Skip to content
Computer vision

diffu-minicpm-ocr-htr

OCR для шведских рукописных архивов

2 min read

Полный файнтюн MiniCPM-OCR-1.0 на 1,55 млрд параметров читает шведские рукописные архивные записи: одну строку или целый текстовый регион со строками в порядке чтения. Модель обучена как reader для проекта diffu и сравнивается с Riksarkivet Swedish Lion TrOCR на тех же данных и тестовых строках.

Ключевые особенности

  • Обучена на GT-Lejonet HTR: 1 481 877 строк и 83 072 региона; тестовый split test_0050 не использовался.
  • Полный файнтюн: vision encoder и языковая модель обучены, aligner заморожен; 28 000 из 36 681 шага (2,3 эпохи) на двух RTX PRO 6000 Blackwell.
  • Нужен vLLM commit fa3b9d13 или новее; кроп строки/региона — белый фон вне полигона, масштаб до примерно 448×448 (регион до 9×).
  • На одной RTX PRO 6000 Blackwell прочитала 5 000 тестовых строк со скоростью 68,55 строк/с; baseline — 7,91 строк/с.
  • Ограничения: хуже читает незнакомые почерки, периоды и типы записей; может заменить старую орфографию на современную или потерять букву; при near-greedy decoding возможны повторы до лимита токенов.

Пример

vllm serve Gabriel/diffu-minicpm-ocr-htr --served-model-name htr --max-model-len 8192 --limit-mm-per-prompt '{"image": 4, "video": 0}' --mm-processor-kwargs '{"downsample_mode": "4x", "max_slice_nums": 9}' --trust-request-mm-kwargs --default-chat-template-kwargs '{"enable_thinking": false}
ПодробнееБенчмарки: 5 000 строк test_0050 — mean CER 5,22%, CER (OpenBMB) 5,64%, WER 20,06%; 6 358 OOD-строк — mean CER 7,21%.
У Swedish Lion TrOCR на тех же наборах mean CER 4,17% и 6,67% соответственно.
Требования: vLLM, max-model-len 8192, до 448×448 на строку; лицензия Apache-2.0. На тесте 3 чтения достигли лимита токенов — повторный хвост можно обрезать.
Discuss Telegram 0 2

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 2.9 GB · ready files3.7 GB
  • FP8 / INT8recommendedweights 1.4 GB2.1 GB
  • 4-bitweights 0.8 GB1.4 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 2060

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.

    # Gabriel/diffu-minicpm-ocr-htr
    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 Gabriel/diffu-minicpm-ocr-htr --local-dir ~/models/diffu-minicpm-ocr-htr
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to diffu-minicpm-ocr-htr
  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

  1. MiniCPM-V-4.6family root
  2. MiniCPM-OCR-1.0
  3. diffu-minicpm-ocr-htr

New posts about versions of MiniCPM-V-4.6 (quants, finetunes, LoRAs) will show up under the bell.

Quantised versions 12All on Hugging Face
Fine-tunes 12All on Hugging Face
LoRAs and adapters 8All on Hugging Face

About the model

Gabriel/diffu-minicpm-ocr-htr
Author
Gabriel
Category
Computer vision
Released
Posted

On NeyroLearn

Hugging Face

20

Ask AI about this model

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

Comments

Language models

PRISM-1B

PRISM-1B