commit 553ec651bfbd5763454e82cec7228927b2e1995d Author: myhloli Date: Thu Mar 26 18:02:27 2026 +0000 Deployed bd46b08 with MkDocs version: 1.5.3 diff --git a/.nojekyll b/.nojekyll new file mode 100644 index 00000000..e69de29b diff --git a/404.html b/404.html new file mode 100644 index 00000000..78869dae --- /dev/null +++ b/404.html @@ -0,0 +1,1605 @@ + + + + + + + + + + + + + + + + + + + MinerU + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

+ +

+ + + + +

+ + +

+ + + + + + +

+ + MinerU + +

+ + + + + +

+ + + + + + + + + +

+ + +

+ + English + +
+ + 中文 + +

+ + + + + + + + + +

+ +

+ + +

+ opendatalab/MinerU +

+ +

+ + +

+ + + + + + + + + + \ No newline at end of file diff --git a/faq/index.html b/faq/index.html new file mode 100644 index 00000000..28f0b3d8 --- /dev/null +++ b/faq/index.html @@ -0,0 +1,1700 @@ + + + + + + + + + + + + + + + + + + + + + + + + + FAQ - MinerU + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

+ + + + Skip to content + + +

+ +

+ + + + +

+ + +

+ + + + + + +

+ + MinerU + +

+ + + FAQ + + +

+ + + + + + + + + +

+ + +

+ + English + +
+ + 中文 + +

+ + + + + + + + + +

+ +

+ + +

+ opendatalab/MinerU +

+ +

+ + + + + +

+ + + + + + Home + + +
+ + + + + + Quick Start + + +
+ + + + + + Usage + + +
+ + + + + + Reference + + +
+ + + + + + FAQ + + +
+ + + + + + Demo + + +

+ + + +

+ +

+ + + + +

+ + + + + + + +

Frequently Asked Questions

If your question is not listed, try using DeepWiki's AI assistant for common issues.

For unresolved problems, join our Discord or WeChat community for support.

Encountered the error ImportError: libGL.so.1: cannot open shared object file: No such file or directory in Ubuntu 22.04 on WSL2

The libgl library is missing in Ubuntu 22.04 on WSL2. You can install the libgl library with the following command to resolve the issue:

sudo apt-get install libgl1-mesa-glx
+

Reference: #388

Missing text information in parsing results when installing and using on Linux systems.

MinerU uses pypdfium2 instead of pymupdf as the PDF page rendering engine in versions >=2.0 to resolve AGPLv3 license issues. On some Linux distributions, due to missing CJK fonts, some text may be lost during the process of rendering PDFs to images. +To solve this problem, you can install the noto font package with the following commands, which are effective on Ubuntu/Debian systems: +

sudo apt update
+sudo apt install fonts-noto-core
+sudo apt install fonts-noto-cjk
+fc-cache -fv
+

+You can also directly use our Docker deployment method to build the image, which includes the above font packages by default. +

Reference: #2915

+ + + + + + + + + + + + + + + +

+ + + + + +

+ + + +

+ + + + + + + + + + \ No newline at end of file diff --git a/images/MinerU-logo.png b/images/MinerU-logo.png new file mode 100644 index 00000000..09ab46b2 Binary files /dev/null and b/images/MinerU-logo.png differ diff --git a/images/datalab_logo.png b/images/datalab_logo.png new file mode 100644 index 00000000..5019ae7c Binary files /dev/null and b/images/datalab_logo.png differ diff --git a/images/flowchart_en.png b/images/flowchart_en.png new file mode 100644 index 00000000..b490011e Binary files /dev/null and b/images/flowchart_en.png differ diff --git a/images/flowchart_zh_cn.png b/images/flowchart_zh_cn.png new file mode 100644 index 00000000..32e0a142 Binary files /dev/null and b/images/flowchart_zh_cn.png differ diff --git a/images/layout_example.png b/images/layout_example.png new file mode 100644 index 00000000..4a57dffe Binary files /dev/null and b/images/layout_example.png differ diff --git a/images/logo.png b/images/logo.png new file mode 100644 index 00000000..01818084 Binary files /dev/null and b/images/logo.png differ diff --git a/images/logo.svg b/images/logo.svg new file mode 100644 index 00000000..65539780 --- /dev/null +++ b/images/logo.svg @@ -0,0 +1,22 @@ + diff --git a/images/poly.png b/images/poly.png new file mode 100644 index 00000000..14af7726 Binary files /dev/null and b/images/poly.png differ diff --git a/images/project_panorama_en.png b/images/project_panorama_en.png new file mode 100644 index 00000000..19616da6 Binary files /dev/null and b/images/project_panorama_en.png differ diff --git a/images/project_panorama_zh_cn.png b/images/project_panorama_zh_cn.png new file mode 100644 index 00000000..3cd6843e Binary files /dev/null and b/images/project_panorama_zh_cn.png differ diff --git a/images/spans_example.png b/images/spans_example.png new file mode 100644 index 00000000..14de87ed Binary files /dev/null and b/images/spans_example.png differ diff --git a/index.html b/index.html new file mode 100644 index 00000000..8b648f7d --- /dev/null +++ b/index.html @@ -0,0 +1,1742 @@ + + + + + + + + + + + + + + + + + + + + + + + MinerU - MinerU + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

+ + + + Skip to content + + +

+ +

+ + + + +

+ + +

+ + + + + + +

+ + MinerU + +

+ + + MinerU + + +

+ + + + + + + + + +

+ + +

+ + English + +
+ + 中文 + +

+ + + + + + + + + +

+ +

+ + +

+ opendatalab/MinerU +

+ +

+ + +

+ + + + + + + + + + \ No newline at end of file diff --git a/quick_start/docker_deployment/index.html b/quick_start/docker_deployment/index.html new file mode 100644 index 00000000..7e4a98f3 --- /dev/null +++ b/quick_start/docker_deployment/index.html @@ -0,0 +1,1963 @@ + + + + + + + + + + + + + + + + + + + + + + + + + Docker Deployment - MinerU + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

+ + + + Skip to content + + +

+ +

+ + + + +

+ + +

+ + + + + + +

+ + MinerU + +

+ + + Docker Deployment + + +

+ + + + + + + + + +

+ + +

+ + English + +
+ + 中文 + +

+ + + + + + + + + +

+ +

+ + +

+ opendatalab/MinerU +

+ +

+ + + + + +

+ + + + + + Home + + +
+ + + + + + Quick Start + + +
+ + + + + + Usage + + +
+ + + + + + Reference + + +
+ + + + + + FAQ + + +
+ + + + + + Demo + + +

+ + + +

+ +

+ + + + +

+ + + + + + + +

Deploying MinerU with Docker

MinerU provides a convenient Docker deployment method, which helps quickly set up the environment and solve some tricky environment compatibility issues.

Build Docker Image using Dockerfile

wget https://gcore.jsdelivr.net/gh/opendatalab/MinerU@master/docker/global/Dockerfile
+docker build -t mineru:latest -f Dockerfile .
+

Docker Description

MinerU's Docker uses vllm/vllm-openai as the base image, so it includes the vllm inference acceleration framework and necessary dependencies by default. Therefore, on compatible devices, you can directly use vllm to accelerate VLM model inference.

Note

Requirements for using vllm to accelerate VLM model inference:

Device must have Volta architecture or later graphics cards with 8GB+ available VRAM.
The host machine's graphics driver should support CUDA 12.9.1 or higher; You can check the driver version using the nvidia-smi command.
Docker container must have access to the host machine's graphics devices.

Start Docker Container

docker run --gpus all \
+  --shm-size 32g \
+  -p 30000:30000 -p 7860:7860 -p 8000:8000 \
+  --ipc=host \
+  -it mineru:latest \
+  /bin/bash
+

After executing this command, you will enter the Docker container's interactive terminal with some ports mapped for potential services. You can directly run MinerU-related commands within the container to use MinerU's features. +You can also directly start MinerU services by replacing /bin/bash with service startup commands. For detailed instructions, please refer to the Start the service via command.

Start Services Directly with Docker Compose

We provide a compose.yaml file that you can use to quickly start MinerU services.

# Download compose.yaml file
+wget https://gcore.jsdelivr.net/gh/opendatalab/MinerU@master/docker/compose.yaml
+

Note

The compose.yaml file contains configurations for multiple services of MinerU, you can choose to start specific services as needed.
Different services might have additional parameter configurations, which you can view and edit in the compose.yaml file.
Due to the pre-allocation of GPU memory by the vllm inference acceleration framework, you may not be able to run multiple vllm services simultaneously on the same machine. Therefore, ensure that other services that might use GPU memory have been stopped before starting the vlm-openai-server service or using the vlm-vllm-engine backend.

Start OpenAI-compatible server service

connect to openai-server via vlm-http-client backend +

docker compose -f compose.yaml --profile openai-server up -d
+

Tip

In another terminal, connect to openai server via http client (only requires CPU and network, no vllm environment needed) +

mineru -p <input_path> -o <output_path> -b vlm-http-client -u http://<server_ip>:30000
+

Start Web API service

docker compose -f compose.yaml --profile api up -d
+

Tip

Access http://<server_ip>:8000/docs in your browser to view the API documentation.

Start Gradio WebUI service

docker compose -f compose.yaml --profile gradio up -d
+

Tip

Access http://<server_ip>:7860 in your browser to use the Gradio WebUI.

+ + + + + + + + + + + + + + + +

+ + + + + +

+ + + +

+ + + + + + + + + + \ No newline at end of file diff --git a/quick_start/extension_modules/index.html b/quick_start/extension_modules/index.html new file mode 100644 index 00000000..8b948c35 --- /dev/null +++ b/quick_start/extension_modules/index.html @@ -0,0 +1,1921 @@ + + + + + + + + + + + + + + + + + + + + + + + + + Extension Modules - MinerU + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

+ + + + Skip to content + + +

+ +

+ + + + +

+ + +

+ + + + + + +

+ + MinerU + +

+ + + Extension Modules + + +

+ + + + + + + + + +

+ + +

+ + English + +
+ + 中文 + +

+ + + + + + + + + +

+ +

+ + +

+ opendatalab/MinerU +

+ +

+ + + + + +

+ + + + + + Home + + +
+ + + + + + Quick Start + + +
+ + + + + + Usage + + +
+ + + + + + Reference + + +
+ + + + + + FAQ + + +
+ + + + + + Demo + + +

+ + + +

+ +

+ + + + +

+ + + + + + + +

MinerU Extension Modules Installation Guide

MinerU supports installing extension modules on demand based on different needs to enhance functionality or support specific model backends.

Common Scenarios

Core Functionality Installation

The core module is the core dependency of MinerU, containing all functional modules except vllm/lmdeploy. Installing this module ensures the basic functionality of MinerU works properly. +

uv pip install "mineru[core]"
+

Using `vllm` to Accelerate VLM Model Inference

Note

vllm and lmdeploy have nearly identical VLM inference acceleration effects and usage methods. You can choose one of them to install and use based on your actual needs, but it is not recommended to install both modules simultaneously to avoid potential dependency conflicts.

The vllm module provides acceleration support for VLM model inference, suitable for graphics cards with Volta architecture and later (8GB+ VRAM). Installing this module can significantly improve model inference speed.

uv pip install "mineru[core,vllm]"
+

Tip

If exceptions occur during installation of the extra package including vllm, please refer to the vllm official documentation to try to resolve the issue, or directly use the Docker deployment method.

Using `lmdeploy` to Accelerate VLM Model Inference

Note

The lmdeploy module provides acceleration support for VLM model inference, suitable for graphics cards with Volta architecture and later (8GB+ VRAM). Installing this module can significantly improve model inference speed.

uv pip install "mineru[core,lmdeploy]"
+

Tip

If exceptions occur during installation of the extra package including lmdeploy, please refer to the lmdeploy official documentation to try to resolve the issue.

Installing Lightweight Client to Connect to OpenAI-compatible servers (for vlm-http-client mode)

If you need to install a lightweight client on edge devices to connect to an OpenAI-compatible server for using VLM mode, you can install the basic mineru package, which is very lightweight and suitable for devices with only CPU and network connectivity. +

uv pip install mineru
+mineru -p <input_path> -o <output_path> -b vlm-http-client -u http://127.0.0.1:30000
+

Installing Lightweight Client to Connect to OpenAI-compatible servers (for hybrid-http-client mode)

If you need to install a lightweight client on edge devices to connect to an OpenAI-compatible server for using hybrid mode, you can install the mineru pipeline extension package, which is relatively lightweight and can be used on devices with only CPU and network connectivity, while running faster on devices that support GPU acceleration. +

uv pip install "mineru[pipeline]"
+mineru -p <input_path> -o <output_path> -b hybrid-http-client -u http://127.0.0.1:30000
+

+ + + + + + + + + + + + + + + +

+ + + + + +

+ + + +

+ + + + + + + + + + \ No newline at end of file diff --git a/quick_start/index.html b/quick_start/index.html new file mode 100644 index 00000000..97adaec1 --- /dev/null +++ b/quick_start/index.html @@ -0,0 +1,1800 @@ + + + + + + + + + + + + + + + + + + + + + + + + + Quick Start - MinerU + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

+ + + + Skip to content + + +

+ +

+ + + + +

+ + +

+ + + + + + +

+ + MinerU + +

+ + + Quick Start + + +

+ + + + + + + + + +

+ + +

+ + English + +
+ + 中文 + +

+ + + + + + + + + +

+ +

+ + +

+ opendatalab/MinerU +

+ +

+ + + + + +

+ + + + + + Home + + +
+ + + + + + Quick Start + + +
+ + + + + + Usage + + +
+ + + + + + Reference + + +
+ + + + + + FAQ + + +
+ + + + + + Demo + + +

+ + + +

+ +

+ + + + +

+ + + + + + + +

Quick Start

If you encounter any installation issues, please check the FAQ first.

Online Experience

Official online web application

The official online version has the same functionality as the client, with a beautiful interface and rich features, requires login to use

Gradio-based online demo

A WebUI developed based on Gradio, with a simple interface and only core parsing functionality, no login required

Local Deployment

Warning

Prerequisites - Hardware and Software Environment Support

To ensure the stability and reliability of the project, we have optimized and tested only specific hardware and software environments during development. This ensures that users can achieve optimal performance and encounter the fewest compatibility issues when deploying and running the project on recommended system configurations.

By concentrating our resources and efforts on mainstream environments, our team can more efficiently resolve potential bugs and timely develop new features.

In non-mainstream environments, due to the diversity of hardware and software configurations, as well as compatibility issues with third-party dependencies, we cannot guarantee 100% usability of the project. Therefore, for users who wish to use this project in non-recommended environments, we suggest carefully reading the documentation and FAQ first, as most issues have corresponding solutions in the FAQ. Additionally, we encourage community feedback on issues so that we can gradually expand our support range.

+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

Parsing Backend	pipeline	*-auto-engine		*-http-client
Parsing Backend	pipeline	hybrid	vlm	hybrid	vlm
Backend Features	Good Compatibility	High Hardware Requirements		For OpenAI Compatible Servers²
Accuracy¹	82+	90+
Operating System	Linux³ / Windows⁴ / macOS⁵
Pure CPU Support	✅	❌		✅
GPU Acceleration	Volta and later architecture GPUs or Apple Silicon				Not Required
Min VRAM	6GB	10GB	8GB	3GB	Not Required
RAM	Min 16GB+, Recommended 32GB+			8GB
Disk Space	20GB+, SSD Recommended			2GB
Python Version	3.10-3.13

+ +

¹ Accuracy metrics are the End-to-End Evaluation Overall scores from OmniDocBench (v1.5), based on the latest version of MinerU.
+² Servers compatible with OpenAI API, such as local model servers or remote model services deployed via inference frameworks like vLLM/SGLang/LMDeploy.
+³ Linux only supports distributions from 2019 and later.
+⁴ Since the key dependency ray does not support Python 3.13 on Windows, only versions 3.10~3.12 are supported.
+⁵ macOS requires version 14.0 or later.

Install MinerU

Install MinerU using pip or uv

pip install --upgrade pip
+pip install uv
+uv pip install -U "mineru[all]"
+

Install MinerU from source code

git clone https://github.com/opendatalab/MinerU.git
+cd MinerU
+uv pip install -e .[all]
+

Tip

mineru[all] includes all core features, compatible with Windows / Linux / macOS systems, suitable for most users. +If you need to specify the inference framework for the VLM model, or only intend to install a lightweight client on an edge device, please refer to the documentation Extension Modules Installation Guide.

Deploy MinerU using Docker

MinerU provides a convenient Docker deployment method, which helps quickly set up the environment and solve some tricky environment compatibility issues. +You can get the Docker Deployment Instructions in the documentation.

Using MinerU

If your device meets the GPU acceleration requirements in the table above, you can use a simple command line for document parsing: +

mineru -p <input_path> -o <output_path>
+

+If your device does not meet the GPU acceleration requirements, you can specify the backend as pipeline to run in a pure CPU environment: +

mineru -p <input_path> -o <output_path> -b pipeline
+

You can use MinerU for PDF parsing through various methods such as command line, API, and WebUI. For detailed instructions, please refer to the Usage Guide.

+ + + + + + + + + + + + + + + +

+ + + + + +

+ + + +

+ + + + + + + + + + \ No newline at end of file diff --git a/reference/changelog/index.html b/reference/changelog/index.html new file mode 100644 index 00000000..4441c2e1 --- /dev/null +++ b/reference/changelog/index.html @@ -0,0 +1,3229 @@ + + + + + + + + + + + + + + + + + + + + + + + + + Changelog - MinerU + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

+ + + + Skip to content + + +

+ +

+ + + + +

+ + +

+ + + + + + +

+ + MinerU + +

+ + + Changelog + + +

+ + + + + + + + + +

+ + +

+ + English + +
+ + 中文 + +

+ + + + + + + + + +

+ +

+ + +

+ opendatalab/MinerU +

+ +

+ + +

+ + + + + + + + + + \ No newline at end of file diff --git a/reference/index.html b/reference/index.html new file mode 100644 index 00000000..691ff0d4 --- /dev/null +++ b/reference/index.html @@ -0,0 +1,1701 @@ + + + + + + + + + + + + + + + + + + + + + + + + + Reference - MinerU + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

+ + + + Skip to content + + +

+ +

+ + + + +

+ + +

+ + + + + + +

+ + MinerU + +

+ + + Reference + + +

+ + + + + + + + + +

+ + +

+ + English + +
+ + 中文 + +

+ + + + + + + + + +

+ +

+ + +

+ opendatalab/MinerU +

+ +

+ + +

+ + + + + + + + + + \ No newline at end of file diff --git a/reference/output_files/index.html b/reference/output_files/index.html new file mode 100644 index 00000000..44af4e96 --- /dev/null +++ b/reference/output_files/index.html @@ -0,0 +1,3342 @@ + + + + + + + + + + + + + + + + + + + + + + + + + Output File Format - MinerU + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

+ + + + Skip to content + + +

+ +

+ + + + +

+ + +

+ + + + + + +

+ + MinerU + +

+ + + Output File Format + + +

+ + + + + + + + + +

+ + +

+ + English + +
+ + 中文 + +

+ + + + + + + + + +

+ +

+ + +

+ opendatalab/MinerU +

+ +

+ + + + + +

+ + + + + + Home + + +
+ + + + + + Quick Start + + +
+ + + + + + Usage + + +
+ + + + + + Reference + + +
+ + + + + + FAQ + + +
+ + + + + + Demo + + +

+ + + +

+ +

+ + + + +

+ + + + + + + +

MinerU Output Files Documentation

Overview

After executing the mineru command, in addition to the main markdown file output, multiple auxiliary files are generated for debugging, quality inspection, and further processing. These files include:

Visual debugging files: Help users intuitively understand the document parsing process and results
Structured data files: Contain detailed parsing data for secondary development

The following sections provide detailed descriptions of each file's purpose and format.

Visual Debugging Files

Layout Analysis File (layout.pdf)

File naming format: {original_filename}_layout.pdf

Functionality:

Visualizes layout analysis results for each page
Numbers in the top-right corner of each detection box indicate reading order
Different background colors distinguish different types of content blocks

Use cases:

Check if layout analysis is correct
Verify if reading order is reasonable
Debug layout-related issues

layout page example

Text Spans File (span.pdf)

Note

Only applicable to pipeline backend

File naming format: {original_filename}_span.pdf

Functionality:

Uses different colored line boxes to annotate page content based on span type
Used for quality inspection and issue troubleshooting

Use cases:

Quickly troubleshoot text loss issues
Check inline formula recognition
Verify text segmentation accuracy

span page example

Structured Data Files

Important

The VLM backend output has significant changes in version 2.5 and is not backward-compatible with the pipeline backend. If you plan to build secondary development on structured outputs, please read this document carefully.

Pipeline Backend Output Results

Model Inference Results (model.json)

File naming format: {original_filename}_model.json

Data Structure Definition

from pydantic import BaseModel, Field
+from enum import IntEnum
+
+class CategoryType(IntEnum):
+    """Content category enumeration"""
+    title = 0               # Title
+    plain_text = 1          # Text
+    abandon = 2             # Including headers, footers, page numbers, and page annotations
+    figure = 3              # Image
+    figure_caption = 4      # Image caption
+    table = 5               # Table
+    table_caption = 6       # Table caption
+    table_footnote = 7      # Table footnote
+    isolate_formula = 8     # Interline formula
+    formula_caption = 9     # Interline formula number
+    embedding = 13          # Inline formula
+    isolated = 14           # Interline formula
+    text = 15               # OCR recognition result
+
+class PageInfo(BaseModel):
+    """Page information"""
+    page_no: int = Field(description="Page number, first page is 0", ge=0)
+    height: int = Field(description="Page height", gt=0)
+    width: int = Field(description="Page width", ge=0)
+
+class ObjectInferenceResult(BaseModel):
+    """Object recognition result"""
+    category_id: CategoryType = Field(description="Category", ge=0)
+    poly: list[float] = Field(description="Quadrilateral coordinates, format: [x0,y0,x1,y1,x2,y2,x3,y3]")
+    score: float = Field(description="Confidence score of inference result")
+    latex: str | None = Field(description="LaTeX parsing result", default=None)
+    html: str | None = Field(description="HTML parsing result", default=None)
+
+class PageInferenceResults(BaseModel):
+    """Page inference results"""
+    layout_dets: list[ObjectInferenceResult] = Field(description="Page recognition results")
+    page_info: PageInfo = Field(description="Page metadata")
+
+# Complete inference results
+inference_result: list[PageInferenceResults] = []
+

Coordinate System Description

poly coordinate format: [x0, y0, x1, y1, x2, y2, x3, y3]

Represents coordinates of top-left, top-right, bottom-right, bottom-left points respectively
Coordinate origin is at the top-left corner of the page

poly coordinate diagram

Sample Data

[
+    {
+        "layout_dets": [
+            {
+                "category_id": 2,
+                "poly": [
+                    99.1906967163086,
+                    100.3119125366211,
+                    730.3707885742188,
+                    100.3119125366211,
+                    730.3707885742188,
+                    245.81326293945312,
+                    99.1906967163086,
+                    245.81326293945312
+                ],
+                "score": 0.9999997615814209
+            }
+        ],
+        "page_info": {
+            "page_no": 0,
+            "height": 2339,
+            "width": 1654
+        }
+    },
+    {
+        "layout_dets": [
+            {
+                "category_id": 5,
+                "poly": [
+                    99.13092803955078,
+                    2210.680419921875,
+                    497.3183898925781,
+                    2210.680419921875,
+                    497.3183898925781,
+                    2264.78076171875,
+                    99.13092803955078,
+                    2264.78076171875
+                ],
+                "score": 0.9999997019767761
+            }
+        ],
+        "page_info": {
+            "page_no": 1,
+            "height": 2339,
+            "width": 1654
+        }
+    }
+]
+

Intermediate Processing Results (middle.json)

File naming format: {original_filename}_middle.json

Top-level Structure

+ + + + + + + + + + + + + + + + + + + + + + + + + +

Field Name	Type	Description
`pdf_info`	`list[dict]`	Array of parsing results for each page
`_backend`	`string`	Parsing mode: `pipeline` or `vlm`
`_version_name`	`string`	MinerU version number

Page Information Structure (pdf_info)

+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

Field Name	Description
`preproc_blocks`	Unsegmented intermediate results after PDF preprocessing
`page_idx`	Page number, starting from 0
`page_size`	Page width and height `[width, height]`
`images`	Image block information list
`tables`	Table block information list
`interline_equations`	Interline formula block information list
`discarded_blocks`	Block information to be discarded
`para_blocks`	Content block results after segmentation

Block Structure Hierarchy

Level 1 blocks (table | image)
+└── Level 2 blocks
+    └── Lines
+        └── Spans
+

Level 1 Block Fields

+ + + + + + + + + + + + + + + + + + + + + +

Field Name	Description
`type`	Block type: `table` or `image`
`bbox`	Rectangular box coordinates of the block `[x0, y0, x1, y1]`
`blocks`	List of contained level 2 blocks

Level 2 Block Fields

+ + + + + + + + + + + + + + + + + + + + + +

Field Name	Description
`type`	Block type (see table below)
`bbox`	Rectangular box coordinates of the block
`lines`	List of contained line information

Level 2 Block Types

+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

Type	Description
`image_body`	Image body
`image_caption`	Image caption text
`image_footnote`	Image footnote
`table_body`	Table body
`table_caption`	Table caption text
`table_footnote`	Table footnote
`text`	Text block
`title`	Title block
`index`	Index block
`list`	List block
`interline_equation`	Interline formula block

Line and Span Structure

Line fields: +- bbox: Rectangular box coordinates of the line +- spans: List of contained spans

Span fields: +- bbox: Rectangular box coordinates of the span +- type: Span type (image, table, text, inline_equation, interline_equation) +- content | img_path: Text content or image path

Sample Data

{
+    "pdf_info": [
+        {
+            "preproc_blocks": [
+                {
+                    "type": "text",
+                    "bbox": [
+                        52,
+                        61.956024169921875,
+                        294,
+                        82.99800872802734
+                    ],
+                    "lines": [
+                        {
+                            "bbox": [
+                                52,
+                                61.956024169921875,
+                                294,
+                                72.0000228881836
+                            ],
+                            "spans": [
+                                {
+                                    "bbox": [
+                                        54.0,
+                                        61.956024169921875,
+                                        296.2261657714844,
+                                        72.0000228881836
+                                    ],
+                                    "content": "dependent on the service headway and the reliability of the departure ",
+                                    "type": "text",
+                                    "score": 1.0
+                                }
+                            ]
+                        }
+                    ]
+                }
+            ],
+            "layout_bboxes": [
+                {
+                    "layout_bbox": [
+                        52,
+                        61,
+                        294,
+                        731
+                    ],
+                    "layout_label": "V",
+                    "sub_layout": []
+                }
+            ],
+            "page_idx": 0,
+            "page_size": [
+                612.0,
+                792.0
+            ],
+            "_layout_tree": [],
+            "images": [],
+            "tables": [],
+            "interline_equations": [],
+            "discarded_blocks": [],
+            "para_blocks": [
+                {
+                    "type": "text",
+                    "bbox": [
+                        52,
+                        61.956024169921875,
+                        294,
+                        82.99800872802734
+                    ],
+                    "lines": [
+                        {
+                            "bbox": [
+                                52,
+                                61.956024169921875,
+                                294,
+                                72.0000228881836
+                            ],
+                            "spans": [
+                                {
+                                    "bbox": [
+                                        54.0,
+                                        61.956024169921875,
+                                        296.2261657714844,
+                                        72.0000228881836
+                                    ],
+                                    "content": "dependent on the service headway and the reliability of the departure ",
+                                    "type": "text",
+                                    "score": 1.0
+                                }
+                            ]
+                        }
+                    ]
+                }
+            ]
+        }
+    ],
+    "_backend": "pipeline",
+    "_version_name": "0.6.1"
+}
+

Content List (content_list.json)

File naming format: {original_filename}_content_list.json

Functionality

This is a simplified version of middle.json that stores all readable content blocks in reading order as a flat structure, removing complex layout information for easier subsequent processing.

Content Types

+ + + + + + + + + + + + + + + + + + + + + + + + + +

Type	Description
`image`	Image
`table`	Table
`text`	Text/Title
`equation`	Interline formula

Text Level Identification

Text levels are distinguished through the text_level field:

No text_level or text_level: 0: Body text
text_level: 1: Level 1 heading
text_level: 2: Level 2 heading
And so on...

Common Fields

All content blocks include a page_idx field indicating the page number (starting from 0).
All content blocks include a bbox field representing the bounding box coordinates of the content block [x0, y0, x1, y1], mapped to a range of 0-1000.

Sample Data

[
+        {
+        "type": "text",
+        "text": "The response of flow duration curves to afforestation ",
+        "text_level": 1, 
+        "bbox": [
+            62,
+            480,
+            946,
+            904
+        ],
+        "page_idx": 0
+    },
+    {
+        "type": "image",
+        "img_path": "images/a8ecda1c69b27e4f79fce1589175a9d721cbdc1cf78b4cc06a015f3746f6b9d8.jpg",
+        "image_caption": [
+            "Fig. 1. Annual flow duration curves of daily flows from Pine Creek, Australia, 1989–2000. "
+        ],
+        "image_footnote": [],
+        "bbox": [
+            62,
+            480,
+            946,
+            904
+        ],
+        "page_idx": 1
+    },
+    {
+        "type": "equation",
+        "img_path": "images/181ea56ef185060d04bf4e274685f3e072e922e7b839f093d482c29bf89b71e8.jpg",
+        "text": "$$\nQ _ { \\% } = f ( P ) + g ( T )\n$$",
+        "text_format": "latex",
+        "bbox": [
+            62,
+            480,
+            946,
+            904
+        ],
+        "page_idx": 2
+    },
+    {
+        "type": "table",
+        "img_path": "images/e3cb413394a475e555807ffdad913435940ec637873d673ee1b039e3bc3496d0.jpg",
+        "table_caption": [
+            "Table 2 Significance of the rainfall and time terms "
+        ],
+        "table_footnote": [
+            "indicates that the rainfall term was significant at the $5 \\%$ level, $T$ indicates that the time term was significant at the $5 \\%$ level, \\* represents significance at the $10 \\%$ level, and na denotes too few data points for meaningful analysis. "
+        ],
+        "table_body": "<html><body><table><tr><td rowspan=\"2\">Site</td><td colspan=\"10\">Percentile</td></tr><tr><td>10</td><td>20</td><td>30</td><td>40</td><td>50</td><td>60</td><td>70</td><td>80</td><td>90</td><td>100</td></tr><tr><td>Traralgon Ck</td><td>P</td><td>P,*</td><td>P</td><td>P</td><td>P,</td><td>P,</td><td>P,</td><td>P,</td><td>P</td><td>P</td></tr><tr><td>Redhill</td><td>P,T</td><td>P,T</td><td>，*</td><td>**</td><td>P.T</td><td>P,*</td><td>P*</td><td>P*</td><td>*</td><td>，*</td></tr><tr><td>Pine Ck</td><td></td><td>P,T</td><td>P,T</td><td>P,T</td><td>P,T</td><td>T</td><td>T</td><td>T</td><td>na</td><td>na</td></tr><tr><td>Stewarts Ck 5</td><td>P,T</td><td>P,T</td><td>P,T</td><td>P,T</td><td>P.T</td><td>P.T</td><td>P,T</td><td>na</td><td>na</td><td>na</td></tr><tr><td>Glendhu 2</td><td>P</td><td>P,T</td><td>P,*</td><td>P,T</td><td>P.T</td><td>P,ns</td><td>P,T</td><td>P,T</td><td>P,T</td><td>P,T</td></tr><tr><td>Cathedral Peak 2</td><td>P,T</td><td>P,T</td><td>P,T</td><td>P,T</td><td>P,T</td><td>*,T</td><td>P,T</td><td>P,T</td><td>P,T</td><td>T</td></tr><tr><td>Cathedral Peak 3</td><td>P.T</td><td>P.T</td><td>P,T</td><td>P,T</td><td>P,T</td><td>T</td><td>P,T</td><td>P,T</td><td>P,T</td><td>T</td></tr><tr><td>Lambrechtsbos A</td><td>P,T</td><td>P</td><td>P</td><td>P,T</td><td>*,T</td><td>*,T</td><td>*,T</td><td>*,T</td><td>*,T</td><td>T</td></tr><tr><td>Lambrechtsbos B</td><td>P,T</td><td>P,T</td><td>P,T</td><td>P,T</td><td>P,T</td><td>P,T</td><td>P,T</td><td>P,T</td><td>T</td><td>T</td></tr><tr><td>Biesievlei</td><td>P,T</td><td>P.T</td><td>P,T</td><td>P,T</td><td>*,T</td><td>*,T</td><td>T</td><td>T</td><td>P,T</td><td>P,T</td></tr></table></body></html>",
+        "bbox": [
+            62,
+            480,
+            946,
+            904
+        ],  
+        "page_idx": 5
+    }
+]
+

VLM Backend Output Results

Model Inference Results (model.json)

File naming format: {original_filename}_model.json

File format description

Two-level nested list: outer list = pages; inner list = content blocks of that page
Each block is a dict with at least: type, bbox, angle, content (some types add extra fields like score, block_tags, content_tags, format)
Designed for direct, raw model inspection

Supported content types (type field values)

{
+  "text": "Plain text",
+  "title": "Title",
+  "equation": "Display (interline) formula",
+  "image": "Image",
+  "image_caption": "Image caption",
+  "image_footnote": "Image footnote",
+  "table": "Table",
+  "table_caption": "Table caption",
+  "table_footnote": "Table footnote",
+  "phonetic": "Phonetic annotation",
+  "code": "Code block",
+  "code_caption": "Code caption",
+  "ref_text": "Reference / citation entry",
+  "algorithm": "Algorithm block (treated as code subtype)",
+  "list": "List container",
+  "header": "Page header",
+  "footer": "Page footer",
+  "page_number": "Page number",
+  "aside_text": "Side / margin note",
+  "page_footnote": "Page footnote"
+}
+

Coordinate system

bbox = [x0, y0, x1, y1] (top-left, bottom-right)
Origin at top-left of the page
All coordinates are normalized percentages in [0,1]

Sample data

[
+  [
+    {
+      "type": "header",
+      "bbox": [0.077, 0.095, 0.18, 0.181],
+      "angle": 0,
+      "score": null,
+      "block_tags": null,
+      "content": "ELSEVIER",
+      "format": null,
+      "content_tags": null
+    },
+    {
+      "type": "title",
+      "bbox": [0.157, 0.228, 0.833, 0.253],
+      "angle": 0,
+      "score": null,
+      "block_tags": null,
+      "content": "The response of flow duration curves to afforestation",
+      "format": null,
+      "content_tags": null
+    }
+  ]
+]
+

Intermediate Processing Results (middle.json)

File naming format: {original_filename}_middle.json

Structure is broadly similar to the pipeline backend, but with these differences:

list becomes a second‑level block, a new field sub_type distinguishes list categories:
- text: ordinary list
- ref_text: reference / bibliography style list
+
New code block type with sub_type(a code block always has at least a code_body, it may optionally have a code_caption):
- code
- algorithm
+
discarded_blocks may contain additional types:
- header
- footer
- page_number
- aside_text
- page_footnote
+
All blocks include an angle field indicating rotation (one of 0, 90, 180, 270).

Examples

Example: list block +

{
+  "bbox": [174,155,818,333],
+  "type": "list",
+  "angle": 0,
+  "index": 11,
+  "blocks": [
+    {
+      "bbox": [174,157,311,175],
+      "type": "text",
+      "angle": 0,
+      "lines": [
+        {
+          "bbox": [174,157,311,175],
+            "spans": [
+              {
+                "bbox": [174,157,311,175],
+                "type": "text",
+                "content": "H.1 Introduction"
+              }
+            ]
+        }
+      ],
+      "index": 3
+    },
+    {
+      "bbox": [175,182,464,229],
+      "type": "text",
+      "angle": 0,
+      "lines": [
+        {
+          "bbox": [175,182,464,229],
+          "spans": [
+            {
+              "bbox": [175,182,464,229],
+              "type": "text",
+              "content": "H.2 Example: Divide by Zero without Exception Handling"
+            }
+          ]
+        }
+      ],
+      "index": 4
+    }
+  ],
+  "sub_type": "text"
+}
+

Example: code block with optional caption: +

{
+  "type": "code",
+  "bbox": [114,780,885,1231],
+  "blocks": [
+    {
+      "bbox": [114,780,885,1231],
+      "lines": [
+        {
+          "bbox": [114,780,885,1231],
+          "spans": [
+            {
+              "bbox": [114,780,885,1231],
+              "type": "text",
+              "content": "1 // Fig. H.1: DivideByZeroNoExceptionHandling.java  \n2 // Integer division without exception handling.  \n3 import java.util.Scanner;  \n4  \n5 public class DivideByZeroNoExceptionHandling  \n6 {  \n7 // demonstrates throwing an exception when a divide-by-zero occurs  \n8 public static int quotient( int numerator, int denominator )  \n9 {  \n10 return numerator / denominator; // possible division by zero  \n11 } // end method quotient  \n12  \n13 public static void main(String[] args)  \n14 {  \n15 Scanner scanner = new Scanner(System.in); // scanner for input  \n16  \n17 System.out.print(\"Please enter an integer numerator: \");  \n18 int numerator = scanner.nextInt();  \n19 System.out.print(\"Please enter an integer denominator: \");  \n20 int denominator = scanner.nextInt();  \n21"
+            }
+          ]
+        }
+      ],
+      "index": 17,
+      "angle": 0,
+      "type": "code_body"
+    },
+    {
+      "bbox": [867,160,1280,189],
+      "lines": [
+        {
+          "bbox": [867,160,1280,189],
+          "spans": [
+            {
+              "bbox": [867,160,1280,189],
+              "type": "text",
+              "content": "Algorithm 1 Modules for MCTSteg"
+            }
+          ]
+        }
+      ],
+      "index": 19,
+      "angle": 0,
+      "type": "code_caption"
+    }
+  ],
+  "index": 17,
+  "sub_type": "code"
+}
+

Content List (content_list.json)

File naming format: {original_filename}_content_list.json

Based on the pipeline format, with these VLM-specific extensions:

New code type with sub_type (code | algorithm):
- Fields: code_body (string), optional code_caption (list of strings)
+
New list type with sub_type (text | ref_text):
- Field: list_items (array of strings)
+
All discarded_blocks entries are also output (e.g., headers, footers, page numbers, margin notes, page footnotes).
Existing types (image, table, text, equation) remain unchanged.
bbox still uses the 0–1000 normalized coordinate mapping.

Examples

Example: code (algorithm) entry +

{
+  "type": "code",
+  "sub_type": "algorithm",
+  "code_caption": ["Algorithm 1 Modules for MCTSteg"],
+  "code_body": "1: function GETCOORDINATE(d)  \n2:  $x \\gets d / l$ ,  $y \\gets d$  mod  $l$   \n3: return  $(x, y)$   \n4: end function  \n5: function BESTCHILD(v)  \n6:  $C \\gets$  child set of  $v$   \n7:  $v' \\gets \\arg \\max_{c \\in C} \\mathrm{UCTScore}(c)$   \n8:  $v'.n \\gets v'.n + 1$   \n9: return  $v'$   \n10: end function  \n11: function BACK PROPAGATE(v)  \n12: Calculate  $R$  using Equation 11  \n13: while  $v$  is not a root node do  \n14:  $v.r \\gets v.r + R$ ,  $v \\gets v.p$   \n15: end while  \n16: end function  \n17: function RANDOMSEARCH(v)  \n18: while  $v$  is not a leaf node do  \n19: Randomly select an untried action  $a \\in A(v)$   \n20: Create a new node  $v'$   \n21:  $(x, y) \\gets \\mathrm{GETCOORDINATE}(v'.d)$   \n22:  $v'.p \\gets v$ ,  $v'.d \\gets v.d + 1$ ,  $v'.\\Gamma \\gets v.\\Gamma$   \n23:  $v'.\\gamma_{x,y} \\gets a$   \n24: if  $a = -1$  then  \n25:  $v.lc \\gets v'$   \n26: else if  $a = 0$  then  \n27:  $v.mc \\gets v'$   \n28: else  \n29:  $v.rc \\gets v'$   \n30: end if  \n31:  $v \\gets v'$   \n32: end while  \n33: return  $v$   \n34: end function  \n35: function SEARCH(v)  \n36: while  $v$  is fully expanded do  \n37:  $v \\gets$  BESTCHILD(v)  \n38: end while  \n39: if  $v$  is not a leaf node then  \n40:  $v \\gets$  RANDOMSEARCH(v)  \n41: end if  \n42: return  $v$   \n43: end function",
+  "bbox": [510,87,881,740],
+  "page_idx": 0
+}
+

Example: list (text) entry +

{
+  "type": "list",
+  "sub_type": "text",
+  "list_items": [
+    "H.1 Introduction",
+    "H.2 Example: Divide by Zero without Exception Handling",
+    "H.3 Example: Divide by Zero with Exception Handling",
+    "H.4 Summary"
+  ],
+  "bbox": [174,155,818,333],
+  "page_idx": 0
+}
+

Example: discarded blocks output +

[
+  {
+    "type": "header",
+    "text": "Journal of Hydrology 310 (2005) 253-265",
+    "bbox": [363,164,623,177],
+    "page_idx": 0
+  },
+  {
+    "type": "page_footnote",
+    "text": "* Corresponding author. Address: Forest Science Centre, Department of Sustainability and Environment, P.O. Box 137, Heidelberg, Vic. 3084, Australia. Tel.: +61 3 9450 8719; fax: +61 3 9450 8644.",
+    "bbox": [71,815,915,841],
+    "page_idx": 0
+  }
+]
+

Summary

The above files constitute MinerU's complete output results. Users can choose appropriate files for subsequent processing based on their needs:

+
Model outputs (Use raw outputs):
+
- model.json
+
+
Debugging and verification (Use visualization files):
+
- layout.pdf
- span.pdf
+
+
Content extraction: (Use simplified files):
+
- *.md
- content_list.json
+
+
Secondary development: (Use structured files):
+
- middle.json
+

+ + + + + + + + + + + + + + + +

+ + + + + +

+ + + +

+ + + + + + + + + + \ No newline at end of file diff --git a/requirements.txt b/requirements.txt new file mode 100644 index 00000000..6669adad --- /dev/null +++ b/requirements.txt @@ -0,0 +1,4 @@ +mkdocs +mkdocs-static-i18n +markdown-gfm-admonition +mkdocs-video \ No newline at end of file diff --git a/search/search_index.json b/search/search_index.json new file mode 100644 index 00000000..d598a198 --- /dev/null +++ b/search/search_index.json @@ -0,0 +1 @@ +{"config":{"lang":["en","zh"],"separator":"[\\s\\-]+","pipeline":["stopWordFilter"]},"docs":[{"location":"","title":"MinerU","text":"

\ud83d\ude80MinerU Official Website\u2192\u2705 Zero-Install Online Version \u2705 Full-Featured Client \u2705 Developer API Online Access, skip deployment hassles, get all product formats with one click, go fast!

\ud83d\udc4b join us on Discord and WeChat

"},{"location":"#project-introduction","title":"Project Introduction","text":"

MinerU is a tool that converts PDFs into machine-readable formats (e.g., markdown, JSON), allowing for easy extraction into any format. MinerU was born during the pre-training process of InternLM. We focus on solving symbol conversion issues in scientific literature and hope to contribute to technological development in the era of large models. Compared to well-known commercial products domestically and internationally, MinerU is still young. If you encounter any issues or if the results are not as expected, please submit an issue on GitHub Issues and attach the relevant PDF.

"},{"location":"#key-features","title":"Key Features","text":"

Remove headers, footers, footnotes, page numbers and other elements to ensure semantic coherence
Output text in human reading order, suitable for single-column, multi-column and complex layouts
Retain the original document structure, including titles, paragraphs, lists, etc.
Extract images, image descriptions, tables, table titles and footnotes
Automatically identify and convert formulas in documents to LaTeX format
Automatically identify and convert tables in documents to HTML format
Automatically detect scanned PDFs and garbled PDFs, and enable OCR functionality
OCR supports detection and recognition of 109 languages
Support multiple output formats, such as multimodal and NLP Markdown, reading-order-sorted JSON, and information-rich intermediate formats
Support multiple visualization results, including layout visualization, span visualization, etc., for efficient confirmation of output effects and quality inspection
Support pure CPU environment operation, and support GPU(CUDA)/NPU(CANN)/MPS acceleration
Compatible with Windows, Linux and Mac platforms

"},{"location":"#user-guide","title":"User Guide","text":"

Quick Start Guide
Detailed Usage Instructions

"},{"location":"demo/","title":"Demo","text":""},{"location":"faq/","title":"FAQ","text":""},{"location":"faq/#frequently-asked-questions","title":"Frequently Asked Questions","text":"

If your question is not listed, try using DeepWiki's AI assistant for common issues.

For unresolved problems, join our Discord or WeChat community for support.

Encountered the error ImportError: libGL.so.1: cannot open shared object file: No such file or directory in Ubuntu 22.04 on WSL2

The libgl library is missing in Ubuntu 22.04 on WSL2. You can install the libgl library with the following command to resolve the issue:

sudo apt-get install libgl1-mesa-glx\n

Reference: #388

Missing text information in parsing results when installing and using on Linux systems.

MinerU uses pypdfium2 instead of pymupdf as the PDF page rendering engine in versions >=2.0 to resolve AGPLv3 license issues. On some Linux distributions, due to missing CJK fonts, some text may be lost during the process of rendering PDFs to images. To solve this problem, you can install the noto font package with the following commands, which are effective on Ubuntu/Debian systems:

sudo apt update\nsudo apt install fonts-noto-core\nsudo apt install fonts-noto-cjk\nfc-cache -fv\n

You can also directly use our Docker deployment method to build the image, which includes the above font packages by default.

Reference: #2915

"},{"location":"quick_start/","title":"Quick Start","text":""},{"location":"quick_start/#quick-start","title":"Quick Start","text":"

If you encounter any installation issues, please check the FAQ first.

"},{"location":"quick_start/#online-experience","title":"Online Experience","text":""},{"location":"quick_start/#official-online-web-application","title":"Official online web application","text":"

The official online version has the same functionality as the client, with a beautiful interface and rich features, requires login to use

"},{"location":"quick_start/#gradio-based-online-demo","title":"Gradio-based online demo","text":"

A WebUI developed based on Gradio, with a simple interface and only core parsing functionality, no login required

"},{"location":"quick_start/#local-deployment","title":"Local Deployment","text":"

Warning

Prerequisites - Hardware and Software Environment Support

By concentrating our resources and efforts on mainstream environments, our team can more efficiently resolve potential bugs and timely develop new features.

Parsing Backend pipeline *-auto-engine *-http-client hybrid vlm hybrid vlm Backend Features Good Compatibility High Hardware Requirements For OpenAI Compatible Servers² Accuracy¹ 82+ 90+ Operating System Linux³ / Windows⁴ / macOS⁵ Pure CPU Support \u2705 \u274c \u2705 GPU Acceleration Volta and later architecture GPUs or Apple Silicon Not Required Min VRAM 6GB 10GB 8GB 3GB RAM Min 16GB+, Recommended 32GB+ 8GB Disk Space 20GB+, SSD Recommended 2GB Python Version 3.10-3.13

¹ Accuracy metrics are the End-to-End Evaluation Overall scores from OmniDocBench (v1.5), based on the latest version of MinerU. ² Servers compatible with OpenAI API, such as local model servers or remote model services deployed via inference frameworks like vLLM/SGLang/LMDeploy. ³ Linux only supports distributions from 2019 and later. ⁴ Since the key dependency ray does not support Python 3.13 on Windows, only versions 3.10~3.12 are supported. ⁵ macOS requires version 14.0 or later.

"},{"location":"quick_start/#install-mineru","title":"Install MinerU","text":""},{"location":"quick_start/#install-mineru-using-pip-or-uv","title":"Install MinerU using pip or uv","text":"

pip install --upgrade pip\npip install uv\nuv pip install -U \"mineru[all]\"\n

"},{"location":"quick_start/#install-mineru-from-source-code","title":"Install MinerU from source code","text":"

git clone https://github.com/opendatalab/MinerU.git\ncd MinerU\nuv pip install -e .[all]\n

Tip

mineru[all] includes all core features, compatible with Windows / Linux / macOS systems, suitable for most users. If you need to specify the inference framework for the VLM model, or only intend to install a lightweight client on an edge device, please refer to the documentation Extension Modules Installation Guide.

"},{"location":"quick_start/#deploy-mineru-using-docker","title":"Deploy MinerU using Docker","text":"

MinerU provides a convenient Docker deployment method, which helps quickly set up the environment and solve some tricky environment compatibility issues. You can get the Docker Deployment Instructions in the documentation.

"},{"location":"quick_start/#using-mineru","title":"Using MinerU","text":"

If your device meets the GPU acceleration requirements in the table above, you can use a simple command line for document parsing:

mineru -p <input_path> -o <output_path>\n

If your device does not meet the GPU acceleration requirements, you can specify the backend as pipeline to run in a pure CPU environment:

mineru -p <input_path> -o <output_path> -b pipeline\n

You can use MinerU for PDF parsing through various methods such as command line, API, and WebUI. For detailed instructions, please refer to the Usage Guide.

"},{"location":"quick_start/docker_deployment/","title":"Docker Deployment","text":""},{"location":"quick_start/docker_deployment/#deploying-mineru-with-docker","title":"Deploying MinerU with Docker","text":"

MinerU provides a convenient Docker deployment method, which helps quickly set up the environment and solve some tricky environment compatibility issues.

"},{"location":"quick_start/docker_deployment/#build-docker-image-using-dockerfile","title":"Build Docker Image using Dockerfile","text":"

wget https://gcore.jsdelivr.net/gh/opendatalab/MinerU@master/docker/global/Dockerfile\ndocker build -t mineru:latest -f Dockerfile .\n

"},{"location":"quick_start/docker_deployment/#docker-description","title":"Docker Description","text":"

Note

Requirements for using vllm to accelerate VLM model inference:

Device must have Volta architecture or later graphics cards with 8GB+ available VRAM.
The host machine's graphics driver should support CUDA 12.9.1 or higher; You can check the driver version using the nvidia-smi command.
Docker container must have access to the host machine's graphics devices.

"},{"location":"quick_start/docker_deployment/#start-docker-container","title":"Start Docker Container","text":"

docker run --gpus all \\\n  --shm-size 32g \\\n  -p 30000:30000 -p 7860:7860 -p 8000:8000 \\\n  --ipc=host \\\n  -it mineru:latest \\\n  /bin/bash\n

After executing this command, you will enter the Docker container's interactive terminal with some ports mapped for potential services. You can directly run MinerU-related commands within the container to use MinerU's features. You can also directly start MinerU services by replacing /bin/bash with service startup commands. For detailed instructions, please refer to the Start the service via command.

"},{"location":"quick_start/docker_deployment/#start-services-directly-with-docker-compose","title":"Start Services Directly with Docker Compose","text":"

We provide a compose.yaml file that you can use to quickly start MinerU services.

# Download compose.yaml file\nwget https://gcore.jsdelivr.net/gh/opendatalab/MinerU@master/docker/compose.yaml\n

Note

The compose.yaml file contains configurations for multiple services of MinerU, you can choose to start specific services as needed.
Different services might have additional parameter configurations, which you can view and edit in the compose.yaml file.
Due to the pre-allocation of GPU memory by the vllm inference acceleration framework, you may not be able to run multiple vllm services simultaneously on the same machine. Therefore, ensure that other services that might use GPU memory have been stopped before starting the vlm-openai-server service or using the vlm-vllm-engine backend.

"},{"location":"quick_start/docker_deployment/#start-openai-compatible-server-service","title":"Start OpenAI-compatible server service","text":"

connect to openai-server via vlm-http-client backend

docker compose -f compose.yaml --profile openai-server up -d\n

Tip

In another terminal, connect to openai server via http client (only requires CPU and network, no vllm environment needed)

mineru -p <input_path> -o <output_path> -b vlm-http-client -u http://<server_ip>:30000\n

"},{"location":"quick_start/docker_deployment/#start-web-api-service","title":"Start Web API service","text":"

docker compose -f compose.yaml --profile api up -d\n

Tip

Access http://<server_ip>:8000/docs in your browser to view the API documentation.

"},{"location":"quick_start/docker_deployment/#start-gradio-webui-service","title":"Start Gradio WebUI service","text":"

docker compose -f compose.yaml --profile gradio up -d\n

Tip

Access http://<server_ip>:7860 in your browser to use the Gradio WebUI.

"},{"location":"quick_start/extension_modules/","title":"Extension Modules","text":""},{"location":"quick_start/extension_modules/#mineru-extension-modules-installation-guide","title":"MinerU Extension Modules Installation Guide","text":"

MinerU supports installing extension modules on demand based on different needs to enhance functionality or support specific model backends.

"},{"location":"quick_start/extension_modules/#common-scenarios","title":"Common Scenarios","text":""},{"location":"quick_start/extension_modules/#core-functionality-installation","title":"Core Functionality Installation","text":"

The core module is the core dependency of MinerU, containing all functional modules except vllm/lmdeploy. Installing this module ensures the basic functionality of MinerU works properly.

uv pip install \"mineru[core]\"\n

"},{"location":"quick_start/extension_modules/#using-vllm-to-accelerate-vlm-model-inference","title":"Using vllm to Accelerate VLM Model Inference","text":"

Note

uv pip install \"mineru[core,vllm]\"\n

Tip

If exceptions occur during installation of the extra package including vllm, please refer to the vllm official documentation to try to resolve the issue, or directly use the Docker deployment method.

"},{"location":"quick_start/extension_modules/#using-lmdeploy-to-accelerate-vlm-model-inference","title":"Using lmdeploy to Accelerate VLM Model Inference","text":"

Note

uv pip install \"mineru[core,lmdeploy]\"\n

Tip

If exceptions occur during installation of the extra package including lmdeploy, please refer to the lmdeploy official documentation to try to resolve the issue.

"},{"location":"quick_start/extension_modules/#installing-lightweight-client-to-connect-to-openai-compatible-servers-for-vlm-http-client-mode","title":"Installing Lightweight Client to Connect to OpenAI-compatible servers (for vlm-http-client mode)","text":"

uv pip install mineru\nmineru -p <input_path> -o <output_path> -b vlm-http-client -u http://127.0.0.1:30000\n

"},{"location":"quick_start/extension_modules/#installing-lightweight-client-to-connect-to-openai-compatible-servers-for-hybrid-http-client-mode","title":"Installing Lightweight Client to Connect to OpenAI-compatible servers (for hybrid-http-client mode)","text":"

uv pip install \"mineru[pipeline]\"\nmineru -p <input_path> -o <output_path> -b hybrid-http-client -u http://127.0.0.1:30000\n

"},{"location":"reference/","title":"Reference","text":""},{"location":"reference/#reference-documentation","title":"Reference Documentation","text":"

This section provides detailed reference materials for MinerU project. Here you can find technical specifications, API documentation, output file formats, and version history.

"},{"location":"reference/#table-of-contents","title":"Table of Contents","text":"

Output Files Documentation - Detailed explanation of all output files and their formats
Changelog - Version update history and release notes

"},{"location":"reference/#documentation-overview","title":"Documentation Overview","text":""},{"location":"reference/#output-files-documentation","title":"Output Files Documentation","text":"

Understanding the output files generated by MinerU is crucial for effective use of the tool. The output files documentation provides:

Visual debugging files: Help you understand the document parsing process
Structured data files: Contain detailed parsing results for further processing
File format specifications: Detailed descriptions of each output file type

"},{"location":"reference/#changelog","title":"Changelog","text":"

The changelog documents the evolution of MinerU, including:

Version updates: New features and improvements for each release
Bug fixes: Issues resolved in each version
Breaking changes: Important changes that may affect your usage
Deprecations: Features that are being phased out

"},{"location":"reference/changelog/","title":"Changelog","text":""},{"location":"reference/changelog/#changelog","title":"Changelog","text":"

This document records the update history of MinerU project for version 2.6.7 and earlier. For the latest version updates, please check the project README.

"},{"location":"reference/changelog/#26-series-versions","title":"2.6 Series Versions","text":""},{"location":"reference/changelog/#267-20251212","title":"2.6.7 (2025/12/12)","text":"

Bug fix: #4168

"},{"location":"reference/changelog/#266-20251202","title":"2.6.6 (2025/12/02)","text":"

mineru-api tool optimizations

Added descriptive text to mineru-api interface parameters to improve API documentation readability.
You can use the environment variable MINERU_API_ENABLE_FASTAPI_DOCS to control whether the auto-generated interface documentation page is enabled (enabled by default).
Added concurrency configuration options for the vlm-vllm-async-engine, vlm-lmdeploy-engine, and vlm-http-client backends. Users can use the environment variable MINERU_API_MAX_CONCURRENT_REQUESTS to set the maximum number of concurrent API requests (unlimited by default).

"},{"location":"reference/changelog/#265-20251126","title":"2.6.5 (2025/11/26)","text":"

Added support for a new backend vlm-lmdeploy-engine. Its usage is similar to vlm-vllm-(async)engine, but it uses lmdeploy as the inference engine and additionally supports native inference acceleration on Windows platforms compared to vllm.

"},{"location":"reference/changelog/#264-20251104","title":"2.6.4 (2025/11/04)","text":"

Added timeout configuration for PDF image rendering, default is 300 seconds, can be configured via environment variable MINERU_PDF_RENDER_TIMEOUT to prevent long blocking of the rendering process caused by some abnormal PDF files.
Added CPU thread count configuration options for ONNX models, default is the system CPU core count, can be configured via environment variables MINERU_INTRA_OP_NUM_THREADS and MINERU_INTER_OP_NUM_THREADS to reduce CPU resource contention conflicts in high concurrency scenarios.

"},{"location":"reference/changelog/#263-20251031","title":"2.6.3 (2025/10/31)","text":"

Added support for a new backend vlm-mlx-engine, enabling MLX-accelerated inference for the MinerU2.5 model on Apple Silicon devices. Compared to the vlm-transformers backend, vlm-mlx-engine delivers a 100%\u2013200% speed improvement.
Bug fixes: #3849, #3859

"},{"location":"reference/changelog/#262-20251024","title":"2.6.2 (2025/10/24)","text":"

pipeline backend optimizations

Added experimental support for Chinese formulas, which can be enabled by setting the environment variable export MINERU_FORMULA_CH_SUPPORT=1. This feature may cause a slight decrease in MFR speed and failures in recognizing some long formulas. It is recommended to enable it only when parsing Chinese formulas is needed. To disable this feature, set the environment variable to 0.
OCR speed significantly improved by 200%~300%, thanks to the optimization solution provided by @cjsdurj
OCR models optimized for improved accuracy and coverage of Latin script recognition, and updated Cyrillic, Arabic, Devanagari, Telugu (te), and Tamil (ta) language systems to ppocr-v5 version, with accuracy improved by over 40% compared to previous models

vlm backend optimizations

table_caption and table_footnote matching logic optimized to improve the accuracy of table caption and footnote matching and reading order rationality in scenarios with multiple consecutive tables on a page
Optimized CPU resource usage during high concurrency when using vllm backend, reducing server pressure
Adapted to vllm version 0.11.0

General optimizations

Cross-page table merging effect optimized, added support for cross-page continuation table merging, improving table merging effectiveness in multi-column merge scenarios
Added environment variable configuration option MINERU_TABLE_MERGE_ENABLE for table merging feature. Table merging is enabled by default and can be disabled by setting this variable to 0

"},{"location":"reference/changelog/#25-series-versions","title":"2.5 Series Versions","text":""},{"location":"reference/changelog/#254-20250926","title":"2.5.4 (2025/09/26)","text":"

\ud83c\udf89\ud83c\udf89 The MinerU2.5 Technical Report is now available! We welcome you to read it for a comprehensive overview of its model architecture, training strategy, data engineering and evaluation results.
Fixed an issue where some PDF files were mistakenly identified as AI files, causing parsing failures

"},{"location":"reference/changelog/#253-20250920","title":"2.5.3 (2025/09/20)","text":"

Dependency version range adjustment to enable Turing and earlier architecture GPUs to use vLLM acceleration for MinerU2.5 model inference.
pipeline backend compatibility fixes for torch 2.8.0.
Reduced default concurrency for vLLM async backend to lower server pressure and avoid connection closure issues caused by high load.
More compatibility-related details can be found in the announcement

"},{"location":"reference/changelog/#252-20250919","title":"2.5.2 (2025/09/19)","text":"

We are officially releasing MinerU2.5, currently the most powerful multimodal large model for document parsing.

With only 1.2B parameters, MinerU2.5's accuracy on the OmniDocBench benchmark comprehensively surpasses top-tier multimodal models like Gemini 2.5 Pro, GPT-4o, and Qwen2.5-VL-72B. It also significantly outperforms leading specialized models such as dots.ocr, MonkeyOCR, and PP-StructureV3.

The model has been released on HuggingFace and ModelScope platforms. Welcome to download and use!

Core Highlights

SOTA Performance with Extreme Efficiency: As a 1.2B model, it achieves State-of-the-Art (SOTA) results that exceed models in the 10B and 100B+ classes, redefining the performance-per-parameter standard in document AI.
Advanced Architecture for Across-the-Board Leadership: By combining a two-stage inference pipeline (decoupling layout analysis from content recognition) with a native high-resolution architecture, it achieves SOTA performance across five key areas: layout analysis, text recognition, formula recognition, table recognition, and reading order.

Key Capability Enhancements

Layout Detection: Delivers more complete results by accurately covering non-body content like headers, footers, and page numbers. It also provides more precise element localization and natural format reconstruction for lists and references.
Table Parsing: Drastically improves parsing for challenging cases, including rotated tables, borderless/semi-structured tables, and long/complex tables.
Formula Recognition: Significantly boosts accuracy for complex, long-form, and hybrid Chinese-English formulas, greatly enhancing the parsing capability for mathematical documents.

Repository Adjustments

Additionally, with the release of vlm 2.5, we have made some adjustments to the repository:

The vlm backend has been upgraded to version 2.5, supporting the MinerU2.5 model and no longer compatible with the MinerU2.0-2505-0.9B model. The last version supporting the 2.0 model is mineru-2.2.2.
VLM inference-related code has been moved to mineru_vl_utils, reducing coupling with the main mineru repository and facilitating independent iteration in the future.
The vlm accelerated inference framework has been switched from sglang to vllm, achieving full compatibility with the vllm ecosystem, allowing users to use the MinerU2.5 model and accelerated inference on any platform that supports the vllm framework.
Due to major upgrades in the vlm model supporting more layout types, we have made some adjustments to the structure of the parsing intermediate file middle.json and result file content_list.json. Please refer to the documentation for details.

Other Repository Optimizations

Removed file extension whitelist validation for input files. When input files are PDF documents or images, there are no longer requirements for file extensions, improving usability.

"},{"location":"reference/changelog/#22-24-series-versions","title":"2.2 - 2.4 Series Versions","text":""},{"location":"reference/changelog/#222-20250910","title":"2.2.2 (2025/09/10)","text":"

Fixed the issue where the new table recognition model would affect the overall parsing task when some table parsing failed

"},{"location":"reference/changelog/#221-20250908","title":"2.2.1 (2025/09/08)","text":"

Fixed the issue where some newly added models were not downloaded when using the model download command.

"},{"location":"reference/changelog/#220-20250905","title":"2.2.0 (2025/09/05)","text":"

Major Updates

In this version, we focused on improving table parsing accuracy by introducing a new wired table recognition model and a brand-new hybrid table structure parsing algorithm, significantly enhancing the table recognition capabilities of the pipeline backend.
We also added support for cross-page table merging, which is supported by both pipeline and vlm backends, further improving the completeness and accuracy of table parsing.

Other Updates

The pipeline backend now supports 270-degree rotated table parsing, bringing support for table parsing in 0/90/270-degree orientations
pipeline added OCR capability support for Thai and Greek, and updated the English OCR model to the latest version. English recognition accuracy improved by 11%, Thai recognition model accuracy is 82.68%, and Greek recognition model accuracy is 89.28% (by PPOCRv5)
Added bbox field (mapped to 0-1000 range) in the output content_list.json, making it convenient for users to directly obtain position information for each content block
Removed the pipeline_old_linux installation option, no longer supporting legacy Linux systems such as CentOS 7, to provide better support for uv's sync/run commands

"},{"location":"reference/changelog/#21-series-versions","title":"2.1 Series Versions","text":""},{"location":"reference/changelog/#2110-20250801","title":"2.1.10 (2025/08/01)","text":"

Fixed an issue in the pipeline backend where block overlap caused the parsing results to deviate from expectations #3232

"},{"location":"reference/changelog/#219-20250730","title":"2.1.9 (2025/07/30)","text":"

transformers 4.54.1 version adaptation

"},{"location":"reference/changelog/#218-20250728","title":"2.1.8 (2025/07/28)","text":"

sglang 0.4.9.post5 version adaptation

"},{"location":"reference/changelog/#217-20250727","title":"2.1.7 (2025/07/27)","text":"

transformers 4.54.0 version adaptation

"},{"location":"reference/changelog/#216-20250726","title":"2.1.6 (2025/07/26)","text":"

Fixed table parsing issues in handwritten documents when using vlm backend
Fixed visualization box position drift issue when document is rotated #3175

"},{"location":"reference/changelog/#215-20250724","title":"2.1.5 (2025/07/24)","text":"

sglang 0.4.9 version adaptation, synchronously upgrading the dockerfile base image to sglang 0.4.9.post3

"},{"location":"reference/changelog/#214-20250723","title":"2.1.4 (2025/07/23)","text":"

Bug Fixes

Fixed the issue of excessive memory consumption during the MFR step in the pipeline backend under certain scenarios #2771
Fixed the inaccurate matching between image/table and caption/footnote under certain conditions #3129

"},{"location":"reference/changelog/#211-20250716","title":"2.1.1 (2025/07/16)","text":"

Bug fixes

Fixed text block content loss issue that could occur in certain pipeline scenarios #3005
Fixed issue where sglang-client required unnecessary packages like torch #2968
Updated dockerfile to fix incomplete text content parsing due to missing fonts in Linux #2915

Usability improvements

Updated compose.yaml to facilitate direct startup of sglang-server, mineru-api, and mineru-gradio services
Launched brand new online documentation site, simplified readme, providing better documentation experience

"},{"location":"reference/changelog/#210-20250705","title":"2.1.0 (2025/07/05)","text":"

This is the first major update of MinerU 2, which includes a large number of new features and improvements, covering significant performance optimizations, user experience enhancements, and bug fixes. The detailed update contents are as follows:

Performance Optimizations

Significantly improved preprocessing speed for documents with specific resolutions (around 2000 pixels on the long side).
Greatly enhanced post-processing speed when the pipeline backend handles batch processing of documents with fewer pages (<10 pages).
Layout analysis speed of the pipeline backend has been increased by approximately 20%.

Experience Enhancements

Built-in ready-to-use fastapi service and gradio webui. For detailed usage instructions, please refer to Documentation.
Adapted to sglang version 0.4.8, significantly reducing the GPU memory requirements for the vlm-sglang backend. It can now run on graphics cards with as little as 8GB GPU memory (Turing architecture or newer).
Added transparent parameter passing for all commands related to sglang, allowing the sglang-engine backend to receive all sglang parameters consistently with the sglang-server.
Supports feature extensions based on configuration files, including custom formula delimiters, enabling heading classification, and customizing local model directories. For detailed usage instructions, please refer to Documentation.

New Features

Updated the pipeline backend with the PP-OCRv5 multilingual text recognition model, supporting text recognition in 37 languages such as French, Spanish, Portuguese, Russian, and Korean, with an average accuracy improvement of over 30%. Details
Introduced limited support for vertical text layout in the pipeline backend.

"},{"location":"reference/changelog/#20-series-versions","title":"2.0 Series Versions","text":""},{"location":"reference/changelog/#206-20250620","title":"2.0.6 (2025/06/20)","text":"

Fixed occasional parsing interruptions caused by invalid block content in vlm mode
Fixed parsing interruptions caused by incomplete table structures in vlm mode

"},{"location":"reference/changelog/#205-20250617","title":"2.0.5 (2025/06/17)","text":"

Fixed the issue where models were still required to be downloaded in the sglang-client mode
Fixed the issue where the sglang-client mode unnecessarily depended on packages like torch during runtime.
Fixed the issue where only the first instance would take effect when attempting to launch multiple sglang-client instances via multiple URLs within the same process

"},{"location":"reference/changelog/#203-20250615","title":"2.0.3 (2025/06/15)","text":"

Fixed a configuration file key-value update error that occurred when downloading model type was set to all
Fixed the issue where the formula and table feature toggle switches were not working in command line mode, causing the features to remain enabled.
Fixed compatibility issues with sglang version 0.4.7 in the sglang-engine mode.
Updated Dockerfile and installation documentation for deploying the full version of MinerU in sglang environment

"},{"location":"reference/changelog/#200-20250613","title":"2.0.0 (2025/06/13)","text":"

New Architecture

MinerU 2.0 has been deeply restructured in code organization and interaction methods, significantly improving system usability, maintainability, and extensibility.

Removal of Third-party Dependency Limitations: Completely eliminated the dependency on pymupdf, moving the project toward a more open and compliant open-source direction.
Ready-to-use, Easy Configuration: No need to manually edit JSON configuration files; most parameters can now be set directly via command line or API.
Automatic Model Management: Added automatic model download and update mechanisms, allowing users to complete model deployment without manual intervention.
Offline Deployment Friendly: Provides built-in model download commands, supporting deployment requirements in completely offline environments.
Streamlined Code Structure: Removed thousands of lines of redundant code, simplified class inheritance logic, significantly improving code readability and development efficiency.
Unified Intermediate Format Output: Adopted standardized middle_json format, compatible with most secondary development scenarios based on this format, ensuring seamless ecosystem business migration.

New Model

MinerU 2.0 integrates our latest small-parameter, high-performance multimodal document parsing model, achieving end-to-end high-speed, high-precision document understanding.

Small Model, Big Capabilities: With parameters under 1B, yet surpassing traditional 72B-level vision-language models (VLMs) in parsing accuracy.
Multiple Functions in One: A single model covers multilingual recognition, handwriting recognition, layout analysis, table parsing, formula recognition, reading order sorting, and other core tasks.
Ultimate Inference Speed: Achieves peak throughput exceeding 10,000 tokens/s through sglang acceleration on a single NVIDIA 4090 card, easily handling large-scale document processing requirements.
Online Experience: You can experience our brand-new VLM model on MinerU.net, Hugging Face, and ModelScope.

Incompatible Changes Notice

To improve overall architectural rationality and long-term maintainability, this version contains some incompatible changes:

Python package name changed from magic-pdf to mineru, and the command-line tool changed from magic-pdf to mineru. Please update your scripts and command calls accordingly.
For modular system design and ecosystem consistency considerations, MinerU 2.0 no longer includes the LibreOffice document conversion module. If you need to process Office documents, we recommend converting them to PDF format through an independently deployed LibreOffice service before proceeding with subsequent parsing operations.

"},{"location":"reference/changelog/#1x-series-historical-versions","title":"1.x Series Historical Versions","text":""},{"location":"reference/changelog/#1312-20250524","title":"1.3.12 (2025/05/24)","text":"

Added support for PPOCRv5 models, updated ch_server model to PP-OCRv5_rec_server, and ch_lite model to PP-OCRv5_rec_mobile (model update required)

In testing, we found that PPOCRv5(server) has some improvement for handwritten documents, but has slightly lower accuracy than v4_server_doc for other document types, so the default ch model remains unchanged as PP-OCRv4_server_rec_doc.
Since PPOCRv5 has enhanced recognition capabilities for handwriting and special characters, you can manually choose the PPOCRv5 model for Japanese-Traditional Chinese mixed scenarios and handwritten documents
You can select the appropriate model through the lang parameter lang='ch_server' (Python API) or --lang ch_server (command line):
ch: PP-OCRv4_server_rec_doc (default) (Chinese/English/Japanese/Traditional Chinese mixed/15K dictionary)
ch_server: PP-OCRv5_rec_server (Chinese/English/Japanese/Traditional Chinese mixed + handwriting/18K dictionary)
ch_lite: PP-OCRv5_rec_mobile (Chinese/English/Japanese/Traditional Chinese mixed + handwriting/18K dictionary)
ch_server_v4: PP-OCRv4_rec_server (Chinese/English mixed/6K dictionary)
ch_lite_v4: PP-OCRv4_rec_mobile (Chinese/English mixed/6K dictionary)

Added support for handwritten documents through optimized layout recognition of handwritten text areas

This feature is supported by default, no additional configuration required
You can refer to the instructions above to manually select the PPOCRv5 model for better handwritten document parsing results

The huggingface and modelscope demos have been updated to versions that support handwriting recognition and PPOCRv5 models, which you can experience online

"},{"location":"reference/changelog/#1310-20250429","title":"1.3.10 (2025/04/29)","text":"

Added support for custom formula delimiters, which can be configured by modifying the latex-delimiter-config section in the magic-pdf.json file in your user directory.

"},{"location":"reference/changelog/#139-20250427","title":"1.3.9 (2025/04/27)","text":"

Optimized formula parsing functionality, improved formula rendering success rate

"},{"location":"reference/changelog/#138-20250423","title":"1.3.8 (2025/04/23)","text":"

The default ocr model (ch) has been updated to PP-OCRv4_server_rec_doc (model update required)

PP-OCRv4_server_rec_doc is trained on a mixture of more Chinese document data and PP-OCR training data based on PP-OCRv4_server_rec, adding recognition capabilities for some traditional Chinese characters, Japanese, and special characters. It can recognize over 15,000 characters and improves both document-specific and general text recognition abilities.
Performance comparison of PP-OCRv4_server_rec_doc/PP-OCRv4_server_rec/PP-OCRv4_mobile_rec
After verification, the PP-OCRv4_server_rec_doc model shows significant accuracy improvements in Chinese/English/Japanese/Traditional Chinese in both single language and mixed language scenarios, with comparable speed to PP-OCRv4_server_rec, making it suitable for most use cases.
In some pure English scenarios, PP-OCRv4_server_rec_doc may have word adhesion issues, while PP-OCRv4_server_rec performs better in these cases. Therefore, we've kept the PP-OCRv4_server_rec model, which users can access by adding the parameter lang='ch_server' (Python API) or --lang ch_server (command line).

"},{"location":"reference/changelog/#137-20250422","title":"1.3.7 (2025/04/22)","text":"

Fixed the issue where the lang parameter was ineffective during table parsing model initialization
Fixed the significant speed reduction of OCR and table parsing in cpu mode

"},{"location":"reference/changelog/#134-20250416","title":"1.3.4 (2025/04/16)","text":"

Slightly improved OCR-det speed by removing some unnecessary blocks
Fixed page-internal sorting errors caused by footnotes in certain cases

"},{"location":"reference/changelog/#132-20250412","title":"1.3.2 (2025/04/12)","text":"

Fixed dependency version incompatibility issues when installing on Windows with Python 3.13
Optimized memory usage during batch inference
Improved parsing of tables rotated 90 degrees
Enhanced parsing of oversized tables in financial report samples
Fixed the occasional word adhesion issue in English text areas when OCR language is not specified (model update required)

"},{"location":"reference/changelog/#131-20250408","title":"1.3.1 (2025/04/08)","text":"

Fixed several compatibility issues

Added support for Python 3.13
Made final adaptations for outdated Linux systems (such as CentOS 7) with no guarantee of continued support in future versions, installation instructions

"},{"location":"reference/changelog/#130-20250403","title":"1.3.0 (2025/04/03)","text":"

Installation and compatibility optimizations

Resolved compatibility issues caused by detectron2 by removing layoutlmv3 usage in layout
Extended torch version compatibility to 2.2~2.6 (excluding 2.5)
Added CUDA compatibility for versions 11.8/12.4/12.6/12.8 (CUDA version determined by torch), solving compatibility issues for users with 50-series and H-series GPUs
Extended Python compatibility to versions 3.10~3.12, fixing the issue of automatic downgrade to version 0.6.1 when installing in non-3.10 environments
Optimized offline deployment process, eliminating the need to download any model files after successful deployment

Performance optimizations

Enhanced parsing speed for batches of small files by supporting batch processing of multiple PDF files (script example), with formula parsing speed improved by up to 1400% and overall parsing speed improved by up to 500% compared to version 1.0.1
Reduced memory usage and improved parsing speed by optimizing MFR model loading and usage (requires re-running the model download process to get incremental updates to model files)
Optimized GPU memory usage, requiring only 6GB minimum to run this project
Improved running speed on MPS devices

Parsing effect optimizations

Updated MFR model to unimernet(2503), fixing line break loss issues in multi-line formulas

Usability optimizations

Completely replaced the paddle framework and paddleocr in the project by using paddleocr2torch, resolving conflicts between paddle and torch, as well as thread safety issues caused by the paddle framework
Added real-time progress bar display during parsing, allowing precise tracking of parsing progress and making the waiting process more bearable

"},{"location":"reference/changelog/#121-20250303","title":"1.2.1 (2025/03/03)","text":"

Fixed some issues

Fixed the impact on punctuation marks during full-width to half-width conversion of letters and numbers
Fixed caption matching inaccuracies in certain scenarios
Fixed formula span loss issues in certain scenarios

"},{"location":"reference/changelog/#120-20250224","title":"1.2.0 (2025/02/24)","text":"

This version includes several fixes and improvements to enhance parsing efficiency and accuracy:

Performance Optimization

Increased classification speed for PDF documents in auto mode.

Parsing Optimization

Improved parsing logic for documents containing watermarks, significantly enhancing the parsing results for such documents.
Enhanced the matching logic for multiple images/tables and captions within a single page, improving the accuracy of image-text matching in complex layouts.

Bug Fixes

Fixed an issue where image/table spans were incorrectly filled into text blocks under certain conditions.
Resolved an issue where title blocks were empty in some cases.

"},{"location":"reference/changelog/#110-20250122","title":"1.1.0 (2025/01/22)","text":"

In this version we have focused on improving parsing accuracy and efficiency:

Model capability upgrade (requires re-executing the model download process to obtain incremental updates of model files)

The layout recognition model has been upgraded to the latest doclayout_yolo(2501) model, improving layout recognition accuracy.
The formula parsing model has been upgraded to the latest unimernet(2501) model, improving formula recognition accuracy.

Performance optimization

On devices that meet certain configuration requirements (16GB+ VRAM), by optimizing resource usage and restructuring the processing pipeline, overall parsing speed has been increased by more than 50%.

Parsing effect optimization

Added a new heading classification feature (testing version, enabled by default) to the online demo (mineru.net/huggingface/modelscope), which supports hierarchical classification of headings, thereby enhancing document structuring.

"},{"location":"reference/changelog/#101-20250110","title":"1.0.1 (2025/01/10)","text":"

This is our first official release, where we have introduced a completely new API interface and enhanced compatibility through extensive refactoring, as well as a brand new automatic language identification feature:

New API Interface

For the data-side API, we have introduced the Dataset class, designed to provide a robust and flexible data processing framework. This framework currently supports a variety of document formats, including images (.jpg and .png), PDFs, Word documents (.doc and .docx), and PowerPoint presentations (.ppt and .pptx). It ensures effective support for data processing tasks ranging from simple to complex.
For the user-side API, we have meticulously designed the MinerU processing workflow as a series of composable Stages. Each Stage represents a specific processing step, allowing users to define new Stages according to their needs and creatively combine these stages to customize their data processing workflows.

Enhanced Compatibility

By optimizing the dependency environment and configuration items, we ensure stable and efficient operation on ARM architecture Linux systems.
We have deeply integrated with Huawei Ascend NPU acceleration, providing autonomous and controllable high-performance computing capabilities. This supports the localization and development of AI application platforms in China. Ascend NPU Acceleration

Automatic Language Identification

By introducing a new language recognition model, setting the lang configuration to auto during document parsing will automatically select the appropriate OCR language model, improving the accuracy of scanned document parsing.

"},{"location":"reference/changelog/#0x-series-historical-versions","title":"0.x Series Historical Versions","text":""},{"location":"reference/changelog/#0100-20241122","title":"0.10.0 (2024/11/22)","text":"

Introducing hybrid OCR text extraction capabilities:

Significantly improved parsing performance in complex text distribution scenarios such as dense formulas, irregular span regions, and text represented by images.
Combines the dual advantages of accurate content extraction and faster speed in text mode, and more precise span/line region recognition in OCR mode.

"},{"location":"reference/changelog/#093-20241115","title":"0.9.3 (2024/11/15)","text":"

Integrated RapidTable for table recognition, improving single-table parsing speed by more than 10 times, with higher accuracy and lower GPU memory usage.

"},{"location":"reference/changelog/#092-20241106","title":"0.9.2 (2024/11/06)","text":"

Integrated the StructTable-InternVL2-1B model for table recognition functionality.

"},{"location":"reference/changelog/#090-20241031","title":"0.9.0 (2024/10/31)","text":"

This is a major new version with extensive code refactoring, addressing numerous issues, improving performance, reducing hardware requirements, and enhancing usability:

Refactored the sorting module code to use layoutreader for reading order sorting, ensuring high accuracy in various layouts.
Refactored the paragraph concatenation module to achieve good results in cross-column, cross-page, cross-figure, and cross-table scenarios.
Refactored the list and table of contents recognition functions, significantly improving the accuracy of list blocks and table of contents blocks, as well as the parsing of corresponding text paragraphs.
Refactored the matching logic for figures, tables, and descriptive text, greatly enhancing the accuracy of matching captions and footnotes to figures and tables, and reducing the loss rate of descriptive text to near zero.
Added multi-language support for OCR, supporting detection and recognition of 84 languages. For the list of supported languages, see OCR Language Support List.
Added memory recycling logic and other memory optimization measures, significantly reducing memory usage. The memory requirement for enabling all acceleration features except table acceleration (layout/formula/OCR) has been reduced from 16GB to 8GB, and the memory requirement for enabling all acceleration features has been reduced from 24GB to 10GB.
Optimized configuration file feature switches, adding an independent formula detection switch to significantly improve speed and parsing results when formula detection is not needed.
Integrated PDF-Extract-Kit 1.0:
Added the self-developed doclayout_yolo model, which speeds up processing by more than 10 times compared to the original solution while maintaining similar parsing effects, and can be freely switched with layoutlmv3 via the configuration file.
Upgraded formula parsing to unimernet 0.2.1, improving formula parsing accuracy while significantly reducing memory usage.
Due to the repository change for PDF-Extract-Kit 1.0, you need to re-download the model. Please refer to How to Download Models for detailed steps.

"},{"location":"reference/changelog/#081-20240927","title":"0.8.1 (2024/09/27)","text":"

Fixed some bugs, and providing a localized deployment version of the online demo and the front-end interface.

"},{"location":"reference/changelog/#080-20240909","title":"0.8.0 (2024/09/09)","text":"

Supporting fast deployment with Dockerfile, and launching demos on Huggingface and Modelscope.

"},{"location":"reference/changelog/#071-20240830","title":"0.7.1 (2024/08/30)","text":"

Add paddle tablemaster table recognition option

"},{"location":"reference/changelog/#070b1-20240809","title":"0.7.0b1 (2024/08/09)","text":"

Simplified installation process, added table recognition functionality

"},{"location":"reference/changelog/#062b1-20240801","title":"0.6.2b1 (2024/08/01)","text":"

Optimized dependency conflict issues and installation documentation

"},{"location":"reference/changelog/#initial-open-source-release-20240705","title":"Initial Open-Source Release (2024/07/05)","text":"

MinerU project's first open-source release

"},{"location":"reference/output_files/","title":"Output File Format","text":""},{"location":"reference/output_files/#mineru-output-files-documentation","title":"MinerU Output Files Documentation","text":""},{"location":"reference/output_files/#overview","title":"Overview","text":"

Visual debugging files: Help users intuitively understand the document parsing process and results
Structured data files: Contain detailed parsing data for secondary development

The following sections provide detailed descriptions of each file's purpose and format.

