Why care about "Monad"?
This is monad https://huggingface.co/PleIAs/Monad
In my personal opinion, I believe it will change text classification in the long term as it is just as good as bert at half the size.
-This the accuracy of training it on my gender text classification dataset (https://huggingface.co/datasets/breadlicker45/gender-classification-v4.5).
This isn't over fitting or under fitting, most text classification models (like over 150m) get to 100% accuracy very fast and under 100m don't really. Monad is a bit slower at getting but overall gets to the same accuracy as a 150m text classification model. I can't wait to see this used in emotion classification in text.