Unlimited-OCR 百度开源长文档 OCR 深度实战：vLLM 加速 + Docker 部署 + 企业级应用

# R-SWA 核心逻辑伪代码
class ReferentialSlidingWindowAttention(nn.Module):
    def __init__(self, window_size=2048, ref_size=512):
        self.window_size = window_size  # 当前窗口（局部细节）
        self.ref_size = ref_size        # 参考窗口（全局上下文）

    def forward(self, hidden_states, past_kv):
        # 1. 当前窗口注意力（局部细节）
        local_attn = self.local_attention(hidden_states[-self.window_size:])

        # 2. 参考窗口注意力（全局上下文）
        ref_attn = self.ref_attention(past_kv[-self.ref_size:])

        # 3. 门控融合：局部 + 全局
        gate = self.sigmoid(self.gate_proj(hidden_states))
        fused = gate * local_attn + (1 - gate) * ref_attn
        return fused

# 安装 vLLM
pip install vllm

# 启动 vLLM 服务
python -m vllm.entrypoints.openai.api_server \
  --model baidu/Unlimited-OCR \
  --tensor-parallel-size 1 \
  --max-model-len 32768 \
  --port 8000

# 调用 API（兼容 OpenAI 格式）
curl http://localhost:8000/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "baidu/Unlimited-OCR",
    "messages": [
      {
        "role": "user",
        "content": [
          {"type": "image_url", "image_url": {"url": "file:///path/to/document.pdf"}},
          {"type": "text", "text": "请解析这份文档"}
        ]
      }
    ]
  }'

# 安装依赖
pip install transformers torch pillow

# Python 代码
from transformers import AutoModelForVision2Seq, AutoProcessor
from PIL import Image
import torch

# 加载模型
model = AutoModelForVision2Seq.from_pretrained(
    "baidu/Unlimited-OCR",
    torch_dtype=torch.float16,
    device_map="auto"
)
processor = AutoProcessor.from_pretrained("baidu/Unlimited-OCR")

# 处理文档
image = Image.open("document.pdf")  # 支持 PDF、图片
inputs = processor(images=image, text="解析这份文档", return_tensors="pt")
inputs = {k: v.to(model.device) for k, v in inputs.items()}

# 推理
with torch.no_grad():
    outputs = model.generate(**inputs, max_new_tokens=32768)

result = processor.decode(outputs[0], skip_special_tokens=True)
print(result)

# 安装 vLLM
pip install vllm

# 启动服务
python -m vllm.entrypoints.openai.api_server \
  --model baidu/Unlimited-OCR \
  --tensor-parallel-size 1 \
  --max-model-len 32768 \
  --port 8000

# 批量处理脚本
import requests
import base64
import os

def ocr_document(pdf_path):
    with open(pdf_path, "rb") as f:
        pdf_base64 = base64.b64encode(f.read()).decode()

    response = requests.post(
        "http://localhost:8000/v1/chat/completions",
        json={
            "model": "baidu/Unlimited-OCR",
            "messages": [{
                "role": "user",
                "content": [
                    {"type": "image_url", "image_url": {"url": f"data:application/pdf;base64,{pdf_base64}"}},
                    {"type": "text", "text": "解析这份文档为 Markdown 格式"}
                ]
            }]
        }
    )
    return response.json()["choices"][0]["message"]["content"]

# 批量处理
for pdf_file in os.listdir("./documents"):
    if pdf_file.endswith(".pdf"):
        result = ocr_document(f"./documents/{pdf_file}")
        with open(f"./results/{pdf_file}.md", "w") as f:
            f.write(result)
        print(f"Processed: {pdf_file}")

# 拉取官方镜像
docker pull baidu/unlimited-ocr:latest

# 启动服务
docker run -d \
  --name unlimited-ocr \
  --gpus all \
  -p 8000:8000 \
  -v $(pwd)/models:/models \
  -v $(pwd)/results:/results \
  baidu/unlimited-ocr:latest

# 处理文档
docker exec unlimited-ocr python infer.py \
  --model /models/Unlimited-OCR \
  --input /results/document.pdf \
  --output /results/output.md \
  --mode multi

# Docker Compose 配置
cat > docker-compose.yml << 'EOF'
version: '3.8'
services:
  unlimited-ocr:
    image: baidu/unlimited-ocr:latest
    ports:
      - "8000:8000"
    volumes:
      - ./models:/models
      - ./results:/results
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: 1
              capabilities: [gpu]
    environment:
      - MODEL_PATH=/models/Unlimited-OCR
      - MAX_MODEL_LEN=32768
EOF

docker-compose up -d

# 输入：scanned_manual.pdf (100 页)
# 输出：editable_manual.md

python infer.py \
  --model baidu/Unlimited-OCR \
  --input scanned_manual.pdf \
  --output editable_manual.md \
  --mode multi \
  --format markdown \
  --context-length 32768

# 耗时：约 90 秒（vLLM 加速）
# 输出质量：标题层级、代码块、表格结构均正确识别

from unlimited_ocr import UnlimitedOCR

ocr = UnlimitedOCR(model_path="baidu/Unlimited-OCR")

# R-SWA 的连续上下文确保跨页表格完整提取
result = ocr.extract_tables(
    input_path="financial_report.pdf",
    output_format="csv",
    context_length=32768
)

# 输出：完整的跨页表格，无截断
result.to_csv("extracted_tables.csv")

#!/bin/bash
# batch_parse_papers.sh

INPUT_DIR="./papers"
OUTPUT_DIR="./parsed"
MODEL="baidu/Unlimited-OCR"

mkdir -p "$OUTPUT_DIR"

for pdf in "$INPUT_DIR"/*.pdf; do
    filename=$(basename "$pdf" .pdf)
    echo "Processing: $filename"

    python infer.py \
      --model "$MODEL" \
      --input "$pdf" \
      --output "$OUTPUT_DIR/${filename}.md" \
      --mode multi \
      --extract-structure  # 提取标题、摘要、章节

    echo "Done: $filename"
done

echo "All papers processed."