songkaisong/Graph_Toolformer

Graph-Toolformer

Graph-ToolFormer: To Empower LLMs with Graph Reasoning Ability via Prompt Augmented by ChatGPT

Paper URL at IFMLab: http://www.ifmlab.org/files/paper/graph_toolformer.pdf

Paper description in Chinese: 文章中文介绍

Project Conda Environment

See the shared environment.yml file. Create a local environment at your computer with command

conda env create -f environment.yml

For the packages cannot be instead with the above conda command, you may consider to install via pip instead.

References

(also add \usepackage{hyperref} into your paper for adding the url address in the reference)

@article{zhang2023graph,
  title={Graph-ToolFormer: To Empower LLMs with Graph Reasoning Ability via Prompt Augmented by ChatGPT},
  author={Zhang, Jiawei},
  note = {\url{http://www.ifmlab.org/files/paper/graph_toolformer.pdf}},
  year={2023},
  note = "[Online; Published 10-April-2023]"
}

---

Tasks to be done

• Polish the framework: 8/9 done
  • add working memory module
  • add query parser module
  • add query excutor module
  • add graph dataset hub
  • add graph model hub
  • add graph reasoning task hub
  • add llm model hub
  • add llm trainer module
  • add cli for interaction
• Expand the framework: 0/4
  • Add and test more LLMs
  • Include more graph datasets
  • Add more pre-trained graph models
  • Include more graph reasoning tasks
• Release the framework and service: 0/4
  • Implement the CLI with GUI panel
  • Provide the demo for graph reasoning
  • Release web access and service
  • Add API for customerized graph reasoning

---

Organization of the code?

By adding the needed dataset, the model can run by calling "python gtoolformer_gptj_script.py".

A simpler template of the code is also available at http://www.ifmlab.org/files/template/IFM_Lab_Program_Template_Python3.zip

The whole program is divided into five main parts:

(1) data.py (for data loading and basic data organization operators, defines abstract method load() )

(2) method.py (for complex operations on the data, defines abstract method run() )

(3) result.py (for saving/loading results from files, defines abstract method load() and save() )

(4) evaluate.py (for result evaluation, defines abstract method evaluate() )

(5) setting.py (for experiment settings, defines abstract method load_run_save_evaluate() )

The base class of these five parts are defined in ./code/base_class/, they are all abstract class defining the templates and architecture of the code.

The inherited class are provided in ./code, which inherit from the base classes, implement the abstract methonds.

Detailed information on funtional classes?

a. data

(1) DatasetLoader.py (for dataset loading)

b. method

(1) Method_Graph_Toolformer_GPTJ.py (Graph-Toolformer Implementation with GPT-J 6b 8bit)

(2) Module_GPTJ_8bit.py (The GPT-J module quantization with 8 bit)

c. result

(1) ResultSaving.py (for saving results to file)

(2) Result_Loader.py (for loading result from file)

d. evaluate

(1) Evaluate_Metrics.py (implementation of evaluation metrics used in this experiment, include rouge scores, bleu scores, etc.)

e. setting

(1) Setting_Regular.py (defines the interactions and data exchange among the above classes)

---