← back to catalog · registered 2026-08-22 13:56

canyrtcn/Gemma_E4B_Kizagan_Abliterated

canyrtcn Gemma 8.0B
Your rig guess connected
? Why do I need an app?
Reading your rig…

This is a rough estimate. Install the free app - we'll show exact numbers.

Reading real hardware from your app right now. Numbers below are exact.

Below is the per-quantization compatibility for this model.

curl -H "Authorization: Bearer $ABL_KEY" \
     "https://abliteration.org/api/v1/models/canyrtcn%2FGemma_E4B_Kizagan_Abliterated"
Response includes
  • classification m1
  • files 9
  • hub_downloads_all_time 452
  • author_summary 2 models
  • readme_text full
10 credits · hourly refresh · ~4 KB payload Get an API key →
Abliteration classifier · v1.0.0
M1
Primary method

Direct removal

No other method signals detected in this model.
Confidence
MEDIUM
Why this label 3 signals
Method inferred from partial signals - repository name, related files, or tag patterns. Producer identity not confirmed; label may sharpen or shift as we gather more evidence.
  • 'abliterated' in name/tags
  • is_gguf=0 (base model)
  • no specific method indicators - defaulting to M1 (most common)
Refusal direction extracted via
Extraction technique

Difference-of-means

Confidence
MEDIUM
Why we say so
primary_method=M1; difference-of-means is the reference extraction for M1/M3 (Arditi 2024)
Downloads · lifetime
452
16 last 30d - cooling
Likes
3
Descendants
2
in 2 direct forks
Model age
3mo ago
created 2026-07-07

Training datasets

Corpora the author lists in the model card. Datasets tracked in our /datasets catalog carry a category badge linking to the workflow stage. Others open on Hugging Face.

Downloads over time
Now461→from298↑55%
290352415477298 on Jul 15461 on Oct 11JulAugSepOct
Jul 15 → Oct 11 · 53 snapshots · spans 88 days

Genealogy 2 direct forks

Full fork graph →

This model's place in the market. Above: what it was derived from. Below: the tree of everything derived from it.

Variants by this author 2 formats · 145 downloads combined

The same weights this author released in different packaging. Pick the format that matches your runtime.

Metadata

License
apache-2.0
Languages
tr en
Tags
transformers safetensors gemma4 image-text-to-text turkish abliterated gemma gemma-4 kizagan text-generation reasoning türkçe

Related

Total size
14.9 GB
Files
9
Quantizations
1
Registered
2026-08-22 13:56
Last updated on HF
2026-07-07 03:02

Files by quantization

Auxiliary files 9 files 14.9 GB
model.safetensors 14.9 GB 8cdb9dfd download
tokenizer.json 30.7 MB cc8d3a0c download
chat_template.jinja 16.9 KB c19999a3 download
README.md 5.19 KB 6bf3eaeb download
config.json 5.02 KB d68960fd download
tokenizer_config.json 2.05 KB 375b25dc download
processor_config.json 1.65 KB 5465974d download
.gitattributes 1.53 KB 52373fe2 download
generation_config.json 208 B e605bb45 download

README current version from Hugging Face


language:

  • tr
  • en
    license: apache-2.0
    license_name: apache-2.0
    license_link: https://www.apache.org/licenses/LICENSE-2.0
    tags:
  • turkish
  • abliterated
  • gemma
  • gemma-4
  • kizagan
  • text-generation
  • reasoning
  • türkçe
    pipeline_tag: text-generation
    base_model: AlicanKiraz0/Kizagan-E4B-Turkish-Reasoning-Model
    library_name: transformers
    datasets:
  • AlicanKiraz0/Kizagan-E4B-Turkish-Reasoning-Model

🔓 Gemma E4B Kızagan Abliterated

Türkçe muhakeme yeteneğine sahip, reddetme davranışı kaldırılmış (abliterated) açık kaynak dil modeli.

Bu model, AlicanKiraz0/Kizagan-E4B-Turkish-Reasoning-Model üzerinde refusal (reddetme) davranışını kaldırmak için abliteration işlemi uygulanmış versiyonudur.


🧬 Model Soy Ağacı

google/gemma-4-E4B-it
    └── AlicanKiraz0/Kizagan-E4B-Turkish-Reasoning-Model  (540K satır Türkçe SFT)
            └── canyrtcn/Gemma_E4B_Kizagan_Abliterated  ← BU MODEL (abliterated)

⚡ Özellikler

Özellik Değer
🇹🇷 Dil Türkçe (optimize), İngilizce
🧠 Temel Model Gemma 4 E4B-it (8B ham / 4B efektif PLE)
📚 Eğitim Kızagan SFT (540K satır, ~2B token Türkçe)
🔓 Abliteration Reddetme davranışı kaldırıldı
📐 Parametre 7.5B toplam / 4B efektif (MoE-style PLE)
📏 Context 131.072 token
⚖️ Lisans Apache 2.0 (Gemma 4 şartlarına tabi)

🔓 Abliterasyon Hakkında

Bu model, reddetme (refusal) davranışını hedef alan katmanlardaki aktivasyonları modifiye ederek, modelin içerik filtreleme eğilimini kaldıracak şekilde abliterate edilmiştir. Modelin genel yetenekleri korunmuştur.

