Skip to content

Latest commit

 

History

History
474 lines (315 loc) · 22.9 KB

README.md

File metadata and controls

474 lines (315 loc) · 22.9 KB
Raw

Marker API

Important

Marker API provides a simple endpoint for converting PDF documents to Markdown quickly and accurately. With just one click, you can deploy the Marker API endpoint and start converting PDFs seamlessly.

Features 🚀

  • PDF to Markdown Conversion 📄 ➡️ 📝: Converts PDF documents to Markdown efficiently.
  • Asynchronous Processing ⏳🔄: Supports both sync and async task processing.
  • Distributed Architecture 🌐📡: Enables distributed processing using Celery and Redis.
  • Monitoring 📊👁️: Integrated with Flower for monitoring the Celery tasks.
  • Scalable 📈🛠️: Easily scalable with distributed workers.
  • GPU/CPU Support 💻⚡: Supports both CPU and GPU for intensive processing.
  • Multi-Language Support 🌍💬: Handles documents in any language.
  • Advanced Formatting 🎨📋: Extracts tables, images, and code blocks accurately.
  • Equation Handling ✏️➗: Converts most mathematical equations to LaTeX.
  • Docker and Kubernetes Support 🐳🛠️: Full Docker support for both CPU and GPU setups, and Kubernetes (coming soon).

Comparison

Original PDF Marker-API PyPDF
Original PDF Marker-API PyPDF

Here’s an improved version of the features section, along with the comparison table and setup details:

Simple Server vs. Distributed Server

Criteria Simple Server Distributed Server
Use Case Small scale, local deployments Large scale, production environments
Scalability Limited (single server) Highly scalable with multiple workers
Performance Synchronous (one file at a time) Asynchronous with parallel task processing
Setup Complexity Easy (local or Docker setup) Moderate (multiple services to manage)
Monitoring No built-in monitoring Flower for task monitoring
Failure Tolerance Low (single server) High (distributed worker nodes)
Docker Support Yes, CPU and GPU Yes, CPU and GPU
Resource Utilization Limited to local resources Efficient resource distribution with Celery
Task Distribution No task distribution Uses Celery for task distribution
Ideal For Testing, small local deployments Production environments with high traffic

Installation and Setup

Simple Server Setup

Local Setup

  1. Clone the repository:

    git clone https://github.com/adithya-s-k/marker-api
cd marker-api

  2. Set up the Python environment:

    cp .env.example .env
poetry install  # or
pip install -e .

  3. Run the server:

    python server.py --host 0.0.0.0 --port 8080
Docker Setup (Simple Server)

  • For CPU:

    docker build -f docker/Dockerfile.cpu.server -t marker-api-cpu .
docker run -p 8080:8080 marker-api-cpu

  • For GPU:

    docker build -f docker/Dockerfile.gpu.server -t marker-api-gpu .
docker run --gpus all -p 8080:8080 marker-api-gpu

Distributed Server Setup

The distributed server setup uses several services like FastAPI, Redis, Celery workers, and Flower, which are designed for large-scale deployments. You can either set it up locally or use Docker Compose for more seamless scaling.

Local Setup ⚙️

Step 1: Clone the repository and set up the environment
git clone https://github.com/adithya-s-k/marker-api
cd marker-api
cp .env.example .env
poetry install  # or
pip install -e .
Step 2: Run Redis using Docker

Since Redis is essential for Celery to function, let's run Redis in a Docker container. This will ensure it's up and running on port 6379.

docker run -d -p 6379:6379 redis

Now, update your .env file with the correct Redis host URL:

REDIS_HOST=redis://localhost:6379/0
Step 3: Open three terminals for each service

You will need three separate terminals for Redis, Celery, and FastAPI:

  • Terminal 1: Start a Celery worker

    celery -A marker_api.celery_worker.celery_app worker --pool=solo --loglevel=info

  • Terminal 2: Start Flower for Celery task monitoring

    celery -A marker_api.celery_worker.celery_app flower --port=5555

  • Terminal 3: Run FastAPI server

    python distributed_server.py --host 0.0.0.0 --port 8080
Adding More Celery Workers

To scale Celery workers manually, you can open more terminals and run the same Celery worker command:

celery -A marker_api.celery_worker.celery_app worker --pool=solo --loglevel=info

Each new terminal will spin up a new worker, allowing the system to handle more tasks concurrently.

