wesamhamad/ qassim-gpt

3 Downloads Updated 2 weeks ago

Fine-tuned Qwen2.5-1.5B for Qassim University student advisory. Arabic-native RAG-powered assistant trained on 3,919 Q&A pairs from university policies, regulations, and student guides.

ollama run wesamhamad/qassim-gpt

curl http://localhost:11434/api/chat \
  -d '{
    "model": "wesamhamad/qassim-gpt",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

from ollama import chat

response = chat(
    model='wesamhamad/qassim-gpt',
    messages=[{'role': 'user', 'content': 'Hello!'}],
)
print(response.message.content)

import ollama from 'ollama'

const response = await ollama.chat({
  model: 'wesamhamad/qassim-gpt',
  messages: [{role: 'user', content: 'Hello!'}],
})
console.log(response.message.content)

Models

Name

1 model

Size

Context

Input

qassim-gpt:latest

986MB · 32K context window · Text · 2 weeks ago

qassim-gpt:latest

986MB

32K

Text

Readme

Qassim-GPT

A fine-tuned Arabic language model for Qassim University student services. Built on Qwen2.5-1.5B-Instruct with LoRA fine-tuning on university-specific data.

Quick Start

ollama run wesamhamad/qassim-gpt

Model Details

Spec	Value
Base Model	Qwen2.5-1.5B-Instruct
Fine-tuning	LoRA (rank=16, 3 epochs)
Training Data	3,919 curated Q&A pairs
Quantization	Q4_K_M (GGUF)
Languages	Arabic (primary), English
License	MIT

Training Data

777 HTML pages scraped from qu.edu.sa
262 PDF documents (policies, regulations, student guides)
3,919 filtered Q&A pairs covering admissions, registration, exams, transfers, scholarships, and more

Use Cases

Student advisory chatbot
University policy Q&A
Academic regulation lookup
Admissions guidance

Example

>>> ما شروط القبول في جامعة القصيم؟
>>> كيف أقدر أعتذر عن الفصل الدراسي؟
>>> ما عقوبة الغش في الاختبارات؟

RAG Integration

For best results, pair this model with the full RAG pipeline: - 18,505 document chunks in ChromaDB - BGE-M3 embeddings for Arabic semantic search - Gradio chat UI with citations

Built By

Qassim University — Deanship of Information Technology

Acknowledgments

Qwen2.5 by Alibaba
Unsloth for efficient fine-tuning
Ollama for local model serving

# Qassim-GPT

A fine-tuned Arabic language model for Qassim University student services. Built on Qwen2.5-1.5B-Instruct with LoRA fine-tuning on university-specific data.

## Quick Start

```bash
ollama run wesamhamad/qassim-gpt
```

## Model Details

| Spec | Value |
|------|-------|
| Base Model | Qwen2.5-1.5B-Instruct |
| Fine-tuning | LoRA (rank=16, 3 epochs) |
| Training Data | 3,919 curated Q&A pairs |
| Quantization | Q4_K_M (GGUF) |
| Languages | Arabic (primary), English |
| License | MIT |

## Training Data

- **777 HTML pages** scraped from qu.edu.sa
- **262 PDF documents** (policies, regulations, student guides)
- **3,919 filtered Q&A pairs** covering admissions, registration, exams, transfers, scholarships, and more

## Use Cases

- Student advisory chatbot
- University policy Q&A
- Academic regulation lookup
- Admissions guidance

## Example

```
>>> ما شروط القبول في جامعة القصيم؟
>>> كيف أقدر أعتذر عن الفصل الدراسي؟
>>> ما عقوبة الغش في الاختبارات؟
```

## RAG Integration

For best results, pair this model with the full RAG pipeline:
- **18,505 document chunks** in ChromaDB
- **BGE-M3 embeddings** for Arabic semantic search
- **Gradio chat UI** with citations

## Built By

Qassim University — Deanship of Information Technology

## Acknowledgments

- [Qwen2.5](https://github.com/QwenLM/Qwen2.5) by Alibaba
- [Unsloth](https://github.com/unslothai/unsloth) for efficient fine-tuning
- [Ollama](https://ollama.com) for local model serving

Paste, drop or click to upload images (.png, .jpeg, .jpg, .svg, .gif)