I understand that simpleTransformers offers a simpler API, so we don't have to tokenize our data prior to training and it is done automatically based on the model we select,
But I was wondering, how is it done and how can I manipulate it?
In the case of binary classification, if I have 1 sentence then it is simply:
[CLS] This is my original sentence.
What if I want to add another 2 sentences, we have 2 options either:
Option 1: [CLS] This is my original sentence. [SEP] This is my 1st extra sentence. [SEP] This is my 2nd extra sentence
Option 2: [CLS] This is my original sentence. This is my 1st extra sentence. This is my 2nd extra sentence
Usually transformer model would learn better from option 1, but since simpleTransformer does it automatically, I was wondering how do I ensure that I am doing option 1?