Question in the sampling stage

[medcx]

The style key of the validation set should be unknown, but you generate it directly during data preprocessing and use it directly in the sampling stage. Is this reasonable?

[CKBling]

Hello,

What you mentioned refers to the "Ours_best" option in the paper, and you're right—it isn't a true validation (test). The code is set up so that you can test using other options, like "Ours_avg," which only uses the style key from the training set.