wzic
wzic
Hi! Thanks for the code. When I used the "Random" initialization to do the training, the program reported error "assert (boxes[:, 2:] >= boxes1[:, :2]).all()" for the "generalized_box_iou" function in...
Hi! May I know why there are 20 keypoints of the pose for the reference? The conventional number of keypoints is 18. Also, it seems that the input image is...
So that means, when I want to use VACE 14B, I first need to have Wan 14B, and then load them together when using them ?
I also have the same problem when loading google/siglip-so400m-patch14-384 as the vision tower 
anyone knows about the 729 / 730 discrepancy issue ? It seems to have something to do withe the CLS token (just my guess)