centersl/MinerU
1
0
Fork 0
mirror of https://github.com/opendatalab/MinerU.git synced 2026-03-27 02:58:54 +07:00
Code Issues Packages Projects Releases Wiki Activity

feat: add Dockerfile for corex environment setup and update vllm server configurations

Browse Source
This commit is contained in:
myhloli
2026-01-29 22:05:03 +08:00
parent dc572f4c30
commit bdbee2b3ba
8 changed files with 198 additions and 27 deletions
Download Patch File Download Diff File Expand all files Collapse all files

27
docker/china/corex.Dockerfile Normal file
View File

@@ -0,0 +1,27 @@
# Base image containing the vLLM inference environment, requiring amd64(x86-64) CPU + iluvatar GPU.
FROM
# Install Noto fonts for Chinese characters
RUN apt-get update && \
apt-get install -y \
fonts-noto-core \
fonts-noto-cjk \
fontconfig && \
fc-cache -fv && \
apt-get clean && \
rm -rf /var/lib/apt/lists/*
# Install mineru latest
RUN python3 -m pip install -U pip -i https://mirrors.aliyun.com/pypi/simple && \
python3 -m pip install 'mineru[core]>=2.7.4' \
numpy==1.26.4 \
opencv-python==4.11.0.86 \
-i https://mirrors.aliyun.com/pypi/simple && \
python3 -m pip cache purge
# Download models and update the configuration file
RUN /bin/bash -c "mineru-models-download -s modelscope -m all"
# Set the entry point to activate the virtual environment and run the command line tool
ENTRYPOINT ["/bin/bash", "-c", "export MINERU_MODEL_SOURCE=local && exec \"$@\"", "--"]

2
docker/china/dcu.Dockerfile
View File

@@ -2,7 +2,7 @@
FROM harbor.sourcefind.cn:5443/dcu/admin/base/vllm:0.9.2-ubuntu22.04-dtk25.04.2-1226-das1.7-py3.10-20251226
# Install libgl for opencv support & Noto fonts for Chinese characters
# Install Noto fonts for Chinese characters
RUN apt-get update && \
apt-get install -y \
fonts-noto-core \

2
docker/china/gcu.Dockerfile
View File

@@ -2,7 +2,7 @@
FROM crpi-vofi3w62lkohhxsp.cn-shanghai.personal.cr.aliyuncs.com/opendatalab-mineru/gcu:docker_images_topsrider_i3x_3.6.20260106_vllm0.11_pytorch2.8.0
# Install libgl for opencv support & Noto fonts for Chinese characters
# Install Noto fonts for Chinese characters
RUN echo 'deb http://mirrors.aliyun.com/ubuntu/ noble main restricted universe multiverse\n\
deb http://mirrors.aliyun.com/ubuntu/ noble-updates main restricted universe multiverse\n\
deb http://mirrors.aliyun.com/ubuntu/ noble-backports main restricted universe multiverse\n\

120
docs/zh/usage/acceleration_cards/Iluvatar.md Normal file
View File

@@ -0,0 +1,120 @@
## 1. 测试平台
以下为本指南测试使用的平台信息，供参考：
```
os: Ubuntu 22.04.5 LTS
cpu: Intel x86-64
gcu: Iluvatar BI-V150
driver: 4.4.0
docker: 28.1.1
```
## 2. 环境准备
### 2.1 使用 Dockerfile 构建镜像
```bash
wget https://gcore.jsdelivr.net/gh/opendatalab/MinerU@master/docker/china/corex.Dockerfile
docker build --network=host -t mineru:corex-vllm-latest -f corex.Dockerfile .
```
## 3. 启动 Docker 容器
```bash
docker run --name mineru_docker \
-v /usr/src:/usr/src \
-v /lib/modules:/lib/modules \
-v /dev:/dev \
--privileged \
--cap-add=ALL \
--pid=host \
--group-add video \
--network=host \
--shm-size '400gb' \
--ulimit memlock=-1 \
--security-opt seccomp=unconfined \
--security-opt apparmor=unconfined \
-e VLLM_ENFORCE_CUDA_GRAPH=1 \
-e MINERU_MODEL_SOURCE=local \
-e MINERU_LMDEPLOY_DEVICE=corex \
-it mineru:corex-vllm-latest \
/bin/bash
```
执行该命令后，您将进入到Docker容器的交互式终端，您可以直接在容器内运行MinerU相关命令来使用MinerU的功能。
您也可以直接通过替换`/bin/bash`为服务启动命令来启动MinerU服务，详细说明请参考[通过命令启动服务](https://opendatalab.github.io/MinerU/zh/usage/quick_usage/#apiwebuihttp-clientserver)。
## 4. 注意事项
不同环境下，MinerU对Enflame加速卡的支持情况如下表所示：
<table border="1">
<thead>
<tr>
<th rowspan="2" colspan="2">使用场景</th>
<th colspan="2">容器环境</th>
</tr>
<tr>
<th>vllm</th>
</tr>
</thead>
<tbody>
<tr>
<td rowspan="3">命令行工具(mineru)</td>
<td>pipeline</td>
<td>🟢</td>
</tr>
<tr>
<td>&lt;vlm/hybrid&gt;-auto-engine</td>
<td>🟢</td>
</tr>
<tr>
<td>&lt;vlm/hybrid&gt;-http-client</td>
<td>🟢</td>
</tr>
<tr>
<td rowspan="3">fastapi服务(mineru-api)</td>
<td>pipeline</td>
<td>🟢</td>
</tr>
<tr>
<td>&lt;vlm/hybrid&gt;-auto-engine</td>
<td>🟢</td>
</tr>
<tr>
<td>&lt;vlm/hybrid&gt;-http-client</td>
<td>🟢</td>
</tr>
<tr>
<td rowspan="3">gradio界面(mineru-gradio)</td>
<td>pipeline</td>
<td>🟢</td>
</tr>
<tr>
<td>&lt;vlm/hybrid&gt;-auto-engine</td>
<td>🟢</td>
</tr>
<tr>
<td>&lt;vlm/hybrid&gt;-http-client</td>
<td>🟢</td>
</tr>
<tr>
<td colspan="2">openai-server服务（mineru-openai-server）</td>
<td>🟢</td>
</tr>
<tr>
<td colspan="2">数据并行 (--data-parallel-size)</td>
<td>🔴</td>
</tr>
</tbody>
</table>
注：
🟢: 支持，运行较稳定，精度与Nvidia GPU基本一致
🟡: 支持但较不稳定，在某些场景下可能出现异常，或精度存在一定差异
🔴: 不支持，无法运行，或精度存在较大差异
>[!TIP]
>GCU加速卡指定可用加速卡的方式与NVIDIA GPU类似，请参考[使用指定GPU设备](https://opendatalab.github.io/MinerU/zh/usage/advanced_cli_parameters/#cuda_visible_devices)章节说明,
>将环境变量`CUDA_VISIBLE_DEVICES`替换为`TOPS_VISIBLE_DEVICES`即可。

