JSL-MedPhi2-2.7B Open-Source Medical Large Language Model - Empowering Medical Scenario Applications with an Optimized Architecture

JSL MedPhi2 2.7B

Developed by johnsnowlabs

A large language model for the medical domain developed by John Snow Labs, optimized based on Microsoft Phi-2 architecture

Large Language Model

Transformers

Other#Medical Q&A #Medical Knowledge Reasoning #Clinical Decision Support

Downloads 3,050

Release Time : 4/30/2024

Model Overview

A 2.7B-parameter large language model specialized in the medical field, suitable for tasks like medical Q&A and clinical knowledge reasoning

Model Features

Medical Domain Optimization

Specially trained and optimized for medical professional content

Efficient Reasoning

Based on Phi-2 architecture, maintaining good performance with a smaller parameter count

Professional Evaluation

Comprehensively evaluated on multiple medical professional datasets

Model Capabilities

Medical text generation

Medical Q&A systems

Clinical knowledge reasoning

Medical literature comprehension

Use Cases

Medical Education

Medical Exam Assistance

Helping medical students prepare for professional exams

Achieved 38.99% accuracy on the medmcqa dataset

Clinical Decision Support

Clinical Knowledge Q&A

Answering clinical-related professional questions

Achieved 63.4% accuracy on the MMLU clinical knowledge subset

Medical Research

Literature Q&A

Understanding and answering medical literature-related questions

Achieved 73% accuracy on the PubMedQA dataset

🚀 JSL-MedPhi2-2.7B

A medical large language model developed by John Snow Labs based on the microsoft/phi2 base model.

This model is developed by John Snow Labs.

This model is available under a CC-BY-NC-ND license and must also conform to this Acceptable Use Policy. If you need to license this model for commercial use, please contact us at info@johnsnowlabs.com.

🚀 Quick Start

This section provides a quick guide on how to use the JSL-MedPhi2-2.7B model.

💻 Usage Examples

Basic Usage

!pip install -qU transformers accelerate

from transformers import AutoTokenizer
import transformers
import torch

model = "johnsnowlabs/JSL-MedPhi2-2.7B"
messages = [{"role": "user", "content": "What is a large language model?"}]

tokenizer = AutoTokenizer.from_pretrained(model)
prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
pipeline = transformers.pipeline(
    "text-generation",
    model=model,
    torch_dtype=torch.float16,
    device_map="auto",
)

outputs = pipeline(prompt, max_new_tokens=256, do_sample=True, temperature=0.7, top_k=50, top_p=0.95)
print(outputs[0]["generated_text"])

🏆 Evaluation

The following tables show the evaluation results of the JSL-MedPhi2-2.7B model on various tasks.

Tasks	Version	Filter	Metric	Value		Stderr
stem	N/A	none	acc_norm	0.3904	±	0.0066
		none	acc	0.4402	±	0.0058
- medmcqa	Yaml	none	acc	0.3899	±	0.0075
		none	acc_norm	0.3899	±	0.0075
- medqa_4options	Yaml	none	acc	0.3920	±	0.0137
		none	acc_norm	0.3920	±	0.0137
- anatomy (mmlu)	0	none	acc	0.4815	±	0.0432
- clinical_knowledge (mmlu)	0	none	acc	0.6340	±	0.0296
- college_biology (mmlu)	0	none	acc	0.6181	±	0.0406
- college_medicine (mmlu)	0	none	acc	0.5665	±	0.0378
- medical_genetics (mmlu)	0	none	acc	0.6300	±	0.0485
- professional_medicine (mmlu)	0	none	acc	0.4522	±	0.0302
- pubmedqa	1	none	acc	0.7300	±	0.0199

Groups	Version	Filter	n-shot	Metric	Value		Stderr
stem	N/A	none	0	acc_norm	0.3904	±	0.0066
		none	0	acc	0.4402	±	0.0058

📄 License

This model is released under the CC-BY-NC-ND 4.0 license.

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