PneumoLLM: Harnessing the Power of Large Language Model for Pneumoconiosis Diagnosis

Updated on 2023.12.06

Key Features

This repository provides the official implementation of PneumoLLM: Harnessing the Power of Large Language Model for Pneumoconiosis Diagnosis.

  • New paradigm in applying large language models to diagnose data-scarce occupational diseases
  • The novel contextual multi-token engine and information emitter module to meticulously draw out knowledge from LLMs
  • Superiority in diagnosing pneumoconiosi, effectiveness of each designed module

Links

Preparation

  • Download LLaMA-7B from HuggingFace (unofficial).

Get Started

Training

  • Before training, please modify related parameters, e.g., exp_name, and check the related parameters, e.g., epochs, lr, batch_size, accum_iter.
  • Replace the data file and llama model path
--data_root
--llama_model_path
run train_acc.py

Validation

  • Before validation, please modify related parameters.
run eval_acc.py

[Checkpoint]Baidu Drive or Google Drive

📝 Citation

@article{song2023pneumollm,
  title={PneumoLLM: Harnessing the Power of Large Language Model for Pneumoconiosis Diagnosis},
  author={Song, Meiyue and Yu, Zhihua and Wang, Jiaxin and Wang, Jiarui and Lu, Yuting and Li, Baicun and Wang, Xiaoxu and Huang, Qinghua and Li, Zhijun and Kanellakis, Nikolaos I and others},
  journal={arXiv preprint arXiv:2312.03490},
  year={2023}
}