Open zml110120 opened 1 year ago
Dear author, I have a question.
What is the difference between "-1" in "TARGET_SENT" and "0" in "INPUT_SENT"? Just like the first image and second image? "-1" means padding and "0" means the end of sentence?
Right! And: "-1" in "TARGET_SENT" is convenient for computing loss.
XE loss
label smoothing
Dear author, I have a question.
What is the difference between "-1" in "TARGET_SENT" and "0" in "INPUT_SENT"? Just like the first image and second image? "-1" means padding and "0" means the end of sentence?