使用异步的上下文管理器，启动加速

api.py

@@ -1,9 +1,9 @@
 import datetime
-import json
+from contextlib import asynccontextmanager
 
 import torch
 import uvicorn
-from fastapi import FastAPI, Request
+from fastapi import FastAPI
 from pydantic import BaseModel
 from transformers import AutoModel, AutoTokenizer
 
@@ -11,6 +11,21 @@ DEVICE = "cuda"
 DEVICE_ID = "0"
 CUDA_DEVICE = f"{DEVICE}:{DEVICE_ID}" if DEVICE_ID else DEVICE
 
+models = {}
+
+@asynccontextmanager
+async def lifespan(app: FastAPI):
+    models['chat'] = AutoModel.from_pretrained(
+        "THUDM/models", 
+        trust_remote_code=True).half().cuda()
+    models['chat'].eval()
+    models['tokenizer'] = AutoTokenizer.from_pretrained(
+        "THUDM/models", 
+        trust_remote_code=True)
+    yield
+    for model in models.values():
+        del model
+        torch_gc()
 
 def torch_gc():
     if torch.cuda.is_available():
@@ -18,8 +33,7 @@ def torch_gc():
             torch.cuda.empty_cache()
             torch.cuda.ipc_collect()
 
-
-app = FastAPI()
+app = FastAPI(lifespan=lifespan)
 
 class Item(BaseModel):
     prompt: str
@@ -35,38 +49,18 @@ class Answer(BaseModel):
     time: str
 
 @app.post("/")
-async def create_item(request: Request):
-    global model, tokenizer
-    json_post_raw = await request.json()
-    json_post = json.dumps(json_post_raw)
-    json_post_list = json.loads(json_post)
-    prompt = json_post_list.get('prompt')
-    history = json_post_list.get('history')
-    max_length = json_post_list.get('max_length')
-    top_p = json_post_list.get('top_p')
-    temperature = json_post_list.get('temperature')
-    response, history = model.chat(tokenizer,
-                                   prompt,
-                                   history=history,
-                                   max_length=max_length if max_length else 2048,
-                                   top_p=top_p if top_p else 0.7,
-                                   temperature=temperature if temperature else 0.95)
+async def create_item(item: Item):
+    response, history = models['chat'].chat(
+        models['tokenizer'],
+        item.prompt,
+        history=item.history,
+        max_length=item.max_length,
+        top_p=item.top_p,
+        temperature=item.temperature)
     now = datetime.datetime.now()
     time = now.strftime("%Y-%m-%d %H:%M:%S")
-    answer = {
-        "response": response,
-        "history": history,
-        "status": 200,
-        "time": time
-    }
-    log = "[" + time + "] " + '", prompt:"' + prompt + '", response:"' + repr(response) + '"'
-    print(log)
-    torch_gc()
-    return answer
-
+    print(f"[{time}] prompt: '{item.prompt}', response: '{response}'")
+    return Answer(response=response, history=history, status=200, time=time)
 
 if __name__ == '__main__':
-    tokenizer = AutoTokenizer.from_pretrained("THUDM/chatglm-6b", trust_remote_code=True)
-    model = AutoModel.from_pretrained("THUDM/chatglm-6b", trust_remote_code=True).half().cuda()
-    model.eval()
     uvicorn.run(app, host='0.0.0.0', port=8000, workers=1)