"},{"location":"reference/output_files/#visual-debugging-files","title":"Visual Debugging Files","text":""},{"location":"reference/output_files/#layout-analysis-file-layoutpdf","title":"Layout Analysis File (layout.pdf)","text":"

File naming format: {original_filename}_layout.pdf

Functionality:

Visualizes layout analysis results for each page
Numbers in the top-right corner of each detection box indicate reading order
Different background colors distinguish different types of content blocks

Use cases:

Check if layout analysis is correct
Verify if reading order is reasonable
Debug layout-related issues

"},{"location":"reference/output_files/#text-spans-file-spanpdf","title":"Text Spans File (span.pdf)","text":"

Note

Only applicable to pipeline backend

File naming format: {original_filename}_span.pdf

Functionality:

Uses different colored line boxes to annotate page content based on span type
Used for quality inspection and issue troubleshooting

Use cases:

Quickly troubleshoot text loss issues
Check inline formula recognition
Verify text segmentation accuracy

"},{"location":"reference/output_files/#structured-data-files","title":"Structured Data Files","text":"

Important

"},{"location":"reference/output_files/#pipeline-backend-output-results","title":"Pipeline Backend Output Results","text":""},{"location":"reference/output_files/#model-inference-results-modeljson","title":"Model Inference Results (model.json)","text":"

File naming format: {original_filename}_model.json

"},{"location":"reference/output_files/#data-structure-definition","title":"Data Structure Definition","text":"

from pydantic import BaseModel, Field\nfrom enum import IntEnum\n\nclass CategoryType(IntEnum):\n    \"\"\"Content category enumeration\"\"\"\n    title = 0               # Title\n    plain_text = 1          # Text\n    abandon = 2             # Including headers, footers, page numbers, and page annotations\n    figure = 3              # Image\n    figure_caption = 4      # Image caption\n    table = 5               # Table\n    table_caption = 6       # Table caption\n    table_footnote = 7      # Table footnote\n    isolate_formula = 8     # Interline formula\n    formula_caption = 9     # Interline formula number\n    embedding = 13          # Inline formula\n    isolated = 14           # Interline formula\n    text = 15               # OCR recognition result\n\nclass PageInfo(BaseModel):\n    \"\"\"Page information\"\"\"\n    page_no: int = Field(description=\"Page number, first page is 0\", ge=0)\n    height: int = Field(description=\"Page height\", gt=0)\n    width: int = Field(description=\"Page width\", ge=0)\n\nclass ObjectInferenceResult(BaseModel):\n    \"\"\"Object recognition result\"\"\"\n    category_id: CategoryType = Field(description=\"Category\", ge=0)\n    poly: list[float] = Field(description=\"Quadrilateral coordinates, format: [x0,y0,x1,y1,x2,y2,x3,y3]\")\n    score: float = Field(description=\"Confidence score of inference result\")\n    latex: str | None = Field(description=\"LaTeX parsing result\", default=None)\n    html: str | None = Field(description=\"HTML parsing result\", default=None)\n\nclass PageInferenceResults(BaseModel):\n    \"\"\"Page inference results\"\"\"\n    layout_dets: list[ObjectInferenceResult] = Field(description=\"Page recognition results\")\n    page_info: PageInfo = Field(description=\"Page metadata\")\n\n# Complete inference results\ninference_result: list[PageInferenceResults] = []\n

"},{"location":"reference/output_files/#coordinate-system-description","title":"Coordinate System Description","text":"

poly coordinate format: [x0, y0, x1, y1, x2, y2, x3, y3]

Represents coordinates of top-left, top-right, bottom-right, bottom-left points respectively
Coordinate origin is at the top-left corner of the page

"},{"location":"reference/output_files/#sample-data","title":"Sample Data","text":"

[\n    {\n        \"layout_dets\": [\n            {\n                \"category_id\": 2,\n                \"poly\": [\n                    99.1906967163086,\n                    100.3119125366211,\n                    730.3707885742188,\n                    100.3119125366211,\n                    730.3707885742188,\n                    245.81326293945312,\n                    99.1906967163086,\n                    245.81326293945312\n                ],\n                \"score\": 0.9999997615814209\n            }\n        ],\n        \"page_info\": {\n            \"page_no\": 0,\n            \"height\": 2339,\n            \"width\": 1654\n        }\n    },\n    {\n        \"layout_dets\": [\n            {\n                \"category_id\": 5,\n                \"poly\": [\n                    99.13092803955078,\n                    2210.680419921875,\n                    497.3183898925781,\n                    2210.680419921875,\n                    497.3183898925781,\n                    2264.78076171875,\n                    99.13092803955078,\n                    2264.78076171875\n                ],\n                \"score\": 0.9999997019767761\n            }\n        ],\n        \"page_info\": {\n            \"page_no\": 1,\n            \"height\": 2339,\n            \"width\": 1654\n        }\n    }\n]\n

"},{"location":"reference/output_files/#intermediate-processing-results-middlejson","title":"Intermediate Processing Results (middle.json)","text":"

File naming format: {original_filename}_middle.json

"},{"location":"reference/output_files/#top-level-structure","title":"Top-level Structure","text":"Field Name Type Description pdf_info list[dict] Array of parsing results for each page _backend string Parsing mode: pipeline or vlm _version_name string MinerU version number"},{"location":"reference/output_files/#page-information-structure-pdf_info","title":"Page Information Structure (pdf_info)","text":"Field Name Description preproc_blocks Unsegmented intermediate results after PDF preprocessing page_idx Page number, starting from 0 page_size Page width and height [width, height] images Image block information list tables Table block information list interline_equations Interline formula block information list discarded_blocks Block information to be discarded para_blocks Content block results after segmentation"},{"location":"reference/output_files/#block-structure-hierarchy","title":"Block Structure Hierarchy","text":"

Level 1 blocks (table | image)\n\u2514\u2500\u2500 Level 2 blocks\n    \u2514\u2500\u2500 Lines\n        \u2514\u2500\u2500 Spans\n

"},{"location":"reference/output_files/#level-1-block-fields","title":"Level 1 Block Fields","text":"Field Name Description type Block type: table or image bbox Rectangular box coordinates of the block [x0, y0, x1, y1] blocks List of contained level 2 blocks"},{"location":"reference/output_files/#level-2-block-fields","title":"Level 2 Block Fields","text":"Field Name Description type Block type (see table below) bbox Rectangular box coordinates of the block lines List of contained line information"},{"location":"reference/output_files/#level-2-block-types","title":"Level 2 Block Types","text":"Type Description image_body Image body image_caption Image caption text image_footnote Image footnote table_body Table body table_caption Table caption text table_footnote Table footnote text Text block title Title block index Index block list List block interline_equation Interline formula block"},{"location":"reference/output_files/#line-and-span-structure","title":"Line and Span Structure","text":"

Line fields: - bbox: Rectangular box coordinates of the line - spans: List of contained spans

Span fields: - bbox: Rectangular box coordinates of the span - type: Span type (image, table, text, inline_equation, interline_equation) - content | img_path: Text content or image path

"},{"location":"reference/output_files/#sample-data_1","title":"Sample Data","text":"

{\n    \"pdf_info\": [\n        {\n            \"preproc_blocks\": [\n                {\n                    \"type\": \"text\",\n                    \"bbox\": [\n                        52,\n                        61.956024169921875,\n                        294,\n                        82.99800872802734\n                    ],\n                    \"lines\": [\n                        {\n                            \"bbox\": [\n                                52,\n                                61.956024169921875,\n                                294,\n                                72.0000228881836\n                            ],\n                            \"spans\": [\n                                {\n                                    \"bbox\": [\n                                        54.0,\n                                        61.956024169921875,\n                                        296.2261657714844,\n                                        72.0000228881836\n                                    ],\n                                    \"content\": \"dependent on the service headway and the reliability of the departure \",\n                                    \"type\": \"text\",\n                                    \"score\": 1.0\n                                }\n                            ]\n                        }\n                    ]\n                }\n            ],\n            \"layout_bboxes\": [\n                {\n                    \"layout_bbox\": [\n                        52,\n                        61,\n                        294,\n                        731\n                    ],\n                    \"layout_label\": \"V\",\n                    \"sub_layout\": []\n                }\n            ],\n            \"page_idx\": 0,\n            \"page_size\": [\n                612.0,\n                792.0\n            ],\n            \"_layout_tree\": [],\n            \"images\": [],\n            \"tables\": [],\n            \"interline_equations\": [],\n            \"discarded_blocks\": [],\n            \"para_blocks\": [\n                {\n                    \"type\": \"text\",\n                    \"bbox\": [\n                        52,\n                        61.956024169921875,\n                        294,\n                        82.99800872802734\n                    ],\n                    \"lines\": [\n                        {\n                            \"bbox\": [\n                                52,\n                                61.956024169921875,\n                                294,\n                                72.0000228881836\n                            ],\n                            \"spans\": [\n                                {\n                                    \"bbox\": [\n                                        54.0,\n                                        61.956024169921875,\n                                        296.2261657714844,\n                                        72.0000228881836\n                                    ],\n                                    \"content\": \"dependent on the service headway and the reliability of the departure \",\n                                    \"type\": \"text\",\n                                    \"score\": 1.0\n                                }\n                            ]\n                        }\n                    ]\n                }\n            ]\n        }\n    ],\n    \"_backend\": \"pipeline\",\n    \"_version_name\": \"0.6.1\"\n}\n

"},{"location":"reference/output_files/#content-list-content_listjson","title":"Content List (content_list.json)","text":"

File naming format: {original_filename}_content_list.json

"},{"location":"reference/output_files/#functionality","title":"Functionality","text":"

This is a simplified version of middle.json that stores all readable content blocks in reading order as a flat structure, removing complex layout information for easier subsequent processing.

"},{"location":"reference/output_files/#content-types","title":"Content Types","text":"Type Description image Image table Table text Text/Title equation Interline formula"},{"location":"reference/output_files/#text-level-identification","title":"Text Level Identification","text":"

Text levels are distinguished through the text_level field:

No text_level or text_level: 0: Body text
text_level: 1: Level 1 heading
text_level: 2: Level 2 heading
And so on...

"},{"location":"reference/output_files/#common-fields","title":"Common Fields","text":"

All content blocks include a page_idx field indicating the page number (starting from 0).
All content blocks include a bbox field representing the bounding box coordinates of the content block [x0, y0, x1, y1], mapped to a range of 0-1000.

"},{"location":"reference/output_files/#sample-data_2","title":"Sample Data","text":"

[\n        {\n        \"type\": \"text\",\n        \"text\": \"The response of flow duration curves to afforestation \",\n        \"text_level\": 1, \n        \"bbox\": [\n            62,\n            480,\n            946,\n            904\n        ],\n        \"page_idx\": 0\n    },\n    {\n        \"type\": \"image\",\n        \"img_path\": \"images/a8ecda1c69b27e4f79fce1589175a9d721cbdc1cf78b4cc06a015f3746f6b9d8.jpg\",\n        \"image_caption\": [\n            \"Fig. 1. Annual flow duration curves of daily flows from Pine Creek, Australia, 1989\u20132000. \"\n        ],\n        \"image_footnote\": [],\n        \"bbox\": [\n            62,\n            480,\n            946,\n            904\n        ],\n        \"page_idx\": 1\n    },\n    {\n        \"type\": \"equation\",\n        \"img_path\": \"images/181ea56ef185060d04bf4e274685f3e072e922e7b839f093d482c29bf89b71e8.jpg\",\n        \"text\": \"$$\\nQ _ { \\\\% } = f ( P ) + g ( T )\\n$$\",\n        \"text_format\": \"latex\",\n        \"bbox\": [\n            62,\n            480,\n            946,\n            904\n        ],\n        \"page_idx\": 2\n    },\n    {\n        \"type\": \"table\",\n        \"img_path\": \"images/e3cb413394a475e555807ffdad913435940ec637873d673ee1b039e3bc3496d0.jpg\",\n        \"table_caption\": [\n            \"Table 2 Significance of the rainfall and time terms \"\n        ],\n        \"table_footnote\": [\n            \"indicates that the rainfall term was significant at the $5 \\\\%$ level, $T$ indicates that the time term was significant at the $5 \\\\%$ level, \\\\* represents significance at the $10 \\\\%$ level, and na denotes too few data points for meaningful analysis. \"\n        ],\n        \"table_body\": \"<html><body><table><tr><td rowspan=\\\"2\\\">Site</td><td colspan=\\\"10\\\">Percentile</td></tr><tr><td>10</td><td>20</td><td>30</td><td>40</td><td>50</td><td>60</td><td>70</td><td>80</td><td>90</td><td>100</td></tr><tr><td>Traralgon Ck</td><td>P</td><td>P,*</td><td>P</td><td>P</td><td>P,</td><td>P,</td><td>P,</td><td>P,</td><td>P</td><td>P</td></tr><tr><td>Redhill</td><td>P,T</td><td>P,T</td><td>\uff0c*</td><td>**</td><td>P.T</td><td>P,*</td><td>P*</td><td>P*</td><td>*</td><td>\uff0c*</td></tr><tr><td>Pine Ck</td><td></td><td>P,T</td><td>P,T</td><td>P,T</td><td>P,T</td><td>T</td><td>T</td><td>T</td><td>na</td><td>na</td></tr><tr><td>Stewarts Ck 5</td><td>P,T</td><td>P,T</td><td>P,T</td><td>P,T</td><td>P.T</td><td>P.T</td><td>P,T</td><td>na</td><td>na</td><td>na</td></tr><tr><td>Glendhu 2</td><td>P</td><td>P,T</td><td>P,*</td><td>P,T</td><td>P.T</td><td>P,ns</td><td>P,T</td><td>P,T</td><td>P,T</td><td>P,T</td></tr><tr><td>Cathedral Peak 2</td><td>P,T</td><td>P,T</td><td>P,T</td><td>P,T</td><td>P,T</td><td>*,T</td><td>P,T</td><td>P,T</td><td>P,T</td><td>T</td></tr><tr><td>Cathedral Peak 3</td><td>P.T</td><td>P.T</td><td>P,T</td><td>P,T</td><td>P,T</td><td>T</td><td>P,T</td><td>P,T</td><td>P,T</td><td>T</td></tr><tr><td>Lambrechtsbos A</td><td>P,T</td><td>P</td><td>P</td><td>P,T</td><td>*,T</td><td>*,T</td><td>*,T</td><td>*,T</td><td>*,T</td><td>T</td></tr><tr><td>Lambrechtsbos B</td><td>P,T</td><td>P,T</td><td>P,T</td><td>P,T</td><td>P,T</td><td>P,T</td><td>P,T</td><td>P,T</td><td>T</td><td>T</td></tr><tr><td>Biesievlei</td><td>P,T</td><td>P.T</td><td>P,T</td><td>P,T</td><td>*,T</td><td>*,T</td><td>T</td><td>T</td><td>P,T</td><td>P,T</td></tr></table></body></html>\",\n        \"bbox\": [\n            62,\n            480,\n            946,\n            904\n        ],  \n        \"page_idx\": 5\n    }\n]\n

"},{"location":"reference/output_files/#vlm-backend-output-results","title":"VLM Backend Output Results","text":""},{"location":"reference/output_files/#model-inference-results-modeljson_1","title":"Model Inference Results (model.json)","text":"

File naming format: {original_filename}_model.json

"},{"location":"reference/output_files/#file-format-description","title":"File format description","text":"

Two-level nested list: outer list = pages; inner list = content blocks of that page
Each block is a dict with at least: type, bbox, angle, content (some types add extra fields like score, block_tags, content_tags, format)
Designed for direct, raw model inspection

"},{"location":"reference/output_files/#supported-content-types-type-field-values","title":"Supported content types (type field values)","text":"

{\n  \"text\": \"Plain text\",\n  \"title\": \"Title\",\n  \"equation\": \"Display (interline) formula\",\n  \"image\": \"Image\",\n  \"image_caption\": \"Image caption\",\n  \"image_footnote\": \"Image footnote\",\n  \"table\": \"Table\",\n  \"table_caption\": \"Table caption\",\n  \"table_footnote\": \"Table footnote\",\n  \"phonetic\": \"Phonetic annotation\",\n  \"code\": \"Code block\",\n  \"code_caption\": \"Code caption\",\n  \"ref_text\": \"Reference / citation entry\",\n  \"algorithm\": \"Algorithm block (treated as code subtype)\",\n  \"list\": \"List container\",\n  \"header\": \"Page header\",\n  \"footer\": \"Page footer\",\n  \"page_number\": \"Page number\",\n  \"aside_text\": \"Side / margin note\",\n  \"page_footnote\": \"Page footnote\"\n}\n

"},{"location":"reference/output_files/#coordinate-system","title":"Coordinate system","text":"

bbox = [x0, y0, x1, y1] (top-left, bottom-right)
Origin at top-left of the page
All coordinates are normalized percentages in [0,1]

"},{"location":"reference/output_files/#sample-data_3","title":"Sample data","text":"

[\n  [\n    {\n      \"type\": \"header\",\n      \"bbox\": [0.077, 0.095, 0.18, 0.181],\n      \"angle\": 0,\n      \"score\": null,\n      \"block_tags\": null,\n      \"content\": \"ELSEVIER\",\n      \"format\": null,\n      \"content_tags\": null\n    },\n    {\n      \"type\": \"title\",\n      \"bbox\": [0.157, 0.228, 0.833, 0.253],\n      \"angle\": 0,\n      \"score\": null,\n      \"block_tags\": null,\n      \"content\": \"The response of flow duration curves to afforestation\",\n      \"format\": null,\n      \"content_tags\": null\n    }\n  ]\n]\n

"},{"location":"reference/output_files/#intermediate-processing-results-middlejson_1","title":"Intermediate Processing Results (middle.json)","text":"

File naming format: {original_filename}_middle.json

Structure is broadly similar to the pipeline backend, but with these differences:

list becomes a second\u2011level block, a new field sub_type distinguishes list categories:
- text: ordinary list
- ref_text: reference / bibliography style list
New code block type with sub_type(a code block always has at least a code_body, it may optionally have a code_caption):
- code
- algorithm
discarded_blocks may contain additional types:
- header
- footer
- page_number
- aside_text
- page_footnote
All blocks include an angle field indicating rotation (one of 0, 90, 180, 270).

"},{"location":"reference/output_files/#examples","title":"Examples","text":"

Example: list block

{\n  \"bbox\": [174,155,818,333],\n  \"type\": \"list\",\n  \"angle\": 0,\n  \"index\": 11,\n  \"blocks\": [\n    {\n      \"bbox\": [174,157,311,175],\n      \"type\": \"text\",\n      \"angle\": 0,\n      \"lines\": [\n        {\n          \"bbox\": [174,157,311,175],\n            \"spans\": [\n              {\n                \"bbox\": [174,157,311,175],\n                \"type\": \"text\",\n                \"content\": \"H.1 Introduction\"\n              }\n            ]\n        }\n      ],\n      \"index\": 3\n    },\n    {\n      \"bbox\": [175,182,464,229],\n      \"type\": \"text\",\n      \"angle\": 0,\n      \"lines\": [\n        {\n          \"bbox\": [175,182,464,229],\n          \"spans\": [\n            {\n              \"bbox\": [175,182,464,229],\n              \"type\": \"text\",\n              \"content\": \"H.2 Example: Divide by Zero without Exception Handling\"\n            }\n          ]\n        }\n      ],\n      \"index\": 4\n    }\n  ],\n  \"sub_type\": \"text\"\n}\n

Example: code block with optional caption:

{\n  \"type\": \"code\",\n  \"bbox\": [114,780,885,1231],\n  \"blocks\": [\n    {\n      \"bbox\": [114,780,885,1231],\n      \"lines\": [\n        {\n          \"bbox\": [114,780,885,1231],\n          \"spans\": [\n            {\n              \"bbox\": [114,780,885,1231],\n              \"type\": \"text\",\n              \"content\": \"1 // Fig. H.1: DivideByZeroNoExceptionHandling.java  \\n2 // Integer division without exception handling.  \\n3 import java.util.Scanner;  \\n4  \\n5 public class DivideByZeroNoExceptionHandling  \\n6 {  \\n7 // demonstrates throwing an exception when a divide-by-zero occurs  \\n8 public static int quotient( int numerator, int denominator )  \\n9 {  \\n10 return numerator / denominator; // possible division by zero  \\n11 } // end method quotient  \\n12  \\n13 public static void main(String[] args)  \\n14 {  \\n15 Scanner scanner = new Scanner(System.in); // scanner for input  \\n16  \\n17 System.out.print(\\\"Please enter an integer numerator: \\\");  \\n18 int numerator = scanner.nextInt();  \\n19 System.out.print(\\\"Please enter an integer denominator: \\\");  \\n20 int denominator = scanner.nextInt();  \\n21\"\n            }\n          ]\n        }\n      ],\n      \"index\": 17,\n      \"angle\": 0,\n      \"type\": \"code_body\"\n    },\n    {\n      \"bbox\": [867,160,1280,189],\n      \"lines\": [\n        {\n          \"bbox\": [867,160,1280,189],\n          \"spans\": [\n            {\n              \"bbox\": [867,160,1280,189],\n              \"type\": \"text\",\n              \"content\": \"Algorithm 1 Modules for MCTSteg\"\n            }\n          ]\n        }\n      ],\n      \"index\": 19,\n      \"angle\": 0,\n      \"type\": \"code_caption\"\n    }\n  ],\n  \"index\": 17,\n  \"sub_type\": \"code\"\n}\n

"},{"location":"reference/output_files/#content-list-content_listjson_1","title":"Content List (content_list.json)","text":"

File naming format: {original_filename}_content_list.json

Based on the pipeline format, with these VLM-specific extensions:

New code type with sub_type (code | algorithm):
- Fields: code_body (string), optional code_caption (list of strings)
New list type with sub_type (text | ref_text):
- Field: list_items (array of strings)
All discarded_blocks entries are also output (e.g., headers, footers, page numbers, margin notes, page footnotes).
Existing types (image, table, text, equation) remain unchanged.
bbox still uses the 0\u20131000 normalized coordinate mapping.

"},{"location":"reference/output_files/#examples_1","title":"Examples","text":"

Example: code (algorithm) entry

{\n  \"type\": \"code\",\n  \"sub_type\": \"algorithm\",\n  \"code_caption\": [\"Algorithm 1 Modules for MCTSteg\"],\n  \"code_body\": \"1: function GETCOORDINATE(d)  \\n2:  $x \\\\gets d / l$ ,  $y \\\\gets d$  mod  $l$   \\n3: return  $(x, y)$   \\n4: end function  \\n5: function BESTCHILD(v)  \\n6:  $C \\\\gets$  child set of  $v$   \\n7:  $v' \\\\gets \\\\arg \\\\max_{c \\\\in C} \\\\mathrm{UCTScore}(c)$   \\n8:  $v'.n \\\\gets v'.n + 1$   \\n9: return  $v'$   \\n10: end function  \\n11: function BACK PROPAGATE(v)  \\n12: Calculate  $R$  using Equation 11  \\n13: while  $v$  is not a root node do  \\n14:  $v.r \\\\gets v.r + R$ ,  $v \\\\gets v.p$   \\n15: end while  \\n16: end function  \\n17: function RANDOMSEARCH(v)  \\n18: while  $v$  is not a leaf node do  \\n19: Randomly select an untried action  $a \\\\in A(v)$   \\n20: Create a new node  $v'$   \\n21:  $(x, y) \\\\gets \\\\mathrm{GETCOORDINATE}(v'.d)$   \\n22:  $v'.p \\\\gets v$ ,  $v'.d \\\\gets v.d + 1$ ,  $v'.\\\\Gamma \\\\gets v.\\\\Gamma$   \\n23:  $v'.\\\\gamma_{x,y} \\\\gets a$   \\n24: if  $a = -1$  then  \\n25:  $v.lc \\\\gets v'$   \\n26: else if  $a = 0$  then  \\n27:  $v.mc \\\\gets v'$   \\n28: else  \\n29:  $v.rc \\\\gets v'$   \\n30: end if  \\n31:  $v \\\\gets v'$   \\n32: end while  \\n33: return  $v$   \\n34: end function  \\n35: function SEARCH(v)  \\n36: while  $v$  is fully expanded do  \\n37:  $v \\\\gets$  BESTCHILD(v)  \\n38: end while  \\n39: if  $v$  is not a leaf node then  \\n40:  $v \\\\gets$  RANDOMSEARCH(v)  \\n41: end if  \\n42: return  $v$   \\n43: end function\",\n  \"bbox\": [510,87,881,740],\n  \"page_idx\": 0\n}\n

Example: list (text) entry

{\n  \"type\": \"list\",\n  \"sub_type\": \"text\",\n  \"list_items\": [\n    \"H.1 Introduction\",\n    \"H.2 Example: Divide by Zero without Exception Handling\",\n    \"H.3 Example: Divide by Zero with Exception Handling\",\n    \"H.4 Summary\"\n  ],\n  \"bbox\": [174,155,818,333],\n  \"page_idx\": 0\n}\n

Example: discarded blocks output

[\n  {\n    \"type\": \"header\",\n    \"text\": \"Journal of Hydrology 310 (2005) 253-265\",\n    \"bbox\": [363,164,623,177],\n    \"page_idx\": 0\n  },\n  {\n    \"type\": \"page_footnote\",\n    \"text\": \"* Corresponding author. Address: Forest Science Centre, Department of Sustainability and Environment, P.O. Box 137, Heidelberg, Vic. 3084, Australia. Tel.: +61 3 9450 8719; fax: +61 3 9450 8644.\",\n    \"bbox\": [71,815,915,841],\n    \"page_idx\": 0\n  }\n]\n

"},{"location":"reference/output_files/#summary","title":"Summary","text":"

The above files constitute MinerU's complete output results. Users can choose appropriate files for subsequent processing based on their needs:

Model outputs (Use raw outputs):
- model.json
Debugging and verification (Use visualization files):
- layout.pdf
- span.pdf
Content extraction: (Use simplified files):
- *.md
- content_list.json
Secondary development: (Use structured files):
- middle.json

"},{"location":"usage/","title":"Usage","text":""},{"location":"usage/#usage-guide","title":"Usage Guide","text":"

This section provides comprehensive usage instructions for the project. We will help you progressively master the project's usage from basic to advanced through the following sections:

"},{"location":"usage/#table-of-contents","title":"Table of Contents","text":"

Quick Usage - Quick setup and basic usage
Model Source Configuration - Detailed configuration instructions for model sources
Command Line Tools - Detailed parameter descriptions for command line tools
Advanced Optimization Parameters - Advanced parameter descriptions for command line tool adaptation

"},{"location":"usage/#getting-started","title":"Getting Started","text":"

We recommend reading the documentation in the order listed above, which will help you better understand and use the project features.

If you encounter issues during usage, please check the FAQ

"},{"location":"usage/advanced_cli_parameters/","title":"Advanced CLI Parameters","text":""},{"location":"usage/advanced_cli_parameters/#advanced-command-line-parameters","title":"Advanced Command Line Parameters","text":""},{"location":"usage/advanced_cli_parameters/#pass-through-of-inference-engine-parameters","title":"Pass-through of inference engine parameters","text":""},{"location":"usage/advanced_cli_parameters/#vllm-acceleration-parameter-optimization","title":"vllm Acceleration Parameter Optimization","text":"

Tip

If you can already use vllm normally for accelerated VLM model inference but still want to further improve inference speed, you can try the following parameters:

If you have multiple graphics cards, you can use vllm's multi-card parallel mode to increase throughput: --data-parallel-size 2

"},{"location":"usage/advanced_cli_parameters/#parameter-passing-instructions","title":"Parameter Passing Instructions","text":"

Tip

All officially supported vllm/lmdeploy parameters can be passed to MinerU through command line arguments, including the following commands: mineru, mineru-openai-server, mineru-gradio, mineru-api
If you want to learn more about vllm parameter usage, please refer to the vllm official documentation
If you want to learn more about lmdeploy parameter usage, please refer to the lmdeploy official documentation

"},{"location":"usage/advanced_cli_parameters/#gpu-device-selection-and-configuration","title":"GPU Device Selection and Configuration","text":""},{"location":"usage/advanced_cli_parameters/#cuda_visible_devices-basic-usage","title":"CUDA_VISIBLE_DEVICES Basic Usage","text":"

Tip

In any situation, you can specify visible GPU devices by adding the CUDA_VISIBLE_DEVICES environment variable at the beginning of the command line. For example:
```
CUDA_VISIBLE_DEVICES=1 mineru -p <input_path> -o <output_path>\n
```
This specification method is effective for all command line calls, including mineru, mineru-openai-server, mineru-gradio, and mineru-api, and applies to both pipeline and vlm backends.

"},{"location":"usage/advanced_cli_parameters/#common-device-configuration-examples","title":"Common Device Configuration Examples","text":"

Tip

Here are some common CUDA_VISIBLE_DEVICES setting examples:

CUDA_VISIBLE_DEVICES=1  # Only device 1 will be seen\nCUDA_VISIBLE_DEVICES=0,1  # Devices 0 and 1 will be visible\nCUDA_VISIBLE_DEVICES=\"0,1\"  # Same as above, quotation marks are optional\nCUDA_VISIBLE_DEVICES=0,2,3  # Devices 0, 2, 3 will be visible; device 1 is masked\nCUDA_VISIBLE_DEVICES=\"\"  # No GPU will be visible\n

"},{"location":"usage/advanced_cli_parameters/#practical-application-scenarios","title":"Practical Application Scenarios","text":"

Tip

Here are some possible usage scenarios:

If you have multiple graphics cards and need to specify cards 0 and 1, using multi-card parallelism to start openai-server, you can use the following command:
```
CUDA_VISIBLE_DEVICES=0,1 mineru-openai-server --engine vllm --port 30000 --data-parallel-size 2\n
```
If you have multiple graphics cards and need to start two fastapi services on cards 0 and 1, listening on different ports respectively, you can use the following commands:
```
# In terminal 1\nCUDA_VISIBLE_DEVICES=0 mineru-api --host 127.0.0.1 --port 8000\n# In terminal 2\nCUDA_VISIBLE_DEVICES=1 mineru-api --host 127.0.0.1 --port 8001\n
```

"},{"location":"usage/cli_tools/","title":"CLI Tools","text":""},{"location":"usage/cli_tools/#command-line-tools-usage-instructions","title":"Command Line Tools Usage Instructions","text":""},{"location":"usage/cli_tools/#view-help-information","title":"View Help Information","text":"

To view help information for MinerU command line tools, you can use the --help parameter. Here are help information examples for various command line tools:

mineru --help\nUsage: mineru [OPTIONS]\n\nOptions:\n  -v, --version                   Show version and exit\n  -p, --path PATH                 Input file path or directory (required)\n  -o, --output PATH               Output directory (required)\n  --api-url TEXT                  MinerU FastAPI base URL; if omitted, `mineru` starts a temporary local `mineru-api`\n  -m, --method [auto|txt|ocr]     Parsing method: auto (default), txt, ocr (pipeline and hybrid* backend only)\n  -b, --backend [pipeline|hybrid-auto-engine|hybrid-http-client|vlm-auto-engine|vlm-http-client]\n                                  Parsing backend (default: hybrid-auto-engine)\n  -l, --lang [ch|ch_server|ch_lite|en|korean|japan|chinese_cht|ta|te|ka|th|el|latin|arabic|east_slavic|cyrillic|devanagari]\n                                  Specify document language (improves OCR accuracy, pipeline and hybrid* backend only)\n  -u, --url TEXT                  OpenAI-compatible backend URL passed through to the server when using http-client\n  -s, --start INTEGER             Starting page number for parsing (0-based)\n  -e, --end INTEGER               Ending page number for parsing (0-based)\n  -f, --formula BOOLEAN           Enable formula parsing (default: enabled)\n  -t, --table BOOLEAN             Enable table parsing (default: enabled)\n  --help                          Show help information\n

mineru-api --help\nUsage: mineru-api [OPTIONS]\n\nOptions:\n  --host TEXT     Server host (default: 127.0.0.1)\n  --port INTEGER  Server port (default: 8000)\n  --reload        Enable auto-reload (development mode)\n  --help          Show this message and exit.\n

mineru-gradio --help\nUsage: mineru-gradio [OPTIONS]\n\nOptions:\n  --enable-example BOOLEAN        Enable example files for input. The example\n                                  files to be input need to be placed in the\n                                  `example` folder within the directory where\n                                  the command is currently executed.\n  --enable-http-client BOOLEAN    Enable http-client backend to link openai-\n                                  compatible servers.\n  --enable-api BOOLEAN            Enable gradio API for serving the\n                                  application.\n  --max-convert-pages INTEGER     Set the maximum number of pages to convert\n                                  from PDF to Markdown.\n  --server-name TEXT              Set the server name for the Gradio app.\n  --server-port INTEGER           Set the server port for the Gradio app.\n  --latex-delimiters-type [a|b|all]\n                                  Set the type of LaTeX delimiters to use in\n                                  Markdown rendering: 'a' for type '$', 'b' for\n                                  type '()[]', 'all' for both types.\n  --help                          Show this message and exit.\n

"},{"location":"usage/cli_tools/#environment-variables-description","title":"Environment Variables Description","text":"

Note

Starting from this version, mineru is an orchestration client built on top of mineru-api: - Without --api-url, the CLI launches a temporary local mineru-api - With --api-url, the CLI connects to that FastAPI service directly - --url is no longer the MinerU API address; it is the OpenAI-compatible backend URL used by server-side vlm/hybrid-http-client

Some parameters of MinerU command line tools have equivalent environment variable configurations. Generally, environment variable configurations have higher priority than command line parameters and take effect across all command line tools. Here are the environment variables and their descriptions:

MINERU_TOOLS_CONFIG_JSON:
- Used to specify configuration file path
- defaults to mineru.json in user directory, can specify other configuration file paths through environment variables.
MINERU_FORMULA_ENABLE:
- Used to enable formula parsing
- defaults to true, can be set to false through environment variables to disable formula parsing.
MINERU_FORMULA_CH_SUPPORT:
- Used to enable Chinese formula parsing optimization (experimental feature)
- Default is false, can be set to true via environment variable to enable Chinese formula parsing optimization.
- Only effective for pipeline backend.
MINERU_TABLE_ENABLE:
- Used to enable table parsing
- Default is true, can be set to false via environment variable to disable table parsing.
MINERU_TABLE_MERGE_ENABLE:
- Used to enable table merging functionality
- Default is true, can be set to false via environment variable to disable table merging functionality.
MINERU_PDF_RENDER_TIMEOUT:
- Used to set the timeout (in seconds) for rendering PDFs to images.
- Default is 300 seconds; you can set a different value via an environment variable to adjust the rendering timeout.
- Only effective on Linux and macOS systems.
MINERU_PDF_RENDER_THREADS:
- Used to set the number of threads used when rendering PDFs to images.
- Default is 4; you can set a different value via an environment variable to adjust the number of threads for image rendering.
- Only effective on Linux and macOS systems.
MINERU_INTRA_OP_NUM_THREADS:
- Used to set the intra_op thread count for ONNX models, affects the computation speed of individual operators
- Default is -1 (auto-select), can be set to other values via environment variable to adjust the thread count.
MINERU_INTER_OP_NUM_THREADS:
- Used to set the inter_op thread count for ONNX models, affects the parallel execution of multiple operators
- Default is -1 (auto-select), can be set to other values via environment variable to adjust the thread count.
MINERU_HYBRID_BATCH_RATIO:
- Used to set the batch ratio for small model processing in hybrid-* backends.
- Commonly used in hybrid-http-client, it allows adjusting the VRAM usage of a single client by controlling the batch ratio of small models.
- Single Client VRAM Size MINERU_HYBRID_BATCH_RATIO <= 6 GB 8 <= 4.5 GB 4 <= 3 GB 2 <= 2.5 GB 1
MINERU_HYBRID_FORCE_PIPELINE_ENABLE:
- Used to force the text extraction part in hybrid-* backends to be processed using small models.
- Defaults to false. Can be set to true via environment variable to enable this feature, thereby reducing hallucinations in certain extreme cases.
MINERU_VL_MODEL_NAME:
- Used to specify the model name for the vlm/hybrid backend, allowing you to designate the model required for MinerU to run when multiple models exist on a remote openai-server.
MINERU_VL_API_KEY:
- Used to specify the API Key for the vlm/hybrid backend, enabling authentication on the remote openai-server.

"},{"location":"usage/model_source/","title":"Model Source","text":""},{"location":"usage/model_source/#model-source-documentation","title":"Model Source Documentation","text":"

MinerU uses HuggingFace and ModelScope as model repositories. Users can switch model sources or use local models as needed.

HuggingFace is the default model source, providing excellent loading speed and high stability globally.
ModelScope is the best choice for users in mainland China, providing seamlessly compatible hf SDK modules, suitable for users who cannot access HuggingFace.

"},{"location":"usage/model_source/#methods-to-switch-model-sources","title":"Methods to Switch Model Sources","text":""},{"location":"usage/model_source/#switch-via-command-line-parameters","title":"Switch via Command Line Parameters","text":"

Currently, only the mineru command line tool supports switching model sources through command line parameters. Other command line tools such as mineru-api, mineru-gradio, etc., do not support this yet.

mineru -p <input_path> -o <output_path> --source modelscope\n

"},{"location":"usage/model_source/#switch-via-environment-variables","title":"Switch via Environment Variables","text":"

You can switch model sources by setting environment variables in any situation. This applies to all command line tools and API calls.

export MINERU_MODEL_SOURCE=modelscope\n

import os\nos.environ[\"MINERU_MODEL_SOURCE\"] = \"modelscope\"\n

Tip

Model sources set through environment variables will take effect in the current terminal session until the terminal is closed or the environment variable is modified. They have higher priority than command line parameters - if both command line parameters and environment variables are set, the command line parameters will be ignored.

"},{"location":"usage/model_source/#using-local-models","title":"Using Local Models","text":""},{"location":"usage/model_source/#1-download-models-to-local-storage","title":"1. Download Models to Local Storage","text":"

mineru-models-download --help\n

or use the interactive command line tool to select model downloads:

mineru-models-download\n

Note

After download completion, the model path will be output in the current terminal window and automatically written to mineru.json in the user directory.
You can also create it by copying the configuration template file to your user directory and renaming it to mineru.json.
After downloading models locally, you can freely move the model folder to other locations while updating the model path in mineru.json.
If you deploy the model folder to another server, please ensure you move the mineru.json file to the user directory of the new device and configure the model path correctly.
If you need to update model files, you can run the mineru-models-download command again. Model updates do not support custom paths currently - if you haven't moved the local model folder, model files will be incrementally updated; if you have moved the model folder, model files will be re-downloaded to the default location and mineru.json will be updated.

"},{"location":"usage/model_source/#2-use-local-models-for-parsing","title":"2. Use Local Models for Parsing","text":"

mineru -p <input_path> -o <output_path> --source local\n

or enable through environment variables:

export MINERU_MODEL_SOURCE=local\nmineru -p <input_path> -o <output_path>\n

"},{"location":"usage/quick_usage/","title":"Quick Usage","text":""},{"location":"usage/quick_usage/#using-mineru","title":"Using MinerU","text":""},{"location":"usage/quick_usage/#quick-model-source-configuration","title":"Quick Model Source Configuration","text":"

MinerU uses huggingface as the default model source. If users cannot access huggingface due to network restrictions, they can conveniently switch the model source to modelscope through environment variables:

export MINERU_MODEL_SOURCE=modelscope\n

For more information about model source configuration and custom local model paths, please refer to the Model Source Documentation in the documentation."},{"location":"usage/quick_usage/#quick-usage-via-command-line","title":"Quick Usage via Command Line","text":"

MinerU has built-in command line tools that allow users to quickly use MinerU for PDF parsing through the command line:

mineru -p <input_path> -o <output_path>\n

Tip

<input_path>: Local PDF/image file or directory
<output_path>: Output directory
Without --api-url, the CLI launches a temporary local mineru-api
With --api-url, the CLI connects to an existing local or remote FastAPI service directly

For more information about output files, please refer to Output File Documentation.

Note

The command line tool will automatically attempt cuda/mps acceleration on Linux and macOS systems. Windows users who need cuda acceleration should visit the PyTorch official website to select the appropriate command for their cuda version to install acceleration-enabled torch and torchvision.

If you need to adjust parsing options through custom parameters, you can also check the more detailed Command Line Tools Usage Instructions in the documentation.

"},{"location":"usage/quick_usage/#advanced-usage-via-api-webui-http-clientserver","title":"Advanced Usage via API, WebUI, http-client/server","text":"

Direct Python API calls: Python Usage Example
FastAPI calls:
```
mineru-api --host 0.0.0.0 --port 8000\n
```
Tip

Access http://127.0.0.1:8000/docs in your browser to view the API documentation.
- Health endpoint: GET /health Returns protocol_version, processing_window_size, max_concurrent_requests, and task stats
- Asynchronous task submission endpoint: POST /tasks
- Synchronous parsing endpoint: POST /file_parse
- Task query endpoints: GET /tasks/{task_id}, GET /tasks/{task_id}/result
- API outputs are controlled by the server and written to ./output by default
POST /tasks returns immediately with a task_id. POST /file_parse uses the same task manager internally, waits for the task to finish, and then returns the final result synchronously. Tasks are tracked only in-process for a single mineru-api instance. Task status is not preserved across service restarts, --reload, or multi-process deployments. Completed or failed tasks are retained for 24 hours by default, then their task state and output directory are cleaned automatically. After cleanup, task status and result endpoints return 404. Use MINERU_API_TASK_RETENTION_SECONDS and MINERU_API_TASK_CLEANUP_INTERVAL_SECONDS to adjust retention and cleanup polling intervals.

Asynchronous task submission example:
```
curl -X POST http://127.0.0.1:8000/tasks \\\n  -F \"files=@demo/pdfs/demo1.pdf\" \\\n  -F \"return_md=true\"\n
```
Synchronous parsing example:
```
curl -X POST http://127.0.0.1:8000/file_parse \\\n  -F \"files=@demo/pdfs/demo1.pdf\" \\\n  -F \"return_md=true\" \\\n  -F \"response_format_zip=true\" \\\n  -F \"return_original_file=true\"\n
```
Poll task status and fetch results:
```
curl http://127.0.0.1:8000/tasks/<task_id>\ncurl http://127.0.0.1:8000/tasks/<task_id>/result\ncurl http://127.0.0.1:8000/health\n
```
Start Gradio WebUI visual frontend:
```
mineru-gradio --server-name 0.0.0.0 --server-port 7860\n
```
Tip
- Access http://127.0.0.1:7860 in your browser to use the Gradio WebUI.

Using http-client/server method:

# Start openai compatible server (requires vllm or lmdeploy environment)\nmineru-openai-server --port 30000\n

Tip

In another terminal, connect to openai server via http client

mineru -p <input_path> -o <output_path> -b hybrid-http-client -u http://127.0.0.1:30000\n

Note

All officially supported vllm/lmdeploy parameters can be passed to MinerU through command line arguments, including the following commands: mineru, mineru-openai-server, mineru-gradio, mineru-api. We have compiled some commonly used parameters and usage methods for vllm/lmdeploy, which can be found in the documentation Advanced Command Line Parameters.

"},{"location":"usage/quick_usage/#extending-mineru-functionality-with-configuration-files","title":"Extending MinerU Functionality with Configuration Files","text":"

MinerU is now ready to use out of the box, but also supports extending functionality through configuration files. You can edit mineru.json file in your user directory to add custom configurations.

Important

The mineru.json file will be automatically generated when you use the built-in model download command mineru-models-download, or you can create it by copying the configuration template file to your user directory and renaming it to mineru.json.

Here are some available configuration options:

latex-delimiter-config:
- Used to configure LaTeX formula delimiters
- Defaults to $ symbol, can be modified to other symbols or strings as needed.
llm-aided-config:
- Used to configure parameters for LLM-assisted title hierarchy
- Compatible with all LLM models supporting openai protocol, defaults to using Alibaba Cloud Bailian's qwen3-next-80b-a3b-instruct model.
- You need to configure your own API key and set enable to true to enable this feature.
- If your API provider does not support the enable_thinking parameter, please manually remove it.
  - For example, in your configuration file, the llm-aided-config section may look like:
```
\"llm-aided-config\": {\n   \"api_key\": \"your_api_key\",\n   \"base_url\": \"https://dashscope.aliyuncs.com/compatible-mode/v1\",\n   \"model\": \"qwen3-next-80b-a3b-instruct\",\n   \"enable_thinking\": false,\n   \"enable\": false\n}\n
```
  - To remove the enable_thinking parameter, simply delete the line containing \"enable_thinking\": false, resulting in:
```
\"llm-aided-config\": {\n   \"api_key\": \"your_api_key\",\n   \"base_url\": \"https://dashscope.aliyuncs.com/compatible-mode/v1\",\n   \"model\": \"qwen3-next-80b-a3b-instruct\",\n   \"enable\": false\n}\n
```
models-dir:
- Used to specify local model storage directory
- Please specify model directories for pipeline and vlm backends separately.
- After specifying the directory, you can use local models by configuring the environment variable export MINERU_MODEL_SOURCE=local.

"},{"location":"zh/","title":"MinerU","text":"

\ud83d\ude80MinerU \u5b98\u7f51\u5165\u53e3\u2192\u2705 \u514d\u88c5\u5728\u7ebf\u7248 \u2705 \u5168\u529f\u80fd\u5ba2\u6237\u7aef \u2705 \u5f00\u53d1\u8005API\u5728\u7ebf\u8c03\u7528\uff0c\u7701\u53bb\u90e8\u7f72\u9ebb\u70e6\uff0c\u591a\u79cd\u4ea7\u54c1\u5f62\u6001\u4e00\u952eget\uff0c\u901f\u51b2\uff01

\ud83d\udc4b join us on Discord and WeChat

"},{"location":"zh/#_1","title":"\u9879\u76ee\u7b80\u4ecb","text":"

MinerU\u662f\u4e00\u6b3e\u5c06PDF\u8f6c\u5316\u4e3a\u673a\u5668\u53ef\u8bfb\u683c\u5f0f\u7684\u5de5\u5177\uff08\u5982markdown\u3001json\uff09\uff0c\u53ef\u4ee5\u5f88\u65b9\u4fbf\u5730\u62bd\u53d6\u4e3a\u4efb\u610f\u683c\u5f0f\u3002 MinerU\u8bde\u751f\u4e8e\u4e66\u751f-\u6d66\u8bed\u7684\u9884\u8bad\u7ec3\u8fc7\u7a0b\u4e2d\uff0c\u6211\u4eec\u5c06\u4f1a\u96c6\u4e2d\u7cbe\u529b\u89e3\u51b3\u79d1\u6280\u6587\u732e\u4e2d\u7684\u7b26\u53f7\u8f6c\u5316\u95ee\u9898\uff0c\u5e0c\u671b\u5728\u5927\u6a21\u578b\u65f6\u4ee3\u4e3a\u79d1\u6280\u53d1\u5c55\u505a\u51fa\u8d21\u732e\u3002 \u76f8\u6bd4\u56fd\u5185\u5916\u77e5\u540d\u5546\u7528\u4ea7\u54c1MinerU\u8fd8\u5f88\u5e74\u8f7b\uff0c\u5982\u679c\u9047\u5230\u95ee\u9898\u6216\u8005\u7ed3\u679c\u4e0d\u53ca\u9884\u671f\u8bf7\u5230issue\u63d0\u4ea4\u95ee\u9898\uff0c\u540c\u65f6\u9644\u4e0a\u76f8\u5173PDF\u3002

"},{"location":"zh/#_2","title":"\u4e3b\u8981\u529f\u80fd","text":"

\u5220\u9664\u9875\u7709\u3001\u9875\u811a\u3001\u811a\u6ce8\u3001\u9875\u7801\u7b49\u5143\u7d20\uff0c\u786e\u4fdd\u8bed\u4e49\u8fde\u8d2f
\u8f93\u51fa\u7b26\u5408\u4eba\u7c7b\u9605\u8bfb\u987a\u5e8f\u7684\u6587\u672c\uff0c\u9002\u7528\u4e8e\u5355\u680f\u3001\u591a\u680f\u53ca\u590d\u6742\u6392\u7248
\u4fdd\u7559\u539f\u6587\u6863\u7684\u7ed3\u6784\uff0c\u5305\u62ec\u6807\u9898\u3001\u6bb5\u843d\u3001\u5217\u8868\u7b49
\u63d0\u53d6\u56fe\u50cf\u3001\u56fe\u7247\u63cf\u8ff0\u3001\u8868\u683c\u3001\u8868\u683c\u6807\u9898\u53ca\u811a\u6ce8
\u81ea\u52a8\u8bc6\u522b\u5e76\u8f6c\u6362\u6587\u6863\u4e2d\u7684\u516c\u5f0f\u4e3aLaTeX\u683c\u5f0f
\u81ea\u52a8\u8bc6\u522b\u5e76\u8f6c\u6362\u6587\u6863\u4e2d\u7684\u8868\u683c\u4e3aHTML\u683c\u5f0f
\u81ea\u52a8\u68c0\u6d4b\u626b\u63cf\u7248PDF\u548c\u4e71\u7801PDF\uff0c\u5e76\u542f\u7528OCR\u529f\u80fd
OCR\u652f\u6301109\u79cd\u8bed\u8a00\u7684\u68c0\u6d4b\u4e0e\u8bc6\u522b
\u652f\u6301\u591a\u79cd\u8f93\u51fa\u683c\u5f0f\uff0c\u5982\u591a\u6a21\u6001\u4e0eNLP\u7684Markdown\u3001\u6309\u9605\u8bfb\u987a\u5e8f\u6392\u5e8f\u7684JSON\u3001\u542b\u6709\u4e30\u5bcc\u4fe1\u606f\u7684\u4e2d\u95f4\u683c\u5f0f\u7b49
\u652f\u6301\u591a\u79cd\u53ef\u89c6\u5316\u7ed3\u679c\uff0c\u5305\u62eclayout\u53ef\u89c6\u5316\u3001span\u53ef\u89c6\u5316\u7b49\uff0c\u4fbf\u4e8e\u9ad8\u6548\u786e\u8ba4\u8f93\u51fa\u6548\u679c\u4e0e\u8d28\u68c0
\u652f\u6301\u7eafCPU\u73af\u5883\u8fd0\u884c\uff0c\u5e76\u652f\u6301 GPU(CUDA)/NPU(CANN)/MPS \u52a0\u901f
\u517c\u5bb9Windows\u3001Linux\u548cMac\u5e73\u53f0

"},{"location":"zh/#_3","title":"\u4f7f\u7528\u6307\u5357","text":"

\u5feb\u901f\u4e0a\u624b\u6307\u5357
\u8be6\u7ec6\u4f7f\u7528\u8bf4\u660e

"},{"location":"zh/demo/","title":"\u5728\u7ebf\u6f14\u793a","text":""},{"location":"zh/faq/","title":"\u5e38\u89c1\u95ee\u9898\u89e3\u7b54","text":""},{"location":"zh/faq/#_1","title":"\u5e38\u89c1\u95ee\u9898\u89e3\u7b54","text":"

\u5982\u679c\u672a\u80fd\u5217\u51fa\u60a8\u7684\u95ee\u9898\uff0c\u60a8\u4e5f\u53ef\u4ee5\u4f7f\u7528DeepWiki\u4e0eAI\u52a9\u624b\u4ea4\u6d41\uff0c\u8fd9\u53ef\u4ee5\u89e3\u51b3\u5927\u90e8\u5206\u5e38\u89c1\u95ee\u9898\u3002

\u5982\u679c\u60a8\u4ecd\u7136\u65e0\u6cd5\u89e3\u51b3\u95ee\u9898\uff0c\u60a8\u53ef\u901a\u8fc7Discord\u6216WeChat\u52a0\u5165\u793e\u533a\uff0c\u4e0e\u5176\u4ed6\u7528\u6237\u548c\u5f00\u53d1\u8005\u4ea4\u6d41\u3002

\u5728WSL2\u7684Ubuntu22.04\u4e2d\u9047\u5230\u62a5\u9519ImportError: libGL.so.1: cannot open shared object file: No such file or directory

WSL2\u7684Ubuntu22.04\u4e2d\u7f3a\u5c11libgl\u5e93\uff0c\u53ef\u901a\u8fc7\u4ee5\u4e0b\u547d\u4ee4\u5b89\u88c5libgl\u5e93\u89e3\u51b3\uff1a

sudo apt-get install libgl1-mesa-glx\n

\u53c2\u8003\uff1a#388

\u5728 Linux \u7cfb\u7edf\u5b89\u88c5\u5e76\u4f7f\u7528\u65f6\uff0c\u89e3\u6790\u7ed3\u679c\u7f3a\u5931\u90e8\u4efd\u6587\u5b57\u4fe1\u606f\u3002

MinerU\u5728>=2.0\u7684\u7248\u672c\u4e2d\u4f7f\u7528pypdfium2\u4ee3\u66ffpymupdf\u4f5c\u4e3aPDF\u9875\u9762\u7684\u6e32\u67d3\u5f15\u64ce\uff0c\u4ee5\u89e3\u51b3AGPLv3\u7684\u8bb8\u53ef\u8bc1\u95ee\u9898\uff0c\u5728\u67d0\u4e9bLinux\u53d1\u884c\u7248\uff0c\u7531\u4e8e\u7f3a\u5c11CJK\u5b57\u4f53\uff0c\u53ef\u80fd\u4f1a\u5728\u5c06PDF\u6e32\u67d3\u6210\u56fe\u7247\u7684\u8fc7\u7a0b\u4e2d\u4e22\u5931\u90e8\u4efd\u6587\u5b57\u3002 \u4e3a\u4e86\u89e3\u51b3\u8fd9\u4e2a\u95ee\u9898\uff0c\u60a8\u53ef\u4ee5\u901a\u8fc7\u4ee5\u4e0b\u547d\u4ee4\u5b89\u88c5noto\u5b57\u4f53\u5305\uff0c\u8fd9\u5728Ubuntu/debian\u7cfb\u7edf\u4e2d\u6709\u6548\uff1a

sudo apt update\nsudo apt install fonts-noto-core\nsudo apt install fonts-noto-cjk\nfc-cache -fv\n

\u4e5f\u53ef\u4ee5\u76f4\u63a5\u4f7f\u7528\u6211\u4eec\u7684Docker\u90e8\u7f72\u65b9\u5f0f\u6784\u5efa\u955c\u50cf\uff0c\u955c\u50cf\u4e2d\u9ed8\u8ba4\u5305\u542b\u4ee5\u4e0a\u5b57\u4f53\u5305\u3002

\u53c2\u8003\uff1a#2915

"},{"location":"zh/quick_start/","title":"\u5feb\u901f\u5165\u95e8","text":""},{"location":"zh/quick_start/#_1","title":"\u5feb\u901f\u5165\u95e8","text":"

\u5982\u679c\u9047\u5230\u4efb\u4f55\u5b89\u88c5\u95ee\u9898\uff0c\u8bf7\u5148\u67e5\u8be2 FAQ

"},{"location":"zh/quick_start/#_2","title":"\u5728\u7ebf\u4f53\u9a8c","text":""},{"location":"zh/quick_start/#_3","title":"\u5b98\u7f51\u5728\u7ebf\u5e94\u7528","text":"

\u5b98\u7f51\u5728\u7ebf\u7248\u529f\u80fd\u4e0e\u5ba2\u6237\u7aef\u4e00\u81f4\uff0c\u754c\u9762\u7f8e\u89c2\uff0c\u529f\u80fd\u4e30\u5bcc\uff0c\u9700\u8981\u767b\u5f55\u4f7f\u7528

"},{"location":"zh/quick_start/#gradiodemo","title":"\u57fa\u4e8eGradio\u7684\u5728\u7ebfdemo","text":"

\u57fa\u4e8egradio\u5f00\u53d1\u7684webui\uff0c\u754c\u9762\u7b80\u6d01\uff0c\u4ec5\u5305\u542b\u6838\u5fc3\u89e3\u6790\u529f\u80fd\uff0c\u514d\u767b\u5f55

"},{"location":"zh/quick_start/#_4","title":"\u672c\u5730\u90e8\u7f72","text":"

Warning

\u5b89\u88c5\u524d\u5fc5\u770b\u2014\u2014\u8f6f\u786c\u4ef6\u73af\u5883\u652f\u6301\u8bf4\u660e

\u4e3a\u4e86\u786e\u4fdd\u9879\u76ee\u7684\u7a33\u5b9a\u6027\u548c\u53ef\u9760\u6027\uff0c\u6211\u4eec\u5728\u5f00\u53d1\u8fc7\u7a0b\u4e2d\u4ec5\u5bf9\u7279\u5b9a\u7684\u8f6f\u786c\u4ef6\u73af\u5883\u8fdb\u884c\u4f18\u5316\u548c\u6d4b\u8bd5\u3002\u8fd9\u6837\u5f53\u7528\u6237\u5728\u63a8\u8350\u7684\u7cfb\u7edf\u914d\u7f6e\u4e0a\u90e8\u7f72\u548c\u8fd0\u884c\u9879\u76ee\u65f6\uff0c\u80fd\u591f\u83b7\u5f97\u6700\u4f73\u7684\u6027\u80fd\u8868\u73b0\u548c\u6700\u5c11\u7684\u517c\u5bb9\u6027\u95ee\u9898\u3002

\u901a\u8fc7\u96c6\u4e2d\u8d44\u6e90\u548c\u7cbe\u529b\u4e8e\u4e3b\u7ebf\u73af\u5883\uff0c\u6211\u4eec\u56e2\u961f\u80fd\u591f\u66f4\u9ad8\u6548\u5730\u89e3\u51b3\u6f5c\u5728\u7684BUG\uff0c\u53ca\u65f6\u5f00\u53d1\u65b0\u529f\u80fd\u3002

\u5728\u975e\u4e3b\u7ebf\u73af\u5883\u4e2d\uff0c\u7531\u4e8e\u786c\u4ef6\u3001\u8f6f\u4ef6\u914d\u7f6e\u7684\u591a\u6837\u6027\uff0c\u4ee5\u53ca\u7b2c\u4e09\u65b9\u4f9d\u8d56\u9879\u7684\u517c\u5bb9\u6027\u95ee\u9898\uff0c\u6211\u4eec\u65e0\u6cd5100%\u4fdd\u8bc1\u9879\u76ee\u7684\u5b8c\u5168\u53ef\u7528\u6027\u3002\u56e0\u6b64\uff0c\u5bf9\u4e8e\u5e0c\u671b\u5728\u975e\u63a8\u8350\u73af\u5883\u4e2d\u4f7f\u7528\u672c\u9879\u76ee\u7684\u7528\u6237\uff0c\u6211\u4eec\u5efa\u8bae\u5148\u4ed4\u7ec6\u9605\u8bfb\u6587\u6863\u4ee5\u53caFAQ\uff0c\u5927\u591a\u6570\u95ee\u9898\u5df2\u7ecf\u5728FAQ\u4e2d\u6709\u5bf9\u5e94\u7684\u89e3\u51b3\u65b9\u6848\uff0c\u9664\u6b64\u4e4b\u5916\u6211\u4eec\u9f13\u52b1\u793e\u533a\u53cd\u9988\u95ee\u9898\uff0c\u4ee5\u4fbf\u6211\u4eec\u80fd\u591f\u9010\u6b65\u6269\u5927\u652f\u6301\u8303\u56f4\u3002

\u89e3\u6790\u540e\u7aef pipeline *-auto-engine *-http-client hybrid vlm hybrid vlm \u540e\u7aef\u7279\u6027 \u517c\u5bb9\u6027\u597d \u786c\u4ef6\u914d\u7f6e\u8981\u6c42\u8f83\u9ad8 \u9002\u7528\u4e8eOpenAI\u517c\u5bb9\u670d\u52a1\u5668² \u7cbe\u5ea6\u6307\u6807¹ 82+ 90+ \u64cd\u4f5c\u7cfb\u7edf Linux³ / Windows⁴ / macOS⁵ \u7eafCPU\u5e73\u53f0\u652f\u6301 \u2705 \u274c \u2705 GPU\u52a0\u901f\u652f\u6301 Volta\u53ca\u4ee5\u540e\u67b6\u6784GPU\u6216Apple Silicon \u4e0d\u9700\u8981 \u663e\u5b58\u6700\u4f4e\u8981\u6c42 6GB 10GB 8GB 3GB \u5185\u5b58\u8981\u6c42 \u6700\u4f4e16GB\u4ee5\u4e0a,\u63a8\u835032GB\u4ee5\u4e0a 8GB \u78c1\u76d8\u7a7a\u95f4\u8981\u6c42 20GB\u4ee5\u4e0a,\u63a8\u8350\u4f7f\u7528SSD 2GB python\u7248\u672c 3.10-3.13

¹ \u7cbe\u5ea6\u6307\u6807\u4e3aOmniDocBench (v1.5)\u7684End-to-End Evaluation Overall\u5206\u6570\uff0c\u57fa\u4e8eMinerU\u6700\u65b0\u7248\u672c\u6d4b\u8bd5 ² \u517c\u5bb9OpenAI API\u7684\u670d\u52a1\u5668\uff0c\u5982\u901a\u8fc7vLLM/SGLang/LMDeploy\u7b49\u63a8\u7406\u6846\u67b6\u90e8\u7f72\u7684\u672c\u5730\u6a21\u578b\u670d\u52a1\u5668\u6216\u8fdc\u7a0b\u6a21\u578b\u670d\u52a1 ³ Linux\u4ec5\u652f\u63012019\u5e74\u53ca\u4ee5\u540e\u53d1\u884c\u7248 ⁴ \u7531\u4e8e\u5173\u952e\u4f9d\u8d56ray\u672a\u80fd\u5728windows\u5e73\u53f0\u652f\u6301Python 3.13\uff0c\u6545\u4ec5\u652f\u6301\u81f33.10~3.12\u7248\u672c ⁵ macOS \u9700\u4f7f\u752814.0\u4ee5\u4e0a\u7248\u672c

Tip

\u9664\u4ee5\u4e0a\u4e3b\u6d41\u73af\u5883\u4e0e\u5e73\u53f0\u5916\uff0c\u6211\u4eec\u4e5f\u6536\u5f55\u4e86\u4e00\u4e9b\u793e\u533a\u7528\u6237\u53cd\u9988\u7684\u5176\u4ed6\u5e73\u53f0\u652f\u6301\u60c5\u51b5\uff0c\u8be6\u60c5\u8bf7\u53c2\u8003\u5176\u4ed6\u52a0\u901f\u5361\u9002\u914d\u3002 \u5982\u679c\u60a8\u6709\u610f\u5c06\u81ea\u5df1\u7684\u73af\u5883\u9002\u914d\u7ecf\u9a8c\u5206\u4eab\u7ed9\u793e\u533a\uff0c\u6b22\u8fce\u901a\u8fc7show-and-tell\u63d0\u4ea4\u6216\u63d0\u4ea4PR\u81f3\u5176\u4ed6\u52a0\u901f\u5361\u9002\u914d\u6587\u6863\u3002

"},{"location":"zh/quick_start/#mineru","title":"\u5b89\u88c5 MinerU","text":""},{"location":"zh/quick_start/#pipuvmineru","title":"\u4f7f\u7528pip\u6216uv\u5b89\u88c5MinerU","text":"

pip install --upgrade pip -i https://mirrors.aliyun.com/pypi/simple\npip install uv -i https://mirrors.aliyun.com/pypi/simple\nuv pip install -U \"mineru[all]\" -i https://mirrors.aliyun.com/pypi/simple \n

"},{"location":"zh/quick_start/#mineru_1","title":"\u901a\u8fc7\u6e90\u7801\u5b89\u88c5MinerU","text":"

git clone https://github.com/opendatalab/MinerU.git\ncd MinerU\nuv pip install -e .[all] -i https://mirrors.aliyun.com/pypi/simple\n

Tip

mineru[all]\u5305\u542b\u6240\u6709\u6838\u5fc3\u529f\u80fd\uff0c\u517c\u5bb9Windows / Linux / macOS\u7cfb\u7edf\uff0c\u9002\u5408\u7edd\u5927\u591a\u6570\u7528\u6237\u3002 \u5982\u679c\u60a8\u9700\u8981\u6307\u5b9avlm\u6a21\u578b\u7684\u63a8\u7406\u6846\u67b6\uff0c\u6216\u662f\u4ec5\u51c6\u5907\u5728\u8fb9\u7f18\u8bbe\u5907\u5b89\u88c5\u8f7b\u91cf\u7248client\u7aef\uff0c\u53ef\u4ee5\u53c2\u8003\u6587\u6863\u6269\u5c55\u6a21\u5757\u5b89\u88c5\u6307\u5357\u3002

"},{"location":"zh/quick_start/#dockermineru","title":"\u4f7f\u7528docker\u90e8\u7f72Mineru","text":"

MinerU\u63d0\u4f9b\u4e86\u4fbf\u6377\u7684docker\u90e8\u7f72\u65b9\u5f0f\uff0c\u8fd9\u6709\u52a9\u4e8e\u5feb\u901f\u642d\u5efa\u73af\u5883\u5e76\u89e3\u51b3\u4e00\u4e9b\u68d8\u624b\u7684\u73af\u5883\u517c\u5bb9\u95ee\u9898\u3002 \u60a8\u53ef\u4ee5\u5728\u6587\u6863\u4e2d\u83b7\u53d6Docker\u90e8\u7f72\u8bf4\u660e\u3002

"},{"location":"zh/quick_start/#mineru_2","title":"\u4f7f\u7528 MinerU","text":"

Tip

\u9ed8\u8ba4\u4f7f\u7528\u6258\u7ba1\u5728huggingface\u7684\u6a21\u578b\u8fdb\u884c\u89e3\u6790\uff0c\u9996\u6b21\u4f7f\u7528\u65f6\u4f1a\u81ea\u52a8\u4e0b\u8f7d\u6240\u9700\u6a21\u578b\u6587\u4ef6\uff0c\u540e\u7eed\u4f7f\u7528\u5c06\u76f4\u63a5\u52a0\u8f7d\u672c\u5730\u7f13\u5b58\u7684\u6a21\u578b\u3002\u5982\u679c\u60a8\u65e0\u6cd5\u8bbf\u95eehuggingface\uff0c\u53ef\u4ee5\u901a\u8fc7\u4ee5\u4e0b\u547d\u4ee4\u5207\u6362\u81f3\u56fd\u5185\u955c\u50cf\u6e90:

export MINERU_MODEL_SOURCE=modelscope\n

\u5982\u679c\u60a8\u7684\u8bbe\u5907\u6ee1\u8db3\u4e0a\u8868\u4e2dGPU\u52a0\u901f\u7684\u6761\u4ef6\uff0c\u53ef\u4ee5\u4f7f\u7528\u7b80\u5355\u7684\u547d\u4ee4\u884c\u8fdb\u884c\u6587\u6863\u89e3\u6790:

mineru -p <input_path> -o <output_path>\n

\u5982\u679c\u60a8\u7684\u8bbe\u5907\u4e0d\u6ee1\u8db3GPU\u52a0\u901f\u6761\u4ef6\uff0c\u53ef\u4ee5\u6307\u5b9a\u540e\u7aef\u4e3apipeline\uff0c\u4ee5\u5728\u7eafCPU\u73af\u5883\u4e0b\u8fd0\u884c:

mineru -p <input_path> -o <output_path> -b pipeline\n

\u60a8\u53ef\u4ee5\u901a\u8fc7\u547d\u4ee4\u884c\u3001API\u3001WebUI\u7b49\u591a\u79cd\u65b9\u5f0f\u4f7f\u7528MinerU\u8fdb\u884cPDF\u89e3\u6790\uff0c\u5177\u4f53\u4f7f\u7528\u65b9\u6cd5\u8bf7\u53c2\u8003\u4f7f\u7528\u6307\u5357\u3002

"},{"location":"zh/quick_start/docker_deployment/","title":"Docker\u90e8\u7f72","text":""},{"location":"zh/quick_start/docker_deployment/#dockermineru","title":"\u4f7f\u7528docker\u90e8\u7f72Mineru","text":"

"},{"location":"zh/quick_start/docker_deployment/#dockerfile","title":"\u4f7f\u7528 Dockerfile \u6784\u5efa\u955c\u50cf","text":"

wget https://gcore.jsdelivr.net/gh/opendatalab/MinerU@master/docker/china/Dockerfile\ndocker build -t mineru:latest -f Dockerfile .\n

"},{"location":"zh/quick_start/docker_deployment/#docker","title":"Docker\u8bf4\u660e","text":"

Mineru\u7684docker\u4f7f\u7528\u4e86vllm/vllm-openai\u4f5c\u4e3a\u57fa\u7840\u955c\u50cf\uff0c\u56e0\u6b64\u5728docker\u4e2d\u9ed8\u8ba4\u96c6\u6210\u4e86vllm\u63a8\u7406\u52a0\u901f\u6846\u67b6\u548c\u5fc5\u9700\u7684\u4f9d\u8d56\u73af\u5883\u3002\u56e0\u6b64\u5728\u6ee1\u8db3\u6761\u4ef6\u7684\u8bbe\u5907\u4e0a\uff0c\u60a8\u53ef\u4ee5\u76f4\u63a5\u4f7f\u7528vllm\u52a0\u901fVLM\u6a21\u578b\u63a8\u7406\u3002

Note

\u4f7f\u7528vllm\u52a0\u901fVLM\u6a21\u578b\u63a8\u7406\u9700\u8981\u6ee1\u8db3\u7684\u6761\u4ef6\u662f\uff1a

\u8bbe\u5907\u5305\u542bVolta\u53ca\u4ee5\u540e\u67b6\u6784\u7684\u663e\u5361\uff0c\u4e14\u53ef\u7528\u663e\u5b58\u5927\u4e8e\u7b49\u4e8e8G\u3002
\u7269\u7406\u673a\u7684\u663e\u5361\u9a71\u52a8\u5e94\u652f\u6301CUDA 12.9.1\u6216\u66f4\u9ad8\u7248\u672c\uff0c\u53ef\u901a\u8fc7nvidia-smi\u547d\u4ee4\u68c0\u67e5\u9a71\u52a8\u7248\u672c\u3002
docker\u4e2d\u80fd\u591f\u8bbf\u95ee\u7269\u7406\u673a\u7684\u663e\u5361\u8bbe\u5907\u3002

"},{"location":"zh/quick_start/docker_deployment/#docker_1","title":"\u542f\u52a8 Docker \u5bb9\u5668","text":"

docker run --gpus all \\\n  --shm-size 32g \\\n  -p 30000:30000 -p 7860:7860 -p 8000:8000 \\\n  --ipc=host \\\n  -it mineru:latest \\\n  /bin/bash\n

\u6267\u884c\u8be5\u547d\u4ee4\u540e\uff0c\u60a8\u5c06\u8fdb\u5165\u5230Docker\u5bb9\u5668\u7684\u4ea4\u4e92\u5f0f\u7ec8\u7aef\uff0c\u5e76\u6620\u5c04\u4e86\u4e00\u4e9b\u7aef\u53e3\u7528\u4e8e\u53ef\u80fd\u4f1a\u4f7f\u7528\u7684\u670d\u52a1\uff0c\u60a8\u53ef\u4ee5\u76f4\u63a5\u5728\u5bb9\u5668\u5185\u8fd0\u884cMinerU\u76f8\u5173\u547d\u4ee4\u6765\u4f7f\u7528MinerU\u7684\u529f\u80fd\u3002 \u60a8\u4e5f\u53ef\u4ee5\u76f4\u63a5\u901a\u8fc7\u66ff\u6362/bin/bash\u4e3a\u670d\u52a1\u542f\u52a8\u547d\u4ee4\u6765\u542f\u52a8MinerU\u670d\u52a1\uff0c\u8be6\u7ec6\u8bf4\u660e\u8bf7\u53c2\u8003\u901a\u8fc7\u547d\u4ee4\u542f\u52a8\u670d\u52a1\u3002

"},{"location":"zh/quick_start/docker_deployment/#docker-compose","title":"\u901a\u8fc7 Docker Compose \u76f4\u63a5\u542f\u52a8\u670d\u52a1","text":"

\u6211\u4eec\u63d0\u4f9b\u4e86compose.yml\u6587\u4ef6\uff0c\u60a8\u53ef\u4ee5\u901a\u8fc7\u5b83\u6765\u5feb\u901f\u542f\u52a8MinerU\u670d\u52a1\u3002

# \u4e0b\u8f7d compose.yaml \u6587\u4ef6\nwget https://gcore.jsdelivr.net/gh/opendatalab/MinerU@master/docker/compose.yaml\n

Note

compose.yaml\u6587\u4ef6\u4e2d\u5305\u542b\u4e86MinerU\u7684\u591a\u4e2a\u670d\u52a1\u914d\u7f6e\uff0c\u60a8\u53ef\u4ee5\u6839\u636e\u9700\u8981\u9009\u62e9\u542f\u52a8\u7279\u5b9a\u7684\u670d\u52a1\u3002
\u4e0d\u540c\u7684\u670d\u52a1\u53ef\u80fd\u4f1a\u6709\u989d\u5916\u7684\u53c2\u6570\u914d\u7f6e\uff0c\u60a8\u53ef\u4ee5\u5728compose.yaml\u6587\u4ef6\u4e2d\u67e5\u770b\u5e76\u7f16\u8f91\u3002
\u7531\u4e8evllm\u63a8\u7406\u52a0\u901f\u6846\u67b6\u9884\u5206\u914d\u663e\u5b58\u7684\u7279\u6027\uff0c\u60a8\u53ef\u80fd\u65e0\u6cd5\u5728\u540c\u4e00\u53f0\u673a\u5668\u4e0a\u540c\u65f6\u8fd0\u884c\u591a\u4e2avllm\u670d\u52a1\uff0c\u56e0\u6b64\u8bf7\u786e\u4fdd\u5728\u542f\u52a8vlm-openai-server\u670d\u52a1\u6216\u4f7f\u7528vlm-vllm-engine\u540e\u7aef\u65f6\uff0c\u5176\u4ed6\u53ef\u80fd\u4f7f\u7528\u663e\u5b58\u7684\u670d\u52a1\u5df2\u505c\u6b62\u3002

"},{"location":"zh/quick_start/docker_deployment/#openai","title":"\u542f\u52a8 openai\u517c\u5bb9\u63a5\u53e3 \u670d\u52a1","text":"

\u5e76\u901a\u8fc7vlm-http-client\u540e\u7aef\u8fde\u63a5openai-server

docker compose -f compose.yaml --profile openai-server up -d\n

Tip

\u5728\u53e6\u4e00\u4e2a\u7ec8\u7aef\u4e2d\u901a\u8fc7http client\u8fde\u63a5openai server\uff08\u53ea\u9700cpu\u4e0e\u7f51\u7edc\uff0c\u4e0d\u9700\u8981vllm\u73af\u5883\uff09

mineru -p <input_path> -o <output_path> -b vlm-http-client -u http://<server_ip>:30000\n

"},{"location":"zh/quick_start/docker_deployment/#web-api","title":"\u542f\u52a8 Web API \u670d\u52a1","text":"

docker compose -f compose.yaml --profile api up -d\n

Tip

\u5728\u6d4f\u89c8\u5668\u4e2d\u8bbf\u95ee http://<server_ip>:8000/docs \u67e5\u770bAPI\u6587\u6863\u3002

"},{"location":"zh/quick_start/docker_deployment/#gradio-webui","title":"\u542f\u52a8 Gradio WebUI \u670d\u52a1","text":"

docker compose -f compose.yaml --profile gradio up -d\n

Tip

\u5728\u6d4f\u89c8\u5668\u4e2d\u8bbf\u95ee http://<server_ip>:7860 \u4f7f\u7528 Gradio WebUI\u3002

"},{"location":"zh/quick_start/extension_modules/","title":"\u6269\u5c55\u6a21\u5757\u5b89\u88c5","text":""},{"location":"zh/quick_start/extension_modules/#mineru","title":"MinerU \u6269\u5c55\u6a21\u5757\u5b89\u88c5\u6307\u5357","text":"

MinerU \u652f\u6301\u6839\u636e\u4e0d\u540c\u9700\u6c42\uff0c\u6309\u9700\u5b89\u88c5\u6269\u5c55\u6a21\u5757\uff0c\u4ee5\u589e\u5f3a\u529f\u80fd\u6216\u652f\u6301\u7279\u5b9a\u7684\u6a21\u578b\u540e\u7aef\u3002

"},{"location":"zh/quick_start/extension_modules/#_1","title":"\u5e38\u89c1\u573a\u666f","text":""},{"location":"zh/quick_start/extension_modules/#_2","title":"\u6838\u5fc3\u529f\u80fd\u5b89\u88c5","text":"

core \u6a21\u5757\u662f MinerU \u7684\u6838\u5fc3\u4f9d\u8d56\uff0c\u5305\u542b\u4e86\u9664vllm/lmdeploy\u5916\u7684\u6240\u6709\u529f\u80fd\u6a21\u5757\u3002\u5b89\u88c5\u6b64\u6a21\u5757\u53ef\u4ee5\u786e\u4fdd MinerU \u7684\u57fa\u672c\u529f\u80fd\u6b63\u5e38\u8fd0\u884c\u3002

uv pip install \"mineru[core]\"\n

"},{"location":"zh/quick_start/extension_modules/#vllm-vlm","title":"\u4f7f\u7528vllm\u52a0\u901f VLM \u6a21\u578b\u63a8\u7406","text":"

Note

vllm\u548clmdeploy\u5bf9vlm\u7684\u63a8\u7406\u52a0\u901f\u6548\u679c\u548c\u4f7f\u7528\u65b9\u5f0f\u51e0\u4e4e\u76f8\u540c\uff0c\u60a8\u53ef\u4ee5\u6839\u636e\u5b9e\u9645\u60c5\u51b5\u9009\u62e9\u5176\u4e2d\u4e4b\u4e00\u8fdb\u884c\u5b89\u88c5\u548c\u4f7f\u7528\uff0c\u4f46\u4e0d\u5efa\u8bae\u540c\u65f6\u5b89\u88c5\u8fd9\u4e24\u4e2a\u6a21\u5757\uff0c\u4ee5\u907f\u514d\u6f5c\u5728\u7684\u4f9d\u8d56\u51b2\u7a81\u3002

vllm \u6a21\u5757\u63d0\u4f9b\u4e86\u5bf9 VLM \u6a21\u578b\u63a8\u7406\u7684\u52a0\u901f\u652f\u6301\uff0c\u9002\u7528\u4e8e\u5177\u6709 Volta \u53ca\u4ee5\u540e\u67b6\u6784\u7684\u663e\u5361\uff088G \u663e\u5b58\u53ca\u4ee5\u4e0a\uff09\u3002\u5b89\u88c5\u6b64\u6a21\u5757\u53ef\u4ee5\u663e\u8457\u63d0\u5347\u6a21\u578b\u63a8\u7406\u901f\u5ea6\u3002

uv pip install \"mineru[core,vllm]\"\n

Tip

\u5982\u5728\u5b89\u88c5\u5305\u542bvllm\u7684\u6269\u5c55\u5305\u8fc7\u7a0b\u4e2d\u53d1\u751f\u5f02\u5e38\uff0c\u8bf7\u53c2\u8003 vllm \u5b98\u65b9\u6587\u6863 \u5c1d\u8bd5\u89e3\u51b3\uff0c\u6216\u76f4\u63a5\u4f7f\u7528 Docker \u65b9\u5f0f\u90e8\u7f72\u955c\u50cf\u3002

"},{"location":"zh/quick_start/extension_modules/#lmdeploy-vlm","title":"\u4f7f\u7528lmdeploy\u52a0\u901f VLM \u6a21\u578b\u63a8\u7406","text":"

Note

lmdeploy \u6a21\u5757\u63d0\u4f9b\u4e86\u5bf9 VLM \u6a21\u578b\u63a8\u7406\u7684\u52a0\u901f\u652f\u6301\uff0c\u9002\u7528\u4e8e\u5177\u6709 Volta \u53ca\u4ee5\u540e\u67b6\u6784\u7684\u663e\u5361\uff088G \u663e\u5b58\u53ca\u4ee5\u4e0a\uff09\u3002\u5b89\u88c5\u6b64\u6a21\u5757\u53ef\u4ee5\u663e\u8457\u63d0\u5347\u6a21\u578b\u63a8\u7406\u901f\u5ea6\u3002

uv pip install \"mineru[core,lmdeploy]\"\n

Tip

\u5982\u5728\u5b89\u88c5\u5305\u542blmdeploy\u7684\u6269\u5c55\u5305\u8fc7\u7a0b\u4e2d\u53d1\u751f\u5f02\u5e38\uff0c\u8bf7\u53c2\u8003 lmdeploy \u5b98\u65b9\u6587\u6863 \u5c1d\u8bd5\u89e3\u51b3\u3002

"},{"location":"zh/quick_start/extension_modules/#clientopenai-vlm-http-client","title":"\u5b89\u88c5\u8f7b\u91cf\u7248client\u8fde\u63a5\u517c\u5bb9openai\u670d\u52a1\u5668\u4f7f\u7528 (\u9002\u7528vlm-http-client\u6a21\u5f0f)","text":"

\u5982\u679c\u60a8\u9700\u8981\u5728\u8fb9\u7f18\u8bbe\u5907\u4e0a\u5b89\u88c5\u8f7b\u91cf\u7248\u7684 client \u7aef\u4ee5\u8fde\u63a5\u517c\u5bb9 openai \u63a5\u53e3\u7684\u670d\u52a1\u7aef\u6765\u4f7f\u7528vlm\u6a21\u5f0f\uff0c\u53ef\u4ee5\u5b89\u88c5mineru\u7684\u57fa\u7840\u5305\uff0c\u975e\u5e38\u8f7b\u91cf\uff0c\u9002\u5408\u5728\u53ea\u6709cpu\u548c\u7f51\u7edc\u8fde\u63a5\u7684\u8bbe\u5907\u4e0a\u4f7f\u7528\u3002

uv pip install mineru\nmineru -p <input_path> -o <output_path> -b vlm-http-client -u http://127.0.0.1:30000\n

"},{"location":"zh/quick_start/extension_modules/#clientopenai-hybrid-http-client","title":"\u5b89\u88c5\u8f7b\u91cf\u7248client\u8fde\u63a5\u517c\u5bb9openai\u670d\u52a1\u5668\u4f7f\u7528 (\u9002\u7528hybrid-http-client\u6a21\u5f0f)","text":"

\u5982\u679c\u60a8\u9700\u8981\u5728\u8fb9\u7f18\u8bbe\u5907\u4e0a\u5b89\u88c5\u8f7b\u91cf\u7248\u7684 client \u7aef\u4ee5\u8fde\u63a5\u517c\u5bb9 openai \u63a5\u53e3\u7684\u670d\u52a1\u7aef\u6765\u4f7f\u7528hybrid\u6a21\u5f0f\uff0c\u53ef\u4ee5\u5b89\u88c5mineru\u7684pipeline\u6269\u5c55\u5305\uff0c\u76f8\u5bf9\u8f83\u8f7b\u91cf\uff0c\u53ef\u4ee5\u5728\u53ea\u6709cpu\u548c\u7f51\u7edc\u8fde\u63a5\u7684\u8bbe\u5907\u4e0a\u4f7f\u7528\uff0c\u540c\u65f6\u5728\u652f\u6301gpu\u52a0\u901f\u7684\u8bbe\u5907\u4e0a\u53ef\u4ee5\u66f4\u5feb\u8fd0\u884c\u3002

uv pip install \"mineru[pipeline]\"\nmineru -p <input_path> -o <output_path> -b hybrid-http-client -u http://127.0.0.1:30000\n

"},{"location":"zh/reference/","title":"\u53c2\u8003\u8d44\u6599","text":""},{"location":"zh/reference/#_1","title":"\u53c2\u8003\u6587\u6863","text":"

\u672c\u7ae0\u8282\u63d0\u4f9b\u4e86 MinerU \u9879\u76ee\u7684\u8be6\u7ec6\u53c2\u8003\u8d44\u6599\u3002\u5728\u8fd9\u91cc\u60a8\u53ef\u4ee5\u627e\u5230\u6280\u672f\u89c4\u8303\u3001API \u6587\u6863\u3001\u8f93\u51fa\u6587\u4ef6\u683c\u5f0f\u8bf4\u660e\u4ee5\u53ca\u7248\u672c\u5386\u53f2\u8bb0\u5f55\u3002

"},{"location":"zh/reference/#_2","title":"\u76ee\u5f55","text":"

\u8f93\u51fa\u6587\u4ef6\u8bf4\u660e - \u8be6\u7ec6\u4ecb\u7ecd\u6240\u6709\u8f93\u51fa\u6587\u4ef6\u53ca\u5176\u683c\u5f0f
\u66f4\u65b0\u65e5\u5fd7 - \u7248\u672c\u66f4\u65b0\u5386\u53f2\u548c\u53d1\u5e03\u8bf4\u660e

"},{"location":"zh/reference/#_3","title":"\u6587\u6863\u6982\u89c8","text":""},{"location":"zh/reference/#_4","title":"\u8f93\u51fa\u6587\u4ef6\u8bf4\u660e","text":"

\u7406\u89e3 MinerU \u751f\u6210\u7684\u8f93\u51fa\u6587\u4ef6\u5bf9\u4e8e\u6709\u6548\u4f7f\u7528\u5de5\u5177\u81f3\u5173\u91cd\u8981\u3002\u8f93\u51fa\u6587\u4ef6\u6587\u6863\u63d0\u4f9b\u4e86\uff1a

\u53ef\u89c6\u5316\u8c03\u8bd5\u6587\u4ef6\uff1a\u5e2e\u52a9\u60a8\u7406\u89e3\u6587\u6863\u89e3\u6790\u8fc7\u7a0b
\u7ed3\u6784\u5316\u6570\u636e\u6587\u4ef6\uff1a\u5305\u542b\u8be6\u7ec6\u7684\u89e3\u6790\u7ed3\u679c\uff0c\u53ef\u7528\u4e8e\u8fdb\u4e00\u6b65\u5904\u7406
\u6587\u4ef6\u683c\u5f0f\u89c4\u8303\uff1a\u6bcf\u79cd\u8f93\u51fa\u6587\u4ef6\u7c7b\u578b\u7684\u8be6\u7ec6\u8bf4\u660e

"},{"location":"zh/reference/#_5","title":"\u66f4\u65b0\u65e5\u5fd7","text":"

\u66f4\u65b0\u65e5\u5fd7\u8bb0\u5f55\u4e86 MinerU \u7684\u6f14\u8fdb\u5386\u7a0b\uff0c\u5305\u62ec\uff1a

\u7248\u672c\u66f4\u65b0\uff1a\u6bcf\u4e2a\u7248\u672c\u7684\u65b0\u529f\u80fd\u548c\u6539\u8fdb
\u9519\u8bef\u4fee\u590d\uff1a\u6bcf\u4e2a\u7248\u672c\u4e2d\u89e3\u51b3\u7684\u95ee\u9898
\u91cd\u5927\u53d8\u66f4\uff1a\u53ef\u80fd\u5f71\u54cd\u60a8\u4f7f\u7528\u7684\u91cd\u8981\u53d8\u66f4
\u529f\u80fd\u5f03\u7528\uff1a\u6b63\u5728\u9010\u6b65\u6dd8\u6c70\u7684\u529f\u80fd

"},{"location":"zh/reference/changelog/","title":"\u66f4\u65b0\u65e5\u5fd7","text":""},{"location":"zh/reference/changelog/#_1","title":"\u66f4\u65b0\u65e5\u5fd7","text":"

\u672c\u6587\u6863\u8bb0\u5f55\u4e86MinerU\u9879\u76ee2.6.7\u53ca\u66f4\u65e9\u7248\u672c\u7684\u66f4\u65b0\u5386\u53f2\u3002\u6700\u65b0\u7248\u672c\u7684\u66f4\u65b0\u8bf7\u67e5\u770b\u9879\u76eeREADME\u3002

"},{"location":"zh/reference/changelog/#26","title":"2.6 \u7cfb\u5217\u7248\u672c","text":""},{"location":"zh/reference/changelog/#267-20251212","title":"2.6.7 (2025/12/12)","text":"

bug\u4fee\u590d\uff1a #4168

"},{"location":"zh/reference/changelog/#266-20251202","title":"2.6.6 (2025/12/02)","text":"

Ascend\u9002\u914d\u4f18\u5316

\u4f18\u5316\u547d\u4ee4\u884c\u5de5\u5177\u521d\u59cb\u5316\u6d41\u7a0b\uff0c\u4f7fAscend\u9002\u914d\u65b9\u6848\u4e2dvlm-vllm-engine\u540e\u7aef\u5728\u547d\u4ee4\u884c\u5de5\u5177\u4e2d\u53ef\u7528\u3002
\u4e3aAtlas 300I Duo(310p)\u8bbe\u5907\u66f4\u65b0\u9002\u914d\u6587\u6863\u3002

mineru-api\u5de5\u5177\u4f18\u5316

\u4e3amineru-api\u63a5\u53e3\u53c2\u6570\u589e\u52a0\u63cf\u8ff0\u6027\u6587\u672c\uff0c\u4f18\u5316\u63a5\u53e3\u6587\u6863\u53ef\u8bfb\u6027\u3002
\u53ef\u901a\u8fc7\u73af\u5883\u53d8\u91cfMINERU_API_ENABLE_FASTAPI_DOCS\u63a7\u5236\u662f\u5426\u542f\u7528\u81ea\u52a8\u751f\u6210\u7684\u63a5\u53e3\u6587\u6863\u9875\u9762\uff0c\u9ed8\u8ba4\u4e3a\u542f\u7528\u3002
\u4e3avlm-vllm-async-engine\u3001vlm-lmdeploy-engine\u3001vlm-http-client\u540e\u7aef\u589e\u52a0\u5e76\u53d1\u6570\u914d\u7f6e\u9009\u9879\uff0c\u7528\u6237\u53ef\u901a\u8fc7\u73af\u5883\u53d8\u91cfMINERU_API_MAX_CONCURRENT_REQUESTS\u63a7\u5236api\u63a5\u53e3\u7684\u6700\u5927\u5e76\u53d1\u8bf7\u6c42\u6570\uff0c\u9ed8\u8ba4\u4e3a\u4e0d\u9650\u5236\u6570\u91cf\u3002

"},{"location":"zh/reference/changelog/#265-20251126","title":"2.6.5 (2025/11/26)","text":"

