LM Studio 本地推理¶

# Linux 安装示例
chmod +x LM-Studio-*-x86_64.AppImage
./LM-Studio-*-x86_64.AppImage

# 测试API
curl http://localhost:1234/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "loaded-model-name",
    "messages": [
      {"role": "user", "content": "Hello!"}
    ]
  }'

from openai import OpenAI

# 指向本地LM Studio服务器
client = OpenAI(
    base_url="http://localhost:1234/v1",
    api_key="lm-studio"  # 任意字符串即可
)

response = client.chat.completions.create(
    model="any-model-name",  # LM Studio中加载的模型
    messages=[
        {"role": "system", "content": "你是一个有帮助的助手"},
        {"role": "user", "content": "用Python写一个快速排序"}
    ],
    temperature=0.7,
)

print(response.choices[0].message.content)

# 推理参数
Temperature: 0.7      # 控制随机性 (0=确定, 2=高随机)
Top P: 0.9            # 核采样概率阈值
Top K: 40             # 只从概率最高的K个token中采样
Max Tokens: 2048      # 最大生成长度
Repeat Penalty: 1.1   # 重复惩罚系数

# 系统配置
Context Length: 4096   # 上下文窗口
GPU Layers: -1         # -1=全部offload到GPU
Threads: 8             # CPU线程数
Batch Size: 512        # 批处理大小

# 代码生成预设
Name: "Code Generation"
Temperature: 0.2
Top P: 0.95
Repeat Penalty: 1.0
System Prompt: "You are an expert programmer..."

# 创意写作预设
Name: "Creative Writing"
Temperature: 0.9
Top P: 0.95
Repeat Penalty: 1.2
System Prompt: "You are a creative writer..."

# 使用LM Studio的embedding端点
response = client.embeddings.create(
    model="nomic-embed-text-v1.5",
    input="这是一段测试文本"
)

embedding = response.data[0].embedding
print(f"维度: {len(embedding)}")

from langchain_openai import ChatOpenAI

llm = ChatOpenAI(
    base_url="http://localhost:1234/v1",
    api_key="lm-studio",
    model="local-model",
    temperature=0.3
)

response = llm.invoke("解释什么是RAG")
print(response.content)

from llama_index.llms.openai_like import OpenAILike

llm = OpenAILike(
    api_base="http://localhost:1234/v1",
    api_key="lm-studio",
    model="local-model",
    is_chat_model=True
)

import asyncio
from openai import AsyncOpenAI

client = AsyncOpenAI(
    base_url="http://localhost:1234/v1",
    api_key="lm-studio"
)

async def process_batch(prompts):
    tasks = [
        client.chat.completions.create(
            model="local-model",
            messages=[{"role": "user", "content": p}]
        )
        for p in prompts
    ]
    return await asyncio.gather(*tasks)

prompts = ["翻译: Hello", "翻译: World", "翻译: AI"]
results = asyncio.run(process_batch(prompts))

stream = client.chat.completions.create(
    model="local-model",
    messages=[{"role": "user", "content": "写一首诗"}],
    stream=True
)

for chunk in stream:
    if chunk.choices[0].delta.content:
        print(chunk.choices[0].delta.content, end="", flush=True)

# 某些框架需要指定model名称为LM Studio中显示的名称
# 查看当前加载的模型
import requests
r = requests.get("http://localhost:1234/v1/models")
print(r.json())