Skip to content
Computer vision

Unlimited OCR 2

разбор документов без ограничения по длине

1 min read

Мультиязычная OCR-модель от Baidu на 3,34 млрд параметров для распознавания изображений и PDF. Вторая версия развивает идею «unlimited-horizon parsing»: документ можно подавать целиком, а не отдельными страницами.

Ключевые особенности

  • SR-SWA сжимает text-visual attention до фиксированного набора рабочих reference-токенов (например, 512), поэтому latency и память почти не растут с длиной документа.
  • Поддерживает многостраничный inference в одном запросе — модель может использовать контекст между страницами.
  • Заявлена SOTA на бенчмарках длинных документов MPDocBench и экстремально длинном MarathonDocBench.
  • Есть нативный запуск через Transformers для быстрых тестов и SGLang для serving.
  • В комплекте — отдельный бенчмарк MarathonDocBench для сверхдлинных документов.

Пример

from transformers import AutoModel, AutoTokenizer
import torch
model_name = 'baidu/Unlimited-OCR-2'
tokenizer = AutoTokenizer.from_pretrained(model_name, trust_remote_code=True)
model = AutoModel.from_pretrained(model_name, trust_remote_code=True,
    torch_dtype=torch.bfloat16).eval().cuda()
pages = [f'doc/page_{i}.jpg' for i in range(1, 8)]
text, n = model.infer_multi(tokenizer, image_files=pages, output_path='out')
ПодробнееБенчмарки: SOTA на MPDocBench и MarathonDocBench; точные цифры в карточке не приводятся.
Тестировали Transformers на NVIDIA A800-80GB; для SGLang рекомендуют mem-fraction-static 0.7.
SR-SWA ограничивает вычисления фиксированным окном reference-токенов; страничное окно в SGLang снижает пиковую память, но может ухудшать межстраничное объединение.
Примеры
Discuss Telegram 591 189.9K

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 6.2 GB · ready files7.4 GB
  • FP8 / INT8recommendedweights 3.1 GB3.9 GB
  • 4-bitweights 1.8 GB2.4 GB

Tip: go 8-bit.FP8 / INT8 is within about a percent of full precision in quality, takes half the memory and runs faster on RTX 40/50 and H100. Full precision only pays off for fine-tuning.

What to get for this model

Massed ComputeOpen Massed Compute

Your own GPU: FP8 / INT8 — from RTX 2060 · BF16 — from RTX 5060 Ti 8GB

Massed Compute hourly prices, November 2026. The discounted price applies with the coupon at checkout.

Estimate: weights plus working memory. Long videos, high resolutions and long contexts need more.

Run it on Massed Compute

The script downloads the model to the rented machine. How to run it is in the model card.

  1. 1. Rent a machine with a fitting GPU

    Open Massed Compute
  2. 2. Paste the script into the machine's terminal

    Connect over SSH (the Massed Compute panel shows the command) or open a terminal on the machine's desktop.

    HF_TOKEN / CIVITAI_TOKENThe script starts with token lines: paste yours between the quotes. The Hugging Face token is for gated models, the Civitai one for Civitai models.

    # baidu/Unlimited-OCR-2
    mkdir -p ~/models
    curl -LsSf https://astral.sh/uv/install.sh | sh
    source "$HOME/.local/bin/env"
    uv tool install "huggingface_hub[cli]"
    hf download baidu/Unlimited-OCR-2 --local-dir ~/models/Unlimited-OCR-2
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to Unlimited-OCR-2
  3. 3. Use it

    The model is in ~/models. How to run it is in the model card (link above).

Stop the machine in the panel when you are done: it bills by the minute while it runs.

How to speed it up

Picked for this model type. Choose your hardware above to see what fits you exactly.

  • FlashAttention 44.0.0b40from November 18, 2026new release

    fastest exact attention on B200

    Kernels for Blackwell (beta). Exact attention without quantisation — fine for everything, LLMs included.

    How to enable: pip install flash-attn-4; vLLM and SGLang pick it up themselves.

    Earlier generations (2)
    • FlashAttention 3 — The Hopper (H100/H200) version with FP8.
    • FlashAttention 2 — The standard for RTX 30/40/50 and A100: exact attention without the memory cost of long contexts.
  • torch.compile2.15.0from October 28, 2026new release

    10–30% for free

    Compiles the model for your GPU. Slower first run, faster afterwards.

    How to enable: model = torch.compile(model); in ComfyUI the TorchCompileModel node.

Model family

New posts about versions of Unlimited-OCR-2 (quants, finetunes, LoRAs) will show up under the bell.

Quantised versions 12All on Hugging Face
Fine-tunes 5All on Hugging Face

About the model

baidu/Unlimited-OCR-2
Author
baidu
Category
Computer vision
Released
Posted

On NeyroLearn

Hugging Face

189.9K591

Ask AI about this model

Answered by Gemini 4 · the chat is kept only in this tab

Comments

Computer vision

d1-2B

d1-2B