cross entropy reshaping lines
brucemuller opened this issue · 0 comments
brucemuller commented
Hi, thank you for your efforts!
In cross_entropy2d function in loss.py I noticed there are these reshaping lines:
input = input.transpose(1, 2).transpose(2, 3).contiguous().view(-1, c)
target = target.view(-1)
Before passing into the loss calculation. Are these for efficiency? What's the main purpose?