vLLM 高性能推理系列——多模态模型部署

from openai import OpenAI
import base64

client = OpenAI(base_url="http://localhost:8000/v1", api_key="not-needed")

# 方式 1：使用图片 URL
response = client.chat.completions.create(
    model="Qwen/Qwen2.5-VL-7B-Instruct",
    messages=[
        {
            "role": "user",
            "content": [
                {"type": "text", "text": "请详细描述这张图片的内容"},
                {
                    "type": "image_url",
                    "image_url": {"url": "https://example.com/image.jpg"}
                }
            ]
        }
    ],
    max_tokens=512
)
print(response.choices[0].message.content)

# 方式 2：使用 Base64 编码（本地图片）
def encode_image(image_path: str) -> str:
    with open(image_path, "rb") as f:
        return base64.b64encode(f.read()).decode("utf-8")

image_data = encode_image("/path/to/local/image.jpg")

response = client.chat.completions.create(
    model="Qwen/Qwen2.5-VL-7B-Instruct",
    messages=[
        {
            "role": "user",
            "content": [
                {"type": "text", "text": "这是什么产品？请生成一段商品描述。"},
                {
                    "type": "image_url",
                    "image_url": {"url": f"data:image/jpeg;base64,{image_data}"}
                }
            ]
        }
    ],
    max_tokens=512
)

response = client.chat.completions.create(
    model="Qwen/Qwen2.5-VL-7B-Instruct",
    messages=[
        {
            "role": "user",
            "content": [
                {"type": "text", "text": "对比这两张图片，有什么不同？"},
                {
                    "type": "image_url",
                    "image_url": {"url": "https://example.com/before.jpg"}
                },
                {
                    "type": "image_url",
                    "image_url": {"url": "https://example.com/after.jpg"}
                }
            ]
        }
    ],
    max_tokens=512
)

from PIL import Image
import io
import base64

def preprocess_image(image_path: str, max_size: int = 1024) -> str:
    """压缩图像到合适的分辨率"""
    img = Image.open(image_path)
    
    # 保持宽高比缩放
    ratio = min(max_size / img.width, max_size / img.height)
    if ratio < 1:
        new_size = (int(img.width * ratio), int(img.height * ratio))
        img = img.resize(new_size, Image.LANCZOS)
    
    # 转换为 JPEG（压缩率更高）
    buffer = io.BytesIO()
    img.convert("RGB").save(buffer, format="JPEG", quality=85)
    
    return base64.b64encode(buffer.getvalue()).decode("utf-8")

from openai import OpenAI
from PIL import Image
import io
import base64

class ProductDescriptionGenerator:
    def __init__(self, base_url: str = "http://localhost:8000/v1"):
        self.client = OpenAI(base_url=base_url, api_key="not-needed")
        self.model = "Qwen/Qwen2.5-VL-7B-Instruct"
    
    def preprocess_image(self, image_path: str) -> str:
        """预处理图片"""
        img = Image.open(image_path)
        
        # 限制最大尺寸
        max_size = 800
        ratio = min(max_size / img.width, max_size / img.height)
        if ratio < 1:
            new_size = (int(img.width * ratio), int(img.height * ratio))
            img = img.resize(new_size, Image.LANCZOS)
        
        buffer = io.BytesIO()
        img.convert("RGB").save(buffer, format="JPEG", quality=85)
        return base64.b64encode(buffer.getvalue()).decode("utf-8")
    
    def generate_description(self, image_path: str, category: str = None) -> dict:
        """生成商品描述"""
        image_data = self.preprocess_image(image_path)
        
        prompt = """请分析这张商品图片，生成以下内容：

1. **商品名称**：简洁准确的产品名
2. **核心卖点**：3-5 个产品亮点，用于搜索标签
3. **商品描述**：100-150 字的营销文案，突出产品特色和使用场景
4. **目标人群**：适合的消费群体

请以 JSON 格式返回。"""

        if category:
            prompt += f"\n\n商品类目：{category}"
        
        response = self.client.chat.completions.create(
            model=self.model,
            messages=[
                {
                    "role": "user",
                    "content": [
                        {"type": "text", "text": prompt},
                        {
                            "type": "image_url",
                            "image_url": {"url": f"data:image/jpeg;base64,{image_data}"}
                        }
                    ]
                }
            ],
            max_tokens=800,
            temperature=0.7
        )
        
        return response.choices[0].message.content
    
    def batch_generate(self, image_paths: list, category: str = None) -> list:
        """批量生成（串行，避免显存溢出）"""
        results = []
        for path in image_paths:
            try:
                result = self.generate_description(path, category)
                results.append({"path": path, "description": result, "error": None})
            except Exception as e:
                results.append({"path": path, "description": None, "error": str(e)})
        return results

# 使用示例
generator = ProductDescriptionGenerator()
result = generator.generate_description("/path/to/product.jpg", category="女装/连衣裙")
print(result)