Llama 3.1 – Islamic Education (Yemen, Grade 7) – LoRA
Model Description
This model is a fine-tuned version of Meta Llama 3.1 (8B) using LoRA adapters.
It is specialized in Islamic Education content for Yemeni 7th grade students.
Training Data
The model was trained on a custom local dataset manually collected from
the Yemeni Islamic Education textbook (Grade 7).
The dataset was converted into instruction-based question–answer format (JSON).
Training Method
- Fine-tuning method: LoRA (PEFT)
- Library: Unsloth
- Base model: Meta Llama 3.1 8B
- Quantization: 4-bit
- Epochs: 1
- Hardware: Google Colab (Tesla T4)
Purpose
The purpose of this model is educational support and answering questions
related to Islamic Education curriculum in Yemen.
Limitations
The model is domain-specific and should not be used outside the intended
educational context.
