Is there an issue with the code in the validate section?
Opened this issue · 0 comments
Le-up commented
caps = caps.to(device)
caplens = caplens.to(device)
scores, scores_d, caps_sorted, decode_lengths, sort_ind = decoder(imgs, caps, caplens)
The input of the model uses a true captions，I don't think this is the correct approach.