In experiments/train.py，training labels problem

As shown over here, the 2 labels creating method only masked the starting special tokens until the start of user content, not the whole query inputs until the start assistant content like what regular SFT does.

Does this make sense? train not just the answer but the query content as well? Or what it wants is not only, just train for instruction following, answer generation ability, but also a bit like pretrain for latent sentiment space alignment for the adapters?

And what does these comment notes means? "Not sure... Possibly what they want is...", "...Not 100% this is correct"
Aren't the authors of 《KBLaM》 provided this code for paper reproduction? What labels they used for training exactly? Does anyone have any ideas?

Start in experiments/train.py and locate the method that creates the two training labels. Compare its masking boundaries with the issue's description of regular SFT, then review the KBLaM paper or reproduction notes for the intended labels. Done means the intended masking behavior and the meaning of the comments are established, with any discrepancy clearly identified.