Why do we need to apply mask while fine tuning?

[pranoy-k]

n attention class, you have the following code for masking. I understand the logic for pre training, but in fine tuning if we dont include language model loss we should have a check here for not applying the mask. Do we have to always apply the masking because the model was trained that way, is there an intuitive idea for this, because I dont see a necessity to do it experimentally