PulseGPT
A 123M parameter GPT-style language model trained from scratch on PubMed medical abstracts, built by Sola Technologies.
Model Details
- Architecture: GPT-2 style decoder-only transformer
- Parameters: 123M
- Training data: 85M tokens of PubMed medical abstracts
- Fine-tuned on: PubMedQA medical Q&A dataset
- Hardware: Apple MacBook Pro M5 Pro, 48GB RAM
- Framework: PyTorch + nanoGPT
Intended Use
- Medical text understanding
- Clinical Q&A
- Medical document summarisation
Built by
Sola Technologies — AI for active ageing and healthcare.
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support