metadata
license: apache-2.0
datasets:
- pietrolesci/pubmed-200k-rct
metrics:
- accuracy
base_model:
- openai-community/gpt2
tags:
- medical
- biology
- research
- pubmed
MedGPT — GPT-2 Fine-Tuned on PubMed RCT
MedGPT is a GPT-2 model fine-tuned on the pubmed-200k-rct dataset. It classifies individual sentences from biomedical abstracts into one of five standard sections:
- Background
- Objective
- Methods
- Results
- Conclusion
This model is useful for tasks requiring structured understanding or summarization of scientific literature.
Training Details
- Base Model:
gpt2(124M parameters) - Dataset:
pietrolesci/pubmed-200k-rct - Task: Sentence classification
- Labels: Background, Objective, Methods, Results, Conclusion
- Epochs: 1 (partial training)
- Loss Function: CrossEntropy
- Optimizer: AdamW