Instructions to use halilneed/turkish-pii-detection-v01 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use halilneed/turkish-pii-detection-v01 with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="halilneed/turkish-pii-detection-v01")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("halilneed/turkish-pii-detection-v01")
model = AutoModelForCausalLM.from_pretrained("halilneed/turkish-pii-detection-v01", device_map="auto")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use halilneed/turkish-pii-detection-v01 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "halilneed/turkish-pii-detection-v01"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "halilneed/turkish-pii-detection-v01",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/halilneed/turkish-pii-detection-v01

SGLang

How to use halilneed/turkish-pii-detection-v01 with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "halilneed/turkish-pii-detection-v01" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "halilneed/turkish-pii-detection-v01",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "halilneed/turkish-pii-detection-v01" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "halilneed/turkish-pii-detection-v01",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use halilneed/turkish-pii-detection-v01 with Docker Model Runner:
```
docker model run hf.co/halilneed/turkish-pii-detection-v01
```

halilatasoy/turkish-pii-detection-v01

Gemma-3-270m-it tabanlı Türkçe PII maskeleme modeli; zayıf dilimleri hedefleyen sentetik Türkçe bankacılık/ERP verisiyle devam eğitimi (continued full fine-tune) ile üretildi. 53 PII etiketi, talimat-koşullu maskeleme (tam / beyaz liste / kara liste / kapsam dışı).

Kod, veri üreteci ve değerlendirme: https://github.com/halilatasoy/turkish-pii-guard

Benchmark (halka açık 1000 satırlık Türkçe PII maskeleme benchmark'ı, satır düzeyi tam eşleşme)

Model	Tam eşleşme (1000)	Şema-nötr (903)
taban model (270m)	0.743	0.773
bu model	0.882	0.902

Kategori: beyaz_liste 0.900 · kapsam_disi 0.960 · kara_liste 0.893 · negatif 0.865 · tam 0.854

Zor dilimler: caps 0.802 · cok_kisi 0.600 · duz 0.930 · ekli 0.758 · kayit 0.976 · olmayan_etiket_talebi 0.917 · sozle 0.946 · tuzak 0.865 · uzun 0.656

Eğitim: 24.000 sentetik örnek, 1 epoch, full fine-tune, lr 5e-5,dk. Benchmark yalnızca test için kullanıldı; eğitim kalıpları benchmark cümle ve talimat kalıplarından bağımsız yazıldı, benchmark girdileriyle çakışan satırlar üretimde atıldı.

Kullanım

import torch
from transformers import AutoModelForCausalLM, AutoTokenizer

MODEL = "halilatasoy/turkish-pii-detection-v01"
tokenizer = AutoTokenizer.from_pretrained(MODEL)
model = AutoModelForCausalLM.from_pretrained(MODEL, dtype=torch.bfloat16, device_map="auto").eval()

def maskele(metin, talimat="Metindeki tüm kişisel verileri uygun etiketlerle maskele."):
    prompt = (f"{tokenizer.bos_token}<start_of_turn>user\n{talimat}\n\n"
              f"Metin: {metin}<end_of_turn>\n<start_of_turn>model\n")
    girdi = tokenizer(prompt, return_tensors="pt", add_special_tokens=False).to(model.device)
    with torch.inference_mode():
        cikti = model.generate(**girdi, max_new_tokens=512, do_sample=False,
                               eos_token_id=tokenizer.convert_tokens_to_ids("<end_of_turn>"),
                               pad_token_id=tokenizer.eos_token_id)
    return tokenizer.decode(cikti[0, girdi["input_ids"].shape[1]:], skip_special_tokens=True).strip()

print(maskele("müşteri Ayşe Yılmaz tc 12345678901 tel 0532 111 22 33"))
# -> müşteri [AD] tc [TCKN] tel [TEL]

Sınırlar

Değerlendirme sentetik veriyle yapılmıştır; gerçek kullanıcı metninde ölçmeden üretime alınmamalıdır. Şema dışı etiket üretebilir (53 etiketlik whitelist önerilir). KVKK/GDPR sorumluluğu kullanandadır. Lisans: Gemma Terms of Use.

Downloads last month: 418

Safetensors

Model size

0.3B params

Tensor type

BF16

Model tree for halilneed/turkish-pii-detection-v01

Base model

google/gemma-3-270m

Finetuned

google/gemma-3-270m-it

Finetuned

(1154)

this model