Instructions to use Alamori/GoldenNet-Qwen2.5-0.5B-QLoRA-v1 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Alamori/GoldenNet-Qwen2.5-0.5B-QLoRA-v1 with Transformers:

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

pipe = pipeline("text-generation", model="Alamori/GoldenNet-Qwen2.5-0.5B-QLoRA-v1")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Alamori/GoldenNet-Qwen2.5-0.5B-QLoRA-v1")
model = AutoModelForCausalLM.from_pretrained("Alamori/GoldenNet-Qwen2.5-0.5B-QLoRA-v1", 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 Alamori/GoldenNet-Qwen2.5-0.5B-QLoRA-v1 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Alamori/GoldenNet-Qwen2.5-0.5B-QLoRA-v1"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Alamori/GoldenNet-Qwen2.5-0.5B-QLoRA-v1",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Alamori/GoldenNet-Qwen2.5-0.5B-QLoRA-v1

SGLang

How to use Alamori/GoldenNet-Qwen2.5-0.5B-QLoRA-v1 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 "Alamori/GoldenNet-Qwen2.5-0.5B-QLoRA-v1" \
    --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": "Alamori/GoldenNet-Qwen2.5-0.5B-QLoRA-v1",
		"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 "Alamori/GoldenNet-Qwen2.5-0.5B-QLoRA-v1" \
        --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": "Alamori/GoldenNet-Qwen2.5-0.5B-QLoRA-v1",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Alamori/GoldenNet-Qwen2.5-0.5B-QLoRA-v1 with Docker Model Runner:
```
docker model run hf.co/Alamori/GoldenNet-Qwen2.5-0.5B-QLoRA-v1
```

GoldenNet-Qwen2.5-0.5B-QLoRA-v1

Model Description

GoldenNet-Qwen2.5-0.5B-QLoRA-v1 is a fine-tuned version of Qwen/Qwen2.5-0.5B-Instruct specialized for Iraqi Government Correspondence Processing.

This model performs two key tasks:

Document Classification - Classifies government correspondence into 8 categories
Named Entity Recognition - Extracts entities like persons, organizations, locations, dates, monetary values, and legal references

Supported Categories (التصنيفات)

Arabic	English	Description
طلب	Request	Formal requests for approval, resources, or actions
شكوى	Complaint	Grievances and complaints from citizens or departments
تقرير	Report	Status reports, statistics, and progress updates
إعلام	Notification	Official announcements and notifications
استفسار	Inquiry	Questions seeking information or clarification
دعوة	Invitation	Invitations to events, meetings, or conferences
تعميم	Circular	Directives and circulars from higher authorities
إحالة	Referral	Document referrals to other departments

Training Details

Configuration

Parameter	Value
Base Model	Qwen/Qwen2.5-0.5B-Instruct
Fine-tuning Method	QLoRA (4-bit quantization + LoRA)
LoRA Rank	64
LoRA Alpha	128
LoRA Dropout	0.05
Learning Rate	2e-4
Epochs	3
Batch Size	2 (effective: 16 with gradient accumulation)
Max Sequence Length	2048
Precision	BF16

Training Results

Metric	Value
Training Loss	0.448
Evaluation Loss	0.2998
Training Time	~49 seconds
Hardware	NVIDIA RTX 5070 (8GB VRAM)

Loss Progression

Epoch 1: 0.912
Epoch 2: 0.319
Epoch 3: 0.200

Usage

With Transformers (Python)

from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained(
    "Alamori/GoldenNet-Qwen2.5-0.5B-QLoRA-v1",
    device_map="auto",
    torch_dtype="auto"
)
tokenizer = AutoTokenizer.from_pretrained(
    "Alamori/GoldenNet-Qwen2.5-0.5B-QLoRA-v1"
)

# Example: Classification
correspondence = """جمهورية العراق
وزارة التربية
مديرية تربية بغداد
العدد: 4521/ت/2025
التاريخ: 2025/05/15

إلى/ السيد مدير عام التعليم العام المحترم

م/ طلب تعيين معلمين

تحية طيبة...

