license: apache-2.0
language:
- en
metrics:
- Rouge
pipeline_tag: summarization
tags:
- t5
- t5-small
- summarization
- medical-research
Model Card for Model ID
This model create summarizes text paragraphs from medical research journals into brief titles.
This modelcard aims to be a base template for new models. It has been generated using this raw template.
Model Details
Model Description
This is a text generative model to summarize long abstract from medical jourals into one liners. These one liners can be used as titles in the journal.
- Developed by: Tushar Joshi
- Shared by [optional]: Tushar Joshi
- Model type: t5-small
- Language(s) (NLP): English
- License: Apache 2.0
- Finetuned from model [optional]: t5-small baseline
Model Sources [optional]
- Repository: https://huggingface.co/t5-small
Uses
- As a text summarizer for medical journals paragraphs
- As a tunable language model for other medical tasks
Direct Use
- As a text summarizer for medical abstracts and journals.
Out-of-Scope Use
Should not be used as a text summarizer for very long tasks. Maximum token size of 1024.
Bias, Risks, and Limitations
- Max input token size of 1024
- Max output token size of 24
Recommendations
Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
How to Get Started with the Model
Use the code below to get started with the model.
from transformers import pipeline
text = """Text that needs to be summarized"""
summarizer = pipeline("summarization", model="path-to-model")
summary = summarizer(text)[0]["summary_text"]
print (summary)
Training Details
Training Data
The training data is internally curated and canot be exposed.
Training Procedure
None
Preprocessing [optional]
None
Training Hyperparameters
- Training regime: [More Information Needed]
- None
Speeds, Sizes, Times [optional]
The training was done using GPU T4x 2. The task took 4:09:47 to complete. The dataset size of 10,000 examples was used for training the generative model.
Evaluation
The quality of summarization was tested on 5000 medical journals created over last 20 years. The data of medical jounals is scraped from various sources.
Testing Data, Factors & Metrics
Test Data Size: 5000 examples
Testing Data
The testing data is internally generated and curated.
Factors
[More Information Needed]
Metrics
The model was evaluated on Rouge Metrics below are the baseline results achieved
Results
| Epoch | Training Loss | Validation Loss | Rouge1 | Rouge2 | Rougel | Rougelsum | Gen Len |
|---|---|---|---|---|---|---|---|
| 18 | 2.442800 | 2.375408 | 0.313700 | 0.134600 | 0.285400 | 0.285400 | 16.414100 |
| 19 | 2.454800 | 2.372553 | 0.312900 | 0.134100 | 0.284900 | 0.285000 | 16.445100 |
| 20 | 2.438900 | 2.372551 | 0.312300 | 0.134000 | 0.284500 | 0.284600 | 16.435500 |
Summary
Model Examination [optional]
[More Information Needed]
Environmental Impact
Carbon emissions can be estimated using the Machine Learning Impact calculator presented in Lacoste et al. (2019).
- Hardware Type: GPU T4 x 2
- Hours used: 4.5
- Cloud Provider: GCP
- Compute Region: Ireland
- Carbon Emitted: Unknown
Technical Specifications [optional]
Model Architecture and Objective
[More Information Needed]
Compute Infrastructure
[More Information Needed]
Hardware
[More Information Needed]
Software
[More Information Needed]
Citation [optional]
BibTeX:
[More Information Needed]
APA:
[More Information Needed]
Glossary [optional]
[More Information Needed]
More Information [optional]
[More Information Needed]
Model Card Authors [optional]
Tushar Joshi
Model Card Contact
Tushar Joshi LinkedIn - https://www.linkedin.com/in/tushar-joshi-816133100/