【AI革命】RouteLLM:让大语言模型降本85%的神奇技术!
【AI革命】RouteLLM:让大语言模型降本85%的神奇技术!
pip install "routellm[serve,eval]"
export OPENAI_API_KEY="sk-proj-wDPaUDu1I"
export ANYSCALE_API_KEY="secret_r4jh"
#test question
#what is the square root of 1787569?
import os
from routellm.controller import Controller
os.environ["OPENAI_API_KEY"] = "sk-proj-wDPaUDu1Iimxi8SG3BfwT3B5M"
# Replace with your model provider, we use Anyscale's Mixtral here.
os.environ["ANYSCALE_API_KEY"] = "esecret_r4jhxj9jx75mewwt3"
client = Controller(
routers=["mf"],
strong_model="gpt-4-1106-preview",
weak_model="anyscale/mistralai/Mixtral-8x7B-Instruct-v0.1",
)
response = client.chat.completions.create(
# This tells RouteLLM to use the MF router with a cost threshold of 0.11593
model="router-mf-0.11593",
messages=[
{"role": "user", "content": "Hello!"}
]
)
ollama
import os
from routellm.controller import Controller
# 设置 OpenAI API 密钥
client = Controller(
routers=["mf"],
strong_model="gpt-4-1106-preview",
weak_model="ollama_chat/gemma2",
)
response = client.chat.completions.create(
model="router-mf-0.11593",
messages=[
{"role": "user", "content": "Hello!"}
]
)
print(response.choices[0].message.content)
👉👉👉如有问题请联系我的徽信 stoeng
🔥🔥🔥本项目代码由AI超元域频道制作,观看更多大模型微调视频请访问我的频道⬇
👉👉👉 ** 我的哔哩哔哩频道 **
👉👉👉 ** 我的YouTube频道 **
👉👉👉我的开源项目 ** https://github.com/win4r/AISuperDomain **
stream
import os
from routellm.controller import Controller
client = Controller(
routers=["mf"],
strong_model="gpt-4-1106-preview",
weak_model="ollama_chat/gemma2",
)
# 创建流式响应
stream = client.chat.completions.create(
model="router-mf-0.11593",
messages=[
{"role": "user", "content": "who are you?"}
],
stream=True # 启用流式输出
)
# 遍历流并打印每个部分的内容
for chunk in stream:
if chunk.choices[0].delta.content is not None:
print(chunk.choices[0].delta.content, end='', flush=True)
# 打印换行符,使下一个提示符出现在新行
print()
###