Not: Reddetme davranışı kaldırıldığı için model, verilen prompt'lara göre filtrelenmemiş içerik üretebilir. Kullanım sorumluluğu size aittir.


🚀 Kullanım

🤗 Transformers ile

from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

model_id = "canyrtcn/Gemma_E4B_Kizagan_Abliterated"

tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    torch_dtype=torch.bfloat16,
    device_map="auto",
)

messages = [
    {"role": "user", "content": "Bir yapay zeka modeli nasıl eğitilir? Adım adım açıkla."},
]

inputs = tokenizer.apply_chat_template(
    messages,
    add_generation_prompt=True,
    return_tensors="pt",
).to(model.device)

outputs = model.generate(
    inputs,
    max_new_tokens=1024,
    temperature=0.7,
    top_p=0.9,
    do_sample=True,
)

print(tokenizer.decode(outputs[0][inputs.shape[-1]:], skip_special_tokens=True))

🦙 llama.cpp (GGUF)

# GGUF versiyonu için: canyrtcn/Gemma_E4B_Kizagan_Abliterated-GGUF
llama-cli -hf canyrtcn/Gemma_E4B_Kizagan_Abliterated-GGUF \
  --chat-template gemma

🖥️ LM Studio

Repoyu aratıp indirebilirsiniz.

🐳 Ollama

ollama run hf.co/canyrtcn/Gemma_E4B_Kizagan_Abliterated-GGUF

📊 GGUF Quantization Seçenekleri

GGUF versiyonlar şu repoda: canyrtcn/Gemma_E4B_Kizagan_Abliterated-GGUF

Quant Dosya Boyutu Kalite Önerilen Kullanım
Q4_K_M ~5 GB Çok iyi 8GB VRAM, dengeli
F16 ~15 GB Kayıpsız 16GB+ VRAM, maksimum kalite

⚙️ Önerilen Parametreler

Parametre Değer
temperature 0.3
top_p 0.95
min_p 0.05
top_k 20
repeat_penalty 1.05

⚠️ Sınırlamalar

  • Türkçe odaklıdır — diğer dillerde performans garanti edilmez
  • Küçük model (E4B) — çok uzun bağlamlarda ve ileri uzmanlık alanlarında büyük modellerin yerini tutmaz
  • Reddetme davranışı kaldırıldı — filtrelenmemiş içerik üretebilir, kullanım sorumluluğu size ait
  • Halüsinasyon — tüm LLM'ler gibi yanlış bilgi üretebilir

📜 Lisans

Bu model Apache License 2.0 altında yayınlanmıştır.

Temel model olarak kullanılan google/gemma-4-E4B-it Google'ın Gemma Terms of Use koşullarına tabidir.


🙏 Atıf ve Teşekkür

@misc{kiraz2026kizagan,
  author       = {Alican Kiraz},
  title        = {Kızagan-E4B: A Turkish Reasoning Model Fine-Tuned from Gemma 4 E4B},
  year         = {2026},
  publisher    = {Hugging Face},
  howpublished = {\url{https://huggingface.co/AlicanKiraz0/Kizagan-E4B-Turkish-Reasoning-Model}},
}

@misc{yurtacan2026kizagan_abliterated,
  author       = {Can Yurtaçan},
  title        = {Gemma E4B Kızagan Abliterated — Reddetme Davranışı Kaldırılmış Türkçe Muhakeme Modeli},
  year         = {2026},
  publisher    = {Hugging Face},
  howpublished = {\url{https://huggingface.co/canyrtcn/Gemma_E4B_Kizagan_Abliterated}},
}

Ok yaydan çıktıysa, isabetle dönecek. 🔓

README history 6 versions

The author's README evolved over time. Click a version to see its content at that point.

  1. 2026-07-07Upload README.md with huggingface_hub5c42b875.2 KB
    Loading...
  2. 2026-07-07Upload README.md with huggingface_huba39d3d35.2 KB
    Loading...
  3. 2026-07-07Upload README.md with huggingface_hubb17f3555.2 KB
    Loading...
  4. 2026-07-07Upload README.md with huggingface_hub530bf785.1 KB
    Loading...
  5. 2026-07-07Upload README.md with huggingface_hubd42e6eb5 KB
    Loading...
  6. 2026-07-07Upload folder using huggingface_hubdc8cf635.5 KB
    Loading...
Catalog is the map. Apps are the tools.

Run models on your own machine, not in the cloud.

Every model page has an "Open in Abliteration" button that hands the model directly to the first-party desktop client, at the quantization your rig can actually run. No API keys, no subscription, no prompt leakage.

Open in Abliteration