PaddleOCR-VL-For-Manga

# 推荐使用官方Docker镜像
docker pull quay.io/ascend/vllm-ascend:v0.17.0rc1

# 或使用国内镜像
docker pull m.daocloud.io/quay.io/ascend/vllm-ascend:v0.17.0rc1

export VLLM_USE_MODELSCOPE=true
export TASK_QUEUE_ENABLE=1
export CPU_AFFINITY_CONF=1
export PYTORCH_NPU_ALLOC_CONF="expandable_segments:True"

docker run --rm \
  --name paddleocr-vl-npu \
  --net=host \
  --shm-size=1g \
  --device /dev/davinci0 \
  --device /dev/davinci_manager \
  --device /dev/devmm_svm \
  --device /dev/hisi_hdc \
  -v /usr/local/dcmi:/usr/local/dcmi \
  -v /usr/local/Ascend/driver/tools/hccn_tool:/usr/local/Ascend/driver/tools/hccn_tool \
  -v /usr/local/bin/npu-smi:/usr/local/bin/npu-smi \
  -v /usr/local/Ascend/driver/lib64/:/usr/local/Ascend/driver/lib64/ \
  -v ~/.cache/modelscope:/root/.cache/modelscope \
  -it quay.io/ascend/vllm-ascend:v0.17.0rc1 bash

# 安装vLLM-Ascend
pip install vllm-ascend

# 验证安装
python -c "import vllm; print(vllm.__version__)"

#!/bin/bash
export VLLM_USE_MODELSCOPE=true
export MODEL_PATH="PaddlePaddle/PaddleOCR-VL"
export TASK_QUEUE_ENABLE=1
export CPU_AFFINITY_CONF=1
export PYTORCH_NPU_ALLOC_CONF="expandable_segments:True"

vllm serve ${MODEL_PATH} \
    --max-num-batched-tokens 16384 \
    --served-model-name PaddleOCR-VL-0.9B \
    --trust-remote-code \
    --no-enable-prefix-caching \
    --mm-processor-cache-gb 0 \
    --compilation-config '{"cudagraph_mode":"FULL_DECODE_ONLY"}' \
    --additional_config '{"enable_cpu_binding":true}' \
    --port 8000

# 检查服务状态
curl -sf http://127.0.0.1:8000/v1/models

# 文本推理测试
curl -s http://127.0.0.1:8000/v1/chat/completions \
    -H 'Content-Type: application/json' \
    -d '{
        "model": "PaddleOCR-VL-0.9B",
        "messages": [{"role": "user", "content": "say hi"}],
        "temperature": 0,
        "max_tokens": 16
    }'

from openai import OpenAI

client = OpenAI(
    api_key="EMPTY",
    base_url="http://localhost:8000/v1",
    timeout=3600
)

# OCR任务
messages = [{
    "role": "user",
    "content": [
        {
            "type": "image_url",
            "image_url": {"url": "https://example.com/image.png"}
        },
        {
            "type": "text",
            "text": "OCR:"
        }
    ]
}]

response = client.chat.completions.create(
    model="PaddleOCR-VL-0.9B",
    messages=messages,
    temperature=0.0,
)

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

# 安装依赖
pip install requests pillow

# 执行推理
python inference.py \
    --image path/to/image.png \
    --task ocr \
    --backend npu \
    --output result.json

# 运行精度对比脚本
python accuracy_eval.py \
    --backend npu \
    --test-data ./test_images/ \
    --output accuracy_report.json

# 运行性能对比脚本
python performance_eval.py \
    --backends cpu,npu \
    --input-len 512 \
    --output-len 128 \
    --num-iterations 100 \
    --output performance_report.json

--max-model-len 8192
--max-num-seqs 4

vllm serve ${MODEL_PATH} \
    --enforce-eager \
    --dtype float16 \
    --port 8000