\u589e\u52a0\u65b0\u540e\u7aefvlm-lmdeploy-engine\u652f\u6301\uff0c\u4f7f\u7528\u65b9\u5f0f\u4e0evlm-vllm-(async)engine\u7c7b\u4f3c\uff0c\u4f46\u4f7f\u7528lmdeploy\u4f5c\u4e3a\u63a8\u7406\u5f15\u64ce\uff0c\u4e0evllm\u76f8\u6bd4\u989d\u5916\u652f\u6301Windows\u5e73\u53f0\u539f\u751f\u63a8\u7406\u52a0\u901f\u3002
\u65b0\u589e\u56fd\u4ea7\u7b97\u529b\u5e73\u53f0\u6607\u817e/npu\u3001\u5e73\u5934\u54e5/ppu\u3001\u6c90\u66e6/maca\u7684\u9002\u914d\u652f\u6301\uff0c\u7528\u6237\u53ef\u5728\u5bf9\u5e94\u5e73\u53f0\u4e0a\u4f7f\u7528pipeline\u4e0evlm\u6a21\u578b\uff0c\u5e76\u4f7f\u7528vllm/lmdeploy\u5f15\u64ce\u52a0\u901fvlm\u6a21\u578b\u63a8\u7406\uff0c\u5177\u4f53\u4f7f\u7528\u65b9\u5f0f\u8bf7\u53c2\u8003\u5176\u4ed6\u52a0\u901f\u5361\u9002\u914d\u3002
\u56fd\u4ea7\u5e73\u53f0\u9002\u914d\u4e0d\u6613\uff0c\u6211\u4eec\u5df2\u5c3d\u91cf\u786e\u4fdd\u9002\u914d\u7684\u5b8c\u6574\u6027\u548c\u7a33\u5b9a\u6027\uff0c\u4f46\u4ecd\u53ef\u80fd\u5b58\u5728\u4e00\u4e9b\u7a33\u5b9a\u6027/\u517c\u5bb9\u95ee\u9898\u4e0e\u7cbe\u5ea6\u5bf9\u9f50\u95ee\u9898\uff0c\u8bf7\u5927\u5bb6\u6839\u636e\u9002\u914d\u6587\u6863\u9875\u9762\u5185\u7ea2\u7eff\u706f\u60c5\u51b5\u81ea\u884c\u9009\u62e9\u5408\u9002\u7684\u73af\u5883\u4e0e\u573a\u666f\u8fdb\u884c\u4f7f\u7528\u3002
\u5982\u5728\u4f7f\u7528\u56fd\u4ea7\u5316\u5e73\u53f0\u9002\u914d\u65b9\u6848\u7684\u8fc7\u7a0b\u4e2d\u9047\u5230\u4efb\u4f55\u6587\u6863\u672a\u63d0\u53ca\u7684\u95ee\u9898\uff0c\u4e3a\u4fbf\u4e8e\u5176\u4ed6\u7528\u6237\u67e5\u627e\u89e3\u51b3\u65b9\u6848\uff0c\u8bf7\u5728discussions\u7684\u6307\u5b9a\u5e16\u5b50\u4e2d\u8fdb\u884c\u53cd\u9988\u3002

"},{"location":"zh/reference/changelog/#264-20251104","title":"2.6.4 (2025/11/04)","text":"

\u4e3apdf\u6e32\u67d3\u56fe\u7247\u589e\u52a0\u8d85\u65f6\u914d\u7f6e\uff0c\u9ed8\u8ba4\u4e3a300\u79d2\uff0c\u53ef\u901a\u8fc7\u73af\u5883\u53d8\u91cfMINERU_PDF_RENDER_TIMEOUT\u8fdb\u884c\u914d\u7f6e\uff0c\u9632\u6b62\u90e8\u5206\u5f02\u5e38pdf\u6587\u4ef6\u5bfc\u81f4\u6e32\u67d3\u8fc7\u7a0b\u957f\u65f6\u95f4\u963b\u585e\u3002
\u4e3aonnx\u6a21\u578b\u589e\u52a0cpu\u7ebf\u7a0b\u6570\u914d\u7f6e\u9009\u9879\uff0c\u9ed8\u8ba4\u4e3a\u7cfb\u7edfcpu\u6838\u5fc3\u6570\uff0c\u53ef\u901a\u8fc7\u73af\u5883\u53d8\u91cfMINERU_INTRA_OP_NUM_THREADS\u548cMINERU_INTER_OP_NUM_THREADS\u8fdb\u884c\u914d\u7f6e\uff0c\u4ee5\u51cf\u5c11\u9ad8\u5e76\u53d1\u573a\u666f\u4e0b\u7684\u5bf9cpu\u8d44\u6e90\u7684\u62a2\u5360\u51b2\u7a81\u3002

"},{"location":"zh/reference/changelog/#263-20251031","title":"2.6.3 (2025/10/31)","text":"

\u589e\u52a0\u65b0\u540e\u7aefvlm-mlx-engine\u652f\u6301\uff0c\u5728Apple Silicon\u8bbe\u5907\u4e0a\u652f\u6301\u4f7f\u7528MLX\u52a0\u901fMinerU2.5\u6a21\u578b\u63a8\u7406\uff0c\u76f8\u6bd4vlm-transformers\u540e\u7aef\uff0cvlm-mlx-engine\u540e\u7aef\u901f\u5ea6\u63d0\u5347100%~200%\u3002
bug\u4fee\u590d: #3849 #3859

"},{"location":"zh/reference/changelog/#262-20251024","title":"2.6.2 (2025/10/24)","text":"

pipline\u540e\u7aef\u4f18\u5316

\u589e\u52a0\u5bf9\u4e2d\u6587\u516c\u5f0f\u7684\u5b9e\u9a8c\u6027\u652f\u6301\uff0c\u53ef\u901a\u8fc7\u914d\u7f6e\u73af\u5883\u53d8\u91cfexport MINERU_FORMULA_CH_SUPPORT=1\u5f00\u542f\u3002\u8be5\u529f\u80fd\u53ef\u80fd\u4f1a\u5bfc\u81f4MFR\u901f\u7387\u7565\u5fae\u4e0b\u964d\u3001\u90e8\u5206\u957f\u516c\u5f0f\u8bc6\u522b\u5931\u8d25\u7b49\u95ee\u9898\uff0c\u5efa\u8bae\u4ec5\u5728\u9700\u8981\u89e3\u6790\u4e2d\u6587\u516c\u5f0f\u7684\u573a\u666f\u4e0b\u5f00\u542f\u3002\u5982\u9700\u5173\u95ed\u8be5\u529f\u80fd\uff0c\u53ef\u5c06\u73af\u5883\u53d8\u91cf\u8bbe\u7f6e\u4e3a0\u3002
OCR\u901f\u5ea6\u5927\u5e45\u63d0\u5347200%~300%\uff0c\u611f\u8c22 @cjsdurj \u63d0\u4f9b\u7684\u4f18\u5316\u65b9\u6848
OCR\u6a21\u578b\u4f18\u5316\u62c9\u4e01\u6587\u8bc6\u522b\u7684\u51c6\u5ea6\u548c\u5e7f\u5ea6\uff0c\u5e76\u66f4\u65b0\u897f\u91cc\u5c14\u6587(cyrillic)\u3001\u963f\u62c9\u4f2f\u6587(arabic)\u3001\u5929\u57ce\u6587(devanagari)\u3001\u6cf0\u5362\u56fa\u8bed(te)\u3001\u6cf0\u7c73\u5c14\u8bed(ta)\u8bed\u7cfb\u81f3ppocr-v5\u7248\u672c\uff0c\u7cbe\u5ea6\u76f8\u6bd4\u4e0a\u4ee3\u6a21\u578b\u63d0\u534740%\u4ee5\u4e0a

vlm\u540e\u7aef\u4f18\u5316

table_caption\u3001table_footnote\u5339\u914d\u903b\u8f91\u4f18\u5316\uff0c\u63d0\u5347\u9875\u5185\u591a\u5f20\u8fde\u7eed\u8868\u573a\u666f\u4e0b\u7684\u8868\u683c\u6807\u9898\u548c\u811a\u6ce8\u7684\u5339\u914d\u51c6\u786e\u7387\u548c\u9605\u8bfb\u987a\u5e8f\u5408\u7406\u6027
\u4f18\u5316\u4f7f\u7528vllm\u540e\u7aef\u65f6\u9ad8\u5e76\u53d1\u65f6\u7684cpu\u8d44\u6e90\u5360\u7528\uff0c\u964d\u4f4e\u670d\u52a1\u7aef\u538b\u529b
\u9002\u914dvllm0.11.0\u7248\u672c

\u901a\u7528\u4f18\u5316

\u8de8\u9875\u8868\u683c\u5408\u5e76\u6548\u679c\u4f18\u5316\uff0c\u65b0\u589e\u8de8\u9875\u7eed\u8868\u5408\u5e76\u652f\u6301\uff0c\u63d0\u5347\u5728\u591a\u5217\u5408\u5e76\u573a\u666f\u4e0b\u7684\u8868\u683c\u5408\u5e76\u6548\u679c
\u4e3a\u8868\u683c\u5408\u5e76\u529f\u80fd\u589e\u52a0\u73af\u5883\u53d8\u91cf\u914d\u7f6e\u9009\u9879MINERU_TABLE_MERGE_ENABLE\uff0c\u8868\u683c\u5408\u5e76\u529f\u80fd\u9ed8\u8ba4\u5f00\u542f\uff0c\u53ef\u901a\u8fc7\u8bbe\u7f6e\u8be5\u53d8\u91cf\u4e3a0\u6765\u5173\u95ed\u8868\u683c\u5408\u5e76\u529f\u80fd

"},{"location":"zh/reference/changelog/#25","title":"2.5 \u7cfb\u5217\u7248\u672c","text":""},{"location":"zh/reference/changelog/#254-20250926","title":"2.5.4 (2025/09/26)","text":"

\ud83c\udf89\ud83c\udf89 MinerU2.5\u6280\u672f\u62a5\u544a\u73b0\u5df2\u53d1\u5e03\uff0c\u6b22\u8fce\u9605\u8bfb\u5168\u9762\u4e86\u89e3\u5176\u6a21\u578b\u67b6\u6784\u3001\u8bad\u7ec3\u7b56\u7565\u3001\u6570\u636e\u5de5\u7a0b\u548c\u8bc4\u6d4b\u7ed3\u679c\u3002
\u4fee\u590d\u90e8\u5206pdf\u6587\u4ef6\u88ab\u8bc6\u522b\u6210ai\u6587\u4ef6\u5bfc\u81f4\u65e0\u6cd5\u89e3\u6790\u7684\u95ee\u9898

"},{"location":"zh/reference/changelog/#253-20250920","title":"2.5.3 (2025/09/20)","text":"

\u4f9d\u8d56\u7248\u672c\u8303\u56f4\u8c03\u6574\uff0c\u4f7f\u5f97Turing\u53ca\u66f4\u65e9\u67b6\u6784\u663e\u5361\u53ef\u4ee5\u4f7f\u7528vLLM\u52a0\u901f\u63a8\u7406MinerU2.5\u6a21\u578b\u3002
pipeline\u540e\u7aef\u5bf9torch 2.8.0\u7684\u4e00\u4e9b\u517c\u5bb9\u6027\u4fee\u590d\u3002
\u964d\u4f4evLLM\u5f02\u6b65\u540e\u7aef\u9ed8\u8ba4\u7684\u5e76\u53d1\u6570\uff0c\u964d\u4f4e\u670d\u52a1\u7aef\u538b\u529b\u4ee5\u907f\u514d\u9ad8\u538b\u5bfc\u81f4\u7684\u94fe\u63a5\u5173\u95ed\u95ee\u9898\u3002
\u66f4\u591a\u517c\u5bb9\u6027\u76f8\u5173\u5185\u5bb9\u8be6\u89c1\u516c\u544a

"},{"location":"zh/reference/changelog/#252-20250919","title":"2.5.2 (2025/09/19)","text":"

\u6211\u4eec\u6b63\u5f0f\u53d1\u5e03 MinerU2.5\uff0c\u5f53\u524d\u6700\u5f3a\u6587\u6863\u89e3\u6790\u591a\u6a21\u6001\u5927\u6a21\u578b\u3002\u4ec5\u51ed 1.2B \u53c2\u6570\uff0cMinerU2.5 \u5728 OmniDocBench \u6587\u6863\u89e3\u6790\u8bc4\u6d4b\u4e2d\uff0c\u7cbe\u5ea6\u5df2\u5168\u9762\u8d85\u8d8a Gemini2.5-Pro\u3001GPT-4o\u3001Qwen2.5-VL-72B\u7b49\u9876\u7ea7\u591a\u6a21\u6001\u5927\u6a21\u578b\uff0c\u5e76\u663e\u8457\u9886\u5148\u4e8e\u4e3b\u6d41\u6587\u6863\u89e3\u6790\u4e13\u7528\u6a21\u578b\uff08\u5982 dots.ocr, MonkeyOCR, PP-StructureV3 \u7b49\uff09\u3002

\u6a21\u578b\u5df2\u53d1\u5e03\u81f3HuggingFace\u548cModelScope\u5e73\u53f0\uff0c\u6b22\u8fce\u5927\u5bb6\u4e0b\u8f7d\u4f7f\u7528\uff01

\u6838\u5fc3\u4eae\u70b9

\u6781\u81f4\u80fd\u6548\uff0c\u6027\u80fdSOTA: \u4ee5 1.2B \u7684\u8f7b\u91cf\u5316\u89c4\u6a21\uff0c\u5b9e\u73b0\u4e86\u8d85\u8d8a\u767e\u4ebf\u4e43\u81f3\u5343\u4ebf\u7ea7\u6a21\u578b\u7684SOTA\u6027\u80fd\uff0c\u91cd\u65b0\u5b9a\u4e49\u4e86\u6587\u6863\u89e3\u6790\u7684\u80fd\u6548\u6bd4\u3002
\u5148\u8fdb\u67b6\u6784\uff0c\u5168\u9762\u9886\u5148: \u901a\u8fc7 \"\u4e24\u9636\u6bb5\u63a8\u7406\" (\u89e3\u8026\u5e03\u5c40\u5206\u6790\u4e0e\u5185\u5bb9\u8bc6\u522b) \u4e0e \u539f\u751f\u9ad8\u5206\u8fa8\u7387\u67b6\u6784 \u7684\u7ed3\u5408\uff0c\u5728\u5e03\u5c40\u5206\u6790\u3001\u6587\u672c\u8bc6\u522b\u3001\u516c\u5f0f\u8bc6\u522b\u3001\u8868\u683c\u8bc6\u522b\u53ca\u9605\u8bfb\u987a\u5e8f\u4e94\u5927\u65b9\u9762\u5747\u8fbe\u5230 SOTA \u6c34\u5e73\u3002

\u5173\u952e\u80fd\u529b\u63d0\u5347

\u5e03\u5c40\u68c0\u6d4b: \u7ed3\u679c\u66f4\u5b8c\u6574\uff0c\u7cbe\u51c6\u8986\u76d6\u9875\u7709\u3001\u9875\u811a\u3001\u9875\u7801\u7b49\u975e\u6b63\u6587\u5185\u5bb9\uff1b\u540c\u65f6\u63d0\u4f9b\u66f4\u7cbe\u51c6\u7684\u5143\u7d20\u5b9a\u4f4d\u4e0e\u66f4\u81ea\u7136\u7684\u683c\u5f0f\u8fd8\u539f\uff08\u5982\u5217\u8868\u3001\u53c2\u8003\u6587\u732e\uff09\u3002
\u8868\u683c\u89e3\u6790: \u5927\u5e45\u4f18\u5316\u4e86\u5bf9\u65cb\u8f6c\u8868\u683c\u3001\u65e0\u7ebf/\u5c11\u7ebf\u8868\u3001\u4ee5\u53ca\u957f\u96be\u8868\u683c\u7684\u89e3\u6790\u80fd\u529b\u3002
\u516c\u5f0f\u8bc6\u522b: \u663e\u8457\u63d0\u5347\u4e2d\u82f1\u6df7\u5408\u516c\u5f0f\u53ca\u590d\u6742\u957f\u516c\u5f0f\u7684\u8bc6\u522b\u51c6\u786e\u7387\uff0c\u5927\u5e45\u6539\u5584\u6570\u5b66\u7c7b\u6587\u6863\u89e3\u6790\u80fd\u529b\u3002

\u4ed3\u5e93\u8c03\u6574

\u6b64\u5916\uff0c\u4f34\u968fvlm 2.5\u7684\u53d1\u5e03\uff0c\u6211\u4eec\u5bf9\u4ed3\u5e93\u505a\u51fa\u4e00\u4e9b\u8c03\u6574\uff1a

vlm\u540e\u7aef\u5347\u7ea7\u81f32.5\u7248\u672c\uff0c\u652f\u6301MinerU2.5\u6a21\u578b\uff0c\u4e0d\u518d\u517c\u5bb9MinerU2.0-2505-0.9B\u6a21\u578b\uff0c\u6700\u540e\u4e00\u4e2a\u652f\u63012.0\u6a21\u578b\u7684\u7248\u672c\u4e3amineru-2.2.2\u3002
vlm\u63a8\u7406\u76f8\u5173\u4ee3\u7801\u5df2\u79fb\u81f3mineru_vl_utils,\u964d\u4f4e\u4e0emineru\u4e3b\u4ed3\u5e93\u7684\u8026\u5408\u5ea6\uff0c\u4fbf\u4e8e\u540e\u7eed\u72ec\u7acb\u8fed\u4ee3\u3002
vlm\u52a0\u901f\u63a8\u7406\u6846\u67b6\u4ecesglang\u5207\u6362\u81f3vllm,\u5e76\u5b9e\u73b0\u5bf9vllm\u751f\u6001\u7684\u5b8c\u5168\u517c\u5bb9\uff0c\u4f7f\u5f97\u7528\u6237\u53ef\u4ee5\u5728\u4efb\u4f55\u652f\u6301vllm\u6846\u67b6\u7684\u5e73\u53f0\u4e0a\u4f7f\u7528MinerU2.5\u6a21\u578b\u5e76\u52a0\u901f\u63a8\u7406\u3002
\u7531\u4e8evlm\u6a21\u578b\u7684\u91cd\u5927\u5347\u7ea7\uff0c\u652f\u6301\u66f4\u591alayout type\uff0c\u56e0\u6b64\u6211\u4eec\u5bf9\u89e3\u6790\u7684\u4e2d\u95f4\u6587\u4ef6middle.json\u548c\u7ed3\u679c\u6587\u4ef6content_list.json\u7684\u7ed3\u6784\u505a\u51fa\u4e00\u4e9b\u8c03\u6574\uff0c\u8bf7\u53c2\u8003\u6587\u6863\u4e86\u89e3\u8be6\u60c5\u3002

\u5176\u4ed6\u4ed3\u5e93\u4f18\u5316

\u79fb\u9664\u5bf9\u8f93\u5165\u6587\u4ef6\u7684\u540e\u7f00\u540d\u767d\u540d\u5355\u6821\u9a8c\uff0c\u5f53\u8f93\u5165\u6587\u4ef6\u4e3aPDF\u6587\u6863\u6216\u56fe\u7247\u65f6\uff0c\u5bf9\u6587\u4ef6\u7684\u540e\u7f00\u540d\u4e0d\u518d\u6709\u8981\u6c42\uff0c\u63d0\u5347\u6613\u7528\u6027\u3002

"},{"location":"zh/reference/changelog/#22-24","title":"2.2 - 2.4 \u7cfb\u5217\u7248\u672c","text":""},{"location":"zh/reference/changelog/#222-20250910","title":"2.2.2 (2025/09/10)","text":"

\u4fee\u590d\u65b0\u7684\u8868\u683c\u8bc6\u522b\u6a21\u578b\u5728\u90e8\u5206\u8868\u683c\u89e3\u6790\u5931\u8d25\u65f6\u5f71\u54cd\u6574\u4f53\u89e3\u6790\u4efb\u52a1\u7684\u95ee\u9898

"},{"location":"zh/reference/changelog/#221-20250908","title":"2.2.1 (2025/09/08)","text":"

\u4fee\u590d\u4f7f\u7528\u6a21\u578b\u4e0b\u8f7d\u547d\u4ee4\u65f6\uff0c\u90e8\u5206\u65b0\u589e\u6a21\u578b\u672a\u4e0b\u8f7d\u7684\u95ee\u9898

"},{"location":"zh/reference/changelog/#220-20250905","title":"2.2.0 (2025/09/05)","text":"

\u4e3b\u8981\u66f4\u65b0

\u5728\u8fd9\u4e2a\u7248\u672c\u6211\u4eec\u91cd\u70b9\u63d0\u5347\u4e86\u8868\u683c\u7684\u89e3\u6790\u7cbe\u5ea6\uff0c\u901a\u8fc7\u5f15\u5165\u65b0\u7684\u6709\u7ebf\u8868\u8bc6\u522b\u6a21\u578b\u548c\u5168\u65b0\u7684\u6df7\u5408\u8868\u683c\u7ed3\u6784\u89e3\u6790\u7b97\u6cd5\uff0c\u663e\u8457\u63d0\u5347\u4e86pipeline\u540e\u7aef\u7684\u8868\u683c\u8bc6\u522b\u80fd\u529b\u3002
\u53e6\u5916\u6211\u4eec\u589e\u52a0\u4e86\u5bf9\u8de8\u9875\u8868\u683c\u5408\u5e76\u7684\u652f\u6301\uff0c\u8fd9\u4e00\u529f\u80fd\u540c\u65f6\u652f\u6301pipeline\u548cvlm\u540e\u7aef\uff0c\u8fdb\u4e00\u6b65\u63d0\u5347\u4e86\u8868\u683c\u89e3\u6790\u7684\u5b8c\u6574\u6027\u548c\u51c6\u786e\u6027\u3002

\u5176\u4ed6\u66f4\u65b0

pipeline\u540e\u7aef\u589e\u52a0270\u5ea6\u65cb\u8f6c\u7684\u8868\u683c\u89e3\u6790\u80fd\u529b\uff0c\u73b0\u5df2\u652f\u63010/90/270\u5ea6\u4e09\u4e2a\u65b9\u5411\u7684\u8868\u683c\u89e3\u6790
pipeline\u589e\u52a0\u5bf9\u6cf0\u6587\u3001\u5e0c\u814a\u6587\u7684ocr\u80fd\u529b\u652f\u6301\uff0c\u5e76\u66f4\u65b0\u4e86\u82f1\u6587ocr\u6a21\u578b\u81f3\u6700\u65b0\uff0c\u82f1\u6587\u8bc6\u522b\u7cbe\u5ea6\u63d0\u534711%\uff0c\u6cf0\u6587\u8bc6\u522b\u6a21\u578b\u7cbe\u5ea6 82.68%\uff0c\u5e0c\u814a\u6587\u8bc6\u522b\u6a21\u578b\u7cbe\u5ea6 89.28%\uff08by PPOCRv5\uff09
\u5728\u8f93\u51fa\u7684content_list.json\u4e2d\u589e\u52a0\u4e86bbox\u5b57\u6bb5(\u6620\u5c04\u81f30-1000\u8303\u56f4\u5185)\uff0c\u65b9\u4fbf\u7528\u6237\u76f4\u63a5\u83b7\u53d6\u6bcf\u4e2a\u5185\u5bb9\u5757\u7684\u4f4d\u7f6e\u4fe1\u606f
\u79fb\u9664pipeline_old_linux\u5b89\u88c5\u53ef\u9009\u9879\uff0c\u4e0d\u518d\u652f\u6301\u8001\u7248\u672c\u7684Linux\u7cfb\u7edf\u5982Centos 7\u7b49\uff0c\u4ee5\u4fbf\u5bf9uv\u7684sync/run\u7b49\u547d\u4ee4\u8fdb\u884c\u66f4\u597d\u7684\u652f\u6301

"},{"location":"zh/reference/changelog/#21","title":"2.1 \u7cfb\u5217\u7248\u672c","text":""},{"location":"zh/reference/changelog/#2110-20250801","title":"2.1.10 (2025/08/01)","text":"

\u4fee\u590dpipeline\u540e\u7aef\u56e0block\u8986\u76d6\u5bfc\u81f4\u7684\u89e3\u6790\u7ed3\u679c\u4e0e\u9884\u671f\u4e0d\u7b26 #3232

"},{"location":"zh/reference/changelog/#219-20250730","title":"2.1.9 (2025/07/30)","text":"

transformers 4.54.1 \u7248\u672c\u9002\u914d

"},{"location":"zh/reference/changelog/#218-20250728","title":"2.1.8 (2025/07/28)","text":"

sglang 0.4.9.post5 \u7248\u672c\u9002\u914d

"},{"location":"zh/reference/changelog/#217-20250727","title":"2.1.7 (2025/07/27)","text":"

transformers 4.54.0 \u7248\u672c\u9002\u914d

"},{"location":"zh/reference/changelog/#216-20250726","title":"2.1.6 (2025/07/26)","text":"

\u4fee\u590dvlm\u540e\u7aef\u89e3\u6790\u90e8\u5206\u624b\u5199\u6587\u6863\u65f6\u7684\u8868\u683c\u5f02\u5e38\u95ee\u9898
\u4fee\u590d\u6587\u6863\u65cb\u8f6c\u65f6\u53ef\u89c6\u5316\u6846\u4f4d\u7f6e\u6f02\u79fb\u95ee\u9898 #3175

"},{"location":"zh/reference/changelog/#215-20250724","title":"2.1.5 (2025/07/24)","text":"

sglang 0.4.9 \u7248\u672c\u9002\u914d\uff0c\u540c\u6b65\u5347\u7ea7dockerfile\u57fa\u7840\u955c\u50cf\u4e3asglang 0.4.9.post3

"},{"location":"zh/reference/changelog/#214-20250723","title":"2.1.4 (2025/07/23)","text":"

bug\u4fee\u590d

\u4fee\u590dpipeline\u540e\u7aef\u4e2dMFR\u6b65\u9aa4\u5728\u67d0\u4e9b\u60c5\u51b5\u4e0b\u663e\u5b58\u6d88\u8017\u8fc7\u5927\u7684\u95ee\u9898 #2771
\u4fee\u590d\u67d0\u4e9b\u60c5\u51b5\u4e0bimage/table\u4e0ecaption/footnote\u5339\u914d\u4e0d\u51c6\u786e\u7684\u95ee\u9898 #3129

"},{"location":"zh/reference/changelog/#211-20250716","title":"2.1.1 (2025/07/16)","text":"

bug\u4fee\u590d

\u4fee\u590dpipeline\u5728\u67d0\u4e9b\u60c5\u51b5\u53ef\u80fd\u53d1\u751f\u7684\u6587\u672c\u5757\u5185\u5bb9\u4e22\u5931\u95ee\u9898 #3005
\u4fee\u590dsglang-client\u9700\u8981\u5b89\u88c5torch\u7b49\u4e0d\u5fc5\u8981\u7684\u5305\u7684\u95ee\u9898 #2968
\u66f4\u65b0dockerfile\u4ee5\u4fee\u590dlinux\u5b57\u4f53\u7f3a\u5931\u5bfc\u81f4\u7684\u89e3\u6790\u6587\u672c\u5185\u5bb9\u4e0d\u5b8c\u6574\u95ee\u9898 #2915

\u6613\u7528\u6027\u66f4\u65b0

\u66f4\u65b0compose.yaml\uff0c\u4fbf\u4e8e\u7528\u6237\u76f4\u63a5\u542f\u52a8sglang-server\u3001mineru-api\u3001mineru-gradio\u670d\u52a1
\u542f\u7528\u5168\u65b0\u7684\u5728\u7ebf\u6587\u6863\u7ad9\u70b9\uff0c\u7b80\u5316readme\uff0c\u63d0\u4f9b\u66f4\u597d\u7684\u6587\u6863\u4f53\u9a8c

"},{"location":"zh/reference/changelog/#210-20250705","title":"2.1.0 (2025/07/05)","text":"

\u8fd9\u662f MinerU 2 \u7684\u7b2c\u4e00\u4e2a\u5927\u7248\u672c\u66f4\u65b0\uff0c\u5305\u542b\u4e86\u5927\u91cf\u65b0\u529f\u80fd\u548c\u6539\u8fdb\uff0c\u5305\u542b\u4f17\u591a\u6027\u80fd\u4f18\u5316\u3001\u4f53\u9a8c\u4f18\u5316\u548cbug\u4fee\u590d\uff0c\u5177\u4f53\u66f4\u65b0\u5185\u5bb9\u5982\u4e0b\uff1a

\u6027\u80fd\u4f18\u5316

\u5927\u5e45\u63d0\u5347\u67d0\u4e9b\u7279\u5b9a\u5206\u8fa8\u7387\uff08\u957f\u8fb92000\u50cf\u7d20\u5de6\u53f3\uff09\u6587\u6863\u7684\u9884\u5904\u7406\u901f\u5ea6
\u5927\u5e45\u63d0\u5347pipeline\u540e\u7aef\u6279\u91cf\u5904\u7406\u5927\u91cf\u9875\u6570\u8f83\u5c11\uff08<10\uff09\u6587\u6863\u65f6\u7684\u540e\u5904\u7406\u901f\u5ea6
pipeline\u540e\u7aef\u7684layout\u5206\u6790\u901f\u5ea6\u63d0\u5347\u7ea620%

\u4f53\u9a8c\u4f18\u5316

\u5185\u7f6e\u5f00\u7bb1\u5373\u7528\u7684fastapi\u670d\u52a1\u548cgradio webui\uff0c\u8be6\u7ec6\u4f7f\u7528\u65b9\u6cd5\u8bf7\u53c2\u8003\u6587\u6863
sglang\u9002\u914d0.4.8\u7248\u672c\uff0c\u5927\u5e45\u964d\u4f4evlm-sglang\u540e\u7aef\u7684\u663e\u5b58\u8981\u6c42\uff0c\u6700\u4f4e\u53ef\u57288G\u663e\u5b58(Turing\u53ca\u4ee5\u540e\u67b6\u6784)\u7684\u663e\u5361\u4e0a\u8fd0\u884c
\u5bf9\u6240\u6709\u547d\u4ee4\u589e\u52a0sglang\u7684\u53c2\u6570\u900f\u4f20\uff0c\u4f7f\u5f97sglang-engine\u540e\u7aef\u53ef\u4ee5\u4e0esglang-server\u4e00\u81f4\uff0c\u63a5\u6536sglang\u7684\u6240\u6709\u53c2\u6570
\u652f\u6301\u57fa\u4e8e\u914d\u7f6e\u6587\u4ef6\u7684\u529f\u80fd\u6269\u5c55\uff0c\u5305\u542b\u81ea\u5b9a\u4e49\u516c\u5f0f\u6807\u8bc6\u7b26\u3001\u5f00\u542f\u6807\u9898\u5206\u7ea7\u529f\u80fd\u3001\u81ea\u5b9a\u4e49\u672c\u5730\u6a21\u578b\u76ee\u5f55\uff0c\u8be6\u7ec6\u4f7f\u7528\u65b9\u6cd5\u8bf7\u53c2\u8003\u6587\u6863

\u65b0\u7279\u6027

pipeline\u540e\u7aef\u66f4\u65b0 PP-OCRv5 \u591a\u8bed\u79cd\u6587\u672c\u8bc6\u522b\u6a21\u578b\uff0c\u652f\u6301\u6cd5\u8bed\u3001\u897f\u73ed\u7259\u8bed\u3001\u8461\u8404\u7259\u8bed\u3001\u4fc4\u8bed\u3001\u97e9\u8bed\u7b49 37 \u79cd\u8bed\u8a00\u7684\u6587\u5b57\u8bc6\u522b\uff0c\u5e73\u5747\u7cbe\u5ea6\u6da8\u5e45\u8d8530%\u3002\u8be6\u60c5
pipeline\u540e\u7aef\u589e\u52a0\u5bf9\u7ad6\u6392\u6587\u672c\u7684\u6709\u9650\u652f\u6301

"},{"location":"zh/reference/changelog/#20","title":"2.0 \u7cfb\u5217\u7248\u672c","text":""},{"location":"zh/reference/changelog/#206-20250620","title":"2.0.6 (2025/06/20)","text":"

\u4fee\u590dvlm\u6a21\u5f0f\u4e0b\uff0c\u67d0\u4e9b\u5076\u53d1\u7684\u65e0\u6548\u5757\u5185\u5bb9\u5bfc\u81f4\u89e3\u6790\u4e2d\u65ad\u95ee\u9898
\u4fee\u590dvlm\u6a21\u5f0f\u4e0b\uff0c\u67d0\u4e9b\u4e0d\u5b8c\u6574\u7684\u8868\u7ed3\u6784\u5bfc\u81f4\u7684\u89e3\u6790\u4e2d\u65ad\u95ee\u9898

"},{"location":"zh/reference/changelog/#205-20250617","title":"2.0.5 (2025/06/17)","text":"

\u4fee\u590d\u4e86sglang-client\u6a21\u5f0f\u4e0b\u4f9d\u7136\u9700\u8981\u4e0b\u8f7d\u6a21\u578b\u7684\u95ee\u9898
\u4fee\u590d\u4e86sglang-client\u6a21\u5f0f\u9700\u8981\u4f9d\u8d56torch\u7b49\u5b9e\u9645\u8fd0\u884c\u4e0d\u9700\u8981\u7684\u5305\u7684\u95ee\u9898
\u4fee\u590d\u4e86\u540c\u4e00\u8fdb\u7a0b\u5185\u5c1d\u8bd5\u901a\u8fc7\u591a\u4e2aurl\u542f\u52a8\u591a\u4e2asglang-client\u5b9e\u4f8b\u65f6\uff0c\u53ea\u6709\u7b2c\u4e00\u4e2a\u751f\u6548\u7684\u95ee\u9898

"},{"location":"zh/reference/changelog/#203-20250615","title":"2.0.3 (2025/06/15)","text":"

\u4fee\u590d\u4e86\u5f53\u4e0b\u8f7d\u6a21\u578b\u7c7b\u578b\u8bbe\u7f6e\u4e3aall\u65f6\uff0c\u914d\u7f6e\u6587\u4ef6\u51fa\u73b0\u952e\u503c\u66f4\u65b0\u9519\u8bef\u7684\u95ee\u9898
\u4fee\u590d\u4e86\u547d\u4ee4\u884c\u6a21\u5f0f\u4e0b\u516c\u5f0f\u548c\u8868\u683c\u529f\u80fd\u5f00\u5173\u4e0d\u751f\u6548\u5bfc\u81f4\u529f\u80fd\u65e0\u6cd5\u5173\u95ed\u7684\u95ee\u9898
\u4fee\u590d\u4e86sglang-engine\u6a21\u5f0f\u4e0b\uff0c0.4.7\u7248\u672csglang\u7684\u517c\u5bb9\u6027\u95ee\u9898
\u66f4\u65b0\u4e86sglang\u73af\u5883\u4e0b\u90e8\u7f72\u5b8c\u6574\u7248MinerU\u7684Dockerfile\u548c\u76f8\u5173\u5b89\u88c5\u6587\u6863

"},{"location":"zh/reference/changelog/#200-20250613","title":"2.0.0 (2025/06/13)","text":"

\u5168\u65b0\u67b6\u6784

MinerU 2.0 \u5728\u4ee3\u7801\u7ed3\u6784\u548c\u4ea4\u4e92\u65b9\u5f0f\u4e0a\u8fdb\u884c\u4e86\u6df1\u5ea6\u91cd\u6784\uff0c\u663e\u8457\u63d0\u5347\u4e86\u7cfb\u7edf\u7684\u6613\u7528\u6027\u3001\u53ef\u7ef4\u62a4\u6027\u4e0e\u6269\u5c55\u80fd\u529b\u3002

\u53bb\u9664\u7b2c\u4e09\u65b9\u4f9d\u8d56\u9650\u5236\uff1a\u5f7b\u5e95\u79fb\u9664\u5bf9 pymupdf \u7684\u4f9d\u8d56\uff0c\u63a8\u52a8\u9879\u76ee\u5411\u66f4\u5f00\u653e\u3001\u5408\u89c4\u7684\u5f00\u6e90\u65b9\u5411\u8fc8\u8fdb\u3002
\u5f00\u7bb1\u5373\u7528\uff0c\u914d\u7f6e\u4fbf\u6377\uff1a\u65e0\u9700\u624b\u52a8\u7f16\u8f91 JSON \u914d\u7f6e\u6587\u4ef6\uff0c\u7edd\u5927\u591a\u6570\u53c2\u6570\u5df2\u652f\u6301\u547d\u4ee4\u884c\u6216 API \u76f4\u63a5\u8bbe\u7f6e\u3002
\u6a21\u578b\u81ea\u52a8\u7ba1\u7406\uff1a\u65b0\u589e\u6a21\u578b\u81ea\u52a8\u4e0b\u8f7d\u4e0e\u66f4\u65b0\u673a\u5236\uff0c\u7528\u6237\u65e0\u9700\u624b\u52a8\u5e72\u9884\u5373\u53ef\u5b8c\u6210\u6a21\u578b\u90e8\u7f72\u3002
\u79bb\u7ebf\u90e8\u7f72\u53cb\u597d\uff1a\u63d0\u4f9b\u5185\u7f6e\u6a21\u578b\u4e0b\u8f7d\u547d\u4ee4\uff0c\u652f\u6301\u5b8c\u5168\u65ad\u7f51\u73af\u5883\u4e0b\u7684\u90e8\u7f72\u9700\u6c42\u3002
\u4ee3\u7801\u7ed3\u6784\u7cbe\u7b80\uff1a\u79fb\u9664\u6570\u5343\u884c\u5197\u4f59\u4ee3\u7801\uff0c\u7b80\u5316\u7c7b\u7ee7\u627f\u903b\u8f91\uff0c\u663e\u8457\u63d0\u5347\u4ee3\u7801\u53ef\u8bfb\u6027\u4e0e\u5f00\u53d1\u6548\u7387\u3002
\u7edf\u4e00\u4e2d\u95f4\u683c\u5f0f\u8f93\u51fa\uff1a\u91c7\u7528\u6807\u51c6\u5316\u7684 middle_json \u683c\u5f0f\uff0c\u517c\u5bb9\u591a\u6570\u57fa\u4e8e\u8be5\u683c\u5f0f\u7684\u4e8c\u6b21\u5f00\u53d1\u573a\u666f\uff0c\u786e\u4fdd\u751f\u6001\u4e1a\u52a1\u65e0\u7f1d\u8fc1\u79fb\u3002

\u5168\u65b0\u6a21\u578b

MinerU 2.0 \u96c6\u6210\u4e86\u6211\u4eec\u6700\u65b0\u7814\u53d1\u7684\u5c0f\u53c2\u6570\u91cf\u3001\u9ad8\u6027\u80fd\u591a\u6a21\u6001\u6587\u6863\u89e3\u6790\u6a21\u578b\uff0c\u5b9e\u73b0\u7aef\u5230\u7aef\u7684\u9ad8\u901f\u3001\u9ad8\u7cbe\u5ea6\u6587\u6863\u7406\u89e3\u3002

\u5c0f\u6a21\u578b\uff0c\u5927\u80fd\u529b\uff1a\u6a21\u578b\u53c2\u6570\u4e0d\u8db3 1B\uff0c\u5374\u5728\u89e3\u6790\u7cbe\u5ea6\u4e0a\u8d85\u8d8a\u4f20\u7edf 72B \u7ea7\u522b\u7684\u89c6\u89c9\u8bed\u8a00\u6a21\u578b\uff08VLM\uff09\u3002
\u591a\u529f\u80fd\u5408\u4e00\uff1a\u5355\u6a21\u578b\u8986\u76d6\u591a\u8bed\u8a00\u8bc6\u522b\u3001\u624b\u5199\u8bc6\u522b\u3001\u7248\u9762\u5206\u6790\u3001\u8868\u683c\u89e3\u6790\u3001\u516c\u5f0f\u8bc6\u522b\u3001\u9605\u8bfb\u987a\u5e8f\u6392\u5e8f\u7b49\u6838\u5fc3\u4efb\u52a1\u3002
\u6781\u81f4\u63a8\u7406\u901f\u5ea6\uff1a\u5728\u5355\u5361 NVIDIA 4090 \u4e0a\u901a\u8fc7 sglang \u52a0\u901f\uff0c\u8fbe\u5230\u5cf0\u503c\u541e\u5410\u91cf\u8d85\u8fc7 10,000 token/s\uff0c\u8f7b\u677e\u5e94\u5bf9\u5927\u89c4\u6a21\u6587\u6863\u5904\u7406\u9700\u6c42\u3002
\u5728\u7ebf\u4f53\u9a8c\uff1a\u60a8\u53ef\u4ee5\u5728MinerU.net\u3001Hugging Face, \u4ee5\u53caModelScope\u4f53\u9a8c\u6211\u4eec\u7684\u5168\u65b0VLM\u6a21\u578b

\u4e0d\u517c\u5bb9\u53d8\u66f4\u8bf4\u660e

\u4e3a\u63d0\u5347\u6574\u4f53\u67b6\u6784\u5408\u7406\u6027\u4e0e\u957f\u671f\u53ef\u7ef4\u62a4\u6027\uff0c\u672c\u7248\u672c\u5305\u542b\u90e8\u5206\u4e0d\u517c\u5bb9\u7684\u53d8\u66f4\uff1a

Python \u5305\u540d\u4ece magic-pdf \u66f4\u6539\u4e3a mineru\uff0c\u547d\u4ee4\u884c\u5de5\u5177\u4e5f\u7531 magic-pdf \u6539\u4e3a mineru\uff0c\u8bf7\u540c\u6b65\u66f4\u65b0\u811a\u672c\u4e0e\u8c03\u7528\u547d\u4ee4\u3002
\u51fa\u4e8e\u5bf9\u7cfb\u7edf\u6a21\u5757\u5316\u8bbe\u8ba1\u4e0e\u751f\u6001\u4e00\u81f4\u6027\u7684\u8003\u8651\uff0cMinerU 2.0 \u5df2\u4e0d\u518d\u5185\u7f6e LibreOffice \u6587\u6863\u8f6c\u6362\u6a21\u5757\u3002\u5982\u9700\u5904\u7406 Office \u6587\u6863\uff0c\u5efa\u8bae\u901a\u8fc7\u72ec\u7acb\u90e8\u7f72\u7684 LibreOffice \u670d\u52a1\u5148\u884c\u8f6c\u6362\u4e3a PDF \u683c\u5f0f\uff0c\u518d\u8fdb\u884c\u540e\u7eed\u89e3\u6790\u64cd\u4f5c\u3002

"},{"location":"zh/reference/changelog/#1x","title":"1.x \u7cfb\u5217\u5386\u53f2\u7248\u672c","text":""},{"location":"zh/reference/changelog/#1312-20250524","title":"1.3.12 (2025/05/24)","text":"

\u589e\u52a0ppocrv5\u6a21\u578b\u7684\u652f\u6301\uff0c\u5c06ch_server\u6a21\u578b\u66f4\u65b0\u4e3aPP-OCRv5_rec_server\uff0cch_lite\u6a21\u578b\u66f4\u65b0\u4e3aPP-OCRv5_rec_mobile\uff08\u9700\u66f4\u65b0\u6a21\u578b\uff09

\u5728\u6d4b\u8bd5\u4e2d\uff0c\u53d1\u73b0ppocrv5(server)\u5bf9\u624b\u5199\u6587\u6863\u6548\u679c\u6709\u4e00\u5b9a\u63d0\u5347\uff0c\u4f46\u5728\u5176\u4f59\u7c7b\u522b\u6587\u6863\u7684\u7cbe\u5ea6\u7565\u5dee\u4e8ev4_server_doc\uff0c\u56e0\u6b64\u9ed8\u8ba4\u7684ch\u6a21\u578b\u4fdd\u6301\u4e0d\u53d8\uff0c\u4ecd\u4e3aPP-OCRv4_server_rec_doc\u3002
\u7531\u4e8eppocrv5\u5f3a\u5316\u4e86\u624b\u5199\u573a\u666f\u548c\u7279\u6b8a\u5b57\u7b26\u7684\u8bc6\u522b\u80fd\u529b\uff0c\u56e0\u6b64\u60a8\u53ef\u4ee5\u5728\u65e5\u7e41\u6df7\u5408\u573a\u666f\u4ee5\u53ca\u624b\u5199\u6587\u6863\u573a\u666f\u4e0b\u624b\u52a8\u9009\u62e9\u4f7f\u7528ppocrv5\u6a21\u578b
\u60a8\u53ef\u901a\u8fc7lang\u53c2\u6570lang='ch_server'(python api)\u6216--lang ch_server(\u547d\u4ee4\u884c)\u81ea\u884c\u9009\u62e9\u76f8\u5e94\u7684\u6a21\u578b\uff1a
ch \uff1aPP-OCRv4_rec_server_doc\uff08\u9ed8\u8ba4\uff09\uff08\u4e2d\u82f1\u65e5\u7e41\u6df7\u5408/1.5w\u5b57\u5178\uff09
ch_server \uff1aPP-OCRv5_rec_server\uff08\u4e2d\u82f1\u65e5\u7e41\u6df7\u5408+\u624b\u5199\u573a\u666f/1.8w\u5b57\u5178\uff09
ch_lite \uff1aPP-OCRv5_rec_mobile\uff08\u4e2d\u82f1\u65e5\u7e41\u6df7\u5408+\u624b\u5199\u573a\u666f/1.8w\u5b57\u5178\uff09
ch_server_v4 \uff1aPP-OCRv4_rec_server\uff08\u4e2d\u82f1\u6df7\u5408/6k\u5b57\u5178\uff09
ch_lite_v4 \uff1aPP-OCRv4_rec_mobile\uff08\u4e2d\u82f1\u6df7\u5408/6k\u5b57\u5178\uff09

\u589e\u52a0\u624b\u5199\u6587\u6863\u7684\u652f\u6301\uff0c\u901a\u8fc7\u4f18\u5316layout\u5bf9\u624b\u5199\u6587\u672c\u533a\u57df\u7684\u8bc6\u522b\uff0c\u73b0\u5df2\u652f\u6301\u624b\u5199\u6587\u6863\u7684\u89e3\u6790

\u9ed8\u8ba4\u652f\u6301\u6b64\u529f\u80fd\uff0c\u65e0\u9700\u989d\u5916\u914d\u7f6e
\u53ef\u4ee5\u53c2\u8003\u4e0a\u8ff0\u8bf4\u660e\uff0c\u624b\u52a8\u9009\u62e9ppocrv5\u6a21\u578b\u4ee5\u83b7\u5f97\u66f4\u597d\u7684\u624b\u5199\u6587\u6863\u89e3\u6790\u6548\u679c

huggingface\u548cmodelscope\u7684demo\u5df2\u66f4\u65b0\u4e3a\u652f\u6301\u624b\u5199\u8bc6\u522b\u548cppocrv5\u6a21\u578b\u7684\u7248\u672c\uff0c\u53ef\u81ea\u884c\u5728\u7ebf\u4f53\u9a8c

"},{"location":"zh/reference/changelog/#1310-20250429","title":"1.3.10 (2025/04/29)","text":"

\u652f\u6301\u4f7f\u7528\u81ea\u5b9a\u4e49\u516c\u5f0f\u6807\u8bc6\u7b26\uff0c\u53ef\u901a\u8fc7\u4fee\u6539\u7528\u6237\u76ee\u5f55\u4e0b\u7684magic-pdf.json\u6587\u4ef6\u4e2d\u7684latex-delimiter-config\u9879\u5b9e\u73b0\u3002

"},{"location":"zh/reference/changelog/#139-20250427","title":"1.3.9 (2025/04/27)","text":"

\u4f18\u5316\u516c\u5f0f\u89e3\u6790\u529f\u80fd\uff0c\u63d0\u5347\u516c\u5f0f\u6e32\u67d3\u7684\u6210\u529f\u7387

"},{"location":"zh/reference/changelog/#138-20250423","title":"1.3.8 (2025/04/23)","text":"

ocr\u9ed8\u8ba4\u6a21\u578b(ch)\u66f4\u65b0\u4e3aPP-OCRv4_server_rec_doc\uff08\u9700\u66f4\u65b0\u6a21\u578b\uff09

PP-OCRv4_server_rec_doc\u662f\u5728PP-OCRv4_server_rec\u7684\u57fa\u7840\u4e0a\uff0c\u5728\u66f4\u591a\u4e2d\u6587\u6587\u6863\u6570\u636e\u548cPP-OCR\u8bad\u7ec3\u6570\u636e\u7684\u6df7\u5408\u6570\u636e\u8bad\u7ec3\u800c\u6210\uff0c\u589e\u52a0\u4e86\u90e8\u5206\u7e41\u4f53\u5b57\u3001\u65e5\u6587\u3001\u7279\u6b8a\u5b57\u7b26\u7684\u8bc6\u522b\u80fd\u529b\uff0c\u53ef\u652f\u6301\u8bc6\u522b\u7684\u5b57\u7b26\u4e3a1.5\u4e07+\uff0c\u9664\u6587\u6863\u76f8\u5173\u7684\u6587\u5b57\u8bc6\u522b\u80fd\u529b\u63d0\u5347\u5916\uff0c\u4e5f\u540c\u65f6\u63d0\u5347\u4e86\u901a\u7528\u6587\u5b57\u7684\u8bc6\u522b\u80fd\u529b\u3002
PP-OCRv4_server_rec_doc/PP-OCRv4_server_rec/PP-OCRv4_mobile_rec \u6027\u80fd\u5bf9\u6bd4
\u7ecf\u9a8c\u8bc1\uff0cPP-OCRv4_server_rec_doc\u6a21\u578b\u5728\u4e2d\u82f1\u65e5\u7e41\u5355\u79cd\u8bed\u8a00\u6216\u591a\u79cd\u8bed\u8a00\u6df7\u5408\u573a\u666f\u5747\u6709\u660e\u663e\u7cbe\u5ea6\u63d0\u5347\uff0c\u4e14\u901f\u5ea6\u4e0ePP-OCRv4_server_rec\u76f8\u5f53\uff0c\u9002\u5408\u7edd\u5927\u90e8\u5206\u573a\u666f\u4f7f\u7528\u3002
PP-OCRv4_server_rec_doc\u5728\u5c0f\u90e8\u5206\u7eaf\u82f1\u6587\u573a\u666f\u53ef\u80fd\u4f1a\u53d1\u751f\u5355\u8bcd\u7c98\u8fde\u95ee\u9898\uff0cPP-OCRv4_server_rec\u5219\u5728\u6b64\u573a\u666f\u4e0b\u8868\u73b0\u66f4\u597d\uff0c\u56e0\u6b64\u6211\u4eec\u4fdd\u7559\u4e86PP-OCRv4_server_rec\u6a21\u578b\uff0c\u7528\u6237\u53ef\u901a\u8fc7\u589e\u52a0\u53c2\u6570lang='ch_server'(python api)\u6216--lang ch_server(\u547d\u4ee4\u884c)\u8c03\u7528\u3002

"},{"location":"zh/reference/changelog/#137-20250422","title":"1.3.7 (2025/04/22)","text":"

\u4fee\u590d\u8868\u683c\u89e3\u6790\u6a21\u578b\u521d\u59cb\u5316\u65f6lang\u53c2\u6570\u5931\u6548\u7684\u95ee\u9898
\u4fee\u590d\u5728cpu\u6a21\u5f0f\u4e0bocr\u548c\u8868\u683c\u89e3\u6790\u901f\u5ea6\u5927\u5e45\u4e0b\u964d\u7684\u95ee\u9898

"},{"location":"zh/reference/changelog/#134-20250416","title":"1.3.4 (2025/04/16)","text":"

\u901a\u8fc7\u79fb\u9664\u4e00\u4e9b\u65e0\u7528\u7684\u5757\uff0c\u5c0f\u5e45\u63d0\u5347\u4e86ocr-det\u7684\u901f\u5ea6
\u4fee\u590d\u90e8\u5206\u60c5\u51b5\u4e0b\u7531footnote\u5bfc\u81f4\u7684\u9875\u9762\u5185\u6392\u5e8f\u9519\u8bef

"},{"location":"zh/reference/changelog/#132-20250412","title":"1.3.2 (2025/04/12)","text":"

\u4fee\u590d\u4e86windows\u7cfb\u7edf\u4e0b\uff0c\u5728python3.13\u73af\u5883\u5b89\u88c5\u65f6\u4e00\u4e9b\u4f9d\u8d56\u5305\u7248\u672c\u4e0d\u517c\u5bb9\u7684\u95ee\u9898
\u4f18\u5316\u6279\u91cf\u63a8\u7406\u65f6\u7684\u5185\u5b58\u5360\u7528
\u4f18\u5316\u65cb\u8f6c90\u5ea6\u8868\u683c\u7684\u89e3\u6790\u6548\u679c
\u4f18\u5316\u8d22\u62a5\u6837\u672c\u4e2d\u8d85\u5927\u8868\u683c\u7684\u89e3\u6790\u6548\u679c
\u4fee\u590d\u4e86\u5728\u672a\u6307\u5b9aOCR\u8bed\u8a00\u65f6\uff0c\u82f1\u6587\u6587\u672c\u533a\u57df\u5076\u5c14\u51fa\u73b0\u7684\u5355\u8bcd\u9ecf\u8fde\u95ee\u9898\uff08\u9700\u8981\u66f4\u65b0\u6a21\u578b\uff09

"},{"location":"zh/reference/changelog/#131-20250408","title":"1.3.1 (2025/04/08)","text":"

\u4fee\u590d\u4e86\u4e00\u4e9b\u517c\u5bb9\u95ee\u9898

\u652f\u6301python 3.13
\u4e3a\u90e8\u5206\u8fc7\u65f6\u7684linux\u7cfb\u7edf\uff08\u5982centos7\uff09\u505a\u51fa\u6700\u540e\u9002\u914d\uff0c\u5e76\u4e0d\u518d\u4fdd\u8bc1\u540e\u7eed\u7248\u672c\u7684\u7ee7\u7eed\u652f\u6301\uff0c\u5b89\u88c5\u8bf4\u660e

"},{"location":"zh/reference/changelog/#130-20250403","title":"1.3.0 (2025/04/03)","text":"

\u5b89\u88c5\u4e0e\u517c\u5bb9\u6027\u4f18\u5316

\u901a\u8fc7\u79fb\u9664layout\u4e2dlayoutlmv3\u7684\u4f7f\u7528\uff0c\u89e3\u51b3\u4e86\u7531detectron2\u5bfc\u81f4\u7684\u517c\u5bb9\u95ee\u9898
torch\u7248\u672c\u517c\u5bb9\u6269\u5c55\u52302.2~2.6(2.5\u9664\u5916)
cuda\u517c\u5bb9\u652f\u630111.8/12.4/12.6/12.8\uff08cuda\u7248\u672c\u7531torch\u51b3\u5b9a\uff09\uff0c\u89e3\u51b3\u90e8\u5206\u7528\u623750\u7cfb\u663e\u5361\u4e0eH\u7cfb\u663e\u5361\u7684\u517c\u5bb9\u95ee\u9898
python\u517c\u5bb9\u7248\u672c\u6269\u5c55\u52303.10~3.12\uff0c\u89e3\u51b3\u4e86\u5728\u975e3.10\u73af\u5883\u4e0b\u5b89\u88c5\u65f6\u81ea\u52a8\u964d\u7ea7\u52300.6.1\u7684\u95ee\u9898
\u4f18\u5316\u79bb\u7ebf\u90e8\u7f72\u6d41\u7a0b\uff0c\u90e8\u7f72\u6210\u529f\u540e\u4e0d\u9700\u8981\u8054\u7f51\u4e0b\u8f7d\u4efb\u4f55\u6a21\u578b\u6587\u4ef6

\u6027\u80fd\u4f18\u5316

\u901a\u8fc7\u652f\u6301\u591a\u4e2apdf\u6587\u4ef6\u7684batch\u5904\u7406\uff08\u811a\u672c\u6837\u4f8b\uff09\uff0c\u63d0\u5347\u4e86\u6279\u91cf\u5c0f\u6587\u4ef6\u7684\u89e3\u6790\u901f\u5ea6 (\u4e0e1.0.1\u7248\u672c\u76f8\u6bd4\uff0c\u516c\u5f0f\u89e3\u6790\u901f\u5ea6\u6700\u9ad8\u63d0\u5347\u8d85\u8fc71400%\uff0c\u6574\u4f53\u89e3\u6790\u901f\u5ea6\u6700\u9ad8\u63d0\u5347\u8d85\u8fc7500%)
\u901a\u8fc7\u4f18\u5316mfr\u6a21\u578b\u7684\u52a0\u8f7d\u548c\u4f7f\u7528\uff0c\u964d\u4f4e\u4e86\u663e\u5b58\u5360\u7528\u5e76\u63d0\u5347\u4e86\u89e3\u6790\u901f\u5ea6(\u9700\u91cd\u65b0\u6267\u884c\u6a21\u578b\u4e0b\u8f7d\u6d41\u7a0b\u4ee5\u83b7\u5f97\u6a21\u578b\u6587\u4ef6\u7684\u589e\u91cf\u66f4\u65b0)
\u4f18\u5316\u663e\u5b58\u5360\u7528\uff0c\u6700\u4f4e\u4ec5\u97006GB\u5373\u53ef\u8fd0\u884c\u672c\u9879\u76ee
\u4f18\u5316\u4e86\u5728mps\u8bbe\u5907\u4e0a\u7684\u8fd0\u884c\u901f\u5ea6

\u89e3\u6790\u6548\u679c\u4f18\u5316

mfr\u6a21\u578b\u66f4\u65b0\u5230unimernet(2503)\uff0c\u89e3\u51b3\u591a\u884c\u516c\u5f0f\u4e2d\u6362\u884c\u4e22\u5931\u7684\u95ee\u9898

\u6613\u7528\u6027\u4f18\u5316

\u901a\u8fc7\u4f7f\u7528paddleocr2torch\uff0c\u5b8c\u5168\u66ff\u4ee3paddle\u6846\u67b6\u4ee5\u53capaddleocr\u5728\u9879\u76ee\u4e2d\u7684\u4f7f\u7528\uff0c\u89e3\u51b3\u4e86paddle\u548ctorch\u7684\u51b2\u7a81\u95ee\u9898\uff0c\u548c\u7531\u4e8epaddle\u6846\u67b6\u5bfc\u81f4\u7684\u7ebf\u7a0b\u4e0d\u5b89\u5168\u95ee\u9898
\u89e3\u6790\u8fc7\u7a0b\u589e\u52a0\u5b9e\u65f6\u8fdb\u5ea6\u6761\u663e\u793a\uff0c\u7cbe\u51c6\u628a\u63e1\u89e3\u6790\u8fdb\u5ea6\uff0c\u8ba9\u7b49\u5f85\u4e0d\u518d\u75db\u82e6

"},{"location":"zh/reference/changelog/#121-20250303","title":"1.2.1 (2025/03/03)","text":"

\u4fee\u590d\u4e86\u4e00\u4e9b\u95ee\u9898

\u4fee\u590d\u5728\u5b57\u6bcd\u4e0e\u6570\u5b57\u7684\u5168\u89d2\u8f6c\u534a\u89d2\u64cd\u4f5c\u65f6\u5bf9\u6807\u70b9\u7b26\u53f7\u7684\u5f71\u54cd
\u4fee\u590d\u5728\u67d0\u4e9b\u60c5\u51b5\u4e0bcaption\u7684\u5339\u914d\u4e0d\u51c6\u786e\u95ee\u9898
\u4fee\u590d\u5728\u67d0\u4e9b\u60c5\u51b5\u4e0b\u7684\u516c\u5f0fspan\u4e22\u5931\u95ee\u9898

"},{"location":"zh/reference/changelog/#120-20250224","title":"1.2.0 (2025/02/24)","text":"

\u8fd9\u4e2a\u7248\u672c\u6211\u4eec\u4fee\u590d\u4e86\u4e00\u4e9b\u95ee\u9898\uff0c\u63d0\u5347\u4e86\u89e3\u6790\u7684\u6548\u7387\u4e0e\u7cbe\u5ea6\uff1a

\u6027\u80fd\u4f18\u5316

auto\u6a21\u5f0f\u4e0bpdf\u6587\u6863\u7684\u5206\u7c7b\u901f\u5ea6\u63d0\u5347

\u89e3\u6790\u4f18\u5316

\u4f18\u5316\u5bf9\u5305\u542b\u6c34\u5370\u6587\u6863\u7684\u89e3\u6790\u903b\u8f91\uff0c\u663e\u8457\u63d0\u5347\u5305\u542b\u6c34\u5370\u6587\u6863\u7684\u89e3\u6790\u6548\u679c
\u6539\u8fdb\u4e86\u5355\u9875\u5185\u591a\u4e2a\u56fe\u50cf/\u8868\u683c\u4e0ecaption\u7684\u5339\u914d\u903b\u8f91\uff0c\u63d0\u5347\u4e86\u590d\u6742\u5e03\u5c40\u4e0b\u56fe\u6587\u5339\u914d\u7684\u51c6\u786e\u6027

\u95ee\u9898\u4fee\u590d

\u4fee\u590d\u5728\u67d0\u4e9b\u60c5\u51b5\u4e0b\u56fe\u7247/\u8868\u683cspan\u88ab\u586b\u5145\u8fdbtextblock\u5bfc\u81f4\u7684\u5f02\u5e38
\u4fee\u590d\u5728\u67d0\u4e9b\u60c5\u51b5\u4e0b\u6807\u9898block\u4e3a\u7a7a\u7684\u95ee\u9898

"},{"location":"zh/reference/changelog/#110-20250122","title":"1.1.0 (2025/01/22)","text":"

\u5728\u8fd9\u4e2a\u7248\u672c\u6211\u4eec\u91cd\u70b9\u63d0\u5347\u4e86\u89e3\u6790\u7684\u7cbe\u5ea6\u4e0e\u6548\u7387\uff1a

\u6a21\u578b\u80fd\u529b\u5347\u7ea7\uff08\u9700\u91cd\u65b0\u6267\u884c \u6a21\u578b\u4e0b\u8f7d\u6d41\u7a0b \u4ee5\u83b7\u5f97\u6a21\u578b\u6587\u4ef6\u7684\u589e\u91cf\u66f4\u65b0\uff09

\u5e03\u5c40\u8bc6\u522b\u6a21\u578b\u5347\u7ea7\u5230\u6700\u65b0\u7684 doclayout_yolo(2501) \u6a21\u578b\uff0c\u63d0\u5347\u4e86layout\u8bc6\u522b\u7cbe\u5ea6
\u516c\u5f0f\u89e3\u6790\u6a21\u578b\u5347\u7ea7\u5230\u6700\u65b0\u7684 unimernet(2501) \u6a21\u578b\uff0c\u63d0\u5347\u4e86\u516c\u5f0f\u8bc6\u522b\u7cbe\u5ea6

\u6027\u80fd\u4f18\u5316

\u5728\u914d\u7f6e\u6ee1\u8db3\u4e00\u5b9a\u6761\u4ef6\uff08\u663e\u5b5816GB+\uff09\u7684\u8bbe\u5907\u4e0a\uff0c\u901a\u8fc7\u4f18\u5316\u8d44\u6e90\u5360\u7528\u548c\u91cd\u6784\u5904\u7406\u6d41\u6c34\u7ebf\uff0c\u6574\u4f53\u89e3\u6790\u901f\u5ea6\u63d0\u534750%\u4ee5\u4e0a

\u89e3\u6790\u6548\u679c\u4f18\u5316

\u5728\u7ebfdemo\uff08mineru.net / huggingface / modelscope\uff09\u4e0a\u65b0\u589e\u6807\u9898\u5206\u7ea7\u529f\u80fd\uff08\u6d4b\u8bd5\u7248\u672c\uff0c\u9ed8\u8ba4\u5f00\u542f\uff09\uff0c\u652f\u6301\u5bf9\u6807\u9898\u8fdb\u884c\u5206\u7ea7\uff0c\u63d0\u5347\u6587\u6863\u7ed3\u6784\u5316\u7a0b\u5ea6

"},{"location":"zh/reference/changelog/#101-20250110","title":"1.0.1 (2025/01/10)","text":"

\u8fd9\u662f\u6211\u4eec\u7684\u7b2c\u4e00\u4e2a\u6b63\u5f0f\u7248\u672c\uff0c\u5728\u8fd9\u4e2a\u7248\u672c\u4e2d\uff0c\u6211\u4eec\u901a\u8fc7\u5927\u91cf\u91cd\u6784\u5e26\u6765\u4e86\u5168\u65b0\u7684API\u63a5\u53e3\u548c\u66f4\u5e7f\u6cdb\u7684\u517c\u5bb9\u6027\uff0c\u4ee5\u53ca\u5168\u65b0\u7684\u81ea\u52a8\u8bed\u8a00\u8bc6\u522b\u529f\u80fd\uff1a

\u5168\u65b0API\u63a5\u53e3

\u5bf9\u4e8e\u6570\u636e\u4fa7API\uff0c\u6211\u4eec\u5f15\u5165\u4e86Dataset\u7c7b\uff0c\u65e8\u5728\u63d0\u4f9b\u4e00\u4e2a\u5f3a\u5927\u800c\u7075\u6d3b\u7684\u6570\u636e\u5904\u7406\u6846\u67b6\u3002\u8be5\u6846\u67b6\u5f53\u524d\u652f\u6301\u5305\u62ec\u56fe\u50cf\uff08.jpg\u53ca.png\uff09\u3001PDF\u3001Word\uff08.doc\u53ca.docx\uff09\u3001\u4ee5\u53caPowerPoint\uff08.ppt\u53ca.pptx\uff09\u5728\u5185\u7684\u591a\u79cd\u6587\u6863\u683c\u5f0f\uff0c\u786e\u4fdd\u4e86\u4ece\u7b80\u5355\u5230\u590d\u6742\u7684\u6570\u636e\u5904\u7406\u4efb\u52a1\u90fd\u80fd\u5f97\u5230\u6709\u6548\u7684\u652f\u6301\u3002
\u9488\u5bf9\u7528\u6237\u4fa7API\uff0c\u6211\u4eec\u5c06MinerU\u7684\u5904\u7406\u6d41\u7a0b\u7cbe\u5fc3\u8bbe\u8ba1\u4e3a\u4e00\u7cfb\u5217\u53ef\u7ec4\u5408\u7684Stage\u9636\u6bb5\u3002\u6bcf\u4e2aStage\u4ee3\u8868\u4e86\u4e00\u4e2a\u7279\u5b9a\u7684\u5904\u7406\u6b65\u9aa4\uff0c\u7528\u6237\u53ef\u4ee5\u6839\u636e\u81ea\u8eab\u9700\u6c42\u81ea\u7531\u5730\u5b9a\u4e49\u65b0\u7684Stage\uff0c\u5e76\u901a\u8fc7\u521b\u9020\u6027\u5730\u7ec4\u5408\u8fd9\u4e9b\u9636\u6bb5\u6765\u5b9a\u5236\u4e13\u5c5e\u7684\u6570\u636e\u5904\u7406\u6d41\u7a0b\u3002

\u66f4\u5e7f\u6cdb\u7684\u517c\u5bb9\u6027\u9002\u914d

\u901a\u8fc7\u4f18\u5316\u4f9d\u8d56\u73af\u5883\u548c\u914d\u7f6e\u9879\uff0c\u786e\u4fdd\u5728ARM\u67b6\u6784\u7684Linux\u7cfb\u7edf\u4e0a\u80fd\u591f\u7a33\u5b9a\u9ad8\u6548\u8fd0\u884c\u3002
\u6df1\u5ea6\u9002\u914d\u534e\u4e3a\u6607\u817eNPU\u52a0\u901f\uff0c\u79ef\u6781\u54cd\u5e94\u4fe1\u521b\u8981\u6c42\uff0c\u63d0\u4f9b\u81ea\u4e3b\u53ef\u63a7\u7684\u9ad8\u6027\u80fd\u8ba1\u7b97\u80fd\u529b\uff0c\u52a9\u529b\u4eba\u5de5\u667a\u80fd\u5e94\u7528\u5e73\u53f0\u7684\u56fd\u4ea7\u5316\u5e94\u7528\u4e0e\u53d1\u5c55\u3002 NPU\u52a0\u901f\u6559\u7a0b

\u81ea\u52a8\u8bed\u8a00\u8bc6\u522b

\u901a\u8fc7\u5f15\u5165\u5168\u65b0\u7684\u8bed\u8a00\u8bc6\u522b\u6a21\u578b\uff0c \u5728\u6587\u6863\u89e3\u6790\u4e2d\u5c06 lang \u914d\u7f6e\u4e3a auto\uff0c\u5373\u53ef\u81ea\u52a8\u9009\u62e9\u5408\u9002\u7684OCR\u8bed\u8a00\u6a21\u578b\uff0c\u63d0\u5347\u626b\u63cf\u7c7b\u6587\u6863\u89e3\u6790\u7684\u51c6\u786e\u6027\u3002

"},{"location":"zh/reference/changelog/#0x","title":"0.x \u7cfb\u5217\u5386\u53f2\u7248\u672c","text":""},{"location":"zh/reference/changelog/#0100-20241122","title":"0.10.0 (2024/11/22)","text":"

\u901a\u8fc7\u5f15\u5165\u6df7\u5408OCR\u6587\u672c\u63d0\u53d6\u80fd\u529b\uff1a

\u5728\u516c\u5f0f\u5bc6\u96c6\u3001span\u533a\u57df\u4e0d\u89c4\u8303\u3001\u90e8\u5206\u6587\u672c\u4f7f\u7528\u56fe\u50cf\u8868\u73b0\u7b49\u590d\u6742\u6587\u672c\u5206\u5e03\u573a\u666f\u4e0b\u83b7\u5f97\u89e3\u6790\u6548\u679c\u7684\u663e\u8457\u63d0\u5347
\u540c\u65f6\u5177\u5907\u6587\u672c\u6a21\u5f0f\u5185\u5bb9\u63d0\u53d6\u51c6\u786e\u3001\u901f\u5ea6\u66f4\u5feb\u4e0eOCR\u6a21\u5f0fspan/line\u533a\u57df\u8bc6\u522b\u66f4\u51c6\u7684\u53cc\u91cd\u4f18\u52bf

"},{"location":"zh/reference/changelog/#093-20241115","title":"0.9.3 (2024/11/15)","text":"

\u4e3a\u8868\u683c\u8bc6\u522b\u529f\u80fd\u63a5\u5165\u4e86RapidTable,\u5355\u8868\u89e3\u6790\u901f\u5ea6\u63d0\u534710\u500d\u4ee5\u4e0a\uff0c\u51c6\u786e\u7387\u66f4\u9ad8\uff0c\u663e\u5b58\u5360\u7528\u66f4\u4f4e

"},{"location":"zh/reference/changelog/#092-20241106","title":"0.9.2 (2024/11/06)","text":"

\u4e3a\u8868\u683c\u8bc6\u522b\u529f\u80fd\u63a5\u5165\u4e86StructTable-InternVL2-1B\u6a21\u578b

"},{"location":"zh/reference/changelog/#090-20241031","title":"0.9.0 (2024/10/31)","text":"

\u8fd9\u662f\u6211\u4eec\u8fdb\u884c\u4e86\u5927\u91cf\u4ee3\u7801\u91cd\u6784\u7684\u5168\u65b0\u7248\u672c\uff0c\u89e3\u51b3\u4e86\u4f17\u591a\u95ee\u9898\uff0c\u63d0\u5347\u4e86\u6027\u80fd\uff0c\u964d\u4f4e\u4e86\u786c\u4ef6\u9700\u6c42\uff0c\u5e76\u63d0\u4f9b\u4e86\u66f4\u4e30\u5bcc\u7684\u6613\u7528\u6027\uff1a

\u91cd\u6784\u6392\u5e8f\u6a21\u5757\u4ee3\u7801\uff0c\u4f7f\u7528 layoutreader \u8fdb\u884c\u9605\u8bfb\u987a\u5e8f\u6392\u5e8f\uff0c\u786e\u4fdd\u5728\u5404\u79cd\u6392\u7248\u4e0b\u90fd\u80fd\u5b9e\u73b0\u6781\u9ad8\u51c6\u786e\u7387
\u91cd\u6784\u6bb5\u843d\u62fc\u63a5\u6a21\u5757\uff0c\u5728\u8de8\u680f\u3001\u8de8\u9875\u3001\u8de8\u56fe\u3001\u8de8\u8868\u60c5\u51b5\u4e0b\u5747\u80fd\u5b9e\u73b0\u826f\u597d\u7684\u6bb5\u843d\u62fc\u63a5\u6548\u679c
\u91cd\u6784\u5217\u8868\u548c\u76ee\u5f55\u8bc6\u522b\u529f\u80fd\uff0c\u6781\u5927\u63d0\u5347\u5217\u8868\u5757\u548c\u76ee\u5f55\u5757\u8bc6\u522b\u7684\u51c6\u786e\u7387\u53ca\u5bf9\u5e94\u6587\u672c\u6bb5\u843d\u7684\u89e3\u6790\u6548\u679c
\u91cd\u6784\u56fe\u3001\u8868\u4e0e\u63cf\u8ff0\u6027\u6587\u672c\u7684\u5339\u914d\u903b\u8f91\uff0c\u5927\u5e45\u63d0\u5347 caption \u548c footnote \u4e0e\u56fe\u8868\u7684\u5339\u914d\u51c6\u786e\u7387\uff0c\u5e76\u5c06\u63cf\u8ff0\u6027\u6587\u672c\u7684\u4e22\u5931\u7387\u964d\u81f3\u63a5\u8fd10
\u589e\u52a0 OCR \u7684\u591a\u8bed\u8a00\u652f\u6301\uff0c\u652f\u6301 84 \u79cd\u8bed\u8a00\u7684\u68c0\u6d4b\u4e0e\u8bc6\u522b\uff0c\u8bed\u8a00\u652f\u6301\u5217\u8868\u8be6\u89c1 OCR \u8bed\u8a00\u652f\u6301\u5217\u8868
\u589e\u52a0\u663e\u5b58\u56de\u6536\u903b\u8f91\u53ca\u5176\u4ed6\u663e\u5b58\u4f18\u5316\u63aa\u65bd\uff0c\u5927\u5e45\u964d\u4f4e\u663e\u5b58\u4f7f\u7528\u9700\u6c42\u3002\u5f00\u542f\u9664\u8868\u683c\u52a0\u901f\u5916\u7684\u5168\u90e8\u52a0\u901f\u529f\u80fd(layout/\u516c\u5f0f/OCR)\u7684\u663e\u5b58\u9700\u6c42\u4ece16GB\u964d\u81f38GB\uff0c\u5f00\u542f\u5168\u90e8\u52a0\u901f\u529f\u80fd\u7684\u663e\u5b58\u9700\u6c42\u4ece24GB\u964d\u81f310GB
\u4f18\u5316\u914d\u7f6e\u6587\u4ef6\u7684\u529f\u80fd\u5f00\u5173\uff0c\u589e\u52a0\u72ec\u7acb\u7684\u516c\u5f0f\u68c0\u6d4b\u5f00\u5173\uff0c\u65e0\u9700\u516c\u5f0f\u68c0\u6d4b\u65f6\u53ef\u5927\u5e45\u63d0\u5347\u901f\u5ea6\u548c\u89e3\u6790\u6548\u679c
\u96c6\u6210 PDF-Extract-Kit 1.0
\u52a0\u5165\u81ea\u7814\u7684 doclayout_yolo \u6a21\u578b\uff0c\u5728\u76f8\u8fd1\u89e3\u6790\u6548\u679c\u60c5\u51b5\u4e0b\u6bd4\u539f\u65b9\u6848\u63d0\u901f10\u500d\u4ee5\u4e0a\uff0c\u53ef\u901a\u8fc7\u914d\u7f6e\u6587\u4ef6\u4e0e layoutlmv3 \u81ea\u7531\u5207\u6362
\u516c\u5f0f\u89e3\u6790\u5347\u7ea7\u81f3 unimernet 0.2.1\uff0c\u5728\u63d0\u5347\u516c\u5f0f\u89e3\u6790\u51c6\u786e\u7387\u7684\u540c\u65f6\uff0c\u5927\u5e45\u964d\u4f4e\u663e\u5b58\u9700\u6c42
\u56e0 PDF-Extract-Kit 1.0 \u66f4\u6362\u4ed3\u5e93\uff0c\u9700\u8981\u91cd\u65b0\u4e0b\u8f7d\u6a21\u578b\uff0c\u6b65\u9aa4\u8be6\u89c1 \u5982\u4f55\u4e0b\u8f7d\u6a21\u578b

"},{"location":"zh/reference/changelog/#081-20240927","title":"0.8.1 (2024/09/27)","text":"

\u4fee\u590d\u4e86\u4e00\u4e9bbug\uff0c\u540c\u65f6\u63d0\u4f9b\u4e86\u5728\u7ebfdemo\u7684\u672c\u5730\u5316\u90e8\u7f72\u7248\u672c\u548c\u524d\u7aef\u754c\u9762

"},{"location":"zh/reference/changelog/#080-20240909","title":"0.8.0 (2024/09/09)","text":"

\u652f\u6301Dockerfile\u5feb\u901f\u90e8\u7f72\uff0c\u540c\u65f6\u4e0a\u7ebf\u4e86huggingface\u3001modelscope demo

"},{"location":"zh/reference/changelog/#071-20240830","title":"0.7.1 (2024/08/30)","text":"

\u96c6\u6210\u4e86paddle tablemaster\u8868\u683c\u8bc6\u522b\u529f\u80fd

"},{"location":"zh/reference/changelog/#070b1-20240809","title":"0.7.0b1 (2024/08/09)","text":"

\u7b80\u5316\u5b89\u88c5\u6b65\u9aa4\u63d0\u5347\u6613\u7528\u6027\uff0c\u52a0\u5165\u8868\u683c\u8bc6\u522b\u529f\u80fd

"},{"location":"zh/reference/changelog/#062b1-20240801","title":"0.6.2b1 (2024/08/01)","text":"

\u4f18\u5316\u4e86\u4f9d\u8d56\u51b2\u7a81\u95ee\u9898\u548c\u5b89\u88c5\u6587\u6863

"},{"location":"zh/reference/changelog/#20240705","title":"\u9996\u6b21\u5f00\u6e90 (2024/07/05)","text":"

MinerU\u9879\u76ee\u9996\u6b21\u5f00\u6e90\u53d1\u5e03

"},{"location":"zh/reference/output_files/","title":"\u8f93\u51fa\u6587\u4ef6\u683c\u5f0f","text":""},{"location":"zh/reference/output_files/#mineru","title":"MinerU \u8f93\u51fa\u6587\u4ef6\u8bf4\u660e","text":""},{"location":"zh/reference/output_files/#_1","title":"\u6982\u89c8","text":"

mineru \u547d\u4ee4\u6267\u884c\u540e\uff0c\u9664\u4e86\u8f93\u51fa\u4e3b\u8981\u7684 markdown \u6587\u4ef6\u5916\uff0c\u8fd8\u4f1a\u751f\u6210\u591a\u4e2a\u8f85\u52a9\u6587\u4ef6\u7528\u4e8e\u8c03\u8bd5\u3001\u8d28\u68c0\u548c\u8fdb\u4e00\u6b65\u5904\u7406\u3002\u8fd9\u4e9b\u6587\u4ef6\u5305\u62ec\uff1a

\u53ef\u89c6\u5316\u8c03\u8bd5\u6587\u4ef6\uff1a\u5e2e\u52a9\u7528\u6237\u76f4\u89c2\u4e86\u89e3\u6587\u6863\u89e3\u6790\u8fc7\u7a0b\u548c\u7ed3\u679c
\u7ed3\u6784\u5316\u6570\u636e\u6587\u4ef6\uff1a\u5305\u542b\u8be6\u7ec6\u7684\u89e3\u6790\u6570\u636e\uff0c\u53ef\u7528\u4e8e\u4e8c\u6b21\u5f00\u53d1

\u4e0b\u9762\u5c06\u8be6\u7ec6\u4ecb\u7ecd\u6bcf\u4e2a\u6587\u4ef6\u7684\u4f5c\u7528\u548c\u683c\u5f0f\u3002

"},{"location":"zh/reference/output_files/#_2","title":"\u53ef\u89c6\u5316\u8c03\u8bd5\u6587\u4ef6","text":""},{"location":"zh/reference/output_files/#layoutpdf","title":"\u5e03\u5c40\u5206\u6790\u6587\u4ef6 (layout.pdf)","text":"

\u6587\u4ef6\u547d\u540d\u683c\u5f0f\uff1a{\u539f\u6587\u4ef6\u540d}_layout.pdf

\u529f\u80fd\u8bf4\u660e\uff1a

\u53ef\u89c6\u5316\u5c55\u793a\u6bcf\u4e00\u9875\u7684\u5e03\u5c40\u5206\u6790\u7ed3\u679c
\u6bcf\u4e2a\u68c0\u6d4b\u6846\u53f3\u4e0a\u89d2\u7684\u6570\u5b57\u8868\u793a\u9605\u8bfb\u987a\u5e8f
\u4f7f\u7528\u4e0d\u540c\u80cc\u666f\u8272\u5757\u533a\u5206\u4e0d\u540c\u7c7b\u578b\u7684\u5185\u5bb9\u5757

\u4f7f\u7528\u573a\u666f\uff1a

\u68c0\u67e5\u5e03\u5c40\u5206\u6790\u662f\u5426\u6b63\u786e
\u786e\u8ba4\u9605\u8bfb\u987a\u5e8f\u662f\u5426\u5408\u7406
\u8c03\u8bd5\u5e03\u5c40\u76f8\u5173\u95ee\u9898

"},{"location":"zh/reference/output_files/#spanpdf","title":"\u6587\u672c\u7247\u6bb5\u6587\u4ef6 (span.pdf)","text":"

Note

\u4ec5\u9002\u7528\u4e8e pipeline \u540e\u7aef

\u6587\u4ef6\u547d\u540d\u683c\u5f0f\uff1a{\u539f\u6587\u4ef6\u540d}_span.pdf

\u529f\u80fd\u8bf4\u660e\uff1a

\u6839\u636e span \u7c7b\u578b\u4f7f\u7528\u4e0d\u540c\u989c\u8272\u7ebf\u6846\u6807\u6ce8\u9875\u9762\u5185\u5bb9
\u7528\u4e8e\u8d28\u91cf\u68c0\u67e5\u548c\u95ee\u9898\u6392\u67e5

\u4f7f\u7528\u573a\u666f\uff1a

\u5feb\u901f\u6392\u67e5\u6587\u672c\u4e22\u5931\u95ee\u9898
\u68c0\u67e5\u884c\u5185\u516c\u5f0f\u8bc6\u522b\u60c5\u51b5
\u9a8c\u8bc1\u6587\u672c\u5206\u5272\u51c6\u786e\u6027

"},{"location":"zh/reference/output_files/#_3","title":"\u7ed3\u6784\u5316\u6570\u636e\u6587\u4ef6","text":"

Important

2.5\u7248\u672cvlm\u540e\u7aef\u7684\u8f93\u51fa\u5b58\u5728\u8f83\u5927\u53d8\u5316\uff0c\u4e0epipeline\u7248\u672c\u5b58\u5728\u4e0d\u517c\u5bb9\u60c5\u51b5\uff0c\u5982\u9700\u57fa\u4e8e\u7ed3\u6784\u5316\u8f93\u51fa\u8fdb\u884c\u4e8c\u6b21\u5f00\u53d1\uff0c\u8bf7\u4ed4\u7ec6\u9605\u8bfb\u672c\u6587\u6863\u5185\u5bb9\u3002

"},{"location":"zh/reference/output_files/#pipeline","title":"pipeline \u540e\u7aef \u8f93\u51fa\u7ed3\u679c","text":""},{"location":"zh/reference/output_files/#modeljson","title":"\u6a21\u578b\u63a8\u7406\u7ed3\u679c (model.json)","text":"

\u6587\u4ef6\u547d\u540d\u683c\u5f0f\uff1a{\u539f\u6587\u4ef6\u540d}_model.json

"},{"location":"zh/reference/output_files/#_4","title":"\u6570\u636e\u7ed3\u6784\u5b9a\u4e49","text":"

from pydantic import BaseModel, Field\nfrom enum import IntEnum\n\nclass CategoryType(IntEnum):\n    \"\"\"\u5185\u5bb9\u7c7b\u522b\u679a\u4e3e\"\"\"\n    title = 0               # \u6807\u9898\n    plain_text = 1          # \u6587\u672c\n    abandon = 2             # \u5305\u62ec\u9875\u7709\u9875\u811a\u9875\u7801\u548c\u9875\u9762\u6ce8\u91ca\n    figure = 3              # \u56fe\u7247\n    figure_caption = 4      # \u56fe\u7247\u63cf\u8ff0\n    table = 5               # \u8868\u683c\n    table_caption = 6       # \u8868\u683c\u63cf\u8ff0\n    table_footnote = 7      # \u8868\u683c\u6ce8\u91ca\n    isolate_formula = 8     # \u884c\u95f4\u516c\u5f0f\n    formula_caption = 9     # \u884c\u95f4\u516c\u5f0f\u7684\u6807\u53f7\n    embedding = 13          # \u884c\u5185\u516c\u5f0f\n    isolated = 14           # \u884c\u95f4\u516c\u5f0f\n    text = 15               # OCR \u8bc6\u522b\u7ed3\u679c\n\nclass PageInfo(BaseModel):\n    \"\"\"\u9875\u9762\u4fe1\u606f\"\"\"\n    page_no: int = Field(description=\"\u9875\u7801\u5e8f\u53f7\uff0c\u7b2c\u4e00\u9875\u7684\u5e8f\u53f7\u662f 0\", ge=0)\n    height: int = Field(description=\"\u9875\u9762\u9ad8\u5ea6\", gt=0)\n    width: int = Field(description=\"\u9875\u9762\u5bbd\u5ea6\", ge=0)\n\nclass ObjectInferenceResult(BaseModel):\n    \"\"\"\u5bf9\u8c61\u8bc6\u522b\u7ed3\u679c\"\"\"\n    category_id: CategoryType = Field(description=\"\u7c7b\u522b\", ge=0)\n    poly: list[float] = Field(description=\"\u56db\u8fb9\u5f62\u5750\u6807\uff0c\u683c\u5f0f\u4e3a [x0,y0,x1,y1,x2,y2,x3,y3]\")\n    score: float = Field(description=\"\u63a8\u7406\u7ed3\u679c\u7684\u7f6e\u4fe1\u5ea6\")\n    latex: str | None = Field(description=\"LaTeX \u89e3\u6790\u7ed3\u679c\", default=None)\n    html: str | None = Field(description=\"HTML \u89e3\u6790\u7ed3\u679c\", default=None)\n\nclass PageInferenceResults(BaseModel):\n    \"\"\"\u9875\u9762\u63a8\u7406\u7ed3\u679c\"\"\"\n    layout_dets: list[ObjectInferenceResult] = Field(description=\"\u9875\u9762\u8bc6\u522b\u7ed3\u679c\")\n    page_info: PageInfo = Field(description=\"\u9875\u9762\u5143\u4fe1\u606f\")\n\n# \u5b8c\u6574\u7684\u63a8\u7406\u7ed3\u679c\ninference_result: list[PageInferenceResults] = []\n

"},{"location":"zh/reference/output_files/#_5","title":"\u5750\u6807\u7cfb\u7edf\u8bf4\u660e","text":"

poly \u5750\u6807\u683c\u5f0f\uff1a[x0, y0, x1, y1, x2, y2, x3, y3]

\u5206\u522b\u8868\u793a\u5de6\u4e0a\u3001\u53f3\u4e0a\u3001\u53f3\u4e0b\u3001\u5de6\u4e0b\u56db\u70b9\u7684\u5750\u6807
\u5750\u6807\u539f\u70b9\u5728\u9875\u9762\u5de6\u4e0a\u89d2

"},{"location":"zh/reference/output_files/#_6","title":"\u793a\u4f8b\u6570\u636e","text":"

[\n    {\n        \"layout_dets\": [\n            {\n                \"category_id\": 2,\n                \"poly\": [\n                    99.1906967163086,\n                    100.3119125366211,\n                    730.3707885742188,\n                    100.3119125366211,\n                    730.3707885742188,\n                    245.81326293945312,\n                    99.1906967163086,\n                    245.81326293945312\n                ],\n                \"score\": 0.9999997615814209\n            }\n        ],\n        \"page_info\": {\n            \"page_no\": 0,\n            \"height\": 2339,\n            \"width\": 1654\n        }\n    },\n    {\n        \"layout_dets\": [\n            {\n                \"category_id\": 5,\n                \"poly\": [\n                    99.13092803955078,\n                    2210.680419921875,\n                    497.3183898925781,\n                    2210.680419921875,\n                    497.3183898925781,\n                    2264.78076171875,\n                    99.13092803955078,\n                    2264.78076171875\n                ],\n                \"score\": 0.9999997019767761\n            }\n        ],\n        \"page_info\": {\n            \"page_no\": 1,\n            \"height\": 2339,\n            \"width\": 1654\n        }\n    }\n]\n

"},{"location":"zh/reference/output_files/#middlejson","title":"\u4e2d\u95f4\u5904\u7406\u7ed3\u679c (middle.json)","text":"

\u6587\u4ef6\u547d\u540d\u683c\u5f0f\uff1a{\u539f\u6587\u4ef6\u540d}_middle.json

"},{"location":"zh/reference/output_files/#_7","title":"\u9876\u5c42\u7ed3\u6784","text":"\u5b57\u6bb5\u540d \u7c7b\u578b \u8bf4\u660e pdf_info list[dict] \u6bcf\u4e00\u9875\u7684\u89e3\u6790\u7ed3\u679c\u6570\u7ec4 _backend string \u89e3\u6790\u6a21\u5f0f\uff1apipeline \u6216 vlm _version_name string MinerU \u7248\u672c\u53f7"},{"location":"zh/reference/output_files/#pdf_info","title":"\u9875\u9762\u4fe1\u606f\u7ed3\u6784 (pdf_info)","text":"\u5b57\u6bb5\u540d \u8bf4\u660e preproc_blocks PDF \u9884\u5904\u7406\u540e\u7684\u672a\u5206\u6bb5\u4e2d\u95f4\u7ed3\u679c page_idx \u9875\u7801\uff0c\u4ece 0 \u5f00\u59cb page_size \u9875\u9762\u7684\u5bbd\u5ea6\u548c\u9ad8\u5ea6 [width, height] images \u56fe\u7247\u5757\u4fe1\u606f\u5217\u8868 tables \u8868\u683c\u5757\u4fe1\u606f\u5217\u8868 interline_equations \u884c\u95f4\u516c\u5f0f\u5757\u4fe1\u606f\u5217\u8868 discarded_blocks \u9700\u8981\u4e22\u5f03\u7684\u5757\u4fe1\u606f para_blocks \u5206\u6bb5\u540e\u7684\u5185\u5bb9\u5757\u7ed3\u679c"},{"location":"zh/reference/output_files/#_8","title":"\u5757\u7ed3\u6784\u5c42\u6b21","text":"

\u4e00\u7ea7\u5757 (table | image)\n\u2514\u2500\u2500 \u4e8c\u7ea7\u5757\n    \u2514\u2500\u2500 \u884c (line)\n        \u2514\u2500\u2500 \u7247\u6bb5 (span)\n

"},{"location":"zh/reference/output_files/#_9","title":"\u4e00\u7ea7\u5757\u5b57\u6bb5","text":"\u5b57\u6bb5\u540d \u8bf4\u660e type \u5757\u7c7b\u578b\uff1atable \u6216 image bbox \u5757\u7684\u77e9\u5f62\u6846\u5750\u6807 [x0, y0, x1, y1] blocks \u5305\u542b\u7684\u4e8c\u7ea7\u5757\u5217\u8868"},{"location":"zh/reference/output_files/#_10","title":"\u4e8c\u7ea7\u5757\u5b57\u6bb5","text":"\u5b57\u6bb5\u540d \u8bf4\u660e type \u5757\u7c7b\u578b\uff08\u8be6\u89c1\u4e0b\u8868\uff09 bbox \u5757\u7684\u77e9\u5f62\u6846\u5750\u6807 lines \u5305\u542b\u7684\u884c\u4fe1\u606f\u5217\u8868"},{"location":"zh/reference/output_files/#_11","title":"\u4e8c\u7ea7\u5757\u7c7b\u578b","text":"\u7c7b\u578b \u8bf4\u660e image_body \u56fe\u50cf\u672c\u4f53 image_caption \u56fe\u50cf\u63cf\u8ff0\u6587\u672c image_footnote \u56fe\u50cf\u811a\u6ce8 table_body \u8868\u683c\u672c\u4f53 table_caption \u8868\u683c\u63cf\u8ff0\u6587\u672c table_footnote \u8868\u683c\u811a\u6ce8 text \u6587\u672c\u5757 title \u6807\u9898\u5757 index \u76ee\u5f55\u5757 list \u5217\u8868\u5757 interline_equation \u884c\u95f4\u516c\u5f0f\u5757"},{"location":"zh/reference/output_files/#_12","title":"\u884c\u548c\u7247\u6bb5\u7ed3\u6784","text":"

\u884c (line) \u5b57\u6bb5\uff1a - bbox\uff1a\u884c\u7684\u77e9\u5f62\u6846\u5750\u6807 - spans\uff1a\u5305\u542b\u7684\u7247\u6bb5\u5217\u8868

\u7247\u6bb5 (span) \u5b57\u6bb5\uff1a - bbox\uff1a\u7247\u6bb5\u7684\u77e9\u5f62\u6846\u5750\u6807 - type\uff1a\u7247\u6bb5\u7c7b\u578b\uff08image\u3001table\u3001text\u3001inline_equation\u3001interline_equation\uff09 - content | img_path\uff1a\u6587\u672c\u5185\u5bb9\u6216\u56fe\u7247\u8def\u5f84

"},{"location":"zh/reference/output_files/#_13","title":"\u793a\u4f8b\u6570\u636e","text":"

{\n    \"pdf_info\": [\n        {\n            \"preproc_blocks\": [\n                {\n                    \"type\": \"text\",\n                    \"bbox\": [\n                        52,\n                        61.956024169921875,\n                        294,\n                        82.99800872802734\n                    ],\n                    \"lines\": [\n                        {\n                            \"bbox\": [\n                                52,\n                                61.956024169921875,\n                                294,\n                                72.0000228881836\n                            ],\n                            \"spans\": [\n                                {\n                                    \"bbox\": [\n                                        54.0,\n                                        61.956024169921875,\n                                        296.2261657714844,\n                                        72.0000228881836\n                                    ],\n                                    \"content\": \"dependent on the service headway and the reliability of the departure \",\n                                    \"type\": \"text\",\n                                    \"score\": 1.0\n                                }\n                            ]\n                        }\n                    ]\n                }\n            ],\n            \"layout_bboxes\": [\n                {\n                    \"layout_bbox\": [\n                        52,\n                        61,\n                        294,\n                        731\n                    ],\n                    \"layout_label\": \"V\",\n                    \"sub_layout\": []\n                }\n            ],\n            \"page_idx\": 0,\n            \"page_size\": [\n                612.0,\n                792.0\n            ],\n            \"_layout_tree\": [],\n            \"images\": [],\n            \"tables\": [],\n            \"interline_equations\": [],\n            \"discarded_blocks\": [],\n            \"para_blocks\": [\n                {\n                    \"type\": \"text\",\n                    \"bbox\": [\n                        52,\n                        61.956024169921875,\n                        294,\n                        82.99800872802734\n                    ],\n                    \"lines\": [\n                        {\n                            \"bbox\": [\n                                52,\n                                61.956024169921875,\n                                294,\n                                72.0000228881836\n                            ],\n                            \"spans\": [\n                                {\n                                    \"bbox\": [\n                                        54.0,\n                                        61.956024169921875,\n                                        296.2261657714844,\n                                        72.0000228881836\n                                    ],\n                                    \"content\": \"dependent on the service headway and the reliability of the departure \",\n                                    \"type\": \"text\",\n                                    \"score\": 1.0\n                                }\n                            ]\n                        }\n                    ]\n                }\n            ]\n        }\n    ],\n    \"_backend\": \"pipeline\",\n    \"_version_name\": \"0.6.1\"\n}\n

"},{"location":"zh/reference/output_files/#content_listjson","title":"\u5185\u5bb9\u5217\u8868 (content_list.json)","text":"

\u6587\u4ef6\u547d\u540d\u683c\u5f0f\uff1a{\u539f\u6587\u4ef6\u540d}_content_list.json

"},{"location":"zh/reference/output_files/#_14","title":"\u529f\u80fd\u8bf4\u660e","text":"

\u8fd9\u662f\u4e00\u4e2a\u7b80\u5316\u7248\u7684 middle.json\uff0c\u6309\u9605\u8bfb\u987a\u5e8f\u5e73\u94fa\u5b58\u50a8\u6240\u6709\u53ef\u8bfb\u5185\u5bb9\u5757\uff0c\u53bb\u9664\u4e86\u590d\u6742\u7684\u5e03\u5c40\u4fe1\u606f\uff0c\u4fbf\u4e8e\u540e\u7eed\u5904\u7406\u3002

"},{"location":"zh/reference/output_files/#_15","title":"\u5185\u5bb9\u7c7b\u578b","text":"\u7c7b\u578b \u8bf4\u660e image \u56fe\u7247 table \u8868\u683c text \u6587\u672c/\u6807\u9898 equation \u884c\u95f4\u516c\u5f0f"},{"location":"zh/reference/output_files/#_16","title":"\u6587\u672c\u5c42\u7ea7\u6807\u8bc6","text":"

\u901a\u8fc7 text_level \u5b57\u6bb5\u533a\u5206\u6587\u672c\u5c42\u7ea7\uff1a

\u65e0 text_level \u6216 text_level: 0\uff1a\u6b63\u6587\u6587\u672c
text_level: 1\uff1a\u4e00\u7ea7\u6807\u9898
text_level: 2\uff1a\u4e8c\u7ea7\u6807\u9898
\u4ee5\u6b64\u7c7b\u63a8...

