[🌐 Website] •
[📜 Paper] •
[🤗 HF Models] •
[🐱 GitHub]
[🐦 Twitter] •
[💬 Reddit] •
[🍀 Unofficial Blog]
Repo for "ToRA: A Tool-Integrated Reasoning Agent for Mathematical Problem Solving"
Figure 1: Comparing ToRA with baselines on LLaMA-2 base models from 7B to 70B.
- [2023/10/08] 🔥🔥🔥 All ToRA models released at 🤗 HuggingFace!!!
- [2023/09/29] ToRA paper, repo, and website released.
ToRA is a series of Tool-integrated Reasoning Agents designed to solve challenging mathematical reasoning problems by interacting with tools, e.g., computation libraries and symbolic solvers. ToRA series seamlessly integrate natural language reasoning with the utilization of external tools, thereby amalgamating the analytical prowess of language and the computational efficiency of external tools.
Model | Size | GSM8k | MATH | AVG@10 math tasks† |
---|---|---|---|---|
GPT-4 | - | 92.0 | 42.5 | 78.3 |
GPT-4 (PAL) | - | 94.2 | 51.8 | 86.4 |
![]() |
7B | 68.8 | 40.1 | 62.4 |
![]() |
7B | 72.6 | 44.6 | 66.5 |
![]() |
13B | 72.7 | 43.0 | 65.9 |
![]() |
13B | 75.8 | 48.1 | 71.3 |
![]() |
34B | 80.7 | 51.0 | 74.8 |
![]() |
70B | 84.3 | 49.7 | 76.9 |
-
*ToRA-Code-34B is currently the first and only open-source model to achieve over 50% accuracy (pass@1) on the MATH dataset, which significantly outperforms GPT-4’s CoT result (51.0 vs. 42.5), and is competitive with GPT-4 solving problems with programs. By open-sourcing our codes and models, we hope more breakthroughs will come!
-
†10 math tasks include GSM8k, MATH, GSM-Hard, SVAMP, TabMWP, ASDiv, SingleEQ, SingleOP, AddSub, and MultiArith.
Figure 2: A basic example of single-round tool interaction, which interleaves rationales with program-based tool use.
Figure 3: Training ToRA contains ① Imitation Learning, and ② output space shaping.
We recommend using Conda to manage your environment. We use vLLM (0.1.4) to accelerate inference. Run the following commands to setup your environment:
git clone https://github.com/microsoft/ToRA.git && cd ToRA/src
conda create -n tora python=3.10
conda activate tora
pip install torch==2.0.1 --index-url https://download.pytorch.org/whl/cu118 # CUDA 11.8 for example
pip install -r requirements.txt
We provide a script for inference, simply config the MODEL_NAME_OR_PATH
and DATA
in src/scripts/infer.sh and run the following command:
bash scritps/infer.sh
We also open-source the model outputs from our best models (ToRA-Code-34B and ToRA-70B) in the src/outputs/
folder.
The src/eval/grader.py file contains the grading logic that assesses the accuracy of the predicted answer by comparing it to the ground truth. This logic is developed based on the Hendrycks' MATH grading system, which we have manually verified on the MATH dataset to minimize false positives and false negatives.
To evaluate the predicted answer, run the following command:
python -m eval.evaluate \
--data_name "math" \
--prompt_type "tora" \
--file_path "outputs/llm-agents/tora-code-34b-v1.0/math/test_tora_-1_seed0_t0.0_s0_e5000.jsonl" \
--execute
then you will get:
Num samples: 5000
Num scores: 5000
Timeout samples: 0
Empty samples: 2
Mean score: [51.0]
Type scores: {'Algebra': 67.3, 'Counting & Probability': 42.2, 'Geometry': 26.1, 'Intermediate Algebra': 40.0, 'Number Theory': 59.3, 'Prealgebra': 63.8, 'Precalculus': 34.2}
Due to some restrictions, ToRA-Corpus 16k is under review and cannot be released immediately. However, we open-source our complete training scripts as well as output space shaping pipelines for the community, and you may construct your own dataset for training.
To train a model, run the following command:
bash scritps/train.sh codellama 7b
If you find this repository helpful, please consider citing our paper:
@misc{gou2023tora,
title={ToRA: A Tool-Integrated Reasoning Agent for Mathematical Problem Solving},
author={Zhibin Gou and Zhihong Shao and Yeyun Gong and yelong shen and Yujiu Yang and Minlie Huang and Nan Duan and Weizhu Chen},
year={2023},
eprint={2309.17452},
archivePrefix={arXiv},
primaryClass={cs.CL}
}
This project welcomes contributions and suggestions. Most contributions require you to agree to a Contributor License Agreement (CLA) declaring that you have the right to, and actually do, grant us the rights to use your contribution. For details, visit https://cla.opensource.microsoft.com.
This project has adopted the Microsoft Open Source Code of Conduct. For more information see the Code of Conduct FAQ or contact opencode@microsoft.com with any additional questions or comments.