Hi Thank you .
1.If sentences is shown as a 768-dimensional vector for each sentence. what is the use of padding in this case ? we have set the PAD to 12 so we should be expecting the sentences with dimension as 12 right instead of 748 ?
PAD is applied for sentences length restriction right ? so you mean that is not applied at dimension level ?
2.Is the output of the sentence normalized ? In google USC getting sentence vector was simple but in berth there are lot of hard coded steps .
3.If i want to find the sentence similarity between 2 sentences then how can i do it ? inorder to verify that BERT gives a good representation of the sentences.