"},{"location":"zh/reference/output_files/#_17","title":"\u901a\u7528\u5b57\u6bb5","text":"

\u6240\u6709\u5185\u5bb9\u5757\u90fd\u5305\u542b page_idx \u5b57\u6bb5\uff0c\u8868\u793a\u6240\u5728\u9875\u7801\uff08\u4ece 0 \u5f00\u59cb\uff09\u3002
\u6240\u6709\u5185\u5bb9\u5757\u90fd\u5305\u542b bbox \u5b57\u6bb5\uff0c\u8868\u793a\u5185\u5bb9\u5757\u7684\u8fb9\u754c\u6846\u5750\u6807 [x0, y0, x1, y1] \u6620\u5c04\u57280-1000\u8303\u56f4\u5185\u7684\u7ed3\u679c\u3002

"},{"location":"zh/reference/output_files/#_18","title":"\u793a\u4f8b\u6570\u636e","text":"

[\n        {\n        \"type\": \"text\",\n        \"text\": \"The response of flow duration curves to afforestation \",\n        \"text_level\": 1, \n        \"bbox\": [\n            62,\n            480,\n            946,\n            904\n        ],\n        \"page_idx\": 0\n    },\n    {\n        \"type\": \"image\",\n        \"img_path\": \"images/a8ecda1c69b27e4f79fce1589175a9d721cbdc1cf78b4cc06a015f3746f6b9d8.jpg\",\n        \"image_caption\": [\n            \"Fig. 1. Annual flow duration curves of daily flows from Pine Creek, Australia, 1989\u20132000. \"\n        ],\n        \"image_footnote\": [],\n        \"bbox\": [\n            62,\n            480,\n            946,\n            904\n        ],\n        \"page_idx\": 1\n    },\n    {\n        \"type\": \"equation\",\n        \"img_path\": \"images/181ea56ef185060d04bf4e274685f3e072e922e7b839f093d482c29bf89b71e8.jpg\",\n        \"text\": \"$$\\nQ _ { \\\\% } = f ( P ) + g ( T )\\n$$\",\n        \"text_format\": \"latex\",\n        \"bbox\": [\n            62,\n            480,\n            946,\n            904\n        ],\n        \"page_idx\": 2\n    },\n    {\n        \"type\": \"table\",\n        \"img_path\": \"images/e3cb413394a475e555807ffdad913435940ec637873d673ee1b039e3bc3496d0.jpg\",\n        \"table_caption\": [\n            \"Table 2 Significance of the rainfall and time terms \"\n        ],\n        \"table_footnote\": [\n            \"indicates that the rainfall term was significant at the $5 \\\\%$ level, $T$ indicates that the time term was significant at the $5 \\\\%$ level, \\\\* represents significance at the $10 \\\\%$ level, and na denotes too few data points for meaningful analysis. \"\n        ],\n        \"table_body\": \"<html><body><table><tr><td rowspan=\\\"2\\\">Site</td><td colspan=\\\"10\\\">Percentile</td></tr><tr><td>10</td><td>20</td><td>30</td><td>40</td><td>50</td><td>60</td><td>70</td><td>80</td><td>90</td><td>100</td></tr><tr><td>Traralgon Ck</td><td>P</td><td>P,*</td><td>P</td><td>P</td><td>P,</td><td>P,</td><td>P,</td><td>P,</td><td>P</td><td>P</td></tr><tr><td>Redhill</td><td>P,T</td><td>P,T</td><td>\uff0c*</td><td>**</td><td>P.T</td><td>P,*</td><td>P*</td><td>P*</td><td>*</td><td>\uff0c*</td></tr><tr><td>Pine Ck</td><td></td><td>P,T</td><td>P,T</td><td>P,T</td><td>P,T</td><td>T</td><td>T</td><td>T</td><td>na</td><td>na</td></tr><tr><td>Stewarts Ck 5</td><td>P,T</td><td>P,T</td><td>P,T</td><td>P,T</td><td>P.T</td><td>P.T</td><td>P,T</td><td>na</td><td>na</td><td>na</td></tr><tr><td>Glendhu 2</td><td>P</td><td>P,T</td><td>P,*</td><td>P,T</td><td>P.T</td><td>P,ns</td><td>P,T</td><td>P,T</td><td>P,T</td><td>P,T</td></tr><tr><td>Cathedral Peak 2</td><td>P,T</td><td>P,T</td><td>P,T</td><td>P,T</td><td>P,T</td><td>*,T</td><td>P,T</td><td>P,T</td><td>P,T</td><td>T</td></tr><tr><td>Cathedral Peak 3</td><td>P.T</td><td>P.T</td><td>P,T</td><td>P,T</td><td>P,T</td><td>T</td><td>P,T</td><td>P,T</td><td>P,T</td><td>T</td></tr><tr><td>Lambrechtsbos A</td><td>P,T</td><td>P</td><td>P</td><td>P,T</td><td>*,T</td><td>*,T</td><td>*,T</td><td>*,T</td><td>*,T</td><td>T</td></tr><tr><td>Lambrechtsbos B</td><td>P,T</td><td>P,T</td><td>P,T</td><td>P,T</td><td>P,T</td><td>P,T</td><td>P,T</td><td>P,T</td><td>T</td><td>T</td></tr><tr><td>Biesievlei</td><td>P,T</td><td>P.T</td><td>P,T</td><td>P,T</td><td>*,T</td><td>*,T</td><td>T</td><td>T</td><td>P,T</td><td>P,T</td></tr></table></body></html>\",\n        \"bbox\": [\n            62,\n            480,\n            946,\n            904\n        ],  \n        \"page_idx\": 5\n    }\n]\n

"},{"location":"zh/reference/output_files/#vlm","title":"VLM \u540e\u7aef \u8f93\u51fa\u7ed3\u679c","text":""},{"location":"zh/reference/output_files/#modeljson_1","title":"\u6a21\u578b\u63a8\u7406\u7ed3\u679c (model.json)","text":"

\u6587\u4ef6\u547d\u540d\u683c\u5f0f\uff1a{\u539f\u6587\u4ef6\u540d}_model.json

"},{"location":"zh/reference/output_files/#_19","title":"\u6587\u4ef6\u683c\u5f0f\u8bf4\u660e","text":"

\u8be5\u6587\u4ef6\u4e3a VLM \u6a21\u578b\u7684\u539f\u59cb\u8f93\u51fa\u7ed3\u679c\uff0c\u5305\u542b\u4e24\u5c42\u5d4c\u5957list\uff0c\u5916\u5c42\u8868\u793a\u9875\u9762\uff0c\u5185\u5c42\u8868\u793a\u8be5\u9875\u7684\u5185\u5bb9\u5757
\u6bcf\u4e2a\u5185\u5bb9\u5757\u90fd\u662f\u4e00\u4e2adict\uff0c\u5305\u542b type\u3001bbox\u3001angle\u3001content \u5b57\u6bb5

"},{"location":"zh/reference/output_files/#_20","title":"\u652f\u6301\u7684\u5185\u5bb9\u7c7b\u578b","text":"

{\n    \"text\": \"\u6587\u672c\",\n    \"title\": \"\u6807\u9898\", \n    \"equation\": \"\u884c\u95f4\u516c\u5f0f\",\n    \"image\": \"\u56fe\u7247\",\n    \"image_caption\": \"\u56fe\u7247\u63cf\u8ff0\",\n    \"image_footnote\": \"\u56fe\u7247\u811a\u6ce8\",\n    \"table\": \"\u8868\u683c\",\n    \"table_caption\": \"\u8868\u683c\u63cf\u8ff0\",\n    \"table_footnote\": \"\u8868\u683c\u811a\u6ce8\",\n    \"phonetic\": \"\u62fc\u97f3\",\n    \"code\": \"\u4ee3\u7801\u5757\",\n    \"code_caption\": \"\u4ee3\u7801\u63cf\u8ff0\",\n    \"ref_text\": \"\u53c2\u8003\u6587\u732e\",\n    \"algorithm\": \"\u7b97\u6cd5\u5757\",\n    \"list\": \"\u5217\u8868\",\n    \"header\": \"\u9875\u7709\",\n    \"footer\": \"\u9875\u811a\",\n    \"page_number\": \"\u9875\u7801\",\n    \"aside_text\": \"\u88c5\u8ba2\u7ebf\u65c1\u6ce8\", \n    \"page_footnote\": \"\u9875\u9762\u811a\u6ce8\"\n}\n

"},{"location":"zh/reference/output_files/#_21","title":"\u5750\u6807\u7cfb\u7edf\u8bf4\u660e","text":"

bbox \u5750\u6807\u683c\u5f0f\uff1a[x0, y0, x1, y1]

\u5206\u522b\u8868\u793a\u5de6\u4e0a\u3001\u53f3\u4e0b\u4e24\u70b9\u7684\u5750\u6807
\u5750\u6807\u539f\u70b9\u5728\u9875\u9762\u5de6\u4e0a\u89d2
\u5750\u6807\u4e3a\u76f8\u5bf9\u4e8e\u539f\u59cb\u9875\u9762\u5c3a\u5bf8\u7684\u767e\u5206\u6bd4\uff0c\u8303\u56f4\u57280-1\u4e4b\u95f4

"},{"location":"zh/reference/output_files/#_22","title":"\u793a\u4f8b\u6570\u636e","text":"

[\n    [\n        {\n            \"type\": \"header\",\n            \"bbox\": [\n                0.077,\n                0.095,\n                0.18,\n                0.181\n            ],\n            \"angle\": 0,\n            \"score\": null,\n            \"block_tags\": null,\n            \"content\": \"ELSEVIER\",\n            \"format\": null,\n            \"content_tags\": null\n        },\n        {\n            \"type\": \"title\",\n            \"bbox\": [\n                0.157,\n                0.228,\n                0.833,\n                0.253\n            ],\n            \"angle\": 0,\n            \"score\": null,\n            \"block_tags\": null,\n            \"content\": \"The response of flow duration curves to afforestation\",\n            \"format\": null,\n            \"content_tags\": null\n        }\n    ]\n]\n

"},{"location":"zh/reference/output_files/#middlejson_1","title":"\u4e2d\u95f4\u5904\u7406\u7ed3\u679c (middle.json)","text":"

\u6587\u4ef6\u547d\u540d\u683c\u5f0f\uff1a{\u539f\u6587\u4ef6\u540d}_middle.json

"},{"location":"zh/reference/output_files/#_23","title":"\u6587\u4ef6\u683c\u5f0f\u8bf4\u660e","text":"

vlm \u540e\u7aef\u7684 middle.json \u6587\u4ef6\u7ed3\u6784\u4e0e pipeline \u540e\u7aef\u7c7b\u4f3c\uff0c\u4f46\u5b58\u5728\u4ee5\u4e0b\u5dee\u5f02\uff1a

list\u53d8\u6210\u4e8c\u7ea7block\uff0c\u589e\u52a0sub_type\u5b57\u6bb5\u533a\u5206list\u7c7b\u578b:
- text\uff08\u6587\u672c\u7c7b\u578b\uff09
- ref_text\uff08\u5f15\u7528\u7c7b\u578b\uff09
\u589e\u52a0code\u7c7b\u578bblock\uff0ccode\u7c7b\u578b\u5305\u542b\u4e24\u79cd\"sub_type\":
- \u5206\u522b\u662fcode\u548calgorithm
- \u81f3\u5c11\u6709code_body, \u53ef\u9009code_caption
discarded_blocks\u5185\u5143\u7d20type\u589e\u52a0\u4ee5\u4e0b\u7c7b\u578b:
- header\uff08\u9875\u7709\uff09
- footer\uff08\u9875\u811a\uff09
- page_number\uff08\u9875\u7801\uff09
- aside_text\uff08\u88c5\u8ba2\u7ebf\u6587\u672c\uff09
- page_footnote\uff08\u811a\u6ce8\uff09
\u6240\u6709block\u589e\u52a0angle\u5b57\u6bb5\uff0c\u7528\u6765\u8868\u793a\u65cb\u8f6c\u89d2\u5ea6\uff0c0\uff0c90\uff0c180\uff0c270

"},{"location":"zh/reference/output_files/#_24","title":"\u793a\u4f8b\u6570\u636e","text":"

list block \u793a\u4f8b

{\n    \"bbox\": [\n        174,\n        155,\n        818,\n        333\n    ],\n    \"type\": \"list\",\n    \"angle\": 0,\n    \"index\": 11,\n    \"blocks\": [\n        {\n            \"bbox\": [\n                174,\n                157,\n                311,\n                175\n            ],\n            \"type\": \"text\",\n            \"angle\": 0,\n            \"lines\": [\n                {\n                    \"bbox\": [\n                        174,\n                        157,\n                        311,\n                        175\n                    ],\n                    \"spans\": [\n                        {\n                            \"bbox\": [\n                                174,\n                                157,\n                                311,\n                                175\n                            ],\n                            \"type\": \"text\",\n                            \"content\": \"H.1 Introduction\"\n                        }\n                    ]\n                }\n            ],\n            \"index\": 3\n        },\n        {\n            \"bbox\": [\n                175,\n                182,\n                464,\n                229\n            ],\n            \"type\": \"text\",\n            \"angle\": 0,\n            \"lines\": [\n                {\n                    \"bbox\": [\n                        175,\n                        182,\n                        464,\n                        229\n                    ],\n                    \"spans\": [\n                        {\n                            \"bbox\": [\n                                175,\n                                182,\n                                464,\n                                229\n                            ],\n                            \"type\": \"text\",\n                            \"content\": \"H.2 Example: Divide by Zero without Exception Handling\"\n                        }\n                    ]\n                }\n            ],\n            \"index\": 4\n        }\n    ],\n    \"sub_type\": \"text\"\n}\n

code block \u793a\u4f8b

{\n    \"type\": \"code\",\n    \"bbox\": [\n        114,\n        780,\n        885,\n        1231\n    ],\n    \"blocks\": [\n        {\n            \"bbox\": [\n                114,\n                780,\n                885,\n                1231\n            ],\n            \"lines\": [\n                {\n                    \"bbox\": [\n                        114,\n                        780,\n                        885,\n                        1231\n                    ],\n                    \"spans\": [\n                        {\n                            \"bbox\": [\n                                114,\n                                780,\n                                885,\n                                1231\n                            ],\n                            \"type\": \"text\",\n                            \"content\": \"1 // Fig. H.1: DivideByZeroNoExceptionHandling.java  \\n2 // Integer division without exception handling.  \\n3 import java.util.Scanner;  \\n4  \\n5 public class DivideByZeroNoExceptionHandling  \\n6 {  \\n7 // demonstrates throwing an exception when a divide-by-zero occurs  \\n8 public static int quotient( int numerator, int denominator )  \\n9 {  \\n10 return numerator / denominator; // possible division by zero  \\n11 } // end method quotient  \\n12  \\n13 public static void main(String[] args)  \\n14 {  \\n15 Scanner scanner = new Scanner(System.in); // scanner for input  \\n16  \\n17 System.out.print(\\\"Please enter an integer numerator: \\\");  \\n18 int numerator = scanner.nextInt();  \\n19 System.out.print(\\\"Please enter an integer denominator: \\\");  \\n20 int denominator = scanner.nextInt();  \\n21\"\n                        }\n                    ]\n                }\n            ],\n            \"index\": 17,\n            \"angle\": 0,\n            \"type\": \"code_body\"\n        },\n        {\n            \"bbox\": [\n                867,\n                160,\n                1280,\n                189\n            ],\n            \"lines\": [\n                {\n                    \"bbox\": [\n                        867,\n                        160,\n                        1280,\n                        189\n                    ],\n                    \"spans\": [\n                        {\n                            \"bbox\": [\n                                867,\n                                160,\n                                1280,\n                                189\n                            ],\n                            \"type\": \"text\",\n                            \"content\": \"Algorithm 1 Modules for MCTSteg\"\n                        }\n                    ]\n                }\n            ],\n            \"index\": 19,\n            \"angle\": 0,\n            \"type\": \"code_caption\"\n        }\n    ],\n    \"index\": 17,\n    \"sub_type\": \"code\"\n}\n

"},{"location":"zh/reference/output_files/#content_listjson_1","title":"\u5185\u5bb9\u5217\u8868 (content_list.json)","text":"

\u6587\u4ef6\u547d\u540d\u683c\u5f0f\uff1a{\u539f\u6587\u4ef6\u540d}_content_list.json

"},{"location":"zh/reference/output_files/#_25","title":"\u6587\u4ef6\u683c\u5f0f\u8bf4\u660e","text":"

vlm \u540e\u7aef\u7684 content_list.json \u6587\u4ef6\u7ed3\u6784\u4e0e pipeline \u540e\u7aef\u7c7b\u4f3c\uff0c\u4f34\u968f\u672c\u6b21middle.json\u7684\u53d8\u5316\uff0c\u505a\u4e86\u4ee5\u4e0b\u8c03\u6574\uff1a

\u65b0\u589ecode\u7c7b\u578b\uff0ccode\u7c7b\u578b\u5305\u542b\u4e24\u79cd\"sub_type\":
- \u5206\u522b\u662fcode\u548calgorithm
- \u81f3\u5c11\u6709code_body, \u53ef\u9009code_caption
\u65b0\u589elist\u7c7b\u578b\uff0clist\u7c7b\u578b\u5305\u542b\u4e24\u79cd\"sub_type\":
- text
- ref_text
\u589e\u52a0\u6240\u6709\u6240\u6709discarded_blocks\u7684\u8f93\u51fa\u5185\u5bb9
- header
- footer
- page_number
- aside_text
- page_footnote

"},{"location":"zh/reference/output_files/#_26","title":"\u793a\u4f8b\u6570\u636e","text":"

code \u7c7b\u578b content

{\n    \"type\": \"code\",\n    \"sub_type\": \"algorithm\",\n    \"code_caption\": [\n        \"Algorithm 1 Modules for MCTSteg\"\n    ],\n    \"code_body\": \"1: function GETCOORDINATE(d)  \\n2:  $x \\\\gets d / l$ ,  $y \\\\gets d$  mod  $l$   \\n3: return  $(x, y)$   \\n4: end function  \\n5: function BESTCHILD(v)  \\n6:  $C \\\\gets$  child set of  $v$   \\n7:  $v' \\\\gets \\\\arg \\\\max_{c \\\\in C} \\\\mathrm{UCTScore}(c)$   \\n8:  $v'.n \\\\gets v'.n + 1$   \\n9: return  $v'$   \\n10: end function  \\n11: function BACK PROPAGATE(v)  \\n12: Calculate  $R$  using Equation 11  \\n13: while  $v$  is not a root node do  \\n14:  $v.r \\\\gets v.r + R$ ,  $v \\\\gets v.p$   \\n15: end while  \\n16: end function  \\n17: function RANDOMSEARCH(v)  \\n18: while  $v$  is not a leaf node do  \\n19: Randomly select an untried action  $a \\\\in A(v)$   \\n20: Create a new node  $v'$   \\n21:  $(x, y) \\\\gets \\\\mathrm{GETCOORDINATE}(v'.d)$   \\n22:  $v'.p \\\\gets v$ ,  $v'.d \\\\gets v.d + 1$ ,  $v'.\\\\Gamma \\\\gets v.\\\\Gamma$   \\n23:  $v'.\\\\gamma_{x,y} \\\\gets a$   \\n24: if  $a = -1$  then  \\n25:  $v.lc \\\\gets v'$   \\n26: else if  $a = 0$  then  \\n27:  $v.mc \\\\gets v'$   \\n28: else  \\n29:  $v.rc \\\\gets v'$   \\n30: end if  \\n31:  $v \\\\gets v'$   \\n32: end while  \\n33: return  $v$   \\n34: end function  \\n35: function SEARCH(v)  \\n36: while  $v$  is fully expanded do  \\n37:  $v \\\\gets$  BESTCHILD(v)  \\n38: end while  \\n39: if  $v$  is not a leaf node then  \\n40:  $v \\\\gets$  RANDOMSEARCH(v)  \\n41: end if  \\n42: return  $v$   \\n43: end function\",\n    \"bbox\": [\n        510,\n        87,\n        881,\n        740\n    ],\n    \"page_idx\": 0\n}\n

list \u7c7b\u578b content

{\n    \"type\": \"list\",\n    \"sub_type\": \"text\",\n    \"list_items\": [\n        \"H.1 Introduction\",\n        \"H.2 Example: Divide by Zero without Exception Handling\",\n        \"H.3 Example: Divide by Zero with Exception Handling\",\n        \"H.4 Summary\"\n    ],\n    \"bbox\": [\n        174,\n        155,\n        818,\n        333\n    ],\n    \"page_idx\": 0\n}\n

discarded \u7c7b\u578b content

[{\n    \"type\": \"header\",\n    \"text\": \"Journal of Hydrology 310 (2005) 253-265\",\n    \"bbox\": [\n        363,\n        164,\n        623,\n        177\n    ],\n    \"page_idx\": 0\n},\n{\n    \"type\": \"page_footnote\",\n    \"text\": \"* Corresponding author. Address: Forest Science Centre, Department of Sustainability and Environment, P.O. Box 137, Heidelberg, Vic. 3084, Australia. Tel.: +61 3 9450 8719; fax: +61 3 9450 8644.\",\n    \"bbox\": [\n        71,\n        815,\n        915,\n        841\n    ],\n    \"page_idx\": 0\n}]\n

"},{"location":"zh/reference/output_files/#_27","title":"\u603b\u7ed3","text":"

\u4ee5\u4e0a\u6587\u4ef6\u4e3a MinerU \u7684\u5b8c\u6574\u8f93\u51fa\u7ed3\u679c\uff0c\u7528\u6237\u53ef\u6839\u636e\u9700\u8981\u9009\u62e9\u5408\u9002\u7684\u6587\u4ef6\u8fdb\u884c\u540e\u7eed\u5904\u7406\uff1a

\u6a21\u578b\u8f93\u51fa(\u4f7f\u7528\u539f\u59cb\u8f93\u51fa):
- model.json
\u8c03\u8bd5\u548c\u9a8c\u8bc1(\u4f7f\u7528\u53ef\u89c6\u5316\u6587\u4ef6):
- layout.pdf
- span.pdf
\u5185\u5bb9\u63d0\u53d6(\u4f7f\u7528\u7b80\u5316\u6587\u4ef6):
- *.md
- content_list.json
\u4e8c\u6b21\u5f00\u53d1(\u4f7f\u7528\u7ed3\u6784\u5316\u6587\u4ef6):
- middle.json

"},{"location":"zh/usage/","title":"\u4f7f\u7528\u6307\u5357","text":""},{"location":"zh/usage/#_1","title":"\u4f7f\u7528\u6307\u5357","text":"

\u672c\u7ae0\u8282\u63d0\u4f9b\u4e86\u9879\u76ee\u7684\u5b8c\u6574\u4f7f\u7528\u8bf4\u660e\u3002\u6211\u4eec\u5c06\u901a\u8fc7\u4ee5\u4e0b\u51e0\u4e2a\u90e8\u5206\uff0c\u5e2e\u52a9\u60a8\u4ece\u57fa\u7840\u5230\u8fdb\u9636\u9010\u6b65\u638c\u63e1\u9879\u76ee\u7684\u4f7f\u7528\u65b9\u6cd5\uff1a

"},{"location":"zh/usage/#_2","title":"\u76ee\u5f55","text":"

\u672c\u5730\u90e8\u7f72
- \u57fa\u7840\u4f7f\u7528 - \u5feb\u901f\u4e0a\u624b\u548c\u57fa\u672c\u4f7f\u7528
- \u6a21\u578b\u6e90\u914d\u7f6e - \u6a21\u578b\u6e90\u7684\u8be6\u7ec6\u914d\u7f6e\u8bf4\u660e
- \u547d\u4ee4\u884c\u5de5\u5177 - \u547d\u4ee4\u884c\u5de5\u5177\u7684\u8be6\u7ec6\u53c2\u6570\u8bf4\u660e
- \u547d\u4ee4\u884c\u8fdb\u9636\u53c2\u6570 - \u4e00\u4e9b\u9002\u914d\u547d\u4ee4\u884c\u5de5\u5177\u7684\u8fdb\u9636\u53c2\u6570\u8bf4\u660e
\u5176\u4ed6\u52a0\u901f\u5361\u9002\u914d\uff08\ud83d\ude80\u5b98\u65b9\u652f\u6301/\u2764\ufe0f\u793e\u533a\u8d21\u732e\uff09
- \u6607\u817e Ascend \ud83d\ude80
- \u5e73\u5934\u54e5 T-Head \ud83d\ude80
- \u6c90\u66e6 METAX \ud83d\ude80
- \u6d77\u5149 Hygon \ud83d\ude80
- \u71e7\u539f Enflame \ud83d\ude80
- \u6469\u5c14\u7ebf\u7a0b MooreThreads \ud83d\ude80
- \u5929\u6570\u667a\u82af IluvatarCorex \ud83d\ude80
- \u5bd2\u6b66\u7eaa Cambricon \ud83d\ude80
- \u6606\u4ed1\u82af Kunlunxin \ud83d\ude80
- \u592a\u521d\u5143\u7881 Tecorigin \u2764\ufe0f
- \u58c1\u4ede Biren \u2764\ufe0f
- AMD #3662 \u2764\ufe0f
- \u701a\u535a VastAI #4237 \u2764\ufe0f
\u63d2\u4ef6\u4e0e\u751f\u6001
- Cherry Studio
- Sider
- Dify
- n8n
- Coze
- FastGPT
- ModelWhale
- DingTalk
- DataFlow
- BISHENG
- RagFlow

"},{"location":"zh/usage/#_3","title":"\u5f00\u59cb\u4f7f\u7528","text":"

\u5efa\u8bae\u6309\u7167\u4e0a\u8ff0\u987a\u5e8f\u9605\u8bfb\u6587\u6863\uff0c\u8fd9\u6837\u53ef\u4ee5\u5e2e\u52a9\u60a8\u66f4\u597d\u5730\u7406\u89e3\u548c\u4f7f\u7528\u9879\u76ee\u529f\u80fd\u3002

\u5982\u679c\u60a8\u5728\u4f7f\u7528\u8fc7\u7a0b\u4e2d\u9047\u5230\u95ee\u9898\uff0c\u8bf7\u67e5\u770b FAQ

"},{"location":"zh/usage/advanced_cli_parameters/","title":"\u547d\u4ee4\u884c\u8fdb\u9636\u53c2\u6570","text":""},{"location":"zh/usage/advanced_cli_parameters/#_1","title":"\u547d\u4ee4\u884c\u53c2\u6570\u8fdb\u9636","text":""},{"location":"zh/usage/advanced_cli_parameters/#_2","title":"\u63a8\u7406\u5f15\u64ce\u53c2\u6570\u900f\u4f20","text":""},{"location":"zh/usage/advanced_cli_parameters/#vllm","title":"vllm \u52a0\u901f\u53c2\u6570\u4f18\u5316","text":"

Tip

\u5982\u679c\u60a8\u5df2\u7ecf\u53ef\u4ee5\u6b63\u5e38\u4f7f\u7528vllm\u5bf9vlm\u6a21\u578b\u8fdb\u884c\u52a0\u901f\u63a8\u7406\uff0c\u4f46\u4ecd\u7136\u5e0c\u671b\u8fdb\u4e00\u6b65\u63d0\u5347\u63a8\u7406\u901f\u5ea6\uff0c\u53ef\u4ee5\u5c1d\u8bd5\u4ee5\u4e0b\u53c2\u6570\uff1a

\u5982\u679c\u60a8\u6709\u8d85\u8fc7\u591a\u5f20\u663e\u5361\uff0c\u53ef\u4ee5\u4f7f\u7528vllm\u7684\u591a\u5361\u5e76\u884c\u6a21\u5f0f\u6765\u589e\u52a0\u541e\u5410\u91cf\uff1a--data-parallel-size 2

"},{"location":"zh/usage/advanced_cli_parameters/#_3","title":"\u53c2\u6570\u4f20\u9012\u8bf4\u660e","text":"

Tip

\u6240\u6709vllm/lmdeploy\u5b98\u65b9\u652f\u6301\u7684\u53c2\u6570\u90fd\u53ef\u7528\u901a\u8fc7\u547d\u4ee4\u884c\u53c2\u6570\u4f20\u9012\u7ed9 MinerU\uff0c\u5305\u62ec\u4ee5\u4e0b\u547d\u4ee4:mineru\u3001mineru-openai-server\u3001mineru-gradio\u3001mineru-api
\u5982\u679c\u60a8\u60f3\u4e86\u89e3\u66f4\u591a\u6709\u5173vllm\u7684\u53c2\u6570\u4f7f\u7528\u65b9\u6cd5\uff0c\u8bf7\u53c2\u8003 vllm\u5b98\u65b9\u6587\u6863
\u5982\u679c\u60a8\u60f3\u4e86\u89e3\u66f4\u591a\u6709\u5173lmdeploy\u7684\u53c2\u6570\u4f7f\u7528\u65b9\u6cd5\uff0c\u8bf7\u53c2\u8003 lmdeploy\u5b98\u65b9\u6587\u6863

"},{"location":"zh/usage/advanced_cli_parameters/#gpu","title":"GPU \u8bbe\u5907\u9009\u62e9\u4e0e\u914d\u7f6e","text":""},{"location":"zh/usage/advanced_cli_parameters/#cuda_visible_devices","title":"CUDA_VISIBLE_DEVICES \u57fa\u672c\u7528\u6cd5","text":"

Tip

\u4efb\u4f55\u60c5\u51b5\u4e0b\uff0c\u60a8\u90fd\u53ef\u4ee5\u901a\u8fc7\u5728\u547d\u4ee4\u884c\u7684\u5f00\u5934\u6dfb\u52a0CUDA_VISIBLE_DEVICES \u73af\u5883\u53d8\u91cf\u6765\u6307\u5b9a\u53ef\u89c1\u7684 GPU \u8bbe\u5907\uff1a
```
CUDA_VISIBLE_DEVICES=1 mineru -p <input_path> -o <output_path>\n
```
\u8fd9\u79cd\u6307\u5b9a\u65b9\u5f0f\u5bf9\u6240\u6709\u7684\u547d\u4ee4\u884c\u8c03\u7528\u90fd\u6709\u6548\uff0c\u5305\u62ec mineru\u3001mineru-openai-server\u3001mineru-gradio \u548c mineru-api\uff0c\u4e14\u5bf9pipeline\u3001vlm\u540e\u7aef\u5747\u9002\u7528\u3002

"},{"location":"zh/usage/advanced_cli_parameters/#_4","title":"\u5e38\u89c1\u8bbe\u5907\u914d\u7f6e\u793a\u4f8b","text":"

Tip

\u4ee5\u4e0b\u662f\u4e00\u4e9b\u5e38\u89c1\u7684 CUDA_VISIBLE_DEVICES \u8bbe\u7f6e\u793a\u4f8b\uff1a

CUDA_VISIBLE_DEVICES=1  # Only device 1 will be seen\nCUDA_VISIBLE_DEVICES=0,1  # Devices 0 and 1 will be visible\nCUDA_VISIBLE_DEVICES=\"0,1\"  # Same as above, quotation marks are optional\nCUDA_VISIBLE_DEVICES=0,2,3  # Devices 0, 2, 3 will be visible; device 1 is masked\nCUDA_VISIBLE_DEVICES=\"\"  # No GPU will be visible\n

"},{"location":"zh/usage/advanced_cli_parameters/#_5","title":"\u5b9e\u9645\u5e94\u7528\u573a\u666f","text":"

Tip

\u4ee5\u4e0b\u662f\u4e00\u4e9b\u53ef\u80fd\u7684\u4f7f\u7528\u573a\u666f\uff1a

\u5982\u679c\u60a8\u6709\u591a\u5f20\u663e\u5361\uff0c\u9700\u8981\u6307\u5b9a\u53610\u548c\u53611\uff0c\u5e76\u4f7f\u7528\u591a\u5361\u5e76\u884c\u6765\u542f\u52a8openai-server\uff0c\u53ef\u4ee5\u4f7f\u7528\u4ee5\u4e0b\u547d\u4ee4\uff1a
```
CUDA_VISIBLE_DEVICES=0,1 mineru-openai-server --engine vllm --port 30000 --data-parallel-size 2\n
```
\u5982\u679c\u60a8\u6709\u591a\u5f20\u663e\u5361\uff0c\u9700\u8981\u5728\u53610\u548c\u53611\u4e0a\u542f\u52a8\u4e24\u4e2afastapi\u670d\u52a1\uff0c\u5e76\u5206\u522b\u76d1\u542c\u4e0d\u540c\u7684\u7aef\u53e3\uff0c\u53ef\u4ee5\u4f7f\u7528\u4ee5\u4e0b\u547d\u4ee4\uff1a
```
# \u5728\u7ec8\u7aef1\u4e2d\nCUDA_VISIBLE_DEVICES=0 mineru-api --host 127.0.0.1 --port 8000\n# \u5728\u7ec8\u7aef2\u4e2d\nCUDA_VISIBLE_DEVICES=1 mineru-api --host 127.0.0.1 --port 8001\n
```

"},{"location":"zh/usage/cli_tools/","title":"\u547d\u4ee4\u884c\u5de5\u5177","text":""},{"location":"zh/usage/cli_tools/#_1","title":"\u547d\u4ee4\u884c\u5de5\u5177\u4f7f\u7528\u8bf4\u660e","text":""},{"location":"zh/usage/cli_tools/#_2","title":"\u67e5\u770b\u5e2e\u52a9\u4fe1\u606f","text":"

\u8981\u67e5\u770b MinerU \u547d\u4ee4\u884c\u5de5\u5177\u7684\u5e2e\u52a9\u4fe1\u606f\uff0c\u53ef\u4ee5\u4f7f\u7528 --help \u53c2\u6570\u3002\u4ee5\u4e0b\u662f\u5404\u4e2a\u547d\u4ee4\u884c\u5de5\u5177\u7684\u5e2e\u52a9\u4fe1\u606f\u793a\u4f8b\uff1a

mineru --help\nUsage: mineru [OPTIONS]\n\nOptions:\n  -v, --version                   \u663e\u793a\u7248\u672c\u5e76\u9000\u51fa\n  -p, --path PATH                 \u8f93\u5165\u6587\u4ef6\u8def\u5f84\u6216\u76ee\u5f55\uff08\u5fc5\u586b\uff09\n  -o, --output PATH               \u8f93\u51fa\u76ee\u5f55\uff08\u5fc5\u586b\uff09\n  --api-url TEXT                  MinerU FastAPI \u670d\u52a1\u5730\u5740\uff1b\u4e0d\u4f20\u65f6\u81ea\u52a8\u62c9\u8d77\u672c\u5730\u4e34\u65f6 mineru-api\n  -m, --method [auto|txt|ocr]     \u89e3\u6790\u65b9\u6cd5\uff1aauto\uff08\u9ed8\u8ba4\uff09\u3001txt\u3001ocr\uff08\u4ec5\u7528\u4e8e pipeline \u4e0e hybrid* \u540e\u7aef\uff09\n  -b, --backend [pipeline|hybrid-auto-engine|hybrid-http-client|vlm-auto-engine|vlm-http-client]\n                                  \u89e3\u6790\u540e\u7aef\uff08\u9ed8\u8ba4\u4e3a hybrid-auto-engine\uff09\n  -l, --lang [ch|ch_server|ch_lite|en|korean|japan|chinese_cht|ta|te|ka|th|el|latin|arabic|east_slavic|cyrillic|devanagari]\n                                  \u6307\u5b9a\u6587\u6863\u8bed\u8a00\uff08\u53ef\u63d0\u5347 OCR \u51c6\u786e\u7387\uff0c\u4ec5\u7528\u4e8e pipeline \u4e0e hybrid* \u540e\u7aef\uff09\n  -u, --url TEXT                  \u5f53\u4f7f\u7528 http-client \u65f6\uff0c\u4f20\u7ed9\u670d\u52a1\u7aef\u540e\u7aef\u7684 OpenAI \u517c\u5bb9\u5730\u5740\n  -s, --start INTEGER             \u5f00\u59cb\u89e3\u6790\u7684\u9875\u7801\uff08\u4ece 0 \u5f00\u59cb\uff09\n  -e, --end INTEGER               \u7ed3\u675f\u89e3\u6790\u7684\u9875\u7801\uff08\u4ece 0 \u5f00\u59cb\uff09\n  -f, --formula BOOLEAN           \u662f\u5426\u542f\u7528\u516c\u5f0f\u89e3\u6790\uff08\u9ed8\u8ba4\u5f00\u542f\uff09\n  -t, --table BOOLEAN             \u662f\u5426\u542f\u7528\u8868\u683c\u89e3\u6790\uff08\u9ed8\u8ba4\u5f00\u542f\uff09\n  --help                          \u663e\u793a\u5e2e\u52a9\u4fe1\u606f\n

mineru-api --help\nUsage: mineru-api [OPTIONS]\n\nOptions:\n  --host TEXT     \u670d\u52a1\u5668\u4e3b\u673a\u5730\u5740\uff08\u9ed8\u8ba4\uff1a127.0.0.1\uff09\n  --port INTEGER  \u670d\u52a1\u5668\u7aef\u53e3\uff08\u9ed8\u8ba4\uff1a8000\uff09\n  --reload        \u542f\u7528\u81ea\u52a8\u91cd\u8f7d\uff08\u5f00\u53d1\u6a21\u5f0f\uff09\n  --help          \u663e\u793a\u6b64\u5e2e\u52a9\u4fe1\u606f\u5e76\u9000\u51fa\n

mineru-gradio --help\nUsage: mineru-gradio [OPTIONS]\n\nOptions:\n  --enable-example BOOLEAN        \u542f\u7528\u793a\u4f8b\u6587\u4ef6\u8f93\u5165(\u9700\u8981\u5c06\u793a\u4f8b\u6587\u4ef6\u653e\u7f6e\u5728\u5f53\u524d\n                                  \u6267\u884c\u547d\u4ee4\u76ee\u5f55\u4e0b\u7684 `example` \u6587\u4ef6\u5939\u4e2d)\n  --enable-http-client BOOLEAN    \u5728\u540e\u7aef\u9009\u9879\u4e2d\u542f\u7528 HTTP \u5ba2\u6237\u7aef\u9009\u9879\n  --enable-api BOOLEAN            \u542f\u7528 Gradio API \u4ee5\u63d0\u4f9b\u5e94\u7528\u7a0b\u5e8f\u670d\u52a1\n  --max-convert-pages INTEGER     \u8bbe\u7f6e\u4ece PDF \u8f6c\u6362\u4e3a Markdown \u7684\u6700\u5927\u9875\u6570\n  --server-name TEXT              \u8bbe\u7f6e Gradio \u5e94\u7528\u7a0b\u5e8f\u7684\u670d\u52a1\u5668\u4e3b\u673a\u540d\n  --server-port INTEGER           \u8bbe\u7f6e Gradio \u5e94\u7528\u7a0b\u5e8f\u7684\u670d\u52a1\u5668\u7aef\u53e3\n  --latex-delimiters-type [a|b|all]\n                                  \u8bbe\u7f6e\u5728 Markdown \u6e32\u67d3\u4e2d\u4f7f\u7528\u7684 LaTeX \u5206\u9694\u7b26\u7c7b\u578b\n                                  ('a' \u8868\u793a '$' \u7c7b\u578b\uff0c'b' \u8868\u793a '()[]' \u7c7b\u578b\uff0c\n                                  'all' \u8868\u793a\u4e24\u79cd\u7c7b\u578b\u90fd\u4f7f\u7528)\n  --help                          \u663e\u793a\u6b64\u5e2e\u52a9\u4fe1\u606f\u5e76\u9000\u51fa\n

"},{"location":"zh/usage/cli_tools/#_3","title":"\u73af\u5883\u53d8\u91cf\u8bf4\u660e","text":"

Note

\u4ece\u5f53\u524d\u7248\u672c\u5f00\u59cb\uff0cmineru \u662f\u57fa\u4e8e mineru-api \u7684\u7f16\u6392\u5ba2\u6237\u7aef\uff1a - \u672a\u4f20 --api-url \u65f6\uff0cCLI \u4f1a\u81ea\u52a8\u62c9\u8d77\u672c\u5730\u4e34\u65f6 mineru-api - \u4f20\u5165 --api-url \u65f6\uff0cCLI \u4f1a\u76f4\u8fde\u8be5 FastAPI \u670d\u52a1 - --url \u4e0d\u518d\u8868\u793a MinerU API \u5730\u5740\uff0c\u800c\u662f\u670d\u52a1\u7aef vlm/hybrid-http-client \u6240\u9700\u7684 OpenAI \u517c\u5bb9\u5730\u5740

MinerU\u547d\u4ee4\u884c\u5de5\u5177\u7684\u67d0\u4e9b\u53c2\u6570\u5b58\u5728\u76f8\u540c\u529f\u80fd\u7684\u73af\u5883\u53d8\u91cf\u914d\u7f6e\uff0c\u901a\u5e38\u73af\u5883\u53d8\u91cf\u914d\u7f6e\u7684\u4f18\u5148\u7ea7\u9ad8\u4e8e\u547d\u4ee4\u884c\u53c2\u6570\uff0c\u4e14\u5728\u6240\u6709\u547d\u4ee4\u884c\u5de5\u5177\u4e2d\u90fd\u751f\u6548\u3002 \u4ee5\u4e0b\u662f\u5e38\u7528\u7684\u73af\u5883\u53d8\u91cf\u53ca\u5176\u8bf4\u660e\uff1a

MINERU_TOOLS_CONFIG_JSON\uff1a
- \u7528\u4e8e\u6307\u5b9a\u914d\u7f6e\u6587\u4ef6\u8def\u5f84
- \u9ed8\u8ba4\u4e3a\u7528\u6237\u76ee\u5f55\u4e0b\u7684mineru.json\uff0c\u53ef\u901a\u8fc7\u73af\u5883\u53d8\u91cf\u6307\u5b9a\u5176\u4ed6\u914d\u7f6e\u6587\u4ef6\u8def\u5f84\u3002
MINERU_FORMULA_ENABLE\uff1a
- \u7528\u4e8e\u542f\u7528\u516c\u5f0f\u89e3\u6790
- \u9ed8\u8ba4\u4e3atrue\uff0c\u53ef\u901a\u8fc7\u73af\u5883\u53d8\u91cf\u8bbe\u7f6e\u4e3afalse\u6765\u7981\u7528\u516c\u5f0f\u89e3\u6790\u3002
MINERU_FORMULA_CH_SUPPORT\uff1a
- \u7528\u4e8e\u542f\u7528\u4e2d\u6587\u516c\u5f0f\u89e3\u6790\u4f18\u5316\uff08\u5b9e\u9a8c\u6027\u529f\u80fd\uff09
- \u9ed8\u8ba4\u4e3afalse\uff0c\u53ef\u901a\u8fc7\u73af\u5883\u53d8\u91cf\u8bbe\u7f6e\u4e3atrue\u6765\u542f\u7528\u4e2d\u6587\u516c\u5f0f\u89e3\u6790\u4f18\u5316\u3002
- \u4ec5\u5bf9pipeline\u540e\u7aef\u751f\u6548\u3002
MINERU_TABLE_ENABLE\uff1a
- \u7528\u4e8e\u542f\u7528\u8868\u683c\u89e3\u6790
- \u9ed8\u8ba4\u4e3atrue\uff0c\u53ef\u901a\u8fc7\u73af\u5883\u53d8\u91cf\u8bbe\u7f6e\u4e3afalse\u6765\u7981\u7528\u8868\u683c\u89e3\u6790\u3002
MINERU_TABLE_MERGE_ENABLE\uff1a
- \u7528\u4e8e\u542f\u7528\u8868\u683c\u5408\u5e76\u529f\u80fd
- \u9ed8\u8ba4\u4e3atrue\uff0c\u53ef\u901a\u8fc7\u73af\u5883\u53d8\u91cf\u8bbe\u7f6e\u4e3afalse\u6765\u7981\u7528\u8868\u683c\u5408\u5e76\u529f\u80fd\u3002
MINERU_PDF_RENDER_TIMEOUT\uff1a
- \u7528\u4e8e\u8bbe\u7f6e\u5c06PDF\u6e32\u67d3\u4e3a\u56fe\u7247\u7684\u8d85\u65f6\u65f6\u95f4\uff08\u79d2\uff09
- \u9ed8\u8ba4\u4e3a300\u79d2\uff0c\u53ef\u901a\u8fc7\u73af\u5883\u53d8\u91cf\u8bbe\u7f6e\u4e3a\u5176\u4ed6\u503c\u4ee5\u8c03\u6574\u6e32\u67d3\u56fe\u7247\u7684\u8d85\u65f6\u65f6\u95f4\u3002
- \u4ec5\u5728linux\u548cmacOS\u7cfb\u7edf\u4e2d\u751f\u6548\u3002
MINERU_PDF_RENDER_THREADS\uff1a
- \u7528\u4e8e\u8bbe\u7f6e\u5c06PDF\u6e32\u67d3\u4e3a\u56fe\u7247\u65f6\u4f7f\u7528\u7684\u7ebf\u7a0b\u6570
- \u9ed8\u8ba4\u4e3a4\uff0c\u53ef\u901a\u8fc7\u73af\u5883\u53d8\u91cf\u8bbe\u7f6e\u4e3a\u5176\u4ed6\u503c\u4ee5\u8c03\u6574\u6e32\u67d3\u56fe\u7247\u65f6\u7684\u7ebf\u7a0b\u6570\u3002
- \u4ec5\u5728linux\u548cmacOS\u7cfb\u7edf\u4e2d\u751f\u6548\u3002
MINERU_INTRA_OP_NUM_THREADS\uff1a
- \u7528\u4e8e\u8bbe\u7f6eonnx\u6a21\u578b\u7684intra_op\u7ebf\u7a0b\u6570\uff0c\u5f71\u54cd\u5355\u4e2a\u7b97\u5b50\u7684\u8ba1\u7b97\u901f\u5ea6
- \u9ed8\u8ba4\u4e3a-1\uff08\u81ea\u52a8\u9009\u62e9\uff09\uff0c\u53ef\u901a\u8fc7\u73af\u5883\u53d8\u91cf\u8bbe\u7f6e\u4e3a\u5176\u4ed6\u503c\u4ee5\u8c03\u6574\u7ebf\u7a0b\u6570\u3002
MINERU_INTER_OP_NUM_THREADS\uff1a
- \u7528\u4e8e\u8bbe\u7f6eonnx\u6a21\u578b\u7684inter_op\u7ebf\u7a0b\u6570\uff0c\u5f71\u54cd\u591a\u4e2a\u7b97\u5b50\u7684\u5e76\u884c\u6267\u884c
- \u9ed8\u8ba4\u4e3a-1\uff08\u81ea\u52a8\u9009\u62e9\uff09\uff0c\u53ef\u901a\u8fc7\u73af\u5883\u53d8\u91cf\u8bbe\u7f6e\u4e3a\u5176\u4ed6\u503c\u4ee5\u8c03\u6574\u7ebf\u7a0b\u6570\u3002
MINERU_HYBRID_BATCH_RATIO\uff1a
- \u7528\u4e8e\u8bbe\u7f6e hybrid-* \u540e\u7aef\u4e2d \u5c0f\u6a21\u578b\u5904\u7406\u7684batch\u500d\u7387
- \u5728hybrid-http-client\u4e2d\u8f83\u4e3a\u5e38\u7528\uff0c\u53ef\u4ee5\u901a\u8fc7\u63a7\u5236\u5c0f\u6a21\u578b\u7684batch\u500d\u7387\u6765\u8c03\u6574\u5355\u4e2a\u5ba2\u6237\u7aef\u7684\u663e\u5b58\u5360\u7528\u91cf
- \u5355\u4e2aclient\u7aef\u663e\u5b58\u5927\u5c0f MINERU_HYBRID_BATCH_RATIO <= 6 GB 8 <= 4.5 GB 4 <= 3 GB 2 <= 2.5 GB 1
MINERU_HYBRID_FORCE_PIPELINE_ENABLE\uff1a
- \u7528\u4e8e\u5f3a\u5236\u5c06 hybrid-* \u540e\u7aef\u4e2d\u7684 \u6587\u672c\u63d0\u53d6\u90e8\u5206\u4f7f\u7528 \u5c0f\u6a21\u578b \u8fdb\u884c\u5904\u7406
- \u9ed8\u8ba4\u4e3afalse\uff0c\u53ef\u901a\u8fc7\u73af\u5883\u53d8\u91cf\u8bbe\u7f6e\u4e3atrue\u6765\u542f\u7528\u8be5\u529f\u80fd\uff0c\u4ece\u800c\u5728\u67d0\u4e9b\u6781\u7aef\u60c5\u51b5\u4e0b\u51cf\u5c11\u5e7b\u89c9\u7684\u53d1\u751f\u3002
MINERU_VL_MODEL_NAME\uff1a
- \u7528\u4e8e\u6307\u5b9a vlm/hybrid \u540e\u7aef\u4f7f\u7528\u7684\u6a21\u578b\u540d\u79f0\uff0c\u8fd9\u5c06\u5141\u8bb8\u60a8\u5728\u540c\u65f6\u5b58\u5728\u591a\u4e2a\u6a21\u578b\u7684\u8fdc\u7a0bopenai-server\u4e2d\u6307\u5b9a MinerU \u8fd0\u884c\u6240\u9700\u7684\u6a21\u578b\u3002
MINERU_VL_API_KEY:
- \u7528\u4e8e\u6307\u5b9a vlm/hybrid \u540e\u7aef\u4f7f\u7528\u7684API Key\uff0c\u8fd9\u5c06\u5141\u8bb8\u60a8\u5728\u8fdc\u7a0bopenai-server\u4e2d\u8fdb\u884c\u8eab\u4efd\u9a8c\u8bc1\u3002

"},{"location":"zh/usage/model_source/","title":"\u6a21\u578b\u6e90\u914d\u7f6e","text":""},{"location":"zh/usage/model_source/#_1","title":"\u6a21\u578b\u6e90\u8bf4\u660e","text":"

MinerU\u4f7f\u7528 HuggingFace \u548c ModelScope \u4f5c\u4e3a\u6a21\u578b\u4ed3\u5e93\uff0c\u7528\u6237\u53ef\u4ee5\u6839\u636e\u9700\u8981\u5207\u6362\u6a21\u578b\u6e90\u6216\u4f7f\u7528\u672c\u5730\u6a21\u578b\u3002

HuggingFace \u662f\u9ed8\u8ba4\u7684\u6a21\u578b\u6e90\uff0c\u5728\u5168\u7403\u8303\u56f4\u5185\u63d0\u4f9b\u4e86\u4f18\u5f02\u7684\u52a0\u8f7d\u901f\u5ea6\u548c\u6781\u9ad8\u7a33\u5b9a\u6027\u3002
ModelScope \u662f\u4e2d\u56fd\u5927\u9646\u5730\u533a\u7528\u6237\u7684\u6700\u4f73\u9009\u62e9\uff0c\u63d0\u4f9b\u4e86\u65e0\u7f1d\u517c\u5bb9\u7684SDK\u6a21\u5757\uff0c\u9002\u7528\u4e8e\u65e0\u6cd5\u8bbf\u95eeHuggingFace\u7684\u7528\u6237\u3002

"},{"location":"zh/usage/model_source/#_2","title":"\u6a21\u578b\u6e90\u7684\u5207\u6362\u65b9\u6cd5","text":""},{"location":"zh/usage/model_source/#_3","title":"\u901a\u8fc7\u547d\u4ee4\u884c\u53c2\u6570\u5207\u6362","text":"

\u76ee\u524d\u4ec5mineru\u547d\u4ee4\u884c\u5de5\u5177\u652f\u6301\u901a\u8fc7\u547d\u4ee4\u884c\u53c2\u6570\u5207\u6362\u6a21\u578b\u6e90\uff0c\u5176\u4ed6\u547d\u4ee4\u884c\u5de5\u5177\u5982mineru-api\u3001mineru-gradio\u7b49\u6682\u4e0d\u652f\u6301\u3002

mineru -p <input_path> -o <output_path> --source modelscope\n

"},{"location":"zh/usage/model_source/#_4","title":"\u901a\u8fc7\u73af\u5883\u53d8\u91cf\u5207\u6362","text":"

\u5728\u4efb\u4f55\u60c5\u51b5\u4e0b\u53ef\u4ee5\u901a\u8fc7\u8bbe\u7f6e\u73af\u5883\u53d8\u91cf\u6765\u5207\u6362\u6a21\u578b\u6e90\uff0c\u8fd9\u9002\u7528\u4e8e\u6240\u6709\u547d\u4ee4\u884c\u5de5\u5177\u548cAPI\u8c03\u7528\u3002

export MINERU_MODEL_SOURCE=modelscope\n

\u6216

import os\nos.environ[\"MINERU_MODEL_SOURCE\"] = \"modelscope\"\n

Tip

\u901a\u8fc7\u73af\u5883\u53d8\u91cf\u8bbe\u7f6e\u7684\u6a21\u578b\u6e90\u4f1a\u5728\u5f53\u524d\u7ec8\u7aef\u4f1a\u8bdd\u4e2d\u751f\u6548\uff0c\u76f4\u5230\u7ec8\u7aef\u5173\u95ed\u6216\u73af\u5883\u53d8\u91cf\u88ab\u4fee\u6539\u3002\u4e14\u4f18\u5148\u7ea7\u9ad8\u4e8e\u547d\u4ee4\u884c\u53c2\u6570\uff0c\u5982\u540c\u65f6\u8bbe\u7f6e\u4e86\u547d\u4ee4\u884c\u53c2\u6570\u548c\u73af\u5883\u53d8\u91cf\uff0c\u547d\u4ee4\u884c\u53c2\u6570\u5c06\u88ab\u5ffd\u7565\u3002

"},{"location":"zh/usage/model_source/#_5","title":"\u4f7f\u7528\u672c\u5730\u6a21\u578b","text":""},{"location":"zh/usage/model_source/#1","title":"1. \u4e0b\u8f7d\u6a21\u578b\u5230\u672c\u5730","text":"

mineru-models-download --help\n

\u6216\u4f7f\u7528\u4ea4\u4e92\u5f0f\u547d\u4ee4\u884c\u5de5\u5177\u9009\u62e9\u6a21\u578b\u4e0b\u8f7d\uff1a

mineru-models-download\n

Note

\u4e0b\u8f7d\u5b8c\u6210\u540e\uff0c\u6a21\u578b\u8def\u5f84\u4f1a\u5728\u5f53\u524d\u7ec8\u7aef\u7a97\u53e3\u8f93\u51fa\uff0c\u5e76\u81ea\u52a8\u5199\u5165\u7528\u6237\u76ee\u5f55\u4e0b\u7684 mineru.json\u3002
\u60a8\u4e5f\u53ef\u4ee5\u901a\u8fc7\u5c06\u914d\u7f6e\u6a21\u677f\u6587\u4ef6\u590d\u5236\u5230\u7528\u6237\u76ee\u5f55\u4e0b\u5e76\u91cd\u547d\u540d\u4e3a mineru.json \u6765\u521b\u5efa\u914d\u7f6e\u6587\u4ef6\u3002
\u6a21\u578b\u4e0b\u8f7d\u5230\u672c\u5730\u540e\uff0c\u60a8\u53ef\u4ee5\u81ea\u7531\u79fb\u52a8\u6a21\u578b\u6587\u4ef6\u5939\u5230\u5176\u4ed6\u4f4d\u7f6e\uff0c\u540c\u65f6\u9700\u8981\u5728 mineru.json \u4e2d\u66f4\u65b0\u6a21\u578b\u8def\u5f84\u3002
\u5982\u60a8\u5c06\u6a21\u578b\u6587\u4ef6\u5939\u90e8\u7f72\u5230\u5176\u4ed6\u670d\u52a1\u5668\u4e0a\uff0c\u8bf7\u786e\u4fdd\u5c06 mineru.json\u6587\u4ef6\u4e00\u540c\u79fb\u52a8\u5230\u65b0\u8bbe\u5907\u7684\u7528\u6237\u76ee\u5f55\u4e2d\u5e76\u6b63\u786e\u914d\u7f6e\u6a21\u578b\u8def\u5f84\u3002
\u5982\u60a8\u9700\u8981\u66f4\u65b0\u6a21\u578b\u6587\u4ef6\uff0c\u53ef\u4ee5\u518d\u6b21\u8fd0\u884c mineru-models-download \u547d\u4ee4\uff0c\u6a21\u578b\u66f4\u65b0\u6682\u4e0d\u652f\u6301\u81ea\u5b9a\u4e49\u8def\u5f84\uff0c\u5982\u60a8\u6ca1\u6709\u79fb\u52a8\u672c\u5730\u6a21\u578b\u6587\u4ef6\u5939\uff0c\u6a21\u578b\u6587\u4ef6\u4f1a\u589e\u91cf\u66f4\u65b0\uff1b\u5982\u60a8\u79fb\u52a8\u4e86\u6a21\u578b\u6587\u4ef6\u5939\uff0c\u6a21\u578b\u6587\u4ef6\u4f1a\u91cd\u65b0\u4e0b\u8f7d\u5230\u9ed8\u8ba4\u4f4d\u7f6e\u5e76\u66f4\u65b0mineru.json\u3002

"},{"location":"zh/usage/model_source/#2","title":"2. \u4f7f\u7528\u672c\u5730\u6a21\u578b\u8fdb\u884c\u89e3\u6790","text":"

mineru -p <input_path> -o <output_path> --source local\n

\u6216\u901a\u8fc7\u73af\u5883\u53d8\u91cf\u542f\u7528\uff1a

export MINERU_MODEL_SOURCE=local\nmineru -p <input_path> -o <output_path>\n

"},{"location":"zh/usage/quick_usage/","title":"\u57fa\u7840\u4f7f\u7528","text":""},{"location":"zh/usage/quick_usage/#mineru","title":"\u4f7f\u7528 MinerU","text":""},{"location":"zh/usage/quick_usage/#_1","title":"\u5feb\u901f\u914d\u7f6e\u6a21\u578b\u6e90","text":"

MinerU\u9ed8\u8ba4\u4f7f\u7528huggingface\u4f5c\u4e3a\u6a21\u578b\u6e90\uff0c\u82e5\u7528\u6237\u7f51\u7edc\u65e0\u6cd5\u8bbf\u95eehuggingface\uff0c\u53ef\u4ee5\u901a\u8fc7\u73af\u5883\u53d8\u91cf\u4fbf\u6377\u5730\u5207\u6362\u6a21\u578b\u6e90\u4e3amodelscope\uff1a

export MINERU_MODEL_SOURCE=modelscope\n

\u6709\u5173\u6a21\u578b\u6e90\u914d\u7f6e\u548c\u81ea\u5b9a\u4e49\u672c\u5730\u6a21\u578b\u8def\u5f84\u7684\u66f4\u591a\u4fe1\u606f\uff0c\u8bf7\u53c2\u8003\u6587\u6863\u4e2d\u7684\u6a21\u578b\u6e90\u8bf4\u660e\u3002"},{"location":"zh/usage/quick_usage/#_2","title":"\u901a\u8fc7\u547d\u4ee4\u884c\u5feb\u901f\u4f7f\u7528","text":"

MinerU\u5185\u7f6e\u4e86\u547d\u4ee4\u884c\u5de5\u5177\uff0c\u7528\u6237\u53ef\u4ee5\u901a\u8fc7\u547d\u4ee4\u884c\u5feb\u901f\u4f7f\u7528MinerU\u8fdb\u884cPDF\u89e3\u6790\uff1a

mineru -p <input_path> -o <output_path>\n

Tip

<input_path>\uff1a\u672c\u5730 PDF/\u56fe\u7247 \u6587\u4ef6\u6216\u76ee\u5f55
<output_path>\uff1a\u8f93\u51fa\u76ee\u5f55
\u672a\u4f20 --api-url \u65f6\uff0cCLI \u4f1a\u81ea\u52a8\u62c9\u8d77\u672c\u5730\u4e34\u65f6 mineru-api
\u4f20\u5165 --api-url \u65f6\uff0cCLI \u4f1a\u76f4\u8fde\u8fdc\u7aef\u6216\u5df2\u6709\u672c\u5730 FastAPI \u670d\u52a1

\u66f4\u591a\u5173\u4e8e\u8f93\u51fa\u6587\u4ef6\u7684\u4fe1\u606f\uff0c\u8bf7\u53c2\u8003\u8f93\u51fa\u6587\u4ef6\u8bf4\u660e\u3002

Note

\u547d\u4ee4\u884c\u5de5\u5177\u4f1a\u5728Linux\u548cmacOS\u7cfb\u7edf\u81ea\u52a8\u5c1d\u8bd5cuda/mps\u52a0\u901f\u3002Windows\u7528\u6237\u5982\u9700\u4f7f\u7528cuda\u52a0\u901f\uff0c \u8bf7\u524d\u5f80 Pytorch\u5b98\u7f51 \u9009\u62e9\u9002\u5408\u81ea\u5df1cuda\u7248\u672c\u7684\u547d\u4ee4\u5b89\u88c5\u652f\u6301\u52a0\u901f\u7684torch\u548ctorchvision\u3002

\u5982\u679c\u9700\u8981\u901a\u8fc7\u81ea\u5b9a\u4e49\u53c2\u6570\u8c03\u6574\u89e3\u6790\u9009\u9879\uff0c\u60a8\u4e5f\u53ef\u4ee5\u5728\u6587\u6863\u4e2d\u67e5\u770b\u66f4\u8be6\u7ec6\u7684\u547d\u4ee4\u884c\u5de5\u5177\u4f7f\u7528\u8bf4\u660e\u3002

"},{"location":"zh/usage/quick_usage/#apiwebuihttp-clientserver","title":"\u901a\u8fc7api\u3001webui\u3001http-client/server\u8fdb\u9636\u4f7f\u7528","text":"

\u901a\u8fc7python api\u76f4\u63a5\u8c03\u7528\uff1aPython \u8c03\u7528\u793a\u4f8b
\u901a\u8fc7fast api\u65b9\u5f0f\u8c03\u7528\uff1a
```
mineru-api --host 0.0.0.0 --port 8000\n
```
Tip

\u5728\u6d4f\u89c8\u5668\u4e2d\u8bbf\u95ee http://127.0.0.1:8000/docs \u67e5\u770bAPI\u6587\u6863\u3002
- \u5065\u5eb7\u68c0\u67e5\u63a5\u53e3\uff1aGET /health \u8fd4\u56de protocol_version\u3001processing_window_size\u3001max_concurrent_requests \u7b49\u670d\u52a1\u4fe1\u606f
- \u5f02\u6b65\u4efb\u52a1\u63d0\u4ea4\u63a5\u53e3\uff1aPOST /tasks
- \u540c\u6b65\u89e3\u6790\u63a5\u53e3\uff1aPOST /file_parse
- \u4efb\u52a1\u67e5\u8be2\u63a5\u53e3\uff1aGET /tasks/{task_id}\u3001GET /tasks/{task_id}/result
- API \u8f93\u51fa\u76ee\u5f55\u7531\u670d\u52a1\u7aef\u56fa\u5b9a\u63a7\u5236\uff0c\u9ed8\u8ba4\u5199\u5165 ./output
POST /tasks \u4f1a\u7acb\u5373\u8fd4\u56de task_id\uff1bPOST /file_parse \u4f1a\u5728\u5185\u90e8\u63d0\u4ea4\u5230\u540c\u4e00\u4e2a\u4efb\u52a1\u7ba1\u7406\u5668\uff0c\u7b49\u5f85\u4efb\u52a1\u5b8c\u6210\u540e\u540c\u6b65\u8fd4\u56de\u6700\u7ec8\u7ed3\u679c\u3002 \u4efb\u52a1\u4e3a\u5355\u8fdb\u7a0b\u3001\u8fdb\u7a0b\u5185\u72b6\u6001\u5b9e\u73b0\uff0c\u670d\u52a1\u91cd\u542f\u3001--reload \u70ed\u91cd\u8f7d\u6216\u591a\u8fdb\u7a0b\u90e8\u7f72\u540e\u4e0d\u4fdd\u8bc1\u4ecd\u53ef\u67e5\u8be2\u5386\u53f2\u4efb\u52a1\u72b6\u6001\u3002 \u9ed8\u8ba4\u4efb\u52a1\u5b8c\u6210\u6216\u5931\u8d25\u540e\u4fdd\u7559 24 \u5c0f\u65f6\uff0c\u968f\u540e\u81ea\u52a8\u6e05\u7406\u4efb\u52a1\u72b6\u6001\u548c\u8f93\u51fa\u76ee\u5f55\uff1b\u6e05\u7406\u540e\u8bbf\u95ee\u4efb\u52a1\u72b6\u6001\u6216\u7ed3\u679c\u4f1a\u8fd4\u56de 404\u3002 \u53ef\u901a\u8fc7\u73af\u5883\u53d8\u91cf MINERU_API_TASK_RETENTION_SECONDS \u548c MINERU_API_TASK_CLEANUP_INTERVAL_SECONDS \u8c03\u6574\u4fdd\u7559\u65f6\u957f\u4e0e\u6e05\u7406\u8f6e\u8be2\u95f4\u9694\u3002

\u5f02\u6b65\u4efb\u52a1\u63d0\u4ea4\u793a\u4f8b\uff1a
```
curl -X POST http://127.0.0.1:8000/tasks \\\n  -F \"files=@demo/pdfs/demo1.pdf\" \\\n  -F \"return_md=true\"\n
```
\u540c\u6b65\u89e3\u6790\u793a\u4f8b\uff1a
```
curl -X POST http://127.0.0.1:8000/file_parse \\\n  -F \"files=@demo/pdfs/demo1.pdf\" \\\n  -F \"return_md=true\" \\\n  -F \"response_format_zip=true\" \\\n  -F \"return_original_file=true\"\n
```
\u8f6e\u8be2\u4efb\u52a1\u72b6\u6001\u4e0e\u7ed3\u679c\uff1a
```
curl http://127.0.0.1:8000/tasks/<task_id>\ncurl http://127.0.0.1:8000/tasks/<task_id>/result\ncurl http://127.0.0.1:8000/health\n
```
\u542f\u52a8gradio webui \u53ef\u89c6\u5316\u524d\u7aef\uff1a
```
mineru-gradio --server-name 0.0.0.0 --server-port 7860\n
```
Tip
- \u5728\u6d4f\u89c8\u5668\u4e2d\u8bbf\u95ee http://127.0.0.1:7860 \u4f7f\u7528 Gradio WebUI\u3002

\u4f7f\u7528http-client/server\u65b9\u5f0f\u8c03\u7528\uff1a

# \u542f\u52a8openai\u517c\u5bb9\u670d\u52a1\u5668(\u9700\u8981\u5b89\u88c5vllm\u6216lmdeploy\u73af\u5883)\nmineru-openai-server --port 30000\n

Tip

\u5728\u53e6\u4e00\u4e2a\u7ec8\u7aef\u4e2d\u901a\u8fc7http client\u8fde\u63a5openai server

mineru -p <input_path> -o <output_path> -b hybrid-http-client -u http://127.0.0.1:30000\n

Note

\u6240\u6709vllm/lmdeploy\u5b98\u65b9\u652f\u6301\u7684\u53c2\u6570\u90fd\u53ef\u7528\u901a\u8fc7\u547d\u4ee4\u884c\u53c2\u6570\u4f20\u9012\u7ed9 MinerU\uff0c\u5305\u62ec\u4ee5\u4e0b\u547d\u4ee4:mineru\u3001mineru-openai-server\u3001mineru-gradio\u3001mineru-api\uff0c \u6211\u4eec\u6574\u7406\u4e86\u4e00\u4e9bvllm/lmdeploy\u4f7f\u7528\u4e2d\u7684\u5e38\u7528\u53c2\u6570\u548c\u4f7f\u7528\u65b9\u6cd5\uff0c\u53ef\u4ee5\u5728\u6587\u6863\u547d\u4ee4\u884c\u8fdb\u9636\u53c2\u6570\u4e2d\u83b7\u53d6\u3002

"},{"location":"zh/usage/quick_usage/#mineru_1","title":"\u57fa\u4e8e\u914d\u7f6e\u6587\u4ef6\u6269\u5c55 MinerU \u529f\u80fd","text":"

MinerU \u73b0\u5df2\u5b9e\u73b0\u5f00\u7bb1\u5373\u7528\uff0c\u4f46\u4e5f\u652f\u6301\u901a\u8fc7\u914d\u7f6e\u6587\u4ef6\u6269\u5c55\u529f\u80fd\u3002\u60a8\u53ef\u901a\u8fc7\u7f16\u8f91\u7528\u6237\u76ee\u5f55\u4e0b\u7684 mineru.json \u6587\u4ef6\uff0c\u6dfb\u52a0\u81ea\u5b9a\u4e49\u914d\u7f6e\u3002

Important

mineru.json \u6587\u4ef6\u4f1a\u5728\u60a8\u4f7f\u7528\u5185\u7f6e\u6a21\u578b\u4e0b\u8f7d\u547d\u4ee4 mineru-models-download \u65f6\u81ea\u52a8\u751f\u6210\uff0c\u4e5f\u53ef\u4ee5\u901a\u8fc7\u5c06\u914d\u7f6e\u6a21\u677f\u6587\u4ef6\u590d\u5236\u5230\u7528\u6237\u76ee\u5f55\u4e0b\u5e76\u91cd\u547d\u540d\u4e3a mineru.json \u6765\u521b\u5efa\u3002

\u4ee5\u4e0b\u662f\u4e00\u4e9b\u53ef\u7528\u7684\u914d\u7f6e\u9009\u9879\uff1a

latex-delimiter-config\uff1a
- \u7528\u4e8e\u914d\u7f6e LaTeX \u516c\u5f0f\u7684\u5206\u9694\u7b26
- \u9ed8\u8ba4\u4e3a$\u7b26\u53f7\uff0c\u53ef\u6839\u636e\u9700\u8981\u4fee\u6539\u4e3a\u5176\u4ed6\u7b26\u53f7\u6216\u5b57\u7b26\u4e32\u3002
llm-aided-config\uff1a
- \u7528\u4e8e\u914d\u7f6e LLM \u8f85\u52a9\u6807\u9898\u5206\u7ea7\u7684\u76f8\u5173\u53c2\u6570\uff0c\u517c\u5bb9\u6240\u6709\u652f\u6301openai\u534f\u8bae\u7684 LLM \u6a21\u578b
- \u9ed8\u8ba4\u4f7f\u7528\u963f\u91cc\u4e91\u767e\u70bc\u7684qwen3-next-80b-a3b-instruct\u6a21\u578b
- \u60a8\u9700\u8981\u81ea\u884c\u914d\u7f6e API \u5bc6\u94a5\u5e76\u5c06enable\u8bbe\u7f6e\u4e3atrue\u6765\u542f\u7528\u6b64\u529f\u80fd
- \u5982\u679c\u60a8\u7684api\u4f9b\u5e94\u5546\u4e0d\u652f\u6301enable_thinking\u53c2\u6570\uff0c\u8bf7\u624b\u52a8\u5c06\u8be5\u53c2\u6570\u5220\u9664
  - \u4f8b\u5982\uff0c\u5728\u60a8\u7684\u914d\u7f6e\u6587\u4ef6\u4e2d\uff0cllm-aided-config \u90e8\u5206\u53ef\u80fd\u5982\u4e0b\u6240\u793a\uff1a
```
\"llm-aided-config\": {\n   \"api_key\": \"your_api_key\",\n   \"base_url\": \"https://dashscope.aliyuncs.com/compatible-mode/v1\",\n   \"model\": \"qwen3-next-80b-a3b-instruct\",\n   \"enable_thinking\": false,\n   \"enable\": false\n}\n
```
  - \u8981\u79fb\u9664enable_thinking\u53c2\u6570\uff0c\u53ea\u9700\u5220\u9664\u5305\u542b\"enable_thinking\": false\u7684\u90a3\u4e00\u884c\uff0c\u7ed3\u679c\u5982\u4e0b:
```
\"llm-aided-config\": {\n   \"api_key\": \"your_api_key\",\n   \"base_url\": \"https://dashscope.aliyuncs.com/compatible-mode/v1\",\n   \"model\": \"qwen3-next-80b-a3b-instruct\",\n   \"enable\": false\n}\n
```
models-dir\uff1a
- \u7528\u4e8e\u6307\u5b9a\u672c\u5730\u6a21\u578b\u5b58\u50a8\u76ee\u5f55\uff0c\u8bf7\u4e3apipeline\u548cvlm\u540e\u7aef\u5206\u522b\u6307\u5b9a\u6a21\u578b\u76ee\u5f55\uff0c
- \u6307\u5b9a\u76ee\u5f55\u540e\u60a8\u53ef\u901a\u8fc7\u914d\u7f6e\u73af\u5883\u53d8\u91cfexport MINERU_MODEL_SOURCE=local\u6765\u4f7f\u7528\u672c\u5730\u6a21\u578b\u3002

"},{"location":"zh/usage/acceleration_cards/AMD/","title":"AMD","text":""},{"location":"zh/usage/acceleration_cards/AMD/#tritonrocm-vllmpipelinelayoutdoclayout-yolo","title":"\u57fa\u4e8eTriton\u7684ROCm \u4e0d\u540c\u540e\u7aef\u5b9e\u73b0\u4f18\u5316\uff0c\u57fa\u672c\u5b9e\u73b0vllm\u540e\u7aef\u6b63\u5e38\u63a8\u7406\uff0c\u4ee5\u53capipeline\u540e\u7aef\u4e2d\u7b2c\u4e00\u6b65layout\u7528\u7684DocLayout-YOLO","text":"

\u5df2\u6709\u5b8c\u6574python vllm\u548cmineru\u73af\u5883\u76f4\u63a5\u8df3\u8f6c\u7b2c\u4e94\u6b65\uff01\uff01\uff01 \u5176\u4ed6GPU\u6267\u884c\u95ee\u9898\u53ef\u4ee5\u53c2\u8003\uff0c\u5148prof\u67e5\u770b\u5b9a\u4f4d\u627e\u5230\u54ea\u4e2a\u7b97\u5b50\u95ee\u9898\uff0c\u7136\u540etriton\u540e\u7aef\u5b9e\u73b0\u5373\u53ef \u6d4b\u8bd5\u4e86\u4e00\u4e0b\uff0c\u57fa\u672c\u548cMinerU\u5b98\u7f51\u6548\u679c\u5dee\u4e0d\u591a\uff0c\u7528AMD\u7684\u4eba\u4e5f\u4e0d\u662f\u5f88\u591a\uff0c\u5c31\u5728\u8bc4\u8bba\u533a\u5206\u4eab\u7ed9\u5927\u5bb6\u4e86

"},{"location":"zh/usage/acceleration_cards/AMD/#1","title":"1.\u7ed3\u679c\u4ecb\u7ecd","text":"

\u8865\u5145\u4e00\u4e2a200\u9875\u7684PDF python\u7f16\u7a0b\u4e66\u6d4b\u8bd5\u4e00\u4e0b\u901f\u5ea6\uff0c\u53ef\u4ee5\u52301.99it/s\uff1a Two Step Extraction: 100%|\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588| 200/200 [01:40<00:00, 1.99it/s]

\u4e0b\u9762\u4e3a\u4e4b\u524d14\u5b66\u672f\u8bba\u6587\u6d4b\u8bd5\u7ed3\u679c\uff1a 7900xtx mineru-gradio --server-name 0.0.0.0 --server-port 7860 --enable-vllm-engine true \u901f\u5ea6\u5927\u6982\u4e3a1.6-1.8s/it\uff0c\u6ca1\u6709\u4ed4\u7ec6\u6d4b\u8bd5\uff0c\u7b80\u5355\u8bd5\u4e86\u4e24\u4e2a\u6587\u6863\u3002\u7b2c\u4e8c\u79cd\u77e9\u9635\u4e58\u6cd5\u4ee3\u66ff\u539f\u6765\u7684dots\u70b9\u4e58\u53ef\u4ee5\u8fdb\u4e00\u6b65\u63d0\u901f\u52301.3s/it\uff0c\u4f18\u5316\u540e\u7684\u4e3b\u8981\u7b97\u5b50\u8017\u65f6\u5728hipblast(\u8fd9\u4e2a\u6ca1\u6cd5\u63d0\u5347\u4e86)\u548cvllm triton\u540e\u7aef\uff0c\u5404\u536025%\u8017\u65f6\u5427\uff0cvllm tirion\u540e\u7aef\u8fd9\u4e2a\u8fd9\u4e2a\u53ea\u80fd\u7b49\u5b98\u65b9\u4f18\u5316\u4e86\u3002\u3002\u3002\u3002 doclayout-yolo\u7684layout\u901f\u5ea6\u4ece\u539f\u6765\u76841.6it/s\u63d0\u9ad8\u523015it/s\uff0c\u6ce8\u610f\u9700\u8981\u7f13\u5b58\u4e00\u4e0b\u8f93\u5165\u7684pdf\u5c3a\u5bf8\u540e\uff0ctriton\u5fc5\u987b\u8981\u7f13\u5b58\u5c3a\u5bf8\u6ca1\u529e\u6cd5\u3002\u4e3b\u8981\u662f\u4e3a\u4e86\u4fdd\u7559\u6a21\u578b\u8f93\u5165\u8f93\u51fa\u63a5\u53e3\uff0c\u6700\u5c0f\u4ee3\u7801\u6539\u52a8\u3002 \u91c7\u7528-b vlm-vllm-engine\u6a21\u5f0f\u4e3e\u4e2a\u4f8b\u5b50