Docker Compose Setup (Distributed Server) 🐳

For an even more seamless setup, you can use Docker Compose, which helps run all services within containers. Here’s how to do it.

Step 1: Start the distributed server with Docker Compose

Use Docker Compose to orchestrate all the services.

  • Without Scaling:

    sudo docker-compose -f docker-compose.gpu.yml up --build

This command will start:

  • FastAPI on port 8080.
  • Redis for managing task queue messages.
  • Celery Worker to handle background tasks.
  • Flower on port 5556 to monitor the Celery tasks.
Step 2: Scale Celery Workers with Docker Compose

To scale up Celery workers, you can use the --scale option:

sudo docker-compose -f docker-compose.gpu.yml up --build --scale celery_worker=3
  • This command will start 3 Celery workers, increasing the task processing capacity of the system.
  • When you use scaling, Docker Compose will automatically spin up multiple instances of the celery_worker service.

Explanation of Scaling in Docker Compose

When running without scaling, you get one instance of the Celery worker service. However, adding the --scale celery_worker=3 flag creates three instances of the worker, meaning tasks will be processed concurrently by three separate workers, which improves the throughput and helps distribute the load across multiple workers.

Kubernetes Support

(Coming Soon)

Why Distributed?

The distributed server architecture offers several advantages over the simple server:

  1. Scalability: By using Celery workers, the system can easily scale horizontally by adding more worker instances to handle increased load.

  2. Reliability: The distributed system is more resilient to failures. If one worker crashes, others can continue processing tasks.

  3. Non-blocking: Asynchronous processing allows the API to handle many requests concurrently without blocking, improving overall throughput.

  4. Resource Management: The distributed architecture allows for better management of resources, particularly important for GPU utilization in PDF processing.

  5. Monitoring: With tools like Flower, it's easier to monitor the system's performance and identify bottlenecks.

  6. Flexibility: The distributed setup allows for more complex workflows and integration with other services.

By using the distributed server, you can process multiple PDFs in parallel, handle high loads more effectively, and provide a more responsive service to your users.

Marker Readme

Marker converts PDF to markdown quickly and accurately.

  • Supports a wide range of documents (optimized for books and scientific papers)
  • Supports all languages
  • Removes headers/footers/other artifacts
  • Formats tables and code blocks
  • Extracts and saves images along with the markdown
  • Converts most equations to latex
  • Works on GPU, CPU, or MPS

How it works

Marker is a pipeline of deep learning models:

  • Extract text, OCR if necessary (heuristics, surya, tesseract)
  • Detect page layout and find reading order (surya)
  • Clean and format each block (heuristics, texify
  • Combine blocks and postprocess complete text (heuristics, pdf_postprocessor)

It only uses models where necessary, which improves speed and accuracy.

Examples

PDF Type Marker Nougat
Think Python Textbook View View
Think OS Textbook View View
Switch Transformers arXiv paper View View
Multi-column CNN arXiv paper View View

Performance

Benchmark overall

The above results are with marker and nougat setup so they each take ~4GB of VRAM on an A6000.

See below for detailed speed and accuracy benchmarks, and instructions on how to run your own benchmarks.

Commercial usage

I want marker to be as widely accessible as possible, while still funding my development/training costs. Research and personal usage is always okay, but there are some restrictions on commercial usage.

The weights for the models are licensed cc-by-nc-sa-4.0, but I will waive that for any organization under $5M USD in gross revenue in the most recent 12-month period AND under $5M in lifetime VC/angel funding raised. If you want to remove the GPL license requirements (dual-license) and/or use the weights commercially over the revenue limit, check out the options here.

Community

Discord is where we discuss future development.

Limitations

PDF is a tricky format, so marker will not always work perfectly. Here are some known limitations that are on the roadmap to address:

  • Marker will not convert 100% of equations to LaTeX. This is because it has to detect then convert.
  • Tables are not always formatted 100% correctly - text can be in the wrong column.
  • Whitespace and indentations are not always respected.
  • Not all lines/spans will be joined properly.
  • This works best on digital PDFs that won't require a lot of OCR. It's optimized for speed, and limited OCR is used to fix errors.

