Check out some of the pretrained networks available

[RickyY1689]

Muck around on google colab + figure out how the flow is to download and use networks

[jaycehaoranwang]

Additional Useful links to possible guides to follow:

• https://www.machinecurve.com/index.php/2020/12/21/easy-text-summarization-with-huggingface-transformers-and-machine-learning/#performing-summarization-with-a-pretrained-transformer

The above provides an extremely simple grab and go code to get a summarizer already pretrained on news article with variable output summary length. Alot of applications and guides use pretrained models directly without retraining, some pretrained models are already pretrained on new articles, will need to find out how to further train them

[jaycehaoranwang]

Will need to determine the fidelity of our model/pipeline, as we can go super low level to process and load and retrain models like (https://blog.paperspace.com/generating-text-summaries-gpt-2/) does or we can simply use pretrained models on news articles and continuously retrain them if needed, for early prototyping and for the sake of getting things up and running maybe a very simple 5-liner summarizer code is sufficient, maybe for flex it will not be, but getting into lower levels will probably take exponentially more time.

It will probably be the case that this summarizer is not where we can best exercise our machine learning skills per se as low level work will basically turn into an infeasible research project and there are already so many pretrained models/networks out there to train from. better not reinvent the wheel.