\u6d4b\u8bd5\u7ed3\u679c\u4e3a\u4f18\u5316\u4e3a5d\u77e9\u9635\u4e58\u4ee3\u66ff\u539f\u6765\u7684\u70b9\u79ef\u7ed3\u679c\uff1a 2025-10-05 15:45:12.985 | INFO | mineru.backend.vlm.vlm_analyze:get_model:128 - get vllm-engine predictor cost: 18.45s Adding requests: 100%|\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588| 14/14 [00:01<00:00, 12.20it/s] Processed prompts: 100%|\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588| 14/14 [00:08<00:00, 1.56it/s, est. speed input: 2174.18 toks/s, output: 791.87 toks/s] Adding requests: 100%|\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588| 278/278 [00:00<00:00, 323.03it/s] Processed prompts: 100%|\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588| 278/278 [00:07<00:00, 37.63it/s, est. speed input: 5264.66 toks/s, output: 2733.31 toks/s]

mineru-gradio --server-name 0.0.0.0 --server-port 7860 --enable-vllm-engine true\u6d4b\u8bd5\uff1a 2025-10-05 15:46:55.953 | WARNING | mineru.cli.common:convert_pdf_bytes_to_bytes_by_pypdfium2:54 - end_page_id is out of range, use pdf_docs length Two Step Extraction: 100%|\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588\u2588| 14/14 [00:18<00:00, 1.30s/it]

"},{"location":"zh/usage/acceleration_cards/AMD/#2","title":"2.\u539f\u56e0\u4ecb\u7ecd","text":"

AMD RDNA\u4f7f\u7528vllm\u540e\u7aef\u6709\u4e25\u91cd\u7684\u6027\u80fd\u95ee\u9898\uff0c\u539f\u56e0\u662f\u56e0\u4e3avllm\u7684qwen2_vl.py\u4e2d\u6709\u4e00\u4e2a\u7b97\u5b50\u5728rocm kernel\u4e0a\u6ca1\u6709\u5bf9\u5e94\u7684\u5b9e\u73b0\uff0c\u5bfc\u81f4\u6027\u80fd\u51fa\u73b0\u4e25\u91cd\u7684\u5377\u79ef\u8ba1\u7b97\u56de\u9000\uff0c\u4e00\u6b21\u6267\u884c\u82b1\u4e8612s\uff0c\u3002\u3002\u3002\u3002\u3002\u3002\u3002\u3002\u4e00\u8a00\u96be\u5c3d\u3002\u5373MIOpen \u5e93\u4e2d\u7f3a\u5c11\u6a21\u578b\u4e2d\u7279\u5b9a Conv3d(bfloat16) \u7684\u4f18\u5316\u5185\u6838\u3002 DocLayout-YOLO\u7684g2l_crm.py\u7a7a\u6d1e\u5377\u79ef\u4e5f\u662f\u8fd9\u4e2a\u95ee\u9898\uff0c\u4e13\u4e1a\u7684CDNA MI210\u4e5f\u6ca1\u89e3\u51b3\u8fd9\u4e2a\u95ee\u9898 \u6b63\u597d\u4e00\u8d77\u5904\u7406\u4e86\u3002

"},{"location":"zh/usage/acceleration_cards/AMD/#3","title":"3.\u73af\u5883\u4ecb\u7ecd","text":"

System: Ubuntu 24.04.3 Kernel: Linux 6.14.0-33-generic ROCm version: 7.0.1 python\u73af\u5883\uff1a python 3.12 pytorch-triton-rocm 3.5.0+gitbbb06c03 torch 2.10.0.dev20251001+rocm7.0 torchvision 0.25.0.dev20251003+rocm7.0 vllm 0.11.0rc2.dev198+g736fbf4c8.rocm701 \u4e0d\u540c\u7248\u672c\u65e0\u6240\u8c13\uff0c\u5904\u7406\u65b9\u6cd5\u662f\u4e00\u6837\u7684\u3002

"},{"location":"zh/usage/acceleration_cards/AMD/#4","title":"4.\u524d\u7f6e\u73af\u5883\u5b89\u88c5","text":"

uv venv --python python3.12\nsource .venv/bin/activate\nuv pip install --pre torch torchvision   -i https://pypi.tuna.tsinghua.edu.cn/simple/   --extra-index-url https://download.pytorch.org/whl/nightly/rocm7.0\nuv pip install pip\n# \u907f\u514d\u8986\u76d6\u6211\u4eec\u672c\u5730\u7684pytorch\uff0c\u6539\u7528pip\u800c\u6ca1\u6709\u7ee7\u7eed\u4f7f\u7528uv pip\npip install -U \"mineru[core]\" -i https://pypi.mirrors.ustc.edu.cn/simple/\n

vllm \u5b89\u88c5\u53c2\u8003\u5b98\u65b9\u624b\u518cVllm

#\u624b\u52a8\u5b89\u88c5aiter\uff0cvllm\uff0camd-smi\u7b49\uff0c\u81ea\u884c\u627e\u4e00\u4e2a\u4f4d\u7f6eclone\uff0c\u7136\u540e\u8fdb\u5165\u8be5\u76ee\u5f55\u5427\ngit clone --recursive https://github.com/ROCm/aiter.git\ncd aiter\ngit submodule sync; git submodule update --init --recursive\npython setup.py develop\ncd ..\ngit clone https://github.com/vllm-project/vllm.git\ncd vllm/\ncp -r /opt/rocm/share/amd_smi ~/Pytorch/vllm/\npip install amd_smi/\npip install --upgrade numba \\\n    scipy \\\n    huggingface-hub[cli,hf_transfer] \\\n    setuptools_scm\npip install -r requirements/rocm.txt\nexport PYTORCH_ROCM_ARCH=\"gfx1100\"   #\u6839\u636e\u81ea\u5df1\u7684GPU\u67b6\u6784 rocminfo | grep gfx\npython setup.py develop\n

"},{"location":"zh/usage/acceleration_cards/AMD/#5vllmtriton","title":"5.vllm\u4e2d\u5173\u952etriton\u7b97\u5b50\u6dfb\u52a0","text":""},{"location":"zh/usage/acceleration_cards/AMD/#1518sit7900xtx13sitamd-gpu","title":"\u8fd9\u91cc\u6211\u7ed9\u51fa\u4e24\u79cd\u89e3\u51b3\u65b9\u6cd5\uff0c\u7b2c\u4e00\u79cd\u89e3\u51b3\u65b9\u6cd5\u5c31\u662f\u524d\u9762\u63d0\u5230\u7684\u4f18\u5316\u52301.5\u52301.8s/it\uff0c\u7b2c\u4e8c\u79cd\u65b9\u6cd5\u6709\u624b\u52a8\u4f18\u5316\u7b97\u5b50\u5230\u77e9\u9635\u4e58\u6cd5\uff0c7900xtx\u80af\u5b9a\u9002\u7528\uff0c\u5927\u69821.3s/it\uff0c\u5176\u4ed6AMD GPU\u76f8\u5bf9\u65b9\u6848\u4e00\u4e5f\u6709\u63d0\u901f\uff0c\u4f46\u662f\u4e0d\u4e00\u5b9a\u662f\u6700\u4f73\u901f\u5ea6\u5b9e\u73b0\uff0c\u91cc\u9762\u7684\u624b\u52a8\u90e8\u5206\u53ef\u80fd\u9700\u8981\u5fae\u8c03\u3002","text":"

\u6ce8\u610fpip\u628atriton \u540e\u7aef\u7684flash_attn\u5378\u8f7d\u4e86\uff0c\u641e\u4e86\u534a\u5929\u5404\u79cd\u5c1d\u8bd5\u8fd8\u662f\u62a5\u9519\uff0c\u95ee\u9898\u6bd4\u8f83\u5927\uff0c\u76f4\u63a5\u4e0d\u7528\u5c31\u884c\u4e86

#\u5b9a\u4f4d\u81ea\u5df1vllm\u4f4d\u7f6eXXX\npip show vllm\n

\u5173\u952e\u66f4\u6539 XXX/vllm/model_executor/models/qwen2_vl.py\u6587\u4ef6\uff1a 1.qwen2_vl.py\u6587\u4ef633\u884c\u4e0b\u589e\u52a0from .qwen2_vl_vision_kernels import triton_conv3d_patchify

from collections.abc import Iterable, Mapping, Sequence\nfrom functools import partial\nfrom typing import Annotated, Any, Callable, Literal, Optional, Union\n\nimport torch\nimport torch.nn as nn\nimport torch.nn.functional as F\nfrom .qwen2_vl_vision_kernels import triton_conv3d_patchify\n

\u63a5\u4e0b\u6765\u5206\u4e3a\u65b9\u6848\u4e00(2.1\u548c3.1)\u548c\u65b9\u6848\u4e8c(2.2\u548c3.2)\uff0c\u9009\u53d6\u4e00\u79cd\u5b9e\u73b0\u5373\u53ef

\u65b9\u68481 2.1qwen2_vl.py\u6587\u4ef6498\u884cclass Qwen2VisionPatchEmbed(nn.Module),PS.\u5c31\u662f\u8fd9\u73a9\u610fAMD\u6ca1\u6709\u73b0\u6210\u7684\u5185\u6838\u7b97\u5b50\u5bfc\u81f4\u56de\u9000

class Qwen2VisionPatchEmbed(nn.Module):\n\n    def __init__(\n        self,\n        patch_size: int = 14,\n        temporal_patch_size: int = 2,\n        in_channels: int = 3,\n        embed_dim: int = 1152,\n    ) -> None:\n        super().__init__()\n        self.patch_size = patch_size\n        self.temporal_patch_size = temporal_patch_size\n        self.embed_dim = embed_dim\n\n        kernel_size = (temporal_patch_size, patch_size, patch_size)\n        self.proj = nn.Conv3d(in_channels,\n                              embed_dim,\n                              kernel_size=kernel_size,\n                              stride=kernel_size,\n                              bias=False)\n    def forward(self, x: torch.Tensor) -> torch.Tensor:\n        L, C = x.shape\n        x_reshaped = x.view(L, -1, self.temporal_patch_size, self.patch_size,\n                            self.patch_size)\n\n        # Call your custom Triton kernel instead of self.proj\n        x_out = triton_conv3d_patchify(x_reshaped, self.proj.weight)\n\n        # The output of our kernel is already the correct shape [L, embed_dim]\n        return x_out\n

3.1XXX/vllm/model_executor/models/\u76ee\u5f55\u4e0b\u521b\u5efaqwen2_vl_vision_kernels.py\u6587\u4ef6\uff0c\u7528triton\u5b9e\u73b0

import torch\nfrom vllm.triton_utils import tl, triton\n\n@triton.jit\ndef _conv3d_patchify_kernel(\n    # Pointers to tensors\n    X, W, Y,\n    # Tensor dimensions\n    N, C_in, D_in, H_in, W_in,\n    C_out, KD, KH, KW,\n    # Stride and padding for memory access\n    stride_xn, stride_xc, stride_xd, stride_xh, stride_xw,\n    stride_wn, stride_wc, stride_wd, stride_wh, stride_ww,\n    stride_yn, stride_yc,\n    # Triton-specific metaparameters\n    BLOCK_SIZE: tl.constexpr,\n):\n    \"\"\"\n    Triton kernel for a non-overlapping 3D patching convolution.\n    Each kernel instance computes one output value for one patch.\n    \"\"\"\n    # Get the program IDs for the N (patch) and C_out (output channel) dimensions\n    pid_n = tl.program_id(0)  # The index of the patch we are processing\n    pid_cout = tl.program_id(1) # The index of the output channel we are computing\n\n    # --- Calculate memory pointers ---\n    # Pointer to the start of the current input patch\n    x_ptr = X + (pid_n * stride_xn)\n    # Pointer to the start of the current filter (weight)\n    w_ptr = W + (pid_cout * stride_wn)\n    # Pointer to where the output will be stored\n    y_ptr = Y + (pid_n * stride_yn + pid_cout * stride_yc)\n\n    # --- Perform the convolution (element-wise product and sum) ---\n    # This is a dot product between the flattened patch and the flattened filter.\n    accumulator = tl.zeros((BLOCK_SIZE,), dtype=tl.float32)\n\n    # Iterate over the elements of the patch/filter\n    for c_offset in range(0, C_in):\n        for d_offset in range(0, KD):\n            for h_offset in range(0, KH):\n                # Unrolled loop for the innermost dimension (width) for performance\n                for w_offset in range(0, KW, BLOCK_SIZE):\n                    # Create masks to handle cases where KW is not a multiple of BLOCK_SIZE\n                    w_range = w_offset + tl.arange(0, BLOCK_SIZE)\n                    w_mask = w_range < KW\n\n                    # Calculate offsets to load data\n                    patch_offset = (c_offset * stride_xc + d_offset * stride_xd +\n                                    h_offset * stride_xh + w_range * stride_xw)\n                    filter_offset = (c_offset * stride_wc + d_offset * stride_wd +\n                                     h_offset * stride_wh + w_range * stride_ww)\n\n                    # Load patch and filter data, applying masks\n                    patch_vals = tl.load(x_ptr + patch_offset, mask=w_mask, other=0.0)\n                    filter_vals = tl.load(w_ptr + filter_offset, mask=w_mask, other=0.0)\n\n                    # Multiply and accumulate\n                    accumulator += patch_vals.to(tl.float32) * filter_vals.to(tl.float32)\n\n    # Sum the accumulator block and store the single output value\n    output_val = tl.sum(accumulator, axis=0)\n    tl.store(y_ptr, output_val)\n\n\ndef triton_conv3d_patchify(x: torch.Tensor, weight: torch.Tensor) -> torch.Tensor:\n    \"\"\"\n    Python wrapper for the 3D patching convolution Triton kernel.\n    \"\"\"\n    # Get tensor dimensions\n    N, C_in, D_in, H_in, W_in = x.shape\n    C_out, _, KD, KH, KW = weight.shape\n\n    # Create the output tensor\n    # The output of this specific conv is (N, C_out, 1, 1, 1), which we squeeze\n    Y = torch.empty((N, C_out), dtype=x.dtype, device=x.device)\n\n    # Define the grid for launching the Triton kernel\n    # Each kernel instance handles one patch (N) for one output channel (C_out)\n    grid = (N, C_out)\n\n    # Launch the kernel\n    # We pass all strides to make the kernel flexible\n    _conv3d_patchify_kernel[grid](\n        x, weight, Y,\n        N, C_in, D_in, H_in, W_in,\n        C_out, KD, KH, KW,\n        x.stride(0), x.stride(1), x.stride(2), x.stride(3), x.stride(4),\n        weight.stride(0), weight.stride(1), weight.stride(2), weight.stride(3), weight.stride(4),\n        Y.stride(0), Y.stride(1),\n        BLOCK_SIZE=16, # A reasonable default, can be tuned\n    )\n\n    return Y\n

\u65b9\u68482 2.2qwen2_vl.py\u6587\u4ef6498\u884cclass Qwen2VisionPatchEmbed(nn.Module)\u51fd\u6570,PS.\u5c31\u662f\u8fd9\u73a9\u610fAMD\u6ca1\u6709\u73b0\u6210\u7684\u5185\u6838\u7b97\u5b50\u5bfc\u81f4\u56de\u9000\uff0c\u8fd9\u91cc\u6211\u4eec\u76f4\u63a55D\u5f20\u91cf\u4e00\u6b65\u5230\u4f4d\uff0c\u6539\u4e3a\u77e9\u9635\u4e58\u6cd5

class Qwen2VisionPatchEmbed(nn.Module):\n\n    def __init__(\n        self,\n        patch_size: int = 14,\n        temporal_patch_size: int = 2,\n        in_channels: int = 3,\n        embed_dim: int = 1152,\n    ) -> None:\n        super().__init__()\n        self.patch_size = patch_size\n        self.temporal_patch_size = temporal_patch_size\n        self.embed_dim = embed_dim\n\n        kernel_size = (temporal_patch_size, patch_size, patch_size)\n\n        self.proj = nn.Conv3d(in_channels,\n                              embed_dim,\n                              kernel_size=kernel_size,\n                              stride=kernel_size,\n                              bias=False)\n\n    def forward(self, x: torch.Tensor) -> torch.Tensor:\n        L, C = x.shape\n        x_reshaped_5d = x.view(L, -1, self.temporal_patch_size, self.patch_size,\n                               self.patch_size)\n\n        return triton_conv3d_patchify(x_reshaped_5d, self.proj.weight)\n

3.2XXX/vllm/model_executor/models/\u76ee\u5f55\u4e0b\u521b\u5efaqwen2_vl_vision_kernels.py\u6587\u4ef6\uff0c\u7528triton\u5b9e\u73b0

import torch\nfrom vllm.triton_utils import tl, triton\n\n@triton.jit\ndef _conv_gemm_kernel(\n    A, B, C, M, N, K,\n    stride_am, stride_ak,\n    stride_bk, stride_bn,\n    stride_cm, stride_cn,\n    BLOCK_M: tl.constexpr, BLOCK_N: tl.constexpr, BLOCK_K: tl.constexpr,\n):\n    pid_m = tl.program_id(0)\n    pid_n = tl.program_id(1)\n    offs_m = pid_m * BLOCK_M + tl.arange(0, BLOCK_M)\n    offs_n = pid_n * BLOCK_N + tl.arange(0, BLOCK_N)\n    offs_k = tl.arange(0, BLOCK_K)\n    a_ptrs = A + (offs_m[:, None] * stride_am + offs_k[None, :] * stride_ak)\n    b_ptrs = B + (offs_k[:, None] * stride_bk + offs_n[None, :] * stride_bn)\n    accumulator = tl.zeros((BLOCK_M, BLOCK_N), dtype=tl.float32)\n    for k in range(0, K, BLOCK_K):\n        a = tl.load(a_ptrs, mask=(offs_m[:, None] < M) & (offs_k[None, :] < K), other=0.0)\n        b = tl.load(b_ptrs, mask=(offs_k[:, None] < K) & (offs_n[None, :] < N), other=0.0)\n        accumulator += tl.dot(a, b)\n        a_ptrs += BLOCK_K * stride_ak\n        b_ptrs += BLOCK_K * stride_bk\n        offs_k += BLOCK_K\n    c = accumulator.to(C.dtype.element_ty)\n    offs_cm = pid_m * BLOCK_M + tl.arange(0, BLOCK_M)\n    offs_cn = pid_n * BLOCK_N + tl.arange(0, BLOCK_N)\n    c_ptrs = C + stride_cm * offs_cm[:, None] + stride_cn * offs_cn[None, :]\n    c_mask = (offs_cm[:, None] < M) & (offs_cn[None, :] < N)\n    tl.store(c_ptrs, c, mask=c_mask)\n\ndef triton_conv3d_patchify(x_5d: torch.Tensor, weight_5d: torch.Tensor) -> torch.Tensor:\n    N_patches, _, _, _, _ = x_5d.shape\n    C_out, _, _, _, _ = weight_5d.shape\n    A = x_5d.view(N_patches, -1)\n    B = weight_5d.view(C_out, -1).transpose(0, 1).contiguous()\n    M, K = A.shape\n    _K, N = B.shape\n    assert K == _K\n    C = torch.empty((M, N), device=A.device, dtype=A.dtype)\n\n    # --- \u9488\u5bf97900xtx\u7684\u624b\u52a8\u8c03\u4f18\u914d\u7f6e\uff0c\u5176\u4ed6GPU\u7684\u6700\u4f18\u7ec4\u5408\u53ef\u80fd\u9700\u8981\u81ea\u884c\u5bfb\u627e\uff0cAMD\u7684autotune\u6548\u679c\u5c31\u662f\u6ca1\u6709\u6548\u679c ---\n    best_config = {\n        'BLOCK_M': 128,\n        'BLOCK_N': 128,\n        'BLOCK_K': 32,\n    }\n    num_stages = 4\n    num_warps = 8\n\n    grid = (triton.cdiv(M, best_config['BLOCK_M']),\n            triton.cdiv(N, best_config['BLOCK_N']))\n\n    _conv_gemm_kernel[grid](\n        A, B, C,\n        M, N, K,\n        A.stride(0), A.stride(1),\n        B.stride(0), B.stride(1),\n        C.stride(0), C.stride(1),\n        **best_config,\n        num_stages=num_stages,\n        num_warps=num_warps\n    )\n\n    return C\n

4.\u5173\u95ed\u7ec8\u7aef\u540e\u518d\u6b21\u4f7f\u7528mineru-gradio\u4f1a\u62a5\u4e00\u4e2aLora\u9519\u8bef\uff0c\u4fee\u6539\u4ee3\u7801\u8df3\u8fc7\u5b83

pip show mineru_vl_utils\n

\u6253\u5f00\u8be5\u6587\u4ef6XXX/mineru_vl_utils/vlm_client/vllm_async_engine_client.py\u4fee\u6539\u7b2c58\u884cself.tokenizer = vllm_async_llm.tokenizer.get_lora_tokenizer()\u4e3a\uff1a

        try:\n            self.tokenizer = vllm_async_llm.tokenizer.get_lora_tokenizer()\n        except AttributeError:\n            # \u5982\u679c\u6ca1\u6709 get_lora_tokenizer \u65b9\u6cd5\uff0c\u76f4\u63a5\u4f7f\u7528\u539f\u59cb tokenizer\n            self.tokenizer = vllm_async_llm.tokenizer\n

\u6700\u540e\u6574\u4e24\u4e2a\u73af\u5883\u53d8\u91cf\u540e\u6109\u5feb\u73a9\u800d\u5373\u53ef

export MINERU_MODEL_SOURCE=modelscope\nexport TORCH_ROCM_AOTRITON_ENABLE_EXPERIMENTAL=1\n

"},{"location":"zh/usage/acceleration_cards/AMD/#6vllmpipeline-layoutdoclayout-yolo","title":"6.vllm\u540e\u7aef\u5df2\u7ecf\u6ca1\u6709\u95ee\u9898\uff0c\u4e0b\u9762\u662fpipeline \u4e2dlayout\u7528\u7684doclayout-yolo\u6a21\u578b\u7a7a\u6d1e\u5377\u79ef\u95ee\u9898","text":""},{"location":"zh/usage/acceleration_cards/AMD/#doclayout-yolo-pipeline","title":"\u6211\u5728 DocLayout-YOLO \u4e0b\u505a\u4e86\u4e00\u4e2a\u56de\u7b54\uff0c\u56e0\u6b64 pipeline \u7684\u7a7a\u6d1e\u5377\u79ef\u95ee\u9898\u4e0d\u5728\u8fd9\u91cc\u8d58\u8ff0\uff0c\u76f4\u63a5\u70b9\u51fb\u94fe\u63a5\u67e5\u770b\u5373\u53ef\u3002","text":"

\u67e5\u770b\u81ea\u5df1doclayout-yolo\u5b89\u88c5\u4f4d\u7f6e\u5982\u4e0b\uff0c\u7136\u540e\u8fdb\u5165\u4fee\u6539\u94fe\u63a5\u4e2d\u56de\u590d\u4ecb\u7ecd\u7684\u6587\u4ef6\u5373\u53ef

pip show doclayout-yolo\n

"},{"location":"zh/usage/acceleration_cards/Ascend/","title":"Ascend","text":""},{"location":"zh/usage/acceleration_cards/Ascend/#1","title":"1. \u6d4b\u8bd5\u5e73\u53f0","text":"

\u4ee5\u4e0b\u4e3a\u672c\u6307\u5357\u6d4b\u8bd5\u4f7f\u7528\u7684\u5e73\u53f0\u4fe1\u606f\uff0c\u4f9b\u53c2\u8003\uff1a

os: CTyunOS 22.06  \ncpu: Kunpeng-920 (aarch64)  \nnpu: Ascend 910B2  \ndriver: 23.0.3 \ndocker: 20.10.12\n

"},{"location":"zh/usage/acceleration_cards/Ascend/#2","title":"2. \u73af\u5883\u51c6\u5907","text":"

Note

Ascend\u52a0\u901f\u5361\u652f\u6301\u4f7f\u7528vllm\u6216lmdeploy\u8fdb\u884cVLM\u6a21\u578b\u63a8\u7406\u52a0\u901f\u3002\u8bf7\u6839\u636e\u5b9e\u9645\u9700\u6c42\u9009\u62e9\u5b89\u88c5\u548c\u4f7f\u7528\u5176\u4e2d\u4e4b\u4e00:

"},{"location":"zh/usage/acceleration_cards/Ascend/#21-dockerfile-vllm","title":"2.1 \u4f7f\u7528 Dockerfile \u6784\u5efa\u955c\u50cf \uff08vllm\uff09","text":"

Tip

ascend-vllm\u652f\u6301\u8bbe\u5907\u5982\u4e0b:

Atlas A2 training series (Atlas 800T A2, Atlas 900 A2 PoD, Atlas 200T A2 Box16, Atlas 300T A2)
Atlas 800I A2 inference series (Atlas 800I A2)
Atlas A3 training series (Atlas 800T A3, Atlas 900 A3 SuperPoD, Atlas 9000 A3 SuperPoD)
Atlas 800I A3 inference series (Atlas 800I A3)
[Experimental] Atlas 300I inference series (Atlas 300I Duo)

Dockerfile\u6587\u4ef6\u7b2c\u4e09\u884c\u4e3aascend-vllm\u57fa\u7840\u955c\u50cf\u4fe1\u606f,\u9ed8\u8ba4tag\u4e3aA2\u9002\u914d\u7684\u7248\u672c,\u4f8b\u5982 v0.11.0

\u5982\u9700\u4f7f\u7528A3\u9002\u914d\u7684\u7248\u672c,\u8bf7\u5c06\u7b2c\u4e09\u884c\u7684tag\u4fee\u6539\u4e3a v0.11.0-a3,\u7136\u540e\u518d\u6267\u884cbuild\u64cd\u4f5c\u3002
\u5982\u9700\u4f7f\u7528Atlas 300I Duo\u9002\u914d\u7684\u7248\u672c,\u8bf7\u5c06\u7b2c\u4e09\u884c\u7684tag\u4fee\u6539\u4e3a v0.10.0rc1-310p,\u7136\u540e\u518d\u6267\u884cbuild\u64cd\u4f5c\u3002

wget https://gcore.jsdelivr.net/gh/opendatalab/MinerU@master/docker/china/npu.Dockerfile\ndocker build --network=host -t mineru:npu-vllm-latest -f npu.Dockerfile .\n

"},{"location":"zh/usage/acceleration_cards/Ascend/#22-dockerfile-lmdeploy","title":"2.2 \u4f7f\u7528 Dockerfile \u6784\u5efa\u955c\u50cf \uff08lmdeploy\uff09","text":"

Tip

ascend-lmdeploy\u652f\u6301\u8bbe\u5907\u5982\u4e0b:

Atlas A2 training series (Atlas 800T A2, Atlas 900 A2 PoD, Atlas 200T A2 Box16, Atlas 300T A2)
Atlas 800I A2 inference series (Atlas 800I A2)

\u5982\u679c\u60a8\u7684\u8bbe\u5907\u4e3aAtlas A3\u7cfb\u5217\u6216Atlas 300I Duo\u7cfb\u5217\uff0c\u8bf7\u4f7f\u7528vllm\u7248\u672c\u7684\u955c\u50cf\u3002

wget https://gcore.jsdelivr.net/gh/opendatalab/MinerU@master/docker/china/npu.Dockerfile\n# \u5c06\u57fa\u7840\u955c\u50cf\u4ece vllm \u5207\u6362\u4e3a lmdeploy\nsed -i '3s/^/# /' npu.Dockerfile && sed -i '5s/^# //' npu.Dockerfile\ndocker build --network=host -t mineru:npu-lmdeploy-latest -f npu.Dockerfile .\n

"},{"location":"zh/usage/acceleration_cards/Ascend/#3-docker","title":"3. \u542f\u52a8 Docker \u5bb9\u5668","text":"

docker run -u root --name mineru_docker --privileged=true \\\n    --ipc=host \\\n    --network=host \\\n    --device=/dev/davinci0 \\\n    --device=/dev/davinci1 \\\n    --device=/dev/davinci2 \\\n    --device=/dev/davinci3 \\\n    --device=/dev/davinci4 \\\n    --device=/dev/davinci5 \\\n    --device=/dev/davinci6 \\\n    --device=/dev/davinci7 \\\n    --device=/dev/davinci_manager \\\n    --device=/dev/devmm_svm \\\n    --device=/dev/hisi_hdc \\\n    -v /var/log/npu/:/usr/slog \\\n    -v /usr/local/dcmi:/usr/local/dcmi \\\n    -v /usr/local/bin/npu-smi:/usr/local/bin/npu-smi \\\n    -v /usr/local/Ascend/driver:/usr/local/Ascend/driver \\\n    -e VLLM_WORKER_MULTIPROC_METHOD=spawn \\\n    -e MINERU_MODEL_SOURCE=local \\\n    -e MINERU_LMDEPLOY_DEVICE=ascend \\\n    -it mineru:npu-vllm-latest \\\n    /bin/bash\n

Tip

\u8bf7\u6839\u636e\u5b9e\u9645\u60c5\u51b5\u9009\u62e9\u4f7f\u7528vllm\u6216lmdeploy\u7248\u672c\u7684\u955c\u50cf\uff0c\u5982\u9700\u4f7f\u7528lmdeploy\uff0c\u66ff\u6362\u4e0a\u8ff0\u547d\u4ee4\u4e2d\u7684mineru:npu-vllm-latest\u4e3amineru:npu-lmdeploy-latest\u5373\u53ef\u3002

\u6267\u884c\u8be5\u547d\u4ee4\u540e\uff0c\u60a8\u5c06\u8fdb\u5165\u5230Docker\u5bb9\u5668\u7684\u4ea4\u4e92\u5f0f\u7ec8\u7aef\uff0c\u60a8\u53ef\u4ee5\u76f4\u63a5\u5728\u5bb9\u5668\u5185\u8fd0\u884cMinerU\u76f8\u5173\u547d\u4ee4\u6765\u4f7f\u7528MinerU\u7684\u529f\u80fd\u3002 \u60a8\u4e5f\u53ef\u4ee5\u76f4\u63a5\u901a\u8fc7\u66ff\u6362/bin/bash\u4e3a\u670d\u52a1\u542f\u52a8\u547d\u4ee4\u6765\u542f\u52a8MinerU\u670d\u52a1\uff0c\u8be6\u7ec6\u8bf4\u660e\u8bf7\u53c2\u8003\u901a\u8fc7\u547d\u4ee4\u542f\u52a8\u670d\u52a1\u3002

Note

\u7531\u4e8e310p\u52a0\u901f\u5361\u4e0d\u652f\u6301\u56fe\u6a21\u5f0f\u4e0ebf16\u7cbe\u5ea6\uff0c\u56e0\u6b64\u5728\u4f7f\u7528\u8be5\u52a0\u901f\u5361\u65f6\uff0c\u6267\u884c\u4efb\u610f\u4e0evllm\u76f8\u5173\u547d\u4ee4\u9700\u8ffd\u52a0--enforce-eager --dtype float16\u53c2\u6570\u3002 \u4f8b\u5982:

mineru-openai-server --port 30000 --enforce-eager --dtype float16\n

"},{"location":"zh/usage/acceleration_cards/Ascend/#4","title":"4. \u6ce8\u610f\u4e8b\u9879","text":"

\u4e0d\u540c\u73af\u5883\u4e0b\uff0cMinerU\u5bf9Ascend\u52a0\u901f\u5361\u7684\u652f\u6301\u60c5\u51b5\u5982\u4e0b\u8868\u6240\u793a\uff1a

\u4f7f\u7528\u573a\u666f \u5bb9\u5668\u73af\u5883 vllm lmdeploy \u547d\u4ee4\u884c\u5de5\u5177(mineru) pipeline \ud83d\udfe2 \ud83d\udfe2 <vlm/hybrid>-auto-engine \ud83d\udfe2 \ud83d\udfe2 <vlm/hybrid>-http-client \ud83d\udfe2 \ud83d\udfe2 fastapi\u670d\u52a1(mineru-api) pipeline \ud83d\udfe2 \ud83d\udfe2 <vlm/hybrid>-auto-engine \ud83d\udfe2 \ud83d\udfe2 <vlm/hybrid>-http-client \ud83d\udfe2 \ud83d\udfe2 gradio\u754c\u9762(mineru-gradio) pipeline \ud83d\udfe2 \ud83d\udfe2 <vlm/hybrid>-auto-engine \ud83d\udfe2 \ud83d\udfe2 <vlm/hybrid>-http-client \ud83d\udfe2 \ud83d\udfe2 openai-server\u670d\u52a1\uff08mineru-openai-server\uff09 \ud83d\udfe2 \ud83d\udfe2 \u6570\u636e\u5e76\u884c (--data-parallel-size/--dp) \ud83d\udfe2 \ud83d\udd34

\u6ce8\uff1a \ud83d\udfe2: \u652f\u6301\uff0c\u8fd0\u884c\u8f83\u7a33\u5b9a\uff0c\u7cbe\u5ea6\u4e0eNvidia GPU\u57fa\u672c\u4e00\u81f4 \ud83d\udfe1: \u652f\u6301\u4f46\u8f83\u4e0d\u7a33\u5b9a\uff0c\u5728\u67d0\u4e9b\u573a\u666f\u4e0b\u53ef\u80fd\u51fa\u73b0\u5f02\u5e38\uff0c\u6216\u7cbe\u5ea6\u5b58\u5728\u4e00\u5b9a\u5dee\u5f02 \ud83d\udd34: \u4e0d\u652f\u6301\uff0c\u65e0\u6cd5\u8fd0\u884c\uff0c\u6216\u7cbe\u5ea6\u5b58\u5728\u8f83\u5927\u5dee\u5f02

Tip

NPU\u52a0\u901f\u5361\u6307\u5b9a\u53ef\u7528\u52a0\u901f\u5361\u7684\u65b9\u5f0f\u4e0eNVIDIA GPU\u7c7b\u4f3c\uff0c\u8bf7\u53c2\u8003ASCEND_RT_VISIBLE_DEVICES
\u5728Ascend\u5e73\u53f0\u53ef\u4ee5\u901a\u8fc7npu-smi info\u547d\u4ee4\u67e5\u770b\u52a0\u901f\u5361\u7684\u4f7f\u7528\u60c5\u51b5\uff0c\u5e76\u6839\u636e\u9700\u8981\u6307\u5b9a\u7a7a\u95f2\u7684\u52a0\u901f\u5361ID\u4ee5\u907f\u514d\u8d44\u6e90\u51b2\u7a81\u3002

"},{"location":"zh/usage/acceleration_cards/Biren/","title":"Biren","text":""},{"location":"zh/usage/acceleration_cards/Biren/#1","title":"1. \u6d4b\u8bd5\u5e73\u53f0","text":"

\u4ee5\u4e0b\u4e3a\u672c\u6307\u5357\u6d4b\u8bd5\u4f7f\u7528\u7684\u5e73\u53f0\u4fe1\u606f\uff0c\u4f9b\u53c2\u8003\uff1a

os: Ubuntu 22.04.4 LTS\ncpu: Intel x86-64\ngpu: Biren 106C\ndriver: 1.10.0\ndocker: 28.0.4\n

"},{"location":"zh/usage/acceleration_cards/Biren/#2","title":"2. \u73af\u5883\u51c6\u5907","text":""},{"location":"zh/usage/acceleration_cards/Biren/#21-vllm","title":"2.1 \u4e0b\u8f7d\u5e76\u52a0\u8f7d\u955c\u50cf \uff08vllm\uff09","text":"

wget http://birentech.com/xxx/MinerU/mineru-vllm.tar \u94fe\u63a5\u83b7\u53d6\u8bf7\u8054\u7cfb\u58c1\u4ede\u5185\u90e8\u4eba\u5458\uff08\u90ae\u7bb1\uff1aMonaLiu@birentech.com\uff09\ndocker load -i mineru-vllm.tar\n

"},{"location":"zh/usage/acceleration_cards/Biren/#3-docker","title":"3. \u542f\u52a8 Docker \u5bb9\u5668","text":"

docker run -it --name mineru_docker \\\n    --privileged \\\n    --network=host \\\n    --shm-size=100G \\\n    -e MINERU_MODEL_SOURCE=local \\\n    -e MINERU_DEVICE_MODEL=supa \\\n    -e SHAPE_TRANSFORM_GRANK=true \\\n    mineru:biren-vllm-latest \\\n    /bin/bash\n

"},{"location":"zh/usage/acceleration_cards/Biren/#4","title":"4. \u6ce8\u610f\u4e8b\u9879","text":"

\u4e0d\u540c\u73af\u5883\u4e0b\uff0cMinerU\u5bf9Biren\u52a0\u901f\u5361\u7684\u652f\u6301\u60c5\u51b5\u5982\u4e0b\u8868\u6240\u793a\uff1a

\u4f7f\u7528\u573a\u666f \u5bb9\u5668\u73af\u5883 vllm \u547d\u4ee4\u884c\u5de5\u5177(mineru) pipeline \ud83d\udfe2 <vlm/hybrid>-auto-engine \ud83d\udfe2 <vlm/hybrid>-http-client \ud83d\udfe2 fastapi\u670d\u52a1(mineru-api) pipeline \ud83d\udfe2 <vlm/hybrid>-auto-engine \ud83d\udfe2 <vlm/hybrid>-http-client \ud83d\udfe2 gradio\u754c\u9762(mineru-gradio) pipeline \ud83d\udfe2 <vlm/hybrid>-auto-engine \ud83d\udfe2 <vlm/hybrid>-http-client \ud83d\udfe2 openai-server\u670d\u52a1\uff08mineru-openai-server\uff09 \ud83d\udfe2 \u6570\u636e\u5e76\u884c (--data-parallel-size) \ud83d\udd34

Tip

Biren\u52a0\u901f\u5361\u6307\u5b9a\u53ef\u7528\u52a0\u901f\u5361\u7684\u65b9\u5f0f\u4e0eNVIDIA GPU\u7c7b\u4f3c\uff0c\u8bf7\u53c2\u8003\u4f7f\u7528\u6307\u5b9aGPU\u8bbe\u5907\u7ae0\u8282\u8bf4\u660e, \u5c06\u73af\u5883\u53d8\u91cfCUDA_VISIBLE_DEVICES\u66ff\u6362\u4e3aSUPA_VISIBLE_DEVICES\u5373\u53ef\u3002
\u5728\u58c1\u4ede\u5e73\u53f0\u53ef\u4ee5\u901a\u8fc7brsmi\u547d\u4ee4\u67e5\u770b\u52a0\u901f\u5361\u7684\u4f7f\u7528\u60c5\u51b5\uff0c\u5e76\u6839\u636e\u9700\u8981\u6307\u5b9a\u7a7a\u95f2\u7684\u52a0\u901f\u5361ID\u4ee5\u907f\u514d\u8d44\u6e90\u51b2\u7a81\u3002

"},{"location":"zh/usage/acceleration_cards/Cambricon/","title":"Cambricon","text":""},{"location":"zh/usage/acceleration_cards/Cambricon/#1","title":"1. \u6d4b\u8bd5\u5e73\u53f0","text":"

\u4ee5\u4e0b\u4e3a\u672c\u6307\u5357\u6d4b\u8bd5\u4f7f\u7528\u7684\u5e73\u53f0\u4fe1\u606f\uff0c\u4f9b\u53c2\u8003\uff1a

os: Ubuntu 22.04.5 LTS  \ncpu: Hygon Hygon C86 7490\nmlu: MLU590-M9D\ndriver: v6.2.11\ndocker: 28.3.0\n

"},{"location":"zh/usage/acceleration_cards/Cambricon/#2","title":"2. \u73af\u5883\u51c6\u5907","text":"

Note

Cambricon\u52a0\u901f\u5361\u652f\u6301\u4f7f\u7528lmdeploy\u6216vllm\u8fdb\u884cVLM\u6a21\u578b\u63a8\u7406\u52a0\u901f\u3002\u8bf7\u6839\u636e\u5b9e\u9645\u9700\u6c42\u9009\u62e9\u5b89\u88c5\u548c\u4f7f\u7528\u5176\u4e2d\u4e4b\u4e00:

"},{"location":"zh/usage/acceleration_cards/Cambricon/#21-dockerfile-lmdeploy","title":"2.1 \u4f7f\u7528 Dockerfile \u6784\u5efa\u955c\u50cf \uff08lmdeploy\uff09","text":"

wget https://gcore.jsdelivr.net/gh/opendatalab/MinerU@master/docker/china/mlu.Dockerfile\ndocker build --network=host -t mineru:mlu-lmdeploy-latest -f mlu.Dockerfile .\n

"},{"location":"zh/usage/acceleration_cards/Cambricon/#22-dockerfile-vllm","title":"2.2 \u4f7f\u7528 Dockerfile \u6784\u5efa\u955c\u50cf \uff08vllm\uff09","text":"

wget https://gcore.jsdelivr.net/gh/opendatalab/MinerU@master/docker/china/mlu.Dockerfile\n# \u5c06\u57fa\u7840\u955c\u50cf\u4ece lmdeploy \u5207\u6362\u4e3a vllm\nsed -i -e '3,4s/^/# /' -e '6,7s/^# //' mlu.Dockerfile\ndocker build --network=host -t mineru:mlu-vllm-latest -f mlu.Dockerfile .\n

"},{"location":"zh/usage/acceleration_cards/Cambricon/#3-docker","title":"3. \u542f\u52a8 Docker \u5bb9\u5668","text":"

docker run --name mineru_docker \\\n   --privileged \\\n   --ipc=host \\\n   --network=host \\\n   --shm-size=400g \\\n   --ulimit memlock=-1 \\\n   -v /dev:/dev \\\n   -v /lib/modules:/lib/modules:ro \\\n   -v /usr/bin/cnmon:/usr/bin/cnmon \\\n   -e MINERU_MODEL_SOURCE=local \\\n   -e MINERU_LMDEPLOY_DEVICE=camb \\\n   -it mineru:mlu-lmdeploy-latest \\\n   /bin/bash\n

Tip

\u8bf7\u6839\u636e\u5b9e\u9645\u60c5\u51b5\u9009\u62e9\u4f7f\u7528vllm\u6216lmdeploy\u7248\u672c\u7684\u955c\u50cf\uff0c\u5982\u9700\u4f7f\u7528vllm,\u8bf7\u6267\u884c\u4ee5\u4e0b\u64cd\u4f5c\uff1a

\u66ff\u6362\u4e0a\u8ff0\u547d\u4ee4\u4e2d\u7684mineru:mlu-lmdeploy-latest\u4e3amineru:mlu-vllm-latest
\u8fdb\u5165\u5bb9\u5668\u540e\uff0c\u901a\u8fc7\u4ee5\u4e0b\u547d\u4ee4\u5207\u6362venv\u73af\u5883\uff1a
```
source /torch/venv3/pytorch_infer/bin/activate\n
```
\u5207\u6362\u6210\u529f\u540e\uff0c\u60a8\u53ef\u4ee5\u5728\u547d\u4ee4\u884c\u524d\u770b\u5230(pytorch_infer)\u7684\u6807\u8bc6\uff0c\u8fd9\u8868\u793a\u60a8\u5df2\u6210\u529f\u8fdb\u5165vllm\u7684\u865a\u62df\u73af\u5883\u3002

"},{"location":"zh/usage/acceleration_cards/Cambricon/#4","title":"4. \u6ce8\u610f\u4e8b\u9879","text":"

Note

\u517c\u5bb9\u6027\u8bf4\u660e\uff1a\u7531\u4e8e\u5bd2\u6b66\u7eaa\uff08Cambricon\uff09\u76ee\u524d\u5bf9 vLLM v1 \u5f15\u64ce\u7684\u652f\u6301\u5c1a\u5f85\u5b8c\u5584\uff0cMinerU \u73b0\u9636\u6bb5\u91c7\u7528 v0 \u5f15\u64ce\u4f5c\u4e3a\u9002\u914d\u65b9\u6848\u3002 \u53d7\u6b64\u9650\u5236\uff0cvLLM \u7684\u5f02\u6b65\u5f15\u64ce\uff08Async Engine\uff09\u529f\u80fd\u5b58\u5728\u517c\u5bb9\u6027\u95ee\u9898\uff0c\u53ef\u80fd\u5bfc\u81f4\u90e8\u5206\u4f7f\u7528\u573a\u666f\u65e0\u6cd5\u6b63\u5e38\u8fd0\u884c\u3002 \u6211\u4eec\u5c06\u6301\u7eed\u8ddf\u8fdb\u5bd2\u6b66\u7eaa\u5bf9 vLLM v1 \u5f15\u64ce\u7684\u652f\u6301\u8fdb\u5c55\uff0c\u5e76\u53ca\u65f6\u5728 MinerU \u4e2d\u8fdb\u884c\u76f8\u5e94\u7684\u9002\u914d\u4e0e\u4f18\u5316\u3002

\u4e0d\u540c\u73af\u5883\u4e0b\uff0cMinerU\u5bf9Cambricon\u52a0\u901f\u5361\u7684\u652f\u6301\u60c5\u51b5\u5982\u4e0b\u8868\u6240\u793a\uff1a

Tip

lmdeploy\u9ec4\u706f\u95ee\u9898\u4e3a\u4e0d\u80fd\u8f93\u5165\u6587\u4ef6\u5939\u4f7f\u7528\u6279\u91cf\u89e3\u6790\u529f\u80fd\uff0c\u8f93\u5165\u5355\u4e2a\u6587\u4ef6\u65f6\u8868\u73b0\u6b63\u5e38\u3002
vllm\u9ec4\u706f\u95ee\u9898\u4e3a\u5728\u7cbe\u5ea6\u672a\u5bf9\u9f50\uff0c\u5728\u90e8\u5206\u573a\u666f\u4e0b\u53ef\u80fd\u51fa\u73b0\u9884\u671f\u5916\u7ed3\u679c\u3002

\u4f7f\u7528\u573a\u666f \u5bb9\u5668\u73af\u5883 vllm lmdeploy \u547d\u4ee4\u884c\u5de5\u5177(mineru) pipeline \ud83d\udfe2 \ud83d\udfe2 <vlm/hybrid>-auto-engine \ud83d\udfe1 \ud83d\udfe1 <vlm/hybrid>-http-client \ud83d\udfe1 \ud83d\udfe2 fastapi\u670d\u52a1(mineru-api) pipeline \ud83d\udfe2 \ud83d\udfe2 <vlm/hybrid>-auto-engine \ud83d\udd34 \ud83d\udfe2 <vlm/hybrid>-http-client \ud83d\udfe1 \ud83d\udfe2 gradio\u754c\u9762(mineru-gradio) pipeline \ud83d\udfe2 \ud83d\udfe2 <vlm/hybrid>-auto-engine \ud83d\udd34 \ud83d\udfe2 <vlm/hybrid>-http-client \ud83d\udfe1 \ud83d\udfe2 openai-server\u670d\u52a1\uff08mineru-openai-server\uff09 \ud83d\udfe1 \ud83d\udfe2 \u6570\u636e\u5e76\u884c (--data-parallel-size/--dp) \ud83d\udd34 \ud83d\udd34

Tip

Cambricon\u52a0\u901f\u5361\u6307\u5b9a\u53ef\u7528\u52a0\u901f\u5361\u7684\u65b9\u5f0f\u4e0eNVIDIA GPU\u7c7b\u4f3c\uff0c\u8bf7\u53c2\u8003\u4f7f\u7528\u6307\u5b9aGPU\u8bbe\u5907\u7ae0\u8282\u8bf4\u660e, \u5c06\u73af\u5883\u53d8\u91cfCUDA_VISIBLE_DEVICES\u66ff\u6362\u4e3aMLU_VISIBLE_DEVICES\u5373\u53ef\u3002
\u5728Cambricon\u5e73\u53f0\u53ef\u4ee5\u901a\u8fc7cnmon\u547d\u4ee4\u67e5\u770b\u52a0\u901f\u5361\u7684\u4f7f\u7528\u60c5\u51b5\uff0c\u5e76\u6839\u636e\u9700\u8981\u6307\u5b9a\u7a7a\u95f2\u7684\u52a0\u901f\u5361ID\u4ee5\u907f\u514d\u8d44\u6e90\u51b2\u7a81\u3002

"},{"location":"zh/usage/acceleration_cards/Enflame/","title":"Enflame","text":""},{"location":"zh/usage/acceleration_cards/Enflame/#1","title":"1. \u6d4b\u8bd5\u5e73\u53f0","text":"

\u4ee5\u4e0b\u4e3a\u672c\u6307\u5357\u6d4b\u8bd5\u4f7f\u7528\u7684\u5e73\u53f0\u4fe1\u606f\uff0c\u4f9b\u53c2\u8003\uff1a

os: Ubuntu 22.04.4 LTS  \ncpu: Intel x86-64\ngcu: Enflame S60 \ndriver: 1.7.0.9\ndocker: 28.0.1\n

"},{"location":"zh/usage/acceleration_cards/Enflame/#2","title":"2. \u73af\u5883\u51c6\u5907","text":""},{"location":"zh/usage/acceleration_cards/Enflame/#21-dockerfile","title":"2.1 \u4f7f\u7528 Dockerfile \u6784\u5efa\u955c\u50cf","text":"

wget https://gcore.jsdelivr.net/gh/opendatalab/MinerU@master/docker/china/gcu.Dockerfile\ndocker build --network=host -t mineru:gcu-vllm-latest -f gcu.Dockerfile .\n

"},{"location":"zh/usage/acceleration_cards/Enflame/#3-docker","title":"3. \u542f\u52a8 Docker \u5bb9\u5668","text":"

docker run -u root --name mineru_docker \\\n    --network=host \\\n    --ipc=host \\\n    --privileged \\\n    -e MINERU_MODEL_SOURCE=local \\\n    -it mineru:gcu-vllm-latest \\\n    /bin/bash\n

"},{"location":"zh/usage/acceleration_cards/Enflame/#4","title":"4. \u6ce8\u610f\u4e8b\u9879","text":"

\u4e0d\u540c\u73af\u5883\u4e0b\uff0cMinerU\u5bf9Enflame\u52a0\u901f\u5361\u7684\u652f\u6301\u60c5\u51b5\u5982\u4e0b\u8868\u6240\u793a\uff1a

Tip

GCU\u52a0\u901f\u5361\u6307\u5b9a\u53ef\u7528\u52a0\u901f\u5361\u7684\u65b9\u5f0f\u4e0eNVIDIA GPU\u7c7b\u4f3c\uff0c\u8bf7\u53c2\u8003\u4f7f\u7528\u6307\u5b9aGPU\u8bbe\u5907\u7ae0\u8282\u8bf4\u660e, \u5c06\u73af\u5883\u53d8\u91cfCUDA_VISIBLE_DEVICES\u66ff\u6362\u4e3aTOPS_VISIBLE_DEVICES\u5373\u53ef\u3002
\u5728Enflame\u5e73\u53f0\u53ef\u4ee5\u901a\u8fc7efsmi\u547d\u4ee4\u67e5\u770b\u52a0\u901f\u5361\u7684\u4f7f\u7528\u60c5\u51b5\uff0c\u5e76\u6839\u636e\u9700\u8981\u6307\u5b9a\u7a7a\u95f2\u7684\u52a0\u901f\u5361ID\u4ee5\u907f\u514d\u8d44\u6e90\u51b2\u7a81\u3002

"},{"location":"zh/usage/acceleration_cards/Hygon/","title":"Hygon","text":""},{"location":"zh/usage/acceleration_cards/Hygon/#1","title":"1. \u6d4b\u8bd5\u5e73\u53f0","text":"

\u4ee5\u4e0b\u4e3a\u672c\u6307\u5357\u6d4b\u8bd5\u4f7f\u7528\u7684\u5e73\u53f0\u4fe1\u606f\uff0c\u4f9b\u53c2\u8003\uff1a

os: Ubuntu 22.04.3 LTS  \ncpu: Hygon C86-4G(x86-64)\ndcu: BW200\ndriver: 6.3.13-V1.12.0a\ndocker: 20.10.24\n

"},{"location":"zh/usage/acceleration_cards/Hygon/#2","title":"2. \u73af\u5883\u51c6\u5907","text":""},{"location":"zh/usage/acceleration_cards/Hygon/#21-dockerfile","title":"2.1 \u4f7f\u7528 Dockerfile \u6784\u5efa\u955c\u50cf","text":"

wget https://gcore.jsdelivr.net/gh/opendatalab/MinerU@master/docker/china/dcu.Dockerfile\ndocker build --network=host -t mineru:dcu-vllm-latest -f dcu.Dockerfile .\n

"},{"location":"zh/usage/acceleration_cards/Hygon/#3-docker","title":"3. \u542f\u52a8 Docker \u5bb9\u5668","text":"

docker run -u root --name mineru_docker \\\n    --network=host \\\n    --ipc=host \\\n    --shm-size=16G \\\n    --device=/dev/kfd \\\n    --device=/dev/mkfd \\\n    --device=/dev/dri \\\n    -v /opt/hyhal:/opt/hyhal \\\n    --group-add video \\\n    --cap-add=SYS_PTRACE \\\n    --security-opt seccomp=unconfined \\\n    -e MINERU_MODEL_SOURCE=local \\\n    -it mineru:dcu-vllm-latest \\\n    /bin/bash\n

"},{"location":"zh/usage/acceleration_cards/Hygon/#4","title":"4. \u6ce8\u610f\u4e8b\u9879","text":"

\u4e0d\u540c\u73af\u5883\u4e0b\uff0cMinerU\u5bf9Hygon\u52a0\u901f\u5361\u7684\u652f\u6301\u60c5\u51b5\u5982\u4e0b\u8868\u6240\u793a\uff1a

Tip

DCU\u52a0\u901f\u5361\u6307\u5b9a\u53ef\u7528\u52a0\u901f\u5361\u7684\u65b9\u5f0f\u4e0eAMD GPU\u7c7b\u4f3c\uff0c\u8bf7\u53c2\u8003GPU isolation techniques
\u5728Hygon\u5e73\u53f0\u53ef\u4ee5\u901a\u8fc7hy-smi\u547d\u4ee4\u67e5\u770b\u52a0\u901f\u5361\u7684\u4f7f\u7528\u60c5\u51b5\uff0c\u5e76\u6839\u636e\u9700\u8981\u6307\u5b9a\u7a7a\u95f2\u7684\u52a0\u901f\u5361ID\u4ee5\u907f\u514d\u8d44\u6e90\u51b2\u7a81\u3002

"},{"location":"zh/usage/acceleration_cards/IluvatarCorex/","title":"IluvatarCorex","text":""},{"location":"zh/usage/acceleration_cards/IluvatarCorex/#1","title":"1. \u6d4b\u8bd5\u5e73\u53f0","text":"

\u4ee5\u4e0b\u4e3a\u672c\u6307\u5357\u6d4b\u8bd5\u4f7f\u7528\u7684\u5e73\u53f0\u4fe1\u606f\uff0c\u4f9b\u53c2\u8003\uff1a

os: Ubuntu 22.04.5 LTS  \ncpu: Intel x86-64\ngpu: Iluvatar BI-V150\ndriver: 4.4.0\ndocker: 28.1.1\n

"},{"location":"zh/usage/acceleration_cards/IluvatarCorex/#2","title":"2. \u73af\u5883\u51c6\u5907","text":""},{"location":"zh/usage/acceleration_cards/IluvatarCorex/#21-dockerfile","title":"2.1 \u4f7f\u7528 Dockerfile \u6784\u5efa\u955c\u50cf","text":"

wget https://gcore.jsdelivr.net/gh/opendatalab/MinerU@master/docker/china/corex.Dockerfile\ndocker build --network=host -t mineru:corex-vllm-latest -f corex.Dockerfile .\n

"},{"location":"zh/usage/acceleration_cards/IluvatarCorex/#3-docker","title":"3. \u542f\u52a8 Docker \u5bb9\u5668","text":"

docker run --name mineru_docker \\\n   -v /usr/src:/usr/src \\\n   -v /lib/modules:/lib/modules \\\n   -v /dev:/dev \\\n   --privileged \\\n   --cap-add=ALL \\\n   --pid=host \\\n   --group-add video \\\n   --network=host \\\n   --shm-size '400gb' \\\n   --ulimit memlock=-1 \\\n   --security-opt seccomp=unconfined \\\n   --security-opt apparmor=unconfined \\\n   -e VLLM_ENFORCE_CUDA_GRAPH=1 \\\n   -e MINERU_MODEL_SOURCE=local \\\n   -e MINERU_VLLM_DEVICE=corex \\\n   -it mineru:corex-vllm-latest \\\n   /bin/bash\n

"},{"location":"zh/usage/acceleration_cards/IluvatarCorex/#4","title":"4. \u6ce8\u610f\u4e8b\u9879","text":"

Tip

\u76ee\u524dIluvatar\u65b9\u6848\u4f7f\u7528vllm\u4f5c\u4e3a\u63a8\u7406\u5f15\u64ce\u65f6\uff0c\u53ef\u80fd\u51fa\u73b0\u670d\u52a1\u505c\u6b62\u540e\u663e\u5b58\u65e0\u6cd5\u6b63\u5e38\u91ca\u653e\u7684\u95ee\u9898\uff0c\u5982\u679c\u9047\u5230\u8be5\u95ee\u9898\uff0c\u8bf7\u91cd\u542fDocker\u5bb9\u5668\u4ee5\u91ca\u653e\u663e\u5b58\u3002

\u4e0d\u540c\u73af\u5883\u4e0b\uff0cMinerU\u5bf9Iluvatar\u52a0\u901f\u5361\u7684\u652f\u6301\u60c5\u51b5\u5982\u4e0b\u8868\u6240\u793a\uff1a

Tip

Iluvatar\u52a0\u901f\u5361\u6307\u5b9a\u53ef\u7528\u52a0\u901f\u5361\u7684\u65b9\u5f0f\u4e0eNVIDIA GPU\u7c7b\u4f3c\uff0c\u8bf7\u53c2\u8003\u4f7f\u7528\u6307\u5b9aGPU\u8bbe\u5907\u7ae0\u8282\u8bf4\u660e
\u5728Iluvatar\u5e73\u53f0\u53ef\u4ee5\u901a\u8fc7ixsmi\u547d\u4ee4\u67e5\u770b\u52a0\u901f\u5361\u7684\u4f7f\u7528\u60c5\u51b5\uff0c\u5e76\u6839\u636e\u9700\u8981\u6307\u5b9a\u7a7a\u95f2\u7684\u52a0\u901f\u5361ID\u4ee5\u907f\u514d\u8d44\u6e90\u51b2\u7a81\u3002

"},{"location":"zh/usage/acceleration_cards/Kunlunxin/","title":"Kunlunxin","text":""},{"location":"zh/usage/acceleration_cards/Kunlunxin/#1","title":"1. \u6d4b\u8bd5\u5e73\u53f0","text":"

\u4ee5\u4e0b\u4e3a\u672c\u6307\u5357\u6d4b\u8bd5\u4f7f\u7528\u7684\u5e73\u53f0\u4fe1\u606f\uff0c\u4f9b\u53c2\u8003\uff1a

os: Ubuntu 22.04.5 LTS  \ncpu: Intel x86-64\nxpu: P800\ndriver: 515.58\ndocker: 20.10.5\n

"},{"location":"zh/usage/acceleration_cards/Kunlunxin/#2","title":"2. \u73af\u5883\u51c6\u5907","text":""},{"location":"zh/usage/acceleration_cards/Kunlunxin/#21-dockerfile-vllm","title":"2.1 \u4f7f\u7528 Dockerfile \u6784\u5efa\u955c\u50cf \uff08vllm\uff09","text":"

wget https://gcore.jsdelivr.net/gh/opendatalab/MinerU@master/docker/china/kxpu.Dockerfile\ndocker build --network=host -t mineru:kxpu-vllm-latest -f kxpu.Dockerfile .\n

"},{"location":"zh/usage/acceleration_cards/Kunlunxin/#3-docker","title":"3. \u542f\u52a8 Docker \u5bb9\u5668","text":"

docker run -u root --name mineru_docker \\\n    --device=/dev/xpu0:/dev/xpu0 \\\n    --device=/dev/xpu1:/dev/xpu1 \\\n    --device=/dev/xpu2:/dev/xpu2 \\\n    --device=/dev/xpu3:/dev/xpu3 \\\n    --device=/dev/xpu4:/dev/xpu4 \\\n    --device=/dev/xpu5:/dev/xpu5 \\\n    --device=/dev/xpu6:/dev/xpu6 \\\n    --device=/dev/xpu7:/dev/xpu7 \\\n    --device=/dev/xpuctrl:/dev/xpuctrl \\\n    --net=host \\\n    --cap-add=SYS_PTRACE --security-opt seccomp=unconfined \\\n    --tmpfs /dev/shm:rw,nosuid,nodev,exec,size=32g \\\n    --cap-add=SYS_PTRACE \\\n    -v /home/users/vllm-kunlun:/home/vllm-kunlun \\\n    -v /usr/local/bin/xpu-smi:/usr/local/bin/xpu-smi \\\n    -w /workspace \\\n    -e MINERU_MODEL_SOURCE=local \\\n    -e MINERU_FORMULA_CH_SUPPORT=true \\\n    -e MINERU_VLLM_DEVICE=kxpu \\\n    -it mineru:kxpu-vllm-latest \\\n    /bin/bash\n

"},{"location":"zh/usage/acceleration_cards/Kunlunxin/#4","title":"4. \u6ce8\u610f\u4e8b\u9879","text":"

\u4e0d\u540c\u73af\u5883\u4e0b\uff0cMinerU\u5bf9Kunlunxin\u52a0\u901f\u5361\u7684\u652f\u6301\u60c5\u51b5\u5982\u4e0b\u8868\u6240\u793a\uff1a

Tip

Kunlunxin\u52a0\u901f\u5361\u6307\u5b9a\u53ef\u7528\u52a0\u901f\u5361\u7684\u65b9\u5f0f\u4e0eNVIDIA GPU\u7c7b\u4f3c\uff0c\u8bf7\u53c2\u8003\u4f7f\u7528\u6307\u5b9aGPU\u8bbe\u5907\u7ae0\u8282\u8bf4\u660e, \u5c06\u73af\u5883\u53d8\u91cfCUDA_VISIBLE_DEVICES\u66ff\u6362\u4e3aXPU_VISIBLE_DEVICES\u5373\u53ef\u3002
\u5728Kunlunxin\u5e73\u53f0\u53ef\u4ee5\u901a\u8fc7xpu-smi\u547d\u4ee4\u67e5\u770b\u52a0\u901f\u5361\u7684\u4f7f\u7528\u60c5\u51b5\uff0c\u5e76\u6839\u636e\u9700\u8981\u6307\u5b9a\u7a7a\u95f2\u7684\u52a0\u901f\u5361ID\u4ee5\u907f\u514d\u8d44\u6e90\u51b2\u7a81\u3002

"},{"location":"zh/usage/acceleration_cards/METAX/","title":"METAX","text":""},{"location":"zh/usage/acceleration_cards/METAX/#1","title":"1. \u6d4b\u8bd5\u5e73\u53f0","text":"

\u4ee5\u4e0b\u4e3a\u672c\u6307\u5357\u6d4b\u8bd5\u4f7f\u7528\u7684\u5e73\u53f0\u4fe1\u606f\uff0c\u4f9b\u53c2\u8003\uff1a

os: Ubuntu 22.04   \ncpu: INTEL x86_64\ngpu: C500  \ndriver: 2.12.13\ndocker: 28.1.1\n

"},{"location":"zh/usage/acceleration_cards/METAX/#2","title":"2. \u73af\u5883\u51c6\u5907","text":"

Note

maca\u52a0\u901f\u5361\u652f\u6301\u4f7f\u7528vllm\u6216lmdeploy\u8fdb\u884cVLM\u6a21\u578b\u63a8\u7406\u52a0\u901f\u3002\u8bf7\u6839\u636e\u5b9e\u9645\u9700\u6c42\u9009\u62e9\u5b89\u88c5\u548c\u4f7f\u7528\u5176\u4e2d\u4e4b\u4e00:

"},{"location":"zh/usage/acceleration_cards/METAX/#21-metaxvllm","title":"2.1 \u4f7f\u7528metax\u5b98\u65b9\u955c\u50cf\u4f5c\u4e3a\u57fa\u7840\u955c\u50cf\u6784\u5efavllm\u73af\u5883\u955c\u50cf","text":"

\u4ecemetax\u5b98\u65b9\u4ed3\u5e93\u62c9\u53d6\u57fa\u7840\u955c\u50cf
- 1.1 \u955c\u50cf\u83b7\u53d6\u5730\u5740\uff1ahttps://developer.metax-tech.com/softnova/docker
- 1.2 \u5728\u955c\u50cf\u7f51\u7ad9\u9009\u62e9AI\u5206\u7c7b\uff0c\u8f6f\u4ef6\u5305\u7c7b\u578b\u9009\u62e9vllm\uff0c\u64cd\u4f5c\u7cfb\u7edf\u9009\u62e9ubuntu
- 1.3 \u627e\u5230vllm:maca.ai3.1.0.7-torch2.6-py310-ubuntu22.04-amd64\u955c\u50cf\uff0c\u590d\u5236\u62c9\u53d6\u547d\u4ee4\u5e76\u5728\u672c\u5730\u7ec8\u7aef\u6267\u884c

\u4f7f\u7528 Dockerfile \u6784\u5efa\u955c\u50cf \uff08vllm\uff09

wget https://gcore.jsdelivr.net/gh/opendatalab/MinerU@master/docker/china/maca.Dockerfile\ndocker build --network=host -t mineru:maca-vllm-latest -f maca.Dockerfile .\n

"},{"location":"zh/usage/acceleration_cards/METAX/#22-dockerfile-lmdeploy","title":"2.2 \u4f7f\u7528 Dockerfile \u6784\u5efa\u955c\u50cf \uff08lmdeploy\uff09","text":"

wget https://gcore.jsdelivr.net/gh/opendatalab/MinerU@master/docker/china/maca.Dockerfile\n# \u5c06\u57fa\u7840\u955c\u50cf\u4ece vllm \u5207\u6362\u4e3a lmdeploy\nsed -i '3s/^/# /' maca.Dockerfile && sed -i '5s/^# //' maca.Dockerfile\ndocker build --network=host -t mineru:maca-lmdeploy-latest -f maca.Dockerfile .\n

"},{"location":"zh/usage/acceleration_cards/METAX/#3-docker","title":"3. \u542f\u52a8 Docker \u5bb9\u5668","text":"

docker run --ipc host \\\n   --cap-add SYS_PTRACE \\\n   --privileged=true \\\n   --device=/dev/mem \\\n   --device=/dev/dri \\\n   --device=/dev/mxcd \\\n   --device=/dev/infiniband \\\n   --group-add video \\\n   --network=host \\\n   --shm-size '100gb' \\\n   --ulimit memlock=-1 \\\n   --security-opt seccomp=unconfined \\\n   --security-opt apparmor=unconfined \\\n   --name mineru_docker \\\n   -v /datapool:/datapool \\\n   -e MINERU_MODEL_SOURCE=local \\\n   -e MINERU_LMDEPLOY_DEVICE=maca \\\n   -it mineru:maca-vllm-latest \\\n   /bin/bash\n

Tip

\u8bf7\u6839\u636e\u5b9e\u9645\u60c5\u51b5\u9009\u62e9\u4f7f\u7528vllm\u6216lmdeploy\u7248\u672c\u7684\u955c\u50cf\uff0c\u5982\u9700\u4f7f\u7528lmdeploy\uff0c\u66ff\u6362\u4e0a\u8ff0\u547d\u4ee4\u4e2d\u7684mineru:maca-vllm-latest\u4e3amineru:maca-lmdeploy-latest\u5373\u53ef\u3002

"},{"location":"zh/usage/acceleration_cards/METAX/#4","title":"4. \u6ce8\u610f\u4e8b\u9879","text":"

\u4e0d\u540c\u73af\u5883\u4e0b\uff0cMinerU\u5bf9maca\u52a0\u901f\u5361\u7684\u652f\u6301\u60c5\u51b5\u5982\u4e0b\u8868\u6240\u793a\uff1a

Tip

MACA\u52a0\u901f\u5361\u6307\u5b9a\u53ef\u7528\u52a0\u901f\u5361\u7684\u65b9\u5f0f\u4e0eNVIDIA GPU\u7c7b\u4f3c\uff0c\u8bf7\u53c2\u8003\u4f7f\u7528\u6307\u5b9aGPU\u8bbe\u5907\u7ae0\u8282\u8bf4\u660e\u3002
\u5728METAX\u5e73\u53f0\u53ef\u4ee5\u901a\u8fc7mx-smi\u547d\u4ee4\u67e5\u770b\u52a0\u901f\u5361\u7684\u4f7f\u7528\u60c5\u51b5\uff0c\u5e76\u6839\u636e\u9700\u8981\u6307\u5b9a\u7a7a\u95f2\u7684\u52a0\u901f\u5361ID\u4ee5\u907f\u514d\u8d44\u6e90\u51b2\u7a81\u3002

"},{"location":"zh/usage/acceleration_cards/MooreThreads/","title":"MooreThreads","text":""},{"location":"zh/usage/acceleration_cards/MooreThreads/#1","title":"1. \u6d4b\u8bd5\u5e73\u53f0","text":"

\u4ee5\u4e0b\u4e3a\u672c\u6307\u5357\u6d4b\u8bd5\u4f7f\u7528\u7684\u5e73\u53f0\u4fe1\u606f\uff0c\u4f9b\u53c2\u8003\uff1a

os: Ubuntu 22.04.4 LTS  \ncpu: Intel x86-64\ndcu: MTT S4000\ndriver: 3.0.0-rc-KuaE2.0\ndocker: 24.0.7\n

"},{"location":"zh/usage/acceleration_cards/MooreThreads/#2","title":"2. \u73af\u5883\u51c6\u5907","text":""},{"location":"zh/usage/acceleration_cards/MooreThreads/#21-dockerfile","title":"2.1 \u4f7f\u7528 Dockerfile \u6784\u5efa\u955c\u50cf","text":"

wget https://gcore.jsdelivr.net/gh/opendatalab/MinerU@master/docker/china/musa.Dockerfile\ndocker build --network=host -t mineru:musa-vllm-latest -f musa.Dockerfile .\n

"},{"location":"zh/usage/acceleration_cards/MooreThreads/#3-docker","title":"3. \u542f\u52a8 Docker \u5bb9\u5668","text":"

docker run -u root --name mineru_docker \\\n    --network=host \\\n    --ipc=host \\\n    --shm-size=80g \\\n    --privileged \\\n    -e MTHREADS_VISIBLE_DEVICES=all \\\n    -e MINERU_VLLM_DEVICE=musa \\\n    -e MINERU_MODEL_SOURCE=local \\\n    -it mineru:musa-vllm-latest \\\n    /bin/bash\n

"},{"location":"zh/usage/acceleration_cards/MooreThreads/#4","title":"4. \u6ce8\u610f\u4e8b\u9879","text":"

\u4e0d\u540c\u73af\u5883\u4e0b\uff0cMinerU\u5bf9MooreThreads\u52a0\u901f\u5361\u7684\u652f\u6301\u60c5\u51b5\u5982\u4e0b\u8868\u6240\u793a\uff1a

Note

\u517c\u5bb9\u6027\u8bf4\u660e\uff1a\u7531\u4e8e\u6469\u5c14\u7ebf\u7a0b\uff08MooreThreads\uff09\u76ee\u524d\u5bf9 vLLM v1 \u5f15\u64ce\u7684\u652f\u6301\u5c1a\u5f85\u5b8c\u5584\uff0cMinerU \u73b0\u9636\u6bb5\u91c7\u7528 v0 \u5f15\u64ce\u4f5c\u4e3a\u9002\u914d\u65b9\u6848\u3002 \u53d7\u6b64\u9650\u5236\uff0cvLLM \u7684\u5f02\u6b65\u5f15\u64ce\uff08Async Engine\uff09\u529f\u80fd\u5b58\u5728\u517c\u5bb9\u6027\u95ee\u9898\uff0c\u53ef\u80fd\u5bfc\u81f4\u90e8\u5206\u4f7f\u7528\u573a\u666f\u65e0\u6cd5\u6b63\u5e38\u8fd0\u884c\u3002 \u6211\u4eec\u5c06\u6301\u7eed\u8ddf\u8fdb\u6469\u5c14\u7ebf\u7a0b\u5bf9 vLLM v1 \u5f15\u64ce\u7684\u652f\u6301\u8fdb\u5c55\uff0c\u5e76\u53ca\u65f6\u5728 MinerU \u4e2d\u8fdb\u884c\u76f8\u5e94\u7684\u9002\u914d\u4e0e\u4f18\u5316\u3002

\u4f7f\u7528\u573a\u666f \u5bb9\u5668\u73af\u5883 vllm \u547d\u4ee4\u884c\u5de5\u5177(mineru) pipeline \ud83d\udfe2 <vlm/hybrid>-auto-engine \ud83d\udfe2 <vlm/hybrid>-http-client \ud83d\udfe2 fastapi\u670d\u52a1(mineru-api) pipeline \ud83d\udfe2 <vlm/hybrid>-auto-engine \ud83d\udd34 <vlm/hybrid>-http-client \ud83d\udfe2 gradio\u754c\u9762(mineru-gradio) pipeline \ud83d\udfe2 <vlm/hybrid>-auto-engine \ud83d\udd34 <vlm/hybrid>-http-client \ud83d\udfe2 openai-server\u670d\u52a1\uff08mineru-openai-server\uff09 \ud83d\udfe2 \u6570\u636e\u5e76\u884c (--data-parallel-size) \ud83d\udd34

Tip

MooreThreads\u52a0\u901f\u5361\u6307\u5b9a\u53ef\u7528\u52a0\u901f\u5361\u7684\u65b9\u5f0f\u4e0eNVIDIA GPU\u7c7b\u4f3c\uff0c\u8bf7\u53c2\u8003GPU \u679a\u4e3e
\u5728MooreThreads\u5e73\u53f0\u53ef\u4ee5\u901a\u8fc7mthreads-gmi\u547d\u4ee4\u67e5\u770b\u52a0\u901f\u5361\u7684\u4f7f\u7528\u60c5\u51b5\uff0c\u5e76\u6839\u636e\u9700\u8981\u6307\u5b9a\u7a7a\u95f2\u7684\u52a0\u901f\u5361ID\u4ee5\u907f\u514d\u8d44\u6e90\u51b2\u7a81\u3002

"},{"location":"zh/usage/acceleration_cards/THead/","title":"THead","text":""},{"location":"zh/usage/acceleration_cards/THead/#1","title":"1. \u6d4b\u8bd5\u5e73\u53f0","text":"

\u4ee5\u4e0b\u4e3a\u672c\u6307\u5357\u6d4b\u8bd5\u4f7f\u7528\u7684\u5e73\u53f0\u4fe1\u606f\uff0c\u4f9b\u53c2\u8003\uff1a

os: Ubuntu 22.04   \ncpu: INTEL x86_64\nppu: ZW810E  \ndriver: 1.4.0\ndocker: 26.1.4\n

"},{"location":"zh/usage/acceleration_cards/THead/#2","title":"2. \u73af\u5883\u51c6\u5907","text":"

Note

ppu\u52a0\u901f\u5361\u652f\u6301\u4f7f\u7528vllm\u6216lmdeploy\u8fdb\u884cVLM\u6a21\u578b\u63a8\u7406\u52a0\u901f\u3002\u8bf7\u6839\u636e\u5b9e\u9645\u9700\u6c42\u9009\u62e9\u5b89\u88c5\u548c\u4f7f\u7528\u5176\u4e2d\u4e4b\u4e00:

"},{"location":"zh/usage/acceleration_cards/THead/#21-dockerfile-vllm","title":"2.1 \u4f7f\u7528 Dockerfile \u6784\u5efa\u955c\u50cf \uff08vllm\uff09","text":"

wget https://gcore.jsdelivr.net/gh/opendatalab/MinerU@master/docker/china/ppu.Dockerfile\ndocker build --network=host -t mineru:ppu-vllm-latest -f ppu.Dockerfile .\n

"},{"location":"zh/usage/acceleration_cards/THead/#22-dockerfile-lmdeploy","title":"2.2 \u4f7f\u7528 Dockerfile \u6784\u5efa\u955c\u50cf \uff08lmdeploy\uff09","text":"

wget https://gcore.jsdelivr.net/gh/opendatalab/MinerU@master/docker/china/ppu.Dockerfile\n# \u5c06\u57fa\u7840\u955c\u50cf\u4ece vllm \u5207\u6362\u4e3a lmdeploy\nsed -i '3s/^/# /' ppu.Dockerfile && sed -i '5s/^# //' ppu.Dockerfile\ndocker build --network=host -t mineru:ppu-lmdeploy-latest -f ppu.Dockerfile .\n

"},{"location":"zh/usage/acceleration_cards/THead/#3-docker","title":"3. \u542f\u52a8 Docker \u5bb9\u5668","text":"

docker run --privileged=true \\\n  --name mineru_docker \\\n  --device=/dev/alixpu \\\n  --device=/dev/alixpu_ctl \\\n  --ipc=host \\\n  --network=host \\\n  --ulimit memlock=-1 \\\n  --ulimit stack=67108864 \\\n  --shm-size=500g \\\n  -v /mnt:/mnt \\\n  -v /datapool:/datapool \\\n  -v /var/run/docker.sock:/var/run/docker.sock \\\n  -e MINERU_MODEL_SOURCE=local \\\n  -it mineru:ppu-vllm-latest \\\n  /bin/bash\n

Tip

\u8bf7\u6839\u636e\u5b9e\u9645\u60c5\u51b5\u9009\u62e9\u4f7f\u7528vllm\u6216lmdeploy\u7248\u672c\u7684\u955c\u50cf\uff0c\u5982\u9700\u4f7f\u7528lmdeploy\uff0c\u66ff\u6362\u4e0a\u8ff0\u547d\u4ee4\u4e2d\u7684mineru:ppu-vllm-latest\u4e3amineru:ppu-lmdeploy-latest\u5373\u53ef\u3002

"},{"location":"zh/usage/acceleration_cards/THead/#4","title":"4. \u6ce8\u610f\u4e8b\u9879","text":"

\u4e0d\u540c\u73af\u5883\u4e0b\uff0cMinerU\u5bf9ppu\u52a0\u901f\u5361\u7684\u652f\u6301\u60c5\u51b5\u5982\u4e0b\u8868\u6240\u793a\uff1a

Tip

PPU\u52a0\u901f\u5361\u6307\u5b9a\u53ef\u7528\u52a0\u901f\u5361\u7684\u65b9\u5f0f\u4e0eNVIDIA GPU\u7c7b\u4f3c\uff0c\u8bf7\u53c2\u8003\u4f7f\u7528\u6307\u5b9aGPU\u8bbe\u5907\u7ae0\u8282\u8bf4\u660e\u3002
\u5728T-Head\u5e73\u53f0\u53ef\u4ee5\u901a\u8fc7ppu-smi\u547d\u4ee4\u67e5\u770b\u52a0\u901f\u5361\u7684\u4f7f\u7528\u60c5\u51b5\uff0c\u5e76\u6839\u636e\u9700\u8981\u6307\u5b9a\u7a7a\u95f2\u7684\u52a0\u901f\u5361ID\u4ee5\u907f\u514d\u8d44\u6e90\u51b2\u7a81\u3002

"},{"location":"zh/usage/acceleration_cards/Tecorigin/","title":"Tecorigin","text":""},{"location":"zh/usage/acceleration_cards/Tecorigin/#1","title":"1. \u6d4b\u8bd5\u5e73\u53f0","text":"

\u4ee5\u4e0b\u4e3a\u672c\u6307\u5357\u6d4b\u8bd5\u4f7f\u7528\u7684\u5e73\u53f0\u4fe1\u606f\uff0c\u4f9b\u53c2\u8003\uff1a

os: Ubuntu 22.04.5 LTS  \ncpu: AMD EPYC (amd64)\ngpu: T100\ndriver: 3.0.0\ndocker: 28.0.4\n

"},{"location":"zh/usage/acceleration_cards/Tecorigin/#2","title":"2. \u73af\u5883\u51c6\u5907","text":""},{"location":"zh/usage/acceleration_cards/Tecorigin/#21-vllm","title":"2.1 \u4e0b\u8f7d\u5e76\u52a0\u8f7d\u955c\u50cf \uff08vllm\uff09","text":"

wget http://wb.tecorigin.com:8082/repository/teco-customer-repo/Course/MinerU/mineru-vllm.tar\n\ndocker load -i mineru-vllm.tar\n

"},{"location":"zh/usage/acceleration_cards/Tecorigin/#3-docker","title":"3. \u542f\u52a8 Docker \u5bb9\u5668","text":"

docker run -dit --name mineru_docker \\\n    --privileged \\\n    --cap-add SYS_PTRACE \\\n    --cap-add SYS_ADMIN \\\n    --network=host \\\n    --shm-size=500G \\\n    mineru:sdaa-vllm-latest \\\n    /bin/bash\n

Tip

\u5982\u9700\u4f7f\u7528vllm\u73af\u5883,\u8bf7\u6267\u884c\u4ee5\u4e0b\u64cd\u4f5c\uff1a - \u8fdb\u5165\u5bb9\u5668\u540e\uff0c\u901a\u8fc7\u4ee5\u4e0b\u547d\u4ee4\u5207\u6362\u5230conda\u73af\u5883\uff1a

conda activate vllm_env_py310\n

\u5207\u6362\u6210\u529f\u540e\uff0c\u60a8\u53ef\u4ee5\u5728\u547d\u4ee4\u884c\u524d\u770b\u5230(vllm_env_py310)\u7684\u6807\u8bc6\uff0c\u8fd9\u8868\u793a\u60a8\u5df2\u6210\u529f\u8fdb\u5165vllm\u7684\u865a\u62df\u73af\u5883\u3002

"},{"location":"zh/usage/acceleration_cards/Tecorigin/#4","title":"4. \u6ce8\u610f\u4e8b\u9879","text":"

\u4e0d\u540c\u73af\u5883\u4e0b\uff0cMinerU\u5bf9Tecorigin\u52a0\u901f\u5361\u7684\u652f\u6301\u60c5\u51b5\u5982\u4e0b\u8868\u6240\u793a\uff1a

Tip

Tecorigin\u52a0\u901f\u5361\u6307\u5b9a\u53ef\u7528\u52a0\u901f\u5361\u7684\u65b9\u5f0f\u4e0eNVIDIA GPU\u7c7b\u4f3c\uff0c\u8bf7\u53c2\u8003\u4f7f\u7528\u6307\u5b9aGPU\u8bbe\u5907\u7ae0\u8282\u8bf4\u660e, \u5c06\u73af\u5883\u53d8\u91cfCUDA_VISIBLE_DEVICES\u66ff\u6362\u4e3aSDAA_VISIBLE_DEVICES\u5373\u53ef\u3002
\u5728\u592a\u521d\u5e73\u53f0\u53ef\u4ee5\u901a\u8fc7teco-smi -c\u547d\u4ee4\u67e5\u770b\u52a0\u901f\u5361\u7684\u4f7f\u7528\u60c5\u51b5\uff0c\u5e76\u6839\u636e\u9700\u8981\u6307\u5b9a\u7a7a\u95f2\u7684\u52a0\u901f\u5361ID\u4ee5\u907f\u514d\u8d44\u6e90\u51b2\u7a81\u3002

"},{"location":"zh/usage/acceleration_cards/VastAI/","title":"VastAI","text":""},{"location":"zh/usage/acceleration_cards/VastAI/#1","title":"1. \u701a\u535a\u534a\u5bfc\u4f53","text":"

\u5b98\u65b9\u7f51\u5740\uff1ahttps://www.vastaitech.com
\u6a21\u578b\u4e2d\u5fc3\uff1ahttps://github.com/Vastai/VastModelZOO

"},{"location":"zh/usage/acceleration_cards/VastAI/#2","title":"2. \u6d4b\u8bd5\u5e73\u53f0","text":"

\u4ee5\u4e0b\u4e3a\u672c\u6307\u5357\u6d4b\u8bd5\u4f7f\u7528\u7684\u5e73\u53f0\u4fe1\u606f\uff0c\u4f9b\u53c2\u8003

os: Ubuntu-22.04.3-LTS-x86_64\ncpu: Hygon C86-4G\ngpu: VA16 / VA1L / VA10L\ntorch: 2.8.0+cpu\ntorch-vacc: 1.3.3.777\nvllm: 0.11.1.dev0+gb8b302cde.d20251030.cpu\nvllm-vacc: 0.11.0.777\ndriver: 00.25.12.30 d3_3_v2_9_a3_1 a76bf37 20251230\ndocker: 28.1.1\n

"},{"location":"zh/usage/acceleration_cards/VastAI/#3","title":"3. \u73af\u5883\u51c6\u5907","text":"

\u83b7\u53d6vllm_vacc\u57fa\u7840\u955c\u50cf

sudo docker pull harbor.vastaitech.com/ai_deliver/vllm_vacc:VVI-25.12.SP2\n

\u542f\u52a8\u5bb9\u5668

sudo docker run -it \\\n    --privileged=true \\\n    --shm-size=256g \\\n    --name vllm_service \\\n    --ipc=host \\\n    --network=host \\\n    harbor.vastaitech.com/ai_deliver/vllm_vacc:VVI-25.12.SP2 bash\n

