Skip to content
Other

看帖识字 Kantie Shizi v1

4,96 млн параметров для 4 167 иероглифов

1 min read

Модель распознаёт один вырезанный иероглиф китайской каллиграфии среди 4 167 типов знаков в пяти стилях: 楷, 行楷, 行, 草 и 隶. Это CNN на 4,96 млн параметров и 20 МБ, обученная с нуля на оцифрованных работах Национального дворца-музея в Тайбэе. Запускается в браузере и на телефоне.

Ключевые особенности

  • Всего 4,96 млн параметров и 20 МБ: есть ONNX и safetensors, инференс через onnxruntime.
  • 4 167 типов знаков; на вход подаётся одно изображение, на выходе — вероятности по всем классам.
  • Обучена с нуля на 213 753单字 из 1 385 произведений; ещё 346 работ держали отдельно для оценки.
  • Без шрифтов, без pretrained backbone и без выходов других моделей; данные музея — CC BY 4.0.
  • Принимает цветное или灰度 изображение любого размера; preprocessing уже включён в kantie.py.

Пример

from kantie import Kantie
k = Kantie()
print(k.predict('一个字.png', topk=5))
# [('書', 0.93), ('晝', 0.02), ...]
ПодробнееНа 500 знаках от 61 каллиграфа, не встречавшегося в обучении: 94,0% точности, для 草书 — 82,0%.
На отложенных работах: 95,9% top-1 и 98,5% top-5.
Обучение: 60 эпох, AdamW, OneCycle, около 5 часов на DGX Spark.
Ограничения: только вырезанный одиночный знак; печатные шрифты и ручка распознаются плохо; только 4 167 известных знаков; 草书 слабее всего.
Discuss Telegram 0

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 0.0 GB · ready files0.5 GB
  • FP8 / INT8recommendedweights 0.0 GB0.5 GB
  • 4-bitweights 0.0 GB0.5 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.

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

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

About the model

svoic/kantie-shizi
Author
svoic
Category
Other
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