llama-3.2-3B-GGK(Grpo-Gsm8k): GRPO-Finetuned Llama-3.2-3B on GSM8K

This repository contains a Llama-3.2-3B model fine-tuned on GSM8K using the GRPO method, as described in DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models.

Checkpoints

• checkpoint-250/
• checkpoint-500/ (latest, recommended)

Each checkpoint contains:

• Adapter weights (adapter_model.safetensors)
• Tokenizer files
• Training state and configs

Model Card for outputs

This model is a fine-tuned version of unsloth/Llama-3.2-3B-Instruct. It has been trained using TRL.

Quick Start

Download the files first, then run the below code in inference.py

from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

tokenizer = AutoTokenizer.from_pretrained("grpo_gsm8k/checkpoint-500")
model = AutoModelForCausalLM.from_pretrained(
    "grpo_gsm8k/checkpoint-500",
    torch_dtype=torch.float16,
    device_map="auto"
)

prompt = "What is the sqrt of 101?"
inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
outputs = model.generate(
    **inputs,
    max_new_tokens=1024,
    temperature=0.4,
    top_p=0.95,
    repetition_penalty=1.15
)
response = tokenizer.decode(outputs[0], skip_special_tokens=True)
print(response)

Training procedure

This model was trained with GRPO, a method introduced in DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models.

Citations

@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},
}

@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 Gallou{\'e}dec},
    year         = 2020,
    journal      = {GitHub repository},
    publisher    = {GitHub},
    howpublished = {\url{https://github.com/huggingface/trl}}
}