YAML Metadata
Warning:
empty or missing yaml metadata in repo card
(https://huggingface.co/docs/hub/model-cards#model-card-metadata)
T5-Small Fine-tuned for Clinical Summarization of FHIR Document Reference Clinical Notes
This model is a fine-tuned version of the t5-small
model from Hugging Face, specifically tailored for the clinical summarization of FHIR Document Reference Clinical Notes.
Model Details
- Original Model: T5-Small
- Fine-tuned Model: dlyog/t5-small-finetuned
- License: Apache-2.0 (same as the original T5 license)
Fine-tuning Process
The model was fine-tuned using a synthetic dataset created with tools like Synthea. This dataset was used to simulate real-world clinical notes, ensuring the model understands the nuances and intricacies of medical terminology and context.
Only the last two layers of the t5-small
model were fine-tuned to retain most of the pre-trained knowledge while adapting it for better clinical summarization.
Usage
Using the model is straightforward with the Hugging Face Transformers library:
from transformers import T5ForConditionalGeneration, T5Tokenizer
model = T5ForConditionalGeneration.from_pretrained("dlyog/t5-small-finetuned")
tokenizer = T5Tokenizer.from_pretrained("dlyog/t5-small-finetuned")
def summarize(text):
input_text = "summarize: " + text
input_ids = tokenizer.encode(input_text, return_tensors="pt")
summary_ids = model.generate(input_ids)
summary = tokenizer.decode(summary_ids[0])
return summary
# Example
text = "Your clinical note here..."
print(summarize(text))
# Acknowledgements
A big thanks to the creators of the original t5-small model and the Hugging Face community. Also, gratitude to tools like Synthea that enabled the creation of high-quality synthetic datasets for fine-tuning purposes.
# License
This model is licensed under the Apache-2.0 License, the same as the original T5 model.
- Downloads last month
- 411
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social
visibility and check back later, or deploy to Inference Endpoints (dedicated)
instead.