\u5b89\u88c5MinerU

\u53c2\u8003\u5b98\u65b9\u6587\u6863\u5b89\u88c5\uff1aREADME_zh-CN.md#\u5b89\u88c5-mineru

```bash\n# \u542f\u52a8\u5bb9\u5668\n# sudo docker exec -it vllm_service bash\n\n# \u53ef\u9009pypi\u6e90\n# https://mirrors.163.com/pypi/simple/\n# https://mirrors.aliyun.com/pypi/simple/\n# https://pypi.mirrors.ustc.edu.cn/simple/\n# https://pypi.tuna.tsinghua.edu.cn/simple/\n# https://mirror.baidu.com/pypi/simple\n\n# \u901a\u8fc7\u6e90\u7801\u5b89\u88c5MinerU\ngit clone https://github.com/opendatalab/MinerU.git\ngit checkout 8c4b3ef3a20b11ddac9903f25124d24ea82639b5\npip install -e .[core] -i https://mirrors.aliyun.com/pypi/simple\n\n# \u6216\u4f7f\u7528pip\u5b89\u88c5MinerU\npip install -U \"mineru[core]==2.7.0\" -i https://mirrors.aliyun.com/pypi/simple\n```\n

Note

vllm_vacc\u57fa\u7840\u955c\u50cf\u5185\u5df2\u5305\u542btorch/vllm\u7b49\u76f8\u5173\u4f9d\u8d56
\u622a\u81f32025/12/31\uff0cVastAI\u5df2\u652f\u6301MinerU\u81f3\u6700\u65b0\u7248\u672c2.7.0\uff0cmaster\u5206\u652f8c4b3ef3
\u548cNVIDIA\u786c\u4ef6\u4e0bCUDA_VISIBLE_DEVICES\u7c7b\u4f3c\uff1b\u5728VastAI\u786c\u4ef6\u4e2d\u53ef\u4ee5\u4f7f\u7528VACC_VISIBLE_DEVICES\u6307\u5b9a\u53ef\u89c1\u8ba1\u7b97\u5361ID\uff0c\u5982-e VACC_VISIBLE_DEVICES=0,1,2,3
\u9700\u6307\u5b9a\u9002\u5f53\u7684--shm-size\u865a\u62df\u5185\u5b58

"},{"location":"zh/usage/acceleration_cards/VastAI/#4-mineru","title":"4. MinerU\u529f\u80fd","text":"

Note

VastAI\u52a0\u901f\u5361\u4ec5\u652f\u6301\u4f7f\u7528vlm-auto-engine\u548cvlm-http-client\u5f62\u5f0f\u8fdb\u884cVLM\u6a21\u578b\u63a8\u7406\u52a0\u901f

\u8fdb\u5165\u5bb9\u5668

sudo docker exec -it vllm_service bash\n

\u4f7f\u7528MinerU

\u6a21\u578b\u51c6\u5907\uff0c\u53c2\u8003\u5b98\u65b9\u4ecb\u7ecd\uff1amodel_source.md

\u65b9\u5f0f\u4e00\uff1avlm-auto-engine

export MINERU_MODEL_SOURCE=modelscope\n\n# step1, \u4ee5`vlm-auto-engine`\u65b9\u5f0f\u542f\u52a8MinerU\u89e3\u6790\u4efb\u52a1\nmineru -p image.png \\\n-o ./output \\\n-b vlm-auto-engine \\\n--http-timeout 1200 \\\n--tensor-parallel-size 2 \\\n--enforce_eager \\\n--trust-remote-code \\\n--max-model-len 16384\n

\u65b9\u5f0f\u4e8c\uff1avlm-http-client

# step1, \u542f\u52a8vLLM API server\nvllm serve /root/.cache/modelscope/hub/models/OpenDataLab/MinerU2.5-2509-1.2B \\\n--tensor-parallel-size 2 \\\n--trust-remote-code \\\n--enforce_eager \\\n--port 8090 \\\n--max-model-len 16384 \\\n--served-model-name MinerU2.5-2509-1.2B\n\n# step2\uff0c\u4ee5`vlm-http-client`\u65b9\u5f0f\u542f\u52a8MinerU\u89e3\u6790\u4efb\u52a1\nmineru -p demo/pdfs/demo1.pdf \\\n-o ./output \\\n-b vlm-http-client \\\n-u http://127.0.0.1:8090 \\\n--http-timeout 1200\n

Note

\u6ce8\u610f\u5728\u6267\u884c\u4efb\u610f\u4e0evllm\u76f8\u5173\u547d\u4ee4\u9700\u8ffd\u52a0--enforce_eager\u53c2\u6570

"},{"location":"zh/usage/acceleration_cards/VastAI/#5","title":"5. \u6ce8\u610f\u4e8b\u9879","text":"

VastAI\u52a0\u901f\u5361\u5bf9MinerU\u7684\u652f\u6301\u60c5\u51b5\u5982\u4e0b\u8868\u6240\u793a\uff1a

\u4f7f\u7528\u573a\u666f \u652f\u6301\u60c5\u51b5 \u547d\u4ee4\u884c\u5de5\u5177(mineru) pipeline \ud83d\udd34 hybrid-http-client \ud83d\udd34 hybrid-auto-engine \ud83d\udd34 vlm-auto-engine \ud83d\udfe2 vlm-http-client \ud83d\udfe2 fastapi\u670d\u52a1(mineru-api) pipeline \ud83d\udd34 hybrid-http-client \ud83d\udd34 hybrid-auto-engine \ud83d\udd34 vlm-auto-engine \ud83d\udfe2 vlm-http-client \ud83d\udfe2 gradio\u754c\u9762(mineru-gradio) pipeline \ud83d\udd34 hybrid-http-client \ud83d\udd34 hybrid-auto-engine \ud83d\udd34 vlm-auto-engine \ud83d\udfe2 vlm-http-client \ud83d\udfe2 openai-server\u670d\u52a1\uff08mineru-openai-server\uff09 \ud83d\udfe2 Tensor\u5e76\u884c (--tensor-parallel-size) \ud83d\udfe2 \u6570\u636e\u5e76\u884c (--data-parallel-size) \ud83d\udd34

Note

\ud83d\udfe2: \u652f\u6301\uff0c\u8fd0\u884c\u8f83\u7a33\u5b9a\uff0c\u7cbe\u5ea6\u4e0eNVIDIA GPU\u57fa\u672c\u4e00\u81f4
\ud83d\udfe1: \u652f\u6301\u4f46\u8f83\u4e0d\u7a33\u5b9a\uff0c\u5728\u67d0\u4e9b\u573a\u666f\u4e0b\u53ef\u80fd\u51fa\u73b0\u5f02\u5e38\uff0c\u6216\u7cbe\u5ea6\u5b58\u5728\u4e00\u5b9a\u5dee\u5f02
\ud83d\udd34: \u4e0d\u652f\u6301\uff0c\u65e0\u6cd5\u8fd0\u884c\uff0c\u6216\u7cbe\u5ea6\u5b58\u5728\u8f83\u5927\u5dee\u5f02
vlm-auto-engine\uff1aVastAI\u4ec5\u652f\u6301vLLM\u540e\u7aef

"},{"location":"zh/usage/plugin/BISHENG/","title":"BISHENG \u7b80\u4ecb","text":""},{"location":"zh/usage/plugin/BISHENG/#bisheng","title":"BISHENG \u7b80\u4ecb","text":"

BISHENG\u6bd5\u6607 \u662f\u4e00\u6b3e\u5f00\u6e90 LLM\u5e94\u7528\u5f00\u53d1\u5e73\u53f0\uff0c\u4e3b\u653b\u4f01\u4e1a\u573a\u666f\uff0c \u5df2\u6709\u5927\u91cf\u884c\u4e1a\u5934\u90e8\u7ec4\u7ec7\u53ca\u4e16\u754c500\u5f3a\u4f01\u4e1a\u5728\u4f7f\u7528\u3002\u201c\u6bd5\u6607\u201d\u662f\u6d3b\u5b57\u5370\u5237\u672f\u7684\u53d1\u660e\u4eba\uff0c\u6d3b\u5b57\u5370\u5237\u672f\u4e3a\u4eba\u7c7b\u77e5\u8bc6\u7684\u4f20\u9012\u8d77\u5230\u4e86\u5de8\u5927\u7684\u63a8\u52a8\u4f5c\u7528\u3002BISHENG\u6bd5\u6607\u56e2\u961f\u5e0c\u671b\u201cBISHENG\u6bd5\u6607\u201d\u540c\u6837\u80fd\u591f\u4e3a\u667a\u80fd\u5e94\u7528\u7684\u5e7f\u6cdb\u843d\u5730\u63d0\u4f9b\u6709\u529b\u652f\u6491\u3002

\u5b98\u7f51\u5730\u5740\uff1ahttps://bisheng.dataelem.com/
Miner \u5728BISHENG\u6bd5\u6607 \u9879\u76ee\u4e2d\u7684\u63d2\u4ef6\u9879\u76ee\uff1ahttps://github.com/dataelement/bisheng/pulls

\u7279\u522b\u9e23\u8c22 @pzc163

"},{"location":"zh/usage/plugin/Cherry_Studio/","title":"Cherry Studio \u7b80\u4ecb","text":""},{"location":"zh/usage/plugin/Cherry_Studio/#cherry-studio","title":"Cherry Studio \u7b80\u4ecb","text":"

Cherry Studio \u662f\u4e00\u6b3e\u529f\u80fd\u5f3a\u5927\u7684\u591a\u6a21\u578b AI \u5ba2\u6237\u7aef\u8f6f\u4ef6\uff0c\u652f\u6301 Windows\u3001macOS \u548c Linux \u7b49\u591a\u5e73\u53f0\u8fd0\u884c\uff0c\u96c6\u6210\u4e86 OpenAI\u3001DeepSeek\u3001Gemini\u3001Anthropic \u7b49\u4e3b\u6d41 AI \u4e91\u670d\u52a1\uff0c\u540c\u65f6\u652f\u6301\u672c\u5730\u6a21\u578b\u8fd0\u884c\uff0c\u7528\u6237\u53ef\u4ee5\u7075\u6d3b\u5207\u6362\u4e0d\u540c\u7684AI\u6a21\u578b\u3002

\u76ee\u524d\uff0cMinerU \u5f3a\u5927\u7684\u6587\u6863\u89e3\u6790\u80fd\u529b\u5df2\u6df1\u5ea6\u96c6\u6210\u5230 Cherry Studio \u7684\u77e5\u8bc6\u5e93\u4e0e\u5bf9\u8bdd\u4ea4\u4e92\u4e2d\uff0c\u4e3a\u7528\u6237\u5e26\u6765\u66f4\u4fbf\u6377\u7684\u6587\u6863\u5904\u7406\u4e0e\u4fe1\u606f\u83b7\u53d6\u4f53\u9a8c\u3002

Cherry Studio \u5b98\u7f51\u5730\u5740\uff1ahttps://www.cherry-ai.com/

"},{"location":"zh/usage/plugin/Cherry_Studio/#mineru-cherry-studio","title":"MinerU \u5728 Cherry Studio \u4e2d\u7684\u4f7f\u7528\u65b9\u6cd5","text":""},{"location":"zh/usage/plugin/Cherry_Studio/#cherry-studio_1","title":"\u8fdb\u5165 Cherry Studio \u8bbe\u7f6e","text":"

a. \u6253\u5f00 Cherry Studio \u5e94\u7528\u7a0b\u5e8f

b. \u70b9\u51fb\u5de6\u4e0b\u89d2\u7684\"\u8bbe\u7f6e\"\u6309\u94ae\uff0c\u8fdb\u5165\u8bbe\u7f6e\u9875\u9762

c. \u5728\u5de6\u4fa7\u83dc\u5355\u4e2d\uff0c\u9009\u62e9\"MCP \u670d\u52a1\u5668\"

\u5728\u53f3\u4fa7\u7684 MCP \u670d\u52a1\u5668\u914d\u7f6e\u754c\u9762\u4e2d\uff0c\u60a8\u53ef\u4ee5\u770b\u5230\u5df2\u6709\u7684 MCP \u670d\u52a1\u5668\u5217\u8868\u3002\u70b9\u51fb\u53f3\u4e0a\u89d2\u7684\"\u6dfb\u52a0\u670d\u52a1\u5668\"\u6309\u94ae\u6765\u521b\u5efa\u65b0\u7684 MCP \u670d\u52a1\uff0c\u6216\u8005\u70b9\u51fb\u73b0\u6709\u670d\u52a1\u6765\u7f16\u8f91\u914d\u7f6e\u3002

"},{"location":"zh/usage/plugin/Cherry_Studio/#mineru-mcp","title":"\u6dfb\u52a0 MinerU-MCP \u914d\u7f6e","text":"

\u70b9\u51fb\"\u6dfb\u52a0\u670d\u52a1\u5668\"\u540e\uff0c\u60a8\u5c06\u770b\u5230\u4e00\u4e2a\u914d\u7f6e\u8868\u5355\u3002\u8bf7\u6309\u4ee5\u4e0b\u6b65\u9aa4\u586b\u5199\uff1a

a. \u540d\u79f0\uff1a\u8f93\u5165\"MinerU-MCP\"\u6216\u60a8\u559c\u6b22\u7684\u5176\u4ed6\u540d\u79f0

b. \u63cf\u8ff0\uff1a\u53ef\u9009\uff0c\u5982\"\u6587\u6863\u8f6c\u6362\u4e3aMarkdown\u5de5\u5177\"

c. \u7c7b\u578b\uff1a\u9009\u62e9\"\u6807\u51c6\u8f93\u5165/\u8f93\u51fa\uff08stdio\uff09\"

d. \u547d\u4ee4\uff1a\u8f93\u5165 uvx

e. \u53c2\u6570\uff1a\u8f93\u5165 mineru-mcp

f. \u73af\u5883\u53d8\u91cf\uff1a\u6dfb\u52a0\u4ee5\u4e0b\u73af\u5883\u53d8\u91cf

MINERU_API_BASE=https://mineru.net\nMINERU_API_KEY=\u60a8\u7684API\u5bc6\u94a5\nOUTPUT_DIR=./downloads\nUSE_LOCAL_API=false\nLOCAL_MINERU_API_BASE=http://localhost:8888\n

\u4f7f\u7528 uvx \u547d\u4ee4\u53ef\u4ee5\u81ea\u52a8\u5904\u7406 mineru-mcp \u7684\u5b89\u88c5\u548c\u8fd0\u884c\uff0c\u65e0\u9700\u9884\u5148\u624b\u52a8\u5b89\u88c5 mineru-mcp \u5305\u3002\u8fd9\u662f\u6700\u7b80\u5355\u7684\u914d\u7f6e\u65b9\u5f0f\u3002

"},{"location":"zh/usage/plugin/Cherry_Studio/#_1","title":"\u4fdd\u5b58\u914d\u7f6e","text":"

\u786e\u8ba4\u65e0\u8bef\u540e\uff0c\u70b9\u51fb\u754c\u9762\u53f3\u4e0a\u89d2\u7684\"\u4fdd\u5b58\"\u6309\u94ae\u5b8c\u6210\u914d\u7f6e\u3002\u4fdd\u5b58\u540e\uff0cMCP \u670d\u52a1\u5668\u5217\u8868\u4e2d\u4f1a\u663e\u793a\u60a8\u521a\u521a\u6dfb\u52a0\u7684 MinerU-MCP \u670d\u52a1\u3002

"},{"location":"zh/usage/plugin/Cherry_Studio/#cherry-studio-mineru-mcp","title":"\u4f7f\u7528 Cherry Studio \u4e2d\u7684 MinerU MCP","text":"

\u4e00\u65e6\u914d\u7f6e\u5b8c\u6210\uff0c\u60a8\u53ef\u4ee5\u5728 Cherry Studio \u4e2d\u7684\u5bf9\u8bdd\u4e2d\u4f7f\u7528 MinerU MCP \u5de5\u5177\u3002\u5728 Cherry Studio \u4e2d\uff0c\u60a8\u53ef\u4ee5\u4f7f\u7528\u5982\u4e0b\u63d0\u793a\u8ba9\u6a21\u578b\u8c03\u7528 MinerU MCP \u5de5\u5177\u3002\u6a21\u578b\u4f1a\u81ea\u52a8\u8bc6\u522b\u4efb\u52a1\u5e76\u8c03\u7528\u76f8\u5e94\u7684\u5de5\u5177\u3002

"},{"location":"zh/usage/plugin/Cherry_Studio/#1-url","title":"\u793a\u4f8b 1: \u4f7f\u7528 URL \u8f6c\u6362\u6587\u6863","text":"

\u7528\u6237\u8f93\u5165:

\u8bf7\u4f7f\u7528 MinerU MCP \u5c06\u4ee5\u4e0b URL \u7684 PDF \u6587\u6863\u8f6c\u6362\u4e3a Markdown \u683c\u5f0f\uff1ahttps://example.com/sample.pdf\n

\u6a21\u578b\u5c06\u6267\u884c\u7684\u6b65\u9aa4\uff1a

\u6a21\u578b\u8bc6\u522b\u8fd9\u662f\u6587\u6863\u8f6c\u6362\u4efb\u52a1\uff0c\u5e76\u8c03\u7528 parse_documents \u5de5\u5177\uff0c\u53c2\u6570\u4e3a:

{\"file_sources\": \"https://example.com/sample.pdf\"}\n

\u5de5\u5177\u5904\u7406\u5b8c\u6210\u540e\uff0c\u6a21\u578b\u4f1a\u544a\u77e5\u60a8\u8f6c\u6362\u7ed3\u679c\u3002

"},{"location":"zh/usage/plugin/Cherry_Studio/#2","title":"\u793a\u4f8b 2: \u8f6c\u6362\u672c\u5730\u6587\u6863","text":"

\u7528\u6237\u8f93\u5165:

\u8bf7\u4f7f\u7528 MinerU-MCP \u5c06\u672c\u5730\u7684 D://sample.pdf \u6587\u4ef6\u8f6c\u6362\u4e3a Markdown \u683c\u5f0f\n

\u6a21\u578b\u5c06\u6267\u884c\u7684\u6b65\u9aa4\uff1a

\u6a21\u578b\u8bc6\u522b\u8fd9\u662f\u672c\u5730\u6587\u6863\u8f6c\u6362\u4efb\u52a1\uff0c\u8c03\u7528 parse_documents \u5de5\u5177\uff0c\u53c2\u6570\u4e3a:

{\"file_sources\": \"D://sample.pdf\"}\n

"},{"location":"zh/usage/plugin/Cherry_Studio/#3-ocr","title":"\u793a\u4f8b 3: \u542f\u7528 OCR \u5904\u7406\u626b\u63cf\u6587\u6863","text":"

\u7528\u6237\u8f93\u5165:

\u8bf7\u4f7f\u7528 MinerU-MCP \u5c06\u4ee5\u4e0b URL \u7684\u626b\u63cf\u7248 PDF \u6587\u6863\u8f6c\u6362\u4e3a Markdown \u683c\u5f0f\uff0c\u5e76\u542f\u7528 OCR \u529f\u80fd\uff1a\nhttps://example.com/scanned.pdf\n

\u6a21\u578b\u5c06\u6267\u884c\u7684\u6b65\u9aa4\uff1a

\u6a21\u578b\u8bc6\u522b\u8fd9\u662f\u9700\u8981 OCR \u5904\u7406\u7684\u6587\u6863\u8f6c\u6362\u4efb\u52a1\uff0c\u8c03\u7528 parse_documents \u5de5\u5177\uff0c\u5e76\u542f\u7528 OCR \u53c2\u6570:

{\"file_sources\": \"https://example.com/scanned.pdf\",\"enable_ocr\": true}\n

"},{"location":"zh/usage/plugin/Cherry_Studio/#4","title":"\u793a\u4f8b 4: \u5b8c\u6574\u5bf9\u8bdd\u6d41\u7a0b","text":"

\u4ee5\u4e0b\u662f\u4e00\u4e2a\u5b8c\u6574\u7684\u5bf9\u8bdd\u6d41\u7a0b\u793a\u4f8b\uff1a

\u7528\u6237:

\u6211\u6709\u4e00\u4efd\u5b66\u672f\u8bba\u6587\u7684 PDF\uff0c\u7f51\u5740\u662f https://arxiv.org/pdf/2303.08774.pdf\uff0c\u80fd\u5e2e\u6211\u8f6c\u6362\u6210 Markdown \u683c\u5f0f\u5417\uff1f\n

\u6a21\u578b:

\u6211\u53ef\u4ee5\u5e2e\u60a8\u5c06\u8fd9\u4efd\u5b66\u672f\u8bba\u6587\u8f6c\u6362\u4e3a Markdown \u683c\u5f0f\u3002\u6211\u5c06\u4f7f\u7528 MinerU-MCP \u5de5\u5177\u6765\u5904\u7406\u8fd9\u4e2a\u4efb\u52a1\u3002\n[\u6a21\u578b\u8c03\u7528 parse_documents \u5de5\u5177\uff0c\u5904\u7406\u8fc7\u7a0b\u4e2d...]\n\u8bba\u6587\u5df2\u6210\u529f\u8f6c\u6362\u4e3a Markdown \u683c\u5f0f\u3002\u8f6c\u6362\u7ed3\u679c\u5982\u4e0b\uff1a\n# The Capacity of Diffusion Models to Memorize and Generate Training Data\n## Abstract\nRecent diffusion models can generate high-quality images that are nearly indistinguishable from real ones...\n[\u663e\u793a\u8bba\u6587\u5185\u5bb9...]\n

"},{"location":"zh/usage/plugin/Cherry_Studio/#_2","title":"\u5de5\u5177\u53c2\u6570\u8be6\u89e3","text":"

\u5728\u4f7f\u7528\u8fc7\u7a0b\u4e2d\uff0c\u6a21\u578b\u4f1a\u6839\u636e\u60a8\u7684\u6307\u4ee4\u81ea\u52a8\u9009\u62e9\u5408\u9002\u7684\u5de5\u5177\u548c\u53c2\u6570\u3002\u4ee5\u4e0b\u662f\u4e3b\u8981\u5de5\u5177\u7684\u53c2\u6570\u8bf4\u660e\uff1a

"},{"location":"zh/usage/plugin/Cherry_Studio/#parse_documents","title":"\u25cf parse_documents \u5de5\u5177\u53c2\u6570","text":""},{"location":"zh/usage/plugin/Cherry_Studio/#get_ocr_languages","title":"\u25cf get_ocr_languages \u5de5\u5177\u53c2\u6570","text":"

\u65e0\u9700\u53c2\u6570\uff0c\u7528\u4e8e\u83b7\u53d6OCR\u652f\u6301\u7684\u8bed\u8a00\u5217\u8868\u3002

"},{"location":"zh/usage/plugin/Cherry_Studio/#_3","title":"\u9ad8\u7ea7\u7528\u6cd5","text":""},{"location":"zh/usage/plugin/Cherry_Studio/#_4","title":"\u6307\u5b9a\u8bed\u8a00\u548c\u9875\u7801\u8303\u56f4","text":"

\u7528\u6237\u8f93\u5165:

\u8bf7\u4f7f\u7528 MinerU MCP \u5c06\u4ee5\u4e0b URL \u7684\u6587\u6863\u8f6c\u6362\u4e3a Markdown \u683c\u5f0f\uff0c\u53ea\u5904\u7406\u7b2c 5-10 \u9875\uff0c\u5e76\u6307\u5b9a\u8bed\u8a00\u4e3a\u4e2d\u6587\uff1ahttps://example.com/document.pdf\n

\u6a21\u578b\u4f1a\u4f7f\u7528 parse_documents \u5de5\u5177\uff0c\u5e76\u8bbe\u7f6e language \u53c2\u6570\u4e3a \"ch\"\uff0cpage_ranges \u53c2\u6570\u4e3a \"5-10\"\u3002

"},{"location":"zh/usage/plugin/Cherry_Studio/#_5","title":"\u6279\u91cf\u5904\u7406\u591a\u4e2a\u6587\u6863","text":"

\u7528\u6237\u8f93\u5165:

\u8bf7\u4f7f\u7528 MinerU-MCP \u5c06\u4ee5\u4e0b\u591a\u4e2a URL \u7684\u6587\u6863\u8f6c\u6362\u4e3a Markdown \u683c\u5f0f\uff1a\nhttps://example.com/doc1.pdf\nhttps://example.com/doc2.pdf\nhttps://example.com/doc3.pdf\n

\u6a21\u578b\u4f1a\u8c03\u7528 parse_documents \u5de5\u5177\uff0c\u5e76\u5c06\u591a\u4e2a URL \u4ee5\u9017\u53f7\u5206\u9694\u4f20\u5165 file_sources \u53c2\u6570\u3002

"},{"location":"zh/usage/plugin/Cherry_Studio/#_6","title":"\u6ce8\u610f\u4e8b\u9879","text":"

\u25cf \u5f53\u8bbe\u7f6e USE_LOCAL_API=true \u65f6\uff0c\u4f7f\u7528\u672c\u5730\u914d\u7f6e\u7684API\u8fdb\u884c\u89e3\u6790

\u25cf \u5f53\u8bbe\u7f6e USE_LOCAL_API=false \u65f6\uff0c\u4f1a\u4f7f\u7528 MinerU \u5b98\u7f51\u7684API\u8fdb\u884c\u89e3\u6790

\u25cf \u5904\u7406\u5927\u578b\u6587\u6863\u53ef\u80fd\u9700\u8981\u8f83\u957f\u65f6\u95f4\uff0c\u8bf7\u8010\u5fc3\u7b49\u5f85

\u25cf \u5982\u679c\u9047\u5230\u8d85\u65f6\u95ee\u9898\uff0c\u8bf7\u8003\u8651\u5206\u6279\u5904\u7406\u6587\u6863\u6216\u4f7f\u7528\u672c\u5730API\u6a21\u5f0f

"},{"location":"zh/usage/plugin/Cherry_Studio/#_7","title":"\u5e38\u89c1\u95ee\u9898\u4e0e\u89e3\u51b3\u65b9\u6848","text":""},{"location":"zh/usage/plugin/Cherry_Studio/#mcp","title":"\u65e0\u6cd5\u542f\u52a8 MCP \u670d\u52a1","text":"

\u95ee\u9898\uff1a\u8fd0\u884c uv run -m mineru.cli\u65f6\u62a5\u9519\u3002

\u89e3\u51b3\u65b9\u6848\uff1a

\u25cf \u786e\u4fdd\u5df2\u6fc0\u6d3b\u865a\u62df\u73af\u5883

\u25cf \u68c0\u67e5\u662f\u5426\u5df2\u5b89\u88c5\u6240\u6709\u4f9d\u8d56

\u25cf \u5c1d\u8bd5\u4f7f\u7528 python -m mineru.cli\u547d\u4ee4\u66ff\u4ee3

"},{"location":"zh/usage/plugin/Cherry_Studio/#_8","title":"\u6587\u4ef6\u8f6c\u6362\u5931\u8d25","text":"

\u95ee\u9898\uff1a\u6587\u4ef6\u4e0a\u4f20\u6210\u529f\u4f46\u8f6c\u6362\u5931\u8d25\u3002

\u89e3\u51b3\u65b9\u6848\uff1a

\u25cf \u68c0\u67e5\u6587\u4ef6\u683c\u5f0f\u662f\u5426\u53d7\u652f\u6301

\u25cf \u786e\u8ba4API\u5bc6\u94a5\u662f\u5426\u6b63\u786e

\u25cf \u67e5\u770bMCP\u670d\u52a1\u65e5\u5fd7\u83b7\u53d6\u8be6\u7ec6\u9519\u8bef\u4fe1\u606f

"},{"location":"zh/usage/plugin/Cherry_Studio/#_9","title":"\u6587\u4ef6\u8def\u5f84\u95ee\u9898","text":"

\u95ee\u9898\uff1a\u4f7f\u7528 parse_documents \u5de5\u5177\u5904\u7406\u672c\u5730\u6587\u4ef6\u65f6\u62a5\u627e\u4e0d\u5230\u6587\u4ef6\u9519\u8bef\u3002

\u89e3\u51b3\u65b9\u6848\uff1a\u8bf7\u786e\u4fdd\u4f7f\u7528\u7edd\u5bf9\u8def\u5f84\uff0c\u6216\u8005\u76f8\u5bf9\u4e8e\u670d\u52a1\u5668\u8fd0\u884c\u76ee\u5f55\u7684\u6b63\u786e\u76f8\u5bf9\u8def\u5f84\u3002

"},{"location":"zh/usage/plugin/Cherry_Studio/#mcp_1","title":"MCP \u670d\u52a1\u8c03\u7528\u8d85\u65f6\u95ee\u9898","text":"

\u95ee\u9898\uff1a\u8c03\u7528 parse_documents \u5de5\u5177\u65f6\u51fa\u73b0 Error calling tool 'parse_documents': MCP error -32001: Request timed out \u9519\u8bef\u3002

\u89e3\u51b3\u65b9\u6848\uff1a\u8fd9\u4e2a\u95ee\u9898\u5e38\u89c1\u4e8e\u5904\u7406\u5927\u578b\u6587\u6863\u6216\u7f51\u7edc\u4e0d\u7a33\u5b9a\u7684\u60c5\u51b5\u3002\u5728\u67d0\u4e9b MCP \u5ba2\u6237\u7aef\uff08\u5982 Cursor\uff09\u4e2d\uff0c\u8d85\u65f6\u540e\u53ef\u80fd\u5bfc\u81f4\u65e0\u6cd5\u518d\u6b21\u8c03\u7528 MCP \u670d\u52a1\uff0c\u9700\u8981\u91cd\u542f\u5ba2\u6237\u7aef\u3002\u6700\u65b0\u7248\u672c\u7684 Cursor \u4e2d\u53ef\u80fd\u4f1a\u663e\u793a\u6b63\u5728\u8c03\u7528 MCP\uff0c\u4f46\u5b9e\u9645\u4e0a\u6ca1\u6709\u771f\u6b63\u8c03\u7528\u6210\u529f\u3002\u5efa\u8bae\uff1a

\u25cf \u7b49\u5f85\u5b98\u65b9\u4fee\u590d\uff1a\u8fd9\u662fCursor\u5ba2\u6237\u7aef\u7684\u5df2\u77e5\u95ee\u9898\uff0c\u5efa\u8bae\u7b49\u5f85Cursor\u5b98\u65b9\u4fee\u590d

\u25cf \u5904\u7406\u5c0f\u6587\u4ef6\uff1a\u5c3d\u91cf\u53ea\u5904\u7406\u5c11\u91cf\u5c0f\u6587\u4ef6\uff0c\u907f\u514d\u5904\u7406\u5927\u578b\u6587\u6863\u5bfc\u81f4\u8d85\u65f6

\u25cf \u5206\u6279\u5904\u7406\uff1a\u5c06\u591a\u4e2a\u6587\u4ef6\u5206\u6210\u591a\u6b21\u8bf7\u6c42\u5904\u7406\uff0c\u6bcf\u6b21\u53ea\u5904\u7406\u4e00\u4e24\u4e2a\u6587\u4ef6

\u25cf \u589e\u52a0\u8d85\u65f6\u65f6\u95f4\u8bbe\u7f6e\uff08\u5982\u679c\u5ba2\u6237\u7aef\u652f\u6301\uff09

\u25cf \u5bf9\u4e8e\u8d85\u65f6\u540e\u65e0\u6cd5\u518d\u6b21\u8c03\u7528\u7684\u95ee\u9898\uff0c\u9700\u8981\u91cd\u542f MCP \u5ba2\u6237\u7aef

\u25cf \u5982\u679c\u53cd\u590d\u51fa\u73b0\u8d85\u65f6\uff0c\u8bf7\u68c0\u67e5\u7f51\u7edc\u8fde\u63a5\u6216\u8003\u8651\u4f7f\u7528\u672c\u5730 API \u6a21\u5f0f

"},{"location":"zh/usage/plugin/Coze/","title":"Coze \u7b80\u4ecb","text":""},{"location":"zh/usage/plugin/Coze/#coze","title":"Coze \u7b80\u4ecb","text":"

Coze\uff08\u4e2d\u6587\u7248\u540d\u79f0\uff1a\u6263\u5b50\uff09 \u662f\u5b57\u8282\u8df3\u52a8\u63a8\u51fa\u7684\u96f6\u4ee3\u7801 AI \u5e94\u7528\u5f00\u53d1\u5e73\u53f0\u3002\u65e0\u8bba\u7528\u6237\u662f\u5426\u6709\u7f16\u7a0b\u7ecf\u9a8c\uff0c\u90fd\u53ef\u4ee5\u901a\u8fc7\u8be5\u5e73\u53f0\u5feb\u901f\u521b\u5efa\u5404\u79cd\u7c7b\u578b\u7684\u804a\u5929\u673a\u5668\u4eba\u3001\u667a\u80fd\u4f53\u3001AI \u5e94\u7528\u548c\u63d2\u4ef6\uff0c\u5e76\u5c06\u5176\u90e8\u7f72\u5728\u793e\u4ea4\u5e73\u53f0\u548c\u5373\u65f6\u804a\u5929\u5e94\u7528\u7a0b\u5e8f\u4e2d\u3002

\u76ee\u524d\uff0cMinerU \u63d2\u4ef6\u5df2\u5728 Coze \u63d2\u4ef6\u5546\u5e97\u4e0a\u7ebf\uff0c\u901a\u8fc7\u5176\u5f3a\u5927\u7684\u6587\u6863\u89e3\u6790\u80fd\u529b\uff0c\u4e3a\u7528\u6237\u642d\u5efa\u667a\u80fd\u4f53\u4e0e\u5de5\u4f5c\u6d41\u63d0\u4f9b\u6587\u6863\u89e3\u6790\u80fd\u529b\uff0c\u52a0\u5feb\u7528\u6237 AI \u5e94\u7528\u7684\u5f00\u53d1\u3002

\u6263\u5b50\u5b98\u7f51\u5730\u5740\uff1ahttps://www.coze.cn/
MinerU \u6263\u5b50\u63d2\u4ef6\u4e0b\u8f7d\u5730\u5740\uff1ahttps://www.coze.cn/store/plugin/7527957359730360354

"},{"location":"zh/usage/plugin/Coze/#mineru-coze","title":"MinerU \u5728 Coze \u4e2d\u7684\u4f7f\u7528\u65b9\u6cd5","text":""},{"location":"zh/usage/plugin/Coze/#coze_1","title":"Coze\uff1a\u96c6\u6210\u5e94\u7528","text":"

\u8fdb\u5165 https://www.coze.cn/home coze \u5f00\u53d1\u5e73\u53f0

"},{"location":"zh/usage/plugin/Coze/#_1","title":"\u667a\u80fd\u4f53","text":""},{"location":"zh/usage/plugin/Coze/#-","title":"\u5de5\u4f5c\u7a7a\u95f4 -> \u9879\u76ee\u5f00\u53d1 -> \u521b\u5efa -> \u521b\u5efa\u667a\u80fd\u4f53 -> \u521b\u5efa -> \u8f93\u5165\u9879\u76ee\u540d","text":""},{"location":"zh/usage/plugin/Coze/#-mineru","title":"\u63d2\u4ef6\u914d\u7f6e -> \u6dfb\u52a0 \u63d2\u4ef6 -> \u641c\u7d22 MinerU","text":""},{"location":"zh/usage/plugin/Coze/#parse_file","title":"\u6dfb\u52a0 parse_file \u5de5\u5177\uff08\u5728\u7ebf\u7248\uff09","text":""},{"location":"zh/usage/plugin/Coze/#mineru-api-key","title":"\u9009\u62e9 MinerU \u63d2\u4ef6 -> \u7f16\u8f91\u53c2\u6570 -> \u586b\u5199 api key","text":"

\u8bb0\u5f97\u5173\u95ed url \u548c token \u663e\u793a

"},{"location":"zh/usage/plugin/Coze/#_2","title":"\u8c03\u8bd5 \u667a\u80fd\u4f53","text":""},{"location":"zh/usage/plugin/Coze/#_3","title":"\u5de5\u4f5c\u6d41","text":"

\u7528\u5de5\u4f5c\u6d41\u7684\u65b9\u5f0f\u4f7f\u7528 minerU

"},{"location":"zh/usage/plugin/Coze/#-_1","title":"\u5de5\u4f5c\u6d41 -> \u521b\u5efa\u5de5\u4f5c\u6d41","text":""},{"location":"zh/usage/plugin/Coze/#-mineru-","title":"\u5de5\u4f5c\u6d41\u63d2\u4ef6\u914d\u7f6e -> \u6dfb\u52a0 \u63d2\u4ef6 -> \u641c\u7d22 MinerU -> \u6dfb\u52a0","text":""},{"location":"zh/usage/plugin/Coze/#mineru-api-key_1","title":"\u9009\u62e9MinerU \u63d2\u4ef6 -> \u7f16\u8f91\u53c2\u6570 -> \u586b\u5199 api key","text":""},{"location":"zh/usage/plugin/Coze/#-input-mineru","title":"\u9009\u62e9\u5f00\u59cb\u8282\u70b9 -> \u914d\u7f6e input \u7c7b\u578b\u4e3a\u6587\u4ef6\u7c7b\u578b -> \u8fde\u63a5\u5230 mineru \u8282\u70b9","text":""},{"location":"zh/usage/plugin/Coze/#-mineru-output-mineru-parse_filetext","title":"\u9009\u62e9\u7ed3\u675f\u8282\u70b9 -> \u8fde\u63a5\u5230 mineru \u8282\u70b9 -> \u914d\u7f6e output \u8f93\u51fa\u4e3a mineru \u8282\u70b9\u7684 parse_file.text","text":""},{"location":"zh/usage/plugin/Coze/#-_2","title":"\u4e0a\u4f20\u6587\u4ef6 -> \u8bd5\u8fd0\u884c","text":""},{"location":"zh/usage/plugin/Coze/#-_3","title":"\u53d1\u5e03 -> \u6dfb\u52a0\u5230\u5f53\u524d\u667a\u80fd\u4f53","text":""},{"location":"zh/usage/plugin/Coze/#mineru-","title":"\u79fb\u9664 mineru \u63d2\u4ef6 -> \u8c03\u8bd5","text":""},{"location":"zh/usage/plugin/DataFlow/","title":"\u5143\u67a2\u667a\u6c47 ADP \u667a\u80fd\u6570\u636e\u5e73\u53f0 \u7b80\u4ecb","text":""},{"location":"zh/usage/plugin/DataFlow/#adp","title":"\u5143\u67a2\u667a\u6c47 ADP \u667a\u80fd\u6570\u636e\u5e73\u53f0 \u7b80\u4ecb","text":"

\u5143\u67a2\u667a\u6c47 ADP \u667a\u80fd\u6570\u636e\u5e73\u53f0\u57fa\u4e8e\u81ea\u7814 AI \u6570\u636e\u5e93\u548c DataFlow\u6570\u636e\u51c6\u5907\u6846\u67b6\u6253\u9020\uff0c\u65e8\u5728\u5e2e\u52a9\u4f01\u4e1a\u9ad8\u6548\u7ba1\u7406\u3001\u68c0\u7d22\u3001\u5904\u7406\u6d77\u91cf\u6570\u636e\uff0c\u5e76\u901a\u8fc7\u4f53\u7cfb\u5316\u3001\u81ea\u52a8\u5316\u6570\u636e\u6cbb\u7406\u964d\u4f4e\u6a21\u578b/\u667a\u80fd\u4f53\u8bad\u7ec3\u7684\u4e13\u4e1a\u95e8\u69db\uff0c\u5e2e\u52a9\u4f01\u4e1a\u7ed3\u5408\u4e1a\u52a1\u573a\u666f\u53d1\u6325\u79c1\u6709\u6570\u636e\u7684\u4ef7\u503c\uff0c\u771f\u6b63\u843d\u5730AI\u5e94\u7528\u3002

\u76ee\u524d\uff0cMinerU \u5df2\u6df1\u5ea6\u96c6\u6210\u4e8e\u5143\u67a2\u667a\u6c47 ADP \u667a\u80fd\u6570\u636e\u5e73\u53f0\u7684 DataFlow \u6a21\u5757\u4e2d\uff0c\u5176\u6570\u636e\u89e3\u6790\u670d\u52a1\u7531\u6587\u6863\u8bed\u6599\u63d0\u53d6\u5f15\u64ce MinerU \u63d0\u4f9b\u652f\u6301\u3002

\u5b98\u7f51\u5730\u5740\uff1ahttps://adp.originhub.tech/agent
Miner fastGPT \u63d2\u4ef6\u4e0b\u8f7d\u5730\u5740\uff1ahttps://cloud.fastgpt.io/dashboard/systemPlugin?type=productivity

"},{"location":"zh/usage/plugin/Dify/","title":"Dify \u7b80\u4ecb","text":""},{"location":"zh/usage/plugin/Dify/#dify","title":"Dify \u7b80\u4ecb","text":"

Dify \u662f\u4e00\u4e2a\u5f00\u6e90\u7684\u5927\u8bed\u8a00\u6a21\u578b\uff08LLM\uff09\u5e94\u7528\u5f00\u53d1\u5e73\u53f0\uff0c\u65e8\u5728\u7b80\u5316\u548c\u52a0\u901f\u751f\u6210\u5f0f AI \u5e94\u7528\u7684\u521b\u5efa\u548c\u90e8\u7f72\u3002\u5b83\u7ed3\u5408\u4e86\u540e\u7aef\u5373\u670d\u52a1\uff08BaaS\uff09\u548c LLMOps \u7684\u7406\u5ff5\uff0c\u4e3a\u5f00\u53d1\u8005\u63d0\u4f9b\u4e86\u7528\u6237\u53cb\u597d\u7684\u754c\u9762\u548c\u5f3a\u5927\u7684\u5de5\u5177\uff0c\u6709\u6548\u964d\u4f4e\u4e86 AI \u5e94\u7528\u5f00\u53d1\u7684\u95e8\u69db\u3002

\u76ee\u524d MinerU \u4e0e Dify \u8054\u5408\u7814\u53d1\u7684 MinerU \u63d2\u4ef6\u5df2\u5728 Dify \u5e02\u573a\u4e0a\u67b6\uff0c\u5e2e\u52a9\u7528\u6237\u642d\u5efa\u5de5\u4f5c\u6d41\uff0c\u63d0\u4f9b\u6587\u6863\u89e3\u6790\u7684\u5de5\u4f5c\u3002

Dify \u5b98\u7f51\u5730\u5740\uff1ahttps://dify.ai/zh
MinerU Dify \u63d2\u4ef6\u4e0b\u8f7d\u5730\u5740\uff1ahttps://marketplace.dify.ai/plugins/langgenius/mineru

"},{"location":"zh/usage/plugin/Dify/#mineru-dify","title":"MinerU \u5728 Dify \u4e2d\u7684\u4f7f\u7528\u65b9\u6cd5","text":""},{"location":"zh/usage/plugin/Dify/#mineru-dify-v040","title":"\u4e00\u3001\u65b0\u7248MinerU Dify\u63d2\u4ef6\u4eae\u70b9 (v0.4.0)","text":"

\u5b8c\u7f8e\u9002\u914dMinerU2\uff1a\u5168\u9762\u517c\u5bb9MinerU2\u7684\u6700\u65b0\u529f\u80fd\uff0c\u91ca\u653e\u9876\u5c16\u7684\u6587\u6863\u89e3\u6790\u80fd\u529b\u3002
\u8d85\u9ad8\u7075\u6d3b\u6027\uff1a\u540c\u65f6\u652f\u6301\u5b98\u65b9\u5728\u7ebfAPI\u548c\u672c\u5730\u5316\u90e8\u7f72\u7684API\uff08\u5e76\u5411\u4e0b\u517c\u5bb9 1.x \u7248\u672c\uff09\u3002
\u8d4b\u80fd\u5de5\u4f5c\u6d41\uff1a\u8ba9Dify\u7684Agent\u62e5\u6709\u5f3a\u5927\u7684\u6587\u6863\u201c\u8bfb\u5199\u201d\u80fd\u529b\uff0c\u8f7b\u677e\u5904\u7406\u590d\u6742\u4efb\u52a1\u3002

"},{"location":"zh/usage/plugin/Dify/#_1","title":"\u4e8c\u3001\u5b9e\u6218\u6f14\u7ec3\uff1a\u4e24\u4e2a\u6848\u4f8b\u5e26\u4f60\u5feb\u901f\u4e0a\u624b","text":"

\u7a7a\u8c08\u4e0d\u5982\u5b9e\u6218\u3002\u4e0b\u9762\u6211\u4eec\u901a\u8fc7\u4e24\u4e2a\u5178\u578b\u573a\u666f\uff0c\u5411\u4f60\u5c55\u793a\u65b0\u7248\u63d2\u4ef6\u7684\u5f3a\u5927\u4e4b\u5904\u3002

"},{"location":"zh/usage/plugin/Dify/#_2","title":"\u51c6\u5907","text":"

\u5728Dify\u63d2\u4ef6\u9875\u9762\u5b89\u88c5MinerU\u63d2\u4ef6\uff08\u79c1\u6709\u5316\u90e8\u7f72\u7684Dify\u540c\u7406\uff09
\u586b\u5199API URL\u7b49\u4fe1\u606f

\u4f7f\u7528\u5b98\u65b9API\u65f6\u4ee4\u724c\uff08Token\uff09\u5fc5\u987b\u63d0\u4f9b\ud83d\udc46\uff0c\u4f7f\u7528\u672c\u5730\u90e8\u7f72API\u65f6\u4ee4\u724c\u53ef\u4e0d\u586b\u5199\ud83d\udc47

"},{"location":"zh/usage/plugin/Dify/#chat-pdf","title":"\u6848\u4f8b\u4e00\uff1a\u89e3\u6790\u5355\u6587\u4ef6\uff0c\u642d\u5efaChat PDF\u5e94\u7528","text":"

\u60f3\u501f\u52a9AI\u4e0e\u4f60\u7684\u6587\u6863\u5bf9\u8bdd\u5417\uff1f\u8ddf\u7740\u4e0b\u9762\u51e0\u6b65\uff0c\u8f7b\u677e\u5b9e\u73b0

"},{"location":"zh/usage/plugin/Dify/#chatflow","title":"\u7b2c\u4e00\u6b65\uff1a\u521b\u5efa\u7a7a\u767d\u5e94\u7528\uff0c\u9009\u62e9\u201cChatflow\u201d","text":"

\u8f93\u5165\u5e94\u7528\u540d\u79f0\u4e0e\u63cf\u8ff0

"},{"location":"zh/usage/plugin/Dify/#_3","title":"\u7b2c\u4e8c\u6b65\uff1a\u521b\u5efa\u7684\u521d\u59cb\u6a21\u677f\u4e2d\uff0c\u9009\u62e9\u201c\u5f00\u59cb\u201d\u8282\u70b9","text":"

\u5b57\u6bb5\u7c7b\u578b\u9009\u4e3a\u5355\u6587\u4ef6\uff0c\u586b\u5199\u53d8\u91cf\u540d\u79f0\uff08\u6b64\u5904\u586b\u4e3ainput_file\uff09,\u652f\u6301\u6587\u6863\u7c7b\u578b\u9009\u4e3a\u6587\u6863\u4e0e\u56fe\u7247

"},{"location":"zh/usage/plugin/Dify/#mineru","title":"\u7b2c\u4e09\u6b65\uff1a\u6dfb\u52a0\u5de5\u5177\u8282\u70b9\u2014\u2014MinerU\u63d2\u4ef6\u6765\u89e3\u6790\u4e0a\u4e00\u6b65\u5f00\u59cb\u8282\u70b9\u4e0a\u4f20\u7684\u6587\u4ef6","text":""},{"location":"zh/usage/plugin/Dify/#mineru-input_file","title":"\u7b2c\u56db\u6b65\uff1a\u8bbe\u7f6eMinerU\u7684\u8f93\u5165\u53d8\u91cf\uff0c\u9009\u62e9\u4e0a\u4e00\u6b65\u5f00\u59cb\u8282\u70b9\u6dfb\u52a0\u7684 input_file","text":""},{"location":"zh/usage/plugin/Dify/#llm","title":"\u7b2c\u4e94\u6b65\uff1a\u914d\u7f6eLLM\u6a21\u578b","text":"

\u9009\u62e9\u201cLLM\u201d\u8282\u70b9\u540e\uff0c\u5982\u679c\u6ca1\u6709\u6a21\u578b\u53ef\u7528\uff0c\u9700\u8981\u5355\u72ec\u5728\u63d2\u4ef6\u5e02\u573a\u5b89\u88c5\uff08\u8fd9\u91cc\u4f7f\u7528 Deepseek\u4f5c\u4e3a\u793a\u4f8b\uff09

\u201c\u4e0a\u4e0b\u6587\u201d\u9009\u62e9MinerU\u7684\u8f93\u51fa\u53d8\u91cf text\uff08MinerU\u89e3\u6790\u6587\u6863\u540e\u7684markdown\u683c\u5f0f\uff09

\u5728\u201cSYSTEM\u201d\u533a\u57df\u6839\u636e\u5b9e\u9645\u9700\u6c42\u586b\u5199\u63d0\u793a\u8bcd\uff0c\u53ef\u5982\u56fe\u586b\u5199\u201c\u5728Parse File text\u4e2d\u63d0\u53d6\u7528\u6237\u7684\u95ee\u9898\u7b54\u6848\u201d

"},{"location":"zh/usage/plugin/Dify/#_4","title":"\u7b2c\u516d\u6b65\uff1a\u9884\u89c8\uff0c\u4e0a\u4f20\u6587\u4ef6\u5e76\u63d0\u95ee\u673a\u5668\u4eba\u5173\u4e8e\u6587\u6863\u7684\u5185\u5bb9","text":"

\u81f3\u6b64\u4e00\u4e2a\u7b80\u5355\u7684\u6587\u6863\u95ee\u7b54\u5e94\u7528Chat PDF\u642d\u5efa\u5b8c\u6210\uff0c\u70b9\u51fb\u201c\u9884\u89c8\u201d\uff0c\u67e5\u770b\u6548\u679c\u5982\u4f55\ud83d\udc47

\u7ed3\u679c\u5982\u4e0b\uff1a

"},{"location":"zh/usage/plugin/Dify/#_5","title":"\u7b2c\u4e03\u6b65\uff1a\u53d1\u5e03\u4e0e\u6d4b\u8bd5","text":"

\u4fdd\u5b58\u5e76\u53d1\u5e03\u4f60\u7684\u5e94\u7528\u3002\u73b0\u5728\uff0c\u4e0a\u4f20\u4e00\u4efdPDF\u6216\u56fe\u7247\uff0c\u4f60\u5c31\u53ef\u4ee5\u548c\u5b83\u81ea\u7531\u5bf9\u8bdd\u4e86\uff01

"},{"location":"zh/usage/plugin/Dify/#s3","title":"\u6848\u4f8b\u4e8c\uff1a\u81ea\u52a8\u5316\u6279\u91cf\u5904\u7406\u6587\u6863\uff0c\u5e76\u4e0a\u4f20\u81f3\u4e91\u7aefS3","text":"

\u9700\u8981\u5904\u7406\u5927\u91cf\u6587\u6863\u5e76\u5f52\u6863\uff1fMinerU \u63d2\u4ef6\u540c\u6837\u80fd\u80dc\u4efb

"},{"location":"zh/usage/plugin/Dify/#botos3","title":"\u7b2c\u4e00\u6b65\uff1a\u5b89\u88c5 botos3 \u63d2\u4ef6","text":""},{"location":"zh/usage/plugin/Dify/#s3-bucket","title":"\u7b2c\u4e8c\u6b65\uff1a\u914d\u7f6e S3 bucket","text":""},{"location":"zh/usage/plugin/Dify/#_6","title":"\u7b2c\u4e09\u6b65\uff1a\u521b\u5efa\u5de5\u4f5c\u6d41","text":"

\u9009\u62e9\u5b57\u6bb5\u7c7b\u578b\u4e3a\u201c\u6587\u4ef6\u5217\u8868\u201d\uff0c\u586b\u5199\u53d8\u91cf\u540d\u79f0\uff08\u6b64\u5904\u586b\u4e3ainput_files\uff09,\u652f\u6301\u7684\u6587\u6863\u7c7b\u578b\u9009\u4e3a\u6587\u6863\u4e0e\u56fe\u7247

"},{"location":"zh/usage/plugin/Dify/#_7","title":"\u7b2c\u56db\u6b65\uff1a\u6dfb\u52a0\u201c\u8fed\u4ee3\u201d","text":"

\u5728\u201c\u5f00\u59cb\u201d\u8282\u70b9\u540e\u6dfb\u52a0\u201c\u8fed\u4ee3\u201d\uff0c\u5e76\u914d\u7f6e\u8fed\u4ee3\u5185\u7684MinerU\u8282\u70b9,\u8bbe\u7f6e\u8fed\u4ee3\u7684\u8f93\u5165\u4e3a\u4e0a\u4e00\u6b65\u5f00\u59cb\u8282\u70b9\u7684upload_files\uff0c\u8f93\u51fa\u8282\u70b9\u6682\u65f6\u4e0d\u586b\u5199\uff0c\u518d\u6574\u4e2a\u8fed\u4ee3\u914d\u7f6e\u5b8c\u6210\u540e\u9009\u62e9MinerU\u8282\u70b9Parse File\u7684full_zip_url

\u5c06MinerU\u7684\u8f93\u5165\u53c2\u6570file\u9009\u62e9\u4e3a\u8fed\u4ee3\u5668\u7684 item

"},{"location":"zh/usage/plugin/Dify/#mineru_1","title":"\u7b2c\u4e94\u6b65\uff1a\u589e\u52a0\u4e2d\u95f4\u8282\u70b9\u201c\u4ee3\u7801\u6267\u884c\u201d\u6765\u8f6c\u6362MinerU\u7684\u89e3\u6790\u7ed3\u679c","text":"

\u8f93\u5165\u53d8\u91cf(\u53d8\u91cf\u540d\u79f0\u9700\u4e0e\u4ee3\u7801\u5b9a\u4e49\u4e00\u81f4)

text\uff1a\u9009\u62e9MinerU Parse File\u7684\u8f93\u51fa\u53d8\u91cftext
uploadFiles\uff1a\u9009\u62e9\u201c\u5f00\u59cb\u201d\u8282\u70b9\u7684\u6587\u4ef6\u5217\u8868upload_files\uff0c\u7528\u6765\u6839\u636e\u8fed\u4ee3\u7684index\u7d22\u5f15\u4e0b\u6807\u627e\u5230\u5bf9\u5e94\u7684\u539f\u59cb\u6587\u4ef6\u540d
index\uff1a\u8fed\u4ee3\u7684\u4e0b\u6807\u7d22\u5f15\uff0c\u9009\u62e9\u8fed\u4ee3\u5668\u7684index

\u8f93\u51fa\u53d8\u91cf(\u53d8\u91cf\u540d\u79f0\u9700\u4e0e\u4ee3\u7801\u5b9a\u4e49\u4e00\u81f4)

fileName\uff1aString
base64\uff1aString

\u4ee3\u7801\u9009\u62e9JavaScript\uff0c\u7f16\u5199\u8f6c\u6362\u4ee3\u7801\uff1a

\u6682\u65f6\u65e0\u6cd5\u5728\u98de\u4e66\u6587\u6863\u5916\u5c55\u793a\u6b64\u5185\u5bb9

\u4ee5\u4e0b\u4e3aPython\u7248\u672c\uff1a

\u6682\u65f6\u65e0\u6cd5\u5728\u98de\u4e66\u6587\u6863\u5916\u5c55\u793a\u6b64\u5185\u5bb9

"},{"location":"zh/usage/plugin/Dify/#botos3_1","title":"\u7b2c\u516d\u6b65\uff1a\u914d\u7f6e Botos3 \u63d2\u4ef6\u6765\u4e0a\u4f20\u5185\u5bb9","text":"

\u6dfb\u52a0\u5de5\u5177\u8282\u70b9Botos3\uff0c\u9009\u62e9\u201c\u901a\u8fc7s3\u4e0a\u4f20base64\u201d

\u6587\u4ef6base64\u9009\u62e9\u4ee3\u7801\u6267\u884c\uff08\u56fe\u4e2d\u4e3a\u8f6c\u6362MINERU MD\u6587\u672c\uff09\u8f93\u51fa\u7684base64\u5b57\u6bb5

S3\u5bf9\u8c61key\uff0cS3 \u5bf9\u8c61key\u586b\u5199\u6587\u4ef6\u5b58\u50a8\u7684\u8def\u5f84\uff0c\u5728 botos3 \u63d2\u4ef6\u914d\u7f6e\u754c\u9762\u5df2\u7ecf\u586b\u5199\u4e86 bucket \u540d\u79f0\uff0c\u8fd9\u91cc\u53ea\u9700\u8981\u586b\u5199\u5728bucket\u4e0b\u5b58\u50a8\u7684\u76ee\u5f55\u5373\u53ef\u3002\u9009\u62e9\u4ee3\u7801\u6267\u884c\uff08\u56fe\u4e2d\u4e3a\u8f6c\u6362MINERU MD\u6587\u672c\uff09\u7684fileName

"},{"location":"zh/usage/plugin/Dify/#_8","title":"\u7b2c\u4e03\u6b65\uff1a\u9884\u89c8\u6548\u679c","text":"

\u8fde\u63a5\u7ed3\u675f\u8282\u70b9\uff0c\u81f3\u6b64\uff0c\u4e00\u4e2a\u7b80\u5355\u7684\u4e0a\u4f20\u5230s3\u7684\u5de5\u4f5c\u6d41\u914d\u7f6e\u5b8c\u6210\uff0c\u70b9\u51fb\u201c\u8fd0\u884c\u201d\u770b\u770b\u6548\u679c\ud83d\udc47\uff1a

"},{"location":"zh/usage/plugin/Dify/#vis3","title":"\u7b2c\u516b\u6b65\uff1aVis3\u67e5\u770b\u6587\u6863","text":"

\u8fd0\u884c\u7ed3\u675f\uff0c\u53ef\u901a\u8fc7vis3\u6765\u67e5\u770bS3\u6876\u5185\u662f\u5426\u5df2\u4e0a\u4f20\u89e3\u6790\u540e\u7684md\u6587\u4ef6\uff0cVis3\u4f7f\u7528\u53ef\u53c2\u8003

\u65b0\u5de5\u5177\u5f00\u6e90\uff01Vis3\u5927\u6a21\u578b\u6570\u636e\u53ef\u89c6\u5316\u5229\u5668\uff1a\u586b AK/SK \u76f4\u63a5\u9884\u89c8 S3 \u6570\u636e\uff0cJSON/\u89c6\u9891/\u56fe\u7247\u79d2\u5f00\uff01\u672c\u5730\u6587\u4ef6\u4e5f\u53ef\u7528

"},{"location":"zh/usage/plugin/DingTalk/","title":"\u9489\u9489\u7b80\u4ecb","text":""},{"location":"zh/usage/plugin/DingTalk/#_1","title":"\u9489\u9489\u7b80\u4ecb","text":"

\u9489\u9489\uff08DingTalk\uff09\u662f\u963f\u91cc\u5df4\u5df4\u96c6\u56e2\u6253\u9020\u7684\u4f01\u4e1a\u7ea7\u667a\u80fd\u79fb\u52a8\u529e\u516c\u5e73\u53f0\uff0c\u662f\u6570\u5b57\u7ecf\u6d4e\u65f6\u4ee3\u7684\u4f01\u4e1a\u7ec4\u7ec7\u534f\u540c\u529e\u516c\u548c\u5e94\u7528\u5f00\u53d1\u5e73\u53f0\u3002\u9489\u9489\u6574\u5408\u4e86 IM \u5373\u65f6\u6c9f\u901a\u3001\u9489\u9489\u6587\u6863\u3001\u9489\u95ea\u4f1a\u3001\u9489\u76d8\u3001Teambition\u3001OA\u5ba1\u6279\u3001\u667a\u80fd\u4eba\u4e8b\u3001\u9489\u5de5\u724c\u3001\u5de5\u4f5c\u53f0\u7b49\u529f\u80fd\uff0c\u65e8\u5728\u5b9e\u73b0\u7b80\u5355\u3001\u9ad8\u6548\u3001\u5b89\u5168\u3001\u667a\u80fd\u7684\u6570\u5b57\u5316\u5de5\u4f5c\u65b9\u5f0f\u3002\u5b83\u652f\u6301\u4f01\u4e1a\u7ec4\u7ec7\u6570\u5b57\u5316\u548c\u4e1a\u52a1\u6570\u5b57\u5316\uff0c\u8986\u76d6\u201c\u4eba\u3001\u8d22\u3001\u7269\u3001\u4e8b\u3001\u4ea7\u3001\u4f9b\u3001\u9500\u3001\u5b58\u201d\u7684\u5168\u94fe\u8def\u7ba1\u7406\u3002

\u901a\u8fc7\u9489\u9489\u5f00\u653e\u5e73\u53f0\u4e0a\u7684SaaS\u8f6f\u4ef6\uff0c\u4f01\u4e1a\u53ef\u4f4e\u6210\u672c\u642d\u5efa\u6570\u5b57\u5316\u5e94\u7528\uff0c\u6574\u5408\u6240\u6709\u6570\u5b57\u5316\u7cfb\u7edf\u3002\u6b64\u5916\uff0c\u9489\u9489\u63d0\u4f9b\u8d85\u8fc72000\u4e2aAPI\u63a5\u53e3\uff0c\u4e3a\u4f01\u4e1a\u6570\u5b57\u5316\u8f6c\u578b\u63d0\u4f9b\u5f00\u653e\u517c\u5bb9\u73af\u5883\u3002\u4e0d\u4f1a\u4ee3\u7801\u7684\u7528\u6237\u4e5f\u53ef\u5229\u7528\u4f4e\u4ee3\u7801\u5de5\u5177\u6784\u5efaCRM\u3001ERP\u3001OA\u3001\u9879\u76ee\u7ba1\u7406\u3001\u8fdb\u9500\u5b58\u7b49\u7cfb\u7edf\u3002

\u76ee\u524d\uff0c\u9489\u9489\u6587\u6863\u3001AI \u8868\u683c\u7b49\u4ea7\u54c1\u6b64\u524d\u5df2\u6df1\u5ea6\u96c6\u6210 MinerU \u80fd\u529b\uff0c\u5e76\u901a\u8fc7\u5f00\u653e\u5e73\u53f0\u5411\u751f\u6001\u5f00\u53d1\u8005\u5f00\u653e\u6587\u6863\u89e3\u6790\u529f\u80fd\uff0c\u4e3a DLU \u7684\u8054\u5408\u7814\u53d1\u63d0\u4f9b\u4e86\u624e\u5b9e\u7684\u6280\u672f\u4e0e\u573a\u666f\u57fa\u7840\u3002

\u9489\u9489\u5b98\u7f51\uff1ahttps://www.dingtalk.com/

"},{"location":"zh/usage/plugin/FastGPT/","title":"FastGPT \u7b80\u4ecb","text":""},{"location":"zh/usage/plugin/FastGPT/#fastgpt","title":"FastGPT \u7b80\u4ecb","text":"

FastGPT \u662f\u4e00\u4e2a\u57fa\u4e8e LLM \u5927\u8bed\u8a00\u6a21\u578b\u7684\u77e5\u8bc6\u5e93\u95ee\u7b54\u7cfb\u7edf\uff0c\u5c06\u667a\u80fd\u5bf9\u8bdd\u4e0e\u53ef\u89c6\u5316\u7f16\u6392\u5b8c\u7f8e\u7ed3\u5408\uff0c\u8ba9 AI \u5e94\u7528\u5f00\u53d1\u53d8\u5f97\u7b80\u5355\u81ea\u7136\u3002\u65e0\u8bba\u60a8\u662f\u5f00\u53d1\u8005\u8fd8\u662f\u4e1a\u52a1\u4eba\u5458\uff0c\u90fd\u80fd\u8f7b\u677e\u6253\u9020\u4e13\u5c5e\u7684 AI \u5e94\u7528\u3002

\u5b98\u7f51\u5730\u5740\uff1ahttps://fastgpt.cn
Miner fastGPT \u63d2\u4ef6\u4e0b\u8f7d\u5730\u5740\uff1ahttps://cloud.fastgpt.io/dashboard/systemPlugin?type=productivity

"},{"location":"zh/usage/plugin/ModelWhale/","title":"ModelWhale \u7b80\u4ecb","text":""},{"location":"zh/usage/plugin/ModelWhale/#modelwhale","title":"ModelWhale \u7b80\u4ecb","text":"

ModelWhale\u662f\u4e00\u6b3e\u9ad8\u6548\u7387\u7684\u6570\u636e\u79d1\u5b66\u4e91\u7aef\u534f\u4f5c\u5de5\u5177\uff0c\u4e3a\u6570\u636e\u5de5\u4f5c\u8005\u63d0\u4f9b\u4e86\u5373\u5f00\u5373\u7528\u7684\u4e91\u7aef\u5206\u6790\u73af\u5883\uff0cJupyter Notebook \u4ea4\u4e92\u5f0f\u548cCanvas \u62d6\u62fd\u5f0f\u4e24\u79cd\u5206\u6790\u754c\u9762\uff0c\u5e2e\u52a9\u79d1\u7814\u8005\u3001\u6559\u80b2\u5de5\u4f5c\u8005\u89e3\u51b3\u5e95\u5c42\u5de5\u7a0b\u7e41\u590d\u3001\u6570\u636e\u96be\u4ee5\u5b89\u5168\u5e94\u7528\u3001\u6210\u679c\u6d41\u8f6c\u590d\u73b0\u56f0\u96be\u7b49\u95ee\u9898\u3002\u57fa\u4e8e\u4e0d\u540c\u4f7f\u7528\u573a\u666f\uff0cModelWhale \u4e3a\u7528\u6237\u63d0\u4f9b\u4e09\u4e2a\u4ea7\u54c1\u7248\u672c\uff0c\u5206\u522b\u662f\u57fa\u7840\u7248\u3001\u4e13\u4e1a\u7248\u3001\u56e2\u961f\u7248\u3002

\u76ee\u524d\uff0cMinerU \u63d2\u4ef6\u5df2\u5728 ModelWhale \u5de5\u4f5c\u4e2d\uff0c\u901a\u8fc7\u5176\u5f3a\u5927\u7684\u6587\u6863\u89e3\u6790\u80fd\u529b\uff0c\u4e3a\u7528\u6237\u642d\u5efa\u667a\u80fd\u4f53\u4e0e\u5de5\u4f5c\u6d41\u63d0\u4f9b\u6587\u6863\u89e3\u6790\u80fd\u529b\uff0c\u52a0\u5feb\u7528\u6237 AI \u5e94\u7528\u7684\u5f00\u53d1\u3002

images/DingTalk_01.png

ModelWhale \u5b98\u7f51\uff1aMohttps://www.modelwhale.com/pricing?scroll=1
MinerU \u5728ModelWhale \u7684\u4f7f\u7528\u5730\u5740\uff1ahttps://www.heywhale.com/org/7b38d/workspace/iframe?url=https://www.heywhale.com/api/model/services/68089d360b1519a862ccb9b4/app/

"},{"location":"zh/usage/plugin/RagFlow/","title":"RagFlow","text":""},{"location":"zh/usage/plugin/RagFlow/#ragflow","title":"RAGFlow","text":"

RAGFlow \u662f\u4e00\u6b3e\u5f00\u6e90 RAG\uff08Retrieval-Augmented Generation\uff09\u5f15\u64ce\u4e0e\u5e94\u7528\u5e73\u53f0\uff0c\u6df1\u5ea6\u878d\u5408\u4e86\u6df1\u5ea6\u6587\u6863\u7406\u89e3\u3001\u81ea\u52a8\u5316 RAG \u5de5\u4f5c\u6d41\u4e0e\u5927\u6a21\u578b\u8c03\u7528\uff0c\u6253\u901a\u4e86\u590d\u6742\u6570\u636e\u5904\u7406\u3001\u77e5\u8bc6\u68c0\u7d22\u3001\u589e\u5f3a\u751f\u6210\u7684\u5168\u6d41\u7a0b\uff0c\u65e8\u5728\u4e3a\u4f01\u4e1a\u53ca\u5f00\u53d1\u8005\u63d0\u4f9b\u4e00\u7ad9\u5f0f\u667a\u80fd\u95ee\u7b54\u5f00\u53d1\u670d\u52a1\uff0c\u5e76\u652f\u6301\u5404\u7c7b\u590d\u6742\u573a\u666f\u4e0b\u5927\u6a21\u578b\u7684\u6784\u5efa\u4e0e\u5e94\u7528\u843d\u5730\u3002

\u76ee\u524d\uff0cMinerU \u5df2\u6df1\u5ea6\u96c6\u6210\u81f3 RAGFlow \u77e5\u8bc6\u5e93\u5728\u7ebf\u7248\u672c\uff0c\u4f5c\u4e3a\u5185\u7f6e PDF \u6587\u6863\u89e3\u6790\u5668\uff0c\u4e3a\u7528\u6237\u77e5\u8bc6\u5e93\u642d\u5efa\u63d0\u4f9b\u4e13\u4e1a\u3001\u53ef\u9760\u7684\u6587\u6863\u89e3\u6790\u652f\u6301\u3002\u672c\u5730\u90e8\u7f72\u7248\u672c\u90e8\u7f72\u4f7f\u7528\u65b9\u5f0f\u8be6\u89c1\u4e0b\u65b9\u4f7f\u7528\u6559\u7a0b\u3002

\u4f7f\u7528\u53ef\u8bbf\u95ee\uff1ahttps://demo.ragflow.io/

"},{"location":"zh/usage/plugin/RagFlow/#ragflow-mineru","title":"\u4f7f\u7528\u6559\u7a0b\uff1a\u5982\u4f55\u5728 RAGFlow \u4e2d\u4f7f\u7528 MinerU","text":""},{"location":"zh/usage/plugin/RagFlow/#_1","title":"\u4e00\u3001\u5b89\u88c5\u914d\u7f6e","text":"

\u9996\u5148\uff0c\u6211\u4eec\u5efa\u8bae\u60a8\u901a\u8fc7 docker \u7684\u5f62\u5f0f\u5728\u672c\u5730\u90e8\u7f72 RagFlow \u4ee5\u65b9\u4fbf\u4f7f\u7528 MinerU \u63d2\u4ef6\u4f5c\u4e3a\u89e3\u6790\u5de5\u5177\u3002\u5728\u5b89\u88c5\u5b8c RagFlow \u540e\u6267\u884c\uff1a

\u7248\u672c\u68c0\u67e5\uff1a

\u786e\u4fdd\u4f60\u7684RAGFlow\u7248\u672c >= v0.21.1\u3002

\u66f4\u65b0 .env \u6587\u4ef6\uff1a

\u4e3a\u4e86\u786e\u4fdd\u670d\u52a1\u80fd\u88ab\u5e73\u7a33\u4fee\u6539\uff0c\u5efa\u8bae\u5148\u5728 cmd \u8fd0\u884c docker compose down \u505c\u6389\u670d\u52a1\u3002

\u6253\u5f00 .env \u6587\u4ef6\uff0c\u5728\u6587\u4ef6\u7684\u672b\u5c3e\uff0c\u6dfb\u52a0\u8fd9\u4e24\u884c\u4ee3\u7801\uff0c\u4fdd\u5b58\u6587\u4ef6\u3002

HF_ENDPOINT=https://hf-mirror.com\nMINERU_EXECUTABLE=/ragflow/uv_tools/.venv/bin/mineru\n

\u542f\u52a8\u5e76\u8fdb\u5165\u5bb9\u5668\uff1a

\u5728 cmd \u4e2d\uff0c\u91cd\u65b0\u542f\u52a8\u670d\u52a1\uff1adocker compose up -d

\u7b49\u5f85\u670d\u52a1\u5168\u90e8 Running \u6216 Healthy \u540e\uff0c\u8fd0\u884c\u4ee5\u4e0b\u547d\u4ee4\u8fdb\u5165RAGFlow\u7684\u6838\u5fc3\u5bb9\u5668\uff1a

docker compose exec ragflow-cpu bash\n

\uff08\u4f60\u7684\u547d\u4ee4\u884c\u63d0\u793a\u7b26\u4f1a\u4ece C:\\...> \u53d8\u4e3a root@...\uff09

\u5728\u5bb9\u5668\u5185\u4e0b\u8f7d MinerU \u6a21\u578b\uff1a

\u5728\u5bb9\u5668\u5185\u90e8\uff0c\u4f9d\u6b21\u8fd0\u884c\u4ee5\u4e0b 5 \u6761\u547d\u4ee4

mkdir uv_tools\ncd uv_tools\nuv venv .venv\nsource .venv/bin/activate\nuv pip install -U \"mineru[core]\" -i https://mirrors.aliyun.com/pypi/simple\n

\u9000\u51fa\u5e76\u91cd\u542f\uff1a

\u5b89\u88c5\u5b8c\u6210\u540e\uff0c\u8f93\u5165 exit \u5e76\u6309\u56de\u8f66\u3002

\u8fd0\u884c\u91cd\u542f\u547d\u4ee4\uff0c\u8ba9 RAGFlow \u52a0\u8f7d\u521a\u88c5\u597d\u7684 MinerU

docker compose restart ragflow-cpu\n

"},{"location":"zh/usage/plugin/RagFlow/#_2","title":"\u4e8c\u3001\u4f7f\u7528\u5165\u53e3","text":"

\u5728\u672c\u5730\u90e8\u7f72\u5b8c\u6bd5\u540e\uff0c\u8981\u542f\u7528 MinerU\uff0c\u60a8\u9700\u8981\u5728\u8fdb\u5165 RagFlow \u7279\u5b9a\u77e5\u8bc6\u5e93\u7684\u914d\u7f6e\u9875\u9762\u5e76\u9009\u62e9 MinerU \u4f5c\u4e3a\u9ed8\u8ba4\u7684 PDF \u89e3\u6790\u5668\u3002\uff08\u6ce8\uff1aRagFlow \u5728\u7ebf\u7248\u4e2d\u5df2\u7ecf\u5185\u7f6e\u4e86 MinerU \u63d2\u4ef6\u4e3a\u60a8\u63d0\u4f9b\u4e86\u9ad8\u7ea7\u7684 PDF \u6587\u4ef6\u89e3\u6790\u80fd\u529b\uff0c\u4f7f\u7528\u65b9\u5f0f\u4e0e\u6b64\u4e00\u81f4\u3002\uff09

\u5165\u53e3\u548c\u914d\u7f6e\u6b65\u9aa4\uff1a

\u8fdb\u5165\u77e5\u8bc6\u5e93\u914d\u7f6e\uff1a
\u9996\u5148\uff0c\u5728\u60a8\u7684\u77e5\u8bc6\u5e93\u7ba1\u7406\u754c\u9762\uff0c\u9009\u62e9\u60a8\u9700\u8981\u914d\u7f6e\u7684\u7279\u5b9a\u77e5\u8bc6\u5e93\uff08\u4f8b\u5982\u56fe\u793a\u4e2d\u7684 \"content\" \u77e5\u8bc6\u5e93\uff09\u3002
\u5728\u77e5\u8bc6\u5e93\u8be6\u60c5\u9875\u9762\u7684\u5de6\u4fa7\u5bfc\u822a\u680f\u4e2d\uff0c\u70b9\u51fb\u3010\u914d\u7f6e\u3011\u9009\u9879\u5361\u3002
\u5b9a\u4f4d PDF \u89e3\u6790\u5668\u8bbe\u7f6e\uff1a
\u5411\u4e0b\u6eda\u52a8\u9875\u9762\uff0c\u627e\u5230\u201cIngestion pipeline\u201d\uff08\u6444\u53d6\u7ba1\u9053\uff09\u8bbe\u7f6e\u90e8\u5206\u3002
\u5728\u6b64\u90e8\u5206\u4e2d\uff0c\u60a8\u4f1a\u770b\u5230\u4e00\u4e2a\u540d\u4e3a\u3010PDF\u89e3\u6790\u5668\u3011\uff08PDF Parser\uff09\u7684\u9009\u9879\u3002
\u9009\u62e9 MinerU\uff1a
\u70b9\u51fb\u3010PDF\u89e3\u6790\u5668\u3011\u65c1\u8fb9\u7684\u4e0b\u62c9\u83dc\u5355\u3002
\u4ece\u53ef\u7528\u9009\u9879\u4e2d\uff0c\u9009\u62e9\u3010MinerU\u3011\u3002
\u4fdd\u5b58\u4fee\u6539\uff1a
\u5b8c\u6210\u9009\u62e9\u540e\uff0c\u8bf7\u52a1\u5fc5\u70b9\u51fb\u9875\u9762\u5e95\u90e8\u7684\u3010\u4fdd\u5b58\u3011\u6309\u94ae\uff0c\u4ee5\u4f7f\u66f4\u6539\u751f\u6548\u3002

"},{"location":"zh/usage/plugin/Sider/","title":"Sider \u7b80\u4ecb","text":""},{"location":"zh/usage/plugin/Sider/#sider","title":"Sider \u7b80\u4ecb","text":"

Sider \u662f\u4e00\u6b3e\u6d4f\u89c8\u5668\u4fa7\u8fb9\u680f\u7c7b\u7684 AI \u52a9\u624b\u6269\u5c55\uff0c\u4e3b\u8981\u5728\u7f51\u9875\u53f3\u4fa7\u5f00\u542f\u4e00\u4e2a\u201c\u968f\u5904\u53ef\u7528\u201d\u7684\u667a\u80fd\u9762\u677f\uff0c\u5c06\u5bf9\u8bdd\u5f0f AI\uff08\u5982 GPT\u3001Claude\u3001Gemini \u7b49\uff09\u5e26\u5230\u4f60\u6b63\u5728\u6d4f\u89c8\u7684\u4efb\u4f55\u9875\u9762\u4e2d\u3002\u5b83\u7684\u6838\u5fc3\u5b9a\u4f4d\u662f\uff1a\u63d0\u5347\u9605\u8bfb\u3001\u5199\u4f5c\u3001\u7ffb\u8bd1\u3001\u68c0\u7d22\u4e0e\u603b\u7ed3\u6548\u7387\uff0c\u5e76\u4e0e\u7f51\u9875\u5185\u5bb9\u6df1\u5ea6\u8054\u52a8\u3002

\u76ee\u524d\uff0cSider\u5728 Wisebase \u6a21\u5757\u4e2d\u6df1\u5ea6\u96c6\u6210\u4e86 MinerU \u7684\u76f8\u5173\u529f\u80fd\u3002\u8be5\u6a21\u5757\u662f\u4e00\u4e2a\u7531AI\u9a71\u52a8\u7684\u77e5\u8bc6\u5e93\uff0c\u60a8\u53ef\u4ee5\u901a\u8fc7\u4e0a\u4f20 PDF \u7b49\u5404\u7c7b\u578b\u6587\u4ef6\uff0c\u6784\u5efa\u4e2a\u4eba\u56fe\u4e66\u9986\u4ee5\u5b9e\u73b0\u9ad8\u6548\u7684\u77e5\u8bc6\u7ba1\u7406\uff0cMinerU \u53ef\u4ee5\u5e2e\u52a9\u60a8\u66f4\u597d\u5730\u89e3\u6790\u6b64\u7c7b\u6587\u4ef6\uff0c\u7cbe\u51c6\u5730\u63d0\u53d6\u6587\u4ef6\u4e2d\u7684\u4fe1\u606f\u3002

Sider \u5b98\u7f51\u5730\u5740\uff1ahttps://sider.ai/zh-CN/chat
\u4f7f\u7528\u96c6\u6210 MinerU \u76f8\u5173\u529f\u80fd\u7684 Sider \u5730\u5740\uff1ahttps://sider.ai/zh-CN/wisebase

"},{"location":"zh/usage/plugin/n8n/","title":"n8n \u7b80\u4ecb","text":""},{"location":"zh/usage/plugin/n8n/#n8n","title":"n8n \u7b80\u4ecb","text":"

n8n \u662f\u4e00\u6b3e\u4ee5\u4f4e\u4ee3\u7801\uff08Low-code\uff09\u3001\u5de5\u4f5c\u6d41\u81ea\u52a8\u5316\u4e3a\u6838\u5fc3\u7684\u5e94\u7528\u5f00\u53d1\u5e73\u53f0\uff0c\u8bb8\u591a\u4f01\u4e1a\u90fd\u501f\u52a9\u4e8e\u5176\u7075\u6d3b\u7684\u8282\u70b9\uff08Node\uff09\u914d\u7f6e\uff0c\u5b9e\u73b0\u4e1a\u52a1\u6d41\u7a0b\u7684\u81ea\u52a8\u5316\u6267\u884c\u3002\u5b83\u901a\u8fc7\u53ef\u89c6\u5316\u754c\u9762\u548c\u4ee3\u7801\u6269\u5c55\u80fd\u529b\uff0c\u5e2e\u52a9\u7528\u6237\u8fde\u63a5\u5404\u79cd\u5e94\u7528\u7a0b\u5e8f\u548c\u670d\u52a1\uff0c\u6784\u5efa\u590d\u6742\u7684\u81ea\u52a8\u5316\u6d41\u7a0b,\u964d\u4f4e\u7528\u6237\u4f7f\u7528\u95e8\u69db\u3002

\u76ee\u524d\uff0cMinerU \u5df2\u5c06\u5176\u5f3a\u5927\u7684\u6587\u6863\u89e3\u6790\u80fd\u529b\u5c01\u88c5\u4e3a n8n \u8282\u70b9\uff0c\u7528\u6237\u5728\u642d\u5efa\u5de5\u4f5c\u6d41\u65f6\uff0c\u53ef\u4ee5\u66f4\u52a0\u4fbf\u6377\u5730\u5904\u7406\u590d\u6742\u7684\u6587\u6863\u89e3\u6790\u4efb\u52a1\u3002

n8n \u5b98\u7f51\u5730\u5740\uff1ahttps://n8n.io/
MinerU n8n \u63d2\u4ef6\u4e0b\u8f7d\u5730\u5740\uff1ahttps://www.npmjs.com/package/n8n-nodes-mineru

"},{"location":"zh/usage/plugin/n8n/#mineru-n8n","title":"MinerU \u5728 n8n \u4e2d\u7684\u4f7f\u7528\u65b9\u6cd5","text":""},{"location":"zh/usage/plugin/n8n/#step1-node","title":"step1 \u8fdb\u5165\u793e\u533anode\u5b89\u88c5\u754c\u9762","text":""},{"location":"zh/usage/plugin/n8n/#step2-n8n-nodes-mineru","title":"step2 \u5b89\u88c5 n8n-nodes-mineru \u8282\u70b9","text":"

\u2248assets/images/n8n_2.png)

"},{"location":"zh/usage/plugin/n8n/#step3-n8n-nodes-mineru-api-key","title":"step3 \u65b0\u5efa\u5de5\u4f5c\u6d41\uff0c\u6dfb\u52a0 n8n-nodes-mineru \u8282\u70b9\uff0c\u5e76\u8bbe\u7f6e api key","text":""},{"location":"zh/usage/plugin/n8n/#n8n_1","title":"n8n\u4f7f\u7528\u8282\u70b9\u6587\u6863","text":"

https://www.npmjs.com/package/n8n-nodes-mineru

"},{"location":"zh/usage/plugin/n8n/#_1","title":"\u5728\u5de5\u4f5c\u6d41\u5185\u96c6\u6210\u89e3\u538b\u529f\u80fd","text":""},{"location":"zh/usage/plugin/n8n/#json","title":"\u5bfc\u5165 json \u6a21\u677f","text":"

\u6682\u65f6\u65e0\u6cd5\u5728\u98de\u4e66\u6587\u6863\u5916\u5c55\u793a\u6b64\u5185\u5bb9

"},{"location":"zh/usage/plugin/n8n/#url","title":"\u914d\u7f6e \u51ed\u8bc1\u548c\u6587\u6863url","text":""},{"location":"zh/usage/plugin/n8n/#_2","title":"\u6839\u636e\u5404\u81ea\u7684\u9700\u6c42\u914d\u7f6e\u6240\u9700\u7684\u8f93\u51fa","text":""},{"location":"zh/usage/plugin/n8n/#_3","title":"\u8c03\u8bd5","text":""}]} \ No newline at end of file diff --git a/sitemap.xml b/sitemap.xml new file mode 100644 index 00000000..86de3f1c --- /dev/null +++ b/sitemap.xml @@ -0,0 +1,343 @@ + + + + https://opendatalab.github.io/MinerU/ + 2026-03-26 + daily + + + + + https://opendatalab.github.io/MinerU/demo/ + 2026-03-26 + daily + + + + + https://opendatalab.github.io/MinerU/faq/ + 2026-03-26 + daily + + + + + https://opendatalab.github.io/MinerU/quick_start/ + 2026-03-26 + daily + + + + + https://opendatalab.github.io/MinerU/quick_start/docker_deployment/ + 2026-03-26 + daily + + + + + https://opendatalab.github.io/MinerU/quick_start/extension_modules/ + 2026-03-26 + daily + + + + + https://opendatalab.github.io/MinerU/reference/ + 2026-03-26 + daily + + + + + https://opendatalab.github.io/MinerU/reference/changelog/ + 2026-03-26 + daily + + + + + https://opendatalab.github.io/MinerU/reference/output_files/ + 2026-03-26 + daily + + + + + https://opendatalab.github.io/MinerU/usage/ + 2026-03-26 + daily + + + + + https://opendatalab.github.io/MinerU/usage/advanced_cli_parameters/ + 2026-03-26 + daily + + + + + https://opendatalab.github.io/MinerU/usage/cli_tools/ + 2026-03-26 + daily + + + + + https://opendatalab.github.io/MinerU/usage/model_source/ + 2026-03-26 + daily + + + + + https://opendatalab.github.io/MinerU/usage/quick_usage/ + 2026-03-26 + daily + + + + + https://opendatalab.github.io/MinerU/zh/ + 2026-03-26 + daily + + + + + https://opendatalab.github.io/MinerU/zh/demo/ + 2026-03-26 + daily + + + + + https://opendatalab.github.io/MinerU/zh/faq/ + 2026-03-26 + daily + + + + + https://opendatalab.github.io/MinerU/zh/quick_start/ + 2026-03-26 + daily + + + + + https://opendatalab.github.io/MinerU/zh/quick_start/docker_deployment/ + 2026-03-26 + daily + + + + + https://opendatalab.github.io/MinerU/zh/quick_start/extension_modules/ + 2026-03-26 + daily + + + + + https://opendatalab.github.io/MinerU/zh/reference/ + 2026-03-26 + daily + + + + + https://opendatalab.github.io/MinerU/zh/reference/changelog/ + 2026-03-26 + daily + + + + + https://opendatalab.github.io/MinerU/zh/reference/output_files/ + 2026-03-26 + daily + + + + + https://opendatalab.github.io/MinerU/zh/usage/ + 2026-03-26 + daily + + + + + https://opendatalab.github.io/MinerU/zh/usage/advanced_cli_parameters/ + 2026-03-26 + daily + + + + + https://opendatalab.github.io/MinerU/zh/usage/cli_tools/ + 2026-03-26 + daily + + + + + https://opendatalab.github.io/MinerU/zh/usage/model_source/ + 2026-03-26 + daily + + + + + https://opendatalab.github.io/MinerU/zh/usage/quick_usage/ + 2026-03-26 + daily + + + + + https://opendatalab.github.io/MinerU/zh/usage/acceleration_cards/AMD/ + 2026-03-26 + daily + + + + https://opendatalab.github.io/MinerU/zh/usage/acceleration_cards/Ascend/ + 2026-03-26 + daily + + + + https://opendatalab.github.io/MinerU/zh/usage/acceleration_cards/Biren/ + 2026-03-26 + daily + + + + https://opendatalab.github.io/MinerU/zh/usage/acceleration_cards/Cambricon/ + 2026-03-26 + daily + + + + https://opendatalab.github.io/MinerU/zh/usage/acceleration_cards/Enflame/ + 2026-03-26 + daily + + + + https://opendatalab.github.io/MinerU/zh/usage/acceleration_cards/Hygon/ + 2026-03-26 + daily + + + + https://opendatalab.github.io/MinerU/zh/usage/acceleration_cards/IluvatarCorex/ + 2026-03-26 + daily + + + + https://opendatalab.github.io/MinerU/zh/usage/acceleration_cards/Kunlunxin/ + 2026-03-26 + daily + + + + https://opendatalab.github.io/MinerU/zh/usage/acceleration_cards/METAX/ + 2026-03-26 + daily + + + + https://opendatalab.github.io/MinerU/zh/usage/acceleration_cards/MooreThreads/ + 2026-03-26 + daily + + + + https://opendatalab.github.io/MinerU/zh/usage/acceleration_cards/THead/ + 2026-03-26 + daily + + + + https://opendatalab.github.io/MinerU/zh/usage/acceleration_cards/Tecorigin/ + 2026-03-26 + daily + + + + https://opendatalab.github.io/MinerU/zh/usage/acceleration_cards/VastAI/ + 2026-03-26 + daily + + + + https://opendatalab.github.io/MinerU/zh/usage/plugin/BISHENG/ + 2026-03-26 + daily + + + + https://opendatalab.github.io/MinerU/zh/usage/plugin/Cherry_Studio/ + 2026-03-26 + daily + + + + https://opendatalab.github.io/MinerU/zh/usage/plugin/Coze/ + 2026-03-26 + daily + + + + https://opendatalab.github.io/MinerU/zh/usage/plugin/DataFlow/ + 2026-03-26 + daily + + + + https://opendatalab.github.io/MinerU/zh/usage/plugin/Dify/ + 2026-03-26 + daily + + + + https://opendatalab.github.io/MinerU/zh/usage/plugin/DingTalk/ + 2026-03-26 + daily + + + + https://opendatalab.github.io/MinerU/zh/usage/plugin/FastGPT/ + 2026-03-26 + daily + + + + https://opendatalab.github.io/MinerU/zh/usage/plugin/ModelWhale/ + 2026-03-26 + daily + + + + https://opendatalab.github.io/MinerU/zh/usage/plugin/RagFlow/ + 2026-03-26 + daily + + + + https://opendatalab.github.io/MinerU/zh/usage/plugin/Sider/ + 2026-03-26 + daily + + + + https://opendatalab.github.io/MinerU/zh/usage/plugin/n8n/ + 2026-03-26 + daily + + + \ No newline at end of file diff --git a/sitemap.xml.gz b/sitemap.xml.gz new file mode 100644 index 00000000..ce5bd17f Binary files /dev/null and b/sitemap.xml.gz differ diff --git a/usage/advanced_cli_parameters/index.html b/usage/advanced_cli_parameters/index.html new file mode 100644 index 00000000..8113b49a --- /dev/null +++ b/usage/advanced_cli_parameters/index.html @@ -0,0 +1,1967 @@ + + + + + + + + + + + + + + + + + + + + + + + + + Advanced CLI Parameters - MinerU + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

+ + + + Skip to content + + +

+ +

+ + + + +

+ + +

+ + + + + + +

+ + MinerU + +

+ + + Advanced CLI Parameters + + +

+ + + + + + + + + +

+ + +

+ + English + +
+ + 中文 + +

+ + + + + + + + + +

+ +

+ + +

+ opendatalab/MinerU +

+ +

+ + + + + +

+ + + + + + Home + + +
+ + + + + + Quick Start + + +
+ + + + + + Usage + + +
+ + + + + + Reference + + +
+ + + + + + FAQ + + +
+ + + + + + Demo + + +

+ + + +

+ +

+ + + + +

+ + + + + + + +

Advanced Command Line Parameters

Pass-through of inference engine parameters

vllm Acceleration Parameter Optimization

Tip

If you can already use vllm normally for accelerated VLM model inference but still want to further improve inference speed, you can try the following parameters:

If you have multiple graphics cards, you can use vllm's multi-card parallel mode to increase throughput: --data-parallel-size 2

Parameter Passing Instructions

Tip

All officially supported vllm/lmdeploy parameters can be passed to MinerU through command line arguments, including the following commands: mineru, mineru-openai-server, mineru-gradio, mineru-api
If you want to learn more about vllm parameter usage, please refer to the vllm official documentation
If you want to learn more about lmdeploy parameter usage, please refer to the lmdeploy official documentation

GPU Device Selection and Configuration

CUDA_VISIBLE_DEVICES Basic Usage

Tip

In any situation, you can specify visible GPU devices by adding the CUDA_VISIBLE_DEVICES environment variable at the beginning of the command line. For example: +
```
CUDA_VISIBLE_DEVICES=1 mineru -p <input_path> -o <output_path>
+
```
This specification method is effective for all command line calls, including mineru, mineru-openai-server, mineru-gradio, and mineru-api, and applies to both pipeline and vlm backends.

Common Device Configuration Examples

Tip

Here are some common CUDA_VISIBLE_DEVICES setting examples: +

CUDA_VISIBLE_DEVICES=1  # Only device 1 will be seen
+CUDA_VISIBLE_DEVICES=0,1  # Devices 0 and 1 will be visible
+CUDA_VISIBLE_DEVICES="0,1"  # Same as above, quotation marks are optional
+CUDA_VISIBLE_DEVICES=0,2,3  # Devices 0, 2, 3 will be visible; device 1 is masked
+CUDA_VISIBLE_DEVICES=""  # No GPU will be visible
+

Practical Application Scenarios

Tip

Here are some possible usage scenarios:

+
If you have multiple graphics cards and need to specify cards 0 and 1, using multi-card parallelism to start openai-server, you can use the following command: +
```
CUDA_VISIBLE_DEVICES=0,1 mineru-openai-server --engine vllm --port 30000 --data-parallel-size 2
+
```
+

If you have multiple graphics cards and need to start two fastapi services on cards 0 and 1, listening on different ports respectively, you can use the following commands: +

# In terminal 1
+CUDA_VISIBLE_DEVICES=0 mineru-api --host 127.0.0.1 --port 8000
+# In terminal 2
+CUDA_VISIBLE_DEVICES=1 mineru-api --host 127.0.0.1 --port 8001
+

+ + + + + + + + + + + + + + + +

+ + + + + +

+ + + +

+ + + + + + + + + + \ No newline at end of file diff --git a/usage/cli_tools/index.html b/usage/cli_tools/index.html new file mode 100644 index 00000000..1d626d27 --- /dev/null +++ b/usage/cli_tools/index.html @@ -0,0 +1,1978 @@ + + + + + + + + + + + + + + + + + + + + + + + + + CLI Tools - MinerU + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

+ + + + Skip to content + + +

+ +

+ + + + +

+ + +

+ + + + + + +

+ + MinerU + +

+ + + CLI Tools + + +

+ + + + + + + + + +

+ + +

+ + English + +
+ + 中文 + +

+ + + + + + + + + +

+ +

+ + +

+ opendatalab/MinerU +

+ +

+ + + + + +

+ + + + + + Home + + +
+ + + + + + Quick Start + + +
+ + + + + + Usage + + +
+ + + + + + Reference + + +
+ + + + + + FAQ + + +
+ + + + + + Demo + + +

+ + + +

+ +

+ + + + +

+ + + + + + + +

Command Line Tools Usage Instructions

View Help Information

To view help information for MinerU command line tools, you can use the --help parameter. Here are help information examples for various command line tools: +

mineru --help
+Usage: mineru [OPTIONS]
+
+Options:
+  -v, --version                   Show version and exit
+  -p, --path PATH                 Input file path or directory (required)
+  -o, --output PATH               Output directory (required)
+  --api-url TEXT                  MinerU FastAPI base URL; if omitted, `mineru` starts a temporary local `mineru-api`
+  -m, --method [auto|txt|ocr]     Parsing method: auto (default), txt, ocr (pipeline and hybrid* backend only)
+  -b, --backend [pipeline|hybrid-auto-engine|hybrid-http-client|vlm-auto-engine|vlm-http-client]
+                                  Parsing backend (default: hybrid-auto-engine)
+  -l, --lang [ch|ch_server|ch_lite|en|korean|japan|chinese_cht|ta|te|ka|th|el|latin|arabic|east_slavic|cyrillic|devanagari]
+                                  Specify document language (improves OCR accuracy, pipeline and hybrid* backend only)
+  -u, --url TEXT                  OpenAI-compatible backend URL passed through to the server when using http-client
+  -s, --start INTEGER             Starting page number for parsing (0-based)
+  -e, --end INTEGER               Ending page number for parsing (0-based)
+  -f, --formula BOOLEAN           Enable formula parsing (default: enabled)
+  -t, --table BOOLEAN             Enable table parsing (default: enabled)
+  --help                          Show help information
+

mineru-api --help
+Usage: mineru-api [OPTIONS]
+
+Options:
+  --host TEXT     Server host (default: 127.0.0.1)
+  --port INTEGER  Server port (default: 8000)
+  --reload        Enable auto-reload (development mode)
+  --help          Show this message and exit.
+

mineru-gradio --help
+Usage: mineru-gradio [OPTIONS]
+
+Options:
+  --enable-example BOOLEAN        Enable example files for input. The example
+                                  files to be input need to be placed in the
+                                  `example` folder within the directory where
+                                  the command is currently executed.
+  --enable-http-client BOOLEAN    Enable http-client backend to link openai-
+                                  compatible servers.
+  --enable-api BOOLEAN            Enable gradio API for serving the
+                                  application.
+  --max-convert-pages INTEGER     Set the maximum number of pages to convert
+                                  from PDF to Markdown.
+  --server-name TEXT              Set the server name for the Gradio app.
+  --server-port INTEGER           Set the server port for the Gradio app.
+  --latex-delimiters-type [a|b|all]
+                                  Set the type of LaTeX delimiters to use in
+                                  Markdown rendering: 'a' for type '$', 'b' for
+                                  type '()[]', 'all' for both types.
+  --help                          Show this message and exit.
+

Environment Variables Description

Note

Starting from this version, mineru is an orchestration client built on top of mineru-api: +- Without --api-url, the CLI launches a temporary local mineru-api +- With --api-url, the CLI connects to that FastAPI service directly +- --url is no longer the MinerU API address; it is the OpenAI-compatible backend URL used by server-side vlm/hybrid-http-client

Some parameters of MinerU command line tools have equivalent environment variable configurations. Generally, environment variable configurations have higher priority than command line parameters and take effect across all command line tools. +Here are the environment variables and their descriptions:

+
MINERU_TOOLS_CONFIG_JSON:
+
- Used to specify configuration file path
- defaults to mineru.json in user directory, can specify other configuration file paths through environment variables.
+
+
MINERU_FORMULA_ENABLE:
+
- Used to enable formula parsing
- defaults to true, can be set to false through environment variables to disable formula parsing.
+
+
MINERU_FORMULA_CH_SUPPORT:
+
- Used to enable Chinese formula parsing optimization (experimental feature)
- Default is false, can be set to true via environment variable to enable Chinese formula parsing optimization.
- Only effective for pipeline backend.
+
+
MINERU_TABLE_ENABLE:
+
- Used to enable table parsing
- Default is true, can be set to false via environment variable to disable table parsing.
+
+
MINERU_TABLE_MERGE_ENABLE:
+
- Used to enable table merging functionality
- Default is true, can be set to false via environment variable to disable table merging functionality.
+
+
MINERU_PDF_RENDER_TIMEOUT:
+
- Used to set the timeout (in seconds) for rendering PDFs to images.
- Default is 300 seconds; you can set a different value via an environment variable to adjust the rendering timeout.
- Only effective on Linux and macOS systems.
+
+
MINERU_PDF_RENDER_THREADS:
+
- Used to set the number of threads used when rendering PDFs to images.
- Default is 4; you can set a different value via an environment variable to adjust the number of threads for image rendering.
- Only effective on Linux and macOS systems.
+
+
MINERU_INTRA_OP_NUM_THREADS:
+
- Used to set the intra_op thread count for ONNX models, affects the computation speed of individual operators
- Default is -1 (auto-select), can be set to other values via environment variable to adjust the thread count.
+
+
MINERU_INTER_OP_NUM_THREADS:
+
- Used to set the inter_op thread count for ONNX models, affects the parallel execution of multiple operators
- Default is -1 (auto-select), can be set to other values via environment variable to adjust the thread count.
+
+
MINERU_HYBRID_BATCH_RATIO:
+
- Used to set the batch ratio for small model processing in hybrid-* backends.
- Commonly used in hybrid-http-client, it allows adjusting the VRAM usage of a single client by controlling the batch ratio of small models.
- + + + + + + + + + + + + + + + + + + + + + + + + + +
  Single Client VRAM Size MINERU_HYBRID_BATCH_RATIO
  <= 6 GB 8
  <= 4.5 GB 4
  <= 3 GB 2
  <= 2.5 GB 1
  +
+
+
MINERU_HYBRID_FORCE_PIPELINE_ENABLE:
+
- Used to force the text extraction part in hybrid-* backends to be processed using small models.
- Defaults to false. Can be set to true via environment variable to enable this feature, thereby reducing hallucinations in certain extreme cases.
+
+
MINERU_VL_MODEL_NAME:
+
- Used to specify the model name for the vlm/hybrid backend, allowing you to designate the model required for MinerU to run when multiple models exist on a remote openai-server.
+
+
MINERU_VL_API_KEY:
+
- Used to specify the API Key for the vlm/hybrid backend, enabling authentication on the remote openai-server.
+

+ + + + + + + + + + + + + + + +

Single Client VRAM Size	MINERU_HYBRID_BATCH_RATIO
<= 6 GB	8
<= 4.5 GB	4
<= 3 GB	2
<= 2.5 GB	1

+ + + + + +

+ + + +

+ + + + + + + + + + \ No newline at end of file diff --git a/usage/index.html b/usage/index.html new file mode 100644 index 00000000..6b27b733 --- /dev/null +++ b/usage/index.html @@ -0,0 +1,1690 @@ + + + + + + + + + + + + + + + + + + + + + + + + + Usage - MinerU + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

+ + + + Skip to content + + +

+ +

+ + + + +

+ + +

+ + + + + + +

+ + MinerU + +

+ + + Usage + + +

+ + + + + + + + + +

+ + +

+ + English + +
+ + 中文 + +

+ + + + + + + + + +

+ +

+ + +

+ opendatalab/MinerU +

+ +

+ + +

+ + + + + + + + + + \ No newline at end of file diff --git a/usage/model_source/index.html b/usage/model_source/index.html new file mode 100644 index 00000000..e88d5dfe --- /dev/null +++ b/usage/model_source/index.html @@ -0,0 +1,1935 @@ + + + + + + + + + + + + + + + + + + + + + + + + + Model Source - MinerU + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

+ + + + Skip to content + + +

+ +

+ + + + +

+ + +

+ + + + + + +

+ + MinerU + +

+ + + Model Source + + +

+ + + + + + + + + +

+ + +

+ + English + +
+ + 中文 + +

+ + + + + + + + + +

+ +

+ + +

+ opendatalab/MinerU +

+ +

+ + + + + +

+ + + + + + Home + + +
+ + + + + + Quick Start + + +
+ + + + + + Usage + + +
+ + + + + + Reference + + +
+ + + + + + FAQ + + +
+ + + + + + Demo + + +

+ + + +

+ +

+ + + + +

+ + + + + + + +

Model Source Documentation

MinerU uses HuggingFace and ModelScope as model repositories. Users can switch model sources or use local models as needed.

HuggingFace is the default model source, providing excellent loading speed and high stability globally.
ModelScope is the best choice for users in mainland China, providing seamlessly compatible hf SDK modules, suitable for users who cannot access HuggingFace.

Methods to Switch Model Sources

Switch via Command Line Parameters

mineru -p <input_path> -o <output_path> --source modelscope
+

Switch via Environment Variables

You can switch model sources by setting environment variables in any situation. This applies to all command line tools and API calls. +

export MINERU_MODEL_SOURCE=modelscope
+

+or +

import os
+os.environ["MINERU_MODEL_SOURCE"] = "modelscope"
+

Tip

Using Local Models

1. Download Models to Local Storage

mineru-models-download --help
+

+or use the interactive command line tool to select model downloads: +

mineru-models-download
+

Note

After download completion, the model path will be output in the current terminal window and automatically written to mineru.json in the user directory.
You can also create it by copying the configuration template file to your user directory and renaming it to mineru.json.
After downloading models locally, you can freely move the model folder to other locations while updating the model path in mineru.json.
If you deploy the model folder to another server, please ensure you move the mineru.json file to the user directory of the new device and configure the model path correctly.
If you need to update model files, you can run the mineru-models-download command again. Model updates do not support custom paths currently - if you haven't moved the local model folder, model files will be incrementally updated; if you have moved the model folder, model files will be re-downloaded to the default location and mineru.json will be updated.

2. Use Local Models for Parsing

mineru -p <input_path> -o <output_path> --source local
+

+or enable through environment variables: +

export MINERU_MODEL_SOURCE=local
+mineru -p <input_path> -o <output_path>
+

+ + + + + + + + + + + + + + + +

+ + + + + +

+ + + +

+ + + + + + + + + + \ No newline at end of file diff --git a/usage/quick_usage/index.html b/usage/quick_usage/index.html new file mode 100644 index 00000000..de84d1a1 --- /dev/null +++ b/usage/quick_usage/index.html @@ -0,0 +1,1974 @@ + + + + + + + + + + + + + + + + + + + + + + + + + Quick Usage - MinerU + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

+ + + + Skip to content + + +

+ +

+ + + + +

+ + +

+ + + + + + +

+ + MinerU + +

+ + + Quick Usage + + +

+ + + + + + + + + +

+ + +

+ + English + +
+ + 中文 + +

+ + + + + + + + + +

+ +

+ + +

+ opendatalab/MinerU +

+ +

+ + + + + +

+ + + + + + Home + + +
+ + + + + + Quick Start + + +
+ + + + + + Usage + + +
+ + + + + + Reference + + +
+ + + + + + FAQ + + +
+ + + + + + Demo + + +

+ + + +

+ +

+ + + + +

+ + + + + + + +

Using MinerU

Quick Model Source Configuration

export MINERU_MODEL_SOURCE=modelscope
+

+For more information about model source configuration and custom local model paths, please refer to the Model Source Documentation in the documentation. +

Quick Usage via Command Line

MinerU has built-in command line tools that allow users to quickly use MinerU for PDF parsing through the command line: +

mineru -p <input_path> -o <output_path>
+

Tip

<input_path>: Local PDF/image file or directory
<output_path>: Output directory
Without --api-url, the CLI launches a temporary local mineru-api
With --api-url, the CLI connects to an existing local or remote FastAPI service directly

For more information about output files, please refer to Output File Documentation.

Note

The command line tool will automatically attempt cuda/mps acceleration on Linux and macOS systems. +Windows users who need cuda acceleration should visit the PyTorch official website to select the appropriate command for their cuda version to install acceleration-enabled torch and torchvision.

If you need to adjust parsing options through custom parameters, you can also check the more detailed Command Line Tools Usage Instructions in the documentation.

Advanced Usage via API, WebUI, http-client/server

Direct Python API calls: Python Usage Example
FastAPI calls: +
```
mineru-api --host 0.0.0.0 --port 8000
+
```
+
Tip
+
Access http://127.0.0.1:8000/docs in your browser to view the API documentation.
+
- Health endpoint: GET /health + Returns protocol_version, processing_window_size, max_concurrent_requests, and task stats
- Asynchronous task submission endpoint: POST /tasks
- Synchronous parsing endpoint: POST /file_parse
- Task query endpoints: GET /tasks/{task_id}, GET /tasks/{task_id}/result
- API outputs are controlled by the server and written to ./output by default
+
POST /tasks returns immediately with a task_id. POST /file_parse uses the same task manager internally, waits for the task to finish, and then returns the final result synchronously. +Tasks are tracked only in-process for a single mineru-api instance. Task status is not preserved across service restarts, --reload, or multi-process deployments. +Completed or failed tasks are retained for 24 hours by default, then their task state and output directory are cleaned automatically. After cleanup, task status and result endpoints return 404. +Use MINERU_API_TASK_RETENTION_SECONDS and MINERU_API_TASK_CLEANUP_INTERVAL_SECONDS to adjust retention and cleanup polling intervals.
+
Asynchronous task submission example: +
```
curl -X POST http://127.0.0.1:8000/tasks \
+  -F "files=@demo/pdfs/demo1.pdf" \
+  -F "return_md=true"
+
```
+
Synchronous parsing example: +
```
curl -X POST http://127.0.0.1:8000/file_parse \
+  -F "files=@demo/pdfs/demo1.pdf" \
+  -F "return_md=true" \
+  -F "response_format_zip=true" \
+  -F "return_original_file=true"
+
```
+
Poll task status and fetch results: +
```
curl http://127.0.0.1:8000/tasks/<task_id>
+curl http://127.0.0.1:8000/tasks/<task_id>/result
+curl http://127.0.0.1:8000/health
+
```
+
+
+
Start Gradio WebUI visual frontend: +
```
mineru-gradio --server-name 0.0.0.0 --server-port 7860
+
```
+
+
Tip
+
- Access http://127.0.0.1:7860 in your browser to use the Gradio WebUI.
+
+

Using http-client/server method: +

# Start openai compatible server (requires vllm or lmdeploy environment)
+mineru-openai-server --port 30000
+

Tip

In another terminal, connect to openai server via http client +

mineru -p <input_path> -o <output_path> -b hybrid-http-client -u http://127.0.0.1:30000
+

Note

All officially supported vllm/lmdeploy parameters can be passed to MinerU through command line arguments, including the following commands: mineru, mineru-openai-server, mineru-gradio, mineru-api. +We have compiled some commonly used parameters and usage methods for vllm/lmdeploy, which can be found in the documentation Advanced Command Line Parameters.

Extending MinerU Functionality with Configuration Files

MinerU is now ready to use out of the box, but also supports extending functionality through configuration files. You can edit mineru.json file in your user directory to add custom configurations.

Important

Here are some available configuration options:

+
latex-delimiter-config:
+
- Used to configure LaTeX formula delimiters
- Defaults to $ symbol, can be modified to other symbols or strings as needed.
+
+
llm-aided-config:
+
- Used to configure parameters for LLM-assisted title hierarchy
- Compatible with all LLM models supporting openai protocol, defaults to using Alibaba Cloud Bailian's qwen3-next-80b-a3b-instruct model.
- You need to configure your own API key and set enable to true to enable this feature.
- If your API provider does not support the enable_thinking parameter, please manually remove it.
  - For example, in your configuration file, the llm-aided-config section may look like: +
```
"llm-aided-config": {
+   "api_key": "your_api_key",
+   "base_url": "https://dashscope.aliyuncs.com/compatible-mode/v1",
+   "model": "qwen3-next-80b-a3b-instruct",
+   "enable_thinking": false,
+   "enable": false
+}
+
```
  - To remove the enable_thinking parameter, simply delete the line containing "enable_thinking": false, resulting in: +
```
"llm-aided-config": {
+   "api_key": "your_api_key",
+   "base_url": "https://dashscope.aliyuncs.com/compatible-mode/v1",
+   "model": "qwen3-next-80b-a3b-instruct",
+   "enable": false
+}
+
```
  +
+
+
models-dir:
+
- Used to specify local model storage directory
- Please specify model directories for pipeline and vlm backends separately.
- After specifying the directory, you can use local models by configuring the environment variable export MINERU_MODEL_SOURCE=local.
+

+ + + + + + + + + + + + + + + +

+ + + + + +

+ + + +

+ + + + + + + + + + \ No newline at end of file diff --git a/zh/demo/index.html b/zh/demo/index.html new file mode 100644 index 00000000..367b65b8 --- /dev/null +++ b/zh/demo/index.html @@ -0,0 +1,1658 @@ + + + + + + + + + + + + + + + + + + + + + + + 在线演示 - MinerU + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

+ +

+ + + + +

+ + +

+ + + + + + +

+ + MinerU + +

+ + + 在线演示 + + +

+ + + + + + + + + +

+ + +

+ + English + +
+ + 中文 + +

+ + + + + + + + + +

+ +

+ + +

+ opendatalab/MinerU +

+ +

+ + +

+ + + + + + + + + + \ No newline at end of file diff --git a/zh/faq/index.html b/zh/faq/index.html new file mode 100644 index 00000000..868ac295 --- /dev/null +++ b/zh/faq/index.html @@ -0,0 +1,1700 @@ + + + + + + + + + + + + + + + + + + + + + + + + + 常见问题解答 - MinerU + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

+ + + + 跳转至 + + +

+ +

+ + + + +

+ + +

+ + + + + + +

+ + MinerU + +

+ + + 常见问题解答 + + +

+ + + + + + + + + +

+ + +

+ + English + +
+ + 中文 + +

+ + + + + + + + + +

+ +

+ + +

+ opendatalab/MinerU +

+ +

+ + + + + +

+ + + + + + 主页 + + +
+ + + + + + 快速入门 + + +
+ + + + + + 使用指南 + + +
+ + + + + + 参考资料 + + +
+ + + + + + 常见问题解答 + + +
+ + + + + + 在线演示 + + +

+ + + +

+ +

+ + + + +

+ + + + + + + +

常见问题解答

如果未能列出您的问题，您也可以使用DeepWiki与AI助手交流，这可以解决大部分常见问题。

如果您仍然无法解决问题，您可通过Discord或WeChat加入社区，与其他用户和开发者交流。

在WSL2的Ubuntu22.04中遇到报错ImportError: libGL.so.1: cannot open shared object file: No such file or directory

WSL2的Ubuntu22.04中缺少libgl库，可通过以下命令安装libgl库解决：

sudo apt-get install libgl1-mesa-glx
+

参考：#388

在 Linux 系统安装并使用时，解析结果缺失部份文字信息。

MinerU在>=2.0的版本中使用pypdfium2代替pymupdf作为PDF页面的渲染引擎，以解决AGPLv3的许可证问题，在某些Linux发行版，由于缺少CJK字体，可能会在将PDF渲染成图片的过程中丢失部份文字。 +为了解决这个问题，您可以通过以下命令安装noto字体包，这在Ubuntu/debian系统中有效： +

sudo apt update
+sudo apt install fonts-noto-core
+sudo apt install fonts-noto-cjk
+fc-cache -fv
+

+也可以直接使用我们的Docker部署方式构建镜像，镜像中默认包含以上字体包。 +

参考：#2915

+ + + + + + + + + + + + + + + +

+ + + + + +

+ + + +

