self, temperature: Optional[float] = None, top_p: Optional[float] = None) -> tuple[float, float]:

ktransformers/server/main.py

@@ -5,7 +5,6 @@ from fastapi.staticfiles import StaticFiles
 import uvicorn.logging
 import uvicorn
 import sys
-import atexit
 project_dir = os.path.dirname(os.path.dirname(os.path.dirname(__file__)))
 from fastapi.middleware.cors import CORSMiddleware
 from ktransformers.server.args import ArgumentParser
@@ -17,8 +16,7 @@ from fastapi.middleware.cors import CORSMiddleware
 from ktransformers.server.api import router, post_db_creation_operations
 from ktransformers.server.utils.sql_utils import Base, SQLUtil
 from ktransformers.server.config.log import logger
-import subprocess
+
-import tempfile
 
 def mount_app_routes(mount_app: FastAPI):
     sql_util = SQLUtil()
@@ -108,27 +106,6 @@ def main():
     arg_parser = ArgumentParser(cfg)
 
     args = arg_parser.parse_args()
-    if args.backend_type == "balance_serve":
-        import pickle
-        def cleanup():
-            if sched_process.poll() is None:
-                sched_process.terminate()
-
-        with tempfile.NamedTemporaryFile(delete=False) as temp_file:
-            pickle.dump(args, temp_file)
-            temp_file_path = temp_file.name
-        current_file = __file__
-        target_file = os.path.join(os.path.dirname(current_file), "balance_serve", "sched_rpc.py")
-        target_file = os.path.normpath(target_file)
-        log_path = os.path.join(args.log_dir, "rpc.log")
-        log = open(log_path, "a") 
-        sched_process = subprocess.Popen(
-            ["python3", target_file, "--config", temp_file_path], 
-            stdout=log, 
-            stderr=log
-        )
-        print("sched_rpc started with PID:", sched_process.pid)
-        atexit.register(cleanup)
     create_interface(config=cfg, default_args=cfg)
     app = create_app()
     custom_openapi(app)