vllm代码提交

llm/README.md

@@ -25,6 +25,10 @@ curl -X POST "http://127.0.0.1:8101/v1/completions" \
 -H "Content-Type: application/json" \
 -d "{\"model\": \"THUDM/chatglm3-6b\",\"prompt\": \"你叫什么名字\", \"history\": [{\"role\": \"user\", \"content\": \"你出生在哪里.\"}, {\"role\": \"assistant\", \"content\": \"出生在北京\"}]}"
 
+多轮对话
+curl -X POST "http://127.0.0.1:8101/v1/chat/completions" \
+-H "Content-Type: application/json" \
+-d "{\"model\": \"THUDM/chatglm3-6b\", \"messages\": [{\"role\": \"system\", \"content\": \"You are ChatGLM3, a large language model trained by Zhipu.AI. Follow the user's instructions carefully. Respond using markdown.\"}, {\"role\": \"user\", \"content\": \"你好，给我讲一个故事，大概100字\"}], \"stream\": false, \"max_tokens\": 100, \"temperature\": 0.8, \"top_p\": 0.8}"
 
 
 

llm/VllmGPT.py

@@ -2,37 +2,80 @@ import json
 import requests
 # from core import content_db
 
+class VllmGPT:
 
-def question(cont):
-    chat_list = []
-    # contentdb = content_db.new_instance()
-    # list = contentdb.get_list('all','desc',11)
-    # answer_info = dict()
-    # chat_list = []
-    # i = len(list)-1
-    # while i >= 0:
-    #     answer_info = dict()
-    #     if list[i][0] == "member":
-    #         answer_info["role"] = "user"
-    #         answer_info["content"] = list[i][2]
-    #     elif list[i][0] == "fay":
-    #         answer_info["role"] = "bot"
-    #         answer_info["content"] = list[i][2]
-    #     chat_list.append(answer_info)
-    #     i -= 1
-    content = {
-        "model": "THUDM/chatglm3-6b",
-        "prompt":"请简单回复我。" +  cont,
-        "history":chat_list}
-    url = "http://192.168.1.3:8101/v1/completions"
-    req = json.dumps(content)
-    
-    headers = {'content-type': 'application/json'}
-    r = requests.post(url, headers=headers, data=req)
-    res = json.loads(r.text)
-    
-    return res['choices'][0]['text']
+    def __init__(self, host="127.0.0.1",
+                 port="8000",
+                 model="THUDM/chatglm3-6b",
+                 max_tokens="1024"):
+        self.host = host
+        self.port = port
+        self.model=model
+        self.max_tokens=max_tokens
+        self.__URL = "http://{}:{}/v1/completions".format(self.host, self.port)
+        self.__URL2 = "http://{}:{}/v1/chat/completions".format(self.host, self.port)
 
+    def question(self,cont):
+        chat_list = []
+        # contentdb = content_db.new_instance()
+        # list = contentdb.get_list('all','desc',11)
+        # answer_info = dict()
+        # chat_list = []
+        # i = len(list)-1
+        # while i >= 0:
+        #     answer_info = dict()
+        #     if list[i][0] == "member":
+        #         answer_info["role"] = "user"
+        #         answer_info["content"] = list[i][2]
+        #     elif list[i][0] == "fay":
+        #         answer_info["role"] = "bot"
+        #         answer_info["content"] = list[i][2]
+        #     chat_list.append(answer_info)
+        #     i -= 1
+        content = {
+            "model": self.model,
+            "prompt":"请简单回复我。" +  cont,
+            "history":chat_list}
+        url = self.__URL
+        req = json.dumps(content)
+        
+        headers = {'content-type': 'application/json'}
+        r = requests.post(url, headers=headers, data=req)
+        res = json.loads(r.text)
+        
+        return res['choices'][0]['text']
+
+    def question2(self,cont):
+        chat_list = []
+        # contentdb = content_db.new_instance()
+        # list = contentdb.get_list('all','desc',11)
+        # answer_info = dict()
+        # chat_list = []
+        # i = len(list)-1
+        # while i >= 0:
+        #     answer_info = dict()
+        #     if list[i][0] == "member":
+        #         answer_info["role"] = "user"
+        #         answer_info["content"] = list[i][2]
+        #     elif list[i][0] == "fay":
+        #         answer_info["role"] = "bot"
+        #         answer_info["content"] = list[i][2]
+        #     chat_list.append(answer_info)
+        #     i -= 1
+        content = {
+            "model": self.model,
+            "prompt":"请简单回复我。" +  cont,
+            "history":chat_list}
+        url = self.__URL2
+        req = json.dumps(content)
+        
+        headers = {'content-type': 'application/json'}
+        r = requests.post(url, headers=headers, data=req)
+        res = json.loads(r.text)
+        
+        return res['choices'][0]['message']['content']
+    
 if __name__ == "__main__":
-    req = question("你叫什么名字啊今年多大了")
-    print(req)
+    vllm = VllmGPT('192.168.1.3','8101')
+    req = vllm.question("你叫什么名字啊今年多大了")
+    print(req)