+ + + + + + + + + + \ No newline at end of file diff --git a/zh/index.html b/zh/index.html new file mode 100644 index 00000000..b2651521 --- /dev/null +++ b/zh/index.html @@ -0,0 +1,1741 @@ + + + + + + + + + + + + + + + + + + + + + + + MinerU - MinerU + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

+ + + + 跳转至 + + +

+ +

+ + + + +

+ + +

+ + + + + + +

+ + MinerU + +

+ + + MinerU + + +

+ + + + + + + + + +

+ + +

+ + English + +
+ + 中文 + +

+ + + + + + + + + +

+ +

+ + +

+ opendatalab/MinerU +

+ +

+ + +

+ + + + + + + + + + \ No newline at end of file diff --git a/zh/quick_start/docker_deployment/index.html b/zh/quick_start/docker_deployment/index.html new file mode 100644 index 00000000..3cdf1cf3 --- /dev/null +++ b/zh/quick_start/docker_deployment/index.html @@ -0,0 +1,1963 @@ + + + + + + + + + + + + + + + + + + + + + + + + + Docker部署 - MinerU + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

+ + + + 跳转至 + + +

+ +

+ + + + +

+ + +

+ + + + + + +

+ + MinerU + +

+ + + Docker部署 + + +

+ + + + + + + + + +

+ + +

+ + English + +
+ + 中文 + +

+ + + + + + + + + +

+ +

+ + +

+ opendatalab/MinerU +

+ +

+ + + + + +

+ + + + + + 主页 + + +
+ + + + + + 快速入门 + + +
+ + + + + + 使用指南 + + +
+ + + + + + 参考资料 + + +
+ + + + + + 常见问题解答 + + +
+ + + + + + 在线演示 + + +

+ + + +

+ +

+ + + + +

+ + + + + + + +

使用docker部署Mineru

MinerU提供了便捷的docker部署方式，这有助于快速搭建环境并解决一些棘手的环境兼容问题。

使用 Dockerfile 构建镜像

wget https://gcore.jsdelivr.net/gh/opendatalab/MinerU@master/docker/china/Dockerfile
+docker build -t mineru:latest -f Dockerfile .
+

Docker说明

Mineru的docker使用了vllm/vllm-openai作为基础镜像，因此在docker中默认集成了vllm推理加速框架和必需的依赖环境。因此在满足条件的设备上，您可以直接使用vllm加速VLM模型推理。

Note

使用vllm加速VLM模型推理需要满足的条件是：

设备包含Volta及以后架构的显卡，且可用显存大于等于8G。
物理机的显卡驱动应支持CUDA 12.9.1或更高版本，可通过nvidia-smi命令检查驱动版本。
docker中能够访问物理机的显卡设备。

启动 Docker 容器

docker run --gpus all \
+  --shm-size 32g \
+  -p 30000:30000 -p 7860:7860 -p 8000:8000 \
+  --ipc=host \
+  -it mineru:latest \
+  /bin/bash
+

执行该命令后，您将进入到Docker容器的交互式终端，并映射了一些端口用于可能会使用的服务，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。

通过 Docker Compose 直接启动服务

我们提供了compose.yml文件，您可以通过它来快速启动MinerU服务。

# 下载 compose.yaml 文件
+wget https://gcore.jsdelivr.net/gh/opendatalab/MinerU@master/docker/compose.yaml
+

Note

compose.yaml文件中包含了MinerU的多个服务配置，您可以根据需要选择启动特定的服务。
不同的服务可能会有额外的参数配置，您可以在compose.yaml文件中查看并编辑。
由于vllm推理加速框架预分配显存的特性，您可能无法在同一台机器上同时运行多个vllm服务，因此请确保在启动vlm-openai-server服务或使用vlm-vllm-engine后端时，其他可能使用显存的服务已停止。

启动 openai兼容接口服务

并通过vlm-http-client后端连接openai-server +

docker compose -f compose.yaml --profile openai-server up -d
+

Tip

在另一个终端中通过http client连接openai server（只需cpu与网络，不需要vllm环境） +

mineru -p <input_path> -o <output_path> -b vlm-http-client -u http://<server_ip>:30000
+

启动 Web API 服务

docker compose -f compose.yaml --profile api up -d
+

Tip

在浏览器中访问 http://<server_ip>:8000/docs 查看API文档。

启动 Gradio WebUI 服务

docker compose -f compose.yaml --profile gradio up -d
+

Tip

在浏览器中访问 http://<server_ip>:7860 使用 Gradio WebUI。

+ + + + + + + + + + + + + + + +

+ + + + + +

+ + + +

+ + + + + + + + + + \ No newline at end of file diff --git a/zh/quick_start/extension_modules/index.html b/zh/quick_start/extension_modules/index.html new file mode 100644 index 00000000..2b470aac --- /dev/null +++ b/zh/quick_start/extension_modules/index.html @@ -0,0 +1,1921 @@ + + + + + + + + + + + + + + + + + + + + + + + + + 扩展模块安装 - MinerU + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

+ + + + 跳转至 + + +

+ +

+ + + + +

+ + +

+ + + + + + +

+ + MinerU + +

+ + + 扩展模块安装 + + +

+ + + + + + + + + +

+ + +

+ + English + +
+ + 中文 + +

+ + + + + + + + + +

+ +

+ + +

+ opendatalab/MinerU +

+ +

+ + + + + +

+ + + + + + 主页 + + +
+ + + + + + 快速入门 + + +
+ + + + + + 使用指南 + + +
+ + + + + + 参考资料 + + +
+ + + + + + 常见问题解答 + + +
+ + + + + + 在线演示 + + +

+ + + +

+ +

+ + + + +

+ + + + + + + +

MinerU 扩展模块安装指南

MinerU 支持根据不同需求，按需安装扩展模块，以增强功能或支持特定的模型后端。

常见场景

核心功能安装

core 模块是 MinerU 的核心依赖，包含了除vllm/lmdeploy外的所有功能模块。安装此模块可以确保 MinerU 的基本功能正常运行。 +

uv pip install "mineru[core]"
+

使用`vllm`加速 VLM 模型推理

Note

vllm和lmdeploy对vlm的推理加速效果和使用方式几乎相同，您可以根据实际情况选择其中之一进行安装和使用，但不建议同时安装这两个模块，以避免潜在的依赖冲突。

vllm 模块提供了对 VLM 模型推理的加速支持，适用于具有 Volta 及以后架构的显卡（8G 显存及以上）。安装此模块可以显著提升模型推理速度。 +

uv pip install "mineru[core,vllm]"
+

Tip

如在安装包含vllm的扩展包过程中发生异常，请参考 vllm 官方文档尝试解决，或直接使用 Docker 方式部署镜像。

使用`lmdeploy`加速 VLM 模型推理

Note

lmdeploy 模块提供了对 VLM 模型推理的加速支持，适用于具有 Volta 及以后架构的显卡（8G 显存及以上）。安装此模块可以显著提升模型推理速度。 +

uv pip install "mineru[core,lmdeploy]"
+

Tip

如在安装包含lmdeploy的扩展包过程中发生异常，请参考 lmdeploy 官方文档尝试解决。

安装轻量版client连接兼容openai服务器使用 (适用vlm-http-client模式)

如果您需要在边缘设备上安装轻量版的 client 端以连接兼容 openai 接口的服务端来使用vlm模式，可以安装mineru的基础包，非常轻量，适合在只有cpu和网络连接的设备上使用。 +

uv pip install mineru
+mineru -p <input_path> -o <output_path> -b vlm-http-client -u http://127.0.0.1:30000
+

安装轻量版client连接兼容openai服务器使用 (适用hybrid-http-client模式)

如果您需要在边缘设备上安装轻量版的 client 端以连接兼容 openai 接口的服务端来使用hybrid模式，可以安装mineru的pipeline扩展包，相对较轻量，可以在只有cpu和网络连接的设备上使用，同时在支持gpu加速的设备上可以更快运行。 +

uv pip install "mineru[pipeline]"
+mineru -p <input_path> -o <output_path> -b hybrid-http-client -u http://127.0.0.1:30000
+

+ + + + + + + + + + + + + + + +

+ + + + + +

+ + + +

+ + + + + + + + + + \ No newline at end of file diff --git a/zh/quick_start/index.html b/zh/quick_start/index.html new file mode 100644 index 00000000..09f507d5 --- /dev/null +++ b/zh/quick_start/index.html @@ -0,0 +1,1811 @@ + + + + + + + + + + + + + + + + + + + + + + + + + 快速入门 - MinerU + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

+ + + + 跳转至 + + +

+ +

+ + + + +

+ + +

+ + + + + + +

+ + MinerU + +

+ + + 快速入门 + + +

+ + + + + + + + + +

+ + +

+ + English + +
+ + 中文 + +

+ + + + + + + + + +

+ +

+ + +

+ opendatalab/MinerU +

+ +

+ + + + + +

+ + + + + + 主页 + + +
+ + + + + + 快速入门 + + +
+ + + + + + 使用指南 + + +
+ + + + + + 参考资料 + + +
+ + + + + + 常见问题解答 + + +
+ + + + + + 在线演示 + + +

+ + + +

+ +

+ + + + +

+ + + + + + + +

快速入门

如果遇到任何安装问题，请先查询 FAQ

在线体验

官网在线应用

官网在线版功能与客户端一致，界面美观，功能丰富，需要登录使用

基于Gradio的在线demo

基于gradio开发的webui，界面简洁，仅包含核心解析功能，免登录

本地部署

Warning

安装前必看——软硬件环境支持说明

为了确保项目的稳定性和可靠性，我们在开发过程中仅对特定的软硬件环境进行优化和测试。这样当用户在推荐的系统配置上部署和运行项目时，能够获得最佳的性能表现和最少的兼容性问题。

通过集中资源和精力于主线环境，我们团队能够更高效地解决潜在的BUG，及时开发新功能。

在非主线环境中，由于硬件、软件配置的多样性，以及第三方依赖项的兼容性问题，我们无法100%保证项目的完全可用性。因此，对于希望在非推荐环境中使用本项目的用户，我们建议先仔细阅读文档以及FAQ，大多数问题已经在FAQ中有对应的解决方案，除此之外我们鼓励社区反馈问题，以便我们能够逐步扩大支持范围。

+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

解析后端	pipeline	*-auto-engine		*-http-client
解析后端	pipeline	hybrid	vlm	hybrid	vlm
后端特性	兼容性好	硬件配置要求较高		适用于OpenAI兼容服务器²
精度指标¹	82+	90+
操作系统	Linux³ / Windows⁴ / macOS⁵
纯CPU平台支持	✅	❌		✅
GPU加速支持	Volta及以后架构GPU或Apple Silicon				不需要
显存最低要求	6GB	10GB	8GB	3GB	不需要
内存要求	最低16GB以上,推荐32GB以上			8GB
磁盘空间要求	20GB以上,推荐使用SSD			2GB
python版本	3.10-3.13

+ +

¹ 精度指标为OmniDocBench (v1.5)的End-to-End Evaluation Overall分数，基于MinerU最新版本测试
+² 兼容OpenAI API的服务器，如通过vLLM/SGLang/LMDeploy等推理框架部署的本地模型服务器或远程模型服务
+³ Linux仅支持2019年及以后发行版
+⁴ 由于关键依赖ray未能在windows平台支持Python 3.13，故仅支持至3.10~3.12版本
+⁵ macOS 需使用14.0以上版本

Tip

除以上主流环境与平台外，我们也收录了一些社区用户反馈的其他平台支持情况，详情请参考其他加速卡适配。
+如果您有意将自己的环境适配经验分享给社区，欢迎通过show-and-tell提交或提交PR至其他加速卡适配文档。

安装 MinerU

使用pip或uv安装MinerU

pip install --upgrade pip -i https://mirrors.aliyun.com/pypi/simple
+pip install uv -i https://mirrors.aliyun.com/pypi/simple
+uv pip install -U "mineru[all]" -i https://mirrors.aliyun.com/pypi/simple 
+

通过源码安装MinerU

git clone https://github.com/opendatalab/MinerU.git
+cd MinerU
+uv pip install -e .[all] -i https://mirrors.aliyun.com/pypi/simple
+

Tip

mineru[all]包含所有核心功能，兼容Windows / Linux / macOS系统，适合绝大多数用户。 +如果您需要指定vlm模型的推理框架，或是仅准备在边缘设备安装轻量版client端，可以参考文档扩展模块安装指南。

使用docker部署Mineru

MinerU提供了便捷的docker部署方式，这有助于快速搭建环境并解决一些棘手的环境兼容问题。 +您可以在文档中获取Docker部署说明。

使用 MinerU

Tip

默认使用托管在huggingface的模型进行解析，首次使用时会自动下载所需模型文件，后续使用将直接加载本地缓存的模型。如果您无法访问huggingface，可以通过以下命令切换至国内镜像源: +

export MINERU_MODEL_SOURCE=modelscope
+

如果您的设备满足上表中GPU加速的条件，可以使用简单的命令行进行文档解析: +

mineru -p <input_path> -o <output_path>
+

+如果您的设备不满足GPU加速条件，可以指定后端为pipeline，以在纯CPU环境下运行: +

mineru -p <input_path> -o <output_path> -b pipeline
+

您可以通过命令行、API、WebUI等多种方式使用MinerU进行PDF解析，具体使用方法请参考使用指南。

+ + + + + + + + + + + + + + + +

+ + + + + +

+ + + +

+ + + + + + + + + + \ No newline at end of file diff --git a/zh/reference/changelog/index.html b/zh/reference/changelog/index.html new file mode 100644 index 00000000..29ed67d7 --- /dev/null +++ b/zh/reference/changelog/index.html @@ -0,0 +1,3236 @@ + + + + + + + + + + + + + + + + + + + + + + + + + 更新日志 - MinerU + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

+ + + + 跳转至 + + +

+ +

+ + + + +

+ + +

+ + + + + + +

+ + MinerU + +

+ + + 更新日志 + + +

+ + + + + + + + + +

+ + +

+ + English + +
+ + 中文 + +

+ + + + + + + + + +

+ +

+ + +

+ opendatalab/MinerU +

+ +

+ + +

+ + + + + + + + + + \ No newline at end of file diff --git a/zh/reference/index.html b/zh/reference/index.html new file mode 100644 index 00000000..16959e37 --- /dev/null +++ b/zh/reference/index.html @@ -0,0 +1,1701 @@ + + + + + + + + + + + + + + + + + + + + + + + + + 参考资料 - MinerU + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

+ + + + 跳转至 + + +

+ +

+ + + + +

+ + +

+ + + + + + +

+ + MinerU + +

+ + + 参考资料 + + +

+ + + + + + + + + +

+ + +

+ + English + +
+ + 中文 + +

+ + + + + + + + + +

+ +

+ + +

+ opendatalab/MinerU +

+ +

+ + +

+ + + + + + + + + + \ No newline at end of file diff --git a/zh/reference/output_files/index.html b/zh/reference/output_files/index.html new file mode 100644 index 00000000..b8b8bd97 --- /dev/null +++ b/zh/reference/output_files/index.html @@ -0,0 +1,3497 @@ + + + + + + + + + + + + + + + + + + + + + + + + + 输出文件格式 - MinerU + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

+ + + + 跳转至 + + +

+ +

+ + + + +

+ + +

+ + + + + + +

+ + MinerU + +

+ + + 输出文件格式 + + +

+ + + + + + + + + +

+ + +

+ + English + +
+ + 中文 + +

+ + + + + + + + + +

+ +

+ + +

+ opendatalab/MinerU +

+ +

+ + + + + +

+ + + + + + 主页 + + +
+ + + + + + 快速入门 + + +
+ + + + + + 使用指南 + + +
+ + + + + + 参考资料 + + +
+ + + + + + 常见问题解答 + + +
+ + + + + + 在线演示 + + +

+ + + +

+ +

+ + + + +

+ + + + + + + +

MinerU 输出文件说明

概览

mineru 命令执行后，除了输出主要的 markdown 文件外，还会生成多个辅助文件用于调试、质检和进一步处理。这些文件包括：

可视化调试文件：帮助用户直观了解文档解析过程和结果
结构化数据文件：包含详细的解析数据，可用于二次开发

下面将详细介绍每个文件的作用和格式。

可视化调试文件

布局分析文件 (layout.pdf)

文件命名格式：{原文件名}_layout.pdf

功能说明：

可视化展示每一页的布局分析结果
每个检测框右上角的数字表示阅读顺序
使用不同背景色块区分不同类型的内容块

使用场景：

检查布局分析是否正确
确认阅读顺序是否合理
调试布局相关问题

layout 页面示例

文本片段文件 (span.pdf)

Note

仅适用于 pipeline 后端

文件命名格式：{原文件名}_span.pdf

功能说明：

根据 span 类型使用不同颜色线框标注页面内容
用于质量检查和问题排查

使用场景：

快速排查文本丢失问题
检查行内公式识别情况
验证文本分割准确性

span 页面示例

结构化数据文件

Important

2.5版本vlm后端的输出存在较大变化，与pipeline版本存在不兼容情况，如需基于结构化输出进行二次开发，请仔细阅读本文档内容。

pipeline 后端输出结果

模型推理结果 (model.json)

文件命名格式：{原文件名}_model.json

数据结构定义

from pydantic import BaseModel, Field
+from enum import IntEnum
+
+class CategoryType(IntEnum):
+    """内容类别枚举"""
+    title = 0               # 标题
+    plain_text = 1          # 文本
+    abandon = 2             # 包括页眉页脚页码和页面注释
+    figure = 3              # 图片
+    figure_caption = 4      # 图片描述
+    table = 5               # 表格
+    table_caption = 6       # 表格描述
+    table_footnote = 7      # 表格注释
+    isolate_formula = 8     # 行间公式
+    formula_caption = 9     # 行间公式的标号
+    embedding = 13          # 行内公式
+    isolated = 14           # 行间公式
+    text = 15               # OCR 识别结果
+
+class PageInfo(BaseModel):
+    """页面信息"""
+    page_no: int = Field(description="页码序号，第一页的序号是 0", ge=0)
+    height: int = Field(description="页面高度", gt=0)
+    width: int = Field(description="页面宽度", ge=0)
+
+class ObjectInferenceResult(BaseModel):
+    """对象识别结果"""
+    category_id: CategoryType = Field(description="类别", ge=0)
+    poly: list[float] = Field(description="四边形坐标，格式为 [x0,y0,x1,y1,x2,y2,x3,y3]")
+    score: float = Field(description="推理结果的置信度")
+    latex: str | None = Field(description="LaTeX 解析结果", default=None)
+    html: str | None = Field(description="HTML 解析结果", default=None)
+
+class PageInferenceResults(BaseModel):
+    """页面推理结果"""
+    layout_dets: list[ObjectInferenceResult] = Field(description="页面识别结果")
+    page_info: PageInfo = Field(description="页面元信息")
+
+# 完整的推理结果
+inference_result: list[PageInferenceResults] = []
+

坐标系统说明

poly 坐标格式：[x0, y0, x1, y1, x2, y2, x3, y3]

分别表示左上、右上、右下、左下四点的坐标
坐标原点在页面左上角

poly 坐标示意图

示例数据

[
+    {
+        "layout_dets": [
+            {
+                "category_id": 2,
+                "poly": [
+                    99.1906967163086,
+                    100.3119125366211,
+                    730.3707885742188,
+                    100.3119125366211,
+                    730.3707885742188,
+                    245.81326293945312,
+                    99.1906967163086,
+                    245.81326293945312
+                ],
+                "score": 0.9999997615814209
+            }
+        ],
+        "page_info": {
+            "page_no": 0,
+            "height": 2339,
+            "width": 1654
+        }
+    },
+    {
+        "layout_dets": [
+            {
+                "category_id": 5,
+                "poly": [
+                    99.13092803955078,
+                    2210.680419921875,
+                    497.3183898925781,
+                    2210.680419921875,
+                    497.3183898925781,
+                    2264.78076171875,
+                    99.13092803955078,
+                    2264.78076171875
+                ],
+                "score": 0.9999997019767761
+            }
+        ],
+        "page_info": {
+            "page_no": 1,
+            "height": 2339,
+            "width": 1654
+        }
+    }
+]
+

中间处理结果 (middle.json)

文件命名格式：{原文件名}_middle.json

顶层结构

+ + + + + + + + + + + + + + + + + + + + + + + + + +

字段名	类型	说明
`pdf_info`	`list[dict]`	每一页的解析结果数组
`_backend`	`string`	解析模式：`pipeline` 或 `vlm`
`_version_name`	`string`	MinerU 版本号

页面信息结构 (pdf_info)

+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

字段名	说明
`preproc_blocks`	PDF 预处理后的未分段中间结果
`page_idx`	页码，从 0 开始
`page_size`	页面的宽度和高度 `[width, height]`
`images`	图片块信息列表
`tables`	表格块信息列表
`interline_equations`	行间公式块信息列表
`discarded_blocks`	需要丢弃的块信息
`para_blocks`	分段后的内容块结果

块结构层次

一级块 (table | image)
+└── 二级块
+    └── 行 (line)
+        └── 片段 (span)
+

一级块字段

+ + + + + + + + + + + + + + + + + + + + + +

字段名	说明
`type`	块类型：`table` 或 `image`
`bbox`	块的矩形框坐标 `[x0, y0, x1, y1]`
`blocks`	包含的二级块列表

二级块字段

+ + + + + + + + + + + + + + + + + + + + + +

字段名	说明
`type`	块类型（详见下表）
`bbox`	块的矩形框坐标
`lines`	包含的行信息列表

二级块类型

+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

类型	说明
`image_body`	图像本体
`image_caption`	图像描述文本
`image_footnote`	图像脚注
`table_body`	表格本体
`table_caption`	表格描述文本
`table_footnote`	表格脚注
`text`	文本块
`title`	标题块
`index`	目录块
`list`	列表块
`interline_equation`	行间公式块

行和片段结构

行 (line) 字段： +- bbox：行的矩形框坐标 +- spans：包含的片段列表

片段 (span) 字段： +- bbox：片段的矩形框坐标 +- type：片段类型（image、table、text、inline_equation、interline_equation） +- content | img_path：文本内容或图片路径

示例数据

{
+    "pdf_info": [
+        {
+            "preproc_blocks": [
+                {
+                    "type": "text",
+                    "bbox": [
+                        52,
+                        61.956024169921875,
+                        294,
+                        82.99800872802734
+                    ],
+                    "lines": [
+                        {
+                            "bbox": [
+                                52,
+                                61.956024169921875,
+                                294,
+                                72.0000228881836
+                            ],
+                            "spans": [
+                                {
+                                    "bbox": [
+                                        54.0,
+                                        61.956024169921875,
+                                        296.2261657714844,
+                                        72.0000228881836
+                                    ],
+                                    "content": "dependent on the service headway and the reliability of the departure ",
+                                    "type": "text",
+                                    "score": 1.0
+                                }
+                            ]
+                        }
+                    ]
+                }
+            ],
+            "layout_bboxes": [
+                {
+                    "layout_bbox": [
+                        52,
+                        61,
+                        294,
+                        731
+                    ],
+                    "layout_label": "V",
+                    "sub_layout": []
+                }
+            ],
+            "page_idx": 0,
+            "page_size": [
+                612.0,
+                792.0
+            ],
+            "_layout_tree": [],
+            "images": [],
+            "tables": [],
+            "interline_equations": [],
+            "discarded_blocks": [],
+            "para_blocks": [
+                {
+                    "type": "text",
+                    "bbox": [
+                        52,
+                        61.956024169921875,
+                        294,
+                        82.99800872802734
+                    ],
+                    "lines": [
+                        {
+                            "bbox": [
+                                52,
+                                61.956024169921875,
+                                294,
+                                72.0000228881836
+                            ],
+                            "spans": [
+                                {
+                                    "bbox": [
+                                        54.0,
+                                        61.956024169921875,
+                                        296.2261657714844,
+                                        72.0000228881836
+                                    ],
+                                    "content": "dependent on the service headway and the reliability of the departure ",
+                                    "type": "text",
+                                    "score": 1.0
+                                }
+                            ]
+                        }
+                    ]
+                }
+            ]
+        }
+    ],
+    "_backend": "pipeline",
+    "_version_name": "0.6.1"
+}
+

内容列表 (content_list.json)

文件命名格式：{原文件名}_content_list.json

功能说明

这是一个简化版的 middle.json，按阅读顺序平铺存储所有可读内容块，去除了复杂的布局信息，便于后续处理。

内容类型

+ + + + + + + + + + + + + + + + + + + + + + + + + +

类型	说明
`image`	图片
`table`	表格
`text`	文本/标题
`equation`	行间公式

文本层级标识

通过 text_level 字段区分文本层级：

无 text_level 或 text_level: 0：正文文本
text_level: 1：一级标题
text_level: 2：二级标题
以此类推...

通用字段

所有内容块都包含 page_idx 字段，表示所在页码（从 0 开始）。
所有内容块都包含 bbox 字段，表示内容块的边界框坐标 [x0, y0, x1, y1] 映射在0-1000范围内的结果。

示例数据

[
+        {
+        "type": "text",
+        "text": "The response of flow duration curves to afforestation ",
+        "text_level": 1, 
+        "bbox": [
+            62,
+            480,
+            946,
+            904
+        ],
+        "page_idx": 0
+    },
+    {
+        "type": "image",
+        "img_path": "images/a8ecda1c69b27e4f79fce1589175a9d721cbdc1cf78b4cc06a015f3746f6b9d8.jpg",
+        "image_caption": [
+            "Fig. 1. Annual flow duration curves of daily flows from Pine Creek, Australia, 1989–2000. "
+        ],
+        "image_footnote": [],
+        "bbox": [
+            62,
+            480,
+            946,
+            904
+        ],
+        "page_idx": 1
+    },
+    {
+        "type": "equation",
+        "img_path": "images/181ea56ef185060d04bf4e274685f3e072e922e7b839f093d482c29bf89b71e8.jpg",
+        "text": "$$\nQ _ { \\% } = f ( P ) + g ( T )\n$$",
+        "text_format": "latex",
+        "bbox": [
+            62,
+            480,
+            946,
+            904
+        ],
+        "page_idx": 2
+    },
+    {
+        "type": "table",
+        "img_path": "images/e3cb413394a475e555807ffdad913435940ec637873d673ee1b039e3bc3496d0.jpg",
+        "table_caption": [
+            "Table 2 Significance of the rainfall and time terms "
+        ],
+        "table_footnote": [
+            "indicates that the rainfall term was significant at the $5 \\%$ level, $T$ indicates that the time term was significant at the $5 \\%$ level, \\* represents significance at the $10 \\%$ level, and na denotes too few data points for meaningful analysis. "
+        ],
+        "table_body": "<html><body><table><tr><td rowspan=\"2\">Site</td><td colspan=\"10\">Percentile</td></tr><tr><td>10</td><td>20</td><td>30</td><td>40</td><td>50</td><td>60</td><td>70</td><td>80</td><td>90</td><td>100</td></tr><tr><td>Traralgon Ck</td><td>P</td><td>P,*</td><td>P</td><td>P</td><td>P,</td><td>P,</td><td>P,</td><td>P,</td><td>P</td><td>P</td></tr><tr><td>Redhill</td><td>P,T</td><td>P,T</td><td>，*</td><td>**</td><td>P.T</td><td>P,*</td><td>P*</td><td>P*</td><td>*</td><td>，*</td></tr><tr><td>Pine Ck</td><td></td><td>P,T</td><td>P,T</td><td>P,T</td><td>P,T</td><td>T</td><td>T</td><td>T</td><td>na</td><td>na</td></tr><tr><td>Stewarts Ck 5</td><td>P,T</td><td>P,T</td><td>P,T</td><td>P,T</td><td>P.T</td><td>P.T</td><td>P,T</td><td>na</td><td>na</td><td>na</td></tr><tr><td>Glendhu 2</td><td>P</td><td>P,T</td><td>P,*</td><td>P,T</td><td>P.T</td><td>P,ns</td><td>P,T</td><td>P,T</td><td>P,T</td><td>P,T</td></tr><tr><td>Cathedral Peak 2</td><td>P,T</td><td>P,T</td><td>P,T</td><td>P,T</td><td>P,T</td><td>*,T</td><td>P,T</td><td>P,T</td><td>P,T</td><td>T</td></tr><tr><td>Cathedral Peak 3</td><td>P.T</td><td>P.T</td><td>P,T</td><td>P,T</td><td>P,T</td><td>T</td><td>P,T</td><td>P,T</td><td>P,T</td><td>T</td></tr><tr><td>Lambrechtsbos A</td><td>P,T</td><td>P</td><td>P</td><td>P,T</td><td>*,T</td><td>*,T</td><td>*,T</td><td>*,T</td><td>*,T</td><td>T</td></tr><tr><td>Lambrechtsbos B</td><td>P,T</td><td>P,T</td><td>P,T</td><td>P,T</td><td>P,T</td><td>P,T</td><td>P,T</td><td>P,T</td><td>T</td><td>T</td></tr><tr><td>Biesievlei</td><td>P,T</td><td>P.T</td><td>P,T</td><td>P,T</td><td>*,T</td><td>*,T</td><td>T</td><td>T</td><td>P,T</td><td>P,T</td></tr></table></body></html>",
+        "bbox": [
+            62,
+            480,
+            946,
+            904
+        ],  
+        "page_idx": 5
+    }
+]
+

VLM 后端输出结果

模型推理结果 (model.json)

文件命名格式：{原文件名}_model.json

文件格式说明

该文件为 VLM 模型的原始输出结果，包含两层嵌套list，外层表示页面，内层表示该页的内容块
每个内容块都是一个dict，包含 type、bbox、angle、content 字段

支持的内容类型

{
+    "text": "文本",
+    "title": "标题", 
+    "equation": "行间公式",
+    "image": "图片",
+    "image_caption": "图片描述",
+    "image_footnote": "图片脚注",
+    "table": "表格",
+    "table_caption": "表格描述",
+    "table_footnote": "表格脚注",
+    "phonetic": "拼音",
+    "code": "代码块",
+    "code_caption": "代码描述",
+    "ref_text": "参考文献",
+    "algorithm": "算法块",
+    "list": "列表",
+    "header": "页眉",
+    "footer": "页脚",
+    "page_number": "页码",
+    "aside_text": "装订线旁注", 
+    "page_footnote": "页面脚注"
+}
+

坐标系统说明

bbox 坐标格式：[x0, y0, x1, y1]

分别表示左上、右下两点的坐标
坐标原点在页面左上角
坐标为相对于原始页面尺寸的百分比，范围在0-1之间

示例数据

[
+    [
+        {
+            "type": "header",
+            "bbox": [
+                0.077,
+                0.095,
+                0.18,
+                0.181
+            ],
+            "angle": 0,
+            "score": null,
+            "block_tags": null,
+            "content": "ELSEVIER",
+            "format": null,
+            "content_tags": null
+        },
+        {
+            "type": "title",
+            "bbox": [
+                0.157,
+                0.228,
+                0.833,
+                0.253
+            ],
+            "angle": 0,
+            "score": null,
+            "block_tags": null,
+            "content": "The response of flow duration curves to afforestation",
+            "format": null,
+            "content_tags": null
+        }
+    ]
+]
+

中间处理结果 (middle.json)

文件命名格式：{原文件名}_middle.json

文件格式说明

vlm 后端的 middle.json 文件结构与 pipeline 后端类似，但存在以下差异：

+
list变成二级block，增加sub_type字段区分list类型:
+
- text（文本类型）
- ref_text（引用类型）
+
+
增加code类型block，code类型包含两种"sub_type":
+
- 分别是code和algorithm
- 至少有code_body, 可选code_caption
+
+
discarded_blocks内元素type增加以下类型:
+
- header（页眉）
- footer（页脚）
- page_number（页码）
- aside_text（装订线文本）
- page_footnote（脚注）
+
所有block增加angle字段，用来表示旋转角度，0，90，180，270

示例数据

list block 示例 +

{
+    "bbox": [
+        174,
+        155,
+        818,
+        333
+    ],
+    "type": "list",
+    "angle": 0,
+    "index": 11,
+    "blocks": [
+        {
+            "bbox": [
+                174,
+                157,
+                311,
+                175
+            ],
+            "type": "text",
+            "angle": 0,
+            "lines": [
+                {
+                    "bbox": [
+                        174,
+                        157,
+                        311,
+                        175
+                    ],
+                    "spans": [
+                        {
+                            "bbox": [
+                                174,
+                                157,
+                                311,
+                                175
+                            ],
+                            "type": "text",
+                            "content": "H.1 Introduction"
+                        }
+                    ]
+                }
+            ],
+            "index": 3
+        },
+        {
+            "bbox": [
+                175,
+                182,
+                464,
+                229
+            ],
+            "type": "text",
+            "angle": 0,
+            "lines": [
+                {
+                    "bbox": [
+                        175,
+                        182,
+                        464,
+                        229
+                    ],
+                    "spans": [
+                        {
+                            "bbox": [
+                                175,
+                                182,
+                                464,
+                                229
+                            ],
+                            "type": "text",
+                            "content": "H.2 Example: Divide by Zero without Exception Handling"
+                        }
+                    ]
+                }
+            ],
+            "index": 4
+        }
+    ],
+    "sub_type": "text"
+}
+

code block 示例 +

{
+    "type": "code",
+    "bbox": [
+        114,
+        780,
+        885,
+        1231
+    ],
+    "blocks": [
+        {
+            "bbox": [
+                114,
+                780,
+                885,
+                1231
+            ],
+            "lines": [
+                {
+                    "bbox": [
+                        114,
+                        780,
+                        885,
+                        1231
+                    ],
+                    "spans": [
+                        {
+                            "bbox": [
+                                114,
+                                780,
+                                885,
+                                1231
+                            ],
+                            "type": "text",
+                            "content": "1 // Fig. H.1: DivideByZeroNoExceptionHandling.java  \n2 // Integer division without exception handling.  \n3 import java.util.Scanner;  \n4  \n5 public class DivideByZeroNoExceptionHandling  \n6 {  \n7 // demonstrates throwing an exception when a divide-by-zero occurs  \n8 public static int quotient( int numerator, int denominator )  \n9 {  \n10 return numerator / denominator; // possible division by zero  \n11 } // end method quotient  \n12  \n13 public static void main(String[] args)  \n14 {  \n15 Scanner scanner = new Scanner(System.in); // scanner for input  \n16  \n17 System.out.print(\"Please enter an integer numerator: \");  \n18 int numerator = scanner.nextInt();  \n19 System.out.print(\"Please enter an integer denominator: \");  \n20 int denominator = scanner.nextInt();  \n21"
+                        }
+                    ]
+                }
+            ],
+            "index": 17,
+            "angle": 0,
+            "type": "code_body"
+        },
+        {
+            "bbox": [
+                867,
+                160,
+                1280,
+                189
+            ],
+            "lines": [
+                {
+                    "bbox": [
+                        867,
+                        160,
+                        1280,
+                        189
+                    ],
+                    "spans": [
+                        {
+                            "bbox": [
+                                867,
+                                160,
+                                1280,
+                                189
+                            ],
+                            "type": "text",
+                            "content": "Algorithm 1 Modules for MCTSteg"
+                        }
+                    ]
+                }
+            ],
+            "index": 19,
+            "angle": 0,
+            "type": "code_caption"
+        }
+    ],
+    "index": 17,
+    "sub_type": "code"
+}
+

内容列表 (content_list.json)

文件命名格式：{原文件名}_content_list.json

文件格式说明

vlm 后端的 content_list.json 文件结构与 pipeline 后端类似，伴随本次middle.json的变化，做了以下调整：

+
新增code类型，code类型包含两种"sub_type":
+
- 分别是code和algorithm
- 至少有code_body, 可选code_caption
+
+
新增list类型，list类型包含两种"sub_type":
+
- text
- ref_text
+
+
增加所有所有discarded_blocks的输出内容
+
- header
- footer
- page_number
- aside_text
- page_footnote
+

示例数据

code 类型 content +

{
+    "type": "code",
+    "sub_type": "algorithm",
+    "code_caption": [
+        "Algorithm 1 Modules for MCTSteg"
+    ],
+    "code_body": "1: function GETCOORDINATE(d)  \n2:  $x \\gets d / l$ ,  $y \\gets d$  mod  $l$   \n3: return  $(x, y)$   \n4: end function  \n5: function BESTCHILD(v)  \n6:  $C \\gets$  child set of  $v$   \n7:  $v' \\gets \\arg \\max_{c \\in C} \\mathrm{UCTScore}(c)$   \n8:  $v'.n \\gets v'.n + 1$   \n9: return  $v'$   \n10: end function  \n11: function BACK PROPAGATE(v)  \n12: Calculate  $R$  using Equation 11  \n13: while  $v$  is not a root node do  \n14:  $v.r \\gets v.r + R$ ,  $v \\gets v.p$   \n15: end while  \n16: end function  \n17: function RANDOMSEARCH(v)  \n18: while  $v$  is not a leaf node do  \n19: Randomly select an untried action  $a \\in A(v)$   \n20: Create a new node  $v'$   \n21:  $(x, y) \\gets \\mathrm{GETCOORDINATE}(v'.d)$   \n22:  $v'.p \\gets v$ ,  $v'.d \\gets v.d + 1$ ,  $v'.\\Gamma \\gets v.\\Gamma$   \n23:  $v'.\\gamma_{x,y} \\gets a$   \n24: if  $a = -1$  then  \n25:  $v.lc \\gets v'$   \n26: else if  $a = 0$  then  \n27:  $v.mc \\gets v'$   \n28: else  \n29:  $v.rc \\gets v'$   \n30: end if  \n31:  $v \\gets v'$   \n32: end while  \n33: return  $v$   \n34: end function  \n35: function SEARCH(v)  \n36: while  $v$  is fully expanded do  \n37:  $v \\gets$  BESTCHILD(v)  \n38: end while  \n39: if  $v$  is not a leaf node then  \n40:  $v \\gets$  RANDOMSEARCH(v)  \n41: end if  \n42: return  $v$   \n43: end function",
+    "bbox": [
+        510,
+        87,
+        881,
+        740
+    ],
+    "page_idx": 0
+}
+

list 类型 content +

{
+    "type": "list",
+    "sub_type": "text",
+    "list_items": [
+        "H.1 Introduction",
+        "H.2 Example: Divide by Zero without Exception Handling",
+        "H.3 Example: Divide by Zero with Exception Handling",
+        "H.4 Summary"
+    ],
+    "bbox": [
+        174,
+        155,
+        818,
+        333
+    ],
+    "page_idx": 0
+}
+

discarded 类型 content +

[{
+    "type": "header",
+    "text": "Journal of Hydrology 310 (2005) 253-265",
+    "bbox": [
+        363,
+        164,
+        623,
+        177
+    ],
+    "page_idx": 0
+},
+{
+    "type": "page_footnote",
+    "text": "* Corresponding author. Address: Forest Science Centre, Department of Sustainability and Environment, P.O. Box 137, Heidelberg, Vic. 3084, Australia. Tel.: +61 3 9450 8719; fax: +61 3 9450 8644.",
+    "bbox": [
+        71,
+        815,
+        915,
+        841
+    ],
+    "page_idx": 0
+}]
+

总结

以上文件为 MinerU 的完整输出结果，用户可根据需要选择合适的文件进行后续处理：

+
模型输出(使用原始输出):
+
- model.json
+
+
调试和验证(使用可视化文件):
+
- layout.pdf
- span.pdf
+
+
内容提取(使用简化文件):
+
- *.md
- content_list.json
+
+
二次开发(使用结构化文件):
+
- middle.json
+

+ + + + + + + + + + + + + + + +

+ + + + + +

+ + + +

+ + + + + + + + + + \ No newline at end of file diff --git a/zh/usage/acceleration_cards/AMD/index.html b/zh/usage/acceleration_cards/AMD/index.html new file mode 100644 index 00000000..ab0e0c80 --- /dev/null +++ b/zh/usage/acceleration_cards/AMD/index.html @@ -0,0 +1,1969 @@ + + + + + + + + + + + + + + + + + + + + + AMD - MinerU + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

+ + + + 跳转至 + + +

+ +

+ + + + +

+ + +

+ + + + + + +

+ + MinerU + +

+ + + AMD + + +

+ + + + + + + + + +

+ + +

+ + English + +
+ + 中文 + +

+ + + + + + + + + +

+ +

+ + +

+ opendatalab/MinerU +

+ +

+ + + + + +

+ + + + + + 主页 + + +
+ + + + + + 快速入门 + + +
+ + + + + + 使用指南 + + +
+ + + + + + 参考资料 + + +
+ + + + + + 常见问题解答 + + +
+ + + + + + 在线演示 + + +

+ + + +

+ +

+ + + + +

+ + + + + + + +

AMD

+ +

基于Triton的ROCm 不同后端实现优化，基本实现vllm后端正常推理，以及pipeline后端中第一步layout用的DocLayout-YOLO

已有完整python vllm和mineru环境直接跳转第五步！！！ +其他GPU执行问题可以参考，先prof查看定位找到哪个算子问题，然后triton后端实现即可 +测试了一下，基本和MinerU官网效果差不多，用AMD的人也不是很多，就在评论区分享给大家了

1.结果介绍

补充一个200页的PDF python编程书测试一下速度，可以到1.99it/s： +Two Step Extraction: 100%|████████████████████████████████████████| 200/200 [01:40<00:00, 1.99it/s]

下面为之前14学术论文测试结果： +7900xtx mineru-gradio --server-name 0.0.0.0 --server-port 7860 --enable-vllm-engine true 速度大概为1.6-1.8s/it，没有仔细测试，简单试了两个文档。第二种矩阵乘法代替原来的dots点乘可以进一步提速到1.3s/it，优化后的主要算子耗时在hipblast(这个没法提升了)和vllm triton后端，各占25%耗时吧，vllm tirion后端这个这个只能等官方优化了。。。。 +doclayout-yolo的layout速度从原来的1.6it/s提高到15it/s，注意需要缓存一下输入的pdf尺寸后，triton必须要缓存尺寸没办法。主要是为了保留模型输入输出接口，最小代码改动。 +采用-b vlm-vllm-engine模式举个例子

测试结果为优化为5d矩阵乘代替原来的点积结果： +2025-10-05 15:45:12.985 | INFO | mineru.backend.vlm.vlm_analyze:get_model:128 - get vllm-engine predictor cost: 18.45s +Adding requests: 100%|████████████████████████████████████████████████████████████████████████████████| 14/14 [00:01<00:00, 12.20it/s] +Processed prompts: 100%|█████████████████████| 14/14 [00:08<00:00, 1.56it/s, est. speed input: 2174.18 toks/s, output: 791.87 toks/s] +Adding requests: 100%|█████████████████████████████████████████████████████████████████████████████| 278/278 [00:00<00:00, 323.03it/s] +Processed prompts: 100%|██████████████████| 278/278 [00:07<00:00, 37.63it/s, est. speed input: 5264.66 toks/s, output: 2733.31 toks/s]

mineru-gradio --server-name 0.0.0.0 --server-port 7860 --enable-vllm-engine true测试： +2025-10-05 15:46:55.953 | WARNING | mineru.cli.common:convert_pdf_bytes_to_bytes_by_pypdfium2:54 - end_page_id is out of range, use pdf_docs length +Two Step Extraction: 100%|████████████████████████████████████████████████████████████████████████████| 14/14 [00:18<00:00, 1.30s/it]

2.原因介绍

AMD RDNA使用vllm后端有严重的性能问题，原因是因为vllm的qwen2_vl.py中有一个算子在rocm kernel上没有对应的实现，导致性能出现严重的卷积计算回退，一次执行花了12s，。。。。。。。。一言难尽。即MIOpen 库中缺少模型中特定 Conv3d(bfloat16) 的优化内核。 +DocLayout-YOLO的g2l_crm.py空洞卷积也是这个问题，专业的CDNA MI210也没解决这个问题 +正好一起处理了。

3.环境介绍

System: Ubuntu 24.04.3 Kernel: Linux 6.14.0-33-generic ROCm version: 7.0.1 +python环境： +python 3.12 +pytorch-triton-rocm 3.5.0+gitbbb06c03 +torch 2.10.0.dev20251001+rocm7.0 +torchvision 0.25.0.dev20251003+rocm7.0 +vllm 0.11.0rc2.dev198+g736fbf4c8.rocm701 +不同版本无所谓，处理方法是一样的。

4.前置环境安装

uv venv --python python3.12
+source .venv/bin/activate
+uv pip install --pre torch torchvision   -i https://pypi.tuna.tsinghua.edu.cn/simple/   --extra-index-url https://download.pytorch.org/whl/nightly/rocm7.0
+uv pip install pip
+# 避免覆盖我们本地的pytorch，改用pip而没有继续使用uv pip
+pip install -U "mineru[core]" -i https://pypi.mirrors.ustc.edu.cn/simple/
+

+vllm 安装参考官方手册Vllm +

#手动安装aiter，vllm，amd-smi等，自行找一个位置clone，然后进入该目录吧
+git clone --recursive https://github.com/ROCm/aiter.git
+cd aiter
+git submodule sync; git submodule update --init --recursive
+python setup.py develop
+cd ..
+git clone https://github.com/vllm-project/vllm.git
+cd vllm/
+cp -r /opt/rocm/share/amd_smi ~/Pytorch/vllm/
+pip install amd_smi/
+pip install --upgrade numba \
+    scipy \
+    huggingface-hub[cli,hf_transfer] \
+    setuptools_scm
+pip install -r requirements/rocm.txt
+export PYTORCH_ROCM_ARCH="gfx1100"   #根据自己的GPU架构 rocminfo | grep gfx
+python setup.py develop
+

5.vllm中关键triton算子添加

这里我给出两种解决方法，第一种解决方法就是前面提到的优化到1.5到1.8s/it，第二种方法有手动优化算子到矩阵乘法，7900xtx肯定适用，大概1.3s/it，其他AMD GPU相对方案一也有提速，但是不一定是最佳速度实现，里面的手动部分可能需要微调。

注意pip把triton 后端的flash_attn卸载了，搞了半天各种尝试还是报错，问题比较大，直接不用就行了 +

#定位自己vllm位置XXX
+pip show vllm
+

+关键更改 +XXX/vllm/model_executor/models/qwen2_vl.py文件： +1.qwen2_vl.py文件33行下增加from .qwen2_vl_vision_kernels import triton_conv3d_patchify +

from collections.abc import Iterable, Mapping, Sequence
+from functools import partial
+from typing import Annotated, Any, Callable, Literal, Optional, Union
+
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+from .qwen2_vl_vision_kernels import triton_conv3d_patchify
+

+接下来分为方案一(2.1和3.1)和方案二(2.2和3.2)，选取一种实现即可 +

方案1 +2.1qwen2_vl.py文件498行class Qwen2VisionPatchEmbed(nn.Module),PS.就是这玩意AMD没有现成的内核算子导致回退 +

class Qwen2VisionPatchEmbed(nn.Module):
+
+    def __init__(
+        self,
+        patch_size: int = 14,
+        temporal_patch_size: int = 2,
+        in_channels: int = 3,
+        embed_dim: int = 1152,
+    ) -> None:
+        super().__init__()
+        self.patch_size = patch_size
+        self.temporal_patch_size = temporal_patch_size
+        self.embed_dim = embed_dim
+
+        kernel_size = (temporal_patch_size, patch_size, patch_size)
+        self.proj = nn.Conv3d(in_channels,
+                              embed_dim,
+                              kernel_size=kernel_size,
+                              stride=kernel_size,
+                              bias=False)
+    def forward(self, x: torch.Tensor) -> torch.Tensor:
+        L, C = x.shape
+        x_reshaped = x.view(L, -1, self.temporal_patch_size, self.patch_size,
+                            self.patch_size)
+
+        # Call your custom Triton kernel instead of self.proj
+        x_out = triton_conv3d_patchify(x_reshaped, self.proj.weight)
+
+        # The output of our kernel is already the correct shape [L, embed_dim]
+        return x_out
+

+3.1XXX/vllm/model_executor/models/目录下创建qwen2_vl_vision_kernels.py文件，用triton实现 +

import torch
+from vllm.triton_utils import tl, triton
+
+@triton.jit
+def _conv3d_patchify_kernel(
+    # Pointers to tensors
+    X, W, Y,
+    # Tensor dimensions
+    N, C_in, D_in, H_in, W_in,
+    C_out, KD, KH, KW,
+    # Stride and padding for memory access
+    stride_xn, stride_xc, stride_xd, stride_xh, stride_xw,
+    stride_wn, stride_wc, stride_wd, stride_wh, stride_ww,
+    stride_yn, stride_yc,
+    # Triton-specific metaparameters
+    BLOCK_SIZE: tl.constexpr,
+):
+    """
+    Triton kernel for a non-overlapping 3D patching convolution.
+    Each kernel instance computes one output value for one patch.
+    """
+    # Get the program IDs for the N (patch) and C_out (output channel) dimensions
+    pid_n = tl.program_id(0)  # The index of the patch we are processing
+    pid_cout = tl.program_id(1) # The index of the output channel we are computing
+
+    # --- Calculate memory pointers ---
+    # Pointer to the start of the current input patch
+    x_ptr = X + (pid_n * stride_xn)
+    # Pointer to the start of the current filter (weight)
+    w_ptr = W + (pid_cout * stride_wn)
+    # Pointer to where the output will be stored
+    y_ptr = Y + (pid_n * stride_yn + pid_cout * stride_yc)
+
+    # --- Perform the convolution (element-wise product and sum) ---
+    # This is a dot product between the flattened patch and the flattened filter.
+    accumulator = tl.zeros((BLOCK_SIZE,), dtype=tl.float32)
+
+    # Iterate over the elements of the patch/filter
+    for c_offset in range(0, C_in):
+        for d_offset in range(0, KD):
+            for h_offset in range(0, KH):
+                # Unrolled loop for the innermost dimension (width) for performance
+                for w_offset in range(0, KW, BLOCK_SIZE):
+                    # Create masks to handle cases where KW is not a multiple of BLOCK_SIZE
+                    w_range = w_offset + tl.arange(0, BLOCK_SIZE)
+                    w_mask = w_range < KW
+
+                    # Calculate offsets to load data
+                    patch_offset = (c_offset * stride_xc + d_offset * stride_xd +
+                                    h_offset * stride_xh + w_range * stride_xw)
+                    filter_offset = (c_offset * stride_wc + d_offset * stride_wd +
+                                     h_offset * stride_wh + w_range * stride_ww)
+
+                    # Load patch and filter data, applying masks
+                    patch_vals = tl.load(x_ptr + patch_offset, mask=w_mask, other=0.0)
+                    filter_vals = tl.load(w_ptr + filter_offset, mask=w_mask, other=0.0)
+
+                    # Multiply and accumulate
+                    accumulator += patch_vals.to(tl.float32) * filter_vals.to(tl.float32)
+
+    # Sum the accumulator block and store the single output value
+    output_val = tl.sum(accumulator, axis=0)
+    tl.store(y_ptr, output_val)
+
+
+def triton_conv3d_patchify(x: torch.Tensor, weight: torch.Tensor) -> torch.Tensor:
+    """
+    Python wrapper for the 3D patching convolution Triton kernel.
+    """
+    # Get tensor dimensions
+    N, C_in, D_in, H_in, W_in = x.shape
+    C_out, _, KD, KH, KW = weight.shape
+
+    # Create the output tensor
+    # The output of this specific conv is (N, C_out, 1, 1, 1), which we squeeze
+    Y = torch.empty((N, C_out), dtype=x.dtype, device=x.device)
+
+    # Define the grid for launching the Triton kernel
+    # Each kernel instance handles one patch (N) for one output channel (C_out)
+    grid = (N, C_out)
+
+    # Launch the kernel
+    # We pass all strides to make the kernel flexible
+    _conv3d_patchify_kernel[grid](
+        x, weight, Y,
+        N, C_in, D_in, H_in, W_in,
+        C_out, KD, KH, KW,
+        x.stride(0), x.stride(1), x.stride(2), x.stride(3), x.stride(4),
+        weight.stride(0), weight.stride(1), weight.stride(2), weight.stride(3), weight.stride(4),
+        Y.stride(0), Y.stride(1),
+        BLOCK_SIZE=16, # A reasonable default, can be tuned
+    )
+
+    return Y
+

方案2 +2.2qwen2_vl.py文件498行class Qwen2VisionPatchEmbed(nn.Module)函数,PS.就是这玩意AMD没有现成的内核算子导致回退，这里我们直接5D张量一步到位，改为矩阵乘法 +

class Qwen2VisionPatchEmbed(nn.Module):
+
+    def __init__(
+        self,
+        patch_size: int = 14,
+        temporal_patch_size: int = 2,
+        in_channels: int = 3,
+        embed_dim: int = 1152,
+    ) -> None:
+        super().__init__()
+        self.patch_size = patch_size
+        self.temporal_patch_size = temporal_patch_size
+        self.embed_dim = embed_dim
+
+        kernel_size = (temporal_patch_size, patch_size, patch_size)
+
+        self.proj = nn.Conv3d(in_channels,
+                              embed_dim,
+                              kernel_size=kernel_size,
+                              stride=kernel_size,
+                              bias=False)
+
+    def forward(self, x: torch.Tensor) -> torch.Tensor:
+        L, C = x.shape
+        x_reshaped_5d = x.view(L, -1, self.temporal_patch_size, self.patch_size,
+                               self.patch_size)
+
+        return triton_conv3d_patchify(x_reshaped_5d, self.proj.weight)
+

+3.2XXX/vllm/model_executor/models/目录下创建qwen2_vl_vision_kernels.py文件，用triton实现 +

import torch
+from vllm.triton_utils import tl, triton
+
+@triton.jit
+def _conv_gemm_kernel(
+    A, B, C, M, N, K,
+    stride_am, stride_ak,
+    stride_bk, stride_bn,
+    stride_cm, stride_cn,
+    BLOCK_M: tl.constexpr, BLOCK_N: tl.constexpr, BLOCK_K: tl.constexpr,
+):
+    pid_m = tl.program_id(0)
+    pid_n = tl.program_id(1)
+    offs_m = pid_m * BLOCK_M + tl.arange(0, BLOCK_M)
+    offs_n = pid_n * BLOCK_N + tl.arange(0, BLOCK_N)
+    offs_k = tl.arange(0, BLOCK_K)
+    a_ptrs = A + (offs_m[:, None] * stride_am + offs_k[None, :] * stride_ak)
+    b_ptrs = B + (offs_k[:, None] * stride_bk + offs_n[None, :] * stride_bn)
+    accumulator = tl.zeros((BLOCK_M, BLOCK_N), dtype=tl.float32)
+    for k in range(0, K, BLOCK_K):
+        a = tl.load(a_ptrs, mask=(offs_m[:, None] < M) & (offs_k[None, :] < K), other=0.0)
+        b = tl.load(b_ptrs, mask=(offs_k[:, None] < K) & (offs_n[None, :] < N), other=0.0)
+        accumulator += tl.dot(a, b)
+        a_ptrs += BLOCK_K * stride_ak
+        b_ptrs += BLOCK_K * stride_bk
+        offs_k += BLOCK_K
+    c = accumulator.to(C.dtype.element_ty)
+    offs_cm = pid_m * BLOCK_M + tl.arange(0, BLOCK_M)
+    offs_cn = pid_n * BLOCK_N + tl.arange(0, BLOCK_N)
+    c_ptrs = C + stride_cm * offs_cm[:, None] + stride_cn * offs_cn[None, :]
+    c_mask = (offs_cm[:, None] < M) & (offs_cn[None, :] < N)
+    tl.store(c_ptrs, c, mask=c_mask)
+
+def triton_conv3d_patchify(x_5d: torch.Tensor, weight_5d: torch.Tensor) -> torch.Tensor:
+    N_patches, _, _, _, _ = x_5d.shape
+    C_out, _, _, _, _ = weight_5d.shape
+    A = x_5d.view(N_patches, -1)
+    B = weight_5d.view(C_out, -1).transpose(0, 1).contiguous()
+    M, K = A.shape
+    _K, N = B.shape
+    assert K == _K
+    C = torch.empty((M, N), device=A.device, dtype=A.dtype)
+
+    # --- 针对7900xtx的手动调优配置，其他GPU的最优组合可能需要自行寻找，AMD的autotune效果就是没有效果 ---
+    best_config = {
+        'BLOCK_M': 128,
+        'BLOCK_N': 128,
+        'BLOCK_K': 32,
+    }
+    num_stages = 4
+    num_warps = 8
+
+    grid = (triton.cdiv(M, best_config['BLOCK_M']),
+            triton.cdiv(N, best_config['BLOCK_N']))
+
+    _conv_gemm_kernel[grid](
+        A, B, C,
+        M, N, K,
+        A.stride(0), A.stride(1),
+        B.stride(0), B.stride(1),
+        C.stride(0), C.stride(1),
+        **best_config,
+        num_stages=num_stages,
+        num_warps=num_warps
+    )
+
+    return C
+

4.关闭终端后再次使用mineru-gradio会报一个Lora错误，修改代码跳过它 +

pip show mineru_vl_utils
+

打开该文件XXX/mineru_vl_utils/vlm_client/vllm_async_engine_client.py修改第58行self.tokenizer = vllm_async_llm.tokenizer.get_lora_tokenizer()为： +

        try:
+            self.tokenizer = vllm_async_llm.tokenizer.get_lora_tokenizer()
+        except AttributeError:
+            # 如果没有 get_lora_tokenizer 方法，直接使用原始 tokenizer
+            self.tokenizer = vllm_async_llm.tokenizer
+

最后整两个环境变量后愉快玩耍即可 +

export MINERU_MODEL_SOURCE=modelscope
+export TORCH_ROCM_AOTRITON_ENABLE_EXPERIMENTAL=1
+

6.vllm后端已经没有问题，下面是pipeline 中layout用的doclayout-yolo模型空洞卷积问题

我在 DocLayout-YOLO 下做了一个回答，因此 pipeline 的空洞卷积问题不在这里赘述，直接点击链接查看即可。

查看自己doclayout-yolo安装位置如下，然后进入修改链接中回复介绍的文件即可 +

pip show doclayout-yolo
+

+ + + + + + + + + + + + + + + +

+ + + + + +

+ + + +

+ + + + + + + + + + \ No newline at end of file diff --git a/zh/usage/acceleration_cards/Ascend/index.html b/zh/usage/acceleration_cards/Ascend/index.html new file mode 100644 index 00000000..71837cdf --- /dev/null +++ b/zh/usage/acceleration_cards/Ascend/index.html @@ -0,0 +1,1805 @@ + + + + + + + + + + + + + + + + + + + + + Ascend - MinerU + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

+ + + + 跳转至 + + +

+ +

+ + + + +

+ + +

+ + + + + + +

+ + MinerU + +

+ + + Ascend + + +

+ + + + + + + + + +

+ + +

+ + English + +
+ + 中文 + +

+ + + + + + + + + +

+ +

+ + +

+ opendatalab/MinerU +

+ +

+ + + + + +

+ + + + + + 主页 + + +
+ + + + + + 快速入门 + + +
+ + + + + + 使用指南 + + +
+ + + + + + 参考资料 + + +
+ + + + + + 常见问题解答 + + +
+ + + + + + 在线演示 + + +

+ + + +

+ +

+ + + + +

+ + + + + + + +

Ascend

+ +

1. 测试平台

以下为本指南测试使用的平台信息，供参考： +

os: CTyunOS 22.06  
+cpu: Kunpeng-920 (aarch64)  
+npu: Ascend 910B2  
+driver: 23.0.3 
+docker: 20.10.12
+

2. 环境准备

Note

Ascend加速卡支持使用vllm或lmdeploy进行VLM模型推理加速。请根据实际需求选择安装和使用其中之一:

2.1 使用 Dockerfile 构建镜像（vllm）

Tip

ascend-vllm支持设备如下:

Atlas A2 training series (Atlas 800T A2, Atlas 900 A2 PoD, Atlas 200T A2 Box16, Atlas 300T A2)
Atlas 800I A2 inference series (Atlas 800I A2)
Atlas A3 training series (Atlas 800T A3, Atlas 900 A3 SuperPoD, Atlas 9000 A3 SuperPoD)
Atlas 800I A3 inference series (Atlas 800I A3)
[Experimental] Atlas 300I inference series (Atlas 300I Duo)

Dockerfile文件第三行为ascend-vllm基础镜像信息,默认tag为A2适配的版本,例如 v0.11.0

如需使用A3适配的版本,请将第三行的tag修改为 v0.11.0-a3,然后再执行build操作。
如需使用Atlas 300I Duo适配的版本,请将第三行的tag修改为 v0.10.0rc1-310p,然后再执行build操作。

wget https://gcore.jsdelivr.net/gh/opendatalab/MinerU@master/docker/china/npu.Dockerfile
+docker build --network=host -t mineru:npu-vllm-latest -f npu.Dockerfile .
+

2.2 使用 Dockerfile 构建镜像（lmdeploy）

Tip

ascend-lmdeploy支持设备如下:

Atlas A2 training series (Atlas 800T A2, Atlas 900 A2 PoD, Atlas 200T A2 Box16, Atlas 300T A2)
Atlas 800I A2 inference series (Atlas 800I A2)

如果您的设备为Atlas A3系列或Atlas 300I Duo系列，请使用vllm版本的镜像。

wget https://gcore.jsdelivr.net/gh/opendatalab/MinerU@master/docker/china/npu.Dockerfile
+# 将基础镜像从 vllm 切换为 lmdeploy
+sed -i '3s/^/# /' npu.Dockerfile && sed -i '5s/^# //' npu.Dockerfile
+docker build --network=host -t mineru:npu-lmdeploy-latest -f npu.Dockerfile .
+

3. 启动 Docker 容器

docker run -u root --name mineru_docker --privileged=true \
+    --ipc=host \
+    --network=host \
+    --device=/dev/davinci0 \
+    --device=/dev/davinci1 \
+    --device=/dev/davinci2 \
+    --device=/dev/davinci3 \
+    --device=/dev/davinci4 \
+    --device=/dev/davinci5 \
+    --device=/dev/davinci6 \
+    --device=/dev/davinci7 \
+    --device=/dev/davinci_manager \
+    --device=/dev/devmm_svm \
+    --device=/dev/hisi_hdc \
+    -v /var/log/npu/:/usr/slog \
+    -v /usr/local/dcmi:/usr/local/dcmi \
+    -v /usr/local/bin/npu-smi:/usr/local/bin/npu-smi \
+    -v /usr/local/Ascend/driver:/usr/local/Ascend/driver \
+    -e VLLM_WORKER_MULTIPROC_METHOD=spawn \
+    -e MINERU_MODEL_SOURCE=local \
+    -e MINERU_LMDEPLOY_DEVICE=ascend \
+    -it mineru:npu-vllm-latest \
+    /bin/bash
+

Tip

请根据实际情况选择使用vllm或lmdeploy版本的镜像，如需使用lmdeploy，替换上述命令中的mineru:npu-vllm-latest为mineru:npu-lmdeploy-latest即可。

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。

Note

由于310p加速卡不支持图模式与bf16精度，因此在使用该加速卡时，执行任意与vllm相关命令需追加--enforce-eager --dtype float16参数。 +例如: +

mineru-openai-server --port 30000 --enforce-eager --dtype float16
+

4. 注意事项

不同环境下，MinerU对Ascend加速卡的支持情况如下表所示：

+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

使用场景		容器环境
使用场景		vllm	lmdeploy
命令行工具(mineru)	pipeline	🟢	🟢
	<vlm/hybrid>-auto-engine	🟢	🟢
	<vlm/hybrid>-http-client	🟢	🟢
fastapi服务(mineru-api)	pipeline	🟢	🟢
	<vlm/hybrid>-auto-engine	🟢	🟢
	<vlm/hybrid>-http-client	🟢	🟢
gradio界面(mineru-gradio)	pipeline	🟢	🟢
	<vlm/hybrid>-auto-engine	🟢	🟢
	<vlm/hybrid>-http-client	🟢	🟢
openai-server服务（mineru-openai-server）		🟢	🟢
数据并行 (--data-parallel-size/--dp)		🟢	🔴

+ +

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

NPU加速卡指定可用加速卡的方式与NVIDIA GPU类似，请参考ASCEND_RT_VISIBLE_DEVICES
在Ascend平台可以通过npu-smi info命令查看加速卡的使用情况，并根据需要指定空闲的加速卡ID以避免资源冲突。

+ + + + + + + + + + + + + + + +

+ + + + + +

+ + + +

+ + + + + + + + + + \ No newline at end of file diff --git a/zh/usage/acceleration_cards/Biren/index.html b/zh/usage/acceleration_cards/Biren/index.html new file mode 100644 index 00000000..f36ef3a4 --- /dev/null +++ b/zh/usage/acceleration_cards/Biren/index.html @@ -0,0 +1,1734 @@ + + + + + + + + + + + + + + + + + + + + + Biren - MinerU + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

+ + + + 跳转至 + + +

+ +

+ + + + +

+ + +

+ + + + + + +

+ + MinerU + +

+ + + Biren + + +

+ + + + + + + + + +

+ + +

+ + English + +
+ + 中文 + +

+ + + + + + + + + +

+ +

+ + +

+ opendatalab/MinerU +

+ +

+ + + + + +

+ + + + + + 主页 + + +
+ + + + + + 快速入门 + + +
+ + + + + + 使用指南 + + +
+ + + + + + 参考资料 + + +
+ + + + + + 常见问题解答 + + +
+ + + + + + 在线演示 + + +

+ + + +

+ +

+ + + + +

+ + + + + + + +

Biren

+ +

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 22.04.4 LTS
+cpu: Intel x86-64
+gpu: Biren 106C
+driver: 1.10.0
+docker: 28.0.4
+

2. 环境准备

2.1 下载并加载镜像（vllm）

wget http://birentech.com/xxx/MinerU/mineru-vllm.tar 链接获取请联系壁仞内部人员（邮箱：MonaLiu@birentech.com）
+docker load -i mineru-vllm.tar
+

3. 启动 Docker 容器

docker run -it --name mineru_docker \
+    --privileged \
+    --network=host \
+    --shm-size=100G \
+    -e MINERU_MODEL_SOURCE=local \
+    -e MINERU_DEVICE_MODEL=supa \
+    -e SHAPE_TRANSFORM_GRANK=true \
+    mineru:biren-vllm-latest \
+    /bin/bash
+

4. 注意事项

不同环境下，MinerU对Biren加速卡的支持情况如下表所示：

+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

使用场景		容器环境
使用场景		vllm
命令行工具(mineru)	pipeline	🟢
	<vlm/hybrid>-auto-engine	🟢
	<vlm/hybrid>-http-client	🟢
fastapi服务(mineru-api)	pipeline	🟢
	<vlm/hybrid>-auto-engine	🟢
	<vlm/hybrid>-http-client	🟢
gradio界面(mineru-gradio)	pipeline	🟢
	<vlm/hybrid>-auto-engine	🟢
	<vlm/hybrid>-http-client	🟢
openai-server服务（mineru-openai-server）		🟢
数据并行 (--data-parallel-size)		🔴

+ +

Tip

Biren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设备章节说明, +将环境变量CUDA_VISIBLE_DEVICES替换为SUPA_VISIBLE_DEVICES即可。
在壁仞平台可以通过brsmi命令查看加速卡的使用情况，并根据需要指定空闲的加速卡ID以避免资源冲突。

+ + + + + + + + + + + + + + + +

+ + + + + +

+ + + +

+ + + + + + + + + + \ No newline at end of file diff --git a/zh/usage/acceleration_cards/Cambricon/index.html b/zh/usage/acceleration_cards/Cambricon/index.html new file mode 100644 index 00000000..41eb5aac --- /dev/null +++ b/zh/usage/acceleration_cards/Cambricon/index.html @@ -0,0 +1,1790 @@ + + + + + + + + + + + + + + + + + + + + + Cambricon - MinerU + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

+ + + + 跳转至 + + +

+ +

+ + + + +

+ + +

+ + + + + + +

+ + MinerU + +

+ + + Cambricon + + +

+ + + + + + + + + +

+ + +

+ + English + +
+ + 中文 + +

+ + + + + + + + + +

+ +

+ + +

+ opendatalab/MinerU +

+ +

+ + + + + +

+ + + + + + 主页 + + +
+ + + + + + 快速入门 + + +
+ + + + + + 使用指南 + + +
+ + + + + + 参考资料 + + +
+ + + + + + 常见问题解答 + + +
+ + + + + + 在线演示 + + +

+ + + +

+ +

+ + + + +

+ + + + + + + +

Cambricon

+ +

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: Hygon Hygon C86 7490
+mlu: MLU590-M9D
+driver: v6.2.11
+docker: 28.3.0
+

2. 环境准备

Note

Cambricon加速卡支持使用lmdeploy或vllm进行VLM模型推理加速。请根据实际需求选择安装和使用其中之一:

2.1 使用 Dockerfile 构建镜像（lmdeploy）

wget https://gcore.jsdelivr.net/gh/opendatalab/MinerU@master/docker/china/mlu.Dockerfile
+docker build --network=host -t mineru:mlu-lmdeploy-latest -f mlu.Dockerfile .
+

2.2 使用 Dockerfile 构建镜像（vllm）

wget https://gcore.jsdelivr.net/gh/opendatalab/MinerU@master/docker/china/mlu.Dockerfile
+# 将基础镜像从 lmdeploy 切换为 vllm
+sed -i -e '3,4s/^/# /' -e '6,7s/^# //' mlu.Dockerfile
+docker build --network=host -t mineru:mlu-vllm-latest -f mlu.Dockerfile .
+

3. 启动 Docker 容器

docker run --name mineru_docker \
+   --privileged \
+   --ipc=host \
+   --network=host \
+   --shm-size=400g \
+   --ulimit memlock=-1 \
+   -v /dev:/dev \
+   -v /lib/modules:/lib/modules:ro \
+   -v /usr/bin/cnmon:/usr/bin/cnmon \
+   -e MINERU_MODEL_SOURCE=local \
+   -e MINERU_LMDEPLOY_DEVICE=camb \
+   -it mineru:mlu-lmdeploy-latest \
+   /bin/bash
+

Tip

请根据实际情况选择使用vllm或lmdeploy版本的镜像，如需使用vllm,请执行以下操作：

+
替换上述命令中的mineru:mlu-lmdeploy-latest为mineru:mlu-vllm-latest
+
+
进入容器后，通过以下命令切换venv环境： +
```
source /torch/venv3/pytorch_infer/bin/activate
+
```
+
+
切换成功后，您可以在命令行前看到(pytorch_infer)的标识，这表示您已成功进入vllm的虚拟环境。
+

4. 注意事项

Note

兼容性说明：由于寒武纪（Cambricon）目前对 vLLM v1 引擎的支持尚待完善，MinerU 现阶段采用 v0 引擎作为适配方案。 +受此限制，vLLM 的异步引擎（Async Engine）功能存在兼容性问题，可能导致部分使用场景无法正常运行。 +我们将持续跟进寒武纪对 vLLM v1 引擎的支持进展，并及时在 MinerU 中进行相应的适配与优化。

不同环境下，MinerU对Cambricon加速卡的支持情况如下表所示：

Tip

lmdeploy黄灯问题为不能输入文件夹使用批量解析功能，输入单个文件时表现正常。
vllm黄灯问题为在精度未对齐，在部分场景下可能出现预期外结果。

+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

使用场景		容器环境
使用场景		vllm	lmdeploy
命令行工具(mineru)	pipeline	🟢	🟢
	<vlm/hybrid>-auto-engine	🟡	🟡
	<vlm/hybrid>-http-client	🟡	🟢
fastapi服务(mineru-api)	pipeline	🟢	🟢
	<vlm/hybrid>-auto-engine	🔴	🟢
	<vlm/hybrid>-http-client	🟡	🟢
gradio界面(mineru-gradio)	pipeline	🟢	🟢
	<vlm/hybrid>-auto-engine	🔴	🟢
	<vlm/hybrid>-http-client	🟡	🟢
openai-server服务（mineru-openai-server）		🟡	🟢
数据并行 (--data-parallel-size/--dp)		🔴	🔴

+ +

Tip

Cambricon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设备章节说明, +将环境变量CUDA_VISIBLE_DEVICES替换为MLU_VISIBLE_DEVICES即可。
在Cambricon平台可以通过cnmon命令查看加速卡的使用情况，并根据需要指定空闲的加速卡ID以避免资源冲突。

+ + + + + + + + + + + + + + + +

+ + + + + +

+ + + +

+ + + + + + + + + + \ No newline at end of file diff --git a/zh/usage/acceleration_cards/Enflame/index.html b/zh/usage/acceleration_cards/Enflame/index.html new file mode 100644 index 00000000..60bfbc98 --- /dev/null +++ b/zh/usage/acceleration_cards/Enflame/index.html @@ -0,0 +1,1732 @@ + + + + + + + + + + + + + + + + + + + + + Enflame - MinerU + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

+ + + + 跳转至 + + +

+ +

+ + + + +

+ + +

+ + + + + + +

+ + MinerU + +

+ + + Enflame + + +

+ + + + + + + + + +

+ + +

+ + English + +
+ + 中文 + +

+ + + + + + + + + +

+ +

+ + +

+ opendatalab/MinerU +

+ +

+ + + + + +

+ + + + + + 主页 + + +
+ + + + + + 快速入门 + + +
+ + + + + + 使用指南 + + +
+ + + + + + 参考资料 + + +
+ + + + + + 常见问题解答 + + +
+ + + + + + 在线演示 + + +

+ + + +

+ +

+ + + + +

+ + + + + + + +

Enflame

+ +

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 22.04.4 LTS  
+cpu: Intel x86-64
+gcu: Enflame S60 
+driver: 1.7.0.9
+docker: 28.0.1
+

2. 环境准备

2.1 使用 Dockerfile 构建镜像

wget https://gcore.jsdelivr.net/gh/opendatalab/MinerU@master/docker/china/gcu.Dockerfile
+docker build --network=host -t mineru:gcu-vllm-latest -f gcu.Dockerfile .
+

3. 启动 Docker 容器

docker run -u root --name mineru_docker \
+    --network=host \
+    --ipc=host \
+    --privileged \
+    -e MINERU_MODEL_SOURCE=local \
+    -it mineru:gcu-vllm-latest \
+    /bin/bash
+

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：

+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

使用场景		容器环境
使用场景		vllm
命令行工具(mineru)	pipeline	🟢
	<vlm/hybrid>-auto-engine	🟢
	<vlm/hybrid>-http-client	🟢
fastapi服务(mineru-api)	pipeline	🟢
	<vlm/hybrid>-auto-engine	🟢
	<vlm/hybrid>-http-client	🟢
gradio界面(mineru-gradio)	pipeline	🟢
	<vlm/hybrid>-auto-engine	🟢
	<vlm/hybrid>-http-client	🟢
openai-server服务（mineru-openai-server）		🟢
数据并行 (--data-parallel-size)		🔴

+ +

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设备章节说明, +将环境变量CUDA_VISIBLE_DEVICES替换为TOPS_VISIBLE_DEVICES即可。
在Enflame平台可以通过efsmi命令查看加速卡的使用情况，并根据需要指定空闲的加速卡ID以避免资源冲突。

+ + + + + + + + + + + + + + + +

+ + + + + +

+ + + +

+ + + + + + + + + + \ No newline at end of file diff --git a/zh/usage/acceleration_cards/Hygon/index.html b/zh/usage/acceleration_cards/Hygon/index.html new file mode 100644 index 00000000..d1b3f867 --- /dev/null +++ b/zh/usage/acceleration_cards/Hygon/index.html @@ -0,0 +1,1738 @@ + + + + + + + + + + + + + + + + + + + + + Hygon - MinerU + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

+ + + + 跳转至 + + +

+ +

+ + + + +

+ + +

+ + + + + + +

+ + MinerU + +

+ + + Hygon + + +

+ + + + + + + + + +

+ + +

+ + English + +
+ + 中文 + +

+ + + + + + + + + +

+ +

+ + +

+ opendatalab/MinerU +

+ +

+ + + + + +

+ + + + + + 主页 + + +
+ + + + + + 快速入门 + + +
+ + + + + + 使用指南 + + +
+ + + + + + 参考资料 + + +
+ + + + + + 常见问题解答 + + +
+ + + + + + 在线演示 + + +

+ + + +

+ +

+ + + + +

+ + + + + + + +

Hygon

+ +

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 22.04.3 LTS  
+cpu: Hygon C86-4G(x86-64)
+dcu: BW200
+driver: 6.3.13-V1.12.0a
+docker: 20.10.24
+

2. 环境准备

2.1 使用 Dockerfile 构建镜像

wget https://gcore.jsdelivr.net/gh/opendatalab/MinerU@master/docker/china/dcu.Dockerfile
+docker build --network=host -t mineru:dcu-vllm-latest -f dcu.Dockerfile .
+

3. 启动 Docker 容器

docker run -u root --name mineru_docker \
+    --network=host \
+    --ipc=host \
+    --shm-size=16G \
+    --device=/dev/kfd \
+    --device=/dev/mkfd \
+    --device=/dev/dri \
+    -v /opt/hyhal:/opt/hyhal \
+    --group-add video \
+    --cap-add=SYS_PTRACE \
+    --security-opt seccomp=unconfined \
+    -e MINERU_MODEL_SOURCE=local \
+    -it mineru:dcu-vllm-latest \
+    /bin/bash
+

4. 注意事项

不同环境下，MinerU对Hygon加速卡的支持情况如下表所示：

+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

使用场景		容器环境
使用场景		vllm
命令行工具(mineru)	pipeline	🟢
	<vlm/hybrid>-auto-engine	🟢
	<vlm/hybrid>-http-client	🟢
fastapi服务(mineru-api)	pipeline	🟢
	<vlm/hybrid>-auto-engine	🟢
	<vlm/hybrid>-http-client	🟢
gradio界面(mineru-gradio)	pipeline	🟢
	<vlm/hybrid>-auto-engine	🟢
	<vlm/hybrid>-http-client	🟢
openai-server服务（mineru-openai-server）		🟢
数据并行 (--data-parallel-size)		🟢

+ +

Tip

DCU加速卡指定可用加速卡的方式与AMD GPU类似，请参考GPU isolation techniques
在Hygon平台可以通过hy-smi命令查看加速卡的使用情况，并根据需要指定空闲的加速卡ID以避免资源冲突。

+ + + + + + + + + + + + + + + +

+ + + + + +

+ + + +

+ + + + + + + + + + \ No newline at end of file diff --git a/zh/usage/acceleration_cards/IluvatarCorex/index.html b/zh/usage/acceleration_cards/IluvatarCorex/index.html new file mode 100644 index 00000000..706cd3c2 --- /dev/null +++ b/zh/usage/acceleration_cards/IluvatarCorex/index.html @@ -0,0 +1,1746 @@ + + + + + + + + + + + + + + + + + + + + + IluvatarCorex - MinerU + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

+ + + + 跳转至 + + +

+ +

+ + + + +

+ + +

+ + + + + + +

+ + MinerU + +

+ + + IluvatarCorex + + +

+ + + + + + + + + +

+ + +

+ + English + +
+ + 中文 + +

+ + + + + + + + + +

+ +

+ + +

+ opendatalab/MinerU +

+ +

+ + + + + +

+ + + + + + 主页 + + +
+ + + + + + 快速入门 + + +
+ + + + + + 使用指南 + + +
+ + + + + + 参考资料 + + +
+ + + + + + 常见问题解答 + + +
+ + + + + + 在线演示 + + +

+ + + +

+ +

+ + + + +

+ + + + + + + +

IluvatarCorex

+ +

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
+gpu: 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

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

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_VLLM_DEVICE=corex \
+   -it mineru:corex-vllm-latest \
+   /bin/bash
+

4. 注意事项

Tip

目前Iluvatar方案使用vllm作为推理引擎时，可能出现服务停止后显存无法正常释放的问题，如果遇到该问题，请重启Docker容器以释放显存。

不同环境下，MinerU对Iluvatar加速卡的支持情况如下表所示：

+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

使用场景		容器环境
使用场景		vllm
命令行工具(mineru)	pipeline	🟢
	<vlm/hybrid>-auto-engine	🟢
	<vlm/hybrid>-http-client	🟢
fastapi服务(mineru-api)	pipeline	🟢
	<vlm/hybrid>-auto-engine	🟢
	<vlm/hybrid>-http-client	🟢
gradio界面(mineru-gradio)	pipeline	🟢
	<vlm/hybrid>-auto-engine	🟢
	<vlm/hybrid>-http-client	🟢
openai-server服务（mineru-openai-server）		🟢
数据并行 (--data-parallel-size)		🟢

+ +

Tip

Iluvatar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设备章节说明
在Iluvatar平台可以通过ixsmi命令查看加速卡的使用情况，并根据需要指定空闲的加速卡ID以避免资源冲突。

+ + + + + + + + + + + + + + + +

+ + + + + +

+ + + +

+ + + + + + + + + + \ No newline at end of file diff --git a/zh/usage/acceleration_cards/Kunlunxin/index.html b/zh/usage/acceleration_cards/Kunlunxin/index.html new file mode 100644 index 00000000..a2e06871 --- /dev/null +++ b/zh/usage/acceleration_cards/Kunlunxin/index.html @@ -0,0 +1,1747 @@ + + + + + + + + + + + + + + + + + + + + + Kunlunxin - MinerU + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

+ + + + 跳转至 + + +

+ +

+ + + + +

+ + +

+ + + + + + +

+ + MinerU + +

+ + + Kunlunxin + + +

+ + + + + + + + + +

+ + +

+ + English + +
+ + 中文 + +

+ + + + + + + + + +

+ +

+ + +

+ opendatalab/MinerU +

+ +

+ + + + + +

+ + + + + + 主页 + + +
+ + + + + + 快速入门 + + +
+ + + + + + 使用指南 + + +
+ + + + + + 参考资料 + + +
+ + + + + + 常见问题解答 + + +
+ + + + + + 在线演示 + + +

+ + + +

+ +

+ + + + +

+ + + + + + + +

Kunlunxin

+ +

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
+xpu: P800
+driver: 515.58
+docker: 20.10.5
+

2. 环境准备

2.1 使用 Dockerfile 构建镜像（vllm）

wget https://gcore.jsdelivr.net/gh/opendatalab/MinerU@master/docker/china/kxpu.Dockerfile
+docker build --network=host -t mineru:kxpu-vllm-latest -f kxpu.Dockerfile .
+

3. 启动 Docker 容器

docker run -u root --name mineru_docker \
+    --device=/dev/xpu0:/dev/xpu0 \
+    --device=/dev/xpu1:/dev/xpu1 \
+    --device=/dev/xpu2:/dev/xpu2 \
+    --device=/dev/xpu3:/dev/xpu3 \
+    --device=/dev/xpu4:/dev/xpu4 \
+    --device=/dev/xpu5:/dev/xpu5 \
+    --device=/dev/xpu6:/dev/xpu6 \
+    --device=/dev/xpu7:/dev/xpu7 \
+    --device=/dev/xpuctrl:/dev/xpuctrl \
+    --net=host \
+    --cap-add=SYS_PTRACE --security-opt seccomp=unconfined \
+    --tmpfs /dev/shm:rw,nosuid,nodev,exec,size=32g \
+    --cap-add=SYS_PTRACE \
+    -v /home/users/vllm-kunlun:/home/vllm-kunlun \
+    -v /usr/local/bin/xpu-smi:/usr/local/bin/xpu-smi \
+    -w /workspace \
+    -e MINERU_MODEL_SOURCE=local \
+    -e MINERU_FORMULA_CH_SUPPORT=true \
+    -e MINERU_VLLM_DEVICE=kxpu \
+    -it mineru:kxpu-vllm-latest \
+    /bin/bash
+

4. 注意事项

不同环境下，MinerU对Kunlunxin加速卡的支持情况如下表所示：

+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

使用场景		容器环境
使用场景		vllm
命令行工具(mineru)	pipeline	🟢
	<vlm/hybrid>-auto-engine	🟢
	<vlm/hybrid>-http-client	🟢
fastapi服务(mineru-api)	pipeline	🟢
	<vlm/hybrid>-auto-engine	🟢
	<vlm/hybrid>-http-client	🟢
gradio界面(mineru-gradio)	pipeline	🟢
	<vlm/hybrid>-auto-engine	🟢
	<vlm/hybrid>-http-client	🟢
openai-server服务（mineru-openai-server）		🟢
数据并行 (--data-parallel-size)		🔴

+ +

Tip

Kunlunxin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设备章节说明, +将环境变量CUDA_VISIBLE_DEVICES替换为XPU_VISIBLE_DEVICES即可。
在Kunlunxin平台可以通过xpu-smi命令查看加速卡的使用情况，并根据需要指定空闲的加速卡ID以避免资源冲突。

+ + + + + + + + + + + + + + + +

+ + + + + +

+ + + +

+ + + + + + + + + + \ No newline at end of file diff --git a/zh/usage/acceleration_cards/METAX/index.html b/zh/usage/acceleration_cards/METAX/index.html new file mode 100644 index 00000000..d5f187f6 --- /dev/null +++ b/zh/usage/acceleration_cards/METAX/index.html @@ -0,0 +1,1778 @@ + + + + + + + + + + + + + + + + + + + + + METAX - MinerU + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

+ + + + 跳转至 + + +

+ +

+ + + + +

+ + +

+ + + + + + +

+ + MinerU + +

+ + + METAX + + +

+ + + + + + + + + +

+ + +

+ + English + +
+ + 中文 + +

+ + + + + + + + + +

+ +

+ + +

+ opendatalab/MinerU +

+ +

+ + + + + +

+ + + + + + 主页 + + +
+ + + + + + 快速入门 + + +
+ + + + + + 使用指南 + + +
+ + + + + + 参考资料 + + +
+ + + + + + 常见问题解答 + + +
+ + + + + + 在线演示 + + +

+ + + +

+ +

+ + + + +

+ + + + + + + +

METAX

+ +

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   
+cpu: INTEL x86_64
+gpu: C500  
+driver: 2.12.13
+docker: 28.1.1
+

2. 环境准备

Note

maca加速卡支持使用vllm或lmdeploy进行VLM模型推理加速。请根据实际需求选择安装和使用其中之一:

2.1 使用metax官方镜像作为基础镜像构建vllm环境镜像

从metax官方仓库拉取基础镜像
- 1.1 镜像获取地址：https://developer.metax-tech.com/softnova/docker
- 1.2 在镜像网站选择AI分类，软件包类型选择vllm，操作系统选择ubuntu
- 1.3 找到vllm:maca.ai3.1.0.7-torch2.6-py310-ubuntu22.04-amd64镜像，复制拉取命令并在本地终端执行
+

使用 Dockerfile 构建镜像（vllm） +

wget https://gcore.jsdelivr.net/gh/opendatalab/MinerU@master/docker/china/maca.Dockerfile
+docker build --network=host -t mineru:maca-vllm-latest -f maca.Dockerfile .
+

2.2 使用 Dockerfile 构建镜像（lmdeploy）

wget https://gcore.jsdelivr.net/gh/opendatalab/MinerU@master/docker/china/maca.Dockerfile
+# 将基础镜像从 vllm 切换为 lmdeploy
+sed -i '3s/^/# /' maca.Dockerfile && sed -i '5s/^# //' maca.Dockerfile
+docker build --network=host -t mineru:maca-lmdeploy-latest -f maca.Dockerfile .
+

3. 启动 Docker 容器

docker run --ipc host \
+   --cap-add SYS_PTRACE \
+   --privileged=true \
+   --device=/dev/mem \
+   --device=/dev/dri \
+   --device=/dev/mxcd \
+   --device=/dev/infiniband \
+   --group-add video \
+   --network=host \
+   --shm-size '100gb' \
+   --ulimit memlock=-1 \
+   --security-opt seccomp=unconfined \
+   --security-opt apparmor=unconfined \
+   --name mineru_docker \
+   -v /datapool:/datapool \
+   -e MINERU_MODEL_SOURCE=local \
+   -e MINERU_LMDEPLOY_DEVICE=maca \
+   -it mineru:maca-vllm-latest \
+   /bin/bash
+

Tip

请根据实际情况选择使用vllm或lmdeploy版本的镜像，如需使用lmdeploy，替换上述命令中的mineru:maca-vllm-latest为mineru:maca-lmdeploy-latest即可。

4. 注意事项

不同环境下，MinerU对maca加速卡的支持情况如下表所示：

+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

使用场景		容器环境
使用场景		vllm	lmdeploy
命令行工具(mineru)	pipeline	🟢	🟢
	<vlm/hybrid>-auto-engine	🟢	🟢
	<vlm/hybrid>-http-client	🟢	🟢
fastapi服务(mineru-api)	pipeline	🟢	🟢
	<vlm/hybrid>-auto-engine	🟢	🟢
	<vlm/hybrid>-http-client	🟢	🟢
gradio界面(mineru-gradio)	pipeline	🟢	🟢
	<vlm/hybrid>-auto-engine	🟢	🟢
	<vlm/hybrid>-http-client	🟢	🟢
openai-server服务（mineru-openai-server）		🟢	🟢
数据并行 (--data-parallel-size/--dp)		🔴	🔴

+ +

Tip

MACA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设备章节说明。
在METAX平台可以通过mx-smi命令查看加速卡的使用情况，并根据需要指定空闲的加速卡ID以避免资源冲突。

+ + + + + + + + + + + + + + + +

+ + + + + +

+ + + +

+ + + + + + + + + + \ No newline at end of file diff --git a/zh/usage/acceleration_cards/MooreThreads/index.html b/zh/usage/acceleration_cards/MooreThreads/index.html new file mode 100644 index 00000000..d46e3c81 --- /dev/null +++ b/zh/usage/acceleration_cards/MooreThreads/index.html @@ -0,0 +1,1740 @@ + + + + + + + + + + + + + + + + + + + + + MooreThreads - MinerU + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

+ + + + 跳转至 + + +

+ +

+ + + + +

+ + +

+ + + + + + +

+ + MinerU + +

+ + + MooreThreads + + +

+ + + + + + + + + +

+ + +

+ + English + +
+ + 中文 + +

+ + + + + + + + + +

+ +

+ + +

+ opendatalab/MinerU +

+ +

+ + + + + +

+ + + + + + 主页 + + +
+ + + + + + 快速入门 + + +
+ + + + + + 使用指南 + + +
+ + + + + + 参考资料 + + +
+ + + + + + 常见问题解答 + + +
+ + + + + + 在线演示 + + +

+ + + +

+ +

+ + + + +

+ + + + + + + +

MooreThreads

+ +

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 22.04.4 LTS  
+cpu: Intel x86-64
+dcu: MTT S4000
+driver: 3.0.0-rc-KuaE2.0
+docker: 24.0.7
+

2. 环境准备

2.1 使用 Dockerfile 构建镜像

wget https://gcore.jsdelivr.net/gh/opendatalab/MinerU@master/docker/china/musa.Dockerfile
+docker build --network=host -t mineru:musa-vllm-latest -f musa.Dockerfile .
+

3. 启动 Docker 容器

docker run -u root --name mineru_docker \
+    --network=host \
+    --ipc=host \
+    --shm-size=80g \
+    --privileged \
+    -e MTHREADS_VISIBLE_DEVICES=all \
+    -e MINERU_VLLM_DEVICE=musa \
+    -e MINERU_MODEL_SOURCE=local \
+    -it mineru:musa-vllm-latest \
+    /bin/bash
+

4. 注意事项

不同环境下，MinerU对MooreThreads加速卡的支持情况如下表所示：

Note

兼容性说明：由于摩尔线程（MooreThreads）目前对 vLLM v1 引擎的支持尚待完善，MinerU 现阶段采用 v0 引擎作为适配方案。 +受此限制，vLLM 的异步引擎（Async Engine）功能存在兼容性问题，可能导致部分使用场景无法正常运行。 +我们将持续跟进摩尔线程对 vLLM v1 引擎的支持进展，并及时在 MinerU 中进行相应的适配与优化。

+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

使用场景		容器环境
使用场景		vllm
命令行工具(mineru)	pipeline	🟢
	<vlm/hybrid>-auto-engine	🟢
	<vlm/hybrid>-http-client	🟢
fastapi服务(mineru-api)	pipeline	🟢
	<vlm/hybrid>-auto-engine	🔴
	<vlm/hybrid>-http-client	🟢
gradio界面(mineru-gradio)	pipeline	🟢
	<vlm/hybrid>-auto-engine	🔴
	<vlm/hybrid>-http-client	🟢
openai-server服务（mineru-openai-server）		🟢
数据并行 (--data-parallel-size)		🔴

+ +

Tip

MooreThreads加速卡指定可用加速卡的方式与NVIDIA GPU类似，请参考GPU 枚举
在MooreThreads平台可以通过mthreads-gmi命令查看加速卡的使用情况，并根据需要指定空闲的加速卡ID以避免资源冲突。

+ + + + + + + + + + + + + + + +

+ + + + + +

+ + + +

+ + + + + + + + + + \ No newline at end of file diff --git a/zh/usage/acceleration_cards/THead/index.html b/zh/usage/acceleration_cards/THead/index.html new file mode 100644 index 00000000..a9b47115 --- /dev/null +++ b/zh/usage/acceleration_cards/THead/index.html @@ -0,0 +1,1765 @@ + + + + + + + + + + + + + + + + + + + + + THead - MinerU + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

+ + + + 跳转至 + + +

+ +

+ + + + +

+ + +

+ + + + + + +

+ + MinerU + +

+ + + THead + + +

+ + + + + + + + + +

+ + +

+ + English + +
+ + 中文 + +

+ + + + + + + + + +

+ +

+ + +

+ opendatalab/MinerU +

+ +

+ + + + + +

+ + + + + + 主页 + + +
+ + + + + + 快速入门 + + +
+ + + + + + 使用指南 + + +
+ + + + + + 参考资料 + + +
+ + + + + + 常见问题解答 + + +
+ + + + + + 在线演示 + + +

+ + + +

+ +

+ + + + +

+ + + + + + + +

THead

+ +

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   
+cpu: INTEL x86_64
+ppu: ZW810E  
+driver: 1.4.0
+docker: 26.1.4
+

2. 环境准备

Note

ppu加速卡支持使用vllm或lmdeploy进行VLM模型推理加速。请根据实际需求选择安装和使用其中之一:

2.1 使用 Dockerfile 构建镜像（vllm）

wget https://gcore.jsdelivr.net/gh/opendatalab/MinerU@master/docker/china/ppu.Dockerfile
+docker build --network=host -t mineru:ppu-vllm-latest -f ppu.Dockerfile .
+

2.2 使用 Dockerfile 构建镜像（lmdeploy）

wget https://gcore.jsdelivr.net/gh/opendatalab/MinerU@master/docker/china/ppu.Dockerfile
+# 将基础镜像从 vllm 切换为 lmdeploy
+sed -i '3s/^/# /' ppu.Dockerfile && sed -i '5s/^# //' ppu.Dockerfile
+docker build --network=host -t mineru:ppu-lmdeploy-latest -f ppu.Dockerfile .
+

3. 启动 Docker 容器

docker run --privileged=true \
+  --name mineru_docker \
+  --device=/dev/alixpu \
+  --device=/dev/alixpu_ctl \
+  --ipc=host \
+  --network=host \
+  --ulimit memlock=-1 \
+  --ulimit stack=67108864 \
+  --shm-size=500g \
+  -v /mnt:/mnt \
+  -v /datapool:/datapool \
+  -v /var/run/docker.sock:/var/run/docker.sock \
+  -e MINERU_MODEL_SOURCE=local \
+  -it mineru:ppu-vllm-latest \
+  /bin/bash
+

Tip

请根据实际情况选择使用vllm或lmdeploy版本的镜像，如需使用lmdeploy，替换上述命令中的mineru:ppu-vllm-latest为mineru:ppu-lmdeploy-latest即可。

4. 注意事项

不同环境下，MinerU对ppu加速卡的支持情况如下表所示：

+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

使用场景		容器环境
使用场景		vllm	lmdeploy
命令行工具(mineru)	pipeline	🟢	🟢
	<vlm/hybrid>-auto-engine	🟢	🟢
	<vlm/hybrid>-http-client	🟢	🟢
fastapi服务(mineru-api)	pipeline	🟢	🟢
	<vlm/hybrid>-auto-engine	🟢	🟢
	<vlm/hybrid>-http-client	🟢	🟢
gradio界面(mineru-gradio)	pipeline	🟢	🟢
	<vlm/hybrid>-auto-engine	🟢	🟢
	<vlm/hybrid>-http-client	🟢	🟢
openai-server服务（mineru-openai-server）		🟢	🟢
数据并行 (--data-parallel-size/--dp)		🔴	🔴

+ +

Tip

PPU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设备章节说明。
在T-Head平台可以通过ppu-smi命令查看加速卡的使用情况，并根据需要指定空闲的加速卡ID以避免资源冲突。

+ + + + + + + + + + + + + + + +

+ + + + + +

+ + + +

+ + + + + + + + + + \ No newline at end of file diff --git a/zh/usage/acceleration_cards/Tecorigin/index.html b/zh/usage/acceleration_cards/Tecorigin/index.html new file mode 100644 index 00000000..48d014fe --- /dev/null +++ b/zh/usage/acceleration_cards/Tecorigin/index.html @@ -0,0 +1,1744 @@ + + + + + + + + + + + + + + + + + + + + + Tecorigin - MinerU + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

+ + + + 跳转至 + + +

+ +

+ + + + +

+ + +

+ + + + + + +

+ + MinerU + +

+ + + Tecorigin + + +

+ + + + + + + + + +

+ + +

+ + English + +
+ + 中文 + +

+ + + + + + + + + +

+ +

+ + +

+ opendatalab/MinerU +

+ +

+ + + + + +

+ + + + + + 主页 + + +
+ + + + + + 快速入门 + + +
+ + + + + + 使用指南 + + +
+ + + + + + 参考资料 + + +
+ + + + + + 常见问题解答 + + +
+ + + + + + 在线演示 + + +

+ + + +

+ +

+ + + + +

+ + + + + + + +

Tecorigin

+ +

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: AMD EPYC (amd64)
+gpu: T100
+driver: 3.0.0
+docker: 28.0.4
+

2. 环境准备

2.1 下载并加载镜像（vllm）

wget http://wb.tecorigin.com:8082/repository/teco-customer-repo/Course/MinerU/mineru-vllm.tar
+
+docker load -i mineru-vllm.tar
+

3. 启动 Docker 容器

docker run -dit --name mineru_docker \
+    --privileged \
+    --cap-add SYS_PTRACE \
+    --cap-add SYS_ADMIN \
+    --network=host \
+    --shm-size=500G \
+    mineru:sdaa-vllm-latest \
+    /bin/bash
+

Tip

如需使用vllm环境,请执行以下操作： +- 进入容器后，通过以下命令切换到conda环境： +

conda activate vllm_env_py310
+

切换成功后，您可以在命令行前看到(vllm_env_py310)的标识，这表示您已成功进入vllm的虚拟环境。

4. 注意事项

不同环境下，MinerU对Tecorigin加速卡的支持情况如下表所示：

+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

使用场景		容器环境
使用场景		vllm
命令行工具(mineru)	pipeline	🟢
	<vlm/hybrid>-auto-engine	🟢
	<vlm/hybrid>-http-client	🟢
fastapi服务(mineru-api)	pipeline	🟢
	<vlm/hybrid>-auto-engine	🟢
	<vlm/hybrid>-http-client	🟢
gradio界面(mineru-gradio)	pipeline	🟢
	<vlm/hybrid>-auto-engine	🟢
	<vlm/hybrid>-http-client	🟢
openai-server服务（mineru-openai-server）		🟢
数据并行 (--data-parallel-size)		🔴

+ +

Tip

Tecorigin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设备章节说明, +将环境变量CUDA_VISIBLE_DEVICES替换为SDAA_VISIBLE_DEVICES即可。
在太初平台可以通过teco-smi -c命令查看加速卡的使用情况，并根据需要指定空闲的加速卡ID以避免资源冲突。

+ + + + + + + + + + + + + + + +

+ + + + + +

+ + + +

+ + + + + + + + + + \ No newline at end of file diff --git a/zh/usage/acceleration_cards/VastAI/index.html b/zh/usage/acceleration_cards/VastAI/index.html new file mode 100644 index 00000000..df57bbb9 --- /dev/null +++ b/zh/usage/acceleration_cards/VastAI/index.html @@ -0,0 +1,1866 @@ + + + + + + + + + + + + + + + + + + + + + VastAI - MinerU + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

+ + + + 跳转至 + + +

+ +

+ + + + +

+ + +

+ + + + + + +

+ + MinerU + +

+ + + VastAI + + +

+ + + + + + + + + +

+ + +

+ + English + +
+ + 中文 + +

+ + + + + + + + + +

+ +

+ + +

+ opendatalab/MinerU +

+ +

+ + + + + +

+ + + + + + 主页 + + +
+ + + + + + 快速入门 + + +
+ + + + + + 使用指南 + + +
+ + + + + + 参考资料 + + +
+ + + + + + 常见问题解答 + + +
+ + + + + + 在线演示 + + +

+ + + +

+ +

+ + + + +

+ + + + + + + +

VastAI

+ +

1. 瀚博半导体

官方网址：https://www.vastaitech.com
模型中心：https://github.com/Vastai/VastModelZOO

2. 测试平台

以下为本指南测试使用的平台信息，供参考 +

os: Ubuntu-22.04.3-LTS-x86_64
+cpu: Hygon C86-4G
+gpu: VA16 / VA1L / VA10L
+torch: 2.8.0+cpu
+torch-vacc: 1.3.3.777
+vllm: 0.11.1.dev0+gb8b302cde.d20251030.cpu
+vllm-vacc: 0.11.0.777
+driver: 00.25.12.30 d3_3_v2_9_a3_1 a76bf37 20251230
+docker: 28.1.1
+

3. 环境准备

获取vllm_vacc基础镜像 +

sudo docker pull harbor.vastaitech.com/ai_deliver/vllm_vacc:VVI-25.12.SP2
+

启动容器 +

sudo docker run -it \
+    --privileged=true \
+    --shm-size=256g \
+    --name vllm_service \
+    --ipc=host \
+    --network=host \
+    harbor.vastaitech.com/ai_deliver/vllm_vacc:VVI-25.12.SP2 bash
+

+
安装MinerU
+

参考官方文档安装：README_zh-CN.md#安装-mineru

