BioGPT logo

BioGPT

Free

4.2k

FreeFree tier
Inputs: textOutputs: text
Type
Open Source
Company
Microsoft

About BioGPT

BioGPT is a domain-specific generative pre-trained transformer model for biomedical text generation and mining, developed by Microsoft Research. It is pre-trained on large-scale biomedical literature and fine-tuned for various downstream tasks including question answering (PubMedQA), relation extraction (BC5CDR, DDI, KD-DTI), and document classification (HoC). The repository provides pre-trained checkpoints and fine-tuned models, along with code built on the fairseq framework. BioGPT achieves state-of-the-art performance on several biomedical NLP benchmarks and is available as open source on GitHub with Hugging Face Hub integration.

Key Features

Pre-trained on large-scale biomedical literature
Fine-tuned models for PubMedQA question answering
Fine-tuned models for relation extraction on BC5CDR, DDI, KD-DTI
Fine-tuned model for document classification on HoC
Available in base and large model variants
Open source implementation using fairseq framework
Pre-trained checkpoints downloadable via URL and Hugging Face Hub
Supports both text generation and information extraction tasks

Pros & Cons

Pros
  • State-of-the-art performance on several biomedical NLP benchmarks
  • Domain-specific pre-training improves accuracy on biomedical tasks
  • Open source and freely available with permissive license
  • Multiple fine-tuned models readily available for different tasks
  • Integrates with Hugging Face Hub for easy access
  • Well-documented usage examples and code
Cons
  • Requires significant computational resources for training and inference
  • Limited to biomedical domain; may not perform well on general text
  • Requires technical expertise to set up dependencies (fairseq, Moses, fastBPE)
  • Model size may be large for deployment on resource-constrained environments

Best For

Question answering on biomedical literatureDrug-target interaction extractionDrug-drug interaction extractionChemical-disease relation extractionDocument classification in the biomedical domainBiomedical text generation and summarization

FAQ

What is BioGPT?
BioGPT is a generative pre-trained transformer model specifically designed for biomedical text generation and mining, developed by Microsoft Research.
What tasks can BioGPT perform?
BioGPT can be fine-tuned for question answering, relation extraction, and document classification in the biomedical domain. Pre-trained and fine-tuned models are provided for PubMedQA, BC5CDR, DDI, KD-DTI, and HoC.
How do I use BioGPT?
You can download pre-trained checkpoints from the repository or Hugging Face Hub and use them with the provided code. The README includes example usage for generation and fine-tuned tasks.
Is BioGPT free?
Yes, BioGPT is open source and available for free on GitHub under the MIT license (as indicated by the repository license).