Open mkhoshle opened 1 year ago
These masks specify where padding was added to the image and are used to ignore those regions during the attention computation. If you do not have padding just initialize masks with zeros/False. See here for where the masks are generated.
Hello,
In the Deformable Transformer, there is a variable called
valid_ratios
which is used based on the masks.valid_ratios = torch.stack([self.get_valid_ratio(m) for m in masks], 1)
. If the masks are None in my case how am I supposed to calculate it?Also, what is the purpose of
valid_ratios
? I could not find anything in the Trackformer and Original Deformable Detr paper.I would appreciate it if you could clarify this.