# Fine-tuned BLOOMZ medical
The model is based on [BLOOMZ-3B](https://huggingface.co/bigscience/bloomz) a large language model trained by the BigScience team. <br>
The full paper is available [here](https://arxiv.org/abs/2211.01786).<br>
The model is a created using peft, transformers, and bitsandbytes.<br>

# The goal of the model
Using a transcript of a patient visit with his/her general health care provider (GP) recommend where that patient should be referred to and extract the main keywords from the transcript.
<br>
For example, if a patient has allergy-related concerns, then GP needs to refer that patient to the allergy/immunology department or an immunology doctor available.
<br><br>
Extracted keywords are there to help the doctor in making the final decision of what to recommend to the patient.
<br><br>

---
**dataset for fine-tuning:** 
- tchebonenko/MedicalTranscriptions
---