mtma / q6-k

Q6_K

تم التحديث منذ أسبوعين

جاري جلب البيانات...
بطاقة البيانات (README)
# 🧠 Suri-Qwen 3.1 4B (Uncensored) | نسخة Q6_K عالية الدقة ## 📄 Model Summary | ملخص النموذج This is the **GGUF** version of the **Suri-Qwen-3.1-4B-Uncensored** model, quantized using **Q6_K** method. هذا النموذج هو نسخة مضغوطة بصيغة `GGUF` من الموديل القوي "Suri-Qwen". يتميز هذا الإصدار بأنه **غير مقيد (Uncensored)**، مما يمنحه حرية أكبر في الإجابة على الأسئلة دون القيود الصارمة المعتادة في النماذج التجارية. تم اختيار دقة **Q6_K** تحديداً لضمان أعلى مستوى من "الذكاء" وفهم اللغة العربية، مع الحفاظ على سرعة التشغيل. --- ## 💎 Why Q6_K? | لماذا هذه النسخة؟ Because this is a **4B parameter model** (small & fast), using standard compression (like Q4) might degrade its reasoning capabilities. * **Q6_K (This File):** Retains **99.9%** of the original model's intelligence. It is the perfect balance for running a "smart" assistant on consumer hardware. * **Size:** ~3.3 GB (Lightweight). --- ## 📊 Technical Details | تفاصيل تقنية * **Model Architecture:** Qwen 2.5/3.1 (4 Billion Parameters). * **Format:** GGUF (Compatible with `llama.cpp`, `LM Studio`, `Ollama`). * **Quantization:** Q6_K (6-bit High Precision). * **Type:** Uncensored / Instruction Following. * **Languages:** Arabic, English. --- ## 🛠 How to Run | طريقة التشغيل You can load this model easily using `llama-cpp-python` or any GGUF runner: ```python from llama_cpp import Llama # Load the model (Run on CPU or GPU) llm = Llama( model_path="./Suri-Qwen-3.1-4B-Uncensored-Hard.i1-Q6_K.gguf", n_ctx=4096, # Context window n_gpu_layers=-1 # Set to -1 to offload to GPU ) # Chat example response = llm.create_chat_completion( messages=[ {"role": "system", "content": "أنت مساعد ذكي تتحدث العربية بطلاقة."}, {"role": "user", "content": "اشرح لي نظرية النسبية باختصار."} ] ) print(response['choices'][0]['message']['content'])
Installation
$ pip install oneurai
تم النسخ
download.py
from oneurai import load_dataset
import os

# اسم الداتا ست
repo_id = "mtma/q6-k"

print("🚀 Starting the download process...")

dataset_path = load_dataset(repo_id)

if dataset_path:
    print(f"\n✅ تمت العملية بنجاح!")
    print(f"📂 مسار البيانات: {dataset_path}")
else:
    print("\n❌ فشل التحميل.")
تم نسخ الكود بنجاح!

نظرة عامة

التحميلات 0
الحجم 3.08 GB
نوع المهمة Text Classification
MTMA
ناشر معتمد