|
|
--- |
|
|
license: apache-2.0 |
|
|
language: |
|
|
- en |
|
|
base_model: |
|
|
- Qwen/Qwen2.5-Math-1.5B |
|
|
pipeline_tag: text-generation |
|
|
library_name: transformers |
|
|
tags: |
|
|
- math |
|
|
- qwen |
|
|
- gsm8k |
|
|
- lora |
|
|
--- |
|
|
# OpenMath |
|
|
Fine-tuning a Small Language Model (SLM) for Step-by-Step Math Reasoning |
|
|
|
|
|
## Overview |
|
|
**OpenMath** is an open-source project focused on fine-tuning a **small language model (SLM)** to solve **math word problems** with clear, step-by-step reasoning. |
|
|
The project uses **LoRA/QLoRA fine-tuning** on popular math reasoning datasets and provides a benchmarking pipeline to compare performance against other open-source SLMs/LLMs. |
|
|
|
|
|
This project is designed to be reproducible on **free Colab (T4) GPU**. |
|
|
|
|
|
--- |
|
|
|
|
|
## What’s Included |
|
|
- QLoRA fine-tuning code (4-bit) |
|
|
- GSM8K subset training (example: 1k samples) |
|
|
- GSM8K evaluation script (accuracy) |
|
|
- Saved LoRA adapter weights |
|
|
|
|
|
--- |
|
|
|
|
|
## Base Model |
|
|
- **Qwen2.5-Math-1.5B** |
|
|
|
|
|
--- |
|
|
|
|
|
## Dataset |
|
|
- **GSM8K** (Grade School Math 8K) |
|
|
- Training used: **1000 samples** |
|
|
- Evaluation: GSM8K test split |
|
|
|
|
|
--- |
|
|
|
|
|
## Results |
|
|
### Training Setup (Current) |
|
|
- Samples: 1000 |
|
|
- Epochs: 6 |
|
|
- Max length: 1024 |
|
|
- LoRA rank: 16 |
|
|
- Loss masking: trained mainly on the **solution portion** to improve reasoning |
|
|
|
|
|
### Accuracy |
|
|
- **GSM8K Accuracy (100-sample test subset): 41%** |
|
|
|
|
|
> Note: The 41% score was measured on a **100-question subset** of the GSM8K test set for faster evaluation on Colab. |
|
|
|
|
|
--- |
|
|
|
|
|
## GSM8K Leaderboard (Baseline) |
|
|
|
|
|
| Model | Params | GSM8K Accuracy (%) | |
|
|
|------|--------|---------------------| |
|
|
| LLaMA 2 | 13B | 28.7 | |
|
|
| Gemma 2 (PT) | 2B | 23.9 | |
|
|
| Mistral (Base) | 7B | 36.5 | |
|
|
| ERNIE 4.5 | 21B | 25.2 | |
|
|
| Baichuan (Base) | 13B | 26.6 | |
|
|
| Gemma | 7B | 46.4 | |
|
|
| Zephyr-7b-gemma-v0.1 | 7B | 45.56 | |
|
|
| LLaMA 3.2 Instruct (CoT) | 1B | 39.04 | |
|
|
| Gemma 3 IT | 1B | 42.15 | |
|
|
| Qwen 3 (Instruct mode) | 1.7B | 33.66 | |
|
|
| **OpenMath (Qwen2.5-Math-1.5B + LoRA)** | **1.5B** | **41.0** | |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
<img width="1090" height="590" alt="image" src="https://github.com/user-attachments/assets/662ea336-8946-4542-b2f2-eb78712d5a2d" /> |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
--- |
|
|
|
|
|
## Repository Files |
|
|
### LoRA Adapter Folder |
|
|
This project provides the fine-tuned adapter weights: |
|
|
|
|
|
- `adapter_model.safetensors` → LoRA weights |
|
|
- `adapter_config.json` → LoRA configuration |
|
|
|
|
|
> Note: This is **not a full model**. |
|
|
> You must load the **base model** and then attach the adapter. |
|
|
|
|
|
--- |
|
|
|
|
|
## Disclaimer |
|
|
OpenMath is an educational/research project. |
|
|
|
|
|
The fine-tuned model may produce incorrect, incomplete, or misleading answers. |
|
|
|
|
|
Always verify solutions independently before using them for exams, assignments, or real-world decisions. |
|
|
|
|
|
This project does **not** guarantee correctness and should not be used as a substitute for professional advice. |
|
|
|
|
|
--- |
|
|
|
|
|
## Contributing |
|
|
Contributions are welcome! 🎉 |
|
|
|
|
|
If you’d like to contribute: |
|
|
1. Fork the repository |
|
|
2. Create a new branch (`feature/your-feature-name`) |
|
|
3. Commit your changes |
|
|
4. Open a Pull Request |
|
|
|
|
|
### Contribution Ideas |
|
|
- Run full GSM8K test evaluation (1319 samples) and report results |
|
|
- Train on larger GSM8K subsets (3k–5k samples) |
|
|
- Add SVAMP / ASDiv datasets for better generalization |
|
|
- Improve decoding to reduce repetition |
|
|
- Add a Streamlit demo for interactive testing |
|
|
- Benchmark against more open-source SLMs/LLMs |
|
|
- Improve evaluation scripts and metrics |
|
|
|
|
|
--- |
|
|
|
|
|
## Note |
|
|
OpenMath is a **fun and practical side project** built to explore **efficient fine-tuning (QLoRA)** and **math reasoning evaluation** on limited compute. |
|
|
|
|
|
The goal is to learn, experiment, and share reproducible results — while keeping the code clean and open for community improvements. |
|
|
|
|
|
--- |
|
|
|
|
|
## License |
|
|
This project is licensed under the **Apache License 2.0**. |
