license: apache-2.0
datasets:
- pubmed
language:
- fr
- en
- de
- nl
- es
- pt
- pl
- ro
- it
pipeline_tag: text-generation
tags:
- medical
- biology
BioMistral: A Collection of Open-Source Pretrained Large Language Models for Medical Domains
Abstract:
Large Language Models (LLMs) have demonstrated remarkable versatility in recent years, offering potential applications across specialized domains such as healthcare and medicine. Despite the availability of various open-source LLMs tailored for health contexts, adapting general-purpose LLMs to the medical domain presents significant challenges. In this paper, we introduce BioMistral, an open-source LLM tailored for the biomedical domain, utilizing Mistral as its foundation model and further pre-trained on PubMed Central. We conduct a comprehensive evaluation of BioMistral on a benchmark comprising 10 established medical question-answering (QA) tasks in English. We also explore lightweight models obtained through quantization and model merging approaches. Our results demonstrate BioMistral's superior performance compared to existing open-source medical models and its competitive edge against proprietary counterparts. Finally, to address the limited availability of data beyond English and to assess the multilingual generalization of medical LLMs, we automatically translated and evaluated this benchmark into 7 other languages. This marks the first large-scale multilingual evaluation of LLMs in the medical domain. Datasets, multilingual evaluation benchmarks, scripts, and all the models obtained during our experiments are freely released.
1. BioMistral models
BioMistral is a suite of Mistral-based further pre-trained open source models suited for the medical domains and pre-trained using textual data from PubMed Central Open Access (CC0, CC BY, CC BY-SA, and CC BY-ND). All the models are trained using the CNRS (French National Centre for Scientific Research) Jean Zay French HPC.
| Model Name | Base Model | Model Type | Sequence Length | Download |
|---|---|---|---|---|
| BioMistral-7B | Mistral-7B-Instruct-v0.1 | Further Pre-trained | 2048 | HuggingFace |
| BioMistral-7B-DARE | Mistral-7B-Instruct-v0.1 | Merge DARE | 2048 | HuggingFace |
| BioMistral-7B-TIES | Mistral-7B-Instruct-v0.1 | Merge TIES | 2048 | HuggingFace |
| BioMistral-7B-SLERP | Mistral-7B-Instruct-v0.1 | Merge SLERP | 2048 | HuggingFace |
2. Quantized Models
| Base Model | Method | q_group_size | w_bit | version | VRAM GB | Time | Download |
|---|---|---|---|---|---|---|---|
| BioMistral-7B | FP16/BF16 | 15.02 | x1.00 | HuggingFace | |||
| BioMistral-7B | AWQ | 128 | 4 | GEMM | 4.68 | x1.41 | HuggingFace |
| BioMistral-7B | AWQ | 128 | 4 | GEMV | 4.68 | x10.30 | HuggingFace |
| BioMistral-7B | BnB.4 | 4 | 5.03 | x3.25 | HuggingFace | ||
| BioMistral-7B | BnB.8 | 8 | 8.04 | x4.34 | HuggingFace | ||
| BioMistral-7B-DARE | AWQ | 128 | 4 | GEMM | 4.68 | x1.41 | HuggingFace |
| BioMistral-7B-TIES | AWQ | 128 | 4 | GEMM | 4.68 | x1.41 | HuggingFace |
| BioMistral-7B-SLERP | AWQ | 128 | 4 | GEMM | 4.68 | x1.41 | HuggingFace |
2. Using BioMistral
You can use BioMistral with Hugging Face's Transformers library as follow.
Loading the model and tokenizer :
from transformers import AutoModel, AutoTokenizer
tokenizer = AutoTokenizer.from_pretrained("BioMistral/BioMistral-7B")
model = AutoModel.from_pretrained("BioMistral/BioMistral-7B")
3. Supervised Fine-tuning Benchmark
| Clinical KG | Medical Genetics | Anatomy | Pro Medicine | College Biology | College Medicine | MedQA | MedQA 5 opts | PubMedQA | MedMCQA | Avg. | |
|---|---|---|---|---|---|---|---|---|---|---|---|
| BioMistral 7B | 59.9 | 64.0 | 56.5 | 60.4 | 59.0 | 54.7 | 50.6 | 42.8 | 77.5 | 48.1 | 57.3 |
| Mistral 7B Instruct | 62.9 | 57.0 | 55.6 | 59.4 | 62.5 | 57.2 | 42.0 | 40.9 | 75.7 | 46.1 | 55.9 |
| BioMistral 7B Ensemble | 62.8 | 62.7 | 57.5 | 63.5 | 64.3 | 55.7 | 50.6 | 43.6 | 77.5 | 48.8 | 58.7 |
| BioMistral 7B DARE | 62.3 | 67.0 | 55.8 | 61.4 | 66.9 | 58.0 | 51.1 | 45.2 | 77.7 | 48.7 | 59.4 |
| BioMistral 7B TIES | 60.1 | 65.0 | 58.5 | 60.5 | 60.4 | 56.5 | 49.5 | 43.2 | 77.5 | 48.1 | 57.9 |
| BioMistral 7B SLERP | 62.5 | 64.7 | 55.8 | 62.7 | 64.8 | 56.3 | 50.8 | 44.3 | 77.8 | 48.6 | 58.8 |
| MedAlpaca 7B | 53.1 | 58.0 | 54.1 | 58.8 | 58.1 | 48.6 | 40.1 | 33.7 | 73.6 | 37.0 | 51.5 |
| PMC-LLaMA 7B | 24.5 | 27.7 | 35.3 | 17.4 | 30.3 | 23.3 | 25.5 | 20.2 | 72.9 | 26.6 | 30.4 |
| MediTron-7B | 41.6 | 50.3 | 46.4 | 27.9 | 44.4 | 30.8 | 41.6 | 28.1 | 74.9 | 41.3 | 42.7 |
| BioMedGPT-LM-7B | 51.4 | 52.0 | 49.4 | 53.3 | 50.7 | 49.1 | 42.5 | 33.9 | 76.8 | 37.6 | 49.7 |
| GPT-3.5 Turbo 1106* | 74.71 | 74.00 | 65.92 | 72.79 | 72.91 | 64.73 | 57.71 | 50.82 | 72.66 | 53.79 | 66.0 |
Supervised Fine-Tuning (SFT) performance of BioMistral 7B models compared to baselines, measured by accuracy (↑) and averaged across 3 random seeds of 3-shot. DARE, TIES, and SLERP are model merging strategies that combine BioMistral 7B and Mistral 7B Instruct. Best model in bold, and second-best underlined. *GPT-3.5 Turbo performances are reported from the 3-shot results without SFT.
Citation BibTeX
Coming Soon