wesamhamad/ qu-llm-assistant:latest

4 2 weeks ago

Fine-tuned Arabic student advisor for Qassim University. Two variants: Qwen2.5-1.5B (lightweight, 986MB) and ALLaM-7B by SDAIA (higher quality, 4.3GB). Trained on 15,903 Q&A pairs from official university documents. Apache-2.0 licensed. 

ollama run wesamhamad/qu-llm-assistant

Details

2 weeks ago

48a5091c4dbe · 986MB ·

model
qwen2
·
1.54B
·
Q4_K_M
template
{{ if .System }}<|im_start|>system {{ .System }}<|im_end|> {{ end }}{{ if .Prompt }}<|im_start|>user
system
أنت مساعد جامعة القصيم الذكي. أجب على أسئلة الطلاب بالل
params
{ "num_ctx": 4096, "stop": [ "<|im_end|>", "<|im_start|>" ], "temper

Readme

qu-llm-assistant

Arabic-native AI assistant fine-tuned to answer student questions about Qassim University policies, regulations, and services.

Model Variants

Variant Base Model Size Training Data Ollama
v1 (default) Qwen2.5-1.5B-Instruct 986 MB 15,903 Q&A pairs wesamhamad/qu-llm-assistant
v2 (ALLaM) ALLaM-7B-Instruct (SDAIA) 4.3 GB 12,320 Q&A pairs (2024-2026) wesamhamad/qu-llm-assistant-allam
  • v1 is lightweight and fast — ideal for quick responses on consumer hardware
  • v2 uses Saudi Arabia’s ALLaM model by SDAIA — better Arabic quality, trained exclusively on recent (2024-2026) university documents

Both variants use GGUF Q4_K_M quantization and are licensed under Apache 2.0.

Quick Start

# Lightweight variant (Qwen 1.5B — fast, ~1 GB)
ollama run wesamhamad/qu-llm-assistant

# Arabic-native variant (ALLaM 7B — higher quality, ~4.3 GB)
ollama run wesamhamad/qu-llm-assistant-allam

Training Data

  • Source corpus: 262 PDF documents + 800+ HTML pages from qu.edu.sa
  • 15,903 Q&A pairs (v1) generated with multi-stage quality filtering
  • 12,320 Q&A pairs (v2) filtered to 2024-2026 documents only
  • Quality pipeline: Deduplication, OCR artifact removal, garbled text detection, semantic similarity filtering, length checks
  • Coverage: Admissions, study regulations, examinations, discipline, cooperative training, student services, graduation, academic advising, transfers, complaints, and graduate studies

Fine-tuning

Property v1 (Qwen) v2 (ALLaM)
Method LoRA via Unsloth LoRA via Unsloth
LoRA config rank=16, alpha=32 rank=16, alpha=32
Epochs 3 1
Hardware Google Colab T4 (16 GB) Google Colab T4 (16 GB)
Training time ~2 hours ~30 minutes

Example Prompts

Prompt Category
ما هي شروط القبول في جامعة القصيم؟ Admissions
كم ساعة أقدر أسجل في الفصل كحد أقصى؟ Study Regulations
ما عقوبة الغش في الاختبارات النهائية؟ Discipline
وش متطلبات التدريب التعاوني وكيف أسجل فيه؟ Cooperative Training
كيف أقدم على التخرج وما هي الشروط المطلوبة؟ Graduation
ما هي شروط التحويل بين الكليات داخل الجامعة؟ Transfers
كيف أقدم شكوى أو تظلم على درجة في مادة؟ Complaints
ما هي الخدمات التي تقدمها عمادة شؤون الطلاب؟ Student Services

RAG Integration

For best results, pair this model with the full RAG pipeline:

  • 18,505 document chunks indexed in ChromaDB
  • BGE-M3 embeddings for multilingual semantic search
  • Gradio chat UI with source citations and confidence scores

Also Available On

Limitations

  • Answers are based on training data from 2024-2026 documents and may not reflect the latest policy changes
  • The model may occasionally generate plausible-sounding but inaccurate information
  • Best used as a supplementary tool alongside official university channels
  • Optimized for Qassim University context; may not generalize well to other institutions
  • Responses should be treated as informational guidance, not official university decisions

License

Apache 2.0

Author

Qassim University — Deanship of Information Technology / wesamhamad