Installation

You'll need python 3.9+ and PyTorch. You may need to install the CPU version of torch first if you're not using a Mac or a GPU machine. See here for more details.

Install with:

pip install marker-pdf

Optional: OCRMyPDF

Only needed if you want to use the optional ocrmypdf as the ocr backend. Note that ocrmypdf includes Ghostscript, an AGPL dependency, but calls it via CLI, so it does not trigger the license provisions.

See the instructions here

Usage

First, some configuration:

  • Inspect the settings in marker/settings.py. You can override any settings with environment variables.
  • Your torch device will be automatically detected, but you can override this. For example, TORCH_DEVICE=cuda.
    • If using GPU, set INFERENCE_RAM to your GPU VRAM (per GPU). For example, if you have 16 GB of VRAM, set INFERENCE_RAM=16.
    • Depending on your document types, marker's average memory usage per task can vary slightly. You can configure VRAM_PER_TASK to adjust this if you notice tasks failing with GPU out of memory errors.
  • By default, marker will use surya for OCR. Surya is slower on CPU, but more accurate than tesseract. If you want faster OCR, set OCR_ENGINE to ocrmypdf. This also requires external dependencies (see above). If you don't want OCR at all, set OCR_ENGINE to None.

Convert a single file

marker_single /path/to/file.pdf /path/to/output/folder --batch_multiplier 2 --max_pages 10 --langs English
  • --batch_multiplier is how much to multiply default batch sizes by if you have extra VRAM. Higher numbers will take more VRAM, but process faster. Set to 2 by default. The default batch sizes will take ~3GB of VRAM.
  • --max_pages is the maximum number of pages to process. Omit this to convert the entire document.
  • --langs is a comma separated list of the languages in the document, for OCR

Make sure the DEFAULT_LANG setting is set appropriately for your document. The list of supported languages for OCR is here. If you need more languages, you can use any language supported by Tesseract if you set OCR_ENGINE to ocrmypdf. If you don't need OCR, marker can work with any language.

Convert multiple files

marker /path/to/input/folder /path/to/output/folder --workers 10 --max 10 --metadata_file /path/to/metadata.json --min_length 10000
  • --workers is the number of pdfs to convert at once. This is set to 1 by default, but you can increase it to increase throughput, at the cost of more CPU/GPU usage. Parallelism will not increase beyond INFERENCE_RAM / VRAM_PER_TASK if you're using GPU.
  • --max is the maximum number of pdfs to convert. Omit this to convert all pdfs in the folder.
  • --min_length is the minimum number of characters that need to be extracted from a pdf before it will be considered for processing. If you're processing a lot of pdfs, I recommend setting this to avoid OCRing pdfs that are mostly images. (slows everything down)
  • --metadata_file is an optional path to a json file with metadata about the pdfs. If you provide it, it will be used to set the language for each pdf. If not, DEFAULT_LANG will be used. The format is:
{
  "pdf1.pdf": {"languages": ["English"]},
  "pdf2.pdf": {"languages": ["Spanish", "Russian"]},
  ...
}

You can use language names or codes. The exact codes depend on the OCR engine. See here for a full list for surya codes, and here for tesseract.

Convert multiple files on multiple GPUs

MIN_LENGTH=10000 METADATA_FILE=../pdf_meta.json NUM_DEVICES=4 NUM_WORKERS=15 marker_chunk_convert ../pdf_in ../md_out
  • METADATA_FILE is an optional path to a json file with metadata about the pdfs. See above for the format.
  • NUM_DEVICES is the number of GPUs to use. Should be 2 or greater.
  • NUM_WORKERS is the number of parallel processes to run on each GPU. Per-GPU parallelism will not increase beyond INFERENCE_RAM / VRAM_PER_TASK.
  • MIN_LENGTH is the minimum number of characters that need to be extracted from a pdf before it will be considered for processing. If you're processing a lot of pdfs, I recommend setting this to avoid OCRing pdfs that are mostly images. (slows everything down)

Note that the env variables above are specific to this script, and cannot be set in local.env.