```bash
+# 启动容器
+# sudo docker exec -it vllm_service bash
+
+# 可选pypi源
+# https://mirrors.163.com/pypi/simple/
+# https://mirrors.aliyun.com/pypi/simple/
+# https://pypi.mirrors.ustc.edu.cn/simple/
+# https://pypi.tuna.tsinghua.edu.cn/simple/
+# https://mirror.baidu.com/pypi/simple
+
+# 通过源码安装MinerU
+git clone https://github.com/opendatalab/MinerU.git
+git checkout 8c4b3ef3a20b11ddac9903f25124d24ea82639b5
+pip install -e .[core] -i https://mirrors.aliyun.com/pypi/simple
+
+# 或使用pip安装MinerU
+pip install -U "mineru[core]==2.7.0" -i https://mirrors.aliyun.com/pypi/simple
+```
+

Note

vllm_vacc基础镜像内已包含torch/vllm等相关依赖
截至2025/12/31，VastAI已支持MinerU至最新版本2.7.0，master分支8c4b3ef3
和NVIDIA硬件下CUDA_VISIBLE_DEVICES类似；在VastAI硬件中可以使用VACC_VISIBLE_DEVICES指定可见计算卡ID，如-e VACC_VISIBLE_DEVICES=0,1,2,3
需指定适当的--shm-size虚拟内存

4. MinerU功能

Note

VastAI加速卡仅支持使用vlm-auto-engine和vlm-http-client形式进行VLM模型推理加速

进入容器 +

sudo docker exec -it vllm_service bash
+

使用MinerU

+
模型准备，参考官方介绍：model_source.md
+

方式一：vlm-auto-engine

export MINERU_MODEL_SOURCE=modelscope
+
+# step1, 以`vlm-auto-engine`方式启动MinerU解析任务
+mineru -p image.png \
+-o ./output \
+-b vlm-auto-engine \
+--http-timeout 1200 \
+--tensor-parallel-size 2 \
+--enforce_eager \
+--trust-remote-code \
+--max-model-len 16384
+

方式二：vlm-http-client

# step1, 启动vLLM API server
+vllm serve /root/.cache/modelscope/hub/models/OpenDataLab/MinerU2.5-2509-1.2B \
+--tensor-parallel-size 2 \
+--trust-remote-code \
+--enforce_eager \
+--port 8090 \
+--max-model-len 16384 \
+--served-model-name MinerU2.5-2509-1.2B
+
+# step2，以`vlm-http-client`方式启动MinerU解析任务
+mineru -p demo/pdfs/demo1.pdf \
+-o ./output \
+-b vlm-http-client \
+-u http://127.0.0.1:8090 \
+--http-timeout 1200
+

Note

注意在执行任意与vllm相关命令需追加--enforce_eager参数

5. 注意事项

VastAI加速卡对MinerU的支持情况如下表所示：

+ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

使用场景		支持情况
使用场景		命令行工具(mineru)	pipeline	🔴
hybrid-http-client	🔴
hybrid-auto-engine	🔴
vlm-auto-engine	🟢
vlm-http-client	🟢
fastapi服务(mineru-api)	pipeline	🔴
	hybrid-http-client	🔴
	hybrid-auto-engine	🔴
	vlm-auto-engine	🟢
	vlm-http-client	🟢
gradio界面(mineru-gradio)	pipeline	🔴
	hybrid-http-client	🔴
	hybrid-auto-engine	🔴
	vlm-auto-engine	🟢
	vlm-http-client	🟢
openai-server服务（mineru-openai-server）		🟢
Tensor并行 (--tensor-parallel-size)		🟢
数据并行 (--data-parallel-size)		🔴

+ +

Note

🟢: 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
vlm-auto-engine：VastAI仅支持vLLM后端

+ + + + + + + + + + + + + + + +

+ + + + + +

+ + + +

+ + + + + + + + + + \ No newline at end of file diff --git a/zh/usage/advanced_cli_parameters/index.html b/zh/usage/advanced_cli_parameters/index.html new file mode 100644 index 00000000..ea4b9514 --- /dev/null +++ b/zh/usage/advanced_cli_parameters/index.html @@ -0,0 +1,1967 @@ + + + + + + + + + + + + + + + + + + + + + + + + + 命令行进阶参数 - MinerU + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

+ + + + 跳转至 + + +

+ +

+ + + + +

+ + +

+ + + + + + +

+ + MinerU + +

+ + + 命令行进阶参数 + + +

+ + + + + + + + + +

+ + +

+ + English + +
+ + 中文 + +

+ + + + + + + + + +

+ +

+ + +

+ opendatalab/MinerU +

+ +

+ + + + + +

+ + + + + + 主页 + + +
+ + + + + + 快速入门 + + +
+ + + + + + 使用指南 + + +
+ + + + + + 参考资料 + + +
+ + + + + + 常见问题解答 + + +
+ + + + + + 在线演示 + + +

+ + + +

+ +

+ + + + +

+ + + + + + + +

命令行参数进阶

推理引擎参数透传

vllm 加速参数优化

Tip

如果您已经可以正常使用vllm对vlm模型进行加速推理，但仍然希望进一步提升推理速度，可以尝试以下参数：

如果您有超过多张显卡，可以使用vllm的多卡并行模式来增加吞吐量：--data-parallel-size 2

参数传递说明

Tip

所有vllm/lmdeploy官方支持的参数都可用通过命令行参数传递给 MinerU，包括以下命令:mineru、mineru-openai-server、mineru-gradio、mineru-api
如果您想了解更多有关vllm的参数使用方法，请参考 vllm官方文档
如果您想了解更多有关lmdeploy的参数使用方法，请参考 lmdeploy官方文档

GPU 设备选择与配置

CUDA_VISIBLE_DEVICES 基本用法

Tip

任何情况下，您都可以通过在命令行的开头添加CUDA_VISIBLE_DEVICES 环境变量来指定可见的 GPU 设备： +
```
CUDA_VISIBLE_DEVICES=1 mineru -p <input_path> -o <output_path>
+
```
这种指定方式对所有的命令行调用都有效，包括 mineru、mineru-openai-server、mineru-gradio 和 mineru-api，且对pipeline、vlm后端均适用。

常见设备配置示例

Tip

以下是一些常见的 CUDA_VISIBLE_DEVICES 设置示例： +

CUDA_VISIBLE_DEVICES=1  # Only device 1 will be seen
+CUDA_VISIBLE_DEVICES=0,1  # Devices 0 and 1 will be visible
+CUDA_VISIBLE_DEVICES="0,1"  # Same as above, quotation marks are optional
+CUDA_VISIBLE_DEVICES=0,2,3  # Devices 0, 2, 3 will be visible; device 1 is masked
+CUDA_VISIBLE_DEVICES=""  # No GPU will be visible
+

实际应用场景

Tip

以下是一些可能的使用场景：

+
如果您有多张显卡，需要指定卡0和卡1，并使用多卡并行来启动openai-server，可以使用以下命令： +
```
CUDA_VISIBLE_DEVICES=0,1 mineru-openai-server --engine vllm --port 30000 --data-parallel-size 2
+
```
+

如果您有多张显卡，需要在卡0和卡1上启动两个fastapi服务，并分别监听不同的端口，可以使用以下命令： +

# 在终端1中
+CUDA_VISIBLE_DEVICES=0 mineru-api --host 127.0.0.1 --port 8000
+# 在终端2中
+CUDA_VISIBLE_DEVICES=1 mineru-api --host 127.0.0.1 --port 8001
+

+ + + + + + + + + + + + + + + +

+ + + + + +

+ + + +

+ + + + + + + + + + \ No newline at end of file diff --git a/zh/usage/cli_tools/index.html b/zh/usage/cli_tools/index.html new file mode 100644 index 00000000..5cb81bcb --- /dev/null +++ b/zh/usage/cli_tools/index.html @@ -0,0 +1,1973 @@ + + + + + + + + + + + + + + + + + + + + + + + + + 命令行工具 - MinerU + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

+ + + + 跳转至 + + +

+ +

+ + + + +

+ + +

+ + + + + + +

+ + MinerU + +

+ + + 命令行工具 + + +

+ + + + + + + + + +

+ + +

+ + English + +
+ + 中文 + +

+ + + + + + + + + +

+ +

+ + +

+ opendatalab/MinerU +

+ +

+ + + + + +

+ + + + + + 主页 + + +
+ + + + + + 快速入门 + + +
+ + + + + + 使用指南 + + +
+ + + + + + 参考资料 + + +
+ + + + + + 常见问题解答 + + +
+ + + + + + 在线演示 + + +

+ + + +

+ +

+ + + + +

+ + + + + + + +

命令行工具使用说明

查看帮助信息

要查看 MinerU 命令行工具的帮助信息，可以使用 --help 参数。以下是各个命令行工具的帮助信息示例： +

mineru --help
+Usage: mineru [OPTIONS]
+
+Options:
+  -v, --version                   显示版本并退出
+  -p, --path PATH                 输入文件路径或目录（必填）
+  -o, --output PATH               输出目录（必填）
+  --api-url TEXT                  MinerU FastAPI 服务地址；不传时自动拉起本地临时 mineru-api
+  -m, --method [auto|txt|ocr]     解析方法：auto（默认）、txt、ocr（仅用于 pipeline 与 hybrid* 后端）
+  -b, --backend [pipeline|hybrid-auto-engine|hybrid-http-client|vlm-auto-engine|vlm-http-client]
+                                  解析后端（默认为 hybrid-auto-engine）
+  -l, --lang [ch|ch_server|ch_lite|en|korean|japan|chinese_cht|ta|te|ka|th|el|latin|arabic|east_slavic|cyrillic|devanagari]
+                                  指定文档语言（可提升 OCR 准确率，仅用于 pipeline 与 hybrid* 后端）
+  -u, --url TEXT                  当使用 http-client 时，传给服务端后端的 OpenAI 兼容地址
+  -s, --start INTEGER             开始解析的页码（从 0 开始）
+  -e, --end INTEGER               结束解析的页码（从 0 开始）
+  -f, --formula BOOLEAN           是否启用公式解析（默认开启）
+  -t, --table BOOLEAN             是否启用表格解析（默认开启）
+  --help                          显示帮助信息
+

mineru-api --help
+Usage: mineru-api [OPTIONS]
+
+Options:
+  --host TEXT     服务器主机地址（默认：127.0.0.1）
+  --port INTEGER  服务器端口（默认：8000）
+  --reload        启用自动重载（开发模式）
+  --help          显示此帮助信息并退出
+

mineru-gradio --help
+Usage: mineru-gradio [OPTIONS]
+
+Options:
+  --enable-example BOOLEAN        启用示例文件输入(需要将示例文件放置在当前
+                                  执行命令目录下的 `example` 文件夹中)
+  --enable-http-client BOOLEAN    在后端选项中启用 HTTP 客户端选项
+  --enable-api BOOLEAN            启用 Gradio API 以提供应用程序服务
+  --max-convert-pages INTEGER     设置从 PDF 转换为 Markdown 的最大页数
+  --server-name TEXT              设置 Gradio 应用程序的服务器主机名
+  --server-port INTEGER           设置 Gradio 应用程序的服务器端口
+  --latex-delimiters-type [a|b|all]
+                                  设置在 Markdown 渲染中使用的 LaTeX 分隔符类型
+                                  ('a' 表示 '$' 类型，'b' 表示 '()[]' 类型，
+                                  'all' 表示两种类型都使用)
+  --help                          显示此帮助信息并退出
+

环境变量说明

Note

从当前版本开始，mineru 是基于 mineru-api 的编排客户端： +- 未传 --api-url 时，CLI 会自动拉起本地临时 mineru-api +- 传入 --api-url 时，CLI 会直连该 FastAPI 服务 +- --url 不再表示 MinerU API 地址，而是服务端 vlm/hybrid-http-client 所需的 OpenAI 兼容地址

MinerU命令行工具的某些参数存在相同功能的环境变量配置，通常环境变量配置的优先级高于命令行参数，且在所有命令行工具中都生效。 +以下是常用的环境变量及其说明：

+
MINERU_TOOLS_CONFIG_JSON：
+
- 用于指定配置文件路径
- 默认为用户目录下的mineru.json，可通过环境变量指定其他配置文件路径。
+
+
MINERU_FORMULA_ENABLE：
+
- 用于启用公式解析
- 默认为true，可通过环境变量设置为false来禁用公式解析。
+
+
MINERU_FORMULA_CH_SUPPORT：
+
- 用于启用中文公式解析优化（实验性功能）
- 默认为false，可通过环境变量设置为true来启用中文公式解析优化。
- 仅对pipeline后端生效。
+
+
MINERU_TABLE_ENABLE：
+
- 用于启用表格解析
- 默认为true，可通过环境变量设置为false来禁用表格解析。
+
+
MINERU_TABLE_MERGE_ENABLE：
+
- 用于启用表格合并功能
- 默认为true，可通过环境变量设置为false来禁用表格合并功能。
+
+
MINERU_PDF_RENDER_TIMEOUT：
+
- 用于设置将PDF渲染为图片的超时时间（秒）
- 默认为300秒，可通过环境变量设置为其他值以调整渲染图片的超时时间。
- 仅在linux和macOS系统中生效。
+
+
MINERU_PDF_RENDER_THREADS：
+
- 用于设置将PDF渲染为图片时使用的线程数
- 默认为4，可通过环境变量设置为其他值以调整渲染图片时的线程数。
- 仅在linux和macOS系统中生效。
+
+
MINERU_INTRA_OP_NUM_THREADS：
+
- 用于设置onnx模型的intra_op线程数，影响单个算子的计算速度
- 默认为-1（自动选择），可通过环境变量设置为其他值以调整线程数。
+
+
MINERU_INTER_OP_NUM_THREADS：
+
- 用于设置onnx模型的inter_op线程数，影响多个算子的并行执行
- 默认为-1（自动选择），可通过环境变量设置为其他值以调整线程数。
+
+
MINERU_HYBRID_BATCH_RATIO：
+
- 用于设置 hybrid-* 后端中小模型处理的batch倍率
- 在hybrid-http-client中较为常用，可以通过控制小模型的batch倍率来调整单个客户端的显存占用量
- + + + + + + + + + + + + + + + + + + + + + + + + + +
  单个client端显存大小 MINERU_HYBRID_BATCH_RATIO
  <= 6 GB 8
  <= 4.5 GB 4
  <= 3 GB 2
  <= 2.5 GB 1
  +
+
+
MINERU_HYBRID_FORCE_PIPELINE_ENABLE：
+
- 用于强制将 hybrid-* 后端中的文本提取部分使用小模型进行处理
- 默认为false，可通过环境变量设置为true来启用该功能，从而在某些极端情况下减少幻觉的发生。
+
+
MINERU_VL_MODEL_NAME：
+
- 用于指定 vlm/hybrid 后端使用的模型名称，这将允许您在同时存在多个模型的远程openai-server中指定 MinerU 运行所需的模型。
+
+
MINERU_VL_API_KEY:
+
- 用于指定 vlm/hybrid 后端使用的API Key，这将允许您在远程openai-server中进行身份验证。
+

+ + + + + + + + + + + + + + + +

单个client端显存大小	MINERU_HYBRID_BATCH_RATIO
<= 6 GB	8
<= 4.5 GB	4
<= 3 GB	2
<= 2.5 GB	1

+ + + + + +

+ + + +

+ + + + + + + + + + \ No newline at end of file diff --git a/zh/usage/index.html b/zh/usage/index.html new file mode 100644 index 00000000..5d9d9b35 --- /dev/null +++ b/zh/usage/index.html @@ -0,0 +1,1723 @@ + + + + + + + + + + + + + + + + + + + + + + + + + 使用指南 - MinerU + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

+ + + + 跳转至 + + +

+ +

+ + + + +

+ + +

+ + + + + + +

+ + MinerU + +

+ + + 使用指南 + + +

+ + + + + + + + + +

+ + +

+ + English + +
+ + 中文 + +

+ + + + + + + + + +

+ +

+ + +

+ opendatalab/MinerU +

+ +

+ + +

+ + + + + + + + + + \ No newline at end of file diff --git a/zh/usage/model_source/index.html b/zh/usage/model_source/index.html new file mode 100644 index 00000000..e9fe7539 --- /dev/null +++ b/zh/usage/model_source/index.html @@ -0,0 +1,1935 @@ + + + + + + + + + + + + + + + + + + + + + + + + + 模型源配置 - MinerU + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

+ + + + 跳转至 + + +

+ +

+ + + + +

+ + +

+ + + + + + +

+ + MinerU + +

+ + + 模型源配置 + + +

+ + + + + + + + + +

+ + +

+ + English + +
+ + 中文 + +

+ + + + + + + + + +

+ +

+ + +

+ opendatalab/MinerU +

+ +

+ + + + + +

+ + + + + + 主页 + + +
+ + + + + + 快速入门 + + +
+ + + + + + 使用指南 + + +
+ + + + + + 参考资料 + + +
+ + + + + + 常见问题解答 + + +
+ + + + + + 在线演示 + + +

+ + + +

+ +

+ + + + +

+ + + + + + + +

模型源说明

MinerU使用 HuggingFace 和 ModelScope 作为模型仓库，用户可以根据需要切换模型源或使用本地模型。

HuggingFace 是默认的模型源，在全球范围内提供了优异的加载速度和极高稳定性。
ModelScope 是中国大陆地区用户的最佳选择，提供了无缝兼容的SDK模块，适用于无法访问HuggingFace的用户。

模型源的切换方法

通过命令行参数切换

目前仅mineru命令行工具支持通过命令行参数切换模型源，其他命令行工具如mineru-api、mineru-gradio等暂不支持。 +

mineru -p <input_path> -o <output_path> --source modelscope
+

通过环境变量切换

在任何情况下可以通过设置环境变量来切换模型源，这适用于所有命令行工具和API调用。 +

export MINERU_MODEL_SOURCE=modelscope
+

+或 +

import os
+os.environ["MINERU_MODEL_SOURCE"] = "modelscope"
+

Tip

通过环境变量设置的模型源会在当前终端会话中生效，直到终端关闭或环境变量被修改。且优先级高于命令行参数，如同时设置了命令行参数和环境变量，命令行参数将被忽略。

使用本地模型

1. 下载模型到本地

mineru-models-download --help
+

+或使用交互式命令行工具选择模型下载： +

mineru-models-download
+

Note

下载完成后，模型路径会在当前终端窗口输出，并自动写入用户目录下的 mineru.json。
您也可以通过将配置模板文件复制到用户目录下并重命名为 mineru.json 来创建配置文件。
模型下载到本地后，您可以自由移动模型文件夹到其他位置，同时需要在 mineru.json 中更新模型路径。
如您将模型文件夹部署到其他服务器上，请确保将 mineru.json文件一同移动到新设备的用户目录中并正确配置模型路径。
如您需要更新模型文件，可以再次运行 mineru-models-download 命令，模型更新暂不支持自定义路径，如您没有移动本地模型文件夹，模型文件会增量更新；如您移动了模型文件夹，模型文件会重新下载到默认位置并更新mineru.json。

2. 使用本地模型进行解析

mineru -p <input_path> -o <output_path> --source local
+

+或通过环境变量启用： +

export MINERU_MODEL_SOURCE=local
+mineru -p <input_path> -o <output_path>
+

+ + + + + + + + + + + + + + + +

+ + + + + +

+ + + +

+ + + + + + + + + + \ No newline at end of file diff --git a/zh/usage/plugin/BISHENG/index.html b/zh/usage/plugin/BISHENG/index.html new file mode 100644 index 00000000..aba95cce --- /dev/null +++ b/zh/usage/plugin/BISHENG/index.html @@ -0,0 +1,1639 @@ + + + + + + + + + + + + + + + + + + + + + BISHENG 简介 - MinerU + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

+ + + + 跳转至 + + +

+ +

+ + + + +

+ + +

+ + + + + + +

+ + MinerU + +

+ + + BISHENG 简介 + + +

+ + + + + + + + + +

+ + +

+ + English + +
+ + 中文 + +

+ + + + + + + + + +

+ +

+ + +

+ opendatalab/MinerU +

+ +

+ + +

+ + + + + + + + + + \ No newline at end of file diff --git a/zh/usage/plugin/Cherry_Studio/index.html b/zh/usage/plugin/Cherry_Studio/index.html new file mode 100644 index 00000000..a66676d1 --- /dev/null +++ b/zh/usage/plugin/Cherry_Studio/index.html @@ -0,0 +1,1759 @@ + + + + + + + + + + + + + + + + + + + + + Cherry Studio 简介 - MinerU + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

+ + + + 跳转至 + + +

+ +

+ + + + +

+ + +

+ + + + + + +

+ + MinerU + +

+ + + Cherry Studio 简介 + + +

+ + + + + + + + + +

+ + +

+ + English + +
+ + 中文 + +

+ + + + + + + + + +

+ +

+ + +

+ opendatalab/MinerU +

+ +

+ + + + + +

+ + + + + + 主页 + + +
+ + + + + + 快速入门 + + +
+ + + + + + 使用指南 + + +
+ + + + + + 参考资料 + + +
+ + + + + + 常见问题解答 + + +
+ + + + + + 在线演示 + + +

+ + + +

+ +

+ + + + +

+ + + + + + + +

Cherry Studio 简介

Cherry Studio 是一款功能强大的多模型 AI 客户端软件，支持 Windows、macOS 和 Linux 等多平台运行，集成了 OpenAI、DeepSeek、Gemini、Anthropic 等主流 AI 云服务，同时支持本地模型运行，用户可以灵活切换不同的AI模型。

目前，MinerU 强大的文档解析能力已深度集成到 Cherry Studio 的知识库与对话交互中，为用户带来更便捷的文档处理与信息获取体验。

Cherry Studio 官网地址：https://www.cherry-ai.com/

MinerU 在 Cherry Studio 中的使用方法

进入 Cherry Studio 设置

a. 打开 Cherry Studio 应用程序

b. 点击左下角的"设置"按钮，进入设置页面

c. 在左侧菜单中，选择"MCP 服务器"

在右侧的 MCP 服务器配置界面中，您可以看到已有的 MCP 服务器列表。点击右上角的"添加服务器"按钮来创建新的 MCP 服务，或者点击现有服务来编辑配置。

添加 MinerU-MCP 配置

点击"添加服务器"后，您将看到一个配置表单。请按以下步骤填写：

a. 名称：输入"MinerU-MCP"或您喜欢的其他名称

b. 描述：可选，如"文档转换为Markdown工具"

c. 类型：选择"标准输入/输出（stdio）"

d. 命令：输入 uvx

e. 参数：输入 mineru-mcp

f. 环境变量：添加以下环境变量

MINERU_API_BASE=https://mineru.net
+MINERU_API_KEY=您的API密钥
+OUTPUT_DIR=./downloads
+USE_LOCAL_API=false
+LOCAL_MINERU_API_BASE=http://localhost:8888
+

使用 uvx 命令可以自动处理 mineru-mcp 的安装和运行，无需预先手动安装 mineru-mcp 包。这是最简单的配置方式。

保存配置

确认无误后，点击界面右上角的"保存"按钮完成配置。保存后，MCP 服务器列表中会显示您刚刚添加的 MinerU-MCP 服务。

使用 Cherry Studio 中的 MinerU MCP

一旦配置完成，您可以在 Cherry Studio 中的对话中使用 MinerU MCP 工具。在 Cherry Studio 中，您可以使用如下提示让模型调用 MinerU MCP 工具。模型会自动识别任务并调用相应的工具。

示例 1: 使用 URL 转换文档

用户输入:

请使用 MinerU MCP 将以下 URL 的 PDF 文档转换为 Markdown 格式：https://example.com/sample.pdf
+

模型将执行的步骤：

模型识别这是文档转换任务，并调用 parse_documents 工具，参数为:

{"file_sources": "https://example.com/sample.pdf"}
+

工具处理完成后，模型会告知您转换结果。

示例 2: 转换本地文档

用户输入:

请使用 MinerU-MCP 将本地的 D://sample.pdf 文件转换为 Markdown 格式
+

模型将执行的步骤：

模型识别这是本地文档转换任务，调用 parse_documents 工具，参数为:

{"file_sources": "D://sample.pdf"}
+

示例 3: 启用 OCR 处理扫描文档

用户输入:

请使用 MinerU-MCP 将以下 URL 的扫描版 PDF 文档转换为 Markdown 格式，并启用 OCR 功能：
+https://example.com/scanned.pdf
+

模型将执行的步骤：

模型识别这是需要 OCR 处理的文档转换任务，调用 parse_documents 工具，并启用 OCR 参数:

{"file_sources": "https://example.com/scanned.pdf","enable_ocr": true}
+

示例 4: 完整对话流程

以下是一个完整的对话流程示例：

用户:

我有一份学术论文的 PDF，网址是 https://arxiv.org/pdf/2303.08774.pdf，能帮我转换成 Markdown 格式吗？
+

模型:

我可以帮您将这份学术论文转换为 Markdown 格式。我将使用 MinerU-MCP 工具来处理这个任务。
+[模型调用 parse_documents 工具，处理过程中...]
+论文已成功转换为 Markdown 格式。转换结果如下：
+# The Capacity of Diffusion Models to Memorize and Generate Training Data
+## Abstract
+Recent diffusion models can generate high-quality images that are nearly indistinguishable from real ones...
+[显示论文内容...]
+

工具参数详解

在使用过程中，模型会根据您的指令自动选择合适的工具和参数。以下是主要工具的参数说明：

● parse_documents 工具参数

● get_ocr_languages 工具参数

无需参数，用于获取OCR支持的语言列表。

高级用法

指定语言和页码范围

用户输入:

请使用 MinerU MCP 将以下 URL 的文档转换为 Markdown 格式，只处理第 5-10 页，并指定语言为中文：https://example.com/document.pdf
+

模型会使用 parse_documents 工具，并设置 language 参数为 "ch"，page_ranges 参数为 "5-10"。

批量处理多个文档

用户输入:

请使用 MinerU-MCP 将以下多个 URL 的文档转换为 Markdown 格式：
+https://example.com/doc1.pdf
+https://example.com/doc2.pdf
+https://example.com/doc3.pdf
+

模型会调用 parse_documents 工具，并将多个 URL 以逗号分隔传入 file_sources 参数。

注意事项

● 当设置 USE_LOCAL_API=true 时，使用本地配置的API进行解析

● 当设置 USE_LOCAL_API=false 时，会使用 MinerU 官网的API进行解析

● 处理大型文档可能需要较长时间，请耐心等待

● 如果遇到超时问题，请考虑分批处理文档或使用本地API模式

常见问题与解决方案

无法启动 MCP 服务

问题：运行 uv run -m mineru.cli时报错。

解决方案：

● 确保已激活虚拟环境

● 检查是否已安装所有依赖

● 尝试使用 python -m mineru.cli命令替代

文件转换失败

问题：文件上传成功但转换失败。

解决方案：

● 检查文件格式是否受支持

● 确认API密钥是否正确

● 查看MCP服务日志获取详细错误信息

文件路径问题

问题：使用 parse_documents 工具处理本地文件时报找不到文件错误。

解决方案：请确保使用绝对路径，或者相对于服务器运行目录的正确相对路径。

MCP 服务调用超时问题

问题：调用 parse_documents 工具时出现 Error calling tool 'parse_documents': MCP error -32001: Request timed out 错误。

解决方案：这个问题常见于处理大型文档或网络不稳定的情况。在某些 MCP 客户端（如 Cursor）中，超时后可能导致无法再次调用 MCP 服务，需要重启客户端。最新版本的 Cursor 中可能会显示正在调用 MCP，但实际上没有真正调用成功。建议：

● 等待官方修复：这是Cursor客户端的已知问题，建议等待Cursor官方修复

● 处理小文件：尽量只处理少量小文件，避免处理大型文档导致超时

● 分批处理：将多个文件分成多次请求处理，每次只处理一两个文件

● 增加超时时间设置（如果客户端支持）

● 对于超时后无法再次调用的问题，需要重启 MCP 客户端

● 如果反复出现超时，请检查网络连接或考虑使用本地 API 模式

+ + + + + + + + + + + + + + + +

+ + + + + +

+ + + +

+ + + + + + + + + + \ No newline at end of file diff --git a/zh/usage/plugin/Coze/index.html b/zh/usage/plugin/Coze/index.html new file mode 100644 index 00000000..af397639 --- /dev/null +++ b/zh/usage/plugin/Coze/index.html @@ -0,0 +1,1686 @@ + + + + + + + + + + + + + + + + + + + + + Coze 简介 - MinerU + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

+ + + + 跳转至 + + +

+ +

+ + + + +

+ + +

+ + + + + + +

+ + MinerU + +

+ + + Coze 简介 + + +

+ + + + + + + + + +

+ + +

+ + English + +
+ + 中文 + +

+ + + + + + + + + +

+ +

+ + +

+ opendatalab/MinerU +

+ +

+ + +

+ + + + + + + + + + \ No newline at end of file diff --git a/zh/usage/plugin/DataFlow/index.html b/zh/usage/plugin/DataFlow/index.html new file mode 100644 index 00000000..a31ab2a6 --- /dev/null +++ b/zh/usage/plugin/DataFlow/index.html @@ -0,0 +1,1640 @@ + + + + + + + + + + + + + + + + + + + + + 元枢智汇 ADP 智能数据平台简介 - MinerU + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

+ + + + 跳转至 + + +

+ +

+ + + + +

+ + +

+ + + + + + +

+ + MinerU + +

+ + + 元枢智汇 ADP 智能数据平台简介 + + +

+ + + + + + + + + +

+ + +

+ + English + +
+ + 中文 + +

+ + + + + + + + + +

+ +

+ + +

+ opendatalab/MinerU +

+ +

+ + +

+ + + + + + + + + + \ No newline at end of file diff --git a/zh/usage/plugin/Dify/index.html b/zh/usage/plugin/Dify/index.html new file mode 100644 index 00000000..076c22f1 --- /dev/null +++ b/zh/usage/plugin/Dify/index.html @@ -0,0 +1,1733 @@ + + + + + + + + + + + + + + + + + + + + + Dify 简介 - MinerU + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

+ + + + 跳转至 + + +

+ +

+ + + + +

+ + +

+ + + + + + +

+ + MinerU + +

+ + + Dify 简介 + + +

+ + + + + + + + + +

+ + +

+ + English + +
+ + 中文 + +

+ + + + + + + + + +

+ +

+ + +

+ opendatalab/MinerU +

+ +

+ + +

+ + + + + + + + + + \ No newline at end of file diff --git a/zh/usage/plugin/DingTalk/index.html b/zh/usage/plugin/DingTalk/index.html new file mode 100644 index 00000000..6c40e236 --- /dev/null +++ b/zh/usage/plugin/DingTalk/index.html @@ -0,0 +1,1639 @@ + + + + + + + + + + + + + + + + + + + + + 钉钉简介 - MinerU + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

+ + + + 跳转至 + + +

+ +

+ + + + +

+ + +

+ + + + + + +

+ + MinerU + +

+ + + 钉钉简介 + + +

+ + + + + + + + + +

+ + +

+ + English + +
+ + 中文 + +

+ + + + + + + + + +

+ +

+ + +

+ opendatalab/MinerU +

+ +

+ + +

+ + + + + + + + + + \ No newline at end of file diff --git a/zh/usage/plugin/FastGPT/index.html b/zh/usage/plugin/FastGPT/index.html new file mode 100644 index 00000000..3a89d01e --- /dev/null +++ b/zh/usage/plugin/FastGPT/index.html @@ -0,0 +1,1640 @@ + + + + + + + + + + + + + + + + + + + + + FastGPT 简介 - MinerU + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

+ + + + 跳转至 + + +

+ +

+ + + + +

+ + +

+ + + + + + +

+ + MinerU + +

+ + + FastGPT 简介 + + +

+ + + + + + + + + +

+ + +

+ + English + +
+ + 中文 + +

+ + + + + + + + + +

+ +

+ + +

+ opendatalab/MinerU +

+ +

+ + +

+ + + + + + + + + + \ No newline at end of file diff --git a/zh/usage/plugin/ModelWhale/index.html b/zh/usage/plugin/ModelWhale/index.html new file mode 100644 index 00000000..80b9d9f7 --- /dev/null +++ b/zh/usage/plugin/ModelWhale/index.html @@ -0,0 +1,1641 @@ + + + + + + + + + + + + + + + + + + + + + ModelWhale 简介 - MinerU + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

+ + + + 跳转至 + + +

+ +

+ + + + +

+ + +

+ + + + + + +

+ + MinerU + +

+ + + ModelWhale 简介 + + +

+ + + + + + + + + +

+ + +

+ + English + +
+ + 中文 + +

+ + + + + + + + + +

+ +

+ + +

+ opendatalab/MinerU +

+ +

+ + +

+ + + + + + + + + + \ No newline at end of file diff --git a/zh/usage/plugin/RagFlow/index.html b/zh/usage/plugin/RagFlow/index.html new file mode 100644 index 00000000..731b3182 --- /dev/null +++ b/zh/usage/plugin/RagFlow/index.html @@ -0,0 +1,1697 @@ + + + + + + + + + + + + + + + + + + + + + RagFlow - MinerU + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

+ + + + 跳转至 + + +

+ +

+ + + + +

+ + +

+ + + + + + +

+ + MinerU + +

+ + + RagFlow + + +

+ + + + + + + + + +

+ + +

+ + English + +
+ + 中文 + +

+ + + + + + + + + +

+ +

+ + +

+ opendatalab/MinerU +

+ +

+ + + + + +

+ + + + + + 主页 + + +
+ + + + + + 快速入门 + + +
+ + + + + + 使用指南 + + +
+ + + + + + 参考资料 + + +
+ + + + + + 常见问题解答 + + +
+ + + + + + 在线演示 + + +

+ + + +

+ +

+ + + + +

+ + + + + + + +

RagFlow

+ +

RAGFlow

RAGFlow 是一款开源 RAG（Retrieval-Augmented Generation）引擎与应用平台，深度融合了深度文档理解、自动化 RAG 工作流与大模型调用，打通了复杂数据处理、知识检索、增强生成的全流程，旨在为企业及开发者提供一站式智能问答开发服务，并支持各类复杂场景下大模型的构建与应用落地。

目前，MinerU 已深度集成至 RAGFlow 知识库在线版本，作为内置 PDF 文档解析器，为用户知识库搭建提供专业、可靠的文档解析支持。本地部署版本部署使用方式详见下方使用教程。

使用可访问：https://demo.ragflow.io/

使用教程：如何在 RAGFlow 中使用 MinerU

一、安装配置

首先，我们建议您通过 docker 的形式在本地部署 RagFlow 以方便使用 MinerU 插件作为解析工具。在安装完 RagFlow 后执行：

版本检查：

确保你的RAGFlow版本 >= v0.21.1。

更新 .env 文件：

为了确保服务能被平稳修改，建议先在 cmd 运行 docker compose down 停掉服务。

打开 .env 文件，在文件的末尾，添加这两行代码，保存文件。

HF_ENDPOINT=https://hf-mirror.com
+MINERU_EXECUTABLE=/ragflow/uv_tools/.venv/bin/mineru
+

启动并进入容器：

在 cmd 中，重新启动服务：docker compose up -d

等待服务全部 Running 或 Healthy 后，运行以下命令进入RAGFlow的核心容器：

docker compose exec ragflow-cpu bash
+

（你的命令行提示符会从 C:\...> 变为 root@...）

+
在容器内下载 MinerU 模型：
+
在容器内部，依次运行以下 5 条命令
+

mkdir uv_tools
+cd uv_tools
+uv venv .venv
+source .venv/bin/activate
+uv pip install -U "mineru[core]" -i https://mirrors.aliyun.com/pypi/simple
+

退出并重启：

安装完成后，输入 exit 并按回车。

运行重启命令，让 RAGFlow 加载刚装好的 MinerU

docker compose restart ragflow-cpu
+

二、使用入口

在本地部署完毕后，要启用 MinerU，您需要在进入 RagFlow 特定知识库的配置页面并选择 MinerU 作为默认的 PDF 解析器。（注：RagFlow 在线版中已经内置了 MinerU 插件为您提供了高级的 PDF 文件解析能力，使用方式与此一致。）

入口和配置步骤：

进入知识库配置：
首先，在您的知识库管理界面，选择您需要配置的特定知识库（例如图示中的 "content" 知识库）。
在知识库详情页面的左侧导航栏中，点击【配置】选项卡。
定位 PDF 解析器设置：
向下滚动页面，找到“Ingestion pipeline”（摄取管道）设置部分。
在此部分中，您会看到一个名为【PDF解析器】（PDF Parser）的选项。
选择 MinerU：
点击【PDF解析器】旁边的下拉菜单。
从可用选项中，选择【MinerU】。
保存修改：
完成选择后，请务必点击页面底部的【保存】按钮，以使更改生效。

+ + + + + + + + + + + + + + + +

+ + + + + +

+ + + +

+ + + + + + + + + + \ No newline at end of file diff --git a/zh/usage/plugin/Sider/index.html b/zh/usage/plugin/Sider/index.html new file mode 100644 index 00000000..59c76026 --- /dev/null +++ b/zh/usage/plugin/Sider/index.html @@ -0,0 +1,1639 @@ + + + + + + + + + + + + + + + + + + + + + Sider 简介 - MinerU + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

+ + + + 跳转至 + + +

+ +

+ + + + +

+ + +

+ + + + + + +

+ + MinerU + +

+ + + Sider 简介 + + +

+ + + + + + + + + +

+ + +

+ + English + +
+ + 中文 + +

+ + + + + + + + + +

+ +

+ + +

+ opendatalab/MinerU +

+ +

+ + +

+ + + + + + + + + + \ No newline at end of file diff --git a/zh/usage/plugin/n8n/index.html b/zh/usage/plugin/n8n/index.html new file mode 100644 index 00000000..65ed215b --- /dev/null +++ b/zh/usage/plugin/n8n/index.html @@ -0,0 +1,1661 @@ + + + + + + + + + + + + + + + + + + + + + n8n 简介 - MinerU + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

+ + + + 跳转至 + + +

+ +

+ + + + +

+ + +

+ + + + + + +

+ + MinerU + +

+ + + n8n 简介 + + +

+ + + + + + + + + +

+ + +

+ + English + +
+ + 中文 + +

+ + + + + + + + + +

+ +

+ + +

+ opendatalab/MinerU +

+ +

+ + +

+ + + + + + + + + + \ No newline at end of file diff --git a/zh/usage/quick_usage/index.html b/zh/usage/quick_usage/index.html new file mode 100644 index 00000000..0ece0c7e --- /dev/null +++ b/zh/usage/quick_usage/index.html @@ -0,0 +1,1973 @@ + + + + + + + + + + + + + + + + + + + + + + + + + 基础使用 - MinerU + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

+ + + + 跳转至 + + +

+ +

+ + + + +

+ + +

+ + + + + + +

+ + MinerU + +

+ + + 基础使用 + + +

+ + + + + + + + + +

+ + +

+ + English + +
+ + 中文 + +

+ + + + + + + + + +

+ +

+ + +

+ opendatalab/MinerU +

+ +

+ + + + + +

+ + + + + + 主页 + + +
+ + + + + + 快速入门 + + +
+ + + + + + 使用指南 + + +
+ + + + + + 参考资料 + + +
+ + + + + + 常见问题解答 + + +
+ + + + + + 在线演示 + + +

+ + + +

+ +

+ + + + +

+ + + + + + + +

使用 MinerU

快速配置模型源

MinerU默认使用huggingface作为模型源，若用户网络无法访问huggingface，可以通过环境变量便捷地切换模型源为modelscope： +

export MINERU_MODEL_SOURCE=modelscope
+

+有关模型源配置和自定义本地模型路径的更多信息，请参考文档中的模型源说明。 +

通过命令行快速使用

MinerU内置了命令行工具，用户可以通过命令行快速使用MinerU进行PDF解析： +

mineru -p <input_path> -o <output_path>
+

Tip

<input_path>：本地 PDF/图片文件或目录
<output_path>：输出目录
未传 --api-url 时，CLI 会自动拉起本地临时 mineru-api
传入 --api-url 时，CLI 会直连远端或已有本地 FastAPI 服务

更多关于输出文件的信息，请参考输出文件说明。

Note

命令行工具会在Linux和macOS系统自动尝试cuda/mps加速。Windows用户如需使用cuda加速， +请前往 Pytorch官网选择适合自己cuda版本的命令安装支持加速的torch和torchvision。

如果需要通过自定义参数调整解析选项，您也可以在文档中查看更详细的命令行工具使用说明。

通过api、webui、http-client/server进阶使用

通过python api直接调用：Python 调用示例
通过fast api方式调用： +
```
mineru-api --host 0.0.0.0 --port 8000
+
```
+
Tip
+
在浏览器中访问 http://127.0.0.1:8000/docs 查看API文档。
+
- 健康检查接口：GET /health + 返回 protocol_version、processing_window_size、max_concurrent_requests 等服务信息
- 异步任务提交接口：POST /tasks
- 同步解析接口：POST /file_parse
- 任务查询接口：GET /tasks/{task_id}、GET /tasks/{task_id}/result
- API 输出目录由服务端固定控制，默认写入 ./output
+
POST /tasks 会立即返回 task_id；POST /file_parse 会在内部提交到同一个任务管理器，等待任务完成后同步返回最终结果。 +任务为单进程、进程内状态实现，服务重启、--reload 热重载或多进程部署后不保证仍可查询历史任务状态。 +默认任务完成或失败后保留 24 小时，随后自动清理任务状态和输出目录；清理后访问任务状态或结果会返回 404。 +可通过环境变量 MINERU_API_TASK_RETENTION_SECONDS 和 MINERU_API_TASK_CLEANUP_INTERVAL_SECONDS 调整保留时长与清理轮询间隔。
+
异步任务提交示例： +
```
curl -X POST http://127.0.0.1:8000/tasks \
+  -F "files=@demo/pdfs/demo1.pdf" \
+  -F "return_md=true"
+
```
+
同步解析示例： +
```
curl -X POST http://127.0.0.1:8000/file_parse \
+  -F "files=@demo/pdfs/demo1.pdf" \
+  -F "return_md=true" \
+  -F "response_format_zip=true" \
+  -F "return_original_file=true"
+
```
+
轮询任务状态与结果： +
```
curl http://127.0.0.1:8000/tasks/<task_id>
+curl http://127.0.0.1:8000/tasks/<task_id>/result
+curl http://127.0.0.1:8000/health
+
```
+
+
+
启动gradio webui 可视化前端： +
```
mineru-gradio --server-name 0.0.0.0 --server-port 7860
+
```
+
+
Tip
+
- 在浏览器中访问 http://127.0.0.1:7860 使用 Gradio WebUI。
+
+

使用http-client/server方式调用： +

# 启动openai兼容服务器(需要安装vllm或lmdeploy环境)
+mineru-openai-server --port 30000
+

Tip

在另一个终端中通过http client连接openai server +

mineru -p <input_path> -o <output_path> -b hybrid-http-client -u http://127.0.0.1:30000
+

Note

所有vllm/lmdeploy官方支持的参数都可用通过命令行参数传递给 MinerU，包括以下命令:mineru、mineru-openai-server、mineru-gradio、mineru-api， +我们整理了一些vllm/lmdeploy使用中的常用参数和使用方法，可以在文档命令行进阶参数中获取。

基于配置文件扩展 MinerU 功能

MinerU 现已实现开箱即用，但也支持通过配置文件扩展功能。您可通过编辑用户目录下的 mineru.json 文件，添加自定义配置。

Important

mineru.json 文件会在您使用内置模型下载命令 mineru-models-download 时自动生成，也可以通过将配置模板文件复制到用户目录下并重命名为 mineru.json 来创建。

以下是一些可用的配置选项：

+
latex-delimiter-config：
+
- 用于配置 LaTeX 公式的分隔符
- 默认为$符号，可根据需要修改为其他符号或字符串。
+
+
llm-aided-config：
+
- 用于配置 LLM 辅助标题分级的相关参数，兼容所有支持openai协议的 LLM 模型
- 默认使用阿里云百炼的qwen3-next-80b-a3b-instruct模型
- 您需要自行配置 API 密钥并将enable设置为true来启用此功能
- 如果您的api供应商不支持enable_thinking参数，请手动将该参数删除
  - 例如，在您的配置文件中，llm-aided-config 部分可能如下所示： +
```
"llm-aided-config": {
+   "api_key": "your_api_key",
+   "base_url": "https://dashscope.aliyuncs.com/compatible-mode/v1",
+   "model": "qwen3-next-80b-a3b-instruct",
+   "enable_thinking": false,
+   "enable": false
+}
+
```
  - 要移除enable_thinking参数，只需删除包含"enable_thinking": false的那一行，结果如下: +
```
"llm-aided-config": {
+   "api_key": "your_api_key",
+   "base_url": "https://dashscope.aliyuncs.com/compatible-mode/v1",
+   "model": "qwen3-next-80b-a3b-instruct",
+   "enable": false
+}
+
```
  +
+
+
models-dir：
+
- 用于指定本地模型存储目录，请为pipeline和vlm后端分别指定模型目录，
- 指定目录后您可通过配置环境变量export MINERU_MODEL_SOURCE=local来使用本地模型。
+

+ + + + + + + + + + + + + + + +

+ + + + + +

+ + + +

+ + + + + + + + + + \ No newline at end of file

404 - Not found

Frequently Asked Questions

MinerU

Project Introduction

Key Features

User Guide

Deploying MinerU with Docker

Build Docker Image using Dockerfile

Docker Description

Start Docker Container

Start Services Directly with Docker Compose

Start OpenAI-compatible server service

Start Web API service

Start Gradio WebUI service

MinerU Extension Modules Installation Guide

Common Scenarios

Core Functionality Installation

Using vllm to Accelerate VLM Model Inference

Using lmdeploy to Accelerate VLM Model Inference

Installing Lightweight Client to Connect to OpenAI-compatible servers (for vlm-http-client mode)

Installing Lightweight Client to Connect to OpenAI-compatible servers (for hybrid-http-client mode)

Quick Start

Online Experience

Official online web application

Gradio-based online demo

Local Deployment

Install MinerU

Install MinerU using pip or uv

Install MinerU from source code

Deploy MinerU using Docker

Using MinerU

Changelog

2.6 Series Versions

2.6.7 (2025/12/12)

2.6.6 (2025/12/02)

2.6.5 (2025/11/26)

2.6.4 (2025/11/04)

2.6.3 (2025/10/31)

2.6.2 (2025/10/24)

2.5 Series Versions

2.5.4 (2025/09/26)

2.5.3 (2025/09/20)

2.5.2 (2025/09/19)

2.2 - 2.4 Series Versions

2.2.2 (2025/09/10)

2.2.1 (2025/09/08)

2.2.0 (2025/09/05)

2.1 Series Versions

2.1.10 (2025/08/01)

2.1.9 (2025/07/30)

2.1.8 (2025/07/28)

2.1.7 (2025/07/27)

2.1.6 (2025/07/26)

2.1.5 (2025/07/24)

2.1.4 (2025/07/23)

2.1.1 (2025/07/16)

2.1.0 (2025/07/05)

2.0 Series Versions

2.0.6 (2025/06/20)

2.0.5 (2025/06/17)

2.0.3 (2025/06/15)

2.0.0 (2025/06/13)

1.x Series Historical Versions

1.3.12 (2025/05/24)

1.3.10 (2025/04/29)

1.3.9 (2025/04/27)

1.3.8 (2025/04/23)

1.3.7 (2025/04/22)

1.3.4 (2025/04/16)

1.3.2 (2025/04/12)

1.3.1 (2025/04/08)

1.3.0 (2025/04/03)

1.2.1 (2025/03/03)

1.2.0 (2025/02/24)

1.1.0 (2025/01/22)

1.0.1 (2025/01/10)

0.x Series Historical Versions

0.10.0 (2024/11/22)

0.9.3 (2024/11/15)

0.9.2 (2024/11/06)

Using `vllm` to Accelerate VLM Model Inference

Using `lmdeploy` to Accelerate VLM Model Inference