OpenMath2-Llama3.1-8B-nemo

NeMo checkpoint for OpenMath2-Llama3.1-8B is obtained by finetuning Llama3.1-8B-Base with OpenMathInstruct-2.

The model outperforms Llama3.1-8B-Instruct on all the popular math benchmarks we evaluate on, especially on MATH by 15.9%.

Model	GSM8K	MATH	AMC 2023	AIME 2024	Omni-MATH
Llama3.1-8B-Instruct	84.5	51.9	9/40	2/30	12.7
OpenMath2-Llama3.1-8B (nemo | HF)	91.7	67.8	16/40	3/30	22.0
+ majority@256	94.1	76.1	23/40	3/30	24.6
Llama3.1-70B-Instruct	95.8	67.9	19/40	6/30	19.0
OpenMath2-Llama3.1-70B (nemo | HF)	94.9	71.9	20/40	4/30	23.1
+ majority@256	96.0	79.6	24/40	6/30	27.6

The pipeline we used to produce the data and models is fully open-sourced!

• Code
• Models
• Dataset

See our paper to learn more details!

How to use the models?

Our models are trained with the same "chat format" as Llama3.1-instruct models (same system/user/assistant tokens). Please note that these models have not been instruction tuned on general data and thus might not provide good answers outside of math domain.

This is a NeMo checkpoint, so you need to use NeMo Framework to run inference or finetune it. We also release a HuggingFace checkpoint and provide easy instructions on how to convert between different formats or run inference with these models using our codebase.

Reproducing our results

We provide all instructions to fully reproduce our results.

Citation

If you find our work useful, please consider citing us!

@article{toshniwal2024openmath2,
  title   = {OpenMathInstruct-2: Accelerating AI for Math with Massive Open-Source Instruction Data},
  author  = {Shubham Toshniwal and Wei Du and Ivan Moshkov and  Branislav Kisacanin and Alexan Ayrapetyan and Igor Gitman},
  year    = {2024},
  journal = {arXiv preprint arXiv:2410.01560}
}

Terms of use

By accessing this model, you are agreeing to the LLama 3.1 terms and conditions of the license, acceptable use policy and Meta’s privacy policy