2 part component Streamlit app with text translation using Google Translate (or DeepL) and biomedical text generation using the BioGPT generative transformer model pre-trained on specific data.
The code is not well structured and it results with a bit of a clunkiness in the app. It was more of an exercise to use the Streamlit API and test BioGPT.
An free tier app can be deployed on Streamlit but only the translator will work since running the text generation obviously requires too much memory and GPU power for a free tier.
Using the Python translators API, the app allows you to use either DeepL or Google with respective available choices or their find auto feature.
Choice of 3 pre-trained model for biomedical text generation and mining. All parameters were based on the official Microsoft/BioGPT model published on the huggingface platform.
Examples are output from locally ran queries.
More info can be found in the official publication :
- doi : https://doi.org/10.1093/bib/bbac409
- eprint = https://academic.oup.com/bib/article-pdf/23/6/bbac409/47144271/bbac409.pdf
- Install requirements
pip install -r requirements.txt
- Run it locally on Streamlit
streamlit run main_app.py
Non-functional (translation OK, text generation will not work, see comments above) live demo can be checked here : https://simlal-biogpt-streamlit-main-app-uct3jy.streamlit.app/