nvidia/OpenMath-CodeLlama-34b-Python-hf

1---2license: llama23base_model:4- codellama/CodeLlama-34b-Python-hf5datasets:6- nvidia/OpenMathInstruct-17language:8- en9tags:10- nvidia11- code12- math13---14 15 16# OpenMath-CodeLlama-34b-Python-hf17 18OpenMath models were designed to solve mathematical problems by integrating text-based reasoning with code blocks19executed by Python interpreter. The models were trained on [OpenMathInstruct-1](https://huggingface.co/datasets/nvidia/OpenMathInstruct-1),20a math instruction tuning dataset with 1.8M problem-solution pairs generated using permissively licensed21[Mixtral-8x7B](https://huggingface.co/mistralai/Mixtral-8x7B-v0.1) model.22 23<table border="1">24  <tr>25    <td></td>26    <td colspan="2" style="text-align: center;">greedy</td>27    <td colspan="2" style="text-align: center;">majority@50</td>28  </tr>29  <tr>30    <td style="text-align: center;">model</td>31    <td style="text-align: center;">GSM8K</td>32    <td style="text-align: center;">MATH</td>33    <td style="text-align: center;">GMS8K</td>34    <td style="text-align: center;">MATH</td>35  </tr>36  <tr>37    <td style="text-align: right;">OpenMath-CodeLlama-7B (<a href="https://huggingface.co/nvidia/OpenMath-CodeLlama-7b-Python">nemo</a> | <a href="https://huggingface.co/nvidia/OpenMath-CodeLlama-7b-Python-hf">HF</a>)</td>38    <td style="text-align: center;">75.9</td>39    <td style="text-align: center;">43.6</td>40    <td style="text-align: center;">84.8</td>41    <td style="text-align: center;">55.6</td>42  </tr>43  <tr>44    <td style="text-align: right;">OpenMath-Mistral-7B (<a href="https://huggingface.co/nvidia/OpenMath-Mistral-7B-v0.1">nemo</a> | <a href="https://huggingface.co/nvidia/OpenMath-Mistral-7B-v0.1-hf">HF</a>)</td>45    <td style="text-align: center;">80.2</td>46    <td style="text-align: center;">44.5</td>47    <td style="text-align: center;">86.9</td>48    <td style="text-align: center;">57.2</td>49  </tr>50  <tr>51    <td style="text-align: right;">OpenMath-CodeLlama-13B (<a href="https://huggingface.co/nvidia/OpenMath-CodeLlama-13b-Python">nemo</a> | <a href="https://huggingface.co/nvidia/OpenMath-CodeLlama-13b-Python-hf">HF</a>)</td>52    <td style="text-align: center;">78.8</td>53    <td style="text-align: center;">45.5</td>54    <td style="text-align: center;">86.8</td>55    <td style="text-align: center;">57.6</td>56  </tr>57  <tr>58    <td style="text-align: right;">OpenMath-CodeLlama-34B (<a href="https://huggingface.co/nvidia/OpenMath-CodeLlama-34b-Python">nemo</a> | <a href="https://huggingface.co/nvidia/OpenMath-CodeLlama-34b-Python-hf">HF</a>)</td>59    <td style="text-align: center;">80.7</td>60    <td style="text-align: center;">48.3</td>61    <td style="text-align: center;">88.0</td>62    <td style="text-align: center;">60.2</td>63  </tr>64  <tr>65    <td style="text-align: right;">OpenMath-Llama2-70B (<a href="https://huggingface.co/nvidia/OpenMath-Llama-2-70b">nemo</a> | <a href="https://huggingface.co/nvidia/OpenMath-Llama-2-70b-hf">HF</a>)</td>66    <td style="text-align: center;"><b>84.7</b></td>67    <td style="text-align: center;">46.3</td>68    <td style="text-align: center;">90.1</td>69    <td style="text-align: center;">58.3</td>70  </tr>71  <tr>72    <td style="text-align: right;">OpenMath-CodeLlama-70B (<a href="https://huggingface.co/nvidia/OpenMath-CodeLlama-70b-Python">nemo</a> | <a href="https://huggingface.co/nvidia/OpenMath-CodeLlama-70b-Python-hf">HF</a>)</td>73    <td style="text-align: center;">84.6</td>74    <td style="text-align: center;"><b>50.7</b></td>75    <td style="text-align: center;"><b>90.8</b></td>76    <td style="text-align: center;"><b>60.4</b></td>77  </tr>78</table>79 80The pipeline we used to produce these models is fully open-sourced!81 82- [Code](https://github.com/Kipok/NeMo-Skills)83- [Models](https://huggingface.co/collections/nvidia/openmath-65c5619de2ba059be0775014)84- [Dataset](https://huggingface.co/datasets/nvidia/OpenMathInstruct-1)85 86See our [paper](https://arxiv.org/abs/2402.10176) for more details!87 88# How to use the models?89 90Try to [run inference with our models](https://github.com/Kipok/NeMo-Skills/blob/main/docs/inference.md) with just a few commands!91 92# Reproducing our results93 94We provide [all instructions](https://github.com/Kipok/NeMo-Skills/blob/main/docs/reproducing-results.md) to fully reproduce our results.95 96# Improving other models97 98To improve other models or to learn more about our code, read through the docs below.99 100- [NeMo-Skills Pipeline](https://github.com/Kipok/NeMo-Skills)101    - [Generating synthetic data](https://github.com/Kipok/NeMo-Skills/blob/main/docs/synthetic-data-generation.md)102    - [Finetuning models](https://github.com/Kipok/NeMo-Skills/blob/main/docs/finetuning.md)103    - [Evaluating models](https://github.com/Kipok/NeMo-Skills/blob/main/docs/evaluation.md)104 105In our pipeline we use [NVIDIA NeMo](https://www.nvidia.com/en-us/ai-data-science/generative-ai/nemo-framework/),106an end-to-end, cloud-native framework to build, customize, and deploy generative AI models anywhere.107It includes training and inferencing frameworks, guardrailing toolkits, data curation tools, and pretrained models,108offering enterprises an easy, cost-effective, and fast way to adopt generative AI.109 110# Citation111 112If you find our work useful, please consider citing us!113 114```bibtex115@article{toshniwal2024openmath,116  title   = {OpenMathInstruct-1: A 1.8 Million Math Instruction Tuning Dataset},117  author  = {Shubham Toshniwal and Ivan Moshkov and Sean Narenthiran and Daria Gitman and Fei Jia and Igor Gitman},118  year    = {2024},119  journal = {arXiv preprint arXiv: Arxiv-2402.10176}120}121```122 123# License124 125The use of this model is governed by the [Llama 2 Community License Agreement](https://ai.meta.com/llama/license/)