之前将ChatGLM6B模型下载到本地运行起来了:ChatGLM3-6B上手体验;如果想要用在项目中,那么可以使用API调用的方式进行操作,尤其当你的项目还是不同语言的异构的场景下,其他服务需要调用的时候就可以直接通过请求服务来获取了。
1.运行API模式
ChatGLM3-6B的官方代码里有运行API的bat文件,可以直接点击运行即可,如果在Linux系统或者通过docker部署的方式运行,可以运行openai_api_demo目录下的 openai_api.py文件。
模型默认运行在8000端口
2.接口介绍
接口url :http://127.0.0.1:8000/v1/chat/completions
请求参数:
"functions": functions, # 函数定义
"model": model, # 模型名称
"messages": messages, # 会话历史
"stream": use_stream, # 是否流式响应
"max_tokens": 100, # 最多生成字数
"temperature": 0.8, # 温度
"top_p": 0.8, # 采样概率
请求的方式有流式和非流式请求,对应请求参数中stream的True和False,请求方式的处理可以参照官方给出的方法
import requests
import json
base_url = "http://127.0.0.1:8000"
def create_chat_completion(model, messages, functions, use_stream=False):
data = {
"functions": functions, # 函数定义
"model": model, # 模型名称
"messages": messages, # 会话历史
"stream": use_stream, # 是否流式响应
"max_tokens": 100, # 最多生成字数
"temperature": 0.8, # 温度
"top_p": 0.8, # 采样概率
}
response = requests.post(f"{base_url}/v1/chat/completions", json=data, stream=use_stream)
if response.status_code == 200:
if use_stream:
# 处理流式响应
for line in response.iter_lines():
if line:
decoded_line = line.decode('utf-8')[6:]
try:
response_json = json.loads(decoded_line)
content = response_json.get("choices", [{}])[0].get("delta", {}).get("content", "")
print(content)
except:
print("Special Token:", decoded_line)
else:
# 处理非流式响应
decoded_line = response.json()
content = decoded_line.get("choices", [{}])[0].get("message", "").get("content", "")
print(content)
else:
print("Error:", response.status_code)
return None
3.不同的调用方式
ChatGLM3支持工具调用,api接口也支持了这一特性
- 简单对话
def simple_chat(use_stream=True):
functions = None
chat_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"
}
]
create_chat_completion("chatglm3-6b", messages=chat_messages, functions=functions, use_stream=use_stream)
- 工具调用
def function_chat(use_stream=True):
functions = [
{
"name": "get_current_weather",
"description": "Get the current weather in a given location.",
"parameters": {
"type": "object",
"properties": {
"location": {
"type": "string",
"description": "The city and state, e.g. Beijing",
},
"unit": {"type": "string", "enum": ["celsius", "fahrenheit"]},
},
"required": ["location"],
},
}
]
chat_messages = [
{
"role": "user",
"content": "波士顿天气如何?",
},
{
"role": "assistant",
"content": "get_current_weather\n ```python\ntool_call(location='Beijing', unit='celsius')\n```",
"function_call": {
"name": "get_current_weather",
"arguments": '{"location": "Beijing", "unit": "celsius"}',
},
},
{
"role": "function",
"name": "get_current_weather",
"content": '{"temperature": "12", "unit": "celsius", "description": "Sunny"}',
},
# ... 接下来这段是 assistant 的回复和用户的回复。
# {
# "role": "assistant",
# "content": "根据最新的天气预报,目前北京的天气情况是晴朗的,温度为12摄氏度。",
# },
# {
# "role": "user",
# "content": "谢谢",
# }
]
create_chat_completion("chatglm3-6b", messages=chat_messages, functions=functions, use_stream=use_stream)
- 命令行调用测试
def chatincmd(use_stream=True):
while True:
functions = None
content = input("Question:")
chat_messages = [
{
"role": "user",
"content": content
}
]
create_chat_completion("chatglm3-6b", messages=chat_messages, functions=functions, use_stream=use_stream)