Skip to content
Other

BERT для детекции токсичности

бинарный классификатор комментариев

2 min readupdated

Это bert-base-uncased, дообученный на английских комментариях из google/civil_comments для бинарной классификации: токсично или нет. Модель создана в рамках университетского проекта Universidad Bernardo O'Higgins (2026) как компонент блога с модерацией комментариев. Размер — около 0,11 млрд параметров, язык — только английский.

Ключевые особенности

  • Бинарная метка: toxico = 1, если исходная оценка токсичности ≥ 0,5; иначе 0.
  • Обучение на 49 925 комментариях после очистки; класс токсичных — 7,91%.
  • 5 эпох, learning rate 2e-5, max_length 256, fp16 на GPU T4, early stopping.
  • Лучшая модель выбиралась по F1 macro на валидации.
  • Предназначена для помощи модераторам, а не для автоматических санкций.

Пример

import torch
from transformers import AutoTokenizer, AutoModelForSequenceClassification
mid = 'bmerino10/vigia-bert-toxicidad'
tok = AutoTokenizer.from_pretrained(mid)
model = AutoModelForSequenceClassification.from_pretrained(mid).eval()
enc = tok('You are a complete idiot and nobody wants you here.', truncation=True, max_length=256, return_tensors='pt')
with torch.no_grad(): probs = torch.softmax(model(**enc).logits, dim=-1)[0]
print(f'Prob. tóxico: {probs[1]:.4f}')  # порог 0.5
ПодробнееНа тесте из 7 489 комментариев: accuracy 0,9424, F1 macro 0,8043, F1 токсичного клас��а 0,6399.
Класс токсичных: precision 0,63, recall 0,65 — примерно каждый третий флаг может быть ложным.
Только английский; возможны ошибки на сарказме, цитатах и контексте; аудит смещений не проводился.
Обучение: 5 эпох с early stopping, lr 2e-5, batch 8/16, weight decay 0,01, seed 42.
Discuss Telegram 0 62

Will it run?

Runs in F32

Fits without quantisation.

  • F32weights 0.4 GB · ready files0.9 GB
  • BF16weights 0.2 GB0.7 GB
  • FP8 / INT8recommendedweights 0.1 GB0.6 GB
  • 4-bitweights 0.1 GB0.6 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 · F32 — 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.

    # bmerino10/vigia-bert-toxicidad
    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 bmerino10/vigia-bert-toxicidad --local-dir ~/models/vigia-bert-toxicidad
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to vigia-bert-toxicidad
  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 bert-base-uncased (quants, finetunes, LoRAs) will show up under the bell.

Quantised versions 7All on Hugging Face
Fine-tunes 12All on Hugging Face
LoRAs and adapters 3All on Hugging Face

About the model

bmerino10/vigia-bert-toxicidad
Author
bmerino10
Category
Other
Released
Posted

On NeyroLearn

Hugging Face

620

Ask AI about this model

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

Comments

Other

d3-edge

d3-edge