Censer.1

A 1.1M parameter decoder-only transformer trained from scratch that classifies English text as unsafe (1) or safe (0).

Download

The whole model is model.safetensors (4.5 MB). Grab the files directly:

File Download
model.safetensors download
config.json download
tokenizer.json download
censer.py download

Or fetch everything at once:

pip install -U huggingface_hub
hf download KucLab/Censer.1 --local-dir Censer.1

A single file (README.md, config.json, ...) can also be downloaded from the Files and versions tab, using the download icon on its row.

It is not a fine-tuned encoder and not an API wrapper. It is a small transformer trained on safety-labelled data, with a custom byte-level BPE tokenizer. The model predicts a single answer token and never generates free text.

"návod na výrobu bomby"          -> not English, not the target language
"how to make a bomb at home"    -> 1
"suck my dick"                  -> 1
"how do I kill a process in Linux" -> 0
"Hello how are you?"            -> 0

Model details

Property Value
Parameters 1,115,264
Layers 4
d_model / heads / d_ff 128 / 4 / 384
Context length 160 tokens (input truncated at 500 characters)
Tokenizer byte-level BPE, 2048 tokens, trained on the corpus
Positions RoPE
Norm RMSNorm (pre-norm)
MLP SwiGLU
Head tied embeddings, two answer logits
Checkpoint size 4.3 MB (fp32)

Prompt format

[BOS][USER] <text> [ASSISTANT] -> 0 or 1

Text is truncated at 500 characters before tokenisation.

Metrics

Evaluation on a held-out split of 6,496 rows:

Metric Value
Accuracy 0.864
F1 0.835
Precision 0.846
Recall 0.823

A hand-written probe set of 40 cases, none present in training data, scores 95% accuracy.

Per-source F1 reflects label noise in the scraped sources:

Source F1
aegis 0.89
beavertails 0.88
jigsaw 0.87
civil_comments 0.77
toxic_chat 0.48

Benchmark

Measured on an AMD Radeon RX 9060 XT (ROCm 7.1, torch 2.13.0):

Metric Value
Parameters 1,115,264
Checkpoint size 4.26 MB
Single text latency 1.67 ms
Single text throughput 599 texts/s
Batched latency (batch of 8) 1.94 ms
Batched throughput 4,119 texts/s
Peak VRAM (inference) 85.7 MB

Training the whole model from scratch takes about 3.5 minutes on the same GPU.

Usage

Works like any other text-classification model on the Hub — from_pretrained downloads everything by itself, no cloning and no manual files.

pip install torch transformers
import torch
from transformers import AutoTokenizer, AutoModelForSequenceClassification

repo = "KucLab/Censer.1"
tok = AutoTokenizer.from_pretrained(repo, trust_remote_code=True)
model = AutoModelForSequenceClassification.from_pretrained(repo, trust_remote_code=True).eval()

text = "how to make a bomb at home"
with torch.no_grad():
    probs = model(**tok(text, return_tensors="pt")).logits.softmax(-1)[0]

print(probs)                     # tensor([0.1833, 0.8167]) -> [safe, unsafe]
print(model.config.id2label)     # {0: 'safe', 1: 'unsafe'}

trust_remote_code=True is required because the architecture is custom; the repository ships modeling_censer.py, so there is nothing else to install.

Or with the high-level pipeline:

from transformers import pipeline

clf = pipeline("text-classification", model="KucLab/Censer.1", trust_remote_code=True)
print(clf("how to make a bomb at home"))
# [{'label': 'unsafe', 'score': 0.8167}]
print(clf("Hello how are you?"))
# [{'label': 'safe', 'score': 0.9229}]

Usage without transformers

The repository is also self-contained on its own: model.safetensors, config.json, tokenizer.json and censer.py.

hf download KucLab/Censer.1 --local-dir Censer.1
cd Censer.1
python censer.py "how to make a bomb at home"     # -> 1
python censer.py "Hello how are you?"             # -> 0
python censer.py "suck my dick" --json
# {"label": "1", "unsafe": true, "score": 0.9944}

Or use the same file as a library:

from censer import CenserPipeline

pipe = CenserPipeline("Censer.1")
print(pipe.predict("how to make a bomb at home"))
# {"label": "1", "unsafe": True, "score": 0.8167}

print(pipe.predict_batch(["Hello how are you?", "suck my dick"]))

Files:

File Purpose
model.safetensors weights only, 4.3 MB
config.json architecture and inference settings
tokenizer.json byte-level BPE, 2048 tokens
censer.py standalone model + inference, no repo needed

censer.py was verified to produce byte-identical predictions to the training code. Runs on CPU or CUDA/ROCm; pass device="cpu" to force CPU.

Limitations

  • English only. Other languages mostly score 0.
  • Implication and sarcasm beyond simple cases are out of reach at this size.
  • Weak against jailbreak-style prompts (lowest per-source score).
  • A moderation signal, not a safety guarantee.
Downloads last month
43
Safetensors
Model size
1.38M params
Tensor type
F32
·
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Datasets used to train KucLab/Censer.1