LFM2-1.2B-RAG-ARABIC-LoRA

LFM2-1.2B-RAG Arabic (LoRA Fine-tuned)

Fine-tuned version of LiquidAI/LFM2-1.2B-RAG for Arabic reading comprehension and question answering tasks using LoRA (Low-Rank Adaptation) technique.

📋 Model Description

This model specializes in extractive question answering for Arabic text. It has been fine-tuned using LoRA on the Arabic Reading Comprehension Dataset (ARCD) to improve its ability to answer questions based on provided context in Modern Standard Arabic.

Key Features:

• Optimized for Arabic extractive QA
• Context-based question answering
• Maintains faithfulness to source documents
• Efficient fine-tuning via LoRA (rank=16)

🎯 Intended Use

Direct Use

• Arabic question answering systems
• RAG (Retrieval-Augmented Generation) applications for Arabic content
• Information extraction from Arabic documents
• Educational tools for Arabic reading comprehension

Downstream Use

Can be further fine-tuned for:

• Domain-specific QA (medical, legal, financial)
• Multi-turn conversational QA
• Document summarization with Q&A

Out-of-Scope Use

Not recommended for:

• Open-domain question answering without context
• Creative writing or content generation
• Translation tasks
• Code generation

🚀 How to Use

Basic Usage

import torch
from transformers import AutoModelForCausalLM, AutoTokenizer

# Load model and tokenizer
model_id = "azeddinShr/LFM2-1.2B-RAG-ARABIC-LoRA"
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    device_map="auto",
    torch_dtype=torch.bfloat16,
)
tokenizer = AutoTokenizer.from_pretrained(model_id)

# Prepare input
context = "نيوم هو مشروع ضخم في شمال غرب السعودية بتكلفة 500 مليار دولار."
question = "ما هي تكلفة مشروع نيوم؟"

prompt = f"استخدم السياق التالي للإجابة على السؤال:\n\n{context}\n\nالسؤال: {question}"

# Generate answer
messages = [{"role": "user", "content": prompt}]
input_ids = tokenizer.apply_chat_template(
    messages,
    add_generation_prompt=True,
    return_tensors="pt"
).to(model.device)

with torch.no_grad():
    outputs = model.generate(
        input_ids,
        max_new_tokens=150,
        temperature=0.0,
        do_sample=False,
        pad_token_id=tokenizer.eos_token_id
    )

answer = tokenizer.decode(outputs[0][input_ids.shape[1]:], skip_special_tokens=True)
print(answer)  # Output: 500 مليار دولار

📊 Training Details

Training Data

• Dataset: hsseinmz/arcd
• Training samples: 693
• Validation samples: 351
• Test samples: 351
• Language: Modern Standard Arabic
• Task: Extractive question answering

Training Procedure

Fine-tuning method: LoRA (Low-Rank Adaptation)

Hyperparameters:

• Base model: LiquidAI/LFM2-1.2B-RAG
• Epochs: 10
• Batch size: 16 (4 per device × 4 gradient accumulation)
• Learning rate: 2e-4
• Optimizer: AdamW (8-bit paged)
• LR scheduler: Cosine
• Warmup steps: 50
• Weight decay: 0.01
• LoRA rank (r): 16
• LoRA alpha: 32
• LoRA dropout: 0.05
• Target modules: w1, w2, w3, q_proj, k_proj, v_proj, out_proj, in_proj

Training infrastructure:

• Precision: bfloat16
• Gradient checkpointing: Enabled
• Framework: Hugging Face Transformers + PEFT + TRL

🔒 Ethical Considerations

• This model should not be used for generating misleading or false information
• Users should verify factual claims, especially for sensitive topics
• The model's responses reflect patterns in training data and may not represent complete or unbiased information

📜 Citation

If you use this model in your research or application, please cite:

@misc{lfm2-rag-arabic-lora,
  author = {Azeddin sahir},
  title = {LFM2-1.2B-RAG Arabic (LoRA Fine-tuned)},
  year = {2025},
  publisher = {Hugging Face},
  howpublished = {\url{https://huggingface.co/azeddinShr/lfm2-1.2b-arabic-qa-lora}}
}

👍🏻 Acknowledgments

• Base Model: LiquidAI for LFM2-1.2B-RAG
• Dataset: ARCD - Arabic Reading Comprehension Dataset
• Framework: Hugging Face Transformers, PEFT, TRL

📄 License

Same as based model

📧 Contact

For questions, issues, or collaboration opportunities, please open an issue in the model repository, contact via Hugging Face, or email me directly at [email protected].