نرجو الموافقة على تعيين 50 معلماً في المدارس الابتدائية.

مع التقدير
مدير التربية"""

instruction = "صنّف المراسلة الحكومية التالية إلى إحدى الفئات: طلب، شكوى، تقرير، إعلام، استفسار، دعوة، تعميم، إحالة. أجب بصيغة JSON تتضمن الفئة ودرجة الثقة والتبرير."

messages = [
    {"role": "user", "content": f"{instruction}\n\n{correspondence}"}
]

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

outputs = model.generate(**inputs, max_new_tokens=256, temperature=0.1)
response = tokenizer.decode(outputs[0][inputs['input_ids'].shape[1]:], skip_special_tokens=True)
print(response)
# Output: {"category": "طلب", "confidence": 0.96, "reasoning": "المراسلة تطلب تعيين موظفين..."}

With Ollama

# Create the model
ollama create goldennet-iraqi-gov -f Modelfile

# Run inference
ollama run goldennet-iraqi-gov

With vLLM

vllm serve Alamori/GoldenNet-Qwen2.5-0.5B-QLoRA-v1 --port 8000

Example Outputs

Classification Task

Input:

صنّف المراسلة الحكومية التالية...
[تعميم من مجلس الوزراء بشأن الدوام الرسمي]

Output:

{"category": "تعميم", "confidence": 0.97, "reasoning": "المراسلة تعميم قانوني يتضمن توجيهات إلزامية"}

Entity Extraction Task

Input:

استخرج جميع الكيانات المسماة من المراسلة الحكومية التالية...
[تقرير صحي من دائرة صحة البصرة]

Output:

{
  "persons": ["السيد وزير الصحة", "د. سعاد الموسوي"],
  "organizations": ["وزارة الصحة", "دائرة صحة البصرة"],
  "locations": ["محافظة البصرة", "الزبير", "الفاو"],
  "dates": ["2025/06/10"],
  "reference_numbers": ["7823/ص/2025"],
  "monetary_values": ["5 مليار دينار"],
  "quantities": ["3 مراكز صحية", "120 طبيباً"],
  "projects": [],
  "laws_regulations": []
}

Limitations

Optimized specifically for Iraqi government correspondence format
Best performance on formal Arabic administrative documents
May require adaptation for other Arabic dialects or document types
Recommended max input length: 2048 tokens

Intended Use

Government document processing and automation
Administrative workflow optimization
Document routing and prioritization
Metadata extraction from official correspondence
Research on Arabic NLP for government applications

Ethical Considerations

This model is designed for legitimate government administrative purposes. Users should:

Ensure compliance with data privacy regulations
Use appropriate access controls for sensitive documents
Validate model outputs before making critical decisions
Not use for surveillance or unauthorized data collection

Citation

@misc{goldennet-qwen-qlora-v1,
  author = {Golden Net AI},
  title = {GoldenNet-Qwen2.5-0.5B-QLoRA-v1: Iraqi Government Correspondence Classifier},
  year = {2025},
  publisher = {Hugging Face},
  url = {https://huggingface.co/Alamori/GoldenNet-Qwen2.5-0.5B-QLoRA-v1}
}

About Golden Net AI

Golden Net AI is dedicated to developing AI solutions for Arabic language processing, with a focus on government and enterprise applications in Iraq and the MENA region.

License

This model is released under the Apache 2.0 License.

Developed by Golden Net AI
Empowering Iraqi Government Digital Transformation

Downloads last month: 12

Safetensors

Model size

0.5B params

Tensor type

BF16

Model tree for Alamori/GoldenNet-Qwen2.5-0.5B-QLoRA-v1

Base model

Qwen/Qwen2.5-0.5B

Finetuned

Qwen/Qwen2.5-0.5B-Instruct

Finetuned

(937)

this model

Evaluation results

Eval Loss
self-reported

0.300