Skip to content
Computer vision

NetraOCR-1.0

единый OCR на 1,27B для Indic-скриптов и рукописных таблиц

1 min readupdated

NetraOCR-1.0 — мультимодальная OCR-модель от dipankar500 на 1 273 967 936 параметров, построенная на Qwen3.5-0.8B с гибридной архитектурой Gated-DeltaNet + full attention. Она объединяет IndicOCR и OvisOCR3 в один чекпоинт через Branch-Train-MiX и Dual Interleaved Vision Token Fusion.

Ключевые особенности

  • Два замороженных vision-энкодера: OvisOCR3 ViT + merger и IndicOCR ViT + merger; токены чередуются с 2-frame M-RoPE.
  • 24-слойный декодер: 18 слоёв linear-attention GatedDeltaNet и 6 слоёв full attention.
  • В каждом слое FFN — soft-routed MoE с двумя замороженными экспертами (IndicOCR и OvisOCR3) и обучаемым роутером.
  • Единый словарь на 303 101 токен: 248 070 токенов OvisOCR3/Qwen3.5 и 55 031 токен чистых Indic-скриптов.
  • Режим two-pass region-crop: split_page_regions() режет плотные рукописные страницы на горизонтальные полосы перед area_clamp.

Пример

import torch
from PIL import Image
from modeling_netra import NetraForConditionalGeneration
from processing_netra import NetraProcessor, area_clamp
model = NetraForConditionalGeneration.from_pretrained("dipankar500/NetraOCR-1.0", torch_dtype=torch.bfloat16).cuda().eval()
proc = NetraProcessor.from_pretrained("dipankar500/NetraOCR-1.0", frames=2)
batch = {k: v.cuda() for k, v in proc([proc.prompt("Extract all text from the image.")], [area_clamp(Image.open("sample.png").convert("RGB"))]).items()}
out = model.generate(**batch, max_new_tokens=512, do_sample=False)
ПодробнееПараметров — 1 273 967 936; база — Qwen3.5-0.8B.
ViT-бэкбоны обоих энкодеров полностью заморожены, обучаются MLP-мерджеры и роутеры.
Код использует CUDA и torch.bfloat16; данных о VRAM и бенчмарках в карточке нет.
Discuss Telegram 0

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.

    # dipankar500/NetraOCR-1.0
    # ===== 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 dipankar500/NetraOCR-1.0 --local-dir ~/models/NetraOCR-1.0
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to NetraOCR-1.0
  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 NetraOCR-1.0 (quants, finetunes, LoRAs) will show up under the bell.

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

About the model

dipankar500/NetraOCR-1.0
Author
dipankar500
Category
Computer vision
Released
Posted

On NeyroLearn

Hugging Face

00

Ask AI about this model

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

Comments