Troubleshooting

There are some settings that you may find useful if things aren't working the way you expect:

  • OCR_ALL_PAGES - set this to true to force OCR all pages. This can be very useful if the table layouts aren't recognized properly by default, or if there is garbled text.
  • TORCH_DEVICE - set this to force marker to use a given torch device for inference.
  • OCR_ENGINE - can set this to surya or ocrmypdf.
  • DEBUG - setting this to True shows ray logs when converting multiple pdfs
  • Verify that you set the languages correctly, or passed in a metadata file.
  • If you're getting out of memory errors, decrease worker count (increased the VRAM_PER_TASK setting). You can also try splitting up long PDFs into multiple files.

In general, if output is not what you expect, trying to OCR the PDF is a good first step. Not all PDFs have good text/bboxes embedded in them.

Benchmarks

Benchmarking PDF extraction quality is hard. I've created a test set by finding books and scientific papers that have a pdf version and a latex source. I convert the latex to text, and compare the reference to the output of text extraction methods. It's noisy, but at least directionally correct.

Benchmarks show that marker is 4x faster than nougat, and more accurate outside arXiv (nougat was trained on arXiv data). We show naive text extraction (pulling text out of the pdf with no processing) for comparison.

Speed

Method Average Score Time per page Time per document
marker 0.613721 0.631991 58.1432
nougat 0.406603 2.59702 238.926

Accuracy

First 3 are non-arXiv books, last 3 are arXiv papers.

Method multicolcnn.pdf switch_trans.pdf thinkpython.pdf thinkos.pdf thinkdsp.pdf crowd.pdf
marker 0.536176 0.516833 0.70515 0.710657 0.690042 0.523467
nougat 0.44009 0.588973 0.322706 0.401342 0.160842 0.525663

Peak GPU memory usage during the benchmark is 4.2GB for nougat, and 4.1GB for marker. Benchmarks were run on an A6000 Ada.

Throughput

Marker takes about 4.5GB of VRAM on average per task, so you can convert 10 documents in parallel on an A6000.

Benchmark results

Running your own benchmarks

You can benchmark the performance of marker on your machine. Install marker manually with:

git clone https://github.com/VikParuchuri/marker.git
poetry install

Download the benchmark data here and unzip. Then run benchmark.py like this:

python benchmark.py data/pdfs data/references report.json --nougat

This will benchmark marker against other text extraction methods. It sets up batch sizes for nougat and marker to use a similar amount of GPU RAM for each.

Omit --nougat to exclude nougat from the benchmark. I don't recommend running nougat on CPU, since it is very slow.

Thanks

This work would not have been possible without amazing open source models and datasets, including (but not limited to):

  • Surya
  • Texify
  • Pypdfium2/pdfium
  • DocLayNet from IBM
  • ByT5 from Google

Thank you to the authors of these models and datasets for making them available to the community!

To Do

  • Create server
  • Add support for single PDF upload
  • Add support for multi PDF upload
  • Docker support and Skypilot support
  • Implement handling for multiple PDF uploads simultaneously.
  • Introduce a toggle mode to generate Markdown without including images in the output.
  • Enhance GPU utilization and optimize performance for efficient processing.
  • Implement dynamic adjustment of batch size based on available VRAM.
  • Add celery and flower for distributed queue and monitoring
  • Locust for load testing
  • Distibuted server for scaling workloads
  • Clinet package init
  • add K8s and helm chart
  • Benchmarks from production load testing

Built With

License

Marker-api is licensed under the GPL-3.0 license. See LICENSE for more information. The project uses Marker under the hood, which has a commercial license that needs to be followed. Here are the details:

Commercial Usage

Marker and Surya OCR Models are designed to be as widely accessible as possible while still funding development and training costs. Research and personal usage are always allowed, but there are some restrictions on commercial usage. The weights for the models are licensed under cc-by-nc-sa-4.0. However, this restriction is waived for any organization with less than $5M USD in gross revenue in the most recent 12-month period AND less than $5M in lifetime VC/angel funding raised. To remove the GPL license requirements (dual-license) and/or use the weights commercially over the revenue limit, check out the options provided. Please refer to Marker for more Information about the License of the Model weights

Acknowledgements

This project is a fork of marker project created by VikParuchuri.

Star History Chart