Skip to content
Computer vision

Wang Qwen3.5 2B DocTags

компактный файнтюн для DocTags

1 min read

Модель от wangcheng на архитектуре Qwen3.5: 2,21 млрд параметров, BF16, задача image-text-to-text. Предназначена для распознавания документов и структурированного вывода DocTags; это PyTorch-checkpoint, а ONNX- и WebGPU-варианты проверялись отдельно.

Ключевые особенности

  • 2,21 млрд параметров, BF16, 24 текстовых слоя с гибридом линейного и полного внимания, hidden size 2048, словарь 248 654.
  • На выходе — DocTags: теги <doctag>, <text>, таблицы и разметка layout.
  • Используются Qwen2VLImageProcessorFast / Qwen3VLProcessor; промпт из карточки — <image>Convert this page to docling..
  • Для инференса предложены greedy, temperature=0.0, top_p=1.0, repetition_penalty=1.02, max_pixels=2073600, max_model_len=32768.
  • ONNX-конвертация завершена; WebGPU-тест в Chrome 157 / A800 прошёл, 26 токенов совпали с CPU-версией.
  • Подтверждённых accuracy/бенчмарков нет: проверка загрузки не равна проверке OCR-точности.

Пример

Промпт из карточки:

<image>Convert this page to docling.

Ожидаемый формат ответа — DocTags с текстом, таблицами и layout-тегами.

ПодробнееПубличный тест: синтетическая короткая страница на CPU — 26 токенов, около 19,3 с.
WebGPU: Chrome 157 / A800, 26 токенов совпали с CPU-оригиналом.
Веса — PyTorch/BF16; для ONNX и браузера нужна отдельная конвертация.
Полных бенчмарков точности нет; описание обучающих данных в карточке не приведено.
Apache-2.0, автор wangcheng; checkpoint step 72 000.
Discuss Telegram 0 21

Will it run?

Working out the memory needs

The bot is still collecting this model's weight sizes; the estimate will appear here in a few minutes. Meanwhile you can rent a GPU with headroom.

Massed ComputeOpen Massed Compute

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.

    # multimodal-vision-ai/Wang-Qwen3.5-2B-DocTags
    # ===== Tokens: paste yours between the quotes =====
    export HF_TOKEN=""        # huggingface.co/settings/tokens, for gated models
    # ==========================================
    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 multimodal-vision-ai/Wang-Qwen3.5-2B-DocTags --local-dir ~/models/Wang-Qwen3.5-2B-DocTags
    # The model is in ~/models. How to run it is in the model card (link above).
  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 Wang-Qwen3.5-2B-DocTags (quants, finetunes, LoRAs) will show up under the bell.

No known relatives of this model yet. All derivatives on Hugging Face

About the model

multimodal-vision-ai/Wang-Qwen3.5-2B-DocTags
Author
multimodal-vision-ai
Category
Computer vision
Released
Posted

On NeyroLearn

Hugging Face

210

Ask AI about this model

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

Comments

Other

QEV-0.8B

QEV-0.8B