Skip to content
Other

NSFW Detector v2

крошечный фильтр с правкой 0,01% весов

2 min read

Pankaj8922/experiment — экспериментальный бинарный классификатор изображений: выдаёт вероятность NSFW. Это MobileNetV3-small на 1,53 млн параметров; по сути — отредактированная версия Pankaj8922/nsfw-detector-v2: изменено 151 число из 1 532 018 (0,01%), без переобучения сети.

Ключевые особенности

  • Оценка NSFW: 0 = sfw, 1 = nsfw; вход 224×224 RGB.
  • Правка только финального слоя: 150 весов + 1 bias; остальные параметры бит-в-бит как у оригинала.
  • На двух независимых тестах AUC вырос: 0,768 → 0,946 и 0,865 → 0,944.
  • Меньше ложных срабатываний: обычные фото 87% → 20%, пустые кадры 41% → 0%, neutral 30% → 9% при p>0,5.
  • Слабое место — аниме и рисунки: около половины всё ещё уходит в флаг или ревью.
  • Три режима по порогам: allow / human review / block; пороги в config.json.

Пример

import torch, timm
from PIL import Image
import torchvision.transforms as T
model = timm.create_model('mobilenetv3_small_100', pretrained=False, num_classes=2)
model.load_state_dict(torch.load('pytorch_model.bin', map_location='cpu')); model.eval()
tf = T.Compose([T.Resize(224), T.CenterCrop(224), T.ToTensor(), T.Normalize([0.485,0.456,0.406],[0.229,0.224,0.225])])
x = tf(Image.open('image.jpg').convert('RGB')).unsqueeze(0)
with torch.no_grad(): p = torch.softmax(model(x), dim=1)[0, 1].item()  # P(nsfw)
ПодробнееAUC: 0,768 → 0,946 (dataset 1) и 0,865 → 0,944 (dataset 2).
Модель крошечная: 1,53 млн параметров, вход 224×224, запуск в примере на CPU.
Оценка на двух датасетах, а не на реальном трафике; выборка скромная: 380 и около 630 изображений.
Ограничения: center crop, не оценена fairness, не юридический инструмент для поиска незаконного контента; аниме и рисунки — слабое место.
Discuss Telegram 0 28

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.

    # Pankaj8922/experiment
    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 Pankaj8922/experiment --local-dir ~/models/experiment
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to experiment
  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. mobilenetv3_small_100.lamb_in1kfamily root
  2. nsfw-detector-v2
  3. experiment

New posts about versions of mobilenetv3_small_100.lamb_in1k (quants, finetunes, LoRAs) will show up under the bell.

Quantised versions 6All on Hugging Face
Fine-tunes 12All on Hugging Face

About the model

Pankaj8922/experiment
Author
Pankaj8922
Category
Other
Released
Posted

On NeyroLearn

Hugging Face

280

Ask AI about this model

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

Comments