--- base_model: QCRI/Fanar-1-9B-Instruct datasets: AI-MO/NuminaMath-TIR library_name: transformers model_name: Fanar-0.5B-GRPO-test tags: - generated_from_trainer - trl - grpo - math - reasoning - R1 licence: license license: apache-2.0 language: - ar - en --- # ๐Ÿง  Fanar-Math-R1-GRPO **Fanar-Math-R1-GRPO** is a reasoning-optimized language model built on [`QCRI/Fanar-1-9B-Instruct`](https://huggingface.co/QCRI/Fanar-1-9B-Instruct). This version is fine-tuned using **Group Relative Policy Optimization (GRPO)** from the DeepSeekMath framework on the [`AI-MO/NuminaMath-TIR`](https://huggingface.co/datasets/AI-MO/NuminaMath-TIR) dataset. It is designed for step-by-step mathematical problem-solving with structured reasoning in both English and Arabic. --- ## ๐Ÿš€ Model Highlights - ๐Ÿ” Fine-tuned with **GRPO**, a sample-efficient reinforcement learning method - ๐Ÿงฎ Specializes in **multi-step mathematical reasoning** - ๐Ÿ’ฌ Outputs responses in a structured conversational format using `` and `` tags - ๐Ÿง  Trained using **TRL** (`transformers`, `peft`, and `math_verify`) - ๐Ÿท๏ธ Useful for both instruction-following and math-heavy dialogue generation --- ## ๐Ÿ“ฆ Model Details | Component | Description | |------------------|-----------------------------------------------------------------------------| | **Base Model** | [`QCRI/Fanar-1-9B-Instruct`](https://huggingface.co/QCRI/Fanar-1-9B-Instruct) | | **Fine-Tuning** | GRPO via Hugging Face [TRL](https://github.com/huggingface/trl) | | **Dataset** | [`AI-MO/NuminaMath-TIR`](https://huggingface.co/datasets/AI-MO/NuminaMath-TIR) | | **Format** | ` ... ... ` tagged reasoning structure | | **LoRA** | Enabled (modules: `q_proj`, `v_proj`, rank=8) | | **Epochs** | 1 (lightweight test configuration) | | **Tokenizer** | Same as base model | --- ## ๐Ÿงช Inference Example ```python from transformers import AutoTokenizer, AutoModelForCausalLM import torch import time model_id = "Omartificial-Intelligence-Space/Fanar-Math-R1-GRPO" model = AutoModelForCausalLM.from_pretrained(model_id, torch_dtype="auto", device_map="auto") tokenizer = AutoTokenizer.from_pretrained(model_id) def generate_with_reasoning(prompt_text): inputs = tokenizer(prompt_text, return_tensors="pt").to(model.device) start = time.time() with torch.no_grad(): output = model.generate(**inputs, max_length=1024) end = time.time() generated = tokenizer.decode(output[0], skip_special_tokens=True) duration = end - start num_input_tokens = inputs["input_ids"].shape[1] num_generated_tokens = output.shape[1] - num_input_tokens return generated, duration, num_generated_tokens # Example Arabic math problem prompt = """A conversation between User and Assistant. The user asks a question, and the Assistant solves it. The assistant first thinks about the reasoning process in the mind and then provides the user with the answer either in Arabic or English based on user's language. The reasoning process and answer are enclosed within and tags, respectively, i.e., reasoning process here answer here ููŠ ู…ุฏูŠู†ุฉ ูŠุจู„ุบ ุนุฏุฏ ุณูƒุงู†ู‡ุง 1 ู…ู„ูŠูˆู† ู†ุณู…ุฉุŒ ุฅุฐุง ูƒุงู† 60% ู…ู† ุงู„ุณูƒุงู† ุจุงู„ุบูŠู†ุŒ ูˆ40% ู…ู† ุงู„ุจุงู„ุบูŠู† ูŠุนู…ู„ูˆู†ุŒ ููƒู… ุนุฏุฏ ุงู„ุนุงู…ู„ูŠู† ููŠ ุงู„ู…ุฏูŠู†ุฉุŸ""" result, time_taken, tokens = generate_with_reasoning(prompt) print(result) ``` --- ## ๐Ÿ› ๏ธ Training Setup ### Configuration Summary - **learning_rate**: 1e-5 - **epochs**: 1 - **max_completion_length**: 64 - **num_generations**: 4 - **gradient_accumulation_steps**: 16 - **logging_steps**: 10 ### Reward Functions - **accuracy_reward**: validates correctness of the answer using `math_verify` - **format_reward**: checks for proper usage of `` and `` tags ### Libraries & Versions ``` transformers==4.47.1 trl==0.14.0 peft==0.14.0 datasets==2.21.0 math_verify==0.3.3 torch==2.4.1 ``` --- ## ๐Ÿ“Š Training Metrics (Snapshot) | Step | Reward (avg) | Accuracy Reward | Format Reward | Loss | KL Divergence | |------|--------------|-----------------|---------------|-------|---------------| | 10 | 0.029 | 0.029 | 0.0 | 0.0 | 0.00024 | | 100 | 0.039 | 0.039 | 0.0 | 0.0001| 0.00188 | | 200 | 0.033 | 0.033 | 0.0 | 0.0001| 0.00183 | | 300 | 0.045 | 0.045 | 0.0 | 0.0001| 0.00127 | *Note: Training was run with a small config for notebook-friendly experimentation.* --- ## ๐Ÿ“š Output Format The model is trained to follow a reasoning-first format: ``` First, we calculate 60% of 1 million, which is 600,000. Then, 40% of that is 240,000. 240,000 ``` --- ## ๐Ÿ”ฌ Citations ### GRPO โ€“ DeepSeekMath ```bibtex @article{zhihong2024deepseekmath, title={DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models}, author={Shao, Zhihong and Wang, Peiyi and Zhu, Qihao and Xu, Runxin and Song, Junxiao and Zhang, Mingchuan and Li, Y.K. and Wu, Y. and Guo, Daya}, journal={arXiv preprint arXiv:2402.03300}, year={2024} } ``` ### TRL Library ```bibtex @misc{vonwerra2022trl, title={TRL: Transformer Reinforcement Learning}, author={von Werra, Leandro and Belkada, Younes and Tunstall, Lewis and Beeching, Edward and Thrush, Tristan and Lambert, Nathan and Huang, Shengyi and Rasul, Kashif and Gallouรฉdec, Quentin}, year={2022}, howpublished={\url{https://github.com/huggingface/trl}} } ``` --- ## ๐Ÿ”— Resources - [DeepSeekMath Paper](https://arxiv.org/abs/2402.03300) - [TRL Documentation](https://huggingface.co/docs/trl) - [Open-R1 Project](https://github.com/huggingface/open-r1) --- ## ๐Ÿง‘โ€๐Ÿ”ฌ Authors Developed and trained by **Omar Paniego** with adaptation of the DeepSeek-R1 training recipe using Hugging Face's open tools and datasets. --- ## ๐Ÿ“ข License Refer to the license file in the repository. --- ## โค๏ธ Acknowledgements Thanks to: - **Hugging Face Science Team** for `trl` and `math_verify` - **AI-MO** for the NuminaMath-TIR dataset - **DeepSeek Team** for releasing their methodology and insights Happy reasoning! ๐Ÿ”โœจ ## Citations Cite GRPO as: ```bibtex @article{zhihong2024deepseekmath, title = {{DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models}}, author = {Zhihong Shao and Peiyi Wang and Qihao Zhu and Runxin Xu and Junxiao Song and Mingchuan Zhang and Y. K. Li and Y. Wu and Daya Guo}, year = 2024, eprint = {arXiv:2402.03300}, } ``` Cite TRL as: ```bibtex @misc{vonwerra2022trl, title = {{TRL: Transformer Reinforcement Learning}}, author = {Leandro von Werra and Younes Belkada and Lewis Tunstall and Edward Beeching and Tristan Thrush and Nathan Lambert and Shengyi Huang and Kashif Rasul and Quentin Gallouรฉdec}, year = 2020, journal = {GitHub repository}, publisher = {GitHub}, howpublished = {\url{https://github.com/huggingface/trl}} } ```