PULSE-7bv5 Open-source Medical Large Language Model - Free for Full-scenario Medical Language Processing

PULSE 7bv5

Developed by OpenMEDLab

Medical large language model optimized based on Bloom 7B architecture, covering full-scenario medical language processing tasks

Large Language Model

Transformers

Chinese#Medical NLP Full Stack #Chinese Medical Instruction Fine-tuning #Medical Licensing Exam Support

Downloads 112

Release Time : 6/25/2023

Model Overview

PULSE is an open-source large language model focused on the medical field, deeply optimized with approximately 4 million medical and general domain instruction data points, possessing comprehensive medical NLP capabilities.

Model Features

Full-stack Medical NLP Capabilities

Covers full-scenario medical language processing tasks including health education, medical licensing exams, report parsing, electronic medical record structuring, and diagnostic simulation.

Massive Training Data

Deeply optimized with approximately 4 million medical and general domain instruction fine-tuning data points.

Leading in Chinese Medical Evaluation

Performs excellently in Chinese medical evaluations, with Elo scores surpassing similar medical large models.

Model Capabilities

Medical text generation

Medical Q&A

Report parsing

Electronic medical record structuring

Diagnostic simulation

Health education

Medical licensing exam assistance

Use Cases

Medical Education

Medical Student Exam Assistance

Helps medical students prepare for medical licensing qualification exams

Achieves high accuracy in mock exams

Clinical Assistance

Electronic Medical Record Structuring

Converts unstructured electronic medical records into structured data

🚀 PULSE

PULSE is a model that supports various natural language processing tasks in the medical field, fine - tuned with a large amount of SFT data.

🚀 Quick Start

This README provides detailed information about the PULSE model, including its features, limitations, evaluation results, inference requirements, and usage examples.

✨ Features

Large - scale Training: The PULSE model is further fine - tuned on the basis of the Bloom 7B model using approximately 4,000,000 SFT data from the medical and general domains.
Comprehensive Medical NLP Tasks: PULSE supports a variety of natural language processing tasks in the medical field, including health education, physician exam questions, report interpretation, medical record structuring, and simulated diagnosis and treatment.

📦 Installation

Hardware Requirements

The following table provides the GPU memory required for local deployment of PULSE for inference with a batch size of 1.

Property	Details
Quantization Level	Memory for Loading Model
FP16	14GB

Steps

Download the repository to your local/remote server.

git clone https://github.com/openmedlab/PULSE
cd PULSE

Create a conda environment and install dependencies.

conda env create -f llm.yml
conda activate llm

It is not recommended to use versions of torch and transformers lower than the recommended versions.

💻 Usage Examples

Web Demo

Gradio

python web_demo_gradio.py

Command - line Demo

You can run cli_demo.py in the repository to start a simple command - line demo:

python cli_demo.py

📚 Documentation

Limitations

Due to the relatively small number of model parameters and the autoregressive generation paradigm, although the model provides inference results regarding disease diagnosis and treatment, these results cannot replace the advice and treatment plans of offline professional doctors. All responses are for reference only and should not be used as a basis for diagnosis or treatment. We strongly recommend that users seek the help and advice of professional doctors when they need to diagnose or treat diseases.

Elo Evaluation

model_name	model_size	ALL	MedQA_Mainland	PromptCBLUE	webMedQA
GPT4	220B*8(?)	1195	1087	1134	1107
ChatGPT	175B(?)	1123	1053	1089	1067
PULSE_7b with prompt	7B	1074	1019	1047	1060
PULSE_14b	14B	1055	1001	1037	1056
PULSE_7b	7B	1054	1028	1037	1030
BianQue	6B	926	939	920	1011
QiZhenGPT	13B	918	949	935	974
Med - ChatGLM	6B	864	988	921	859
BenTsao	7B	846	966	913	859
DoctorGLM	6B	812	935	891	856

📄 License

The code in this project is licensed under the Apache 2.0 license, and the model weights are licensed under the GNU AGPL 3.0 license. If the model and its modified versions in this project are used to provide services that generate misleading or harmful remarks and cause adverse effects, the service provider shall be responsible, and it has nothing to do with this project.

Acknowledgments

Shanghai Artificial Intelligence Laboratory
Qingyuan Research Institute, Shanghai Jiao Tong University
Natural Language Processing and Big Data Mining Laboratory, East China University of Science and Technology

Open - Source Models

[PULSE - 7bv5](https://huggingface.co/OpenMEDLab/PULSE - 7bv5)

Featured Recommended AI Models

Empowering the Future, Your AI Solution Knowledge Base

English 简体中文繁體中文にほんご