通义千问Qwen2.5-MAX多语言应用实践指南

pip install dashscope

from dashscope import Generation

response = Generation.call(
    model='qwen2.5-max',
    prompt='请将以下内容翻译为泰语：
"这款产品支持7天无理由退换货"',
    result_format='message'
)
print(response.output.text)

# 使用vLLM部署（推荐生产环境）
pip install vllm
python -m vllm.entrypoints.openai.api_server     --model Qwen/Qwen2.5-14B-Instruct     --host 0.0.0.0 --port 8000

# 使用Ollama
ollama run qwen2.5:14b

from peft import LoraConfig, get_peft_model
lora_config = LoraConfig(r=16, lora_alpha=32,
    target_modules=["q_proj", "v_proj"])
model = get_peft_model(base_model, lora_config)
trainer = Trainer(model=model, train_dataset=dataset)
trainer.train()

# Docker Compose部署Qwen服务
version: '3.8'
services:
  qwen-api:
    image: qwen-api:latest
    ports:
      - "8000:8000"
    environment:
      - MODEL_NAME=Qwen/Qwen2.5-72B-Instruct
      - CUDA_VISIBLE_DEVICES=0,1,2,3
    volumes:
      - ./models:/app/models
    deploy:
      resources:
        reservations:
          devices:
            - capabilities: [gpu]