Med42 70b

🚀 Med42 - Clinical Large Language Model

Med42 is an open - access clinical large language model (LLM) developed by M42. It aims to expand access to medical knowledge. Built on LLaMA - 2 with 70 billion parameters, this generative AI system offers high - quality answers to medical questions.

🚀 Quick Start

🚨 Update: Version 2 of Med42 Released! 🚨

Please find the models here: [Med42 - v2 - 70B](https://huggingface.co/m42 - health/Llama3 - Med42 - 70B) and [Med42 - v2 - 8B](https://huggingface.co/m42 - health/Llama3 - Med42 - 8B)

Access Med42 on Hugging Face

This is a form to enable access to Med42 on Hugging Face. Please read the [Med42 License](https://huggingface.co/spaces/m42 - health/License) and accept our license terms and acceptable use policy before submitting this form. Requests will be processed by the M42 Team within 2 working days. Submit

Required Information

✨ Features

• Medical Knowledge Expansion: Med42 helps expand access to medical knowledge, providing high - quality answers to medical questions.
• Competitive Performance: Achieves competitive results on various medical benchmarks, such as MedQA, MedMCQA, and more.

📦 Installation

To use Med42, you need to follow these steps:

1. Read the [Med42 License](https://huggingface.co/spaces/m42 - health/License) and accept the license terms.
2. Request access to download the model weights (and tokenizer).

💻 Usage Examples

Basic Usage

from transformers import AutoModelForCausalLM, AutoTokenizer

model_name_or_path = "m42 - health/med42 - 70b"

model = AutoModelForCausalLM.from_pretrained(model_name_or_path,
                                             device_map="auto")

tokenizer = AutoTokenizer.from_pretrained(model_name_or_path)

prompt = "What are the symptoms of diabetes ?"
prompt_template=f'''
<|system|>: You are a helpful medical assistant created by M42 Health in the UAE.
<|prompter|>:{prompt}
<|assistant|>:
'''

input_ids = tokenizer(prompt_template, return_tensors='pt').input_ids.cuda()
output = model.generate(inputs=input_ids, temperature=0.7, do_sample=True,eos_token_id=tokenizer.eos_token_id, pad_token_id=tokenizer.pad_token_id, max_new_tokens=512)
print(tokenizer.decode(output[0]))

📚 Documentation

Model Details

• Model Developers: M42 Health AI Team
• Finetuned from model: Llama - 2 - 70B
• Context length: 4k tokens
• Input: Text only data
• Output: Model generates text only
• Status: This is a static model trained on an offline dataset. Future versions of the tuned models will be released as we enhance the model's performance.
• License: A custom license is available [here](https://huggingface.co/spaces/m42 - health/License)
• Research Paper: Med42 - Evaluating Fine - Tuning Strategies for Medical LLMs: Full - Parameter vs. Parameter - Efficient Approaches

Intended Use

Med42 is available for further testing and assessment as an AI assistant to enhance clinical decision - making and improve access to an LLM for healthcare use. Potential use cases include:

• Medical question answering
• Patient record summarization
• Aiding medical diagnosis
• General health Q&A

Hardware and Software

The training process was performed on the Condor Galaxy 1 (CG - 1) supercomputer platform.

Evaluation Results

Med42 achieves competitive performance on various medical benchmarks, including MedQA, MedMCQA, PubMedQA, HeadQA, and Measuring Massive Multitask Language Understanding (MMLU) clinical topics.

**We note that 0 - shot performance is not reported for Med - PaLM 2. Further details can be found at https://github.com/m42health/med42*.

Key performance metrics:

• Med42 achieves a 72% accuracy on the US Medical Licensing Examination (USMLE) sample exam, surpassing the prior state of the art among openly available medical LLMs.
• 61.5% on MedQA dataset (compared to 50.8% for GPT - 3.5)
• Consistently higher performance on MMLU clinical topics compared to GPT - 3.5.

Limitations & Safe Use

• Med42 is not ready for real clinical use. Extensive human evaluation is undergoing as it is required to ensure safety.
• Potential for generating incorrect or harmful information.
• Risk of perpetuating biases in training data.

⚠️ Important Note Use this model responsibly! Do not rely on it for medical usage without rigorous safety testing.

📄 License

The use of this model is governed by the M42 Health license. Please read the [Med42 License](https://huggingface.co/spaces/m42 - health/License).

📋 Accessing Med42 and Reporting Issues

• Reporting issues with the model: https://github.com/m42health/med42
• Reporting risky content generated by the model, bugs and/or any security concerns: https://forms.office.com/r/YMJu3kcKat
• M42’s privacy policy: [https://m42.ae/privacy - policy/](https://m42.ae/privacy - policy/)
• Reporting violations of the Acceptable Use Policy or unlicensed uses of Med42: med42@m42.ae

📖 Citation

Our paper has been published at AAAI 2024 Spring Symposium - Clinical Foundation Models and is available on arXiv: https://arxiv.org/abs/2404.14779

@article{christophe2024med42,
      title={Med42 -- Evaluating Fine-Tuning Strategies for Medical LLMs: Full-Parameter vs. Parameter-Efficient Approaches}, 
      author={Clément Christophe and Praveen K Kanithi and Prateek Munjal and Tathagata Raha and Nasir Hayat and Ronnie Rajan and Ahmed Al-Mahrooqi and Avani Gupta and Muhammad Umar Salman and Gurpreet Gosal and Bhargav Kanakiya and Charles Chen and Natalia Vassilieva and Boulbaba Ben Amor and Marco AF Pimentel and Shadab Khan},
      year={2024},
      eprint={2404.14779},
      archivePrefix={arXiv},
      primaryClass={cs.CL}
}