9
mineru/backend/pipeline/model_init.py
View File

@@ -297,7 +297,14 @@ def ocr_det_batch_setting(device):
# 检测torch的版本号
import torch
from packaging import version
if version.parse(torch.__version__) >= version.parse("2.8.0") or str(device).startswith('mps'):
device_type = os.getenv("MINERU_LMDEPLOY_DEVICE", "")
if (
version.parse(torch.__version__) >= version.parse("2.8.0")
or str(device).startswith('mps')
or device_type.lower() in ["corex"]
):
enable_ocr_det_batch = False
else:
enable_ocr_det_batch = True

7
mineru/backend/pipeline/pipeline_analyze.py
View File

@@ -193,7 +193,12 @@ def batch_image_analyze(
# 检测torch的版本号
import torch
from packaging import version
if version.parse(torch.__version__) >= version.parse("2.8.0") or str(device).startswith('mps'):
device_type = os.getenv("MINERU_LMDEPLOY_DEVICE", "")
if (
version.parse(torch.__version__) >= version.parse("2.8.0")
or str(device).startswith('mps')
or device_type.lower() in ["corex"]
):
enable_ocr_det_batch = False
else:
enable_ocr_det_batch = True

33
mineru/backend/vlm/vlm_analyze.py
View File

@@ -101,20 +101,27 @@ class ModelSingleton:
except ImportError:
raise ImportError("Please install vllm to use the vllm-engine backend.")
"""
# musa vllm v1 引擎特殊配置
device = get_device()
if device.startswith("musa"):
import torch
if torch.musa.is_available():
compilation_config = {
"cudagraph_capture_sizes": [1, 2, 3, 4, 5, 6, 7, 8, 10, 12, 14, 16, 18, 20, 24, 28, 30],
"simple_cuda_graph": True
}
block_size = 32
kwargs["compilation_config"] = compilation_config
kwargs["block_size"] = block_size
"""
# device = get_device()
# if device_type.startswith("musa"):
# import torch
# if torch.musa.is_available():
# compilation_config = {
# "cudagraph_capture_sizes": [1, 2, 3, 4, 5, 6, 7, 8, 10, 12, 14, 16, 18, 20, 24, 28, 30],
# "simple_cuda_graph": True
# }
# block_size = 32
# kwargs["compilation_config"] = compilation_config
# kwargs["block_size"] = block_size
# corex vllm v1 引擎特殊配置
device_type = os.getenv("MINERU_LMDEPLOY_DEVICE", "")
if device_type.lower() == "corex":
compilation_config = {
"cudagraph_mode": "FULL_DECODE_ONLY",
"level": 0
}
kwargs["compilation_config"] = compilation_config
if "compilation_config" in kwargs:
if isinstance(kwargs["compilation_config"], str):

25
mineru/model/vlm/vllm_server.py
View File

@@ -56,17 +56,22 @@ def main():
model_path = auto_download_and_get_model_root_path("/", "vlm")
if (not has_logits_processors_arg) and custom_logits_processors:
args.extend(["--logits-processors", "mineru_vl_utils:MinerULogitsProcessor"])
"""
# musa vllm v1 引擎特殊配置
device = get_device()
if device.startswith("musa"):
import torch
if torch.musa.is_available():
if not has_block_size_arg:
args.extend(["--block-size", "32"])
if not has_compilation_config:
args.extend(["--compilation-config", '{"cudagraph_capture_sizes": [1,2,3,4,5,6,7,8,10,12,14,16,18,20,24,28,30], "simple_cuda_graph": true}'])
"""
# device = get_device()
# if device.startswith("musa"):
# import torch
# if torch.musa.is_available():
# if not has_block_size_arg:
# args.extend(["--block-size", "32"])
# if not has_compilation_config:
# args.extend(["--compilation-config", '{"cudagraph_capture_sizes": [1,2,3,4,5,6,7,8,10,12,14,16,18,20,24,28,30], "simple_cuda_graph": true}'])
# corex vllm v1 引擎特殊配置
device_type = os.getenv("MINERU_LMDEPLOY_DEVICE", "")
if device_type.lower() == "corex":
if not has_compilation_config:
args.extend(["--compilation-config", '{"cudagraph_mode": "FULL_DECODE_ONLY", "level": 0}'])
# 重构参数，将模型路径作为位置参数
sys.argv = [sys.argv[0]] + ["serve", model_path] + args